CO2 Fixation Kinetics of Halothiobacillus neapolitanus Mutant Carboxysomes Lacking Carbonic Anhydrase Suggest the Shell Acts as a Diffusional Barrier for CO2*

The widely accepted models for the role of carboxysomes in the carbon-concentrating mechanism of autotrophic bacteria predict the carboxysomal carbonic anhydrase to be a crucial component. The enzyme is thought to dehydrate abundant cytosolic bicarbonate and provide ribulose 1.5-bisphosphate carboxylase/oxygenase (RubisCO) sequestered within the carboxysome with sufficiently high concentrations of its substrate, CO2, to permit its efficient fixation onto ribulose 1,5-bisphosphate. In this study, structure and function of carboxysomes purified from wild type Halothiobacillus neapolitanus and from a high CO2-requiring mutant that is devoid of carboxysomal carbonic anhydrase were compared. The kinetic constants for the carbon fixation reaction confirmed the importance of a functional carboxysomal carbonic anhydrase for efficient catalysis by RubisCO. Furthermore, comparisons of the reaction in intact and broken microcompartments and by purified carboxysomal RubisCO implicated the protein shell of the microcompartment as impeding diffusion of CO2 into and out of the carboxysome interior.

Many bacteria form intracellular polyhedral microcompartments that act as microbial organelles. They sequester metabolically important enzymes and enhance or regulate their activity. Several molecular mechanisms have been postulated for the way in which microcompartments function (1)(2)(3)(4); all of these assume that the bounding proteinaceous shell of the microcompartment acts as a selective diffusion barrier, effectively separating the enclosed enzymes and the reactions they catalyze from the cell cytoplasm (5). Two families of small shell proteins appear to be the only common genetic and structural elements among the bacterial microcompartments formed by such metabolically diverse prokaryotes as heterotrophs and autotrophs (1,6). By far the best studied microcompartments are the carboxysomes of cyanobacteria and chemolithoautotrophic bacteria, which contain ribulose 1,5-bisphosphate carbox-ylase/oxygenase (RubisCO), 2 the CO 2 -fixing enzyme of the Calvin-Bensen-Bassham cycle, and are thought to act as the terminal element of the carbon-concentrating mechanism in these bacteria (reviewed in Refs. 3,5). Genetic and physiological studies suggest that in ␤-cyanobacteria (7) inorganic carbon is actively transported into the cell interior and concentrated in the cytoplasm as bicarbonate, which must first be converted to CO 2 by a carboxysome-associated carbonic anhydrase before it can be fixed by RubisCO (8). Direct biochemical studies of ␤-carboxysomes from cyanobacteria have been hampered by difficulties with the purification of intact particles (1,9). Carboxysomes of the ␣-type found in chemolithoautotrophs and ␣-cyanobacteria (7) and exemplified by those of the sulfur bacterium Halothiobacillus neapolitanus, have been purified to homogeneity and shown to be composed of eight major proteins (10). The CbbL and CbbS polypeptides, which account for 60 -70% of the total carboxysome protein (3,10), represent the large and small subunit, respectively, of RubisCO. The balance of the carboxysome mass is accounted for by polypeptides that are associated with the thin protein boundary membrane, also known as shell. The major structural shell elements are CsoS1A, CsoS1B, and CsoS1C, highly conserved small proteins that are the products of duplicated genes. Recent crystallographic studies of CsoS1A and its cyanobacterial homologs, CcmK2 and CcmK4, have shown that the proteins form hexamers. These pack into sheets and likely form the facets of the icosahedral carboxysome (11,12). Each of the hexamers contains a pore that might be involved in the controlled passage of substrates and products into and out of the carboxysome. The other conserved microcompartment constituents, OrfA, OrfB, and CcmL, crystallize as pentamers and may be located on the vertices of the icosahedral particle (13). The CsoS2A and CsoS2B polypeptides, which together account for nearly 12% of the carboxysomal protein by weight (3), are products of a single gene in the cso operon (14). Their high calculated pI value has led to the speculation that they may act as rudimentary transporters for the phosphorylated RubisCO substrates and products (15). The remaining shell protein CsoSCA, which is encoded by the csoS3 gene and only a minor component by weight (15), is the carbonic anhydrase of ␣-carboxysomes (16). Its lack of sequence homology with known carbonic anhydrases led to its original categorization into a novel class (⑀) (16). Subsequent crystallographic analysis, however, revealed that CsoSCA constitutes a subclass of ␤-carbonic anhydrases that appears to be populated exclusively by ␣-carboxysomal enzymes (17). The CsoSCA protein is very tightly associated with the shell and is released only under conditions that result in disassembly of the shell. The likely functional form of the carboxysomal carbonic anhydrase is a dimer that is present in only 40 copies/carboxysome. Despite this low number, kinetic characterization of recombinant CsoSCA (rCsoSCA) revealed that the enzyme could supply the ϳ270 holoenzyme molecules of carboxysomal RubisCO with a sufficient amount of CO 2 to maintain saturating substrate conditions (15).
