Tribal Relationships in the : Evidence from DNA Sequences of the Chloroplast Gene ndhF

J. F. Smith; J. C. Wolfram; K. D. Brown; C. L. Carroll; D. S. Denton

Annals of the Missouri Botanical Garden, Vol. 84, No. 1. (1997), pp. 50-66.

Stable URL: http://links.jstor.org/sici?sici=0026-6493%281997%2984%3A1%3C50%3ATRITGE%3E2.0.CO%3B2-T

Annals of the Missouri Botanical Garden is currently published by Missouri Botanical Garden Press.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/mobot.html.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact [email protected].

http://www.jstor.org Tue Nov 13 08:53:31 2007 TRIBAL RELATIONSHIPS IN J. F. Smith2, J. C. Wolfram2," K. D. THE GESNERIACEAE: BrownZ, C. L. Carroll2, and D. S. DentonZ EVIDENCE FROM DNA SEQUENCES OF THE CHLOROPLAST GENE ndhF1

The tribal relationships of tlie Gesneriaceae are investigated using ndhF sequences. A full analysis of 70 taxa including 16 species froni the Scrophulariaceae, Bignoniaceae, and canthaceae as outgroups, resulted in two most- parsinionious of 5610 steps each. In all trees the Gesneriaceae were a nionophyletic group and Paulownia was tlie closest single-species outgroup for tlie analysis. Further analyses eliminated all but the members of the Gesneriaceae and Paulownia in order to better assess relationships within the family. The smaller analysis resulted in a single most- parsimonious of 4613 steps. The Klugieae are identified as tlie sister to tlie remainder of the family and could be separated as a distinct subfamily. The subfanlilies Cyrtandroideae (excluding Klugieae) and are monophyletic. The placement of Coronanthereae in Cyrtandroideae does not have suppol-t from this analysis, whereas its placement in Gesnerioideae is suppol-ted. Alternatively, Coronanthereae could be segregated as a separate subfamily 11ut in order to avoid a paraphyletic Gesnerioideae 15-auld either include the Sapeantheae and Beslerieae or elevate these two tribes to an additional subfamily. within Gesnerioideae the Sinningia is removed from tlie tribe Gloxinieae into tlie Sinningieae, which also contains the recently combined species Sinningia brasiliensis (Lietzia), as \+-ell as Paliavana and Vanhouttea. The Episcieae, Gesnerieae, Napeantheae, and Beslerieae are identified as mono- phyletic groups, as are the remainder of the Gloxinieae with Sinningia sensu lato removed. Within Cyrtandroideae, several well-supported, monophyletic lineages within the large, heterogeneous tribe Didyniocarpeae are identified, and with the current data tlie tribe Trichosporeae appears to be polyphyletic. The distribution of chromosome numbers. nodal anatomy, placental structure, and stem ~nodificationare examined based on these niolecular trees.

Investigations" of higher- level cladistic relation- several -genera not having-. any strict affinity to the ships (generic, familial, and above) have recently Scrophulariaceae or other related families included drawn a great deal of attention (Annals of the Mis- in the analysis (Olmstead & Reeves, 1995). Like- souri Botanical Garden Vol. 80(3); Olmstead et al., wise, Olmstead and Reeves (1995) found that sev- 1992, 1993; Donoghue et al., 1992; Cantino, 1992; eral families traditionally segregated from the Judd et al., 1994). These analyses have provided Scrophulariaceae are best included as members of tremendous insights toward our classification sys- one of the two major lineages (e.g., Plantaginaceae). tem and process of classification, frequently draw- Although most members of the s.1. are ing attention to families that have been separated temperate, there are some primarily tropical groups on the basis of primarily woody versus herbaceous (Gesneriaceae, , Bignoniaceae). In or- taxa (Cantino, 1992; Judd et al., 1994) or tropical der to better assess whether the division between versus temperate (Judd et al., 1994). More recently these families represents another artificial segre-- an investigation of the Lamiales sensu lato has in- gation based on distribution (tropical vs. temperate) dicated that the largest family in this order, Scroph- or woody versus herbaceous (e.g., Bignoniaceae vs. ulariaceae, is unlikely to be a monophyletic group Gesneriaceae), a thorough investigation of the Ges- (Olmstead et al., 1992, 1993; Olmstead & Reeves, neriaceae was deemed necessary to complement 1995). A thorough investigation of the Scrophular- the investigations that have already demonstrated iaceae utilizing DNA sequences from both the rbcL monophyly of Acanthaceae (Scotland et al., 1995) and ndhF -genes has indicated that the family is and Bignoniaceae- (R. Olmstead, pers. comm.), but comprised of at least two monophyletic groups with have not sampled widely in the Gesneriaceae.

'Ki are indebted to the following for sharing material: L. E. Skog, '81. L. Wagner, J. K. Boggan, Strybing rboretum. M. Kallersjo, B. Sordenstam, R. Dunn. D. Turley, J. Katzenstein, B. Stewart. M. Evans, and tlie American Gesneriad and Gloxinia Society (AGGS) seed fund. we also thank Richard Ol~nsteadand Michael Kielin for helpful comments on the manuscript. Funding for this project. was provided by NSF grant DEB-9317775 and a grant from AGGS to JFS. ' Department of Biolog); Boise State University, 1910 University Drive, Boise, Idaho, 83725, US-I ? Current address: 1263 Londondery, Idaho Falls, Idaho, 83404, U.S.A. Volume 84, Number 1 Smith et al. 51 1997 Tribal Relationships in Gesneriaceae

The Gesneriaceae are a mid-sized to large plant 1962). Another character that has been useful in family comprising approximately 2500-3500 spe- separating the subfamilies is the presence (Gesner- cies in 120-135 genera, distributed primarily in ioideae) or absence (Cyrtandroideae) of endosperm the tropics with a few temperate species in Europe, in the seed. In addition, the Gesnerioideae have a China, and Japan (Heywood, 1978; A. Weber, pers. neotropical distribution and most species have in- comm.). The majority of species in the Gesneri- ferior or semi-inferior ovaries, whereas the Cyrtan- aceae are herbaceous perennials, but can be an- droideae are primarily paleotropical with superior nuals, , lianas, and trees. Many species ovaries. However, the geographic distribution and (20%) are epiphytic, and the Gesneriaceae rank ovary position are not consistent within the subfam- among the top ten plant families in terms of abso- ilies. Therefore, although the Cyrtandroideae can lute numbers of epiphytic taxa (Madison, 1977; be defined by a synapomorphic character (uneven Kress, 1986). Given the diverse habits of the Ges- cotyledon development), the Gesnerioideae have neriaceae, it is not surprising that there is a wide been characterized by a symplesiomorphic char- array of morphological variation within the family. acter common to dicotyledons in general. Corolla tubes may be long and prominent as in Col- The two subfamilies have been divided further umnea L., or short as in Saintpaulia Wendl. Leaves into 9-17 tribes (Bentham, 1876; Burtt, 1962, are opposite in the majority of the family, but ani- 1977; Fritsch, 1893, 1894; Ivanina, 1965; Wiehler, sophylly, leading to an alternate arrangement with 1983; Burtt & Wiehler, 1995). The classification abscission of the smaller leaf, is common. Many of schemes differ due to the characters emphasized. these morphologically diverse features of the Ges- For example, Fritsch (1893, 1894) placed the Col- neriaceae are hypothesized as adaptations to the umneae in the Cyrtandroideae based on their su- epiphytic habit (Ackerman, 1986). perior ovary. Later, the Columneae were moved to The Gesneriaceae are a member of the Lamialee the Gesnerioideae due to the presence of uniform s.1. and are distinguished from other families in the cotyledons (Burtt, 1962, 1977) and combined into order by the combination of five-lobed corollas, pa- the Episcieae based on nodal anatomy (Wiehler, rietal placentation, and presence of endosperm in 1983). This paper presents a cladistic analysis of most taxa (Cronquist, 1981). However, because DNA sequences in order that phylogenetic relation- many of these characters vary within some mem- ships among taxa may be more clearly resolved, bers of the Gesneriaceae (including variation within and a more stable classification scheme proposed. individuals of some species), there has been con- The gene ndhF is a chloropfast gene that in to- siderable confusion regarding the placement of bacco encodes a protein of 740 amino acids pre- some genera. For example, members with axile pla- sumed to be a subunit of an NADH dehydrogenase centation can be classified incorrectly with the (Sugiura, 1992). The use of ndhF sequences for ScropKulariaceae, and those genera lacking endo- systematic studies has provided a far greater num- sperm potentially may be classified with the Acan- ber of characters to resolve relationships than stud- thaceae and Bimoniaceae." There have been relatively few cladistic analyses ies using rbcL. The reasons for the increased num- performed within the Gesneriaceae (Kvist, 1990; ber of characters are that the gene is approximately Crisci et al., 1991; Boggan, 1991; Smith & Sytsma, 50% longer than rbcL (2103 vs. 1431 bp in tobacco 1994a, b, c; Smith, 1996), and only one (Smith, [Wolfe, 19911) and has a nucleotide substitution rate that is approximately two times higher than 1996), performed. at the tribal level. A cladistic analysis is desirable to help resolve relationships, rbcL based on comparisons of rice and tobacco to determine if the family is monophyletic, and to (Sugiura, 1989). In recent studies using this gene improve classification within the family by rear- in the Acanthaceae, Scotland et al. (1995) found ranging tribes and subfamilies to reflect phyloge- three times the number of characters compared to netic relationships. rbcL, and Olmstead and Sweere (1994) discovered Classifications of the Gesneriaceae traditionally 60% more variable characters with ndhF in the So- recognize two subfamilies (Gesnerioideae and Cyr- lanaceae. Likewise, Clark et af. (1995) have found tandroideae) (Bentham, 1876; Burtt, 1962, 1977; that ndhF sequences are informative for resolving Fritsch, 1893, 1894), but others have included an relationships within the Poaceae, and Olmstead and additional subfamily (Coronantheroideae: Wiehler, Reeves (1995) have resolved several clades in a 1983; Episcioideae: Ivanina, 1965). The division of polyphyletic Scrophulariaceae. The larger number the family is largely based on the uniform (Ges- of variable characters makes ndhF sequences ideal nerioideae), or uneven (Cyrtandroideae) enlarge- for taxonomic groups that have not been resolved ment of the cotyledons after germination (Burtt, well using rbcL data, such as members of the As- Annals of the Missouri Botanical Garden

