The Minimal CO₂-Concentrating Mechanism of Prochlorococcus spp. MED4 Is Effective and Efficient^,,

A model of a simple CO₂-concentrating mechanism matches physiological data showing that, despite its simplicity, the system is functional and highly efficient.

Abstract

As an oligotrophic specialist, Prochlorococcus spp. has streamlined its genome and metabolism including the CO₂-concentrating mechanism (CCM), which serves to elevate the CO₂ concentration around Rubisco. The genomes of Prochlorococcus spp. indicate that they have a simple CCM composed of one or two HCO₃⁻ pumps and a carboxysome, but its functionality has not been examined. Here, we show that the CCM of Prochlorococcus spp. is effective and efficient, transporting only two molecules of HCO₃⁻ per molecule of CO₂ fixed. A mechanistic, numerical model with a structure based on the CCM components present in the genome is able to match data on photosynthesis, CO₂ efflux, and the intracellular inorganic carbon pool. The model requires the carboxysome shell to be a major barrier to CO₂ efflux and shows that excess Rubisco capacity is critical to attaining a high-affinity CCM without CO₂ recovery mechanisms or high-affinity HCO₃⁻ transporters. No differences in CCM physiology or gene expression were observed when Prochlorococcus spp. was fully acclimated to high-CO₂ (1,000 µL L⁻¹) or low-CO₂ (150 µL L⁻¹) conditions. Prochlorococcus spp. CCM components in the Global Ocean Survey metagenomes were very similar to those in the genomes of cultivated strains, indicating that the CCM in environmental populations is similar to that of cultured representatives.

The marine picocyanobacteria genus Prochlorococcus along with its sister group the marine genus Synechococcus dominate primary production in oligotrophic marine environments (Partensky et al., 1999). Prochlorococcus spp. is an oligotrophic specialist with several key adaptations allowing it to outcompete other phytoplankton in the stable, low-nutrient regions where it thrives. These adaptations include small cell size (less than 1 μm), allowing it to effectively capture nutrients and light, and genome streamlining, which minimizes nutrient requirements (Partensky and Garczarek, 2010). At approximately 1,900 genes, the genomes of high-light-adapted Prochlorococcus spp. are the smallest known among photoautotrophs, suggesting that this is about the minimum number of genes needed to make a cell from inorganic constituents and light (Rocap et al., 2003). Genome reduction has been accomplished by both the loss of entire pathways and complexes, such as the phycobilisomes and many regulatory capabilities, and the paring down of systems to their minimal components, as is the case for the circadian clock and the photosynthetic complexes (Rocap et al., 2003; Kettler et al., 2007; Partensky and Garczarek, 2010).

As part of this genome streamlining, the CO₂-concentrating mechanism (CCM), which enhances the efficiency of photosynthesis by elevating the concentration of CO₂ around Rubisco, has been reduced to what appears to be the minimal number of components necessary for a functional CCM (Badger and Price, 2003; Badger et al., 2006). In typical cyanobacteria, the CCM is composed of HCO₃⁻ transporters, CO₂ uptake systems, and the carboxysome, a protein microcompartment in which Rubisco and carbonic anhydrase (CA) are enclosed. HCO₃⁻ is accumulated in the cytoplasm by direct import from the environment and by the active conversion of CO₂ to HCO₃⁻ via an NADH-dependent process, which constitutes the CO₂ uptake mechanism (Shibata et al., 2001). The accumulated HCO₃⁻ then diffuses into the carboxysome, where CA converts it to CO₂, elevating the concentration of CO₂ around Rubisco (Reinhold et al., 1987; Price and Badger, 1989).

Whereas some cyanobacteria have up to three different families of HCO₃⁻ transporters with differing affinities for use under different environmental conditions, Prochlorococcus spp. has only one or two families (Badger et al., 2006)_. Most cyanobacteria have low-affinity and high-affinity CO₂ uptake systems, but no CO₂ uptake systems are apparent in Prochlorococcus spp. genomes. The carboxysome of Prochlorococcus spp. and other α-cyanobacteria has apparently been laterally transferred from chemoautotrophs, but all of the required components of the carboxysome are present and it is functional (Badger et al., 2002; Roberts et al., 2012). Despite its simplicity, this CCM is likely functional. HCO₃⁻ can be accumulated in the cytoplasm by the HCO₃⁻ transporters and then diffuse into the carboxysome for conversion to CO₂ and subsequent fixation by Rubisco. However, the functionality of the CCM in Prochlorococcus spp. has not yet been tested. Prochlorococcus spp. is a representative of the α-cyanobacteria, a group with distinct CCMs, which have been much less well studied than the CCMs of β-cyanobacteria (Rae et al., 2011, 2013; Whitehead et al., 2014).

We characterized inorganic carbon (C_i) acquisition and processing in Prochlorococcus spp. MED4, examined the effect of long-term acclimation to different CO₂ concentrations on CCM physiology and gene expression, and searched metagenomes for Prochlorococcus spp. CCM genes to determine if CCMs in the natural populations are similar to cultured strains.

RESULTS

Photosynthesis and C_i Uptake versus C_i

Rates of net photosynthesis, CO₂ flux, and HCO₃⁻ flux were measured as C_i was gradually supplied to Prochlorococcus spp. MED4 acclimated to 150 and 1,000 µL L⁻¹ CO₂. The net photosynthesis and HCO₃⁻ flux data were fit with Michaelis-Menten functions to summarize the data (Fig. 1). Net photosynthetic rates began to decline slightly above approximately 1 mm C_i, most likely due to extended time in the assay chamber, so these data were excluded from the fits. The one-half-saturation constant of net photosynthesis for inorganic carbon (K_P) was low (approximately 30 μm C_i) and unaffected by culture CO₂ concentration (Fig. 1; Table I; 95% confidence intervals overlap based on se in the fit). Maximal photosynthetic rates (P_max) were also not significantly different between the CO₂ treatments. The one-half-saturation constant of HCO₃⁻ uptake for inorganic carbon (K_B) was significantly higher than K_P (95% confidence intervals do not overlap), but neither K_B nor the maximal HCO₃⁻ uptake rate (B_max) was significantly different between CO₂ treatments. Net growth rates were not affected by CO₂ (150 µL L⁻¹ CO₂, 0.37 ± 0.05 d⁻¹; 1,000 µL L⁻¹ CO₂, 0.36 ± 0.01 d⁻¹).

Figure 1.

Rates of photosynthesis (P), CO₂ uptake or efflux (Cup), and HCO₃⁻ uptake (Bup) in Prochlorococcus spp. MED4 as a function of C_i concentration for cells grown at 150 µL L⁻¹ CO₂ (A) and 1,000 µL L⁻¹ CO₂ (B). Positive rates indicate uptake into the cell, and negative rates indicate efflux from the cell. C_i concentrations in the culture medium were approximately 1,850 μm at 150 µL L⁻¹ CO₂ and approximately 2,200 μm at 1,000 µL L⁻¹ CO₂.

