Citation
Fiber-coupled microscopy for 3D neuronal imaging

Material Information

Title:
Fiber-coupled microscopy for 3D neuronal imaging
Creator:
Ozbay, Baris N.
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English

Thesis/Dissertation Information

Degree:
Doctorate ( Doctor of philosophy)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
Department of Bioengineering, CU Denver
Degree Disciplines:
Bioengineering
Committee Chair:
Benninger, Richard K.P.
Committee Members:
Gibson, Emily A.
Restrepo, Diego
Dell'Acqua, Mark L.
Felsen, Gidon

Notes

Abstract:
In this dissertation I describe the design and implementation of miniature fiber-coupled microscopes (FCMs) with active focusing for three-dimensional (3D) neuronal imaging. The goal is to provide neuroscience researchers with versatile microscopy tools to perform neuronal optical imaging of awake and mobile mice. This research is motivated by the recent advancements of powerful genetically-encoded optical proteins, including fluorescent activity sensors as well as optogenetic actuators, which permit the functional interrogation of in vivo neuronal circuits. Newly developed miniature microscope tools allow optical imaging in freely-behaving mice, but current designs do not combine optical sectioning capabilities with active focusing for full 3D-imaging. The first three chapters in this dissertation serve as a background for current state of intravital fluorescence microscopy in neuroscience research. I argue for the importance of miniaturizing the microscope technologies that enable high-contrast imaging with optical sectioning, combined with axial focusing, to enable 3D-imaging at the rodent-scale. I present two FCM designs that achieve full 3D-imaging using a coherent imaging fiber bundle (CIFB) for lateral imaging and an electrowetting tunable lens (EWTL) to enable electrically tunable axial focusing. The first design is a confocal FCM (C-FCM) that takes advantage of the optical sectioning capability of the CIFB to acquire high-contrast images. The second design is a two-photon FCM (2P-FCM), in which pre-compensated ultrashort pulses are propagated through the CIFB for two-photon excitation microscopy. In each section, I characterize the 3D optical performance of the FCM. Finally, as a proof-of-principle using the 2P-FCM, I show in vivo 3D-imaging of neurons and Ca2+-activity in the motor cortex of a freely-behaving mouse.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
Copyright Baris N. Ozbay. Permission granted to University of Colorado Denver to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.

Downloads

This item has the following downloads:


Full Text
FIBER-COUPLED MICROSCOPY FOR 3D NEURONAL IMAGING
by
BARIS N OZBAY
B.S., University of Colorado Denver, 2010
A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Doctor of Philosophy Bioengineering Program
2017


©2017
BARIS OZBAY ALL RIGHTS RESERVED
11


This thesis for the Doctor of Philosophy degree by Baris N Ozbay has been approved for the Bioengineering Program
by
Richard KP Benninger, Chair Emily A Gibson, Advisor Diego Restrepo, Co-advisor Mark L Dell’ Acqua Gidon Felsen
Date: December 16, 2017
m


Ozbay, Baris N (Ph.D., Bioengineering Program)
Fiber-Coupled Microscopy for 3D Neuronal Imaging
Thesis directed by Assistant Professor Emily A Gibson and Professor Diego Restrepo
ABSTRACT
In this dissertation I describe the design and implementation of miniature fiber-coupled microscopes (FCMs) with active focusing for three-dimensional (3D) neuronal imaging. The goal is to provide neuroscience researchers with versatile microscopy tools to perform neuronal optical imaging of awake and mobile mice. This research is motivated by the recent advancements of powerful genetically-encoded optical proteins, including fluorescent activity sensors as well as optogenetic actuators, which permit the functional interrogation of in vivo neuronal circuits. Newly developed miniature microscope tools allow optical imaging in freely-behaving mice, but current designs do not combine optical sectioning capabilities with active focusing for full 3D-imaging.
The first three chapters in this dissertation serve as a background for current state of intravital fluorescence microscopy in neuroscience research. I argue for the importance of miniaturizing the microscope technologies that enable high-contrast imaging with optical sectioning, combined with axial focusing, to enable 3D-imaging at the rodent-scale.
I present two FCM designs that achieve full 3D-imaging using a coherent imaging fiber bundle (CIFB) for lateral imaging and an electrowetting tunable lens (EWTL) to enable electrically tunable axial focusing. The first design is a confocal FCM (C-FCM) that takes advantage of the optical sectioning capability of the CIFB to acquire high-contrast images. The second design is a two-photon FCM (2P-FCM), in which precompensated ultrashort pulses are propagated through the CIFB for two-photon excitation
IV


microscopy. In each section, I characterize the 3D optical performance of the FCM. Finally, as a proof-of-principle using the 2P-FCM, I show in vivo 3D-imaging of neurons and Ca2+-activity in the motor cortex of a freely-behaving mouse.
The form and content of this abstract are approved. I recommend its publication.
Approved: Emily A Gibson and Diego Restrepo
v


Dedicated to the research animals
and to the researchers who treat them with compassion and respect


ACKNOWLEDGEMENTS
The completion of this dissertation is only possible because of the support and patience of my advisors, Emily Gibson and Diego Restrepo. I am grateful for their enduring belief in me through periods of adversity and for giving me the autonomy to build my own path.
I thank the contributors to this work, including Gregory Futia, Justin Losacco, Ming Ma, Bob Cormack, and Ethan Hughes who gave their skills and time to important parts of these projects. Thanks also to the labs of Juliet Gopinath and Victor Bright at the University of Colorado Boulder, who have shared their expertise and resources. Thanks to the Advanced Light Microscopy Core for trusting me to tinker with and not break their instruments. My sincere gratitude goes to everyone has come through our labs for the discussions, assistance, and friendships.
I acknowledge those who worked to advocate for the larger concepts that germinated my own projects and funding from the National Science Foundation that made this work a reality. I am appreciative for those who critically reviewed my work and my thesis committee for their patient examination of my progress.
I thank my parents, for everything.
Finally, my heartfelt gratitude goes to Katie, for being a resilient and guiding light through these many years of study.
Vll


TABLE OF CONTENTS
Abstract.....................................................................iv
Acknowledgements............................................................vii
Table of Contents..........................................................viii
List of Figures...............................................................x
List of Abbreviations.......................................................xvi
I INTRODUCTION...............................................................1
1.1 Thesis organization...................................................3
II MINIATURIZED MICROSCOPY IN NEUROSCIENCE....................................5
2.1 Fluorescence microscopy in the neurosciences..........................5
2.2 Modern microscopy techniques for studying the brain...................9
2.3 In vivo brain imaging methods........................................17
2.4 Miniaturized microscope designs......................................21
2.5 Where we are in miniaturized neuronal imaging?.......................28
III AXIAL FOCUSING IN MINIATURIZED MICROSCOPES...............................30
3.1 Benefits of minimizing depth-of-field in fluorescence imaging........32
3.2 Active axial focusing methods........................................41
3.3 Tunable focus by liquid lens technology..............................48
3.4 Optical design considerations for axial scanning with a tunable lens.50
IV CONFOCAL FIBER-COUPLED MICROSCOPE FOR 3D IMAGING.........................54
viii


4.1 Miniaturizing confocal laser-scanning microscopy.................55
4.2 Methods..........................................................56
4.3 Results..........................................................60
4.4 Summary..........................................................65
V TWO-PHOTON FIBER-COUPLED 3D MICROSCOPY................................67
5.1 Miniaturizing two-photon laser-scanning microscopy...............68
5.2 Methods..........................................................69
5.3 Results..........................................................81
5.4 Summary..........................................................95
VI CONCLUSIONS AND OUTLOOK..............................................96
6.1 Revisiting the confocal FCM......................................97
6.2 Moving forward with the two-photon FCM..........................100
6.3 Fiber-bundles: Simple, versatile, needs improvement.............102
6.4 Finding a need for the 3D imaging FCM...........................105
APPENDIX - FIBER-BUNDLE DE-PIXELATION METHOD............................107
References..............................................................116
IX


LIST OF FIGURES
FIGURE
2.1: (a) Purkinje neurons as represented in the 19th century by Ramon y Cajal
(Sotelo, 2003). (b) Modern fluorescence intensity projection image of a mouse Purkinje neuron filled with Lucifer Yellow fluorescent dye (Martone, 2002)...........................................................6
2.2: Illustration of the epi-fluorescence microscope. Monochromatic excitation light, selected by an excitation filter, is reflected by a dichroic beamsplitter. An objective lens focuses the excitation light onto the sample and collects the emitted fluorescence from the fluorophores. The emission passes through the dichroic beam-splitter and emission filter before being captured on a detector, such as a camera, to form an image of the fluorescence...............................................................7
2.3: A simplified Jablonski-diagram showing electronic and vibrational states of a common fluorophore illustrates how a photon of the correct energy, related to wavelength (k), is absorbed by a fluorophore to increase the electronic state from the ground state (SO) to the excited state (SI).
Spontaneous vibrational relaxation is followed by fluorescence emission
as the fluorophore recovers to SO, with the emitted photon having less
energy and longer wavelength..............................................8
2.4: Confocal microscopy basic setup with stage-scanning. An optical pinhole in the conjugate focal plane is adjusted to reject out-of-focus fluorescence emission from a thick fluorescent sample. The laser is scanned across the sample, in this illustration with a translation stage..........11
2.5: 2P-LSM basic setup with stage-scanning. A pulsed ultrashort excitation laser is focused onto the sample, resulting in a non-linear excitation of fluorescence limited to the focal volume. The fluorescence is filtered and recorded for each pixel by a single-element detector..............................13
2.6: One- and two-photon absorption process are shown to both cause a similar transition to the excited electronic state, resulting in similar fluorescence
output. The two-photon process requires two photons to arrive near-simultaneously with approximately twice the wavelength, or half of the energy, of the energy gap..............................................14
2.7: Cranial window implanted in mouse to achieve optical access to cortical
structures. An objective lens is used to image directly into the window........17
2.8: GRIN lens illustrations. Top: Single GRIN lens with pitch ~0.5 with a
magnification of 1. Bottom: Two-element GRIN lens with low NA relay at pitch -0.75 and high NA objective with pitch -0.25..................19
x


2.9: GRIN lens assembly implanted into deep-brain of mouse and imaged with
a low NA objective lens to couple into the low NA GRIN lens element.........20
2.10: Example widefield miniature microscope design from miniscope.org. (a)
Internal components of miniature microscope (weight 2-g), with
integrated LED, CMOS camera, filters, and lenses, (b) Photo of mouse
with head-attached miniature microscope.....................................22
2.11: Fiber-scanning techniques, (a) Piezoelectric resonance-scanning of the fiber tip, resulting in the scanning of resonant modes, such as the spiral shown inset, (b) MEMS actuated-mirror scanners, resulting in arbitrary scan trajectories, such as the raster shown inset....................................24
2.12: Illustration of proximal to distal image coherence of CIFB, as well as the pixelation of an image formed through the bundle because of the discrete fiber cores...........................................................................26
2.13: Fujikura CIFB fluorescence image of a uniform target showing core
distribution.................................................................27
3.1: (a) Sketch representation of the human motor cortex neuronal column by
Ramon y Cajal (1899). (b) Modern 3D-reconstruction of a three-photon image of a cortical column in an intact mouse brain (adapted from (Horton et al., 2013)).......................................................31
3.2: (a) Gaussian profile of the w(z) function, indicating the x-axis parameter
coo and the z-axis parameter zr. (b) Gaussian transverse intensity profile at beam waist (z=0), indicating the wo value and the related FWHMr. (c)
Axial intensity profile along the optical axis (x=0), indicating the zR
value and the related FWHMz..................................................36
3.3: 2-pm fluorescent microsphere imaged under blue light illumination, (a) A widefield image of the fluorescent microsphere through the focus, showing the transverse (top) and axial (bottom) focus and the extended DOF. (b) A LSCM image of an identical bead showing the improved lateral resolution and confined axial spread..........................................37
3.4: Power encircled at radius 5FWHMr through focus of 2 pm beads capture with widefield (solid line, no optical sectioning) and LSCM (dashed line, with optical sectioning)........................................................38
3.5: Fluorescent pollen grains under green light excitation taken with a 1.42 NA 60x objective, (a) The pollen grains imaged under widefield illumination, showing the loss of contrast in the lateral dimension and a large DOF fetching out-of-focus grains, (b) A LSCM optical stack of several thin focal planes from the same pollen grains showing much higher contrast.......................................................................39
xi


3.6: Three fundamental axial focusing options. Object field scan changes the location of the imaging target relative to the optical system. Image field scan changes the imaging optics to change the focal plane. Tunable focus scanning changes the effective focal length of the optics to change the focal plane......................................................................42
3.7: Examples of published lens-movement-based axials scanning, (a) Axial scanning by small focusing-motor actuation with 1.1-mm axial motion (Flusberg et al., 2008a). (b) Lens actuation by shape-memory alloy contraction, with 150-pm of axial motion (Wu et al., 2010)...........................43
3.8: Examples of published MEMS mirror devices with axial focusing, (a)
Thin-film piezoelectric actuator for 190-pm of vertical translation with parabolic mirrors for excitation and collection (Qiu et al., 2014). (b)
Monolithic 3-axis MEMS scanning mirror assembly, with vertical translation photos shown in (c), achieving 546-pm of vertical translation (Lietal., 2016)..............................................................45
3.9: Two varieties of commercially available liquid ETLs. (a) Optotune EL-10-30-C-MV ETL based on a shape-changing polymer surface, with outer diameter of 30-mm and 10-mm aperture, (b) Varioptic Arctic 316, based on electrowetting lens technology, with outer diameter of 7.8-mm and 2.5-mm aperture.......................................................................49
3.10: Axial focusing by wavefront curvature shaping at the objective back focal plane, (a) An input converging or diverging wavefront is transformed into an axial translation away from the designed focal length of the lens.
(b) Spherical aberrations as a result of a lens obeying the sine condition..52
4.1: Laser scanning confocal microscope coupled to distal imaging optics
through fiber bundle (IL: Imaging lens, EWTL: Electrowetting tunable
lens, OL: Objective lenses)..................................................56
4.2: (a) CAD model of adapter with optics with lenses aligned in polyimide
(PI) tubing, (c) Photo of assembled adapter and EWTL.........................58
4.3: Lateral imaging characteristics of C-FCM at different objective depths, (a)
GFP labeled neuroglia imaged with 20X, 0.8 NA objective, (b) Same
region imaged with C-FCM at shortest and (c) longest objective lens
length.......................................................................60
4.4: Imaging power loss over the full FOV with 1-mm distance objective lens
(black line) and 4-mm distance objective lens (grey line)....................61
4.5: Zoom-in showing the cell bodies indicated by white arrow in Figure 4.3 (a). Left: Raw C-FCM image showing pixilation due to fiber-cores.
Middle: C-FCM image after filtering. Right: Comparison with standard confocal microscope..........................................................61
xii


4.6: Axial scan range of electrowetting C-FCM (a) Orthogonal projection
(inverted grayscale) of 1-pm diameter red fluorescent beads in agarose imaged with scanning Z-stage (left) and scanning the EWTL (right), (b)
Black dots: 40 beads mapped from EWTL optical power to relative Z-position. Gray line: Simulated C-FCM focal length with varying EWTL power......................................................................62
4.7: Lateral and axial resolution of C-FCM. (a) Lateral and axial images of fluorescent beads. Left: 1-pm bead with open pinhole (PH), Middle: 1-pm bead with closed pinhole, Right: 2-pm bead with closed pinhole, (b)
Top: Averaged line profiles of 1- and 2-pm beads (black) compared with diffraction-limited resolution (grey) Bottom: Averaged axial profile of several 1 and 2-pm beads (black) compared with theoretical axial resolution (grey)...........................................................63
4.8: 3D-imaging of mouse nerve tissue, (a) Four optical sections that were
taken at specific EWTL optical power settings. Scale bar is 100-pm. (b)
Maximum intensity projection of an image stack of intact olfactory
neuron axons labeled with YFP...............................................65
5.1: 2P-FCM imaging system. Pulses from a Ti:Sapphire laser source are spectrally pre-broadened through polarization maintaining fiber (PM fiber) and pre-chirped using a grating-based pulse stretcher. Output pulses are scanned onto the surface of the CIFB through scanning mirrors, scan/tube lens relay, and lOx Objective. Fluorescence emission is collected by the CIFB and directed to a PMT through a dichroic filter.
The collected pulses are amplified and transformed to logic levels to be detected by the DAQ and PC..................................................70
5.2: Optics of the 2P-FCM miniature microscope head that focuses excitation light from CIFB cores onto the tissue. The CIFB -coupling asphere collects the light from the cores of the CIFB, which are then passed through the aperture of the EL. The plano-convex lens and the objective asphere focus the light onto the tissue through a #1 coverglass with 0.15-
mm thickness..............................................................73
5.3: NA comparison of forward excitation light at 910 nm (top) and backward emission light at 532-nm (bottom). Largest emission and excitation field positions are matched at 220-pm FOV...............................................74
5.4: 3D-printed enclosure (a) A two-part 2P-FCM snaps together to secure the EWTL and electrode, (b) Photo of 3D-printed parts, top: before any processing with supports still attached and bottom: assembled 2P-FCM..............75
5.5: Mouse attachment, (a) The CIFB is attached to a coupling objective on the proximal end and the 2P-FCM on the distal end. (a) 2P-FCM is attached to the permanent baseplate on the mouse with a quarter-turn.......................79
xm


5.6: Image processing of two-photon imaging through CIFB. (a) Flat-field map showing enhanced non-uniformity due to heterogeneity in CIFB. (b)
Unprocessed image of cells in mouse cortex with fiber-pixelation. (c) Post-processed image after fiber-cores were corrected with flat-field mask and re-gridded into a typical square pattern. Grid lines added to emphasize pixels..............................................................80
5.7: Axial and lateral resolution tested by imaging fluorescent beads with a 20x
Olympus objective (dashed lines) or with the 2P-FCM (solid lines).............82
5.8: Testing of axial scan range, (a). Side-projection of ~2-pm diameter
fluorescent beads suspended in clear agarose and imaged with a 20x 0.8 NA Olympus objective (green) using 910 nm excitation light and a motorized stage or with the 2P-FCM while varying the EWTL power (red), (b) Predicted scan range as the EWTL optical power is changed modeled in Zemax (grey line) and Z-positions of measured beads (black
circles)......................................................................83
5.9: Measurements of magnification of 2P-FCM. Elements of a USAF 1951 resolution target were imaged through the CIFB at three different focus settings. The known size of the elements is known precisely is indicated...............84
5.10: Measurement of pulses for propagation through 1,0-m long CIFB using FROG at 910 nm. Each figure: Top: Measured spectrogram. Middle:
Spectrum retrieval from FROG spectrogram overlaid with phase plot (orange). Separately measured spectrum is shown as well (dashed line).
Bottom: Temporal pulse trace retrieved from FROG spectrogram
overlaid with phase plot (orange), (a) Pulse measured directly from
ultrafast laser, (b) Pulse measured after propagating through CIFB..........86
5.11: Fixed tissue with GFP-labeled oligodendrocytes imaged with 2P-FCM. (a)
3D volume acquired by the 2P-FCM (220 x 220 pm lateral x 180 pm
axial) with over 200 cells in the image, (a). Processed image of a single
slice in the stack after filtering to remove pixelation pattern.............88
5.12: Tilted-field imaging enabled by rapid focusing of the EWTL. (a)
Maximum intensity projection of fixed mouse brain tissue expressing
GCaMP6s in neurons, acquired by the 2P-FCM. Arrows indicate cell
bodies retained in fields, (b) Side projection of the volume. The same
cell bodies are indicated by the arrows. The planes for the horizontal and
angled scan limits are indicated (c) Images showing the largest tilted-
field scan acquired in this experiment, indicating the same cell bodies
that are shown to intersect with the red or white planes in (b).............89
5.13: Multi-color imaging, (a) Two-color maximum intensity projection acquired with a 20x 0.75 NA Objective A. Maximum intensity projection of a region of brain tissue acquired with a 20x 0.75 NA
xiv


objective. Yellow cells are oligodendrocytes (Green and Red), while
red-only cells are astrocytes and oligodentrocytes cells. Two astrocytes
are marked with arrows, and are easily identified by the characteristic
bushy morphology and two are marked with arrows, (b) Same tissue
imaged with the 2P-FCM, using the same detectors, filters, and
excitation wavelength. Green and red cells are visible in the field, with
likely astrocytes marked by arrows........................................90
5.14: Mouse 2P-FCM attachment photos, (a) Baseplate implanted on mouse
with cranial window, (b) Mouse behaving with 2P-FCM attached..............91
5.16: Histological coronal section of mouse injected with GCaMP6s virus,
showing good expression in layers 2-5 of motor cortex.....................92
5.15: Implant stability over 17 days imaged with widefield epi-fluorescence.......92
5.17: 3D-imaging of cells labeled with GCaMP6s in behaving mouse, (a)
Volume view of cells (b) Side-projection of cells.........................94
5.18: Awake-behaving mouse Ca2+ imaging with GCaMP6s. (a) Three different focal planes showing distinct cellular populations. The colors and regions represent the peak fluorescent spatial intensity from the corresponding transients. Scale bar 25 pm. (b) Traces showing change in GCaMP6s fluorescence for the 5 regions of distinct activity indicated in panel (a)...........................................................................95
6.1: Coupling of green emission light into CIFB is indifferent to spot size
unless coupling is in single-core..........................................97
6.2: Conceptual experiments for confocal imaging with FCM and GRIN lens.
(a) 2P-FCM coupled to a 0.8 NA GRIN lens assembly to image GFP-labeled oligodendrocytes in fixed mouse brain tissue, (b) Maximum intensity projection of lateral and side-views after processing to remove fiber-pixelation (scale bar is 30-pm)......................................99
6.3: Core-to-core coupling imaged at distal end of fiber while single-core is illuminated by 910 nm light at proximal end. (a) Raw camera image of fibers showing higher order modes coupling into nearby fibers, (b) Same image with coupled cores circled...................................................104
xv


LIST OF ABBREVIATIONS
2D...........Two-dimensional
2P-LSM.......Two-photon laser-scanning microscopy
3D...........Three-dimensional
C-FCM........Confocal fiber-coupled microscope
CIFB.........Coherent imaging fiber-bundle
CMOS.........Complementary metal-oxide semiconductor
CW...........Continuous wave
DAQ..........Data acquisition
DOF..........Depth-of-field
ETL..........Electrically tunable lens
EWTL.........Electro wetting tunable lens
FFT..........Fast Fourier transform
FL...........Focal length
FROG.........Frequency-resolved optical gating
FWHM.........Full-width at half-maximum
GDD..........Group-delay dispersion
GECI.........Genetically encoded calcium indicator
GFP..........Green fluorescent protein
LSM..........Laser-scanning microscopy
LSCM.........Laser-scanning confocal microscopy
MTF..........Modulation transfer function
NA...........Numerical aperture
NIR
.Near-infrared


PMT...........Photomultiplier tube
PSF...........Point spread function
ROI...........Region of interest
SLM...........Spatial light modulator
SNR...........Signal-to-noise ratio
SPIM..........Selective plane illumination microscopy
YFP...........Yellow fluorescent protein


CHAPTER I
INTRODUCTION
The topic of this thesis is the development, assembly, and testing of miniature fiber-coupled microscopes with tunable focus for mouse brain imaging. These microscopes are designed specifically to aid neuroscience researchers in investigating the functioning of the mammalian brain in a natural behavior paradigm with true three-dimensional (3D) imaging capabilities.
In the last two decades, fluorescence microscopy has become an indispensable tool for neuroscience research. Recently, there have been important innovations in the engineering of fluorescent proteins to enable genetically-targeted optical sensing and actuation of cellular activity (Akerboom et al., 2012; Chen et al., 2013b; Lin and Schnitzer, 2016; Specht et al., 2017; Zhang et al., 2007). These optical proteins can be made to express in specific neuronal cell types of live animals, such as mice. Modern fluorescence microscopes and surgical techniques have allowed researchers to interrogate in vivo functioning neuronal networks in a much larger temporal and spatial regime than ever before. In contrast, electrophysiological recordings from implanted brain electrodes can record from only a small number of cells and the recorded signals are not intrinsically cell-type specific. With the continued development of faster optical sensors and spectrally distinct optical actuators, the hope is that an “all-optical” interrogation of large neuronal networks in awake and freely-behaving animals can be achieved (Emiliani et al., 2015).
Despite the benefits of in vivo optical imaging, it must still overcome some big hurdles to match the maturity of electrophysiological techniques. Some of the weaknesses of optical imaging are: the lower temporal resolution, the limitation on imaging depth due to
1


brain tissue opacity, and the difficulty to miniaturize optics for attachment to small subjects, such as mice, for minimally invasive recording of behavior. Meanwhile, electrode-based recordings have been miniaturized for freely-behaving mouse imaging for several decades, leading to the innovation of the tetrode for gathering rich data from behaving mice (Harris et al., 2000). These tools have allowed for the validation of many neuronal circuit models that were developed using in vitro techniques. To obtain similarly functionally relevant data from optical imaging, it is desirable to combine optical imaging with behavioral monitoring. Tethered but freely-moving mouse electrophysiology (Anikeeva et al., 2011; Li et al., 2014) and microscopy (Ziv et al., 2013) have demonstrated the importance of minimizing behavioral restraints for studying the awake brain in action. Further development of behaviorally-related optical imaging would open up new avenues for investigating animal-models of disease and the effects of drugs and therapies on the mammalian brain (Heys et al., 2014; Hillman, 2007). However, optics are difficult to miniaturize to the scales that would minimally disrupt or restrict the behavioral environment.
Along with the improvements in the biological tools, microscope technologies have generally improved to meet the demand. Head-fixed animal imaging has provided an avenue for utilizing existing large microscope optics for imaging awake mice during a limited set of behaviors (Dombeck and Tank, 2014). Cutting-edge techniques, such as mesoscopic imaging, have begun to show that very large-scale data, (upwards of thousands of spatially-related neurons), can be retrieved through optical imaging of the brain (Sofroniew et al., 2016). But there are many behavioral paradigms that are not well replicated in a severely restrained setup, such as those used in head-fixed imaging. Examples
2


are social and mating behaviors, spatial maze exploration, and the complex responses to anxiety or fear stimuli. The reliance on existing large microscope designs for head-fixed imaging has meant that the transition to in vivo optical imaging for small subjects, such as mice, has been under-addressed by available microscope technologies.
As optical interrogation tools are becoming an undeniably powerful methodology, there is a push in neuroscience to close the gap in microscope scale. Many established neuronal circuit models of function and disease are being tested using genetically-targeted fluorescent tools (Grewe et al., 2017; Tantirigama et al., 2017). Simultaneously, head-attached miniature microscopes for mice are becoming more available and increasingly useful for neuroscience researchers, but are narrow in the available modalities (Chen et al., 2013a). Versatile miniature microscopes that incorporate more advanced microscope technologies are increasingly in-demand. In this thesis I will summarize my approach to addressing this technological gap.
1.1 Thesis organization
The work that I present here is focused on the development of two types of miniature fiber-coupled microscopes (FCM) for fluorescent neuronal imaging of tethered but otherwise freely-moving mice.
This technique leverages two existing technologies:
1) The high-density coherent imaging fiber-bundle (CIFB)
2) The electrowetting tunable-focus lens (EWTL).
The CIFB is employed for lateral tissue imaging while the tunable lens adds axial focusing capabilities, adding a third dimension to the data. I show their integration into a
3


miniaturized imaging system that allows for the first demonstration of true 3D neuronal imaging in freely-moving mice.
I describe the development and validation of two types of 3D FCMs:
1) A confocal fiber-coupled microscope (C-FCM)
2) A two-photon fiber-coupled microscope (2P-FCM)
In Chapter 2,1 explore the history and current state of fluorescence in vivo neuronal imaging in mammals, with a focus on the development of miniature microscopes and endoscopes. In Chapter 3,1 discuss the advantages and methods of achieving rapidly tunable focus in both bench-top and miniature microscope systems. In Chapter 4,1 detail the development and testing of the C-FCM with an integrated EWTL, which was the first demonstration of using tunable lenses for 3D imaging in a miniature microscope. In Chapter 5,1 discuss the design and testing of the 2P-FCM with integrated EWTL, including results of awake and freely-behaving mouse neuronal imaging. I also discuss the challenges of performing two-photon microscopy through the CIFB. In the concluding Chapter 6,1 discuss the potential improvements and promising applications of the FCMs, including deep-brain imaging through miniaturized GRIN lenses and opportunities for increasing resolution and imaging speed.
I am fortunate to be working in an era of responsible animal research. All work presented here that involved live animals or animal tissue was approved by the Institutional Animal Care and Use Committee of the University of Colorado Anschutz Medical Campus, protocol # B-39615(05)1E.
4


CHAPTER II
MINIATURIZED MICROSCOPY IN NEUROSCIENCE
2.1 Fluorescence microscopy in the neurosciences
Image contrast
Arguably the most important property in microscopy, contrast quantifies the amount to which one can separate the foreground objects of interest from the background of an image. Because of the very small differences in refractive index found in whole tissues, cells such as neurons cannot be readily distinguished from surrounding brain tissues without some contrast agent or technique. Many such techniques have been developed in the past centuries to visualize neurons. Golgi’s method of silver staining was used by Santiago Ramon y Cajal in the 19th century to perform sparse labeling to allow visualizing neuronal processes in detail (Sotelo, 2003). Crucially, the silver that fills the cells does not impart contrast, but rather contrast is the result of the sparse cellular labeling, which allowed Cajal to see the expansive branching of a single Purkinje cell. An example of such a neuron is shown in Figure 2.1 (a). Today, a variety of histological stains achieve straightforward contrast in tissue slices to aid in disease diagnosis, research, and forensics. However, they are not suited for providing in vivo contrast because they often require trans-illumination, are rarely compatible with living tissues, and are difficult to target to specific cellular populations.
Although simple trans-illumination stains continue to be useful for histological studies, fluorescent probes have become the primary source of contrast for biological micros-
5


Figure 2.1: (a) Purkinje neurons as represented in the 19th century by Ramon y Cajal (Sotelo, 2003). (b) Modem fluorescence intensity projection image of a mouse Purkinje neuron filled with Lucifer Yellow fluorescent dye (Martone, 2002).
copy in the past several decades (Specht et al., 2017). Figure 2.1 (b) shows a fluorescence image of a Purkinje neuron as a modern comparison, showing the highly branched dendritic arbor and detailed spines along the processes.
Fluorescence microscopy
Fluorescence microscopy can be performed in epi-illumination mode, in which a single microscope objective lens is the conduit for both light illumination and collection. Fluorescent probes are excited using wavelengths that can be spectrally separated from the resulting fluorescence emission. Any non-fluorescing objects in the imaging field are dark, resulting in very large contrast ratios. Figure 2.2 shows a basic fluorescence microscope setup. The excitation and detection paths are combined spectrally with a dichroic beam-splitter, allowing the use of a single objective for epi-fluorescence imaging. This allows a greater variety of thick specimens, including whole animals, to be placed under
6


Camera
iiiiiiiiiiiiiiiiiiiii
Excitation-
Filter

Tube Lens Emission
Figure 2.2: Illustration of the epi-fluorescence microscope. Monochromatic excitation light, selected by an excitation filter, is reflected by a dichroic beam-splitter. An objective lens focuses the excitation light onto the sample and collects the emitted fluorescence from the fluorophores. The emission passes through the dichroic beam-splitter and emission filter before being captured on a detector, such as a camera, to form an image of the fluorescence.
the objective for imaging. Importantly, the imaging optics must be optimized for both exciting the fluorescence and retrieving the emission light.
The mechanism of fluorescence is dependent on the transition between electronic states of the fluorophore molecule. Figure 2.3 shows the Jablonski diagram for singlephoton fluorescence. A higher energy, shorter wavelength photon excites a fluorophore from the ground state (SO) to the higher energy state (SI), which will decay spontaneously by emitting a lower energy, longer wavelength photon that can be spectrally iso-
7


Figure 2.3: A simplified Jablonski-diagram showing electronic and vibrational states of a common fluorophore illustrates how a photon of the correct energy, related to wavelength (X ), is absorbed by a fluorophore to increase the electronic state from the ground state (SO) to the excited state (SI). Spontaneous vibrational relaxation is followed by fluorescence emission as the fluorophore recovers to SO, with the emitted photon having less energy and longer wavelength.
lated with colored filters. This shift in wavelength is called the Stokes shift, and is a fundamental property of fluorescent probes.
A powerful feature of fluorescent probes is that they can be combined to be multiplexed by color, allowing multiple independently addressable contrast agents to be accessible with the correct choice of color filters. There are thousands of fluorescent probes available spanning the ultraviolet to the near-infrared spectra of light (Specht et al.,
2017). They can be delivered to cells using a wide range of techniques, and can be made to localize to specific sub-regions of a cell. The molecular versatility of fluorescent molecules has allowed for groundbreaking innovations, such as super-resolution optical imaging and probes that have activity-dependent fluorescence.
8


Genetically-targetedfluorescent proteins
Fluorescent genetically-encoded Ca2+-indicators (GECIs) and optogenetic actuators (Deisseroth, 2015), continue to improve and they have become an indispensable part of the in vivo neuronal research toolbox. Prior to the recent functional advancement in GECIs, the primary tool of in vivo imaging were organic Ca2+-sensitive membrane-permeable dyes injected into the brain region of interest (Stosiek et al., 2003). The invention of GCaMP6 was a milestone that propelled the use of GECIs for in vivo imaging because it achieved a sensitivity and dynamic range greater than most organic dyes (Chen et al., 2013b). GECIs can be targeted to specific cell types, and can be delivered by virus injection that can be stable in expression for many months. Transgenic animals that constitu-tively express GECIs have also been made available (Chen et al., 2012; Dana et al.,
2014). In addition, optogenetic actuators, such as Channelrhodopsin and its variants, have been used to leverage these imaging systems to allow for patterned simultaneous stimulation of spatially selective networks (Papagiakoumou et al., 2013). Together, researchers can use the spatial and genetic specificity of these proteins to use a single optical system to control and read-out information from the intact brain (Emiliani et al., 2015). The remaining potential impact of these tools in neuroscience is in closing-the-loop between theoretical models developed from in vitro neuronal research and reality of the lucid and normally functional mammalian brain.
2.2 Modern microscopy techniques for studying the brain
There are many microscopy techniques available, and more are being continuously developed. In this section, I will describe two of the most commonly used for achieved
9


3D-imaging in thick samples. These laser-scanning techniques are also the basis of my FCM designs.
Laser-scanning confocal microscopy
Confocal microscopy was invented by Marvin Minsky in 1957, but only achieved widespread adoption and use in the 1990s, then known as laser-scanning confocal microscopy (LSCM), because of advances in laser technologies, fluorescent dyes, fast beam scanners, and sensitive photo-detectors (Minsky, 1957; Webb, 1999). LSCM is commonly used for reducing background signal during the imaging of fixed and fluorescently labeled tissue slices, but is also used for intravital microscopy in the context of clinical and research endoscopy (Jabbour et al., 2012).
The simplified LSCM setup is illustrated in Figure 2.4. LSCM involves the use of a spatial filter, usually a pinhole, in the conjugate focal plane to reject out-of-focus light. This dramatically improves imaging contrast by shrinking the effective depth-of-field (DOF) to close-to the diffraction limited axial focus of the imaging system (Webb, 1999). The size of the pinhole is optimized either automatically or manually to minimize the loss of signal light but also to reject the out-of-focus background that limits the imaging contrast. The pinhole size is chosen based on the size of the diffraction-limited spot magnified by the de-scanning optics (Conchello and Lichtman, 2005). This technique not only improves the axial resolution but also improves overall contrast by reducing the amount of background emission light collected from outside of the focal plane. An improvement in contrast is reported, resulting in a greater ability to resolve lateral features in some samples (Inoue, 2006). Details of the benefits of optical sectioning for brain tissue imaging are discussed in Chapter 3. Another version of confocal microscopy is spinning disk
10


Excitation
Light
C...^ >••• Pinhole
CT .I>-Tube Lens
XY-Translation
Stage
Figure 2.4: Confocal microscopy basic setup with stage-scanning. An optical pinhole in the conjugate focal plane is adjusted to reject out-of-focus fluorescence emission from a thick fluorescent sample. The laser is scanned across the sample, in this illustration with a translation stage.
confocal microscope, in which disk of pinholes and a camera is used to parallelize acquisition and increase frame rate (Toomre and Pawley, 2006).
In practice, LSCM requires additional mechanical and optical complexity to scan and image the laser spot across the sample. Typically, LSCM uses a mechanical lateral scanning system, such as galvanometric mirrors, to change the angle of the laser light entering an objective lens, which translates into a lateral movement of the focus spot across the sample. The fluorescence emission light must be de-scanned through the same scanning system, after which it is spectrally separated from the excitation light with optical filters.
11


The point-by-point detection is acquired by a single element detector, such as a photomultiplier tube (PMT), and mapped to a pixel value in the image. Fluorophores with spectrally-separable emission colors may be excited and acquired at the same time using different excitation lasers and emission filters along with additional detectors. Two-photon laser-scanning microscopy
The two-photon absorption effect was first described theoretically by Maria Goppert-Mayer in her 1930 doctoral thesis, but it was not verified until the invention of the laser later in the 20th century (Goppert-Mayer, 1931). After the technologies that gave rise to LSCM and further advancement of lasers, two-photon laser-scanning microscopy (2P-LSM) was ultimately shown by Winfred Denk et al. in 1990 (Denk et al., 1990). Since then, it has become one of the most important optical techniques in neuroscience. 2P-LSM allows researchers to reach targets in intact tissues that may not otherwise be accessible with other optical techniques. Like LSCM, 2P-LSM can also recover thin optical sections, but can do so with far less background excitation, a less complicated optical light path, and maintaining discernable contrast at remarkable tissue depths (typically ~500-pm depth in brain tissue).
2P-LSM uses a point-scanned focus to capture each pixel and construct an image. Figure 2.5 shows a basic 2P-LSM setup. Fluorescence excitation is achieved through the non-linear multiphoton absorption process. The excited electronic state of a fluorophore may be reached by the multiphoton absorption process. Multiphoton absorption occurs when multiple photons of fractional energies required to change electronic states impact the fluorophore near-simultaneously, resulting in exciting the molecule followed by nor-
12


mal fluorescence (Helmchen and Denk, 2005; Svoboda and Yasuda, 2006). The two-pho-ton process is shown alongside the normal one-photon process in Figure 2.6. The multiphoton absorption process is very unlikely at normal intensities, so the instantaneous photonic flux at the fluorophores must be made to be extremely high. To achieve this an ultrashort pulsed laser with pulse width <100 femtoseconds (fs) and high pulse energy is focused onto the sample.
Because of the non-linearity of the multiphoton process, the excitation efficiency is proportional to the square of the instantaneous peak pulse intensity. The time averaged
Ultrashort Pulsed Laser
Objective Lens Fluorophores
XY-Translation
Stage
Figure 2.5: 2P-LSM basic setup with stage-scanning. A pulsed ultrashort excitation laser is focused onto the sample, resulting in a non-linear excitation of fluorescence limited to the focal volume. The fluorescence is filtered and recorded for each pixel by a single-element detector.
13


S1
2A
excitation
w\*
excitation
emission
emission
w\*

\AAA>
2k
excitation
W\/>
so
One-photon process Two-photon process
Figure 2.6: One- and two-photon absorption process are shown to both cause a similar transition to the excited electronic state, resulting in similar fluorescence output. The two-photon process requires two photons to arrive near-simultaneously with approximately twice the wavelength, or half of the energy, of the energy gap.
fluorescence intensity generated by a pulsed laser source with average power Pavg, pulse width tp, and pulse repetition rate fP is approximated as (Diaspro et al., 2005):
where S2 is the fluorophore’s molecular two-photon cross-section, t] is the quantum efficiency of the fluorophore, h is Planck’s constant, and c is the speed of light in vacuum.
The numerical aperture (NA) is dependent on the acceptance angle, 6, of the imaging system in a medium with refractive index nx:
Rays at higher angles contribute more to higher spatial frequencies, so NA is the de facto descriptor for a system’s imaging resolution. However, in two-photon excitation terms,
(2.1)
NA = nx sin 6
(2.2)
14


