| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics and Phytogeography |
2Department of Botany, University of Wisconsin, Madison, Wisconsin 53706 USA
Received for publication December 7, 2005. Accepted for publication May 1, 2006.
ABSTRACT
Recent molecular phylogenetic studies of Solanaceae have identified many well-supported clades within the family and have permitted the creation of a phylogenetic system of classification. Here we estimate the phylogeny for Iochrominae, a clade of Physaleae sensu Olmstead et al. (1999)
, which contains 34 Andean species encompassing an immense diversity of floral forms and colors. Using three nuclear regions, ITS, the second intron of LEAFY, and exons 2 to 9 of the granule-bound starch synthase gene (waxy), we evaluated the monophyly of the traditional genera comprising Iochrominae and assessed the extent of interspecific hybridization within the clade. Only one of the six traditionally recognized genera of Iochrominae was supported as monophyletic. Further, comparison of the individual nuclear data sets revealed two interspecific hybrid taxa and a third possible case. These hybrid taxa occur in the Amotape–Huancabamba zone, a region between the northern and central Andes that has the greatest diversity of Iochroma species and offers frequent opportunities for hybridization in areas of sympatry. We postulate that periodic hybridization events in this area coupled with pollinator-mediated selection and the potential for microallopatry may have acted together to promote diversification in montane Andean taxa, such as Iochrominae.
Key Words: floral evolution • granule-bound starch synthase • interspecific hybridization • LEAFY • phylogeny • pollination • reticulate evolution • speciation
The tropical Andes comprise the pre-eminent hotspot of plant biodiversity, with approximately 15% of all plant species native to that region (Myers et al., 2000
). Many plant families, though cosmopolitan, have centers of diversity in western South America, for example, Ericaceae, Orchidaceae, and Solanaceae (Dressler, 1981
; D'Arcy, 1991
; Luteyn, 2002
). An important contributor to the origin of this diversity is the topological and environmental variation resulting from the uplift of the Andes (Gentry, 1982
; Hooghiemstra et al., 2002
). Phylogenetic studies support an association between the diversification of Andean plants (von Hagen and Kadereit, 2003
; Kay et al., 2005
) and animals (Patton and Smith, 1992
; Bates and Zink, 1994
; Brower, 1994
) and the major episodes of Andean uplift, beginning in the early Miocene (ca. 20 mya) and ending in the Pliocene (ca. 3 mya) (Hoorn et al., 1995
; Hooghiemstra and van der Hammen, 1998
). Indeed, the parallel invasions of higher elevations by numerous plant groups and the coincident radiations of pollinating animals, e.g., hummingbirds (Bleiweiss, 1998
), may explain the "explosive" speciation seen in some Andean groups (Gentry, 1982
; Luteyn, 2002
). Here we investigate the phylogenetic history of Iochrominae, a group of Andean Solanaceae, which have radiated in floral morphology and pollination system (Cocucci, 1999
) and which may serve as a model system for other Andean radiations.
Recent phylogenetic analyses using plastid genes have greatly clarified relationships within Solanaceae and allowed for the creation of a phylogenetic system of classification (Olmstead and Sweere, 1994
; Olmstead et al., 1999
; Martins and Barkman, 2005
; Olmstead et al., University of Washington, personal communication). Iochrominae sensu Olmstead et al. (1999)
is a clade of Physaleae comprising around 34 mainly Andean species traditionally assigned to six genera: Acnistus Schott, Dunalia H.B.K, Eriolarynx (Hunz.) Hunz., Iochroma Benth., Saracha R. and P., and Vassobia Rusby (Table 1). In the Olmstead et al. (1999; R. G. Olmstead, University of Washington, unpublished manuscript) scheme, Iochrominae together with Physalinae and Withaninae form the large clade Physaleae, which is sister to Capsiceae. Although the phylogenetic classification was not accompanied by a morphological reassessment, Iochrominae can be distinguished from other subtribes in Physaleae by the fact that they are all woody shrubs or small trees and often have showy tubular flowers. In a recent morphological phylogenetic analysis (Sawyer, 2005
), all Iochrominae genera except one, Acnistus, were found to be monophyletic, united most notably by the rounded-mucronate shape of the fruiting calyx margin and the presence of sclerosomes in the fruit wall.
|
; Hunziker, 2001
; Table 1). In contrast, a vast majority of taxa within Physalinae, Withaninae, and Capsiceae have small, white or yellow, rotate flowers, and there are no instances of long, tubular, red or purple corollas in these three clades. Thus, the brightly colored tubular flowers likely represent a derived feature that arose within or at the base of Iochrominae.
The great floral diversity of Iochrominae sensu Olmstead et al. (1999)
has misled classifications based on morphology. For example, Hunziker's (2001) morphologically delimited Iochrominae included Oryctes S. Watson, a monotypic tubular-flowered genus native to California and Nevada. Oryctes has since been shown to be nested within Physalinae, probably sister to Leucophysalis (Whitson and Manos, 2005
; Olmstead et al., University of Washington, personal communication). Similarly, Sawyer's (2005/
) morphological cladistic analysis of Physaleae identified an Iochrominae clade that included all genera except Acnistus, which appeared with Tubocapsicum in a distant clade. Although Acnistus and the monotypic Japanese Tubocapsicum share small campanulate-infundibuliform flowers with valvate bud aestivation, molecular studies strongly suggest that Tubocapsicum is more closely related to other Physaleae (e.g., Nothocestrum and Withania) than to Acnistus and other Iochrominae (Olmstead et al., 1999
; Olmstead et al., University of Washington, personal communication).
Another challenge in the systematics of Iochrominae is the potential for hybridization among species and across generic boundaries. Horticulturists have generated several hybrids (e.g., I. australe x I. cyaneum), and botanists have occasionally encountered hybrid populations in nature (Shaw, 1998
; S. D. Smith, personal observation). The ease of crossing, the overlapping species ranges of many Iochrominae, and the observation of natural hybrids suggest that hybridization may have been important in the evolutionary history of Iochrominae. Combined with external sources of information such as morphology, biogeography, and cytology, phylogenetic estimation using multiple genetic markers can help identify instances of hybridization.
In this study, we used three nuclear regions, the internal transcribed spacer (ITS), exons 2 through 9 of the nuclear granule-bound starch synthase gene (GBSSI or waxy), and the second intron of LEAFY (LFY) to estimate the phylogeny of Iochrominae. Both ITS and waxy have been useful in clarifying specific and generic relationships in Solanaceae (e.g., Marshall et al., 2001
; Peralta and Spooner, 2001
; Whitson and Manos, 2005
). LFY introns are increasingly utilized for resolving interspecific relationships and identifying hybrid taxa (e.g., Oh and Potter, 2003
; Howarth and Baum, 2005
), although this is the first study to use LFY in Solanaceae systematics. Our specific objectives were to evaluate the monophyly of the six traditional genera of Iochrominae and to assess the extent of interspecific hybridization. We close by considering our results in a biogeographical context.
