Carboxysome encapsulation of the CO2-fixing enzyme Rubisco in tobacco chloroplasts

A long-term strategy to enhance global crop photosynthesis and yield involves the introduction of cyanobacterial CO2-concentrating mechanisms (CCMs) into plant chloroplasts. Cyanobacterial CCMs enable relatively rapid CO2 fixation by elevating intracellular inorganic carbon as bicarbonate, then concentrating it as CO2 around the enzyme Rubisco in specialized protein micro-compartments called carboxysomes. To date, chloroplastic expression of carboxysomes has been elusive, requiring coordinated expression of almost a dozen proteins. Here we successfully produce simplified carboxysomes, isometric with those of the source organism Cyanobium, within tobacco chloroplasts. We replace the endogenous Rubisco large subunit gene with cyanobacterial Form-1A Rubisco large and small subunit genes, along with genes for two key α-carboxysome structural proteins. This minimal gene set produces carboxysomes, which encapsulate the introduced Rubisco and enable autotrophic growth at elevated CO2. This result demonstrates the formation of α-carboxysomes from a reduced gene set, informing the step-wise construction of fully functional α-carboxysomes in chloroplasts.

P hotosynthetic CO 2 fixation via ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) is the primary input of carbon into crop biomass. However, Rubisco-mediated CO 2 fixation in C 3 chloroplasts is catalytically slow, competitively inhibited by oxygen 1 and, from an agricultural stand-point, makes inefficient use of water and combined nitrogen 2 . These latter inefficiencies are driven by passive acquisition of CO 2 from the air (leading to water loss via open stomata) and by large investment in Rubisco (up to 50% of leaf protein) to overcome its poor kinetics 3 . A suggested approach to increase CO 2 fixation, minimize water-loss and decrease investment in Rubisco is to translate essential components of the cyanobacterial CO 2 -concentrating mechanism (CCM) into C 3 crops 4-6 .
The cyanobacterial CCM is a single-cell, bipartite system that first generates a high intracellular bicarbonate (HCO 3 − ) pool through action of membrane-bound inorganic carbon (C i ) transporters and CO 2 -converting complexes [7][8][9] (Fig. 1a). This HCO 3 − pool is then utilized by subcellular micro-compartments called carboxysomes, which encapsulate the cell's complement of Rubisco 10 . The carboxysome's outer protein shell enables diffusional influx of HCO 3 − and RuBP, where the former is converted to CO 2 by a localized carbonic anhydrase (CA). Physiological evidence 11 and mathematical models 12,13 suggest carboxysomes resist CO 2 efflux, resulting in concentration of CO 2 around Rubisco. Cyanobacterial carboxysomes possess high-catalyticturnover, but low-CO 2 -specificity Rubisco enzymes 1 . When the intracellular HCO 3 − pool is elevated, a high CO 2 environment can be generated inside the carboxysome, overcoming this low specificity and enabling rapid CO 2 fixation with reduced inhibition by oxygen 14 .
An engineering strategy to generate a chloroplastic CCM in crop plants (Fig. 1b) relies on transfer of genes encoding HCO 3 − transporters, directed to the chloroplast inner-envelope membrane (IEM), to generate an elevated stromal HCO 3 − pool, and genes encoding the carboxysome and its Rubisco 4,6 . Active HCO 3 − transporters in the chloroplast IEM alone are expected to improve photosynthesis due to the elevation of CO 2 concentrations around Rubisco 15,16 . Notably, either a carboxysome-   Fig. 1 Cyanobacterial CCM components for improved photosynthesis. Cyanobacterial CCMs (a) use bicarbonate (HCO 3 − ) and CO 2 pumps on the plasma and thylakoid membranes, respectively, to elevate cytosolic HCO 3 − . This HCO 3 − pool is utilized by icosahedral-shaped Rubisco microcompartments called carboxysomes (yellow icosahedron), where HCO 3 − is converted to CO 2 by localized carbonic anhydrase (CA) and accumulates due to resistive CO 2 efflux. The carboxysome encapsulates the cell's complement of a high-catalytic-rate Rubisco, which operates at close to V max , converting ribulose-1,5bisphosphate (RuBP) to 3-phosphoglycerate (3-PGA) within the Calvin cycle. This mechanism leads to efficient photosynthesis and reduced nitrogen allocation compared to C 3 plants. A strategy to improve C 3 photosynthesis in plant cells (b, represented here as rectangular structures containing dualmembrane chloroplasts) using the cyanobacterial CCM recommends independent transfer of carboxysomes containing their cognate Rubisco to the chloroplast stroma (i) and HCO 3 − pumps (ii) to the chloroplast inner-envelope membrane. Successful transfer of HCO 3 − transporters alone (ii) should generate a moderately elevated stromal HCO 3 − pool (indicated by the change in colour shading of the chloroplast) and has a predicted photosynthetic improvement of 9% 15 due to the resulting net elevation of CO 2 near Rubisco. Expression of functional carboxysomes (i), or just their cognate Rubisco, in the chloroplast stroma should lead to a high CO 2 -requiring (HCR) phenotype due to the characteristically high K M and low specificity for CO 2 of carboxysomal Rubiscos and absence of an elevated HCO 3 − or CO 2 pool. However, in combination (iii), generation of high stromal HCO 3 − pool in the presence of functional carboxysomes, with stromal CA eliminated, is predicted to generate a stromal HCO 3 − concentration approaching 5 mM 16 and to increases in CO 2 fixation and yield of up to 60% 15 . c Carboxysomes of Cyanobium PCC7001, used in this study, consist of many thousands of polypeptides, arranged in an icosahedral structure. In this model, a single layer of shell-bound Rubisco (CbbLS, green) is shown, with carboxysomal CA (orange). CsoS2 (yellow/brown) interlinks Rubisco and the shell made predominantly of CsoS1A hexamers (light blue). These and ancillary shell proteins (CsoS1D and CsoS1E, dark blue) enable substrate transport via central pores. Pentameric vertex proteins (CsoS4AB, purple) complete the structure encapsulated or free cyanobacterial Rubisco in C 3 plant chloroplasts will effectively lead to high CO 2 requirement for growth because cyanobacterial Rubiscos have low affinity and specificity for CO 2 1 . Stromal HCO 3 − pools in C 3 plants grown in air approximate 0.5 mM 17 , but the cyanobacterial cytoplasm reaches concentrations between 5 and 20 mM 18 , despite low external C i 19 , to drive the CCM. In combination, a high stromal HCO 3 − pool generated by active HCO 3 − transporters and a fully functional carboxysome where CO 2 can be elevated could improve C 3 plant CO 2 fixation and yield up to 60% 15,20 . This improvement would provide savings in energy costs for the plant and both nitrogen and carbon investment in the CO 2 fixation machinery 5 . Elimination of the native stromal CA and C 3 Rubisco to further improve the accumulation of HCO 3 − within the stroma is required to realize an optimal functioning chloroplastic CCM 4,6,16 . Within this proposed engineering strategy, construction of the carboxysome is particularly challenging due to genetic and protein-organizational complexity and requirements for functionality; some carboxysomes require coordinated expression of 13 genes 4 . Carboxysomes are a subset of proteinaceous bacterial microcompartments (BMCs 21 ), with specialized CO 2 anabolic function 22 . Two carboxysome types have arisen through convergent evolution: α-carboxysomes encapsulate Form-1A Rubisco in proteobacteria and some unicellular cyanobacteria, and βcarboxysomes encapsulate the plant-like Form-1B Rubisco in the remaining cyanobacteria 10,23 . Noting that the biogenesis and composition of each carboxysome type is unique 10 , components of the β-type lumen have been successfully expressed in Nicotiana tabacum (hereafter tobacco) chloroplasts 24 . This showed that cyanobacterial Form-1B Rubisco could be successfully expressed and cross-linked with CcmM35 25 to form large aggregates in the chloroplast 24 . Additionally, transient expression studies showed that carboxysome shell proteins could interact and form structures within chloroplasts 26 . However, these attempts could not generate structural carboxysomes nor encapsulate Rubisco, key requirements to generate CO 2 around Rubisco and for overall CCM functionality 5 . While carboxysomes have been heterologously expressed in bacterial systems 27,28 , there are currently no reports of αor β-carboxysome biogenesis in eukaryotic systems.
