Cytogeography of the Humifusa clade of Opuntia s.s. Mill. 1754 (Cactaceae, Opuntioideae, Opuntieae): correlations with pleistocene refugia and morphological traits in a polyploid complex

Abstract Ploidy has been well studied and used extensively in the genus Opuntia to determine species boundaries, detect evidence of hybridization, and infer evolutionary patterns. We carried out chromosome counts for all members of the Humifusa clade to ascertain whether geographic patterns are associated with differences in ploidy. We then related chromosomal data to observed morphological variability, polyploid formation, and consequently the evolutionary history of the clade. We counted chromosomes of 277 individuals from throughout the ranges of taxa included within the Humifusa clade, with emphasis placed on the widely distributed species, Opuntia humifusa (Raf.) Raf., 1820 s.l. and Opuntia macrorhiza Engelm., 1850 s.l. We also compiled previous counts made for species in the clade along with our new counts to plot geographic distributions of the polyploid and diploid taxa. A phylogeny using nuclear ribosomal ITS sequence data was reconstructed to determine whether ploidal variation is consistent with cladogenesis. We discovered that diploids of the Humifusa clade are restricted to the southeastern United States (U.S.), eastern Texas, and southeastern New Mexico. Polyploid members of the clade, however, are much more widely distributed, occurring as far north as the upper midwestern U.S. (e.g., Michigan, Minnesota, Wisconsin). Morphological differentiation, although sometimes cryptic, is commonly observed among diploid and polyploid cytotypes, and such morphological distinctions may be useful in diagnosing possible cryptic species. Certain polyploid populations of Opuntia humifusa s.l. and Opuntia macrorhiza s.l., however, exhibit introgressive morphological characters, complicating species delineations. Phylogenetically, the Humifusa clade forms two subclades that are distributed, respectively, in the southeastern U.S. (including all southeastern U.S. diploids, polyploid Opuntia abjecta Small, 1923, and polyploid Opuntia pusilla (Haw.) Haw., 1812) and the southwestern U.S. (including all southwestern U.S. diploids and polyploids). In addition, tetraploid Opuntia humifusa s.l., which occurs primarily in the eastern U.S., is resolved in the southwestern diploid clade instead of with the southeastern diploid clade that includes diploid Opuntia humifusa s.l. Our results not only provide evidence for the polyphyletic nature of Opuntia humifusa and Opuntia macrorhiza, suggesting that each of these represents more than one species, but also demonstrate the high frequency of polyploidy in the Humifusa clade and the major role that genome duplication has played in the diversification of this lineage of Opuntia s.s. Our data also suggest that the southeastern and southwestern U.S. may represent glacial refugia for diploid members of this clade and that the clade as a whole should be considered a mature polyploid species complex. Widespread polyploids are likely derivatives of secondary contact among southeastern and southwestern diploid taxa as a result of the expansion and contraction of suitable habitat during the Pleistocene following glacial and interglacial events.

tetraploid O. humifusa s.l., which occurs primarily in the eastern U.S., is resolved in the southwestern diploid clade instead of with the southeastern diploid clade that includes diploid O. humifusa s.l. Our results not only provide evidence for the polyphyletic nature of O. humifusa and O. macrorhiza, suggesting that each of these represents more than one species, but also demonstrate the high frequency of polyploidy in the Humifusa clade and the major role that genome duplication has played in the diversification of this lineage of Opuntia s.s. Our data also suggest that the southeastern and southwestern U.S. may represent glacial refugia for diploid members of this clade and that the clade as a whole should be considered a mature polyploid species complex. Widespread polyploids are likely derivatives of secondary contact among southeastern and southwestern diploid taxa as a result of the expansion and contraction of suitable habitat during the Pleistocene following glacial and interglacial events.
