Minimal cobalt metabolism in the marine cyanobacterium Prochlorococcus

Significance

Photosynthetic phytoplankton are the foundation of marine ecosystems. Their growth in the sunlit ocean depends on ample supply of over a dozen essential elements. Of these elemental nutrients, the metal cobalt is found at the lowest concentrations in seawater, but it is unknown whether cobalt scarcity impacts phytoplankton growth. We have measured minimum cobalt requirements of the photosynthetic bacterium Prochlorococcus, which flourishes in nutrient-poor regions of the ocean where many other phytoplankton cannot survive. Prochlorococcus can grow with less than 50 cobalt atoms per cell, an extraordinarily small requirement that explains how this organism can persist in low-cobalt environments. These results enable predictions of how marine ecosystems respond to climate-driven changes in nutrient supply.

Abstract

Despite very low concentrations of cobalt in marine waters, cyanobacteria in the genus Prochlorococcus retain the genetic machinery for the synthesis and use of cobalt-bearing cofactors (cobalamins) in their genomes. We explore cobalt metabolism in a Prochlorococcus isolate from the equatorial Pacific Ocean (strain MIT9215) through a series of growth experiments under iron- and cobalt-limiting conditions. Metal uptake rates, quantitative proteomic measurements of cobalamin-dependent enzymes, and theoretical calculations all indicate that Prochlorococcus MIT9215 can sustain growth with less than 50 cobalt atoms per cell, ∼100-fold lower than minimum iron requirements for these cells (∼5,100 atoms per cell). Quantitative descriptions of Prochlorococcus cobalt limitation are used to interpret the cobalt distribution in the equatorial Pacific Ocean, where surface concentrations are among the lowest measured globally but Prochlorococcus biomass is high. A low minimum cobalt quota ensures that other nutrients, notably iron, will be exhausted before cobalt can be fully depleted, helping to explain the persistence of cobalt-dependent metabolism in marine cyanobacteria.

Cobalt is the least abundant elemental nutrient in seawater, but this may not have always been the case. In the modern ocean, cobalt scarcity develops from the oxidation of soluble Co²⁺ to insoluble Co³⁺ by Mn-oxidizing bacteria (1, 2), leaving behind a small pool of cobalt bound to strong organic ligands. Mn- and Co-oxidizing reactions are inhibited at low O₂, leading to the buildup of cobalt observed in oxygen-depleted waters (3–6). Similar reasoning has suggested that much higher cobalt inventories were found prior to the oxygenation of the oceans ∼300 million years ago (7, 8). Greater access to cobalt in the ancient anoxic ocean may have encouraged cobalt use in metalloenzymes, which evolved early in Earth’s history (9, 10).

Since the oxidation of the oceans, there has been evolutionary selection against cobalt use over time, consistent with its decreasing availability (7, 11, 12). Eukaryotic genomes use cobalt less frequently than those of bacteria and archaea, and all eukaryotic organisms lack the ability for the de novo synthesis of cobalamin (vitamin B₁₂) from cobalt ion (13–15). The adoption of cobalamin-independent isoforms of methionine synthase, ribonucleotide reductase, and other enzymes has liberated land plants from cobalt dependence altogether (14, 16). Despite its scarcity and the availability of substitutes, there are benefits to cobalt use, exemplified among marine phytoplankton. For instance, 100-fold faster catalysis by cobalamin-dependent methionine synthase, MetH, compared to cobalamin-independent methionine synthase, MetE, allows for economization of nitrogen and other limiting resources (15, 17). Furthermore, Co²⁺ ions can activate Zn-dependent metalloenzymes like carbonic anhydrase, improving growth under Zn scarcity (18–20).

In the surface ocean, light inhibits bacterial Mn oxidation (21), but cobalt uptake by phytoplankton leads to the lowest concentrations at any depth, which can fall below 10⁻¹¹ mol L⁻¹ (10 pM; refs. 3, 6, and 22). There is growing evidence that the addition of cobalt or cobalamin to surface waters can increase phytoplankton growth rates when added in conjunction with other nutrients like iron and nitrogen (23–26). Yet, it is difficult to extrapolate these observations, because the cobalt requirements of some key phytoplankton taxa have not been quantified.