In this study, we designed simplified α-carboxysomes inspired by those from Cyanobium marinum PCC7001 (hereafter Cyanobium). Cyanobium carboxysomes likely consist of a protein shell primarily made up of CsoS1A, interspersed with proteins CsoS1D and/or CsoS1E 29 (Fig. 1c). Together, these proteins are envisaged to provide a selectively permeable shell, allowing HCO 3 − and RuBP into the carboxysome and 3-PGA release 30 but limiting CO 2 efflux 13,31 . Within the carboxysome, CsoSCA, a CA on the inner shell surface 32 , converts accumulated HCO 3 − to CO 2 . Rubisco (comprising CbbL and CbbS subunits) is likely anchored to the shell via CsoS2 33 , which arises as two isoforms from one gene in many α-carboxysomal species but only one isoform in Cyanobium 29,33 . The pentameric vertex proteins (CsoS4A and CsoS4B) complete the icosahedral structure, since their absence can lead to elongated carboxysomes with aberrant function 34 .
While a complete Cyanobium carboxysome requires at least nine polypeptides, generating plants containing genes for all these proteins would increase the risk of unforeseen errors in expression, transgene stability and carboxysome biogenesis. Instead, a bottom-up approach to carboxysome construction in a eukaryotic host seems a more practical proposal. Given the selforganizing nature of CsoS1A 35,36 , and the carboxysomeorganizing role of CsoS2 in complex with both CsoS1A and Rubisco 33 , we hypothesized that these components alone may provide a minimal set of proteins for a simplified carboxysome assembly design, with the potential for Rubisco encapsulation upon their co-expression. We constructed multigene cassettes for tobacco chloroplast transformation that contained genes for Cyanobium Rubisco large subunit (LSU, cbbL) and small subunit (SSU, cbbS) or these genes in combination with those for αcarboxysome proteins CsoS1A and CosS2 (Fig. 2). These genetic expression constructs were introduced into the tobacco plastome where they replaced the endogenous Rubisco LSU gene.
Here we report an example of structural carboxysomes, encapsulating a cyanobacterial Form-1A Rubisco, expressed in plant chloroplasts. The primary outcome is the formation of structurally identifiable and purifiable carboxysome structures, formed with Rubisco and just two shell proteins (CsoS1A and CsoS2). This provides a proof-of-concept for the construction of complete and functional carboxysomes within the chloroplast.

Results
Generation of transgenic plants. Chloroplasts of the Rhodosprillum rubrum-tobacco (cm trL ) master line 37 carrying the singlesubunit Rubisco (RbcM) were transformed with gene expression cassettes for Cyanobium Rubisco only (CyLS) or CyLS in concert with the carboxysome proteins CsoS1A and CsoS2 (CyLS-S 1 S 2 ; Fig. 2) using biolistic bombardment. Transgene expression cassettes also carried the spectinomycin selection marker gene, aadA, and transformants were recovered through selective tissue culture. Successful plantlets were subsequently grown in soil at 2% (v/v) CO 2 to flowering and seed collection. Southern blot analysis using a DNA probe specific to common sequence in wild-type, cm trL and transgenic plants confirmed the presence of DNA fragments of the anticipated size (Fig. 2). In addition, the complete loss of the R. rubrum rbcM DNA fragment indicated that transformant lines were homoplasmic (Fig. 2).
Protein content in transformed tobacco leaves. To determine whether successful introduction of transgenes led to production of cyanobacterial proteins, we conducted sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE) and immunoblot analysis of protein in leaf extracts. This confirmed the presence of the LSU and SSU of Cyanobium Rubisco in CyLS plant leaves, along with both carboxysome structural proteins CsoS1A and CsoS2 in CyLS-S 1 S 2 leaves (Fig. 3a). Using antibodies specific to the R. rubrum Rubisco (RbcM) in the the cm trL host plants, we confirmed the absence of RbcM from CyLS and CyLS-S 1 S 2 leaves. An antibody which detects both the tobacco RbcL and Cyanobium CbbL proteins confirmed that CyLS and CyLS-S 1 S 2 lines contained only the Cyanobium Rubisco. We also noted the presence of small quantities of the nuclear-encoded tobacco SSU (RbcS) in CyLS-S 1 S 2 leaf material and the successful expression of Cyanobium CbbS in transformants.
Leaf ultrastructure reveals carboxysome formation. To determine whether carboxysome protein expression led to ultrastructural chloroplast changes, we carried out transmission electron microscopy (TEM) of ultrathin leaf sections. Chloroplasts of CyLS plants showed no observable abnormalities, with chloroplasts of these plants typical of those found in wild-type leaves (Fig. 3b). CyLS-S 1 S 2 chloroplasts, however, contained multiple electron-dense particles (Fig. 3c, d), similar to Cyanobium α-carboxysomes 18 . Closely packed, localized clusters of geometric structures, approximately 100 nm in diameter, were observed in most chloroplast sections of CyLS-S 1 S 2 plants.
Presuming that the structures observed in CyLS-S 1 S 2 chloroplasts were carboxysomes, a variation of an α-carboxysome purification technique 38 was used to isolate particles from leaf tissue. The same method was also used to isolate carboxysomes from Cyanobium cells grown in liquid culture. Using this approach, ca. 100 nm particles were purified from CyLS-S 1 S 2 leaves by sucrose density-gradient centrifugation (Supplementary Fig. 1) and were found to be isometric with native Cyanobium carboxysomes, as determined by TEM and particle size analysis ( Fig. 3e-g, Table 1). Notably, the particles isolated from CyLS-S 1 S 2 leaves had slightly less defined vertices than their authentic Cyanobium counterparts ( Fig. 3e-f).
