Aliso: a Journal of Systematic and Evolutionary Botany Species Status of Sclerocactus Brevispinus, S. Wetlandicus, and S. Glaucus: Inferences from Morphology, Chloroplast Dna Sequences, and Aflp Markers

We examine patterns of variation in 12 continuous morphological traits, chloroplast DNA sequences from 10 intergenic spacer regions and Amplified Fragment Length Polymorphism (AFLP) genetic markers in Sclerocactus glaucus sensu lato (5S. brevispinus, S. glaucus, and S. wetlandicus), a complex that historically has been considered conspecific and afforded protection under the Endangered Species Act. This complex is considered to represent three different species by some authors. We describe the expected patterns of morphological, DNA, and AFLP variation under the conditions that (a) the complex is a single species, and (b) that there are three antonymous species. We show that morphological evidence is consistent with the presence of three significantly different morphological species. Chloroplast DNA sequences provide evidence that the populations of S. glaucus (restricted to Colorado) are a lineage distinct from the populations of S. brevispinus and S. wetlandicus (restricted to Utah). AFLP genetic markers reveal significant genetic divergence among S. brevispinus, S. glaucus, and S. wetlandicus. Equally important, there is greater divergence among species than among populations within the species. The three sources of evidence all support the presence of three species and not a single species. These results indicate that protection of S. glaucus as a threatened species under the Endangered Species Act, as historically prescribed, includes populations of three species, two in Utah (S. brevispinus and S. wetlandicus) and one in Colorado (S. glaucus).

Near Myton, Utah, is found S. brevispinus (Fig. 1), possessing the most extreme morphological form of the three taxa. Sclerocactus brevispinus is depressed-globose, either lacking all central spines or if present they are solitary, very short (ca. 0.5-5.0 mm) and hooked, and have small, broad, pale pink to purple flowers (Heil andPorter 1994, 2003; see also Hochstä tter 1990Hochstä tter , 1992Hochstä tter , 1993. It has been purported by Welsh (1987) that these individuals were the consequence of phenotypic plasticity, long remain in a juvenile stage, and providing the suggestion that these populations represent a pedomorphic form (a developmental mutation in which sexual maturity, in this case flowering, occurs when plants appear morphologically similar to juvenile individuals).
Sclerocactus glaucus was described from plants collected by C. A. Purpus on adobe clay hills in Delta County, Colorado, in 1892 (Fig. 2). The original description was very brief, only describing the flowers as pink; however, the name has consistently been applied to the Sclerocactus growing at the foot of Grand Mesa above the Gunnison River. These plants are moderately sized and have globose to sub-cylindrical stems (3-28 cm) with 1-4 straight or hooked central spines and narrow, red-purple flowers. This species has been particularly controversial both nomenclaturally and taxonomically (see below).
The third species in this group is S. wetlandicus (Fig. 3). This species has stems that are globose to cylindrical (3-15 cm) bearing 3-5 straight, unhooked or curved central spines. It was distinguished from S. glaucus based upon seed coat features. The testa of S. wetlandicus has cells that are clearly flattened, whereas those of S. glaucus are rounded and often referred to as papillate (Hochstätter 1989). In addition, S. wetlandicus is geographically isolated from S. glaucus, being restricted to the Uintah Basin of Utah, along the Green, White, and Strawberry rivers.
Historically, the distribution of S. glaucus was considered to incorporate two disjunct areas: (1) the Colorado and Gunnison River valleys of west-central Colorado and (2) Uintah Basin of northeastern Utah, on the Colorado Plateau (Atwood and Reveal 1975;Colorado Native Plant Society 1989;Weber 1987;Welsh et al. 1987Welsh et al. , 1993. That is, all three species were treated as a single taxon. In fact, some treatments (e.g., Welsh et al. 1987) considered the entire collective to be unworthy of taxonomic recognition, treating them as conspecific with S. whipplei.
