| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
2Conservatoire & Jardin botaniques, CH-1292 Chambésy, Genève, Switzerland; 3Biological Interactions and 4Molecular Systematics Sections, Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond TW9 3DS, UK
Received for publication May 17, 2002. Accepted for publication October 10, 2002.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: linalool • non-coding plastid DNA • Paliavana • plastid-expressed glutamine synthetase • pollination syndrome • Sinningia • species-level phylogenetics • Vanhouttea
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Fritsch, 1894
; Wiehler, 1983
; Chautems, 1990
). Such confusion and the extensive synonymy arose as a result of their extensive variation in habit, inflorescence form, and corolla shape, which in turn exemplify their morphological diversification in the Neotropics. The genera Sinningia Nees (which includes the commonly cultivated florist gloxinia), Vanhouttea Lemaire, and Paliavana Vandelli were first included in Gloxinieae by Wiehler (1983)
together with 17 other genera. However, the lack of rhizomes and the presence of a tuber in most Sinningia distinguish the former three genera from the rest of the tribe (Wiehler, 1983
; Boggan, 1991
). Then, based on molecular data, Smith et al. (1997)
and Zimmer et al. (2002)
proposed to remove Sinningia, Vanhouttea, and Paliavana from Gloxinieae and include them in a redefined Sinningieae. Major progress in generic circumscription has also been made based on hybridization experiments, which has led to the inclusion of Rechsteineria Regel and Lietzia Regel & Schmidt in Sinningia (Clayberg, 1968
; Moore, 1973
; Wiehler and Chautems, 1995
). Finally, 61 species of Sinningia, four of Paliavana and four of Vanhouttea are currently recognized (Wiehler, 1983
, 1984
; Chautems, 1990
, 1991
, 1995
; Chautems and Weber, 1999
; Chautems et al., 2000
), but six new taxa still await formal publication (A. Chautems, unpublished manuscript). These species are distributed from Central America to northern Argentina, with the highest diversity occurring in southeastern Brazil. Habit ranges from herbaceous species with a perennial tuber in most Sinningia species to woody shrubs without tubers in Paliavana and Vanhouttea. Stems are annual in most Sinningia species with new shoots growing every year from the tuber or perennial in Paliavana, Vanhouttea, and nine species of Sinningia (Chautems and Weber, 1999
). They occur mostly on rocky outcrops within the Atlantic rain forest between sea level and 2000 m a.s.l. (e.g., Meirelles et al., 1999
; Safford and Martinelli, 2000
), but some are epiphytic or terrestrial (Chautems and Weber, 1999
). Floral shapes, colors, nectar sugar composition, and fragrance indicate adaptations to various pollinators including hummingbirds, bees, bats, and hawk moths (Vogel [1969]
under Gesneria and Lietzia; Silberbauer-Gottsberger and Gottsberger, 1975
; Proctor et al. [1996]
under Rechsteineria; Perret et al., 2001
).
At lower taxonomic levels, molecular phylogenetic analyses are the most powerful tool to identify parallelisms/reversals, character correlations, and factors that may promote speciation (e.g., see Harvey and Purvis, 1991
; Larson and Losos, 1996
; Givnish, 1997
; Barraclough et al., 1999
; Barraclough and Nee, 2001
). In plants, the use of species-level phylogenetic studies has provided insight to our understanding of the evolutionary pathways leading to the diversification of traits associated with flower evolution and life history (e.g., Armbruster, 1993
; Silvertown and Dodd, 1997
; Givnish et al., 2000
; Möller and Cronk, 2001
).
Here, we have used a portion of the nuclear plastid-expressed glutamine synthetase gene (ncpGS, introns 7–10), as well as four plastid intergenic spacers (trnT-trnL, trnL-trnF, trnS-trnG, atpB-rbcL) and two introns (within trnL and rpl16) to produce a nearly complete species-level phylogenetic tree for the members of Sinningieae. We sampled all known species (Wiehler, 1983
, 1984
; Chautems, 1990
, 1991
, 1995
; Chautems and Weber, 1999
; Chautems et al., 2000
) with the exception of six taxa for which we could not obtain suitable material (i.e., S. helleri, S. schomburkiana, S. sulcata, and three unpublished new species). Phylogenetic utility of the ncpGS introns was described by Emshwiller and Doyle (1999)
, whereas the trnL-trnF region, the atpB-rbcL spacer, and the rpl16 intron have been widely used to assess infrafamilial relationships (e.g., see Taberlet et al., 1991
; Manen et al., 1994a
; Jordan et al., 1996
; Kelchner and Clark, 1997
; Samuel et al., 1997
; MacDade and Moody, 1999
). The trnT-trnL spacer has been used less often for phylogenetic analyses (but see Böhle et al., 1996
), whereas the trnS-trnG spacer was suggested to be more useful at the population level (Hamilton, 1999
). Based on separate and simultaneous analyses, we have compared congruence and phylogenetic usefulness of these DNA regions. We also propose a new taxonomic treatment for the tribe as well as scenarios for diversification in pollination syndromes and habits.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). Tissues were mostly sampled from cultivated specimens held in the Geneva Botanical Garden (Switzerland), the Gesneriad Research Foundation (USA), or at the Instituto Plantarum Ltda and associated private collections (Brazil).
