IMPACT OF PLOIDY ON MORPHOLOGICAL VARIATION IN ARIZONA PHLOX, PHLOX AMABILIS
(POLEMONIACEAE)
By
Matthew Thomas Chansler

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Plant Biology—Master of Science
2015

ABSTRACT
IMPACT OF PLOIDY ON MORPHOLOGICAL VARIATION IN ARIZONA PHLOX, PHLOX AMABILIS
(POLEMONIACEAE)
By
Matthew Thomas Chansler
Polyploidy is an important factor in the evolution and ecology of flowering plants. A better
understanding of the kind and degree of morphological differentiation among ploidy levels within a
species can help explain further how polyploidy affects biodiversity. How widespread is the impact of
ploidy across the phenotype of a species? Which aspects of morphology vary, and do they vary
consistently? How does ploidy relate to overall morphological diversity? Do ploidy levels have
detectable phenotypic profiles? Finally, are there morphological differences between populations,
potentially due to environment or evolutionary changes since formation apparent in natural
populations? I assessed morphological variation within Arizona phlox, Phlox amabilis. This species of
conservation concern is endemic to Arizona, and prior work has detected diploid, tetraploid, and
hexaploid populations. I sampled 11 populations of P. amabilis, covering a large portion of the species’
range. A wide array of morphological features, including characters that described cell size, overall
habit, leaf dimensions, and floral dimensions, were measured for up to 25 plants at each population.
Significant differences were detected in 15 out of 27 characters using mixed GLM. A large amount of
overall morphological variation is explained by the differences between ploidy levels, and each ploidy
level can be described by a specific multivariate phenotype with 95% accuracy. Finally, although overall
structuring was influenced by ploidy, differences among populations still contributed a high degree of
variation in the morphospace of Phlox amabilis. This morphological assessment will be integrated with
ecological and genetic data to build a more complete understanding of the interplay between these
factors and ploidy in Phlox amabilis.

ACKNOWLEDGEMENTS
I would like to thank the members of my committee: Alan Prather (PI, MSU), Carolyn Ferguson
(KSU), and Barbara Lundrigan (MSU) for their guidance throughout experimental design, sampling,
analysis, and writing. Additionally, Shannon Fehlberg and Kevin Fehlberg (Desert Botanical Garden,
Phoenix, AZ), and Mark Mayfield (KSU) deserve special mention for their contribution to field support
and cytotypification. Additional cytotypification was done by Bethany Wright and Jordan Mcguinn. A
big thank you goes to Anna Busch, Sam Rachwal, Sam Stockwell, Alex Vantill, Sarah Pelkey, Mike
Niezgoda, and Josh Polito, all undergraduates who helped me immensely during data collection. For
analysis, thank you to Moslem Ladoni for his SAS expertise. Thank you to the National Science
Foundation and the Michigan State University Herbarium Endowment for funding. I gratefully
acknowledge permitting assistance by personnel of USDA Forest Service Region 3. Finally, I thank my
friends and family for their constant support and advice during this journey.

iii

TABLE OF CONTENTS
LIST OF TABLES .............................................................................................................................................. v
LIST OF FIGURES ........................................................................................................................................... vi
INTRODUCTION ............................................................................................................................................. 1
How broad is the impact of ploidy across phenotype in flowering plants? .............................................. 3
Is ploidy diversity associated with overall morphological diversity? ........................................................ 5
Do different ploidy levels have different multivariate phenotypes? ......................................................... 6
Are populations different morphologically? ............................................................................................. 7
Focal taxon ................................................................................................................................................ 8
Objectives .................................................................................................................................................. 9
MATERIAL AND METHODS .......................................................................................................................... 10
Sampling and field measurement ........................................................................................................... 10
Laboratory measurement ....................................................................................................................... 12
Univariate diagnostics ............................................................................................................................ 15
Univariate analysis .................................................................................................................................. 15
Principle components analysis ................................................................................................................ 16
Discriminant function analysis ................................................................................................................ 17
RESULTS ...................................................................................................................................................... 19
DISCUSSION................................................................................................................................................. 31
The effects of ploidy were widespread across the phenotype of naturally occurring populations of
Phlox amabilis plants .............................................................................................................................. 31
Ploidy explained a considerable amount of total phenotypic diversity .................................................. 32
Each ploidy level had a distinctive multivariate phenotype .................................................................... 33
Morphological differences among Phlox amabilis populations were apparent, but grouping is primarily
by ploidy .................................................................................................................................................. 35
LITERATURE CITED ...................................................................................................................................... 37

iv

LIST OF TABLES
Table 1. Sites of Phlox amabilis sampled for morphology in May 2013………………………………………….

10

Table 2. Sets of correlated characters and whether they were removed or retained for principal
components analysis.…………………………………………………………………………………………………………………….

16

Table 3. Results of univariate analyses of all morphological characters and their association with
ploidy in natural populations of Phlox amabilis diploids, tetraploids and hexaploids…………………….

20

Table 4. Eigenvectors indicating correlation of each character with principal component axes 1
and 2 for diploid, tetraploid, and hexaploid Phlox amabilis individuals…...…………………………………….

25

Table 5. F-values for 14 characters that contributed significantly (p<0.05) to overall
discrimination between ploidy levels in Phlox amabilis in a stepwise discriminant analysis……………

26

Table 6. Total canonical structure associated with the discriminant function which maximized
differences between diploid, tetraploid and hexaploid Phlox amabilis…………………………………………..

28

Table 7. Results of linear discriminant analysis, using cross-validation to assess accuracy of the
canonical discriminant function used to classify ploidy levels of 253 Phlox amabilis individuals…….

28

Table 8. Eigenvectors indicating correlation of each character to PC axes 1 and 2 for diploid,
tetraploid, and hexaploid Phlox amabilis populations….…………………………………………………………………

30

v

LIST OF FIGURES
Figure 1. Geographic locations of populations.......………………………………………………………………………..

11

Figure 2. Example landmarked corolla lobe of P. amabilis.…………………………………………………………….

14

Figure 3. Squared Mahalanobis distance for each Phlox amabilis individual plotted against the
predicted Chi-square distribution..…………………………………………………………………………………………………

17

Figure 4. Distributions of measurements of characters found to differ significantly between at
least two ploidy levels across natural populations of Phlox amabilis………………………………………………

21

Figure 5. Principle component analysis of morphological data collected from natural populations
of diploid, tetraploid, and hexaploid Phlox amabilis plants…………………………………………………………….

24

Figure 6. Individual Phlox amabilis plants plotted across canonical variates (CVs), generated using
14 characters retained in the stepwise analysis……………………………………………………………………………..

27

Figure 7. Populations of Phlox amabilis plants plotted across principal components (PCs) using
uncorrelated characters for diploids, tetraploids, and hexaploids.…………………………………………………

29

vi

INTRODUCTION
Since the discovery of polyploidy in the early 1900s and the effort made by Stebbins to elucidate
its origins, incidence, and evolution (Stebbins, 1950, 1971), our understanding of polyploidy has
developed considerably (Tate et al., 2005; Soltis et al., 2014). One of the most important milestones has
been the realization that variation in ploidy is more common than previously assumed (reviews by Soltis
et al., 2007; Wood et al., 2009). Otto & Whitton (2000) assert that between 2 and 4% of speciation
events are associated with polyploidization. Recent research indicates that the relationship between
polyploidization and speciation in vascular plants is stronger, with 12-13% of angiosperm species and
17% of fern species containing ploidy level variation (Wood et al., 2009). Rice et al. (2015) generated an
even higher estimate; 16% of angiosperm species have multiple ploidy levels by their analysis, with a
significant portion of chromosome variation explained by autopolyploidization.
Neopolyploidy results in an instantaneous increase in genome size (Bennett and Leitch, 2005)
and increasing genome size has a significant positive relationship with cell size (Müntzing, 1936;
Stebbins, 1950; Cavalier-Smith, 1978, 2005; Gregory, 2001; Beaulieu et al., 2008). A large body of
research indicates that guard cell length and pollen grain diameter, direct measurements of cell size,
increase with increasing ploidy (reviewed in Otto and Whitton, 2000; Tate et al., 2005; Beaulieu et al.,
2008), though naturally-occurring polyploids tend to have smaller cells than synthetic polyploids
(Butterfass, 1987).
Increased cell size in flowering plants can manifest as an overall increase in the size of
anatomical features; Stebbins called this the “Gigas Effect” (Stebbins, 1971), named after the Oenothera
lamarckiana Ser. mutant in which polyploidy was discovered (Lutz, 1907). A cell-size increase during
polyploidization is sometimes associated with differences in both vegetative and floral morphological
features between ploidy levels (Otto and Whitton, 2000; Mizukami, 2001; Tate et al., 2005; Beaulieu et
al., 2008). Measurements of vegetative traits, including both gross plant habit and specific tissues (e.g.
1

leaves), and floral traits (reviewed in Vamosi et al., 2007) have also been seen to increase in polyploids
(Otto and Whitton, 2000; Tate et al., 2005), with some specific exceptions (e.g. Mandáková and
Münzbergová, 2008; Španiel et al., 2008).
Morphological differentiation between plants of different ploidy levels may have ecological
impacts (Husband and Schemske, 1998; Johnson et al., 2003; Baack, 2005; Tate et al., 2005; Ståhlberg,
2009; Ramsey, 2011; Thompson et al., 2014). However, in other instances ploidy is not associated with
habitat differentiation (Suda et al., 2004; Halverson and Heard, 2008; Šingliarová et al., 2011; Mráz et
al., 2014; Visser and Molofsky, 2015). Common garden experiments have been used to directly show
whether ploidy has a direct impact on ecological adaptation (Ramsey, 2011) or not (Münzbergová, 2007;
Ramsey, 2011). Whether morphological differences between ploidy levels result from polyploidization
directly, phenotypic plasticity (Macdonald et al., 1988), or selection after the fact (Jordan et al., 2015) is
not always clear.
Drought tolerance can be an ecological impact of polyploidy (reviewed in Visser and Molofsky,
2015). Water use efficiency may interact with stomatal characters to influence niche differentiation
based on drought tolerance (van Laere et al., 2011; Manzaneda et al., 2012; Mráz et al., 2014), and has
been assessed in polyploid crops such as wheat (reviewed in Khazaei et al., 2009).
Morphological change as a result of polyploidy has the potential to alter pollinator interactions.
For instance, pollinators are able to distinguish between diploid and tetraploid fireweed Chamerion
angustifolium (L.) Holub (Husband and Schemske, 2000; Husband and Sabara, 2004; Kennedy et al.,
2006). Polyploidization was associated with both pollinator differences in autotetraploid Heuchera
grossulariifolia Rydb. and differential attack by herbivores between ploidy levels (reviewed in Douglas E.
Soltis, Soltis, & Tate, 2003). On the other hand, polyploidization and any associated morphological
changes were not associated with ecological differences in Heuchera cylindrica Douglas (Godsoe et al.,

2

2013). In some cases, whether ploidy influences morphology depends on the trait and the environment
(Segraves & Thompson 1999, Ståhlberg 2009).
Investigations that test the relationship between polyploidy and morphology in nature are
relatively uncommon. Balao et al (2011) were comprehensive in their collection of many characters
across several ploidy levels and natural populations. Ståhlberg (2009) analyzed 35 vegetative and
reproductive characters across two mixed diploid-tetraploid localities. Pettigrew et al (2012) analyzed
seven reproductive and micromorphological characters across a number of populations spanning most
of southern Africa. Examples of studies that combine these aspects with high sample size and characterlevel analysis are uncommon, however.

