HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via ISI Web of Science (4) Citing Articles via Google Scholar Google Scholar Articles by Wolfe, A. D. Articles by Diaz, J. Search for Related Content PubMed Articles by Wolfe, A. D. Articles by Diaz, J. Agricola Articles by Wolfe, A. D. Articles by Diaz, J.

Phylogeny, taxonomic affinities, and biogeography of Penstemon (Plantaginaceae) based on ITS and cpDNA sequence data¹

²Department of Evolution, Ecology, and Organismal Biology, The Ohio State University, 318 West 12th Avenue, Columbus, Ohio 43210 USA; ³Department of Ecology and Evolutionary Biology, University of Kansas, 1200 Sunnyside Ave, Lawrence, Kansas 66045 USA; and ⁴Department of Biological Sciences, California State University Sacramento, 6000 J Street, Sacramento, California 95819 USA

Received for publication December 21, 2005. Accepted for publication August 24, 2006.

ABSTRACT

The large and diverse genus Penstemon (ca. 271 species) is endemic to North America and has been divided into six subgenera primarily based on anther dehiscence patterns. Species of Penstemon are known to be pollinated by a variety of insects (hymenopterans, lepidopterans, dipterans) and hummingbirds. Nucleotide sequence data from ITS and two noncoding regions of chloroplast DNA were used to reconstruct the phylogeny of Penstemon. Trees generated from nuclear and chloroplast DNA sequences are incongruent, which is probably the result of hybridization, and not fully resolved, which is likely due to a rapid evolutionary radiation. Penstemon represents a recent continental radiation where speciation has resulted primarily from evolutionary adaptations to ecological niches such as pollinator specialization. The results from these analyses show that the current circumscription of subgenera and sections needs revision to reflect more closely the evolutionary relationships of species. Specifically, species in subgenera Saccanthera, Habroanthus, and Penstemon are polyphyletic. These results also confirm the independent origin of hummingbird floral morphology in 10 clades.

Key Words: biogeography • Cheloneae • hummingbird floral morphology • ITS • Scrophulariaceae • trnC-D, trnT-L

Penstemon Mitchell, with ca. 271 species (Lodewick and Lodewick, 1999 ), is the largest plant genus endemic to North America. It is distributed from Alaska to Guatemala and from coast to coast. Most of the species occur in the western cordillera with the Intermountain Region representing the center of diversity. Most species have narrow distributions, and more than 100 are endemic to a single state. Straw (1966) proposed that Penstemon originated in the central Rocky Mountains and adjacent Columbia Plateau in the late Tertiary. Wolfe et al. (2002) supported this hypothesis in a molecular phylogenetics study of Cheloneae, but the sampling within Penstemon was very limited. Wolfe et al. (2002) also proposed a post-Tertiary radiation of the genus given its remarkable floral and vegetative diversity. Much of the diversity within Penstemon is associated with evolutionary adaptations such as specialization for pollinators or ecological niches. Thus, Penstemon represents a rare example of a continental evolutionary radiation.

The genus has been divided into six subgenera (Lodewick and Lodewick, 1999 ), two of which are monotypic (subg. Cryptostemon and subg. Dissecti). The other four subgenera are Penstemon (ca. 182 spp.), Habroanthus (ca. 50 spp.), Saccanthera (ca. 28 spp.), and Dasanthera (ca. 9 spp.). There is considerable morphological diversity within and among subgenera involving habit, floral, leaf, and stem characters. Flowers of most Penstemon species attract a wide variety of insect pollinators including bees and wasps, moths, or bee flies (Pennell, 1935 ; Straw, 1955 , 1956a , b, 1963a ; Crosswhite and Crosswhite, 1966 ; Tepedino et al., 1999 ; Thomson et al., 2000 ), whereas ca. 40 species have floral syndromes typical of hummingbird pollination (Wilson et al., 2006 ).

In addition to floral divergence, the architecture of vegetative characters varies dramatically among species (Holmgren, 1984 ). Stems are woody, suffrutescent, or herbaceous. They are either simple or branched and arise from a woody caudex, from a thick crown subtended by a taproot or from a basal rosette subtended by a taproot. Some species form dense mats, while others have one or few stems above a basal rosette; some are long-lived shrubs, while others are short-lived perennials. Many of these morphological features characterize taxonomic groupings within Penstemon (e.g., most of the taxa within subg. Dasanthera are suffrutescent, subg. Saccanthera has species with a woody caudex, sect. Ericopsis subsect. Caespitosi consists of diminutive mat formers).

Variable characters in leaf morphology; inflorescence architecture; stem, leaf, and floral vestiture; and numerous floral structures also distinguish species of Penstemon. However, many species are differentiated from their closest relatives by only a few characters (Keck, 1932 , 1936a , b , 1937a , b , 1938 , 1940 , 1945 , 1957 ; Pennell, 1935 ; Straw, 1959 , 1962b ; Crosswhite, 1965b , c , 1967 a, c ; Holmgren, 1984 ), and hybridization has obscurred taxonomic differences (Straw, 1955 , 1956a , b ; Viehmyer, 1958 ; Every, 1977 ; Wolfe and Elisens, 1995 ; Wolfe et al., 1998a , b ; Datwyler and Wolfe, 2004 ).

Penstemon is a member of Cheloneae, a tribe characterized by a cymose inflorescence and flowers having a staminode (Wettstein, 1891 ; Straw, 1966 ; Thieret, 1967 ; Wolfe et al., 2002 ). Pennellianthus, Nothochelone, and Keckiella were segregated from Penstemon by Straw (1966) . The primary synapomorphies for Straw's (1966) circumscription of Penstemon are (1) the presence of epistaminal nectaries as opposed to a nectary consisting of a hypogynous disc, (2) glabrous anther-bearing filaments, (3) presence of a staminode, and (4) seeds without wings. Chionophila also has epistaminal nectaries, and Straw (1966) inferred a close relationship of the two genera based on this character

Penstemon has a range of other dehiscence patterns and morphologies that characterize subgenera. For example, subg. Saccanthera is named for the saccate anther morphology possessed by its species. Saccate anthers dehisce across the connective, but not to the distal ends. However, three species not within subg. Saccanthera (P. baccharifolius, P. multiflorus, and P. dissectus) also have saccate anther morphology. Subgenus Habroanthus, as traditionally circumscribed (Crosswhite, 1967a –c ), has species with anthers dehiscing from the distal end of the anther sac but not across the connective. All other species of Penstemon have anthers that dehisce across the connective and to the distal ends, including species of subg. Dasanthera, which have anther sacs densely covered in wooly hairs. Anther dehiscence patterns play a role in pollination (Thomson et al., 2000 ; Castellanos et al., 2003 , 2006 ), with bee-pollinated species restricting the pollen presentation to small quantities over a long period, and bird-pollinated species dispensing pollen in larger quantities over a short period.

