Introduction

The plastid genome of land plants is highly conserved in size, structure, gene content, and synteny (Palmer 1985, 1991; Palmer and Stein 1986; Downie et al. 1991, 1994; Downie and Palmer 1992). Any changes may be important phylogenetic markers (Palmer 1985, 1990), but perhaps more importantly, they may also be used to study molecular evolutionary and genetic processes involving the structure, function, and evolution of the plant plastid genome (Palmer 1985, 1990, 1991; dePamphilis and Palmer 1990).

The morning-glory family, Convolvulaceae, is one of the few angiosperm families exhibiting substantial plastid genome rearrangement (Downie and Palmer 1992). There is evidence from a small number of sampled species indicating structural differences in the Convolvulaceae plastid DNA (ptDNA) relative to other flowering plants. For example, the rpl2 gene is interrupted by an intron in most but not all angiosperms. Based on plastid genome sequences of diverse land plants, including one liverwort (Ohyama et al. 1986), one gymnosperm (Wakasugi et al. 1994), two basal angiosperms (Goremykin et al. 2003, 2004), five eudicots (Shinozaki et al. 1986a; Spielmann et al. 1988; Sato et al. 1999; Hupfer et al. 2000; Kato et al. 2000), and five monocots (Posno et al. 1986; Hiratsuka et al. 1989; Maier et al. 1995; Ogihara et al. 2002; Stefanovic’ et al. 2004), it became evident that this intron was present in the common ancestor of flowering plants (and probably land plants), but was subsequently lost in some angiosperm lineages (e.g., Spinacia oleracea; Zurawski et al. 1984; Sehmitz-Linneweber et al. 1999). In their large-scale filter hybridization survey across angiosperms, Downie et al. (1991) reported at least six independent losses of the rpl2 intron, including in the ancestor of two representatives of green Convolvulaceae and one Cuscuta species. Outside of Convolvulaceae, no other member of the Asteridae was found to lack this intron. In another example, Downie and Palmer (1992) reported the loss of an open reading frame (ORF). ycf1 (= ORF1244 in Nicotiana tabacum), of unknown function, but common to plastid genomes of land plants in three nonparasitic Convolvulaceae and one Cuscuta species (as well as in Poaceae, Fabaceae, and some other families not closely related to Convolvulaceae). Given the conserved nature and distribution of this ORF across the land plants, it is most parsimonious to conclude that ycf1 was lost several times independently within angiosperms, similar to the rpl2 intron. A third example is the atpB gene which, like rbcL, is very conserved in its length across all angiosperms (except at the very end of the gene). However, in the two members of green Convolvulaceae for which sequences were known, an in-frame deletion encoding two amino acids has been reported at the 5′ end of this gene (Savolainen et al. 2000).

Cuscuta, the only parasitic genus associated with Convolvulaceae, has been the subject of more extensive molecular analyses than the rest of this family (Machado and Zetche 1990; Haberhausen et al. 1992; Bömmer et al. 1993; Haberhausen and Zetsche 1994; van der Kooij et al. 2000; Krause et al. 2003). Most major synoptic works on flowering plants (e.g., Cronquist 1988; Takhtajan 1997) accept Cuscuta at the family level, thus implying that it is only distantly related to other species in Convolvulaceae, but recent phylogenetic results indicate it belongs within Convolvulaceae (Stefanovic’ et al. 2002; Stefanovic’ and Olmstead 2004). Cytological, anatomical, and morphological characters indicate the presence of three well-defined groups in Cuscuta and classification in three subgenera (Cuscuta, Grammica, and Monogyna) was proposed by Engelmann (1859; formalized by Yuncker 1932). Members of this genus are characterized by twining, slender, pale stems, with reduced, scale-like leaves, no roots, and are attached to the host by haustoria, therefore depending entirely or almost entirely on their hosts to supply water and nutrients (Kuijt 1969; Dawson et al. 1994). Many Cuscuta species are characterized also by the reduced amounts or absence of chlorophylls (van der Kooij et al. 2000). However, some species produce significant amounts of chlorophylls, especially in the tips of seedlings not attached to the host and in fruiting sepals and ovaries (Panda and Choudhury 1992; Dawson et al. 1994). Thus, both holoparasitic and hemiparasitic species occur in this genus. In holoparasites, found mostly in subgenera Cuscuta and Grammica, all carbohydrates are provided by the host. In herniparasites, confined predominantly to subgenus Monogyna, carbohydrates supplied by hosts are supplemented, to a small extent, by the plant’s own photosynthesis (Hibberd et al. 1998). This diversity of photosynthetic ability among Cuscuta species initiated several physiological studies of photosynthetic enzymes ( Machado and Zetche 1990; van der Kooij et al. 2000) and molecular evolution studies of the plastid genome (Haberhausen et al. 1992; Bömmer et al. 1993; Haberhausen and Zetsche 1994; Krause et al. 2003).

Even though the plastids of Cuscuta reflexa (subgenus Monogyna) have no visible grana and the number of thylakoids is highly reduced compared to normal plastids, Machado and Zetche (1990) showed the presence of both chlorophyll a and b, detected residual light- and CO2-dependent photosynthetic activity, and demonstrated incorporation of 14CO2 into carbohydrates. In contrast, the plastids of C. europaea (subgenus Cuscuta) lack both grana and thylakoids and appear to lack chlorophylls completely. Consequently, neither ribulose-1,5-bisphosphate carboxylase-oxygenase (Rubisco) activity nor light-dependent CO2 fixation could be detected (Machado and Zetche 1990). Further studies, focusing mainly on species of subgenus Grammica, demonstrated that in most cases the rbcL gene, thylakoids, and chlorophylls are present, and low amounts of the large subunit of Rubisco could be detected immunologically (van der Kooij et al. 2000). However, two species from the same subgenus have been shown to lack thylakoids and chlorophylls, and neither the rbcL gene nor its protein product, Rubisco large subunit, could be detected (van der Kooij et al. 2000).

