Integrating livestock and aquatic plant towards mitigating antibiotic resistance transmission from swine wastewater

Introduction

The expeditive spread of antibiotic resistance genes (ARGs) has led to the widespread dissemination of bacterial resistance, posing a pressing challenge to global health and sustainable development. It is alarming that nearly 5 million people died from infections caused by bacterial antimicrobial resistance in 2019, making it the third leading cause of death worldwide¹. The World Bank estimates that antimicrobial resistance would result in losses of 1 trillion to 3.4 trillion USD in gross domestic product per year by 2030². From a “One Health” perspective, livestock and poultry farms are closely associated with the global prevalence of antimicrobial resistance and have emerged as hotspots for the dissemination of ARGs^3,4,5. Multiple studies have highlighted that livestock wastewater harbors a diverse array of ARGs (over 300 subtypes), with absolute levels ranging from 1.0 × 10⁶ to 1.0 × 10¹² copies mL⁻¹ ^6,7. Notably, the discharge or irrigation of livestock wastewater directly elevates the diversity and burden of ARGs in the receiving environment^8,9. Furthermore, these wastewater-borne ARGs could be transferred into pathogenic bacteria, potentially increasing their environmental exposure risk to both humans and animals¹⁰. Such notorious hazards concerning ARGs in livestock wastewater have underscored the imperative action to mitigate risks.

Mitigating the load of ARGs is essential for controlling the transmission risk posed by livestock wastewater^11,12. Anaerobic digestion, a prevalent technology in treating such wastewater, effectively eliminates traditional organic compounds and demonstrates potential in reducing ARG abundance^13,14. However, this process can sometimes increase the abundance of specific ARGs by enriching potential ARG hosts, such as denitrifying bacteria^15,16. Advanced oxidation processes, including Fenton and Fenton-like processes, ozonation, photocatalysis, and persulfate oxidation, are gaining popularity for ARG elimination due to their ability to damage cell surfaces and DNA structures through free radical reactions¹⁷. Despite their potential in removing ARGs from livestock wastewater, these processes are limited by the energy-intensive nature, high implementation costs in treatment facilities, and the generation of harmful byproducts¹⁸. Recently, artificial wetlands have gained attention as an effective ecological engineering solution¹⁹. Studies have shown that the absolute abundance of ARGs in pig wastewater significantly decreases after treatment in constructed wetlands, yet their relative abundance remains higher than in the influent²⁰. Given that selective pressures from antibiotics and heavy metals can promote the proliferation and transfer of ARGs, artificial wetlands are also regarded as potential hotspots for the secondary transmission of ARGs²¹. Consequently, it is still lacking for green, effective, and economical technical means to reduce ARGs in livestock wastewater within agricultural production.

Phytoremediation has emerged as a cost-effective, eco-friendly, and sustainable approach for treating livestock wastewater, leveraging the extensive root and foliage systems of aquatic plants²². These plants not only thrive on the high contents of organic matter, ammonia nitrogen, nitrate nitrogen, phosphorus, and other nutrients present in livestock wastewater, but also remove pollutants such as antibiotics residues and heavy metals through uptake and accumulation^23,24. Currently, the integration of aquatic plants and livestock production systems has become a productive and sustainable strategy in China, addressing both resource recycling and pollution control in livestock wastewater^25,26,27. While the role of aquatic plant in reducing ARGs has been noted in constructed wetlands²⁸, the specific mechanism by which the plant reduce ARGs in wastewater under crop-livestock integration systems remains elusive.

To address this knowledge gap regarding the reduction of ARGs in swine wastewater, Myriophyllum elatinoides was selected as a candidate aquatic plant for its potential to mitigate ARGs. Firstly, we assessed the reduction in ARG abundance in the field-scale ecological ponds planted with M. elatinoides and explored the underlying microbial mechanisms, focusing on shifts in microbial community composition and the role of mobile genetic elements (MGEs). Secondly, a controlled laboratory hydroponic experiment was established to verify the ability of M. elatinoides to remove ARGs and antibiotic-resistant bacteria (ARB) from livestock wastewater and to trace the migration of ARGs and ARB within the wastewater-M. elatinoides system. Additionally, the post-treatment transmission risk of ARGs in livestock wastewater was evaluated using simulated environment-receiving microcosms. We hypothesized that (1) ARG abundance can be significantly reduced by M. elatinoides through direct uptake and accumulation; and (2) the proliferation and transfer of ARGs in swine wastewater may be indirectly curbed by alleviating the selective pressures from residual antibiotics.

