Identification of the Photorespiratory 2-Phosphoglycolate Phosphatase, PGLP1, in Arabidopsis 1

The chloroplastidal enzyme 2-phosphoglycolate phosphatase (PGLP), PGLP1, catalyzes the first reaction of the photorespiratory C 2 cycle, a major pathway of plant primary metabolism. Thirteen potential PGLP genes are annotated in the Arabidopsis genome; however, none of these genes has been functionally characterized and the gene encoding the photorespiratory PGLP is not known. Here, we report on the identification of the PGLP1 gene in a higher plant and provide functional evidence for a second, non-photorespiratory PGLP, PGLP2. Two candidate genes, At5g36700 ( AtPGLP1 ) and At5g47760 ( AtPGLP2 ), were selected by sequence similarity to known PGLPs from microorganisms. The two encoded proteins were overexpressed in E. coli and both show PGLP activity. T-DNA knockout of one of these genes, At5g36700 , results in very low leaf PGLP activity. The mutant is unviable in normal air, but grows well in air enriched with 0.9% CO 2 . In contrast, deletion of At5g47760 does not result in a visible phenotype and leaf PGLP activity is unaltered. Sequencing of genomic DNA from another PGLP-deficient mutant (Somerville and Ogren, 1979) revealed a combined missense and missplicing point mutation in At5g36700 . These combined data establish At5g36700 as the gene encoding the photorespiratory PGLP, PGLP1.


INTRODUCTION
2-Phosphoglycolate (2PG) phosphatase (PGLP; EC 3.1.3.18) is one of the core enzymes of plant photosynthetic carbon assimilation. Chloroplasts of photosynthesizing leaves, especially those of C 3 plants, synthesize very large daily amounts of 2PG by the oxygenase reaction of Rubisco (Andrews et al., 1971;Bowes et al., 1971). This, in addition to the severe withdrawal of metabolites from the Calvin cycle, potentially results in several negative effects on metabolism, for example, the inhibition of the Calvin cycle enzyme triosephosphate isomerase (Anderson, 1971) or of the glycolytic enzyme phosphofructokinase (Kelly and Latzko, 1976). Therefore, 2PG must be rapidly degraded and reconverted to Calvin cycle metabolites. This reconversion of 2PG occurs in the photorespiratory C 2 cycle and requires the combined action of a number of enzymes distributed over three different organelles (Tolbert, 1971;Ogren, 1984;Tolbert, 1997). Briefly, the 2PG is dephosphorylated in chloroplasts by PGLP. The peroxisomal-located glycolate oxidase then catalyzes the irreversible oxidation of the formed glycolate to glyoxylate which, by the action of two aminotransferases and using glutamate and serine as amino donors, is subsequently transaminated to glycine. The combined action of two mitochondrial enzymes, glycine decarboxylase and serine hydroxymethyltransferase, then converts two molecules of glycine to one molecule each of serine, NH 3 and CO 2 . Serine leaves the mitochondria and becomes deaminated to 3hydroxypyruvate in a second peroxisomal aminotransferase reaction. 3-hydroxypyruvate reductase is reduced to glycerate, which becomes phosphorylated by the chloroplastidal glycerate 3-kinase to finally yield the Calvin cycle intermediate 3-phosphoglycerate.
PGLP is a light-inducible and light-regulated enzyme (for example, Baldy et al., 1989) which was purified from higher plants (for example, Richardson and Tolbert, 1961;Belanger and Ogren, 1987), green algae (Mamedov et al., 2001), and other sources. The enzyme is highly specific for 2PG (Husic and Tolbert, 1984) and requires Cl - (Seal and Rose, 1987) and Mg 2+ ions (Christeller and Tolbert, 1978a) for its full activity. Reported subunit molecular masses for the enzyme of higher plants vary between 21 and 32 kDa and both homo-dimeric and homo-tetrameric structures were suggested for the native enzyme (Christeller and Tolbert, 1978b;Hardy and Baldy, 1986;Hall et al., 1987;Belanger and Ogren, 1987).
In light of the position of the photorespiratory PGLP as the entry enzyme into one of the major pathways of primary plant metabolism, it is remarkable that neither the structure of the protein nor the encoding gene(s) are known. Earlier genetic data suggested that the photorespiratory PGLP is encoded by single genes in Arabidopsis thaliana (Arabidopsis) and barley. This was concluded from the preliminary characterization of a PGLP-deficient ethylmethane sulfonate (EMS) mutant of Arabidopsis, CS119, which could not grow under 6 ambient conditions, but was viable under the non-photorespiratory condition of 1% CO 2 (Somerville and Ogren, 1979). Similar growth properties were reported for a PGLP-deficient barley mutant (Hall et al., 1987); however, isoenzyme studies performed by the same authors indicated the presence of two isoforms of the enzyme in barley (Hall et al., 1987). The genetic defect is not known for either of these two mutants.
We therefore set out to identify this key enzyme of plant photosynthetic metabolism and its encoding gene(s) in Arabidopsis. The respective AraCyc reaction currently lists a total of 13 putative PGLP-encoding genes (http://www.arabidopsis.org/biocyc/index.jsp), but none of these genes has been functionally characterized. Several of the translated proteins show homology to prokaryotic PGLPs and seven putative PGLPs are predicted to be chloroplast proteins (At1g56500, At3g10970, At3g48420, At4g25840, At4g39970, At5g36700, and At5g36790). By highest similarity to the PGLP amino acid sequence from the unicellular green alga Chlamydomonas reinhardtii (Mamedov et al., 2001), which represents the only as yet identified PGLP from a photosynthetic eukaryote, we selected two candidate genes for overexpression and knockout studies. Both genes encode functional PGLPs; however, the knockout of only one gene, At5g36700 (AtPGLP1), resulted in leaf PGLP-deficiency in combination with a conditional lethal high-CO 2 requiring phenotype. The knockout of the second gene, At5g47760 (AtPGLP2), had no effect on leaf PGLP activity or plant growth and development in normal air. Sequencing of the At5g36700 gene from the CS119 Arabidopsis mutant revealed a point mutation in the immediate upstream context of the splice donor site of exon eight. This mutation, as its major effect, causes aberrant splicing of the Atpglp1 primary transcript. These findings thus ultimately identify the photorespiratory PGLP1 of plants and show that this enzyme is encoded by gene At5g36700 in Arabidopsis.

