Chaperonin Cofactors, Cpn10 and Cpn20, of Green Algae and Plants Function as Hetero-oligomeric Ring Complexes*♦

Background: The chloroplast chaperonin system is encoded by multiple genes. Results: The chaperonin cofactors of the green alga C. reinhardtii and the plant A. thaliana form hetero-oligomeric ring complexes containing seven ∼10-kDa modules. Conclusion: The hetero-oligomeric cofactors were able to interact with chaperonin and assist protein folding. Significance: Formation of hetero-oligomers can explain the occurrence of multiple chaperonin cofactor genes in chloroplasts. The chloroplast chaperonin system of plants and green algae is a curiosity as both the chaperonin cage and its lid are encoded by multiple genes, in contrast to the single genes encoding the two components of the bacterial and mitochondrial systems. In the green alga Chlamydomonas reinhardtii (Cr), three genes encode chaperonin cofactors, with cpn10 encoding a single ∼10-kDa domain and cpn20 and cpn23 encoding tandem cpn10 domains. Here, we characterized the functional interaction of these proteins with the Escherichia coli chaperonin, GroEL, which normally cooperates with GroES, a heptamer of ∼10-kDa subunits. The C. reinhardtii cofactor proteins alone were all unable to assist GroEL-mediated refolding of bacterial ribulose-bisphosphate carboxylase/oxygenase but gained this ability when CrCpn20 and/or CrCpn23 was combined with CrCpn10. Native mass spectrometry indicated the formation of hetero-oligomeric species, consisting of seven ∼10-kDa domains. The cofactor “heptamers” interacted with GroEL and encapsulated substrate protein in a nucleotide-dependent manner. Different hetero-oligomer arrangements, generated by constructing cofactor concatamers, indicated a preferential heptamer configuration for the functional CrCpn10-CrCpn23 complex. Formation of heptamer Cpn10/Cpn20 hetero-oligomers was also observed with the Arabidopsis thaliana (At) cofactors, which functioned with the chloroplast chaperonin, AtCpn60α7β7. It appears that hetero-oligomer formation occurs more generally for chloroplast chaperonin cofactors, perhaps adapting the chaperonin system for the folding of specific client proteins.

The folding of many newly synthesized proteins is assisted by molecular chaperones, which function to prevent protein misfolding and aggregation and to maintain protein homeostasis (1)(2)(3). Among the best studied molecular chaperones are the chaperonins, a class of large cylindrical complexes consisting of two 7-9-membered rings that are stacked back to back. The Escherichia coli chaperonin GroEL (Hsp60 in mitochondria and Cpn60 in chloroplasts), consisting of two heptameric rings of ϳ57-kDa subunits, cooperates with the cofactor GroES (Hsp10 in mitochondria and Cpn10 in chloroplasts), a heptameric ring of ϳ10-kDa subunits (4). Each GroEL subunit is composed of three structurally distinct domains as follows: an apical, an intermediate, and an equatorial domain. The apical domains form the ring openings and expose hydrophobic amino acids toward the central cavity for the binding of nonnative protein substrate and for the association with GroES (5). The equatorial domains mediate most intra-and inter-ring contacts and contain the MgATP and potassium ion-binding sites. GroES binding causes the displacement of substrate protein into a now hydrophilic, cage-like space formed by one ring of GroEL and the dome-shaped GroES. The GroEL ATPase acts as a timer, regulating the binding and release of GroES and allowing the substrate to fold in the chaperonin cavity for ϳ10 s (at 25°C) (2,3,6,7).
In contrast to the E. coli chaperonin system, which is encoded by only two genes, groEL and groES, the situation in the plastids of green algae and plants is considerably more complex, with both Cpn60 and Cpn10 being encoded by more than one gene (8 -10). The Cpn60 proteins can be phylogenetically divided into two groups, Cpn60␣ and Cpn60␤, which share only about 50% sequence identity. In plants, Cpn60 has been shown to exist as a hetero-oligomeric complex of equimolar amounts of Cpn60␣ and Cpn60␤ (Cpn60␣ 7 ␤ 7 ). The subunit arrangement in the chloroplast chaperonin is still unknown (11). Both native (12) and recombinant (13)(14)(15) plant chloroplast chaperonins have been characterized biochemically and shown to assist protein folding in vitro. Recently, it has been shown that the low abundance chaperonin subunit Cpn60␤4 is specifically required to fold the chloroplast protein NdhH, suggesting that substrate specificity may exist in vivo based on the subunit composition of the chaperonin (16).