Table I. Michaelis-Menten fits to photosynthesis and C_i fluxes as a function of C_i and se values in the fitted parameters.

For the net photosynthesis fits, the data at C_i greater than 1 mm were excluded from the fit.

Parameter	150 µL L−1 CO2		1,000 µL L−1 CO2
	Fitted Value	se	Fitted Value	se
KP (μm)	28	8	37	8
Pmax (×10−20 mol cell−1 s−1)	1.60	0.09	1.32	0.06
KB (μm)	82	15	127	23
Bmax (×10−20 mol cell−1 s−1)	2.54	0.12	2.39	0.12

CO₂ concentrations increased above equilibrium during photosynthesis, except at low C_i concentrations (200 μm or less), indicating CO₂ efflux from Prochlorococcus spp. rather than CO₂ uptake. We verified that CO₂ was above equilibrium concentrations by adding bovine CA, which led to a rapid decline in the CO₂ concentration to equilibrium levels (Fig. 2A). In a separate experiment, ¹³C¹⁸O-labeled C_i was added as the C_i source. During photosynthesis, an increase in ¹³C¹⁶O¹⁶O was observed, which indicates that the CO₂ originates from C_i taken up for photosynthesis rather than respiratory ¹²CO₂ and that the C_i has undergone many hydration/dehydration cycles leading to the removal of the ¹⁸O label. CO₂ fluxes were near zero at low C_i (200 μm or less) but became negative (representing net CO₂ efflux) beyond that, reaching their one-half-maximal level at approximately 400 to 500 μm C_i.

Figure 2.

A, Sample data showing increases in CO₂ concentrations upon illumination in an assay chamber containing Prochlorococcus spp. MED4 and 2 mm C_i. That the CO₂ concentration exceeds equilibrium with HCO₃⁻ is illustrated by the addition of bovine CA just after 9 min (dashed line), which establishes equilibrium between CO₂ and HCO₃⁻. Oxygen data show that photosynthesis was not affected by the addition of bovine CA. B, A light/dark cycle with 2 mm ¹³C¹⁸O-labeled C_i added to the assay chamber (darkness = gray background; illumination at 250 μmol photons m⁻² s⁻¹ = white background).

Intracellular C_i Pools

Intracellular C_i pools were estimated from the CO₂ efflux that occurred at the beginning of each dark cycle. These data were highly variable, and there were no significant differences between the two CO₂ treatments, so the data from the two treatments were pooled. The intracellular C_i pool increased as the extracellular C_i increased, reaching a maximal value of approximately 15 mm around 1,000 μm extracellular C_i (Fig. 3A). Cyanobacteria typically have somewhat larger intracellular C_i pools (Kaplan et al., 1980; Badger et al., 1985), but it should be noted that the calculated intracellular C_i pool is highly dependent on the cell diameter used to determine the volume of the spherical cells. We used a cell diameter of 0.7 μm, most commonly reported for the MED4 strain (Bertilsson et al., 2003; Ting et al., 2007), but values as low as 0.5 μm have been used (Morris et al., 2011), which would result in 3-fold lower volume and so 3-fold higher intracellular C_i. A Michaelis-Menten fit to the data had a one-half-saturation constant of 550 ± 210 μm C_i and a saturation value of 21 ± 3 mm. The intracellular C_i pool was linearly related to the rate of CO₂ efflux (Fig. 3B; r² = 0.94, P < 0.001) but was not directly related to the net photosynthetic rate (Fig. 3C). The CO₂ concentration in the carboxysome was calculated from the intracellular C_i concentration by assuming that the cytoplasmic pH was 7.35 (Falkner et al., 1976; Kallas and Dahlquist, 1981; Belkin et al., 1987), that the carboxysome pH is the same as that of the cytoplasm (Menon et al., 2010), and that CO₂ and HCO₃⁻ are in equilibrium in the carboxysome. At seawater C_i (2 mm), the internal C_i pool was approximately 15 mm and the estimated CO₂ concentration in the carboxysome was approximately 500 μm.

Figure 3.

A, Intracellular C_i pool as a function of external C_i concentration as inferred from CO₂ efflux following each light phase. The CO₂ concentration in the carboxysome ([CO₂]_x) was calculated assuming that the intracellular pH is 7.35. A Michaelis-Menten fit to the data had a one-half-saturation constant of 550 ± 210 μm C_i and a saturation value of 21 ± 3 mm. B, Relationship between CO₂ efflux and the intracellular C_i pool. C, Net photosynthesis (NP) of Prochlorococcus spp. MED4 acclimated to 150 µL L⁻¹ CO₂ as a function of the intracellular C_i pool.

Photosynthesis and C_i Uptake versus Irradiance

Rates of net photosynthesis, CO₂ flux, and HCO₃⁻ flux were measured as irradiance was gradually increased on Prochlorococcus spp. MED4 cultures acclimated to 150 µL L⁻¹ CO₂ (Fig. 4; Table II). The net photosynthesis and HCO₃⁻ flux data were fit with Michaelis-Menten functions allowing an offset to account for negative net photosynthesis (respiration) in the dark (zero irradiance). The one-half-saturation of net photosynthesis for irradiance (I_k) was 63 μmol photons m⁻² s⁻¹ and P_max was 1.62 × 10⁻²⁰ mol cell⁻¹ s⁻¹. The one-half-saturation of HCO₃⁻ uptake for irradiance (I_B) was 42 μmol photons m⁻² s⁻¹ and B_max was 2.57 × 10⁻²⁰ mol cell⁻¹ s⁻¹. The maximal rates of net photosynthesis and HCO₃⁻ uptake are indistinguishable from those measured in the photosynthesis versus C_i experiments, showing that rates were light saturated in these experiments, which were conducted at 200 μmol photons m⁻² s⁻¹. Additionally, I_k and I_B are significantly below the incubator irradiance (100 μmol photons m⁻² s⁻¹), showing that these rates are light saturated, or nearly so, under growth conditions.

Figure 4.

Rates of net photosynthesis (P), CO₂ uptake or efflux (Cup), and HCO₃⁻ uptake (Bup) in Prochlorococcus spp. MED4 as a function of irradiance at saturating C_i concentration (1,000 μm).

Table II. Michaelis-Menten fits to photosynthesis and C_i fluxes as a function of irradiance and se values in the fitted parameters.

An offset, representing the irradiance at which rates are zero, was included in the fits. I_k, I_B, and the offsets are in μmol photons m⁻² s⁻¹, and P_max and B_max are in ×10⁻²⁰ mol cell⁻¹ s⁻¹.

