HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

What's this?

This Article Abstract Full Text (PDF) Supplementary Data Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Web of Science (10) Citing Articles via Google Scholar Google Scholar Articles by Stefanovic, S. Articles by Costea, M. Search for Related Content PubMed Articles by Stefanovic, S. Articles by Costea, M. Agricola Articles by Stefanovic, S. Articles by Costea, M. Social Bookmarking                            What's this?

Delimitation of major lineages within Cuscuta subgenus Grammica (Convolvulaceae) using plastid and nuclear DNA sequences¹

Saa Stefanovi⁴, Maria Kuzmina and Mihai Costea

²Department of Biology, University of Toronto at Mississauga, Mississauga, Ontario L5L 1C6, Canada; and ³Department of Biology, Wilfrid Laurier University, Waterloo, Ontario N2L 3C5, Canada

Received for publication September 20, 2006. Accepted for publication February 15, 2007.

ABSTRACT

Subgenus Grammica, the largest and most diverse group in the parasitic genus Cuscuta, includes 130 species distributed primarily throughout the New World, with Mexico as its center of diversity. To circumscribe the subgenus and assess the relationships among its major lineages, we conducted the first phylogenetic study of Grammica using plastid trnL-F and nrITS sequences from a wide taxonomic sampling covering its morphological, physiological, and geographical diversity. With the exception of one species belonging elsewhere, the subgenus was found to be monophyletic. The results further indicate the presence of 15 well-supported major clades within Grammica. Some of those lineages correspond partially to earlier taxonomic treatments, but the majority of groups are identified in this study for the first time. The backbone relationships among major clades, however, remain weakly supported or unresolved in some cases. The phylogenetic results indicate that the fruit dehiscence character is homoplastic, thus compromising its value as a major taxonomic and evolutionary feature. While several striking cases of long-distance dispersal are inferred, vicariance emerges as the most dominant biogeographical pattern for Cuscuta. Species placed within one of the clades with a predominantly South American distribution are hypothesized to have substantially altered plastid genomes.

Key Words: Convolvulaceae • Cuscuta • Grammica • molecular phylogeny • nuclear ribosomal ITS • parasitic plants • plastid trnL-F

The parasitic mode of life arose at least 11 times independently during the evolution of flowering plants (Nickrent, 2002 ; APG II, 2003 ) and is found in approximately 4000 plant species (Nickrent et al., 1998 ) representing 1% of the known angiosperm diversity. Parasitism is frequently associated with the extreme reduction or modification of vegetative structures as well as rampant convergence with other parasitic taxa, rendering an assessment of homology with other plant lineages quite hard (Kuijt, 1969 ). For these reasons, parasitic plants in general, and holoparasites in particular have been notoriously difficult to study from a systematic and taxonomic point of view. The resulting lack of knowledge of relationships within parasitic lineages as well as their precise relationships to autotrophic relatives hampers our ability to conduct detailed comparative studies and to understand the sequence of events that have shaped the evolution of these fascinating plants (Nickrent et al., 1998 ; Futuyma, 2004 ).

The genus Cuscuta represents one such taxonomically problematic group. Comprising some 165–175 currently described species, Cuscuta is nearly cosmopolitan in distribution with its species found on every continent (except Antarctica), ranging from the 60th parallel north in Europe and Asia, to the Cape region of South Africa, and as far south as the 47th parallel in Argentina and Chile (Yuncker, 1932 ; Hunziker, 1950 ; Mabberley, 1997 ). All members of this genus are vines with twining, slender, pale stems, with reduced, scale-like leaves, and no roots. These stem parasites are attached to the host by haustoria and depend entirely (or nearly so) on their hosts to supply water and nutrients (Kuijt, 1969 ; Dawson et al., 1994 ). Most Cuscuta species are also characterized by reduced amounts or the complete absence of chlorophylls (van der Kooij et al., 2000 ) even though some species are capable of limited and localized photosynthesis (Dawson et al., 1994 ; Hibberd et al., 1998 ). Various species (commonly known as dodders) are capable of parasitizing a wide range of herbaceous and woody crop plants, but for the most part they do not cause significant agricultural losses due to the effectiveness of currently available methods of control (reviewed by Dawson et al., 1994 ; Costea and Tardif, 2006 ). Members of this genus were recently implicated as vectors in the horizontal transfer of mitochondrial genes in plants (Mower et al., 2004 ).

Traditional classifications largely ignored the question of Cuscuta's precise relationships with nonparasitic relatives, owing mainly to the lack of useful taxonomic characters. An association with the Convolvulaceae was recognized early on, based on reproductive morphology, but few attempts were made to propose a more detailed scheme of relationships between Cuscuta and nonparasitic members of the family. The approaches taken have fallen into two categories, either recognition of Cuscuta as a separate monotypic family, implying a sister-group relationship to the rest of Convolvulaceae or placement of Cuscuta within Convolvulaceae under various taxonomic ranks (but without any further implications on its possible relationships). Examples of the former approach include the classifications by Dumortier (1829) and Roberty (1952 , 1964 ), followed by most major synoptic works on flowering plants (e.g., Cronquist, 1988 ; Takhtajan, 1997 ). The latter approach includes recognition of Cuscuta as tribe Cuscuteae (Choisy, 1845 ; Bentham and Hooker, 1873 ; Baillon, 1891 ; Hallier, 1893 ; Peter, 1897 ; Austin, 1998 ) or as subfamily Cuscutoideae (Peter, 1891 ; Melchior, 1964 ). Molecular phylogenetic studies conducted on a broad sampling of Solanales indicated that not only was Cuscuta a part of the Convolvulaceae clade (Stefanovi et al., 2002 ), but also that it was nested within that family, with at least two nonparasitic lineages diverging before Cuscuta (Stefanovi and Olmstead, 2004 ).

Within Cuscuta, Engelmann (1859) recognized three groups based primarily on stigma and style morphology. These groups were formally adopted by Peter (1897) and later by Yuncker (1932) as subgenera (Fig. 1). Subgenus Monogyna has a single style, partially to completely undivided, with a variety of stigma shapes. Subgenera Cuscuta and Grammica are characterized by two distinct styles and can be distinguished by their stigma morphology (elongated and linear vs. short and capitate, respectively). Plastid and nuclear sequence data obtained for a limited number of taxa identified three lineages consistent with the traditionally proposed subgenera and resolved subgenus Monogyna as the sister to the rest of the genus and subgenera Cuscuta and Grammica as sister to each other (Stefanovi et al., 2002 ; Revill et al., 2005 ). However, there is an indication, based also on a limited number of species, that the South African members of subgenus Cuscuta from section Africanae are in fact more closely related to subgenus Grammica than to the other species from subgenus Cuscuta (McNeal, 2005 ). To date, Cuscuta has not been the subject of broad molecular phylogenetic analyses.

View larger version (19K): [in this window] [in a new window]   Fig. 1. Precladistic scheme of classification for Cuscuta, adopted and modified from Yuncker (1932) . The scheme is based primarily on style and stigma morphology as well as capsule dehiscence. Thick lines delimit three subgenera (Grammica, Cuscuta, and Monogyna). Within subgenus Grammica, thin lines encircle sections Eugrammica and Cleistogrammica, members of which are characterized by dehiscent and indehiscent capsules, respectively. Putative relationships among sections and subsections according to Yuncker (1932) are depicted by arrows and the five-letter name abbreviations are indicated for each subsection.

Any attempt to resolve longstanding controversies and to nurture a greater understanding of the numerous changes that have affected Cuscuta, is in large part dependent on addressing these problems within Cuscuta subgenus Grammica, a group that epitomizes the complexity of the genus as a whole. Grammica is by far the largest group of Cuscuta, accounting for approximately three-fourths of the species diversity of the genus (130–135 spp.). While few members of this subgenus are widespread, the vast majority of species occur only in the Americas, with Mexico and adjacent regions as a center of diversity (Yuncker, 1932 ). Following the most recent and comprehensive monograph of the genus (Yuncker, 1932 ), subgenus Grammica is divided into two sections, Cleistogrammica and Eugrammica, based on indehiscent or dehiscent capsules, respectively. Yuncker (1932) further subdivided each of these sections into 12 subsections, based on a combination of characters, and proposed a scheme of phylogenetic relationships among them (Fig. 1). Characters used to distinguish the various subsections include the number, size, texture, and shape of flower parts, pedicel length, type and density of inflorescences, presence and shape of infrastaminal scales, ovary/capsule shape, embryo shape, and others. Unfortunately, many of these features are quantitative rather than qualitative, difficult to discern, or subjective and open to interpretation. The general difficulty with these characters, combined with the large number of species in subgenus Grammica, explains in part why an updated revision of Cuscuta has not appeared in more than 70 years since Yuncker's (1932) seminal work on this subject.

Given the size and complexity of Cuscuta subgenus Grammica, the present study takes a "bottom-up" phylogenetic approach and focuses primarily on the first three of the aforementioned goals, i.e., the circumscription of major lineages within the subgenus, relationships among them, as well as an assessment of the monophyly of this group overall. To answer these questions, we generated a new molecular data set consisting of plastid and nuclear noncoding DNA sequences. This study presents the first phylogenetic analysis that includes members of all sections and subsections of Cuscuta subgenus Grammica.

