Role of Small Subunit in Mediating Assembly of Red-type Form I Rubisco

Background: Rubisco, a key photosynthetic enzyme of eight large and eight small subunits, is phylogenetically divided into green and red types. Results: The small subunits of red-type Rubisco contain a C-terminal β-hairpin extension that mediates efficient assembly of the holoenzyme. Conclusion: The C-terminal β-hairpin renders red-type Rubisco independent of specialized assembly chaperones. Significance: These findings can help in bioengineering red-type Rubisco into crop plants. Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) is the key enzyme involved in photosynthetic carbon fixation, converting atmospheric CO2 to organic compounds. Form I Rubisco is a cylindrical complex composed of eight large (RbcL) subunits that are capped by four small subunits (RbcS) at the top and four at the bottom. Form I Rubiscos are phylogenetically divided into green- and red-type. Some red-type enzymes have catalytically superior properties. Thus, understanding their folding and assembly is of considerable biotechnological interest. Folding of the green-type RbcL subunits in cyanobacteria is mediated by the GroEL/ES chaperonin system, and assembly to holoenzyme requires specialized chaperones such as RbcX and RAF1. Here, we show that the red-type RbcL subunits in the proteobacterium Rhodobacter sphaeroides also fold with GroEL/ES. However, assembly proceeds in a chaperone-independent manner. We find that the C-terminal β-hairpin extension of red-type RbcS, which is absent in green-type RbcS, is critical for efficient assembly. The β-hairpins of four RbcS subunits form an eight-stranded β-barrel that protrudes into the central solvent channel of the RbcL core complex. The two β-barrels stabilize the complex through multiple interactions with the RbcL subunits. A chimeric green-type RbcS carrying the C-terminal β-hairpin renders the assembly of a cyanobacterial Rubisco independent of RbcX. Our results may facilitate the engineering of crop plants with improved growth properties expressing red-type Rubisco.

The enzyme Rubisco 3 catalyzes the carboxylation of the fivecarbon sugar ribulose 1,5-bisphosphate in the dark reactions of photosynthesis, thereby mediating the conversion of atmospheric CO 2 into usable sugars. In its most abundant form (form I), Rubisco is a hexadecameric complex composed of eight large (RbcL, ϳ50 kDa) and eight small subunits (RbcS, ϳ12-15 kDa) (1,2). The RbcL subunits form a cylindrical tetramer of anti-parallel dimers that is capped by four RbcS subunits at the top and four RbcS subunits at the bottom. Based on the alignment of RbcL sequences, form I Rubisco is phylogenetically divided into green and red branches (3)(4)(5). Greentype Rubisco occurs in cyanobacteria, green algae and plants, whereas the red type is found in photosynthetic bacteria, nongreen algae and phytoplankton.
Despite its central role in the global carbon cycle, Rubisco is a slow catalyst. Moreover, it discriminates poorly between CO 2 and O 2 (6). Although the fixation of CO 2 results in the synthesis of usable sugars, the competing reaction with O 2 (photorespiration) decreases the efficiency of carbon fixation (7,8). Extensive efforts have focused on engineering a Rubisco enzyme with improved specificity for CO 2 (1, 9 -12). Interestingly, compared with their green-type homologs, several of the red-type enzymes display a Ͼ2-fold higher specificity for CO 2 coupled with a lower Michaelis constant for CO 2 (13). Specificity for CO 2 over O 2 (S c/o ) determines preferential carboxylation over oxygenation of ribulose-1,5-bisphosphate during Rubisco catalysis. A comparison of the kinetic properties of different forms of Rubisco suggests an inverse relation between S c/o and turnover rate. So far, the highest specificity factor (S c/o, 167) is reported for the Rubisco from the filamentous red algae Griffithsia monilis, which has