Morphological Evolution and Systematics of Synthyris and Besseya (Veronicaceae): A Phylogenetic Analysis

Phylogenetic analyses are used to examine the morphological diversity and systematics of Synthyris and Besseya. The placement of Synthyris and Besseya in Veronicaceae is strongly supported in parsimony analyses of nuclear ribosomal ITS DNA sequences. Parsimony and maximum likelihood (ML) criteria provide consistent hypotheses of clades of Synthyris and Besseya based on the ITS data. The combination of morphological characters and ITS data resolve additional clades of Synthyris and Besseya. The results show that Synthyris is paraphyletic to Besseya. In the monophyletic Synthyris clade, Besseya forms part of a Northwest clade that also includes the alpine S. canbyi, S. dissecta, and S. lanuginosa and mesic forest S. cordata, S. reniformis, S. platycarpa, and S. schizantha. The Northwest clade is the sister of S. borealis. An Intermountain clade, comprising S. ranunculina, S. laciniata, S. pinnatifida, and S. missurica, is the sister to the rest of the Synthyris clade. Constraint topologies are used to test prior hypotheses of relationships and morphological similarities. Parametric bootstrapping is used to compare the likelihood values of the best trees obtained in searches under constraints to that of the best tree found without constraints. These results indicate that topologies in which a monophyletic Synthyris is the sister of Besseya are significantly worse than the best ML tree in which Synthyris is paraphyletic to Besseya. Similarly, forcing either the monophyly of all taxa that have deeply incised leaf margins or those that have reniform laminas and broadly rounded apices results in trees that are significantly worse than the best ML tree, in which leaf margin incision and reniform laminas are homoplastic. We propose a new classification for Synthyris that emphasizes monophyletic groups. The new combination Synthyris oblongifolia is proposed. Synthyris Benth. and Besseya Rydb. are North American members of the Veronica L. alliance (Scrophulariaceae tribe Veroniceae or Veronicaceae sensu Olmstead et al. 2001). They are rhizomatous perennials that form rosettes of foliage leaves and axillary, racemose inflorescences annually (Hufford 1992a, b). All Synthyris and Besseya are distributed in western North America, except for the disjunct B. bullii that occurs in north central United States (Pennell 1933). Synthyris borealis, which is restricted to unglaciated regions of the Yukon and Alaska (Hultén 1937), is also disjunct from other Synthyris and Besseya (Pennell 1933). The systematics of Synthyris and Besseya has received considerable attention, including three taxonomic revisions in the 20th century. Initially, taxa recognized today as Synthyris and Besseya were discussed primarily in terms of alternative generic placements and circumscriptions. Bentham (1846) originally circumscribed Synthyris as a group of four species that had been placed in Gymnandra Pall. (G. bullii Eaton, G. rubra Douglas), Veronica (V. plantaginea James), and Wulfenia Jacq. (W. reniformis Douglas). Greene (1894) preferred to include this group of species in Wulfenia. Rydberg (1903) characterized Synthyris s. str. as having a corolla like that of Veronica and segregated the genus Besseya, which he characterized as possessing ‘‘an altogether different corolla, cleft to near the base into two distinct lips’’ (p. 278). Rydberg (1903) included also in Besseya a group of species (B. gymnocarpa, B. rubra, and B. wyomingensis) that he observed to lack a corolla. Nieuwland (1914, p. 188) regarded presence/ absence of a corolla as ‘‘a character deserving generic consideration’’ and segregated B. gymnocarpa, B. rubra, and B. wyomingensis as the new genus Lunellia Nieuwl. In the first comprehensive revision of this complex, Pennell (1933) recognized Wulfenia as Eurasian and distinct from the North American Synthyris-Besseya complex, which he emphasized has flowers and fruits like those of Veronica. His approach to generic circumscription largely followed Rydberg’s proposal in recognizing Synthyris (14 species) and Besseya (8 species; including Lunellia as a subgenus). Subsequent revisions of the complex, which have contributed to our understanding of character state variation and species circumscriptions, have continued to argue for the recognition of both Synthyris and Besseya (Hedglin 1959; Schaack 1983). When framed as an