It is of great concern worldwide that gaseous emissions from management of organic solid waste contribute to regional and global-scale environmental processes, such as eutrophication, acidification, and climate change (Naylor et al., 2005; Hou et al., 2015; Owen and Silver, 2015; Pardo et al., 2015). Organic waste management has been identified as an important source of anthropogenic greenhouse gas (GHG) emissions, particularly methane (CH₄) and nitrous oxide (N₂O) [Intergovernmental Panel on Climate Change (IPCC), 2006]. Global CH₄ and N₂O emissions contribute considerably to the radiative forcing of the atmosphere, as their global warming potentials are 298 and 25 times that of carbon dioxide (CO₂) on mass basis over the 100-year time horizon, respectively [Intergovernmental Panel on Climate Change (IPCC), 2013]. Manures from livestock production account for 30–50% of the global agricultural N₂O emissions and 12–41% of the total agricultural CH₄ emissions for most countries (Oenema and Tamminga, 2005; Chadwick et al., 2011).

Manure composting is an alternative agricultural strategy for organic waste management that produces organic fertilizer to improve soil structure and fertility in croplands (Larney and Hao, 2007). The CH₄ and N₂O fluxes from manure composting have been studied extensively, contributing to a comprehensive assessment of CH₄ and N₂O emissions from manure composting worldwide [Czepiel et al., 1996; Hao et al., 2001; Intergovernmental Panel on Climate Change (IPCC), 2006; Xu et al., 2007; Mulbry and Ahn, 2014; Jiang et al., 2015; Pardo et al., 2015]. Yet, little is known about the interaction between pile physicochemical parameters and bacterial community, which has a key role in CH₄ and N₂O emissions from manure composting (Sharma et al., 2011; Angnes et al., 2013; Zhang et al., 2015). In particular, few studies have simultaneously focused on quantitative analysis of bacterial community composition and CH₄ and N₂O fluxes from composting manure (Maeda et al., 2010a,b; Chen et al., 2014; Zhang et al., 2015). Comparative quantitative analysis of specific bacterial functional groups and their interplay in CH₄ and N₂O emissions during manure composting are still limited.

To date, genes encoding enzymes involved in CH₄ and N₂O emissions have been targets of choice for studies focusing on functional groups of bacteria. This focus is fundamental for understanding mechanisms of carbon and nitrogen biogeochemistry and strategies for GHGs mitigation (Hu et al., 2015). Morales et al. (2010) illustrated denitrifying gene abundances as proxies for predicting N₂O emissions from soils as a response to different long-term land management regimes. Regan et al. (2011) found the evidence that differences in microbial abundances can help explain enhanced N₂O emissions in permanent grasslands under elevated atmospheric carbon dioxide. Nevertheless, a trade-off between CH₄ and N₂O fluxes has frequently been found in rice paddies and manure composting (Hou et al., 2001; Zou et al., 2005; Ahn et al., 2011; Shen et al., 2011; Mulbry and Ahn, 2014), simultaneous comparisons of the abundance of multiple CH₄- and N₂O-related genes and their interactions would be highly needed, especially when targeting functional bacteria to mitigate GHGs emission from agriculture (Hu et al., 2015).

Manure composting system is suggested as a good study model to examine the role of microbial abundances in shaping dynamics of CH₄ and N₂O fluxes due to their sensitive responses to changes in pile physicochemical properties, nitrogen transformation, and organic carbon decomposition during composting (Chadwick et al., 2011). Methane is produced by methanogenic organisms during the anaerobic degradation of organic materials; and the final key step, being reduction of CO₂ using H₂ to generate CH₄, is catalyzed by methyl-coenzyme M reductase (MCR, EC 2.8.4.1) (Kim et al., 2008). The highly conserved mcrA gene encoding the α-subunit of MCR has been widely used for analysis and quantification of methanogen communities (Pereyra et al., 2010; Sonoki et al., 2013). The generated CH₄ could be oxidized to methanol with the catalysis of particulate membrane bound methane monooxygenase (EC 1.14.13.25) (Xin et al., 2004). The pmoA gene encoding the α-subunit is widely used as the indicator for quantification of the methanotrophs from environmental samples (Wasmund et al., 2009; Sharma et al., 2011). The CH₄ flux is the net outcome and combined action of methanogen and methanotrophs that are closely related to changes in physicochemical parameters and environmental factors during manure composting process (Sonoki et al., 2013).

