Environmentally driven evolution of Rubisco and improved photosynthesis and growth within the C3 genus Limonium (Plumbaginaceae)



  • Carbon assimilation by most ecosystems requires ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco). Its kinetic parameters are likely to have evolved in parallel with intracellular CO2 availability, with the result that faster forms of Rubisco occur in species with CO2-concentrating mechanisms.
  • The Rubisco catalytic properties were determined and evaluated in relation to growth and carbon assimilation capacity in Mediterranean Limonium species, inhabiting severe stress environments.
  • Significant kinetic differences between closely related species depended on two amino acid substitutions at functionally important residues 309 and 328 within the Rubisco large subunit. The Rubisco of species facing the largest CO2 restrictions during drought had relatively high affinity for CO2 (low Michaelis–Menten constant for CO2 (Kc)) but low maximum rates of carboxylation (math formula ), while the opposite was found for species that maintained higher CO2 concentrations under similar conditions. Rubisco kinetic characteristics were correlated with photosynthetic rate in both well-watered and drought-stressed plants. Moreover, the drought-mediated decrease in plant biomass accumulation was consistently lower in species with higher Rubisco carboxylase catalytic efficiency (math formula /Kc).
  • The present study is the first demonstration of Rubisco adaptation during species diversification within closely related C3 plants, revealing a direct relationship between Rubisco molecular evolution and the biomass accumulation of closely related species subjected to unfavourable conditions.


The photosynthetic processes that are most vulnerable to high temperature and drought remain unclear and somewhat controversial (Salvucci & Crafts-Brandner, 2004; Schrader et al., 2004; Yamori et al., 2010). However, factors related to the performance of a key photosynthetic enzyme, ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco), undoubtedly contribute to the restricted yields associated with elevated temperature and water deficit (Sharkey, 2005; Sage & Kubien, 2007). Rubisco provides a ubiquitous gateway for the entry of inorganic carbon (as CO2) into plant metabolism, yet cannot fully discriminate between CO2 and O2 as it catalyses the chemical combination of either gaseous substrate with an enediolate form of the Calvin cycle intermediate ribulose-1,5-bisphosphate (RuBP). The oxygenation of RuBP gives rise to the energy-dissipating process of photorespiration, and increases steeply in extent with increasing temperature and water stress, reducing plant productivity. Significant progress has been made in understanding the components that determine the fundamental properties of Rubisco at a molecular level (Spreitzer & Salvucci, 2002; Parry et al., 2013), such as the driving forces behind the evolutionary adaptation of Rubisco (Raven, 2000; Tcherkez et al., 2006), the trade-offs involved in determining the magnitude of its kinetic parameters (Savir et al., 2010; Tcherkez, 2013; Galmés et al., 2014), and the underlying structural dynamics (Bracher et al., 2011; van Lun et al., 2011, 2014). Strong evidence for widespread amino acid sequence coevolution and for the role of positive selection in the evolutionary fine-tuning of Rubisco from both C3 and C4 plants has been reported in recent phylogenetic studies (Kapralov & Filatov, 2007; Christin et al., 2008; Kapralov et al., 2011, 2012; Wang et al., 2011; Young et al., 2012). Genetic transformation of Chlamydomonas has helped to elucidate the roles of particular Rubisco domains (Meyer et al., 2012; Esquivel et al., 2013), while new molecular tools have facilitated the modification of Rubisco in planta, enabling the effect of specific amino acid substitutions on Rubisco kinetics to be evaluated (Whitney et al., 2011). However, there are still insufficient data to establish the way in which the kinetic properties of Rubisco impact upon the capacity for CO2 assimilation and the overall plant carbon balance under challenging environmental conditions.

Extreme environments are particularly useful in the study of adaptive evolution in plants. Intense selection pressure not only increases the frequency of adaptation but can also lead to the emergence of alternative survival strategies (Agrawal et al., 2009; Adler et al., 2014). Species of the genus Limonium Mill. (Plumbaginaceae, Caryophyllales) inhabit Mediterranean coastal habitats, which are characterized by low soil osmotic potential as a result of negative water balance and the continuous influx of salt from marine spray. Low atmospheric and soil water potentials require strict control of water loss through transpiration, by means of stomatal closure. Diminished stomatal conductance (gs) is often accompanied by decreased leaf mesophyll conductance to CO2 (gm), which in turn limits the availability of CO2 to Rubisco in the chloroplast stroma, favouring RuBP oxygenation (Flexas et al., 2013). The high ambient temperatures and elevated solar irradiance typical of these xeric environments further increase photorespiration, which may compromise plant carbon balance and, ultimately, survival.

