Isoleucine 309 acts as a C₄ catalytic switch that increases ribulose-1,5-bisphosphate carboxylase/oxygenase (rubisco) carboxylation rate in Flaveria

Abstract

Improving global yields of important agricultural crops is a complex challenge. Enhancing yield and resource use by engineering improvements to photosynthetic carbon assimilation is one potential solution. During the last 40 million years C₄ photosynthesis has evolved multiple times, enabling plants to evade the catalytic inadequacies of the CO₂-fixing enzyme, ribulose-1,5-bisphosphate carboxylase/oxygenase (rubisco). Compared with their C₃ ancestors, C₄ plants combine a faster rubisco with a biochemical CO₂-concentrating mechanism, enabling more efficient use of water and nitrogen and enhanced yield. Here we show the versatility of plastome manipulation in tobacco for identifying sequences in C₄-rubisco that can be transplanted into C₃-rubisco to improve carboxylation rate (V_C). Using transplastomic tobacco lines expressing native and mutated rubisco large subunits (L-subunits) from Flaveria pringlei (C₃), Flaveria floridana (C₃-C₄), and Flaveria bidentis (C₄), we reveal that Met-309-Ile substitutions in the L-subunit act as a catalytic switch between C₄ (³⁰⁹Ile; faster V_C, lower CO₂ affinity) and C₃ (³⁰⁹Met; slower V_C, higher CO₂ affinity) catalysis. Application of this transplastomic system permits further identification of other structural solutions selected by nature that can increase rubisco V_C in C₃ crops. Coengineering a catalytically faster C₃ rubisco and a CO₂-concentrating mechanism within C₃ crop species could enhance their efficiency in resource use and yield.

The future uncertainties of global climate change and estimates of unsustainable population growth have increased the urgency of improving crop yields (1). One possible solution is to “supercharge” photosynthesis by improving the C₃ cycle (2, 3). Although a simple idea, this is a complex challenge that involves several possible alternatives. Many of these alternatives focus on enhancing the performance of the CO₂-fixing enzyme ribulose-1,5-bisphosphate (RuBP) carboxylase/oxygenase (rubisco), which catalyses the first step in the synthesis of carbohydrates. Despite its pivotal role, rubisco is a slow catalyst, completing only one to four carboxylation reactions per catalytic site per second in plants (4, 5). Moreover CO₂ not only is fixed through a complex catalytic process but also must compete with O₂. The oxygenation of RuBP produces 2-phosphoglycolate, whose recycling by photorespiration requires energy and results in the futile loss of fixed carbon [∼30% of fixed CO₂ in many C₃ plants (6)].

To compensate for rubisco's catalytic limitations, plants invest as much as 25% of their leaf nitrogen in rubisco (7). This value is much lower in C₄ plants, where a biochemical CO₂-concentrating mechanism (CCM) elevates CO₂ around rubisco. This optimized microenvironment allows rubisco to operate close to its maximal activity, reducing O₂ competition. This CCM has enabled C₄ plants to evolve rubiscos with substantially improved carboxylation rates (V_C) relative to their C₃ ancestors, albeit at the expense of reducing CO₂ affinity [i.e., a higher apparent K_m for CO₂ (K_C)] (8, 9). As a consequence, C₄ plants require less rubisco, thereby enhancing nitrogen use with improvements in V_C correlating with improved efficiency in nitrogen use (10). In addition, the high concentration of CO₂ around rubisco allows C₄ plants to operate at lower CO₂ pressures within their leaf air spaces, thereby reducing their stomatal conductance requirements and the associated H₂O loss by transpiration. Indicative of these growth advantages, C₄ photosynthesis has evolved independently several times from multiple C₃ lineages during the last 20–40 million years (11, 12).

Following nature's example, a number of CO₂ transgenic approaches have been designed to emulate CCM strategies in C₃ plastids and improve rubisco performance (2). These approaches include elevating the CO₂ concentration within chloroplasts using recombinant CO₂/HCO₃⁻ transporters from cyanobacteria or engineering alternative pathways to bypass photorespiration and release CO₂ within the stroma (13, 14). Although each strategy faces continuous challenges in its fine tuning and integration into crops, further improvements in yield and in the efficiency of water and nitrogen use are likely by concurrently “speeding up” rubisco (9).

