A New Rubisco-like Protein Coexists with a Photosynthetic Rubisco in the Planktonic Cyanobacteria Microcystis*

Two genes encoding proteins related to large subunits of Rubisco were identified in the genome of the planktonic cyanobacterium Microcystis aeruginosa PCC 7806 that forms water blooms worldwide. The rbcLI gene belongs to the form I subfamily typically encountered in cyanobacteria, green algae, and land plants. The second and newly discovered gene is of the form IV subfamily and widespread in the Microcystis genus. In M. aeruginosa PCC 7806 cells, the expression of both rbcLI and rbcLIV is sulfur-dependent. The purified recombinant RbcLIV overexpressed in Escherichia coli cells did not display CO2 fixation activity but catalyzed enolization of 2,3-diketo-5-methylthiopentyl-1-phosphate, and the rbcLIV gene rescued a Bacillus subtilis MtnW-deficient mutant. Therefore, the Microcystis RbcLIV protein functions both in vitro and in vivo and might be involved in a methionine salvage pathway. Despite variations in the amino acid sequences, RbcLIV shares structural similarities with all members of the Rubisco superfamily. Invariant amino acids within the catalytic site may thus represent the minimal set for enolization, whereas variations, especially located in loop 6, may account for the limitation of the catalytic reaction to enolization. Even at low protein concentrations in vitro, the recombinant RbcLIV assembles spontaneously into dimers, the minimal unit required for Rubisco forms I–III activity. The discovery of the coexistence of RbcLI and RbcLIV in cyanobacteria, the ancestors of chloroplasts, enlightens episodes of the chaotic evolutionary history of the Rubiscos, a protein family of major importance for life on Earth.

In photosynthetic organisms, the ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) 4 enzyme enables the fixation of inorganic atmospheric carbon dioxide into organic matter for use as a source of energy after incorporation into cellular components (1,2). Despite considerable efforts directed toward the crop improvement issue, the in vivo study of plant Rubisco has proven arduous, and it has not been possible to reconstitute flowering plant Rubisco from dissociated subunits in vitro (3).
Different types of Rubisco proteins have been described so far (3,4). The form I enzyme, typically encountered in plants, eukaryotic algae, and cyanobacteria, contains large (RbcL) and small (RbcS) subunits that have been shown to assemble into a hexadecameric structure, (L 2 ) 4 (S 4 ) 2 . The catalytic site contains active amino acids from two neighboring large subunits, and studies of site-specific enzymes indicate that the small subunit is required for maximal catalysis and contributes to CO 2 /O 2 specificity (5). Eukaryotic dinoflagellates, sulfur bacteria, and several chemoautotrophic bacteria synthesize, sometimes concomitantly with a form I enzyme, the form II Rubisco, which is constituted solely of large subunits (3). Crystallization of the Rhodospirillum rubrum form I Rubisco revealed a dimeric association (L 2 ), and larger multiples of L 2 associations have also been reported (2,6).
The sequencing of the genome of various microorganisms has unveiled the presence of genes encoding distantly related large subunits of Rubiscos. This finding led to the definition of two new classes of Rubiscos (3). Archaeal members of form III are bona fide Rubiscos because all of them revealed Rubisco activity albeit very weak in some cases (7)(8)(9)(10). For Thermococcus kodakaraensis KOD1, the three-dimensional structure proved to be an (L 2 ) 5 decamer (10,11).
Members of the form IV subfamily (Rubisco-like proteins or RLPs) do not display Rubisco activity, but the two eubacterial members of this family have been shown to play a role in sulfur metabolism. Indeed disruption of the RLP gene of Chlorobium tepidum, a green sulfur bacterium, provoked accumulation of sulfur inclusions (12). The mtnW (previously ykrW) gene of Bacillus subtilis encodes an enzyme of the methionine salvage pathway that permits recycling of methylthioadenosine, a toxic end product of polyamine synthesis, into methionine through methylthioribose (13,14). This metabolic route, first deciphered in Klebsiella pneumoniae (15), was also shown to occur in plants (16), yeast (17), and protozoal parasites (18). Although the bona fide function of Rubisco enzymes consists of a threestep sequential reaction (enolization, carboxylation/oxygenation, and final hydrolysis of the substrate ribulose 1,5-bisphosphate), 1 H NMR experiments have shown that the B. subtilis MtnW RLP only catalyzed enolization of 2,3-diketo-5-methylthiopentyl-1-phosphate, a compound with structural similarities to ribulose 1,5-bisphosphate (19). Based on the fact that the form II Rubisco of R. rubrum retained the enolase function in the methionine salvage pathway of B. subtilis, form IV RLPs have been suggested to be the ancestors of the photosynthetic Rubisco (4,19).