Protein content in isolated carboxysomes. Both the carboxysomes isolated from CyLS-S 1 S 2 plants and genuine Cyanobium carboxysomes were subjected to SDS-PAGE and immunoblots to determine protein presence and identity. The protein content of plant-derived carboxysomes was consistent with the protein complement of wild-type Cyanobium carboxysomes (Fig. 3h). Since wild-type Cyanobium carboxysomes consist of at least nine polypeptides and those of CyLS-S 1 S 2 plants only four, there was relatively more of each protein in plant-derived carboxysomes as a proportion of total protein (Fig. 3h). We also noted that the CyLS-S 1 S 2 carboxysomes were generally of higher purity than those isolated from Cyanobium ( Supplementary Fig. 2). Nonetheless, both CyLS-S 1 S 2 leaves and their isolated particles contained CbbL, CbbS, CsoS1A and CsoS2 in similar proportion to carboxysomes from Cyanobium (Fig. 3a, h). The complete suite of Cyanobium carboxysome proteins was absent from those in CyLS-S 1 S 2 plants due to the minimized gene set utilized. Despite the presence of small quantities of tobacco Rubisco SSU leaf extracts of CyLS-S 1 S 2 plants (Fig. 3a), it was absent from isolated particles, indicating no hybrid Cyanobium-tobacco Rubisco formation within the carboxysomes (Fig. 3h).
We identified CsoS2 as a single protein isoform in both Cyanobium carboxysomes and CyLS-S 1 S 2 particles (Fig. 3a, h). The SDS-PAGE banding pattern of CsoS2 in CyLS-S 1 S 2 plant extracts is indicative of potential degradation in the leaf (Fig. 3a), but notably the relatively clean band in isolated carboxysomes suggests that degraded protein is not incorporated into the carboxysome (Fig. 3h).
Using current knowledge of carboxysome protein interactions and the minimal protein set used to generate the structures found in CyLS-S 1 S 2 plants, we formulated a structural model of the carboxysomes formed in these plants (Fig. 3i). This model highlights the absence of specific shell components (i.e. proposed facet proteins CsoS1D and CSoS1E and vertex proteins CsoS4A and CsoS4B) and the internal carbonic anhydrase, CsoSCA, compared to the model of the Cyanobium carboxysome (Fig. 1c).
Immuno-gold localization and aberrant carboxysome formation. Immunogold labelling of CyLS-S 1 S 2 chloroplast TEM sections, using an antibody specific to the carboxysomal shell protein CsoS1A, showed an association of gold particles with the carboxysome structures (Fig. 4a-c). This confirmed that the structures observed in situ contained CsoS1A. This analysis highlighted the presence of occasional rod-shaped particles in chloroplasts that also reacted positively to CsoS1A antibodies. Approximately 4% of carboxysomes observed in chloroplast sections appeared to be elongated rods ( Fig. 4c-f, Table 1). These rod-shaped structures were also present in isolated carboxysome fractions (Fig. 4) at a rate of approximately 16% of purified carboxysome particles (Table 1). They were of variable length but of regular diameter [59 ± 5 (s.d.) nm], with sub-structural particles of~12 nm in diameter, which we interpreted as Rubisco (Fig. 4f).
Cyanobacterial Form-1A Rubisco-dependent plant growth. Presuming that inclusion of the Cyanobium Rubisco in tobacco chloroplasts, either alone or within carboxysomes, should lead to a high CO 2 requirement for growth ( Fig. 1), we characterized the    (Table 2 and ref. 18 ). Both CyLS and CyLS-S 1 S 2 plants demonstrated autotrophic growth at 2% (v/v) CO 2 (Fig. 5). Both transgenic plant types contained similar quantities of Rubisco but ten-fold less than wild-type tobacco ( Table 2). For CyLS plants, leaf

plants (e) and carboxysomes purified from
Cyanobium cyanobacterial cells (f). Scale bars for purified carboxysomes 100 nm. g Diameters of carboxysomes from wild-type Cyanobium cells (cyan line) and carboxysomes purified from CyLS-S 1 S 2 plants (magenta line) determined using a Nanosight particle analyser. h SDS-PAGE and immunoblots of proteins in particles isolated from CyLS-S 1 S 2 leaves and genuine carboxysomes from cultured Cyanobium cells (Cyanobium cbx). i A model of the carboxysome structures produced in transgenic CyLS-S 1 S 2 plant chloroplasts indicating the four protein components required to generate the structure. Comparison can be made with the complete wild-type structural model presented in Fig. 1c in which the pentameric vertex proteins (CsoS4AB), ancillary shell proteins (CsoS1D and CsoS1E) and carbonic anhydrase (CsoSCA) are present photosynthetic CO 2 response curves were consistent with the quantity and catalytic performance of the Cyanobium Rubisco, confirmed by modelling of photosynthetic rates using the Rubisco catalytic properties of the Cyanobium enzyme (Fig. 5, Table 2). However, the very low rates of CO 2 fixation by both transformants at the CO 2 concentrations provided in gas exchange experiments ( Fig. 5a, b), coupled with the high K C of their Rubisco and its potential progressive deactivation as CO 2 decreases in the gas-exchange chamber, provided potentially misleading information about true photosynthetic performance of these plants. To gather more information regarding the CO 2 assimilation phenotypes of CyLS and CyLS-S 1 S 2 plants, we conducted measurements at high CO 2 using a membrane inlet mass spectrometer (MIMS). Photosynthetic CO 2 responses of leaf discs by MIMS enabled analysis of photosynthetic assimilation rates up to 20 mbar CO 2 (~2% v/v, Table 2; Fig. 5c), mimicking physiological conditions that are predicted in a fully functional chloroplastic CCM. These conditions indicated a capacity for CyLS plants to reach similar photosynthetic rates to their wild-type counterparts ( Table 2; Fig. 5c) despite lower Rubisco content, consistent with the catalytic properties and content of the cyanobacterial Rubisco. MIMS analysis of CyLS-S 1 S 2 plants revealed lower CO 2 assimilation rates at 20 mbar CO 2 ( Table 2; Fig. 5c). This was consistently lower than mathematical modelling would indicate, based on Rubisco content and catalysis. CyLS-S 1 S 2 plants also took considerably longer to reach maturity ( Fig. 5d-j).
Catalytic parameters of isolated carboxysomes. To investigate the role that Rubisco encapsulation might play in the CO 2 assimilation phenotype of the CyLS-S 1 S 2 plants, carboxysomes were isolated from plant tissue and their functionality compared with those of isolated Cyanobium carboxysomes and the free CyLS Rubisco. In this analysis, an approximately two-fold higher Michaelis-Menten constant for RuBP (K MRuBP ) was observed in both the plant-derived and wild-type Cyanobium carboxysomes compared with the free Rubisco form ( Table 2, Supplementary  Fig 3), consistent with resistance to access of RuBP via the carboxysome shell.
