A mating system conundrum

Material Information

A mating system conundrum hybridization in Apocynum (Apocynaceae)
Johnson, Samuel A
Publication Date:
Physical Description:
viii, 43 leaves : illustrations ; 29 cm


Subjects / Keywords:
Apocynaceae ( lcsh )
Hybridization ( lcsh )
Apocynaceae ( fast )
Hybridization ( fast )
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )


Includes bibliographical references (leaves 40-43).
General Note:
Submitted in partial fulfillment of the requirements for the degree, Master of Arts, Biology.
General Note:
Department of Integrative Biology
Statement of Responsibility:
by Samuel A. Johnson.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
37853958 ( OCLC )
LD1190.L45 1997m .J64 ( lcc )

Full Text
Samuel A. Johnson
B. S., Dominican College, 1970
A thesis submitted to the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Master of Arts

This thesis for the Master of Arts
degree by
Samuel A. Johnson
has been approved
Diana F. T omback
Leo P. Bruederle
^ # /??<>
v Date/

Johnson, Samuel A. (M.A., Biology)
A Mating System Conundrum: Hybridization in Apocynum (Apocynaceae)
Thesis directed by Professor Diana F. Tomback
Apocynum x-floribundum Greene (= medium Greene) has long been
considered a hybrid of A. androsaemifolium L. and A. cannabinum L. This
assertion is based on intermediate morphological characters in A. x-
floribundum. The floral morphology in this genus, however, strongly
suggests obligate selfing, and reproductive success (fruit set) in most
populations is low to zero. Studies of insect visitors to flowers indicate that
pollen is not generally accessible and is not carried from flower to flower.
Hybridization was thus called into question. Allozyme evidence is used to
test the hypothesis that A. x-floribundum is a hybrid. Six diagnostic and
two strongly supportive loci were resolved for the parental species, and all
populations of A. x-floribundum studied were heterozygous at these loci.
Evidence from these and additional loci indicate that populations of all three
taxa tend to be strongly clonal. Observed heterozygosity is very low in the
parental species, suggesting a history of inbreeding or a severe bottleneck.
There is no support for earlier assumptions that some intermediates are
derived from backcrosses or "secondary hybrids;" hybrids appear to be Fj.

Morphological studies of plant height, leaf shape, petal length, sepal length,
follicle length, seed length, and number of seeds per follicle support the
conclusions drawn from allozyme analysis. The persistence and spread of
hybrid clones may contribute to an illusion that hybridization is common.
In spite of extensive investigation, the pollen vectors remain unidentified.
This abstract accurately represents the content of the candidates thesis.
I recommend its publication.
Diana F. T omback

I would like to thank Diana F. Tomback for her patience and refusal to
allow me to abandon a project that at first seemed hopeless; Leo P.
Bruederle for his wisdom and enthusiasm in the laboratory; and The
Colorado Springs School Parents' Association for financial support of this

Tables ..............................................viii
1. INTRODUCTION....................................1
2. MATERIALS & METHODS.............................7
Study Populations..............................7
Studies of Reproductive Biology................9
Morphology................................ 10
Population Genetics...........................11
3. RESULTS........................................13
Pollination Studies...........................13
Morphological Studies.........................18
Population Genetics...........................26
4. DISCUSSION.................................... 31

1.1 Floral Structure of Apocvnum x-floribundum..............3
a. The flower from above, showing connivent anthers
b. Arrangement of reproductive structures with one petal
c. Relative position of anthers and stigma
d. The gynoecium subtended by nectaries
1.2 Localities of Study Populations..........................5
3.1 Venn Diagram Illustrating the Intersection of Floral Visitor
Communities Among the Three Taxa of Apocvnum
Examined in this Study...............................16
3.2 Box-and-Whisker Diagrams of Seed Number Data for
Apocvnum Species.....................................21
3.3 Follicle Length in Apocvnum Populations.................23
3.4 Petal Length in Apocvnum Populations....................24
3.5 Plant Height in Apocvnum Populations....................25

3.1 Estimated Percent Fruit Set for Apocvnum Populations..14
3.2 Germination Percentages of Selected Apocvnum Populations 17
3.3 ANOVA and Tukeys Test Results of Analysis of Morphological
Characters Among Apocvnum cannabinum. A. x-
floribundum, and A. androsaemifolium..............19
3.4 Kruskal-Wallis Test Results for Seed Number per Follicle in
Apocvnum Species..................................20
3.5 Mean, Standard Deviation, and Range Values for Morphological
Measures from Apocvnum Species....................22
3.6 Allozyme Data for Apocvnum Populations................28
3.7 Allele Frequencies for Diagnostic and Supportive Loci for
Apocvnum Populations..............................29

This study began as an effort to quantify the efficiency of nocturnal pollinators
(especially moths) versus diurnal pollinators on Apocvnum L. flowers (see Chapter 2,
Methods and Materials). At the end of the flowering period, however, all experimental
treatments had no fmit set, and scrutiny of many other patches of Apocvnum in the
region showed that many of them also had no fruit set. Those that did set fruit
produced very few follicles. The question arose as to why a plant genus that attracts so
many insect visitors representing so many orders and species fails to get pollinated. A
monograph on the genus by Woodson (1930) suggested that one explanation might be
hybrid sterility, but I was not convinced of the hybrid status of many of these
populations because on close examination, the floral structure appears to be that of an
obligate self-pollinator. The focus of this study therefore shifted to an examination of
the status of the putative hybrid populations. Field studies, examination of insect
visitors, reproductive studies, morphological measurements, and allozyme studies were
all employed to answer questions about pollen vectors, population genetics, and
hybridization in the genus Apocvnum.
The taxonomy, phylogeny, pollination system, and ecological role of the strictly
North American genus Apocvnum (Apocynaceae) have long been in debate. In 1930
Woodson revised the genus, reducing over 80 described species to seven: A. pumilum
(A. Gray), A. androsaemifolium L., A. medium Greene, A. jonesii Woodson, A.
suksdorfii Greene, A. cannabinum L., and A. hvpericifolium Ait. The rest were
relegated to subspecific or hybrid rank. He assumed that most of his subspecies had