The ptDNA sequence of Cuscuta reflexa, a member of the predominantly hemiparasitic subgenus Monogyna, has been studied in most detail from a molecular evolutionary standpoint. Results indicate that this species retains an affected, yet functional, plastid genome (Haberhausen et al. 1992; Bömmer et al. 1993; Haberhausen and Zetsche 1994). Epifagus virginiana (Orobanchaceae, Lamiales) is the only parasitic plant species for which the entire plastid genome has been mapped (dePamphilis and Palmer 1990) and subsequently sequenced (Wolfe et al. 1992). In contrast to this holoparasitic species, which lacks all photosynthetic and most putatively chlororespiratory ndh genes (dePamphilis and Palmer 1990). C. reflexa 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, chlororespiratory (ndh) genes seem to be either altered to the point of becoming pseudogenes (e.g., ndhB) or are lost from the plastid genome (Haberhausen and Zetsche 1994). In addition, a large deletion (∼6.5 kb) at the junction between the inverted repeat (IR) and the large single-copy (LSC) region, comprising two ribosomal protein genes, rpl2 and rpl23, one tRNA (trnI-CAA), as well as a large portion of the ycf2 gene (= ORF2280 in tobacco), has been reported in C. reflexa (Bömmer et al. 1993). Given this sequence information and the absence of detectable Southern hybridization to total C. reflexa cellular DNA using tobacco probes for affected genes, Bömmer et al. (1993) postulated the total loss of rpl2 and rpl23 and hypothesized further that the whole translation apparatus might be nonfunctional in this Cuscuta species.

Utilizing a combination of heterologous and homologous Southern hybridization, in conjunction with PCR amplification using internal primers, Krause et al. (2003) showed evidence for parallel loss of the rpoA and rpoB genes coding for the plastid-encoded RNA polymerase (PEP) in three holoparasitic Cuscuta species from subgenus Grammica. These genes were, however, retained as ORFs in hemiparasitic C. reflexa, and one examined species of autotrophic Convolvulaceae showed strong hybridization signal for both genes under investigation (Krause et al. 2003).

Most of these findings of plastid genome structural rearrangements in Cuscuta were attributed to its parasitic lifestyle, but without comparison to related nonparasitic members of the family. Also, sequence analysis of Cuscuta species other than C. reflexa are quite limited (but see Freyer et al. 1995; Krause et al. 2003). The plastid genome of C. europaea (subgenus Cuscuta) was shown to retain an rbcL ORF, as well as plastid-encoded psbA and nucleus-encoded rbcS genes, although this species contains no chlorophyll and is unable to photosynthesize (Machado and Zetsche 1990). In addition, more extensive ptDNA sequences from species belonging to the subgenus Grammica which could have potentially the most affected plastid genome (van der Kooij et al. 2000; Krause et al. 2003), have not been reported.

Collectively, all of these observations make Convolvulaceae one of the very few angiosperm families showing substantial structural variation in the plastid genome. This is even more intriguing considering that the sister family. Solanaceae, has virtually no rearrangements and represents a “typical” chloroplast genome (Olmstead and Palmer 1992; Sugiura 1992). However, in order to accurately trace the sequence of events in plastid molecular evolution within Convolvulaceae, a reconstruction of phylogenetic relationships among parasitic and autotrophic members of the family is necessary. Convolvulaceae have been the subject of only two family-wide molecular phylogenetic studies (Stefanovic’ et al. 2002; Stefanovic’ and Olmstead 2004). In aggregate, these two studies used seven genes drawn from all three plant genomes (five chloroplast, one mitochondrial, and one nuclear) with taxonomic sampling ranging from 32 to 109 species, representing the diversity of both parasitic and nonparasitic Convolvulaceae. Cuscuta was shown to be nested well within Convolvulaceae with at least two autotrophic lineages diverging before Cuscuta. However, the exact sister group of Cuscuta could not be ascertained, even though many alternatives were tested and rejected with confidence (Stefanovic’ et al. 2002; Stefanovic’ and Olmstead 2004).

The purpose of the present study is four-pronged: (1) to introduce new ptDNA sequence data for the underinvestigated Cuscuta species belonging to the holoparasitic subgenus Grammica, (2) to summarize the patterns of ptDNA variation observed in sequence data drawn from a large sample of heterotrophic and autotrophic species of Convolvulaceae, (3) to offer an interpretation for newly obtained data and for data obtained in previous studies in light of the current estimate of phylogenetic relationships within Convolvulaceae, and (4) to investigate implications of observed ptDNA variation on the molecular processes of plastid genome evolution in Convolvulaceae.

Materials and Methods

Seed of Cuscuta sandwichiana Choisy (subgenus Grammica) was obtained from a herbarium specimen deposited at the WTU herbarium (Degener & Degener 36596, collected in 1984 in Hawaii, USA), and were germinated in the greenhouse. The seedlings were grown on Beta vulgaris or Coleus species as host plants. Total genomic DNA was isolated from fresh C. sandwichiana tissue by the modified CTAB procedure (Doyle and Doyle 1987), followed by ultracentrifugation in CsCl-ethidium bromide gradient. DNA samples from other Cuscuta species and from other Convolvulaceae, obtained for phylogenetic studies, were also used (for extraction and voucher information see Stefanovic’ et al. 2002 and Stefanovic’ and Olmstead 2004).

A number of long polymerase chain reaction (PCR) amplifications were designed to amplify extensive portions of C. sandwichiana plastid genome. Also, a series of PCR experiments was designed to assay for two types of plastid rearrangements previously reported in Cuscuta and/or Convolvulaceae, namely the loss of genes and/or introns and the presence of large inversions. Presence/absence of introns was determined by comparing the size of amplified products of organisms under investigation with those of a reference species (in most cases tobacco). Some representative products were also sequenced to find out whether the intron loss was complete or partial, and to examine its boundaries. To assay for the inversion, two pairs of forward and reverse primers were used, each pair spanning one of the putative inversion endpoints, resulting in four diagnostic amplifications (Wolfe and Liston 1998). The presence/absence of inversions is determined from the resulting pattern of successful and failed amplifications. Tobacco was used as a control.

