Proximal Cys172 and Cys192 in the large subunit of the photosynthetic enzyme Rubisco (ribulose-1,5-bisphosphate carboxylase/oxygenase; EC 18.104.22.168) are evolutionarily conserved among cyanobacteria, algae and higher plants. Mutation of Cys172 has been shown to affect the redox properties of Rubisco in vitro and to delay the degradation of the enzyme in vivo under stress conditions. Here, we report the effect of the replacement of Cys172 and Cys192 by serine on the catalytic properties, thermostability and three-dimensional structure of Chlamydomonas reinhardtii Rubisco. The most striking effect of the C172S substitution was an 11% increase in the specificity factor when compared with the wild-type enzyme. The specificity factor of C192S Rubisco was not altered. The Vc (Vmax for carboxylation) was similar to that of wild-type Rubisco in the case of the C172S enzyme, but approx. 30% lower for the C192S Rubisco. In contrast, the Km for CO2 and O2 was similar for C192S and wild-type enzymes, but distinctly higher (approximately double) for the C172S enzyme. C172S Rubisco showed a critical denaturation temperature approx. 2 °C lower than wild-type Rubisco and a distinctly higher denaturation rate at 55 °C, whereas C192S Rubisco was only slightly more sensitive to temperature denaturation than the wild-type enzyme. X-ray crystal structures reveal that the C172S mutation causes a shift of the main-chain backbone atoms of β-strand 1 of the α/β-barrel affecting a number of amino acid side chains. This may cause the exceptional catalytic features of C172S. In contrast, the C192S mutation does not produce similar structural perturbations.
- Chlamydomonas reinhardtii
- chloroplast mutant
- proximal cysteine
- ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco)
- specificity factor
- X-ray crystallography
The bifunctional enzyme Rubisco (ribulose-1,5-bisphosphate carboxylase/oxygenase; EC 22.214.171.124) operates at a metabolic branch point channelling carbon either to the photosynthetic carbon fixation (Calvin–Benson) cycle or to the photorespiratory pathway . The enzyme converts RuBP (ribulose 1,5-bisphosphate) into a reactive enediolate form, which is prone to attack by either CO2 or O2. The resulting carboxylated or oxygenated intermediate is hydrolysed in two further steps (hydration and C2–C3 bond cleavage), producing two molecules of 3-phosphoglycerate or one each of 3-phosphoglycerate and 2-phosphoglycolate respectively as final products . Because the carboxylation and oxygenation reactions share the same enediol substrate, the oxygenation reaction may be an unavoidable consequence of the catalytic mechanism of the enzyme [3,4], which evolved under anoxic conditions. The intrinsic preference of Rubisco for one or the other reaction is measured by the so-called specificity factor [Ω (CO2/O2 specificity factor)], which is defined as the ratio of the Vc (Vmax for carboxylation)] to the Km for CO2 [Kc (Michaelis constant for CO2)], divided by the ratio of the Vo (Vmax for oxygenation) to the Km for O2 [Ko (Michaelis constant for O2)] . The value of Ω (VcKo/VoKc) is known to vary among Rubisco enzymes from different species . Because of the relevance to plant growth of the photosynthesis/photorespiration ratio, the beneficial manipulation of crop-plant Ω has been an unfulfilled biotechnological goal.
The crystal structures of the active enzyme have been solved for a number of species including prokaryotes, algae and land plants . The most common form of Rubisco (type I present in cyanobacteria, green algae and higher plants) is composed of eight chloroplast-encoded large subunits (∼55 kDa) and eight nuclear-encoded small subunits (∼15 kDa). Two large subunits assemble into functional dimers with two active sites at the dimer interface. Catalytic competence requires carbamoylation by CO2 of the ϵ-amino group of a specific lysine residue in the active site, and subsequent binding of an Mg2+ ion to the carbamoyl group . Four small subunits cap the top and bottom of four pairs of the large subunits to form the hexadecameric holoenzyme. The knowledge of Rubisco structure has fostered an understanding of the central features of the catalytic mechanism and other properties of Rubisco, but the structural determinants of the variations in Ω have remained elusive.
