Changes in Rubisco Kinetics during the Evolution of C4 ...

Changes in Rubisco Kinetics during the Evolution of C4 Photosynthesis in Flaveria (Asteraceae) Are Associated with Positive Selection on Genes Encoding the Enzyme Maxim V. Kapralov, ,1 David S. Kubien,* ,2 Inger Andersson,3 and Dmitry A. Filatov1 1

Department of Plant Sciences, University of Oxford, Oxford, United Kingdom Department of Biology, The University of New Brunswick, Fredericton, New Brunswick, Canada 3 Department of Molecular Biology, Swedish University of Agricultural Sciences, Uppsala, Sweden  Both authors contributed equally to this work. *Corresponding author: E-mail: [email protected]. Associate editor: Neelima Sinha 2

Rubisco, the primary photosynthetic carboxylase, evolved 3–4 billion years ago in an anaerobic, high CO2 atmosphere. The combined effect of low CO2 and high O2 levels in the modern atmosphere, and the inability of Rubisco to distinguish completely between CO2 and O2, leads to the occurrence of an oxygenation reaction that reduces the efficiency of photosynthesis. Among land plants, C4 photosynthesis largely solves this problem by facilitating a high CO2/O2 ratio at the site of Rubisco that resembles the atmosphere in which the ancestral enzyme evolved. The prediction that such conditions favor Rubiscos with higher kcatCO2 and lower CO2/O2 specificity (SC/O) is well supported, but the structural basis for the differences between C3 and C4 Rubiscos is not clear. Flaveria (Asteraceae) includes C3, C3–C4 intermediate, and C4 species with kinetically distinct Rubiscos, providing a powerful system in which to study the biochemical transition of Rubisco during the evolution from C3 to C4 photosynthesis. We analyzed the molecular evolution of chloroplast rbcL and nuclear rbcS genes encoding the large subunit (LSu) and small subunit (SSu) of Rubisco from 15 Flaveria species. We demonstrate positive selection on both subunits, although selection is much stronger on the LSu. In Flaveria, two positively selected LSu amino acid substitutions, M309I and D149A, distinguish C4 Rubiscos from the ancestral C3 species and statistically account for much of the kinetic difference between the two groups. However, although Flaveria lacks a characteristic ‘‘C4’’ SSu, our data suggest that specific residue substitutions in the SSu are correlated with the kinetic properties of Rubisco in this genus.

Introduction Photosynthesis is the primary route through which energy and carbon enter the biosphere. During photosynthesis, Rubisco (ribulose-1, 5-bisphosphate [RuBP] carboxylase/ oxygenase, EC 4.1.1.39) catalyzes the addition of CO2 to RuBP, producing two molecules of 3-phosphoglycerate (3-PGA). In phototrophic eukaryotes and cyanobacteria, the thylakoidal light-harvesting reactions deliver the energy required to reduce 3-PGA and regenerate RuBP. The associated water-splitting reaction has produced the modern aerobic atmosphere but has also reduced the efficiency of photosynthesis because Rubisco cannot distinguish completely between CO2 and O2 as substrates for RuBP fixation. The oxygenation of RuBP produces a 3-PGA and a 2-phosphoglycolate (2-PGO); photorespiration recovers some of the carbon in 2-PGO, but this consumes light energy and reduces both the rate and efficiency of CO2 assimilation. Hence, the efficiency of photosynthesis is strongly dependent on the relative specificity of Rubisco for CO2 versus O2 (SC/O), which is the ratio of the catalytic efficiencies (kcat/Km) of carboxylation (e.g., kcatCO2/Kc) and oxygenation (kcatO2/Ko) (Laing et al. 1974). In plants, SC/O ranges 70–110 mol mol–1 (Jordan and Ogren 1981;

Kane et al. 1994; Galmes et al. 2005; Tcherkez et al. 2006), which favors CO2 fixation. However, because oxygen is about 500
more abundant than CO2 in the modern atmosphere, carboxylation proceeds at only about four times the rate of oxygenation in C3 plants under optimal physiological conditions. Exacerbating the poor substrate specificity is a low turnover rate (kcatCO2), in the range of 2.5–5 s–1 in land-plant Rubiscos. Rubisco is likely the most abundant enzyme on Earth (Ellis 1979), an investment necessitated by it being a relatively feeble catalyst (Roy and Andrews 2000). The oxygenation reaction is an inevitable consequence of the Rubisco reaction mechanism (Tcherkez et al. 2006), but natural variation in SC/O and other biochemical parameters does occur. For example, SC/O is typically lower in Rubisco from cyanobacteria than from green plants (Badger et al. 1998). Among land plants, C4 species typically have Rubiscos with higher kcatCO2 and Kc and lower SC/O than enzymes from C3 plants (Yeoh et al. 1980, 1981; Hudson et al. 1990; Sage 2002; Kubien et al. 2008). The CO2-concentrating mechanism of C4 photosynthesis includes a suite of modified biochemical, and usually structural, changes that largely suppress RuBP oxygenation, by

© The Author 2010. Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution. All rights reserved. For permissions, please e-mail: [email protected]

Mol. Biol. Evol. 28(4):1491–1503. 2011 doi:10.1093/molbev/msq335

Advance Access publication December 16, 2010

1491

Research article

Abstract

MBE

Kapralov et al. · doi:10.1093/molbev/msq335

providing supra-atmospheric CO2 levels to Rubisco (Sage 2004). This modified CO2 environment enables more efficient in situ function of Rubisco in C4 species, and selects for faster, less-specific Rubiscos (Sage 2002; Kubien et al. 2008). Form 1 Rubiscos from plants and algae are hexadecamers of eight large subunits (LSu) and eight small subunits (SSu) (L8S8); the LSu includes the catalytic domain, and so the holoenzyme has eight catalytic sites (for review, see Spreitzer and Salvucci 2002). In plants and green algae, the LSu and SSu are encoded by the chloroplast rbcL gene and a family of nuclear rbcS genes, respectively. Positive selection on protein structure and function is a common occurrence in the land-plant Rubisco LSu (Kapralov and Filatov 2006, 2007; Christin et al. 2008). In a survey of .3,000 rbcL sequences, Kapralov and Filatov (2007) identified 98 residues that potentially evolved under positive selection in the LSu, including residues involved with dimer–dimer, intradimer, intersubunit, activase, and catalytic–site interactions. In a subsequent study, Christin et al. (2008) identified eight LSu residues under positive selection that were potentially associated with the transition from C3 to C4 photosynthesis; five of these amino acid sites correspond with those identified in C3 plants by Kapralov and Filatov (2007), suggesting that biochemical adaptation of Rubisco in both C3 and C4 plants may involve the same sites. However, because no measurements of Rubisco’s kinetic parameters were associated with these studies, they cannot assess the kinetic effect of individual amino acid replacements. The biochemical role of the SSu is currently uncertain, and levels of selection acting on the rbcS gene product have not yet been estimated (Andrews and Ballment 1983; Spreitzer and Salvucci 2002; Spreitzer 2003; Genkov and Spreitzer 2009). Although some bacterial and dinoflagellate Rubiscos are LSu dimers (e.g., L2), both the LSu and the SSu are needed to achieve the catalytic efficiency observed in green plants (Andrews and Ballment 1983; Spreitzer and Salvucci 2002). There is evidence that the SSu can affect SC/O; hybrid enzymes with cyanobacterial LSu (SC/O 41 mol mol–1) and diatom SSu (SC/O 107 mol mol–1) had SC/O of 65 mol mol–1 but greatly reduced carboxylation capacity (Read and Tabita 2002). However, hybrid Rubiscos consisting of cyanobacterial LSu and spinach SSu had SC/O similar to the prokaryote (Andrews and Lorimer 1985). Hybrid Rubiscos consisting of LSu from sunflower and SSu from tobacco have kinetic properties of the native enzymes, which are kinetically equivalent (Sharwood et al. 2008), although a previous report on such hybrids indicated lower kcatCO2, SC/O, and higher Kc (Kanevski et al. 1999). Most of this evidence supports the notion that the kinetically crucial residues are on the LSu. However, none of these studies utilize very closely related enzymes, thus the context within which either subunit operates in the hybrid Rubiscos differs significantly from its native state. The biochemistry (reviewed by Spreitzer and Salvucci 2002) and molecular evolution (e.g., Pasternak and Glick 1992; Kapralov and Filatov 2007; Christin et al. 2008) of 1492

