Bending of Protonema Cells in a Plastid Glycolate/Glycerate Transporter Knockout Line of Physcomitrella patens

Arabidopsis LrgB (synonym PLGG1) is a plastid glycolate/glycerate transporter associated with recycling of 2-phosphoglycolate generated via the oxygenase activity of ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO). We isolated two homologous genes (PpLrgB1 and B2) from the moss Physcomitrella patens. Phylogenetic tree analysis showed that PpLrgB1 was monophyletic with LrgB proteins of land plants, whereas PpLrgB2 was divergent from the green plant lineage. Experiments with PpLrgB–GFP fusion proteins suggested that both PpLrgB1 and B2 proteins were located in chloroplasts. We generated PpLrgB single (∆B1 and ∆B2) and double (∆B1/∆B2)-knockout lines using gene targeting of P. patens. The ∆B1 plants showed decreases in growth and photosynthetic activity, and their protonema cells were bent and accumulated glycolate. However, because ∆B2 and ∆B1/∆B2 plants showed no obvious phenotypic change relative to the wild-type or ∆B1 plants, respectively, the function of PpLrgB2 remains unclear. Arabidopsis LrgB could complement the ∆B1 phenotype, suggesting that the function of PpLrgB1 is the same as that of AtLrgB. When ∆B1 was grown under high-CO2 conditions, all novel phenotypes were suppressed. Moreover, protonema cells of wild-type plants exhibited a bending phenotype when cultured on media containing glycolate or glycerate, suggesting that accumulation of photorespiratory metabolites caused P. patens cells to bend.


Introduction
Photorespiration is essential for the viability of all oxygen-producing photosynthetic organisms (reviewed in [1]). The process commences with generation of 2-phosphoglycolate (2-PG) via the oxygenase activity of ribulose 1,5-biphosphate (RuBP) carboxylase/oxygenase (RuBisCO). After conversion of 2-PG to glycolate by 2-phosphoglycolate phosphatase (PGLP) in the Many photorespiratory mutants have been isolated from different plant species including Arabidopsis, tobacco, rice, and maize, as well as green algae and cyanobacteria (reviewed in [10]), but not bryophytes those diverged from vascular plants early after land colonization [12]. The data show that photorespiration is essential not only for C 3 and C 4 plants, but also for green algae and cyanobacteria, growing in ambient air [10]. The moss Physcomitrella patens is used as a model plant due to a high frequency of homologous recombination [13] and availability of the entire genome sequence [12]. Similar to other bryophytes, the life cycle of P. patens is dominated by a haploid gametophyte phase. A spore germinates into chloronema, one type of protonema. Caulonema, the other type of protonema, arises from chloronema cells, and can form gametophores. Both female and male organs form at the apex of the gametophore and, after fertilization, sporophytes develop.
Land plants, including mosses, must be able to deal with variable light intensities because excess light energy channeled into photosynthesis generates reactive oxygen species (ROS), causing photodamage and photoinhibition. The fastest response to high light stress is provided by non-photochemical quenching (NPQ) that is a mechanism dissipating excess energy as heat. The second protection mechanism includes photorespiration, water-water cycle, cyclic electron transport within Photosystem I (PSI) and so on. In vascular plants, NPQ relies on the activity of S subunit of Photosystem II (PSBS), while algae use light-harvesting complex (Lhc)like polypeptide, LHCSR for NPQ. P. patens occupies an evolutionary intermediate position between algae and vascular plants, and exhibits high-level NPQ (in contrast to Arabidopsis) via both algal-type LHCSR-dependent and plant-type PSBS-dependent mechanisms [14]. This high-level NPQ may influence other photoprotection mechanisms including photorespiration. In the present paper, we explored the LrgB gene of the moss and analyzed the phenotypes of mutants generated via gene-targeting techniques.

