Introduction

Elemental S is used as a S fertiliser but is not readily available for plant uptake as it has to be first oxidised to sulphate. For agronomic purposes, elemental S is generally not applied as a powder because of safety and practical considerations (Rothbaum and Groom 1961) but is often co-granulated with a macronutrient carrier such as ammonium phosphate or urea. The chemical environment around a fertiliser granule in soil can be very different to that in the bulk soil. For instance, the pH around the granule can be lower due to acidic compounds or compounds that undergo acid-forming reactions, such as oxidation of ammonium (nitrification) or oxidation of elemental S. Furthermore, the soluble compounds within the granule result in high salt concentrations in and near the granule (Sample et al. 1980).

Soil pH has a strong effect on bacterial diversity and community composition, as has been seen both for soils where pH was altered due to long-term management and for soils with natural pH variations across terrestrial ecosystems (Rousk et al. 2010; Fierer et al. 2012). Soil pH has also been reported to be positively related to bacterial abundance (Rousk et al. 2010). The effect of soil pH on microbial abundance may be due to the effect of pH on the chemical form and solution concentration, and hence availability, of substrates (Kemmitt et al. 2006). It has been found that elemental S oxidation is positively related with soil pH (Lawrence and Germida 1988; Zhao et al. 2015), which is likely because pH could affect both the abundance and activity of S oxidisers in soil (Zhao et al. 2016a, 2016b).

Changes in ionic strength in soil solution may also affect microbial activity or growth, e.g. due to osmotic stress (Galinski and Trüper 1994). In a culture-based study assessing the effect of pH and salts on S oxidation by Thiobacillus thiooxidans, it was found that low salt concentrations increased elemental S oxidation likely because of charge neutralisation of the cell surface, but high concentrations (∼0.2 M) inhibited elemental S oxidation, which was attributed to high osmotic pressures (Suzuki et al. 1999). In addition, high ionic concentration was also found to encourage the adherence of bacteria to solid surfaces by decreasing the thickness of electrical membrane layers (Tan et al. 1991, 1992).

We hypothesised that the lowered pH and/or the elevated ionic strength around a fertiliser granule may affect elemental S oxidation through effects on the microbial abundance and activity. To test this hypothesis, we modified soil pH and ionic strength to different levels to reflect the changes occurring around a macronutrient fertiliser granule. The objectives of this study were to (1) investigate the effects of acidification or an increase in ionic strength on elemental S oxidation, (2) examine the changes of bacterial abundance and bacterial community composition under these conditions and (3) examine if changes in elemental S oxidation as a result of chemical modifications can be related to shifts in bacterial abundance in soil.

Material and methods

Experiment 1: effects of acidification or increased ionic strength on elemental S oxidation in SA soil

A sandy soil from South Australia (SA soil) was used for this experiment. The soil properties are listed in Table 1(for details of soil analysis, see Zhao et al. 2015).

Table 1 Physico-chemical properties of three soils from South Australia (SA), Victoria (VIC) and New South Wales (NSW)
Full size table

To assess the effect of acidification, air-dried soil in four containers (200 g each) was acidified with 0, 4, 7 or 10 mL 1 M HNO3, respectively. The acidified soils were equilibrated overnight and leached to remove excess salt. To that end, the soil of each treatment (200 g in dry weight) was transferred to a pot sealed at the bottom with nylon cloth and 400 mL of artificial rainwater was added on the top, after which the excess water was left to drain through the soil. The artificial rainwater consisted of CaCl2 (0.015 mM), KNO3 (0.008 mM), MgCl2 (0.010 mM), Na2SO4 (0.010 mM), NaCl (0.064 mM) and NH4Cl (0.006 mM) (Donn and Menzies 2005). The leached soil was air-dried for 1 week. Thereafter, the soil pH was determined in a 1:5 (kg L−1) water suspension and was 6.7 (p1 or control), 6.1 (p2), 5.2 (p3) and 4.4 (p4) for the four treatments. The air-dried soil of each treatment was separated into four replicates, and each replicate was rewetted with 5.2 mL of deionised water (corresponding to 70% of field capacity) and amended with elemental S at a rate of 200 mg kg−1 air-dried soil. The elemental S particles had an average diameter of 48 μm (see the distribution of elemental S particle sizes in the study of Zhao et al. 2015).