Equation (2.1) shows a heavily non-linear dependence of fluorescence generation on NA. This is because the NA both contributes to the excitation intensity at the focal spot and is also the primary factor responsible for collecting the fluorescence light.
Typically, objectives used for two-photon excitation microscopy in deep tissues have NA >0.8 (Helmchen and Denk, 2005). This is a problem for the manufacture of small imaging systems, because the system aperture is constrained by the diameter of the physical optics, which in turns forces the optical design to compromise on other important parameters, such as working distance and field-of-view (Liang et al., 2002). Fortunately, the excitation and collection NA can be decoupled. The time-averaged intensity of fluorescence collected geometrically is:
(If,col) = (2-3)
where fIf is the fractional collection efficiency of an imaging system, summarized as:
When imaging into scattering tissue, the excitation NA may not be as important as the collection NA. This is both because high angle rays have longer optical paths, so have more chances to be scattered and affected by aberrations, and because there are diminishing returns as the focal volume decreases with high NA, resulting in too few fluorophores fluorescing. Therefore, a good efficiency two-photon imaging system can be designed that has a relatively low NA (-0.4-0.5) excitation path if the collection NA path can be increased independently. Such systems are often employed by using light collection paths
15


that involve large-area detectors, which can be replicated with large effective-area optical fibers in miniaturized microscopes (Wu et al., 2009; Zong et al., 2017).
2P-LSM has other pros and cons compared to other imaging techniques. The intrinsic spatial-confinement of the fluorescence excitation allows the rest of the optical system to be simplified, compared to LSCM, in the following ways:
• No de-scanning of the emission light. Fluorescence emission is spatially confined by the excitation, so no pinhole is needed to remove out-of-focus fluorescence.
• Imaging optics do not need to be achromatically corrected as carefully.
• Excitation wavelengths are much longer than emission, allowing larger bandwidth spectral filters.
On the other hand, 2P-LSM creates the following additional optical design challenges:
• An ultrashort pulsed excitation laser is needed, which cannot be miniaturized and may be deleteriously affected by propagating through long distances in glass.
• Requires special glass-coatings and optics that are efficiently transmissive in both the visible and near-infrared (NIR) wavelengths are uncommon.
• Detectors in the optical setup are much more sensitive to external light because of their wide detection apertures, so the optical path must be light-tight.
In summary, 2P-LSM is the most common optical technique for imaging deep into scattering while maintaining good contrast and optical sectioning. The unique challenges for two-photon fluorescence excitation and collection warrant designing an imaging system specifically for 2P-LSM.
16


2.3 In vivo brain imaging methods
Both LSCM and 2P-LSM are promising tools for microscopy because they offer good contrast at some depth in brain tissue. But in either case direct optical access to the tissues of interest needs to be attained surgically. Two common methods are described below.
Cranial windows
Cranial windows are used to provide optical access to surface-level brain structures with traditional microscope objectives. Typically, 2P-LSM is performed on a head-fixed mouse with a coverslip implanted in a small craniotomy above the brain region of interest, normally 3-mm2 or less in size, illustrated in Figure 2.7. There are several variations on windows that show the flexibility of this technique. A similar procedure is skull-thinning and polishing, which can be healthier and stable for the brain tissue, but is time-consuming and may not be as optically clear (Shih et al., 2012). Removable windows and permeable windows have also been described (Goldey et al., 2014).
P----... Glass window
V---
Objective Lens
Figure 2.7: Cranial window implanted in mouse to achieve optical access to cortical structures. An objective lens is used to image directly into the window.
17


Examples of common regions accessed by cranial windows are olfactory bulb (Wachowiak et al., 2013), barrel cortex (Peron et al., 2015b), and visual cortex (Andermann, 2010). These regions display high levels of behavior-dependent neuronal activity relatively close to the surface. The extended depth-limit of 2P-LSM, compared with widefield microscopy or LSCM, allows access to neurons about 500-pm below the brain surface. In terms of cortical layers in mouse, 500-pm is approximately the depth of the cell bodies found in layers 4/5 of structures such as the motor and somatosensory cortex, but does not reach deeper structures such as the hippocampus, located at > 1,000-pm below the dorsal brain surface (Paxinos and Franklin, 2012). To access deeper structures, it is necessary to excavate the tissue above the target structure or to implant relay optics, such as GRIN lenses.
Graded-refractive index (GRIN) lenses
GRIN lenses have been classically used for simple and space-efficient collimation of the divergent light exiting a single fiber core. In the past decade, GRIN lenses have been re-purposed to be used for many miniaturized imaging applications (Jung and Schnitzer, 2003; Knittel et al., 2001). The radial profile of the refractive index, ng, in a GRIN lens varies according to (Jung, 2004):
where g is the gradient constant. Rays launched into one end of the GRIN lens will be focused based on the axial length of the lens, defined by the pitch length, pL.
(2.5)
18


(2.6)
A pitch of 1 is a full sinusoidal period at a given wavelength. GRIN lenses are also available in a variety of NA values, which are defined by the focus angle in the external medium with index n1. This value can be approximated as (Jung, 2004):
NA,
grin
ng,ogd
nicsc(gL)
(2.7)
where d is the GRIN lens diameter. These expressions show that GRIN lenses with smaller diameter, d, need a shorter pitch length to achieve the same NA. To achieve a long and thin GRIN lens for deep-brain implants at high NA, it may require a very high pitch. To get around this constraint, a relay lens of lower NA is used in front of a high NA objective lens. Figure 2.8 shows an illustration of two kinds of GRIN lens assemblies used for imaging. A singlet lens is shown compared to a two-element lens with a low NA relay. By increasing the pitch number of the low NA relay, GRIN lenses of various lengths can be constructed for reaching deep-brain structures.
^~1.5-5 mm Single-element GRIN
Two-element GRIN
Figure 2.8: GRIN lens illustrations. Top: Single GRIN lens with pitch ~0.5 with a magnification of 1. Bottom: Two-element GRIN lens with low NA relay at pitch -0.75 and high NA objective with pitch -0.25.
19


GRIN lenses are engineered to be highly space-efficient, achieving relatively high NA (up to 0.5) for the small diameters (0.35 - 1.8-mm). The disadvantages are that they are highly wavelength-dependent and have rapidly deteriorating off-axis performance for imaging. This is mainly due to the small aperture, which vignettes the rays of off-axis light, effectively reducing the NA towards the edges of the lens. Longer GRIN lenses also suffer from accumulated aberrations that reduce imaging quality.
Even with the disadvantages, GRIN lenses offer a tool for providing optical access to otherwise deep-brain structures. Deep-brain implants of GRIN lenses have been used with both widefield fluorescence microscopy as well as 2P-LSM in structures such as the hippocampus (Barretto and Schnitzer, 2012a). Animals may be head-restrained with an appropriate objective used to image near the top surface of the GRIN lens, as shown in Figure 2.9. Usually a baseplate is implanted on the skull to provide stability to the fragile GRIN lens assembly.
Figure 2.9: GRIN lens assembly implanted into deep-brain of mouse and imaged with a low NA objective lens to couple into the low NA GRIN lens element.
20


2.4 Miniaturized microscope designs
It is becoming more important to perform physiological (both electrophysiological and opto-physiological) neuronal recordings in awake animals. This is because of the difficulty of assessing the functioning of neuronal circuits in anesthetized mice, in which the neuronal circuitry may function very differently compared to the awake state (Haider et al., 2012). In addition, to study a large range of behaviors it is necessary to record from animals that move and function normally during the experiment. This gap has been somewhat addressed by placing head-fixed animals on platforms or surfaces that allow them to move their limbs in a somewhat natural way. This also allows them to perform simple behavioral tests that are placed directly in front of them. Some researchers have also exploited virtual reality environments to increase the breadth of tasks that the mice may perform (Dombeck et al., 2010). However, it is more ideal (and sometimes crucial to the experimental question) to allow the mice to physically traverse an environment. To address this gap, the past 15 years or so has seen attempts to create lightweight, miniaturized microscopes that are designed to be affixed to the skull of a rodent for stable imaging during awake, mobile behavior.
Widefield miniature microscopes
A successful technology, in terms of adoption by researchers, market presence, and publications, has been the miniature widefield epifluorescence microscope. The basic design is a head-mounted microscope setup that incorporates a miniature camera, usually a CMOS sensor, a dichroic optical filter for spectral separation of the excitation and emission light, a single lens to form an image on the camera, and either on-board LEDs or
21


separately fiber-coupled excitation light sources. Notable examples of the widefield miniature microscope are the UCLA Miniscope (Cai et al., 2016) and the Inscopix nVista microscope (Ziv et al., 2013). Figure 2.10 shows the UCLA Miniscope. Images shown are from miniscope.org.
These widefield head attached miniature microscopes are capable of imaging the activity hundreds of cells using GECIs, owing to the large field-of-view of-500-pm and high DOF, which allows collection from many cells in multiple planes at once. Deep-brain structures are also accessible with the use of longer GRIN lens relays. This technique requires a temporally varying fluorescence to reliably resolve cells because of the large-amount of light scattering at the visible wavelengths and because of the excitation of a large amount of out-of-focus fluorescence. The low contrast of these microscopes means that morphological imaging is sacrificed to obtain high throughput imaging.
(a) (b)
Figure 2.10: Example widefield miniature microscope design from miniscope.org. (a) Internal components of miniature microscope (weight 2-g), with integrated LED, CMOS camera, filters, and lenses, (b) Photo of mouse with head-attached miniature microscope.
22


Laser-scanning miniature microscopes
LSCM and 2P-LSM are more suitable techniques for structural and high-resolution imaging. To take advantage of the benefits of laser-scanning microscopy techniques, a laser-source and point-scanning system are required. For the laser source, high intensity laser light must be brought into the miniature microscope enclosure using an optical fiber. Fibers can be selected for the most efficient propagation of the laser light, generally such that the fiber-core supports only the fundamental transverse electromagnetic (TEMoo) mode, in which case it is known as a single-mode fiber. In the special case of ultrashort laser pulses required for 2P-LSM, the maintenance of the peak pulse power is dependent on accounting for temporal dispersion, modal dispersion, polarization dispersion, and non-linear effects (Agrawal, 2000). These effects can be minimized by pre-compensating the pulses such that they cancel out through the fiber, and by choosing the correct fiber (Agrawal and Potasek, 1986; Clark et al., 2001). For example, hollow-core fiber, in which the electric field exists mostly in an air-filled core, is preferable for minimizing these effects for 1 - 5 m of fiber. The choice of fiber should also depend on its weight, flexibility, and cost for the practical purposes of mouse imaging. Regardless of the choice, a single fiber core must be focused onto the sample and actively scanned to form an image.
Fiber-scanning microscopy is a relatively well-developed field, owed to the progress in endoscopic surgical and diagnostic clinical tools that use miniaturized imaging heads coupled with fiber-optics to a detection system (Jabbour et al., 2012). There are many techniques for accomplishing this, two of the most common being piezoelectric resonance scanning of the fiber-tip and integrated microelectromechanical systems (MEMS)
23


actuated mirror scanning. Figure 2.11 (a) shows a simplified piezoelectric resonance scanner, in which the fiber tip is mechanically coupled to one or more piezoelectric actuators and forms a mechanical system that can be solved to predict the fiber-tip bending dynamics (Helmchen et al., 2013). Groups have shown the ability to create a variety of scan-patterns, including a spiral pattern, (shown in inset), (Engelbrecht et al., 2008; Myaing et al., 2006), a Lissajous pattern (Flusberg et al., 2005a), and even raster/arbitrary patterns (Rivera and Brown, 2011). Figure 2.11 (b) shows a simplified MEMS-based scanner with two actuated mirrors, each scanning one of the two orthogonal axes. MEMS actuated mirrors can be controlled precisely for arbitrary beam-scanning, and groups have
Figure 2.11: Fiber-scanning techniques, (a) Piezoelectric resonance-scanning of the fiber tip, resulting in the scanning of resonant modes, such as the spiral shown inset, (b) MEMS actuated-mirror scanners, resulting in arbitrary scan trajectories, such as the raster shown inset.
24


shown a variety of implementations to reduce size, power requirements, and overall complexity (Qiu and Piyawattanamatha, 2017). Recently, a miniature microscope for mouse imaging, (weight of 2-g), with a two-axis MEMS-actuated mirror was able to visualize dendritic spines in an awake-behaving mouse (Zong et al., 2017).
An extra consideration for these single-fiber delivery systems is that the single-mode fiber-core may be ineffective for the collection of fluorescence emission, which must usually be detected separately. Some solutions include the use of large mode-area fibers flanking the excitation core (Wu et al., 2009), miniature single-element detectors mounted on the microscope head (Helmchen et al., 2001), and a separate emission path for large-area collection fibers (Helmchen et al., 2013; Zhao et al., 2010).
Overall, these single-fiber-scanning techniques allow access to high-efficiency laser-transmission, which is especially useful for 2P-LSM because of the difficulty required in maintaining ultrashort pulse integrity at the focus. The cost of these techniques is the added mechanical complexity required to implement miniaturized laser-scanning. These may result in challenges in the optical design that limit imaging performance. Frequently, it is difficult to achieve large scan ranges or maintain a large beam size through small mirrors. So far, these systems have been discussed only in the context of transverse scanning. Miniature fiber-scanning microscopes that implement a third axial scanning dimension can become overly cumbersome in their complexity.
Coherent imaging fiber-bundle (CIFB) miniature microscopes
A CIFB is made from a longitudinally ordered bundle of optical fiber preforms that are drawn into dense canes with only a small amount of cladding between the cores (Wood et al., 2017). The resulting CIFB is spatially coherent on both ends, allowing a
25


pattern of light on one end to be represented by the cores on the other end, illustrated in Figure 2.12.
CIFBs can be coupled with a variety of simple imaging lenses to allow for full-field imaging. One of the simplest methods is to use optical epoxy to adhere an imaging GRIN lens (pitch ~0.5) to the distal end of the fiber bundle for use as a handheld imaging probe (Pierce et al., 2011). A fluorescence image of a Fujikura CIFB surface is shown in Figure 2.13. Lateral-laser scanning microscopy using a CIFB can be accomplished by simply scanning the excitation laser focus across the proximal surface of the fiber. The light is sequentially coupled into each core and transmitted to the distal surface. The distal surface can then be imaged onto a sample to create targeted excitation of fluorophores. The emission light can be collected through the same optical path. This setup is convenient because all the filters, detectors, and scanning optics can be located proximal to the fiber bundle, and can potentially be those of a standard bench-top microscope.
Proximal End
Distal End
Figure 2.12: Illustration of proximal to distal image coherence of CIFB, as well as the pixelation of an image formed through the bundle because of the discrete fiber cores.
26


0.55 mm
Figure 2.13: Fujikura CIFB fluorescence image of a uniform target showing core distribution.
Even with the gained simplicity, there are some disadvantages to using a CIFB compared to a single-core fiber for miniature laser-scanning microscopy:
• The CIFB introduces an unavoidable pixelation to the image due to the necessary cladding between the fiber-cores. Some techniques can be used to process the images to reduce this artifact, which can be summarized by some forms of low-pass spatial filtering (Cheon et al., 2014; Han and Yoon, 2011; Han et al., 2010;
Shinde and Matham, 2014). However, the fundamental lateral resolution limit of the imaging system is determined by the core-to-core spacing and magnification of the imaging system.
• To achieve small mode-field area sizes for single-mode propagation, it is necessary that the core size is small. Difficulties in manufacturing mean that most
27


small-core CIFBs have cores of different shapes, sizes, and propagation properties. This may result in random signal heterogeneity that shows up in the captured images.
• Core-to-core coupling is a phenomenon that occurs because of the close inter-core spacing in fiber bundles (Chen et al., 2008). This artefactual spreading of light intensity is most critical with broadband light in the NIR wavelengths, as used in multiphoton microscopy. This also limits the density of fiber-cores that can be used, and consequently the imaging resolution and field-of-view.
• Some CIFBs are much stiffer than a thin, single-core fiber, with the mechanical stiffness depending primarily on the diameter of the CIFB (and thus the number and density of the cores). This poses a challenge for developing a system for freely behaving rodents, as it limits their mobility in their behaving environment. Leached fiber-bundles, in which the cores are only fused at the ends of the bundle, may be a promising solution to this problem (Kostuk and Carriere, 2000).
While appreciating these challenges, CIFBs are used in the work in this dissertation because they greatly simplify the distal optical setup, which is desirable when implementing the distal axial focusing mechanism. Additionally, the use of a CIFB may greatly reduce the cost of adoption for fiber-coupled microscopes because it may allow researchers to use an existing bench-top microscope for fiber-coupled imaging with few or no modifications.
2.5 Where we are in miniaturized neuronal imaging?
Since the rise of in vivo imaging, as a result of the development of 2P-LSM, there has been a concerted effort to miniaturize these technologies for head-attached behaving
28


mouse neuronal imaging. Several designs had been proposed but the cost of adoption was not offset by the value of the results (Wilt et al., 2009). Meanwhile, the field of medical endoscopy has become a powerful technological and innovative force as a result of the desire for less invasive surgical tools and a greater emphasis on diagnostic and preventive healthcare (Qiu and Piyawattanametha, 2015). New tools, such as GCaMP6 and Chan-nelrhodopsin, motivated the neuroscience field to appeal for more usable miniaturized optical solutions. Fiber photometry and miniaturized wide-field microscopes have occupied that gap in the short term, because of the inherent simplicity of the approaches that lead to consistent results (Warden et al., 2014). Only a small number of laser-scanning fiber-coupled microscopes currently exist for neuronal imaging, but none have seen similar upward trends in usage. Further, none of the existing designs truly leverage the optical sectioning capabilities for 3D imaging without introducing cumbersome complexity. In the Chapter 3,1 argue that EWTLs offer an elegant solution for dramatically increasing the versatility of these devices by adding axial focusing.
29


CHAPTER III
AXIAL FOCUSING IN MINIATURIZED MICROSCOPES
Scientists have historically gravitated towards reducing the dimensionality of 3D objects down to a 2D-plane. Illustrations like Santiago Ramon y Cajal's famous sketches of neurons stained with Golgi's method were the original imaging modality, limited by the pen and paper medium (Ramon y Cajal, 1888, 1899). Today, we can reconstruct volumetric representations of tissues in virtual space. In Figure 3.1 (a), Ramon y Cajal illustrates the familiar layered cortical column structure. Figure 3.1 (b) is a modern 3D-representa-tion of a similar cortical column acquired with fluorescent labeling of neurons, three-photon excitation microscopy, and advanced software for 3D-reconstruction (Horton et al., 2013). The latter was acquired in 2013 at Cornell University, over a century following Ramon y Cajal's work. Neuronal networks exist in complex 3D-structures that cannot be fully represented by 2D-imaging. Beyond the conceptual benefits, 3D-imaging provides access to a larger number of cells and connections over greater spatial distances.
In this chapter, I discuss efforts to acquire 3D representations of neuronal structure and activity from intact brains using optical methods to actively tune the focal plane of a fluorescence imaging system. First, I justify the need for an active focusing solution in optical-sectioning microscopy. I briefly review the technologies that exist at the scale of the bench-top microscope, including the standard motorized-stages, remote-focusing, actuated objectives, and electrically tunable-lenses (ETLs). Thenceforth I specifically examine the technologies that are compatible with miniaturized microscopes and discuss the current implementations of these techniques in the literature.
30


Figure 3.1: (a) Sketch representation of the human motor cortex neuronal column by Ramon y Cajal (1899). (b) Modem 3D-reconstruction of a three-photon image of a cortical column in an intact mouse brain (adapted from (Horton et al., 2013)).
31


3.1 Benefits of minimizing depth-of-field in fluorescence imaging
Most optical imaging systems require light to be focused to a single surface to form a
cus, while other sources of light are defocused. In this first section, I present the motivation for an imaging system that discards the defocused light from the field in favor of increasing contrast for a single thin focal plane, also known as optical sectioning.
Lateral and axial properties of a focused Gaussian beam
A discussion of the focal plane of an imaging system can be simplified by decomposing the plane into an array of focused light beams. When light is focused with a lens, it is straightforward to intuit the confinement of the focus-spot in the 2D plane, (lateral dimensions), perpendicular to the optical-axis, (axial dimension), as a radially symmetric spot with peak intensity in the center. In the axial dimension, however, the shape of the focused light intensity is not as clearly confined.
To describe the confinement of the axial focus, I present a review of the 3D shape of a simple Gaussian beam. Below is the equation for the electric field (E) of a Gaussian beam propagating in free-space, the lowest order solution to the scalar wave equation, derived in-depth elsewhere (Bom and Wolf, 1999; Li and Kogelnik, 1966).
Here, Imax is the normalized peak intensity of the beam focus. The wavenumber constant, k, is equal to 2tt/A, where A is the wavelength of light. The beam width function, w(z), is
2D image. The focal distance is selected to choose a region of the field that is in sharp fo-
, VV Q
E(r,z) = fl
max ^yexp —i(kz - 4>(z))
max
(3.1)
32


the envelope at which the intensity drops to 1/e2 of the axial value. R(z) is the radius of curvature of the phase front and 4>(z) is the scalar phase. w(z), R(z), and 4>(z) are defined by the equations below:
w(z) = w0
R(z) = z
(3.2)
(3.3)
d)(z) = arctan ( ——7 I vnkw0 J
f 2z \
V
(3.4)
The factor zR in Equation (3.2) is the Rayleigh range, and is defined as the z-distance from the optical axis that the beam width, w(z), is V2 larger than w0. It is also the distance from the beam waist at which the beam intensity drops to half of its maximum value, Imax. zR can be related to the beam waist by way of the expression:
Zr =
Wo*
2k
ttWq
“T
(3.5)
The intensity distributions functions along the lateral and axial axes can be derived from Equation (3.1) by computing the magnitude squared at z = 0 and r = 0:
Iiat(r) = E(r,0):
= E
exp
j.2 ^
2 W0 V
(3.6)
33


I axial (z) = E(0, z)‘
= I
w l
maxw2(z) Imax
(3.7)

J
The lateral intensity function Iiat can be approximated as a Gaussian function. The lateral focus-spot full-width at half maximum (FWHM) radial size is approximately:
FWHMr = w02V2 In 2 s 2.35w0
(3.8)
These expressions can be related to the NA of the focused beam, which is defined by the angular beam divergence, 0, at z » zR given a specific index of refraction, nx, for the focusing external medium:
0
kw0
NA = n1sin(0)
(3.9)
(3.10)
These expressions can be rearranged to represent the system resolution in the axial and transverse dimensions as a function of the imaging system NA. First, the beam waist size w0 is related to the NA of the focusing system by reordering and substituting (3.10) into (3.8):
w0
nA
ttNA
(3.11)
34


Replacing this result into (3.8) and (3.5) yields:
2.35 nA
FWHMr
TT NA
(3.12)
FWHMZ - ^
(3.13)
Equations (3.12) and (3.13) are used to define the transverse and axial resolution of a Gaussian beam focused by an imaging system with a certain NA. The profile for w(z) is
shown in Figure 3.2 (b), showing the transverse FWHMr. The axial profile of the beam is shown in Figure 3.2 (c), showing the axial Rayleigh criterion, zR.
Importantly, the lateral imaging numbers for FWHMr are valid for the entire plane because those dimensions are radially symmetric. However, the axial imaging numbers for FWHMZ are only valid at r = 0, on the optical axis. Integrating the field at any axial position with a circle of increasing radius will eventually enclose the whole beam, meaning that the axial focus spot is never completely confined. The practical implication is that any source of light that intersects with this beam in the axial dimension may be detected, and may contribute to a low contrast, defocused background.
shown in Figure 3.2 (a). The transverse profile of the beam at the Gaussian beam waist is
35


w(z)
iz
I
(b)
Figure 3.2: (a) Gaussian profile of the w(z) function, indicating the x-axis parameter coo and the z-axis parameter zr. (b) Gaussian transverse intensity profile at beam waist (z=0), indicating the wo value and the related FWHMr. (c) Axial intensity profile along the optical axis (x=0), indicating the zR value and the related FWHMZ.
Benefits of optical sectioning for improved contrast
The 3D Gaussian focus described above describes the impulse-response of an imaging system, also known as the excitation point-spread function (PSF) in 3D space. The 3D excitation PSF can also be shown by the illumination pattern of a unitary fluorescent element, or one in which the diameter is « zR. An approximation of this is shown by imaging a 2-pm fluorescent microsphere embedded in clear agarose gel with blue light in Figure 3.3 (a). The light source is incoherent, so may not behave precisely according to the above functions, but the approximate beam waist and envelope are still clear. Notable is the large amount of out-of-focus fluorescence that is spread out laterally from the optical axis, even far from the diffraction limited zR. The Gaussian beam definitions require
36


(a) (b)
Figure 3.3: 2-pm fluorescent microsphere imaged under blue light illumination, (a) A widefield image of the fluorescent microsphere through the focus, showing the transverse (top) and axial (bottom) focus and the extended DOF. (b) A LSCM image of an identical bead showing the improved lateral resolution and confined axial spread.
that the power enclosed at the radius w(z) through the axial focus is constant, so the confinement of the axial light intensity is regulated by the slowly varying beam diameter. Although the focus spot along the optical axis is confined to |z| < zR , there is still significant fluorescence excitation far away from the beam waist. In other words, the widefield imaging system has a large depth-of-field (DOF), which is the distance in both directions from the optical axis at which there is still appreciable fluorescence from a focal plane. The DOF is dependent on optical parameters, such as NA and wavelength, but it is also dependent on the environment, such as density of fluorescent molecules and signal level.
Optical sectioning techniques bring the DOF closer to the diffraction-limited case.
For example, Figure 3.3 (b) shows an LSCM optically sectioned version of an identical
37


Figure 3.4: Power encircled at radius 5FWHMr through focus of 2 pm beads capture with wide-field (solid line, no optical sectioning) and LSCM (dashed line, with optical sectioning).
bead, in which the axial intensity spread is very similar along the optical axis, but the out-of-focus signal away from the optical axis is eliminated. Note that the lateral intensity profile of the bead also shows improved contrast compared to the widefield case, because of the rejection of light out-of-focus axial planes. Figure 3.4 shows a plot of power encircled at a radius of 5 times the FWHMr through the axial focus of each bead. In the widefield case, the total power along the optical axis is almost constant, showing that the small bead contributes backgrounds for a large axial range. In the optically sectioned bead, the power is much more confined, meaning that it will not create out-of-focus background after a much shorter axial range.
In actual tissue imaging there may be many fluorescent particles in the path of the excitation beam that can contribute to a loss of contrast. If the goal is to recover high contrast details from a focal plane in tissue, it is critical to use an optical sectioning method. Figure 3.5 shows a similar imaging setup but with a slide containing fluorescent pollen grains taken with a 1.42 NA 60x oil-immersion objective lens (Olympus) and green excitation light at ~532-nm wavelength. Each pollen grain is ~10-pm in diameter. Using these parameters, FWHMZ « 390 nm. This means that the pollen is very thick compared
38


Widefield (single plane) LSCM (max projection) LSCM (single plane)
(a) (b) (c)
Figure 3.5: Fluorescent pollen grains under green light excitation taken with a 1.42 NA 60x objective. (a) The pollen grains imaged under widefield illumination, showing the loss of contrast in the lateral dimension and a large DOF fetching out-of-focus grains, (b) A LSCM optical stack of several thin focal planes from the same pollen grains showing much higher contrast.
to the diffraction-limited focal plane. Although many of the fine features of the grains are still resolved under widefield illumination in Figure 3.5 (a), the contrast is much higher and the extraneous background signal is absent with LSCM, shown in Figure 3.5 (b).
The shallow DOF from the LSCM technique also allows remarkable cross-sections of the pollen grain, shown in Figure 3.5 (c). Meanwhile, the large DOF of the widefield technique fills in these gaps with lower resolution out-of-focus light, suggested by the lack of fine features resolved in the center of the pollen grain.
Focal plane size in LSCM and 2P-LSM
The tests above show the empirical results of optical sectioning with LSCM, but it is important to quantify the 3D PSF for the two primary modalities discussed here. In LSCM the axial intensity confinement is dependent on the diameter of the pinhole relative to the size of the diffraction-limited spot. The following expressions are the final derivations for that resolution in terms of the Gaussian focus (Cheng, 2006):
39


(3.14)
wO,CF
0.511
NA
ZR,CF
( 1.96A
(3.15)
The pinhole diameter is dph, and it is a separate term under the radical in Equation (3.15). For an ideal system the pinhole is infinitely small, and the axial spread is just the first term under the radical. In a realistic system, the pinhole is set to about 80% of w0 CF.
The two-photon 3D-PSF is summarized by the following expressions:
W0,2P
0.611
NA
(3.16)
ZR,2P
0.37ttA
NA2
(3.17)
Here, w0 2p is the Gaussian beam waist, related to the lateral spread, and zR2P is the Rayleigh range, related to the axial spread. Because of the longer wavelengths used for two-photon excitation, the lateral resolution ends up being about twice the value of the confocal lateral resolution. Although the axial resolution for LSCM is better with an ideal pinhole, when considering realistic parameters, the two modalities have very similar optical section thickness (Diaspro et al., 2005).
In summary, optical sectioning is technique that is used to reduce the DOF of an imaging system. Widefield imaging techniques are still able to resolve similar small features compared to LSCM, but the larger DOF results in reduced contrast. DOF depends heavily on the sample environment being imaged and is increased in the presence of many fluorescent sources. Notably, even in the absence of extra fluorescent molecules, the DOF of
40


widefield imaging is still much larger than the diffraction limited Rayleigh FWHMZ.
There are many layered sources of out-of-focus fluorescence in the brain. Diffuse fluores-cently labeled neuronal processes, such as in the cortical pia, can generate large amounts of unstructured background. In general, as the effective DOF becomes larger, performing optical-sectioning microscopy will have a greater impact on imaging contrast. Therefore, optical sectioning is an especially powerful tool for contrast-enhancement for in vivo brain imaging.
The cost of this contrast enhancement and shallower DOF is that an imaging system with a single focal distance will be unable to gather any information from regions outside of the focal plane. This limitation can be resolved by taking multiple images through the axial focus to generate a larger volume with high contrast.
3.2 Active axial focusing methods
Active focusing involves changing the focal plane during imaging. The number of discrete focal planes through which one can resolve meaningfully distinct structures is the thickness of the tissue (or the depth limit of the microscope) divided by the DOF. For tissue imaging, optical sectioning greatly reduces the DOF and therefore increases the number of distinct focal planes. This can both be useful, because it allows for better 3D separation of tissue structure. Active axial focusing combined with optical sectioning can be used to generate a 3D-representation of the tissues being imaged. Fast axial-focusing methods may also be used to create laser-scanning trajectories in 3D-space, allowing efficient sampling of cells in a volume rather than relying on a single plane-of-interest (Gobel et al., 2007; Nadella et al., 2016).
41


Scanning object Scanning imaging system Tuning focal length
Figure 3.6: Three fundamental axial focusing options. Object field scan changes the location of the imaging target relative to the optical system. Image field scan changes the imaging optics to change the focal plane. Tunable focus scanning changes the effective focal length of the optics to change the focal plane.
Three fundamental techniques that can be used to change the focal plane are shown in Figure 3.6. Most focusing techniques can be categorized under these methods. For inclusion in a mouse-attached device it must be miniaturizable and stable, which means object-scanning techniques are likely not workable. An additional consideration is that changing the focal plane may induce unwanted optical aberrations, since microscope optics are designed to function at fixed focal lengths. Considering these requirements, in this section I describe some of axial focusing methods that can be adapted for miniature microscopes and their methods of implementation.
Objective lens scanning
Axial scanning of the focal plane for laser-scanning microscopy can be achieved by moving the objective lens. Rapid mechanical movement of an infinitely-conjugated objective lens produces the fewest aberrations since objective maintains its designed focal length. Moreover, piezo-actuated objective lenses been shown with extremely fast axial focusing, with speeds up to 80 Hz for rapid sampling of hundreds of active neurons in a
42


volume (Katona et al., 2012). Several objective scanning techniques have been demonstrated for miniaturized imaging:
• Miniaturized motors for GRIN lens actuation, Figure 3.7 (a) (Flusberg et al., 2005a, 2008a)
• Shape memory alloys for lens actuation, Figure 3.7 (b) (Wu et al., 2010)
• Piezo-actuated lenses adapted for endoscopes (Sherlock et al., 2017).
• Linear motors with high speeds in hand-held endoscopes (Xie et al., 2006). Because of their relatively slow speed and larger size, lens actuators are typically
used to optimize the focal plane prior to initiating imaging by moving a lens axially in the imaging system. Higher speed motors and piezo-electric actuators of optics are fast but tend to be heavy relative to the weight of the miniature microscope, which makes them a challenge to include in a lightweight design. Motors may become a more viable choice as high-speed motors at small scales become more accessible.
Fiber
bundle
Focusing
motor

Gear
assembly
1 cm
, W
Focusing lens
.
Objective
Double-clad
fiber
Fixed
tube
Spring
Moveable
Tube
(a)
(b)
Figure 3.7: Examples of published lens-movement-based axials scanning, (a) Axial scanning by small focusing-motor actuation with 1.1-mm axial motion (Flusberg et al., 2008a). (b) Lens actuation by shape-memory alloy contraction, with 150-pm of axial motion (Wu et al., 2010).
43


MEMS devices
MEMS devices are nano-fabricated mechanical actuators that can be designed for complex behavior at a sub-millimeter scale. Qiu and Piyawattanamatha comprehensively reviewed existing MEMS-based endoscope devices, including current solutions for MEMS-based axial focusing (Qiu and Piyawattanamatha, 2017). Examples of MEMS-based axial-focusing methods include:
• Axial actuation of a miniature lens (Liu et al., 2010).
• Refocusing with MEMS variable focus membrane mirrors (Shao et al., 2004).
• Thin-film piezoelectric actuator with rotational scanning, shown in Figure 3.8 (a) (Qiu et al., 2014).
• A dedicated Z-axis mirror for axial actuation, shown in Figure 3.8 (b,c) (Li et al., 2016)
Although MEMS-based microendoscopes for small animal neuronal imaging have been developed, (see Chapter 2), none of the above axial integration techniques have been demonstrated in this application. MEMS-based 3D miniature microscopes for neuroscience work are very promising due to their versatility and small size. However, the optical difficulty of refocusing with a mirror in a small space can be cumbersome for implementation.
44


Figure 3.8: Examples of published MEMS mirror devices with axial focusing, (a) Thin-film piezoelectric actuator for 190-pm of vertical translation with parabolic mirrors for excitation and collection (Qiu et al., 2014). (b) Monolithic 3-axis MEMS scanning mirror assembly, with vertical translation photos shown in (c), achieving 546-pm of vertical translation (Li et al., 2016).
Tunable lenses
Optical focusing with tunable lens is different from mechanical translation of the imaging optics relative to the specimen. Most commonly, optical focusing is accomplished with an electrically tunable lens (ETL) placed near the back-aperture of the objective lens. Changing the phase curvature of the beam entering the back of an objective lens with an ETL translates the focal plane. In most ETL, there is a minimal inertial mass that moves during refocusing, which makes them ideal for high-speed applications. Large-aperture ETLs are becoming a common addition to an objective lens for rapid focus scanning with no moving parts (Chen et al., 2014; Fahrbach et al., 2013; Grewe et al., 2011; Jabbour et al., 2014; Jeong et al., 2016; Ryan et al., 2017).
Examples of the use of ETLs in miniature microscopes include:
• IR light responsive lenses for fiber-coupled focusing (Zeng and Jiang, 2009)
45


• A liquid filled electrically tunable membrane attached to a piezo-actuated fiber-coupled endoscope (Meinert et al., 2014)
• Shape-changing polymer tunable-lens used in a laparoscopic imaging device (Volpi et al., 2017)
Using ETLs has the distinct advantage of having a minimal amount of moving parts and can be incorporated into most optical paths. Careful placement of the tunable lens near the aperture stop of the imaging system will minimize the aberrations introduced to the light rays arriving from different angles for lateral scanning. Aside from needing to consider the effects of the divergent beams on the imaging properties of the objective lens, ETLs generally introduce their own aberrations which need to be considered.
Other axial scanning methods
This section will introduce several other techniques that have not been implemented fully in miniaturized microscope systems, but may be compatible with such a system.
Temporal focusing is uniquely useful for multiphoton microscopy. This method involves engineering the spectral chirp such that the highest intensity focus is at a defined focal plane (Oron and Silberberg, 2015). This focal plane can be modulated rapidly, and the spectral properties can be calibrated and maintained for transport through an optical fiber, which obviates the need for a distal scanning mechanism. It also minimally disrupts the optical properties of the imaging objective. Temporal focusing is still limited in actual scan range, defined by the bandwidth of the incoming pulsed laser light.
Wavefront shaping can produce exotic laser focuses and can be tuned rapidly with a high refresh-rate spatial light modulator (SLM) (Horstmeyer et al., 2015). Some examples of focusing patterns include being able to focus at multiple axial and lateral positions
46


simultaneous and creating arbitrary spatially selected patterns for simultaneous optoge-netic excitation and imaging (Quirin et al., 2014). Wavefront shaping is also powerful since it can be propagated through some optic fibers. Some groups have shown three-dimensional scanning of a focal spot through a fibers with proximal wavefront shaping such that a specific focus pattern is form on the other end (Kim et al., 2014; Porat et al., 2016; Stasio et al., 2015). The primary challenges in wavefront shaping for a fiber-coupled is that it is heavily affected by the mechanical bending of the fiber, and must use optical feedback to rapidly alter the shaping pattern to compensate for changes imposed by the fiber and the tissue. If this problem can be solved it can obviate the need for any distal optics, which would be the ideal system.
Fixed multifocal systems implement multiple focal plane imaging intrinsically in the design. Several focal planes can be resolved by staggering optic fibers axially, so each fiber addresses a different focal plane (Rivera et al., 2012). Another demonstrated approach is to design the lenses such that multiple focuses are form from a single optical path (Ouzounov et al., 2013). These systems are mechanically simple and space-saving but they are limited in that there is no easy way for continuous axial scanning.
Remote focusing is a promising technique that has already become an effective solution for larger microscope systems. Botcherby et al. described the use of a second objective lens in a conjugate region of the primary objective to form a 3D image of the sample, and then to use a moveable mirror at the second objective to select the focal plane dynamically (Botcherby et al., 2008). This system is attractive for miniaturization because the second lens only needs to be as large as and move as far as the secondary image, which makes small technologies like MEMS-based mirrors an attractive option.
47


3.3 Tunable focus by liquid lens technology
Liquid lenses are a common variety of ETL that use one or more liquids as a shapechanging refractive interface to allow for focus adjustment with minimal mechanical motion (Chiu et al., 2012). Liquid ETLs have been used in several microscope systems to accomplish high-speed scanning. One example is their use in selective-plane illumination microscopy (SPIM), which deals with large, high-resolution imaging volumes acquired rapidly and directly benefits from the speed of the tunable lens (Fahrbach et al., 2013). Liquid ETLs have also been used as an alternative to mechanical scanning in live-imaging systems for bench-top laser-scanning microscopes.
There are several common liquid ETL types that may be used for performing highspeed optical focusing for laser-scanning microscopy (Blum et al., 2011). Large-aperture liquid ETLs are particularly suited for bench-top systems because of their large focal range, high speed, and repeatability. An example of a commercially available 10-mm aperture, 30-mm diameter ETL (EL-10-30-C-MV, Optotune AG, Switzerland) is shown in Figure 3.9 (a). This ETL functions by rapidly changing the volume of liquid in a container with a flexible polymer surface, which results in a focal length shift. Large aperture liquid ETLs have enabled rapid axial focusing, up to 100 Hz, in compact LSCM, 2P-LSM, and SPIM systems without mechanical movement of the objective or sample (Jabbour et al., 2014; Jiang et al., 2015; Rickgauer et al., 2014; Ryan et al., 2017). However, shape-changing polymer ETLs are not suitable for miniaturized head-attached microscopes because of their large size and susceptibility to orientation and vibration.
48