Identifying the natural changes that result in the faster C₄ rubiscos is far from simple, given the complex structure and biogenesis pathway of the hexadecameric rubisco (L₈S₈) in vascular plants, whose assembly requirements cannot be met by conventional bacterial or in vitro expression systems (15). The catalytic core of L₈S₈ rubisco comprises four 52-kDa large (L)-subunit pairs which are capped by two sets of 15-kDa small (S)-subunit tetramers that provide structural stability and influence catalysis (16, 17). Although supplementing rice rubisco with S-subunits from the C₄ plant sorghum was found to improve V_C of the heterologous L₈S₈ enzyme (16), crosses between C₃ and C₄ Flaveria and Atriplex species showed C₄ catalysis to be maternally inherited (18, 19) and hence defined by the chloroplast-encoded L-subunit gene (rbcL). Therefore, changes in both Rubisco L- and S-subunits can influence catalysis.

Although phylogenetic studies have identified potential L-subunit residues involved in the transition from C₃-like to C₄-like rubisco, it is uncertain which residues are catalytically determinant (11, 20). Here we undertake a transplastomic approach to identify such residues in vivo. By manipulating the rbcL gene in tobacco to produce hybrid L₈^FS₈^t rubiscos (containing variant Flaveria L- and tobacco S-subunits), we demonstrate that Met-309-Ile substitutions in the L-subunit act as a catalytic switch between C₄ (³⁰⁹Ile) and C₃ (³⁰⁹Met) catalysis in Flaveria rubisco.

Results

Flaveria Rubisco L-Subunit Expression in Tobacco Chloroplasts.

The rbcL genes from Flaveria pringlei (C₃), Flaveria floridana (a C₃-C₄ intermediate), and Flaveria bidentis (C₄) were chosen for transformation into tobacco plastids because of their diverse catalytic properties (ref. 21 and below) despite their high sequence similarity (Fig. 1A). As in all Flaveria rbcL genes, nonsilent nucleotide changes occur only at residues 149, 265, and 309 (20). The Flaveria L-subunits show >95% identity with the tobacco L-subunit, with 22–24 amino acid differences in addition to a highly charged TDKDKDKKR extension at the C terminus (Fig S1). The Flaveria rbcL genes were cloned into the plastome-transforming plasmid pLEV4, where the expression of the transgenes is regulated by the native tobacco rbcL gene regulatory sequences [i.e., its promoter, 5′-, and 3′-untranslated sequences (22)]. The transforming plasmids, including the control pLEV4, were introduced biolistically into ^cmtrL, a tailor-made tobacco master line for integrating rbcL transgenes into the tobacco chloroplast by homologous recombination (Fig. 1B) (23). Two independent transplastomic lines for each rbcL transgene were grown to maturity in soil in air supplemented with 0.5% (vol/vol) CO₂. The transformed tobacco lines that incorporated the F. floridana, F. bidentis, and F. pringlei rbcL genes were called “tob^flo,” “tob^bid,” and “tob^pring,” respectively.

Fig. 1.

Flaveria rubisco L-subunit sequence and expression in tobacco chloroplasts. (A) Comparison of Flaveria rbcL sequences that differ only in L-subunit substitutions at amino acids 149, 265, and 309 (20). (B) The transforming plasmid pLEV4 contains a homologous plastome flanking sequence (indicated by dashed lines; numbering indicates region of sequence integration relative to plastome sequence; GenBank ID Z00444) that directed integration of the rbcL transgenes and a promoterless aadA-selectable marker gene into the plastome of the tobacco master line, ^cmtrL (23). The L-subunit amino acid differences at residues 149 (white triangle), 265 (black triangle), and 309 (gray triangle) in the tob^LEV4 (tobacco rbcL control), tob^pring (F. pringlei rbcL), tob^flo (F. floridana rbcL), tob^bid (F. bidentis rbcL), tob^flo-bid (chimeric F. floridana–F. bidentis rbcL), tob^bid-flo (chimeric F. bidentis–F. floridana rbcL), and tob^149A,265I (mutated tobacco rbcL) transplastomic tobacco (tob) lines are shown. Annealing locations of primers LSH, LSE, and the 5′ ΔrbcL probe (24) are shown. N, NheI; S, SalI cloning sites. (C) Nondenaturing PAGE analysis of soluble protein from comparable leaves of independent T₀ transplastomic lines, ^cmtrL, and wild-type tobacco (protein from 1.5 mm² of leaf was loaded per lane). Homoplasmic transformants produce only L₈S₈ rubisco (∼520 kDa) and not the ∼100-kDa R. rubrum L₂ rubisco produced in ^cmtrL (23). (D) Detection of rbcL and rbcL-aadA mRNA in total RNA from 3 mm² (for wild type) or 6 mm² (other samples) of the leaves sampled in C.

