Horizontal gene transfer in plants

Abstract

Horizontal gene transfer (HGT) has played a major role in bacterial evolution and is fairly common in certain unicellular eukaryotes. However, the prevalence and importance of HGT in the evolution of multicellular eukaryotes remain unclear. Recent studies indicate that plant mitochondrial genomes are unusually active in HGT relative to all other organellar and nuclear genomes of multicellular eukaryotes. Although little about the mechanisms of plant HGT is known, several studies have implicated parasitic plants as both donors and recipients of mitochondrial genes. Most cases uncovered thus far have involved a single transferred gene per species; however, recent work has uncovered a case of massive HGT in Amborella trichopoda involving acquisition of at least a few dozen and probably hundreds of foreign mitochondrial genes. These foreign genes came from multiple donors, primarily eudicots and mosses. This review will examine the implications of such massive transfer, the potential mechanisms and consequences of plant-to-plant mitochondrial HGT in general, as well as the limited evidence for HGT in plant chloroplast and nuclear genomes.

Amborella trichopoda, genome evolution, horizontal (lateral) gene transfer, plant mitochondria

Introduction

Horizontal gene transfer (HGT), the transfer of genes between non-mating species, is remarkably common and important in prokaryotes. In many prokaryotes, HGT has contributed 10–20% of the genes (Koonin et al., 2001; Lawrence and Ochman, 2001; Nakamura et al., 2004). HGT often critically influences prokaryotic evolution, leading to acquisition or modification of such important traits as antibiotic resistance, virulence, photosynthesis, and nitrogen fixation. Some authors suggest that HGT ‘may be the dominant force [in prokaryotic evolution], affecting most genes in most prokaryotes’ (Doolittle et al., 2003).

With very rare exception, HGT occurs much less frequently in eukaryotes than in bacteria, although the process may have been more common early in eukaryotic evolution. Several groups have inferred that the eukaryotic nuclear genome derives from HGT through the fusion of archaebacterial and eubacterial genomes (Moreira and Lopez-Garcia, 1998; Rivera et al., 1998; Rivera and Lake, 2004), but this interpretation has been called into question (Kurland et al., 2006). Following the endosymbiotic origin of mitochondria and chloroplasts, many genes of eubacterial origin migrated to the nucleus from these organelles via intracellular gene transfer (IGT) (reviewed in Lang et al., 1999; Adams and Palmer, 2003; Timmis et al., 2004). Functional IGT from the mitochondrial genome has, based on current evidence, entirely ceased in animals and virtually ceased in fungi. In contrast, it occurs relatively frequently in flowering plants (Adams et al., 2002).

Among the eukaryotes, unicellular eukaryotes generally experience the most HGT (Keeling and Palmer, 2001; Andersson et al., 2003; Richards et al., 2003; Andersson, 2005), perhaps because they lack a sequestered germline and because they often engulf their prey, releasing DNA near the nucleus (Doolittle, 1998). Most of the foreign genes detected in these protists were acquired from bacterial donors. Although frequent in eukaryotic terms, the amount of HGT in unicellular eukaryotes ranges from a single gene to several dozen, accounting for <1% of the genome (Andersson, 2005).

Nuclear HGT is rare in multicellular eukaryotes (animals, fungi, and plants). Nearly all known cases involve bacteria as donors (Garcia-Vallve et al., 2000; Rosewich and Kistler, 2000; Screen and St Leger, 2000; Intrieri and Buiatti, 2001; Veronico et al., 2001; Watts et al., 2001; Wolf and Koonin, 2001; Kondo et al., 2002; Zardoya et al., 2002; Hall et al., 2005). Among the plants, Agrobacterium rhizogenes has donated genes, some functional, to members of its host genus Nicotiana (Furner et al., 1986; Aoki and Syono, 1999a, b; Intrieri and Buiatti, 2001; Suzuki et al., 2002). During pathogenesis, Agrobacterium transforms its host with several plasmid-encoded genes, with HGT as a natural consequence. Additional putative cases of bacterium-to-plant nuclear genome HGT (outside of organelle-to-nucleus IGT) include the acquisition of aquaglyceroporins from a eubacterium ∼1200 million years ago (Zardoya et al., 2002) and of glutathione biosynthesis genes from an alpha-proteobacterium (Copley and Dhillon, 2002).

