Rampant gene loss in the underground orchid Rhizanthella gardneri highlights evolutionary constraints on plastid genomes.

Since the endosymbiotic origin of chloroplasts from cyanobacteria 2 billion years ago, the evolution of plastids has been characterized by massive loss of genes. Most plants and algae depend on photosynthesis for energy and have retained ∼110 genes in their chloroplast genome that encode components of the gene expression machinery and subunits of the photosystems. However, nonphotosynthetic parasitic plants have retained a reduced plastid genome, showing that plastids have other essential functions besides photosynthesis. We sequenced the complete plastid genome of the underground orchid, Rhizanthella gardneri. This remarkable parasitic subterranean orchid possesses the smallest organelle genome yet described in land plants. With only 20 proteins, 4 rRNAs, and 9 tRNAs encoded in 59,190 bp, it is the least gene-rich plastid genome known to date apart from the fragmented plastid genome of some dinoflagellates. Despite numerous differences, striking similarities with plastid genomes from unrelated parasitic plants identify a minimal set of protein-encoding and tRNA genes required to reside in plant plastids. This prime example of convergent evolution implies shared selective constraints on gene loss or transfer.

Introduction

Plastid-based photosynthesis is the primary energy source for life on Earth providing oxygen and sugars as the basis of almost every “food chain” scaling up from photosynthetic cells to whole ecosystems. Because the endosymbiosis of cyanobacteria by a nonphotosynthetic eukaryote cell, the evolution of plastids have been characterized by massive gene loss and transfer of most of the remainder to the nucleus (Martin et al. 1998; Archibald 2009). However, nonphotosynthetic plants, algae, and apicomplexan parasites such as Plasmodium sp. and Toxoplasma sp. have retained a plastid and a plastid genome or “plastome” (Wolfe et al. 1992; Wilson et al. 1996; Funk et al. 2007) suggesting that the loss or transfer of DNA to the nucleus is limited by one or more essential nonphotosynthetic functions of plastids. Analyzing these functions in photosynthetic organisms is complex because perturbations of plastid metabolism have pleiotropic effects and are often deleterious or lethal (Ahlert et al. 2003; de Longevialle et al. 2008). The comparison of the plastid genomes from nonphotosynthetic plants should highlight these essential functions as well as the constraints preventing the transfer of genes from the plastid to the nucleus.

There are two distinctly different functional categories of plants that have partly or fully lost the capacity for photosynthesis: 1) parasites that exploit other plants via direct connections and 2) myco-heterotrophs that indirectly exploit other plants via mycorrhizal fungi (Brundrett 2009). Both categories are referred to as parasites here. In total, there are about 13 evolutionary lineages of parasitic plants and at least 30 of myco-heterotrophs, almost all the latter within the orchid family (Brundrett 2009; Westwood et al. 2010). To date, the analysis of complete plastomes from parasitic plants has been restricted to dicots which are direct parasites of either roots (Orobanchaceae; dePamphilis and Palmer 1990) or shoots (Cuscuta; Funk et al. 2007) of other plants. No similar analysis has been published on the plastid genomes from nonphotosynthetic orchids that have a distinct evolutionary origin (the last common ancestor of orchids and dicots lived at least 120 Ma; Ramirez et al. 2007), with a very different growth habit and a different mode of parasitism.

For this study, we chose the Western Underground Orchid (Rhizanthella gardneri; fig. 1 and ) an iconic West Australian species. Remarkably, and unlike land plants from any other genus, the entire life cycle of R. gardneri occurs underground, with the flowers opening several centimeters below the soil surface. Its subterranean lifestyle ensures that it is incapable of photosynthesis; instead, it gains its energy and nutrients as a myco-heterotroph via Ceratobasidium fungi that form ectomycorrhizas with roots of broom bush, Melaleuca uncinata (fig. 1; Bougoure et al. 2009, 2010). R. gardneri is critically endangered and known from only five locations in Western Australia (fig. 1), but related (and even rarer) species have been collected in Queensland and New South Wales, thousands of kilometers to the east, suggesting that these are ancient relictual species from a time before the center of Australia became arid, a process that started ∼15 Ma (Byrne et al. 2008). Here, we report the full plastome sequence of R. gardneri. It is the smallest and most derived plastome described in land plants but shows strong evolutionary convergences with plastomes of other unrelated parasitic organisms.