The ionic strength was changed using KH2PO4 as elemental S is often co-granulated with P fertilisers. The effect of ionic strength was assessed for conditions relevant for granular fertilisers, at 0.1–0.7 M KH2PO4 in soil solution, which corresponded to the range of P concentrations in the soil solution around a diammonium phosphate (DAP) granule (circa 38 mg) at distances of 0–4, 4–10 and 10–15 mm after 1-day incubation, as determined in a preliminary experiment (data not shown). To obtain treatments with 0 (E1 or control), 0.1 (E2), 0.4 (E3) or 0.7 M (E4) KH2PO4, 21 mL of a solution with a KH2PO4 concentration of 0, 14, 54 or 95 g L−1, respectively, was added to four aliquots of 200-g soil and mixed thoroughly. The soil of each treatment was divided into four replicates and amended with 200 mg elemental S kg−1 air-dried soil.

Following the addition of elemental S, the soil from each replicate was evenly distributed into four plastic containers. The soil was incubated at 25 °C, and soil water content was adjusted on a weekly basis. The soils were sampled and analysed (see Sample analysis section) at weeks 2, 4, 8 and 14.

Experiment 2: effect of acidification on elemental S oxidation in VIC and NSW soils

From experiment 1, we found that elemental S oxidation rate was not related to short-term changes in pH or ionic strength or gene abundances (16S ribosomal deoxyribonucleic acid (rRNA), soxB and 18S rRNA genes), which seemed inconsistent with the finding that elemental S oxidation across soils correlated well with soil pH and 16S rRNA or soxB abundance (Zhao et al. 2016b). Therefore, a second experiment was conducted to verify the effect of acidification on elemental S oxidation in two other soils, one from Victoria (VIC soil, a loamy sand soil) and another from New South Wales (NSW soil, a sandy loam soil).

Before incubation, 10 and 7.5 mL of a 1 M HNO3 were applied to 250 g of each soil to bring the soil pH within the range of 4–5, as determined in a preliminary experiment. The soils were then leached with artificial rainwater (see previous section). The control soil without acid amendment was also leached in the same way. The pH was determined in water after the soil was air-dried. The pH in the control soil remained the same as that prior to leaching (Table 1), and the pH in the acidified treatments was 4.4 for the VIC soil and 4.7 for the NSW soil. Elemental S was mixed through the soils at a rate of 200 mg elemental S kg−1 air-dried soil. Water was added to the soils to reach 70% field capacity, and the samples were incubated at 25 °C. Sampling and analysis was carried out at 2, 4 and 6 weeks.

Sample analysis

Analyses of pH, electric conductivity and remaining elemental S

All samples were analysed for pH, electrical conductivity (EC, as an indicator of ionic strength) and concentrations of elemental S. Soil pH and EC were determined at a soil/water ratio of 1:5 kg L−1. Elemental S analysis was carried out by extracting 2.5 g of fresh soil sample with 5 mL of chloroform, and 2 mL of water was also added to disperse soil aggregates effectively (Zhao et al. 2015). The samples were shaken for 2 h before centrifugation at 4600g for 20 min. The clear chloroform supernatant was transferred to a glass vial and diluted with methanol for analysis by high-performance liquid chromatography (HPLC, Agilent, USA) with methanol as the mobile phase. The diluted chloroform supernatant (50 μL) was injected into a reverse-phase HPLC column (C18, 250 by 4.6 mm). The flow rate was 1 mL min−1, and the detection wavelength was set at 254 nm. The retention time for elemental S was 8 min.

Analyses of gene abundances

DNA was extracted from soil samples incubated for 2, 4 and 14 weeks in experiment 1 or for 2, 4 and 6 weeks in experiment 2. The amounts of 16S rRNA, soxB and 18S rRNA genes were quantified and used as indicators of the abundance of total bacteria, S-oxidising bacteria and soil fungi, respectively. Quantitative polymerase chain reaction (qPCR) analysis was performed on a Stratagene Mx3000P system (Stratagene, Australia). soxB gene abundance was determined using soxB 432F/693B primers (Petri et al. 2001). The reaction mixtures consisted of 1× iQTM SYBR® Green Supermix (Bio-Rad, USA), 600 nM of each primer and 2 μL of DNA templates. Quantitative PCR was run at 95 °C for 15 min and 40 cycles at 95 °C for 40 s, 54 °C for 40 s and 72 °C for 40 s. A melting curve was generated between 60 and 95 °C with readings every 0.5 °C. The abundance of 16S rRNA genes was determined according to Gupta et al. (2014). The abundance of 18S rRNA genes was quantified using FR1/FF390 primers. The PCR amplification consisted of 40 cycles at 95 °C for 15 s, at 50 °C for 30 s and at 72 °C for 1 min (Prevost-Boure et al. 2011).