How broad is the impact of ploidy across phenotype in flowering plants?
Morphological characters that directly measure cell size, such as guard cell length (e.g. Nesom,
1983; Mishra, 1997; Buggs and Pannell, 2007; Khazaei et al., 2009; de las Mercedes Sosa et al., 2012;
Kim et al., 2012; Pettigrew et al., 2012; Marinho et al., 2014) and pollen grain size (e.g. de las Mercedes
Sosa et al., 2012; Pettigrew et al., 2012; Marinho et al., 2014) are almost always observed to increase
with ploidy. Beaulieu et al. (2008) found a positive relationship between guard and epidermal cell size,
and genome size across 101 species of angiosperms. Exceptions are rare, but they do occur: Tetraploid
species of Tarassa Phil. produced smaller pollen grains than their diploid counterparts (Tressens, 1970;
Tate and Simpson, 2004). This result was associated with loss of outcrossing in tetraploid species (Tate
and Simpson, 2004). Measurements of structure density are included here as semi-direct
measurements of cell size; larger cells combined with a fixed number of structures leads to higher
density. Stomatal density is probably the most frequently recorded measurement of this type, and it is
consistently observed to decrease with increasing ploidy in natural systems (e.g. Chen et al., 2009;
Khazaei et al., 2009; Pettigrew et al., 2012). Studies that measure trichome density are much rarer, but
in Buddleja macrostachya Benth., trichome density was significantly lower in dodecaploids than in
3

hexaploids (Chen et al., 2009), while it was higher in tetraploid compared to diploid Stemodia hyptoides
Cham. & Schltdl. (de las Mercedes Sosa et al., 2012).
A relationship between ploidy and overall plant habit has been detected for polyploid systems
in nature (Ståhlberg, 2009; Šingliarová et al., 2011). Plant height seems to have a complicated
relationship with ploidy: Sometimes, it increased with ploidy, (Müntzing, 1936; Ståhlberg, 2009), while in
other scenarios plants with higher ploidy were shorter (Buggs and Pannell, 2007; Fialová et al., 2014), or
no difference was detected at all (Treier et al., 2009). Treier et al (2009) also reported an interaction
between “geocytotype” and environment: In disturbed environments, North American tetraploid
Centaurea stoebe L. plants produced more stems than European tetraploid or diploid plants. Polyploid
plants have often been observed to produce fewer flowers than diploids (reviewed in Otto and Whitton,
2000; Segraves and Thompson, 1999; Husband and Schemske, 2000; Ståhlberg, 2009), although in
Spartina pectinata Link, number of spikelets per stem did not differ between tetraploids and hexaploids
(Kim et al., 2012).
Leaf dimensions are a commonly measured vegetative structure not associated with plant habit,
and often differ between ploidy levels. Longer (Kannenberg and Elliott, 1962; Sugiyama, 2005; Chen et
al., 2009; Ståhlberg, 2009; Li et al., 2010), broader (Chen et al., 2009; Ståhlberg, 2009; Li et al., 2010)
leaves with increasing ploidy level are commonly documented in natural populations.
Floral characters tend to increase in size between diploids and polyploids (e.g. Müntzing, 1936;
Kannenberg and Elliott, 1962; Brunsfeld and Johnson, 1990; Segraves and Thompson, 1999; Kennedy et
al., 2006; Vamosi et al., 2007; Ståhlberg, 2009; Pettigrew et al., 2012), but this is not always true (Tate
and Simpson, 2004; de las Mercedes Sosa et al., 2012). Notably, Kim et al.(2012) found no aspect of
reproductive anatomy that differed between tetraploid and hexaploid Spartina pectinata Link.
Ploidy appears to affect individual morphological features, particularly those which describe cell
size directly, in many genera and species. However, questions remain: In particular, little work has been

4

done to determine how multiple ploidy levels affect morphology in a taxon, in the context of broad
sampling across the range of the species.
Generally, results of morphological assessment of synthetic polyploids corroborate those found
in natural systems: polyploidy increases cell size and thus the size of many vegetative and floral features.
This is important because synthetic polyploid systems describe the contribution ploidy makes to
morphology while controlling for genetics and environment. In particular, direct measurements of cell
size increased in synthetic polyploids (Tan and Dunn, 1973; Saraswathy Amma et al., 1984; Hahn et al.,
1990; Fassuliotis and Nelson, 1992; Vandenhout et al., 1995; Breuer et al., 2007; Tate et al., 2009; Ye et
al., 2010; Majdi et al., 2010; van Laere et al., 2011; Oates et al., 2012), supporting the results detected in
natural systems. Synthetic polyploids tend to be taller (Tai and Dewey, 1966; Rowe, 1967; Saraswathy
Amma et al., 1984), though sometimes the opposite is true (Jaskani et al., 2005), or there is no
difference between ploidy levels (Oates et al., 2012). Differences in leaf length and thickness are more
commonly observed, and increased in the synthetic tetraploid Phlox subulata L. (Zhang et al., 2008), as
well as in other synthetic tetraploids (e.g. Saraswathy Amma et al., 1984; Hahn et al., 1990; Bretagnolle
and Lumaret, 1995; Majdi et al., 2010). Floral structures also exhibit a general increase in size when
polyploidy is induced, as observed in synthetic tetraploid Phlox subulata (Zhang et al., 2008) as well as in
other systems (Saraswathy Amma et al., 1984; Bretagnolle and Lumaret, 1995; Jaskani et al., 2005;
Majdi et al., 2010; Oates et al., 2012).

Is ploidy diversity associated with overall morphological diversity?
Studies that analyze the relationship between ploidy and overall morphological diversity across
individuals in a multivariate framework, especially using ordination methods such as principal
components analysis (PCA) are not common. PCA allows a determination of whether or not ploidy adds
to the overall pattern of morphological diversity to be made. Morphological diversity added by
polyploidy may be highly apparent (e.g. Ståhlberg, 2009), dispersed with overlap along a gradient (e.g. Li
5

et al., 2010), or absent (e.g. Hodálová et al., 2007; Mandáková and Münzbergová, 2008; Španiel et al.,
2008; Šingliarová et al., 2011). Balao et al. (2011) took an interesting approach by using genome size,
thereby treating ploidy as a covariate rather than a factor. In this case, principal components describing
overall morphological diversity were significantly associated with increasing genome size. Many studies
do not present a strong case for a direct contribution of ploidy to overall morphological diversity,
possibly because characters used in these analyses are only those which would be used for creating
taxonomic keys, rather than a broad array of characters (e.g. Hodálová et al., 2007; Mandáková and
Münzbergová, 2008; Španiel et al., 2008; Šingliarová et al., 2011).

Do different ploidy levels have different multivariate phenotypes?
Canonical structure loadings that describe the contribution of individual characters to
discrimination (e.g. Šingliarová et al., 2011) are rarely reported. Although ploidy levels were not distinct,
Španiel et al (2008) were able to identify that number of stems and several floral characters, including
pappus length and number of florets, contributed more information than the rest of the characters
(which were all floral). In other systems, discrimination between ploidy levels (Hardy and
Vanderhoeven, 2000) and/or species of different ploidy is possible (Shore and Barrett, 1985; Hardy and
Vanderhoeven, 2000; Šingliarová et al., 2011), and sometimes not (Hodálová et al., 2007; Mandáková
and Münzbergová, 2008; Španiel et al., 2008). This pattern could be attributed to ploidy having a weak
morphological signal, or it could be due to the types of characters used. The use of canonical
discriminant analysis to discriminate between three ploidy levels in a species while also examining
morphological features individually is very uncommon. Li et al. (2010) conducted a study with these
attributes in Actinidia deliciosa C.F.Liang & A.R.Ferguson. and, although some differentiation occurred
between tetraploid, pentaploid, and hexaploid plants occurred, it was not perfect.

6

Are populations different morphologically?
The potential of ploidy to affect morphology occurs in the context of climatic and genetic
variation at the level of population. Populations of different ploidy levels may occupy different climate
niches (e.g. Ramsey, 2011; Manzaneda et al., 2012; Thompson et al., 2014; McAllister et al., 2015) or not
(e.g. Godsoe et al., 2013). The genetic structure of populations may be associated with ploidy (see
references in Soltis et al., 2007; Balao et al., 2010). Others have detected that population genetic
structure between ploidy levels is often relatively invariant (Brown and Young, 2000; Halverson and
Heard, 2008; Mandáková and Münzbergová, 2008; Ramsey et al., 2008; Fehlberg et al., 2014), although
Mandáková and Münzbergová were able to differentiate diploids and tetraploids in Aster amellus L.
through a small set of alleles unique to each ploidy level (Mandáková and Münzbergová, 2008).
Additionally, gene flow may be limited or absent between populations of different ploidy levels
(Münzbergová et al., 2013). Sometimes, neither morphology nor population genetic structure differ
between ploidy levels (Lihová et al., 2003). Does the commonly recorded pattern of morphological
differentiation between ploidy levels hold true for population genetic variation?
Not only is there variation in the types of characters observed across the literature that
describes morphology and ploidy, but also in the number of ploidy levels assessed and the number of
populations sampled. For example, many examine two ploidy levels (e.g. Nesom, 1983; Brunsfeld and
Johnson, 1990; Segraves and Thompson, 1999; Sugiyama, 2005; Kennedy et al., 2006; Buggs and Pannell,
2007; Vamosi et al., 2007; Treier et al., 2009; Chen et al., 2009; Ståhlberg, 2009; Kim et al., 2012;
Pettigrew et al., 2012), some three (e.g. Khazaei et al., 2009; Li et al., 2010; de las Mercedes Sosa et al.,
2012; Marinho et al., 2014), and some assess morphology among more than three (e.g. Mishra, 1997;
Brochmann et al., 2004; Tate and Simpson, 2004; Balao et al., 2011).
There is also wide variation in the number of populations assessed; Marinho et al. (2014)
sampled four populations total, while Brochmann (1992) sampled 54. Very generally, the number of
7

populations sampled tends to be fewer than 20. Studies that tease apart the impact on morphology
across multiple ploidy levels and multiple naturally-occurring populations of a species, with a high
sample size collected from a broad set of morphological characters, though they exist (Li et al., 2010;
Balao et al., 2011), are rare; more studies of this structure are needed.