evolutionary hypothesis, however, both Pennell (1933) and Schaack (1983) suggested that Besseya was derived from Synthyris. Our understanding of evolutionary relationships is based on shared-derived features. Hufford (1993) noted that Synthyris lacks derived features independent of those it shares with Besseya; thus, a provisional hypothesis of their monophyly is warranted. An important objective of phylogenetic systematics is to identify monophyletic groups in classification and to provide revisions that reflect our understanding of monophyly. If Synthyris is paraphyletic to Besseya, then taxonomic revision to recognize monophyletic groups is warranted. The provisional hypothesis that Synthyris 2004] 717 HUFFORD & MCMAHON: SYNTHYRIS PHYLOGENY and Besseya are monophyletic requires testing in a phylogenetic analysis that includes their close relatives. Synthyris and Besseya have consistently been allied to Veronica and its relatives, a group recognized as the tribe Veroniceae (Bentham 1846; Pennell 1933, 1935; Thieret 1955). The first explicit evolutionary hypotheses for the relationships of Synthyris and Besseya were presented by Pennell (1933). He suggested an evolutionary sequence from Picrorhiza Royle ex Benth. to Wulfenia and, subsequently, to Veronica (and the related Veronicastrum Heist. ex Fabr. and Hebe Comm. ex Juss.; Pennell 1921). He questioned whether Synthyris and Besseya were more closely related to the more primitive Picrorhiza and Wulfenia or to Veronica; he argued that the closer evolutionary relationship was to the latter. Yamazaki (1957) illustrated Synthyris and Besseya as equally closely related to Veronica and a Hebe-Detzneria Schlt. ex Diels clade. Went (1958) suggested that Synthyris and Besseya were derived directly from New World Veronica. In a phylogenetic study of Veroniceae, Hong (1984) placed Synthyris as the sister of Besseya, and their clade was placed in his Veronica group as the sister of a clade comprising Pseudolysimachion Opiz, Veronica, Oligospermum D. Y. Hong (5 Veronica sect. Diplophyllum (Lehm.) Walp.), and Cochlidosperma (5 Veronica sect. Diplophyllum subsect. Cymbalariae (Benth.) Elenevskij). Hong (1984) recognized the paraphyly of Veronica and the problems it can create for a taxonomy based on monophyly. Kampny and Dengler (1997) found Synthyris and Besseya to be the sister clade of Veronica, but their phylogenetic analysis sampled few Veroniceae. Albach and Chase (2001) used sequence data from the internal transcribed spacers (ITS) of nuclear ribosomal DNA to infer phylogenetic relationships in Veroniceae. Their results placed Synthyris (Besseya was not sampled) as the sister of a clade that included species of Veronica and the Hebe complex, which bears a striking similarity to the earlier suggestion of Yamazaki (1957). In order to challenge existing hypotheses that Synthyris and Besseya are monophyletic, selected species of both genera are included here in a phylogenetic analysis of broadly sampled members of Veroniceae that represent major clades found by Wagstaff and Garnock-Jones (1998; Wagstaff et al. 2002) and Albach and Chase (2001). Synthyris has not been the subject of previous phylogenetic analysis. Hedglin (1959, p. 5) recognized ‘‘three major lines of evolution’’ in Synthyris, including: 1) S. reniformis (including S. cordata); 2) S. schizantha and S. platycarpa, and 3) the other species of the genus. Schaack (1983) proposed detailed phylogenetic scenarios for Synthyris and Besseya. Hufford (1993) provided a phylogenetic analysis of Besseya based on morphological data. The clades found in that analysis had limited robustness; for example, none had a bootstrap proportion greater than 90%. The most parsimonious trees identified a grade at the base of Besseya comprising B. bullii, B. rubra, and B. wyomingensis. Trees in which B. rubra and B. wyomingensis were constrained to be monophyletic (Nieuwland’s Lunellia and Pennell’s subgenus Lunellia) were only two steps longer than the most parsimonious. Results of that analysis also identified a monophyletic Southern Rocky Mountain clade that consisted of B. alpina, B. oblongifolia, B. plantaginea, and B. ritteriana. In addition to challenging the hypothesis of monophyly for Synthyris and Besseya in a broad phylogenetic analysis of Veroniceae, we provide a separate analysis to examine support for major clades and sister species in the two genera, in which we have sampled extensively among species. We