Table 1. Species sequenced in this study with Genbank submission numbers and voucher specimens. JFS - James F. Smith, IVLW - marren L. Wagner, DEB - Dennis E. Breedlove, SI - Smithsonian Institution, LG - Longwood Gardens. Letters in parentheses indicate herbarium where vouchers are deposited.

Genbank Species Voucher number Achimenes skinneri Lindl. SI 94-606 Aeschynanthus micranthus C. B. Clarke JFS 643 (KIS) Agalmyla parasitica (Lam.) Kuntze SI 94-570 Alloplectus meridensis Klotzsch JFS 1182 (wIS) Anna mollifolia (m: T. Wang) 'K T. Wang & K. Y. Pan Skog 94-498 Asteranthera ovata (Cav.) Hanst. Stewart 12234 (SRP) Besleria afJink Morton LG870575 Boea hygroscopica F. Muell. SI 89-041 Chirita sinensis Lindl. SI 94-111 Codonanthe elegans Wiehler SI 82-45 Columnea schiedeana Schleclit. JFS 288 ('KIS) Cyrtandra hawaiensis C. B. Clarke RrLW 6753 (BISH) Cyrtandra umbellijera Merr. WLW 6701 (BISH) Cyrtandromoea acuminata Benth. & Hook. JFS 3539 (SRP) Diastema racemijerum Bentli. SI 85-98 Didissandra frutescens Clarke SI 94-512 Didymocarpus albomargjnata Hemsl. SI 94-509 Drymonia stenophylla (J. D. Smith) H. E. Moore JFS 2248 ('KIS) australis Cunn. Stewart s.n. (SRP) Gasteranthus corallinus (Fritsch) Kiehler SI 94-243 Gesneria pedicellaris lain SI 94-567 Gesneria christii Urban SI 94-507 Gloxinia sylvatica (HBK) Kunth Dunn 9012051 (SRP) Hemiboea henryi C. B. Clarke SI 85-157 Kohleria spicata (Kunth) Oerst. SI 94-552 SI 94-158 Mitraria coecinea Cav. Stewart s.n. (SRP) Monophyllaea hirticalyx Franch. no voucher Monopyle macrocarpa Benth. no voucher Nap.eanthus costaricensis Wiehler no voucher Napeanthus macrostoma Leeuwenberg Feuillet (US) Negria rhabdothamnoides F. Muell. Nordenstam 8608 (S) Nematanthus hirsutus (Mart.) Wiehler Ol~nstead&. Reeves. 1995 Niphaea oblonga Lindl. SI 78-354 Opithandra primuloides (Miq.) B. L. Bul-tt SI 93-073 Ornithoboea wildeana Craib. SI 93-075 Paliavana prasinata (Ker-Gawl.) Fritsch SI 78-368 Paraboea rufescens (Franch.) Bul-tt Skog s.n. (US) Petrocosmeajaccida Craib SI 85-196 Primulina tabacum Hance SI 93-040 Ramonda myconi (L.) Rchh. Katzenstein s.n. (SRP) Rhynchoglossum notonianum (Wall.) B. L. Bul-tt SI 94-378 Rytldophyllum tomentosum (L.) Mart. SI 77-235 R.ltidophrllum ., aurlculatum Hook. SI 94-524 Saintpaulia rupicola B. L. Burtt SI 94-492 Sarmienta repens Ruiz &. Pav6n Stewal-t s.n. (SRP) SinningGz (htzia)brasiliensk (Regel &. Schmidt) Wiehler Dunn 9104014 (SRP) Sinningia cooperi (Past.) Wiehler SI 94-340 Sinningia richii Clayb. SI 94-554 Solenophora obliqua D. L. Denham &. D. N. Gibson DEB 71542 (CAS) Streptocarpus holstii Engl. Ol~nstead& Reeves. 1995 Streptocarpus saxorum Engl. JFS s.n. (wIS) Titanotrichum oldhamii (Hemsl.) Soler. SI 86-106 Volume 84, Number 1 Smith et al. 1997 Tribal Relationships in Gesneriaceae

Table 1. Continued.

Genbank Species Voucher number Vanhouttea lanata Fritsch

Antirrhinum majus L. Ol~nstead& Reeves, 1995 Brillantaisia lamium Benth. Scotland et al.. 1995 Catalpa sp. Ol~nstead& Reeves, 1995 Celsia arcturus Jacq. Olrnstead & Reeves, 1995 Crabbea reticulata C. B. Clarke Scotland et al., 1995 Crossandra nilotica Oliv. Scotland et al., 1995 Digitalis grandgora Mill. Olnlstead & Reeves, 1995 corymbosa Lindau Scotland et al., 1995