Opuntia humifusa s.l. is composed of numerous morphological entities that have been recognized in certain taxonomic treatments as different species (see Small 1933). Throughout its range, O. humifusa s.l. has been divided into as many as 14 taxa (Britton and Rose 1920, Small 1933, Benson 1982, Majure and Ervin 2008. Thus, O. humifusa s.l. is occasionally referred to as a species complex (Doyle 1990). Currently, two taxa are recognized in O. humifusa s.l. (O. humifusa var. ammophila (Small) L.D. Benson and O. humifusa var. humifusa;Pinkava 2003). Likewise, Opuntia macrorhiza has been divided into as many as 11 taxa (see Benson 1982). Opuntia macrorhiza was previously considered a variety of O. humifusa (see Benson 1962; see Table 1 (Benson 1982).
Opuntia pusilla has been divided into several species: O. drummondii Graham, 1841, O. frustulenta Gibbes, 1858, O. impedita Small, 1923, O. pes-corvi LeConte, 1857, and O. tracyi Britton, 1911(Britton and Rose 1920, Small 1933; however, Benson (1982) placed them in synonymy under the name O. pusilla. Opuntia triacantha (Willd.) Sweet, 1826, also has been divided into several species, i.e., O. abjecta of the Florida Keys, O. militaris Britton & Rose, 1919, of Cuba, and O. triacantha from different parts of the Greater and Lesser Antilles (Britton and Rose 1920), but all of these have since been placed in synonymy within O. triacantha (Benson 1982). Phylogenetic and morphological studies have indicated that O. abjecta is not even in the same clade as O. triacantha (LCM, WSJ unpubl. data) and so here is treated as O. abjecta.
Contributing to the confusing taxonomic history of this clade is the high degree of morphological variation exhibited by most taxa, the lack of complete sampling throughout the range of the clade, the absence of cytological and phylogenetic evidence, reliance on poorly prepared and sparse herbarium collections (Majure and Ervin 2008, LCM unpubl. data), and hybridization and polyploidy (Benson 1982, Rebman andPinkava 2001). Careful examination of morphological characters across the geographic range of the widely distributed O. humifusa s.l. and O. macrorhiza s.l. reinforces the hypothesis that hybridization may have preceded the origin of geographical morphotypes, because morphological characters displayed by certain taxa appear to be introgessive between O. humifusa s.l. and O. macrorhiza s.l. ( Table 2). For instance, O. cespitosa Raf., 1830, from the eastern U.S. and recently recognized by Majure and Ervin (2008), has yellow tepals that are basally tinged crimson-to orange-red, a characteristic typical of O. macrorhiza and occasionally O. tortispina from western North America (Benson 1982, Pinkava 2003, Powell and Weedin 2004, but the spine characters of O. cespitosa are typical of O. humifusa s.l. (see Majure and Ervin 2008).
Chromosome counts reported for species in the Humifusa clade do not encompass all of the taxa within the range of the clade nor the wide distributions exhibited by sev- eral of the more common species. To further our understanding of species complexes and the evolution of polyploids within those complexes, cytological data are needed from the entire distribution of a given species (Babcock and Stebbins 1938, Stebbins 1942, Stebbins 1950. Thus, an in-depth study of the distribution of cytotypes and correlations between cytotypes and morphology is desperately needed in order to aid in the delimitation of potentially unrecognized and cryptic species and to elucidate relationships in the Humifusa clade. Here we present chromosome counts for all taxa considered to be part of the O. humifusa complex and all taxa of the Humifusa clade (LCM, WSJ, PSS, DES, unpubl. data) and provide counts throughout most of the known ranges of all taxa to determine the geographic structure of ploidy and differences in ploidy among morphologically distinct taxa. We also reconstruct a phylogeny of diploid and polyploid members of the Humifusa clade based on nrITS data to investigate the relationship between geographic distribution and evolutionary relationships. We provide counts for another common species in the southeastern U.S., O. stricta (Haw.) Haw., 1812, because it has been hypothesized to hybridize with members of the Humifusa clade (Benson 1982). In addition, ploidy of the putative hybrid between O. abjecta and O. stricta, i.e., O. ochrocentra Small, 1923, was analyzed. Ploidy determinations of the Humifusa clade, coupled with morphological character analysis and further molecular phylogenetics, will aid in the delimitation of species in the group and in determining the origin and evolutionary significance of polyploidy in this clade.