MATERIALS AND METHODS

Taxon sampling
A total of 265 accessions representing 99 species were used in this study. Species names, sources, voucher information, and corresponding DNA extraction numbers are provided in the Appendix. Yunckers's (1932) intrageneric classification is followed here because it is the most widely used one and represents the only comprehensive work at the generic level. The 96 ingroup taxa, on which our analyses are primarily focused, include members of two traditionally recognized sections within Cuscuta subgenus Grammica and all 24 recognized subsections. Due to the difficulties in distinguishing many of these species morphologically, effort was made to sample multiple accessions of each. Approximately two-thirds of the species examined here are represented by more than one individual. Special attention was paid to morphologically variable species containing more than one subspecies/variety (e.g., C. salina, C. indecora, C. umbellata) and to those with wide geographic range (e.g., C. campestris, C. gronovii, C. californica). These species were represented by upward of 7–10 individuals from across their respective morphological/geographical range. The remaining one-third of the species is represented by a single individual mainly because they are either rare or locally abundant but known only from their type localities or otherwise restricted areas or because they are underrepresented in collections. The latter is an especially significant factor for many South American species. Three species (C. nitida, C. europaea, and C. approximata) from the putative sister subgenus Cuscuta were selected as outgroup taxa.

Molecular techniques
Total genomic DNA from silica-dried or herbarium material was extracted using a modified hexadecyltrimethylammonium bromide (CTAB) technique from Doyle and Doyle (1987) and purified using Wizard minicolumns (Promega, Madison, Wisconsin, USA). The polymerase chain reaction (PCR) was used to obtain the double-stranded DNA fragments of interest. The plastid genome (ptDNA) region containing the trnL intron, 3' trnL exon, and intergenic spacer between this exon and trnF (hereafter called trnL-F) was amplified using the C and F primers described by Taberlet et al. (1991) . The internal transcribed spacer (ITS) region of nuclear ribosomal DNA (nrDNA) containing ITS1, 5.8S, and ITS2 (hereafter called nrITS) was obtained using primers ITS5 and ITS4 described by White et al. (1990) . PCR was carried out in 50 µL volumes with annealing temperatures of 50–55°C. Amplified products were cleaned by polyethylene glycol/NaCl precipitations. Cleaned products were sequenced directly, including both strands to ensure accuracy, using the DYEnamic ET dye terminator sequencing kit (GE Healthcare, Baie-d'Urfé, Quebec, Canada) on an Applied Biosystems model 377 automated DNA sequencer (PE Biosystems, Foster City, California, USA). PCR products for which polymorphism was detected during direct sequencing (mostly ITS) were cloned into the pSTBlue-1 AccepTor vector (EMD Biosciences, San Diego, California, USA), and multiple clones were sequenced. Sequence data were proofed, edited, and contigs assembled using Sequencher v.3.0 (Gene Codes Corp., Ann Arbor, Michigan, USA). Sequences generated in this study are in GenBank (accession numbers EF194288–EF194718 and EF202557–EF202563; Appendix).

Phylogenetic analyses
Sequences were aligned manually using the program Se-Al v.2.0a11 (Rambaut, 2002 ). Although numerous gaps had to be introduced in the alignments, the sequences were readily alignable among the ingroup taxa in both plastid and nuclear matrices. Regions that could not be unambiguously aligned were excluded from subsequent analyses. Gaps in the alignments were treated as missing data. Phylogenetic analyses were conducted using parsimony and Bayesian inference methods.

Parsimony analyses
Heuristic searches and estimates of clade support were conducted for each matrix separately as well as for a combined data set. Nucleotide characters were treated as unordered, and all changes were equally weighted. Searches for most parsimonious (MP) trees were performed using a two-stage strategy with PAUP* version 4.0b10 (Swofford, 2002 ). First, the analyses involved 1000 replicates with stepwise random taxon addition, tree-bisection-reconnection (TBR) branch swapping saving no more than 10 trees per replicate, and MULTREES option off. The second round of analyses was performed on all trees in memory with the same settings except with the MULTREES option on. Both stages were conducted to completion or until 100 000 trees were found. In addition, other searches were conducted using the parsimony "ratchet" analysis (Nixon, 1999 ) as implemented in NONA (Goloboff, 1999 ) with the WinClada interface (Nixon, 2002 ). Ten consecutive tree searches were conducted using 200 iterations per search, one tree held for each iteration, 10% of total characters sampled, and amb-poly= (no swapping on ambiguously supported nodes), but they did not find shorter trees. Relative support for clades was inferred by nonparametric bootstrapping (Felsenstein, 1985 ) as implemented in PAUP* using 500 pseudoreplicates, each with 20 random sequence addition cycles, TBR branch swapping, and MULTREES option off (DeBry and Olmstead, 2000 ). Conflict between data sets was evaluated by visual inspection, looking for the presence of strongly supported yet conflicting topologies from individual matrices.

Bayesian analyses
The general time-reversible (GTR) model (Yang, 1994 ) of DNA substitution, with rate variation among nucleotides following a discrete gamma distribution and assuming a portion of invariant sites (GTR + G + I), was selected as the best-fit by both the hierarchical likelihood ratio test (hLRT) and Akaike information criterion (AIC), as implemented in ModelTest version 3.7 (Posada and Crandall, 1998 ). Bayesian phylogenetic inferences were performed using MrBayes version 3.1.2 (Ronquist and Huelsenbeck, 2003 ) on the combined data set only. Two runs starting from random trees were carried out using the GTR + G + I substitution model. All model parameters were treated as unknown variables with uniform prior probabilities and were estimated as part of the analysis together with tree topologies. Metropolis-coupled Markov chain Monte Carlo algorithm was used with four simultaneous chains, set at two million generations, and sampled every 100 generations. To determine the burn-in cut-off point, we plotted the –ln likelihood scores against generation time for both runs. After discarding all preasymptotic samples, remaining data points were analyzed separately in PAUP* to compute the 50% majority-rule consensus tree. Because no significant differences between the two runs were detected, the reported topologies and posterior probabilities (PP) are based on trees pooled from both independent Bayesian analyses. Only the nodes receiving 0.95 PP were considered statistically significantly supported, given the assumptions of DNA sequence evolution (Rannala and Yang, 1996 ).

Testing of alternative topologies
Alternative topologies, mainly designed to investigate the evolution of characters defining some traditional taxonomic groups, were constructed and their cost in parsimony was assessed using PAUP* (Swofford, 2002 ). To statistically compare resulting alternative phylogenetic hypotheses, we conducted one-tailed Shimodaira–Hasegawa tests (SH tests; Shimodaira and Hasegawa, 1999 ; Goldman et al., 2000 ) using the aforementioned substitution model and likelihood settings. The SH tests were conducted with PAUP* using 1000 replicates and full parameter optimization of the model.

RESULTS

Sequences and alignments
Characteristics of the sequenced regions as well as statistics of MP trees derived from separate and combined analyses are summarized in Table 1.

Aligned sequences of nrITS used here were 717 bp in length with the individual sequences varying between 560 and 600 bp. The length variation was more or less equally distributed throughout the entire region. The nrITS sequences could not be obtained for all investigated individuals/species. However, the unsuccessful amplifications were randomly distributed (i.e., not part of any particular taxonomic group, unlike in the case described for trnL-F) and were probably due to the poor quality of the DNA extracted from older herbarium specimens. For the majority of DNA accessions, the direct sequencing approach yielded results without apparent polymorphism. In some cases, however, it becomes clear that the PCR product contained more than one type of nrITS sequence, and for those the cloning approach was followed. In most of those cases, the "polymorphism" was caused by the presence of fungal DNA resulting either from the natural presence of fungal epi- and endophytes in Cuscuta species or from fungal tissue contamination. Fungal sequences were easily separated from Cuscuta nrITS sequences and were excluded from further analyses. In several cases a genuine polymorphism within Cuscuta nrDNA was detected, caused by point mutations and/or length variants. However, preliminary phylogenetic analyses in all of those cases (results not shown) indicated that these paralogous sequences were most closely related to each other, suggesting relatively recent duplication events or minor DNA polymerase error, and only one, randomly chosen, sequence was used to represent the given individual. The nrITS sequences were relatively easily alignable across all ingroup species, and this whole region was included in the phylogenetic analyses. However, in a way similar to that described for the trnL-F sequences, none of the three a priori chosen outgroup taxa from subgenus Cuscuta could be aligned with ingroup species for the more variable ITS1 and ITS2 regions. Only the highly conservative (and least informative) 5.8S was alignable between ingroups and outgroups, and consequently only this region could be used to root the nrITS trees. The same was true for a putative ingroup species, C. appendiculata, which had the most sequence similarity with C. nitida for nrITS region as well.