Analysis of Rubisco mutants has been helpful in evaluating the functional role of the substituted residues or deleted sequences [1,9–12]. Many of these studies have been carried out in Chlamydomonas reinhardtii, a unicellular green alga that allows for straightforward genetic manipulation of both the chloroplastic and nuclear genomes. Because of their singular redox properties, cysteine residues are potential candidates for harbouring regulatory effects and may therefore be considered as primary targets for site-directed mutagenesis. In fact, it is known that the specific oxidative modification of cysteine residues by disulfide exchange causes inactivation and increases proteolytic susceptibility of the purified Rubisco in vitro [13–15]. These findings suggest a physiological role of cysteine oxidation in Rubisco catabolism. Whereas the critical cysteine residues have not been unambiguously identified, the conserved Cys172 residue from the Rubisco large subunit has attracted some attention [16,17]. Furthermore, specific modification of this cysteine residue by an affinity label has been shown to abolish enzymatic activity , suggesting a yet undefined role for this residue in catalytic performance. Replacement of Cys172 by serine in C. reinhardtii yields an active enzyme but eliminates arsenite sensitivity, shifts the critical redox potential that controls proteolytic susceptibility in vitro, and increases resistance to stress-induced degradation in vivo . A slower catabolism of Rubisco under stress has also been reported after replacement of the same residue in a prokaryotic Rubisco , whereas the replacement of the adjacent Cys192 or other conserved cysteine residues (Cys247) did not delay the stress-induced degradation .
Analysis of the crystal structures of Rubisco enzymes from different sources reveals that Cys172 and Cys192 are close to each other, but in most cases, they are not close enough to form a disulfide bond . In one of the structures of the non-carbamoylated enzyme from tobacco , a disulfide bond is observed, whereas a second structure, obtained under similar conditions, does not feature the 172–192 disulfide bond . Structures of activated (carbamoylated) Rubisco [21–23] show residues 172 and 192 as free cysteine residues. In any case, the proximity of the cysteine residues is likely to support redox interactions that may have structural, catalytic and/or regulatory relevance. In an attempt to define the structural and functional roles of these proximal cysteine residues, they have been replaced by site-directed mutagenesis and chloroplast transformation of C. reinhardtii. The C172S and C192S mutant Rubisco enzymes have been analysed in detail with respect to catalytic function and their X-ray crystal structures have been solved.
Strains and cell cultures
C. reinhardtii strain C172S, with a Cys172 to serine replacement in the Rubisco large subunit, has already been described . Strain C192S was created by site-directed mutagenesis and chloroplast transformation as described below. Strain 2137 mt+ is the wild-type . Non-photosynthetic strain 18-7G, which has a nonsense mutation in the rbcL large-subunit gene, was used as the host for transformation . All strains were kept on standard solid medium supplemented with acetate  at 25 °C in the dark. For Rubisco purification, 1.5 litre flasks of standard liquid culture medium supplemented with acetate were inoculated with the appropriate strains and the cells were grown at 28 °C in the light (3×30 W white light tubes at 30 cm, supplying approx. 10 μE·m−2·s−1) with moderate (120 rev./min) shaking up to a density of 107 cells·ml−1. The cells were collected by centrifugation (3000 g for 5 min) and the pellet was frozen in liquid nitrogen and stored at −80 °C.
Directed mutagenesis and chloroplast transformation
Plasmid prbclP containing an HpaI fragment (bases −743 to +1927), which encompasses the entire rbcL coding region (bases +1 to +1428), cloned into the SmaI site of a pBluescript SK- (Strata gene) vector, was used as a source of the C. reinhardtii Rubisco large-subunit chloroplast gene. Site-directed mutagenesis of Cys192 to serine was achieved by changing TGT (Cys192) to AGC, using the QuikChange® kit of Stratagene. This substitution created a new HindIII restriction site (between bases +573 and +574) that expedited the screening of transformants. The mutated plasmid (called pLS-C192S) was used for chloroplast transformation via projectile bombardment of C. reinhardtii strain 18-7G according to published procedures [16,26]. Transformation of the non-photosynthetic 18-7G mutant with pLS-C192S yielded photosynthetically competent colonies on minimal medium. The transformants became homoplasmic (as checked by HindIII restriction analysis of PCR-amplified rbcL) after successive rounds of single-colony isolation, replating in selective minimal medium and growth under light. Finally, the entire coding region of rbcL was sequenced to confirm that only the intended mutation was present. The strain thus created was named C192S.