a turnover rate of 2.6 s Ϫ1 (13). By comparison, the bacterial red-type enzyme from Rhodobacter sphaeroides possess a similar CO 2 turnover rate (ϳ2.0 to 2.9 s Ϫ1 ), but a significantly lower S c/o (ϳ55) (12,14,15). It has been proposed that transplantation of the G. monilis Rubisco into C3 plant chloroplasts would raise crop carbon gain by up to ϳ30% without any increase in the amount of Rubisco per unit leaf area at ambient conditions by improving the efficiency of light, water and nitrogen usage (11,12). However, when the rbcLS operon from the red algae Galdieria sulfuraria (Gs) or the diatom Phaeodactylum tricornutum (Pt) was inserted into the tobacco plastid genome, the heterologous Rubisco subunits were expressed abundantly but failed to assemble into an active holoenzyme (13), suggesting incompatibility with the chloroplast chaperones involved in the folding and assembly of greentype Rubisco. Folding of the RbcL subunit is mediated by the chaperonin system: GroEL/ES in proteobacteria and cyanobacteria and Cpn60/Cpn10/Cpn20 in algae and plants (16 -18). Moreover, several factors have recently been reported to be required for the efficient assembly of the RbcL 8 S 8 complex. These factors include the assembly chaperone, RbcX, and the Rubisco accumulation factors 1 and 2 (RAF1; RAF2), which are structurally unrelated but conserved in most organisms harboring green-type Rubisco (19 -25). A combination of GroEL/ES and RbcX allowed the in vitro reconstitution of cyanobacterial Rubisco from Synechococcus sp. PCC 6301 (Syn6301) (20). Interestingly, organisms containing red-type Rubisco lack the rbcX or RAF1 genes (26 -29), suggesting an alternative mode of assembly.
Here, we analyzed the chaperone requirement for the in vitro folding and assembly of a prokaryotic red-type form I Rubisco from the proteobacterium Rhodobacter sphaeroides (Rs). We show that the folding of the RbcL subunit requires the GroEL/ES chaperonin, but the assembly to holoenzyme is independent of accessory factors. Instead, assembly is mediated by the RbcS subunits in a manner dependent on red-type specific C-terminal ␤-hairpin extensions. The extensions of the four RbcS subunits that cap the RbcL 8 complex form an eight stranded ␤-barrel structure, partially occupying the central solvent channel of the holoenzyme (2,30,31). Mutational analysis and the finding that red-type RbcS can mediate green-type Syn6301RbcL assembly in the absence of RbcX, suggest that the red-type RbcS fulfills the role of an assembly factor.

EXPERIMENTAL PROCEDURES
Molecular Biology-The rbcL and rbcS genes from R. sphaeroides were amplified from the plasmid pET30bRsrbcLS (15) using restriction sites NdeI/BamHI sites to generate the plasmids pET30bRsrbcL and pET30bRsrbcS, respectively. A C-terminally truncated version of the rbcS gene comprising residues 1-106 was also generated (pET30bRsrbcS⌬EF). The QuikChange protocol (Stratagene) was employed to introduce point mutations in RsRbcS. P. tricornutum rbcS was amplified from the total genomic DNA and cloned into pET11a plasmid using the NdeI/BamHI restriction sites resulting in pET11aPtrbcS. G. sulfuraria rbcS was cloned into pET11a from the pGs plasmid comprising the plastome fragment encoding rbcL-rbcS-cbbX. A hybrid version of Syn6301RbcS containing the C-terminal extension of RsRbcS (residues 106 -129) was generated as follows: the three penultimate amino acids Gly-109, Arg-110, and Tyr-111 of Syn6301RbcS in pET11aSyn6301rbcS (19) were deleted using QuikChange protocol (Stratagene) to generate the plasmid pET11aSyn6301rbcS⌬GRY. Primers were designed to introduce the RsRbcS C-terminal extension resulting in pET11aSyn6301rbcS⌬GRY-RsrbcSEF. The plasmids for expression of the cyanobacterial proteins Syn6301RbcL, Syn6301RbcS, Syn7002RbcS, and N-terminally His 6 -tagged AnaCARbcX are described by Saschenbrecker et al. (19). Coding regions of all constructs were verified by DNA sequencing.