MATERIALS AND METHODS
Taxon sampling
This study includes a nearly complete sampling of Iochrominae (Table 1) and a broad sampling of related lineages in the Solanoid radiation. Thirty-three of the 34 commonly recognized species of Iochrominae (all but Eriolarnyx iochromoides) were sampled in this study, as well as three as yet undescribed taxa (Appendix 1). The status of these unnamed taxa is under review by S. Leiva G., and for the purposes of this study, we will use temporary names, indicated by quote marks, based on their likely species epithets (S. Leiva G., Herbario Antenor Orrego, personal communication). For Iochroma peruvianum, a species known only from the type collection, our determination remains tentative because we were unable to find the species in its type locality and have here sampled individuals from another locality that closely resemble the type but have some small differences. For one ingroup species, A. arborescens, multiple individuals were included because the species is extremely widespread and variable.
Three ingroup taxa were suspected to have recent hybrid ancestry: Iochroma "sagasteguii," I. ayabacense, and I. stenanthum. These species are endemic to northern Peru and are often found in sympatry with other species of Iochroma and Acnistus. They share some characteristics of Iochroma (e.g., tubular flowers, purple coloration in the latter two) and some of Acnistus (e.g., yellow-green markings inside the corolla lobes), making their taxonomic affinity unclear. Preliminary chromosome counts for one of these three species, I. ayabacense suggest that it is n = 12 (S. D. Smith and V. Kolberg, University of Wisconsin, unpublished data) as are other species and genera of Iochrominae (Hunziker, 2001
, and references therein). We, therefore, considered these taxa to be possible homoploid hybrids.
The 10 outgroup taxa were selected by reference to the plastid phylogeny of Solanaceae (Olmstead et al., 1999
) and included Nicandreae (Nicandra), Solaneae (Solanum), Capsiceae (Capsicum and Lycianthes) and other members of Physaleae (Leucophysalis, Physalis, Salpichroa, Tubocapsium, and Witheringia) (Appendix 1). Also, included were the Andean genera Cuatresia and Larnax, which have not yet been incorporated into the phylogenetic classification scheme for the family, but appear to belong in Physaleae (R. G. Olmstead et al., University of Washington, unpublished manuscript).
Data collection
Total genomic DNA was extracted from silica-dried leaf material (Chase and Hills, 1991
) using a modified 2x CTAB protocol (Doyle and Doyle, 1987
). ITS was amplified as described in Baum et al. (1998)
with primers ITS leu.1 (Andreasen et al., 1999
) and ITS4 (White et al., 1990
) and sequenced with these two primers plus ITS2 (White et al., 1990
) and ITS3B (Baum et al., 1994
).
The waxy region was amplified using primers 5' and 3' and sequenced using primers GBSSI-A, -B, -CR, and -DR designed by Peralta and Spooner (2001)
. For difficult taxa, four Iochrominae specific primers were designed: F41, F420, R991, and R1235 (Appendix 2). Each 25 µL waxy PCR reaction contained 2.5 µL 10x PCR Buffer (Qiagen, Valencia, California, USA), 2.5 µL of 25 mM MgCl2, 1.0 µL of 10 mM dNTPs, 1.0 µL of each primer (10 µM solutions), 0.125 µL Taq polymerase (5 units/µL), and approximately 100 ng of template DNA. The PCR program was 95°C for 2 min, then 35 cycles of 95°C for 45 s, 56°C for 30 s, 72°C for 3 min, followed by a final extension of 72°C for 5 min.
|
). These sequences were used to create Solanoid specific primers (LFYSOL-F7, -F68, -R700, -R1000 in Appendix 2). The PCR reactions for LFY differed from the waxy reactions in that they contained 2.0 µL MgCl2 and 1.0 µL Q-solution (Qiagen). The PCR program for LFY amplification was 95°C for 4 min, then 35 cycles of 95°C for 1 min, 48°C for 1 min, 72°C for 3 min, followed by a final extension of 72°C for 5 min. All PCR products were purified with AMPure using manufacturer's protocols (Agencourt Bioscience, Beverly, Massachusetts, USA).
Copy number and allelic variants are of concern when using nuclear genes for phylogenetics. The ITS region, as part of the repeating units of rDNA in the nuclear genome, undergoes concerted evolution, potentially homogenizing the many copies (Hamby and Zimmer, 1992
). This may explain why direct sequencing was possible for ITS for all taxa. With the single or low copy nuclear loci, LFY and waxy, direct sequencing often failed to yield a single sequence. In these cases, PCR products were gel-purified with the QIAquick gel extraction kit (Qiagen) and cloned using the pGEM-T easy vector system (Promega, Madison, Wisconsin, USA) following the manufacturer's protocol. Five to eight clones from each product were sequenced.
Sequencing used ABI sequencing reagents (Applied Biosystems, Foster City, California, USA). Each 10-µL cycle sequencing reaction contained 1.0 µL of purified PCR product, 2 pM of primer, 2 µL Big Dye, and 2 µL of sequencing buffer and was cycled through a program of 94°C for 2 min, then 30 cycles of 94°C for 20 s, 47°C for 20 s, and 60°C for 3 min. Reactions were cleaned using CleanSEQ (Agencourt) and run on ABI PRISM 3700 DNA analyzer at the University of Wisconsin Sequencing Facility. Sequences were edited in Sequencher (Gene Codes Corp., Ann Arbor, Michigan, USA) and aligned manually in MacClade 4.0 (Maddison and Maddison, 2000
). Unique clones were maintained in the alignment until data collection was complete. Minor allelic variants (five or fewer substitutions per kilobase) were combined to create a consensus sequence for the species with differences coded as ambiguities. When alleles from a single species differed markedly in sequence or in length (due to indels) but still formed a clade in phylogenetic analyses, the allele giving the shortest branch in a parsimony tree was kept for analysis. In cases where the two alleles of a given species did not consistently form as a clade, both were kept for analysis. We use the term "divergent alleles" for such sequences for the remainder of the paper.
Sequences were examined for evidence of intragenic recombination by visual examination of the spatial distribution of different site patterns. Additionally, the sequences were analyzed for evidence of recombination using the MaxChi (Maynard Smith, 1992
) and GENECONV (Padidam et al., 1999
) methods (with default settings) in the program RDP (Martin and Rybicki, 2000
). These methods have a limited ability to detect recombination, but are still potentially informative (Posada and Crandall,
2001; Posada, 2002
). Final sequence alignments were deposited in TreeBASE (study accession number S1498, http://www.treebase.org).
