Arsenic Methylation and its Relationship to Abundance and Diversity of arsM Genes in Composting Manure

Although methylation is regarded as one of the main detoxification pathways for arsenic (As), current knowledge about this process during manure composting remains limited. In this study, two pilot-scale compost piles were established to treat manure contaminated with As. An overall accumulation of methylated As occurred during 60 day-composting time. The concentration of monomethylarsonic acid (MMA) increased from 6 to 190 μg kg−1 within 15 days and decreased to 35 μg kg−1 at the end of the maturing phase; while the concentration of dimethylarsinic acid (DMA) continuously increased from 33 to 595 μg kg−1 over the composting time. The arsM gene copies increased gradually from 0.08 × 109 to 6.82 × 109 copies g−1 dry mass over time and correlated positively to the concentrations of methylated As. 16S rRNA gene sequencing and arsM clone library analysis confirmed the high abundance and diversity of arsM genes. Many of these genes were related to those from known As-methylating microbes, including Streptomyces sp., Amycolatopsis mediterranei and Sphaerobacter thermophiles. These results demonstrated that As methylation during manure composting is significant and, for the first time, established a linkage between As biomethylation and the abundance and diversity of the arsM functional genes in composting manure.

and tetramethylarsonium, have been reported in rice grains which in trace amounts originated from soils [14][15][16] . Methylated As species, both volatile (e.g. TMA and TMAO) and non-volatile (e.g. MMA and DMA), are less toxic than their inorganic forms 17 . Therefore, methylation of As is normally regarded as one of the main detoxification pathways for As in the environment 18 .
Many studies that explore As methylation during composting have been published recently 18 , driven by the incentive to detoxify As in wastes. Diaz-Bone et al. reported that metal(loid)s could undergo intensive biomethylation during the composting of As hyperaccumulating fern Pteris vittata 19 . Maňáková et al. observed a slight decrease in DMA and MMA contents during the composting of waste sludge 20 . However, the microbial dynamics of As methylation and resulting speciation during composting of pig manure remains to be explored.
The stepwise microbial methylation of inorganic arsenite is catalyzed into its methylated counterparts (e.g. [As(III)]→[MMA]→[DMA]→[TMA]) by S-adenosylmethionine methyltransferase encoded by arsM genes 18 . Since its first isolation from soil bacterium Rhodopseudomonas palustris 18 , arsM has been identified in many other microbes, including Pseudomonas spp. 21 , methanogens 22 , Halobacterium sp. NRC-1 23 and a number of eukaryotic algae 24 . Recently, arsM was identified even more in phylogenetically-diverse microbial communities including Actinobacteria, Gemmatimonadales, α-Proteobacterales, β-Proteobacterales, δ-Proteobacterales, Firmicutes, Archaea, and other organisms residing in rice rhizosphere soil and roots 25 . The function of the proteins encoded by the arsM genes was first demonstrated in an As-hypersensitive strain of Escherichia coli, where recombinant expression of an arsM gene conferred As resistance 26 . Similarly, transgenic rice expressing an arsM gene from R. palustris was shown to methylate inorganic As into a variety of organic As species 27 . Although the mechanism of microbial As methylation is known and arsM genes have been detected in various environments, there remains a limited understanding of how the abundance and diversity of arsM genes correlate with the methylation process during manure composting.
In this study, two pilot-scale pig manure composting piles were constructed for a systematic investigation of As methylation. Physical and chemical parameters of the piles were monitored during composting process. Microbial community composition and abundance, as well as the abundance and diversity of arsM genes were monitored using real-time PCR (qPCR) and amplicon sequencing. The results have yielded the first clear connection between As methylation and microbial arsM gene abundance in manure composting, providing valuable insights for developing strategic management of manure waste.

