American Journal of Botany 95(5): 608–625. 2008.

F ROM ANNUALS TO PERENNIALS: PHYLOGENY OF SUBTRIBE CASTILLEJINAE (OROBANCHACEAE) 1

David C. Tank2,4 and Richard G. Olmstead2,3

2 Department of Biology, University of Washington, Box 355325, Seattle, Washington 98195-5325 USA; and 3 Herbarium, Burke Museum of Natural History, University of Washington, Box 355325, Seattle, Washington 98195-5325

Variation in life history strategies is a fundamental question in evolutionary biology, and the cooccurrence of annual and peren- nial habits in Castilleja and Castillejinae provides the opportunity to study the evolution of plant life history in a phylogenetic context. Molecular phylogenetic analysis of two chloroplast (rps16 and trnL/F) and two nuclear ribosomal (internal and external transcribed spacers) DNA regions support the monophyly of subtribe Castillejinae (Orobanchaceae). A well-supported phylogeny of the six genera ( Castilleja [~180 spp.], Clevelandia [1 sp.], Cordylanthus [18 spp.], Ophiocephalus [1 sp.], Orthocarpus [9 spp.], and Triphysaria [5 spp.]) comprising the subtribe is presented, and morphological synapomorphies are identifi ed for the major lineages recovered. Orthocarpus and Triphysaria are both monophyletic; Cordylanthus is biphyletic. Clevelandia and Ophio- cephalus are derived from within Castilleja . The perennial Castilleja clade (~160 spp.) is derived from a grade of annual taxa in- cluding Castilleja sect. Oncorhynchus (16 spp.), Cordylanthus , Orthocarpus, and Triphysaria . This suggests that the perennial habit evolved a single time from an annual ancestral lineage that persisted throughout the diversifi cation of Castillejinae, contrary to classical interpretations of life history evolution in plants. Given the prevalence of polyploidy among perennial Castilleja spe- cies, perenniality may have played an important role in the origin and establishment of polyploidy in Castilleja .

Key words: annual; Castilleja ; Castillejinae; Cordylanthus ; Orobanchaceae; Orthocarpus ; perennial; polyploidy.

Species belonging to Castilleja (Orobanchaceae), commonly treated in detail from a morphological, anatomical, and cytoge- referred to as the paintbrushes, are some of the most emblem- netic perspective ( Heckard, 1968 ; Chuang and Heckard, 1972 , atic wildfl owers of western North America. Castilleja consists 1973 , 1975a , b , 1976 ; Heckard and Chuang, 1977 ; Heckard of ~180 species distributed from coastal dunes to alpine mead- et al., 1980 ; Chuang and Heckard, 1982 , 1983 , 1986 , 1992a , b , ows. The large majority of paintbrushes are herbaceous peren- 1993 ). Despite this, the lack of a robust phylogenetic hypothesis nials (~160 species), but Castilleja also includes ~20 annual for the paintbrushes and their relatives precludes the evolution- species. In addition to Castilleja , subtribe Castillejinae includes ary study of the fascinating morphological variation and pat- fi ve small genera of annuals: Clevelandia (1 species), Cordy- terns of speciation both within Castilleja and among the genera lanthus (18 species), Ophiocephalus (1 species), Orthocarpus of Castillejinae. (9 species), and Triphysaria (5 species). There is an abundance The cooccurrence of annual and perennial habits in Castilleja of biosystematic data available for many species (e.g., Heckard, and Castillejinae provides the opportunity to study the evolu- 1968 ; Holmgren, 1970 , 1971 , 1976 ; Heckard and Chuang, tion of plant life history in a phylogenetic context. Variation in 1977 ; Heckard et al., 1980 ; Chuang and Heckard, 1982 , 1983 , life history strategies is a fundamental question in evolutionary 1992a , 1993 ), and Castillejinae have received considerable at- biology, and the dichotomy between semelparity — character- tention taxonomically (e.g., Chuang and Heckard, 1991). Each ized by a single reproductive episode — and iteroparity — char- of the genera (excluding many species of Castilleja ) have been acterized by repeated reproductive output — has received considerable attention from a theoretical standpoint ( Young and Augspurger, 1991 ; Stearns, 1992 ). Perhaps the most com- 1 Manuscript received 24 October 2007; revision accepted 28 February 2008. mon form of semelparity in plants is the annual habit, which is The authors thank J. Ammirati, T. Bradshaw, M. Egger, and three in contrast to the iteroparous life history exhibited by most pe- anonymous reviewers for critical comments on earlier drafts of the rennial plants. Most investigations of life history variation in manuscript; M. Donoghue, K. Karol, B. Moore, T. Near, and S. Smith for plants have focused on the mathematical modeling of ecologi- helpful discussions; and S. Collier, P. Lu-Irving, and P. Reeves for laboratory assistance. This research was supported by a Graduate cal and evolutionary scenarios (e.g., bet-hedging, reproductive Fellowship in Molecular Systematics from the University of Washington effort, demography) for the evolution of semelparity in an Department of Botany, the Karling Graduate Student Research Award from iteroparous background (see Young and Augspurger [1991] for the Botanical Society of America, the Research Award for Graduate a review). However, there have been very few studies that have Students from the American Society of Plant Taxonomists, the Award for investigated the evolution of the annual and perennial habits in Graduate Student Research from the Society of Systematic Biologists, a a phylogenetic context (Bena et al., 1998, Conti et al., 1999, Sigma-Xi Grants in Aid of Research from the University of Washington Andreasen and Baldwin, 2001; Verboom et al., 2004). Despite Chapter, and the Giles Award for Graduate Student Field Research from the lack of phylogenetic studies investigating this fundamental the University of Washington Department of Botany to D.C.T., and the trait, there is a widely held opinion among plant evolutionary NSF Doctoral Dissertation Improvement Grant DEB-0412653 to R.G.O. biologists that annuals are derived from perennial ancestors and for D.C.T. 4 Author for correspondence (e-mail: [email protected]); present that the shift between these two strategies is unidirectional (e.g., address: Division of Botany, Peabody Museum of Natural History, Yale Stebbins, 1957 ). University, P.O. Box 208118, New Haven, CT 06520-8118 USA A robust phylogenetic hypothesis provides a framework in which to explore patterns of phenotypic evolution. In this doi: 10.3732/ajb.2007346 paper, we focus on the evolution of life history variation in 608 May 2008] Tank and Olmstead — Castillejinae Phylogeny 609

Castillejinae, but also consider morphological characteristics respectively ( Castilleja racemosa , C. cusickii , C. peckiana , Cordylanthus capi- that have been important historically for circumscribing taxo- tatus , Triphysaria fl oribunda , Orthocarpus pachystachyus , Clevelandia beldin- nomic groups. In addition, we identify diagnostic traits for the gii , and Ophiocephalus angustifolius ). Amplifi ed PCR products were purifi ed by precipitation from a 20% polyethylene glycol solution and washed in 70% major lineages of Castillejinae and discuss systematic implica- ethanol prior to sequencing. After repeated attempts, we were unable to obtain tions. We report results based on data from both the chloroplast any PCR products for some taxa for the rps16 intron, ITS, and ETS DNA re- and nuclear genome. The chloroplast data are from the trnL / F gions (Table 1); however, all taxa included in this study are represented by at region ( Gielly and Taberlet, 1994 ) and the intron of the ribo- least three of the four DNA regions sequenced. somal protein rps16 ( Oxelman et al., 1997 ). The nuclear ge- To ensure accuracy, we sequenced both strands of the cleaned PCR prod- nome is represented by nrDNA sequences of the internal ucts using the DYEnamic ET Terminator Cycle Sequencing Kit (Amersham Biosciences, Piscataway, New Jersey, USA) on an ABI 377 DNA sequencer transcribed spacer (ITS) and external transcribed spacer (ETS) (Applied Biosystems, Foster City, California, USA). For the trnL / F region, regions (Baldwin et al., 1995; Baldwin and Markos, 1998). Be- the internal primers trn-d and trn-e of Taberlet et al. (1991) and trnL-2C and cause hybridization has been suggested to be important in the trnL-2F (located just internal to the PCR primers) of Beardsley and Olmstead evolution of Castilleja and relatives, obtaining data from both (2002) were used to improve sequence quality. The rps16 intron was sequenced the chloroplast and nuclear genomes may permit the identifi ca- using the external PCR primers, rps16_F and rps16_2R, and two internal tion of potential problems in the resulting phylogenetic recon- primers, rps16_iF (5′ -GGTATGTTGCTGCCATTTTG-3′ ) and rps16_iR (5′ - CAAAATGGCAGCAACATACC-3′ ), designed using Castillejinae sequences struction from one or the other source of data. as a reference. The ITS region was sequenced using the PCR primers and two internal primers, its2 and its3 (Baldwin, 1992). The ETS region was sequenced using the 5′ ETS-B PCR primer, and the sequencing primer 18S-E (Baldwin MATERIALS AND METHODS and Markos, 1998 ), that is slightly internal to the 3′ PCR primer. Sequence data were assembled and edited for each region using the program Sequencher Taxon sampling— In total, 76 species of subtribe Castillejinae and three (Gene Codes, Ann Arbor, Michigan, USA), and consensus sequences were outgroup taxa were used in this study (Table 1). We follow the subtribe classi- generated. fi cation of Chuang and Heckard (1991) because it is the most recent and com- plete reorganization of Castillejinae. The majority of species in the subtribe are Phylogenetic analyses— Sequence alignments for all four gene regions represented by the genus Castilleja , in which the taxonomic complexity is ex- were performed manually using the program Se-Al version 2.0a11 ( Rambaut, tremely challenging. The treatment of Castilleja in preparation for the Flora of 1996 ). For each region, parsimony-informative gaps were coded as presence/ North America project (J. M. Egger [WTU], M. Wetherwax [JEPS], and D. C. absence characters using simple gap coding ( Graham et al., 2000 ; Simmons and Tank, unpublished manuscript) is based on Chuang and Heckard’ s broad view Ochoterena, 2000). Complex gaps were not included as additional characters in of the genus. However, J. M. Egger (unpublished data) has tentatively rein- our analyses. stated an infrageneric classifi cation based on detailed fi eld and herbarium ob- Both the trnL / F and rps16 intron regions are part of the haploid chloroplast servations. Within Castilleja , Egger ’ s classifi cation represents the only recent genome, and thus, their histories are linked, and there is no a priori reason to not and comprehensive infrageneric work and, therefore, was used here as a guide combine the data in one analysis. Likewise, the nrDNA ITS and ETS regions for sampling. The taxa on which our analyses are based represent all six genera are tightly linked in the rDNA repeat and can, like the chloroplast, be treated as comprising subtribe Castillejinae, as well as three outgroups. Within Castilleja , one locus. However, differences in base composition and rates of evolution 46 species representing all fi ve subgenera and 11 of 19 sections were sampled. among loci could result in incongruencies between gene trees produced from Efforts were made to include multiple species for most infrageneric groups different data sets (Bull et al., 1993). To determine whether the individual data within Castilleja, except for small or monotypic groups (Table 1). Given the sets are signifi cantly different from random partitions of the combined data, we goals of this study, special attention was given to Castilleja subg. Colacus sect. used the incongruence length difference test (ILD; Farris et al., 1994) in the Oncorhynchus (11 of 16 species sampled), which includes species previously program PAUP* version 4.0b10 ( Swofford, 2002 ) on the combined cpDNA recognized as Orthocarpus ( Chuang and Heckard, 1991 ). All species of Or- and nrDNA data set. All parsimony uninformative sites were excluded from the thocarpus and Triphysaria were sampled, as well as the two monotypic genera partition homogeneity test ( Cunningham, 1997 ), and 1000 replicates of heuris- Clevelandia and Ophiocephalus . Cordylanthus is represented by 14 of 18 spe- tic searches (each with 10 replicates of stepwise random taxon addition and cies, including representatives from all infrageneric groups (sensu Chuang and tree-bisection-reconnection [TBR] branch swapping) were performed. Heckard, 1973 , 1975a , 1986 ). On the basis of previously published molecular Parsimony analyses were conducted on each of the four data sets individu- systematic studies of Orobanchaceae and Scrophulariaceae s.l. (e.g., Young ally, as well as the cpDNA, nrDNA, and combined data sets as implemented in et al., 1999 ; Olmstead et al., 2001 ; Wolfe et al., 2005 ; Bennett and Mathews, PAUP* ( Swofford, 2002 ). Heuristic searches were performed with 1000 repli- 2006; reviewed in Tank et al., 2006), Lamourouxia , Pedicularis , and Seymeria cates of stepwise random taxon addition and TBR branch swapping. Maximum were chosen as outgroups. likelihood (ML) analyses were performed on the cpDNA, nrDNA, and com- bined data sets using PAUP* ( Swofford, 2002 ). The program MODELTEST Molecular methods— Total genomic DNA was extracted from either silica- version 3.6 (Posada and Crandall, 1998) was used to determine the model of gel dried tissue, herbarium specimens, or fresh tissue using the modifi ed 2× sequence evolution best fi t to the data by the hierarchical likelihood ratio test. CTAB method (Doyle and Doyle, 1987) and purifi ed using QIAquick spin- Sequence parameters were estimated by an iterative approach (e.g., Swofford columns following the protocols of the manufacturer (QIAGEN, Valencia, et al., 1996 ). First, parameters were estimated from one of the most parsimonious California, USA). trees. This tree was then used as a starting tree for a heuristic search with near- DNA for sequencing the regions of interest was generated via polymerase est neighbor interchange (NNI) branch swapping, and the estimated parameter chain reaction (PCR) using the trn-c and trn-f primers (Taberlet et al. 1991) for values fi xed for the search. Parameter values were then estimated from the re- the trnL / F region, rps16_ F and rps16 _2R primers ( Oxelman et al., 1997 ) for the sulting tree, and the process was repeated. Once the parameter values estimated rps16 intron, and its4 and its5 primers (Baldwin, 1992) for the entire ITS re- from the resulting tree stabilized, these values were used for the full ML search. gion. To amplify a portion of the 3′ end of the ETS region, the 5′ primer ETS-B The full ML search was conducted using the fi xed parameter values with 10 ( Beardsley and Olmstead, 2002 ) and the 18S-IGS 3′ primer of Baldwin and replicates of stepwise random taxon addition and TBR branch swapping. Markos (1998) were used. For some taxa, we had diffi culty amplifying the Bayesian phylogenetic analyses were conducted using the program MrBayes trnL / F , rps16 , and ITS regions as one fragment. For these taxa, internal primers version 3.1.2 ( Ronquist and Huelsenbeck, 2003 ) on the cpDNA, nrDNA, and were used to amplify the fragments in two parts. For the trnL / F region, the combined data sets. Each analysis was conducted using the same models as trn-c/trn-d and trn-e/trn-f primer pairs were used to amplify the trnL intron and used for the ML analyses and consisted of two runs of 10 000 000 generations trnL - trnF spacer, respectively (Castilleja tenuis , Cordylanthus eremicus subsp. from a random starting tree using the default priors and four Markov chains eremicus , C. ramosus , Ophiocephalus angustifolius , Orthocarpus purpureo- (using the default heating values) sampled every 100 generations. Convergence albus , and O. imbricatus ). For the rps16 intron, the primer pairs rps16_F/ of the chains was determined by examining the plot of all parameter values and rps16_iR and rps16_iF/rps16_2R were used to amplify the region in two pieces the – ln L against generation using the program Tracer version 1.3 ( Rambaut and for one species (Castilleja racemosa ). The ITS1 and ITS 2 regions were ampli- Drummond, 2004 ). Stationarity was assumed when all parameter values and the fi ed separately for eight taxa using the its5/its2 and its3/its4 primer pairs, – ln L had stabilized. Burn-in trees were then discarded and the remaining trees, 610 American Journal of Botany [Vol. 95

