Selection of Cyanobacterial (Synechococcus sp. Strain PCC 6301) RubisCO Variants with Improved Functional Properties That Confer Enhanced CO2-Dependent Growth of Rhodobacter capsulatus, a Photosynthetic Bacterium

RubisCO catalysis has a significant impact on mitigating greenhouse gas accumulation and CO2 conversion to food, fuel, and other organic compounds required to sustain life. Because RubisCO-dependent CO2 fixation is severely compromised by oxygen inhibition and other physiological constraints, improving RubisCO’s kinetic properties to enhance growth in the presence of atmospheric O2 levels has been a longstanding goal. In this study, RubisCO variants with superior structure-functional properties were selected which resulted in enhanced growth of an autotrophic host organism (R. capsulatus), indicating that RubisCO function was indeed growth limiting. It is evident from these results that genetically engineered RubisCO with kinetically enhanced properties can positively impact growth rates in primary producers.

K c ϭ ϳ20 M and ⍀ ϭ 80) (23,37). The sequestration of Synechococcus RubisCO into CO 2 -concentrating carboxysomes in vivo presumably explains the lack of selective pressure to naturally evolve a "better" kinetic variant (12,15,23). The Synechococcus enzyme has thus been an excellent model enzyme for directed evolution in heterologous hosts that lack a carbon-concentrating mechanism (20,23,38,39). Complementation of the R. capsulatus RubisCO deletion strain with Synechococcus sp. strain PCC 6301 RubisCO genes allowed the selection of several mutant substitutions that both positively and negatively influenced activity and interactions with CO 2 or O 2 , resulting in the identification of a semiconserved hydrophobic region adjacent to the active site (23,38,39). Subsequent studies targeting equivalent residues in R. eutropha form I and archaeal Archaeoglobus fulgidus or Thermococcus kodakarensis form III RubisCOs resulted in the identification of mutants with beneficial changes to the enzymes' oxygen sensitivity (25,40,41), leading to the conclusion that this hydrophobic region in divergent enzymes could be a critical contributor for differential interactions with CO 2 and O 2 during catalysis. In the current study, additional residues in this hydrophobic region of the Synechococcus form I RubisCO were analyzed using site-directed mutagenesis. In addition, random mutagenesis and suppressor selection with negativemutant genes resulted in the identification of second-site suppressor mutations in the structural genes encoding both large and small subunits of the enzyme. Detailed structure-function analyses point to the importance of additional hydrophobic regions and the large-small subunit interface for differential interactions with CO 2 and O 2 . Selection of mutant enzymes with enhanced catalytic properties that confer superior CO 2 -dependent growth phenotypes accentuates the utility of nonnative autotrophic host systems for artificial evolution of RubisCO variants with attendant physiological consequences.

Analysis of residues in hydrophobic regions of Synechococcus form I RubisCO.
Previous studies identified mutant substitutions in residues Phe 342 and Ala 375 (Phe 345 and Ala 378 in spinach RubisCO) that led to improved structure-function properties of the enzyme (23,32,39,42,43). These two residues are in a hydrophobic region near the active site, which shows a striking pattern of conservation among the three forms of RubisCO (Table 1; Fig. 1). Residues in this region can directly impact the movement of invariant catalytic residues Lys 331 and Ser 376 (Lys 334 and Ser 379 in spinach RubisCO) during catalysis, thus affecting substrate RuBP binding and CO 2 /O 2 specificity (8). The identity of Ala 375 in the Synechococcus form I RubisCO, or its equivalent in other RubisCOs, appears to be specifically important for differential interactions with CO 2 and O 2 (25,(39)(40)(41). Hence, other conserved and semiconserved nonpolar residues in van der Waals contact with Ala 375 (within 4 Å) were targeted for mutagenic analysis (Table 1; Fig. 1). The invariant Thr 327 was changed to an alanine (neutral) or a valine or a leucine (nonpolar), Phe 391 was changed to an alanine or a leucine (to reflect its identity in other RubisCOs), and Leu 397 was changed to an alanine. Ala 375 was also changed to a leucine because substitution with a shorter (valine) or a bulkier (isoleucine) branched-chain hydrophobic residue resulted in contrasting CO 2 -dependent growth phenotypes of R. capsulatus SB I/II Ϫ (39). Similarly to the wild type, mutants T327A L and F391L L (superscript L refers to a mutant substitution in the rbcL gene) both could support CO 2 -dependent autotrophic growth of the host strain under anoxic conditions but not in the presence of oxygen. None of the other site-directed mutants could complement for CO 2 -dependent growth (Fig. 2). These growth responses accentuate the importance of the indicated residues in this region for enzyme function in vivo.