The plastid chaperonin cofactors are also diverse, encompassing the conventional ϳ10-kDa protein (Cpn10) as well as ϳ20-kDa cofactors consisting of tandem fusions of Cpn10 domains (17)(18)(19). The functional oligomeric state of these proteins is unclear. Although Arabidopsis thaliana Cpn10 (AtCpn10) 2 has low functional activity in vitro (19), AtCpn20 has been reported to form multiple oligomeric states, including tetramers (20,21), which cooperate with E. coli GroEL in protein refolding (19 -22). How exactly Cpn20 interacts with the heptameric chaperonin cylinder without encountering a symmetry mismatch is still unresolved. It has been suggested that Cpn20 might form a double-ring heptameric complex (23).
In this study, we have cloned and recombinantly expressed the three genes encoding chloroplast-targeted chaperonin cofactors in the unicellular green alga Chlamydomonas reinhardtii. The three proteins purified include CrCpn10 and two isoforms of Cpn20 (CrCpn20 and CrCpn23) (10). In vitro characterization of these cofactors using various biophysical and biochemical methods demonstrates that chaperonin-mediated protein folding requires the formation of mixed complexes consisting of CrCpn10-CrCpn20, CrCpn10-CrCpn23, or CrCpn10-CrCpn20-CrCpn23. Native mass spectrometry (MS) showed that these complexes form a heptameric arrangement of ϳ10-kDa domains. Similar results were obtained with the A. thaliana cofactor proteins, AtCpn10 and AtCpn20. Interestingly, although the C. reinhardtii Cpn20 cofactors must form hetero-oligomers with Cpn10, AtCpn20 can function as a homotetramer with one ϳ10-kDa domain being excluded from interacting with chaperonin. Different subunit configurations of the cofactors may correlate with a differential ability to assist the folding of specific client proteins by modifying the chemical environment of the chaperonin folding cage.
The C. reinhardtii cofactor concatamer constructs were generated by cloning the relevant ORFs described above into pHue vector separated by a flexible three amino acid linker consisting of glycine, serine, and alanine (see Fig. 6A). To generate the desired DNA sequences, the first subunit was inserted using SacII and BamHI restriction sites, the second subunit using BamHI and PstI sites, the third subunit using PstI and XhoI sites, and the fourth subunit using XhoI and HindIII sites. Internal restriction sites that interfered with this strategy were removed by introducing silent mutations using the QuikChange protocol.
The coding regions of AtCpn60␣2 (residues 47-586, at2g28000) and AtCpn60␤3 (residues 55-600, at1g55490) (9) lacking the N-terminal transit peptides were amplified from a cDNA library (gift from J. Soll). The genes of the isoforms AtCpn60␣2 and AtCpn60␤3 were chosen due to their high expression levels in planta (23). Both AtCpn60␣2 and AtCpn60␤3 genes, including the translational ATG start codon, were first cloned separately into pET11a vector (Novagen) using NdeI and BamHI restriction sites. The reverse primer of AtCpn60␣2 contained an additional NheI site between the stop codon and the BamHI site, allowing excision of the gene using SphI and NheI and insertion into SphI/XbaI-digested pET11a vector containing the translational cassette for AtCpn60␤3. This procedure resulted in a bicistronic expression vector, pETAtCpn60␣2␤3.
His 6 -ubiquitin fusion proteins were expressed in E. coli BL21(DE3) by inducing log-phase cells with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside for 20 h at 23°C. The harvested cells were resuspended in buffer A (50 mM Tris-HCl, pH 8.0, 300 mM NaCl) containing 10 mM imidazole and incubated with 0.3 mg ml Ϫ1 lysozyme on ice for 30 min. Cells were then disrupted by ultrasonication after addition of 1 mM phenylmethanesulfonyl fluoride (PMSF) and 1 unit ml Ϫ1 benzonase. After removal of cell debris by centrifugation, the supernatant was applied to Ni 2ϩ -nitrilotriacetic acid resin (Qiagen) equilibrated with buffer A, 10 mM imidazole. The resin was washed with buffer A, 25 mM imidazole before eluting the His 6 -ubiquitin fusion protein with buffer A, 200 mM imidazole. The ubiquitin moiety was cleaved overnight at 20°C using the deubiquitylating enzyme Usp2 (26). The protein was dialyzed against buffer B (20 mM Tris-HCl, pH 7.5, 20 mM NaCl, 1 mM EDTA), applied to a pre-equilibrated Mono Q 16/10 HR column (GE Healthcare), and eluted using a linear salt gradient to 0.5 M NaCl. Fractions containing the protein of interest were combined, concentrated, and applied to a Superdex200 10/300 size-exclusion column equilibrated in buffer C (20 mM Tris-HCl, pH 7.5, 50 mM NaCl). Finally, fractions containing pure protein were concentrated to ϳ10 mg ml Ϫ1 ; glycerol was added to 10%, and the protein was flash-frozen in liquid nitrogen and stored at Ϫ80°C. Cofactor proteins were quantified using Bradford assay with bovine serum albumin as the standard. The purified proteins were Ͼ95% pure as judged by SDS-PAGE (supplemental Figs. S1 and S3).