During manure composting, NH₄⁺ generated from amino acids can be oxidized to NO₂^- by ammonia-oxidizing archaea and bacteria (AOA and AOB, respectively), through ammonia monooxygenase (EC 1.14.99.39, encoding by amoA) and hydroxylamine oxidoreductase. A part of the NO₂^- could be oxidized to NO₃^- by nitrite-oxidizing bacteria (NOB) (Maeda et al., 2011). Bacterial denitrification is a biochemical reaction where oxidized forms of nitrogen, including nitrate, nitrite, nitric oxide, and nitrous oxide, are gradually reduced (Wang et al., 2013). The four steps are generally catalyzed by nitrate reductase (encoding by narG), nitrite reductase (nirS/nirK), nitric oxide reductase (norB), and nitrous oxide reductase (nosZ) (Maeda et al., 2011). Eventually, the N₂O emission is a result of dynamic balance between N₂O production and consumption. In addition, given that a trade-off between CH₄ and N₂O fluxes has been well documented in windrow composting systems (Hao et al., 2001; Zou et al., 2005; Ahn et al., 2011; Shen et al., 2011; Mulbry and Ahn, 2014), some compositional bacterial genes could be multifunctional as proxies for indicating dynamics of CH₄ and N₂O fluxes during windrow composting.

We conducted an in situ measurement of CH₄ and N₂O fluxes from a commercial composting windrow. Molecular techniques were used to quantify the abundances of CH₄- and N₂O-related functional genes. The main objective of this study is to examine whether bacterial gene abundances can indicate dynamics of CH₄ and N₂O fluxes during windrow composting. Specifically, we aimed to test three general hypotheses. The first hypothesis stated that changes in pile physicochemical parameters would shape diverse time course patterns of bacterial functional genes abundance during windrow composting, given that bacterial community response variables are sensitive to environmental change. Second, we predicted that some specific physicochemical parameters and compositional bacterial enzymes encoded by relevant genes would be multifunctional as involved both in CH₄ and N₂O due to a trade-off between CH₄ and N₂O fluxes during windrow composting. Eventually, as both production and consumption of CH₄ and N₂O are controlled by the interplay of enzyme encoding bacterial functional genes, we hypothesized that bacterial functional gene abundances could be used as proxies for indicating dynamics of CH₄ and N₂O fluxes during windrow composting.

There was a trade-off between CH₄ and N₂O fluxes during manure windrow composting (Figure 1). Substantial CH₄ emissions occurred primarily during thermophilic Phase I. During Phase I, CH₄ fluxes ascended rapidly until the peak fluxes were attained approximately 4 days after the onset of composting. Thereafter, CH₄ emission was dramatically decreased by pile turning and then remained lower release rate, which was closely associated with decreases in pile temperature (Table 2). In contrast, N₂O fluxes stayed relatively lower during Phase I, and they gradually increased during phase II. Eventually, N₂O fluxes were highest by the end of manure compositing (Figure 1).

Over the 65-day period of manure composting, CH₄-C fluxes varied from 0.4 to 134.5 g m^-2 d^-1, with an average flux of 17.1 g m^-2 d^-1 (Figure 1). Cumulative CH₄ emissions in terms of initial windrow surface area were 1.1 kg m^-2, being equivalent to 0.8% of total C in initial manure dry weight (MCF). The fluxes of N₂O-N varied within the range of 15.0–95.0 mg m^-2 d^-1 during Phase I, and rapidly increased from 114.2 to 1621.1 mg m^-2 d^-1 during Phase II, dedicating to an average of 383.3 mg m^-2 d^-1 over the whole composting process. Cumulative N₂O emissions in terms of initial windrow surface area were 25.1 g m^-2, representing 0.18 kg per ton of MDW. The emission factor of N₂O (EF, percentage of initial N in manure compost pile emitted as N₂O-N) was estimated to be 1.2% for composting windrow.

