Familial placement and relations of Rehmannia and Triaenophora (Scrophulariaceae s.l.) inferred from five gene regions^†

Abstract

Accurate classification systems based on evolution are imperative for biological investigations. The recent explosion of molecular phylogenetics has resulted in a much improved classification of angiosperms. More than five phylogenetic lineages have been recognized from Scrophulariaceae sensu lato since the family was determined to be polyphyletic; however, questions remain about the genera that have not been assigned to one of the segregate families of Scrophulariaceae s.l. Rehmannia Liboschitz and Triaenophora Solereder are such genera with uncertain familial placement. There also is debate whether Triaenophora should be segregated from Rehmannia. To evaluate the phylogenetic relations between Rehmannia and Triaenophora, to find their closest relatives, and to verify their familial placement, we conducted phylogenetic analyses of the sequences of one nuclear DNA (ITS) region and four chloroplast DNA gene regions (trnL-F, rps16, rbcL, and rps2) individually and combined. The analyses showed that Rehmannia and Triaenophora are each strongly supported as monophyletic and together are sister to Orobanchaceae. This relation was corroborated by phytochemical and morphological data. Based on these data, we suggest that Rehmannia and Triaenophora represent the second nonparasitic branch sister to the remainder of Orobanchaceae (including Lindenbergia).

Nothing in biology makes sense except in the light of evolution. —Theodosius 20

As a corollary to Theodosius Dobzhansky's famous quote, understanding the evolutionary history of organisms can improve our understanding of biology. Recent molecular phylogenetic analyses have resulted in a major rearrangement of angiosperm classification that now better reflects the evolutionary history of these plants. However, many species remain unsampled and thus unplaced in the new classification scheme. A series of molecular systematic studies of Scrophulariaceae s.l. have revealed that the traditionally circumscribed Scrophulariaceae is polyphyletic (51; 19; 50; 53). These studies have resulted in recircumscriptions and new descriptions of families to encompass the monophyletic lineages that were recovered. However, questions remain about the genera that have not been assigned to one of the segregate families of Scrophulariaceae s.l., such as Rehmannia and Triaenophora.

The genus Rehmannia Liboschitz consists of six species endemic to China (17), in which R. glutinosa is widely distributed in central China and cultivated in Japan and Korea (59) and is an important species in traditional Chinese medicine. Rehmannia has longstanding controversies surrounding its systematic placement at both the generic and familial levels. It was originally included within Digitalis (26) and was established by Liboschitz in 1835 because of its corolla shape and fruit dehiscence (24). Since then, Rehmannia has usually been placed in the tribe Digitaleae in Scrophulariaceae s.l. (12; 65; 36; 17). Others have suggested Rehmannia to be part of subfamily Cyrtandroideae in Gesneriaceae based on reports of its unilocular ovary (18; 29; 65; 36; 15).

Some species initially described as Rehmannia have been segregated as monotypic, i.e., Triaenophora and Titanotrichum (65). Titanotrichum was transferred to Gesneriaceae (65), a move that is supported by both morphological and molecular data (15; 78, 79, 80, 75; 63, 64; 54). On the contrary, Triaenophora, which now contains three species (65; 17; 38), has received almost no attention besides 36 who returned it to Rehmannia, and 17 who later segregated it.

As for the phylogenetic relations of Rehmannia and Triaenophora, one issue is whether there is any phylogenetic affinity with Digitalis (26). Digitalis, and Rehmannia/Triaenophora are remarkably different from each other in their corolla shapes and a series of morphological characters such as inflorescence morphology and fruit dehiscence, as well as geographic distribution (17; 77). All phylogenetic analyses of Scrophulariaceae s.l. have placed Digitalis in Plantaginaceae (50; 7; 5; 53; 68). A second, equally likely probability is that Rehmannia and Triaenophora are members of Gesneriaceae because Titanotrichum, a segregate from Rehmannia, has been convincingly placed there (15; 78, 79, 80, 75; 63, 64; 54).

Recently, Rehmannia was included in a cladistic analysis of DNA sequence data for the further disintegration of Scrophulariaceae, in which a single species of Rehmannia was sister to Lancea and Mazus (subfamily Mazoideae of Phyrmaceae sensu 8) and in a clade with Paulownia, Orobanchaceae, and Phrymaceae (53). Sampling all six species of Rehmannia, 3, using nuclear ITS and chloroplast trnL-F and rps16 sequences, placed Lindenbergia as sister to Rehmannia. However, the low sampling of species other than Rehmannia weakens their result (only four species including outgroups are sampled outside Rehmannia).