Emerging prior to the oxygenation of the oceans, marine cyanobacteria in the Prochlorococcus and Synechococcus lineage differ from eukaryotic phytoplankton in that they possess an absolute requirement for cobalt, independent of other nutrients like Zn (7, 18, 27, 28). Despite their minimal genomes, which contain nearly half as many genes as other cyanobacterial species (e.g., Synechocystis), all sequenced Prochlorococcus genomes have retained pathways for the de novo biosynthesis of pseudocobalamin (psB₁₂), a structural analog of vitamin B₁₂ (29–31). The dominant phytoplankton group in the subtropical and tropical oceans, Prochlorococcus contributes a significant fraction of global primary production (32, 33), but the amount of cobalt needed to support this productivity is unknown. Here, we show that Prochlorococcus cells can grow with a very small number of cobalt atoms, enabling this organism to thrive in cobalt-poor environments like the equatorial Pacific Ocean without replacing cobalt-dependent metabolism.

Results and Discussion

Low Cobalt and Iron Requirements of Prochlorococcus MIT9215.

To quantify cellular cobalt requirements, we conducted growth experiments in chemically defined media with axenic cultures of Prochlorococcus strain MIT9215, a member of the abundant high-light II clade that was originally isolated from the equatorial Pacific Ocean (34). While scarce iron in this region limits the growth of Prochlorococcus and other phytoplankton (35, 36), the tropical Pacific Ocean also hosts low cobalt concentrations [<10 pM (3, 23, 37)].

The influence of nutrient limitation on fitness is a function of two distinct traits: 1) the ability of a cell to acquire nutrients from seawater and 2) the ability to apply acquired nutrients toward the biosynthesis of compounds needed for growth and cell division (38). In our experiments, division rates of Prochlorococcus MIT9215 decreased when inorganic cobalt concentrations (Co′; see Methods) fell below 3 pM, and no growth was observed at a Co′ concentration below 0.1 pM (Fig. 1A). However, thresholds of Co′ in these experiments were highly sensitive to the concentrations of Zn and Mn present in the growth medium, due to competitive inhibition at the uptake site, and are difficult to relate to typical oceanic concentrations directly (39). Regardless, we observed a robust relationship between intracellular cobalt contents (the cobalt quota, Q_Co) and cell growth rate (μ; Fig. 1B and SI Appendix, Fig. S1). Growth of Prochlorococcus MIT9215 was well described by the Droop Equation,

$$\mu ={\mu }_{max}\left(1-\frac{q_{min}}{Q_{Co}}\right),$$ [1]

which derives an absolute minimum quota, q_min, as growth rate (μ) approaches 0 (38). Least-squares fitting of q_min yielded extraordinarily low values: 14.8 ± 1.3 atoms per cell (R² = 0.96; equivalent to a cellular concentration of 220 nM). Implicit in this model is the expectation that increasing growth rates must be fueled by higher cobalt quotas, which is supported by our observations (Fig. 1B). From Eq. 1, a growth-limited range (μ < 95% μ_max; SI Appendix, Table S1) can be identified when the cellular cobalt quota is ∼300 atoms per cell or less, equal to a Co:Carbon ratio of 1.2 ± 0.1 × 10⁻⁷ (atom:atom). This value is similar to the Co:C ratio of a coastal Synechococcus under cobalt limitation [0.8 × 10⁻⁷ (18)], but an order of magnitude lower than the composite Co:C of eukaryotic phytoplankton [1.5 × 10⁻⁶ (40)], which is biochemically distinct from the cyanobacterial requirement and reflects variable substitution with Zn (18).

Fig. 1.

Growth experiments with Prochlorococcus MIT9215, originally isolated from the equatorial Pacific Ocean. (A) The exponential growth rate of Prochlorococcus MIT9215 as a function of iron (white) and cobalt (blue) concentrations (calculated as the sum of inorganic metal species, Fe′ and Co′, in equilibrium with ethylenediaminetetraacetic acid [EDTA]; see Methods). (B) Dependence of specific growth rate on acquired Fe and Co (i.e., the cell quota), measured by ICP-MS. Lines show best fit curves for Eq. 1 and derived values of the minimum cell quota, q_min (R² = 0.96 for both elements). (C) Cell quotas (±1 SD) of cobalt, the cobalamin-dependent ribonucleotide reductase (NrdJ, blue), and the cobalamin-dependent methionine synthase (MetH, pink). Proteins were quantified by selected reaction monitoring of four tryptic peptides (colored bars) in large volume cultures of cobalt-limited Prochlorococcus MIT9215 (μ = 0.20 d⁻¹; SI Appendix, Table S7). Gray lines show predicted levels of these enzymes needed to support maximum growth rates (0.6 d⁻¹) observed in these experiments via use efficiency calculations (SI Appendix, Table S4).