Analysis of TRFLP of total bacterial communities

Terminal restriction fragment length polymorphism (TRFLP) was performed on soils sampled after weeks 2 and 14 of experiment 1 to assess the shift in bacterial community composition in acidified soils and in soils with increased ionic strength. The analysis of TRFLP was conducted with 16S primers: 8F-FAM (5′-AGAGTTTGATCCTGGCTCAG-3′) and 1522R-HEX (5′-AAGGAGGTGATCCAGCCGCA-3′) (Edwards et al. 1989) to target 16S rRNA in soil bacteria. Polymerase chain reaction (PCR) procedures were run in analytical duplicates. Single PCR reaction mixtures contained 1× PCR buffer, 0.4 mM of MgCl2, 1 unit of HotStarTaq DNA polymerase (Qiagen, Australia), 400 nM of each primer, 0.2 mM of dNTPs (Promega, Australia) and 5 μL of DNA templates. The PCR conditions consisted of enzyme activation at 95 °C for 15 min, 35 cycles of denaturation at 95 °C for 30 s, primer annealing at 55 °C for 1 min and elongation at 72 °C for 2 min, and final extension at 72 °C for 10 min. The size of PCR products was examined by gel electrophoresis on a gel (0.8% w/v) stained with GelRed Nucleic Acid Stain (Biotium, Australia) and visualised by transilluminator (Bio-Rad, USA). Amplicons from analytical duplicates were mixed and cleaned with a MinElute 96 UF PCR Purification Kit (Qiagen) according to the manufacturer’s protocol. The concentrations of purified DNA were determined using a Quant-iT PicoGreen™ dsDNA Assay Kit (Invitrogen, USA) on a Stratagene Mx3000P. Approximately 150 ng of PCR products was digested with restriction enzymes AluI and CfoI (Promega) in separate reactions. The digestion reaction was performed at 37 °C for 3 h, followed by denaturation at 65 °C for 20 min. The digested PCR products were analysed for size by the Australian Genome Research Facility (Adelaide, SA) using capillary separation on an ABI 3730 DNA analyser with a LIZ500 standard.

Data analysis

Analysis of variance was performed with SPSS (IBM, version 20) followed by Duncan’s multiple range test to determine the differences in log-transformed gene abundance, percentage of elemental S oxidised and Shannon index (H′) between four treatments or over four incubation times for experiment 1. With regard to experiment 2, independent sample test (SPSS, IBM, version 20) was carried out to determine the difference in log-transformed gene abundance and percentage of elemental S oxidised between the control and the acidified treatment.

The TRFLP fragments were analysed using the GeneMarker software (SoftGenetics Inc., State College, PA, USA) to obtain band intensity (peak height) data. Bands with intensities <100 were excluded to distinguish restriction fragments from background noise. The band intensities (heights) were imported to Microsoft Excel 2010 for the calculation of relative intensity, which was expressed as the intensity of each terminal restriction fragment (TRF) divided by the total intensities of TRFs in the profile. Bands with relative intensities ≥0.5% were included for diversity index calculation and multivariate statistical analysis using the Primer 6 software package (Primer-E Ltd., Plymouth, UK). The Shannon index (H′) (Shannon and Weaver 1948) was calculated to estimate bacterial diversity. Similarity between bacterial communities was analysed using the Bray-Curtis algorithm based on square-root-transformed abundance data. Permutation-based multivariate analysis of variance (PERMANOVA) (Anderson 2001) was performed to test if acidification or increases in ionic strength or incubation time had any effect on bacterial community composition as estimated by the square root of components of variation (√CV). Multivariate distances between samples were visualised by generating non-metric multidimensional scaling (nMDS) plots. Linkage between bacterial community composition and environmental variables was examined by the BIO-ENV test (Clarke and Ainsworth 1993).