As we have outlined, the familial placement of Rehmannia and Triaenophora has not been well resolved either from a morphological or a molecular perspective. The debate regarding the familial placement of Rehmannia and Triaenophora in morphology is mainly based on the selection of characters used for classification (65; 15), many of which have been revealed to be convergent (50) and thus provide little insight regarding the evolutionary relations for the classification system. Meanwhile, the controversy in molecular data lies in the lack of sufficient sampling among the putative relatives of Rehmannia and Triaenophora and the relation between these two genera.

Greater sampling of taxa putatively close to Rehmannia and Triaenophora and the analysis of additional DNA regions are necessary to determine the familial placement of these two genera and their closest relatives in Lamiales s.l. This study is thus conducted with a comprehensive sampling of putative relatives of Rehmannia and Triaenophora in Lamiales s.l. and the use of five DNA regions (ITS, trnL-F, rps16, rbcL, and rps2) that have been shown to be particularly informative in the Lamiales s.l. (19; 63, 64; 49; 92; 50; 8; 5; 53; 89; 68). The goal of this study was to (1) evaluate the phylogenetic relation between Rehmannia and Triaenophora, (2) find their closest relatives, and thereby (3) verify their familial placement.

MATERIALS AND METHODS

Taxon sampling

We sampled all six species of Rehmannia and two of three species of Triaenophora. To fully examine the putative relatives of Rehmannia and Triaenophora, we selected one species of Paulownia, two genera of Gesneriaceae, six genera of Plantaginaceae (including Digitalis), seven genera of Scrophulariaceae sensu stricto, six genera of Phyrmaceae (four in Phrymoideae and two in Mazoideae including four species of Mazus in addition to the one species sampled in previous studies), 12 representative genera of four major clades in Orobanchaceae sensu lato (including the nonparasitic genus Lindenbergia), three genera of Acanthaceae, two genera of Bignoniaceae, one genus of Lamiaceae, and one genus of Pedaliaceae. Taxon sampling was based on recent molecular systematic studies (19; 63, 64; 49; 92; 50; 8; 5; 53; 89; 10; 68). Voucher specimens are deposited in the Herbarium of the Institute of Botany, Chinese Academy of Sciences (PE). The materials studied and details of voucher specimens are shown in Table 1.

Table 1. Species, voucher with collection locality and GeneBank accession number for taxa included in this study (new sequences are in boldface). NA: not applicable; PE: Herbarium, Institute of Botany, Chinese Academy of Sciences where the voucher specimens were deposited.

