Volume 98, Issue 4 p. 731-753
Systematics and Phytogeography
Free Access

Phylogenetic perspectives on diversification, biogeography, and floral evolution of Collinsia and Tonella (Plantaginaceae)

Bruce G. Baldwin

Corresponding Author

Bruce G. Baldwin

Jepson Herbarium and Department of Integrative Biology, 1001 Valley Life Sciences Building #2465, University of California, Berkeley, California 94720-2465 USA

Author for correspondence ([email protected])Search for more papers by this author
Susan Kalisz

Susan Kalisz

Department of Biological Sciences, University of Pittsburgh, 4249 Fifth Avenue, Pittsburgh, Pennsylvania 15260 USA

Search for more papers by this author
W. Scott Armbruster

W. Scott Armbruster

School of Biological Sciences, King Henry Building, King Henry I Street, University of Portsmouth, Portsmouth PO1 2DY, UK

Department of Biology, NTNU, Realfagbygget, Gløshaugen, N-7491, Trondheim, Norway

Institute of Arctic Biology, University of Alaska, P.O. Box 757000, Fairbanks, Alaska 99775-7000 USA

Search for more papers by this author
First published: 01 April 2011
Citations: 40

This paper is dedicated to the memory of E. C. Neese, who was a major source of encouragement and help with collecting for this project. The authors especially thank M. S. Park, A. M. Randle, and C. Richey for their invaluable collections and interpretations of Collinsia and Tonella populations, M. S. Park for technical assistance, and B. L. Wessa and T. Yi for extensive help with generating molecular data. The authors also thank S. J. Bainbridge, B. A. Bennett, E. Elle, K. M. Hanley, C. Ivey, Å. Lankinen, W. Legard, J. McGraw, A. Moore, E. Painter, J. Paul, M. J. Sanderson, S. P. Schechter, M. Wetherwax, M. F. Wojciechowski, and J. W. Wright for help with collecting; the Land Trust of Napa County and Wantrup Wildlife Sanctuary for assistance in fieldwork; C. Christie, J. Davis, S. Matson, K. Morse, K. Robertson, and A. Schusteff for permission to reproduce their plant images in Figs. 1–6; and M. P. Simmons and two anonymous reviewers for helpful comments on the manuscript. This research was supported by grants from the National Science Foundation (DEB-0324733) to B.G.B., (DEB-0324764) and (DEB-0709638) to S.K., and (DEB-0324808) to W.S.A. and the Lawrence R. Heckard Endowment Fund of the Jepson Herbarium (to B.G.B.).

Abstract

Premise of the study: Collinsia was the subject of classic biosystematic studies by Garber and colleagues and is increasingly investigated to address major evolutionary questions. Lack of phylogenetic data from more than one gene region and one taxonomic exemplar has left relationships, diversity, and phytogeography of Collinsia in question and has limited understanding of its diversification.

Methods: Phylogenetic analyses representing 179 populations of Collinsia and closely related Tonella were conducted based on DNA sequences of nuclear ribosomal transcribed spacers, the single-copy nuclear gene CYCLOIDEA-1, and part of the chloroplast matK/trnK intron region to reexamine systematic hypotheses and extend understanding of the importance of floral characters, chromosome evolution, interfertility, crossability, hybridization, edaphic factors, and ecogeographic barriers to diversification in the group.

Key results: Informal “sections” of Collinsia are artificial, although pedicel length and other traditional deep-level taxonomic characters are more conservative evolutionarily than flower size. Evolutionary loss of crossability and interfertility in Collinsia appears to be largely a byproduct of divergence. Although most taxa appear to have arisen by divergent evolution, multiple lines of evidence indicate a homoploid hybrid constitution of C. tinctoria, possibly explaining an occurrence of convergent chromosome evolution. Phylogeographic and cryptic diversity is extensive.

Conclusions: Diversity in Collinsia is greater than previously documented. Recently divergent lineages are often associated with distinct habitat (including soil) and geographic factors, different flower sizes, and contrasting chromosomal arrangements. Evidence for a hybrid constitution of diploid C. tinctoria is consistent with lack of strong intersterility barriers between closely related taxa.

The need for a detailed phylogenetic understanding of Collinsia Nutt. (Plantaginaceae) has grown with increasing attention to the genus as a subject of diverse studies in mating-system biology and evolution (100; 85; 53; 13; 69; 55, 21; 3; 22; 54; 21; 75; 61; 63; 62, 60; 58, 59; 60; 68; 80) and in adaptation to serpentine and nonserpentine habitats (102, 104; 103; 65; 87). A preliminary phylogenetic hypothesis, based on nuclear rDNA internal transcribed spacer region (ITS) sequences, indicated that floral evolution of Collinsia has been dynamic, with repeated shifts in flower size and development (3) associated with differences in ability to autonomously self-pollinate (80). Additional sampling of populations across Collinsia revealed sufficient morphological and ecological variation to warrant a more extensive phylogenetic investigation and a reconsideration of the systematics and historical biogeography of the genus, which was the subject of intensive biosystematic and cytogenetic studies by Garber and colleagues in the 1950s and 1960s (see 30).

Members of Collinsia are spring-flowering annuals that occur in a wide diversity of habitats from sea level to ∼4000 m a.s.l. in temperate North America, with most taxa confined to the California Floristic Province (CA-FP) (81) and only three that occur east of the Rocky Mountains (C. parviflora, C. verna, and C. violacea). All species except for the two central and eastern North American endemics (C. verna and C. violacea) occur at least in part in California. 72, in the most recent, comprehensive taxonomic treatment of Collinsia, recognized 28 minimum-rank taxa (17 species plus 11 additional varieties). In subsequent treatments of the genus for the Pacific states or for California only, 76, 70, and 71 adopted taxonomies similar to Newsom's, with 22–26 minimum-rank taxa (16–19 species), not counting the two exclusively central and eastern North American species. The only species of Collinsia described since 72 monograph, C. antonina, was treated as a synonym of C. parryi by 71 and as a distinct species by 6 based on morphological considerations and molecular (ITS) evidence for a closer relationship of C. parryi to C. concolor than to C. antonina.

Tonella Nutt. ex A. Gray, the putative sister genus of Collinsia, comprises two species of spring-flowering annuals of the far-western United States. One species occurs widely in the Pacific states (T. tenella; Fig. 1); the other is restricted to the Snake River drainage of the Pacific Northwest (T. floribunda). Collinsia and Tonella differ most conspicuously in whether the central lower lobe of the corolla is keeled and encloses the stamens and style (in Collinsia), much as in a papillionoid legume (see Figs. 2–6), or not (in Tonella).

Details are in the caption following the image

Flowers and inflorescences of Collinsia and Tonella. 1. T. tenella (scale bar = 1 mm). Image by John Davis. 2. C. wrightii (scale bar = 2 mm). Image by Keir Morse. 3. C. verna (scale bar = 10 mm). Image by Kenneth Robertson. 4. C. corymbosa (scale bar = 10 mm). Image by Steve Matson. 5. C. tinctoria (scale bar = 10 mm). Image by Aaron Schusteff. 6. C. heterophylla (scale bar = 10 mm). Image by Christopher Christie.

Although no formal subgenera or sections of Collinsia have been described, relationships among the species have been implied or suggested on the basis of morphological and biosystematic criteria. Gray's (1880, 39) primary separation of Collinsia into two informal groups differing in pedicel length in his key to species was adhered to by subsequent authors including 72, p. 262), who noted that “… length of the pedicels forms the basis for the most conspicuous division of the genus into the sessile- and pedicel-flowered groups, and these in turn are easily arranged in series according to bearding of staminal filaments or size and shape of corollas.” 38, 39) and 72 also referred to differences in seed size and shape (flattened vs. thickened) as important characteristics for distinguishing species and, in part, for diagnosing the sessile- vs. pediceled-flowered species groups.

Garber and colleagues (e.g., 27, 28, 29; 33; 1; 10; 32; 34) based their biosystematic and cytogenetic studies of Collinsia on the informal taxonomic framework of sessile- and pediceled-flowered “sections” and discovered that grouping of taxa by chiasma frequencies, patterns of crossability between taxa, and levels of chromosomal association, chiasma formation, and interfertility in hybrids were to some extent consistent with previously suggested relationships or groups in the mostly diploid (x = 7) genus. They found evidence for extensive chromosomal evolution by reciprocal translocations and paracentric inversions, with most widely recognized taxa distinguished by at least one chromosomal interchange. Notwithstanding a sometimes striking lack of correlation between levels of interfertility and degree of chromosomal structural divergence between taxa, 32, p. 291) noted that crossability, interfertility, and chromosomal associations at meiosis I “… have, in most cases, provided a reasonable basis for recognizing taxa in this genus” and for judging the relative merits of previous taxonomies of Collinsia.

Garber and colleagues inferred explicit chromosomal arrangements for most studied taxa of Collinsia on the basis of cytogenetic evidence for arm interchanges (29; 1; 10; 32), yet an overall phylogenetic hypothesis based on those and other biosystematic data was not proposed. 29 and 1 suggested an incipient role for chromosomal repatterning in evolutionary divergence leading to speciation, although they cautioned about an absence of evidence for variation in chromosomal arrangements within natural populations, with one relatively recently documented exception (15). Polyploidy was regarded by Garber and colleagues as of relatively minor importance in speciation within Collinsia; despite repeated documentation of spontaneous polyploids or polyploid branches among progeny of artificial hybrids (e.g., 48; 33; 1; 44), stable polyploids have been documented in the wild only in the hexaploid C. torreyi s.s. (2n = 21II; 28) and, subsequently, in Canadian (tetraploid) populations of C. parviflora (2n = 14II; 26) and other members of the C. grandifloraC. parviflora clade (E. Elle, Simon Fraser University, personal communication).

Here, we examine Collinsia and its putative sister genus, Tonella, from a molecular phylogenetic perspective to resolve evolutionary patterns and processes and historical biogeography in the two genera, with attention to hypotheses from prior evolutionary studies based on morphological, biosystematic, and ecological data. In particular, we focus on the following questions: (1) Does the keeled floral lobe in Collinsia diagnose a monophyletic group or does Tonella belong within Collinsia, as suggested by 40 on the basis of floral polymorphism? (2) Does the traditional division of Collinsia into sessile- and pediceled-flowered groups reflect phylogeny or has pedicel length been less conservative evolutionarily? (3) Does crossability between taxa or levels of chromosomal association or interfertility in hybrids reflect relationships within Collinsia, as suggested by Garber and colleagues, or have rates of chromosomal evolution or of loss of crossability or interfertility been too irregular to reflect phylogeny? (4) Has hybridization had evolutionary consequences in Collinsia (5) Does the history of chromosomal interchanges estimated by Garber and colleagues contain phylogenetic signal? (6) Have previously undetected “cryptic species” evolved within Collinsia, as suspected by 34 (7) Do patterns and timing of diversification in Collinsia provide any insights into processes of speciation or the biogeographic history of the genus?

MATERIALS AND METHODS

Taxon sampling

We sampled representatives of 179 populations of Collinsia and Tonella, with attention to capturing ecological, geographic, and morphological variation and taxonomic diversity in the two genera. All species and most or all varieties recognized by 72, 76, 70, and 71 were represented by samples from 2–21 populations each. The outgroup consisted of up to three additional taxa in Chelone L., Keckiella Straw, and Penstemon Schmidel that represent the three principal clades of the sister group to Collinsia and (putatively) Tonella in tribe Cheloneae, based on results of 101. Collection and voucher data are presented in Appendix 1. All vouchers are at JEPS and UC.

Gene regions sampled

For all samples, we examined 18S-26S nuclear ribosomal DNA (nrDNA) sequences of the internal transcribed spacer region (ITS), i.e., ITS-1, 5.8S subunit, and ITS-2, and a 399–405 bp portion of the 3′ end of the external transcribed spacer (ETS) immediately upstream of the 18S subunit. Chloroplast DNA (cpDNA) sequences of the 3′ intron region of trnK and a portion of matK were obtained for the same, complete set of samples. For a more limited subset of samples representing major clades identified with ITS, ETS, and chloroplast data, we obtained sequences of a CYCLOIDEA-like locus, CYCLOIDEA-1 (CYC1), a single-copy nuclear gene that is putatively homologous with TCP transcription factors involved in floral symmetry in Antirrhinum majus L. (see 47).

DNA sequencing

Total DNA was extracted from fresh, frozen, or silica-dried leaves of a single plant using a modification of 17 method (adding a phenol extraction, RNase digestion, and two ethanol precipitations of DNA) or the DNeasy Plant Mini Kit (Qiagen, Valencia, California), following the manufacturer's protocol except with 1–2 h (rather than 10 min) of incubation (cell lysis step). Polymerase chain reaction (PCR) amplifications of the ITS region followed the methods of 9 except for use of AccuPower PCR Premix (K-2016; Bioneer Corp., Chungbuk, Korea). Methods of 7 were used to amplify the intergenic spacer of nrDNA and sequence upstream from the 18S subunit to construct a primer, Col-ETS (5′-GGCATATTGGATCCCTGCT-3′), used for subsequent amplification and sequencing (with primer 18S-ETS; 7) of 399–405 bp at the 3′ end of the ETS. PCR conditions for amplification and sequencing of ETS sequences were identical to those for the ITS region except for annealing temperature (60°C). The cpDNA 3′ intron segment of trnK and the 3′ end of matK were amplified and sequenced using primers matK 8 and trnK 2r designed by 95, with corrected reporting of trnK 2r by 52. PCR conditions for the cpDNA region were identical to those for the ITS region.

CYC-like sequences were amplified and sequenced using primers P1 and P2 (98) and methods of 46. Amplification products were cloned using the Zero Blunt TOPO PCR Cloning Kit for Sequencing (K2875, Invitrogen Corp., Carlsbad, California, USA). Cloning of PCR products was undertaken for ITS and ETS sequences only when sequences from pooled PCR products could not be resolved unambiguously or were highly polymorphic. All sequences of CYC-like genes were from individual cloned sequences. Of the two CYC-like paralogues identified among the cloned sequences of Collinsia and relatives by subsequent phylogenetic analysis (CYC1 and CYC2), only CYC1 yielded sufficient clones and variation for study here. Both DNA strands were sequenced for all samples. Exonuclease I and shrimp alkaline phosphatase were used to remove excess nucleotides from PCR products using the PCR Product Pre-Sequencing Kit (70995, United States Biochemical Corp., Cleveland, Ohio, USA). Sanger sequencing of PCR products was conducted at the UC Berkeley DNA Sequencing Facility (Barker Hall) with the same primers used for PCR (except ITS5 was used instead of ITS-I for sequencing of the ITS region). GenBank accession numbers for sequences are provided in Appendix 1.