examine previous evolutionary hypotheses, especially those for morphological characters. Finally, we include a taxonomy for the Synthyris-Besseya complex that emphasizes monophyly, as inferred from our results. MATERIALS AND METHODS Taxon Sampling. We examined first whether Synthyris and Besseya are monophyletic, which required a broad sampling of Veroniceae to test for alternative sister group relationships for the two genera. We used the earlier systematic and phylogenetic studies of Pennell (1921, 1933), Thieret (1955), Hong (1984), Wagstaff and Garnock-Jones (1998), Albach and Chase (2001), and Wagstaff et al. (2002) to guide taxon sampling. We included multiple exemplars from the Veronica I-IV clades found by Albach and Chase (2001). We used Wagstaff et al. (2002) to select multiple taxa from the Hebe, Heliohebe Garn.-Jones, Parahebe W.R.B. Oliv., Chionohebe B.G. Briggs & Ehrend. A, Chionohebe B, Leonohebe Heads, and Derwentia Raf. s.l. clades identified in their results. For this broad analysis, we sampled five species of both Synthyris and Besseya. These selected species of Synthyris and Besseya encompassed the subgeneric groups recognized by Pennell (1933) and Schaack (1983) as well as most of the geographic range and morphological diversity of the two genera. As the outgroups for this broad analysis of Veroniceae, we applied Aragoa abietina Kunth, Erinus alpinus L., Globularia salicina Lam., and Plantago lanceolata L., which were placed close to Veroniceae in the phylogenetic studies of Albach and Chase (2001), Wagstaff et al. (2002), and Rønsted et al. (2002) and in the Veronicaceae of Olmstead et al. (2001). Our second goal was to identify the major clades and sister species in Synthyris and Besseya. For parsimony analysis we sampled 20 accessions of Besseya and 18 accessions of Synthyris (Table 1) that encompassed the species recognized in the revisions of Pennell (1933), Hedglin (1959), and Schaack (1983). We sampled multiple accessions of several species of Synthyris and Besseya (Table 1) to assess the monophyly of species and to examine issues of circumscription raised in those revisions. As outgroups, we used seven other Veroniceae based on the results of our broad analysis of the tribe. DNA Sequences. New ITS sequences were obtained for all accessions of Synthyris and Besseya (Table 1). Total DNA was extracted from either herbarium or silica dried specimens of leaves using a standard CTAB procedure (Doyle and Doyle 1987). The ITS region was amplified and sequenced using the primers Nnc18s10 and C26A using a ‘‘touchdown’’ profile: 4 min at 948C, 5 cycles of 1 min at 948C, 1 min at 528C and 2 min at 728C, decreasing the annealing temperature by one degree each cycle, followed by 30 cycles using a 488C annealing temperature, and ending with a final extension of 5 min at 728C. The ITS sequences for Veroniceae other than Synthyris and Besseya as well as for outgroups outside of Veroniceae were obtained 718 [Volume 29 SYSTEMATIC BOTANY TABLE 1. Accessions of Synthyris and Besseya sampled for ITS sequences, with the collection sampled (herbarium voucher or publication) and GenBank accession number. Besseya alpina (A. Gray) Rydb. Hufford 242 (WS), AY483210; Hufford 244 (WS), AY483179. B. bullii (Eaton) Rydb. Hufford 277 (WS), AY48320; Hufford 335 (WS), AY483211; Hufford 609 (WS), AY483184. B. oblongifolia Pennell Hufford 235 (WS), AY483180. B. plantaginea (E. James) Rydb. Hufford 234 (WS), AY483181; Hufford 241 (WS), AY483208. B. ritteriana (Eastw.) Rydb. Hufford 237 (WS), AY483182. B. rubra (Douglas) Rydb. Hufford 298 (WS), AY483196; Hufford 451 (WS), AY483201; Hufford 1691 (WS), AY483183. B. wyomingensis (A. Nelson) Rydb. Hufford 282 (WS), AY483212; Hufford 1909 (WS), AY483204; Hufford 1936 (WS), AY483213; Hufford 2168 (WS), AY483205; Hufford 2194 (WS) (two individuals from population), AY483203, AY483206; Hufford 2195 (WS), AY483207; Hufford 2197 (WS), AY483214. Synthyris borealis Pennell Moran 47 (ALA), AY483177; Parker 1132 (ALA), AY483188. S. canbyi Pennell Gilbert 3 (MONTU), AY483185; Stickney 1740 (MONTU), AY483186. S. cordata (A. Gray) A. Heller Schenk 269D (OSC), AY483197. S. dissecta Rydb. Atwood 14179 (BRY), AY483189. S. laciniata Rydb. Atwood 16148 (BRY), AY483190. S. lanuginosa (Piper) Pennell and J. W. Thomps. McMahon 683 (WS), AY483202. S. missurica (Raf.) Pennell subsp. missurica Hufford 866 (WS), AY483199; Hufford 1209 (WS), AY483178. S. missurica (Raf.) Pennell