arisen by hybridization, noting numerous intermediate morphological characters, a high
degree of pollen sterility, and low to no fruit set. He assumed that hybridization was
common, but failed to explain how hybrids could arise in view of the curious floral
structure in this genus (Fig. 1.1, a-d). Anderson (1936) showed that the offspring of
putative parental species (A. cannabinum L. and A. androsaemifolium L.) are very
similar to the parents, while at least some of the intermediates produce variable
offspring. He regarded this as evidence of hybridization, but at that time neither
Anderson nor Woodson had the benefit of allozyme data to test the hybridization
Colorado's Front Range hosts many forms of Apocvnum. Contemporary field
guides (e.g. Weber 1990) distinguish only three "species:" A. cannabinum. A.
androsaemifolium. and A. x-medium Greene, the last named being an intermediate,
highly variable, putative hybrid taxon. Greene also described A. lividum (1901) and
A. floribundum (1893) from among the same general group of intermediates, but the
characters that have been used to separate A. medium from A. floribundum and A.
lividum (mainly extent and placement of pubescence), are considered inconsistent.
Woodson (1930) considered both A. floribundum and A. lividum as varieties of A.
medium. McGregor et al (1986) also considered these names synonymous, but noted
that the specific epithet A. floribundum has priority. I found that the characters that
separate these and other taxa are indeed unreliable, and that even forms that seem to be
distinct are not consistently assignable outside of A. x-floribundum. This paper will
therefore follow McGregor et al (1986). Apocvnum sibiricum Jacq. is recorded from
the Front Range also. It is supposedly distinguished from A. cannabinum by having
sessile cordate-clasping leaves, smaller follicles, and sub-foliaceous inflorescence
bracts (Harrington, 1954), but I was unable to distinguish it. This taxon is regarded by
McGregor et al (1986) and Weber (1990) as synonymous with A. cannabinum.
Apocvnum flowers appear to be ideally suited to selfing (Fig. 1.1, a-d). A full,

Fig. 1.1 Floral Structure of Apocvnum x-floribundum. a. The flower from above,
showing connivent anthers, b. Arrangement of reproductive structures with one petal
removed, c. Relative position of anthers and stigma, d. The gynoecium subtended by

detailed description of floral morphology in the family can be found in Woodson
(1930) and Cronquist (1981). The most striking feature of the flower is that its
morphology appears to prevent both the import and export of pollen. The androecium
is connivent about the gynoecium, shielding it from contact with potential pollinators.
The anthers dehisce introrsely and are essentially glued together at the edges, so that
neither the stigma nor the pollen is accessible to flower visitors. The pollen sheds only
onto its own stigmatal surface. Below the stigma there is a mucus ring that pastes the
anthers onto the style and catches any pollen that may fall. Anthers never splay upon
maturation of the flower, but remain fused to the pistil until the corolla falls from the
fertilized ovary. This presents a troubling conundrum: how can such a mating system
produce hybrids? And the obvious question arises: are the intermediates really hybrids?
Apocvnum is apparently a diploid genus with n = 11 (Darlington 1956). It is
known to reproduce asexually by cloning (Woodson 1930). Dates of flowering and
fruit set vary widely, probably as a function of temperature and rainfall (personal
observations, 1990-1996). In the Front Range of Colorado A. x-floribundum is
normally the first to bloom, with flowers first opening during the middle of June (but
as early as 18 May), and all species in full flower by the end of June. Flowering may
continue into mid September, but usually peaks in early July and finishes by the first
week of August. In dry years, populations of this genus appear to be the leading
producers of nectar in some areas during late June and July, a role taken over in August
by species of Melilotus officinalis L., Medicago sativa L., Monarda fistulosa L. and
numerous others. Some extraordinarily dense populations of A. x-floribundum seem
to act as ecological keystones, attracting huge numbers of insect visitors when the
flowers of other species are few and scattered. Woodson (1930) noted that in the
western states some bee keepers allow plants of this genus to occupy large fields for
nectar production. Similarly, Wilson et al (1958) list Apocvnum as an important nectar
source for honey bees. The enormous list of insect visitors (Appendix) is curious in

Figure 1.2 Localities of Study Populations. Large circles represent Apocynum
cannabinum. squares represent A. androsaemifolium. and triangles represent A. Xz
floribundum. Cities are located with small circles.

view of the floral anatomy.
Although the putative parental species are broadly sympatric along the Front
Range, populations are separated by habitat preference (Fig. 1.2). Apocynum
cannabinum occurs on prairie river flood plains, terraces, and roadside ditches, usually
in fine sandy or loamy soils. Apocynum androsaemifolium inhabits valleys and slopes
in the Transition Zone, usually on gravelly soils in the partial shade of oaks or pines.
A. x-floribundum inhabits roadsides and streambeds, usually on sandy soils. Although
A. x-floribundum can occasionally be found in proximity to the other species, the two
putative parental species were never found within a half kilometer of each other, and are
usually much more distant than this. It is difficult to imagine that a visitor that
somehow contacted pollen on a mountain stream terrace would later fly out to a prairie
river bottom and somehow transfer the pollen to the protected stigma of a flower there.
For the above reasons I questioned the hybrid hypothesis. A polytypic genus
of highly inbred species might respond to a selection gradient (from the prairie to the
higher foothills) in the same way. The fact that intermediate morphs are found
geographically between the extreme morphs could be explained in either way as
hybrids thriving in marginal habitats, or as several species adapted specifically to a
selection gradient. I addressed the hybrid hypothesis by examining morphological
characters and allozyme genotypes for parental and putative hybrid populations. The
objectives of this study are to determine whether A. x-floribundum is a hybrid of A.
cannabinum and A. androsaemifolium. or whether it is a related species of more distant
origin. If it is a hybrid, is there evidence for introgression? Population dynamics of
these taxa are also described.

Study Populations
I have located over 60 populations of this genus between the towns of
Fountain, CO to the south and Boulder, CO to the north. Of these, 11 from the
southern portion of this range were selected for study (Fig. 1.2). Three stands are
clearly assignable to A. cannabinum [according to Woodson (1930) and subsequent
keys (Rydberg 1932, McGregor et al 1986, and Weber 1990)]; three stands are clearly
assignable to A. androsaemifolium according to the same authorities; and five stands
were selected to illustrate a wide range of A. x-floribundum variants. Populations
numbered cl, c2 and c3 are A. cannabinum: al, a2, and a3 are A. androsaemifolium:
and xl, x2, x3, x4, and x5 are A. x-floribundum.
Population cl is in Fountain Creek Regional Park in the town of Fountain (elev.
1740m). The population spans about 160m2. The habitat is disturbed floodplain with
widely scattered Populus sp.and Salix sp. trees. Population c2 lies along a severely
disturbed floodplain of Monument Creek in downtown Colorado Springs (elev.
1830m). The population spans about 900m2- The upstream portion is evenly mixed
with Salix exigua Nutt., Robinia pseudoacacia L., and other shrubs and sapling trees,
while the lower portion is a near monoculture that interdigitates with Equisetum arvense
L., grasses, and forbs. Population c3 is at the intersection of Fountain Creek and
Nevada Avenue in Colorado Springs (elev. 1800m). It is much smaller than the others,
about 120m2, and associates mainly with Salix exigua.