Total genomic DNA was isolated from silica gel dried leaf tissue using the cetyltrimethyl ammonium bromide (CTAB) method of Doyle and Doyle (1987)
with the addition of 1% polyphenol polyvinylpyrolidon in the extraction buffer to bind phenolic compounds that might inhibit DNA amplification (Savolainen et al., 1995
). DNA was then purified using silica particles according to the manufacturer's protocol (Prep-A-Gene; Bio-Rad Laboratories, Hercules, California, USA). Double-stranded DNA was amplified via the polymerase chain reaction (PCR); each reaction contained 10 µL of GeneAmp 10x reaction buffer (100 mmol/L Tris-HCl, pH 8.3 at 25°C, 500 mmol/L KCl), 10 µL of 25 mmol/L MgCl2, 10 µL of 2 mmol/L dNTPs (Amersham Pharmacia Biotech, Piscataway, New Jersey, USA), 0.005% bovine serum albumin (BSA), 0.4 µL of 5U/µL AmpliTaq DNA polymerase (Applied Biosystems, Branchburg, New Jersey, USA), 1 µL of each 100 mmol/L primer (MWG-Biotech, Ebersberg, Germany) and 2 µL of genomic DNA (approximately 200 ng). The PCR included an initial denaturation step of 5 min at 94°C, followed by 35 cycles with 1 min denaturation at 94°C, 1 min annealing at 50°C, 2 min extension at 72°C, and a final extension of 5 min at 72°C. Following Taberlet et al. (1991)
, primer pairs a/b were used to amplify the trnT-trnL intergenic spacer and primer pairs c/f were used to amplify the trnL intron and intergenic spacer between trnL and trnF. The atpB-rbcL spacer was amplified using forward primer atpB-107 (5'-GGAAARGCTACATCTAGTAC-3') and reverse primer rbcL-326 (5'-CATRTTAGTAACAGAACCTTATT-3'). For the rpl16 intron, amplification was performed using primers F71 and R1661 (Jordan et al., 1996
); the trnS-trnG spacer was amplified using primers trnS (GCU) and trnG (UCC) from Hamilton (1999)
. A portion of the glutamine synthetase (ncpGS) was amplified with primers Gscp687f and Gscp994r (Emshwiller and Doyle, 1999
). The PCR products were run in 1% agarose gels stained with ethidium bromide and purified from the gel using Prep-A-Gene. DNA sequencing was performed using BigDye Sequence kits (Applied Biosystems, Foster City, California, USA) and then run on an ABI 377 DNA sequencer according to the manufacturer's protocols (Applied Biosystems). Both strands were sequenced using the amplification primers; for the atpB-rbcL spacer the internal primers atpB-49 (5'-TTTCAAGCGTGGAAACCCCA-3') and rbcL-188 (5'-TACAGTTGTCCATGTACCAG-3') were used as well, and for the rpl16 intron we used the additional primer R1516 designed by Kelchner and Clark (1997)
.
Direct sequencing of ncpGS produced heterogeneous electropherograms in 28 accessions. Therefore, purified PCR products were cloned by ligation into pGEM-T Vector Systems (Promega, Madison, Wisconsin, USA); XL2-Blue MRF ultra-competent cells were transformed according to the manufacturer's protocols (Stratagene, La Jolla, California, USA). Resulting colonies were screened for plasmids with inserts by PCR using pGEM/M13 primers, and five amplified inserts were then sequenced using primer Gscp687f as described above.
Sequences have been deposited in EMBL/GenBank (trnS-trnG accessions AJ438352–AJ438434; trnT-trnL accessions AJ439249–AJ439331; trnL-trnF accessions AJ439745–AJ439829; atpB-rbcL accessions AJ439900–AJ439984; rpl16 intron accessions AJ487702–AJ487786; ncpGS accessions AJ459606–AJ459691).
Phylogenetic analyses
All sequences were aligned by eye; the aligned matrix is available from M. Perret and V. Savolainen. Three regions in the trnS-trnG (positions 69–116, 416–560, and 702–880 in the alignment of trnS-trnG), two regions in the trnT-trnL (positions 370–395 and 608–630) and two regions in the rpl16 intron (positions 151–160 and 811–825) were unsatisfactorily aligned due to poly(AT) and poly(T) regions; therefore they were excluded from the analysis. Insertions/deletions (indels) were coded as binary characters using question marks for nested indels (Barriel, 1994
).
Phylogenetic analyses were performed using PAUP* 4.0b8 (Swofford, 1999
). Most-parsimonious trees were found with heuristic searches including 1000 replicates of random taxon additions and tree bisection-reconnection (TBR) branch swapping with only 10 trees held at each step; the trees found in these replicates were then used as starting trees in another search without a tree limit until 10 000 most parsimonious trees were found. Internal support was evaluated with the bootstrap (BS; Felsenstein, 1985
): 1000 replicates were performed using TBR swapping, SIMPLE addition of taxa, and a tree limit of 500. The searches described above were performed separately for each DNA region, all plastid regions combined, and plastid DNA regions plus ncpGS combined.
Tests of incongruence
To compare the pattern of phylogenetic signal present in the different DNA regions, we used the "incongruence length difference test" (ILD test, Farris et al., 1995
) implemented in PAUP* 4.0b8 as the "partition homogeneity test" (Swofford, 1999
). The collection of most parsimonious trees obtained from analyses of each partition separately were also compared and conflicting nodes supported by a minimum of 70% bootstrap (BS) support were identified (i.e., hard incongruences; Johnson and Soltis, 1998
). When hard incongruences were found between plastid-based and ncpGS-based topologies, we evaluated their significance using the "Shimodaira-Hasegawa test" (SH test; Shimodaira and Hasegawa, 1999
; Goldman et al., 2000
) implemented in PAUP* 4.0b8 (Swofford, 1999
). These conflicting nodes were used as constraints with the rival data and a new heuristic search was performed as described above. Then, two subsets of 100 trees obtained with and without constraint were compared simultaneously with the RELL option of the SH test and 1000 bootstrap replicates as implemented in PAUP* 4.0b8 (Swofford, 1999
). Because maximum likelihood calculations are necessary to achieve an SH test, we used the HKY85 model of evolution (Hasegawa et al., 1985
) with a gamma distribution to accomodate for rate heterogeneity among sites (Yang, 1996
); all parameters (i.e., base frequencies, transition/transversion ratios, alpha shape, proportion of variable sites) were estimated for plastid and ncpGS data sets from DNA sequences and a 50% majority-rule bootstrap consensus tree.
