Evolving Methanococcoides burtonii archaeal Rubisco for improved photosynthesis and plant growth

In photosynthesis Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) catalyses the often rate limiting CO2-fixation step in the Calvin cycle. This makes Rubisco both the gatekeeper for carbon entry into the biosphere and a target for functional improvement to enhance photosynthesis and plant growth. Encumbering the catalytic performance of Rubisco is its highly conserved, complex catalytic chemistry. Accordingly, traditional efforts to enhance Rubisco catalysis using protracted “trial and error” protein engineering approaches have met with limited success. Here we demonstrate the versatility of high throughput directed (laboratory) protein evolution for improving the carboxylation properties of a non-photosynthetic Rubisco from the archaea Methanococcoides burtonii. Using chloroplast transformation in the model plant Nicotiana tabacum (tobacco) we confirm the improved forms of M. burtonii Rubisco increased photosynthesis and growth relative to tobacco controls producing wild-type M. burtonii Rubisco. Our findings indicate continued directed evolution of archaeal Rubisco offers new potential for enhancing leaf photosynthesis and plant growth.


Figure 1. Selecting for improvements in MbR catalysis using Rubisco Dependent E. coli (RDE).
(a) Simplified schematic of the RDE selection system that uses a glycolysis/gluconeogenesis interrupted glyceraldehyde-3-phosphate dehydrogenase deletion (gapA − ) strain of E. coli (MM1). Ectopic expression of phosphoribulokinase (PRK) and Rubisco in MM1 acts as a bypass shunt for glycolysis to enable carbon flow from hexose carbon to the TCA cycle for energy and growth. PRK catalyses the conversion of ribulose-5phosphate (rib-5-P) produced by the pentose phosphate pathway (PPP) to RuBP, which is toxic to cell growth.
Scientific RepoRts | 6:22284 | DOI: 10.1038/srep22284 bacterial L 2 Rubisco from Rhodospirillum rubrum as well as mutations that significantly enhance the assembly (solubility) of cyanobacteria L 8 S 8 Rubisco and, in one instance, marginally improved all catalytic parameters 12,24 . More recently, the MM1-prk RDE identified a Synechocystis PCC6803 L 8 S 8 Rubisco mutant with 3-fold improvements in carboxylation efficiency that improved photosynthesis rates by > 50% when re-integrated into the high CO 2 carboxysome compartment within the cyanobacterium 11 . In contrast, it is unlikely that these improvements would be of benefit in plant leaves as the high CO 2 levels needed to account for the low CO 2 affinity and poor S C/O of cyanobacteria Rubisco are not met by chloroplasts 25 .
The non-photosynthetic role of the ancient L 2 /L 10 Rubisco isoforms in archaea implies they have likely undergone alternative selection pressures to photosynthetic Rubisco during evolution. For example, Rubisco from archaea have a high affinity for RuBP and high thermostability but low carboxylation rates (k cat C ) and S C/O 26-28 . This catalytic distinctiveness arises from the alternative biological role of archaeal Rubisco in the pentose bisphosphate pathway where it functions to metabolize the RuBP produced during nucleoside metabolism 29 . Despite its non-photosynthetic function, archaeal Rubisco can still support plant photosynthesis and growth. For example, the Methanococcoides burtonii L 10 Rubisco (MbR) is highly expressed in leaf chloroplasts and shown to support the growth to fertile maturity of tobacco under the high CO 2 levels needed to accommodate the low k cat C and S C/O of MbR 26 .
The significantly poorer carboxylase properties and alternative function of Rubisco in archaea suggest this form of the enzyme has adapted to alternative evolutionary pressures compared with Rubisco in photosynthetic organisms. This questions whether the carboxylation properties of the archaeal L 10 Rubisco might be more amenable for improvement towards those required for enhanced photosynthetic potential. To address this question we used the MM1-prk RDE system (Fig. 1a) to select evolved MbR mutants with improvements in catalytic properties that are required to enhance C 3 -plant photosynthesis 5,30 . These properties include increasing k cat C , carboxylation efficiency (k cat C divided by K C

21%O2
; the K m for CO 2 under ambient O 2 ) and S C/O . Using chloroplast genome (plastome) transformation we introduce mbR genes into tobacco to demonstrate successful translation of improved MbR properties selected in E. coli into leaf chloroplasts. The enhanced photosynthesis and growth of the transformed plants producing improved MbR mutants relative to control lines producing non-mutated MbR provides novel proof of concept on the utility of improving Rubisco catalysis by directed evolution in E. coli to improve the CO 2 -assimilation rate in leaves.