The planktonic Microcystis is one of the most common cyanobacterial genera found in water blooms worldwide. Sequence analysis of the partial genome of Microcystis aeruginosa PCC 7806 has unraveled the presence of two genes encoding large subunits of the Rubisco superfamily. The rbcL I XS operon encodes a typical form I Rubisco. The second gene, rbcL IV , encodes a form IV Rubisco-like protein widespread in the Microcystis genus. This study reports the expression of the rbcL I and rbcL IV genes and the functional and structural characterization of RbcL IV , a new form of Rubisco-like protein in cyanobacteria.

EXPERIMENTAL PROCEDURES
Strains, Plasmids, and Media-Stock cultures of the Microcystis strains of the Pasteur Culture Collection (PCC) were maintained in BG11 0 (20) supplemented with 2 mM NaNO 3 and 10 mM NaHCO 3 at 22°C under a light/dark regime (14 h/10 h). Light was provided by Osram Universal White fluorescent tubes under a photon flux density of 20 mol/m 2 /s (LICOR LI-185B quantum/radiometer/photometer equipped with a LICOR LI-190SB quantum sensor). Precultures were supplemented with 20 mM NaHCO 3 and bubbled with 1% (v/v) CO 2 in air at 28°C under continuous light (30 mol/m 2 /s). For transcription analyses, experimental cultures of 500 ml were obtained by dilution (1:10) of a preculture (0D 750 ϭ 0.8 -0.9) in the same medium and were incubated under the same conditions. Cells in exponential growth phase (OD 750 ϭ 0.5-0.6) were harvested by centrifugation (10,000 ϫ g, 10 min, 25°C). Cell pellets were washed and resuspended in the same volume of complete medium as described for precultures or in the same volume of sulfur-depleted medium prepared by replacing the sulfate salts with equimolar amounts of chloride ones. After a further incubation period of 15 h under the same conditions, cells were harvested by centrifugation (10,000 ϫ g, 10 min, 25°C). Cell pellets were immediately frozen in liquid nitrogen and kept at Ϫ80°C until RNA extractions.
To construct the B. subtilis mtnW deletion strain, a SmaIrestricted spectinomycin resistance cassette (21) was used. Two DNA fragments were amplified by PCR, one upstream from the mtnW gene (nucleotides Ϫ731 to Ϫ138 relative to the translational start point of mtnW, amplified with the following primers: 5Ј-CCGGAATTCCTTTTTGCAGACTGAAGGTGC-3Ј (EcoRI site underlined) and 5Ј-TCCCCCGGGCTCTGCCTG-CTCTTGATAAG-3Ј (SmaI site underlined)) and the second one downstream from the mtnW gene (nucleotides Ϫ34 to ϩ437 relative to the mtnW stop codon, amplified with the following primers: 5Ј-TCCCCCGGGCGTATGACGACTCGAA-AACC-3Ј (SmaI site underlined) and 5Ј-CGCGGATCCGCA-TTGATTGCTGCATGTGC-3Ј (BamHI site underlined)). PCR products and the spectinomycin cassette were ligated and inserted into the EcoRI and BamHI sites of pUC19 (Roche Applied Science) producing plasmid pHPP7027. Prior to transformation in B. subtilis, this plasmid was linearized at its unique ScaI site. Complete deletion of the gene was obtained by a double crossover event, giving strain BSHP7082 (trpC2 mtnW::spc).
The mtnXYZ region (starting at nucleotide Ϫ31 relative to the mtnX translation start point and ending 3 bp after the stop codon of mtnZ) was amplified by PCR using a forward primer (5Ј-GGACTAGTCGCTAGATAAATGGGGAAAG-G-3Ј) introducing a SpeI cloning site at the 5Ј-end and a reverse primer (5Ј-CGCGGATCCCGCTTATTGATTCAC-GCTGTC-3Ј) introducing a BamHI cloning site at the 3Ј-end of the fragment. This fragment was inserted into the SpeI and BamHI sites of the xylose-inducible pX plasmid (22) producing plasmid pHPP7015. Prior to transformation in B. subtilis, this plasmid was linearized at its unique ScaI site. Complete integration of the plasmid was obtained by a double crossover event at the amyE locus in the BSHP7082 strain, giving strain BSHP7075 (trpC2 mtnW::spc amyE::mtnXYZ).
The replicative plasmid pDG148 (23) with its isopropyl ␤-Dthiogalactopyranoside (IPTG)-inducible promoter was used to clone the rbcL IV gene from M. aeruginosa PCC 7806. The rbcL IV gene was amplified from chromosomal DNA by PCR using the forward primer 5Ј-ACGCGTCGACAAGGAGGTACCTTTTA-TGACTATAATTGTCG-3Ј (SalI site underlined and introduced ribosome binding site in bold) and the reverse primer 5Ј-ACAT-GCATGCGAATCTAGCTTAACCCCACTC-3Ј (SphI site underlined). The amplified fragment after digestion was inserted into the SalI and SphI sites of pDG148, producing plasmid pHPP7032. The BSHP7075 strain was transformed with this plasmid giving strain BSHP7080 (trpC2 mtnW::spc amyE:mt-nXYZ pDG148-rbcL IV ). All restriction enzymes used were from Roche Applied Science.