Arabidopsis genes At5g36700 and At5g47760 encode PGLPs
Among the 13 putative PGLPs present in the Arabidopsis genome (http://www.arabidopsis.org/biocyc/index.jsp), At5g36700, At5g36790, and At5g47760 are most similar to the Chlamydomonas PGLP (Mamedov et al., 2001). Two of these proteins, At5g36700 and At5g36790, harbour predicted chloroplast targeting sequences. Closer inspection of the corresponding region on chromosome 5, however, indicated double annotation in this region including the corresponding two PGLP genes. Due to the lack of functional evidence for any of these 13 putative PGLPs, we selected At5g36700 and and, after sub-cloning, ligated into the bacterial expression vector pBAD/His-A. Correctness of both fusion constructs was confirmed by DNA sequencing. While the region encoding the putative chloroplast targeting sequence was omitted in case of At5g36700 (PGLP1), the fragment for At5g47760 (PGLP2) covered the entire coding region. Affinity purification of the expressed recombinant proteins resulted in one major band of the correct size for each PGLP and several minor bands in SDS-polyacrylamide electrophoresis (not shown). Both semi-purified proteins showed PGLP activity, 420 µmol min -1 mg -1 for PGPL1 and 690 µmol min -1 mg -1 for PGPL2. Due to the different degrees of contamination by other proteins, these specific activities represent approximations.

Knockout of At5g36700 but not of At5g47760 results in leaf PGLP-deficiency
Earlier data suggested that the photorespiratory PGLP is encoded by a single gene in Arabidopsis (Somerville and Ogren, 1979). To identify this gene, we next isolated knock-out mutants from T-DNA insertion lines (SALK institute lines SALK 130835 and SALK 130837 for Better growth was achieved by elevating the CO 2 concentration to 0.3% ( Figure 1C and D), but near-to-normal growth rates require an even higher CO 2 concentration (0.9%). Under these non-photorespiratory conditions, the knockout mutant flowers and produces seeds. This is very similar to reported features of the Arabidopsis CS119 mutant (Somerville and Ogren, 1979) and strongly suggests that AtPGLP1 encodes the photorespiratory PGLP. In contrast, Atpglp2-1 knockout plants grow normally in air and do not show any visual phenotypic difference in comparison with wildtype plants ( Figure 1E).
Enzyme measurements with protein extracts from mutant plants grown under 0.9% CO 2 revealed a 97% reduction of leaf PGLP activity in Atpglp1-1 plants (0.036 ± 0.006 µmol min -1 mg protein -1 ) relative to wildtype plants (1.278 ± 0.211 µmol min -1 mg protein -1 ), which is similar to reported data for the CS119 mutant (Somerville and Ogren, 1979). In contrast, leaf PGLP activity in Atpglp2-1 (1.254 ± 0.076 µmol min -1 mg protein -1 ) was not affected by the knockout. These data correspond well with publicly available gene expression data (GENEVESTIGATOR, Zimmermann et al., 2004) which show more than twenty-fold higher transcript levels for AtPGLP1 in leaves, but much lower levels in non-photosynthesising organs, in comparison with AtPGLP2 (Supplemental Figure S1). Hence, these expression data further substantiate the role of AtPGLP1 as the photorespiratory PGLP in Arabidopsis.
The protein harbours a predicted 62 amino acid chloroplast targeting peptide (Emanuelsson et al., 2000), and the calculated molecular mass of the mature protein approximately corresponds with some (Belanger and Ogren, 1987) although not all biochemical data (for example, Hall et al., 1987).