Directed evolution of M. burtonii Rubisco (MbR) in E. coli. The native mbiiL gene coding M. burtonii
Rubisco is efficiently translated in E. coli and assembles into abundantly expressed (> 6% (w/w) of soluble cell protein) as functional L 2 Rubisco (MbR) 26 . In the presence of substrate RuBP (or structurally comparable sugar phosphate ligand) the L 2 units assemble into a stable L 10 MbR complex. Random mutations were introduced into mbiiL using error prone PCR (averaging 2 mutations per kb) and the mutant genes ligated into a lac inducible vector pTrcHisB 20 .
Three mbiiL libraries (each comprising ~180 k variants) were transformed into MM1-prk cells (Fig. 1a) and grown at 23 °C as described 20 . The initial selection was performed under high-Rubisco inducing (0.5 mM IPTG) and low-PRK inducing conditions (0.05% (w/v) arabinose) in air supplemented with 2.5% (v/v) CO 2 . After 9-16 days 80 colonies showing improved growth relative to MM1-prk producing wild-type MbR were identified (Supplementary Table 1). The Rubisco-containing plasmid from each colony was sequenced revealing substitutions in 78 of the 474 amino acids (Supplementary Table 1). Each mbiiL mutant was cloned back into pTrcHisB and re-transformed into MM1-prk RDE cells and separately grown under higher Rubisco activity selection (i.e. on media containing 0.1% (w/v) arabinose to elevate PRK expression; Fig. 1b). Colony growth was scored relative to MM1-prk cells expressing wild-type MbR that could not grow on media containing 0.1% (w/v) arabinose (Fig. 1b). Seven mbiiL mutant genes were found to convey a distinct selective advantage to MM1-prk E. coli growth (Fig. 1c).
The seven mutant and wild-type pTrc-mbiiL genes were expressed in XL1-Blue E. coli without co-expressing PRK. This resulted in only L 2 MbR oligomers being formed as the cells made no RuBP that is required for the formation of L 10 MbR complexes 26 . The cellular content and catalytic properties of each L 2 MbR enzyme was measured ( Fig. 2a). Maximal rates of CO 2 -fixation (k cat C ) at 25 °C were determined under ambient O 2 levels (~252 μM O 2 ) and at pH 7.2 due to the low pH favoured by MbR catalysis 26 . Under these conditions MbR mutant isolates #1 (MbR-K332E), #10A (MbR-E138V) and #63 (MbR-T421A) showed significant 40% to 90% improvements in k cat C and corresponding 10% to 25% increases in S C/O (Fig. 2a). Quantification of MbR expression in E. coli by 14 C-CABP binding and confirmation by SDS PAGE (Fig. 2b) showed that most mutations had little effect on the level of MbR expression. The MbR-T421A mutant (#63) showed a modest, but significant, increase in expression while the mutations in MbR mutants #14, #23 and #45 significantly impeded MbR production (Fig. 2b).
