Functional Diversification of Thylakoidal Processing Peptidases in Arabidopsis thaliana

Thylakoidal processing peptidase (TPP) is responsible for removing amino-terminal thylakoid-transfer signals from several proteins in the thylakoid lumen. Three TPP isoforms are encoded by the nuclear genome of Arabidopsis thaliana. Previous studies showed that one of them termed plastidic type I signal peptidase 1 (Plsp1) was necessary for processing three thylakoidal proteins and one protein in the chloroplast envelope in vivo. The lack of Plsp1 resulted in seedling lethality, apparently due to disruption of proper thylakoid development. The physiological roles of the other two TPP homologs remain unknown. Here we show that the three A. thaliana TPP isoforms evolved to acquire diverse functions. Phylogenetic analysis revealed that TPP may have originated before the endosymbiotic event, and that there are two groups of TPP in seed plants: one includes Plsp1 and another comprises the other two A. thaliana TPP homologs, which are named as Plsp2A and Plsp2B in this study. The duplication leading to the two groups predates the gymnosperm-angiosperm divergence, and the separation of Plsp2A and Plsp2B occurred after the Malvaceae-Brassicaceae diversification. Quantitative reverse transcription-PCR assay revealed that the two PLSP2 genes were co-expressed in both photosynthetic tissues and roots, whereas the PLSP1 transcript accumulated predominantly in photosynthetic tissues. Both PLSP2 genes were expressed in the aerial parts of the plsp1-null mutant at levels comparable to those in wild-type plants. The seedling-lethal phenotype of the plsp1-null mutant could be rescued by a constitutive expression of Plsp1 cDNA but not by that of Plsp2A or Plsp2B. These results indicate that Plsp1 and Plsp2 evolved to function differently, and that neither of the Plsp2 isoforms is necessary for proper thylakoid development in photosynthetic tissues.


Introduction
Oxygenic photosynthesis supports the lives of most organisms on earth. Capture of light energy and electron transport utilizing water as an electron donor occur in thylakoids, the internal membrane structures of cyanobacteria and chloroplasts in photosynthetic eukaryotes. As an endosymbiotic organelle, the chloroplast contains its own genome although most of its protein constituents are encoded in the nuclear genome [1]. Many of these nuclear-encoded proteins destined to chloroplasts are synthesized on cytosolic ribosomes as precursors with amino-terminal extensions called transit peptides. These precursor proteins first traverse the double-membrane envelope via the TOC and TIC (for translocon at the outer-and inner-envelope-membrane of chloroplasts) complexes [2][3][4]. Transit peptides are essential for proper protein targeting to the stroma and are removed by a soluble metallopeptidase [5]. Four distinct pathways have been identified to target proteins from the stroma to thylakoids: the cpSec (for chloroplast Sec) and cpTat (for chloroplast twinarginine translocation) pathways direct proteins to the thylakoid lumen, whereas the cpSRP (for chloroplast signal recognition particle) and non-assisted spontaneous pathways catalyze targeting of thylakoid membrane proteins [6,7]. All known nuclear-encoded cpSec and cpTat substrates, as well as some proteins that use the spontaneous pathway, carry bipartite transit peptides, which consist of thylakoid-transfer signals following the stroma-targeting transit peptides in their amino termini. The proteins with bipartite transit peptides include plastocyanins, the 33-, 23-, and 17-kD subunits of the oxygen-evolving-complex (OEC; PsbO, PsbP and PsbQ, respectively), some other photosystem components, lumenlocated proteases and several other enzymes [8,9]. The thylakoidtransfer signal is also found in chloroplast-encoded cytochrome f, which is sorted to the thylakoid membrane by the cpSec pathway [10][11][12]. During or soon after the translocation, the thylakoidtransfer signals are removed by thylakoidal processing peptidase (TPP) in the lumen.