Cellular cobalt demands were compared to those of iron, the nutrient that limits Prochlorococcus growth in the equatorial Pacific Ocean (36). Growth limitation occurred when inorganic iron concentrations (Fe′) fell below 30 pM (Fig. 1A), similar to dissolved iron concentrations in the equatorial Pacific (41). To avoid iron limitation, Prochlorococcus MIT9215 required upward of 31,000 ± 3,400 Fe atoms per cell (μ < 95% μ_max, R² = 0.96; SI Appendix, Table S1), and estimated Fe:C ratios of iron-limited cells (1.6 ± 0.2 × 10⁻⁵) were similar to Fe:C ratios of bulk particulate digestions from the equatorial Pacific, the closest approximation of Prochlorococcus metal quotas in situ (42). Iron requirements in Prochlorococcus MIT9215 were much lower than those reported for Prochlorococcus strain MED4 (Fe:C > 4.1 × 10⁻⁵), which was isolated from the iron-rich Mediterranean Sea (43, 44), suggesting evolutionary pressure within this group to adapt to iron-poor regions of the oceans. The low Fe quotas of Prochlorococcus MIT9215 are still 100-fold larger than minimum cobalt requirements measured here (Fig. 1B and SI Appendix, Table S1).

Cobalt Use Efficiency in Prochlorococcus.

Raven (45) originally defined a “metal use efficiency” by scaling in vitro enzyme rates with protein:metal stoichiometries to derive maximum rates of biomass production per atom of catalytic metal (45, 46). For example, given light-saturated rates of O₂ evolution at photosystem II and a stoichiometry of 20 iron atoms per photosynthetic unit (SI Appendix, Table S3), a photoautotrophic cell could be expected to grow at a rate of 1 mol C s⁻¹, per mol Fe (46). At this iron use efficiency, a cell the size of Prochlorococcus growing at a rate of 0.4 d⁻¹ would need ∼8,200 iron atoms per cell, similar to measured iron quotas at this growth rate (5,100 ± 1,100 atoms per cell; SI Appendix, Table S2). This is consistent with photosynthetic and respiratory enzymes being the dominant use for acquired iron in Prochlorococcus.

We sought to extend this metal use efficiency framework to cobalt. All sequenced Prochlorococcus genomes encode two cobalamin-dependent enzymes: 1) the ribonucleotide triphosphate reductase NrdJ, which generates deoxyribonucleotide bases (dNTPs) for DNA replication from ribonucleotide precursors, and 2) the methionine synthase MetH, which methylates homocysteine to generate the amino acid methionine, needed for protein synthesis (30, 47). Unlike plants and many eukaryotic phytoplankton (15), Prochlorococcus lacks cobalamin-independent versions of ribonucleotide reductase and methionine synthase, indicating that there is no substitute for cobalt requirements associated with NrdJ and MetH. Other characterized cobalamin- or Co²⁺-dependent enzymes, such as methylmalonyl CoA mutase and nitrile hydratase, are not present in the genome of Prochlorococcus MIT9215 or other sequenced Prochlorococcus strains.

Prior to dividing, Prochlorococcus MIT9215 must replicate its genome of 1.7 million base pairs, requiring dNTP synthesis via NrdJ. To grow at a rate of 0.6 d⁻¹, the maximum observed in our experiments and similar to in situ growth rates in the equatorial Pacific (36), at least 2.04 million dNTP bases per cell must be synthesized daily. In vitro characterizations of NrdJ show a maximum turnover time of two cycles per second: unimpeded, a single enzyme can produce 170,000 dNTP bases per day (48, 49). Assuming that the Prochlorococcus NrdJ can achieve similar rates in vivo, cellular demand for dNTPs would require at least 12 functioning copies of this enzyme to support maximum growth rates (SI Appendix, Table S4).

A much faster enzyme, MetH, can operate at over 18 cycles per second (17, 50). A single MetH enzyme can produce over 1.5 million methionine molecules per day. While a complex methionine cycle makes it challenging to predict cellular MetH demands (51), the cellular sulfur quota of roughly 12 million atoms (composed of methionine, cysteine and homocysteine thiols, sulfates, and sulfones, etc.) can serve as an upper bound (SI Appendix, Table S4). By this calculation, five copies of MetH would fulfill even this overestimated demand.

Given a 1:1 protein:cobalt stoichiometry evident from crystal structures of NrdJ and MetH, this analysis can account for 17 Co atoms in cobalamin-dependent enzymes for Prochlorococcus growing maximally, equivalent to a metal use efficiency of 6.6 × 10⁷ mol C d⁻¹ per mol Co. This estimate is fourfold higher than an empirical Co use efficiency of 1.6 × 10⁷ calculated from growth experiments (SI Appendix, Tables S1 and S4), which may indicate the presence of additional Co associated with unknown enzymes, chaperones, and other transient reservoirs within the cell. To a first order, however, these calculations suggest that the quotas measured in Prochlorococcus cells are biochemically reasonable.