The segregation of S. brevispinus and S. wetlandicus from S. glaucus has found support from comparative trnL-trnF DNA sequencing (Porter et al. 2000). That study found S. glaucus to share more recent common ancestry with S. whipplei and S. wrightiae L.D.Benson than with S. brevispinus or S. wetlandicus that were sister taxa. At the same time it is important to recognize that a morphological cline has been suggested to exist along Pariette Wash, from the Myton populations of S. brevispinus to the type locality of S. wetlandicus. Across this cline, morphology has been suggested to shift from the typical S. brevispinus morphology to typical S. wetlandicus morphology. Whether this purported clinal variation represents secondary contact and hybridization between two formerly isolated species, or primary contact of a peripheral, diverging portion of a single species, is not known.  (Fig. 2a,b) and S. wetlandicus Hochstä tter (Fig. 3a,b) A comparative study of quantitative morphology in Sclerocactus (which did not include the seed traits discussed by Hochstä tter 1989) revealed complex patterns of morphological similarity (Heil and Porter 1987). Although nearly all sampled populations showed some differences from one another, no significant differences in stem, spine, and floral features of S. glaucus and S. wetlandicus were found (Heil and Porter 1987). At the same time, there were significant differences in these same traits between S. brevispinus and both S. glaucus and S. wetlandicus. This evidence was used to support species status of S. brevispinus (Heil and Porter 1994).
Currently, those who work with the genus are left with a variety of alternative treatments provided by contemporary systematists. Heil andPorter (1994, 2003) believe that S. glaucus s.l. represents three different species. They suggest that S. glaucus s.s. is restricted to Colorado. In the Uintah Basin of Utah are two species: S. brevispinus and S. wetlandicus (Heil and Porter 2003). Hochstä tter (1989,1993) recognizes two species, S. glaucus (of Colorado) and S. wetlandicus (of Utah). The taxon that Heil and Porter treat as S. brevispinus is considered by Hochstä tter to represent a different subspecies of S. wetlandicus, i.e., S. wetlandicus subsp. ilseae Hochstä tter (Hochstä tter 1995). Welsh et al. (2003) provide another alternative treatment in which there are two taxa, but both are varieties of S. whipplei. One taxon is S. whipplei var. ilseae (Hochstä tter) S.Welsh, which corresponds to S. brevispinus and/or S. wetlandicus subsp. ilseae. The other is S. whipplei var. glaucus (K.Schum.) S.Welsh, which corresponds to S. glaucus and S. wetlandicus (subsp. wetlandicus sensu Hochstä tter 1993). This conflict in species boundaries presents a further problem, given that these taxa all have protection under the Endangered Species Act.

Study Goals
During this recent period of taxonomic re-evaluation and change, the United States Fish and Wildlife Service (USFWS) has been charged with the protection and recovery of S. glaucus, a species protected under the Endangered Species Act as a threatened species (USFWS 1979(USFWS , 1985. All of the recently named species, i.e., S. brevispinus and S. wetlandicus, have until recently been treated under the rubric of ''S. glaucus,'' as has been the tradition of Utah botanists (Atwood and Reveal 1975;Welsh et al. 1987Welsh et al. , 1993. This has afforded federal protection to all three named taxa, without the need of petitioning for federal listing of S. brevispinus and/or S. wetlandicus. The difficulty with this approach is that the numbers of populations of the three species combined may be high enough to question the need for protection; or, mitigations may impact one taxon more severely. In 2007(USFWS 2007, 2009a, S. brevispinus and S. wetlandicus were designated threatened species. However, the difference in opinion concerning taxonomy has left open to challenge the very existence of some of the taxa. Sound conservation and species management requires sound taxonomy. It is frequently argued that taxonomy is largely the opinion of those who practice the naming of species. Both scientists and nonscientists alike have often suggested that whether a species is carved away from another (splitting) or two species are agglomerated together (lumping) is more an art than a science, being prone to subjectivity. However, speciation events result in characteristic patterns among populations, providing testable expectations for species. Species are evolutionarily independent, cohesive groups of populations, which are genetically differentiated from one another. Given this, we would expect that: (1) different species would be significantly different genetically and minimally possess diagnostic differences in allele frequencies, and (2) as a consequence we would usually observe significant differences in morphology, physiology, and/or reproductive features. Such properties of species can be tested (and potentially falsified) using comparative, population genetic, and phylogenetic methodologies.