Subcellular localization of PpLrgBs
The TargetP program predicted that both PpLrgB1 and B2 encoded plastid-targeting sequences of 39 and 38 amino acids, respectively, with corresponding scores of 0.87 and 0.51. To explore the subcellular locations of PpLrgB1 and B2, we constructed two plasmids in which the Cauliflower mosaic virus (CaMV) 35S promoter directed expression of the putative transit peptide (TP) fused to green fluorescent protein (GFP). After polyethylene glycol (PEG)-mediated transformation of the plasmids, the GFP fusion proteins, which bore the N-terminal regions of either PpLrgB1 or B2, were observed in chloroplasts of P. patens, corroborating the computer predictions ( Fig. 1).

Generation of transformants of P. patens
We constructed plasmids to generate PpLrgB-knockout transformants (S4 Fig.). The 5 0 and 3 0 genomic regions of the PpLrgB1 and B2 genes were amplified via genomic polymerase chain reaction (PCR) and cloned. The neomycin phosphotransferase (NPTII) gene, or the zeocin-resistance gene, driven by the CaMV35S promoter and terminated using the CaMV35S polyadenylation signal, was inserted between the 5 0 and 3 0 genomic regions of PpLrgB1 or PpLrgB2, respectively, followed by PEG-mediated transformation into P. patens. Southern hybridization experiments were performed to determine plasmid copy number. After PpLrgB1 gene targeting, line #2 showed a single insertion of the NPTII gene (S4c Fig.). Single insertions of the PpLrgB2 gene were found in two transformants (S4d Fig.). To generate PpLrgB1/B2 (ΔB1/ΔB2) double-knockout lines, we disrupted the PpLrgB2 gene in PpLrgB1 (ΔB1) knockout line #2. Southern hybridization revealed disruption of both PpLrgB1 and PpLrgB2 in two lines (S4e Fig.). Reverse transcription (RT)-PCR showed that PpLrgB transcripts were not detected in knockout transformants (S3b Fig.).
To compare the functions of the LrgB genes of P. patens and A. thaliana, stable transformants expressing AtLrgB were generated using the ΔB1-knockout line. First, AtLrgB cDNA was cloned between the rice actin promoter and the pea rbcS terminator. Next, this construct was inserted into the cloned PpDRP5B-2 gene together with a hygromycin-resistance (HPT) gene (S5 Fig.) because disruption of PpDRP5B-2 has no effect on P. patens [15]. PEG-mediated transformation of the ΔB1 knockout line followed, and stable transformants were generated. RT-PCR confirmed the expression of AtLrgB in the transformants (S5c Fig.).