+3.5 rrr1
7.8 mm
ter—^ +36 rm1
B>°m-1
\ l /_
--15 m-1
t t |
(b)
Figure 3.9: Two varieties of commercially available liquid ETLs. (a) Optotune EL-10-30-C-MV ETL based on a shape-changing polymer surface, with outer diameter of 30-mm and 10-mm aperture. (b) Varioptic Arctic 316, based on electrowetting lens technology, with outer diameter of 7.8-mm and 2.5-mm aperture.
Electro wetting tunable lenses (EWTLs) are another type of liquid ETL that has applications in microscopy (Zhao and Wang, 2013). An example of an EWTL (Arctic 316, Varioptic, France) with outer diameter 7.8-mm and 2.5-mm aperture is shown in Figure 3.9 (b). Electrowetting is a method for changing the wettability of a liquid on a dielectric surface by applying a voltage across the interface, effectively changing the contact angle of the liquid with the surface. To form a lens, two fluids with dissimilar refractive-indices and hydrophobicity are placed in a container with dielectric side-walls. The hydrophilic polar liquid has dissolved impurities to allow it to react to an external electric field. The contact angle of the liquid interface to the side-walls of the lens can be changed by applying an electric field across the lens. Eventually an equilibrium is reached between the electrostatic forces acting on the polar liquid and the surface tension of the system to create a lens surface with a stable curvature.
Carefully matching the density of the liquids at these scales makes EWTLs very resistant to the effects of gravity due to orientation and vibrations, which make it ideally
49


suited for a miniature head-mounted system. Commercial EWTLs have also demonstrated very large tuning ranges, on the order of 50 diopters. Further, the responses of EWTLs to input voltages are well-described by simple oscillators. Although EWTL focusing speed is not as rapid as some other ETLs, engineering the voltage input function, the lens response time can be brought to less than 20 milliseconds, which is within the range of what is required for an active scanning system (Supekar et al., 2017a). EWTLs also have the potential to perform extended optical functions, such as beam steering and wavefront shaping (Smith et al., 2006; Supekar et al., 2017b). Because of the small aperture size, these lenses are not frequently used in microscopy applications. However, EWTLs are good candidates for robust axial focusing solution for miniature FCMs.
3.4 Optical design considerations for axial scanning with a tunable lens
Unlike the method of physically moving the samples or the imaging lenses, tuning the axial focus changes the way light propagates through the optical system to change the focusing distance. More specifically, the curvature at the back focal plane of the objective lens is modified. When transformed through a lens, this curvature of the wavefront gets transformed into an axial displacement of the paraxial focus spot. This is analogous to how a tilted wavefront becomes transformed by a lens into a transverse displacement of the focus.
One important consideration when designing such a system is the magnitude of curvature that is needed to achieve a desired axial scan range. An ETL placed in the back focal plane of a lens directly shapes the quadratic wavefront entering the transformation.
50


The wavefront phase term cj) can be written as a function of the pupil radius p
(Botcherby et al., 2007):
NA2
2M2
(3.18)
where M is the objective magnification and Zc is the distance at which the focus is formed with a given quadratic phase. Rearranging this in terms of the amount of axial change for a given change in wavefront phase curvature gives:
Zr
2M2
kNA2
(3.19)
This last expression shows that the ability of a tunable lens to change the axial focus of a lens system is dependent inversely on the square of the objective magnification. This means that designing for high magnification with the purpose of increasing resolution or NA will result in a decrease in axial scan range for a given ETL.
Another consideration is that any imaging system that produces magnification cannot perfectly represent both the axial and transverse focus transformations simultaneously. This effect is illustrated in Figure 3.10. Botcherby et al. describe the consequences of this on imaging properties (Botcherby et al., 2008). In real lens design scenarios, optical engineers will optimize for either transverse or axial focusing. Most commonly transverse imaging is preserved because imaging lenses are expected to have uniform performance throughout their FOV. The consequence is an accumulation of spherical aberrations along the axial range of the lens, shown in Figure 3.10 (c). One can also design for the Her-schel condition, which instead forms perfect focus spots along the optical axis, but accumulates aberrations in the transverse dimension.
51


f
Designed focus
Figure 3.10: Axial focusing by wavefront curvature shaping at the objective back focal plane, (a) An input converging or diverging wavefront is transformed into an axial translation away from the designed focal length of the lens, (b) Spherical aberrations as a result of a lens obeying the sine condition.
In the FCM imaging system designs presented here, I optimize primarily for the Her-schel condition, by constraining the optimization parameters in Zemax optical design software to maintain consistent imaging properties throughout the full focal range of the imaging system. One important characteristic that is maintained is telecentricity, which ensures that there is nearly no magnification change as the focal length is tuned. However, this choice sacrifices some imaging performance in the transverse dimension, and so imaging performance is maintained only up to the expected FOV prescribed by the CIFB diameter. Further, it is necessary to balance high magnification, which is required to achieve high resolution with the fiber-core spacing, with axial scanning range. In the
52


end the relationships between aperture size, focal length, transverse and axial consistency, and magnification become a careful interplay of compromises to reach a useful solution for miniaturized imaging. In this work the actual optical designs are sought empirically based on initial designs and guiding principles. The designs are restricted by the parameters laid out in the following chapters for size, resolution, FOV, and axial scan range. I also restrict the designs to commercially available optics to allow for easier replication and more affordable designs.
53


CHAPTER IV
CONFOCAL FIBER-COUPLED MICROSCOPE FOR 3D IMAGING
Confocal fluorescence microscopy is a technique that intentionally rejects fluorescence emission light outside of the desired focal plane. This effectively reduces the DOF to a close-to diffraction-limited axial sheet. LSCM is the laser-scanning variant of confocal fluorescence microscopy. When LSCM is combined with axial scanning, multiple layers can be combined to produce a 3D-representation of the sample being imaged.
More details about the background of LSCM can be found in Chapter 2. Although this method has limited imaging depth in tissue compared with 2P-LSM, the unique optical path offered by the CIFB may allow for a practical 3D neuronal imaging solution.
In this chapter I describe the design, construction, and testing of a miniature confocal fiber-coupled microscope (C-FCM). The C-FCM design takes advantage of the pinholelike structure of the CIFB to replicate confocal imaging in a miniature imaging system. An integrated EWTL allows the C-FCM to capture multiple axial planes to extend its capability to 3D imaging. The optical components of the device are combined in a custom 3D-printed adapter with an assembled weight of ~2 g that can be mounted onto the head of a mouse. Confocal sectioning provides an axial resolution of ~12-pm and an axial scan range of ~80-pm. The lateral field-of-view is 300-pm and the lateral resolution is 1.8-pm. Results of testing the C-FCM show bead images to quantify the resolution and scan range of the system, as well as ex vivo tissue 3D-imaging of mouse peripheral olfactory nerve labeled with yellow-fluorescent protein (YFP).
Some of the content in this chapter was previously published in the journal Optics Letters in 2015 (Ozbay et al., 2015).
54


4.1 Miniaturizing confocal laser-scanning microscopy
The primary advantage of LSCM is the ability to extract thin optical sections; important for imaging cells in thick tissue, as argued in Chapter 3. Briefly, optical sectioning combined with axial scanning enables the collection of 3D data from tissue and increases contrast that is required for imaging structural details of densely labeled samples in thick tissues, such as neurons in the intact brain. LSCM typically requires a complicated light path, which makes it difficult to miniaturize.
Confocal imaging through a fiber-bundle
For most LSCM systems, emission light from the sample must be de-scanned via the excitation path and then spatially filtered through a pinhole at a conjugate focal plane (Paddock, 2000). Optical fiber cores have been used as confocal pinholes in many designs because they can simplify the detection path (Maitland et al., 2006; Meyer et al., 2016; Stelzer, 2006). To reduce the size and number of optics, it has been shown that a fiber-bundle can be used as an array of pinholes for spatial filtering, greatly simplifying the number of optics required distal to the fiber (Gmitro and Aziz, 1993).
Axial scanning with EWTL
ETLs have been shown to be capable providing axial scanning with no mechanical actuation for standard microscopy applications (Chen et al., 2014; Fahrbach et al., 2013; Jabbour et al., 2014; Koukourakis et al., 2014; Meinert et al., 2014). EWTLs are a type of liquid ETL that are highly resistant to motion are small enough to incorporate in a miniature C-FCM system for active axial scanning (Zhao and Wang, 2013).
55


4.2 Methods
Experimental setup
Figure 4.1 shows the imaging setup designed to incorporate a commercial EWTL in the focusing optics. A laser-scanning confocal microscope (Leica SP5 II) provides the continuous-wave (CW) laser for fluorescence excitation, steering of the beam laterally using a resonance beam scanner, and spectrally filtered detectors for fluorescence imaging. To fiber-couple the device to the microscope, I use a high-density CIFB with 0.5 m length, 30,000 count fiber cores, total effective imaging diameter of 0.8-mm, ~2.9-pm core diameter, and 4.5-pm inter-core spacing (Fujikura, FIGH-30-850N). The excitation laser is focused onto the proximal surface of the fiber-bundle using a 10X 0.4 NA Olympus objective lens. Imaging over a single focal plane is performed by raster-scanning the
Image
Plane
0.4NA,10X
Objective
Imaging Fiber Bundle
0.21 -0.29
Object Plane
Figure 4.1: Laser scanning confocal microscope coupled to distal imaging optics through fiber bundle (IL: Imaging lens, EWTL: Electrowetting tunable lens, OL: Objective lenses).
56


laser into individual fiber cores. At the distal end of the CIFB, the light from the cores is collimated with a 3-mm focal-length achromatic lens (Edmund Optics, 65-566), passes through the EWTL, and is focused onto the sample using two 2-mm focal length achromatic lenses (Edmund Optics, 65-565). The imaging system has -2.5X magnification.
For each fiber core, fluorescence emission is transmitted back through the optics into the same fiber core and registered on the confocal microscope internal photomultiplier detector. The EWTL is placed in the infinity space of the telescopic imaging system, where changing the focal length of the EWTL by voltage results in a shift of the front focal length of the C-FCM. To control focus is a commercial EWTL (Arctic 316, Varioptic Inc.), which has a focal length that ranges from -57 to +29-mm (-17 to 36 m"1 in diopters) corresponding to a voltage input from 25 to 60 Vrms provided through a flexible lens cable. Typical power draw of the EWTL is ~15 mW.
A custom-fabricated two-part plastic adapter was designed (shown in cross section in Figure 4.2 (a)) to provide a lightweight and rigid enclosure for the imaging optics, including the 7.8-mm diameter EWTL package. The adapter is designed to be surgically attached on a mouse head with the 1-mm objective lens to be inserted for deep brain imaging. The top section is easily detachable, and the bottom section is designed with a low profile < 5-mm height) such that it may remain implanted for long-term imaging. The EWTL and electrode are clamped between the two adapter sections and held in place with an O-ring. The EWTL is separated from the skull by ~2-mm, which provides adequate insulation from the electrical contact. The objective lenses are glued with cyanoacrylate adhesive into a length (3 - 7-mm) of polyimide tubing with 1.06-mm outer diameter and fit into the bottom of the holder. Figure 4.2 (b) shows a photograph of the
57


Figure 4.2: (a) CAD model of adapter with optics with lenses aligned in polyimide (PI) tubing, (c) Photo of assembled adapter and EWTL.
assembled C-FCM after being manufactured via a 0.1-mm tolerance 3D-printing process followed by minor reworking. The adapter maintained good optical alignment over repeated assemblies. The final assembly, including the lenses, EL, and electrode, weighs approximately 1.9 g, comparable in size and weight to current miniature head-mounted microscopes for awake-behaving imaging (Piyawattanametha, 2014).
Optical design
I used Zemax optical design software to compute the optical performance of the C-FCM with the objective lens extending 1-mm below the adapter, for surface level imaging, and extending 4-mm below the adapter, for deep-tissue imaging. The adjustable design permits access to regions of interest in the brain. The Zemax models for the doublet lenses were obtained from Edmund Optics. The Zemax model of the electrowetting lens was obtained from Varioptic. There is low chromatic aberration in this system due to the
58


selection of achromatic lenses and there are minimal changes for wavelengths ranging from 500-600 nm. Results from the design are summarized in Table 4.1.
Table 4.1: C-FCM optical parameters from Zemax model
Objective length (mm) Scan range (jim) Magnification Lateral resolution (lim) Field-of-view (jim) NA
1 211-288 2.4-2.6x 1.8- 1.7 -300 0.36-0.35
4 212-290 2.2-2.7x i o cs -240 0.41-0.35
The lateral resolution of 1.8-pm is determined by the fiber-bundle core-spacing (4.5-pm) de-magnified by the imaging system. The optics are required to have good performance only up to the maximum spatial frequency resolvable by the fiber-bundle. The spatial frequency is represented as line pairs per mm (lp/mm), which is calculated as iooo [im/mm _ ^qq ip jmm The effective field-of-view (FOV) is calculated by finding
1.8 [1771
the minimum object field diameter at which the modulus of the optical transfer function (MTF) at 300-lp/mm fails to meet the Rayleigh criterion (MTF > 0.2) (Born and Wolf, 1999). The model shows a constant field curvature that results in a ~ 20-pm axial focus shift at the edge of the FOV compared to the center, which is managed by imaging thick tissue (> 20-pm thickness) or by combining multiple optical sections. As the objective lens distance is increased, there is an increase in off-axis vignetting and aberrations, most significantly astigmatism and distortion. The FOV at the longer objective length is predicted to be smaller due to the increased aberrations and vignetting.
59


4.3 Results
Optical performance testing
I evaluated the FOV experimentally by imaging a 15-pm-thick section of mouse brain showing neuroglial oligodendrocyte cells expressing green fluorescent protein (GFP). Figure 4.3 (a) shows a standard confocal fluorescence image of the sample using 488 nm excitation light. To compensate for field curvature, I performed a maximum intensity projection of 3 - 4 axial planes at once, obtained by varying the EWTL focus. Figure 4.3 (b) and Figure 4.3 (c) show images for the 1-mm and 4-mm objective lens distances. At the 4-mm distance there is a reduction in FOV due to off-axis vignetting and aberrations. The predicted FOV from Table 4.1 is marked with a dashed circle. I experimentally measured the power loss across the FOV using a thick fluorescence test slide (Chroma) shown in Figure 4.4. The decreasing signal is because of vignetting and aberrations towards the edge of the FOV and was unchanged when varying the axial focus with the EWTL. From these results I posit that increasing the objective length only reduces the FOV while otherwise maintaining similar image quality.
(a) (b) (c)
Figure 4.3: Lateral imaging characteristics of C-FCM at different objective depths, (a) GFP labeled neuroglia imaged with 20X, 0.8 NA objective, (b) Same region imaged with C-FCM at shortest and (c) longest objective lens length.
60


-1
Radial fraction of fiber surface
H
1
Figure 4.4: Imaging power loss over the full FOV with 1-mm distance objective lens (black line) and 4-mm distance objective lens (grey line).
The images were processed by band-stop filtering using a spatial fast Fourier transform (Matlab) to remove the fiber-pixelation artifact, illustrated in Figure 4.5. Individual cell bodies and neuronal processes can be resolved in the image, showing sub-cellular resolution, comparable to the standard confocal microscope images obtained with a 10X 0.3 NA objective lens. The effects of distortion and NA variations are not visible within the defined FOV.
I experimentally verified the axial scan range of the C-FCM. A nanopositioner stage (Mad City Labs, LP100) with 100-pm z-scan range was used as an axial ruler. I imaged a thick test sample consisting of 1 -pm diameter red fluorescent beads (Invitrogen, F8887), embedded in agarose gel, using 561-nm excitation light. Initially, the EWTL was fixed at
Raw FFT 10x
. : Fiber • Filter 0.3NA
* %
4
• . • •
Figure 4.5: Zoom-in showing the cell bodies indicated by white arrow in Figure 4.3 (a). Left: Raw C-FCM image showing pixilation due to fiber-cores. Middle: C-FCM image after filtering. Right: Comparison with standard confocal microscope.
61


the shortest focal length and I obtained axial image sections at 1-pm intervals using the nano-positioner stage. Next, I fixed the sample position and obtained 36 optical sections of the same regions by varying the EWTL across the full focal range. Figure 4.6 (a) shows an orthogonal projection of several beads resulting from varying the stage position and the EWTL focal length. Figure 4.6 (b) shows the mapping of bead axial centroids from the stage position to the EWTL focal setting. Data collected from 40 beads agree with the simulated focal length dependence obtained in the Zemax model. I conclude that the C-FCM provides a scan range of approximately 80-pm.
I tested the lateral and axial resolution by imaging agarose samples containing 2-pm or 1-pm red fluorescent beads. The confocal pinhole on the microscope was set either to the open position, allowing light collection from multiple fibers, or closed to a setting of
Z-Stage EWT Lens
Cl
Q
N
0
20
40
60
80
100
100 pm 100 pm
fa)
£
1—
CD
$
O
CL
03
O
Cl
o
(b)
Figure 4.6: Axial scan range of electrowetting C-FCM (a) Orthogonal projection (inverted grayscale) of 1-pm diameter red fluorescent beads in agarose imaged with scanning Z-stage (left) and scanning the EWTL (right), (b) Black dots: 40 beads mapped from EWTL optical power to relative Z-position. Gray line: Simulated C-FCM focal length with varying EWTL power.
62


2 Airy patterns, which allows light collection from only one core. Figure 4.7 (a) shows lateral and axial bead images. With the pinhole open, out-of-focus fluorescence emission leaks into adjacent fibers, but is eliminated by closure of the pinhole due to the confocal sectioning capability provided by the fiber cores. Figure 4.7 (b) shows the profiles of beads that were imaged with the closed-pinhole configuration. The theoretical resolution limit is shown in grey using the equations for lateral resolution in Equation (3.14) and the pinhole-limited axial resolution in Equation (3.15), repeated below in terms of FWHM.
FWHMrCF
0.511
NA
(3.14)
fwhmzCF
0.881
n — Vn2 — NA2
(3.15)
Open 1 pm 2 pm
Pinhole bead bead
10 pm 10 pm 10 pm
Lateral
(a)
(b)
Figure 4.7: Lateral and axial resolution of C-FCM. (a) Lateral and axial images of fluorescent beads. Left: I -pm bead with open pinhole (PH), Middle: I -pm bead with closed pinhole, Right: 2-pm bead with closed pinhole, (b) Top: Averaged line profiles of 1- and 2-pm beads (black) compared with diffraction-limited resolution (grey) Bottom: Averaged axial profile of several 1 and 2-pm beads (black) compared with theoretical axial resolution (grey).
63


The calculations use object-space NA = 0.35, refractive index of air n = 1, wavelength X = 600-nm, and pinhole diameter dPh = 1-pm (2.5-pm fiber-core size de-magni-fied by 2.5x). As shown in Figure 4.7 (a) the 1-pm diameter beads occupy a single fiber core while the 2-pm beads are sampled by multiple cores. This indicates that the C-FCM lateral resolution is fundamentally limited by the fiber core spacing. The lateral resolution is estimated to be 1.8-pm, determined by de-magnifying the 4.5-pm core spacing by 2.5X, compared to the theoretical 0.87-pm resolution from Equation (3.14). As shown in Figure 4.7 (b), the FWHM axial resolution is determined to be ~10- and ~12-pm for the 1- and 2-pm diameter beads respectively, compared to the theoretical 9.3-pm resolution. Cross-talk of multiple fiber light collection may explain the difference in axial resolution. Tissue-sample testing
To demonstrate imaging in tissue with the C-FCM, I imaged intact mouse olfactory nerve fibers expressing yellow-fluorescent protein in olfactory sensory neurons (Li et al., 2014). The mouse was sacrificed by CO2 inhalation, according to existing protocols, and the head was bisected sagittally to expose the olfactory epithelium and nerve. The C-FCM was held in position adjacent to the tissue using a manipulator arm with an aqueous saline solution interface between the objective lens and tissue. Imaging was performed using a 488-nm laser at a resolution of 1024 by 1024 pixels at 1.7 seconds/frame. 36 image slices were taken while varying EWTL optical power from 13 to -5 m"1. The images were post-processed by band-stop filtering. Figure 4.8 (a) shows four separate optical sections spanning ~50-pm, limited by light scattering in the tissue. Each of these four images represents a ~12-pm optical section with distinct morphological features, demonstrating efficient optical sectioning with this device. Figure 4.8 (b) shows a maximum-
64


intensity projection of a stack of optical sections. The diameter of each axonal bundle is 10-pm and is easily resolved.
Figure 4.8: 3D-imaging of mouse nerve tissue, (a) Four optical sections that were taken at specific EWTL optical power settings. Scale bar is lOO-pm. (b) Maximum intensity projection of an image stack of intact olfactory neuron axons labeled with YFP.
4.4 Summary
This is the first demonstration for the use of electrowetting tunable focus lens technology in a fluorescence C-FCM to allow full 3D tissue imaging. I validated a FOV of 300-pm, and ~1.8-pm lateral and ~12-pm axial resolution. I verified experimentally an axial scan range of 80-pm. I showed 3D images of detailed nerve fibers showing axonal networks. The C-FCM enclosure is a simple and lightweight adapter that can be modified for use as a head-mounted device for brain imaging. Further improvement of low-voltage EWTL technology will allow for more diverse endoscopic applications (Niederriter et al., 2013). This approach is promising for use as an implantable device because it has no mechanical fatigue, no vibration, and low power consumption.
65


This C-FCM concept with a tunable lens can also be applied to multiphoton imaging of fluorescent indicators, but require dispersion compensation to achieve good signal levels. The next chapter will focus on the adaptations that are made to the C-FCM design for two-photon imaging.
66


CHAPTER V
TWO-PHOTON FIBER-COUPLED 3D MICROSCOPY
2P-LSM is a widely-used fluorescence imaging technique in neuroscience because it yields optical access deeper into tissue compared to standard single-photon fluorescence microscopy (Denk et al., 1990; Peron et al., 2015a). Like LSCM, 2P-LSM allows for the recovery of a small DOF, which can be combined with axial scanning to achieve 3D-im-aging. 2P-LSM surpasses LSCM with increased imaging depth in solid tissue and reduced phototoxicity, making it the preferred imaging technique for in vivo neuronal imaging. When used in combination with rapid axial scanning, it is possible to generate 3D representations of neuronal structure and activity (Gobel et al., 2007; Nadella et al., 2016). Additional details on the history and methods of 2P-LSM can be found in Chapter 2. Though 2P-LSM is powerful for in vivo imaging, the size and complexity of the ultrafast laser excitation and unique optical requirements for efficiency light collection make it challenging to miniaturize.
In this chapter, I describe the design, construction, and testing of the two-photon fiber-coupled microscope (2P-FCM), with axial scanning enabled by an integrated EWTL and lateral scanning achieved with the use of a CIFB. Use of an EWTL is ideal for this application as it is lightweight, compact, has low-power requirements, and is immune to motion and orientation (Berge and Peseux, 2000; Blum et al., 2011). The optics are packaged in a light-weight 3D-printed enclosure.
Pulse propagation through the CIFB is controlled by careful pre-compensation of the dispersion of glass and verified by spectrally resolved auto-correlation measurements. Verification of the 2P-FCM performance is shown by imaging resolution test targets and
67


fluorescent beads in thick agarose preparations. Finally, in vivo neuronal activity is shown by 3D-imaging of neurons in the motor cortex of a freely-behaving mouse with the head-attached 2P-FCM. A baseplate is permanently affixed to the mouse skull for attachment and alignment for repeated imaging over the same brain region. The 2P-FCM presented here is the first device to attain 3D-imaging of neural activity in a 240-pm diameter by 180-pm depth in the brain of a freely-moving mouse with no mechanically moving parts.
5.1 Miniaturizing two-photon laser-scanning microscopy
2P-FCMs use ultrashort pulsed lasers coupled through an optical fiber. The light is typically focused and scanned across the sample to form an image. The fluorescence emission is collected and delivered to a high-sensitivity photodetector, such as a photomultiplier tube (PMT) or avalanche photodiode (APD).
There are two main challenges when fiber-coupling 2P-LSM:
• Ultrashort pulse propagation
The excitation source must reach the tissue sample with high peak-pulse power to efficiently result in multiphoton excitation (see Chapter 2 for more details). Propagation through a solid-core fiber may result in temporal broadening due to chromatic dispersion of the glass, modal dispersion in the fiber-cores, and Kerr-in-duced non-linear effects that diminish the peak pulse power to a point where multiphoton excitation is not possible.
• Efficient emission collection
Emission light from 2P-LSM must be efficiently collected from as large a field as possible, since it is assumed that the excitation region is spatially restricted for
68


each focus-spot. In fiber-coupled systems, there must usually be a second optical path for emission collection.
Several types of laser-scanning FCMs and microendoscopes for two-photon imaging have been developed that have solutions for these challenges (Flusberg et al., 2008b; Helmchen et al., 2013; Sawinski et al., 2009; Szabo et al., 2014). In addition to in vivo neuroscience research, these devices have clinical endoscopy applications where they can deliver a variety of advanced imaging techniques in miniature packages (Liu et al., 2015; Qiu and Piyawattanametha, 2015).
Lateral scanning and axial focusing methods for miniature laser-scanning microscopes are explored in Chapters 2 and 3, respectively. For lateral scanning in the 2P-FCM described here I use a CIFB because of its proven effectiveness performing laserscanning microscopy while maintain mechanical simplicity to allow for experimentation with tunable focus optics (Flusberg et al., 2005b; Gobel et al., 2004). For active axial focusing, I implement an EWTL because of the compact size, large focusing range, and resistance to motion (Berge and Peseux, 2000; Bower et al., 2012; Hendriks et al., 2005).
5.2 Methods Overall system design
The experimental setup for imaging with the 2P-FCM is shown in Figure 5.1. The distal optics of the 2P-FCM are housed in a two-part 3D-printed enclosure, which can be repeatedly attached to a baseplate affixed to the animal’s head. A CIFB couples the distal imaging optics to a custom 2P-LSM system to relay the excitation laser and collect emitted fluorescence from the sample. Laser-scanning over the CIFB forms an image of the
69


sample while the fluorescence collected back through the fiber cores is detected by photodetectors housed in the 2P-LSM after spectral filtering. The individual components of the system are discussed in detail below.
Imaging
Fiber-bundle
Fiber-coupling
Asphere
Electrowetting
Lens
Plano-convex
Lens
Objective
Asphere
Coverglass
Figure 5.1: 2P-FCM imaging system. Pulses from a Ti:Sapphire laser source are spectrally prebroadened through polarization maintaining fiber (PM fiber) and pre-chirped using a grating-based pulse stretcher. Output pulses are scanned onto the surface of the CIFB through scanning mirrors, scan/tube lens relay, and lOx Objective. Fluorescence emission is collected by the CIFB and directed to a PMT through a dichroic filter. The collected pulses are amplified and transformed to logic levels to be detected by the DAQ and PC.
Laser-source
The excitation source is a Spectra-Physics MaiTai HP Ti: Sapphire pulsed laser, with ~80-fs duration pulses tuned to a center wavelength of 910-nm and operating at 80-MHz repetition rate. The beam power is controlled by a half-wave plate on a rotation mount (Newport Conex-AG-PRIOOP) followed by a Glan-Taylor polarizer (Thorlabs GT10-B).
70


Spectral and temporal pre-compensation
A polarization maintaining (PM) single-mode fiber (Thorlabs) is used for spectral pre-broadening to counter the spectral narrowing that occurs while propagating the high-power light through the fiber-bundle cores. After propagating through the PM fiber, the beam size is expanded by 5x to reduce the average intensity on the grating-pair. The grating pair is used to apply ~40,000-fs2 of negative group delay dispersion (GDD) on the laser pulse to compensate for the positive GDD that occurs when propagating through the 1.0-m length of optical fiber. The gratings are reflective ruled gold with a density of 300-grooves/mm (Edmund Optics 49-572) separated by 160-mm. Following the gratings, the beam size is reduced by 4x.
2P-LSM bench-top system
The beam is routed through a galvanometric mirror scanning system (Cambridge Technologies, 6215H) and relayed through a 50-mm FL scan lens and 180-mm FL tube lens in an Olympus 1X71 research microscope. The beam is scanned over the surface of the CIFB through a 10X 0.35 NA Olympus UPLANSApo objective lens. A XYZ-transla-tion stage (Thorlabs, CXYZ05) is used to accurately align the fiber to the focus of the objective lens.
The emitted fluorescence is collected by the distal imaging optics and propagates through the fiber cores of the CIFB. The fluorescence passes through a low pass filter (Chroma T670LPXR) and is split by a dichroic filter (Semrock FF562-Di02) for dual detection of GCaMP6s and tdTomato. The signal is detected on non-descanned photoncounting large area PMTs (Hamamatsu H7422PA-40) for the two channels. The output pulses from the PMT pass through a high-bandwidth amplifier (Becker & Hickl GmbH
71


ACA-4-35db) and are converted to logic-level pulses by a timing discriminator (6915, Phillips Scientific). The pulses are counted by a data-acquisition (DAQ) card (National Instruments PCIe-6259) at a rate of 20-MHz. The counts are sampled and binned by pixels and converted into image in custom software in Labview (National Instruments). The software also controls the EWTL driver.
2P-FCM miniature optical system design
The imaging system for the 2P-FCM was designed in Zemax optical design software. Models for the stock lenses and for the EWTL were obtained from the manufacturers (Edmund Optics, Thorlabs, and Varioptic). The CIFB (Fujikura Ltd. FIGH-15-600N) has an outer diameter of 700-pm, an active image diameter of 550-pm, and a length of 1.0 m. There are -15,000 cores with core-to-core spacing of 4.5-pm and core diameter of 3.2-pm, as previously reported by Chen et al. measured with scanning electron microscopy (Chen et al., 2008).
The miniature optics contained in the head-mounted 2P-FCM imaging system are shown in Figure 5.2. The fiber-coupling lens is an aspheric lens (FL: 6.2-mm, diameter: 4.7-mm, Edmund Optics 83-710) that collimates the light diverging from the fiber bundle. The electrowetting lens is placed in the collimated beam and the light is refocused onto the sample by an objective lens consisting of a plano-convex lens (FL: 7.5-mm, diameter: 3.0-mm, Edmund Optics 49-177), and an aspheric lens (FL: 2.0-mm, diameter: 3.0-mm, Thorlabs 355151-B). The nominal magnification of this imaging system is 0.4x and field-of-view (FOV) is -220-pm, corresponding to the de-magnified CIFB imaging diameter. Similarly, the core sampling resolution is the de-magnified core spacing, which is ~1.8-pm.
72


#1 Coverglass
Figure 5.2: Optics of the 2P-FCM miniature microscope head that focuses excitation light from CIFB cores onto the tissue. The CIFB -coupling asphere collects the light from the cores of the CIFB, which are then passed through the aperture of the EL. The plano-convex lens and the objective asphere focus the light onto the tissue through a #1 coverglass with 0.15-mm thickness.
A commercially available EWTL (Varioptic Arctic 316) is used to control axial focusing of the 2P-FCM imaging system. The EWTL contains two immiscible fluids with different refractive index and equal density, which interact via surface tension. An applied electric field adjusts the curvature of the interface and thus the focal length of the EWTL (Berge and Peseux, 2000). The predicted working distance and other imaging properties from Zemax at these three settings are summarized in Table 5.1. The optical power range of the EWTL is specified as -16 to +36 diopters. The optical system is optimized through the full focal range of the EWTL to minimize magnification change, maximize the axial scan range, and maximize the working distance of the 2P-FCM. Chapter 3 discusses the requirements for maintaining imaging performance throughout the full focal range.
73


Table 5.1: 2P-FCM optical parameters from Zemax through range of focal lengths.
EWTL control (Vrms) EWTL optical power (nr1) Working distance (pm) Magnification NA
60 35 450 0.41 0.43
42 0 570 0.40 0.44
25 -16 690 0.39 0.45
The 2P-FCM has low excitation NA compared to what one would typically use for two-photon excitation, which is usually > 0.8 NA. It has been shown that low NA objective lenses may be well-suited for efficient two-photon imaging because of the larger volume of the 3D focus spot (Singh et al., 2015). High NA objectives are desired more for capturing as many emitted fluorescence photons as possible, which can arrive from unpredictable field locations after being heavily scattered through deep tissue. In the 2P-FCM, the large effective area CIFB combined with achromatic misfocusing result in a
910 nm Excitation: 0.45 NA
532 nm Collection: 0.60 NA
Figure 5.3: NA comparison of forward excitation light at 910-nm (top) and backward emission light at 532-nm (bottom). Largest emission and excitation field positions are matched at 220-pm FOV.
74


higher collection NA of ~0.6, averaged over the EWTL focuses. This is illustrated in Figure 5.3 showing either 910-nm forward excitation or 532-nm backward emission. 3D-printed enclosure design
The enclosure for the 2P-FCM optics is designed in Solidworks 3D CAD software (Dassault Systemes). The packaging is split into three sections: top, bottom, and baseplate, shown in Figure 5.4: ,3D-printed enclosure (a) A two-part 2P-FCM snaps together to secure the EWTL and electrode, (b) Photo of 3D-printed parts, top: before any processing with supports still attached and bottom: assembled 2P-FCM.. The top-section contains the CIFB ferrule, held in place by two set-screws, and the fiber-collimating as-phere. The bottom-section contains the objective lenses. The unmounted lenses are held in by friction in precisely sized openings. The top-section has two curved tabs that interface with slots in the bottom-section, which help to ensure reproducible alignment. The
(a)
(b)
Figure 5.4: ,3D-printed enclosure (a) A two-part 2P-FCM snaps together to secure the EWTL and electrode, (b) Photo of 3D-printed parts, top: before any processing with supports still attached and bottom: assembled 2P-FCM.
75


EWTL and the electrode are sandwiched between the bottom-section and the top-section with an O-ring that ensures good electrical contact. The flat-flex electrode cable exits the enclosure through a small slot between the sections. The top-section tabs have a single thread at the end, which interfaces with the baseplate as shown in Figure 5.5 (a). In this way, the baseplate is pulled up against the bottom-section by the top-section. This greatly improves rigidity when attached to a moving animal. The baseplate is designed with ridges and holes to improve adhesion of the cement for attachment to the animal skull. The entire enclosure is 3D-printed using a high-resolution projection-based resin printer (Kudo3D Titan 1), with a resolution of 50-pm. The material used is a photo-curable resin (3DM-XGreen) dyed with 0.5% molybdenum disulfide to decrease light scattering and thus increase feature resolution. A photo of the top- and bottom-sections immediately after printing are shown in Figure 5.5 (b). 3D printing allows optimization of the prototype and easily enables design changes, such as the inclusion of GRIN lenses for deep-brain imaging (Barretto and Schnitzer, 2012a).
Test sample preparation
Resolution and axial scan range measurements were performed by imaging fluorescent beads embedded in agarose (Sigma-Aldrich A9414) and a USAF 1951 resolution target (Edmund Optics 38-257). 2-pm yellow-green fluorescent beads (Invitrogen F8853) were used to measure axial scanning extent as well as lateral and axial resolution.
Low melting-point agarose was prepared at a concentration of 0.5% in water. The 2-pm diameter fluorescent yellow-green beads were diluted in the agarose to a concentration of ~2.0 x 107 beads/mL. Approximately 2.0-mL of solution was placed on a #1 coverglass and allowed to set at room temperature. The beads were imaged in sequential
76


axial planes by a 20x 0.75 NA Olympus UPLANSApo objective with a motorized stage and separately by the 2P-FCM by changing the voltage applied to the EWTL.
Mouse imaging setup
All experiments were approved and conducted in accordance with the Institutional Animal Care and Use Committee of the University of Colorado Anschutz Medical Campus. Male 3-month old C57BL/6 mice were anesthetized by intraperitoneal ketamine-xylazine injection. The skin above the target site was numbed by lidocaine injection and retracted to expose the skull. The mouse was injected with an adeno-associated virus driving expression of GCaMP6s under the synapsin promoter (AAV5.Syn.GCaMP6s), similar to procedures in (Chen et al., 2013b). The coordinates of the injection targeted the hindlimb somatosensory cortex, 0.2-mm posterior to bregma and 1.5-mm lateral to the midline, at a depth of 300-pm (Paxinos and Franklin, 2012). The injection volume was 0.66 microliters delivered with a glass micro-pipette through a 0.5-mm hole drilled at the target site.
One month after injection the mice were implanted with an optical cranial window near the injection site, using standard techniques as previously described. Briefly, mice were anesthetized by isoflurane inhalation and the skin under the scalp was numbed by subcutaneous lidocaine injection. The skin above the skull was removed to expose the injection site and skull surface. A 2-mm square window of skull was removed immediately anterior to the injection hole to expose the dura mater. The opening was covered with a 2-mm square #1 coverglass and secured in place with cyanoacrylate glue. Dental acrylic cement (C&B-Metabond) was used to cover the skull surface. The presence of fluorescence
77


signal was confirmed with standard 2P-LSM using a 20x 1.0 NA Zeiss Plan-Apochromat water-immersion objective.
The baseplate attachment procedure is similar to what has been described for other miniature head-attached microscopes. While the mouse was still anesthetized, the 2P-FCM was held and positioned above the window with a micromanipulator (Sutter MP-285) until fluorescent signal could be observed with widefield epifluorescence through the bundle. The target region was chosen with two-photon imaging and the 2P-FCM was positioned to the region with the baseplate attached. The baseplate was then secured to the existing acrylic with black acrylic cement (Lang Dental Jet Acrylic). After allowing to set for ~30 minutes, the 2P-FCM was removed, leaving the baseplate in place, and the mouse was allowed to recover.
The imaging setup is illustrated in Figure 5.5. The mouse was lightly anesthetized with isoflurane inhalation. The baseplate was carefully gripped by thumb and forefinger and the 2P-FCM was inserted and secured with a quarter-turn. The EWTL electrode was connected to light-gauge wires that were draped, along with the CIFB, over a horizontal metal post above the behavior cage. The mouse was allowed to recover in the behavior cage for imaging. The cage was illuminated by red light to minimize coupling into the fluorescence detection path and a camera (Logitech C615) was positioned above the cage to monitor behavior during imaging.
Image processing
The images from the 2P-FCM show a honeycomb pixelation pattern due to the packing of the cores of the CIFB. Several methods have been described to de-pixelate images from CIFBs (Lee and Han, 2013a, 2013b; Shinde and Matham, 2014). The simplest
78


(a) (b)
Figure 5.5: Mouse attachment, (a) The CIFB is attached to a coupling objective on the proximal end and the 2P-FCM on the distal end. (a) 2P-FCM is attached to the permanent baseplate on the mouse with a quarter-turn.
methods involve low-pass filtering with either a blurring function(Gobel et al., 2004) or masking the image in the frequency domain(Winter et al., 2006). However, two-photon imaging through a CIFB has the additional complication of amplifying the non-uniformity of the fiber cores. Each core is assumed to have a unique sensitivity, due to the variability in diameter, shape, NA, and amount of cladding between adjacent cores. This manifests as discrete variations in image intensity across the FOV.
This was addressed by programmatically dividing out the sensitivity of each core and interpolating the core values to remove the pixelation pattern. This process is described in detail in the Appendix. Briefly, I used a flat map of the full field CIFB fiber-cores by imaging a fluorescent test slide (Chroma Technologies) with the 2P-FCM, with an example shown in Figure 5.6 (a). The flat-map stores the centroid coordinates of the cores and their corresponding sensitivity. The processing was performed with custom software (Matlab, Mathworks). Each image to be analyzed was registered to the flat-map, which
79


allowed identification of the cores. An example of a raw pixelated image is shown in Figure 5.6 (b). The relative sensitivity of each core was compensated by dividing by the flat map values. The honeycomb pattern was eliminated by using the nearest neighbor interpolation method (Rupp et al., 2009). A Savitsky-Golay filter was used to reduce the added single-pixel noise introduced by the core multiplication factor during flat-normalization. During the interpolation, the fiber-cores were registered to a square pixel grid for straightforward analysis. An example output frame is shown in Figure 5.6 (c).
For processing of temporal scans, each frame was processed with the same flat-map alignment so that the cores are static in the field. Once the honeycomb pattern and CIFB-induced intensity variation was removed, a clustering algorithm was used to identify regions of interest (ROIs) of high-correlation (Ozden et al., 2008). Significant changes in cytosolic Ca2+ were identified as changes in fluorescence larger than 3 standard deviations above baseline within each ROT
Figure 5.6: Image processing of two-photon imaging through CIFB. (a) Flat-field map showing enhanced non-uniformity due to heterogeneity in CIFB. (b) Unprocessed image of cells in mouse cortex with fiber-pixelation. (c) Post-processed image after fiber-cores were corrected with flat-field mask and re-gridded into a typical square pattern. Grid lines added to emphasize pixels.
80