The mutation in CS119 affects the gene At5g36700, AtPGLP1
These data from T-DNA knockout mutants strongly indicated that At5g36700 is the affected gene in CS119. To further substantiate this finding and to exactly identify the genetic lesion in CS119, we isolated genomic DNA from seeds. This source of DNA was chosen because available seeds from two different sources did not germinate anymore. By PCR, three overlapping fragments were amplified which covered the entire At5g36700 gene. Sequencing revealed a G-to-A transition in the ultimate nucleotide of exon 8 (Figure 2A). No other mutation was identified over the entire At5g36700 gene. The observed point mutation leads to an amino acid exchange from Gly 260 (GGT) to Ser 260 (AGT). We therefore wanted to exclude the possibility that this missense mutation was already present in the genetic background of the Arabidopsis plants used for the production of CS119. To this end, genomic DNA was isolated from seeds of an independent line from the same screen, CS116 (later renamed to CS8012, Somerville and Ogren, 1982). The respective region of the At5g36700 gene was amplified and sequenced (not shown); however, the obtained sequence did not show the G-to-A transition observed in CS119. This confirms our sequence evidence that the chemically induced mutation in CS119 affects the At5g36700 gene, AtPGLP1.

CS119 is a combined missense and missplicing mutant
It is not necessarily to be expected that the exchange of Gly 260 to Ser will completely abolish PGLP activity. To examine the effects of this amino acid exchange on the activity of recombinant PGLP1, we introduced the same point mutation into the corresponding wildtype cDNA by site-specific mutagenesis. Again, the use of wildtype instead of CS119 cDNA was necessary because the CS119 seeds did not germinate after storage for more than 20 years. The mutagenized PGLP1 260Ser was overexpressed in E. coli and bacterial lysates analyzed for PGLP activity. PGLP1 260Ser showed ~40% (2.462 ± 0.187 µmol min -1 mg protein -1 ) of wildtype PGLP1 260Gly activity (6.544 ± 0.27 µmol min -1 mg protein -1 ) and a ~130-fold higher PGLP activity in comparison with an empty-vector control (0.019 ± 0.003 µmol min -1 mg protein -1 ).
This shows that the Gly 260 -to-Ser exchange does not lead to full inactivation of the mature enzyme and thus cannot be responsible for the very low residual PGLP activity in CS119. On the pre-mRNA level, the mutation does not affect the canonical 5' splice donor site (GU) as it does, for example, in another photorespiratory mutant, shm1 (Voll et al., 2006). Instead, the mutation alters the first nucleotide upstream from GU at a highly conserved position of the 5' splice site consensus C(A)AG/GUAAGU (Brown, 1986;McCullough et al., 1993). The importance of the immediate upstream context of the splice donor site is apparent not only from the high degree of conservation of the respective nucleotides, but also from the analysis of other mutants (for example, Sablowski and Meyerowitz, 1998). Moreover, a NetGene2 Server (http://www.cbs.dtu.dk/services/NetGene2) analysis (Hebsgaard et al., 1996) failed to recognize the altered donor splice site in the CS 119 AtPGLP1 gene, whereas all splice sites were correctly predicted in the wildtype gene.
Because missplicing of the Atpglp1 pre-mRNA could not be directly tested with Next, four individual transgenic plants were subjected to RT-PCR analysis with genespecific primers. cDNA signals were obtained with a primer combination for exon 2 (R797) and exon 9 (R798) of AtPGLP1, which confirms expression of the transgene (left of Figure   2B). Notably, more smears can be seen with the transgenic lines in comparison with wildtype, suggesting irregular splicing. The use of a primer pair for exon 8 (R391) and exon 11 (R494) did not result in visible PCR products for most of the examined lines (right of Figure 2B). To provide direct evidence for aberrant splicing of the transgene's primary transcript, we eluted the PCR-amplified cDNA from a relatively large portion of lane t3 (square, left of Figure 2B). This procedure ensured recovery of some of the DNA smear that surrounds the major fraction of PCR products obtained with the transgenic lines. The eluted DNA was then cloned into vector pGEM-T and subjected to restriction analysis. Sequencing of three selected clones of different size (similar, larger, and smaller size in comparison with the corresponding wildtype sequence) revealed that they all represent improperly spliced transcript variants of the expressed transgene (Supplemental Figure S2). In the smaller PCR product, the entire exon 8 sequence was missing, whereas intron 8 had not been spliced out in case of the larger PCR product. In the third case, representing a PCR product of similar size relative to wildtype, the deletion of four 3'-terminal nucleotides from exon 8 would result in a frameshift. This specific splice artifact could be related to the very faint band observed for the same transgenic line with the second primer pair, R391 and R494 (right of Figure 2B lane t3). It shall be noted that sense primer R391 binds very close to the 3' end of exon 8 (Supplemental Figure S2). Absence of additional 9-to-10 nucleotides would abolish PCR amplification with this second primer pair. The comparison of the results obtained with the two different primer pairs therefore indicates that a slightly larger 3'-terminal fraction of exon 8, including at least part of the R391 binding site, is missing in a significant fraction of transcripts. Such a situation would be compatible with the presence (primers R797 and R798) and absence (primers R391 and R494), respectively, of a major fraction of RT-PCR products. While it is beyond the scope of this article to provide a more comprehensive analysis of the existing splice variants, all three sequenced transcripts showed aberrant splicing which would result in truncated PGLP1 proteins. These indirect results obtained with transgenic lines cannot absolutely exclude the possible presence of a small fraction of correctly spliced AtPGLP1 mRNA and hence some residual PGLP1 in CS119. However, they fully explain the reported very low PGLP activity that leads to the conditional lethal phenotype of this mutant.