Table 1. Taxa and voucher information for plant material from which DNA was extracted. Classifi cation of the subtribes are based on Chuang and Heckard (1986 , 1991 , 1992a ); infrageneric classifi cation of Castilleja is based on J. M. Egger (unpublished data). Estimates of the numbers of species in each group are given in brackets.

GenBank accession number Taxon* [no. spp.] DNA voucher or source a ITS ETSrps16 trnL/F

Castilleja Mutis ex L.f. Subg. Castilleja Sect. Castilleja [29] Castilleja auriculata Eastw. var. auriculata Egger & Tank 1187 (WTU) EF103715 EF103646 EF103787 EF103865 C. integrifolia L.f. subsp. integrifolia Egger & Tank 1207 (WTU) EF103716 EF103648 EF103789 EF103867 C. linariifolia Benth. Tank 01 – 49 (WTU) — EF103647 EF103788 EF103866 C. tenuifl ora Benth. var. tenuifl ora Egger & Tank 1197 (WTU) EF103714 EF103645 EF103786 EF103864 Subg. Colacus (Jeps.) T. I. Chuang & Heckard Sect. Oncorhynchus (Lehm.) T. I. Chuang & Heckard [16] Castilleja ambigua Hook. & Arn. subsp. ambigua Halse 4905 (WTU) EF103686 EF103611 EF103752 EF103830 C. attenuata (A. Gray) T. I. Chuang & HeckardEgger 550 (WTU) EF103687 EF103612 EF103753 EF103831 C. brevistyla (Hoover) T. I. Chuang & HeckardEgger 562 (WTU) — EF103613 EF103754 EF103832 C. campestris (Benth.) T. I. Chuang & Heckard subsp. Beardsley s.n. (WTU) EF103681 EF103605 EF103746 EF103824 campestris C. densifl ora (Benth.) T. I. Chuang & Heckard subsp. Egger 559 (WTU) EF103689 EF103615 EF103756 EF103834 densifl ora C. exserta (A. Heller) T. I. Chuang & Heckard subsp. Egger 623 (WTU) EF103688 EF103614 EF103755 EF103833 exserta C. lacera (Benth.) T. I. Chuang & HeckardEgger 400 (WTU) EF103683 EF103608 EF103749 EF103827 C. lasiorhyncha (A. Gray) T. I. Chuang & HeckardBeardsley 98 – 27 (WTU) EF103684 EF103609 EF103750 EF103828 C. lineariloba (Benth.) T. I. Chuang & HeckardEgger 555 (WTU) — EF103606 EF103747 EF103825 C. rubicundula (Jeps.) T. I. Chuang & Heckard subsp. Egger 570 (WTU) EF103685 EF103610 EF103751 EF103829 rubicundula C. tenuis (A. Heller) T. I. Chuang & HeckardTank 01 – 13 (WTU) EF103682 EF103607 EF103748 EF103826 Sect. Pilosae T. I. Chuang & Heckard [8] Castilleja arachnoidea Greenm. Tank 01 – 34 (WTU) EF103691 EF103618 EF103759 EF103837 C. nana Eastw. Tank 01 – 48 (WTU) — EF103617 EF103758 EF103836 C. pilosa Rydb.Colwell s.n. (WTU) EF103690 EF103616 EF103757 EF103835 Subg. Euchroma Sect. Affi nes [2] Castilleja affi nis Benth. subsp. affi nis Beardsley 98 – 48 (WTU) EF103708 EF103638 EF103779 EF103857 Sect. Euchroma (Nutt.) Benth. [30] Castilleja arvensis Cham. & Schltdl. Egger & Tank 1185 (WTU) EF103698 EF103626 EF103767 EF103845 C. conzattii Fernald Egger & Tank 1208 (WTU) EF103697 EF103625 EF103766 EF103844 C. nivibractea G. L. Nesom Egger & Tank 1209 (WTU) — EF103633 EF103774 EF103852 C. scorzonerifolia Kunth Egger & Tank 1195 (WTU) EF103696 EF103624 EF103765 EF103843 C. zempoaltepetlensis G. L. Nesom Egger & Tank 1199 (WTU) EF103703 EF103632 EF103773 EF103851 Sect. Hispidae [13] Castilleja chromosa A. Nelson Myers s.n. (WTU) — EF103629 EF103770 EF103848 C. hispida Benth. var. hispida Tank 01 – 2 (WTU) EF103699 EF103627 EF103768 EF103846 C. parvifl ora Bong. Olmstead 01 – 83 (WTU) EF103701 EF103630 EF103771 EF103849 C. peckiana Pennell Tank 01 – 29 (WTU) EF103700 EF103628 EF103769 EF103847 C. peirsonii Eastw. Tank 01 – 52 (WTU) EF103702 EF103631 EF103772 EF103850 Sect. Septentrionales [22] Castilleja elata Piper Tank 01 – 37 (WTU) EF103713 EF103643 EF103784 EF103862 C. elmeri FernaldOlmstead 01 – 78 (WTU) EF103709 EF103639 EF103780 EF103858 C. integra A. Gray var. integra Tank 01 – 58 (WTU) EF103704 EF103634 EF103775 EF103853 C. lutescens (Greenm.) Rydb.Myers s.n. (WTU) EF103711 EF103641 EF103782 EF103860 C. miniata Douglas ex Hook. Colwell s.n (WTU) EF103712 EF103642 EF103783 EF103861 C. occidentalis Torr. Colwell s.n . (WTU) EF103710 EF103640 EF103781 EF103859 C. septentrionalis Lindl. Beardsley and Olmstead, 2002 AF478944 AF478977 — AF479008 (as C. sulphurea Rydb.) Sect. Stenanthae [1] Castilleja exilis A. Nelson Egger 763 (WTU) — EF103644 EF103785 EF103863 Sect. Viscidulae [11] Castilleja applegatei Fernald subsp. pinetorum (Fernald) RH 40 (WTU) EF103705 EF103635 EF103776 EF103854 T. I. Chuang & Heckard C. pruinosa Fernald Donahue 007 (WTU) EF103707 EF103637 EF103778 EF103856 C. xanthotricha Pennell Olmstead 00 – 39 (WTU) EF103706 EF103636 EF103777 EF103855 Subg. Gentrya (Breedlove & Heckard) T.I. Chuang & Heckard [1] Castilleja racemosa (Breedlove & Heckard) Breedlove 19200 (JEPS) EF103680 EF103604 EF103745 EF103823 T. I. Chuang & Heckard Subg. Pallescentes Sect . Pallescentes (Rydb.) T. I. Chuang & Heckard [6] Castilleja oresbia Greenm. Tank 01 – 27 (WTU) EF103692 EF103620 EF103761 EF103839 C. plagiotoma A. Gray Ertter 3430 (WTU) — EF103619 EF103760 EF103838 May 2008] Tank and Olmstead — Castillejinae Phylogeny 611

Table 1. Continued.

GenBank accession number Taxon* [no. spp.] DNA voucher or source a ITS ETSrps16 trnL/F C. praeterita Heckard & Bacig.Tank 01 – 56 (WTU) EF103695 EF103623 EF103764 EF103842 Sect . Pallidae [17] Castilleja cusickii Greenm. Tank 01 – 28 (WTU) EF103694 EF103622 EF103763 EF103841 C. lemmonii A. Gray Tank 01 – 51 (WTU) EF103693 EF103621 EF103762 EF103840 Cordylanthus Nutt. ex Benth. Subg. Cordylanthus Sect. Anisochelia A. Gray [6] Cordylanthus capitatus Nutt. ex Benth. Arnot 739 (WTU) EF103723 EF103658 EF103799 EF103877 C. eremicus (Coville & C. V. Morton) Munz J. L. 1781 (WTU) — EF103661 EF103802 EF103880 subsp. eremicus C. kingii S. Watson subsp. kingii Egger 712 (WTU) — EF103659 EF103800 EF103878 C. wrightii A. Gray subsp. wrightii Egger 748 (WTU) EF103724 EF103660 EF103801 EF103879 Sect. Cordylanthus [6] Cordylanthus pilosus A. Gray subsp. pilosus Ertter 3937 (WTU) EF103722 EF103657 EF103798 EF103876 C. pringlei A. Gray Smith 9395 (WTU) — EF103654 EF103795 EF103873 C. rigidus (Benth.) Jeps. subsp. rigidus Collum 408 (WTU) — EF103655 EF103796 EF103874 C. tenuis A. Gray subsp. tenuis Egger 862a (WTU) EF103721 EF103656 EF103797 EF103875 Sect. Ramosi (Pennell) T. I. Chuang & Heckard [1] Cordylanthus ramosus Nutt. ex Benth.Smith 3567 (WTU) EF103725 EF103662 EF103803 EF103881 Subg. Hemistegia (A. Gray) Jeps. [4] Cordylanthus maritimus Nutt. subsp. canescens Tiehm 12253 (WTU) EF103717 EF103649 EF103790 EF103868 (A. Gray) T. I. Chuang & Heckard C. mollis A. Gray subsp. hispidus (Pennell) Egger 860 (WTU) EF103718 EF103650 EF103791 EF103869 T. I. Chuang & Heckard C. palmatus (Ferris) J. F. Macbr. Egger 725 (WTU) EF103719 EF103651 EF103792 EF103870 C. tecopensis Munz & J. C. Roos Egger 588 (WTU) EF103720 EF103652 EF103793 EF103871 Subg. Dicranostegia (A .Gray) T. I. Chuang & Heckard [1] Cordylanthus orcuttianus A. Gray Egger 887b (WTU) — EF103653 EF103794 EF103872 Orthocarpus Nutt. [9] Orthocarpus tolmiei Hook. & Arn. subsp. tolmiei Egger 749 (WTU) EF103726 EF103663 EF103804 EF103882 O. luteus Nutt. Colwell s.n. (WTU) EF103727 EF103664 EF103805 EF103883 O. purpureo-albus A. Gray ex S. Watson Holmgren 7389 (WTU) EF103728 EF103665 EF103806 EF103884 O. bracteosus Benth. Olmstead 98 – 07 (WTU) EF103729 EF103666 EF103807 EF103885 O. barbatus J. S. Cotton Caplow & Beck 95072 (WTU) EF103730 EF103667 EF103808 EF103886 O. cuspidatus Greene subsp. cuspidatus Egger 637 (WTU) EF103731 EF103668 EF103809 EF103887 O. pachystachyus A. Gray Taylor 15721 (JEPS) EF103732 EF103669 EF103810 EF103888 O. imbricatus Torr. ex S. Watson Weber 12222 (WTU) EF103733 EF103670 EF103811 EF103889 O. tenuifolius (Pursh) Benth. Egger 960 (WTU) EF103734 EF103671 EF103812 EF103890 Triphysaria Fisch. & C. A. Mey. [5] Triphysaria eriantha (Benth.) T. I. Chuang & Heckard Egger 551 (WTU) EF103735 EF103672 EF103813 EF103891 subsp. rosea (A. Gray) T. I. Chuang & Heckard T. fl oribunda (Benth.) T. I. Chuang & HeckardEgger 1254 (WTU) EF103736 EF103673 EF103814 EF103892 T. micrantha (Greene ex A. Heller) T. I. Egger 953b (WTU) EF103737 — EF103815 EF103893 Chuang & Heckard T. pusilla (Benth.) T. I. Chuang & HeckardNass 5702 (WTU) EF103738 EF103674 EF103816 EF103894 T. versicolor Fisch. & C. A. Mey. subsp. Olmstead 541 (WTU) EF103739 EF103675 EF103817 EF103895 faucibarbata (A. Gray) T. I. Chuang & Heckard Clevelandia Greene [1] Clevelandia beldingii (Greene) Greene Heckard 3958 (JEPS) EF103740 EF103676 EF103818 EF103896 Ophiocephalus Wiggins [1] Ophiocephalus angustifolius Wiggins Moran 15394 (JEPS) EF103741 EF103677 EF103819 EF103897 Seymeria Pursh Seymeria laciniata Standl. Egger & Tank 1201 (WTU) EF103742 EF103678 EF103820 EF103898 Pedicularis L. Pedicularis attolens A. Gray Tank 01 – 50 (WTU) EF103743 EF103679 EF103821 EF103899 Lamourouxia Kunth Lamourouxia rhinanthifolia Kunth Egger & Tank 1190 (WTU) EF103744 — EF103822 EF103900 a References are for accessions used in previous studies; vouchers are for accessions fi rst used here (herbarium where deposited in parentheses). *Unsampled infrageneric groups of include Castilleja subg. Euchroma sects. Lanatae [9], Latifoliae [3], and Bryantii [1]; Castilleja subg. Castilleja sects. Epichroma Benth. [5], Ortegae G. L. Nesom [4], and Spiranthoides [5]; Castilleja subg. Pallescentes sects. Flavae [4] and Callichroma Benth. [3].