The purpose of this study is to examine the morphology, phylogenetic relationships, and patterns of genetic variation within and among populations of S. brevispinus, S. glaucus, and S. wetlandicus. If these three taxa represent a single, undifferentiated species, then we expect that chloroplast gene phylogenies will display all samples coalescing together, without respect to taxon names, rather than forming three different clades. If, on the other hand, they represent different species and have been reproductively isolated for a sufficiently long period of time, then we would expect populations of S. glaucus to coalesce (form a clade), those of S. brevispinus to coalesce, and those of S. wetlandicus to coalesce. We would further expect to find fixed mutations, unique to each of the species, provided sufficient time has occurred since speciation for mutations to become fixed in all populations. Similarly, if these three taxa represent a single, undifferentiated species, then we expect genetic variation to be uncorrelated with species assignment and be highly similar across all of the populations. If they represent different species, then we would expect genetic variation to be highly correlated with species assignment. In addition, we expect genetic divergence among the species. Here, we test these expectations.

METHODS
Floral buds from S. glaucus s.l. (including S. brevispinus, S. glaucus, and S. wetlandicus) were collected from eight wild populations located in Utah and Colorado (Table 1, Fig. 4). At the time of collection, latitude and longitude were recorded and a color digital photograph was made of each sampled plant. Floral tissues were dried in silica gel. Samples were stored in a 220uC freezer at Rancho Santa Ana Botanic Garden until DNA extraction.
Floral buds of S. brevispinus were collected from two small populations (N 5 16-60 individuals; NF and GW, see Brand. Two individuals of Sclerocactus wetlandicus cooccurred with S. brevispinus at the GW site. Samples of S. wetlandicus were collected from three large populations (N . 200 individuals) in Utah. The first population (PW) was located at the type location for S. wetlandicus on the slopes above Pariette Wetland, southwest of Myton, Utah. The second S. wetlandicus population (GR) was on an oil shale bench on the west bank of the Green River. The third population (BPP) of S. wetlandicus was outside Bonanza, Utah, southwest of the power plant. Samples of S. glaucus were collected from three large populations (N . 600 plants) in western Colorado. The first population (GP) was above a gravel pit along Gunnison River in the Escalante Canyon east of Grand Junction, Colorado. The second population (RM) was adjacent to a roadside and power line cut at Reeder Mesa. The third population (PR) grew on white sandy soil with a pediment of black volcanic rock at Pyramid Rock on the slopes above the Colorado River.
DNA was extracted from both ovary walls and perianth using a modified CTAB protocol. Extractions included three washes, the first with CTAB and then Nucleon PhytoPure Resin (Tepnel Life Sciences plc for Amersham Biosciences, Little Chalfont, UK) and chloroform, the second with CTAB and 1% w/v caylase (Cayla-InvivoGen, Toulouse, France) and then chloroform, and the third with 75% ethanol. DNAs from five individuals were used to screen 12 chloroplast markers for and blue squares identify sites of S. wetlandicus. Each site is denoted using the population codes described in Table 1. genetic variation. The five individuals were: one sample of S. brevispinus from Pariette Wash, one sample of S. wetlandicus from GR, and three samples of S. glaucus representing populations from GP, RM, and PR. The 12 rapidly evolving chloroplast DNA regions surveyed for variation included 10 intergenic spacer regions [IGSRs] (petA-psbJ, psbk-trnS, psbM-trnD, rpob-trnC, trnC-trnD, trnGCU-trnG2S, trnFM-trnUGA, atpF-atpH, trnT-trnD, and trnQ-psbk), as well as atpF, and rpl16.