Focal taxon
Phlox amabilis Brand is an upright, 8-25cm tall, deep-rooted perennial herb (Wherry, 1955;
Wilken and Porter, 2005). Compared to related taxa, P. amabilis grows from a tough, compact rhizome,
and has a relatively long style (7-15mm) and flat calyx membranes (Wherry, 1955; Wilken and Porter,
2005). A species of special concern due to limited distribution (NatureServe 2011), P. amabilis is
endemic to northwestern Arizona, in particular the Colorado Plateau (Wherry, 1955). It is typically
found in open Pinus ponderosa Douglas ex C.Lawson stands, or less commonly mixed stands of Pinus
edulis Engelm. and Juniperus spp. L. (pers. obs.). Like most Phlox species, P. amabilis is probably
outcrossing (Fehlberg and Ferguson, 2012).
Ploidy variation in Phlox is well-documented (Flory, Jr, 1934; Meyer, 1944; Levin, 1966; Smith
and Levin, 1967; Levy and Levin, 1974; Worcester et al., 2012). In P. amabilis, Fehlberg and Ferguson
(2012) detected diploid, tetraploid, and hexaploid plants. Each population was found to contain only one
ploidy level. Phlox amabilis polyploids probably have an autopolyploid origin, although one population
of hexaploids was genetically distinct, possibly indicating an allopolyploid origin (Fehlberg and Ferguson,
2012). Besides this hexaploid population, no association was detected between ploidy and genetic
diversity.
Although synthetic systems are useful for assessing the direct morphological influence of
polyploidization, I was interested in the final result of polyploidization after environmental and genetic
effects. Can the signal of polyploidy be detected under complex circumstances? How does polyploidy
impact morphological diversity In natural populations?
8

Aside from helping explain the link between ploidy and morphology in natural systems at large,
understanding the variation associated with polyploidy in this species might be important in establishing
conservation priorities. I attempted to help fill this gap by focusing on the relationship between ploidy
and morphological diversity in a species with a limited geographic distribution.

Objectives
This study incorporates a large sample size, collected across most known populations of P.
amabilis, and examines a broad array of morphological characters to answer how ploidy impacts
morphological diversity in this species. How widespread are the effects of ploidy across the phenotype?
Which features of morphology vary with ploidy, and do they vary the same way between all three ploidy
levels? Does ploidy contribute greatly to overall morphological diversity of P. amabilis? Does each
ploidy level have a distinctive phenotypic profile? Are morphological differences between populations,
potentially due to environment or evolutionary changes since formation, apparent in natural
populations? I hope to contribute a better understanding of these fundamental questions about the
nature of polyploidy and the relationship it has to morphological diversity.

9

MATERIAL AND METHODS
Sampling and field measurement
Samples of P. amabilis were collected in May 2013 from 12 sites in northwestern Arizona (Table
1). The Ferguson lab at Kansas State University (KSU) determined the ploidy level for these sites using a
combination of chromosome counts and flow cytometry (some previously published in Fehlberg and
Ferguson, 2012).
Table 1. Sites of Phlox amabilis sampled for morphology in May 2013. Two ploidy levels were present
at Table Top Mountain; plants of different ploidy were treated as separate sites. Unless otherwise
indicated, ploidy determinations are from Fehlberg and Ferguson (2012). We sampled 25 individuals per
site, except for Table Top Mountain, which consisted of 11 diploids and six tetraploids.
Population
Ploidy Voucher**
Date
County
GPS coordinates
Black Rock Summit*
2x
AP3812
14-May-13 Mojave
N36°47.025' W113°48.416'
Camp Wood
2x
AP3804
9-May-13
Yavapai
N34°47.533' W112°57.345'
Death Valley Springs
4x
AP3813
15-May-13 Mojave
N36°22.015' W113°15.225'
Hobble Mountain
6x
AP3807
11-May-13 Coconino N35°29.153' W111°53.191'
Kaibab Lake
6x
AP3806
10-May-13 Coconino N35°16.317' W112°09.535'
Lost Spring Tank*
6x
AP3808
11-May-13 Coconino N35°29.878' W112°01.111'
Mingus Mountain
6x
AP3801
8-May-13
Yavapai
N34°40.338' W112°08.733'
Pine Creek*
2x
AP3810
12-May-13 Yavapai
N34°48.415' W112°44.398'
Table Top Mountain*
2x
AP3805
10-May-13 Yavapai
N34°52.588' W112°57.070'
Table Top Mountain*
4x
AP3805
10-May-13 Yavapai
N34°52.588' W112°57.070'
Thumb Butte
2x
AP3802
9-May-13
Yavapai
N34°32.448' W112°32.905'
Watson Lake
4x
AP3803
8-May-13
Yavapai
N34°34.021' W112°26.124'
* Indicates new sites relative to Fehlberg and Ferguson (2012), with ploidy level determination
conducted in 2013.
**Voucher specimens were collected by Alan Prather and deposited in the Michigan State University
Herbarium (MSC).

10

Figure 1. Geographic locations of populations. Diploid populations are represented with black circles,
tetraploid with red squares, and hexaploid with blue triangles. Population codes represent vouchers
associated with the locations in Table 1.
At each site except Table Top Mountain, where 11 diploids and six tetraploids were sampled, we
collected data and material from 25 individuals and labeled them A-Y. When possible, we preferentially
selected plants that had two or more flowering stems so that we could collect two data points per plant
for most characters. Populations consisted of reproductive and non-reproductive individuals; plants in
full flower were seldom abundant or densely spaced. Thus, we haphazardly chose plants with two or
more flowering stems until we reached 25 individuals. In several cases we sampled plants with only one
flowering stem because we were not able to locate 25 individuals with two.
We recorded five characters in the field. Stem length (analogous to what others refer to as
“plant height”) describes the distance from soil level to the tip of one of the corolla lobes of the longest
stem when gently pulled taught. The number of flowering stems and sterile stems was recorded. The
number of flowers per stem, and internode length between the uppermost pairs of leaves (not bracts),
were recorded from two stems. “Stems” were distinguished by locating offshoots from the plant’s main
axis that had at least two consecutive sets of leaves before any vegetative or floral branching.
We collected four kinds of material for later analysis from each stem. The uppermost leaf
(defined as the most apical foliar structure with no flower arising from the axil) was collected. A leaf
11

peel was made on the leaf opposite the one collected. To create leaf peels, nail polish (Nutra Nail Nail
Strengthener™) was applied to the abaxial side of the leaf near the middle of the blade (or along the
entire leaf if it was short). Peels were allowed to dry for approximately 20-25 minutes before they were
removed. The youngest expanded flower and at least one anther from an unopened flower (or opened
flower if no un-opened flowers were available) were also collected. Leaves and flowers were stored in
preservative (70% ethanol, 5.25% glycerin) while leaf peels and anthers were collected and stored dry in
microcentrifuge tubes.

Laboratory measurement
Length, width, and thickness were measured from preserved leaves using calipers. Width was
measured at the widest location on each leaf. Additionally, we assessed leaf trichome density by
counting the number of trichomes for three 1.25x1.25mm areas at 40x, evenly spaced on the adaxial
surface of the leaves using an ocular grid. One count was taken near the leaf tip, one in the middle of
the blade, and one near the leaf base. The grid was positioned to capture hairs between the midvein
and the margin of the leaves.
Leaf peels were mounted on microscope slides and examined using Leica Application Suite 2.0.0
(LAS EZ, Leica Microsystems 2010) to characterize guard cell length and stomatal density. We used a
magnification setting of 50x (100x objective x 0.5x camera). Stomata were located by first centering the
microscope on one side of the peel, between the midrib and edge, and between the peel tip and peel
base. In this description, “stoma” refers to the guard cells in combination with the aperture, as opposed
to just the aperture. The first viewable stoma was photographed, and then two more stomata were
selected by rotating the stage adjustment knob about a quarter turn until a new stoma came into view.
A stoma was rejected if it was distorted or part of it was out of the focal plane. Average guard cell
length was calculated for each plant. Stomatal density was assessed by viewing with LAS EZ 2.0.0 and
counting three haphazardly chosen fields of 130μm x 98μm, located between the edge of the peel and
12

the imprint of the leaf’s midrib, at 200x. Average number of stomata per mm2 was calculated for
individual plants using six fields.
Within no more than ten days after collection, anthers were dried for 24 hours at 40°C to halt
fungal growth and to prepare pollen for staining. Anthers were torn apart using forceps in a drop of
cotton blue in lactophenol solution (Sigma-Aldrich) on a microscope slide and a cover slip was added.
Slides were heated for at least three hours and sealed with two coats of nail enamel (Sally Hansen Hard
as Nails™). LAS EZ was used to view grains in order to assess size. At 40x magnification, 10 viable grains
were haphazardly chosen to be photographed from the left or right edge of the slide to its center. Their
diameters were then measured in LAS EZ.
Calyx length, corolla tube length, corolla tube width, corolla lobe length, corolla lobe width,
notch size, longest and shortest stamen lengths, anther length, gynoecium length, ovary length, style
length, and stigma length were measured from preserved flowers. Calipers were used to measure calyx
length, corolla tube length, and corolla tube width. Corolla tube length was measured as the distance
from the base of the corolla to the point where corolla becomes divided into lobes. Corolla tube width
was measured as close to where the corolla lobes separate as possible. Corolla lobes were cut from both
flowers of a plant and mounted on slides with Permount (Fisher Scientific). Measurements were taken
at 10x with an ocular micrometer. The distance between the base and tip of the corolla lobe was
measured by running a straight edge between the two tips of the corolla lobe and measuring from it to
the base of the corolla lobe. Additionally, the width of the corolla lobe, measured at the widest point;
and the size of the notch in the corolla lobe, measured between the straight edge and the base of the
notch, were collected). An ocular unit to millimeter conversion factor was obtained by measuring a
millimeter ruler.
Corollas were dissected by slitting the tube longitudinally. The open corolla and stamens were
photographed under a dissecting microscope at 8x and measured using LAS EZ software. The lengths of

13

the longest and shortest stamen (from the tips of their anthers to the base of the corolla tube) and the
length of the anther from the shortest stamen were measured. The gynoecium was photographed
separately, and total gynoecium length, style length, and stigma length were measured in LAS EZ. Ovary
length was calculated by subtracting style length and stigma length from that of the total gynoecium.
Conversions were obtained by measuring a millimeter ruler in LAS EZ
We used thin-plate spline to quantify shape variation in corolla lobes (typically referred to as
“petals”). Corolla lobe slides were imaged using a scanner. tpsDig v2.17 (Rohlf 2013a) was used to add
curves and scale to corolla lobe photographs (Figure 2). True landmarks were located at both edges at
the base of each lobe, and also in the notch at the distal end of each lobe. Landmarks and
semilandmarks were initially placed together, as two curves using resampling: the first curve with 32
points and the second with 31. The curves were then converted in tpsUtil v1.58 (Rohlf 2013b) to true
landmarks. tpsUtil was then used to generate a “sliders” file, allowing the specification of which
“landmarks” were actually semilandmarks. tpsRelW v1.53 (Rohlf 2013c) was used to compute relative
warp scores and corolla lobe centroid (a measurement of overall size) for each individual. Additionally,
tpsRelW was used to compute the percent of shape variation explained by each relative warp.