Nondenaturing PAGE (ndPAGE) analysis of the soluble leaf protein was used to confirm the production of the hybrid L₈^FS₈^t rubisco (comprising Flaveria L-subunits and tobacco S-subunits) and to assesses the homoplasmicity of the tob^pring, tob^flo, and tob^bid lines (Fig. 1C). Each of the transformed lines examined was deemed to be homoplasmic, because no L₂ Rhodospirillum rubrum rubisco from the parental ^cmtrL line was detected; homoplasmicity was further confirmed by DNA blot analysis (Fig S2).

Differential Expression of the Hybrid L₈^FS₈^t Rubiscos.

Differences in the intensities of the L₈^FS₈^t bands in ndPAGE indicated that the amounts of rubisco produced in the T₀ lines varied (Fig. 1C). This variation was confirmed by quantitative [¹⁴C]2-carboxyarabinitol-1,5-bisphosphate (CABP) binding. The amount of L₈^FS₈^t produced in the tob^flo, tob^bid, and tob^pring leaves was reduced by approximately 50%, 65%, and 75% relative to wild-type, respectively (Fig. 1C). In contrast, rubisco content in the tob^Lev4 control transformants matched that in wild-type, indicating that the additional genome changes around rbcL were not the cause of the reduced L₈^FS₈^t expression. SDS PAGE-immunoblot analysis showed no unassembled Flaveria L-subunits accumulated as insoluble aggregates.

Contrary to the reduced L₈^FS₈^t content in the T₀ leaves, there was little or no difference in the Flaveria rbcL mRNA levels in the same leaves relative to wild type (Fig. 1D), indicating that L₈^FS₈^t synthesis probably is perturbed posttranscriptionally. As shown previously (23–25), a less abundant rbcL-aadA dicistronic mRNA (∼10% that of the rbcL mRNA) was produced in all transformants as a result of inefficient transcription termination by the tobacco rbcL 3′UTR. The stages that hinder L₈^FS₈^t expression during rubisco biogenesis or degradation remain to be identified fully.

Plant Growth and Leaf Photosynthesis.

The disparity in L₈^FS₈^t levels in leaves of the tob^flo, tob^bid, and tob^pring lines persisted to the T₂ progeny and correlated with their relative differences in photosynthesis and growth rates. For the tob^flo and tob^bid plants, the higher L₈^FS₈^t levels produced were sufficient for them to survive through to maturity in air (without CO₂ enrichment), although they grew more slowly than wild-type plants (Fig. 2A). In contrast, the tob^pring transformants grew poorly in air. As seen previously in tobacco^Rst lines producing hybrid L₈^sS₈^t rubisco (comprising sunflower L- and tobacco S-subunits) (24), the juvenile tob^pring plants displayed a pale green leaf phenotype with marginal curling and dimpling (Fig. 2A Right). This phenotype is likely a consequence of the very low L₈^FS₈^t content during early vegetative growth (<3 μmol sites m⁻²·s⁻¹) and in the young mature leaves (<6 μmol sites m⁻²·s⁻¹; Fig. 2C) of the tob^pring plants. Consistent with their different growth rates and varied L₈^FS₈^t contents, the leaf photosynthetic CO₂ assimilation rates at varying CO₂ partial pressures (pCO₂) were slowest for tob^pring [but still slightly higher than in tobacco^Rst (24)] and were successively better for the tob^bid and tob^flo lines, albeit still slower than in wild-type tobacco (Fig. 2C). Measurements of the ratio of variable fluorescence to maximal fluorescence (F_v/F_m) in wild-type leaves (0.82 ± 0.01) were identical to those in the three tob^Flaveria genotypes (0.83 ± 0.01), indicating no difference in photochemical efficiency under the growth conditions.

Fig. 2.