E. coli BL21(DE3) cells transformed with pETAtCpn60␣2␤3 were grown and induced as described above. The cells were resuspended in buffer D (50 mM Tris-HCl, pH 8.0, 20 mM NaCl, 1 mM EDTA), incubated with lysozyme (0.3 mg ml Ϫ1 ) on ice for 30 min, followed by ultrasonication as above after supplemen-tation with protease inhibitors (Complete TM , Roche Applied Science) and benzonase (1 unit ml Ϫ1 ). The cell debris was removed by centrifugation, and the supernatant was applied onto a Source 30Q column equilibrated with buffer D. After eluting with a linear salt gradient to 0.4 M NaCl, fractions containing the chaperonin were collected, and (NH 4 ) 2 SO 4 was added to a final concentration of 1 M. The resulting precipitate was removed by centrifugation, and the supernatant was applied to a HiLoad 16/10 phenyl-Sepharose HP column (GE Healthcare) equilibrated with 20 mM Tris-HCl, pH 7.5, 1 M (NH 4 ) 2 SO 4 . The proteins were eluted using a descending linear salt gradient from 650 to 0 mM (NH 4 ) 2 SO 4 , and the chaperonin containing peak fractions were combined, dialyzed against buffer B, and applied to a Mono Q 16/10 HR column as above. The column was eluted using a linear salt gradient to 0.5 M NaCl, and fractions containing the chaperonin protein were pooled, concentrated, and applied onto a Superdex 200 10/300 size-exclusion column equilibrated with buffer C. The fractions containing chaperonin protein were concentrated, glycerol was added to 10%, flash-frozen in liquid nitrogen, and stored at Ϫ80°C. Protein concentration was determined using Bradford assay as above.
Mass Spectrometry-The chaperonin cofactors were bufferexchanged into 100 or 200 mM ammonium acetate (Fractopur, Merck) using micro Bio-Spin 6 or 30 chromatography columns (Bio-Rad). Protein concentrations used are given in the figure legends. Native mass spectrometry (MS) analyses were performed in positive ion mode using an electrospray ionization quadrupole time-of-flight (ESI-TOF) instrument (Synapt TM HDMS TM system, Waters Corp., Manchester, UK) equipped with a Z-spray nano-ESI source. Capillaries were gold-plated 10 m nano-ESI pipettes purchased from Mascom (Bremen, Germany). Pressure in the first vacuum stage was increased to ϳ2.5 mbars for the cofactor proteins and values ranging from 4.5 to 5.2 mbars for the chaperonin complexes to cool the ions collisionally and maintain intact complexes (29 -31). Nano-electrospray voltages were optimized for generation and transmission of the macromolecular protein complexes; the needle voltage varied between 1200 and 1500 V, and the sample cone voltage varied between 100 and 200 V. All spectra were calibrated using an aqueous solution of cesium iodide (30 g liter Ϫ1 ).
ATPase Activity-ATP hydrolysis was measured spectrophotometrically at 25°C using a coupled enzymatic assay (32). Briefly, the measurements were carried out in a mixture of 50 mM BisTris/NaOH, pH 7.5, 100 mM KCl, 10 mM MgCl 2 , 2 mM phosphoenolpyruvate, 20 units ml Ϫ1 pyruvate kinase, 30 units ml Ϫ1 lactate dehydrogenase, and 0.5 mM ␤-nicotinamide adenine dinucleotide, reduced disodium salt hydrate (NADH). Chaperonin cofactors and 1 mM ATP were added, followed by an ϳ3-min preincubation to remove any ADP present. The reaction was started by the addition of 0.4 M GroEL or SR-EL, and the decrease in absorbance at 340 nm was monitored over 10 min.
Rubisco Refolding Assay-GroEL-assisted refolding of R. rubrum Rubisco (final concentration of 200 nM) was performed at 25°C as described previously (33,34) with minor modifications. Rubisco (20 M) was denatured in 20 mM MOPS/KOH, pH 7.5, 100 mM KCl, 5 mM MgCl 2 , 1 mM EDTA, 10 mM DTT, 6 M guanidinium hydrochloride for 1 h at 25°C before it was diluted 100-fold into ice-cold refolding buffer (20 mM MOPS/ KOH, pH 7.5, 100 mM KCl, 5 mM MgCl 2 , 5 mM DTT, 1 mg ml Ϫ1 BSA) containing 400 nM GroEL or AtCpn60␣ 7 ␤ 7 . The reaction mixture was incubated at 25°C for 5 min, and aggregates were removed by centrifugation (5 min at 16,000 ϫ g), followed by addition of chaperonin cofactors as indicated in the figure legends. Refolding of Rubisco was initiated by addition of 5 mM ATP. Aliquots were removed at indicated time points, and refolding reaction was stopped with glucose (100 mM) and hexokinase (1 unit l Ϫ1 ). Rubisco enzyme activity was determined after incubation at 25°C for 25 min essentially as described previously (35).