Character mapping
Floral and vegetative types were mapped onto the trees resulting from parsimony analyses of combined plastid DNA regions, ncpGS alone, and all data sets combined. Recontruction of ancestral states were performed after pruning 18 accessions to include only one representative per species and Smithiantha as outgroup (77 accessions in total). Using MacClade 3.1 (Maddison and Maddison, 1992
), unambiguous changes were reconstructed with maximum parsimony and both accelerated (ACCTRAN) and delayed (DELTRAN) character state optimizations by considering all alternative topologies to accomodate for polytomies.
Four broad classes of pollination syndromes have been recognized in Sinningieae based on corolla shape, size, color, fragrance, nectar volume and sugar compositions. These are hummingbird (51 species), bee (20 species), bat (four species), and moth (one species, S. tubiflora).
The tuber of Sinningia has a hypocotyledonary origin and grows by incorporating the basal shoot internodia over the years (Weber, 1936
under Corytholoma). We recognized here three different habits: (1) tuberous plants with annual growth and with basal internodes giving rise to new axillary shoots every year (53 Sinningia species); (2) tuberous evergreen plants in which the tuber is small and growth restarts from the upper nodes of the perennial stem (six Sinningia species); (3) non-tuberous plants with perennial stem, which is suffrutescent to woody in Paliavana spp., Vanhouttea spp., S. lindleyi, S. schiffneri, and S. sp. nov. 4 (Wiehler, 1983
; Chautems and Weber, 1999
).
Odor collection and analysis
The moth-pollinated flowers of S. tubiflora produce a sweet fragrance to attract its pollinators. To identify the volatile compounds, a flower of S. tubiflora growing in a glasshouse at the Royal Botanic Gardens, Kew (accession number 1985-5527) was partially enclosed in a transparent forensic sample bag, and air from near the flower was drawn through an odor trap (100 mm x 3 mm diameter; SGE, Victoria, Australia) packed with 100 mg of Tenax TA (35–60 mesh, Jones Chromatography, Hengoed, Mid-Glamorgan, UK) at 25 mL/min for 3 h (21:00–23:00) by means of a portable pump. The trap was analyzed by thermal desorption-gas chromatography-mass spectrometry (TD-GC-MS: TD injector, SGE; GC, Perkin Elmer 8500, Shelton, Connecticut, USA; MS, Finnigan Ion Trap Detector 800 Series, San Jose, California, USA). The odor trap was desorbed at 250°C for 3 min in a flow of helium and chromatography was performed on a 25 m x 0.22 mm (internal diameter) x 0.25 µm BPX5 capillary GC column (SGE) using an oven temperature program of 40–260°C at 4°C/min with a helium carrier gas pressure of 68947.6 Pa. During desorption of the trap, the front end of the GC column was looped through a polystyrene cup filled with liquid nitrogen so as to focus the volatile compounds. Electron ionization mass spectra of the column eluate were recorded every second using a scan range of m/z 38–400 (mass/charge).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
and Emshwiller and Doyle (1999)
. Total length of the introns varied from 399 to 427 bp (intron 7, 95 bp; intron 8; 131–153 bp; intron 9; 78–81 bp; intron 10, 90–95 bp). In the ingroup, mean pairwise divergence for the ncpGS introns was 4.4% (maximum of 10%) with 23.9% of potentially informative characters, which is about 2–5 times higher than in the non-coding plastid DNAs (Table 1). When outgroup taxa were included, mean pairwise divergence for the introns increased to 5.5% (maximum of 13%) with about a third of the characters being potentially informative. Out of the 28 accessions with heterogeneous electrophenograms, 10 of these had a unique sequence for each of the five clones sequenced. Heterogeneity was therefore probably due to different sequences present in a minority of the PCR products. Within the 18 remaining plants, more than one sequence of ncpGS was found and eight of them display even more than two sequences (up to five). The sequences found within individuals differ by 1–2 bp except for S. richii (3 bp), S. villosa (5 bp), and V. gardneri (11 bp). Phylogenetic analysis including all clones showed all species to be monophyletic except for V. gardneri, where the most divergent clone clustered with V. leonii/V. pendula (data not shown). For subsequent analyses only one clone per individual was used, except for V. gardneri for which we discarded the most divergent clone.