Measurements of growth and leaf gas exchange in the transformants producing the variant L^F₈S^t₈ rubiscos. (A) (Left) Comparatively slower growth in air of the tob^Flaveria transformants as a function of plant height relative to wild-type. (Right) The tob^pring lines grew extremely poorly in air. pce, postcotyledon emergence. (B) Phenotype of the plants from A at age pce as shown. (Scale bars: 4 cm.) Air supplemented with 0.5% (vol/vol) CO₂ was used to grow tob^pring plants to maturity. (C) Comparative differences in gas-exchange measurements of CO₂ assimilation rates at 25 °C under varying chloroplast CO₂ pressures (C_c) at growth illumination (400 μmol quanta m⁻²·s⁻¹). Measurements were made on young mature leaves located at similar canopy positions (fifth leaf from the apical meristem) of physiologically comparable mature plants analogous to those shown in B. Leaf rubisco contents were 25.0 and 30.5; 12.4 and 11.9; 10.1 and 10.8; 4.1 and 4.3 μmol rubisco sites m⁻² in the independent wild-type (circles), tob^flo (squares; line 1 white, line 2 black), tob^bid (triangles; line 1 white, line 2 black), and tob^pring (diamonds; line 1 white, line 2 black) plants analyzed, respectively. The lines show the rubisco limited CO₂ assimilation rates for wild-type (– –), tob^flo (—),tob^bid (– · –), and tob^pring (····) modeled according to ref. 26 using the catalytic parameters for the respective hybrid L^F₈S^t₈ rubiscos in Fig. 3 and assuming rubisco was fully activated and a value of 0.3 mol m⁻²·s⁻¹·bar⁻¹ for mesophyll conductance.

Catalysis by Each L₈^FS₈^t Rubisco Matches the Source Flaveria Enzyme.

The catalytic properties of the recombinant L₈^FS₈^t were compared with the source L₈^FS₈^F enzymes from the corresponding Flaveria species (Fig. 3). As seen previously for the L₈^sS₈^t rubisco produced in tobacco^Rst (24), the L₈^FS₈^t and equivalent Flaveria L₈^FS₈^F enzymes were catalytically comparable with respect to their carboxylation (V_C) and oxygenation (V_O) rates, their apparent Michaelis constants (K_m) for CO₂ (K_C) and O₂ (K_O), and CO₂/O₂ specificities (S_C/O). Of particular interest was the faster carboxylation rate (∼35% higher V_C relative to the C₃ rubiscos) and lower CO₂ affinity (∼50% higher K_C) of the hybrid L₈^FS₈^t rubisco from tob^bid that matched the F. bidentis L₈^FS₈^F enzyme (Fig. 3). As expected, when the catalytic properties and content of each L₈^FS₈^t hybrid enzyme were used to model CO₂ assimilation rates according to Farquhar et al (26), the final values closely matched those measured by whole-leaf gas exchange (Fig. 2C). These results provide confidence both in the accuracy of the measured catalytic properties of the L₈^FS₈^t hybrid enzymes and the rubisco-limited CO₂-assimilation model.

Fig. 3.

Comparative catalysis at 25 °C of the wild-type and Q¹⁴⁹A-V²⁶⁵I mutated tobacco (L^t₈S^t₈) rubisco, the source Flaveria (L^F₈S^F₈) rubiscos, the hybrid (L^F₈S^t₈, comprising Flaveria L- and tobacco S-subunits), and chimeric (L^cF₈S^t₈, comprising chimeric Flaveria L- and tobacco S-subunits) rubisco variants produced in the transplastomic tobacco plants. The L-subunit amino acid residues at codons 149, 265, and 309 in each rubisco type are shown. Values shown are the average ± SD of independent assays (n = 4–24; see Materials and Methods for details). The maximal oxygenation rate (V_O) was calculated using the equation S_C/O = (V_C/K_C)/(V_O/K_O).

Interchanging C₃–C₄ Catalysis via ³⁰⁹Met-³⁰⁹Ile Substitutions in Flaveria Rubisco.