Phylogenetic reconstruction
For parsimony analyses, all characters were equally weighted, and gaps were treated as missing characters. Heuristic searches were conducted in PAUP*, version 4.0b10 (Swofford, 2002
), using 1000 random taxon addition sequences (holding two trees at each step) with tree-bisection-reconnection (TBR) branch swapping and keeping up to 100 most parsimonious trees (MPTs) per random addition replicate. Similar to Catalán et al. (1997)
, we next completed a heuristic search using the same settings but with 5000 random taxon additions and retaining only trees not compatible with the strict consensus of the first parsimony search (by enforcing the strict consensus as a reverse constraint). If the second search returned only trees longer than first search, then we considered the MPTs from the first search an adequate sample of parsimony tree space. If we found shorter trees, we repeated the process until no additional MPTs were recovered. To estimate clade support, heuristic searches were completed for 1000 bootstrap replicates with 10 random sequence additions (holding one tree at each step), TBR branch swapping, and maxtrees set to 100.
For likelihood analyses, the best fitting model was chosen by hierarchical likelihood ratio tests. Likelihood scores were calculated in PAUP* (Swofford, 2002
) for the following models (in order of increasing complexity): JC, K2P, HKY, HKY+
, HKY++I, GTR+, and GTR++I (Swofford et al., 1996
, and references therein). The most-parsimonious tree (MPT) with the highest likelihood under the JC model was used for calculating likelihoods under more complex models. Likelihood searches were carried out in PAUP* (Swofford, 2002
) using the best fitting model with all the MPTs used as starting trees, TBR branch swapping, and model parameters estimated during the hierarchical likelihood ratio tests.
Bayesian analyses were performed with MrBayes, version 3.1.1 (Ronquist and Huelsenbeck, 2003
). The ITS and LFY intron data sets were each treated as a single data partition, whereas the waxy data set was divided into three partitions: first and second codon positions, third codon positions, and introns. Thus, the combined data set had five total data partitions. Each partition was assigned the best fitting model as suggested by likelihood ratio tests using MPTs from each partition as described previously. Transition/transversion ratio, substitution rates, state frequencies, gamma shape parameters, and proportion of invariant sites were unlinked across partitions and estimated during the Markov Chain Monte Carlo (MCMC) runs. For the individual and combined data sets, we conducted four independent MCMC runs, each with two internal runs (nruns = 2), to give eight tree files for each data set. Each run was initiated with a different starting seed and comprised four linked chains with temperature of 0.2. The chains were run for 5 000 000 generations, sampling every 100 generations, except for ITS, for which we used 15 000 000 generations, sampling every 150 generations. Adequate mixing (sampling of tree and parameter space) was judged by movement among chains and acceptance rates, which should be between 10 and 70%, and, most importantly, by convergence among independent runs with different starting points (Huelsenbeck et al., 2002
). Inadequate mixing in some initial runs was corrected by adjusting the temperature and re-running the analysis. We considered that the runs had converged when the convergence diagnostics provided in sump output approached 1 and when clade credibilities (post burn-in), branch lengths, and topologies were similar across the four independent runs. We discarded 10% of trees as our burn-in period, which appeared to be very conservative given visual inspection of likelihood-by-generation plots. Posterior probabilities (PP) were averaged across runs.
Statistical tests
We estimated the g1 statistic, a measure of phylogenetic signal, for each data set in PAUP* using 10 000 random trees. Significance of the statistic was assessed following Hillis and Huelsenbeck (1992)
.
Incongruence between the three data sets was estimated with the incongruence length difference (ILD) test (Farris et al., 1994
), implemented as the partition homogeneity test in PAUP*. The test was conducted with 1000 replicate partitions, each subjected to heuristic parsimony searches, comprising 10 random taxon addition replicates with TBR branch swapping and keeping no more than 100 trees per random addition replicate. The difference in phylogenetic signal from the three data sets as manifested in differing tree topologies was further examined using Wilcoxon signed-ranks (WSR) tests, also known as Templeton tests (Templeton, 1983
), implemented in PAUP*. A detailed description of the use of WSR tests to compare phylogenetic hypotheses is given in Larson (1994)
. Constrained searches completed in conjunction with WSR tests were carried out with the same settings as unconstrained parsimony searches (described previously).
We also examined incongruence between data sets in a Bayesian framework as described in Buckley et al. (2002)
. We determined whether the combined topology existed within the 95% credible set of trees from each gene. If not, we assumed that the gene in question evolved under a different topology or that the model of evolution was inappropriate. In these cases, we attempted to localize areas of discordance by comparing individual clade credibilities between the individual and combined analyses.
RESULTS
Phylogenetic analyses of individual data sets
ITS—Sequences were completed for all taxa (Appendix 1) and easily aligned to provide a matrix of 803 characters (described in Table 2). We found no evidence of intragenic recombination, either by visual inspection or by use of MaxChi and GENECONV methods in the program RDP (P > 0.05). Relative to LFY and waxy, ITS had low consistency and low phylogenetic signal (Table 2), resulting in many more MPTs. Our initial Bayesian analyses showed variation in clade credibilities across runs, but lengthening the runs to 15 000 000 generations (sampling every 150 generations) produced identical majority-rule consensus trees with less than 5% difference in PP across independent runs for clades with over 50% PP. Also, the convergence diagnostic was 1.0 for each run and acceptance rates were between 10 and 70%, indicating good mixing. Despite thorough exploration of tree space, ITS provided little resolution at any level and showed only a few strongly supported clades, which appeared consistently in the MP, ML, and Bayesian analyses. At the broader level, ITS data suggested that the mostly closely related taxa to Iochrominae are other Physaleae (bootstrap support [BS] 59%, PP = 1.0; Fig. 1), but it did not provide support for any specific Physaleae lineage being sister to Iochrominae. Iochrominae appeared monophyletic (BS = 86%, PP = 1.0; Fig. 1), excluding the spiny Bolivian endemic I. cardenasianum, which apparently is not a member of Iochroma. Within Iochrominae, ITS resolved only small clades, such as Vassobia (BS = 76%, PP = 1.0; Fig. 2) and the C clade (BS = 93%, PP = 1.0, see Fig. 2 legend for explanation of clade names). One interesting feature of the ITS phylogeny was the placement of the U group, which appeared as a clade (BS = 52%; not shown) sister to the rest of Iochrominae (BS = 33%; not shown) in parsimony analyses or as a basal grade in likelihood and Bayesian analyses (Fig. 2).
|
|
|
Bayesian analyses of the waxy data set mixed well as indicated by the convergence diagnostics and the low variation among independent runs. The waxy analyses strongly supported the monophyly of Iochrominae (BS = 100, PP = 1.0; Fig. 1) and its inclusion in Physaleae (BS = 83, PP = 1.0; Fig. 1), perhaps as sister to Physalinae plus Tubocapsicum (BS = 75, PP = 1.0; Fig. 1). Like ITS, waxy showed I. cardenasianum to be distantly related from other iochromas. Further, all waxy analyses divided Iochrominae into a principally northern Andean clade containing Acnistus and Iochroma (A, C, L, F, and U clades; Fig. 2) and a mixed northern, central, and southern Andean clade containing members of Dunalia, Eriolarynx, Saracha, and Vassobia (D, E, S, and V; Fig. 2).