The staminode of Penstemon may be bearded or glabrous, exserted or included, and also plays a role in pollination (Walker-Larsen and Harder, 2001 ; Dieringer and Cabrera R., 2002 ). Hummingbird-pollinated species of Penstemon tend to have glabrous staminodes that are reduced in size and do not function in pollination (Walker-Larsen and Harder, 2001 ). In contrast, the staminode of bee-pollinated species varies in size, structure, and pubescence, and contributes to pollination by enhancing contact of floral sexual organs by insects (Walker-Larsen and Harder, 2001 ; Dieringer and Cabrera R., 2002 ).

The most distinctive group of species within Penstemon is subg. Dasanthera. In addition to the wooly anthers characteristic of the group, most species are woody or suffrutescent (except P. lyallii), and the flowers have a distinct pleating along the base of the throat and a petal fusion ridge along the top of the flower. The species in the Sierra Nevada and Cascades have evergreen foliage, whereas the species in the Rocky Mountains and Intermountain region have deciduous leaves.

Other taxonomic groups with morphological features that are strikingly distinctive include sect. Elmigera of subg. Habroanthus and sect. Fasciculus and sect. Ericopsis subsect. Caespitosi of subg. Penstemon. Species within sect. Elmigera have flowers that are typical of hummingbird pollination (e.g., red, narrow corollas; exserted anthers; reflexed petals). Taxa within Fasciculus have fascicles at each stem node, and species of subsection Caespitosi are diminutive wiry mat formers with small leaves and pleated flowers.

The base chromosome number of Penstemon is x = 8 (Keck, 1945 ; Straw, 1966 ). Polyploidy occurs in several sections. The nature of polyploid species has not been extensively studied in Penstemon. However, hypotheses of hybrid origin for polyploid species have been made for species in subg. Saccanthera and Penstemon sect. Penstemon (Keck, 1932 , 1945 ). Hybridization and hybrid speciation at the diploid level have been hypothesized (Straw, 1955 , 1956b ) and are supported by molecular data (Wolfe et al., 1998a , b ).

The objectives of our study were to use nucleotide sequence data to infer the phylogeny of Penstemon and to use the resulting phylogeny to examine (1) taxonomic definitions, (2) patterns of morphological and chromosome evolution within Penstemon, (3) transitions from insect to bird pollination, and (4) prior biogeographic hypotheses.

MATERIALS AND METHODS

Sampling
Twelve genera comprising 196 taxa were included in our survey (Appendix 1), including all members of Cheloneae (Chelone, Chionophila, Collinsia, Keckiella, Nothochelone, Pennellianthus, Penstemon; Wolfe et al., 2002 ) and representatives from all taxonomic sections of Penstemon. The total number of Penstemon species surveyed was 163 (Appendix 1, Table 1), 11 of which included two or more varieties for a total of 178 taxa. Uroskinnera (Central America), which has traditionally been placed in Cheloneae, and Tetranema and Russelia, shown to be sister taxa to Cheloneae (Wolfe et al., 2002 ; Albach et al., 2005 ; Oxelman et al., 2005 ), were also included in the survey. Antirrhinum and Linaria were used as outgroup taxa.

View this table: [in this window] [in a new window]   Table 1. Taxonomy of Penstemon as currently recognized by the American Penstemon Society or treated in the most recent field guides (Lodewick and Lodewick, 1999 ; Nold, 1999 ). Sampling of species for this study is indicated in the numerator, and abbreviations are those used in Figs. 1–4

Di-deoxy termination sequencing was performed using manual and automated techniques. Manual sequencing was performed using the dITP terminators of the USB (Cleveland, Ohio) DNA sequencing kit (Wolfe et al., 2002 ; Datwyler and Wolfe, 2004 ). Automated cycle sequencing reactions were performed using the Big Dye terminator chemistry (ABI, Foster City, California, USA), and reactions were run on either an ABI 310 or ABI 3100 automated DNA sequencer. Double-stranded sequences were generated for ITS, trnT-L, and trnC-D using internal primers appropriate for each locus (Taberlet et al., 1991 ; Demesure et al., 1995 ; Wolfe and Randle, 2001 ).

Data analyses
Sequences were assembled using Sequencher (GeneCodes, Ann Arbor, Michigan, USA). Sequences were aligned in ClustalX (Thompson et al., 1997 ) with manual adjustments as necessary. Analyses were performed separately for ITS, trnC-D, and trnT-L, in addition to a combined data set of the plastid loci. Gaps within the ITS matrix and the combined plastid sequence matrix were treated as missing data. Eleven and 22 indel characters were concatenated onto the end of the ITS and plastid sequence matrices, respectively. Of these indels, 12 (three for ITS and nine for trnC-D/T-L) were coded as complex gap characters (Simmons and Ochoterena, 2000 ), and treated as ordered characters with step matrices constructed in MacClade 3.08a (Maddison and Maddison, 1999 ).

Phylogenetic analyses were conducted using PAUP* version 4.0b10 (Swofford, 2002 ). All analyses were performed with uninformative characters excluded. Trees were generated using the heuristic search option with 10 random addition sequences (MaxTrees = 2000, Mulpars, tree-bisection-reconnection (TBR) branch swapping, 10 shortest trees held at each step). Our initial analyses were run with MaxTrees set to "automatically increase by 100." This resulted in 11 500 trees, the limit supported by available computer memory. Strict consensus trees for the entire set of 11 500 trees and subsets of 2000 trees were compared to examine topological congruence among the saved trees. The differences among strict consensus trees were only in the extreme tips so subsequent analyses were done with MaxTrees set to 2000 to expedite the analyses. Support for each node was assessed using 2000 "fast" bootstrap replicates.

Topological incongruence was examined by visually comparing trees from each data set. All nodes resolved in the strict consensus trees were considered in this analysis, and, particularly, nodes with bootstrap values of at least 70% (Kellogg et al., 1996 ; Mason-Gamer and Kellogg, 1996 ; Archibald et al., 2005 ). In addition to the visual analysis, the partition homogeneity test (incongruence length difference test of Farris et al., 1994 ) in PAUP* was used (100 replicates, TBR branch swapping, 100 replicates of random taxon addition, TBR branch swapping, and MaxTrees = 500).

The ITS and cpDNA consensus trees were used to assess morphological and chromosome evolution in Penstemon as well as geographic distribution of taxa. The following traits were mapped onto the consensus trees: Anther morphology, chromosome number(s), and occurrence of bird pollination. In addition, the taxonomic category (subgenus, section, subsection) and geographic distributions were mapped onto the trees.