		GenBank accession number
Taxon	Voucher, collection locality and citation	ITS	trnL-F	rps16	rbcL	rps2
Ingroups
Rehmannia
R. chingii H. L. Li	XZ-2004–04–004; Zhejiang, China. (PE)	EF363673	EF363679	FJ172696	FJ172724	FJ172710
R. elata N. E. Brown	3	DQ069315	DQ856496	DQ856490	—	—
R. glutinosa (Gaert.) Libosch. ex Fisch. et Mey.	XZ-2004–04–005; Beijing, China. (PE)	EF363674	EF363680	FJ172697	FJ172725	FJ172711
R. henryi N. E. Brown	XZ-2004–04–002; Hubei, China. (PE)	EF363671	EF363677	FJ172694	FJ172722	FJ172708
R. piasezkii Maximowicz	XZ-2004–04–001; Hubei, China. (PE)	EF363670	EF363676	FJ172693	FJ172721	FJ172707
Rehmannia solanifolia Tsoong et Chin	XZ-2004–04–003; Chongqing, China. (PE)	EF363672	EF363678	FJ172695	FJ172723	FJ172709
Triaenophora
T. rupestris (Hemsl.) Solereder	XZ-2004–04–009; Hubei, China. (PE)	EF363675	EF363681	FJ172698	FJ172726	FJ172712
T. shennongjiaensis X. D. Li	XZ-2007–0522; Hubei, China. (PE)	FJ172741	FJ172690	FJ172704	FJ172732	FJ172717
Acanthaceae
Barleria lupulina Lindl.	45	AF169751	—	—	—	—
B. prionitis Lindl.	50	—	—	—	L01886	AF248247
Elytraria crenata Vahl.	50	—	—	—	AF188127	—
E. imbricata Vahl.	45	AF169852	AF061819	—	—	—
Thunbergia alata Bojer ex Sims	45; 50; 53	AF169850	AJ608564	AJ609131	—	AF248248
T. usumbarica Lindau	50	—	—	—	L12596	—
Bignoniaceae
Catalpa speciosa Warder ex Engelm	53; 37	AY486307	AJ608599	AJ609197	—	—
Catalpa sp.	50	—	—	—	L11679	AF248256
Kigelia africana Benth.	50; Gutierrez and Freeman, unpublished data	AY178638	—	—	AF102648	U48764
Gesneriaceae
Streptocarpus caulescens Vatke	53	NA	AJ608601	AJ609135	—	—
Streptocarpus holstii Engl.	50	NA	—	—	L14409	—
Titanotrichum oldhamii (Hemsl) Solereder.	75	NA	AY423129	—	AF206829	—
Lamiaceae
Lamium purpureum L.	50; 53; Sudarmono and Okada, unpublished data	AB266244	AJ608588	AJ609175	U75702	AF248259
Orobanchaceae
Alectra sessiliflora Benth.	50; 89	AY911210	—	—	AF026820	U48742
Boschniakia strobilacea A. Gray	50; 89	AY911215	—	—	AF26818	U48758
Buchnera glabrata Benth	89	AY911216	NA	NA	NA	NA
Castilleja linariifolia Benth	50; 69	—	—	EF103788	AF026823	U48739
Castilleja sulphurea Rydberg	8	AF478944	AF479008	—	—	—
Lindenbergia philippensis Benth.	50; 53; 89	AY911231	AJ608586	AJ609169	AF123664	AF055151
Melampyrum lineare Lam.	50; 34	—	AF482608	—	AF026834	—
M. sylvaticum L.	50; 89	AY911232	—	—	—	AF055148
Melasma scabrum Berg.	50; 89	AY911233	—	—	AF190904	U48743
Orobanche corymbosa (Rydb.) Ferris	50; 89	AY911236	—	—	U73969	U48760
Orobanche hederae Duby	14	—	—	AJ431050	—	—
O. minor Sm.	41	—	AJ007724	—	—	—
Pedicularis attollens (A.) Gray	69	—	EF103899	EF103821	—	—
P. foliosa L.	50; 58	AY949679	—	—	AF026836	U48740
Schwalbea americana L.	89	AY911252	NA	NA	NA	NA
Seymeria laciniata Standl.	69	—	EF103898	EF103820.	—	—
S. pectinata Pursh	50; 89	AY911253	—	—	AF026837	AF055141
Tozzia alpina L.	50; 89	AY911258	—	—	AF026843	U48754
Paulownia
P. tomentosa (Thunb.) Steud	8; 53	AF478941	AF479005	AJ609153	L36447	AF055155
Pedaliaceae
Sesamum indicum L.	50; 8; 53	AF478946	AF479010	AJ609226	L14408	AF248261
Phrymaceae
Berendtia laevigata B. L. Rob et Greenm.	53	—	AJ608615	AJ609208	—	—
B. rugosa (Benth.) Gray	9	AY575398	—	—	—	—
Hemichaena fruticosa Benth.	8; 53	AF478921	AJ608632	AJ609179	—	—
Lancea tibetica Hook. f. et Thoms.	XZ-2007–0525; Sichuan, China. (PE)	FJ172736	FJ172685	FJ172699	FJ172727	FJ172713
Mazus gracilis Hemsl.	XZ-2007–058; Henan, China. (PE)	FJ172738	FJ172687	FJ172701	FJ172729	FJ172715
M. japonicus (Thunb.) O. Kuntze.	XZ-2007–051; Beijing, China. (PE)	FJ172737	FJ172686	FJ172700	FJ172728	FJ172714
M. omeiensis Li.	XZ-2007–0515; Sichuan, China. (PE)	FJ172739	FJ172688	FJ172702	FJ172731	—
M. reptans N. E. Br.	8	AF478940	NA	NA	NA	NA
M. spicatus Vaniot.	XZ-2007–0514; Henan, China. (PE)	FJ172740	FJ172689	FJ172703	FJ172730	FJ172716
M. stachydifolius (Turcz.) Maxim	53	—	AJ607432 AJ607433	AJ609167	—	—
Mimulus aurantiacus Curtis	50; 8; 53	AF478917	AF478982	AJ609163	AF026835	AF055154
M. tenellus var. tenellus Bunge	XZ-2007–053; Henan, China. (PE)	FJ172742	FJ172691	FJ172705	FJ172733	FJ172718
Mimulus szechuanensis Pai	XZ-2007–0523; Sichuan, China. (PE)	FJ172743	FJ172692	FJ172706	FJ172734	FJ172719
Phryma leptostachya L.	53	—	AJ430928	AJ609150	—	—
P. leptostachya L. var asiatica Hara	8; XZ-2007–061; Henan, China. (PE)	AF478924	—	—	FJ172735	FJ172720
Plantaginaceae
Antirrhinum majus L.	50; 53	NA	AJ608634	AJ609218	L11688	U48766
Chelone obliqua L.	50; 53; 88	NA	DQ531203	AJ609220	AF026824	U48770
Collinsia grandiflora Lindley	50	NA	—	—	AF026825	AF248252
C. heterophylla R.Grah	88	NA	DQ531198	—	—	—
C. tinctoria Hartw. ex. Benth	5	NA	—	AY492200	—	—
Digitalis obscura L.	4; 2	NA	AF486418	AY218799	—	—
D.s purpurea L.	50	NA	—	—	L01902	U48767
Plantago coronopus L.	2	NA	—	AY218801	—	—
P. lanceolata L.	50; 62	NA	AY101952	—	L36454	—
P. major L.	50	NA	—	—	—	AF248254
Veronica arvensis L.	50	NA	—	—	—	U48768
V. persica Poir.	50; 4	NA	AF513336	—	L36453	—
V. campylopoda Boiss.	2	NA	—	AY218811	—	—
Scrophulariaceae s.s.
Alonsoa unilabiata (L. f.) Steud.	50; 53	NA	AJ608620	AJ609217	AF026821	AF248262
Buddleja davidii Franchet	50; 53	NA	AJ608612	AJ609204	L14392	AF248264
Leucophyllum frutescens I. M. Johnston	50; 53	NA	AF380873	AJ609171	AF123665	AF055156
Myoporum mauritianum A. DC.	50; 53	NA	AJ608582	AJ609161	L36445	—
M. parvifolium R.Br.	50	NA	—	—	—	AF055157
Nemesia strumosa Benth.	50; 53	NA	AJ608631	AJ609159	AF123663	AF248265
Scrophularia californica Cham.& Schldl.	50; 53	NA	—	AJ609224	L36449	U48762
Scrophularia ningpoensis Hemsl.	16	NA	AY695886	—	—	—
Verbascum arcturus L.	53	NA	—	AJ609128	—	—
V. blattaria L.	50	NA	—	—	—	U48763
V. thapsus L.	50	NA	—	—	L36452	—
V. speciosum Schrad.	44	NA	AJ492271	—	—	—
Outgroup
Calceolariaceae
Calceolaria mexicana Benth	50; 53	NA	AJ608611	AJ609202	AF123669	AF055162