Protein Purification-Unless stated otherwise, all purification procedures were performed at 4°C. Concentrations of pure proteins were determined spectrophotometrically at 280 nm. E. coli GroEL and GroES were purified as described previously (32,33). The Syn6301RbcL 8 was purified as described by Saschenbrecker et al. (19). N-terminally tagged AnaCARbcX was purified according to the protocol of Liu et al. (20). Expression and purification of Rubisco from R. sphaeroides was performed as described by Mueller-Cajar et al. (15). Purification of denatured RsRbcL and Syn6301RbcL from inclusion bodies was performed essentially as described previously (19,20). The proteins were stored in buffer (40 mM Tris, pH 8.0, 250 mM sucrose, 10 mM EDTA, 6 M GuHCl) at Ϫ80°C. RbcS proteins were also purified from inclusion bodies but were refolded during dialysis against 25 mM Tris, pH 8.0, 50 mM NaCl (19). Purified RbcS proteins (see Fig. 4A) were concentrated and stored in buffer containing 25 mM Tris pH 8.0, 50 mM NaCl, 10% (v/v) glycerol, and 5 mM DTT.
Rubisco Refolding-GroEL-assisted refolding of GuHCl-denatured RsRbcL was performed at 25°C as described previously (20) with minor modifications. Denatured RsRbcL was diluted to 0.5 M final concentration into ice-cold refolding buffer (20 mM MOPS-KOH, pH 7.5, 100 mM KCl, 5 mM Mg(OAc) 2 , 10 mM DTT, 0.5 mg ml Ϫ1 BSA) containing 1 M GroEL. Final concentration of GuHCl in the refolding assay was maintained at Ͻ30 mM. RbcL not captured by GroEL and aggregated was removed by centrifugation at 16,200 ϫ g for 5 min at 4°C followed by addition of 2 M GroES. Refolding was initiated by the addition of 10 mM ATP and stopped by addition of apyrase (0.25 units/ l). Addition of RbcS and other proteins was as specified in the figure legends. CABP was added at a concentration of 0.1 mM when indicated. Reactions were analyzed by 6% native PAGE or NuPAGE® Novex® 3-8% Tris acetate (Invitrogen) or SDS-PAGE followed by immunoblotting with anti-RbcL antibody where indicated. RsRbcL was detected with antibody from Agrisera and Syn6301RbcL with antibodies raised in rabbits (20). Rubisco activity was analyzed as described previously with modifications (20,34). Activities of in vitro reconstituted enzyme are expressed in percent of native control, which refers to the activity of recombinant, purified holoenzyme.
CABP-binding Assay-RsRbcL was expressed in E. coli with or without overexpression of GroEL/ES. Soluble cell lysate was obtained by centrifugation of crude cell lysate at 16,200 ϫ g for 15 min at 4°C. Equivalent amounts of lysate containing RsRbcL were incubated with purified RsRbcS and holoenzyme formation was allowed to proceed for 15 min at 25°C. The concentration of Rubisco active sites was measured by stoichiometric binding of the tight binding inhibitor [2-14 C]carboxyarabinitol-P2 after preincubation with 25 mM NaHCO 3 and 15 mM MgCl 2 , as described previously (35).
Thermal Stability Assay-The apparent melting temperature (T m ) of RsRubisco (RsRbcL 8 S 8 ) and mutant variants was determined as described (10,36). Equal concentrations of soluble protein from E. coli lysates expressing RsRbcL were incubated with wild-type or mutant RbcS for 30 min at 25°C to allow formation of holoenzyme. Rubisco activities were measured as above (at 25°C) after incubation of lysate containing holoenzyme at increasing temperatures from 30 to 70°C for 10 min, followed by incubation on ice for 5 min. Activities measured at 30°C were set to 100%.