Purification of Rubisco
Rubisco was extracted and purified by 35–60% (w/v) ammonium sulfate precipitation, sucrose-gradient centrifugation and anion-exchange chromatography according to a published procedure . Alternatively, for the determination of kinetic constants, Rubisco was purified following a somewhat different procedure described elsewhere .
Rubisco activity assays
Rubisco carboxylase activity was routinely determined using a 14C-fixation method previously described . Kinetic constants were measured according to a reported procedure , and Ω was independently determined using the dual-labelling method .
Crystallization and structure determination
Crystals of C. reinhardtii Rubisco-mutant enzymes were grown by using hanging-drop vapour diffusion at 20 °C. The drops were prepared by mixing a solution containing 10 mg/ml Rubisco (in 50 mM Hepes, pH 7.5) and 1 mM 2-CABP (2-carboxyarabinitol 1,5-bisphosphate) with an equal volume of a well solution containing 50 mM Hepes (pH 7.5), 0.05–0.2 M NaCl, 7–12% (w/v) PEG [poly(ethylene glycol)] 4000, 10 mM NaHCO3 and 5 mM MgCl2. Crystals grew within 1 week. Prior to data collection, crystals were transferred to a solution containing the well solution with 30% (v/v) ethylene glycol as a cryoprotectant and frozen in liquid nitrogen. Data were collected from single crystals at 100 K on beam line ID14-3 at the ESRF (European Synchrotron Radiation Facility; Grenoble, France) (C172S), and on beam line I711 at Max-lab, Lund, Sweden (C192S). The data were processed using DENZO and SCALEPACK  to a resolution of 2.30 Å (1 Å=0.1 nm) for C172S and 2.65 Å for C192S (see Supplementary Table S1 at http://www.BiochemJ.org/bj/411/bj4110241add.htm).
The mutant crystal structures were solved by molecular replacement using the program AMoRe [31,32]. The search model consisted of one set of large and small subunits of wild-type C. reinhardtii Rubisco (PDB code 1GK8). In each case, eight solutions corresponding to eight different orientations of the search unit within one hexadecamer in the asymmetric unit were found.
Refinement was performed using REFMAC v. 5 . For cross-validation, 5% of the data were excluded from the refinement for Rfree calculations. Initial electron-density maps calculated after one round of rigid body refinement showed clear density for the substituted residues 172 and 192 respectively. After building the corresponding residues into density, the mutant enzyme structures were further refined using a maximum-likelihood target function and overall B-factor refinement. Tight NCS (non-crystallographic symmetry) restraints were imposed for the eight copies of large and small subunits in the asymmetric unit at the beginning of the refinement. Towards the end of refinement, NCS restraints were relaxed or removed for regions of the structure that clearly deviated from NCS. Solvent molecules were added using ARP/wARP . Throughout the refinement, 2mFo–DFc and mFo–DFc σA weighted maps  were inspected and the models manually adjusted using the program O . The structures of C172S and C192S were refined to final Rcryst/Rfree values of 0.171/0.204 and 0.205/0.234 respectively. The co-ordinates and structure factors have been deposited in the PDB with accession codes 2VDH (for C172S) and 2VDI (for C192S).
Isolation and growth of the C192S mutant
Chloroplast transformation of the C. reinhardtii 18-7G mutant lacking Rubisco  with the pLS-C192S plasmid, containing the rbcL gene with a C192S substitution, yielded transformant colonies that grew on minimal medium in the light. Amplification and sequencing of the rbcL gene confirmed that the transformants carried the single C192S mutation. This showed that the C192S mutant Rubisco was catalytically active. The C192S strain grew at a similar rate to the wild-type, both in liquid culture and on solid media, under photoautotrophic, photoheterotrophic or strictly heterotrophic conditions (results not shown). In general, the growth of the C192S strain presented no particular features when compared with the wild-type, except for a somewhat reduced survival of cells in old cultures.
