Genes that encode the Flaveria trinervia PEPC gene and the Flaveria pringlei PEPC gene in the pTrc 99A plasmid were provided by Peter Westhoff (Dusseldorf). The PEPC genes were sub cloned into the pET-1B His6 TEV LIC vector plasmid, provided by Scott Gradia (Berkley; Addgene plasmid #29653). Genes were sub cloned using the ligation independent cloning method with Q5 DNA polymerase and T4 DNA polymerase (NEB). Cloned plasmids were isolated using a Miniprep DNA kit (Qiagen). Plasmids were Sanger sequenced to confirm the sequence identity (GATC Biotech).

Leaf samples were collected from P. queenslandicum at midday in full daylight and flash frozen in liquid nitrogen. Leaf samples were homogenized with a pestle and mortar in liquid nitrogen. RNA was extracted from ground leaves using the RNeasy Kit (Qiagen). Libraries of cDNA were generated with SuperScript II Reverse Transcriptase (Thermo Fischer Scientific). The PEPC from P. queenslandicum was amplified using the primers PquFor1B and PquRev1B ( Supplementary Table 1 ), and Q5 polymerase. The amplified gene was Sanger sequenced (GATC Biotech) with the PCR primers and with the primers Pqu_1323_Seq_For and Pqu_1752_Seq_Rev (Primers synthesized by Sigma, summarized in Supplementary Table 1 ). The gene was then cloned into the pET-1B His6 TEV LIC vector plasmid as above.

Because non-C₄ PEPC from C₄ grasses generally represent distant paralogs resulting from ancient duplications that predate the origin of the family (Christin et al., 2007), the most closely related non-C₄ PEPC are in most cases those from related C₃ species. We consequently selected a gene from a C₃ species from the same tribe as P. queenslandicum. The sequence for PEPC from P. pygmeaum has been previously obtained via leaf transcriptome sequencing (Dunning et al., 2017). The sequence was codon optimized for expression in E. coli and synthesized by GenArt Gene Synthesis in the pTRCC Plasmid. The synthesized gene was sub-cloned into the pET-1B His6 TEV LIC plasmid and verified by Sanger sequencing.

For protein expression the BL21λ (DE3) strain of E. coli (NEB) was used. Chemically competent E. coli cells were transformed with each of the plasmids. Eight liters of cultures were grown in LB media at 37°C to OD₆₀₀ 0.8. Cultures were cooled to 4°C for 1 h prior to recombinant protein induction with 0.5 mM IPTG (Fischer). Cultures were then incubated at 18°C for 18 h. Cells were harvested by centrifugation at 5,400×g for 25 min and stored at −80°C.

Cells were suspended in IMAC buffer (25 mM Tris, 0.5 M NaCl, 0.3 M glycerol, 20 mM imidazole (Acros Scientific)), 10 ml per 2 L of culture with 50 µl of 50 mgml⁻¹ DNase I and 100 µl of 100 mgml⁻¹ Pefabloc. Cells were passed twice through a cell disruptor (Constant Systems) before centrifugation at 276,000×g for 40 min. The supernatant was passed through a 0.45 µm pore filter (Elkay Labs.). PEPC was separated from soluble protein with a prepacked 1 ml nickel affinity column using an ÄKTA Pure 25 L Chromatography System. The loaded column was washed with 50 column volumes of IMAC buffer, then 50 column volumes of IMAC buffer containing 150 mM imidazole. Pure PEPC was eluted with 10 column volumes of IMAC buffer containing 400 mM imidazole.

Protein eluted from IMAC purification was loaded onto a Sephadex G50 desalting column (Amersham Biosciences) and rebuffered in storage buffer (20 mM Tris, 5% v/v glycerol, 150 mM KCl, 1 mM DTT (AnaSpec. Inc.)). Protein was aliquoted and frozen at −80°C until use.

PEPC enzyme concentration was quantified by absorption at 280 nm. Enzyme extinction coefficients were calculated using the ExPASy protein parameter tool and corrected by determining the absorbance of the protein denatured in 6 M guanidine hydrochloride (Gill and von Hippel, 1989). The extinction coefficients for the F. trinervia, F. pringlei, P. queenslandicum and P. pygmaeum PEPC were 120,480 M⁻¹cm⁻¹, 117,030 M⁻¹cm⁻¹, 105,810 M⁻¹cm⁻¹ and 111,510 M⁻¹cm⁻¹, respectively. Gel based protein quantification was not used.