RESULTS

Analysis of the ITS data set
The aligned sequences and 11 gap characters yielded a matrix with 695 characters, 202 of which were potentially informative, with 0.2% of the matrix cells coded as missing data. The ITS1, 5.8S, and ITS2 sequence lengths ranged from 560 to 618 bp for Collinsia heterophylla and Penstemon californicus, respectively, with most sequences ranging from 575–605 bp.

From the maximum parsimony analysis, 2000 equally parsimonious trees were saved (L = 877, CI = 0.414, RI = 0.726). The strict consensus tree (Figs. 1, 2) was rooted using Antirrhinum and Linaria as outgroup taxa. Clades for genera within Cheloneae generally had high bootstrap values, but only a few terminal lineages of sister taxa had bootstrap values above 70% within the Penstemon clade (Figs. 1, 2).

View larger version (39K): [in this window] [in a new window]   Fig. 1. Strict consensus of 2000 equally parsimonious ITS trees, part I. Outlined box in the tree is shown in Fig. 2. The line thickness shown in the tree reflects bootstrap (BP) values of 70–100%, with BP >90% shown with thick black lines and values 70–80% shown with thick gray lines. Taxon abbreviations are in Table 1, and distribution designations are in Table 2. P. = Penstemon. Anther morphology: 1 = dehisces across the connective and from end-to-end of the anther sac; 2 = dehisces across the connective but not to the distal end of the anther sac; and 3 = dehisces from the distal end toward the connective, but not across the connective. Bird pollination = hummingbird-pollinated species. Chromosome numbers were obtained from references cited throughout the manuscript

View larger version (38K): [in this window] [in a new window]   Fig. 2. Strict consensus of 2000 equally parsimonious ITS trees, part II. Abbreviations and designations are the same as in Fig. 1

Taxonomic groupings (subgenera, sections, and subsections) within Penstemon corresponded sometimes, but certainly not always, to clades (Figs. 1, 2). For example, all of the species in subg. Dasanthera group together with P. personatus from subg. Cryptostemon, all species in subg. Saccanthera are found in the basal clades of Penstemon, and most species in sections Ericopsis, Aurator, and Fasciculus form clades.

Analysis of the chloroplast data set
The chloroplast data set consisted of 2212 characters, including 22 indel characters coded as simple or complex gaps, with 289 positions being potentially informative and 1.0% of the cells, mostly at the beginning and ends of the aligned sequences, coded as missing data. Nucleotide sequences from two noncoding regions, trnC-D and trnT-L, were combined for this anlaysis. The trnC-D region yielded an aligned sequence matrix of 663 characters plus 10 gap characters, whereas the trnT-L region yielded an aligned sequence matrix of 1527 characters plus 12 gap characters. The number of potentially informative characters was 81 and 186 for trnC-D and trnT-L, respectively. For either data set, 12.2% of the characters were parsimony-informative. The trnC-D sequence lengths ranged 259 to 556 bp for Tetranema mexicanum and Penstemon speciosus, respectively, with most sequence lengths ranging 500–530 bp. The trnT-L sequence lengths ranged from 669 to 1268 bp for Penstemon aridus and Tetranema mexicanum, respectively, with most sequence lengths in the range of 1150–1200 bp.

From the maximum parsimony analysis, 2000 equally parsimonious trees were saved (L = 763, CI = 0.510, RI = 0.803). The strict consensus tree (Figs. 3, 4) was rooted using Antirrhinum and Linaria as outgroup taxa. Similar to the results from the ITS study, clades for genera within Cheloneae generally had high bootstrap values, but only a few terminal lineages of sister taxa had bootstrap values above 70 within the Penstemon clade (Figs. 3, 4).

View larger version (33K): [in this window] [in a new window]   Fig. 3. Strict consensus of 2000 equally parsimonious combined trnC-D/T-L trees, part I. Outlined box in the tree is shown in Fig. 4. Abbreviations and designations are the same as in Fig. 1

View larger version (32K): [in this window] [in a new window]   Fig. 4. Strict consensus of 2000 equally parsimonious trnC-D/T-L trees, part II. Abbreviations and designations are the same as in Fig. 1

Similar to the ITS tree (Figs. 1, 2), subgenera, sections, and subsections within Penstemon only sometimes seemed to represent clades (Figs. 3, 4). Most of the species in subg. Dasanthera group except for P. montanus, which appears as sister to the rest of Penstemon, all species in subg. Saccanthera are found in the basal clades of Penstemon, and most species in sect. Aurator group together in the same clade.

Congruence testing and comparison of ITS and cpDNA trees
Major topological differences observed visually between the ITS and cpDNA trees (Figs. 1–4) include the placement of Chionophila tweedyi, P. campanulatus, P. clevelandii, P. dasyphyllus, P. dissectus, P. dolius, P. gentryi, P. hartwegii, P. heterophyllus, P. incertus, P. isophyllous, P. kunthii, P. lanceolatus, P. leonensis, P. linarioides, P. montanus, P. neotericus, P. personatus, P. pinifolius, P. strictus, P. teucroides, and P. virgatus var. asa-grayii. Additional topological differences include members of clades containing P. caespitosus and P. gentianoides.

The partition homogeneity test (ILD) revealed significant incongruence (P = 0.01) between the ITS and cpDNA trees. Successive deletions of taxa and a rerun of the analysis did not improve the scores. The incongruence could be caused by a history of introgression and chloroplast capture events (Wolfe and Elisens, 1993 , 1994 , 1995 ; Wolfe et al., 1998a , b ; Datwyler and Wolfe, 2004 ). Thus, we did not combine these data sets.

Despite conflicts in the nuclear and cpDNA data sets, there are elements of MP trees inferred from each data set that are congruent. For example, species in subg. Dasanthera group together in a clade that is sister to the rest of Penstemon (Figs. 1, 3, 5), and the species in subg. Saccanthera are found at the base of the Penstemon clade associated with certain species of subg. Penstemon sect. Penstemon. Species in sections Aurator, Ericopsis, and Fasciculus show similar topological patterns in each analysis in terms of terminal clades and species groupings, which reflects their close relationships within their respective sections. Also, species of subg. Habroanthus are mixed in a group with several sections of subg. Penstemon toward the tips of the Penstemon clade (Figs. 2, 4).