Phylogenetic analyses

DNA sequences were aligned manually based on the similarity criterion (91), with special attention to codon structure in the length-variable CYC1 and 3′ end of matK. Gaps were treated as missing data. A concatenated matrix was produced for the entire data set. Lack of allelic or copy-type variation in more than one of the four gene regions (i.e., ITS, ETS, cpDNA, or CYC1) for all but one sample allowed for duplicate copies of sequences of three regions to be concatenated with each distinct cloned sequence from the fourth region (the lone exception, Collinsia tinctoria from Napa County, California was included only in separate analyses of ITS and ETS data; data not shown). Separate and simultaneous (combined) phylogenetic analyses of the four data sets were conducted to examine congruent and potentially conflicting signal in the different gene regions and to allow for analyses that either maximized taxon sampling or maximized sequence data (as noted above, the CYC1 data set included samples from only a subset of populations examined for the other three gene regions). The incongruence length difference (ILD) test (23), as implemented in the program package PAUP* version 4.0b10 (97), as the partition homogeneity test, also was conducted to assess potentially conflicting phylogenetic signal across the ITS, ETS, cpDNA, and CYC1 data sets, using 1000 replicates, with 10 random addition sequences per replicate and MulTrees set to 1.

Phylogenetic criteria used to estimate relationships and clade support included maximum parsimony (MP), using PAUP*, and Bayesian Markov chain Monte Carlo (MCMC) inference (BI; 105), using the programs MrBayes 3.1 or 3.2 (84) and BEAST v1.5.4 (19). For computational efficiency, samples with identical sequences were treated as one operational taxonomic unit (OTU). For BI analyses, parameters for each data partition (nrDNA, cpDNA, or CYC1) were obtained using the program MrModeltest 2.3 (73) or, for BEAST runs, Modeltest 3.7 (77) using the Akaike information criterion (2). Chosen parameters were as follows: nrDNA and CYC1: GTR + I + Γ; cpDNA: GTR + G for MrBayes or TIM + G for BEAST, with editing of the BEAST XML file). For each MCMC analysis of separate or combined (partitioned) data in MrBayes, four independent runs were conducted, each using three “heated” chains and one “cold” chain, for 10–15 million generations (saving one tree every 1000 generations). All trees obtained prior to a decrease in the standard deviation of split frequencies below 0.01 were discarded as burn-in. To obtain ultrametric trees, sequence data were analyzed using BEAST under a relaxed clock (uncorrelated log normal [18]) and Yule process of speciation, with four gamma categories and constrained monophyly of the ingroup. In the absence of a known fossil record for Collinsia or Tonella, we estimated maximum ages for nodes by calibrating the most recent common ancestor (MRCA) of Collinsia and Tonella at 15 million years ago (Ma), under the assumption that diversification of this principally western North American clade of drought-avoiding spring annuals would not have begun prior to the mid-Miocene onset of summer-drying in the region (5; 25; see 8). To do so, we selected a uniform prior distribution for the ingroup MRCA, with a lower setting of 14.999 and an upper setting of 15.001. Each of four independent MCMC analyses were run for 10 million generations (saving one tree every 1000 generations) when the program Tracer v1.5 (78) indicated that the effective sample size of the posterior distribution was >1000 across runs, with a burn-in of 25%. (Note: Veracity of a mid-Miocene calibration as a maximal age for the MRCA of Collinsia and Tonella was upheld by results of BI analysis of the expanded nrDNA + cpDNA data set using BEAST under a relaxed clock [75 million generations × 4 chains, sampling every 10000 generations, with 10% burn-in for each chain], with specification of a normally distributed prior for the mean substitution rate of the ITS-1 + ITS-2 partition [instead of a nodal calibration] that corresponded to the mean [4.13 × 10−9 substitutions/site/year or sub/site/y] and standard deviation [1.81 × 10−9 sub/site/y] of ITS rates estimated from calibrated trees for a wide diversity of herbaceous angiosperms [57]: The resulting mean age estimate for the MRCA of Collinsia and Tonella was 12.5 Ma [95% highest probability density or HPD = 5.8–20.6 Ma], with an estimated mean ITS rate of 5.92 × 10−9 sub/site/y [95% HPD = 2.92–8.95 × 10−9 sub/site/y]). For MP analyses of the nrDNA + cpDNA + CYC1 (all partitions) data set, full heuristic searches were conducted using 1000 random addition sequences, with tree-bisection-reconnection (TBR) branch swapping and MulTrees on. All characters were given equal weight except for higher weights in some analyses for the single, ordered character of recoded chromosomal-interchange data (see below), which involved a step matrix. MP analyses of the expanded taxon data sets of nrDNA, cpDNA, or nrDNA + cpDNA were conducted using 1000 random addition sequences, with TBR branching swapping and MulTrees off, followed by TBR branch-swapping on saved trees with MulTrees on. Clade reliability based on MP was estimated using nonparametric bootstrapping (24), with 10000 replicates, 10 random addition sequences per replicate, and MaxTrees set to 1. For the BI analyses, posterior probabilities for each clade were obtained from a 50% majority-rule consensus of retained trees (minus burn-in). Sequence matrices and trees are deposited at TreeBASE (http://purl.org/phylo/treebase/phylows/study/TB2:S11078).

Character evolution and biogeographic history were estimated in the program package Mesquite 2.6 (67), using parsimony mapping for those polymorphic (in part) or multistate data. Pediceled/sessile flowers, stamen bearding, and seed shape for each species were scored as binary characters, as by 72 for seed and stamen traits and by Garber and collaborators for pedicels (e.g., 30), except for scoring of C. greenei and C. multicolor as polymorphic for sessile and pediceled flowers (see Discussion). The same characters were scored for C. antonina and C. “metamorphica” (not studied by Newsom or Garber) from herbarium specimens. Species floral size was determined by growing plants of all species from seed to flowering in a common greenhouse environment at the University of Pittsburgh. Floral measurements of corolla size were taken on 3–6 individuals/species. Mean flower size was calculated for each species and a grand mean flower size for the genus was determined. Flower size was also treated as a binary character: species means that were greater or smaller than the grand mean of the genus were scored as large- or small-flowered, respectively. Chromosome-arm arrangements proposed for different species of Collinsia by Garber and colleagues (29; 1; 10; 31) were treated as states of an ordered character and recoded in a step matrix for assessing genomic evolution by reciprocal translocations in a parsimony context. Evolution of intersterility was also explored in a phylogenetic context by considering pairwise data from crosses of Garber and colleagues (see 30) with respect to depth of divergence of taxa in the smaller taxon-set (all-partition) trees. For biogeographic analysis, two coding schemes were used, one with taxa scored as occurring in one or more of six areas (CA-FP, Pacific Northwest outside of CA-FP, Great Basin, Mojave/Sonoran Desert, Rocky Mountains, or Central-Eastern North America) and the other with taxa included in one or more of 11 areas (northwestern CA-FP, Cascade Range of CA-FP, Sierra Nevada, Great Central Valley, central western California, southwestern California, Great Basin, Mojave/Sonoran Desert, Pacific Northwest outside of CA-FP, Rocky Mountains, or Central-Eastern North America), with boundaries of regions within the CA-FP based on geographic subdivisions in The Jepson Manual (45) and extra-Californian CA-FP boundaries based on 81 delimitation of the region. Coding of each OTU was based on collection locality of that sample for the expanded-taxon (nrDNA and cpDNA) data set.

RESULTS

Phylogenetic analyses

Partition homogeneity test results indicated that ITS and ETS data are not significantly heterogeneous (P = 0.412), so the two nrDNA regions were treated as a common partition in most subsequent analyses. All pairwise comparisons among nrDNA, cpDNA, and CYC1 data suggested significant heterogeneity (P < 0.05), so the three regions were treated as separate partitions in subsequent phylogenetic analyses of separate and combined data. Incongruence in clade composition between trees based on different partitions was limited to clades with <90% MP bootstrap support in one or both trees. Clades supported by two or more partitions with ≥90% MP bootstrap support included each of the following groups: Collinsia, Tonella, C. bartsiifolia + C. corymbosa, C. concolor + C. parryi s.s. (i.e., excluding C. antonina), C. grandiflora + C. parviflora, C. linearis s.s. (i.e., excluding the C. “metamorphica” complex) + C. rattanii, the C. “metamorphica” clade, C. sparsiflora var. collina + C. sparsiflora var. sparsiflora (including var. arvensis), and C. verna + C. violacea. Simultaneous phylogenetic analysis of nrDNA, cpDNA, and CYC1 data yielded trees (Fig. 7) with generally higher MP bootstrap and BI posterior probabilities than trees based on individual gene regions (Figs. 8109). Groups resolved at ≥90% MP bootstrap support in the all-partitions trees (Fig. 7) that were less robust in results of separate data partitions (Figs. 8109) include a west–east clade, (C. grandiflora + C. parviflora) + (C. verna + C. violacea), and a clade containing the sessile-flowered taxa, i.e., C. bartsiifolia, C. corymbosa, C. heterophylla, C. multicolor, and C. tinctoria, plus C. antonina, C. greenei, C. parryi, and C. sparsiflora. Groups of diminished bootstrap support in the all-partitions trees that were supported at ≥90% MP bootstrap support in trees based on one or more individual gene regions were (1) the clade containing all sessile-flowered taxa + C. antonina and C. parryi (see Fig. 11) and (2) a (nrDNA and nrDNA+cpDNA) clade containing the montane species C. childii and the C. “metamorphica” complex (Figs. 8, 12), not recovered in the all-partitions maximum-clade-credibility (MCC) tree (Fig. 7). Posterior probabilities of BI trees from simultaneous analysis of all partitions were ≥95% for all clades supported by ≥90% bootstrap support in separate or combined-data MP analyses except for the nrDNA clade containing C. childii and the C. “metamorphica” clade; the BI combined all-partition analysis instead resolved a clade of 96% posterior probability with the C. “metamorphica” clade sister to a clade that included the sessile-flowered taxa plus C. antonina, C. greenei, C. parryi, and C. sparsiflora.

Details are in the caption following the image

Chronogram and maximum-clade-credibility tree of Collinsia and Tonella based on Bayesian phylogenetic analysis of combined sequences of nrDNA external and internal transcribed spacer (ETS and ITS) regions, 3′ matK/3′ trnK intron chloroplast DNA (cpDNA), and the single-copy nuclear gene CYCLOIDEA 1 (CYC1) using BEAST. Branch lengths are scaled to time, in million years, with maximal basal calibration at 15 Ma (see Materials and Methods); estimated maximum ages are given to the right of the node (blue-green bars delimit the 95% highest probability density for ages). Numbers following taxon names refer to population number (see Appendix 1). Bayesian posterior probabilities (left of slash) and maximum parsimony (MP) bootstrap values (right of slash) are along or left of tree nodes, for nodes supported at ≥0.95 posterior probability. One asterisk (*) = clade with <90% support in 0.5 majority-rule maximum parsimony bootstrap tree. Two asterisks (**) = clade not resolved in 0.5 majority-rule MP bootstrap tree. Outgroup not shown. Abbreviations: b. = bartsiifolia. C. = Collinsia. s. = sparsiflora. T. = Tonella.

Details are in the caption following the image

Phylogram of 0.5 majority-rule consensus tree for Collinsia and Tonella based on Bayesian phylogenetic analysis of nrDNA ETS and ITS sequences using MrBayes (scale in expected substitutions per site). Numbers following taxon names (or, for C. tinctoria, at branch tips) refer to population number (see Appendix 1); all sampled populations are represented for taxa without indicated numbers. Numbers in parentheses refer to different cloned sequences for C. tinctoria (ETS clones for sample 1; ITS clones for samples 4 and 8). Bayesian posterior probabilities (left of slash) and maximum parsimony (MP) bootstrap values (right of slash) are given for nodes supported at ≥0.95 posterior probability. One asterisk (*) = clade with <90% support in 0.5 majority-rule MP bootstrap tree. Two asterisks (**) = clade not resolved in 0.5 majority-rule MP bootstrap tree. Outgroup not shown. Taxa in colored font pertain to Hybridization and evolution section of Discussion. Abbreviations: C. = Collinsia. c. = clone(s). T. = Tonella.

Details are in the caption following the image

Phylogram of 0.5 majority-rule consensus tree for Collinsia and Tonella based on Bayesian phylogenetic analysis of cpDNA sequences using MrBayes (scale in expected substitutions per site). Numbers following taxon names refer to population number (see Appendix 1); all sampled populations are represented for taxa without indicated numbers. Bayesian posterior probabilities (left of slash) and maximum parsimony (MP) bootstrap values (right of slash) are given for nodes supported at ≥0.95 posterior probability. One asterisk (*) = clade with <90% support in 0.5 majority-rule MP bootstrap tree. Two asterisks (**) = clade not resolved in 0.5 majority-rule MP bootstrap tree. Outgroup not shown. Taxa in colored font pertain to Hybridization and evolution section of Discussion. Abbreviations: C. = Collinsia. s. = sparsiflora. T. = Tonella.

Details are in the caption following the image

Phylogram of 0.5 majority-rule consensus tree for Collinsia and Tonella based on Bayesian phylogenetic analysis of CYCLOIDEA-1 (CYC1) sequences using MrBayes (scale in expected substitutions per site). Numbers following taxon names (or, for C. heterophylla, at branch tips) refer to population number (see Appendix 1); all sampled populations are represented for taxa without indicated numbers. Numbers in parentheses refer to different cloned sequences of CYC1. Bayesian posterior probabilities (left of slash) and maximum parsimony (MP) bootstrap values (right of slash) are given for nodes supported at ≥0.95 posterior probability. One asterisk (*) = clade with <90% support in 0.5 majority-rule MP bootstrap tree. Two asterisks (**) = clade not resolved in 0.5 majority-rule MP bootstrap tree. Outgroup not shown. Taxa in colored font pertain to Hybridization and evolution section of Discussion. Abbreviations: c. = clone(s). C. = Collinsia. s. = sparsiflora. T. = Tonella.