For phylogenies aimed at resolving species-level relationships, it is of paramount importance to incorporate within-species variability and take into account possible biological phenomena that can confound results (such as lineage sorting, deep coalescence). For these reasons, most of the species in the study were represented by multiple individuals, sampled from geographically distinct areas and encompassing morphological variability. However, the addition of terminal taxa results in a sharp increase of computational burden (Felsenstein, 1978 ). Therefore, to facilitate the phylogenetic analyses, individuals of the same species having both the trnL-F and nrITS sequences identical to each other were grouped into a single operational taxonomic unit (OTU). Following this procedure, the 223 individuals from trnL-F matrix were aggregated into 141 OTUs, 207 from nrITS matrix into 153, and 265 individuals used in the combined data set were aggregated into 161 OTUs (Table 1). No significant heterogeneity in base composition was detected within any of these data matrices across all taxa.

Tree topologies
A number of distinct phylogenetic analyses were conducted using parsimony and Bayesian approaches to explore the distribution of phylogenetic signal in the different matrices. All analyses produced trees of remarkably similar topology although resolution and branch support varied. Tree characteristics from MP searches are summarized in Table 1.

Individual data set analyses
The trnL-F and nrITS matrices produced >100 000 trees, 790 and 1965 steps in length, respectively. Schematic consensus trees from parsimony analyses are presented in Fig. 2. The overview of relationships among the major groups also allows for topological comparison of results between the two data sets (Fig. 2). The detailed trees obtained from separate analyses of the data sets are presented in Appendices S1 and S2 (see Supplemental Data accompanying online version of this article). A total of 15 major clades, labeled A–O, were resolved within Cuscuta subgenus Grammica with nrITS sequences. Fourteen of the same groups, A–N, were also recovered with trnL-F data. However, none of the sequences belonging to clade O, a lineage consisting almost exclusively of South American species from subsections Odoratae, Grandiflorae, and Acutilobae, could be obtained for trnL-F. This plastid region could not be amplified either for several species within clade K (e.g., C. erosa, C. boldinghii), even though the same DNA accessions yielded good PCR products for nrITS. Nevertheless, other members of the K clade were sequenced for trnL-F and were available as placeholders in the phylogenetic analyses. Most of the 15 major clades received moderate (70–85%) to strong bootstrap support (>85%) from both of the individual matrices. However, some groups were found to be weakly supported (<70%) by one of the data sets while receiving moderate to strong support from the other in a mutually complementary fashion. For example, clade N was supported only by 59% BS with nrITS data, but it received 80% BS from trnL-F data. In a complementary fashion, clade C obtained <50% BS for with plastid sequences, yet the same clade was supported with 96% BS with nuclear data. The overall strong support for the circumscription of these 15 major clades stands in contrast to the less-resolved backbone relationships within Cuscuta subgenus Grammica based on separate analyses. The trnL-F phylogeny has only two well-supported backbone relationships, a group consisting of A–C clades (receiving 83% BS) and a group consisting of L–N clades (which received 96% BS). The nrITS consensus tree was somewhat more resolved, having three highly supported backbone nodes (100% BS for a group composed of A–E clades, 100% BS for a group consisting of A–I clades, and 96% BS for a group composed of A–K clades). Given the current taxonomic sampling, the only topological disagreement observed between the plastid and nuclear phylogenies involved clades J and K. With plastid data, these two clades were placed as each other's sister-group, whereas nuclear data resolved them as a successively diverging grade. However, these alternative topologies are weakly supported (<70% BS) in both cases. In addition, this difference represents only a slight topological distortion (a nearest-neighbor interchange) most likely caused by sampling discrepancies between trnL-F and nrITS matrices within the K clade (as described before). Taking all of this into account, we deemed these two matrices congruent and combined them into one data set.

View larger version (22K): [in this window] [in a new window]   Fig. 2. Overview and comparison of strict consensus trees derived from separate trnL-F and nrITS parsimony analyses. Fifteen major groups are labeled A–O, and their parsimony bootstrap support values are indicated above branches. Plastid sequences could not be obtained for members of the O clade. Species relationships within the major clades are not shown (see Appendices S1 and S2 for detailed trees in Supplemental Data accompanying online version of this article). Trees are tentatively rooted using the L–O clades as functional outgroups (see Results for full explanation).

View larger version (33K): [in this window] [in a new window]   Fig. 3. Distal portion of the strict consensus tree derived from the combined trnL-F and nrITS maximum parsimony analysis comprising species from all traditionally recognized Cuscuta subgenus Grammica sections and subsections. Major clades referred to in the text are labeled and their bootstrap support is encircled. Bootstrap values for relationships within major clades as well as for the backbone relationships are indicated above branches (for support 50%). Asterisk indicates branches with Bayesian posterior probability <0.95; all other interior branches have posterior probability 0.95. Capsule indehiscence (INDE) and dehiscence (DE), corresponding to species' circumscription into sections Cleistogrammica and Eugrammica, respectively, are indicated before the species name, along with the subsection to which it traditionally belongs. The five-letter abbreviations of subsection names follow those from Fig. 1. Solid bars depict position of changes in fruit dehiscence optimized on the tree. Open bars indicate two equally parsimonious reconstructions, and their alternatives are indicated with dotted arrows. Numbers in parentheses following species names correspond to DNA accessions (see Appendix).

View larger version (37K): [in this window] [in a new window]   Fig. 4. Proximal portion of the strict consensus tree derived from the combined trnL-F and nrITS maximum parsimony analysis comprising species from all traditionally recognized Cuscuta subgenus Grammica sections and subsections. For further explanations, see Fig. 3.

View larger version (68K): [in this window] [in a new window]   Fig. 5. One of equally parsimonious trees derived from the combined trnL-F and nrITS analysis, including species from all traditionally recognized Cuscuta subgenus Grammica sections and subsections, chosen to illustrate branch lengths. Branch lengths are drawn proportionally to the number of changes. Shaded boxes indicate the primary geographic distribution of clades by regions. Asterisks depict species (or groups of species) that are inferred to have a major disjunction in distribution compared to the other members of their respective clades (indicative of long-distance dispersal events). Numbers in parentheses following species names correspond to DNA accessions (see Appendix).

Neither of the two groups historically defined within subgenus Grammica, sections Cleistogrammica and Eugrammica, were found to be monophyletic. The presence of indehiscent vs. dehiscent capsules, the defining character states for these sections, was inferred to have occurred at least 10 times independently (solid bars; Figs. 3, 4). To segregate all the species into two groups based on the type of dehiscence, multiple well-supported nodes, most of them at 100% BS and 0.95 PP, would have to be collapsed. Not surprisingly, the enforced monophyly of Cleistogrammica and Eugrammica resulted in trees 460 steps longer than the most parsimonious trees and was rejected as a significantly worse solution by the SH test (P < 0.01). Another alternative topology concerning the fruit dehiscence was tested as well. Members of the C. indecora complex (i.e., clade M) all have indehiscent capsules, yet they were found on the MP trees surrounded by clades with predominantly dehiscent fruits. We wanted to determine the cost in parsimony and its significance for the alternative in which the M clade would be in closer association with the clades that have almost exclusively indehiscent capsules (clades A–E). Although much less stringent than the previous one, this constraint nevertheless yielded trees 46 steps longer than the most parsimonious trees and was also rejected as significantly different from the best solution by the SH test (P < 0.01). Even though the two character states clearly originated more than once (and hence their corresponding sections are not monophyletic), a boundary can still be drawn between two groups, one predominantly with indehiscent capsules and the other composed mostly of species with dehiscent capsules. The former group includes clades A–E and the latter clades G–O. Clade F cannot be unequivocally assigned to either of these groups at present due to the uncertainties regarding the capsule character states (open bars; Fig. 3).

Of 24 subsections defined within Grammica by Yuncker (1932 ; Fig. 1), three are monotypic (Cephalanthe, Lobostigmae, and Prismaticae). Only one subsection containing more than one species, subsection Californicae, was found to be a monophyletic group, nested in clade A, although a couple of its morphologically distinct species were not included in present analyses. None of the remaining 20 subsections were found to be monophyletic. In some instances, albeit not forming a monophyletic group, all members of a particular subsection were still found within one of the labeled clades and hence relatively closely related. For example, Arvenses is not monophyletic, but all species classified in this subsection are found only in clade B. Correspondingly, species classified in Subulatae are found only in clade G, species classified in Leptanthae are found only in clade L, etc. However, most of the subsections are divided among different major groups, with their species dispersed throughout the tree, and clearly do not constitute natural assemblages of taxa.

Monophyly and rooting of the subgenus Grammica
As indicated earlier, the sequence differences between ingroup and outgroup taxa were too great to unambiguously assess the primary homology between the two. This was the case for both trnL-F and nrITS sequences except for their highly conservative portions. The preliminary analyses, designed to verify the monophyly of the subgenus Grammica and to explore alternative placements of the root, were conducted using only the regions alignable between the ingroup and outgroup taxa (resulting trees not shown). When plastid data were restricted to include only the trnL and trnF exons, the results strongly supported the monophyly of Grammica (100% BS; excluding only C. appendiculata) and suggested placement of the root on the branch separating A–K clades on one side and L–N clades on the other (Fig. 2). Analyses limited to the more conservative 5.8S rDNA sequences indicated the L–O grade as sister to the reminder of the subgenus Grammica (Fig. 2). The monophyly of subgenus Grammica received 85% BS, a moderate yet significant value given the conservative nature of the region included in this analysis. Finally, the midpoint rooting conducted with ingroup taxa only but encompassing the entire trnL-F region and/or entire nrITS region indicated the same basal split into two major groups, the first including species from clades A–K and the second including members of clades L–O. Taking all of these results in aggregate, we used species from the L-O clades (L-M for trnL-F data alone) as functional outgroups to provide directionality for the inferred underlying phylogenetic network (Figs. 2--5).

