Fresh pig manure was procured from the Changping Test Base at the Institute of Animal Sciences, Chinese Academy of Agricultural Sciences, located in Beijing, China. N-Acetyl-D-glucosamine and chitin were produced and supplied by Wuhan Lanabai Pharmaceutical Chemical Co., Ltd. Cornstalks were utilized to adjust the carbon-to-nitrogen ratio (C/N), sourced from Shangzhuang Experimental Station at China Agricultural University, Beijing, China. They underwent air-drying and were subsequent meticulously sieved through 2 mm sieves. Supplementary Table S1 displayed the basic properties of the raw materials for composting. Composting experiments were conducted in the reactors described by Chen et al. (2022). Pig manure was mixed with cornstalks in a ratio of 5:1 (wet weight basis). The initial C/N of the composting process was approximately 25, and the initial wet weight of each pile was 20 kg. Three groups were established: two experimental groups (PMNG and PMCH) and a control group (PM). In the PMNG group, materials were uniformly blended with 5% N-Acetyl-D-glucosamine, while in the PMCH group, materials were mixed with 5% chitin; the PM group did not receive any additional treatments. Each treatment had three replicates. The initial moisture content of the piles was adjusted to approximately 60%, and no additional water would be added throughout the composting process. Throughout composting, air was supplied from the bottom with an average airflow rate of approximately 0.2 L kg⁻¹ min⁻¹. To maintain a consistent oxygen supply, the piles were turned over at each sampling time. Sampling was conducted on days 0, 5, 12, 19, 26, 33, 40, and 47, with five subsamples collected from different locations within the pile profile and homogeneously mixed each time. Approximately 400 g of samples were collected from each pile. A portion of the samples was stored at 4°C for subsequent physical–chemical analyses, while the remaining samples were stored at −20°C for molecular extraction.

Composting temperature was continuously monitored and recorded using a digital thermometer. Fresh samples were extracted with deionized water (DW) at a solid: liquid ratio of 1:10 (w/v, dry mass basis). Subsequently, pH and electrical conductivity (EC) were measured using a digital pH meter and an electrical conductivity meter (MP521, Sanxin, China). Total carbon and total nitrogen were quantified using a Vario MACRO cube elemental analyzer (Elementar Analysensystem, Germany), as described in Chen et al. (2023). The germination index (GI) was determined in accordance with the Chinese Standard for Organic Fertilizers (NY/T 525–2021). Total organic carbon (TOC) in the samples was analyzed using an organic carbon analyzer (TOC-Vcp, Shimadzu).

Total genomic DNA was extracted using the FastDNA SPIN Kit for Soil (MP Biomedicals, Solon, OH, United States) following the manufacturer’s instructions. The purity and concentration of DNA for each sample were measured using a NanoDrop2000 (Thermo Fisher Scientific, Waltham, MA, United States). DNA integrity was verified through 1% agarose gel electrophoresis. The V3–V4 region of the 16S rRNA gene was amplified using primers 341F (F: 5′-CCTAYGGGRBGCASCAG-3′) and 806R (R: 5′-GGACTACNNGGGTATCTAAT-3′). Sequencing was performed on the Illumina MiSeq PE250 platform (Novogene Co., Ltd., Beijing, China). Raw 16S rRNA gene sequences were processed using QIIME2 (Quantitative Insights into Microbial Ecology, version 2022.2) as described by Bolyen et al. (2019). Reads were trimmed based on the Q30 minimum value and denoised using the Deblur method. All amplicon sequence variants were aligned with mafft, and a phylogenetic tree was constructed using fasttree2.