Proteinase K Protection Assay-Rubisco (20 M) was denatured as above and diluted 100-fold into ice-cold assay buffer (20 mM MOPS-KOH, pH 7.5, 5 mM Mg(OAc) 2 ) containing 0.8 M SR-EL, 10 mM KCl and incubated at 25°C for 5 min. Reactions were spun at 16,000 ϫ g for 5 min to remove aggregates. The supernatant was incubated for 5 min at 25°C with or without chaperonin cofactors in the presence of 4 mM AMP-PNP followed by addition of 2 g ml Ϫ1 proteinase K (PK) and incubated for the times indicated. Proteolysis was stopped by addition of 1 mM PMSF, and samples were analyzed by 12.5% SDS-PAGE and immunoblotting.
Analytical Gel Filtration-GroEL and cofactors at concentrations as indicated in the figure legends were incubated in 20 mM MOPS-KOH, pH 7.5, 50 mM KCl, 5 mM MgCl 2 , 2 mM ADP. After 10 min incubation at 25°C, 50 l of the reaction mixture was applied to a Superdex200 PC3.2/10 column (GE Healthcare) equilibrated with the same buffer (flow rate 0.5 ml min Ϫ1 ). Fractions were collected and analyzed by 16% SDS-PAGE and Coomassie staining.

Structural Features of Bacterial and Eukaryotic Chaperonin
Cofactors-A sequence alignment of the ϳ10-kDa domains of the chaperonin cofactors of E. coli (GroES), T4 phage (gp31), human mitochondria (Hsp10), A. thaliana (AtCpn10 and AtCpn20), and C. reinhardtii (CrCpn10, CrCpn20, and CrCpn23) reveals that AtCpn10 and the N-terminal ϳ10-kDa domain of CrCpn20 (CrCpn20N) are lacking the complete sequence corresponding to a ␤-hairpin that forms the roof of the dome-shaped oligomer ( Fig. 1) (4,36,37). In contrast, AtCpn20 and CrCpn23 possess this sequence in both of their ϳ10-kDa tandem domains. The interaction with the barrelshaped chaperonin is mediated by mobile loop segments of the cofactor (Fig. 1), which adopt a ␤-hairpin loop conformation when bound (38). Mutational analysis suggests that the balance between loop flexibility and formation of the ␤-hairpin loop determines the affinity of the interaction and that an intermediate affinity is required for function (39 -41). In the case of E. coli GroES, the double mutation S21T/T28P was shown to improve the interaction with the heterologous mitochondrial chaperonin Hsp60, presumably due to a reduction in loop flexibility and thus increased affinity (39). The CrCpn10 and CrCpn23 mobile loops all encode Thr and Pro at positions 21 and 28, respectively (numbering corresponding to GroES), predicting an increased affinity for interaction with the chaperonin (Fig. 1). In contrast, both the ϳ10-kDa tandem domains of CrCpn20 do not display this pattern, but instead the N-terminal domain encodes Ser-21 and Pro-28 (similar to AtCpn10) and the C-terminal domain encodes Thr-21 and Ala-28, suggesting greater loop flexibility and possibly lower affinity to chaperonin. Furthermore, CrCpn23 possesses a longer linker sequence between the tandem ϳ10-kDa domains than CrCpn20 or AtCpn20 (Fig. 1).
Algal Cofactor Proteins Are Unable to Functionally Interact with GroEL-CrCpn10, CrCpn20, and CrCpn23 (10) were recombinantly expressed in E. coli and purified in soluble form (supplemental Fig. S1). E. coli GroES was used as the control protein. Native MS was used to investigate the stoichiometry and quaternary structure of the algal chaperonin cofactors. The gentle nature of electrospray ionization (ESI) employed in this method of mass spectrometry allows the analysis of intact noncovalent structures, such as the chaperonin complexes (42)(43)(44)(45)(46)(47). As expected, E. coli GroES was heptameric (ϳ73 kDa) by native MS ( Fig. 2A and supplemental Table 1). In contrast, CrCpn10 and CrCpn23 existed predominantly as monomers (ϳ11 and ϳ23 kDa, respectively), whereas CrCpn20 exclusively formed a tetramer of ϳ82 kDa (Fig. 2, B-D, and supplemental Table 1). The respective oligomeric states of the cofactors were also observed by size-exclusion chromatography (Fig. 2E).