Material and methods
Chromosome counts -Methods follow those of Majure and Ribbens (in press). Briefly, root tips were collected from early morning throughout early afternoon and table 2. Selected taxa of O. humifusa s.l. and O. macrorhiza s.l. with morphological characters and corresponding ploidy. Polyploids often exhibit characters from more than one diploid taxon or characters of other polyploids, although certain characters (e.g., red glochids) have not been observed in any diploids analyzed thus far.  (Soltis 1980) for up to 8 hours at 4°C or in N 2 O (Kato 1999) for 1 hour and then fixed in a 3:1 solution of absolute ethanol: glacial acetic acid for 2 to 24 hours. Root tips then were placed in 70% ethanol for at least 2 hours and digested in 40% HCl for 5-10 minutes (depending on the size of the root) at room temperature. Squashes were performed in 60% acetic acid and stained with 1% aceto-orcein dye and viewed on a Zeiss Photomicroscope III (Carl Zeiss, Oberkochen, Germany). To confirm each count, at least three to five metaphase cells were counted per specimen. These multiple counts per sample alleviated concerns regarding endomitosis, which has been reported in the allopolyploid (4x), Opuntia spinosibacca M.S. Anthony, 1956, (Weedin and Powell 1978), tetraploid O. pusilla (Bowden 1945b), as well as in many other angiosperms (e.g., Barrow andMeister 2003, Tate et al. 2009 (Table 1). Thus, we have arranged our counts of plants within these two species (see Appendix 1) according to their various segregates to determine whether the morphological variation of these segregate entities ( Table 2) is correlated with cytotype and/or geographical and phylogenetic patterns.
Cytogeographic analysis -We mapped the localities for all of the individuals for which we determined ploidy (277 in number) and incorporated previous counts (n = 41) (Bowden 1945a, Pinkava and McLeod 1971, Pinkava et al. 1973, Weedin and Powell 1978, Pinkava and Parfitt 1982, Pinkava et al. 1985, Weedin et al. 1989, Doyle 1990, Pinkava et al. 1992, Pinkava et al. 1998, Powell and Weedin 2001, Baker et al. 2009a to cover the majority of the geographic distribution of each taxon. This allowed us to explore the geographic boundaries of the different ploidal levels encountered in this clade and construct hypotheses regarding polyploid formation and speciation. Phylogenetic analysis -We generated sequences from the nuclear ribosomal internal transcribed spacer (nrITS: White et al. 1990) for a sample of diploid (n = 6) and polyploid taxa (n = 8) of the Humifusa clade from the eastern and western U.S. (Table  3). Opuntia basilaris Engelm. & J.M. Bigelow, 1856, was used as an outgroup based on previous analyses of Opuntia (LCM unpubl. data). A phylogenetic analysis of these data was carried out to determine whether the geographic distribution of ploidy (as determined here) was correlated with the evolutionary history of the clade. We carried out a Maximum Likelihood analysis using RAxML (Stamatakis 2006) running 10000 bootstrap pseudoreplicates under 25 rate categories and the GTR+Γ model of molecular evolution.