In an effort to further assess the biological role of the carboxysomal CA, we have taken advantage of the homogeneous carboxysome preparations that can be obtained from H. neapolitanus to directly compare the biochemical properties of wild type microcompartments with those isolated from a csoS3 insertion mutant that is devoid of carboxysomal carbonic anhydrase activity and requires elevated CO 2 levels to grow at normal rates.

EXPERIMENTAL PROCEDURES
Cultures-Wild type H. neapolitanus (ATCC 23641) and mutant csoS3::Km cultures were maintained in a bioreactor (INFORS-HT, Switzerland) at a dilution rate of 0.08 h Ϫ1 and a pH of 6.4, as previously described (16). The csoS3::Km mutant was grown in air supplemented with 5% CO 2 . Growth of wild type and mutant culture over time was monitored by determining the optical density of batch cultures at 600 nm.
Construction of H. neapolitanus csoS3::Km-Plasmid pTnEP2.2 (18), which contains a 2.2-kb EcoRI/PstI fragment encompassing nucleotides 4192-6364 of the cso operon (Gen-Bank TM accession number AF038430) in the vector pT7T3␣13, was digested with ClaI. The kanamycin resistance cassette was excised from the plasmid pUC4K (19,20) by digestion with BamHI. The overhangs generated by the restriction enzymes were filled in with Escherichia coli PolI Klenow fragment, and the cassette was blunt end-ligated into the linearized pTnEP2.2 to generate pHncsoS3::Km. The wild type csoS3 gene of H. neapolitanus was replaced by the mutant version via homologous recombination following electroporation with pHncsoS3::Km (21). Transformants were allowed to recover in medium without antibiotic for 24 h before being transferred to selective medium containing 50 g⅐ml Ϫ1 kanamycin. All of the cultures were maintained in a 5% CO 2 -enriched atmosphere. Gene replacement in H. neapolitanus csoS3::Km was verified by sequence determination of genomic DNA and by Southern hybridization.
Complementation of H. neapolitanus csoS3::Km-A single colony of the H. neapolitanus csoS3::Km mutant was used to inoculate a liquid culture that was grown at 30°C in medium containing 50 g/ml kanamycin. Cells from this culture were transformed with pcsoS3ProEx plasmid DNA (17) by electroporation (21). This expression construct contains a mutated csoS3 gene that yields enzymatically active rCsoSCA with a His residue instead of Tyr 92 . Following a 24-h recovery period at elevated CO 2 levels, cells that had regained the ability to grow at ambient CO 2 levels were selected in medium without antibiotics.
Genomic DNA was isolated as described previously (22) and digested with EcoRI or ClaI. Following gel electrophoresis the DNA fragments were blotted onto a nylon support via alkaline transfer (23). The blot was probed with 32 P-labeled csoS3 DNA at 68°C overnight. Replacement of csoS3::Km by an uninterrupted copy of the gene in H. neapolitanus was confirmed by sequencing of genomic DNA.
Isolation of Carboxysomes from Wild Type and Mutant H. neapolitanus-Wild type and csoS3::Km mutant cells from 6 -8-liter cultures were used to purify carboxysomes as described before (16). Because elevated CO 2 levels are known to suppress carboxysome formation (24), it was necessary to switch csoS3::Km cells to ambient CO 2 levels once the culture had reached the desired density during growth in CO 2 -supplemented air. All of the protein samples were desalted using Micro-Spin 6 columns (Bio-Rad) equilibrated in 50 mM Bicine, pH 8.0.