DISCUSSION

This study represents the most complete molecular phylogenetic hypothesis for Cuscuta subgenus Grammica yet made. It is based on plastid and nuclear noncoding sequences obtained for an extensive sampling of species from a broad taxonomic and geographic range. The resulting phylogenetic inferences are well resolved and robust, including significant support for some of the higher level relationships along the spine of the tree.

Circumscription of major clades within Cuscuta subgenus Grammica
Combined data analyses resulted in a phylogenetic hypothesis (Figs. 3--5) featuring many resolved and well-supported clades. We circumscribe here 15 of those groups, labeled informally as clades A–O, and discuss them in some detail later. A formal classification will be provided elsewhere, along with the taxonomic revisions for the whole genus. Several factors were taken into account in deciding which clades are well enough defined to warrant labeling. Priority was given to molecular analyses in which labeled clades were both well supported and distinct (Figs. 3--4), as evidenced by their relative branch lengths (Fig. 5). Morphological distinctiveness, geographical distribution, as well as correspondence with previously circumscribed taxa were also considered.

Current knowledge of morphology, ecology, distribution, and other biologically relevant information on these 15 different species groups is uneven. Species from clades that occur primarily in North America are collected more frequently; thus their taxonomy is best understood. Several recent treatments have provided a wealth of information on North American species but have focused either on a particular geographic area (Beliz, 1986 , 1993 ; Austin, 1986 ; Musselman, 1986 ; Prather and Tyrl, 1993 ) or a taxonomic subset (Costea et al., 2005 , 2006a –d ). Treatments covering Central and South America are comparatively fewer in number and older (Hunziker, 1947a , b , 1949 , 1950 ), despite the fact that these regions are the centers of biodiversity for Cuscuta (and in particular subgenus Grammica). By defining a number of previously unsuspected relationships among South American species and pointing out their overall importance for understanding the systematics and biogeography of the genus, our present study provides new impetus to rectify this obvious disparity.

Clade A
This clade corresponds largely to the C. salina-californica complex, as defined by Costea et al., (2006d) and groups together species distributed in North America, west of the Rockies, from the Pacific Northwest to the northern Mexican states. Molecular data revealed four distinct lineages within this clade. The first group contains two species, C. californica and C. occidentalis, both belonging to the subgenus Californicae (Yuncker, 1932 ). In most cases, these two species can be distinguished by a combination of characters (Costea et al., 2006d ), but morphologically intermediate individuals are consistently, if rarely, encountered. Most of the sequences obtained from individuals representing C. californica and C. occidentalis segregated into two separate groups. One group corresponds entirely to C. californica, including C. brachycalyx, an entity described by Yuncker (1932 , 1965 ) as a separate species but treated here as conspecific with C. californica. The other corresponds predominantly to C. occidentalis. However, the intermediate plants included into our analyses grouped also with C. occidentalis. It is not clear if this could be attributed to introgression/hybridization between these sympatric species or to incomplete lineage sorting at both the molecular and morphological level. Subsection Californicae, characterized by a complete reduction of infrastaminal scales, are the only traditionally defined subsection (Yuncker, 1932 ) that appears monophyletic in our results. It should be noted, however, that two morphologically distinct species also classified in Californicae, C. sandwichiana and C. jepsonii, are not included in the present analyses. At least one of them, C. sandwichiana, is believed to belong to the C. pentagona complex (our clade B) based on its branching characteristics (Costea et al., 2006c ). If true, this would render Californicae nonmonophyletic as well. The second and third groups within the A clade belong traditionally to subsection Subinclusae. Here, two species, C. salina and C. suksdorfii, with infrastaminal scales reduced to ridges in one and lateral wings in the other, form a well-supported group. They have C. howelliana, a species with well-developed infrastaminal scales, as their sister, supporting the hypothesis that these scales underwent a gradual reduction in this group (Costea et al., 2006b ). The exact relationship of C. subinclusa to the other groups is not resolved. The fourth group in clade A consists of several individuals belonging to C. decipiens. This Mexican species, traditionally circumscribed in subsection Racemosae, is found as an isolated and well-defined lineage in a sister-group position to the rest of the A clade.

Clade B
Taxonomically, this clade contains all sampled members of subsection Arvenses (C. pentagona- complex) with several additional species traditionally classified into Platycarpae (Yuncker, 1932 ) nested within it. Together, this group is characterized by depressed-globose capsules, with mostly short and subulate styles, and relatively large interstylar apertures (Costea et al., 2006c ). Its monophyly is strongly supported by molecular data (100% BS) in a position as sister group to C. stenolepis, a species from Ecuador originally classified in Indecorae (Yuncker, 1932 ). Albeit morphologically entangled with other species from subsection Arvenses to the point of being treated sometimes as conspecific with C. pentagona (e.g., Beliz, 1986 ), C. campestris was found to be molecularly quite distinct from this group and more closely related to those members of Platycarpae found in the B clade.

Several species of clade B have their distributions spanning multiple continents and represent some of the most frequently encountered and best-known dodders worldwide. For example, the native distributional range of C. australis includes Asia, Australia, and Europe, while that of C. obtusiflora spans the entire western hemisphere. In addition, C. campestris is one of the most successful parasitic weeds, spreading worldwide through contaminated crop seed, especially that of forage legumes (e.g., alfalfa, clovers; Dawson et al., 1994 ). Originally, however, this species is native to North America, where it is second in abundance only to C. gronovii (Yuncker, 1932 ). Even though the B clade is today essentially cosmopolitan (due to C. australis and C. obtusiflora with worldwide distributions), its origin and diversification is still deduced to be primarily North American. This inference is supported by the derived position of taxa with wide geographic distribution, nested as a monophyletic group within a grade of species with largely North American distribution. It is worth noting that the species with the widest geographical distributions and ecological amplitudes (C. australis, C. obtusiflora, C. campestris) also form a monophyletic group, nested within clades containing species with narrow (C. runyonii, C. plattensis, C. harperi) to moderate (C. glabrior, C. pentagona, C. stenolepis) distribution ranges. This implies that their physiological capability to use hundreds of genera as hosts probably arose once, in their common ancestor. This capability thus allowed them to spread over large geographic areas, either naturally or as consequence of anthropogenic influences.

Clade C
The existence and composition of this clade, first identified here, emerges as one of the biggest surprises in this study. Based on substantial morphological differences, the eight currently sampled species belonging to this clade were previously dispersed among five different subsections. The unusual level of heterogeneity is best illustrated by an example. Two species with dehiscent capsules, C. corniculata and C. xanthochortos, are not only nested within the C clade, where all other species have indehiscent capsules, but are also nested within an even larger group, comprising five clades (A–E), all members of which are characterized by indehiscent fruits. In a context of the rooted phylogeny (discussed later), this represents the only case of a reversal from indehiscent to dehiscent fruit in Cuscuta. All the species found in the C clade are distributed exclusively in South America, mainly east of the Andes.

Clade D
All of the species traditionally assigned to subsections Oxycarpae, Lepidanchae, and Cephalanthae (Yuncker, 1932 ) are found only within this clade. Furthermore, according to Yuncker's (1932) phylogenetic scheme, these three subsections are each other's closest relatives (i.e., they are expected to be monophyletic, using modern terminology). Morphologically, this entire group is supported by the presence of an ovary/capsule that thickens apically, forming, in some species in its most pronounced form, a beak-like stylopodium, and also corresponds to a broadly circumscribed subsection Oxycarpae sensu Engelmann (1859) . The monophyly of this entire group is well supported in all molecular analyses and under all optimality criteria, and the long branch leading to this clade further highlights its molecular distinctiveness (Fig. 5). In these respects, clade D represents one of the very few cases in which molecular and traditional classifications are in full accord (compare with Fig. 1). The distinctiveness and support for the D clade, however, is in sharp contrast with very short branch lengths and almost complete lack of resolution within it. It is therefore not clear whether either of the subsections (other than monotypic Cephalanthae) is monophyletic or not. Additional, faster evolving data will be necessary to answer this question. Most species of this clade are found from the Great Plains to the Rockies. In addition, some common species, such as C. gronovii, are distributed throughout North America (Yuncker, 1932 ).