Here we integrate data on Rubisco kinetics, leaf photosynthesis and plant growth under irrigated and severe drought conditions for a number of species of the genus Limonium. Limonium species from the Balearic Islands are an ideal model system in which to study the link between drought tolerance and Rubisco kinetics for several reasons. First, the genus Limonium, with up to 300 species, represents one of the largest radiations in the Mediterranean Basin (Erben, 1993). The Balearic Islands are home to 42 Limonium species, most of them being endemics with a narrow distribution (Erben, 1993; Rosselló & Sáez, 2001) that have evolved as a recent radiation (reviewed in Lledó et al., 2005). Secondly, drought-tolerant Limonium species have adapted to the hostile environments of coastal margins (Galmés et al., 2011a), where biotic competition is virtually absent as a result of extreme abiotic stress factors such as lack of fresh water, high temperature and excess solar radiation. Finally, there is significant diversity in Rubisco kinetics among Limonium, with Limonium gibertii (Sennen) Sennen having among the highest specificities for CO2 reported in terrestrial plants (Galmés et al., 2005). Furthermore, biochemical modelling predicts that significant yield increases would ensue if L. gibertii Rubisco were to replace the native wheat counterpart (Parry et al., 2011).

In the present work, we sampled all Limonium species inhabiting the Balearic Islands and assessed the amino acid sequence polymorphisms in the Rubisco large subunit (L-subunit). Full kinetic characterization of Rubisco was performed for 14 of these species, representing all of the amino acid polymorphisms found. Further, photosynthetic performance and plant growth were assessed for these species under well-watered (WW) and severe water stress (WS) conditions. Our results demonstrate a correlation between drought tolerance, photosynthetic performance and Rubisco catalytic efficiency. We also identified two L-subunit amino acid substitutions that were probably responsible for changes in Rubisco kinetic properties, and assessed the ecological success of Limonium species with different L-subunits.

Materials and Methods

Species sampling

All cited and described Limonium species reliably recognized in the Balearic Islands (Erben, 1993; Rosselló & Sáez, 2001) were included in the survey, resulting in sampling plants from 47 populations which represented 42 Limonium species (Fig. 1). Several populations were sampled for species occurring in different islands, occurring in extremes of an island, or showing notorious differences in plant morphology. The locations of sampled populations are shown in Fig. 1. In addition to the 47 populations of Limonium, we also sampled Myriolimon ferulaceum (L.) Lledó, Erben & M.B. Crespo, the monotypic genus sister to Limonium, which served as an outgroup. Fresh leaf tissue and seeds from five plants per population were sampled in the field and dried in silica gel for further DNA sequencing and the drought experiment, respectively.

Figure 1.

Map of the Balearic Islands showing locations of the sampled populations for each species. Rubisco large subunit (L-subunit) haplotypes are indicated as circles for haplotype I, triangles, apex down for haplotype II, and triangles, apex up for haplotype III. Populations used in the drought experiment are indicated in bold. The outgroup Myriolimon ferulaceum, with an L-subunit haplotype not corresponding to I, II or III, is indicated in grey (*).

rbcL sequences and phylogenetic analyses

Total DNA was extracted from fresh leaves using the DNeasy Plant MiniKit (Qiagen). The rbcL gene coding for the Rubisco L-subunit was amplified and sequenced using the forward primer rbcL1F (5′-ATGTCACCACAAACAGAAACT-3′) and a newly designed reverse primer (5′-AGTATCCATTGCGGCGAATT-3′), following the protocol in Kapralov & Filatov (2006). All polymorphic sites were checked against the original sequence chromatograms and doubtful regions were re-sequenced; sequences were compared with homologues from GenBank, and open reading frame (ORF) integrity was confirmed. Novel sequences have been submitted to GenBank under accession numbers KJ608007KJ608054. Sequences were aligned and phylogenetic reconstructions were performed using maximum likelihood (ML) and maximum parsimony (MP) algorithms, both with 1000 bootstrap resamplings, implemented in mega 5.1 (Tamura et al., 2011), and Bayesian analyses (Markov chain Monte Carlo method), with 106 bootstrap generations, implemented in MrBayes 3.1.2 (Huelsenbeck & Ronquist, 2001). The trees were rooted using M. ferulaceum. All methods produced trees with identical topology but minor differences in bootstrap values.

Identification of Rubisco residues under positive selection

We analysed variation in the ratio of nonsynonymous to synonymous substitution rates (dN/dS) across rbcL codons and/or particular branches of phylogenies using models M8 and A from the codeml program of the paml package v.4 (Yang, 1997, 2007). Posterior probabilities that a certain amino acid belongs to a class with dN/dS > 1 and hence is under positive selection were calculated using the Bayes Empirical Bayes approaches implemented in paml (Yang et al., 2005). Because all Balearic Limonium rbcL sequences were clustered into three groups without nonsynonymous variation within them, we only used one sequence per group plus the sequence of M. ferulaceum. One needs at least seven sequences or preferably more to overcome the problem of false negatives in codeml (Anisimova et al., 2001). Therefore, in addition to four rbcL sequences representing Balearic Plumbaginaceae, we used seven rbcL sequences from GenBank which represented Plumbaginaceae species occurring in communities of drought-resistant plants in regions with a Mediterranean climate similar to that experienced by Balearic Limonium species (Kubitzki et al., 1993): Armeria bottendorfensis A. Schulz (Z97640), Armeria splendens Boiss. (Y16908), Limoniastrum monopetalum Boiss. (Z97642), Limonium sinuatum (L.) Mill. (Y16900), Limonium spectabile (Svent.) G. Kunkel & Sunding (Z97646), Plumbago auriculata Lam. (EU002283) and Plumbago capensis Thunb. (Y16906). Using such a broad outgroup also allowed the establishment of ancestral states for polymorphic amino acid residues. paml analyses were run on an unrooted phylogeny (Fig. 2a; spinach (Spinacia oleracea L.) was not included in the analyses) which was reconstructed using ML and only synonymous sites to avoid distortion caused by possible positive selection on rbcL (Kapralov & Filatov, 2007). Models M8 and A assumed selection across all branches and selected foreground branches only, respectively. Branches leading to all Balearic Limonium haplotypes and to derived haplotypes II and III only were selected as foreground branches in two separate tests for positive selection allowing dN/dS ratios to vary both among sites and among lineages.

