A Multiprotein Bicarbonate Dehydration Complex Essential to Carboxysome Function in Cyanobacteria

Abstract

Carboxysomes are proteinaceous biochemical compartments that constitute the enzymatic “back end” of the cyanobacterial CO₂-concentrating mechanism. These protein-bound organelles catalyze HCO₃⁻ dehydration and photosynthetic CO₂ fixation. In Synechocystis sp. strain PCC6803 these reactions involve the β-class carbonic anhydrase (CA), CcaA, and Form 1B ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco). The surrounding shell is thought to be composed of proteins encoded by the ccmKLMN operon, although little is known about how structural and catalytic proteins integrate to form a functional carboxysome. Using biochemical activity assays and molecular approaches we have identified a catalytic, multiprotein HCO₃⁻ dehydration complex (BDC) associated with the protein shell of Synechocystis carboxysomes. The complex was minimally composed of a CcmM73 trimer, CcaA dimer, and CcmN. Larger native complexes also contained RbcL, RbcS, and two or three immunologically identified smaller forms of CcmM (62, 52, and 36 kDa). Yeast two-hybrid analyses indicated that the BDC was associated with the carboxysome shell through CcmM73-specific protein interactions with CcmK and CcmL. Protein interactions between CcmM73 and CcaA, CcmM73 and CcmN, or CcmM73 and itself required the N-terminal γ-CA-like domain of CcmM73. The specificity of the CcmM73-CcaA interaction provided both a mechanism to integrate CcaA into the fabric of the carboxysome shell and a means to recruit this enzyme to the BDC during carboxysome biogenesis. Functionally, CcaA was the catalytic core of the BDC. CcmM73 bound H¹⁴CO₃⁻ but was unable to catalyze HCO₃⁻ dehydration, suggesting that it may potentially regulate BDC activity.

The cyanobacteria are globally important contributors to the biogeochemical cycling of carbon and to primary productivity. The ability of this diverse group of photoautotrophs to efficiently assimilate CO₂ from the environment relies upon the concurrent activity of a biophysical CO₂-concentrating mechanism (CCM) (4, 11, 16). This unique evolutionary adaptation, in effect, enhances the catalytic output of ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) by increasing the CO₂ concentration near the enzyme's active sites. The CCM consists of two principal and multifaceted components: (i) light-dependent active transport systems that concentrate and retain Ci (CO₂ + HCO₃^-) in the cytosol and (ii) distinctive protein machines, called carboxysomes, where HCO₃^- dehydration and photosynthetic CO₂ fixation occur (5, 30).

Most of the cellular complement of Rubisco is found within carboxysomes, surrounded by a number of characteristic proteins that form a boundary shell (6, 25, 30). Phylogenetic analysis indicates that there are two distinct and mutually exclusive types of carboxysomes (alpha and beta) in cyanobacteria, characterized by the presence of either Form 1A or Form 1B Rubisco (3). The two types are morphologically similar but differ in that the putative shell proteins are encoded either by the cso operon (csoS123AB) or by the ccm operon (ccmKLMN), respectively (3, 7). Bicarbonate dehydration is catalyzed by a specific, carboxysome-localized carbonic anhydrase (CA) (2, 41). In some marine cyanobacteria, such as Prochlorococcus sp. with Form 1A Rubisco, the ɛ-class CA, CsoSCA (formerly CsoS3), likely fulfills this catalytic role (42), while in freshwater and some other marine strains with Form 1B Rubisco, the β-class CA, CcaA (also known as IcfA) catalyzes the formation of CO₂ (10, 28, 40, 43, 48). A role for the putative γ-class CA, CcmM, in carboxysomal HCO₃^- dehydration has not been established.

Our understanding of carboxysome structure-function relationships has been advanced through conceptual modeling (31, 32) and through the study of targeted and random mutations within ccm and other genes (16, 30). Mutants with defects in ccmK, ccmL, ccmM, or ccmN, for instance, express aberrantly shaped carboxysomes or lack them completely. These structural defects prevent normal carboxysome function and restrict photoautotrophic growth of mutant strains to high-CO₂-containing environments (9, 26, 29). Mutants completely lacking CcaA, however, express carboxysomes that are morphologically indistinguishable from wild type, but their catalytic ability is compromised, leading to the same high-CO₂-requiring phenotype (43).

The nature of the integration of structural and catalytic components within carboxysomes is unknown, but its elucidation will likely provide insight into the molecular basis of carboxysome function. To this end, we have recently shown that the protein shell of alpha-carboxysomes is not solely a structural entity but participates directly in catalysis through CsoSCA-mediated HCO₃^- dehydration (42). Further understanding of carboxysome architecture has come from high-resolution electron cryotomography studies (13, 35) and from X-ray crystal structure analysis of CcmK and CsoS1 (18, 47). These proteins are members of an evolutionarily related group of proteins generally implicated in the formation of a range of bacterial microcompartments. Structural analysis reveals that both CcmK and CsoS1 form hexameric units, which organize into higher-order, 1.8-nm-thick sheet-like assemblies reminiscent of carboxysome polyhedral surfaces. Central pores located in each hexameric unit may serve as portals for the inward flux of substrate and the outward flux of product (47). An X-ray crystal structure of CsoSCA has also been recently obtained (34), but bacterial two-hybrid analysis was unable to identify carboxysome protein partners that facilitated its integration into the shell (12).

To provide insight into the organization of catalytic components involved in HCO₃^- dehydration, we used yeast two-hybrid and affinity-precipitation techniques to identify interacting protein partners and complexes associated with the beta-carboxysomes of Synechocystis sp. strain PCC6803. The data revealed that CcaA, CcmM, and CcmN form a novel multiprotein HCO₃^- dehydration complex that associates with the carboxysome shell through protein interactions with CcmK and CcmL. Based on HCO₃^- binding studies, fractionation, and CA activity assays, we propose that this shell-localized complex controls a pathway for the entry of cytosolic HCO₃^- to the carboxysome interior, where it is subsequently dehydrated to CO₂ by CcaA and channeled to Rubisco for fixation.

MATERIALS AND METHODS

Strains and growth conditions.