Total genomic DNA for detecting ARGs, MGEs, and 16S rRNA was extracted using the FastDNA Spin Kit for Soil (MP Biomedicals) following the manufacturer’s instructions. Forty-nine ARGs for aminoglycoside (Aac6-Aph2, aac(3)-xa, aac(6′)-Ib(aka aacA4)-01, aac(6′)-II, aac(6′)-Ib, aacA/aphD, aadD, aadE, aphA1, strB), miltidrug (acrA, acrB, arsA, mdtA, qnrD, qnrS2), macrolide-lincosamide-streptogramin B (MLSB) (ere(A), erm(A), lmrA, lnuA, mefA, mphA, msr(A), vat(A), vgaA), sulfonamide (dfrA1, folA, sul1, sul2), tetracycline (tetA, tetB, tetC, tetD, tetE, tetH, tetJ, tetK, tetM, tetO, tetR, tetS, tetT, tetW, tetX), vancomycin (vanA, vanB, vanC) and trimethoprim (dfrBmulti, dfrC), eight MGEs, including integrase (intl1, intl2, intl3), plasmid (tra-A), Insertional_sequence (IS3, IS26, ISCR1) and transposase (tnpA-1) and one 16S rRNA were quantified using the WaferGen SmartChip Real-Time qPCR System (WaferGen Bio-systems, Fremont, CA, USA) with primers sourced from Anhui Microanaly Genetech Co., Ltd. The qPCR reactions were conducted in a 100 nL reaction system comprising 1x LightCycler 480 SYBR Green I Master Mix, 500 nM of each primer, and 2 ng μL⁻¹ DNA template. The reaction mixtures were loaded into a SmartChip Cycler (WaferGen Biosystems) using a multi-sample dispenser (WaferGen Biosystems) following the 144 (assays) × 36 (samples) model. Each sample was measured in triplicate. The reaction protocol involved pre-denaturation at 95°C for 10 min, denaturation at 95°C for 30 s, and annealing at 60°C for 30 s. The qPCR results were automatically analyzed using qPCR software, with CT = 31 set as the detection threshold. No detections in any of the three replicates were considered negative results.

The gene copy numbers of ARGs, MGEs, and 16S rRNA were determined using the formula: gene copy number = 10^{(31 − CT) × (10/3)}. The relative abundance of ARGs or MGEs was calculated as the ratio of the gene copy number of ARGs or MGEs to the gene copy number of 16S rRNA. Quantification of the 16S rRNA genes was performed using a QuantStudio 6 Flex Real-Time PCR analyzer (Applied Biosystems, Waltham, MA, USA). The qPCR reaction system for 16S rRNA genes consisted of 10 μL, comprising 5 μL 2x Mix, 0.5 μL forward primer, 0.5 μL reverse primer, 0.2 μL 5x ROX dye, 2.8 μL nucleotide-free water, and 1 μL DNA template. The reaction conditions involved enzyme activation at 95°C for 10 min, followed by 40 cycles of denaturation at 95°C for 15 s, and annealing at 60°C for 1 min. Each PCR reaction was performed in triplicate for technical consistency (Chen et al., 2024).

The physicochemical data were visualized using Origin Pro 2022. Heatmaps were generated in R Version 4.1.3 using the pheatmap package. Alpha diversity estimates, including Shannon diversity, Pielou’s evenness, and Chao1 richness, were calculated using the Vegan package. Co-occurrence network analysis between bacteria and ARGs/MGEs was conducted using the psych package with the Spearman correlation coefficient (r ≥ 0.96, p ≤ 0.01) and visualized using Gephi 0.9.6. Redundancy Analysis (RDA) was performed using Canoco 5 for Windows to identify the key physicochemical factors influencing the variations in ARGs.

