The injection of CO₂ gas into deep subsurface environments has emerged as one possible process to reduce anthropogenic greenhouse gas (GHG) emissions to the atmosphere (Benson and Cole, 2008). This technology involves the pumping of CO₂ into suitable geologic environments such as depleted gas and oil fields, saline aquifers, and basalt formations (Holloway, 2005; McGrail et al., 2006). In many of these environments, elevated pressures and temperatures result in a CO₂ phase change, from a gaseous to a supercritical state. Within deep subsurface environments, supercritical CO₂ (scCO₂) can undergo dissolution and disassociation reactions in formation fluids, resulting in pH decreases between 1.5 and 4 units (Mitchell et al., 2010). This pH change can subsequently cause mineral dissolution (and pH buffering) and re-precipitation reactions; the formation of new carbonate minerals may sequester the injected carbon over extended time periods (Krevor and Lackner, 2011; Qafoku et al., 2014).

A significant fraction of microbial biomass on Earth is present in the deep subsurface (Whitman et al., 1998; Kallmeyer et al., 2012; McMahon and Parnell, 2014), in many terrestrial and oceanic locations that could potentially be used for CO₂ sequestration (Stevens and McKinley, 1995; Krumholz, 2000; Basso et al., 2009; Edwards et al., 2012). scCO₂ has been shown to significantly reduce bacterial membrane potential and membrane integrity (Tamburini, 2013), and is commercially used as a microbial sterilizing agent in the food industry (Amanatidou et al., 1999). The injection of scCO₂ into the subsurface could therefore potentially kill a significant fraction of biomass in the storage formation (White et al., 2006). However, the heterogeneity of the subsurface may provide microenvironments where less harsh conditions can be found; indeed, a field test site in Ketzin, Germany reported active populations of sulfate-reducing bacteria (SRB) and methanogenic Archaea following a CO₂ injection (Morozova et al., 2010). Similarly, studies in an Australian sandstone aquifer revealed increases in Comamonadaceae and Sphingomonadaceae following CO₂ injection (Mu et al., 2014). The activity of such microbial populations in subsurface regions receiving injected CO₂ has both beneficial and deleterious implications for the efficiency of this process. Biofilm formation might act to block small fractures and pores, preventing upward migration of CO₂ into overlying aquifers (Mitchell et al., 2009), while microbially-enhanced mineral precipitation could accelerate rates of CO₂ incorporation into insoluble phases (Mitchell et al., 2010). Conversely, microbial growth around injection wells can cause localized clogging (Morozova et al., 2010), corrosion (Pitonzo et al., 2004), and reductions in injection rates, while the potential for certain microbial strains to metabolize CO₂ and H₂ to methane, a more potent GHG than CO₂, remains poorly understood under relevant conditions.

Although in situ microbial activity has been reported following CO₂ injection, little is known about how CO₂ injection and possible upward migration of CO₂ into overlying formations can affect microbial physiology, metabolism, and stress responses. Studies on a range of different microorganisms have reported reductions in growth rates and cell yields associated with increasing CO₂ exposures, while a combination of increased CO₂ at high pressures caused large reductions in viable cells (Coyne, 1933; Eklund, 1984; Wu et al., 2010; Schulz et al., 2012). Mitchell et al. (2008) reported that biomass present in a biofilm were more resistant to scCO₂ than similar cells in a planktonic state, and hypothesized that differences in cell viability might be due to protective effects of extracellular polymeric substances (EPS) that accumulate within a biofilm. However, despite these studies there is currently little understanding of the stress response mechanisms that may be utilized by microbial populations that are suddenly exposed to CO₂ at elevated pressures. Sulfate reducing bacteria (SRB) inhabit a wide range of diverse ecosystems, under conditions that are frequently considered to be “extreme” for microbial growth, including the deep subsurface in terrestrial and ocean environments (Bale et al., 1997; Sass and Cypionka, 2004; Basso et al., 2005; Klouche et al., 2009). Here we exposed a model subsurface SRB species, Desulfovibrio vulgaris, to CO₂ under a range of pressures, using a novel combination of high-pressure nuclear magnetic resonance spectroscopy (HP-NMR) and RNA-Seq to track both physiological and transcriptional responses to CO₂ stresses that may occur in deep subsurface injection sites.