Electron Microscopy-Formvar/carbon-coated copper grids (EMS, Fort Washington, PA) were floated on a solution containing purified carboxysomes for 3 min. The samples were allowed to dry in air for 10 s before being stained with 1% (w/v) ammonium molybdate in 10 mM Tris-HCl, pH 8.0, for 40 s. The grids were air-dried for 10 min prior to being observed under a Zeiss EM-109 transmission electron microscope. Photographs of carboxysomes were taken by exposing Kodak EM film 4489 for 2 s.
SDS-PAGE and Immunoblot Analysis-Carboxysome proteins were separated by electrophoresis in pre-cast Criterion SDS, 4 -20% polyacrylamide gradient gels (Bio-Rad) and stained with Gelcode Blue (Pierce). Protein blots were probed with polyclonal antiserum raised against H. neapolitanus rCsoSCA (16) as primary antibody and with goat anti-rabbit, alkaline phosphatase-conjugated IgG as secondary antibody. The blots were developed with One-Step NBT-BCIP reagent (Pierce). Densitometric analysis was performed using the Quantity One program supplied with the VersaDoc imaging system (model 4000 MP; Bio-Rad).
Radiometric RubisCO Assays-Desalted protein samples were activated in 50 mM Bicine-NaOH, pH 8.0, 10 mM MgCl 2 , and 10 mM Na 14 HCO 3 (0.1 Ci⅐mol Ϫ1 ) for 10 min. Each 0.5-ml assay contained 50 mM Bicine-NaOH, pH 8.0, 20 mM MgCl 2 , 0.5 mM ribulose 1,5-bisphosphate (RubP) (Fluka brand; Sigma), and the appropriate amount of NaHCO 3 to give a final concentration range from 1.45 to 80.2 mM (this includes the contribution of the bicarbonate used to activate RubisCO). In parallel assays, the NaHCO 3 concentration was kept at 60 mM, and the RubP concentration was varied from 12.5 to 800 M. CO 2 fixation was initiated by adding 10 l of activated protein sample. To determine the relative carbon assimilation rates in intact wild type and mutant cells, 25-ml aliquots of bioreactorgrown wild type and mutant cultures were centrifuged at 12,000 ϫ g for 10 min, and the pelleted cells were resuspended in 1 ml of sterile medium. To start the reaction, 20-l aliquots of the cell suspension were added to growth medium that contained radioactive bicarbonate at various concentrations (1.0 Ci⅐mol Ϫ1 ). The reactions were stopped at 1-min intervals by adding 100-l assay aliquots to 300 l of glacial acetic acid. The vials containing the acidified samples were heated to remove excess 14 CO 2 , cooled, and mixed with 4 ml of ScintiVerse II mixture (Fisher). Radioactivity was counted in a LS 6000SC scintillation counter (Beckman Coulter, Fullerton, CA). Quench and counting efficiency were determined from a standard curve of 14 C-labeled n-hexadecane in 0.3 ml of glacial acetic acid.
The initial velocities were calculated from linear regression lines plotted using Prism 4. The V max and K m values for CO 2 (K C ) and RubP were obtained by fitting the results to the Michaelis-Menten equation. Kinetic assays were performed in triplicate and repeated with a minimum of three independent preparations.
Protein Determination-Protein concentrations in carboxysome and RubisCO preparations were determined with the BCA assay (Pierce); those in intact cell samples were estimated with a modified Lowry assay (Pierce). Bovine serum albumin served as standard.

RESULTS
Generation of the csoS3::Km Mutant-The wild type csoS3 gene of H. neapolitanus, which encodes CsoSCA, was interrupted by insertion of a kanamycin resistance cassette between the ClaI sites in the csoS3 coding sequence to produce plasmid pHncsoS3::Km. Following electroporation of mid-exponential phase, H. neapolitanus cells with this construct and selection of kanamycinresistant transformants in 5% CO 2enriched air colonies with a high CO 2 -requiring phenotype were recovered that grew at close-to-wild type rates only at elevated CO 2 levels. Gene replacement was confirmed by Southern blotting and sequence analysis of genomic DNA (data not shown). A single high CO 2 -requiring colony was used to establish the working culture of csoS3::Km. Comparison of mutant and wild type growth rates demonstrated extremely slow growth of csoS3::Km at ambient CO 2 levels but nearly wild type rates when the culture was aerated with 5% CO 2 in air (Fig. 1). Immunoblot analysis of crude cell extracts with anti-rCsoSCA antibodies revealed a strong signal for the wild type but undetectable CsoSCA levels in the mutant (Fig. 2C).
