ABSTRACT
Glacial comminution of bedrock generates fresh mineral surfaces capable of sustaining chemotrophic microbial communities under the dark conditions that pervade subglacial habitats. Geochemical and isotopic evidence suggests that pyrite oxidation is a dominant weathering process generating protons that drive mineral dissolution in many subglacial systems. Here, we provide evidence correlating pyrite oxidation with chemosynthetic primary productivity and carbonate dissolution in subglacial sediments sampled from Robertson Glacier (RG), Alberta, Canada. Quantification and sequencing of ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) transcripts suggest that populations closely affiliated with Sideroxydans lithotrophicus, an iron sulfide-oxidizing autotrophic bacterium, are abundant constituents of microbial communities at RG. Microcosm experiments indicate sulfate production during biological assimilation of radiolabeled bicarbonate. Geochemical analyses of subglacial meltwater indicate that increases in sulfate levels are associated with increased calcite and dolomite dissolution. Collectively, these data suggest a role for biological pyrite oxidation in driving primary productivity and mineral dissolution in a subglacial environment and provide the first rate estimate for bicarbonate assimilation in these ecosystems. Evidence for lithotrophic primary production in this contemporary subglacial environment provides a plausible mechanism to explain how subglacial communities could be sustained in near-isolation from the atmosphere during glacial-interglacial cycles.
INTRODUCTION
Mechanical comminution of bedrock by glacial ice exposes fresh mineral surfaces that, in wet portions of the glacial bed, are poised for chemical weathering and release of solutes to downstream environments (1–3). Evidence from microcosm experiments suggests that rates of mineral denudation beneath Bodalsbreen Glacier, Norway, and Haut Glacier d'Arolla, Switzerland, are accelerated up to 8-fold by the activity of microorganisms compared to abiological rates (4). Montross et al. (4) attributed the increased rate of solute production in sediment-containing microcosms to mineral dissolution resulting from the production of acid from either (i) biologically catalyzed oxidation of pyrite (FeS2) and concomitant production of protons or (ii) dissolution of carbon dioxide from heterotrophic metabolism to form carbonic acid. When considered in the light of recent geochemical analyses that indicate significant production of solute transported by meltwaters from a number of globally distributed glaciers (see, e.g., references 3, 4, 5, 6, 7, 8, and 9), these data imply a central role for biological activity in driving the solute flux from sub-ice environments.
Molecular analyses of subglacial communities from a variety of systems also suggest a role for microbial activity in the weathering of subglacial minerals, in particular, FeS2 (8, 10, 11). For example, using RNA-based approaches, the active component of the bacterial community associated with subglacial sediment sampled from Robertson Glacier (RG), Alberta, Canada, was shown to include populations inferred to be involved in iron or iron sulfide oxidation, including putatively autotrophic Sideroxydans spp. (10) and Thiobacillus spp. (11). Consistent with this observation, a recent molecular analysis of minerals incubated in the outwash channel of RG for 6 months showed that FeS2 (a trace mineral [<1.0% dry weight] in RG carbonate bedrock) harbored a bacterial community most similar in composition to native subglacial sediment-associated bacterial communities compared to other iron-containing minerals such as iron oxides and silicate minerals such as olivine (11). The recovery of populations closely affiliated with the mineral sulfide-oxidizing and autotrophic genera Thiobacillus, Acidithiobacillus, and Sideroxydans from FeS2 incubated in the RG outwash environment (11) suggests that these organisms utilize this mineral as an energy source for chemolithoautotrophic production and growth. Consistent with this hypothesis, chemolithoautotrophic iron or sulfur oxidizers were shown to contribute to the oxidation of FeS2 in a glacial moraine at Midtre Lovenbreen Glacier, Svalbard, Norway (12, 13), suggesting that such processes are also of importance in recently deglaciated environments. Though it has been demonstrated that FeS2 is widespread in glacial catchments, evidence for the assimilation of inorganic carbon using energy derived from the oxidation of this mineral sulfide has not yet been documented.