View larger version (10K): [in this window] [in a new window]   Fig. 5. Phylograms of one of 2000 equally parsimonious trees for ITS (on left) and cpDNA (on right) analyses. The arrows point to the node defining the Cheloneae clade in each analysis. The Dasanthera clade of Penstemon is marked with an asterisk

DISCUSSION

Cheloneae and Russelieae
The clade definition of Cheloneae sensu Wolfe et al. (2002) is "the least inclusive clade that contains Chelone, Collinsia, and Pennellianthus." Synapomorphies for the tribe include a cymose inflorescence, the presence of a staminode, simple hairs, and stems with pith. The previous molecular study by Wolfe et al. (2002) did not include Uroskinnera, and it was proposed that this genus was key to elucidating relationships among Tetranema, Russelia, and Cheloneae. We were able to amplify and sequence ITS from a herbarium specimen of Uroskinnera, but were unsuccessful in our attempts to do the same for the noncoding chloroplast regions used in this study. In the ITS phylogeny (Fig. 1), Uroskinnera was sister to Cheloneae. The node supporting Cheloneae had high bootstrap support, but there was no support for including Uroskinnera as a member of Cheloneae. Burtt (1965) proposed that Uroskinnera (distributed in Mexico and Guatemala) and Brookea (endemic to Borneo) were closest allies. We were unable to sample Brookea for this study, but it will be interesting to elucidate the relationships among these genera and Russelieae and Cheloneae when material becomes available for sequencing.

Wolfe et al. (2002) used ITS and matK sequences to infer the phylogenetic patterns of Cheloneae and related genera within Plantaginaceae and found strong support for a clade consisting of Russelia and Tetranema as sister taxa. Two studies since then have used additional plastid genes and independent tissue collections to verify this relationship and the position of this clade as sister, or near to, Cheloneae (Albach et al., 2005 ; Oxelman et al., 2005 ). Our current study also reveals a strongly supported clade of Russelia and Tetranema. Taken together, these results support a redefinition of tribe Russelieae (Pennell, 1920 b) to include Tetranema.

Chionophila
Straw (1966) hypothesized that Chionophila and Penstemon were sister taxa based on a shared epistaminal nectary character. The two species of Chionophila are morphologically distinct with non-overlapping ranges. Chionophila jamesii occurs in the Rocky Mountains of Colorado and Medicine Bow Mountains of Wyoming. It is found at the edges of melting snowfields in the alpine zone. This diminutive plant has an inflorescence of several tightly packed, cream-colored flowers. Chionophila tweedyi occurs in subalpine open woodlands in Idaho and Montana, and has a loosely organized raceme of four to 10 lavender flowers.

The placement of Chionophila tweedyi differs in the ITS and cpDNA trees (Figs. 1, 3). It occurs within a polytomy within the Penstemon clade in the ITS strict consensus tree, but as sister to Chionophila jamesii in the cpDNA strict consensus tree (bootstrap = 70) with this clade grouped with Chelone, Keckiella, and Nothochelone. An analysis of combined data from ITS and matK (Datwyler and Wolfe, 2004 ) using a subset of the taxa found in this study revealed a strongly supported clade of the two species of Chionophila (bootstrap = 97, Bremer support = 5). Similar results were reported in Wolfe et al. (2002) using a larger subset of the taxa sampled here. These results are consistent with Chionophila being more closely related to other members of Cheloneae rather than as a sister taxon to Pentemon. However, further studies are needed to investigate the anomalous pattern in the ITS tree reported here (Fig. 1).

Taxonomic implications for subgenera of Penstemon
This is the first phylogenetic study of Penstemon with a thorough sampling of subgenera, sections, and subsections. We were able to sample all but one monotypic subsection for this survey for a total of 163 of 271 species (Table 1). Penstemon has not been examined in its entirety, but has been the subject of investigation in regional floras (e.g., Pennell, 1920b , 1935 ; Holmgren, 1984 ) or treatments of particular groups of species (Keck, 1932 , 1936a , b , 1937a , b , 1938 , 1940 , 1945 ; Keck and Cronquist, 1957 ; Crosswhite, 1965a - c , 1966 , 1967a – c , 1970 ; Straw, 1959 , 1962 , 1963b ). The taxonomy of Penstemon has been summarized in publications sponsored by the American Penstemon Society (Lodewick and Lodewick, 1999 ) and in field and garden guide books (Way and James, 1998 ; Nold, 1999 ). However, the genus has not yet been treated as a whole in the scientific literature. Because the Lodewick and Lodewick (1999) publication represents a summary of all the taxonomic literature available for Penstemon, we followed the nomenclature and classification scheme presented there.

Our study falls short of providing the phylogenetic basis of a revised taxonomy for Penstemon, but does indicate that some infrageneric taxa may be natural groupings while others are hopelessly paraphyletic. Further, it highlights a number of taxonomic anomalies that may be fixed by moving a species from one taxonomic group to another.

Subgenera Dasanthera and Cryptostemon
Of the six subgenera of Penstemon, only subg. Dasanthera appears distinctive in the ITS and cpDNA phylogenies (Figs. 1–5). However, P. personatus of the monotypic subg. Cryptostemon groups with the Dasanthera species in the ITS tree (Fig. 1), and P. montanus falls outside the clade of Dasanthera penstemon in the cpDNA tree (Fig. 3). There are many molecular synapomorphies for the Dasanthera clade (Fig. 5), and the morphology of these species is very distinctive compared to the rest of the genus (Datwyler and Wolfe, 2004 ).

Keck (1936b) discussed the taxonomic affinities of P. personatus in terms of its floral morphology. The most striking floral feature of P. personatus is the short staminode. Keck (1936b) pointed out that several of the Dasanthera species also have short staminodes and that the staminode of P. rupicola is shorter than that of P. personatus. Morphological synapomorphies for P. personatus and species of subg. Dasanthera include a short staminode and dense pubescence within the corolla throat. Our results combined with these morphological characters support adding P. personatus to subg. Dasanthera, and eliminating subg. Cryptostemon.

Subgenera Saccanthera, Habroanthus, and Dissecti
The other three subgenera of Penstemon identified by distinctive anther dehiscence patterns are Saccanthera, Habroanthus, and Dissecti. Species in subg. Saccanthera and Dissecti have anther sacs that dehisce across the connective, but not to the distal ends, and species in subg. Habroanthus have anther sacs that dehisce from the distal end toward the connective, but not across the connective. Anther dehiscence patterns and taxonomic affiliations were scored next to the strict consensus trees from each analysis (Figs. 1–4).

Keck (1932) classified all but three species of Penstemon with saccate anther morphology as members of sect. Saccanthera. Several hypotheses regarding taxonomic affinities were proposed by Keck (1932) , incuding (1) P. gracilentus is most closely allied to species in subg. Penstemon sect. Penstemon; (2) P. azureus is a derivative of P. laetus; (3) P. neotericus is a hybrid between P. laetus and P. azureus; (4) P. serrulatus is closely allied to P. ovatus of sect. Penstemon; (5) P. heterophyllus is a derivative of P. laetus, P. azureus, or a hybrid derivative of both; (6) P. triphyllus is related most closely to P. richardsonii; (7) P. glandulosus is not closely allied to other species of the section; and (8) P. rostriflorus appears to be more closely related to members of sect. Elmigera than to other species of Saccanthera. Our results can address most of these hypotheses, beginning with the affinities of subg. Saccanthera to other groups of Penstemon.