Details are in the caption following the image

Parsimony mapping of pediceled- and sessile-flowered conditions based on the topology of the combined all-partitions tree in Fig. 7. Abbreviations: C. = Collinsia. T. = Tonella. If sessile-flowers are interpreted as ancestral in the clade of C. antonina + (C. concolor + C. parryi) (see Fig. 12B), then sessile flowers are reconstructed as evolving once, with two reversals to pediceled flowers, in C. antonina and C. parryi (see Discussion).

Details are in the caption following the image

Phylogram of 0.5 majority-rule consensus tree for Collinsia and Tonella based on Bayesian phylogenetic analysis of combined nrDNA and cpDNA sequence data using MrBayes (scale in expected substitutions per site). Geographic areas assigned to branches are based on results from parsimony mapping on the maximum-clade-credibility (MCC) tree, found using TreeAnnotator in the BEAST package (19). Topology of the MCC tree (not shown) is congruent with resolved structure in the consensus tree except for only slightly differing positions of two samples (C. tinctoria 6 and C. wrightii 7). For each branch of unequivocal area assignment, the estimate shown is also the predominant unequivocal estimate found throughout the posterior distribution of trees, based on results using Trace Character Over Trees in Mesquite (67). Areas indicated above biogeographically equivocal nodes indicate equally parsimonious estimates for that node (and for more apical, equivocal nodes on contiguous branches). Estimated maximal divergence times, with 95% highest probability density in parentheses, indicated for some nodes are from a nrDNA + cpDNA chronogram (using BEAST, not shown) of a subset of samples from well-supported clades resolved here. Numbers following taxon names refer to population number (see Appendix 1). Numbers in parentheses following population number refer to different cloned sequences of nuclear ribosomal DNA (all of the ITS region except for C. tinctoria sample 1 clones, of ETS). Bayesian posterior probabilities (left of slash) and maximum parsimony (MP) bootstrap values (right of slash) are given for nodes supported at ≥0.95 posterior probability. One asterisk following a slash (*) = clade with <90% support in 0.5 majority-rule MP bootstrap tree. Two asterisks (**) = clade not resolved in 0.5 majority-rule MP bootstrap tree. Clade support values in parentheses indicate (higher) support for nodes if the putatively hybrid taxon C. tinctoria was removed from the analyses. Outgroup not shown. Abbreviations: c. = clone(s). C. = Collinsia. s. = sparsiflora. T. = Tonella.

Details are in the caption following the image
Figure 12 (continued)

Phylogram of 0.5 majority-rule consensus tree for Collinsia and Tonella based on Bayesian phylogenetic analysis of combined nrDNA and cpDNA sequence data using MrBayes (scale in expected substitutions per site). Geographic areas assigned to branches are based on results from parsimony mapping on the maximum-clade-credibility (MCC) tree, found using TreeAnnotator in the BEAST package (19). Topology of the MCC tree (not shown) is congruent with resolved structure in the consensus tree except for only slightly differing positions of two samples (C. tinctoria 6 and C. wrightii 7). For each branch of unequivocal area assignment, the estimate shown is also the predominant unequivocal estimate found throughout the posterior distribution of trees, based on results using Trace Character Over Trees in Mesquite (67). Areas indicated above biogeographically equivocal nodes indicate equally parsimonious estimates for that node (and for more apical, equivocal nodes on contiguous branches). Estimated maximal divergence times, with 95% highest probability density in parentheses, indicated for some nodes are from a nrDNA + cpDNA chronogram (using BEAST, not shown) of a subset of samples from well-supported clades resolved here. Numbers following taxon names refer to population number (see Appendix 1). Numbers in parentheses following population number refer to different cloned sequences of nuclear ribosomal DNA (all of the ITS region except for C. tinctoria sample 1 clones, of ETS). Bayesian posterior probabilities (left of slash) and maximum parsimony (MP) bootstrap values (right of slash) are given for nodes supported at ≥0.95 posterior probability. One asterisk following a slash (*) = clade with <90% support in 0.5 majority-rule MP bootstrap tree. Two asterisks (**) = clade not resolved in 0.5 majority-rule MP bootstrap tree. Clade support values in parentheses indicate (higher) support for nodes if the putatively hybrid taxon C. tinctoria was removed from the analyses. Outgroup not shown. Abbreviations: c. = clone(s). C. = Collinsia. s. = sparsiflora. T. = Tonella.

Separate analyses of the expanded taxon-set for nrDNA or cpDNA data (Figs. 8, 9) yielded trees congruent with results from the smaller taxon-set analyses of each of those molecular regions (results not shown). All species of Collinsia as treated by 72 and 70 were resolved as clades except for species in each of three large- and small-flowered species pairs: C. grandiflora + C. parviflora, C. linearis + C. rattanii, and C. concolor + C. parryi. Plants referable to C. austromontana, recognized as a species by 76 and originally as a variety of C. heterophylla by 72, were nested among samples of C. heterophylla s.s. in the expanded taxon-set trees. Collinsia parryi sensu 71 was resolved as nonmonophyletic, as in the all-partitions trees (Fig. 7), with Monterey County populations (= C. antonina) in a clade outside the C. concolor + C. parryi clade. Collinsia linearis sensu 71 also was resolved as polyphyletic, as in the all-partitions trees (Fig. 7), with populations from the Sierra Nevada in a separate, distantly related lineage (= C. “metamorphica” clade).

Sequences of Collinsia tinctoria in the expanded taxon-set trees were placed in distinct clades in cpDNA and nrDNA analyses. In the cpDNA trees (Fig. 9), one sample of C. tinctoria (population 7) was resolved within C. heterophylla, and the others were resolved as a clade closely related to C. bartsiifolia and C. corymbosa (plus other taxa). Extensive polymorphism of directly sequenced PCR products of C. tinctoria ITS or ETS was reflected by placement of subsets of clones (from single plants) of C. tinctoria nrDNA in different clades in the ITS, ETS, and ITS + ETS (nrDNA) trees. The nrDNA sequences of C. tinctoria were resolved either as successively diverging along a branch connecting C. bartsiifolia + C. corymbosa with C. heterophylla (in BI trees; Fig. 8), or some sequences were resolved together in a clade with C. bartsiifolia and C. corymbosa and the rest were in a clade with C. heterophylla (in MP trees; see Fig. 8). Such polymorphism for nrDNA copy types that belong to distinct clades was not detected in any other taxon of Collinsia except in one population of C. heterophylla, from Santa Barbara County, California (see Appendix 1 and Discussion).

Character-evolution analyses and biogeographic history

Mapping of morphological characters on the all-partition MCC tree (Fig. 7) using the parsimony criterion yielded unambiguous evidence of the following evolutionary changes: (1) pediceled to sessile flowers (Fig. 11), with either two origins of sessile flowers (one represented by Collinsia concolor) or one reversal to pediceled flowers (represented by C. antonina and C. parryi), (2) glabrous to bearded stamens (at least twice; see Discussion), and (3) thickened seeds to flattened seeds (once; see Discussion). Flower size was mapped onto the same tree, with six unequivocal shifts from larger to smaller flowers that yielded the small-flowered lineages represented by Tonella tenella, C. callosa + C. childii, C. parviflora, C. rattanii, C. sparsiflora var. collina, and C. wrightii. Equivocal reconstructions involving an additional clade indicated either one shift to smaller flowers in the MRCA of C. antonina and C. parryi and an associated reversal to larger flowers in C. concolor or, instead, separate shifts to smaller flowers represented by C. antonina and C. parryi and no reversal to larger flowers. Estimates of changes in flower size in the expanded-taxon tree based on nrDNA + cpDNA data were uniformly from larger- to smaller-flowers except for reconstructions of smaller- to larger-flowers involving poorly supported clades, i.e., within C. sparsiflora and within the C. grandiflora + C. parviflora clade (see Fig. 12).

Chromosomal interchanges were mapped onto the all-partitions MCC tree with only one instance of homoplasy (14 steps). Reanalysis of the all-partitions data set using MP with weighting of ≥5 for the ordered, multistate chromosomal character resulted in a shift in the position of C. tinctoria to the branch between C. greenei and C. sparsiflora (i.e., with C. tinctoria sister to C. sparsiflora), with no homoplasy (13 steps; Fig. 13).

Details are in the caption following the image

Nonhomoplasious parsimony mapping of chromosomal interchanges (treated as an ordered character, based on chromosomal arm arrangements estimated by Garber and colleagues; see Discussion) on the tree obtained from weighting the chromosomal character at ≥5 steps in maximum parsimony analysis of the combined, three-partition molecular data set. The hatched arrow indicates the position of C. tinctoria from analyses without the chromosomal character, as in Fig. 7, or with reduced weighting of that character, which requires one extra chromosomal interchange (14 rather than 13 interchanges). A shared arm arrangement on chromosomes 3 and 4 in C. sparsiflora var. sparsiflora and C. tinctoria, otherwise undocumented in taxa of Collinsia, is homoplasious based on molecular data and may have arisen in C. tinctoria as a consequence of hybrid history involving an ancestor of C. bartsiifolia + C. corymbosa and an ancestor of C. heterophylla (see Chromosome evolution section in Discussion).

Two interfertility groups in Collinsia were found to correspond closely to two clades (Fig. 7). One was the mostly sessile-flowered clade, based on evidence of at least partially fertile progeny from crossing data involving one or more hybrid combinations linking C. bartsiifolia, C. corymbosa, C. concolor, C. heterophylla, and C. tinctoria. Collinsia parryi is part of the same clade and was included by Garber and colleagues in relatively few crossing attempts, which yielded one hybrid (a spontaneous amphidiploid) with C. concolor (44). No crossing data are available for the most recently described species of the clade, C. antonina. Estimated maximal timing of divergence between the most distantly related interfertile taxa in the mostly sessile-flowered clade was 4 Ma (95% HPD = 3–5.6 Ma) (Fig. 7). The other interfertility group in Collinsia corresponds to C. sparsiflora s.l., with divergence of taxa at ≤0.2 Ma (95% HPD = 0–0.6 Ma). Production of sterile hybrids by Garber and colleagues extended to a deeper node in the all-partitions MCC tree (Fig. 7); namely, to the MRCA of C. greenei + C. sparsiflora and the mostly sessile-flowered taxa, at ≤4.8 Ma (95% HPD = 3.8–6.6 Ma). All reported crosses between taxa separated by older nodes failed except one (see Discussion).

Parsimony mapping of biogeographic regions on the all-partitions MCC tree (Fig. 7), with terminal taxa coded by overall taxon distribution, yielded unequivocal estimates for nodal states under the broad area-coding (six-area) scheme, with the CA-FP estimated for each internal node except for the MRCAs of (1) Collinsia + Tonella (Pacific Northwest), Tonella (Pacific Northwest), and C. verna + C. violacea (Central-Eastern North America). Mapping under the finer-scale area-coding (11-area) scheme on the all-partitions tree yielded equivocal estimates of areas for most internal nodes. With coding of OTUs by collection locality in the expanded taxon-set trees (Fig. 12), the MRCA of Collinsia was estimated as having occurred in the northwestern CA-FP, regardless of whether the deep structure of the tree was constrained to conform to that of the all-partitions tree or not. Most deep nodes within Collinsia were estimated as ancestral occurrences in the northwestern CA-FP under the nrDNA + cpDNA tree topology (Fig. 12) or in either the northwestern CA-FP or the Sierra Nevada with the all-partitions deep tree-structure imposed (results not shown). Ancestral occurrence in central and eastern North America was recovered under both tree topologies for the MRCA of C. verna and C. violacea and for the MRCA of the larger clade also including C. grandiflora and C. parviflora (Fig. 12). Well-supported phylogeographic structure was recovered within T. tenella, C. heterophylla, C. latifolia, the C. linearis + C. rattanii clade, the C. “metamorphica” clade, and C. wrightii.

DISCUSSION

Monophyly of Collinsia and Tonella

Robust support for a sister-group relationship between Collinsia and Tonella from phylogenetic analyses of combined data (Fig. 7) and of separate nrDNA (Fig. 8) and CYC1 (Fig. 10) data sets reinforces the previous hypothesis that the two genera are closest relatives (e.g., 39; 72; 88; 30; 81; 3; 101). Based on the tree topology, enfolding of the stamens and pistil by conduplication of the central, lower corolla lobe evolved only once, at the base of Collinsia, and is possibly a key innovation associated with the evolution of ∼10-fold greater species diversity than has evolved in Tonella, which has a planar, central, lower corolla lobe and exposed stamens and pistil (Fig. 1). (Note, however, that a key innovation cannot be defensibly concluded from only one comparison [see 86; 49].) The keeled, central, lower corolla lobe in flowers of Collinsia (see Figs. 2–6) enforces sternotribic pollination (85) and protects pollen from theft by syrphid flies, small pollen-collecting bees, and other pollen-feeding insects (W. S. Armbruster, unpublished observations) and is part of a complex of floral characters that can affect reproductive success in Collinsia (55, 21; 3; 22; 54; 75).

40 placement of both species of Tonella within Collinsia was based on his discovery of wild plants of C. heterophylla (= C. bicolor Benth.) with aberrant flowers that resembled those of Tonella; each flower had a planar, as opposed to keeled, central, lower corolla lobe. 40, p. 55) concluded that “… since a collinsia of the bilabiate type can sportively array itself partly in tonella blossoms … there seems to be no support left for Tonella … They are perfect collinsias in all but what is now shown to be the mere accident, of a plane rather than folded lower corolla-lobe.” Spontaneous open-keeled mutants also have been observed in several species of Collinsia grown under greenhouse conditions from field-collected seed (S. Kalisz, unpublished data). Our data show that despite the occurrence of such mutant floral forms in Collinsia, evolutionary shifts from keeled to nonkeeled floral morphology have not occurred in the ingroup clade; continued recognition of Tonella is justified.