A recent analysis of a number of geographically distinct glacial systems found that while sulfide oxidation (presumably from FeS2) represents the primary driver of mineral weathering, 14% to 37% of the total solute flux from these environments is likely attributable to microbial CO2 production generated through oxidation of organic carbon (9). Consistent with this calculation, numerous heterotrophic archaea, bacteria, and eukarya have been identified in subglacial ecosystems (10, 14–17) where high concentrations of dissolved organic carbon (DOC) and particulate organic carbon (POC) are also observed (18–24). Previous studies that examined the composition of DOC pools in meltwaters sampled from the Greenland ice sheet during high discharge (July) revealed signatures corresponding primarily to lignin and relict plant material likely sourced from the flushing of overridden soils (25). In contrast, analysis of DOC pools from early season meltwaters (May) revealed signatures indicative of more labile compounds such as protein and lipid, which were interpreted to reflect autochthonous in situ microbial production. An alternative possibility is that DOC from primary production on glacial surfaces (26–28) is introduced to the subglacial bed through moulins or crevasses. While the source of DOC in subglacial ecosystems remains enigmatic, it is likely that acidity generated through both its oxidation and the oxidation of sulfide minerals contributes to mineral dissolution in subglacial environments.
Here, using a sensitive microcosm-based radiotracer approach, we evaluated the potential for primary production in subglacial sediment-associated microbial communities and evaluated the potential for this activity to be driven by FeS2 oxidation. Quantitative RNA-based tools and geochemical analyses of RG meltwater were used to evaluate the extent to which the observed microcosm activities reflect putative processes occurring in situ at RG. Collectively, these data indicate that a potential energy source for driving primary productivity in subglacial ecosystems is the oxidation of sulfide minerals such as FeS2. Lithoautotrophic primary production could supply chemosynthate capable of supporting the diverse and abundant secondary consumers (heterotrophs) that have been identified in previous analyses in the subglacial environment at RG (10) and other glacial systems (14–16).
MATERIALS AND METHODS
Sample collection.Fine-grained basal sediments were collected aseptically from within ice caves that formed at the terminus of the glacier due to subglacial discharge at 12:00 p.m. in 2009 (samples Robinson East 09a [RE09a], RE09b, RW09a, and RW09b), 2010 (RE10), and 2011 (RE11) (Fig. 1) for use in RNA- and microcosm-based analyses. Sample locations were accessed through two ice caves that developed from two melt streams (Robertson East [RE] and Robertson West [RW]) in 2009. Samples were collected from RE only in 2010 and 2011, since that ice cave existed only during that time period. Sample collection was conducted wherein multiple samples from a given location (∼0.5 m2) were collected and pooled in order to minimize the influence of spatial heterogeneity on process rate measurements as well on as quantification and sequencing of transcripts as described previously (10). Samples (∼1 g) for DNA-based analyses were collected in sterile 1.5-ml microcentrifuge tubes with flame-sterilized spatulas and were immediately flash-frozen in a dry ice-ethanol slurry. Sediment aliquots (∼1 g) for RNA were collected in sterile 2-ml tubes containing 0.5 ml RNALater (Qiagen, Valencia, CA) and flash-frozen in a dry ice-ethanol slurry. Samples were stored on dry ice during transport to the field station and during transport back to Montana State University, where they were stored at −80°C until they were further processed. Sediments for microcosm assays (described below) were collected from the same locations in 2010 as those for RNA analyses using a flame-sterilized spoon. These sediments were placed in a sterile 500-ml screw cap container and were frozen immediately using a dry ice-ethanol slurry. These samples were also stored on dry ice during transport to the field station and during transport back to Montana State University, where they were stored at −80°C until they were further processed.
Location of subglacial and supraglacial sampling sites at Robertson Glacier (Table 1). Samples are labeled to indicate their respective environments as follows: RE, Robertson East; RW, Robertson West, SUP, supraglacial. The last two digits indicate the year the samples were collected and include “a” and “b” to denote multiple samples taken from a given location. Samples were collected in 2009, 2010, and 2011.