Protein samples were analyzed for purity using SDS PAGE analysis. Samples of cell lysate or pure protein (25 µg or 5 µg protein respectively; BCA assay from Pierce) were denatured in 2 × SDS PAGE loading dye (200 mM Tris HCl pH 6.8, 2% SDS, 20% Glycerol, 0.01% Bromophenol blue (BDH Laboratory Supplies) and 7% β-mercaptoethanol). Protein was loaded onto an 8% acrylamide SDS gel with 2 µl of Blue Prestained Protein Standard Broad Range (11–190 kDa) (NEB). Gels were run for 50 min at 200 V with 1 × Tris/Glycine/SDS running buffer (Geneflow). Gels were stained with InstantBlue (Expedeon) and imaged with a ChemiDoc MP (BioRad).

PEPC activity was measured spectroscopically at 340 nm by coupling to NADH-malate dehydrogenase. Assays with a high fixed concentration of bicarbonate were observed using a FLUOstar plate reader (BMG Labtech) through a 340 nm ± 5 nm filter in absorbance mode. These assays were conducted in a reaction volume of 150 µl at 25°C. A typical reaction mixture contained 50 mM Tricine.KOH pH 8.0, 10 mM MgCl₂ (Fluka), 5 mM KHCO₃, 0.2 mM NADH (Fischer) and 0.1 U µl⁻¹ malate dehydrogenase. Assays were initiated with the addition of PEPC enzyme. Rates were calculated with a NADH calibration curve; this method takes account of the short pathlength in microtiter plates.

Assays at a range of bicarbonate concentrations were observed with a Cary spectrophotometer (Agilent Technologies) in the same reaction buffer, in a total reaction volume of 600 µl. Initial rates were calculated using the Cary analysis software. To remove background bicarbonate, the water and tricine buffer were sparged with nitrogen for 18 h prior to use in assays. These assays were constructed under a nitrogen flow and performed in a sealed cuvette. The reaction was initiated with the addition of 50 nM PEPC, delivered with a gastight syringe (Hamilton). Bicarbonate concentrations were controlled with the addition of freshly prepared potassium bicarbonate.

The background bicarbonate was determined using an endpoint assay with no potassium bicarbonate (30 min). This procedure determines the total concentration of dissolved and hydrated CO₂, (i.e. CO_{2 (aq)}, H₂CO₃, HCO₃ ⁻ and CO₃ ²⁻), at this pH over 97% is in the form of bicarbonate. Reported bicarbonate concentrations are the sum of the background and the added bicarbonate.

Kinetic parameters were evaluated by non-linear regression analysis in Igor Pro (Version 7.0.8.1; Wavemetrics Inc., Lake Oswego, Oregon). In all cases, the enzyme was assumed to be fully active. Primary plots were analyzed using Equation (1).

Analysis of secondary plots (i.e. of k c a t a p p or k c a t a p p / k m a p p vs [PEP]) with Equation (2) allowed determination of the steady-state kinetic parameters, k c a t / K m H C O 3 − and k c a t / K m P E P ,   K i P E P and k _cat.

Where k and k ^app are the true and apparent values of k _cat or k c a t / K m H C O 3 − and K is K m P E P or K i P E P .

In the case of inhibition data, secondary plots were analyzed using Equation (3), where in the case of competitive inhibition k ^app is k c a t a p p / k m a p p and K _i is the competitive inhibition constant K _ic or in the case of non-competitive inhibition k ^app is k c a t a p p and K _i is the non-competitive inhibition constant K _iu.

The non-competitive inhibition constant (K _iu) was determined by the secondary plot of k c a t a p p against inhibitor concentration. The competitive inhibition constant (K _ic) was determined by the secondary plot of k c a t a p p / k m a p p against inhibitor concentration.

All data points shown on plots of initial rate against substrate concentration are individual measurements. Standard errors are provided for every parameter estimate. In secondary plots of apparent kinetic parameters against substrate or inhibitor concentration the standard error of those parameter estimates are shown. These standard errors are provided directly by the nonlinear regression analysis routine implemented within Igor Pro.