Results

Attenuation of ARG diversity and abundance in wastewater at field-scale ponds

During the operation of the swine farm, we examined the dynamics of ARGs in water and sediment samples collected from the five sequential ecological ponds (Fig. 1a). Metagenomic analysis revealed a significant reduction in the number of ARG subtypes in the water samples, from 234 in P0 to 80 in P4, with a corresponding decrease in total relative abundance from 0.52 to 0.08 copies per 16S rRNA gene (Fig. 1b). In contrast, sediment samples showed no significant changes across the five ecological ponds. These findings were corroborated by principal coordinates analysis (PCoA) results (Fig. 1c). Notably, substantial decreases were observed in water samples for ARGs associated with resistance to tetracycline, sulfonamide, aminoglycoside, macrolide-lincomycin-streptomycin, chloramphenicol, and beta-lactam, with attenuation rates surpassing 85% (Fig. 1d). At the subtype level, the relative abundance of predominant ARGs (sul1, sul2, aadA, tetM, tetC, ermF, etc.) dropped from 10⁻² copies per 16S rRNA gene in P0 to 10⁻³ copies per 16S rRNA gene in P4 or were undetectable, with reduction rates exceeding 85% and even reaching 100% (Supplementary Fig. 1). To validate these changes, several dominant ARG subtypes were selected for absolute abundance quantification using qPCR method. Compared to P1, the absolute abundance of these representative dominant ARGs exhibited a significant decreasing trend by orders of magnitude (Supplementary Fig. 2). Overall, these results suggest that M. elatinoides-planting ponds have a great potentiality in mitigating ARGs derived from swine breeding wastewater.

Integrating livestock and aquatic plant towards mitigating antibiotic resistance transmission from swine wastewater — **Fig. 1: Mitigation of ARGs in the wastewater at field planting ponds.**

Potential microbial mechanisms underlying ARG attenuation in field-scale ponds

To explore the microbial cues that potentially linked with ARG attenuation in the ecological ponds, we focused on the compositions of bacterial communities and MGEs, which are key in the vertical and horizontal gene transfer of ARGs, respectively (Fig. 2). The Shannon and Evenness indices of the water bacterial community in ponds P3 and P4 were significantly higher than that in P0 and P1, suggesting an increase in alpha-diversity as the water passed through the system (Supplementary Fig. 3). This trend was supported by PCoA analysis with ADONIS (Water R² = 0.94, p < 0.001; Sediment: R² = 0.94, p < 0.001) and ANOSIM analysis (Water R = 1, p < 0.001; Sediment: R = 1, p < 0.001) (Supplementary Fig. 4). The relative abundance of dominant phyla Proteobacteria, Firmicutes, Bacteroidetes, and Elusimicrobia in water samples decreased by 15.42%, 19.65%, 10.84%, and 38.93%, respectively; while Cyanobacteria, Actinobacteria, and Acidobacteria increased by 3.10-, 2.44-, and 1.41-fold, respectively (Fig. 2a). No obvious changes were observed in sediment samples across different ponds. At the genus level, Meanwhile, the relative abundance of dominant bacterial genera such as Serratia, Xanthomonas, and Dechloromonas decreased by 76.92%, 77.19%, and 83.08%, respectively, in water samples across ponds; while no significant changes in sediment samples (Supplementary Fig. 5). Furthermore, we characterized the changes in MGEs across the five ponds. The diversity and abundance of MGEs in water samples showed a downward trend, with reductions of 37.62% and 56.72% in P4 compared to P0, while sediment samples exhibited a slight variation (15.87% and 35.36%) (Fig. 2b and Supplementary Fig. 6). Transposase was the predominant MGE type, followed by insertion element, integrase, and plasmid, with their abundances in water samples decreasing by 57.82%, 44.91%, 76.15%, and 43.61%, respectively, across the ponds (Fig. 2b). In sediment samples, the abundance of transposase and insert sequence decreased, while integrase and plasmid increased. The majority of dominant MGE subtypes showed a decreasing trend in water samples, with smaller or non-significant changes in sediments (Supplementary Fig. 7).