B. subtilis cells were transformed with plasmid DNA following the two-step protocol described previously (24). Transformants were selected on Luria-Bertani (LB) solid medium containing the appropriate antibiotics at the following concentrations: 100 g/ml for spectinomycin, 5 g/ml for chloramphenicol, or 5 g/ml for kanamycin.
Escherichia coli Strains-The E. coli strains used were TG1 and XL1-blue (laboratory collection) for the construction of pHPP7032 and of pHPP7027 and pHPP7015, respectively; JM109 or GM48 (Promega Corp., Madison, WI) for the expression vector constructs; and BL21 or Rosetta(DE3)pLysS (Novagen, Darmstadt, Germany) for protein expression. E. coli strains were grown at 37°C on LB solid medium and in aerobic liquid cultures following standard procedures (25) with the appropriate antibiotics at the following concentrations: 100 g/ml for ampicillin, 25 g/ml for kanamycin, or 100 g/ml for spectinomycin. The expression vectors were pET-43.1a(ϩ) and pET-28a(ϩ) (Novagen, Darmstadt, Germany).
Gene Identification-Using the protein sequence of form I Rubisco from Synechocystis sp. PCC 6803 as a query, comparison analysis using the BLAST algorithm was performed against a data base containing the partial genome sequence of M. aeruginosa PCC 7806. This non-public data base forms part of a current sequencing genome project at the Pasteur Genopole-Ile de France (Institut Pasteur, France) and is the preliminary assembly of shotgun sequencing of short fragments of whole genomic DNA cloned in the pcDNA-2.1 vector (Invitrogen).
Nucleic Acid Extraction, Screening by PCR, and Transcription Analysis-M. aeruginosa PCC 7806 genomic DNA and RNA were extracted as described previously (26). Isolation of plasmid DNA from recombinant E. coli cells was performed with the alkaline extraction method (25) for checking the transformants or using the AX500 Nucleobond kit (Qiagen, Hilden, Germany) for large scale purification.
Twenty-five Microcystis strains of the PCC were assayed for the presence of the rbcL I and rbcL IV genes by using PCR on 1 l of cryolysates prepared as follows. Cell pellets (2 ml of stock cultures at OD 750 of 0.5) were frozen for 1 min in liquid nitrogen followed with 1-min incubation in boiling water. This was repeated 8 -10 times to obtain cell lysis. The oligonucleotide primers used were: 5Ј-GGTATCCACTTCCGCGTTTT-3Ј and 5Ј-GGTTCCACCACCGAACTGTA-3Ј for rbcL I and 5Ј-AAA-ACCCGACGACAACAATC-3Ј and 5Ј-CTGGAGCAGGAA-AAGTGGTG-3Ј for rbcL IV . One microliter of cryolysate, 10 pmol of each primer, a 250 M concentration of each deoxynucleoside triphosphate, and 1 unit of Taq polymerase (Promega Corp.) in 1ϫ buffer were mixed and subjected to an initial step of 94°C for 5 min followed by 30 cycles of 94°C for 1 min, 60°C for 30 s, 72°C for 2 min, and a final elongation step of 72°C for 7 min in a 9700 PerkinElmer Life Sciences thermocycler (Applied Biosystems, Foster City, CA). Ten microliters of PCR product were analyzed by gel electrophoresis on 1.8% (w/v) agarose in 1ϫ Tris-borate-EDTA buffer and stained with ethidium bromide. The gels were photographed under UV light with an ImageMaster VDS-Amersham Biosciences FTI-500 imaging system.