subsp. stellata (Pennell) Kartesz & Gandhi Hufford 3910 (WS), AY483200. S. missurica (Raf.) Pennell subsp. major (Hook.) Pennell Hufford 3901 (WS), AY483198. S. pinnatifida S. Watson Huber 1078 (BRY), AY483191; Brasher and Bates 1907 (BRY), AY483187. S. platycarpa Gail & Pennell Daubenmire 6351 (WS), AY483192. S. ranunculina Pennell Knight 1955 (UNLV), AY483194. S. reniformis (Douglas) Bentham Hufford 435 (WS), AY483195. S. schizantha Piper Hufford 428 (WS), AY483193. Aragoa abietina Kunth Bello et al. (2002), AJ459404. Chionohebe densifolia (F. Muell.) B. G. Briggs & Ehrend. Wagstaff and Garnock-Jones (1998), AY034849. C. thomsonii (Buchanan) B. G. Briggs & Ehrend. Wagstaff and Garnock-Jones (2000), AF229039. Derwentia perfoliata (R. Br.) Raf. Wagstaff and Garnock-Jones (2000), AY034850. Globularia salicina Lam. Albach and Chase (2001), AF313039. Erinus alpinus L. Albach and Chase (2001), AF313032. Hebe ciliolata (Hook. f.) Cockayne & Allen Wagstaff et al. (2002), AY034851. H. cupressoides (Hook. f.) Cockayne & Allen Wagstaff and Garnock-Jones (1998), AY037378. H. formosa (R. Br.) Cockayne Wagstaff and Garnock-Jones (1998), AF037383. H. lycopodioides (Hook. f.) Cockayne & Allen Wagstaff and Garnock-Jones (1998), AF037383. H. macrantha (Hook. f.) Cockayne & Allen Wagstaff and Wardle (1999), AF069456. Heliohebe hulkeana (F. Muell.) Garn.-Jones Wagstaff and Garnock-Jones (1998), AF037379. H. raoulii (Hook. f.) Garn.-Jones Wagstaff and Garnock-Jones (1998), AF037380. Lagotis angustibracteata Tsoong & Yang Albach and Chase (2001), AF313028. L. brachystachya Maxim. Albach and Chase (2001), AF313027. Paederota lutea L. Albach and Chase (2001), AF313024. Parahebe brevistylis (Garn.-Jones) Heads Wagstaff and Garnock-Jones (2000), AF229045. H. lyallii (Hook. f.) W. R. B. Oliv. Wagstaff and Garnock-Jones (1998), AF037395. H. planopetiolata (G. Simpson & J. S. Thomson) W. R. B. Oliv. Wagstaff and Garnock-Jones (2000), AF229050. H. trifida W. R. B. Oliv. Wagstaff and Garnock-Jones (1998), AF037376. H. vandewateri (Pennell) P. Royen Wagstaff and Garnock-Jones (2000), AF229052. Plantago lanceolata L. Rønsted et al. (2002), AY101898. Pseudolysimachion spicatum (L.) Opiz Albach and Chase (2001), AF313022. P. dahuricum (Steven) T. Yamaz. Albach and Chase (2001), AF313023. Veronica alpina L. Albach and Chase (2001), AF313013. V. calycina R. Br. Wagstaff et al. (2002), AY034863. V. chamaedrys L. Albach and Chase (2001), AF313003. V. fruticulosa L. Albach and Chase (2001), AF313004. V. gentianoides Vahl Albach and Chase (2001), AF313018. V. montana L. Albach and Chase (2001), AF313014. V. peregrina L. Albach and Chase (2001), AF313016. V. persica Poir Albach and Chase (2001), AF313001. V. saturejoides Vis. Albach and Chase (2001), AF313005. Veronicastrum virginicum (L.) Farw. Albach and Chase (2001), AF313030. V. stenostachyum (Helmsl.) T. Yamaz. Albach and Chase (2001), AF313031. Wulfenia blechicii Lakusic Albach and Chase (2001), AF313026. W. carinthiaca Jacq. Albach and Chase (2001), AF313025. from GenBank (Table 1). Sequences were aligned manually in SeAl (Rambaut 1996). Morphological Characters. Twenty morphological characters (Tables 2, 3) were defined based on the examination of herbarium specimens and data from Hufford (1992a, 1992b, 1993). The Hufford (1992a, 1992b, 1993) data were obtained from dissections of fluid preserved specimens, leaf clearings, preparations of microtomed sections, and scanning electron microscopy. Phylogenetic Analyses. Data matrices have been accessioned in TreeBASE (study accession number S1099, matrix accession numbers M1879–1881). All phylogenetic analyses were conducted using PAUP* 4.0 (Swofford 2002). Parsimony analyses used heuristic search procedures that included 1000 starting trees built by random taxon addition followed by TBR branch-swapping. All character state transitions were equally weighted. Indels were treated as missing data. Tree statistics and measures of homoplasy were calculated using PAUP* with uninformative characters removed. Multiple most parsimonious trees were combined in PAUP* to construct strict consensus cladograms. The robustness of clades was assessed using bootstrap analysis (Felsenstein 1985) implemented using heuristic procedures in PAUP*, including random taxon addition and TBR branch-swapping for 1000 pseudoreplicates in which Maxtrees was set at 50,000. A maximum likelihood (ML) analysis was conducted on the ITS data set that included only