We assessed the potential contribution of the carboxysome shell and the carboxysomal CA (CsoSCA) to Rubisco performance by carrying out CO 2 response assays on the free CyLS Rubisco, isolated CyLS-S 1 S 2 carboxysomes and Cyanobium carboxysomes, in the presence or absence of the CA inhibitor acetazolamide (AZ; Supplementary Fig. 3). This showed that there was no observable resistance to CO 2 influx by the carboxysome shell, with the free Cyanobium enzyme and the enzyme in both carboxysome types having similar K C values ( Table 2, Supplementary Fig. 3). The inhibitor AZ also had no observable effect on CO 2 fixation rates in the Cyanobium carboxysomes that contain a CA enzyme and no effect on either the free Rubisco or the CyLS-S 1 S 2 Rubisco ( Supplementary  Fig. 3).
We found with measurements of both CO 2 and RuBP supply to the enzyme that the plant-derived carboxysomes showed a significantly lower k cat than either of its counterparts ( Table 2, Supplementary Fig. 3), indicating that a sizable percentage of internalized Rubisco active sites were not capable of the expected rate of catalysis compared to naked Rubisco.
To determine whether the low k cat of the CyLS-S 1 S 2 carboxysomal Rubisco resulted from incorrect formation of L 8 S 8 holoenzymes, we assessed the relative stoichiometry of Rubisco isolated from CyLS plants with that found in enriched CyLS-S 1 S 2 and wild-type Cyanobium carboxysomes using western blots ( Supplementary Fig. 4). This confirmed that the plant-derived carboxysomes contained Rubisco with stoichiometry not significantly different to that of either the free enzyme from CyLS plants or that of Cyanobium carboxysomes.
Assuming that the catalytic phenotype of the plant-derived carboxysomes might result from inactivation of CyLS-S 1 S 2 carboxysomal Rubisco, we attempted to compare the catalytic performance of Rubisco in intact carboxysomes with that of the free enzyme after freeze-thaw treatment 38 commonly used to break the structures and release free Rubisco. While this was extremely successful for Cyanobium carboxysomes, we could not achieve significant rupture of carboxysomes from CyLS-S 1 S 2 plants ( Supplementary Fig. 2). Instead, to examine the relative performance of free Rubisco compared to that of carboxysomeencapsulated Rubisco from CyLS-S 1 S 2 leaf extracts, we used the supernatant fraction of crude leaf homogenates after high-speed centrifugation that contained Rubisco but was depleted in carboxysome proteins ( Supplementary Fig. 5). This revealed that the free Rubisco from CyLS-S 1 S 2 plants had both a lower K MRuBP and a higher k cat than its encapsulated counterpart ( Table 2). Rubisco in broken Cyanobium carboxysome preparations maintained a high k cat , but its K MRuBP was lower in the absence of the carboxysome shell (Table 2, Supplementary Fig. 3). As K C values did not appear to be diagnostic of carboxysome function in our assays, it was not determined for broken Cyanobium carboxysomes (Table 2).

Discussion
The expression of structurally intact carboxysomes within a C 3 plant chloroplast is a critical and complex engineering milestone towards the longer-term goal of attaining a functional chloroplastic CCM in C 3 crop plants 4,5 . The structures reported here mimic the gross structure of carboxysomes from Cyanobium but lack specific components expected to be required for full functionality in an operating CCM (viz. the CsoSCA, vertex proteins-CsoS4A and CsoS4B and potential metabolite-pore shell components CsoS1D and CsoS1E; cf. Figs. 1c and 3i). Despite these missing components, we show that a simplified set of proteins (consisting of the major a Proportions in each particle type in chloroplast sections and purified extracts were determined from analysis of TEM images. Elongated rods were defined as structures with average widths <75 nm and disproportionate length b Carboxysome dimensions were determined as the mode (±s.e.m.) of particle distributions measured using a Nanosight NS300 from 12 analyses of two independent preparations of Cyanobium carboxysomes and 19 analyses of three independent preparations of CyLS-S 1 S 2 carboxysomes c Elongated rod dimensions were determined as mean widths and lengths (±s.d.) of particles analysed in TEM images of purified carboxysome preparations from CyLS-S 1 S 2 leaves. Note that there is a greater proportion of elongated particles observed in purified extracts, which we assume to be due to a combination of: breakage of very long particles during isolation, giving rise to a greater proportion of rods and, difficulty in accurately identifying elongated carboxysomes in random cross-sections of chloroplasts. n.a. not applicable outer-shell component CsoS1A, the Rubisco linker CsoS2 and Rubisco itself) are capable of self-organizing to produce structural carboxysomes. This finding is an advance on previous attempts to generate either β-carboxysome shells 26 or interlinked Rubiscos 24 in chloroplasts and demonstrates a way forward to functional carboxysome construction in C 3 plants.
Simplified carboxysomes in CyLS-S 1 S 2 plants resembled those from Cyanobium but with predictable differences. Occasional elongated carboxysome structures in tobacco chloroplasts are consistent with observed structures resulting from the deletion of csoS4 vertex protein genes in Halothiobacillus neapolitanus 34 . A similar phenotype is found in β-carboxysomes where a genetic lesion interrupts expression of the vertex protein homologue CcmL 39 . Vertex protein mutants in both carboxysome types are functionally hindered, probably due to leakiness of the normally gas-tight shell 34,39 . We therefore conclude that the elongated structures observed in CyLS-S 1 S 2 plants and extracts result from the absence of csoS4A and csoS4B vertex protein genes in the expression cassette. The observed slightly rounded shape of CyLS-S 1 S 2 carboxysomes (Fig. 3e) is possibly attributable to the absence of the CsoS1D protein, which is predicted to be a large metabolite, gated pore in the mature α-carboxysome shell 30 and a likely contributor to shell rigidity 27 .
The generation of carboxysomes with just four proteins represents a pivotal step in our understanding of α-carboxysome biogenesis. Despite the need for additional proteins to produce structurally identical carboxysomes to those of Cyanobium, as few as four proteins are required to make an encapsulating body. Thus, relatively simple rules lead to the self-assembly of icosahedral protein bodies, of the correct size, containing Rubisco. This confirms our hypothesis that CsoS1A, CsoS2 and Rubisco alone are required for simple carboxysome formation. More complex requirements for α-carboxysome structure have been assumed in the past. For example, construction of the αcarboxysomes of H. neapolitanus in Escherichia coli utilized an almost complete carboxysome operon of ten genes to achieve structural formation 27 . Our results add further weight to a streamlined approach to carboxysome construction in plants, such as the synthetic domain-fusion approach described for βcarboxysomes 40 , but highlights the existence of already relatively simple gene sets in some cyanobacteria for this purpose.