Rubisco have been studied extensively, but the structural differences that account for kinetic changes remain uncertain. Hudson et al. (1990) found only one rbcL replacement substitution that occurred in each of three congeneric pairs of C3 and C4 species (Met309Ile). However, this is not a universal difference between C3 and C4 Rubiscos; both wheat (C3) and maize (C4) have Met-309 (Christin et al. 2008). One problem is that relatively few studies examine both the structure and the kinetic behavior of the enzyme simultaneously. Kinetic information is widely available but usually from distantly related taxa (Yeoh et al. 1980, 1981; Hudson et al. 1990; Sage 2002; but see Kubien et al. 2008). A tremendous amount of sequence data has been collected for phylogenetic reconstruction (Kellogg and Juliano 1997), particularly for rbcL, but rarely is such data associated with any biochemical information. Resolving the structure– function relationships of Rubisco likely requires detailed biochemical and structural studies on enzymes from closely related taxa. Such an approach should minimize the number of nonadaptive changes, which may mask adaptive mutations. In this study, we use closely related angiosperm species with different photosynthetic pathways to explore the evolution of genes encoding Rubisco. We attempt to relate observed amino acid substitutions with established biochemical variation. To achieve this, we sequenced the rbcL and rbcS genes from 15 species of Flaveria (Asteraceae, Asteroideae) in which key Rubisco kinetic parameters have been previously determined (Kubien et al. 2008). Flaveria includes C3, C3–C4 intermediate, C4-like, and C4 species; intermediate and C4-like species vary in the degree to which the oxygenation reaction inhibits photosynthesis (Ku et al. 1991) and can include biochemical and/or structural intermediacy (McKown and Dengler 2007). The intermediate steps are thought necessary for the evolution of complete C4 photosynthesis (Sage 2004). In Flaveria, the C4 species have 45% higher Kc, 23% higher kcatCO2, and 5% lower SC/O than the ancestral C3 species in the genus, whereas the intermediate taxa have largely C3-type Rubiscos (Wessinger et al. 1989; Kubien et al. 2008). Hudson et al. (1990) reported three amino acid substitutions between the LSus of F. pringlei (C3) and F. bidentis (C4), but did not report kinetic parameters, nor any structural information about the SSu. Because of the low molecular divergence between species, the relatively young age of the genus (Kopriva et al. 1996; McKown et al. 2005), and the presence of kinetically distinct Rubiscos, Flaveria is an ideal system in which to study the molecular bases of the transition from C3 to C4 photosynthesis and to examine the association between amino acid substitutions and the kinetic parameters of Rubisco.

Materials and Methods Plant Material and Sequencing of Chloroplast and Nuclear Genes Plants were grown from the same seed families or cuttings that produced the individuals used for Rubisco kinetic

MBE

Evolution of rbcL and rbcS in Flaveria · doi:10.1093/molbev/msq335

measurements by Kubien et al. (2008). Genomic DNA was isolated from fresh leaf material using a DNeasy Plant Mini Kit (Qiagen Ltd., Crawley, UK) in accordance with the manufacturer’s protocol. Additionally, we obtained DNA from another C4 Asteraceae species, Pectis papposa, using a specimen from the Oxford University Herbarium (OXF; voucher number 2585). The primers from Panero and Funk (2008) were used for amplification and sequencing of the chloroplast genes ndhF, psbA, rbcL, and trnL-F. Primers FS1F (5#-GTGTGGCCACCASTTGGAAA-3#) and FS688R (5#-GGATCCATGCCTGAGGGTAC-3#) were designed and used in this study for amplification and sequencing of the nuclear rbcS genes. For polymerase chain reaction (PCR) amplification of all regions, we used BioMix Red (Bioline Ltd., London, UK) with the following PCR conditions: one cycle of 95 °C, 2 min; 55 °C, 30 s; 72 °C, 4 min followed by 36 cycles of 93 °C, 30 s; 53 °C, 30 s; 72 °C, 3.5 min. The PCR products were extracted from agarose gels using the Qiagen gel extraction kit. Chloroplast genes were sequenced directly; nuclear rbcS was sequenced after cloning in Escherichia coli (TA-cloning kit; Invitrogen Ltd., Paisley, UK). Twenty to forty clones were sequenced for each Flaveria species. Sequencing was performed using ABI BigDye v3.1 system on an ABI3700 automated sequencer. Sequence chromatograms were checked and corrected, and the contigs were assembled using ProSeq3 software (Filatov 2009). All polymorphic sites were checked against the original sequence chromatograms, and doubtful regions were resequenced; sequences were compared with homologs from GenBank, and open reading frame integrity was confirmed for protein-coding sequences. Novel sequences have been submitted to GenBank under accession numbers HQ534103–HQ534293.

Positive Selection Analysis in rbcL Alignments of chloroplast genes were performed and manually checked using ProSeq3 software (Filatov 2009). Alignments were unambiguous given the conservatism of selected chloroplast regions. Bayesian estimations of phylogeny assuming a general time reversible þ C model of DNA sequence evolution (MrBayes, Ronquist and Huelsenbeck 2003) were performed for every chloroplast region investigated, as well as for concatenated data sets using Flaveria species only or Flaveria and outgroup species. Analyses, each of four chains, were run for 5,000,000 generations. A tree was sampled each 100 generations after a burn-in period, required to reach apparent stationarity. The outgroup consisted of the C4 Asteraceae, P. papposa, obtained in this study, and two C3 Asteraceae species (Guizotia abyssinica [EU549769] and Helianthus annuus [DQ383815]) obtained from GenBank. The topologies of trees for different chloroplast genes were similar according to the approximately unbiased (AU) and weighted Shimodaira–Hasegawa (WSH) tests (Shimodaira 2002) performed in the Treefinder package (Jobb et al. 2004). Thus, for further phylogenetic analyses of positive selection in Flaveria rbcL, we used the tree based on concatenated data set of all regions (fig. 1).

We tested for positive selection using nested maximum likelihood models, allowing for variation in the ratio of nonsynonymous to synonymous substitutions rates (dN/dS) across codons and/or branches of phylogenies, implemented in the codeml program of the PAML package version 4 (Yang 1997; Yang et al. 2005). We performed two likelihood ratio tests (LRTs). 1) M1a–M2a LRT for positive selection: the null M1a model assumes purifying or nearly neutral evolution without positive selection and allows codons with dN/dS , 1 and/or dN/dS 5 1, but no codons with dN/dS . 1; in addition, the more general M2a model allows for codons under positive selection (dN/dS . 1). 2) Test 2 or branchsite test of positive selection (Yang et al. 2000, 2005; Zhang et al. 2005): the alternative A model allows dN/dS ratios to vary both among sites and among lineages; it allows 0 , dN/dS ,1 and dN/dS 5 1 for both foreground and background branches, and also two additional classes of codons with dN/dS . 1 on foreground branches while restricted as 0 , dN/dS ,1 and dN/dS 5 1 on background branches, whereas the null model is model A1 with fixed dN/ dS 5 1 for two additional classes of codons on foreground branches (the same codons that have dN/dS . 1 in the A model). This is the most robust test for positive selection on foreground branches. The branches leading to C4 and C4-like Flaveria species, as well as to C4 P. papposa (fig. 1), were selected as foreground branches. For both LRTs, the first, null model, is a simplified version of the second, with fewer parameters, and is thus expected to provide a poorer fit to the data (lower maximum likelihood). The significance of the LRTs was calculated assuming that twice the difference in the log of maximum likelihood between the two models is distributed as a chi-square distribution, with the degrees of freedom (df) given by the difference in the numbers of parameters in the two nested models. For the M1a–M2a comparison, the df 5 2, and for Test 2, the null distribution should be a 50:50 mixture of point mass 0 and v2, so the P values were calculated using df 5 1 and then halved (Yang et al. 2000). To identify amino acid sites potentially under positive selection, the parameter estimates from models M2a and A were used to calculate the posterior probabilities that an amino acid belongs to a class with dN/dS . 1, using the Bayes Empirical Bayes (BEB) approaches implemented in PAML (Yang et al. 2005).