In this study, metabolic and physiological indicators were monitored during exposure of an actively growing SRB model organism, D. vulgaris, to CO₂ over a range of environmentally relevant pressures. Using a high-pressure batch incubation reactor, biomass was incubated under CO₂ at atmospheric, 10 bar, 25 bar, 50 bar, and 80 bar pressures at 37°C, with samples recovered for kinetic, microscopic, and transcriptomic analyses. High-pressure NMR techniques were applied concurrently to measure microbial oxidation of lactate under similar pressure and temperature conditions.

Despite the use of HEPES-buffered media, incubation under CO₂ (≥5 bar) resulted in pH decreases over the 4 h incubation period from 7.2 to values ranging from 5.5 to 6.5. Atmospheric pressure incubations at pH 6 (N₂ headspace) indicated that the decrease in media pH had little effect on cell growth or sulfide production (data not shown). CO₂ exposure had a dramatic effect on cell respiration however, as inferred from sulfide measurements, cell density measurements (Figures 1A,B), and time-course NMR data collection (Figure 2).

Sulfide generation was greatly reduced at CO₂ pressures above atmosphere (1 bar), with any S²⁻ increases only observed over the first hour of exposure to 10, 25, 50, or 80 bar CO₂. By contrast, sulfide was consistently generated over the 4-h measurement period under CO₂ at atmospheric pressure (Figure 1A). These trends were mirrored when measuring cell biomass increases via OD₆₀₀ readings (Figure 1B), with consistent increases in cell density observed at atmospheric pressure under a CO₂ headspace. At elevated CO₂ pressures however (≥10 bar), small increases in cell density were generally only observed during the first hour (Figure 1B), with cessation of growth at later time periods. No additional cell growth was observed in these experiments over longer time periods (~48 h) (data not shown). Biomass samples were also recovered at each time point and stained for cell-viability measurements using live-dead cell staining techniques. These data indicated that little cell death occurred at 10 bar and 25 bar CO₂ pressures during the 4 h incubation period (Figure 1C). Similar trends were initially observed at 50 bar and 80 bar CO₂. Cell viability began to decrease in these samples at later time points (>3 h), with the greatest proportion of cells with compromised membranes measured after 4 h incubation at 80 bar CO₂ (Figure 1C). To decouple the effects of elevated pressure and CO₂ stress, parallel incubations were carried out using N₂ gas as the pressurizing phase rather than CO₂. These experiments were carried out at pH 6 to mirror the pH change associated with CO₂ exposure. Results from these incubations demonstrated that pressure alone did not impact D. vulgaris activity. The growth rate in cultures exposed to 80 bar N₂ (μ = 0.19 h⁻¹) was even greater than the growth rate measured in cultures exposed to 1 bar CO₂ (μ = 0.13 h⁻¹) during the 4-h incubation period (Figures 1A,D).

13-Carbon-NMR was used to track the oxidation of ¹³C-3-lactate, and formation of ¹³C-2-acetate by D. vulgaris across a range of CO₂ pressures at 37°C. At 10 bar CO₂, incomplete oxidation of lactate was observed, indicating inhibition of microbial respiration and supporting observations made in batch reactions (Figure 2). Similar observations were made via NMR at pressures of 25, 50, and 80 bar CO₂. However, the complete conversion of lactate to acetate was observed at both 1 and 5 bar CO₂ over similar reaction times, suggesting that some exposure to increased CO₂ pressures could be tolerated by D. vulgaris (Figure 2). In contrast, 1-H proton NMR measurements revealed complete lactate oxidation to acetate by D. vulgaris under 25 bar N₂. These data support batch incubation results that indicated no negative effects of pressure alone on growth rate (Figure 2).

Biomass samples taken at 0 h and 4 h time points during 50 bar CO₂ exposure were imaged using TEM. Initial TEM images showed healthy D. vulgaris cells at the beginning of the experiment (Figures 3A–C). After 4 h incubation under 50 bar CO₂, there was visual evidence for compromised cells, as observed by increasingly shriveled cell membranes (Figures 3D–F). While there are sometimes artifacts from negative strain preparations, here the images may indicate the secretion of large quantities of EPS around cells following 50 bar CO₂ incubations (Figures 3D–F). To investigate the EPS chemistry, we modified a biofilm EPS extraction protocol to isolate two “operational” EPS fractions, laEPS and bEPS, from the 0 and 4 h time points at 50 bar CO₂ for chemical characterization. EPS yields were low from all samples. This was not unexpected since biomass in the high pressure batch reactor were generally planktonic. Interestingly, ATR-FTIR spectroscopy did indicate chemical changes in lipid chemistry of the bEPS (EDTA-extractable) fraction after 4 h exposure to CO₂ (Figure 4). Predominant peaks representing CH₂- stretches of lipids were observed at 2846 and 2915 wavenumber (cm⁻¹) in the 0 h exposure. After the 4 h CO₂ exposure, these peaks were replaced with CH₃- lipid peaks at 2888 and 2944 wavenumber (cm⁻¹). No chemical differences could be conclusively identified within other classes of biomolecules (proteins, polysaccharides, nucleic acids) in the EPS fractions using ATR-FTIR spectroscopy (not shown).