We next tested whether the algal cofactors were functional in assisting the E. coli chaperonin GroEL to refold the model substrate, Rubisco, of the bacterium R. rubrum. No Rubisco activity was detected in the presence of the algal cofactors, whereas the cognate cofactor GroES yielded ϳ80% of native Rubisco activity within 10 min of refolding (Fig. 3A). The lack of activity of the algal cofactors could reflect the inability of these proteins to  form a complex with GroEL. Gel filtration analysis in the presence of ADP, which favors formation of a stable GroEL-GroES complex (48), showed that GroES and CrCpn10 co-migrated with GroEL, whereas CrCpn20 and CrCpn23 eluted in their free forms (Fig. 3B). The binding of CrCpn10 to GroEL was surprising, given that CrCpn10 is monomeric (Fig. 2B), and no free CrCpn10 was detected (Fig. 3B), suggesting that this protein bound to GroEL in a substrate-like manner rather than as a cofactor. To investigate whether complex formation results in encapsulation of the substrate protein, we used the noncycling single-ring variant of GroEL, SR-EL (49). SR-EL contains mutations in four residues (R452E, E461A, S463A, and V464A) of the equatorial domain that mediate inter-ring contacts. Because of the absence of the second ring, GroES remains bound to SR-EL, and substrate is stably encapsulated. Binding of GroES to preformed complexes of SR-EL and denatured Rubisco resulted in ϳ95% protease protection of Rubisco (Fig. 3C) (27). In contrast, CrCpn10 afforded only a minimal ϳ5% protection from proteinase K digestion, consistent with the notion that CrCpn10 binds to GroEL as a substrate. In the presence of CrCpn20 and CrCpn23, the Rubisco protein was completely digested (Fig.  3C), consistent with these algal cofactors having low or no affinity for GroEL (Fig. 3B). The interaction with GroEL was further probed by measuring the effect of the cofactors on the ATPase activity of GroEL and SR-EL. As expected, GroES inhibited the GroEL ATPase rate by about 50% (50), although the ATPase activity of SR-EL was almost completely inhibited (Fig. 3D) (49). Note that SR-EL undergoes only one round of ATP hydrolysis upon GroES binding. CrCpn20 caused only a mild inhibition of the GroEL/SR-EL ATPase rate (ϳ20 -30%), indicating low binding affinity to chaperonin. Both CrCpn23 and CrCpn10 even caused a slight stimulation of the ATPase, which would be consistent with some substrate-like binding of these cofactors (Fig. 3D) (51,52). In summary, the algal cofactors were unable to function with the E. coli chaperonin GroEL in client protein folding, in contrast to the plant proteins studied previously (12, 18 -21).
Algal Chaperonin Cofactors Form Hetero-oligomers-The diversity of the C. reinhardtii chloroplast chaperonin cofactors and their atypical behavior was intriguing. We next explored the possibility that hetero-oligomeric complexes could result in functional cofactors. When CrCpn10 was mixed with CrCpn20 at an equimolar concentration, a prominent species of molecular mass 72,991.9 Ϯ 48.2 Da was detected by native MS (Fig.  4A). This mass corresponds to a complex of one CrCpn10 and three CrCpn20 molecules and is incompatible with other possible stoichiometries (Fig. 4A and supplemental Table 1). Similarly, a mixture of CrCpn10 and CrCpn23 yielded a species of 79,327.7 Ϯ 0.4 Da, corresponding to three CrCpn10 in complex with two CrCpn23 (Fig. 4B and supplemental Table 1). Notably, both of these mixed complexes contain seven ϳ10-kDa domains, equivalent to the functional heptameric E. coli GroES (see Figs. 2A and 3A). However, when CrCpn20 and CrCpn23 were mixed, three new species of 43,621.9 Ϯ 0.3, 66,688.9 Ϯ 45.5, and 87,363.5 Ϯ 29.2 Da were detected. These masses correspond to complexes with four, six, or eight ϳ10-kDa domains, consisting of a CrCpn20/CrCpn23 heterodimer, one CrCpn20 combined with two CrCpn23 subunits, or two CrCpn20 complexed with two CrCpn23 subunits, respectively ( Fig. 4C and supplemental Table 1). Surprisingly, when all three algal cofactors were mixed, only one prominent species of 75,449.4 Ϯ 60.9 Da was detected, which consisted of a heterooligomeric complex containing one CrCpn10, two CrCpn20, and one CrCpn23, again corresponding to a complex of seven ϳ10-kDa domains ( Fig. 4D and supplemental Table 1). The formation of hetero-oligomeric complexes containing seven ϳ10-kDa domains would present a solution to the symmetry problem encountered by the homo-oligomeric CrCpn20 or CrCpn23 complexes (see Fig. 2) (23). To test whether the ability to form hetero-oligomers is specific to the algal cofactors, we mixed CrCpn20 or CrCpn23 with GroES. Native MS clearly showed that no hetero-oligomers were formed, indicating the absence of subunit exchange with the GroES heptamer (supplemental Fig. S2).