5.3 Results
Resolution, magnification, and scan range
I characterized the lateral and axial resolution and the axial focusing range of the 2P-FCM by imaging 2-pm diameter yellow-green fluorescent micro-beads embedded in clear agarose. The lateral resolution is fundamentally limited by the average spacing of the fiber cores in the CIFB. Using a calibrated objective lens, I measured the inter-core spacing to be ~4.5-pm, which agrees with previously published results (Chen et al.,
2008). With the 2P-FCM magnification factor of 0.4x, the theoretical lateral sampling at the object is ~1.8-pm.
The beads were imaged with the 2P-FCM at sequential focal planes by tuning the EWTL focus in discrete steps to obtain a Z-stack. The images were processed to remove the fiber-pixelation pattern as described in the methods section. Figure 5.7 compares processed images of the beads imaged with a 20x 0.75 NA objective and the 2P-FCM. The average lateral and axial line profiles of 5 beads measured from different focus positions were fit to a Gaussian function. The axial bead size measured by the 20x objective was 4.5-pm FWHM and with the 2P-FCM it was 9.9-pm FWHM. The lateral bead size measured by the 20x objective to be 1.7-pm FWHM (dotted grey line) and with the 2P-FCM it was 2.6-pm. The lateral bead size is larger than diffraction limit as it is limited by the fiber bundle spacing. With a bead size of ~2-pm, non-uniform sampling of the bead with
81


multiple fiber cores causes a larger effective lateral profile. The axial profile measurements of both the 20x objective at 0.75 NA and the 2P-FCM at 0.45 NA are similar to what is expected from the diffraction-limited calculations.
Axial Bead Profiles Lateral Bead Profiles
Figure 5.7: Axial and lateral resolution tested by imaging fluorescent beads with a 20x Olympus objective (dashed lines) or with the 2P-FCM (solid lines).
Figure 5.8 (a) shows side-projections of beads as imaged by a 20x 0.75 NA objective and 2P-FCM overlaid in green and red, respectively. The same region of the agarose-bead sample was imaged in both cases, such that most of the same beads appear in both Z-stacks. This made it possible to compare directly their apparent size and axial location to determine the 2P-FCM scan range. The predicated axial focus plane through the EWTL focusing range from Zemax and actual bead positions are shown in Figure 5.8 (b) (grey line and black dots, respectively). The full focusing range did not span the range predicted (240-pm predicted vs. 180-pm measured), likely due to under-performance by the EWTL at the high-end of the optical power range. I was able to rule out optical alignment issues by tolerancing analysis in Zemax, which indicated that any physical alignment error or static lens discrepancies could not account for the change in focus range. The focal range of the EWTL was estimated by measuring the distance from the lens to
82


(a) (b)
Figure 5.8: Testing of axial scan range, (a). Side-projection of ~2-pm diameter fluorescent beads suspended in clear agarose and imaged with a 20x 0.8 NA Olympus objective (green) using 910 nm excitation light and a motorized stage or with the 2P-FCM while varying the EWTL power (red), (b) Predicted scan range as the EWTL optical power is changed modeled in Zemax (grey line) and Z-positions of measured beads (black circles).
the focus spot when imaging a distant light source. The optical power at the highest voltage setting (60V) was measured to be ~ 30-m"1, lower than the specified 36-m"1.
The optical magnification from the CIFB to the target was evaluated by imaging the group 6, element 2 square on the USAF 1951 resolution target with the 2P-FCM as shown in Figure 5.9. The imaging diameter of the CIFB was measured to be 550-pm. The magnification at three different optical power settings for the EWTL was measured by comparing the scaled size of the resolution target through the CIFB with the actual size. The results are summarized in Table 5.2. The magnification is measured to be ~0.4x, varying by less than 5% through the focusing range. This agrees closely with the predictions from the Zemax model.
83


Full Text

PAGE 1

i FIBER COUPLED MICROS COPY FOR 3D NEURONAL IMAGING by BARIS N OZBAY B.S., University of Colorado Denver, 2010 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Doctor of Philosophy Bioengineering Program 2017

PAGE 2

ii 2017 BARIS OZBAY ALL RIGHTS RESERVED

PAGE 3

iii This thesis for the Doctor of Philosophy degree by Baris N Ozbay has been approved for the Bioengineering Program by Richard KP Benninger, Chair Emily A Gibson, Advisor Diego Restrepo , Co advisor Mark L Dell’Acqua Gidon Felsen Date: December 16, 2017

PAGE 4

iv Ozbay, Baris N (Ph.D., Bioengineering Program ) Fiber Coupled Microscopy for 3D Neuronal Imaging Thesis directed by Assistant Professor Emily A Gibson and Professor Diego Restrepo ABSTRACT In this dissertation I describe the design and implementation of miniature fiber coupled microscopes (FCM s ) with active focusing for three dimensional (3D) neuronal imaging. The goal is to provide neuroscience researchers with versatile microscopy tools to perform neuronal optical imaging of awake and mobile mice . This research is motivated by the r ecent advancements of powerful genetically encoded optical proteins , including fluorescent activity sensors as well as optogenetic actuators, which permit the functional interrogation of in vivo neuronal circuits . Newly developed miniature microscope tools allow optical imaging i n freely behaving mice, but current designs do not combine optical sectioning capabilities with active focusing for full 3D i maging. The first three chapters in this dissertation serve as a background for current state of intravital fluorescence microscopy in neuroscience research. I argue for the importance of miniaturizing the microscope technologies that enable high contrast imaging with optical sectioning, combined with axial focusing, to enable 3D imaging at the rodent scale. I present two FCM designs that achieve full 3D imaging using a coherent imaging fiber bundle (CIFB) for lateral imaging and an electrowetting tunable lens (EWTL) to enable electrically tunable axial focusing. The first design is a confocal FCM (C FCM) that takes advantage of the op tical sectioning capability of the CIFB to acquire highcontrast images. The second design is a twophoton FCM (2P FCM), in which pre compensated ultrashort pulses are propagated through the CIFB for twophoton excitation

PAGE 5

v microscopy. In each section, I cha racterize the 3D optical performance of the FCM . Finally, as a proof of principle using the 2P FCM, I show in vivo 3D imaging of neurons and Ca2+activity in the motor cortex of a freely behaving mouse. The form and content of this abstract are approved. I recommend its publication. Approved: Emily A Gibson and Diego Restrepo

PAGE 6

vi Dedicated t o the research animals and to the researchers who treat them with compassion and respect

PAGE 7

vii ACKNOWLEDGEMENTS The completion of this dissertation is only possible because of the support and patience of my advisors, Emily Gibson and Diego Restrepo. I am grateful for their enduring belief in me through periods of adversity and for giving me the autonomy to build my own path. I thank the c ontributors to this work , including Gregory Futia, Justin Losacco, Ming Ma, Bob Cormack, and Ethan Hughes wh o gave their skills and time to important parts of these projects. Thanks also to the labs of Juliet Gopinath and Victor Bright at the University of Colorado Boulder, who have shared their expertise and resources. Thanks to the Advanced Light Microscopy Core for trusting me to tinker with and not break their instruments. My sincere gratitude goes to everyone has come through our labs for the discussions, assistance, and friendships . I acknowledge those who worked to advocate for the larger concepts that germinated my own projects and funding from the National Science Foundation that made this work a reality. I am appreciative for those who critically reviewed my work and my thesis committee for their patient examination of my progress. I thank my parents, for everything. Finally, my heartfelt gratitude goes to Katie, f or be ing a resilient and guiding light through these many years of study.

PAGE 8

viii TABLE OF CONTENTS Abstract .............................................................................................................................. iv Acknowledgements ........................................................................................................... vii Table of Contents ............................................................................................................. viii List of Figures ..................................................................................................................... x List of Abbreviations ....................................................................................................... xvi INTRODUCTION ........................................................................................................... 1 1.1 Thesis organization .............................................................................................. 3 MINIATURIZED MICROSCOPY IN NEUROSCIENCE ........................................... 5 2.1 Fluorescence microscopy in the neurosciences ................................................... 5 2.2 Modern microscopy techniques for studying the brain ........................................ 9 2.3 In vivo brain imaging methods .......................................................................... 17 2.4 Miniaturized microscope designs ...................................................................... 21 2.5 Where we are in miniaturized neuronal imaging? ............................................. 28 AXIAL FOCUSING IN MINIATURIZED MICROSCOPES ................................... 30 3.1 Benefits of minimizing depth-of-field in fluorescence imaging ........................ 32 3.2 Active axial focusing methods ........................................................................... 41 3.3 Tunable focus by liquid lens technology ........................................................... 48 3.4 Optical design considerations for axial scanning with a tunable lens ............... 50 CONFOCAL FIBER COUPLED MICROSCOPE FOR 3D IMAGING ................... 54

PAGE 9

ix 4.1 Miniaturizing confocal laser scanning microscopy ........................................... 55 4.2 Methods.............................................................................................................. 56 4.3 Results ................................................................................................................ 60 4.4 Summary ............................................................................................................ 65 TWO PHOTON FIBE RCOUPLED 3D MICROSCOPY .......................................... 67 5.1 Miniaturizing two -photon laser-scanning microscopy ...................................... 68 5.2 Methods.............................................................................................................. 69 5.3 Results ................................................................................................................ 81 5.4 Summary ............................................................................................................ 95 CONCLUSIONS AND OUTLOOK .......................................................................... 96 6.1 Revisiting the confocal FCM ............................................................................. 97 6.2 Moving forward with the two-photon FCM .................................................... 100 6.3 Fiber -bundles: Simple, versatile, needs improvement ..................................... 102 6.4 Finding a need for the 3D imaging FCM ......................................................... 105 APPENDIX FIBER BUNDLE DE PIXELATION METHOD ................................... 107 References ....................................................................................................................... 116

PAGE 10

x LIST OF FIGURES FIGURE 2.1: (a) Purkinje neurons as represented in the 19th century by Ramn y Cajal (Sotelo, 2003). (b) Modern fluorescence intensity projection image of a mouse Purkinje neuron filled with Lucifer Yellow fluorescent dye (Martone, 2002). .................................................................................................... 6 2.2: Illustration of the epifluorescence microscope. Monochromatic excitation light, selected by an excitation filter, is reflected by a dichroic beam splitter. An objective lens focuses the excitation light onto the sample and collects the emitted fluorescence from the fluorophores. The emission passes through the dichroic beam splitter and emission filter before being captured on a detector, such as a camera, to form an image of the fluorescence. ................................................................................................ 7 2.3: A simplified Jablonski diagram showing electronic and vibrational states of a common fluorophore illustrates how a photon of the correct energy, electronic state from the ground state (S0) to the excited state (S1). Spontaneous vibrational relaxation is followed by fluorescence emission as th e fluorophore recovers to S0, with the emitted photon having less energy and longer wavelength. .............................................................................. 8 2.4: Confocal microscopy basic setup with stage scanning. An optical pinhole in the conjugate focal plane is adjusted to reject out of focus fluorescence emission from a thick fluorescent sample. The laser is scanned across the sample, in this illustration with a translation stage. .............. 11 2.5: 2P LSM basic setup with stagescanning. A pulsed ultrashort excitation laser is focused onto the sample, resulting in a non linear excitation of fluorescen ce limited to the focal volume. The fluorescence is filtered and recorded for each pixel by a single element detector. ......................................... 13 2.6: One and twophoton absorption process are shown to both cause a similar transition to the excited electronic state, resulting in similar fluores cence output. The twophoton process requires two photons to arrive near simultaneously with approximately twice the wavelength, or half of the energy, of the energy gap. ................................................................................... 14 2.7: Cranial window implanted in mouse to achieve optical access to cortical structures. An objective lens is used to image directly into the window. ........... 17 2.8: GRIN lens illustrations. Top: Single GRIN lens with pitch ~0.5 with a magnification of 1. Bottom: Twoelement GRIN lens with low NA relay at pitch ~0.75 and high NA objective with pitch ~0.25. ..................................... 19

PAGE 11

xi 2.9: GRIN lens assembly impl anted into deepbrain of mouse and imaged with a low NA objective lens to couple into the low NA GRIN lens element. ........... 20 2.10: Example widefield miniature microscope design from miniscope.org. (a) Internal components of miniature microscope (weight 2g), with integrated LED, CMOS camera, filters, and lenses. (b) Photo of mouse with headattached miniature microscope. .......................................................... 22 2.11: Fiber scanning techniques. (a) Piezoelectric resonancescanning of the fiber tip, resulting in the scanning of resonant modes, such as the spiral shown inset. (b) MEMS actuated mirror scanners, resulting in arbitrary scan trajectories, such as the raster shown inset. ................................................. 24 2.12: Illustration of proximal to distal image coherence of CIFB, as well as the pixelation of an image formed through the bundle because of the discrete fiber cores. ........................................................................................................... 26 2.13: Fujikura CIFB fluorescence image of a uniform target showing core distribution. .......................................................................................................... 27 3.1: (a) Sketch representation of the human motor cortex neuronal column by Ramn y Cajal (1899). (b) Modern 3D reconstruction of a threephot on image of a cortical column in an intact mouse brain (adapted from (Horton et al., 2013)). .......................................................................................... 31 3.2: (a) Gaussian profile of the w(z) function, indicating the x axis parameter 0 and the z axis parameter zR. (b) Gaussian transverse intensity profile at beam waist (z=0), i ndicating the w0 value and the related FWHMr . (c) Axial intensity profile along the optical axis (x=0), indicating the zR value and the related FWHMz . ............................................................................ 36 3.3: 2m fluorescent microsphere imaged under blue light illumination. (a) A widefield image of the fluorescent microsphere through the focus, showing the transverse (top) and axial (bottom) focus and the extended DOF. (b) A LSCM image of an identical bead showing the improved lateral resolution an d confined axial spread. ....................................................... 37 3.4: Power encircled at radius 5FWHMr through focus of 2 m beads capture with widefield (solid line, no optical sectioning) and LSCM (dashed line, with optical sectioning). .............................................................................. 38 3.5: Fluorescent pollen grains under green light excitation taken with a 1.42 NA 60x objective. (a) The pollen grains imaged under widefield illumination, showing the loss of contrast in the lateral dimension and a large DOF fetching out of focus grains. (b) A LSCM optical stack of several thin focal planes from the same pollen grains sho wing much higher contrast. .................................................................................................... 39

PAGE 12

xii 3.6: Three fundamental axial focusing options. Object field scan changes the location of the imaging target relative to the optical system. Image field scan changes the imaging optics to change the focal plane. Tunable focus scanning changes the effective focal length of the optics to change the focal plane. .................................................................................................... 42 3.7: Examples of published lens movement based axials scanning. (a) Axial scanning by small focusingmotor actuation with 1.1mm axial motion (Flusberg et al., 2008a). (b) Lens actuation by shape memory alloy contraction, with 150m of axial motion (Wu et al., 2010). .............................. 43 3.8: Examples of published MEMS mirror devices with axial focusing. (a) Thin film piezoelectric actuator for 190 m of vertical translation with parabolic mirrors for excitation and collection (Qiu et a l., 2014). (b) Monolithic 3 axis MEMS scanning mirror assembly, with vertical translation photos shown in (c), achieving 546m of vertical translation (Li et al., 2016). ................................................................................................... 45 3.9: Two varieties of commercially available liquid ETLs. (a) Optotune EL 10 30C MV ETL based on a shapechanging polymer surface, with outer diameter of 30mm and 10mm aperture. (b) Varioptic Arctic 316, based on electrowetting lens technology, with outer diameter of 7.8 mm and 2.5mm aperture. ................................................................................................. 49 3.10: Axial focusing by wavefront curvature shaping at the objective back focal plane. (a) An input converging or diverging wavefront is transformed into an axial translation away from the designed focal length of the lens. (b) Spherical aberrations as a result of a lens obeying the sine condition. .......... 52 4.1: Laser scanning confocal microscop e coupled to distal imaging optics through fiber bundle (IL: Imaging lens, EWTL: Electrowetting tunable lens, OL: Objective lenses). ................................................................................ 56 4.2: (a) CAD model of adapter with optics with lenses aligned in polyimide (PI) tubing. (c) Photo of assembled adapter and EWTL. .................................... 58 4.3: Lateral imaging characteristics of C FCM at different objective depths. (a) GFP labeled neuroglia imaged with 20X, 0.8 NA objective. (b) Same region imaged with C FCM at shortest and (c) longest objective lens length. .................................................................................................................. 60 4.4: Imaging power loss over the full FOV with 1mm distance objective lens (black line) and 4mm distance objective lens (grey line). ................................. 61 4.5: Zoom in showing the cell bodies indicated by white arrow in Figure 4.3 (a). Left: Raw C FCM image showing pixilation due to fiber cores. Middle: C FCM image after filtering. Right: Comparis on with standard confocal microscope. ........................................................................................... 61

PAGE 13

xiii 4.6: Axial scan range of electrowetting C FCM (a) Orthogonal projection (inverted grayscale) of 1 m diameter red fluorescent beads in agarose imaged with scanning Z stage (left) and scanning the EWTL (right). (b) Black dots: 40 bea ds mapped from EWTL optical power to relative Z position. Gray line: Simulated C FCM focal length with varying EWTL power. .................................................................................................................. 62 4.7: Lateral and axial resolution of C FCM. (a) Lateral and axial images of fluorescent beads. Left: 1 m bead with open pinhole (PH), Middle: 1m bead with closed pinhole, Right: 2m bead with closed pinhole. (b) Top: Averaged line profiles of 1and 2 m beads (black) compared with diffraction limited resolution (grey) Bottom: Averaged axial profile of several 1 and 2 m beads (black) compared with theoretical axial resolution (grey). ................................................................................................. 63 4.8: 3D imaging of mouse nerve tissue. (a) Four optical sections that were taken at specific EWTL optical power settings. Scale bar is 100 m. (b) Maximum intensity projection of an image stack of intact olfactory neuron axons labeled with YFP. .......................................................................... 65 5.1: 2P FCM imaging system. Pulses from a Ti:Sapphire laser source are spectrally prebroadened through polarization maintaining fiber (PM fiber) and pre chirped using a grating based pulse stretcher. Output pulses are scanned onto the surface of the CIFB through scanning mirrors, scan/tube lens relay, and 10x Objective. Fluorescence emission is collected by the CIFB and directed to a PMT through a dichroic filter. The collected pulses are amplified and transformed to logic levels to be detected by the DAQ and PC. ............................................................................. 70 5.2: Optics of the 2P FCM miniature microscope head that focuses excitation light from CIFB cores onto the tissue. The CIFB coupling asphere collects the light from the cores of the CIFB, which are then passed through the aperture of the EL. The planoconvex lens and the objective asphere focus the light onto the tissue through a #1 coverglass with 0.15mm thickness. ...................................................................................................... 73 5.3: NA comparison of forward excitation light at 910 nm (top) and backward emission light at 532 nm (bottom). Largest emission and excitation field posit ions are matched at 220 m FOV. ............................................................... 74 5.4: 3D printed enclosure (a) A two part 2P FCM snaps together to secure the EWTL and electrode. (b) Photo of 3D printed parts, top: before any processing with supports still attached and bottom: assembled 2P FCM. .......... 75 5.5: Mouse attachment. (a) The CIFB is attached to a coupling objective on the proximal end and the 2P FCM on the distal end. (a) 2P FCM is attached to the permanent baseplate on the mouse with a quarter turn. ............................ 79

PAGE 14

xiv 5.6: Image processing of twophoton imaging through CIFB. (a) Flat field map showing enhanced nonuniformity due to heterogeneity in CIFB. (b) Unprocessed image of cells in mouse cortex with fiber pixelation. (c) Post processe d image after fiber cores were corrected with flat field mask and re gridded into a typical square pattern. Grid lines added to emphasize pixels. ................................................................................................. 80 5.7: Axial and lateral resolution tested by imaging fluorescent beads with a 20x Olympus objective (dashed lines) or with th e 2P FCM (solid lines). ................. 82 5.8: Testing of axial scan range. (a). Side projection of ~2m diameter fluorescent beads suspended in clear agarose and imaged with a 20x 0.8 NA Olympus objective (green) using 910 nm excitation light and a motorized stage or with the 2P FCM while varying the EWTL power (red). (b) Predicted scan range as the EWTL optical power is changed modeled in Zemax (grey li ne) and Z positions of measured beads (black circles). ................................................................................................................ 83 5.9: Measurements of magnification of 2P FCM. Elements of a USAF 1951 resolution target were imaged through the CIFB at three different focus settings. The known size of the elements is known precisely is indicated. ......... 84 5.10: Measurement of pulses for propagation through 1.0m long CIFB using FROG at 910 nm. Each figure: Top: Measured spectrogram. Middle: Spectrum retrieval from FR OG spectrogram overlaid with phase plot (orange). Separately measured spectrum is shown as well (dashed line). Bottom: Temporal pulse trace retrieved from FROG spectrogram overlaid with phase plot (orange). (a) Pulse measured directly from ultrafast laser. (b) Pulse measured after propagating through CIFB. .................. 86 5.11: Fixed tissue with GFP labeled oligodendrocytes imaged with 2P FCM. (a) 3D volume acquired by the 2P FCM (220 x 220 m lateral x 180 m axial) with over 200 cells in the image. (a). Processed image of a single sli ce in the stack after filtering to remove pixelation pattern. ............................. 88 5.12: Tilted field imaging enabled by rapid focusing of the EWTL. (a) Maximum intensity projection of fixed mouse brain tissue expressing GCaMP6s in neurons, acquired by the 2P FCM. Arrows indicate cell bodie s retained in fields. (b) Side projection of the volume. The same cell bodies are indicated by the arrows. The planes for the horizontal and angled scan limits are indicated (c) Images showing the largest tilted field scan acquired in this experiment, indi cating the same cell bodies that are shown to intersect with the red or white planes in (b). ........................... 89 5.13: Multi color imaging. (a) Twocolor maximum intensity projection acquired with a 20x 0.75 NA Objective A. Maximum intensity projection of a region of brain tissue acquired with a 20x 0.75 NA

PAGE 15

xv objective. Yellow cells are oligodendrocytes (Green and Red), while red only cells are astrocytes and oligodentrocytes cells. Two astrocytes are marked with arrows, and ar e easily identified by the characteristic bushy morphology and two are marked with arrows. (b) Same tissue imaged with the 2P FCM, using the same detectors, filters, and excitation wavelength. Green and red cells are visible in the field, with likely astro cytes marked by arrows. .................................................................... 90 5.14: Mouse 2P FCM attachment photos. (a) Baseplate implanted on mouse with cranial window. (b) Mouse behaving with 2P FCM attached. ................... 91 5.16: Histological coronal section of mouse injected with GCaMP6s virus, showing good expression in layers 2 5 of motor cortex. .................................... 92 5.15: Implan t stability over 17 days imaged with widefield epifluorescence. ............... 92 5.17: 3D imaging of cells labeled with GCaMP6s in behaving mouse. (a) Volume view of cells (b) Side projection of cells. .............................................. 94 5.18: Awake behaving mouse Ca2+ imaging with GCaMP6s. (a) Three different focal planes showing distinct cellular populations. The colors and regions represent the peak fluorescent spatial intensity from the corresponding transients. Scale bar 25 m. (b) Traces showing change in GCaMP6s fluorescence for the 5 regions of distinct activity indicated in panel (a). .............................................................................................................. 95 6.1: Coupling of green emission light into CIFB is indifferent to spot size unless coupling is in single core. ........................................................................ 97 6.2: Conceptual experiments for confocal imaging with FCM and GRIN lens. (a) 2P FCM coupled to a 0.8 NA GRIN lens assembly to image GFP labeled oligodendrocytes in fixed mouse brain tissue. (b) Maximum intensity projection of lateral and side views after processing to remove fiber pixelation (scale bar is 30 m). .................................................................. 99 6.3: Core to core coupling imaged at distal end of fiber while single core is illuminated by 910 nm light at proximal end. (a) Raw camera image of fibers showing higher order modes coupling into nearby fibers. (b) Same image with coupled cores circled. ..................................................................... 104

PAGE 16

xvi LIST OF ABBREVIATIONS 2D ...................Two dimensional 2P LSM ..........Two photon laser scanning microscopy 3D ...................Three dimensional C FCM ...........Confocal fiber coupled microscope CIFB ...............Coherent imaging fiber bundle CMOS ............Complementary metal oxide semiconductor CW .................Continuous wave DAQ ...............Data acquisition DOF ................Depth of field ETL ................Electrically tunable lens EWTL ............Electrowetting tunable lens FFT .................Fast Fourier transform FL ...................Focal length FROG .............Frequencyresolved optical gating FWHM ...........Full width at half maximum GDD ...............Groupdelay dispersion GECI ..............Genetically encoded calcium indicator GFP ................Green fluorescent protein LSM ...............Laser scanning microscopy LSCM .............Laser scanning confocal microscopy MTF ...............Modulation transfer func tion NA ..................Numerical aperture NIR .................Near infrared

PAGE 17

xvii PMT ...............Photomultiplier tube PSF .................Point spread function ROI .................Region of interest SLM ...............Spatial light modulator SNR ................Signal to noise ratio SPIM ..............Selective plane illumination microscopy YFP ................Yellow fluorescent protein

PAGE 18

1 INTRODUCTION The topic of this thesis is the development, assembly, and testing of miniature fibercoupled microscopes with tunable focus for mouse brain imaging. These microscopes are designed specifically to aid neuroscience researchers in investigating the functioning of the mammalian brain in a natural behavior paradigm with true three -dimensional (3D) imaging capabilities. In the last two decades, fluorescence microscopy has become an indispensable tool for neuroscience research. Recently, there have been important innovations in the engineering of fluorescent proteins to enable genetically -targeted optical sensing and actuation of cellular activity (Akerboom et al., 2012; Chen et al., 2013b; Lin and Schnitzer, 2016; Specht et al., 2017; Zhang et al., 2007). These optical proteins can be made to express in specific neuronal cell types of live animals, suc h as mice. Modern fluorescence microscopes and surgical techniques have allowed researchers to interrogate in vivo functioning neuronal networks in a much larger temporal and spatial regime than ever before. In contrast, electrophysiological recordings from implanted brain electrodes can record from only a small number of cells and the recorded signals are not intrinsically cell -type specific. With the continued development of faster optical sensors and spect rally distinct optical actuators, the hope is that an “all -optical” interrogation of large neuronal networks in awake and freely -behaving animals can be achieved (Emiliani et al., 2015 ). Despite the benefits of in vivo optical imaging, it must still overcome some big hurdles to match the maturity of electrophysiological techniques. Some of the weaknesses of optical imaging are : the lower temporal resolution, the limitation on imaging depth due to

PAGE 19

2 brain tissue opacity , and the difficulty to miniaturize optics for attachment to small subjects, such as mice, for minimally invasive recording of behavior. Meanwhile, electrode based recordings have been miniaturized for freelybehaving mous e imaging for several decades, leading to the innovation of the tetrode for gathering rich data from behaving mice (Harris et al., 2000) . These tools have allowed for the validation of many neuronal circuit models that were developed using in vitro techniques. To obtain similarly functionally relevant data from optical imaging, it is desirable to combine optical imaging with behavioral monitoring. Tethered but freely moving mouse electrophysiology (Anikeeva et al., 2011; Li et al., 2014) and microscopy (Ziv et al., 2013) have demonstrated the importance of minimizing behavioral restraints for studying the awake brain in action. Further development of behaviorally related optical imaging would open up new avenues for investigating animal models of disease and the effects of drugs and therapies on the mammalian brain (Heys et al., 2014; Hillman, 2007) . However, optics are difficult to miniaturize to the scales that would minimally disrupt or restrict the be havioral environment . Along with the improvements in the biological tools, microscope technologies have generally improved to meet the demand. Headfixed animal imaging has provided an avenue for utilizing existing large microscope optics for imaging awake mice during a limited set of behaviors (Dombeck and Tank, 2014) . Cuttingedge techniques, such as mesoscopic imaging, ha ve begun to show that very large scale data, (upwards of thousan ds of spatially related neurons ) , can be retrieved through optical imaging of the brain (Sofroniew et al., 2016) . But t here are many behavioral paradigms that are not well replicated in a severely restrained setup , such as those used in headfixed imaging. Examples

PAGE 20

3 are social and mating behaviors, spatial maze exploration, and the complex responses to anxiety or fear stimuli. The reliance on existing large microscope designs for head fixed imaging has meant that t he transition to in vivo optical imaging for small subjects, such as mice, has been under addressed by available microscope technologies. As optical inte rrogation tools are becoming an undeniably powerful methodology, there is a push in neuroscience to close the gap in microscope scale. Many established neuronal circuit models of function and disease are being tested using gen etically targeted fluorescent tools (Grewe et al., 2017; Tantirigama et al., 2017) . Simultaneously, head attached miniature microscopes for mice are becoming more available and increasingly useful for neuroscience researchers, but are narrow in the available modalities (Chen et al., 2013a) . Versatile miniature microscopes that inco rporate more advanced microscope technologies are increasingly in demand. In this thesis I will summarize my approach to addressing this technological gap. 1.1 Thesis organization The work that I present here is focused on the development of two types of min i ature fiber coupled microscope s (FCM) for fluorescent neuronal imaging of tethered but otherwise freely moving mice. This technique leverages two existing technologies: 1) T he highdensity coherent imaging fiberbundle ( CI FB) 2) T he electrowetting tunable focus lens (EWTL) . The CIFB is empl oyed for lateral tissue imaging while the tunable lens adds axial focusing capabilities, adding a third dimension to the data. I show their integration into a

PAGE 21

4 miniaturized imaging system that allows for the first demonstration of true 3D neuronal imaging in freely moving mice. I describe the development and validation of two types of 3D FCMs : 1) A confocal fiber coupled microscope ( C FCM) 2) A two photon fiber coupled microscope (2P FCM) In C hapter 2, I explore the history and current state of fluorescence in vivo neuronal imaging in mammals, with a focus on the development of miniature microscopes and endoscopes. In C hapter 3, I discuss the advantages and methods of achieving rapidly tunable focus in both benchtop and miniature microscope systems. In C hapter 4, I detail the development and testing of the C FCM with an integrated EWTL , which was the first demonstration of using tunable lenses for 3D imaging in a miniature microscope. In C hapter 5, I discuss the design and testing of the 2P FCM with integrated EWTL, including results of awake and fr eely behaving mouse neuronal imaging. I also discuss the challenges of performing twophoton microscopy through the C I FB . In the concluding C hapter 6, I discuss the potential improvements and promising applications of the FCMs , including deep brain imaging t hrough miniaturized GRIN lenses and opportunities for increasing resolution and imaging speed. I am fortunate to be working in an era of responsible animal research. All work presented here that involved live animals or animal tissue was approved by the Institutional Animal Care and Use Committee of the University of Colorado Anschutz Medical Campus , protocol # B 39615(05)1E .

PAGE 22

5 MINIATURIZED MICROSCOPY IN NEUROSCIENCE 2.1 Fluorescence microscopy in the neurosciences Image contrast Arguably the most important property in microscopy, contrast quantifies the amount to which one can separate the foreground objects of interest from the background of an image. Because of the very small differences in refractive index found in whole tissues , cells such as neurons cannot be readily distinguished from surrounding brain tissues without some contrast agent or technique. Many such techn iques have been developed in the past centuries to visualize neurons. Golgi’s method of silver staining was used by Santiago Ramn y Cajal in the 19th century to perform sparse labeling to allow visualizing neuronal processes in detail (Sotelo, 2003) . Crucially, the silver that fills the cells does not impart contrast, but rather contrast is th e result of the sparse cellular labeling , which allowed Cajal to see the expansive branching of a single Purkinje cell. An example of such a neuron is shown in Figure 2.1 (a) . Today, a variety of histological stains achieve straightforward contrast in tissue slices to aid in disease diagnosis, research, and forensics. However, they are not suited for providing in vivo contrast because they often require trans -illumination, are rarely compatible with living tissues, and are difficult to target to specific cellular populations. Although simple trans-illumination stains continue to be useful for histological studies, fluorescent probes have become the primary source of contrast for biological micros-

PAGE 23

6 copy in the past several decades (Specht et al., 2017). Figure 2.1 (b) shows a fluorescence image of a Purkinje neuron as a modern comparison, showing the highly branched dendritic arbor and detailed spines along the processes. Fluorescence microscopy Fluorescence microscopy can be performed in epiillumination mode, in which a single microscope objective lens is the conduit for both light illumination and collection . Fluorescent probes are excited using wavelengths that can be spectrally separated from the resulting fluorescence emission. Any non-fluorescing objects in the imaging field are dark, resulting in very large contrast ratios. Figure 2.2 shows a basic fluorescence microscope setup. The excitation and detection paths are combined spectrally with a dichroic beam -splitter, allowing the use of a single objective for epifluorescence imaging. This allows a greater variety of thick specimens, i ncluding whole animals, to be placed under Figure 2.1: (a) Purkinje neurons as represented in the 19th century by Ramn y Cajal (Sotelo, 2003) . (b) Modern fluorescence intensity projection image of a mouse Purkinje neuron filled with Lucifer Yellow fluorescent dye (Martone, 2002) .

PAGE 24

7 the objective for imaging. Importantly, the imaging optics must be optimized for both exciting the fluorescence and retrieving the emission light . The mechanism of fluorescence is dependent on the transition between electronic states of the fluorophore molecule. Figure 2.3 shows the Jablonski diagram for singlephoton fluorescence. A higher energy, shorter wavelength photon excites a fluorophore from the ground state (S0) to the higher energy state (S1), which will decay spontaneously by emitting a lower energy, longer wavelength photon that can be spectrally isoFigure 2.2: Illustration of the epi -fluorescence microscope. Monochromatic excitation light, selected by an excitati on filter, is reflected by a dichroic beam -splitter. An objective lens focuses the excitation light onto the sample and collects the emitted fluorescence from the fluorophores. The emission passes through the dichroic beam splitter and emission filter befo re being captured on a detector, such as a camera, to form an image of the fluorescence.

PAGE 25

8 lated with colored filters. This shift in wavelength is called the Stokes shift, and is a fundamental property of fluorescent probes. A powerful feature of f luorescent probes is that they can b e combined to be multiplexed by color, allowing multiple independently addressable contrast agents to be accessible with the correct choice of color filters. There are thousands of fluorescent probes available spanning the ultraviolet to the near-infrared spectra of light (Specht et al., 2017). They can be delivered to cells using a wide range of techniques, and can be made to localize to specific sub regions of a cell. The molecular versatility of fluorescent molecules has allowed for groundbreaking innovations, such as superresolution optical imaging and probes that have ac tivity dependent fluorescence. Figure 2.3: A simplified Jablonski -diagram showing electronic and vibrational states of a common fluoroph ore illustrates how a photon of the correct energy, related to wavelength ( ), is absorbed by a fluorophore to increase the electronic state from the ground state (S0) to the excited state ( S1 ). Spontaneous vibrational relaxation is followed by fluorescenc e emission as the fluorophore recovers to S0, with the emitted photon having less energy and longer wavelength.

PAGE 26

9 Geneticallytargeted fluorescent proteins Fluorescent genetically encoded Ca2+indicators (GECIs) and optogenetic actuators (Deisseroth, 2015) , continue to improve and they have become an indispensable part of the in vivo neuronal research toolbox . Prior to the recent functional advancement in GECIs, the primary tool of in vivo imaging were organic Ca2+sensitive membranepermeable dyes injected into the brain region of interest (Stosiek et al., 2003) . The invention of GCaMP6 was a milestone that propelled the use of GECIs for in vivo imaging because it achieved a sensitivity and dynam ic range greater than most organic dyes (Chen et al., 2013b) . GECIs can be targeted to specific cell types, and can be delivered by virus injection that can be stable in expression for many mont hs. Transgenic animals that constitutively express GECIs have also been made available (Chen et al., 2012; Dana et al., 2014) . In addition, optogenetic actuators, such as Channe lrhodopsin and its variants , have been used to leverage these imaging systems to allow for patterned simultaneous stimulation of spatially selective networks (Papagiakoumou et al., 2013) . Together, researchers can use the spatial and genetic specificity of these proteins to use a single op tical system to control and readout information from the intact brain (Emiliani et al., 2015). The remaining potential impact of these tools in neuroscience is in closing the loop between theoretical models developed from in vitro neuronal research and reality of the lucid and normally functional mammalian brain. 2.2 Modern m icroscopy techniques for studying the brain There are many microscopy techniques available, and more are being continuously developed. In this section, I will describe two of the most commonly used for achieved

PAGE 27

10 3D imaging in thick samples. These laser scanning techniques are also the basis of my FCM designs. Laserscanning c onfocal microscopy Confocal microscopy was invented by Marvin Minsky in 1957, but only achieved widespread adoption and use in the 1990s, then known as laser scanning confocal microscopy (LSCM), because of advances in laser techno logies, fluorescent dyes, fast beam scanner s, and sensitive photodetector s (Minsky, 1957; Webb, 1999) . LSCM is commonly used for reducing background signal during the imaging of fixed and fluorescently labeled tissue slices, but is also used for intravital microscopy in the context of clinical a nd research endoscopy (Jabbour et al., 2012) . The simplified LSCM setup is illustrated i n Figure 2.4. LSCM involves the use of a spatial filter, usually a pinhole, in the conjugate focal plane to reject out of focus light. This dramatically improves imaging contrast by shrinking the effective depthof field (DOF) to close to the diffraction limited axial focus of the imaging system (Webb, 1999) . The size of the pinhole is optimized either automatically or manually to minimize the loss of signal light but also to reject the out of focus background that limits the imaging contrast. The pinhole size is chosen based on the size of the diffraction limited spot magnified by the de scanning optics (Conchello and Lichtman, 2005) . This technique not only improves the axial resolution but also improves overall contrast by reducing the amount of background emission light collected from outside of the focal plane. An improvement in contrast is reported, resulting in a greater ability to resolve lateral features in some samples (Inou, 2006) . Details of the benefits of optical sectioning for brain tissue imaging are discussed in Chapter 3. Another version of confocal microscopy is spinning disk

PAGE 28

11 confocal microscope, in which disk of pinholes and a camera is used to parallelize acquisition and increase frame rate (Toomre and Pawley, 2006). In practice, LSCM requires additional mechanical and optical complexity to scan and image the laser spot across the sample. Typically, LSCM uses a mechanical lateral scanning system , such as galvanometric mirrors, to change the angle of the laser light entering an objective lens, which translates into a lateral movement of the focus spot across the sample. The fluorescence emission light must be de-scanned through the same scanning system, after which it is spectrally separated from the excitation light with optical filters. Figure 2.4: Confocal microscopy basic setup with stage-scanning. An optical pinhole in the conjugate focal plane is adjusted to reject out-of -focus fluorescence emission from a thick fluorescent sample. The laser is scanned across the sample, in this illustration with a translation stage.