Coupled to these physiological measurements, RNA was extracted from D. vulgaris biomass that had been exposed to CO₂ at atmospheric pressure, 25 bar, 50 bar, and 80 bar. Duplicate sets of biomass were generated for each condition. Following conversion of the extracted RNA to cDNA, this material was sequenced using the SOLiD platform. Sequenced reads were mapped to the D. vulgaris genome to identify gene expression patterns characteristic of a stress response. Despite the measurement of only limited metabolic activity at elevated pressures, shifts in transcript abundance were observed across these conditions, indicating that cells were responding to CO₂ stresses (Supplementary Material). Genes associated with chemotaxis (flagella subunits, methyl-accepting chemotaxis proteins) were clearly up-regulated across all CO₂ exposures, suggesting that cells were attempting to find more suitable environmental conditions (Table 1). However, more genes associated with two-component environmental sensing systems were down-regulated across the same set of conditions (Table 1). The regulation and expression of certain amino acids appears to play a key role in the D. vulgaris stress response to pressurized CO₂ conditions; relative to abundances at atmospheric pressure, transcripts for leucyl and isoleucyl-tRNA synthetases (DVU1196; DVU1927) were both up-regulated at 25 bar and 50 bar CO₂ pressures, while expression of a tRNA modification GTPase (TrmE; DVU1079) exhibited a similar pattern. Further evidence for a significant amino-acid based response to these stresses was inferred from genes associated with leucine biosynthesis. In addition to the leucyl-tRNA synthetase, transcripts from a 2-isopropylmalate synthase (DVU2981), and both the large and small subunits of a 3-isopropylmalate dehydratase (DVU2982 and DVU2983, respectively) were similarly up-regulated. This pattern was maintained in the 50 bar and 80 bar CO₂ exposures (Figure 5). Other general stress responses were observed during CO₂ exposure at 25 bar. The phage shock protein (Psp) system responds to extracytoplasmic stresses, and is induced by dissipation of proton motive force (PMF) in the periplasm (Darwin, 2005). However, although PsPA is the most abundant protein upon induction of the PsP system, little is currently known as to its physiological function. PspA was up-regulated at both 25 bar and 50 bar CO₂ exposures, possibly suggesting a loss of membrane integrity in some cells. In addition, a periplasmic Fe-hydrogenase that has been shown to play a key role in protecting against oxidative stress (Fournier et al., 2004) was up-regulated over the same conditions. Interestingly, the transcription of carbon starvation protein A (CstA) was up regulated under all pressurized conditions; given that lactate was provided in excess as a carbon and electron source, the cstA gene may play a role in a more generalized stress response. The role of hypothetical genes in many cellular functions was also emphasized via transcriptomic results. Across all three pressurized conditions approximately 25% of up- and down-regulated genes were hypothetical, hinting at unknown stress mechanisms and responses to CO₂ exposure.

In this study, we illustrate the negative effects of CO₂ exposure on microbial activity in the model organism D. vulgaris. As CO₂ injection into subsurface systems occurs, microorganisms are exposed both within the reservoir, and in shallower regions via CO₂ leakage and migration through cap rock (Harvey et al., 2012). Our findings indicate that even at relatively low pressures (10 bar) some microbial activity is significantly impacted by exposure to CO₂. Growth of D. vulgaris is significantly limited under all pressurized CO₂ conditions, and respiration (inferred from the measurements of microbial oxidation of lactate, and accumulation of sulfide) is similarly impacted. The rapid growth of D. vulgaris under N₂ pressurized conditions (up to 80 bar) suggests that pressure itself has no negative impact on cell health. Instead, toxic effects from CO₂ exposure linked to compromised membrane integrity would appear to be responsible for decreases in D. vulgaris activity in certain experiments. Kinetic measurements during pressurized CO₂ exposure experiments suggest that some microbial activity persisted for approximately the first hour, before respiration and cell growth ceased. Despite this cessation of growth under lower pressure conditions (10 bar and 25 bar CO₂), epifluorescence microscopy observations reveal that cells remained viable. In addition, the continued production of mRNA transcripts at all CO₂ pressures studied here further supports the observation that D. vulgaris cells remain viable during CO₂ exposures. These results support observations from environmental samples recovered from a saline aquifer in Germany, where SRB biomass remained viable for up to 30 days, despite the cessation of active sulfate reduction following CO₂ injection (Frerichs et al., 2013).