Biochemical properties of TPP activity were extensively studied in the mid-1980's to the early 1990's [13][14][15][16][17][18][19][20]. The presence of TPP activity in complex chloroplasts of a heterokont alga was also reported [21]. The results revealed that TPP belongs to a group of membrane-bound serine proteases called the type I signal peptidase (SPase I) family. Members of the SPase I family are found in both prokaryotes and eukaryotes, cleaving intracellular or intraorganellar sorting signals in the amino termini of the translocated proteins [22]. In prokaryotes, SPases I are often called leader peptidases, which exist in the plasma membrane, apparently as monomeric forms, and remove the amino-terminal export signals from a number of proteins at the periplasmic space. Leader peptidases were shown to be essential for the viability of several Gram-negative and -positive bacteria [23][24][25][26][27][28]. In eukaryotes, there are two distinct SPase I activities in addition to TPP [22]. One is present in the endoplasmic reticulum as an oligomeric complex, cleaving the signal peptides either cotranslationally or posttranslationally [29]. Another activity is found in the mitochondria inner membrane. The enzyme responsible for this activity is called Imp (for inner membrane protease) and removes the intramitochondrial sorting signals from a subset of proteins in the space between the outer and inner membranes [30]. Imp was shown to comprise two subunits, each of which had catalytic activities with distinct specificities [31]. Biochemical studies revealed that SPases I recognize several short sequence motifs in the substrate proteins, notably small hydrophobic residues that are present at the 23 and 21 positions to the cleavage site. TPP showed more stringent requirements for these residues than other types of SPases I in vitro [20]. Beyond the conserved motifs, however, the substrate specificity of SPase I is relatively broad. For example, a bacterial SPase I could process thylakoid-transfer peptides whereas TPP could cleave bacterial export signals in vitro [18].
In 1998, the first TPP cDNA (At2g30440) was cloned from Arabidopsis thaliana based on its similarity to cyanobacterial SPases I in the coding sequence [32]. The carboxyl-terminal soluble domain of At2g30440 (residues 177-340) comprised catalytic residues conserved among SPases I. The antibody against this domain was shown to recognize a 30-kD protein in the thylakoid membrane. Furthermore, when produced in E. coli, this domain could process a cpTat substrate (the 23-kD subunit of OEC from wheat) although its activity was very low compared to the E. coli enzyme [30,32]. Based on these results, At2g30440 was defined as the TPP although its in vivo function has not been demonstrated. Later, two additional TPP homologs (At1g06870 and At3g24590) were found to be encoded in the A. thaliana genome [30,33]. A genetic study showed that one of them (At3g24590, which was termed as Plsp1 for plastidic SPase I 1) was required for proper chloroplast development [34]. PLSP1 was originally found by screening for a gene encoding a protein responsible for complete maturation of Toc75, the protein translocation channel in the chloroplast outer envelope membrane. It turned out that the plsp1null mutant accumulated unprocessed forms of not only Toc75 but also two cpSec substrates (PsbO and plastocyanin) and one cpTat substrate (PsbP) [34,35]. The catalytic activity of Plsp1 could not be directly demonstrated in vitro [35]. Nonetheless, results of biochemical and electron microscopy-immunolocalization studies support the physical involvement of Plsp1 in protein maturation in both the envelope and thylakoids [36]. Inhibition of the complete maturation of Toc75 by the combination of site-directed mutagenesis and genetic complementation with the presence of Plsp1 did not affect proper chloroplast biogenesis [35]. Hence it was suggested that the accumulation of unprocessed lumenal proteins led to disruption of thylakoid development [35,37]. These findings revealed the importance of protein maturation for thylakoid development. However, physiological roles of the other two TPP homologs in A. thaliana remain unknown.
In this work, we aimed to address the significance of gene duplications that gave rise to multiple TPP homologs in A. thaliana.
We examined phylogenetic relationships of the TPP homologs, compared patterns of their gene expression, and used a genetic complementation assay with the seedling-lethal plsp1-null mutant to test their functional interchangeability. The obtained results revealed functional diversification of the TPP homologs.

Results
Gene duplication events that gave rise to multiple TPP homologs The three TPP homologs in A. thaliana are relatively diverse in their amino termini. However, At1g06870 has a higher overall sequence identity to At2g30440 (62%) than to Plsp1 (41%) ( Figure 1A). In addition, as was reported [30], AT1G06870 and AT2G30440 genes are more similar to each other than they are to PLSP1 in the exon-intron structure ( Figure 1B). These data indicate that diversification of PLSP1 occurred before that of the other two genes. Based on this, we named At1g06870 and At2g30440 as two Plsp2 isoforms, Plsp2A and Plsp2B, respectively.
To gain insights into the gene duplication events that gave rise to the three proteins in A. thaliana, we analyzed phylogenetic relationships between potential TPP homologs from a diverse set of land plants, green algae, a red alga, a diatom and cyanobacteria, with Escherichia coli SPase I as an outgroup. The tree was constructed based on the alignment of six common domains (A, B, C, C9, D and E) [22,38] and their flanking regions that are conserved among SPases I ( Figure S1). The resulting tree ( Figure 2) provided three findings.