Results

Experiment 1—SA soil

Elemental S oxidation and changes in pH and EC over time

For the acidification series, the pH prior to incubation ranged from 6.7 to 4.4. In the acidified treatment with the lowest acid rate (p2), there was no effect of acidification on the oxidation rate (Fig. 1a). For the most acidified treatment (p4), there was a small significant difference with the control treatment (p1) at week 4, but not at the other sampling occasions, indicating that the effect of acidification was small and transient. However, for the intermediate acid rate (p3), there was a more pronounced and consistent decrease in oxidation rate (8–18% reduction). The reason why the effect of acidification was more pronounced in treatment p3 than in p4 is unclear. Overall, approximately 40% of elemental S was oxidised at week 2 and 90% at week 14 (Fig. 1a). By the end of the incubation period, soil pH decreased by 0.1–1.2 units as compared to the initial pH (Table 2). Soil EC increased over the incubation period but was still ≤0.28 mS cm−1 in all treatments by week 14 (Table 2).

Fig. 1
figure 1

Percentage of elemental S oxidised at different incubation times in soil treatments (a) adjusted to different pH values with HNO3 (pH 6.7 (p1), 6.1 (p2), 5.2 (p3) and 4.4 (p4)) or (b) increased ionic strength with KH2PO4 by 0 M (E1), 0.1 M (E2), 0.4 M (E3) and 0.7 M (E4). Different letters indicate significant differences between treatments at each sampling time (P ≤ 0.05). Error bars indicate the standard deviation (n = 4)

Full size image
Table 2 pH and electrical conductivity (EC) at different incubation times in soils acidified with HNO3 or added with KH2PO4 to increase ionic strength
Full size table

In the treatments in which ionic strength was varied by KH2PO4 addition, no treatment effect on elemental S oxidation was observed, with about 50% of the elemental S oxidised at week 2 and 95% at week 14 (Fig. 1b). The EC in the control treatment (E1) strongly increased in time due to the elemental S oxidation, but the ECs in the soils treated with KH2PO4 (E2–E4) showed relatively less change because of the higher starting values of EC in these soils (Table 2). Compared to the initial values, the soil pH dropped by 1.2 units in the control treatment (E1) after 14-week incubation, but only by 0.3–0.6 units in the KH2PO4-treated soils (E2–E4) because of the additional pH buffering by phosphate (Table 2).

Abundances of 16S rRNA, soxB and 18S rRNA genes

The abundance of 16S rRNA genes was higher in the p2 treatment than in the control treatment at weeks 2 and 4, but lower in the more acidified treatments (p3, p4) at all sampling times (Fig. 2a). The abundance of soxB genes was lower in all acidified treatments (p2–p4) than in the control treatment (p1) (Fig. 2b). Over time, the soxB gene abundance showed an increase between weeks 2 and 4 in the acidified treatments (p2–p4), followed by a slight decrease at week 14 in the p2 and p4 treatments (Fig. 2b). The abundance of 18S rRNA genes was higher in the most acidified treatment (p4) but not different for the other two treatments (p2, p3) as compared to the control treatment (p1) (Fig. 2c).

Fig. 2
figure 2

Abundances of 16S rRNA, soxB and 18S rRNA genes at different incubation times in soils treatments (A) adjusted to different pH values with HNO3 (pH 6.7 (p1), 6.1 (p2), 5.2 (p3) and 4.4 (p4)) or (B) increased ionic strength with KH2PO4 by 0 M (E1), 0.1 M (E2), 0.4 M (E3) and 0.7 M (E4). Different small letters indicate significant differences between treatments at each sampling time, and different capital letters indicate significant differences between sampling times for each treatment (P ≤ 0.05). Error bars indicate the standard deviation (n = 4)

Full size image

Increased ionic strength and incubation time had no, or minor, effects on the 16S rRNA gene abundance (Fig. 2d), while the increase in ionic strength significantly reduced soxB gene abundances compared to the control (Fig. 2e), though the changes were smaller than those induced by acidification. Over time, soxB gene abundance showed a significant increase at week 4, followed by a significant decrease at week 14 for all treatments (Fig. 2e). Increases in ionic strength had no effect on, or increased, 18S rRNA gene abundances, except for a slight decrease in E2 at week 2 compared to the control (E1) (Fig. 2f).