Four PEPC isoforms were characterized. In grasses, the C₄ and non-C₄ forms of ppc-1P3 genes were isolated from the C₄ P. queenslandicum and synthesized based on the sequence of the C₃ species P. pygmaeum, respectively. The cloned genes were 962 and 969 codons long, respectively. They have an 86.2% identity in amino acids and a 93.2% similarity, including on the two positions that have been linked in C₄ Flaveria to K m P E P and decreased inhibition (positions 774 and 884, respectively; Blasing et al., 2000; DiMario and Cousins, 2018). In Flaveria, the C₄ and non-C₄ ppc-1E2 genes corresponding to the C₄ F. trinervia and the C₃ species F. pringlei were analyzed [ppcA as described in Svensson et al. (1997)]. The two genes are both 967 codons long, with a 94.7% identity and a 97.5% similarity. The orthologous relationships between these pairs of genes were confirmed by phylogenetic analyses ( Supplementary Figure 1 ).

All four genes were prepared in vectors for over-expression in E. coli with an N-terminal His tag. In all cases, expressed protein was purified to >95% purity as assessed by SDS PAGE with a single immobilized metal column ( Supplementary Figure 2 ).

Assays at saturating bicarbonate and variable concentrations of PEP ( Supplementary Figure 3 ) showed that both His tagged Flaveria PEPCs behaved similarly to untagged proteins previously described (Svensson et al., 1997; Blasing et al., 2000; Jacobs et al., 2008). Specifically, at pH 8.0, 10 mM MgCl₂, 10 mM KHCO₃, coupled to malate dehydrogenase, the His₆-PEPC from F. trinervia catalyses the formation of oxaloacetate with a K m P E P of 0.61 ± 0.05 mM and a k _cat of 47.99 ± 1.22 s⁻¹. Literature values are K m P E P ranging from 0.278 to 0.652 mM and V _max of 29 U mg⁻¹, allowing for the different protein concentration, our k _cat would be equivalent to a V _max of 25.56 U mg⁻¹ (Svensson et al., 1997; Blasing et al., 2000). Under the same conditions, the His₆-PEPC from F. pringlei catalyses the formation of oxaloacetate with a K m P E P of 0.05 ± 0.01 mM and a k _cat of 52.65 ± 1.37 s⁻¹; literature values are K m P E P ranging from 0.029 to 0.061 mM and V _max of 27 U mg⁻¹, and allowing for the different protein concentration, our k _cat would be equivalent to a V _max of 28.02 U mg⁻¹ (Svensson et al., 1997; Blasing et al., 2000). This confirms previous reports (Paulus et al., 2013a) that the presence of an N-terminal poly-histidine tag does not adversely affect the activity of these proteins.

The specificity for bicarbonate of all four enzymes was determined using a gas-tight assay system. Background bicarbonate was reduced to ca. 50 µM by sparging with nitrogen gas. Assays were performed at five PEP concentrations, while varying the concentration of bicarbonate ( Figure 1 ). The analysis of secondary plots ( Supplementary Figures 4 and 5 ) provided estimates of k _cat and the specificity constant, k _cat/K _m, for both substrates ( Table 1 ).

The specificity for bicarbonate (k _cat/K _m) of the C₄ P. queenslandicum PEPC is 1.09 × 10⁶ M⁻¹s⁻¹, almost twice as large as that of the non-C₄ P. pygmaeum enzyme ( Table 1 ). The specificity of this non-C₄ enzyme is comparable to that of the C₄ PEPC of Flaveria at 0.69 × 10⁶ M⁻¹s⁻¹ ( Table 1 ), which again is slightly higher than that of the Flaveria non-C₄ PEPC ( Table 1 ). In both cases the specificity constant for PEP is smaller in the C₄ form of the enzyme ( Table 1 ). In terms of bicarbonate K _m values these are within the range previously reported for C₄ and non-C₄ plant PEPC isoforms in work with reasonably careful control of background bicarbonate (O'Leary, 1982; Bauwe, 1986; Janc et al., 1992; Dong et al., 1998; DiMario and Cousins, 2018).

For both non-C₄ and C₄ enzymes, we investigated inhibition by the two feedback inhibitors, malate ( Supplementary Figure 6 ) and aspartate ( Supplementary Figure 7 ) across a range of PEP concentrations. These two structurally related inhibitors show different kinetic characteristics; unlike aspartate, malate remains an inhibitor at saturating concentration of PEP ( Table 2 ).