and their associated parameter values, were saved. To increase the chance of (i.e., two data partitions corresponding to the cpDNA and nrDNA data sets, exploring more of tree space and decrease the chance of obtaining stationarity where each is modeled under different parameter values) analyses were per- on local optima, we ran two independent analyses for each data set. In addition, formed. Following Nylander et al. (2004) , Brandley et al. (2005) , and Brown for the combined data set both single-model (i.e., where the two data sets, cp- and Lemmon (2007) we used the Bayes factor to choose between the single- DNA and nrDNA, are modeled by the same parameter values) and partitioned model and partitioned Bayesian analyses. 612 American Journal of Botany [Vol. 95

Nonparametric bootstrapping (Felsenstein, 1985) was used to evaluate rela- 2001 ) and has been criticized as too conservative (e.g., Graham tive support for particular clades recovered in the phylogenetic analyses. Parsimony et al., 1998; Barker and Lutzoni, 2002; Darlu and Lecointre, bootstrapping was performed with 500 replicates, each with 20 replicates of 2002). Furthermore, the trees resulting from separate analy- stepwise random taxon addition and TBR branch swapping with MULTREES off ( DeBry and Olmstead, 2000 ). Maximum likelihood bootstrapping was ses of the cpDNA and nrDNA data sets are largely consistent performed with 250 replicates, each with three replicates of stepwise ran- (Figs. 1 and 2). When the nrDNA topology was constrained dom taxon addition and NNI branch swapping. In addition to parsimony and to include the well-supported nodes recovered by phyloge- ML bootstrap values, posterior probabilities (PP) resulting from Bayesian phy- netic analyses of the cpDNA data (i.e., Bayesian PP values logenetic analyses were also used to evaluate relative branch support. A major- ≥ 0.95 and bootstrap values ≥ 70%; see Fig. 1), the Shimodaira– ity rule consensus tree showing all compatible partitions from the resulting Hasegawa (SH) test indicated that the resulting likelihood posterior distribution of tree topologies was used to recover the Bayesian PP values for each clade. The Shimodaira – Hasegawa test (SH; Shimodaira and values were not signifi cantly different (P = 0.12). Likewise, Hasegawa, 1999), as implemented in PAUP* (Swofford, 2002), was used to when the cpDNA topology was constrained to include the evaluate whether major topological differences between results of the separate well-supported nodes recovered by analyses of the nrDNA cpDNA and nrDNA analyses were signifi cant, as well as to evaluate whether data, the SH test indicated no signifi cant difference between alternative topological hypotheses based on traditional taxonomy were signifi - the two topologies (P = 0.08). Therefore, in the absence of any cantly different than the topology resulting from phylogenetic analyses of the indication that such an analysis is compromised by confl icting combined cpDNA and nrDNA data (100 000 bootstrap replicates using RELL optimization). histories in the two genomes, we conducted a combined analy- sis to take advantage of the greater resolution that the larger data set can provide. RESULTS Results of the parsimony analyses for the cpDNA, nrDNA, and combined data are shown in Table 2, and the resulting strict Phylogenetic analyses— The two noncoding chloroplast re- consensus trees for the cpDNA and nrDNA data are shown in gions, trnL / F and rps16 , aligned unambiguously, although nu- Fig. 1. For the cpDNA, nrDNA, and combined data, model se- merous short gaps were introduced. For the trnL / F region, all lection as implemented in the program MODELTEST ( Posada taxa were included, but Castilleja septentrionalis was missing and Crandall, 1998) resulted in the GTR+I+Γ (GTR = general for the rps16 region (sequences for the other three regions for time reversible, I = proportion of invariable sites, Γ = gamma C. septentrionalis were downloaded from GenBank; Table 1). distributed variable sites) model of sequence evolution. The it- The tightly linked nrDNA ITS and ETS regions were more dif- erative approach taken to estimate parameters for the maximum fi cult to align, but we were still able to align the majority of likelihood (ML) analyses (see Materials and Methods, Phylo- both data sets. Regions that could not be unambiguously aligned genetic analyses ) took three rounds of parameter estimates for were excluded from the phylogenetic analyses (positions 258– each of the three data sets before the parameter values stabi- 265 in the ITS alignment and positions 192 – 195 in the ETS lized; the parameter values fi xed for the ML analyses are shown alignment). Characteristics of the cpDNA and nrDNA regions in Table 3 . Maximum likelihood analyses of the cpDNA data sequenced for this study are summarized in Table 2 . resulted in two equally optimal trees that only differ in the Because of their linked histories, the cpDNA trnL /F and placement of one taxon in a clade where there is little or no rps16 regions were treated as one locus, and the nrDNA ITS branch support. The resulting ML trees for each of the data sets and ETS regions were treated as one locus; therefore, we only (not shown) were congruent in overall topology with those present results for the combined cpDNA data and the com- resulting from parsimony and Bayesian analyses. Bayesian bined nrDNA data. The partition homogeneity test for cpDNA analysis of the cpDNA and nrDNA data achieved apparent sta- vs. nrDNA resulted in a signifi cant difference (P = 0.01); how- tionarity after ~1 000 000 generations, however, because this ever, this result is not surprising given the conservative nature analysis contained long chains (10 000 000 generations) and a of the test and the different base composition and sequence high sampling frequency (every 100 generations), a conserva- divergence values between the two data sets ( Tables 2 and 3 ). tive burn-in of 2 000 000 generations was used. The majority Nevertheless, the partition homogeneity test should not be rule consensus trees calculated from the posterior distribution considered a test of combinability of the data (Yoder et al., (excluding burn-in trees) with mean branch lengths are shown

Table 2. Summary descriptions and parsimony results for sequences included in individual and combined analyses of two chloroplast regions (trnL / F and rps16 ) and two nuclear ribosomal regions (ITS and ETS) (cpDNA = chloroplast combined, nrDNA = nuclear ribosomal combined).

Results trnL / F rps16 cpDNA ITS ETS nrDNA Combined

Number of taxa included 79 78 79 66 77 79 79 Sequence characteristics Length of sequenced regions (range) 921– 1033 862– 889 1783– 1922 683– 698 425– 442 1108– 1140 2891– 3062 Excluded characters 0 0 0 8 4 12 12 Aligned length 1147 952 2099 729 468 1197 3296 Gaps coded as binary characters 10 8 18 13 6 19 37 Variable sites 171 173 344 305 324 629 973 Pairwise distances (%) 0 – 6.1 0 – 8.9 — 0 – 31.9 0.2 – 59.0 — — Parsimony analyses Informative sites 77 63 140 226 240 466 606 Number of trees — — 137 — — 1650 58 Number of steps — — 480 — — 1956 2523 Consistency index (CI) — — 0.82 — — 0.51 0.55 Retention index (RI) — — 0.92 — — 0.74 0.76 Rescaled consistency index (RC) — — 0.75 — — 0.38 0.42 May 2008] Tank and Olmstead — Castillejinae Phylogeny 613

Fig. 1. Strict consensus trees recovered from parsimony analysis of the nuclear ribosomal DNA data (nrDNA; ITS + ETS) and the chloroplast DNA data (cpDNA; trnL/F + rps16 ). The lineages corresponding to the six genera of Castillejinae, as well as the three subgenera of Cordylanthus , are specifi ed. Numbers along the branches indicate parsimony bootstrap percentages, maximum likelihood (ML) bootstrap percentages, and Bayesian posterior probabil- ity (PP) values for the major lineages of Castillejinae, respectively (i.e., parsimony/ML/PP). Branches marked by an asterisk received Bayesian PP values ≥ 0.95 and parsimony and ML bootstrap percentages ≥ 70%. in Fig. 2. To avoid the pitfall of achieving apparent stationarity Although model selection indicated that both the cpDNA and on a local optimum, we ran two independent Bayesian analyses nrDNA data sets are best modeled under the same model for each data set; in both analyses, all parameters reached (GTR+I+Γ ), results of the partition homogeneity test and ML stationarity at the same level. The results presented here (Fig. 2, parameter estimates (e.g., widely differing base composi- Table 4 ) are those of one of the independent runs. tions and substitution rates between the two data sets; Table 3 )

Table 3. Summary of fi xed parameter values used in maximum likelihood analyses for the chloroplast (cpDNA), nuclear ribosomal (nrDNA), and combined data sets.