Figure 2.

Limonium Rubisco large subunit (L-subunit) sequence variation and structural location of polymorphic residues. (a) Maximum likelihood tree inferred from the rbcL gene coding for the L-subunit for Plumbaginaceae species used in the analyses of positive selection. Spinacia oleracea was added as an outgroup and for comparative purposes; it was not included in the positive selection analysis. Bootstrap values (from 1000 replicates) are indicated at the nodes. Amino acid polymorphism of six Rubisco L-subunit residues that varied in Balearic Limonium is shown on the right. Residue numbers are shown above the table. Residues identical to those of the first sequence are shown as dots. (b) Location of the six polymorphic Rubisco L-subunit residues shown on the structure of spinach Rubisco (1UPM). Two L-subunits forming a functional dimer are highlighted in green and cyan. Residues forming two active sites are shown in orange; six polymorphic residues are shown as green and cyan spheres for L-subunits B and L, respectively.

We also inferred ancestor Rubisco L-subunit protein sequences for each interior node of the analysed Plumbaginaceae rbcL tree using the paml package v.4 (Yang, 1997, 2007) and looked for parallel or convergent amino acid changes, in which changes along independent lineages have occurred from the same or different ancestral amino acids, respectively (Zhang & Kumar, 1997).

Drought treatment and growth analysis

Based on the three different Rubisco L-subunit haplotypes detected within Limonium, the following 14 species were selected for the drought experiment: Limonium balearicum, Limonium barceloi, Limonium companyonis and Limonium echioides representing haplotype I; Limonium ejulabilis, Limonium grosii, Limonium leonardi-llorensii, Limonium magallufianum and Limonium retusum representing haplotype II; Limonium antonii-llorensii, Limonium artruchium, Limonium biflorum, Limonium gibertii and Limonium gymnesicum representing haplotype III. Populations used for each species and their locations are indicated in Fig. 1 (including species authorities). In these populations, seeds were collected from several plants. Thereafter, 10 plants per species were grown from seeds outdoors in 3-l pots with water supplied at field capacity under typical Mediterranean summer conditions. Three months after germination, and to the end of the experiment, five plants per species were subjected to severe water stress by maintaining the pots at 10–30% of field capacity (WS), whereas the remaining plants were maintained at field capacity (WW). Pots were weighed every 2–3 d to enable replacement of lost water. Three months after drought treatment application, plants were harvested to obtain total plant biomass.

Leaf gas exchange and chlorophyll fluorescence measurements

Leaf gas exchange and chlorophyll fluorescence were measured in all species and treatments included in the drought experiment except for L. artruchium and L. gymnesicum, because of their small leaves. Measurements were performed 2 months after the onset of the drought treatment, that is, in leaves completely developed under treatment. The net CO2 assimilation rate (AN) was measured at mid-morning in fully expanded leaves using a gas-exchange analyser equipped with a leaf chamber fluorometer (Li-6400-40; Li-Cor Inc., Lincoln, NE, USA). Leaf chamber conditions consisted of a photosynthetic photon flux density of 1500 μmol m−2 s−1 (with 10% blue light to induce stomatal aperture) and an ambient CO2 concentration of 400 μmol mol−1. Leaf temperature during measurements was maintained at 25°C.

The quantum efficiency of the photosystem II (PSII)-driven electron transport was determined using the equation:

display math(Eqn 1)

(Fs, the steady-state fluorescence in the light; math formula, the maximum fluorescence obtained with a light-saturating pulse (10 000 μmol quanta m−2 s−1) (Genty et al., 1989)). As ϕPSII represents the number of electrons transferred per photon absorbed by PSII, the rate of electron transport (ETR) can be calculated as:

display math(Eqn 2)

(α, the leaf absorbance; β, the distribution of absorbed energy between the two photosystems). The product α·β was determined from the relationship between ϕPSII and math formula obtained by varying light intensity under nonphotorespiratory conditions in a nitrogen atmosphere containing < 1% (v/v) O2 (Valentini et al., 1995). math formula was calculated from gas-exchange measurements as (AN + RL)/PPFD, assuming that the rate of nonphotorespiratory CO2 evolution in the light (RL) was half of the dark mitochondrial respiration rate, Rdark (Galmés et al., 2011b). Rdark was measured at pre-dawn at an ambient CO2 concentration (Ca) of 400 μmol mol−1 and an air temperature of 25°C. The gross CO2 assimilation rate (AG) was obtained as AN + RL.