**Fig. 2: Potential microbial mechanisms underlying the attenuation of ARGs at field planting ponds.**

Considering the variations of bacterial community and MGEs in water and sediment samples across the five ponds, their associations with ARGs were further explored with Procrustes analysis, Pearson analysis, and variance partitioning analysis (VPA). Procrustes analysis showed a significant correlation between ARG content and bacterial community in both water (M² = 0.0637, p < 0.001) and sediment (M² = 0.4385, p < 0.001) samples (Fig. 2c), indicating that bacterial community composition is an important determinant of ARGs, with a more pronounced role in water samples. Pearson correlation between the abundances of MGEs and ARGs revealed significant relationships in water samples for transposase (R² = 0.74, p < 0.001) and integrase (R² = 0.66, p < 0.001), but not in sediment samples (Fig. 2d). VPA indicated that approximately 70% of the observed variation in ARGs in water samples was explained by bacterial community composition and MGEs, while this proportion was less than 30% in sediment samples (Fig. 2e, f).

A total of 101 non-redundant high-quality ARG-carrying metagenome-assembled genomes (MAGs) were identified from genome survey (Fig. 3a). Proteobacteria was the predominant host of ARGs (50.50%), followed by Actinobacteriota (9.90%), Desulfobacterota (8.91%), and Firmicutes (7.92%). W-P0-1.bin.16 (Aeromonas caviae) harbored the highest number of ARGs, including multidrug (n = 5), tetracycline (n = 2), peptide (n = 2), rifamycin (n = 2), aminoglycoside (n = 1), and others (n = 3). The number of ARG-carrying MAGs in P3 and P4 decreased by 29.20% and 28.80% (p < 0.05), respectively, compared to P0, with corresponding abundances decreasing by 65.28% and 67.83% (p < 0.05), respectively (Fig. 3b). These results indicate a significant reduction in ARG-carrying bacteria during the treatment processes in the M. elatinoides-planted ponds.

**Fig. 3: Distribution and genome survey of ARGs in metagenome-assembled genomes.**

Elimination of antibiotics, ARGs, and ARB in wastewater planted with M. elatinoides

To validate the efficacy of M. elatinoides in eliminating ARGs from wastewater, a hydroponic system was established to quantify residual antibiotic concentrations, the abundance of selected ARGs, and the amount of typical ARB, respectively, by comparing treated samples (T) with control samples (CK) (Fig. 4a). The degradation dynamics showed that the seven antibiotics including sulfamethoxazole (SMX), sulfadimidine (SMD), tetracycline (TET), chloramphenicol (CHL), tylosin (TYL), ciprofloxacin (CIP), and norfloxacin (NOR) degraded more rapidly in T than CK over the 12-d treatment period (Supplementary Fig. 8). By 12 d, the residual concentrations of these antibiotics in T were significantly lower than that in CK, with reduction percentages ranging from 64.82% to 87.83% (Fig. 4b). Moreover, the absolute abundance of 16S rRNA gene and the selected ARGs (tetA, tetC, tetG, tetM, tetW, sul1, sul2, intI1, aadA, and ermF) decreased by 93.63−99.82% in T over the treatment period (Supplementary Fig. 9). At 12 d, the absolute abundance of the seven ARGs in T was significantly lower than that in CK by 75.39−94.28% (p < 0.05) (Fig. 4c). Relative to CK, the relative abundance of the seven ARGs in T decreased by 74.06−94.29% (p < 0.05) at 12 d (Supplementary Fig. 10). Meanwhile, the potential ARB conferring resistance to the seven antibiotics such as SMX, TET, CIP, CHL, ampicillin (AMP), kanamycin (KAN) and sulfadiazine (SDZ), were enumerated. The amount of ARB on various antibiotic-containing plates in T was notably lower than that in CK at 12 d (Fig. 4d). Statistical analysis demonstrated that the count of the seven ARB in T decreased by 66.27−98.39% (p < 0.05) compared with CK (Fig. 4e). In addition, the individual ARB on each plate were randomly selected for the taxonomic annotation based on the 16S rRNA gene sequences. By 12 d, 27 ARB out of 29 isolates in CK were successfully identified as 14 species, while 13 ARB out of 16 isolates in T were identified to 12 species (Fig. 4f), suggesting that M. elatinoides planting could decrease the diversity of ARB in wastewater.