Positive Selection Analysis in rbcS RbcS genes were aligned in PRANK, using a method that recognizes insertions and deletions (indels) as distinct evolutionary events (Loytynoja and Goldman 2005). This method has been shown to improve the quality of sequence alignments and downstream analyses over a wide range of realistic alignment problems (Loytynoja and Goldman 2008). In addition to the Flaveria rbcS genes obtained in this study, we used rbcS expressed sequence tags from the genera Helianthus, Guizotia, Senecio, and Stevia obtained from GenBank (supplementary table S1, Supplementary Material online). Like Flaveria, these genera belong to the subfamily Asteroideae (Panero and Funk 2008). Bayesian estimations of phylogeny (MrBayes, 1493

Kapralov et al. · doi:10.1093/molbev/msq335

MBE

FIG. 1. Chloroplast phylogeny of Flaveria and ‘‘Asteraceae’’ outgroup based on rbcL, ndhF, psbA, and trnL-F sequences. Bayesian posterior probabilities are shown for each node. Gray- and black-dashed lines indicate the M309I and DE149A substitutions, respectively. Photosynthetic pathway is indicated for each species (C3, C3–C4, C4-like, and C4), as assigned previously (McKown et al. 2005; Sudderth et al. 2007). Inset: relationships between kcatCO2 and Kc for the Rubiscos of 15 species of Flaveria; data from Kubien et al. (2008). Note that the kinetics of the outgroup Rubiscos are unknown.

Ronquist and Huelsenbeck 2003) were performed for two data sets: 1) Flaveria sequences containing both exons and introns and 2) Flaveria and other Asteroideae containing exons only. Multiple copies of rbcS genes in the same species and gene conversion between these copies (see Results) could be two potential sources of false positives in PAML analyses (Casola and Hahn 2009). To overcome these problems, we repeated the M1a–M2a test (see above) 1,000 times using only one randomly chosen sequence per species and a species tree with and without outgroup, as recommended by Casola and Hahn (2009).

Phylogenetically Independent Contrasts of Rubisco Structure and Function We assess the impact of structural changes to the LSu and SSu by using parameters published by Kubien et al. (2008), who reported kcatCO2, SC/O, and the Michaelis–Menten constants for CO2 and O2 from 16 species of Flaveria; 1494

F. brownii (C4-like) was included by Kubien et al, but is absent from the current data set. The coding matrix is available as a supplemental file (supplementary table S2, Supplementary Material online). Closely related taxa are not independent data points, and they violate the assumptions of conventional statistical methods (Felsenstein 1985). We tested the effect of changes to amino acid substitutions on Rubisco biochemistry, using equivalent phylogenetically independent contrasts (PICs, Felsenstein 1985), and conventional statistical approaches, on those residues shown to be under positive selection. In essence, our analysis asks if Rubisco kinetic parameters differ when alternative amino acids are present at single residues; does kcatCO2 change if the LSu has Met-309 or Ile-309? By contrast, Kubien et al. (2008) asked if photosynthetic pathway was related to altered kinetic states. Note that species is not considered in either analysis. The PIC approach is more conservative than conventional statistics; the difference in trait values is calculated

Evolution of rbcL and rbcS in Flaveria · doi:10.1093/molbev/msq335

at each node of the phylogeny, resulting in n-1 contrasts, where n is the number of species. We used the Flaveria chloroplast–gene phylogeny obtained in this study (see fig. 1), and the three-gene consensus tree reported by McKown et al. (2005, their fig. 5). We generated Felsenstein’s independent contrasts in Mesquite (version 2.72) using the PDTREE module in PDAP (version 6.0) (Garland et al. 2005a, 2005b). PDAP permits two character states only, which is sufficient for the observed variability in the LSu. For the SSu, we assigned each taxon the amino acid residue that we observed most commonly at positions 24 and 57. Residue SSu-20 was omitted from the analysis because four amino acids were possible at this position. We examined the statistical adequacy of two types of branch lengths: those from the Flaveria phylogeny obtained here, and branch lengths set to unity. Plots of absolute values of standardized contrasts against their standard deviations indicated that uniform branch lengths were adequate, and these were used for the PIC analysis. The direction of the correlation coefficients indicates the effect of replacing the F. cronquistii residue with the alternative amino acid; F. cronquistii is an ancestral in the genus (McKown et al. 2005). To illustrate the importance of considering the relationship between species, we also calculated correlation coefficients between amino acid substitutions and Rubisco biochemistry using the lm function in R (http://www. r-project.org). This is the conventional statistical approach and assumes that the ‘‘tips’’ of the phylogeny (e.g., each species) are independent.

Structural Analysis of Rubisco In the absence of a crystal structure for Rubisco from Flaveria, the structure of activated spinach Rubisco in complex with 2-CABP (8RUC) was used for analysis.

Results Diversity and Phylogeny of Rubisco-Encoding Genes in Flaveria We obtained complete sequences of the chloroplast rbcL gene, and partial sequences of the chloroplast ndhF and psbA genes, and the nuclear trnL-F intergenic region, from 15 Flaveria species and P. papposa (over 4.6 Kb per species). In Flaveria and Pectis, rbcL includes 1455 protein-coding nucleotides, making the mature LSu polypeptide ten amino acid residues longer than in spinach and most other dicots, but consistent with other members of the Asteroideae (Panero and Funk 2008). Synonymous divergence within Flaveria for the concatenated chloroplast data set varies from 0% to 1.22% (mean 0.57%; modified Nei–Gojobori p-distance calculated in Mega). The trnL-F sequences (not shown) are identical to those of McKown et al. (2005), confirming correct species identification. Phylogenetic trees constructed from different chloroplast genes agreed with each other, so we used a combined alignment to build a chloroplast tree (fig. 1).

MBE The resulting phylogeny is similar to one based on morphology (Powell 1978) and to phylogenies based on DNA sequence data (Kopriva et al. 1996; McKown et al. 2005), except for the positions of F. palmeri and F. angustifolia. The three-gene consensus tree of McKown et al. (2005, their fig. 5) places the ‘‘C4-like’’ F. palmeri in clade A, which includes C3–C4, C4-like, and C4 species; in our chloroplast tree, F. palmeri belongs to clade B and is a sister taxon to the C3–C4 intermediate F. chloraefolia (fig. 1). As delineated by McKown et al., clade B includes C3–C4 intermediates and the C4-like F. brownii. Our chloroplast tree indicates that F. angustifolia is a sister taxa to other clade A species, whereas the three-gene tree of McKown et al. places it as sister to clade B. It should be noted that the trnL-F tree of McKown et al. (2005, their fig. 1) places F. angustifolia and F. palmeri in clades A and B, respectively. Both the AU and WSH tests (Shimodaira 2002) show significant topological incongruence between our chloroplast tree and the consensus tree of McKown et al. when F. palmeri and F. angustifolia were present on the trees (P , 0.01), but the trees agree when these taxa are excluded. The mature Flaveria SSu consists of 123 amino acid residues, and the rbcS gene encoding has three exons and two introns (Adams et al. 1987). We sequenced 291 nucleotides of exons two and three (coding for 97 amino acids from the N-terminus), and the second intron. We identified 145 unique rbcS sequences, with an average synonymous divergence of 7.9% (modified Nei–Gojobori p-distance) and found between 5 and 16 unique rbcS sequences in any Flaveria species. These numbers may not reflect the number of rbcS genes in a given species because some clone sequences that differ by one or two nucleotides could be different alleles of one heterozygous gene. Further, we may have missed some copies simply by not sequencing the entire length of the gene. Premature stop codons were found in four of eight rbcS sequences in F. kochiana, and in one of six copies in F. angustifolia; note that in most cases more than one premature codon was detected, which likely cannot be explained by a Taq polymerase error given the relatively short sequence length. There are 35 variable residues and 42 unique SSu polypeptides (ignoring pseudogenes); each species has at least two distinct polypeptides (supplementary fig. S1, Supplementary Material online). The second intron varies from 192 to 466 bp, but the sequences are highly divergent and hardly alignable, due to numerous indels. The rbcS sequences separate into two distinctive groups, which we call ‘‘rbcS-long’’ and ‘‘rbcS-short,’’ based on the intron length (fig. 2). Most Flaveria species are present in both groups, except for F. australasica, F. palmeri, F. trinervia, and F. vaginata, which are absent in the rbcS-short group. Species are monophyletic within rbcS-long, with the exceptions of F. angustifolia and F. pringlei, which are mixed together. Within the rbcS-short group, F. angustifolia and F. pringlei are intermixed, and the sequences of F. chloraefolia, F. floridana, and F. pringlei occur in two distant groups (fig. 2). The Bayesian posterior probabilities for 1495

Kapralov et al. · doi:10.1093/molbev/msq335

MBE monophyletic species are 0.9–1.0. However, within-species relationships are not resolvable for most rbcS copies. Within each group, species–specific clusters of rbcS resemble the consensus phylogeny of McKown et al. (2005), and we find that Flaveria forms a monophyletic group within the Asteroideae (fig. 3).