The chloroplast regions of 25 samples were amplified using the polymerase chain reaction (PCR) on a PTC-100 Thermal Controller (MJ Research, Inc., Watertown, MA): 94uC for 4 min, then 35 cycles of 94uC for 45 s, 56uC for 45 s, and 72uC for 2 min-30 s, concluding with 94uC for 45 s and 72uC for 5 min. A negative control excluding DNA was used in each set of reaction to detect contamination or false positives. The samples included four individuals from Pariette Wash, three individuals from GW, two from PW, four from BPP, four from GP, four from RM, and four from NF. The PCR reactions were cleaned using PEG precipitation, then subjected to the following Big-Dye TM (Applied Biosystems/Life Technologies, Foster City, CA) cycle sequence program for 35 cycles: 96uC for 30 s, 48uC for 15 s, 60uC for 4 min. The cycle sequencing product was placed on a 96-well plate and sequenced in a 3130xl sequencer (Applied Biosystems/Life Technologies). Sequences were aligned by eye in Se-Al vers. 2.0a11 (http://tree.bio.ed.ac.uk/).
Four samples representing each population of Sclerocactus were used to screen Amplified Fragment Length Polymorphism (AFLP) primers for population genetic analysis. Each DNA sample was subjected to restriction digestion using the AFLP Core Reagent Kit, Invitrogen, and EcoR1/Mse1 endonucleases. The restriction digestion was accomplished using the PTC-100 Thermal Controller following the suggested incubation period of 2 hrs at 37uC. Restriction digestion was inactivated by subjecting the mixture to 15 min at 70uC. Ligation of the adapters was accomplished using the restriction digest mixture subjected to 20uC for 2 hrs using the PTC-100 Thermal Controller. A 1:10 dilution of the ligation mixture was made and then subjected to a pre-amplification run using the PTC-100 Thermal Controller with the following reaction with 20 cycles: 94uC for 30 s, 56uC for 60 s, 72uC for 60 s. A 1:50 dilution of the pre-amplification mixture was made for the final selective AFLP amplification. The diluted mixture was then paired with a fluorescently labeled EcoR1-AAC primer and either Mse1-CAC or Mse1-CAG, in separate reaction. The AFLP amplification products were then run out on an ABI 3130xl sequencer and analyzed using GeneMapper software (Applied Biosystems/Life Technologies). Three replicons of each of the above reactions were completed and compared to ensure the alleles were present in each replicon, and that the results were reproducible. A 600 bp-standard (DG2611; Promega, Madison, WI) was used along with the AFLP amplifications for the sequencer run. Only fragments between 60 bp and 600 bp long were called as peaks by the GeneMapper software. The cutoff for allele calls was set at a peak height of 100. Sixteen individuals from each of the eight populations were surveyed using both Mse1-CAC and Mse1-CAG.
Allelic variation was also investigated to estimate the likely number of ancestral populations giving rise to the standing genetic variation, using Bayesian model-based clustering for multilocus genotype data in Structure vers. 2.3.2 (Pritchard et al. 2000;Falush et al. 2003Falush et al. , 2007Hubisz et al. 2009). Data were analyzed both as diploid recessive data and as haploid recessive data to contrast the results, given that the AFLP markers may represent both plastid and nuclear markers. Populations were analyzed with both naïve and populationinformed clustering, k 5 1-10, with 50,000 generations burn-in and posterior sampling of 50,000 generations, running 10 replicates of all analyses.
Twelve morphological characters (Table 2) were measured from 35 individuals at populations NF, PW, BBP, RM, and PR. Only continuous measurements (i.e., no meristic or qualitative traits) were used in analyses. Measurements were analyzed by PCA, DFA, and MANOVA using SPSS 11.0.2.