Fig. 4: Elimination of antibiotics, ARGs, and ARB in wastewater treated by *M. elatinoides.*

Migration of ARGs and ARB from wastewater to M. elatinoides

To elucidate the underlying mechanisms by which M. elatinoides eliminate ARGs from wastewater, the migration of ARGs and ARB from wastewater to plant were characterized. In root samples, the absolute abundance of 16S rRNA genes remained stable between 12 d and 0 d, whereas the abundance of specific ARGs (tetC, tetG, tetW, sul1, intI1, aadA, and ermF) increased significantly by 12 d, with fold changes ranging from 2.01 to 22.93 compared to 0 d (Fig. 5a). In shoot samples, the absolute abundance of the 16S rRNA gene and the seven ARGs at 12 d was 4.25 to 327.93 folds higher than at 0 d (Fig. 5b). Relative abundances of these ARGs in root and shoot samples at 12 d were 1.24 to 25.02 and 1.49 to 19.31 times greater than at 0 d, respectively (Supplementary Fig. 11, Supplementary Fig. 12). Furthermore, the presence of ARB resistant to six antibiotics (SMX, TET, CIP, KAN, CHL, and AMP) in root and shoot samples was significantly higher at 12 d compared to 0 d (Fig. 5c, d). Statistical analysis suggested that the amounts of ARB in root and shoot samples at 12 d were 1.58−13.33 and 2.51−15.62 folds that at 0 d, respectively (Fig. 5e, f). Additionally, taxonomic identification results showed that the ARB isolated from the root samples were affiliated with 16 and 21 species at 0 d and 12 d, respectively (Supplementary Fig. 13). The corresponding taxonomic annotation results in the shoot samples were 11 and 20 species at 0 d and 12 d, respectively (Supplementary Fig. 14).

Using confocal microscopy, we confirmed the migration of antibiotic resistance from wastewater to M. elatinoides using EGFP-labeled ARB in a simulated pot hydroponic system (Fig. 5g). No green fluorescence was observed in Pot1 throughout the treatment period, indicating no presence of EGFP-labeled bacteria in the wastewater, root, or shoot (Fig. 5h). However, consistent green fluorescence was detected in the wastewater of Pot2 and Pot3. Initially, no green fluorescence was detected in the root and shoot samples of Pot2 and Pot3 at 0 d, but it became evident at 4 d and 12 d, indicating the successful migration of the green fluorescent bacteria M. schleiferi S07::EGFP from wastewater to the roots and shoots of M. elatinoides (Fig. 5h).

M. elatinoides mitigates the transmission risk of ARGs from swine wastewater to the receiving environments

To evaluate the transmission risk of ARGs following treatment with M. elatinoides, microcosms were simulated to determine the impact of wastewater-borne ARGs on receiving water and soil. The absolute abundance of tetG, tetW, sul1, intI1, aadA, and ermF in untreated water samples (wastewater-amended water without M. elatinoides treatment) was 35.80, 68.09, 94.24, 4.23, 100.29, and 60.57 folds of that in the blank water samples (wastewater-free water). In contrast, the treated water samples (wastewater-amended water with M. elatinoides treatment) showed a markedly lower increase, with ARG levels only 3.23 to 14.39 times those in the blank water samples (Fig. 6a). Compared with the blank soil samples (wastewater-free soil), the fold changes in the untreated soil samples (wastewater-amended soil without M. elatinoides treatment) ranged from 3.49 to 42.98, while it ranged from 1.60 to 17.51 in the treated soil samples (wastewater-amended soil with M. elatinoides treatment) (Fig. 6b). Overall, M. elatinoides planting could reduce 71.40–96.68% and 36.81–85.69% of the transmission of the target ARGs from wastewater to the receiving water and soils. These findings underscore the effectiveness of M. elatinoides in mitigating the spread of ARGs from swine wastewater to the environment.