Most of the Saccanthera species are found in a single clade in the ITS tree (Fig. 1), with the exception of P. gracilentus, which is at the base of the larger clade containing the Saccanthera clade. This pattern differs in the cpDNA tree (Figs. 3, 4) in that the Saccanthera species do not form a single clade, and the placement of P. rostriflorus is very different in each phylogeny. Our results support Keck's (1932) hypothesis 1, that P. gracilentus is closely allied with species of sect. Penstemon.

Close relationships among most of the Saccanthera species (Keck's [1932 ] hypotheses 2, 3, 5, and 6) are supported by our results, whereas P. serrulatus does not appear to be closely related to P. ovatus (Keck's [1932 ] hypothesis 4; Figs. 1, 3). Our results also do not support Keck's (1932) hypotheses 8 about P. rostriflorus as more closely related to species in sect. Elmigera than to species in Saccanthera, nor is P. glandulosus more closely related to P. ovatus of sect. Penstemon (Keck's [1932 ] hypothesis 7; Fig. 1).

Crosswhite (1967a) designated Habroanthus as a subg. having a synapomorphy of anther dehiscence from the distal end, but not across the connective of the anther sacs. Subg. Habroanthus has two sections, Habroanthus and Elmigera. The last section is comprised of species with red, tubular corollas. Neither sect. Habroanthus nor Elmigera were monophyletic in either of our analyses (Figs. 2, 4). Rather, species in both sections are scattered among species from subg. Penstemon sections Anularius and Peltanthera. The anther dehiscence pattern for subg. Penstemon is complete dehiscence, but it is not known whether dehiscence is initiated from the distal end or connective.

The anther dehiscence patterns within Penstemon would appear to be insufficient synapomorphies for defining subgenera, although this character has been useful for field identification and grouping species in identification keys. Outside of Saccanthera, three species of Penstemon have saccate anthers: P. multiflorus and P. baccharifolius from subg. Penstemon, and P. dissectus from the monotypic subg. Dissecti (Figs. 1–4). None of these three species group with the Saccanthera members. Penstemon dissectus is an unusual species, narrowly endemic to granite outcrops in Georgia, and the only species with dissected leaves. Pennell (1935) was unable to define any close relationships of this species with others in eastern North America; and its placement in the ITS and cpDNA tree differs greatly (Figs. 2, 4). However, P. dissectus is clearly not of a basal lineage and appears to be sufficiently derived from within subg. Penstemon so that its designation as a monotypic subgenus should be questioned.

The taxonomy of Penstemon would be improved by revising sections to reflect evolutionary relationships more faithfully. The results from our molecular phylogenetics study indicate the need for a major revision of Penstemon to include two expanded subgenera, Dasanthera and Penstemon, and many taxonomic changes to accommodate new species affiliations. For example, given the lack of monophyly for species in subg. Saccanthera together with the convergent evolution of saccate anther morphology in the genus (Figs. 1–4), a revised taxonomy including a reduced and monophyletic subset of Saccanthera as a section within subg. Penstemon is warranted. Similarly, the taxonomic status of subg. Habroanthus should be re-evaluated.

Taxonomic implications for sections of Penstemon
With the lack of resolution in the ITS and cpDNA phylogenies, our comments will focus on sections Penstemon, Ericopsis, Aurator, and Fasciculus to supplement what we have already discussed regarding the reassignment of subg. Habroanthus species within sections Anularius and Peltanthera. Additional comments on individual species of Penstemon and their phylogenetic affinities are in Castellanos et al. (2006) .

Sect. Penstemon as currently circumscribed has nine subsections (Table 1). Subsections Proceri and Humiles represent early-branching lineages in both cpDNA and ITS analyses (Figs. 1, 3), with affinities to members of Saccanthera. Species in subsections Deusti and Gairdneriani are in even closer proximity to Saccanthera species in both analyses. However, species in subsections Penstemon, Tubaeflori, Multiflori, Harbouriani, and Arenarii are placed in more terminal clades (Figs. 2, 4) compared to the other four subsections, with the first three grouping together. Penstemon harbourii is closely allied to species in sect. Ericopsis subsection Caespitosi in both analyses (Figs. 1, 4), whereas P. albomarginatus of subsect. Arenarii groups with species from sect. Ambiguii (Fig. 2).

Most of the species of sect. Ericopsis subsections Caespitosi and Linarioides form a clade in both analyses (Figs. 1–4); this clade, however, differs in phylogenetic position between trees. In the ITS tree (Fig. 1), this group of species is sister to the Saccanthera clade, but is in a larger terminal clade containing species from sect. Penstemon subsect. Penstemon, and sections Aurator and Peltanthera in the cpDNA tree (Fig. 4). Penstemon laricifolius (sect. Ericopsis subsect. Ericopsis) does not group with others in its section in either analysis, but appears more closely related to species in sect. Aurator (Figs. 2, 4). Penstemon acaulis also appears as more closely related to species in sect. Aurator than to species in sect. Ericopsis (Fig. 1). Most of the Aurator species occur in a clade with species of sect. Ericopsis (P. acaulis and P. laricifolius in the ITS tree and P. laricifolius in the cpDNA tree, which did not include P. acaulis) in either analysis (Figs. 2, 4). Keck (1937a) and Penland (1958) proposed affinities of sections Ericopsis and Aurator, primarily through P. dolius, which is in clades of Ericopsis species in each analysis (Figs. 1, 4). The transfer of this species to sect. Ericopsis seems appropriate, and it would be practical to reexamine the relationships of P. acaulis and P. laricifolius to species of Ericopsis and Aurator.

Penstemon pinifolius is another species that should be examined more closely. It is currently placed in sect. Fasciculus subsect. Fasciculi based on its possession of axillary fascicles (Straw, 1962 ). However, Straw (1962) hypothesized that P. pinifolius was anomalous in this section because its floral characteristics differed greatly from other species in sect. Fasciculus. Crosswhite and Crosswhite (1981) hypothesized that P. pinifolius has a greater affinity with species of sect. Ericopsis than Fasciculus. Penstemon pinifolius groups with members of sect. Ericopsis in both analyses of this study (Figs. 2, 4), and we propose that it be moved from sect. Fasciculus to sect. Ericopsis.

Incongruence, hybridization, and polyploidy
The topologies of the ITS and cpDNA strict consensus trees (Figs. 1–4) differ in the placement of many taxa. Some of these differences are minor and may reflect the lack of characters to resolve the topologies (Fig. 5). Other differences are apparently large (e.g., the placement of Chionophila tweedyi in Fig. 1 vs. Fig. 3, the Ericopsis clade in Fig. 1 vs. Fig. 4).