CONCLUSIONS
In this report, we show that the entry enzyme into the photorespiratory cycle, PGLP1, is encoded by gene At5g36700 (AtPGLP1) in Arabidopsis. We have also identified the genetic defect in CS119, a PGLP-deficient Arabidopsis mutant described in an earlier landmark paper (Somerville and Ogren, 1979). As we show, this EMS mutant carries a combined missense and missplicing point mutation in the AtPGLP1 gene. The missense mutation does not inactivate PGLP1. Instead, the high-CO 2 requiring phenotype of CS119 is due to aberrant splicing of the Atpglp1 pre-mRNA. The position of the point mutation highlights the invariability of the immediate upstream nucleotide of 5'-splice sites. Most recently, protein At5g36700 has been detected in the chloroplast proteome (Peltier et al., 2006), which fully supports our conclusion.
Among the other more than ten potential PGLP genes of Arabidopsis, At5g47760 (AtPGLP2) represents the far most closely related gene to At5g36700. It encodes a putatively cytosolic enzyme, AtPGLP2, which shows PGLP activity. Knockout of this gene, however, neither affects leaf PGLP activity nor results in any apparent phenotype. Leakage of 2PG from the chloroplasts is not very likely. We therefore conclude that PGLP2 does not contribute to photorespiratory metabolism. The function of this PGLP is probably related to the metabolism of minor amounts of 2PG as they can originate from other processes than photorespiration in the cytosol of most if not all plant cells.

Plant Growth
Seeds were incubated at 4ºC for at least two days to break dormancy prior to germination.
Unless otherwise stated, plants were grown under a 12/12 h light/dark cycle (22/18°C) at 150-200 µE m -2 s -1 in Percival growth chambers. Homozygous Atpglp1-1 mutants were grown at 0.3% or 0.9% CO 2 , respectively, in a Sanyo growth cabinet equipped with a WMA-4 CO 2 control unit (PP Systems, Hitchin, UK) and adjusted to the same conditions with respect to the other growth parameters.

Site-directed Mutagenesis
The Gly 260 -to-Ser 260 exchange was introduced into AtPGLP1 cDNA by amplification with primer R821 and the mutagenizing primer R830. The resulting PCR fragment, after subcloning into pGEM-T, was excised via XhoI und BamHI restriction sites and used to replace the corresponding wildtype fragment in the AtPGLP1:pGEM-T construct carrying an internal the corresponding restriction sites of vector pBAD/His-A as above. The protein-encoding regions of all three overexpression constructs were sequenced.