Thermal stability of the mutant enzymes
The C172S Rubisco has already been shown to be thermosensitive . In order to compare the thermal stability of the C192S mutant protein with the C172S and the wild-type enzymes, purified Rubisco from all three strains was incubated at different temperatures (in the 48–62 °C range) for 30 min (Figure 1A) and at 55 °C for different times (Figure 1B), and the carboxylase activity was subsequently determined. The C172S enzyme showed a critical denaturation temperature (54 °C) approx. 2 °C lower than the wild-type (Figure 1A) and a distinctly higher denaturation rate at 55 °C (Figure 1B). In contrast, the denaturation temperature of the C192S Rubisco (near 55.5 °C, Figure 1A) was only slightly (approx. 0.5 °C) lower than the wild-type, and C192S denatured at a similar rate as the wild-type at 55 °C (Figure 1B). In a previous report , it was suggested that the lower stability of the C172S mutant protein could result from the inability of the mutant to form a Cys172 to Cys192 disulfide bond, which might be present in wild-type Rubisco. Because the C192S enzyme is relatively stable, it can now be concluded that the thermosensitivity of the C172S mutant Rubisco is not caused by the absence of a putative Cys172 to Cys192 disulfide bridge, but reveals instead a significant structural role played by the Cys172 residue, which cannot be fully complemented by serine at the same position. In contrast, the replacement of Cys192 by serine seems to have a lesser effect on structural stability.
Catalytic performance of the mutant enzymes
Table 1 displays the mean and S.D. values of the kinetic constants and Ω values determined for wild-type and mutant Rubisco enzymes. Vc was similar to that of the wild-type enzyme in the case of C172S, but somewhat lower (approx. 30%) for C192S Rubisco. In contrast, Kc and Ko were similar for C192S and wild-type Rubisco, but distinctly higher (approximately double) for the C172S mutant enzyme. This indicates that the lower carboxylation-specific activity reported for C172S Rubisco previously  was likely to be due to the increased Kc value (Table 1).
Whereas the Ω value of C192S Rubisco was identical with that of wild-type Rubisco (Ω=61), the C172S mutant enzyme showed a significantly (P>0.95) higher value of 68 (Table 1). With respect to the C192S enzyme, a lower Vc was compensated for by a lower Kc, resulting in an unchanged Ω value. However, in the case of C172S Rubisco, an even greater increase in Ko relative to an increased Kc is apparently responsible for an 11% increase in Ω (Table 1).
X-ray crystal structures of the mutant enzymes
To determine the molecular basis for their altered properties, the purified mutant enzymes were crystallized and their structures were determined. The crystallographic data collection and refinement statistics for the mutant-enzyme crystal structures are summarized in Supplementary Table S1. The final electron-density maps showed density for all 140 residues of the small subunit. The N-terminus of the large subunit displayed disorder that varied in extent from subunit to subunit. Density was observed for residues 11–475 (from a total of 475 residues) of the large subunit. In several of the subunits, an additional 2–4 N-terminal residues could be built into density. Clear density was observed for the substituted residues in each case. However, because the substituted residue is isosteric, it cannot be verified by inspection of difference Fourier maps in this case.
In general, no dramatic differences in large-subunit Cα backbone atoms were observed among the wild-type, C172S and C192S structures. A superposition of the wild-type large subunit with the large subunits of C172S and C192S by using the algorithms encoded in the program O  gave rmsd (root mean square deviation) values of 0.198 and 0.200 Å for all Cα atoms respectively. This is close to the co-ordinate errors of the respective structures. Inspection of the region around residue 172 in C172S Rubisco reveals that the main-chain atoms of residues 170–174 are displaced significantly (∼0.4 Å) compared with wild-type Rubisco (Figures 2A and 2B), resulting in a shift of β-strand 1 of the α/β-barrel. This is a larger displacement than the average for this β-strand and also for the average of all strands of the α/β-barrel (mean displacement 0.2 Å). Cys192 is located in α-helix 6 of the α/β-barrel facing Cys172. Replacement of Cys192 with serine did not cause a similar displacement of the backbone atoms of α-helix 6 or β-strand 6. In the wild-type enzyme, the distance between the sulfur atoms of residues 172 and 192 is 3.9 Å, which is too long for establishing a disulfide bond. In the C172S and C192S mutant enzymes, the oxygen-to-sulfur distances are even somewhat longer, 4.1 Å. In contrast, the distance between the sulfur atoms of Cys449 and Cys459 (2.0 Å) and their conformations in the wild-type enzyme are ideal for disulfide formation . In C172S, the sulfur-to-sulfur distance between Cys449 and Cys459 was similar to that of wild-type Rubisco (2.1 Å). In C192S this distance was lengthened to 2.9 Å (Figure 3), resulting in a shift of the α-helix harbouring Cys449 (residues 438–452), breaking the disulfide bond.