Figure 2. Example landmarked corolla lobe of P. amabilis. True landmarks are indicated by black
arrows; the remaining points are semilandmarks. Landmarking was conducted using tpsDig v2.17.

14

Univariate diagnostics
Diagnostics were carried out using SAS v9.3. (SAS Institute, Inc., Cary, NY, USA). Data from
different plants were assumed to be independent. Normality of residuals was assessed by examining
normal probability plots generated with PROC UNIVARIATE. Most characters had residual distributions
that were close to normal. Stem length, total number of stems, percentage of stems flowering, number
of flowers per stem, and internode length had residuals that were not normally distributed, so they
were log-transformed. Another assumption of GLM, equal variances between groups (ploidy levels in
this case) was accounted for by using model fitting during the tests. The REPEATED option in PROC
MIXED function has the ability to (1) generate a model under the assumption of unequal variances and
(2) assess the fit of the model using the Akaike Information Criterion (AIC). I enabled the REPEATED
option at the ploidy level in order to calculate separate variances if doing so reduced AIC.

Univariate analysis
To determine whether a series of univariate tests was appropriate, a MANOVA that included all
characters was performed using PROC GLM. This step allowed us to determine whether or not ploidy
had an overall association with morphology.
A univariate GLM for each character, including characters removed due to correlation (Table 2)
was carried out using PROC MIXED. Degrees of freedom for each GLM were obtained using a
Satterthwaite calculation. Aside from Table Top Mountain (Table 1), each population was assumed to
consist of only one ploidy level; that is, population was nested within ploidy. Because of this structure,
perceived morphological differences between ploidy levels could actually be due to differences between
populations. To separate the effect of ploidy from that of population, we defined population as a
random effect within PROC MIXED. Because of the penalty applied to the model by differences between
the populations within one ploidy level, the mixed model is necessarily conservative in its estimates of
significance between ploidy levels. Post-hoc analysis was performed using Tukey’s HSD.
15

Principle components analysis
In order to assess overall morphological variation without respect to ploidy, we used principal
components analysis. Individuals with one or more characters with missing measurements (14 in total)
were excluded from the PCA, leaving a sample size of n=253. Because including strongly correlated
characters in a PCA is not appropriate, PROC CORR was first used to identify characters correlated above
r=0.6. If two characters were correlated, one of them was removed (see below). Multivariate analysis
assumes a multivariate normal distribution, although PCA and discriminant analysis are robust to small
deviations from normality. This was assessed using the multnorm SAS macro (Figure 3, SAS Institute
2015). Log percent stems flowering was removed because it had a large non-normal influence on the
multivariate distribution. To equalize the influence of each character, variables used to generate the
PCA were converted to a scale of mean zero and standard deviation one. PCA was performed using
PROC PRINCOMP. To assess how populations grouped, a second PCA was performed on population
means standardized using the same method as the PCA for individual observations.
Table 2. Sets of correlated characters and whether they were removed or retained for principal
components analysis. Characters were considered correlated when they had a Pearson’s correlation of
r>0.6. Obviously, corolla lobe centroid was retained because it describes both corolla lobe length and
corolla lobe width at once. One character was randomly chosen from the remaining sets of correlated
characters.
Removed character
Corolla lobe length
Corolla lobe width
Corolla notch size
Shortest stamen length
Longest stamen length
Style length

Retained character
Corolla lobe centroid
Corolla lobe centroid
Relative warp 2
Corolla tube length
Corolla tube length
Corolla tube length

r
0.86
0.65
0.82
0.80
0.89
0.63

16

MULTNORM macro: Chi-square Q-Q plot

Squared distance

50
40
30
20
10
10

20

30

40

Chi-square quantile
Squared Distance

LineParm

Figure 3. Squared Mahalanobis distance for each Phlox amabilis individual plotted against the
predicted Chi-square distribution. A distribution is multivariate normal if it follows a line with slope of
one and intercept of zero.

Discriminant function analysis
Multiple discriminant analysis was performed in SAS 9.3 to quantify and visualize how well
individual P. amabilis plants can be distinguished by ploidy. The data were determined to approximate a
multivariate normal distribution. PROC STEPDISC was used to determine the significance and relative
contribution to ploidy differentiation made by individual characters. Thus, characters that did not
significantly improve discrimination were removed from the primary discriminant analysis. PROC
DISCRIM was used to calculate maximum squared canonical correlation between optimally correlated
linear combinations of characters in order to create new canonical variables, CV1 and CV2.
The three primary options for evaluating how well groups are separated in SAS are resubstitution, a
variant of resubstitution known as data partitioning, and jackknife cross-validation. In resubstitution,
the algorithm computes a single discriminant function that describes the contribution of each character
to ploidy differentiation. This method is biased to overestimate group (Lance et al., 2000). Data
partitioning, in which part of the dataset may be removed and used to train the algorithm that classifies
the remaining data, can avoid this problem. This technique is questionable with small sample sizes,
however. Finally, jackknife cross-validation sequentially removes each observation, calculates a new
17

function, and classifies the observation. This classification algorithm was found to be practically
unbiased by Lance et al. (2000) in comparison to resubstitution. Therefore, parametric jackknife crossvalidation was enabled in PROC DISCRIM to create a classification model, using method=normal. The
results included posterior probabilities for ploidy membership for all individuals, and overall model
classification efficacy.

18

RESULTS
The MANOVA of all characters yielded a significant Wilks’ Lambda statistic (p<0.0001).
Subsequent univariate GLM found significant associations between morphology and ploidy in 15 out of
27 characters (Table 3). Six characters differed significantly between diploids and polyploids, but the
two polyploid levels did not differ from one another. The tetraploids differed from the other two levels
for one character and the hexaploids different for one. For the remaining seven characters, diploids
differed from one of the polyploidy levels while the third was intermediate (Table 3). No differences
were detected between ploidy levels for measurements of plant habit (characters shaded blue in Table
3), but two of four leaf characters (green), ten of 16 perianth characters (pink) differed between at least
two ploidy levels (Table 3).

19

Table 3. Results of univariate analyses of all morphological characters and their association with
ploidy in natural populations of Phlox amabilis diploids, tetraploids and hexaploids. The overall Type
III significance levels generated using the mixed model are reported. Significant Tukey’s HSD results are
shown as letters next to ploidy level averages: ploidy levels sharing a letter in a particular character were
not found to be significantly different by the post-hoc analysis. Characters are sorted by type: cell size
characters are shown in yellow, habit characters in blue, leaf characters in green, and reproductive
characters in pink. Significant differences between two or more ploidy levels (p<0.05) are presented in
bold font.
Character

2x

Guard cell length (μm)
Pollen diameter (μm)
2

Trichome density (#/mm )
2

Stomatal density (#/mm )
log Stem length (cm)
log Number of stems
log Percentage of stems flowering
log Number of flowers per stem
log Internode length (mm)
Leaf length (mm)
Leaf width (mm)
Leaf thickness (mm)
Calyx length (mm)
Corolla tube length (mm)

4x

36.66

a

39.89

a

4.75

a

6x

42.68
44.03
4.10

a

b

ab

b
b

Type III test statistic

42.48

b

<0.001

44.82

b

0.011

2.97

b

0.003

ab

112.70
1.1
0.39
1.83
0.81
0.9
25.01a
3.38
0.40a
8.43
14.16a

82.13
1.15
0.36
1.84
0.88
1.02
31.92b
3.08
0.36a
9.3
15.90b

92.98
1.037
0.2
1.91
0.71
0.85
24.42a
3.51
0.57b
8.78
15.16ab

0.023
0.343
0.183
0.065
0.462
0.286
0.01
0.429
0.009
0.114
0.019

Corolla tube width (mm)

1.77a

1.85ab

2.00b

0.025

Corolla lobe length (mm)

8.27

a

9.71

b

b

0.002

5.30

a

6.57

b

ab

0.042

1.13

a

b

0.031

Corolla lobe width (mm)
Corolla notch size (mm)

a

1.70

ab
b

9.49
6.16

1.68

b

Corolla lobe centroid
Relative warp 1*
Relative warp 2*
Length of shortest stamen (mm)
Length of longest stamen (mm)

25.9
0.008
-0.014
10.45
14.01a

29.76
0.003
0.012
12.19
16.54b

29.49
-0.003
0.007
11.72
15.69ab

0.007
0.767
0.116
0.055
0.02

Anther length (mm)

2.20a

2.52b

2.56b

0.005

a

b

b

Ovary length (mm)
1.51
1.79
1.79
0.009
Style length (mm)
9.08
10.81
10.64
0.268
Stigma length (mm)
0.84
0.98
1.01
0.128
*Together, relative warps 1 and 2 explained ~72% of the variation in corolla shape among individual
plants

20

The degree of overlap of morphological measurements between ploidy levels varied by
character (Figure 4). Distributions of guard cell length (Figure 4D), for example, were relatively welldefined between diploids and polyploids, while distributions of other characters, such as trichome
density (Figure 4C) were not as clearly delineated. Generally, characters that were significantly
associated with ploidy, and were some measure of size, increased between diploids and polyploids
(Figure 4).

a

b
b

b

b

b

a

ab

a

A

B

C

Figure 4. Distributions of measurements of characters found to differ significantly between at least
two ploidy levels across natural populations of Phlox amabilis. Horizontal lines through the boxes
represent median values, while the lower and upper box limits represent the distribution of 25th and 75th
percentiles. Whiskers indicate the distribution of 5th and 95th percentiles. Points beyond the whiskers
represent values above and below 5th and 95th percentiles. Letters above the distributions represent
results of Tukey’s HSD; ploidy levels sharing a letter in a particular character were not found to be
significantly different by the post-hoc analysis.