**Fig. 6: *M. elatinoides* decreases the transmission risk of ARGs from wastewater to the receiving environments.**

Discussion

The integration of crop-livestock system is essential for promoting sustainable and environmentally friend development in rural regions²⁶. Aquatic plants, such as M. elatinoides, are widely implemented in these systems to enhance the recycling of nutrients preserved in animal wastes, thereby mitigating the environmental impact of wastewater²⁴. In this study, bioinformatics-guided analyses showed that M. elatinoides planting significantly decreased the diversity and abundance of ARGs from wastewater under field conditions (Fig. 1, Supplementary Fig. 1, Supplementary Fig. 2). Experimental evidence confirmed that M elatinoides effectively mitigated both the absolute and relative abundance of the dominant ARGs present in wastewater (Fig. 4, Supplementary Figs. 9 and 10). Since the diversity and abundance of ARGs in the sediments did not show significant changes across the five field ponds (Fig. 1), the direct uptake capacity of M. elatinoides, as well as indirect microbial cues hidden in the alleviation of selective pressures, were hypothesized to act together to contribute to the mitigation of ARGs in wastewater.

Noted that the direct cue associated with migration of ARGs had been evidenced in soil-plant continuum^29,30, it is reasonable to assume the direct uptake and accumulation of M. elatinoides for ARGs from wastewater. As expected, the absolute abundances of the typical ARGs and the amounts of typical ARB in the roots and shoots of M. elatinoides significantly increased after being planted in wastewater (Fig. 5). Such a finding indicates the strong capacity for uptake and accumulation possessed by the aquatic plant, whether this role explicitly participates in the mitigation of antibiotic resistance was further investigated using an EGFP tracing vector. The results showed the EGFP-labeled indigenous antibiotic-resistant bacterium M. schleiferi S07 successfully migrated from wastewater to the roots and shoots of M. elatinoides, which demonstrates that ARGs could be removed from swine wastewater by phytoextraction. Overall, the direct uptake and accumulation of ARGs by M. elatinoides play a pivotal role in the mitigation of ARGs in the livestock wastewater.

Guided by the processes of vertical gene transfer and horizontal gene transfer^31,32, the associations between ARGs and microbial community and/or MGEs are frequently examined as potential evidence of the microbial mechanisms responsible for the environmental transmission of ARGs. VPA results showed that 69.34% of ARG variations in water could be explained by bacterial community composition and MGEs (Fig. 2e), indicating that both factors jointly determine the reduction of ARGs in M. elatinoides-planted ponds. Firstly, the bacterial communities were distinct in water samples but were similar in sediment samples across the five ponds (Fig. 2a and Supplementary Fig. 4), which was consistent with that of ARGs. Procrustes analysis suggested that the coefficients between ARG profiles and bacterial community compositions lower than 0.5 (Fig. 2c), indicating that bacterial community compositions might be a potential determinant of ARGs in both water and sediment samples^32,33. The discrepancy of bacterial community composition may cause the differences of ARGs between water and sediment samples. Meanwhile, the diversity and abundance of ARG-carrying MAGs gradually decreased in the five continuous ponds (Fig. 2b), suggesting that the narrowing of potential ARG hosts in water samples planted with M. elatinoides. In this regard, we proposed that the reduction of ARGs in the wastewater were significantly related to the changes of bacterial communities in M. elatinoides-planted systems.