A contrasting time course pattern of mcrA and pmoA genes abundance was detected during windrow composting. The measured abundances of mcrA were the highest at the onset of composting, being average 6.26 log copy numbers⋅g^-1 (Figure 2). Thereafter, the mcrA gene abundance gradually decreased until it remained stable around ∼5.0 log copy numbers⋅g^-1. Relative to a smaller variation of mcrA gene abundance, pmoA gene abundance showed larger variation (variation range: 5.11–8.05 log copy numbers⋅g^-1) over the composting process. The measured abundance of pmoA decreased in the first week and then kept an ascending trend over the composting process, reaching the greatest abundance by the end of windrow composting (Figure 2).

For the functional genes involved in N₂O emissions, the abundance of amoA, narG, nirK, and nirS genes shared a rising pattern over the composting process, in contrast to a declining trend for norB and nosZ genes abundance (Figure 2). Over the composting process, predominant genes in bacterial community shifted from norB and nosZ genes (6.76–6.91 log copy numbers⋅g^-1) during Phase I to nirK gene (6.78 log copy numbers⋅g^-1) during Phase II (Figure 2). The amoA gene abundance stayed the lowest (∼3.10 log copy numbers⋅g^-1) over the whole windrow composting.

The CH₄ fluxes showed strong positive correlations with compost material parameters including moisture, C/N ratio, NH₄⁺-N, and TOC during the composting process (Figure 3). The CH₄ fluxes were also positively correlated with mcrA and nosZ gene numbers, but negatively correlated with pmoA and narG gene numbers. For the relative abundance of functional genes group, CH₄ fluxes were positively correlated with mcrA/pmoA (r = 0.78, p = 0.01). Besides strong negative correlations between N₂O fluxes and pile moisture, temperature and C/N ratio, N₂O fluxes showed significant positive correlations with pmoA, narG, and nirK genes abundance, but negatively correlated with nosZ gene abundance (Figure 3). The N₂O fluxes were positively correlated with relative abundances of functional gene group (nirK+nirS)/nosZ (r = 0.90, p = 0.001).

During the composting process, TOC, TN, moisture and C/N ratio were significantly correlated with each other, acting as a group to correlate with pmoA, narG, nirK, and nirS genes abundance and CH₄ and N₂O fluxes (Figures 3, 4). Among functional genes, the abundance of pmoA, narG, and nirS genes were correlated with each other. Besides, the narG gene abundance was correlated with amoA and nirK genes, but amoA gene abundance was not significantly correlated with nirK gene (Figure 4).

A stepwise regression analysis was conducted for modeling CH₄ and N₂O fluxes with bacterial functional gene abundances and physicochemical parameters (Table 3). The model based on bacterial functional genes mcrA/pmoA alone can explain 55% of the variance in CH₄ fluxes over the composting process (Y_C1, Table 3). While integrating mcrA gene abundance together with pile temperature (T), however, the simulated regression model explained as high as 94% of the variance in CH₄ fluxes (Y_C2, Table 3). Furthermore, the regression model including mcrA gene, T and NH₄⁺-N significantly lowered the model error and increased model efficiency, which can almost fully project the time course of CH₄ fluxes (Y_C3, Table 3 and Figure 5). Among the regression models, the model based on pile temperature (T), mcrA and NH₄⁺-N appeared to be the best fit for CH₄ flux variance when the statistics R², P, RMSE, and MEF were comprehensively evaluated (Table 3).