Plastid-based phylogenetic analyses
Each data set (i.e., atpB-rbcL, trnT-trnL, trnL-trnF and trnS-trnG spacers, trnL and rpl16 introns) was first analyzed separately, resulting in relatively poorly resolved/supported trees (Fig. 1). The highest number of nodes supported by at least 70% BS was retrieved with rpl16 intron, whereas the lowest was provided with trnL-trnF spacer (Fig. 1). The ILD test between all pairs of individual plastid data sets revealed significant incongruences (P < 0.01) except between trnL and rpl16 introns (P = 0.86). However, as we have not found any hard incongruence (BS > 70%) between these topologies, we performed a combined analysis of all plastid regions (hereafter "plastid data set"; Fig. 2). In comparison with partitioned analyses, simultaneous analysis of plastid sequences increased from 1.8 to 14 times the number of branches supported by at least 70% BS (Fig. 1). Sinningieae were monophyletic (BS 100%). Within Sinningieae, three major clades are observed and denoted clade A (BS 91%), B (BS 67%), and C (BS < 50%). Clade C includes several Paliavana species and the majority of the species of Vanhouttea, which are embedded in Sinningia. Not assigned to a clade are P. tenuiflora, S. richii, and V. lanata, sisters to clades B, C, and B/C respectively. The remaining species fall into two small clades denoted clade D (P. gracilis, V. hilariana, V. leonii, and V. pendula; BS 98%) and clade E (P. plumerioides, S. schiffneri, and S. sp. nov. 4; BS 91%). When two individuals per species were used, they came out as sisters with the exception of the individuals of S. speciosa and S. villosa, which were embedded in a same clade together with S. guttata (Fig. 2). Infraspecific divergence ranged between 0.03 and 0.43%, which is below interspecific divergence within Sinningieae (0.79–2.06%; Table 1).
|
|
70% than any single plastid regions (Fig. 1). The ncpGS tree depicted in Fig. 3 also supports the monophyly of Sinningieae (BS 100%). Three major clades emerge (A, B, and F), and P. plumerioides is sister to the rest of the species (BS < 50%). Species included in clades C, D, and E based on plastid DNAs are lumped together with P. tenuiflora and V. lanata in a clade denoted F (BS < 50%). Clade A (BS 91%) includes the same species as in the plastid-based analysis, but clade B (BS 77%) differs by the inclusion of S. richii, which is sister to clade C based on plastid DNAs (BS < 50%; Fig. 2). In addition, relationships among clades are different between the nuclear and plastid-based analyses, with clade B sister to A based on ncpGS (BS 83%; Fig. 3), whereas it is sister to C based on plastid DNAs (BS 78%; Fig. 2). When we used two individuals per species, they came in sister positions (Fig. 3). Intraspecific pairwise divergences in introns was 0–1.69%, which is 2.5 times lower than mean interspecific divergence within Sinningieae (4.4%; Table 1).
|
symbol on Figs. 2 and 3). In the plastid-based trees, S. warmingii is sister to S. allagophylla; S. tubiflora is sister to S. aggregata; S. concinna is sister to S. harleyi; P. prasinata clusters with S. guttata, S. kautskyi, S. macrophylla, S. speciosa, and S. villosa; and clades B and C together with V. lanata are monophyletic (Fig. 2). However, in the ncpGS-based tree, S. warmingii is sister to S. tubiflora; S. concinna is included in a clade with S. aghensis/S. pusilla; P. prasinata is sister to P. tenuiflora; and clade B is sister to clade A (Fig. 3). In most cases, constraining these conflicting nodes onto the rival data set retrieved trees with maximum likelihood scores that were significantly worse than those derived from an unconstrained search, according to the SH test (significant conflicted nodes at P < 0.05 were denoted by an asterisk in Figs. 2 and 3). The null hypothesis was, however, not rejected in four cases when ncpGS data was constrained by the sister relationship of S. warmingii/S. allagophylla and B/C clades and when plastid data was constrained by the sister relationship of P. prasinata/P. tenuiflora and A/B clades.
Total evidence
We also combined these two data sets in a "total evidence" approach (Kluge, 1989
; Fig. 4). Combining data sets did not decrease global bootstrap support as the number of nodes with at least 70% BS was slightly higher than in the plastid-based trees (47 vs. 43; Fig. 1). Resulting topologies are similar to those obtained from the plastid data set and include five major clades (A to E). Due to the higher number of informative characters in the plastid data set, conflicting relationships with ncpGS are mainly resolved towards the plastid-based topologies with the exception of a few relationships that follow the ncpGS-based topologies (i.e., position of S. richii within clade B [BS 61%], sister positions of P. tenuiflora/P. prasinata [BS 57%] and P. sericiflora/P. werdermanii [BS 98%], position of V. fruticulosa clustered with S. gigantifolia and S. cochlearis [BS 87%], and V. lanata clustered with other Vanhouttea [BS 68%]). Within the subclade that includes S. aggregata, S. allagophylla, S. amambayensis, S. curtiflora, S. sellovii, S. tubiflora, and S. warmingii (BS 100%), only two nodes were present in the strict consensus of the total evidence analysis (Fig. 4), whereas four are found in the plastid-based strict consensus (Fig. 2); this decrease of resolution is certainly the consequence of conflicts between plastid and ncpGS data sets in the relationships of these species.
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Randall et al., 1998
). Among the plastid regions we sequenced, the trnT-trnL and trnS-trnG spacers and the rpl16 intron provided more variable sites than the commonly used atpB-rbcL, trnL-trnF spacers, and the trnL intron; this result corroborates results of Small et al. (1998)
for trnT-trnL spacer and the rpl16 intron. As observed by Gielly and Taberlet (1994)
, the evolutionary rate of the trnL intron is only slightly lower than the adjacent trnL-trnF spacer despite conserved stem regions and secondary structures of this group I intron (Kuhsel et al., 1990
). Similarly, the number of variable positions in the rpl16 intron range within values obtained for intergenic spacers, despite the fact that this group II intron was suspected to be strongly constrained by structural features (Clegg et al., 1994
). Among intergenic spacers, atpB-rbcL spacer exhibits the lowest pairwise divergence. Unlike other spacer between trn genes, atpB-rbcL spacer contains conserved elements involved in gene regulation like atpB and rbcL promoters and leader sequences of both genes; these could account for the low evolution rates comparable to the rbcL exon, as observed in Rubieae (Manen et al., 1994b
; Manen and Natali, 1995
).