Positive Selection in Rubisco-Encoding Genes in Flaveria

FIG. 2. Phylogeny of Flaveria rbcS genes. The second rbcS intron varies from 192 to 466 bp, by which we delineate the ‘‘rbcS-short’’ and ‘‘rbcS-long’’ subfamilies.

1496

There are three replacement substitutions in Flaveria rbcL. The replacement of methionine by isoleucine at position 309 occurs along two independent branches leading to C4 and/or C4-like Flaveria, and along the branch leading to C4 Pectis (fig. 1). Replacement of glutamate/aspartate by alanine at position 149 occurs along the branch leading to the C4 species and F. vaginata (C4-like, sensu McKown et al. 2005) and is Flaveria specific. The third nonsynonymous mutation is a replacement of isoleucine by valine at position 265 in F. angustifolia and F. pringlei. In the consensus tree of McKown et al. (2005), F. cronquistii is basal, making V265 the ancestral state for Flaveria and the Asteroideae. LRTs for positive selection (Yang 1997, 2007) show that the model assuming positive selection on the chloroplast rbcL gene (M2a) fit the data better than the nested model without positive selection (M1a), whether or not the outgroup is included (table 1). The A model, which allows positive selection on branches leading to C4 and C4-like taxa, is a better fit than the model assuming no positive selection (table 1; Yang 1997, 2007). Our site-specific Bayesian tests uncovered three sites putatively under positive selection in the M2a model with posterior probabilities .0.95 (Yang et al. 2005), but only M309I was identified to be under positive selection in the A model (table 1). For Flaveria rbcS genes, we found 42 replacement mutations and gene conversion between the various copies was identified using GENECONV (P , 0.05, Sawyer 1989). Gene conversion is a neutral process that can lead to spurious false positives in phylogenetic analysis of positive selection. We attempted to overcome this problem by sampling just one copy per species under the M1a–M2a model (Casola and Hahn 2009), substituting a single additional copy in a randomized framework (
1000) and obtained similar results at each step. The model M2a assuming positive selection fits better (P , 0.05) than the model without selection (M1a) in 8% and 24% of cases when we used Flaveria with or without the outgroup, respectively. Three SSu amino acid sites, 20, 24, and 57, are putatively under positive selection, with posterior probabilities .0.5, in the M2a model by BEB analyses, in 95%, 98%, and 70% of PAML runs using the Flaveria-only data set, respectively. SSu residue 20 is potentially under positive selection, with posterior probabilities .0.95, in 23% of PAML runs. In Flaveria SSu, four amino acids can occur at residue 20 (Asn, Asp, Glu, or Thr), three at residue 24 (Thr, Ser, or Ala), and two at residue 57 (Arg or Ser). Only the C3 species F. pringlei has Ser-24; Ala-24 occurs in all the clade A species (e.g., the ‘‘C4’’ clade) and in F. chloraefolia and F. palmeri.

MBE

Evolution of rbcL and rbcS in Flaveria · doi:10.1093/molbev/msq335

Effects of Observed Amino Acid Substitutions on Rubisco Biochemistry Variation in the LSu residues 149 and 309 is correlated to the function of Rubisco, but residue 265 has no obvious biochemical impact (table 2; fig. 1). Kinetic values reported here are from Kubien et al. (2008) but calculated on the basis of amino acids rather than on photosynthetic pathway (see supplementary table S2, Supplementary Material online). Flaveria Rubiscos with Asp/Glu-149 have lower kcatCO2 (3.2 vs. 4.0 s–1) and Kc (12 vs. 21 lM) than enzymes with Ala-149 (P , 0.05, table 2). The same substitution is correlated with lower CO2/O2 specificity (from 81.5 to 77.1 mol mol–1), but this difference is significant only if the consensus tree of McKown et al. (2005) is used. The M309I substitution does not explain variation in SC/O between Flaveria Rubiscos but is associated with a faster enzyme (kcatCO2 3.9 s–1 vs. 3.2 s–1, P,0.05). Enzymes with Ile-309 have higher Kc than those with Met-309 (20 vs. 12 lM), but this difference is significant only using our chloroplast tree (table 2). The Michaelis–Menten constant for O2 (Ko) does not vary systematically with either residue. None of the substitutions in the SSu correlate with detectable effects on function, although the statistical conclusions are affected by the phylogeny used to generate the PICs (table 2). A PIC analysis using our chloroplast tree indicates that the R57S substitution is correlated strongly with higher Kc (18.7 vs. 13.5 lM, P , 0.05) and weakly (P , 0.09) with higher kcatCO2 (3.9 vs. 3.3 s–1). However, analysis based on the McKown et al. (2005) consensus tree indicates that this substitution has no observable effect (table 2). The trees differ in the placement of F. palmeri, which has kcatCO2 and Kc of 3.5 s–1 and 13.5 lM, respectively. Flaveria palmeri has Ser-57, but our chloroplast tree places it in the clade where Arg-57 is most common. Studies analyzing Rubisco biochemical parameters with consideration of the phylogenetic relationships between species are relatively rare (but see Kubien et al. 2008). To illustrate the importance of this approach, we also analyzed the effect of amino acid substitutions on Rubisco biochemistry without considering the enzymes’ evolution (e.g., considering the ‘‘tips’’ of the phylogeny as independent observations, using a conventional statistical approach). Both D/E149A and M309I Rubiscos have higher kcatCO2 and Kc and lower SC/O. The Type-I error rates are lower than in the PIC analysis, and these are essentially the differences between C3 and C4 species, reported by Kubien et al. (2008). The differences between PICs and ‘‘tips‘‘ analyses are most apparent in the SSu substitutions. If phylogeny is ignored, then Rubiscos with Ser-57 are faster and have higher Kc than Arg-57 enzymes (P , 0.05, table 2), and the T24A substitution is associated with a higher Kc (11.5 vs. 16.7 lM, standard error , 10% of the mean).

Discussion FIG. 3. Phylogeny of Flaveria and Asteraceae outgroup based on rbcS exons. Bayesian posterior probabilities are shown.

Flaveria As an Experiment Performed by Nature Flaveria is a young genus containing closely related C3, C4, and intermediate species and is therefore a convenient 1497

MBE

Kapralov et al. · doi:10.1093/molbev/msq335 Table 1. Results of LRTs for positive selection in Rubisco-encoding genes. Branch-site Model

Site Models M0 Data set N Length, bp vc Flaveria-rbcL 15 1455 0.21 Asteraceae-rbcL 18 1455 0.14 15 291 0.16 6 0.04 Flaveria-rbcSa b

M2a d

P2 0.007 0.014 0.06 6 0.01

v2e 43.55 10.71 2.25 6 0.50

A M2a Versus M1a Test A Versus A1 Test d P value P2 v 2e P value 0.000 0.003 47.50 0.024 0.000 0.009 28.57 0.000 NA NA NA NA

NA, not available. a Thousand Flaveria-rbcS data sets were generated by random sampling of one sequence per species from the pool of all Flaveria rbcS sequences. b Number of species. c dN/dS ratio averaged across all branches and codons; x and P values for Flaveria-rbcS are averaged across 1,000 LRT replicates. d Proportion of codons in a class under positive selection; x and P values for Flaveria-rbcS are averaged across LRT 1,000 replicates. e dN/dS ratio in a class under positive selection; x and P values for Flaveria-rbcS are averaged across LRT 1,000 replicas.

model to study the evolution of C4 photosynthesis (Powell 1978; Kopriva et al. 1996; McKown et al. 2005) and the adaptation of Rubisco to different CO2 availability (Wessinger et al. 1989; Kubien et al. 2008). In Flaveria, Rubiscos from C4 plants have higher kcatCO2 and Kc– and lower SC/O than the C3 or intermediate species (Kubien et al. 2008). Our data suggest that the kinetic differences between Flaveria Rubiscos are explained mainly by two LSu residues and may be influenced by SSu variation. Both Rubisco subunits are under positive selection, but the signal is much stronger in the LSu. The structure–function effects of the observed residue substitutions are not obvious because the catalytic site is unaltered, and it is possible that the SSu plays an unknown effect on catalysis.

