Change in microbial community in landfill refuse contaminated with antibiotics facilitates denitrification more than the increase in ARG over long-term

In this study, the addition of sulfamethazine (SMT) to landfill refuse decreased nitrogen intermediates (e.g. N2O and NO) and dinitrogen (N2) gas fluxes to <0.5 μg-N/kg-refuse·h−1, while the N2O and N2 flux were at ~1.5 and 5.0 μg-N/kg-refuse·h−1 respectively in samples to which oxytetracycline (OTC) had been added. The ARG (antibiotic resistance gene) levels in the refuse increased tenfold after long-term exposure to antibiotics, followed by a fourfold increase in the N2 flux, but SMT-amended samples with the largest resistome facilitated the denitrification (the nitrogen accumulated as NO gas at ~6 μg-N/kg-refuse·h−1) to a lesser extent than OTC-amended samples. Further, deep sequencing results show that long-term OTC exposure partially substituted Hyphomicrobium, Fulvivirga, and Caldilinea (>5%) for the dominant bacterial hosts (Rhodothermus, ~20%) harboring nosZ and norB genes that significantly correlated with nitrogen emission pattern, while sulfamethazine amendment completely reduced the relative abundance of the “original inhabitants” functioning to produce NOx gas reduction. The main ARG carriers (Pseudomonas) that were substantially enriched in the SMT group had lower levels of denitrifying functional genes, which could imply that denitrification is influenced more by bacterial dynamics than by abundance of ARGs under antibiotic pressures.


Results
Effects of antibiotics on nitrogen and nitrogen oxides gas emission. The emissions of NO, N 2 O and N 2 were all impacted during the period of the whole experiment (Fig. 1). The complete denitrification rate (N 2 flux) decreased from ~3.5 μ g-N/kg-refuse·h −1 (Fig. S1) to 1.1 and 0.07 μ g-N/kg-refuse·h −1 (SMT) after the addition with OTC and SMT, respectively (one-way ANOVA, p < 0.05). Meanwhile, refuse added with OTC brought about a higher emissions flux of nitrogen intermediates where the dominant contributor was N 2 O (5.2 ± 1.8 μ g-N/kg-refuse·h −1 ). However, the average emission fluxes of target gases were all below 0.5 μ g-N/kg-refuse·h −1 in the SMT group and occurred at significantly lower rates than in the OTC-amendment group (p < 0.01). The intrinsic denitrification of raw samples appears to be more susceptible to SMT than to OTC. Figure 1 shows that after the long-term antibiotics incubation, N 2 fluxes increased significantly from 1.1 ± 0.2 to 4.2 ± 0.2 μ g-N/kg-refuse·h −1 of OTC and 0.09 ± 0.02 to 0.32 ± 0.1 μ g-N/kg-refuse·h −1 of SMT (p < 0.01, paired t-test). Notably, longterm exposure of refuse to OTC-mitigated emission flux of both two target nitrogen oxide gases (paired t-test, p < 0.01) was essentially consistent at ~0.01 μ g-N/kg-refuse·h −1 . However, the SMT stimulated emission of NO from < 0.05 to 5.6 ± 0.6 μ g-N/kg-refuse·h −1 (paired t-test, p < 0.01). Denitrification could be partially streamlined with long-term incubation (compared to the short-term exposure), which can be more pronouncedly reflected in the SMT group. The total nitrogen (sum of NO, N 2 O and N 2 ) gas flux is increased significantly from < 0.1 to ~6.5 μ g-N/kg-refuse·h −1 (Wilcoxon tests (WT), p < 0.01).  4). The error bars indicate standard deviation (± 2 SD); a statistical significance (p < 0.01, one-way ANOVA) between the mean values was labeled with double asterisks (**). The dash line indicates the mean N 2 flux of raw landfill refuses after the injection of leachate (without addition with antibiotics, 3.51 ± 0.61 μ g-N/kg-refuse·h −1 ); the emission fluxes of NO and N 2 O from this control group were both < 0.1 μ g-N/kg-refuse·h −1 ). Influence of antibiotics on the distribution of ARGs and denitrification functional genes. A total of 14 ARG types were detected among all samples via metagenomics sequencing, and an average of 108, 138 and 120 ARG subtypes were further assessed in the A (previously unexposed), B (OTC) and C (SMT) samples, respectively (Table S1). Figure 2 shows that the most prominent ARG types included bacitracin (44.2%), sulfonamides (SAs, 15.1%), multi-drug (13.9%), macrolide− lincosamide− streptogramin (MLS, 6.1%), tetracyclines (TCs, 5.1%), fluoroquinolones (FQs, 4.0%), chloramphenicol (CPs, 3.1%), beta-lactam (3.4%) and aminoglycoside (2.8%). Although bacitracin was not added to any samples in this experiment, the sequences (ppm) of genes coding for this type of resistance (baca and bcra) were significantly higher than any other ARGs (one-way ANOVA, n = 12, p < 0.05). As shown on the right side of Fig. 2, the overall ARG sequences were more distributed to long-term SMT-exposed samples (in B1 and B2, 50.9%). The percentages of sul R (genes resisting to SAs) sequences increased substantially from 2.1% to 11.3%, albeit with no statistical significance detected (WT, p = 0.1). However, the addition of OTC imposed little impact on the abundance of tet R (0.4% to 0.6%) and multidrug (2.8% to 3.2%) genes (WT, p = 0.55). The tet R (genes resistant to TCs) sequences in the long-term OTC-exposed samples were at a significantly lower level than in the SMT groups (WT, n = 23, p < 0.05). Similarly, a higher abundance of multidrug resistance genes was detected in the SMT group compared to both OTC and unexposed groups ((WT), n = 26, p < 0.05). These results may imply that ARGs are more prone to disseminate in SMT-amended refuse versus the OTC.