The temperature was consistently maintained above 50°C for more than 6 days in all three treatments (Supplementary Figure S1), effectively eliminating the majority of pathogens and weed seeds. The temperature changes in the PM and PMNG treatments exhibited similar patterns, while in the PMCH treatment, a distinct lagged thermophilic phase was observed. With the progression of the composting process, the gradual decomposition of chitin facilitated the transition of PMCH into the thermophilic phase. After day 38, all piles reached the maturation phase, and the temperatures stabilized. On the 12th day of composting, the pH value in the PMCH treatment was 7.0, significantly lower than the other two treatments. As composting progressed, the pH values in all treatments increased to a range of 7.8–8.3 (Supplementary Figure S1). The EC values in the PM and PMNG treatments showed a declining trend, whereas in the PMCH treatment, the EC value initially increased and then decreased. By the end of composting, the EC values in all treatments were below 4 mS/cm (Supplementary Figure S1). The C/N ratios in PM, PMNG, and PMCH decreased, respectively, to 13.4, 11.2, and 13.3 at the end of composting. The GI for all three piles exceeded 90%, indicating the maturation of the composting process (Supplementary Figure S1). Composting is a process of organic mineralization and humification, and the TOC content in all three treatments exhibited a noticeable decrease during composting (Supplementary Figure S1), suggesting the microbial mineralization and transformation of organic matter.

In this study, 57 common ARGs and MGEs subtypes were quantified, encompassing resistance to aminoglycoside (10), multidrug (6), MLSB (9), sulfonamide (4), tetracycline (15), vancomycin (3), trimethoprim (2) and MGEs (8) (Figure 1). The resistance mechanisms of ARGs were assessed. Efflux pumps and antibiotic deactivation were identified as dominant mechanisms in all samples, both in terms of detected gene numbers and gene abundances (Figure 2).

At the initial stage of composting, the relative abundance of total ARGs and MGEs per 16S rRNA in PM, PMNG, and PMCH ranged from 4.7 × 10⁻¹ to 6.0 × 10⁻¹ and 1.1 × 10⁻¹ to 3.4 × 10⁻¹, respectively (Figure 1I). Throughout the composting process, substantial differences were observed among the three treatments in the dynamics of ARGs and MGEs (Figure 1I). Specifically, in the PM and PMNG treatments, the total abundance of ARGs and MGEs reached its minimum on day 12, while in the PMCH treatment, the minimum abundance occurred on day 47 (Figure 1I). During the thermophilic phase (from day 0 to day 12), the removal efficiencies of total ARGs and MGEs in PM, PMNG, and PMCH were 91.4% for PM, 96.0% for PMNG, and 41.9% for PMCH; subsequently, the efficiencies were 96.2, 98.1, and 68.1%, respectively (Supplementary Figure S2). At the same time, the removal efficiency of ARGs and MGEs in the PMCH treatment was significantly lower than in the PM and PMNG treatments. The delayed onset of the thermophilic phase in PMCH resulted in lower removal rates of total ARGs and MGEs on the twelfth day. Additionally, the PM treatment exhibited varying degrees of enrichment for vancomycin-resistant genes, while the PMCH treatment showed enrichment for vancomycin, tetracycline, and MLSB resistance genes (Supplementary Figure S2). After 26 days of composting in the cooling phase, the removal rates of total ARGs and MGEs in PM, PMNG, and PMCH decreased to 17.7, 5.3, 7.3, and 50.7%, 23.7, 36.5%, respectively (Supplementary Figure S2). Simultaneously, all three treatments exhibited enrichment for tetracycline, sulfonamide, and multidrug resistance genes, while PM and PMCH treatments continued to show enrichment for vancomycin-resistant genes. At the end of composting, the removal rates of total ARGs in PM, PMNG, and PMCH were 40.2, −59.9%, and 54.5%, respectively (Supplementary Figure S2). For MGEs, the removal rates in PM, PMNG, and PMCH were 62.1, −135.4%, and 78.5%, respectively. The addition of chitin significantly enhanced the removal rates of total ARGs and MGEs in composting. It is noteworthy that PMNG significantly enriched ARGs and MGEs, primarily including tetracycline (tetA, tetD, tetE, tetK, tetR, tetX), sulfonamide (dfrA1, sul1, sul2), and MGEs (intl1, intl2, intl3, IS3, ISCR1) (Figures 1, 2). PM and PMCH also exhibited varying degrees of enrichment for tetracycline (tetA, tetD, tetR, tetX) and sulfonamide (sul1, sul2) (Figures 1, 2).