21

Figure 4 cont’d
b

a

a
b

b
ab

a

a

a

D

E
b

F
b

ab

ab

a

a

b

a

G

I

H
b

b

ab

a

a

ab

a

J

b

K

L
22

b
b

Figure 4 cont’d
b
a

ab
a

b
b

b

b

a

M

a

N

O

23

Principal component analysis revealed that morphological variation associated with ploidy is a
large component of morphological diversity between individual plants. Principal component 1 explained
26% of the morphological variation between individuals while principal component 2 explained 11% of
the variation between individuals. Principal components 3-6 explained only 5-9% of the variation each.
Principal component 1 separated diploids from polyploids (tetraploids plus hexaploids, which broadly
overlapped on this axis). Principal component 2 separated tetraploids from hexaploids, but diploids
were broadly overlapping with both along this axis.

PC2 (11%)

2
1
0
-1
-2

-2

-1

0

1

2

PC1 (26%)
Ploidy:

Diploid

Tetraploid

Hexaploid

Figure 5. Principle component analysis of morphological data collected from natural populations of
diploid, tetraploid, and hexaploid Phlox amabilis plants. PC1 explains 26% of the variation between
individuals, while PC2 explains 11%.
Many characters were correlated with PC1 roughly equally. Nine had correlations between 0.27
and 0.33 (Table 4): Guard cell length, corolla lobe centroid, ovary length, calyx length, stomatal density,
pollen diameter, corolla tube width, corolla tube length, and anther length (in order of decreasing
correlation). Only four characters were more highly correlated with PC2 than all others: Leaf thickness,
trichome density, and log internode length.

24

Table 4. Eigenvectors indicating correlation of each character with principal component axes 1 and 2
for diploid, tetraploid, and hexaploid Phlox amabilis individuals. Sets of characters with the highest
correlations are highlighted.
Character
Log number of stems
Log number of flowers per stem
Log internode length
Leaf length
Leaf width
Leaf thickness
Trichome density
Guard cell length
Stomatal density
Calyx length
Corolla tube length
Corolla tube width
Corolla lobe centroid
Relative warp 1
Relative warp 2
Anther length
Pollen diameter
Ovary length
Stigma length

PC1
-0.13
-0.06
0.15
0.12
0.08
0.14
-0.05
0.33
-0.3
0.31
0.28
0.29
0.33
0
0.17
0.27
0.3
0.32
0.23

PC2
0.25
0.04
0.38
0.27
-0.23
-0.46
0.42
-0.11
-0.26
0.23
0.16
-0.22
0.02
-0.16
-0.03
-0.05
0.01
0.05
-0.22

Stepwise discriminant analysis revealed that 14 characters contributed significantly to general
discrimination between ploidy levels (Table 5). Guard cell length was by far the most robust overall
discriminator; leaf length and pollen diameter also discriminated strongly compared to the rest of the
characters.

25

Table 5. F-values for 14 characters that contributed significantly (p<0.05) to overall discrimination
between ploidy levels in Phlox amabilis in a stepwise discriminant analysis. Characters with higher Fvalues added more new information about discrimination than others.
Character
Guard cell length
Leaf thickness
Pollen diameter
Length of longest stamen
Leaf length
Trichome density
Calyx length
Corolla tube width
Leaf width
Relative warp 1
Ovary length
Log number of flowers/stem
Log number of stems
Anther length

F Value
161.24
85.42
50.95
19.88
15.01
11.65
8.05
7.78
7.54
6.39
6.09
5.42
5.12
3.13

Canonical discriminant analysis indicated that four out of the fourteen characters that
significantly contributed to discrimination overall were especially important for discrimination parsed
between ploidy levels. Morphological discrimination between diploids and polyploids occurred on CV1,
while CV2 describes discrimination between tetraploids and hexaploids (Figure 6). Guard cell length and
pollen diameter were the strongest discriminators along canonical variate 1 (CV1), although other
characters also contributed information to discrimination on this axis (Table 6). Leaf thickness and leaf
length discriminated best along canonical variate 2 (CV2). Both canonical variates were significantly
correlated with variation between ploidy levels (p<0.001).
Canonical discriminant analysis generated a well-defined, but slightly overlapping, group for
each ploidy level (Figure 6). No more than three individuals fell outside of their respective 95%
confidence limit for diploids, tetraploids, or hexaploids. As in principal component analysis, canonical
discriminant analysis also differentiated diploids from polyploids along CV1, and tetraploids from
hexaploids along CV2.

26

4

CV2

2
0
-2
-4
-5.0

-2.5

0.0

2.5

5.0

CV1
Ploidy:

Diploid

Tetraploid

Hexaploid

Figure 6. Individual Phlox amabilis plants plotted across canonical variates (CVs), generated using 14
characters retained in the stepwise analysis. Diploids are shown in black, tetraploids in red, and
hexaploids in blue. Ellipses represent 95% confidence limits of each ploidy level.
The 253 Phlox amabilis individuals were classified with an average error rate of 4.74%; 12
individuals were misclassified, based on cross-validation classification (Table 7). Four of these
individuals had a probability greater than 0.7 of misclassification. The rest of the individuals had a
moderate probability of belonging to their original group (Table 7).

27

Table 6. Total canonical structure associated with the discriminant function which maximized
differences between diploid, tetraploid and hexaploid Phlox amabilis. Values >0.5 are indicated.
Variable
Log number of stems
Log # flowers/stem
Leaf length
Leaf width
Leaf thickness
Trichome density
Guard cell length
Calyx length
Corolla tube width
Relative warp 1
Length of longest stamen
Anther length
Pollen diameter
Ovary length

CV1
-0.30
-0.16
0.04
0.05
0.52
-0.31
0.81
0.23
0.57
-0.08
0.60
0.53
0.75
0.59

CV2
0.20
0.28
0.60
-0.27
-0.65
0.12
0.27
0.30
-0.23
0.03
0.41
0.08
0.10
0.17

Table 7. Results of linear discriminant analysis, using cross-validation to assess accuracy of the
canonical discriminant function used to classify ploidy levels of 253 Phlox amabilis individuals.
Samples with posterior probabilities of misclassification >0.70 are highlighted.

Specimen

Assigned
Ploidy

Posterior Probability of
Membership in Ploidy

2
4
6
3801T*
6
0.60
0.02
0.38
3802P*
4
0.00
0.44
0.56
3802W
4
0.04
0.34
0.62
3802X
4
0.84
0.11
0.05
3803N*
2
0.47
0.49
0.04
3803O*
2
0.39
0.02
0.59
3805Y*
4
1.00
0.00
0.00
3812B
2
0.29
0.70
0.01
3812V
2
0.35
0.40
0.25
3812X
2
0.00
0.00
1.00
3813T
4
0.01
0.42
0.57
3813V
4
0.02
0.46
0.52
*indicates individuals for which flow cytometry data were acquired. For the remainder of individuals in
the table, ploidy was inferred from sampling other individuals in the respective population.

28

Principal component analysis of population means revealed general grouping of populations by
ploidy level (Figure 7). Principal component 1 explained 38% of the variation between populations,
while PC2 explained 18% (Figure 7). Diploid populations did not group as tightly as either group of
polyploid populations. While there is no overlap in the distributions of populations of different ploidy
levels, it should be noted that some populations were morphologically closer to populations of other
ploidy levels (e.g. population 3812, a diploid, is more similar to 3813, a tetraploid, than it is to some
other diploids such as 3810). Eigenvectors for most characters that loaded on PC1 were close to each
other, while on PC2 values were more stratified.
3802

3813

PC2 (18%)

1

3803

3812
3805-2x

0

3805-4x

3804

3807
3801

-1

3808
3806

3810

-1

0

1

PC1 (38%)
Ploidy:

Diploid

Tetraploid

Hexaploid

Figure 7. Populations of Phlox amabilis plants plotted across principal components (PCs) using
uncorrelated characters for diploids, tetraploids, and hexaploids. PC1 explains 38% of the variation
between populations, while PC2 explains 18%. Labels represent AP voucher numbers in Table 1. Table
Top Mountain, population 3805, is separated into diploid (38052) and tetraploid (38054) components.
Many characters contribute nearly equally to PC1; guard cell length, stomatal density, and nine
floral characters (all except Relative Warp 1) were nearly evenly correlated to PC1. In comparison, plant
habit characters and the leaf characters tended to be correlated to PC2.