Parameter	Fitted Value	se
Ik	63	18
Pmax	1.62	0.22
Photosynthetic rate offset	27	5
IB	42	10
Bmax	2.57	0.17
HCO3− uptake offset	22	5

Rubisco Content and Kinetics

Cellular Rubisco content as measured by quantitative western blots was 8.25 ± 0.7 × 10⁻²² mol Rubisco hexadecamer cell⁻¹ (6.6 × 10⁻²¹ mol active site cell⁻¹) at 150 µL L⁻¹ CO₂. The one-half-saturation constant of Rubisco for CO₂ was 263 ± 5 μm (Supplemental Fig. S1), similar to a previously measured value of 295 μm (Roberts et al., 2012). Both of these K_m values come from measurements made at pH 8, and the Roberts et al. (2012) measurements were made on purified carboxysomes, whereas ours were with a crude extract that likely contained a mix of intact carboxysomes and single Rubisco molecules. They are lower than the value reported for purified Prochlorococcus spp. Rubisco (750 μm) by Scott et al. (2007) made at pH 7.5. These differences may be accounted for by pH, since lower pH reduces the K_m of cyanobacterial Rubisco (Badger 1980), or by differences between intact carboxysomes and purified Rubisco either due to actual changes in K_m of the enzyme or environmental effects. The consequences of uncertainty in K_m for the CCM model are discussed below.

Modeling

A mechanistic model of the Prochlorococcus spp. CCM (see “Materials and Methods”) was parameterized using data obtained in this study and values from the literature (Table III; Fig. 5A). The model was used to assess our understanding of CCM structure based on its ability to match rates of photosynthesis, CO₂ efflux, and internal C_i concentrations observed under culture conditions. While the model structure should be applicable to a broader range of environmental conditions, the specific implementation used here should not be applied without modification to more general environmental conditions, since the model was parameterized based on cultures grown under a narrow range of conditions (constant temperature, light, etc.).

Table III. Model parameters.

Symbola	Definition	Value	Unit	Source
cx	CO2 concentration	Varies	μm
bx	HCO3− concentration	Varies	μm
pHe	Extracellular pH	8	–	Measurement
pHc	Cytoplasmic and carboxysomal pH	7.35	–	“Results”
kcf	CO2 hydration rate constant in cytoplasm	3 × 10−2	s−1	Uncatalyzed; Johnson (1982)
kxf	CO2 hydration rate constant in carboxysome	1,000	s−1	Sufficient for equilibration
mR	Rubsico content	6.6 × 10−21	mol active site cell−1	This study
Km-R	Rubisco one-half-saturation constant for CO2	263	μm	This study
kcat-R	Rubisco maximal turnover rate	10.6	s−1	Tcherkez et al. (2006)
Km-B	One-half-saturation constant for HCO3− uptake	82	μm	This study
Vmax-B	Maximal HCO3− uptake rate	2.5 × 10−20	mol cell−1 s−1	This study
Nx	Number of carboxysomes per cell	6		Ting et al. (2007)
fc-c	Cellular transfer coefficient for CO2	1 × 10−8	cm3 s−1	Assume diffusion limited (membrane is no barrier); Pasciak and Gavis (1974)
fc-x	Carboxysome transfer coefficient for CO2	2 × 10−15	cm3 s−1	Fitted to data
fb-x	Carboxysome transfer coefficient for HCO3−	6 × 10−10	cm3 s−1	Assume diffusion limited (no barrier); Pasciak and Gavis (1974)
Vc	Cytoplasmic volume	1.8 × 10−13	cm3	Ting et al. (2007)
Vx	Total carboxysome volume	2.3 × 10−15	cm3	Ting et al. (2007)

^aSubscripts are as follows: b, bicarbonate; c, cytoplasm; e, external environment; f, forward rate constant; and x, carboxysome.

Figure 5.

Model results. The model structure diagram (A) illustrates the active fluxes (solid arrows), passive fluxes (dashed arrows), and CO₂/HCO₃⁻ interconversion (double-headed arrows) in the CCM, with the parameters controlling these fluxes indicated above each arrow (for notation, see Table III). Agreement of the various models with photosynthesis (P) and CO₂ flux (Cup) data (B) and the internal C_i pool (C) is shown. Inferred carboxysome CO₂ concentrations were calculated as described in “Results.”

The only critical parameter that has no literature constraints and was not measured in this study was the CO₂ transfer coefficient of the carboxysome. This parameter was varied to optimize the fit of the model to observed photosynthetic rates, CO₂ efflux rates, and internal C_i concentrations. The optimized model fit the data quite well and, in particular, captured the major regimes present in the data: a C_i-limited regime (less than 200 μm external C_i) where photosynthesis is increasing with C_i, CO₂ efflux is low, and the internal C_i pool is small; and a C_i-replete regime (greater than 200 μm external C_i) where photosynthesis is saturated for C_i but the internal C_i pool accumulates and CO₂ efflux becomes significant (Fig. 5). Two notable features of this best model are that the carboxysome CO₂ transfer coefficient is low (2.4 ± 0.5 × 10⁻¹⁵ cm³ s⁻¹) and there is excess Rubisco capacity (i.e. the CO₂-saturated Rubisco fixation rate per cell exceeds the observed maximal photosynthetic rate). To show that the best model predicts several features of the data better than other models, we ran two alternative models: the first with no excess Rubisco capacity (low Rubisco in Fig. 5) and the second with a 10-fold higher carboxysome CO₂ transfer coefficient (10× f_c−x in Fig. 5). Both of these alternative models do not fit the data as well, lacking the clear distinction between the two regimes evident in the best model. Instead, in these models, photosynthesis, CO₂ efflux, and the intracellular C_i pool increase gradually until saturation around 1,000 μm external C_i (Fig. 5, B and C).

The absolute values of the modeled intracellular C_i are sensitive to internal pH and to the carboxysome CO₂ transfer coefficient. Because the only loss processes for C_i inside the cell are photosynthesis and leakage of CO₂, the internal pool will build up until these loss rates match the HCO₃⁻ uptake rate. Both these loss rates are dependent on the carboxysome CO₂ concentration, which in turn is dependent on pH and the carboxysome CO₂ transfer coefficient. The intracellular pH used in the model (7.35) was chosen from the range of reported values to obtain reasonable agreement between the modeled and observed absolute values of internal C_i. Nonetheless, agreement between the shape of the modeled intracellular C_i curve and the data is meaningful and distinguishes the best model from alternative models (Fig. 5C).