Results
The base chromosome number for Cactaceae has been well established as x = 11 (Remski 1954, Pinkava and McLeod 1971, Lewis 1980, Pinkava et al. 1985, Pinkava 2002, and we saw no deviation from this in our counts (Appendix 1). Out of 318 counts of the Humifusa clade, including 41 from the literature, 210 (66%) were polyploid and 108 (34%) were diploid. Diploid (2n = 2x = 22) and tetraploid (2n = 4x = 44) O. humifusa s.l. and O. macrorhiza s.l. were discovered ( Fig. 1A-D, I-J, L). Diploid O. humifusa s.l. is restricted entirely to the southeastern U.S., whereas diploid O. macrorhiza s.l. is restricted entirely to the southwestern U.S. (eastern Texas (see Appendix 1) and southeastern New Mexico (M. Baker and D.J. Pinkava pers. comm.)). Tetraploid members of O. humifusa s.l. and O. macrorhiza s.l. are much more widely distributed throughout the U.S. than are their diploid relatives (Fig. 2). Tetraploids of O. humifusa s.l. are found from Massachusetts south to the southeastern U.S. where they abut the distribution of diploid taxa and throughout the eastern and midwestern U.S. Tetraploid O. macrorhiza s.l. is distributed throughout parts of the Great Plains through the midwestern U.S., most of the southwestern U.S., parts of the Rocky Mountains, and the upper Sierra Madre Occidental in Sonora, Mexico (Fig. 2).
Diploid, triploid, and tetraploid populations of O. pusilla were discovered ( Fig.  1E-G) throughout its restricted range in the southeastern U.S. (Fig. 3). Interestingly, with the exception of two populations, polyploid individuals (3x and 4x) were mostly confined to the coastline, although diploid populations were much more widespread throughout the interior part of the distribution of the species (Fig. 3). Of the three examples of O. abjecta sampled from the Florida Keys, one was dip- loid (Fig. 1H), and two were tetraploid. Opuntia tortispina (southwestern U.S.) was hexaploid in six and tetraploid in one of the populations examined (see Fig. 2 for hexaploid distribution). Individuals of O. stricta sampled from the southeastern U.S. were all hexaploid. Samples included members of the taxa considered by some (Anderson 2001) to be O. dillenii (Ker-Gawl.) Haw., 1819, and O. stricta. Three individuals of the putative hybrid O. ochrocentra from two localities in the Florida Keys were pentaploid (Fig. 1K), and the putative hybrid O. alta was hexaploid.
Maximum likelihood analysis of ITS data reveals that the Humifusa clade is made up of two well-supported subclades. One is restricted to the southeastern U.S. and includes polyploid members of O. pusilla and O. abjecta, and the other includes southwestern diploid O. macrorhiza and all other polyploids pertaining to O. humifusa s.l., O. macrorhiza s.l., and O. tortispina. There is no further resolution within the tree at the species level using ITS (Fig. 4). Species relationships within these two clades are further resolved with the addition of other loci (LCM unpubl. data), however, that is beyond the scope of this study.

Discussion
Opuntia macrorhiza has only been recorded previously as tetraploid (Pinkava et al. 1971(Pinkava et al. , 1973(Pinkava et al. , 1977(Pinkava et al. , 1992(Pinkava et al. , 1998Weedin 2001, 2004;Pinkava 2003). These are the first reports of diploid O. macrorhiza and likely represent descendants of those progenitors from which tetraploid O. macrorhiza s.l. and other polyploids arose. Likewise, this is the first report of diploid and triploid O. pusilla, which was formerly known only from tetraploid counts (Bowden 1945a).
Diploid members of O. humifusa s.l. (e.g., represented by the segregate taxa O. ammophila Small, 1919, O. austrina Small, 1903, O. lata Small, 1919, in this study; see also Appendix 1) exhibit high levels of morphological variability but each is diagnosable morphologically, which suggests that these segregate taxa may need to be recognized at the species level. Likewise, diploid material of O. macrorhiza s.l. from eastern Texas (e.g., O. xanthoglochia Griffiths, 1910, in this study; see also Appendix 1) and southeastern New Mexico is morphologically distinct from tetraploid material of O. macrorhiza s.l., which may also justify the recognition of O. xanthoglochia and O. macrorhiza as separate species.
Our hexaploid counts of O. stricta are consistent with those of Pinkava et al. (1992) and Negrón-Ortiz (2007). In contrast, Spencer (1955) reported O. stricta from Puerto Rico to be diploid. Other authors have also found Spencer's counts from Puerto Rico to be inconsistent with more recent counts (e.g., Negrón-Ortiz 2007 for Consolea Lem., 1862).
