An exceptional horizontal gene transfer in plastids: gene replacement by a distant bacterial paralog and evidence that haptophyte and cryptophyte plastids are sisters

Abstract

Background: Horizontal gene transfer (HGT) to the plant mitochondrial genome has recently been shown to occur at a surprisingly high rate; however, little evidence has been found for HGT to the plastid genome, despite extensive sequencing. In this study, we analyzed all genes from sequenced plastid genomes to unearth any neglected cases of HGT and to obtain a measure of the overall extent of HGT to the plastid.

Results: Although several genes gave strongly supported conflicting trees under certain conditions, we are confident of HGT in only a single case beyond the rubisco HGT already reported. Most of the conflicts involved near neighbors connected by long branches (e.g. red algae and their secondary hosts), where phylogenetic methods are prone to mislead. However, three genes--clpP, ycf2, and rpl36--provided strong support for taxa moving far from their organismal position. Further taxon sampling of clpP and ycf2 resulted in rejection of HGT due to long-branch attraction and a serious error in the published plastid genome sequence of Oenothera elata, respectively. A single new case, a bacterial rpl36 gene transferred into the ancestor of the cryptophyte and haptophyte plastids, appears to be a true HGT event. Interestingly, this rpl36 gene is a distantly related paralog of the rpl36 type found in other plastids and most eubacteria. Moreover, the transferred gene has physically replaced the native rpl36 gene, yet flanking genes and intergenic regions show no sign of HGT. This suggests that gene replacement somehow occurred by recombination at the very ends of rpl36, without the level and length of similarity normally expected to support recombination.

Conclusion: The rpl36 HGT discovered in this study is of considerable interest in terms of both molecular mechanism and phylogeny. The plastid acquisition of a bacterial rpl36 gene via HGT provides the first strong evidence for a sister-group relationship between haptophyte and cryptophyte plastids to the exclusion of heterokont and alveolate plastids. Moreover, the bacterial gene has replaced the native plastid rpl36 gene by an uncertain mechanism that appears inconsistent with existing models for the recombinational basis of gene conversion.

Figure 1

rpl36 tree and alignment. The M3 codon model in MrBayes was used to calculate the tree using the alignment shown. Nodes with posterior probability <0.95 are collapsed. Posterior probabilities (left) and PROML BP values >50% (right) are shown on the remaining nodes. The PROML bootstraps were run with four rate categories (estimated with PUZZLE) and global rearrangements. Nucleotide and amino-acid based ML analyses using PAUP* and MrBayes also gave 100% support for the division between the c-type and p-type rpl36 genes. This support is maintained when all positions containing gaps are removed. Because the 3' extension unique to some c-type rpl36 genes (see Additional File 2) was excluded from this phylogenetic analysis, it is not shown in the alignment. In the alignment, each base is colored according to the key. Taxa in red include the red algae and their secondary plastid containing relatives. A subset of the many proteobacterial species which contain both the p-type and c-type genes is shown in purple. The p-type Pseudomonas, Photobacterium, and Vibrio genes are not shown here.

Figure 2

clpP phylogeny before and after taxon addition. ML analysis was performed on an all-position nucleotide alignment using PAUP* as described in Methods with the TVM+G model used for both trees. A 60-bp 3' extension with questionable homology across taxa was removed in this analysis but was included in the original analysis. This is probably responsible for the change in the BP from 84% originally (not shown) to 70% here for Oenothera going with grasses in tree (A). Bootstrap values <50% are not shown. (A) Original taxon sampling; (B) after new taxa added.

Figure 3

Error in the published Oenothera elata ycf2. (A) Alignment of ycf2 nucleotide sequences: The top two sequences, Oenothera biennis and O. elata, were sequenced as part of this study. We did not sequence the first ~1600 bp. The bottom two sequences correspond to the published Oenothera elata and Nicotiana tabacum sequences. The bottom three sequences were used to determine a consensus base at each position, and positions that did not match this consensus are colored as denoted in the key. (B) All codon position ML tree using the TVM+G model in PAUP* with 100 bootstrap replicates. Only the 3' region of ycf2 starting at position 4023 of the published Oenothera sequence [31] was obtained for the four Myrtales taxa (Eucalyptus, Fuchsia, Clarkia and Epilobium) and the analysis was performed using this region of aligned positions as indicated in (A). Within this region, gappy positions were removed prior to phylogenetic analysis, which resulted in 2567 positions. When the entire gene was used with the published Oenothera sequence excluded, the topology was the same except that the Lotus and Arabidopsis branches were switched. When the published Oenothera is included in the full-length analysis, its strong chimerism pulled Clarkia, Epilobium and our elata sequence into an artifactual clade with the published elata gene at the base of the Solanaceae.

Figure 4

Phylogenetic tree of red-like and green-like rbcL sequences. The amino-acid Bayesian tree was generated using MrBayes with the following parameters: rates = invgamma; aamodelpr = mixed; ngen = 500000; nchains = 4. The burnin was set to 100 to generate the tree and this burnin gave a convergence diagnostic of 0.017. The nodal support values are PROML bootstrap support values obtained using global rearrangements, and four rate categories and an invariant category estimated using PUZZLE. Support values are shown on nodes with BP ≥ 50.