The amounts of rbcL I and rbcL IV transcripts relative to the amounts of transcripts of rnpB (catalytic subunit of RNase P) in sulfur-replete and -depleted cells of M. aeruginosa PCC 7806 were determined by semiquantitative RT-PCR. Each RT primer (2 pmol) was annealed to 5 g of RNA in the presence of 1 mM dNTP mixture under a volume of 10 l. The mixture was heated at 70°C for 5 min, slowly cooled down to 42°C in about 15 min (Ϫ2°C/min), and then placed on ice for at least 1 min. Reverse transcription, using oligonucleotide primers 5Ј-GGT-ATCCACTTCCGCGTTTT-3Ј for rbcL I , 5Ј-AAAACCCGAC-GACAACAATC-3Ј for rbcL IV , and 5Ј-AACCTTTGTCCCTC-CACCTT-3Ј for rnpB, was carried out in SuperScript II buffer containing MgSO 4 (1.25 mM), dithiothreitol (10 mM), and RNaseOUT recombinant RNase inhibitor (Invitrogen). The mixture was heated at 42°C for 2 min. After adding 200 units of SuperScript II reverse transcriptase (Invitrogen) the mixture was incubated for 60 min, and the reaction was inactivated by heating at 70°C for 15 min. The control for DNA contamination consisted of the same reaction mixture except that Super-Script II reverse transcriptase was replaced by H 2 O. A total of 2 l (0.5 g of initial RNA) of the RT reaction mixture or 20 ng of genomic DNA as positive controls were used for subsequent PCR using the following oligonucleotide primers: 5Ј-CCGCG-TTTTAGCTAAGTGCT-3Ј and 5Ј-GGAAGCGTAGTCTTG-GGTGA-3Ј for rbcL I , 5Ј-GAAAGCTACGGCACAAAAGC-3Ј and 5Ј-TAAGCCAGCAAAAGCCACTT-3Ј for rbcL IV , and 5Ј-ACCCTTACCCCCAATCAGTT-3Ј and 5Ј-CGTGAGGATA-GTGCCACAGA-3Ј for rnpB. Samples were taken at successive PCR cycles. Each sample was analyzed by gel electrophoresis on 1.5% (w/v) agarose in 1ϫ Tris-borate-EDTA buffer, the gels were photographed under UV light as described above, and the gel image was quantified using the free software ImageJ (Millersville University).
Plasmid Constructs, Overexpression, and Purification of the Recombinant Proteins-The rbcL IV gene from M. aeruginosa PCC 7806 was amplified from chromosomal DNA, cloned in both the pET-43.1a(ϩ) and pET-28a(ϩ) systems by PCR amplification using a proofreading polymerase, and verified by sequencing. For cloning in the expression vector pET-28a(ϩ) (Novagen, Darmstadt, Germany), the primers used were 5Ј-AAACCCCGGGGCAGCCATATGAGAATTTGTATTT-TCAGGGTGCTAGCATGACTATAATTGTCGATTAT-CGC-3Ј (SmaI, NdeI, and NheI sites underlined; tobacco etch virus protease cleavage site in bold) and 5Ј-TGCG-GTCGACTTTACTCGAGACCCCACTCTAGAATTGC-CTG-3Ј (SalI and XhoI sites underlined). The PCR product and the plasmid pET-28a(ϩ), digested with NdeI and SalI, were ligated. The ligation mixture was transformed by electroporation in E. coli GM48 cells, and the construct was then transferred by electroporation to Rosetta(DE3)pLysS (Novagen, Darmstadt, Germany), a strain provided for "universal" translation. The recombinant plasmid pET-28a:rbcL IV was introduced into BL21(DE3). After IPTG induction, Histagged RbcL IV was purified in one step on nickel-nitrilotriacetic acid resin (Novagen, Madison, WI) according to the manufacturer's instruction.
For cloning in the expression vector pET-43.1a(ϩ) (Novagen, Darmstadt, Germany), the rbcL IV gene from M. aeruginosa PCC 7806 was amplified from chromosomal DNA by PCR using the forward primers 5Ј-AATGACTATAATTGTCGATTATCGC-3Ј (no restriction site) and the reverse primer 5Ј-CGGGATCCTTA-ACCCCACTCTAGAAT-3Ј (BamHI site underlined). The PCR product was digested with BamHI (Invitrogen) and ligated to pET-43.1a(ϩ) digested with BamHI and PshAI (Amersham Biosciences). The ligation mixture was transformed in E. coli JM109 by a heat shock procedure (Promega Corp.), and the construct was transferred by electroporation to E. coli BL21 (Novagen, Darmstadt, Germany). Two liters of LB medium were inoculated at A 600 ϭ 0.1 with E. coli cells containing the recombinant plasmid pET-43.1a:rbcL IV . Upon reaching A 600 ϭ 0.5, IPTG was added to a final concentration of 1 mM to induce gene expression for 4 h. The cells were harvested by centrifugation (8,000 ϫ g, 15 min, 4°C) and kept at Ϫ20°C until purification of the protein. The cell pellet was resuspended in 25 mM Tris-HCl buffer, pH 8.3, 5 mM dithiothreitol and sonicated on ice. After centrifugation (17,000 ϫ g, 15 min, 4°C), the soluble supernatant was adjusted to a concentration of 800 mM (NH 4 ) 2 SO 