Martinella obovata IHBK), , Bureau & K. Schum Olnlstead & Reeves, 1995 Paulownia tomentosa Steud. Olmstead & Reeves, 1995 Selago thunbergii Choisy Ol~nstead& Reeves, 1995 Schlegelia parvlflora (Oerst.) Monachino Olmstead & Reeves, 1995 Scrophularia sp. Ol~nstead& Reeves, 1995 Tabebuia heterophylla (A. de Candolle) Britton Olmstead & Reeves, 1995 Verbascum thapsus L. Ol~nstead& Reeves, 1995 Veronica catenata Pennell Olmstead & Reeves, 1995 teridae and Lamiales s.1. (Olmstead et al., 1992, fications within the family and, whenever possible, 1993; Chase et al., 1993). to match genera used in the morphological analysis. In some instances, a genus that had been used with MATERIALSANDMETHODS the morphological analysis (Smith, 1996) was not readily available for the molecular analysis. There- The gene sequences used in this analysis were fore this molecular analysis contains many taxa that generated by thermal cycle sequencing (Innis et al., have not been included in the morphological anal- 1988) of previously amplified ndhF regions. The ysis, and direct comparisons will be made with a ndhF gene was amplified in two overlapping sec- reduced data set at a future date. The species used tions (positions 1-1350, and 972-2044) from ge- in the analysis, voucher information, and Genbank nomic'DNA isolated from fresh, frozen, or silica gel accession numbers are included in Table 1. dried material (Smith et al., 1992). Once amplifi- Genera have been selected to represent the most cation products were obtained, the sample was pu- recent tribal classifications with two to ten genera rified using PCR wizard purification preps (Pro- from each tribe (Tables 2 and 3). In order to rep- mega) according to the manufacturer's instructions. resent current classification systems along with ear- The purified DNA then was subjected to cycle se- lier systems, 48 genera were selected (Tables 2 and quencing using the Silver Sequence method (Pro- mega). This sequencing method resulted in se- 3). quences that could be read within six to eight hours after amplification was completed. The products of OUTGROUP SELECTIOK one round of amplification provided sufficient ma- Outgroups were selected to root the tree repre- terial for completing the gene sequences described senting tribal relationships within Gesneriaceae. here. The advantages of the silver staining proce- The best method for doing this is by outgroup com- dure over radioactive methods are safety, minimal parison (Donoghue & Cantin4 1984; Maddison et waste disposal, and speed. al., 1984). The most appropriate outgroup for the tribes of the Gesneriaceae should be the most CHOICE OF TAXA closely related plant family or clade. The Gesner- The focus of this analysis was on the tribal re- iaceae have been placed in the order Lamiales s.1. lationships of the Gesneriaceae and comparison of in the subclass Asteridae (or equivalent groups of the results of this analysis with one based on mor- families) in numerous taxonomic treatments (Dahl- phological data (Smith, 1996). Genera were select- gren, 1975; Thorne, 1976, 1983, 1992; Heywood, ed to represent current and previous tribal classi- 1978; Takhtajan, 1980; Cronquist, 1981). However, Annals of the Missouri Botanical Garden

Tahle 2. Genera of Gesnerioideae (Burtt & Kiehler, 1995) used and their classification status. ST = not treated. Subfanlilial names are underlined to readily distinguish them from tribal names. Suhtrihal names are abbreviated as Colum. - Columneineae, Codon. - Codonanthinae.

Genus R'iehler. 1983 Ivanina, 1965 Fritsch. 1893-94 Gesnerioideae Gesnerioideae Gesnerioideae Achlmenes Gloxinieae Gloxinieae Gloxinieae Gloxznza Gloxinieae Gloxinieae Gloxinieae Monopr le Gloxinieae Bellonieae Bellonieae ,Vzphaea Gloxinieae Bellonieae Bellonieae Kohlerza Glosinieae Kohlerieae Kohlerieae Dzastema Gloxinieae Kohlerieae Kohlerieae Slnnzngla Glosinieae Kohlerieae Sinningieae Vanhouttea Gloxinieae Kohlerieae Kohlerieae Pallavana Glosinieae Reichsteinerieae Kohlerieae Lzetzza Gloxinieae Reichsteinerieae Sinningieae Solenophora Glosinieae Solenophoreae Solenophoreae Gesnerza Gloxinieae Gesnerieae Gesnerieae R~tldophrllum =Gesneria Gesnerieae Gesnerieae Episcioideae Cyrtandroideae Columnea Episcieae Columneae Columneae-Colum. Codonantlle Episcieae Columneae Columneae-Codon. ~Vematanthus Episcieae Columneae Columneae-Colum. Alloplectus Episcieae Episcieae Columneae-Colum. Drjmonia Episcieae Episcieae Columneae-Colurn. Beslerza Beslerieae Episcieae Beslerieae Gasteranthus Beslerieae NT ST ~Vapeanthus Napeantheae Episcieae Klugieae Coronantheroideae Cyrtandroideae Asteranthera Coronanthereae Mitrarieae Coronanthereae Sarmienta Coronanthereae Mitrarieae Coronanthereae Mitraria Coronanthereae Mitrarieae Coronanthereae Fieldia Coronanthereae Mitrarieae Coronanthereae Negria Coronanthereae Coronanthereae Coronanthereae

the relationships among these families are some- PHYLOGENETIC ANALYSIS what ambiguous. A recent cladistic analysis of Phylogenetic divergence was reconstructed using these families based on DNA sequencing of the PAUP version 3.1.1 (Swofford, 1993) to implement chloroplast encoded rbcL gene resulted in poor res- Wagner parsimony (Farris, 1970; Farris et al., olution of the relationships of these families (Olm- 1970; Swofford & Maddison, 1987). This program stead et al., 1993), although these relationships allows parallelisms and reversals (homoplasy), and have been more resolved with the addition of ndhF provides an option for missing data. In this analy- sequences (Olmstead & Reeves, 1995). sis, trees were generated using the general heuristic Three families from the Lamiales s.1. were used option, saving minimal trees only, with the collapse as outgroups for this analysis. These were the zero-length branches, and ignore uninformative Acanthaceae, Bignoniaceae, and Scrophulariaceae. characters options in effect. Because of the large Sequences for sixteen species of these three fami- number of taxa in this analysis, the branch and lies were obtained via Genbank (Table 1; Olmstead bound and exhaustive search options would have & Reeves, 1995; Scotland et al., 1995) and includ- consumed an excessive amount of time. Therefore, ed representatives from three lineages identified the trees presented here are best approximations within the Scrophulariaceae (Olmstead & Reeves, and not exact solutions. The manner in which the 1995). Initial analyses used all 16 species as the program reconstructs phylogenetic sequences is outgroup. Subsequent analyses used only Gesneri- sensitive to the order of taxa presentation in the aceae with Paulownza Sieb. & Zucc. as the out- data matrix, frequently finding islands of equally group. parsimonious trees depending on the order (Mad- Volume 84, Number 1 Smith et al. 1997 Tribal Relationships in Gesneriaceae

Table 3. Genera of Cyrtandroideae (Burtt & wieliler, 1995) used and their classification status. ST = not treated. The tribe, Didymocarpeae, is abbreviated Didy. in order to show the subtribal classification system of Ivanina (1965) and Fritsch (1893, 1894).

Genus Bul-tt, 1962.77 Ivanina. 1965 Fritsch. 1893-94 Ramonda Didymocarpeae Ramondeae Ramondeae Saintpaulia Didymocarpeae Saintpaulieae Ramondeae Opithandra Didymocarpeae Didy.-Roettlerineae ST Didymocarpus Didymocarpeae Didy.-Roettlerineae Didy.-Roettlerineae Didissandra Didymocarpeae Didy.-Roettlerineae Didy.-Oreacliarineae Anna Didymocarpeae Didy.-Roettlerineae NT Chirita Didymocarpeae Didy.-Roettlerineae Didy.-Roettlerineae Petrocosmea Didymocarpeae Didy.-Roettlerineae Ramondeae Titanotrichum anomalous NT ST Cyrtandromoea Loxonieae/Scropli. Klugieae Beslerieae Paraboea Didymocarpeae Didy.-Roettlerineae Didy.-Roettlerineae Boea Didymocarpeae ST Streptocarpeae Hemiboea Didymocarpeae Didy.-Roettlerineae NT Primulina Didymocarpeae Didy.-Roettlerineae Klugieae Streptocarpus Didymocarpeae Didy.-Streptocarp. Streptocarpeae Ornithoboea Didymocarpeae Didy.-Streptocarp. Streptocarpeae Aeschynanthus Tricliosporeae Trichosporeae Trichosporeae Agalmyla Trichosporeae Tricliosporeae Trichosporeae Lysionotus Trichosporeae Tricliosporeae Tricliosporeae Cyrtandra Cyl-tandreae Cyl-tandreae Cyl-tandreae Rhynchoglossum Klugieae Klugieae Klugieae Monophyllaea Klugieae Klugieae Beslerieae

