GRAMMICA (CONVOLVULACEAE) USING PLASTID AND NUCLEAR DNA SEQUENCES1

22 pages
52 views
of 22
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Share
Description
GRAMMICA (CONVOLVULACEAE) USING PLASTID AND NUCLEAR DNA SEQUENCES1
Tags
Transcript
  D ELIMITATION OF MAJOR LINEAGES WITHIN  C  USCUTA  SUBGENUS G  RAMMICA  (C ONVOLVULACEAE )  USING PLASTID AND NUCLEAR DNA  SEQUENCES 1 S ASˇA  S TEFANOVIC´ , 2,4 M ARIA  K UZMINA , 2 AND  M IHAI  C OSTEA 3 2 Department of Biology, University of Toronto at Mississauga, Mississauga, Ontario L5L 1C6, Canada; and 3 Department of Biology, Wilfrid Laurier University, Waterloo, Ontario N2L 3C5, Canada Subgenus  Grammica , the largest and most diverse group in the parasitic genus  Cuscuta , includes  ; 130 species distributedprimarily throughout the New World, with Mexico as its center of diversity. To circumscribe the subgenus and assess therelationships among its major lineages, we conducted the first phylogenetic study of   Grammica  using plastid  trnL-F   and nrITSsequences from a wide taxonomic sampling covering its morphological, physiological, and geographical diversity. With theexception ofone species belonging elsewhere, thesubgenus was found to be monophyletic. Theresults furtherindicate thepresenceof 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 ishomoplastic, 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 withinone 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 timesindependently during the evolution of flowering plants(Nickrent, 2002; APG II, 2003) and is found in approximately4000 plant species (Nickrent et al., 1998) representing ; 1% of the known angiosperm diversity. Parasitism is frequentlyassociated 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 havebeen notoriously difficult to study from a systematic andtaxonomic point of view. The resulting lack of knowledge of relationships within parasitic lineages as well as their preciserelationships to autotrophic relatives hampers our ability toconduct detailed comparative studies and to understand thesequence of events that have shaped the evolution of thesefascinating plants (Nickrent et al., 1998; Futuyma, 2004).The genus  Cuscuta  represents one such taxonomicallyproblematic group. Comprising some 165–175 currentlydescribed species,  Cuscuta  is nearly cosmopolitan in distribu-tion with its species found on every continent (except Antarctica), ranging from the 60th parallel north in Europeand Asia, to the Cape region of South Africa, and as far southas the 47th parallel in Argentina and Chile (Yuncker, 1932;Hunziker, 1950; Mabberley, 1997). All members of this genusare vines with twining, slender, pale stems, with reduced, scale-like leaves, and no roots. These stem parasites are attached tothe 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 reducedamounts 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 asdodders) are capable of parasitizing a wide range of herbaceousand woody crop plants, but for the most part they do not causesignificant agricultural losses due to the effectiveness of currently available methods of control (reviewed by Dawsonet al., 1994; Costea and Tardif, 2006). Members of this genuswere 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. Anassociation with the Convolvulaceae was recognized early on,based on reproductive morphology,but few attempts were madeto propose a more detailed scheme of relationships between Cuscuta  and nonparasitic members of the family. Theapproaches taken have fallen into two categories, either recognition of   Cuscuta  as a separate monotypic family,implying a sister-group relationship to the rest of Convolvula-ceae or placement of   Cuscuta  within Convolvulaceae under various taxonomic ranks (but without any further implicationson its possible relationships). Examples of the former approachinclude the classifications by Dumortier (1829) and Roberty(1952, 1964), followed by most major synoptic works onflowering plants (e.g., Cronquist, 1988; Takhtajan, 1997). Thelatter approach includes recognition of   Cuscuta  as tribeCuscuteae (Choisy, 1845; Bentham and Hooker, 1873; Baillon, 1 Manuscript received 20 September 2006; revision accepted 15 February2007. The authors warmly thank A. Colwell, T. Van Devender, T. Deroin, I.Garcı´a, 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, andan anonymous reviewer for critical comments on the manuscript. Thiswork was supported by NSERC of Canada grants to S.S. and M.C. 4 Author for correspondence (e-mail: sstefano@utm.utoronto.ca) 568 American Journal of Botany 94(4): 568–589. 