© 2004 Society of Systematic Biologists
Testing the Phylogenetic Position of a Parasitic Plant (Cuscuta, Convolvulaceae, Asteridae): Bayesian Inference and the Parametric Bootstrap on Data Drawn from Three Genomes
Edited by Pam Soltis: Associate Editor
a Stefanovi
1
Department of Botany, University of Washington Box 355325, Seattle, Washington 98195–5325, USA
 |
Abstract |
|---|
 TopNotes Abstract Materials and MethodsResults and DiscussionConclusionsAppendix 1AcknowledgmentsReferences |
|---|
Previous findings on structural rearrangements in the chloroplast genome of Cuscuta (dodder), the only parasitic genus in the morning-glory family, Convolvulaceae, were attributed to its parasitic life style, but without proper comparison to related nonparasitic members of the family. Before molecular evolutionary questions regarding genome evolution can be answered, the phylogenetic problems within the family need to be resolved. However, the phylogenetic position of parasitic angiosperms and their precise relationship to nonparasitic relatives are difficult to infer. Problems are encountered with both morphological and molecular evidence. Molecular data have been used in numerous studies to elucidate relationships of parasitic taxa, despite accelerated rates of sequence evolution. To address the question of the position of the genus Cuscuta within Convolvulaceae, we generated a new molecular data set consisting of mitochondrial (atpA) and nuclear (RPB2) genes, and analyzed these data together with an existing chloroplast data matrix (rbcL, atpB, trnL-F, and psbE-J), to which an additional chloroplast gene (rpl2) was added. This data set was analyzed with an array of phylogenetic methods, including Bayesian analysis, maximum likelihood, and maximum parsimony. Further exploration of data was done by using methods of phylogeny hypothesis testing. At least two nonparasitic lineages are shown to diverge within the Convolvulaceae before Cuscuta. However, the exact sister group of Cuscuta could not be ascertained, even though many alternatives were rejected with confidence. Caution is therefore warranted when interpreting the causes of molecular evolution in Cuscuta. Detailed comparisons with nonparasitic Convolvulaceae are necessary before firm conclusions can be reached regarding the effects of the parasitic mode of life on patterns of molecular evolution in Cuscuta.
Keywords: Bayesian analysis; Convolvulaceae; cpDNA; Cuscuta; maximum likelihood; maximum parsimony; mitochondrial DNA; molecular systematics; parametric bootstrap; RPB2
Received August 13, 2002; Revised March 29, 2003; Accepted January 4, 2004
Understanding the fascinating changes that have shaped the evolution of parasitic plants would be greatly facilitated, from both morphological and molecular standpoints, by detailed comparative studies with their closest nonparasitic relatives. However, the phylogenetic position of parasitic angiosperms and their precise relationship to autotrophic relatives are not easy to deduce (Nickrent et al., 1998). In general, problems are encountered with both morphological and molecular evidence. Parasitism is associated with extreme reduction and/or modification of vegetative structures, and convergence with other parasitic taxa is common. Both phenomena are encountered in Cuscuta (dodder; Convolvulaceae; asterids). The morphology of this parasitic genus is characterized by loss of roots, significant reduction of chlorophyll synthesis, almost complete reduction of leaves and cotyledons, and the evolution of haustoria, organs that enable these plants to connect to the hosts. Both Cuscuta and Cassytha (Lauraceae; magnoliids) are pale, twining, stem parasites, and provide an excellent example of convergent evolution in parasitic plants (Kuijt, 1969).
Molecular data have been used in numerous studies seeking to elucidate the relationships of parasitic taxa (e.g., Nickrent and Starr, 1994; Wolfe and dePamphilis, 1995, 1997; dePamphilis et al., 1997; Duff and Nickrent, 1997; Young et al., 1999). Despite this effort, the phylogenetic affiliation of many parasitic groups, especially the so-called "nonasterid holoparasites" (Balanophoraceae, Cynomoriaceae, and Cytinaceae), are not known (Nickrent et al., 1998). In many cases the chloroplast and nuclear genes typically used to deduce large-scale flowering plant relationships (e.g., Chase et al., 1993; Soltis et al., 1997) are lost, significantly altered, or evolving at greatly accelerated rates, making phylogenetic inferences extremely difficult. In recent progress, the holoparasitic family Hydnoraceae was placed as sister to Aristolochiaceae (Nickrent and Duff, 1996; Nickrent et al., 2002) and enigmatic Rafflesia, genus with the largest known flowers, was found to be a member of rosids, most closely related to the order Malpighiales (Barkman et al., 2004). In contrast to nonasterid holoparasites, the general position of most hemiparasites (e.g., Cassytha, Krameriaceae, Santalales) as well as the "asterid holoparasites" (e.g., Lennoaceae, Orobanchaceae in part) in the global angiosperm phylogeny is not in dispute. However, even in these cases the precise relationships to nonparasitic taxa remain uncertain. For example, there is little doubt, based both on reproductive morphology and molecular data, that the hemiparsitic genus Cassytha is closely associated with Lauraceae, but its placement, either as a sister-group to Lauraceae or nested deeper within Lauraceae, remains uncertain (Rohwer, 2000; Renner and Chanderbali, 2000). Likewise, the small holoparasitic family Lennoaceae was recognized early on, based on floral and pollen morphology, to be closely related to Boraginaceae. Preliminary results of molecular analyses indicated that this family is indeed related to the Boraginaceae subfamily Ehretioideae, but its closest relatives remain uncertain (Smith et al., 2000). In certain instances, however, the closest nonparasitic relatives of parasitic plants were ascertained with high support using molecular data. For example, Lindenbergia, a nonparasitic genus, is clearly sister to the rest of Orobanchaceae s.l., which comprises both hemi- and holoparasitic species of the traditional Scrophulariaceae and Orobanchaceae (Young et al., 1999; Olmstead et al., 2001). Also, the parasitic species of Santalales are clearly nested within autotrophic members of the paraphyletic family Olacaceae (Nickrent and Duff, 1996; Nickrent et al., 1998).
The main source of difficulties for precise inference of relationships of parasitic angiosperms is the widespread homoplasy in molecular data due to accelerated rates of sequence evolution (reviewed in Nickrent et al., 1998), and associated analytical problems. Apparently, all three genomes in parasitic plants can be affected by this phenomenon (e.g., Colwell, 1994; Nickrent and Starr, 1994; Duff and Nickrent, 1997; Wolfe and dePamphilis, 1998). The relaxation of selection on genes involved in photosynthesis is one hypothesis for the increased nucleotide substitution rates in the chloroplast genome of parasitic plants (Wolfe and dePamphilis, 1998). However, it is not fully understood why the parasitic habit causes the acceleration observed in nuclear (Colwell, 1994; Nickrent and Starr, 1994; Neyland, 2000), and, to a lesser extent, mitochondrial genomes (Duff and Nickrent, 1997; Barkman et al., 2004), although small effective population size and molecular drive have been proposed as possible explanations (Nickrent and Starr, 1994; Nickrent et al., 1998).
The genus Cuscuta, consisting of some 160 to 170 species, is nearly cosmopolitan in distribution, and occurs in a wide range of habitats (Yuncker, 1932; Mabberley, 1987). Engelmann (1859) recognized three groups within Cuscuta, based primarily on the morphology of styles and stigma, which were assigned subgeneric ranks by Yuncker (1932). Subgenera Cuscuta and Grammica are characterized by two distinct styles, and are distinguishable by their stigma morphology (elongated or short and capitate, respectively). Subgenus Monogyna has a single style, partially or completely united, with capitate, conical, or ovate stigmas.
Even though vegetative characters are altered in association with its eccentric mode of life, Cuscuta floral morphology is quite similar to that of the Convolvulaceae, the morning-glory family, and the clear association with this family was recognized early on. Many classifications recognize a separate tribe (Choisy, 1845; Bentham and Hooker, 1873; Baillon, 1891; Hallier, 1893; Peter, 1897; Austin, 1998) or subfamily (Peter, 1891; Melchior, 1964) within Convolvulaceae for Cuscuta. However, some botanists (e.g., Roberty, 1952, 1964; Austin, 1973) adopted Dumortier's (1829) view that Cuscuta should be recognized as a separate family. This opinion is reflected in some major synoptic works on flowering plants (e.g., Cronquist, 1988; Takhtajan, 1997) that place Cuscuta in its own family, Cuscutaceae.