Mass Balance of Cobalt Metabolism.

To validate the above calculations, we constructed ¹⁵N-labeled peptide standards to quantify intracellular NrdJ and MetH levels by selected reaction monitoring mass spectrometry. From large volume cultures of cobalt-limited Prochlorococcus MIT9215 (μ = 0.2 d⁻¹), measurements of four tryptic peptides indicated that there were 15 ± 2 copies of NrdJ per cell (Fig. 1C and SI Appendix, Tables S5–S7). In the same sample, MetH was measured to an abundance of 7 ± 2 copies per cell. These measurements are remarkably consistent with predictions of 5 MetH and 12 NrdJ copies per cell based on the above use efficiency calculations at peak growth rates.

Still, the maximum amount of cobalt that can be attributed to use in NrdJ and MetH enzymes (22 ± 3 atoms per cell, assuming a 1:1 stoichiometry between Co and enzyme) is significantly less than the total cobalt quota in these cells, as measured by inductively coupled plasma mass spectrometry (ICP-MS; 62 ± 7 atoms per cell; Fig. 1C). Other reservoirs may account for more cobalt than NrdJ and MetH, even at these low intracellular concentrations. Before being used in NrdJ and MetH, acquired Co²⁺ ions must be transformed into pseudocobalamin (psB₁₂) cofactors, whose biosynthesis requires over a dozen enzymatic steps, beginning with the insertion of Co²⁺ into the corrinoid ring by the cobaltochelatase CbiK (14, 52). Given that many of the psB₁₂ biosynthetic intermediates would be present at a steady-state concentration of a few copies per cell or less, the random walks of single molecules from the active site of one enzyme in the biosynthetic sequence to the next may be the rate-limiting step to Prochlorococcus growth under cobalt limitation. The sheer number of enzymatic alterations required to produce pseudocobalamin [>15 cobalt-bearing intermediates (12, 46)] probably acts as a limit to growth efficiency at low cobalt.

Perhaps related to these physical limits, there appears to be a lack of regulatory control on Prochlorococcus MIT9215 cobalt metabolism, at least by inorganic Co species. The abundance of NrdJ and MetH did not increase significantly at higher Co′, despite a 14-fold increase in cellular Co (54 to 780 atoms per cell; SI Appendix, Fig. S2A). Cobalt use efficiency calculations imply that the measured cellular inventory of NrdJ and MetH in cobalt-limited cells is sufficient to sustain the flux of dNTP and methionine needed to support maximum growth rates. When cellular cobalt quotas are growth limiting, it is likely that these enzymes are unsaturated with respect to their pseudocobalamin cofactors, decreasing their metabolic production and delaying cell division.

Furthermore, comparisons between the cellular proteomes of replete and cobalt-limited Prochlorococcus MIT9215 did not show large differences in protein abundance, with only 12% of detected proteins significantly different (SI Appendix, Fig. S2B). Stress-related proteins—the photosynthetic proteins PsaD, PsbO, and some ribosomal proteins—were somewhat higher in cobalt-limited cells, while adenosine 5′-triphosphate synthase and sugar metabolism genes were less abundant (SI Appendix, Table S8). As these proteins are associated with basic metabolic functions, it is likely that changes in the cell division rate, not cobalt itself, drive differences in the proteome under cobalt-limiting conditions. This is consistent with the absence of many cyanobacterial metal-sensing and regulatory genes in Prochlorococcus genomes (39, 53).

Prochlorococcus Co Nutrition in the Equatorial Pacific Ocean.

All sequenced Prochlorococcus isolates harbor genes for psB₁₂ synthesis, as well as nrdJ and metH. It is expected that the cobalt requirements defined for the MIT9215 strain are broadly representative of all Prochlorococcus cells—at least as a lower limit—and can be used to investigate whether wild Prochlorococcus populations can meet their nutritional cobalt requirements.