Standing cultures of Synechocystis sp. strain PCC6803 (hereafter referred to as simply Synechocystis) were grown photoautotrophically at 30°C in BG 11 medium with continuous illumination supplied at a photosynthetic photon flux density of 20 μmol m⁻² s⁻¹ (39, 43). Transformed cultures of Escherichia coli strains were grown on Luria-Bertani medium supplemented with the appropriate antibiotic (100 μg of ampicillin or 50 μg of zeocin ml⁻¹) (33). The yeast strain Saccharomyces cerevisiae L40 was maintained as described in the Hybrid Hunter yeast two-hybrid kit (Invitrogen). Transformants harboring bait or prey plasmids were selected on YPAD medium containing 300 μg of zeocin ml⁻¹ or on defined medium lacking tryptophan and uracil, respectively. Transformants harboring both bait and prey plasmids were selected on defined medium supplemented with 300 μg of zeocin ml⁻¹ but lacking histidine, lysine, tryptophan, and uracil.

Expression and purification of recombinant proteins.

Genomic DNA from Synechocystis was extracted, as previously described (40), and served as a template for the PCR to generate recombinant forms of ccaA, ccmM, and ccmN. A plasmid encoding a T7 epitope-tagged form of CcaA was generated previously (39). In the present study, a second construct encoding CcaA without an epitope tag was generated for the in vitro binding experiments. An NdeI-XhoI DNA fragment containing the entire ccaA gene was amplified by PCR using the CcaAF and CcaAΔ0R primers (see Table S1 in the supplemental material). This fragment was subsequently ligated into the pET-21a expression vector (Novagen) at a site within the T7 tag coding sequence that prevented epitope expression. A BamHI-XhoI fragment was also amplified by using CcmMbaitF and CcmMpreyR primers and ligated into pET-21a, thereby producing a construct that would express a T7 tag-CcmM fusion protein. Similarly, a plasmid encoding a histidine-tagged form of CcmM was generated for the bicarbonate binding assay. In this construct, a fragment containing NdeI and XhoI restriction sites was amplified by using CcmMhisF and CcmMbaitR primers and subsequently ligated into the pET-15b vector (Novagen).

To generate an epitope-tagged protein that could be distinguished from those generated with the T7 tag epitope for the in vitro binding assay, the pFlag expression vector (Sigma, Canada) was used to produce Flag-tagged CcmN. A recombinant form of CcmN fused to the Flag tag epitope was generated by ligating the HindIII-XhoI fragment amplified with the CcmNflagF and CcmNflagR primers into the pFLAG expression vector (Sigma, Canada).

The constructs were electrotransformed into E. coli BL21(DE3) cells by using the Gene Pulser II system (Bio-Rad, Canada). All constructs were sequenced using an ABI 373A automated sequencer (Applied Biosciences) to confirm the orientation and reading frame of the cloned fragments. Standard molecular biology techniques were performed as described by Sambrook and Russell (33).

Induction of recombinant protein expression was carried out by incubating 1-liter cultures with 1 mM IPTG (isopropyl-β-d-thiogalactopyranoside) for 4 h at 30°C. Total protein extracts were prepared by sonicating cells suspended in T7 binding buffer (4.29 mM Na₂HPO₄, 1.47 mM KH₂PO₄, 2.7 mM KCl, 137 mM NaCl, 0.1% Tween 20, 0.002% sodium azide [pH 7.3]) or in His binding buffer (20 mM NaH₂PO₄, 0.5 M NaCl, 30 mM imidazole [pH 7.4]). The soluble protein fraction was collected after centrifugation (10,000 × g, 10 min, 4°C) for downstream applications.

For the affinity purification of T7 tagged protein, soluble protein extracts containing T7 tag-CcaA or T7 tag-CcmM were incubated at 4°C with T7 tag antibody coupled to agarose beads (Novagen) for 1 h. For the in vitro binding experiments, soluble protein extracts expressing CcaA without an epitope tag or Flag tag-CcmN were incubated at 4°C with the resin-bound T7 tag-CcmM for 1 h. The resin-bound protein complexes were pelleted at 500 × g for 5 min and washed four times with 40 ml of T7 binding buffer.

For the purification of histidine-tagged CcmM, soluble protein extract containing His-CcmM was incubated at 4°C with a Ni²⁺ affinity resin (GE Healthcare) for 1 h. The resin-bound CcmM was subsequently pelleted at 800 × g for 5 min and washed once with 20 ml of His binding buffer. To prepare the protein for H¹⁴CO₃^- binding experiments, the resin-bound CcmM was incubated for 1 h at 4°C in a buffer containing 50 mM EPPS [4-(2-hydroxyethyl)-1-piperazine propanesulfonic acid]-NaOH and 1 μM ZnSO₄. To lower the Ci concentration in this buffer, the buffer was sparged with N₂ gas for 2 h, followed by autoclaving the solution twice under liquid cycle.

Preparation of Synechocystis protein extracts.

Cell lysates of Synechocystis were initially prepared by passing a cell suspension through a chilled French pressure cell at 1.24 MPa. Carboxysome-enriched fractions were prepared from this extract using Percoll/Mg²⁺ (GE Healthcare) method as described previously (21, 28). Intact Percoll-bound carboxysomes were disrupted by freezing at −20°C for 30 min (42). The disrupted carboxysomes were thawed, resuspended in EM buffer (100 mM EPPS-NaOH [pH 8.0] and 20 mM MgSO₄), and centrifuged at 13,800 × g for 2 min.

For affinity-precipitation experiments, the soluble protein fraction from Synechocystis cell lysates was collected after centrifugation (12,000 × g, 15 min, 4°C). The extract was incubated at 4°C with resin-bound T7 Tag-CcaA, T7 Tag-CcmM, T7 Tag-CcmM NT, or T7 Tag-CcmM CT for 2 h with constant agitation. The resin-bound protein complexes were then pelleted and washed four times with 40 ml of EM buffer.

Immunoblot analysis.

Denaturing polyacrylamide gel electrophoresis (i.e., sodium dodecyl sulfate-polyacrylamide gel electrophoresis [SDS-PAGE]) and immunoblot analyses of recombinant and cyanobacterial proteins were carried out as described previously (40). Immunodetection of epitope-tagged proteins was performed using anti-T7 Tag monoclonal antibody conjugated to alkaline phosphatase (Invitrogen) and anti-Flag M2 monoclonal antibody (Sigma, Canada) at dilutions of 1:10,000 (vol/vol) and 1:400 (vol/vol), respectively. For detection of the anti-Flag M2 primary antibody, goat anti-mouse immunoglobulin G-conjugated to alkaline phosphatase (Bio-Rad) was used at a dilution of 1:2,000 (vol/vol).