Cuscuta has been the focus of many scientific studies for several reasons. Many species are recognized to be pests on an array of important agricultural crops such as alfalfa, clover, beans, soy, cranberry, and, most importantly, members of the grass family. Infested crops now can be treated efficiently both by herbicides (Dawson, 1987, 1990) and by fungal bioagents (Bewick et al., 1987; Li, 1987). A substantial body of literature deals with the life history, ecology, and pest control of different dodder species (reviewed by Dawson et al., 1994, and references therein). Because this branch parasite is amenable to culture and direct experimental manipulation, it is also frequently used as a model system for developmental research, especially of haustorial initiation and formation (e.g., Dörr, 1987; Heide-Jörgensen, 1987; Lee and Lee, 1989; Subramaniam and Mahadevan, 1994).
In addition, Cuscuta has been the subject of extensive molecular analyses. Both hemiparasitic (e.g., C. reflexa) and holoparasitic (e.g., C. europaea) species occur in the genus. This diversity of photosynthetic ability among species prompted several physiological studies of photosynthetic enzymes and molecular evolutionary studies of the chloroplast genome. The results indicate that the hemiparasitic C. reflexa retains an affected, yet functional, chloroplast genome. This species retains most of the plastid genes generally found in autotrophic land plants, including both those involved in photosynthesis and ‘house-keeping’ functions (Haberhausen et al., 1992). However, putatively chlororespiratory (ndh) genes seem to be either altered to the point of becoming pseudogenes (ndhB) or are lost from the plastid genome (Haberhausen and Zetsche, 1994). The plastid genome of holoparasite C. europaea has sustained greater losses (Freyer et al., 1995), and the crude extract of this species shows no ribulose-1,5-bisphosphate carboxylase-oxygenase (Rubisco) activity nor light-dependent CO2 fixation, but cpDNA retains an rbcL open reading frame (Machado and Zetsche, 1990). The findings of chloroplast genome structural rearrangements in Cuscuta were attributed to its parasitic life style, but without proper comparison to related nonparasitic members of the family. The phylogenetic problems within the family, especially the placement of Cuscuta, as well as the relative placement of hemi- and holoparasitic species within this genus, need to be resolved satisfactorily before many evolutionary questions can be answered.
Convolvulaceae have been the subject of only one broad molecular phylogenetic study (Stefanovi et al., 2002). That study was based on sequences from four chloroplast loci—rbcL, atpB, psbE-J operon, and the trnL-F region—obtained from 112 taxa, including 109 species from all 10 traditionally recognized tribes (Austin, 1973; modified 1998) as well as three outgroups. Those results found that two groups that have been proposed as segregate families (Dumortier, 1829), Cuscuta and tribe Dichondreae, were nested within the Convolvulaceae. The exact position of Cuscuta could not be elucidated, however, mainly due to its highly divergent sequences. One alternative pertinent for circumscription of the family, the position of Cuscuta as sister to the rest of Convolvulaceae, was rejected. This result was further corroborated by the distribution of deletions in the atpB gene and trnL intron found in Cuscuta species as well as in all nonparasitic Convolvulaceae except Humbertia, which is sister to the rest of the family (Stefanovi et al., 2002).
To address the question of the position of Cuscuta within Convolvulaceae, we generated a new molecular data set consisting of mitochondrial (atpA), nuclear (RPB2), and chloroplast (rpl2) genes, and analyzed these data together with the existing (Stefanovi et al., 2002) chloroplast data matrix (rbcL, atpB, trnL-F, and psbE-J). The majority of data in this study are derived from organellar sequences. The chloroplast/mitochondrial-haplotype tree has a substantially higher probability of accurately inferring short internodes (e.g., those resulting from recent and/or rapid radiations) than does a nuclear gene tree due to more rapid coalescence time and lower subsequent substitution rates (Moore, 1995). The only nuclear data used in this study are derived from partial RPB2 gene sequences. The product of this gene forms a part of the catalytic core of the RNA polymerase II. This protein is highly conserved across the angiosperms at the amino-acid level, but the nucleotide sequences are quite variable, which enables its use at the lower phylogenetic level (Denton et al., 1998; Oxelman and Bremer, 2000).
This molecular data set, derived from all three plant genomes, is analyzed by three methods of character-based phylogenetic reconstruction: maximum likelihood (ML), maximum parsimony (MP), and Bayesian inference (BI). ML provides an objective way to incorporate important aspects of molecular evolution, such as unequal base frequencies, complex substitution models, and among-site rate variation. In addition, ML is a more consistent estimator of phylogenies (Felsenstein, 1981, 1988) and is less sensitive to the effects of rate variation (Huelsenbeck and Hillis, 1993; Huelsenbeck, 1995) than MP. However, taking advantage of complex models of DNA evolution imposes severe computational constraints, especially when a larger number of taxa is sampled (Sanderson and Kim, 2000). MP, on the other hand, is less impaired by large number of taxa (Hillis, 1996), but more realistic models of DNA sequence evolution are difficult to implement within this framework.
Bayesian phylogenetic inference (Larget and Simon, 1999; Huelsenbeck and Ronquist, 2001) is, like ML, a probabilistic method that relies on explicit models of sequence evolution. However, because it does not attempt to find the global optimum likelihood and uses Markov Chain Monte Carlo (MCMC) to estimate the posterior distribution of parameters (Huelsenbeck and Bollback, 2001), the BI method is, in comparison with ML, computationally much less intensive, and can be employed even with a relatively large number of taxa. An additional advantage of BI is that the interpretation of Bayesian posterior probabilities is considered to be straightforward, unlike the nonparametric bootstrap analyses, employed in both MP and ML contexts, where the relationship between bootstrap and statistical probability has been debated (e.g., Hillis and Bull, 1993). BI has recently been used successfully in addressing some of the most difficult phylogenetic problems such as the origin of land plants (Karol et al., 2001) and the early mammalian radiation (Murphy et al., 2001).
The main goal of this study is to narrow down the phylogenetic position of Cuscuta, the only parasitic genus associated with Convolvulaceae. In addition, we compare the performance of BI with ML and MP methods in the notoriously difficult task of assessing parasitic plant relationships. Finally, we explore an empirical implementation of the parametric bootstrap method for testing different alternative phylogenetic hypotheses using MP, which should be applicable to large data sets not only in Convolvulaceae, but also for sequence-based phylogenies in general.
|
Materials and Methods |
|---|
|
TopNotesAbstractMaterials and MethodsResults and DiscussionConclusionsAppendix 1AcknowledgmentsReferences |
|---|
Taxon Sampling
The 35 species included in this study (Appendix 1) represent a subset of taxa used by Stefanovi
et al. (2002) in their broad analysis of Convolvulaceae. Members of all but 1 out of 10 tribes sensu Austin (1973, modified 1998) were sampled. To represent the diversity of the family, efforts were made to include two or more species, selected to span the root node of each well supported clade, for all except very small clades circumscribed in our previous study. In addition, many species of unresolved relationships were sampled, especially those found within the "bifid style" clade (Stefanovi et al., 2002). Some genera, found to be monophyletic and strongly supported, but exhibiting long branches and uncertain placement are represented by two species (e.g., Jacquemontia, Erycibe). Cuscuta is also represented by two species, C. japonica (subgenus Monogyna) and C. europaea (subgenus Cuscuta), chosen as place-holders for the genus, because they exemplify the morphological (united style versus bifid style) and physiological (hemiparasite versus holoparasite) diversity within the genus. These two species showed the least amount of sequence divergence compared to the photosynthetic members of the family, and their sequences were fully alignable throughout the trnL-F region, which was not the case with any member of the more highly divergent subgenus Grammica. Relying on previously published molecular systematic studies of the asterids (e.g., Olmstead and Palmer, 1992; Chase et al., 1993; Soltis et al., 1997) and Convolvulaceae (Stefanovi et al., 2002), we selected two taxa (Nicotiana tabacum and Schizanthus pinnatus) spanning the root node of the Solanaceae, the sister family, as well as one additional species (Montinia caryophyllacea) belonging to the Solanales as outgroups.
The same species, and in many cases the same DNA isolate, was used to represent each taxon for each gene sequence whenever possible. However, certain gene regions that could not be obtained from a given species, due to the poor quality or lack of DNA, were sequenced from their respective closest relatives, as inferred from our previous analysis (Stefanovi et al., 2002). Those taxa are labeled on trees by genus name only.