Diploid refugia and polyploid formation -Polyploidy is very common within the Humifusa clade, occurring in 66% of the samples reported here. Most researchers that have studied Opuntia cytologically have found polyploid taxa (e.g., Bowden 1945a, Weedin and Powell 1978, Pinkava et al. 1985, Doyle 1990, Segura et al. 2007  but see Spencer 1955). All diploids in our analysis were restricted to either the southeastern or southwestern (eastern Texas and southeastern New Mexico) U.S., and the polyploid individuals were found nearly everywhere in between as well as north of these two diploid "refugia." The disjunct pattern observed here in the Humifusa clade and in other studies between the southeastern U.S. and the southwestern U.S. is thought to have occurred as a result of the disruption of a semi-arid zone along the Gulf Coast region during the mid-Pleistocene (Webb 1990, Althoff andPellmyr 2002). These two areas likely served as glacial refugia for a variety of animals and plants (e.g., Remington 1968, Davis and Shaw 2001, Al-Rabab'ah and Williams 2002, Althoff and Pellmyr 2002, Soltis et al. 2006, Waltari et al. 2007, Whittemore and Olsen 2011 and may have promoted current species richness and genetic diversity in southern populations (Hewitt 2000). Specifically, Swenson and Howard (2005) identified southeastern Texas and northern Florida as Pleistocene refugia for animal and plant species. Species from these regions subsequently came into contact following the last glacial maximum and formed hybrid zones at contact areas expand- ing out from these refugia. Swenson and Howard (2005) also hypothesized "postglacial routes of expansion" from these proposed diploid refugia (e.g., Fig. 1 Swenson and Howard 2005). Those post-glacial routes and diploid contact zones are consistent with the current distributions of polyploid taxa within O. humifusa s.l. and O. macrorhiza s.l. The restricted diploid and widespread polyploid distribution pattern has been recorded in many other plants and is a common pattern seen in polyploid complexes (Babcock and Stebbins 1938, Stebbins 1950, DeWet 1971, Lewis 1980, Grant 1981, Parfitt 1991.

, G & H in
The seemingly disjunct southeastern New Mexico diploid population of O. macrorhiza s.l. may represent a mere extension of the eastern Texas diploid refugium, which has since been mostly replaced by polyploid taxa. Alternatively, a diploid extension may still exist but was not detected due to the lack of cytological data for populations from east Texas to southeastern New Mexico (Fig. 2). Diploid taxa of other clades (e.g., O. polyacantha Haw. var. arenaria (Engelm.) Parfitt, 1819) are coincidentally found near the same region (Pinkava 2002(Pinkava , 2003, however, suggesting that a third diploid refugium, i.e., in southeastern New Mexico-western Texas, may need to be recognized. Pinkava (2003) Webb 1990), it is likely that the Humifusa clade originated in the southwestern U.S. and adjacent northern Mexico, then dispersed eastward into the southeastern U.S. The arid habitat along the coast of the Gulf of Mexico during the mid-Pliocene to early Pleistocene would have been interrupted during the mid-Pleistocene, creating the disjunction and promoting the genetic divergence among diploid populations we see today (Fig. 4). Taxa from these two diploid refugia would have come back into contact and formed the widely successful polyploids of the Midwest and eastern U.S. (Fig. 5). This scenario is further corroborated by phylogenetic analyses, where eastern U.S. polyploids of O. humifusa s.l. are resolved in a clade with the southwestern diploid O. macrorhiza (Fig. 4). The lower frequency of diploids encountered in western populations of the Humifusa clade also suggest that those diploid populations may be older (see Stebbins 1971, p. 157) than those of the southeastern U.S.; however, this could merely be a bias resulting from more limited sampling of western populations.