4 and loaded on a HiPrep 16/10 Butyl FF column (Amersham Biosciences) pre-equilibrated in 25 mM Tris-HCl buffer, pH 8.3. A step gradient of decreasing (NH 4 ) 2 SO 4 concentration (800-0 mM (NH 4 ) 2 SO 4 in 25 mM Tris-HCl, pH 8.3) at a flow rate of 5 ml/min was applied, and the NusA-RbcL IV fusion protein (100 kDa) was eluted at 100 mM (NH 4 ) 2 SO 4 . Digestion with 1 unit of thrombin protease (Novagen, Madison, WI)/mg of protein was performed overnight in 1ϫ buffer (Novagen, Madison, WI) at room temperature. The sample was desalted and concentrated using an Amicon Ultra-15 centrifugal filter device (Millipore) prior to separation of the cleaved NusA (58 kDa) and RbcL IV proteins (42 kDa) on a Mono Q HR 5/5 column (Amersham Biosciences) using an increasing gradient of (NH 4 ) 2 SO 4 (0 -500 mM in 25 mM Tris-HCl, pH 8.3) at a flow rate of 1 ml/min. RbcL IV and NusA were eluted at 75 and 500 mM (NH 4 ) 2 SO 4 , respectively. Purified RbcL IV was finally applied to a Superdex 75 HR 10/30 column (Amersham Biosciences) in 25 mM Tris-HCl, pH 8.3, 150 mM (NH 4 ) 2 SO 4 at a flow rate of 0.2 ml/min. The protein content and purity of each fraction were visualized by SDS-PAGE on 12% acrylamide (acrylamide/bisacrylamide, 37.5:1) stained with Coomassie Blue in 40% (v/v) methanol, 10% (v/v) acetic acid (27).
The enolase activity of M. aeruginosa PCC 7806 RbcL IV was also analyzed by complementation of the mtnW-disrupted B. subtilis mutant by rbcL IV . The wild type, mtnW-disrupted, and MArbcL IV ϩ B. subtilis were grown on 3 ml of LB medium until A 600 ϭ 1.0. The cells were collected by centrifugation at 3,000 ϫ g, washed five times with 10 mM Tris-HCl (pH 7.1) containing 10 mM MgCl 2 , 10 mM CaCl 2 , and 100 mM NaCl, and transferred to 5 ml of the sulfur-free minimal medium (19) containing 1 mM IPTG. Antibiotics were added at the following concentrations: spectinomycin, 100 g/ml; chloramphenicol, 5 g/ml; and kanamycin, 5 g/ml. Cultures were shaken at 250 rpm and 37°C. Cell growth was monitored by measuring A 600 .
Spectroscopic Analysis of the Recombinant RbcL IV -The far-UV circular dichroism spectra were monitored at 20°C in a circular cell (path length ϭ 0.1 mm) on a Jasco JA-810 spectropolarimeter between 180 and 260 nm at 2-nm constant bandwidth with a 0.5-nm step and an integration time of 2 s. The pure RbcL IV (1 mg/ml) was extensively dialyzed against phosphate buffer (20 mM Na 2 HPO 4 , 20 mM NaH 2 PO 4 , pH 8.0) prior to analysis. The data were an aver-age of five scans recorded at a scanning speed of 10 nm/min. The base line was acquired on dialysis buffer under the same conditions and subtracted from the sample spectrum. Quantitative secondary structure was determined using the CD-Pro software (OlisGlobal Works, Oxford, UK) including SELCON3, CONTIN/LL, and CDSSTR algorithms.
The fluorescence emission of Trp following excitation at 280 or 290 nm was monitored on a Kontron SFM25 spectrofluorometer at 25°C in 25 mM Tris-HCl buffer, pH 8.3. The excitation and emission slit widths were set at 15 and 10 nm, respectively, and the spectra were recorded between 300 and 400 nm. Fluorescence emission was measured for a concentration of pure RbcL IV of 10 g/ml in the absence or presence of denaturant (8 M urea). A blank spectrum of buffer alone was subtracted from all spectra.
Sedimentation equilibrium experiments were carried out in a Beckman XL-A analytical ultracentrifuge with an AN60Ti rotor at 20°C in 25 mM Tris-HCl (pH 8.2), 150 mM (NH 4 ) 2 SO 4 . Different protein concentrations (75, 125, 250, 500, and 1000 g/ml) were investigated at two speeds: 13,000 and 15,000 rpm for the lower protein concentrations, and 12,000 and 15,000 rpm for the higher protein concentrations (density, 1.003). Each speed was maintained until reaching equilibrium (over 20 h). The sedimentation profiles were recorded using absorbance optics simultaneously at 230 and 235 nm for 75 and 125 g of protein/ml, 235 and 280 nm for 250 g of protein/ml, 280 nm for 500 g of protein/ml, and 260 nm for 1000 g of protein/ml. The base line was measured at 55,000 rpm. Sedimentation equilibrium data were evaluated with the programs provided by Beckman and fitted with model ideal1, assuming a unique molecular species and using a partial specific volume of the protein (v ) of 0.744 ml/g as calculated from protein sequence.