Characterization of PpLrgB knockout and AtLrgB complemented lines
During generation of knockout lines lacking each PpLrgB gene, we found that protonemal colonies of the ΔB1 line were smaller than those of WT plants (S6 Fig.). Examination of growth confirmed that the ΔB1 line exhibited a lower growth rate than WT plants (Fig. 2a). However, this was not true of the ΔB2 line. Moreover, the ΔB1/ΔB2 line grew at the same rate as the ΔB1 line. These results suggested that PpLrgB2 knockout did not affect the growth rate. Stable transformation of AtLrgB into the ΔB1 line complemented the growth reduction phenotype (Fig. 2,  S6 Fig.), suggesting that PpLrgB1 is a plastidic glycolate/glycerate transporter in P. patens. To confirm the existence of a relationship between the PpLrgB1 gene and photorespiration, the ΔB1 line was grown under high-CO 2 [~0.3% (v/v) = 3,000 ppm] conditions. The growth rate of the ΔB1 line increased to that of WT plants under high-CO 2 conditions (Fig. 2). Although some chlorotic cell death was observed in leaves of the AtLrgB-knockout line of A. thaliana [4,5,6], we did not notice this phenotype in the ΔB1 line, even when cells grown under high-CO 2 conditions were transferred to ambient air (S7 Fig.).
As mutations in photorespiration-related genes affect photosynthesis in Arabidopsis [10], we measured the maximum photochemical efficiencies of Table 1). Although the F v /F m value of the ΔB2 line (0.72 for protonemata and 0.76 for gametophores) was unchanged from that of the WT (0.73 for protonemata and 0.77 for gametophores), that of the ΔB1 line decreased slightly to 0.63 for protonemata and 0.73 for gametophores. The value of the ΔB1/ΔB2 line was the same as that of ΔB1. As with the growth rate, the AtLrgB gene complemented this phenotype. The F v /F m values of ΔB1 cells were normalized under high-CO 2 conditions (Table 1). These results indicated that the ΔB1 line exhibited a photorespiratory phenotype. P. patens is known to exhibit high-level NPQ, in contrast to Arabidopsis [14]. Our measurements also confirmed high NPQ activity in P. patens (Fig. 2b). In both the ΔB1 and ΔB1/ΔB2 lines, a increase in NPQ capacity was observed in comparison with other lines, including the WT. The chlorophyll contents of the protonemata were determined for the ΔB1, ΔB2, and ΔB1/ΔB2 mutants, and the ΔB1 mutant complemented with AtLrgB (S8 Fig.). The levels of chlorophylls a and b were unchanged in all mutants.
Microscopic observation showed that both chloronema and caulonema cells of the ΔB1 and ΔB1/ΔB2 mutants were bent, whereas knockout of the PpLrgB2 gene caused no detectable changes in cell shape (Fig. 3). Measurement of the bending angles of protonemal tip cells confirmed that knockout of the PpLrgB1 gene caused cells to become curved ( Table 2). The AtLrgB gene and high-CO 2 conditions complemented the bending phenotypes (Fig. 3, Table 2). Electron microscopy revealed no obvious difference in the shape of chloroplasts between WT and ΔB1 plants (Fig. 4). However, the ΔB1 plants revealed a significant decrease (level of significance, 1%; t-test) in the number of grana (7.18 ± 1.81 grana/μm 2 ) in comparison to wild type plants (9.15 ± 1.60 grana/μm 2 ). Chloroplasts both in WT and ΔB1 mutant lines grown under high-CO 2 conditions had many thylakoids and starch granules.

Glycolate content and effects of photorespiratory metabolites
If PpLrgB1 were a plastid glycolate/glycerate transporter, glycolate would be expected to accumulate in the ΔB1 line because transport of glycolate from chloroplasts to the cytosol must necessarily be affected by the knockout. We determined that the glycolate content of the ΔB1 line Data were analyzed using the SPSS software by one-way ANOVA followed by the post hoc Tukey test to identify subgroups (a, b and c; P < 0.01), indicated by different letters. Data for protonema and gametophores were analyzed separately.
To explore whether bending of protonema cells required accumulation of photorespiratory metabolites, WT and ΔB1 cells were grown on media containing glycolate or glycerate ( Fig. 5 and Table 3). Bending of protonema cells of WT plants was evident under such conditions, suggesting that these metabolites caused cell bending.