Shannon diversity, Pielou’s evenness, and Chao1 richness were employed to characterize the α-diversity of bacterial communities. At the end of composting, the PMNG treatment exhibited significantly higher Shannon diversity and Pielou’s evenness compared to the other two treatments, indicating that the addition of N-Acetyl-D-glucosamine significantly enhances the α-diversity of bacteria, providing a more diverse bacterial hosts for ARGs and MGEs, leading to gene enrichment (Figure 3A). Moreover, the dynamics of bacterial communities exhibited consistency with changes in ARGs and MGEs (Figure 3), suggesting a strong positive correlation between bacterial communities and ARGs/MGEs. Principal Coordinate Analysis (PCoA) based on the Bray-Curtis distance matrix revealed that bacterial communities in all samples during composting could be categorized into three groups (Figure 3B). However, on day 47, the bacterial communities in PM and PMCH tended to be similar. This result could explain the trends in the changes of ARGs and MGEs (Figure 1). At the phylum level, significant differences in bacterial composition were observed among PM, PMNG, and PMCH during the composting process (Figure 3C). Firmicutes, Proteobacteria, Gemmatimonadota, and Myxococcota were predominant phyla shared by all three treatments, constituting over 95% of all phyla. At the end of composting, the abundance of Firmicutes in PMNG was significantly higher than in the other two treatments. This provides a reasonable explanation for the enrichment of ARGs in the PMNG treatment and the eventual rebound in ARGs removal rates in the other treatments. The hierarchical clustering heatmap illustrates the dynamics of the top 30 bacterial communities at the genus level, standardized using the log2 function (Figure 3D). Bacterial communities across all treatments clustered into three groups, with higher abundances in the group comprising Longimicrobiaceae (Gemmatimonadota), S0134_terrestrial_group (Gemmatimonadota), Sinibacillus (Firmicutes), Bacillus (Firmicutes), Pseudogracilibacillus (Firmicutes), and Gracilibacillus (Firmicutes). It can be inferred that on the 12th day of composting, the high abundance of Bacillus (3.4–30.0%) across all treatments contributed to a significant reduction in the levels of ARGs and MGEs in composting (Figure 3D).

The co-occurrence networks among ARGs, MGEs, and microbial communities (OTUs) for the three treatments are depicted in Figure 4. The network exhibits nodes (110, 89, and 62) and edges (142, 119, and 72) in PM, PMNG, and PMCH, respectively (Supplementary Table S2). Across PM (8), PMNG (5), and PMCH (6) bacterial phyla, the network identifies potential hosts for 22 ARGs and 8 MGEs, suggesting their crucial interactions. The predominant bacterial phyla associated with ARGs and MGEs are Proteobacteria, Firmicutes, Gemmatimonadota, and Myxococcota, highlighting their significance in the transformation and dissemination of ARGs and MGEs. Network analysis also allows for the identification of the most prevalent genes in different treatments. Noteworthy ARGs across all treatments include vanA, IS3, and intl3 in PM; acrB, arsA, and tra-A in PMNG; and vat(A) and qnrS2 in PMCH. IS3 is identified as the most abundant host across all three treatments (Figure 4). Therefore, specific attention should be directed toward subtypes vanA, IS3, intl3, acrB, arsA, tra-A, vat(A), and qnrS2 during the composting of pig manure.

Based on redundancy analysis, we investigated the relationships among physicochemical factors, MGEs, and ARGs (Figure 5). The analysis revealed that the carbon-to-nitrogen ratio (C/N) emerged as a crucial factor influencing ARGs and MGEs across the three treatments (p < 0.05). In other words, C/N ratio was identified as a key variable shaping antibiotic resistance during the composting process.