As shown in the stepwise regression model, a functional genes group, (nirK+nirS)/nosZ acted as a good proxy for predicting dynamics of N₂O fluxes (Y_N1, Table 1). While taking nirK minus nosZ genes into account but excluding nirS genes, the performance of simulated model was significantly improved, explaining as high as 92% of the variance in N₂O fluxes during windrow composting (Y_N2, Table 1). Besides nirK and nosZ genes, pmoA gene was also responsible for the variance in N₂O fluxes as shown in pairwise correlation (Figure 3). Indeed, about 95% of the variance in N₂O fluxes can be explained by the model based on linear regression of nirK, nosZ, and pmoA genes abundance (Y_N3, Table 1). Compared to the Y_N1 and Y_N2 models, the Y_N3 model including pmoA gene abundance as an additional predictor was able to minimize the uncertainty in N₂O flux estimates (Table 1 and Figure 5).

The simulated OLS models were also applicable to paddy rice cropping systems (Table 3 and Figures 6, 7). In rice paddies, about 75% of the seasonal variance in CH₄ fluxes as a response to elevated atmospheric CO₂ concentration and rising temperature can be explained by re-parameterized Y_C3 model based on linear combination of soil temperature, mcrA and NH₄⁺-N (Figure 7). Similarly, the re-parameterized nirK, nosZ, and pmoA genes abundance in Y_N3 model can largely reflect seasonal CH₄ fluxes response to elevated atmospheric CO₂ concentration and rising temperature in paddy rice cropping systems (Figure 7).

Manure composting has been recognized as an important source of CH₄ and N₂O to atmosphere [Intergovernmental Panel on Climate Change (IPCC), 2006; Chadwick et al., 2011; Hou et al., 2015; Owen and Silver, 2015; Pardo et al., 2015]. The IPCC guidelines introduce the terms of MCF (methane conversion factor) and EF (emission factor of N for N₂O) for accounting CH₄ and N₂O emissions from manure composting, respectively [Intergovernmental Panel on Climate Change (IPCC), 2006]. In the present study, total CH₄ and N₂O emissions were estimated to be 1.1 kg m^-2 and 25.1 g m^-2, being equivalent to a MCF of 0.8% and an EF of 1.2% for composting windrow, respectively. The value of MCF in this study falls well within the IPCC default value range of 0.5–1.5% in composting windrows [Intergovernmental Panel on Climate Change (IPCC), 2006]. The estimated EF (1.2%) in this study is slightly greater than the IPCC default EF of 1.0% [Intergovernmental Panel on Climate Change (IPCC), 2006], but highly close to the recent estimates (mean of EF: 1.2%) based on a summary of available data on composting windrow by Pardo et al. (2015).

Measurements of GHG fluxes showed a trade-off between CH₄ and N₂O fluxes, as previously found in rice paddy soils and windrow compost (Cai et al., 1997; Hou et al., 2001; Zou et al., 2005; Ahn et al., 2011; Shen et al., 2011). Consistent with previous studies (Hou et al., 2001; Sánchez-Monedero et al., 2010; Sharma et al., 2011; Chen et al., 2014), substantial CH₄ emissions occurred mainly in the early stage of manure composting process. In contrast, remarkable N₂O emissions were triggered around the middle stage of the composting when pile temperature started to decline and oxygen availability was limited (Fukumoto et al., 2003; Xu et al., 2007; Ahn et al., 2011; Tsutsui et al., 2013; Wang et al., 2013). In addition, norB and nosZ genes were relatively predominant during the early stage of composting (Figure 2), suggesting that much N₂O was further transformed into N₂ as the final product in denitrification.