RuBP toxicity can be alleviated by Rubisco catalysed carboxylation followed by utilisation of the 3PGA product in glycolysis. Expression of PRK and Rubisco is carefully regulated by varying the inducing agents arabinose and IPTG respectively 20 . Early cell division is facilitated by addition of trace levels of glycerol (0.4% v/v; upstream C-source), 30 mM malate and 0.5% w/v cas amino-acids (downstream C-and N-sources). (b) Comparative growth of MM1-prk cells expressing "improved" mutant and wild type MbR under increasing levels of arabinose induced PRK expression. (− ) no growth; (+ to + + + + ) relative increase in colony size. Expression of improved MbR in tobacco chloroplasts. The tobacco genotype cm trL has been genetically tailored for Rubisco engineering using chloroplast transformation 31 . In the plastome of cm trL chloroplasts the wild type rbcL gene has been replaced with a synthetic, codon modified version of the R. rubrum bacterial rbcM gene ( cm rbcM) that codes for Form II L 2 Rubisco in place of tobacco L 8 S 8 Rubisco (Fig. 3a). The increased O 2 sensitivity of R. rubrum L 2 Rubisco reduces both S C/O and carboxylation efficiency under ambient O 2 (k cat C /K C 21%O2 ) by ~7-fold relative to tobacco L 8 S 8 Rubisco (Table 1). These poorer catalytic properties result in the cm trL genotype requiring high CO 2 for growth in soil 31 . As shown by Alonso et al., (2009) the catalytic properties of MbR (both in L 2 and L 10 complexes) are even more impeded than R. rubrum L 2 Rubisco, especially with increasing alkaline pH. Despite this impairment, transplastomic replacement of the cm rbcM gene in cm trL with the wild-type mbiiL gene generated the L 10 MbR producing tobacco genotype tob mbiiL that could survive under elevated CO 2 (2.5% v/v) in soil 26 . The tob mbiiL lines took more than 300 days to reach fertile maturity compared with ~30 days for wild type tobacco and ~32 days for cm trL under the same growth conditions.
To test whether the evolved MbR enzymes translated to improved tobacco photosynthesis and growth, synthetic mbR genes were made that incorporated the codon use of the tobacco rbcL gene (Fig. 3a). In addition, the native MbR N-terminal MSLIYEDLV sequence was replaced with the MSPQTETKASVGF sequence of the tobacco L-subunit that undergoes a range of post-translational modifications that tentatively provide protection from proteolysis 13 . Three mbR genes coding wild-type MbR, MbR-K332E and MbR-E138V were cloned into the pLEV4 plastome transforming plasmid and transformed into cm trL leaves 31 . Transplastomic tobacco lines producing L 10 MbR were identified by native PAGE (Fig. 3b). At least two independent lines for each of the tob MbR , tob MbRE138V and tob MbRK332E genotypes were continuously propagated on spectinomycin-containing media until homoplasmic (i.e. no longer producing L 2 R. rubrum Rubisco) before growing the T 0 plants to maturity in soil in air supplemented with 2.5% [v/v] CO 2 .
Only L 10 MbR was detected in leaves ( Fig. 3b) due to the continuous production of RuBP under illumination and the relative stability of the decameric complex 26 . While the T 0 tob MbR-E138V and tob MbR-K332E plants grew substantially quicker than tob MbR , little difference was detected in the L 10 MbR content in comparable upper canopy leaves of the juvenile (~21 cm tall) T 0 plants (Fig. 3c). When at ~60 cm in height, 3-6-fold higher levels of L 10 MbR were measured in the newly emerging upper canopy leaves with significantly higher amounts detected in the faster growing, healthier looking, tob MbRE138V and tob MbRK332E T 0 plants (Fig. 3c). At both development stages, the leaf MbR levels were generally 3-4-fold lower than the L 2 and L 8 S 8 Rubisco content in the cm trL and wild-type tobacco controls growing alongside.  14 C-CABP quantification of Rubisco active site content in comparable young upper canopy leaves of each genotype during early (~20 cm in height, colored bars) and late (~60 cm in height, black bars) exponential growth. N. tabacum [2] (81) Limitations in the steady state mbR mRNA levels in each genotype contributed to the deficiency in MbR ( Supplementary Fig. 1). As indicated in Fig. 3a, both a monocistronic mbR and a (50-70% less abundant) discistronic mbR-aadA transcript were made in each T 0 tob MbR genotype. In the young upper leaves of T 0 plants at ~21 cm in height the total mbR mRNA abundance was 30-70% lower in abundance than the rbcL mRNA levels in wild-type ( Supplementary Fig. 1). As seen previously in Rubisco-modified tobacco genotypes with reduced photosynthetic potential 8,26,30,32 , these reduced mRNA levels correlate with the impaired viability of the thinner, smaller sized, pale green leaves of each transplastomic genotype (see below).