From combined gas-exchange and chlorophyll a fluorescence measurements, mesophyll conductance for CO2 (gm) was estimated according to Harley et al. (1992) as:

display math(Eqn 3)

The chloroplast CO2 compensation point (Γ*) was calculated from the in vitro Rubisco specificity factor (Sc/o) values for each species as:

display math(Eqn 4)

(O, the partial pressure of oxygen concentration in air (200 mbar)). Finally, from gm estimates, the substomatal CO2 concentration (Ci) values were transformed into chloroplastic CO2 concentration (Cc) as:

display math(Eqn 5)

Rubisco catalytic characterization

Rubisco kinetic parameters were measured in the 14 species used in the growth experiment. Rates of Rubisco 14CO2 fixation using fresh leaf protein extract were measured in 7-ml septum-capped scintillation vials, containing reaction buffer (yielding final concentrations of 100 mM Bicine-NaOH, pH 8.0, 20 mM MgCl2, 0.4 mM RuBP and c. 100 W-A units of carbonic anhydrase) and one of nine different concentrations of CO2 (0–80 μM, each with a specific radioactivity of 3.7 ×1010 Bq mol−1), each at two concentrations of O2 (0 and 21% (v/v)), as described previously (Parry et al., 2007). Assays (1.0 ml total volume) were started by the addition of activated leaf extract, and the maximum velocity for carboxylase activity (Vmax) together with the Michaelis–Menten constant (Km) for CO2 (Kc) determined from the fitted data. The Km for the oxygenase activity was calculated from the relationship math formula . The [O2] was assumed to be 265 μM, but corrected for partial pressure by taking account of the atmospheric pressure and water saturated vapour pressure. Replicate measurements (n = 3–5) were made using protein preparations from two to 10 different leaves of different individuals. For each sample, the maximum rate of carboxylation (math formula) was extrapolated from the corresponding Vmax value after allowance was made for the Rubisco active site concentration, as determined by [14C]CPBP binding (Yokota & Canvin, 1985). Rubisco CO2/O2 specificity (Sc/o) was measured as described previously (Galmés et al., 2005) using enzyme purified by polyethylene glycol (PEG) precipitation and ion exchange chromatography, and the values given for each species were the mean of five to 10 repeated determinations. The maximum oxygenation rate (math formula) was calculated using the equation Sc/o = (math formula /Kc)/( math formula /Ko). All kinetic measurements were performed at 25°C.


The molecular evolution of Rubisco in Balearic Limonium species

The 42 Limonium species (47 populations; Fig. 1) from the Balearic Islands possessed three Rubisco L-subunit amino acid haplotypes (haplotypes I, II and III) occurring in similar island habitats. Haplotype III, found in 29 species, was by far the most frequent in geographical distribution, while haplotypes I and II were found only in five and eight species, respectively (Fig. 1 and Supporting Information Fig. S1). The differences among haplotypes corresponded to polymorphisms at six L-subunit amino acid residues (Fig. 2a). Haplotypes I and II differed in four positions: F10Y, I309M, S328A and D340E; while haplotypes I and III differed in five positions: N95S, G245A, I309M, S328A and D340E (Fig. 2a). Phylogenetically, haplotype I appeared more basal than haplotypes II and III, when eight Plumbaginaceae species were included as the outgroup (Figs 2a, S1).

The implementing of Bayes Empirical Bayes analysis (Yang et al., 2005) revealed that residues 309 and 328 were under positive selection along the branch leading to the Balearic Limonium haplotypes II and III, with posterior probabilities exceeding 0.95 (Fig. 2a). Amino acid substitutions I309M and S328A were unique to haplotypes II and III. No residues were shown to be under positive selection along the branch leading to all Balearic Limonium haplotypes (Fig. 2a). Site-specific tests revealed that residue 340 was under positive selection across all investigated Plumbaginaceae, with a posterior probability of 0.93 (Fig. 2a).

Reconstruction of ancestral sequences for interior nodes was reliable; mean posterior probabilities for the entire protein exceeded 99% for all nodes. It confirmed that amino acid substitutions I309M and S328A occurred only once within analysed phylogeny (Fig. 2a) along the branch leading to the Balearic Limonium haplotypes II and III. Amino acid substitution G245A also occurred only once along the branch leading to the Balearic Limonium haplotype III. However, three other amino acid substitutions within the Balearic Limonium were not unique to this group. Parallel substitution F10Y occurred three times: along the branch leading to Armeria species, along the branch leading to L. spectabile and L. sinuatum, and also along the branch leading to the Balearic Limonium haplotype II. Parallel substitution N95S occurred twice along the branch leading to Plumbago species and the branch leading to the Balearic Limonium haplotype III. Parallel substitution E340D occurred four times: along the branch leading to the Balearic Limonium haplotype I, along the branch leading to L. sinuatum, along the branch leading to L. monopetalum and along the branch leading to the Plumbago species.