Kinetic Differences between Flaveria Rubiscos Are Associated with Two LSu Substitutions Four C4 species of Flaveria and one C4-like species (F. vaginata) have replacements of methionine by isoleucine at position 309 and glutamate/aspartate by alanine at position 149. These species have Rubiscos with higher kcatCO2 and Kc and lower SC/O than enzymes from other Flaveria species (Kubien et al. 2008). Another C4-like species, F. palmeri, has Ile-309, Asp-149, and its Rubisco has C3-like kinetics. Interestingly, our chloroplast phylogeny is topologically incongruent with a phylogeny based on one chloroplast and two nuclear genes (McKown et al. 2005) in the position of F. palmeri. The likely explanation is an interspecific hybridization event, with the ancestor of F. palmeri ‘‘capturing’’ a chloroplast from a clade B Flaveria species; as a result, its Rubisco is under positive selection to evolve a C4 form. In F. palmeri I309 is coded by ATC, whereas in the other C4 and C4-like Flaveria species I309 is coded as ATT. Hence, F. palmeri acquired this substitution independent of the other C4-like and C4 Flaveria species within a relatively short time (dS 5 0.1% between F. palmeri and F. anomala, fig. 1). The ancestor of the ‘‘C4 clade’’ (i.e., clade A sensu McKown et al. 2005, including F. palmeri) obtained amino acid replacements at LSu residues 149 and 309, but F. palmeri lost the ‘‘C4 rbcL haplotype’’ and subsequently obtained Ile-309 but not Ala-149. The M309I substitution is under positive selection on phylogenetic branches leading to C4 and C4-like Flaveria 1498

species, as well as C4 P. papposa. Hudson et al. (1990) showed that M309I is the only replacement shared by all C4 species in three congeneric C3/C4 species pairs in Flaveria (Asteraceae), Atriplex (Chenopodiaceae), and Neurachne (Poaceae). The substitution was shown to be positively selected in numerous groups of terrestrial and aquatic C3 plants (Kapralov and Filatov 2007) and in C4 grasses and sedges (Christin et al. 2008). There is positive association between the M309I substitution and the kcatCO2 and Kc of Flaveria Rubisco (table 2). Flaveria palmeri has Ile-309 but Asp-149 and its Rubisco is biochemically similar to that of C3 species (kcatCO2 3.5 s–1, Kc 13.5 lM, Kubien et al. 2008). The C3 kinetics of the F. palmeri Rubisco indicate that the M309I substitution alone does not account for the observed differences between Flaveria Rubiscos. Within the Asteraceae, the D/E149A substitution is unique to the C4 Flaveria clade (Panero and Funk 2008), which explains why we could not detect positive selection on this residue using branch-specific analyses. This substitution is also unique to Flaveria when C3 and C4 species from different families are compared (Hudson et al. 1990; Christin et al. 2008). However, Ala-149 is associated with higher kcatCO2 and Kc in Flaveria (table 2), and Flaveria Rubiscos have fully C4 kinetics only if both Ile-309 and Ala-149 are present (see Discussion on F. palmeri above). The presence of two mutations leading to increased kcatCO2 and Kc is consistent with the observation that several amino acids are affected simultaneously by positive selection during Rubisco fine-tuning (Kapralov and Filatov 2006, 2007; Christin et al. 2008). It should be emphasized that although the presence of Ala-149 and Ile-309 gives a ‘‘C4’’ Rubisco phenotype in Flaveria, the presence of these two residues is not a universal C4 trait; maize and spinach are both Glu-149 and Met-309, but their Rubiscos are kinetically C4 and C3, respectively (Kubien et al. 2008). Clearly, the structural context within which these substitutions occur influences their kinetic effects. LSu residue 265 is also variable in Flaveria. Thirteen of the 15 Flaveria species sequenced here have Ile-265, whereas F. pringlei and F. angustifolia have Val-265; these two species hybridize in nature (Powell 1978; McKown et al. 2005). Val-265 was previously assumed to be the

MBE

Evolution of rbcL and rbcS in Flaveria · doi:10.1093/molbev/msq335

Table 2. PIC and Conventional (tips) analysis of Correlation (r) between Positively Selected Amino Acids and Rubisco Kinetic Parameters in Flaveria. Analysis PICs

Tree chloroplast (fig. 1)

Residue LSu-149

LSu-265

LSu-309

SSu-24

SSu-57

PICs

Three-gene consensus (McKown et al. 2005)

LSu-149

LSu-265

LSu-309

SSu-24

SSu-57

tips

n.a.

LSu-149

LSu-265

LSu-309

SSu-24

SSu-57

Variable SC/O kcat Kc Ko SC/O kcat Kc Ko SC/O kcat Kc Ko SC/O kcat Kc Ko SC/O kcat Kc Ko SC/O kcat Kc Ko SC/O kcat Kc Ko SC/O kcat Kc Ko SC/O kcat Kc Ko SC/O kcat Kc Ko SC/O kcat Kc Ko SC/O kcat Kc Ko SC/O kcat Kc Ko SC/O kcat Kc Ko SC/O kcat Kc Ko

r 20.279 0.636 0.810 0.017 0.169 20.136 0.022 20.015 0.083 0.474 0.646 20.432 0.150 20.083 0.282 20.076 0.183 0.375 0.446 20.242 20.487 0.453 0.766 0.135 0.102 0.048 0.197 20.144 0.048 0.467 0.430 20.366 0.067 20.094 0.302 0.015 0.317 0.229 0.009 0.152 0.726 0.762 0.953 0.172 0.270 0.374 0.218 0.111 0.520 0.758 0.871 0.426 0.279 0.293 0.568 0.170 0.179 0.550 0.517 0.195

df 13 13 13 12 13 13 13 12 13 13 13 12 13 13 13 12 13 13 13 12 13 13 13 12 13 13 13 12 13 13 13 12 13 13 13 12 13 13 13 12 13 13 13 12 13 13 13 12 13 13 13 12 13 13 13 12 13 13 13 12

F 1.09 8.84 24.77 0.00 0.38 0.25 0.01 0.00 0.09 3.76 9.33 2.75 0.30 0.09 1.13 0.07 0.45 2.13 3.23 0.74 4.03 3.35 18.40 0.22 0.14 0.03 0.53 0.27 0.03 3.62 2.96 1.85 0.06 0.12 1.31 0.00 1.46 0.72 0.01 0.29 14.51 18.00 129.90 0.37 1.02 2.11 0.65 0.15 4.82 17.59 41.00 2.66 1.10 1.22 6.18 0.36 0.43 5.63 4.73 0.47

P 0.1574 0.0054 0.0001 0.4768 0.2741 0.3140 0.4696 0.4804 0.3849 0.0373 0.0046 0.0616 0.2968 0.3845 0.1539 0.3980 0.2571 0.0842 0.0478 0.2026 0.0329 0.0450 0.0004 0.3225 0.3593 0.4327 0.2406 0.3050 0.4327 0.0398 0.0547 0.0992 0.4057 0.3697 0.1369 0.4798 0.1246 0.2060 0.4879 0.3015 0.0022 0.0010 0.0000 0.5560 0.3308 0.1701 0.4341 0.7049 0.0470 0.0011 0.0000 0.1290 0.3133 0.2899 0.0273 0.5604 0.5232 0.0337 0.0487 0.5047

NOTE.—The data matrix is available as supplementary table S2, Supplementary Material Online. Numerator df are 1 in all cases; denominator values are shown. The PIC analysis uses the chloroplast gene phylogeny obtained in this study (fig. 1), and the consensus chloroplast/nuclear gene tree of McKown et al. (2005, fig. 5). For the analysis of LSu-149, we considered aspartate and glutamate equivalent amino acids (see table 3). For the SSu, we assigned to each taxon the amino acid most commonly observed at each residue. In F. pringlei, serine and alanine were equally observed at residue SSu-24; we assigned it Ala-24, giving two character states across the phylogeny, in order to implement PDAP (see Materials and Methods). n.a., not applicable.