Consistent with the first hypothesis prediction, changes in physicochemical parameters shaped different time course patterns of CH₄- and N₂O-related functional genes (Gödde and Conrad, 1999; Holtan-hartwig et al., 2002; Kandeler et al., 2006; Zhou et al., 2011; Zhang et al., 2015). Correlation analyses showed that the abundance of pmoA, narG, nirK, and nirS genes shared negative correlations with pile temperature (r = ∼-0.70, p < 0.05), moisture, TOC, and C/N ratio, and positive correlations with TN (Figure 4), suggesting that changes in interacting physicochemical parameters in the pile shaped the dynamic pattern of the pmoA, narG, nirK, and nirS genes (Figure 4), in line with the results obtained by Hallin et al. (2009) showing that nitrate reducers and denitrifiers were closely related to soil TOC, TN, and C/N ratio in a 50-year-old fertilization experiment. Similar relationships between methanotrophs abundances and abiotic parameters were found in landfill cover soils (Kumaresan et al., 2009; Lin et al., 2009). Some studies also reported the relationships between denitrifiers and physicochemical parameters, such as correlations of nirK and nirS genes abundance with pile temperature (Zhang et al., 2015), and correlations of narG, nirK and nosZ genes abundance with soil TOC (Kandeler et al., 2006; Zhou et al., 2011).

Significant correlations of nosZ gene abundance with NH₄⁺ and NO₂^- suggested that nosZ gene abundance dynamics was mainly associated with substrate availabilities. The mcrA and amoA genes abundance did not show significant correlations with any of physicochemical parameters, and amoA gene abundance was, on average, much lower than nitrate reducers and denitrifiers (narG, nirK nirS, norB, and nosZ) abundance, suggesting that denitrification was much stronger than nitrification during windrow composting (Hao et al., 2001). Partially due to pH remaining stable around 8.0 during windrow composting, pile pH did not show significant correlations with any of bacterial functional genes in this study, consistent with Kandeler et al. (2006) but contrary to other previous studies (Deiglmayr et al., 2004; Bárta et al., 2010; Zhang et al., 2015). In addition, NH₄⁺ was slightly correlated with narG, nirK, and nirS genes abundance (p = ∼0.07), while NO₃^- did not show significant correlations with nitrate reducer and denitrifier abundances, which might suggest that nitrate in pile manure is not important for denitrifiers (Tiedje, 1988; Mergel et al., 2001; Avrahami et al., 2002; Liu et al., 2003; Kandeler et al., 2006; Zhang et al., 2015).

The NH₄⁺ and pmoA gene were involved in CH₄ and N₂O emissions, respectively, which partially supported the second hypothesis that some specific physicochemical parameters and compositional bacterial enzymes encoded by relevant genes would be multifunctional as involved both in CH₄ and N₂O. The CH₄ fluxes showed a positive correlation with NH₄⁺, and NH₄⁺ was selected as an indicator in the stepwise regression model (Table 3 and Figure 3). A great many studies have revealed that NH₄⁺ has an inhibitory effect on CH₄ oxidization through either competition for methane monooxygenase or generation of toxic hydroxylamine and nitrite from ammonium oxidation (Steudler et al., 1989; Bosse et al., 1993; Dunfield and Knowles, 1995; Hanson and Hanson, 1996; Duan et al., 2013; Dam et al., 2014; Karbin et al., 2015), although stimulation effects or no effects of NH₄⁺ on methanotrophs were reported in some other studies (Dunfield et al., 1995; Delgado and Mosier, 1996; Dan et al., 2001; Krüger and Frenzel, 2003; Shrestha et al., 2010; Hu and Lu, 2015). The CH₄ fluxes were negatively related with narG but positively related with nosZ genes abundances, which might be due to the significant correlations of narG with pmoA genes (r = 0.84, p = 0.005) and of nosZ genes with NH₄⁺ (r = 0.76, p = 0.02). In addition, N₂O fluxes were correlated with pmoA gene abundances, and pmoA gene abundances were included in the regression model (Table 3 and Figure 3), which might suggest denitrification with methane as external carbon source. Some studies reported that aerobic methane-oxidation coupled to denitrification is accomplished by aerobic methanotrophs oxidizing methane and releasing soluble organics that are used by coexisting denitrifiers as electron donors for denitrification (Modin et al., 2007). Indeed, the pmoA, narG, and nirS gene abundances were correlated in this study (Figure 4).