All denitrification functional genes, except for norC and napB, were sequenced and annotated in this experiment (Table S2). Overall, PCA indicates broad changes in the main denitrification functional genes among the unexposed (raw refuse), SMT-and OTC-exposed samples (Fig. S2). As summarized in Table 1, the dominant genes in each denitrification step are narG, nirS, norB and nosZ with relative abundances of 0.031%, 0.021%, 0.019% and 0.016%, respectively, on average. Although long-term antibiotics incubation reduced the total abundance of genes encoding nitrate reductase (nar and nap) from ~0.07% to 0.05% and 0.04% in OTC-and SMT-amended samples, respectively, the levels of genes encoding nitrite reductases (nirS and nirK) remained at 0.004% in the OTC group. Statistically, the relative abundances of nap, nar and nir genes (functioning NO production) decreased significantly by 0.005% in the SMT group (WT; n = 12, p < 0.01). However, no significance was detected between unexposed and OTC-exposed groups (WT; p = 0.07). Interestingly, the relative abundance of norB was enriched in the OTC group from 0.018% to 0.23%, while the relative abundance of norB decreased by 0.003% in the SMT group. In contrast, the nosZ decreased in both of the two antibiotics-exposed samples, which was seemingly more pronounced in the SMT group from 0.02% to 0.01% on average. Figure 2. Distribution of ARG types in previously unexposed (raw refuse for short-term exposure) and exposed landfill cover refuse samples. The annotated sequences were visualized by using Circos (http://circos.ca/). The A, B and C denote previously unexposed, SMT-and OTC-amended (exposed) groups, respectively; 1 and 2 are replicate samples. The length of the bars on the outer ring represents the percentage of ARGs in each sample (right side in diagram), and the number of inner rings indicates relative abundance of annotated ARGs sequenced in each sample (ppm).  (Table S3). However, the PCA results show that antibiotics exposure affected the microbial community structures substantially, particularly of potential denitrifiers (Fig. S3). Under no antibiotic stress conditions (Table S3), the predominant bacterial genera were Sphingomonas (6.30%), Gemmatimonas (3.6%), and Ignavibacterium (2.7%). Under the SMT-and OTC-exposed conditions, Hyphomicrobium (6.4%), Anaerolinea (5.2%) and Pseudomonas (1.9%) had the highest abundance. Among these genera, 20 genera of potential denitrifiers were assembled as detectable regarding carrying genes of interest (denitrification functional genes, tet R , sul R and multi-drug resistance). Concretely, as shown in Table S4, 18 genera of potential denitrifiers were detectable (> 0.1%), among which the predominant denitrifiers were Sphingomonas (5.8%) and Bradyrhizobium (1.4%). After the antibiotics amended, the main denitrifying bacteria were Hyphomicrobium (6.2% of SMT, 2.0% of OTC) and Anaerolinea (2.0% of SMT, 4.0% of OTC). Other dominant genera included Caldilinea (1.7%) Nitrospira (1.2%), Rhodothermus (0.7%), Fulvivirga (0.5%) and Thauera (0.3%).