Geochemical methods.Collection of melt water for geochemical characterization was performed as previously described (22). Briefly, water temperature and pH were measured at the time of sample collection with a WTW 330i meter and probe. Conductivity and temperature were measured with a YSI 30 conductivity meter. Water for geochemical analysis was collected and filtered using 0.8- and 0.2-μm-pore-size Supor syringe filters (Acrodisc 32-mm-diameter PF syringe filter with 0.8- and 0.2-μm Supor membranes). Membranes were flushed with 20 ml of sample prior to collection to minimize contamination. Samples for ion chromatography (IC) were collected in acid-cleaned and sterile 60-ml Nalgene bottles. Analysis of the major cation and anion concentrations via ion chromatography was performed as previously described (22). Major cations were analyzed with a Dionex DX 120 IC system (Dionex, Sunnyvale, CA) and major anions with a Dionex DX 600 dual IC system. The certified standards Dionex Combined 6 Cations Standard II and Alltech Multicomponent Certified Anion Standard Mix 6 were used to quantify cations and anions, respectively. Samples were analyzed in duplicate for major cations and anions. Meltwater DOC and dissolved inorganic carbon (DIC) concentrations were measured with an OI Analytical model 1010 wet oxidation total organic carbon (TOC) analyzer. Samples were analyzed in duplicate (analytical variability) for major cations and anions, and the average values are presented. Values from replicate analyses did not differ by greater than 2.1% (data not shown).
Dissolved inorganic carbon uptake potential.A microcosm approach was used to assess the potential for the subglacial-sediment-associated microbial populations sampled in 2010 (sample site RE10) to assimilate inorganic carbon, as described previously (29, 30). As described above, sediments for use in microcosms were collected on 14 October 2010 and were stored at −80°C until they were thawed overnight at 4°C (4 March 2011) for use in microcosm assays. Approximately 10 g of fine-grained sediments (∼82% dry solid content as determined by drying at 90°C for 24 h) was added aseptically to preautoclaved 70-ml serum bottles. Sediments were overlaid with 30 ml of bicarbonate buffered distilled water (5 mM final concentration, pH 8.0) and capped with butyl rubber stoppers. Triplicate biological and triplicate killed controls were prepared. Two treatments of autoclave sterilization (121°C, 30 min), with a 24-h incubation (4°C) between treatments, were used to sterilize sediments. Reactions were initiated by addition of NaH14CO3 (MP Biomedicals) (422 MBq/mmol) to a final concentration of 7.6 × 103 Bq ml−1, and the reaction mixtures were incubated at 4°C in the dark.
The amount of 14C incorporated into biomass was determined every 2 weeks for the first 2 months and bimonthly thereafter for a total of 6 months, using previously described methods (29). Briefly, microcosms were shaken to uniformly resuspend sediments and a 2-ml subsample was removed aseptically. Samples were acidified by adding 1 ml concentrated (1 N) HCl to volatize unreacted NaH14CO3 (the final solution pH was <2.0). The acidified samples were shaken vigorously and allowed to vent in a fume hood for ∼1 h under a stream of N2. Samples were filtered through white 0.22-μm-pore-size polycarbonate filters, and the filters were washed with sterile distilled water, dried at 90°C for 24 h, and weighed. Filters were transferred to scintillation vials containing 10 ml of CytoScint scintillation cocktail (MP Biomedicals, Irvine, CA), and the radioactivity associated with each sample was quantified with a 1900CA Tri-Carb liquid scintillation counter (Packard, Downers Grove, IL).
Two additional replicate microcosms that did not contain added NaH14CO3 were prepared as described above in order to monitor the concentration of dissolved inorganic carbon (DIC) in the microcosms during incubation. DIC determinations were made at 64 and 176 days postinoculation using an on-line TOC-VCSH, TOC/total inorganic carbon (TIC), and total nitrogen (TN) analyzer (Shimadzu Scientific Instruments, Columbia, MD). The concentrations of DIC did not differ significantly between replicates, and did not differ significantly between samples collected at 64 and 176 days postinoculation, and averaged 8.2 mM among these four samples. Conversion of the rates of carbon assimilation or mineralization based on calculation of the uptake of 14C tracers compared to the total uptake (14C plus 12C) was performed using the methods of Lizotte et al. (31). Briefly, uptake rates (biological minus killed controls) were calculated by multiplying the value for the uptake of 14C-labeled substrate by the total effective concentration (14C-labeled substrate plus native substrate). Recognizing that isotopic discrimination factors differ for different autotrophic processes (as summarized in Havig et al. [32]), we adopted the uniform value of 1.06 as the isotopic discrimination factor for this study, as previously described (31). All uptake values were multiplied by 1.06 and were then normalized to the grams of dry mass sediment in each 2-ml aliquot. The averages and standard deviations of the results of three replicate treatments are presented.