A sensitivity analysis of the model was conducted, the details of which are presented in “Supplemental Data.” Given the estimated parameter uncertainties for random error, the model is most sensitive to the internal pH and the turnover rate of Rubisco, although even the worst case scenarios for these parameters do not cause major issues with the model fit (Supplemental Text S1; Supplemental Tables S1–S3). There is also systematic uncertainty in the K_m for Rubisco, which potentially ranges between 263 and 750 μm. First, we considered increasing K_m for Rubisco from our measured value of 263 to 520 μm, the value estimated for pH 7.35, assuming that the pH sensitivity is similar to that of Rubisco from Anabaena variabilis (Badger, 1980). This change does reduce model performance, primarily through increasing the internal C_i pool at low external C_i, but the two regimes in the data are still resolved and the overall fit is reasonable (Supplemental Fig. S2). A further increase of K_m for Rubisco to 750 μm, the value obtained by Scott et al. (2007), degrades the model fit further, with the most notable discrepancy being a higher modeled internal C_i pool at low external C_i (Supplemental Fig. S2). There is still some evidence for a distinction between the C_i-limited and C_i-replete regimes, but as K_m for Rubisco increases, the model begins to look more similar to the low-Rubisco model, which lacks excess Rubisco capacity. The sensitivity of the model to K_m for Rubisco is primarily a concern with respect to the effects of pH on K_m for Rubisco. If confinement to the carboxysome imparts a lower apparent K_m for Rubisco through environmental or allosteric effects, then the lower K_m for Rubisco is more appropriate for our model, since such effects are not treated explicitly in the model.

CCM Gene Expression

The expression of putative CCM genes and two housekeeping genes was measured in cultures acclimated to 150 and 1,000 µL L⁻¹ CO₂ (Fig. 6). None of the genes examined was significantly up- or down-regulated by the CO₂ treatments. The CCM genes assessed include components of the carboxysome shell (csoS1, csoS2, and csoSCA), the large and small subunits of Rubisco (cbbL and cbbS), a strong homolog to the low-affinity Na⁺-dependent HCO₃⁻ transporter A (bicA2-1), a weaker homolog to this transporter (bicA2-2), and a weak homolog to the high-affinity Na⁺-dependent HCO₃⁻ transporter A (sbtA2). The two housekeeping genes assessed, whose regulation was not expected to be affected by CO₂, were RNA polymerase σ factor 70 (rpoD) and DNA gyrase subunit A (gyrA).

Figure 6.

Effects of long-term acclimation to different CO₂ concentrations on the expression of genes involved in the CCM of Prochlorococcus spp. MED4 and selected housekeeping genes, as assessed using reverse transcription-qPCR. Although there is some indication that CCM genes are slightly up-regulated at low CO₂, none of the differences are statistically significant (Student’s t test, n = 3). CCM genes include carboxysome shell proteins (csoS1, csoS2, and csoSCA), Rubisco (cbbL and cbbS), and potential HCO₃⁻ transporters (bicA2-1, bicA2-2, and sbtA2). Housekeeping genes are σ factor 70 (rpoD) and DNA gyrase (gyrA).

Prochlorococcus spp. CCM Components in Marine Metagenomes

The Global Ocean Survey (GOS) metagenomes were searched for Prochlorococcus spp. CCM components using a reciprocal best BLAST hit approach. Carboxysome components were all present at reasonable abundances, although some components differed from the frequencies expected from the sequenced Prochlorococcus spp. genomes (Fig. 7). In particular, the abundances of CsoS4A, CsoS4B, and CbbS appear to be more than two times more common in the field than in the sequenced culture genomes. While we cannot rule out the possibility that their higher abundances are real, these genes are the smallest genes examined (CsoS4A, 261 nucleotides; CsoS4B, 246; and CbbS, 333), which leads us to suspect that the normalization procedures artificially increased abundance. A manual inspection of selected BLAST results for these genes confirmed that there was no difficulty in distinguishing Prochlorococcus spp. sequences from those of its closest relative, genus Synechococcus spp. We also searched for components of CO₂ uptake systems, focusing on CO₂ hydration protein X (ChpX) and CO₂ hydration protein Y (ChpY) genes, since these genes are unique to cyanobacterial CO₂ uptake systems (Badger and Price, 2003; Ogawa and Mi, 2007). ChpX and ChpY sequences from marine and freshwater Synechococcus spp. were used to query the metagenomes in this case, since these genes are not present in sequenced Prochlorococcus spp. genomes. Only ChpX genes were found in the GOS metagenomes using a reciprocal best BLAST hit analysis. The mate pairs of these sequences were examined to see if any hit to Prochlorococcus spp. Out of 134 ChpX sequences found, 127 mate pairs had best hits to genes in Synechococcus spp. genomes and three hit to components of a heme uptake system in Prochlorococcus spp. MIT9202, with the remaining mate pairs (four) hitting heterotrophic bacteria.

Figure 7.

Frequency of Prochlorococcus spp. CCM components in genomes (A) and GOS metagenomes (B). Prochlorococcus spp. CCM components in the genomes were identified by BLAST analysis. In the metagenomes, Prochlorococcus spp. CCM components were identified by reciprocal BLAST analysis, and the counts were length normalized relative to RecA and then divided by the mean counts of four single-copy genes to estimate CCM genes per Prochlorococcus spp. genome (gray numbers next to protein names).

DISCUSSION

The CCM of Prochlorococcus spp. Is Functional and Efficient

The K_P was approximately 30 μm, much lower than the approximately 2 mm C_i available in seawater, showing that photosynthesis is fully saturated with C_i in the ocean. This low K_P is similar to that of cyanobacteria with a greater repertoire of HCO₃⁻ and CO₂ uptake systems (Badger and Andrews, 1982; Price et al., 2004; Rae et al., 2011). As discussed in more detail below, neither K_P nor any other parameter was affected by growth at either 150 or 1,000 µL L⁻¹ CO₂, so the two treatments will not be distinguished here. HCO₃⁻ uptake as a function of C_i availability followed Michaelis-Menten kinetics, consistent with a single transporter dominating HCO₃⁻ uptake. The K_B was between 80 and 130 μm, similar to K_B values for BicA transporters from several different cyanobacteria (38–171 μm) and, most notably, close to the K_B for BicA from the marine Synechococcus sp. WH8102 (approximately 75 μm; Price et al., 2004), suggesting that the BicA homolog in Prochlorococcus spp. is the primary HCO₃⁻ transporter. The homolog of the high-affinity SbtA may be active in Prochlorococcus spp., but it would likely only be responsible for a small fraction of overall HCO₃⁻ uptake under the conditions examined, since the K_B of SbtA is much lower (2–16 μm; Shibata et al., 2002; Price et al., 2004). HCO₃⁻ uptake continues to increase after photosynthesis is saturated for C_i, and this additional HCO₃⁻ uptake simply leaks back out of the cell as CO₂. However, it does increase the internal C_i pool, and thus the CO₂ concentration in the carboxysome, which could serve to reduce photorespiration by increasing the CO₂-to-oxygen ratio.