Bacterial community composition

The Shannon index (H′) indicated that acidification significantly decreased bacterial TRF diversity in all treatments (P ≤ 0.05) (Table 3). Over time, the bacterial TRF diversity increased in the control (p1) and p3 treatments but showed no significant change in the other treatments (p2, p4) (Table 3). The PERMANOVA analysis showed that acidification and incubation significantly altered the composition of the bacterial community (PERMANOVA √CV = 36.7, P = 0.001 for acidification, and √CV = 15.0, P = 0.001 for the incubation). The results from nMDS ordination corroborated that obtained from PERMANOVA analysis. The nMDS plot showed clear shifts in bacterial community composition caused by acidification (Fig. 3a). In the most acidified treatment (p4), there was also an evident shift in the composition over time (week 2 versus 14), but not in the other treatments (Fig. 3a). BIO-ENV analysis revealed a close link between the shifts in bacterial community composition and soil variables, with pH and EC the main contributors (BIO-ENV test ρ = 0.571; P = 0.01).

Table 3 Shannon index (H′) of bacterial TRF diversity at two incubation times in soils treated with or without HNO3 to generate a pH gradient of 6.7 (p1), 6.1 (p2), 5.2 (p3) and 4.4 (p4) or with or without KH2PO4 to increase ionic strength by 0 M (E1), 0.1 M (E2), 0.4 M (E3) and 0.7 M (E4)
Full size table
Fig. 3
figure 3

Non-metric multidimensional scaling (nMDS) plots showing variation in bacterial community composition in elemental S-amended soils incubated for 2 and 14 weeks. The soil was adjusted to (a) different pH values with HNO3 (pH 6.7 (p1), 6.1 (p2), 5.2 (p3) and 4.4 (p4)) or (b) increased ionic strength with KH2PO4 by 0 M (E1), 0.1 M (E2), 0.4 M (E3) and 0.7 M (E4). We manually grouped the points resulted from the two sampling times for each treatment. Increasing distance between points reflects increasing dissimilarities in bacterial community composition

Full size image

In the soils with increased ionic strength, bacterial TRF diversity was decreased compared to the control treatment (E1), but the adverse effect was weaker at week 14 than week 2 (Table 3). Incubation increased TRF diversity in the control (E1) and E3 treatments over time (Table 3). The PERMANOVA analysis showed that increases in ionic strength and incubation altered the bacterial community composition significantly (PERMANOVA √CV = 28.7, P = 0.001 for increases in ionic strength, and √CV = 22.8, P = 0.001 for the incubation). The results from nMDS ordination corroborated those obtained from PERMANOVA analysis. The nMDS plots suggested clear separation of the bacterial communities between treatments or at incubation times (Fig. 3b). A close link was observed between the shifts in bacterial community composition and soil pH and EC (BIO-ENV test ρ = 0.562; P = 0.01).

Overall, acidification and increased ionic strength decreased the total and S-oxidising bacterial abundances and changed the bacterial community composition. However, acidification and increased ionic strength had no or little effect on the oxidation rate of elemental S. These results suggest that elemental S oxidation is independent of short-term changes in the abundance and composition of bacterial population in soil. The lack of relationship between elemental S oxidation and bacterial population was unexpected given the results of our previous study, in which we found a strong positive relationship between elemental S oxidation and soil pH (Zhao et al. 2015) or bacterial abundance (Zhao et al. 2016b). To confirm the results from this experiment, a second experiment was conducted using two other soils with different physico-chemical properties.

Experiment 2—VIC and NSW soils

Elemental S oxidation over time

After 6-week incubation, approximately 60% of elemental S was oxidised in both the VIC and NSW soils (Fig. 4a). Acidification did not show any effect on elemental S oxidation over the incubation period in the NSW soil (Fig. 4a). In the VIC soil, no difference in elemental S oxidation was observed between the control and the acidified treatments at weeks 2 and 4, but a small decrease was found at week 6 (12%; P ≤ 0.01) (Fig. 4a).