The relative simplicity of the Cyanobium carboxysome makes it a suitable candidate for expression in chloroplasts. The utility of Cyanobium as a genetic donor is further emphasized by the formation of a single gene product from csoS2 (Fig. 3a, h 41 ). CsoS2 plays a critical role in carboxysome formation in α-cyanobacterial carboxysomes 33 and is a highly disordered protein that contains recognizable repeat domains at the Nterminus and the middle (M) region of the protein and a unique C-terminal region that possibly protrudes through to the carboxysome exterior 33 . In some α-carboxysomes, CsoS2 occurs as two isoforms (CsoS2A and 2B). Cyanobium, however, produces only one form of CsoS2, due to the lack of an internal frameshifting motif in the native gene sequence, which leads to C-terminal truncation during peptide synthesis in homologous sequences 41 . Additionally, Cyanobium carboxysomes require only one CsoS1 shell protein (Fig. 1c 29 ). The model α-carboxysome from H. neapolitanus has three CsoS1 shell proteins (CsoS1A, B and C) 36 , as does its β-carboxysome counterpart from Synechococcus elongatus PCC7942 42 . The α-carboxysome operons of the oceanic α-cyanobacteria are all relatively simple 29 , while the Cyanobium Rubisco has kinetic parameters that approach those of the β-cyanobacteria 18 , and outpace those found in its αcyanobacterial Prochlorococcus relatives 43 . Taken together, our results point to Cyanobium carboxysomes as ideal practical components for use in a chloroplastic CCM. The ability to isolate and analyse purified plant-generated carboxysomes with relative ease is important. The isolation of highly pure carboxysome fractions of α-carboxysomes 38 enables aspects of their structure and functionality to be addressed in vitro 44 . This ensures that analysis can confirm carboxysome function in C 3 chloroplasts in the absence of the functional HCO 3 − transporters required to generate a working chloroplastic CCM 4 . This is likely to save considerable time within the chloroplastic CCM engineering strategy. Here we have demonstrated the capability to analyse and evaluate aspects of function in our plant-expressed carboxysomes in comparison with those of Cyanobium.
Autotrophic growth of our transformed plants demonstrates the first example of C 3 plants reliant on a high-catalytic-rate Form-1A Rubisco. The α-carboxysomal Form-1A Rubisco is phyletically distant from the Form-1B isoform utilized by terrestrial plants, having arisen outside the plant lineage 45 . Limited data are currently available for Form-1A Rubisco catalytic properties, and this report highlights that further research in this area may give rise to alternative sources of Rubisco for C 3 photosynthesis augmentation. Furthermore, similar rates of photosynthesis in wild-type and CyLS leaf discs at 20 mbar CO 2 ( Table 2; Fig. 5c), despite 90% less Rubisco protein in CyLS leaves, highlights the potential to achieve photosynthetic performance with appreciable reduction in nitrogen investment towards Rubisco.
While there was no significant difference in Cyanobium Rubisco content between the two transformant lines, we consistently found a slightly lower content in the leaves of CyLS-S 1 S 2 plants (Table 2) and lower assimilation rates at 20 mbar CO 2 in MIMS assays (Fig. 5c). However, the assimilation rates of CyLS-S 1 S 2 plants was lower than predicted by modelling (Fig. 5c), based on the measured Rubisco content in crude leaf homogenates, and maximum catalytic properties of the naked enzyme (Table 2). Importantly, the density of carboxysomes in leaf extracts leads to the separation of two populations of Rubisco in CyLS-S 1 S 2 plants upon centrifugation (viz. insoluble carboxysomal Rubisco and free Rubisco). Thus observed maximum catalytic properties for CyLS-S 1 S 2 Rubisco represents those of the naked enzyme, resulting from either a population of unpacked Rubisco or loss from carboxysomes during homogenization. On the other hand, our Rubisco content measurements from crude homogenates are a true representation of the total active sites. We find, however, that the CO 2 assimilation rate of the CyLS-S 1 S 2 plants does not achieve expected rates based on both Rubisco content and maximum kinetics, leading to the conclusion that there is some kinetic impairment of the encapsulated Rubisco population. Currently, we cannot easily determine the relative quantities of Rubisco in each population due to difficulties in knowing the relative breakage of carboxysomes in extracts. In cyanobacteria, a proportion of Rubisco does appear to be in the free form, but this is difficult to measure with any degree of confidence 46 .
Based on MIMS analysis, it seems unlikely that the CyLS-S 1 S 2 photosynthetic phenotype results from carboxysome-mediated CO 2 diffusional limitation, since atmospheric CO 2 concentrations around 2% (v/v) are generally sufficient to overcome mutations  49 . We also found that both CyLS and CyLS-S 1 S 2 plants achieved maximum photosynthetic rates more rapidly than wild-type tobacco ( Supplementary Fig. 6), indicating rapid activation of Cyanobium Rubisco in tobacco chloroplasts. Thus we reasoned that Rubisco activation status was both unlikely to provide an explanation for CyLS-S 1 S 2 plant CO 2 assimilation rate nor provide meaningful results in this case. Instead, we determined whether the catalytic properties of the Rubisco in plantderived carboxysomes were impaired relative to the naked enzyme or that of wild-type carboxysomes isolated from Cyanobium. This revealed that Rubisco in plant-derived carboxysomes, lacking many of the protein factors found in wild-type carboxysomes, had a lower k cat than either their naked counterpart or that found in Cyanobium carboxysomes (Table 2). In addition, a lower k cat was predicted from modelling observed CO 2 assimilation in MIMS assays (Fig. 5). A similar observation was made by Occhialini and co-workers 50 where they successfully  Table 2, and a J max of 140 µmol m −2 s −1 , at 25°C. The model predicts Rubisco active site concentrations of 1.8 ± 0.4 (s.e.m.) and 1.7 ± 0.4 (s.e. m.) µmol sites m −2 for CyLS and CyLS-S 1 S 2 leaves, respectively, within the range determined from leaf tissue ( Table 2). Equations are outlined in Methods. c CO 2 assimilation rates of leaf discs from each plant line from plants grown at 2% (v/v) CO 2 in membrane inlet mass spectrometer (MIMS) assays, carried out as described in Methods. Solid lines (WT, magenta; CyLS, cyan; CyLS-S 1 S 2 , yellow) are averaged data from n = 3 independent leaf discs (±s.d., shaded areas). The dashed lines for CyLS and CyLS-S 1 S 2 are modelled assimilation rates using the same parameters as a, b. Using the model to estimate k cat for the Rubisco in CyLS-S 1 S 2 leaf discs gives an estimate of 6.46 s −1 ± 0.01 (s.e.m.) based on the measured Rubisco content in Table 2. d Growth measured as plant height post germination (±s.e.m., n = 4-6) for wild-type, CyLS and CyLS-S 1 S 2 plants grown at 2% (v/v) CO 2 . Growth phenotypes at 30 days after germination (e-g) and at maturity (h-j) of wild-type (e, h), CyLS (f, i) and CyLS-S 1 S 2 (g, j) plants grown in soil at 2% (v/v) CO 2 in 20 cm pots. Note the delayed germination and time to reach maturity in both transformant lines. Scale bars 5 cm co-expressed the cyanobacterial Form-1B Rubisco from S. elongatus with its cognate carboxysome-binding protein CcmM35 in tobacco chloroplasts. In that instance, a decrease in k cat was also observed, although they attributed this to an association with tobacco Rubisco SSUs 50 . We did not find any tobacco SSU associated with our purified CyLS-S 1 S 2 carboxysomes (Fig. 3h). However, when we attempted to isolate free Rubisco from CyLS-S 1 S 2 carboxysomes using the freeze-thaw technique 38 we were not successful (Supplemental Fig. 2). We speculate that this is due to incomplete composition of our minimal carboxysomes leading to different physical characteristics and changed Rubisco kinetics. This implies that our minimal carboxysomes lack a factor, or factors, required to ensure correct carboxysome structure and Rubisco catalytic performance. Future step-wise addition of other carboxysome proteins should enable an increase in catalytic performance to levels observed for the wild-type Cyanobium carboxysomes.