Results
Physical-chemical properties and total As changes during composting. Two pilot-scale pig manure compost piles (MC1 and MC2) were established and monitored over 60 days. Based on the measured temperature profile, both compost piles progressed through mesophilic (day 0~4), thermophilic (day 5~42), and maturing phases (day 43~60) as defined in other composting processes 28 . A rapid increase in temperature from ~31 °C to ~60 °C within the first 4 days was observed (Supplementary Figure S1), and this high temperature was maintained throughout the thermophilic phase for approximately 42 days before dropping to ~40 °C in the subsequent maturing phase. Over the composting period, samples were collected to track their activity. The organic matter (OM) content decreased gradually from ~60% to ~41% (g VS/g dry mass) after 60 days, and the moisture content decreased from ~64% to ~43% in the composting piles (Supplementary Table S1). In the first 15 days, the pH decreased from 6.9 to 3.5 and then gradually increased back up to around 6.0. Bulk oxidation-reduction potential (Eh) decreased from 174 mV to 60-65 mV after 60 days and NH 4 + -N concentrations also decreased substantially as the composting proceeded (Supplementary Table S1). The total As content on a per kg dry solids basis increased from 1,270 ± 62 μg kg −1 to 1,800 ± 68 μg kg −1 in MC1 and from 1,240 ± 67 μg kg −1 to 1,720 ± 180 μg kg −1 in MC2 over time, amounting to a ~1.4 fold increase in concentration in both compost piles by day 60 (Supplementary Figure S2).
The change of As species during composting. The concentrations of two major methylated As species, MMA and DMA, were shown in Fig. 1. MMA content in both compost piles peaked during the thermophilic phase (day 5-42) and dropped rapidly during the maturing phase (day 43-60). The MMA concentrations in MC1 peaked at 175 ± 21 μg kg −1 at the middle of the thermophilic phase (day 25) and dropped subsequently rapidly to 43 ± 3 μg kg −1 in the maturing phase (day 60). The MMA concentrations in MC2 exhibited a similar trend, peaking at 222 ± 14 μg kg −1 on day 15, and subsequently dropping from day 15 onwards to a final concentration of 28 ± 4 μg kg −1 on day 60. In contrast, DMA concentrations increased steadily in both compost piles over the composting period (Fig. 1), from 30 ± 4 to 620 ± 10 μg kg −1 in MC1 and from 37 ± 3 to 570 ± 20 μg kg −1 in MC2. The total concentrations of methylated As species increased more rapidly during the mesophilic and thermophilic phases, while only small increases during the maturing phase. Moreover, there were no significant differences in the concentrations of methylated As species at most sampling times between the two composting piles (MC1 and MC2) as determined using a two-way ANOVA test (Supplementary Table S2).
Copy numbers of bacterial 16S rRNA and arsM genes. The copy numbers of the bacterial 16S rRNA gene (copies g −1 dry mass) in the compost piles during the experimental period is shown in Fig. 2a. The abundance of 16S rRNA genes decreased from ~3 × 10 11 to 0.8 × 10 11 copies g −1 dry mass during the thermophilic phase. The abundance of arsM genes in MC1 and MC2 as a function of composting time increased gradually from ~0.1 × 10 9 to ~6.8 × 10 9 copies g −1 dry mass in both piles (Fig. 2b). There were also no significant differences in arsM gene copies between MC1 and MC2 except for those samples on day 25 and 35 (Supplementary Table S2). Significantly, a steady and significant increase in the ratio of arsM genes to 16S rRNA gene numbers was observed. The highest value (~6% on day 45, MC1; ~8% on day 60, MC2) occurred during the maturing phase (Fig. 2c). Further, the sum of MMA and DMA concentrations at different time points was found to correlate strongly with the arsM gene copy numbers (Fig. 3).
Dynamics in microbial community structure. The V6-V8 region 16S rRNA gene was amplified and sequenced, returning an average of ~23,000 reads per sample. All samples showed relatively high coverage (0.74-0.89) and high diversity as indicated by Shannon, ACE, and Chao1 indices (Supplementary Table S3).  A diverse set of OTUs from 18 orders from 14 phyla were detected in MC1 and MC2 with major shifts over the composting period (Supplementary Figure S3). The relative abundance of Firmicutes decreased markedly while significant increases in the proportions of Actinobacteria, Proteobacteria and Bacteroidetes were observed (Supplementary Figure S3). At the beginning of the composting period, Firmicutes accounted for ~93% of the total population; by the end, they accounted for only 46.8% (MC1) and 41.2% (MC2). As shown in Supplementary Figure S3, the relative abundance of Actinobacteria, Proteobacteria and Bacteroidetes at the end were 18.7%, 9.4% and 16.0% in MC1, and 30.0%, 9.6% and 4.5% in MC2, respectively. The relative abundance of Chloroflexi increased from 0.04% to 1.4% in MC1 and 0.01% to 3.2% in MC2 over the composting process, respectively. Considering the OTU sequences clustered at the order level, the samples from MC1 and MC2 shared similar profiles over time (Supplementary Figure S3). At the beginning of the composting, Clostridiales, Lactobacillales and Erysipelotrichales were dominant, accounting for ~90% of the total population. Halanaerobiales and Bacillales significantly increased within the first 15 days, and then decreased markedly, while Flavobacteriales and Burkholderiales increased in the maturing phase. Pseudonocardiales and Sphaerobacterales were not dominant in the initial microbial community structure (0.25% and 0.65%), however, they both increased during the thermophilic and maturing phases and increased to ~1-2% of the population by day 60, respectively (Supplementary Figure S3).