The low levels of horizontal transfer of nuclear genes in multicellular eukaryotes contrasts with evidence that their nuclear transposable elements have moved horizontally on numerous occasions, although relatively few such transfers have yet been documented in plants (Kidwell and Lisch, 2001; Feschotte and Wessler, 2002; Diao et al., 2006). Like nuclear genomes and yeast mitochondrial genomes (Goddard and Burt, 1999), plant mitochondria have been subject to horizontal transfer of mobile genetic elements. Most notably, the discovery of high frequency angiosperm-to-angiosperm horizontal transfer of a homing group I intron in the mitochondrial cox1 gene (Cho et al., 1998; Cho and Palmer, 1999) foreshadowed the recent discovery of widespread horizontal transfer of plant mitochondrial genes.

HGT in plant mitochondria

Two papers appearing in 2003 showed that unlike other eukaryotic genomes, plant mitochondrial genomes experience frequent and evolutionarily widespread horizontal transfer of genes acquired from other eukaryotes, plants in particular. Won and Renner (2003) showed that an intron-containing portion of the mitochondrial nad1 gene had been transferred from an angiosperm to the gymnosperm Gnetum (also see Table 1). Their discovery arose from a study of Gnetum phylogeny: the authors showed that this transfer was recent (2–5 million years ago) and restricted to a single Asian clade within Gnetum. Following the acquisition of the angiosperm-derived sequence in the common ancestor of this Asian clade of Gnetum, several species in the clade have lost the foreign sequence, while all have retained the corresponding native sequence.

Table 1

Published accounts of horizontally acquired genes shown or thought to be located in plant mitochondrial genomes

Citation	Recipienta	Donorb	Gene	Statec
Bergthorsson et al. (2003)	Actinidia	Monocot	rps2	R
	Amborella	Eudicot	atp1	D
	Betulaceae	Unclear	rps11	R
	Caprifoliaceae	Ranunculales	rps11	R
	Sanguinaria	Monocot	3′ rps11	C
Won and Renner (2003)	Gnetum	Asterid	nad1B-C	D
Davis and Wurdack (2004)	Rafflesiaceae	Vitaceae	nad1B-C	?
Mower et al. (2004)	Plantago	Orobanchaceae	atp1	D
	Plantago	Convolvulaceae	atp1	D
Nickrent et al. (2004)	Apodanthaceae	Fabales	atp1	?
Woloszynska et al. (2004)	Phaseolus	Angiosperm	cp pvs-trnA	N
Bergthorsson et al. (2004)	Amborella	Angiospermd	atp4, atp6, atp8, atp9, ccmB, ccmC, ccmFN1, cox2 (2×), cox3, nad1, nad2, nad4, nad5, nad7, rpl16, rps19, sdh4	D
		Moss	cox2, nad2, nad3, nad4, nad5, nad6, nad7	D
Schönenberger et al. (2005)	Ternstroemia	Ericaceae	atp1	?
	Bruinsmia	Cyrillaceae	atp1	?
Davis et al. (2005)	Botrychium	Santalales	nad1B-C, matR	D

a

Recipient lineages are indicated by the genus examined, or when multiple related genera were found to share the same foreign gene, the family name. Parasitic plants are in bold.

b

Donor lineages as best defined by current data. Parasitic plants are in bold.

c

State indicates whether the addition of the foreign sequence resulted in duplication (D), recapture (R) of a gene previously lost to the nucleus, chimericism (C), a novel gene (N), or is unknown (?).

d

All but atp9, nad5, nad7, and cox3 are from eudicots, a derived group within angiosperms.