Net CO₂ efflux was observed from the cells at C_i concentrations of 200 μm and above (Fig. 1). As shown by the addition of ¹³C¹⁸O-labeled C_i, the CO₂ originates from freshly transported C_i rather than from respiration, which would produce predominantly ¹²CO₂. The effluxed CO₂ is entirely depleted of ¹⁸O label, indicating that the C_i has undergone many hydration/dehydration cycles prior to exiting the cell (Tchernov et al., 1997). Since the only known CA in Prochlorococcus spp. is in the carboxysome (So et al., 2004), this shows that the leaked CO₂ originates in the carboxysome, where CA removes the ¹⁸O label and increases the CO₂ concentration, driving efflux. Unlike many cyanobacteria that have active CO₂-to-HCO₃⁻ conversion mechanisms to take up and recover leaked CO₂, Prochlorococcus spp. has no known CO₂ recovery mechanisms. The observation of CO₂ efflux from the cell is consistent with this, as is the pattern of ¹⁸O exchange in the light. Upon illumination, cyanobacteria that possess CO₂ recovery mechanisms draw down ¹³C¹⁸O¹⁸O and ¹³C¹⁸O¹⁶O while ¹³C¹⁶O¹⁶O increases, whereas cyanobacteria with inactivated CO₂ recovery systems only show increases of ¹³C¹⁶O¹⁶O, as was observed here for Prochlorococcus spp. (Maeda et al., 2002; Whitehead et al., 2014).

If there are no CO₂ recovery mechanisms, HCO₃⁻ uptake into the cytoplasm is the only active transport step, and the efficiency of the CCM can be estimated as the rate of HCO₃⁻ uptake divided by the photosynthetic rate (Tchernov et al., 1997; Hopkinson et al., 2011). Under growth conditions, which are similar to typical conditions in the ocean, the CCM is remarkably efficient, with only two molecules of HCO₃⁻ transported per CO₂ molecule fixed. For comparison, the diatom Phaeodactylum tricornutum transports 3.5 molecules of HCO₃⁻ per CO₂ fixed (Hopkinson et al., 2011) and Synechococcus sp. WH7803 transports six molecules of HCO₃⁻ per CO₂ fixed (Tchernov et al., 1997), making the Prochlorococcus spp. CCM the most efficient of the few that have been examined.

The Minimal CCM Model Is Consistent with Physiological Data

The simple CCM indicated by an analysis of Prochlorococcus spp. genomes implies that there should be straightforward relationships between the intracellular C_i pool, photosynthesis, and CO₂ efflux. Without a CO₂ recovery mechanism, CO₂ efflux should be linearly related to the CO₂ gradient between the carboxysome and the extracellular solution. Because the CO₂ concentration in the carboxysome was always much greater than the CO₂ in the external solution (maximum of 15 μm), to first order CO₂ efflux should be directly related to the CO₂ concentration in the carboxysome, which is proportional to the intracellular C_i pool. A linear relationship was observed between the CO₂ efflux rate and the intracellular C_i pool, consistent with the simple CCM model (Fig. 3B). However, a closer examination of the data shows that there are two distinct regimes to the data. In the first regime (less than 200 μm external C_i), photosynthesis increases with external C_i while little to no CO₂ efflux is observed and the intracellular C_i pool remains small (Figs. 1 and 3). In the second regime, at higher external C_i (greater than 200 μm), photosynthesis is saturated for C_i but the internal C_i pool continues to increase and CO₂ efflux becomes significant.

The CCM model is able to explain these two regimes and shows that the key characteristics producing these two regimes are excess Rubisco capacity and a low permeability of the carboxysome to CO₂, as illustrated by the failure of alternative models lacking these characteristics to match the data (Fig. 5). In many algae, there is little excess Rubisco capacity, so photosynthesis should follow a Michaelis-Menten relationship with the CO₂ concentration around Rubisco (Badger et al., 1980; Losh et al., 2013; Hopkinson, 2014). In Prochlorococcus spp., CO₂ efflux should be linearly related to the CO₂ concentration in the carboxysome and so would also show a Michaelis-Menten relationship if there was no excess Rubisco capacity. This is clearly not the case here: photosynthesis became saturated at an intracellular C_i concentration of approximately 5 mm when CO₂ in the carboxysome was approximately 200 μm, and only minimal CO₂ efflux was observed (Fig. 3C). Further increase in the carboxysome CO₂ concentration (to 500 μm at 2 mm external C_i) did not lead to higher rates of photosynthesis, meaning that a factor other than CO₂, such as electron transport or ribulose 1,5-bisphosphate (RuBP) regeneration, was limiting carbon fixation in this regime.

Consistent with this, an estimate of the maximal CO₂ fixation rate by Rubisco at saturating CO₂ based on quantitative measurements of Rubisco abundance combined with estimated turnover rates (10.6 s⁻¹ active site⁻¹ and eight active sites per molecule) shows that there was approximately 4-fold excess Rubisco capacity above the measured rates of CO₂ fixation. This excess Rubisco capacity contributes to the low K_P of Prochlorococcus spp. and may explain how the low K_P is achieved in the absence of CO₂ recovery mechanisms or high-affinity HCO₃⁻ transporters. Other cyanobacteria with a greater repertoire of C_i transporters have little excess Rubisco capacity (Whitehead et al., 2014). Once photosynthesis is saturated for CO₂, further HCO₃⁻ import leads to the buildup of the intracellular C_i pool and carboxysome CO₂ concentration, such that the rate of CO₂ leakage matches the rate of HCO₃⁻ import in excess of photosynthesis.

Given the reduced features of the Prochlorococcus spp. CCM, the carboxysome must be a significant barrier to CO₂ efflux. The best model fit was obtained with a carboxysome CO₂ transfer coefficient of 2.4 × 10⁻¹⁵ cm³ s⁻¹ or, approximating the carboxysomes as spheres of 45-nm radius (Ting et al., 2007), a permeability of 1 × 10⁻⁵ cm s⁻¹. The mass transfer coefficient is approximately 4 orders of magnitude lower than the diffusion-controlled transfer coefficient (9 × 10⁻¹⁰ cm³ s⁻¹), obtained assuming that the carboxysome structure does not hinder CO₂ diffusion (Pasciak and Gavis, 1974). A maximal estimate of the carboxysome CO₂ transfer coefficient can be made using our measurements of Rubisco content and one-half-saturation constant for CO₂ combined with literature values for the maximum turnover rate of cyanobacterial Rubisco (Table III) to calculate the minimal CO₂ concentration in the carboxysome (84 μm) required to match the maximal rate of photosynthesis. Since membranes are highly permeable to CO₂, the resistance of the carboxysome should be the major control on CO₂ efflux, in which case the efflux rate can be described by the CO₂ transfer coefficient for the carboxysome (f_c-x) and the CO₂ concentration difference between the carboxysome and the extracellular solution: efflux = f_c-xN_x([CO₂]_x − [CO₂]_e), where N_x = the number of carboxysomes per cell. Using the minimal carboxysome CO₂ concentration and the observed efflux rates, the maximal carboxysome CO₂ transfer coefficient was estimated as 2.4 × 10⁻¹⁴ cm³ s⁻¹, or permeability of 1 × 10⁻⁴ cm s⁻¹, still several orders of magnitude lower than the diffusion-controlled estimate. These best and maximal permeability values are very similar to values that have been found necessary to obtain efficient CCMs in previous models (1 × 10⁻⁵ to 2.5 × 10⁻⁴ cm s⁻¹; Reinhold et al., 1987, 1991). If the carboxysome did not act as a CO₂ barrier, the efficiency of the CCM would be massively reduced, limiting its usefulness. The carboxysome has long been postulated to be a barrier to CO₂ efflux (Reinhold et al., 1987), and analysis of isolated carboxysomes lacking CA has provided support for this hypothesis (Dou et al., 2008), but the simple structure of the Prochlorococcus spp. CCM has allowed us to quantitatively assess the extent to which the carboxysome prevents CO₂ efflux.