Overexpression Experiments
Overexpression in recombinant E. coli cells, strain LMG194, was induced with 0.02% Larabinose. After over night-incubation at room temperature, cells were pelleted by centrifugation, resuspended in 30 mM Na-cacodylate (pH 7.6) and sonicated on ice. The supernatant obtained after centrifugation for 15 min at 20,000 g (4°C) was applied to an affinity column (Ni-NTA ProBond Slurry Matrix, Invitrogen, Karlsruhe, Germany). The column was washed with 30 mM Na-cacodylate, pH 7.6, 50 mM imidazole, and His-tagged proteins were stepwise eluted with 3 times 1 ml 250 mM imidazole in the same buffer. Proteins were analyzed in standard 12% SDS polyacrylamide gels (Laemmli, 1970).
The obtained fragments were directly sequenced to verify the insertion sites. Homozygous plants were identified by PCR with genomic DNA using two gene-specific primer pairs (R391 and R392 for Atpglp1-1, R515 and R519 for Atpglp2-1) encompassing the respective T-DNA insertion. The knock-out of AtPGLP1 and AtPGLP2, respectively, in homozygous plants of both mutant lines was verified by RT-PCR using 2.5 µg of leaf RNA for cDNA synthesis as described above. Primers R797 (sense) and R798 (antisense) resulted in a ~550 bp PCR fragment for AtPGLP1 transcripts, and primers R519 (sense) and R515 (antisense) gave a ~350 bp PCR fragment for AtPGLP2. Prior to PCR analysis, cDNA amounts were calibrated according to 435 bp signals obtained with primers R176 and R177 from the constitutively expressed At2g09990 gene encoding the 40S ribosomal protein S16.

Sequencing of Genomic DNA
Genomic DNA was isolated from CS116 and CS119 seeds according to Baumbusch et al. (2001). In short, 30 seeds were ground in liquid nitrogen and extracted in 700 reading Pwo polymerase, Peqlab, Erlangen, Germany) for CS119 using primer combinations R757/R796, R493/R494, and R391/R756 (for primer sequences compare Table 1). Primers R797 and R798 were used to amplify a corresponding genomic fragment from CS116 genomic DNA. The PCR products were cloned into plasmids pGEM-T or pUC19 and sequenced in both directions.

Transgenic Atpglp1-1 Plants for Complementation and Splicing Experiments
A 2500 bp DNA fragment was PCR-amplified (primers R813 and R814) from CS119 seed DNA and subcloned into pGEM-T. The region between (and including) start and stop codon of the Atpglp1 gene was excised with NcoI and SpeI and ligated into the corresponding sites between the 35S promoter and the nos terminator of the plant transformation vector pCAMBIA1302 (Cambia, Canberra, Australia), which was modified by excision of the originally present reporter gene. Homozygous Atpglp1-1 plants were grown in 0.9% CO 2 and transformed by the floral dip method using Agrobacterium tumefaciens GV3101 (Clough and Bent, 1998). Transgenic lines were selected from T1 seeds by their resistance against hygromycin on plates with solidified Murashige-Skoog medium. The presence of the Atpglp1 gene from CS119 was verified by restriction patterns and PCR analyses and homozygousity with respect to the T-DNA was confirmed as described above. PCR products obtained with cDNA using primers R797 and R798 (square in lane 3 of Figure 2B) were cloned into vector pGEM-T, and three clones of different size were sequenced from both sides.

Assay of PGLP Activity
For PGLP activity measurements in plants, rosette leaves from three individual plants per knockout line and wildtype, respectively, all grown at 0.9% CO 2 and sampled in the middle of the light period, were pooled and extracted in about 3 ml ice-cold 10 mM HEPES-NaOH (pH 7.0) per g fresh weight. After centrifugation for 15 minutes at 4°C and 20,000 g, protein concentration was determined (Bradford, 1976). Three independent extracts using a total of nine individual plants per line were analysed. Bacterial lysates and fractions from Ni-NTA affinity chromatography were used directly for PGLP activity assays as described by Somerville and Ogren (1979). Reactions were initiated at 25ºC with 2 mM 2PG and, after 0, 2, 4, 6, 8, and 10 minutes, aliquots were terminated by the addition of acid molybdate reagent and the released phosphate was determined photometrically (Ames, 1966).

Supplemental Data
The following materials are available in the online version of this article.  For the PCR experiment shown in the right panel, primers R391 and R494 were used. A PCR product of the expected size (363 bp) was obtained for wildtype cDNA, but not for most of the transgenic lines. Size markers are shown (one hundred bp ladder, Fermentas, St. Leon-Rot, Germany). R-codes refer to primers listed in Table 1