Cyanobacterial proteins were detected using antisera directed against CcaA from Synechocystis sp. (40), CcmM from Synechococcus sp. strain PCC7942 (kindly provided by G. D. Price, Australian National University), RbcL from Synechococcus sp. strain PCC6301 (kindly provided by T. J. Andrews, Australian National University), and RbcS+RbcL (kindly provided by S. M. Whitney, Australian National University) at a 1:2,000 (vol/vol) dilution.

To quantify the amount of CcaA, CcmM, and RbcL associated with each carboxysome fraction, the blots were developed using an ECL-Plus Western blotting detection system (GE Healthcare) and scanned for fluorescence by using a Storm 840 optical scanner (Molecular Dynamics) according to the manufacturer's instruction. The resulting band intensities were analyzed by using ImageQuant analysis software.

Yeast two-hybrid analysis.

Carboxysome protein interactions were also examined by using the Hybrid Hunter Yeast two-hybrid system (Invitrogen). Bait constructs were generated which expressed LexA fusion proteins with CcaA (39), with the entire Synechocystis CcmM polypeptide and with truncated forms of CcmM consisting of the first 249 N-terminal amino acids (CcmM NT) or the 455 C-terminal residues (CcmM CT). The sequences encoding these polypeptides were amplified by PCR (see Table S1 in the supplemental material) and inserted into the EcoRI and XhoI sites immediately downstream of the LexA DNA-binding domain of the pHybLex/Zeo expression vector.

Prey constructs were generated by ligating the coding sequences of Synechocystis carboxysomal proteins into the pYESTrp2 prey expression vector. The BamHI-XhoI fragments containing ccmA, ccmL, ccmN, ccmK1, ccmK2, ccmK3, ccmK4, ccmM (intact and truncated forms), and ccmO were PCR amplified (see Table S1 in the supplemental material), and each was ligated into pYESTrp2, producing fusion proteins with the B42 transcriptional activation domain. Prey constructs expressing CcaA (intact and truncated forms), RbcL, and RbcS were generated previously (39).

Yeast transformations were carried out by using the S.c. EasyComp transformation kit (Invitrogen). Transformants capable of prototrophic growth on histidine-deficient medium were further assayed for the presence of β-galactosidase activity using the chromogenic substrate X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside).

Mass spectrometric determination of CA activity.

The presence of low levels of CA activity in bacterial cell lysates and carboxysome-enriched preparations was detected by using a sensitive mass spectrometric technique. The reversible hydration of CO₂ that is catalyzed by CA was directly measured as an enhancement in the rate of ¹³C¹⁶O¹⁶O (m/z = 45) formation from ¹³C¹⁸O¹⁸O (m/z = 49) (37, 42) by using a magnetic sector mass spectrometer equipped with a membrane inlet (model MM 14-80SC; VG Gas Analysis, United Kingdom), as described previously (44).

H¹⁴CO₃^- binding assay.

Resin-bound His₆-tagged CcmM (1.0 ml) was incubated with 0.9 ml of Ci-free buffer supplemented with 2.1 μmol of NaH¹⁴CO₃ for 16 to 18 h at 4°C. Aliquots of resin slurry (80 μl) were applied to a glass microanalysis filtration system (Millipore) containing 15 ml of Ci-free buffer and filtered under vacuum through a 0.65-μm-pore-size hydrophilic Durapore membrane (Millipore, Canada). Membranes were subsequently washed four times with 5 ml of Ci-free buffer. The radioactivity of ¹⁴C retained on the filter membrane was measured with a scintillation counter (Wallac model 1409).

RESULTS

Proteins putatively associated with Synechocystis carboxysomes include those encoded by the central ccm gene cluster, CcmK2 (Sll1028), CcmK1 (Sll1029), CcmL (Sll1030), CcmM (Sll1031), and CcmN (Sll1032) (Fig. 1A). CcmO (Slr0436), RbcS (Slr0012), and RbcL (Slr0009) (Form 1B) are encoded 1.24 to 1.3 Mbp downstream of the central ccm cluster, while CcaA (Slr1347) is a further 0.74 Mbp downstream. CcmA (Sll0934) is encoded 0.12 Mbp upstream of the cluster, followed by CcmK3 (Slr1838) and CcmK4 (Slr1839) a further 0.62 Mbp upstream (Fig. 1A) (see also http://bacteria.kazusa.or.jp/cyanobase/ and reference 15).

FIG. 1.

(A) Arrangement of genes encoding putative proteins associated with Synechocystis carboxysomes. Arrows indicate the direction of transcription. (B) Targeted yeast two-hybrid analysis of carboxysome protein interactions. Genes used for prey constructs are listed across the top in panel A. Proteins fused to bait constructs are listed on the left side of panel B. NT refers to the N-terminal 249 amino acids of CcmM, and CT refers to the C-terminal 455 amino acids of CcmM. S. cerevisiae (5 ml) cells harboring both bait and prey constructs were spotted onto nitrocellulose and lysed by immersion in liquid nitrogen for 30 s. Cell lysates were assayed for β-galactosidase activity by incubating the membrane in Z buffer supplemented with 1 mg of X-Gal ml⁻¹ for 30 min at 30°C. Positive reactions are indicated as dark spots.

CcaA interacts with the N-terminal region of CcmM.

To determine whether the 274-amino-acid (30.7 kDa) β-class CA, CcaA, associates specifically with other carboxysome components (Fig. 1A), we used a targeted yeast two-hybrid analysis to detect potential protein-protein interactions (Fig. 1B). In yeast transformants expressing recombinant CcaA as bait, significant β-galactosidase reporter activity was detected in cells coexpressing, as prey, the full-length 687-amino-acid (73.1 kDa) CcmM polypeptide (CcmM73), indicating a positive protein-protein interaction. A similar positive reaction was obtained when CcaA and CcmM were alternatively coexpressed as prey and bait. In yeast transformants coexpressing CcaA and the first 249 amino acids of the N-terminal region of CcmM73 (hereafter referred to as CcmM NT), β-galactosidase reporter activity was also readily detected. However, when the C-terminal 455 amino acids of the protein (hereafter referred to as CcmM CT) were expressed as prey, no detectable activity was found (Fig. 1B). Thus, specific elements within CcmM NT mediate the CcaA-CcmM73 interaction. Interestingly, the N-terminal region of CcmM73 that interacts with CcaA shares significant (60%) amino acid sequence similarity with Cam, a γ-class CA from the archaeon Methanosarcina thermophila (1, 19, 22, 41). The C-terminal region of Synechocystis CcmM contains four repeated stretches of about 85 amino acids that possess up to 30% sequence similarity to RbcS (41).

