Molecular Studies of Parasitic Plants Using Ribosomal rna
|
|  Yüklə 155.5 Kb.
|
| Recent sequencing efforts have greatly expanded the number of available 18S rDNA sequences in plants (Nickrent and Soltis, 1995; Soltis et al., submitted.). At present, over 400 complete 18S sequences exist for angiosperms, i. e. sufficient density to avoid erroneous topologies stemming from insufficient taxon sampling (Chase et al., 1993). This large dataset has allowed a re-examination of the placement of the above three holoparasite families. The results presented below should be considered extremely preliminary since no statistical tests (e.g. bootstrap analyses) have yet been conducted.
18S rDNA sequences from 223 angiosperms representing all major lineages were obtained (Soltis et al., submitted) and aligned. The following holoparasites were then added: Balanophora fungosa J. R. & G. Forst., Corynaea crassa Hook. f., Cynomorium coccineum Mich., Helosis cayennensis (Schwartz) Spreng., Ombrophytum subterraneum (Asplund) B. Hansen, Rhopalocnemis phalloides Jungh., and Scybalium jaimaicense (Sw.) Schott. & Endl. (Balanophoraceae), Cytinus ruber Fritsch, Rafflesia keithii Meijer, and Rhizanthes zippelii (Blume) Spach (Rafflesiaceae s. lat.), and Hydnora africana Thunb. and Prosopanche americana (R. Br.) Baillon (Hydnoraceae) for a total of 245 sequences. Given the size of this dataset, only heuristic search strategies could be used. Since branch swapping did not go to completion, the tree discussed below is not optimal (minimum length). The tree resulting from this analysis retained the major topological features found in the more extensive analysis reported in Soltis et al. (submitted). For example, the clade composed of Santalales (strict sense) was allied with Polygalaceae and Fabaceae within a large group composed of Rosid and Dilleniid taxa. The placement of Santalales as sister to Polygalales is of interest since parasitism has also evolved in one family of the latter (Krameriaceae). None of the holoparasite families (Balanophoraceae, Hydnoraceae, and Rafflesiaceae) were associated with the santalalean clade. The two genera of Hydnoraceae, Prosopanche and Hydnora, were placed as sister taxa in a clade composed of several families referred to as paleoherbs (sensu Donoghue and Doyle, 1989), i.e. Aristolochiaceae, Chloranthaceae, Lactoridaceae, Piperaceae, and Saruraceae. Evidence from several molecular and morphological datasets supports the basal position of the paleoherbs within the angiosperms. It is of interest that the relationship between Hydnoraceae and Aristolochiaceae is fully concordant with the concepts proposed by Solms-Laubach (1894) and Harms (1935). More recently, Cocucci (1983) used embryological and floral characters to derive Aristolochiaceae, Rafflesiaceae, and Hydnoraceae from annonaceous stock. The molecular analyses reported here provide additional evidence linking Hydnoraceae to magnoliid ancestors. When analyzed together, Rafflesiaceae and Balanophoraceae exhibit artifactual long-branch attractions. The clade composed of Rafflesia and Rhizanthes is intercalated in the middle of Balanophoraceae, an unlikely relationship. In this analysis, this composite clade was placed with taxa allied with Saxifragales, an equally unlikely relationship. The 18S rDNA analysis does not support a relationship between Cytinus and Rafflesia/Rhizanthes. These taxa were separated at the tribal level by Harms (1935) and, on the basis of pollen morphology, Cytinus has been segregated to Cytinaceae by Takhtajan et al. (1985). The classification of Cytinus in its own family is concordant with results from these molecular analyses that place it in Rosidae. These results indicate that Rafflesiaceae sensu latu is a paraphyletic group composed of at least two and possibly three distinct families. Preliminary molecular analyses of 18S rDNA sequences indicates that the small-flowered Rafflesiaceae (Apodanthes, Bdallophyton, Mitrastemma, and Pilostyles - Berlinianche yet to be sampled) are not closely related to the large-flowered members (Rafflesia and Rhizanthes - Sapria yet to be sampled). Since 18S rDNA analyses strongly suggest placement of Hydnoraceae in Magnoliidae, and since all the major angiosperm classifications have placed Rafflesiaceae and Hydnoraceae together in Rafflesiales, it is tempting to follow tradition and place Rafflesiaceae in Magnoliidae, despite lack of direct support from molecular data. The position of Cytinus in Rosidae counteracts this inclination since this family may provide a link to the more derived members of the family. All systematic discussion must be considered speculative at this point since the extreme rate heterogeneity displayed by these plants is clearly