Sessile- and pediceled-flowered groups in Collinsia

The long-standing, informal, infrageneric classification of Collinsia into two groups or “sections” (39; 51; 72; 76; 70; 30; 71) distinguished by pedicel length relative to calyx length—that is, “sessile-flowered,” with pedicels absent or generally shorter than calyces, or “pediceled-flowered,” with pedicels generally longer than calyces—does not precisely reflect phylogeny, based on our findings. Pediceled flowers are unequivocally ancestral in Collinsia (and Tonella) and probably evolved secondarily from a sessile-flowered state in C. antonina and C. parryi, which are nested within the otherwise mostly sessile-flowered clade and are resolved here as the two closest relatives of the sessile-flowered C. concolor. Although parsimony mapping of sessile- and pediceled-flowered states based on the small taxon-set (all-partitions) MCC tree (Fig. 7) did not resolve unequivocally the evolution of pediceled flowers in C. antonina and C. parryi (Fig. 11), the well-supported, nested position of C. parryi in a grade of C. concolor lineages in the expanded-taxon tree (Fig. 12B) is consistent with separate origins of pediceled flowers in C. antonina and C. parryi. Under that scenario, presence of sessile-flowers in C. concolor and other sessile-flowered taxa of Collinsia would be strictly homologous.

Of the two pediceled-flowered taxa of Collinsia that are nested in the otherwise mostly sessile-flowered clade, only one, C. parryi, was included in previous biosystematic studies (44; see 30). Collinsia parryi successfully crossed only with C. concolor, its closest relative based on our molecular trees. The resulting single hybrid obtained was reported to be a spontaneous allotetraploid (44). 72, p. 287) regarded C. parryi as “… probably a segregate of the sparsiflora group” and to our knowledge C. parryi was not suggested to be a close relative of any of the sessile-flowered taxa prior to availability of molecular data (6). Lack of detected molecular divergence between some samples of the large-flowered C. concolor and short-flowered C. parryi in nrDNA, cpDNA, and CYC1 sequences indicates a striking example of rapid morphological evolution in Collinsia.

Two species of Collinsia (C. greenei and C. multicolor) have pedicels that can be longer or shorter than calyces on the same plant and therefore have been of uncertain placement. Collinsia greenei, treated by 39 and 51 in the sessile-flowered group, by 76 in the pediceled-flowered group, and by 72, 70, and 71 in both groups, was resolved as sister to C. sparsiflora in the all-partitions and CYC1 trees (Figs. 7, 10) or to C. sparsiflora and the least-inclusive clade that contains all of the undisputed sessile-flowered taxa in the expanded taxon-set nrDNA + cpDNA tree (Fig. 12B). In studies by Garber and colleagues, C. greenei yielded (sterile) hybrids only with C. sparsiflora as circumscribed here, i.e., including C. bruceae M. E. Jones and C. solitaria Kellogg (29; 1; 34; 44; see 30). Collinsia multicolor, treated by 51 and 72 (as C. franciscana Bioletti) in the pediceled-flowered group and by 76, 70, and 71 in the sessile-flowered group, was resolved in the molecular trees within the predominantly sessile-flowered clade, in keeping with high levels of fertility and chromosomal association in hybrids between C. multicolor and undisputed sessile-flowered taxa (29; 10; 31; see 30).

Evolution of other taxonomically important morphological characters

Flower size

Our phylogenetic results from multiple gene regions reinforce the evidence of 3 that much of the diversification of Collinsia and Tonella has been accompanied by shifts in flower size, with at least seven or eight shifts from larger to smaller flowers and likely at least one shift from smaller to larger flowers (represented by population 6 of C. grandiflora). (Note that in C. parviflora + C. grandiflora, shifts in flower size do not appear to correspond with shifts from diploidy to tetraploidy, as is sometimes evident [see 90]). The following large- and small-flowered sister-taxa or terminal clades are strongly supported by all lines of molecular evidence (Figs. 710, 12): C. bartsiifolia var. bartsiifolia (larger) + C. bartsiifolia var. davidsonii (smaller), C. concolor (larger) + C. parryi (smaller), C. grandiflora (larger) + C. parviflora (smaller), C. linearis (larger) + C. rattanii (smaller), C. sparsiflora var. sparsiflora (larger) + C. sparsiflora var. collina (smaller), and Tonella floribunda (larger) + T. tenella (smaller). All but cpDNA data also support a monophyletic C. torreyi s.l. (including C. wrightii), with large- and small-flowered clades. All tested members of Collinsia and Tonella are self-compatible and autonomously self-pollinate during anthesis (see 30; 55; 3); in large- and small-flowered sister-taxa pairs, self-pollination generally occurs later during floral development in large-flowered taxa (3; 80) and autonomous fruit set is generally higher in small-flowered taxa (80). Experimental findings on mating-system evolution across Collinsia based on genetic estimates of natural selfing rates and their relationship to flower size evolution will be presented elsewhere (S. Kalisz et al., unpublished data).

Flower-size and pedicel-length shifts can evidently occur exceedingly rapidly in Collinsia; cpDNA, nrDNA, and CYC1 sequences do not reliably distinguish small- and pediceled-flowered C. parryi from large- and sessile-flowered C. concolor. These two species were not considered closely related prior to the molecular analyses. Nested placement of C. parryi within a grade of C. concolor lineages leads us to conclude that the small- and pediceled-flowered C. antonina and C. parryi, recently regarded as conspecific, probably represent independent shifts to a small-flowered state, with C. concolor retaining an ancestrally large- and sessile-flowered condition rather than being an example of a small- to large-flowered shift. The expanded nrDNA + cpDNA trees (Fig. 12) indicate two lineages that warrant further investigation as possible examples of small- to large-flower size shifts, of special evolutionary interest (e.g., 94): the northern C. rattanii + C. linearis clade and C. parviflora + C. grandiflora. Considerations of chromosomal evolution (see above) in C. sparsiflora that indicate that the genomic arrangement in the small-flowered var. collina descended from the genomic arrangement unique to the large-flowered var. sparsiflora are in keeping with a large- to small-flowered shift, although further characterization of relationships in C. sparsiflora is desirable in light of (weak) phylogenetic patterns indicating a small- to large-flowered shift (Fig. 12B).

Stamen bearding

72, p. 263) noted that presence or absence of bearding on the proximal half of the upper pair of staminal filaments was useful to delimit “rather large and important groupings” in Collinsia, within the sessile- and pediceled-flowered groups. Character mapping on the all-partitions MCC tree (Fig. 7) indicates ancestrally glabrous filaments in Collinsia, with as few as two origins of filament bearding in the genus: one in the MRCA of the two central-eastern North American endemics, C. verna (Fig. 3) and C. violacea, and another in the MRCA of the clade that includes all of the undisputed sessile-flowered taxa plus C. greenei and C. sparsiflora, with a possible separate origin of bearded filaments in the ancestry of C. sparsiflora (C. greenei lacks filament bearding).

Seed shape

72 also regarded seed shape (thickened and bean-shaped as opposed to flattened and often winged or cup-shaped) as a taxonomically important character. 39 noted earlier that sessile-flowers in Collinsia are associated with flattened (“meniscoidal”) seeds. Notwithstanding seed-shape intermediacy or variation in some species, thickened seeds were mapped as ancestral in Collinsia, with a single shift to flattened seeds in the MRCA of the undisputed sessile-flowered species plus C. greenei, the C. “metamorphica” clade, and C. sparsiflora.

Evolutionary divergence and interfertility

On the basis of the all-partitions chronogram (Fig. 7), members of Collinsia with sessile flowers descended from a much more recent common ancestor (≤4 [3–5.6] Ma) than those with pediceled flowers (≤11.7 [9–12.2] Ma), in accord with biosystematic and cytogenetic findings of Garber and colleagues (see 30) and 41. In those earlier studies, Collinsia hybrids of at least low fertility were generally obtained from crosses between sessile-flowered taxa. In contrast, crosses between most pediceled-flowered taxa or between sessile- and pediceled-flowered taxa either failed or yielded sterile hybrids. The major exception to those patterns was the recovery of moderately to highly fertile hybrids from crosses in various combinations between the pediceled-flowered taxa C. sparsiflora var. arvensis, C. bruceae, and C. solitaria (1), which all have been subsequently treated within the circumscription of C. sparsiflora (e.g., 71) and are estimated here to have diverged from a common ancestor since the Pleistocene (≤0.21 [0–0.6] Ma).

On the basis of the all-partitions chronogram (Fig. 7), no pair of Collinsia taxa estimated to have diverged from a common ancestor more than 4 (3–5.6) Ma yielded fertile hybrids and only one species pair estimated to have diverged from a common ancestor more than 4.8 (3.8–6.6) Ma yielded any hybrids in studies by Garber and colleagues (see 30) or 41. The exceptionally wide, successful hybrid combination was between C. heterophylla and C. verna, with complete sterility and lack of chromosomal pairing at meiosis I reported (41).

Although divergence in chromosomal arrangements can have major effects on interfertility and most examined species of Collinsia are differentiated by chromosomal interchanges (see 30), recency of common ancestry is evidently roughly correlated with crossability and interfertility of taxa in the genus, as expected if sterility barriers have arisen largely as a byproduct of evolutionary divergence rather than as a driver of diversification (see also 79). A frequent bias toward alternate segregation at meiosis I resulted in higher than expected fertility in interspecific hybrids produced by Garber and colleagues (see 29, 15) and may limit the effect of chromosomal rearrangements on postzygotic reproductive isolation in Collinsia. 29, p. 243) nonetheless regarded such rearrangements to be “a major factor in speciation” in the genus, primarily in limiting recombination between rearranged chromosomal segments of divergent lineages (37).

Hybridization and evolution

The strongest evidence from molecular data for a hybrid constitution of any taxon in Collinsia was found in C. tinctoria (Fig. 5), a diploid sessile-flowered species generally of rocky habitats. Collinsia tinctoria has been placed in taxonomic proximity to two species of sandy soils, C. bartsiifolia and C. corymbosa (Fig. 4), in some treatments of Collinsia (72; 76, 70), based in part on their sharing of sessile flowers, flattened seeds, and a short upper corolla lip. 71 suggested morphological intergradation among those three species and C. heterophylla (Fig. 6), which includes populations with a short upper corolla lip in southwestern California (sometimes recognized as var. austromontana), well outside the geographic distribution of C. tinctoria. We are unaware of any verified natural hybrids between any of the four species, although 71, p. 1026) suggested that C. heterophylla “… may hybridize with C. bartsiifolia, C. multicolor, and C. tinctoria (more study needed)”.

Sequences of C. tinctoria were placed in distinct clades in both nrDNA and cpDNA trees, with a subset of C. tinctoria sequences in a clade with C. bartsiifolia and C. corymbosa in the cpDNA trees (Fig. 9) and in the nrDNA MP trees (see Fig. 8) and a different subset of sequences in a clade with C. heterophylla (Figs. 8, 9). In the cpDNA trees, sequences from seven of the eight sampled populations of C. tinctoria constituted a clade in an unresolved polytomy with C. bartsiifolia + C. corymbosa, C. greenei, C. multicolor, and C. sparsiflora; the eighth population of C. tinctoria yielded a cpDNA sequence that is identical to the sequence found in most sampled populations of C. heterophylla. On the basis of parsimony reconstruction of character evolution, the cpDNA sequence shared by the one C. tinctoria sample and most samples of C. heterophylla is ancestral for the C. heterophylla lineage and is widespread across modern populations of C. heterophylla; either recent chloroplast capture or ancient hybridization could explain occurrence of that chloroplast haplotype in C. tinctoria. In nrDNA trees, some sequences of C. tinctoria were resolved within a grade of lineages diverging just basal to a lineage containing all sequences of C. heterophylla; other sequences of C. tinctoria were resolved either in a clade with C. bartsiifolia and C. corymbosa (in MP trees) or as a lineage diverging along a branch connecting C. bartsiifolia and C. corymbosa with C. heterophylla and other sequences of C. tinctoria (BI trees; Fig. 8). Cloned ITS or ETS sequences from the same individual plants of C. tinctoria were placed in each of those contrasting positions. Association of C. tinctoria sequences with the same two clades (i.e., with C. bartsiifolia + C. corymbosa and with C. heterophylla) in both cpDNA and nrDNA trees and lack of comparable detected variation in other species lead us to conclude that hybridization, rather than lineage sorting or slow concerted evolution, caused the observed molecular patterns. Concerns that inclusion of a hybrid taxon might disrupt phylogenetic analyses were mostly unrealized and affected only weakly supported clades; exclusion of C. tinctoria from analyses resulted in nrDNA BI trees conforming to the nrDNA MP topology (with C. heterophylla in a clade with C. antonina, C. concolor, and C. parryi rather than with C. bartsiifolia and C. corymbosa) and resulted in an all-partition MCC tree with C. childii sister to C. “metamorphica” rather than sister to C. callosa, and with C. multicolor sister to C. bartsiifolia and C. corymbosa rather than to C. antonina, C. concolor, C. heterophylla, and C. parryi (results not shown).

Molecular evidence for a hybrid constitution of populations of C. tinctoria that are widely separated geographically—in the San Francisco Bay Area, North Coast Ranges (results not shown, for plant from Chiles Valley, Napa County; both ITS and ETS clonal diversity is extensive), and central and southern Sierra Nevada—is consistent with the hypothesis that C. tinctoria represents a species of homoploid hybrid origin rather than a recent participant in introgressive hybridization with other species. Presence of cpDNA haplotypes of each putative parental lineage across populations of C. tinctoria could reflect genetic contribution by reciprocal hybrids in the common ancestry of C. tinctoria or conceivably polyphyly of C. tinctoria, with separate hybrid speciation events based on similar parent lineages, as suggested for Helianthus anomalus (89). Lack of nrDNA mutations diagnostic for resolved sublineages of C. heterophylla within sequences of C. tinctoria indicates that the variation and polymorphism observed in C. tinctoria are best explained by hybridization that predated the divergence of sampled members of C. heterophylla from a common ancestor (≤1.4 [0.4–2.3] Ma; see Fig. 12B). Origin of C. tinctoria also apparently predated divergence of C. bartsiifolia and C. corymbosa from a common ancestor, at ≤1.4 (0.3–1.5) Ma (Fig. 7), similar to the divergence time for C. heterophylla lineages.