Regulation of the CCM by CO₂

In most microalgae, CO₂ availability is the primary factor regulating CCM expression, with decreased CO₂ availability inducing up-regulation of the CCM (Woodger et al., 2005; Matsuda et al., 2011). One common manifestation of CCM up-regulation is a decline in K_P and K_B (Kaplan et al., 1980; Rost et al., 2003). Despite some reduction in K_P and K_B at 150 µL L⁻¹ CO₂, these parameters were not significantly different from their values at 1,000 µL L⁻¹ CO₂, nor were any other physiological characteristics (P_max, B_max, and CO₂ efflux; Table I). At the genetic level, no CCM genes were differentially expressed between the two long-term CO₂ acclimation treatments (Fig. 6). Consistent with this lack of a long-term response to CO₂ treatment is the absence of canonical CCM transcription factors (CcmR and CmpR) in the Prochlorococcus spp. genome (Omata et al., 2001; Woodger et al., 2007; Nishimura et al., 2008). However, Prochlorococcus spp. does have a P_II protein, which contributes to CCM regulation in some cyanobacteria (Palinska et al., 2002), and we did not explore the potential for short-term responses to CO₂ deficiency, which could induce transient changes in gene and protein expression as observed in response to nitrogen starvation in Prochlorococcus spp. (Tolonen et al., 2006). Another α-cyanobacterium, Synechococcus sp. WH5701, also showed no physiological changes when acclimated to different CO₂ concentrations, but some changes in gene expression were observed (Rae et al., 2011). This is not the case for all α-cyanobacteria, as some have strong physiological responses to changes in long-term CO₂ availability (Hassidim et al., 1997; Whitehead et al., 2014).

Some algae are able to take advantage of rising ocean CO₂ concentrations by down-regulating their CCMs and redirecting energy toward increased growth (Rost et al., 2008; Hopkinson et al., 2011). The lack of response to CO₂ suggests that Prochlorococcus spp. cannot take advantage of rising ocean CO₂ concentrations in this way. Consistent with this, the growth of Prochlorococcus spp. does not increase with increasing CO₂ (Fu et al., 2007). Although, in principle, the increased extracellular CO₂ concentration should decrease CO₂ leakage, conserving some energy, the CO₂ concentration in the carboxysome is so high (approximately 500 μm) compared with external CO₂ (5 μm at 150 µL L⁻¹ CO₂ and 32 μm at 1,000 µL L⁻¹ CO₂) that the change in the absolute gradient, which controls the leakage rate, would be small.

Prochlorococcus spp. CCM Genes in the Environment

The set of CCM genes found in Prochlorococcus spp. genomes obtained from cultured isolates is nearly constant, with the exception of some small variations in the carboxysome proteins (Badger et al., 2006; Roberts et al., 2012). The genetic complement and physiological characteristics of isolated Prochlorococcus spp. strains are generally similar to those in the environment, but some differences have been identified (Martiny et al., 2009; Rusch et al., 2010). To determine if the complement of CCM genes found in natural Prochlorococcus spp. populations is similar to that of cultured isolates, we searched the GOS metagenomes for core CCM genes using a reciprocal BLAST approach to ensure that only Prochlorococcus spp. sequences were matched. The natural populations show a similar gene frequency to cultured isolates, suggesting that the CCM in the environment is similar to that in cultured strains (Fig. 7). The inferred high abundance of short carboxysome genes (CsoS4A, CsoS4B, and CbbS) is thought to be an artifact due to the short length of these genes. We also searched the metagenomes for genes diagnostic for CO₂ uptake systems in cyanobacteria (ChpX and ChpY; Shibata et al., 2001; Maeda et al., 2002). Using mate-pair analysis, three sequences were identified that may have been in Prochlorococcus spp., but in all three cases, the mate pairs hit components of a heme uptake system, which itself is thought to be horizontally acquired and rare in Prochlorococcus spp. (Hopkinson and Barbeau, 2012). These results suggest that environmental Prochlorococcus spp., like their cultured representatives, generally cannot take up CO₂, although on extremely rare occasions they may have this ability.

CONCLUSION

The agreement between the CCM model and physiological data suggests we have a good understanding of the mechanics of the Prochlorococcus spp. CCM and, to the extent that it is similar to the CCMs in other cyanobacteria, a good understanding of those CCMs as well. However, some questions about the function and ecological role of the Prochlorococcus spp. CCM remain. The CCM is quite efficient, pumping two molecules of HCO₃⁻ per CO₂ fixed, compared with a few other marine algae that have been studied. This may help Prochlorococcus spp. conserve energy and outcompete other algae, especially deep in the ocean where light intensity is low, but the efficiency of the CCM has only been assessed in a few selected algae. Although this efficiency is impressive, Prochlorococcus spp. shows the potential to be even more efficient. Photosynthesis becomes saturated at low C_i concentrations (approximately 200 μm), when essentially no CO₂ efflux is observed, and the CCM is perfectly efficient in that every HCO₃⁻ molecule imported is fixed. However, as C_i increases further, the cell imports more HCO₃⁻, which is not fixed but instead leaks back out of the cell. The increased rates of HCO₃⁻ uptake do increase the intracellular C_i pool and carboxysomal CO₂ concentration, which would help to reduce photorespiration by further increasing the CO₂-to-oxygen ratio. Nonetheless, the Prochlorococcus spp. CCM is surprisingly effective and efficient despite its simplicity, another example of the remarkable adaptations that allow this simple organism to dominate large parts of the ocean. More generally, this study confirms that the CCMs of α-cyanobacteria, even with the minimal components present in Prochlorococcus spp., function similarly to the well-studied CCMs of β-cyanobacteria (Whitehead et al., 2014).