In order to gain more specific knowledge about organisms responsible for As methylation, we aligned the 16S rRNA sequences from the compost piles against 16S rRNA sequences from microbes containing an arsM gene (Supplementary Table S5). Eighty-three OTUs from the compost piles match to previously report hosts of arsM  . Changes and taxa of selected 16S OTUs related to As methylation. A custom database of microorganisms contained arsM genes was constructed by compiling all 16S rRNA gene sequences from NCBI. All 16S rRNA gene sequences of samples were checked against the database, and sequences that have a similarity ≥95% and Read depth ≥10 were retained. Changes of these OTUs with composting time are shown. Note abundance is in per mil, not percent. Maximum abundance is about 6%, which is similar to the max ratio of arsM to 16S rRNA genes recovered (see Fig. 2c). Refer to Supplementary Table S4 for specific OTUs per group. genes (similarity ≥ 95%) and their relative abundance clearly increased with composting time (Fig. 4). At the beginning of composting, these matching OTUs were primarily Bacillales within the Firmicutes and made up only 0.8 to 1.3% of the total population (note: Y-axis in Fig. 4 is in per mil, not percent). Bacillales, Hydrogenophilales and Chromatiales all increased in relative abundance over time. Pseudonocardiales and Corynebacteriales were also relatively highly represented in the samples. Sphaerobacterales (Chloroflexi) increased to 0.4% (MC1) and 0.9% (MC2) on day 60. Streptomycetales could only be detected on day 60 (0.15% in MC1 and 0.01% in MC2).
The abundance and biodiversity of arsM genes. The arsM genes encoding S-adenosylmethionine methyltransferase were amplified from DNA samples collected on day 15 and day 60 using previously published primers 25 targeting most known arsM genes. Resulting clone libraries were sequenced to assess the diversity of these genes. Sequences were clustered and results were visualized in a heat map (Supplementary Figure S4). This sequencing data confirmed that a high abundance and diversity of arsM genes could be found in compost piles. The most abundance partial sequence clones (PSCs) were PSC001, accounting for 58.3% of total clones. Besides, PSC002 (7.6%), PSC003 (1.2%), PSC005 (9.7%), PSC006 (8.3%), PSC008 (2.1%), PSC010 (6.7%), PSC013 (1.4%) were also abundant. PSC001, PSC008 and PSC013 were most abundant in the samples collected on day 15, while PSC005, PSC002, PSC010 and PSC003 were most abundant in those samples collected on day 60. PSC006 was abundant across all samples. All PSCs were further compared with the database of referenced arsM gene sequences from NCBI (Supplementary Table S5). A Neighbor-joining tree of the 8 most abundant arsM PSCs (>1% relative abundance) was constructed with selected references (Fig. 6). PSC001, PSC002 and PSC005 did Figure 5. Neighbor-joining analysis of representative 16S OTUs obtained from the composting samples using MEGA 6.05. A custom database of microorganisms contained arsM genes was constructed by downloading corresponding 16S rRNA gene sequences from NCBI. All 16S rRNA gene sequences from composting samples were checked against this database, and sequences that have a similarity ≥95% and read depth ≥10 were retained. Bootstrap values >50% are shown on nodes. The scale bar indicates sequence dissimilarity between nodes. The taxonomic assignment of OUTs of the compost piles is indicated in parentheses.