RESULTS
PCA analysis of 12 continuous morphological features finds six factors, each of which explain a significant proportion of variance ( Table 2). The greatest variance proportion is associated with a factor that is characterized by a high correlation among stem length, stem width, spine length, and flower length (e.g., Factor 1, Table 2). However, this factor does not strongly aid in discriminating the three taxa. MANOVA of factor loadings demonstrates that there is a significant difference in the factors among the species (Table 3). Most of the variance associated with species differences is attributed to Factors 2 and 3. These differences are evident in Fig. 5, which shows the morphological isolation of S. brevispinus, based on Factors 2 and 3. Table 4 provides the results of a Bonferroni analysis of the factors used in the MANOVA. The Bonferroni analysis reveals which species are significantly different based on particular factors. For example, S. brevispinus and S. glaucus are significantly different only in Factor 2 ( Table 4).
The stepwise DFA of S. brevispinus, S. glaucus, and S. wetlandicus required the addition of only seven continuous characters to discriminate among the three species (Table 5, Fig. 6). Even so, the separation among the three species is similar to the PCA, but with greater separation of S. wetlandicus and S. glaucus.
Of the 12 chloroplast regions examined, only one of the markers showed genetic variation. This was the petA-psbJ IGSR (Appendix 1). This region ranges between 570 and 598 nucleotides and includes a 29-base pair long indel (insertiondeletion feature). All of the individual samples of S. glaucus possess this 29-base pair segment of DNA; but in both S. brevispinus and S. wetlandicus it is absent. This indel feature was included in the phylogenetic analysis of petA-psbJ by adding a single binary character at the end of the DNA sequence matrix (see Appendix 1).
Parsimony analysis of the petA-psbJ region resulted in a single most-parsimonious tree (Fig. 7) of five steps, CI 5 1.000, RI 5 1.000. The consistency index (CI) and retention index (RI) indicate that there is no homoplasy in this data set. The tree unambiguously separates all S. glaucus samples from those of S. brevispinus and S. wetlandicus. This region does not differentiate S. brevispinus from S. wetlandicus; however, two S. brevispinus individuals from population GW share a unique mutation.
Three replicons (replicate runs) of fluorescently labeled EcoR1-AAC/Mse1-CAC primers and fluorescently labeled EcoR1-AAC/Mse1-CAG primers were completed and compared for the 10 individuals from each of the sampled populations ( Table 1). The replication ensures that the alleles compared were consistently present, and the results are reproducible. We found 167 alleles. Analysis of molecular variance (AMOVA) of 167 AFLP markers, sampled from populations of S. brevispinus, S. glaucus, and S. wetlandicus, reveals that there is a significant degree of genetic divergence among the three species (Tables 5,  6). Although the greatest genetic diversity lies within species, there is greater divergence among species than among populations of the same species (Table 7).
The estimation of the number of populations using Structure vers. 2.3.2 produced different inferences depending upon the Table 2. Factor loadings from the orthogonal principal component analysis of 12 morphological characters measured for the samples of Sclerocactus brevispinus, S. glaucus, and S. wetlandicus. The bold values call attention to the traits that primarily contribute to each of the factors. Eigenvalues (c) and proportion of variance (s 2 Prop) contributed by each of the factors is provided for each factor. Morphological characters surveyed (measured in mm): stem length (stemL), stem diameter at 1/2 length (stemW), lower (''hooked'') central spine length (cspineL), flower length (flrL), flower diameter at anthesis (flrdia), outer perianth lobe length (sepL), outer perianth lobe width (sepW), inner perianth lobe length (petL), fruit length (frtL), fruit diameter at 1/2 length (frtW), seed long axis length (seedL), and seed short axis length (seedW). a Design: Intercept + species; b Exact statistic; c R-squared 5 0.055 (adjusted R-squared 5 0.038); d R-squared 5 0.587 (adjusted R-squared 5 0.580); e R-squared 5 0.406 (adjusted R-squared 5 0.396); f R-squared 5 0.058 (adjusted R-squared 5 0.041) assumptions associated with analyses. General patterns found in all estimations are: (1) most members of populations of Sclerocactus brevispinus cluster together in naïve clustering and most members