Nature Biotechnology
(2026) Cite this article
Silicate mineral weathering (dissolution) is a scalable strategy for capture and storage of CO2 but is too slow for industrial deployment. Bacteria can accelerate mineral dissolution by secreting siderophores, molecules that solubilize iron released from the mineral. Here we investigate how to deploy siderophore-producing bacteria at scale to continuously enhance dissolution of the mineral olivine. We demonstrate that natural genetic regulation precludes continuous siderophore production in mineral bioreactors. To overcome this limitation, we engineer the marine bacterium Alteromonas macleodii for enhanced siderophore production, conferring a 2.6-fold increase in the rate of olivine dissolution. Life-cycle analysis indicated that renewable feedstocks and minimal replenishment of modified cells are critical to achieve net CO2 removal at scale. With these guidelines, we constructed pilot-scale continuous mineral bioreactors that use unprocessed seawater and a renewable acetate feedstock to weather 4 kg of olivine. In reactors with engineered cells, we directly measured removal of 0.50 g CO2 per day from the air through alkalinity generation.
Enhanced weathering of silicate rocks could enable removal and long-term storage of atmospheric CO2 at large scales1,2,3. Mafic and ultramafic minerals such as olivine generate alkalinity upon dissolution. In marine and terrestrial mediums, this alkalinity is primarily comprised of stable bicarbonate ions HCO3−, derived from atmospheric CO2 (Fig. 1). In ongoing field trials in which mafic rocks were distributed into soils or coastal environments, the extent of CO2 removal has been challenging to measure because of slow rock weathering rates4,5. Tank-based mineral processing would allow control over hydrology and mineral surface area, simplifying measurement of mineral dissolution and CO2 capture and storage. Existing unit operations to dissolve minerals leverage chemical or biological acids, however, precluding generation of alkalinity6. A tank-based unit operation to accelerate the dissolution of silicate minerals in a low-cost environmental medium such as seawater without reliance on acids could enable large-scale, measurable CO2 storage on the decadal timescale.
Siderophores solubilize ferric iron, preventing mineral passivation by iron oxides. After dissolution, silicate ions act as proton acceptors to stabilize carbonic acid in solution.
Bacterial siderophores, secreted secondary metabolites that chelate and solubilize ferric iron, can accelerate the dissolution of silicate minerals at neutral pH7. Iron is abundant in mafic and ultramafic rocks and forms iron oxides in aqueous, oxidative conditions. During natural weathering, iron oxides precipitate back onto the mineral surface and prevent further mineral dissolution7. Siderophores may prevent this passivation by solubilizing ferric iron. They have also been shown to directly extract iron from iron oxide minerals, a process that likely occurs in soils8,9. In batch reactions, the siderophore desferrioxamine accelerated the dissolution of olivine by solubilization of passivating iron oxides10. It was suggested, however, that siderophore production may be limited by the presence of olivine, preventing biologically accelerated mineral dissolution at large scales.
In this study, we developed continuous mineral bioreactors and experimentally confirmed that siderophore production is limited at scale. Then, we engineered a strain of the marine bacterium Alteromonas macleodii with enhanced siderophore production that can continuously accelerate olivine dissolution. We performed a life-cycle analysis (LCA) to assess the potential of mineral bioreactors to capture and store CO2 at industrial scale. Informed by our LCA, we constructed pilot-scale mineral bioreactors and demonstrated sustained acceleration of olivine dissolution by engineered A. macleodii in unprocessed seawater.
We sought to develop a scalable, tank-based unit operation for biologically accelerated olivine dissolution (Fig. 1). We reasoned that seawater was the optimal process medium; seawater is abundant, inexpensive and can be returned to the environment carrying alkaline ions for long-term CO2 storage. Siderophore-producing bacteria are also ubiquitous in the ocean11. Here, we used the marine bacterium A. macleodii because it naturally produces the siderophore petrobactin12. The genes for petrobactin synthesis were previously shown to be expressed in response to iron limitation13, as is typical for most siderophore-producing bacteria14,15 (Supplementary Fig. 1). In batch cultures with synthetic seawater medium, A. macleodii acquired iron from olivine for growth (Supplementary Fig. 2). We could only achieve iron-limited growth, however, with 0.4 g l−1 olivine. Ions released from this small quantity of olivine were not detectable in the seawater medium. To overcome this measurement challenge, we developed a platform for continuous cultures.
We developed parallelized bioreactors that enable continuous cell growth on a mineral substrate. In these modified eVOLVER chemostats16, A. macleodii culture is continuously diluted while a fixed quantity of olivine sand is retained in the vessel by settling (Fig. 2a). After 1–2 days of continuous dilution, reactors with olivine sand reached steady-state growth, while reactors without olivine exhibited washout (drop to near-zero cell density), supporting the assumption that cells were dependent on iron acquired from the olivine (Supplementary Fig. 3).
a, Continuous mineral bioreactors. Seawater medium was added and removed at a constant rate while olivine sand remained in the vial by sinking. Stirring ensured well-mixed culture fluid. b, Predictions for mineral reactors at steady state according to the theoretical chemostat model. Each point in this phase diagram represents a hypothetical reactor at steady state with a given dilution rate and mineral mass. Color spectrum represents the ratio of glucose to iron limitation according to the model. The purple region indicates process conditions that are predicted to yield a high steady-state concentration of siderophores. In c–f, real reactors were operated with different dilution rates and mineral masses. The same set of 16 reactors were operated in replicate several weeks apart. Each triangle represents one chemostat at steady state. c, OD600. d, Expression of siderophore synthesis genes asbA and asbC, measured by qPCR and normalized to a housekeeping gene rpoB. e, Expression of iron storage genes ftnA1 and ftnA2. f, Concentration of the siderophore petrobactin measured by LC–MS and normalized within each experiment. g, Iron released by olivine inhibits siderophore production.