2007.  1891; Hallier, 1893; Peter, 1897; Austin, 1998) or as subfamilyCuscutoideae (Peter, 1891; Melchior, 1964). Molecular phylo-genetic studies conducted on a broad sampling of Solanalesindicated that not only was  Cuscuta  a part of the Convolvu-laceae clade (Stefanovic´ et al., 2002), but also that it was nestedwithin that family, with at least two nonparasitic lineagesdiverging before  Cuscuta  (Stefanovic´ and Olmstead, 2004).Within  Cuscuta , Engelmann (1859) recognized three groupsbased primarily on stigma and style morphology. These groupswere formally adopted by Peter (1897) and later by Yuncker (1932) as subgenera (Fig. 1). Subgenus  Monogyna  has a singlestyle, partially to completely undivided, with a variety of stigma shapes. Subgenera   Cuscuta  and  Grammica  arecharacterized by two distinct styles and can be distinguishedby their stigma morphology (elongated and linear vs. short andcapitate, respectively). Plastid and nuclear sequence data obtained for a limited number of taxa identified three lineagesconsistent with the traditionally proposed subgenera andresolved subgenus  Monogyna  as the sister to the rest of thegenus and subgenera   Cuscuta  and  Grammica  as sister to eachother (Stefanovic´ 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 tosubgenus  Grammica  than to the other species from subgenus Cuscuta  (McNeal, 2005). To date,  Cuscuta  has not been thesubject of broad molecular phylogenetic analyses.Our research on Cuscuta was initiated with several goals inmind: (1) to test the monophyly of traditionally proposedsubgenera; (2) to circumscribe major lineages within subgen-era, particularly within the largest subgenus  Grammica ; (3) todevelop a well-supported phylogenetic hypothesis for   Cuscuta as a whole; (4) to investigate scenarios of morphologicalcharacter evolution within the genus; (5) to assess the relativeimportance of long-distance dispersal versus vicariance for thebiogeography of the genus; (6) to develop, in conjunction witha reevaluation of traditional taxonomic characters, a compre-hensive, phylogeny-based classification; and (7) to investigatein-depth the molecular processes of plastid genome evolutionwithin this group of parasitic plants.Any attempt to resolve longstanding controversies and tonurture a greater understanding of the numerous changes that have affected  Cuscuta , is in large part dependent on addressingthese problems within  Cuscuta  subgenus  Grammica , a groupthat epitomizes the complexity of the genus as a whole. Grammica  is by far the largest group of   Cuscuta , accountingfor approximately three-fourths of the species diversity of thegenus (130–135 spp.). While few members of this subgenus arewidespread, the vast majority of species occur only in theAmericas, with Mexico and adjacent regions as a center of diversity (Yuncker, 1932). Following the most recent andcomprehensive monograph of the genus (Yuncker, 1932),subgenus  Grammica  is divided into two sections,  Cleistogram-mica  and  Eugrammica , based on indehiscent or dehiscent capsules, respectively. Yuncker (1932) further subdivided eachof these sections into 12 subsections, based on a combinationof characters, and proposed a scheme of phylogeneticrelationships among them (Fig. 1). Characters used todistinguish the various subsections include the number, size,texture, and shape of flower parts, pedicel length, type anddensity of inflorescences, presence and shape of infrastaminalscales, ovary/capsule shape, embryo shape, and others.Unfortunately, many of these features are quantitative rather  Fig. 1. Precladistic scheme of classification for   Cuscuta , adopted and modified from Yuncker (1932). The scheme is based primarily on style andstigma 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 areindicated for each subsection. April 2007] S TEFANOVIC´ ET AL .—C IRCUMSCRIPTION OF  C USCUTA  SUBGENUS  G  RAMMICA  569  than qualitative, difficult to discern, or subjective and open tointerpretation. 