Both mcrA and pmoA genes abundances were correlated with CH₄ fluxes, and the balance of mcrA/pmoA genes abundance was selected as a good indicator in the regression model (Table 3 and Figure 3), indicating that both methanogens and methanotrophs played important roles in CH₄ fluxes from composting windrow. The N₂O fluxes were positively correlated with narG and nirK genes abundance, but negatively correlated with nosZ gene abundance. However, N₂O fluxes were not correlated with nirS gene abundances in this study. The nitrite reducers with Cu-containing enzyme encoded by nirK gene are generally believed to be more important than those with cytochrome cd1 nitrite reductase encoded by nirS gene in the nitrite reduction step during manure composting (Yoshida et al., 2009; Bárta et al., 2010; Chen et al., 2010; Zhou et al., 2011; Zhang et al., 2015).

Based on physicochemical and biological variables measurements during the composting and their correlation and regression analyses, we developed a schematic model that explains the dynamics of CH₄ and N₂O fluxes associated with bacterial functional genes and physicochemical parameters during manure composting (Figure 8). The schematic model shows how the prevalence of bacteria is involved in key steps in the process of CH₄ and N₂O emissions, and the CH₄ and N₂O fluxes during windrow composting are controlled by the interplay of enzyme encoding bacterial functional genes (Figure 8). In the schematic model, some physicochemical parameters are correlated with each other and interacting to shape the dynamics of bacterial functional gene abundance. Besides bacterial functional genes are directly involved in CH₄ or N₂O emissions, CH₄ oxidization and denitrification processes interact together, where NH₄⁺ has inhibitory effects on CH₄ oxidization and pmoA gene abundance can facilitate denitrification with methane as external carbon source (Figure 8). Some studies stated that the aerobic methanotrophic bacteria are particularly useful for discovering and analyzing diverse mechanisms for nitrification and denitrification processes (Stein and Klotz, 2011; Zhou et al., 2014). By testing against samples in paddy rice cropping systems (Table 3 and Figures 6, 7), the simulated models can also be applicable to predicting seasonal dynamics of CH₄ and N₂O fluxes as responses to elevated atmospheric CO₂ and rising temperature.

Bacterial genes key functional to CH₄ and N₂O fluxes identified in this study may be used as strategies for mitigating GHGs. For example, biochar application can significantly reduce both CH₄ and N₂O emissions by depressing nirK and mcrA while stimulating nosZ and pmoA genes abundances during manure composting (Sonoki et al., 2013; Wang et al., 2013). We noted that limitations on the use of qPCR for investigation of targeted genes also exist as results of PCR-bias, disturbance by DNA from dead cell, detecting only DNA copy numbers but not RNA transcriptional activity, lacking information regarding detailed community structures of specific microorganisms. The cDNA-based technologies and high throughout strategies, such as reverse transcription quantitative PCR, Illumina sequencing, and Gene Chip, will be very useful for a deeper understanding the characteristics of the functional genes and specific microbial groups, as well as their relationships with GHG emissions.

We presented the quantitative study illustrating interactions between different bacterial activities and their role in controlling CH₄ and N₂O fluxes as a response to changes in physicochemical parameters during windrow composting. This study also presented the quantitative assessment of CH₄ and N₂O fluxes based on multiple microbial gene abundances at the functional levels in composting windrow. Additional studies in this area are highly needed to extend such capabilities and allow us to quantitatively address microbial contributions to GHG fluxes from soils and manure management systems. This is particularly important, as it is widely believed that microorganisms play important roles in global carbon and nitrogen biogeochemical cycles, yet they are generally not included in current biogeochemical models for carbon and nitrogen cycles.

SqL and JZ conceived this study. SqL has the main responsibility for microbial sampling and microbial analyses for this study. SwL and QS provided valuable input for the design and data analyses of this study. LS, XG, and YJ performed qPCR analyses and gas sampling. SwL, SqL, and JZ performed the statistical analyses and wrote the paper. All authors edited and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.