Secondly, considering the indispensable role of MGEs in the dissemination of ARGs³⁴, the reduction of MGEs in wastewater samples showed a decreasing potential of horizontal gene transfer (Fig. 2b). This was further verified by the reduction in the absolute abundance of intI1 in the laboratory experiment (Fig. 4b). Meanwhile, the relative abundances of transposase and integrase significantly correlated with the relative abundance of ARGs in the water while not in the sediments, indicating these MGEs may play an important role in the mitigation of ARGs in the ecological ponds. Moreover, it is well-known that the MGE-mediated horizontal transfer incident of ARGs is usually driven by selective pressures, especially antibiotic residues and metals^3,35. In addition to being able to directly remove ARGs, aquatic plant has been reported to reduce the concentrations of the pollutants such as residual antibiotics and heavy metals in the wastewater samples³⁶. As expected, the dissipation of residual antibiotics in the wastewater accelerated following planting with M. elatinoides (Fig. 4a), which support the reduction of selection pressures on the spread of ARGs. In a word, M. elatinoides could mitigate the horizontal transfer incidence of ARGs by reducing the load of MGEs, and this process can be further weakened by indirectly alleviating selection pressure.

As ARGs possess both physicochemical properties of persistence and biological properties of rapid proliferation^3,37, whether the mitigation of wastewater-borne ARGs by M. elatinoides eventually results in the reduction of ARG load in the receiving environments remains elusive. Compared with the groups amended with the untreated wastewater, the amendment with the wastewater treated by M. elatinoides reduced 71.40–96.68% and 36.81–85.69% of the load of the typical ARGs into the receiving water and soils (Fig. 6). Such a significant reduction in ARG contamination may be attributed to the following three aspects: (1) M. elatinoides can directly and effectively reduce the load of the predominant ARGs and ARB in the treated wastewater; (2) M. elatinoides can mitigate the selective pressures such as residual antibiotics in the treated wastewater; (3) M. elatinoides can alleviate the horizontal transfer of ARGs as the reduction in the amount of MGEs such as intI1. Overall, integrating aquatic plants with livestock wastewater treatment offers an outstanding advantage in preventing the spread of wastewater-borne ARGs in livestock production systems.

Our findings indicate that the diversity and abundance of ARGs in swine wastewater could be significantly diminished in M. elatinoide-planted ponds, offering an effective mitigation strategy. The attenuation of ARGs is attributed to two underlying mechanisms: the direct accumulation of ARGs by M. elatinoide; and the indirect reduction of proliferation and horizontal transfer events of ARGs due to the alleviation of selective stresses from residual antibiotics. Additionally, planting with M. elatinoide in swine wastewater significantly mitigate the transmission risk of ARGs into the receiving soil and water. Overall, the integrated system of aquatic plants and livestock production presents a practical and eco-friendly solution to curb the spread of antibiotic resistance in agricultural settings. Further research is necessary to corroborate these findings under contrasting types and environmental conditions due to the specificity that exists in livestock wastewater.

Methods

Chemicals and reagents

Standard substances, including TET, SMD, SMX, CHL, TYL, CIP and NOR, were purchased from Dr. Ehrenstorfer GmbH (Germany). Analytic substances such as TET, SMX, CIP, CHL, AMP, KAN and SDZ were purchased from Shanghai Yuanye Bio-Technology Co., Ltd. (China). Phosphate Buffer Solution (PBS, pH 7.2‒7.4), as well as Luria Bertani (LB) broth and agar broth, were provided by Sangon Biotech (Shanghai) Co., Ltd. (China). Analytical and chromatographic grade methanol (MeOH), acetonitrile (ACN), calcium chloride (CaCl₂), sodium chloride (NaCl), phosphoric acid (H₃PO₄), and ethylene diamine tetraacetic acid disodium (EDTA2Na) were purchased from Sigma-Aldrich Pyt. Ltd. (Germany). Ultrapure water was generated using a Milli-Q water purification system (USA).

Field sampling

Field samples were collected from Anhui Runhang Agri-Tech Development Co., Ltd. located in Ganghu Village, Fengtai County, Anhui Province, China (32.75 N, 116.73 E), a company dedicated to the utilization of agricultural waste resources within an integrated crop-livestock system. A series of ponds (approximately 20 m × 15 m × 0.8 ~ 1 m for each pond) were constructed to harness the excessive carbon and nitrogen nutrients in swine wastewater for the cultivation and growth of M. elatinoides. This study focused on five consecutive M. elatinoides-planting ponds (P0, P1, P2, P3, and P4) to track the removal efficacy of ARGs as the swine wastewater is sequentially purified through the ecological ponds. A schematic diagram of these five ponds is depicted in Fig. 1a. In August 2021, approximately 2.0 L of water and 500 g of sediment samples were collected for each pond with three biological replicates. All samples were carefully packed in sterile, sealed bags with dry ice and promptly transported to the laboratory.

