Large perturbations in CO2 flux and subsequent chemosynthesis are induced in agricultural soil by the addition of elemental sulfur

The microbial contribution to soil organic matter has been shown to be much larger than previously thought and thus it plays a major role in carbon cycling. Among soil microorganisms, chemoautotrophs can fix CO2 without sunlight and can glean energy through the oxidation of reduced elements such as sulfur. Here we show that the addition of sulfur to soil results in an initial surge in production of CO2 through microbial respiration, followed by an order of magnitude increase in the capture of carbon from the atmosphere as elemental sulfur is oxidised to sulfate. Thiobacillus spp., take advantage of specific conditions to become the dominant chemoautotrophic group that consumes CO2. We discern the direct incorporation of atmospheric carbon into soil carbohydrate, protein and aliphatic compounds and differentiate these from existing biomass. These results suggest that chemoautotrophs can play a large role in carbon cycling and that this carbon is heavily influenced by land management practises.


CO 2 efflux.
The net change in atmospheric CO 2 produced by the incubated soil over a period of twelve weeks was continuously measured. All external influences of CO 2 were removed after careful measurements of background artefacts. The remaining data represented the net change in CO 2 within the internal chamber atmosphere, due solely to soil gas flux activity. Treatments S U and S SA were measured to a high degree of resolution (ng min -1 ) to determine the average rate of change (See Supplementary Information Table 2) and hence the efflux rates for each week (Fig. 1A). Average efflux rates per week (n = 12) of CO 2 for the sulfur amended soil (S SA ) and control (S U ) were, 0.63 ± 0.24 and 0.32 ±0.09 µg CO 2 week -1 kg -1 , respectively (Fig. 1A, Supplementary Information Table 3). The total mass of CO 2 produced by both treatments were 7.59 and 3.86 µg CO 2 kg -1 , respectively. S SA produced an initial rapid increase in CO 2 efflux, almost double that of S U , up to week nine (Fig. 1A). The average CO 2 efflux rate then fell after week nine, with the period of lowest CO 2 efflux at week ten (0.01 µg CO 2 week -1 kg -1 ), which was concurrent with other data presented here. The rate of efflux for S U represented normal background respiration under constant temperature and water holding capacity. For the first four weeks of incubation the rate of CO 2 efflux was on average 0.48 ± 0.08 µg CO 2 week -1 kg -1 . For the remaining eight weeks the rate of CO 2 efflux settled down to a mean value of 0.24 ± 0.05 µg CO 2 week -1 kg -1 . Therefore, after homogenisation and wetting, S U appeared to establish a relatively steady state of efflux after ~four weeks incubation.
At week one, both treatments had very similar efflux rates (4.5% difference), indicative of their shared origin from a single homogenised soil sample. The rates of efflux between treatments diverged after week one by 50.8%, most likely due to the stimulatory effect of added S 0 . The box and whisker plot (Fig.S1) illustrates the difference between the two treatment medians and clearly shows the variability induced by sulfur addition. Outlier for the sulfur amended soil (S SA) was identified as week ten, demonstrating the extreme reduction of efflux rates during that time period. Overall, the two treatments were significantly different over the course of the experiment (p = < 0.001) and we estimate that sulfur addition produced the lowest rate of CO 2 efflux at week ten before returning to background levels of respiration.