PAGE 29

12 The point by point detection is acquired by a single element detector , such as a photomultiplier tube (PMT), and ma pped to a pixel value in the image. Fluorophores with spectrally separable emission colors may be excited and acquired at the same time using different excitation lasers and emission filters along with additional detectors. Two photon laser scanning microscopy The two photon absorption effect was first described theoretically by Mari a G ppert Mayer in her 1930 doctoral thesis, but it was not verified until the invention of the laser later in the 20th century (Gppert Mayer, 1931) . After the technologies that gave rise to LSCM and further advancement of lasers, two photon laser scanning microscopy ( 2P LSM ) was ultimately shown by Winfred Denk et al. in 1990 (Denk et al., 1990) . S ince then, it has become one of the most important opti cal techniques in neuroscience. 2P LSM allows researchers to reach targets in intact tissues that may not otherwise be accessible with other optical techniques. Like LSCM, 2P LSM can also recover thin optical sections, but can do so with far less background excitation, a less complicated optical light path, and maintaining discernable cont rast at remarkable tissue depths ( typically ~500m depth in brain tissue) . 2P LSM uses a point scanned focus to capture each pixel and construct an image . Figure 2.5 shows a basic 2P LSM setup. Fluorescence excitation is achieved through the nonlinear multiphoton absorption process. The excited electronic state of a fluorophore may be reached by the multiphoton absorption process. Multiphoton absorption occurs when multiple photons of fractional energies required to change electronic states impact the fluorophore near simultaneously, resulting in exciting the molecule followed by nor-

PAGE 30

13 mal fluorescence (Helmchen and Denk, 2005; Svoboda and Yasuda, 2006). The t wo -photon process is shown alongside the normal one-photon process in Figure 2.6. The multiphoton absorption process is very unlikely at normal intens ities, so the instantaneous photonic flux at the fluorophores must be made to be extremely high. To achieve this an ultrashort pulsed laser with pulse width < 100 femtoseconds (fs) and high pulse energy is focused onto the sample. Because of the non linearity of the multiphoton process, the excitation efficiency is proportional to the square of the instantaneous peak pulse intensity. The time averaged Figure 2.5: 2P -LSM basic setup with stage-scanning. A pulsed ultrashort excitation laser is focused onto the sample, resulting in a non -linear excitation of fluorescence limited to the focal volume. The fluorescence is filtered and recorded for each pixel by a single -element detector.

PAGE 31

14 fluorescence intensity generated by a pulsed laser source with average power , pulse width , and pulse repetition rate is approximated as (Diaspro et al., 2005): I , ( 2 . 1 ) where is the fluorophore’s molecular two photon cross section, is the quantum efficiency of the fluorophore, is Planck’s constant, and is the speed of light in vacuum. The numerical aperture (NA) is dependent on the acceptance angle, , of the imaging system in a medium with refractive index : NA = n sin ( 2 . 2 ) Rays at higher angles contribute more to higher spatial frequencies, so NA is the de facto descriptor for a system’s imaging resolution. However, in two photon excitation terms, Figure 2.6: Oneand two -photon absorption process are shown to both cause a similar transition to the excited electronic state, resulting in similar fluorescence output. The two -photon process requires two photons to arrive near-simultaneously with approximately twice the wavelength, or half of the energy, of the energy gap.

PAGE 32

15 Equation ( 2.1 ) shows a heavily non linear dependence of fluorescence generation on NA. This is because the NA both contributes to the excitation intensity at the focal spot and is also the primary factor responsible for collectin g the fluorescence light. Typically, objectives used for twophoton excitation microscopy in deep tissues have NA > 0. 8 (Helmchen and Denk, 2005) . This is a problem for the manufacture o f small imaging systems, because t he system aperture is constrained by the diameter of the physical optics, which in turns forces the optical design to compromise on other important parameters, such as working distance and field of view (Liang et al., 2002) . Fortunately, the excitation and collection NA can be decoupled. The time averaged intensity of fluorescence collected geometrically is : I , , ( 2 . 3 ) where is the fractional collection efficiency of an imaging system , summarized as: = 1 rF 1 r 1 rF @ # J A 2 ( 2 . 4 ) When imaging into scattering tissue, the excit ation NA may not be as important as the collection NA. This is both because high angle rays have longer optical paths, so have more chances to be scattered and affected by aberrations, and because there are diminishing returns as the focal volume decreases with high NA, resulting in too few fluorophores fluorescing. Therefore, a good efficiency twophoton imaging system can be designed that has a relatively low NA (~0.4 0.5) excitation path if the collection NA path can be increased independently. Such syst ems are often employed by using light collection paths

PAGE 33

16 that involve large area detectors, which can be replicated with large effective area optical fibers in miniaturized microscopes (Wu et al., 2009; Zong et al., 2017) . 2P LSM has other pros and cons compared to other imaging techniques . The intrinsic spatial confinement of the fluorescence excitation allows the rest of the optical system to be simplified, compared to LSCM, in the following ways: No de scanning of the emission light . Fluorescence emission is spatially confined by the excitation, so no pinhole is needed to remove out of focus fluorescence. Imaging optics do not need to be achromatically corrected as carefully. Excitation wavelengths are much longer than emission, allowing larger bandwidth spectral filters. On the other hand, 2P LSM creates the following additional optic al design challenges: An ultrashort pulsed excitation laser is needed, which cannot be miniaturized and may be deleteriously affected by propagating through long distances in glass. Requires special g lass coatings and optics that are efficiently trans missive in both the visible and near infrared (NIR) wavelengths are uncommon. Detectors in the optical setup are much more sensitive to external light because of their wide detection apertures , so the optical path must be light tight. In summary, 2P LSM is the most common optical technique for imaging deep into scattering while maintaining good contrast and optical sectioning. The unique challenges for two photon fluorescence excitation and collection warrant designing an imaging system specifically for 2P LSM.

PAGE 34

17 2.3 In vivo brain imaging methods Both LSCM and 2P-LSM are promising tools for microscopy because they offer good contrast at some depth in brain tissue. But in either case direct optical access to the tissues of interest needs to be attained surgically. Two common methods are described below. Cranial windows Cranial windows are used to provide optical access to surfacelevel brain structures with traditional microsco pe objectives. Typically, 2PLSM is performed on a headfixed mouse with a coverslip implanted in a small craniotomy above the brain region of interest, normally 3mm2 or less in size, illustrated in Figure 2.7. There are several variations on windows that show the flexibility of this technique. A similar procedure is skull-thinning and polishing, which can be healthier and stable for the brain tissue, but is time -consuming and may not be as optically clear (Shih et al., 2012). Removable windows and permeable windows have also been described (Goldey et al., 2014). Figure 2.7: Cranial window implanted in mouse to achieve optical access to cortical structures. An objective lens is used to image directly into the window.

PAGE 35

18 Examples of c ommon regions accessed by cranial windows are olfactory bulb (Wach owiak et al., 2013) , barrel cortex (Peron et al., 2015b) , and visual corte x (Andermann, 2010) . These regions display high levels of behavior dependent neuronal activity relatively close to the surface. The extended depthlimit of 2P LSM, compared with widefield microscopy or LSCM, allows access to neurons about 500 m below the brain surface. In terms of cortical layers in mouse , 500m is approximately the depth of the cell bodies found in layers 4/5 of structures such as the motor and somatosensory cortex , but does not reach deeper structures such as the hippocampus , located at > 1,000 m below the dorsal brain surface (Paxinos and Franklin, 2012) . To access deeper structures, it is necessary to excavate the tissue above the target structure or to implant relay optics, such as GRIN lenses. Gradedrefractiv e index (GRIN) lenses GRIN lenses have been classically used for simple and space efficient collimation of the divergent light exiting a single fiber core. In the past decade, GRIN lenses have been re purposed to be used for many miniaturized imaging applications (Jung and Schnitzer, 2003; Knittel et al., 2001) . The radial profile of the refractive index , , in a GRIN lens varies according to (Jung, 2004) : n ( r ) = n e . 1 rF g r 2 G ( 2 . 5 ) where is the gradient constant. R ays launched into one end of the GRIN lens will be focused based on the axial length of the lens , defined by the pitch length, .

PAGE 36

19 p = g ( 2 . 6 ) A pitch of 1 is a full sinusoidal period at a given wavelength. GRIN lenses are also available in a variety of NA values, which are defined by the focus angle in the external medium with index . This value can be approximated as (Jung, 2004): N A n e , gd n csc ( gL ) ( 2 . 7 ) where d is the GRIN lens diameter. These expressions show that GRIN lenses with smaller diameter, d, need a shorter pitch length to achieve the same NA. To achieve a long and thin GRIN lens for deep-brain implants at high NA, it may require a very high pitch. To get around this constraint, a relay lens of lower NA is used in front of a high NA objective lens. Figure 2.8 shows an illustration of two kinds of GRIN lens assemblies used for imaging. A singlet lens is shown compared to a twoelement lens with a low NA relay. By increasing the pitch number of the low NA relay, GRIN lenses of various lengths can be constructed for reaching deepbrain structures. Figure 2.8: GRIN lens illustrations. Top: Single GRIN lens with pitch ~0.5 with a magnification of 1. Bottom: Two -element GRIN lens with low NA relay at pitch ~0.75 and high NA objective with pitch ~0.25.

PAGE 37

20 GRIN lenses are engineered to be highly spaceefficient, achieving relatively high NA (u p to 0.5) for the small diameters (0.35 1.8 mm ). The disadvantages are that they are highly wavelength-dependent and have rapidly deteriorating offaxis performance for imaging. This is mainly due to the small aperture, which vignettes the rays of offaxis light, effectively reducing the NA towards the edges of the lens. Longer GRIN lenses also suffer from accumulated aberrations that reduce imaging quality. Even with the disadvantages, GRIN lenses offer a tool for providing optical access to otherwise deep brain structure s. Deep brain implants of GRIN lenses have been used with both widefield fluorescence microscopy as well as 2P LSM in structures such as the hippocampus (Barretto and Schnitzer, 2012a) . Animals may be head restrained with an appropriate objective used to image near the top surface of the GRIN lens, as shown in Figure 2.9. Usually a baseplate is implanted on the skull to provide stability to the fragile GRIN lens assembly. Figure 2.9: GRIN lens assembly implanted into deep -brain of mouse and imaged with a low NA objective lens to couple into the low NA GRIN lens element.

PAGE 38

21 2.4 Miniaturized micro scope designs It is becoming more important to perform physiological (both electrophysiological and optophysiological) neuronal recordings in awake animals. This is because of the difficulty of assessing the functioning of neuronal circuits in anesthetized mice, in whi ch the neuronal circuitry may function very differently compared to the awake state (Haider et al., 2012) . In addition, to study a large range of behaviors it is necessary to record from animals that move and function normally during the e xperiment. This gap has been somewhat addressed by placing headfixed animals on platforms or surfaces that allow them to move their limbs in a somewhat natural way. This also allows them to perform simple behavioral tests that are placed directly in front of them. Some researchers have also exploited virtual reality environments to increase the breadth of tasks that the mice may perform (Dombeck et al., 2010) . However, it is more ideal (and sometimes crucial to the experimental question) to allow the mice to physically traverse an environment. To address this gap, the past 15 years or so has seen attempts to create lightweight, miniaturized microscopes that are designed to be affixed to the skull of a rodent for stable imaging during awake, mobile behavior. Widefield miniature microscopes A successful technology, in terms of adoption by researchers, market presence, and publications, has been the miniature widefield epifluorescence microscope. The basic design is a head mounted microscope setup that incorporates a miniature camera , usually a CMOS sensor , a dichroic optical filter for spectral separation of the excitation and emission light , a single lens to form an image on the camera, and either onboard LEDs or

PAGE 39

22 separately fiber coupled excitation light sources. Notable examples of the widefield miniature microscope are the UCLA Miniscope (Cai et al., 2016) and the Inscopix nVista microscope (Ziv et al., 2013). Figure 2.10 shows the UCLA Miniscope . I mages shown are from miniscope.org. These widefield head attached miniature microscopes are capable of imaging the activity hundreds of cells using GECIs, owing to the large field-of-view of ~500m and high DOF , which allows collection from many cells in multiple planes at once. D eep brain structures a re also accessible wi th the use of longer GRIN lens relays. This technique requires a temporally varying fluorescence to reliably resolve cells because of the large -amount of light scattering at the visible wavelengths and because of the excitation of a la rge amount of out-offocus fluorescence. The low contrast of these microscopes means that morphological imaging is sacrificed to obtain high throughput imaging. Figure 2. 10: Example widefield miniature microscope design from miniscope.org. (a) Internal components of miniature microscope (weight 2 g), with integrated LED, CMOS camera, filters, and lenses. (b) Photo of mouse with head -attached miniature microscope.

PAGE 40

23 Laserscanning miniature microscopes LSCM and 2P LSM are more suitable techniques for structura l and high resolution imaging. T o take advantage of the benefits of laser scanning microscopy techniques, a laser source and point scanning system are required. For the laser source, high intensity laser light must be brought into the miniature microscope enclosure using an optical fiber. Fibers can be selected for the most efficient propagation of the laser light, generally such that the fiber core supports only the fundamental transverse electromagnetic (TEM00) mode , in which case it is known as a single mode fiber . In the special case of ultrashort laser pulses required for 2P LSM, the maintenance o f the peak pulse power is dependent on accounting for temporal dispersion, modal dispersion, polarization dispersion, and nonlinear effects (Agrawal, 2000) . These effects can be minimized by precompensating the pulses such that they c ancel out through the fiber, and by choosing the correct fiber (Agrawal and Potasek, 1986; Clark et al., 2001) . For example, hollow core fiber, in which the electric field exists mostly in an air filled core, is preferable fo r minimizing these effects for 1 – 5 m of fiber. The choice of fiber should also depend on its weight, flexibility, and cost for the practical purposes of mouse im aging. Regardless of the choice, a single fiber core must be focused onto the sample and actively scanned to form an image. Fiber scanning microscopy is a relatively well developed field, owed to the progress in endoscopic surgical and diagnostic clinical tools that use miniaturized imaging heads coupled with fiber optics to a detection system (Jabbour et al., 2012) . There are many techniques for acco mplishing this, two of the most common being piezoelectric resonance scanning of the fiber tip and integrated microelectromechanical systems (MEMS)

PAGE 41

24 actuated mirror scanning. Figure 2.11 (a) shows a simplified piezoelectric resonance scanner, in which the fiber tip is mechanically coupled to one or more piezoelectric actuators and forms a mechanical system that can be solved to predict the fiber -tip bending dynamics (Helmchen et al., 2013). Groups have shown the ability to create a variety of scan -patterns, including a spiral pattern, (shown in inset), (Engelbrecht et al., 2008; Myaing et al., 2006) , a Lissajous pattern (Flusberg et al., 2005a), and eve n raster/arbitrary patterns (Rivera and Brown, 2011). Figure 2.11 (b) shows a simplified MEMS -based scanner with two actuated mirrors, each scanning one of the two orthogonal axes. MEMS actuated mirrors can be controlled precisely for arbitrary beam -scanning, and groups have Figure 2. 11: Fiber-scanning techniques. (a) Piezoelectric resonance -scanning of the fiber tip, resulting in the scanning of resonant modes, such as the spiral shown inset. (b) MEMS actuated mirror scanners, resulting in arbitrary scan trajectories, such as the raster shown inset .

PAGE 42

25 shown a variety of implementations to reduce size, power requirements, and overall complexity (Qiu and Piyawattanamatha, 2017) . R ecently, a miniature microscope for mouse imaging, (weight of 2g), with a two axis MEMS actuated mirror was able to visualize dendritic spines in an awakebehaving mouse (Zong et al., 2017) . An extra consideration for these single fiber delivery systems is that the single mode fiber core may be ineffective for the collection of fluorescence emission, which must usually be detected separately. Some solutions include the use of l arge mode area fibers flanking the excitation core (Wu et al., 2009) , miniature single element detectors mounted on the microscope head (Helmchen et al., 2001) , and a separate emission path for large area coll ection fibers (Helmchen et al., 2013; Zhao et al., 2010) . Overall, these single fiber scanning techniques allow access to high efficiency laser transmission, which is especially useful for 2P LSM b ecause of the difficulty required in maintaining ultrashort pulse integrity at the focus. The cost of these techniques is the added mechanical complexity required to implement miniaturized laser scanning. These may result in challenges in the optical design that limit imaging performance. Frequently, it is difficult to achieve large scan ranges or maintain a large beam size through small mirrors. So far, these systems have been discussed only in the context of transverse scanning. Miniature fiberscanning microscopes that implement a third axial scanning dimension can become overly cumbersome in their complexity. Coherent i maging fiber bundle (CIFB) miniature micro scopes A CIFB is made from a longitudinally ordered bundle of optical fiber preforms that are drawn into dense canes with only a small amount of cladding between the cores (Wood et al., 2017) . The resulting CIFB is spatially coherent on both ends, allowing a

PAGE 43

26 pattern of light on one end to be represented by the cores on the other end, illustrated in Figure 2.12. CIFBs can be coupled with a variety of simple imaging lenses to allow for fullfield imaging. One of the simplest methods is to use optical epoxy to adhere an imaging GRIN lens (pitch ~0.5) to the distal end of the fiber bundle for use as a handheld imaging probe (P ierce et al., 2011) . A fluorescence image of a Fujikura CIFB surface is shown in Figure 2.13. Lateral -laser scanning microscopy using a CIFB can be accomplished by simply scanning the excitation laser focus across the proximal surface of the fiber. The light is sequentially coupled into each core and transmitted to the distal surface. The distal surface can then be imaged onto a sample to create targeted excitation of fluorophores. The emission light can be collected through the same optical path. This setup is convenient because all the filters, detectors, and scanning optics can be located proximal to the fiber bundle, and can potentially be those of a standard bench top microscope. Figure 2. 12: Illustration of proximal to distal image coherence of CIFB, as well as the pixelati on of an image formed through the bundle because of the discrete fiber cores.

PAGE 44

27 Even with the gained simplicity, t here are some disadvantages to using a CIFB compared to a singlecore fiber for miniature laser-scanning microscopy: The CIFB introduces an unavoidable pixelation to the image due to the necessary cladding between the fiber cores. Some techniques can be used to process the images to reduce t his artifact, which can be summarized by some forms of lowpass spatial filtering (Cheon et al., 2014; Han and Yoon, 2011; Han et al., 2010; Shinde and Matham, 2014). However, the fundamental lateral resolution limit of the imaging system is determined by the coreto -core spacing and magnification of the imaging system. To achieve small modefield area sizes for single mode propagation, it is necessary that the core size is small. Difficulties in manufacturing mean that most Figure 2. 13: Fujikura CIFB fluorescence image of a uniform target showing core distribution.

PAGE 45

28 small core CIFBs have cores of different shapes, sizes, and propagation properties. This may result in random signal heterogeneity that shows up in the captured images. Core to core coupling is a phenomenon that occurs because of the close inter core spacing in fiber bundles (Chen et al., 2008) . This artefactual spreading of light intensity is most critical with broadband light in the NIR wavelengths , as used in multiphoton microscopy. This also limits the density of fiber cores that can be used, and consequently the imaging resolution and fieldof view . Some CIFBs are much stiffer than a thin, single core fiber, with the mechanical stiffness depending primarily on the diamet er of the CIFB (and thus the number and density of the cores). This poses a challenge for developing a system for freely behaving rodents, as it limits their mobility in their behaving environment. Leached fiber bundles, in which the cores are only fused a t the ends of the bundle, may be a promising solution to this problem (Kostuk and Carriere, 2000) . While appreciating these challenges, CIFBs are used in th e work in this dissertation because they greatly simplify the distal optical setup, which is desirable when implementing the distal axial focusing mechanism. Additionally, the use of a CIFB may greatly reduce the cost of adoption for fiber coupled microscopes because it may allow researchers to use an existing b ench top microscope for fibercoupled imaging with few or no modifications. 2.5 Where we are in miniaturized neuronal imaging ? Since the rise of in vivo imaging, as a result of the development of 2P LSM, there has been a concerted effort to miniaturize these technologies for headattached behaving

PAGE 46

29 mouse neuronal imaging. Several designs had been proposed but the cost of adoption was not offset by the value of the results (Wilt et al., 2009) . Meanwhile, the field of medical endoscopy has become a powerful technological and innovative force as a result of the desire for less invasive surgical tools and a gr eater emphasis on diagnostic and preventive healthcare (Qiu and Piyawattanametha, 2015) . New tools, such as GCaMP6 and Channelrhodopsin, motivated the neuroscience field to appeal for more usable miniaturized optical solutions. Fiber photometry and miniaturized wide field microscopes have occupied that gap in the short term, because of the inherent simplicity of the approaches that lead to consistent results (Warden et al., 2014) . Only a small number of laser scanning fiber coupled microscopes currently exist for neuronal imaging, but none have seen similar upward trends in usage. Further, none of the existing designs truly leverage the optical sectioning capabilities for 3D imaging without introducing cumbersome complexity. In the Chapter 3, I argue that EWTLs offer an elegant solution for dramatically increasing the versatility of these devices by adding axial focusing.

PAGE 47

30 AXIAL FOCUSING IN MINIATURIZED MICROSCOPES Scientists have historically gravitated towards red ucing the dimensionality of 3D objects down to a 2D-plane. Illustrations like Santiago Ramn y Cajal's famous sketches of neurons stained with Golgi's method were the original imaging modality , limited by the pen and paper medium (Ramn y Cajal, 1888, 1899). Today, we can reconstruct volumetric represen tations of tissues in virtual space. In Figure 3.1 (a) , Ramn y Cajal illustrates the familiar layered cortical column structure. Figure 3.1 (b) is a modern 3D representation of a similar cortical column acquired with fluorescent labeling of neurons, three -photon excitation microscopy, and advanced software for 3Dreconstruction (Horton et al., 2013). The latter was acquired in 2013 at Cornell University, over a century following Ramn y Cajal's work . Neuronal networks exist in complex 3D-structures that cannot be fully represented by 2D-imaging. Beyond the conceptual benefits, 3D-imaging provides access to a larger number of cells and connections over greater spatial distances. In this chapter, I discuss efforts to acquire 3D representations of neuronal structure and activity from intact brains using optical methods to actively tune the focal plane of a fluorescen ce imaging system. First, I justify the need for an active focusing solution in optical-sectioning microscopy. I briefly review the technologies that exist at the scale of the benchtop microscope, including the standard motorized stages, remote-focusing, actuated objectives, and electrically tunablelenses (ETLs). Thenceforth I specifically examine the technologies that are compatible with miniaturized microscopes and discuss the current implementations of these techniques in the literature.

PAGE 48

31 Figure 3.1: (a) Sketch representation of the human motor cortex neuronal column by Ramn y Cajal (1899). (b) Modern 3D -reconstruction of a three -photon image of a cortical column in an intact mouse brain (adapted fr om (Horton et al., 2013) ).

PAGE 49

32 3.1 Benefits of minimizing d epthof field in fluorescence imaging Most optical imaging systems require light to be focused to a single surface to form a 2D image. The focal distance is selected to choose a region of the field that is in sharp focus, while other sources of light are defocused. In this first section, I present the motivation for an imaging system that discards the defocused light from the field in favor of increasing contrast for a single t hin focal plane, also known as optical sectioning. Lateral and axial proper ties of a focused Gaussian beam A discussion of the focal plane of an imaging system can be simplified by decomposing the plane into an array of focused light beams . When l ight is focused with a lens, it is straightforward to intuit the confinement of the focus spot in the 2D plane, ( lateral dimensions ) , perpendicular to the optical axis , (axial dimension) , as a radially symmetric spot with peak intensity in the center. In the axial dimension, however, the shape of the focused light intensity is not as clearly confined. To describe the confinement of the axial focus, I present a review of the 3D shape of a simple Gaussian beam. Below is the equation for the electric field ( E) of a G aussian beam propagating in free space, the lowest order solution to the scalar wave equation, derived in depth elsewhere (Born and Wolf, 1999; Li and Kogelnik, 1966) . E ( r , z ) = I w w ( z ) exp i ( kz ( z ) ) r 1 w ( z ) + ik 2R ( z ) Here, I is the normalized peak intensity of the beam focus. The wavenumber constant, k , is equal to / , where is the wavelength of light . The beam width function, w (z), is

PAGE 50

33 the envelope at which the intensity drops to 1 /e of the axial value. R (z) is the radius of curvature of the phase front and ( z ) is the scalar phase. w (z), R (z), and ( z ) are defined by the equations below: w ( z ) = w 1 + rl z z V rp ( 3 . 2 ) R ( z ) = z re 1 + F k w 2z G ri ( 3 . 3 ) ( z ) = arctan F 2z N k w G ( 3 . 4 ) The factor z in Equation ( 3.2) is the Rayleigh range, and is defined as the zdistance from the optical axis that the beam width, w (z), is 2 larger than w. It is also the distance from the beam waist at which the beam int ensity drops to half of its maximum value, I. z can be related to the beam waist by way of the expression: z = w 2k = N w ( 3 . 5 ) The intensity distributions functions along the lateral and axial axes can be derived from Equation ( 3.1) by computing the magnitude squared at z = 0 and r = 0: I ( r ) = E ( r , 0 ) = Iexp 2 r w G ( 3 . 6 )

PAGE 51

34 I ( z ) = E ( 0 , z ) = I w w( z ) = I 1 + @ z z (3 .7 ) The lateral intensity function I can be approximated as a Gaussian function. The lateral focus spot fullwidth at half maximum ( FWHM ) radial size is approximately: FWH M = w 2 2 ln 2 2 . 35 w ( 3 . 8 ) T hese expressions can be related to the NA of the focused beam, which is defined by the angular beam divergence , , at z z given a specific index of refraction, n, for the focusing external medium : 2 k w ( 3 . 9 ) NA = n sin ( ) ( 3 . 10 ) These expressions can be rearranged to represent the system resolution in the axial and transverse dimensions as a function of the imaging system NA. First, the beam waist size w is related to the NA of the focusing system by reordering and substituting ( 3.10) into ( 3.8 ) : w = n I N NA ( 3 . 11 )

PAGE 52

35 Replacing this result into ( 3.8) and ( 3.5 ) yields: FWH M 2 . 35 N n I NA 0 . 75 n NA ( 3 . 12 ) FWH M x = 2z V = 2 w N A (3 .13 ) Equations ( 3.12) and ( 3.13) are used to define the transverse and axial resolution of a Gaussian beam focused by an imaging system with a certain NA. The profile for w (z) is shown in Figure 3.2 (a) . The transverse profile of the beam at the Gaussian beam waist is shown in Figure 3.2 (b) , showing the transverse FWH Mr. The axial profile of the beam is shown in Figure 3.2 (c) , showing the axial Rayleigh criterion, z. Importantly, the lateral imaging numbers for FWH M are valid for the entire plane because those dimensions are radially symmetric. However, the axial imaging numbers for FWH M are only valid at r = 0, on the optical axis. Integrating the field at any axial position with a circle of increasing radius will eventually enclose the whole beam, meaning that the axial focus spot is never completely confi ned. The practical implication is that any source of light that intersects with this beam in the axial dimension may be detected, and may contribute to a low contrast, defocused background.

PAGE 53

36 Benefits of optical sectioning for improved contrast The 3D Gaussian focus described above describes the impulseresponse of an imaging system, also known as the excitation pointspread function (PSF) in 3D space. The 3D excitation PSF can also be shown by the illumination pattern of a unitary fluorescent element, or one in which the diameter is z. An approximation of this is shown by imaging a 2-m fluorescent microsphere embedded in clea r agarose gel with blue light in Figure 3.3 (a) . T he light source is incoherent , so may not behave precisely according to the above functions, but the approximate beam waist and envelope are still clear . Notable is the large amount of out-offocus fluorescence that is spread out laterally from the optical axis, even far from the diffraction limited z. The Gaussian beam definitions require Figure 3.2: (a) Gaussian profile of the w(z) function, indicating the x -axis parameter 0 and the z -axis parameter zR. (b) Gaussian transverse intensity profile at beam waist (z=0), indicating the w 0 value and the related FWH M. (c) Axial intensity profile along the optical axis (x=0), indicating the z FWH M

PAGE 54

37 that the power enclosed at the radius w(z) through the axial focus is constant, so the confinement of the axial light intensity is regulated by the slowly varying beam diameter. Although the focus spot along the optical axis is confined to |z|
PAGE 55

38 bead, in which the axial intensity spread is very similar along the optical axis, but the out offocus signal away from the optical axis is eliminated. Note that the lateral intensity profile of the bead also shows improved contrast compared to the widefield case, because of the rejection of light out-offocus axial planes. Figure 3.4 shows a plot of power encircled at a radius of 5 times the through the axial focus of each bead. In the widefield case, the total power along the optical axis is almost constant , showing that the small bead contributes backg rounds for a large axial range. In the optically sectioned bead, the power is much more confined, meaning that it will not create out-of-focus background after a much shorter axial range. In actual tissue imaging there may be many fluorescent particles in the path of the excitation beam that can contribute to a loss of contrast. If the goal is to recover high contrast details from a focal plane in tissue, it is critical to use an optical sectioning method. Figure 3.5 shows a similar imaging setup but with a slide containing fluorescent pollen grains taken with a 1.42 NA 60x oil-immersion objective lens (Olympus) and green excitation light at ~532nm wavelength. Each pollen grain is ~10m in diameter. Using these parameters, FWH M 390 nm . This means that the pollen is very thick compared Figure 3.4: Power encircled at radius 5FWHMr through focus of 2 m beads capture with widefield (solid line, no optical sectioning) and LSCM (dashed line, with optical sectioning).

PAGE 56

39 to the diffraction -limited focal plane. Although many of the fine features of the grains are still resolved under widefield illumination in Figure 3.5 (a) , the contrast is much higher and the extraneous background signal is absent with LSCM, shown in Figure 3.5 (b) . The shallow DOF from the LSCM technique also allows remarkable cross sections of the pollen grain , shown in Figure 3.5 (c) . Meanwhile, the large DOF of the widefield technique fills in these gaps with lower resolution out-of-focus light, suggested by the lack of fine features resolved in the center of the pollen grain. Focal plane size in LSCM and 2PLSM The tests above show the empirical results of op tical sectioning with LSCM, but it is important to quantify the 3D PSF for the two primary modalities discussed here. In LSCM the axial intensity confinement is dependent on the diameter of the pinhole relative to the size of the diffraction limited spot. The following expressions are the final derivations for that resolution in terms of the Gaussian focus (Cheng, 2006): Figure 3.5: Fluorescent pollen grains under green light excitation taken with a 1.42 NA 60x objective. (a) The pollen grains imaged under widefield illumination, show ing the loss of contrast in the lateral dimension and a large DOF fetching out of -focus grains. (b) A LSCM optical stack of several thin focal planes from the same pollen grains showing much higher contrast.

PAGE 57

40 w , 0 . 51 # ( 3 . 14 ) z V , 1 . {xI n rF n rF N A rp + F 2 d nf NA G ( 3 . 15 ) The pinhole diameter is d, and it is a separate term under the radical in Equation ( 3.15) . For an ideal system the pinhole is infinitely small, and the axial spread is just the first term under the radical. In a realistic system, the pinhole is set to about 80 % of w , . The two photon 3D PSF is summarized by the following expressions: w , 0 . 61 # ( 3 . 16 ) z V , 0 . uyNI N A ( 3 . 17 ) Here , w , is th e Gaussian beam waist, related to the lateral spread, and , is the Rayleigh range, related to the axial spread . Because of the longer wavelengths used for two photon excitation, the lateral resolution ends up being about twice the value of the confocal lateral resolution. Although the axial resolution for LSCM is better with an ideal pinhole, when considering realistic parameters, the two modalities have very similar optical section thickness (Diaspro et al., 2005) . In s ummary, optical sectioning is technique that is used to reduce the DOF of an imaging system . Widefield imaging techniques are still able to resolve similar small features compared to LSCM, but the larger DOF results in reduced contrast. DOF depends heavily on the sample environment being imaged and is increased in the presence of many fluorescent sources. Notably, even in the absence of extra fluorescent molecules, the DOF of

PAGE 58

41 widefield imaging is still much larger than the diffraction limite d Rayleigh FWH M. There are many layered sources of out of focus fluorescence in the brain. Diffuse fluorescently labeled neuronal processes, such as in the cortical pia, can generate large amounts of unstructured background. In general, as the effective D OF becomes larger , performing optical sectioning microscopy will have a greater impact on imaging contrast. Therefore, optical sectioning is an especially powerful tool for contrast enhancement for in vivo brain imaging. The cost of this contrast enhancement and shallower DOF is that an imaging system with a single focal distance will be unable to gather any information from regions outside of the focal plane. This limitation can be resolved by taking multiple images through the axial focus to generate a larger volume with high contrast. 3.2 Active a xial focusing methods Active focusing involves changing the focal plane during imaging. The number of discrete focal planes through which one can resolve meaningfully distinct structures is the thickness of the tissu e (or the depth limit of the microscope) divided by the DOF. For tissue imaging, optical sectioning greatly reduces the DOF and therefore increases the number of distinct focal planes. This can both be useful, because it allows for better 3D separation of tissue structure. Active axial focusing combined with optical sectioning can be used to generate a 3D representation of the tissues being imaged. Fast axial focusing methods may also be used to create laser scanning trajectories in 3D space, allowing effic ient sampling of cells in a volume rather than relying on a single plane of interest (Gbel et al., 2007; Nadella et al., 2016) .

PAGE 59

42 Three fundamental techniques that can be used to change the focal plane are shown in Figure 3.6. Most focusing techniques can be categorized under these methods. For inclusion in a mouseattached device it must be miniaturizable and stable, which means object -scanning techniques are likely not workable. An additional consideration is that changing the focal plane may induce unwanted optical aberrations, since microscope optics are designed to function at fixed focal lengths. Considering these requirements, in this section I describe some of axial focusing methods that can be adapted for miniature microscopes and their methods of implementation. Objective lens scanning Axial scanning of the focal plane for laser scanning microscopy can be achieved by moving the objective lens. Rapid mechanical movement of an infinitely -conjuga ted objective lens produces the fewest aberrations since objective maintains its designed focal length. Moreover, piezo actuated objective lenses been shown with extremely fast axial focusing, with speeds up to 80 Hz for rapid sampling of hundreds of active neurons in a Figure 3.6: Three fundamental axial focusing options. Object field scan changes the location of the imaging target relative to the optical system. Image field scan changes the imaging optics to change the focal plane. Tunable focus scanning changes the effective foca l length of the optics to change the focal plane.

PAGE 60

43 volume (Katona et al., 2012). Several objective scanning techniques have been demonstrated for miniaturized imaging: Miniaturized motors for GRIN lens actuation , Figure 3.7 (a) (Flusberg et al., 2005a, 2008a) Shape memory alloys for lens actuatio n, Figure 3.7 (b) (Wu et al., 2010) Piezo actuated lenses adapted for endoscopes (Sherlock et al., 2017). L inear motors with high speeds in hand-held endoscopes (Xie et al., 2006). Because of their relatively slow speed and lar ger size, lens actuators are typically used to optimize the focal plane prior to initiating imaging by moving a lens axially in the imaging system. Higher speed motors and piezoelectric actuators of optics are fast but tend to be heavy relative to the weight of the miniature microscope, which makes them a challenge to include in a lightweight design. Motors may become a more viable choice as highspeed motors at small scales become more accessible. Figure 3.7: Examples of published lens movement -based axials scanning. (a) Axial scanning by small focusing -motor actuation with 1.1 mm axial motion (Flusberg et al., 200 8a) . (b) L ens actuation by shape memory alloy contraction, with 150 -m of axial motion (Wu et al., 2010) .

PAGE 61

44 MEMS devices MEMS devices are nano fabricated mechanical actuators that can be designed for complex behavior at a submillimeter scale . Qiu and Piyawattanamatha comprehensively reviewed existing MEMS based endoscope devices, including current solutions for MEMS based axial focusing (Qiu and Piyawattanamatha, 2017) . Examples of MEMS based axial focusing methods include : A xial actuation of a miniature lens (Liu et al., 2010) . Refocusing with MEMS variable focus membr ane mirrors (Shao et al., 2004) . Thin film piezoelectric actuator with rotational scanning, shown in Figure 3.8 (a) (Qiu et al., 2014) . A dedicated Z axis mirror for axial actuation, shown in Figure 3.8 (b,c) (Li et al., 2016) Although MEMS based microendoscopes for small animal neuronal imaging have been developed, (see Chapter 2), none of the above axial integration techniques have been demonstrated in this application. MEMS based 3D miniature microscopes for neuroscience work are very promising due to their versatility and small size. However, the optical difficulty of refocusing with a mirror in a small space can be cumbersome for implementation.

PAGE 62

45 Tunable lenses Optical focusin g with tunable lens i s different from mechanical translation of the imaging optics relative to the specimen. Most commonly, optical focusing is accomplished with an electrically tunable lens (ETL) placed near the back -aperture of the objective le ns. Changing the phase curvature of the beam entering the back of an objective lens with an ETL translates the focal plane . In most ETL, there is a minimal inertial mass that moves during refocusing, which makes them ideal for highspeed applications. Large aperture ETLs are becoming a common addition to an objective lens for rapid focus scanning with no moving parts (Chen et al., 2014; Fahrbach et al., 2013; Grewe et al., 2011; Jabbour et al., 2014; Jeong et al., 2016; Ryan et al., 2017). Examples of the use of ETLs in miniature microscopes include: IR light responsive lenses for fiber -coupled focusing (Zeng and Jiang, 2009) Figure 3.8: Examples of published MEMS mirror devices with axial focusing. (a) Thin-film piezoelectric actuator for 190 m of vertical translation with parabolic mirrors for excitation and collection (Qiu et al., 2014). (b) Monolithic 3 -axis MEMS scanning mirror assembly, with vertical translation photos shown in (c), achieving 546 -m of vertical translation (Li et al., 2016) .

PAGE 63

46 A liquid filled electrically tunable membrane attached to a piezo actuated fiber coupled endoscope (Meinert et al., 2014) Shape changing polymer tunable len s used in a laparoscopic imaging device (Volpi et al., 2017) Using ETLs has the distinct advantage of having a minimal amount of moving parts and can be incorporated into most optical paths. Careful placement of the tunable lens near th e aperture stop of the imaging system will minimize the aberrations introduced to the light rays arriving from different angles for lateral scanning. Aside from needing to consider the effects of the divergent beams on the imaging properties of the objecti ve lens, ETLs generally introduce their own aberrations which need to be considered. Other axial scanning methods This section will introduce several other techniques that have not been implemented fully in miniaturized microscope systems, but may be compatible with such a system. Temporal focusing is uniquely useful for multiphoton microscopy. This method involves engineering the spectral chirp such that the highest intensity focus is at a defined focal plane (Oron and Silberberg, 2015) . This focal plane can be modulated rapidly, and the spectral properties can be calibrated and maintained for transport through an optical fiber, which obviates the need for a distal scanning mechanism. It also minimally disrupts the optical properties of the imaging objective. Temporal focusing is stil l limited in actual scan range, defined by the bandwidth of the incoming pulsed laser light. Wavefront shaping can produce exotic laser focuses and can be tuned rapidly with a high refresh rate spatial light modulator (SLM) (Horstmeyer et al., 2015) . Some examples of focusing patterns include being able to focus at multi ple axial and lateral positions

PAGE 64

47 simultaneous and creating arbitrary spatially selected patterns for simultaneous optogenetic excitation and imaging (Quirin e t al., 2014) . W avefront shaping is also powerful since it can be propagated through some optic fibers. Some groups have shown three dimensional scanning of a focal spot through a fibers with proximal wavefront shaping such that a specific focus pattern is form on the other end (Kim et al., 2014; Porat et al., 2016; Stasio et al., 2015) . The primary challenges in wavefront shaping for a f iber coupled is that it is heavily affected by the mechanical bending of the fiber, and must use optical feedback to rapidly alter the shaping pattern to compensate for changes imposed by the fiber and the tissue. If this problem can be solved it can obvia te the need for any distal optics, which would be the ideal system. Fixed multifocal systems implement multiple focal plane imaging intrinsically in the design. Several focal planes can be resolved by staggering optic fibers axially, so each fiber addresse s a different focal plane (Rivera et al., 2012) . Another demonstrated approach is to design the lenses such that multiple focuses are form from a single o ptical path (Ouzounov et al., 2013) . These systems are mechanically simple and spacesaving but they are limited in that there is no easy way for continuous axial scanning. Remote focusing is a promi sing technique that has already become a n effective solution for larger microscope systems. Botcherby et al. described the use of a second objective lens in a conjugate region of the primary objective to form a 3D image of the sample, and then to use a moveable mirror at the second objective to select the focal plane dynamically (Botcherby et al., 2008) . This system is attractive for miniaturization because the second lens only needs to be as large as and move as far as the secondary image, which makes small technologies like MEMS based mirrors an attractive option.