Plastid non-coding regions were rich in strings of mononucleotides repeats and small tandem repeat units, which are known to be subject to slipped-strand mispairing, a major mechanism in length mutations (Levinson and Gutman, 1987
). Sequence repeats and extensive length variation were particularly high in the spacers trnT-trnL and trnS-trnG, requiring the introduction of a large number of gaps in their alignments, which potentially limited their use to closely related species between which indel types can be recognized and used for alignment (Kelchner, 2000
). Although indels were recorded as being as frequent as nucleotide substitution between closely related species (Zurawski and Clegg, 1987
; Gielly and Taberlet, 1994
), our results indicate that potentially informative indel characters represent only one-sixth (in rpl16 intron) to one-third (in trnL-trnF spacer) of the total number of informative characters (Table 1).
The four introns of ncpGS sequenced in Sinningieae amounted to a length slightly longer than in Oxalis (360–394 bp). They occupied the same locations as in Oxalis and Medicago, indicating that the intron positions among rosids and asterids are conserved (Tischer et al., 1986
; Emshwiller and Doyle, 1999
). Restricted length variation in ncpGS and a maximum pairwise divergence of 13% between Sinningieae and outgroups allow an unambiguous alignment, and we assume that the phylogenetic utility of this gene might be not only at the generic or tribal level but also at higher taxonomic levels. The mean pairwise divergence found in the ncpGS intron among Sinningieae species (averaging 4.4%) is slightly higher than among closely related Oxalis species according to the results of Emshwiller and Doyle (1999)
and comparable at the same taxonomic level with intron divergence detected in other low copy nuclear genes like Adh, pistillata, or cycloidea (Randall et al., 1998
; Bailey and Doyle, 1999
; Citerne et al., 2000
).
The presence of more than one ncpGS sequence within individual plants revealed by cloning indicates apparent heterozygosity for this locus. The level of ncpGS variation does not however exceed intra-individual polymorphism detected within paralogous Adh genes (Sang et al., 1997b
) or between multiple copies of the internal transcribed spacers of nuclear ribosomal DNA (nrITS; Denduangboripant and Cronk, 2000
; Kita and Ito, 2000
). Presence of multiple ncpGS sequences within individuals was also found in Oxalis, in which variants were interpreted as either normal allelic polymorphism, homeologous loci, or artifacts due to recombination during PCR amplification (Emshwiller and Doyle, 1999
). Contrary to Oxalis, a maximum of two variants is expected in Sinningia, in which cytological data report only diploids with 2n = 26 (Clayberg, 1967
). Further investigation are thus required to determine the origin of more than two sequences in eight individuals by examining alternatives such as the presence of undetected polyploids, weakly divergent paralogous copies, or putative PCR artifacts.
Incongruence between data sets
Separate analyses of each of the six plastid DNA regions were shown to be significantly incongruent according to the ILD test (except between trnT-trnL and rpl16). However, the topologies obtained were congruent at a confidence limit of 70% BS, exemplifying the difference between characters and topological incongruences (Allard and Carpenter, 1996
). Because no hard incongruence was found, the inclusion of all linked plastid DNA regions into a single analysis should maximize the explanatory power regardless of the level of character incongruences between data sets detected with the ILD (Nixon and Carpenter, 1996
; Yoder et al., 2001
). Sequence variation and resolution of ncpGS are higher than for any plastid non-coding regions, and several topological incongruences were found between the plastid and ncpGS-based phylogenetic trees, as inferred from bootstrap analyses and SH test (Figs. 2 and 3). Some of the taxa included in the conflicting clades probably have separate evolutionary histories for plastid and ncpGS gene, and thus introgression, lineage sorting, or gene duplication/deletion processes should be considered (Doyle, 1992
; Maddison, 1997
; Wendel and Doyle, 1998
). The probability that independent duplication of a single locus has occurred among closely related species is low (Maddison, 1997
; Sang and Zhang, 1999
). However, the presence of multiple sequences detected within individuals of Sinningieae (e.g., V. gardneri) does not exclude completely that paralogous copies are responsible for incongruences between ncpGS and plastid-based analyses. On the other hand, at a low taxonomic level, many discrepancies between plastid and nuclear phylogenetic analyses were attributed to presumed introgression or lineage sorting events in a wide range of plant groups like Ceanothus, Gossypium, Helianthus, Heuchera, Paeonia, and Triticeae (Rieseberg, 1991
; Wendel et al., 1991
; Soltis and Kuzoff, 1995
; Mason-Gamer and Kellogg, 1996
; Sang et al., 1997a
; Seelanan et al., 1997
; Hardig et al., 2000
). In Sinningieae, our results show significant topological incongruences as identified by the SH test at nodes including terminal branches, and few individual species can be identified as the cause of the conflict. At this within-clade level, experimental hybridization studies and cytological analyses demonstrated a weak post-zygotic reproductive barrier between Sinningia species (Clayberg, 1968
, 1970
, 1996
; Wiehler and Chautems, 1995
), and it is therefore possible that introgression explains the incongruences observed between the plastid and nuclear-based phylogenetic trees.