F

Description of Rhizanthella gardneri. (A) Uncovered capitulum of R. gardneri. Picture courtesy of Susumu Yamaguchi. (B) Each capitulum encloses up to ∼50 flowers. (C) Typical habitat of R. gardneri, showing the host plant, Melaleuca uncinata. (D) Approximate known sites for R. gardneri in south-western Australia. Map from Google Earth.

Materials and Methods

DNA was extracted from a capitulum of R. gardneri collected in the Munglinup area on the south coast of Western Australia. Due to the status of this endangered species, precise locations cannot be published. It was purified as described (Lang and Burger 2007) except that plastid DNA could not be isolated from nuclear DNA by ultracentrifugation. Shotgun sequencing of total DNA was carried out on a Roche FLX system using the titanium kit (Roche Diagnostics, Castle Hill, NSW, Australia) at the Lotterywest State Biomedical Facility (Perth, WA, Australia). De novo assembly was performed with wgs (Miller et al. 2008), which yielded four large plastid DNA contigs. After alignment on the Phalaenopsis aphrodite subsp. formosana plastid genome (AY916449) and manual editing of the assembly using the Seqman module of Lasergene 8 (DNASTAR, Madison, WI), the assembly of the contigs, gaps, and low coverage areas were verified by polymerase chain reaction (PCR) using primer pairs described in supplementary table 1, Supplementary Material online, and PrimeSTAR DNA polymerase (Takara, Japan). tRNA detection was performed with Aragorn (Laslett and Canback 2004) and by systematic alignment of P. aphrodite subsp. formosona plastid tRNA exons.

Total RNA from an aliquot of the sample used for DNA extraction was purified with acid phenol bromochloropropane (1/1) and precipitated with ethanol. RNAs were precipitated with LiCl before treatment with DNAfree DNAse (Ambion, Austin, TX). One microgram of total RNA was reverse transcribed with random primers and Superscript III (Invitrogen, Mount Waverley, VIC, Australia). rpl2, rpl16, and clpP cDNAs were amplified with the primers pairs described in supplementary table 1, Supplementary Material online, and PrimeSTAR DNA polymerase (Takara, Japan).

To assess the variability of the plastid genome, DNA was extracted from a bract of R. gardneri collected in the Corrigin area (WA, Australia) and purified as described above. Fragments covering 32.9% of the plastid genome (accession numbers GU066242–GU066252) were sequenced using the primers pairs described in supplementary table 1, Supplementary Material online, and the PrimeSTAR DNA polymerase (Takara, Japan). R. gardneri mitochondrial sequences among the library of reads from the FLX system were identified by similarity to rice mitochondrial DNA. For each open reading frame (ORF), the matching reads were assembled with Seqman (Lasergene 8, DNASTAR, Madison, WI). Because of low coverage, only partial sequences could be recovered for most coding sequences (CDS) except atp4, cox3, and nad9 genes. The corresponding sequences from samples collected in the Munglinup area (accession numbers GU066236, GU066239, and GU066241) and the Corrigin area (accession numbers GU066237, GU066238, and GU066240) were produced as described above using primer pairs described in supplementary table 1, Supplementary Material online. Sequence alignments were performed using MUSCLE (Edgar 2004) and manually edited. The gene lists of 172 complete plastid genomes were collected from National Center for Biotechnology Information and manually curated (supplementary table 2, Supplementary Material online). Duplicates within each genome were discarded as well as hypothetical ORFs without homologs. Using this matrix, the genomes were clustered by Bayesian inference using MrBayes 3.1 (Ronquist and Huelsenbeck 2003).

Results

Given the rarity of R. gardneri, biological material is extremely limited and insufficient for isolation of plastids. We estimated the proportion of plastid DNA in a total R. gardneri DNA preparation by quantitative PCR and calculated that it should be possible to achieve sufficient coverage of the plastid genome from a single sequencing run on total DNA. This prediction turned out to be well founded, and we assembled the 59,190 bp genome sequence (GenBank accession GQ413967) with an average of 15.7-fold coverage by shotgun deep sequencing of total DNA and targeted PCR amplifications to fill three short gaps and verify homopolymer runs.