PAGE 65

48 3.3 Tunable focus by liquid lens technology Liquid lenses are a common variety of ETL that use one or more liquids as a shape changing refractive interface to allow for focus adjustment with minimal mechanical motion (Chiu et al., 2012) . Liquid ETLs have been used in several microscope systems to accomplish highspeed scanning. One example is their use in selective plane illumination microscopy (SPIM), which deals with large, high resolution imaging volumes acquired rapidly and directly benefits from the speed of the tunable lens (Fahrbach et al., 2013) . Liquid ETLs have also been used as an alternative to mechanical scanning in liveimaging systems for benchtop laser scanning microscopes . There are several common liquid ETL types that may be used for performing highspeed optical focusing for laser scanning microscopy (Blum et al., 2011) . Large aperture liquid ETLs are particularly suited for bench top systems because of their large focal range, high speed, and repeatability. An example of a commercially available 10mm aperture, 30mm diameter ETL (E L 1030C MV, Optotune AG, Switzerland) is shown in Figure 3.9 ( a) . This ETL functions by rapidly changing the volume of liquid in a container with a flexible polymer surface, which results in a focal length shift . Large aperture liquid ETLs have enabled rapid axial focusing, up to 100 Hz, in compact LSCM, 2P LSM, and SPIM systems without mechanical movement of the objective or sample (Jabbour et al., 2014; Jiang et al., 2015; Rickgauer et al., 2014; Ryan et al., 2017) . However, shapechanging polymer ETLs are not suitable for miniaturized headattached microscopes because of their large size and susceptibility to orientation and vibration.

PAGE 66

49 Electrowetting tunable lenses (EWTLs) are another type of liquid ETL that has applications in microscopy (Zhao and Wang, 2013). An example of an EWTL (Arctic 316, Varioptic, France) with outer diameter 7.8 mm and 2.5mm aperture is shown in Figure 3.9 (b). Electrowetting is a method for changing the wettability of a liquid on a dielectric surface by applying a voltage across the interface, effectively changing the contact angle of the liquid with the surface. To form a lens, two fluids with dissimilar refractive indices and hydrophobicity are placed in a cont ainer with dielectric side walls . The hydrophilic polar liquid has dissolved impurities to allow it to react to an external electric field. The contact angle of the liquid interface to the side-walls of the lens can be changed by applying an electric field across the lens. Eventually an equilibrium is reached between the electrostatic forces acting on the polar liquid and the surface tension of the system to create a lens surface with a stable curvature. Carefully matching the density of the liquids at the se scales makes EWTLs very resistant to the effects of gravity due to orientation and vibrations, which make it ideally Figure 3.9: Two varieties of commercially available liquid ETLs. (a) Optotune EL-10-30-CMV ETL based on a shape -changing polymer surface, with outer diameter of 30-mm and 10-mm aperture. (b) Varioptic Arctic 316, based on electrowetting lens technolo gy, with outer diameter of 7.8 -mm and 2.5-mm aperture.

PAGE 67

50 suited for a miniature head mounted system . Commercial EWTLs have also demonstrated very large tuning ranges, on the order of 50 diopter s. Further, the responses of EWTLs to input voltages are well described by simple oscillators. Although EWTL focusing speed is not as rapid as some other ETLs, engineering the voltage input function, the lens response time can be brought to less than 20 mi lliseconds, which is within the range of what is required for an active scanning system (Supekar et al., 2017a) . EWTLs also have the potential to perform extended optical functions, such as beam steering and wavefront shaping (Smith et al., 2006; Supekar et al., 2017b) . Because of the small aperture size, these lenses are not frequently used i n microscopy applications. However, EWTLs are good candidates for robust axial focusing solution for miniature FCMs. 3.4 Optical design considerations for axial scanning with a tunable lens Unlike the method of physically moving the samples or the imaging lens es, tuning the axial focus changes the way light propagates through the optical system to change the focusing distance. More specifically, the curvature at the back focal plane of the objective lens is modified. When transformed through a lens, this curvat ure of the wavefront gets transformed into an axial displacement of the paraxial focus spot. This is analogous to how a tilted wavefront becomes transformed by a lens into a transverse displacement of the focus. One important consideration when designing such a system is the magnitude of curvature that is needed to achieve a desired axial scan range. An ETL placed in the back focal plane of a lens directly shapes the quadratic wavefront entering the transformation.

PAGE 68

51 The wavefront phase term can be writ ten as a function of the pupil radius (Botcherb y et al., 2007) : ( ) = k Z NA 2 M O ( 3 . 18 ) where M is the objective magnification and Z is the distance at which the focus is formed with a given quadratic phase. Rearranging this in terms of the amount of axial change for a given change in wavefront phase curvature gives: Z = 2M kN A ( 3 . 19 ) This last expression shows that the ability of a tunable lens to change the axial focus of a lens system is dep endent inversely on the square of the objective magnification. This means that designing for high magnification with the purpose of increasing resolution or NA will result in a decrease in axial scan range for a given ETL. Another consideration is that any imaging system that produces magnification cannot perfectly represent both the axial and transverse focus transformations simultaneously. This effect is illustrated in Figure 3. 10. Botcherby et al. describe the consequences of this on imaging properties (Botcherby et al., 2008) . In real lens design scenarios, optical engineers will optimize for either transverse or axial focusing. Most commonly transverse imaging is preserved because imaging lenses are exp ected to have uniform performance throughout their FOV. The consequence is an accumulation of spherical aberrations along the axial range of the lens, shown in Figure 3.10 (c). One can also design for the Herschel condition, which instead forms perfect focus spots along the optical axis, but accumulates aberrations in the transverse dimension.

PAGE 69

52 In the FCM imaging system designs presented here, I optimize primarily for the Herschel condition, by constraining the optimization parameters in Zemax optical design software to maintain consistent imaging properties throughout the full focal range of the imaging system. One important characteristic that is maintained is telecentricity, which ensures that there is nearly no magnification change as the focal length is tuned. However, this choice sacrifices some imaging performance in the transverse dimension, and so imaging performance is maintained only up to the expected FOV prescribed by the CIFB diameter. Further, it is necessary to balance high magnification, which is required to achieve high resolution with the fiber-core spacing, with axial scanning range. In the Figure 3. 10: Axial focusing by wavefront curvature shaping at the objective back focal plane. (a) An input converging or diverging wavefront is transformed into an axial translation away from the designed focal length of the lens. (b) Spherical aberrations as a result of a lens obeying the sine condition.

PAGE 70

53 end the relationships between aperture size, focal length, transverse and axial consistency, and magnification become a careful interplay of compromises to reach a useful solution for miniaturized imaging . In this work the actual optical designs are sought empirically based on initial designs and guiding principles. The designs are restricted by the parameters laid out in the following chapters for size, resolution, FOV, and axial scan range. I also restri ct the designs to commercially available optics to allow for easier replication and more affordable designs.

PAGE 71

54 CONFOCAL FIBERCOUPLED MICROSCOPE FOR 3D IMAGING Confocal fluorescence microscopy is a technique that intentionally rejects fluorescence emission light outside of the desired focal plane . This effectively reduces the DOF to a closeto d iffraction limited axial sheet. LSCM is the laser -scanning variant of confocal fluorescence microscopy. When LSCM is combined with axial scanning , multiple layers can be combined to produce a 3D-representation of the sample being imaged. More details about the background of LSCM can be found in Chapter 2. Although this method has limited imaging depth in tissue compared with 2P-LSM, the unique optical path offered by the CIFB may allow for a practical 3D neuronal imaging solution. In this chapter I describe the design, construction, and testing of a miniature confocal fiber -coupled microscope (CFCM). The C FCM design takes advantage of the pinholelike structure of the CIFB to replicat e confocal imaging in a miniature imaging system. An integrated EWTL allows the C FCM to capture multiple axial planes to extend its capability to 3D imaging. The optical components of the device are combined in a custom 3Dprinted adapter with an assemble d weight of ~2 g that can be mounted onto the head of a mouse. Confocal sectioning provides an axial resolution of ~12m and an axial scan range of ~80m . The lateral field -of-view is 300m and the lateral resolution is 1.8m . Results of testing the C -FCM show bead images to quantify the resolution and scan range of the system, as well as ex vivo tissue 3D-imaging of mouse peripheral olfactory nerve labeled with yellowfluorescent protein ( YFP ). Some of the content in this chapter was previously published in the journal Optics Letters in 2015 (Ozbay et al., 2015).

PAGE 72

55 4.1 Miniaturiz ing c onfocal laser scanning microscopy The primary advantage of LSCM is the ability to extract thin optical sections; important for imaging cells in thick tissue , as argued in Chapter 3. Briefly, o ptical sectioning combined with axial scanning enables the col lection of 3D data from tissue and increases contrast that is required for imaging structural details of densely labeled samples in thick tissues, such as neurons in the intact brain. LSCM typically requires a complicated light path, which makes it difficult to miniaturize. Confocal imaging through a fiber bundle For most LSCM systems, emission light from the sample must be de scanned via the excitation path and then spatially filtered through a pinhole at a conjugate focal plane (Paddock, 2000) . Optical fiber cores have been used as confocal pinholes in many designs because they can simplify the detection path (Maitland et al., 2006; Meyer et al., 2016; Stelzer, 2006) . To reduce the size and number of optics, it has been shown that a fiber bundle can be used as an array of pinholes for spatial filtering, greatly simplifying the number of optics required distal to the fiber (Gmitro and Aziz, 1993) . Axial scanning with EWTL ETLs have been shown to be capable providing axial scanning with no mechanical actuation for s tandard microscopy applications (Chen et al., 2014; Fahrbach et al., 2013; Jabbour et al., 2014; Koukourakis et al., 2014; Meinert et al., 2014) . EWTLs are a type of liquid ETL that are highly resistant to motion are small enough to incorporate in a miniature C FCM system for active axial scanning (Zhao and Wang, 2013) .

PAGE 73

56 4.2 Methods Experimental setup Figure 4.1 shows the imaging setup designed to incorporate a commercial EWTL in the focusing optics. A laser-scanning confocal microscope (Leica SP5 II) provides the continuouswave (CW) laser for fluorescence excitation, steering of the be am laterally using a resonance beam scanner, and spectrally filtered detectors for fluorescence imaging. To fibercouple the device to the microscope, I use a high-density CIFB with 0.5 m length, 30,000 count fiber cores, total effective imaging diameter of 0.8mm, ~2.9m core diameter, and 4.5m inter-core spacing (Fujikura, FIGH-30-850N). The excitation laser is focused onto the proximal surface of the fiber -bundle using a 10X 0.4 NA Olympus objective lens. Imaging over a single focal plane is performed by raster -scanning the Figure 4.1: Laser scanning confocal microscope coupled to distal imaging optics through fiber bundle (IL: Imaging lens, EWTL: Electrowetting tunable lens, OL: Objective lenses).

PAGE 74

57 laser into individual fiber cores. At the distal end of the CIFB, the light from the cores is collimated with a 3mm focal length achromatic lens (Edmund Optics, 65 566), passes through the E WT L, and is focused onto the sample using two 2 mm focal length achromatic lenses (Edmund Optics, 65 565). The imaging system has ~2.5X magnification. For each fiber core, fluorescence emission is transmitted back through the optics into the same fiber core and registered on the confocal microsco pe internal photomultiplier detector. The EWTL is placed in the infinity space of the telescopic imaging system, where changing the focal length of the EWTL by voltage results in a shift of the front focal length of the C FCM. To control focus is a commercial EWTL (Arctic 316, Varioptic Inc.), which has a focal length that ranges from 57 to +29mm ( 17 to 36 m1 in diopters) corresponding to a voltage input from 25 to 60 VRMS provided through a flexible lens cable. Typical power draw of the EWTL is ~15 mW. A custom fabricated two part plastic adapter was designed (s hown in cross section in Figure 4.2 (a) ) to provide a lightweight and rigid enclosure for the imaging optics, including the 7.8mm diameter EWTL package. The adapter is designed to be surgically attached on a mouse head with the 1 mm objective lens to be inserted for deep brain imaging. The top section is easily detachable, and the bottom section is designed with a low profile < 5 mm height) such that it may remain implanted for longterm imaging. The EWTL and electrode are clamped between the two adapter sections and held in place with an O ring. The EWTL is separated from the skull b y ~2 mm, which provides adequate insulation from the electrical contact. The objective lenses are glued with cyanoacrylate adhesive into a length (3 – 7mm ) of polyimide tubing with 1.06mm outer diameter and fit into the bottom of the holder. Figure 4.2 (b) shows a photograph of the

PAGE 75

58 assembled C-FCM after being manufactured via a 0.1mm tolerance 3D -printing process followed by minor reworking. The adapter maintained good optical alignment over repeated assemblies. The final assembly, including the lenses, EL, and electrode, weighs approximately 1.9 g, comparable in size and weight to current miniature head-mounted microscopes for awake-behaving imaging (Piyawattanametha, 2014). Optical design I used Zemax optical design software to compute the optical performance of the CFCM with the objective lens extend ing 1-mm below the adapter, for surface level imaging, and extending 4mm below the adapter, for deep-tissue imaging. The adjustable design permits access to regions of interest in the brain. The Zemax models for the doublet lenses were obtained from Edmun d Optics. The Zemax m odel of the electrowetting lens was obtained from Varioptic. There is low chromatic aberration in this system due to the Figure 4.2: (a) CAD model of adapter with optics with lenses aligned in polyimide (PI) tubing. (c) Photo of assembled adapter and EWTL.

PAGE 76

59 selection of achromatic lenses and there are minimal changes for wavelengths ranging from 500600 nm. Results from the design are summarized in Table 4.1. The lateral resolution of 1.8m is determined by the fiber bundle core spacing (4.5 m ) de magnified by the imaging system. The optics are required to have good performance only up to the maximum spatial frequency resolvable by the fiber bundle. The spatial frequency is repr esented as l ine pairs per mm (lp/mm ), which is calculated as / . 300 / . The effective field of view (FOV) is calculated by finding the minimum object field diameter at which the modulus of the optical transfer function (MTF) at 300 lp/mm fails to meet the Rayleigh criterion (MTF > 0.2) (Born and Wolf, 1999) . The model shows a constant field curvature that results in a ~ 20m axial focus shift at the edge of the FOV compared to the center, which is managed by imaging thick tissue (> 20m thickness ) or by combining multiple optical sections. As the objective lens distance is increased, there is an increase in off axis vignetting and aberrations, most significantly astigmatism and distortion. The FOV at the longer objective length is predicted to be smaller due to the increased aberrations and vignetting. Table 4.1: C FCM optical parameters from Zemax model Objective length (mm) Scan range (m) Magnification Lateral resolution (m) Field of view (m) NA 1 211 – 288 2.4 – 2.6 1.8 – 1.7 ~300 0.36 – 0.35 4 212 – 290 2.2 – 2.7 2.0 – 1.7 ~240 0.41 – 0.35

PAGE 77

60 4.3 Results Optical perfor mance testing I evaluated the FOV experimentally by imaging a 15m -thick section of mouse brain showing neuroglial oligodendrocyte cells expressing green fluorescent protein ( GFP ). Figure 4.3 (a) shows a standard confocal fluorescence image of the sample using 488 nm excitation light. To compensate for field curvature, I performed a maximum intensity projection of 3 – 4 axia l planes at once, obtained by varying the EWTL focus. Figure 4.3 (b) and Figure 4.3 (c) show images for the 1mm and 4 mm objective lens distances. At the 4-mm distance there is a reduction in FOV due to offaxis vignetting and aberrations. The predicted FOV from Table 4.1 is marked with a dashed circle. I experimentally measured the power loss across the FOV using a thick fluorescence test slide (Chroma) shown in Figure 4.4. The decreasing signal is because of vignetting and aberrations towards the edge of the FOV and was unchanged when varying the axial focus with the EWTL. From these results I posit that i ncreasing the objective length only reduces the FOV while otherwise maintaining similar image quality. Figure 4.3: Lateral imaging characteristics of C-FCM at different objective depths. (a) GFP labeled neuroglia imaged with 20X, 0.8 NA objective. (b) Same region imaged with C-FCM at shortes t and (c) longest objective lens length.

PAGE 78

61 The images were processed by band stop filtering using a spatial fast Fourier transform (Matlab) to remove the fiber pixelati on artifact, illustrated in Figure 4.5. Individual cell bodies and neuronal processes can be resolved in the image, showing subcellular resolution, comparable to the standard confocal microscope images obtained with a 10X 0.3 NA objective lens. The effects of distortion and NA variations are not visible within the defined FOV. I experimentall y verified the axial scan range of the C-FCM. A nanopositioner stage (Mad City Labs, LP100) with 100m zscan range was used as an axial ruler. I imaged a thick test sample consisting of 1m diameter red fluorescent beads (Invitrogen, F8887), embedded in agarose gel, using 561nm excitation light. Initially, the EWTL was fixed at Figure 4.4: Imaging power loss over the full FOV with 1mm distance objective lens (black line) and 4 -mm distance objective lens (grey line). Figure 4.5 : Zoom -in showing the cell bodies indicated by white arrow in Figure 4.3 (a). Left: Raw C-FCM image showing pixilation due to fiber -cores. Middle: C-FCM image after filtering. Right: Comparison with standard confocal microscope.

PAGE 79

62 the shortest focal length and I obtained axial image sections at 1m intervals using the nanopositioner stage. N ext, I fixed the sample position and obtained 36 optical sections of the same regions by varying the EWTL across the full focal range. Figure 4.6 (a) shows an orthogonal projection of several beads resulting from varying the stage position an d the EWTL focal length. Figure 4.6 (b) shows the mapping of bead axial centroids from the stage position to the EWTL focal setting. Data collected from 40 beads agree with the simulated focal length dependence obtained in the Zemax model. I conclude that the C-FCM provides a scan range of approximately 80m . I tested the lateral and axial resolution by imaging agarose samples containing 2 m or 1m red fluorescent beads. The confocal pinhole on the microscope was set either to the open position, allowi ng light collection from multiple fibers, or closed to a setting of Figure 4.6: Axial scan range of electrowetting C-FCM (a) Orthogonal projection (inverted grayscale) of 1 -m diameter red fluorescent beads in agarose imaged with scanning Z -stage (left) and scanning the EWTL (right). (b) Black dots: 40 beads mapped from EWTL optical power to relative Z -position. Gray lin e: Simulated CFCM focal length with varying EWTL power.

PAGE 80

63 2 Airy patterns, which allows light collection from only one core. Figure 4.7 (a) s hows lateral and axial bead images. With the pinhole open, out-offocus fluorescence emission leaks into adjacent fibers, but is eliminated by closure of the pinhole due to the confocal sectioning capability provided by the fiber cores. Figure 4.7 (b) shows the profiles of beads that were imaged with the closed-pinhole configuration. The theoretical resolution limit is shown in grey using the equations for lateral resolution in Equation (3.14) and the pinholelimited axial resolution in Equation (3.15) , repeated below in terms of FWHM. , 0 . 51 þØ5ÜAØ5ÜA# ( 3 . 14 ) (* / , 0 . 88 J rF J rF # rp + rl @ þØ5ÜAØ5ÜA# rp ( 3 . 15 ) Figure 4.7: Lateral and axial resolution of C-FCM. (a) Lateral and axial images of fluorescent beads. Left: 1 -m bead with open pinhole (PH), Middle: 1-m bead with closed pinhole, Right: 2 -m bead with closed pinhole. (b) Top: Averaged line profiles of 1 and 2-m beads (black) compared with diffraction -limited resolution (grey) Bottom: Averaged axial profile of several 1 and 2 -m beads (black) compared with theoretical axial resolution (grey).

PAGE 81

64 The calculations use object space NA = 0.35, refractive index of air n = 1, wavelength = 600 nm, and pinhole diameter dph = 1m (2.5m fiber core size de magnified by 2.5x). As shown in Figure 4.7 (a) the 1 m diameter beads occupy a single fiber core while the 2 m beads are sampled by multiple cores. This indicates that the C FCM lateral resolution is fundamentally limited by the fib er core spacing. The lateral resolution is estimated to be 1.8m , determined by de magnifying the 4.5m core spacing by 2.5X, compared to the theoretical 0.87m resolution from Equation (3.14) . As shown in Figure 4.7 (b) , the FWHM axial resolution is determined to be ~10and ~12m for the 1and 2m diameter beads respectively, compared to the theoretical 9.3 m resolution. Cross talk of multiple fiber light collection may explain the difference in axial resolution. Tissue sample testing To demonstrate imaging in tissue with the C FCM, I im aged intact mouse olfactory nerve fibers expressing yellow fluorescent protein in olfactory sensory neurons (Li et al., 2014) . The mouse was sacrificed by CO2 inhalation, according to existing protocols, and the head was bisected sagittally to expose the olfactory epithelium and nerve. The C FCM was held in position adjacent to the tissue using a manipulator arm with an aqueous saline solution interface bet ween the objective lens and tissue. Imaging was performed using a 488nm laser at a resolution of 1024 by 1024 pixels at 1.7 seconds/frame. 36 image slices were taken while varying EWTL optical power from 13 to 5 m1. The images were post processed by ban ds top filtering . Figure 4.8 (a) shows four separate optical sections spanning ~50m , limited by light scattering in the tissue. Each of these four i mages represents a ~12 m optical section with distinct morphological features, demonstrating efficient optical sectioning with this device. Figure 4.8 (b) shows a maximum -

PAGE 82

65 intensity projection of a stack of optical sections. The diameter of each axonal bundle is 10m and is easily resolved. 4.4 Summary This is the first demonstration for the use of electrowetting tunable focus lens technology in a fluorescence C-FCM to allow full 3D tissue imaging. I validated a FOV of 300m , and ~1.8m lateral and ~12 m axial resolution. I verified experimentally an axial scan range of 80m . I showed 3D images of detailed nerve fibers showing axonal networks. The C-FCM enclosure is a simple and lightweight adapter that can be modified for use as a head -mounted device for brain imaging. Further improvement of low-voltage EWTL technology will allow for more diverse endoscopic applications (Niederriter et al., 2013) . This approach is promising for use as an implantable device because it has no mechanical fatigue, no vibration, and low power consumption. Figure 4.8 : 3D -imaging of mouse nerve tissue. (a) Four optical sections that were taken at specific EWTL optical power settings. Scale bar is 100 -m. (b) Maximum intensity projection of an image stack of intact olfactory neuron axons labeled with YFP.

PAGE 83

66 This C FCM concept with a tunable lens can also be applied to multiphoton imaging of fluorescent indicators, but require dispersion compensation to achieve good signal levels. The next chapter will focus on the adaptations that are made to the C FCM design for two photon imaging.

PAGE 84

67 TWO PHOTON FIBERCOUPLED 3D MICROSCOPY 2PLSM is a widely -used fluorescence imaging technique in neuroscience because it yields optical access deeper into tissue compared to standard single -photon fluorescence microscopy (Denk et al., 1990; Peron et al., 2015a). Like LSCM , 2PLSM allows for the recovery of a small DOF, which can be combined with axial scanning to achieve 3Dimaging. 2PLSM surpasses LSCM with increased imaging depth in solid tissue and reduced phototoxicity, making it the preferred imaging technique for in vivo ne uronal imaging. When used in combination with rapid axial scanning, it is possible to generate 3D representations of neuronal structure and activity (Gbel et al., 2007; Nadella et al., 2016). Additional details on the history and methods of 2P-LSM can be found in Chapter 2. Though 2PLSM is powerful for in vivo imaging, the size and complexity of the ultrafast laser excitation and unique optical requirements for efficiency light collection make it challenging to miniaturize. In this chapter, I describe the design, construction, and testing of the two -photon fibercoupled microscope ( 2P-FCM), with axial scanning enabled by an integrated EWTL and lateral scanning achieved with the use of a CIFB. Use of an EWTL is ideal for this application as it is lightweight, compact, has low-power requirements, and is immune to motion and orientation (Berge and Peseux, 2000; Blum et al., 2011). The optics are packaged in a lightweight 3D printed enclosure. Pulse propagation through the CIFB is controlled by careful pre-compensation of the dispersion of glass and verified by spectral ly resolved autocorrelation measurements. Verification of the 2P-FCM performance is shown by imaging resolution test targets and

PAGE 85

68 fluorescent beads in th ick agarose preparations . Finally, in vivo neuronal activity is shown by 3D imaging of neurons in the motor cortex of a freely behaving mouse with the head attached 2P FCM. A baseplate is permanently affixed to the mouse skull for attachment and alignment for repeated imaging over the same brain region. The 2P FCM presented here is the first device to attain 3D imaging of neural activity in a 240m diameter by 180m depth in the brain of a freelymoving mouse with no mechanically moving parts. 5.1 Miniaturizi ng t wo photon laser scanning microscopy 2P FCMs use ultrashort pulsed lasers coupled through an optical fiber. The light is typically focused and scanned across the sample to form an image. The fluorescence emission is collected and delivered to a highsensitivity photodetector, such as a photomultiplier tube (PMT) or avalanche photodiode (APD). There are two main challenges when fiber coupling 2P LSM: Ultrashort pulse propagation The excitation source must reach the tissue sample with high peak pul se power to efficiently result in multiphoton excitation (see Chapter 2 for more details). Propagation through a solid core fiber may result in tem poral broadening due to chromatic dispersion of the glass, modal dispersion in the fiber cores , and Kerr induced nonlinear effects that diminish the peak pulse power to a point where multiphoton excitation is not possible. Efficient emission collection Em ission light from 2P LSM must be efficiently collected from as large a field as possible, since it is assumed that the excitation region is spatially restricted for

PAGE 86

69 each focusspot. In fiber coupled systems, there must usually be a second optical path for emission collection. Several types of laser scanning FCMs and microendoscopes for twophoton imaging have been developed that have solutions for these challenges (Flusberg et al., 2008b; Helmchen et al., 2013; Sawinski et al., 2009; Szabo et al., 2014) . In addition to in vivo neuroscience research, these devices have clinical endoscopy applications where they can deliver a variety of advanced imaging techniques in miniature packages (Liu et al., 2015; Qiu and Piyawattanametha, 2015) . Lateral scanning and axial focusing methods for miniature laser scanning microscopes are explored in Chapte rs 2 and 3, respectively. For lateral scanning in the 2P FCM described here I use a CIFB because of its proven effectiveness performing laser scanning microscopy while maintain mechanical simplicity to allow for experimentation with tunable focus optics (Flusberg et al., 2005b; Gbel et al., 2004) . For active axial focus ing, I implement an EWTL because of the compact size, large focusing range, and resistance to motion (Berge and Peseux, 2000; Bower et al., 2012; Hendriks et al., 2005) . 5.2 Methods Overall system design The experimental setup for imaging with the 2P FCM is shown in Figure 5.1. The distal optics of the 2P FCM are housed in a two part 3D printed enclosure, which can be repeatedly attached to a baseplate affixed to the animal’s head. A CIFB couples the distal imaging optics to a custom 2P LSM system to relay the excitation laser and collect emitted fluorescence from the sample. Laser scanning over the CIFB forms an image of the

PAGE 87

70 sample while the fluorescence collected back through the fiber cores is detected by photodetectors housed in the 2PLSM after spectral filtering. The individual components of the system are discussed in detail below. Lasersource The excitation source is a SpectraPhysics MaiTai HP Ti:Sapphire pulsed laser, with ~80 fs duration pulses tuned to a center wavelength of 910-nm and operating at 80MHz repetition rate. The beam power is controlled by a halfwave plate o n a rotation mount (Newport ConexAG-PR100P) followed by a GlanTaylor polarizer (Thorlabs GT10 B). Figure 5.1: 2P -FCM imaging system. Pulses from a Ti:Sapphire laser source are spectrally prebroadened through polarization maintaining fiber (PM fiber) and pre -chirped using a gratingbased pulse stretcher. Output pulses are scanned onto the surface of the CIFB thro ugh scanning mirrors, scan/tube lens relay, and 10x Objective. Fluorescence emission is collected by the CIFB and directed to a PMT through a dichroic filter. The collected pulses are amplified and transformed to logic levels to be detected by the DAQ and PC.

PAGE 88

71 Spectral and temporal pre compensation A polarization maintain ing (PM) single mode fiber (Thorlabs) is used for spectral pre broadening to counter the spect ral narrowing that occurs while propagating the highpower light through the fiber bundle cores. After propagating through the PM fiber, the beam size is expanded by 5x to reduce the average intensity on the gratingpair . The grating pair is used to apply ~40,000fs2 of negative group delay dispersion (GDD) on the laser pulse to compensate for the positive GDD that occurs when propagating through the 1.0m length of optical fiber. The gratings are reflective ruled gold with a density of 300grooves/mm (Edmund Optics 49 572) separated by 160mm. Following the gratings, the beam size is reduced by 4x. 2P LSM benchtop system The beam is routed through a galvanometric mirror scanning system (Cambridge Technologies, 6215H) and relayed through a 50mm FL scan len s and 180mm FL tube lens in an Olympus IX71 research microscope. The beam is scanned over the surface of the CIFB through a 10X 0.35 NA Olympus UPLANSApo objective lens. A XYZ translation stage (Thorlabs, CXYZ05) is used to accurately align the fiber to t he focus of the objective lens. The emitted fluorescence is collected by the distal imaging optics and propagates through the fiber cores of the CIFB. The fluorescence passes through a low pass filter (Chroma T670LPXR) and is split by a dichroic filter (Se mrock FF562Di02) for dual detection of GCaMP6s and tdTomato. The signal is detected on nondescanned photon counting large area PMTs (Hamamatsu H7422PA 40) for the two channels. The output pulses from the PMT pass through a highbandwidth amplifier (Becke r & Hickl GmbH

PAGE 89

72 ACA435db) and are converted to logic level pulses by a timing discriminator (6915, Phillips Scientific). The pulses are counted by a data acquisition (DAQ) card (National Instruments PCIe 6259) at a rate of 20MHz. The counts are sampled and binned by pixels and converted into image in custom software in Labview (National Instruments). The software also controls the EWTL driver. 2P FCM miniature optical system design The imaging system for the 2P FCM was designed in Zemax optical design sof tware. Models for the stock lenses and for the EWTL were obtained from the manufacturers (Edmund Optics, Thorlabs, and Varioptic). The CIFB (Fujikura Ltd. FIGH 15600N) has an outer diameter of 700 m , an active image diameter of 550 m , and a length of 1.0 m. There are ~15,000 cores with core to core spacing of 4.5m and core diameter of 3.2 m , as previously reported by Chen et al. measured with scanning electron microscopy (Chen et al., 2008) . The miniature optics contained in the headmounted 2P FCM im aging system are show n in Figure 5.2. The fiber coupling lens is an aspheric lens ( FL: 6.2mm, diameter : 4.7mm, Edmund Optics 83710) that collimates the light diverging f rom the fiber bundle. The electrowetting lens is placed in the collimated beam and the light is refocused onto the sample by an objective lens consisting of a planoconvex lens (FL: 7.5mm , diameter : 3.0mm, Edmund Optics 49177), and an aspheric lens (FL : 2.0mm, diameter : 3.0mm, Thorlabs 355151B). The nominal magnification of this imaging system is 0.4x and fieldof view (FOV) is ~ 220m , corresponding to the de magnified CIFB imaging diameter. Similarly, the core sampling resolution is the de magnified core spacing, which is ~1.8m .

PAGE 90

73 A commercially available EWTL (Varioptic Arctic 316) is used to control axial focusing of the 2P-FCM imaging system. The EWTL contains two immiscible fluids with different refractive index and equal density, which interact via surface tension. An applied electric field adjusts the curvature of the interface and thus the focal length of the EWTL (Berge and Peseux, 2000). The predicted working distance and other imaging properties from Zemax at these three settings are summarized in Table 5.1 . The optical power range of the EWTL is specified as -16 to +36 diopters. The optical system is optimized through the full focal range of the EWTL to minimize magnification change, maximize t he axial scan range, and maximize the working distance of the 2PFCM. Chapter 3 discusses the requirements for maintaining imaging performance throughout the full focal range. Figure 5.2 : Optics of the 2P-FCM miniature microscope head that focuses excitation light from CIFB cores onto the tissue. The CIFB -coupling asphere collects the light from the cores of the CIFB, which are then passed through the aperture of the EL. The plano -convex lens and the objective asphere focus the light onto the tissue through a #1 coverglass with 0.15 mm thickness.

PAGE 91

74 Table 5.1: 2P-FCM optical parameters from Zemax through range of focal lengths. EWTL control (VRMS) EWTL optical power (m-1) Working distance (m) Magnification NA 60 35 450 0.41 0.43 42 0 570 0.40 0.44 25 -16 690 0.39 0.45 The 2P-FCM has low excitation NA compared to what one would typically use for two -photon excitation, which is usually > 0.8 NA. It has been shown that low NA objective lenses may be well suited for efficient two -photon imaging because of the larger volume of the 3D focus spot (Singh et al., 2015). High NA objectives are desired more for capturing as many emitted fluorescence photons as possible, which can arrive from unpredictable field locations after being heavily scattered through deep tissue. In the 2PFCM, the large effective area CIFB combined with achromatic misfocusing result in a Figure 5.3: NA comparison of forward excitation light at 910nm (top) and backward emission light at 532 -nm (bottom). Largest emission and excitation field positions are matched at 220m FOV.

PAGE 92

75 higher collection NA of ~0.6, averaged over the EWTL focuses. This is illustrated in Figure 5.3 showing either 910-nm forward excitation or 532nm backward emission. 3Dprinted enclosure design The enclosure for the 2P-FCM optics is designed in Solidworks 3D CAD soft ware (Dassault Systemes). The packaging is split into three sections: top, bottom, and baseplate, shown in Figure 5.4: .3D-printed enclosure (a) A two-part 2PFCM snaps together to secure the EWTL and electrode. (b) Photo of 3D-printed parts, top: before any processing with supports still attached and bottom: assembled 2PFCM. . The top-section contains the CIFB ferrule, held in place by two set-screws, and the fibercollimating asphere. The bottomsection contains the objective lenses. The unmounted lenses are held in by friction in precisely sized openings. The topsection has two curved tabs that interface with slots in the bottom -section, which help to ensure reproducible alignment. The Figure 5.4: .3D -printed enclosure (a) A two -part 2P -FCM snaps together to secure the EWTL and electrode. (b) Photo of 3D -printed parts, top: before any processing with supports still attached and bottom: assembled 2P -FCM.

PAGE 93

76 EWTL and the electrode are sandwiched between the bottom section and the topsection with an O ring that ensures good e lectrical contact. The flat flex electrode cable exits the enclosure through a small slot between the sections. The topsection tabs have a single thread at the end, which interfaces with the baseplate as shown in Figure 5.5 (a) . In this way, the baseplate is pulled up against the bottom section by the topsection. This greatly improves rigidity when attached to a moving animal. The baseplate is designe d with ridges and holes to improve adhesion of the cement for attachment to the animal skull. The entire enclosure is 3D printed using a high resolution projection based resin printer (Kudo3D Titan 1), with a resolution of 50m . The material used is a pho to curable resin (3DM XGreen) dyed with 0.5% molybdenum disulfide to decrease light scattering and thus increase feature resolution. A photo of the top and bottom sections immediately after printing are shown in Figur e 5.5 (b) . 3D printing allows optimization of the prototype and easily enables design changes, such as the inclusion of GRIN lenses for deep brain imaging (Barretto and Schnitzer, 2012a) . Test sample preparation Resolution and axial scan range measurements were performed by imaging fluorescent beads embedded in agarose (SigmaAldrich A9414) and a USAF 1951 resolution target (Edmund Optics 38257). 2m yellow green fluorescent beads (Invitrogen F8853) were used to measure axial scanning extent as well as lateral and axial resolution. Low melting point agarose was prepared at a concentration of 0.5% in water. The 2 m diameter fluorescent yellow green beads were diluted in the agarose to a concentration of ~2.0 x 107 beads/mL. Approximately 2.0mL of solution was placed on a #1 coverglass and allowed to set at room temperature. The beads were imaged in sequential

PAGE 94

77 axial planes by a 20x 0.75 NA Olympus UPLANSApo objective with a motorized stage and separately by the 2P FCM by changing the voltage applied to the EWTL. Mouse imaging setup All experiments were approved and conducted in accordance with the Institutional Animal Care and Use Committee of the University of Colorado Anschutz Medical Campus. Male 3 month old C57BL/6 mice were anesthetized by intraperitoneal ketamine xylazine injection. The skin above the target site was numbed by lidocaine injection and retracted to expose the skull. The mouse was injected with an adenoassociated virus driving expression of GCaMP6s under the synapsin promoter (AAV5.Syn.GCaMP6s), similar to procedu res in (Chen et al., 2013b) . The coordinates of the injection targeted the hindlimb somatosensory cortex, 0.2 mm posterior to bregma and 1.5mm lateral to the midline, at a depth of 300m (Paxinos and Franklin, 2012) . The injection volume was 0.66 microliters delivered with a glass micro pipette through a 0.5mm hole drilled at the target site. One month after injection the mic e were implanted with an optical cranial window near the injection site, using standard techniques as previously described. Briefly, mice were anesthetized by isoflurane inhalation and the skin under the scalp was numbed by subcutaneous lidocaine injection. The skin above the skull was removed to expose the injection site and skull surface. A 2 mm square window of skull was removed immediately anterior to the injection hole to expose the dura mater. The opening was covered with a 2mm square #1 coverglass a nd secured in place with cyanoacrylate glue. Dental acrylic cement (C&B Metabond) was used to cover the skull surface. The presence of fluorescence

PAGE 95

78 signal was confirmed with standard 2P LSM using a 20x 1.0 NA Zeiss PlanApochromat water immersion objective . The baseplate attachment procedure is similar to what has been described for other miniature head attached microscopes. While the mouse was still anesthetized, the 2P FCM was held and positioned above the window with a micromanipulator (Sutter MP 285) until fluorescent signal could be observed with widefield epifluorescence through the bundle. The target region was chosen with twophoton imaging and the 2P FCM was positioned to the region with the baseplate attached. The baseplate was then secured to the existing acrylic with black acrylic cement (Lang Dental Jet Acrylic). After allowing to set for ~30 minutes, the 2P FCM was removed, leaving the baseplate in place, and the mouse was allowed to recover. The imaging setup is illustrated in Figure 5.5. The mouse was lightly anesthetized with isoflurane inhalation. The baseplate was carefully gripped by thumb and forefinger and the 2P FCM was inserted and secured with a quarter turn. The EWTL electrode was connected to light gauge wires that were draped, along with the CIFB, over a horizontal metal post above the behavior cage. The mouse was allowed to recover in the behavior cage for imaging. The cage was illuminated by red light to minimize coupling into the fluorescence detection path and a camera (Logitech C615) was positioned above the cage to monitor behavior during imaging. Image processing The images from the 2P FCM show a honeycomb pixelation patter n due to the packing of the cores of the CIFB. Several methods have been described to de pixelate images from C I FBs (Lee and Han, 2013a, 2013b; Shinde and Matham, 2014) . The simplest

PAGE 96

79 methods involve lowpass filtering with either a blurring function (Gbel et al., 2004) or masking the image in the frequency domain(Winter et al., 2006). However, two-photon imaging through a CIFB has the additional complication of amplifying the non-uniformity of the fiber cores. Each core is assumed to have a unique sensitivity, due to the variability in diameter, shape, NA, and amount of cladding between adjacent cores. This manifests as discrete variations in image intensity across the F OV. This was addressed by programmatically dividing out the sensitivity of each core and interpolating the core values to remove the pixelation pattern. This process is described in detail in the Appendix . Briefly, I used a flat map of the full field CI FB fiber-cores by imaging a fluorescent test slide (Chroma Technologies) with the 2PFCM , with an example shown in Figure 5.6 (a) . The flat map stores the centroid coordinates of the cores and their corresponding sensitivity. The processing was performed with custom software (Matlab, Mathworks). Each image to be ana lyzed was registered to the flat -map, which Figure 5.5: Mouse attachment. (a) The CIFB is attached to a coupling objective on the proximal end and the 2P -FCM on the distal end. (a) 2P-FCM is attached to the permanent baseplate on the mouse wit h a quarter -turn.