Discussion
The rapid and sustained increase of temperature in the thermophilic stage in the two compost piles is an indicator of rapid establishment of microbial activity (Supplementary Figure S1), as biodegradation of organic matter resulted in a substantial heat production 20,29 . During the manure composting, As became more concentrated in the compost (Supplementary Figure S2), which was consistent with bulk mass loss via respiration and mineralization of organic matter during composting 30 . Song et al. reported that As concentrations increased by 57% in cow manure compost and 36% in pig manure compost 31 . However, the total As concentration is an overall pollution indicator but provides little information about which species are present, or about their potential of mobility and bioavailability in the environment. The total concentrations of methylated As species increased in the composting piles especially during the mesophilic and thermophilic phases. The concentrations of methylated As species represented 37% (MC1) and 35% (MC2) of total As concentration by day 60, clearly indicating As methylation during manure composting. Similarly, the conversion rates up to 50% for As have been reported in alfalfa hay composting 19 . Methylated As species analysis also showed that MMA concentrations peaked in the thermophilic phase and dropped subsequently rapidly in the maturing phase (Fig. 1). However, a continuous increase was found in DMA concentrations, indicating the transformation of MMA to DMA (Fig. 1). Diaz-Bone et al. reported that the concentrations of methylated species, first MMA, then DMA, peaked during the thermophilic phase of dry alfalfa hay composting, while at the end of the thermophilic phase both MMA and DMA decreased 19 . Higher temperature could considerably accelerate the rate of As methylation by enhancing microbial or enzymatic activities. Algal enzymes were shown to convert As(III) to DMA within 30 min at 70 °C, but not at 37 °C 18 . Purified recombinant CmArsMs enzymes were shown to transform As(III) into MMA, DMA and TMAO with an optimum temperature of 60-70 °C 18 . Bas Vriens et al. reported an increased rate of As methylation with increasing temperatures in both surface water and air suggesting that the methylation of As is temperature dependent 32 . It is also reported that at high pH, the presence of organic matter and moderate moisture and temperature would favor Bootstrap values >50% are shown on nodes. The scale bar indicates sequence dissimilarity between nodes. The taxonomy note shown at the right is based on known sequences and corresponding species. microbial-mediated As biotransformation 33 . Methylated As concentrations were found to correlate significantly negatively with soil pH, but positively with dissolved organic carbon 34 . Furthermore, the addition of organic matter stimulated the methylation of As species 35,36 . A recent study has shown that As methylation increased with the decrease of Eh values in a soil environment 37 . Thus, the lower Eh values in the maturing phase (shown in Supplementary Table S1) may have also enhanced As methylation during manure composting in this study. The stepwise As methylation was indicated by successive maxima of methylated As, while continuously increase of DMA was observed, TMA was not detected in the current study. The conversion of DMA to TMA is the rate limiting step in As methylation 38 . Therefore, DMA commonly accumulates in the environment. In the present study, the total concentrations of methylated As species increased little in maturing phases , reflecting the balance between As methylation and the losses of methylated As. MMA transformed to DMA and a fraction of DMA may be reduced to dimethylarsine, or further methylated to form trimethylarsine, both of which are volatile. It was reported that up to 35% of As was volatilized from anaerobic digestion of cow dung using cultures of methanogenic bacteria 39 . Moreover, As can be volatilized into methylarsines, mainly trimethylarsine and some monoand dimethylarsine in model biogas digesters when the major substrates were rice straw and animal manure 40 . Therefore, we postulate that a portion of As in the manure volatilized into the atmosphere during composting in our study and the risk of the volatile As in the compost should be paid more attention. However, the species and levels of volatile As in this experiment remains unknown, which will be determined in our further studies.