Sequence Data and Alignment
In addition to the DNA samples used in our previous study (Stefanovi et al., 2002), total genomic DNA was isolated from herbarium specimens or silica-gel dried tissue (0.05 to 0.2 g), or from fresh (1 to 2 g) tissue by the modified hexadecyltrimethylammonium bromide (CTAB) procedure (Rogers and Bendich, 1985; Doyle and Doyle, 1987) and purified using Qiagen columns following protocols provided by the manufacturer.
Double-stranded DNA fragments for the regions of interest were obtained by polymerase chain reaction (PCR) from total genomic DNA using the primers described by Olmstead et al. (1992) for rbcL, by Hoot et al. (1995) for atpB, by Graham and Olmstead (2000) for the psbE-J operon and rpl2 gene, by Taberlet et al. (1991) for the trnL-F region, and by Davis et al. (1998) for the mitochondrial atpA gene. The nuclear RPB2 gene was shown to be duplicated in Gentianales (RPB2-i and RPB2-d; Oxelman and Bremer, 2000), and current research indicates that the duplication is found in the euasterid I clade (sensu APG, 1998) and in Ericales (Oxelman et al., 2004). Although multiple copies of similar, paralogous sequences can confound phylogenetic interpretations, the lack of introns in the RPB2-d copy in euasterid I plants makes a priori homology assessment easy. Plant-specific RPB2 primers P10F and P11aR (Denton et al., 1998) were used for initial amplifications of RPB2-d, targeting a region that corresponds to exons 18–24 in the Arabidopsis thaliana RPB2 gene. Based on these initial sequences, two more Convolvulaceae-specific primers were designed and used for PCR and sequencing (Conv-f: 5'-GCCATYGCMTGTYAYTCRGG-3'; and Conv-r: 5'-CGCCCTTGTGAATCTTGTCATCCACC-3'). Some PCR products, mainly those involving the low-copy nuclear RPB2 gene, were cloned (pCR2.1 vector; Invitrogen, Carlsbad, California, USA) and three to five clones were sequenced. Amplified PCR products were cleaned using Qiagen columns (Valencia, California, USA). Cleaned products were then directly sequenced, including both strands to ensure accuracy, using the BigDye Terminator cycle sequencing kit (PE Applied Biosystem, Foster City, California, USA) on an ABI 377 DNA automated sequencer (PE Applied Biosystem). Sequence data were edited and assembled using Sequencher 4.1 (Gene Codes Corporation, Ann Arbor, Michigan, USA). The alignments were obtained manually using the edit option of the MUST package (Philippe, 1993) or directly with PAUP* 4.0b10 (Swofford, 2002).
Phylogenetic Analyses
Bayesian inference
We used MrBayes v2.01 (Huelsenbeck and Ronquist, 2001) to perform the Bayesian phylogenetic analyses. This software uses a Metropolis-coupled Markov chain Monte Carlo, or "(MC)3," algorithm that runs several chains at the same time to permit a more thorough exploration of data space. Five separate runs were carried out, using the GTR+I+
model. The model parameters were treated as unknown variables with uniform prior probabilities and were estimated as part of the analyses together with tree topologies. We ran four chains simultaneously, three heated and one cold, enabling the random exchange of parameters between them, thereby minimizing the chance of being trapped in local optima. Each Markov chain was initiated from a random starting tree and run for 1 x 106 generations. For the first of the five runs, the sampling was done every 100th generation resulting in 10,000 samples. The remaining four runs were sampled every 50th generation for a total of 20,000 sample points per run. In order to determine whether and where stationarity was achieved, and to decide on the cut-off value ("burn-in," i.e., data points sampled before the chain reaches stationarity), the –log likelihood scores of each run were plotted against generation time. After discarding all samples preceding stationarity, the remaining data points were first analyzed separately for each run and than combined into a single file. Separate and combined files containing tree topologies were analyzed using PAUP* to compute the 50% majority rule consensus tree. The percentage of samples recovering any particular node represents the posterior probability of that node (Huelsenbeck and Ronquist, 2001). These values are considered to be the true probabilities of the nodes given the assumptions of the model of DNA sequence evolution (Rannala and Yang, 1996), and therefore the nodes receiving 
0.95 could be considered statistically significantly supported.
Maximum likelihood
The most complex model of nucleotide substitution currently available, the GTR+I+ model (Yang, 1994), was selected as the best-fit by ModelTest v3.04 (Posada and Crandall, 1998), by both the LR test and Akaike Information Criterion (AIC; Akaike, 1973). The gamma distribution was separated into four discrete rate classes. A heuristic analysis was done, using PAUP* (Swofford, 2002), with 20 replicates with stepwise random taxon addition, TBR branch swapping, and model parameters set to the values estimated by ModelTest v3.04. Because ML tree searches are computationally intensive, this procedure was conducted on the combined data set only.
To facilitate the nonparametric bootstrap analysis (Felsenstein, 1985) under the ML criterion, the topological constraint option in PAUP* was used to constrain certain taxonomic groupings that had been identified as monophyletic and strongly supported in our previous study (Stefanovi et al., 2002). This approach effectively reduces the number of terminal taxa in the analysis while maintaining all of the sequence data, thereby enabling the optimal assessments of substitutions on the tree (Olmstead et al., 1992). Internal nodes on the tree, where branching patterns are critical to the questions addressed by this analysis, were left unconstrained. Fourteen nodes that were constrained in the bootstrap analysis are indicated by asterisks in Figure 3. ML bootstrap analysis incorporated 100 pseudoreplicates, SPR branch swapping, starting trees obtained by neighbor-joining (with uncorrected "p" distances), and the same DNA sequence parameters estimated by ModelTest v3.04 as in the original ML search.
|
|
|
Maximum parsimony
For the parametric bootstrap analyses the heuristic searches for most parsimonious (MP) trees were performed using PAUP* (Swofford, 2002). Parsimony analyses of the data were conducted for each region separately and in combination using 1000 replicates (Maddison, 1991) with stepwise random taxon addition and TBR branch swapping and multrees on.
Testing of Alternative Hypotheses
Shimodaira-Hasegawa test
To compare alternative phylogenetic hypotheses statistically, the one-tailed Shimodaira-Hasegawa nonparametric tests (SH tests; Shimodaira and Hasegawa, 1999; Goldman et al., 2000) were conducted, using the aforementioned substitution model and likelihood settings. These tests are recommended for evaluation when the number of candidate trees is not very large (Shimodaira, 2002). The SH tests were conducted with PAUP* using 1000 bootstrap replicates and full parameter optimization of the model. In this fashion, a particular version of the SH test is implemented, referred to as the posNPfcd by Goldman et al. (2000). Using this approach, we tested the placement of Cuscuta by evaluating the ML topology against a set of topologies differing in the phylogenetic placement of this parasitic genus. All together, we tested seven alternative hypotheses against the optimal ML topology (depicted on Fig. 5).
|
|
Parametric bootstrap
The likelihood ratio (LR) test is frequently used to distinguish between competing hypotheses. As long as the tested hypotheses are nested, i.e., special cases of one another, the LR approximates a
2 statistic (G statistics) with degrees of freedom (df) equal to the difference in the number of parameters in the two models. In a phylogenetic context, this test is implemented to evaluate different models of DNA substitution and/or molecular clock hypotheses. However, one important common goal in phylogenetics—evaluating and choosing between competing tree topologies—cannot be tested using the 2 approximation, because different topologies are not nested within each other, and consequently, the df = 0. The parametric bootstrap (PB) is shown to be a statistically sound method of evaluating different alternative topological hypotheses (Huelsenbeck et al., 1996; Swofford et al., 1996; Goldman et al., 2000). This procedure uses a simulation to generate the null distribution from which statistical significance is deduced, thereby avoiding the need to rely on 2 statistics. However, despite the demonstrated power of the PB test (Goldman et al., 2000), it has been used very rarely on larger data sets (e.g., Knowles, 2000; Fishbein et al., 2001; Zanis et al., 2002), mainly due to the enormous computational cost this approach involves when implemented in the ML context (up to 200-fold the initial ML estimation time per one PB analysis). This time is further multiplied by the number of different hypotheses that one might wish to evaluate, because each alternative hypothesis requires not only the new test statistics but also the new null distribution for differences between the optimal tree and the model tree to be generated. The implementation of the PB for testing different a posteriori topologies is equally valid under the MP criterion (Goldman et al., 2000). Under this criterion some well known biases in the data are difficult to accommodate (see above), but it offers the ability to perform statistically sound, and thorough, tests in significantly less time (Sanderson and Kim, 2000). To compare support for the optimal tree against alternative branching hypotheses designed to investigate the placement of the parasitic genus Cuscuta, and to assess the significance of the observed differences between those trees given our data set, a series of PB tests was conducted (depicted on Fig. 5).