A superposition of the wild-type small subunit with the small subunits of C172S and C192S gave rmsd values of 0.235 and 0.224 Å respectively for Cα atoms 1–126. This is close to the co-ordinate errors of the respective structures, indicating that the C172S and C192S substitutions in the large subunit had no noticeable effect on the small-subunit structure.
Alterations in the active site
Residue 172 is located in β-strand 1 of the α/β barrel and is close to the active site. The most immediate effect of the C172S substitution is to shift the main-chain atoms of β-strand 1 by 0.4 Å relative to wild-type. This shifted residue 172 by 0.4 Å and resulted in an increase (by 0.2 Å) in the distance between the side chains of residues 172 and 192. It also resulted in a significant increase in the distance (by more than 0.3 Å) between Thr173 and CABP relative to wild-type Rubisco (Figure 2). All other shifts of active-site residues were close to the co-ordinate error.
Cys172 and Cys192 of the large subunit are highly conserved among both eukaryotic and cyanobacterial Rubiscos. Their structural vicinity, potentially allowing for a disulfide bond to form between them, could be a possible source of redox regulatory effects on the enzyme. Analysis of the high-resolution (1.4 Å) structure of the wild-type C. reinhardtii enzyme  clearly shows the presence of free Cys172 and Cys192 residues in the electron-density maps. The distance and geometries of the two residues in activated Rubisco are also not compatible with the geometry required for a disulfide bond. This implies that, in order for a disulfide to form under oxidative conditions, the main-chain backbone around the residues would have to move considerably. Nevertheless, disulfide bonding between Cys172 and Cys192 appears to be possible if Rubisco becomes decarbamoylated, as observed in the unactivated tobacco enzyme . On the other hand, conformational changes are known to be induced in Rubisco by oxidative modification of cysteine residues . Therefore Cys172 and Cys192 offer, in principle, a potential for redox modulation of Rubisco through structural changes induced by oxidation.
We have addressed the structural and functional role of the conserved cysteine residues by replacement with serine via site-directed mutagenesis and chloroplast transformation of C. reinhardtii. As a result of this single-atom substitution (of oxygen for sulfur), the redox-sensitivity of the residue is abolished. We have shown that the replacement of Cys172 with serine increases Rubisco thermosensitivity (Figure 1), perturbs the structure near the substitution site (Figure 2) and alters the kinetic constants of the enzyme (Table 1). Because the symmetric replacement of Cys192 does not induce the same changes, it appears that the effects described are not directly related to the presence/absence of a disulfide bond between the proximal cysteine residues. Thus it seems that Cys172 plays by itself a remarkable role in the structure and catalytic properties of Rubisco.
The replacement of Cys172 by serine increases the Km for both gaseous substrates (Table 1). Because CO2 and O2 do not bind to the enzyme before reacting with the enediolate form of RuBP, the increased Km values might result from a structural change that causes a restricted accessibility of the bound enediolate RuBP to CO2 and O2. However, because the Ω value of the C172S enzyme is also altered (Table 1), it is more likely that conformational changes occur that affect the stabilities of the carboxylation and oxygenation transition states differently . Cys172 is situated in β-strand 1 of the α/β-barrel, which forms the scaffold for the active site . Thus replacement of this residue in the barrel may affect active-site residues. Such perturbations, however subtle, may have dramatic consequences for the kinetics of the reaction. Indeed, replacement of either of the flanking residues, Gly171 or Thr173, in C. reinhardtii Rubisco eliminates carboxylase activity . Our X-ray diffraction analysis of the C172S mutant Rubisco indicates that the replacement with serine at position 172 causes a significant shift of the main-chain backbone atoms of β-strand 1 of the α/β-barrel. This shift could disturb the geometry of the active site as well as the position of individual residues in the active site. In particular, the interactions of the proximal Thr173, hydrogen-bonding to the C2-hydroxyl of the CABP transition-state analogue , and perhaps Lys175, which is thought to participate as a secondary proton acceptor in generating the reactive enediolate , are likely to be strained by the displacement of the β-strand. In contrast, the C192S mutation, which produces relatively minor alterations of enzymatic properties (Table 1), does not involve structural changes around the catalytic site.