The R. gardneri plastome (fig. 2) contains two inverted repeats (IR) of 9,767 bp, a large single-copy region of 26,361 bp, and a small single-copy region of 13,295 bp. This is the smallest known land plant organelle genome, being smaller than the plastid genome of the nonphotosynthetic dicot parasite Epifagus virginiana (Wolfe et al. 1992). The synteny with the plastome of the photosynthetic orchid P. aphrodite (Chang et al. 2006) is perfect (supplementary fig. 1, Supplementary Material online) despite rampant gene loss. Unlike any other plastome with large IRs (Strauss et al. 1988; Cai et al. 2008), the ribosomal RNA (rRNA) genes are not included within them (fig. 2). The R. gardneri plastome has a high proportion of noncoding DNA (table 1). This is not primarily due to the presence of pseudogenes as there are only two obvious pseudogenes (rpl33 and trnL) and two gene fragments (psaB and ndhK) covering 1.9% of the genome. Although easily detected by sequence alignments, the three rps12 exons predict an RPS12 protein lacking some highly conserved amino acids at the C-terminus and we failed to detect any transspliced rps12 mRNA, leading us to consider rps12 as a third pseudogene.

F

Rhizanthella gardneri plastid genome. Exons are displayed as blue arrows, introns as lines joining exons, tRNAs as black triangles, rRNAs in red, IRs as yellow arrows, and fragments or pseudogenes in orange. Ψ: pseudogenes. frag: fragment.

Table 1.

Global Features of Selected Plastomes.

	Size (kb)	% Noncoding DNAa	Coding Capacityb	Number of Pseudogenes	Number of tRNAsc	Accession Number
Angiosperms
Rhizanthella gardneri	59.2	45.1	33	5	9	GQ413967
Phalaenopsis aphrodite	149	49.8	95	7	27	NC_007499
Epifagus virginiana	70	41.2	40	18	15	NC_001568
Cuscuta obtusiflora	85.3	29.2	85	1	23	NC_009949
C. exaltata	125.4	44.8	96	7	27	NC_009963
Nicotiana tabacum	155.9	53.5	110	1	28	NC_001879
Chlorophyta
Helicosporidium sp.	37.5	5.1	53	0	24	NC_008100
Euglena longa	73.3	47.7	56	1	26	NC_002652
E. gracilis	143.2	57.4	86	7	26	NC_001603
Containing secondarily acquired red algal plastids
Eimeria tenella	34.8	8.4	40	7	23	NC_004823
Theileria parva	39.6	38.3	43	0	21	NC_007758
Toxoplasma gondii	35	19.3	44	0	21	NC_001799
Chromera velia	119.8	31.4	84	5	25	NC_014340
Alveolata sp. CCMP3155	85.5	13.6	99	2	25	NC_014345
Porphyra purpurea	191	24.2	218	0	35	NC_000925
Peridin-containing dinoflagellatesd
Amphidinium operculatum	34.2	64.2	18	0	1	Barbrook, Santucci, et al. (2006)

NOTE.—P. aphrodite, N. tabacum, E. gracilis, C. velia, Alveolata sp., P. purpurea, and A. operculatum are photosynthetic organisms while others are parasitic or heterotrophic.

Including structural RNAs but not introns. Only ORFs conserved in at least two plastid genomes were included.

Number of genes including structural RNAs and ORFs conserved in at least two plastid genomes. Duplicated genes were counted as one.

Duplicated tRNAs were counted as one.

The plastome of these organisms is fragmented into several plasmids.

The R. gardneri plastome contains only 37 genes (including duplicates within the repeats) encoding 20 proteins, 4 rRNAs, and 9 transfer RNAs (table 2). In comparison, the P. aphrodite plastome contains 110 genes (Chang et al. 2006), the plastome of E. virginiana contains 53 (Wolfe et al. 1992) and even the highly reduced genomes of Toxoplasma gondii (35 kb; NC 001799) or the parasitic green alga Helicosporidium sp. (37.5 kb; de Koning and Keeling 2006) contain 65 and 54, respectively. With the exception of the very peculiar plastid genomes of the peridin-containing dinoflagellates that are fragmented into several plasmids (Barbrook, Santucci, et al. 2006; Howe et al. 2008), the R. gardneri plastid is the most gene-poor plastid genome characterized so far (table 1, supplementary table 2, Supplementary Material online). It contains fewer genes than any other characterized genetic system in land plants. In comparison with the plastome of P. aphrodite, which can be taken to resemble that of the photosynthetic ancestor of R. gardneri, an estimated 70% of the original genes were lost or transferred to the nucleus after the switch to a parasitic nonphotosynthetic lifestyle. These missing genes include those coding for the plastid-encoded RNA polymerase (PEP), the maturase-like protein MatK, all the genes required for photosynthesis (encoding subunits of photosystem I, photosystem II, cytochrome b6f complex, and ATP synthase), as well as 6 genes encoding ribosomal proteins and 27 genes encoding tRNAs. Some of these missing genes may have been transferred to the nucleus. Although no unambiguous examples of this could be found in the sequencing data set, we cannot rule out the eventuality because of the low coverage of the nuclear genome.