CrCpn10-containing Hetero-oligomers Are Functional-We next investigated the ability of the various hetero-oligomer complexes to functionally interact with GroEL and assist folding of Rubisco. The hetero-oligomers containing CrCpn10 and CrCpn23 were able to support efficient refolding of Rubisco with final yields of ϳ70 -80%, albeit at ϳ3-5-fold slower rates than the cognate GroES (Fig. 5A). The combination of CrCpn10.20.23, forming a defined oligomeric species by MS (Fig. 4D), also resulted in highly efficient refolding but at an ϳ2-fold slower rate compared with CrCpn10.23. A final yield of only ϳ40% and an ϳ50-fold slower rate of folding was observed for CrCpn10.20 (Fig. 5A). All these cofactor combinations can form complexes with seven ϳ10-kDa domains (see Fig. 4). In contrast, the CrCpn20.23 hetero-oligomers, containing four, six, or eight ϳ10-kDa domains (see Fig. 4C), were essentially unable to assist refolding (Fig. 5A). Protease protection assays showed that the CrCpn10-containing complexes protected SR-EL-bound Rubisco from digestion with varying efficiency (Fig.  5B). CrCpn10.20 afforded significantly less protection than the other CrCpn10-containing hetero-oligomers, consistent with the lower folding yield and rate observed above (Fig. 5A). This suggests that the interaction of CrCpn10.20 with GroEL is more transient (Fig. 5B). Essentially, no protection was observed with CrCpn20.23 (Fig. 5B), consistent with the inability of this complex to support refolding (Fig. 5A). Furthermore, all heterooligomeric complexes containing CrCpn10 caused an ϳ50% inhibition of the GroEL ATPase rate, similar to GroES (Fig. 5C). However, unlike GroES, they were unable to inhibit the SR-EL ATPase rate completely, reflecting lower binding affinity for GroEL (Fig. 5C).
Cofactor Concatamers Suggest Preferred Subunit Arrangements-To eliminate uncertainties concerning the heterogeneity of molecular species in the hetero-oligomeric cofactor complexes, various combinations of physically linked subunits were constructed based on the previously reported strategy to produce active GroES as a concatamer of seven subunits (53). In the concatamers, subunits are connected from the C terminus of one subunit to the N terminus of the next subunit via different three amino acid linkers (GSG, AAG, GSS or ASS) (Fig. 6A). Except for the concatamer containing three CrCpn20 subunits, all other constructs were designed to contain seven ϳ10-kDa domains, corresponding to the subunit combinations as detected by native MS (see Fig. 4). The proteins were purified in soluble form (supplemental Fig. S1) and analyzed for their ability to functionally interact with GroEL. As expected, the concatamer conCrCpn20 -20-20, consisting of three fused CrCpn20 subunits, was completely inactive in assisting Rubisco refolding (Fig. 6B). Function in refolding not only required seven ϳ10-kDa domains per concatamer but was also dependent on the exact arrangement of the cofactor proteins. Strikingly, fully efficient refolding, both with respect to yield and rate, was observed with conCrCpn23-10-10-23-10, whereas conCrCpn10-10-23-23-10 allowed only residual refolding (Fig.  6B). In the active configuration, the three CrCpn10 subunits are arranged nonconsecutively, separating the two CrCpn23 subunits (Fig. 6A). Apparently, this configuration is not exclusively present when CrCpn10 and CrCpn23 are combined in unlinked form, as evidenced by the slower rate of folding observed under these conditions (see Fig. 5A). Conversely, the combination of unlinked CrCpn10, CrCpn20, and CrCpn23 was more efficient than the two concatamers of these proteins tested (see Figs. 5A and 6B). Thus, for these concatamers the specific covalent linkage either enforced less active configurations or restricted con- formational flexibility necessary for function. Alternatively, it cannot be excluded that the complex formed by unlinked CrCpn10, CrCpn20, and CrCpn23 changes composition in the presence of GroEL, allowing folding to proceed efficiently (Figs. 4D and 5A).