CcaA-CcaA self-interaction is required for CcaA-CcmM interactions.

Like CcmM73, CcaA has two distinct and asymmetric domains. The catalytic region of CcaA is associated with the N-terminal 197 amino acids and has strong amino acid sequence similarity with plant and other bacterial β-CAs (39, 41). The C-terminal region features an unusual 75-amino-acid extension that is only found in putative carboxysomal β-CAs. To evaluate the involvement of the CcaA C-terminal region in mediating interaction with CcmM73, C-terminal truncated forms of CcaA (CcaAΔ0 to CcaAΔ127) were coexpressed with CcmM73 as bait (Fig. 2). As indicated by β-galactosidase activity, species up to and including CcaAΔ60 were capable of interacting with CcmM73. This interaction ceased after the truncation of 70 or more residues. We have previously shown that the truncation of 63 or more C-terminal amino acids led to the loss of CcaA-CcaA self-interaction (dimerization) (Fig. 2) and to the complete loss of CcaA catalytic activity (39). Since SDS-PAGE and Western blot analysis revealed the presence of a stable and abundant supply of the truncated form of CcaA in all of the strains tested, the inability to form catalytically active enzyme was attributed to an inability to assemble the CcaA dimer (39). Thus, the simultaneous loss of CcaA catalytic activity, CcaA-CcaA self-interaction and CcaA-CcmM interaction after the truncation of the 10-amino-acid segment between CcaAΔ60 and CcaAΔ70 indicates that the CcaA dimer itself is the minimal unit necessary for CcaA-CcmM73 interaction.

FIG. 2.

Determination of CcaA-CcaAΔCT and CcmM-CcaAΔCT interactions. The ability of full-length CcaA or CcmM, as baits, to interact with C-terminal truncated forms of CcaA (CcaAΔ0 to CcaAΔ127), as prey, was determined by using a targeted yeast two-hybrid analysis. Lysates of cells coexpressing bait and prey constructs were assayed for β-galactosidase activity as described in Fig. 1.

CcmM73, a focal point for carboxysome-protein interaction.

When CcmM73 was expressed as bait and the other carboxysome components were individually expressed as prey, significant levels of β-galactosidase activity were observed in cells expressing CcmL, CcmN, and the K1, K2, and K4 variants of CcmK (Fig. 1B). CcmM73 also interacts with itself when the protein was coexpressed as both bait and prey, suggesting that, like CcaA, CcmM73 exists in an oligomeric state in vivo. Surprisingly, no apparent interactions were detected between CcmM73 and either of the Rubisco subunits despite the conserved RbcS-like domains contained in CcmM.

To further characterize the interacting regions of CcmM73, two-hybrid analysis was performed using either the 249-amino-acid CcmM NT construct or the 455-amino-acid CcmM CT construct as bait (Fig. 1B). In cells expressing CcmM NT and each of the other carboxysome proteins as prey, pairwise interactions were detected between CcmM NT and CcaA, CcmM NT and CcmN, CcmM NT and CcmM73, and CcmM NT and itself, but not between CcmM NT and CcmM CT. Interestingly, CcmM NT failed to interact with CcmL or any of the variants of CcmK, reflecting a loss of protein interaction and suggesting a possible requirement for the C-terminal region of CcmM73. When the analysis was carried out with CcmM CT as bait, however, protein interactions were not restored, suggesting that either the entire CcmM73 polypeptide is required and/or that CcmM73 oligomers are required to facilitate the interactions with these carboxysomal shell proteins. The lack of interaction between CcmM CT and each of CcaA, CcmM NT, and CcmN indicates that all of these proteins specifically interact with CcmM73 through its N-terminal γ-CA-like domain.

In vitro protein binding assays indicate that CcaA-CcmM73-CcmN form a complex.

To investigate the significance of protein interactions detected by the two-hybrid analysis, in vitro binding experiments were devised using recombinant CcaA, T7 tag-CcmM73, and Flag tag-CcmN polypeptides. The T7 tag-CcmM73 was initially affinity purified using anti-T7 tag antibody conjugated to agarose resin. The washed resin with bound T7 tag-CcmM73 was then incubated with a soluble protein extract from E. coli cells expressing either recombinant CcaA or Flag tag-CcmN. The agarose resin was subsequently centrifuged and washed several times to remove unbound and loosely bound proteins. Association of recombinant CcaA or Flag tag-CcmN with resin-bound CcmM73 was determined by using immunoblot analysis (Fig. 3). When the blot was developed using anti-T7 tag antibody (Fig. 3, lanes 1 and 3), a 75-kDa polypeptide corresponding to the entire recombinant T7 tag-CcmM73 protein was detected, demonstrating the successful expression and purification of this protein. When the resin-bound material was further probed with anti-CcaA antibody (Fig. 3, lane 2), a prominent 30.7-kDa recombinant CcaA polypeptide was detected, indicating that CcmM73 was able to interact with and specifically pull-down this β-CA from the highly complex E. coli protein extract. A minor 30.1-kDa protein was also detected that may be a degradation product of recombinant CcaA. Similarly, immunoblot analysis revealed that CcmM73 was also able to interact with and precipitate CcmN from E. coli protein extracts (Fig. 3, lane 4). In experiments in which both CcaA- and CcmN-containing extracts were incubated either simultaneously or sequentially with resin-bound T7 tag-CcmM73, immunoblots revealed that CcmM73 was able to pull down both proteins together, presumably in the same complex. Furthermore, only proteins corresponding in size to T7 tag-CcmM73, CcaA, and Flag tag-CcmN were evident after SDS-PAGE and Coomassie staining (results not shown). In contrast, both CcaA and CcmN were absent from immunoblots when the extracts were incubated with agarose resin conjugated to the T7 tag alone (not shown). Collectively, these results are consistent with the yeast two-hybrid analysis and provide additional evidence that complexes involving CcmM73-CcaA, CcmM73-CcmN, and CcaA-CcmM73-CcmN form through specific protein-protein interactions.