One probable explanation for the differences seen in the ITS and cpDNA trees is hybridization (Viehmeyer, 1958; Straw, 1955 , 1956a , b ; Crosswhite, 1965a ; Wilson and Valenzuela, 2002 ). Hybrid speciation and introgression have been documented in Penstemon (Wolfe and Elisens, 1993 , 1994 , 1995 ; Wolfe et al., 1998a , b ; Datwyler and Wolfe, 2004 ), as well as polyploidy (Figs. 1–4). Most of the polyploid taxa are basal lineages within the genus in sections Penstemon and Saccanthera. Whether these are allopolyploid or autopolyploid has not yet been examined using a molecular approach, but given the ability of many species of Penstemon to hybridize, we hypothesize that many of the polyploid species are allopolyploids. Keck (1945) examined the cytology of many species in sections Penstemon and Saccanthera and hypothesized that P. attenuatus, P. subserratus, and P. wilcoxii are allopolyploids. He (Keck, 1932 ) also proposed allopolyploidy in the origin of P. neotericus, P. azureus, and P. heterophyllus in his monograph of the Saccanthera species.

Penstemon appears to have recently undergone a rapid evolutionary radiation (Fig. 5). Reproductive boundaries between species seem to be imperfect, which allows for hybridization where species occur in sympatry. Chloroplast-capture events have been found in at least one hybrid complex of sect. Peltanthera (Wolfe and Elisens, 1995 ). Thus, it is unsurprising that there is incongruence between the nuclear and plastid gene trees.

Floral evolution in Penstemon
Pennell (1935 , p. 198) hypothesized that bee pollination was the pleisiomorphic condition in Penstemon and that "evolution has clearly progressed from pollination by bees to pollination by moths, butterflies or hummingbirds." Our results support this hypothesis. Each of the traditional subgenera of Penstemon, characterized by different anther dehiscence patterns, has species pollinated by hummingbirds (Figs. 1–4). Many of the taxonomic anomalies resulted from past researchers placing together hummingbird-pollinated species (e.g., sect. Elmigera) or from uncertainty as to what to do with them (e.g., P. pinifolius). We scored the occurrence of hummingbird floral morphology onto the ITS and cpDNA trees (Fig. 1–4). Species of Penstemon with this syndrome occur in at least 10 separate clades in each of the phylogenies presented here (Figs. 1–4), and, accounting for incomplete taxon sampling, many more originations are possible. These results are congruent with and an expansion of those reported in Castellanos et al. (2006) , and Wilson et al. (2006) . Several recent studies (Thomson et al., 2000 ; Castellanos et al., 2003 , 2004 , 2006 ; Wilson et al., 2006 ) have documented the characteristics of Penstemon flowers that accompany the shift from bee- to bird-pollination syndromes. These studies have demonstrated that pollen presentation and pollination efficiency differ in species pollinated primarily by bees vs. birds. Thus, the taxonomic use of anther dehiscence characters is called into question by their apparent convergence.

The majority of Penstemon species are pollinated by hymenopterans, and numerous groups of bees are specialized for a limited suite of flowers (Crosswhite and Crosswhite, 1966 ; Tepedino et al., 1995). Very few species of Penstemon have moths as their primary pollinator (e.g., P. albidus, P. guadalupensis, and P. tubaeflorus; Pennell, 1935 ). We included P. albidus and P. tubaeflorus in this survey, and they are placed toward the tips of the tree in their respective clades (Figs. 2 and 4). Pennell (1935) hypothesized that the large variety of pollinators in Penstemon accounts for the diversity of flowers in the genus. We concur and hypothesize that pollinators, and the selection for particular floral traits, are important driving forces for the diversification of the genus (Wilson et al., 2006 ).

Biogeography of Penstemon
Wolfe et al. (2002) examined the biogeography of Cheloneae, concluding that the North American members of the tribe diversified initially in the Klamath Region with subsequent migration of elements to the Rocky Mountains. These findings were in agreement with Straw's (1966) hypotheses as to the origins of Chionophila and Penstemon in the Rocky Mountain/Columbia Plateau region. Datwyler and Wolfe (2004) examined the biogeography of subg. Dasanthera and supported a Rocky Mountain origin for this subgenus with a subsequent radiation of species through the Cascade–Sierra Nevada cordillera. Subg. Dasanthera appears as a basal lineage in the genus (Figs. 1–5). Thus our biogeographic hypotheses for the genus as a whole will be based on a Rocky Mountain–Cascade-Sierra Nevada radiation of the basal lineage.

Geographic distributions (Table 2) of each species are shown on the ITS and cpDNA strict consensus trees (Figs. 1–4). The basal lineages of Penstemon are distributed in the Rocky Mountains, Pacific Northwest, Boreal Northwest, California and Baja California, and the Intermountain Region. Species with distributions into the southwestern USA, Mexico east of the Gulf of California, and Guatemala appear at the base of the terminal clade of Penstemon in both analyses, and then scattered throughout the most terminal clades (Figs. 2, 4). Species with distributions east of the Rockies are found predominantly in the most terminal clades of each analysis (Figs. 2, 4). Pennell (1935) considered the species of Penstemon in eastern temperate North America to be among the most derived of the genus, and our study supports this hypothesis. We propose the following biogeographic scenario for Penstemon: (1) an origin for the genus in the Rocky Mountains with subsequent migration to the Cascade–Sierra Nevada cordillera; (2) radiation of elements throughout North America west of the Cascade–Sierra Nevada cordillera; (3) migration throughout the Intermountain Region and then into southwestern North America; and (4) migration of elements over the Rocky Mountain cordillera into the Great Plains and then throughout eastern North America. The apparent, recent evolutionary radiation (Fig. 5) in Penstemon, suggests that this biogeographic scenario is correlated with events in the Pleistocene.

APPENDIX

Voucher information and GenBank accession numbers. # = ITS, = trnT-L, and = trnC-D sequences. Underlined names and voucher information indicates that ITS only was sequenced for a particular taxon

Taxon—Voucher specimen (Herbarium); GenBank nos.

Antirrhinum majus L.—Wolfe s.n. (OS); #AF375150, DQ531193, DQ531338.

Chelone lyonii Pursh—Nelson 131 (OKL); #AF375163, DQ531202, DQ531347. C. obliqua L.—Wolfe 586 (OS); #AF375164, DQ531203, DQ531348.

Chionophila jamesii Benth.—Wolfe 473 (OS); #AF375167, DQ531204, DQ531349. C. tweedyi (Canby & Rose) L.F. Hend.—Datwyler 110 (OS); #AF375166, DQ531205, DQ531350.