Population al occurs near the northbound rest area on Interstate Highway 25
near Larkspur (elev. 2040m). It is largely shaded by Ouercus gambelli Nutt, and Pinus
ponderosa Laws., and occurs at low density across about 1500m2. Population a2 is in
Bear Creek Canyon (elev. 1980m), along the southwest edge of Colorado Springs. It
lies in a narrow 50m-long strip along the partially shaded north side of the road on a
sand and gravel slope, and associates mainly with Q. gambelli. Population a3 is in
South Cheyenne Canyon (elev. 1935m) near Seven Falls. It is also in sandy soil
shaded by Q. gambelli. but is on level ground. It occupies both deep shade (under P.
ponderosa) and slight shade along about 40 meters on the northeast side of the road.
Populations numbered xl and x3 are similarly situated in the northeast quadrant
of Colorado Springs in Austin Bluffs (elev. 1905m) and Palmer Park (elev. 1960m),
respectively. Both occupy unstable sandy soils in intermittent stream bottoms in
association with Q. gambelli. Padus virginiana (L.) Miller, Ribes sp., and P.
ponderosa. Both are subject to occasional flash flooding, and both follow the narrow
streambeds for about 100 meters. Population x2 is along Mesa Road (elev. 1975m)
near Garden of the Gods Park in Colorado Springs. This habitat hosts Yucca glauca
Nutt., Rhus trilobata (Nutt.) Weber, Ribes sp., Cercocarpus montanus Rafinesque,
and Eriogonum sp. Population x4 is the most extreme variant of A. x-floribundum
encountered. It is beside Bear Creek Road about a hundred meters west of the entrance
to Bear Creek Regional Park (elev. 1950m). It occupies about 15 meters of stream
terrace roadside on gentle south slopes. It is mostly shaded by large Populus sp. trees.
The plants are unusually tall (mean of about 160 cm versus 94 cm for other A. Xz
floribundum stands) and flower more profusely than any other Apocvnum stands
found. Population x5 is on the south end of Castle Rock on the west side of interstate
highway 25 (elev. 1890m). It is the smallest population studied, occupying a drainage
depression about ten meters across. This site is the most heavily disturbed of all,

hosting a few invasive weed species.
Studies of Reproductive Biology
Exclusionary cages were made to enclose 0.44m2 plots of plants or ramets to
study the efficiency of diurnal versus nocturnal pollinators. Three replicates each of
diumally-available, nocturnal ly-available, never-available, and always-available plots
were caged or, in the cases of always-available controls, marked with string. Each
morning and evening the cages enclosing the diumally- and nocturnally-available plots
were switched to exclude the unwanted pollinator sets. This experiment was only done
in the largest and densest patch of flowering Apocvnum found, population x 1.
Reproductive success was measured in numbers of fruit set in each plot.
I attempted to artificially cross-pollinate plants in the field using a hand lens,
forceps, and a fine brush. An anther was removed from the flower and pollen applied
to the exposed stigma. Out of over 65 attempts, about 30 flowers were pollinated in
this way.
Records were made of fruit-set as a percentage of the estimated total numbers of
flowers in a given population. Numbers of flowers were estimated as:
total flowers = flowers per ramet X ramets per m2 X population area in m2.
The actual (or estimated) total number of developed or developing follicles found was
divided by this number to derive percent reproductive success.
Seed viability was verified by germination of seeds in petrie dishes with damp
filter papers. Seeds were taken from representative populations of each taxon. In each

population tested, follicles were collected from numerous ramets, and seed samples
were amassed by taking three or four seeds from each follicle.
Field studies also included observation, collection, and identification of flower
visitors; examination of visitors for pollen load assessment; and insect behavior on
inflorescences. Insects were collected between 0800h and 2300h in June and July from
1993 through 1996. All specimens were taken by aerial net or killing jar directly from
flowers. All flower-visiting insects were examined, even though some (especially
butterflies) appeared to be nectar thieves, and others (mainly predators) appeared to
make only incidental contact with flowers while searching for prey or perching.
A few ramets of A. androsaemifolium and A. x-floribundum were dug up to
determine the depth of the connecting rhizomes. Clonal growth was verified in this
In order to compare morphological evidence with allozyme evidence (see
below), I measured both reproductive and vegetative characters. Characters chosen
were petal length, the ratio of petal length to sepal length, follicle length, number of
seeds per follicle, seed length, plant height, and the ratio of leaf length to leaf width.
The first three of these and plant height are used in identification keys as diagnostic
features of the taxa. The remainder are measures that appear to vary greatly among
populations, but which have apparently not been reported. Data were collected from
each population along haphazardly placed transects that criss-crossed stands, so that all
extremes of each population were represented. This was done to increase the
probability of obtaining different individuals rather than clonal ramets.
Morphological data were analyzed by one-way ANOVA whenever the data were