dison, 1991). Therefore, it is important to repeat inferred from gaps in the sequence alignments, the analysis several times. To do this, the search which in previous analyses had been re-scored as strategy of Olmstead and Palmer (1994) was imple- binary characters and used a$ either an indepen- mented: searching for 1000 trees each in five sub- dent data set or combined with the sequence data sequent analyses with the nearest neighbor inter- (Scotland et al., 1995). These indels were viewed change (NNI) search option in effect and mulpars as having phylogenetic importance (Scotland et al., "off." Each of the results from the five NNI search- 1995); therefore indels found in the Gesneriaceae es was used as the starting tree(s) for a search with were removed and examined independently of se- tree bisection reconnection (TBR) and mulpars quence data for their phylogenetic utility. "on." This strategy was used in the full analysis The monophyly of various tribal relationships not with all 16 non-Gesneriaceae taxa designated as obtained in the most-parsimonious trees was ex- outgroups. Likewise, the same strategy was used amined by using the constraints option of PAUP. with only the members of the Gesneriaceae and These included the Trichosporeae, the Didymocar- Paulownia as the outgroup, and with constraints peae, the inclusion of Klugieae in Cyrtandroideae, options. and Sinningieae in Gloxinieae. Also, since the Branch support analysis was performed to ex- analysis with all 16 outgroup taxa resulted in the amine trees that were six or fewer steps longer than the most-parsimonious tree (Bremer, 1988; Dono- placement of Nematanthus Schrader and Klugieae ghue et al., 1992; Bremer, 1994). This type of anal- in discrepant positions from traditional classifica- ysis provides an indication of the robustness of the tions, an analysis with all 16 outgroups constrained data by determining which clades persist in a con- Nematanthus to the Gesnerioideae, and the Klu- sensus tree as parsimony is relaxed. This analysis gieae from the Gesnerioideae. The position of Klu- was performed by saving all trees six steps longer gieae and Nematanthus was also examined by con- than the most-parsimonious trees and then exam- structing a user-defined tree with a topology of one ining subsets of trees one to six steps longer with of the two most-parsimonious trees except that Ne- the filter option of PAUP. matanthus was placed in the Episcieae, and Klu- The ndhF sequences used here had several six gieae was placed as sister to the remainder of the to twelve base pair insertions or deletions (indels) Gesneriaceae. This user-defined tree was then the Annals of the Missouri Botanical Garden

starting tree for a search using TBR and mulpars construct these trees The analysis required four "on." addltlonal steps to create a monophyletlc Trichos- poreae, fi~efor a monophyletic D~d~mocarpeae,two to Include the Klugieae In the Cyrtandroideae, and four to Include the Slnningieae In the Gloxlnleae Complete sequences for the ndhF gene were ob- Constralnlng the anal~slsof all 70 taxa to lace tained for 52 species of Gesneriaceae (Table 1). Klugieae as the slster to the Gesnerlaceae and Are- These sequences were supplemented with sequenc- matanthus w~thinthe Eplscleae resulted in four es from an additional 18 species (2 within Gesner- trees 58 steps longer than the most-parsimonious iaceae and 16 from related families) from Genbank trees regardless of whether the constraint optlon of (Table 1). The complete sequences resulted in 849 PAUP, or user defined trees aere implemented. phylogenetically informative characters among all 70 species in the full analysis. A smaller analysis focused on only the Gesneriaceae species with Paulownia as the outgroup. Within this smaller The cladistic analysis of 54 species of Gesneri- analysis 690 nucleotide positions were found to be aceae with 16 species of Scrophulariaceae, Bigno- phYlogenetically informative. Indels were found at niaceae, and Acanthaceae as outgroups resulted in several positions in the Gesneriaceae from the se- a monophyletic Gesneriaceae with the single genus quences used in this analysis. Two widespread in- Paulownia (Scrophulariaceae) indicated as the sertions were a 12 bp insertion at position 1440 closest outgroup (Figs. 1, 2). These results verified and a 6 bp insertion at 1548. Other insertions were that the Gesneriaceae are distinct from other mem- autapo~nor~hicfor species or genera used in the bers of the Lamiales s.1. and not an artificial unit analysis (unpublished results). No insertions were based on their largely tropical distribution and her- used in the analysis. The 6 bp insertion was sym- baceous habit as has been seen for some family plesiomorphic for the Gesneriaceae. The 12 bp in- pairs (Judd et al., 1994). The full analysis is largely sertion was also symplesiomorphic for the Gesner- in agreement with the position of the taxa in the iaceae; however, sequence divergence within this reduced analysis with the exception of the positions insertion provides an additional synapomorphy for of Nemntanthus and the tribe Klugieae. The place- the clade comprised of Columnea, Drymonin Mart., ment of ~Vematanthusas the sister to the remainder and Alloplectus Mart. (Fig. 4), where a single base of the family is very far removed from its traditional pair transition characterizes these three genera. classification within the Episcieae (Fig. 1). Like- Other base pair substitutions and insertions were wise the Klugieae are placed unusually in the sub- found within this 12 bp insertion but, with the cur- family Gesnerioideae (Fig. 2). The most likely ex- rent level of sampling, were autapomorphic. planation for the anomalous placement of these taxa Cladistic analysis was performed initially with all is the high level of homoplasy between the Ges- 70 taxa of the four families (Gesneriaceae, Scroph- neriaceae and the outgroups. This is exemplified ulariaceae, Acanthaceae, and Bignoniaceae) and all when 15 of the 16 outgroup species are removed taxa in the three outgroup families designated as from the analysis. In the reduced analysis both Ne- the outgroup. This analysis resulted in two trees of matnnthus and Klugieae are in more expected po- 5610 steps each (consistency index (CI) = 0.30, sitions regarding relationships to the remainder of retention index (RI) = 0.48), all of which indicated the family. An alternative explanation is that be- the Gesneriaceae were a monophvletic family and cause of the size of the data set, PAUP did not find that the genus Paulozunia (Scrophulariaceae) was the shortest tree and that a shorter tree with all 70 the closest outgroup (Figs. 1, 2). species exists that places fiematanthus and the Subsequent analyses were performed to minimize Klugieae in their more expected relationships. This computer analysis time that utilized only the Ges- latter explanation is unlikelv since searches con- neriaceae and Paulowniu as a designated outgroup. straining these taxa to their more traditional posi- This reduced analysis resulted in a single most- tions, or a user-defined tree that placed them there, parsimonious tree of 4613 steps (CI = 0.27, RI = resulted in four trees that were 58 steps longer. 0.38 ) (Figs. 3, 4). Some taxa that have been The reduced analysis resulted in a single most- thought to be monophvletic, or comprised tribes, parsimonious tree (Figs. 3, 4). Three major mono- were examined using the constraints option of phyletic divisions within the family coi-respond to PAUP to determine the impact of the monophyletic subfamilies Gesnerioideae and Cvrtandroideae (mi- grouping on the remainder of the data and to de- nus tribe Klugieae) and tribe Klugieae in a separate ternline the number of additional steps required to position as a potential third subfamily. Traditional Volume 84, Number 1 Smith et al. 1997 Tribal Relationships in Gesneriaceae

Gesnerioideae

Cyrtandra hawaiensis Cyrtandra umbellifera Hemiboea Lysionotus Aeschynanthus Petrocosmea Opithandra Anna Didissandra Didymocarpus CY Ornithoboea Prirnulina Agalmyla Boea Paraboea Streptocarpus saxorurn Saintpaulia Streptocarpus holstii Ramonda Chirita Titanotrichup - Nematanthus -GE Paulownia 1 Antirrhinum Digitalis Veronica Jsc Verbascum Celsia Scrophularia - Selago lsc Catalpa Tabebuia Martinella IBI - Brillantaisia Hygrophila AC Crabbea Crossandra Schlegelia -BI Figure 1. Strict consensus of two most-parsimonious trees of 5610 steps each (CI = 0.30. RI = 0.48) displaying the outgroup taxa, BI-Bignoniaceae, AC-Acanthaceae, SC-Scrophulariaceae, and the subfamilies of the Gesneri- aceae. GE-Gesnerioideae and CY-Cyrtandroideae. The remainder of the Gesnerioideae are displayed in Figure 2. See text for explanation of position of Nematanthus in this cladogram. Annals of the Missouri Botanical Garden

Diastema C Monopyle - Achirnenes Kohleria - Sinningia brasiliensis (Lietzia) C Sinningia richii Gesneria christii -

17(12) Gesnerioideae 4

Cyrtandra hawaiensis

Cyrtandra 18(16)

-28(24) -(14)