1499

MBE

Kapralov et al. · doi:10.1093/molbev/msq335 Table 3. Physical Properties of Amino Acid Substitutions in Flaveria Rubisco. Subunit LSu

SSu

Residuea 149 149 265 309 20 20 24 57

Substitution E 5> A E 5> D I 5> V M 5> I N 5> D N 5> E T 5> A R 5> S

Type of Changeb UA 5> HN UA 5> UA HN 5> HN HN 5> HN UP 5> UA UP 5> UA UP 5> HN UB 5> UP

DHc 5.3 0.0 20.3 2.6 0.0 0.0 2.5 3.7

DH% 151.4 0.0 26.7 136.8 0.0 0.0 357.1 82.2

DPd 24.2 0.7 0.7 20.5 1.4 0.7 20.5 21.3

DP% 234.1 5.7 13.5 28.8 12.1 6.0 25.8 212.4

DVe 249.8 –27.3 226.7 3.8 23.0 24.3 227.5 284.4

DV% 236.0 219.7 216.0 2.3 22.6 21.3 223.7 248.7

NOTE.—For the SSu, only the replacements under positive selection are shown. a Numbering is based on the spinach sequence (Knight et al. 1990). b Side chain type changes: A, acidic (negatively charged); B, basic (positively charged); H, hydrophobic; N, nonpolar aliphatic; P, polar uncharged; U, hydrophilic. c Hydropathicity difference (Kyte and Doolittle 1982). d Polarity difference (Grantham 1974). e Van der Waals volume difference taken from (Zamyatnin 1972, http://www.imb-jena.de/IMAGE_AA.html).

C3 condition in Flaveria (Hudson et al. 1990), and our data suggest that it is ancestral in the Asteroideae, but F. cronquistii (C3) has Ile-265. Substitutions at residue 265 do not correlate with Rubisco kinetics in Flaveria (table 2, and see Hudson et al. 1990) and may be functionally neutral given the similar nature of isoleucine and valine (table 3).

Lesser Role of the SSu The nuclear rbcS genes are more polymorphic and show more complex evolution than chloroplast rbcL. There are two subfamilies of rbcS genes in Flaveria (fig. 2). Within these groups, all genes cluster into monophyletic clades at the species level, with a few exceptions indicating known cases of interspecific hybridization (e.g., between F. pringlei and F. angustifolia). RbcS phylogenies within the two subfamilies agree with the three-gene consensus phylogeny of McKown et al. (2005), based on AU and WHS tests. We sampled six species that have C4 or C4-like photosynthesis; four of them (F. australasica, F. palmeri, F. trinervia, and F. vaginata) have rbcS genes from the rbcS-long subfamily only. The other taxa, including the C3 species and two C4 plants (F. bidentis and F. kochiana) have rbcS genes belonging to both subfamilies. The presence of pseudogenes in the C4 species F. kochiana is further evidence of the asymmetrical decay in rbcS copy numbers between C4 and C4-like Flaveria species and the rest of the genus. There are two possible explanations for the degradation and elimination of rbcS from one gene subfamily, and the resulting reduction in copy number, in C4 and C4-like species. First, decreasing the number of rbcS genes may be one of the mechanisms reducing Rubisco quantity in leaves of C4 Flaveria. Because Rubisco is localized to a high CO2 environment that enhances the efficiency of the enzyme in situ, C4 plants require less Rubisco than C3 species (Sage 2004). Spreitzer (2003) suggested that differential expression of rbcS genes could be a means of regulating Rubisco abundance; perhaps, the loss of rbcS copies reflects the reduced need for the holoenzyme. The second hypothesis is that rbcS genes may play a regulatory role, enabling the plant to tailor the biochemical properties of Rubisco to best suit environmental conditions. In C3 species, the CO2 levels at Rubisco are far below the current atmospheric CO2 level 1500

and are impacted severely by abiotic factors, such as drought and temperature (von Caemmerer and Quick 2000; Sage 2002). A plant might maximize carbon gain and fitness by being able to fine-tune Rubisco, and perhaps carrying a range of rbcS genes provides this capacity. By contrast, the CO2 environment experienced by a C4 Rubisco is stable and greatly above current atmospheric levels, as long as there is sufficient light to drive the mesophyll reactions. Perhaps the need to fine-tune the enzyme is less important in C4 plants, and so Rubisco does not need to be as ‘‘flexible’’ as in C3 species. However, in the absence of a specific role for the SSu, this remains speculative, and further work is necessary to establish a role of the SSu in altering the kinetic parameters of Rubisco. Flaveria’s multiple rbcS copies, and the gene conversion between them, agree with previous findings in other angiosperms (e.g., Meagher et al. 1989) and could lead to false positives in selection analysis (Casola and Hahn 2009). To exclude false positives, we repeated PAML tests for rbcS 1,000 times using only one randomly chosen sequence per species, which reduced our analytical power significantly. Computer simulations predict that including an outgroup will increase analytical specificity when site methods are used (Casola and Hahn 2009). Consistent with this, the proportion of cases with significant positive selection on rbcS (P , 0.05) increased from 8% to 24% when an outgroup was included. However, even 24% is likely an underestimate of the number of subsamples in which selection was detected within rbcS. This is because of the small number of sequences analyzed, and the fact that the C4 and C4-like species are in a single clade on the nuclear gene tree. The reduced kinetic role of replacement substitutions on the SSu versus the LSu is also confirmed by weaker results from tests for positive selection, and by the 19-fold difference between dN/dS values for codons under putative positive selection (table 1). Although we have relatively little statistical power, due to our small sample size, the Bayes Empirical Bayes approaches implemented in PAML (Yang et al. 2005) show that SSu residues 20, 24, and 57 are under weak positive selection in 70–98% of cases. None of these residues are among the seven highly conserved SSu residues that affect Rubisco’s catalytic

Evolution of rbcL and rbcS in Flaveria · doi:10.1093/molbev/msq335

FIG. 4. Structural implications of the observed amino acid replacements in Flaveria Rubisco, modeled on the spinach enzyme. Residues that are variable within Flaveria are colored green. (a) The LSu, showing the reaction-intermediate analog CABP and a Mg2þ ion bound to the active site. The analog of the adjacent LSu in the L2 dimer is also shown, but the LSu itself is not included because ˚ ) from the C2 it is located in front of the picture. Distances (in A atoms of CABP to the Ca atoms of Asn-149 and Met-309 are also shown. (b) The SSu (yellow) and the adjacent LSu (depicted as a gray surface). The position of the active site is indicated by the presence of CABP and a Mg2þ ion. The figure was prepared with PyMOL (DeLano 2002).

properties (Genkov and Spreitzer 2009). However, residue 57 occurs in the bA-bB loop of the SSu, one of the most variable regions in the enzyme (Spreitzer and Salvucci 2002) and shows an interesting substitution pattern. In most of the genus, this residue is arginine, but 11 of the 42 unique SSu polypeptides have serine. Ser-57 is present in all but one of the C4 and C4-like Flaveria. We cannot say that it is always or solely expressed in C4 and C4-like species like F. bidentis that have both types of genes, but four C4 and C4-like species (F. australasica, F. palmeri, F. trinervia, and F. vaginata) have only rbcS polypeptides with Ser-57. Depending on the phylogeny used, this substitution is positively correlated with increasing Kc and has a positive but nonsignificant correlation with kcatCO2.

Analysis of Flaveria Substitutions Using the Crystal Structure of Spinach Rubisco LSu residue 149 is located in the amino-terminal domain of the LSu, in a loop between two helices (aD and aE, follow-