Bergthorsson et al. (2003) reported five cases of mitochondrion-to-mitochondrion HGT occurring among flowering plants (Table 1). This study was guided by results of previous work on mitochondrion-to-nucleus IGT that identified Actinidia arguta, Betula nigra, and Lonicera sp. as each possessing a mitochondrial gene that was not present in the mitochondrial genomes of nearly 200 related taxa (Adams et al., 2002). For example, rps11 is absent from all 182 core eudicots examined except Lonicera and Betula (Fig. 1). Phylogenetic analyses (Bergthorsson et al., 2003) provided strong evidence that the Actinidia and Lonicera/Caprifoliaceae genes were acquired from distantly related angiosperms (Table 1), but are neutral on whether the Betula gene is of vertical or horizontal descent. The overall evidence for HGT in Betula is thus weaker than in the other cases reported by Bergthorsson et al. (2003), resting in the gene's anomalous presence in Betula mitochondrial DNA (mtDNA).

Fig. 1

Distribution of mitochondrial rps11 among 280 flowering plants. Red and blue branches indicate the presence and absence, respectively, of rps11 in the mitochondrial genome. Blue circles with minus signs indicate gene losses as inferred by parsimony; red circles with plus signs indicate putative recaptures of rps11 by mtDNA. Modified from Bergthorsson et al. (2003). See Adams et al. (2002) for names of all 280 angiosperms represented by the tree.

The Actinidia, Lonicera, and Betula cases of HGT all represent examples of recapture HGT, where horizontal transfer repopulates a mitochondrial genome with a gene that had previously been lost as a consequence of its functional transfer to the nucleus. The two remaining cases reported in Bergthorsson et al. (2003), as well as the case in Won and Renner (2003), represent other outcomes of HGT with respect to the number and structure of members of a mitochondrial gene family. The transfers found in Amborella and Gnetum represent duplicative HGT, whereby foreign and native copies of a gene co-exist in the same genome (see next section for many such cases in Amborella reported in a subsequent study), while chimeric HGT occurred in Sanguinaria, giving rise to a hybrid gene that is half native/half foreign (see below for details). No other cases of chimeric HGT have been reported among the many subsequently discovered transfers listed in Table 1 and discussed below; most of the known transfers have created a state of partial to complete gene ‘duplication’ (Table 1). However, often only a short length of what is potentially a much larger—and possibly chimeric—region of transfer has been characterized, and a recent study has uncovered evidence for substantial chimeric HGT at one particular mitochondrial locus (HC Ong and JD Palmer, unpublished data).

The Bergthorsson et al. (2003) and Won and Renner (2003) papers serve as proof of principle: plants clearly exchange genes via HGT. The last 2 years have seen a small explosion of studies reporting additional cases of horizontal transfer of genes in plants (Bergthorsson et al., 2004; Davis and Wurdack, 2004; Mower et al., 2004; Woloszynska et al., 2004; Nickrent et al., 2004; Davis et al., 2005; Schönenberger et al., 2005) (Table 1). In most, but not all, cases the putative HGTs receive strong support from phylogenetic analysis. In virtually all cases, investigators have taken care to rule out the possibility that a ‘misplaced’ gene in a phylogeny merely represents experimental error, i.e. DNA contamination or mix-up. That is, results were verified by analysis of multiple DNA extractions, including some conducted in other laboratories.

In all but one of the 40 plant-to-plant HGT cases reported thus far, the transferred gene is a mitochondrial gene (encoding a housekeeping respiratory or ribosomal protein), and thus the dominant mode of HGT in plants reported thus far is mitochondrion to, apparently, mitochondrion. Moreover, the one apparent exception, involving chloroplast pvs-trnA in Phaseolus (Woloszynska et al., 2004), may actually represent mitochondrion-to-mitochondrion transfer too. This is because chloroplast sequences frequently become incorporated into mitochondrial genomes via IGT (Unseld et al., 1997; Kubo et al., 2000; Notsu et al., 2002; Handa, 2003; Clifton et al., 2004; Knoop, 2004; Sugiyama et al., 2005), and thus the possibility remains that Phaseolus acquired this chloroplast sequence via intermediate transfer through the donor's mitochondrial genome.