The likelihood of recent introgression involving C. tinctoria appears low based on Garber and colleagues’ finding that C. tinctoria was exceptionally difficult to hybridize to other sessile-flowered species of Collinsia except to the allopatric, coastal-dune endemic C. corymbosa (29, 15; 1; 10). All crosses of C. tinctoria to C. bartsiifolia failed and crosses to C. heterophylla yielded only shriveled, inviable seeds except for one that was germinated to produce a hybrid with abortive anthers and some fertile ovules (10).

Other evidence that may reflect a deep history of evolutionarily significant hybridization in Collinsia comes from conflicts in clade composition between cpDNA and nuclear trees. Although results from the partition homogeneity test indicated significant heterogeneity among cpDNA, nrDNA, and CYC1, clade support from trees based on analysis of individual regions (Figs. 810) indicate conflicts of only weak to moderate support based on (conservative) MP bootstrap values, primarily between cpDNA and nuclear trees. In particular, cpDNA trees (Fig. 9) indicate that the mostly sessile-flowered C. bartsiifolia, C. corymbosa, C. multicolor, and C. tinctoria are more closely related to C. greenei and C. sparsiflora than to C. antonina, C. concolor, C. heterophylla, and C. parryi, in contrast to the nrDNA trees (Fig. 8) and to patterns of interfertility in Collinsia (assuming gradual decay of interfertility through lineages). In addition, cpDNA trees indicate that the pediceled-flowered species C. linearis, C. rattanii, and C. torreyi s.l. constitute a clade with C. grandiflora, C. parviflora, C. verna, and C. violacea rather than a basal grade of lineages in Collinsia, as in the CYC1 trees (Fig. 10). Delayed lineage/allelic sorting or slow concerted evolution could contribute to the observed incongruities in clade composition. We also cannot rule out homoplasy as the sole cause of the deep-clade incongruities between trees based on each of the different gene regions; MP bootstrap support for one or both clades is below 90% in the cpDNA–nuclear DNA “conflicts” and below 80% in nrDNA–CYC1 “conflicts”.

Chromosome evolution

Analysis of chromosome associations at meiosis I in a wide diversity of hybrid combinations involving taxa resolved here as a clade—the mostly sessile-flowered clade plus C. greenei and C. sparsiflora—allowed Garber and colleagues to propose distinct chromosome arrangements resulting from one or more reciprocal translocations (29; 1; 10; see 30). Each taxon studied was proposed to differ from others by at least one chromosomal-arm interchange except for C. concolor and C. heterophylla s.s., which evidently shared the same arm arrangement. Recoding and mapping of the chromosome-arm arrangements as an ordered character (with number of interchanges between arrangements treated as steps) onto the all-partitions MCC tree (Fig. 7) optimized character-state changes to place the arm arrangement shared by C. concolor and C. heterophylla s.s. as unequivocally ancestral for the mostly sessile-flowered clade, with independent evolution of only one arm arrangement—for chromosomes 3 and 4—in C. sparsiflora and in C. tinctoria. Inclusion of the ordered chromosomal character in the combined-data phylogenetic analysis and assignment of a weight of ≥5 to that character resulted in placement of the C. tinctoria lineage between C. sparsiflora and C. greenei (Fig. 13), thereby eliminating the redundant interchange involving chromosomes 3 and 4; the overall genomic arrangement in C. tinctoria is one interchange removed from the C. sparsiflora var. sparsiflora arrangement and a different single interchange removed from the arrangement in C. heterophylla s.s.

The potential for convergent evolution of chromosome arrangements via hybridization may explain the presence of a chromosome arrangement in C. tinctoria that was not predicted based on the molecular trees and morphological considerations. Although C. tinctoria evidently descended from hybridization between ancestors of C. heterophylla and C. bartsiifolia + C. corymbosa (see above), the arm-arrangements of chromosomes 3 and 4 in (sessile-flowered) C. tinctoria are like those in (pediceled-flowered) C. sparsiflora var. sparsiflora, as noted above, and unlike those of any of the other sessile-flowered taxa, including C. bartsiifolia, C. corymbosa, and C. heterophylla. 16 reported the presence of “new” chromosome-arm arrangements in F3 hybrid lines involving sessile-flowered taxa, including C. corymbosa and C. tinctoria. A chromosome-arm arrangement otherwise known only from C. multicolor was recovered from a hybrid line from C. corymbosa × C. tinctoria; a different hybrid line from C. corymbosa × C. multicolor yielded plants with a chromosome-arm arrangement otherwise known only from C. heterophylla s.s. and C. concolor (see also 30). Evolution of new chromosome arrangements associated with homoploid hybrid speciation and recurrent evolution of similar arrangements in artificial hybrid lines has been studied in great detail in Helianthus (83, 69) and warrants further investigation in Collinsia.

With or without weighting of the ordered chromosomal character in molecular phylogenetic analyses, one genomic arrangement was unequivocally reconstructed as evolving directly from a documented arrangement other than the one shared by C. concolor and C. heterophylla s.s., namely, the arrangement proposed for C. sparsiflora var. collina, which evidently evolved by two interchanges from the arrangement in C. sparsiflora var. sparsiflora (Fig. 13). Assuming that the arrangements proposed for taxa by Garber and colleagues are uniform throughout those taxa and were associated with initial evolutionary divergence, the presence of a chromosome-arm arrangement in var. collina that is evolutionarily derived relative to the arm arrangement in C. sparsiflora var. sparsiflora is evidence against descent of var. sparsiflora from an ancestor referable to var. collina. In other words, the chromosomal data are consistent with a large- to small-flowered evolutionary shift in C. sparsiflora, in contrast to the small-to large-flowered pattern weakly resolved for C. sparsiflora in the expanded-taxon nrDNA + cpDNA tree (see Fig. 12B). Examples of taxa other than C. sparsiflora var. collina that are distinguished from the most chromosomally similar taxa by multiple interchanges are C. greenei (three interchanges), C. heterophylla “austromontana” (two interchanges), and C. multicolor (two interchanges) (Fig. 13).

Garber and colleagues also detected up to two paracentric inversions in hybrids between taxa of Collinsia (heterozygosity for each inversion was resolved by a dicentric bridge and acentric fragment at anaphase I), with no such inversion found to distinguish species now treated within the circumscription of C. sparsiflora (29; 1). Absence of detected inversions in hybrids between C. heterophylla and C. concolor (29; 10) is consistent with the interchange data and the close relationship resolved between those two species in the molecular trees. Overall, the inversion data are too preliminary to allow for firm phylogenetic conclusions.

Cryptic diversity and phylogeography

34 suggestion that ecologically or intrinsically isolated “cryptic species” may occur within Collinsia gains support from our results. Although sequences of most traditionally recognized species or varieties either constituted clades or were unresolved within more inclusive clades, some taxa as recently treated appear to be nonmonophyletic or to contain internal phylogenetic structure corresponding to geographic or (minor) morphological variation. The most conspicuous examples of nonmonophyly involve widely disjunct, ecogeographically distinct populations that have been misassigned to the same species.

A new Sierran clade

Central Sierra Nevada Collinsia populations in the Merced River drainage assigned previously to C. linearis, a taxon otherwise known from the Klamath Ranges of the northwestern CA-FP, constitute an ancient lineage that is only distantly related to C. linearis based on all of the molecular data (Figs. 710, 12). The Sierran Collinsia appears to be an edaphic endemic, known from isolated populations on schist exposures (thus the informal name C. “metamorphica”), and contains morphologically, ecologically, and molecularly divergent lineages that warrant taxonomic recognition (to be described elsewhere). Noise-free ITS + ETS structure within the C. “metamorphica” clade and congruent cpDNA signal indicate bluish-purple- or magenta-flowered, mostly higher-elevation lineages and a white-flowered, lower-elevation lineage. The Sierran plants, which initially appeared to be a possible example of a recently diverged peripheral isolate comparable to the famous Merced drainage species Clarkia lingulata (66; 35), instead represent an ancient lineage that has undergone diversification within the Yosemite region. Confusion of the Sierran plants with C. linearis was evidently the result of convergent evolution in floral characters and the difficulty of interpreting subtle floral variation from (pressed) herbarium specimens.

Klamath lineages of Collinsia linearis

Within C. linearis s.s., from the Klamath Ranges, two major lineages resolved with strong support from nrDNA and nrDNA + cpDNA analyses (Figs. 8 and 12) are geographically distinct and include another long-recognized taxon, C. rattanii. One lineage includes all samples of C. linearis from the Klamath River drainage, where the type was collected; the other lineage includes all samples from more northerly populations (in Oregon) in addition to all sampled populations (from California and Oregon) referable to the smaller-flowered C. rattanii, which 72 treated as conspecific with C. linearis (with recognition of C. linearis as a variety of C. rattanii). Most (but not all) of the sampled populations of C. linearis in the Klamath drainage have much paler corollas than those of populations of C. linearis sampled in the sister group. Although the large-flowered members of the two clades that constitute C. linearis appear to have remained geographically distinct since diverging ≤1.9 (0.5–3.9) Ma (see Fig. 12A), smaller-flowered populations referable to C. rattanii have evidently evolved and dispersed throughout and beyond the range of C. linearis during that time, in keeping with the general distributional patterns of closely related large- and small-flowered taxa in Collinsia (80).

A phylogeographic break in C. linearis near the northern limit of the Klamath River drainage approximates the pattern seen in some other plant taxa, such as the sword fern, Polystichum munitum (Kaulf.) C. Presl, and western white pine, Pinus monticola Douglas ex D. Don, each with northern and southern groups that break near the California–Oregon state borders (96; 92). These and other north–south phylogeographic breaks in plants involving northwestern CA-FP and the Pacific Northwest have been attributed to Pleistocene vicariance (see 92), in accord with estimated timing of the divergence in C. linearis and other evidence for the Klamath Ranges serving as a glacial refugium (see 81).

Collinsia antonina

Collinsia antonina is an example of a species that was described before our study (42) but was of uncertain taxonomic status prior to molecular phylogenetic analysis. 71 treated C. antonina as a synonym of a long-recognized species, C. parryi, which is morphologically similar and geographically disjunct. Data from ETS, cpDNA, and CYC1 sequences corroborate the conclusion of 6, based in part on the ITS results presented here, that the small-flowered, southern Californian C. parryi, previously treated within the “pediceled-flowered” group, is more closely related to the large- and “sessile-flowered,” southern Californian C. concolor than to the small- and pediceled-flowered, central California C. antonina (Figs. 710, 12B). As with C. “metamorphica,” C. antonina is evidently an edaphic endemic, known only from isolated populations on semibarren exposures (of silicious-shale talus).

Transverse Range break in Collinsia heterophylla

The widespread, large-flowered Collinsia heterophylla contains phylogenetic structure that corresponds to a previously recognized phylogeographic boundary in California. Populations sampled north of the Transverse Ranges—in the Sierra Nevada, North Coast Ranges, and South Coast Ranges—constitute a robust nrDNA and nrDNA + cpDNA lineage; populations from southwestern California, in the Transverse and Peninsular ranges, constitute a set of basally unresolved lineages within C. heterophylla (Figs. 12B). Although the Transverse Ranges appear to contain an important phylogeographic boundary in a diversity of Californian animals (see 12; 14), examples of plants showing such phylogeographic structure are lacking.

Our results for Collinsia heterophylla do not appear to conform to a finer-scale phylogeographic boundary within the Transverse Ranges between the Sierra Pelona and San Gabriel Mountains, as seen in various animal groups (14). We did not resolve the short-bannered var. austromontana, described from the Eastern Transverse Ranges (i.e., San Gabriel and San Bernardino mountains) by 72, as a monophyletic group, and one population referable to var. austromontana, from the San Bernardino Mountains (14 in Appendix 1), was identical in nrDNA and cpDNA sequences to a population referable to var. heterophylla from the western edge of the Central Transverse Ranges, at Wheeler Gorge (19 in Appendix 1), west of the Sierra Pelona. Our results instead indicate a phylogeographic break between the Transverse Ranges and South Coast Ranges. The northern population sampled closest to the Transverse Ranges, from the Sierra Madre in Santa Barbara County (population 18 in Appendix 1), was polymorphic for ITS copy types with either all diagnostic mutations of the northern clade or only a subset of those mutations (clone 6).

Estimated age of the Transverse–South Coast ranges split in C. heterophylla (≤1.2 [0.2–1.5] Ma based on nrDNA + cpDNA data; see Fig. 12B)—and in C. antonina vs. C. concolor and C. parryi (≤2.1 [1.2–3.1] Ma in the all-partitions chronogram; Fig. 7), discussed above—is in keeping with geological evidence for major Late Pliocene through Pleistocene uplift in these ranges (e.g., 74; 11; 20; see 50). Phylogeographic resolution in C. heterophylla in general does not correlate well with previously noted morphological variation, including striking variation in corolla coloration across populations.

Southern Sierra Nevada break in Collinsia wrightii

Phylogeography of the small-flowered, montane taxon Collinsia wrightii [= C. torreyi var. wrightii (S. Watson) I. M. Johnst.] (Fig. 2) indicates another pattern relevant to a Transverse Range break. Both cpDNA and nrDNA trees resolve a basal split within C. wrightii between a lineage of the Western and Eastern Transverse Ranges plus the southernmost Sierra Nevada (Tulare and Kern counties) and a lineage of more northerly montane regions (central and northern Sierra Nevada, Klamath Ranges, High North Coast Ranges, and Warner Mountains). The main topological difference between cpDNA and nrDNA trees is the position of the only sample of C. wrightii referable to C. torreyi var. brevicarinata, from the type locality in the southern High Sierra Nevada, which is weakly placed within the northerly lineage in nrDNA trees and strongly placed in the southerly lineage in cpDNA trees; combined nrDNA + cpDNA data resolve var. brevicarinata as sister to southerly members of C. wrightii, from the Western and Eastern Transverse Ranges and the southern Sierra Nevada (Fig. 12A). The results from C. wrightii (including C. torreyi var. brevicarinata) again appear to reflect a somewhat different phylogeographic pattern than the bulk of animal phylogeographic studies, wherein a major break is often estimated to lie between the Central and Eastern Transverse Ranges (14), as noted above. As for C. heterophylla, timing of the phylogeographic break in C. wrightii (≤ 1.9 [0.5–3.9] Ma; see Fig. 12) is in line with geological evidence for major uplift in the Sierra Nevada and Transverse Ranges during the Late Pliocene and Pleistocene (11; 99; see 50).