The C₄ cycle of Flaveria produces both malate and aspartate (Moore and Edwards, 1986; Meister et al., 1996), while Panicum species are expected to produce mainly aspartate around PEPC (Rao and Dixon, 2016). The two molecules have however been shown to inhibit PEPC in a variety of C₄ species (Huber and Edwards, 1975). In our analyses, all the PEPC enzymes are inhibited by malate at both limiting and saturating concentrations of PEP, and malate is a mixed inhibitor ( Figures 2 and 3 ). This mixed inhibition can be characterized by two inhibition constants; K i c M a l a t e at limiting PEP and K i u M a l a t e at saturating PEP. In all cases, K i c M a l a t e  >> K i u M a l a t e , which means that malate can be viewed as a predominantly competitive inhibitor. The two C₄ forms of the enzyme are both less sensitive to malate than the two non-C₄ forms ( Table 2 ). Unlike malate, aspartate is solely a competitive inhibitor for all of these enzymes ( Figure 4 ). Increasing concentrations of aspartate do not affect k _cat ( Supplementary Figure 8 ). Once again, the two C₄ forms of the enzyme are much less sensitive to aspartate than the two non-C₄ forms ( Table 2 ). Overall, our analyses indicate that the C₄ forms are much less sensitive to both inhibitors, independently of the taxonomic group and C₄ subtype, confirming previous reports (Huber and Edwards, 1975).

The non-C₄ genes encoding the PEPC enzymes of the C₃ plants P. pygmaeum and F. pringlei diverged about 150 million years ago and since then have accumulated numerous mutations and undergone multiple gene duplications (Christin et al., 2007; Christin et al., 2015). They share an 83.5% identity and a 91.2% similarity, and greater than 93% similarity with their respective C₄ proteins. While the exact function of each non-C₄ isoform is unknown, they are transcribed at similarly moderate levels (Moreno-Villena et al., 2018). Our investigation shows that the two non-C₄ enzymes characterized here exhibit functionally similar kinetic characteristics, including high sensitivity to competitive inhibition by malate and aspartate and a similar sensitivity to bicarbonate. However, the two non-C₄ isoforms differ in their K m P E P which is three-fold lower in the F. pringlei enzyme ( Table 1 ). While systematic screens of non-C₄ PEPC are missing, those of Alternanthera and a distant root paralog from Z. mays have a similar K m P E P to the F. pringlei enzyme (Dong et al., 1998; Gowik et al., 2006). These data suggest that despite hundreds of million years of divergence, non-C₄ PEPC are generally associated with high sensitivity to inhibitors and low (<0.2 mM) K _m for both substrates. These properties are likely required for a tight regulation and fast response of isoforms involved in anaplerotic functions, where the concentrations of substrates and products are low.

In both Flaveria and Panicum, the C₄ PEPC shows a markedly reduced sensitivity to both malate and aspartate as compared with the non-C₄ ortholog ( Table 2 ). This reduction in sensitivity, reported before in Flaveria (Blasing et al., 2002; Paulus et al., 2013b; DiMario and Cousins, 2018) and a variety of grasses from different C₄ subtypes (Huber and Edwards, 1975), is observed at all concentrations of PEP ( Figures 2 – 4 ). Our observations are thus consistent with the conclusion that the same selective pressures act in C₄ eudicots and at least some grasses to decrease the sensitivity to the inhibitors malate and aspartate. In C₄ plants the concentration of malate and aspartate are high, so this reduced sensitivity prevents PEPC being inhibited by downstream metabolites (Arrivault et al., 2017). The respective amounts of malate and aspartate vary among C₄ species (Moore and Edwards, 1986; Meister et al., 1996; Rao and Dixon, 2016), and concerted reduction of inhibition by both species is consistent with them sharing a binding site (Paulus et al., 2013a).

The adaptation of PEPC to the demands of the C₄ pathway involved qualitatively similar changes in substrate specificity between Flaveria and the grasses considered here ( Table 1 ). In both cases the specificity for PEP decreases and the specificity for bicarbonate increases. The C₄ form of Zea mays, an independent C₄ origin within grasses, has an affinity for PEP that is similar to P. queenslandicum (Dong et al., 1998). In addition, changes of K _m for PEP in the same direction have been reported in Alternanthera (Gowik et al., 2006), suggesting that decreases in PEP K _m happened convergently across C₄ origins. The functional value of these changes remains speculative and might be a side-effect of adaptation of other properties of the enzyme or a direct target of selection for tighter regulation when PEP concentrations are higher (Svensson et al., 2003). The differences in K _m for bicarbonate are less marked than those of PEP ( Table 1 ). The K _m for bicarbonate parameter is much higher in the non-C₄ root isoform from Z. mays (Dong et al., 1998), indicating it varies tremendously among non-C₄ PEPC. Data from more species are needed to determine whether the qualitative convergence observed here between Flaveria and Panicum is universal, or depends on the co-opted gene or the details of the C₄ phenotype (e.g. biochemical and anatomical subtypes). Indeed, the cellular concentration of bicarbonate depends on the action of the enzyme carbonic anhydrase, in addition to the cell pH, and it is thus possible that variation in these factors changes the adaptive value of bicarbonate affinity.