The evolved MbR have improved carboxylase activity. The catalytic properties of the wild type and mutant L 10 MbR isoforms produced in the T 0 progenies were measured at pH 8 (the approximate pH of the chloroplast stroma, Table 1). While the S C/O values matched those measured for the L 2 enzymes produced in E. coli (Fig. 2b), the k cat C rates were lower than those measured at pH7.2 due to the increased activity of MbR at low pH 26 . Nevertheless, even at pH 8 both k cat C and the carboxylation efficiencies (k cat C /K C 21%O2 ) of the MbR-E138V and MbR-K332E enzymes were between 2 and 3.4-fold higher than MbR, with an accompanying ~15% increase in S C/O for the MbR-E138V enzyme (Table 1). Importantly, these improvements in CO 2 affinity, specificity and fixation speed came without expense to the natural high affinity of MbR for RuBP (i.e. a low K m RuBP , Table 1).
Enhancing MbR catalysis improves tobacco photosynthesis and growth. The improved growth and healthier phenotype of the tob MbRK332E and tob MbRE138V genotypes relative to the tob MbR lines was evident in the T 1 progeny. In tissue culture germination trials all the T 1 progeny emerged as green cotyledons on spectinomycin media after 1 week confirming all were transplastomic (Fig. 4). After 5 weeks it was evident that addition of sucrose to the tissue culture media was required for the germinated tob MbR and tob MbRE138V lines to survive under elevated (2.5% v/v) CO 2 (Fig. 4). In contrast, the tob MbRK332E plants survived under high CO 2 without sucrose supplementation and grew quicker under all tissue culture conditions tested.  (Table 1), the tob MbRE138V and tob MbRK332E genotypes supported faster leaf photosynthetic CO 2 assimilation rates relative to the tob MbR lines (Fig. 5a). To compensate for the lower leaf levels of MbR and the poorer catalytic properties of the L 10 MbR relative to tobacco L 8 S 8 Rubisco (Fig. 3c and Table 1), measurements of photosynthetic CO 2 -assimilation rates in all the MbR transformed leaves were performed under 1% (v/v) O 2 . Even under these low O 2 pressures, photosynthesis remained limited by MbR-activity over the full range of intercellular leaf CO 2 pressures (C i ) tested (Fig. 5a) with the highest assimilation rate of 2.4 μmol CO 2 fixed.m 2 .s −1 measured in tob MbRK332E leaves under the highest leaf gas exchange C i of 2000 μbar CO 2 (Fig. 4b). As this rate is more than 10-fold slower than the 26-30 μmol CO 2 fixed.m 2 .s −1 rates measured in high CO 2 grown wild type leaves 33 the tob MbRK332E grew ~5-fold slower than wild-type under high CO 2 (Fig. 5b,c).
The relative differences in the leaf CO 2 -assimilation rates of each transplastomic genotype (Fig. 5a) correlated with their growth rate (Fig. 5b). The tob MbRK332E plants grew faster and reached fertile maturity before the tob MbRE138V lines while the growth of the tob MbR controls were substantially impaired (Fig. 5b). The faster growth by the tob MbRK332E plants contrasts with the better carboxylation properties of MbR-E138 V relative to MbR-K332E (Table 1). This discrepancy can be attributed to the ~50% lower levels of MbR-E138 V produced in tob MbRE138V leaves relative to MbR levels produced in comparable leaves from both the tob MbR and tob MbRK332E genotypes (Fig. 5a). Identifying if the E138 V mutation impedes the translation, biogenesis or/and stability of MbR remains to be tested.

The structural location of the E138 V and K332E mutations in MbR. Phylogenetic analysis of
MbR reveals it shares closer sequence homology with R. rubrum Form II Rubisco than other archaeal Rubisco (Supplemental Fig. 2) 26 . These alignments showed that E138 and K332 in MbR align with A134 and E331 in R. rubrum Rubisco and R134 and E324 in the T. kodakorensis archaeal L 10 Rubisco (Fig. 6a). As shown in Fig. 6b, A134 in the R. rubrum L 2 crystal structure is solvent exposed and located distil to the active site. In contrast R134 in the T. kodakarensis L 10 structure is one of only ten amino acids that form a highly ionic network between adjoining dimers (i.e. at each L 2 -L 2 interface) 27 .
