Glycine 176 Affects Catalytic Properties and Stability of the Synechococcus sp. Strain PCC6301 Ribulose-1,5-bisphosphate Carboxylase/Oxygenase* for

A previously described system biological selection of randomly mutagenized ribulose-1,5-bisphosphate carboxyl-ase/oxygenase (Rubisco) employing the phototrophic bacterium Rhodobacter capsulatus (Smith, S. A., and Tabita, F. R. (2003) J. Mol. Biol. 331, 557–569) was used to select a catalytically altered form of a cyanobacterial ( Synechococcus sp. strain PCC6301) enzyme. This mutant Rubisco, in which conserved glycine 176 was replaced with an aspartate residue, was not able to support CO 2 -dependent growth of the host strain. Site-directed mutant proteins were also constructed, e.g. asparagine and alanine residues replaced the native glycine with the result that these mutant proteins either greatly reduced the ability of R. capsulatus to support growth or had little effect, respectively. Growth phenotypes were consistent with the Rubisco activity levels associated with these proteins, and this was also borne out with purified recombinant proteins. Despite being catalytically challenged, the G176D and G176N mutant proteins were found to exhibit a more favorable inter-action with CO 2 than the wild type protein but exhibited a reduced affinity for the substrate ribulose 1,5-bisphos-phate. The G176A enzyme differed little from the wild type protein in these properties. None of the mutants had CO 2 /O 2 specificities that differed markedly from

Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco), 1 the key enzyme of the Calvin-Benson-Bassham reductive pentose phosphate pathway, catalyzes the fixation of CO 2 onto the fivecarbon substrate ribulose 1,5-bisphosphate (RuBP) yielding two molecules of 3-phosphoglycerate. Rubisco also catalyzes a competing monooxygenase reaction, whereby RuBP reacts with molecular O 2 yielding one molecule of 3-phosphoglycerate and one molecule of 2-phosphoglycolate. The rates at which Rubisco catalyzes the competing carboxylase and oxygenase reactions are determined by the inherent substrate specificity factor of the protein, ⍀, and may be expressed by: where V is V max and K represents the Michaelis constants for carboxylation and oxygenation (c and o, respectively). The production of 2-phosphoglycolate is the first reaction of a respiratory pathway, which ultimately leads to the loss of carbon from cells. It is not understood why different yet structurally similar Rubisco proteins possess inherent and variable catalytic efficiencies and substrate selectivities (1,2). The inherent kinetic properties of this bifunctional enzyme, particularly ⍀, and the relative affinities for its gaseous substrates constitute a major limiting factor governing biomass yields of many industrially and ecologically significant microorganisms that fix CO 2 as well as several economically important plants. For these reasons, there is considerable interest in elucidating the molecular factors that influence the variability in kinetic properties exhibited by structurally similar Rubisco enzymes. There have been multiple attempts to change or engineer Rubisco; however, no alterations have significantly influenced the physiology of the host organism so that some distinct benefit might be achieved.
Recently a system was described for directed evolution of prokaryotic Rubisco genes using the purple, nonsulfur, phototrophic bacterium Rhodobacter capsulatus as a selective host (3). R. capsulatus is capable of aerobic growth using organic carbon (chemoheterotrophy) and can synthesize its own organic carbon through CO 2 fixation via the Calvin-Benson-Bassham cycle either photosynthetically (photoautotrophy) or during dark aerobic growth (chemoautotrophy). The strain of R. capsulatus used in this system, strain SBI-II Ϫ , is a Rubisco deletion strain incapable of either photoautotrophic/chemoautotrophic or photoheterotrophic growth (4) unless complemented in trans by exogenous Rubisco genes. Indeed, this system has the inherent advantage that growth of strain SBI-II Ϫ under these conditions is directly dependent on the properties of the Rubisco gene(s) employed for complementation. Thus, by expressing heterologous Rubisco genes in strain SBI-II Ϫ followed by complementation to CO 2 -dependent growth, the enzyme from the cyanobacterium Synechococcus PCC6301 was chosen as a useful target to initiate bioselection studies. A positive selection condition was identified in which the wild type Synechococcus Rubisco genes were unable to support growth at levels of atmospheric CO 2 lower than 1.5% but could support growth at concentrations of CO 2 greater that 5%. Using this selection system a variety of mutant Rubisco proteins were isolated (some better able to support growth at low CO 2 levels), and enzymes with inherent difficulties to support growth at higher levels of CO 2 were also obtained (3). In addition to the phenotype that Synechococcus Rubisco genes confer to R. capsulatus SBI-II Ϫ , these genes may also be conveniently expressed to produce recombinant enzyme in Escherichia coli. The previously solved x-ray structure (5) then allows one to better understand the positioning of mutated residues in the cyanobacterial enzyme. Furthermore, the cyanobacterial (Synechococcus) enzyme is a "green-like" Rubisco with high sequence and structural identity to plant enzymes, making it a suitable model to understand plant and eukaryotic Rubisco (1).