It is important to emphasize that few of the studies firmly demonstrate the mitochondrial provenance of the foreign sequences, although two provide convincing indirect evidence from patterns of mitochondrial RNA editing, Southern blot hybridization intensity, and substitution rates (Bergthorsson et al., 2003, 2004). In cases lacking such data, a cautious interpretation of the results is that the genes, while of foreign mitochondrial origin, may reside in either the mitochondrial or nuclear genomes (the chloroplast genome is so unlikely that it can be dismissed out of hand; see below).

Most of the published cases of HGT in plants involve transfer from one angiosperm to another. The exceptions involve the angiosperm-to-gymnosperm transfer described above (Won and Renner, 2003), a transfer from angiosperms to a fern (Davis et al., 2005), and seven transfers from mosses to an angiosperm (Bergthorsson et al., 2004). This apparent bias may simply reflect the fact that most mitochondrial genes sequenced in plants are from angiosperms. Many transfers are quite recent in evolutionary terms, being restricted to a single genus of recipient plants, or even a limited number of species within a genus (Bergthorsson et al., 2003; Won and Renner, 2003; Mower et al., 2004; Davis et al., 2005; Schönenberger et al., 2005).

It is clear that plant mtDNA is exceptionally active in HGT. Orders of magnitude more sequence data are available for land plant chloroplast DNA and animal mtDNA than for plant mtDNA, yet no cases of HGT have been found in land plant chloroplast or animal mitochondrial genomes. Although plant mtDNAs usually contain numerous nuclear- and chloroplast-derived sequences (Unseld et al., 1997; Kubo et al., 2000; Notsu et al., 2002; Handa, 2003; Clifton et al., 2004; Knoop, 2004; Sugiyama et al., 2005), there is no good evidence of a plant chloroplast genome containing DNA from other cellular compartments (Rice and Palmer, 2006). Plant mitochondria possess an active DNA uptake system (Koulintchenko et al., 2003), but no such system is known in chloroplasts. This uptake system may be critically important in the incorporation of both foreign and native DNA. A major, well-documented difference between the two organelles that may account, at least in part, for their differential propensity for HGT is their propensity to fuse. Plant mitochondria regularly fuse (Arimura et al., 2004; Sheahan et al., 2005), promoting recombination between parental mitochondrial genomes in the case of somatic hybrid plants generated by protoplast fusion, whereas chloroplasts virtually never fuse under similar conditions (Kanno et al., 1997; Mohapatra et al., 1998).

DNA transfer facilitated by direct plant-to-plant contact through parasitism has emerged as a common mechanism of HGT; several of the reported HGT events involve parasitic plants as donors (Mower et al., 2004; Davis et al., 2005) or as recipients (Davis and Wurdack, 2004; Nickrent et al., 2004; Table 1, see taxa in bold). However, this mechanism is unlikely to explain all the transfers, as many of the donor and recipient groups do not have a host–parasite relationship. Parasitic plants could still play a role in HGT, given that a generalist parasite could serve as a vector between two unrelated species. At this point, alternative mechanisms await empirical evidence. Illegitimate pollination, herbivory, bacterial or viral transfer, uptake of naked DNA in the soil, and fungal pathogens or symbionts have all been postulated (Bergthorsson et al., 2003; Won and Renner, 2003; Davis et al., 2005).