In both R. rubrum and T. kodakorensis Rubisco, the corresponding E331 and E324 residues are located near the hinge of the conserved flexible loop 6 structure of the C-terminal α /β -barrel (Fig. 6b). A glutamate at this position in loop 6 is highly conserved among photosynthetic L 8 S 8 Rubisco isoforms (e.g. E336 in plants like tobacco, E339 in red algae such as Griffithsia monilis) and is in close vicinity to the strictly conserved K334 catalytic residue (tobacco Rubisco numbering) whose side-chain interactions with RuBP and gaseous substrate are critical determinants of catalytic efficiency (i.e. k cat C and S C/O ) 10 . The increased catalytic turnover rate and  (Table 1) imply a glutamate at this position in loop 6 may benefit Rubisco catalysis in photosynthetic organisms but pose no benefit for the non-photosynthtic role of archaea Rubisco.

Discussion
He we uniquely demonstrate the potential of directed evolution using RDE selection to successfully deliver more efficient forms of the non-photosynthetic M. burtonii archaeal Rubisco (MbR). The derived improvements in CO 2 -fixation speed, CO 2 -affinity and specificity for CO 2 of the evolved MbR-E138 V and MbR-K332E mutant enzymes translated to supporting faster rates of CO 2 assimilation and growth in tobacco relative to the control tob MbR genotype producing wild-type MbR. This finding provides the first proof of concept that directed evolution of non-photosynthetic Rubisco in E. coli can deliver mutants with improvements in all the catalytic parameters needed to stimulate photosynthesis in leaf chloroplasts. This contrasts with prior success in evolving improved catalytic mutants of cyanobacterial Rubisco that either show only marginal (<5%) overall improvements in catalysis 24 or a significant enhancement (>50%) in carboxylation efficiency that came at the expense of an unwanted parallel increase in inhibitory oxygenation efficiency 11 . An additional challenge with cyanobacterial rubrum (Rr) and T. kodakorensis (Tk) Rubisco large subunit sequences adjoining the E138 and K332 mutation sites in MbR. Only amino acids differing from MbR are shown. Secondary structure information is relative to that in Tk L 10 Rubisco 27 . (b) Location of the mutated residues in R. rubrum L 2 Rubisco (A134 and E321: PDB 9RUB) and Tk L 10 Rubisco (R134 and E324: PDB 3A12) are highlighted in cyan. Rr L-subunits are shaded red and green, active site bound RuBP or CABP in yellow. In the Tk structure R134 is located at the L 2 -L 2 interface (differentially coloured red and blue). E321 and E324 are located within the flexible loop 6 region in both Rr and Tk Rubisco respectively. Diagrams constructed using PyMOL.
Scientific RepoRts | 6:22284 | DOI: 10.1038/srep22284 L 8 S 8 Rubisco is its limited biogenesis potential in tobacco chloroplasts (~10% of wild-type 34 ) compared with MbR L 10 which is produced at ~25-50% of wild-type tobacco Rubisco (Fig. 3c). The high solubility and overall success with evolving MbR catalysis inspires continued effort to evolve properties along evolutionary trajectories that further enhance its photosynthetic potential.
Exploration of Rubisco sequence space towards mutations that improve its efficiency in crop plants is an ongoing challenge 13 . Our continued inability after 50 years to rationally predict what sequence changes can improve Rubisco function steered our attention towards the potential of directed evolution to explore Rubisco sequence space for improved catalysis. A common requirement of successful directed evolution studies is identifying a suitable starting point for mutagenesis and appropriate selection system 14,17,35 . The ease by which the carboxylase activities of MbR could be enhanced by single amino acid changes (Table 1) likely stems from it having undergone specialisation to an alternative metabolic role during its non-photosynthetic evolution 29 . This implies that archaeal Rubisco may occupy an alternative position to photosynthetic Rubisco within the evolutionary landscape of sequence space diversity in relation to catalysis. Consistent with this, archaeal Rubisco catalysis is typically distinct relative to contemporary (photosynthetic) Rubisco 26 . Archaeal Rubisco can sustain functionality at extreme temperatures, under which thermotolerant archaea grow, and exhibit the heightened affinity for RuBP required to metabolise the finite levels made during nucleotide metabolism 29 . Offsetting these beneficial features, archaeal Rubisco show low k cat C and S C/O 26 . Improving our fundamental understanding of the structural features that determine these unconventional kinetics requires a more comprehensive survey of archaeal Rubisco structure, catalytic, and sequence diversity.