The catalytic similarity between native F. bidentis rubisco and the hybrid tob^bid L₈^FS₈^t enzyme indicated that the introduced L-subunit determined the catalytic phenotype. Because the F. bidentis and F. floridana L-subunits differ only at residues 149 and 309 but show significant differences in V_C and K_C (Fig. 3), domain swapping of their rbcL was used to identify which residue(s) imparted the C₄ catalysis of F. bidentis rubisco. The chimeric rbcL gene in the transforming plasmid pLEV^flo-bid introduced a Met-309-Ile substitution into the F. floridana rbcL gene, whereas in pLEV^bid-flo the chimeric rbcL gene coded an Ile-309-Met substitution in the F. bidentis rbcL gene (Fig. 1B). Both plasmids were transformed into ^cmtrL, and independent tob^flo-bid- and tob^bid-flo-transformed tobacco lines were produced that grew to reproductive maturity in air.

Catalytic analysis of the chimeric (L₈^cFS₈^t) rubisco produced in the tob^flo-bid T₁ progeny showed that the Met-309-Ile substitution increased V_C and K_C, matching that measured for the F. bidentis and tob^bid C₄-like rubiscos (Fig. 3). In contrast, introducing an Ile-309-Met mutation into the F. bidentis L-subunit (tob^bid-flo lines) reduced V_C and K_C, resulting in a rubisco with C₃-like catalysis. These results demonstrate that ³⁰⁹Ile confers Flaveria rubisco with C₄-like catalysis. Although a comparison of higher plant L-subunits shows that ³⁰⁹Met is highly conserved in most C₃-plant rubiscos, the tobacco L-subunit encodes ³⁰⁹Ile (Fig. 4A).

Fig. 4.

Conservation and location of the L-subunit residues 149 and 309 in higher plant rubisco. (A) Coding matrix summary of ClustalW-aligned residues 300–319 in L-subunit sequences from data sets with corresponding C₃ and C₄ speciation data (11, 20, 38). (B) Structure of spinach L₈S₈ rubisco and L₈ core (L-subunits are shown in green; S-subunits are shown in blue) viewed from the top, showing central solvent channel (Left and Center), and from the side (Right). The relative locations of ¹⁴⁹Gln (white triangle) and ³⁰⁹Met (yellow triangle) in one L-subunit pair (L₂) is shown. The ³⁰⁹Met residues are located at the interface of L-subunits; the ¹⁴⁹Gln residues are positioned at the L₂–L₂ interface toward the surface of the central solvent channel. (C) View of an L-subunit pair (L₁ in green showing ribbon structural detail; L₂ in blue) showing the positioning of ¹⁴⁹Gln and ³⁰⁹Met relative to each Mg²⁺ (black sphere) and the reaction-intermediate analog 2-CABP (yellow and red ball and stick) bound to the two active sites in the dimer. The conserved active-site residues are shown for one active site. Distances (in Å) from the S atom of ³⁰⁹Met in L₂ to each Mg²⁺ and the Cα atom of the closest conserved active-site residue, ²⁹⁵Arg, are shown. (This figure was prepared with PyMOL using the PDB co-ordinates 8RUC.)

Amino Acid 149 Is Catalytically Neutral but Can Influence Rubisco Expression.

The matching C₄-like catalysis of rubisco from tob^flo-bid and tob^bid and C₃-like catalysis of the tob^bid-flo, tob^flo, and tob^pring rubiscos suggests that changes to amino acid 149 in Flaveria rubisco are catalytically neutral and possibly account for the amino acid heterogeneity at this position (Fig. 1A) (20). Likewise, conservation of ²⁶⁵Ile in F. floridana and F. bidentis rubisco indicates that this residue also is catalytically neutral. The influence of Gln-149-Ala and Val-265-Ile L-subunit substitutions (to match those in F. bidentis rubisco; Fig. 1B) on tobacco rubisco were tested by transforming ^cmtrL with the pLEV^149A,265I. Catalysis by rubisco in the tob^149A,265I lines matched that of the wild-type enzyme, demonstrating that both substitutions are catalytically neutral and are not able to impart C₄-like catalysis on tobacco rubisco (Fig. 3).

Despite the apparent neutrality of amino acid 149 on catalysis, changes at this position affected the level of the hybrid L₈^FS₈^t expression. In young mature leaves of both tob^flo-bid and tob^flo, whose L-subunits share ¹⁴⁹Asp (and ²⁶⁵Ile), the rubisco levels were comparable (10–13 μmol sites m⁻²·s⁻¹). In contrast, the rubisco content in equivalent leaves from tob^bid-flo and tob^bid (whose L-subunits code ¹⁴⁹Ala and ²⁶⁵Ile) were lower (6–8 μmol sites m⁻²·s⁻¹) (Fig S3). These results suggest that changes to the amino acid (or its mRNA sequence) at residue 149 might be responsible for the variations in hybrid rubisco expression. However, this did not appear to be the case for tobacco Rubisco as the leaf Rubisco levels in the tob^149A,265I lines matched that in the wild-type leaves (Fig. 1C). How changes at residue 149 in the Flaveria L-subunit might differentially influence its translation, folding, and/or assembly with tobacco S-subunits or the stability of L₈^FS₈^t complexes remains to be examined.