Other fine-scale phylogenetic structure

Collinsia grandiflora + C. parviflora and the small-flowered Tonella tenella also warrant further exploration for cryptic diversity and phylogeographic boundaries. Studies of the C. grandiflora + parviflora clade may help to reveal cryptic diversity associated with polyploidy (E. Elle, Simon Fraser University, personal communication), which in Collinsia is otherwise known only from C. torreyi. 26 found that populations of the C. grandiflora + C. parviflora clade in British Columbia, Canada (treated here in C. parviflora), are tetraploid, unlike previously studied populations (see 30). We found that British Columbian populations of the clade are identical in nrDNA and cpDNA sequences to one another and to populations of robust C. parviflora from the Outer North Coast Ranges of California and to a population from central-western Oregon (Lane County) with flowers as large as those of C. grandiflora but falling outside the well-supported C. grandiflora clade (Fig. 12A) and reportedly tetraploid (E. Elle, personal communication). Tonella tenella also contained phylogenetic structure, with a population from southern Oregon (Rogue River Valley) placed sister to a well-supported lineage including populations from California's North Coast Ranges and the San Francisco Bay Area (Fig. 12A), which may be comparable to phylogeographic patterns resolved for C. linearis and C. rattanii (see above).

Edaphic factors

Taxa endemic to a limited range of harsh substrates are widely dispersed across lineages of Collinsia and have evidently evolved repeatedly in California. Three such edaphic endemics are the most narrowly distributed species or terminal clades in Collinsia (C. antonina, C. corymbosa, and the C. “metamorphica” complex) and therefore might be suspected to be of relatively recent origin compared to more widespread taxa. Based on the molecular trees, however, none of these narrow endemics is nested phylogenetically within another taxon and each belongs to a stem lineage that diverged prior to closely related taxa that are not as geographically or edaphically restricted. Collinsia antonina, endemic to silicious-shale talus in the Outer South Coast Ranges of California, and C. corymbosa, endemic to coastal dunes in northern California, each have sister lineages that underwent divergence into large- and small-flowered taxa after splitting from a MRCA with the edaphic endemics: C. concolor and C. parryi (in the sister group of C. antonina) and C. bartsiifolia var. bartsiifolia and C. bartsiifolia var. davidsonii (in the sister group of C. corymbosa). The C. “metamorphica” complex, endemic to quartzite-rich schist in the Central High Sierra Nevada, appears to be older than C. antonina or C. corymbosa, both in timing of divergence from a MRCA with other taxa and in crown age (1.1 [0.3–1.3] Ma; see Fig. 12B), in keeping with resolution of morphologically and ecologically distinct lineages within that Sierran complex (unlike in C. antonina or C. corymbosa).

Collinsia greenei and C. sparsiflora are members of the principal clade in Collinsia that has invaded seasonally xeric, low-elevation habitats in California (see below) and represent different stages of serpentine adaptation. Collinsia greenei, a serpentine endemic of California's North Coast Ranges, is estimated to have diverged from a common ancestor with living relatives since the Late Miocene or Pliocene (Fig. 7), in accord with the age of serpentine exposures in the region (43) and with intersterility of C. greenei and other members of Collinsia (see 30). Collinsia sparsiflora occurs mostly on nonserpentine soils across much of the central and northern CA-FP, with genetically and phenotypically distinct serpentine and nonserpentine populations in the Knoxville region of the Inner North Coast Ranges (102, 104; 103). The presumed recent timeframe for initial, ongoing divergence between serpentine and nonserpentine ecotypes of C. sparsiflora var. sparsiflora is reflected by a lack of lineage structure across samples of these populations in our trees based on nrDNA and cpDNA mutations.

Broad-scale biogeographic and ecological history

Parsimony-based mapping of distributional history onto the all-partitions MCC tree (Fig. 7) indicates a general pattern of north-to-south and west-to-east dispersal in the ingroup clade, with the MRCA of Collinsia and Tonella in either the Pacific Northwest or northwestern CA-FP and the MRCA of Collinsia in the northwestern CA-FP. 101 inferred the Klamath Ranges of the northwestern CA-FP as a probable place of origin for the MRCA of the tribe that includes Collinsia (Cheloneae), as well, based on molecular phylogenetic and distributional considerations.

The longest dispersal event in Collinsia that led to diversification of modern species evidently occurred from the northwestern CA-FP to central-eastern North America (Fig. 12A), where the sister-species C. verna and C. violacea are endemic and diverged since the Pliocene (≤3.6 [1.6–5] Ma; Fig. 7). Dispersal to the eastern half of North America from the northwestern CA-FP was inferred by 101 for the common ancestor of the closely related Chelone as well and is consistent with evidence for long-term occurrence of mesic conditions (including some summer rain) in the Klamath Ranges of the northwestern CA-FP (81). The possibility that the sister clade of C. verna and C. violacea (namely, the C. grandiflora + C. parviflora clade) may have originated in central or eastern North America and dispersed west is consistent with the basally divergent position of central-eastern samples of C. parviflora in the expanded taxon nrDNA + cpDNA trees (Fig. 12A) and warrants further study, with additional sampling of the C. grandiflora + C. parviflora clade throughout its wide range.

An ecological shift since the Late Miocene or Pliocene (≤4.8 [3.8–6.6] Ma; Fig. 7) from montane to low-elevation, seasonally xeric habitats was mapped to the base of the clade that contains all of the sessile-flowered taxa of Collinsia plus C. antonina, C. greenei, C. parryi, and C. sparsiflora (results not shown). On the basis of parsimony mapping, the primarily low-elevation clade descended either from a MRCA in the northwestern CA-FP (Fig. 12B) or Sierra Nevada, with subsequent colonization of more xeric central-western and southwestern California. All taxa of the low-elevation clade have occurrences well below 500 m in elevation and only the evidently hybrid species C. tinctoria in that clade reaches altitudes above 2000 m a.s.l. Other members of Collinsia that occur within a similar elevational range are found in more mesic environments of the northwestern CA-FP. The estimated timing of origin of the low-elevation clade is consistent with the time frame for development of Mediterranean-climatic conditions in the CA-FP (4; 36).

Taxonomic implications

Our results indicate the need for some taxonomic changes in Collinsia to address problems in taxon circumscription, rank, and position. Minimally, members of the C. “metamorphica” clade must be treated outside C. linearis, and no names are currently available to do so (M. S. Park, B. G. Baldwin, and W. S. Armbruster, unpublished manuscript). Collinsia linearis may be best treated in an even narrower sense to exclude populations north of the Klamath River drainage that are evidently more closely related to C. rattanii than to typical C. linearis.

Collinsia sparsiflora, as broadly circumscribed here, has been treated variously as a complex of several species or varieties that vary in flower size, flower shape, and inflorescence architecture, and additional variation has been found to be associated with recent evolution of serpentine ecotypes (102, 104). Tentatively, we retain recognition of var. collina and var. sparsiflora for small- and large-flowered populations that pass the test of sympatry and justifiably could be regarded as separate species, although phylogenetic resolution is lacking. Difficulty in resolving relationships within C. sparsiflora leaves open the question of whether floral characters associated with each of the two taxa have evolved independently in different lineages within the complex.

Our sampling to date of the C. grandiflora + C. parviflora clade suggests problems with the current taxonomy that may reflect lability in evolution of flower size, which has been a key character for distinguishing the two taxa. More sampling across the exceptionally broad geographic distribution of the C. grandiflora + C. parviflora clade, including ongoing studies of ploidy (E. Elle, Simon Fraser University, personal communication), should help to refine classification of taxa in the group.

In C. torreyi s.l., three taxa usually treated as varieties (var. latifolia, var. torreyi, and var. wrightii) represent lineages that evidently diverged prior to diversification of the largely sessile-flowered clade in Collinsia (Fig. 7). Sympatry between C. torreyi var. latifolia and C. wrightii (= C. torreyi var. wrightii) and between C. torreyi s.s. (= C. torreyi var. torreyi) and C. wrightii is widespread, and hybrids are unknown. Recognition of C. torreyi var. latifolia, C. torreyi s.s., and C. wrightii (now including C. torreyi var. brevicarinata) as different species is therefore warranted, and a new combination at species rank is needed for var. latifolia:

Collinsia latifolia (Newsom) B. G. Baldwin, Kalisz & Armbr., comb. et stat. nov.—Collinsia torreyi A. Gray var. latifolia Newsom, Botanical Gazette 87: 299. 1929. TYPE: USA, Oregon, Ashland Butte, Cusick 2893 (holotype: POM (no. 42879)!; isotype: GH).

Conclusions

Diversity in Collinsia is greater than previously documented and includes narrowly endemic lineages that appear vulnerable to habitat loss and warrant conservation attention. Recently diverged lineages in Collinsia are often associated with distinct habitats, including harsh substrates (e.g., serpentine), and/or geographic boundaries that have been implicated in evolutionary divergence in other organisms as consequences of Plio-Pleistocene geoclimatic factors, such as mountain building and glaciation. Repeated shifts in flower size, mostly in the direction from larger to smaller flowers, have also characterized evolution of Collinsia (and Tonella), as has chromosome evolution by structural rearrangements. Rapid evolution of strong intersterility barriers evidently has not been a dominant theme in Collinsia evolution and has left open the potential for homoploid hybrid speciation, which appears to have occurred in the evolution of C. tinctoria.

Appendix 1

Voucher information and GenBank accession numbers for Collinsia, Tonella, and outgroup taxa sampled in this study. Californian collections are deposited at JEPS unless otherwise specified; non-Californian collections are at UC. Voucher information is presented in alphabetical order by taxon and, within taxa, by state and county occurrences. Sequential numbering of voucher entries corresponds to numbers following taxon names in figures. GenBank accession numbers for DNA sequences are given after voucher information in the following order: ITS (internal transcribed spacer region=ITS-1, 5.8S subunit, ITS-2), ETS (3′ end of external transcribed spacer upstream of 18S gene), 3′ matK/3′ trnK intron, and CYC1 (CYCLOIDEA-1). Abbreviations: AMR = April M. Randle, BGB = Bruce G. Baldwin, Co. = County, CR = Christian Richey, ECN = Elizabeth C. Neese, MSP = Michael S. Park, SK = Susan Kalisz, WSA = W. Scott Armbruster.

Taxon: Collection location (voucher entry), voucher, GenBank accessions: ITS, ETS, 3′ matK/3′ trnK intron, CYC1.

Collinsia antonina Hardham: USA, California, Monterey Co., Outer South Coast Ranges, San Antonio Hills (1) ECN 21500, HQ653130, HQ653326, HQ653529, —, (2) Fort Hunter Liggett Military Reservation, E side of County Road G14, ca. 0.2 km N of southern Reservation boundary, MSP, WSA, BGB, & CR 1063, HQ653131, HQ653327, HQ653530, HQ653394, (3) Fort Hunter Liggett, Sulphur Springs Road, 3.9 road km N from Infantry Road junction, MSP & BGB 930, HQ653132, HQ653328, HQ653531, —.

C. bartsiifolia Benth. var. bartsiifolia: USA, California, (1) Lake Co., Inner North Coast Ranges, Clear Lake vicinity, S of Lakeport, BGB & WSA 858, AF385344, HQ653267, HQ653479, —; Santa Cruz Co., Outer South Coast Ranges, Santa Cruz Mountains, Sandhills, Ben Lomond vicinity, (2) Quail Hollow Ranch, J. McGraw s.n., HQ653072, HQ653271, HQ653483, HQ653384, (3) Olympia Wellfield, J. McGraw s.n., HQ653073, HQ653272, HQ653484, —.

C. bartsiifolia Benth. var. davidsonii (Parish) Newsom: USA, California, (1) Madera Co., = C. stricta Greene, Central Sierra Nevada Foothills, Wahlberg Ranch, B. Brock 401 (UC), HQ653074, HQ653273, HQ653485, —; Monterey Co., Outer South Coast Ranges, (2) Fort Hunter Liggett, 3.2 km W of Cosio Knob, E. Painter & ECN HL2546, HQ653069, HQ653268, HQ653480, —, (3) Fort Hunter Liggett, Mission Road, Training Area 6, MSP 1212, HQ653070, HQ653269, HQ653481, —; (4) San Benito Co., Inner South Coast Ranges, Coalinga Road, BGB, WSA, & MSP 1378, HQ653071, HQ653270, HQ653482, —; (5) Santa Barbara Co., Outer South Coast Ranges, below Santa Barbara Canyon, BGB & S. J. Bainbridge 1033, HQ653075, HQ653274, HQ653486, —.

C. callosa Parish: USA, California, (1) Kern Co., Western Transverse Ranges, NW side of Cerro Noroeste, BGB & WSA 947, AF385354, HQ653250, HQ653462, HQ653375; (2) High Southern Sierra Nevada, Piute Mountain Road, ∼4.8 km from Bodfish Road, D. Gowen 356, HQ653055, HQ653252, HQ653464, —; (3) San Bernardino Co., San Bernardino Mountains, near Hesperia, A. Lankinen 03-01, HQ653054, HQ653251, HQ653463, HQ653374.

C. childii Parry ex A. Gray: USA, California, (1) Fresno Co., Southern High Sierra Nevada, E edge of Pinehurst, ECN 21501, AF385355, HQ653253, HQ653465, —; (2) Mariposa Co., Central High Sierra Nevada, Iron Creek, MSP 1608, HQ653057, HQ653255, HQ653467, —; (3) Monterey Co., Outer South Coast Ranges, Santa Lucia Range, Chew's Ridge, BGB & WSA 1201a, HQ653060, HQ653258, HQ653470, —; (4) San Bernardino Co., San Bernardino Mountains, along State Route 138, 0.3 road km S of Old Mill Road (near town of Crestline), MSP 1840, HQ653059, HQ653257, HQ653469, —; (5) San Diego Co., Peninsular Ranges, Palomar Mountain, Fry Creek Campground, MSP 1825, HQ653058, HQ653256, HQ653468, —; (6) Santa Barbara Co., Outer South Coast Ranges, Sierra Madre Ridge, Bates Canyon, BGB, WSA, & SK 1375, HQ653056, HQ653254, HQ653466, HQ653377–HQ653379 (CYC1 clones 1, 2, 4).