Long PCRs were conducted using the Expand™ Long Template PCR System kit (Roche Applied Science), following instructions provided by the manufacturer. Other PCRs were conducted using Taq polymerase (Promega). Initially, sets of primers designed by Stefanovic’ et al. (2004), which cover a large portion of the plastid genome, were used for amplifications and/or sequencing. Based on these initial sequences, a number of Cuscuta-specific sequencing primers were designed and used for chromosome walking with long PCR products. Primer sequences are available upon request from SS. PCR products were separated by electrophoresis using 1% agarose gels and were visualized with ethidium-bromide. Amplified PCR products from both long and regular amplifications, intended for sequencing, were cleaned using Qiagen columns. Cleaned products were then directly sequenced using the BigDye™ Terminator cycle sequencing kit (Applied Biosystems) on an ABI 377 DNA automated sequencer (Applied Biosystems). Sequence data were edited and assembled using Sequencher™ 4.1 (Gene Codes Corporation). All newly obtained sequences reported in this study are deposited in GenBank (accession numbers AY936335-AY9336360).

Slot-blot DNA hybridizations followed standard procedures described in Doyle et al. (1995) and Adams et al. (1999). Immobilon-Ny+ membranes (Millipore) were prehybridized and hybridized at 60–62°C and filters were washed at the hybridization temperature. PCR-derived probes were labeled with 32P using random oligonucleotide primers. Autoradiography was carried out using intensifying screens at room temperature overnight (for positive control probe) or −80°C for 18–48 h (for other probes).

Results and Discussion

PtDNA Variation Within Cuscuta and Its Comparison With Tobacco Plastid Genome

Newly obtained sequence data from Cuscuta subgenus Grammica (this study) and existing sequences from Cuscuta subgenus Monogyna (Haberhausen et al. 1992; Bömmer et al. 1993; Haberhausen and Zetsche 1994) were compiled and compared with corresponding regions of Nicotiana tabacum cpDNA (Shinozaki et al. 1986a). These sequence results, together with the results from a series of experiments involving only PCR amplification, using Cuscuta species from all three subgenera, are summarized in Fig. 1.

Figure 1
figure 1

Comparison of the size, structure, gene content, and synteny in the plastid genomes of tobacco and different species of the parasitic genus Cuscuta. The newly obtained sequence data from C. sandwchiana (subgenus Grammica) and existing sequences from C. reflexa (subgenus Monogyna; Haberhausen et al. 1992; Bömmer et al. 1993; Haberhausen and Zetsche 1994) are compiled and compared with corresponding regions of tobacco chloroplast DNA (cpDNA) shown in the box between the Cuscuta genome maps. Shaded boxes depict coding regions; open boxes indicate introns (asterisks additionally denote genes that have introns), and thick dotted lines depict unsequenced regions. Genes shown above the line are transcribed from right to left; genes below the line are transcribed from left to right. Gene nomenclature follows Wakasugi et al. (1998). (A) Structural organization of tobacco cpDNA according to Wakasugi et al. (1998). The inverted repeat regions (IR A and IRB) divide the rest of the genome into large single copy (LSC) and small single copy (SSC). Arrows indicate location of the LSC and the IR regions used for detailed comparison with parasitic ptDNA as shown in B and C. (B) Comparison of C. sandwichiana sequences with the corresponding LSC regions of tobacco (∼40 kb). The ∼15-kb inversion in Cuscuta subgenus Monogyna is depicted with dashed lines as well as the lack of it in the other two Cuscuta subgenera (as determined by the PCR experiments and/or sequencing; see text for full explanation). (C) Comparison of C. sandwichiana and C. reflexa sequences with the corresponding IRA regions of tobacco (∼15 kb).

Full size image

It was demonstrated by Southern hybridization that the rpl2 intron is missing in five species belonging to tribes Ipomoeeae and Convolvuleae as well as in one species of Cuscuta (Downie et al. 1991). We have documented further a family-wide loss of this intron, using primers located in the conserved regions of the surrounding gene. Intron absence is confirmed for 30 representatives selected from throughout the Convolvulaceae (see Fig. 3), including Humbertia, for all of which the PCR bands are ∼700 bp shorter than the control (tobacco). This PCR-based determination was investigated further by DNA sequencing in order to determine the nature of the intron loss in rpl2 (i.e., partial or complete deletion, precise or imprecise intron excision). This approach confirmed the complete loss of the rpl2 intron in Convolvulaceae (see Fig. 3; see below), resulting in combination of the two exons into a single ORF, similar to the intron-deletion cases encountered in previous studies (e.g., rpoC1, clpP, rpl16, trnI; Hiratsuka et al. 1989; Zurawski et al. 1984; Downie et al. 1991).

The situation with Cuscuta is more complicated. Bömer et al. (1993) reported the complete loss of the rpl2 gene, along with rpl23, trnI, and a big portion of the ycf2 homologue in C. reflexa (=ORF2280 in tobacco). These authors argued that the translation apparatus must, therefore, be nonfunctional in this parasitic genus. However, we were able to amplify the rpl2 gene, with the same set of primers used for the nonparasitic taxa, in several Cuscuta species representing all three subgenera including C. japonica, a close relative of C. reflexa (subgenus Monogyna; Fig. 1; see also Fig. 3). These conflicting results can be reconciled in two ways: (1) either this large deletion may be a taxon-specific loss in C. reflexa, thus an autapomorphy for this species rather than a genus-wide feature linked to the parasitic lifestyle, or more likely (2) this represents a contraction of the inverted repeat (IR) in the subgenus Monogyna and possibly in the entire genus (see also Plunkett and Downie 2000). The reported deletion for C. reflexa occurs precisely at the tobacco IRA-LSC junction (i.e., J LA ) and could also be explained by a contraction of the IR A in this species. If the J LA in C. reflexa were to fall within ycf2, the region containing rpl2, rpl23, trnI, and a 5′ end of the ycf2 would still be found in the LSC region adjacent to the other IR (i.e., IRB). Given that only one junction was sequenced (J LA , as indicated by the presence of trnH, which is always neighboring the IRA in land plants), this possibility might have been overlooked. Bömer et al. (1993) reported negative Southern hybridization results of C. reflexa total cellular DNA hybridized against rpl2/rpl23 and 5′-ycf2 probes from tobacco, thus implying the total absence of these regions from all C. reflexa genomes. However, the authors failed to provide a positive control demonstrating the presence of detectable levels of C. reflexa ptDNA on these blots. Our result, demonstrating the presence of rpl2 in closely related C. japonica (subgenus Monogyna), further supports the notion of the IR contraction and retention of only one copy of rpl2 (and probably also one copy of rpl23, trnI, and ycf2 for which sequence confirmations are not available).