DNA extraction, PCR amplification, and sequencing

Total DNA was extracted from silica-gel-dried leaf material using the CTAB method following the protocol of 60 and used as the template in the polymerase chain reaction (PCR). The entire ITS region, comprising ITS1, 5.8S rDNA, and ITS2, was amplified with primers ITS1 and ITS4 (84). The trnL-F region was amplified with primers c and f of 67. The rps16 intron was amplified with primers rps16-2F and rps16-R3 (14). The rbcL and rps2 gene regions were amplified with primers RH1 and Z1352R (51; 86; 50), and rps2-18F and rps2-661R (19), respectively. PCR products were purified with the UNIQ-10 PCR purification kit (Sangon, Shanghai, China). Sequencing primers were the same as amplification primers. Automated sequencing was performed on a MegaBACE 1000 automatic sequencer (Amersham Biosciences, Sunnyvale, California, USA) using manufacturer's protocols. The DNA sequences reported in the paper have been deposited in GenBank with accession numbers shown in Table 1. For some genera, sequences of the five regions sampled here were available only for different species. Rather than limit our sampling of either genera or sequences, we combined sequences from different species into a single genus in our analyses provided that there was evidence that the genus was monophyletic and the genus was not the primary focus of our study.

Sequence alignment and phylogenetic analysis

Sequence alignments were made with the program CLUSTAL_X (72) and refined manually for the maximization of sequence homology using the program BioEdit 5.0.9 (28).

Parsimony analysis for each matrix was carried out using maximum parsimony (MP) methods in the program PAUP* version 4.0b10 (66). All characters and character-state changes were specified as unordered and weighted equally, and gaps were coded as missing data. Heuristic searches were performed with 1000 replicates of random addition, one tree held at each step during stepwise addition, tree-bisection-reconnection (TBR) branch swapping, Multrees in effect, and steepest descent off. To examine the robustness of various clades, we ran a bootstrap analysis (22) with 500 replicates using a heuristic search with 1000 replicates of random sequence addition and TBR branch swapping.

Bayesian inference (BI) was conducted using the program MrBayes version 3.0b4 (61). The program Modeltest 3.06 (56) was employed to determine the appropriate model of sequence evolution for each DNA data set. Four chains of Markov chain Monte Carlo (MCMC) were each run for 10000000 generations and were sampled every 100000 generations, starting with a random tree. For each run, the first 50 samples before the chains reached stationarity were discarded as burn-in. Posterior probability (PP) was used to estimate robustness.

For combined sequence data, the incongruence-length-difference (ILD) test (21) was conducted, using the partition homogeneity test in PAUP* 4.0b10 (66), to examine the congruence between nuclear ITS and chloroplast data sets. Test settings were 100 random stepwise additions and 1000 replicates of heuristic search with TBR branch swapping (21). The resulting P value was used to determine whether the two data sets had significant incongruence (P < 0.05). We excluded species from the combined data set if they only had ITS or cpDNA sequences available.