A series of controls were performed to account for quenching due to the solubilization of organic compounds present in the filtered sediments by the scintillation cocktail. Here, the same mass of subglacial sediments present on filters was added to scintillation vials, acidified, and dried as described above. A 10-ml volume of scintillation cocktail was added to each vial, and NaH14CO3 was added in the amount of 4, 8, or 12 MBq per vial. Each treatment was performed in triplicate, and the result was counted by liquid scintillation as described above. The dpms in vials lacking sediments were on average a factor of 1.67 greater than the dpms in vials that contained sediments, regardless of the amount of NaH14CO3 added. To account for quenching of dpms by leached organic materials in microcosm assays, uptake rates were multiplied by a factor of 1.67.
Incorporation of 14C into different cellular carbon pools was quantified following the method of Brock and Brock (33) as previously described (29) following the termination of the microcosm incubation at 176 days. Details of the protocol used are provided in the supplemental material. The remaining supernatant was used to quantify sulfate concentrations in the microcosm samples using Hach Sulfate reagent powder pillows and a Hach DR/200 spectrophotometer (Hach Company, Loveland, CO).
RNA extraction and cDNA synthesis.RNA was extracted from subglacial sediment with a FastRNA Pro Soil-Direct kit (MP Biomedical, Solon, OH) and further purified using a High Pure RNA isolation kit (Roche, Indianapolis, IN) as previously described (10). RNA was extracted in triplicate from approximately 400 mg of wet subglacial sediment. The concentration of purified RNA was determined using a Qubit RNA assay kit (Molecular Probes) and a Qubit 2.0 Fluorometer (Invitrogen). cDNA was synthesized from 15 ng of purified RNA with an iScript cDNA synthesis kit (Bio-Rad, Hercules, CA) using the following reaction conditions: 5 min at 25°C, 30 min at 42°C, and 5 min at 85°C. Following synthesis of cDNA, samples were purified by ethanol precipitation and resuspended in nuclease-free water for further analyses, as described previously (10).
PCR amplification of cbbL from cDNA.The cbbL gene, which encodes the ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) protein that functions in the Calvin-Benson-Bassham (CBB) cycle, was selected for use in characterizing the abundance and composition of autotrophs in subglacial sediments, since 16S rRNA transcript sequence data from the same sediments (10) indicated a prevalence of bacterial phylotypes inferred to utilize this autotrophic pathway, including numerous Fe- and S-oxidizing proteobacteria (e.g., Sideroxydans and Thiobacillus spp.). An approximately 1,100-bp fragment of the large subunit of the “green-like” form I RuBisCO was amplified using primers cbbL-G1F (5′-GGCAACGTGTTCGGSTTCAA-3′) and cbbL-G1R (5′-TTGATCTCTTTCCACGTTTCC-3′) and an annealing temperature of 57°C (34). An approximately 800-bp fragment of the large subunit of the “red-like” form I RuBisCO was amplified using primers cbbL-R1F (5′-AAGGAYGACGAGAACATC-3′) and cbbL-R1R (5′-TCGGTCGGSTAGTTGAA-3′) and an annealing temperature of 62°C (34). For each set of primers, ∼1 ng of purified cDNA was subjected to 35 cycles of PCR in triplicate using reaction conditions and PCR conditions reported previously (22). Equal volumes of the triplicate reactions were pooled, purified, and cloned, and the resultant plasmids were purified for use as standards in quantitative reverse transcription-PCR (qRT-PCR) assays as previously described (35).
qRT-PCR.qRT-PCRs were performed with a Power SYBR green RNA-to-CT one-step kit from Invitrogen (Carlsbad, CA) according to the manufacturer's protocol, and these were assayed on a RotorGene-Q real-time PCR detection system from Qiagen (Valencia, CA). Reactions were performed in triplicate, with 10 ng of total RNA quantified as described above, with 500 nM forward primer and reverse primer, in a final reaction volume of 20 μl using the following cycling conditions: reverse transcription at 48°C (30 min) followed by initial activation of the DNA polymerase at 95°C (10 min) followed by 40 cycles of denaturation at 95°C (15 s), annealing, and extension (at the optimal annealing temperature for the cbbL red or cbbL green templates as described above for 1 min). The specificity of the qRT-PCR assays was verified by melt curve analysis. Control reaction mixtures contained either no reverse transcriptase or no template RNA. The dynamic range and sensitivity of the qRT-PCR assays are reported in the accompanying supplemental material.