29

Table 8. Eigenvectors indicating correlation of each character to PC axes 1 and 2 for diploid,
tetraploid, and hexaploid Phlox amabilis populations.
Character
Log number of stems
Log # flowers per stem
Log internode length
Leaf length
Leaf width
Leaf thickness
Trichome density
Guard cell length
Stomatal density
Calyx length
Corolla tube length
Corolla tube width
Corolla lobe centroid
Relative warp 1
Relative warp 2
Anther length
Pollen diameter
Ovary length
Stigma length

PC1
-0.2
-0.14
0.15
0.14
0.01
0.12
-0.13
0.32
-0.3
0.26
0.28
0.29
0.33
0.01
0.25
0.24
0.26
0.29
0.26

PC2
0.3
0.25
0.29
0.41
-0.26
-0.42
0.36
-0.04
-0.22
0.25
0.13
-0.2
-0.01
-0.14
0.07
-0.01
0.04
0.05
-0.13

30

DISCUSSION
The effects of ploidy were widespread across the phenotype of naturally occurring
populations of Phlox amabilis plants
I quantified variation in a broad sample of traits spanning four different categories of
morphology: overall plant habit, leaves, the perianth, and reproductive structures (Table 3).
Measurements covered an array of features, from cell size proxies, to dimensions of macroscopic traits
that describe habit, to leaf and floral dimensions, including multivariate estimation of corolla shape. Of
the 27 morphological characters, 15 of them had a significant association with ploidy. In every category
except those features of overall plant habit, variation in most traits was associated with ploidy.
Because of the aforementioned relationship between ploidy and cell size, special attention was
given to the features that are direct measures of cell size, in particular guard cell size and pollen grain
diameter. In P. amabilis, we detected significant differences between ploidy levels, particularly for
guard cell length, which was larger in polyploids (Table 3, Figure 4D). Tetraploids and hexaploids did not
differ from each other, however. For pollen diameter, I found a significant difference between diploids
and hexaploids; tetraploids were intermediate (Table 3, Figure 4N).
Measurements of number of structures per unit area have also been a focus of research into the
link between ploidy and morphology. These characters are likely dependent on cell size, but are not
direct measurements of it. Both density measurements we collected, stomatal density and trichome
density had significant associations with ploidy. Stomatal density was significantly lower in tetraploids
than in diploids, though in hexaploids it was statistically intermediate (Figure 4E, Table 3). Trichome
density was significantly lower in polyploids compared to diploids (Figure 4C, Table3).
I did not detect a relationship between ploidy and characters related to overall plant habit (e.g.
stem length, a measure of overall plant size, Table 3). In contrast, I detected an association between
ploidy and 11 out of 18 leaf, perianth, and reproductive characters of organ size (Table 3). Only two
31

characters in our analysis distinguished polyploids from each other: Leaf length and leaf thickness.
Tetraploids had longer and thinner leaves than hexaploids (Figure 4A, 3B).
I detected morphological divergence between ploidy levels in 9 out of 15 floral characters (Table
3, Figure 4), though some floral characters were correlated with each other (Table 2). Corolla lobe
length, corolla lobe centroid, anther length, and ovary length were all significantly larger in polyploids,
but once again, tetraploids and hexaploids never differed from each other for any floral trait measured.
For the remaining five characters, one of the polyploids was intermediate between diploid plants and
the other polyploid. For example, diploids had significantly shorter corolla tube length than tetraploids,
while hexaploids were intermediate between them (Table 3, Figure 4F).

Ploidy explained a considerable amount of total phenotypic diversity
As the principal components analysis (Figure 5) showed, even when no assumptions were made
about group membership (in comparison to the CVA, Figure 6), individuals of like ploidy mostly grouped
together, indicative of a contribution to morphological diversity by ploidy. The morphological difference
between diploids and polyploids explained more variation between individual plants than any other
component, as evidenced by their relationship with the axis of the first PCA, with the difference
between tetraploids and hexaploids explaining the next highest amount of total variation, as evidence
by their relationship with the second PCA. Furthermore, besides calyx length and log internode length,
features that contributed to differentiation between ploidy levels (Table 4) all had a significant
relationship with ploidy in the univariate analysis (Tables 3, 4), further suggesting that characters
associated with ploidy explain much of the total morphological diversity in natural populations of P.
amabilis. The amount of total variation in morphology explained by the first two principal components
tends to be low in this context (Hodálová et al., 2007; Španiel et al., 2008; Ståhlberg, 2009; Li et al.,
2010; Šingliarová et al., 2011), a pattern I also found (Figure 5). The high number of characters used in

32

the PCA, many of them associated with ploidy in a similar way (Table 3) creates a situation where
several combinations of the variables explain morphological variation in a similar way.

Each ploidy level had a distinctive multivariate phenotype
Canonical discriminant analysis generated groups of individuals with similar phenotypes based
on ploidy level (Figure 6). Some characters contributed more discriminatory information than others; in
particular, guard cell length was by far the best discriminating character (Table 5). Guard cell length and
pollen diameter discriminated diploids from polyploids better than other characters (Table 6). Five of
the six remaining characters that contributed to CV1 were perianth or reproductive features. Leaf
thickness and leaf length contributed most of the discriminatory information to CV2, which
distinguished tetraploids from hexaploids (Table 6).
Some characters that were removed by the stepwise discriminant analysis were significant in
the univariate analysis, but a different character contributed the same type of information in a way that
better discriminated ploidy levels (Tables 3, 5). All characters that were strongly correlated with CV1 or
CV2 were significant in the univariate analysis (Table 6), corroborating those results. Interestingly,
number of stems, number of flowers per stem, and calyx length contributed unique information to
discrimination because they were retained by stepwise analysis (Table 5), but they did not differentiate
ploidy levels well on either canonical axis compared to other characters (Table 6) and they were not
significantly associated with ploidy in the univariate analysis (Table 3).
The canonical functions associated with each ploidy level were able to identify the ploidy of
individual plants with a success rate of ~95% (Table 7). However, discriminant analysis of ploidy levels is
generated from a combination of univariate characters: On a single-character basis, even characters
with high loadings on the CV axes overlap greatly (Figure 4), a trend that others have also detected
(Hodálová et al., 2007; Španiel et al., 2008).

33

Many of the characters that differentiated ploidy levels in principal components analysis also did
so in canonical discriminant analysis (Tables 4, 6), reiterating that those characters which are most
related to discriminating ploidy level are often the same as those that explain most of the morphological
diversity of the species. However, the importance of each character was differently loaded onto the
canonical axes than on the principal component axes because for CVA morphological distance between
ploidy levels, rather than individuals, was maximized (Table 5).
We tried to relate ploidy to the univariate and multivariate morphological profile of P. amabilis.
One of the most important aspects of the sampling we used to do this was collection of many samples
from multiple populations of each ploidy level. This allowed detection of significance for many
characters in spite of the variability between observations and between populations. Even for
characters where a significant difference between ploidy levels was detected, there was still
considerable overlap between individual samples (Figure 4) or populations; no character was perfectly
reliable for determining the ploidy level of an individual plant. Additionally, single population
comparisons for some characters did not necessarily follow the same pattern as all populations
together. For example, the diploid population 3803 had leaves that were shorter, but not significantly
so when compared to the tetraploid population 3813 (Student’s t-test, p=0.11), while the overall trend
was significantly longer leaves in tetraploids. Literature exists that summarizes morphological variation
across many populations and ploidy levels, but with a focus on differences across between species (e.g.
Brochmann, 1992; Marcussen and Borgen, 2010; Šingliarová et al., 2011).
Studies that sample morphological characters that describe many facets of a naturally-occurring
species’ morphology (e.g. Mandáková and Münzbergová, 2008; Ståhlberg, 2009; Balao et al., 2011; de
las Mercedes Sosa et al., 2012; Kim et al., 2012; Pettigrew et al., 2012) are particularly important. Do
characters with different developmental histories, which are possibly under different selective
influences, respond the same way to polyploidization? The answer seems to be a resounding “it

34

depends!” The response to polyploidy in our system seems to be unique: The overall size increase
often associated with genome size only occurred in the jump from diploidy to polyploidy for most
aspects of morphology in P. amabilis. Surprisingly, only leaf morphology contributed significant
information to differentiating tetraploid and hexaploid plants. Tetraploids were intermediate for corolla
tube width and corolla notch size; hexaploids were intermediate for corolla tube length, corolla tube
width, and length of longest stamen, further supporting our result that organ size does not always
increase proportionally with ploidy, or perhaps, as in other systems, phenotype is plastic (Macdonald et
al., 1988) and/or responsive to selection after polyploidization (Jordan et al., 2015).

Morphological differences among Phlox amabilis populations were apparent, but
grouping is primarily by ploidy
Principal components analysis of character population averages demonstrated that populations
did not group randomly, but rather by ploidy (Figure 7). However, ploidy levels did not group tightly, and
overall morphology for a population was sometimes more similar to populations of different ploidy
levels than populations of the same ploidy level. This is tentative evidence that the environment,
genetics, or environment x genetics interaction is/are responsible for some variation among natural
populations, without overshadowing the influence of ploidy. Interestingly, in terms of population
genetic structure, the morphological diversity observed between populations of P. amabilis did not
correspond to the pattern of genetic diversity observed by Fehlberg and Ferguson (2012), who
cytotyped many of the same populations sampled for morphology. The genetically unusual population
detected by Fehlberg and Ferguson (2012) was not different morphologically; Hobble Mountain,
AP3807, grouped tightly with the other hexaploid populations (Figure 7).
An important characteristic of this system was the presence of three ploidy levels. Studies with
more than two ploidy levels that are focused on ploidy variation within one species (e.g. Li et al., 2010;
Balao et al., 2011; de las Mercedes Sosa et al., 2012) are uncommon. This could be due to the rarity of
35

species with representation from multiple ploidy levels. A three-ploidy level system combined with
information from multiple populations with high sample sizes and a wide array of morphological
characters contributes to our focus on capturing the effects of increased ploidy on morphology.
With a comprehensive set of morphological measurements, I found that ploidy contributes to
diversity in a relatively unique way in this species, with the main differences occurring between diploids
and polyploids. This high degree of associative variation in ploidy and morphology is an essential
consideration in determining conservation priorities. The relationship between conservation and ploidy
has not been well studied (Soltis et al., 2007), which is why we need a comprehensive understanding of
the interaction between ploidy, morphology, genetic structure, phylogeny, and ecology to fully
determine the impact of habitat change on the conservation of this species of special concern.
Further research will include assessment of the relationship between morphological, ecological,
and genetic distances: Specifically, how strongly do climate and genetics contribute to the morphology
of P. amabilis? Do ploidy levels have habitat requirements or genetic profiles? If so, are those
requirements related to aspects of their morphology? With the morphological framework established by
this research, we can integrate ideas and help answer fundamental questions about the impact of ploidy
on diversity in natural populations.