PAGE 97

80 allowed identification of the cores. An example of a raw pixelated image is shown in Figure 5.6 (b) . The re lative sensitivity of each core was compensated by dividing by the flat map values. The honeycomb pattern was eliminated by using the nearest neighbor interpolation method (Rupp et al., 2009) . A SavitskyGolay filter was used to redu ce the added single-pixel noise introduced by the core multiplication factor during flatnormalization. During the interpolation, the fibercores were registered to a square pixel grid for straightforward analysis. An example output frame is shown in Figure 5.6 (c) . For processing of temporal scans, each frame was processed with the same flat map alignment so that the cores are static in the field. Once the honeycomb pattern and CI FB induced intensity variation was removed, a clustering algorithm was used to identify regions of interest (ROIs ) of highcorrelation (Ozden et al., 2008). Significant changes in cytosolic Ca2+ were identified as changes in fluorescence larger than 3 standard deviations above baseline within each ROI. Figure 5.6: Image processing of twophoton imaging through CIFB. (a) Flat -field map showing enhanced non -uniformity due to heterogeneity in CIFB. (b) Unprocessed image of cells in mouse cortex with fiber -pixelation. (c) Post-processed image after fiber-cores were corrected with flat field mask and re -gridded into a typical square pattern. Grid lines added to emphasize pixels.

PAGE 98

81 5.3 Results Resolution, magnification, and scan range I characterized the lateral and axial resolution and the axial focusing range of the 2P FCM by imaging 2m diameter yellow green fluorescent micro beads embedd ed in clear agarose. The lateral resolution is fundamentally limited by the average spacing of the fiber cores in the CIFB. Using a calibrated objective lens, I measured the inter core spacing to be ~4.5m, which agrees with previously published results (Chen et al., 2008) . With the 2P FCM magnification factor of 0.4x, the theoretical lateral sampling at the object is ~1.8m . The beads were imaged with the 2P FCM at sequential focal planes by tuning the EWTL focus in discrete steps to obtain a Z stack. The images were processed to remove th e fiber pixelation pattern as described in the methods section. Figure 5.7 compares processed images of the beads imaged with a 20x 0.75 NA objective and the 2P FCM. The average lateral and axial line profiles of 5 beads measured from different focus positions were fit to a Gaussian function. The axial bead size measured by the 20x objective was 4.5m FWHM and with the 2P FCM it was 9.9 m FWHM. The la teral bead size measured by the 20x objective to be 1.7m FWHM (dotted grey line) and with the 2P FCM it was 2.6m. The lateral bead size is larger than diffraction limit as it is limited by the fiber bundle spacing. With a bead size of ~2m, nonunifor m sampling of the bead with

PAGE 99

82 multiple fiber cores causes a larger effective lateral profile. The axial profile measurements of both the 20x objective at 0.75 NA and the 2PFCM at 0.45 NA are similar to what is expected from the diffraction limited calculations. Figure 5.8 (a) shows side-projections of beads as imaged by a 20x 0.75 NA objective and 2P-FCM overlaid in green and red, respectively. The same region of the agarosebead sample was imaged in both case s, such that most of the same beads appear in both Zstacks. This made it possible to compare directly their apparent size and axial location to determine the 2PFCM scan range. The predicated axial focus plane through the EWTL focusing range from Zemax and actual bead positions are shown in Figure 5.8 (b) (grey line and black dots, respectively). The full focusing range did not span the range predicted (240 m predicted vs. 180m measured), likely due to under-performance by the EWTL at the high -end of the optical power range. I was able to rule out optical alignment issues by tolerancing analysis in Zemax, which indicated that any physical alignment er ror or static lens discrepancies could not account for the change in focus range. The focal range of the EWTL was estimated by measuring the distance from the lens to Figure 5.7: Axial and lateral resolution tested by imaging fluorescent beads with a 20x Olympus objective (dashed lines) or with the 2P -FCM (solid lines).

PAGE 100

83 the focus spot when imaging a distant light source. The optical power at the highest voltage setting (60V) was measured to be ~ 30 -m-1, lower than the specified 36 -m-1. The optical magnification from the C IFB to the target was evaluated by imaging the group 6, element 2 square on the USAF 1951 resolution target with the 2PFCM as shown in Figure 5.9. The imaging diameter of the CI FB was measured to be 550 m . The magnification at three different optical power settings for the EWTL was measured by comparing the scaled size of the resolution target through the CIFB with the actual size. The results are summarized in Table 5.2. The magnification is measured to be ~0.4x, varying by less than 5% through the focusing range. This agrees closely with the predictions from the Zemax model. Figure 5.8: Testing of axial scan range. (a). Side-projection of ~2m diameter fluorescent beads suspended in clear agarose and imaged with a 20x 0.8 NA Olympus objective (green) using 910 nm excitation light and a motorized stage or with the 2P -FCM while varying the EWTL power (red). (b) Predicted scan range as the EWTL optical power is changed modeled in Zemax (grey line) and Z -positions of measured beads (black circles).

PAGE 101

84 Table 5.2: Measured magnification variation through focus EWTL control (VRMS) Predicted magnification Measured magnification 60 0.41 0.405 42 0.40 0.393 25 0.39 0.388 Ultrafast laser pulse propagation through fiber bundle Chromatic dispersion through the glass of the individual fiber of the CIFB results in broadening of the pulse-width which can reduce the fluorescent signal from two-photon microscopy by decreasing peak pulse power (Lefort et al., 2011) . Further, temporally focused laser pulses in glass can result in unwanted nonlinear effects, such as two -photon Figure 5.9: Measurements of magnification of 2P -FCM. Elements of a USAF 1951 resolution target were imaged through the CIFB at three different focus settings. The known size of the elements is known precisely is indicated.

PAGE 102

85 autofluorescence from the gla ss surface and self phase modulation (Planas et al., 1993) . Although compensation has been applied in CIFBs, these effects have not been explored in detail for CIFBs (Gbel et al., 2004; Lelek et al., 2007) . The nonuniform cross sectional core spacing, size, and shape warrant an investigation into the feasibility of propagating ultrashort pulses in a CIFB for twophoton imaging. In single core optical fibers, the unwanted effects can be reduced by pre chirping the pulse before it enters the fiber (Agrawal, 2000) . The setup for pre compensation is shown in the left side of Fi gure 5.1. A single mode PM fiber was used to broaden the spectrum by approximately twotimes to compensate for the spectral narrowing which would otherwise occur at highpower in the CIFB (Lefort et al., 2011; Lelek et al., 2007) . A grating pair with a retro reflector was used to apply approximately 40,000 fs2 of negative GDD to the beam before coupling it into the fibercores. Frequencyresolved optical gating ( FROG) was used to measure the temporal and spectral properties of the pulses (Trebino e t al., 1997) . The FROG technique recovers both the pulse amplitude and phase, which allows the direct examination of effects such as temporal and frequencychirp, spectral broadening or narrowing, and higher order dispersive effects. The ultrashort pul sed beam was measured both after the laser and after propagating through the CIFB and the results are shown in Figure 5.10. The power output of the CI FB was measured to be 30 milliwatts. Overall, the second order dispersion compensation by the gratingpair is sufficient to compensate for the pulse well at the power levels used, with the pulse width only encountering a broadening from 65to 81fs. The r esults of measurements at various points along the laser path are shown in Table 5.3. Without the pulse pre compensation, the

PAGE 103

86 pulse width was measured to be ~600fs after the CIFB, resulting in a large loss of peak pulse intensity. The small increase in pulse width in the current setup is likely attributable to third -order dispersion, which may become more prominent at higher power l evels. A solution to reduce third order dispersive effects has been demonstrated with the use of grism pair pulse stretchers (Gibson et al., 2006; Kalashyan et al., 2012). Figure 5. 10: Measurement of pulses for propagation through 1.0-m long CIFB using FROG at 910 nm. Each figure: Top: Measured spectrogram. Mi ddle: Spectrum retrieval from FROG spectrogram overlaid with phase plot (orange). Separately measured spectrum is shown as well (dashed line). Bottom: Temporal pulse trace retrieved from FROG spectrogram overlaid with phase plot (orange). (a) Pulse measure d directly from ultrafast laser. (b) Pulse measured after propagating through CIFB.

PAGE 104

87 Table 5.3: FROG pulse measurements along laser beam path Measurement site Spectral width (nm) Pulse width (fs) Laser output 18 65 After PM fiber 34 305 After pulse stretcher 34 410 After CIFB 15 81 Tissue sample imaging The following experiments on large volume, tiltedfield scanning, and twocolor imaging demonstrate the potential features of the 2P FCM as an extremely versatile device. Volumetric imaging: The 3D imaging potential of the 2P FCM may not be fully demonstrated when i maging in mouse cortex through a cranial window. To demonstrate volumetric imaging in an ideal setting, a ~1 mm thick sample of fixed mouse cortical tissue was imaged with the 2P FCM . The sample was fixed with 4% paraformaldehyde and has cells that express green fluorescent protein (GFP) driven by proteolipid protein promoter. The GFP labels oligodendrocyte cell bodies in the cortex. Axial scanning was accomplished by sequentially tuning the EWTL across the full focal range in 36 discrete steps, acquiring 36 image planes to form a Z stack. The images were processed to remove the fiber core pattern as described in the Methods section. A 3D volume image was created using ImageJ/Fiji software and is shown in Figure 5.11. Over 200 cells were observed in this volume.

PAGE 105

88 Tilted field imaging: This is a simple but effective way of increasing the functionality of a system that has a rapid axial focusing mechanism. Further, because there are no moving parts there is no concern for the stability of the specimen during the scan. The tilted field scan is accomplished by driving the EWTL with a control signal that varies across the scan field. Figure 5.12 shows the results of performing this test on fixed, thick brain tissue with sparse fluorescence signal from neurons expressing GCaMP6s. 25 total angles were acquired, 12 in each direction, varying the voltage on the EWTL from 40 to 50 VRMS across the slow axis of the raster scan. The waveform was a simple sawtooth function and could likely be improved according to evidence shown in voltage waveform function shaping research for electrowetting lenses (Supekar et al., 2017a). Multi color imaging: Separating fluorescence colors is difficult in many miniaturized microscopes because the detection path is exclusive and often also needs to b e miniaturized. This can be an important feature for researchers who want to investigate the co localization or relative abundance of certain physiological markers. The 2PFCM offers an Figure 5. 11: Fixed tissue with GFP-labeled oligodendrocytes imaged with 2P FCM. (a) 3D volume acquired by the 2P -FCM (220 x 220 m lateral x 180 m axial) with over 200 cells in the image. (a). Processed image of a single slice in the stack after filtering to remove p ixelation pattern.

PAGE 106

89 advantage in that two-photon excitation can often overlap a large amount with dyes that have separated emission spectra. A common example is GFP and tdTomato, which are both efficiently excited at approximately 900 nm with pulsed light, but have spectrally separate emission bands. Another feature of a fiber coupled system is that the emission signal can be spectrally de-mixed after returning through the fiber-bundle. This means that the same filters and detectors used for normal two -color microscopy can be re-used in place during 2P-FCM imaging. Figure 5. 12: Tilted -field imaging enabled by rapid focusing of the EWTL. (a) Maximum intensity projection of fixed mouse brain tissue expressing GCaMP6s in neurons, acquired by the 2P FCM. Arrows indicate cell bodies retained in fields. (b) Side projection of the volume. T he same cell bodies are indicated by the arrows. The planes for the horizontal and angled scan limits are indicated (c) Images showing the largest tilted -field scan acquired in this experiment, indicating the same cell bodies that are shown to intersect wi th the red or white planes in (b).

PAGE 107

90 In Figure 5.13, I show two-color imaging with 900 nm two-photon excitation of tdTomato and GFP simultaneously with both a 20x Olympus objective and the 2PFCM. The cells are GFP -expressing mature oligodendr ocytes and tdTomato expressing oligodendrocyte lineage cells and sparsely labeled astrocytes (MobpEGFP; Olig2 Cre; R26 lsl tdTomato triple transgenic mice). The two experiments were performed using the same detectors and filters. This provides an example of how the 2PFCM can act as a swappable objective lens for mouse imaging, without needing to change the filters or detection path to acquire spectrally similar images. Each mouse for 2P-FCM imaging has a 3Dprinted baseplate permanently attached as descri bed in the methods. A photo of a baseplate attached to a mouse with a cranial window is shown in Figure 5.14 (a) .For each imaging session, the 2PFCM is attached to the baseplate with little force on the skull, as shown in Figure 5.5 . Figure 5.14 (b) shows a Figure 5. 13: Multi-color imaging. (a) Two-color maximum intensity projection acquired with a 20x 0.75 NA Objective A. Maximum i ntensity projection of a region of brain tissue acquired with a 20x 0.75 NA objective. Yellow cells are oligodendrocytes (Green and Red), while red -only cells are astrocytes and oligodentrocytes cells. Two astrocytes are marked with arrows, and are easily identified by the characteristic bushy morphology and two are marked with arrows. (b) Same tissue imaged with the 2P -FCM, using the same detectors, filters, and excitation wavelength. Green and red cells are visible in the field, with likely astrocytes mar ked by arrows.

PAGE 108

91 photo of a mouse with the 2PFCM attached. The C I FB and electrode wire for the E WT L are draped passively over a horizontal post to reduce the weight on the mouse. When attached, the mouse is able to move around freely in a small area (about 12” square). The movement is somewhat restricted by the length of the CIFB (1.0 m) and torsional resistance of the C IFB, which prevents rotation of more than about 180. It was found that, after short accli mation (~30 minutes), mice could traverse the entire behavioral area and habituated to the restrictions . Future implementation may include a commutator with rotation encoder for realignment, which has been shown previously (Flusberg et al., 2008a). Figure 5.16 shows a wide field epifluorescence image of the region for an implanted mouse through the 2PFCM. The left image shows the background fluorescence and vasculature on the day of implantation. T he right image shows that the baseplate is stable over 17 days, showing only a slight lateral shift in alignment. Figure 5. 14: Mouse 2P -FCM attachment photos. (a) Baseplate implanted on mouse with cranial window. (b) Mouse behaving with 2P -FCM attached.

PAGE 109

92 The expression level in the neurons virally transfected with GCaMP6s in the cortex was verified by histology. Figure 5.15 shows a 20m -thick confocal stack of a coronal section of a perfused mouse, 12 weeks after injection. Neuronal cell bodies are seen in layers 4/5, as well as layers 2/3 in lower density, which is where I performed imaging with the 2PFCM. Figure 5. 16: Implant stability over 17 days imaged with widefield epi -fluorescence. Figure 5. 15: Histological coronal section of mouse injected with GCaMP6s virus, showing g ood expression in layers 2 -5 of motor cortex.

PAGE 110

93 In vivo mouse imaging during behavior I performed in vivo two photon imaging of neuronal activity in an awake and mobile mouse expressing GCaMP6s in cortical neurons. The mouse was allowed to wander freely i n a 7” by 11” plastic cage. The cage was filled with sawdust as well as food and various novel objects, such as raisins and tissue paper, to motivate the mouse to explore the environment while attached to the 2P FCM. F igure 5.17 (a) is a 3D projection of a Z stack taken by tuning the E WT L focus, showing the imaging volume acquired. The bright cells were located at the depth range between 80 and 160m , likely corresponding to cortical layers 2/3 because of the expected imaging depth of the 2P FCM . Figure 5.17 (b) is a side projection of the same volume, where distinct cells are in different focal planes. Time courses were acquired at various focal planes, with three planes shown in Figure 5.18 (a) corresponding to the depths annotated in Figure 5.17. Distinct static cell populations are visible (green) in each of the three panels. Temporal scans were performed at each of these three planes. The movies were acquired with a FOV of 200x200m at 1.5 Hz or 80x80m at 4 Hz. In some of these scans, Ca2+dependent bursts of varying height and duration were recorded a nd are represented as fluorescence traces in Figure 5.18 (b) . Each spike is indicated by a colored arrow, which corresponds to change in average fluor escence bounded by the colored regions in Figure 5.18 (a) . The regions are identified algorithmically as clusters of cores that have high crosscorrelation values with their neighbors over the course of the temporal scan.

PAGE 111

94 The mouse was recorded with a camera during the imaging session to correlate the imaging results to the motion of the mouse. The cranial window was in a region of the somatosensory cortex, but none of the large spikes were directly correlated with overt behavior. Likely a more targeted behavioral paradigm is needed to extract behavioral correlations in these shallower cortical layers. Some motion artifacts occurred during image and caused a shifting of the image by not more than one fibercore’s distance (< 2 m) The motion artifacts were greatest during specific behaviors, especially those that involved rapid, short movements such as grooming or eating. Motion artifacts were rarely seen while the mouse traversed the cage and were uncommon in general. Motion did not result in permanent lateral shifts or changes in the focal plane. The motion artifacts were much less than that reported in head-fixed two-photon imaging studies (Dombeck and Tan k, 2014; Dombeck et al., 2007) and similar to the artifacts seen in wide -field imaging, which benefits from a much larger DOF (Ghosh et al., 2011). The 2P-FCM did not become loose or dislodged during the imaging sessions, which lasted between 1 to 3 hours. Figure 5. 17: 3D imaging of cells labeled with GCaMP6s in behaving mouse. (a) Volume view of cells (b) Side -projection of cells.

PAGE 112

95 5.4 Summary In this chapter I described t he first head -mounted 2PFCM that achieves 3D imaging of neural activity in a freely -behaving mouse. The imaging volume is 220m diameter by 180m depth. The device is optimized for resolving neuronal somata with a lateral resolu tion of 2.6m and axial resolution of 9 m . The 2PFCM recorded neuronal activity in different planes of cortical layer 2/3 of a freely moving mouse with minimal motion artifacts. Th e 2PFCM differs from previous headmounted microscopes (Flusberg et al., 2008a; Ghosh et al., 2011; Helmchen et al., 2001, 2013; Sawinski et al., 2009; Ziv et al., 2013) because it allows li ve-focusing for a range of ~200m depth suitable for imaging neurons in layer 2/3 of cortex. Figure 5. 18: Awake-behaving mouse Ca2+ imaging with GCaMP6s. (a) Three different focal planes showing distinct cellular populations. The colors and regions represent the peak fluorescent spatial intensity from the corresponding transients. Scale bar 25 m. (b) Traces showing change in GCaMP6s fluorescence for the 5 regions of distinct activity indicated in panel (a).

PAGE 113

96 CONCLUSIONS AND OUTLOOK The focus of this dissertation has been the optical, mechanical, and surgical design of versatile mouse neuronal imaging systems. The goal has been to create fiber coupled microscopes (FCMs) that have the capability to compete with head -fixed imaging setups for awake and behaving mouse brain imaging. There have been two key technologies that have driven the progress of this work. First, the coherent imaging fiberbundle (CIFB), which is a tool that has large adoption in the field of clinical endoscopy, but has been displaced in the past decade by more advanced scanning technologies in the neuroscience field. In the experiments I show here I make the case for its use as an easily adoptable passive imaging device that functions remarkably well for both singleand multi -photon imaging. Sec ond, the electrowetting tunable lens (EWTL), which is also not commonly adopted by basic researchers. The main reason for this is the small aperture size of electrowetting lenses, which limits them from many table -top microscopy applications. However, because of its lightweight and mechanical simplicity, it is an ideal tool for miniaturized applications. Laser -scanning confocal and two-photon microscopy (LSCM and 2PLSM) were enabled in 3D with the combination of the CIFB and EWTL. Using these technologies , I was able to show imaging of a variety of targets in 3D, including an awake-behaving mouse brain. However, there are obviously still limitations in the demonstrated devices and the map forward must be planned out.

PAGE 114

97 6.1 Revisiting the confocal FCM The confoca l FCM (CFCM) and two-photon FCM (2PFCM) were developed and tested in that order. Since adopting two-photon excitation microscopy as the modality of choice for the FCM designs, no additional dedicated CFCMs have been designed. As described in C hapter 3, a critical component of the C-FCM is the ability to focus both the excitation and the emission light achromatically, because chromatic aberrations w ould result in a blurring of the emission fluorescence on the fiber-bundle resulting in lessefficient filtering of out-of-focus light. The 2P-FCM designs abandon this optimization because of the greater difficulty in compensating for chromatic aberrations with such a large difference in excitation and emission colors. Also, removing this restriction allow s more freedom in the optical design to optimize the system for efficient light collection on multiple fibers. There was concern that coupling emission light on large areas would reduce collection efficiency. Although there is a benefit in lightcollection efficiency to refocusing the emission on a single fibercore, an experiment of illuminating progressively larger spots showed that as soon as the single-core coupling in lost the size of the illumination spot makes no difference, summarized in Figure 6.1. Figure 6.1: Coupling of green emission light into CIFB is indifferent to spot size unless coupling is in single -core.

PAGE 115

98 Because of this realization , the 2P FCM was not designed to be useful for confocal microscopy but just to collect as much light onto the fiber surface as possible. Simulations in Zemax of point source emission light traveling back through the FCM would focus onto spots as large as 20 – 30m in diameter on the CIFB, well outside the ideal confocal pinhole size. Unexpect edly, the 2P FCM managed to demonstrate a role for the confocal in the future. Because of the additional complexity of setting up the system for two photon excitation, it is frequently convenient to use the 2P FCM on a single photon LSCM system for remote imaging experiments, especially on samples that do not have to be imaged through heavily scattering tissue. Imaging peripheral nerves and ex vivo tissues with the 2P FCM using confocal microscopy has yielde d remarkably highquality images. Also, some optic al sectioning is maintained even with the larger pinhole, because it reduces overall background light. These results have encouraged a revisiting of the C FCM. Although the LSCM modality does not have the ability to see as far into opaque tissues, compared to the two photon imaging, confocal imaging through the CIFB has the following advantages: Less heterogeneity of fiber cores in single photon imaging No detectable core to core coupling effects Readily attachable to LSCM systems for imaging at many facili ties without modifications to excitation source The C FCM may also be useful for implanted GRIN lens imaging, where the cells of interest may be located much closer to the optics and so there would be less tissue to image through. I nitial experiments of imaging ex vivo fluorescently labeled tissues with the

PAGE 116

99 2P-FCM through a 4.1mm-long 0.8 NA GRIN lens (GRINTech) in confocal imaging mode have been successful, shown in Figure 6.2. The higher magnification of the GRIN lens results in better resolved cellprocesses of the GFP -labeled oligodendrocytes and the side-projections show some optical sectioning capability. Widefield imaging has already been shown t o be effective at visualizing cells through an implanted GRIN lens in the mouse brain (Barretto and Schnitzer, 2012a; Lee and Yun, 2011). Updated FCM designs for efficient coupling into longer GRIN lens relays may allow for deeper brain access. These revelations on the usefulness of the C -FCM have been encouraging enough to warrant revisiting the optical design. The C -FCM could become a useful tool for peripheral nerve imaging. Also, i t will be worth exploring the use of a redesigned CFCM for Figure 6.2: Conceptual experiments for confocal imaging with FCM and GRIN lens. (a) 2P FCM coupled to a 0.8 NA GRIN lens assembly to image GFP -labeled oligodendrocytes in fixed mouse brain tissue. (b) Maximum intensity projection of lateral and side -views after processing to remove fiber -pixelation (scale bar is 30 -m).

PAGE 117

100 imaging with implanted GRIN lenses. Minimally, the 2P FCM design has been shown to be capable of some confocal imaging which is a useful modality to have available. 6.2 Moving forward with the two photon FC M Head fixed 2P LSM has been a powerful technique because of the ability to image cortical neuronal and dendritic activity in awake animals (Peron et al., 2015b) . Importantly, recent advances have allowed mesoscopic measurements in large numbers of neurons (Peron et al., 2015a; Sofroniew et al., 2016) . As these tools become more ubiquitous, researchers are demanding tools for imaging animals i n natural behavioral contexts. Some aspects of free movement can be simulated by placing the animal in a virtual reality situation (Rickgauer et al., 2014) . However, restriction of head movement limits neuronal imaging to a subset of behaviors. The 2P FCM can be a useful tool to study neural activity during behaviors that involve the movement of the animal such as social behaviors, prey capture, behaviors involving vestibular stimulation, odorant plume tracking, and interaction with physical objects. It is im portant for these experiments is an imaging system that is resistant to motion and vibration. The 2P FCM showed minimal motion artifacts , even with the shallow DOF provided by twophoton imaging. This can be owed to a rigid and compact design with no mecha nically moving parts and that t he EWTL is insensitive to motion and vibration (Hendriks et al., 2005) . This 2P FCM design also provides a great degree of flexibility for the user. The CIFB permits the selection of arbitrary ROIs within the FOV, which can be widened for slower and larger region scanning, or tightened down for faster and smaller r egions when targeting small clusters or single cells. The active focusing afforded by the EWTL allows live

PAGE 118

101 readjustment of the focal plane. This is essential for focal alignment with the shallow DOF , and allows for easy imaging from multiple focal planes. Compared with other methods of axial focusing in miniature microscopes, the EWTL is simple in implementation, small and low w eight, and immune to orientation and motion. The EWTL is also able to operate at high speeds (Up to 50 Hz (Burger et al., 2015; Murade et al., 2012) . In this w ork I demonstrate only slow imaging, because the current implementation relies on acquiring data while the focus is following a linear trajectory. It is more ideal to drive the lens closer to resonance, similar to resonant laser scanning setups (Hoover and Squier, 2014) . F urther development will enable more complex 3D scans, such as random access scanning for truly volumetric data collection (Gbel et al., 2007; Nadella et al., 2016; Szalay et al., 2016) . There are still h urdles to overcome for the 2P FCM to compete with head fixed 2P LSM directly. The 3D imaging volume (a cylinder of 240m diameter x 180m depth) is smaller than the volume imaged by head fixed 2P LSM. In addition, the lateral resolution is limited by the fiber bundle spacing and is worse than 2P LSM with comparable NA objectives. The CIFB limits mouse movement and rotation compared with other miniature fiber coupled microscopes for twophoton imaging (Ducourthial et al., 2016) . Nonetheless, even with these limitations it is possible to record cellular activity with smaller motion artifacts than head fixed systems and from different focal planes with no moving parts. This 2P FCM may be a solution to brain imaging of freely moving mice, but will need to expand its capabilities to become competitive. In the future the 2P FCM can be paired with GRIN lenses to image deep brain regions (Barretto and Schnitzer, 2012b) .

PAGE 119

102 6.3 Fiber bundles: Simple, versatile, needs improvement In many ways, the CIFB is an incredibly versatile tool that also greatly simplifies the laser scanning setup. By using the CIFB there is no need for miniaturized scanning optics or a dedicated detection path for fluorescence emission. It is also a multi modality tool that can be used with both confocal and twophoton modalities. In addition, both FCM designs allowed for widefield epi fluorescence imaging, which was often used with the microscope oculars to locate regions to image . Because the imaging system is ignorant of the source of the light, it could be used for arbitrary region imaging and similarly for spatially patterned or targe ted stimulation, as previously shown in static focus FCM (Szabo et al., 2014) . This kind of versatility is the reason that the CIFB is still relevant in the discussion of mouse attached microscopes. T ethering to the CI FB restricts movement of the mouse. Although mice acclimated quickly to the restriction s and were able to navigate the whole cage during the experiments, this is a limitation that can be fixed in future versions. For example incorporation of a commutator w ill allow turning of the animal (Flusberg et al., 2008a) , but will have to be controllable so that the resulting image is not blurred due to rapid rotation. L eached fiber bundles are another potential solution to the inflexibility of the CIFB demonstrated here. In leached bundles, the fibers are fabricated with sacrificial acid soluble glass between individual fibers (Kostuk and Carriere, 2000) . This inter fiber glass is dissolved after the fiberbundle is made so that only the ends are fused, and the rest of the fiber is extremely flexible. The disadvantage of the leached fiber bundle is that none are currently available with single mode core sizes. This limits their usefulness for techniques

PAGE 120

103 such as two photon microscopy, in which modal dispersion can severely reduce fluorescence excitation efficiency (Kim et al., 2008) . The CIFB also has only a ~50% effective imaging area. As the laser is scanned across the bundle surface, it is only coupling into a fiber core about half of the dwell time, with the other half spent on the cladding between the cores. Additionally, the solid glass core fibers require more compensation than one would need with a hollow core fiber, in which the electric field exists primarily in a gasfilled waveguide, reducing the effects from propagation in glass (Zong et al., 2017) . To increase signal lev els with the CIFB , better pulse compensation systems can be implemented such that higher pulse power can be transmitted through the CIFB without higher order dispersive effects resulting in diminishing returns (Gibson et al., 2006; Lefort et al., 2014) . It may be worthwhile to consider other options for developing the lateral imaging in future FCM designs. The EWTL focusing implementations are almost completely independent of the chosen lateral scanning mechanism in the FCM. Piezoactuated fiber scanners and MEMS actuated micro mirrors are promising options (Qiu and Piyawattanamatha, 2017 ) . Recently, we showed that electrowetting prism devices can be used for lateral scanning in benchtop microscopes, and may be implemented in the FCM in the near future (Supekar et al., 2017b) . Even small amounts of lateral displacement can reduce the number of cores required to achieve similar fields of view by laterally scanning the entire fiber bundle across the sample. The issue of poor NIR light transmission, especially at large bandwidths, in CIFBs is well documented (Chen et al., 2008) . This is understood to be a result of core to core coupling due to t he close inter core spacing. In my experiments imaging the transmitted

PAGE 121

104 light at the distal end of the CIFB I was able to see that the power coupling into cores other than the primary. Further these fibers appear to carry only higher-order modes. This can b e seen in Figure 6.3. These higher order modes would be ineffective for two-photon excitation, so they would not likely add unwanted noise to the signal. However, a l arge amount of power may be lost into these cores that would not be useful for imaging. However, there are solutions that may allow CIFBs to eliminate core to core coupling while simultaneously decreasing inter-core spacing, making the resolution higher. Inhomogeneity of core sizes and shapes has been shown to reduces coreto -core coupling, and groups making CIFBs with intentionally heterogeneous core distributions have shown some success (Koshiba et al., 2009; Stone et al., 2017; Wood et al., 2017). Regardless of the solution for lateral imaging, the CIFB continues to be a useful tool for designing and testing the 3D scanning FCMs as we investigate the correct applications for them. Figure 6.3: Core -to -core coupling imaged at distal end of fiber while single -core is illuminated by 910 nm light at proxima l end. (a) Raw camera image of fibers showing higher order modes coupling into nearby fibers. (b) Same image with coupled cores circled.

PAGE 122

105 6.4 Finding a need for the 3D imaging FCM As discussed above, it is clearly useful to neuroscientists to have a two photon imaging system 3D imaging microscope that also allows for free movement and behavior of the animal. Compared to other readily available widefield miniature microscopes, the 2P FCM allows for higher contrast imaging in brain tissues. Widefield systems often rely on fluorescence changes to inform the location of the cells, because static cell signals are hardly distinguishable from the high background fluorescence (Ghosh et al., 2011) . The optical sectioning capabilities of the FCMs allow for morphological imaging in the absence of activity. However, the limitations of the FCM approaches mean t hat the usefulness of the FCM system for a particular problem must outweigh the costs in adopting, implementing, and using the technology. In addition to the disadvantages outlined above, transplanting the 2P FCM imaging system to a remote facility requires additional modifications. For example, the pre compensation system for the excitation source is obstructive in its complexity. The 2P LSM requires both compensation for temporal dispersion as well as non linear spectral narrowing. These are accomplished by inserting a gratingpair pulse stretcher and a singlemode fiber into the optical path (Lefort et al., 2014) . This requires vigilant alignment and reduces overall available excitation power for imaging . This will be a challenge to implement for imaging on different microscopes and laser sources. Hence, the application that requires the 3D imaging capabilities of the 2P FCM must outweigh these costs. Speculatively, applications uniquely suited for the FCM include: Tracking of cell migration in 3D over time courses of hours. It has been shown that cells in the central nervous system involved in homeostatic maint enance may

PAGE 123

106 migrate on the timescales of hours or days (Hughes et al., 2013a) . Such longterm are difficult to track in head fixed animals and it is difficult to study behavioral influences on those changes. Investiga ting the 3D dynamics of spatially complex structures such as the hippocampus. Current headfixed setups have shown that multi focal imaging of neurons in the hippocampus with thin optical sectioning can reveal dynamics such as the turnover of dendritic spi nes in short time frames (Attardo et al., 2015) . Widefield imaging cannot resolve such structural detai ls, so current miniature microscopes do not provide a way to visualize events that require high contrast imaging in these tissues. Further exploration of options for awake behaving 3D imaging will help inform the future path of the 2P FCM for neuroscience research . The work presented here is only a first step in an unknown trajectory that will be influenced by good communication with researchers in the neuroscience field. In conclusion, this work is a demonstration of two miniaturized 3D neuronal imaging sy stems using confocal and twophoton imaging modalities to achieve optical sectioning. Axial focusing is achieved by integrating a lightweight and motionresistant electrowetting tunable lens into the miniature imaging system. An imaging fiber bundle is use d for lateral laserscanning and for fluorescence emission collection. The efficacy of this combination is demonstration by quantitative characterization and imaging of Ca2+dependent fluorescence changes in the brain of a freely moving mouse. The results shown here may be important in the advancement of miniaturized imaging for studying the mammalian brain in vivo .

PAGE 124

107 APPENDIX F IBERBUNDLE DEPIXELATION METHOD In Chapter 4 , I described the use of a spatial frequency filter to remove the pixelation of the coherent imaging fiber-bundle (CIFB). In Chapter 5 I discussed the use of an interpolation based depixelation method to improve contrast and reduce the fibercore heterogeneity artifact. In this appendix I will detail these de -pixelation methods. CIFB de pixelation background and techniques CIFBs are frequently used in clinical endoscopic imaging applications for surgical guidance and optical biopsies (Hughes et al., 2013b; Jabbour et al., 2012) . There are various fibercore geometries, but the most space efficient packing of circular cores results in the characteristic “honeycomb” pattern, shown in Figure 2.13 and replicated below. Figure A. 1 (2.13) : Fujikura CIFB fluorescence image of a uniform target showing core distribution.

PAGE 125

108 This pixelation can be distracting for clinicians and reduces effective imaging contrast (Han and Yoon, 2011; Han et al., 2010; Liu et al., 2011). For the purposes of fluorescence neuronal imaging the honeycomb artifact may also confuse image-processing algorithms that re ly on edge-finding and uniform spatial sampling, such as algorithms for motion correction and temporal crosscorrelation. There are many techniques that have been reported for reducing this pixelation (Shinde and Matham, 2014) . The simplest algorithms involve basic spatial filtering, such as the use of a spatial Gaussian filter or a spatial frequency filtering technique to eliminate the spatial frequencies associated with the cores (Gbel et al., 2004; Winter et al., 2006) . The effectiveness of spatial filtering techniques is reduced if the cores are not in an entirely predictable structure. More advanced iterative and learning algorithms have also been described (Han and Yoon, 2011; Lee and Han, 2013b; Liu et al., 2016). Also shown are spatial interpolation algorithms, in which each fiber-core and the signal riding on it is mapped to a spatial location and the gaps are filled by some interpolation method (Rupp et al., 2009). Enhanced core heterogeneity under two -photon excitation In the specific case of two -photon excitation microscopy through the CIFBs presented in this thesis (Fujikura, FIGH -N type), there is a distinct heterogeneity in the fiber honeycomb pixelation pattern. This is likely because twophoton fluorescence excitat ion depends on the arrival of a pulse’s wavefront within a short time window (~10 fs), which could be heavily disrupted by modal dispersion and coreto -core coupling (Chen et al., 2008; Warren et al., 2016) . The same effect would instead sli ghtly lower the imaging resolution in one -photon fluorescence microscopy. CIFBs similar to the ones used here have

PAGE 126

109 been shown to exhibit inconsistent modal propagation characteristics (Han and Kang, 2012). Figure A. 2 below is an example of a 0.55mm diameter CIFB imaged with a 920 nm ultrashort pulse (~ 100 fs) laser (Spectraphysics MaiTai), with distal optics to image either a NIR phosphorescent card (Newport) to demonstrate onephoton excitation efficiency or a unif orm green fluorescent slide (Chroma) to demonstrate two -photon efficiency. Under two -photon excitation, the cores have an unpredictable sensitivity to uniform fluorescence, which will inaccurately represent the image of a sample. Further, lower sensitivity cores will have greatly reduced signal to noise because the lower core sensitivity will be applied multiplicatively to the signal but not to various noise sources intrinsic to the imaging system. Therefore, two-photon imaging through the CIFB demands additional processing to minimize this artifact. Figure A. 2 : Fujikura CIFB used for imaging of a NIR phosphorescent detector card to demonstrate one -photon excitation and a uniform green fluorescent test slide to demonstrate t wo photon excitation.

PAGE 127

110 Hybrid interpolation and noise damping algorithm for CIFB de pixelation The goals of the method I describe here are to: 1. Eliminate the pixelation pattern of the cores to improve the overall contrast. 2. Maintain spatial coherence such that objects the FOV are not distorted. This is important for applying motion correction and crosscorrelation algorithms to time traces. 3. Reduce the heterogeneity of fiber-core sensitivity under two-photon excitation without simultaneously multiplying or s preading the noise functions. Regarding point 2: A primary concern when imaging with a CIFB is that a small object may only be sampled by a few cores at a time. In a timeseries, small motion artifacts may result in the appearance of the object changing si ze or intensity when it is really just translating underneath the sampling pattern. Techniques that simply perform spatial filtering the pixelation or that use a priori learning assumptions have a lower chance of discriminating these kinds of motion artifa cts from a change in size or shape. An example of this artifact because of some minor displacement is illustrated in Figure A. 3. The displaced heart on the right appears thinner because of the sparser sampling of the edges when shifted relative to the cores. Interpolation based methods take the spacing of the cores into account, and so are more effective at accounting for spatial shifts.

PAGE 128

111 R egarding point 3: The fibercores under two -photon excitation have highly variable sensitivities. Here I define core sensitivity by a scalar multiplier . The time -dependent intensity from each fiber core in the CIFB is expressed as: þØ5ÜPØ5ÜP þØ5ÜPØ5ÜPØ5ÜPØ5ÜP ( ) = þØ5ÞýØ5Þý þØ5Ü`Ø5Ü` ( ) + þØ5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[ ( ) w h ere þØ5Ü`Ø5Ü` is the fluorescence signal that is illuminating the core on the distal end and þØ5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[ is the noise originating from sources other than the sample, such as shot noise, background emission from the fiber, and light leakage to the detectors. As decreases, the ratio of the signal intensity to the noise also drops. Attempting to compensate for this by dividing the core intensity by the sensitivity factor will disproportionately increase the Figure A. 3: Demonstration of object shape changing artifacts resulting from different sampling of same shape by fiber cores and simple spatial filtering algorithms.