Nucleotide frequencies Rate matrix Proportion of Gamma shape Data set (A, C, G, T) (A → C, A → G, A → T, C→ G, C → T ) invariable sites parameter (℘ ) ln Likelihood cpDNA 0.35, 0.17, 0.17, 0.31 0.810, 0.720, 0.280, 0.489, 1.058 0.249 1.012 − 6079.722 nrDNA 0.19, 0.27, 0.28, 0.26 1.210, 1.980, 1.600, 0.316, 3.594 0.165 0.806 − 11558.972 Combined 0.29, 0.21, 0.21, 0.29 0.857, 1.080, 0.620, 0.440, 2.777 0.395 0.575 − 18764.840 614 American Journal of Botany [Vol. 95

thocarpus forms a clade with the rest of the subtribe (parsimony bootstrap = 73%; ML bootstrap = 58%; Bayesian PP = 0.86). When the nrDNA topology was constrained to match the cpDNA topology with respect to these lineages (i.e., Cordylan- thus subg. Cordylanthus sister to Orthocarpus ), the Shimodaira – Hasegawa test indicated that the resulting likelihood values were not signifi cantly different (P = 0.41). Likewise, when the cpDNA topology was constrained to that of the nrDNA analy- ses, the SH test indicated no signifi cant difference between the two topologies ( P = 0.26). Therefore, the results of the cpDNA and nrDNA analyses are not in strong confl ict with each other. In all analyses, the large genus Castilleja also was paraphyletic with the two monotypic genera Clevelandia and Ophiocephalus nested within Castilleja. Monophyly of Castilleja including Clevelandia and Ophiocephalus ( Castilleja s.l.) was well sup- ported in the analyses of the nrDNA data set (parsimony boot- strap = 98%; ML bootstrap = 86%; Bayesian PP = 1.0), but only weakly supported by the cpDNA data (parsimony bootstrap = 59%; ML bootstrap = 60%; Bayesian PP = 0.60). In contrast, results from both DNA regions provided strong support for the clade comprised of Castilleja s.l. and Triphysaria . Figures 3 and 4 show the majority rule consensus tree result- ing from the partitioned Bayesian analysis of the combined cpDNA and nrDNA data with PP values and mean branch lengths, respectively. This tree was largely congruent with those Fig. 2. Majority rule consensus trees (excluding burn-in trees) with resulting from the separate analyses of the cpDNA and nrDNA mean branch lengths from the separate Bayesian analyses of the nuclear regions (Figs. 1 and 2) with all analytical methods, as well as the ribosomal DNA data (nrDNA; ITS + ETS) and the chloroplast DNA data parsimony, ML, and single-model Bayesian analyses of the (cpDNA; trnL/F + rps16). Terminal taxon names have been removed and combined data (trees not shown), and represents our best hy- the lineages corresponding to the six genera of Castillejinae, as well as the pothesis for the phylogeny of subtribe Castillejinae. In agree- three subgenera of Cordylanthus , are specifi ed. Branch lengths for both ment with the results from the cpDNA data set ( Figs. 1 and 2 ), trees are proportional to the mean number of substitutions per site as mea- Cordylanthus subg. Cordylanthus, and Orthocarpus were re- sured by the scale bar. solved as monophyletic and sister to the remainder of the sub- tribe. However, as a result of the disagreement between the suggest that the two DNA regions may be more appropriately nrDNA and cpDNA data, support for this relationship decreased modeled as two data partitions. In addition, the Bayes factor in the combined analyses relative to the cpDNA results. Within result from the comparison of the single-model and partitioned Castilleja s.l., some relationships that were unresolved or poorly analyses provides very strong evidence (i.e., > 10; Kass and supported by the individual analyses were recovered with in- Raftery, 1995) for the partitioned analysis (Bayes factor = creased support in the combined analysis. This was most evident 371.42). Therefore, we have chosen to show the resulting trees with respect to the annual lineages of Castilleja s.l., including from the partitioned analysis only; Figs. 3 and 4 show the ma- Castilleja subg. Colacus sect. Oncorhynchus ( Table 1 , Fig. 5 ) jority rule consensus tree from the partitioned Bayesian analy- and the monotypic annual genera Clevelandia and Ophiocepha- sis with PP values and mean branch lengths, respectively. lus . In the combined analyses, these taxa were resolved as a Figures 1 and 2 show comparisons of the phylogenetic analy- basal grade of annual lineages leading to a large clade of peren- ses of the separate cpDNA and nrDNA data sets. Figure 1 shows nial Castilleja species (Figs. 3 and 4). Of these annual lineages, a comparison of the strict consensus trees resulting from Clevelandia and Ophiocephalus were resolved as the sister parsimony analyses; bootstrap values (parsimony and ML) and group to the perennial Castilleja clade, albeit with low support. Bayesian PP values resulting from the individual analyses are Nevertheless, a strongly supported clade including Clevelandia, shown for the major lineages and clades receiving Bayesian PP Ophiocephalus, and the perennial Castilleja species was recov- values ≥ 0.95 and bootstrap values ≥ 70% are indicated with an ered by all analytical methods (parsimony bootstrap = 81%; ML asterisk. Figure 2 shows a comparison of the Bayesian majority bootstrap = 99%; Bayesian PP = 1.0). When the topology recov- rule consensus trees with mean branch lengths. In all analyses, ered in the combined analyses ( Figs. 3 and 4 ) was constrained to the monophyly of subtribe Castillejinae as well as the two gen- correspond to traditional taxonomic groups (i.e., monophyly of era Orthocarpus and Triphysaria was well supported. However, Castilleja, monophyly of Castilleja sect. Oncorhynchus, and the genus Cordylanthus was biphyletic, forming two well- monophyly of Cordylanthus), the Shimodaira– Hasegawa test supported clades, with one clade corresponding to subg. Cordy- indicated that these alternative topologies were signifi cantly lanthus and the other comprised of the monotypic subg. less likely ( Table 4 ). Dicranostegia sister to subg. Hemistegia . Results from analy- ses of the cpDNA data provided moderate support for a sister- group relationship between Cordylanthus subg. Cordylanthus DISCUSSION and the genus Orthocarpus (parsimony bootstrap = 85%; ML bootstrap = 63%; Bayesian PP = 0.95). However, this relation- Throughout the history of the subtribe, there has been ship was not recovered in the nrDNA analyses, in which Or- diffi culty defi ning generic boundaries in Castillejinae. This May 2008] Tank and Olmstead — Castillejinae Phylogeny 615

Table 4. Summary of morphological characteristics traditionally used to recognize taxonomic groups in Castillejinae and whether the group was recovered as monophyletic in our analyses. Characters in boldface represent morphological synapomorphies, as noted on Fig. 5. The Shimodaira– Hasegawa (SH) test was used to evaluate whether constraining traditional taxonomic groups to be monophyletic resulted in signifi cantly different topologies in combined cpDNA and nrDNA analyses.

Genus Traditional diagnostic morphological characters a Monophyletic?P -value in SH test

Castilleja Annual or perennial; upper corolla lip (beak) open at tip; stigma expanded; hilum of seed terminal; No 0.009 seed coat loose-fi tting; x = 12 Castilleja subg. Colacus Annual; upper corolla lip (beak) open at tip, ≤ 3 times length of lower lip; stigma expanded; hilum of No 0.015 sect. Oncorhynchus seed terminal; seed coat loose-fi tting; x = 12 Clevelandia Annual; corolla only slightly bilabiate, funnel-shaped, upper lip not beaked or galeate; stigma — — expanded; hilum of seed terminal; seed coat loose-fi tting; N = 12 Cordylanthus Annual; upper corolla lip (galea) closed at tip, lower lip not toothed; calyx spathe-like (cleft No 0.031 abaxially completely, or nearly so); stigma unexpanded; hilum of seed lateral Ophiocephalus Annual; corolla only slightly bilabiate, club-shaped, upper lip not beaked or galeate; stamens — — strongly exseted; anther sacs nearly equal and attached medially to fi lament; N = 12 Orthocarpus Annual; upper corolla lip (galea) closed at tip, lower lip minutely 3-toothed ; calyx unequally Yes— 4-cleft with deeper cut adaxially ; stigma unexpanded; hilum of seed lateral Triphysaria Annual; upper corolla lip (beak) open at tip, corolla throat indented forming a fold ; stigma Yes— expanded; hilum of seed terminal; seed coat tight-fi tting; single anther sac per stamen ; N = 11 a sensu Chuang and Heckard, 1991 diffi culty has been especially evident between Castilleja and these phylogenetic results suggest and identify diagnostic mor- Orthocarpus, where numerous species have been shifted be- phological traits ( Fig. 5 ) that will serve as a guide for a taxo- tween the two genera (e.g., Gray, 1862; Watson, 1871; Eastwood, nomic revision of Castillejinae. 1909 ; Jepson, 1925 ; Keck, 1927 ; Chuang and Heckard, 1991 , 1992b ). The most recent treatment of Castillejinae, based on Evolution of perenniality — Among the genera of Castilleji- extensive morphological study including fl oral morphology, nae, perenniality is limited to Castilleja , in which perennial seed and seed-coat morphology, and cytological analyses, led species are the large majority of the approximately 180 species. to a major realignment of generic boundaries in the subtribe Although Chuang and Heckard (1991) recognized the unique ( Chuang and Heckard, 1991 ). In this treatment, Orthocarpus Castilleja fl oral structure (i.e., elongate upper corolla lip, lower was redefi ned to include only nine annual species, elevating corolla lip reduced to teeth) as a derived morphology in the Orthocarpus subg. Triphysaria to genus, and assigning 12 other subtribe, rather than interpreting the evolution of perennial annual Orthocarpus species to Castilleja ( Chuang and Heckard, habit as derived, they hypothesized at least fi ve independent 1991 , 1992b ). They also concluded that the monotypic genus origins of annual lineages from a perennial ancestor ( Fig. 6 ). In Gentrya did not differ enough from Castilleja to warrant ge- their generic realignment, Chuang and Heckard (1991) moved neric status and reassigned it to Castilleja (C. racemosa) as the 12 annual species from Orthocarpus (sensu Keck, 1927 ) and monotypic subg. Gentrya ( Chuang and Heckard, 1991 ). the monotypic annual genus Gentrya into their expanded Cas- In addition to their realignment of generic boundaries within tilleja, breaking from the tradition of Castilleja as a strictly pe- Castillejinae, Chuang and Heckard (1991) proposed an “ intui- rennial genus. However, Castilleja sect. Oncorhynchus , Castilleja tive phylogeny” depicting relationships among the major lin- racemosa , Clevelandia , Ophiocephalus, and Triphysaria were eages of the subtribe ( Fig. 6 ) . A morphological cladistic analysis all viewed as annual lineages independently derived from an (Chuang, 1993) with 11 terminal taxa (genera and major infrage- ancestral perennial lineage (Cordylanthus and Orthocarpus neric groups) resulted in a nearly identical topology. They relied were left unresolved with respect to the other lineages; Fig. 6). on these estimates of phylogenetic relationship to interpret the This interpretation likely stems from the conventional wisdom evolution of some major morphological characters, including that annual taxa are derived from perennial ancestors. Histori- the evolution of the annual and perennial life history strategies. cally, this belief is seated in two lines of evidence. First, mor- The phylogenetic analyses presented here resulted in a robust phologists and plant evolutionary biologists have long held that hypothesis of phylogeny for the major lineages comprising the woody habit, and thus the perennial life history strategy, Castillejinae (Figs. 3– 5 ). The nearly complete sampling of the is the ancestral condition in angiosperms (e.g., Jeffrey, 1916; annual groups of Castillejinae (e.g., Castilleja subg. Gentrya , Ames, 1939 ; Stebbins, 1957 ), and therefore an herbaceous Castilleja sect. Oncorhynchus , Cordylanthus , Orthocarpus , annual would necessarily be derived. Second, the majority of and Triphysaria ), in addition to representative sampling of the studies investigating the annual habit in plants are devoted to perennial Castilleja species, allows us to investigate the evolu- the evolution of the annual habit as a bet-hedging strategy with tion of life history strategies and explore the association of an- respect to their perennial relatives in extreme and unpredictable nual and perennial life histories with traits associated with them environments. Multiple authors have discussed the origin of in Castillejinae. Specifi cally, we are interested in the direction desert annuals from perennial ancestors (e.g., Gleason and of evolutionary change between the annual and perennial life Cronquist, 1964 ; Johnson, 1968 ; Axelrod, 1979 ), and this com- histories in Castilleja . In addition, we discuss in detail the evo- mon view may be an extension of some of these ideas. Surpris- lution of a number of fl oral characteristics that have been im- ingly, there have been few studies on the evolution of the annual portant historically for guiding taxonomic treatments in and perennial habits in a phylogenetic context (e.g., Bena et al., Castillejinae (e.g., Chuang and Heckard, 1973 , 1975a , 1986 , 1998; Conti et al., 1999; Andreasen and Baldwin, 2001; 1991 , 1992b ). Finally, we discuss systematic implications that Verboom et al., 2004 ). 616 American Journal of Botany [Vol. 95

Fig. 3. Majority rule consensus tree (excluding burn-in trees) from the partitioned Bayesian analysis of the combined chloroplast and nuclear ribo- somal DNA data. Numbers above the branches indicate Bayesian posterior probability (PP) values. Numbers below the branches indicate maximum likeli- hood (ML) and parsimony bootstrap percentages (BS), respectively. Bootstrap percentages for clades that were not recovered in the ML and/or parsimony bootstrap consensus trees are indicated with dashes. May 2008] Tank and Olmstead — Castillejinae Phylogeny 617