Amino Acid Sequence Analysis/Molecular Modeling-Sequence analysis was performed using PSI-BLAST (28) as well as fold recognition methods (29,30). The most similar sequences found in the Protein Data Bank were Rubisco-like proteins from the archaeum T. kodakaraensis (1GEH; 27% over 442 amino acids) and from the bacterium C. tepidum (1TEL; 28% over 336 amino acids). As this last sequence also corresponds to a type IV RLP, it was chosen as a template for modeling the dimeric structure of M. aeruginosa PCC 7806 RbcL IV . Alignments were refined considering the sequences of forms I and II Rubiscos and using the sensitive hydrophobic cluster analysis method (31,32) and were submitted to Modeler 6.2 (33) for comparative modeling. Resulting models were evaluated using Verify3D (34) and PROSA (35). Both RbcL I and RbcL IV are simultaneously synthesized in M. aeruginosa PCC 7806 cells (data not shown). Members of form IV RLPs have been proposed to play a role in sulfur metabolism (12)(13)(14)19); the possibility of a sulfur regulation of the rbcL genes was thus explored by using semiquantitative RT-PCR. As shown in Fig. 1, the rbcL IV transcripts are 22-fold more abundant in sulfur-depleted than in sulfur-replete cultures. Conversely a 5-fold decrease of the rbcL I transcript abundance is observed under the same conditions.

M. aeruginosa PCC 7806 Genome Carries Two Rubisco Large
Oligonucleotide primers designed on strain PCC 7806 rbcL I and rbcL IV sequences were used to test by PCR whether other strains of the Microcystis genus also contained two rbcL genes. Indeed a rbcL I product was amplified for all the strains tested, and a rbcL IV product was amplified for 22 of the 25 Microcystis strains of the PCC. The results for each strain are listed in supplemental Table S2. Some sequence divergence may have prevented amplification for the three strains PCC 9804, PCC 9805, and PCC 10025. The presence of counterparts of rbcL IV was also investigated by BLAST comparison in all the available cyanobacterial genome data bases. In each of them, a unique sequence displayed 22-24% identity to RbcL IV and more than 90% identity to RbcLs of form I Rubiscos. This renders the planktonic water bloom former of the genus Microcystis rather unique in this respect.
The Recombinant RbcL IV Protein Catalyzes an Enolase Reaction-To determine the enzyme activity of the M. aeruginosa PCC 7806 RbcL IV , the corresponding gene was cloned in the pET-28a(ϩ) system, and the purified His-tagged protein was tested in in vitro assays. The M. aeruginosa PCC 7806 RbcL IV does not catalyze carboxylation of ribulose 1,5-bisphosphate (data not shown) but is capable of 2,3-diketo-5-methylthiopentyl-1-phosphate enolization (Fig. 2). The K m value for the substrate was calculated as 13 mM and was similar to that for the Bacillus counterpart. 5 To confirm that M. aeruginosa PCC 7806 RbcL IV can function in vivo, we performed a complementation test using a B. subtilis RLP-deficient mutant (mtnW Ϫ ). Making use of the methionine salvage pathway, wild type B. subtilis can grow on a medium with methylthioadenosine as the sole source of sulfur, but a mtnW Ϫ mutant cannot (Fig. 3). After transformation, the M. aeruginosa PCC 7806 rbcL IV gene rescues the growth of the mtnW Ϫ mutant on methylthioadenosine to some extent (Fig. 3). The observed inefficient rescue of the mtnW Ϫ mutant by the rbcL IV gene from M. aeruginosa PCC 7806 might be due to a low level of expression of the rbcL IV gene in the mutant. These results demonstrate that M. aeruginosa PCC 7806 RbcL IV can catalyze the 2,3-diketo-5methylthiopentyl-1-phosphate enolase reaction both in vitro and in vivo.
Because enolization of 2,3-diketo-5-methylthiopentyl-1phosphate is a key step in the methionine salvage pathway in bacteria (14), a BLAST search for putative counterparts of the genes encoding other enzymes of this pathway in bacteria was conducted using the genome sequence of M. aeruginosa PCC 7806 presently available. With the exception of speD, all the genes have been found (see supplemental Table S3). The methionine salvage pathway is thus most likely to be present in M. aeruginosa PCC 7806.
The Recombinant RbcL IV Is a Dimer-Because the expression of the RbcL IV protein is low with the pET-28a(ϩ) system, the M. aeruginosa PCC 7806 rbcL IV gene was cloned in the pET-43.1a(ϩ) system. The resulting NusA-RbcL IV protein of 100 kDa was recovered in the soluble fraction and purified (data not shown). The pure RbcL IV devoid of NusA was then used for further biochemical studies.
