Pig manure was obtained from a large-scale pig farm in Shunyi District (Beijing, China), with prior assurance that no antibiotics were applied to the sampled pigs during the breeding process. Wheat straw was collected from Shangzhuang Experimental Station of China Agricultural University and cut into 3–5 cm segments (Liu et al., 2017a,b, 2019). AMX (98%) was gained from Huamaike Biological Technology Company, Ltd. (Beijing, China). The chemicals used in HM extraction were of analytical grade.

In this study, two groups of fresh pig manure (about 5.2 kg) were fully mixed with 100 mg/kg (dry weight) AMX aqueous solution or an equivalent amount of ultrapure water (control), and then a certain amount of wheat straw and ultrapure water were mixed uniformly, according to the ratio of total carbon and total nitrogen (C/N) of 20:1 and moisture content (MC) of 65% (Liu et al., 2019). After being completely mixed and equilibrated for about 6 h (Qian et al., 2016; Liu et al., 2019), the two groups of initial compost mixture were placed in two parallel sets of in-laboratory aerobic composting reactor systems (Supplementary Figure S1), as the experimental group (AMX group) and the control group without AMX (CK group). Both groups of reactors were continuously ventilated by 0.2 L/(kg-VS-min). The basic physicochemical properties of initial composting materials have been determined and listed in Supplementary Table S1. The above settings of composting conditions and AMX concentration were based on the existing reports and our previous studies (Kotzerke et al., 2011; Liu et al., 2018, 2019).

According to the aerobic composting periods (Liu et al., 2018), about 120 g of the compost mixture was sampled on days 0, 2, 4, 6, 9, 15, and 21, and the basic physicochemical, biological, and heavy metal-related indicators were determined. Furthermore, the viable counts of AMX-resistant bacteria and total culturable bacteria (TCB) were performed immediately on the day of sampling. The presence of AMX resistance genes (bla_TEM, bla_VIM), Cu resistance gene (copA), Zn resistance gene (czrC), and intI1 were determined by real-time fluorescence quantitative PCR (qPCR) for day 0, 2, 6, 15, and 21 samples. Combined with the existing reports and our previous studies, TEM- and VIM-type β-lactam resistance genes, as the most common and highly abundant ARGs in animal breeding and clinical environments, can significantly enhance the zoonotic potential (Pomba et al., 2004). Together with copA- and czrC-type MRGs, they are often used as indicator genes of bacterial resistance contamination (Narciso-da-Rocha et al., 2014). Moreover, intI1 is widely used as an important universal marker for human pollutants (Gillings et al., 2015; Yin et al., 2017). The bacterial community of compost on days 2, 6, and 15 were analyzed according to different representative composting stages based on 16S rRNA gene sequencing. Samples used for molecular biological determinations were first lyophilized with a vacuum freeze dryer (Martin Christ Alpha 1–2 LD plus, Germany) for the same low MC, then ground with a 1-mm sieve and frozen at −80°C for DNA extraction. Other samples were kept at −20°C for physicochemical parameters analysis. The compost piles were fully mixed manually before and after each sampling.

Repeated measurements were expressed as mean ± standard deviation (SD, indicated by the error bar in the figures). Excel 2016 (Microsoft, United States), SPSS 25 (IBM, United States), and OriginProV.8.5 SR1 (OriginLab Corp., United States) were used for basic statistical analysis. Shannon-Wiener curves, principal coordinates analysis (PCoA), heatmap, correlation analysis, and genus-level phylogenetic tree were performed by R3.4.1 software. Redundancy analysis (RDA) was carried out based on CANOCO 5.0. Correlation network analysis (Spearman, r > 0.6, p < 0.05) was conducted by R3.4.1 software and Gephi0.9.3 software.

As important indicators of environmental HMs potential risks, exchangeable HMs (bio-Cu and bio-Zn) can sensitively reflect the dynamic changes of HM morphology and toxicity during composting. The concentrations of bio-Cu and bio-Zn in the two groups leveled off after first decreasing during composting, presenting a relatively consistent change trend (Figures 1A,B). The results reflected that aerobic composting treatment will reduce the migration rate and bioavailability of Cu and Zn to a certain extent, thereby effectively reducing the toxicity and potential risks of HMs in the products. Comparatively, the concentrations of bio-Cu and bio-Zn in the AMX group were slightly higher than those in the CK group during the early stage of composting. This difference indicated that the addition of AMX likely hindered the passivation process of HMs and, thus, increased the risks of HMs contamination and the co-selection pressure of inducing bacterial resistance, which was also confirmed by previous reports (Song et al., 2017; Vos et al., 2019).