Fig. 4
figure 4

(a) The percentage of elemental S oxidised, (b) 16S rRNA gene and (c) soxB gene abundances in the VIC and NSW soils for the control and acidified treatments. One asterisk indicates a significant difference at P ≤ 0.05, and two indicate a significant difference at P ≤ 0.01. Error bars indicate the standard deviation (n = 4)

Full size image

Abundances of 16S rRNA and soxB genes

Acidification decreased the 16S rRNA gene abundance in the VIC soil at all sampling times but slightly increased it in the NSW soil at weeks 2 and 6 (Fig. 4b). The soxB gene abundance in the acidified treatments was also lower than that in the control treatment in the VIC soil at all sampling times, while no treatment effects on soxB gene abundance were observed for the NSW soil (Fig. 4c).

Discussion

Elemental S oxidation in soil has been reported to be related to soil pH and the abundance and diversity of bacterial oxidisers (Li and Caldwell 1966; Nor and Tabatabai 1977; Janzen and Bettany 1987; Lawrence and Germida 1988; Tourna et al. 2014; Zhao et al. 2015). In our previous study, we found that pH and bacterial abundance explained 80% of the variation in the oxidation rate of elemental S across soils (Zhao et al. 2016b). In the present study, we modified soil pH by acid addition and did not find a clear relationship between pH and bacterial abundance on one hand and elemental S oxidation rate on the other. To reconcile these seemingly contradictory results regarding the relationship between bacterial abundance and elemental S oxidation, the differences in experimental design between this and previous studies need to be considered. Our previous study and other literature studies that found a positive relationship used different soils with different pH values, in which case pH is associated with many other soil physico-chemical properties, vegetation differences and variations in the composition of soil microbial communities. In the present study, however, we used one soil that was modified to different pH and ionic strength levels and these modifications took place 1 week before or at the commencement of the incubation.

A decrease in pH had no or little effect on the elemental S oxidation rate in the present study, while increases in pH due to the amendment with CaCO3 (Adamczyk-Winiarska et al. 1975) or phosphate rock (Lee et al. 1987) have been proved to increase elemental S oxidation rate. The lack of agreement between our study and the previous studies may be due to the difference in elemental S application rate, which was 200 mg kg−1 in our study compared to 5000–10,000 mg kg−1 in these literature studies. The high application rates of elemental S in these two literature studies are hypothesised to cause severe acidification and high sulphate concentration in the control soil, which may have decreased the activity of enzymes involved in elemental S oxidation. The addition of CaCO3 (Adamczyk-Winiarska et al. 1975) or of phosphate rock (Lee et al. 1987) would have prevented this severe acidification, explaining the higher elemental S oxidation in the amended soils. We also observed a slight decrease in elemental S oxidation in the acidified VIC soil at week 6 (Fig. 4a) and some of the acidified treatments of the SA soil (Fig. 1a), but not in the NSW soil. This may be related to the lower buffering capacity of the VIC soil (38.5 mmol (H+) kg−1 ΔpH−1) and the SA soil (26.9 mmol (H+) kg−1 ΔpH−1) than that of the NSW soil (50.5 mmol (H+) kg−1 ΔpH−1). Thus, the effect of changes in pH on elemental S oxidation is likely soil-type dependent. When used as a fertiliser source, the elemental S rates (circa 30 mg kg−1 soil, depending on the S requirement of crops) are much lower than in most laboratory studies, and acidification is not likely to cause significant effects in the short term. However, the acidifying effect of elemental S cannot be neglected in the long term and may need to be offset by lime application.

Abundances of a specific functional group have been used to predict the rate of a biochemical process as they can provide an integrated measurement of the genetic potential of a soil reflecting the physico-chemical environment (Petersen et al. 2012). This was reflected in our previous study using multiple soils with varying characteristics, where pH and bacterial abundances were effective predictors of elemental S oxidation (Zhao et al. 2016b). The lack of agreement between elemental S oxidation and bacterial abundance in the present study indicates that elemental S oxidation was not necessarily limited by the size of the microbial population. Also, other studies have proved that microbial abundance is generally not limiting the elemental S oxidation rate except at very high application rates (e.g. >5000 mg elemental S kg−1) (Li and Caldwell 1966; Nor and Tabatabai 1977; Lettl et al. 1981; Janzen and Bettany 1987). For instance, in a short-term (6 days) incubation experiment using Canadian soils, Janzen and Bettany (1987) reported that the oxidation rate of elemental S was independent of its application rate ranging from 0 to 4000 mg elemental S kg−1. Similar results have also been observed over longer time incubations (up to 60 days) in other studies (Li and Caldwell 1966; Nor and Tabatabai 1977; Lettl et al. 1981; Janzen and Bettany 1987).