affecting their position relative to slower-rate plants. Despite long-branch attractions, relationships among the seven genera of Balanophoraceae can be addressed. When Balanophoraceae are analyzed without Rafflesiaceae, using the gnetophytes Ephedra and Gnetum as outgroups, the phylogram shown in Fig. 4 results. One clade is composed of the New World genera Corynaea, Helosis, Ombrophytum and Scybalium and the other clade of the Old World genera Balanophora, Rhopalocnemis, and Langsdorffia. Langsdorffia occurs basal to the Old World branch, a position concordant with biogeographic evidence since it is the only genus in the family present in both the Old and New World (Central and South America and Madagascar). These relationships are partly in accord with the classification of Harms (1935) who proposed six subfamilies. Contrary to that classification, where the Indonesian Rhopalocnemis was placed with the New World genera in Helosidoideae, rDNA sequence data show it is very closely related to Balanophora. The most distantly related component of this clade is Cynomorium. Given the amount and types of mutations present in 18S rDNA from this genus and other Balanophoraceae indicate it is distinct. For this reason, segregation at the family level as proposed by Tahktajan (1987) and Thorne (1992) is favored. Molecular Phylogenetic Studies of Scrophulariales Within Scrophulariales, two families are traditionally defined that contain parasitic members: Scrophulariaceae and Orobanchaceae. These families contain greater than 45 genera and 1650 parasitic species and also exhibit the entire range of trophic conditions found in parasitic plants (hemiparasites to holoparasites). The traditional concept maintaining these two families as separate is not being followed in recent morphologically-based (e.g. Takhtajan, 1987; Minkin and Eshbaugh, 1989; Thorne, 1992; Judd et al., 1994) and molecularly-based (e.g. Olmstead and Reeves, 1995; dePamphilis, 1995 ) classifications. The study by Olmstead and Reeves (1995) utilized sequences from the chloroplast genes rbcL and ndhF (separately and in combination); however, they did not include any members of the hemiparasitic rhinanthoid Scrophulariaceae or Orobanchaceae. Wolfe and dePamphilis (1995) have generated over 35 new rbcL sequences for these families and have discovered that intact rbcL genes are present in some nonphotosynthetic species. This contrasts with the condition in Epifagus (Orobanchaceae) where rbcL is present as a pseudogene (434 bp vs. 1452 bp in length). They also observed that pseudogene formation can occur independently within a genus (Orobanche). A second chloroplast gene, rps2 (ribosomal small-subunit 2) has been sequenced and analyzed by dePamphilis (1995). Sampling in this study included parasitic and nonparasitic Scrophulariaceae, Orobanchaceae and several additional taxa from Scrophulariales. Results from this analysis showed that: 1) Orobanchaceae is derived from within Scrophulariaceae making the latter family paraphyletic, 2) taxa traditionally placed in other families are related to Scrophulariaceae (e.g. Bignoniaceae, Verbenaceae), and 3) parasitism arose just once within this lineage whereas loss of photosynthesis has occurred on many independent occasions. The rps2 tree was not fully resolved, however, some relationships received strong support such as a clade containing Epifagus, Conopholis and Boschniakia, a second with Harveya and Hyobanche, a third with two species of Orobanche, and a fourth with Pedicularis and Castilleja. Lathraea is the last member to emerge from this large polytomy. Colwell (1994) conducted phylogenetic analyses of hemiparasitic Scrophulariaceae and holoparasitic Orobanchaceae using nuclear 18S rDNA sequences. Although this molecule usually provides insufficient characters for analysis at the rank of family and below, increased rates of nucleotide substitution in Orobanchaceae provided enough variation to address relationships at this level. Colwell (1994) showed results of analyses of nine ingroup (parasitic) plants, however, subsequent sequencing (Colwell and Nickrent, unpublished) has increased this number to 18. Phylogenetic analysis was conducted using 18 ingroup 18S rDNA sequences (Scrophulariaceae and Orobanchaceae) and three outgroup sequences (Glycine, Lycopersicon and Ipomoea). Two equally parsimonious minimum-length trees were obtained and the consensus is shown in Fig. 5. A clade containing Linaria and Chamophila is sister to the remaining Scrophulariaceae plus Orobanchaceae suggesting (as with rps2, above) that parasitism arose just once. The topology of this tree is fully concordant with that obtained with rps2 and in fact provides further resolution of relationships. The