We also found that both CyLS-S 1 S 2 and Cyanobium carboxysome Rubisco had an elevated K MRuBP (Table 2, Supplementary  Fig. 3), indicative of resistance to substrate influx in both carboxysome types. This suggests plant-derived carboxysomes are relatively intact and that the absence of the vertex proteins does not necessarily result in substantial changes in permeability to RuBP. The observed increase in K MRuBP is unlikely to explain the CyLS-S 1 S 2 plant CO 2 assimilation phenotype alone, although we cannot rule out the possibility that this diffusional resistance effect might be magnified in the observed clusters of carboxysomes inside chloroplasts (Figs. 3 and 4). This clustering is reminiscent of carboxysomes in β-cyanobacterial mutants lacking the ability to properly distribute their carboxysomes 42 . We hypothesize that an even distribution of carboxysomes within the stroma is preferable to prevent localized concentration gradients that could limit substrate access to carboxysomes.
We could not detect a predicted resistance to CO 2 flux across the carboxysome shell 12,13 as assessed by no observable change in K C for intact carboxysomes compared with the free enzyme ( Table 2). We also found that the CA inhibitor AZ had no effect on carboxysomal Rubisco kinetics in our assays ( Supplementary  Fig. 3). We propose that this results from an inability to reproduce, in vitro, the CO 2 :HCO 3 − disequilibrium, which is normally achieved in cyanobacterial cells [11][12][13] . Previous modelling suggested that, under standard Rubisco assay conditions at pH 8.0, the CO 2 :HCO 3 − ratio (~0.015) would lead to negligible differences in the observed carboxylation rates by naked and carboxysomal Rubisco 51 . Thus we assume that, even at low CO 2 concentrations in a typical Rubisco CO 2 response assay, there is sufficient CO 2 present to overcome carboxysome shell resistance and the need for an internal CA in the carboxysome preparations described here. Further development of functional assays is needed to investigate the true properties of the carboxysome shell.
Both the marginally lower Rubisco content and poor growth phenotype of CyLS-S 1 S 2 plants might also be indicative of impairment of protein expression. A recent study revealed the detrimental effects of overusing intercistronic expression elements (IEEs) in plastome transformation constructs 52 and here we have utilized multiple IEE sequences in the CyLS-S 1 S 2 gene cassette (Fig. 2), based on successful gene constructs used for βcarboxysomal protein expression 24 . While this possibility does not explain the CO 2 assimilation phenotype of the CyLS-S 1 S 2 plants, it may contribute to the slow germination and overall poor growth phenotype. This knowledge will inform future construct design to improve protein expression and maintain correct chloroplast function.
To generate fully functional carboxysomes within the chloroplast, further careful engineering is required. While there have been some reports of successful insertion and expression of small multi-transgene operons in chloroplasts (e.g. refs. 53,54 ), our stepwise construction of a simplified synthetic carboxysome operon minimizes the potential for errors in expression, transgene stability and carboxysome biogenesis. This also allows for a bottom-up approach to carboxysome construction, enabling analysis and evaluation of component functionality in a step-wise fashion. Addition of the minor shell proteins CsoS1D/E may be required to facilitate metabolite transport 30 and fine-tuning of structure 27 , while both vertex proteins (CsoS4A and B) will be required to minimize rod-shaped carboxysome formation and ensure closure of a functional shell 34,55 . The CsoSCA must also be included to maintain the high carboxysomal CO 2 concentrations to support cyanobacterial Rubisco catalysis 38 . Requirement for Rubisco folding and activation chaperones (acRAF-α-carboxysome Rubisco Accumulation Factor 56 ; and CbbX-a member of a Rubisco activase family whose proteobacterial functional homologue CbbQ/O 57,58 is found associated with proteobacterial αcarboxysomes 59 ) is not yet known. We hypothesize that inclusion of one or all these factors should enhance the observed Rubisco catalysis to that found for Cyanobium carboxysomes. Beyond the construction of a functional carboxysome, a working CCM is ultimately dependent upon transport proteins, which can boost stromal HCO 3 − concentrations to levels that enable carboxysome operation 5 . Recent advances in correct localization of inorganic carbon pumps to the chloroplast inner membrane 60,61 provide encouraging progress towards the achievement of this goal.
The work presented here not only sets a molecular baseline for the construction of a complete carboxysome as part of a chloroplastic CCM but also informs the assembly of α-carboxysomes and shows that simple structures can form with as few as four proteins. Coupled with results from a recent study on the principles of micro-compartment shell assembly 62 , the work presented here also generates alternative avenues to tailor microcompartment design based on simplified sets of genes. Our results draw attention to the use of α-carboxysomes (whose origins lie in already simplified gene operons) as a less complicated alternative to β-carboxysomes for the encapsulation of Rubisco in a working chloroplastic CCM. Despite the apparent complex nature of carboxysome construction, this advance indicates a feasible way forward to the engineering of a chloroplastic CCM and improved photosynthesis in C 3 crops.

Methods
Tobacco chloroplast transformation. Chloroplasts of the R. rubrum-Rubisco tobacco master line were transformed with CyLS and CyLS-S 1 S 2 constructs through biolistic bombardment according to Maliga and Tungsuchat-Huang 63 using of 2.5 mg of tungsten particles coated with freshly prepared plasmid DNA (10 µg). Each leaf was bombarded with 0.5 mg DNA-coated tungsten particles. This enabled recombination of the genes of interest in place of the R. rubrum Rubisco LSU (rbcM) gene locus in the plastid genome and selection using a spectinomycinresistance marker gene (aadA) downstream of the genes of interest, under the control of the tobacco psbA promoter (Fig. 2). Successful explants were cultured on regeneration medium (Murashige and Skoog medium supplemented with 3% (w/v) sucrose, 0.5 mg mL −1 spectinomycin, 1.0 mg L −1 6-benzylaminopurine [BAP], 1.0 mg L −1 1-naphthaleneacetic acid [NAA], 1.0 mg L −1 thiamine-HCl, 100 mg L −1 myo-inositol and solidified with 0.6% [w/v] agar) in controlled temperature cabinets (Thermoline, Wetherill Park, NSW, Australia) supplied with 2% (v/v) CO 2 . Plantlets were assessed with western immunoblot and PCR to identify successful transformants. Homoplasmic transformants were obtained by subculturing the plantlets in regeneration medium. Homoplasmic plantlets were transferred to rooting medium (regeneration medium lacking BAP, NAA, thiamine and myoinositol) and then to soil to grow to maturity for seed harvest.