Haberhausen and Zetsche (1994) showed that the ndhB gene is reduced to a pseudogene in C. reflexa due to many frameshift mutations, while the other ten ndh genes are either lost or strongly altered in this species, as suggested by lack of hybridization signals using heterologous gene probes derived from tobacco. We have documented, using PCR amplification, the absence of at least one cluster of ndh genes (ndhCJK) in several other Cuscuta species, sampled from all three subgenera. The same cluster is present, however, in five representatives sampled throughout nonparasitic Convolvulaceae (Operculina aequisepala, Falkia repens, Evolvulus glomeratus, Jacquemontia tamnifolia, and Dinetus truncatus) and sequencing confirmed that these three ndh genes are present as ORFs in all five cases. In addition, the amplification of the ndhF gene, using primers both within the coding and flanking regions, failed in multiple Cuscuta species examined (using the same DNA samples that yielded PCR products with other primer combinations), while ndhF was easily amplifiable and is present as ORF in the same five above-mentioned nonparasitic Convolvulaceae species. Even though negative PCR results (i.e., the lack of amplification) cannot be explained unequivocally, this result is consistent with the one obtained by Southern hybridization (Haberhausen and Zetsche 1994).

The function of the ndh genes in the plant plastid genome was controversial in the past, but their involvement in chlororespiration now appears to be well established (dePamphilis and Palmer 1990; Sugiura 1992; Guedeney et al. 1996; Casano et al. 2000; Nixon 2000; reviewed in Pettier and Cournac 2002). The conserved nature of these genes suggests that a functional constraint is present throughout the angiosperms. Based on the genome map for the Epifagus virginiana plastid, which shows the complete loss of photosynthetic and chlororespiratory genes in this holoparasitic flowering plant, dePamphilis and Palmer (1990) proposed that the ndh genes are involved in a metabolic pathway associated with photosynthesis in autotrophic plants. However, in all Cuscuta species under investigation, including both hemi- and holoparasitic ones, most, if not all, photosynthetic genes are present in a more or less functional form, while all ndh genes are either lost or significantly altered, suggesting the decoupling of photosynthetic and chlororespiratory functions (see also Bungard 2004).

Loss of the rpoA and rpoB genes (coding for the plastid-encoded RNA polymerase; PEP) in three holoparasitic Cuscuta species from subgenus Grammica as well as retention of these genes as ORFs in one hemiparasitic species from subgenus Monogyna was demonstrated recently by Krause et al. (2003). In order to further investigate the phylogenetic extent of these findings, we conducted slot-blot hybridization on five Cuscuta species from all three subgenera (including subgenus Cuscuta, previously not assayed), as well as five autotrophic members of Convolvulaceae. In addition, our survey included probes for two remaining rpo genes, rpoC 1 and rpoC 2 . Results of the hybridization survey are presented in Fig. 2. In five nonparasitic species, hybridization signal is of similar intensity among all four rpo probes and the 16S rRNA positive control indicating a high level of similarity with tobacco rpo sequences. Likewise, Cuscuta species from subgenera Cuscuta and Monogyna (C. europaea and C. japonica, respectively) show little, if any, evidence of signal diminution. The presence of intact ORFs deduced from hybridization in these two Cuscuta species was further corroborated by complete or partial sequencing of all four rpo genes, and no evidence of frameshift deletion or premature stop codons was found. In contrast, three remaining parasitic species belonging to subgenus Grammica show a highly diminished signal for rpoA and complete absence of hybridization in three other genes (Fig. 2). PCR amplification and sequencing of rpoA region using primes in adjacent plastid genes (petD and rps11) revealed the presence of rpoA pseudogenes in C. sandwichiana and C. gronovii, while multiple primer combinations in each of three remaining rpo genes failed to produce amplicons in these species, consistent with the negative hybridization results. Together with previously published data (Krause et al. 2003), our hybridization survey of rpo genes indicates that loss of these genes is concerted and within Convolvulaceae confined only to Cuscuta species belonging to subgenus Grammica.

Figure 2
figure 2

Autoradiographs showing slot-blot hybridization results of four probes derived from rpo genes to DNAs from ten Convolvulaceae species. Small plastid ribosomal subunit (16S rRNA) was used as positive control. Species 1–5 are from parasitic genus Cuscuta (1–3 belong to subgenus Grammica, 4 to subgenus Cuscuta, and 5 to subgenus Monogyna), whereas species 6–10 represent diverse autotrophic members of the family. Note that the absence of rpoB, rpoC 1 , and rpoC 2 hybridization as well as significant signal intensity diminution for rpoA are restricted to three Cuscuta subgenus Grammica representatives.

Full size image

A large inversion in the C. reflexa plastid genome, ∼15 kb in length (Fig. 1B), was hypothesized based on the position of the petG gene relative to the rbcL gene and its transcriptional orientation (Haberhausen et al. 1992). Our PCR assay indicates that the same inversion reported for C. reflexa exists in C. japonica (both are species of subgenus Monogyna). However, this inversion is absent from the other species of Cuscuta belonging to subgenera Cuscuta and Grammica (Fig. 1B), as well as from all nonparasitic Convolvulaceae investigated. The arrangement of genes in ptDNAs from these species are syntenic with that of tobacco ptDNA, indicating that the 15-kb inversion appears to be a synapomorphy for Cuscuta subgenus Monogyna.