To this end, we mapped the distribution of cobalt in the equatorial Pacific Ocean—the region where Prochlorococcus MIT9215 was originally isolated—from samples collected on the Metzyme expedition in October 2011 (Fig. 2A). Dissolved cobalt (dCo) was depleted in the surface ocean and accumulated in mesopelagic waters, consistent with the distribution of other phytoplankton nutrients like phosphate (Fig. 2 B–D). In the subsurface, cobalt concentrations greater than 100 pM were associated with very low oxygen waters (<50 μM; Fig. 2E and SI Appendix, Fig. S3A) that extend all of the way to the eastern margin (3). Sustained anoxia along the continental shelves of Peru and Mexico promotes the dissolution of cobalt bound in sedimentary Mn oxides (54, 55), and, at the basin scale, cobalt input from these reducing sediments is balanced by its removal onto Mn oxides formed in oxic waters (56). Inhibition of Mn and Co oxidation at low O₂, combined with significant regeneration of phytoplankton-derived cobalt, leads to plumes of high cobalt within low-oxygen waters in both the North and South Pacific (Fig. 2C).

Fig. 2.

The distribution of dissolved cobalt, dCo, in the equatorial Pacific Ocean. (A) Station locations (pink circles) from the Metzyme expedition (KM1128) relative to low-oxygen water masses originating on the Peruvian and Mexican margins. Contours mark 25-μM intervals in dissolved oxygen (O₂) between 25 μM and 100 μM at 200-m depth from the World Ocean Circulation Experiment (WOCE) gridded atlas (82). Blue shading highlights low dissolved oxygen. (B) A latitudinal section of dissolved oxygen concentrations from the Metzyme expedition. (C) dCo (operationally defined as the concentration passing through a 0.2-μm filter) and (D) dissolved phosphate (PO₄) along the same section. (E) The full depth distribution of dCo overlain with dissolved oxygen contours at 25-μM intervals between 0 and 100 μM O₂, which highlight their correlation throughout the water column (SI Appendix, Fig. S3).

At the equator, upwelling of subsurface waters from the Equatorial Undercurrent (EUC) leads to high nutrient concentrations and productivity. Originating in the oxygenated waters of the western Pacific, the EUC transports a large iron source from Papua New Guinea into the equatorial upwelling region (41, 57). As it travels eastward, the EUC splits low-oxygen water masses, leading to a downward deflection of O₂ contours between 2°N and 2°S (Fig. 2B) (58). Despite its importance as an iron source, cobalt concentrations within the EUC were not elevated (30 ± 2 pM between 100 and 250 m), especially compared to the surrounding low-oxygen waters (Fig. 2C). Low cobalt in the EUC probably reflects the high O₂ concentrations near Papua New Guinea and the predominance of oxic (rather than reducing) sedimentary sources (57, 59). Instead, entrainment of low-oxygen, high-cobalt waters into the EUC leads to elevated cobalt concentrations in the euphotic zone at the equator (23 ± 6 pM between 0 and 100 m; Fig. 2C).

As upwelled equatorial waters are advected poleward, nutrient concentrations throughout the euphotic zone decrease due to uptake by phytoplankton. Interestingly, cobalt is more strongly depleted in the South Pacific Gyre than in the North Pacific, opposite to the trend in phosphate (compare, for instance, the 5-pM contour for cobalt in Fig. 2B and the 0.25-μM contour for PO₄ in Fig. 2C). Greater contact with coastal sources from East Asia, a larger oxygen minimum zone, and orders of magnitude greater deposition of desert dust in the Northern Hemisphere probably contribute to higher cobalt inventories in the North Pacific Subtropical Gyre (60–62). These sources are weaker in the South Pacific, allowing surface concentrations to fall below the 2-pM detection limit of our electrochemical method at both 9.2°S and 12°S (Fig. 3A). These stations also contained the largest populations of Prochlorococcus observed on the transect, evident in divinyl chlorophyll a concentrations [which are specific to Prochlorococcus (63)] that exceeded chlorophyll a produced by other phytoplankton (SI Appendix, Fig. S4). At a threshold growth rate of 95% μ_max, cobalt requirements of Prochlorococcus MIT9215 are estimated to be 300 atoms per cell, equivalent to a biomass Co:P ratio of 26 × 10⁻⁶ (SI Appendix, Table S1). Thus, in order to take up the 300 nM surface PO₄ found at 9.2°S and 12°S, Prochlorococcus must also acquire at least 7.8 pM Co in order to avoid cobalt-limited growth rates: more than triple the dCo concentration found at these stations. Indeed, dissolved Co:PO₄ ratios in the euphotic zone (0 to 100 m) fall below this threshold for all stations south of the equator (Fig. 3E), suggesting that the PO₄ reservoir here cannot be depleted without forcing cobalt-limited growth rates.

Fig. 3.