The implemented parametric bootstrapping procedure is summarized by the flow chart in Fig. 1. The original combined data matrix was used to obtain the optimal (MP) tree (Ha), as described above. An alternative topology constraint was constructed using MacClade (Maddison and Maddison, 1992). The best tree given this constraint and the original data set (H0) was assessed by implementing the topological constraints function in PAUP*. This analysis involved 100 replicates with stepwise random taxon addition and TBR branch swapping. The observed difference between H0 and Ha (
observed) represents the test statistic. In order to determine whether the observed is significant, i.e., whether it is larger than expected under the null hypotheses, the null distribution of differences must be generated via simulated data matrices. We simulated 99 data matrices of the same size as the original one using Seq-Gen v1.2.4.1 (Rambault and Grassly, 1997). The GTR+I+ model was used to simulate data sets, with parameters, including its branch lengths, estimated from the original data matrix and using the null hypothesis (constraint) topology.
For each of the 99 simulated matrices two searches were performed: first without any constraints, producing the h0 score, and the second with the constraint compatible with the original optimal (unconstrained) tree, resulting in ha score. Those searches were done using the same strategy as on the original data set. The difference for each pair (i.e., h01-ha1 through h099-ha99) was calculated. To these 's, the original observed is added (for a total of 100), and the histogram of the null distribution of 's was generated to determine the rejection region. The hypothesis that the difference observed between the original optimal tree and the constraint topology was due to chance alone is tested by comparison with this null distribution directly. The significance level is calculated as the proportion of times that the observed exceeds the values obtained in simulations (Huelsenbeck and Crandall, 1997; Goldman et al., 2000).
Because in parametric bootstrapping the simulated data sets are generated under the assumption that the particular null hypothesis (H0) is correct, neither the model parameters nor the simulated matrices could be reused. Therefore, this procedure was repeated for each alternative position of Cuscuta.
|
Results and Discussion |
|---|
|
TopNotesAbstractMaterials and MethodsResults and DiscussionConclusionsAppendix 1AcknowledgmentsReferences |
|---|
Sequences and Alignments
Characteristics of the sequenced regions as well as statistics of MP trees derived from each of the seven loci are summarized in Table 1. Descriptions of sequences derived from chloroplast genes used in this study that were published previously (i.e., rbcL, atpB, psbE-J, and trnL-F) are discussed in more detail in the original study (Stefanovi
et al., 2002). The alignments of three newly obtained genes, rpl2 (chloroplast), atpA (mitochondrion), and partial RPB2 (nucleus), were straightforward. All of these three protein coding regions exhibited some length variation, always within the open reading frame (ORF). Only one gap was needed to align sequences for all taxa for atpA and RPB2 (deletions in Cuscuta japonica and Erycibe glomerata, respectively). The rpl2 alignment required two gaps; one was an insertion in Cuscuta japonica and the other was an insertion in Montinia relative to all other sequences (but cannot be polarized). In addition, an intron usually found in the rpl2 gene of angiosperms is deleted in all Convolvulaceae, including Humbertia and Cuscuta, representing a unique event within asterids and a synapomorphy for Convolvulaceae (Stefanovi et al., 2002). Either because of incomplete sequences or gaps, the analyses involving sequence simulations may be biased owning to the presence of a large amount of missing information in the alignment. Thus, the portions of alignments where substantial sequence data are missing (e.g., 5' and 3' termini, and/or gaps) were excluded, resulting in a total analyzed length of 6842 bp (Table 1). No significant heterogeneity in base composition was observed within any of the separate matrices across all taxa (Table 1). Also, no significant difference in base composition was encountered among Cuscuta sequences alone. Due to the poor quality of the DNA extracted from herbarium specimens, sequences for the low copy nuclear RPB2 gene could not be obtained for two species, Humbertia madagascariensis and Wilsonia backhousei. Alignments in Nexus format are available on request from the first author, and have been archived also at the Systematic Biology (http://ag.arizona.edu/systbiol/SSBWeb/) website.
|
Phylogenetic Analyses and Implications for the Placement of Cuscuta
The separate equally weighted MP analyses of seven DNA matrices, conducted to detect potential areas of strongly supported incongruence, gave remarkably similar results (results not shown). All analyses identified several well-supported monophyletic groups. Differences mainly involved the number of resolved nodes and their bootstrap support. Visual inspection of the resulting cladograms revealed no topological incongruences that were, at the same time, conflicting and well-supported by different data partitions. Because these independent analyses gave congruent results, albeit quite unresolved, we combined all seven matrices. The trees produced by combined analysis had better resolution and overall support relative to those produced by independent analyses. Therefore, we have based our discussion on the analyses of the combined data set.
All five Bayesian analyses, each initiated from a random starting tree, converged on similar log-likelihood scores and reached stationarity at no later than 200,000 generations (Fig. 2). The initial 2000 samples from the first run, and 4000 from each of four more densely sampled runs, were discarded. In both cases the discarted samples accounted for 20% of the total sample points. The burn-in of data points accumulated before stationarity left a total of 72,000 combined samples. A majority-rule consensus of the 72,000 trees resulted in the phylogenetic hypothesis depicted in Figures 3 and 4
. When analyzed separately, all five independent runs found essentially identical tree topologies and posterior probabilities (results not shown), indicating that the sample number was sufficient to permit the algorithm to converge on a global solution.
The relationships inferred through the BI analysis (Figs. 3 and 4) are topologically congruent with results derived from a data set with larger taxon sampling (Stefanovi et al., 2002) under the MP criterion. According to the BI results Humbertia forms the sister to the rest of the family. The next two lineages to diverge within the Convolvulaceae are two small clades, one comprising some members of Poraneae and the other the genus Erycibe. Monophyly of each of these two lineages is strongly supported, but their progressively more terminal placements on the optimal tree is not (P = 0.5). The rest of the family is split into two major clades (Fig. 3; clade 1, clade 2). Clade 1 includes tribes Argyreieae, Ipomoeeae, Convolvuleae, and Merremieae and comprises the majority of species in the family. Within this clade the relationships are largely resolved and well supported except for the relationships of Merremieae and Convolvuleae. Clade 2 consists mostly of taxa that have a more or less deeply divided style ("bifid style" clade) and includes several strongly supported subclades, but the backbone relationships are largely unresolved. Overall, 24 out of 32 nodes (75%) were supported with a significance level 95%. However, the position of Cuscuta as a sister-group of clade 1 on the optimal tree is not significantly supported (P = 0.58). The remaining 42% of bipartitions place Cuscuta either as sister-group to clade 2 (26%) or as sister to the large clade combining both clades 1 and 2 (16%).
Under the ML criterion, with the GTR+I+ model of sequence evolution, one optimal tree was obtained. This tree differed from the optimal BI solution only in branching patterns for two weakly supported nodes (Fig. 3; dotted lines). The position of Cuscuta is the same as in the BI tree, with similarly weak support (52% BS).
Taken overall, the combined data analysis, using two probabilistic methods, ML and BI, recovered highly congruent topologies (Figs. 3 and 4). The points of disagreement included only those nodes that have not received substantial support. The basal position of Humbertia, subdivision of the rest of the family into two major clades, the relationships within clade 1 as well as most of the relationships within clade 2, are all points of complete agreement. The results from the present analyses are also quite similar, both in inferred patterns of evolution and support, as well as lack of support in certain regions, with the MP analysis based on a larger taxon sampling and chloroplast data only (Stefanovi et al., 2002). However, the overall support for the position of Cuscuta remains weak under all criteria, including BI, which has proved effective in resolving some other difficult phylogenetic issues such as the early mammalian radiation and land plant phylogeny (Murphy et al., 2001; Karol et al., 2001). The consensus places this parasitic genus in the general vicinity of clades 1 and 2, without further bearing on exact patterns among these three groups. The current evolutionary hypothesis for Convolvulaceae based on these different analyses is summarized in Figure 5.