LFY
This region was more variable than waxy (Table 2), and could not be directly sequenced for many taxa. Nevertheless, most clones constituted minor sequence variants that were represented in the final matrix by consensus sequences. However, three of 49 taxa (Eriolarynx lorentzii, I. ayabacense, and I. "sagasteguii") contained two alleles that did not form a clade with others from the same accession. These divergent alleles were kept in the final matrix for phylogenetic analysis. We found no evidence of intragenic recombination in Iochrominae, either by visual inspection or with MaxChi and GENECONV methods implemented in RDP (P > 0.05).
Although LFY sequences were completed for all taxa, this intron could not be aligned outside Physaleae due to the enormous length variation (2.2 kb in Capsiceae vs. 1.4 kb in Iochrominae). Characteristics of the final data set of 43 taxa are given in Table 2. Similar to waxy, final Bayesian analyses mixed well as judged by acceptance rates and agreement among runs. Although LFY was too variable to be informative outside of Physaleae, it provided a good resolution within Iochrominae. Using other Physaleae as outgroup taxa, as indicated by waxy and ITS (Fig. 1), LFY produced an ingroup topology with many of the same well-supported clades that appear in ITS and waxy, but with some differences in relationships among the groups. Unlike waxy but similar to ITS, LFY placed the U clade sister to the rest of Iochrominae (BS 70%, PP 1.0; Fig. 2). As in waxy analyses, LFY supported a northern Andean clade with Acnistus and most of Iochroma (A, C, L, and F; Fig. 2) and a clade with Dunalia, Eriolarynx, Saracha, and Vassobia (D, E, S, and V; Fig. 2). LFY supported a monophyletic group of Acnistus and Acnistus-like iochromas (the A clade, Fig. 2) sister to a clade comprising other Iochroma subclades (C, L, and F; Fig. 2). This is in contrast with waxy, which placed the F clade sister to a clade comprising A, L, and C (but with A unresolved).
Divergent alleles in LFY and waxy
Three species, Eriolarnyx lorentzii, Iochroma "sagasteguii," and I. ayabacense, had divergent LFY alleles, and I. ayabacense also had divergent waxy alleles. In the case of E. lorentzii, one LFY allele formed a clade with E. fasciculata and the other with I. australe (Fig. 2). When the two alleles are constrained to be sister, the resulting trees are significantly longer than the optimal trees (WSR, P = 0.0001–0.0017), suggesting that E. lorentzii alleles are not exclusive and that there may be true genealogical discordance (e.g., due to lineage sorting or hybridization).
One LFY allele of Iochroma "sagasteguii" was sister to a sample of Acnistus arborescens, whereas the other fell in the distantly related U clade (Fig. 2). The LFY alleles of I. ayabacense were split between the C and L clades. When either I. "sagasteguii" or I. ayabacense alleles were forced to form a clade, the resulting trees were significantly longer than unconstrained trees (WSR, P = 0.0001–0.004 for I. ayabacense and P < 0.0001 for I. "sagasteguii"). Iochroma ayabacense also showed divergent waxy alleles, with one allele in the C clade and the other in the L clade, consistent with the LFY analysis (Fig. 2). Constraining the two waxy alleles from I. ayabacense to form a clade resulted in some significantly longer trees (WSR, P = 0.025–0.096). The distant placement of I. "sagasteguii" LFY alleles and I. ayabacense LFY and waxy alleles points to a hybrid origin for these taxa, a possibility that will be explored in more detail in the discussion.
Discordance among genes
The ILD (Farris et al., 1994
) was used as an initial test of "global" congruence among and within data partitions. An ILD test indicated that the assignments of characters to the three waxy partitions (first and second codon positions, third codon positions, and introns) was not significantly different from random (P = 0.70), suggesting that waxy can be treated as a single data partition. In contrast, pairwise comparisons of the ITS, LFY, and waxy (excluding non-Physaleae outgroups and putative hybrids) all yielded significant ILD tests (P < 0.01), indicating that the three data sets are not drawn from the same population of characters (but see Darlu and Lecointre, 2002
; Hipp et al., 2004
). We attempted to localize the discordance by repeating the ILD test with successively pruned data sets (Table 3). We divided the data set into three parts, the ACLF group, the DESV group and the U group, and we found that only the DESV returned significant P-values (Table 3). However, simply deleting the DESV taxa from the larger clade did not result in insignificant ILD results (not shown), suggesting that it was not the sole source of incongruence.
|
) among the three gene trees (Fig. 2). Three of these cases were due to differences in the placement of divergent LFY or waxy alleles. We used Templeton tests to compare the remaining five sources of hard incongruence (Table 4). In all cases, one or the other partition failed to reject the conflicting resolution at the P < 0.05 level. This suggests that the incongruence detected by ILD tests is "diffuse" rather than due to particular points of discordance.
|
Parsimony analysis of the combined data set of 40 taxa and 4023 characters yielded 12 MPTs (Table 2) and increased support for many of the clades observed in individual data sets (Fig. 2). Similar results were obtained for ML and Bayesian analyses. For example, among individual analyses, the A clade only appeared in the LFY tree (BS = 74%; PP = 1.0; Fig. 2), but it appeared in the combined analysis with a BS of 90% and PP of 1.0. Likewise, the placement of the U clade sister to the rest occurred with moderate support in LFY (BS = 70%, PP = 1.0; Fig. 2) and weak support in analyses of ITS (BS = 33%, PP = 0.12; not shown), but appeared strongly supported in the combined analysis (BS = 90%, PP = 0.99; Fig. 2). Nonetheless several areas on the combined tree remain unresolved, most notably within the DESV clade and within the A clade. Also, there were differences among modes of analysis. Clades C, L, and F together formed a clade in parsimony and ML searches of the combined data (Fig. 2), but Bayesian analyses showed clade F as sister to an A, C, and L clade and clade A sister to clade L with high posterior probability (PP = 0.96–1.0) at all relevant nodes (tree not shown, but see Fig. 2 caption). Exploration of pruned data sets (not shown) established that the resolution among the A, C, L, and F clades in a Bayesian framework is very sensitive to the inclusion or exclusion of L and to model choice (e.g., whether data partitions were allowed to evolve under different models or whether they were linked as in traditional likelihood searches).