We found that the abundance of copA and czrC in the AMX group was higher at the initial stage of composting (Figures 1C,D), especially the difference between the two groups of czrC was the most significant (p < 0.05, t-test), further illustrating the vital role of AMX selective pressure on the promotion of HM resistance and adaptation cost of MRGs-carrying bacteria (Imran et al., 2019; Zhang et al., 2020). Notably, AMX may have a stronger promotion effect on some Zn resistome. In the late stage of composting, the abundance of copA in the two groups increased rapidly then fell, while czrC of the two groups was very low and stable, probably because copA-carrying bacteria mainly dominated the composting maturity stage and were strongly affected by microbial succession at this stage. According to Figure 1, AMX selective pressure may significantly inhibit the passivation of HMs while promoting the development of some MRGs, thereby raising the potential risk of HM toxicity and bacterial resistance in the environment.

A total of 397 operational taxonomic units (OTUs) were detected in compost samples. According to the Shannon-Wiener curves of the compost samples (Supplementary Figure S3), the sequencing data were sufficient to reflect the vast majority of microbial information in all samples. After leveling the minimum sequence number of the samples, alpha diversity analysis was performed on the two groups of compost samples (Supplementary Figure S4). The three diversity indices of the AMX group were higher than those of the CK group as a whole, confirming that the bacterial community of the AMX group was relatively rich and uniform with higher community diversity (Supplementary Figures S4A–C). Moreover, there was a greater difference in species lineage among samples in the CK group when considering species abundance and evolutionary distance (Supplementary Figure S4D). In summary, although AMX selective pressure reduced the absolute abundances of culturable bacteria (Supplementary Figure S2B), the richness and evenness of the bacterial community were relatively high (Supplementary Figure S4), which may indicate the effect of AMX on the alteration of microbial community structure during composting and the selective pressure on AMX-sensitive and -resistant bacteria. According to the PCoA, PC1 and PC2 accounted for 85.6% of the total variation of the bacterial community (Figure 4A). Principal coordinates analysis of the bacterial community revealed that the samples of different composting periods were clearly separated along the PCo1 axis, and some clustering of composting samples from the same period was also observed. The difference in bacterial community composition between the two groups decreased gradually as the number of composting days increased, illustrating that the impact of the composting periods and habitat on the bacterial community structure was particularly significant (Yin et al., 2017). Furthermore, AMX addition significantly affected the bacterial community structure in the early stage of composting, which is consistent with the above results (Supplementary Figure S4).

It can be seen from Figures 4B,C that the bacterial community of the two groups continued to change during composting, while the differences in community composition gradually decreased. Among the communities, Firmicutes, Proteobacteria, Actinobacteria, and Bacteroidetes were the dominant phylum bacteria during the whole composting process (Figure 4B), accounting for 95.2–99.9% of the bacterial 16S rRNA gene sequences (Supplementary Figure S5). As shown in Figure 4B, the relative abundance (RA) of Firmicutes was the highest in the early stage of composting and the RA in the AMX group was lower than that in the CK group, but decreased rapidly as the compost matured. This may be since Firmicutes, beneficial bacteria for promoting cellulose degradation, widely exist in animal intestines, compost, and soil environment, and dominate the heating and thermophilic phases of composting (Yin et al., 2017; Liu et al., 2019). Moreover, high concentrations of antibiotics will significantly inhibit their growth and reproduction (Qian et al., 2016; Guo et al., 2017; Liu et al., 2019), resulting in a significant reduction of Firmicutes RA in the AMX group. According to Figure 4B, RA of Bacteroides increased significantly in the mature composting stage of the two groups and was relatively higher in the CK group. In Figure 4C, it can be seen that Sphingobacteriia, Flavobacteriia, and Bacteroidia (which belong to Bacteroides) also displayed the same regularities and cluster. Bacteroides are a kind of bacteria attributed to the degradation of high-molecular-weight compounds and the growth of supporting materials, which mainly grow in the middle and late stages of composting (Zhu et al., 2017b; Liu et al., 2019). The RA of Proteobacteria and Actinobacteria (most notably in Proteobacteria) in the AMX group increased with composting maturity and was significantly higher than that in the CK group during the early stage of composting (Figure 4B). This large difference may be attributed to the dominance of both bacteria in the cooling and mature stages of composting, where the abundance of Actinobacteria is mostly used to mark the degree of maturity (Liu et al., 2019; Zhang et al., 2020). Previous studies have shown that both bacteria have a significant positive correlation with ARGs (Yin et al., 2017), and their RA will also be significantly increased in the presence of high concentrations of antibiotics (Guo et al., 2017). Besides, the driving force of the increased resistance in the compost was mainly attributed to Proteobacteria (Meng et al., 2015), in which γ-Proteobacteria has a strong resistance potential to high-concentration antibiotics (Zhu et al., 2017b). Moreover, β-Proteobacteria is considered to be the original source of intI1, which has become the core of antibiotic resistance due to its ability to capture and express multiple ARGs (Guo et al., 2017; Zhu et al., 2017b). The above conclusions were also further confirmed in Figure 4C, where the variation rules of α, β, γ-Proteobacteria, and Actinobacteria were similar and clustered with each other. Therefore, the high concentration of AMX added to the experimental group can selectively inhibit the growth and reproduction of sensitive bacteria, such as Firmicutes. This provided a more favorable living environment for drug-resistant Proteobacteria and Actinobacteria, thus affecting and changing the bacterial community structure and drug resistance level during composting.