Evaluation of Alternative Placements for Cuscuta
Traditional classifications, for the most part, ignored the question of Cuscuta's precise relationships with its putative nonparasitic relatives, owing mainly to the lack of useful taxonomic characters. Even though the association with the Convolvulaceae was recognized, based on reproductive morphology, no attempts were made to propose a more detailed scheme of relationships between Cuscuta and nonparasitic members of the family. The approaches taken have usually fallen into two categories: (1) recognition of Cuscuta as a monotypic family, implying, in modern terms, a sister-group relationship to the rest of Convolvulaceae, or (2) placing Cuscuta within Convolvulaceae under various taxonomic ranks, but without bearing any further on its possible relationships. Examples of the former approach include classifications by Dumortier (1829) and Roberty (1952, 1964), which were subsequently 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). An exception to this is a study by Austin (1973). Austin system also segregated Cuscuta as a separate family, but a close association with tribe Dichondreae is implicit from his phylogenetic scheme based mainly on chromosome numbers (his Fig. 33). This Cuscuta-Dichondreae connection was supported by some unique shared fruit features as well as similarities in embryo morphology.
As expected, Cuscuta exhibits strong rate acceleration in chloroplast and, to a somewhat lesser extent, in mitochondrial and nuclear DNA evolution (Nickrent et al., 1998; results shown only in aggregate, Fig. 4). Even though optimal trees from all three methods of phylogenetic inference place Cuscuta nested well within the Convolvulaceae, its precise placement is not supported. Therefore, an in-depth analysis of the relationship of this genus with the remainder of the family is warranted. All together, seven hypotheses were tested concerning possible placements of Cuscuta using two different approaches: Shimodaira-Hasegawa test (Shimodaira and Hasegawa, 1999; Goldman et al., 2000) under the ML criterion and parametric bootstrapping (Swofford et al., 1996; Huelsenbeck and Crandall, 1997; Goldman et al., 2000) under the MP criterion. The different tested points of attachment for constrained topologies are marked in Figure 5, and results are summarized in Table 2.
|
One alternative topology was designed to test the relationship proposed implicitly by Austin (1973). According to this hypothesis Cuscuta would be more closely related to Falkia, the place-holder for tribe Dichondreae in the present study (Fig. 5, no. 1). The SH test found significant difference in likelihood between this topology and the ML tree. Similarly, the PB test detected a significant difference. The observed length difference of 25 steps was far beyond the greatest difference (16 steps) observed in any simulated data sets (P < 0.01), rejecting the hypothesis of a closer Cuscuta-Dichondreae connection. It is possible, though unlikely, that Cuscuta could be closely related to some other members of the "bifid clade." The formal tests, however, were not conducted, because the backbone relationships in this clade were not resolved.
The other six tests conducted were designed to explore the limits of significance for alternative hypotheses and to help narrow down the possible placements for Cuscuta (Fig. 5, nos. 2 to 7). Clade 1 contains approximately 2/3 of all species belonging to Convolvulaceae, and the backbone relationships within that clade are resolved and well supported. We wanted to determine the cost in log-likelihood and parsimony and its significance for two alternatives in which Cuscuta would be nested within clade 1 (Fig. 5, nos. 2 and 3). Both SH and PB tests detected a significant difference (Table 2), suggesting that the hypotheses of closer association of Cuscuta with either the clade containing Argyreieae, Ipomoeeae, Convolvuleae, and some Merremieae or the clade containing the rest of Merremieae can be rejected. Even though the current consensus of evolutionary hypotheses for Convolvulaceae based on different analyses places Cuscuta effectively in a polytomy with clades 1 and 2 (Fig. 5), we deemed it important to test this placement formally as well. Not surprisingly, neither the topology with Cuscuta as a sister group to clade 2, nor with Cuscuta as the sister group to clades 1 and 2 together (Fig. 5; nos. 4 and 5, respectively) could be rejected with either SH or PB tests (Table 2). Finally, two alternative hypotheses, bearing the most importance for the circumscription of the family as a whole, were also tested. Both of these alternative positions, Cuscuta diverging within the family, as sister to all, except Humbertia (Fig. 5; no. 6), and Cuscuta as a sister group to the rest of the family, i.e., consistent with its recognition as a distinct family (Fig. 5; no. 7), were found to be significantly worse than the optimal trees according to the SH and PB tests.
The Shimodaira-Hasegawa test (Shimodaira and Hasegawa, 1999) is a statistically appropriate nonparametric method for comparing a posteriori tree topologies (Goldman et al., 2000). The parametric bootstrap (Swofford et al., 1996; Huelsenbeck and Crandall, 1997; Goldman et al., 2000), although shown to be more powerful than the SH test, given that the assumptions of sequence evolution are not significantly violated (Goldman et al., 2000; Buckley, 2002), is used much less frequently (Knowles, 2000; Fishbein et al., 2001; Zanis et al., 2002) due, in part, to the much greater computational time involved. When implemented under the ML criterion, a number of shortcuts can be employed to lower the computational cost: (1) reduction in number of terminals, either by complete exclusion of taxa from analyses or by keeping all taxa, but constraining a number of nodes, which effectively reduces the number of terminals; (2) avoiding more time-consuming repeated random addition of taxa by using a neighbor-joining starting tree; and/or (3) implementing faster, but less thorough, branch swapping algorithms such as SPR or NNI. However, these shortcuts, especially when combined, present risk of missing the global optima. Using less than optimal trees, whatever the optimality criterion might be, to calculate the differences for simulated data matrices ( 's) will potentially result in a biased null distribution of 's from which the P value is calculated. In addition, even when the shortcuts are employed, the computational time might still be prohibitive with ML, rendering the implementation of tests impractical (Buckley, 2002), especially if more than one alternative needs to be tested.
The maximum parsimony framework for parametric bootstrapping is an equally suitable approach to generate the null distribution of 's (Goldman et al., 2000) and this distribution can be used to determine the significance of the observed cost in parsimony for the alternative. Comparatively few statistical assessments of alternative hypotheses using MP have been published (e.g., Ruedi et al., 1998; Carlini et al., 2000; Zanis et al., 2002). The advantages of the MP approach with parametric bootstrapping is twofold: (1) the computational time is greatly reduced enabling the testing of multiple hypotheses, and, more importantly, (2) the reasonably extensive measures can be taken to explore the tree space more thoroughly in search for global optima, thereby preventing potential biases in P value calculations and known tendency of PB test to generate type 1 errors (Buckley, 2002). In order to minimize the bias and to err on the conservative side (by failing to reject H0), at minimum the same effort should be invested in exploration of simulated data matrices as was invested in searches involving the original matrix. In that case, however, the utility of the PB test would be severely restricted because of computational burden imposed by repeated ML topological searches in conjunction with ever increasing size of molecular data sets (Buckley, 2002). Hence, with moderate to large data sets (> 30 terminals), multiple, yet thorough PB tests can be done under the MP criterion, but not under ML. One disadvantage of this approach is that some well-known characteristics of DNA sequence evolution cannot be readily incorporated within an MP framework. Also, the MP criterion will not be suitable to test certain types of hypotheses, notably those involving branch lengths, where the MP approach was demonstrated as underestimating branch lengths (Saitou, 1989). It will be on a case-by-case basis that researchers will evaluate the costs and benefits of different approaches, and decide which optimality criterion might be more suitable for any particular study.
For all seven hypotheses tested in the present study, the results from SH and PB tests were in agreement not only regarding the conclusions drawn from them, i.e., rejecting or failing to reject the null hypothesis, but also in P values associated with the rejection region (Table 2). The PB is putatively more powerful than the SH test for discrimination among different alternative hypotheses (Goldman et al., 2000), and our results are consistent with this idea. However, the extent of the differences between probabilities estimated by our tests was much smaller than in other studies implementing SH and PB tests (e.g., Goldman et al., 2000; Fishbein et al., 2001).