Rescue of Wild Type Phenotype by Homologous Recombination-The sizes of DNA restriction fragments that hybridize to a csoS3 probe on genomic Southern blots were different between mutant and wild type H. neapolitanus. Insertion of the kanamycin resistance cassette   2 and 4). Lanes 1 and 2, stained polypeptides separated by SDS-PAGE; lanes 3 and 4, immunoblots probed with anti-CsoSCA antiserum. As determined by mass spectrometry, the polypeptide bands marked with asterisks are carboxysome protein aggregates that were not disrupted during electrophoresis.
between the two ClaI sites in the csoS3 gene of the mutant csoS3::Km yielded two diagnostic ClaI restriction fragments of ϳ8.3 and 1.1 kilobase pairs, respectively. ClaI-digested genomic DNA from wild type and from clones complemented with plasmid pcsoS3ProEx produced the two wild type size fragments of ϳ0.9 and 7.3 kilobase pairs, respectively, indicative of restored csoS3 ClaI sites in the transformants. Furthermore, the restored csoS3 gene, as expected, encoded a His instead of the wild type Tyr residue at position 92 of CsoSCA. This fortuitous mutation, which was introduced during subcloning of the H. neapolitanus csoS3 gene (17), has no effect on the carbonic anhydrase activity of rCsoSCA (15). This mutation was therefore instrumental in documenting that the wild type phenotype in the complemented H. neapolitanus clones resulted from replacement of the interrupted genomic csoS3 copy with a fully functional plasmid-borne gene, presumably by homologous recombination.
Isolation of Carboxysomes Lacking CsoSCA-H. neapolitanus csoS3::Km cells were grown in a chemostat under conditions identical to those used to cultivate wild type cells, except that the mutant culture was aerated with air supplemented with 5% CO 2 . Under these conditions, a dilution rate of 0.08 h Ϫ1 was maintained. Removal of the supplemental CO 2 resulted in rapid wash-out of the mutant cells and confirmed that a wild type growth rate was supported only at elevated CO 2 concentrations. Chemostat efflux was collected and aerated with 5% CO 2 in air until ϳ6 liters of culture had accumulated. Because the expression of high cellular carboxysome levels is known to be suppressed by high CO 2 concentrations (24), the mutant culture in the collection tank was switched to ambient CO 2 for 18 -24 h prior to cell harvest and subsequent carboxysome purification. Yields of mutant carboxysomes purified by the standard procedure (16) were comparable with those from wild type cells (ϳ10 mg of carboxysome protein/8 g of wet weight of cells). Furthermore, isolated csoS3::Km carboxysomes did not appear to be morphologically different from wild type microcompartments, as judged by electron microscopic analysis (Fig.  2, A and B). The mutant microcompartments were, however, consistently found to be ϳ15-20% larger in diameter than their wild type counterparts when analyzed by dynamic light scattering (data not shown).
The stopped flow changing indicator assays (25) detected no carbonic anhydrase activity in purified mutant carboxysomes at protein concentrations that yielded clear evidence of enzyme activity in wild type microcompartments (Fig. 3). Comparison of the polypeptide composition of both types of carboxysomes by denaturing SDS-PAGE revealed similar but not identical patterns. The most obvious difference was the absence of the CsoSCA band at ϳ60 kDa in mutant carboxysomes (Fig. 2D). Immunoblot analysis using polyclonal anti-rCsoSCA antibodies (16) confirmed the absence of detectable amounts of CsoSCA from the mutant microcompartments. Somewhat surprisingly, the molar ratio of CsoS2A to CsoS2B, which is close to 1:1 in wild type carboxysomes (3,10), was shifted to favor the larger form 2:1 in mutant carboxysomes (Fig. 2D).