Species with different L-subunit haplotypes have different Rubisco kinetics

In order to investigate the impact of L-subunit substitutions on the catalytic properties of Rubisco, full kinetic characterization was undertaken for four to five species per haplotype (Table 1). Significant differences were observed between species. The Michaelis–Menten constant (Km) for CO2 (Kc) varied between 7.0 μM for L. magallufianum and 10.7 μM for L. echioides (Table 1). The species with the lowest Michaelis–Menten constant for O2 (Ko) was L. magallufianum (297 μM), while the highest values were measured in L. leonardi-llorensii (438 μM). The maximum rate of carboxylation (math formula ) varied 2-fold, with the lowest values measured in L. ejulabilis (2.0 s−1) and the highest in L. echioides (3.9 s−1). Less variation was found in the carboxylation catalytic efficiency (math formula/Kc), indicative of the inverse correlation between the rate of carboxylation and the affinity for CO2. The Rubisco specificity factor (Sc/o) ranged between 103.3 mol mol−1 in L. balearicum and 121.2 mol mol−1 in L. gymnesicum (Table 1).

Table 1. Comparison of Rubisco kinetics at 25°C for 14 species of Limonium, representing the three haplotypes identified
HaplotypeSpeciesRubisco kinetic constants
Kc (μM)Ko (μM)math formula (s−1)math formula/Kc (s−1 μM−1)Sc/o (mol mol−1)math formula (s−1)math formula/Ko (s−1 nM−1)
  1. Rubisco kinetic parameters describe the Michaelis–Menten constants for CO2 (Kc) and O2 (Ko), maximum rate of carboxylation (math formula ) and oxygenation (math formula ), carboxylation catalytic efficiency (math formula/Kc), specificity factor (Sc/o), maximum rate of oxygenation (math formula ) and oxygenation catalytic efficiency (math formula/Ko). Data are means ± SE of 3–10 replicates per species (n = 13–37 replicates per haplotype). Different lowercase and uppercase letters denote statistically significant differences by Duncan analysis (P < 0.05) among species and haplotypes, respectively. The maximum rate of oxygenation (math formula ) was calculated using the equation Sc/o = (math formula /Kc)/( math formula/Ko).

I L. balearicum 9.7 ± 0.5ef346 ± 193.6 ± 0.1fg0.38 ± 0.01d103.3 ± 1.5a1.34.6
L. barceloi 9.3 ± 0.6cde346 ± 173.4 ± 0.1f0.37 ± 0.03cd106.3 ± 2.0ab1.24.4
L. companyonis 8.9 ± 0.3bcde429 ± 283.3 ± 0.2f0.37 ± 0.01cd111.8 ± 2.1bc1.44.2
L. echioides 10.7 ± 0.2f427 ± 383.9 ± 0.1g0.37 ± 0.01cd106.4 ± 1.8ab1.54.4
Average9.7 ± 0.2B390 ± 183.6 ± 0.1B0.37 ± 0.01B106.7 ± 1.1A1.4 ± 0.1B4.4 ± 0.1B
II L. ejulabilis 7.6 ± 0.2ab415 ± 232.0 ± 0.2a0.26 ± 0.03a116.0 ± 3.2cd0.92.9
L. grosii 8.1 ± 0.4abcd328 ± 522.9 ± 0.1de0.36 ± 0.01bcd113.2 ± 3.3bc1.04.0
L. leonardi-llorensii 8.8 ± 0.4bcde438 ± 302.8 ± 0.1cde0.32 ± 0.02abc109.7 ± 1.1abc1.33.6
L. magallufianum 7.0 ± 0.3a297 ± 262.6 ± 0.1cde0.37 ± 0.01cd110.6 ± 2.9bc1.04.2
L. retusum 7.1 ± 0.4a396 ± 102.1 ± 0.1ab0.29 ± 0.01a121.0 ± 1.4d1.03.0
Average7.7 ± 0.2A380 ± 172.4 ± 0.1A0.31 ± 0.01A114.3 ± 1.3B1.0 ± 0.1A3.5 ± 0.3B
III L. antonii-llorensii 8.7 ± 0.3bcde397 ± 202.4 ± 0.1bc0.27 ± 0.01a112.1 ± 1.8bc1.13.1
L. artruchium 9.4 ± 0.3def321 ± 503.0 ± 0.1e0.31 ± 0.02abc112.8 ± 2.0bc0.93.5
L. biflorum 8.0 ± 0.3abc316 ± 192.4 ± 0.1bc0.30 ± 0.01ab111.9 ± 1.8bc0.83.4
L. gibertii 9.1 ± 0.7cde431 ± 812.5 ± 0.1cd0.28 ± 0.02a112.1 ± 2.5bc1.03.1
L. gymnesicum 8.2 ± 0.1abcd388 ± 152.4 ± 0.1bc0.29 ± 0.01a121.2 ± 2.3d0.93.0
Average8.8 ± 0.2AB368 ± 252.6 ± 0.1A0.29 ± 0.01A113.9 ± 1.1B0.9 ± 0.1A3.2 ± 0.1A

Significant kinetic differences were also found between the three haplotypes (Table 1), suggesting that the observed amino acid substitutions had catalytic consequences. Haplotype I Rubisco was faster (higher math formula), but less specific for CO2 (higher Kc and lower Sc/o) compared with haplotypes II and III (Table 1). Based on its higher value for math formula/Kc, haplotype I represented superior carboxylation efficiency.

Highly significant relationships were found between the maximum rates of carboxylation (math formula) and oxygenation (math formula) (Table 1; r2 = 0.67; < 0.001), as well as between the carboxylase (math formula/Kc) and the oxygenase (math formula/Ko) catalytic efficiencies (r2 = 0.98; P < 0.001). The greater kcat and Km values of haplotype I than of haplotypes II and III were paralleled by greater catalytic efficiencies (kcat/Km) for both the carboxylase and oxygenase reactions (Table 1). However, the relative change in catalytic efficiency for the oxygenase activity always exceeded that for the carboxylase activity, leading to a decline in Sc/o. Hence, math formula was negatively correlated with the affinity for CO2 (i.e. 1/Kc) and Sc/o (Fig. 3). In other words, slower Limonium Rubiscos have increased specificity for CO2. Overall, the kinetic data presented in this study are in agreement with the kinetic interrelationships described in the analysis of Savir et al. (2010).

Figure 3.

Relationship between Rubisco kinetic parameters for representative species of Balearic Limonium. Correlation between the maximum rate of carboxylation (math formula ) and (a) the specificity factor (Sc/o) and (b) the Michaelis–Menten constant for CO2 (Kc) is shown. All measurements were made at 25°C. Circles, haplotype I; triangles, apex down, haplotype II; triangles, apex up, haplotype III. Data are means ± SE (n = 3–10 replicates).

The diversity in Rubisco kinetics was related to the availability of CO2 and influenced the photosynthetic rate and biomass accumulation

The concentration of CO2 in the chloroplast (Cc), estimated from in vivo measurements of leaf photosynthesis, was affected by water availability in most, but not all the species. Hence, the drought-induced decrease in Cc between WW and WS plants (CcWWCcWS) was bigger in species with haplotypes II and III, and smaller in species with haplotype I Rubiscos (Fig. 4). Note that CcWWCcWS correlated negatively with math formula (r2 = 0.67; P < 0.001) and positively with Sc/o (r2 = 0.72; P < 0.001). Hence, species with haplotype I Rubisco, which showed the smallest change in Cc between WW and WS plants, had higher math formula and lower Sc/o.

Figure 4.

Adaptation of Rubisco kinetic parameters of Limonium species according to the availability of CO2 at the site of carboxylation. The relationship of the difference in the CO2 concentration at the site of carboxylation between well-watered (CcWW) and water-stressed plants (CcWS) and (a) the maximum rate of carboxylation (math formula) and (b) the specificity factor (Sc/o) for the 12 Limonium species for which in vivo gas exchange was measured is shown. Circles, haplotype I; triangles, apex down, haplotype II; triangles, apex up, haplotype III. Both in vitro and in vivo measurements were performed at 25°C. Data are means ± SE (n = 3–10 replicates).

The application of severe drought stress affected both the photosynthetic capacity and biomass accumulation (B) in a species-specific manner (Fig. 5). Within each irrigation treatment, the variability in the light-saturated, gross CO2 assimilation rate (AG) among species correlated with differences in Rubisco kinetics. Hence, species with higher AG had Rubisco with higher math formula under both WW and WS (Fig. 5a). In agreement with the trade-off between math formula and Sc/o (Fig. 3a), AG was inversely correlated to Sc/o (Fig. 5b). Under WW conditions, the observed trend in the relationship between B and the same Rubisco kinetic parameters was the inverse of that with AG, showing a negative correlation with math formula and a positive correlation with Sc/o (Fig. 5c,d, respectively). However, the correlation between AG and B was not significant, either under control or under stress conditions. This apparent discrepancy serves to illustrate that isolated CO2 exchange rates measured on individual, light-saturated leaves are not necessarily representative of the carbon flux at either whole-plant or canopy scales over the applicable growth period (120 d), particularly in plants with dense cushion-like architecture and considerable self-shading.

Figure 5.

Influence of Rubisco biochemistry on in vivo leaf photosynthesis and plant growth capacity in Limonium. The relationships between (a) gross CO2 assimilation rate (AG) and the maximum rate of carboxylation (math formula), (b) AG and the specificity factor (Sc/o), (c) dry biomass accumulation (B) and math formula, and (d) B and Sc/o are shown for the 12 Limonium species for which in vivo gas exchange was measured, under well-watered (filled symbols, solid line) and water-stress (empty symbols, dashed line) conditions. Circles, haplotype I; triangles, apex down, haplotype II; triangles, apex up, haplotype III. Both in vitro and in vivo measurements were performed at 25°C. AG was obtained by adding half of the dark respiration rate to the net CO2 assimilation rate. Data are means ± SE (n = 3–10 replicates). n.s., not significant.

The Rubisco carboxylase catalytic efficiency (math formula/Kc) has frequently been referred to as an indicator of enzyme performance. There was a positive correlation between math formula/Kc and the ratio of plant biomass under WS and WW conditions (Fig. 6), suggesting that the detrimental effects of drought on plant carbon gain were less in those species with higher math formula/Kc.

Figure 6.

Relationship between carboxylase catalytic efficiency (math formula /Kc) and the ratio of biomass accumulation under water stress and well-watered treatments (BWS/BWW) for the 12 Limonium species for which in vivo gas exchange was measured. Circles, haplotype I; triangles, apex down, haplotype II; triangles, apex up, haplotype III. Data are means ± SE (n = 3–10 replicates).


The present study shows that the stressful environments of the Mediterranean coast facilitated diversification of Rubisco within the genus Limonium, giving rise to three different Rubisco haplotypes coexisting in the Balearic Islands. Significantly different Rubisco kinetic properties among haplotypes were related to the differences in the concentration of CO2 within the chloroplast stroma, and correlated not only with the photosynthetic CO2 assimilation rate, but also with the response of plant growth to severe water stress. Rubisco L-subunit amino acid substitutions at positions 309 and 328 are likely to be responsible for the observed differences in kinetic properties, leading either to more CO2-specific (low Kc) but slower (low math formula) enzymes, or to less CO2-specific (high Kc) but faster (high math formula) forms of Rubisco. Overall, this work represents the first integrative approach linking natural polymorphisms occurring in Rubisco, its kinetic parameters, in vivo leaf photosynthesis and whole-plant growth, under specific environmental conditions.

Three of the observed amino acid substitutions were unique to the Balearic Limonium species, while the other three occurred as parallel substitutions along other branches too. The presence of the parallel amino acid substitutions is not surprising, given that all included Plumbaginaceae are drought-resistant species from Mediterranean climates with relatively similar genetic backgrounds. However, the existence of six polymorphic residues, three of which are unique to the western Mediterranean Limonium species, is surprisingly high given the conservative nature of Rubisco and the short period of divergence of Balearic Limonium species. Complete plastome replacements via interspecific introgression have been documented for numerous plant groups (reviewed in Avise, 2004), and strong positive selection acting on chloroplast genes may facilitate this process. Interspecific hybridization is common in the Balearic Limonium species (Erben, 1993) resulting in cytonuclear discordance (M. À. Conesa et al., unpublished data). Positive selection acting on the rbcL gene might have caused the three observed haplotypes to spread across all Balearic Limonium species. A similar situation was described for another island group, the Hawaiian Schiedea (Kapralov & Filatov, 2006), showing that evolution in small populations within confined geographical regions, such as islands, might be facilitated by the sharing of adaptive mutations by several closely related species.

Unlike Flaveria (Kubien et al., 2008), which contains C3 and C4 species, all Limonium species are strictly C3, and therefore the presence of a carbon-concentrating mechanism cannot be the factor driving Rubisco diversification, although there could be significant variation in Cc between C3 plants (see below). Two amino acid substitutions at positions 309 and 328 were under positive selection along the branch leading to the Limonium species with derived haplotypes II and III, and gave rise to different Rubisco kinetics compared with the basal haplotype I. Phylogenetic analyses of the chloroplast rbcL gene encoding the Rubisco L-subunit showed that residues 309 and 328 were among the most frequently subjected to positive selection in both C3 and C4 plants (Kapralov & Filatov, 2007; Christin et al., 2008; Kapralov et al., 2011, 2012). Hybrid Rubisco assembling in tobacco (Nicotiana tabacum) plastids established that substitution M309I acted as the catalytic switch that caused the kinetic differences (similar to those reported here) between C3 (309Met) and C4 (309Ile) Flaveria species (Whitney et al., 2011). Residue 309 is at the C-terminal end of a short β-strand (βF) at the C-terminal side of the α/β-barrel between strand 5 and helix 5 of the barrel (Fig. 2b, Table S1). This location places residue 309 on the interface of L-subunits forming a catalytic L2 dimer. In the Limonium species with haplotype I (higher math formula ), residue 309 is Ile, but in haplotypes II and III (lower math formula ), it is replaced by the larger and less hydrophobic 309Met (Table S1). Residue 328 is located within the loop-6 region of the α/β-barrel, which is close to the active site and has been shown to be important for catalysis and specificity (Chen & Spreitzer, 1989; Parry et al., 1992). In the Limonium species with haplotypes II and III, polar and uncharged 328Ser was replaced by the nonpolar and hydrophobic 328Ala (Table S1) – a change that in other species was accompanied by an increase in Sc/o (Chen & Spreitzer, 1989; Parry et al., 1992).

The finding that haplotype I Rubiscos (309Ile and 328Ser) had higher math formula (Table 1), albeit a lower affinity for CO2 (i.e. lower Sc/o and higher Kc) compared with the other haplotypes, is in agreement (1) with the observation of frequent positive selection at these sites (Kapralov & Filatov, 2007); and (2) with the fact that 309Ile and 328Ser are found in many (although not all) C4 plants characterized by high math formula and low affinity for CO2 (Christin et al., 2008; Kapralov et al., 2011, 2012). Thus, 309Ile and 328Ser are likely to be among the catalytic switches for faster and less specific Rubiscos in many groups of C3 and C4 plants. However, their effect on the kinetic properties may vary, depending on amino acid identity at other positions, as coevolution (the attainment of similar properties by means of alternative amino acid substitutions) has been shown to be a common phenomenon in Rubisco (Wang et al., 2011). Expression of recombinant Limonium Rubiscos in tobacco, incorporating directed mutations at these positions, may confirm their specific role in Rubisco catalysis, similar to the recent study involving Flaveria (Whitney et al., 2011).

Although kinetic differences among haplotypes are significant, there are cases when species belonging to different haplotypes have similar kinetics despite having different Rubisco L-subunit sequences, as well as species within the same haplotype (i.e. identical L-subunit sequences) having different kinetics. One of the possible explanations for this phenomenon is differences in small subunit (S-subunit) identity and the combinations of distinct S-subunits within the Rubisco holoenzyme. Although Rubisco active sites are located in the L-subunits, S-subunits can also affect catalysis (Spreitzer, 2003). Ishikawa et al. (2011) showed that significant changes in Rubisco kinetics occurred when the rice (Oryza sativa) S-subunit was replaced with the sorghum (Sorghum bicolor) S-subunit. More recently, Morita et al. (2014) showed that Rubisco catalytic properties were influenced by the differential expression of S-subunits in rice. Hexaploid wheat, Triticum aestivum, possesses at least 22 rbcS genes encoding different S-subunits (Galili et al., 1998). Many Balearic Limonium species are also recent polyploids (Erben, 1993), and may also possess large numbers of divergent rbcS genes.

The kinetic differences among haplotypes are explained on the basis of the inverse relationship between Rubisco velocity and CO2 affinity (Fig. 3). This trade-off confirms that Rubisco evolution is flexible but essentially one-dimensional (Savir et al., 2010), that is, that Rubisco may specialize towards either increased math formula or increased Sc/o, because concomitant optimization of velocity and specificity is structurally impeded (Tcherkez, 2013). Among the potential forces driving the evolutionary adaptation of Rubisco, the availability of CO2 for RuBP carboxylation has been crucial (Raven, 2000; Tcherkez et al., 2006). This trend was confirmed in Limonium and other Mediterranean species inhabiting extremely xeric environments, where the low soil and atmospheric water potentials require the operation of Rubisco at low Cc for most of the growing season (Galmés et al., 2013). Under these highly selective conditions, Rubisco has evolved to increase its specificity for CO2 (Galmés et al., 2005). The results in the present study clearly confirm the role exerted by Cc in shaping Rubisco kinetics. Long-term growth of Limonium species under severe drought stress, similar to the environmental conditions of their natural habitat, caused a significant fall in Cc in most species (Fig. 4). Under these stressful conditions, species showing the largest decrease in Cc have a more CO2-specific Rubisco, but a lower math formula, as a consequence of the trade-off between these properties.

On average, the fast and less CO2-specific haplotype I Rubiscos provided higher photosynthetic capacity, although this did not translate into higher whole-plant growth (Fig. 5). This apparent discrepancy can be explained in terms of the contrasting limitations governing the CO2 assimilation rate (A) at the leaf and whole-plant levels. Leaf-level A, which was measured at saturating light intensities during discrete gas exchange measurements, was limited by Rubisco (Ac-limited photosynthesis) in all studied species (data not shown). According to the biochemical model of leaf photosynthesis (Farquhar et al., 1980), under Ac-limited photosynthesis math formula – as part of Rubisco maximum carboxylation rate, Vcmax – has a decisive role in determining A. By contrast, plant biomass is the integral of temporal and spatial CO2 assimilation and respiration processes. Given the particular plant habit of Limonium plants, at any given moment of the day, a large proportion of the leaves in the canopy are likely to be exposed to subsaturating sunlight. Under these conditions, whole-plant photosynthesis is predominantly limited by RuBP regeneration (Aj-limited photosynthesis), and explains why plant biomass (B) is positively correlated to Sc/o (Fig. 5d). B did not appear to be significantly correlated to either math formula or Sc/o under WS (Cc decreased) but appeared to be so under WW conditions. If the ratio BWS/BWW is indicative of whole-plant performance under drought, it is evident (Fig. 6) that Limonium Rubisco with higher carboxylase catalytic efficiency (math formula/Kc) confers greater resilience to such severe stress conditions, as demonstrated by species from haplotype I (Fig. 6).

Our study is the first to use an interdisciplinary approach to investigate the relationship between Rubisco amino acid polymorphism and catalytic efficiency, plant photosynthetic performance, growth and drought tolerance in a group of closely related C3 species. A wide-ranging assessment of the natural diversity of Rubisco-encoding genes, together with Rubisco kinetics and its consequences for plant photosynthesis and growth under contrasting environmental conditions, will be required if knowledge-based attempts to increase global crop yields by harnessing superior forms of Rubisco are to succeed (Parry et al., 2013).


This research was supported by project AGL2009-07999 (Plan Nacional, Spain) awarded to J.G. P.J.A. and M.A.J.P. are supported by the BBSRC 20:20 Wheat® Institute Strategic Program (BBSRC BB/J/00426X/1) and BBSRC BB/I002545/1 and BB/I017372/1.