Discussion

Using transgenic tobacco lines expressing hybrid rubiscos containing Flaveria L- and tobacco S-subunits (L₈^FS₈^t), we have identified ³⁰⁹Ile as the key residue that imparts C₄-like catalytic properties to Flaveria rubisco. The determinant role of this residue supports observations from prior crossing studies that showed C₄ catalysis to be maternally inherited in Flaveria (19). Linkages between catalysis and sequence phylogenies of different Flaveria rubisco L- and S-subunits indicated that C₄ catalysis was associated with two positively selected L-subunit amino acid substitutions: Asp-149-Ala and Met-309-Ile (20). Here we show that amino acid differences at position 149 in Flaveria rubisco probably are catalytically silent, because interchanging ¹⁴⁹Ala with ¹⁴⁹Asp in the L-subunit from either F. floridana or F. bidentis rubisco had no influence on catalysis (Fig. 3). Similarly, tobacco rubisco catalysis was unaffected by Q149A and V265I substitutions (Fig. 3).

A structural/functional explanation for how ³⁰⁹Ile increases V_C in Flaveria rubisco is unclear. The similar positioning of conserved catalytic site residues in the crystal structures of catalytically different rubiscos makes it difficult to rationalize how distant changes influence catalysis (4, 20). The structure for spinach L₈S₈ rubisco (Fig. 4B) shows ³⁰⁹Met located at the L-interface (i.e., between the L-subunits in each L₂ dimer) more than 17 Å away from the Mg²⁺ bound to each catalytic site and at least 13 Å away from the nearest conserved active site residue, ²⁹⁵Arg (Fig. 4C). In contrast, ¹⁴⁹Gln is located further away from the active sites and close to the interface of the adjoining L₂ dimers that form the L₈ core (Fig. 4B). In the absence of a crystal structure for a Flaveria rubisco, it is difficult to explain how insertion of a more hydrophobic ³⁰⁹Ile this far from the active site might influence V_C (20).

Despite the improvements in V_C imparted by ³⁰⁹Ile in hybrid L₈^FS₈^t rubisco, the accompanying reductions in CO₂ affinity (i.e., increased K_C) precluded gains in carboxylation efficiency. At 25 °C under ambient oxygen levels, the carboxylation efficiency (i.e., V_C/K_C^21%O2) of the C₄-like ³⁰⁹Ile-containing rubiscos in tob^bid and tob^flo-bid (145 mM⁻¹·s⁻¹) were poorer than the C₃-like ³⁰⁹Met-L₈^FS₈^t enzymes in tob^pring and tob^flo (150 and 161 mM⁻¹·s⁻¹, respectively). Thus, because of the low CO₂ levels within (unstressed) C₃ chloroplasts (<10 μM), the faster C₄-like enzymes probably provide no advantage to plant growth within a C₃ plant (at least at 25 °C), as shown recently in rice (16). As modeled recently, optimal CO₂ concentrations required for C₄ rubisco are substantially higher (∼80 μM) (8), indicating that taking full advantage of a faster rubisco in a C₃ plant will require the combinatorial effect of a suitable CCM, for which a number of strategies are being pursued (2). Some of these approaches already have demonstrated that elevating CO₂ pressures within C₃ plastids can improve the capacity for CO₂ assimilation by reducing the energy costs of photorespiration (13, 14).