normally distributed, followed by Tukey's tests. Ratios (petal/sepal length and leaf
length/width) were arcsin transformed. In the case of numbers of seeds per follicle,
data were not normally distributed and could not be improved by transformations.
These data were analyzed with the non-parametric Kruskal-Wallis test. All tests were
run on StatMost (DataMost, 1994).
Population Genetics
Allozyme data were obtained for the 11 populations described above using
horizontal starch gel electrophoresis and substrate-specific staining. Twenty-four
ramets were sampled for each of the populations of A. cannabinum and A.
androsaemifolium (except c3, for which n = 20), with ten ramets from each population
of A. x-floribundum. All samples were collected along haphazard transects that criss-
crossed stands. This increased the probability of obtaining different individuals rather
than clonal shoots. Soluable enzymatic proteins were extracted by grinding mature
leaves in a 0.1M Tris-HCl buffer at pH 7.5, modified from Werth (1985) to 10%
polyvinylpyrrolidone (PVP-40) and 2-mercaptoethanol 0.050mL. Extracts were
adsorbed onto Whatman No. 17 chromatography paper wicks (3x13 mm) and stored
at -70C until electrophoresis.
The gel and electrode buffer systems used in this electrophoretic study were
morpholine citrate, pH 6.1 (Clayton and Tretiak 1972) and lithium borate, gel buffer
pH 8.5/electrode buffer pH 8.1 (Cheliak and Pitel, 1984). Gels of 10.5% starch and
were run at 50mA until a bromophenol blue marker had migrated about 12 cm. Gels
were then sliced and stained to resolve 14 loci: PGI-1, PGI-2, PGM-1, PGM-2, MNR-
1, MNR-2, MDH-3, SDH, 6PGD-1, 6PGD-2, TPI-1, TPI-2, SOD, and A AT.
Morpholine citrate gels were stained for PGI-1, PGI-2, PGM-1, PGM-2, MNR-1,

MNR-2, MDH-3, PGD-1, PGD-2, and SDH. Lithium borate gels were stained for
PGI-1, PGI-2, TPI-1, TPI-2, SOD, AAT, and ADH. Enzymes were stained using
minor modifications of the protocols summarized in Vallejos (1983) and Soltis et al
(1983). Data were collected as individual genotypes based on relative mobility of
enzymes. Genotypic deviations from Hardy-Weinberg expectations were calculated
using a chi-square test to compare expected and observed frequencies of all classes of
homozygotes and heterozygotes. Descriptive statistics were calculated for the number
of alleles per locus (and per polymorphic locus), polymorphism, observed and
expected heterozygosity, and clonality. Clonality values were derived by preparing a
ratio of unique genotypes for each species to the total number of individuals sampled.

Pollination Studies
Caged treatments in population xl all failed to set fruit, as did the population as
a whole. In fact, most populations of the putative hybrid A. x-floribundum failed to set
fruit during the years 1993 through 1996. Follicles were found in small numbers in
populations of the putative parental taxa. Crude estimates of reproductive success were
made by dividing the actual or estimated number of follicles in each population by the
estimated number of flowers. Reproductive success of all populations studied ranged
between zero and 0.017% (Table 3.1). The highest success that I have observed (18%)
was in an A. androsaemifolium population in lower Bear Creek.
Extensive collections of Apocvnum-visiting insects by Robertson (1928),
Wilson et al (1958), and myself (Appendix) indicate that little or no pollen is carried by
floral visitors. Robertson noted pollen on various appendages of visitors, but
apparently did not microscopically analyze it to prove that it was from Apocvnum. The
persistent tetrads of spheroidal Apocvnum pollen grains are distinctive. No such pollen
was found in my examination of 391 insect visitors representing about 176 species.
Many visitors, especially bees and some beetles, carried abundant pollen from other
plants. Apocvnum x-floribundum is by far the most heavily visited of the three taxa, in
terms of both numbers and diversity. I have recorded 163 species from this taxon
alone, mostly from a single population. Apocvnum androsaemifolium also attracts a
diverse community, about 71 recorded species, but in smaller numbers, while A.
cannabinum attracts a very large number of insects representing only about 25 species.

Table 3.1 Estimated Percent Fruit Set for Apocvnum Populations.
Pop. # Flowers per ramet Plants per m2 Population area (m2) Follicles per pop. Percent fruit set
ci 250 12 160 80 0.017
c2 700 23 900 70 < 0.001
c3 220 6 40 9 < 0.001
xl 250 40 300 0 0.000
x2 510 25 140 15 < 0.001
x3 450 40 320 7 < 0.001
x4 1100 25 25 4 < 0.001
x5 230 20 35 0 0.000
al 50 3 150 0 0.000
a2 200 12 100 27 0.011
a3 300 20 90 76 0.014

Figure 3.1 illustrates the approximate intersection of the floral visitor sets among the
three taxa.
My close observation of bees representing many families of Hymenoptera, and
of flies representing several families of Diptera, indicates that on Apocvnum flowers
bees and flies are nectar feeders and seem unconcerned with access to pollen.
Butterflies perch on the cymes and direct their probosces along the inner walls of the
petals. They appear to be nectar thieves. Several predacious insect species visit
flowering Apocvnum patches in search of prey, but their incidental contact with anthers
does not bring them into contact with pollen.
Woodson (1930) discussed an hypothesis that flies and bees that become
trapped by the sagittate bases of the anthers might contact pollen as they struggle to
escape. If they do escape, this pollen might be transferred to other flowers. Woodson
offered no evidence in support of this hypothesis. I found several examples of insects
that had died after being trapped thus in flowers, but there is no pollen available at the
bases of the anthers, and the trapped insects that I found (all bees and flies) did not
carry pollen. The absence of Apocvnum pollen on the insects examined indicates that
these plants are not mutualists but "keystone prey."
Attempts to artificially fertilize flowers in the field were unsuccessful due to the
small size and fragility of the flowers, the necessity of damaging flowers while
exposing the stigmata, and the inherent unlikelihood of fruit set in the genus. Often the
style breaks off when an anther is removed, and even when the flower seems
undamaged in this process, the style may be broken. Due to the low fruit set in general
in the genus, huge numbers of flowers would have to be artificially pollinated in order
to obtain results. From the thirty attempts that seemed not to harm the flowers, no fruit
resulted. Germination experiments revealed that seeds from all taxa are viable (Table
3.2), the rates ranging from 20% to 94%, with the two A. x-floribundum populations
exhibiting the poorest germination.

Figure 3.1 Venn Diagram Illustrating the Intersection of Floral Visitor Communities
Among the Three Taxa of Apocvnum Examined in this Study.