The first finding was that TPP may have originated before the endosymbiosis. A cyanobacterium Synechocystis sp. PCC6803 has two SPases I, LepB1 (sll0716) and LepB2 (slr1377). It was previously shown that LepB1 was predominantly involved in maturation of photosynthetic components, whereas LepB2 removed export signals from translocated proteins at the plasma membrane, similar to the indispensable SPases I in nonphotosynthetic bacteria [28]. All four cyanobacteria examined in our study contained at least one each of the LepB1 and LepB2 homologs. LepB1 homologs were more similar to plant TPP homologs than to LepB2 homologs as shown by node (a) in Figure 2. These data suggest that the SPase I specific for photosynthetic components in the ancient cyanobacterium evolved to become TPP in photosynthetic eukaryotes, whereas the one in the plasma membrane for exported proteins was lost during the evolution of chloroplasts.
The second finding was the presence of two distinct groups of TPP in vascular plants: one includes Plsp1 and the other comprises the two Plsp2 isoforms from A. thaliana as shown by node (b) in Figure 2. All seed plants analyzed were found to contain at least one member in each of the Plsp1 and Plsp2 groups. This suggests that the two groups may have evolved to play diverse roles. In addition, every member in the Plsp1 family contained a unique glycine-rich domain between the predicted transit peptide and the transmembrane domain which was not found in Plsp2 orthologs (Figures 2A and S2). The data clearly indicate that the duplication that led to Plsp1 and Plsp2 predates separation of gymnosperms and angiosperms. However, our analysis did not address whether this duplication occurred prior to or after the diversification of vascular and non-vascular plants. This is because the three TPP homologs in the moss Physicomitrella patens formed their own clade, and the relationship of this clade to the Plsp1 and Plsp2 groups could not be resolved (node (b) in Figure 2).
Finally, the tree implies a relatively recent duplication that gave rise to Plsp2A and Plsp2B as shown by node (c) in Figure 2. This is consistent with their origin resulting from the whole-genome duplication event that appears to have occurred early in the evolution of Brassicaceae after their divergence from Malvaceae [39]. In fact, the PLSP2A and PLSP2B genes were found in a pair of duplicated regions on chromosomes 1 and 2 ( Figure S3) [40]. Interestingly, each of the duplicated segments encodes paralogs of two other chloroplast proteins, the FtsH protease and the 23-kD subunit of OEC, PsbP; PLSP2A coexists with FTSH8 and PSBP1, whereas PLSP2B occurs in the same segment as FTSH2 (VAR2) and PSBP2 ( Figure S3). Previous genetic studies showed that the functions of FtsH2 and FtsH8 are partially redundant: knockout of FTSH2 resulted in a variegated phenotype, which could be rescued by overexpression of FTSH8 [41] although the fths8-null mutant was indistinguishable from wild type [42]. By contrast, PSBP1 may be the only gene encoding the functional 23-kD subunit of OEC because PSBP2 appeared to be silenced [43]. Together, these data suggest that the two PLSP2 genes may encode proteins with redundant functions, or that one of them may not be expressed.

Distinct expression patterns of the PLSP1 and PLSP2 genes
Results of our phylogenetic analysis suggest that Plsp1 and Plsp2 may play distinct roles, and that Plsp2A and Plsp2B may be  [22,38] are indicated. The conserved Ser and Lys residues shown to form the catalytic site are indicated with asterisks under the aligned sequences. The Asp residue predicted to be the first mature residue in Plsp1 by ChloroP [69], and Gly residues within the glycine-rich region of Plsp1 (see Figure S2) are indicated with a number sign (#) and plus sings (+), respectively. B) Comparison of structures of A. thaliana TPP genes. Exons and introns are indicated as boxes and solid lines, respectively. The regions of the genes coding for the conserved A, B, C9, C, D and E boxes are indicated by shading with the box. The coding sequence for box C9 is divided by the first intron in all three genes. doi:10.1371/journal.pone.0027258.g001 functionally redundant or one of them may not be functional. To test these possibilities, we examined expression patterns of the TPP genes in A. thaliana. First, we analyzed publicly available datasets by using the Genevestigator [44]. The existing ATH1 platform with useful datasets comprised oligonucleotide probes for PLSP1 and PLSP2A, but not the one for PLSP2B. However, the PLSP2A probe may cross-react with the PLSP2B transcript as it showed a significant identity to part of the Plsp2B cDNA sequence ( Figure  S4A). When the data were plotted according to developmental stages, both PLSP1 and PLSP2A genes were found to be expressed throughout the plant's life cycle, although their expression patterns were different ( Figure S4B). PLSP1 expression was relatively high from the germinated seed stage and peaked at bolting and young flower stages. By contrast, PLSP2A expression was relatively low in the germinated seed, bolting, and mature silique stages, and it peaked at the young rosettes and young flower stages. The difference between the two genes' expression patterns was also clear when the data were analyzed based on different tissues ( Figure S4C). PLSP1 expression was relatively high in embryos and photosynthetic tissues including cotyledons and leaf primordia, but was low in roots and hypocotyls. PLSP2A expression was, by contrast, high not only in photosynthetic tissues but also in roots, and was low in embryos.