The present study not only showed a change in bacterial abundance, but also a shift in total bacterial community composition due to increases in acidity or ionic strength and time of incubation. They corroborate previous reports showing a significant influence of pH and ionic strength on the diversity and composition of bacterial community in soil environments (Rousk et al. 2010; Ibekwe et al. 2010). However, the changes in bacterial community composition had no or little effect on the elemental S oxidation, indicating that the S-oxidising bacterial community in these soils contains functionally redundant taxa that responded to changing conditions. Specifically, the new community of S-oxidising bacteria contains the taxa that may be functionally redundant in the original community, hence the oxidation process rate, was not altered (Allison and Martiny 2008). This is supported by the broad taxonomic and phylogenetic diversity of S-oxidising bacteria in soils (Anandham et al. 2008; Tourna et al. 2014; Xia et al. 2014; Zhao et al. 2016b). Functional redundancy of microbial community is likely common in soils (Nannipieri et al. 2003; Franklin and Mills 2006; Wertz et al. 2007; Allison and Martiny 2008; Ollivier et al. 2012; Lehman et al. 2015). For example, Ollivier et al. (2012) found functional redundancy of ammonia-oxidising bacteria in heavy metal-contaminated soils, and Wertz et al. (2007) observed a high level of functional redundancy for denitrifiers and nitrite oxidisers in a dilution-led disturbance experiment. Similarly, Marschner et al. (2003) investigated the effect of long-term application of organic and inorganic fertilisers on microbial communities in soil and found changes in the composition of bacterial community, but not in urease, xylanase, alkaline phosphatase and arylsulphatase activities.

A limitation of the present study is that we only tested the changes of overall bacterial communities in soil, which may not entirely reflect the shifts of S-oxidising bacterial community composition as the population of S-oxidising bacteria only represented a small proportion (<3%, data not shown) of the overall populations in soil. The fact that elemental S oxidation rate was not or only slightly affected, despite the evident shifts in bacterial community composition due to changes in pH and EC, may be related to the broad distribution of S-oxidising bacteria in terms of physiology and taxonomy in soil (Anandham et al. 2008; Tourna et al. 2014; Xia et al. 2014; Zhao et al. 2016b). It is assumed that the role of some S-oxidising bacteria at one pH or ionic strength level had been fulfilled by other S-oxidising bacteria adapted to another pH or ionic strength level.

Apart from soil bacteria, fungi are also thought to play a role in elemental S oxidation in soil (Grayston and Wainwright 1988). We found higher abundance of 18S rRNA genes in the most acidified treatment (p4) and the treatment with the highest ionic strength (E4) (Fig. 2c). However, gene abundances, and changes therein, do not necessarily reflect the contribution to a microbial process. Overall, the role of fungi in elemental S oxidation in soil is yet to be fully understood.

It is well known that elemental S in granular fertiliser is slower to oxidise than powdered elemental S mixed through soil. Our results indicate that chemical changes around a fertiliser granule resulting from elemental S oxidation or from dissolution of soluble macronutrients, under the experimental conditions described here, may not have any effect on elemental S oxidation. Indeed, we found that for DAP fertiliser co-granulated with elemental S, none of the changes in chemical properties or bacterial populations could explain the adverse effects of granulation on elemental S oxidation. Instead, the decrease in oxidation of the co-granulated elemental S could be explained in terms of a reduction in surface area in contact with soil microorganisms (Zhao et al. 2016a).

Overall, the present study suggests that elemental S oxidation is not affected by short-term changes in bacterial abundance and community composition brought about by temporary increases in soil acidity or ionic strength. It is hypothesised that the point at which the acidity or ionic strength will start to disturb the functional capacity of S-oxidising microorganisms, and consequently the elemental S oxidation rate, likely depends on elemental S rate applied, soil type (physico-chemical properties), habitat conditions and the time of exposure to the altered conditions.