Selection and phenotypes of second-site suppressors of hydrophobic-pocket residue mutants. Mutant large subunit (rbcL) and wild-type small subunit (rbcS) genes of the negative mutants T327L L , T327V L , A375I L , A375L L , A375S L , F391A L , and L397A L were used as the templates to generate a library of randomly mutagenized rbcLS genes in E. coli, conjugated en masse into R. capsulatus strain SB I/II Ϫ and subjected to direct selection for CO 2 -dependent growth under anaerobic growth conditions. Multiple second-site suppressors were isolated for the negative mutant A375I L , and two secondsite suppressors were isolated for each of the negative mutants A375S L and A375L L ( Fig. 3; see also Table S1 in the supplemental material). A point mutation in mutant T327V L led to the recovery of a pseudosuppressor (V327A L ). No suppressors could be identified for the other negative mutants (i.e., T327L L , F391A L , and L397A L ). For suppressor-mutant genes that had more than one point mutation in the same copy of the rbcLS gene cluster, individual point mutations were created with a template that  carries the original mutation that caused the negative phenotype (Table S1) to determine if one of the individual mutations was sufficient for suppression. Whenever a suppressor mutant was identified, the resultant plasmid that had been reisolated from R. capsulatus and used for DNA sequencing was conjugated back into R. capsulatus

FIG 2
Growth phenotypes of R. capsulatus wild type (strain SB1003) and the RubisCO deletion mutant strains that had been complemented with wild type (WT) or site-directed mutants (labeled with respective residue substitutions) of Synechococcus RubisCO. Mixotrophic growth was assessed on rich (peptone-yeast extract) medium supplemented with tetracycline (to select for plasmid-complemented strains) and a gas mixture comprising 5% CO 2 and 95% H 2 . CO 2 -dependent growth was assessed on minimal medium supplemented with a gas mixture comprising either 5% CO 2 and 95% H 2 gas mixture (photoautotrophy) or 5% CO 2 , 45% H 2 , and 50% air (chemoautotrophy). The RubisCO deletion mutant strain that had been complemented with an empty plasmid was used as a negative control ("-ve ctrl"). Mutant Synechococcus RubisCOs Enhance Growth ® strain SB I/II Ϫ to verify that the phenotype was not an artifact of the selection procedure employed. In some cases, spontaneous mutations arose for negative mutants that were placed under selective growth conditions (i.e., with CO 2 as the sole carbon source) for phenotype verification (Table S1). Spontaneous mutations were also identified with mutants that could support anoxic CO 2 -dependent growth but only when placed in liquid cultures under more stringent selective conditions (i.e., in the presence of oxygen). Three mutants were selected based on their ability to support oxygenic CO 2 -dependent growth of R. capsulatus strain SB I/II Ϫ . Two of them arose from additional mutations in the T327A L mutant rbcL gene background, encoding either an S325L L or a V186I L substitution. Mutant M259T/A375V L //M57I S (superscript "S" refers to a mutation in the rbcS gene; a single shill separates mutations occurring in the same large subunit; a double shill separates large and small subunit mutations) arose from mutant M259T/A375I L //M57I S , which had been selected under anoxic growth conditions (Table S1). Growth in liquid cultures provided a quantitative measure of the CO 2 -dependent growth responses conferred by the suppressors. Suppressors of mutant A375I L , which are able to complement for CO 2 -dependent growth only under anoxic conditions, all had growth rates that were generally lower than that of the wild type ( Fig. S1). The R214H/A375S L suppressor mutant, which was isolated under photoautotrophic (anoxic) growth conditions, and the three mutants selected under chemoautotrophic (oxic) growth conditions conferred better growth than the wild type under both anoxic and oxic conditions (Fig. 4). When created in isolation, V186I L , R214H L , and S325L L single mutants were all able to complement for CO 2 -dependent growth under both anoxic and oxic conditions (Table S1). It is apparent that suppressor selection helped identify multiple regions in the quaternary structure of the enzyme (Fig. 3) that impact the enzyme's interactions with CO 2 and O 2 . Enzymatic properties of recombinant mutant and suppressor enzymes. Net yields of recombinant enzymes with small subunit mutant substitutions were generally low. However, significant levels of soluble RubisCO could be purified from E. coli strains expressing mutant large subunit and wild-type small subunit genes. SDS-PAGE analysis of soluble and insoluble fractions of E. coli lysates