Because the effect of C172S on enzyme kinetics appears to be structural and indirect rather than resulting from the elimination of the redox-active thiol group, the particular impairment imposed on the catalytic site by the substitution might be expected to depend on the volume and nature of the substituting residue. Accordingly, the replacement of Cys172 by alanine in a prokaryotic Rubisco produced only a limited increase in Kc but a significant reduction of Vc . In this regard, the modification of Cys172 with the bulky N-bromoacetylethanolamine phosphate affinity label is likely to cause a stronger structural distortion, resulting in activity loss .
Another pair of proximal cysteine residues, Cys449 and Cys459, are located close to the surface and approx. 20 Å away from the 172/192 pair. A survey of all available C. reinhardtii Rubisco structures to date [23,39–41] revealed that whereas the sulfur-to-sulfur distance of the 172/192 pair remains fairly constant, the distance of the 449/459 pair varies considerably. This variation is also illustrated by the present structures. In C172S, the sulfur-to-sulfur distance of the 449/459 pair was similar to that of wild-type, but in C192S this distance was considerably lengthened and there was no sign of a covalent bond in the electron-density maps. This variation at a location close to the molecular surface may suggest a role for the 449/459 pair as a redox sensor. It is not clear at this stage if a long-distance structural connection exists between both redox centres or if some of the redox effects associated with the C172S substitution  might in fact result from the perturbation of the redox properties of Cys449 and Cys459. However, the in vitro and in vivo redox properties of mutant enzymes in which Cys449 and Cys459 have been replaced by serine are very different from those of C172S . Therefore a physiologically relevant effect resulting from the co-operation between the two cysteine pairs seems unlikely.
It is noteworthy that the C172S substitution produces a modest but significant increase in Ω (Table 1). Changes in Ω have been proposed to result from differences in the differential stabilization of the carboxylation and oxygenation transition states [3,43]. It has been recently hypothesized by Tcherkez et al.  that an increase in Ω should result in a decreased Vc and that a negative correlation between Ω and Kc may also exist . These predictions are, however, not fulfilled in the case of the C172S mutant enzyme, which displays an increased Ω value without alteration of Vc but with a significant increase in Kc (Table 1).
Considering that Cys172 is a critical residue that renders Rubisco inactive when oxidatively modified by a number of agents [16–18], one may wonder why it is so highly conserved and why it has not been replaced by serine during evolution for the gain of substrate specificity and oxidative resistance. Moreover, the only apparent disadvantages of the C172S substitution would be a slight structural destabilization that does not seem to be relevant at physiological temperatures (Figure 1) and increases in Kc and Ko (Table 1). Despite the increased Ω value, the increase in Kc would probably cause a decrease in net CO2 fixation, which is determined by the difference in the velocities of carboxylation and oxygenation at physiological concentrations of CO2 and O2 . Nevertheless, the C. reinhardtii C172S mutant has been shown to grow at the same rate as the wild-type under photoautotrophic conditions . There may be, however, additional reasons for the conservation of Cys172. Perhaps it is precisely its ability to act as an oxidation-activated switch, as a part of a wider design of redox control of CO2 fixation and Rubisco turnover, that has prevented the replacement of this residue during evolution.
This work was supported by grants from the Spanish MCyT (Ministerio de Ciencia y Tecnología) (BMC2003-03209 and BFU2006-07783), the Swedish Research Council for Environment, Agricultural Sciences, and Spatial Planning (FORMAS), the European Union (no. QLK3-CT-2002-01945) and the U.S. Department of Agriculture National Research Initiative (no. 2006-35318-17376).
The structural co-ordinates reported for the mutant Rubiscos of Chlamydomonas reinhardtii will appear in the PDB under accession codes 2VDH (for C172S) and 2VDI (for C192S).
Abbreviations: 2-CABP, 2-carboxyarabinitol 1,5-bisphosphate; Kc, Michaelis constant for CO2; Ko, Michaelis constant for O2; NCS, non-crystallographic symmetry; rmsd, root mean square deviation; Rubisco, ribulose-1,5-bisphosphate carboxylase/oxygenase; RuBP, ribulose 1,5-bisphosphate; Vc, Vmax for carboxylation; Vo, Vmax for oxygenation; Ω, CO2/O2 specificity factor
- © The Authors Journal compilation © 2008 Biochemical Society