Using the bacterial selection system, potentially interesting but defective enzymes were identified by their inability to complement R. capsulatus SBI-II Ϫ under conditions where the wild type enzyme was fully competent (3). One particular residue, glycine 176 (Gly-176), was identified through this selection system. Specifically, a mutation that involved a change of Gly-176 to aspartate (G176D) resulted in an enzyme that was unable to complement strain SBI-II Ϫ to photoautotrophic growth in the presence of high levels of CO 2 . However, in a previous study G176 had not been shown to influence catalysis to any great extent (6). The G176D enzyme and the newly constructed G176N and G176A enzymes were used to study the influence of residue Gly-176 in greater depth. In this report we show that the kinetic properties (K c and K RuBP ) of the G176D and G176N mutant proteins were greatly altered. In addition, the altered enzymes exhibited various degrees of reduced thermal stability and to differing extents showed an unusual protein concentration-dependent alteration in specific activity. These latter properties along with the potential for this residue to influence important associations of catalytic subunits suggest that the Gly-176 mutant proteins might serve as starting points for further mutagenesis and selection.

EXPERIMENTAL PROCEDURES
Strains, Plasmids, and Growth-Construction of the Rubisco deletion strain of R. capsulatus strain SBI-II Ϫ was described previously (4). To maintain cultures, R. capsulatus was grown aerobically on peptone yeast extract plates (7) or in a broth containing Ormerod's basal salts (8), 6 g/liter peptone, 5 g/liter yeast extract, 10 mM NaCl, 3 mM KCl, and 0.1 g/ml biotin. Both media were supplemented with 1 g/ml nicotinic acid and 1 g/ml thiamine hydrochloride. Antibiotics in peptone yeast extract plates were used at the following concentrations: 100 g/ml rifampicin, 2 g/ml tetracycline, 10 g/ml spectinomycin, 5 g/ml kanamycin.
Growth of R. capsulatus under selective conditions lacking organic carbon (photoautotrophic growth) was performed using Ormerod's minimal medium (8) supplemented with 1 g/ml nicotinic acid and 1 g/ml thiamine hydrochloride. Plates for photoautotrophic growth were incubated in jars containing a CO 2 /H 2 -generating system (5-6% CO 2 , BBL GasPak system, BD Biosciences), and liquid cultures were bubbled with 1.5% CO 2 /98.5% H 2 as described previously (3). Growth rates were determined by removing 1-ml samples at 6 -12-h intervals for absorbance readings at 660 nm. When cultures reached an OD 660 between 1.2 and 1.5, where maximum levels of Rubisco activity are reached (data not shown), 40 ml were removed, the cells were washed with TE buffer (100 mM Tris-HCl, 1 mM EDTA, pH 8.0), and the cell pellets were stored at Ϫ80°C until needed for assays.
Es. coli JM109 and pUC19 were used for routine cloning, and JM109 was also the donor strain in conjugations (9,10). Es. coli HB101, which harbors the transfer genes required for conjugation on pRK2013, was used as the "helper" strain in triparental matings with Es. coli JM109 and R. capsulatus SBI-II Ϫ (11,12). Epicurian coli® XL1-Red (Stratagene, La Jolla, CA) was used for random mutagenesis of genes encod-ing Rubisco as described previously (3). All Ep. coli strains were grown in Luria-Bertani medium at 37°C using 100 g/ml ampicillin for cultures containing pUC19 or 12.5 g/ml tetracycline for cultures containing pRPS-MCS3.