indicated that abundantly synthesized subunits of the recombinant L397A L mutant protein were present only in the latter. Supplementing the E. coli expression strain with chaperone DnaK, DnaJ, GrpE, GroEL, GroES, or Tf (TaKaRa) did not help with the assembly of L397A L mutant RubisCO (data not shown). Among the site-directed mutants, T327A L and F391L L mutant substitutions resulted in enzymes that retained about 80% or 40% of the wild-type specific activity, respectively, but other recombinant mutant enzymes were devoid of activity (data not shown). The CO 2 -dependent growth complementation phenotypes of the site-directed mutants are consistent with the in vitro enzymatic activities (Fig. 2). All of the recombinant suppressor-mutant enzymes that could be purified retained lower levels of carboxylation specific activities than the wild type. Thus, the artificial selection procedures utilized in this study did not favor the isolation of mutant enzymes with enhanced carboxylation activity.
The ratio of carboxylase activities measured at limiting CO 2 concentrations under 100% N 2 versus 100% O 2 (N 2 /O 2 ratio) has been previously used to screen for RubisCO enzymes with altered kinetic properties (44). As part of the same assay, parallel determination of carboxylation activities in the presence of excess CO 2 , under 100% N 2 , provides a useful screen and measure of any changes to the enzyme's k cat value for carboxylation (44). Similar assays were performed with mutant Synechococcus Rubis-COs. Several enzymes had lower N 2 /O 2 ratios (Table S2), indicating that these enzymes were likely less inhibited in the presence of 100% O 2 . Further, the activity values measured under 100% N 2 (i.e., absence of O 2 ) with excess CO 2 were reflective of the specific activity values obtained with purified enzymes. Enzymes with substantial levels of carboxylation activities, favorable N 2 /O 2 ratios, and abilities to support better CO 2dependent growth than the wild type (i.e., M259T/A375V L //M57I S , R214H/A375S L , and V186I/T327A L ) were chosen for further analysis of catalytic constants.
Catalytic constants were determined with purified recombinant enzymes. The K m values for CO 2 (K c ) and O 2 (K o ) and the calculated K o /K c ratios were altered for most enzymes ( Table 2), indicating that the residue changes indeed influence interactions with the two gaseous substrates. The significantly higher K o /K c ratios obtained for M259T/A375V L //M57I S triple, V186I L single, V186I/T327A L double, S325 L single, and R214H/A375S L double mutant enzymes likely account for the ability of these enzymes to support vigorous CO 2 -dependent growth under oxic conditions (Fig. 4). Notably, the K c values of M259T L single, M259T/A375V L //M57I S triple, V186I L single, V186I/T327A L double, S325 L single, and S325L/T327A L double mutant enzymes were better than the wild-type value. The V186I/T327A L double mutant enzyme also had a significantly higher K o value than the wild-type enzyme ( Table 2). Despite the enzyme having a reduced k cat value for carboxylation (49% lower than the wild-type value), the superior K o /K c ratio, unaltered values of ⍀ and K RuBP , and superior growth-complementation phenotypes (Fig. 4) indicate that the V186I/T327A L suppressor mutant enzyme may be the best oxygen-tolerant RubisCO to have been artificially evolved thus far, conferring growth enhancement on an autotrophic host under physiologically relevant conditions. Mutant Synechococcus RubisCOs Enhance Growth ® Effect of mutant substitutions on enzyme structure and assembly. A significant number of Synechococcus form I RubisCO mutant enzymes previously isolated by artificial evolution were synthesized in vivo at higher levels or were more stable than the wild-type enzyme (32,39,43). To assess the soluble-protein levels of the selected suppressor mutant enzymes, R. capsulatus cells expressing the corresponding RubisCOs were harvested from photoautotrophically grown cultures, and soluble extracts were prepared using sonication and analyzed via SDS-PAGE and Western blotting. Several mutant enzymes with enhanced synthesis and stability were isolated in this study, including mutants R214H/A375S L and M259T/A375I L //M57I S (Fig. S2). With wild-type levels of RubisCO subunits, M259T/A375V L //M57I S appears to have been preferentially selected for enhanced kinetic properties. No discernible trends were observed with specific activities measured from R. capsulatus cell extracts expressing various mutant proteins, although it could be concluded that RubisCO subunit synthesis and specific activities in cell extracts were significantly reduced in all strains that had been grown in the presence of O 2 (data not shown).