Formation of Plant Cofactor Hetero-oligomers-To test whether the GroES-type cofactor proteins in plants also form functional hetero-oligomers, we recombinantly expressed and purified the A. thaliana chaperonin consisting of AtCpn60␣ and AtCpn60␤ subunits (AtCpn60␣ 7 ␤ 7 ) and the cofactors (AtCpn10 and AtCpn20) (supplemental Fig. S3). Interestingly, the AtCpn20 preparation contained in addition to the fulllength protein a small amount of a lower molecular mass species of ϳ18.8 kDa (supplemental Fig. S3A, asterisk), representing C-terminally clipped AtCpn20 (AtCpn20 clp ), lacking the C-terminal 25 amino acid residues, as determined by MS. Native MS showed that AtCpn10, in addition to heptamers, also formed some octamers. Hexamers were observed as a product of dissociation upon gas phase collision in the MS, as indicated by their higher m/z ratio (Fig. 7A and supplemental Table 1). AtCpn20 formed two complexes of 64,232.1 Ϯ 10.5 and 83,010.3 Ϯ 9.7 Da, the former consisting of three full-length AtCpn20 subunits and the latter containing one AtCpn20 clp in addition ( Fig. 7B and supplemental Table 1). Notably, when AtCpn20 and AtCpn10 preparations were mixed, an additional hetero-oligomeric species consisting of three AtCpn20 and one AtCpn10 was detected (data not shown). To avoid the proteolytic processing of AtCpn20 to AtCpn20 clp , the residues Tyr-176 and Ala-177 at the cleavage site were mutated to Gly and Tyr, respectively, to generate AtCpn20 mut (supplemental Fig.  S3). These amino acid changes reflect the corresponding sequence of E. coli GroES (see Fig. 1). AtCpn20 mut exclusively formed trimers by native MS (Fig. 7C and supplemental Table  1), and when mixed with AtCpn10, an additional predominant species corresponding to three AtCpn20 mut and one AtCpn10 was observed (Fig. 7D and supplemental Table 1).
All the A. thaliana cofactors were functional in assisting Rubisco refolding with either AtCpn60␣ 7 ␤ 7 or GroEL ( Fig. 8A and supplemental Fig. S4A), consistent with their ability to inhibit the ATPase rate of GroEL and SR-EL (supplemental Fig.  S4B). Although AtCpn10 lacks the GroES-typical roof structure, this feature does not appear to be critical for the refolding of Rubisco, in contrast to observations with citrate synthase as the substrate (19). Similarly, the mixture of AtCpn20 mut and AtCpn10 (AtCpn20 mut 10) as well as the concatamer, con-CrCpn23-10-10-23-10, both able to form complexes with seven functional ϳ10-kDa domains, were highly active in terms of rate and yield of folding with both chaperonins (Fig. 8A and supplemental Fig. S4A). Surprisingly, AtCpn20 and AtCpn20 mut alone, each containing six ϳ10-kDa domains by MS (Fig. 7), were also functionally active. In the case of AtCpn20, this may be explained by the complex formation of three AtCpn20 subunits with one AtCpn20 clp subunit (Fig. 7B). AtCpn20 clp may be destabilized in its C-terminal ϳ10-kDa domain, and thus the functional complex would effectively contain seven ϳ10-kDa domains. To understand how AtCpn20 mut would functionally interact with the chaperonin, we performed native MS of AtCpn20 mut bound to AtCpn60␣ 7 ␤ 7 in the presence of ADP. The spectra showed oligomers at high mass to charge (m/z) ratio and monomers at low m/z (Fig. 8B, top panel, ① and ②, respectively). These monomers are not present in solution but rather are formed within the mass spectrometer, according to the general mechanism of protein complex dissociation in the gas phase in which an oligomer of n subunits fragments into complementary monomer and (n-1)-mer frag-ments (54). The main oligomeric species (n-mer) has a mass of 898,840 Ϯ 1311 Da, corresponding to four AtCpn20 mut molecules complexed with AtCpn60␣ 7 ␤ 7 and additional buffer components, likely ADP, magnesium, and potassium ions (Fig.  8B, middle panel and supplemental Table 1). This indicates that the AtCpn20 mut when bound to AtCpn60␣ 7 ␤ 7 , is a tetramer, equivalent to eight ϳ10-kDa domains.