of S. wetlandicus cluster together in a different cluster; (2) admixture is present in Sclerocactus brevispinus, involving S. glaucus and to a lesser extent S. wetlandicus. Admixture is also present in populations of S. wetlandicus, involving S. glaucus; however, the population at Bonanza, Utah, shows significant admixture involving S. brevispinus. The naïve estimation, assuming diploid populations, has a maximum likelihood at k 5 6 ( Fig. 8A) with a mean log-likelihood of 22940.55, averaged over 10 Markov Chain Monte Carlo (MCMC) runs. The mean log-likelihood value at k 5 6 is significantly higher than other values of K; however, the likelihood values begin to plateau at k 5 4 (lnL 5 23018.2). The estimated number of diploid ancestral populations, informed by the hypothesized three-species membership similarly maximizes at k 5 6 (Fig. 8B), with a mean log-likelihood of 23949.9, averaged over 10 MCMC runs. As was the case for the naïve clustering, the likelihood began to plateau at k 5 4 (lnL 5 23988.9), and the actual number of sampled populations, k 5 8 (lnL 5 23982.3), is not significantly different from k 5 6. Fixation indices (w ST ) for the three taxa based on Bayesian inference are relatively high (Table 8), leading to the deduction that the three species are reproductively isolated from one another. DISCUSSION We have examined patterns of morphological variation, divergence in chloroplast sequences, and patterns of genetic variation within and among populations of S. brevispinus, S. glaucus and S. wetlandicus. The null hypothesis, i.e., these three  taxa represent a single, undifferentiated species, leads to three expectations: (1) the three named taxa should be morphologically cohesive, or represent a continuum of morphological variation; (2) chloroplast gene phylogenies should show that all samples coalesce together, without respect to taxon naming, or show a branching pattern independent of taxon naming; (3) genetic variation should be uncorrelated with species assignment and be highly similar across all of the populations or, at least, there should be greater divergence among populations of the same taxon than among the assigned species. By contrast, our alternative hypothesis, i.e., S. brevispinus, S. glaucus, and S. wetlandicus represent differentiated species, leads to three contrary expectations: (1) the three named taxa should be morphologically distinct and thus can be discriminated on the basis of morphological variation; (2) chloroplast gene phylogenies should show populations of S. glaucus coalescing, S. brevispinus coalescing, and S. wetlandicus coalescing, or show a branching pattern that is in some way consistent with taxon naming; (3) genetic variation should be correlated with species assignment, and species should show significant genetic divergence, i.e., there should be greater divergence among species than among populations of the same species. If the alternative hypotheses-and thus the three expectations-are true, then by any criterion used for recognizing species (morphological, phylogenetic, genetic isolation), S. brevispinus, S. glaucus, and S. wetlandicus would be considered different species.
The different markers used in this study possess the potential to provide different information or aspects of information concerning the species status of our three study taxa. Markers such as chloroplast DNA sequences are known to evolve slowly, providing information about more ancient events, but may provide little or no information concerning more recent speciation events, because often there is insufficient or no variation in the DNA sequences. By contrast, rapidly evolving molecular markers such as allozymes, microsatellites, or AFLPs are variable enough to provide information about populations and closely related species, but are often too variable to be useful for understanding relationships beyond closely related species. Morphological data can represent a powerful inference tool for discrimination of taxa; however, failure to discriminate taxa may not necessarily reflect that taxa cannot be discriminated: any morphological analysis is limited by the morphological traits included in the analysis. If the set of included traits fails to separate taxa it may be either because the two taxa do not differ in the particular traits, or that the two taxa are in fact morphologically identical. Even given this reality, our data provide a very consistent picture of phylogenetic, morphological, and genetic relatedness. These patterns are consistent with our alternative hypothesis, that our sample represents three species.