36

LITERATURE CITED

37

LITERATURE CITED
BAACK, E. 2005. Ecological factors influencing tetraploid establishment in snow buttercups (Ranunculus
adoneus, Ranunculaceae): minority cytotype exclusion and barriers to triploid. American Journal of
Botany 92: 1827–1835.
BALAO, F., J. HERRERA, AND S. TALAVERA. 2011. Phenotypic consequences at the microevolutionary scale of
polyploidy and genome size dynamics. Journal of Evolutionary Biology 192: 256–265.
BALAO, F., L.M. VALENTE, P. VARGAS, J. HERRERA, AND S. TALAVERA. 2010. Radiative evolution of polyploid races
of the Iberian carnation Dianthus broteri (Caryophyllaceae). New Phytologist 187: 542–551.
BEAULIEU, J.M., I.J. LEITCH, S. PATEL, A. PENDHARKAR, AND C. A. KNIGHT. 2008. Genome size is a strong predictor
of cell size and stomatal density in angiosperms. New Phytologist 179: 975–986.
BENNETT, M.D., AND I.J. LEITCH. 2005. Genome size evolution in plants. In T. R. Gregory [ed.], The Evolution
of the Genome, 89–162. Elsevier Academic Press, Burlington (MA).
BICKFORD, D., D.J. LOHMAN, N.S. SODHI, P.K.L. NG, R. MEIER, K. WINKER, K.K. INGRAM, AND I. DAS. 2007. Cryptic
species as a window on diversity and conservation. Trends in Ecology and Evolution 22: 148–155.
BRETAGNOLLE, F., AND R. LUMARET. 1995. Bilateral polyploidization in Dactylis glomerata L. subsp. lusitanica:
Occurrence, morphological and genetic characteristics of first polyploids. Euphytica 84: 197–207.
BREUER, C., N.J. STACEY, C.E. WEST, Y. ZHAO, J. CHORY, H. TSUKAYA, Y. AZUMI, ET AL. 2007. BIN4, a novel
component of the plant DNA topoisomerase VI complex, is required for endoreduplication in
Arabidopsis. The Plant Cell 19: 3655–3668.
BROCHMANN, C. 1992. Pollen and seed morphology of Nordic Draba (Brassicaceae): Phylogenetic and
ecological implications. Nordic Journal of Botany 12: 657–673.
BROCHMANN, C., A.K. BRYSTING, I.G. ALSOS, L. BORGEN, AND H.H. GRUNDT. 2004. Polyploidy in arctic plants.
Society 82: 521–536.
BROWN, A. H.D., AND A. G. YOUNG. 2000. Genetic diversity in tetraploid populations of the endangered
daisy Rutidosis leptorrhynchoides and implications for its conservation. Heredity 85: 122–129.
BRUNSFELD, S., AND F. JOHNSON. 1990. Cytological, morphological, ecological and phenological support for
specific status of Crataegus suksdorfii (Rosaceae). Madroño 37: 274–282.
BUGGS, R.J.A., AND J.R. PANNELL. 2007. Ecological differentiation and diploid superiority across a moving
ploidy contact zone. Evolution 61: 125–140.
BUTTERFASS, T. 1987. Cell volume ratios of natural and of induced tetraploid and diploid flowering plants.
Cytologia 52: 309–316.

38

CAVALIER-SMITH, T. 2005. Economy, speed and size matter: Evolutionary forces driving nuclear genome
miniaturization and expansion. Annals of Botany 95: 147–175.
CAVALIER-SMITH, T. 1978. Nuclear volume control by nucleoskeletal DNA, selection for cell volume and cell
growth rate, and the solution of the DNA C-value paradox. Journal of Cell Science 34: 247–278.
CHEN, G., W.B. SUN, AND H. SUN. 2009. Morphological characteristics of leaf epidermis and size variation of
leaf, flower and fruit in different ploidy levels in Buddleja macrostachya (Buddlejaceae). Journal of
Systematics and Evolution 47: 231–236.
FASSULIOTIS, G., AND B. NELSON. 1992. Regeneration of tetraploid muskmelons from cotyledons and their
morphological differences from two diploid muskmelon genotypes. Journal of the American Society
for Horticultural Science 117: 863–866.
FEHLBERG, S.D., AND C.J. FERGUSON. 2012. Intraspecific cytotypic variation and complicated genetic
structure in the Phlox amabilis-P. woodhousei (Polemoniaceae) complex. American Journal of
Botany 99: 865–874.
FEHLBERG, S.D., M.C. TY, AND C.J. FERGUSON. 2014. Reexamination of a putative diploid hybrid taxon using
genetic evidence: The distinctiveness of Phlox pilosa subsp. deamii (Polemoniaceae). International
Journal of Plant Sciences 175: 781–793.
FIALOVÁ, M., M. JANDOVÁ, J. OHRYZEK, AND M. DUCHOSLAV. 2014. Biology of the polyploid geophyte Allium
oleraceum (Amaryllidaceae): Variation in size, sexual and asexual reproduction and germination
within and between tetra-, penta- and hexaploid cytotypes. Flora - Morphology, Distribution,
Functional Ecology of Plants 209: 312–324.
FLORY, JR, W.S. 1934. A cytological study on the genus Phlox. Cytologia 6: 1–18.
GODSOE, W., M. A LARSON, K.L. GLENNON, AND K. A SEGRAVES. 2013. Polyploidization in Heuchera cylindrica
(Saxifragaceae) did not result in a shift in climatic requirements. American Journal of Botany 100:
496–508.
GREGORY, T.R. 2001. Coincidence, coevolution, or causation? DNA content, cell size, and the C-value
enigma. Biological Reviews of the Cambridge Philosophical Society 76: 65–101.
HAHN, S., K. BAI, AND R. ASIEDU. 1990. Tetraploids, triploids, and 2n pollen from diploid interspecific
crosses with cassava. Theoretical and Applied Genetics 79: 433–439.
HALVERSON, K., AND S. HEARD. 2008. Origins, distribution, and local co-occurrence of polyploid cytotypes in
Solidago altissima (Asteraceae). American Journal of Botany 95: 50–58.
HARDY, O., AND S. VANDERHOEVEN. 2000. Ecological, morphological and allozymic differentiation between
diploid and tetraploid knapweeds (Centaurea jacea) from a contact zone in the Belgian Ardennes.
New Phytologist 146: 281–290.

39

HODÁLOVÁ, I., V. GRULICH, L. HOROVÁ, M. VALACHOVIČ, AND K. MARHOLD. 2007. Occurrence of tetraploid and
octoploid cytotypes in Senecio jacobaea ssp. jacobaea (Asteraceae) in Pannonia and the
Carpathians. Botanical Journal of the Linnean Society 153: 231–242.
HUSBAND, B., AND H. SABARA. 2004. Reproductive isolation between autotetraploids and their diploid
progenitors in fireweed, Chamerion angustifolium (Onagraceae). New Phytologist 161: 703–713.
HUSBAND, B., AND D. SCHEMSKE. 1998. Cytotype distribution at a diploid–tetraploid contact zone in
Chamerion (Epilobium) angustifolium (Onagraceae). American Journal of Botany 85: 1688–1694.
HUSBAND, B., AND D. SCHEMSKE. 2000. Ecological mechanisms of reproductive isolation between diploid
and tetraploid Chamerion angustifolium. Journal of Ecology 88: 689–701.
JASKANI, M.J., S.W. KWON, AND D.H. KIM. 2005. Comparative study on vegetative, reproductive and
qualitative traits of seven diploid and tetraploid watermelon lines. Euphytica 145: 259–268.
JOHNSON, M., B. HUSBAND, AND T. BURTON. 2003. Habitat differentiation between diploid and tetraploid
Galax urceolata (Diapensiaceae). International Journal of Plant Sciences 164: 703–710.
JORDAN, G.J., R.J. CARPENTER, A. KOUTOULIS, A. PRICE, AND T.J. BRODRIBB. 2015. Environmental adaptation in
stomatal size independent of the effects of genome size. New Phytologist 205: 608–617.
KANNENBERG, L., AND F. ELLIOTT. 1962. Ploidy in Trifolium ambiguum M. Bieb. in relation to some
morphological and physiological characters. Crop Science 2: 1–4.
KENNEDY, B.F., H. A SABARA, D. HAYDON, AND B.C. HUSBAND. 2006. Pollinator-mediated assortative mating in
mixed ploidy populations of Chamerion angustifolium (Onagraceae). Oecologia 150: 398–408.
KHAZAEI, H., P. MONNEVEUX, S. HONGBO, AND S. MOHAMMADY. 2009. Variation for stomatal characteristics
and water use efficiency among diploid, tetraploid and hexaploid Iranian wheat landraces. Genetic
Resources and Crop Evolution 57: 307–314.
KIM, S., A.L. RAYBURN, A. BOE, AND D.K. LEE. 2012. Neopolyploidy in Spartina pectinata Link: 1.
Morphological analysis of tetraploid and hexaploid plants in a mixed natural population. Plant
Systematics and Evolution 298: 1073–1083.
VAN LAERE, K., S.C. FRANÇA, H. VANSTEENKISTE, J. VAN HUYLENBROECK, K. STEPPE, AND M.C. VAN LABEKE. 2011.
Influence of ploidy level on morphology, growth and drought susceptibility in Spathiphyllum
wallisii. Acta Physiologiae Plantarum 33: 1149–1156.
LANCE, R.F., M.L. KENNEDY, AND P.L. LEBERG. 2000. Classification bias in discriminant function analyses used
to evaluate putatively different taxa. Journal of Mammalogy 81: 245–249.
DE LAS MERCEDES SOSA, M., A. PANSERI, AND M. DEMATTEIS. 2012. Morphometric analysis of Stemodia
hyptoides and S. stricta (Plantaginaceae). Plant Systematics and Evolution 298: 1315–1323.
LEVIN, D. A. 1966. The Phlox Pilosa Complex: Crossing and chromosome relationships. Brittonia 18: 142.
40

LEVY, M., AND D. A. LEVIN. 1974. Novel flavonoids and reticulate evolution in the Phlox pilosa-P.
drummondii complex. American Journal of Botany 61: 156.
LI, D., Y. LIU, C. ZHONG, AND H. HUANG. 2010. Morphological and cytotype variation of wild kiwifruit
(Actinidia chinensis complex) along an altitudinal and longitudinal gradient in central-west China.
Botanical Journal of the Linnean Society 164: 72–83.
LIHOVÁ, J., A. TRIBSCH, AND K. MARHOLD. 2003. The Cardamine pratensis (Brassicaceae) group in the Iberian
Peninsula: taxonomy, polyploidy and distribution. Taxon 52: 783–801.
LUTZ, A. 1907. A preliminary note on the chromosomes of Oenothera lamarckiana and one of its
mutants, O. gigas. Science 26: 151–152.
MACDONALD, S., C. CHINNAPPA, AND D. REID. 1988. Evolution of phenotypic plasticity in the Stellaria longipes
complex: comparisons among cytotypes and habitats. Evolution 42: 1036–1046.
MAJDI, M., G. KARIMZADEH, M. MALBOOBI, R. OMIDBAIGI, AND G. MIRZAGHADERI. 2010. Induction of tetraploidy
to feverfew (Tanacetum parthenium Schulz-Bip.): Morphological, physiological, cytological, and
phytochemical changes. HortScience 45: 16–21.
MANDÁKOVÁ, T., AND Z. MÜNZBERGOVÁ. 2008. Morphometric and genetic differentiation of diploid and
hexaploid populations of Aster amellus agg. in a contact zone. Plant Systematics and Evolution 274:
155–170.
MANZANEDA, A.J., P.J. REY, J.M. BASTIDA, C. WEISS-LEHMAN, E. RASKIN, AND T. MITCHELL-OLDS. 2012.
Environmental aridity is associated with cytotype segregation and polyploidy occurrence in
Brachypodium distachyon (Poaceae). The New Phytologist 193: 797–805.
MARCUSSEN, T., AND L. BORGEN. 2010. Species delimitation in the Ponto-Caucasian Viola sieheana complex,
based on evidence from allozymes, morphology, ploidy levels, and crossing experiments. Plant
Systematics and Evolution 291: 183–196.
MARINHO, R.C., C. MENDES-RODRIGUES, A.M. BONETTI, AND P.E. OLIVEIRA. 2014. Pollen and stomata
morphometrics and polyploidy in Eriotheca (Malvaceae-Bombacoideae). Plant Biology 16: 508–
511.
MCALLISTER, C., R. BLAINE, P. KRON, B. BENNETT, H. GARRETT, J. KIDSON, B. MATZENBACHER, ET AL. 2015.
Environmental correlates of cytotype distribution in Andropogon gerardii (Poaceae). American
Journal of Botany 102: 92–102.
MEYER, J.R. 1944. Chromosome studies of Phlox. Genetics 29: 199–216.
MISHRA, M. 1997. Stomatal characteristics at different ploidy levels in Coffea L. Annals of Botany 80:
689–692.
MIZUKAMI, Y. 2001. A matter of size: Developmental control of organ size in plants. Current Opinion in
Plant Biology 4: 533–539.
41