Our results suggest that the carboxylation rate of rubisco in a C₃ plant might be increased either by direct replacement with L-subunits sourced from C₄ plants (as in the tob^bid plants; Fig. 3) or by tailoring appropriate sequence mutations into related C₃ rubisco L-subunits (as in the tob^flo-bid plants; Fig. 3). Although the first approach suffers from our inability to predict a priori the assembly properties of foreign rubiscos within the chloroplast of the recipient transplastomic line, the second approach would require knowing the catalytic/structural effect of every possible mutation within the context of a particular rubisco enzyme, an understanding that we are still far from achieving. Indeed, the finding that tobacco rubisco encodes ³⁰⁹Ile but shows C₃ catalysis highlights the complex natural variation in the sequence–structure–function relationships among plant L-subunits. Even the sequence diversity at position 309 among C₄ rubiscos (Fig. 4A) indicates that this residue is not the only one that can impart C₄ catalysis. This result is consistent with the polyphyletic evolution of C₄ photosynthesis (12) and with predictions that at least eight L-subunit residues (including residue 309 but not residue 149) have been selected for positively by C₄ catalysis (11). Experimentally testing these predictions, identifying other catalytically determinant L-subunit residues, and exploring which particular rubiscos are affected by these changes have been hampered by the preferential location of rbcL in the plastome (27) and the small range of species whose plastomes can be transformed stably (28). However, as shown in this and previous studies (24, 29, 30), these experimental limitations may be circumvented by expressing hybrid rubiscos in tobacco plastids. The generality of this system for examining sequence–performance relationships within otherwise inaccessible, catalytically diverse foreign L-subunits remains to be explored fully.

Although this study demonstrates the pervasive role of the L-subunit in shaping catalysis in plant rubisco, the important role of the S-subunits on catalysis cannot be overlooked. The apparent catalytic neutrality of the tobacco S-subunit when assembled with heterologous L-subunits (Fig. 3) (24, 29, 30) contrasts with the recent success in shaping rice rubisco toward C₄-like catalysis using heterologous S-subunits from C₄ sorghum (16). Similarly, structural changes to the S-subunit have improved Chlamydomonas rubisco catalysis (17). As highlighted recently (20), differences in rubisco S-subunit sequence also may account for the catalytic deviation of Flaveria palmeri rubisco, whose L-subunit sequence matches that of tob^flo-bid (coding ³⁰⁹Ile) but shows C₃ catalysis.

The similar S_C/O values determined for rubisco from F. bidentis, F. floridana, and F. pringlei (Fig. 3) in this study contrast with the slightly varying values determined previously (S_C/O = 76 ± 1, 84 ± 1, and 81 ± 1 respectively) (21). The reason for this variation is unknown but may lie in alterations in the catalytic competence as a result of different purification processes (ion exchange chromatography versus ammonium sulfate fractionation), the final enzyme purity, and the length of ultra-cold storage (24). By using fresh rubisco rapidly purified to >95% homogeneity by ion exchange chromatography, our measured S_C/O values were highly reproducible between preparations from independent biological replicates.

Here we present an rbcL engineering approach involving hybrid rubisco production in tobacco plastids to unravel the sequence and catalytic diversity of related C₃ and C₄ rubiscos from Flaveria. Future applications of this experimental system are focused on identifying sequence changes that account for the natural diversity of rubisco performance and testing the feasibility of transplanting these catalytic improvements into the rubisco L-subunits of agriculturally relevant crops. In particular, when coengineered with biotechnological strategies to elevate CO₂ around rubisco in C₃ plants, a faster rubisco may translate into improved efficiency in water and nitrogen use and the enhanced yields currently enjoyed by C₄ plants.

Materials and Methods

RuBP and [¹⁴C]2-CABP were synthesized as described (31, 32). Protein content was measured using a dye-binding assay (Pierce) and BSA as a protein standard.

Tobacco Plastome Transformation and Growth.

The transforming plasmid pLEV4 directs the insertion of an rbcL gene and a promoterless aadA gene (coding spectinomycin resistance) into the tobacco plastome in place of the L₂ Rhodospirillum rubrum rubisco-coding ^cmrbcM gene in the plastome of the tobacco master line ^cmtrL (Fig. 1B) (23). The rbcL gene from F. bidentis, F. pringlei, and F. floridana was PCR amplified from leaf genomic DNA [isolated using the DNeasy Plant Mini Kit (Qiagen)] with the primers 5′ NheIrbcL (5′-AGCTAGCGTTGGATTCAAAGCTGGTGTT-3′ [the NheI site that spans the rbcL codons 9 (Ala) and 10 (Ser) is shown in italics] and 3′ SalIrbcL (5′-TGTCGACTGTTTTTATCTCTTCTTATCCTTATCCT-3′ [the reverse complement of the rbcL stop codon is shown in bold, and the SalI site is shown in italics]. The 1,439-bp NheI–SalI rbcL products were cloned into pLEV4 to give the transforming plasmids pLEV^pring, pLEV^flo, and pLEV^bid. The plasmids pLEV^flo-bid and pLEV^bid-flo were made by interchanging the 569-bp SphI–SalI fragments of the F. bidentis and F. floridana rbcL genes (Fig. 1B). Mutations coding substitutions Gln-149-Ala and Val-265-Ile in the tobacco rbcL gene in pLEV4 were introduced using the QuikChange Site-Directed Mutagenesis Kit (Stratagene) to produce plasmid pLEVtob^149A,265I. All plasmids were sequenced using BigDye terminator sequencing at the Biomolecular Resource Facility, Australian National University (Canberra, Australia).