Table 2.

Gene Contents of Selected Plastomes.

	Photosynthesis					Protein Synthesis					Transfer RNA Genes
	Pa	Rg	Cg	Ev		Pa	Rg	Cg	Ev		Pa	Rg	Cg	Ev
atpA	+	−	+	Ψ	infA	+	+	−	+	trnA-ugc	+	−	Ψ	Ψ
atpB	+	−	+	Ψ	rpl2	+	+	+	+	trnC-gca	+	+	+	Ψ
atpE	+	−	+	−	rpl14	+	+	+	Ψ	trnD-guc	+	+	+	+
atpF	+	−	+	−	rpl16	+	+	+	+	trnE-uuc	+	+	+	+
atpH	+	−	+	−	rpl20	+	+	+	+	trnF-gaa	+	+	+	+
atpI	+	−	+	−	rpl22	+	−	+	−	trnfM-cau	+	+	+	+
ccsA	+	−	+	−	rpl23	+	+	−	Ψ	trnG-gcc	+	−	+	−
cemA	+	−	+	−	rpl32	+	−	−	−	trnG-ucc	+	−	Ψ	−
petA	+	−	+	−	rpl33	+	Ψ	+	+	trnH-gug	+	−	+	+
petB	+	−	+	−	rpl36	+	+	+	+	trnI-cau	+	+	+	+
petD	+	−	+	−	rps2	+	+	+	+	trnI-gau	+	−	Ψ	Ψ
petG	+	−	+	−	rps3	+	+	+	+	trnK-uuu	+	−	−	−
petL	+	−	+	−	rps4	+	+	+	+	trnL-caa	+	−	+	+
petN	+	−	+	−	rps7	+	+	+	+	trnL-uaa	+	Ψ	+	−
psaA	+	−	+	−	rps8	+	+	+	+	trnL-uag	+	−	+	+
psaB	+	Ψ	+	−	rps11	+	+	+	+	trnM-cau	+	−	+	+
psaC	+	−	+	−	rps12	+	Ψ	+	+	trnN-guu	+	−	+	+
psaI	+	−	−	−	rps14	+	+	−	+	trnP-ugg	+	−	+	+
psaJ	+	−	+	−	rps15	+	−	+	−	trnQ-uug	+	+	+	+
psbA	+	−	+	Ψ	rps16	+	−	−	−	trnR-acg	+	−	Ψ	+
psbB	+	−	+	Ψ	rps18	+	+	+	+	trnR-ucu	+	−	+	Ψ
psbC	+	−	+	−	rps19	+	+	+	+	trnS-gcu	+	−	+	+
psbD	+	−	+	−		RNA Metabolism				trnS-gga	+	−	+	Ψ
psbE	+	−	+	−		Pa	Rg	Cg	Ev	trnS-uga	+	−	+	+
psbF	+	−	+	−	matK	+	−	−	+	trnT-ggu	+	−	+	−
psbH	+	−	+	−	rpoA	+	−	−	Ψ	trnT-ugu	+	−	+	−
psbI	+	−	+	−	rpoB	+	−	−	−	trnV-gac	+	−	+	−
psbJ	+	−	+	−	rpoC1	+	−	−	−	trnV-uac	+	−	−	−
psbK	+	−	+	−	rpoC2	+	−	−	−	trnW-cca	+	+	+	+
psbL	+	−	+	−		Essential Genes				trnY-gua	+	+	+	+
psbM	+	−	+	−		Pa	Rg	Cg	Ev		Ribosomal RNA Genes
psbN	+	−	+	−	clpP	+	+	+	+		Pa	Rg	Cg	Ev
psbT	+	−	+	−	accD	+	+	+	+	rrn16	+	+	+	+
rbcL	+	−	+	Ψ	ycf1	+	+	+	+	rrn23	+	+	+	+
psbZ	+	−	+	−	ycf2	+	+	+	+	rrn4.5	+	+	+	+
ycf3	+	−	+	−						rrn5	+	+	+	+
ycf4	+	−	+	−