The genes encoding Rubisco of Synechococcus PCC6301, rbcLS, were cloned into pUC19 to generate pUC6301, the plasmid used for mutagenesis in Ep. coli XL1-Red. Plasmid pRPS-MCS3 was constructed specifically for this selection system. It was derived from the broad host range plasmid pBBR1-MCS3 (13) and contained a Rubisco promoter from Rhodospirillum rubrum to express wild type and mutated rbcLS genes in R. capsulatus SBI-II Ϫ . The construction and properties of this plasmid were described recently (3).
Identification of XL1-Red Negative Mutant G176D-Plasmid pUC6301 was mutated by propagation in Ep. coli XL1-Red, reisolated, and digested with PstI and XbaI (MBI Fermentas) for subcloning of rbcLS into pRPS-MCS3 (3). Ligations were used to transform Ep. coli JM109, and the pRPS-MCS3 subclones were introduced into R. capsulatus SBI-II Ϫ by triparental conjugation (12). Mated cells were plated onto peptone yeast extract with rifampicin to select against Ep. coli and tetracycline to select against R. capsulatus that had not received the pRPS-MCS3 plasmid. These plates were incubated for 3 days at 30°C.
Tetracycline-resistant colonies of R. capsulatus SBI-II Ϫ were tested for the ability to grow photoautotrophically. Isolates incapable of photoautotrophic growth were called negative mutants. It was verified by colony hybridization that the negative mutants contained rbcLS on pRPS-MCS3, and Western immunoblots revealed which mutants were capable of synthesizing a large subunit polypeptide. The negative mutant genes were sequenced as described previously (3).
Preparation of Cell Extracts and Assays-Prior to assays, frozen cells were thawed on ice and resuspended in TE containing 5 mM ␤-mercaptoethanol. Cells were disrupted by sonication, and low speed supernatants were obtained by microcentrifugation for 10 min at 4°C. High speed supernatants from Es. coli extracts were obtained by ultracentrifugation at 45,000 rpm for 1.5 h at 4°C. Rubisco activity was measured by incorporation of 14 CO 2 into acid stable products (14). Total protein concentrations in samples were measured by a modification of the Lowry assay (15). The extracts were also used to verify Rubisco polypeptide synthesis by SDS-PAGE and Western blots (16). Western blots were developed and visualized by chemifluorescence as described previously (4). The ability of mutant enzymes to assemble into a holoenzyme was screened by Western blotting of nondenaturing gels containing extracts from some of the cultures and purified enzymes.
Kinetic Measurements-Purified enzymes were required for measurements of k cat , K c , and ⍀ as described previously (3). Briefly, 1-liter cultures of Es. coli JM109 containing pUC19 subclones of the wild type and mutant enzymes were induced with 0.5 mM isopropyl-1-thio-␤-Dgalactopyranoside overnight, and harvested cells were disrupted by sonication. Rubisco was isolated from lysates by anion exchange chromatography using a Green-A dye affinity column (Amicon, Beverly, MA) and eluting with a 0 -1 M NaCl gradient. This was followed by 70% ammonium sulfate precipitation and ultracentrifugation of the products through a 0.2-0.8-M sucrose density gradient. The final sample was dialyzed against TEM buffer (25 mM Tris-HCl, 1 mM EDTA, 10 mM MgCl 2 , pH 8.0), and 1-ml aliquots were stored in TEM with 20% glycerol at Ϫ80°C. The purified enzymes were used to measure k cat , K c , K RuBP , and specificity as described previously (3).
Protein Concentration Dependence and Activity-All enzyme preparations were diluted to 0.5 g/l in TEM with 20% glycerol so that the same volume of enzyme were added to each assay. Thus 10 l of enzyme was added for 5 g of Rubisco per assay, 20 l for 10 g, and so on. Samples were assayed by the standard protocol with 1-min assays for the wild type, K11R, and G176A enzymes and 5-min assays for the K11R/G176D, G176D, and G176N mutants. Relative activity was calculated as the percentage of activity in any given assay relative to the assay that contained 5 g of the same Rubisco.