Table 3.2 Germination Percentages of Selected Apocvnum Populations. N = the
number of seeds tested from each population.
Population % germination N
c2 86 100
a2 94 50
a3 76 58
x2 20 41
x3 65 17

Morphological studies
Five of the seven morphological characters studied provide support for the
proposed hybrid origin of A. x-floribundum (Tables 3.3 -3.5 and Figs. 3.3 3.5).
Only leaf length and leaf length-to-width ratios are not useful in separating the taxa, or
are not suggestive of hybridization. One vegetative character studied, plant height, is
significantly different among the taxa, with A. x-floribundum intermediate (Fig. 3.5).
All of the reproductive characters differed significantly among the taxa. The
ratio of petal length to sepal length, long used to identify Apocvnum taxa in the field, is
a demonstrably good diagnostic character. Tukey's test indicates that all three taxa are
separable in this way, A. x-floribundum being intermediate. Petal length taken alone is
an even better diagnostic character (Fig. 3.4), the differences between taxa being highly
significant, with A. x-floribundum intermediate to the putative parental species.
The putative parent species have large follicles compared with the dimuntive
fruits of A. x-floribundum (Fig. 3.3). Seed numbers and size are also distinctive, with
A. androsaemifolium producing huge numbers of small seeds compared with A.
cannabinum's smaller number of large seeds. Apocvnum x-floribundum produces a
still smaller number of small seeds. The Kruskal-Wallis test indicates that seed number
differences are significant (p = 0.000, Table 3.4 and Fig. 3.2); ANOVA's and Tukey's
tests indicate that seed sizes are also statistically different except between A.
androsaemifolium and A. x-floribundum (Table 3.3).

Table 3.3 ANOVA and Tukeys Test Results of Analysis of Morphological Characters
Among Apocvnum cannabinum. A. x-floribundum. and A. androsaemifolium.

Petal Length 1 2,177 160.58 398.5 0.000
Petal Length/Sepal Length 1 2,177 1.75 89.3 < 0.001
Follicle Length 1 2,147 444.58 75.90 < 0.001
Seed Length 1 2,126 121.20 184.32 0.000
Plant Height1 2,267 121,000.00 296.00 0.000
Leaf Length 2 2,177 64.30 19.15 < 0.001
Leaf Length/Leaf Width 2 2,177 0.84 48.11 < 0.001
Results of Tukeys Test:
1 Indicates that all taxa differ significantly from each other.
2 Indicates that A. cannabinum and A. x-floribundum do not differ significantly.

Table 3.4. Kruskal-Wallis Test Results for Seed Number per Follicle in Apocvnum
Species. Box-and-whisker plots are shown in Figure 3.2.
Mean Std. Dev. Median 25 & 75 Percentiles
A. cannabinum 42.2 22.5 38.5 24.3 60.8
A. x-floribundum 15.6 7.1 13.0 10.0 20.8
A. androsaemifolium 140.5 94.0 111.0 72.31 72.5
H (test value) = 65.9
Degree of Freedom = 2
Probability = 0.000

Figure 3.2. Box-and-Whisker Diagrams of Seed Number Data for Apocvnum Species.
Values of medians and hinges are shown in Table 3.4.
A. androsaemifolium
A. cannabinum
A. x-floribundum
Number of seeds per follicle

Table 3.5 Mean, Standard Deviation, and Range Values for Morphological Measures
from Apocvnum Species.
A. cannabinum A. x-floribundum A. androsaemifolium
petal length (mm) 4.0 0.4 (3.5 4.6) 5.6 0.8 (4.0 7.7) 7.3 0.8 (5.5 9.5)
petal/sepal length ratio 1.8 0.3 (1.4-2.3) 2.3 0.5 (1.2-3.5) 3.7 1.1 (1.9 6.7)
follicle length (cm) 10.4 3.0 (6.0- 19.5) 4.9 1.2 (1.9 7.1) 8.3 2.4 (3.9 14.3)
seeds per follicle 42 23 (6-112) 16 7 (5 33) 141 94 (23 -381)
seed length (mm) 5.5 0.9 (2.8 8.8) 2.4 0.4 (1.5 3.5) 2.5 0.5 (1.3 3.2)
plant height (cm) 139 23 (98 195) 105 23 (52- 164) 66 13 (45 92)
leaf 1/w ratio 1.8 1.1 (1.8 3.4) 2.0 1.0 (1.4-4.1) 2.1 1.1 (1.5 2.3)

Figure 3.3 Follicle Length in Apocvnum Populations. Bars represent the mean (white
line), the mean plus and minus one standard deviation (black bar), and the range (gray
Population No.

Figure 3.4 Petal Length in Apocvnum Populations. Bars represent the mean (white
line), the mean plus and minus one standard deviation (black bar), and the range (gray
1 23123451 23
cannabinum x-floribundum androsaamifolium
Population No.

Figure 3.5 Plant Height in Apocvnum Populations. Bars represent the mean (white
line), the mean plus and minus one standard deviation (black bar), and the range (gray
1231 23451 23
cannabinum x-floribundum andnosaemifolium
Population No.

Population Genetics
Three loci (PGI-1, MNR-2, and 6PGD-1) were monomorphic and therefore
uninformative; six loci (PGM-2, MDH-3, SDH, 6PGD-2, TPI-1, and AAT) were fixed
for different alleles in the parent species; and two loci (PGI-2 and SOD-1) exhibited
highly disparate allele frequencies (Tables 3.6 and 3.7).
The A. cannabinum populations examined all had very low levels of
heterozygosity (H0bs = 0.063) and polymorphism (P = 20%). The species mean
number of alleles per locus (A) was 1.19, with no more than two alleles found at any
polymorphic locus. The largest stand, Monument Creek (c2), exhibited no variation at
11 loci, with one locus (PGM-2) having five out of 24 sampled ramets heterozygous,
and another locus (SOD) with the same five ramets being the only homozygotes. These
loci are in Hardy-Weinberg equilibrium; however, because only two genotypes were
represented, and because the five distinctive ramets were adjacent to one another along
the transect, it is probable that the sample represented only two genets. The second
largest stand, cl, was monomorphic for all loci resolved. The Fountain Creek at
Nevada Avenue population (c3) showed an H0bs = 0.08 and P = 15%. Of the two
polymorphic loci, only PGM-2 was out of Hardy-Weinberg equilibrium (X2 =16.3, p
< 0.001), with an overabundance of heterozygotes and the complete absence of one
homozygote class (aa).
Apocvnum androsaemifolium showed similar trends, with species mean H0bs =
0.063, P = 38% and A = 1.44. The Larkspur population (al) had three polymorphic
loci (PGI-1, SOD-1, and TPI-2). In the South Cheyenne Canyon population (a3), the
only polymorphic locus (SDH) was heterozygous for all twenty-four individuals, and
therefore grossly out of Hardy-Weinberg equilibrium (X2 = 24.0, p < 0.001), again