five outgroup and all 38 ingroup taxa. Lagotis brachystachya and Veronica montana were excluded to reduce the number of outgroups and because they exhibited many autapomorphies in the parsimony analyses. Modeltest (Posada and Crandall 1998) was used to select among 56 alternative substitution models for maximum likelihood analyses. Models selected using likelihood ratio tests (LRTs) and the Akaike information criterion (AIC) were compared. To facilitate the likelihood tree search, we used the values for model parameters estimated during model selection. To evaluate sensitivity to these parameters, we conducted tree searches using only the models selected by the LRTs and the AIC, optimizing each parameter in the models. Additionally, we used the parameter values as estimated by Modeltest, sought an ML tree, re-estimated the parameter values on that tree, and sought again for a new ML tree. The first and second trees were compared for topology and branch lengths to further assess sensitivity to parameter values. Alternative Phylogenetic Hypotheses. Hypotheses of taxonomic groups and taxon relationships can be modelled as cladogram topologies. We designed six constraint topologies based on hypotheses of relationships and morphological similarities (Table 4). The hypotheses modelled include the following: (1) Synthyris and Besseya are both monophyletic and are sister clades, as implied by the classifications of Pennell (1933) and Schaack (1983); (2) taxa that have deeply incised leaf margins, including S. canbyi, S. dissecta, S. lanuginosa, S. pinnatifida, are monophyletic; (3) S. borealis, S. canbyi, S. dissecta, S. laciniata, S. lanuginosa, S. pinnatifida are monophyletic as implied by Schaack’s (1983) circumscription of 2004] 719 HUFFORD & MCMAHON: SYNTHYRIS PHYLOGENY TABLE 2. Morphological characters and character states applied in the phylogenetic analyses of Synthyris and Besseya. Data and descriptions of character states are presented primarily in Hufford (1992a, 1992b, 1993), although selected data were derived from the examination of herbarium specimens and Pennell (1933), Went (1958), and Schaack (1983). 1. Lamina/petiole juncture: 0, obtuse angle; 1, same plane. This character refers to the angle formed by the lamina and petiole on the lower side of the leaf. 2. Leaf lamina shape: 0, ovate; 1, reniform. The definitions of lamina shapes follow Lawrence (1951). 3. Leaf lamina margin: 0, toothed (not incised to midrib); 1, incised to within 2 mm of midrib. 4. Leaf lamina teeth: 0, compound; 1, simple. Simple teeth have only a single lobe and adjacent teeth have lobes of a similar size and shape; whereas, compound teeth have primary lobe as well as secondary and sometimes tertiary lobes that are smaller and sometimes a different shape from the primary lobe. 5. Adaxial surface of leaf midvein: 0, level adaxial surface of leaf; 1, sunken below surface of leaf. In leaf cross sections, the midvein can be sunken below the plane of the adaxial surface of the lamina or largely flush with the surface (as illustrated in Hufford 1992b) 6. Size of lamina midvein: 0, less than/equal to 1.5%; 1, 2–3%; 2, greater than/equal to 4%. The size of the lamina midvein was measured as a percent of total lamina width at the middle of the lamina, following the definition of Hickey (1979). 7. Diameter of secondaries to midvein: 0, greater than/equal to 80%; 1, 25–80%; 2, less than/equal to 25. The diameters of the lowest secondary vein and the midvein were measured just above the point where they diverge. 8. Secondaries from midvein: 0, four or fewer; 1, six or more. 9. Extent of basal secondaries: 0, distal half of lamina; 1, proximal half of lamina. The basal secondary veins are those that diverged from the midvein in the petiole, and the character refers to whether they extend beyond the middle of the lamina or end in proximal half of the lamina. 10. Inflorescence attitude: 0, erect; 1, reclinate. Inflorescences of S. cordata and S. reniformis are lax and with the onset of fruiting lie on the surface of the substrate, which contrasts with the inflorescences of other Synthyris and Besseya that remain stiffly erect from initiation through fruiting. 11. Inflorescence bracts: 0, three or more; 1, two; 2, none. Inflorescences bear either three or more helically arranged bracts, two opposite or subopposite bracts, or no bracts. 