Methylation of As is mediated by microbes and arsM gene plays an important role. The arsM gene copies has been investigated mostly in soil in previous studies, For example, Jia et al. found that the copy numbers of arsM genes in paddy soil were in the order of 10 7 -10 8 copies g −1 dry soil assessed by qPCR assay 25 . In the present study, the copy numbers of the arsM gene (Fig. 2b) indicated that the As methylation potential could be highly detected in composting samples. The increase of arsM gene copies with composting time was consistent with other studies. Zhao et al. reported that the arsM gene copy numbers correlated positively with soil pH 34 . Carbonell-Barrachina et al. also found that As methylation could be measurable at pH 6.5 and 8.0 in a sewage sludge suspension, but was drastically restricted at pH 5.0 41 . In our study, the pH values of our compost piles gradually increased to 6.0 on average, and Eh gradually decreased to 62.0 mV on average from day 15 in both compost piles (Supplementary Table S1), which stimulated As methylation. Furthermore, the high concentration of organic substrates in manure may increase carbon and energy sources for the As methylating microbes, which in turn has a positive effect on As biomethylation 36 . The copy numbers of 16S rRNA gene decreased during the thermophilic phase and rebounded slightly in the maturing phase (Fig. 2a), which was similar to the results reported by Wang et al. 42 . This suggested that the temperature may highly affect the microbial community structure dynamics during manure composting 43,44 . In our case, the proportion of arsM-containing microbes increased steadily over the course of compositing, as shown by the progressive increase in the ratio of arsM genes copy numbers to bacterial 16S rRNA gene copy numbers over time (Fig. 2c). Thus arsM-harboring microbes constituted a substantial (~5-8%) fraction of the bacterial community in the later phase of composting. At the same time, the concentrations of methylated As correlate positively with the copy numbers of arsM genes (Fig. 3). Ma et al. reported a positive relationship between arsM gene copy numbers and the concentrations of organic As in rice grain 45 . Jia et al. also found a positive linear relationship between arsM gene copy numbers and concentrations of methylated As species in a soil solution of rhizosphere and bulk soils 25 . Considering the arsM genes play the key roles in microbial As methylation 18,46 , the increasing ratio of arsM gene copies to bacterial 16S rRNA gene copies along with the composting process and the positive correlation between the concentrations of methylated As with the copy numbers of arsM genes confirmed the As methylation ability in the manure compost piles. The ratio of arsM to bacterial 16S rRNA gene copies appears to be a very useful biomarker and predictor of As methylation in many different systems. Moreover, both compost piles MC1 and MC2 behaved similarly (Supplementary Table S2), further confirming our results.
Much attention has been paid to the bacterial communities during the composting process. Firmicutes, represented primarily by Bacilli, were reported as a major group in the mesophilic-and thermophilic-phases of coffee composting and the microbial diversity was significantly limited by temperature 47,48 . In our study, the overall abundance of Firmicutes decreased (Supplementary Figure S3), but Firmicutes was dominant and increased with composting time in the analysis of 16S OTUs related to As methylation (Fig. 4). This suggested that some Firmicutes might be key microbial members contributing to As methylation in compost piles. Some Firmicutes have been reported to harbor putative As resistance genes 49 . Actinobacteria (containing abundant orders Pseudonocardiales, Streptomycetales, and Corynebacteriales) were also found to increase subsequently as the compost pile entered the maturing phase (Fig. 4), as reported in other compost sites 50 . Firmicutes, Actinobateria, and Proteobacteria were prevalent in high As aquifers and acid mine drainage 51,52 suggesting these microbial members are As resistant.
Further, phylogenetic analysis of representative 16S OTUs revealed the abundance of species likely responsible for As methylation in the compost samples (Fig. 5). For example, As biomethylation has also been observed in anaerobic archaea, such as methanogens. We identified an OTU (OTU4319) similar to Methanosarcina thermophila TM-1 in samples from maturing phases. Streptomyces sp. strain GSRB54, a bacterium responsible for As methylation, was isolated from the roots of rice plants grown in As-contaminated paddy soil under anaerobic conditions. An OTU (OTU205) similar to Streptomyces sp. strain GSRB54, was also found to be more abundant in the samples from the maturing phases, when a higher methylated As content was present (Supplementary Table S4). We also identified OTUs similar to Sphaerobacter thermophilus DSM20745 whose arsM gene has been detected in soil 34 . Amycolatopsis mediterranei has been used to examine multiple alignments of deduced amino acid sequences of microbial arsM 53 . These results indicated that 16S OTUs similar to these strains might have an association with As methylation.