Electron microscopy images suggest that EPS may be produced in response to CO₂ exposure, potentially to buffer against deleterious effects. Given that previous studies have at least partly attributed increased resistance to CO₂ toxicity in biofilm forming bacteria to greater EPS production (Mitchell et al., 2008, 2009; Wu et al., 2010), D. vulgaris may utilize a similar approach. ATR-FTIR measurements on EPS extracted from biomass exposed to CO₂ showed little overall chemical changes in most classes of biomolecules except the lipid region. Notably, bEPS chemical shifts from CH₂ to CH₃ stretches may indicate that membrane lipids were being fragmented as a result of the CO₂ exposure. The relatively short exposure (4 h) of the D. vulgaris cells to 50 bar CO₂ may not have been adequate to completely access chemical changes in EPS composition, or the protective effect of biofilm formation under CO₂ exposure conditions. In contrast, measuring gene expression may provide a better representation of the physiological responses to short-term 50 bar CO₂ exposure. Gene transcript data indicates that D. vulgaris utilizes additional stress responses, including the accumulation of certain amino acids such as leucine. The accumulation of specific amino acids (in addition to other solutes) is a well-recognized osmotic stress response in many bacteria (Csonka, 1989); indeed, the up-regulation of several genes involved in leucine biosynthesis has been reported in response to KCl stress (Mukhopadhyay et al., 2006). Under salt stress conditions, the role of these amino acids is generally to buffer the cytoplasm and prevent hyperosmotic shock; glutamate and proline are two of the more well studied examples (Csonka et al., 1994; Amin et al., 1995). A potential role for leucine/isoleucine in CO₂ stress response is less clear, although these amino acids may be important components of other buffering or chaperone proteins, such as ferritin (Götze et al., 2014). Alternatively, amino acid production may be a more general stress response in this microorganism, stimulated via multiple factors. Likewise, phage shock proteins have been observed to increase in abundance under a range of stress conditions, including nitrate, heat, salt, and oxygen stresses (Chhabra et al., 2006; Mukhopadhyay et al., 2006, 2007; He et al., 2010). Previous studies with E. coli have identified increased bacterial motility in response to low pH (Soutourina et al., 2002). Although these experiments were all performed under similar pH conditions (~pH 6), CO₂ stress may have induced greater expression of genes characteristic of motility in D. vulgaris. However, unlike nitrate stress, there was little evidence of CO₂ stress resulting in up-regulation of genes associated with energy metabolism (He et al., 2010).

Using a model SRB system, we tested the combinations of stresses associated with deep subsurface CO₂ injection (elevated CO₂ concentrations, pH decreases, and high-pressure) to ascertain the impact of CO₂ injection on the physiology and metabolism of indigenous deep-subsurface bacteria. We observed that high pressures (80 bar N₂) had less inhibitory effects on cellular growth than CO₂ exposures at relatively low pressures. Although some of the pressures used in this study are lower than those in deep subsurface systems, these results indicate that upward migration of CO₂ from storage reservoirs would impact microbiology in shallower overlying materials (Harvey et al., 2012), with implications for carbon cycling and metal fate and transport. In addition, the potential for inhibition of sulfate reduction in the presence of pressurized CO₂ may help to limit deleterious subsurface processes such as sulfide-induced corrosion and oil field souring in the subsurface. However, it is worth noting that active microbial processes have been observed in CO₂ sequestration reservoirs (Morozova et al., 2010; Mu et al., 2014), likely catalyzed by microbial species that exhibit greater tolerance for scCO₂. Given the results of this study, injected CO₂ in subsurface systems is likely to act as a strong driving force for shifts in microbial community structure. The effects of these changes on ecosystem function are currently unclear. Therefore, having observed laboratory responses to CO₂ stresses in a model microbial strain (D. vulgaris), future work will attempt to understand the mechanisms that indigenous microorganisms use for survival and growth in CO₂-impacted deep subsurface environments, and the impact of microbial persistence on system biogeochemistry.