cbbL sequencing and analysis.Form green and form red cbbL cDNA amplicons derived from sample site RE10 (Fig. 1) were sequenced with a 454 Genome Sequencer FLX system (Roche, Nutley, NJ) at the Research and Testing Laboratory (Lubbock, TX) using the primers described above. Each sample was sequenced once from the 5′ end of the amplicon (i.e., the N terminus of the encoded protein). Postsequence processing was performed using the mothur (ver. 1.24.1) sequence analysis platform (36). Barcodes and primers were removed, and sequences were trimmed to an empirically defined minimum length of 400 bp and were subjected to a filtering step using the quality score files to remove sequences with anomalous base calls. Trimmed sequences were translated using MEGA5 (37); sequences that did not translate in frame over this length of the gene were deleted without further consideration. MEGA5 was also used to align translated sequences with default alignment parameters specified. The aligned sequences were reverse translated, the DNA alignment was imported back into Mothur, and unique sequences were identified. Operational taxonomic units (OTUs) were assigned at a sequence similarity of 97.0% using the average-neighbor method. The remaining sequences were randomly subsampled in order to normalize the total number of sequences in each library. Collectively, these steps resulted in normalized sizes of 144 and 206 cbbL form green and form red sequences for each library, respectively. Sequences were classified using BLASTx (see Tables S2 and S3 in the supplemental material).
Nucleotide sequence accession number.The raw sequence libraries and quality score files have been deposited in the NCBI SRA database under accession number SRR1037420.
RESULTS AND DISCUSSION
The primary cations present in RG melt waters collected in 2009 to 2011 and in those collected in 1994 to 1999 (6) were Ca2+ and Mg2+. The primary anion was presumed to be HCO3− based on dissolved inorganic carbon measurements and the determination of DIC as predominately HCO3− at pH 8.1 to 8.8 (Table 1), consistent with the presence of both calcite (CaCO3) and dolomite [CaMg(CO3)2] in local bedrock (6, 38) and glacial sediments (39). NH4+, NO2−, and PO43− values were below the detection limit and thus are not reported (Table 1). The combined abundances of Ca2+ and Mg2+ exhibited a positive and significant relationship with SO42− (R2 = 0.98 and 0.99, respectively). Previous sulfur isotope measurements showed that the presence of SO42− in RG glacial meltwaters is the result of FeS2 oxidation and not gypsum dissolution (11). Considering that FeS2 represents up to 1% of the dry weight of subglacial sediment and bedrock and that it is the only sulfide mineral identified in these sediments to date (11, 39), these results suggest that the production of acidity during FeS2 oxidation promotes the dissolution of calcite and dolomite. It is also possible that the biological production of organic acids or CO2 may also promote carbonate dissolution (4, 9); however, this was not investigated in the current study.
Aqueous geochemical measurements associated with samples analyzed in the present studya
The Ca2+/Mg2+ ratio in subglacial melt waters measured in this study and previously by Sharp et al. (6) ranged from 2.4 to 11.9, depending on the season and the time of sampling. Congruent dolomite dissolution would produce subglacial melt waters with a Ca2+/Mg2+ molar ratio of 1.0, whereas congruent dissolution of equimolar amounts of dolomite and calcite would produce a Ca2+/Mg2+ molar ratio of 2.0, and Ca2+/Mg2+ molar ratios of >2.0 would reflect a dominance of calcite dissolution if carbonate mineral dissolution were congruent. Laboratory experiments performed on freshly crushed (impure) calcite designed to simulate subglacial conditions demonstrate incongruent dissolution, with enhanced release of elements such as Mg relative to the bulk mineral composition (40). Thus, it is possible that similar processes occur in the RG subglacial environment, such that higher Mg/Ca ratios may reflect either incongruent dissolution of impure calcite or enhanced congruent dolomite dissolution. A plot of the Ca2+/Mg2+ ratio as a function of the concentration of SO42− reveals an inverse trend (Fig. 2A), and lower Ca2+/Mg2+ ratios are accompanied by increased Mg2+ and DIC concentrations (Table 1 and Fig. 2B). The waters with the highest solute (Ca2+, Mg2+, SO42−, and DIC) concentrations were collected late in the melt season (October) (Table 1 and Fig. 2). This was when discharge was low and most of the outflowing water was routed subglacially, leading to increased water-rock contact times relative to times earlier in the melt season (July to September). The increase in Mg2+ concentrations in waters with greater water-rock residence times suggests that increased dolomite dissolution relative to calcite dissolution is associated with increasing SO42− concentrations in subglacial meltwaters rather than with an increase in incongruent calcite dissolution, which is consistent with the findings of Sharp et al. (6). Taken together, these observations support the hypothesis that FeS2 oxidation and the production of protons are driving the dissolution of both calcite and dolomite in the subglacial system and are consistent with the relationships noted between SO42− production and carbonate dissolution in this system (6) and in other glacial systems (3–5).