The in silico data support the idea that Plsp1 and Plsp2 may have diverse functions, but do not address whether the two PLSP2 genes were co-expressed, and if either of them was silenced. Hence, we used quantitative reverse transcription (qRT)-PCR to estimate the level of transcripts encoding the three TPP homologs in leaves, cotyledons, and roots from plate-grown A. thaliana seedlings. In order to increase the accuracy of gene expression profiling, we included a reference gene, PP2A1 (AT1G59830), which encodes a catalytic subunit of Ser/Thr protein phosphatase 2A and whose expression pattern appeared to be consistent over a wide range of developmental stages by microarray analyses [45]. Our results showed that indeed the expression levels of PP2A1 in the three organs were comparable (P.0.1; Figure 3A (a)). By contrast, consistent with the in silico data, PLSP1 expression was higher in the aboveground photosynthetic organs (leaves and cotyledons) than in roots ( Figure 3A (b)): when normalized with the PP2A1 transcript level (see Materials and Methods), PLSP1 expression was found to be 2.9-fold higher in leaves than in roots (P,0.01). Interestingly, on the contrary, the expression of both PLSP2 genes was higher in roots than in photosynthetic organs ( Figure 3A (c)): when the PP2A1 transcript level was used for normalization, PLSP2A and PLSP2B transcript levels were found to be 4.5-and 2fold higher in roots than in leaves (P,0.01), respectively. The quantitative analysis also revealed that in leaves, the level of the PLSP1 transcript was 3.5 times higher than that of the PLSP2A transcript (P,0.01), but was comparable with that of PLSP2B (P.0.1). In roots, by contrast, PLSP2A and PLSP2B transcripts were 5.65 and 11.6 times more abundant than the PLSP1 transcript (P,0.05). Together, the results showed that both PLSP2A and PLSP2B genes were expressed in a similar pattern. The qRT-PCR data also confirmed the distinct expression profiles of PLSP1 and the two PLSP2 genes.
Suppressing gene expression of one isoform often results in enhanced gene expression of another isoform as a mechanism of compensation, such as enhanced expression of PSBO2 in the psbo1null mutant [46,47]. To test if this was also the case for TPP isoforms, we examined expression of the PLSP2A and PLSP2B genes in the plsp1-null mutant. As previously reported, this mutant was seedling-lethal and its development required supplementation of the media with 3% sucrose [35]. Even in this condition, it was technically difficult to separate leaves and cotyledons from other parts of the mutant plants. Hence we combined the aboveground parts together and used them as the source of RNA for the analysis. As shown in Figure 3B, disruption of PLSP1 expression did not lead to increased expression of either of the PLSP2 genes. Instead, expression the PLSP2 genes in roots was reduced in the plsp1-null mutant. This was clearer if the ratios of the expression in the roots and shoots were considered ( Figure 3B, compare (a) and (b)): when normalized with the PP2A1 transcript level, the ratios of the transcript level in roots to that in shoots of wild type were 4.3:1 for PLSP2A and 2.2:1 for PLSP2B (P,0.01), whereas those in the mutant were 2.2:1 for PLSP2A (P,0.01) and 1.5:1 for PLSP2B (P,0.05). These data may indicate that the lack of Plsp1 affected proper plastid development in roots, generating a retrograde signal to suppress expression of a subset of nuclear genes including PLSP2.