Topological congruence between the trees constraining Phrymaceae as monophyletic and no constraint was evaluated with the 71 test using PAUP* 4.0b10 (66). The phylogenetic analysis herein was divided into two steps as follows. We first conducted cladistic analyses of combined cpDNA (trnL-F, rps16, rbcL, and rps2) from all sampled taxa. The four chloroplast regions (cpDNA) included in this study formed a single linkage group as part of the chloroplast genome, so conflicts that might arise between data partitions from different sources subject to different evolutionary histories should not exist (50). The genus Calceolaria was selected as the outgroup based on 50. Attempts to use ITS sequences at this level resulted in numerous ambiguities, and analyses of these sequences resulted in spurious relations among some taxa. Second, based on the described analyses, together with results of previous molecular systematic studies (8; 53), we selected different taxa for separate and combined analyses of nrDNA ITS and combined cpDNA (e.g., excluding Scrophulariaceae s.s., Plantaginaceae, Gesneriaceae), focusing on clades presumably closest to Rehmannia/Triaenophora and Orobanchaceae. The representative species of Acanthaceae, Bignoniaceae, Lamiaceae, and Pedaliaceae were selected as the outgroups.

RESULTS

A comparison of the sequences we generated with those published from 3 indicates that the ITS sequences are identical to each other, the rps16 sequences were 99.76–100% similar, and the trnL-F sequences were 99.29–100% similar. These comparisons confirm the accuracy of both the sequences generated herein and those of 3 for Rehmannia.

Analyses with all sampled taxa

Data for one or two of the four chloroplast regions are missing for 15 genera, but no genus was missing more than two sequences of the four genes. The entire cpDNA data set consists of 4125 bp, of which 2554 (61.9%) were constant, 721 (17.5%) were variable but uninformative, and 850 (20.6%) were parsimony informative. Parsimony analyses resulted in nine trees of 3511 steps each (consistency index [CI] = 0.623; retention index [RI] = 0.608). One most parsimonious (MP) tree of cpDNA data (Fig. 1) was congruent with the Bayesian tree (the best-fit model GTR + I + G) in topology. The MP tree comprised six main clades labeled A–G. Titanotrichum oldhamii was sister to Streptocarpus in Gesneriaceae (clade A) with high support (bootstrap support [BS] = 94%; posterior probability [PP] = 100%). Clades B and C contained the species of Plantaginaceae (BS = 91%; PP = 100%) and Scrophulariaceae sensu stricto (BS = 87%; PP = 100%), respectively. Clade D was Acanthaceae, Pedaliaceae, Lamiaceae, and Bignoniaceae (BS = 73%, PP = 100%) and clade E was Mazoideae (BS = PP = 100%). Clade F comprised Phrymoideae with BS = PP = 100%. Clade G included Rehmannia, Triaenophora, and Orobanchaceae with BS = 62% and PP = 99%. Rehmannia and Triaenophora formed one strongly supported lineage (BS = 100%; PP = 100%) and was sister to Orobanchaceae, that was likewise strongly supported as monophyletic (BS = 88%; PP = 100%). Paulownia was sister to clade G with low support.

Figure 1

One of nine most parsimonious trees generated from analysis of combined chloroplast for all sampled taxa. Branch lengths are proportional to number of nucleotide substitutions (scales represent 10 substitutions). Bootstrap (BS) values (≥50%) are above the branches; Bayesian posterior probabilities (PP) (≥90%) are below the branches. Maz = Mazus, Mim = Mimulus, Rehm = Rehmannia, Triaen = Triaenophora.

Analyses with selected taxa

ITS analysis

The aligned sequences of ITS had 656 bp, of which 209 (31.9%) were constant, 94 (14.3%) were variable but uninformative, and 353 (53.8%) were parsimony informative. Parsimony analysis resulted in 10 trees of 2004 steps each, CI of 0.445, and RI of 0.591. The MP tree (Fig. 2) was congruent with the Bayesian tree (the best-fit model GTR + I + G) in topology except for the position of Paulownia. Paulownia was sister to Phrymoideae (clade F) in the MP tree, but sister to the group that includes clades E, F, and G in the Bayesian tree. The ITS MP tree comprised three main clades labeled as E, F, and G, that correspond to the clades recovered in Fig. 1. Rehmannia, Triaenophora, Paulownia, Mazoideae, Phrymoideae, and Orobanchaceae formed a monophyletic group with maximum support. Clade E was Mazoideae, and clade F comprised all sampled species of Phrymoideae with BS = 60% and PP = 100%. Clade G was also recovered as monophyletic including Rehmannia, Triaenophora, and Orobanchaceae with BS = 92% and PP = 100%. Rehmannia and Triaenophora, each as a monophyletic group, formed one strongly supported lineage (BS = 89%; PP = 94%), which was sister to Orobanchaceae. Orobanchaceae was likewise strongly supported as monophyletic (BS = 80%; PP = 100%). The low supports for the relations within Orobanchaceae were probably due to the sparse taxon sampling within the family. Templeton's test indicated incongruence between the trees constraining Phrymaceae (Phrymoideae and Mazoideae) as monophyletic and that no constraint was insignificant (P = 0.6103).