Cobalt and phosphate profiles in the South Pacific Ocean from the Metzyme expedition. Removal of (A) dCo and (B) dissolved phosphate in the surface ocean is matched by surface maxima of (C) particulate cobalt and (D) particulate phosphorus. For A–D, colors and shapes indicate station locations. (E) The ratio of cobalt to phosphorus in the euphotic zone (0 to 100 m) along the Metzyme transect in both dissolved (pink) and particulate (blue) phases. Dotted black lines indicate the Co:P composition of cobalt-limited Prochlorococcus MIT9215 biomass at threshold growth rates of 0% and 95% μ_max via Eq. 1 (SI Appendix, Table S1).

However, elemental standing stocks do not always reflect their availability to organisms, many of which are adept at recycling trace nutrients. In the South Pacific, depletion of cobalt from the dissolved phase of seawater observed on the Metzyme expedition is mirrored by a significant accumulation of cobalt in particulate matter, often exceeding 1 pM (Fig. 3C). Correlation between pCo and particulate phosphorus indicates that pCo is primarily associated with biomass (Fig. 3D and SI Appendix, Fig. S3B). In both the North and South Pacific, the Co:P composition of particulate matter (110 ± 30 × 10⁻⁶) exceeds both dissolved Co:PO₄ ratios and the cobalt limitation threshold of Prochlorococcus MIT9215 identified in culture (26 × 10⁻⁶; Fig. 3E). To the extent that the Co:P composition of the total particulate pool reflects the composition of Prochlorococcus cells, it would appear that wild Prochlorococcus populations in the South Pacific have accumulated enough cobalt to satisfy their cobalamin metabolism and avoid cobalt limitation.

To understand how phytoplankton are able to accumulate Co in this region, we analyzed a global model of the marine cobalt cycle (37), which reproduces the very low dCo concentrations observed in the South Pacific but does not contain explicit parameterizations for Co limitation of phytoplankton growth. In the model, the South Pacific Gyre is the only region where dissolved Co:PO₄ ratios fall below the 95% μ_max threshold for Prochlorococcus MIT9215 [excluding the Southern Ocean where waters are too cold to support Prochlorococcus growth (64)]. Similar to our observations, the model also simulates elevated phytoplankton Co:P ratios in the South Pacific Gyre (>100 × 10⁻⁶), along with low dissolved Fe concentrations and Fe limitation of phytoplankton growth (SI Appendix, Figs. S5 and S6). Conceptually, iron-limited growth would allow the same number of cobalt transporters to complete more transport cycles prior to cell division, leading to a buildup of cellular Co. This effect is responsible for elevated phytoplankton Co:P ratios in the model, and was also observed during our iron limitation experiments with Prochlorococcus MIT9215 in culture (SI Appendix, Fig. S1 and Table S2). High levels of iron transporters and other protein biomarkers observed during the Metzyme expedition support model-based conclusions that iron limits Prochlorococcus growth in the South Pacific Gyre (SI Appendix, Fig. S7 and 35). Thus, cobalt accumulation afforded by iron-limited growth may be responsible for the high particulate Co:P ratios also observed on this transect.

Analysis of the measurements within the GEOTRACES intermediate data product (65) yields a similar conclusion: all waters with low Co:PO₄ ratios also contain comparatively lower Fe:PO₄ ratios (with respect to q_min values of Fe and Co determined from culture experiments; Fig. 4). If Prochlorococcus growth thresholds for cobalt and iron are assumed to be independent of one another (as in Liebig’s law of the minimum), then cobalt limitation will be blocked by iron limitation (Fig. 4). Still, this scenario suggests that iron addition on short timescales (e.g., dust deposition) may drive Prochlorococcus and other phytoplankton toward cobalt limitation, consistent with observations of iron and cobalt colimitation in bottle incubations (23, 25).

Fig. 4.

Global analysis of cobalt depletion in the marine habitat of Prochlorococcus, defined as waters with temperature >12 °C. Colors represent the relative pixel density of modeled stoichiometry of total Co:PO₄ and Fe:PO₄ (the sum of dissolved and particulate phases) in the surface ocean, with pink colors indicating a large number of pixels (i.e., model grid cells). Crosses indicate measurements of dissolved Co, Fe, and PO₄ from the GEOTRACES Intermediate Data Product 2017 (65) from the surface mixed layer (0 to 30 m). Minimum Co and Fe requirements of Prochlorococcus MIT9215 are used to define growth limitation domains for Co and Fe (solid black lines). Dashed lines mark the shift in nutrient limitation domains if cobalt requirements increase by 10-fold (a Co:P ratio of 260 × 10⁻⁶ at 95% μ_max), which corresponds to Co:P ratios of E. huxleyi cultures grown in low-Zn media (67).