FIG. 3.

In vitro binding of recombinant carboxysome proteins to resin-bound T7 tag-CcmM73. Soluble E. coli protein extracts containing overexpressed, recombinant Synechocystis CcaA or Flag tag-CcmN were prepared and incubated with T7 tag-CcmM73 immobilized on agarose resin. Immunodetection of the epitope-tagged CcmM73 and CcmN was performed using T7 tag- and Flag tag-specific monoclonal antibodies, respectively. Detection of CcaA lacking an epitope tag was performed using polyclonal antibodies directed against CcaA.

Native carboxysomal protein subcomplexes.

To determine whether CcaA and CcmM73 interact in vivo, affinity precipitation experiments were devised to extract native protein complexes from Synechocystis cell lysate containing intact and fragmented carboxysomes. Resin-bound T7 tag-CcaA, T7 tag-CcmM73, T7 tag-CcmM NT, or T7 tag-CcmM CT was used individually to probe the Synechocystis lysate. Proteins of interest that subsequently bound to the probe were identified by immunoblot analysis (Fig. 4). The presence of the appropriate resin-bound probe was verified by using an antibody that recognizes T7 tag. When the T7 tag alone was used to probe Synechocystis lysate, immunoblots using antibodies directed against CcaA, RbcL, RbcS+RbcL, and CcmM were clear, indicating that the epitope tag and resin alone did not retain the proteins of interest (Fig. 4).

FIG. 4.

Affinity precipitation of carboxysome proteins from Synechocystis cell lysates. Resin-bound T7 tag, T7 tag-CcaA, T7 tag-CcmM73, T7 tag-CcmM NT, and T7 tag-CcmM CT were incubated with Synechocystis soluble protein for 1 h at 4°C. Specific carboxysomal proteins forming complexes with the probes were detected using α-CcaA, α-RbcL, α-RbcL+RbcS, and α-CcmM58 antibodies. Resin-bound T7 tag was used as a control to detect nonspecific binding. For comparison, a protein fraction enriched in carboxysomes (CBX) was also subjected to immunoblot analysis.

When T7 tag-CcaA was used to probe the lysate, CcaA-specific antibodies detected the presence of an immunoreactive protein band broadly centered about 32 kDa (Fig. 4). Further analysis of the band (see Fig. S1 in the supplemental material) revealed the presence of both the recombinant 33.1-kDa T7 tag-CcaA and the 30.7-kDa native form of CcaA, suggesting that resin-bound T7 tag-CcaA could integrate into native carboxysome subcomplex fragments. Also associated with the probe was a small but consistently present amount of RbcL. However, RbcS was not usually detected. When an antibody generated against CcmM58 from Synechococcus sp. strain PCC7942 was used, multiple immunoreactive Synechocystis proteins with diffuse banding patterns were observed with apparent molecular masses centered on73 kDa (CcmM73), 62 kDa (CcmM62), 52 kDa (CcmM52), and 36 kDa (CcmM36) (Fig. 4; see also Fig. S1 in the supplemental material). The largest of these, CcmM73, which corresponds to the full-length 687-amino-acid Synechocystis CcmM protein, appears to be the least abundant and is difficult to resolve using our standard protein loading condition. It was consistently found, however, when double the normal amount of protein was loaded (see Fig. S1 in the supplemental material). The smaller immunoreactive proteins were also consistently retrieved from the lysate with the resin-bound T7 tag-CcaA probe in spite of the use of various protease inhibitor cocktails to reduced protein degradation.

When resin-bound T7 tag-CcmM73 or T7 tag-CcmM NT was used to probe the Synechocystis lysate, CcaA-specific antibodies detected the presence of the 30.7-kDa native form of CcaA associated with both tagged polypeptides (Fig. 4). Rubisco antibodies revealed that both RbcL and RbcS were also associated with the T7 tag-CcmM73-CcaA complex but not with the T7 tag-CcmM NT-CcaA complex. In contrast, the T7 tag-CcmM CT probe did not form a complex with CcaA, but Rubisco antibodies identified a complex of RbcL and RbcS associated with the T7 tag-CcmM CT. Thus, in native carboxysome subcomplexes, CcaA associated specifically with the N-terminal region of CcmM73, while RbcL/RbcS associated with the C-terminal region. It therefore seems likely that RbcL and/or RbcS associates with the C-terminal region of CcmM73 in the T7 tag-CcmM73-CcaA-RbcL-RbcS noted above.

In addition to CcaA, RbcL, and RbcS, the T7 tag-CcmM73 complex also contained native CcmM73, represented as a doublet on the gel with T7 tag-CcmM73, CcmM62, CcmM52, and CcmM27. The smaller immunoreactive forms of CcmM did not appear to associate with the CcmM NT-CcaA complex, suggesting that these species lacked the N-terminal γ-CA-like domain. CcmM52 was identified as a prominent constituent of protein subcomplexes when T7 tag-CcaA, T7 tag-CcmM73, and T7 tag-CcmM CT were used as probes. CcmM62 was observed as a faint band on the gels, suggesting that it, like Ccm73, is present in lower abundance than CcmM52 (Fig. 4 and 5).

FIG. 5.

Carboxysomal CA activity is mediated by CcaA and confined to the shell. (A) Carboxysome-enriched preparations from Synechocystis were separated by SDS-PAGE and visualized after Coomassie blue staining. Molecular mass markers are indicated on the left in kilodaltons. (B) Proteins from the gel representing the intact carboxysome fraction (intact cbx) were transferred onto nitrocellulose, and the blots were probed with CcmM-, RbcL-, or CcaA-specific polyclonal antibodies, revealing the presence of 73-kDa CcmM73, diffuse 48- to 52-kDa CcmM52, 52-kDa RbcL, and 31-kDa CcaA. Intact carboxysomes were also disrupted by freeze-thaw treatment and subsequently separated into pelletable (p) and soluble (s) fractions after centrifugation (13,000 × g, 2 min). Immunoblot analysis indicated that CcmM73 and CcaA occurred only in the pelletable fraction, while RbcL and CcmM52 were present in both fractions. (C) CA activity associated with intact carboxysomes and with the pelletable and soluble fractions was determined by using the ¹⁸O exchange assay (37). The rate of ¹³C¹⁶O¹⁶O (m/z = 45) formation from ¹³C¹⁸O¹⁸O (m/z = 49) substrate was monitored over time by mass spectrometry before and after the addition of the various fractions (extract). The uncatalyzed (uncat) reaction with buffer alone is depicted by the dotted line. (D) ¹⁸O exchange activity in extracts from E. coli cells extracts overexpressing either T7 tag-CcmM73 or T7 tag-CcaA. Expression of recombinant protein was induced at 30°C for 4 h by 1 mM IPTG. Protein expression was verified by immunoblot analysis using T7 tag monoclonal antibody.