These results demonstrate that, contrary to the classical inter- tablishment of polyploidy. Because polyploid complexes are pretation, the perennial Castilleja clade is derived from a grade of common throughout perennial Castilleja (Heckard, 1968; annual taxa including the annual Castilleja species comprising Heckard and Chuang, 1977; Chuang and Heckard, 1993), pe- sect. Oncorhynchus ( Fig. 5 ). A straightforward parsimony recon- renniality could be important for the maintenance of spontane- struction of life history strategy on our best estimate of Castille- ous polyploids in populations from which they otherwise would jinae phylogeny suggests that the perennial habit evolved a single be reproductively isolated. Reproductive isolation of these indi- time from an annual ancestral lineage that persisted throughout viduals could then be alleviated by the spontaneous formation the diversifi cation of Castillejinae. Since the evolution of peren- of other individuals of the same ploidy level. Contrary to clas- niality in Castilleja , there have been some reversions to the sical view, the recurrent formation of polyploids is now consid- annual habit. Castilleja arvensis , C. exilis , and C. racemosa ered the rule, rather than the exception ( Soltis and Soltis, 1993 ; (formerly Gentrya racemosa ) are all annual species sampled in Soltis et al., 2003). Despite the prevalence of polyploidy among this study that are derived from within the perennial Castilleja perennial Castilleja species, the lack of resolution of interspe- clade (Fig. 5), and the annual sect. Epichroma (fi ve spp., not cifi c relationships and limited sampling (interspecifi c and in- sampled here) is likely another. However, with the exception of traspecifi c) within this clade prevents any inference as to the the annual habit, these species are more similar morphologically importance of polyploidy to speciation, but will be the focus of to other perennial Castilleja species, including their fl oral mor- future research. phology, which conforms to the derived fl ower structure seen in In addition, these analyses revealed a marked difference be- the perennial Castilleja clade. In addition, the two Baja Califor- tween the amount of sequence divergence (as evidenced by the nia endemics, Clevelandia and Ophiocephalus , may be derived mean branch length estimates) within each of the annual lin- from within the perennial clade, although the optimal trees place eages of Castillejinae and the clade comprising the perennial them as sister to the perennial clade with weak support ( Fig. 3 ). Castilleja species (Figs. 2 and 4). Although the amount of in- Therefore, it is likely that Clevelandia and Ophiocephalus retain ferred sequence divergence from the most recent common the ancestral annual habit and are either the sister lineage or suc- ancestor of Castillejinae is roughly equivalent between the cessive sister lineages to the perennial Castilleja clade. perennials and any of the annual lineages, the inferred sequence Polyploidy is found only in Castilleja , and the large major- divergence among species in each terminal clade varies dra- ity of polyploid species are found in the perennial Castilleja matically between annuals and perennials. Given the represen- clade ( Fig. 4 ), where the number of polyploid species and in- tative, but incomplete, sampling of perennial Castilleja species, traspecifi c variation in ploidy level is great. Some species are it is clear that the perennial clade has dramatically less diver- reported to have a single ploidy level (e.g., C. arvensis [ N = 12], gence among species than the annual clades. This conspicu- C. pruinosa [n = 24]), while others have both diploid (n = 12) ously lower sequence divergence at both chloroplast and nuclear and tetraploid (n = 24) counts reported (e.g., C. chromosa , loci suggests that the diversifi cation of perennial Castilleja spe- C. linariifolia , C. tenuifl ora ), and some include multiple ploidy cies in western North America represents a recent event in the levels (e.g., C. lutescens [n = 24, 48, 60], C. miniata [n = 12, evolution of Castillejinae. 24, 36, 48, 60], C. peckiana [n = 36, 48, 60], C. septentrionalis [n = 12, 24, 48]) (Heckard, 1968; Heckard and Chuang, 1977; Morphological evolution and systematic implications— The Spellenberg, 1986 ; Chuang and Heckard, 1993 ; Chambers results from molecular phylogenetic analysis of the chloroplast et al., 1998 ). Based on chromosome counts from approximately and nuclear ribosomal DNA regions support the monophyly of 120 species of Castilleja , it was estimated that, in western subtribe Castillejinae. Figure 5 shows a summary of relation- North America, more than 50% of the species have polyploid ships among the major lineages of Castillejinae, their current representatives, while in Mexico 25% of the species were esti- taxonomic circumscription, and a list of diagnostic morphologi- mated to be polyploids ( Chuang and Heckard, 1993 ). cal traits for each clade. Table 4 provides a summary of mor- Stebbins (1938 , 1950 ) noted an association between herba- phological characteristics traditionally used to recognize genera ceous perennials and the frequency of polyploidy, and these in Castillejinae and whether the group was recovered as mono- results suggest a link between perenniality and polyploidy in phyletic in our analyses. In this section, each genus is discussed Castilleja. Although these two traits seem to be correlated, the with reference to the most recent reorganization of generic order in which they occur has been debated. Among herba- boundaries in Castillejinae ( Fig. 6 ; Chuang and Heckard, 1991 ) ceous perennials, polyploidy has been associated with taxa that and the evolution of morphological characteristics important for have effective means of vegetative reproduction and, further- the recognition of the major lineages recovered (sensu Scotland more, that perenniality and vegetative reproduction are the re- et al., 2003 ). sult of polyploidization ( Gustafsson, 1948 ). Alternatively, it has been suggested that polyploidization is easier in a peren- Orthocarpus— Orthocarpus sensu Chuang and Heckard nial background, owing to the possibility that perenniality and (1992b) represents a well-supported monophyletic group (Fig. effective vegetative reproduction may allow the negative ef- 3). Chuang and Heckard (1991) recognized that Orthocarpus fects of polyploidization (e.g., reproductive isolation, meiotic s.l. (Keck, 1927) comprised three distinct groups of species that irregularity) to be buffered for several generations ( Stebbins, did not necessarily belong together. Molecular data support this 1950 ). Neopolyploids commonly go through a period of re- assertion, resolving Orthocarpus s.s. as one of the basal lin- duced fertility, but fertility has been shown to increase rapidly eages of the subtribe, which is not particularly close to either in early generation polyploids (Ramsey and Schemske, 2002), Triphysaria or sect. Oncorhynchus of Castilleja ( Figs. 1 and 3 ). and the formation of tetraploids through triploid intermediates The nine species of Orthocarpus can be recognized by the mor- (triploid bridge) may be facilitated by perenniality (Ramsey phology of the lower corolla lip, which is only minutely three- and Schemske, 1998 ). toothed, and their unequally cleft calyx in which the dorsal cleft In Castilleja (which is primarily outcrossing), it is likely that is cut more deeply than that of the ventral calyx cleft ( Fig. 5 ; see perenniality has played an important role in the origin and es- Figs. 2 and 8 in Chuang and Heckard, 1991 ). 618 American Journal of Botany [Vol. 95

Fig. 4. Majority rule consensus tree (excluding burn-in trees) with mean branch lengths from the partitioned Bayesian analysis of the combined chlo- roplast and nuclear ribosomal DNA data. Branch lengths are proportional to the mean number of substitutions per site as measured by the scale bar. Black boxes indicate inferred polyploid changes in chromosome number. The distribution of the perennial habit and the base chromosome number for Castilleja are indicated with arrows. Lowercase letters following species names in Castilleja indicate the reference(s) for the chromosome numbers. (a) Heckard (1958), (b) Heckard (1968), (c) Reveal and Spellenberg (1976), (d) Heckard and Chuang (1977), (e) Pinkava et al. (1979), (f) Chuang and Heckard (1982), (g) Spellenberg (1986) , (h) Lockwood and Forstner (1991) , (i) Chuang and Heckard (1993) , (j) Chambers et al. (1998) . May 2008] Tank and Olmstead — Castillejinae Phylogeny 619

Castillejinae phylogeny suggests that the spathe-like calyx has arisen independently in the two lineages. Chuang and Heckard (1986) , following Pennell (1947) , hypothesized that the origin of the spathe-like calyx of Cordylanthus is an evolutionary modifi cation of the four-lobed calyx found throughout tribe Pedicularideae, where fusion of the two lateral lobes and the deepening of the abaxial cleft is followed by further fusion of the adaxial cleft to produce the one-parted, spathe-like calyx typical of most Cordylanthus species. Two species, Cordylan- thus orcuttianus of subg. Dicranostegia and C. capitatus of subg. Cordylanthus , have a deeply cleft calyx, which can be interpreted as the incomplete fusion of the adaxial cleft follow- ing the fusion of the lateral calyx lobes and deepening of the abaxial cleft ( Chuang and Heckard, 1975a , 1986 ). Cordylan- thus capitatus is resolved as the sister species to the remainder of Cordylanthus subg. Cordylanthus ( Fig. 3 ), and in addition to its bifi d calyx, it also has a defi nite tubular base, indicating that the deepening of the abaxial calyx cleft is not complete. Within subg. Cordylanthus , nearly all of the species have a calyx with a ± tubular base and a bifi d tip (sometimes only minutely so). Likewise, within the second Cordylanthus clade, C. orcuttianus (subg. Dicranostegia ) is resolved as the sistergroup to the Cor- dylanthus subg. Hemistegia clade (Figs. 1, 3, and 5). Along with its deeply bifi d calyx, C. orcuttianus also has a nearly tu- bular base (Chuang and Heckard, 1975a). The four species of subg. Hemistegia all have the derived, spathe-like calyx that lacks a tubular base, but is usually minutely cleft at the tip (Chuang and Heckard, 1973). Given the phylogenetic relationships within the two Cordylanthus clades presented here (Fig. 3), the evolutionary modifi cation of the calyx hypothesized by Chuang and Heckard (1986) is consistent with the distribution of calyx features within Cordylanthus , but would have occurred two Fig. 5. A summary of relationships among the major lineages of Cas- times independently. tillejinae. The tree is the same as that shown in Fig. 4 with the terminal The two Cordylanthus clades comprise two additional basal taxon names removed and the current taxonomic circumscription of the major clades and important monotypic lineages, specifi ed. Numbers on lineages of Castillejinae. Cordylanthus subg. Cordylanthus is branches indicate morphological synapomorphies:  unequal anther-sacs, resolved as the sister group to Orthocarpus in all analyses of the unequally attached;  corolla minutely 3-toothed; calyx unequally cleft, cpDNA data (Figs. 1 and 2) and the combined Bayesian and with deeper cut adaxially;  infl orescence a reduced spike ( < 2 cm), single- parsimony analyses (Figs. 3 and 4), or as the sister group to the fl owered fl orescence, or synfl orescence; tip of the middle lobe of lower remainder of Castillejinae in analyses of the nrDNA data ( Figs. corolla lip tightly revolute;  androecium reduced to 2 fertile stamens; 1 and 2 ) and the combined ML analysis (tree not shown); how-  plants halophytic; terminal stigmatic surface bent backward; middle lobe ever, neither of these relationships were well supported, and  of lower corolla lip erect; upper corolla lip (beak) open at tip; stigma there are no apparent morphological synapomorphies for the expanded; hilum terminal,;  stamens reduced to 1 anther sac; corolla monophyly of Orthocarpus and Cordylanthus subg. Cordylan- throat abruptly indented, forming a fold; n = 11; basic chromosome thus . Members of subg. Cordylanthus can be recognized by two number of x = 12; perennial habit. Two asterisks (**) denote the annual members of the perennial Castilleja clade, C. arvensis and C. exilis , as morphological synapomorphies. First, the architecture of the discussed in the text. infl orescence has undergone an evolutionary reduction from the basic spike common throughout the subtribe (and the majority of Orobanchaceae) to single-fl owered fl orescences and the sub- Cordylanthus— Cordylanthus (sensu Chuang and Heckard, sequent clustering of these into capitate or spike-like synfl ores- 1986) was not monophyletic in any phylogenetic analysis cences. Members of this clade possess infl orescences that are a ( Table 4 ), but rather was recovered in two distinct, well-supported reduced spike (< 2 cm), single-fl owered fl orescences, or capitate clades corresponding to Cordylanthus subg. Cordylanthus and or spike-like synfl orescences ( Fig. 5 ; see Fig. 2 in Chuang and a clade containing the monotypic Cordylanthus subg. Dicra- Heckard, 1986 ). The second morphological synapomorphy is a nostegia plus Cordylanthus subg. Hemistegia ( Figs. 1, 3, and 5 ). tightly revolute tip of the middle lobe of the lower corolla lip This result was surprising given that Cordylanthus has been ( Fig. 5 ). considered one of the most distinctive genera of Castillejinae, The second Cordylanthus clade, comprised of the monotypic due primarily to the presence of a unique calyx which is cleft subg. Dicranostegia and Cordylanthus subg. Hemistegia , was completely (or nearly so) to the base abaxially and fused adaxi- resolved unambiguously as the sister lineage to the remainder ally, forming a spathe-like structure that surrounds the corolla of Castillejinae (exclusive of subg. Cordylanthus and Orthocar- (Chuang and Heckard, 1973, 1986 ; see Fig. 1 in Chuang and pus) in all analyses (Figs. 1, 3, and 5), although there is no Heckard, 1991 and Fig. 5 in Chuang and Heckard, 1986). How- apparent morphological synapomorphy for this more inclusive ever, based on the results presented here, a parsimony recon- clade. Throughout Cordylanthus , there has been a tendency to- struction of this distinctive feature on our best estimate of ward a reduction of the androecium, putatively correlated with 620 American Journal of Botany [Vol. 95

trast, a reduced androecium in the clade containing subg. Dicranostegia and subg. Hemistegia can be interpreted as a morphological synapomorphy (Fig. 5). In the monotypic subg. Dicranostegia , the androecium of C. orcuttianus consists of two fertile stamens (the longer, anterior stamens), each with two anther sacs, while the posterior pair is reduced to fi laments bearing sterile appendages (Chuang and Heckard, 1975a, 1986 ). In the sister group of C. orcuttianus , Cordylanthus subg. Hemistegia , three of the four species possess two fertile ante- rior stamens (with two anther-sacs each) with the posterior pair of stamens reduced to two rudimentary fi laments (Chuang and Heckard, 1986). The posterior stamens of the fourth species, C. maritimus , each bear one fertile anther sac (rather than just a rudimentary fi lament), and this species is the only exception to the pattern of reduction that unites subg. Dicranostegia and subg. Hemistegia . Cordylanthus subg. Hemistegia is comprised of a special- ized group of species adapted to the coastal and inland saline habitats of western North America (Chuang and Heckard, 1973, 1986 ). The distinctive nature of this group of species has been recognized by numerous authors (e.g., Gray, 1867; Ferris, 1918; Chuang and Heckard, 1973) and has been de- scribed as the distinct genus Chloropyron Behr ( Behr, 1855 ). Subgenus Hemistegia is further distinguished from Cordylan- thus subg. Cordylanthus by having the typical infl orescence type found throughout subtribe Castillejinae, an elongate spike with only one fl oral bract associated with each fl ower (in con- trast to the synfl orescence described earlier; Chuang and Heckard, 1973). In addition to the halophytic nature of this group, Cordylanthus subg. Hemistegia is marked by the following morphological synapomorphies: (1) the terminal stigmatic sur- face is bent backward at maturity, and (2) the middle lobe of the lower corolla lip is erect ( Fig. 5 ; Chuang and Heckard, 1973, 1986 ).