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 sinceYuncker’s (1932) seminal work on this subject.Given the size and complexity of   Cuscuta  subgenus Grammica , the present study takes a   ‘‘ bottom-up ’’  phylogeneticapproach and focuses primarily on the first three of theaforementioned goals, i.e., the circumscription of major lineages within the subgenus, relationships among them, aswell 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 DNAsequences. This study presents the first phylogenetic analysisthat includes members of all sections and subsections of  Cuscuta  subgenus  Grammica .MATERIALS AND METHODS Taxon sampling  —  A total of 265 accessions representing 99 species wereused in this study. Species names, sources, voucher information, andcorresponding DNA extraction numbers are provided in the Appendix.Yunckers’s (1932) intrageneric classification is followed here because it isthe most widely used one and represents the only comprehensive work at thegeneric level. The 96 ingroup taxa, on which our analyses are primarilyfocused, include members of two traditionally recognized sections within Cuscuta  subgenus  Grammica  and all 24 recognized subsections. Due to thedifficulties in distinguishing many of these species morphologically, effort wasmade to sample multiple accessions of each. Approximately two-thirds of thespecies examined here are represented by more than one individual. Specialattention was paid to morphologically variable species containing more thanone subspecies/variety (e.g.,  C. salina ,  C. indecora ,  C. umbellata ) and to thosewith wide geographic range (e.g.,  C. campestris ,  C. gronovii ,  C. californica ).These species were represented by upward of 7–10 individuals from acrosstheir 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 otherwiserestricted areas or because they are underrepresented in collections. The latter isan 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 herbar-ium material was extracted using a modified hexadecyltrimethylammoniumbromide (CTAB) technique from Doyle and Doyle (1987) and purified usingWizard minicolumns (Promega, Madison, Wisconsin, USA). The polymerasechain reaction (PCR) was used to obtain the double-stranded DNA fragments of interest. The plastid genome (ptDNA) region containing the  trnL  intron, 3 0 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) wasobtained using primers ITS5 and ITS4 described by White et al. (1990). PCRwas carried out in 50  l L volumes with annealing temperatures of 50–55 8 C.Amplified products were cleaned by polyethylene glycol/NaCl precipitations.Cleaned products were sequenced directly, including both strands to ensureaccuracy, using the DYEnamic ET dye terminator sequencing kit (GEHealthcare, Baie-d’Urfe´, Quebec, Canada) on an Applied Biosystems model377 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 (EMDBiosciences, San Diego, California, USA), and multiple clones weresequenced. Sequence data were proofed, edited, and contigs assembled usingSequencher v.3.0 (Gene Codes Corp., Ann Arbor, Michigan, USA). Sequencesgenerated in this study are in GenBank (accession numbers EF194288– EF194718 and EF202557–EF202563; Appendix).  Phylogenetic analyses  —  Sequences were aligned manually using theprogram Se-Al v.2.0a11 (Rambaut, 2002). Although numerous gaps had tobe introduced in the alignments, the sequences were readily alignable amongthe ingroup taxa in both plastid and nuclear matrices. Regions that could not beunambiguously aligned were excluded from subsequent analyses. Gaps in thealignments were treated as missing data. Phylogenetic analyses were conductedusing 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 equallyweighted. Searches for most parsimonious (MP) trees were performed using a two-stage strategy with PAUP* version 4.0b10 (Swofford, 2002). First, theanalyses involved 1000 replicates with stepwise random taxon addition, tree-bisection-reconnection (TBR) branch swapping saving no more than 10 treesper replicate, and MULTREES option off. The second round of analyses wasperformed on all trees in memory with the same settings except with theMULTREES option on. Both stages were conducted to completion or until100000 trees were found. In addition, other searches were conducted using theparsimony  ‘‘ ratchet  ’’  analysis (Nixon, 1999) as implemented in NONA(Goloboff, 1999) with the WinClada interface (Nixon, 2002). Ten consecutivetree searches were conducted using 200 iterations per search, one tree held for each iteration, 10% of total characters sampled, and amb-poly ¼ (no swappingon ambiguously supported nodes), but they did not find shorter trees. Relativesupport for clades was inferred by nonparametric bootstrapping (Felsenstein,1985) as implemented in PAUP* using 500 pseudoreplicates, each with 20random sequence addition cycles, TBR branch swapping, and MULTREESoption off (DeBry and Olmstead, 2000). Conflict between data sets wasevaluated by visual inspection, looking for the presence of strongly supportedyet 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 discretegamma distribution and assuming a portion of invariant sites (GTR þ G þ I),was selected as the best-fit by both the hierarchical likelihood ratio test (hLRT)and Akaike information criterion (AIC), as implemented in ModelTest version3.7 (Posada and Crandall, 1998). Bayesian phylogenetic inferences wereperformed using MrBayes version 3.1.2 (Ronquist and Huelsenbeck, 2003) onthe combined data set only. Two runs starting from random trees were carriedout using the GTR  þ  G  þ  I substitution model. All model parameters weretreated as unknown variables with uniform prior probabilities and wereestimated as part of the analysis together with tree topologies. Metropolis-coupled Markov chain Monte Carlo algorithm was used with four simultaneouschains, set at two million generations, and sampled every 100 generations. Todetermine 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 thetwo runs were detected, the reported topologies and posterior probabilities (PP)are based on trees pooled from both independent Bayesian analyses. Only thenodes 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 designedto investigate the evolution of characters defining some traditional taxonomicgroups, were constructed and their cost in parsimony was assessed usingPAUP* (Swofford, 2002). To statistically compare resulting alternativephylogenetic hypotheses, we conducted one-tailed Shimodaira–Hasegawa tests(SH tests; Shimodaira and Hasegawa, 1999; Goldman et al., 2000) using theaforementioned substitution model and likelihood settings. The SH tests wereconducted with PAUP* using 1000 replicates and full parameter optimizationof the model. RESULTS Sequences and alignments  —Characteristics of the se-quenced regions as well as statistics of MP trees derived fromseparate and combined analyses are summarized in Table 1.The total aligned length of the  trnL-F   region is 689 bp, while570 A MERICAN  J OURNAL OF  B OTANY  [Vol. 94  individual sequences varied from 480 to 680 bp in length (480– 510 bp for the ingroups alone). Most of the  Cuscuta  speciesunder investigation were readily amplifiable for the  trnL-F  region with the same universal set of primers (C and F) usedroutinely for many other groups across angiosperms (Taberlet et al., 1991). However, this plastid region could not be obtainedfor a number of species belonging to several closely relatedsubsections sensu Yuncker (1932;  Grandiflorae ,  Odoratae ,  Acutilobae , and  Ceratophorae ), despite the fact that the sameDNA accessions produced nrITS fragments without difficulty.Several attempts to amplify smaller fragments with internaland/or alternative primers failed as well. Sequences were easilyaligned across most of the  trnL-F   region for all but one ingroupspecies sampled in this study. However, the spacer between 3 0 - trnL  and  trnF   is evolving more rapidly than the  trnL  intron interms of length and point mutations (as noted previously for Convolvulaceae in general; Stefanovic´ et al., 2002) and a portion of 120 bp was excluded from analyses. Three outgroupspecies from subgenus  Cuscuta  ( C. nitida ,  C. europaea , and  C.approximata ) yielded sequences significantly longer than thosefound within ingroup taxa. Except for short segmentscorresponding to  trnL  and  trnF   genes themselves, thesesequences could not be aligned with the ingroups and hencecould not be used in analyses to root trees. Surprisingly, oneputative ingroup species,  C. appendiculata  from South Africa,was also found to have a longer sequence, unalignable withother ingroup species. Furthermore, while  C. europaea  and  C.approximata  (both Eurasian in distribution) had significant levels of similarity and were easily alignable with each other, C. nitida  was divergent to the point that it could not be alignedwith the other two outgroups. This South African species hadhigh similarity only to  