Although some of the transferred genes are obvious pseudogenes, others are intact and potentially functional. Analysis of the expression of these foreign genes has been reported in only one case. This case is striking, because the rps11 gene of Sanguinaria (the only copy present in its mtDNA) is transcribed, RNA-edited, and chimeric: its 5′ half is of eudicot, vertical origin, whereas its 3′ half is of monocot, horizontal origin (Bergthorsson et al., 2003; Fig. 2). Whether this and other intact, transcribed, and edited foreign genes actually encode functional protein awaits future studies. Previous work has reported plant mtDNA pseudogenes that are transcribed and edited (Aubert et al., 1992; Brandt et al., 1993; Sandoval et al., 2004; Kim and Kim, 2006), transcribed but not edited (Quinones et al., 1996), or not transcribed at all (Dong et al., 1998).

Fig. 2

Chimeric rps11 in Sanguinaria derived in part by HGT. Maximum likelihood trees reproduced from Bergthorsson et al. (2003) of (a) the 5′ half (219 nucleotides) and (b) the 3′ half (237 nucleotides) of rps11. Bootstrap values >60% from parsimony analysis are given above nodes, Bayesian posterior probability values >90% below. Members of the Papaveraceae are in blue (with black bullets), while monocots are in red. Note that although most nodes in these trees are not well supported owing to the short length of sequence and low substitution rate of plant mitochondrial genes (see references in text), there are several well supported nodes separating the 3′ half of Sanguinaria rps11 from other members of the Papaveraceae family and placing it firmly with monocots. Scale bars represent 0.01 nucleotide substitutions per site. Dendroc., Dendrocalamus. (c) Model for the origin of the chimeric rps11, with approximate dates from Wikström et al. (2001). Photo credits: U Bergthorsson.

Do different genes or different plant lineages differ in rates of HGT? Work on IGT has shown that mitochondrial ribosomal protein genes are much more likely to be functionally transferred to the nucleus than those involved in bioenergetics (Adams et al., 2002). There does not appear to be any such bias in genes subject to HGT (Table 1), perhaps because most of the transfers may be neutral events of no functional consequence. Substantial lineage-specific rate variation in HGT does exist among flowering plant mitochondria. There is no evidence of HGT for the protein-coding genes of the completely sequenced mitochondrial genomes of Arabidopsis, Brassica, rice, corn, or sugar beet (Unseld et al., 1997; Kubo et al., 2000; Notsu et al., 2002; Handa, 2003; Bergthorsson et al., 2004; Clifton et al., 2004). Although most published accounts of HGT involve a single gene from any given plant (Table 1), few additional genes have been examined in most of these cases, and thus the possibility of more extensive HGT must be considered. At the extreme, the mitochondrial genome of A. trichopoda is enormously rich in foreign DNA; in fact, it may contain more foreign than native DNA.

HGT in Amborella trichopoda: promiscuity of unprecedented scale

Amborella trichopoda has acquired, via HGT, partial or full-length copies of 20 of its 31 mitochondrial genes recovered by PCR (Bergthorsson et al., 2004). For 15 genes, at least one extra copy was acquired horizontally, while at least two extra copies were acquired for four genes (e.g. Fig. 3), and three copies for one gene. For all of the 20 genes with a foreign copy, Amborella mtDNA also contains a copy positioned at the base of angiosperms, as expected for vertical transmission. Therefore, HGT in these cases has created a state of duplication, or triplication, or even quadruplication, for a total of 26 transferred genes. Eighteen of the horizontally inherited genes are intact (in the regions sequenced, some genes are much longer than typical PCR products) and potentially functional, while all 31 vertically inherited genes in Amborella are intact (Bergthorsson et al., 2004; and our unpublished data).

Fig. 3

Two transfers of nad5 to Amborella from disparate plant donors. Maximum likelihood tree of nad5 exons [reproduced from Bergthorsson et al. (2004), Copyright 2004 National Academy of Sciences, USA]. Amborella genes are shown in red, core eudicots genes in blue, and moss genes in green. Bootstrap values (100 ML replicates) >50% are shown. H, horizontally acquired gene; V, vertically inherited gene. The scale bar represents 0.01 nucleotide substitutions per site. The number of nucleotides in the alignment varies across genes as not all exons amplified for all sequences; vertically transmitted copy, 1238 nucleotides; angiosperm-derived copy, 601 nucleotides; moss-derived copy, 1062 nucleotides.