13 CO 2 tracking with compound specific stable isotope ratio mass spectrometry (CSIA)
Combining 13 C isotope studies with compound specific stable isotope ratio mass spectrometry (CSIA) of phospholipid fatty acids (PLFA) provides taxonomic and quantitative information on the microorganisms utilising a given 13 C labelled substrate 1,2,3 . PLFA biomarkers are found solely in the cell membrane of living organisms. On cell death the PLFA's degrade rapidly; therefore PLFA's are biomarkers for viable cells only (White et al. 1979) 4 . The specificity of PLFA's as biomarkers for active microbes is improved by stable isotope studies as microbes utilising the 13 C labelled substrate incorporate 13 C into their PLFA's (Evershed et al. 2006) 5 . PLFA's are identified by gas chromatography mass spectrometry (GCMS) and their δ 13 C values are measured by compound specific gas chromatography-combustion -isotope ratio mass spectrometry (GC-C-IRMS). Here we measured δ 13 C values of PLFA's from the incubated soils to assess the impact of sulfur addition on soil autotrophy. The average δ 13 C values of all PLFA's in soils S U and S SA at the different time points are plotted in Figure 1 ( Table S4). The higher δ 13 C values obtained for PLFA's at weeks 4, 8 and 12 confirms incorporation of 13 CO 2 into both the S U and S SA soils. However, the S SA soil (added sulfur) has significantly higher δ 13 C values at T4, T8 and T12 compared to S U (no added sulfur) indicating higher levels of CO 2 sequestration by soil microbes in S SA . The average δ 13 C values of the S U soil increase from -31.74‰ to -21.58‰ between T0 and T4 (P<0.05) but then appear to stabilise and do not increase or decrease significantly (P>0.05) between T4, T8 or T12. The average δ 13 C values of the S SA soil also increase (P<0.001) between T0 and T4. However, in contrast to S U , S SA 's δ 13 C values do not stagnate and 13 CO 2 incorporation continues to increase (P<0.05) from 1.35‰ (T4) to 11.42‰ (T8) and a final value of 29.83‰ (T12). In comparison the S U soil has a much lower δ 13 C value of -22.43‰ at the end point (T12). The abundances of the PLFA's reported do not change significantly from T0 to T12 in either soil, reflecting the sequencing results (Table 4 supplementary data). Of the PLFA's not reported here, there is a notable decrease in the abundance of polyunsaturated fatty acids 20:4ω6, 20:5ω3 and 20:3ω6, which are often used as biomarkers for species of protozoa 6,7,8 . These PLFA's decrease by a factor of 10 between T0 and T4, indicating that the incubation conditions do not favour these organisms.
Over the 12 week incubation of soil S SA the majority of the 13 C label was incorporated into gram negative bacterial PLFA's, 16:1ω7, cy17:0 and 18:1ω7, with a final average δ 13 C value of 173.07‰ in the S SA soil compared to -8.69‰ in S U . 16:1ω7 and 18:ω7 have been reported as key fatty acids in Pseudomonas sp. PS+ 7 . 18:2ω6 and 18:1ω9 are frequently used as fungal molecular markers 9 . While fungi have incorporated 13 C into their fatty acids it is difficult to know if this is due to cross feeding or CO 2 uptake. 16:1ω5 is a signature biomarker for arbuscular mycorrhizal fungi (AMF) 10 and is significantly enriched in the sulfur amended soil (S SA ) with a δ 13 C value of 21.71‰ at T12. AMF are a major functional group in the sequestration of plant derived C to rhizosphere microorganisms 8 but here they appear to be incorporating 13 C into their PLFA's through a different pathway. Iso and anteiso branched C15, C16 and C17 fatty acids are used as biomarkers for gram positive bacteria 11 and incorporation of 13 C into their fatty acids was much lower than gram negative and fungal fatty acids, as observed by Butler et al, 2003 3 . The saturated PLFA's 16:0 and 18:0 also show high levels of enrichment, especially in S SA with δ 13 C values of 184.49‰ and 3.55‰ respectively. Although these biomarkers are non-specific, 16:0 is an important intermediate in the cellular biosynthesis of fatty acids 3 . Iso17:1ω7 has a final δ 13 C value of 16.06‰ and has been reported as a biomarker for sulfate reducing bacteria. Overall, these results show that the addition of sulfur leads to a large increase in 13 C incorporation into the PLFA's of gram negative bacteria and fungi.