There are two factors that complicate this interpretation. First, some Prochlorococcus lineages may retain a higher Co requirement than Prochlorococcus MIT9215, due to a reliance on other putative Co-dependent metalloenzymes, such as the PhoA alkaline phosphatase (30, 66). Second, in addition to Prochlorococcus, the pool of particulate matter in the surface ocean is also composed of small eukaryotic phytoplankton, mixotrophs, and grazers, as well as heterotrophic bacteria and detritus. Differences in cobalt (or P) accumulation between these reservoirs will distort comparisons with axenic Prochlorococcus cultures. Indeed, the cobalt requirements of eukaryotic phytoplankton like the diatom Thalassiosira oceanica and the haptophyte Emiliana huxleyi can be much greater than Prochlorococcus MIT9215 (Co:C > 10⁻⁶ for both species) because these organisms substitute cobalt into their Zn-dependent enzymes (cambialism) when grown under low zinc concentrations found in open ocean environments (18, 19). Under conditions of Zn scarcity, the Co requirements of E. huxleyi can be more than 10-fold higher than Prochlorococcus MIT9215: equivalent to a biomass Co:P ratio greater than 300 × 10⁻⁶ (67). While the degree of Co−Zn interreplacement in the open ocean is currently unknown, dissolved Zn in the South Pacific Gyre can fall below 0.1 nM (65, 68), and free Zn′ concentrations are likely orders of magnitude lower due to complexation by organic ligands (69). When analyzing the Co:P ratio of a mixed assemblage, the potentially high Co:P ratios associated with cambialistic requirements may mask lower Prochlorococcus Co:P. Such high Co requirements must also make eukaryotic phytoplankton more susceptible to Co−Zn limitation. At the Co:P stoichiometry observed in E. huxleyi, the severity of Co depletion would exceed Fe throughout large areas of the North and South Pacific Oceans (Fig. 4 and SI Appendix, Fig. S8), possibly leading to Co−Zn limitation of growth. The simultaneous depletion of iron, nitrogen, cobalt, zinc, and other nutrients in the South Pacific and other subtropical gyres may prevent large-scale increases in ecosystem biomass, even if sources of the “limiting” nutrient are increased.

Evolutionary Drivers of Minimized Cobalt Metabolism.

High cobalt concentrations in the equatorial Pacific are associated with oxygen-poor water masses, matching observations elsewhere in the Pacific and Atlantic (3, 6, 70). The worldwide association between Co and O₂ supports thermodynamic predictions that the anoxic ocean found during the Proterozoic and Archaean also contained much higher cobalt (7). As cyanobacteria evolved during this time, high cobalt availability probably facilitated the incorporation of (pseudo)cobalamin-dependent enzymes like NrdJ and MetH instead of other, cobalamin-independent versions of these proteins. However, genome-based timelines suggest that the Prochlorococcus lineage did not emerge until the late Paleoproterozoic or early Phanerozoic (28, 71). Although marine oxygen concentrations during this period are uncertain, the subsequent diversification of Prochlorococcus ecotypes throughout the Phanerozoic probably occurred in an oxygenated ocean (28, 72), with cobalt concentrations that were probably lower than when the cyanobacteria first evolved. As the evolution of Prochlorococcus ecotypes is characterized by metabolic streamlining and genome minimization (73, 74), it is interesting to note that, because all DNA bases in Prochlorococcus are synthesized by the NrdJ enzyme, decreasing genome size also helps to decrease cobalt requirements.

Our analysis suggests that Prochlorococcus cells can continue to function with a very small number of cobalt atoms, as low as 17 atoms per cell. While some lineages of Prochlorococcus may maintain a broader suite of cobalt-dependent or cambialistic enzymes (and therefore a higher cobalt requirement), the very low cobalt requirement of Prochlorococcus MIT9215 minimizes its exposure to cobalt-limiting conditions (Fig. 4), likely explaining the retention of psB₁₂ biosynthesis in Prochlorococcus genomes. Similarity in proteomes between cobalt-replete and cobalt-limited growth suggests that Prochlorococcus MIT9215 does not actively monitor its cobalt inventory (SI Appendix, Fig. S2), nor does it make significant efforts to increase Co uptake by up-regulating Co′ transport systems. Indeed, Co′ uptake into Prochlorococcus MIT9215 cells appears to be relegated to the periplasmic Mn transport system MntABC, rather than a transporter specific to cobalt (39). These adaptations may reflect efforts to decrease the number of transporters and regulatory systems needed within the cell, a resource and energy savings that may offset the resource/energy expenditure associated with the production of psB₁₂ biosynthetic enzymes.