MATERIALS AND METHODS

Culturing

Prochlorococcus spp. MED4 (CCMP 1986) was obtained from the National Center for Marine Algae and Microbiota and maintained in PRO99 medium using natural Gulf Stream seawater as a base (Moore et al., 2007). The seawater base was sterilized by autoclaving, and the nutrient stocks were 0.2 μm filtered. All culture vessels were acid washed (10% [v/v] HCl) and thoroughly rinsed with deionized water and then ultrapure (greater than 18 MΩ cm⁻¹) water. Cultures were maintained at 18°C and a light intensity of 100 μmol photons m⁻² s⁻¹ on a light/dark cycle of 16/8 h. Growth was tracked by measurement of in vivo chlorophyll fluorescence. Culture CO₂ conditions were controlled by bubbling with premixed CO₂-air mixtures (150 or 1,000 µL L⁻¹ CO₂). To ensure that the bubbling rate was sufficient to achieve equilibrium, medium pH was routinely measured using Thymol Blue (Zhang and Byrne, 1996), and dissolved C_i was measured on occasion using membrane inlet mass spectrometry (MIMS). Cultures were allowed to acclimate for at least 2 weeks to the CO₂ conditions. Cultures were always harvested in midmorning, approximately 4 h after the light period began, to ensure that cells were in a consistent state with respect to diel variability.

Cell Counts

Samples from experimental suspensions were preserved with glutaraldehyde (0.125% [v/v] final concentration), frozen, and stored in liquid nitrogen. Just prior to analysis, samples were defrosted at 35°C and diluted 400-fold in 0.2 μm filtered artificial seawater. Cells were counted flow cytometrically as described previously (Worden and Binder, 2003).

C_i Acquisition

A membrane inlet mass spectrometer (Pfeiffer QMS220) was used to measure net CO₂ and HCO₃⁻ fluxes and photosynthetic oxygen production. Cells were harvested by centrifugation at 11,000g for 15 min, resuspended in 1 to 2 mL of C_i-free artificial seawater, centrifuged again at 7,000g for 3 min, and finally resuspended in 1 mL of C_i-free artificial seawater with 20 mm Tris buffer at pH 8 for experimentation. This cell suspension was then placed in the cuvette of the MIMS system. Fifty micromolar acetazolamide, a CA inhibitor that does not pass through cell membranes, was added to ensure that CO₂ hydration and HCO₃⁻ dehydration rates were at background, uncatalyzed rates. To assess C_i acquisition as a function of C_i availability, any residual C_i in the assay solution was first consumed through photosynthesis by turning on a light-emitting diode light (200 μmol photons m⁻² s⁻¹). ¹³C-labeled C_i was then gradually added back to the sample using alternating dark and light phases to determine net CO₂ and HCO₃⁻ fluxes into or out of the cells and net photosynthesis rates following the methods of Badger et al. (1994). Net photosynthetic rates were determined from the rate of oxygen production. Net CO₂ flux was calculated from the extent to which the CO₂ concentration was drawn down below (uptake) or raised above (efflux) equilibrium. Net HCO₃⁻ uptake was then calculated as the difference between photosynthesis and net CO₂ flux (with positive CO₂ flux indicating uptake and negative indicating efflux). Rates were normalized by cell number as counted using flow cytometry. Michaelis-Menten curves were fit to the net photosynthesis and HCO₃⁻ uptake data.

After each dark cycle, CO₂ concentrations rose above equilibrium levels, which we interpret to be the result of leakage of accumulated intracellular C_i out of the cell. The rate of leakage can be determined by tracking the CO₂ concentration and accounting for net loss due to hydration (Supplemental Fig. S3):

(1)

where k_f is the spontaneous CO₂ hydration rate, [CO₂]_eq is the CO₂ concentration at equilibrium, d indicates differentiation, and t is time. The leakage rate can then be integrated over time until the CO₂ concentration returns to equilibrium, giving the total amount of C_i that was lost from the cell, which is used here as a measure of the intracellular C_i pool (Supplemental Fig S3). This assumes that the main loss process for accumulated C_i is by leakage of CO₂ out of the cell. The approach is similar to that taken by Badger et al. (1985), except that in their experiments, CA was added, so that the CO₂ measurements were a direct measurement of the total external C_i concentration. Recent comparisons of the MIMS method with the more traditional silicone oil centrifugation method found that the approaches were generally similar, but the MIMS method underestimated C_i pools at low external C_i in some cases (Whitehead et al., 2014). The cytoplasmic volume, which is required to calculate the internal C_i concentration, was estimated from the three-dimensional reconstructions of Prochlorococcus spp. MED4 cells of Ting et al. (2007). The total cell volume was calculated assuming that the cell is a sphere of diameter 0.7 μm, and then the thylakoid volume, estimated to take up approximately 30% of the cell volume by analysis of the images in Ting et al. (2007), was subtracted from the total volume to calculate the cytoplasmic volume.

In another set of experiments, the effect of light intensity on C_i uptake was assessed similarly, except that 1 mm C_i was added and light intensity was gradually increased from 25 to 850 μmol photons m⁻² s⁻¹.

Rubisco Quantification and Kinetics

Quantification of Rubisco protein was performed as described by Losh et al. (2013). Briefly, cells were pelleted via centrifugation, and total protein was extracted by vortexing and boiling the pellet for 5 min in SDS buffer (50 mm Tris-HCl, 2% [w/v] SDS, 10% [v/v] glycerol, and 12.5 mm EDTA). Total protein was quantified using the bicinchoninic acid assay against a bovine serum albumin protein standard according to the manufacturer’s instructions (Pierce, Thermo Scientific). Picomoles of CbbL (the large subunit of Rubisco) were determined through quantitative western blotting with global antibodies and standards according to the manufacturer’s instructions (Agrisera). Since there are equimolar concentrations of CbbL and CbbS (the small subunit), total Rubisco protein was calculated from the pmol of CbbL and masses of 52.57 and 12.94 kD for CbbL and CbbS, respectively (Rocap et al., 2003).

The one-half-saturation constant for Rubisco CO₂ fixation was determined at 20°C in crude protein extracts using a ¹⁴C assay. In 2 mL of N₂-sparged gas-tight vials, 10 μg of Rubisco (in 20 μL of crude extract) was added to 500 μL of assay buffer (50 mm Bicine, 20 mm, MgCl₂, 1 mm EDTA, 5 mm dithiothreitol, 0.1 mg mL⁻¹ CA, and 0.4 mm RuBP, pH 8, bubbled with N₂ to remove all CO₂ and oxygen). The reaction was started by the addition of varying amounts of NaH¹⁴CO₃ from 0 to 1 m, and vials were incubated for 4 min. Reaction was stopped by the addition of 0.5 mL of 6 n HCl. Vials were left to degas of inorganic ¹⁴C overnight. Organic ¹⁴C was counted with a scintillation counter (PerkinElmer). To confirm that there was no nonspecific activity, we tested ¹⁴C assays with activated crude extract without RuBP and used these values as blanks

Gene Expression

The relative expression of putative CCM genes at 150 and 1,000 µL L⁻¹ CO₂ was assessed using reverse transcription-quantitative polymerase chain reaction (qPCR). Prochlorococcus spp. cultures were allowed to acclimate to 150 and 1,000 µL L⁻¹ CO₂ as described above, at which point cells were harvested by centrifugation and RNA was isolated using TRIzol (Invitrogen) following the manufacturer’s instructions. The isolated total RNA was treated with 2 units of amplification-grade DNase to remove any potential contaminating DNA. One microgram of total RNA was reverse transcribed using random hexamer primers to complementary DNA (cDNA) with SuperScript III reverse transcriptase (Invitrogen). Parallel reactions were run without reverse transcriptase for use as no-template controls in the qPCR analysis.