Congruence in a Bayesian framework
As an additional assessment of congruence, we compared the results of the Bayesian analysis of the combined data set (described previously) with the results from Bayesian analyses of individual data sets that had been pruned to the same 40 taxa (Buckley et al., 2002
). The results of these runs are provided in Appendix S1 (see Supplemental Data accompanying the online version of this article). We found that there were no trees that were shared between the posterior distributions (post burn-in) of the individual and combined data sets. This is perhaps not surprising given that there are 1.3 x 1055 possible unrooted trees for 40 taxa and thus a fairly small chance that different data sets would sample exactly equivalent topologies.
We next examined localized points of disagreement among data sets in the Bayesian framework. Within the ACLF group, we observed that LFY had a PP of 0.0 for F sister to ACL, whereas waxy had a PP of 0.0 for the F sister to C topology. This suggests that there may be true genealogical discordance between LFY and waxy within the ACLF clade (Table 4, conflict 5). A contrasting result was found with respect to the placement of the U clade (sister to the rest in LFY and sister to A, C, L, and F in waxy). We found that trees with U sister to the rest of Iochrominae (the "U-sister" topology), as suggested by LFY, appeared in waxy posterior distributions with a PP of 0.0078. Similarly U was sister to A, C, L, and F (the "U-nested" topology) in the LFY posterior with a PP of 0.001. While these values are lower than the traditional 0.05 threshold, the fact that both topologies were present in the posterior distributions for both data sets suggests that there may not be hard incongruence (consistent with the WSR tests, Table 4). On the other hand, the fact that the combined analysis supports the U-sister topology more strongly than does LFY alone, suggests that U-sister is a more plausible hypothesis at this time than U-nested.
DISCUSSION
Position of Iochrominae in Solanaceae
Our goal in outgroup sampling was to confirm the monophyly of Iochrominae and verify that it belongs in Physaleae as indicated by plastid data. Indeed, once Iochroma cardenasianum is excluded, Iochrominae appears to be monophyletic. Plastid data (Olmstead et al., University of Washington, personal communication) confirm the distant relationship of I. cardenasianum to Iochrominae and place it within the Datureae.
Our data support the inference that Iochrominae is part of Physaleae, but its relationship to other taxa remains unclear. Of the three markers, the LFY intron could not be readily aligned with the more distant outgroups, and ITS provided little resolution among Physaleae (Fig. 1). However, analysis of waxy alone and combined analysis of waxy and ITS strongly supported Iochrominae as sister to Physalinae sensu Olmstead et al. (1999)
plus Tubocapsicum. This result disagrees with the most recent plastid phylogeny, which places Deprea plus Larnax sister to Iochrominae, albeit with weak support (Olmstead et al., University of Washington, personal communication). Resolving the lineages that comprise Physaleae and the relationships among them will require increased sampling and perhaps additional markers.
Taxonomic implications for genera of Iochrominae
Acnistus
In Hunziker's (1982) revision of Acnistus, he acknowledged that Acnistus has greatest affinity to the genus Iochroma. The important differences he noted between them were the small flowers and anthers of Acnistus, the calyx (accresent in Iochroma but not in Acnistus), and the bud aestivation (induplicate in Iochroma and valvate in Acnistus). Confusing this demarcation are a few species currently placed in Iochroma that have the latter two characteristics of Acnistus. For example, I. ellipticum and I. confertiflorum, two large-flowered species that were transferred from Acnistus by Hunziker (1977
, 1982
), have valvate bud aestivation and lack a strongly accresent calyx. This combination of traits is also found in two recently named species, I. edule and I. salpoanum (Leiva, 1995
; Leiva et al., 2003) and in I. peruvianum. Furthermore, field observations of these five iochromas (S. D. Smith, personal observation) indicate that they share with Acnistus a conspicuous green mark on the inner surface of the corolla lobe, which fades to yellow as the flower ages (Fig. 3). Thus, it is not surprising that Acnistus and these five other species form a well-supported clade in our analyses (clade A), but whether this group should be officially segregated from Iochroma deserves careful consideration and will be discussed further (see Iochroma).
|
). The three accessions representative of this traditional species do not form a monophyletic group on any of the gene trees (though a clade is not contradicted by waxy). One possible explanation is that there has been incomplete lineage sorting within the A clade. This seems unlikely because our screen of clones revealed no allele sharing among other species in the A clade: all alleles from a given group A species formed a clade at all loci. Another interpretation is that A. arborescens refers to a lowland progenitor form that has given rise to multiple novel higher-elevation forms, similar to the case of Lisianthus skinneri (Systma and Schaal, 1985
). Alternatively, because A. arborescens may occasionally hybridize in nature with related higher-elevation taxa such as Iochroma confertiflorum (S. D. Smith, personal observation), it is possible that different A. arborescens populations have acquired different introgressed alleles from other members of the A clade.
Dunalia
Hunziker's (1960) delimitation of Dunalia centered on a single character, the presence of enlarged and showy "stapets," which appear as winged or toothed lateral appendages emerging from the filament bases at the point of their insertion on the corolla tube. Our analyses suggest that Dunalia sensu Hunziker (1960)
is not monophyletic. Notably, the type species, D. solanacea appears more closely related to Saracha than to other Dunalia species. Whereas other Dunalia species are xerophytes of the central and southern Andes, D. solanacea is a northern Andean cloud forest shrub with a dense indumentum of stellate hairs, anisogeminate leaves, and small, yellow-green, trumpet-shaped flowers. Although its placement within Saracha could be a phylogenetic artifact (note that D. solanacea has a long terminal branch for all genes), there is no evidence of an association between this species and the other "Dunalia" species.
The remaining four Dunalia species are similar to each other in morphology, distribution, and habit; however, they do not form a clade in any of the trees. Furthermore, one species traditionally placed in Iochroma, I. parvifolium, appears more closely related to some dunalias. However, the association of I. parvifolium with D. brachyacantha and D. spinosa is reasonable given its spiny xerophytic habit and tubular purple flowers. Iochroma parvifolium was placed in Iochroma as opposed to Dunalia because it lacks the showy stapets (Hunziker, 1977
). Nevertheless, close examination of fresh flowers of I. parvifolium in the field revealed small, tooth-like expansions of the stapets, which are hard to detect in dried specimens (S. D. Smith, personal observation). Also, during the course of collection trips, one population of I. parvifolium was found to be gynodioecious, a condition found in some Dunalia species (S. D. Smith, personal observation). Iochroma species (members of A, C, L, F, and U) are invariably hermaphroditic and never spiny, making I. parvifolium an unlikely Iochroma. The epithet "parvifolia" does not exist in Dunalia, but transferring I. parvifolium to Dunalia is confounded by the fact that D. solanacea, the type species, is not associated with the other "Dunalia" species, making the taxonomic future of Dunalia uncertain.