Collinsia heterophylla Graham—Wolfe s.n. (OS); #AF375153, DQ531198, DQ531343.

Keckiella breviflora (Lindl.) Straw—Wilson 3487 (OS); #AF375161, DQ531199, DQ531344. K. cordifolia (Benth.) Straw—Wilson 3513 (OS); #AF375162. K. corymbosa (Benth.) Straw—Wolfe 437 (OS); #AF375160. K. lemmonii (Gray) Straw—Wolfe 436 (OS); #AF375158. K. rothrockii (Gray) Straw—Wilson 3512 (OS); #AF375159. K. ternata (Torr.) Straw—Valenzuela 43 (OS); #AF375157, DQ531200, DQ531345.

Linaria vulgaris P. Mill.—Nickrent s.n. (SIU); #DQ531053, DQ531194, DQ531139.

Nothochelone nemorosa (Douglas ex Lindl.) Straw—Wolfe 603 (OS); #AF375165, DQ531201, DQ531346.

Pennellianthus frutescens (Lambert) Crosswhite—Wolfe s.n. (OS); #AF375156, DQ531197, DQ531342.

Penstemon acaulis L. O. Williams—Dorn 7961 (OS); #DQ531054. P. acuminatus Douglas ex Lindl.—Datwyler 3 (OS); #DQ534923, DQ531206, DQ531351. P. alamosensis Pennell & Nisbet—Wolfe 813 (OS); #DQ534924. P. albertinis Greene—Walker 261 (OS); #DQ531055, DQ531207, DQ531352. P. albidus Nutt.—Dorn 7949 (OS); #DQ531056, DQ531208, DQ531353. P. albomarginatus M. E. Jones—Anderson 97-15 (ASU); #DQ531057, DQ531209, DQ531354. P. alpinus Torr.—Thomson 96-6 (OS); #DQ534925, DQ531210, DQ531355. P. ambiguus Torr.—Wolfe 837 (OS); #DQ531058, DQ531211, DQ531356. P. amphorellae Crosswh.—Wolfe 825 (OS); #DQ534926, DQ531212, DQ531357. P. anguineus Eastw.—Wolfe 537 (OS); #DQ534927, DQ531213, DQ531358. P. angustifolius Nutt. ex Pursh—Dorn 7947 (OS); #DQ531059, DQ531214, DQ531359. P. arenicola A. Nels.—Pate and Porter 9002 (RM); #DQ534928, DQ531215, DQ531360. P. aridus Rydb.—Lutz s.n. (OS); #D1531060, DQ531216, DQ531361. P. attenuatus Dougl. ex Lindl.—Wolfe 914, Wolfe 637 (OS); #DQ531061, DQ531217, DQ531362. P. auriberbis Pennell—Wolfe 843 (OS); #DQ534929. P. azureus Benth.—Wolfe 514 (OS); #AF375178, DQ531218 , DQ531363.

Penstemon baccharifolius Hook.—Wolfe 839 (OS); #DQ534930, DQ531219, DQ531364. P. barbatus (Cav.) Roth—Wolfe 509, Wolfe 778 (OS); #AF375177, DQ531220, DQ531365. P. barrettiae A. Gray—Wolfe 605 (OS); #AY195634, DQ531221, DQ531366. P. bicolor (Brandeg.) Clokey & Keck—Wolfe 795, Wolfe 796 (OS); #DQ531062, DQ531222, DQ531367. P. breviculus (Keck) Nisbet & R.C. Jackson—Arft 15-51 (COLO); #DQ531063, DQ531223, DQ531368.

Penstemon caesius A. Gray—Wilson 3482 (OS); #DQ531064, DQ531224, DQ531369. P. caespitosus var. caespitosus Nutt. ex A. Gray—Wolfe 782 (OS); #DQ531065, DQ531225, DQ531370. P. caespitosus var. desertipicti Nutt. ex A. Gray—Holmgren 12846 (NY); #DQ531066. P. californicus (Munz & Johnston) Keck—Wolfe 915 (OS); #DQ531067. P. campanulatus (Cav.) Willd.—Wolfe s.n. (OS); #DQ534931, DQ531226, DQ531371. P. canescens (Britt.) Britt.—Wolfe 854 (OS); #DQ534932. P. cardwellii T.J. Howell—Datwyler 11 (OS); #DQ531068, DQ531227, DQ531372. P. carnosus Pennell—Wolfe 757 (OS); #DQ531069, DQ531228, DQ531373. P. caryi Pennell—Lutz C1, Dorn 8005 (OS); #DQ531070, DQ531229, DQ531374. P. centranthifolius Benth.—Wilson 3521, Wolfe 308 (OS); #DQ531071, DQ531230, DQ531375. P. clevelandii A. Gray—Wilson s.n. (OS); #DQ531072, DQ531231, DQ531376. P. clutei A. Nels.—Wolfe 810 (OS); #DQ531073, DQ531232, DQ531377. P. cobaea Nutt.—Wolfe 838 (OS); #DQ534933, DQ531233, DQ531378. P. comarrhenus A. Gray—Wolfe 812 (OS); #DQ531074, DQ531234, DQ531379. P. confertus Dougl. ex Lindl.—Wolfe 635 (OS); #DQ534934, DQ531235, DQ531380. P. confusus M.E. Jones—Wolfe 458, Wolfe 748 (OS); #DQ531075, DQ531236, DQ531381. P. crandallii var. atratus Keck—Holmgren 12824 (NY); #DQ531076. P. crandallii var. crandallii A. Nels.—Thomson 96-20 (OS); #DQ531077, DQ531237, DQ531382. Penstemon crandallii var. glabrascens (Pennell) Keck—Wolfe 840 (OS); #DQ531078. P. cyananthus Hook.—Wolfe 672 (OS); #DQ534935, DQ531238, DQ531383. P. cyaneus Pennell—Datwyler 65 (OS); #DQ531079. P. cyanocaulis Payson—Wolfe 755 (OS); #DQ531080. P. cyathophorus Rydb.—Wolfe 774 (OS); #DQ534936.

Penstemon dasyphyllus A. Gray—Freeman 022 (UTEP); #DQ531081, DQ531239, DQ531384. P. davidsonii Greene—Cultivated, s.n., Datwyler 37 (OS); #AY195637, DQ531240, DQ531385. P. debilis O'Kane & J. Anderson—McMullen s.n. (OS); #AF375180, DQ531241, DQ531386. P. deustus var. suffrutescens L. Henderson—Wilson 3551 (OS); #DQ531082. P. deustus var. variabilis (Suksdorf) Cronq.—Wolfe 628 (OS); #AF375182, DQ531242, DQ531387. P. digitalis Nutt. ex Sims—Lindgren s.n. (NEB); #DQ531083, DQ531243, DQ531388. P. dissectus Ell.—Leege s.n. (OS); #DQ534937, DQ531244, DQ531389. P. dolius M.E. Jones ex Pennell—Wolfe 818 (OS); #DQ531084, DQ531245, DQ531390.