Thermal Stability Assays-All enzymes were diluted to 0.5 g/l in TEM with 20% glycerol, which was the same buffer the enzymes had been stored in at Ϫ80°C. 20 l was used for a standard Rubisco assay, which served as the reference assay for calculation of relative activity. 200-l samples were placed in a thermocycler (PerkinElmer Life Sciences) heated to 55°C. After 10, 20, 30, 40, 60, 80, 100, and 120 min of incubation, 20-l aliquots were removed and immediately assayed by the standard protocol. The wild type, K11R, and G176A enzymes were assayed for 1 min; the other mutants were assayed for 5 min. Different assay times were necessary to stay within the linear range of enzyme activity with respect to time. Relative activity was calculated as the percent of specific activity in the incubated sample relative to the specific activity in the reference assay prior to incubation.

RESULTS
Residue Gly-176 and Growth Phenotypes-In a previous report (3), a system was described whereby potentially interesting mutant forms of cyanobacterial Rubisco could be selected after random mutagenesis and complementation of a Rubiscodepleted strain of R. capsulatus. This strain, SBI-II Ϫ , was unable to grow photoautotrophically, photoheterotrophically, or chemoautotrophically unless genes encoding a functional Rubisco, such as rbcLS from Synechococcus sp. strain PCC6301, were provided in trans. Furthermore, when these genes were randomly mutagenized prior to introduction into R. capsulatus SBI-II Ϫ , altered forms of Rubisco could be identified based on the phenotype of SBI-II Ϫ complemented to photoautotrophic growth on minimal medium plates. Whereas the wild type genes enabled growth of the strain, genes carrying "negative" mutations either would not complement or would poorly complement the strain to photoautotrophic growth. In this manner we isolated a mutant enzyme that had an aspartate residue substituted for a highly conserved glycine residue at position 176 in the cyanobacterial enzyme.
In a previous study (6), all of the glycine residues of the Synechococcus 6301 enzyme were changed, and residue Gly-176 (residue Gly-179 of spinach Rubisco) was found to influence activity; however, no further indications of its potential role in catalysis was reported. In light of our discovery that the G176D mutation grossly affected the photoautotrophic growth phenotype of R. capsulatus SBI-II Ϫ , we decided to focus on this residue further; additional site-directed mutants, G176A and G176N, were prepared to assist in these studies. We also reconstructed the G176D mutation to ensure that none of the effects previously observed were caused by changes in the strain background of SBI-II Ϫ or to mutations elsewhere on the pRPS-MCS3 plasmid.
The growth phenotype of R. capsulatus strain SBI-II Ϫ was affected by all of the mutant enzymes studied, even the conservative G176A mutation. On photoautotrophic plates, none of the mutant enzymes complemented the strain to photoautotrophic growth after a week of incubation. In liquid cultures that were continuously supplied with 1.5% CO 2 /98.5% H 2 , the G176D and G176N enzymes supported growth although with doubling times nearly 3-fold higher than the wild type enzyme. The G176A enzyme was much less affected in its ability to complement strain SBI-II Ϫ under these growth conditions (Table I). Furthermore, Rubisco specific activity levels in crude extracts of photoautotrophically grown cells seemed to reflect the growth rates with the G176D and G176N mutants having activities that were only a fraction of those measured for either the G176A mutant or the wild type enzymes (Table I). To ensure that these low activities were not simply the consequence of poor expression of the mutant genes in R. capsulatus, the steady state levels of large subunit polypeptides were qualitatively assayed on Western blots of these same extracts, and there appeared to be no significant differences (Fig. 1). Fig. 1 also included a K11R mutant that was discovered as part of a different study and found to have no effect on the properties of the enzyme or the phenotype of R. capsulatus SBI-II Ϫ relative to the wild type enzyme (data not shown).
Catalytic Properties of Gly-176 Mutant Enzymes-The kinetic properties of the purified enzymes were studied in vitro. Surprisingly, the G176D and G176N mutant enzymes had a higher affinity for CO 2 than either the wild type or G176A mutant, and they had a lower affinity for the substrate RuBP (Table II). Note that despite the roughly 2-fold difference in the K c for these mutants relative to the wild type there was no significant difference in specificity, at least with the precision of the employed specificity assay, indicating that parameters that were not measured, such as the K o , were also probably affected.