Clade E
Subsection Denticulatae traditionally includes four species (Yuncker, 1932 , 1943 ). Two of those, C. denticulara and C. nevadensis, constitute a distinct and well-supported clade, labeled E in this study. One species, C. veatchii, was not sampled here, but there is strong evidence that it forms a natural group with the former two, based on a few distinctive morphological characters (Costea et al., 2005 ). The most unusual of these is the seed with a "thickened" embryo, where the embryo's radicular end is enlarged in a ball-like structure. This feature is unique among dodder species. Also, these three species are characterized by a distinctively reticular calyx surface. Similarly to the A clade distribution, they occur in North America, west of the Rockies, from the Pacific Northwest to the northern Mexican states. Among them, C. denticulata has the broadest geographical distribution, encompassing this entire region. C. nevadensis is sympatric with C. denticulata, but it has a narrower range and is found in southern California, Nevada, and perhaps Arizona. Cuscuta veatchii is restricted in distribution to Baja California in Mexico and is parapatric with C. denticulata. Morphological differences among these species are subtle, yet discontinuous and consistent (Costea et al., 2005 ), and thus in agreement with the substantial differences in branch lengths observed between sampled species (Fig. 5). Yuncker (1932) also included C. microstyla in Denticulatae, based on its overall similarity to the other species in this subsection. However, only material from the type locality was available to him at the time, and it contained neither capsules nor seed, thus preventing him from observing the defining embryo feature. This Chilean species is found elsewhere on the tree, as a distinct lineage within the O clade, together with the other species with South American distribution.

Clade F
Little is known about the four species found in this clade, the second of three lineages with exclusively South American distribution revealed in this study. Even though Yuncker noted the overall resemblance between C. haughtii and C. partita in his monograph (1932), he nevertheless placed these species into two different sections, based on their fruit dehiscence. The other two species from the F clade, C. longiloba and C. burrellii, both of which were described more recently (Yuncker, 1957 , 1961 ), are known only from very few collections and are even more obscure. Their capsules were not observed, and the state of dehiscence is not clear at present. These uncertainties leave open the question whether this clade is primarily dehiscent, with convergent evolution of indehiscence in one of its species (C. haughtii) or whether it is primarily indehiscent with reversion to dehiscence in C. partita.

Clades G–N
The common theme for these eight clades, discussed here together, is their fruit dehiscence and geographic distribution. Almost all species within these groups, resolved as a grade, feature dehiscent capsules. In a few exceptional cases where capsules are found to be indehiscent, those species were usually nested within the given clade in a derived position (with the exception of clade M). In the context of rooted phylogeny, this indicates a convergent evolution toward an indehiscent character state from a plesiomorphic dehiscent state. Also, all these clades are distributed primarily or exclusively in Mexico and its adjacent regions, comprising Southwest USA bordering Mexico, Central America, and/or West Indes. However, three long-distance dispersals are inferred from within three of these clades (G, H, and L; marked on Fig. 5 with asterisks).

Clade G comprises species traditionally included in subsections Subulatae, Tinctoriae, Lobostigmae, and Platycarpae (Yuncker, 1932 ). It is characterized by relatively large, thick flowers and subulate or thick styles, which become more obvious as the fruit matures. Capsules are primarily dehiscent, and infrastaminal scales are thick and fleshy. While most of its species are distributed in Mexico and Central America, a well-supported subclade, consisting of two species with indehiscent capsules (C. victoriana and C. tasmanica), is found in SW Australia and Tasmania, strongly implying long-distance dispersal.

The four species of the H clade were originally included in three different sections (Yuncker, 1932 ); however, they share some common morphological characteristics. Their calyx lobes are often carinate or with longitudinal protuberances along the midvein, and their capsules are surrounded at the base by the withered corolla. Cuscuta yucatana was described by Yuncker as having an indehiscent fruit, while the remaining species have dehiscent capsules. Three species occur in Mexico and/or the southern USA, whereas C. chinensis is clearly disjunct from the rest of the clade and is found in SE Asia, Australia, and Africa.

Most of the species confined to clades I and J were traditionally included in subsection Americanae (Yuncker, 1932 ). These species have dehiscent capsules and more or less cylindrical flowers and are distributed in Mexico, the West Indies, Central America, and northern South America. Although morphologically very similar, these two clades are well supported by molecular data as distinct and monophyletic.

Clade K, which includes a group of several Mexican and Central American species generally possessing various appendages on their calyx and/or corolla lobes, corresponds closely to section Ceratophorae of Yuncker (1932) . As already indicated, the relative positions of this clade and the J clade represent the only point of topological conflict between the plastid and nuclear data sets, albeit weakly supported in both cases. With trnL-F sequences, these two clades were found as each other's sisters (69% BS; Fig. 2), yet were resolved as diverging consecutively, with the nrITS data (66% BS). Because the trnL-F sequences could not be obtained for several accessions of the K clade (compare Appendices S1 and S2; see Supplemental Data accompanying online version of this article), this conflict can be accounted for, in large part, as an artifact derived from sampling differences between the two data sets.

Clade L includes mostly species circumscribed by Yuncker (1932) in subsections Umbellatae and Leptanthae plus a few species that were traditionally classified elsewhere (subsections Odontolepisae and Acutae). Altogether, these species are characterized by loose, umbellate inflorescences and flowers with acute calyx and corolla lobes. Most species have dehiscent capsules and occur in Mexico and the SW USA. However, C. acuta, endemic to the Galapagos Islands, has indehiscent capsules. Additionally, C. hyalina, with its disjunct populations found in India and W South Africa, is also nested in this clade.

Clade M corresponds very closely to subsection Indecorae, a group of species characterized primarily by fleshy, papillose or glandular flowers. According to Yunker (1932, 1965), Indecorae comprise four species, two North American (C. coryli and C. warnerii), one South American (C. stenolepis), and one widely spread species spanning the Americas (C. indecora with its varieties). An additional species (C. attenuata) was recognized within Indecorae by Prather and Tyrl (1993) . Albeit morphologically very similar to C. indecora, the two species were treated as separate, based on failure of artificial crosses to yield fruit/seed and their apparent reproductive isolation (Prather and Tyrl, 1993 ). However, emphasizing the morphological similarities with C. indecora and taking into account the degree of morphological differences found within this species, C. attenuata was recently reduced to a variety of C. indecora (Costea et al., 2006a ). All taxa traditionally classified into subsection Indecorae (Yuncker, 1932 ), except for C. stenolepis, are found in the M clade, with high support. Flowers of C. stenolepis are not fleshy, and a combination of other characters was used to place it with Indecorae. As previously noted, this species is a member of clade B, where it forms an isolated lineage positioned as sister group to the rest of that clade.

While the composition of clade M is not surprising, its phylogenetic placement is, both from morphological and geographical points of view. All the species grouped here have indehiscent capsules, yet they are nested deeply within several distinct clades all of which are primarily or exclusively with dehiscent capsules. Bringing the M clade in a closer association with other groups characterized by indehiscent capsules (clades A–E) requires dissolving a number of internal branches, some of which are highly supported. This scenario was formally tested and rejected as significantly worse topological solution. A similar, albeit less clear-cut, argument can be made from the biogeographical standpoint. Most taxa in the M clade, as currently delimited, are North American in distribution. Cuscuta warnerii is narrowly limited to the Four Corners region of the USA (i.e., southern Utah and Colorado and northern Arizona and New Mexico). Cuscuta attenuata is also limited in its distribution and is found only in Kansas, Oklahoma, and northern Texas. Cuscuta coryli is much more widespread in distribution, but still confined to North America, occurring throughout most of the USA and southern Canada, east of the Rockies. Yet, all the other groups in the immediate vicinity of the M clade are primarily Mexican in distribution. It has to be pointed out, however, that C. indecora, also a member of the M clade, is distributed throughout the Americas, from southern Canada to Argentina and Chile, and is rendering the geographical distinctiveness of this group with its phylogenetic surroundings more blurry compared to the morphological.

The morphological and geographical uniqueness of the M clade is equaled by its molecular distinctiveness. As indicated in the phylogram (Fig. 5), this group has one of the longest branches on the tree, due primarily to more divergent nrITS sequences. Within this clade, both C. coryli and C. attenuata are nested within C. indecora. Cuscuta coryli seems to have achieved the reciprocal monophyly, based on samples from several individuals, and is distinct molecularly (as well as morphologically) from C. indecora. Cuscuta attenuata is both morphologically and molecularly more similar to C. indecora, in accordance with keeping these two entities conspecific (Costea et al., 2006a ). Cuscuta warnerii, a strikingly distinct dodder with calyx lobes apically extended into horn-like projections, forms a separate lineage, sister to C. indecora s.l.

Clade N represents another segregate from the subsection Umbellatae; hence it is not surprising that it is morphologically very similar to clade L. However, it differs from the L clade by the tendency to have no stems at flowering/fructification time. Consequently, the inflorescences appear as tufts emerging directly from the stems of their hosts. Most species found here are poorly known. For example, C. serruloba, C. mcvaughii, and C. aristeguietae have been known only from their type collections, and C. columbiana is extremely rare. All species have dehiscent capsules except for C. aristeguietae, which has indehiscent fruits. This species is also one of the very few that Yuncker (1932) did not place in any of the 24 subsections he described in the subgenus Grammica.