While differences in substrate specificity and sensitivity to inhibitors are qualitatively convergent between Flaveria and the two grasses considered here, they are more marked in the latter ( Table 1 ). These quantitative differences might be linked to the contrast between the length of time spent as C₄ in each lineage, from more than 16 million years for P. queenslandicum to less than three for Flaveria (Christin et al., 2008; Christin et al., 2011). The C₄ PEPCs share a 76.5% identity and an 88.1% similarity. Indeed, the kinetic properties observed in the PEPC of extant taxa result from adaptive changes accumulated since the initial origin of C₄ photosynthesis. According to current models, an initial C₄ pathway can evolve via enzyme upregulation and limited modifications of the proteins (Sage et al., 2012; Heckmann et al., 2013; Dunning et al., 2019; Heyduk et al., 2019), as observed in C₃–C₄ intermediates (Svensson et al., 2003; Dunning et al., 2017). Once a C₄ pathway is in place, selection will act to improve its efficiency (Heckmann et al., 2013), and variation among members of the same C₄ lineage indicates that such process can take protracted periods of selection on novel mutations (Heyduk et al., 2019).

Because these are likely necessary for a function of PEPC in C₄ cells with high concentrations of metabolites, we suggest that relaxed sensitivity to inhibitors happens early during the evolution of C₄ PEPC. This hypothesis is supported by the fact that changes in sensitivity to inhibitors are observed in intermediates from Flaveria (Engelmann et al., 2003). It has moreover been shown that one single amino acid replacement is sufficient to generate a large decrease in sensitivity (Paulus et al., 2013a). The C₄-specific residue at this site is observed in multiple C₄ lineages of both grasses and eudicots (Paulus et al., 2013a; Paulus et al., 2013b), suggesting that a rapid decrease of inhibition is involved in many origins of C₄ PEPC.

Other properties of C₄-specific PEPC might represent secondary adaptations to the C₄ context, which might happen either to strengthen the early trends or in response to other changes of the plant biochemical phenotype. Over time, sustained diversifying selection on C₄ PEPC would have led to stronger differences between P. queenslandicum and the C₃ grasses. This view is supported by the similar kinetic parameters between the C₄ PEPC of P. queenslandicum and Zea mays, two grass lineages of similar age, as well as similar kinetic parameters observed between the C₄ PEPC in Alternanthera and Flaveria, two comparatively young lineages (Christin et al., 2011). It is however possible that secondary PEPC adaptations vary among and maybe even within old C₄ lineages, as different biochemical and anatomical C₄ subtypes evolved. Data from more lineages are needed to test the hypothesis that such diversifying PEPC secondary adaptation happened.

The molecular basis of the C₄ specific properties reported here are not well understood. Analysis of the evolution of the amino acid sequence of C₄ PEPC has shown that at least 22 sites underwent positive selection in grasses and sedges (Christin et al., 2007). Of these sites, three are also observed in C₄ Flaveria (Christin et al., 2007; Besnard et al., 2009). Some of these mutations have been shown to be responsible for key C₄ specific kinetic properties. Of these, a mutation for alanine to serine at position 774 (Flavaria numbering) has been identified as an important determinant of the low specificity for PEP of the C₄ form of the enzyme (Blasing et al., 2000); interestingly, the effect of this position on bicarbonate specificity depends on the rest of the sequence and concentrations of allosteric regulators (DiMario and Cousins, 2018). Additionally, a mutation at position 884 (Flaveria numbering), in the allosteric inhibitor binding site, has been shown to have a notable effect on the IC₅₀ for malate. An arginine residue in this position, as seen in the non-C₄ form of the enzyme is well placed to directly interact with the inhibitor, increasing the susceptibility to inhibition (Paulus et al., 2013a). These amino acid changes are observed in the C₄ PEPC of Panicum but not its non-C₄ ortholog, and presumably contribute to the kinetic differences between the two. The specific role of other grass mutations has yet to be identified, a task that will be complicated by the large amount of variation among grass PEPC and possible epistasy among sites. These factors make it difficult to associate specific kinetic changes with specific amino acid replacements. Here, we compared the characteristics of PEPC from old, diverse lineages; these efforts now need to be expanded to other C₄ lineages, with the well characterized isoforms of Flaveria continuing to serve as a model to assess the effect of specific sites.