(A) The molar ratio of Ca2+ to Mg2+ in supraglacial and subglacial melt waters sampled from RG plotted as a function of the concentration of SO42−. (B) The molar ratio of Ca2+ to Mg2+ in supraglacial and subglacial melt waters plotted as a function of the DIC. Data from 1994 to 1996, as reported by Sharp et al. in 2002 (6), are also plotted.
Microcosm-based studies indicate that the production of sulfate in glacial sediments is accelerated 8-fold by microbial activity (4). This finding, coupled with (i) the results of recent molecular analyses of RG sediments indicating the presence of active bacterial populations inferred to be involved in iron or sulfur oxidation in the RG subglacial environment (10, 11) and (ii) the general phenotype of iron- or sulfur-oxidizing bacteria being autotrophic (41), prompted a microcosm-based study to quantify the rate of DIC assimilation and to determine if this activity is associated with SO42− production.
Significant incorporation of [14C]bicarbonate into sediment-associated biomass over the first 28 days of incubation, relative to the level seen with killed controls (pairwise Student t test significance [P] of 0.05), provided the first evidence that RG subglacial assemblages are capable of assimilating DIC. The rate of DIC assimilation over this time interval was 22.7 ± 13.4 ng C/gram dry weight sediment (gdws) per day (Fig. 3). The amount of DIC assimilated in killed controls did not change during the course of the incubation (data not shown). A previous characterization of the abundance and composition of archaeal, bacterial, and eukaryal 16S and 18S rRNA templates in RG subglacial sediments collected at the same time and from the same location as those used in the microcosm experiments allows the normalization of the DIC assimilation rate to a per cell level. Using an estimate of 9.2 × 107 16S rRNA gene templates/gdws (10) and making the assumptions that (i) proteobacterial cells on average harbor two 16S rRNA gene templates (42) and (ii) 50% of the active community is autotrophic or mixotrophic (an estimate based on the abundance of inferred autotrophic bacteria and autotrophic methanogens [10]), this rate converts to 0.6 ± 0.3 × 10−17 mol DIC/cell per day. Intriguingly, this rate is only ∼1 order of magnitude lower than the rate of DIC assimilation (6.7 ± 3.9 × 10−17 mol/cell per day) previously estimated in deep marine sediments (43), which further highlights the similarities between the cold, dark, subsurface ecosystems in subglacial and marine environments with respect to metabolic transformations noted previously (23).
Assimilation of DIC in microcosm assays containing RG subglacial sediments (sampled from site RE10) during dark incubation at 4°C. Values represent the average differences in assimilation among triplicate biological and triplicate killed controls.
The rate of DIC assimilation over the 176 days of incubation was 14.2 ± 7.8 ng C/gdws per day. Over the course of this study (176 days incubation), 90.1% of the DIC assimilated was recovered in cellular protein whereas 4.6% was recovered as nucleic acid, with the remaining 4.4% recovered as low-molecular-weight compounds, indicating that chemosynthate is not partitioned equally into cellular pools. Previous studies suggest preferential partitioning of carbon toward protein in chemotrophic communities experiencing nutrient limitation such as those in hot spring environments (91.0% of total C in a cellular protein pool) (33) and photosynthetic communities, the latter of which were interpreted to reflect low growth rates and enhanced protein turnover (44). While RG subglacial communities have previously been shown to be N limited (22), the biochemical basis for chemosynthate partitioning in this system is unclear. The abundance and availability of DOC and other nutrients, which have been demonstrated to vary considerably in glacial systems (18–21), may also impact the partitioning of chemosynthate among various cellular pools.