RubisCO gene (cbbL) copy abundance.
Over the 12 week incubation RubisCO gene copies (cbbL) were monitored (Fig 1E) to establish any variations in abundance over the course of the experiment. The presence of the cbbL gene indicates the potential for atmospheric CO 2 uptake through the Benson-Bassham (CBB) reductive pentose phosphate pathway (Tourova et al, 2010) 12 . Prior to incubation, cbbL gene copies were determined to be 6.2 X 10 3 ng -1 soil DNA. Analysis of cbbL abundance after 2 weeks shows a negligible rise in S SA to 6.4 X 10 3 ng -1 DNA whilst there is a more notable increase in S U to 7.9 X 10 3 ng -1 DNA. The absence of light may encourage microbes that harbour the cbbl gene, but the addition of sulfur appears to have temporarily inhibited their growth. At approximately 2.5 weeks however the cbbL copies increase rapidly in S SA whilst remaining low in S U . The beginning of this large increase coincides with the beginning of the transformation of Sº to SO 4 2strongly indicating that the SO 4 2is acting as an electron donor. The availability of chemical energy has also stimulated the microbial community which is reflected in an increased production of CO 2 through microbial respiration. The peak in cbbl abundance occurs at week 8 (2.3 X 10 4 ng -1 DNA), just as the CO 2 uptake event occurs, again strongly suggesting that SO 4 2facilitated chemosynthesis via the CBB pathway is taking place. By week 12 the cbbL gene copies decrease sharply in both soils, falling to levels much lower than those at the beginning of the experiment. Nutrient exhaustion may cause this collapse but CO 2 fixation may still be possible through other pathways.

Microbial Community Changes
The microbial phyla identified in both incubated soils resemble those already reported for diverse soils 13 . The most notable changes observed in the experiment involve Acidobacteria and Proteobacteria where increases in the number of sequences representing these phyla were observed over the course of the experiment (Figure 2). Acidobacteria is a complex phylum with few known isolates and is divided into 26 subgroups based on 16S rRNA sequence analysis 14,15 . They were consistently higher in S U throughout the incubation period with the population peaking at week 8 when 47% of total microbial sequences generated from S U were attributed to this phylum. Within the phylum, 18 orders were observed over the course of the experiment in both S U and S SA with order iii1-15 and RB 41, group 5 and group 3 subdivisions respectively, dominating. The relative abundance of Acidobacteria is strongly correlated with pH, particularly within a number of sub groups including group 3 that has a negative relationship 14 . While the abundance of Acidobacteria increased in both soils (S U S SA ), the highest increase occurred in S U where the pH was stable at 7. These microorganisms are known to have slow metabolic rates that allow them to survive in low nutrient environments 16 . The unamended soil (S U ) had no new source of nutrients over the 12 week experiment and therefore is relatively nutrient depleted compared to S SA soil. Thiobacillus spp flourished in S SA but not in S U where they remain below detection limits from week 8 onwards. The addition of sulfur has stimulated an expansion of the Thiobacillus spp. The Genera Thiobacillus is a diverse one with species capable of growth at low pH 17 , under aerobic and anaerobic conditions 18,19 , with the capacity to grow heterotrophically 20,21 , chemoautrophically 22 , 23 and the capability of oxidising elemental sulfur 24,25 and iron 26 , they are therefore also considered as mixotrophs. The Thiobacillus genera are flexible and their elevated presence within S SA suggests that they are oxidizing S 0 and/or growing via a chemoautotrophic lifestyle. Less dominant phyla present in both S U and S SA throughout the incubation period include candidate division WS3, recently designated Latescibacteria 27 , Nitrospirae and Planctomycetes detected at low levels, varying from 0.5% -5% of the total microbial population. The phylum Latescibacteria has no isolated representatives 28 but sequences sharing homology with this phylum have been detected in varying environments including marine sediments obtained at depths >400 m 29 , surface sediments of fresh water lakes 30 and hydrocarbon contaminated aquifers 31 . One representative of the latescibacteria phylum has recently been shown to contain the RubisCO gene 27 . The Phylum Nitrospirae is also made up of genera with the capacity to grow under chemoautotrophic conditions 32 .

DNA labelling provides insights into carbon fixing populations
The species identified within the most abundant genus (Thiobacillus spp) were: Thiobacillus thioparus, Thiobacillus denitrificans, Thiobacillus thiophillus and Thiobacillus k6. Thiobacillus species have been identified and isolated from agricultural soils in diverse locations across the world 33 , and their importance is based on the fact that Sº must be oxidised to SO 4 -2 to aid plant growth. Thiobacillus denitrificans is interesting as not only is it one of the first bacteria to be linked to sulfur-based chemoautotrophy 34 , but unlike others of the Thiobacillus genera who perform aerobic respiration, T. denitrificans can use nitrate as the electron acceptor under anaerobic conditions if necessary 35 . The remaining 37.3% of isotopically enriched sequences could not be identified beyond the family level suggesting