Each of the pLEV-derived plasmids was introduced biolistically into three leaves of ^cmtrL1 as described (23), and three to seven independent spectinomycin-resistant plants were obtained for each mutant. Two independent plastome-transformed lines for each introduced rbcL gene were grown to maturity in soil in a growth atmosphere supplemented with 0.5% (vol/vol) CO₂ as described (24). At each generation the plants were fertilized artificially with wild-type pollen.

PAGE and Nucleotide Blot Analyses.

The preparation and analysis of soluble leaf protein by SDS/PAGE, nondenaturing PAGE, and immunoblot analysis was performed as described (33). Total leaf genomic DNA was isolated using the DNeasy Plant Mini Kit and used to PCR amplify and sequence the transformed plastome region using primers 5′- CTATGGAATTCGAACCTGAACTT-3′ (LSH) and 5′- GAGGTGTGATACTTGGCTTGATTC-3′ (LSE) (Fig. 1B) (24). DNA blot analysis of the genomic DNA was used to confirm homoplasmicity (Fig S1). Total RNA was extracted from 0.5 cm² of leaf in 0.8 mL TRIzol (Invitrogen). Six per cent or 12% of the RNA was separated on denaturing formaldehyde gels (34). The RNA was blotted onto Hybond-N nitrocellulose membrane (GE Healthcare) and probed with a ³²P-labeled 5′ ΔrbcL probe (Fig. 1B) as described (24).

Rubisco Content and Catalytic Assessments.

Rates of rubisco ¹⁴CO₂ fixation using soluble leaf protein extract were measured in 7-mL septum-capped scintillation vials in reaction buffer [50 mM Hepes-NaOH (pH 7.8), 10 mM MgCl₂, 0.5 mM RuBP] containing varying concentrations of NaH¹⁴CO₃ (0–67 μM) and O₂ (0–25%) (vol/vol), accurately mixed with nitrogen using Wostoff gas-mixing pumps as described (24, 33). Assays (0.5 mL total volume) were started by the addition of activated leaf protein, and the Michaelis constants (K_m) for CO₂ (K_C) and O₂ (K_O) were determined from the fitted data. Replicate measurements (n = 4–8) were made using protein preparations from two to four different leaves of independently transformed lines. For each sample the maximal rate of carboxylation (V_C) was extrapolated from the Michealis–Menten fit and then normalized by dividing the rate by the number of rubisco-active sites quantified by [¹⁴C]2-CABP binding (35, 36). Rubisco CO₂/O₂ specificity (S_C/O) was measured as described (37), using freshly extracted rubisco, quickly purified by ion exchange chromatography (24), from at least two separate plants for each independently transformed line.

Growth and Photosynthesis Analysis.

Wild-type (Nicotiana tabacum L. Petit Havana) and transplastomic tobacco lines were grown in growth chambers at 25 °C and 400 ± 50 μmol photons m⁻²·s⁻¹ as described (24) in air or 0.5% (vol/vol) CO₂-enriched air. Plant height from the soil surface to the apical meristem was measured until the first floral apertures emerged. Leaf photosynthesis and dark respiration rates in plants of comparable physiological development (45–50 cm in height; 14 or 15 leaves) were made in the growth chamber using an LI-6400 gas-exchange system (Li-COR) (24). The maximum quantum efficiency of PSII in dark-adapted leaves [variable fluorescence (F_v)/maximum fluorescence (F_m)] was measured in the same leaves using an LI-6400–40 Leaf Chamber Fluorometer.