Over the course of the incubation, a total of 12.6 ± 7.2 μmol of SO42− was released, which corresponds to a rate of 1.5 ± 0.9 nmol SO42−/gdws per day. For comparison, when not normalized to dry mass, the amount of SO42− produced by RG sediment microcosms was 1.3 ± 0.7 nmol SO42−/g sediment per day, which is closer to biological SO42− production values observed previously in sediments collected from Haut Glacier d'Arolla, Switzerland (0.7 nmol SO42−/g sediment per day), than to those observed in sediments collected from Bodalsbreen Glacier, Norway (13.5 nmol sulfate/g sediment per day) (4). Sulfate SO42− production levels reported in the Bodalsbreen study, in which the glacial sediments were crushed, suggest that crushing and exposure of fresh FeS2 surfaces are likely to influence rates of oxidation. Alternatively, these data may indicate that RG is biogeochemically more similar to Haut Glacier d'Arolla than to Bodalsbreen.
The amount of DIC assimilated over the 176-day incubation period was 2.46 ± 1.37 μg C/gdws or 0.21 ± 0.11 μmol C/gdws, while the rate of SO42− production was 1.5 ± 0.9 nmol/gdws per day. If the majority of DIC was assimilated using energy obtained from aerobic oxidation of FeS2 persulfide as described using reaction 1 below (2 mol SO42− produced per 1 mol FeS2 oxidized), a reaction stoichiometry that ranges from ∼7 to 24 μmol SO42− produced/μmol C fixed is reached.
It has also been suggested that, under the anoxic conditions which are thought to develop in localized portions of the glacial bed (3, 20, 48), FeS2 oxidation proceeds abiotically with Fe3+ as an oxidant (5). While the potential for abiotic SO42− production through oxidation by Fe3+ cannot be discounted on the basis of the data presented here, its occurrence would further decrease the reaction stoichiometries defined above to levels that would be below the predicted minimum stoichiometry level supporting this reaction. Moreover, the (i) lack of a significant decrease in the rate of CO2 fixation in microcosms over time suggests that the preferred oxidant supporting this activity (i.e., O2) is unlikely to have been depleted and the (ii) lack of significant SO42− production in killed controls indicates that this process is unlikely to have occurred abiotically.
Transcripts (i.e., cDNA) of cbbL, which encode a protein involved in the Calvin Benson Bassham (CBB) reductive pentose phosphate carbon fixation pathway, were quantified and sequenced in order to examine the distribution, abundance, and diversity of populations putatively involved in iron or sulfur oxidation. The cbbL gene was chosen as a functional target since the majority of putative iron- and sulfur-oxidizing bacteria that have been identified in RG subglacial sediments to date (10, 11) are thought to use this pathway to assimilate DIC (41). Two degenerate primer sets that target two major lineages of bacterial cbbL (denoted “form red” and “form green” [34]) were used. Form green cbbL is common in plants, algae, and alpha-, beta-, and gammaproteobacteria, whereas form red cbbL exhibits a more limited distribution among nongreen algae and alpha- and betaproteobacteria (34). The abundance of cbbL red templates was greater than that of cbbL green templates (Fig. 4A) in sites RW09a, RE09a, RE09b, and RE11. cbbL form green transcripts were more abundant than form red transcripts in site RE10, while the abundances of form red and form green cbbL transcripts were similar in site RW09b. It is possible, however, that the long-term storage (<30 months) of sediment samples at −80°C and in the presence of RNAlater led to degradation of RNA, which may have confounded the comparisons identified above. With this being said, there is no obvious trend indicating lower cbbL transcript abundances in samples collected earlier in the study period compared to more recently collected samples, which might be expected if RNA degradation were occurring.
(A) Abundances of cbbL form green and form red transcripts from subglacial sediment samples. (B and C) The composition of cbbL form green (B) and form red (C) in cDNA derived from subglacial sediment collected from the Robertson East location in 2010 (RE10).