Phylogenetic Relationships of Nonparasitic Convolvulaceae
Besides helping to narrow down the phylogenetic position of Cuscuta, the present study also provides additional support for the relationships among nonparasitic taxa of Convolvulaceae. This is most evident with respect to the "bifid style" clade. This clade comprises genera that have a more or less deeply divided style, generally following the concept of Dicranostyleae proposed by Hallier (1893). The "bifid style" clade was first explicitly identified by Stefanovi et al. (2002), but support for its monophyly was weak (39% with Cuscuta species included in analysis; 80% without Cuscuta). The present study, with additional sequence data, offers further evidence, not only for the monophyly of that clade (P = 1 from BI; 89% bootstrap support with ML), but also its composition. The inclusion of Jacquemontia in the "bifid style" clade was one of the most surprising results of molecular phylogenetic study of Convolvulaceae. This genus is traditionally regarded as a member of tribe Convolvuleae due to its undivided, filiform style with elongated stigmas. Even though the defining morphological character, divided style, is not present in Jacquemontia, this genus was found to share a unique synapomorphy with the rest of the "bifid style" clade, reversion to a nonedited start codon for the psbL gene (Stefanovi et al., 2002). This condition is not found anywhere else in Convolvulaceae and its closest relatives. Given the unresolved relationships within the "bifid style" clade, one possible evolutionary scenario is that Jacquemontia is the sister group to the rest of the taxa with divided styles. This would account for a single origin for each of these two characters. The BI analysis lends some support to this scenario, resolving Jacquemontia as the sister-group to the rest of the "bifid style" clade (Fig. 3). However, this relationship on the optimal tree is not well supported, and was not recovered using other methods of reconstruction. Also, the present study offers confirmation and further support for polyphyly of Erycibeae, Poraneae, and Merremieae, as well as close relationships of tribes Hildebrandtieae with Cresseae, Ipomoeeae with Argyreieae, and Dichondreae with some members of Poraneae (Fig. 5).
|
Conclusions |
|---|
|
TopNotesAbstractMaterials and MethodsResults and DiscussionConclusionsAppendix 1AcknowledgmentsReferences |
|---|
In terms of both the quantity of DNA sequence data and range of analytical methods, this study represents an intensive effort to estimate the phylogenetic position of a relatively small clade of parasitic plants. The inability to recover the exact position of Cuscuta with confidence even with relatively large amounts of data exemplifies the magnitude of the problem in inferring phylogenetic relationships of parasitic taxa. However, this approach did help to narrow down the position of Cuscuta and to reject with confidence a number of alternative hypotheses. Bayesian analysis, used with success to resolve other difficult phylogenetic problems, offered no more insight for Cuscuta placement than did maximum likelihood and maximum parsimony. However, the majority-rule consensus tree derived from the Bayesian analysis was very similar to the best phylogeny inferred by maximum likelihood analysis, both in terms of inferred topology and support, while requiring significantly less computational time. Significant computational time saving was achieved also by implementing the parametric bootstrap under the maximum parsimony criterion to test a series of alternative topologies.
The taxonomic implications of this study will have positive impacts on comparative studies of Cuscuta, which presently use the currently available classification as a framework. It is important for future comparative studies, especially those on chloroplast genome evolution, to recognize that the differences observed in different Cuscuta species are not attributable solely to Cuscuta's parasitic mode of life and that a significant proportion of those changes could be better explained as a plesiomorphic condition within the family, i.e., conditions shared with other members of the Convolvulaceae.
|
Appendix 1 |
|---|
|
TopNotesAbstractMaterials and MethodsResults and DiscussionConclusionsAppendix 1AcknowledgmentsReferences |
|---|
|
|
Acknowledgments |
|---|
|
TopNotesAbstractMaterials and MethodsResults and DiscussionConclusionsAppendix 1AcknowledgmentsReferences |
|---|
The authors thank D. F. Austin, A. L. Denton, Th. Deroin, R. Neyland, and G. W. Staples, as well as the curators of A, H, K, MO, NY, P, US, and WTU for supplying plant material. We are also thankful to Sean Graham and Joel McNeal for helpful discussions. For constructive comments on earlier versions of the manuscript, we thank Pam Soltis, Chris Simon, Claude dePamphilis, and Vincent Savolainen. This work was supported by a Karling Graduate Student Research Award from the Botanical Society of America to S. Stefanovi
, a Research Award for Graduate Students from the American Society of Plant Taxonomists to S. Stefanovi, and NSF Doctoral Dissertation Improvement grant DEB-0073396 to R. G. Olmstead for S. Stefanovi. |
Notes |
|---|
|
TopNotesAbstractMaterials and MethodsResults and DiscussionConclusionsAppendix 1AcknowledgmentsReferences |
|---|
1 Present address: Department of Biology, Indiana University, 1001 E. Third Street, Bloomington, IN 47405, USA; E-mail: sstefano{at}bio.indiana.edu
|
References |
|---|
|
TopNotesAbstractMaterials and MethodsResults and DiscussionConclusionsAppendix 1AcknowledgmentsReferences |
|---|
-
Akaike H. Information theory as an extension of the maximum likelihood principle. Pages 267–281 in Second International Symposium on Information Theory—Petrov B. N., Csaki F., eds. (1973) Budapest: Akademiai Kiado.
An ordinal classification for the families of flowering plants. Ann. Mo. Bot. Gard.—Angiosperm Phylogeny Group (APG), ed. (1998) 85:531–553.[CrossRef]
Austin D. F. The American Erycibeae (Convolvulaceae): Maripa, Dicranostyles, and Lysiostyles I. Systematics. Ann. Mo. Bot. Gard. (1973) 60:306–412.[CrossRef]
Austin D. F. Parallel and convergent evolution in the Convolvulaceae. Pages 201–234 in Biodiversity and taxonomy of tropical flowering plants—Mathews P., Sivadasan M., eds. (1998) Culicut, India: Mentor Books.
Baillon H. Convolvulacées. Pages 305–331 in Histoire des plante (1891) Paris, France: Librairie Hachett and Cie.
Barkman T.J, Lim S.-H., Salleh K. M., Nais J. Mitochondrial DNA sequences reveal the photosynthetic relatives of Rafflesia, the world's largest flower. Proc. Natl. Acad. Sci. U.S.A. (2004) 101:787–792.
Bentham G., Hooker J. D. Genera plantarum (1873) 2:865–881.
Bewick T. A., Binning L. K., Stevenson W. R., Stewart J. A mycoherbicide for control of swamp dodder (Cuscuta gronovii Willd.), Cuscutaceae. Pages 93–104 in Parasitic flowering plant—Weber H. C., Forstreuter W., eds. (1987) Marburg, Germany: Philipps-Universität.
Bömmer D., Haberhausen G., Zetsche K. A large deletion in the plastid DNA of the holoparasitic flowering plant Cuscuta reflexa concerning two ribosomal proteins (rpl2, rpl23), one transfer RNA (trn I) and an ORF 2280 homologue. Curr. Genet. (1993) 24:171–176.[CrossRef][Web of Science][Medline]
Buckley T. R. Model misspecification and probablistic tests of topology: Evidence from empirical data sets. Syst. Biol. (2002) 51:509–523.
Carlini B. B., Reece K. S., Graves J. E. Actin gene family evolution and the phylogeny of coleoid caphalopods (Mollusca: Cephalopoda). Mol. Biol. Evol. (2000) 17:1353–1370.
Chase M. W., Soltis D. E., Olmstead R. G., Morgan D., Les D. H., Mishler B. D., Duvall M. R., Price R., Hills H. G., Qui Y.-L., Kron K. A., Rettig J. H., Conti E., Palmer J. D., Manhart J. R., Sytsma K. J., Michaels H. J., Kress W. J., Karol K. G., Clark W. D., Hedren M., Gaut B. S., Jansen R. K., Kim K.-J., Wimpee C. F., Smith J. F., Furnier G. R., Strauss S. H., Xiang Q.-Y., Plunkett G. M., Soltis P. S., Williams S. E., Gadek P. A., Quinn C. J., Eguiarte L. E., Golenberg E., Learn G. H., Graham S. W., Barrett S. C. H., Dayanandam S., Albert V. A. Phylogenetics of seed plants: An analysis of nucleotide sequences from the plastid gene rbcL. Ann. Mo. Bot. Gard. (1993) 80:528–580.[CrossRef]
Choisy J. D. Convolvulaceae. A. DeCandole, Prodromus (1845) 9:323–465.
Colwell A. E. Genome evolution in a non-photosynthetic plant, Conopholis americana (1994) St. Louis, MOUSA: Washington University. Ph.D. dissertation.
Cronquist A. The evolution and classification of flowering plants (1988) Bronx, NY, USA: The New York Botanical Garden.
Davis J. I., Simmons M. P., Stevenson D. W., Wendel J. F. Data decisiveness, data quality, and incongruence in phylogenetic analysis: An example from the monocotyledons using mitochondrial atpA sequences. Syst. Biol. (1998) 47:282–310.
Dawson J. H. Cuscuta (Convolvulaceae) and its control. Pages 137–149 in Parasitic flowering plant—Weber H. C., Forstreuter W., eds. (1987) Marburg, Germany: Philipps-Universität.