Discussion
A major finding in the present study is the contrasting three target gases that were produced during denitrification between the two antibiotics-exposure strategies. The observed results may imply that antibiotics in the landfilled refuse could inhibit denitrification in both the short-and the long-term, and their intrinsic N 2 production capacity could decline by > 50% if OTC-containing leachates were encountered. However, SMT-containing leachates may impede denitrification at an early step of N 2 O reduction. As shown in Fig. 1, this N 2 O processing (reduction) rate in SMT-exposed refuse is two orders of magnitude lower than in OTC-exposed refuse (0.09 ± 0.01 vs. 5.2 ± 1.8 μ g-N/kg-refuse·h −1 ; t-test, p < 0.01) but still increased by approximately tenfold compared to the control group (Fig. S1). These results are similar to previous research results that SMT exposure reportedly elevated N 2 O flux three times to 0.07 ng-N/kg·h −1 12,30,31 . Apparently, the observed deficiency in denitrification of refuse may not be ascribed to a low abundance of resistance genes, given that the unexposed sample (A1 and A2) possesses substantially higher sul R than tet R with respect to average sequencing reads of 370 ppm and 80 ppm (Table S1).
Further, this SMT-vulnerability of landfill refuse was not apparently mended with long-term exposure to antibiotics. The increasing levels of genes encoding SMT resistance in SMT-exposed samples (Fig. 2), particularly of sul1 (100 ppm to 750 ppm) and sul2 (270 ppm to 1200 ppm), were inversely proportional to the NO x and N 2 gas emission matrix (Mental-test; r 2 = 0.31, p < 0.01). By contrast, this emission pattern is positively correlated with the abundances of denitrifying functional gene clusters (Mental-test; r 2 = 0.53, p < 0.001), where the decreased genes encoding NO reductase (norB, Table 1) could probably explain the increased NO flux from 0.05 to 5.6 ± 0.6 μ g-N/kg-refuse·h −1 (paired t-test, p < 0.01). Interestingly, although NO x reduction genes were less affected or even increased in OTC-exposed samples (e.g., norB), NO and N 2 O fluxes in OTC-amended samples both declined by tenfold ( Fig. 1), probably because C1 and C2 steadfastly hold relatively higher levels of nosZ and norB during the whole experiment, which could lead to a higher emission of N 2 32,33 . Similarly to the SMT group, analysis also indicates that the target gas emission matrix is positively correlated with denitrifying functional gene clusters (Mental-test; r 2 = 0.84, p < 0.001) in OTC-exposed samples but not with tet R (Mental-test; p > 0.05) where the dominant functional gene clusters are tetG (50 ppm) and tetX (20 ppm) on average; Table S1).
It is unsurprising that long-term antibiotics exposure resulted in a drastic increase of ARG abundance (Fig. 2), but the increasing trend did not correspondingly streamline the denitrification, especially under SMT-stressed conditions. According to Martinez et al. 34 , it is still unclear whether ARGs integrated and-or propagated via horizontal gene transfer (HGT) could confer new metabolic capabilities to the recipients. Additionally, equating the detected ARG sequences to being indicative of resistance may be biased by issues such as the existence Functional genes Abundance (genes/total of each sample; Mean ± SD, n = 2)

Control
Long-term SMT exposed Long-term OTC exposed of extracellular ARGs, which do not necessarily result in expression 35 . As shown in Fig. 3, assigned sequences regarding crosstalk of the denitrifying process and ARGs in refuse show that most potential denitrifiers are not intrinsic ARG carriers, and Pesudomonas was the only detectable genus owning all target genes. The VPA results show that more than half (53%) of the NO x and N 2 gas emission pattern can be explained by the variation of denitrifying functional genes, particular of nirS, nirK (0.1%) and norB (36%), while the target ARGs encompassing tet R , sul R and multidrug (Fig. 4) and diversity of potential denitrifiers (Shannon-Wiener index) individually contributed < 0.1% (Table S5). The main hosts of nitrate reductase genes (narHG and napA) in unexposed raw refuse are Caldithrix (napA; 37.4%), Caldilinea (narH; 23.5%) and Sphingomonas (narG; 15.6%). The abundance of this microbial cluster experienced a significant decline after exposure to SMT and OTC (one-way ANOVA, p < 0.05). NO-producing gene (nirS) hosts such as Anaerolinea and Bradyrhizobium were numerically enhanced by > 10% in the SMT and OTC groups, which seemingly agrees with the increasing NO flux in the SMT group (Fig. 1). However, no significant correlation was detected (Pearson correlation, r 2 = 0.54, p > 0.05). It is noteworthy that the significantly lower abundances of nap, nar and nir genes in the SMT amended groups (Wilcoxon tests, p < 0.05, Table 1) surprisingly decreased the nitrate and nitrite (37.9 and 3.4 mg-N/kg-refuse) residuals, compared to its OTC counterparts (t-test, p < 0.01, Fig. S4). This may suggest that these genes and their potential functions in substrates supplies (for target NO x gases), albeit negatively affected by the presence of antibiotics (Table 1), do not bottleneck the whole denitrification. Additionally, the decreased abundance of nirK (functioning NO production) hosts (e.g., Rhodanobacter, from 6.4% to < 2.0% in SMT and OTC samples) and the low explanatory portion assigned to nirS(K) (0.1%, Fig. 4) may imply that they are not target gas emission determinants.