NOTE.—Pa, Phalaenopsis aphrodite; Rg, Rhizanthella gardneri; Cg, Cuscuta gronovii; Ev, Epifagus virginiana; Ψ, pseudogene; +, present; −, missing. Genes present in all four plastomes are indicated in bold.

The loss of the rpo genes encoding the major RNA polymerase, PEP, is associated with sequence divergence in some of the remaining promoters. Among the R. gardneri plastid genes, some are transcribed both from nucleus-encoded polymerase (NEP) promoters and PEP promoters in Arabidopsis thaliana (Swiatecka-Hagenbruch et al. 2007). The analysis of the corresponding upstream sequences in R. gardneri (supplementary fig. 2, Supplementary Material online) showed that the putative PEP promoters are not conserved in the sequences upstream of the rps4 and rrn16 genes but are apparently still present upstream of ycf1. NEP promoter motifs are conserved in the promoters of ycf1, rps4, and clpP but have diverged upstream of rrn16.

Sixteen of the R. gardneri genes encode proteins of the translation machinery (6 rpl genes, 9 rps genes, and an initiation factor); the other four protein-encoding genes are accD, ycf1, ycf2, and clpP, all easily identified by homology to plastid genes from other plants. Despite the reduced tRNA set, there is no significant change in codon usage compared with P. aphrodite (supplementary table 3, Supplementary Material online).

The clustering of 172 complete plastid genomes based on their gene content (fig. 3) shows that plastomes deriving from green or red algae form two distinct groups, apart from some heterotrophs. These latter species derive either from green algae (Euglena longa, Helicosporidium sp., E. virginiana, and R. gardneri) or red algae (Eimeria tenella, T. gondii, and Theileria parva; Janouskovec et al. 2010) but cluster together. Some parasitic plants or algae (Cuscuta species, Aneura mirabilis and Cryptomonas paramecium) are not part of this cluster. However, these have either not completely switched to a nonphotosynthetic lifestyle (Hibberd et al. 1998) or did so only recently (Wickett et al. 2008; Donaher et al. 2009).

F

Convergence of plastid genomes from parasites. One-hundred and seventy-two plastid genomes were clustered by Bayesian inference using the data in supplementary table 1, Supplementary Material online. Black triangles indicate collapsed clusters, with the number and taxonomic group of the species shown. Probabilities above 0.6 are shown at the corresponding nodes. Red, green, and blue branches indicate plastids derived from the red, green, and blue lineage, respectively. Supplementary data 1, Supplementary Material online, is the full tree in NEXUS format.

Comparison of DNA sequences from the central and southern populations of R. gardneri showed that the plastome is accumulating mutations at a very high rate, at least 15-fold faster than the mitochondrial genome (supplementary table 4, Supplementary Material online). This result is in line with the increased rates of fixation of plastid DNA mutations described in other parasitic plants (Young and dePamphilis 2005). The sequence divergence observed between the two populations of R. gardneri are similar to those observed between different species of Cuscuta (supplementary table 5, Supplementary Material online). The relative rate of divergence of the R. gardneri rrn23 gene is higher than that of the rps8 and rpl36 genes when compared with the equivalent rates in the Cuscuta species (supplementary table 5, Supplementary Material online).

Despite this high rate of sequence divergence in the R. gardneri plastome, preferential conservation of the coding sequences (supplementary table 4, Supplementary Material online) suggests that the genes are expressed and functional. We cloned rpl2, rpl16, and clpP cDNAs (accession numbers GU066223, GU066224, and GU066222, respectively). These three cDNAs were correctly spliced showing that RNA splicing is occurring in R. gardneri plastids. Within these mRNAs, we detected two C–U editing events: at the start codon of rpl2 and in the second exon of rpl16 (supplementary fig. 3, Supplementary Material online). Hence, R. gardneri has normal plastid RNA metabolism with transcription, splicing, and editing occurring.