Discussion
Our data suggest that PpLrgB1 is a plastid glycolate/glycerate transporter in P. patens, but the function of PpLrgB2 remains unclear. The PpLrgB2 gene was expressed, and the gene product was predicted to localize to chloroplasts. Recent proteomic analysis for plastids of P. patens confirmed the existence of PpLrgB2 in plastids, although PpLrgB1 was not detected [16]. The phenotype of the ΔB2 mutants did not appear to differ from the WT, and the ΔB1/ΔB2 doubleknockout lines did not show severe phenotypes in comparison with ΔB1. There is a possibility that PpLrgB2 is a transporter of other nonessential metabolites. At present, we do not exclude the possibility that PpLrgB2 has low glycolate/glycerate transporter activity.
Although ΔB1 plants exhibited decreased growth and photosynthetic activity, we did not observe a cell death phenotype similar to that of Arabidopsis mutants. The atlrgB mutants showed a chlorotic cell death phenotype upon accumulation of photorespiratory metabolites when grown in ambient air [4,5,6]. In the white sectors of leaves, photoinhibition caused by the loss of transport activity of AtLrgB triggered the cell death cascade. However, as chloroplasts were observed in the green sectors of leaves, the effects of photoinhibition were limited. This may be because photorespiratory metabolites were concentrated to greater than toxic levels in some parts of leaf tissues of A. thaliana. In contrast, because the protonemata of P. patens is a filamentous cell layer, diffusion of photorespiratory metabolites such as glycolate and glycerate to the medium may occur and prevent their accumulation in excess of toxic levels. Another reason for the no-cell-death phenotype in P. patens may be strong NPQ in P. patens cells using both of the algal-type LHCSR-dependent and PSBS-dependent mechanisms. It is reported that proteins encoded by the PSBS, LHCSR1 and LHCSR2 genes are all active in NPQ in P. patens [14]. The triple psbs lhcsr1 lhcsr2 knockout mutant lacks NPQ, while the double knockout mutant lhcsr1 lhcsr2 shows the reduction of NPQ related with LHCSR mechanisms [14]. Gene disruption of the PpLrgB1 gene with these NPQ deficient mutants may be able to demonstrate relationship between photorespiration and NPQ systems. The F v /F m value of ΔB1 (Table 1) confirmed that PpLrgB1 depletion exerted mild effects. Increased NPQ capacity was observed in the ΔB1 line when compared to that of WT (Fig. 2). Decreased photosynthetic activity of ΔB1 must cause reduction of cell growth and colony size. Electron microscopic observation suggested that chloroplasts of the ΔB1 plants had decreased numbers of grana, although the amounts of chlorophyll did not change. Stacked thylakoids were reported to become unstacked under strong illumination to prevent further damage to the D1 protein and facilitate degradation of the photodamaged D1 protein [17]. Therefore, the decrease of grana observed in the ΔB1 plants may be one of the responses to light stress. Measurements of the relative amounts of photosystem I and photosystem II may provide information regarding the phenotypes of thylakoids.
A unique phenotype of the ΔB1 mutant is bending of both cell types of protonema, the chloronemal and caulonemal cells. Protonemal cells of P. patens have been used as a model system for the study of tip cell growth (reviewed in [18]). At the apical domes of apical cells, polarized secretion of vesicles containing cell wall components and membranes occurs continuously in growing cells, mediated by turgor pressure-driven cell expansion. The cytoskeletons are closely associated with tip growth. In ΔB1 mutants, transition of chloronemata to caulonemata was evident, and both types of protonemal cells expanded in a tip-growing fashion, suggesting that the basal mechanism of tip growth was normal in such cells. The bending angle of caulonemal tip cells was larger than that of chloronemal cells ( Table 2). As chloronema cells are chloroplast- Data were analyzed using the SPSS software running the nonparametric Kruskal-Wallis test to identify subgroups (a, b and c; P < 0.01).Data for chloronema and caulonema were analyzed separately.
rich and grow more slowly than caulonema cells, the higher bending angle of the latter cells may reflect slightly faster growth. As WT cells became bent upon growth on media containing glycolate or glycerate, accumulation of photorespiratory metabolites may explain the bending (Fig. 5 and Table 3), although we cannot exclude the possibility that other non-photorespiratory metabolites generated from glycolate or glycerate cause bending. Glycolate and glycerate are thought to be able to leak through lipid bilayers by slow passive diffusion, as shown for other small organic acids. We advance a hypothesis to explain bending in ΔB1 mutants which involves the functions of acids. Although the pH values of media containing glycolate or glycerate were adjusted to that of normal medium, such acids may weaken the cell wall, compromising straight growth, or may affect the cytoskeleton of apical cells.
The photorespiration system may be conserved in land plants. However, the phenotypes of LrgB mutants differ between moss and seed plants. Arabidopsis photorespiratory mutants exhibit various phenotypes from severe lethality to minor physiological changes; detailed explanations of mutant-specific phenotypes are as yet unavailable [10]. Accumulation of data on moss photorespiratory mutants via gene knockout in P. patens may help us to understand the evolution of photorespiration in land plants.
For culture under elevated CO 2 [~0.3% (v/v) = 3,000 ppm] conditions, 17.5 ml of 2 M K 2 CO 3 and 12.5 ml of 2 M KHCO 3 were added to a plastic box (ø95 × 110 mm in height) containing an Erlenmeyer flask in which plants were growing, as described by [21]. The solutions were replenished daily. To observe cell death, plants grown for 3 days under high CO 2 conditions were transferred to ambient air conditions and observed at 2 days after transfer. To observe the effects of glycolate or glycerate on growth, the acids were added at appropriate concentrations to BCDAT medium and the pH adjusted to pH 6.6 with NaOH.