Carbon Fixation by Mutant Cells-Rates of carbon fixation by bioreactor-grown wild type and mutant cells were deter-mined at various DIC concentrations. Although maximal fixation rates were comparable (0.12 Ϯ 0.01 and 0.13 Ϯ 0.01 mol⅐min Ϫ1 ⅐mg protein Ϫ1 for wild type and mutant, respectively), the concentration of dissolved inorganic carbon (bicarbonate, carbonate, and CO 2 ) required to produce the half-maximal rate (K DIC ) was more than three times higher for mutant (K DIC ϭ 3.9 Ϯ 1.0 mM) than for wild type cells (K DIC ϭ 1.1 Ϯ 0.2 mM) (Fig. 4). Moreover, the ratio of wild type to mutant CO 2 fixation rate is 3.3 Ϯ 1.0 at ambient CO 2 ([DIC] ϭ 26 M) and 1.5 Ϯ 0.3 at 5% CO 2 ([DIC] ϭ 3.3 mM) (26), consistent with the measured differences in growth rates (Fig. 1).
CO 2 Fixation Kinetics in Isolated Carboxysomes-To judge the effect of the shell-bound carbonic anhydrase on carboxysomal CO 2 fixation rates, the RubisCO activity of mutant and wild type carboxysomes was determined radiometrically (27) (Fig.  5). For these experiments we report K C as M CO 2 , calculated from the input HCO 3 Ϫ concentration at pH 8.0, to allow comparison with the true K m of free RubisCO and of the enzyme in broken carboxysomes. We refer to the concentration of HCO 3 Ϫ required to produce half-maximal velocity as K C to distinguish it from the true K MCO 2 , which can be measured for carboxysome-bound RubisCO only if the flux of CO 2 across the shell does not affect the accessibility of RubisCO to its substrate. CO 2 fixation rates at standard assay conditions (60 mM bicarbonate) and varying RubP concentrations yielded a V max for mutant carboxysomes of 1.1 Ϯ 0.1 mol⅐min Ϫ1 ⅐mg Ϫ1 protein, which is ϳ60% of the maximal rate observed for wild type particles (1.6 Ϯ 0.1 mol⅐min Ϫ1 ⅐mg Ϫ1 protein). The apparent K m for RubP, on the other hand, was essentially the same for mutant and wild type carboxysomes (174.7 Ϯ 16.1 and 146.2 Ϯ 5.6 M, respectively). Likewise, V max obtained at a fixed, saturating RubP concentration of 0.5 mM and increasing bicarbonate concentrations from 1.45 to 80.2 mM was consistently slightly lower for the mutant carboxysomes (1.3 Ϯ 0.1 mol⅐min Ϫ1 ⅐mg Ϫ1 protein) than for wild type particles (1.7 Ϯ 0.1 mol⅐min Ϫ1 ⅐mg Ϫ1 protein) ( Table 1). The K C for mutant carboxysomes (499.3 Ϯ 47.1 M), however, was nearly 3-fold higher than that measured for wild type microcompartments (177.1 Ϯ 16.3 M) ( Table 1).
To assess whether exogenously added carbonic anhydrase could functionally complement the missing carboxysome component, purified rCsoSCA was mixed with purified csoS3::Km carboxysomes at an activity level that was 5-fold higher than that measured for wild type carboxysomes (15). No significant effect on V max and K C values was observed (Table 1).
To further evaluate how the presence and location of CsoSCA within the carboxysome affect carboxysome function, the carboxysome shell was disrupted by a freeze/thaw treatment (15,16,28). Electron microscopic examination of the remaining structures revealed the presence of shell ghosts that retained their icosahedral shape to some degree but contained breaks and tears (3,28). As reported previously, some RubisCO was retained within the ruptured structures but greater than 70% was released into the supernatant, as estimated by differential centrifugation and SDS-PAGE (15). Disruption of the csoS3::Km carboxysomes resulted in an increase in V max (2.0 Ϯ 0.1 mol⅐min Ϫ1 ⅐mg Ϫ1 protein) compared with the value obtained for intact mutant particles (1.3 Ϯ 0.1 mol⅐min Ϫ1 ⅐mg Ϫ1 protein). The K C of broken mutant carboxysomes (167.3 Ϯ 13.2 M) was only slightly higher than that of broken wild type particles (126.7 Ϯ 10.8 M) and within the range of that calculated for intact wild type carboxysomes (177.1 Ϯ 16.3 M) (Fig. 5, A  and B). Control assays with disrupted wild type carboxysomes revealed a small increase in V max (2.0 Ϯ 0.1 mol⅐min Ϫ1 ⅐mg Ϫ1 protein) and a slight decrease in K C (126.7 Ϯ 10.8 M). RubisCO preparations derived from the supernatants of disrupted wild type and mutant carboxysomes, as expected, displayed no significant differences in their kinetic constants (Table 1 and Fig. 5C).