Discussion

The R. gardneri plastid contains a typical quadripartite genome with reduced IRs. This reduction has resulted in the presence of a single copy of the rrn genes as opposed to duplicates or even triplicates in all other plastomes containing large repeats (Strauss et al. 1988; Cai et al. 2008). Genes in the IRs generally display lower substitution rates compared with genes present in the single-copy regions (Wolfe et al. 1987; Perry and Wolfe 2002; Raubeson et al. 2007). The relative divergence rates within the R. gardneri plastid genome (compared with parasitic Cuscuta species) is higher for the rrn23 gene (duplicated in Cuscuta) than for the rps8 and rpl36 genes (single copy in all plastid genomes). This acceleration of mutation rates in the rrn genes in R. gardneri probably results from the loss of the duplicate copies as shown in legumes (Perry and Wolfe 2002).

The reduction of the IRs is the major reason that explains why the R. gardneri plastome is smaller than that of the smallest previously known land plant plastid genome and that of the parasitic dicot E. virginiana. Despite the very small gene set encoded in this genome, it is not the smallest in size. Plastid genomes from apicomplexan parasites such as E. tenella or the parasitic green alga Helicosporidium sp. are much smaller in size, although they contain more genes (table 1).

The R. gardneri plastome contains one of the smallest gene sets characterized in any plastome analyzed so far. Only peridin-containing dinoflagellates possess plastids with less genes (Barbrook, Santucci, et al. 2006; Howe et al. 2008). Despite rampant gene loss, the R. gardneri plastome appears to be the basis of a functioning gene expression system, with transcription, splicing, and RNA editing all detected and translation likely. Although editing requires only nuclear-encoded factors (Schmitz-Linneweber and Small 2008), transcription in land plant plastids usually relies on two distinct transcription machineries: One encoded by the nucleus (NEP) and the other encoded in part by the plastid. The rpo genes coding for the PEP are missing in R. gardneri. In photosynthetic plastids, each polymerase transcribes a distinct but overlapping set of genes (Hajdukiewicz et al. 1997), with PEP preferentially transcribing photosynthesis-related genes. As no gene typically relying exclusively on PEP for its transcription remains in the Rhizanthella plastid genome, the loss of rpo genes in R. gardneri can be understood. The loss of PEP probably explains the divergence noted in the promoter regions of several of the remaining genes (supplementary fig. 2, Supplementary Material online).

Splicing of introns in chloroplasts requires several nuclearly encoded factors that are specific to one or more introns (Asakura and Barkan 2006; de Longevialle et al. 2008). The chloroplast genome also encodes a maturase-like protein, MatK, which is involved in splicing of the trnK intron in which it is embedded (Vogel et al. 1997) but is probably also required for the splicing of the group IIa subset of plastid introns (Jenkins et al. 1997; Vogel et al. 1999; Duffy et al. 2009; McNeal et al. 2009). The loss of matK has been observed in Cuscuta species from the subgenus Grammica, which have also lost the introns thought to require MatK activity. The loss of matK in R. gardneri is more surprising because the R. gardneri plastid genome has retained three group IIa introns, two of which we show to be correctly spliced and which, in tobacco, are bound by MatK (Zoschke et al. 2010). A large proportion of matK genes in orchids are pseudogenes because of unequal insertions/deletions (Kores et al. 2000, 2001) suggesting that the role of matK for splicing in this family of plants is not as essential as in other families. This may explain why matK has been lost in R. gardneri without impairing splicing of the remaining group IIa introns.

Plastid translation is generally required for cell viability from embryogenesis onward in land plants (Berg et al. 2005), and we presume that this is the case in R. gardneri, despite the loss of many genes normally essential for translation to occur. The missing proteins are most probably imported along with other elements of the translation machinery already imported into plastids in other plants (Stengel et al. 2007). Most striking of all, the R. gardneri plastome is characterized by the loss of a large proportion of tRNA genes: Only 10 genes coding for 9 different tRNAs are present (table 2). This is by far the smallest tRNA set in an unfragmented plastid genome (table 1, supplementary table 2, Supplementary Material online) and manifestly insufficient for translation. Despite this reduced tRNA set, there is no significant change in codon usage compared with P. aphrodite (supplementary table 3, Supplementary Material online).