suggesting that the population is a single genet. The Bear Creek population (a2)
showed heterozygotes at two loci (PGI-2 and SDH). These loci were in Hardy-
Weinberg equilibrium, but as in all previous cases the heterozygotes were adjacent
ramets on the transect.
The A. x-floribundum populations, in spite of their great morphological
diversity, were very similar genetically. Relative to A. androsaemifolium and A.
cannabinum. heterozygosity was elevated in all populations (H0bs = 0.657) and
polymorphic loci accounted for 63% of the loci studied. Number of alleles per locus
averaged 1.77, considerably higher than the other populations. Four of the loci that
were diagnostically fixed for A. cannabinum and A. androsaemifolium (MDH-3, PGD-
2, TPI-1, and AAT) were entirely heterozygous in A. x-floribundum. with each allele
also present in one or the other of the other two taxa, the only exception being the
appearance of a "c" allele in population x2. Additionally, four supportive loci (PGI-2,
PGM-2, SDH, and SOD-1), at which allele frequencies were highly disparate between
the parental taxa, were also 100% heterozygous for the alleles most common in A.
cannabinum and A. androsaemifolium. At every locus at which A. x-floribundum
populations were heterozygous, X2 equals 10.0 (p < 0.01), indicating significant
deviations from Hardy-Weinberg equilibrium.

TABLE 3.6 Allozyme Data for Apocvnum Populations. A = number of alleles per
locus, Ap = number of alleles per polymorphic locus, P = polymorphism, H0 =
observed heterozygosity, He = expected heterozygosity, and G = clonality as the ratio
of unique genotypes to number sampled.
Pop. A A? P H0 He G
A. cannabinum
cl 1.00 NA 0.00 0.000 0.000 0.042
c2 1.14 2.0 0.14 0.074 0.048 0.083
c3 1.15 2.0 0.15 0.077 0.042 0.083
A. x-floribundum
xl 1.54 2.0 0.54 0.538 0.269 0.100
x2 1.54 2.2 0.54 0.538 0.269 0.100
x3 1.57 2.0 0.54 0.571 0.286 0.100
x4 1.47 2.4 0.54 0.467 0.233 0.100
x5 1.38 2.0 0.54 0.538 0.269 0.100
A. androsaemifolium
al 1.38 2.7 0.23 0.125 0.118 0.208
a2 1.20 2.0 0.20 0.047 0.042 0.125
a3 1.08 2.0 0.08 0.077 0.038 0.042
means for species:
A. cannabinum 1.19 2.0 0.20 0.063 0.110 0.069
A. x-floribundum 1.77 2.6 0.63 0.657 0.316 0.060
A. androsaemifolium 1.44 2.5 0.38 0.063 0.051 0.042

TABLE 3.7 Allele Frequencies for Diagnostic and Supportive Loci for Apocvnum
Populations. Deviations from Hardy-Weinberg with significant X2 scores are
sp.: cannabinum ------ x-floribundum------- androsaemifolium
No.: N= : cl 24 c2 24 c3 20 xl 10 x2 10 x3 10 x4 10 x5 10 al 24 a2 24 a3 24
Loci PGI-2 a 0 0 0 0 0 0 0 0 0.042 0. 0
b 0 0 0 0 0.501 0.501 0 0.501 0.23 0.90 1.00
c 0 0 0 0 0 0 0 0 0.58 0 0
d 1.00 1.00 0.98 1.00 0 0.50 1.00 0.50 0.15 0.10 0
e 0 0 0.02 0 0.50 0 0 0 0 0
PGM-2 a 0 0 0 0.501 0.501 0.501 0.501 0.501 1.00 1.00 1.00
b 1.00 0.90 0.533 0.50 0.50 0.50 0.50 0.50 0 0 0
c 0 0.10 0.47 0 0 0 0 0 0 0 0
MDH-3 a 0 0 0 0.501 0.501 0.501 0.501 0.501 1.00 1.00 1.00
b 1.00 1.00 1.00 0.50 0.50 0.50 0.50 0 0 0
c 0 0 0 0 0 0 0 0.50 0 0 0
SDH a 0 0 0 0 0 0 0 0 0 0.85 0.504
b 0 0 0 0.501 0.501 0.501 0.50 0.501 1.00 0.15 0.50
c 1.00 1.00 1.00 0.50 0.50 0.50 0.50 0.50 0 0 0
6PGD-2 a 0 0 0 0.501 0.501 0.501 0.501 0.501 1.00 1.00 1.00
b 1.00 1.00 1.00 0.50 0.50 0.50 0.50 0.50 0 0 0

TABLE 3.7 Continued.
sp.: cannabinum x-floribundum androsaemifolium
No.: cl c2 c3 xll X21 X31 X41 X51 al a2 a3
N= : 24 24 20 10 10 10 10 10 24 24 24
Loci SOD-1
a 1.00 0.605 1.00 0.501 0 0 0.501 0 0.316 0 0
b TPI-1 0 0.40 0 0.50 1.00 1.00 0.50 1.00 0.69 1.00 1.00
a 0.00 0.00 0.00 0.501 0.501 0.501 0.501 0.501 1.00 1.00 1.00
b AAT 1.00 1.00 1.00 0.50 0.50 0.50 0.50 0.50 0.00 0.00 0.00
a 1.00 1.00 1.00 0.501 0.501 0.501 0.501 0.501 0.00 0.00 0.00
b 0.00 0.00 0.00 0.50 0.50 0.50 0.50 0.50 1.00 1.00 1.00
1 X2 = 10.0, p < 0.01
2 X2 = 41.2, p< 0.001
3 X2 = 16.3, p< 0.001
4 X2 = 24.0, p < 0.001
5 X2= 11.0, p< 0.001
6 X2 = 4.64, p < 0.05

Speciation by hybridization is well documented (Arnold 1992, Abbot 1992,
Rieseberg 1995, Broyles et al 1996), although there have been many misconceptions
about hybrids. Rieseberg (1995) noted that hybridization does not always result in
intermediate morphological characters. In a study of hybridization in Asclepias. Wyatt
and Broyles (1992) showed that morphological analysis of reproductive characters
alone did not detect hybrids, but that vegetative characters and isozyme evidence could
easily identify hybrids. Molecular evidence was far more reliable in detection of
hybrids. The present study indicates that both morphological and molecular evidence
are strongly indicative of hybridization in Apocvnum.
Apocvnum cannabinum and A. androsaemifolium are fixed for alternate alleles
at six loci for which A. x-floribundum is heterozygous. Based upon this sample,
allozyme data leave no doubt that the A. x-floribundum populations are hybrids of A.
androsaemifolium and A. cannabinum. Even if the six parental populations are
interpreted as ten clonal individuals, which is entirely possible (see below), it is highly
unlikely that all five "individual" intermediates would be almost exclusively
heterozygous, while the other ten "individuals" would be homozygous. Furthermore,
all of the intermediates are probably Fi hybrids. Homozygotes at PGI-2 ("dd") and
SOD-1 ("bb") in some A. x-floribundum populations could result from backcrossing
with parent species, but these could just as easily result from Fj hybridization. Thus,
there is no compelling evidence for Woodson's (1930) hypothesis of "secondary
Analyses of morphological data also strongly support these conclusions.