When determining the k cat of the G176D enzyme, it was noted that the k cat was difficult to validate at different concentrations of enzyme. Thus, an experiment was undertaken to examine this property for all of the mutant enzymes. The G176D enzyme consistently showed a pattern of increasing specific activity with respect to protein concentration up to a concentration of 291 nM (40 g/assay). This property was not displayed by the wild type enzyme and perhaps only mildly exhibited by the G176N and G176A enzymes (Fig. 2). When this experiment was repeated with a constant concentration of Rubisco and increasing concentrations of bovine serum albumin, there was no effect (data not shown). Thus, we conclude that the protein concentration dependence effect on enzyme activity exhibited by the G176D enzyme was a property specific to this Rubisco and not solely the consequence of increasing concentrations of just any protein.
Protein Conformation-It is apparent that the ability of the holoenzyme to maintain its active conformation was compromised in the G176D mutant enzyme. The protein concentration effect described above, together with analysis of the known structure (5,17), suggested the potential for glycine 176 to influence large subunit associations into dimers and eventual formation to the catalytically important octameric core; alternatively, Gly-176 potentially influenced the ability of the octameric core to properly associate with small subunits to form  Gly-176 of Cyanobacterial Rubisco the fully assembled and maximally active holoenzyme. As a potential indicator of holoenzyme stability, it was shown that the wild type enzyme retained substantial activity after 2 h of incubation at 55°C. The mutant proteins, by contrast, had all lost more than 50% of their relative activities within the first 20 min of incubation (Fig. 3). Even the G176A mutant, which showed only minor differences relative to the wild type enzyme up to this time, lost significant activity after only 10 min of incubation at 55°C (Fig. 3). Consistent with these results, the mutant enzymes, including the G176A enzyme, displayed aberrant migration on nondenaturing gels relative to the wild type enzyme, whereas qualitatively the stoichiometry of large and small subunits did not appear to differ between the wild type and the mutants under denaturing conditions (Fig. 4). Fig.  4 also contains two mutants in lanes 2 and 3 (K11R and K11R/G176D). As described above, the K11R mutation was found to have no effect on the properties of the enzyme or the phenotype of R. capsulatus SBI-II Ϫ relative to the wild type enzyme. Likewise, in the context of G176D, this mutation had no effect (data not shown), and it was therefore not germane to the experiments discussed here. DISCUSSION The development of a prokaryotic system to enable the biological selection of potentially interesting mutant forms of Rubisco was described previously (3). The inability of the Synechococcus PCC6301 Rubisco to complement R. capsulatus SBI-II Ϫ to photoautotrophic growth, predicted to be a consequence of the extremely high K c (ϳ173 M) of this enzyme (18), was the major rationale for using this enzyme in these initial studies. Although our original hypothesis was that a system of positive selection in which mutant enzymes capable of complementing SBI-II Ϫ to photoautotrophic growth under a low CO 2 atmo-sphere might yield enzymes with improved affinity for CO 2 , the results given in this investigation and the previous study (3) so far indicate that negative selection produces the most useful and interesting mutant enzymes. Thus, in both studies mutant enzymes were selected that were unable to complement strain SBI-II Ϫ to photoautotrophic growth under an atmosphere of 5% CO 2 /95% H 2 ; these are conditions where the wild type enzyme was fully competent. In addition, it was important to verify and choose mutant proteins that possessed no obvious gross folding defects as evidenced by the ability of the proteins to retain solubility and maintain some degree of catalytic competency (3). In this initial selection, the G176D enzyme represents one of the few mutants that does not have a gross folding problem and appears to be affected by some aspect of catalysis.