DISCUSSION
Numerous biochemical and genetic studies support the widely accepted notion that the carboxysome functions as the terminal component of a CO 2 -concentrating mechanism (1, 2). Its carbonic anhydrase is thought to rapidly convert the abundant cytosolic bicarbonate to CO 2 , the sole form of inorganic carbon that can be utilized by RubisCO. In the work reported here, knocking out CsoSCA, the carboxysomal carbonic anhydrase of H. neapolitanus, by insertion mutagenesis resulted in the csoS3::Km mutant that required elevated CO 2 for growth at wild type rates. The high CO 2 -requiring phenotype is common  for mutants of cyanobacteria and chemoautotrophs that are deficient in carboxysome components. Some of these mutants produce carboxysomes of normal appearance that are, however, obviously functionally compromised (8, 21, 29 -33). Maybe cells can assemble microcompartments from a subset of gene products, but to yield a functional carboxysome all proteins found in wild type particles are required. In agreement with this possible scenario is our recent report that in H. neapolitanus all genes for carboxysome proteins, including those that exist in multiple copies in the cso operon, are transcribed (34).
Although negatively stained mutant and wild type carboxysomes were indistinguishable by transmission electron microscopy, dynamic light scattering measurements consistently revealed csoS3::Km carboxysomes to be 15-20% larger than their wild type counterparts. This difference, although barely exceeding the error margins of dynamic light scattering, was reproducible for all carboxysome preparations; a smaller-thanwild type diameter of mutant carboxysomes was never observed. Because CsoSCA accounts for only 2-3% of carboxysomal protein, this difference in size may be attributed to changes in the carboxysome assembly pattern. Alternatively, a different polydispersity of the mutant carboxysome population caused by the absence of CsoSCA may have rendered the apparent diameter artificially high. The larger size of mutant carboxysomes may also be related to their CsoS2B to CsoS2A ratio, which was twice that of wild type particles, in which both CsoS2 variants are represented in approximately equimolar amounts (3,10).
Clearly, the carboxysomes of the csoS3::Km mutant were functional in that they supported CO 2 fixation by the sequestered RubisCO even in the absence of CsoSCA. However, the requirement of csoS3::Km cells for higher CO 2 levels to support efficient growth suggests that mutant carboxysomes were not able to provide the catalytic advantage RubisCO derives from wild type microcompartments. These results support the idea that CsoSCA is instrumental in providing sufficient levels of CO 2 to the sequestered RubisCO and are consistent with the existence of a carbon-concentrating mechanism in H. neapolitanus similar to that shown to operate in Thiomicrospira crunogena and in several cyanobacteria (5,35,36). The path by which the bicarbonate that is concentrated in the cytosol enters the carboxysome and the way in which the resulting CO 2 is distrib-uted to the RubisCO in the microcompartment core are not well understood. Previous quantitative models predict that to maintain effectively higher concentrations of CO 2 at the active sites of RubisCO, either the shell of the carboxysome prevents CO 2 escape, or the carbonic anhydrase is localized in the center of the microcompartment, and the surrounding densely packed RubisCO molecules slow CO 2 leakage (37,38). In H. neapolitanus and other ␣-carboxysomes, the carbonic anhydrase (CsoSCA) is tightly associated with the shell (15,16,39). This makes the model that posits a central location for the enzyme (37) unlikely, at least for ␣-carboxysomes. The results obtained in this study show that to reach half-maximal CO 2 fixation rates intact mutant carboxysomes lacking CsoSCA required nearly 3-fold higher concentrations of bicarbonate than wild type carboxysomes. Furthermore, exogenously added excess rCsoSCA did not affect this requirement for higher bicarbonate concentrations. To be effective in enhancing the carbon fixation rate of RubisCO, the carbonic anhydrase activity must reside on the inside of the microcompartment boundary. If the shell were freely permeable to CO 2 , exogenously added carbonic anhydrase should catalyze the formation of CO 2 rapidly enough to allow diffusion into the particle. This should at least partially overcome the requirement for high inorganic carbon levels to support wild type carbon fixation rates. Our results are consistent with the idea that the protein shell limits the diffusion of CO 2 into and out of the carboxysome.