The various morphotypes of tetraploid O. macrorhiza in the western U.S. likely arose from southwestern diploid populations but subsequently spread in all directions after formation. Tetraploid O. macrorhiza appears to have arisen numerous times, given that several morphotypes exist throughout its range. However, only two diploid mor-photypes are known to exist (eastern Texas and southeastern New Mexico), suggesting that other ancestral diploids may have since gone extinct or have not yet been found, or that polyploid taxa exhibiting unique, derived characters were partly responsible for the origin of certain morphotypes, which have no diploid counterparts. Stebbins (1971) suggested that there are several degrees of maturation of polyploid complex formation (i.e., initial, young, mature, declining, relictual), which may be deduced by comparing the relative geographic distribution of polyploids versus diploids. By these criteria, Opuntia humifusa s.l. and O. macrorhiza s.l. may represent a mature polyploid complex. The diploid taxa are less common than polyploids and are largely restricted in distribution, whereas the polyploid taxa are much more widespread. Stebbins (1971) also proposed that mature polyploid complexes are relatively young, derived during the Plio-or Pleistocene epochs. This scenario would place polyploid formation in the Humifusa clade at the same time as Pleistocene megafauna. Thus, frequent environmental disturbances associated with glacial and interglacial cycles could have mediated the repeated contact of divergent diploid taxa leading to polyploid formation. Migrating herbivores would have then dispersed those polyploid products over large geographic areas (Jansen 1986). Divergence time estimation of the Humifusa clade places the origin of the clade in the late Pliocene to early Pleistocene (LCM, RP, PG, WSJ, PSS, DES unpubl. data), in agreement with this scenario. The occurrence of only polyploid individuals in previously glaciated areas of the U.S. provides further evidence for their subsequent spread into those available niches following the last glacial maximum.
Many polyploid populations of O. humifusa s.l. and O. macrorhiza s.l., especially in the eastern U.S., are largely isolated from one another and from diploid populations, suggesting that polyploid formation is not ongoing, at least on such a large scale as during the Pleistocene or immediately after the last glacial maximum.  (Parfitt 1991, Loik and Nobel 1993, Ribbens 2008, with diploid relatives (e.g., O. polyacantha var. arenaria) restricted to the southwestern U.S. (Parfitt 1991, Pinkava 2002. Thus, certain polyploid taxa appear to be more cold-resistant than their southerly diploid relatives (and presumed progenitors). Opuntia humifusa s.l. from northern areas of its distribution can withstand temperatures of -20°C (Nobel and Bobich 2002). However, the cold tolerance of diploid taxa has not been tested. Certain polyploid taxa of the Humifusa clade may therefore be better adapted to adverse environmental conditions than their diploid progenitors, which may partly explain their wide distribution relative to their diploid counterparts.
Agamospermy -The tetraploid O. cespitosa (an entity within O. humifusa s.l.; see Table 1) produces viable seed in the absence of outcrossing (Majure pers. obsv.), so this taxon is either self-compatible, which is common in Cactaceae (Rebman and Pinkava 2001), or agamospermous. Agamospermy is commonly associated with polyploidy (Stebbins 1950, DeWet and Stalker 1974, Harlan and DeWet 1975, Lewis 1980, Grant 1981, Whitton et al. 2008) and has been reported in numerous polyploid Opuntia species as well (Reyes-Agüero et al. 2006, Felker et al. 2010), including O. humifusa s.l. and O. stricta (Naumova 1993. Agamospermy would account for the high level of morphological variation observed among polyploid populations, as a result of the maintenance of a specific genotype within a given population through the lack of recombination (DeWet and Stalker 1974). Some agamic complexes also have wider distributions than their diploid progenitors (Babcock andStebbins 1938, Stebbins 1950), as do certain polyploid taxa in this study.