Relative abundances of putative CCM genes were quantified by qPCR using a Bio-Rad iCycler iQ. Primers for the CCM genes and controls were designed using the Prochlorococcus spp. MED4 genome sequence and are listed in Supplemental Table S4. qPCR mixtures (25 μL) consisted of 12.5 μL of Bio-Rad iQ SYBR Green 2× Supermix, 0.75 μL of forward and reverse primers (200 nm final concentration), 9 μL of ultrapure water, and 2 μL of cDNA or no-template control. Temperature profiles for the PCR consisted of an initial 10 min at 50°C and then 5 min at 95°C, followed by 45 cycles of 95°C for 10 s (melting) and 30 s at 58°C (annealing and extension), and finally 1 min at 95°C and 1 min at 55°C. Five dilutions of genomic DNA were analyzed for each gene to produce a standard curve for the quantification of relative gene expression based on the cycle at which fluorescence crossed a threshold level. Following each qPCR, a melting-curve analysis was performed, and selected products were run on a 1% (w/v) agarose gel to verify that a single product of the expected length was amplified. No-template controls amplified much later in the reaction than cDNA samples (worst case three cycles later).

Metagenomic Analyses

Putative CCM genes from the 18 publicly available Prochlorococcus spp. genomes were used to search for homologs in the GOS metagenomic data set using a reciprocal BLAST analysis. These genes include components of the carboxysome shell (CsoS1, CsoS2, CsoSCA, CsoS4A, and CsoS4B; National Center for Biotechnology Information accession numbers WP011819929, WP011132186, WP011132187, WP011132188, and WP011132189), Rubisco large and small subunits (CbbL and CbbS; WP011130576 and WP011132185), and putative bicarbonate transporters (BicA2-1, BicA2-2, and SbtA2; WP011131853, WP011132278, and WP011131852). Sets of protein sequences from the Prochlorococcus spp. genomes were compiled and used as query sequences in a tBLASTn search against the GOS metagenomes, keeping hits with E-values less than 1E-5. These metagenome sequences were then used as query sequences in a BLASTx search against a database of 206 marine bacterial genomes (Hopkinson and Barbeau, 2012). If the top BLASTx hit in this search was a Prochlorococcus spp. sequence used in the initial tBLASTn search (i.e. a reciprocal best BLAST hit), then the metagenome sequence was retained as a member of the protein family of interest. The raw sequence counts were scaled by the average length of genes in the family of interest relative to the length of recombinase A (RecA; i.e. length-corrected counts = raw counts × [RecA length/gene of interest length]) to correct for more frequent sampling of longer genes. CCM gene frequency, in genes per Prochlorococcus spp. genome, was calculated by dividing the length-corrected counts by the average number of single-copy Prochlorococcus spp. genes found (DNA gyrase B, HSP70, RecA, and RNA polymerase B). Because the number of counts at most individual GOS sites was low, counts were summed over the entire GOS data set prior to normalization by single-copy gene number.

Modeling

A numerical model of the Prochlorococcus spp. CCM, similar in structure to that of Reinhold et al. (1987), was used to assess the consistency of the data with the streamlined CCM indicated by genomic analysis (Fig. 5A). CO₂ and HCO₃⁻ concentrations are modeled in the cytoplasm and carboxysome, with CO₃²⁻ treated implicitly as part of the HCO₃⁻ pool, since equilibrium between these species is established very rapidly (Zeebe and Wolf-Gladrow, 2001). The only active transport is the import of HCO₃⁻ into the cytoplasm, which is described by Michaelis-Menten kinetics and parameterized from the HCO₃⁻ uptake data. HCO₃⁻ accumulates in the cytoplasm, which lacks CA, and then diffuses into the carboxysome, which is taken to be highly permeable to HCO₃⁻ as a result of positively charged pores in the carboxysome shell (Klein et al., 2009; Kinney et al., 2011). In the carboxysome, CA catalyzes the conversion of HCO₃⁻ to CO₂, elevating the CO₂ concentration around Rubisco. Rubisco fixes CO₂ in the carboxysome at a rate dependent on the CO₂ concentration and capped at the observed maximal CO₂ fixation rate to simulate limitation by another factor such as electron transport rate or RuBP regeneration. CO₂ can diffuse passively out of the carboxysome and cell, as parameterized by transfer coefficients. The transfer coefficient between the cytoplasm and bulk solution was calculated assuming that the cytoplasmic membrane is not a significant barrier to CO₂ flux, as has been found for many other algae and red blood cells (Silverman et al., 1981; Hopkinson et al., 2011). The CO₂ transfer coefficient for carboxysomes has not been determined and instead was optimized to fit the observed rates of photosynthesis, CO₂ efflux, and the internal C_i concentration. The model is described by a system of differential equations:

(2)

(3)

(4)

(5)

where the HCO₃⁻ uptake (B_up) and photosynthetic (P) rates are described as follows:

(6)

(7)

The notation is detailed in Table III.

Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers WP011819929, WP011132186, WP011132187, WP011132188, WP011132189, WP011130576, WP011132185, WP011131853, WP011132278, and WP011131852.

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Figure S1. Rubisco fixation rate as a function of CO₂.

• Supplemental Figure S2. Effect of Rubisco K_m on model-data agreement.

• Supplemental Figure S3. Determination of the internal C_i pool from CO₂ efflux.

• Supplemental Table S1. Model parameters and variability.

• Supplemental Table S2. Results of sensitivity analysis.

• Supplemental Table S3. Interactions between model parameters.

• Supplemental Table S4. qPCR primers.

• Supplemental Text S1. Model sensitivity analysis.

Glossary

CCM

CO₂-concentrating mechanism

CA

carbonic anhydrase

C_i

inorganic carbon

K_P

one-half-saturation constant of net photosynthesis for inorganic carbon

P_max

maximal photosynthetic rate

K_B

one-half-saturation constant of HCO₃⁻ uptake for inorganic carbon

B_max

maximal HCO₃⁻ uptake rate

I_k

one-half-saturation of net photosynthesis for irradiance

I_B

one-half-saturation of HCO₃⁻ uptake for irradiance

GOS

Global Ocean Survey

RuBP

ribulose 1,5-bisphosphate

MIMS

membrane inlet mass spectrometry

qPCR

quantitative polymerase chain reaction

cDNA

complementary DNA