Characterization of PpLrgB1 and B2 genes
The genomic sequence of P. patens [12] was searched using the tBlastN program and the amino acid sequences encoded by the AtLrgB gene of A. thaliana. We found two genes (Pp1s63_96V6 and Pp1s143_131V6 in the P. patens genome Ver. 1.6: http://www.phytozome. net) that we termed PpLrgB1 and PpLrgB2, respectively. A phylogenetic tree was constructed using the MEGA5 software [22].
RNA was isolated from WT protonemata by a method described previously [23]. Each cDNA was amplified by RT-PCR using the primer sets PpLrgB1/F0-New and PpLrgB1/R0 for PpLrgB1, or PpLrgB2/F0 and PpLrgB2/R0 for PpLrgB2. Probes for Northern hybridization were generated using a PCR DIG Probe Synthesis Kit (Roche Diagnostics) and the appropriate primer sets, PpLrgB1/F1 and PpLrgB1/R0 for PpLrgB1, or PpLrgB2/F2 and PpLrgB2/R0 for PpLrgB2. The primers used are listed in the S9 Fig.

Subcellular localization of PpLrgB-GFP fusion proteins
Computer predictions of protein subcellular localizations were obtained using the TargetP program [24]. To construct the PpLrgB1(TP)-GFP plasmid, in which GFP was fused to the N-terminus of PpLrgB1 driven by the CaMV 35S promoter, we used a DNA fragment that included sequence coding for the N-terminal 82 amino acid residues, amplified via genomic PCR using the PpLrgB1-F0-SalI and PpLrgB1-R3-SalI primers. This DNA was digested with SalI to cut restriction sites in primers and inserted into the SalI-digested sGFP(S65T) plasmid [25]. P. patens was transformed as described previously [23]. The PpLrgB2(TP)-GFP plasmid contained the N-terminal 91 amino acid residues of PpLrgB2.