The hosting bacteria of potential determinant norB (explaining 36% of the emission pattern) in the SMT group included mainly Hypomicrobium (7.3%), Fulvivirga (10.9%) and Rhodopirellula (8.2%), and they were at a significantly lower level than in the OTC group (WT; n = 12, p < 0.01). The long-term OTC exposure substantially increased the hosts of norB genes functioning for NO reduction (Fig. 3) such as Hypomicrobium (16.5%), Fulvivirga (12.4%) and Thauera (5.6%), albeit with no significant difference (WT; n = 12, p > 0.05). Further, even For the reads of the 3 types of resistance genes, their taxonomic assignments were determined using BLAST against NCBI-nr, and the relative abundance of those assigned to the top 20 genera are also shown. The reads number visualized by MEGAN was taken as 100%. The A, B and C denote the mean values of two replicates of previously unexposed (raw refuse for short-term exposure), SMT-and OTC-exposed groups, respectively; and the color intensity in each panel shows the percentage of each target gene carrying a genus in one sample. though the sequenced nosZ reads were less distributed to Rhodothermus (from 18.9% to 12.7%) in both two antibiotics-amended samples, the OTC group was observed to own comparatively diverse hosts such as Caldilinea and Thauera (~5%) and exclusively including 12.4% of Fulvivirga carrying nosZ (Fig. 3). This finding is consistent with other studies that suggest the stepwise processes, such as nitrification and anaerobic digestion, are largely determined by the composition of functional microbial groups more than resistance genes [36][37][38] . Some key functional gene hosts identified in this study, such as Hypomicrobium, Caldilinea and Thauera, were also reportedly reducing NO and N 2 O under antibiotics or toxicants stressed habitats 39,40 , where higher abundances of Hypomicrobium carrying norB and nosZ were reckoned to process a more complete denitrification 41 , probably because these two genes, compared to nir, nap and nar gene clusters, have comparatively low functional redundancy 42 and therefore are likely to be the ceiling factors of the denitrifying process, especially under extreme conditions.
Our research suggests that the mechanisms underlying accumulation of nitrogen intermediates from denitrification in antibiotics-stressed conditions are majorly rooted in the composition of denitrifiers rather than the abundance of ARGs. It is still not very clear how OTC is less inhibitory than SMT to denitrifying bacteria. Nevertheless, the elevated abundances of some bacteria like Hypomicrobium, Caldilinea and Thauera potentially harbouring the norB and nosZ could be responsible for facilitating the dentrification process. The meta-sequencing results show that the enrichment of target resistance genes took place limitedly within Pseudomonas (Fig. 3). Recent studies also indicate that the dissemination of ARGs could not completely decouple from phylogeny and effectively transfer across habitats 43 . It is not uncommon that the resistome sequencing with ARGs integrated or propagated by plasmids could not effectively discern the relationship between these pseudo-genes to their originals 38,44 , future studies concerning the resistance detection of culturable bacteria therefore are highly suggested. Also, to more clearly reveal the mechanisms of the antibiotic-stressed denitrification, a more precisely operated experiment is needed particularly by calculating the N-balance with all nitrogen species included. At current stage, we could conclude that the facilitation of restoring a denitrification in landfilled refuses exposed to antibiotics on long term is more pronounced by the evolving bacterial community than by increasing ARG levels.