Long PCR amplifications with C. sandwichiana (subgenus Grammica) resulted in sequences from three ptDNA regions (Fig. 1B, C). The first sequenced area contains the rbcL gene, atpB-E operon, a series of tRNAs (trnM, trnF, trnL, and trnT), the rps4 gene, trnS, and ycf3 (= ORF168 in tobacco). This region, approximately 7 kb long, is collinear with a 15-kb region of tobacco ptDNA located in the large single copy (LSC) region (Fig. 1B). The difference in size is due to the combined effect of several deletions involving genes, intergenic spacer (IGS) regions, and introns. The most conspicuous deletion in this area of C. sandwichiana ptDNA is the portion that corresponds to 3.5 kb in tobacco containing trnV and the ndhC-J operon. Also lacking is one open reading frame (ORF70A), found in tobacco but not common in other plastid genomes (Wakasugi et al. 1998), and two introns usually present in ycf3, an ORF of unknown function that is found commonly throughout land plants. The juxtaposition of what are, in most plants, three exons into a single, uninterrupted ycf3 gene shows that the introns have been removed precisely from this gene, keeping the ORF intact, similar to the situation found in the rpl2 that was described above. No intermediate cases, i.e., those cases in which reduced intron(s) would be present, were found. This kind of “clean” intron deletion is generally thought to involve through a reverse-transcriptase-mediated mechanism, where the reverse transcription of a spliced transcript is followed by homologous recombination between the intronless cDNA and the original gene, as opposed to illegitimate recombination at a random site in the genome (Fink 1987; Dujon 1989, Hiratsuka et al. 1989; Downie et al. 1991).

The second sequenced region of C. sandwichiana, ∼7 kb long, is collinear with a 10.5-kb region of tobacco ptDNA located also in the LSC (Fig. 1B). It contains, in the following order, the psaB and rps14 genes, trnfM and trnG, ycf9 (= ORF62 in tobacco), trnS, psbC-D operon, a cluster of tRNAs (trnT, trnE, trnY, and trnD) and the psbM gene. Almost all of the ∼3-kb difference in size can be explained by deletions in the IGS regions. Only one ORF, tobacco-specific ORF105, is lacking in this portion of C. sandwichiana ptDNA compared to the tobacco sequence.

Finally, the third region of C. sandwichiana sequenced in this study (Fig. 1C) has similar linear gene arrangement as that reported for the inverted repeat A (IR A ) region of tobacco and C. reflexa (subgenus Monogyna; Bömmer et al. 1993). The portion of IRA, which covers about 15 kb in tobacco, is reduced in this Cuscuta species to approximately 7 kb. The twofold size reduction is due to a large number of reductions/deletions including the following: (1) deletion of the rpl2 gene and its intron as well as loss of the rpl23 gene, (2) significant reduction in length of the largest plastid gene ycf2 (= ORF2280 in tobacco), which has no known function, (3) a series of ORFs of unknown function (ycf15, ORF115, ORF92, and ORF79) are missing, (4) an intron usually found in the in the 3′-rps12 gene is excised precisely, and (5) the ndhB gene (and its intron) is not only greatly reduced in length but also includes a number of frameshift mutations. This transformation of ndhB to a pseudogene and the length reduction, albeit to a somewhat lesser extent, is also reported from C. reflexa (Bömmer et al. 1993; Fig. 1C).

PtDNA Variation Within Convolvulaceae

Because the phylogenetic framework for the morning-glory family, obtained through molecular phylogenetic analyses (Stefanovic’ et al. 2002; Stefanovic’ and Olmstead 2004), shows Cuscuta nested within Convolvulaceae, and because the autotrophic members of this family also exhibit significant ptDNA rearrangements, it became evident that a more appropriate comparison of ptDNA variation observed in the parasitic Cuscuta would be with other members of Convolvulaceae, rather than with the more distantly related tobacco (Solanaceae). Hence, additional insight into the structural mutations in the plastid genome of the Convolvulaceae was provided by the ptDNA sequence data drawn from the large sample of autotrophic species of this family as well as a number of Cuscuta species from all three subgenera, used originally for the phylogenetic reconstructions. The relevant portions of the alignments for four such plastid regions (atpB gene, psbE-J operon, trnL-F intron/spacer, and rpl2 gene) are summarized in Fig. 3 and are accompanied by the current phylogenetic hypothesis for Convolvulaceae.

The atpB gene was found to be missing at least two codons near the 5′’ end (position 163–168 in tobacco atpB) for all members of Convolvulaceae, including Cuscuta species (Fig. 3) The only exception to this is Humbertia madagascariensis, which has a full-length gene, as do all other angiosperms known (Savolainen et al. 2000). In the same region, some nonparasitic species, belonging mainly to the Dicranostyloideae clade (Fig. 3), are lacking three additional amino acids (AA). Given their position in the alignment, it seems that at least four additional and mutually independent deletions occurred (in Dichondreae, Maripeae, one Jacquemontia species within Dicranostyloideae, and in Cardiochlamis within the Cardiochlamyeae clade; Fig. 3) in what appears to be a hotspot for deletion mutations in the atpB gene in Convolvulaceae. Cuscuta species from all three subgenera are missing only two AA in this region as is the case with some 80 additional nonparasitic Convolvulaceae for which atpB sequences are known (Stefanovic’ et al. 2002).