RsRbcL Is Expressed in E. coli in a Soluble
, Assembly-competent Form-We have observed previously that green-type cyanobacterial RbcL is generally highly insoluble upon expression in E. coli, even when GroEL/ES is overexpressed (19). In contrast, red-type RbcL from R. sphaeroides (RsRbcL) was 16 Ϯ 3% soluble upon expression in E. coli cells containing endogenous GroEL/ES levels, and 33 Ϯ 2% soluble protein was recovered when GroEL/ES was transiently overexpressed from an arabinose-inducible plasmid before induction of RsRbcL with IPTG ( Fig. 1A). While at higher concentration (when GroEL/ES was overexpressed), the soluble RsRbcL migrated on native-PAGE as a smear of poorly resolved species consistent with a dynamic ensemble of oligomeric states, it was hardly detectable at lower concentration (Fig. 1B). The soluble RsRbcL was assemblycompetent and formed the RbcL 8 S 8 holoenzyme upon addition of purified RsRbcS to the cell lysate (Fig. 1B). The RbcL 8 S 8 complex was enzymatically active as reflected by its ability to bind the substrate analog [2-14 C]carboxyarabinitol-P2 (35) (Fig. 1C). Overexpression of GroEL/ES resulted in a ϳ2.6-fold increase in holoenzyme protein and a ϳ3.4-fold increase in Rubisco active sites.
To obtain more detailed insight into RsRubisco folding and assembly, we performed reconstitution experiments using purified proteins in vitro. Upon dilution of denatured RsRbcL from 6 M GuHCl into GroEL containing refolding buffer in the absence of ATP, RsRbcL bound stably to GroEL and migrated as a high molecular weight complex on native PAGE ( Fig. 2A, panel i). Addition of GroES and ATP to start the GroEL/ES reaction cycle resulted in the time-dependent release of the folded RsRbcL subunit ( Fig. 2A, panel ii). This RsRbcL remained soluble (Fig. 2B), but was enzymatically inactive and not detectable by native-PAGE as a defined species (Fig. 2, A, panel ii, and C). Thus, RsRbcL differs from the behavior of green-type Syn6301RbcL which retains affinity for GroEL after folding, unless RbcX is present (20). When RsRbcS was added during GroEL/ES-mediated folding, enzymatically active RsRbcL 8 S 8 complex formed at an apparent half-time of ϳ20 min and with a high yield of ϳ80% (Fig. 2, A, panel iii, and C). As the GroEL/ES reaction was stopped with apyrase at different times to deplete ATP, this time course reflects the kinetics of RsRbcL subunit folding and release from GroEL. Folded RsRbcL remained assembly competent as demonstrated by the addition of RsRbcS after refolding ( Fig. 2A, panel iv, and C). These results show that RsRubisco assembly is independent of a separate auxiliary assembly factor. The folding properties of RsRbcL differ from that of Syn6301RbcL. Moreover, RsRbcS appears to have a critical role in shifting a dynamic ensemble of RbcL oligomers to the well defined holoenzyme complex.