Gly-176 (residue 179 in spinach Rubisco) is one of 22 glycine residues that is completely conserved in Rubisco. Note that another glycine, Gly-401 (residue 404 in the spinach enzyme), was identified through this selection process; however, the  2. Protein concentration dependence of Rubisco specific activity for the G176 mutant enzymes. The Rubisco specific activity was measured with concentrations of purified enzyme ranging from 5 to 40 g in each assay, expressed as 36 to 291 nM here. Beyond these concentrations there was a plateau of activity. The specific activity obtained at 36 nM for each enzyme was: wild type (wt), 3.8 mol of CO 2 per fixed/min/mg protein; G176D, 0.07 mol of CO 2 per fixed/min/mg protein; G176N, 0.55 mol/min/mg of protein; and G176A, 1.0 mol/ min/mg of protein. Because of the range of activities, the wild type and G176A enzymes were assayed for 1 min, and the G176D and G176N mutants were assayed for 5 min. The logarithm of the specific activity at each concentration relative to the activity in the 36 nM assay is represented by the y axis.
FIG. 3. Thermal stability of the wild type and G176 mutant enzymes. All samples were diluted to a concentration of 0.5 g/l in TEM ϩ 20% glycerol (the same buffer in which the enzymes were stored). Samples were incubated at 55°C. At the indicated times, samples were withdrawn from the incubations and assayed for Rubisco activity. One assay was performed prior to incubation of the samples (t ϭ 0), and this was the reference assay for calculating relative activity after incubation, represented as a logarithm on the y axis. Lanes 2 and 3 contain K11R and K11R/G176D mutants that were not part of the current study (under "Results"). The gel with SDS contained 15% acrylamide, and the gel without SDS contained 7% acrylamide. Approximately 2 g of each purified sample were loaded onto the SDS gels, and 3 g were loaded onto the nondenaturing gels. Both were stained with Coomassie Blue. LSU, large subunit; SSU, small subunit. G401S enzyme obtained was shown to be folding/assemblyincompetent (3). Gly-176 lies within a stretch of 7 amino acids that are completely conserved; these residues are localized in a loop preceding the first ␣-helix of the C-terminal domain of the enzyme near an interface between large subunits that participate in dimerization (5,17). A broad survey of the effects of substitutions at conserved glycyl residues had previously identified Gly-176 as a potentially interesting residue (6), but the properties of the G176A enzyme prepared in this earlier work differ somewhat from the results obtained for the G176A mutant prepared in the current study. In particular, Cheng and McFadden (6) described their mutant as having ϳ18% wild type activity, although results described here placed it closer to 40 -50% wild type activity. A possible explanation for this discrepancy is the slight concentration-dependent effect on activity shown by this protein (Fig. 2). Although the amount of enzyme used in the assays performed by Cheng and McFadden (6) was not reported, perhaps if higher concentrations had been used the reported activity would more closely resemble that which is reported here. In addition, the G176A enzyme was initially described as mildly defective in its ability to fold properly, but no data were shown. It is conceivable that the slightly altered migration on nondenaturing gels observed here might have been observed previously and leads to classification of the G176A enzyme (G179A in the Cheng and McFadden paper, Ref. 6) as defective for folding/assembly.
The protein concentration-dependent effect on maximal activity, particularly with the G176D enzyme, could be a consequence of forcing an equilibrium toward the association of large subunits into functional dimers, the fundamental units of Rubisco assembly. Interestingly, the nondenaturing gel shown in Fig. 4 may also support this notion. There appears to be significantly less protein in the mutant lanes on this gel (lanes 3-6). A possible explanation for this is that, if the association of large subunits has indeed been affected by substitutions at Gly-176, the decreased stability would result in a lower proportion of holoenzyme relative to total protein in the preparation. The dissociated subunits that may be present would migrate off of the gel, and there would appear to be a lower concentration of enzyme.