Figure 3
figure 3

Partial alignments from four plastid regions (atpB gene, trnL-F intron/spacer, psbE-J operon, and rpl2 gene) showing portions relevant for comparison of plastid DNA structural mutations between parasitic and nonparasitic members of Convolvulaceae. These plastid regions were parts of molecular data matrices used originally for phylogeny reconstructions of Convolvulaceae, results of which are summarized in trees on the far left side (Stefanovic’ et al. 2002; Stefanovic’ and Olmstead 2004). Suprageneric nomenclature follows Stefanovic’ et al. (2003). Approximate position of the alignment within the corresponding gene/region is indicated above each alignment segment. Tobacco plastid DNA is used as the reference species (solid boxes depict coding sequence, open boxes depict introns, and thick lines depict intergenic spacers). Alignments derived from Cuscuta are shaded. Stop codons (thin-lined boxes) and start codons (thick-lined boxes) are indicated. Asterisks and dashes in between arrows represent discontinuations in the trnL intron alignment. Scale bars correspond to ∼100 bp.

Full size image

The absence of a 169-bp portion of the trnL intron (as compared to tobacco) in the Convolvulaceae was first reported by Stefanovic’ et al. (2002). This large deletion has a distribution pattern similar to the one described for the 5′’ atpB deletion. It is found in all Convolvulaceae, including Cuscuta species, with the exception of Humbertia. Cuscuta japonia (subgenus Monogyna) and C. europaea (subgenus Cuscuta), two Cuscuta species showing the least amount of nucleotide divergence compared to the green members of the family, were fully alignable throughout the trnL-F region. Also, the trnL intron deletion boundaries in these two subgenera followed closely those of other Convolvulaceae (Fig. 3). This was not the case with any member of the more highly divergent Cuscuta subgenus Grammica. In this subgenus, additional deletion(s) occurred resulting in an even larger gap (Fig. 3). The only other group that had a substantial departure with respect to the intron gap boundaries is Jacquemontia, the genus that shows the greatest sequence divergence of any nonparasitic Convolvulaceae (Stefanovic’ et al. 2002). The trnL-F spacer region is evolving more rapidly than the trnL intron, in both point mutations and indels, as noted in previous studies dealing with this type of data (e.g., Gielly and Taberlet 1994; McDade and Moody 1999). This region was almost entirely missing in Cuscuta subgenus Grammica (data not shown; see Stefanovic’ et al. 2002). In addition, the small remaining part of the spacer showed great nucleotide divergence and could not be unambiguously aligned with the members of the other Cuscuta subgenera, nor with green Convolvulaceae species.

Most of the length variation of the psbE-J operon is found in its three IGS regions. No size variation was found in the psbF and psbJ genes. However, the psbE and psbL genes exhibited some length variation, while always maintaining the open reading frame. The psbE gene has three independent occurrences of gaps with respect to tobacco, two of them found in green Convolvulaceae (Humbertia and Jacquemontia; not shown) and one in Cuscuta subgenus Grammica (Fig. 3). Due to this deletion in Cuscuta subgenus Grammica, the coding sequence for the psbE gene extends into the psbE-F IGS region and its stop codon overlaps with the start codon for psbF. The psbE gene of C. europaea (subgenus Cuscuta) is truncated due to an early stop codon (not shown), while this gene in C. japonica (subgenus Monogyna) is identical in length to that of nonparasitic relatives. Jacquemontia is the only genus of autotrophic Convolvulaceae that shows length variation due to an A to C substitution in the stop codon (Fig. 3), which is followed immediately by another stop codon found in all Convolvulaceae psbE sequences, in this otherwise very conserved region across angiosperms (Graham and Olmstead 2000). In most flowering plants (with the exception of monocots) for which the sequences are known, the start codon of psbL has an edit site (Kudla et al. 1992; Bock et al. 1993; Graham and Olmstead 2000); editing of C to U is necessary to produce a functional translation initiation codon for this gene in these taxa. This putative edit site is found also in most of the Convolvulaceae, except in a subset of taxa in which the editing is not necessary, and in the outgroup, Montinia. This subset corresponds to the “bifid style” clade (Dicranostyloideae; Fig. 3), first explicitly identified by Stefanovic’ et al. (2002). All examined Cuscuta species have this inferred edited start codon except for C. europaea, which has the nonedited codon (Fig. 3).

As already mentioned, an intron usually found in the rpl2 gene of angiosperms is deleted in all Convolvulaceae, including Humbertia and Cuscuta (Fig. 3), representing a unique event within Asteridae, and a synapomorphy for Convolvulaceae (Stefanovic’ et al. 2002). The rpl2 alignment within Convolvulaceae required only one gap, an in-frame insertion of two codons in Cuscuta japonica (subgenus Monogyna; Fig. 3). A comparison of the Convolvulaceae rpl2 sequences homologous to those immediately flanking the 5′’ intron terminus in the outgroups (Solanaceae, Montiniaceae) reveals that this 6-bp region, the intron-binding sequence 1 (IBSI; Michel et al. 1989), which is essential for intron splicing, is conserved among all taxa examined, including some Cuscuta species (Fig. 3). Alignment of the uninterrupted rpl2 genes from Convolvulaceae with the exons of outgroup taxa demonstrates that these sequences are maintained even after deletion of the intron and are similar to the sequences of other group II introns (Shinozaki et al. 1986b; Michel et al. 1989).

Interpretation of the Observed ptDNA Variation in a Phylogenetic Context

Figure 4 summarizes all of the available data on plastid genome variation in Cuscuta as well as in the rest of the Convolvulaceae, mapped on a simplified family phytogeny. In the context of a rooted phylogenetic hypothesis for the Convolvulaceae, all of the ptDNA structural mutations discussed in this report can be considered either plesiomorphic, autapomorphic, or synapomorphic relative to Cuscuta. The first group consists of those changes that are found in all or almost all members of Convolvulaceae, regardless of whether they are parasitic or not (Fig. 4). For example, the loss of the rpl2 intron, unique among the Asteridae, is confirmed in all Convolvulaceae, indicating that this deletion probably predated the diversification of the family and provides additional evidence for its monophyly, including Humbertia. Also, the absence of the deletions in Humbertia for both the atpB gene (at the 5′’ and 3′ ends) and the trnL intron underlines the isolated position of this genus as a sister to the rest of the family. Given the available Southern hybridization data (Downie and Palmer 1992), the absence (or major alternation) of the ycf1 gene cannot be unambiguously mapped, but it is most likely to be missing in Cuscuta species as well as in some other Convolvulaceae. These findings indicate that the plastid genome of the nonparasitic taxa in this family has undergone a number of structural changes similar to those observed in parasitic plants. Hence, not all of the changes in Cuscuta are correlated with its parasitic habit, and a significant proportion of those, at least three given the presently available data, can be better explained as plesiomorphic characters of the family.