Measures of follicle length, seed number, and seed size are consistent with partial
hybrid sterility in A. x-floribundum. Petal length, petal-to-sepal length ratios, and plant
height all place A. x-floribundum as an intermediate taxon. In general, reproductive
characters provide more definitive evidence than do vegetative characters, although
plant height also clearly places A. x-floribundum as an intermediate taxon. Characters
used in field guides and floras (petal-to-sepal ratios and plant height) appear to be
reliable characters for field determinations. Petal length taken alone, however, is an
even better indicator.
Allozyme evidence suggests that clonal growth is common and important. As
Cook (1983) points out, cloning should be regarded as growth rather than
reproduction. None of the populations examined had more than four genotypes (G =
0.17) among the samples, and seven of the 11 populations revealed only one
(0.04 < G < 0.09). This is consistent with sampling ramets of large genets. The fact
that all variants found in these populations lay adjacent to one another along their
respective transects is further evidence for clonality. That vegetative reproduction is
operative here is most strongly indicated by the fact that when heterozygosity is
encountered [for example, SDH in South Cheyenne Canyon (a3) A. androsaemifolium.
PGM-2 in Upper Fountain Creek (c3) A. cannabinum. and several loci in all of the A.
x-floribundum populations] it occurs at or near 100%, and sometimes at more than one
locus. The extremely high X2 values associated with long series of heterozygotes
verify beyond a doubt that all of these populations consist of one or very few large
That Apocvnum spreads by cloning is well known. Excavating two or more
adjacent shoots from three different populations of A. androsaemifolium and A. X;
floribundum revealed that they were connected underground. The connecting rhizomes
lay 15 32 cm beneath the surface, although on steep, unstable gravel scree they may

be as shallow as 10 cm. Even in populations such as a3, in which sexual reproduction
is demonstrably successful (Table 3.1) and the seeds are viable (Table 3.2), the
population is genetically identical at the loci studied.
Clonality, however, cannot account for the extremely low heterozygosity found
in the parental species. The fact that so many loci are fixed for the same alleles in these
populations suggests a history of inbreeding, or possibly a severe bottleneck
somewhere in the history of these species. Because widely separated populations are
fixed for the same alleles, random genetic drift is an unlikely cause of this effect.
Overdominance selection could account for the occasional cases of excess
heterozygotes, but the total eradication of both homozygotes is a less parsimonious
interpretation than the cloning of a highly inbred species, particularly within the context
of the floral morphology characteristic of the genus.
Wright's F values are not presented because they are misleading in samples
taken from a single genet. When a locus is 100% heterozygous (i. e., where p =q =
0.5), F = -1, indicating obligate outcrossing. But a single genet with a heterozygous
locus will also produce an F value of -1 at that locus. Observations of flower
morphology and floral visitors indicate that obligate outcrossing is an untenable
conclusion. Morphological studies, observation and examination of floral visitors, and
allozyme data all support a history of inbreeding or a severe bottleneck.
It is apparent that the Apocvnum taxa studied herein represent two easily
characterized parent species, and a series of Fj hybrid populations that are genetically
very similar. The extremely low fruit set in these hybrids, lower even than in the parent
species, may be attributed to partial hybrid sterility. The rarity of backcrossed
individuals is probably an artifact of this sterility, allotopy among species, and the
apparent rarity of hybridization events. It is interesting that seeds that are occasionally
produced by some hybrids show some viability (Table 3.2).
Due to the clonal nature of these species, and the fact that the inflorescences of

ramets tend to flower simultaneously, each clone might be interpreted as a single
enormous inflorescence. Nakamura et al (1989) hypothesized that very large
inflorescences tend to reduce male reproductive success due to a high likelihood of self
pollination, and that selection might therefore be expected to favor a reduction in the
size of the inflorescence to enhance outcrossing. If this is so, clonal growth in
Apocvnum may be overpowering selection forces that would reduce inflorescence size.
It is also possible that the floral structure, which severely limits pollen accessibility, is a
response to specialized pollinators that fly long distances to disseminate pollen. This,
too, would enhance male reproductive success.
The problem of pollen vectors remains an inscrutable mystery, wherein must lie
the answer to the mating conundrum in Apocvnum. Vectors of pollen between A.
androsaemifolium and A. cannabinum would have to be highly motile and move great
distances. My collections and observations of floral visitors, while not exhaustive,
have thus far only identified seven species (out of the 176 recognized) on both parent
Apocvnum species. These seven were honeybees (Apis meiifera), a Halictid bee
(genus Lasioglossuml. four species of skipper butterflies (genera Epargvreus. Polites.
Oarisma. and Pmma), and the common European cabbage white butterfly (Pieris
rapae). Of these, only honeybees and cabbage whites are ubiquitous and wide-ranging.
Most Halictid bees are social to some degree, and are known to feed on nectar and
pollen (Wilson 1971). But their tiny size suggests that they do not travel as far as
honeybees in foraging. According to Scott (1986), Epargvreus and Polites are quasi-
territorial, the males adopting perches from which to spot and pursue females. These
would tend to be localized species. By contrast, Oarisma and Piruna males are
patrollers, possibly prone to moving greater distances along paths or in swales. In any
case, none of these seven was found with Apocvnum pollen.
There is a much greater overlap in the floral visitor communities between
hybrids and parental species than between the two parental species (Fig. 3.1). Thus, if