Phylogenetic analysis of arsM clones showed high diversity of arsM genes in the composting samples (Fig. 6). The function of the arsM genes was clearly established when it was cloned from Rhodopseudomonas palustris Scientific RepoRts | 7:42198 | DOI: 10.1038/srep42198 and expressed in an arsenic-hypersensitive strain of Escherichia coli 18 . Rhodopseudomonas palustris has also been a good model organism for studying As detoxification 54 . Clostridiales, Streptomycetales, Desulfovibrionales and Burkholderiales, have also been found to be widely distributed in an As polluted river 55 ; these same groups were also revealed from arsM gene sequences in our compost piles. However, the most abundant arsM partial sequence clones (PSCs), PSC001, PSC002 and PSC005 belonged to unknown taxa indicating arsM diversity begging further investigation. We also identified PSCs with high similarity to arsM genes from Streptomyces sp., Amycolatopsis mediterranei and Sphaerobacter thermophiles. Meanwhile, Streptomyces sp. and Amycolatopsis mediterranei classified in the Actinobacteria phylum were also found in the phylogenetic analysis of 16S OTUs (Fig. 5). The correspondence between 16S rRNA and arsM phylogeny, and increases in the abundance during composting confirm that Streptomyces sp., Amycolatopsis mediterranei and Sphaerobacter thermophiles might be highly involved and more active than other microbes in mediating As methylation during manure composting. Besides, the increased abundance of Actinobacteria in the analysis of 16S OTUs related to As methylation further suggested that Actinobacteria played an important role in the methylation process. However, further studies should be carried out to verify whether these species were the key contributors to As methylation during manure composting.
Furthermore, there remains a discrepancy between the 16S rRNA gene sequencing and arsM clone library analysis results. For the two most abundant 16S OTUs shown in Supplementary Table S4, OTU4960 (similar to arsM-containing Viridibacillus arvi) and OTU1947 (similar to arsM-containing Bacillus sp. FJAT-21945), we could not identify the corresponding arsM gene in our clone analysis (Fig. 6). Conversely, in the arsM clone analysis, beside Rhodopseudomonas palustris, arsM genes from Pelotomaculum thermopropionicum, Rubrivivaxbenzoatilyticus, and Gemmatimonasaurantiaca could not be found in the 16S rRNA gene phylogeny (Fig. 5) and the most abundance PSCs could not be assigned to a known phylogeny. This suggested a good genetic diversity of arsM gene, however, we still don't know exactly what taxa PSC001, PSC002 and PSC005 are from. On the other hand, our results also suggested the inconsistency of 16S rRNA gene sequencing and arsM clone library analysis. There are possible reasons for this inconsistency. Firstly, Depth of sequencing and primer biases may explain some of these discrepancies. It is more likely that the actual distribution and diversity of arsM-containing microbes is much broader than currently known. Secondly, prokaryotes (and some eukaryotes) are asexual and could not form species in a genetic way. Thirdly, horizontal gene transfer contributed to difficulties in assigning phylogenty, also reported by Jia et al. in soil and rice rhizosphere 25 . To date, environmental diversity surveys of As methyltransferase arsM genes remain limited to paddy soils and rice rhizosphere identifying many putative arsM-containing genera, including Actinobacteria, Gemmatimonadales, α-Proteobacterales, δ-Proteobacterales, β-Proteobacterales, Sphaerobacterales, Firmicates, CFB group bacteria, Halobacteriales, Archaea and other 5 unknown groups 25 . Zhang et al. reported that arsM sequences in paddy soils belonged to Gemmatimonadales (16%), Firmicutes (9%), Actinobacteria (11%), Proteobacteria (34%), and Archaea (6%) 56 . The present study showed putative arsM affiliated with a wide range of phylogenetic taxa that were present in all composting samples. The wide distribution of As(III) methyltransferases during composting indicated the potential for As methylation resulting in the accumulation of methylated As species. The abundance and diversity of arsM suggested arsM-carrying bacteria detected by limited molecular approaches may be generalists and As methylation process in pig manure composting may have been overlooked. Therefore, comprehensive analysis or a combination of different analysis should be used for studying the functional microbes in the composting.
In conclusion, two pilot-scale pig manure compost piles were established and an accumulation of methylated As was revealed over the composting process. By qPCR, 16S rRNA sequencing and clone libraries, a diverse group of arsM genes in composting pig manure have been confirmed presented in high abundance and diversity, and they increase along with the composting process. Streptomyces sp., Amycolatopsis mediterranei and Sphaerobacter thermophiles were found in both 16S rRNA and arsM gene sequence libraries, suggesting that these bacteria might be highly involved in the As methylation during manure composting. The inconsistency of 16S rRNA gene sequences and arsM clone library sequences betray a wider diversity than currently appreciated. To our knowledge, this is the first exploration of As methylation and the diversity and distribution of arsM gene during manure composting. Moreover, quantitative analysis of the ratio between arsM to bacterial 16S rRNA gene copies appears to be a very good indicator of As methylation in many different systems. Further work should be conducted to define the connection between phylogeny and arsM gene clone sequence and activity.