The C-terminal ␤-Hairpin Extension of RsRbcS Mediates Assembly and Stability-In the assembly of Syn6301RbcL, RbcX functions in stabilizing antiparallel RbcL dimers and facilitates formation of the RbcL 8 complex. RbcS binding then results in a conformational change in the RbcL subunits, thereby causing the displacement of RbcX (21). In the case of RsRubisco, the RsRbcS subunits might play a similar role in assembly by linking RsRbcL dimers. All red-type RbcS subunits possess a ϳ25-amino acid C-terminal extension that forms a ␤-hairpin structure, consisting of ␤-strands ␤E and ␤F (Fig.  3A). As illustrated by the structure of the red-type Rubisco of Alcaligenes eutrophus, the ␤EF hairpins from four adjacent RbcS subunits are assembled to an eight-stranded ␤-barrel that protrudes into the central solvent channel of the holoenzyme (Fig. 3B). The RbcS ␤-barrel engages in extensive hydrogen bond and salt bridge contacts with the RbcL subunits (2,30,31,37). Tyr-123 in ␤F is able to form a hydrogen bond with Asp-226 in helix ␣8 of RbcL (Fig. 3C). Furthermore, Arg-113, Glu-115, and Arg-119 of the ␤EF hairpin stabilize the conformation of the conserved ␤AB loop (residues 40 -45) (Fig. 3A) in the adjacent RbcS subunit through a hydrogen bond between Arg-119 and Asp-40 and -stacking interactions between the Arg-113 guanidinium group and the imidazole ring of His-42 (Fig.  3C). The ␤AB loop in turn packs against the RbcL subunit of the adjacent RbcL dimer (at helices ␣5 and ␣8) involving a hydrogen bond of Arg-44 (conserved in red-type RbcS) to the C-terminal end of RbcL helix ␣5 at residue Asp-163. In addition, the guanidinium moiety of Arg-44 makes electrostatic interactions with the helix ␣5 dipole and Asp-166 of RbcL. Thus, the RbcS ␤-barrel formed by ␤EF hairpins is expected to confer additional stability to the red-type Rubisco relative to the greentype holoenzyme. The RsRbcS subunits may either first interact with the RsRbcL anti-parallel dimer, followed by holoenzyme formation, or they may first form a tetramer platform on to which RbcL assembles.   To investigate the contribution of the C-terminal ␤-hairpin of RbcS to assembly, we either truncated the C terminus of RsRbcS by 23 residues to generate RsRbcS⌬EF or made point mutations in the highly conserved residues Asp-40, Arg-44, Glu-106, Arg-113, Glu-115, Arg-119, and Tyr-123 to alanine. The mutant proteins were purified from inclusion bodies and refolded upon removal of denaturant by dialysis (Fig. 4A). When added to an in vitro reconstitution reaction, the mutants of the ␤EF hairpin (R113A, R119A, Y123A, and ⌬EF) failed to mediate the formation of holoenzyme detectable on native PAGE, whereas complex formation was preserved with mutants E106A and E115A (Fig. 4B). Similarly, mutation of the conserved residues Asp-40 and Arg-44 in the ␤AB loop (Fig.  3A) impaired holoenzyme formation (Fig. 4B). Surprisingly, the RbcS mutants were nevertheless able to reconstitute RsRubisco activity to wild-type level, except for RsRbcS⌬EF, which resulted in a ϳ30% lower yield (Fig. 4C).
The presence of Rubisco activity, despite the absence of detectable RbcL 8 S 8 complex on native PAGE, may be explained by a stabilizing effect of the sugar substrate ribulose-1,5-bisphosphate, which was absent during reconstitution and native PAGE. Note that the antiparallel RbcL dimer has two sugar binding sites formed by the subunit interface, and substrate binding has been reported to stabilize the dimer (2,38,39). To test this possibility, we added the high affinity substrate analog 2-carboxyarabinitol-1,5-bisphosphate (CABP) during reconstitution (Fig. 4D). In the presence of CABP, the effect of mutations in the C-terminal ␤-hairpin and the ␤AB loop were compensated to varying degrees, and RsRbcL 8 S 8 complexes were clearly observed on native-PAGE (Fig. 4D), confirming a stabilizing effect of the sugar substrate. Note that the migration behavior of some of the mutant complexes varied relative to wild-type, which may reflect differences in charge and shape properties.
We next analyzed the thermal stability of Rubisco holoenzyme containing wild-type (WT) or mutant RbcS. RsRbcL was expressed in E. coli and holoenzyme reconstituted by adding RbcS subunits. The lysates were then incubated for 10 min at increasing temperatures up to 70°C, followed by Rubisco activity assay at 25°C. The apparent melting temperature (T m ) of RsRbcL 8 S⌬EF 8 (ϳ47°C) was reduced by ϳ6°C compared with WT RsRubisco (ϳ53°C) (Fig. 5A    RbcL and RbcS were co-expressed in E. coli, and lysates were incubated at the indicated temperatures for 10 min, followed by cooling on ice and activity assay at 25°C. Activities measured after incubation at 30°C were set to 100%. Shown is S.D. from at least three independent experiments. B, carboxylation activity during in vitro reconstitution of RsRubisco with WT RsRbcS or RsRbcS⌬EF. Experiments were performed essentially described as in Fig. 2C (panel iii).