Figure 4
figure 4

Plastid genome structural rearrangements observed in three Cuscuta subgenera and the rest of Convolvulaceae mapped on a simplified family phylogeny. Solid boxes depict rearrangements that can be unambiguously assigned to different nodes. Open boxes indicate rearrangements inferred from C. sandwichiana that could be potentially shared with other taxa, but cannot be mapped unambiguously at present due to the lack of comparative data (see text). Arrows indicate alternative positions for this type of rearrangement. A number of deletions found in the inverted repeat (IR) of C. sandwichiana, which could potentially be explained by the IR contraction, are encircled. Indels that have occurred in some Cuscuta species and in some, but not all, nonparasitic Convolvulaceae (examples can be found in Fig. 3) are not depicted on the tree.

Full size image

The autapomorphic group consists of events that have occurred in some but not all Cuscuta species/subgenera. For example, the large 15-kb inversion in the LSC is restricted to Cuscuta subgenus Monogyna. None of the other Convolvulaceae investigated, including members of the two other Cuscuta subgenera, have this inversion. However, inversions sometimes occur in photosynthetic plants as well (e.g., large inversions in the LSC of Oryza sativa and Pinus thunbergiana; Hiratsuka et al. 1989; Wakasugi et al. 1994), and there is no reason to suspect a priori an association of this type of structural rearrangement with parasitism. The trnV-UAC gene and its intron, usually found in the LSC of angiosperms, are absent in at least one species of Cuscuta subgenus Grammica (C. sandwichiana). On the other hand, this gene is present in Cuscuta subgenus Monogyna, albeit inverted with respect to its orientation in tobacco (Fig. 1B). The data concerning this tRNA gene are not available for the third subgenus, Cuscuta. However, given the phylogenetic relationships within Cuscuta, and the results for Cuscuta subgenus Monogyna, it is most likely that this deletion represents either a taxon-specific event restricted to subgenus Grammica or a feature shared between subgenera Grammica and Cuscuta. In either case, this tRNA deletion, like the 15-kb inversion in Monogyna, represents an apomorphy for some lineages within the parasitic clade, and cannot be explained entirely by parasitism. Some autapomorphic changes, however, may be related to parasitism. For the most part genomic changes associated with the transition to parasitic habit are expected to be progressive, i.e., found in some clades and not in others. Such an evolutionary transition series cannot be rejected in Cuscuta. Based on ptDNA analyses (Stefanovic’ et al. 2002), the hypothesis of progressive transition to parasitism in this genus is supported by the resulting phylogenetic relationships as well as progressive increase in sequence divergence, as deduced from the branch lengths (Stefanovic’ et al. 2002). The hemiparasitic subgenus Monogyna, showing the least amount of sequence divergence compared to the green members of the family, is found to be the sister to holoparasitic subgenera Cuscuta and Grammica. The branch leading to subgenus Cuscuta is longer, compared to that of subgenus Monogyna, and is followed by even more highly divergent sequences in subgenus Grammica (Stefanovic’ et al. 2002). This stands in contrast with the case of parasitic plants in the Orobanchaceae, where several independent losses of photosynthetic ability were inferred from the phylogenetic relationships among the holo- and hemiparasites, and consequently the hypothesis of evolutionary transition series was rejected (Young et al. 1999; Olmstead et al. 2001).

The third group concerns changes that are not found in any of the nonparasitic Convolvulaceae but are found in all Cuscuta species investigated and represent synapomorphies for this genus. Those are the changes that help define the genus, and could, in some case, be directly implicated in ability to become a parasitic plant. Presently, only the loss or significant modification of all ndh genes is identified as a ptDNA synapomorphy for all Cuscuta members. A number of additional structural changes, mainly deletions and size reductions, are identified in C. sandwichiana (Figs. 1, 3). However, due to the lack of comparative data from other Cuscuta subgenera, and representative samples of the green family members, it is not possible to infer at which point in the evolution of Convolvulaceae these changes occurred (Fig. 4). This is also complicated for the changes observed in the IR region. To distinguish the loss of genes from simple IR contraction, which is frequent in angiosperms (e.g., Goulding et al. 1996; Plunkett and Downie, 2000, and references therein), comparative data are needed to locate the junctions between both IRs and the LSC.

Molecular Implications of the ObservedptDNA Variation

In addition to conclusions regarding the pattern of plastid genome evolution in Convolvulaceae that can be drawn from a phylogenetic perspective, these inferences on structural rearrangements in Convolvulaceae in general, and Cuscuta in particular, have some molecular implications. Regardless of how the changes observed in Cuscuta sandwichiana will be distributed with respect to the two other Cuscuta subgenera and autotrophic taxa, it is already clear that most of the affected regions are those with no function or with unknown function.

Changes that probably have little or no functional consequences include a number of deleted introns, such as those usually found in rpl2, ycf3, and 3′-rps12 genes. Introns are highly stable components of land plant plastid genomes, with no cases of intron gain known during land plant evolution. The lack of these regions, however, does not necessarily affect the functioning of plastids, as evidenced by examples found throughout angiosperms where different introns are lacking in many successfully photosynthetic organisms, e.g., the rpoC1 intron and two clpP introns in Oryza sativa (Poaceae; Hiratsuka et al. 1989), the rpl2 intron in all Caryophyllales (Zurawski et al. 1984; Downie et al. 1991), Menyanthaceae, Saxifragaceae, and Convolvulaceae (Downie et al. 1991; this study), the rpl16 intron in Geraniaceae, and trnI intron in Campanulaceae (Downie et al. 1991).