Plastid introgression was found to occur more frequently than nuclear gene flow as reviewed by Rieseberg and Soltis (1991)
. This process was mainly inferred from discordant distributions between organellar and nuclear characters among interfertile plant species (Whittemore and Schaal, 1991
; Rieseberg et al., 1991
; Soltis and Kuzoff, 1995
; Wolfe and Elisens, 1995
). In Sinningieae, hypothesis of plastid capture could be evoked for morphologically related species (i.e., S. reitzii/S. mauroana, S. tubiflora/S. warmingii, P. prasinata/P. tenuiflora, S. pusilla/S. concinna) that have a sister relationship with the ncpGS tree but significantly divergent positions in the plastid tree (Fig. 2 vs. Fig. 3). In several other cases, the plastid phylogenetic analysis agrees with previous ideas about taxonomic relationships better than does the ncpGS tree. For example, the fusion of two dorsal extended corolla lobes might be seen as a "good" synapomorphy for the clade including S. bulbosa, S. cardinalis, S. magnifica, S. lateritia, S. glaziovana, S. hatschbachii, S. cooperi, S. iarae, and S. micans based on plastid DNAs (BS 64%; Fig. 2). Based on ncpGS, these latter species are oddly embedded in three different clades with species displaying tubular corolla with lobes of equal size (i.e., S. glaziovana/S. cardinalis/S. lateritia with S. douglasii, BS 86%, and S. cooperi with S. piresiana/S. rupicola, BS 99%, Fig. 3). Such incongruences could indicate nuclear introgression events, although the possibility exists that we did not amplify homologues ncpGS in all cases.
Systematics of Sinningieae
The combined analysis of plastid DNAs and ncpGS (Fig. 4) provides insights in the systematics of Sinningieae, but only relationships that are not in significant conflict between plastid and ncpGS single analyses are discussed in this section. All analyses confirm that Sinningieae constitute a monophyletic group in agreement with Smith et al. (1997)
and Zimmer et al. (2002)
, distinct from Gloxinieae within which they were included by Wiehler (1983)
. Paliavana, Vanhouttea, and Sinningia are primarily Brazilian genera currently distinguished by their habit and floral morphology: Paliavana and Vanhouttea are shrubby species with respectively campanulate to bell-shaped flowers or red tubular flowers, whereas Sinningia is characterized by the presence of a tuber irrespective to corolla shape. This generic circumscription is however not congruent with our phylogenetic analyses: both Paliavana and Vanhouttea are found to be polyphyletic and nested within Sinningia. A new circumscription of these genera is thus required. Bearing in mind the need for nomenclatural stability, the inclusion of all taxa in a single redefined genus appears to be the best solution (McNeill, 2000
).
Monophyly of clade A is strongly supported in both separate and combined analyses (BS 91%, 98%, 100% in Figs. 2, 3, and 4, respectively). Most species in this clade are characterized by a verticillate habit corresponding to a pseudo-whorl of leaves at the extremity of an elongated basal internode (Chautems and Weber, 1999
). Exceptions are S. magnifica and S. cooperi, which have internodes of equal size and S. eumorpha and S. conspicua, which have a rosulate habit. Morpholgical synapomorphies for this clade are a calyx base shortly adherent to the ovary, calyx lobes generally short and narrow, two separate enlarged dorsal nectary glands, and five nectary chambers equal in size. Corollas are red and tubular in all species except in the sister species S. conspicua and S. eumorpha, which have respectively yellow or white to light lavender campanulate flowers. Among red tubular species, fusion and expansion of the two dorsal corolla lobes ("hooded" corolla) is a synapomorphy supporting the monophyly of S. bulbosa, S. cardinalis, S. cooperii, S. glazioviana, S. hatschbachii, S. iarae, S. magnifica, S. micans, and S. lateritia (BS 81%; Fig. 4), whereas the remaining species have five lobes of equal size. Based mostly on corolla shape, Fritsch (1894)
classified species within clade A in two different genera: campanulate species were placed in Sinningia, whereas red and tubular species were included either in Corytholoma section Dircaea (hooded corolla) or in section Thamnocaula (corolla lobes equal in size). However, based on tests of crossability, Clayberg (1968
, 1996)
first suggested a close affinity of these species because he was able to produce fully fertile hybrids from all crosses between S. eumorpha, S. cardinalis, S. magnifica, S. cooperi, S. macrorhiza (= bulbosa), S. douglasii, S. macropoda, S. lineata, S. leucotricha, and S. reitzi. Congruence of our results with Clayberg supports that clade A constitutes a taxonomical entity for which the name "Dircaea" is appropriate for nomenclatural reasons.
Clade B (BS 61%) includes species with various floral morphologies and vegetative habit. Basal species are particularly diverse and have habits ranging from verticillate (S. aghensis), rosulate-caulescent (S. barbata) to minute rosulate (S. pusilla), and flowers with tiny funnel-shaped (S. pusilla, S. concinna), bell-shaped (e.g., S. aghensis), or tubular (e.g., S. harleyi) corolla. The 16 species included in a well-supported subclade (BS 97%, from S. amambayensis to S. defoliata in Fig. 4) have a more uniform habit, i.e., annual herbaceous stems with equal size internodes bearing caulescent to spicate florescences (Chautems and Weber, 1999
). A noticeable exception to this above-described morphology is S. defoliata, which displays a stemless, unifoliate habit. Within this subclade corolla size and color are highly labile, ranging from red or orange tubular corolla (
1 cm long in S. allagophyla and S. curtiflora) through long white tubular corolla (7 cm in S. tubiflora) to a greenish gullet-shaped corolla in S. brasiliensis. Despite this diversity, crosses between several species in clade B produced fertile hybrids (Wiehler and Chautems, 1995
; Clayberg, 1996
). Furthemore, we identified some morphological synapomorphies for this clade such as calyx lobes that are foliaceous and non-adherent to the ovary, two enlarged and fused dorsal nectary glands (excepted in S. brasiliensis having a ring composed of five fused glands), and two dorsal nectary chambers more developed than the three remaining ones. Early on, Decaisne (1848)
used this same set of characters to circumscribe the genus Corytholoma Decaisne. Fritsch (1894)
then expanded this definition of Corytholoma and divided it in five sections of which three are part of clade B; "Corytholoma" is therefore an appropriate name for this group.