MBE ing the numbering and secondary structure assignment of spinach Rubisco, Knight et al. 1990). In the spinach enzyme, residue Gln-149 is located close to the surface of the LSu, facing the central solvent channel (figure 4a). The side chain of Gln-149 is stabilized by two intrasubunit contacts: the Oe oxygen hydrogen bonds to the Nd nitrogen of His282 and the Ne nitrogen to the Oc oxygen of Ser-281. Residue 281 is under positive selection in numerous groups of plants (Kapralov and Filatov 2007) and in particular in C4 grasses and sedges (Christin et al. 2008). Residue 149 is also in contact with ordered solvent molecules via its carbonyl oxygen and the Oe oxygen. It makes solvent mediated contacts to the Ng nitrogens of the guanidinium group of Arg285 and to the Od oxygen of Asp-286. In ten of the 15 Flaveria species examined here Gln-149 is replaced by aspartate or glutamate, which should alter at least some of these interactions. However, the C4 Flaveria species have Ala-149, in which case the side-chain interactions are abolished. On the opposite side of the LSus relative to Ala-149 (fig. 4), residue Met-309 is at the carboxy-terminal end of a short b-strand (bF) at the carboxy-terminal side of the a/b-barrel (between strand 5 and helix 5 of the barrel). This location places Met-309 on the interface of LSus in the L2 dimer. The side chain of Met-309 is buried in a hydrophobic cleft and is within van der Waals distance to Phe117, Val-121, Leu-135, Ile-301, Val-313, and Leu-314. Met˚ distance. In 297 of the adjacent LSu is at a distance of 4.5 A P. papposa, the five C4 Flaveria, and F. palmeri, Met-309 is replaced by the smaller and more hydrophobic isoleucine, which may create a small cavity. Residues 149 and 309 are approximately midway between the two catalytic sites in the L2 dimer (fig. 4). Measuring from the position of the a-carbon atom to the C2 atom ˚ and of the two CABP molecules, the distance is 28.1 A ˚ ˚ ˚ 31.3 A for Gln-149, and 22.2 A and 21.2 A for Met-309 in the spinach enzyme, respectively. In the enzyme from Chlamydomonas reinhardtii, residues on the interface between subunits and distant from the active site have been suggested to influence the kinetics of Rubisco (Karkehabadi, Taylor, et al. 2005). It appears that the interfaces are important control surfaces for long-range communication in the hexadecamer. Alanine is a much smaller and more hydrophobic residue than glutamic or aspartic acid (table 3), which will be charged at physiological pH. The M309I substitution introduces a more hydrophobic residue, which has a stabilizing and favorable effect on overall molecule stability according to CUPSAT calculations using spinach pdbstructure (Parthiban et al. 2006; http://cupsat.uni-koeln.de). SSu residues 20 and 24 are located on the surface in contact with the solvent (fig. 4b). In spinach Rubisco residue 20 is proline, and due to the particular rigidity of this residue any other residue in its stead will increase the mobility at this position; Flaveria has at least four possible amino acids at this position, but we did not detect Pro-20. In Flaveria, there are several substitutions in the carboxyterminal end of the bA-bB loop. This loop is among the most divergent structural features of Rubisco (Spreitzer and Salvucci 2002) and is important for both assembly 1501

MBE

Kapralov et al. · doi:10.1093/molbev/msq335

and catalysis (Karkehabadi, Peddi, et al. 2005). In the spinach enzyme, residue Arg-57 lines the surface of the central solvent channel. Most Flaveria species have Arg-57, but in some of the C4 Flaveria Rubiscos the residue is serine, a much smaller and slightly more rigid residue that is uncharged at physiological pH. The R57S replacement may affect interactions between adjacent SSus, but how this affects the structure or catalytic function of Rubisco in Flaveria is uncertain at this stage. LSu substitutions E149A and M309I, and SSu substitutions T24A and R57S, increase hydrophobicity (table 3), a pattern observed for many thermostable proteins, including rbcL from hot spring cyanobacteria (Miller 2003). There is no evidence of differential thermostability between C3 and C4 Rubiscos, but these substitutions may have other subtle unknown effects that produce the distinct enzymatic phenotypes.

Conclusions We sequenced Rubisco-encoding genes from 15 Flaveria species and analyzed their molecular evolution and the correlations between positively selected amino acid substitutions and Rubisco kinetic parameters. To our knowledge, Flaveria is the largest monophyletic lineage within which both Rubisco kinetics and rbcL and rbcS sequences are available for most species. The evidence for positive selection on rbcL is much stronger than for rbcS, and two LSu amino acid substitutions, M309I and DE149A, may explain most of the observed differences in Rubisco kinetics between Flaveria species with different photosynthetic types. However, other substitutions may have similar effects in different taxa, and more effort is required to create a ‘‘periodic table’’ of amino acid substitutions that affect Rubisco kinetics in different phylogenetic lineages. Changes in the Rubisco phenotype are acquired by the evolution of, and interaction between, both subunits, because our data suggest that positive selection on the SSu is probable. Rubisco’s biochemical phenotype should depend on both its genotype (e.g., rbcL and rbcS) and the environment. We suggest that the SSu may alter the environment of the LSu in such a way as to alter the kinetic phenotype of the Rubisco holoenzyme. However, it is still impossible to assign a specific kinetic role to the SSu, thus how this subunit might constrain the phenotype of the holoenzyme cannot yet be ascertained.

Supplementary Material Supplementary tables S1 and S2 and figure S1 are available at Molecular Biology and Evolution online (http:// www.mbe.oxfordjournals.org/).

Acknowledgments Dr Linley Jesson and Dr Antonina Votintseva provided invaluable assistance with the statistical analysis of Rubisco kinetics and laboratory work, respectively. We thank Dr Christopher Dixon and Dr Richard Jobson for stimulating discussion. This work was supported by a Natural Science and 1502

Engineering Research Council of Canada Discovery Grant (327103-06) to D.S.K. and by a grant from Natural Environment Research Council UK (NE/H007741/1) to D.A.F.

References Andrews TJ, Ballment B. 1983. The function of the small subunits of ribulose bisphosphate carboxylase-oxygenase. J Biol Chem. 258:7514–7518. Andrews TJ, Lorimer GH. 1985. Catalytic properties of a hybrid between cyanobacterial large subunits and higher-plant small subunits of ribulose bisphosphate carboxylase/oxygenase. J Biol Chem. 260:4632–4636. Badger MR, Andrews TJ, Whitney SM, Ludwig M, Yellowlees DC, Leggat W, Price GD. 1998. The diversity and coevolution of Rubisco, plastids, pyrenoids, and chloroplast-based CO2concentrating mechanisms in algae. Can J Bot. 76:1051–1071. Casola C, Hahn M. 2009. Gene conversion among paralogs results in moderate false detection of positive selection using likelihood methods. J Mol Evol. 68:679–687. Christin PA, Salamin N, Muasya AM, Roalson EH, Russier F, Besnard G. 2008. Evolutionary switch and genetic convergence on rbcL following the evolution of C4 photosynthesis. Mol Biol Evol. 25:2361–2368. DeLano WL. 2002. PyMOL. [Internet]. Palo Alto (CA): DeLano Scientific. Available from: http://www.pymol.org Ellis RJ. 1979. The most abundant protein in the world. Trends Biochem Sci. 4:241–244. Felsenstein J. 1985. Phylogenies and the comparative method. Am Nat. 125:1–15. Filatov DA. 2009. Processing and population genetic analysis of multigenic datasets with ProSeq3 software. Bioinformatics 25:3189–3190. Galmes J, Flexas J, Keys AJ, Cifre J, Mitchell RAC, Madgwick PJ, Haslam RP, Medrano H, Parry MAJ. 2005. Rubisco specificity factor tends to be larger in plant species from drier habitats and in species with persistent leaves. Plant Cell Environ. 28:571–579. Garland T, Bennett AF, Rezende EL. 2005. Phylogenetic approaches in comparative physiology. J Exp Bot. 208:3015–3035. Garland T, Midford PE, Jones JA, Dickerman AW, Diaz-Uriarte R. 2005. PDAP: phenotypic diversity analysis programs. Version 6. Genkov T, Spreitzer RJ. 2009. Highly conserved small subunit residues influence Rubisco large subunit catalysis. J Biol Chem. 284:30105–30112. Grantham R. 1974. Amino acid difference formula to help explain protein evolution. Science 185:862–864. Hudson GS, Mahon JD, Anderson PA, Gibbs MJ, Badger MR, Andrews TJ, Whitfeld PR. 1990. Comparisons of rbcL genes for the large subunit of ribulose-bisphosphate carboxylase from closely related C3 and C4 plant species. J Biol Chem. 265:808–814. Jobb G, von Haeseler A, Strimmer K. 2004. TREEFINDER: a powerful graphical analysis environment for molecular phylogenetics. BMC Evol Biol. 4:18. Jordan DB, Ogren WL. 1981. Species variation in the specificity of ribulose-bisphosphate carboxylase-oxygenase. Nature 291:513–515. Kane HJ, Viil J, Entsch B, Paul K, Morell MK, Andrews TJ. 1994. An improved method for measuring the CO2/O2 specificity of ribulosebisphosphate carboxylase-oxygenase. Aust J Plant Physiol. 21:449–461. Kanevski I, Maliga P, Rhoades DF, Gutteridge S. 1999. Plastome engineering of ribulose-1,5-bisphosphate carboxylase/oxygenase in tobacco to form a sunflower large subunit and tobacco small subunit hybrid. Plant Physiol. 119:133–142. Kapralov MV, Filatov DA. 2006. Molecular adaptation during adaptive radiation in the Hawaiian endemic genus Schiedea. PLoS ONE. 1(1):e8.