Because many of the kinetically altered suppressor mutant substitution enzymes appeared to be localized to the intersubunit interfaces (Fig. 3), their impact on the strength of large-large and large-small subunit interactions were assessed further using a bacterial two-hybrid system that had been previously utilized to show interactions between RubisCO subunits and regulator proteins (45). The levels of ␤-galactosidase provided a direct measure of the strength of large-small subunit interactions (Fig. S3). Several conclusions could be drawn from the two-hybrid analysis. The interaction strength of various large-small subunit pairs generally correlated with yields of recombinant enzymes that were purified from E. coli. For example, the interaction strengths of each of A375V L , A375I/A411T L , R214H/A375S L , V186I/T327A L , and S325L/T327A L mutant large subunits with wild-type small subunit were better than what was measured with the wild-type large subunit ( Fig. S3B and C). This correlated with consistently higher yields of the corresponding mutant RubisCOs in independent protein purification experiments (data not shown). Mutant substitutions appeared to selectively impact the large-small subunit interactions and not those between the large subunits themselves (Fig. S3A). The interaction strengths of a large-small subunit pair did not seem to be a determinant of the corresponding holoenzyme's ability to support CO 2 -dependent growth of R. capsulatus SB I/II Ϫ (Fig. S3B to D and Table S1). Last, substitutions in the small subunit appeared to generally diminish the strength of their interaction with the corresponding mutant large subunits, whereas the large subunit suppressor mutant substitutions appeared to generally improve the interaction strength with wild-type small subunits (Fig. S3C and D). In summary, the two-hybrid interaction strengths are reflective of the extent of subunit interactions that define protein assembly and turnover in vivo, particularly when the mutant proteins are expressed as recombinant enzymes in E. coli. Despite the enzyme having only ϳ12% of the wild-type level of activity, the improved structural stability was an important determinant of the positive phenotype conferred by the A375V L mutant enzyme (39). Thermal stability assays were performed to further assess the stabilities of recombinant enzymes. Whereas the M259T L single, M259T/A375V L //M57I S triple, and V186I/T327A L double mutant enzymes lost only 2 to 14% activity after a 5-min incubation at 60°C, the wild-type enzyme lost 24% of its initial activity after 5 min (Fig. S4). However, after a 60-min incubation, only the M259T L mutant enzyme retained higher levels of activity than the wild-type sample. Thus, although several suppressor mutant substitutions appear to enhance structural interactions, all of them do not confer physiologically significant phenotypes.

DISCUSSION
In this study, directed evolution resulted in the isolation and selection of cyanobacterial form I RubisCO mutant proteins with kinetic alterations that enhance CO 2dependent growth. The ability to artificially evolve enzymes that improve growth, specifically in the presence of O 2 , such as V186I/T327A L and M259T/A375V L //M57I S , provides direct evidence that RubisCO can be functionally improved to play a physiologically significant role. In addition to positive kinetic variants, the selection procedures described here also allowed for the isolation of proteins with improved structural integrity. Structure-functional divergence and enhanced knowledge of the promiscuity of the RubisCO family of proteins have provided a better understanding of nature's constraints governing the evolution of physiologically relevant enzymatic properties such as oxygen tolerance. Several studies have highlighted the importance of molecular chaperones and other accessory proteins for gene expression (46), functional assembly, activity regulation, and evolvability of divergent RubisCO molecules in heterologous hosts (4,9). Despite the constraints placed by these requirements, heterologous RubisCO genes have been successfully expressed and the resultant proteins functionally assembled in hosts like E. coli, R. capsulatus, R. palustris, and R. eutropha, utilizing only the native host cell's regulatory and assembly machineries (9,20). When the wild-type Synechococcus rbcLS genes are expressed in E. coli or R. capsulatus, the amounts of RubisCO subunits in the soluble fractions are normally small, resulting in the selection and isolation of several suppressor mutants with substitutions that confer higher levels of RubisCO protein in the soluble fraction (see Fig. S3 in the supplemental material) (32). However, the selection of suppressors V186I/T327A L and S325L/T327A L from mutant T327A L and R214H/A375S L from mutant A375S L and the sequential evolution of mutant M259T/A375V L //M57I S from mutant A375I L is direct evidence that physiologically significant growth enhancements are achievable via primary changes to the functional properties of the enzyme, specifically the altered interactions of these enzymes with substrates CO 2 and/or O 2 .