Notably, among the low m/z peaks, we also detected a fragmented monomer of AtCpn20 mut that was generated by the collision energy in the native MS measurement and had a mass of 12,070.4 Ϯ 0.6 Da (AtCpn20 mut ⌬C) (Fig. 8B, bottom panel), corresponding to the removal of close to the complete C-terminal ϳ10-kDa domain. Such fragmentation of the polypeptide backbone during collisional activation of proteins in the gas phase is unusual, and it is indicative of a particularly labile covalent bond (55). Formation of AtCpn20 mut ⌬C was not observed with free AtCpn20 mut (data not shown) but only in the presence of the chaperonin, suggesting that one domain of an AtCpn20 mut subunit is excluded from interacting with  AtCpn60␣ 7 ␤ 7 and is forced to adopt an alternative conformation, thereby rendering it susceptible to fragmentation in the MS measurement. Consistent with this interpretation, one of the species observed in the mass spectra is a complex consisting of AtCpn60␣ 7 ␤ 7 with three intact AtCpn20 mut subunits and one AtCpn20 mut ⌬C subunit (Fig. 8B, middle panel, and supplemental Table 1). The peaks at 9000 -12,000 m/z correspond to complexes containing single-ring AtCpn60␣ 7 ␤ 7 with varying numbers of AtCpn20 subunits. Thus, in contrast to the C. reinhardtii Cpn20 cofactors, which must form hetero-oligomers with Cpn10, AtCpn20 can function as a homotetramer with one ϳ10-kDa domain being excluded from interacting with chaperonin.

DISCUSSION
The structural arrangement of the chloroplast Cpn20 cofactors has been enigmatic. In particular, it has been difficult to reconcile the even number of the ϳ10-kDa domains on a Cpn20 oligomer with the uneven number of binding sites on the heptameric chaperonin rings. In this study, we took advantage of the power of native MS in characterizing the composition of protein assemblies (47, 56 -60). We report the surprising finding that the recombinant algal C. reinhardtii chaperonin cofactors, CrCpn20 and CrCpn23, form heterooligomers with CrCpn10, resulting in functional complexes with seven ϳ10-kDa GroES-type domains. Hence, it appears that for the algal system the symmetry problem is solved by hetero-oligomerization. Notably, the A. thaliana Cpn20 has been shown to functionally interact with chaperonin as a homotetramer (20), raising the question of how the symmetry problem is solved in this case. A possible explanation is provided by our finding that recombinant AtCpn20 contains three intact Cpn20 subunits and one subunit in which the C-terminal ϳ10-kDa domain is partially proteolytically cleaved. If this proteolytic cleavage is prevented by mutation, the ϳ10-kDa domain of one AtCpn20 subunit appears to be excluded from binding to chaperonin, rendering it sensitive to fragmentation by argon gas collision in the mass spectrometer. Alternatively, three AtCpn20 subunits form a hetero-oligomer with one AtCpn10, containing seven ϳ10-kDa domains as in the case of the algal cofactors. Because Cpn10 subunits alone are either inactive (CrCpn10) or tend to have low activity (AtCpn10) (19), we suggest their main role is to form functional hetero-oligomers with Cpn20 (or Cpn23) subunits. The existence of the cofactor hetero-oligomers in vivo remains to be confirmed by isolating and characterizing them from the corresponding chloroplast stroma.
The number of different combinations and arrangements of the C. reinhardtii cofactors that were found to be functional is intriguing, suggesting the possibility that different hetero-oligomers may favor processing of different chaperonin clients. This could be achieved by modulating the size and properties of the chaperonin cavity, as described for the bacteriophage T4 cofactor gp31 (23,37,61). Given the lack of the ␤-hairpin roof structure for the N-terminal ϳ10-kDa domain of CrCpn20, different arrangements would be expected to result in diverse cavity roof properties (Fig. 1), perhaps facilitating the folding of certain client proteins. Similarly, AtCpn10 is also lacking the roof structure. Furthermore, combining Cpn10 and Cpn20 modules would result in cofactors with different affinities for the apical domains of the chaperonin, given that the ϳ10-kDa domains of the Cpn10 and Cpn20 (or Cpn23) chloroplast cofactors vary in the properties of the flexible loop structures mediating the interaction. The CrCpn10 and CrCpn23 mobile loops all encode threonine and proline at positions 21 and 28, respectively (numbering corresponding to GroES). These residues, when present in the mobile loop segment of GroES, have been shown to increase the binding affinity for chaperonin by stabilizing the ␤-hairpin structure of the loop (38,40). Thus, increasing the number of CrCpn10 and CrCpn23 modules in the hetero-oligomer should correlate with higher binding affinity and increased ability to displace certain client proteins for encapsulation and efficient folding. In contrast, both ϳ10-kDa domains of CrCpn20 do not display this pattern, with the N-terminal ϳ10-kDa domain instead encoding Ser at position 21 and the C-terminal ϳ10-kDa domain encoding Ala at position 28 (Fig. 1). Thus, the presence of CrCpn20 in the heterooligomer would confer a lower binding affinity for chaperonin. The cofactor concatamers generated in this study will be useful to test the possible adaptation of cofactor combinations to specific client proteins.