Analysis of Synechocystis carboxysomes.

For comparison with pull-down experiments, preparations enriched with intact carboxysome were also subjected to immunoblot analysis (Fig. 4). The immunoblots indicated the presence of native CcaA, RbcL, and RbcS. An unknown 48-kDa polypeptide was detected by the RbcL+RbcS antibodies, along with unresolved high-molecular-mass components. Also identified in intact carboxysomes were CcmM73, CcmM62, CcmM52, and CcmM36, suggesting that these polypeptides are native components. The CcmM52 band was typically diffuse and spanned a range of about 48 to 52 kDa, perhaps indicating the presence of more than one polypeptide. CcmM27 was not observed in the carboxysome fraction. Of interest is that CcmM36 was additionally found in subcomplexes only when T7 tag-CcaA was used as a probe, perhaps suggesting a preferential relationship. Since CcmM62, CcmM52, and CcmM36 do not interact directly with CcaA, we assume an association with RbcL/RbcS.

CcaA and CcmM73 are carboxysome shell proteins.

Protein fractions enriched with intact carboxysomes were prepared from Synechocystis lysate by Percoll/Mg²⁺ precipitation for SDS-PAGE (Fig. 5A) and immunological analysis (Fig. 5B). To separate the shell proteins from the carboxysome content, intact carboxysomes were subsequently ruptured by a freeze-thaw treatment (6, 42). About 55% of the RbcL was released into the supernatant after the treatment, indicating that a significant number of carboxysomes were disrupted. Immunoblot analysis further revealed that more than 95% of the CcaA and CcmM73 were localized to the pelletable fraction, indicating that these proteins were associated with the carboxysome shell. RbcL was found in both the pelletable and soluble fractions, suggesting the existence of two discrete Rubisco populations, one bound to the shell and the other loosely associated within the carboxysome interior (6). CcmM52 was also found in both the soluble and pelletable fractions. Due to the method used to process the blots we were unable to obtain information regarding CcmM36.

When these fractions were assayed for CA activity (Fig. 5C) by the ¹⁸O exchange method (37, 39, 42), intact carboxysomes accelerated the rate of ¹³CO₂ (m/z = 45) formation well above the level of the uncatalyzed control. Ethoxyzolamide, a classic inhibitor of CA, eliminated the enhancement (not shown). CA activity comparable to the level found in intact carboxysomes was also found associated with the pelletable protein fraction, but activity was not detected in the soluble protein fraction. Thus, the data indicate that CA activity was associated exclusively with the carboxysome shell.

To investigate whether this catalytic activity was attributable to CcaA, CcmM73, or both, recombinant forms of the β- and γ-CAs were expressed and assayed for ¹⁸O exchange activity (Fig. 5D). The rate of ¹³CO₂ formation in the presence of recombinant CcmM73 was nearly identical to the uncatalyzed buffer control. In contrast, recombinant CcaA accelerated the rate of ¹³CO₂ formation 10-fold above the uncatalyzed level (Fig. 5D). These findings indicated that CcmM73 did not catalyze the dehydration of HCO₃^- under our experimental conditions, despite the presence of the highly conserved γ-CA domain and, therefore, it does not have the same catalytic function as CcaA. Similarly, no CA activity was detected in carboxysomes isolated from a Synechocystis mutant completely deficient in CcaA expression (43), indicating that CcmM73 is also catalytically incompetent within the native carboxysome environment. Collectively, these results support the idea that CcaA is the primary CA that catalyzes the dehydration of HCO₃^- within Synechocystis carboxysomes.

CcmM73 binds H¹⁴CO₃^-.

A search of the conserved domain database (http://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi?) (24) at the National Center for Biotechnology Information (NCBI) revealed the presence of a number of putative, overlapping domain structures within the N-terminal region of Synechocystis CcmM73 (accession no. NP_440093). These included the expected carbonic anhydrases/acetyltransferases/isoleucine patch superfamily domain, PaaY/COG0663, and a single-stranded, left-hand beta-helix structure, cd00208. Comparative structural modeling of the N-terminal 200 amino acids of CcmM73 using a variety of homology and fold recognition programs such as 3D-PSSM/PHYRE (http://www.sbg.bio.ic.ac.uk/phyre/) (17) or Swiss Model (http://swissmodel.expasy.org//SWISS-MODEL.html) (36) also illustrated the potential of the N-terminal region of CcmM73 to form a left-hand beta-helix structure that is characteristic of Cam (19, 38) and other trimeric proteins (not shown). Structural and amino acid sequence comparisons indicated that all of the conserved Zn-coordinating histidine residues and other residues identified as important for the binding of HCO₃^- and the formation of the Cam trimer are conserved in CcmM73 (41). These residues mapped to similar spatial locations when computer-generated models of the N-terminal region of CcmM73 were superimposed (http://wishart.biology.ualberta.ca/SuperPose/) (23) onto the experimentally determined crystal structure of Cam (14, 19), with root-mean-square deviations ranging from 1.4 to 4 Å. Thus, the bioinformatics analysis encourages the view that CcmM73 trimers may retain HCO₃^-/CO₂ binding capability, although the experimental evidence indicates that CcmM73 is apparently unable to catalyze HCO₃^- dehydration (Fig. 5).

When incubated with ¹⁴C-labeled HCO₃^-, resin-bound, recombinant His-CcmM73 bound 2.41 × 10⁻⁵ nmol of HCO₃^- ng of protein⁻¹ ± 2.65 × 10⁻⁶ (n = 11). This level of binding was 50% greater than that observed for the resin or resin-bound His tag alone, on an equal volume basis. Potentially, HCO₃^- binding could be mediated nonspecifically by reaction with free amino groups to form stable carbamates. However, when CcmM73 was preincubated with 1 mM KOCN, which itself preferentially forms carbamates with free amino groups (45), no statistical difference in the level of HCO₃^- binding was observed on a protein basis. Whether it is HCO₃^- or CO₂ that binds to CcmM73 has not yet been resolved.