Clade C (BS 56%) includes Paliavana and Vanhoutta species embedded within Sinningia species. Habit is suffrutesent to shrubby in both Paliavana and Vanhouttea (up to 3 m tall), whereas the Sinningia species included in this clade are herbaceous caulescent, rosette-like, or unifoliate (Chautems and Weber, 1999
). Sinningia lindleyi, S. villosa, and S. kautskyi have a reduced tuber producing a perennial stem. Flowers are also highly diversified, ranging from purple and greenish bell-shaped to red and tubular. However, all species in clade C have five nectary glands of equal size surrounding the ovary base, except in S. gigantifolia and S. cochlearis, which display two enlarged and fused nectary glands in dorsal position. Several subclades within clade C are well supported, although their relationships are poorly resolved. The bee-pollinated species S. guttata, S. macrophylla, S. speciosa, and S. villosa are clearly monophyletic (BS 99%). All species of Vanhouttea except one are clustered in a subclade (BS 68%) that includes the type species of the genus, V. calcarata. Not included in this latter subclade is V. fruticulosa, which is clustered with S. gigantifolia and S. cochlearis (BS 87%). Paliavana appears paraphyletic with P. sericiflora/P. werdermannii sister to clade C and P. tenuiflora/P. prasinata embedded within Sinningia, but the position of P. prasinata is discordant between plastid and ncpGS analyses. The positions of the remaining species of Sinningia within clade C are poorly resolved. Some of them (e.g., S. lindleyi and S. guttata) were among the first species recognized in this genus and are morphologically close to the type species S. helleri; we thus propose to include all species of clade C under the name "Sinningia."
Monophyly of clades D and E are well supported in the plastid-based analyses and in the total evidence analyses (BS > 74%); however in the ncpGS analyses, species of clade D and E are lumped together within with clade C under clade F (Fig. 3), but with a BS of <50%. Taxa of both clades D and E share characters with the majority of the members of clade C (clade Sinningia), such as absence of a tuber, suffrutescent/shrubby perennial stem, and five equally sized nectary glands. Clade D includes four shrubby species of which one matches the definition of Paliavana because of its blue campanulate flowers and three were initially considered as Vanhouttea because of their red tubular corolla (Chautems, 2002
). The segregated position of the latter three taxa is surprising as they do not differ from the other Vanhouttea species of clade C, with the exception of calyx lobes that are free in the bud instead of being fused. Based on this character we named this lineage "free calyx lobes Vanhouttea." Clade E includes S. schiffneri, P. plumerioides, and one unpublished new species; their corollas are white and trumpet-shaped in S. schiffneri and S. sp. nov. 4, whereas they are yellowish and gullet-like in P. plumerioides. Fritsch (1908)
early recognized S. schiffneri as being the only member of the S. section Thamnoligeria, and we adopt this name for this clade.
Based on the molecular data, five lineages are now recognized, whereas the lability of flower morphology and habit had made their identification previously difficult. However, nectary glands and adherence of calyx to the ovary characters, already used by Decaisne (1848)
in an attempt to classify Brazilian Sinningieae, turn out to be most useful features for diagnosing the principal lineages within Sinningieae.
Pollination syndromes
Optimization of pollination syndromes in phylogenetic analyses indicates that multiple and rapid changes have occurred. Similar results were found in the Old World Gesneriaceae, Streptocarpus and Saintpaulia (Harrison et al., 1999
), and Plantaginaceae (Reeves and Olmstead, 1998
). Genetic studies in Mimulus and Antirrhinum have furthermore demonstrated that floral color, shapes, symmetry, and nectar production were controlled only by a few loci or were epigenetic (Bradshaw et al., 1995
; Luo et al., 1996
; Cubas et al., 1999
).
Multiple origins of hummingbird and bee syndromes in Sinningieae and their co-occurrence in every major clades apart from E indicate numerous changes in pollination syndromes; however, because in all reconstructions both hummingbird and bee syndromes can be optimized at the root node, the ancestral state was not determined unambiguously. In more recent cladogenesis, unambiguous changes from hummingbird to bee syndromes were reconstructed in contrast to the general view that hummingbird pollination is derived from bee pollination (e.g., Grant, 1994
).
Shifts to the hummingbird from the bee syndrome involved the acquisition of floral features, such as the red color and the straight tubular corolla, that are considered to play a role in deterring insect visitors that may rob nectar (Heinrich and Raven, 1972
; Raven, 1972
; Schemske and Bradshaw, 1999
; Chittka et al., 2001
). Field observations confirm hummingbird specificity for a few red tubular Sinningia and Vanhouttea species (Snow and Teixeira, 1982
; Sazima et al., 1996
; Chautems et al., 2000
; Buzato et al., 2000
; Vasconcelos and Lombardi, 2000
). On the contrary, shifts to bee syndrome involve numerous colors (blue, purple, yellow, and white) independently of campanulate or funnel-shaped flowers.