Eriolarynx
The three species of Eriolarynx, recently segregated from Vassobia, can be distinguished from other Iochrominae by the dense ring of trichomes inside the corolla (Hunziker, 2000
). Our analysis upholds the monophyly of Eriolarynx, with the addition of I. australe. This species was originally described in Iochroma (Grisebach, 1874
), but later transferred to Acnistus (Grisebach, 1879
) and then to Dunalia (Sleumer, 1950
). Iochroma australe was not a good fit in Iochroma because its variable flowers can sometimes be short and funnel-shaped and because the corolla interior is densely pubescent near the base, whereas other Iochromas are typically glabrous. Further, it lacks the valvate aestivation of Acnistus and the characteristic filament appendages of Dunalia. The hairy flowers suggest a better fit with Eriolarynx despite the fact that the three described species typically have rotate or campanulate flowers, while I. australe has a funnel-shaped or tubular corolla. Geography also argues for this placement because both I. australe and Eriolarnyx are restricted to Bolivia and Argentina. There is no good argument against creating the new combination E. australe, except that this may prove to be only a temporary solution if it becomes necessary to combine the entire DESV clade into a single genus (with or without other elements of Iochrominae).
Iochroma
Species currently identified as Iochroma were not found to form a clade, even after the misplaced I. australe and I. parvifolium are ignored. One group of iochromas, the U clade appears as sister to remainder of Iochrominae. We consider this "U-sister" position to be strongly supported by our study for three reasons. First, two of the three loci sampled, ITS and LFY, support or are compatible with the "U-sister" topology. Second, heuristic searches using the waxy data constrained to be consistent with "U-sister" topology do not result in trees that are significantly longer than unconstrained trees (Table 4). Last, despite the differences in topology among loci, support for a "U-sister" relationship is highest in the combined analysis. Specifically the combined analysis of all three genes yielded a 90% bootstrap, as contrasted with a 78% bootstrap support for this relationship in a two-gene combined analysis of LFY and ITS (not shown). This pattern suggests that even though waxy does not return U as sister to the rest of Iochrominae, the waxy data do contain some support for this topology (Olmstead and Sweere, 1994
).
The U group is distinguished from species in the ACLF clade by the form of the corolla and the androecium. Flowers of ACLF (excluding Acnistus) are funnel-shaped or tubular, whereas those of the U group are salverform. Also, the filaments are attached near the base in ACLF, while in the U group they are attached near the middle of the corolla tube (often with a visible bump at the point of attachment, e.g., I. grandiflorum, Fig 3.). The most extreme example of filament adnation in the U group is I. "tingoense" in which the anthers are more or less sessile on the corolla. Thus, even if one doubted the sister group relation between the U clade and other Iochrominae, there is reason to believe that the U clade is divergent from other traditional iochromas.
If the U clade (and I. australe and I. parvifolium) were excluded from Iochroma and if Acnistus were expanded to include the entire A clade (discussed previously), then one could imagine assigning only members of clades C, L, and F to Iochroma. However, this decision would be premature considering that it is not certain from these data that C, L, and F form a clade. Furthermore, there are no clear morphological differences between Acnistus and Iochroma, largely because I. squamosum and I. lehmannii (clade L) possess a mixture of traits from clade A on the one hand and clades C and F on the other; the bud aestivation is induplicate, resulting in wide corolla lobes and plaits in the corolla tube, like C and F, but the yellow flowers lack anthocyanins (Hunziker, 1982
) and have the green markings on the corolla lobes, as in clade A. The other alternative, if we are to only recognize monophyletic groups, is to sink Acnistus into Iochroma. We also note that in a rank-independent system of nomenclature, Acnistus could be defined as a monophyletic group within a monophyletic Iochroma.
Saracha
This small genus of high-elevation treelets is morphologically well defined, including two species of páramo treelets with small coriaceous or subcoriaceous leaves and funnel-shaped or campanulate flowers that can be purple or yellow with purple spots (Alvarez, 1996
). Dunalia solanacea, which often appears nested within Saracha, does not share any obvious features with Saracha except for its high-elevation distribution and occurrence in the northern Andes. As noted, D. solanacea has a long terminal branch for all tree genes, raising the possibility that its placement within Saracha is an artifact. Moreover, although Saracha only appears monophyletic in ITS trees and not in LFY or waxy trees, we note the sister relationship of S. quitensis and S. punctata does appears in Bayesian analyses of LFY with PP 0.08 and in those of waxy with PP 0.33 (Appendix S1, see Supplemental Data accompanying the online version of this article). Thus, we consider it premature to conclude that Saracha is nonmonophyletic.
Vassobia
Among the genera of Iochrominae, Vassobia is the only one that appeared monophyletic in all analyses. Vassobia includes two southern Andean species with small, purple, campanulate, glabrous flowers: V. dichotoma a cloud forest tree restricted to Bolivia, and V. breviflora, a widespread spiny shrub (Hunziker, 1984
, 2001
). The stapets of Vassobia are expanded to form small "auricles" similar to the appendages found in Eriolarynx (Hunziker, 2001
). Considering that species of Eriolarynx formerly belonged to Vassobia, one might have expected a sister relationship between the genera. These data neither support nor strongly contradict this inference.
Hybridization in Iochrominae
Identifying hybrid taxa is a challenge for phylogenetics because reticulation erodes the strictly tree-like process of evolution assumed by most phylogenetic methods (McDade, 1990
). Nonetheless, even when species trees are reticulate, gene trees will be strictly divergent structures so long as the rate of intragenic recombination is low relative to the rate at which lineage sorting occurs. Given that homoploid hybrid taxa potentially carry genetic contributions from one or both parents, we may observe divergent alleles on a single gene tree (with alleles associated with each parent) or disagreement among gene trees, with hybrid alleles appearing related to one parental lineage on one tree and to the other parent on a different gene tree. Thus, we can test hypotheses of hybrid ancestry by identifying divergent alleles or points of conflict among gene trees (Doyle, 1992
; Maddison, 1997
). In this study, our sampling included three taxa, Iochroma ayabacense, I. "sagasteguii," and I. stenanthum, which we had hypothesized to be of hybrid origin due to their distribution and morphology.