(18)- 24(23) 3 1

24(23) >6 7Di lO(9) -1 Di 2 (18) 70(61' Agalmyla Tr -15(11) 2 2

35(31) 34(30) 3

48(38) 3 82(72) Streptocarpus holstii

75 70 Ramonda Di i(22) - 61(571 Chirita Di 60(54) Titanotrichum -Ti 115(102) Monophyllaea

log 93 Rhynchoglossum

Cyrtandromoea 71 62 Paulownia SC Figure 3. Single most-parsimonious tree of 4613 steps (C1 = 0.29. RI = 0.38)frorn the analqsis of the species in the Gesneriaceae with onlq Pauloul~ia(SC-Scr~~llulariaceae) designatetl as the outgroup. Displa!ed in this figure are the tribes of the C>rtantlroideae. K1-Klugieae. Ti-Titanotricheae. Di-Did?mocarpeae. Tr-Trichosporeae. and C?- C)rtantlreae. The Gesrierioideae are tlisplaqed in Figure 4. Numbers along branches are the synaponlorphies that support those clatles. Sumber~in parentheses indicate those ~qnapornor~hiesthat are hornoplastic in this tree. Nurnbers helow ljranches are decal ~alues.Brarlches vith no ~alueintlicatetl ha\-e a decay \.slue of 1. Annals of the Missouri Botanical Garden

-

(3) () Diastema (16) 59(57) Solenophora 57(55) Monopyle 22(21) 74(66) Gloxinia 16(13 23(22) 6 [62(53i Niphaea ~25)65(63) Achimenes 15(14) 4 16(13:) 5 [79(71)19(18)( ) GesneriaKohleria christii

16(15) - >6 30 28 Gesneria pedicellaris 23(21) (I1) Rytidophyllum auriculatum (12) >6 [ (27) Rytidophyllum tomentosum 34(29) 77(6Qoplectu~ 7 27(19) 6 175(68) Drymon$ 4 47(41) 24(21) Columnea 20(17) 4 86(81) Codonanthe 3 15(12) [54(48) Nematanthus 6764) Paliavana 19(17) 25(24) 1 49(43) Vanhouttea 2 60(53) Sinningia cooperi 15(14) Fig* 17(12) 1 60(55) Sinningia brasiliensis (Lietzia) +\4 4(58) sinningia rick 57(53) Napeanthus macrostoma7 28(24) (21) 6 ["(s') Napeanthus costaricensis7 5 41(35) 60(47) Besleria (28) (14) [ 29(27) GasteranthusI 5 Mitraria

27(21) 6 l9(18) >6 [15(14' Fieldia 71(65) Asteranthera

Figure 4. Sirigle most-parsimonious tree of 4613 steps (CI = 0.29. RI = 0.38) from the analysis of the species in the Gesrieriaceae with only Paulotvnia designated as the outgroup. Displayed in this figure are the tribes of the Gesnerioideae, Co--Coronanthereae, Be-Beslerieae, Na-Kapeantheae, Si-Sinningieae, Ep-Episcieae, Ge-Ges- nerieae, and G1-Gloxinieae. The Cyrtandroicleae are displayed in Figure 3. Kumbers along branches are the syna- pomorphies that support those clades. humbers in parentheses indicate those synapomorphies that are homoplastic in this tree. Kumbers below branches are decay values. Branches with no value indicated have a decay value of 1. Volume 84, Number 1 Smith et al. 61 1997 Tribal Relationships in Gesneriaceae

classification schemes have placed tribe Klugieae separate subfamilial status for Coronanthereae due in the Cyrtandroideae (Table 3); however, the in- to the numerous autapomorphic characters pos- clusion of tribe Klugieae within subfamily Cyrtan- sessed by members of this group, such as fusion of droideae would result in a paraphyletic Cyrtandroi- the nectary to the ovary wall and high chromosome deae. The removal of this tribe to a third subfamily numbers (Wiehler, 1983). The morphological data would result in a monophyletic Cyrtandroideae. The would allow the Coronanthereae to be either a monophyletic groups within the subfamily Gesner- monophyletic tribe within the Gesnerioideae or a ioideae correspond highly with traditional classifi- separate monophyletic subfamily without disrupting cation systems for this subfamily (Wiehler, 1983) the of any other group (Smith, 1996). and a previous cladistic analysis based on morpho- However, based on the molecular data presented logical data (Smith, 1996). The relationships within here, if the Coronanthereae were raised to subfam- the Cyrtandroideae are less congruent with previ- ily level, it would either include the tribes Napean- ous taxonomic treatments, mainly due to the limited theae and Beslerieae from the Gesnerioideae or ne- understanding and sampling of the large, hetero- cessitate elevating these two tribes as an additional geneous tribe Didymocarpeae (Burtt, 1962). subfamily (Fig. 4). Therefore it is recommended that the Coronanthereae be treated as a tribe of the

5LBFiMILI4L GROLPINGS Gesnerioideae rather than a separate subfamily.

The separation of the Gesneriaceae into two sub- families (including Coronanthereae in the Gesner- ioideae) has become well accepted during the past 30 years since the discovery of unequal cotyledon Among the relationships within the Gesnerioi- enlargement in the Cyrtandroideae (including mem- deae, the primary lack of congruence between this bers of the tribe Klugieae) and equal cotyledon en- analysis and the most recent classification scheme largement in the Gesnerioideae (Burtt, 1962). How- by Burtt and Wiehler (1995) is the polyphyly of the ever, from a cladistic viewpoint the Cyrtandroideae Gloxinieae. However, the removal of Sinningia are defined by a synapomorphy whereas the Ges- Nees (including Lietzia Regel, but not including nerioideae are defined by a symplesiomorphy. One Paliavana Vandelli or Vanhouttea Lem.) has been problem with this character is that it has not been proposed previously (Fritsch, 1893, 1894) as the examined thoroughly for all members of the differ- tribe Sinningieae. The monoph$ly of Paliavana, ent subfamilies, including many of the taxa used in Lietzia, and Sinningia has been proposed by Bog- this analysis. gan (1991), where all three genera were proposed Although an analysis of morphological data that to be members of Sinningia as the result of a mor- included cotyledon expansion did not support the phologically based cladistic analysis of Sinningia monophyly of the Cyrtandroideae (Smith, 1996), the species and several related genera. These results cladistic analysis of ndhF sequences presented were not supported with a morphology-based cla- here demonstrates both a well-supported monophy- distic analysis (Smith, 1996), most likely due to letic Cyrtandroideae (Klugieae excluded) and Ges- limited sampling among these taxa (Sinningia sen- nerioideae (Figs. 3, 4). The monophyly of the Ges- su stricto was represented only by Sinningia sect. nerioideae is supported in both a morphological Corytholoma and Vanhouttea was not included). analysis (Smith, 1996) and this molecular analysis The results presented here indicate that Sinningia (Fig. 4). The Cyrtandroideae were paraphyletic in (including the recently combined Lietzia), Palia- a cladistic analysis of morphological data (Smith, vana, and Vanhouttea should be removed from 1996) but are well supported with ndhF sequences Gloxinieae and placed in a separate monophyletic (Klugieae excluded), although the position of Titan- tribe Sinningieae. Although Sinningia is paraphy- otrichum Solereder as sister to the remainder of the letic in this analysis (Fig. 4), limited sampling from Cyrtandroideae is supported with only 22 homo- this large genus leads only to a tentative conclusion plastic character state changes (Fig. 3). that both Paliavana and Vanhouttea should be The placement of Coronanthereae within the combined into Sinningia to create a monophyletic Gesnerioideae is well supported with ndhF se- genus. quences (Fig. 4) as it is with morphological data The sister relationship of the Beslerieae and Na- (Smith, 1996). This tribe does not belong in the peantheae has been hinted at based on the overlap subfamily Cyrtandroideae as had been proposed of several diagnostic characters between these earlier (Fritsch, 1894). Wiehler (1983) in his treat- tribes (Skog, 1995; Skog & de Jesus, 1996). How- ment of the neotropical Gesneriaceae suggested a ever, the sister relationship of these two tribes to Annals of the Missouri Botanical Garden