Triphysaria— The elevation of Orthocarpus subg. Triphys- aria (sensu Keck, 1927) to generic status (as Triphysaria ) was Fig. 6. “ Intuitive phylogeny ” of subtribe Castillejinae modifi ed from based on chromosome number, seed and seed-coat morphol- Fig. 34 in Chuang and Heckard (1991). Branches are shaded following ogy, fl oral morphology, and experimental hybridization studies Chuang and Heckard’ s interpretation of the evolution of the annual and ( Chuang and Heckard, 1991 ). In the phylogenetic analyses pre- perennial habits in Castillejinae. sented here, Triphysaria was recovered as a well-supported lin- eage ( Figs. 1 and 3 ) and the sister group of Castilleja s.l. The monophyly of Triphysaria and Castilleja s.l. is supported by a increased specialization in pollination ( Chuang and Heckard, number of morphological synapomorphies (Fig. 5). In both Or- 1986 ). This characteristic is most pronounced in the clade contain- thocarpus and Cordylanthus the galea (formed by the fusion of ing Dicranostegia and Hemistegia ( Fig. 5 ), where the androecium the two petals comprising the upper corolla lip of the zygomor- is reduced to two fertile stamens. The majority of the 13 species phic corolla) is closed at the tip, being the product of the two of subg. Cordylanthus possess four didynamous stamens, each fused corolla lobes that are folded downward forming a true with two anther sacs (the upper, larger anther sac attached me- galea, or hood (Chuang and Heckard, 1991). In both Triphys- dially and the lower, smaller anther sac attached apically, as in aria and Castilleja s.l. the upper corolla lip is open at the tip, most members of Castillejinae; Fig. 5). In subg. Cordylanthus , and there is no true galea ( Fig. 5 ; see Figs. 7 – 12 in Chuang and the four fertile stamens of C. nevinii A. Gray and C. laxifl orus Heckard, 1991). Chuang and Heckard (1991) noted that the use A. Gray (neither sampled in this study) have been reduced to of the term galea throughout the Castilleja literature is in error only the upper anther sac, and C. capitatus , which is sister to and instead promoted the use of the term beaked to describe this the rest of the clade, has been reduced to only two fertile sta- condition. Triphysaria and Castilleja s.l. also share the derived mens, each possessing only the upper anther sac (Chuang and characteristics of an expanded stigma that is either capitate or Heckard, 1986). The remaining 10 species of subg. Cordylan- bilobed ( Fig. 5 ; the stigma is unexpanded in both Cordylanthus thus have undergone no reduction of the androecium. Given the and Orthocarpus) and a terminal attachment of the ovule to the sister-group relationship of C. capitatus and the remainder of placenta (Fig. 5; Cordylanthus and Orthocarpus both have a subg. Cordylanthus (Fig. 3), the reduction of the androecium in lateral hilum). those three species may represent a synapomorphy for a sub- Morphologically, the species of Triphysaria are very similar group or autapomorphies for individual taxa (if C. nevinii , to the annual species of Castilleja subg. Colacus sect. Onco- C. laxifl orus , and C. capitatus are not monophyletic). In con- rhynchus (also removed from Orthocarpus s.l. by Chuang and May 2008] Tank and Olmstead — Castillejinae Phylogeny 621

Heckard, 1991). However, a number of the characteristics used pomorphy for the Castilleja s.l. clade, this group is cytologi- by Chuang and Heckard (1991) as justifi cation for the elevation cally distinct from the remainder of the subtribe with a basic of Triphysaria to generic status provide morphological synapo- haploid chromosome number of x = 12 ( Fig. 5 ). morphies for the genus, given the relationships presented here (Fig. 5). In all fi ve species of Triphysaria, the stamens are re- Clevelandia & Ophiocephalus — Clevelandia and Ophio- duced to a single anther sac, whereas the sister group (Castilleja cephalus have been maintained as distinct genera in Castilleji- s.l.) have no such reduction in the androecium. With the excep- nae despite numerous similarities to Castilleja , including tion of the tiny (4 – 6 mm) corolla of T. pusilla , the throat of the chromosome number (n = 12), vegetative morphology, seed corolla is abruptly indented, forming a distinct fold below the coat morphology, and the ability to make fertile hybrids in ex- lower corolla lip, distinguishing Triphysaria morphologically perimental crosses (only Ophiocephalus was tested for cross from the annual species of Castilleja sect. Oncorhynchus (see compatibility; Chuang and Heckard, 1991 ). Both of these gen- Figs. 9 – 11 in Chuang and Heckard, 1991 ). Lastly, Triphysaria era possess unique corolla morphologies that formed the pri- are cytologically unique among members of Castillejinae with mary justifi cation of their status as separate genera ( Chuang a chromosome number of n = 11 ( Fig. 5 ). and Heckard, 1991 ). Clevelandia has a small corolla (1 – 2 cm) in which the lobes of the upper corolla lip are not fused, and Castilleja— Despite extensive taxonomic study of this iconic the lower corolla lip has three triangular, spreading lobes that group of western North American wildfl owers, infrageneric are wider than the upper corolla lobes, forming a slightly zygo- classifi cation in Castilleja has been diffi cult. Most efforts to morphic corolla (Greene, 1885, 1886 ). Likewise, the infl ated classify this group ( Bentham, 1846 ; Gray, 1862 ; Wettstein, corolla of Ophiocephalus, with its long, exserted stamens, is 1891 ; Eastwood, 1909 ; Rydberg, 1917 ; Pennell, 1935 , 1951 ; only slightly bilabiate; however, the upper corolla lip is fused Ownbey, 1959 ; Holmgren, 1970 , 1971 , 1976 , 1984 ; Nesom, into a beak, as in Castilleja (Wiggins, 1933). Field observa- 1992a– c , 1994 ) have been ignored largely due to their failure to tions of a population of Ophiocephalus angustifolius suggest delimit cohesive units within Castilleja or to their narrow use of that the distinct fl oral morphology of this species is likely the small or monotypic infrageneric groups. Chuang and Heckard result of increased specialization to pollinators (J. M. Egger (1991) proposed a new infrageneric classifi cation for Castilleja [WTU], personal communication), and we postulate the same including the three subgenera, Castilleja , Colacus , and Gen- is true for the unusually small, open corolla of Clevelandia trya. The majority of Castilleja species were included in subg. beldingiii . Castilleja (~150 mostly perennial species), which includes spe- The sister-group relationship of Clevelandia and Ophioceph- cies with fl owers primarily modifi ed for hummingbird pollina- alus received a signifi cant Bayesian PP value (0.99) but low tion. Subgenus Colacus includes the majority of Castilleja ML and parsimony bootstrap support (50% and 41%, respectively) species that have fl owers modifi ed for insect pollination. Cola- and, therefore, may not represent the true relationship of these cus includes two sections of perennials that have been assigned two species ( Fig. 3 ). Based on the combined analyses of the historically to Orthocarpus and Castilleja (sects. Pilosae and cpDNA and nrDNA data (Fig. 3), these two annual genera are Pallescentes) and all of the annual Castilleja species (sect. more closely related to perennial Castilleja than they are to any Oncorhynchus ) previously assigned to Orthocarpus by Keck of the annual species transferred to Castilleja from Orthocarpus (1927) . Subgenus Gentrya comprises the single species Cas- (sensu Keck, 1927) and may represent the sister lineage (or suc- tilleja racemosa (= G. racemosa Breedlove & Heckard). cessive sister lineages) to the perennial Castilleja clade. Castilleja was not monophyletic in any of the separate or combined analyses of the cpDNA and nrDNA regions because Castilleja subgenus Colacus section Oncorhynchus— the two monotypic annual genera Clevelandia and Ophioceph- Chuang and Heckard (1991) were correct in their interpretation alus are derived from within the mostly perennial genus Cas- that the two former sections of the annual genus Orthocarpus tilleja ( Table 4 ). The clade containing Castilleja , Clevelandia , (sects. Castillejoides and Cordylanthoides, sensu Keck, 1927) and Ophiocephalus was recovered by separate and combined belonged with Castilleja , where they were placed in subg. analyses of the cpDNA and nrDNA data (Figs. 1 and 3). The Colacus sect. Oncorhynchus . However, these annuals form a only potential morphological synapomorphy for this clade is basal grade within the Castilleja s.l. clade (Fig. 5), rather than the presence of a loose-fi tting seed coat (i.e., the reticulate outer the monophyletic group hypothesized by Chuang and Heckard seed coat forms a network that encloses a freely suspended ( Fig. 6 ; Table 4 ). The majority of the annual species were re- body made up of the inner seed coat, endosperm, and embryo; covered in two well-supported clades (Fig. 3), which corre- see Figs. 25 – 49 in Chuang and Heckard, 1983 ). Morphological spond to the two sections of Keck (1927) ; the clade containing characteristics of the seed coat have been used extensively Castilleja lineariloba and C. attenuata is equivalent to sect. for infrageneric classifi cation throughout Castillejinae (e.g., Castillejoides (with the exception of Castilleja rubicundula), Chuang and Heckard, 1972, 1983 , 1992b ) and as one of the while the clade containing C. campestris and C. tenuis is major characteristics separating Cordylanthus , Orthocarpus , and equivalent to sect. Cordylanthoides . Castilleja exserta subsp. Triphysaria from the annual species of Castilleja ( Chuang and exserta (formerly of sect. Castillejoides) represents an addi- Heckard, 1991). However, of the fi ve species of Triphysaria , tional independent lineage comprising the grade of annual two species, Triphysaria micrantha and T. eriantha ( Fig. 3 ), have Castilleja species approaching the remainder of the Castilleja a loose-fi tting seed coat like Castilleja ( Chuang and Heckard, s.l. clade ( Fig. 3 ). 1991). Therefore, without knowing the underlying devel- opmental mechanism for this morphological change, whether a Perennial Castilleja and subgenus Gentrya— The majority tight-fi tting seed coat was a reversal in one lineage of Triphys- of the Castilleja species sampled here are included in the large aria or the loose-fi tting seed coat arose independently in Cas- perennial Castilleja clade (Figs. 3 and 5). Because of the lack of tilleja and the two species of Triphysaria is equivocal ( Figs. 3 resolution with the cpDNA and nrDNA regions used in this and 5 ). Although there is no unequivocal morphological syna- study, few conclusions regarding the relationships of perennial 622 American Journal of Botany [Vol. 95