Penstemon eatonii Gray—Wolfe 508, Holmgren 12841 (OS, NY); #DQ534938, DQ531246, DQ531391. P. ellipticus Coult. & Fisher—Datwyler 47 (OS); #AF375168, DQ531247, DQ531392. P. eriantherus var. cleburnei (M.E. Jones) Dorn—Hartman 20034 (RM); #DQ531085, DQ531249, DQ531394.P. eriantherus var. eriantherus Pursh—Dorn 7952. Datwyler 64 (OS); #DQ531086, DQ531248, DQ531393. P. euglaucus English—Wolfe 613 (OS); #DQ531087.

Penstemon fasciculatus A. Gray—Wilson 3590 (OS); #DQ531088. P. fendleri Torr. & A. Gray—Wolfe 836 (OS); #DQ531089, DQ531250, DQ531395. P. floridus Brandeg.—Aldridge and Crandall s.n. (OS); #DQ531090. P. fremontii Torr. & A. Gray ex A. Gray—Wolfe 788 (OS); #DQ531091, DQ531251, DQ531396. P. fruticiformis Coville—Wolfe 819 (OS); #DQ531092, DQ531252, DQ531397. P. fruticosus (Pursh) Greene—Datwyler 29, Datwyler 35 (OS); #AF375171, DQ531253, DQ531398.

Penstemon gairdneri Hook.—Wolfe 925 (OS); #DQ531093. P. gentianoides (Humboldt, Bonpland & Kunth) Poiret—James 99.0017, Wilson 3606 (OS); #DQ531094, DQ531254, DQ531399. P. gentryi Standl.—Wilson 3598 (OS); #DQ531095, DQ531255, DQ531400. P. gibbensii Dorn—Dorn 7953 (OS); #DQ531096. P. glaber Pursh—Dorn 8032 (OS); #DQ531097. P. glandulosus Dougl.—Wolfe 639 (OS); #DQ531098. P. globosus (Piper) Pennell & Keck—Wolfe 643 (OS); #DQ531099, DQ531256, DQ531401. P. gormanii Greene—Armbruster s.n. (ALA); #DQ531100, DQ531257, DQ531402. P. gracilentus Gray—Wilson 3542 (OS); #DQ531101, DQ531258, DQ531403. P. gracilis Nutt.—Wolfe 830 (OS); #DQ531102. P. grandiflorus Nutt.—Wolfe 852 (OS); #DQ534939. P. griffinii A. Nels.—Pate and Porter 10299 (RM); #DQ531103, DQ531259, DQ531404. P. grinnellii Eastw.—Wolfe 306, Wolfe 299 (OS); #DQ531104, DQ531260, DQ531405.

Penstemon harbourii Gray—Thomson s.n. (OS); #DQ531105, DQ531261, DQ531406. P. harringtonii Penl.—Wolfe 784 (OS); #DQ534940, DQ531262, DQ531407. P. hartwegii Benth.—Wolfe s.n. (OS); #DQ531106, DQ531263, DQ531408. P. havardii Gray—Wolfe 849 (OS); #DQ531107, DQ531264, DQ531409. P. haydenii S. Wats.—Stubbendeck s.n. (OS); #DQ531108. P. heterodoxus Gray—Wilson 3502 (OS); #DQ531109. P. heterophyllus var. heterophyllus Lindl.—Wilson 3454 (OS); #DQ531110, DQ531265, DQ531410. P. heterophyllus var. purdyi Keck—Wolfe 574 (OS); #AF375181. P. hirsutus (L.) Willd.—Wolfe s.n. (OS); #DQ531111, DQ531266, DQ531411. P. humilis Nutt. ex Gray—Datwyler 4 (OS); #DQ531112, DQ531267, DQ531412.

Penstemon immanifestus N. Holmgren—Wolfe 785 (OS); #DQ531113. P. incertus Brandeg.—Wolfe 465, Valenzuela s.n. (OS); #DQ534941, DQ531268, DQ531413. P. inflatus Crosswhite—Wolfe 811 (OS); #DQ531114, DQ531269, DQ531414. P. isophyllous Robinson—James 3 (OS); #DQ531115, DQ531270, DQ531415.

Penstemon jamesii Benth.—Wolfe 807 (OS); #DQ531116, DQ531271, DQ531416.

Penstemon kunthii G. Don—Wolfe 834, Wilson 3594 (OS); #DQ531117, DQ531272, DQ531417.

Penstemon labrosus (Gray) Hook. f.—Hogue 87.7 (OS); #DQ531118, DQ531273, DQ531418. P. laetus var. laetus Gray—Wolfe 665 (OS); #DQ531119. P. laetus var. saggitatus Keck—Wilson 3550 (OS); #DQ531120, DQ531274, DQ531419. Penstemon lanceolatus Benth.—Wilson 3608 (OS); #DQ534942, DQ531275, DQ531420. P. laricifolius Hook. & Arn.—Thomson 96-12, Dorn 8004 (OS); #DQ531121, DQ531276, DQ531421. P. laxiflorus Pennell—Freeman 263 (UTEP); DQ534943. P. lemhiensis (Keck) Keck & Cronq.—Crofts s.n. (RM); #DQ531122, DQ531277, DQ531422. P. lentus Pennell—Wilson 3453 (OS); #DQ531123, DQ531278, DQ531423. P. leonardii Rydb.—Wolfe 760 (OS); #DQ531124, DQ531279, DQ531424. P. leonensis Straw—James 99.02 (OS); #DQ531125, DQ531280, DQ531425.P. linarioides Gray—Lindgren s.n. (NEB); #DQ531126, DQ531281, DQ531426. P. lyallii (Gray) Gray—Datwyler 46, Datwyler 44 (OS); #DQ531127, DQ531282, DQ531427.

Penstemon mensarum Pennell—Thomson 96-17 (OS); #DQ531128, DQ531283, DQ531428. P. moffatii Eastw.—Stermitz 515 (CMML); #DQ531129. P. montanus Greene—Datwyler 51 (OS); #AF375169, DQ531284, DQ531429. P. multiflorus Chapman ex Benth.—Wolfe s.n. (OS); #DQ531130, DQ531285, DQ531430.

Penstemon neomexicanus Woot. & Standl—Wolfe 828 (OS); #DQ531131, DQ531286, DQ531431. P. neotericus Keck—Aldridge and Crandall s.n. (OS); #DQ531132, DQ531287, DQ531432. P. newberryi Gray—Wilson 3484 (OS); #DQ531133, DQ531288,