Clade O
This group is almost exclusively of South American distribution and represents the largest and most diverse clade in subgenus Grammica (and perhaps in the entire genus). Its species belong, for the most part, to three relatively closely related subsections, Odoratae, Grandiflorae, and Acutilobae (Fig. 1; Yuncker, 1932 ). Based on good support and sequence divergence, molecular data also reveal three major lineages within the O clade. Those groups, however, have a taxonomic make-up differing from Yuncker's subsections. The first subclade consists of some Odoratae and all of the sampled species belonging to Acutilobae (95% BS; 0.95 PP). With large, thick flowers, cylindric or subulate styles longer than the ovary, and often large, convoluted stigmas, its members superficially resembles species from the G clade. Its species are distributed primarily along the Andes (Ecuador, Peru, Bolivia, Chile). The second subclade includes taxonomically the rest of Odoratae and all of the sampled species belonging to subsection Grandiflorae (100% BS; 0.95 PP). It is characterized by flowers with rotate or globose corollas, anthers often curved toward the ovary, styles absent or shorter than the ovary/capsule, and large, convoluted stigmas. Among the species of this subclade, C. cristata can be interpreted as a case of incomplete evolution of indehiscence. Its pericarp does not split along a definite line in the dehiscence area. Instead, because it is very thin, it breaks when mechanical pressure is applied. Although both Yuncker (1932 , p. 188) and Hunziker (1949, p. 1159) considered the capsule of C. cristata indehiscent, Hunziker labeled it also as rather "pseudoindehiscent." He even described a variety of this species, C. cristata var. chacoensis (not sampled in our study) with almost perfectly dehiscent capsules. The majority of the species confined to this group occur east of the Andes (Argentina, Uruguay, Brazil), and a few are found along the Andes (Colombia to Chile). However, one species represents another striking case of relatively recent long-distance dispersal. Cuscuta kilimanjari, found in eastern Africa, is nested within this otherwise exclusively South American clade. Finally, C. microstyla represents the third distinct lineage within the O clade. Compared to the previous two subclades, this Chilean species has smaller flowers and clearly indehiscent fruits. The relationships among the three major lineages of clade O are not resolved.

Monophyly of Cuscuta subgenus Grammica
In addition to the primary emphasis on the circumscription and relationships of major groups within Grammica, our data are also pertinent to the question of monophyly of this subgenus. According to analyses including portions of sequences alignable between the ingroup and outgroup, subgenus Grammica is monophyletic and supported by high to moderate bootstrap values (100% BS with trnL-F; 85% BS with nrITS). The single origin for this subgenus is further supported by the presence of two distinct styles with globose, nonlinear stigmas. However, there is one notable exception to this result, given our current species sampling. Cuscuta appendiculata features globose stigmas and indehiscent capsules. Both of these characteristics are typical of subgenus Grammica to which this South African species was traditionally assigned. Yet C. appendiculata has both its plastid and nuclear sequences more similar to those of the outgroups, and it is found to group with other South African species currently classified in subgenus Cuscuta with high bootstrap supports (95% BS with trnL-F; 100% with nrITS; resulting trees not shown).

Our results, to a limited extent, also bear on the issue of monophyly of Cuscuta subgenus Cuscuta. They are consistent with the notion that this subgenus is not monophyletic (McNeal, 2005 ) and consists of two quite distinct groups, one South African in distribution, corresponding in large part to subsection Africanae, and the other containing the rest of the species belonging traditionally to this subgenus (García and Martín, in press ).

Multiple origins of fruit indehiscence in Cuscuta
Yuncker (1932) considered species with the indehiscent capsules to be more "primitive" as judged by their central-basal position in his phylogenetic scheme for the genus (Fig. 1). Furthermore, the evolution "from capsules which remain closed to those which are regularly and definitely circumscissile when mature" was explicitly listed as one of the major evolutionary tendencies within Cuscuta (Yuncker, 1932 , p. 115). Our phylogenetic results disagree with both the taxonomic implications as well as the directionality for proposed scenario of evolution of this character.

Regardless of the position of the root, the switch from dehiscence to indehiscence was inferred to have occurred at least 10 times independently within subgenus Grammica (solid bars; Figs. 3--4). An attempt to group all the species according to their fruit characteristics was very costly in terms of parsimony steps (460 additional steps) and was also rejected by the SH test. Even a much less stringent attempt to bring together some of the clades with indehiscent capsules for reasons other than morphology (e.g., for biogeographical reasons) was also rejected. This refutes the taxonomic hypothesis of monophyly of two sections, Cleistogrammica and Eugrammica. In addition, the relative ease by which this character shifts its states, as indicated by the inferred number of changes, suggests relatively simple genetics for this trait, possibly regulated by only one or only a few genes. In some other plant groups featuring similar polymorphisms regarding fruit dehiscence (e.g., Amaranthus, Amaranthaceae), the dehiscent/indehiscent character states were also found to shift easily, and a similarly simple genetic determinism was suggested (Costea et al., 2001 ). As Cronquist (1988) proposed, the dehiscence mechanism may be easily lost because the genetic causes of indehiscence are not selected against and the advantages of indehiscent vs. dehiscent capsules, if any, are obscure.

Because the unrooted phylogenetic networks indicate the location of the changes but not their directions, we used two approaches to assess the polarity of fruit (in)dehiscence. The first line of evidence comes from comparison with the states found in outgroups (the relative apomorphy rule; Wiley et al., 1991 ). While the ingroup species have both indehiscent and dehiscent character states, almost all the outgroups from subgenera Cuscuta and Monogyna feature dehiscent capsules. Hence, the dehiscence is deduced to be a primitive character for Grammica. The second line of evidence comes from the position of the root. The root for Grammica could not be determined unambiguously due to the alignment problems between the ingroup and outgroup taxa. However, the outgroup rooting, using limited conservative portions of alignment, as well as the mid-root rooting approach, indicated that the first split within Grammica occurred among lineages with predominantly dehiscent capsules (clades J–O; Fig. 4). Both of these arguments support the conclusion that dehiscence is a plesiomorphic condition and that the apomorphic indehiscence evolved several times independently. This is contrary to Yuncker's (1932) hypotheses developed for character evolution in Cuscuta, but it supports views put forward for the fruit evolution in flowering plants generally (Cronquist, 1988 ).

Biogeographical implications
While Yunker's (1932) monograph of Cuscuta did take the distribution of species into account in some cases, his overall taxonomic circumscription and proposed phylogenetic arrangement was based almost entirely on morphology with comparatively little attention given to biogeography. Consequently, in subgenus Grammica alone, a minimum of 14 long-distance dispersals has to be implicitly assumed to accommodate this traditional classification. The phylogenetic relationships inferred here portray a significantly different biogeographic scenario. Diversification through vicariance, as opposed to long-distance dispersal, clearly emerges as the more dominant pattern for the species of Grammica. Of its 15 major lineages, four are primarily or entirely of North American distribution (A, B, D, and E), three are South American in distribution (C, F, and O), and the rest (eight major clades in total: G–N) are found in Mexico and adjacent regions. All these inferences are evident from the unrooted phylogenetic network within this subgenus and are not dependent on the exact position of its root. On the other hand, the most closely related outgroup taxon to Grammica, subgenus Cuscuta, and in particular its section Africanae, all occur in the Old World and are most abundant in the Mediterranean region and southern Africa. Hence, there is little doubt that the initial jump had to occur from the Old World, most likely Africa, to the New World. Given the rooting proposed here, the first major split and subsequent diversification of Cuscuta species in the New World most likely occurred between South America (clade O) on one side and Mexico on the other. Mexico (with its adjacent regions), where more than half of Grammica species occur, is clearly the center of biodiversity for Cuscuta. This region also represents a "staging ground" for the next major diversification. Namely, from within this Mexican grade, two clades (C and F) split off, and diversity back in South America and four clades (A, B, D, and E) spread throughout North America.

In addition to the clear preponderance of vicariant events that explain species distribution in Cuscuta, several striking cases of long-distance dispersal are also evident from the inferred phylogeny as well (indicated with asterisks in Fig. 5). For example, C. kilimanjari, an eastern African species, is nested deeply within an otherwise exclusively South American group (clade O). Cuscuta hyalina has a disjunct distribution, with populations found in southern Africa and India, yet this species is nested within the L clade, all other members of which are found in Mexico and adjacent regions. Similar arguments for long-distance dispersal can be made for C. chinensis, distributed in eastern Asia, and C. victoriana and C. tasmanica, both found in Australia. In both of these cases, their closest relatives, grouped in clades H and G, respectively, are of Mexican distribution. Finally, the most cosmopolitan of all dodders today is clade B, with several of its species spreading over multiple continents. However, its ancestral distribution is inferred to be North American, followed by one long-distance dispersal from within it. Nevertheless, all these cases taken together still account for less than half of long-distance dispersals implied by Yuncker's classification (1932) .