Continuous mineral bioreactors also enable measurement of mineral dissolution at steady state. We developed a method to measure the total quantity of metal ions released from the olivine substrate by complete digestion of culture fluid in strong acid (Supplementary Fig. 4a). Testing the method with synthetic metal mixtures revealed that Ni and Fe were good markers of olivine dissolution at small scale (Supplementary Fig. 4b). Mg is a good marker of olivine dissolution at higher concentrations but, at low concentrations, does not overcome the background signal of Mg in the seawater medium. In our pilot experiment, the rate of release of multiple metals from olivine did not change between two dilution rates or in the presence or absence of A. macleodii cells (Supplementary Fig. 3). We hypothesized that cells were not iron limited in these initial conditions and, therefore, did not produce siderophores.
We then sought to identify process conditions in which cells would experience continuous iron limitation and continuously produce siderophores. To guide our experiments, we modeled our continuous mineral bioreactors as chemostats (Appendix 1). In classical chemostats, cell growth is limited by nutrient influx (glucose in our system). Torres et al. previously modeled mineral chemostats, in which cell growth is limited instead by the release of iron from a solid mineral substrate10. Our model builds upon this previous work by allowing cells to be limited by either glucose or iron and by allowing control over both the mass of olivine in the vessel and the dilution rate of the medium. After solving our model, we predicted that cell growth at steady state would be primarily limited by iron only with low olivine mass and a high enough dilution rate (Fig. 2b, red). More specifically, we predicted that continuous siderophore production would only occur in a narrow range of process conditions in which cells experience iron limitation but not so much that the culture washes out (Fig. 2b, purple).
We confirmed in real bioreactors that continuous iron limitation was limited to a narrow range of process conditions. We operated 16 parallel reactors with varied olivine masses and dilution rates in duplicate (Supplementary Fig. 5). As predicted, cultures washed out above a certain dilution rate (Fig. 2c). In reactors with less olivine, washout occurred at lower dilution rates than in chemostats with more olivine, likely because less iron is available. We also measured the expression of genes that are markers for iron limitation13. Siderophore synthesis genes were expressed at a high level with low olivine mass and high dilution rates, suggesting that growth was limited by iron in this regime (Fig. 2d). Conversely, iron storage genes were expressed at a high level with high olivine mass and low dilution rates, suggesting that growth was not limited by iron in this regime (Fig. 2e). We measured the relative quantity of the siderophore petrobactin and observed high concentrations only in conditions with iron limitation and high cell density (Fig. 2f). This observation corroborates our and others’ prediction that olivine could inhibit siderophore production at large scales10 (Fig. 2g). Indeed, we only observed increased release of Ni, Fe and Si from olivine in chemostats with low olivine mass (Supplementary Fig. 3). Together, these observations demonstrate that endogenous regulation in A. macleodii precludes siderophore production in bioreactors under reasonable operating parameters at industrial scale.
To enable siderophore production in the presence of high quantities of olivine, we genetically engineered A. macleodii. To enable this engineering, we characterized a set of synthetic, constitutive promoters (https://parts.igem.org/Promoters/Catalog/Anderson) and observed a >30-fold range of GFP fluorescence (Supplementary Fig. 6). We then cloned the entire petrobactin synthesis operon asb into an episomal plasmid under control of each synthetic promoter and evaluated each transformed strain for petrobactin secretion by chrome azurol S (CAS) agar assay17 (Fig. 3a). While most plasmids conferred little to no increase in CAS assay signal, the strain carrying a plasmid with the asb operon under control of the second strongest promoter (J23100) exhibited a dramatic increase in CAS assay signal, suggesting increased siderophore secretion and cellular iron uptake (Fig. 3b and Supplementary Fig. 7). We denoted this strain ‘asb+P’. Interestingly, we were unable to successfully transform the plasmid with the asb operon under control of the strongest promoter (J23101). We hypothesize, therefore, that the asb+P strain is near the maximal expression level of the asb operon that would confer toxicity.
a, Construction of the asb+P strain. The asb operon was cloned onto an episomal plasmid and conjugated back into A. macleodii under control of a synthetic constitutive promoter. b, Photo of cells growing on CAS agar plates. Yellow signal indicates sequestration of iron from the blue dye by secreted siderophores. Scale bars, 1 cm. c, OD of continuous mineral bioreactors at steady state with different dilution rates and mineral masses. d,e, Expression of the iron storage genes ftnA1 and ftnA2 (d) and the siderophore synthesis genes asbA and asbC (e). Gene expression was measured by qPCR and normalized to a housekeeping gene rpoB. f, Steady-state concentration of the siderophore petrobactin, measured by LC–MS and normalized within the experiment.
Engineered A. macleodii produced siderophores in iron-replete conditions. We continuously cultured the asb+P strain across the same range of reactor conditions we used for the wild-type (WT) strain (Supplementary Fig. 8). The asb+P strain performed like the WT strain; cultures again washed out at high dilution rates (Fig. 3c), the washout rate was lower with lower olivine mass (Fig. 3c) and expression of the ferritin genes ftnA1/2 was lower with low olivine mass and high dilution rates (Fig. 3d). These all suggest that, like for the WT strain, growth was limited by iron in reactors with low olivine mass and high dilution rates. Unlike the WT strain, however, expression of siderophore synthesis genes asbA and asbC was high across all conditions (Fig. 3e). Critically, we observed high concentrations of petrobactin in the reactors with the highest iron availability (Fig. 3f). Petrobactin production appeared to scale with the cell density, thus indicating that the asb+P strain can produce petrobactin even when iron does not limit growth.