While all sequenced Prochlorococcus genomes have retained the genes for psB₁₂ biosynthesis, there are other cyanobacteria that have lost this functionality. For example, the benthic cyanobacterium Synechococcus PCC7002 has lost its cobalamin biosynthetic pathway and replaced its ancestral nrdJ with horizontally acquired nrdAB genes, which encode a B₁₂-independent ribonucleotide reductase (75, 76). As a result, Synechococcus PCC7002 depends on exogenous B₁₂ to power the MetH methionine synthase. Genetic replacement of the metH gene with B₁₂-independent methionine synthase, metE, is sufficient to relieve Synechococcus PCC7002 of its B₁₂ dependence (75). Several other cyanobacteria (e.g., Synechococcus PCC73109, Crocosphaera watsonii) have also acquired B₁₂-independent versions of both methionine synthase and ribonucleotide reductase. Given pervasive signals of horizontal gene acquisition in Prochlorococcus genomes (74), it is surprising that metE and nrdAB genes are absent from sequenced Prochlorococcus isolates.

Prochlorococcus appears to retain its psB₁₂-dependent enzymes out of physiological benefit rather than an insufficient access to B₁₂-independent alternatives. The cobalamin-independent ribonucleotide reductase NrdAB is an iron-dependent enzyme (77), whose incorporation would require an increase in Prochlorococcus’s iron requirement, making it less competitive in iron-limited waters. Similarly, B₁₂-independent MetE can only operate at ∼1% of the rate of MetH, and would therefore be needed at much higher quantity (15, 17), increasing demand for photosynthate (and therefore for Fe). Eukaryotic phytoplankton with both metE and metH genes clearly prefer to use MetH when B₁₂ is present (15). The sheer abundance of Prochlorococcus throughout the subtropics and tropics implies that it is the major synthesizer of cobalamins throughout the sunlit ocean (29, 78, 79), and therefore supports the B₁₂ dependencies of other phytoplankton [the pseudocobalamin produced by Prochlorococcus being convertible to conventional B₁₂ by many species (29, 80)]. The persistence of psB₁₂ biosynthesis in Prochlorococcus makes not only its own metabolism more efficient, but potentially the entire marine ecosystem.

Partially degraded B₁₂ and psB₁₂ molecules are inferred to be the major source of dCo ligands found in seawater, whose stability over hundreds of years protects cobalt in the ocean interior from scavenging by Mn-oxidizing bacteria and burial in marine sediments (23, 37, 56). Therefore, minimization—rather than replacement—of Prochlorococcus cobalamin-dependent metabolism may have a global effect on the marine cobalt cycle.

Methods

Axenic cultures of Prochlorococcus MIT9215 were grown, harvested, and analyzed as described previously (39). Growth rates were determined in multiple 30-mL polycarbonate tubes and pooled to measure cellular metal quotas and protein content. Large (8 L) nonaxenic cultures were used to grow biomass for quantitative protein measurements.

Global and targeted protein measurements were made following previously described protocols (35, 81). Significantly different protein abundance was determined by the Fisher exact test using a cutoff of P < 0.01. Concentrations of Prochlorococcus MIT9215 ribonucleotide reductase (NrdJ) and methionine synthase (MetH) were measured in soluble protein extracts using ¹⁵N-labeled peptide standards generated from Prochlorococcus NrdJ and MetH peptides in an Escherichia coli overexpression system. Labeled peptides were coexpressed with peptide sequences for horse myoglobin which were calibrated relative to a commercially available horse myoglobin protein (Fisher). Presence/absence of nrdA, nrdJ, metE, and metH was determined by Protein-Protein Basic Local Alignment Search Tool (BLASTp) searches with E-value cutoffs of 10⁻²⁰.

Concentrations of dCo (operationally defined as passing through a 0.2-μm filter) from the Metzyme expedition (KM1128) were measured by cathodic stripping voltammetry (3, 35). Sampling procedures and analysis of particulate metals is described by Saito et al. (35). Equations governing the Co biogeochemical model can be found elsewhere (37). Expanded methodological details can be found in SI Appendix.

Data Availability.

Oceanographic data from the Metzyme expedition are accessible at the Biological and Chemical Oceanography Data Management Office repository (https://www.bco-dmo.org/project/2236). Data tables for culture experiments are provided in SI Appendix. Model output can be accessed at Zenodo, https://doi.org/10.5281/zenodo.3880520.