Implications for plastid DNA evolution in Cuscuta
Previous analyses of plastid genomes of Cuscuta species (e.g., Bömmer et al., 1993 ; Haberhausen and Zetsche, 1994 ; Stefanovic et al., 2002 ) have pointed out the unexpectedly conservative nature of Cusucta ptDNA evolution, especially when compared to those of its close nonparasitic relatives (Stefanovic and Olmstead, 2005 ). The profound morphological and physiological modifications that the ancestors of this genus experienced during the transition from an autotrophic to parasitic mode of life are therefore contrasted with their relatively unaltered plastid genomes (Stefanovic and Olmstead, 2005 ). For example, the trnL-F region included in this study was amplified with the same set of primers used for nonparasitic taxa across flowering plants. However, this plastid region was unamplifiable for some DNA accessions. The simple explanation of poor DNA quality could be excluded as unlikely in this case because the very same extractions were used consistently with success to amplify the nrITS region. Furthermore, multiple attempts to amplify smaller fragments using different combinations of internal and/or alternative primers were also unsuccessful. Taken collectively, these data indicate that the trnL-F region either experienced a significant acceleration in substitution and thus attained sequence divergence at priming sites used in PCR, or it was altogether lost from the plastid genome of these species. The phylogenetic analyses, based on nrITS data, revealed that the majority of these species belong to a single group, the O clade (Figs. 2, 4). While the negative PCR results (i.e., the lack of amplifications) are hard to explain unequivocally, in this case, they are consistent with results obtained by van der Kooij et al. (2000) . These authors showed the presence of different degrees of degradation in photosynthetic apparatus across Cuscuta. The variability of plastid functionality was documented using a combination of southern, northern, and western hybridizations on six Cuscuta species (five of which were from Grammica). Their results suggested the absence of the rbcL gene itself in C. odorata as well as the absence of transcripts and proteins in C. odorata and C. grandiflora (van der Kooij et al., 2000 ). It is clear from our phylogenetic results that both of these factors affected species belonging to the O clade, the same group of South American species for which trnL-F could not be amplified. In conjunction with newly inferred phylogenetic relationships, the simultaneous absence of both protein-coding genes and/or their products (rbcL, Rubisco large subunit) as well as noncoding regions (trnL-F), indicates a putatively genome-wide phenomenon for a whole clade rather than a localized exception for a particular species and/or particular plastid region. Based on these two lines of evidence, we hypothesize that most, if not all, species of the O clade will have much more significantly altered plastid genomes in comparison to what has been documented up to this point (for a review, see Stefanovi and Olmstead, 2005 ), with many plastid genes and regions absent or divergent to the point of being undetectable by methods such as PCR and hybridization.

Another, smaller group of species for which the trnL-F sequences could not be obtained was also detected. Both of these species, C. erosa and C. boldinghii, belong traditionally in section Ceratophorae, and are found nested within the K clade, with several well-supported nodes leading to them (Fig. 4). Similar to the situation described for the O clade, the ntITS sequences were easily amplifiable from the same DNA accessions. In this case, however, the independent data regarding the presence or absence of other plastid regions are not currently available. Hence, it remains to be seen if this represents a second lineage of species with highly altered plastid genomes or a simple case of PCR amplification difficulties due to primer mismatches or poor quality DNA.

View this table: [in this window] [in a new window]   Appendix. Taxa, DNA accession numbers, sources of plant material from which DNA was extracted, and GenBank accession numbers for sequences used in this study. Infrageneric classification follows Yuncker (1932) . Numbers in square brackets indicate sampled and estimated numbers of species belonging to corresponding subsections, respectively.

¹ The authors warmly thank A. Colwell, T. Van Devender, T. Deroin, I. García, and D. Tank, as well as the curators/directors of AAU, ALTA, ARIZ, ASU, B, BAB, BOL, BRIT, CANB, CEN, CHR, CIIDIR, CIMI, CTES, DAO, F, GH, H, HUFU, IAC, IND, J, JEPS, LL, LP, LPB, LPS, K, MEL, MERL, MEXU, MICH, NMC, NY, OKLA, OXF, PACA, PRE, QCNE, QFA, P, RB, RSA, SAM, S, SD, SGO, SI, SPF, TEX, TRT, UA, UB, UBC, UCR, UCT, UNB, UNM, UPRRP, UPS, US, USAS, WTU, and XAL for supplying plant material; and L. Goertzen, D. Nickrent, and an anonymous reviewer for critical comments on the manuscript. This work was supported by NSERC of Canada grants to S.S. and M.C.

⁴ Author for correspondence (e-mail: sstefano{at}utm.utoronto.ca )

LITERATURE CITED

Angiosperm Phylogeny Group II (APG II). 2003. An update of the Angiosperm Phylogeny Group classification for the orders and families of flowering plants: APG II. Botanical Journal of the Linnean Society 141: 399-436..[CrossRef][Web of Science]

Austin D. F.. 1986. Cuscutaceae. In T. Barkley, [ed.], Flora of the Great Plains, 661-666. University of Kansas Press, Lawrence, Kansas, USA..

Austin D. F.. 1998. Parallel and convergent evolution in the Convolvulaceae. In P. Mathews, [ed.], Biodiversity and taxonomy of tropical flowering plants, 201-234. Mentor Books, Culicut, India..

Baillon H.. 1891. Convolvulacées. In Histoire des plantes, 305–331. Librairie Hachette et Cie, Paris, France..

Beliz T.. 1986. A revision of Cuscuta section Cleistogrammica using phenetic and cladistic analyses with a comparison of reproductive mechanisms and host preferences in species from California, Mexico, and Central America. Ph.D. dissertation, University of California, Berkeley, California, USA..

Beliz T.. 1993. Cuscuta. In J. Hickman, [ed.], The Jepson manual: higher plants of California, 538-539. University of California Press, Berkeley, California, USA..

Bentham G. Hooker J. D.. 1873. Genera plantarum. 2: 865-881..

Bömmer D. Haberhausen G. Zetsche K.. 1993. A large deletion in the plastid DNA of the holoparasitic flowering plant Cuscuta reflexa concerning two ribosomal proteins (rpl2, rpl23), one transfer RNA (trnI) and an ORF 2280 homologue. Current Genetics 24: 171-176..[CrossRef][Web of Science][Medline]

Choisy J. D.. 1845. Convolvulaceae. In A. P. De Candolle, A. L. P. P. De Candolle, [eds.], Prodromus Systematis Naturalis Regni Vegetabilis, vol. 9, 323-465. Sumptibus Sociorum Treuttel et Würtz, Paris, France..

Costea M. Nesom G. L. Stefanovi S.. 2006a. Taxonomy of Cuscuta gronovii and Cuscuta umbrosa (Convolvulaceae). Sida 22: 197-207..

Costea M. Nesom G. L. Stefanovi S.. 2006b. Taxonomy of the Cuscuta indecora (Convolvulaceae) complex in North America. Sida 22: 209-225..

Costea M. Nesom G. L. Stefanovi S.. 2006c. Taxonomy of the Cuscuta pentagona complex (Convolvulaceae) in North America. Sida 22: 151-175..

Costea M. Nesom G. L. Stefanovi S.. 2006d. Taxonomy of the Cuscuta salina-californica complex (Convolvulaceae). Sida 22: 176-195..

Costea M. Nesom G. L. Tardif F. J.. 2005. Taxonomic status of Cuscuta nevadensis and C. veatchii (Convolvulaceae). Brittonia 57: 264-272..[CrossRef][Web of Science]

Costea M. Sanders A. Waines G.. 2001. Structure of the pericarp in some Amaranthus (Amaranthaceae) species—taxonomic significance. Aliso 20: 51-60..

Costea M. Tardif F. J.. 2006. Biology of Canadian weeds. Cuscuta campestris Yuncker, C. gronovii Willd. ex Schult., C. umbrosa Beyr. ex Hook., C. epithymum (L.) L. and C. epilinum Weihe. Canadian Journal of Plant Sciences 86: 293-316..

Cronquist A.. 1988. The evolution and classification of flowering plants. New York Botanical Garden Press, Bronx, New York, USA..

Dawson J. H. Musselman L. J. Wolswinkel P. Dörr I.. 1994. Biology and control of Cuscuta. Reviews of Weed Science 6: 265-317..

DeBry R. W. Olmstead R. G.. 2000. A simulation study of reduced tree-search effort in bootstrap resampling analysis. Systematic Biology 49: 171-179..[CrossRef][Web of Science][Medline]

Doyle J. J. Doyle J. L.. 1987. A rapid DNA isolation procedure for small quantities of fresh leaf tissue. Phytochemical Bulletin 19: 11-15..

Dumortier B.-C.. 1829. Analyse des plantes. Tournay, Paris, France..

Engelmann G.. 1859. Systematic arrangement of the species of the genus Cuscuta with critical remarks on old species and descriptions of new ones. Transactions of the Academy of Science of St. Louis 1: 453-523..

Felsenstein J.. 1978. The number of evolutionary trees. Systematic Zoology 27: 27-33..[Abstract/Free Full Text]

Felsenstein J.. 1985. Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39: 783-791..[CrossRef][Web of Science]

Futuyma D. J.. 2004. The fruit of the tree of life: insights into evolution and ecology. In J. Cracraft, M. J. Donoghue, [eds.], Assembling the tree of life, 25-39. Oxford University Press, Oxford, UK..

García M. A. Martín M. P.. In press. Phylogeny of Cuscuta subgenus Cuscuta (Convolvulaceae) based on nrDNA ITS and chloroplast trnL intron sequences. Systematic Botany..

Goldman N. Anderson J. P. Rodrigo A. G.. 2000. Likelihood-based tests of topologies in phylogenetics. Systematic Biology 49: 652-670..[Abstract/Free Full Text]

Goloboff P.. 1999. NONA, version 2. Published by the author, Tucumán, Argentina. Website http://www.cladistics.com/ [Accessed 20 June 2006]..