The secondary structure composition of the RbcL IV protein was estimated by deconvolution of the circular dichroism spec-5 Y. Saito, H. Ashida, A. Danchin, and A. Yokota, in preparation. The intensity of the RT-PCR products for rbcL IV (squares), for rbcL I (triangles), and for rnpB (circles) was plotted versus the PCR cycle number. The relative amount of RNA was calculated using the formula A ϭ Y/(1 ϩ E) n in which Y is the relative amount of PCR product, 1 is the theoretical efficiency of the reaction (curve slope), E the experimental efficiency of the reaction, and n the PCR cycle number. trum acquired in the far-UV region (see supplemental Fig. S4 and Table S5). A consensus average of 38% ␣-helix and 12% ␤-strand was obtained.
Information on the tertiary structure of the RbcL IV protein was gathered by steady state fluorescence emission (see supplemental Fig. S6) using as spectroscopic probes the two tryptophan residues present along the sequence (Trp 31 and Trp 385 ). The fluorescence spectra of native and denatured proteins indi-cate that both tryptophan residues of RbcL IV are located within a non-polar environment and that the 12 tyrosine residues present along the amino acid sequence participate with only a minor contribution to the overall fluorescence emission. A potential transfer of energy from tyrosine to tryptophan residues in the monomer or between neighbor monomers, however, cannot be ruled out. Sedimentation equilibrium analyses of pure RbcL IV were performed at different concentrations (see supplemental Fig. S7 and data not shown). Data fitted to model ideal1 implying the presence of a unique protein species. The predicted molecular mass of 87 kDa is consistent with a dimeric state of the protein. Even at concentrations as low as 75 g/ml, no monomer could be detected.
Amino Acid Sequence Comparison and Three-dimensional Structure Modeling-An amino acid alignment of the form IV RLPs was performed and compared with the corresponding part of the amino acid sequences of form I RbcLs of Synechococcus sp. PCC 6301 and M. aeruginosa PCC 7806. As shown in Table 1    To get deeper insights into the M. aeruginosa PCC 7806 RbcL IV protein, alignments of its sequence with those of forms I and II Rubiscos was refined using the hydrophobic cluster analysis method, and its three-dimensional structure was modeled on the basis of the alignment of its sequence with that of the RLP from C. tepidum (36) Fig. 4). Lys 334 has been proposed to play an important role, along with Mg 2ϩ , in the stabilization of one reaction intermediate arising from the reaction of carboxylation/oxygenation of enol-ribulose 1,5bisphosphate (for a review, see Ref. 37). A change in amino acid residue at that position may thus have major consequences on the function of the enzyme. The secondary structural content of the three-dimensional structure model (33% ␣-helices and 20% ␤-strands) was consistent with those obtained from deconvolution of the circular dichroism spectrum.

DISCUSSION
This study is the first report describing a microorganism containing both a typical photosynthetic form I Rubisco and a form IV RLP that catalyzes enolization of 2,3-diketo-5-methylthiopentyl-1-phosphate, a key substrate in the methionine salvage pathway (14). In addition, the genome of M. aeruginosa PCC 7806 carries putative orthologs of the genes involved in this pathway in phylogenetically distant bacteria. To the best of our knowledge, this constitutes the first straightforward evidence for such a metabolic pathway in photosynthetic bacteria. Although a C. tepidum mutant defective for a form IV RLP has revealed impairment of sulfur metabolism (12,38), data so far obtained indicate that RLP in this organism plays a distinct physiological role from that proposed for MtnW in Bacillus. Taken together, our results argue in favor of a role of RbcL IV in Microcystis similar to that of MtnW in Bacillus. However, whether the methionine salvage pathway is functional in Microcystis remains to be demonstrated.
A methionine salvage pathway has not been previously reported in cyanobacteria, but such a pathway is most likely a general property of Microcystis as indicated by the occurrence of the RLP gene rbcL IV in a number of strains belonging to this genus. This may reflect adaptation of Microcystis cellular metabolism to environmental conditions that cells encounter during their life cycle, in particular during water bloom periods and surface scum formation with exposure to high light intensity, O 2 concentration, and temperature. The selective advantage represented by the capacity to recycle methylthioribose, a toxic by-product of polyamine synthesis excreted by bacteria (13), and simultaneously to consume oxygen could protect cells against oxidative stress and contribute to the success of Microcystis in colonizing aquatic ecosystems worldwide (39). The current absence of genes encoding RLPs in the cyanobacterial genomes publicly available indicates that, in this respect, Microcystis is rather unique in this phylum, but it is worth noting that none of these other cyanobacteria share a similar life style. Sequencing of new genomes of cyanobacteria, in particular those forming mats at the surface of which cells are exposed to atmospheric oxygen and intense sunlight in a way similar to Microcystis cells in surface scums, may reveal the presence of new RLP orthologs.