Figure 2

One of 10 most parsimonious trees generated from the ITS data for selected taxa. Branch lengths are proportional to number of nucleotide substitutions (scales represent 10 substitutions). Branches marked by an asterisk indicate the topological discordance between most parsimonious (MP) and Bayesian trees. Bootstrap (BS) values (≥50%) are above the branches; Bayesian posterior probabilities (PP) (≥90%) are below the branches. Maz = Mazus, Mim = Mimulus, Rehm = Rehmannia, Triaen = Triaenophora.

Analysis of combined chloroplast data

The combined chloroplast data set consisted of 3880 positions, of which 2783 (71.7%) were constant, 556 (14.4%) were variable but uninformative, and 541 (13.9%) were parsimony informative. Parsimony analyses resulted in six trees of 1845 steps each (CI = 0.73; RI = 0.737). The MP tree of cpDNA data (Fig. 3) was congruent with the Bayesian tree (the best-fit model GTR + I + G) in topology. Clades E, F, and G, found in the ITS tree (Fig. 2), were also recovered in the cpDNA tree with high support (Fig. 3). The topology of the cpDNA MP tree (Fig. 3) differed from the ITS topology (Fig. 2) mainly in the position of Paulownia. Paulownia was sister to Phrymoideae (Clade F) with low support in the ITS tree (Fig. 2), but was sister to clade G with low support in the cpDNA tree (Fig. 3). Clade G, which includes Rehmannia, Triaenophora, and Orobanchaceae, was recovered as monophyletic with BS = 59% and PP = 100%. Rehmannia and Triaenophora, each as a monophyletic group, formed one maximum supported lineage, which was sister to Orobanchaceae. Orobanchaceae was strongly supported as monophyletic (BS = 78%; PP = 100%) in which Lindenbergia was sister to the remainder of Orobanchaceae. Clade G and Paulownia were further clustered with Phrymoideae (clade F) with moderate support. Templeton's test indicated incongruence between the trees constraining Phrymaceae as monophyletic and that no constraint was insignificant (P = 0.5316).

Figure 3

One of six most parsimonious trees generated from combined chloroplast data for selected taxa. Branch lengths are proportional to number of nucleotide substitutions (scales represent 10 substitutions). Bootstrap (BS) values (≥50%) are shown above the branches, and Bayesian posterior probabilities (PP) values (≥90%) are indicated below the branches. Maz = Mazus, Mim = Mimulus, Rehm = Rehmannia, Triaen = Triaenophora.

Analysis of combined chloroplast and nuclear ITS data

The ILD test gave a value of P = 0.21, indicating that the data sets were not significantly different from random partitions of the combined chloroplast and ITS data. The combined chloroplast and ITS data sets consisted of 4536 positions, of which 2998 (66.1%) were constant, 653 (14.4%) were variable but uninformative, and 885 (19.5%) were parsimony informative. Parsimony analyses resulted in one tree of 3735 steps (CI = 0.592, RI = 0.652). The MP tree (Fig. 4) was congruent with the Bayesian tree (the best-fit model GTR + I + G) in topology. The topology of the MP tree (Fig. 4) from combined cpDNA and nuclear ITS data were completely congruent with the ITS MP tree in the major clades (Fig. 2) and the cpDNA MP tree except for the position of Paulownia in the cpDNA tree (Fig. 3). The ingroup nodes in the topology of the combined chloroplast and ITS data received higher support than those in the separate analyses of either cpDNA or nuclear ITS data alone. Templeton's test indicated incongruence between the trees constraining Phrymaceae as monophyletic and that no constraint was insignificant (P = 0.4054).

Figure 4

Single most parsimonious tree generated from combined chloroplast and ITS data for selected taxa. Bootstrap (BS) values (≥50%) are above the branches; Bayesian posterior probabilities (PP) (≥90%) are below the branches. Maz = Mazus, Mim = Mimulus, Rehm = Rehmannia, Triaen = Triaenophora. ■ part of capsule exserted from the persistent calyx tube; □ capsule included in persistent calyx tube; capsule almost completely exserted from the persistent calyx tube; ● seeds with alveolate and pitted testa; ○ seeds with smooth or anomalous reticulate surfaces; seeds with membranous wings; ▲ calyx of connate sepals with spinescent apices; △ calyx with long triangular and lanceolate calyx lobes; ◆ corolla with two highly reduced and triangular upper lobes; ◇ corolla with two oblong and orbicular upper lobes.