Import of cytosolic tRNAs into plastids has never been directly demonstrated and several lines of evidence suggest that it does not occur in photosynthetic land plants (Lung et al. 2006; Rogalski et al. 2008), but we presume that the missing tRNAs are indeed imported, as also suggested for E. virginiana (Wolfe et al. 1992). Cytosolic tRNAs are imported into mitochondria of land plants and there are intriguing parallels between the tRNA sets encoded in both organelles (Lohan and Wolfe 1998): 25 of 30 tRNAs are either sometimes imported into mitochondria and not conserved in parasite plastomes or never imported into mitochondria and conserved in plastomes (table 3). This implies that the same constraints apply for the import of tRNAs into both mitochondria and plastids. These constraints can be guessed at for all five tRNAs retained in all plant organelle genomes.

Table 3.

Plant Mitochondria and Plastids Tend to Lose and Retain the Same tRNA Genes.

		Mitochondria
		Always Conserved	Sometimes Replaced	Sometimes Lost
Plastids	Always conserved	E-uuc; Q-uug; I-cau; fM-cau; Y-gua	D-guc	F-gaa; W-cca
	Sometimes lost		eM-cau; N-guu; C-gca	S-gga; A-ugc; R-acg; G-gcc; R-ucu; G-ucc; T-ggu; I-gau; T-ugu; K-uuu; V-gac; L-uaa; V-uac; H-gug; L-uag; P-ugg; S-uga; S-gcu; L-caa

NOTE.—Modified from (Lohan and Wolfe 1998). tRNA genes are indicated by single letter amino acid codes and their anticodon sequence. “Sometimes replaced” indicates mitochondrial tRNA genes that have been replaced by orthologous plastid genes in horizontal gene transfer events.

Aminoacylated tRNAGlu(UUC) is a precursor not only for protein synthesis but also for tetrapyrrole synthesis (Tanaka R and Tanaka A 2007). In photosynthetic plants, it is required for chlorophyll synthesis, and in all plants for the synthesis of heme for mitochondrial respiratory complexes and other essential proteins. Imported cytosolic tRNAGlu would therefore need to be recognized by not only glutamyl-tRNA synthetase for aminocylation but also glutamyl-tRNA reductase for tetrapyrrole synthesis. The trnE gene may be the primary justification for the presence of a genome in all nonphotosynthetic plastids (Barbrook, Howe, et al. 2006). Similarly, tRNAfMet is required for initiating translation in prokaryotic systems, including virtually all organelles; imported tRNAMet would need to be recognized by the methionyl-tRNA transformylase or translation initiation factors in addition to being aminoacylated by methionine-tRNA synthetase. For both these tRNAs, the requirement for an imported replacement to be recognized by several distinct enzymes for which they are not normally substrates makes it extremely unlikely for functional replacement to occur.

The other three ubiquitous plant organellar tRNAs share unusual features about their aminoacylation that make them different from all other tRNAs. Organellar tRNAIle (CAU) has a typical tRNAMet anticodon, but this is modified by addition of the amino acid lysidine to C34; the organellar isoleucyl-tRNA synthetase has evolved to recognize this unusual modified base which is not found in cytosolic tRNAs. Organellar tRNATyr differs extensively from its cytosolic counterpart in that it has a very long variable loop. For both these tRNAs, the imported equivalent would fail to be recognized by the corresponding organellar aminoacyl-tRNA synthetase. Plant organellar tRNAGln-Gln is formed by amidation of tRNAGln-Glu; there is no glutaminyl-tRNA synthetase.

Thus for tRNAIle(CAU), tRNATyr, and tRNAGln, import of the cytosolic tRNA alone would be unlikely to lead to functional aminoacyl-tRNA for translation. Only coupled import of a suitable aminoacyl-tRNA synthetase could establish functional replacement of the organellar genes. Incidentally, the convergences noted here between retention of plastid and mitochondrial tRNA genes argues that unlike some early suggestions, coupled transport of tRNAs and aminoacyl-tRNA synthetases is probably not the mechanism by which organellar tRNA import occurs.