Iochroma ayabacense was hypothesized to be an interspecific hybrid between I. cyaneum and I. squamosum. Iochroma ayabacense occurs in at high elevations (2600–2700 m a.s.l.) around the city of Ayabaca in northern Peru, often in proximity to populations of its putative parents, I. squamosum and I. cyaneum. The infrequent I. squamosum favors mildly disturbed habitats like forest gaps or riparian areas, whereas the widespread I. cyaneum tolerates drier conditions and open habitats like roadsides and pastures. The two putative parents are, however, found occasionally in close proximity, for example, when a road passes through a patch of forest. Several morphological features pointed to the possibility that I. ayabacense was a hybrid between these two. It has peculiar yellowish-purple flowers intermediate between the yellow I. squamosum and the purple I. cyaneum, and it has yellow-green markings inside the corolla, which are signatures of clades A and L. Our phylogenetic analyses revealed divergent alleles of I. ayabacense in both waxy and LFY trees, and in each case, one I. ayabacense allele fell in clade C and one in clade L (Fig. 2). In ITS trees, I. ayabacense appeared to be sister to I. squamosum in the L clade (Fig. 2). Considering these gene trees together with its distribution and morphology, we conclude that I. ayabacense is a hybrid between I. cyaneum and I. squamosum. Further field research and genetic data would be needed to determine if I. ayabacense is best interpreted as a hybrid species or a transient hybrid form that lacks sufficient permanence to warrant species status.
Iochroma "sagasteguii" has small white flowers with greenish markings inside the corolla that resemble Acnistus. However, the pubescence on the calyx and corolla, the slightly induplicate bud aestivation, and the extended area of filament adnation are reminiscent of species in the U group. Although the distribution of I. "sagasteguii" is not well known, in some localities in northern Peru, it grows within a few kilometers of populations of I. stenanthum, I. cornifolium, I. grandiflorum, and I. cf. peruvianum and within 15 km of populations of Acnistus arborescens. Similar to I. ayabacense, genetic evidence supported the hypothesis of hybrid ancestry in I. "sagasteguii." We found divergent alleles in the waxy tree, with one allele in the U group and another in clade A. The genetic data, the morphology, and the geography point to I. grandiflorum and A. arborescens as the most likely parental species.
Iochroma stenanthum was the third suspected hybrid. It occurs in northern Peru and has long, tubular, pubescent flowers, most similar to I. cornifolium (Leiva et al., 1998
), but with more triangular corolla lobes and yellow-green markings inside the corolla lobes as in clade A. The corolla color, which fades from cream at the base to purple at the apex, suggests that it is the result of crossing a white-flowered species (e.g., Acnistus arborescens) and a purple-flowered species. Iochroma stenanthum occurs in close proximity of populations of the putative parents, I. cornifolium and A. arborescens. However, our data were insufficient to resolve the relationship of I. stenanthum to other Iochrominae. Its position varied among gene trees and was generally poorly supported. This pattern might be ascribed to lineage sorting, but its morphology is so strongly indicative of a hybrid ancestry that we favor the hypothesis that I. stenanthum is the product of a more ancient hybridization event whose genetic signatures have been blurred by subsequent evolution.
Biogeographical context of the Iochrominae radiation
Simpson (1975)
recognized that phytogeographical distributions in the Andes tend to coincide with the geologically defined structural units of the Cordilleras. Many subsequent authors have observed such a relationship (e.g., Berry, 1982
; Luteyn, 2002
), although the exact delimitation of the structural units and associated phytogeographic zones varies slightly among studies. For instance, Berry (1982)
modified Simpson's (1975) structural units by recognizing the Amotape–Huancabamba zone (A–H zone), an area of low elevation between the northern and central Andes (4–8°S), as a separate unit. Weigend (2002
; Weigend et al., 2004
) supported Berry's distinction, noting the large number of A–H zone endemics, and additionally suggested distinguishing the Andes below 18°S as the southern Andes (as in Fig. 3). As a basis for discussing the biogeography of Iochrominae, it is useful to divide the tropical Andes into northern, central, and southern regions, to recognize the A–H zone as a distinct unit, and to divide the central Andes into a region north of the Pisco deflection (14°S; Berry's Cordillera Central and Occidental) and a region south of the deflection (Berry's Cordillera Oriental).
Similar to other plant groups that have radiated in the Andes (e.g., Fuchsia and Nasa), distribution patterns of Iochrominae species and clades strongly reflect the structural units of the Andes (Fig. 3). The diverse ACLF clade, excluding the weedy Acnistus arborescens and the Galapagos endemic Iochroma ellipticum, is restricted to the Andes from 5°N to 8°S, the southern boundary of the A–H zone. The DEV group contains taxa that only occur below 8°S, while its probable sister group, clade S, is widely distributed from 9°N to 16°S. Clade U straddles the ACLF and the DEV groups, with a distribution from 4°S (the northern limit of A–H zone) to 10°S.
Despite the clear patterns along the latitudinal gradient, we do not observe strong east–west separation of clades as has been the case in many Andean groups (Berry, 1982
; Slade and Moritz, 1998
; Brower, 1994
). Although there is some tendency for greater species richness on the western cordilleras, several taxa, e.g., I. calycinum and Dunalia solanacea, are known to occur on both sides of the Andes. However, the distribution of many species remains poorly characterized. With increased collecting effort, it may eventually be possible to determine if Iochrominae distributions follow east–west structural units as closely as they do north–south units.
Iochrominae show a center of diversity in the A–H zone, where 16 of 33 (48%) species (excluding I. cardenasianum) occur, 11 of which are restricted to this zone. This enhanced diversity can be attributed to the overlap of the ACLF and U clades. The A–H zone is characterized by fragments of the Cordilleras, usually less than 3500 m a.s.l, separated by valleys that dip down to ca. 1000 m a.s.l. Iochrominae prefer cloud forest or Andean scrub forest between 2300 and 2800 m a.s.l and are abundant in the high elevation valleys of the A–H zone. In some areas, as many as five species may occur over the distance of a few kilometers. The proximity coupled with the ease of crossing has resulted in several hybrid taxa, as revealed by this study, all of which are confined to this A–H zone (Fig. 3).
Here we have examined the three putative hybrid taxa with three loci, but this represents only a first attempt at exploring hybridization in Iochrominae. Further investigation into the potential hybrid ancestry of all Iochrominae should include samples of multiple individuals and populations per taxon, additional chromosome counts, statistical morphometric studies, characterization of species distributions, and analysis of mitochondrial or plastid markers. Greater sampling of individuals, taxa, and genes will permit a more fine-tuned estimate of the frequency of hybridization and introgression in Iochrominae history.
As documented in this study, episodes of hybridization have clearly impacted the evolutionary history of Iochrominae. However, considering the amount of agreement among the three nuclear markers, it appears that these events have not entirely obscured the underlying divergent phylogenetic history, having only clouded the branching pattern in some parts of the tree. Furthermore, the presence of leaky species boundaries has not apparently precluded the diversification of Iochrominae. In addition to being the most florally diverse subtribe in Physaleae and perhaps Solanoideae, Iochrominae also boasts the greatest diversity of pollination systems (Cocucci, 1999
). Perhaps the combination of pollinator-mediated selection, microallopatry in dissected Andean habitats and episodic hybridization have together permitted the explosion of floral diversity seen in Iochrominae.