Accumulation of TPP proteins in chloroplasts
To complement the gene expression analysis, we wished to examine the level of TPP proteins. Prior to the present study, two antisera against TPP homologs were available. The first antibody, which was against residues 177-340 of Plsp2B (Plsp2B 177-340 ), was shown to recognize a protein of 30 kD in A. thaliana thylakoids [32]. The second antibody was raised against the unique carboxylterminal sequence of Plsp1 (residues 276-291; Plsp1 276-291 ), and was shown to recognize a protein of 25 kD, which existed predominantly in thylakoids of mature chloroplasts [36]. Because of the similarities between the three TPP homologs, in particular the high sequence identity between Plsp2A and Plsp2B ( Figure 1A), we wished to evaluate the specificity of these antisera. To this end, we examined their cross-reactivity with recombinant forms of the three TPP homologs, which were produced in E. coli. As expected, the aPlsp1 276-291 antibody reacted specifically with Plsp1 ( Figure 4A, compare lanes 1-3). By contrast, the aPlsp2B 177-340 antibody recognized not only Plsp2B, but also Plsp2A ( Figure 4A, lanes 5 and 6), but not Plsp1 ( Figure 4A, lane 4). The crossreactivity of the aPlsp2B 177-340 antibody with Plsp2A is likely due to the high sequence identify (82%) between the two Plsp2 isoforms within the region used as an antigen. The antisera against the unique carboxyl termini of the two Plsp2 isoforms did not appear to recognize the endogenous proteins (Shih-Chi Hsu, Rebecca Roston, and Kentaro Inoue, unpublished). Hence we used the aPlsp2B 177-340 antibody to compare the amounts of Plsp1 and Plsp2 isoforms in chloroplasts isolated from A. thaliana seedlings. Based on the immunoblotting data ( Figure 4B), the levels of Plsp1 and Plsp2 (equivalent to Plsp2B) proteins were calculated to be approximately 0.15 ng and 0.3 ng per mg chlorophyll, respectively. These data corresponded well with the qRT-PCR data, showing that the total amount of PLSP2A and PLSP2B transcripts was slightly larger than the amount of PLSP1 transcripts in leaves ( Figure 3A).
Functional interchangeability of TPP homologs using the seedling lethal plsp1-null mutant Disruption of the PLSP1 gene resulted in seedling lethality and accumulation of a subset of unprocessed proteins in the envelope and thylakoids [34,35]. The available data suggest that the endogenous level of PLSP2A and PLSP2B gene expression could not overcome the lack of the functional Plsp1 protein. We wished to test if this was due to the functional difference between Plsp1 and Plsp2 isoforms, or insufficient expression of the PLSP2 genes, although the level of their transcripts appeared to be significantly high in the plsp1-null mutant ( Figures 3B). To this end, we decided to make use of the genetic complementation system that was used to confirm that the seedling lethal phenotype of the mutant was due to the knockout of PLSP1 [34]. As shown in Table 1 and Figure 5, the construct carrying the Plsp1 coding sequence successfully complemented the mutant phenotype. By contrast, all the plsp1-null plants carrying a coding sequence for either Plsp2A or Plsp2B were indistinguishable from non-transformed plsp1-null plants in their properties, including visible phenotype ( Figure 5A) and the size of PsbO and Toc75 ( Figure 5B). Comparative RT-PCR clearly demonstrated the expression of the inserted transgene encoding Plsp2 proteins in these plants ( Figure S5C). These results support the idea that Plsp2A and Plsp2B were functionally distinct from Plsp1.

Discussion
Proper targeting of many photosynthetic proteins and enzymes located in the thylakoid lumen requires amino-terminal thylakoid-transfer signals, which are removed by TPP upon translocation. The catalytic properties of TPP activity were defined by extensive biochemical studies in the 1980's to 1990's. By the time the first TPP cDNA was cloned from A. thaliana [32], however, research on this enzyme appeared to have diminished, if not completely disappeared. Although two additional TPP homologs were identified in A. thaliana [30], no work had been reported to elucidate their physiological roles until a recent genetic study demonstrating that one of them (Plsp1) was involved in proper thylakoid development [34,35]. The aim of the present work was to elucidate the nature and significance of multiple TPP isoforms.