hemiparasites Pedicularis, Orthocarpus and Castilleja form a clade separate from the holoparasites. Three genera have been placed in either Scrophulariaceae or Orobanchaceae by different taxonomists: Harveya, Hyobanche and Lathraea. This 18S rDNA analysis shows these three genera to be components of a clade containing three species of Orobanche. Although taxon sampling is still incomplete, it appears that members of this clade are undergoing accelerated rates of nucleotide substitution compared with their relatives. The 18S tree also indicates that the genus Orobanche is monophyletic, yet it is not closely related to Epifagus, Conopholis and Boschniakia thus making Orobanchaceae paraphyletic. This study provides evidence that 18S rDNA can be used to address phylogenetic questions in Orobanchaceae. Of the 15 "nontransitional" genera classified in this family, only four have been sampled, hence the inclusion of more Old World representatives is needed to fully address the remaining questions. General Features of Parasite Plastid Genomes Among the three distinct genomes in plant cells, the circular plastid genome is certainly the best characterized in terms of structure and function (Sugiura, 1992). Green plants have from one to several hundred chloroplasts per cell and each chloroplast may contain from 7 to greater than 200 plastid genomes (Maguire et al., 1995). The complete chloroplast DNA molecule has been sequenced for Marchantia (Ohyama et al., 1986), Pinus (Wakasugi et al., 1994), Oryza (Hiratsuka et al., 1989), Nicotiana (Shinozaki et al., 1986) and Epifagus (dePamphilis and Palmer, 1990). The cpDNAs of these plants vary widely in size: 119 kb in pine, 121 kb in liverwort, 134 kb in rice, and 156 kb in tobacco. The genome of Epifagus virginiana (L.) Bart. (beechdrops) is only 71 kb owing to extensive losses of photosynthetic genes that have accompanied its loss of photosynthesis (dePamphilis and Palmer, 1990). Epifagus has a large single copy region (LSC) of 18 kb (vs. 87 kb in tobacco), a small single copy region (SSC) of 3.6 kb (vs. 18.5 kb) that contains only two genes (Wolfe, et al., 1992), but strangely the inverted repeat regions are present and full sized (25 kb). Of the 42 genes remaining in Epifagus, 38 are involved in protein synthesis (e.g. rRNA, tRNA, ribosomal protein genes), yet expression relies upon import of nuclear-encoded tRNAs and RNA polymerase (Morden et al., 1991). The selective maintenance of these specific genes (versus loss of nearly all photosynthetic genes) provided inferential evidence of their functionality, but direct evidence was obtained by Ems et al. (1995) using Northern blot analyses. Conopholis americana (L.) Wallr. (squawroot), another holoparasite in Orobanchaceae, has also been the subject of molecular studies. Using heterologous probes, Wimpee et al. (1991) documented the modification or absence of many photosynthetic genes. Colwell (1994) conducted restriction site mapping of the plastid genome of squawroot documenting its size as 43 kb, i.e. the smallest ptDNA molecule yet observed in plants. Much of this reduction is due to the loss of one copy of the inverted repeat. In addition to parasitic Scrophulariaceae and Orobanchaceae, molecular genetic investigations have also been focused upon Cuscuta. This genus, sometimes placed in its own family (Cuscutaceae), is widely recognized to share a common ancestor with nonparasitic Convolvulaceae. As shown in Fig. 1, Cuscuta (like Scrophulariaceae), includes hemiparasitic species (e.g. C. reflexa Roxb.) as well as holoparasitic species (e.g. C. europaea L.) that lack thylakoids, chlorophyll, Rubisco and light-dependent CO2 fixation but (strangely) retain rbcL (Machado and Zetsche, 1990). Although the plastid genome of Cuscuta is yet to be fully sequenced, significant progress is being made. Bömmer et al. (1993). cloned and sequenced a 9 kb portion of ptDNA from C. reflexa that includes 16S rDNA, psbA, trnH, ORF 740, ORF 77, trnL, and ORF 55. Later, Haberhausen and Zetsche (1994) cloned and sequenced a 9 kb portion of ptDNA from this same species that included a large portion of inverted repeat A spanning a segment from trnA to trnH. Although some sequences were identical to Nicotiana (e.g. trnI), many deletions were observed. For example, rpl2 and rpl23 were both deleted and ORF2280 was reduced to only 740 bp. These results show that, like Epifagus, Cuscuta has experienced major deletions in the plastid genome. The complete loss of ribosomal protein genes such as rpl2 invites questions about how such the translational apparatus of the plastid functions and its relationship to the other two subcellular genomes. Plastid rRNA in Parasitic Plants In plants, the ribosomal cistrons are