Apart from genes involved in the translation machinery (rRNAs, ribosomal proteins, and tRNAs), R. gardneri has retained a very restricted set of other protein-coding genes, namely ycf1, ycf2, accD, and clpP (table 2). accD, ycf1, and ycf2 are conserved in almost all land plants (supplementary table 2, Supplementary Material online) but have been lost in grasses (Katayama and Ogihara 1996) and accD is missing in a few other lineages (Knox and Palmer 1999; Chumley et al. 2006). The accD gene encodes the carboxyltransferase subunit of a multimeric acetyl-CoA carboxylase (ACCase), which provides malonyl-CoA for the biosynthesis of fatty acids. These fatty acids are then used for the synthesis of every cellular membrane (Benning et al. 2006). This fundamental function may explain why mutations in accD (Kode et al. 2005) or in the plastid translation machinery (Berg et al. 2005; Rogalski et al. 2008) are lethal. In the few plants where this is not the case, an imported monomeric ACCase replaces the requirement for the accD gene product (Konishi et al. 1996). The essential functions of ycf1 and ycf2 are unknown (Drescher et al. 2000) but might possibly be linked to expression, assembly, or function of the accD gene product, given that grasses have lost both genes in addition to accD. Other plants that have lost accD have divergent ycf1 and ycf2 sequences.

Finally, clpP is the only protein-coding gene present in all land plant and green algal plastomes except for that of the parasitic alga Helicosporidium sp. (supplementary table 2, Supplementary Material online). It codes for a catalytic subunit of a multimeric protease. In land plants, clpP is essential (Kuroda and Maliga 2003) but which of its varied roles are required remains unclear. Recent studies suggest that a function in the regulation of an array of processes including isoprenoid and tetrapyrrole biosynthesis, lipid body stability, and photosynthesis (Kim et al. 2009; Stanne et al. 2009; Zybailov et al. 2009).

The R. gardneri plastome is also informative on the selective pressures acting to retain certain genes within organelle genomes. The small number of retained genes and relative lack of pseudogenes (5 vs. 18 in E. virginiana) suggests that the R. gardneri plastome has progressed further toward shedding or transferring nonessential genes than any other land plant organelle examined so far. The loss of photosynthetic capacity is associated with a strong reduction of the plastid coding capacity, in line with the evolutionary trend of genome reduction in plastids (Martin et al. 1998). The remaining gene set is clearly not random. The analysis of the plastid gene content in 172 organisms (fig. 3, supplementary table 2, Supplementary Material online) showed the remarkable similarity of the apicomplexans and parasitic plants. We believe this similarity primarily reflects convergent rather than shared evolution given the supposed evolutionary distance between these plastid genomes. We note, however, that although several lines of evidence strongly support a red algal origin for apicoplasts (Wilson et al. 1996; Yoon et al. 2002; Janouskovec et al. 2010), it is still a matter of controversy (Kohler et al. 1997; Lau et al. 2009). It has been noted previously (Stiller et al. 2003) that, when considering ribosomal proteins and tRNAs, similarities in plastid gene content reflect convergent evolution rather than shared descent. The similarity of the plastid gene contents of nonphotosynthetic organisms suggests that they tend to converge toward a distinct shared gene set given enough time. The nature of this set gives clues about the evolutionary constraints on plastid gene losses.

What are these constraints? The most popular previously suggested explanations (Daley and Whelan 2005; Barbrook, Howe, et al. 2006) revolve around constraints on protein import due to high hydrophobicity (de Grey 2005) or essential regulation of gene expression by reactive oxygen signaling cascades (Allen et al. 2005; Howe et al. 2008). However, they cannot easily account for the retention of accD and clpP, which encode typical soluble globular proteins that are not involved in respiration or photosynthesis. An alternative hypothesis suggests that certain genes encoding key products required for controlling the assembly of multiprotein complexes cannot be easily lost or transferred (Zerges 2002). This could be explained by “control by epistasy of synthesis” (Wostrikoff et al. 2004) which is an elegant theory on the regulation of assembly of protein complexes that requires at least one subunit to be organelle encoded. The accD and clpP gene products are both essential components of large protein complexes that must be assembled within the plastid. We postulate that herein lies the explanation for the very small retained gene set of the R. gardneri plastome.

Supplementary Material

Supplementary tables 1–5, figures 1–3, and supplementary data 1 are available at Molecular Biology and Evolution online (http://www.mbe.oxfordjournals.org/).