substantial stabilizing effect of the ␤-hairpin. Moreover, in vitro reconstitution showed that the kinetics of Rubisco assembly with RbcS⌬EF was ϳ3.5-fold slower than with WT RsRbcS (Fig.  5B). The RbcS point mutations in the ␤AB loop and the C-terminal ␤-hairpin also reduced the melting temperature (Table  1), except for E106A (ϳ53°C) and E115A (ϳ54°C), consistent with the ability of these mutants to produce stable RbcL 8 S 8 on native-PAGE in the absence of CABP (Fig. 4B). In summary, the RbcL 8 S 8 complex can form in the absence of the ␤-hairpin loop, but presence of the ␤EF region strongly accelerates the assembly process and stabilizes the holoenzyme. Heterologous Red-type RbcS Mediates Rubisco Assembly but Does Not Confer Catalytic Activity-To investigate whether RbcL 8 S 8 assembly correlates with catalytic activity, we used various red-and green-type RbcS subunits with RsRbcL in the in vitro reconstitution assay. As analyzed by native PAGE, the folded RsRbcL assembled with the RbcS subunits from the red algae G. sulfuraria and the red-type diatom P. tricornutum (Pt) but with none of the tested green-type subunits (Fig. 6A), consistent with the requirement of the C-terminal ␤-hairpin region for assembly. Notably, the heterologous RbcL 8 S 8 complexes were essentially enzymatically inactive (Fig. 6B). This indicates that heterologous red-type RbcS is competent for the assembly but not sufficient for activity. RbcS proteins are generally more divergent (ϳ25-30% identity) in sequence than RbcL subunits (ϳ50 -60% identity) and are apparently critical for holoenzyme activity through long range interactions between RbcS and the active site of the cognate RbcL dimer. Thus, structural competence for assembly does not correlate with the capacity to generate catalytically active enzyme.
Red-type RbcS Subunits Can Overcome the Requirement of Green-type RbcL for the RbcX Assembly Factor-Based on the results above, we speculated that the absence of the ␤-hairpin region in green-type RbcS may explain the requirement of certain cyanobacterial Rubiscos for the RbcX assembly chaperone (19 -21). We tested this using green-type RbcL from Syn6301, which requires RbcX for assembly with RbcS in vitro and remains bound to GroEL in the absence of RbcX (20). In vitro reconstitution confirmed that the formation of enzymatically active Syn6301RbcL 8 S 8 complex was dependent on RbcX (Fig.  7, A and B). Note that the yield of this reaction is limited to ϳ20% due to the inefficient release of Syn6301RbcL from GroEL (20). RbcX-independent assembly was detected by native PAGE with the red-type RbcS subunits from R. sphaeroides, G. sulfuraria, and P. tricornutum (Fig. 7A), but very little activity was measurable (Fig. 7B). Strikingly, a C-terminal fusion of Syn6301RbcS with the ␤EF region of RsRbcS (Syn6301RbcSEF) was able to produce active holoenzyme with similar final yield as Syn6301RbcS in the presence of RbcX (Fig.  7B). However, the apparent rate at which active enzyme was produced was substantially slower than that achieved with AnaCARbcX and Syn6301RbcS (Fig. 7C). This is probably due to a lower affinity of Syn6301RbcSEF for the folded RbcL subunits that tend to rebind to chaperonin. Combination of Syn6301RbcSEF and RbcX again resulted in rapid assembly (Fig. 7C). Thus, the C-terminal ␤-hairpin of red-type RbcS can substitute for the green-type assembly chaperone during in vitro reconstitution but with reduced assembly kinetics.