Plant growth analysis and gas exchange. For growth and gas-exchange experiments, plants were grown from seed germinated in Green Wizard Premium Potting Mix™ (Scotts Australia, Bella Vista, NSW, Australia) in 20 cm pots supplemented with Osmocote Exact™ (ICL Australia & New Zealand, Bella Vista, NSW, Australia) at a rate of 4 g L −1 potting mix. Plants were grown in a growth room located at the Research School of Biology Control Environment Facility, Australian National University, Canberra, Australia, with a 23/22°C day/night temperature regime, 12/12 h light/dark cycle, 60% relative humidity and 500 µmol photons m −2 s −1 , at 2.0% (v/v) CO 2 .
Gas exchange and modelling. Assimilation rates (A; µmol m −2 s −1 ) at 25°C over the range of chloroplastic CO 2 partial pressure (C c ; µbar) were examined during gas exchange experiments using the portable flow-through LI-6400 gas-exchange system (LI-COR, Nebraska, USA). Data were modelled according to Farquhar et al. 64 and von Caemmerer 65 as the minimum of the following equations using the Rubisco catalysis (k cat , s −1 ; K  Table 2. An ambient O 2 partial pressure (O) of 200 mbar, and a non-photorespiratory CO 2 release (R d ) of 1 µmol m −2 s −1 were used. For Eq. 2, an electron transport rate (J) of 140 µmol m −2 s −1 was used. In these simulations, it was assumed that C c ≈ C i , the intercellular CO 2 partial pressure measured by gas exchange.
Southern blot analysis. Cellular DNA was isolated from 10 to 15 mg of leaf tissue using the DNeasy Plant Mini Kit (QIAGEN, Chadstone, VIC, Australia). This DNA (0.5 µg) was digested with FastDigest Bsp119I (Thermo Fisher Scientific, Scoresby, VIC, Australia) at 37°C for 90 min. The digested DNA was separated on 0.9% (w/v) agarose by gel electrophoresis at 90 V for 2 h. The gel was prepared with 0.5× TBE and 1× SYBRSafe (Thermo Fisher Scientific). After electrophoresis, the gel was placed in denaturation buffer (0.5 M NaOH, 1.5 M NaCl) for 30 min with gentle agitation and rinsed with ultrapure water. Next it was placed in neutralization buffer (0.5 M Tris, 1.5 M NaCl, pH 7.5) for 30 min with gentle agitation and rinsed again with ultrapure water. DNA was transferred from the gel to Hybond-N+Membrane (GE Healthcare Life Sciences, Parramatta, NSW, Australia) by capillary blotting overnight as described 66 . DNA was then ultraviolet (UV) cross-linked to the membrane using the Spectrolinker XL-1000 UV Crosslinker (Spectronics Corporation, Westbury, NY, USA) in the Optimal Crosslink mode. The membrane was prehybridized for 2 h in 0.5 M NaCl with 4% (w/v) blocking reagent (GE Healthcare Life Sciences) in a hybridization oven at 55°C. The probe was prepared using the AlkPhos Direct Labelling System (GE Healthcare Life Sciences) and a 483 bp PCR product generated from the accD gene in the 3'flanking region of the plastome insertion locus (Fig. 2) (Primers accD reverse: 5'-AAAGGGCGGCTTCTCCTATG-3', accD forward: 5'-TGCAATTAAAC TCGGCCCAA-3'). The probe was hybridized to the membrane overnight at 55°C. The following day, two 10 min primary washes were performed at 55°C in the hybridization oven. The wash buffer contained 2 M urea, 0.1% SDS (w/v), 50 mM NaH 2 PO 4 pH 7.0, 150 mM NaCl, 1 mM MgCl 2 and 0.2% (w/v) blocking reagent (GE Healthcare Life Sciences). Two 5 min secondary washes were performed at room temperature in 50 mM Tris with 2 mM MgCl 2 . The blot was then placed for 1 h in the dark in 1 mM AttoPhos Fluorescent Substrate (Promega, Alexandria, NSW, Australia) and imaged with a ChemiDoc MP Imaging System (Bio-Rad, Gladesville, NSW, Australia) using Epi-blue illumination and a 530/28 filter.
Carboxysome purification. Carboxysomes were purified from culture-grown Cyanobium cells and CyLS-S 1 S 2 plants essentially as described by So et al. 38 for αcarboxysomes, with some minor modifications. For the purification of carboxysomes from the cyanobacterium Cyanobium, cells were grown in 10 L BG-11 freshwater medium, sparged with air enriched with 2% (v/v) CO 2 . Cells were collected by centrifugation (6000 × g, 10 min) and resuspended in 25 mL TEMB buffer (5 mM Tris-HCl [pH 8.0], 1 mM EDTA, 10 mM MgCl 2 , 20 mM NaHCO 3 ) containing 0.55 M mannitol and 60 kU rLysozyme (Merck-Millipore, Bayswater, VIC, Australia). Cells were incubated at 37°C in the dark for 2-16 h with gentle shaking to enable cell wall degradation and then collected by centrifugation as above. Cells were placed on ice and resuspended in 10 mL ice-cold TEMB for 15 min prior to three passages through an EmulsiFlex-B15 cell disruptor (Avestin, Ottawa, ON, Canada) at a homogenizing pressure of~15,000 psi. IGEPAL CA-630 (Sigma-Aldrich, Castle Hill, NSW, Australia) was added to a final concentration of 1% (v/v), and broken cells were mixed by inversion on a rotating shaker at 4°C for 1 h. Cell debris and unbroken cells were removed by centrifugation at 3000 × g, 1 min, and the supernatant centrifuged at 40,000 × g for 20 min to generate a crude carboxysome pellet. The pellet was washed again in 20 mL TEMB containing 1% (v/v) IGEPAL CA-630. The final pellet was resuspended in 1.5 mL TEMB and clarified by centrifugation at 3000 × g, 1 min prior to loading onto a 10-60% (v/v) linear sucrose gradient in TEMB. Gradients were centrifuged at 105,000 × g for 60 min and the milky-white band towards the bottom of the gradient was collected, diluted in 35 mL TEMB and recentrifuged at 100,000 × g for 60 min. The final carboxysome pellet was resuspended in 500 µL TEMB prior to analysis for protein content, particle size analysis, TEM and Rubisco assays.