Castilleja species can be drawn. Perennial Castilleja are a have undergone a reduction to only one anther sac per stamen morphologically complex group with numerous infrageneric (e.g., Triphysaria and some species of Cordylanthus ), members assemblages and intraspecifi c taxa in which the taxonomic of Castillejinae are characterized by having anther sacs that are complexity is extremely challenging. This complexity stems unequal in size and unequally attached ( Fig. 5 ). from (1) complex morphological variation often attributed to (2) Orthocarpus and Triphysaria are monophyletic and rep- the formation of polyploid complexes (Heckard, 1968; Heckard resent two separate evolutionary lineages within Castillejinae. and Chuang, 1977 ) and from (2) the diffi culty in circumscribing Castilleja subg. Colacus sect. Oncorhynchus, also previously taxa due to overlapping variation in nearly every character recognized as Orthocarpus , is not monophyletic, but rather, (Holmgren, 1984). These two sources of confusion in Castilleja forms a basal grade of annual species within Castilleja s.l. taxonomy are not mutually exclusive. More than half of the pe- Although all three of these groups were previously included in rennial species are known to have polyploid populations, with Orthocarpus (sensu Keck, 1927) as suggested by Chuang and ploidy levels ranging from 4 x to 12 x ( Heckard, 1968 ; Heckard Heckard (1991) , they are more closely related to other members and Chuang, 1977 ), and polyploidy in Castilleja is often attrib- of the subtribe than they are to each other. uted to hybridization (Heckard and Chuang, 1977). Apparent (3) Cordylanthus is biphyletic; subg. Cordylanthus and subg. hybrid swarms and morphological intergradation are common Hemistegia are each monophyletic, and the monotypic subg. between populations of some species, especially those belong- Dicranostegia is sister to subg. Hemistegia . All these clades ing to polyploid complexes (Heckard, 1968; Heckard and can be recognized by morphological synapomorphies ( Fig. 5 ). Chuang, 1977 ; Chuang and Heckard, 1993 ; Hersch and Roy, (4) Castilleja is not monophyletic because the two mono- 2007 ). In experimental hybridization studies, hybrids of vary- typic genera Clevelandia and Ophiocephalus are derived from ing degrees of fertility were obtained from a wide variety of within this clade ( Castilleja s.l.). The majority of Castilleja interspecifi c crosses, including those between parents with dif- species are perennial, and these taxa, along with Clevelandia ferent ploidy levels ( Heckard, 1968 ; Chuang and Heckard, and Ophiocephalus , form a well-supported clade derived from 1991 ). Thus, interspecifi c hybridization may be responsible for the grade of annual taxa in subg. Colacus sect. Oncorhynchus . much of the conspicuous overlapping variation observed be- This clade includes Castilleja racemosa , which was formerly tween many Castilleja species (Holmgren, 1984; Chuang and isolated as the monotypic genus Gentrya . Heckard, 1991). It is clear that to investigate interspecifi c rela- (5) Based on the well-supported phylogenetic hypothesis for tionships within the perennial Castilleja clade, it will be neces- Castillejinae presented here, it will be necessary to revise the sary to generate a much larger cpDNA data set or sequence current circumscription of generic boundaries. In the formal DNA regions that are more variable than cpDNA or nrDNA treatment that will be presented elsewhere, Orthocarpus and (i.e., single or low-copy nuclear genes). Triphysaria will retain their current circumscriptions, Castilleja In their realignment of generic boundaries, Chuang and Heckard will be expanded to include the two monotypic genera Cleve- (1991) included the monotypic Mexican genus Gentrya in Cas- landia and Ophiocephalus , and Cordylanthus will be split into tilleja (as C. racemosa), creating a new monotypic subgenus three genera corresponding to the three previously recognized ( Fig. 6 ). In its original description, the corolla of Gentrya race- subgenera. mosa was described as “ more truly galeate ” than Castilleja be- cause of its curved, helmet-shaped upper corolla lip, rather than Conclusions — The results of this study place the generic re- the straight, beak-like upper corolla lip of Castilleja (Breedlove alignment of Castillejinae presented by Chuang and Heckard and Heckard, 1970, p. 23 ). This trait formed the basis for its (1991) in the context of a robust phylogenetic hypothesis for all status as a distinct genus in Castillejinae, despite the many other of the major lineages comprising the subtribe. In addition to noted similarities to Castilleja . However, Chuang and Heckard providing a basis for the systematics of this group, we were able (1991) felt that this distinction alone did not warrant generic sta- to interpret morphological characters that are important for rec- tus, given their expanded view of Castilleja . In their intuitive ognizing major lineages (sensu Scotland et al., 2003 ) and to view of the phylogenetic relationships of the major lineages of draw inferences regarding the evolution of the perennial habit. Castillejinae, Chuang and Heckard (1991) placed Castilleja subg. The observation of a seemingly tight association between pe- Gentrya as the sister lineage to the remainder of Castilleja and renniality and polyploidy in Castilleja sets the stage for future viewed the species as an isolated lineage in Castilleja ( Fig. 6 ). research focused on the detection of correlates between organ- Our phylogenetic analyses place C. racemosa sister to a group of ismal traits (e.g., chromosomal change and shifts in life history) other Mexican and Central Amer ican Castilleja species ( Fig. 3 ) and shifts in the rate of diversifi cation. The restriction of hum- within the perennial Castilleja clade ( Fig. 5 ). This relationship mingbird pollination to the perennial clade of Castilleja sug- received signifi cant Bayesian PP values (0.98 and 1.0), but rela- gests another possible association that may refl ect underlying tively low ML (66% and 86%) and parsimony (48% and 84%) constraints of perenniality. Thus, the fi ndings presented here bootstrap support ( Fig. 3 ); therefore, the precise relationship of serve as valuable background information for future research this species to the remainder of Castilleja s.l. is still tentative. not only in Castillejinae and the large, complex genus Castilleja , Nevertheless, C. racemosa is a member of the clade containing but also for investigating commonalities that may have been im- perennial Castilleja , Clevelandia , and Ophiocephalus (Fig. 5), portant for plant diversifi cation in western North America. and not an isolated lineage of Castillejinae (e.g., Breedlove and Heckard, 1970 ) or Castilleja (e.g., Chuang and Heckard, 1991 ) as previously hypothesized. LITERATURE CITED Ames , O. 1939 . Economic annuals and human cultures. Botanical Museum Systematic conclusions— (1) Subtribe Castillejinae is mono- of Harvard University, Cambridge, Massachusetts, USA. phyletic and includes the six genera traditionally recognized as Andreasen , K . , and B . G . Baldwin . 2001 . Unequal evolutionary members of this group (sensu Chuang and Heckard, 1991 ). rates between annual and perennial lineages of checker mallows With the exception of Ophiocephalus and those species that (Sidalcea, Malvaceae): Evidence from 18S-26S rDNA internal and May 2008] Tank and Olmstead — Castillejinae Phylogeny 623

external transcribed spacers. Molecular Biology and Evolution 1 8 : Chuang , T . I . , and L . R . Heckard . 1983 . Systematic signifi cance of 936 – 944 . seed-surface features in Orthocarpus (Scrophulariaceae subtribe Axelrod , D. I . 1979 . Age and origin of the Sonoran desert vegetation. Castillejinae). American Journal of Botany 70 : 877 – 890 . Occasional Papers of the California Academy of Sciences 132: Chuang , T . I . , and L . R . Heckard . 1986 . Systematics and evolution of 1 – 74. Cordylanthus (Scrophulariaceae-Pedicularieae). Systematic Botany Baldwin , B . G. 1992 . Phylogenetic utility of the internal transcribed Monographs 10 : 1 – 105 . spacers of nuclear ribosomal DNA in plants: An example from the Chuang , T . I . , and L . R . Heckard . 1991 . Generic realignment and syn- Compositae. Molecular Phylogenetics and Evolution 1 : 3 – 16 . opsis of subtribe Castillejinae (Scrophulariaceae— tribe Pediculareae). Baldwin , B . G . , and S . Markos . 1998 . Phylogenetic utility of the exter- Systematic Botany 16 : 644 – 666 . nal transcribed spacer (ETS) of 18S-26S rDNA: Congruence of ETS Chuang , T . I . , and L . R . Heckard . 1992a . New species of bee- pollinated and ITS trees of Calycadenia (Compositae). Molecular Phylogenetics Castilleja from Peru, with a taxonomic revision of South American and Evolution 10 : 449 – 463 . members of subg. Colacus. Systematic Botany 17 : 417 – 431 . Baldwin , B. G. , M. J. Sanderson , J . M . Porter , M . F . Wojciechowski , Chuang , T . I . , and L . R . Heckard . 1992b . A taxonomic revision of C. S. Campbell , and M. J. Donoghue. 1995 . The ITS region of Orthocarpus (Scrophulariaceae — tribe Pediculareae). Systematic nuclear ribosomal DNA: A valuable source of evidence on angio- Botany 17 : 560 – 582 . sperm phylogeny. Annals of the Missouri Botanical Garden 8 2 : Chuang , T . I . , and L . R . Heckard . 1993 . Chromosome numbers of neo- 247 – 277 . tropical Castilleja (Scrophulariaceae, tribe Pediculareae) and their Barker , F . K . , and F . M . Lutzoni . 2002 . The utility of the incongruence taxonomic implications. Annals of the Missouri Botanical Garden 80 : length difference test. Systematic Biology 51 : 625 – 637 . 974 – 986 . Beardsley , P . M . , and R . G . Olmstead . 2002 . Redefi ning Phrymaceae: Conti , E . , D . E . Soltis , T . M . Hardig , and J . Schneider . The placement of Mimulus , tribe Mimuleae and Phryma. American 1999 . Phylogenetic relationships of the silver saxifrages (Saxifraga , Journal of Botany 89 : 1093 – 1102 . sect. Ligulatae Haworth): Implications for the evolution of substrate Behr , H . H. 1855 . A new genus and species, Chloropyron palustre. specifi city, life histories, and biogeography. Molecular Phylogenetics Proceedings of the California Academy of Sciences 1 : 62 . and Evolution 13 : 536 – 555 . Bena , G. , B. Lejeune , J.-M. Prosperi , and I. Olivieri . 1998 . Molecular Cunningham , C. 1997 . Can three incongruence tests predict when phylogenetic approach for studying life-history evolution: The am- data should be combined? Molecular Biology and Evolution 14 : biguous example of the genus Medicago L. Proceedings of the Royal 733 – 740 . Society of London, B, Biological Sciences 265 : 1141 – 1151 . Darlu , P . , and G . Lecointre . 2002 . When does the incongruence length Bennett , J . R . , and S . Mathews . 2006 . Phylogeny of the parasitic difference test fail? Molecular Biology and Evolution 19 : 432 – 437 . DeBry R W and R G Olmstead plant family Orobanchaceae inferred from phytochrome A. American , . . , . . . 2000 . A simulation study of re- duced tree-search effort in bootstrap resampling analysis. Systematic Journal of Botany 93 : 1039 – 1051 . Biology 49 : 171 – 179 . Bentham , G . 1846 . Scrophulariaceae. In A. de Candolle [ed.], Prodromus Doyle , J . J . , and J. L. Doyle . 1987 . A rapid DNA isolation procedure systematis naturalis regni vegetabilis, vol. 10, 180– 586. Victoris for small quantities of fresh leaf tissue. Phytochemical Bulletin Masson, Paris, France. 19 : 11 – 15 . Brandley , M . C . , A . Schmitz , and T . W . Reeder . 2005 . Partitioned Eastwood , A . 1909 . Synopsis of the Mexican and Central American spe- Bayesian analyses, partition choice, and the phylogenetic relation- cies of Castilleja. Proceedings of the American Academy of Arts 44 : ships of scincid lizards. Systematic Biology 54 : 373 – 390 . 563 – 591 . Breedlove , D . E . , and L . R . Heckard . 1970 . Gentrya , a new genus of Farris , J . S . , M . K ä llersj ö , A . G . Kluge , and C . Bult . 1994 . Testing Scrophulariaceae from Mexico. Brittonia 22 : 20 – 24 . signifi cance of incongruence. Cladistics 10 : 315 – 319 . Brown , J . M . , and A . R . Lemmon . 2007 . The importance of data par- Felsenstein , J . 1985 . Confi dence limits on phylogenies — an approach titioning and the utility of Bayes factors in Bayesian phylogenetics. using the bootstrap. Evolution 39 : 783 – 791 . Systematic Biology 56 : 643 – 655 . Ferris , R . S . 1918 . Taxonomy and distribution of Adenostegia. Bulletin of Bull , J . J . , J . P . Huelsenbeck , C . W . Cunningham , D . L . Swofford , the Torrey Botanical Club 45 : 399 – 423 . and P . J . Waddell. 1993 . Partitioning and combining data in phylo- Gielly , L . , and P . Taberlet. 1994 . The use of chloroplast DNA to genetic analysis. Systematic Biology 42 : 384 – 397 . resolve plant phylogenies— Noncoding versus rbcL sequences. Chambers , K . L . , D . Green , S . Potampa , and L . McMahan . 1998 . Molecular Biology and Evolution 11 : 769 – 777 . IOPB chromosome data 13. International Organization of Plant Gleason , H . A . , and A . Cronquist . 1964 . The natural geography of Biosystematists Newsletter 29 : 18 – 22 . plants. Columbia University Press, New York, New York, USA. Chuang , T . I . 1993 . A cladistic analysis of the subtribe Castillejinae Graham , S . W . , J . R . Kohn , B . R . Morton , J . E . Eckenwalder , and (Scrophulariaceae-tribe Pediculareae). Botanical Bulletin of Academia S . C . H . Barrett . 1998 . Phylogenetic congruence and discor- Sinica 34 : 155 – 161 . dance among one morphological and three molecular data sets from Chuang , T . I . , and L . R . Heckard . 1972 . Seed coat morphology in Pontederiaceae. Systematic Biology 47 : 545 – 567 . Cordylanthus (Scrophulariaceae) and its taxonomic signifi cance. Graham , S . W . , P . A . Reeves , A . C . E . Burns , and R . G . Olmstead . American Journal of Botany 59 : 258 – 265 . 2000 . Microstructural changes in noncoding chloroplast DNA: Chuang , T . I . , and L . R . Heckard . 1973 . Taxonomy of Cordylanthus Interpretation, evolution, and utility of indels and inversions in basal subgenus Hemistegia (Scrophulariaceae). Brittonia 25 : 135 – 158 . angiosperm phylogenetic inference. International Journal of Plant Chuang , T . I . , and L . R . Heckard . 1975a . Taxonomic status of Sciences 161 ( 6 Supplement ): S83 – S96 . Cordylanthus (subg. Dicranostegia ) orcuttianus (Scrophulariaceae). Gray , A . 1862 . Revision of the genus Castilleja. American Journal of Madro ñ o 23 : 88 – 95 . Science II 34 : 335 – 339 . Chuang , T . I . , and L . R . Heckard . 1975b . Re-evaluation of bract mor- Gray , A . 1867 . Characters of new plants of California and elsewhere, phology in taxonomy of Cordylanthus (Scrophulariaceae). Madro ñ o principally of those collected by H. N. Bolander in the state geological 23 : 169 – 173 . survey. Proceedings of the American Academy of Arts 7 : 327 – 402 . Chuang , T . I . , and L . R . Heckard . 1976 . Morphology, evolution, Greene , E . L. 1885 . Studies in the botany of California and parts adja- and taxonomic signifi cance of the infl orescence in Cordylanthus cent. Bulletin of the California Academy of Science 1 : 66 – 155 . (Scrophulariaceae). American Journal of Botany 63 : 272 – 282 . Greene , E . L . 1886 . Studies in the botany of California and parts adjacent Chuang , T . I . , and L . R . Heckard . 1982 . Chromosome numbers of II. Bulletin of the California Academy of Science 4 : 179 – 228 . Orthocarpus and related monotypic genera (Scrophulariaceae, sub- Gustafsson , Å . 1948 . Polyploidy, life-form, and vegetative reproduction. tribe Castillejinae). Brittonia 34 : 89 – 101 . Hereditas 34 : 1 – 22 . 624 American Journal of Botany [Vol. 95