The 26 foreign genes in Amborella mtDNA were acquired from a broad range of plant donors (Table 1; Fig. 3). Seven genes were acquired from mosses and the other 19 from angiosperms. Whether some transfer events may have introduced two or more foreign genes is currently unclear. Most of the 19 angiosperm-derived genes are from eudicots. In two cases, a foreign gene has a strong affiliation with a particular eudicot clade (nad1 with Corylus, a rosid, and ccmF_N1 with Daucus, an asterid), but the limited sampling (∼10 of 175 000 species of eudicots) for most other genes does not permit identification of the donor lineage.

The 26 HGTs inferred by the PCR study of Bergthorsson et al. (2004) is a lower bound; the entire Amborella mitochondrial genome sequence will undoubtedly contain additional cases of HGT, particularly those too short or too divergent to be amplified by primers designed for full-length flowering plant genes. Indeed, current sequencing of the Amborella mitochondrial genome (our unpublished data) reveals many previously undocumented foreign sequences. The available genome sequence contains all of the foreign mitochondrial genes identified in our published PCR study (Bergthorsson et al., 2004), thus putting to rest the concern that the PCR products may have been derived from foreign mtDNA integrated into the nuclear rather than the mitochondrial genome (Martin, 2005). The massive amount of HGT found in Amborella mtDNA contrasts with the absence of HGT in Amborella chloroplast DNA (Goremykin et al., 2003; Stefanovic et al., 2004; Rice and Palmer, 2006).

The unique geographic location of Amborella provides special opportunities for unravelling the factors contributing to its spectacular amounts of HGT. Amborella is a monotypic genus (and family) of subcanopy shrubs and small trees endemic to the South Pacific island of New Caledonia, where it grows on steep slopes in mid-elevation (300–900 m high) montane tropical rain forests (Feild et al., 2001). In this dank, dark environment, Amborella leaves, branches, trunks, and even fruits are often covered with diverse epiphytes, including mosses and other bryophytes. This could readily promote direct, plant-to-plant HGT, especially given the potential for herbivory to introduce epiphytic tissue and exudates within wounded Amborella tissue. Substantial published evidence for direct plant-to-plant HGT occurring in the context of parasitism was discussed above. Epiphytism may offer similar opportunities for HGT.

Nearly 80% of the ∼3400 vascular plants native to the island are endemic to New Caledonia (Lowry, 1998), as are an appreciable fraction of the New Caledonian bryophytes. Therefore, molecular examination of this largely endemic flora growing on and in association with Amborella should prove crucial in efforts to (i) elucidate the factors promoting such extensive HGT; (ii) uncover other cases of extensive HGT; (iii) pinpoint donor identities; (iv) estimate the timing of transfer; and (v) estimate the number of transfers. A major collecting trip to New Caledonia was mounted in late 2005 and DNA from hundreds of species that grow on and in association with Amborella is now being analysed to address these and other issues. In preliminary studies, promising levels of population-level polymorphism in Amborella mtDNA has also been uncovered, including two cases of apparent polymorphism for the presence/absence of horizontally acquired genes. The extraordinary abundance and presumed frequency of HGT in Amborella may well provide special opportunities for studying the evolutionary dynamics of HGT at the population level. Finally, in unpublished studies, it is found that many of the foreign mitochondrial genes in Amborella mtDNA are expressed at the level of detectable steady-state transcripts, some of which are RNA edited. This raises the possibility that some of these genes may actually be functional, although it should be pointed out that numerous instances of transcribed and RNA-edited pseudogenes are known in plant mitochondrial genomes (Aubert et al., 1992; Brandt et al., 1993; Sandoval et al., 2004; Kim and Kim, 2006), as well as even two cases of expressed intact genes that are nonetheless suspected on the basis of their aberrant editing patterns to be ‘cryptic’ pseudogenes (Mundel and Schuster, 1996; Handa, 2003).