Shifts to bat syndrome in S. brasiliensis, P. prasinata, P. sericiflora, and P. werdermannii occurred independently in clade B and C. The greenish, widely opened, gullet-like flower of S. brasiliensis is unambiguously derived from a red tubular flower, which characterizes its closely related species (S. araneosa, S. carangolensis, and S. valsuganensis; Fig. 4). Within clade C, the bat syndrome is restricted to the three Paliavana species displaying green to cream robust bell-shaped corollas. These latter species are embedded with bee- and hummingbird-pollinated species in clade C, and their ancestral state remains ambiguous between these two syndromes. Associated with these morphological features, bat flowers also produce fragrance reminiscent of burnt plastic. The chemical composition of this fragrance has not been identified in Sinningieae, but similar scents were recorded in several bat-pollinated flowers from different angiosperm families; these all consisted of sulfur-containing compounds (Knudsen and Tollsten, 1995
). The production of nectar measured in S. brasiliensis and P. prasinata was 4–6 times higher than in other Sinningieae species pollinated by moths, bees, or hummingbirds (Perret et al., 2001
). Nectar sugar concentration and composition were also different from hummingbird and bee flowers: total sugar concentration in bat flowers averaged 7.1 ± 3.4 % (mass/total mass) and is much lower than in hummingbird (23.9 ± 10.6%) and bee flowers (28.7 ± 10.6%), whereas sucrose proportion in bat flowers is on average 12 times lower than in the other syndromes (Perret et al., 2001
). This sugar concentration in bat-pollinated species is below optimal nectar concentration established from models and feeding experiments with hummingbirds (Heyneman, 1983
; Kingsolver and Daniel, 1983
; Tamm and Gass, 1986
) and below nectar concentration among bee-pollinated flowers in temperate and tropical regions (Pyke and Waser, 1981
). Diluted nectar in bat flowers could consequently deter hummingbirds and bees more efficiently than the morphology of the corolla, paralleling the hypothesis of Bolten and Feinsinger (1978)
.
Moth pollination in S. tubiflora is likely due to features such as the long and slender flower and the white color, but as far as we know, no observations have been reported in the wild (Silberbauer-Gottsberger and Gottsberger, 1975
; Endress, 1994
). Sinningia tubiflora occurs at the southwestern limit of Sinningieae distribution in Paraguay and in northern Argentina. Because all species related to S. tubiflora display hummingbird flowers (Fig. 4), this moth flower is derived unambiguously from a red and tubular morphology. Close relationships between these two pollination syndromes were also recorded in the genera Ipomopsis and Aquilegia in North America, where sphingophilous species occurred at higher elevations and in generally dryer habitats than their ornithophilous relatives (Grant, 1992
). The shift to moth pollination is also associated with a sweet fragrance, which was identified as linalool in this study (Fig. 6); this compound is found in moth-pollinated flowers worldwide (Knudsen and Tollsten, 1993
; Raguso and Pichersky, 1999
). The evolution of a species having a linalool-dominated odor from a non-scented species has also occurred in Clarkia, where it was found that de novo production of linalool in C. breweri was due to the up-regulation of linalool synthase (Dudareva et al. 1996
; Raguso and Pichersky, 1999
). The gene encoding linalool synthase has also arisen frequently during plant evolution, apparently by the duplication of terpene synthase genes followed by divergence and/or domain swapping (Cske et al., 1998
).
Vegetative habit and geographical distribution
The ancestral habit in Sinningieae is likely to comprise suffrutescent stems and a lack of a tuber, as reconstructed in the two clades (D and E) that are successively sister to the rest (Fig. 4). The acquisition of tubers in Sinningia has occurred independently several times and is a feature that was considered a key character defining the genus, as the rest of Gloxinieae have scaly rhizomes (Wiehler, 1983
). This storage organ, characterizing all members of clades A and B, is coincident with the ability of many species to colonize rupicolous habitats in open or forested areas. Cyclic annual growth and tuber dormancy are well adapted with the seasonal climate of southeastern Brazil (i.e., rainfall and warm temperatures in November to March, and dry and cool conditions in June to August). These conditions can be extreme when species grow in rocky habitats. Multiple reversions to perennial growth in association with the reduction of tuber size could then reflect a habitat shift to less seasonally extreme environments, relaxing constraints on the dormancy period.
The geographical distribution of clade A is restricted to the southeast and south regions of Brazil (Fig. 8). These taxa have marked affinities for well-drained substrates, where water is available from rain, fog, or dripping water. Such conditions occur mainly in rupiculous habitats, but two species are also found in epiphytic situations (S. douglasii, S. cooperi). Some taxa occupy impressive cliffs along the coastal range (Serra do Mar) including the well-known granite rock outcrops in Rio de Janeiro (Meirelles et al., 1999
). The popular name "queen of the abyss" was therefore given to S. canescens and S. leucotricha. The largest tubers (over 30 cm for S. lineata and S. macrostachya) were recorded within this clade for the species occurring in southern Brazil, where seasonality is more pronounced than in southeastern Brazil. Conversely, reversion to a perennial suffrutescent stem with a reduced tuber was recorded in S. reitzii and S. mauroana living in rich pockets of humus in high rainfall sections of the Atlantic forest.
|
Species of clades C, D, and E are restricted to the coastal region in Brazil north of the Tropic of Capricorn (Fig. 8). They grow on rocky substrates in various open, mostly montane habitats (Vanhouttea spp. and Paliavana spp.) or in the understory of rain forest at low altitudes (S. schiffneri, S. lindleyi). The co-occurrence of the three habits within clade C (tuberous with annual stem, tuberous with perennial stem, non-tuberous and perennial; Fig. 4) implies multiple shifts in life history <
); bee flowers (
); bat flowers (
); moth flower (
)