Dawson J. H. Dodder (Cuscuta spp.) control with dinitroaniline herbicides in alfalfa (Medicago sativa). Weed Techn (1990) 4:341–348.
Dawson J. H., Musselman L. J., Wolswinkel P., Dörr I. Biology and control of Cuscuta. Rev. Weed Sci. (1994) 6:265–317.
Denton A. L., McConaughy B. L., Hall B. D. Usefulness of RNA polymerase II coding sequence for estimation of green plant phylogeny. Mol. Biol. Evol. (1998) 15:1082–1085.[Web of Science][Medline]
de Pamphilis C. W., Young N. D., Wolfe A. D. Evolution of plastid gene rps2 in a lineage of hemiparasitic and holoparasitic plants: Many losses of photosynthesis and complex patterns of rate variation. Proc. Natl. Acad. Sci. U.S.A. (1997) 94:7367–7372.
Dörr I. The haustorium of Cuscuta—new staructural results. Pages 163–170 in Parasitic flowering plants. Weber H. C., Forstreuter W., eds. (1987) Marburg, Germany: Philipps-Universität.
Doyle J. J., Doyle J. L. A rapid DNA isolation procedure for small quantities of fresh leaf tissue. Phytochem. Bull. (1987) 19:11–15.
Duff R. J., Nickrent D. L. Characterization of mitochondrial small-subunit ribosomal RNAs from holoparasitic plants. J. Mol. Evol. (1997) 45:631–639.[CrossRef][Web of Science][Medline]
Dumortier B.-C. Analyse des plantes (1829) Paris, France: Tournay.
Engelmann G. Systematic arrangement of the species of the genus Cuscuta with critical remarks on old species and descriptions of new ones. Trans. Acad. Sci. St. Louis (1895) 1:453–523.
Felsenstein J. Case in which parsimony or compatibility methods will be positively misleading. Syst. Zool. (1978) 27:401–410.
Felsenstein J. Evolutionary trees from DNA sequences: A maximum likelihood approach. J. Mol. Evol. (1981) 17:368–376.[CrossRef][Web of Science][Medline]
Felsenstein J. Confidence limits on phylogenies: An approach using the bootstrap. Evolution (1985) 39:783–791.[CrossRef][Web of Science]
Felsenstein J. Phylogenies from molecular sequences: Inference and reliability. Annu. Rev. Genet. (1988) 22:521–565.[CrossRef][Web of Science][Medline]
Fishbein M., Hibsch-Jetter C., Soltis D. E., Hufford L. Phylogeny of Saxifragales (angiosperms, eudicots): Analysis of a rapid, ancient radiation. Syst. Biol. (2001) 50:817–847.
Freyer R., Neckermann K., Maier R. M., Kössel H. Structural and functional analysisof plastid genomes from parasitic plants: Loss of an intron within the genus Cuscuta. Curr. Genet. (1995) 27:580–586.[CrossRef][Web of Science][Medline]
Goldman N., Anderson J. P., Rodrigo A. G. Likelihood-based tests of topologies in phylogenetics. Syst. Biol. (2000) 49:652–670.
Graham S. W., Olmstead R. G. Utility of 17 chloroplast genes for inferring the phylogeny of the basal angiosperms. Am. J. Bot. (2000) 87:1712–1730.
Haberhausen G., Valentin K., Zetsche K. Organization and sequence of photosynthetic genes from the plastid genome of the holoparasitic flowering plant Cuscuta reflexa. Mol. Gen. Genet. (1992) 232:154–161.[CrossRef][Web of Science][Medline]
Haberhausen G., Zetsche K. Functional loss of all ndh genes in an otherwise relatively unaltered plastid genome of the holoparasitic flowering plant Cuscuta reflexa. Plant Mol. Biol. (1994) 24:217–222.[CrossRef][Web of Science][Medline]
Hallier H. Versuch einer natürlichen Gleiderung der Convolvulaceen auf morphologischer und anatomischer Grundlage. Bot. Jahrb. Syst. (1893) 16:453–591.
Heide-Jørgensesn H. S. Changes in cuticle structure during development and attachment of the upper haustorium of Cuscuta L. Cassytha L. and Viscum L. Pages 319–334 in Parasitic flowering plants—Weber H. C., Forstreuter W., eds. (1987) Marburg, Germany: Philipps-Universität.
Hendy M. D., Penny D. A framework for the quantitative study of evolutionary trees. Syst. Zool. (1989) 38:297–309.
Hillis D. M. Inferring complex phylogenies. Nature (1996) 383:130–131.[CrossRef][Medline]
Hillis D. M., Bull J. J. An empirical test of bootstrapping as a method for assessing confidence in phylogenetic analysis. Syst. Biol. (1993) 42:182–192.
Hoot S. B., Culham A., Crane P. R. The utility of atpB gene sequences in phylogenetic relationships: Comparison with rbcL and 18S ribosomal DNA sequences in Lardizabalaceae. Ann. Mo. Bot. Gard. (1995) 82:194–207.[CrossRef]
Huelsenbeck J. P. Performance of phylogenetic methods in simulations. Syst. Biol. (1995) 44:17–48.
Huelsenbeck J. P., Bollback J. P. Empirical and hierarchical Bayesian estimation of ancestral states. Sysyt. Biol. (2001) 50:351–366.
Huelsenbeck J. P., Crandall K. A. Phylogeny estimation and hypothesis testing using maximum likelihood. Annu. Rev. Ecol. Syst. (1997) 28:437–466.[CrossRef][Web of Science]
Huelsenbeck J. P., Hillis D. M. Success of phylogenetic methods in the four-taxon case. Syst. Biol. (1993) 42:247–264.
Huelsenbeck J. P., Hillis D. M., Jones R. Parametric bootstrapping in molecular phylogenies: Application and performance. Pages 19–45 in Molecular zoology: Advances, strategies, and protocols—Ferris J. D., Palumbi S. R., eds. (1996) New York, NY, USA: Wiley.
Huelsenbeck J. P., Ronquist F. MrBayes: Bayesian inference of phylogeny. Biometrics (2001) 17:754–755.
Karol K. G., McCourt R. M., Cimino M. T., Delwiche C. F. The closest living relatives of land plants. Science (2001) 294:2351–2353.
Knowles L. L. Tests of Pleistocene speciation in montane grasshoppers (genus Melanoplus) from the Sky Islands of western North America. Evolution (2000) 54:1337–1348.[CrossRef][Web of Science][Medline]
Kuijt J. The biology of parasitic flowering plants (1969) Berkeley, CA, USA: University of California Press.
Larget B., Simon D. L. Markov chain Monte Carlo algorithms for the Bayesian analysis of phylogenetic trees. Mol. Biol. Evol. (1999) 16:750–759.[Web of Science]
Lee K. B., Lee C. D. The structure and development of the haustorium in Cuscuta australis. Canadian J. Bot. (1989) 67:2975–2982.[CrossRef]
Li Y. Parasitism and integrated control of dodder of soybean. Pages 497–500 in Parasitic flowering plants—Weber H. C., Forstreuter W., eds. (1987) Marburg, Germany: Philipps-Universität.
Mabberley D. J. The plant book (1987) Cambridge, UK: Cambridge university press.
Machado M. A., Zetsche K. A structural, functional, and molecular analysis of plastids of the holoparasites Cuscuta reflexa and Cuscuta europaea. Planta (1990) 181:91–96.[Web of Science]
Maddison D. R. The discovery and importance of multiple islands of most-parsimonious trees. Syst. Zool. (1991) 40:315–328.[Abstract]
Maddison W. P., Maddison D. R. MacClade: Analysis of phylogeny and character evolution, version 3.07 (1997) Sunderland, MA, USA: Sinauer.
Melchior H. Engler's Syllabus der Pflanzenfamilien (1964) 12:427–429.
Moore W. S. Inferring phylogenies from mtDNA variation: Mitochondrial-gene trees versus nuclear-gene trees. Evolution (1995) 49:718–726.[CrossRef][Web of Science]
Murphy W. J., Eizirik E., O'Brien S. J., Madsen O., Scally M., Douady C. J., Teeling E., Ryder O. A., Stanhope M. J., W de Jong W., Springer M. S. Resolution of the early placental mammal radiation using Beayesian phylogenetics. Science (2001) 294:2348–2351.