C. appendiculata , and their sequenceswere easily alignable with each other.Aligned sequences of nrITS used here were 717 bp in lengthwith the individual sequences varying between 560 and 600 bp.The length variation was more or less equally distributedthroughout the entire region. The nrITS sequences could not beobtained for all investigated individuals/species. However, theunsuccessful amplifications were randomly distributed (i.e., not part of any particular taxonomic group, unlike in the casedescribed for   trnL-F  ) and were probably due to the poor qualityof the DNA extracted from older herbarium specimens. For themajority of DNA accessions, the direct sequencing approachyielded results without apparent polymorphism. In some cases,however, it becomes clear that the PCR product contained morethan one type of nrITS sequence, and for those the cloningapproach was followed. In most of those cases, the  ‘‘ polymor-phism ’’  was caused by the presence of fungal DNA resultingeither from the natural presence of fungal epi- and endophytesin  Cuscuta  species or from fungal tissue contamination. Fungalsequences were easily separated from  Cuscuta  nrITS sequencesand 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 (resultsnot shown) indicated that these paralogous sequences were most closely related to each other, suggesting relatively recent duplication events or minor DNA polymerase error, and onlyone, randomly chosen, sequence was used to represent the givenindividual. The nrITS sequences were relatively easily alignableacross all ingroup species, and this whole region was includedin the phylogenetic analyses. However, in a way similar to that described for the  trnL-F   sequences, none of the three a priorichosen outgroup taxa from subgenus  Cuscuta  could be alignedwith ingroup species for the more variable ITS1 and ITS2regions. Only the highly conservative (and least informative)5.8S was alignable between ingroups and outgroups, andconsequently only this region could be used to root the nrITStrees. 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 relation-ships, it is of paramount importance to incorporate within-species variability and take into account possible biologicalphenomena that can confound results (such as lineage sorting,deep coalescence). For these reasons, most of the species in thestudy were represented by multiple individuals, sampled fromgeographically distinct areas and encompassing morphologicalvariability. However, the addition of terminal taxa results in a sharp increase of computational burden (Felsenstein, 1978). T ABLE  1. Summary descriptions for sequences included in, and maximum parsimony trees derived from, individual and combined analyses of   Cuscuta subgenus  Grammica . Description  trnL-F   (plastid) nrITS (nuclear) Combined data  Number of individuals sequenced a  223 207 265Number of OTUs analyzed b 141 153 161Sequence characteristics:Aligned length 689 717 1406Analyzed length c 525 676 1201Variable sites 241 450 691Parsimony informative sites 189 402 591Mean AT content ( % ) 63 50 55 d Base frequency homogeneity ( v 2  /df/   P ) 80.1/420/1.0 279.2/456/1.0 146.4/384/1.0 d Tree characteristics:Number of trees  . 100000  . 100000  . 100000Length 790 1965 2776CI/RI 0.52/0.89 0.447/0.886 0.465/0.885 a  Excluding the outgroup taxa that could not be aligned with the ingroup b After individuals with identical sequence for both regions were aggregated into a single terminal taxon c After excluding portions of alignments corresponding to primer sites and ambiguously aligned regions d Including only OTUs for which both sequences are available; CI, consistency index; df, degrees of freedom; OTU, operational taxonomic unit; RI,retention index April 2007] S TEFANOVIC´ ET AL .—C IRCUMSCRIPTION OF  C USCUTA  SUBGENUS  G  RAMMICA  571  Therefore, to facilitate the phylogenetic analyses, individuals of the same species having both the  trnL-F   and nrITS sequencesidentical to each other were grouped into a single operationaltaxonomic unit (OTU). Following this procedure, the 223individuals from  trnL-F   matrix were aggregated into 141OTUs, 207 from nrITS matrix into 153, and 265 individualsused in the combined data set were aggregated into 161 OTUs(Table 1). No significant heterogeneity in base compositionwas detected within any of these data matrices across all taxa. Tree topologies  —A number of distinct phylogenetic anal-yses were conducted using parsimony and Bayesian approach-es to explore the distribution of phylogenetic signal in thedifferent matrices. All analyses produced trees of remarkablysimilar topology although resolution and branch support varied. Tree characteristics from MP searches are summarizedin Table 1.  