Autopolyploidy vs. Allopolyploidy -The mechanism by which Opuntia polyploids are formed (auto-vs. allopolyploidy) is unclear. Unreduced gametes have frequently been found in meiotic analyses of Cactaceae (e.g., Pinkava et al. 1977, Pinkava and Parfitt 1982, Pinkava et al. 1985. Unreduced gamete formation coupled with interspecific hybridization (allopolyploidy) likely is a major factor in polyploid formation within the genus, given that Opuntia is renowned for hybridization (Benson 1982, Grant and Grant 1982, Pinkava 2002, Griffith 2004. It is probable that unreduced gamete formation within a single species (autopolyploidy) also plays a role in the formation of polyploids. Autopolyploids have been discovered in Cactaceae (Pinkava et al. 1985, Sahley 1996, Hamrick et al. 2002 and may be more common than is suspected. Opuntia humifusa as currently circumscribed consists of numerous morphological entities, which are either diploid or tetraploid; those populations differing in ploidy are generally geographically well separated from one another. It is evident from our phylogenetic analysis (Fig. 4) that O. humifusa is polyphyletic. Considering morphological and genetic data, it is likely that tetraploid O. humifusa is of allopolyploid origin. However, the pattern in O. pusilla is different, with populations of diploids found in close proximity to populations of triploids and tetraploids (Fig. 3). This evidence, plus morphological similarity among ploidal levels, suggests possible formation of autopolyploids. This same pattern is seen in other autopolyploid taxa (Lewis 1967, Nesom 1983, although there are exceptions to this pattern (Stebbins 1950, Soltis 1984, Husband and Schemske 1998. Molecular phylogenetic analysis (Fig. 4) and morphological characters (LCM, RP, PG, WSJ, PSS, DES unpubl. data; see Fig. 1E-G) of O. pusilla also do not support an interspecific hybrid origin for the different ploidal levels herein observed for this species, although more variable molecular markers, cytogenetic work, and more detailed morphological analyses are needed to appropriately address this question.
Throughout the distribution of the most common polyploid taxa, there also are polyploid populations that appear to be introgessive products of hybridization with other polyploids. For instance, in Michigan, Wisconsin, and western Illinois, certain populations display characters of both O. cespitosa and tetraploid O. macrorhiza (see Majure 2010, Fig. 1 Each one of the areas in which these intermediate plants occur appears to be a region of secondary contact, where polyploid taxa have introgressed to form new polyploid morphotypes that exhibit characters of both of the putative parents. In the eastern U.S., most populations are represented by only one morphotype and thus appear to be morphologically stable (except for typically variable characters such as spine number; see Rebman and Pinkava 2001), indicating that hybridization is not ongoing among genomically distinct polyploid taxa. In contrast, in central Arkansas and populations farther west, more than one species and/or morphotype may be encountered within a given population. Also, in many coastal populations throughout the southeastern U.S., more than one species may be encountered, and putative hybrid taxa are sometimes observed.

Conclusions
Members of the Humifusa clade are found throughout most of the continental U.S., with no obvious breaks or disjunctions in distribution patterns until detailed analyses of chromosome number were carried out. Our analyses indicate that diploid taxa in the Humifusa clade are presently confined to the southwestern and the southeastern U.S., which likely represent Pleistocene refugia for these taxa. Polyploid taxa of O. humifusa s.l. and O. macrorhiza s.l. were likely formed when diploids from these two refugia came into contact during interglacial cycles of the Pleistocene. This scenario is supported further by phylogenetic analyses, in which two clades correspond to these two diploid refugia, and polyploid taxa are found in either clade. Polyploid taxa likely also contributed to the diversity of polyploid morphotypes through secondary contact and introgression with other polyploids. After the end of the last glacial maximum, open niches would have been readily available for colonization by polyploid taxa produced towards the leading edge of the expansion and distribution of the Humifusa clade. These polyploids subsequently dispersed throughout most of the continent and occupied all suitable habitats available after glacial retreat, accounting for the distribution that we see today. Distributional success was enabled by the extreme cold tolerance displayed by many of the polyploid taxa, which allowed them to colonize more northern areas presumably unsuitable for diploid taxa.