Utility of RubisCO bioselection systems with autotrophic growth capabilities. Autotrophic bacteria such as R. capsulatus, R. palustris, and R. eutropha have been exploited for selection studies with heterologous RubisCOs (23-25, 39, 47, 48). The absolute dependence on RubisCO for CO 2 -dependent growth, coupled with the ability to grow under heterotrophic growth conditions (i.e., RubisCO and the Calvin-Benson-Bassham [CBB] cycle are dispensable), allows for a convenient means to select suppressor mutations in RubisCO genes that overcome an initial negative-growth phenotype. In this study, we identified and selected multiple mutants of Synechococcus sp. PCC 6301 form I RubisCO with the R. capsulatus host strain cultured at various levels of stringency, thereby identifying structural regions of the enzyme (i.e., bulky nonpolar side chains in hydrophobic regions) that were altered to enhance function (i.e., K c or K o values) and support CO 2 -dependent growth. Convergent identification of several residues in the Synechococcus enzyme using both R. capsulatus (Table S1) and E. coli RubisCO selection systems (20,32,49) indicates that the regions surrounding these residues may be the most readily accessible hot spots for targeted improvements in RubisCO's properties under aerobic conditions.
The importance of hydrophobic regions and the large-small subunit interface in form I RubisCOs. In a previous study, molecular dynamics (MD) simulations were performed to investigate the movement of CO 2 and O 2 in and around RubisCO. It was proposed that all form I RubisCOs are able to preferentially sequester CO 2 in hydrophobic regions that are continuous and connect to the active site (50). These simulations also pointed out that the small subunits may act as CO 2 reservoirs. If this is true, the mutant substitutions identified in this study may represent the various CO 2sequestering regions. It is thus reasonable to expect these regions to impact net CO 2 availability in the vicinity of the active site.
Residues Phe 342 and Ala 375 have been identified via selection in this and previous studies (20,23,32,42,43,49). Substitution of Phe 342 or the equivalent residue in other RubisCOs mostly affected the enzyme's structural stability or the K RuBP (20,25,32,49). However, substitution of Ala 375 in the Synechococcus form I enzyme or its equivalent in other enzymes resulted in beneficial changes to the enzymes' K c and K o values (25,39,41). Suppressor selection in R. eutropha with an A380V L negative mutant identified T330A L or Y348C L as a second-site suppressor (25). These amino acid residues (Ser 325 and Val 343 in Synechococcus RubisCO) are part of the same hydrophobic region in which the T327A L and S325L L mutant substitutions are present (Fig. 1). It was thus not surprising that the presence of an S325L L mutant substitution complements the T327A L single mutant enzyme for better CO 2 fixation under oxic conditions. Two independently selected suppressor mutants (M259T/A375I L //M57I S and M259T/ A375S L ) both had an M259T L mutation. M259T L has also been selected in previous studies (23,32,43,49). The side chain of Met 259 is in a second hydrophobic region that is adjacent to the central solvent channel and the large-small subunit interface (Fig. 5). Although far away from the active site, the M259T L mutant substitution confers structural stability and an improved K c value, consistent with what has been reported before (32,43). These changes account for the improved CO 2 -dependent growth phenotype conferred by M259T L . Suppressor selection and targeted modification of large and small subunit residues in this region led to the creation of a Chlamydomonas RubisCO variant with enhanced kinetic properties resembling a plant enzyme (37,(51)(52)(53)58). This provided a rationale for understanding how the M259T L and M57I S substitutions could complement to restore function in the A375I L negative mutant and further lead to the accumulation of an IA375V L pseudosuppressor, resulting in a positive phenotype. Other small subunit suppressors were isolated for the Ala 375 mutants, but they resulted in either reduced recombinant protein yields or low specific activities, precluding them from further analysis.