Our phylogenetic data suggest that TPP originated from the duplication of a SPase I in an ancient cyanobacterium. Sometime between the emergence of land plants and the separation of gymnosperms and angiosperms, TPP evolved further into two groups. The first group includes Plsp1, which is responsible for maturation of PsbO, PsbP, plastocyanin in thylakoids and Toc75 in the envelope membrane [34,35]. The second group comprises Plsp2 isoforms including Plsp2B, the first TPP whose cDNA was cloned from any eukaryotes [32]. The conservation of the glycine- rich domain in the Plsp1 orthologs but not in other TPP-related proteins ( Figure S2) implies that this feature was acquired after the divergence of Plsp1 and Plsp2 paralogs. It remains unknown whether this glycine-rich domain plays a role in targeting of Plsp1 to the chloroplast envelope as in the case of the polyglycine stretch of Toc75 [48,49]. Furthermore, the gene expression profile and the mobility on BN-PAGE suggest that Plsp1 and Plsp2 isoforms may function differently. This idea was further corroborated by the results of the assays using the plsp1-null mutant showing that neither of the Plsp2 isoforms could substitute for Plsp1 in its function. Although many questions remain to be addressed, the data presented in this study clearly indicated that the gene duplication that gave rise to the two TPP groups caused functional diversification.
The available data suggest that Plsp1 functions predominantly in chloroplasts although it should exist in non-photosynthetic plastids, too, to process the general protein translocation channel Toc75 in the envelope membrane [34]. The results obtained in this study also suggest that Plsp2 isoforms' function may be more prominent in roots than in leaves, although their accumulation in chloroplasts was significant. Because their genes are co-expressed, the two Plsp2 isoforms may function redundantly and/or exist in the same oligomeric complexes, similar to the case of the two Imp isoforms in mitochondria [30,31]. Do Plsp2 isoforms function as a TPP in chloroplasts? If so, what are their substrates? Chaal et al. showed a low processing activity of the bacterially-produced catalytic site of Plsp2B against wheat PsbP in vitro [32]. However, the plsp1-null plant, which contained a significant amount of the Plsp2 protein ( Figure 4D), accumulated the unprocessed form of PsbP [35]. One possibility is that Plsp2B requires the presence of Plsp1 to properly process PsbP, although it is also possible that PsbP is not the substrate of Plsp2B in vivo. Interestingly, among TPP substrates examined, the 17-kD subunit of OEC (PsbQ) and FtsH2/8 appeared to accumulate as mature forms in the plsp1-null mutant [35]. Therefore Plsp2 isoforms may be responsible for processing of these thylakoidal proteins instead of PsbP. Another potential Plsp2 substrate is an inner envelope protein, Tic40, as suggested by Firlej-Kwoka et al. [50]. Tic40 carries a bipartite transit peptide and was shown to be processed to its mature size by the E. coli SPase I in vitro [51]. Indeed, the sizes of Tic40 were indistinguishable between the plsp1-null mutant and wild type [34,35]. Hence, similar to the case of Plsp1, Plsp2 isoforms may play roles in processing of proteins in both the envelope and thylakoids. It is noteworthy to mention that genetic data suggest that LepB1 of Synechocystis sp. PCC6803 may exist in both the thylakoid and plasma membranes of the bacterium [28]. Hence, dual localization of TPP may have originated in the ancient cyanobacterium.
What is the function of Plsp2 isoforms in root plastids, which appear to have very limited thylakoid network [52]? A recent study identified the presence of cpSec translocon homologs in root plastids, most probably in their envelope membrane [53]. Similarly, Plsp2 may be located in the envelope membrane of root plastids. Analysis of publicly available microarray databases revealed that expression of genes encoding most of known and putative TPP substrates including Tic40 was relatively low in roots (Kentaro Inoue, unpublished). However, genes for several TPP substrates including FtsH2 and FtsH5, which were recently demonstrated to utilize distinct sorting pathways [54], were found to be expressed at a significant level in roots (Kentaro Inoue, unpublished). We will need to establish a system to determine whether these proteins are located in the scarce thylakoid network or in the envelope membrane within root plastids.
We identified a plsp2b-null mutant (SALK_000738), which was indistinguishable from wild type in its growth phenotypes (Yi-Tze Chen and Kentaro Inoue, unpublished). Interestingly, chloroplasts isolated from this mutant accumulated the protein recognized by the aPlsp2B 177-340 antibody at a level comparable to that in wild type, although the presence of the PLSP2B transcript was under the detection limit (Shih-Chi Hsu, Nicholas Ruppel, Robert Shih and Kentaro Inoue, unpublished). This result may indicate that the immunoreactive protein was derived from the PLSP2A gene, and may also support the idea that the two Plsp2 isoforms were functionally redundant. Future research including generation and analysis of the mutant plant that lacks both Plsp2 isoforms should allow us to develop a specific hypothesis about the functions of Plsp2. Furthermore, defining the localization and physiological roles of TPP isoforms in roots should help us understand the properties and functions of root plastids, which have been under-appreciated in the field of organelle biology.