Materials and Methods
Composting experiments and sampling. Two independent pilot-scale (2.5 m × 1.8 m × 0.75 m in length, width and height) pig manure compost piles were set up in Hangzhou, China, containing on average 1750 ± 60 μg kg −1 As (dw, n = 3). The first manure compost pile (MC1) was composed of 1,200 kg pig manure and 600 kg sawdust for optimal C/N ratio and water content. The second manure compost pile (MC2) contained 1,200 kg pig manure and 600 kg sawdust mixed with burned rice straw. The moisture contents of the composting sites were maintained at approximately 65% by sprinkling water once every two days. Furthermore, compost piles were turned over and mixed once every two days in the first month, and once every four days in the second month for aeration. Determination of bacterial 16S rRNA and arsM genes copy numbers by qPCR. Compost samples were thawed and total DNA was extracted using the FastDNA SPIN Kit for Soil (MP Biomedicals, USA) according to the manufacturer's protocol. The extracted DNA was then quantified using a Nanodrop ND-2000 UV-Vis spectrophotometer (NanoDrop Co., USA) and stored at −20 °C until further analysis. Copy numbers of bacterial 16S rRNA gene in the compost samples were estimated by qPCR using primers 1369F (5′-CGGTGAATACGTTCYCGG) and 1492R (5′-GGWTACCTTGTTACGACTTT). The arsM genes were recovered using universal arsM gene primers as described previously 25 . qPCR assays of the arsM genes used the primers arsMF (5′-TCYCTCGGCTGCGGCAAYCCVAC) and arsMR (5′-CGWCCGCCWGGCTTWAGYACCCG). Details of the amplification conditions of bacterial 16S rRNA and arsM genes are in the Supporting Information.
16S rRNA gene sequencing. The V6-V8 hypervariable region 57 16S rRNA gene was amplified using universal primers 926F (AAA CTY AAA KGA ATT GAC GG) and 1392R (ACG GGC GGT GTG TRC) with the incorporation of multiplex barcodes 58 . The 16S rRNA gene PCR amplifications were then sent for sequencing using Illumina Miseq sequencing platform (Miseq, Illumina Inc., USA) at Zhejing Institute of Microbiology, China. The details of the 16S rRNA gene PCR, sequencing and bioinformatic analysis are presented in the Supporting Information. Further, a database of microorganisms containing an arsM gene was constructed by downloading all 16S rRNA and corresponding arsM gene sequences from NCBI (Supplementary Table S5). All 16S rRNA gene OTUs of samples were checked against the database and 83 sequences that have a nucleotide similarity ≥95% and read depth ≥10 were retained.
Construction of the arsM Clone Library. Four samples (15-MC1, 15-MC2, 60-MC1 and 60-MC2) were selected for the construction of arsM gene clone libraries. The arsM genes were recovered using universal arsM gene primers arsMF and arsMR as described previously 25 . Details of the amplification conditions and the arsM sequence analysis are in the Supporting Information. A total of 23 different arsM partial sequence clones (PSCs), ≥89% similarity level, were recovered from 422 sequenced clones using the MOTHUR program 59,60 . All arsM PSCs were checked against the database shown in Supplementary Table S5.
Phylogenetic analysis. 27 representative 16S rRNA OTUs from 83 retained 16S OTUs and 13 representative 16S rRNA gene sequences of each order from the database were selected and a neighbor-joining tree was constructed using Mega 6.05. 8 most abundance (>1%) arsM PSCs with 26 representative arsM sequences of each order from the database were selected and a neighbor-joining tree was also constructed using Mega 6.05. Heat maps and clustering analyses of arsM PSCs were generated with the R-package (v.3.2.4), which showed the relative abundances of the 23 arsM PSCs.
Statistical Analyses. Statistical analyses were performed with the use of SPSS 20.0 software (SPSS Inc., Chicago, IL). The significant differences in all the measured variables between the composting piles were tested by two-way analysis of variance ANOVA followed by Tukey's test. A p-value less than 0.05 was considered to be significant. Bivariate correlations were conducted to estimate the link among different parameters.