Morphological data (Fig. 5, 6) provide evidence that S. brevispinus, S. glaucus, and S. wetlandicus are morphologically different (F 5 15.771, P 5 0.000) and distinct from one another. In fact, S. brevispinus is the most distinctive of the three, significantly differing from S. glaucus and S. wetlandicus in five of the 12 traits examined: central spine length, flower length, flower diameter, seed length, and seed width. Although Fig. 6. Bivariate plot from discriminant function analysis Functions 1 and 2, based on 12 continuous morphological traits measured from Sclerocactus brevispinus, S. glaucus, and S. wetlandicus. VOLUME 30, NUMBER 2 Species Status in Sclerocactus there has been a longstanding debate concerning the recognition of S. brevispinus as a species, it is one of the most distinctive taxa in the genus in terms of morphology. In addition, we find significant differences between S. glaucus and S. wetlandicus that are morphologically very similar. Sclerocactus glaucus and S. wetlandicus differ from one another in flower length, outer perianth segment length, inner perianth length, fruit length, seed length, and seed width. The patterns of morphological variation are consistent with the hypothesis that the three are different species.
Since the chloroplast genome is maternally inherited and nonrecombining, sequence data can be compared and interpreted to assess the phylogenetic and phylogeographic relationships among S. glaucus, S. wetlandicus, and S. brevispinus. These data reveal a 29-base difference in length between S. glaucus that has the 29-base span and S. brevispinus and S. wetlandicus that lack it. Similarly, mean evolutionary distances, based upon Tamura and Nei (1993) distance of chloroplast DNA sequences, are greatest in comparisons involving S. glaucus, i.e., S. glaucus-S. wetlandicus 5 0.00266; S. glaucus-S. brevispinus 5 0.00318. This is considerably larger than the mean evolutionary distance between S. brevispinus and S. wetlandicus of 0.00051 (see also Fig. 7). The chloroplast DNA sequences unequivocally support the evolutionary separation of the Colorado populations of S. glaucus from the Utah populations of S. brevispinus and S. wetlandicus. The S. glaucus lineage has been reproductively isolated for a sufficiently long period of time that length differences and point mutations could evolve and become fixed in all of the sampled individuals, but remain absent from S. brevispinus and S. wetlandicus. While this is consistent with species status for S. glaucus, the sequence data lack sufficient variation to make any inference concerning species status of S. brevispinus and S. wetlandicus. The chloroplast petA-psbJ IGSR data (Fig. 7) provide nearly identical inference as does the trnL-trnF region (Porter et al. 2000). These new data differ in that two members of S. brevispinus (from population GW) share a unique chloroplast type, derived from the common type in other S. brevispinus and S. wetlandicus. In addition, different chloroplast variants are found in different populations of S. glaucus, suggesting population differentiation in that species.
Our examination of genetic variation using AFLP markers reveals that there is significant genetic divergence among population samples of S. glaucus, S. wetlandicus, and S. brevispinus (Table 6, 7), based on direct measures. Unlike the chloroplast DNA sequence data, AFLP markers show significant (P 5 0.010) evolutionary divergence (w GT 5 0.3018) between S. brevispinus and S. wetlandicus. Further, there is three times the divergence between S. brevispinus and S. wetlandicus as there is among populations within each species. This points to a significant period of isolation between S. brevispinus and S. wetlandicus. It is difficult to imagine such a degree of divergence developing if there were long-term gene flow between the two, given that they are parapatric in distribution and S. brevispinus is represented by a single metapopulation along a 10-mile stretch of Pariette Draw (species census numbers are estimated at 8000-12,000; USFWS 2007). Similar divergences between species are also revealed in the Bayesian estimates of F ST (Table 8). This bolsters the hypothesis that S. glaucus, S. wetlandicus, and S. brevispinus are different, genetically differentiated species.