Changes with unknown functional consequences include a variety of lost ycfs and ORFs. The ycfs abbreviation is reserved usually for hypothetical plastid open reading frames that have no known function but whose homologs are found throughout land plants. The open reading frames, also without known function, but found only in tobacco ptDNA (i.e., there are no homologs in the other known plastid genomes) are referred to with the ORF abbreviation. The ycf1 is reported missing or significantly altered in Cuscuta and in some photosynthetic Convolvulaceae (Downie and Palmer 1992; Fig. 4). The homologs of this hypothetical protein also are lacking from Poaceae (Hiratsuka et al. 1989; Downie and Palmer 1992) and Campanulaceae (Downie and Palmer 1992) without affecting their photosynthetic ability. The largest plastid gene, ycf2, may or may not be significantly reduced in Cuscuta, depending on whether the documented reduction affected only one of the IRs in this genus or both (see above).

Major length mutations ycf2 were detected in several photosynthetic representatives of Passifloraceae, Geraniaceae, Campanulaceae, and Poaceae (Downie et al. 1994), while at the same time the homolog of this gene is present in the otherwise significantly reduced plastid genome of holoparasitic Epifagus virginiana (dePamphilis and Palmer 1990). When combined, these results imply that ycf2 is not involved in photosynthetic metabolism as already suggested by Wolfe et al. (1992). Its function remains unknown, even though a possible proteolytic ATPase activity (Wolfe 1994) and/or a chromoplast-specific function (Richards et al. 1991) have been proposed. Finally, because their homologs are not found in other land plants, photosynthetic or not, the suite of tobacco-specific ORFs (ORF105, ORF70A, ORF115, and ORF79) that are absent in C. sandwichiana are most likely not connected with parasitism.

A third category consists of lost genes, the function of which is either known or suggested. For example, the lack of trnl-CAU and rpl23 in the IR of some Cuscuta species and of trnV-UAC in the LSC of C. sandwichiana (and possibly further in some other Cuscuta species, but definitely present in the subgenus Monogyna; Fig. 1B) might indicate that the translational apparatus is altered in some Cuscuta plastids. The IR deletions could be explained by the IR contraction and retention of only one copy of those genes, which would be sufficient to maintain the function of the plastid translational machinery. This IR-contraction hypothesis is supported by the demonstrated presence of the rpl2 gene in different Cuscuta plastid genomes, previously thought to be completely absent (Bömmer et al. 1993; see above). The lack of trnV-UAC, even if confirmed for the rest of the plastid genome in C. sandwichiana, does not necessarily affect the entire translational mechanism, because it could be compensated by the other valine tRNA (trnV-GAC) located in the IR. Deletion of ndh genes seems to be correlated with evolution of parasitism in Convolvulaceae. Due to the significant amino-acid similarity with mitochondrial genes encoding mitochondrial respiratory chain NADH dehydrogenase, it has been suggested that these genes may be involved in respiratory processes in the plastid (i.e., chlororespiration; Sugiura 1992; Peltier and Cournac 2002). The best evidence for the existence a of chlororespiratory chain comes from investigation of the unicellular green alga Chlamydomonas reinhardtii (Bennuon 1982; Peltier and Schmidt 1991). Additionally, these genes are expressed (Matsubayashi et al. 1987) and conserved across angiosperms (Olmstead and Palmer 1994; Olmstead et al. 2000) supporting the idea that such a plastid respiratory chain also exists in land plants. Because all of the ndh genes are lacking (or altered) in concert with all photosynthetic genes in holoparasitic Epifagus virginiana, it has been hypothesized also that these genes are involved in a metabolism closely connected with photosynthesis in land plants (dePamphilis and Palmer 1990). As pointed out by Haberhausen and Zetsche (1994), this is probably not the case in Cuscuta, which retains all of its photosynthetic genes in unaltered or nearly unaltered form, indicating the lack of a direct connection between photosynthesis and any putative chlororespiration mechanism. This lack of connection is further supported by the loss or significant alteration of all 11 ndh genes encountered in the black pine (Pinus thunbergii; Wakasugi et al. 1994).

In sum, our data indicate that a major trend in Cuscuta plastid evolution, and to a certain extent in other Convolvulaceae as well, is overall genome reduction. This decrease in ptDNA size is achieved through gene/ORF loss, intron loss, and strong reduction in IGS lengths. According to the presently available data, no plastid genes of a known function (with the exception of chlororespiratory genes) are missing in the parasitic genus Cuscuta that could not be potentially replaced by duplicate genes (e.g., in the IR) or genes with similar function (tRNA genes specifying the same AA). Many of the changes in Cuscuta, previously attributed to its parasitic mode of life, could be better explained either as retention of ancestral conditions within the family (i.e., plesiomorphies) or, in most cases, autapomorphies of some Cuscuta species not shared with the rest of the genus. The autapomorphic changes, given their exact phylogenetic distribution and extent, which is yet to be established, could lend further support for the hypothesis of progressive transition to parasitism in Cuscuta, as already has been indicated by the ptDNA sequence analyses. However, the third, synapomorphic, group is most likely to be explained by the parasitic lifestyle alone, because it represents changes found in Cuscuta exclusively.

These mostly unaltered plastid genomes of Cuscuta species stand in sharp contrast with the enormous morphological and physiological modifications that the ancestors of this parasitic genus must have undergone. The unexpectedly conservative nature of the Cuscuta plastids, especially when compared to those of its closest nonparasitic relatives, could be accounted for either by relatively recent origin of this lineage, or, more likely, by relaxed but still present natural selection for photosynthetic capacity, presumably because of the need for such an ability in some stages of Cuscuta’s life cycle, for example during the post-germination/pre-host-attachment period or during endosperm formation.