Heckard , L . R . 1958 . Documented chromosome numbers in plants. Pennell , F . W . 1951 . Scrophulariaceae. In L. Abrams [ed.], Illustrated Madro ñ o 14 : 237 . fl ora of the Pacifi c states, 686 – 859. Stanford University Press, Heckard , L . R. 1968 . Chromosome numbers and polyploidy in Castilleja Stanford, California, USA. (Scrophulariaceae). Brittonia 20 : 212 – 226 . Pinkava , D . J . , T . Reeves , E . Lehto , and L . A . McGill . 1979 . Plants Heckard , L . R . , and T . I . Chuang . 1977 . Chromosome numbers, poly- new to Arizona fl ora. VII. Chromosome counts and new distribu- ploidy, and hybridization in Castilleja (Scrophulariaceae) of the Great tion records of noteworthy species. Journal of the Arizona– Nevada Basin and Rocky Mountains. Brittonia 29 : 159 – 172 . Academy of Science 14 : 37 . Heckard , L . R . , M . I . Morris , and T . I . Chuang . 1980 . Origin and Posada , D . , and K . A . Crandall . 1998 . MODELTEST: Testing the model taxonomy of Castilleja montigena (Scrophulariaceae). Systematic of DNA substitution. Bioinformatics (Oxford, England) 14 : 817 – 818 . Botany 5 : 71 – 85 . Rambaut , A. 1996 . Se-Al: Sequence Alignment Editor. Available at Hersch , E . I . , and B . A . Roy . 2007 . Context-dependent pollinator behav- http://tree.bio.ed.ac.uk/software/ . Institute of Evolutionary Biology, ior: An explanation for patterns of hybridization among three species Edinburgh, UK. of indian paintbrush. Evolution 61 : 111 – 124 . Rambaut , A. , and A. J. Drummond . 2004 . Tracer. Available at Holmgren , N . H . 1970 . Castilleja. In D. S. Corell and M. C. Johnson http:// tree.bio.ed.ac.uk/software/ . Institute of Evolutionary Biology, [eds.], Manual of the vascular plants of Texas, 1439– 1442. Texas Edinburgh, UK. Research Foundation, Renner, Texas, USA. Ramsey , J . , and D . W . Schemske . 1998 . Pathways, mechanisms, and Holmgren , N . H. 1971 . A taxonomic revision of the Castilleja viscidula rates of polyploid formation in fl owering plants. Annual Review of group. Memoirs of the New York Botanical Garden 21 : 1 – 63 . Ecology and Systematics 29 : 467 – 501 . Holmgren , N . H . 1976 . Four new species of Mexican Castilleja (sub- Ramsey , J . , and D . W . Schemske . 2002 . Neopolyploidy in fl owering genus Castilleja , Scrophulariaceae) and their relatives. Brittonia 28 : plants. Annual Review of Ecology and Systematics 33 : 589 – 639 . 195 – 208 . Ronquist , F . , and J . P . Huelsenbeck . 2003 . MrBayes 3: Bayesian phy- Holmgren , N . H . 1984 . Scrophulariaceae. In A. Cronquist, A. H. logenetic inference under mixed models. Bioinformatics (Oxford, Holmgren, N. H. Holmgren, J. L. Reveal, and P. K. Holmgren [eds.], England) 19 : 1572 – 1574 . Intermountain fl ora, 344 – 506. New York Botanical Garden, Bronx, Reveal , J. L. , and R. Spellenberg . 1976 . Miscellaneous chromosome New York, USA. counts of western American plants III. Rhodora 78 : 37 – 52 . Jeffrey , E . C. 1916 . The anatomy of woody plants. University Press, Rydberg , P . A. 1917 . Flora of the Rocky Mountains and adjacent plains. Chicago, Illinois, USA. Published by the author, New York, New York, USA. Jepson , J . L . 1925 . Manual of fl owering plants of California. University Scotland , R . W . , R . G . Olmstead , and J . R . Bennett . 2003 . Phylogeny of California Press, Berkeley, California, USA. reconstruction: The role of morphology. Systematic Biology 52 : Johnson , A . W. 1968 . The evolution of desert vegetation in western 539 – 548 . Shimodaira , H . , and M . Hasegawa . 1999 . Multiple comparisons of log- North America. In G. W. J. Brown [ed.], Desert BIOLOGY, 101– 140. likelihoods with applications to phylogenetic inference. Molecular Academic Press, New York, New York, USA. Biology and Evolution 16 : 1114 – 1116 . Kass , R . , and A . Raftery . 1995 . Bayes factors. Journal of the American Simmons , M . P . , and H . Ochoterena . 2000 . Gaps as characters in Statistical Association 90 : 430 – 773 . sequence-based phylogenetic analyses. Systematic Biology 49 : Keck , D . D. 1927 . A revision of the genus Orthocarpus. Proceedings of 369 – 381 . the California Academy of Science Series 4 16: 517-571. Soltis , D . E . , and P . S . Soltis . 1993 . Molecular data and the dynamic Lockwood , M . , and M . Forstner . 1991 . IOPB chromosome data nature of polyploidy. Critical Reviews in Plant Sciences 12 : 243 – 273 . 3. International Organization of Plant Biosystematists Newsletter Soltis , D. E. , P. S. Soltis , and J. A. Tate. 2003 . Advances in the 17 : 10 . study of polyploidy since Plant speciation. New Phytologist 161 : Nesom , G . L . 1992a . Castilleja section Euchroma (Scrophulariaceae) in 173 – 191 . Mexico: New species and comments on other taxa. Phytologia 73 : Spellenberg , R . 1986 . Chromosome number reports 90. Taxon 35 : 197 . 384 – 388 . Stearns , S . C. 1992 . The evolution of life histories. Oxford University Nesom , G . L . 1992b . New species and taxonomic evaluations of Mexican Press, Oxford, UK. Castilleja (Scrophulariaceae). Phytologia 72 : 231 – 252 . Stebbins , G . L . 1938 . Cytological characteristics associated with differ- Nesom , G . L . 1992c . Taxonomy of the Castilleja tenuifl ora group ent growth habits in the dicotyledons. American Journal of Botany (Scrophulariaceae) in Mexico, with an overview of sect. Castilleja. 25 : 189 – 198 . Phytologia 73 : 389 – 415 . Stebbins , G . L. 1950 . Variation and evolution in plants. Columbia Nesom , G . L . 1994 . A new species of Castilleja (Scrophulariaceae) from University Press, New York, New York, USA. Chihuahua, Mexico. Phytologia 76 : 406 – 409 . Stebbins , G . L . 1957 . Self fertilization and population variability in the Nylander , J. A. A. , F. Ronquist , J. P. Huelsenbeck , and J. L . higher plants. American Naturalist 91 : 337 – 354 . Nieves- Aldrey. 2004 . Bayesian phylogenetic analysis of com- Swofford , D . L . 2002 . PAUP*: Phylogenetic analysis using parsi- bined data. Systematic Biology 53 : 47 – 67 . mony (*and other methods), version 4b10. Sinauer, Sunderland, Olmstead , R . G . , C . W . dePamphilis , A . D . Wolfe , N . D . Young , Massachusetts, USA. W . J . Elisons , and P . A . Reeves . 2001 . Disintegration of the Swofford , D . L . , G . J . Olsen , P . J . Waddell , and D . M . Hillis . Scrophulariaceae. American Journal of Botany 88 : 348 – 361 . 1996 . Phylogenetic inference. In D. M. Hillis, C. Moritz, and B. K. Ownbey , M . 1959 . Castilleja. In C. L. Hitchcock, A. Cronquist, M. Mable [eds.], Molecular systematics, 407– 514. Sinauer, Sunderland, Ownbey, and J. W. Thompson [eds.], Vascular plants of the Pacifi c Massachusetts, USA. Northwest, 295– 326. University of Washington Press, Seattle, Taberlet , P . , L . Gielly , G . Pautou , and J . Bouvet . 1991 . Universal Washington, USA. primers for amplifi cation of three non-coding regions of chloroplast Oxelman , B . , M . Liden , and D . Berglund . 1997 . Chloroplast rps16 DNA. Plant Molecular Biology 17 : 1105 – 1109 . intron phylogeny of the tribe Sileneae (Caryophyllaceae). Plant Tank , D . C . , P . M . Beardsley , S . A . Kelchner , and R . G . Olmstead . Systematics and Evolution 206 : 393 – 410 . 2006 . L. A. S. Johnson review no. 7. Review of the systematics of Pennell , F . W. 1935 . The Scrophulariaceae of eastern temperate North Scrophulariaceae s.l. and its current disposition. Australian Systematic America. Academy of Natural Sciences of Philadelphia Monographs Botany 19 : 289 – 307 . 1 : 1 – 650 . Verboom , G . A . , H . P . Linder , and W . D . Stock . 2004 . Testing the Pennell , F . W . 1947 . Some hitherto undescribed Scrophulariaceae of adaptive nature of radiation: Growth form and life history divergence the Pacifi c states. Proceedings of the Academy of Natural Sciences of in the African genus Ehrharta (Poaceae: Ehrhartoideae). American Philadelphia 99 : 151 – 171 . Journal of Botany 91 : 1364 – 1370 . May 2008] Tank and Olmstead — Castillejinae Phylogeny 625

Watson , S. 1871 . United States geological exploration of the fortieth Yoder , A . D . , J . A . Irwin , and B . A . Payseur . 2001 . Failure of the ILD parallel, Clarence King, Geologist-in-Charge. Government Printing to determine data combinability for slow loris phylogeny. Systematic Offi ce, Washington, D.C., USA. Biology 50 : 408 – 424 . Wettstein , R. 1891 . Scrophulariaceae. In A. Engler and K. Prantl [eds.], Die Young , N . D . , K . E . Steiner , and C . W . dePamphilis . 1999 . The natü rlichen Pfl anzenfamilien, 39– 107. Engelmann, Leipzig, Germany. evolution of parasitism in Scrophulariaceae/Orobanchaceae: plastid Wiggins , I . L . 1933 . New plants from Baja California. Contributions gene sequences refute an evolutionary transition series. Annals of the from the Dudley Herbarium 1 : 161 – 187 . Missouri Botanical Garden 86 : 876 – 893 . Wolfe , A . D . , C . P . Randle , L . Liu , and K . E . Steiner . 2005 . Phylogeny Young , T . P . , and C . K . Augspurger . 1991 . Ecology and evolution and biogeography of Orobanchaceae. Folia Geobotanica 4 0 : of long-lived semelparous plants. Trends in Ecology & Evolution 6 : 115 – 134 . 285 – 289 .