the Coronanthereae (Fig. 4) has not been proposed this species, diverged early in the evolution of the previously. Although the morphological data did not family. The placement of Titanotrichum within the indicate sister group status, the data did indicate a Gesneriaceae is discussed elsewhere (Smith et al., close affinity among these three tribes (Smith, 1997). 1996). The Didymocarpeae are a large heterogeneous- Among the recent classification schemes pro- tribe that includes the majority of genera in the posed for the Gesnerioideae, Burtt and Wiehler's Cyrtandroideae (Burtt, 1962, 1977; Wang et al., (1995) is the closest approximation to the results 1992). In this analysis it is a paraphyletic assem- obtained in this study. The subdivision of Wiehler's blage that includes the Cyrtandreae and Trichos- (1983) Gloxinieae into the Bellonieae, Kohlerieae, poreae (Figs. 1, 3). Because of the large size of the Rechsteinerieae, and Solenophoreae (Ivanina, Didymocarpeae, and the limited sampling of the 1965) (Table 2) is not supported by this cladistic tribe in this analysis, no conclusions regarding its analysis. Likewise separating Wiehler's (1983) monophyly, or potential division into other tribes, Gloxinieae into Bellonieae, Kohlerieae, and Solen- are recommended at this time. Further morpholog- ophoreae (Table 2; Fritsch, 1893, 1894) is not sup- ical investigations in this tribe are under way (B. ported except for the removal of the Sinningieae L. Burtt and A. Weber, pers. comm.), and a cladis- (Fig. 4), which would also necessarily include Pal- tic analysis that focuses on this group will be valu- iavana and Vanhouttea (included in Fritsch's Koh- able toward understanding its relationships. Several lerieae; Table 2). The placement of Napeanthus G. well-supported monophyletic groups within the Di- Gardner in the Klugieae (Cyrtandroideae) (Table 2) dymocarpeae can be identified (Boea Commerson as proposed by Fritsch (1893, 1894) is inappropri- ex LamarcklParaboea (C. B. Clarke) Ridley, Hem- ate. iboea C. B. ClarkelLysionotus D. Don, Didissandra C. B. ClarkelDidymocarpus Wallich, and Strepto- carpus Lindl.lSaintpaulia). It should be noted that Didissandra and Didymocarpus, although forming a Burtt's (1962, 1977) classification system for the monophyletic clade in this analysis, are both large Cyrtandroideae is closer in agreement to this cla- heterogeneous genera and that sampling different distic analysis than previous classification schemes species may have resulted in different placement. (Ivanina, 1965; Fritsch, 1893, 1894). However, the By focusing on morphologica! characters of these monophyly of the largest tribe, the Didymocarpeae, groups it may be possible to identify more inclusive is not supported by this analysis (Fig. 3). Likewise monophyletic tribes out of the paraphyletic Didy- none of the subtribes created by Ivanina (1965) or mocarpeae. Much greater sampling within this large Fritsch (1893, 1894) are supported as monophyletic group will be necessary before any major realign- groups (Fig. 3, Table 3). The Trichosporeae are not ment can begin. supported as a monophyletic clade (Fig. 3). Al- An unexpected result of this analysis is the pa- though this tribe was well supported in the mor- raphyly of Streptocarpus. The most likely explana- phological analysis (Smith, 1996), four additional tion for this paraphyly is limited sampling, with steps beyond the most-parsimonious tree are re- only two species of Streptocarpus and one of Saint- quired to make this clade monophyletic with ndhF paulia. However, it is interesting to note that Saint- data. paulia is one of the few genera within the Gesner- The position of .Titanotrichum has been problem- iaceae to have a chromosome number of n = 15 atic, although this genus has consistently remained (Skog, 1984). The only other genera that share this in the Gesneriaceae (Burtt, 1962, 1977; Wang et number are some species of Streptocarpus, includ- al., 1992; Burtt & Wiehler, 1995). Titanotrichum is ing both S. saxorum and S. holstii, and some spe- a member of the Cvrtandroideae based on these cies of Aeschynanthus Jack (Skog, 1984). The pos- data, and perhaps may be viewed best as a mono- sibility that Saintpaulia is derived from within typic tribe (Titanotricheae; Wang et al., 1992), sis- Streptocarpus, as indicated by ndhF sequences and ter to the remainder of the subfamily. However, the chromosome numbers, currently is being investi- position of Titanotrichum as the sister to the re- gated with greater sampling. mainder of the Cyrtandroideae is only weakly sup- The Trichosporeae traditionally have been ported with 22 homoplastic character state changes, viewed as a monophyletic tribe defined by the pres- and the resolution of its placement is lost in the ence of seed appendages not present elsewhere strict consensus of all trees only one step longer within the family (Burtt, 1962, 1977; Wang et al., than the most-parsimonious tree. Therefore, it is 1992). Based on morphological data, the Trichos- likely that Titanotrichum, or the lineage leading to poreae were one of the most strongly supported Volume 84, Number 1 Smith et al. 1997 Tribal Relationships in Gesneriaceae

tribes in a morphology-based cladistic analysis nerieae from Gloxinieae as seen here with sequence (Smith, 1996). However, it is apparent from this data. analysis of ndhF sequences that the selection of Most Gloxinieae sampled here (excluding Solen- characters that define the Trichosporeae is inappro- ophora Benth., Nzphaea Lindl., and Achimenes priate (e.g., seed appendages are common in the Pers.) have x = 13 (Wiehler, 1983; Skog, 1984). closely related Bignoniaceae). Alternatively, it is In addition, the members of the Sinningieae that possible that inadequate sampling from the Tri- have been examined also have x = 13 (Skog, chosporeae or the large tribe Didymocarpeae may 1984). This similarity in chromosome base number, be causing the separate placement of the three gen- along with other character states, has led previous era sampled from the Trichosporeae. This latter hy- researchers to include the members of the Sinnin- pothesis is unlikely, because one of the more gieae within the Gloxinieae (Wiehler, 1983). How- strongly supported clades in the analysis placed Ly- ever, based on the analysis presented here, the Sin- sionotus (Trichosporeae) with Hemiboea (Didymo- ningieae are best viewed as a tribe separated from carpeae) and away from the other genera of the Tri- the Gloxinieae, and x = 13 is homoplastic. chosporeae (Fig. 3). Other homoplastic chromosome numbers are x = 11 (Niphaea and Achimenes), and x = 9 (Alloplec- KLUGIEAE tus, Drymonia, Columnea, and some Didymocarpus species). Although most of these homoplastic The Klugieae are monophyletic and are the sister counts serve little phylogenetic utility, the count of group to the remainder of the Gesneriaceae (Fig. x = 9 serves to characterize a portion of the Ep- 3). The placement of this tribe in the Gesnerioideae iscieae. Most members of the Episcieae have x = (Fig. 2) in the full data analysis most likely is due 9, but taxa with x = 8 (Codonanthe (Mart.) Hanst. to homoplasy or the result of an incomplete search and Nematanthus) may represent another clade for the shortest tree. The monophyly and sister (Fig. 4). Further sampling within the Episcieae may group status of this tribe also was supported with a reveal if this clade (Fig. 4) continues to be sup- cladistic analysis of morphological data (Smith, ported or is the result of sampling in this analysis. 1996). The Klugieae possess numerous autapo- Other chromosome counts in the Cyrtandroideae morphic characters relative to other Gesneriaceae are highly variable even within genera, and no pat- such as narrow medullary rays, and verrucate edges tern emerges from the counts of the species that of the cells of the seed coat (Smith, 1996). The have been included in the analysis, with the ex- placement of Cyrtandromoea Zoll. in the Klugieae ception of the StreptocarpuslSaintpaulia counts dis- of the Cyrtandroideae was proposed previously by cussed above. Ivanina (1965), although other investigations indi- cated that this -genus should be excluded from the Gesneriaceae on the basis of floral anatomy (Burtt, NODAL IN4TOMI- 1965; Singh & Jain, 1978). The placement of Cyr- Another useful character for the Gesneriaceae is tandromoea in the Gesneriaceae is discussed else- nodal anatomy (Wiehler, 1983). Unfortunately only where (Smith et al., 1997). the subfamily Gesnerioideae has been sampled thoroughly for this character, and the lack of data E\ OLLTION OF NON-MOLECULiR CHiR4CTER ST4TE5 for the Cyrtandroideae necessitated the exclusion CHROMOSOME NLMBERS of this character from the morphological analysis Several chromosome counts are synapomorphic (Smith, 1996). However, if nodal anatomy is and non-homoplastic based on this cladistic anal- mapped onto the trees from this molecular analysis, ysis. Large numbers of chromosomes (n = 30+) are this character can provide useful phylogenetic in- unique to the Coronanthereae and would serve as formation. The tribe Episcieae (Fig. 4) is defined an additional character to separate this tribe from by a three-trace trilacunar node that is unique the remainder of the family (Skog, 1984). A chro- among the Gesnerioideae, although this character mosome base number, x, of 14 characterizes the state is known from the Cyrtandroideae. The unique Gesnerieae (Wiehler, 1983; Skog, 1984). The cla- presence of this character state within the Gesner- distic analysis of morphological data was unable to ioideae adds further support to the monophyly of separate the Gesnerieae from the tribe Gloxinieae the Episcieae. The three-trace trilacunar node may although it represented a monophyletic group with- be symplesiomorphic for the Cyrtandroideae, as all in it (Smith, 1996). The inclusion of chromosome taxa with available data for this character (Saint- numbers (which were excluded due the large num- paulia, Streptocarpus, and Cyrtandra Forster & For- ber of character states) might have removed Ges- ster) possess a three-trace trilacunar node except Annals of the Missouri Botanical Garden