LEfSe analysis (threshold set to 2) was used to further reveal the differences between the two groups of bacterial community during composting (Figure 4D). Compared with the CK group, the species with significantly different abundance in the AMX group were Hahellaceae, Hahella, Enteractinococcus, and Marmoricola (Supplementary Figure S6), where the former two species belong to γ-Proteobacteria and the latter two to Actinobacteria. Proteobacteria is the main host of ARGs and has a high tolerance to antibiotics, while Actinobacteria is the main producer of antibiotics and usually carries a variety of ARGs (Qian et al., 2016; Zhang et al., 2020). Therefore, this finding was consistent with the relevant conclusions in Figures 4B,C and previous literature reports, further confirming the significant role of AMX in reshaping the structure of environmental microbial communities and promoting bacterial resistance. Studies have shown that Enteractinococcus is a type of Micrococcaceae commonly found in animal feces (Cao et al., 2012), while Marmoricola is the gram-positive aerobic bacteria belonging to the Nocardiaceae (Habib et al., 2020). Both of them have been reported to exist in the aerobic composting environment of livestock feces as the main Actinobacteria (Chen et al., 2015).

Redundancy analysis was carried out to further reveal the correlation between the bacterial community structure and its function and environmental factors during composting, which can explain 86.8% of the total species variation (Figure 5A). Differences in the bacterial community structure gradually reduced with the extension of composting time, which intuitively reflected the severe succession and stable formation of the bacterial community structure during composting. A noteworthy correlation was observed between the thermophilic period of composting and the dynamic changes in the basic physiochemical indicators, exchangeable HMs, β-lactam resistance genes, czrC gene, and Firmicutes. Proteobacteria, Actinobacteria, and Bacteroidetes were mainly distributed in the composting maturation stage and were directly related to the dynamic changes of copA gene and intI1 gene. This correlation demonstrated the significant impact of the composting stages (especially the thermophilic period) and bacterial community structure on the composting microenvironment, HM forms, and resistance genes. Furthermore, it can also be suggested that Proteobacteria, Actinobacteria, and Bacteroidetes were likely the main factors directly affecting the development and spread of HM resistance and antibiotic resistance during composting.

To further verify the interaction pattern between the microbial communities in the two groups of different composting environments, the relationships among class-level bacteria, ARGs, MRGs, Bio-HMs, and intI1 were revealed by network analysis (Figure 5B). More positive edges (red, 7) than negative edges (green, 3) were detected in a network of the compost. Seven different class-level bacteria (except for unidentified) had a significant positive correlation with each other (p < 0.05), which was consistent with the results of the RDA analysis. Bacilli (belonging to Firmicutes) and Cytophagia (belonging to Bacteroidetes) were significantly negatively correlated (p < 0.05), which were the dominant bacterial communities in the early composting period and the late composting periods, respectively, (Figure 4B). S0134_terrestrial_group, which belongs to the Gemmatimonadetes, was significantly negatively correlated with bla_TEM gene and czrC gene (p < 0.05). Combined with Figure 4B, Gemmatimonadetes were mainly detected in the middle and late stages of compost samples (AMX6, AMX15, and CK15), and probably play an important role in the reduction of bla_TEM gene and czrC gene. Therefore, it is recommended in future studies to further explore the role of Gemmatimonadetes in removing other ARGs and MRGs during aerobic composting, especially β-lactams and zinc.

Moreover, to identify environmental drivers in our study (Sunagawa et al., 2015), we correlated distance-corrected dissimilarities of taxonomic and functional community composition with those of compost environmental factors (Figure 5C). Overall, MC and NO₃⁻-N were the strongest correlates of both taxonomic and functional composition during composting, while no significant correlations were found for OM and Bio-Zn. Bio-Cu, pH, and NH₄⁺-N were only weakly correlated with functional community, as well as C/N with taxonomic composition, and the other correlations were not statistically significant (Figure 5C). It can be seen that, except for the basic physicochemical (MC, C/N, and pH) and nutritional indicators (NO₃⁻-N, NH₄⁺-N), Bio-Cu may be an important environmental driver affecting taxonomic and functional community composition in composting environments, as well as an important potential factor regulating bacterial resistance.