present on both inverted repeats (when present) and are composed of genes in the following order: 16S, trnI, trnA, 23S, 4.5S, and 5S (Fig. 2B). This arrangement is extremely conserved, yet some variation can be seen in Orobanchaceae. Conopholis retains an intact 16S and 23S rDNA, however, the intervening trnI is absent and trnA has become a pseudogene (Wimpee et al., 1992). The flanking 4.5S and 5S rDNAs are intact, but the spacers between them are only 70% of the typical length. Epifagus also retains intact 16S and 23S rDNA, but both trnI and trnA have become pseudogenes while conserving the length of the entire cistron (Wolfe et al., 1992). Prior to 1995, there were no published data available regarding the presence of a plastid genome (or genes) in Balanophoraceae, Hydnoraceae or Rafflesiaceae. Despite experiencing extensive losses in photosynthetic and other genes, both Epifagus and Conopholis have intact and functional ribosomal cistrons. Given this, it was reasoned that if any genes were present in these holoparasites, very likely they would be rDNA. The goal of these initial studies was to use PCR to amplify 16S rDNA from each of these lineages. A multiple sequence alignment was conducted on 16S rDNA using cyanobacterial outgroups as well as published plant sequences. Conserved regions were identified and oligonucleotide primers developed for PCR and sequencing. PCR was conducted using the 8 forward [GGA GAG TTC GAT CCT GGC TCA G] and the 1461 reverse [GGT GAT CCA GCC GCA CCT TCC AG] primers (numbers based upon position on tobacco). To date, 16S rDNA has been amplified from genomic DNA samples of representatives of Balanophoraceae, Hydnoraceae and Rafflesiaceae (Nickrent et al., 1995). We have also amplified 23S rDNA and the spacer region between the 16S and 23S rDNA in several taxa indicating that a large portion (if not a full) ribosomal cistron is present. Although compelling, these data do not yet prove the existence of a plastid genome since the possibility of migration to nucleus has not been excluded. A secondary structure of the 16S rRNA of Cytinus ruber (Rafflesiaceae or Cytinaceae) is presented in Fig. 6 based upon the model of Gutell et al. (1985) for Zea. This model corresponds well to previously proposed ones for cyanobacteria and green plant plastid rRNAs as evidenced by conservation of all major structural features. Mutations relative to the 16S rRNA of Nicotiana are indicated by highlighted bases. Overall, 130 single base change mutations (8.6%) can be seen on this molecule, i.e. greater than twice as many as would be seen in a comparison of Nicotiana to the more distantly related liverwort Marchantia. Similar (or higher) mutational levels were observed for the other holoparasite 16S sequences. Given the large number of changes in the Cytinus sequence, it is possible that this sequence is that of a pseudogene. To investigate the effect of each type of mutation on the maintenance of helices, mutations were identified and categorized. Fully compensated changes occur when each paired base in a helix changes but canonical or noncanonical base pairing is retained. Fully compensated changes can be found in 12 pairs representing 24 mutational events. Compensated changes, when one member of a base pair in a helix changes, yet canonical or noncanonical pairing is maintained (e.g. G–C G • U), was observed at 37 sites. Changes in unpaired regions (loops and bulges) were the largest class of mutation, constituting 60 changes. Finally, eight changes were found that disrupted a base pair present in the Nicotiana 16S rRNA secondary structure. This exercise shows that, despite frequent mutation, the majority do not result in obvious disruptions of the secondary structure. The conclusion is that the 16S rRNA in Cytinus is very likely functional, although Northern blot and/or in vitro studies are required for additional confirmation. Plastid 16S rDNA sequences have been used infrequently for inferring phylogenetic relationships in plants, owing mainly to the paucity of sufficient numbers of sequences and the overall slow evolutionary rate of this gene. The relationships among seven dicots and two monocots using 16S rRNA sequences was shown in a phylogram reported by Wolfe et al. (1992). That study documented the overall conservative nature of this molecule and the higher nucleotide substitution rates in Conopholis and Epifagus relative to other dicots. The following sequences (with Genbank accession numbers) were used in this phylogenetic analysis: Anacystis nidulans [X03538], Anabaena sp. [X59559], Microcystis aeruginosa [U03402], Marchantia polymorpha [X04465], Pinus thunbergii [D17510], Oryza sativa [X15901], Zea mays [X86563], Pisum sativum [X51598], Glycine max [X06428], Daucus carota [X73670], Cuscuta reflexa [X72584], Nicotiana tabacum [Z00044], Epifagus virginiana [X62099], and Conopholis americana [X58864]. The sequence of Brassica and Vicia were obtained from Gao et al. (1989, 1990, respectively). The 16S sequence of Cytinus ruber is available via Genbank number U47845. The above 16S rDNA sequences were aligned and subjected to cladistic analysis using PAUP (Swofford 1993). Alignment was generally unambigous and secondary structural features were used to guide alignment in variable regions. The branch and bound algorithm found a single most parsimonious minimum length tree of 808 steps. To test the statistical stability of the clades, bootstrap analysis (100 replications using a heuristic search strategy) was conducted and the resulting tree is shown in Fig. 7. The clade containing the two cyanobacteria (Anacystis and Microcystis) is separated from the green plant (chloroplast) clade by 104 substitutions. As expected, the angiosperms form a monophyletic clade, separate from the liverwort and pine, with an 83% bootstrap value. Within angiosperms, a clade containing Asteridae (plus Brassica) is recovered in 79% of the bootstrap trees. The bootstrap value for the legume clade is 70% whereas the monocots (plus Cytinus) is 77%. These two clades, however, are recovered in less than 50% of the bootstrap trees and therefore form a polytomy with the Asteridae clade (above). An unusual feature of the tree is the position of Cytinus with the monocots - clearly a case of long-branch attraction. The branch leading to Cytinus has 102 substitutions, i.e. as great as the branch separating the chloroplast clade from the cyanobacteria. Mapped onto the phylogram shown in Fig. 7 are instances where the plastid inverted repeat has been lost. This has occurred independently in several lineages, i.e. Conopholis, two legumes (Pisum and Vicia but not Glycine) and very likely Cytinus. It is of interest to note that in the legumes, 47 and 34 mutations have occurred since the divergence of Vicia and Pisum (respectively) from their common ancestor with Glycine. These numbers are similar to values obtained when deriving Epifagus and Conopholis from their common ancestor. Strangely, only Conopholis has lost the inverted repeat, not Epifagus. It is possible that the presence of the inverted repeat imparts a stabilizing effect on the dynamics of plastid evolution and that increased rates are seen when this effect is removed. The presence of similar numbers of substitutions for Conopholis and Epifagus may indicate that insufficient time has elapsed for the appearance of major differences. Molecular Evolutionary Studies - Relative Rate Tests The molecular phylogenetic approach to reconstructing evolutionary relationships is closely tied to the concept of a molecular clock (Zuckerkandl and Pauling, 1965). The strict molecular clock infers that for organisms with equal generation times, their respective lineages will have equal rates of nucleotide substitution. For plants, a strict molecular clock cannot be applied universally to all clades and to all gene loci. The generation–time hypothesis is the most frequently cited cause for rate invariance (Bosquet et al., 1992); however, this explanation is often difficult to apply to plants and may be only partially responsible for the phenomenon (Nickrent and Starr, 1994). The most dramatic cases of rate invariance in plants have been detected following relative rate tests using nuclear 18S rDNA sequences from parasitic plants. Five parasitic flowering plants showed unusually high substitution rates, yet these lineages do not have fast generation times (Nickrent and Starr 1994). It was hypothesized that the underlying cause of such divergent rDNA sequences was relaxation of selectional constraints on rRNA structure and function, possibly as a result of small effective population size and molecular drive. These processes, however, are occurring in other eukaryotes that do not show increased substitution rates, hence they do not provide a universal explanation. Accelerated substitution rates are not restricted to a single parasitic plant clade (cf. Viscaceae, Balanophoraceae, Rafflesiaceae) nor are they found universally within a single clade (cf. Viscaceae with Loranthaceae). It appears that rate acceleration of nuclear rDNA occurs only in lineages that exhibit a reduction in photosynthesis and an advanced state of nutritional dependence upon the host. It is not presently understood why these changes in life history are correlated with changes at nuclear rRNA loci. Studies are currently underway (Colwell and Nickrent, in prep.) comparing evolutionary rates of 18S rDNA in all the major parasite and mycotroph lineages. |