Tribe
Subtribe
Taxon—GenBank accession nos.: ITS, LFY, waxy; voucher information.
Nicandreae
Nicandra physaloides (L.) Gaertn.—DQ314155, DQ309515, DQ309465; Peru. Dept. Amazonas. Prov. Chachapoyas, 6.24291°S 77.87443°W, 2250 m, 11-II-04, Smith 369, WIS.
Datureae
Iochroma cardenasianum Hunz.—DQ314156, DQ309516, DQ309466; Bolivia. Dept. Potosí. Carretera Potosi-Orkhola-Tumusla. 20.39638°S 65.56287°W, 3099 m, 18-II-04, Smith 385, WIS.
Solaneae
Solanum lycopersicum L.—DQ314157, DQ309517, DQ309467; UW–Madison, Botany Living Collections s.n.
Capsiceae
Capsicum lycianthoides Bitter—DQ314158, DQ309518, DQ309468; Ecuador. Prov. Pichincha. 0.0157°S 78.680°W, 2250 m, 23-XII-02, Smith 203, WIS.
Lycianthes inaequilatera Bitter—DQ314159, DQ309519, DQ309469; Ecuador. Prov. Pichincha. 0.326°S 79.000°W 800 m, 25-XII-02, Smith 210, WIS.
Physaleae
Salpichroinae
Salpichroa tristis Walp.—DQ314160, DQ309520, DQ309470; Bolivia. Dept. Potosí. Ca. 19.5°S 65.45°W, 4020 m, 18-II-04, Smith 382, WIS.
Physalinae
Physalis peruviana L.—DQ314161, DQ301514, DQ309471; Ecuador. Prov. Pichincha. Gardens of Herbario Nacional (QCNE), 2800 m, 1-I-03, Smith 217, WIS.
Leucophysalis grandiflora (Hook.) Rydb.—DQ314162, DQ301515, DQ309472; Olmstead S-30, WTU.
Witheringia solanacea L'Herit.—DQ314164, DQ301517, DQ309474; D'Arcy 16399, MO.
Withaninae
Tubocapsicum anomalum (Franchet & Savat.) Makino—DQ314163, DQ301516, DQ309473; Chen 231, MO.
[Subtribe not known]
Cuatresia harlingiana Hunz.—DQ314165, DQ301518, DQ309475; Ecuador. Prov. Pichincha. 0.0157°S 78.680°W, 2250 m, 24-XII-02, Smith 204, WIS.
Larnax sachapapa Hunz.—DQ314166, DQ301519, DQ309476; Ecuador. Prov. Pichincha. 0.0157°S 78.680°W, 2250 m, 24-XII-02, Smith 205, WIS.
Iochrominae
Acnistus arborescens (L.) Schlecht.—DQ314173, DQ301528, DQ309483; Costa Rica. Prov. Puntarenas. Las Cruces B. S., 1992, Bohs 2428, UT. A. arborescens (L.) Schlecht.—DQ314181, DQ301536, DQ309491; Ecuador. Prov. Pichincha. 0.3260°S 79.000°W, 750 m, 25-XII-02, Smith 209, WIS. A. arborescens (L.) Schlecht.—DQ314183, DQ301538, DQ309493; Peru. Dept. Cajamarca. 7.42409°W 78.90111°S, 1976 m, 10-I-04, Smith 312, WIS.
Dunalia brachyacantha Miers—DQ314172, DQ301527, DQ309482; Argentina. Prov. Jujuy, 2100 m, 20-IX-02, Nee and Bohs 50811, NY. D. obovata Dammer—DQ314192, DQ301547, DQ309499; Peru. Dept. Junin. 11.34919°S 75.57408°W, 2679 m, 8-III-04, Smith 458, WIS. D. spathulata (Ruíz & Pav.) Braun & Aschers—DQ314198, DQ301554, DQ309506; Peru. Dept. Huanuco. 9.83831°S 76.11503°W, 1842 m, 6-III-04, Smith 452, WIS. D. solanacea H. B. & K.—DQ314174, DQ301529, DQ309484; Ecuador. Dept. Pichincha, ca. 0.23°S 78.75°W, ca. 2200 m, 31-XII-02, Smith 211, WIS. D. spinosa Dammer—DQ314188, DQ301543, DQ309495; Bolivia. Dept. Potosi, ca. 19.6°S 65.6°W, 4020 m, 18-II-04, Smith 379, WIS.
Eriolarynx lorentzii (Dammer) Hunz.—DQ314171, DQ301525/DQ301526 (allele A/allele B); DQ309481; Argentina. Prov. Tucuman, 26.633°S 65.467°W, 1700 m, 12-II-1966, Hawkes et al. 3452, BIRM. E. fasciculata (Miers) Hunz.—DQ314196, DQ301552, DQ309504; Bolivia. Dept. Cochabamba, 17.46477°S 65.75217°W, 3180 m, 26-II-04, Smith 432, WIS.
Iochroma australe Griseb.—DQ314189, DQ301544, DQ309496; Bolivia. Dept. Chuquisaca. 20.78477°S 65.04088°W, 3038 m, 18-II-04, Smith 390, WIS. I. ayabacense S. Leiva—DQ314194, DQ301549/DQ301550 (allele A/allele B), DQ309501/DQ309502 (allele A/allele B); Peru. Dept. Piura. 4.61462°S 79.71178°W, 2701 m, 15-I-04, Smith 337, WIS. I. calycinum Benth.—DQ314201, DQ301557, DQ309512; Ecuador. Prov. Pichincha. 0.24577 S 78.80903 W, 1834 m, 27-XII-04, Smith 471, WIS. I. confertiflorum (Miers) Hunz.—DQ314176, DQ301531, DQ309486; Ecuador. Prov. Loja, 4.1316°S 79.9218°W, 1582m, 15-I-03, Smith 237, WIS. I. cornifolium Miers—DQ314177, DQ301532, DQ309487; Ecuador. Prov. Loja, 4.0869°S 79.9356°W, 2570 m, 15-I-03, Smith 242, WIS. I. cyaneum (Lindl.) M. L. Green—DQ314180, DQ301535, DQ309490; Ecuador. Prov.
brackets on each tree. The labels D, E, S, and V indicate members of the traditional genera Dunalia, Eriolarynx, Saracha and Vassobia, respectively. Iochroma has been divided into smaller clades: A, including Acnistus and Acnistus-like iochromas; C, containing the type I. cyaneum; L, after I. lehmannii; F, after I. fuchsioides; and U after I. umbellatum. Solid bracket lines indicate monophyly; dashed bracket lines indicate non-monophyly. The 
's in the combined tree designate points of difference between the combined likelihood tree and combined Bayesian consensus; PP for clade ACL is 0.97 and PP for clade AL is 1.0 (see text, Results, Combined analysis)