Purification of carboxysomes from CyLS-S 1 S 2 plants was carried out in analogous manner except that 10 g (fresh weight) leaves were extracted in 100 mL TEMB (containing plant protease inhibitor cocktail; Sigma-Aldrich) using an Omni Mixer homogenizer (Kennesaw, GA, United States) on ice with three 15 s pulses. IGEPAL CA-630 was added to a final concentration of 1% (v/v) and extracts were mixed with gentle inversion for 60 min at 4°C. Extracts were then filtered through a single layer of Miracloth (Merck-Millipore, Bayswater, VIC, Australia) prior to removal of heavy leaf debris and starch at 3000 × g for 1 min. The supernatant was subjected to centrifugation at 40,000 × g for 20 min and subsequent isolation procedures carried out on the resulting pellet in an identical manner to those used for wild-type Cyanobium carboxysomes.
Carboxysomes isolated from both CyLS-S 1 S 2 leaves and Cyanobium were subjected to particle size analysis using a Nanosight NS300 apparatus (Malvern Instruments, Malvern, UK) essentially according to the manufacturer's instructions. Samples were diluted 1:10,000 in reverse-osmosis purified, de-ionized, filtered water and delivered to the instrument via syringe at a pump speed of 50 units. Particles were illuminated with a blue laser at 405 nm and the instrument operated at 25°C. A series of five 1 min videos were collected for each sample and subsequently analysed. Video data from multiple, independent carboxysome extractions were analysed and data combined to determine a final particle diameter, reported as the estimated mode ± s.e.m.
For analysis of carboxysome catalytic performance, highly enriched carboxysome fractions obtained immediately prior to application onto sucrose gradients was used for both Cyanobium and CyLS-S 1 S 2 plants. This ensured a high concentration of material with low probability of carboxysome shell breakage sometimes observed after removal from sucrose gradients 67 and enabled same-day Rubisco analysis.
Leaf Rubisco content and kinetic analysis. Leaf disc samples (0.5 cm 2 ), taken at the site of gas exchange immediately after measurement, were used for Rubisco content and kinetic analysis. Radiolabelled [ 14 C] carboxyarabanitol-P 2 (CABP) was used to measure leaf Rubisco content as described 68,69 . In the case of CyLS-S 1 S 2 plants, crude leaf homogenates, clarified only by a low-speed centrifugation step (3000 × g, 1 min), were used in the determination of Rubisco content to avoid excess losses of the insoluble carboxysomes. The same method used to measure leaf Rubisco content was applied to both leaf extracts containing isolated Rubisco and enriched carboxysomes used in activity assays. Radiolabelled 14 CO 2 fixation assays were used to measure maximal carboxylase activities (v c ) and the Michaelis constant for CO 2 at both ambient pO 2 (K C 21%O2 ) and under nitrogen (K C N2 ) using 30-440 µM 14 CO 2 at 25°C, pH 8.04 as described by Sharwood et al. 70 . Samples containing Rubisco were typically activated for 5 min prior to assay. Catalytic turnover rates (k cat ; s −1 ) were determined by dividing the V c max by Rubisco content as determined by 14 C-CABP binding as described 70 . Similar assays were performed using 10-800 µM RuBP at 20 mM MgCl 2 , ambient pO 2 and 20 mM NaHCO 3 to determine K MRuBP . In the case of CyLS-S 1 S 2 plants, analysis of Rubisco NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-06044-0 ARTICLE NATURE COMMUNICATIONS | (2018) 9:3570 | DOI: 10.1038/s41467-018-06044-0 | www.nature.com/naturecommunications kinetics was carried out on what we deemed to be free Rubisco after centrifugation of leaf extracts at 20,000 × g, 1 min, or carboxysomal Rubisco after enrichment of carboxysomes. To generate broken Cyanobium carboxysomes with released Rubisco, enriched carboxysome samples were centrifuged (20,000 × g, 10 min) and the supernatant discarded. Pellets were then frozen at −20°C for at least 30 min and then quickly resuspended in TEMB using a pipette to achieve breakage.
The response of CyLS Rubisco, CyLS-S 1 S 2 carboxysomes and Cyanobium carboxysomes to the CA inhibitor AZ was carried out under normal assay conditions except that 500 µM AZ was included in both activation and assay buffers, allowing incubation of samples with AZ for 5 min at 25°C prior to assay. The same concentration of the more membrane permeable analogue EZ has been shown to reduce CsoSCA activity to 13% of maximum in preparations of affinitypurified, recombinant CsoSCA enzyme from H. neapolitanus carboxysomes, expressed in E. coli 38 .
In vitro S C/O for Cyanobium Rubisco was determined for protein extracted from CyLS plants using 5-10 cm 2 leaf material. Protein was extracted in 50 mM EPPS, pH 8.0, 5 mM MgCl 2 containing plant protease inhibitor cocktail (Sigma-Aldrich) and initially purified using ion-exchange chromatography on 1 mL High Q cartridge columns (Bio-Rad). Rubisco was eluted using 2 mL elution buffer (50 mM EPPS, pH 8.0, 1 M NaCl) and further purified by size-exclusion chromatography on a Superdex 200 column equilibrated with specificity buffer (30 mM triethanolamine pH 8.3, 30 mM Mg acetate) using an ÄKTA Pure chromatography system (GE Healthcare). Fractions containing Rubisco were pooled and concentrated by centrifugation through 30 kDa molecular weight cutoff filters. Purified Rubisco was used to catalyse the production of radiolabelled and 3-PGA and 2-phosphoglycolate from 3 H-RuBP under an atmosphere containing 500 ppm CO 2 in O 2 , generated using Wösthoff precision gas-mixing pumps at 25°C, pH 8.3. The ratio of radiolabelled products was determined after their separation by highperformance liquid chromatography and scintillation counting to estimate specificity 71 .
Membrane inlet mass spectrometry. CO 2 response curves of 1.2 cm 2 leaf discs from wild-type tobacco, CyLS and CyLS-S 1 S 2 plants were determined according to the methods described by Maxwell et al. 72 using a purpose-built cuvette attached to a Micromass membrane inlet mass spectrometer. Discs were taken from third leaf of the same plants used for attached-leaf gas exchange [grown at 2% (v/v) CO 2 ]. At least three discs were analysed for each plant line. CO 2 response was determined up to 20 mbar ambient CO 2 (C a ) and measured as CO 2 consumption by the mass spectrometer.
Source data. The uncropped images of gels and blots are provided as Supplementary Figs. 7-8.

Data availability
All relevant data and plant materials are available from the authors upon request. Raw data corresponding to the figures and results described in this manuscript are available online [https://doi.org/10.25911/5b4edb4bada74]. Additional data reported in this paper are presented in the Supplementary Information. Nucleotide sequences are deposited in GenBank with accession numbers MH051814 and MH051815 for CyLS and CyLS-S1S2 tobacco lines, respectively.