Individual data set analyses  —The  trnL-F   and nrITSmatrices produced  . 100000 trees, 790 and 1965 steps inlength, respectively. Schematic consensus trees from parsimo-ny 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 (seeSupplemental Data accompanying online version of thisarticle). A total of 15 major clades, labeled A–O, wereresolved within  Cuscuta  subgenus  Grammica  with nrITSsequences. Fourteen of the same groups, A–N, were alsorecovered with  trnL-F   data. However, none of the sequencesbelonging to clade O, a lineage consisting almost exclusivelyof South American species from subsections  Odoratae , Grandiflorae , and  Acutilobae , could be obtained for   trnL-F  .This plastid region could not be amplified either for severalspecies within clade K (e.g.,  C. erosa ,  C. boldinghii ), eventhough the same DNA accessions yielded good PCR productsfor nrITS. Nevertheless, other members of the K clade weresequenced for   trnL-F   and were available as placeholders in thephylogenetic analyses. Most of the 15 major clades receivedmoderate (70–85%) to strong bootstrap support ( . 85%) fromboth of the individual matrices. However, some groups werefound to be weakly supported ( , 70%) by one of the data setswhile receiving moderate to strong support from the other in a mutually complementary fashion. For example, clade N wassupported only by 59% BS with nrITS data, but it received80% BS from  trnL-F   data. In a complementary fashion, cladeC obtained , 50% BS for with plastid sequences, yet the sameclade was supported with 96% BS with nuclear data. Theoverall strong support for the circumscription of these 15 major clades stands in contrast to the less-resolved backbonerelationships within  Cuscuta  subgenus  Grammica  based onseparate analyses. The  trnL-F   phylogeny has only two well-supported backbone relationships, a group consisting of A–Cclades (receiving 83% BS) and a group consisting of L–Nclades (which received 96% BS). The nrITS consensus treewas somewhat more resolved, having three highly supportedbackbone nodes (100% BS for a group composed of A–Eclades, 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 involvedclades J and K. With plastid data, these two clades were placedas each other’s sister-group, whereas nuclear data resolvedthem as a successively diverging grade. However, thesealternative topologies are weakly supported ( , 70% BS) inboth cases. In addition, this difference represents only a slight topological distortion (a nearest-neighbor interchange) most likely caused by sampling discrepancies between  trnL-F   andnrITS matrices within the K clade (as described before). Takingall of this into account, we deemed these two matricescongruent and combined them into one data set.  Analyses of combined data sets  —The trees produced by thetotal-evidence approach had better resolution and overallsupport relative to those produced by independent analyses.Therefore, we have based our discussions on the analyses of the combined data sets. The parsimony analysis using thismatrix resulted in  . 100000 MP trees, each 2776 steps inlength. Figures 3–4 present the strict consensus of thoseequally parsimonious trees and one of them, randomly selected,was chosen to illustrate the branch lengths (Fig. 5). BothBayesian analyses, each initiated from a random starting tree, Fig. 2. Overview and comparison of strict consensus trees derivedfrom separate  trnL-F   and nrITS parsimony analyses. Fifteen major groupsare labeled A–O, and their parsimony bootstrap support values areindicated above branches. Plastid sequences could not be obtained for members of the O clade. Species relationships within the major clades arenot shown (see Appendices S1 and S2 for detailed trees in SupplementalData accompanying online version of this article). Trees are tentativelyrooted using the L–O clades as functional outgroups (see Results for fullexplanation). 572 A MERICAN  J OURNAL OF  B OTANY  [Vol. 94
Related Search
Similar documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks