Large variation in the Rubisco kinetics of diatoms reveals diversity among their carbon-concentrating mechanisms

Highlight Broad variations in the CO2 fixation kinetics of diatom Rubisco indicate novel mechanistic diversity and large differences in their carbon-concentrating mechanism.


Introduction
Ribulose-1,5-bisphosphate carboxylase oxygenase (Rubisco) plays a fundamental role in photosynthetic CO 2 assimilation within the global carbon cycle. Rubisco activity within the terrestrial and ocean biospheres contributes approximately equally to the 10 Pmol of CO 2 annually fixed into organic carbon (Raven, 2009). Often the rate of CO 2 fixation is limited by Rubisco activity and, as such, has made the enzyme a primary target to enhance crop photosynthesis and yield through genetic manipulation (Mueller-Cajar and Whitney, 2008;Peterhansel et al., 2008;Carmo-Silva et al., 2015). However, improving Rubisco kinetics has proved difficult as a result of the complex assembly pathway of Rubisco in higher plants (Whitney et al., 2011;Hauser et al., 2015) and apparent trade-offs in its kinetic parameters (Tcherkez et al., 2006;Savir et al., 2010). Widening our understanding of the natural diversity in Rubisco is critical if solutions to improve its performance are to be found and understood (Parry et al., 2013). Of particular interest is Rubisco from organisms adapted to different environments. This includes marine phytoplankton whose efficient carbon-concentrating mechanisms (CCMs) enable them to endure, sometimes thrive in, nutrient-and CO 2 -depleted seawater (Nelson et al., 1995).
In nature, Rubisco is found in a variety of oligomeric forms and within a diverse array of organisms that include archaea, photosynthetic bacteria, cyanobacteria, algae, and plants (Whitney et al., 2011). Form I Rubisco consists of eight large and eight small subunits, and is subdivided into Forms IA-ID depending on its sequence and lineage (Tabita et al., 2008). Most research to date on Rubisco has focused on those sourced from the terrestrial biosphere, with comparatively little characterization of Rubisco from oceanic sources. The terrestrial biosphere is dominated by the 'green' chloroplast lineage pertaining to plants and green algae that contain Form IB Rubisco (Tabita et al., 2008). However, oceanic photosynthesis is primarily carried out by phytoplankton containing chloroplasts from a 'red' lineage that comprise Form ID Rubisco (Delwiche and Palmer, 1997;Yoon et al., 2002;Falkowski et al., 2004). A group of phytoplankton called diatoms are of particular interest due to their importance in ocean primary productivity (estimated to account for ~20% of global primary production; Nelson et al., 1995), thus influencing global biogeochemical cycles, and for producing silicified walls that preserve paleoclimate signals in the fossil record (Armstrong et al., 2001;Egan et al., 2013;Heureux and Rickaby, 2015).
Compared with other photosynthetic enzymes, Rubisco is considered inefficient due to its low CO 2 -saturated CO 2 fixation rate (k cat c ) and low affinity for CO 2 (i.e. an elevated Michaelis constant, K m , for CO 2 ; K C ). The basis of this inefficiency arises from the complex catalytic mechanism of Rubisco that imposes biochemical trade-offs between k cat c , K C , and specificity for CO 2 over its competitive inhibition by O 2 (S C/O ) (Tcherkez et al., 2006). Recent analyses show that the extent of these trade-offs is variable between the Form I Rubisco of plants, red algae, and cyanobacteria (Tcherkez, 2013(Tcherkez, , 2015. Notably Rubisco oxygenation produces 2-phosphoglycolate, which is toxic to the chloroplast (Zelitch et al., 2009), necessitating its removal via the photorespiratory pathway at a cost of energy and fixed carbon (Peterhansel et al., 2008). The loss of CO 2 by photorespiration can be as high as 25% of the total carbon fixed in C 3 flowering plants (Laing et al., 1974).
Despite the catalytic inefficiencies of Rubisco, it appears that it can adapt to the CO 2 :O 2 ratio of its environment. This is particularly evident for Rubisco kinetics in C 3 and C 4 plants. While C 3 plants rely on diffusion of CO 2 from the air to chloroplast stroma, C 4 plants utilize a CCM to elevate CO 2 around Rubisco to avoid photorespiration and its associated cellular resource costs. In response to higher intracellular CO 2 , C 4 Rubisco has evolved improvements in k cat c at the expense of reducing CO 2 affinity (i.e. increasing K C ) (Yeoh et al., 1980;Seemann et al., 1984;Ghannoum et al., 2005). The CCM and higher k cat c allow C 4 plants to reduce their investment in Rubisco, lower the rate of photorespiration, and allow for carbon fixation rates similar to C 3 plants under low stomatal apertures (Sage, 2004;Ghannoum et al., 2005;Way et al., 2014). These features improve the nitrogen, energy, and water use efficiencies of C 4 plants.
In the oceans, the low levels of CO 2 , and its slow diffusion rate in water, have led many photosynthetic organisms to evolve CCMs that utilize the higher concentrations of bicarbonate. These mechanisms are different from the CCMs of C 4 plants that arose during the low CO 2 concentrations of the Oligocene period (Sage, 2001;Osborne and Sack, 2012) and the coupled warmer, arid environments that trigger stomatal closure and N limitation of the soil (Ehleringer et al., 1997;Long, 1999). The C 4 plant CCM fixes HCO 3 − by phosphoenolpyruvate (PEP) carboxylase, leading to production of C 4 organic acids in the Rubisco-lacking mesophyll cells. These C 4 organic acids then diffuse into the Rubisco-containing bundle sheath cells where they are decarboxylated (Sage, 2004). This process facilitates the concentration of CO 2 to levels that effectively saturate Rubisco (Furbank and Hatch, 1987).
While some diatoms may also have a C 4 -like mechanism (Reinfelder et al., 2000(Reinfelder et al., , 2004 or a C 3 -C 4 intermediate-like mechanism (Roberts et al., 2007), they contain a CCM suited to their single-celled physiology and high bicarbonate aquatic environment. In diatoms, the CCM consists of various bicarbonate transporters (Nakajima et al., 2013) and differing forms of carbonic anhydrase that serve to elevate CO 2 levels within the pyrenoid, a low CO 2 -permeable subcellular compartment containing most of the cellular Rubisco (Reinfelder, 2010;Hopkinson et al., 2011). The efficiency of their CCMs allows diatoms to invest their scarce cellular resources conservatively in Rubisco. Accordingly the Rubisco content of diatoms is considerably lower [2-6% (w/w) of total cellular protein] than the 20-50% (w/w) Rubisco content of the soluble protein in plant leaves (Losh et al., 2013;Carmo-Silva et al., 2015).
Our understanding of phytoplankton CCM components and activity regulation remain rudimentary. This is despite the increasing interest in understanding how Form ID Rubisco and CCMs co-evolved and how carbon fixation rates by phytoplankton will respond to rising anthropogenic CO 2 (Raven et al., 2012;Young et al., 2012Young et al., , 2013. From the few Form ID Rubisco kinetics determined, there is a strong signal of positive selection within the evolution of the Rubisco large subunits in red algae, Haptophytes, and diatoms (Young et al., 2012). Rubiscos from red algae have the highest specificities for CO 2 over O 2 (S C/O ; ~130-240 mol mol −1 ) while the lower S C/O diatom Rubisco (~60-115 mol mol −1 ) overlaps with the less diverse S C/O values of C 3 plant and C 4 plant Rubisco (~70-90 mol mol −1 ) (Read and Tabita, 1994;Whitney et al., 2001Whitney et al., , 2011Haslam et al., 2005). K C measurements for diatom Rubisco (~28-40 μM, Badger et al., 1998;Whitney et al., 2001) exceed the concentration of CO 2 in the surface ocean (~13 μM in air-equilibrated surface seawater at 20 °C), exemplifying the requirement for a CCM. Modeling of the CCM in phytoplankton is highly reliant on measurements of their Rubisco kinetics (Hopkinson, 2011(Hopkinson, , 2014. However, the paucity of catalysis measurements for diatom Rubisco-and other phytoplankton-continue to limit reliable assessments of CCM function in microalgae. In this study, we evaluate the diversity of Rubisco kinetics in 11 different diatom species at the common assay temperature of 25 °C. From measurements of S C/O , k cat c , K C , and the K m for O 2 (K O ), we determine the catalytic turnover rate for O 2 (k cat o ) and unveil an unexpected large degree of kinetic variability across the species studied. Uncovered are novel relationships between kinetic parameters not previously observed for other plant and algal Form I Rubisco isoforms. Presented are a novel, robust data set of Rubisco kinetics in marine phytoplankton that provide new insight into potential constraints on microalgal photosynthesis that arise from variations in the effectiveness of their CCM to elevate CO 2 around Rubisco.

Species selection and sampling:
Eleven species of marine diatoms were selected from cultures maintained at Princeton University and the Australian National University (ANU). Strains from Princeton University were grown at 20 °C under continuous light (~150 μmol photons m −2 s −1 ) in 0.2 μm filtered seawater supplemented with Aquil medium (Sunda et al., 2005)

Rubisco extraction
Cells were harvested at or near exponential growth by gentle centrifugation (2000 g for 10 min) and the 0.1-0.5 ml cell pellets were snap-frozen in liquid nitrogen and stored at -80 °C. Pellets were resuspended in 5 ml of ice-cold extraction buffer containing 50 mM EPPS-NaOH, pH 8.0, 1 mM EDTA, 2 mM DTT, and 1% (v/v) plant protease inhibitor cocktail (Sigma-Aldrich, St Louis, MO, USA), and cells were ruptured in a French press. Extract was centrifuged at 14 000 g, 4 °C for 2 min and the supernatant was used to quantify Rubisco K C and k cat c or used to purify Rubisco and measure CO 2 /O 2 specificity (S C/O ) CO 2 /O 2 specificity Rubisco was rapidly purified from ~1 g (~4-5 pooled biological replicates) of the -80 °C stored cells extracted in ice-cold extraction buffer, and lysed using a French press. Polyvinylpolypyrrolidone [1% (w/v)] was added to the lysate to remove secondary metabolites prior to centrifugation (17 600 g, 4 °C, 5 min). The soluble cellular protein was rapidly passed over a 1 ml Bio-Scale Mini Macro-Prep High Q ion exchange column (Biorad, Hercules, CA, USA) equilibrated with column buffer (50 mM EPPS-NaOH, pH 8.0, 1 mM EDTA). Bound Rubisco was eluted in 1.5 ml of column elution buffer (50 mM EPPS-NaOH, pH 8.0, 1mM EDTA 0.8 M NaCl) and concentrated to 0.5 ml using an Amicon Ultra-15 centrifugal filter (30 000 NMWL, Millipore, Billerica, MA, USA). The protein was applied to the Superdex 200 (GE Life Sciences) column to purify and desalt the Rubisco further. Fractions containing Rubisco were pooled and concentrated again by centrifugal filtration to 0.25 ml and glycerol was added to 20% (v/v) final concentration before freezing in liquid nitrogen and storing at -80 °C. Purified Rubisco preparations were used to measure CO 2 /O 2 specificity using the method of Kane et al. (1994).

Statistics
One-way ANOVA was performed to determine whether significant differences (P-value ≤0.05) existed between the Rubisco kinetics parameters measured in the different groups in Fig. 2. Where significant differences were found, Tukey's HSD tests were performed to determine which groups differed from each other. Linear regression was fit by least-square analysis, and the correlation coefficient (r 2 ) was based on standard error of estimate. Analysis of covariance (ANCOVA) was used to determine whether the linear fit was significant to 95% (P-value ≤0.05) in Table 2 and Figs 3 and 4.

Results
Our results for 11 species of diatoms represent the largest data set of Form ID Rubisco kinetics available to date. The diatom species selected in this study represent a wide range of habitats and evolutionary diversity (Table 1), and included multiple strains of the same species (P. tricornutum and C. calcitrans), and one polar pennate (F. cylindrus).

Rubisco activation and stability in diatom cellular protein extract
The maintained stability of Rubisco activity in isolated soluble leaf protein facilitates the accurate and reliable measure of maximum carboxylation rates (k cat c ) and the Michaelis-Menten (half-saturation) constant (K m ) for carboxylation (K C ) without need of a purification step (Sharwood et al., 2008). These measurements require all eight catalytic sites in each L 8 S 8 Rubisco molecule to be primed with CO 2 -Mg 2+ (i.e. activated). In vivo, full activity is prevented by inhibitory binding of sugar phosphate molecules to the catalytic site (Andralojc et al., 2012). In illuminated leaves, the ~4:1 molar ratio of RuBP:catalytic sites leads to RuBP binding to non-activated catalytic sites being the almost exclusive cause of inactivation (Price et al., 1995). Upon cellular protein extraction, the levels of available RuBP deteriorate, facilitating RuBP dissociation (and fixation) and catalytic site activation. As shown in Fig. 1A (time zero), Rubisco from newly expanded upper canopy tobacco leaves is ~80% activated in vivo. At 25 °C, full activation (i.e. dissociation of all inhibitory RuBP) of tobacco Rubisco in all three leaf samples tested occurred within ~5 min in vitro and full activity was maintained over the 20 min test period.
In contrast to the control tobacco Rubisco in vitro activation assays, the activation status of Rubisco in each phytoplankton species was lower, varying between ~20% (Thalassiosira oceania, T. weissflogii, Cylindrotheca fusiformis, and Phaeodactylum tricornutum) to ~70% (Skeletonema ardens). Accordingly, longer incubation times at 25 °C were required to activate their Rubisco fully in vitro (Fig. 1B-H). Nevertheless, in all phytoplankton samples, Rubisco was fully activated within 10 min of extraction at 25 °C and the activity was stable for at least a further 10 min.

Rubisco kinetics are highly variable both between and within diatom species.
Measurements of k cat c , K C , maximum oxygenation rates (k cat o ), K m for O 2 (K O ), and the specificity for carboxylase over oxygenase (S C/O ) were measured at 25 °C to allow for direct comparison with other Rubisco kinetics in the literature. As shown in Table 1, there is significant variation in diatom Rubisco catalysis, with differences even found between Rubisco from the same genus (Chaetoceros). Also included in these analyses were control catalysis measurements for Rubisco from tobacco (C 3 plant) and maize (C 4 plant) whose values match those previously measured (Supplementary  Table S1 at JXB online). In the following, we compare our results with kinetics measured at 25 °C for other Form ID Rubiscos from red algae and Form IB Rubisco from C 3 and C 4 plants (taken from Badger et al., 1998;Savir et al., 2010).
Among the diatom Rubiscos analyzed, there was a <2-fold variation in k cat c , which ranged from 2.1 ± 0.2 s −1 in Bellerochea cf. horologicalis to 3.7 ± 0.2 s −1 in C. fusiformis. As shown in Fig. 2A, k cat c varied significantly between the groups (oneway ANOVA, F=25.1, P<0.001). Further testing with Tukey HSD showed that diatom k cat c values were comparable with those of Rubisco from C 3 plants but statistically lower than those from C 4 plants (P<0.01) and higher than those from red algae (P<0.01). Diatom Rubisco also showed diversity in the oxygenation rates (Fig. 2B). The measured k cat o values of 0.4-1.6 s −1 were significantly lower than those of C 3 plants, but not lower than those of C 4 plants or red algae (one-way  Fig. 2C) and were significantly higher than those of Rubisco from red algae, and C 3 and C 4 plants (one-way ANOVA, F=19.5, P<0.001, Tukey HSD P<0.01 for all three pairs). Only cyanobacteria have a lower CO 2 affinitty than diatoms, with K C values in the range 200-260 μM, which accords with the effectiveness of their CCM (Price et al., 2008;Whitney et al., 2011;Hopkinson et al., 2014). In contrast, diatom Rubisco K O values were not statistically different from those of red algae, C 3 , or C 4 plants (one-way ANOVA, F=1.44, P=0.26). Of particular interest was the low O 2 affinity of T. weissflogii Rubisco whose K O exceeded 2 mM O 2 .
Improving the S C/O of Rubisco, without unfavorably changing its other kinetic parameters, is a prized goal as it has a pervasive influence on Rubisco efficiency in organisms both with and without CCMs (Long et al., 2015). While the S C/O range in C 3 and C 4 plants shows limited diversity, a 2-fold variation was found in the S C/O of diatom Rubisco at 25 °C (i.e. 57-116 mol mol −1 ; Table 1) and encompassed the range found previously in diatoms (Badger et al., 1998;Whitney et al., 2001;Haslam et al., 2005). Despite this large variation, the S C/O of diatom Rubisco was not significantly different from that of C 3 and C 4 plants, and failed to reach the high S C/O of Rubisco from red algae (Fig. 2E). In terms of carboxylation efficiency (k cat c /K C ), we find that diatom Rubisco is low compared with red algae, and C 3 and C 4 plants (one-way ANOVA, F=11.5, P=0.0002, Fig. 2F, Tukey HSD P=0.04, P=0.001, and P=0.005, respectively), even in the presence of ambient O 2 levels (i.e. k cat c /K C 21%O2 ; Table 1).

Novel kinetic relationships of diatom Rubisco
A number of trade-offs between Rubisco kinetic parameters have been observed across a range of primary producers Fig. 1. Measurement of Rubisco activation status, maximal activity, and stability in vitro at 25 °C. Soluble cellular protein rapidly extracted from tobacco and each phytoplankton in CO 2 -free extraction buffer (containing 5 mM MgCl 2 ) was used to measure changes in the Rubisco 14 CO 2 fixation rate after activating the extract for 0-20 min in buffer containing 15 mM MgCl 2 and 15 mM NaHCO 3 . Gray shading indicates the time when protein extract was assayed to quantify k cat c , K C , and K O (Table 1). Data represent measures from duplicate biological samples (± SD). (Tcherkez et al., 2006;Savir et al., 2010). From the relatively small data set examined, relationships between the varied Rubisco kinetic parameters (i.e. K C , k cat c , k cat o , S C/O , and K O ) appear confined to a one-dimensional landscape, with simple power law correlations between parameters (Savir et al., 2010). We examined these relationships by comparing the kinetic parameters of diatom Rubisco with those measured in green algae, red algae, and plants using data sets of Tcherkez et al. (2006) Supplementary Table  S1).
As shown recently by Tcherkez (2015), the well-documented trade-off between K C and k cat c varies between organisms. For plant and algal Form I Rubisco, the correlation between K C and k cat c varies from that seen for Form I Rubisco from cyanobacteria and other prokaryotes (Tcherkez, 2013). With regard to eukaryotic Form I Rubisco, the diatom variants uniquely show no correlation between K C and k cat c (Fig. 3A). The linear correlation of r 2 =0.36, P=1.1 × 10 -6 between k cat c versus K C for plant and eukaryotic algal Form I Rubisco (Fig. 3A, gray circles), became insignificant when diatom data were included (Fig. 3A, black circles, r 2 =0.013, P=0.36). This suggests that the catalytic mechanism of diatom Rubisco may differ relative to other eukaryotic variants-possibly through changes to one or more of the elemental steps in the Rubisco catalytic cycle.  Table 1) compared with previously published values for red algae (red, maroon dashes), C 3 plants (dark green dashes), and C 4 plants (light green dashes) (Badger et al., 1998;Savir et al., 2010;Supplementary Table S1). Kinetic parameters include (A) the maximum rates of carboxylation (k cat c ), (B) oxygenation rate (k cat o ), the K m for (C) CO 2 (K C ) and (D) O 2 (K O ), (E) the specificity for CO 2 over O 2 (S C/O ), and (F) the carboxylation efficiency (k cat c /K C , mM s −1 ).
A positive linear correlation between K C and K O was found among diatom Rubisco (r 2 =0.43, P=0.041; Fig 3B) that was strengthened when the outlying high K O value (i.e. low O 2 affinity) for T. weissflogii Rubisco was removed (r 2 =0.83, P=5.8 × 10 -4 ). Form 1 Rubisco from other plants and eukaryotic algae also displays a significant positive correlation (r 2 =0.24, P=3.73 × 10 -4 ), and the strength of this relationship increases (r=0.41, P=5.0 × 10 -8 ) when the diatom measurements are included.
Diatom Rubiscos also showed no statistically significant relationship between carboxylase efficiencies and S C/O (r 2 <0.002, P=0.5; Fig 3C). Although diatom carboxylation efficiencies are lower than those of higher plant Rubisco (Fig  2F), the diatom S C/O values are within a similar range, indicating that oxygenase efficiencies must also be correspondingly lower than those of plants, probably due to the lower k cat o rates of diatom Rubisco (Fig 2B).
While a weak positive relationship between k cat o and K O has been shown by Tcherkez (2015), we found that such a correlation did not extend to diatom Rubisco (r 2 =0.38, P =0.078) (Fig. 3D, black circles) or when combined with plant and other eukaryotic algal Rubisco (Fig. 3D, gray circles). In addition, no relationship between k cat c and either K C /K O or S C/O was evident in the diatom Rubisco data set, although for these comparisons our values fall within the noise of the data analyzed by Savir et al. (2010). Linear regressions were tested between all kinetic parameters of diatom Rubiscos and are shown in Table 2.

Rubisco carboxylation efficiency versus Rubisco content
Compared with C 3 plants, the faster carboxylation rates of Rubisco in C 4 plants enable them to invest less of their N resources in Rubisco (Seemann et al., 1984;Ghannoum et al., 2005). We therefore compared the Rubisco content measured in a range of diatoms and the Haptophyte Isochrysis galbana in exponentially growing cells (same strains grown under the same conditions; see Losh et al., 2013) against our measured values of k cat c and K C (Table 1). We found no relationship between Rubisco content and k cat c , but observed a positive correlation between Rubisco content and carboxylation efficiency in the presence of 21% O 2 (k cat c /K C 21%O2 ), which is driven by the negative correlation of Rubisco content with K C 21%O2 (Fig. 4). These relationships are also apparent when Rubisco content is plotted against carboxylation efficiency and K C in the absence of O 2 , but with a lower correlation coefficient.

Discussion
In this study, we demonstrate the large catalytic diversity of Rubisco among 11 diatom species. A unique property of diatom Rubisco is that it lacks the relationship between CO 2 -fixing speed (k cat c ) and CO 2 affinity (K C ) shared by many higher plant and algal Form I Rubisco isoforms. Our measured K C values for diatom Rubisco show larger diversity than observed in contemporary plant Form I Rubisco and, as a result, would require different CO 2 concentrations to saturate Rubisco. Our data suggest that the current estimates of CO 2 concentrations at the site of Rubisco in diatoms are significantly underestimated for some species. In addition, there is an unexpected negative relationship between K C 21%O2 and the Rubisco content in diatoms that suggests a trade-off in the allocation of resources (energy and nutrients) between investing in Rubisco production to sustain rates of photosynthesis for competitive growth or enhancing CCM capacity to sustain saturating CO 2 levels for Rubisco function.

No significant relationship exists between k cat c and K C
A positive relationship between k cat c and K C has been found in all Rubiscos studied to date, although there are increasing indications from wider surveys of Rubisco kinetics (Galmes et al., 2014) that the relationship shared by plant and algal Rubisco differs from that observed for prokaryotic Form I Rubisco (Tcherkez, 2013(Tcherkez, , 2015. Remarkably, diatom Rubiscos in this study show no relationship between K C and k cat c , whereby their C 3 plant-like k cat c values ( Fig. 2A) contrast with atypically high K C values that generally exceed those of C 4 plant Rubisco (Fig. 2C). The lack of relationship is surprising as the trade-off between K C and k cat c is thought to be due to a fundamental mechanistic constraint of their inter-related rate constants (Tcherkez et al., 2006(Tcherkez et al., , 2012. However, differences in the relationships of k cat c and K C between different photosynthetic groups may arise from differences in the intrinsic equilibrium of the RuBP enolization reaction (Tcherkez, 2013(Tcherkez, , 2015. More study is needed into the enolization, CO 2 /O 2 addition, hydration, and cleavage reactions of Rubisco from diatoms (and other microalgae) to understand fully the extent and mechanistic foundation for the contrasting relationships between k cat c and K C found in nature.

High K C values require higher concentrations of CO 2 to saturate
Diatom Rubisco K C values are significantly higher than the concentration of CO 2 in seawater (~10 μM at 25 °C). Like C 4 plants, cyanobacteria, and many other photosynthetic organisms, diatoms are rarely limited by CO 2 for growth because they possess a CCM. The CCM varies between species, but all combine both morphological (such as the pyrenoid, carboxysome, and bundle sheath in diatoms, cyanobacteria, and C 4 plants, respectively) and biochemical (e.g. carbonic anhydrases and bicarbonate transporters) specialization to provide high CO 2 concentrations around the site of Rubisco. In C 4 plants, CO 2 concentrations within the bundle sheath chloroplasts are in excess of 160 μM (i.e. >5000 μbar; Furbank and Hatch, 1987) relative to the subsaturating CO 2 concentrations in C 3 chloroplasts (<10 μM).
Similarly, the CCM of cyanobacteria provides highly saturating CO 2 levels of >400 μM within the carboxysome for their Rubisco (assuming a pH of 7.35 and a 15 mM inorganic carbon pool; Hopkinson et al., 2014;Whitehead et al., 2014). Experimental determination of concentrations of CO 2 and O 2 in pyrenoids of diatoms is currently impossible. This Table 2. Linear correlations between various Rubisco kinetic parameters from 11 diatom species Relationships with statistically significant linear correlation (P<0.05) are shown in bold.
a Correlation when all data are included. This is different from the correlation shown in Fig. 2, which does not include the T. weissflogii outlier.
Downloaded from https://academic.oup.com/jxb/article-abstract/67/11/3445/2197668 by guest on 27 July 2018 has necessitated CO 2 levels being modeled according to conceptual understanding of the diatom CCM and its Rubisco kinetics. Based on a K C 21%O2 of ~41 μM for P. tricornutum Rubisco (Whitney et al., 2001), a pyrenoid CO 2 concentration of ~110 μM was modeled (Hopkinson, 2014). While only 9-fold higher than surface seawater CO 2 concentrations, this CO 2 level would give ~75% saturation of Rubisco carboxylase activity. As shown in Table 1, this assumed K C 21%O2 value is at the lower end of the values measured in our study, indicating that most diatom species would require higher concentrations of CO 2 in the pyrenoid to attain similar saturation. For example, from our measures of K C 21%O2 (37-88 μM) the pyrenoid CO 2 concentrations would need to range between ~148 μM and 352 μM for 80% saturation (or ~92-272 μM in the absence of O 2 ). This adjustment would suggest that CO 2 concentrations in the pyrenoid are similar to, and possibly exceed, those measured in C 4 plant bundle sheath cells; see above and Furbank and Hatch (1987).

Energetic investment in the CCM and Rubisco
The single-cell physiology of phytoplankton and variable nutrient availability of marine ecosystems require their expedient use and probably restricts the energy and resources that can be expended on photosynthesis. The investment of cellular resources in Rubisco synthesis is probably limited to the minimum concentration required for growth (Losh et al., 2013) albeit in balance with the requirements of other cellular processes, such as the high energy costs of a CCM. Notably the resource investment by diatoms in Rubisco [e.g. 2-6% (w/w) of cellular protein; Fig. 3] is much smaller than that in the leaf soluble protein of C 3 plants [25-50% (w/w)] and C 4 plants [10-40% (w/w)] where it accounts for 5-25% of leaf N (Ghannoum et al., 2005;Carmo-Silva et al., 2015). The energy cost of a CCM depends on the capacity of phytoplankton to actively take up inorganic carbon at a rate that is proportional to the diffusive loss of CO 2 from the cell, which is influenced by the internal:external CO 2 gradient and the permeability of the pyrenoid to CO 2 (Hopkinson et al., 2011;Raven et al., 2014;Raven and Beardall, 2016). Diatoms appear well adapted to maintaining near-saturating CO 2 levels around Rubisco (Tortell et al., 2000;Rost et al., 2003;Chen and Gao, 2004;Kranz et al., 2015) by regulating CCM activity in response to varying extracellular CO 2 (Chen and Gao, 2004;Hennon et al., 2015;Young et al., 2015). This 'energy minimization' strategy in CCM regulation and Rubisco production probably provides an N minimization strategy and a significant growth advantage to diatoms (Giordano et al., 2005;Raven et al., 2011).
Our study provides evidence for linkages between Rubisco CO 2 affinity and the efficiency of the CCMs in diatoms. Unlike in C 4 plants where improvements in k cat c correlate with reduced N investment in Rubisco (Ghannoum et al., 2005), diatom Rubisco content was positively correlated with carboxylation efficiency (Fig. 4A) not k cat c . This correlation was governed primarily by the strong negative relationship between Rubisco content and K C (both with and without 21% O 2 ; Fig. 4B). This leads us to hypothesize that diatoms balance resource allocation for photosynthesis between Rubisco content or the CCM. Diatoms such as Thalassiosira and Skeletonema species maintain low Rubisco content but require more resource allocation to their CCM to saturate their low-CO 2 affinity (high K C ) Rubiscos. Alternatively, diatoms such as Phaeodactylum and Chaetoceros species have higher Rubisco content and lower K C values, requiring lower CO 2 concentrations for saturation of carboxylation (Fig. 4). As summarized in Fig. 5, this hypothesis infers an evolutionary trade-off between Rubisco kinetics and content in diatoms; that is, the level of energy invested by diatoms in their CCM to attain CO 2 concentrations suited to their Rubisco CO 2 affinity, not its k cat c , influences resource availability for cellular metabolism, that includes Rubisco synthesis.

Co-evolution of Rubisco and CCMs
Variations in the mechanistic chemistry of diatom Rubisco and plasticity in CCM efficiency among diatom species may explain the non-canonical kinetic features identified for diatom Rubisco (Fig. 3). Although the common ancestor of diatoms was thought to have gained a chloroplast from secondary endosymbiosis of a red alga ~1.2 billion years ago (Yoon et al., 2002), diatoms only began to appear in the fossil record ~200 million years ago (Brown and Sorhannus, 2010). When, or how, the subsequent falling CO 2 :O 2 levels triggered the evolution of marine CCMs remains uncertain (Raven et al., 2012). Notably, although sharing similar Form ID Rubiscos, the S C/O and carboxylation efficiencies of red algae are much higher than those of diatoms and, in some cases, those of higher plants (Whitney et al., 2001). Comparative analyses of the large subunit for a small number of Form ID Rubiscos showed a clear signal of positive selection both between and within the major algal groups (Young et al., 2012). In particular, positive selection was detected along the basal branch leading to diatoms, including between the centric diatom Thalassiosirales and Chaetocerales species, which accords with the very different K C values we observed in this study. Further work is needed to understand adaptation of Form ID Rubisco and to elucidate the amino acid changes in the large and/or small subunits responsible for the wide range of kinetic parameters observed in this study.