The bona fide function of the Rubisco enzyme consists of a three-step sequential reaction: 1) enolization, 2) carboxylation/ oxygenation, and 3) hydrolysis of the substrate ribulose 1,5bisphosphate. The first event in catalysis corresponds to the activation of the enzyme through carbamylation of Lys 201 (i.e. binding of CO 2 to the ⑀ amino group). In this reaction, chelation of metal ions, such as Mg 2ϩ , by Asp 203 and Glu 204 cations enables entry and correct positioning of the substrate, ribulose 1,5-bisphosphate, into the catalytic pocket. The C-3 proton of the substrate is subsequently abstracted by the Lys 201 carbamate with Lys 175 and His 294 acting as general bases promoting conversion of ribulose 1,5-bisphosphate into the intermediate enediol (for reviews, see Refs. 37 and 40). The Microcystis and B. subtilis form IV RLPs represent a related enzyme, sharing with Rubiscos some degree of substrate specificity, because 2,3diketo-5-methylthiopentyl-1-phosphate is structurally similar to ribulose 1,5-bisphosphate. These RLPs, however, allow solely enolization (this study and Ref. 19).
Biochemical and structural studies have shown that the secondary structure of the Microcystis RbcL IV is much better conserved than is its primary structure with content in ␣-helices and ␤-strands consistent with those observed within crystal structures of Rubiscos of different bacterial families (see supplemental Fig. S4 and Table S5) (10,36,41,42). At least some elements of the tertiary structure of RbcL IV are also similar to what is observed for other members of forms I and III Rubiscos in which two tryptophan residues (Trp 31 and Trp 385 in the Microcystis amino acid sequence) are present at equivalent positions ( Fig. 4 and Ref. 10). Pure Microcystis RbcL IV spontaneously assembles into dimers whatever the protein concentration (see supplemental Fig. S7) and possesses an enolase activity (Fig. 2) indicating that this oligomerization state represents the functional unit. The form IV RLPs and the form II Rubiscos, despite functional divergence, share the characteristics of associating into stable dimers in vitro, although the existence of higher degrees of oligomerization in vivo cannot be dispelled (36,42). In addition Rubisco forms I and III assemble into L 8 S 8 octamers and (L 2 ) 5 decamers, respectively, but the minimal unit necessary for the carboxylase/oxygenase activity in these enzymes is also a dimer with a catalytic site constituted of amino acids from two neighboring subunits (3,10,43). Taken together these observations argue in favor of the similarity of the mechanisms driving enolization in all forms of Rubiscos and RLPs.
The three-dimensional model we have constructed provides a structural basis for understanding the particular catalytic activity of the Rubisco and RLP enzymes by attempting to identify the set of active residues specifically involved in the enolization reaction. Indeed after a sensitive analysis allowing the accurate alignment of RbcL IV with known three-dimensional templates (the sequences of RbcL IV and three-dimensional templates share less than 30% identity, a level for which automatic alignment procedures do not furnish accurate alignments in many places), we have shown that despite a common structural core with bona fide Rubisco, RbcL IV possesses singularities within its active site that may explain the differences observed in the catalytic activity of the enzyme (Fig. 4). These  8RUC, 9RUB, and 1TEL, respectively). Amino acid numbers on top of the sequence alignment refer to their position in the 1TEL sequence. The observed regular secondary structures of 1TEL and 9RUB are shown above and below their sequences, respectively. Identical amino acids are shown in white on a black background, similarities are boxed, and gray shading is used for positions occupied by hydrophobic amino acids. The amino acids depicted in the three-dimensional view of Fig. 5 are shown with asterisks. This figure was prepared using ESPript (48).
plasts, underscores fundamental implications on the evolutionary history of the Rubisco superfamily and modulates previous discussions on this topic. Indeed the current hypothesis concerning the methionine salvage pathway was that bacteria that emerged early (such as B. subtilis) would function with a RLP, whereas organisms that appeared later in evolution, including cyanobacteria, would use a phylogenetically unrelated enzyme, an enolase/phosphatase (4). In addition correlate regulation by sulfur of both form I Rubisco and form IV RLP in Microcystis may reflect the initial relatedness of the reaction catalyzed by both enzymes and in itself opens new fascinating insights into mechanisms of regulation in cyanobacteria. This is most likely to be different from the control in B. subtilis (47), the only other organism for which regulation by sulfur of a RLP gene, mtnW, has been evidenced and relies on S-box riboswitches not found in Microcystis.
The present work puts forward clues to assess the catalytic relationships between bona fide Rubisco enzymes and RLPs. Although the availability of a primitive type of enzyme offers the rare opportunity to dissect the different steps of a catalytic mechanism, our study also provides the first evidence of the missing link between RLPs and Rubiscos and sheds light on the as yet still obscure evolution of this family of proteins.