Aeschynanthus, which has a one-trace trilacunar biogeographic distribution of these taxa. The Cyr- node common to the Gesnerioideae. tandroideae (excluding the Klugieae) are distrib- uted almost exclusively in the paleotropics with a few temperate European and Asian species. Two African genera were included in this analysis, both The placenta in the Gesneriaceae is either intact of which are in a single clade (Saintpaulia and or divided to the base (Ivanina, 1965). This char- Streptocarpus). Only one of the European taxa (Ra- acter was included in a cladistic analysis of mor- monda L. C. Richard) has been included in this phological data and served as a character state that analysis; therefore nothing can be inferred regard- brought the Episcieae, Beslerieae, and Napean- ing the origin of these taxa at this time. theae together in a single clade (Smith, 1996) as Members of the tribe Klugieae range from India the only taxa sampled that had divided placentae. to south China, Taiwan, the Philippines through Although this character state is consistent with the Malaysia, Indonesia, and into New Guinea. Dis- relationship between the Napeantheae and Besler- crepancies from this distribution include a single ieae, the character state is homoplastic between the species of Rhynchoglossum Blume found in Central Episcieae and Na~eantheaeIBeslerieae based on and South America. The presence of Rhynchoglos- the data presented here (Fig. 4). sum azureum (Schlecht.) B. L. Burtt in the Neo- tropics represents a secondary dispersal event in STEM MODIFIC4TION the family, because all other members of the Klu- Several members of the Gesneriaceae possess gieae are found in the Old World. modifications of the stems (rhizomes and tubers), The Gesnerioideae, are almost exclusively neo- presumably as adaptations to periodic dry seasons tropical, but with the inclusion of the Coronanthe- (Wiehler, 1983). The presence of scaly rhizomes is reae within this subfamily the Gesnerioideae now found almost exclusively, and is widespread, within encompass several Australian and South Pacific is- the Gloxinieae (Wiehler, 1983). Among the taxa land species. sampled here, the presence of scaly rhizomes Literature Cited serves as a synapomorphy for the tribe Gloxinieae, although they are not known from the woody genus Acker~nan,J. D. 1986. Coping with he epiphytic exis- tence: Pollination strategies. Selbyana 9: 5240. Solenophora. Scaly rhizomes also are known from Bentham, G. 1876. Gesneriaceae. Pp. 990-1025 in G. the Cyrtandroideae, including Titanotrichum (Kao Bentham c91 J. D. Hooker (editors), Genera Plantarum. & DeVol, 1972; Wang et al., 1992). Love11 Reeve, London. Tubers are widespread among species of Sinnin- Boggan, J. K. 1991.".A morphological study and cladistic analysis of Sinmingia and associated genera with par- gin indluding Lietzia,which has recently been com- ticular reference to Lembocar~~us,Lietzia, Paliavama, bined into Sinningia (Wiehler & Chautems, 1995). and Vanhouttea (Gesneriaceae: Gloxinieae). M.S. The- Although tubers serve to unite these genera, and to sis, Cornell University, Ithaca, New York. separate them from the Gloxinieae, tubers are not Bremer, K. 1988. The litnits of amino acid serluence data known from Paliavana or Vanhouttea. However, not in angiosperm phylogenetic reconstruction. Evolution 42: 795403. all species of Sinningia are tuberous, and the lack . 1994. Branch support ancl tree stability. Cladis- of tubers in these species can be regarded as intra- tics 10: 295-304. tribal or intra-generic variation. Tubers also are Burtt, B. L. 1962. Studies in the Gesneriaceae of the Old known from several species in the Episcieae (Dry- World XXJL Tentative keys to the tribes ancl genera. Notes Roy. Bot. Garcl. Edinburgh 24: 205-220. monia, Chrysothemis Dcne., Nautilocalyx Lind. ex . 1965. The transfer of Cyrtamd'romoea from Ges- Hanst., Paradrymonia Hanst., and Rhoogeton neriaceae to Scrophulariaceae, with notes on the classi- Leeuwenberg) as well as one member of the Glox- fication of that fanlily. Bull. Bot. SUN. India 7: 734. inieae (Lembocarpus Leeuwenberg). Further studies . 1977. Classification above the genus, as exem- that include these taxa will hopefully resolve the plified by Gesneriaceae, with ~arallelsfrom other groups. P1. Syst. Evol., Supplement 1: 97-109. number of times tubers have originated within the cY H. Wiehler. 1995. Classification of the fanlily Gesneriaceae. Gesneriaceae. Gesneriana 1: 1-4. Cantino, P. D. 1992. Evidence for a polyphyletic origin of the Labiatae. .Ann. Missouri Bot. Garcl. 79: 361-379. BIOGEOGR 4PHI Chase, M. W., D. E. Soltis, R. G. Olmsteacl, D. Morgan, The traditional division of the Gesneriaceae into D. H. Les, B. D. Mishler, M. R. Duvall, R. .A. Price, H. G. Hills, Y.-L. Qiu, K. A. Kron, J. H. Rettig, E. two subfamilies (excluding the Klugieae, which Conti, J. D. Palmer, J. R. Manhart, K. J. Syts~na,H. J. may stand best as a third subfamily) is well sup- Michaels, W. J, Kress, K. G. Karol, W. D. Clark, M. ported in this analysis and is in agreement with the Hedrkn, B. S. Gaut, R. K. Jansen, K.-J. Kiln, C. F. Volume 84, Number 1 Smith et al. 65 1997 Tribal Relationships in Gesneriaceae