DISCUSSION

Phylogenetic relation between Rehmannia and Triaenophora

Triaenophora has been considered closely related to Rehmannia in traditional systematics (25; 65; 36; 17). Our molecular data show that Rehmannia and Triaenophora form a strongly supported clade, in which Triaenophora is sister to Rehmannia. The sister relation between Rehmannia and Triaenophora is also corroborated by allozymic variability (39) and numerical analysis of morphological data (40). This sister relation is not surprising because Rehmannia and Triaenophora share a series of uniform synapomorphies, such as two lateral bracteoles at the base of the pedicel just above the subtending bract (they are aborted early in development in R. chingii, R. solanifolia, and R. glutinosa), four stamens with a gap at the expected site of the adaxial staminode, and the unidirectional initiation of corolla lobes and stamens from the abaxial to the adaxial side (77). Triaenophora has a series of unique traits distinctive from those of Rehmannia, i.e., five trifid calyx lobes; dense, white, spreading, lanose-villous hairs on the stems, leaves, and pedicels; and a bilocular ovary (17; 77). Rehmannia, as a strongly supported monophyletic group distinct from Triaenophora, is characterized by five revolute and undivided calyx lobes; brown or white glandular hairs on stems, leaves, and pedicels; and one ovarian locule (17; 77).

Our results regarding the relations among species within Rehmannia are in agreement with 3 and 39, 33). Albach (D. C. Albach, Johannes Gutenberg-Universität Mainz, Germany, unpublished data) conducted a study similar to ours with the exception that a single species of Triaenophora was included (T. rupestris) and in place of rps2, analyzed ndhF sequences. The results of these two independently conducted studies provide mutual confirmation that Triaenophora and Rehmannia are sister to each other and together are sister to Orobanchaceae. Likewise, both studies find evidence against the monophyly of Phyrmaceae (discussed later). Our results further show that the monotypic genus Titanotrichum initially described as a species of Rehmannia, is not closely related, but better included in Gesneriaceae (15; 78, 79, 80, 75; 63, 64; 54).

Familial placement of Rehmannia and Triaenophora

Rehmannia and Triaenophora were traditionally placed in the Digitaleae (Scrophulariaceae s.l.) with close affinity to Digitalis (11, 12; 25; 85; 36; 17). However, the inclusion of Rehmannia within Digitaleae was questioned when 53 placed one species of Rehmannia as sister to Mazus and Lancea (Scrophulariaceae s.l. or Mazoideae sensu 8). In all trees herein, Rehmannia and Triaenophora are shown to not have a close affinity with any Digitaleae. Our results also show no close relation between Rehmannia and Gesneriaceae including Titanotrichum, a relation that has been traditionally suggested (18; 29; 65; 36; 15).

Rehmannia and Triaenophora, as a monophyletic group, are shown herein to be sister to Orobanchaceae (including Lindenbergia) with moderate to high support BS = 92 for ITS, 59–62 for cpDNA, 94 for combined and high to maximum PP (99–100) from all analyses. Lindenbergia is sister to other parasitic genera of Orobanchaceae with high support in trees of cpDNA and combined ITS and cpDNA data as seen in previous molecular phylogenies (92; 50; 89; 10; 68) rather than sister to Rehmannia and Triaenophora (3). Our results further confirm that Orobanchaceae is a well-supported lineage that includes the holoparasitic members traditionally treated in Orobanchaceae, the hemiparasitic taxa previously treated under Scrophulariaceae s.l. and the nonparasitic genus Lindenbergia (19; 49; 87; 92; 50; 89; 10; 68). The sister relationship between Rehmannia and Mazoideae recognized in 53 is not supported by our molecular data with greater sampling both in Rehmannia and Mazoideae. Rehmannia and Triaenophora together with Orobanchaceae are sister to Paulownia and Phrymoideae.