12. Peduncle pubescence: 0, pilose-villous; 1, sparse-glabrous 13. Calyx/corolla length: 0, corolla well exserted; 1, corolla slightly exserted; 2, corolla not exserted. At anthesis, flowers can have a corolla in which the lobes are well exserted beyond the calyx, a corolla that is approximately the same length as the calyx and only slightly exserted, or, as in B. rubra and B. wyomingensis, a highly reduced corolla not exserted beyond the calyx. 14. Corolla throat: 0, open; 1, closed. At anthesis, the lobes of the corolla can either spread, creating an opening to the corolla throat, or they can remain largely erect and loosely imbricate or even tightly furled around the stamen filaments and style (5 a closed throat). 15. Corolla lobe margins: 0, entire; 1, laciniate 16. Stamen insertion: 0, corolla tube; 1, receptacle. Stamens can be inserted either on corolla tube or on receptacle as shown in Hufford (1992a). 17. Anther color: 0, reddish purple-blue; 1, yellowish white 18. Fruit base shape: 0, rounded; 1, flat, perpendicular to pedicel; 2, flat, acutely angled to pedicel. Fruits vary in shape, especially in the proximal portion of the fruit, and this influences also the angle formed between the fruit base and the pedicel. Although character states were scored on basis of a survey of herbarium specimens, illustrations of fruit shapes are provided by Schaack (1983). 19. Fruit pubescence: 0, glabrous; 1, sparse or pilose only at margins; 2, villous. The definitions of pubescence states follow Lawrence (1951). 20. Seed number: 0, 6 or fewer; 1, 10–16; 2, 20–40 this group as Synthyris section Dissecta; (4) taxa that have reniform leaf laminas that have broadly rounded apices, including S. laciniata, S. missurica, S. platycarpa, S. ranunculina, and S. schizantha, are monophyletic; (5) S. missurica subsp. major is the sister of all other Synthyris and Besseya, which models Schaack’s (1983) proposal that the most primitive extant species of Synthyris is S. major and that all other Synthyris and Besseya are derived from an ancestor most similar to it; and (6) B. wyomingensis is monophyletic. Modelling each of these hypotheses as a cladogram (each with only a few nodes specified), we conducted six constrained ML searches, finding the optimal tree for our data for each alternative. For the constrained analyses, we used SYM1G, which was the less parameter-rich model of the two selected by Modeltest, because the unconstrained analyses demonstrated no differences between the results using either model. Likewise, we used the parameter values as originally estimated during model selection because we found no sensitivity to the model parameters in the unconstrained analyses. We also sought the most parsimonious trees under each constraint and compared the results with the likelihood tree searches, primarily to evaluate whether we could use difference in tree length (instead of difference in likelihood, a more consistent but less efficiently obtained measure) as a test statistic in hypothesis testing. To evaluate the significance in the differences between trees obtained using constraints and the maximum likelihood tree, we conducted parametric bootstrapping (Huelsenbeck and Crandall 1997), also called the SOWH test (Goldman et al. 2000). This test is appropriate for comparing alternatives when one is selected a posteriori (Goldman et al. 2000), as is the ML tree. The null hypothesis is that the alternative tree (i.e., the most likely tree under the constraint) is correct; if so, then we would expect to see the difference in tree scores between the alternative tree and the ML tree frequently. To test this, we simulated data on the alternative tree, sought the best tree under the constraint and the best tree without the constraint, and compared their tree scores. In this way, we apply the same treatment to each simulated data as we did to our observed data. For each alternative tree, we did 500 such simulations. We used Mesquite (v. 0.994, Maddison and Maddison 2003) to simulate the data sets on the alternative trees. To construct the model for data simulation, ten values were estimated from the observed data, corresponding to the GTR1I1G model with unequal base frequencies. 720 [Volume 29 SYSTEMATIC BOTANY TABLE 3. Matrix of morphological character states applied in phylogenetic analyses. Multiple accessions of species sampled for ITS sequence data had the same morphological states, except for Besseya wyomingensis. The character state coding for each sampled population of B. wyomingensis is provided.