Val 186 is a conserved residue in a third hydrophobic region that connects the large-small subunit interface near Met 259 to the other side of the active site via Lys 172 (Lys 175 in spinach RubisCO) (Fig. 3B). A V186I L single mutant was identified and analyzed in a previous study (20), and its enzymatic properties are generally consistent with what is reported here (Table 2). However, the isolation of V186I L as a second-site suppressor of T327A L in this study indicates that the two distal hydrophobic regions ( Fig. 1 and 3) are likely connected via their coordinated interactions with incoming CO 2 . Val 186 is surrounded by other conserved residues that have also been identified via suppressor selection in this study (Fig. 3B). Whereas a G176S L substitution was isolated as a suppressor for the A375I L negative-mutant substitution (Table S1), a previous bioselection screen identified G176D L as a negative mutant with altered CO 2 interactions (38). Substitution of the conserved Cys 192 in the Chlamydomonas form I RubisCO (Cys 189 in the form I Synechococcus RubisCO) resulted in an enzyme with an altered K o /K c ratio (54).
Although several residues that were targeted or identified in this study have been independently identified and analyzed in previous investigations, the unique physiological context provided by the R. capsulatus system underscores the functional significance of complementing structural interactions. This study has also resulted in the concerted identification of other conserved residues in the large-large (Phe 37 , Ala 53 , Lys 249 , and His 307 ) and the large-small subunit interfaces (residues Asn 181 , Gly 192 , Arg 214 , Glu 228 , Ala 411 , and Glu 422 in the large subunit and residues Ser 16 , Gln 29 , Glu 43 , Tyr 54 , Met 57 , and Leu 72 in the small subunit). Suppressor-mutant combinations involving these residues provide new insights regarding complementary structural interactions that may be targeted for evolving RubisCO variants with more predictable structure-function properties.
In conclusion, the R. capsulatus selection strategy has been successfully employed to evolve Synechococcus RubisCO variants with selective improvements in the enzyme's interactions with CO 2 versus O 2 . These results bode well for performing directed evolution and selection studies with other RubisCOs that may be functionally expressed in R. capsulatus. Because prokaryotic RubisCOs function under diverse metabolic contexts (1), it should be possible to learn more about the enzyme's structure-function relationships by tapping into diverse microbial genomes for potentially "evolvable" RubisCO genes. As previously indicated (34,35), the vastly uncultured "microbial dark matter" could be yet another treasure trove for identifying structurally diverse RubisCO genes. Conserved identities of residues identified in this study will facilitate targeted approaches to improve RubisCO's performance in other divergent organisms that contribute significantly to global CO 2 fixation.

MATERIALS AND METHODS
Bacterial strains, culture conditions, and plasmids. Strains SB1003 and SB I/II Ϫ are the wild-type and RubisCO deletion strains, respectively, of Rhodobacter capsulatus (25). R. capsulatus was cultured aerobically under chemoheterotrophic or chemoautotrophic (CO 2 -dependent) growth conditions and anaerobically under photoautotrophic (CO 2 -dependent) growth conditions at 30°C, as described previously (39). Top10 (Thermo Fisher) or XL1-Blue MRF=(Agilent) strains of E. coli were used for cloning procedures. Strain S17-1 (ATCC 47055) was used for mobilizing plasmids into R. capsulatus strain SB I/II Ϫ (47). Strain FW102 (55), which contains the lac operon under the control of the operator, was used as a reporter strain for bacterial two-hybrid assays. Strain BL21(DE3) was used for recombinant protein synthesis. In some cases, this strain was supplemented with chaperone plasmids (TaKaRa) to facilitate assembly of poorly soluble proteins. E. coli cells were cultured aerobically in lysogeny broth (LB) at 37°C with shaking at 250 rpm. For protein purifications, cells were grown, induced, and harvested as described previously (39). A pUC19 clone with Synechococcus rbcLS genes (23) was used as a template for site-directed mutagenesis. A broad-host-range plasmid, pRPS-MCS3, was used for complementation studies with R. capsulatus strain SB I/II Ϫ (39). Plasmids pET28a and pET11a (Novagen) were used for gene expression in E. coli.