DISCUSSION
The RbcS subunits of green-and red-type Rubiscos are distinct in that the latter contain a C-terminal ␤-hairpin extension. The ␤-hairpins of 4 RbcS subunits form an eight-stranded ␤-barrel structure that protrudes into the central solvent channel of the RbcL 8 core. Here we have shown for the Rubisco of R. sphaeroides that this structure confers to the enzyme the ability to assemble the RbcL 8 S 8 complex independent of assembly chaperones such as RbcX. Although the ␤-hairpin is not essential for the assembly of red-type Rubisco and its enzymatic activity, it strongly enhances the rate of assembly and also increases the stability of the holoenzyme. Attachment of the C-terminal ␤-hairpin extension of RsRbcS to the green-type Syn6301RbcS allowed the assembly of Syn6301RbcL 8 SEF 8 without RbcX, albeit with slower kinetics.
In all organisms containing red-type Rubisco, the rbcS gene follows the rbcL gene in an operon (3,40). In eukaryotes with red-type Rubisco (non-green algae and diatoms), both genes are always plastid-encoded and linked (13). Such an organization would ensure that subunit stoichiometry is maintained independently of the overall rates of transcription and translation. In contrast, the genes encoding green-type rbcL and rbcS in cyanobacteria are generally in an operon with the assembly chaperone rbcX located in the intergenic space (41). In green algae and plants, the rbcL is chloroplast encoded, whereas the rbcS and rbcX genes are nuclear encoded (1). The operon linkage of red-type rbcL and rbcS genes, and thus their common regulation, may be a reflection of the function of RbcS in assembly. Given the finding that some red-type Rubiscos have higher CO 2 specificities than the green-type enzymes (30,42,43), it is of interest to ask why organisms with green-type Rubiscos do not utilize an RbcS-assisted assembly mechanism but instead have evolved specialized assembly chaperones (19 -25, 44). One possible reason is that the assembly factors, by providing independence of the rbcL-rbcS operon structure, may have facilitated the transfer of rbcS to the nuclear genome, an organization that presumably provided superior metabolic regulation and outweighed the advantage of RbcS-assisted assembly.
Notably, co-expression of RbcL and RbcS from the red algae G. sulfuraria or the diatom P. tricornutum in tobacco chloroplasts failed to result in assembled red-type Rubisco (13), although these proteins would be expected to be independent of the green-type assembly factors. One possible explanation for the failed assembly is that the green-type RbcS or an assembly factor(s) may have exerted a dominant negative effect. It is also possible that the general chloroplast chaperone systems may be unable to mediate the folding of the red-type Rubisco subunits, considering that the non-green algal plastome also encodes its own Hsp70 (DnaK) and chaperonin (GroEL) (45,46). The exact mechanism by which the ␤-hairpins mediate assembly is not yet clear. Our reconstitution experiments did not provide evidence that red-type RbcS first forms a stable tetramer platform on to which RbcL dimers assemble. Neither do the red-type RbcL subunits assemble independent of RbcS to stable octameric core complexes, in contrast to green-type cyanobacterial RbcL (20,21). Our data indicate that upon GroEL/ ES-mediated folding, the RbcL subunits form a range of soluble oligomeric states which are in dynamic equilibrium (Fig. 1B). Addition of RbcS rapidly shifts this equilibrium toward holoenzyme formation: The RbcS ␤-barrel engages in extensive hydrogen bond and salt bridge contacts with the RbcL subunits (Fig. 3C). Furthermore, the ␤EF hairpin of one RbcS (residues 106 -129) stabilizes the conserved ␤AB loop (residues 40 -45) in the adjacent RbcS subunit ( Fig. 3A and C), which in turn packs against the other RbcL subunit of the RbcL dimer. Thus, the RbcS ␤-barrel stabilizes the assembly at the level of antiparallel RbcL dimers and holds the dimers together. The RbcL dimer can be stabilized to some extent by sugar substrate binding, allowing assembly to occur with C-terminally truncated RbcS (Fig. 4D). However, only formation of the ␤-barrel drives rapid holoenzyme assembly (Fig. 5B). Our findings should facilitate future efforts to increase the yield of crop plants by expressing catalytically superior forms of red-type Rubisco in chloroplasts (6,13).