DISCUSSION

A structure-function model for HCO₃^- dehydration in Synechocystis carboxysomes.

Figure 6 shows a working model that summarizes the protein interactions and associations discovered in the present study. The data indicate that CcmM73 is the focal point of a multiprotein bicarbonate dehydration complex (BDC), the major constituents of which are CcmM73, CcaA, and CcmN. This complex, through CcmM73, also interacts with the major bacterial microcompartment shell proteins CcmK and CcmL. Interactions between CcmM73 and CcmK and between CcmM73 and CcmL required both the N-terminal and the C-terminal domains of the CcmM73 polypeptide (Fig. 1), indicating that it is likely in direct contact with the surface of the carboxysome shell. CcaA-CcmM73-CcmN complexes were formed in vitro when individual components were combined from complex cell lysates (Fig. 3), suggesting that native complexes self-assemble. The ability of these proteins to form a complex in vitro in the presence of numerous other proteins testifies to the specificity of the association. CcmM73-CcaA complexes were also identified as components of protein subcomplexes pulled from Synechocystis cell lysate enriched with carboxysomes and carboxysome fragments, bolstering the notion that this complex is an integral part of native carboxysome structure. Virtually all of the immunodetectable CcmM73 and CcaA, as well as CA catalytic activity, remained in the pelletable fraction after freeze-thaw experiments (Fig. 4), demonstrating that the BDC is associated exclusively with the carboxysome shell. This conclusion is also supported by the observation that the soluble fraction, which contained ca. 55% of the immunoreactive Rubisco and CcmM52, contained no other detectable CA activity.

FIG. 6.

Model of the BDC and surrounding proteins associated with the boundary shell of Synechocystis sp. strain PCC6803 carboxysomes. CcmM73 (M73) is depicted as a trimer, interacting with dimeric CcaA (A) and CcmN (N). This complex, through CcmM73, also interacts with the major carboxysome shell proteins CcmK (K) and CcmL (L). The potential role of CcmM73 in regulating access of cytosolic HCO₃^- to the active site of CcaA is also shown. The Rubisco (Form 1B) complement that is carboxysome shell localized may interact with the C-terminal region of CcmM73, as well as with CcmM52 (M52) and CcmM36 (M36), but the exact nature of these interactions is uncertain. CcmM62 has been omitted. The mechanism of RuBP uptake (dotted line) is unknown.

The interactions between CcmM73 and CcmN and between CcmM73 and CcaA within the BDC are specifically mediated within the N-terminal γ-CA-like region of CcmM73 (Fig. 1). Our working model (Fig. 6) suggests that CcmM73 is a trimer in vivo. Experimentally, this arrangement is supported by yeast two-hybrid analysis, which shows that CcmM73 monomers self-interact specifically via the N-terminal region (Fig. 1). The presence of an N-terminal left-hand parallel β-helix structure comprised of hexapeptide repeats characteristic of and required for the assembly of many trimeric enzymes, including Cam (19), provides further support for this hypothesis. Interestingly, a similar left-hand parallel β-helix is also predicted to occur between amino acids 19 and 90 of CcmN (accession no. NP_440092 [NCBI conserved domain database, cd00208]). This is the same region where weak amino acid sequence similarity exists between CcmN and the N terminus of CcmM73 (30% identity over 110 amino acids). Interaction between the β-helices of the two proteins may provide a mechanism to account for CcmN-CcmM73 interaction and assembly into the BDC, as well as for CcmM73-CcmM73 interactions. We have yet to determine whether CcmN is capable of self-interaction and have elected to illustrate CcmN as a monomer associated with CcmM73 (Fig. 6).

An overlapping PaaY/COG0663 domain (residues 22 to 126) characteristic of the carbonic anhydrases/acetyltransferases/isoleucine patch superfamily of proteins was also identified for CcmN through searches of the conserved domain database at NCBI. Unlike CcmM73, putative residues required for Zn coordination within the domain are completely missing in CcmN. Therefore, it seems unlikely that CcmN plays a direct catalytic role in HCO₃^- dehydration, but it may be required for the structural integrity of the BDC. The close physical proximity of CcmN to CcmM73 and CcaA in the carboxysome, as implied by the yeast two-hybrid and in vitro binding results, bears some similarity to the general structural organization of the alpha-carboxysomal CA, CsoSCA, from Halothiobacillus neapolitanus. This CA has three domains: an N-terminal domain comprised of four α-helices, a central Zn-containing catalytic domain, and a C-terminal domain that structurally resembles the catalytic domain (34). Like CcmN, the C-terminal domain of CsoSCA lacks the residues necessary for Zn coordination and is itself unlikely to be catalytically active. The functional significance of the physical clustering in both alpha- and beta-carboxysomes of active CA protein with inactive CA-related proteins or domains is unclear but may provide a potential to regulate CA activity by providing a binding site for HCO₃^- (8). The universal distribution of CcmN and CcmM73 in beta-carboxysome and of CsoSCA in alpha-carboxysomes suggests that this strategy may be a common theme exploited by evolution to facilitate carboxysome function.

Our model incorporates a CcaA dimer as the minimal catalytic unit of the BDC, located on the inner surface of the carboxysome shell (Fig. 6). This location is consistent with yeast two-hybrid, fractionation, and activity assay data. This arrangement ensures that CcaA is not in direct contact with the cytosolic HCO₃^- pool since such contact has been demonstrated to “short-circuit” the CCM in Synechococcus sp. strain PCC7942 (27). Thus, structural segregation of CcaA from the cytosolic HCO₃^- pool is likely a common feature of beta-carboxysomes. CcaA is encoded 1.8 Mbp downstream of the central ccm gene cluster (Fig. 1). Thus, coordinated expression of these genes is anticipated during carboxysome biogenesis. The specificity of the CcaA-CcmM73 interaction provides a means by which CcaA can be recruited to and integrated within the BDC during carboxysome assembly.

Multiple forms of CcmM.