PAGE 129

112 noise level, resulting in a localized fiber core with very high noise. A workable solution is to temporally filter the core signal relative to the inverse of the sensitivity before multiplicatively normalized for the overall signal level: þØ5ÜPØ5ÜP þØ5ÜPØ5ÜPØ5ÜPØ5ÜP , þØ5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[ ( ) = þØ5Ü9Ø5Ü9Ø5Ü9Ø5Ü9 þØ5ÞýØ5Þý þØ5Ü`Ø5Ü` ( ) + þØ5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[Ø5Ü[ ( ) , This will reduce the temporal sensitivity of the core, but will prevent large localized sources of noise in the image while correcting for the spatial artifacts caused by heterogeneous core sensitivities. Algorithm This code is nicknamed AMBER for Artefactual Multiphoton Bundle Effect Removal. The algorithm is diagrammed in Figure A. 4. The user is expected to input the original raw fiber image stack, which is typically a time series but may also be a zstack or single image, a flat full -fiber image, which is an image taken during the imaging session of a uniform fluorescence source, and optionally a background image, which is an image of the fiber without any fluorescence source. The files are expected to be accompanied by parameters that inform the pixel resolution, which is used to normalize dimensions between all images. The getFiberCentroids function is used to acquire a 2D array of pixel coordinates that correspond to the locations of each identified core of the images. The flat image is used to identify the actual core locations, which are then cropped to the input image dimensions. A binary mask is made of the core locations for each image, which is used by the rigidAlignFiber function to precisely determine where in the full flat FOV the input cores

PAGE 130

113 Figure A. 4: Diagram of AMBER hybrid core interpolation and noise damping algorithm .

PAGE 131

114 are located. This is possible because of the unique spatial structure of this CIFB, which is variant in all directions. Two binary masks are translated with respect to each other, first at two times sub sampl ing to increase speed, and then at high resolution for final overlap determination. The quality of alignment is valued by multiplying the binary masks of two image. The getCentroidValues function is used to sample the values of each of the images at the sp ecified core locations. The cores are sampled by averaging the pixels in a circular area around each centroid location. The result is an array of core values corresponding to the centroids 2D array. The core values array is concatenated with the values fro m each image in the stack, which then becomes an array with dimensions of [number of cores] x [number of images in stack]. At this point the background values are subtracted from the flat and input image values, if it is available. The getCorrectorVals function uses the flat image to develop an array of 1 / values for correcting the heterogeneous sensitivity of the cores. Sometimes very low sensitivity cores may result in unreasonably high 1 / values, so the corrector values are limited to a maximum value equal to the median 1 / value plus the median absolute deviation of the 1 / values. Median absolute deviation is used so an extremely high single value doesn’t influence the limits. The corrector values are then multiplied into the input core values to increase the average value of low sensitivity cores. As discussed, this increases the noise disproportionately, so a weighted temporal filter is used to reduce the noise on the cores with very low sensitivity. A third order Savitsky Golay filter is used to minimize the effects on large transients while greatly cutting down on noise.

PAGE 132

115 The makeFiberImage function recreates the images of the fiber using the core centroid locations and the array of core values corresponding to those centroids. The user may define any structuring element for the fiber, such as circles, squares, or even 2DGaussian functions. The gridFiberCores function is used to obtain an image that is like a traditional camera image, in which the pixels are arranged in a uniform grid. This function uses interpolation based on the method defined by the input parameters, which are nearest-neighbor, natural-nearest neighbor, or linear interpolation. Nearest-neighbor interpolation set each pixel value to the value of the core closest to the right coordinates. This guarantees that none of the pixels will represent more than one core. This is useful if it is a concern that cores with high temporal noise functions may reduce the SNR of less noisy cores by averaging between them. Natural -nearest neighbor interpolation is better for most applications in that it still weights the cores closest more heavily but does better to preserve spatial information about the structures sampled by the cores. The user may specific a sampling factor, in which a factor of 1 will generate a 2D image that has approximately 1 pixel for each core, taking into account the circular shape of the CIFB if necessary. A higher value will create a larger number of pixels for each core to allow for finer interpolation. The final images may be written with the writeFiberImages function, which takes any image generated and outputs it at the set bit depth to the specified file location. Matlab code The code for AMBER, the two -photon CIFB hybrid core image processing algorithm, can be obtained at https://github.com/ozbayb/amber/ in its most current form.

PAGE 133

116 REFERENCES Agrawal, G.P. (2000). Nonlinear Fiber Optics. In Nonlinear Science at the Dawn of the 21st Century, P. Christiansen, M. Sorensen, and A. Scott, eds. (Springer), pp. 195–211. Agrawal, G.P., and Potasek, M.J. (1986). Effect of frequency chirping on the performance of optical communication systems. Opt. Lett. 11 , 318–320. Akerboom, J., Chen, T .W., Wardill, T.J., Tian, L., Marvin, J.S., Mutlu, S., Calderon, N.C., Esposti, F., Borghuis, B.G., Sun, X.R., et al. (2012). Optimization of a GCaMP Calcium Indicator for Neural Activity Imaging. J. Neurosci. 32 , 13819–13840. Andermann, M.L. (2010). Chronic cellular imaging of mouse visual cortex during operant behavior and passive viewing. Front. Cellullar Neurosci. 4, 1–16. Anikeeva, P., Andalman, A.S., Witten, I., Warden, M., Goshen, I., Grosenick, L., Gunaydin, L. a, Frank, L.M., and Deisseroth, K. (2011). Optetrode: a multichannel readout for optogenetic control in freely moving mice. Nat. Neurosci. 15 , 163–170. Attardo, A., Fitzgerald, J.E., and Schnitzer, M.J. (2015). Impermanence of dendritic spines in live adult CA1 hippocampus. Nature 523 , 592–596. Barretto, R.P.J., and Schnitzer, M.J. (2012a). In vivo optical microendoscopy for imaging cells lying deep within live tissue. Cold Spring Harb. Protoc. 7, 1029–1034. Barretto, R.P.J., and Schnitzer, M.J. (2012b). In vivo microendoscopy of the hippocampus. Cold Spring Harb. Protoc. 7, 1092–1099. Berge, B., and Peseux, J. (2000). Variable focal lens controlled by an external voltage: An application of electrowetting. Eur. Phys. J. E 3, 159–163. Blum, M., Beler, M., Grtzel, C., and Aschwanden, M. ( 2011). Compact optical design solutions using focus tunable lenses. In Proceedings of SPIE, p. 81670W. Born, M., and Wolf, E. (1999). Principles of Optics (Cambridge University Press). Botcherby, E.J., Juskaitis, R., Booth, M.J., and Wilson, T. (2007). Abe rration free optical refocusing in high numerical aperture microscopy. Opt. Lett. 32, 2007–2009. Botcherby, E.J., Jukaitis, R., Booth, M.J., and Wilson, T. (2008). An optical technique for remote focusing in microscopy. Opt. Commun. 281, 880–887. Bower, A.J., Bunch, R.M., Leisher, P.O., Li, W., and Christopher, L.A. (2012). Correction of a liquid lens for 3D imaging systems. Proc. SPIE 8429, 842909.

PAGE 134

117 Burger, B., Meimon, S.C., Petit, C., and Nguyen, M.C. (2015). Improvement of Varioptic’s liquid lens based on electrowetting: how to obtain a short response time and its application in the design of a high resolution iris biometric system. Proc. SPIE 9375, 93750S. Cai, D.J., Aharoni, D., Shuman, T., Shobe, J., Biane, J., Song, W., Wei, B., Veshkini, M., La Vu, M., Lou, J., et al. (2016). A shared neural ensemble links distinct contextual memories encoded close in time. Nature 534, 115–118. Chen, J.L., Andermann, M.L., Keck, T., Xu, N.L., and Ziv, Y. (2013a). Imaging Neuronal Populations in Behaving Rodents: Paradigms for Studying Neural Circuits Underlying Behavior in the Mammalian Cortex. J. Neurosci. 33 , 17631–17640. Chen, L., Kumar, S., Kelly, D., Andrews, N., Dallman, M.J., French, P.M.W., and Mcginty, J. (2014). Remote focal scanning optical projecti on tomography with an electrically tunable lens. Biomed. Opt. Express 5, 3367–3375. Chen, Q., Cichon, J., Wang, W., Qiu, L., Lee, S.J.R., Campbell, N.R., DeStefino, N., Goard, M.J., Fu, Z., Yasuda, R., et al. (2012). Imaging Neural Activity Using Thy1GCaM P Transgenic Mice. Neuron 76, 297–308. Chen, T.W., Wardill, T.J., Sun, Y., Pulver, S.R., Renninger, S.L., Baohan, A., Schreiter, E.R., Kerr, R. a, Orger, M.B., Jayaraman, V., et al. (2013b). Ultrasensitive fluorescent proteins for imaging neuronal activit y. Nature 499 , 295–300. Chen, X., Reichenbach, K.L., and Xu, C. (2008). Experimental and theoretical analysis of core to core coupling on fiber bundle imaging. Opt. Express 16, 21598–21607. Cheng, P.C. (2006). The contrast formation in optical microscopy. In Handbook of Biological Confocal Microscopy: Third Edition, pp. 162–206. Cheon, G.W., Cha, J., and Kang, J.U. (2014). Spatial compound imaging for fiber bundle optic microscopy. Proc. SPIE 8938, 893811. Chiu, C.P., Chiang, T.J., Chen, J.K., Chang, F.C., Ko, F.H., Chu, C.W., Kuo, S.W., and Fan, S.K. (2012). Liquid Lenses and Driving Mechanisms: A Review. J. Adhes. Sci. Technol. 26, 1033–1052. Clark, S.W., Ilday, F.., and Wise, F.W. (2001). Fiber delivery of femtosecond pulses from a Ti:sapphire la ser. Opt. Lett. 26 , 1320–1322. Conchello, J., and Lichtman, J.W. (2005). Optical sectioning microscopy. Nat. Methods 2, 920–931. Dana, H., Chen, T.W., Hu, A., Shields, B.C., Guo, C., Looger, L.L., Kim, D.S., and Svoboda, K. (2014). Thy1GCaMP6 Transgenic Mice for Neuronal Population Imaging In Vivo. PLoS One 9, e108697.

PAGE 135

118 Deisseroth, K. (2015). Optogenetics: 10 years of microbial opsins in neuroscience. Nat. Neurosci. 18 , 1213–1225. Denk, W., Strickler, J., and Webb, W. (1990). Twophoton laser scanning fluorescence microscopy. Science (80 . ). 248, 73 –76. Diaspro, A., Chirico, G., and Collini, M. (2005). Twophoton fluorescence excitation and related techniques in biological microsco py. Q. Rev. Biophys. 38, 97–166. Dombeck, D., and Tank, D. (2014). Twophoton imaging of neural activity in awake mobile mice. Cold Spring Harb. Protoc. 2014, 726–736. Dombeck, D., Khabbaz, A., and Collman, F. (2007). Imaging large scale neural activity wi th cellular resolution in awake, mobile mice. Neuron 56, 43 –57. Dombeck, D.A., Harvey, C.D., Tian, L., Looger, L.L., and Tank, D.W. (2010). Functional imaging of hippocampal place cells at cellular resolution during virtual navigation. Nat. Neurosci. 13, 1433 –1440. Ducourthial, G., Leclerc, P., Mansuryan, T., Fabert, M., Brevier, J., Habert, R., Braud, F., Batrin, R., Vever Bizet, C., BourgHeckly, G., et al. (2016). Development of a real time flexible multiphoton microendoscope for label free imaging in a live animal. Sci. Rep. 5, 18303. Emiliani, V., Cohen, A.E., Deisseroth, K., and Hausser, M. (2015). All Optical Interrogation of Neural Circuits. J. Neurosci. 35 , 13917–13926. Engelbrecht, C.J., Johnston, R.S., Seibel, E.J., and Helmchen, F. (2008). Ultra compact fiber optic two photon microscope for functional fluorescence imaging in vivo. Opt. Express 16, 5556–5564. Fahrbach, F.O., Voigt, F.F., Schmid, B., Helmchen, F., and Huisken, J. (2013). Rapid 3D light sheet microscopy with a tunable lens. Opt. Expr ess 21, 21010–21026. Flusberg, B.A., Jung, J.C., Cocker, E.D., Anderson, E.P., and Schnitzer, M.J. (2005a). In vivo brain imaging using a portable 3.9 gram twophoton fluorescence microendoscope. Opt. Lett. 30 , 2272–2274. Flusberg, B.A., Cocker, E.D., Piya wattanametha, W., Jung, J.C., Cheung, E.L.M., and Schnitzer, M.J. (2005b). Fiber optic fluorescence imaging. Nat. Methods 2, 941–950. Flusberg, B.A., Nimmerjahn, A., Cocker, E.D., Mukamel, E.A., Robert, P., Barretto, J., Ko, T., Burns, L., Jung, J.C., and Schnitzer, M.J. (2008a). Highspeed , miniaturized fluorescence microscopy in freely moving mice. Nat. Methods 5, 935–938. Flusberg, B .A., Schnitzer, M.J., and Ko, T.H. (2008b). Microendoscopy with corrective optics (United States).

PAGE 136

119 Ghosh, K.K., Burns, L.D., Cocker, E.D., Nimmerjahn, A., Ziv, Y., Gamal, A. El, and Schnitzer, M.J. (2011). Miniaturized integration of a fluorescence microsc ope. Nat. Methods 8, 871–878. Gibson, E.A., Gaudiosi, D.M., Kapteyn, H.C., Jimenez, R., Kane, S., Huff, R., Durfee, C., and Squier, J. (2006). Efficient reflection grisms for pulse compression and dispersion compensation of femtosecond pulses. Opt. Lett. 3 1, 3363–3365. Gmitro, A.F., and Aziz, D. (1993). Confocal microscopy through a fiber optic imaging bundle. Opt. Lett. 18, 565–567. Gbel, W., Kerr, J.N.D., Nimmerjahn, A., and Helmchen, F. (2004). Miniaturized two photon microscope based on a flexible cohe rent fiber bundle and a gradient index lens objective. Opt. Lett. 29, 2521–2523. Gbel, W., Kampa, B.M., and Helmchen, F. (2007). Imaging cellular network dynamics in three dimensions using fast 3D laser scanning. Nat. Methods 4, 73–79. Goldey, G.J., Roumi s, D.K., Glickfeld, L.L., Kerlin, A.M., Reid, R.C., Bonin, V., and Andermann, M.L. (2014). Removable cranial windows for longterm imaging in awake mice. Nat. Protoc. 9, 2515–2538. Gppert Mayer, M. (1931). ber Elementarakte mit zwei Quantensprngen. Ann. Phys. 401, 273–294. Grewe, B.F., Voigt, F.F., van ’t Hoff, M., and Helmchen, F. (2011). Fast twolayer two photon imaging of neuronal cell populations using an electrically tunable lens. Biomed. Opt. Express 2, 2035–2046. Grewe, B.F., Grndemann, J., Kitc h, L.J., Lecoq, J.A., Parker, J.G., Marshall, J.D., Larkin, M.C., Jercog, P.E., Grenier, F., Li, J.Z., et al. (2017). Neural ensemble dynamics underlying a longterm associative memory. Nature 543, 670– 675. Haider, B., Husser, M., and Carandini, M. (2012). Inhibition dominates sensory responses in the awake cortex. Nature 493, 97–100. Han, J.H., and Yoon, S.M. (2011). Depixelation of coherent fiber bundle endoscopy based on learning patterns of image prior. Opt. Lett. 36, 3212 –3214. Han, J.H., and Kang, J.U. (2012). Effect of multimodal coupling in imaging microendoscopic fiber bundle on optical coherence tomography. In Applied Physics B: Lasers and Optics, pp. 635–643. Han, J.H., Lee, J., and Kang, J.U. ( 2010). Pixelation effect removal from fiber bundle probe based optical coherence tomography imaging. Opt. Express 18, 7427–7439.

PAGE 137

120 Harris, K.D., Henze, D.A., Csicsvari, J., Hirase, H., and Buzski, G. (2000). Accuracy of tetrode spike separation as determined by simultaneous intracellular and extracellular measurements. J. Neurophysiol. 84, 401–414. Helmchen, F., and Denk, W. (2005). Deep tissue twophoton microscopy. Nat. Methods 2, 932–940. Helmchen, F., Fee, M.S., Tank, D.W., and Denk, W. (2001). A miniature head mounted two photon microscope: Highresolution brain imaging in freely moving animals. Neuron 31, 903–912. Helmchen, F., Denk, W., and Kerr, J.N.D. (2013). Miniaturization of TwoPhoton Microscopy for Imaging in Freely Moving Animals. Cold Spring H arb. Protoc. 2013, 904–913. Hendriks, B.H.W., Kuiper, S., Van As, M.A.J., Renders, C.A., and Tukker, T.W. (2005). Electrowetting based variablefocus lens for miniature systems. Opt. Rev. 12, 255–259. Heys, J.G., Rangarajan, K. V, and Dombeck, D. a (2014). The Functional Micro organization of Grid Cells Revealed by CellularResolution Imaging. Neuron 84, 1079– 1090. Hillman, E.M.C. (2007). Optical brain imaging in vivo: techniques and applications from animal to man. J. Biomed. Opt. 12, 51402. Hoover, E.E., and Squier, J.A. (2014). Advances in multiphoton microscopy technology. Nat. Photonics 8, 219–223. Horstmeyer, R., Ruan, H., and Yang, C. (2015). Guidestar assisted wavefront shaping methods for focusing light into biological tissue. Nat. Photonics 9, 563–571. Horton, N.G., Wang, K., Kobat, D., Clark, C.G., Wise, F.W., Schaffer, C.B., and Xu, C. (2013). In vivo three photon microscopy of subcortical structures within an intact mouse brain. Nat. Photonics 7, 205–209. Hughes, E.G., Kang, S.H., Fukaya, M., and Bergles, D.E. (2013a). Oligodendrocyte progenitors balance growth with self repulsion to achieve homeostasis in the adult brain. Nat. Neurosci. 16, 668– 676. Hughes, M., Chang, T.P., and Yang, G.Z. (2013b). Fiber bundle endocytoscopy. Biomed. Opt. Express 4, 2781–2794. Inou, S. (2006). Foundations of Confocal Scanned Imaging in Light Microscopy. In Handbook of Biological Confocal Microscopy: Third Edition, (Boston, MA: Springer US), pp. 1–19.

PAGE 138

121 Jabbour, J.M., Saldua, M. a., Bixler, J.N., and Maitland, K.C. (2012). Confocal endomicroscopy: Instrumentation and medical applications. Ann. Biomed. Eng. 40, 378– 397. Jabbour, J.M., Malik, B.H., Olsovsky, C., Cuenca, R., Cheng, S., Jo, J. a, Cheng, Y.S. L., Wright, J.M., and Maitland, K.C. (2014). Optical axial scanning in confocal microscopy using an electrically tunable lens. Biomed. Opt. Express 5, 645–652. Jeong, H., Yoo, H., and Gweon, D. (2016). Highspeed 3 D measurement with a large field of view based on direct view confocal microscope with an electrically tunable lens. Opt. Express 24, 3806–3816. Jiang, J., Zhang, D., Walker, S., Gu, C., Ke, Y., Yung, W.H., and Chen, S. (2015). Fast 3D temporal focusing microscopy using an electrically tunable l ens. Opt. Express 23, 24362–24368. Jung, J.C. (2004). In Vivo Mammalian Brain Imaging Using One and TwoPhoton Fluorescence Microendoscopy. J. Neurophysiol. 92, 3121–3133. Jung, J.C., and Schnitzer, M.J. (2003). Multiphoton endoscopy. Opt. Lett. 28, 902–904. Kalashyan, M., Lefort, C., Martnez Len, L., Mansuryan, T., Mouradian, L., and Louradour, F. (2012). Ultrashort pulse fiber delivery with optimized dispersion control by reflection grisms at 800 nm. Opt. Express 20, 25624–25635. Katona, G., Szala y, G., Mak, P., Kaszs, A., Veress, M., Hillier, D., Chiovini, B., Vizi, E.S., Roska, B., and Rzsa, B. (2012). Fast twophoton in vivo imaging with three dimensional random access scanning in large tissue volumes. Nat. Methods 9, 201–208. Kim, D., Choi, H., Yazdanfar, S., and So, P.T.C. (2008). Ultrafast optical pulse delivery with fibers for nonlinear microscopy. Microsc. Res. Tech. 71, 887–896. Kim, D., Moon, J., Kim, M., Yang, T.D., Kim, J., Chung, E., and Choi, W. (2014). Toward a miniature endomicros cope: pixelation free and diffraction limited imaging through a fiber bundle. Opt. Lett. 39 , 1921–1924. Knittel, J., Schnieder, L., Buess, G., Messerschmidt, B., and Possner, T. (2001). Endoscope compatible confocal microscope using a gradient index lens system. Opt. Commun. 188, 267–273. Koshiba, M., Saitoh, K., and Kokubun, Y. (2009). Heterogeneous multi core fibers: proposal and design principle. IEICE Electron. Express 6, 98 –103. Kostuk, R.K., and Carriere, J. (2000). Interfaces to fiber image guide bun dles for high density optical interconnects. In Proceedings of SPIE, R.A. Lessard, and T. V. Galstian, eds. pp. 987–992.

PAGE 139

122 Koukourakis, N., Finkeldey, M., Strmer, M., Leithold, C., Gerhardt, N.C., Hofmann, M.R., Wallrabe, U., Czarske, J.W., and Fischer, A. (2014). Axial scanning in confocal microscopy employing adaptive lenses (CAL). Opt. Express 22, 6025–6039. Lee, C. Y., and Han, J.H. (2013a). Elimination of honeycomb patterns in fiber bundle imaging by a superimposition method. Opt. Lett. 38, 2023–2025. Lee, C.Y., and Han, J.H. (2013b). Integrated spatiospectral method for efficiently suppressing honeycomb pattern artifact in imaging fiber bundle microscopy. Opt. Commun. 306, 67–73. Lee, W.M., and Yun, S.H. (2011). Adaptive aberration correction of GRIN lenses for confocal endomicroscopy. Opt. Lett. 36, 4608–4610. Lefort, C., Mansuryan, T., Louradour, F., and Barthelemy, A. (2011). Pulse compression and fiber delivery of 45 fs Fourier transform limited pulses at 830 nm. Opt. Lett. 36, 292– 294. Lefort, C., Kalashyan, M., Ducourthial, G., Mansuryan, T., O’Connor, R.P., and Louradour, F. (2014). Sub30fs pulse compression and pulse shaping at the output of a 2m long optical fiber in the near infrared range. J. Opt. Soc. Am. B 31, 2317–2324. Lelek, M., Suran, E., Louradour, F., Barthelemy, A., Viellerobe, B., and Lacombe, F. (2007). Coherent femtosecond pulse shaping for the optimization of a nonlinear micro endoscope. Opt. Express 15, 10154–10162. Li, T., and Kogelnik, H. (1966). Laser Beams and Resonators. Appl. Opt. 5, 1550–1567. Li, A., Gire, D.H., Bozza, T., and Restrepo, D. (2014). Precise Detection of Direct Glomerular Input Duration by the Olfactory Bulb. J. Neurosci. 34, 16058–16064. Li, H., Duan, X., Qiu, Z., Zhou, Q., Kurabayashi, K., Oldham, K.R., and Wang, T.D. (2016). Integrated monolithic 3D MEMS scanner for switchable real time vertical/horizontal cross sectional imaging. Opt. Express 24, 2145–2155. Liang, C., Sung, K.B., Richards Kortum, R.R., and Descour, M.R. (2002). Design of a highnumeri cal aperture miniature microscope objective for an endoscopic fiber confocal reflectance microscope. Appl. Opt. 41, 4603–4610. Lin, M.Z., and Schnitzer, M.J. (2016). Genetically encoded indicators of neuronal activity. Nat. Neurosci. 19 , 1142–1153. Liu, J.T.C., Mandella, M.J., Loewke, N.O., Haeberle, H., Ra, H., Piyawattanametha, W., Solgaard, O., Kino, G.S., and Contag, C.H. (2015). Micromirror scanned dual axis confocal microscope utilizing a gradient index relay lens for image guidance during brain surge ry. J. Biomed. Opt. 15, 26029.

PAGE 140

123 Liu, L., Wu, L., Zory, P., and Xie, H. (2010). Fiber optic confocal microscope with an electrothermally actuated, large tunable range microlens scanner for depth scanning. Proc. IEEE Int. Conf. Micro Electro Mech. Syst. 2, 827–830. Liu, X., Huang, Y., and Kang, J.U. (2011). Dark field illuminated reflectance fiber bundle endoscopic microscope. J. Biomed. Opt. 16, 46003. Liu, X., Zhang, L., Kirby, M., Becker, R., Qi, S., and Zhao, F. (2016). Iterative l_1 min algorithm for fixed pattern noise removal in fiber bundle based endoscopic imaging. J. Opt. Soc. Am. A 33, 630–636. Maitland, K.C., Shin, H.J., Ra, H., Lee, D., Solgaard, O., and Richards Kortum, R. (2006). Single fiber confocal microscope with a twoaxis gimbaled MEMS scanner for cellular imaging. Opt. Express 14, 8604–8612. Martone, M. (2002). The Cell: An Image Library Image CIL:39952. Meinert, T., Weber, N., Zappe, H., and Seifert, A. (2014). Varifocal MOEMS fiber scanner for confocal endomicroscopy. Opt. Express 22, 31529–31544. Meyer, S.A., Ozbay, B.N., Potcoava, M., Salcedo, E., Restrepo, D., and Gibson, E.A. (2016). Super resolution imaging of ciliary microdomains in isolated olfactory sensory neurons using a custom twocolor stimulated emission depletion microsco pe. J. Biomed. Opt. 21, 66017. Minsky, M. (1957). Microscopy Apparatus. Murade, C.U., van der Ende, D., and Mugele, F. (2012). High speed adaptive liquid microlens array. Opt. Express 20, 18180–18187. Myaing, M.T., MacDonald, D.J., and Li, X. (2006). Fiber optic scanning twophoton fluorescence endoscope. Opt. Lett. 31 , 1076–1077. Nadella, K.M.N.S., Ro, H., Baragli, C., Griffiths, V.A., Konstantinou, G., Koimtzis, T., Evans, G.J., Kirkby, P.A., and Silver, R.A. (2016). Random access scanning microscopy for 3D imaging in awake behaving animals. Nat. Methods 13, 1001–1004. Niederriter, R.D., Watson, A.M., Zahreddine, R.N., Cogswell, C.J., Cormack, R.H., Bright, V.M., and Gopinath, J.T. (2013). Electrowetting lenses for compensating phase and curvature distort ion in arrayed laser systems. Appl. Opt. 52, 3172–3177. Oron, D., and Silberberg, Y. (2015). Temporal focusing microscopy. Cold Spring Harb. Protoc. 2015 , 145–151. Ouzounov, D.G., Rivera, D.R., Webb, W.W., Bentley, J., and Xu, C. (2013). Miniature varifoca l objective lens for endomicroscopy. Opt. Lett. 38, 3103–3106.

PAGE 141

124 Ozbay, B.N., Losacco, J.T., Cormack, R., Weir, R., Bright, V.M., Gopinath, J.T., Restrepo, D., and Gibson, E.A. (2015). Miniaturized fiber coupled confocal fluorescence microscope with an elect rowetting variable focus lens using no moving parts. Opt. Lett. 40, 2553–2556. Ozden, I., Lee, H.M., Sullivan, M.R., and Wang, S.S.H. (2008). Identification and Clustering of Event Patterns From In Vivo Multiphoton Optical Recordings of Neuronal Ensembles. J. Neurophysiol. 100, 495–503. Paddock, S.W. (2000). Principles and practices of laser scanning confocal microscopy. Mol. Biotechnol. 16, 127–149. Papagiakoumou, E., Bgue, A., Leshem, B., Schwartz, O., Stell, B.M., Bradley, J., Oron, D., and Emiliani, V . (2013). Functional patterned multiphoton excitation deep inside scattering tissue. Nat. Photonics 7, 274–278. Paxinos, G., and Franklin, K. (2012). Paxinos and Franklin’s the Mouse Brain in Stereotaxic Coordinates (Elsevier Academic Press). Peron, S., Chen, T.W., and Svoboda, K. (2015a). Comprehensive imaging of cortical networks. Curr. Opin. Neurobiol. 32, 115–123. Peron, S.P., Freeman, J., Iyer, V., Guo, C., and Svoboda, K. (2015b). A Cellular Resolution Map of Barrel Cortex Activity during Tactile Beh avior. Neuron 86, 783–799. Pierce, M., Yu, D., and Richards Kortum, R. (2011). Highresolution Fiberoptic Microendoscopy for in situ Cellular Imaging. J. Vis. Exp. e2306. Piyawattanametha, W. (2014). MEMS based fiber optical microendoscopy. Int. Conf. Opt . MEMS Nanophotonics 107–108. Planas, S. a, Mansur, N.L., Cruz, C.H., and Fragnito, H.L. (1993). Spectral narrowing in the propagation of chirped pulses in single mode fibers. Opt. Lett. 18, 699 –701. Porat, A., Andresen, E.R., Rigneault, H., Oron, D., Gigan, S., and Katz, O. (2016). Widefield lensless imaging through a fiber bundle via speckle correlations. Opt. Express 24, 16835–16855. Qiu, Z., and Piyawattanamatha, W. (2017). New Endoscopic Imaging T echnology Based on MEMS Sensors and Actuators. Micromachines 8, 210. Qiu, Z., and Piyawattanametha, W. (2015). MEMS Based Medical Endomicroscopes. IEEE J. Sel. Top. Quantum Electron. 21, 376–391. Qiu, Z., Rhee, C.H., Choi, J., Wang, T.D., and Oldham, K.R. (2014). Large Stroke Vertical PZT Microactuator With High Speed Rotational Scanning. J. Microelectromechanical Syst. 23, 256–258.

PAGE 142

125 Quirin, S., Jackson, J., Peterka, D.S., and Yuste, R. (2014). Sim ultaneous imaging of neural activity in three dimensions. Front. Neural Circuits 8, 1 –11. Ramn y Cajal, S. (1888). Estructura de los centros nerviosos de las aves. Rev. Trimest. Histol. Norm. Y Patolgica 1, 1–32. Ramn y Cajal, S. (1899). Comparative study of the sensory areas of the human cortex (FORGOTTEN BOOKS). Rickgauer, J.P., Deisseroth, K., and Tank, D.W. (2014). Simultaneous cellular resolution optical perturbation and imaging of place cell firing fields. Nat. Neurosci. 17 , 1816–1824. Rivera, D., and Brown, C. (2011). Compact and flexible raster scanning multiphoton endoscope capable of imaging unstained tissue. Proc. Natl. Acad. Sci. 108 , 17598–17603. Rivera, D.R., Brown, C.M., Ouzounov, D.G., Webb, W.W., and Xu, C. (2012). Multifocal multiphoton endoscope. Opt. Lett. 37 , 1349–1351. Rupp, S., Winter, C., and Elter, M. (2009). Evaluation of spatial interpolation strategies for the removal of combstructure in beroptic images. In Proceedings of the 31st Annual International Conference of the IEEE Engineering in Medicine and Biology Society, pp. 3677–3680. Ryan, D.P., Gould, E.A., Seedorf, G.J., Masihzadeh, O., Abman, S.H., Vijayaraghavan, S., Macklin, W.B., Restrepo, D., and Shepherd, D.P. (2017). Automatic and adaptive heterogeneous refractive index compensation for light sheet microscopy. Nat. Commun. 8, 612. Sawinski, J., Wallace, D.J., Greenberg, D.S., Grossmann, S., Denk, W., and Kerr, J.N.D. (2009). Visually evoked activity in cortical cells imaged in freely moving animals. Proc. Natl. Acad. Sci . 106, 19557–19562. Shao, Y., Dickensheets, D., and Himmer, P. (2004). 3D MOEMS Mirror for Laser Beam Pointing and Focus Control. IEEE J. Sel. Top. Quantum Electron. 10, 528–535. Sherlock, B., Warren, S.C., Alexandrov, Y., Yu, F., Stone, J.M., Knight, J.C ., Neil, M.A.A., Peterson, C., French, P.M.W., and Dunsby, C.W. (2017). In vivo multiphoton microscopy using a handheld scanner with lateral and axial motion compensation. J. Biophotonics. Shih, A.Y., Mateo, C., Drew, P.J., Tsai, P.S., and Kleinfeld, D. (2012). A Polished and Reinforced Thinnedskull Window for Longterm Imaging of the Mouse Brain. J. Vis. Exp. 4–9. Shinde, A., and Matham, M.V. (2014). Pixelate Removal in an Image Fiber Probe Endoscope Incorporating Comb Structure Removal Methods. J. Med. I maging Heal. Informatics 4, 203–211.

PAGE 143

126 Singh, A., McMullen, J.D., Doris, E.A., and Zipfel, W.R. (2015). Comparison of objective lenses for multiphoton microscopy in turbid samples. Biomed. Opt. Express 6, 3113. Smith, N.R., Abeysinghe, D.C., Haus, J.W., and Heikenfeld, J. (2006). Agile wide angle beam steering with electrowetting microprisms. Opt. Express 14, 6557–6563. Sofroniew, N.J., Flickinger, D., King, J., and Svoboda, K. (2016). A large field of view two photon mesoscope with subcellular resolution for in vivo imaging. Elife 5, 1–20. Sotelo, C. (2003). Viewing the brain through the master hand of Ramon y Cajal. Nat. Rev. Neurosci. 4, 71–77. Specht, E.A., Braselmann, E., and Palmer, A.E. (2017). A Critical and Comparative Review of Fluorescent Tools for Live Cell Imaging. Annu. Rev. Physiol. 79, 93–117. Stasio, N., Conkey, D.B., Moser, C., and Psaltis, D. (2015). Light control in a multicore fiber using the memory effect. Opt. Express 23, 30532–30544. Stelzer, E.H. (2006). The intermediate optical system of laser scanning confocal microscopes. In Handbook of Biological Confocal Microscopy: Third Edition, pp. 207 – 208. Stone, J.M., Wood, H.A.C., Harrington, K., and Birks, T.A. (2017). Low index contrast imaging fibers. Opt. Lett. 42, 1484–1487. Stosiek, C., Garaschuk, O., Holthoff, K., and Konnerth, A. (2003). In vivo twophoton calcium imaging of neuronal networks. Proc. Natl. Acad. Sci. 100, 7319–7324. Supekar, O.D., Zohrabi, M., Gopinath, J.T., and Bright, V.M. (2017a). Enhanced Response Time of Electrowet ting Lenses with Shaped Input Voltage Functions. Langmuir 33 , 4863–4869. Supekar, O.D., Ozbay, B.N., Zohrabi, M., Nystrom, P.D., Futia, G.L., Restrepo, D., Gibson, E.A., Gopinath, J.T., and Bright, V.M. (2017b). Twophoton laser scanning microscopy with el ectrowetting based prism scanning. Biomed. Opt. Express 8, 4546– 4556. Svoboda, K., and Yasuda, R. (2006). Principles of TwoPhoton Excitation Microscopy and Its Applications to Neuroscience. Neuron 50, 823–839. Szabo, V., Ventalon, C., DeSars, V., Bradley, J., Emiliani, V., De Sars, V., Bradley, J., and Emiliani, V. (2014). Spatially Selective Holographic Photoactivation and Functional Fluorescence Imaging in Freely Behaving Mice with a Fiberscope. Neuron 84, 1157– 1169.

PAGE 144

127 Szalay, G., Judk, L., Katona, G., ”csai, K., Juhsz, G., Veress, M., Szadai, Z., Fehr, A., Tompa, T., Chiovini, B., et al. (2016). Fast 3D Imaging of Spine, Dendritic, and Neuronal Assemblies in Behaving Animals. Neuron 92, 723–738. Tantirigama, M.L.S., Huang, H.H.Y., and Bekkers, J.M. (2017). Spontaneous activity in the piriform cortex extends the dynamic range of cortical odor coding. Proc. Natl. Acad. Sci. 114, 2407–2412. Toomre, D., and Pawley, J.B. (2006). DiskScanning Confocal Microscopy (Boston, MA: Springer US). Trebino, R., DeLong, K.W., Fittinghoff, D.N., Sweetser, J.N., Krumbgel, M.A., Richman, B. a., and Kane, D.J. (1997). Measuring ultrashort laser pulses in the time frequency domain using frequencyresolved optical gating. Rev. Sci. Instru m. 68, 3277– 3295. Volpi, D., Tullis, I.D.C., Barber, P.R., Augustyniak, E.M., Smart, S.C., Vallis, K.A., and Vojnovic, B. (2017). Electrically tunable fluidic lens imaging system for laparoscopic fluorescence guided surgery. Biomed. Opt. Express 8, 991–998. Wachowiak, M., Economo, M.N., Daz Quesada, M., Brunert, D., Wesson, D.W., White, J. a, and Rothermel, M. (2013). Optical dissection of odor information processing in vivo using GCaMPs expressed in specified cell types of the olfactory bulb. J. Neurosci. 33 , 5285–5300. Warden, M.R., Cardin, J.A., and Deisseroth, K. (2014). Optical Neural Interfaces. Annu. Rev. Biomed. Eng. 16, 103–129. Warren, S.C., Kim, Y., Stone, J.M., Mitchell, C., Knight, J.C., Neil, M.A.A., Paterson, C., French, P.M.W., and Dunsby, C . (2016). Adaptive multiphoton endomicroscopy through a dynamically deformed multicore optical fiber using proximal detection. Opt. Express 24, 21474–21484. Webb, R.H. (1999). Confocal optical microscopy. Reports Prog. Phys. 59, 427–471. Wilt, B.A., Burns, L.D., Ho, E.T., Ghosh, K.K., Mukamel, E.A., and Schnitzer, M.J. (2009). Advances in light microscopy for neuroscience. Annu. Rev. Neurosci. 32, 435 – 506. Winter, C., Rupp, S., Elter, M., Mnzenmayer, C., Gerhuser, H., and Wittenberg, T. (2006). Automatic adaptive enhancement for images obtained with fiberscopic endoscopes. IEEE Trans. Biomed. Eng. 53, 2035–2046. Wood, H.A.C., Harrington, K., Stone, J.M., Birks, T.A., and Knight, J.C. (2017). Quantitative characterization of endoscopic imaging fibers. Opt. Express 25, 1985–1992.

PAGE 145

128 Wu, Y., Xi, J., Cobb, M.J., and Li, X. (2009). Scanning fiber optic nonlinear endomicroscopy with miniature aspherical compound lens and multimode fiber collector. Opt. Lett. 34 , 953–955. Wu, Y., Zhang, Y., Xi, J., Li, M.J., and Li, X. (2010). Fiber optic nonlinear endomicroscopy with focus scanning by using shape memory alloy actuation. J. Biomed. Opt. 15, 60506. Xie, T., Guo, S., Chen, Z., Mukai, D., and Brenner, M. (2006). GRIN lens rod based probe for endoscopic spectral domain optical coherence tomography with fast dynamic focus tracking. Opt. Express 14, 3238–3246. Zeng, X., and Jiang, H. (2009). An endoscope utilizing tunable focus microlenses actuated through infrared light. In TRANSDUCERS 2009 2009 International Solid State Sensors, Actuators and Microsystems Conference, (IEEE), pp. 1214 –1217. Zhang, F., Wang, L.P., Brauner, M., Liewald, J.F., Kay, K., Watzke, N., Wood, P.G., Bamberg, E., Nagel, G., Gottschalk, A., et al. (2007). Multimodal fast optical interrogation of neural circuitry. Nature 446, 633 –639. Zhao, Y.P., and Wang, Y. (2013). Fundamentals and Applications of Electrowetting. Rev. Adhes. Adhes. 1, 114–174. Zhao, Y., Nakamura, H., and Gordon, R.J. (2010). Development of a versatile twophoton endoscope for biological imaging. Biomed. Opt. Express 1, 1159– 1172. Ziv, Y., Burns, L.D., Cocker, E.D., Hamel, E.O., Ghosh, K.K., Kitch, L.J., El Gamal, A., and Schnitzer, M.J. (2013). Longterm dynamics of CA1 hippocampal place codes. Nat. Neurosci. 16 , 264–266. Zong, W., Wu, R., Li, M., Hu, Y., Li, Y., Li, J., Rong, H., Wu, H., Xu, Y., Lu, Y., et al. (2017). Fast highresolution miniature two photon microscopy for brain imaging in freely behaving mice. Nat. Methods 14, 713–719.