Neyland R. A phylogeny inferred from large ribosomal subunit (26S) rDNA sequences suggests that Cuscuta is a derived member of Convolvulaceae. Brittonia (2001) 53:108–115.[Web of Science]
Nickrent D. L., Duff R. J., Colwell A. E., Wolfe A. D., Young N. D., Steiner K. E., dePamphilis C. W. Molecular phylogenetic and evolutionary studies of parasitic plants. In: Molecular Systematics of Plants II: DNA sequencing—Soltis D. E., Soltis P. S., Doyle J. J., eds. (1998) Boston, MA, USA: Kluwer. 211–241. Pages.
Nickrent D. L., Starr E. M. High rates of nucleotide substitution in nuclear small-subunit (18S) rRNA from holoparasitic flowering plants. J. Mol. Evol. (1994) 39:62–70.[Web of Science][Medline]
Nickrent D. L., Blarer A., Qiu Y.-L., Soltis D. E., Soltis P. S., Zanis M. Molecular data place Hydnoraceae with Aristolochiaceae. Am. J. Bot. (2002) 89:1809–1817.
Nickrent D. L., Duff R. J. Molecular studies of parasitic plants using ribosomal RNA. Pages 28–52 in Advances in parasitic plant research—Moreno, ed. (1996) Spain: Dirección General de Investigación Agraria, Servicio de Publicaciones y Divulgación, Córdoba.
Olmstead R. G., de Pamphilis C. W., Wolfe A. D., Young N. D., Elisons W. J., Reeves P. A. Disintegration of the Scrophulariaceae. Am. J. Bot. (2001) 88:348–361.
Olmstead R. G., Michaels H. J., Scott K. M., Palmer J. D. Monophyly of the Asteridae and identification of their major lineages inferred from DNA sequences of rbcL. Ann. Mo. Bot. Gard. (1992) 79:249–265.[CrossRef]
Olmstead R. G., Palmer J. D. A chloroplast DNA phylogeny of the Solanaceae: Subfamilial relationships and characters evolution. Ann. Mo. Bot. Gard. (1992) 79:346–360.[CrossRef]
Oxelman B., Bremer B. Discovery of paralogous nuclear gene sequences coding for the second-largest subunit of RNA polymerase II (RPB2) and their phylogenetic utility in Gentianales of the asterids. Mol. Biol. Evol. (2000) 17:1131–1145.
Oxelman B., Yoshikawa N., McConaughy B. L., Luo J., Denton A. L., Hall B. D. RPB2 gene phylogeny in flowering plants, with particular emphasis on asterids. In: Mol. Phylogenet. Evol. (2004) In press.
Peter A. Convolvulacea—Engler A., Prantl K., eds. (1891) 4. Die Natürlichen Pflanzenfamilien. 1–40.
Peter A. Convolvulaceae—Engler A., Prantl K., eds. (1897) 4. Die Natürlichen Pflanzenfamilien. 375–377. (Appendix).
Philippe H. MUST: A computer package of management utilities for sequences and trees. Nucleic Acids Res. (1993) 21:5264–5272.
Posada D., Crandall K. A. ModelTest: Testing the model of DNA substitution. Bioinformatics (1998) 14:817–818.
Rambault A., Grassly N. C. Seq-Gen: An application of Monte Carlo simulation of DNA sequence evolution along phylogenetic trees. Comput. Appl. Biosci. (1997) 13:235–238.
Rannala B., Yang Z. Probability distribution of molecular evolutionary trees: A new method of phylogenetic inference. J. Mol. Evol. (1996) 43:304–311.[Web of Science][Medline]
Renner S. S., Chanderbali A. S. What is the relationship among Hernandiaceae, Lauraceae, and Monimiaceae, and why is this question so difficult to answer? Int. J. Plant Sci. (2000) 161:S109–S119.[CrossRef][Web of Science]
Roberty G. Genera Convolvulacearum. Candollea (1952) 14:11–60.
Roberty G. Les genres des Convolvulacées (esquisse. Boissiera (1964) 10:129–156.
Rogers S. O., Bendich A. J. Extraction of DNA from milligram amounts of fresh, herbarium and mummified plant tissues. Plant Mol. Biol. (1985) 5:69–76.[CrossRef][Web of Science]
Rohwer J. G. Toward a phylogenetic classification of the Lauraceae: Evidence from mat K sequences. Syst. Biol. (2000) 25:60–71.
Ruedi M., Auberson M., Savolainen V. Biogeography of Sulawesian shrews: Testing for their origin ith a parametric bootstrap on molecular data. Mol. Phylogenet. Evol. (1998) 9:567–571.[CrossRef][Web of Science][Medline]
Saitou N. A theoretical study of the underestimation of branch lengths by the maximum parsimony principle. Syst. Zool. (1989) 38:1–6.[Abstract]
Sanderson M. J., Kim J. Parametric Phylogenies? Syst. Biol. (2000) 49:817–829.
Shimodaira H. An approximately unbiased test of phylogenetic tree selection. Syst. Biol. (2002) 51:492–508.
Shimodaira H., Hasegawa M. Multiple comparisons of log-likelihoods with applications to phylogenetic inferrence. Mol. Biol. Evol. (1999) 16:1114–1116.[Web of Science]
Smith R. A., Ferguson D. M., Barkman T. J., de Pamphilis C. W. Molecular phylogenetic evidence for the origin of Lennoaceae: A case of adelphoparasitism in the angiosperms? Am. J. Bot. S (2000) 87:158.
Soltis D. E., Soltis P. S., Nickrent D. L., Johnson L. A., Hahn W. J., Hoot S. B., Sweere J. A., Kusoff R. K., Kron K. A., Chase M. W., Swenson S. M., Zimmer E. A., Chaw S.-M., Gillespie L. J., Kress W. J., Sytsma K. J. Angiosperm phylogeny inferred from 18S ribosomal DNA sequences. Ann. Mo. Bot. Gard. (1997) 84:1–49.[CrossRef]
Stefanovi S., Krueger L., Olmstead R. G. Monophyly of the Convolvulaceae and circumscription of their major lineages based on DNA sequences of multiple chloroplast loci. Am. J. Bot. (2002) 89:1510–1522.
Subramaniam K., Mahadevan S. The cDNA sequence of cytochrome b5 associated with cytokinin-induced haustoria formation in Cuscuta reflexa. Gene (1994) 149:375–376.[CrossRef][Web of Science][Medline]
Swofford D. L. PAUP*. Phylogenetic analysis using parsimony (*and other methods), version 4.0b2a (1999, 2002) Sunderland, MA: Sinauer.
Swofford D. L., Olsen G. J., Waddell P. J., Hillis D. M. Phylogenetic inference. Pages 407–514 in Molecular systematics—Hillis D. M., Moritz C., Mable B. K., eds. (1996) nd ed. Sunderland, MA: Sinauer.
Taberlet P., Gielly L., Pautou G., Bouvet J. Universal primers for amplification of the noncoding regions of chloroplast DNA. Plant Mol. Biol (1991) 17:1105–1109.[CrossRef][Web of Science][Medline]
Takhtajan A. Diversity and classification of flowering plants (1997) New York, NY, USA: Columbia University Press.
Wolfe A. D., dePamphilis C. W. Systematic implications of relaxed functional constraints on the RUBISCO large subunit in parasitic plants of the Scrophulariaceae and Orobanchaceae. Am. J. Bot. S (1995) 82:6.
Wolfe A. D., dePamphilis C. W. Alternate paths of evolution for the photosynthetic gene rbcL in four nonphotosynthetic species of Orobanche. Plant Mol. Biol. (1997) 33:965–977.[CrossRef][Web of Science][Medline]
Wolfe A. D., dePamphilis C. W. The effect of relaxed functional constraints on the photosynthetic gene rbcL in photosynthetic and nonphotosynthetic parasitic plants. Mol. Biol. Evol. (1998) 15:1243–1258.[Abstract]
Yang Z. Estimating the pattern of nucleotide substitution. J. Mol. Evol. (1994) 39:105–111.[Web of Science][Medline]
Young N. D., Steiner K. E., dePamphilis C. W. The evolution of parasitism in Scrophulariaceae/Orobanchaceae: Plastid gene sequences refute an evolutionary transition series. Ann. Mo. Bot. Gard. (1999) 86:876–893.[CrossRef]
Yuncker T. G. The genus Cuscuta. Memoirs of the Torrey Botanical Club (1932) 18:113–331.
Zanis M. J., Soltis D. E., Soltis P. S., Mathews S., Donoghue M. J. The root of the angiosperms revisited. Proc. Natl. Acad. Sci. U.S.A. (2002) 99:6848–6853.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||