Generation of knockout lines
Plasmid pTN3 carrying the NPTII gene [20] was used to target PpLrgB1 (S4 Fig.). The NPTII gene cassette consisted of the CaMV 35S promoter, the neomycin phosphotransferase gene, and the CaMV 35S polyadenylation sequence. Genomic DNA was isolated from the protonemata of P. patens using the cetyltrimethylammonium bromide (CTAB) method [23]. The 5 0 untranslated region, and the first exon and first intron of PpLrgB1, were amplified from genomic DNA via PCR using PpLrgB1-F4 and PpLrgB1-R1 primers; subjected to blunting using a TaKaRa DNA blunting Kit (TaKaRa Bio); and cloned into the blunted EcoRI site located upstream of the NPTII gene cassette of pTN3. Next, the 3 0 untranslated region was PCR-amplified using the PpLrgB1-F5 and PpLrgB1-R2 primers, subjected to blunting, and inserted into the blunted BamHI site located downstream of the NPTII gene cassette. The plasmid thus constructed was linearized by digestion with KpnI and SacI, and used to transform P. patens. Primary screening of transformants was carried out by genomic PCR using the gene-specific primer and a primer specific for the NTPII gene cassette. Southern hybridization was used to detect additional insertions of transformed DNA into the P. patens genome (S4 Fig.). Probes for Southern hybridization were generated using the PCR DIG Probe Kit (Roche Diagnostics) employing PpLrgB1-F7 and PpLrgB1-R3 primers. The transcribed region of the PpLrgB1 gene had one HindIII and no EcoRV site, while the insert DNA contained no HindIII site and one EcoRV site. Therefore, the sizes of hybridizing bands for transformed genomic DNAs changed from 5.7 (WT) to 10.3 kbp for the HindIII restriction pattern and from 10.7 (WT) to 4.3 kbp for the EcoRV restriction pattern when the plasmid for gene disruption was inserted into the PpLrgB1 gene region (S4 Fig.). If transformants had additional insertions, other hybridized bands were observed. We selected four transformants with cell-bending phenotypes from the primary screened lines and determined that only line #2 had no additional insertions (S4 Fig.).
The p35S-Zeo plasmid carrying a zeocin-resistance gene expression cassette [26] was used to target PpLrgB2 (S4 Fig.). The 3 0 region of PpLrgB2 was amplified by genomic PCR with the PpLrgB2-F7 and PpLrgB2-R3 primers, subjected to blunting, and inserted into the blunted HindIII site of the p35S-Zeo plasmid. Next, the 5 0 region of the PpLrgB2 gene was amplified with the PpLrgB2-F5 and PpLrgB2-R1 primers, subjected to blunting, and inserted into the blunted XbaI site. The constructed plasmid was linearized by digestion with KpnI and SacI, and used to transform wild-type P. patens plants. For generation of PpLrgB1/B2 double-knockout lines, we selected PpLrgB1-knockout line #2 (S4 Fig.). Primary screening of transformants was carried out by genomic PCR using the gene-specific primer and a primer specific for the zeocin gene cassette. The insert DNA copy number was determined by Southern hybridization (S4 Fig.). The transcribed region of the PpLrgB2 gene contained one HindIII site and one EcoO109I site, while the insert DNA contained one HindIII site in the 5 0 untranscribed region of the PpLrgB2 gene and one EcoO109I site. Therefore, sizes of the hybridizing bands for transformed genomic DNAs changed from 4.6 (WT) to 6.3 kbp for the HindIII restriction pattern and from 5.0 (WT) to 4.4 kbp for the Eco0109I restriction pattern when the plasmid for gene disruption was inserted into the PpLrgB2 gene region (S4 Fig.). If transformants had additional insertions, other hybridized bands were observed. We selected several transformants from the primary screened lines and determined that lines #1 and #5 for ΔB2, and lines #1 and #9 for ΔB1/ΔB2 had no other insertions in the transformant genomes. For RT-PCR, we isolated RNAs from mutant and WT plants, treated the RNAs with DNase I, and used the RNAs to generate cDNA from oligo-dT primers. RT-PCR was performed using appropriate primer sets: PpLrgB1/F0-New and PpLrgB1/R0 for PpLrgB1, or PpLrgB2/F0 and PpLrgB2/R0 for PpLrgB2. PpActin served as a control.