If the carboxysomal shell limits access of CO 2 to the carboxysome interior, its disruption should relieve any diffusional limits. Indeed, the observed kinetic constants of RubisCO were restored to near wild type values in mutant carboxysomes whose shell integrity had been compromised. These results are consistent with the previously reported increase in in situ carbonic anhydrase activity upon disruption of the carboxysome shell (15), which was also interpreted as relieving the apparent limitation on diffusion of HCO 3 Ϫ and/or CO 2 across the carboxysome boundary.
Comparison of K C and V max values for intact and disrupted wild type carboxysomes revealed a slight increase in maximal CO 2 fixation rate upon disruption of the shell and no significant difference in K C . This result is in stark contrast to the significant improvement in RubisCO kinetic performance measured for disrupted mutant carboxysomes and affirms the important role CsoSCA plays in facilitating the diffusion of cytosolic bicarbonate across the microcompartment shell. Furthermore, according to one of the quantitative models, the shell should also act as a barrier that reduces leakage of CO 2 from the carboxysome interior before it can be fixed by RubisCO (38).
Although considerable strides have recently been made toward establishing a model of carboxysome architecture that is based on protein structure information (11,12,13,17), our current understanding of shell structure is incomplete, and a satisfactory model that could form the basis for carboxysome function has yet to be formulated. Existing crystallographic data have established nearly seamless interactions between neighboring hexamers of CsoS1A. Because the three CsoS1 proteins make up the bulk of the shell (12), it is difficult to envision how CsoSCA dimers or the filaments they form in crystals (11) can be embedded in or transverse the shell within  Fig. 6). All three CsoS1 paralogs of H. neapolitanus are almost identical in amino acid sequence and are thought to form hexamers of similar, if not identical structure. It is possible that the C-terminal extension of 12 amino acids in CsoS1B, which distinguishes this shell protein from CsoS1A and CsoS1C, permits interaction with CsoSCA. The pores of H. neapolitanus CsoS1A hexamers bear a positive charge and have been hypothesized to facilitate flux of small, negatively charged molecules such as HCO 3 Ϫ across the carboxysome shell (12). If CsoSCA binds tightly to such a pore on the side that faces the inside of the carboxysome, its catalytic activity might create a sink that "pulls" bicarbonate through the pore (Fig. 6). Conversion of HCO 3 Ϫ to CO 2 would create an inorganic carbon species that is slow to diffuse out of the carboxysome interior. Furthermore, because CO 2 is the substrate for RubisCO, the product of the bicarbonate dehydration reaction would be consumed in the carboxysome interior and would drive continued directed "transport" of cytosolic bicarbonate into the microcompartment in a manner that is analogous to a chemical diode.
Such an arrangement would be consistent with the data presented here in that CO 2 fixation in mutant carboxysomes lacking CsoSCA would be dependent on high concentrations of CO 2 to overcome the diffusional resistance offered by the shell. Breeching shell integrity to allow free flux of bicarbonate and CO 2 would produce kinetic parameters similar to those observed for intact wild type carboxysomes, where the 40 dimers of CsoSCA are capable of producing sufficient CO 2 to saturate the carboxysomal RubisCO (15).
The data reported here support this model but at the same time raise a more fundamental question: because ruptured and intact wild type carboxysomes catalyze CO 2 fixation in vitro with similar efficiencies, the advantage these microcompartments offer the cell in vivo is not immediately obvious. Indeed, RubisCO freed from either wild type or csoS3::Km mutant carboxysomes catalyzes the fixation of CO 2 with efficiencies very similar to those measured for intact wild type carboxysomes (10,40). Perhaps this suggests that the substrate concentrations used to measure CO 2 fixation in vitro do not adequately reflect the in vivo conditions. One can envision that carboxysomes may provide an advantage for the cells under conditions of rapidly changing environmental pH and/or bicarbonate concentration. Perhaps, as previously suggested (1,15), in addition to its role in the carbon-concentrating mechanism, the carboxysome fulfills the important function of protecting carboxysomal RubisCO from the detrimental effects of oxygen (Fig. 6).