Mutant Synechococcus RubisCOs Enhance Growth ® Mutagenesis, molecular biology procedures, matings, and selection. Site-directed mutagenesis was performed using a QuikChange kit (Agilent). Random mutagenesis of Synechococcus rbcLS genes was accomplished using either error-prone PCR amplification (39,47) or chemical mutagenesis using N-methyl-N=-nitro-N-nitrosoguanidine (MNNG) as a mutagen and a previously described procedure (59) that was modified as described here. E. coli cells with template plasmids were grown to an optical density (OD) of 0.7 at 600 nm; washed and resuspended in 0.1M citrate buffer, pH 5.5, with MNNG added to a final concentration of 200 g ml Ϫ1 ; and incubated at 30°C for 30 or 60 min. Cells were then washed in 0.1M phosphate buffer, pH 7.0, and grown for 1h in selective LB medium prior to conjugation. Nonmutagenic PCRs were performed with PrimeSTAR GXL DNA polymerase (Clontech). Cloning procedures utilized restriction enzymes and T4 DNA ligase purchased from Thermo Fisher Scientific or New England Biolabs. DNA sequences were verified by Sanger DNA sequencing (Plant-Microbe Genomics Facility, The Ohio State University).
Plasmids were mobilized from E. coli into R. capsulatus using either a triparental or diparental mating strategy (39,47). After mating, the recipient R. capsulatus host cells were selected on either chemoheterotrophic (antibiotic selection) or autotrophic (CO 2 -dependent growth selection) media (39,47). Natural selection for spontaneous mutations occurred in liquid or solid autotrophic media with R. capsulatus strains containing negative-mutant (no activity supported) RubisCO genes.
Bacterial two-hybrid assays. The BacterioMatch II two-hybrid system (Agilent) was used to compare interaction strengths between wild-type and mutant RubisCO subunits. The genes encoding the large (rbcL) and small (rbcS) subunits were cloned into the pTRG (target) and pBT (bait) plasmids, respectively. Reporter assays were carried out as described previously (45).
Preparation of cell extracts, purification of RubisCO, and biochemical and structure analysis. Autotrophically grown R. capsulatus liquid cultures were harvested after reaching the stationary phase (OD, ϳ1.2 to 1.5 at 660 nm) by centrifugation at 8,000 ϫ g for 10 min at 25°C and washed and sonicated in Bicine buffer (50 mM Bicine-NaOH, 10 mM MgCl 2 , 10 mM NaHCO 3 , 1 mM dithiothreitol [DTT], pH 8.0). Supernatants were obtained by centrifugation at 16,100 ϫ g for 10 min at 4°C and used for RubisCOspecific activity measurements, SDS-PAGE, and Western blot analyses (38,39). Identical amounts of soluble protein (5 g) were used in SDS-PAGE and Western blot analyses. RubisCO enzymes were purified as recombinant proteins with or without an N-terminal hexahistidine tag using plasmids pET11a and pET28a, respectively, using a two-step (39) or a three-step (47) procedure. Proteins were dialyzed into a Bicine buffer (50 mM Bicine-NaOH, 10 mM MgCl 2 , 1 mM DTT, 10 mM NaHCO 3 ), concentrated using Amicon filters (MilliporeSigma), mixed with 20% glycerol, and stored as aliquots at Ϫ80°C. Protein concentrations were determined using the Bradford method using a dye reagent (Bio-Rad). All biochemicals were purchased from Sigma-Aldrich. PyMOL was used for protein structural analysis.
Enzyme assays and determination of kinetic constants. RuBP carboxylase activities were measured using radiometric assays that utilized NaH 14 CO 3 (56) (Perkin Elmer). For the initial characterization, assays were performed with excess or limiting amounts of CO 2 in vials flushed with 100% N 2 or 100% O 2 (44). Thermal stabilities of purified recombinant enzymes were determined by incubating aliquots of each enzyme (5 U) at 60°C for various times, cooling on ice, and determining remaining RuBP carboxylation activities at 25°C. Activity values were normalized against the activity in the untreated samples. Substrate specificity (⍀) values were determined with purified enzymes (50 g) in assays performed under saturating (1.23 mM) oxygen concentrations (23). The [1-3 H]RuBP that was required for these assays was synthesized and purified using standard methods (57). Kinetic constants k cat , K c , K o , and K RuBP were determined using procedures described elsewhere (39,47).