C. concolor Greene: USA, California, Riverside Co., Peninsular Ranges (1) W edge San Jacinto Mountains, S-SE of Hemet (Sage Road/Red Mountain Road intersection), ECN 21539A, individual 1, AF385350, HQ653331, HQ653534, —, individual 2, HQ653137, HQ653335, HQ653538, HQ653397 (CYC1 clones 1, 4, 5), (2) San Jacinto Mountains, near Valle Vista, A. Lankinen 03-03, HQ653134, HQ653332, HQ653535, HQ653396, (3) Little Thomas Mountain, MSP 2030, HQ653138, HQ653336, HQ653539, —; San Diego Co., Peninsular Ranges, (4) S of Mesa Grande, Black Canyon, MSP 1817, HQ653135, HQ653333, HQ653536, —, (5) Walker Canyon Ecological Preserve (near Boulevard), WSA s.n. (19 Apr 2009), HQ653136, HQ653334, HQ653537, —.

C. corymbosa Herder: USA, California, Mendocino Co., North Coast, (1) S end of Ten Mile Dunes, ECN 21140, AF385345, HQ653275, HQ653487, —, (2) near Virgin Creek, AMR & K. M. Hanley MD-A, HQ653076, HQ653276, HQ653488, —, (3) north of Mill Creek, AMR & K. M. Hanley MD-B, HQ653077, HQ653277, HQ653489, —, BGB & WSA 859, HQ653078, HQ653278, HQ653490, HQ653385.

C. grandiflora Douglas ex Lindl: USA, California, (1) Humboldt Co., Outer North Coast Ranges, NW side of Horse Mountain, ECN 21359, AF385341, HQ653238, HQ653450, —; Trinity Co., Klamath Ranges, (2) Canyon Creek (5.1 road km N of Junction City), WSA & CR 06-101, HQ653045, HQ653239, HQ653451, —, CR s.n. (24 May 2007), HQ653049, HQ653243, HQ653455, —; Oregon, Jackson Co., Rogue River Valley, (3) Upper Table Rock, BGB & WSA 1191, HQ653046, HQ653240, HQ653452, —, (4) Lower Table Rock, BGB & WSA 1194, HQ653048, HQ653242, HQ653454, HQ653371, (5) Crowfoot Road, AMR s.n. (1 Jun 2005), HQ653047, HQ653241, HQ653453, —; (6) Lane Co., Trout Creek, E. Elle s.n. (Jun 2004), HQ653036, HQ653229, HQ653441, —.

C. greenei A. Gray ex Greene: USA, California, (1) Colusa Co./Lake Co. border, High North Coast Ranges, Snow Mountain vicinity (0.3 trail km N-NE of Summit Spring), BGB 952, HQ653102, HQ653295, HQ653498, —; Lake Co., Inner North Coast Ranges, (2) NE of Middletown (along State Highway 29, 3.7 road km NE of Butts Canyon Road junction), BGB & WSA 851, AF385338, HQ653296, HQ653499, —, (3) Highland Springs Road, AMR, K. M. Hanley, & BGB HSR-A, HQ653104, HQ653298, HQ653501, HQ653387–HQ653388 (CYC1 clones 1, 2, 6, 10); (4) Napa Co. Inner North Coast Ranges, State Highway 128 junction with Berryessa-Knoxville Road (Turtle Rock), CR s.n. (May 2005), HQ653103, HQ653297, HQ653500, —.

C. heterophylla Buist ex Graham: USA, California, (1) Alameda Co., San Francisco Bay Area/San Joaquin Valley border, Tesla Road (1 road km E of divide), BGB & WSA 856, HQ653141, HQ653340, HQ653543, —; (2) El Dorado Co., Northern Sierra Nevada Foothills, South Fork American River at Chili Bar (N of Placerville, at State Highway 193 crossing), CR s.n. (May 2005), individual 1, HQ653152, HQ653351, HQ653554, —, individual 2, HQ653153, HQ653352, HQ653555, —; (3) Los Angeles Co., Western Transverse Ranges, Warm Springs Canyon, MSP 1605, HQ653151, HQ653350, HQ653553, —; Mariposa Co., Central Sierra Nevada Foothills, (4) W. of Mt. Bullion, BGB & S. J. Bainbridge 828, AF385337, HQ653337, HQ653540, —, (5) South Fork Merced River at Hite's Cove, BGB, SK, & T. J. Rosatti 1009, HQ653139, HQ653338, HQ653541, —; Monterey Co., (6) Inner South Coast Ranges, Gabilan Range, 3.6 road km W of San Benito Co. line along Gloria Road, BGB & S. J. Bainbridge 1273b, HQ653143, HQ653342, HQ653545, —, (7) Outer South Coast Ranges, ca. 6.5 road km E of summit of Lockwood-San Ardo Road, BGB & WSA 1377, HQ653146, HQ653345, HQ653548, —, (8) Outer South Coast Ranges, Upper Carmel River watershed, Hastings Natural History Reservation, AMR, K. M. Hanley, & MSP HBR-A, HQ653157, HQ653356, HQ653559, HQ653406–HQ653407 (CYC1 clones 1, 2, 4), (9) Outer South Coast Ranges, Santa Lucia Range, ca. 1.5 km NW of Basket Spring (ca. 2.8 km S of San Miguel Creek), ECN, E. Painter, BGB, M. Wetherwax, T. Morosco, K. Downing, & H. Forbes HL1471, HQ653158, HQ653357, HQ653560, —; Napa Co., Inner North Coast Ranges, (10) State Highway 121, ca. 3.2 road km N of Wooden Valley Road junction, BGB & WSA 1204a, HQ653142, HQ653341, HQ653544, —, (11) Pope Valley vicinity, upper Wantrup Preserve, WSA & CR s.n. (21 Apr 2005), HQ653144, HQ653343, HQ653546, —; (12) Riverside Co., Peninsular Ranges, Santa Ana Mountains, Bear Canyon Trail, MSP 2029, HQ653147, HQ653346, HQ653549, —; San Bernardino Co., (13) San Gabriel Mountains/San Bernardino Mountains border, near Cajon Pass, MSP 1569 (= var. austromontana Newsom), HQ653148, HQ653347, HQ653550, —, (14) San Bernardino Mountains, town of Cedarpines Park, MSP s.n. [no voucher; same population as A. C. Sanders 14938 (UCR)] (= var. austromontana Newsom), HQ653150, HQ653349, HQ653552, —; San Diego Co., Peninsular Ranges, (15) W of Ramona on State Route 78 at Indian Oaks Road, MSP & C. M. Waterman 1797, HQ653155, HQ653354, HQ653557, —, (16) S of Mesa Grande, Black Canyon, MSP 1808, HQ653156, HQ653355, HQ653558, —; (17) San Luis Obispo Co., Outer South Coast Ranges, Santa Lucia Range, Upper Lopez Canyon Road, 3.9 road km from junction with Hi Mountain Road, BGB & S. J. Bainbridge 894, HQ653140, HQ653339, HQ653542, —; (18) Santa Barbara Co., Outer South Coast Ranges, Sierra Madre, Bates Canyon, BGB, SK & WSA 1373, HQ653159–HQ653161 (ITS clones 2, 6, 11, 12), HQ653358, HQ653561, —; (19) Santa Clara Co., San Francisco Bay Area, Mt. Hamilton Range, Kammerer Ranch, S. J. Bainbridge s.n. (17 May 2007), HQ653154, HQ653353, HQ653556, —; Ventura Co., Western Transverse Ranges, (20) Wheeler Gorge, WSA WG-A, HQ653149, HQ653348, HQ653551, HQ653404–HQ653405 (CYC1 clones 1–3), (21) Sisar Canyon, WSA s.n., HQ653145, HQ653344, HQ653547, HQ653398–HQ653403 (CYC1 clones A1–A3, D1–D4, D3P1).

C. latifolia (Newsom) B. G. Baldwin, Kalisz & Armbr.: USA, California, (1) Mendocino Co., High North Coast Ranges, US Forest Service Route 23N60, 0.6 road km N of US Forest Service Route 23N69 junction, BGB 899, HQ652998, HQ653189, HQ653430, —; (2) Shasta Co., High Cascade Range, Noble Pass, G. Gillett 618, HQ653001, HQ653192, HQ653433, —; Siskiyou Co., Klamath Ranges, (3) N Fork Salmon River drainage, Taylor Lake, CR s.n. (9 Jun 2007), HQ652999, HQ653190, HQ653431, —, (4) Marble Mountains, Boulder Creek Trail, CR s.n. (20 Jun 2007), HQ653000, HQ653191, HQ653432, —.

C. linearis A. Gray: USA, California, Humboldt Co., Outer North Coast Ranges, (1) Bair Road, halfway between Hoopa and Pine Ridge summit, CR s.n. (3–4 Jun 2007), HQ653017, HQ653208, HQ653583, —, (2) Bair Road, 1.8 road km NE of Pine Ridge Road junction, CR s.n. (3–4 Jun 2007), HQ653023, HQ653215, HQ653590, —, (3) Bair Road, 1.3 road km NE of Pine Ridge Road junction, CR s.n. (3–4 Jun 2007) HQ653019, HQ653210, HQ653585, —, (4) Bair Road, 24.1 road km NE of State Highway 299 junction, CR s.n. (3–4 Jun 2007), HQ653018, HQ653209, HQ653584, —, (5) W side of Horse Mountain, ECN 21358, AF385343, HQ653211, HQ653586, —; (6) Titlow Hill Road, MSP s.n., individual 1, HQ653021, HQ653213, HQ653588, —, individual 2, HQ653022, HQ653214, HQ653589, —; Siskiyou Co., Klamath Ranges, (7) South Fork Salmon River, along Cecilville Road, 689 m elev., CR s.n. (8 Jun 2007), HQ653024, HQ653216, HQ653591, —, (8) South Fork Salmon River, along Cecilville Road, 1218 m elev., CR s.n. (8 Jun 2007), HQ653026, HQ653218, HQ653593, —, (9) US Forest Service road to Taylor Lake, 0.5 road km S-SE from Sawyers Bar Road junction, CR s.n. (9 Jun 2007), HQ653025, HQ653217, HQ653592, —, (10) Marble Mountains, along Boulder Creek Trail, 1939 m elev., CR s.n. (20 Jun 2007), HQ653027, HQ653219, HQ653594, —; Trinity Co., Klamath Ranges, (11) near Burnt Ranch, along State Highway 299, WSA & CR 06-102 HQ653010, HQ653201, HQ653576, —, (12) Junction City, CR s.n. (24 May 2007), HQ653020, HQ653212, HQ653587, —; Oregon, Jackson Co., (13) Rogue River Valley, Upper Table Rock, BGB & WSA 1192, HQ653029, HQ653221, HQ653596, —, (14) Rogue River Valley, Lower Table Rock, BGB & WSA 1195, HQ653028, HQ653220, HQ653595, HQ653369, (15) Rogue River Valley, Crowfoot Road, AMR s.n. (1 Jun 2005), HQ653011, HQ653202, HQ653577, —, (16) Rogue River Valley, Butte Falls Road, AMR s.n. (3 Jun 2005), HQ653012, HQ653203, HQ653578, —, (17) Applegate Valley, Trillium Mountain, AMR s.n. (2006), HQ653013, HQ653204, HQ653579, —, (18) Applegate Valley, Applegate 500 (Quartz fire burn), AMR s.n. (29 Jun 2005), HQ653014, HQ653205, HQ653580, —, (19) Applegate Valley, Little Applegate Road, AMR s.n. (2006), HQ653016, HQ653207, HQ653582, —, (20) Greensprings area, Grizzly Peak Road, AMR GPR06, HQ653015, HQ653206, HQ653581, —.

C. “metamorphica” complex: USA, California, Mariposa Co., Central High Sierra Nevada, Merced River drainage, (1, 2) Trumbull Peak, MSP 1614, HQ653061, HQ653259, HQ653471, —, MSP 1897, HQ653062, HQ653260, HQ653472, —, (3) south end of Iron Mountain, BGB & WSA 1483, HQ653063, HQ653261, HQ653473, HQ653381–HQ653383 (CYC1 clones 2, 3, 5, 6), (4) South Fork Merced River, MSP 1899, HQ653064, HQ653262, HQ653474, —, (5) Zip Creek, MSP 2147, HQ653068, HQ653266, HQ653478, —; (6) north end of Iron Mountain, MSP 1905, HQ653067, HQ653265, HQ653477, —, (7) Merced River near El Portal (Foresta Road), BGB & WSA 1484, HQ653065, HQ653263, HQ653475, HQ653380, (8) Merced River at Pigeon Gulch, MSP 1871, HQ653066, HQ653264, HQ653476, —.

C. multicolor Lindl. & Paxton: USA, California, (1) San Mateo Co., San Francisco Bay Area, Crystal Springs Road, MSP 1195, HQ653129, HQ653324, HQ653527, —; (2) Santa Cruz Co., San Francisco Bay Area, Swanton Road, ECN 21122, AF385353, HQ653325, HQ653528, HQ653393.

C. parryi A. Gray: San Bernardino Co., (1) W end of San Bernardino Mountains, ECN 21530, AF385349, HQ653329, HQ653532, HQ653395, (2) San Gabriel Mountains, Lone Pine Canyon, A. Lankinen 03-02, HQ653133, HQ653330, HQ653533, —.