On the basis of observations of morphological characters, together with previous reported data (17; 93), we conclude that Rehmannia, Triaenophora, and Orobanchaceae often have capsules that are half or partly exserted from the persistent calyx tubes, whereas in Phrymaceae, the capsules are completely included in the persistent calyx tube (Fig. 4). Seed coat characters are seldom used for systematic analyses at higher taxonomic levels. However, seed coat characters may provide synapomorphies for some of the relations recovered in our molecular based trees. Rehmannia and Triaenophora are characterized by numerous minute seeds with alveolate and reticulate testa similar to seeds of Orobanchaceae with alveolate, pitted and reticulate testa (except for seeds of Melampyrum with smooth testa; Fig. 4) (47; 94; 6; 55). Meanwhile, Phrymaceae possess minute seeds with smooth testa, and Paulownia is characterized by seeds with membranous wings (73; 90; Fig. 4). Along with Lindenbergia, Rehmannia, and Triaenophora have tricolporate pollen with reticulate sculpting of the exine (91; 30; 76), which differs from the characteristic tricolpate pollen (and its variants) with retipilate sculpting of the exine in other parasitic genera in Orobanchaceae (74; 46; 13; 1; 94; 42). The evolutionary trends of pollen exine ornamentation from reticulate to retipilate sculpting have been seen in Dichapetalaceae and in many different groups (57; 46), as well as from tricolporate to tricolpate pollen in some taxa (23; 31; 46; 43). The tricolpate pollen with retipilate sculpting of the exine in parasitic genera of Orobanchaceae may be derived from tricolporate pollen with reticulate sculpting in the nonparasitic genera Lindenbergia, Rehmannia, and Triaenophora.

In addition, phytochemical characters are often used to complement or improve molecular trees (27). Iridoids have been found as natural constituents in most taxa of Lamiales s.l. (32). Comparison of iridoid glycosides distributed among Rehmannia/Triaenophora, Orobanchaceae, Scrophulariaceae s.s and Plantaginaceae shows that catalpol and aucubin are widely present in these taxa (35; 52; 27; 3). However, Rehmannia/Triaenophora and most Orobanchaceae (including Lindenbergia) lack harpagide and 6-rhamnopyranosyl-catalpol and their esters, which are characteristic for many Scrophulariaceae s.s. (33). This distribution pattern supports the sister relationship between Rehmannia/Triaenophora and Orobanchaceae. Meanwhile, the lack of sorbitol as the reserve carbohydrate in Rehmannia and its presence among members of Digitaleae (35; 70; 3) is further evidence against a close affinity between Rehmannia and Digitaleae.

Lastly, Rehmannia and Triaenophora are characterized by two lateral bracteoles borne at the flower pedicel just above the leaf-like subtending bract (77). In some species of Rehmannia, they are aborted early in development and cannot be detected at anthesis (77). Similarly, two lateral bracteoles frequently occur in Orobanchaceae, where they are borne at the pedicel above the leaf-like subtending bract or just below the flower due to a much shortened pedicel (93). Rehmannia/Triaenophora and Orobanchaceae (s.s.) are characterized by simple racemes, while Lindenbergia has compound racemes with each branch composed of flowers and several pairs of bracts, all of which are subtended by a large leaf-like bract (90; 93). The reduction from inflorescence branch to a single or double flower frequently occurs in several major clades of angiosperms, such as Silene in Caryophyllaceae, Salvia in Lamiaceae, and Rhynchoglossum in Gesneriaceae (82; 83; 80). The two lateral bracteoles in Rehmannia/Triaenophora and Orobanchaceae (s.s.) might be the result of a transformation from compound racemes to simple racemes; however, further developmental analyses will be necessary to resolve this.

In comparison to other groups in Lamiales s.l. with respect to the distribution of related phytochemical and morphological characters, the combination of the aforementioned features are synapomorphies for Rehmannia/Triaenophora and Orobanchaceae. Based on the molecular results, corroborated by phytochemical and morphological data, we suggest that Rehmannia and Triaenophora represent the second nonparasitic branch sister to the remainder of Orobanchaceae s.l. (including Lindenbergia) or a clade at the rank of family sister to Orobanchaceae. Our results recognizing this sister relationship represent the first step toward better understanding the relations of Rehmannia and Triaenophora with other segregate families of Scrophulariaceae s.l. Further detailed studies are needed to better understand morphological and anatomical synapomorphies among these species.

The familial status of Paulownia and Mazus/Lancea (Mazoideae) remain uncertain in the results presented here. Paulownia has been placed alternately in the Scrophulariaceae s.l., Bignoniaceae, or assigned to a family of its own (48; 8). It is distinctively different from Orobanchaceae, Scrophulariaceae s.s. and Phrymaceae in its woody habit, capsule with a persistent woody calyx tube, and seeds with membranous wings (Fig. 4). The phylogenetic analyses herein, with increased sampling of Mazus, indicate that Mazus and Lancea (Mazoideae) may not be included in Phrymaceae as previously suggested by 53. However, Templeton's tests (see Results, Analyses with selected taxa) do not reject the inclusion of Mazoideae in Phrymaceae. The systematic position of Paulownia and Mazoideae deserves further detailed studies with greater taxon sampling among their putative relatives and new DNA regions together with genetic or evolutionary developmental methods to gain a comprehensive understanding about their phylogenetic history.