Evolution of rbcL and rbcS in Flaveria · doi:10.1093/molbev/msq335 Kapralov MV, Filatov DA. 2007. Widespread positive selection in the photosynthetic Rubisco enzyme. BMC Evol Biol. 7:73. Karkehabadi S, Peddi SR, Anwaruzzaman M, Taylor TC, Cederlund A, Genkov T, Andersson I, Spreitzer RJ. 2005. Chimeric small subunits influence catalysis without causing global conformational changes in the crystal structure of ribulose-1,5-bisphosphate carboxylase/ oxygenase. Biochemistry 44:9851–9861. Karkehabadi S, Taylor TC, Spreitzer RJ, Andersson I. 2005. Altered intersubunit interactions in crystal structures of catalytically compromised ribulose-1,5-bisphosphate carboxylase/oxygenase. Biochemistry 44:113–120. Kellogg EA, Juliano ND. 1997. The structure and function of Rubisco and their implications for systematic studies. Am J Bot. 84:413–428. Knight S, Andersson I, Bra¨nde´n C-I. 1990. Crystallographic analysis of ˚ ribulose 1,5-bisphosphate carboxylase from spinach at 2.4 A resolution: subunit interactions and the active site. J Mol Biol. 215:113–160. Kopriva S, Chu CC, Bauwe H. 1996. Molecular phylogeny of Flaveria as deduced from the analysis of nucleotide sequences encoding the H-protein of the glycine cleavage system. Plant Cell Environ. 19:1028–1036. Ku MSB, Wu JR, Dai ZY, Scott RA, Chu C, Edwards GE. 1991. Photosynthetic and photorespiratory characteristics of Flaveria species. Plant Physiol. 96:518–528. Kubien DS, Whitney SM, Moore PV, Jesson LK. 2008. The biochemistry of Rubisco in Flaveria. J Exp Bot. 59:1767–1777. Kyte J, Doolittle RF. 1982. A simple method for displaying the hydropathic character of a protein. J Mol Biol. 157:105–132. Laing WA, Ogren WL, Hageman RH. 1974. Regulation of soybean net photosynthetic CO2 fixation by interaction of CO2, O2, and ribulose-1, 5-diphosphate carboxylase. Plant Physiol. 54:678–685. Loytynoja A, Goldman N. 2005. An algorithm for progressive multiple alignment of sequences with insertions. Proc Natl Acad Sci U S A. 102:10557–10562. Loytynoja A, Goldman N. 2008. Phylogeny-aware gap placement prevents errors in sequence alignment and evolutionary analysis. Science 320:1632–1635. McKown AD, Dengler NG. 2007. Key innovations in the evolution of Kranz anatomy and C4 vein pattern in Flaveria (Asteraceae). Am J Bot. 94:382–399. McKown AD, Moncalvo JM, Dengler NG. 2005. Phylogeny of Flaveria (Asteraceae) and inference of C4 photosynthesis evolution. Am J Bot. 92:1911–1928. Meagher RB, Berry-Lowe S, Rice K. 1989. Molecular evolution of the small subunit of ribulose bisphosphate carboxylase: nucleotide substitution and gene conversion. Genetics 123:845–863. Miller SR. 2003. Evidence for the adaptive evolution of the carbon fixation gene rbcL during diversification in temperature tolerance of a clade of hot spring cyanobacteria. Mol Ecol. 12:1237–1246. Panero JL, Funk VA. 2008. The value of sampling anomalous taxa in phylogenetic studies: major clades of the Asteraceae revealed. Mol Phylogenol Evol. 47:757–782. Pasternak JJ, Glick BR. 1992. Molecular evolutionary analyses of the small and large subunits of ribulose-1, 5-bisphosphate carboxylase/oxygenase. Can J Bot. 70:715–723. Pasternak JJ, Glick BR. 1992. Molecular evolutionary analyses of the small and large subunits of ribulose-1, 5-bisphosphate carboxylase/oxygenase. Can J Bot. 70:715–723. Powell AM. 1978. Systematics of Flaveria (Flaveriinae Asteraceae). Ann Mo Bot Gard. 65:590–636. Read BA, Tabita FR. 2002. A hybrid ribulose bisphosphate carboxylase/oxygenase enzyme exhibiting a substantial increase in substrate specificity factor. Biochemistry 31:5553–5560.

MBE Ronquist F, Huelsenbeck JP. 2003. MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19: 1572–1574. Roy H, Andrews T. 2000. Rubisco: assembly and mechanism. In: Leegood R, Sharkey T, Von Caemmerer S, editors. Photosynthesis: physiology and metabolism. Dordrecht (The Netherlands): Kluwer Academic Publishers. p. 53–83. Sage RF. 2002. Variation in the kcat of Rubisco in C3 and C4 plants and some implications for photosynthetic performance at high and low temperature. J Exp Bot. 53:609–620. Sage RF. 2004. The evolution of C4 photosynthesis. New Phytol. 161:341–370. Sawyer S. 1989. Statistical tests for detecting gene conversion. Mol Biol Evol. 6:526–538. Sharwood RE, von Caemmerer S, Maliga P, Whitney SM. 2008. The catalytic properties of hybrid rubisco comprising tobacco small and sunflower large subunits mirror the kinetically equivalent source rubiscos and can support tobacco growth. Plant Physiol. 146:83–96. Shimodaira H. 2002. An approximately unbiased test of phylogenetic tree selection. Syst Biol. 51:492–508. Spreitzer RJ. 2003. Role of the small subunit in ribulose-1,5bisphosphate carboxylase/oxygenase. Arch Biochem Biophys. 414:141–149. Spreitzer RJ, Salvucci ME. 2002. Rubisco: structure, regulatory interactions, and possibilities for a better enzyme. Annu Rev Plant Biol. 53:449–475. Sudderth EA, Muhaidat RM, McKown AD, Kocacinar F, Sage RF. 2007. Leaf anatomy, gas exchange and photosynthetic enzyme activity in Flaveria kochiana. Funct Plant Biol. 34:118–129. Tcherkez GGB, Farquhar GD, Andrews TJ. 2006. Despite slow catalysis and confused substrate specificity, all ribulose bisphosphate carboxylases may be nearly perfectly optimized. Proc Natl Acad Sci U S A. 103:7246–7251. von Caemmerer S, Quick WP. 2000. Rubisco. : physiology in vivo. In: Leegood RC, Sharkey TD, von Caemmerer S, editors. Photosynthesis: physiology and metabolism. Dordrecht (The Netherlands): Kluwer. p. 85–113. Wessinger ME, Edwards GE, Ku MSB. 1989. Quantity and kinetic properties of ribulose 1, 5-bisphosphate carboxylase in C3, C4, and C3-C4 intermediate species of Flaveria (Asteraceae). Plant Cell Physiol. 30:665–671. Yang Z. 1997. PAML: a program package for phylogenetic analysis by maximum likelihood. Comput Appl Biosci. 13:555–556. Yang Z. 2007. PAML 4: phylogenetic analysis by maximum likelihood. Mol Biol Evol. 24:1586–1591. Yang Z, Wong WSW, Nielsen R. 2005. Bayes empirical bayes inference of amino acid sites under positive selection. Mol Biol Evol. 22:1107–1118. Yang Z, Nielsen R, Goldman N, Pedersen AMK. 2000. Codonsubstitution models for heterogeneous selection pressure at amino acid sites. Genetics 155:431–449. Yeoh HH, Badger MR, Watson L. 1980. Variations in Km-CO2 of ribulose-1, 5-bisphosphate carboxylase among grasses. Plant Physiol. 66:1110–1112. Yeoh HH, Badger MR, Watson L. 1981. Variations in kinetic properties of ribulose-1, 5-bisphosphate carboxylase among plants. Plant Physiol. 67:1151–1155. Zamyatnin AA. 1972. Protein volume in solution. Prog Biophys Mol Biol. 24:107–123. Zhang J, Nielsen R, Yang Z. 2005. Evaluation of an improved branchsite likelihood method for detecting positive selection at the molecular levels. Mol Biol Evol. 22:2472–2479.

1503