Methods
Sample collection and pretreatment. Laogang Municipal Waste Disposal Co., Ltd (Laogang Landfill) which is located in the outer southeast suburbs of Shanghai and typically receives approximately 20000 t/d MSWs across greater Shanghai, was selected as the sampling site. The landfilling procedure was operated by a multi-lift method, and each lift was ultimately covered with 0.5 m of sandy clay. The refuse was collected in triplicate using corers (Geoprobe 7822DT System, US). Every 25 g sample of core refuse was preincubated at 37 °C in a 250 ml rubber sealed serum bottle, and N 2 was used to replace the headspace to keep an anaerobic condition. After one month of preincubation (to eliminate background NO 3 − -N and NO 2 − -N), the collected refuse samples were homogenized for the subsequent antibiotic-leachate incubation experiment. Landfill leachates were sampled from the leachate collection pipes of the subject landfill lifts, the composition of pretreated refuse and leachates is listed in Table 2. NH 3 -N in these leachate samples were precipitated by MPA methods as described before 45 , and replaced with the equivalent of NO 3 − , N (450 mg-N/L, Table 2).

Measurement of nitrogen oxides (NO and N 2 O) and nitrogen (N 2 ) gases.
The refuse-antibioticleachate mixture (composition detailed in next section) was put in a 50 ml crimp neck borosilicate vial (CNW, Germany) sealed with a rubber septum and aluminum cap, and the vial was purged with helium (He) for 4 min using a timer-controlled multi-valve air pump (45 s in evacuation mode, 3 min 25 s in He filling mode). This procedure was repeated for 4 cycles prior to gas measurement according to the chamber incubation technique 46 . The headspace gas of each vial was collected by an autosampler every 3 hours, and the headspace samples were measured by gas chromatography (GC, Agilent 7890 A, USA) with two working columns: (i) 10 m poraPLOT U fitted with an electron capture detector (ECD) for N 2 O (detection limit, 0.1 ppb); and (ii) 20 m 5 Å Molsieve fitted with a thermal conductivity detector (TCD) for N 2 (detection limit, 0.1 ppb). Column pressure (He) and temperature are 200 kPa and 36 °C, respectively, for the PLOT U column, and 250 kPa and 50 °C, respectively, for the Molsieve column. The NO concentration was measured by a chemiluminescence analyzer (Sievers Nitric Oxide Analyzer 280i, GE, USA) with a detection limit of 0.1 ppb. The emission fluxes of target gases over the tests were calculated using the following equation:   where the Ef denotes the target gas emission flux (μ g-N/kg-refuse·h −1 ); n and V represent the number of nitrogen atoms (N) in each target gas molecule and headspace volume (V = 50 ml), respectively; and C 1 , C 0 , m and T are indicatives with respect to selected real-time concentration (ppm), initial concentration of target gas (ppm) in the headspace, dry refuse weight (g) and detection time, respectively. The real-time measurement data in the 24 th hour were selected for emission flux calculation when the concentration of target gases became stable (real-time target gases emission data of samples after 60-day antibiotics amendment were provided in the Fig. S5).
Short-term antibiotic exposure experiments. This experiment was designed to mimic the acute effects of high dosage antibiotics on the denitrification in a landfill. The SMT and OTC (99.0%, Aladdin Industrial Inc., Shanghai) were dissolved in methanol (UPLC grade, CNW-GmbH, Germany) with a concentration of 250 mg/L and stored at − 20 °C in a lab freezer. Before the incubation test, two bulk antibiotic solutions were diluted into a final concentration of 25 mg/L with MilliQ water. The combined solution (diluted antibiotic solution and pretreated leachate) were both purged with He for 10 min. The refuse-antibiotic-leachate mixture in each test vial consists of 5 g (dry weight) of homogenized raw (unexposed) refuse, 1 ml of pretreated leachate and 1 ml of antibiotic solution (15 mg/L, 1 ml of MilliQ water for control samples). The nitrogen oxides (NO and N 2 O) and N 2 gases were measured consecutively for 24 hours as described above. The background levels of OTC and SMT in the tested inoculum were 0.24 ± 0.18 and 0.14 ± 0.07 mg/kg (n = 25), respectively (concentrations and method details given in the Supporting Information Table S6).