Phylogenetic analysis
Protein sequences with high similarity (Expect value less than 16e 220 except for Escherichia coli LepB which was used despite having a higher E-value of 26e 210 ) to Plsp1 (AT3G24590.1) were identified by a blastp search against The Arabidopsis Information Resource protein database for Arabidopsis thaliana Columbia (http://www.arabidopsis.org), the National Center for Biotech- nology [60]. The NCBI UniGenes Pta.22678 and Pta.657 from Pinus taeda were also used for analysis. Sequences of putative mitochondrial Imp proteins were identified by blastp against the NCBI non-redundant protein sequences database, and were not included in further analysis. Amino acid sequences were aligned respectively. An asterisk indicates the unidentified bands that ran faster than mature form of PsbO. A double asterisk indicates bands that are larger than the properly processed form of Toc75 as shown before [35]. doi:10.1371/journal.pone.0027258.g005 using the accurate mode of T-COFFEE version 8.99 [61] and minor adjustments were made manually ( Figure S1). Sequences containing the conserved Boxes A, B, C9, C, D and E [22,38], which correspond to the residues 101-162, 164-179 and 184-258 of Plsp1, were used for phylogenetic analysis. To generate Figure  S2, greater than 85 amino acid residues amino terminal of Box A of land plant sequences, except for the Pinus taeda UniGenes which were incomplete, were added. The analysis was performed using the PHYLIP phylogeny package version 3.69 [62]. Mean character distances for 1000 bootstrap datasets were calculated using the Jones-Taylor-Thornton matrix model in PROTDIST. These were then used to build a tree using neighbor-joining in NEIGHBOR. A consensus tree displaying nodes supported by at least 600 bootstrap trees (60%) was produced.

Cloning of cDNA sequences encoding A. thaliana TPP homologs
Subcloning of coding sequence for Plsp1 into the pGEMH-T Easy vector (Promega, Madison, WI) was described previously [34]. The Plsp2A-coding sequence was amplified from 13-day-old A. thaliana seedlings by PCR using a set of primers shown in Table  S1 and subcloned into the pGEMH-T Easy vector. Plsp2B cDNA was obtained from the Arabidopsis Biological Resource Center (The Ohio State University, Columbus, OH) in the pUNI51 vector as clone U69173.
For production of the catalytic domains of TPP homologs in E. coli, the construct encoding residues 177-340 of Plsp2B in vector pET16b-Kan termed pET-TPP558 [32] was a kind gift of Dr. Christopher J. Howe (University of Cambridge). cDNA sequences encoding residues 135-291 of Plsp1 and 207-368 of Plsp2A were amplified by using primers listed in Table S1 and the plasmids described above as templates, and subcloned into the pET16b-kan vector using the BamHI site. The resultant plasmids as well as pET-TPP558 were individually transformed into BL21(DE3)-pLysS cells (Invitrogen, Carlsbad, CA), and proteins recovered in the inclusion bodies were purified by using Ni-NTA column following the manufacturer's instruction (Qiagen, Valencia, CA).

Plant materials, growth conditions, and qRT-PCR
Seeds of wild-type and mutant A. thaliana (Columbia-0) were sown on Gel Drying Film (Promega) placed on top of Murashige-Skoog (MS) media containing Gamborg's vitamins (Caisson Laboratories, North Logan, UT) and 1% (for the assay in Figure 3A) or 3% sucrose (for the assay in Figure 3B) and solidified with 0.7% Phytoagar (Invitrogen) in square plates. After stratification in dark at 4uC for 3 days, plates were transferred to 24uC with 19 hour light per day and incubated for 12 more days. Cotyledons, the first pairs of true leaves, and roots were then separated, frozen in liquid nitrogen, and stored at 280uC before the analysis. RNA was extracted from the stored tissues using RNeasy Plant Mini Kit (Qiagen), quantified spectrophotometrically at 260 nm, and an aliquot of the samples (1.46 mg and 0.9 mg RNA for plants grown on 1% and 3% sucrose media, respectively) was reverse-transcribed using SuperScript TM III with random primers (Invitrogen). The resultant cDNA was used as a template for PCR performed in 7300 Real-Time PCR System (Applied Biosystems, Foster City, CA) with FastStart Universal SYBR Green Master Mix (Roche, Indianapolis, IN) for quantification of PCR products and gene specific primers (Table S1), which were designed using Primer Express v2.0 (Applied Biosystems). Absolute copy number of cDNA for each gene was calculated by Sequence Detection Software v1.2.3 (Applied Biosystems) using external standard curves generated with linearized plasmid DNA. For normalization, the fold difference between the calculated cDNA copy numbers of the TPP isoform gene to that of PP2A1 for each biological replicate was calculated and used.