Konnert (2005) concluded that S. glaucus and S. parviflorus were the most similar, while the differences between S. wetlandicus subsp. ilseae (5S. brevispinus) and S. wetlandicus subsp. wetlandicus were so slight that they could be attributed to different individuals of the same population. However, this was based on examination of a single individual from each of 24 species or subspecies in Sclerocactus, using nine enzyme systems. In contrast, we found greater allele frequency divergence between S. brevispinus and S. wetlandicus (0.0760) than between S. brevispinus and S. glaucus (0.0283) when examining diploid populations (k 5 3), using Structure vers. 2.3.2.
While the AFLP data are supportive of the hypothesis that there are three species, there is also strong evidence for admixture. Perplexingly, the source populations for admixture are not those that are geographically proximal; rather, they are the most distant. For example, populations of S. brevispinus show admixture involving the Colorado populations of S. glaucus. Similarly, the population of S. wetlandicus at Bonanza is characterized by admixture involving S. brevispinus, but other populations of S. wetlandicus do not show admixture. This may be due to the maintenance of ancestral genetic polymorphism rather than recent gene flow. The pollinators of both S. brevispinus and S. wetlandicus are ground-dwelling bees of the family Halictidae (Tepedino et al. 2010). These insects do not have large home ranges. In addition, the fruits of both species are dry (not dispersed by birds), and the seeds fall to the base of the plants (not dispersed by ants) to be moved only by rainfall and wind. As a result, seed dispersal is limited.
While we have argued that the AFLP data are consistent with the three-species hypothesis, it is important to recognize that the number of populations estimated with Structure vers. 2.3.2 was six rather than three (the number of hypothesized species) or eight (the actual number of populations sampled). The six populations identified by an informed population prior (Fig. 8B) discriminates the two populations (NF and GW) of S. brevispinus with evident admixture between these two populations. Similarly, the three populations of S. glaucus (GP, RM, and PR) are found to have significant differentiation. The admixture detected in these populations seems to represent markers from S. brevispinus and S. wetlandicus rather than from other populations of S. glaucus. As noted above, this seems more likely to be the result of the persistence of genetic markers from common ancestors than from recent gene flow between these two species, given the geographic isolation of the populations.
One caveat regarding the AFLP data is that it suffers from limited sampling of individuals at the populations investigated. While reasonable population samples were acquired in the field, funding restrictions reduced that number of individuals analyzed significantly. Moreover, the AFLP markers show great variation, with 167 variable loci. This results in a high degree of noise in the data, more so than would be desirable. Another possible contributing factor to noise in the data is the presence of an unsampled species, S. parviflorus, which may be playing a genetic role.
We have examined patterns of variation in morphology, chloroplast DNA sequences, and AFLP markers in S. brevispinus, S. glaucus, and S. wetlandicus, a group of species that historically have been considered conspecific, under the name S. glaucus. By considering two sets of expected patterns of variation under the conditions that this group represents a  Table 7. Phi (w ) statistics derived from analysis of molecular variance (AMOVA) of Sclerocactus brevispinus, S. glaucus, and S. wetlandicus. We provide estimates of genetic differentiation among species (w GT ), among populations within species (w SG ), and among populations (w ST ).

Statistic
Value P w GT 0.1640 0.002 w SG 20.0774 0.001 w ST 0.0993 0.001 Table 8. F ST statistics derived from Bayesian estimation (Structure vers. 2.3.2) for Sclerocactus brevispinus, S. glaucus, and S. wetlandicus. We provide among-species estimates of genetic differentiation from AFLP markers assumed to represent haploid and diploid data, k 5 3, informed by species membership. Estimates of expected heterozygosity (H E ) are provided parenthetically.  Aligned DNA sequence file of the petA-psbJ chloroplast intergenic spacer region. Sample acronyms correspond to those in Table 1. The length positions of each nucleotide are displayed in brackets above the sequences. Within the sequences, dashes indicate an insertion/deletion, where one or more nucleotides are absent. Following each line of sequence data is the cumulative number of nucleotides, in brackets.