Long-term antibiotic exposure experiment. This experiment was designed to test the hypothesis that long-term exposure (SMT and OTC) will alleviate the inhibition of antibiotics on the denitrification process (compared to the acute effects) in the landfill due to the development of AR. According to previous research 22 , additional antibiotics from new adding layer could transfer to the wastes located in previous sampling surfaces via landfill leachate. The piling-up of one waste-compacts layer above our sampling areas general costs two months, therefore the incubation period is designed as ~60 days. Our field study of lyophilized refuse samples (n = 25) sampled from two large landfills in the Great Shanghai Metropolis (Laogang landfill and Liming landfill) shows that the 25 th percentile, median, 75 th percentile, and maximum SMT and OTC concentrations are 0.13, 0.21, 0.27 and 0.7 mg/kg, respectively; and 0.12, 0.16, 0.21 and 0.27 mg/kg, respectively (Table S6). Thus, another 50 g of collected landfill refuse was initially amended with SMT and OTC at a loading rate equivalent to the median level (0.2 mg/kg for both groups) at 37 °C in a 250 ml rubber-sealed serum bottle. The antibiotics loading on the culture was subsequently increased to 0.5 mg/kg and 5.0 mg/kg on day 20 and day 45, respectively. During the antibiotics-exposure incubation, N 2 was used to replace the headspace of the rubber-sealed serum bottle every day to maintain anaerobic conditions. The incubation procedure was ceased at Day 65 when the leachates were added in, and emission of nitrogen oxides from these long-term antibiotics-amended refuses was measured immediately.

DNA Extraction and Sequencing Analysis.
Samples collected in this study included raw landfill refuse (unexposed) and refuse-amended (long-term exposure) with SMT and OTC. Each sample was subdivided into two equivalent subsamples: one was used to calculate the physical composition of the refuse samples, another was centrifuged at 5000 rpm for 10 min (removing extra liquid), and 200 mg of the pellet was collected for DNA extraction using a PowerSoil DNA Isolation Kit (MOBIO, USA) according to the manufacturer's protocol. The yield and quality of the DNA extractions were verified by spectrophotometry (Nanodrop 1000), requiring that extracts with the OD 260/280 value between 1.8 and 2.0 be used to ensure that no sequencing inhibitors were present. The qualified DNA extracts were stored at − 20 °C for further analysis.
The resulting DNA extracts were provided in duplicate for library construction and high-throughput sequencing with the Illumina HiSeq 2000 platform. Approximately 5 Gb (giga base pairs) of data were generated for each sample. The raw metagenomic data were initially trimmed using Trimmomatic to remove reads with adapter contamination, with three or more ambiguous nucleotides, or with an average quality score below 15 but over 36 bp, and the reads were further filtered and trimmed by using Stickle (Version 1.33) to remove reads that did not meet the Q 20 requirement (> 50% reads; https://github.com/najoshi/sickle). The quality of the filtered data and the number of clean reads are both described in detail in the Supporting Information (Table S7). The remaining clean reads were used for further analysis: the putative ARGs were searched against an antibiotic resistance database (ARDB) with an E value of ≤ 10 −5 . For exploring denitrifying genes in refuse, the quality-filtered Illumina reads of different samples were separately aligned to local nucleotide databases created by downloading DNA sequences of napG(HJ), narA(B), nirK(S), norB and nosZ genes from RDP FunGene (http://fungene.cme.msu.edu/). The identification of the bacterial hosts of the ARGs and denitrifying genes was subjected to BLASTx against the NCBI-nr database with E-value cut-off at 10 −5 . The assigned sequences were then visualized by MEGAN with the Lowest Common Ancestor (using default parameters) algorithm as described in previous research 47 . The quality-filtered Illumina reads of all six refuse samples were submitted to the European Bioinformatics Institute server (accession number: ERS1069313).
Statistics. This study contrasted unexposed (raw refuses), SMT-exposed and OTC-exposed samples relative to the variation of target gases, functional genes, and microbial community using two independent metagenomes per incubation condition (6 samples in total). Prior to conducting statistical comparisons among the three groups, data duplicates of each group were standardized and tested to determine whether they were statistically Scientific RepoRts | 7:41230 | DOI: 10.1038/srep41230 the same by t-test 23 . If the replica were statistically the same, duplicates were clumped as averages to perform more rigorous statistical comparisons including two-sample testing (paired t-test) between unexposed and exposed samples, Wilcoxon tests for non-normally distributed subset data, and analysis of variation (ANOVA) among three incubation samples and/or different target gases/genes. Data analysis and statistical assessments were performed using SPSS 22 software (SPSS, Inc., Chicago, IL), and statistical significance was defined by 95% confidence intervals (p < 0.05, two-tail). Principal component analysis (PCA) indicating difference among samples was employed in PAST (http://folk.uio.no/ohammer/past/). The Mantel test and Variance Partitioning Analysis (VPA) were applied to evaluate the correlations among microbial communities with environmental variables and to evaluate the contributions of functional genes and denitrifying bacterial diversity to target gas emission patterns by using RStudio, respectively. Circos graphs were produced via Circos software on line (http://circos.ca/).