Sources of antibodies and immunoblotting
Antibodies against residues 276-291 of Plsp1 and PsbO (the 33-kD subunit of OEC) were prepared as described [36]. The antibody against residues 177-340 of Plsp2B was raised in rabbits (Zymed Laboratories, San Francisco, CA), and was further purified by using 30 mg of the antigen protein coupled to 100 mL of UltraLink iodoacetyl gel (MicroLink peptide coupling kit; Pierce) in the provided minicolumn according to the manufacturer's instructions as described [63]. Immunoblotting assays were done as described [64]. Amounts of proteins analyzed were quantified by Bradford's method using BSA as standard [65].

Chloroplast isolation
Chloroplasts were isolated from A. thaliana seedlings grown on MS media supplemented with 2% sucrose and 0.7% Phytagar at 24uC with 19 h light per day for 3 to 5 weeks by a grinding method as described [64], except that the grinding buffer was modified to 50 mM HEPES-KOH, 330 mM sorbitol, 2 mM EDTA, and 2% (w/v) BSA, pH 8.0.

Preparation and analysis of plsp1-null mutants transformed with constructs encoding TPP homologs
In the previous work, we used the binary vector pBIG-HYG [66] to complement the plsp1-null mutant [34]. In this work, we used GatewayH technology to facilitate cloning. Briefly, each of the TPP coding sequences was amplified from the plasmid described above with iProof High-Fidelity DNA Polymerase (Bio-Rad, Hercules, CA) using gene specific primers and Gateway attb1 adapter primers (Table S1). The PCR products were subcloned into pDONR TM 221 using the Gateway BP clonase (Invitrogen) to generate entry clones. After confirmation of the sequences, the cloned fragments were transferred to pMDC32 [67] by the Gateway LR clonase reaction (Invitrogen). The resultant plasmids were transformed into plsp1-1 heterozygous plants via the Agrobacterium-mediated floral dip method [68] and putative transformants were selected on the half-strength MS media containing vitamins supplemented with 3% sucrose, 0.8% Phytoblend agar (Caisson Laboratories), and 20 mg per mL of hygromycin.
For genotyping, a subset of plants grown on the media containing hygromycin were transferred to a fresh MS media containing hygromycin and used as the source of genomic PCR with primers listed in Table S1. For RT-PCR assay, RNA was extracted from a pool of leaves from seedlings grown on plates for 21 days and determined to be isogenic. Extraction and quantification of RNA, and reverse transcription of a total of 1 mg RNA were done as described above for qRT-PCR. Primers used for PCR are listed in Table S1, except those for the internal control (QuantumRNA 18S Internal Standards; Ambion, Austin, TX). For immunoblotting assay, total protein was extracted by 0.1 M Tris-HCl pH 6.8, 1% SDS, 15% glycerol, 5% b-ME from approximately 100 mg of the same fresh samples used for RT-PCR assay as described above.  Figure S2 The presence of glycine-rich domains in the Plsp1 orthologs. The sequences of amino-terminal 70 amino acids flanking the conserved Box A are aligned. Numbers correspond to those of the Plsp1 sequence. A polygl stretch was defined as a stretch of ten amino acid residues containing at least six glycine residues. (TIF) Figure S3 A pair of duplicated segments in chromosomes 1 and 2 of A. thaliana nuclear genome (Block 0102031203980) that include three genes PLSP2, FTSH, and PSBP. show the reactions using the plasmid used for transformation. C) RT-PCR profiles of wild-type and mutant A. thaliana seedlings for genes indicated at left. Each reaction contained two sets of primers: one for each cDNA whose size is indicated at right, and another for cDNA derived from 18S RNA indicated with an asterisk. The template used was either total RNA without (2) or with reverse transcription (+, RT). (TIF)

Supporting Information
Table S1 Sequences of oligonucleotide primers used in this study. All sequences are depicted from 59 to 39. F and R indicate forward and reverse primers, respectively. (DOC)