Concluding remarks

Although PCR-based studies have revealed much about HGT, clearly the next major step is to sequence whole mitochondrial genomes of plants known to have experienced HGT. Genome sequencing will uncover transfers that are too short or from such evolutionarily distant donors that they fail to amplify by PCR with flowering plant primers. It should also be established whether multiple genes can be transferred simultaneously, or at least from the same donor lineage. It may even be found that whole donor mitochondria are transferred and fuse with recipient mitochondria to create a state of massive HGT following recombination (as described above, plant mitochondria readily fuse in general). Another clear need is to uncover very recent transfers that could provide further insight into the transfer process and answer several outstanding questions: how long are tracts of HGT; does DNA move back and forth between donors and recipients, or is transfer unidirectional; and do lineages experiencing higher levels of HGT have common ecological or physical characteristics that suggest a transfer mechanism? Addressing these questions will require both broad surveys and dense sampling.

Reported cases of HGT in plants have thus far involved evolutionarily distant donors and recipients, with times since common ancestry between the donor and recipient lineages ranging from 60 million years in the case of the Orobanchaceae-to-Plantago transfer (Mower et al., 2004) to 480 million years in the case of moss-to-Amborella transfer (Wikström et al., 2001; Bergthorsson et al., 2004). Does the preponderance of distant HGT reflect biological reality, or is it a bias in the ability to detect HGT with confidence? Studies to date have relied on the phylogenetic signal of the donated DNA being discordant with that of the host. Given the generally low rate of sequence evolution in plant mitochondrial genes (Wolfe et al., 1987; Palmer and Herbon, 1988; Laroche et al., 1997), if mtDNA were transferred within a plant family, gene sequences may not be divergent enough to provide strong evidence for such HGT. This limitation may present a significant barrier to obtaining a comprehensive view of the tempo and pattern of plant-to-plant HGT if any of the dominant modes of transfer involve mechanisms, such as illegitimate pollination, that favour closely related donors and recipients.

Now that it is clear that plant mitochondria exchange genes relatively frequently, caution is necessary when interpreting plant phylogenies from one or even a few mitochondrial genes as they may not reflect the underlying organismal phylogeny. Fortunately, the majority of past and planned molecular phylogenetic studies in plants (including DNA barcoding; Chase et al., 2005; Rubinoff et al., 2006) have used chloroplast sequences, which seem essentially immune to HGT. Also, in most cases where mitochondrial genes have been used for phylogenetic purposes, this has been done in conjunction with chloroplast and/or nuclear genes.

Although HGT in plant mitochondria is unlikely to confer radically new phenotypes (as it often does in bacteria), the evolutionary consequence of these new genes remains entirely unknown. Finally, if foreign DNA can make it into the mitochondria of one plant from another so readily, might it do the same for the plant nucleus? After all, plant nuclear genomes frequently take up foreign DNA via IGT; they typically contain dozens to hundreds of recently transferred pieces (or sometimes even entire genomes) of endogenous mitochondrial and chloroplast DNA (Arabidopsis Genome Initiative, 2000; Goff et al., 2002; Yu et al., 2002). The recent findings of the horizontal transfer of a nuclear transposable element in grasses (Diao et al., 2006) and of a Poa-like nuclear gene in Festuca (Ghatnekar et al., 2006) suggest that nuclear genomes are the next frontier in plant HGT research.

Abbreviations

Abbreviations
 
• HGT

  horizontal gene transfer

 
• IGT

  intracellular gene transfer

 
• mtDNA

  mitochondrial DNA

We thank two anonymous reviewers for insightful comments, as well as Eric Baack, Ulfar Bergthorsson, Matt Hahn, Nolan Kane, Eric Knox, Jeff Mower, Loren Rieseberg, and Greg Young for critical reading of the manuscript and helpful discussion. Research in our laboratory on HGT is supported by NIH grant RO1-GM-70612 (to JDP) and an NSF graduate research fellowship (to AOR).