At lower protein concentrations, a tendency for weak large subunit associations could also account for the observed thermal instability of enzymes with substitutions at residue 176. If this is true, then the results of the kinetic measurements for the Gly-176 mutant proteins support the hypothesis that the K c and K RuBP are determined, at least in part, by conformational properties at a tertiary and possibly quaternary level. Our previous finding that a D103V mutation located in a region near the interface between large subunit dimers had the opposite effect on the K c further supports this prediction (3). Our attempts to model the G176D mutation in silico (using the Swiss Protein Data Base viewer, version 3.7) indicate that any of the possible conformations assumed by an aspartate at this position would not be accommodated because the aspartate would contact neighboring residues. Disregarding this warning from the software because the G176D enzyme obviously reaches a conformation that has some catalytic competency and using the lowest energy conformation proposed by the software, the model produced (Fig. 5) indicated minor repositioning of some of the local residues with the most apparent change observed in threonine 72 of the opposing large subunit. This model as well as the in vitro data suggest that changing residue 176 to aspartate caused major conformational shifts in a localized region at the interface between large subunits in the dimers. It is conceivable that these changes in monomer association may subsequently be translated to larger conforma-tional changes in the holoenzyme. The properties of decreased thermal stability and altered migration on nondenaturing gels seem to support this conclusion.
An important feature of Rubisco enzymology is the activation of large subunits by a nonsubstrate molecule of CO 2 resulting in the formation of a lysyl carbamate at a completely conserved lysine residue within the active site (Lys-201) (19 -21). In the absence of activation by CO 2 , RuBP and other sugar phosphates will inhibit Rubisco. Activated Rubisco is stabilized by a Mg 2ϩ cofactor, which is essential for the carbamate to act as a general base, abstracting a proton from the substrate RuBP and forming an unstable enediol intermediate (22). Both carboxylation and oxygenation proceed through this intermediate, and it is beyond that point that the chemistry diverges. The same active site residues, therefore, are required for either carboxylation or oxygenation. Rubisco activation was not the focus of this study, but it is worth noting that it was briefly investigated by comparison of the wild type and the original G176D mutant. No difference was noticed in this single activation assay (data not shown), but this property may be worth further investigation in future studies of this enzyme.
It is also interesting to note that the mutations we have found that affected K c (at residues Asp-103 and Gly-176) had no effect on specificity. At the outset of these studies, the K c was the presumed "target" of positive selection, because we predicted the enzyme would require a lower K c to enable R. capsulatus SBI-II Ϫ to grow photoautotrophically with lowered CO 2 in the atmosphere. Furthermore, we hoped that such a mutant would have an increased specificity as a consequence of its decreased K c (see the equation in the Introduction). These results and the results of several mutant studies that precede this reflect that the specificity parameter is recalcitrant to increases even when mutations result in a lower K c as with the G176D and G176N mutants. Perhaps some of the mutant proteins described here will serve as a better starting material for further mutagenesis and selection, allowing us to obtain the long sought highly active improved specificity mutant proteins. For example, the G176D mutant has a better affinity for CO 2 than the wild type enzyme but less than 10% of wild type carboxylase activity. Because this enzyme already possesses one of the desirable characteristics, i.e. low K c , random mutagenesis might lead to a protein with either an enhanced k cat and/or normal K RuBP , which could be identified by the ability of such proteins derived from G176D to support photoautotrophic growth. It seems possible that such a mutant might retain the gains made in K c . It could then be used in further rounds of selection under aerobic chemoautotrophic growth conditions FIG. 5. In silico modeling of the interface between two large subunits in the region of glycine 176. The wild type is pink, and the mutant with the aspartate substitution is blue. Note that residues 175-177 are on one large subunit monomer, whereas residues 71-74 are on the partner large subunit in the active dimer. In the immediate vicinity of this substitution, threonine 72 (Thr72) shows a major repositioning of its side chain. This image was created using DeepView in the Swiss Protein Data Base Viewer, version 3.7. where the levels of CO 2 and O 2 may be changed. Altering the relative concentrations of these gaseous reactants may target specificity mutations more directly. The complex nature of the specificity issue certainly suggests that multiple rounds of mutagenesis and selection will be necessary; the scenario outlined here might allow for such an approach where one aspect of catalysis is targeted (K c ) followed by another (k cat ) and/or K RuBP . This stepwise approach, rather than gross recombination and shuffling, will allow us to keep track of which changes and residues are important. At the very least, biological selection will allow us to learn more of the molecular determinants that influence the affinity and/or association of the enzyme with its substrates and gaseous reactants. Alternatively, isolating suppressor mutations of the G176D, G176N, and G176A enzymes directly by selecting for mutations in those genes that restore normal growth of strain SBI-II Ϫ under photoautotrophic conditions should also prove fruitful.