MRÁZ, P., E. TARBUSH, AND H. MÜLLER-SCHÄRER. 2014. Drought tolerance and plasticity in the invasive
knapweed Centaurea stoebe s.l. (Asteraceae): Effect of populations stronger than those of
cytotype and range. Annals of Botany 114: 289–299.
MÜNTZING, A. 1936. The evolutionary significance of autopolyploidy. Hereditas 21: 363–378.
MÜNZBERGOVÁ, Z. 2007. No effect of ploidy level in plant response to competition in a common garden
experiment. Biological Journal of the Linnean Society 92: 211–219.
MÜNZBERGOVÁ, Z., M. SURINOVÁ, AND S. CASTRO. 2013. Absence of gene flow between diploids and
hexaploids of Aster amellus at multiple spatial scales. Heredity 110: 123–30.
NESOM, G. 1983. Galax (Diapensiaceae): Geographic variation in chromosome number. Systematic
Botany 8: 1–14.
OATES, K.M., T.G. RANNEY, AND D.H. TOUCHELL. 2012. Influence of Induced Polyploidy on Fertility and
Morphology of Rudbeckia Species and Hybrids. HortScience 47: 1217–1221.
OTTO, S.P., AND J. WHITTON. 2000. Polyploid incidence and evolution. Annual Review of Genetics 34: 401–
407.
PETTIGREW, F., D. JACK, AND K. BELL. 2012. Morphology, ploidy and molecular phylogenetics reveal a new
diploid species from Africa in the baobab genus Adansonia (Malvaceae: Bombacoideae). Taxon 61:
1240–1250.
RAMSEY, J. 2011. Polyploidy and ecological adaptation in wild yarrow. Proceedings of the National
Academy of Sciences of the United States of America 108: 7096–7101.
RAMSEY, J., A. ROBERTSON, AND B. HUSBAND. 2008. Rapid adaptive divergence in new world Achillea, an
autopolyploid complex of ecological races. Evolution 62: 639–653.
RICE, A., L. GLICK, S. ABADI, M. EINHORN, N. KOPELMAN, A. SALMAN-MINKOV, J. MAYZEL, ET AL. 2015. The
Chromosome Counts Database ( CCDB ) – a community resource of plant chromosome numbers.
The New Phytologist 206: 19–26.
ROWE, P. 1967. Performance and variability of diploid and tetraploid potato families. American Potato
Journal 44: 263–271.
SARASWATHY AMMA, C., V. MARKOSE, J. LICY, AND A. PANIKKAR. 1984. Cytomorphological studies in an induced
polyploidy of Hevea brasiliensis Muell. Arg. Cytologia 49: 725–729.
SEGRAVES, K., AND J. THOMPSON. 1999. Plant polyploidy and pollination: floral traits and insect visits to
diploid and tetraploid Heuchera grossulariifolia. Evolution 53: 1114–1127.
SHORE, J.S., AND S.C.H. BARRETT. 1985. Morphological differentiation and crossability among populations of
the Turnera ulmifolia L. complex (Turneraceae). Systematic Botany 10: 308–321.

42

ŠINGLIAROVÁ, B., I. HODÁLOVÁ, AND P. MRÁZ. 2011. Biosystematic study of the diploid-polyploid Pilosella
alpicola group with variation in breeding system: patterns and processes. Taxon 60: 450–470.
SMITH, D.M., AND D. A. LEVIN. 1967. Karyotypes of eastern North American phlox. American Journal of
Botany 54: 324–334.
SOLTIS, D., P.S. SOLTIS, AND D.W. SCHEMSKE. 2007. Autopolyploidy in angiosperms: have we grossly
underestimated the number of species? Taxon 56: 13–30.
SOLTIS, D.E., P.S. SOLTIS, AND J. A. TATE. 2003. Advances in the study of polyploidy since Plant Speciation.
New Phytologist 161: 173–191.
SOLTIS, D.E., C.J. VISGER, AND P.S. SOLTIS. 2014. The polyploidy revolution then...and now: Stebbins
revisited. American Journal of Botany 101: 1057–1078.
ŠPANIEL, S., K. MARHOLD, I. HODÁLOVÁ, AND J. LIHOVÁ. 2008. Diploid and tetraploid cytotypes of Centaurea
stoebe (Asteraceae) in central Europe: Morphological differentiation and cytotype distribution
patterns. Folia Geobotanica 43: 131–158.
STÅHLBERG, D. 2009. Habitat differentiation, hybridization and gene flow patterns in mixed populations of
diploid and autotetraploid Dactylorhiza maculata s.l. (Orchidaceae). Evolutionary Ecology 23: 295–
328.
STEBBINS, G.L. 1971. Chromosomal evolution in higher plants. Edward Arnold Ltd., London., London.
STEBBINS, G.L. 1950. Variation and evolution in plants.
SUDA, J., R. MALCOVÁ, D. ABAZID, AND M. BANAŠ. 2004. Cytotype distribution in Empetrum (Ericaceae) at
various spatial scales in the Czech Republic. Folia Geobotanica 39: 161–171.
SUGIYAMA, S.-I. 2005. Polyploidy and cellular mechanisms changing leaf size: Comparison of diploid and
autotetraploid populations in two species of Lolium. Annals of Botany 96: 931–938.
TAI, W., AND D. DEWEY. 1966. Morphology, cytology, and fertility of diploid and colchicine-induced
tetraploid crested wheatgrass. Crop Science 6: 223–226.
TAN, G., AND G. DUNN. 1973. Relationship of stomatal length and frequency and pollen-grain diameter to
ploidy level in Bromus inermis Leyss. Crop Science 13: 332–334.
TATE, J. A, V.V. SYMONDS, A.N. DOUST, R.J. A BUGGS, E. MAVRODIEV, L.C. MAJURE, P.S. SOLTIS, AND D.E. SOLTIS.
2009. Synthetic polyploids of Tragopogon miscellus and T. mirus (Asteraceae): 60 Years after
Ownbey’s discovery. American Journal of Botany 96: 979–988.
TATE, J. A., AND B.B. SIMPSON. 2004. Breeding system evolution in Tarasa (Malvaceae) and selection for
reduced pollen grain size in the polyploid species. American Journal of Botany 91: 207–213.

43

TATE, J. A., P.S. SOLTIS, AND D.E. SOLTIS. 2005. Polyploidy in plants. In T. R. Gregory [ed.], The Evolution of
the Genome, 371–426. Elsevier Academic Press, New York (NY).
THOMPSON, K.A., B.C. HUSBAND, AND H. MAHERALI. 2014. Climatic niche difference between diploid and
tetraploid cytotypes of Chamerion angustifolium (Onagraceae). American Journal of Botany 101:
1868–1875.
TREIER, U. A, O. BROENNIMANN, S. NORMAND, A. GUISAN, U. SCHAFFNER, T. STEINGER, AND H. MÜLLER-SCHÄRER.
2009. Shift in cytotype frequency and niche space in the invasive plant Centaurea maculosa.
Ecology 90: 1366–1377.
TRESSENS, S.G. 1970. Morfologia del polen y evolucion en Tarasa (Malvaceae). Bonplandia 3: 73–100.
VAMOSI, J.C., S.J. GORING, B.F. KENNEDY, R.J. MAYBERRY, C.M. MORAY, L. A NEAME, N.D. TUNBRIDGE, AND E. ELLE.
2007. Pollination, floral display , and the ecological correlates of polyploidy. Functional Ecosystems
and Communities 1: 1–9.
VANDENHOUT, H., R. ORTIZ, D. VUYLSTEKE, R. SWENNEN, AND K. V. BAI. 1995. Effect of ploidy on stomatal and
other quantitative traits in plantain and banana hybrids. Euphytica 83: 117–122.
VISSER, V., AND J. MOLOFSKY. 2015. Ecological niche differentiation of polyploidizaton is not supported by
environmental differences among species in a cosmopolitan grass genus. American Journal of
Botany 102: 36–49.
WHERRY, E. 1955. The genus Phlox.
WILKEN, D., AND J. PORTER. 2005. Polemoniaceae, Phlox Family.
WOOD, T.E., N. TAKEBAYASHI, M.S. BARKER, I. MAYROSE, P.B. GREENSPOON, AND L.H. RIESEBERG. 2009. The
frequency of polyploid speciation in vascular plants. Proceedings of the National Academy of
Sciences of the United States of America 106: 13875–13879.
WORCESTER, L., M.H. MAYFIELD, AND C.J. FERGUSON. 2012. Cytotypic variation in Phlox pilosa ssp. pilosa
(Polemoniaceae) at the western edge of its range in the central United States. Journal of Botanical
Research Institute of Texas 6: 443–451.
YE, Y.M., J. TONG, X.P. SHI, W. YUAN, AND G.R. LI. 2010. Morphological and cytological studies of diploid and
colchicine-induced tetraploid lines of crape myrtle (Lagerstroemia indica L.). Scientia Horticulturae
124: 95–101.
ZHANG, Z., H. DAI, M. XIAO, AND X. LIU. 2008. In vitro induction of tetraploids in Phlox subulata L. Euphytica
159: 59–65.

44