cross pollination were occurring, backcrossed populations would be expected.
Anderson's (1936) assertion that hybridization is very common must be
questioned. The present research indicates that hybridization might be rare, but that the
viability of hybrids is very good. There is evidence that once a hybrid becomes
established it may persist for many decades and grow into large, discontinuous stands,
creating the illusion of multiple hybridization events. The A. x-floribundum population
in Palmer Park (x3) spans a large, improved gravel road that was built several decades
ago, as well as a heavily eroded stream bed that is periodically scoured by flash
flooding. The population, which appears to be genetically homogeneous, must predate
both the road and the deep stream bed. The illusion of frequency of hybrid events
would be further enhanced if hybrid progeny from a single outcross produced
phenotypic variants as they crept across environmental gradients.
Broyles et al (1996) demonstrated that Fi hybridization in milkweeds may be
much rarer than appears, even in taxa in which hybrids are readily diagnosed. They
found that in one site where Asclepias exaltata and A. svriaca are sympatric, only about
0.4% of the stems represented hybrids. They also found that fruits that contained
hybrid seeds had significantly depressed seed set, a result similar to my findings in
Apocvnum (Table 3.4).
It is not surprising that over 80 species were originally described in this genus.
Each hybridization event from the seven currently recognized species may produce a
unique genotype that grows clonally and presents itself as a distinct taxonomic entity.
Despite the curious floral morphology and the lack of pollen on potential pollinators,
morphology and allozymes together show that A. x-floribundum is a result of
hybridization between A. cannabinum and A. androseamifolium.

Table is organized by number of species per family in each insect order on each species
of Apocvnum. All specimens are pinned and in the collection of the author.
Apocvnum cannabinum
Total species.............25

Apocvnum androsaemifolium
Total species..............71

Apocvnum x-floribundum
unidentified nymphs....1

Total species............163

Abbot, R. J. 1992. Plant invasions, interspecific hybridization, and the evolution of
new plant taxa. Trends in Ecology and Evolution. 7: 401-404.
Anderson, E. 1936. An experimental study of hybridization in the genus Apocvnum.
Annals of the Missouri Botanical Garden. 23: 159-169.
Arnold, M. L. 1992. Natural hybridization as an evolutionary process. Annual Review
of Ecology and Svstematics. 23: 237-261.
Bookman, S. S. 1984. Evidence for selective fruit production in Asclepias. Evolution.
38: 72-86.
Broyles, S. B., Vail, C., and Sherman-Broyles, S. 1996. Pollination genetics of
hybridization in sympatric populations of Asclepias exaltata and A. svriaca
(Asclepiadaceae). American Journal of Botany. 83: 1580-1584.
Broyles, S. B. and Wyatt, R. 1991. Paternity analysis in a natural population of
Asclepias exaltata: Multiple paternity, functional gender, and the "pollen
donation hypothesis." Evolution. 44: 1454-1468.

Cheliak, W. M. and Pitel, J. A. 1984. Techniques for starch gel electrophoresis of
enzymes from forest tree species. Information Report Pi X-42. Petawawa
National Forestry Institute, Canadian National Forestry Service, Petawawa,
Clayton, J. W. and Tretiak, D. N. 1972. Amine-citrate buffers for pH control in
starch gel electrophoresis. Journal of the Fisheries Research Board of
Canada 29: 1169-1172.
Cook, R. E. 1983. Clonal plant populations. American Scientist. 71:244-253.
Cronquist, A. 1981. An integrated system of classification of flowering plants.
Columbia University Press, New York. 1262 pp.
Darlington, C. D. and Wylie, A. P. 1956. A chromosome atlas of flowering plants.
Allen and Unwin, London. 519 pp.
Endress, M. E. 1995. A morphological cladistic study of Apocynaceae: trends in
character evolution within a broadened familial circumscription. 1995
Annual Meeting of the Botanical Society of America and the American
Institute of Biological Sciences, San Diego, CA, August 6-10, 1995.
American Journal of Botany. 82(6 suppl.): 127.
Harrington, H. D. 1954. Manual of the Plants of Colorado. Sage Books, Denver,
CO. 666 pp.
McGregor, R. L., Barkley, T. M., Brooks, R. E., and Schofield, E. K. 1986. Flora
of the Great Plains. University of Kansas Press, Lawrence, KS.

Michener, C. D., McGinley, R. J., and Danforth, B. N. 1994. The Bee Genera of
North and Central America (Hvmenoptera: Apoideal Smithsonian
Institution Press, Washington, DC, 209 pp.
Nakamura, R. R., Stanton, M. L., and Mazer, S. J. 1989. Effects of mate size and
mate number on male reproductive success in plants. Ecology. 70: 71-76.
Rieseberg, L. H. 1995. The role of hybridization in evolution: old wine in new skins.
American Journal of Botany. 82: 944-953.
Robertson, C. 1928. Flowers and insects, lists of visitors of four hundred and fifty-
three flowers. Science Press Printing Co., Lancaster, PA.
Rydberg, P. A. 1932. Flora of the prairies and plains of central North America. New
York Botanical Garden, New York, NY.
Scott, J. A. 1986. The Butterflies of North America. Stanford University Press,
Stanford, CA. 583 pp.
Soltis, D. E., Haufler, C. H., Darrow, D. C., and Gastony, G. H. 1983. Starch gel
electrophoresis of ferns: a compilation of grinding buffers, gel and electrode
buffers, and staining schedules. American Fern Journal. 73: 9 27.
Vallejos, C. E. 1983. Enzyme activity staining. In S. D. Tanksley and T. J. Orton
leds.l Isozvmes in plant genetics and breeding. Part A. 469-515. Elsevier,

Weber, W. A. 1990. Colorado Flora: Eastern Slope. University Press of Colorado,
Niwot, CO. 396 pp.
Werth, C. R. 1985. Implementing an isozyme laboratory at the field station. Virginia
Journal of Science. 36: 53-72.
Wilson, E. O. 1971. The Insect Societies. Belknap Press of Harvard University
Press, Cambridge, MA. 548 pp.
Wilson, W. T., Moffett, J. O., and Harrington, H. D. 1958. Nectar and pollen plants
of Colorado. Bulletin 503-S. Colorado State University Experiment Station,
Fort Collins, CO. 72 pp.
Woodson, R. E. 1930. Studies in the Apocynaceae. I. A critical study of the
Apocynoideae (with special reference to the genus Apocvnum).-Annals of
the Missouri Botanical Garden. 17: 1-230.
Wyatt, R. and Broyles, S. B. 1992. North American Asclepias. III. Isozyme
evidence. Systematic Botany. 17: 640-647