Generation of AtLrgB complemented lines
For cross-species complementation testing, we used the AtLrgB gene of A. thaliana. AtLrgB cDNA was amplified by RT-PCR from DNaseI-treated total RNA of A. thaliana using the AtLrgB-F0 and AtLrgB-R0 primers, and cloned into the pBluescript vector (Agilent Technologies). The cDNA region was extracted by digestion with EcoRI and BamHI, blunted, and inserted at the EcoRV site between the rice actin promoter and the pea rbcS terminator of the pTKM1 [27] plasmid. We used the P. patens dynamin-related protein 5B-2 (PpDRP5B-2) genomic region for complementation analysis because disruption of the PpDRP5B-2 gene did not visibly affect P. patens [15]. The hygromycin phosphotransferase (HPT) gene was inserted into the EcoRV fragment region of the cloned PpDRP5B-2 gene [28]. AtLrgB cDNA, with the promoter and terminator, was isolated by digestion with XbaI and KpnI, subjected to blunting, and inserted into the NheI site of the cloned PpDRP5B-2 gene bearing the HPT gene; P. patens transformation followed (S5 Fig.). PEG-mediated transformation was performed using PpLrgB1 knockout line #2. Insertion of AtLrgB cDNA in the PpDRP5B-2 region, and expression thereof, were confirmed by Southern and RT-PCR analyses (S5 Fig.).

Characterization of the PpLrgB knockout and AtLrgB complemented lines
To measure protonemal growth, we selected five small colonies of similar sizes and adjusted the total weight to 2.5 μg/5 colonies. We used one group of 5 colonies for each point in time. Each colony was transferred to fresh BCDAT medium. One group of 5 colonies was chosen at the 1, 2, 3 and 4 weeks after transfer, and the total fresh weights of five colonies in the group were measured. Measurements were repeated five times.
We used a PAM-2500 Chlorophyll Fluorometer (Walz) to measure chlorophyll fluorescence parameters. Protonemata grown for 4-5 days, gametophores grown for 4 weeks, or gametophores grown for 3 weeks under high-CO 2 conditions were used. Samples were dark-treated for 30 min and next subjected to measurements. The intensities of saturating and actinic light were 4,000 and 828 μmol photon m -2 s -1 , respectively. F v /F m and NPQ parameters were calculated as (F m -F 0 )/F m and (F m -F m 0 )/F m 0 , respectively.
To measure the bending angles of the protonemata (Fig. 3), cells cultured for 4-5 days after transfer to new medium were used. First, the locus exhibiting the most bending was determined in a tip cell. The center of the width was positioned at this point. From this position, we drew two lines to the centers of the tip and bottom of the cell in the longitudinal direction and measured bending angles using the Angle Measurement Function of AxioVision (Zeiss).
To measure glycolate contents, the protonemata were cultured in liquid BCDAT medium with aeration for 1 week. Glycolate contents were measured using the quantitative 2,7-dihydroxynaphthalene colorimetric method [30]. To observe the effects of photorespiratory metabolites, protonema cells were cultured on BCDAT media containing glycolate or glycerate (1 μM or 10 μM) for 5 days.

Microscopic observations
Bright-field and epifluorescent cell images were recorded using a charge-coupled device (CCD) camera (Zeiss Axiocam) fitted to a microscope with filter sets for FITC and rhodamine 123 (Zeiss Axioskop 2 plus). For electron microscopy, samples were fixed in 2% glutaraldehyde buffered with 50 mM sodium cacodylate (pH 7.4), exposed to a 2% osmium tetroxide aqueous solution containing 0.1% potassium hexacyanoferrate (II), dehydrated through a graded ethanol series, and embedded in Quetol-651 resin. Thin sections were cut and stained with uranyl acetate and lead citrate, and observed using a JEM-1200EX transmission electron microscope (JEOL). To show the differences between chloroplasts of wild-type and ΔB1 plants, the number of grana and size of chloroplasts was measured for 20 chloroplast sections. The stacked thylakoids with two layers were recognized as grana. Evolutionary history was inferred using the neighbor-joining method [31]. The optimal tree is shown. The percentages of replicate trees in which the associated taxa clustered together in the bootstrap test (1,000 replicates) are given next to the branches [32]. The tree is drawn to scale, and the branch lengths are in the same units employed to describe the evolutionary distances used to infer the phylogenetic tree. Evolutionary distances were computed using the p-distance method [33], and the unit is the number of amino acid differences per site. Analysis featured 61 amino acid sequences. All ambiguous positions were removed for each sequence pair. The final data set contained a total of 649 positions.