Immunoblots of Synechocystis lysates enriched with carboxysomes revealed the presence of multiple forms of CcmM. Using immunological and biochemical techniques, Price and coworkers have also identified two different forms of CcmM in their studies of Synechococcus sp. strain PCC7942 carboxysomes, with calculated masses of 58 and 35 kDa, respectively (21, 30). The larger of the two corresponds to the predicted entire CcmM protein that includes an N-terminal γ-CA-like domain and three C-terminal RbcS-like domains. The smaller and more abundant form (CcmM35) appears to be composed of the RbcS-like repeats alone. Whether the different forms of CcmM arose through specific proteolytic cleavage of full-length CcmM or came about from the use of alternative, in-frame translation start codons (46), as has been advocated by Price and coworkers, has yet to be resolved experimentally. In-frame methionine residues encoded by an ATG are located at residues 1, 19, 153, 154, 174, 201, and 241 within the predicted 687 amino acids of Synechocystis CcmM73. Potential ribosome-binding sites were located on the transcript upstream of residues encoding M1, M19, M201, and M241, yielding potential polypeptides with molecular masses of 73.1, 71.1, 51.8, and 47.6 kDa, respectively. In-frame GTGs encoding V99 and V354 also have putative upstream ribosome-binding sites that could potentially initiate translation of a 62.6-kDa and a 35.5-kDa protein, respectively. The hypothetical 62.6-, 51.8-, and 35.5-kDa proteins may therefore correspond to CcmM62, CcmM52, and CcmM36, respectively. Also, the 47.6-kDa hypothetical protein may contribute to the broad 48- to 52-kDa band observed on Western blots (Fig. 4) that we have called CcmM52. Lindahl and Florencio (20) have previously identified a Synechocystis CcmM by mass spectrometry as a protein migrating on SDS-PAGE immediately below the 52-kDa RbcL polypeptide, providing independent evidence for the existence of CcmM52. In our experiments, the smaller immunoreactive forms of Synechocystis CcmM did not associate with the CcmM NT-CcaA complex (Fig. 4), suggesting that they are all comprised of elements from the C-terminal domain alone. Consistent with their masses, CcmM62 and CcmM52 would be comprised of four RbcS-like repeats, while CcmM36 would be comprised of three. Similarly, CcmM73 could arise from translation initiation from codons for M1 and/or M19, yielding proteins with both the N-terminal domain and the four RbcS-like repeats. Precedence for the use of an internal ribosome entry site by Synechocystis to create an alternative isoform of a protein from a single transcript has recently been established by Thomas et al. (46) in their study of ferredoxin:NADP oxidoreductase.

The role of the smaller CcmM isoforms is not clear. Potentially, their RbcS-like domains may contribute to packing Rubisco holoenzyme within the carboxysome since protein complexes containing CcmM52 or CcmM36 also contained RbcL and RbcS. In addition, CcmM36 has thus far only been found in complexes that also contained CcaA and may therefore participate in the interaction between the BDC and Rubisco (Fig. 6). However, yeast two-hybrid analyses did not provide independent evidence for CcmM73-RbcL or CcmM73-RbcS interactions. Thus, binary protein interactions alone are insufficient to form stable CcmM73-RbcL or CcmM73-RbcS complex. The occurrence of multiple forms of CcmM in both Synechococcus and Synechocystis carboxysomes suggests that this may be a common characteristic of beta-carboxysomes with implications for overall composition and structure.

Functional role of CcmM73.

CcmM73 contributes directly to the structure of the BDC by serving as the focal point for assembly and as the vehicle that anchors it to the carboxysome shell. These structural roles are mediated by the N-terminal γ-CA-like domain. The conservation of the γ-CA-like domain and the ability of CcmM73 to bind HCO₃^-/CO₂ suggest that it may also play a direct role in the biochemical activity and/or regulation of the BDC (Fig. 6). Potentially, CcmM73 may act as a shuttle protein binding cytosolic HCO₃^- and channeling it in a measured way to the nearby active site of CcaA. Alternatively, the binding or release of HCO₃^-/CO₂ by CcmM73 may sterically control the access of cytosolic HCO₃^- to the CcaA active site, thereby controlling the rate of dehydration. These possibilities are currently under investigation.

Multiple forms of the BDC in carboxysomes.

Potential CcaA-CcmM-CcmN complexes similar to the Synechocystis complex are predicted from genome sequencing studies of Cyanothece sp. strain CCY0110; Crocosphaera watsonii WH8501; Nostoc punctiforme PCC73102; and Synechococcus sp. strains PCC6301, PCC7002, and PCC7942. Genes encoding CcmM and CcmN are also found in Anabaena variabilis ATCC 29413, Gloeobacter violaceus PCC7421, Nostoc sp. (Anabaena) strain PCC7120, Trichodesmium erythraeum IMS101, and Thermosynechococcus elongatus BP-1. However, CcaA homologs are missing from the genomes of the latter cyanobacteria. Carboxysomes isolated from Nostoc sp. (Anabaena) strain PCC7120 have CA activity (G. Espie, B. Long, and D. Price, unpublished results), suggesting that a second form of the BDC exists in which CcmM is catalytically active or that they contain an undiscovered phylogenetic lineage of CA. CsoSCA is present and active in the shell of alpha-carboxysomes from the chemolithoautotroph H. neapolitanus and may form the catalytic core of a third form of a BDC. Recombinant CsoSCA from the cyanobacterium Prochlorococcus marinus MED4 and from Synechococcus sp. strain WH8120 have also been shown to be catalytically active enzymes and provide support for this idea (41, 42). All other chemolithoautotrophs and cyanobacteria that express Form 1A Rubisco in carboxysomes also encode CsoSCA homologs. However, the nature of the protein partners that interact with CsoSCA at the shell surface is yet to be resolved (12). In Synechococcus sp. strain WH8102, a putative β-CA, SYNW2467, coseparated with the carboxysome fraction and may represent a second carboxysomal component of the BDC (12). Other evidence, however, suggests that SYNW2467 is a periplasmic CA and is present as a membrane-associated contaminant of the carboxysomal preparations (12, 41). Regardless, it seems likely that a subsurface, catalytic complex that vectorially directs HCO₃^- from the cytosol and concentrates CO₂ in the carboxysome interior is a general feature of carboxysome biochemistry. The apparent evolution of multiple forms of the BDC highlights the importance of this complex to the effective operation of the CCM and to the central process of autotrophic CO₂ fixation in cyanobacteria.