The merits of directed protein evolution are illustrated by the many examples of successfully altering protein solubility, improving catalysis, even enabling promiscuous catalytic function 12,14,15,17 . These outcomes typically require multiple, incremental rounds of mutagenesis. For MbR, an attempt was made to select second generation mutants with improved activity using a combined library of epPCR mutated mbR genes coding MbR, MbR-E138V and MbR-K332E. No MbR mutants were detected that enabled MM1-prk RDE survival under high PRK induction (i.e. 0.2% w/v arabinose). Future goals are to examine the feasibility of capturing the epistasis of the spectrum of first generation mbiiL mutant genes (Supplementary Table 1) using a shuffling approach to identify adaptive trajectories that further improve MbR catalysis. Success however depends on the positive epistatic potential for evolving the carboxylase activity of MbR and ability to avoid or circumvent beneficial mutations that might produce destabilizing effects on structure and function. Such uncertainties are common to directed protein evolution studies. Forecasting the extent to which mutations (via direct or long distance amino acid interactions) influence artificial evolutionary trajectories remains unpredictable 36 .
A significant hurdle is the relatively low selection fidelity and throughput of the MM1-prk selection system 20 . The frequency of success in directed evolution applications depends on the library selection throughput and sensitivity of the selection system to detect a desired trait 14,15 . The reliance and throughput of existing RDE strains suffer from high frequencies of false positives that typically arise through transposon associated PRK escape mutations 11,12 . While relatively immune to false positives, the MM1-prk RDE selection throughput is impeded by a low growth temperature requirement (25 °C), poor transformation efficiency, and reduced cell viability as a result of the gapA − mutation 20 . Improving the selection fidelity of RDE systems is therefore critical to further evolving MbR, and other Rubisco isoforms, with improved photosynthetic properties. One solution might be to tether PRK with an antibiotic resistance protein in an RDE strain thus avoiding selection of "PRK-silenced" false positives as such mutations would also relinquish antibiotic resistance.
Adaptive evolution of archaeal Rubisco in vitro towards one that is more efficient than crop plant L 8 S 8 enzymes is undeniably a significant, long term challenge. Unlike L 8 S 8 Rubisco from plants and algae, the folding and assembly requirements of archaeal Rubisco, like MbR, are met in E. coli (Fig. 1c) 26,27,29 . This property strengthens the suitability of archaeal Rubisco for identifying catalysis enhancing mutants using RDE strains as it curtails selection of mutations that enhance solubility, an outcome that has dominated directed evolution studies with cyanobacteria L 8 S 8 Rubisco 11,12,22 . The amenability of archaeal Rubisco to mutational testing in E. coli has already proven useful to demonstrate how incorporating spinach Rubisco sequence into T. kodakarensis L 10 Rubisco can improve k cat C 27 . Improving archaeal Rubisco catalysis by rational design or directed evolution or a combination of both therefore poses viable future pathways to pursue, particularly given our finding that these benefits can directly translate to improving leaf photosynthesis.
As indicated in Fig. 6, the primary sequences of archaeal Rubisco are highly diverse and their oligomer structures as L 2 or L 10 appears variable. While mass spectrometry analysis infers a mature L 10 quaternary structure for MbR 26 it is uncertain if it forms a comparable toroidal structure to T. kodakarensis archaeal Rubisco (PDB: 1GEH), in particular since they only share 36% amino acid homology and MbR contains a novel 11 amino acid insertion in its C-domain 26 . Ongoing efforts are focused on solving the crystal structures for both L 2 and L 10 MbR to better understand the structural diversity among archaeal Rubisco as well as help interpret how mutations, such as E138V and K332E, functionally impart changes to catalysis.