Haberhausen G. Zetsche K.. 1994. Functional loss of all ndh genes in an otherwise relatively unaltered plastid genome of the holoparasitic flowering plant Cuscuta reflexa. Plant Molecular Biology 24: 217-222..[CrossRef][Web of Science][Medline]

Hallier H.. 1893. Versuch einer natürlichen Gleiderung der Convolvulaceen auf morphologischer und anatomischer Grundlage. Botanische Jahrbücher für Systematik, Pflanzengeschichte und Pflanzgeographie 16: 453-591..

Hibberd J. M. Bungard R. A. Press M. C. Jeschke W. D. Scholes J. D. Quick W. P.. 1998. Localization of photosynthetic metabolism in the parasitic angiosperm Cuscuta reflexa. Planta 205: 506-513..[CrossRef][Web of Science]

Hunziker A. T.. 1947a. Sinopsis de las especies argentino-uruguayas del género Cuscuta. Revista Agronómica del Noroeste Argentino (Tucumán) 14: 123-147..

Hunziker A. T.. 1947b. Tres nuevas especies sudamericanas de Cuscuta. Darwiniana 7: 323-330..

Hunziker A. T.. 1949. Las especies de Cuscuta (Convolvulaceae) de Argentina y Uruguay. Revista de la Facultad de Ciencias Exactas, Físicas y Naturales 12: 1101-1202..

Hunziker A. T.. 1950. Las especies de Cuscuta (Convolvulaceae) de Argentina y Uruguay. Revista de la Facultad de Ciencias Exactas, Físicas y Naturales 13: 178-251..

Kuijt J.. 1969. The biology of parasitic flowering plants. University of California Press, Berkeley, California, USA..

Mabberley D. J.. 1997. The plant-book. Cambridge University Press, Cambridge, UK..

McNeal J. R.. 2005. Systematics and plastid genome evolution in the parasitic plant genus Cuscuta (dodder). Ph.D. dissertation, Pennsylvania State University, University Park, Pennsylvania, USA..

Melchior H.. 1964. Engler's Syllabus der Pflanzenfamilien, vol. 12, 427-429..

Mower J. P. Stefanovi S. Young G. J. Palmer J. D.. 2004. Gene transfer from parasitic to host plants. Nature 432: 165-166..[Medline]

Musselman J. L.. 1986. The genus Cuscuta in Virginia. Castanea 51: 188-196..[Web of Science]

Nickrent D. L.. 2002. Orígenes filogenéticos de las plantas parásitas. In J. A. López-Sáez, P. Catalán, L. Sáez, [eds.], Plantas parásitas de la peninsula Ibérica e Islas Baleares, 29-56. Mundi-Prensa Libros, Madrid, Spain..

Nickrent D. L. Duff R. J. Colwell A. E. Wolfe A. D. Young N. D. Steiner K. E. dePamphilis C. W.. 1998. Molecular phylogenetic and evolutionary studies of parasitic plants. In D. E. Soltis, P. S. Soltis, J. J. Doyle, [eds.], Molecular systematics of plants II: DNA sequencing, 211-241. Kluwer, Boston, Massachusetts, USA..

Nixon K.. 1999. The parsimony ratchet, a new method for rapid parsimony analysis. Cladistics 15: 407-414..[CrossRef][Web of Science]

Nixon K.. 2002. WinClada, version 1.00.08. Published by the author, Ithaca, New York, USA. Website http://www.cladistics.com/ [Accessed 20 June 2006]..

Peter A.. 1891. Convolvulaceae. In A. Engler, K. Prantl, [eds.], Die Natürlichen Pflanzenfamilien, 1-40. Engelmann, Leipzig, Germany..

Peter A.. 1897. Convolvulaceae. In A. Engler, K. Prantl, [eds.], Die Natürlichen Pflanzenfamilien, 375-377 (Appendix). Engelmann, Leipzig, Germany..

Posada D. Crandall K. A.. 1998. ModelTest: testing the model of DNA substitution, version 3.7. Bioinformatics 14: 817-818..[Abstract/Free Full Text]

Prather L. A. Tyrl R. J.. 1993. The biology of Cuscuta attenuata Waterfall (Cuscutaceae). Proceedings of the Oklahoma Academy of Science 73: 7-13..

Rambaut A.. 2002. Se-Al sequence alignment editor, version 2.0a11. University of Oxford, Oxford, UK..

Rannala B. Yang Z.. 1996. Probability distribution of molecular evolutionary trees: a new method of phylogenetic inference. Journal of Molecular Evolution 43: 304-311..[Web of Science][Medline]

Revill M. J. V. Stanley S. Hibberd J. M.. 2005. Plastid genome structure and loss of photosynthetic ability in the parasitic genus Cuscuta. Journal of Experimental Botany 56: 2477-2486..[Abstract/Free Full Text]

Roberty G.. 1952. Genera Convolvulacearum. Candollea 14: 11-60..[Medline]

Roberty G.. 1964. Les genres des Convolvulacées (esquisse). Boissiera 10: 129-156..

Ronquist F. Huelsenbeck J. P.. 2003. MrBayes 3: Bayesian phylogenetic inference under mixed models, version 3.1.2. Bioinformatics 19: 1572-1574..[Abstract/Free Full Text]

Shimodaira H. Hasegawa M.. 1999. Multiple comparisons of log-likelihoods with applications to phylogenetic inferrence. Molecular Biology and Evolution 16: 1114-1116..[Web of Science]

Stefanovi S. Krueger L. Olmstead R. G.. 2002. Monophyly of the Convolvulaceae and circumscription of their major lineages based on DNA sequences of multiple chloroplast loci. American Journal of Botany 89: 1510-1522..[Abstract/Free Full Text]

Stefanovi S. Olmstead R. G.. 2004. Testing the phylogenetic position of a parasitic plant (Cuscuta, Convolvulaceae, Asteridae): Bayesian inference and the parametric bootstrap on data drawn from three genomes. Systematic Biology 53: 384-399..[CrossRef][Web of Science][Medline]

Stefanovi S. Olmstead R. G.. 2005. Down the slippery slope: plastid genome evolution in Convolvulaceae. Journal of Molecular Evolution 61: 292-305..[CrossRef][Web of Science][Medline]

Swofford D. L.. 2002. PAUP*: phylogenetic analysis using parsimony (* and other methods), version 4.0b2a. Sinauer, Sunderland, Massachusetts, USA..

Taberlet P. Gielly L. Pautou G. Bouvet J.. 1991. Universal primers for amplification of the non-coding regions of chloroplast DNA. Plant Molecular Biology 17: 1105-1109..[CrossRef][Web of Science][Medline]

Takhtajan A.. 1997. Diversity and classification of flowering plants. Columbia University Press, New York, New York, USA..

van der Kooij T. A. W. Krause K. Dörr I. Krupinska K.. 2000. Molecular, functional and ultrastructural characterization of plastids from six species of the parasitic flowering plant genus Cuscuta. Planta 210: 701-707..[CrossRef][Web of Science][Medline]

White T. J. Bruns T. Lee S. Taylor J.. 1990. Amplification and direct sequencing of fungal ribosomal RNA genes for phylogenetics. In M. Innis, D. Gelfand, J. Sninsky, T. White, [eds.], PCR protocols: a guide to methods and applications, 315-322. Academic Press, New York, New York, USA..

Wiley E. O. Siegel-Causey D. Brooks D. R. Funk V. A.. 1991. The compleat cladist: a primer of phylogenetic procedures. University of Kansas Museum of Natural History Press, Lawrence, Kansas, USA..

Yang Z.. 1994. Estimating the pattern of nucleotide substitution. Journal of Molecular Evolution 39: 105-111..[Web of Science][Medline]

Yuncker T. G.. 1932. The genus Cuscuta. Memoirs of the Torrey Botanical Club 18: 113-331..

Yuncker T. G.. 1943. Nomenclatural changes in the genus Cuscuta and notes on some American species. Bulletin of the Torrey Botanical Club 70: 61-67..[CrossRef]

Yuncker T. G.. 1957. The Machris Brazilian expedition. Botany: a new dodder from Goias, Cuscuta burrellii. Los Angeles County Museum Contributions in Science 3: 1-2..

Yuncker T. G.. 1961. A new species of Cuscuta from Paraguay. Svensk Botanisk Tidskrift Utgifven af Svenska Botaniska Foreningen (Stockholm) 55: 229-230..

Yuncker T. G.. 1965. Cuscuta. In W. R. Buck, [ed.], North American Flora, series II, vol. 4, 1-51. New York Botanical Garden Press, Bronx, New York, USA..

CiteULike    Complore    Connotea    Del.icio.us    Digg    Facebook    Reddit    Technorati    Twitter    What's this?

This article has been cited by other articles:

				M. Costea and S. Stefanovic Cuscuta jepsonii (Convolvulaceae): An invasive weed or an extinct endemic? Am. J. Botany, September 1, 2009; 96(9): 1744 - 1750. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Supplementary Data Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Web of Science (10) Citing Articles via Google Scholar Google Scholar Articles by Stefanovic, S. Articles by Costea, M. Search for Related Content PubMed Articles by Stefanovic, S. Articles by Costea, M. Agricola Articles by Stefanovic, S. Articles by Costea, M. Social Bookmarking                            What's this?