C. parviflora Douglas ex Lindl: CANADA, British Columbia, (1) Tyhee Lake, Smithers, WSA s.n., HQ653032, HQ653225, HQ653437, —, (2) Vancouver Island, Kin Beach, E. Elle s.n., HQ653033, HQ653226, HQ653438, —; (3) Yukon Territory, hill across Tagish Road from Crag Lake, B. A. Bennett s.n., HQ653038, HQ653231, HQ653443, —; USA, California, Alameda Co., San Francisco Bay Area, Mt. Hamilton Range, (4) Tarraville Creek, BGB & WSA 854, AF385340, HQ653222, HQ653434, —, (5) head of Williams Gulch, BGB, S. J. Bainbridge, & W. Legard 1479, HQ653030, HQ653223, HQ653435, —; Humboldt Co., Outer North Coast Ranges, (6) Titlow Hill Road, 4.3 road km NW of Friday Ridge Road junction, CR s.n. (3–4 Jun 2007), HQ653034, HQ653227, HQ653439, —, (7) near Grouse Mountain, along Friday Ridge Road, 3.5 road km S of Titlow Hill Road junction, CR s.n. (3–4 Jun 2007), HQ653035, HQ653228, HQ653440, —; (8) Siskiyou Co., Klamath Ranges, E side of Taylor Lake, CR s.n. (9 Jun 2007), HQ653039, HQ653232, HQ653444, —; (9) Idaho, Adams Co., Little Salmon River, 20.1 km N of New Meadows along US 95, B. Ertter 19398, HQ653042, HQ653235, HQ653447, —; (10) Michigan, Upper Peninsula, Marquette Co., Mountain Lake (near Big Bay), C. Heckel s.n., individual 1, HQ653043, HQ653236, HQ653448, —, individual 2, HQ653044, HQ653237, HQ653449, —; (11) Nevada, Washoe Co., Modoc Plateau, State Highway 8A, 8 road km W of State Highway 34 junction, BGB, M. J. Sanderson, & M. J. Wojciechowski 866, HQ653031, HQ653224, HQ653436, —; Oregon, Jackson Co., (12) Klamath Ranges, Pilot Rock, near Ashland, AMR & K. M. Hanley s.n. (19 May 2004), HQ653037, HQ653230, HQ653442, HQ653370, (13) Rogue River Valley, Lower Table Rock, summit, BGB & WSA 1196, HQ653029, HQ653234, HQ653446, —, (14) Rogue River Valley, Lower Table Rock, ca. 0.4–0.8 trail km below summit, along main trail, BGB & WSA 1197, HQ653028, HQ653220, HQ653445, —.

C. rattanii A. Gray: USA, California, Glenn Co., High North Coast Ranges, (1) N of Snow Mountain, along US Forest Service Road M3, 1.1 road km S of Ivory Mill Road junction, CR s.n. (25 May 2007), HQ653007, HQ653198, HQ653573, —, (2) US Forest Service Road FH-7, 8.4 road km S of Plaskett Meadows Campground, CR s.n. (25 May 2007), HQ653008, HQ653199, HQ653574, —; (3) Lake Co., Inner North Coast Ranges, Harrington Flat Road, just S of Boggs Lake, BGB & WSA 849, AF385342, HQ653193, HQ653567, HQ653367; Plumas Co., Northern High Sierra Nevada, (4) M. Wetherwax s.n. (22 May 1995), HQ653002, HQ653194, HQ653568, —, (5) State Highway 70, 0.5 road km W of Squirrel Creek (11.3 air km SE of Quincy), L. Ahart 10956, HQ653003, —, HQ653569, —; (6) Siskiyou Co., Klamath Ranges, Marble Mountains, along Boulder Creek Trail, CN s.n. (20 Jun 2007), HQ653009, HQ653200, HQ653575, —; Oregon, Jackson Co., (7) Applegate Valley, Trillium Mountain, AMR s.n. (2006), HQ653004, HQ653195, HQ653570, —, (8) Applegate Valley, Applegate 500 (Quartz fire burn), AMR s.n. (2006), HQ653005, HQ653196, HQ653571, HQ653368, (9) Greensprings area, Keen Creek Road, AMR s.n. (2006), HQ653006, HQ653197, HQ653572, —.

C. sparsiflora Fisch. & C. A. Mey. var. collina (Jeps.) Newsom: USA, California, (1) Alameda Co., San Francisco Bay Area, Mt. Hamilton Range, Tarraville Creek, BGB & WSA 855, AF385339, HQ653299, HQ653502, HQ653389; Napa Co., Inner North Coast Ranges, (2) Pope Valley vicinity, Wantrup Preserve, CR s.n., HQ653119, HQ653314, HQ653517, —, (3) Pope Valley, Ink Grade / Pope Valley Road junction, WSA s.n., HQ653114, HQ653309, HQ653512, —; (4) Nevada Co., Northern Sierra Nevada Foothills, Lime Kiln Road, W of State Route 49, MSP 1769, HQ653127, HQ653322, HQ653525, —; (5) San Benito Co., Inner South Coast Ranges, confluence of Clear Creek and San Benito River, MSP 2125, HQ653128, HQ653323, HQ653526, —.

C. sparsiflora Fisch. & C. A. Mey. var. sparsiflora: USA, California, (1) Contra Costa Co., San Francisco Bay Area, Marsh Creek headwaters, Morgan Territory, MSP 1035, HQ653105, HQ653300, HQ653503, —; Lake Co., Inner North Coast Ranges, (2) Boggs Lake, BGB, WSA, & SK 1067, HQ653106, HQ653301, HQ653504, —, (3) State Route 29, ca. 5 road km NE of Middletown, BGB & WSA 1190, HQ653107, HQ653302, HQ653505, —; Napa Co., Inner North Coast Ranges, (4) 3400 Wooden Valley Road (McQueeney's Ranch entrance), BGB, WSA, & SK 1202b, HQ653108, HQ653303, HQ653506, HQ653390, (5, 6) State Route 121, ca. 3 road km N of Wooden Valley Road junction, BGB, WSA, & SK 1203b, large-flowered, HQ653109, HQ653304, HQ653507, HQ653391, medium-flowered, HQ653115, HQ653310, HQ653513, HQ653392, (7) “Mercury Meadow” (serpentine at Knoxville), J. Wright s.n., HQ653124, HQ653319, HQ653522, —, (8) “Mercury Woods” (non-serpentine at Knoxville), J. Wright s.n., HQ653126, HQ653321, HQ653524, —, (9) “Saddle Ridge” (serpentine near Knoxville), J. Wright s.n., HQ653120, HQ653315, HQ653518, —, (10) Pope Valley vicinity, Wantrup Preserve, WSA & CR s.n. (21 Apr 2005), HQ653110, HQ653305, HQ653508, —, (11–16) Knoxville vicinity, McLaughlin Natural Reserve, Shannon Peters ML16 (30 Apr 2006), (11) on serpentine soil, HQ653121, HQ653316, HQ653519, —, (12) on non-serpentine soil, HQ653122, HQ653317, HQ653520, —, (13) on serpentine soil, HQ653125, HQ653320, HQ653523, —, (14) on serpentine soil, HQ653116, HQ653311, HQ653514, —, (15) on non-serpentine soil, HQ653123, HQ653318, HQ653521, —, (16) on non-serpentine soil, HQ653111, HQ653306, HQ653509, —, (17) Outer North Coast Ranges, Bear Canyon, CR s.n. (18 May 2005), HQ653118, HQ653313, HQ653516, —; (18) Yolo Co., “Gold Pit” (non-serpentine ca. 1.6 km NW of Knoxville), J. Wright s.n., HQ653112, HQ653307, HQ653510, —; Oregon, Jackson Co., Rogue River Valley, Lower Table Rock, (19) summit (white flowers), BGB & WSA 1193, HQ653117, HQ653312, HQ653515, —, (20) base (violet flowers), BGB & WSA 1199, HQ653113, HQ653308, HQ653511, —.

C. tinctoria Hartw. ex Benth: USA, California, (1) Contra Costa Co., San Francisco Bay Area, Mt. Diablo, BGB & WSA 951, HQ653082, HQ653283–HQ653287 (ETS clones 1–12), HQ653495, HQ653386; (2) El Dorado Co., Northern Sierra Nevada Foothills, Placerville, Mosquito Road, CR s.n. (May 2005), HQ653079, HQ653280, HQ653492, —; (3) Fresno Co., Southern High Sierra Nevada, Hume Lake Road, MSP 1326, HQ653080, HQ653281, HQ653493, —; (4) Mariposa Co., Central High Sierra Nevada, Bull Creek Road, WSA & CR 06-103, HQ653089–HQ653093 (ITS clones 1–12), HQ653289, HQ653497, —; (5) Napa Co., Inner North Coast Ranges, Chiles Valley Road, CR s.n. (May 2005), HQ653094–HQ653101 (ITS clones 1–12), HQ653290–HQ653294 (ETS clones 1–12), HQ675014, —; (6) Sonoma Co., Outer North Coast Ranges, Bohemian Highway, 5 road km N of Graton Road junction (N of Camp Meeker), BGB & WSA 857, AF385346, HQ653279, HQ653491, —; Tulare Co., Southern High Sierra Nevada, (7) Greenhorn Mountains, South Creek (near Johnsondale), MSP 1930, HQ653081, HQ653282, HQ653494, —, (8) Kaweah River drainage, BGB 1133, HQ653083–HQ653088 (ITS clones 1–12), HQ653288, HQ653496, —.

C. torreyi A. Gray: USA, California, (1) El Dorado Co., Northern High Sierra Nevada, Wright's Lake vicinity, MSP 767, individual 1, HQ652982, HQ653171, HQ653417, —, individual 2, HQ652985, HQ653175, HQ653421, —; (2) Mariposa Co., Central High Sierra Nevada, Merced River drainage, N end of Iron Mountain, MSP 1907, HQ652983, HQ653172, HQ653418, —; (3, 4) Tuolumne Co., State Route 108, E of Pinecrest, ECN 21345, HQ652984, HQ653174, HQ653420, HQ653365, ECN 21356, AF385347, HQ653173, HQ653419, —.

C. verna Nutt: USA, Illinois, (1) Will Co., Raccoon Grove Forest Preserve (near Monee), SK s.n., AF385351, HQ653244, HQ653456, —; (2) Michigan, Kalamazoo Co., Tu Avenue, SK s.n., HQ653051, HQ653246, HQ653458, —; (3) Pennsylvania, Greene/Washington county line, Enlow Fork, SK s.n., HQ653050, HQ653245, HQ653457, HQ653372.

C. violacea Nutt: USA, (1) Illinois, Shelby Co., Harmon Cemetery, off Highway 128, C. Ivey s.n. (20 May 2005), HQ653053, HQ653249, HQ653461, —; Missouri, (2) Gasconade Co., ca. 1.6 km S of Hermann, J. Paul s.n. (2004), HQ653052, HQ653248, HQ653460, HQ653373; (3) St. Clair Co., ca. 10 km W of Osceola and Highway 13, ECN 21533B, AF385352, HQ653247, HQ653459, —.

C. wrightii S. Watson: USA, California, (1) El Dorado Co., Northern High Sierra Nevada, Wright's Lake vicinity, MSP 775, individual 1, HQ652986, HQ653176, HQ653422, —, individual 2, HQ652987, HQ653177, HQ653423, —; (2) Glenn Co., High North Coast Ranges, saddle just N of Black Butte, CR s.n., HQ652989, HQ653179, HQ653425, —; Kern Co., (3) Western Transverse Ranges, Mt. Pinos Road, ca. 0.8–1.6 road km above Mil Potrero Highway junction, BGB & WSA 948, AF385348, HQ653183, HQ653562, —; (4) High Southern Sierra Nevada, Greenhorn Range, W slope of Sunday Peak, E. Twisselmann 5246, HQ652994, HQ653185, HQ653564, —; (5) Los Angeles Co., San Gabriel Mountains, Table Mountain, MSP 2039, HQ652993, HQ653184, HQ653563, —; (6) Mendocino Co., High North Coast Ranges, S of Anthony Peak, Wells Cabin Campground, MSP 1371b, HQ652997, HQ653188, HQ653429, —; (7) Modoc Co., Warner Mountains, MSP 1945, HQ652991, HQ653181, HQ653427, —; (8) Mono Co., High Central Sierra Nevada, Silver Creek Meadows, D. Taylor 17519, HQ652992, HQ653182, HQ653428, —; (9) Siskiyou Co., Klamath Ranges, N Fork Salmon River drainage, Taylor Lake, CR s.n. (9 Jun 2007), HQ652990, HQ653180, HQ653426, —; (10) Tulare Co., High Southern Sierra Nevada, Kern Plateau, Deadwood Meadow, E. Twisselmann 17923, HQ652995, HQ653186, HQ653565, —; (11) = C. torreyi A. Gray var. brevicarinata Newsom, Southern High Sierra Nevada, Kaweah River Drainage, N of Hockett Meadow, MSP 1933, HQ652996, HQ653187, HQ653566, —; (12) unknown locality, ECN 21416, HQ652988, HQ653178, HQ653424, HQ653366.

Tonella floribunda A. Gray: USA, (1) Idaho, Idaho Co., Rapid River Trail, ∼9.7 km SW of Riggins, B. Ertter 19390, HQ652981, HQ653169, HQ653415, —; (2) Washington, Asotin Co., Snake River drainage, 4.8 km S of Asotin, ECN 21538B, AF385336, HQ653170, HQ653416, HQ653363–HQ653364 (CYC1 clones a, b).

T. tonella (Benth.) A. Heller: USA, California, (1) Lake Co., Inner North Coast Ranges, Boggs Lake, BGB & WSA 852, AF385335, HQ653165, HQ653411, —; (2) Napa Co., Inner North Coast Ranges, Howell Mountain, Las Posadas State Forest, Metcalf Trail, WSA & CR s.n. (20 Apr 2005), HQ652979, HQ653167, HQ653413, —; (3) Santa Clara Co., San Francisco Bay Area, Mt. Hamilton Range, Blue Oak Ranch, S. J. Bainbridge s.n. (16 Apr 2007), HQ652980, HQ653168, HQ653414, —; (4) Oregon, Jackson Co., Rogue River Valley, Lower Table Rock, BGB & WSA 1198, HQ652978, HQ653166, HQ653412, HQ653362.

Outgroup taxa

Chelone glabra L: USA, Massachusetts, Franklin Co., Whately, collector unknown, UC1779237, HQ652976, HQ653163, HQ653409, HQ653360.

Keckiella cordifolia (Benth.) Straw: USA, California, San Luis Obispo Co., Outer South Coast Ranges, Santa Lucia Range, Lopez Canyon, BGB 892, HQ652977, HQ653164, HQ653410, HQ653361.

Penstemon hartwegii Benth: Mexico, Puebla, collector unknown, UC Botanical Garden 2002.0860, HQ652975, HQ653162, HQ653408, HQ653359.