R'itl~pre,J. F. Smith. 6. R. Furtiier, S. H. Strau>s. Q.-Y. R J. A. S~\errr.1994. Conibinillg data in pli!- Xiang. G. RI. Plunkett. P. S. Soltis. S. RI. S~\etison.S. logrnrtic s)stetl~atics: .An rmpirical approach using E. Williams, P. A. Gadrk, C. J. Quinn. L. E. Eguiarte. thrre ti~olrculardata sets iri thr Solanacear. S!st. Biol. E. Golrnberg. (;. H. Lrarn. Jr.. S. R: Grahatl~.S. C. H. 43: 467481. Barrett. S. Da)ananclati cY \., A. .Albert. 1993. Ph!lo- . B. Bremer. Ii. \.I. Scott cY J. D. Paltl~er. 1003. genetics of seed plant>: An anal~siof rluclroticle se- A parhi~tlon! anal)>is of the .Isteridae SCIISU Jato tlueticrs from thr plas~idgrnr r6cL. .Ann. hlissouri Bot. based 011 I-bcL sequrnc.r>. Ann. \lissouri Bot. Gartl. Gard. 80: 528-580. 80: 700-722. Clark. L. (;.. Rl Zhang R J. F. Rktidrl. 1005. A ph)log- , H. J. hlichaels. F;. \.I. Scott R J. D. Pal~nrr. rny of the gra>s l'a~nily (Poacrar) based on r~rliiFse- 1002. hlonoph>l) of ~hrAstericlae arid itlrn~ifica~ionof qucllce data. S!st. Bot. 20: 436460. their major litieagrs inferrrcl from DKA srtlurncrs of Crisci. J. 1;. RI. \.I. Cigliano. J. J. Rlorronr cY S. Roig- I-bcL. Ann. \.li>souri Bot. 6drtl. 70: 240-265. Juileti~. 1001. Historical biogeography of souther11 Scotland. R. R:. J. A. S~\errr.P. .A. Rer\rs R R. G. Olm- South .America. S>st. Zool. 40: 152-171. stracl. 100,5. Highrr lrvrl s)s~rrlia~icsof Acanthacrar Croticl~cist.A. 1981. An Integra~rdSystrm of Clashificalion clrtrrtl~itied by chloropla>~DKA seclurnces. A~ner.J. (IS Flo~rritigPla~its. Colut~ibia Utiiv. Prrss. he!\ York. Bot. 82: 26ri-275. Dahlgren. R. 1975. A system of clashificalion of he an- Singh. 1~.R 1). K. Jain. 1078. Floral atialom) and s!,- gio>per~nsto be used to de~notistratethr distributiori of tematic po>itiori of C\rtarirlr-oniof,n. Proc. hatl. Acad. characters. Bot. Not. 128: 119-147. Sci. India 87: 71-74. Donoghur. hl. J. R P. D. Cantino. 1984. Thr logic ancl Skog. L. E. 1084. -\ rr\ie\\ of chromosome nunil~rrsill liit~ita~iotlsof the outgroup sccb31i1utionapproach lo c.1~1- the (;rsneriacrar. Srlb!ana 7: 8.58-873. distic anal!sis. S!>I. Bot. 9: 192-202. . 1005. .A pos>iblr rralignmrnt of Kc'.sin H. E. . R. 6. Olm>~ead.J. F. Smith R J. D. Paltl~er. hloore (Grsneriacrar). .Atl~er.J. Bot. 82: 161. 1992. Ph!logetie~ic rrlationship> of Dip>acalrs basrtl R F. F. dr Jrsus. 1996. .A r.e\ir\\ of Resicr H. E. oti rhcL secluencrs. Ann. Rlis>ouri Bot. Gard. 70: 333- Rloore (Grsneriacrar). Biollania (in prr>s). 345. Smith. J. F. 1906. Tribal relatioriships ~ithinthr (;e>- Fami>. S. J. 1070. hlethods for co~npu~ingRagnrr trerh. tieriacrar: A claclislic atialysis of morpliological data. S! st. Zool. 19: 83-02. SJ~.Bot. (in press). , A. (;. Iilugr R RI. J. Eckardt. 1970. .A tiu~nrrical R F;. J. Sytsma. 1994a. E\olution in thr .Andean approach to phylogrnrtic s!>~e~natic>.Syst. Zool. 19: epiphytic gem> Coli~nir~ea(Grstirriaceae): Part I. hlor- 172-101. phological \ariatiori. Syst. Bot. 19: 220-235. Fritsch. F;. 1803-1894. Grsnrriaceae. Pp. 133-185 irr: &- . 1004b. Evolution in the Andean rpi- A. Engler R K. Prantl (rditors), llir Natiirlichen Pflan- ph!lic grnus Coli~niric~a((Anrriacrae): Part JJ. Chlo- zrnk~tl~ilien.hl. 4(3b). R-ilhrlt11 Engrlmann. Lripzig. roplast LIKA rrstriction hitr \arialion. Slst. Bot. 10: Hr)wood. \ H. 1078. Flowering Platit> of the R;~rlcl. 3 17-336. Prrntice Hall. Etlgleu-ood Cliffs. New Jersr). Jtinis. RI. A,. Ii. B. hl!a~nho. D. H. Grlfancl R RI. A. D. R- , 1004c. hloleculrs ant1 ~norpholog): Brow. 1988. DNA srquencing and dirrct sequencing Congruencr of data it1 Cllirr~ri~cr(Ge>neriacrar), P1. (IS poly~nerasr chain reac~ioti-a~nplifirclDNA. Proc. S!d. E\ol. 104: 37-52. Katl. Acad. Sci. L.S.A. 8%;: 043G9440. . K. 1). Rrowri, C. L. Carroll R D. S. Drnton. 1997. J\atiina. L. I. 1965. Applica~iotiof thr carpological 111elh- Familial placemrnl oS C,rinr~rlrorrloecr. 7i'icrrroirir~h~irri. ocl to thr laxotiom) of Ge>nrriacrae. Notrs Ro). Bot. and Suncrngo: Threr problematic grnrra of the Lami- (Gard. Edinburgh. 26: 383-402. ales. Taxoti (in prrss). Judcl. R: S.. It. R. Sanders cY \.I. J. Dotioghur. 1994. . K. J. S!tsrlla. J. S. Sliortr~akrrR K. L. Sttcith. Angiosperm l'a~nily pairs: Preliminar! phylogenrtic 1002. A clualitati\r cotl~parihonof total cellular DNA anal!>?>. Harvarcl Pap. Bol. 5: 1-51. rwlraclion protocols. Ph)~ochetl~.Bull. 23: 2-9. Kao, h1.-T, R C. E. Devol. 1072. The Ge>rlrriaceae of Sugiura. 11. 1989. The chloropla>t chro~nosomrsin land Tai~\ati.Taiwania 17: 142-160. . Atiti. Re\. Cell Riol. 5: 51-70. F;re>>.W. J. 1986. Thr >!strnlatic distribution ol'\ascular . 1992. The cliloroplast grnolne. P1. \lolec. Biol. epiph!tr>: ,411 updatr. Se1b)ana 9: 2-22. 10: 149-168. kc is^. L. P. 1990. Revision of Ht,iii)it,//n(Gesneriacear). S~\c~fl'(~rd.D. L. 1Y13. PA4LP:Phylogenrtic .Anal>& Lsing S!>I. Bol. 15: 720-735. Parsimon!. \er>ioti 3.1.1 Compu~rrprogram clistributrtl b! \ladidi>on, D. R. 1091. The di>covrr) atid importance of thr Illinois hatural Histol~Sul-vr!. Clia~npaigti,Illinois. tl~cdliplrislands of mo>t-par>i~notiioustrres. S!ht. Zool. cY R: P. \ladclison. 1987. Rrconstructing atice>- 40: 315-328. tral charac~erslales under Ragtier parsirnon!. Math. Madidison. W. P.. 11. J. Donoghur R 1). R. Rladdison. Biosci. 87: 199-220. 1984. Oulgroup atialysis atid parsimon). Syhl. Zool. 33: Takh~ajau,A. L. 1980. 0u1li11eof a clas>ificalion oi'flo~\-- 83-103. rring platits (\lagnolioph>ta). Bot. Rr\. 46: 225-350. hladi>on. RI. 1077. Lscular epiphytrh: Their s!stematic Thornr. It. F. 1076. A ph>logrnrtic classification of ~hr occurrence ancl salient ka~ure>.Selbyana 2: 1-13. .Angioa[)er~nae. Pp. 3%;-106 iri \I. K. Hecht. R: C. Olmstead. R. G.cY J. D. Palmer. 1994. Chloro1,last 1)N.A Steer? R B. R>llace (editors). tC\olutionary Riolog>.bl. >!>lenlatics: .A re\ir~\of n~e~hodsand data anal>si>. 9. Plenuni Prehs. Kew York. .her. J. Bot. 81: 1205-1224. . 1983. Proposrd neT\. realignnlrn~in the angio- cY P. A. Rre\es. 1995. E\idrnce for the pol!- spermb. horclic J. Bot. 3: 8%;-117. ph)l! of the Scrophulariaceae based on chloropla>~rbcL . 1992. At1 updated ph!logrnrtic cJahhific.ation of and nr//lF srclurnce>. i\tin. Viiiou~i601.Carcl. 82: flo\\eriug platlt> (Rlagnolioph! ta). BOI.Re\. (Lancaster) 170-103. 46: 225-359. Annals of the Missouri Botanical Garden

Pang. El T.. K. Pan cY Z. Li. 1992. keys to the Gesner- Eolfe. K. 1991. Protein-coding genes in chloroplast DNA: iaceae oS China. Edinburgh J. Bot. 49: 5-74. Co~npilationof nucleotide sequences, database entries and Wiehler, H. 1983. A synopsis of the neotropical Gesner- rates of molecular evolution. Pp. 467-482 in L. Bogoracl cY iaceae. Selhyana 6: 1-249. I. K. Iasil (editors). The Photosynthetic Apparatus: Volec- R .A. Chaute~ns. 1995. A reduction of Lietzia to ular Biology and Operation. vol. 7B, Cell Culture and So- Sinmingia. Gesneriana l: 5-7. matic Cell Genetics in Plants. Academic Press, New York.