A requirement for Rubisco activase in diatoms
A unique outcome of this study was the finding of the low and variable activation status of Rubisco in diatoms (~20-70%, Fig. 1) that implies a requirement for accessory factors for functional maintenance. As in higher plants, deactivation of diatom Rubisco probably results from loss of Mg 2+ and the carbamyl prosthetic group in the catalytic site, allowing for autoinhibition by substrate RuBP binding (to form 'ER' complexes). Likewise, binding of other inhibitory sugar phosphate molecules might occur to fully activated Rubisco (Andralojc et al., 2012). Removal of the sugar phosphate is facilitated by Rubisco activase (RCA) via conformational remodeling driven by ATP hydrolysis (Mueller-Cajar et al., 2014). In nature, differing types of RCA have independently evolved among plants, non-green algae (called CbbX;Mueller-Cajar et al., 2013), and photosynthetic prokaryotes (called CbbOQ; Tsai et al., 2016). While an RCA function in diatoms has not been demonstrated, they encode a chloroplast and nuclear cbbX gene (Kroth, 2015), consistent with our findings of a need for Rubisco activity regulation. The low activation status of diatom Rubisco somewhat parallels the low Rubisco activation status in C 4 plants (~40-60%; von Caemmerer and Furbank, 2003) and in C 3 plants grown at high CO 2 (e.g. it is <50% in tobacco grown in air containing 0.3% v/v CO 2 ; Whitney et al., 1999). Understanding how the regulatory properties of diatom RCA and the activation status of Rubisco differ in response to environmental conditions (e.g. temperature, illumination, CO 2 , and nutrients) and impact the resource use efficiency of diatoms have yet to be examined.

Photorespiration in diatoms
Analogous to other Form I Rubisco (Savir et al., 2010;Galmes et al., 2014), diatom Form ID Rubisco showed a positive relationship between K C and K O (Fig. 3B). This indicates that unwanted reductions in CO 2 affinity (i.e. increasing K C ) are complemented by favorable reductions in O 2 affinity (i.e. increasing K O ). These kinetics help to ensure that the rates of photorespiration are not exacerbated in diatoms. The corresponding effect of specificity for CO 2 over O 2 (S C/O ) is quite variable, spanning values that match or exceed the S C/O of C 3 and C 4 plant Rubisco (Fig. 2E). As highlighted by Whitney et al. (2001), the lower carboxylation efficiencies under 21% O 2 (k cat c /K C 21%O2 ) shared by all diatom Rubiscos relative to plant Rubisco (Table 1) precludes them from being more efficient enzymes within the context of a C 3 plant chloroplast-irrespective of the higher S C/O values measured for some diatom Rubisco variants.
As in other CCM-containing organisms, the rates of photorespiration in diatoms will be suppressed under the higher CO 2 concentrations maintained around Rubisco in the pyrenoid. To date, however, little is known about the photorespiration process and rate in diatoms (Schnitzler Parker et al., 2004;Rigobello-Masini et al., 2012). As indicated above, resolving our understanding of diatom CCMs, and the O 2 levels in the pyrenoid, is vital to assessing the susceptibility of diatoms to photorespiration. One interpretation from the wide range in K O and S C/O values measured between the diatom species is that the CO 2 :O 2 pressures within the pyrenoid may differ dramatically. This may arise through variation in the effectiveness of their CCM to concentrate CO 2 or through reduced permeability to O 2 , or both. Particulaly curious is the elevated O 2 insensitivity (i.e. high K O value) of Rubisco from the centric diatom T. weissflogii. This suggests that there may be different biochemical constraints on the Rubisco in this species worthy of further study.

Conclusions
Primary production in the oceans is dominated by phytoplankton, with diatoms accounting for ~20% of global primary production (Falkowski and Raven, 2007). Here we provide the largest data set of diatom Rubisco kinetics that identifies large Fig. 5. Resource allocation hypothesis for the balance between Rubisco content and K C values. Organisms that invest their energy resources heavily in the CCM (e.g. cyanobacteria) to maintain high intracellular CO 2 levels to saturate Rubisco and limit photorespiration are able to reduce their resource investment in Rubisco. At the other end of the spectrum, organisms without a CCM (e.g. C 3 plants) have Rubisco that is undersaturated with CO 2 and require large resource investments in Rubisco content to maintain adequate rates of carbon fixation for survival. Organisms with a CCM fall somewhere along the saturation curve, depending on the carboxylation speed of their Rubisco and their potential to balance the investment of resources in Rubisco biogenesis suitably or maintain elevated intercellular CO 2 levels around Rubisco. catalytic diversity and novel relationships in their properties. The data suggest that the CCM of diatoms is highly diverse and capable of concentrating CO 2 in the pyrenoid to higher levels than currently envisaged. Our findings also suggest that the CO 2 affinity of diatom Rubisco is a key indicator for how these microalgae manage the allocation of their relatively scarce cellular resources between Rubisco biogenesis and components of their CCM. This work highlights the importance of further studying the phylogenetically diverse, non-terrestrial, Rubisco isoforms to decipher potential mechanistic differences in the catalytic chemistry of the Form I Rubisco superfamily.

Supplementary data
Supplementary data are available at JXB online. Table S1. Rubisco kinetics at 25 °C taken from other data sets and used in Fig. 2