In order to identify the active CBB-associated members of the microbial community used for the DIC assay, we sequenced cDNA of cbbL forms green and red from subglacial sediments collected in 2010 (RE10). Following quality screening, which included manual removal of sequences with frameshift errors, a total of 144 cbbL form green and a total of 206 form red cDNA sequences were obtained (see Table S1 in the supplemental material). Rarefaction analysis suggests that 97.7% and 94.1% of the predicted cbbL form green diversity and form red diversity were sampled, respectively. The cbbL green cDNA assemblage is dominated (63.8% of total sequences) by two phylotypes that exhibit close affiliation (95% and 100% sequence identities; see Table S2) with cbbL from Sideroxydans lithotrophicus ES-1 (Fig. 4B). This finding is consistent with the results of our previous transcriptional analysis of the same sediment extract which indicated that the 16S rRNA closely affiliated (97% sequence identity) with S. lithotrophicus ES-1 dominated (∼23% of total sequences) the bacterial community (10). Previous studies indicated that the Sideroxydans genus is comprised of obligate autotrophs that oxidize ferrous iron and solid-phase iron sulfide under microaerobic conditions (49), suggesting that these phylotypes may be responsible for the coupled iron sulfide oxidation and microbial DIC assimilation in RG sediments. A number of autotrophs potentially involved in the oxidation of intermediate FeS2 oxidation products (e.g., thiosulfate) or other reductants were also identified, which may help to explain the low stoichiometric ratio of FeS2 oxidized to DIC assimilated in microcosm assays. For example, two subdominant cbbL form green phylotypes (18.8% of total sequences) were identified that are closely affiliated (both 95% sequence identities) with cbbL from thiosulfate-oxidizing facultative anaerobic autotroph Sulfuricella denitrificans skB26 (50). In addition, a single cbbL green phylotype was identified that is closely affiliated (93% sequence identities) with cbbL from Cupriavidus metallidurans CH34 (6.9% of total sequences), a facultative chemolithotroph capable of driving CO2 assimilation through hydrogen oxidization (51).
The cbbL red cDNA assemblage is dominated (67.5% of the total) by 9 phylotypes that exhibit close affiliation (93% sequence identities) with proteobacterial methanotroph Methylibium petroleiphilum PM1 (Fig. 4; see also Table S3 in the supplemental material). Surprisingly, our prior transcriptional analysis of 16S rRNA from the same sediments did not indicate the presence of sequences affiliated with M. petroleiphilum PM1 (10), which may reflect bias in the bacterial 16S rRNA primers employed in our prior study. While autotrophic growth in the type strain of M. petroleiphilum was not demonstrated during its detailed characterization (52), other proteobacterial methanotrophs have shown to grow autotrophically using the CBB (53). In addition, seven cbbL red phylotypes representing 12.9% of the total sequences that exhibit a close affiliation (95% to 98% sequence identities) with cbbL from autotrophic ammonia oxidizer Nitrosospira multiformis ATCC 25196 were detected in the subglacial assemblage, consistent with previous evidence of ammonia-oxidizing activity in the subglacial environment at RG (22). Nitrification activity has also been detected in other subglacial ecosystems (20, 54), which may indicate a broader role for nitrification activity in generating reductant to drive primary production.
Note that autotrophic pathways other than the Calvin cycle are likely to also contribute chemosynthate to the ecosystem. Indeed, our previous molecular and physiological analyses of subglacial sediment microbial communities (10, 23) indicated the presence of ∼3,000 active hydrogenotrophic methanogen cells per gram dry weight sediment, suggesting an important role for the reductive acetyl coenzyme A pathway of CO2 fixation. The sequencing of subglacial sediment community genomes will provide a wider perspective on the importance of different autotrophic metabolisms in primary production in this and other subglacial sediment ecosystems.
Concluding remarks.The data presented here suggest a potential role for subglacial microbial activity in FeS2 oxidation, carbonate mineral weathering, and solute acquisition in glacial melt waters. The data also suggest that chemical energy generated during the oxidation of FeS2 (and Fe and S reaction intermediates) is used to drive the biological assimilation of DIC, which is likely to be derived from carbonate mineral weathering. It is possible that metabolites from primary production support the diversity of secondary consumers previously identified in the subglacial environment at RG and other glaciers, resulting in the complex microbial food webs thought to exist in these systems (10). If these chemolithoautotrophic processes occur beneath ice sheets as has been suggested previously (9, 55), then these communities may have served as drivers of continental weathering during periods in Earth's history with greater ice cover than in the present.
ACKNOWLEDGMENTS
This work was supported by NASA grants NNX10AT31G and NNA13AA94A. T.L.H. acknowledges support for this work from the NASA Postdoctoral Program.
We thank the Biogeosciences Institute at University of Calgary's Kananaskis Field Station for the use of field and laboratory facilities. We thank Peter Canovas, Kristopher Fecteau, and Natasha Zolotova for help with field work or laboratory analyses.
FOOTNOTES
- Received 13 June 2014.
- Accepted 20 July 2014.
- Accepted manuscript posted online 1 August 2014.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/AEM.01956-14.
- Copyright © 2014, American Society for Microbiology. All Rights Reserved.