Materials and Methods
Evolution, expression and purification of MbR in E. coli. The mbiiL gene from pHUE-mbiiL 26 was cloned into pTrcHisB using NcoI/HindIII and the resulting pTrcMbR plasmid used as template to randomly mutate the mbiiL by error-prone PCR (epPCR) as described 20 . The PCR products were cloned into pTrcHisB and the diversity of the mbiiL mutant library calculated using PEDEL-AA 37 . The library was transformed into the Rubisco Dependent E. coli (RDE) strain MM1-prk and grown under varying selective conditions according to 20 . The mutant mbiiL genes from faster growing colonies were cloned into the 6xhistidine-tagged ubiquitin expression plasmid pHUE and each MbR isoform affinity purified by immobilised metal affinity chromatography (IMAC) as described 26 .
Scientific RepoRts | 6:22284 | DOI: 10.1038/srep22284 Tobacco plastome transformation and growth. A synthetic gene, mbR, coding for M. burtonii Rubisco both with and without mutations coding E138 V or K332E substitutions was synthesised by GenScript. The codon use of mbR matched the tobacco rbcL gene and replaced the native N-terminal coding sequence (MSLIYEDLV) with that for the native tobacco Rubisco large (L-) subunit (MSPQTETKASVGF). The 1,418-bp mbR gene fragments were cloned into NheI/SalI cut pLEV4 32 to produce the plastome transforming plasmids pLEVmbR, pLEVmbR-K332E and pLEVmbR-E138 V and transformed into ten cm trL leaves by biolistic bombardment 31 . Independent positively transformed lines producing L 10 MbR were identified by non-denaturing PAGE (native PAGE) 26 and two independent lines for each MbR genotype were grown to maturity in soil in a growth chamber with the air supplemented with 2.5% (v/v) CO 2 30 . The flowers of the fertile T 0 plants were fertilised with wild type pollen and the seed germinated in tissue culture on RMOP media supplemented with 0% to 3% (w/v) sucrose 30 . The germinated T 1 progeny were carefully transferred to soil and the leaf gas exchange and cell biochemistry of near fully expanded leaves at comparable positions in the upper canopy analysed when the plants were 20-25 cm in height.
DNA, protein and PAGE analyses. Total leaf genomic DNA was isolated using the DNeasy ® Plant Mini Kit and primers LSH and LSE (Fig. 3a) used to PCR amplify and sequence the plastome region transformed in each tobacco genotype as described 32 . The preparation, quantification (against BSA) of soluble leaf protein and analysis by SDS-PAGE, native PAGE and immunoblot analysis was performed as described 38 .
Rubisco content and catalysis. Rates of Rubisco 14 CO 2 fixation were made using soluble protein extracts isolated from bacteria or leaf protein in 50 mM HEPES-NaOH (pH 7.2 or 8.0) containing extraction buffer as described 30 . Protein extract (20 μL) was used to initiate activity in 0.5 mL assays performed in 7 mL septum-capped scintillation vials 38 . Each sample was measured in duplicate under varying concentrations of NaH 14 CO 3 (0-67 μM) and O 2 (0, 2, and 5% (v/v)) to calculate the maximal rate of carboxylation (V C ) and the Michaelis constants (K m ) for CO 2 (K C ) and O 2 (K O ) 38 . The carboxylation turnover rate (k cat C ) was calculated by dividing V C by the Rubisco active sites content quantified by [ 14 C]-2-CABP binding 30 . Rubisco CO 2 /O 2 specificity (S C/O ) and the K m for RuBP were quantified as described 38 using MbR purified from E. coli by immobilised metal affinity chromatography 26 or from tobacco leaves by ion exchange 30 . Growth and photosynthesis analysis. All plants were grown at 25 °C in a growth chamber as described 30 under 200 ± 50 μmol quanta.m 2 .s −1 in air containing 2.5% (v/v) CO 2 . Once approximately 21 cm in height the leaf photosynthesis rates (A) in the 5 th upper canopy leaf were measured using a LI-6400 XT gas exchange system (LI-COR) at varying atmospheric CO 2 partial pressures (C a ; 50-2000 ppm) at a constant leaf temperate of 25 °C and 1000 μmol quanta.m 2 .s −1 . The "A-C i " measurements (C i ; leaf intercellular CO 2 levels) were performed at low O 2 partial pressures (1% (v/v) O 2 in N 2 ) to obtain suitable measures of A.