Isolation of microbial DNA from wastewater and sediment

One hundred milliliters of water samples were filtered through 0.22 μm filter membranes (Millipore, MA, USA) to harvest microbial biomass. Total DNA from water (filtered membranes) and sediment samples was isolated and purified using a FastDNA™ Spin Kit for Soil (MP Biomedicals, USA) following manufacturer’s instructions and then stored at −20 °C for subsequent analysis. The quantity and quality of the extracted DNA were assessed using a NanoDrop™ 2000 spectrophotometer (Thermo Fisher Scientific, USA) and agarose gel electrophoresis, respectively.

Shotgun metagenome library construction and high-throughput sequencing

Prior to library construction, the prepared DNA samples were again measured by Qubit® 3.0 Flurometer (Invitrogen, USA). A total amount of 0.2 μg DNA per sample was used for the library preparations using NEB Next® Ultra™ DNA Library Prep Kit for Illumina (NEB, USA) following manufacturer’s recommendations and index codes were added to each sample. Briefly, genomic DNA sample was fragmented to a size of 350 bp by sonication. Then DNA fragments were endpolished, A-tailed, and ligated with the full-length adapter for Illumina sequencing, followed by PCR amplification. After PCR products were purified by AMPure XP system (Beckman Coulter, Beverly, USA), DNA concentration was measured by Qubit®3.0 Flurometer, libraries were analyzed for size distribution by Agilent 2100 Bioanalyzer (Agilent, USA) and quantified by real-time PCR (3 nM). The clustering of the index-coded samples was performed on a cBot Cluster Generation System using Illumina PE Cluster Kit (Illumina, USA) according to the manufacturer’s instructions. Finally, the generated 350-bp DNA libraries were subjected to high throughput sequencing on Illumina NovaSeq 6000 platform with a paired-end 150-bp strategy.

ARG and MGE annotation and bacterial community profiling

Raw metagenomic reads containing adapter contamination, low-quality nucleotides and unrecognizable nucleotide (N) were trimmed using Trimmomatic software³⁸ with default parameter (Supplementary Table 1). The generated clean reads were searched against high quality reference protein sequences of ARGs obtained from the Comprehensive Antibiotic Resistance Database (http://arpcard.mcmaster.ca) using BLASTX algorithm. The hit read has a cutoff e-value ≤ 1e-5 and a specific identity cut-off value ≥ 80% was classified as an ARG-like sequence. The identified ARG-like sequences were categorized into ARG types (e.g., tetracycline resistance genes), ARG subtypes (e.g., tetC, tetM, etc.) and resistance mechanisms (e.g., antibiotic efflux, antibiotic inactivation, etc.) upon the antibiotic resistance ontology. The relative abundance of ARGs was normalized to the unit of copy of ARGs per 16S-rRNA gene as described by Li et al.³⁹. The profile of MGEs were characterized with a MGE database⁴⁰, with the relative abundance expressed as ARGs. Taxonomic classification of the resulting metagenomic reads was inferred using Kraken2 based on a k-mer database to achieve high accuracy with fast classification speed⁴¹. The classification results were further estimated for the profiling the relative abundance of the taxa in each sample using Bracken software⁴².

Metagenomic assembly and binning

Metagenomic reads were assembled into contigs using megahit with default parameters⁴³. MAGs were generated using three different binning programs (i.e., MetaBAT2, MaxBin, and CONCOCT) within the MetaWRAP pipeline⁴⁴. The bin reassembly module was employed to enhance the quality of MAGs. Only MAGs meeting the criteria of completeness ≥ 50% and contamination ≤ 5 × were retained for further analysis. The retained MAGs were dereplicated using dRep (v3.0.0)⁴⁵. Open reading frames were predicted from each MAG using Prodigal (v2.6.3) and used for ARG characterization using DeepARG with default parameter^46,47. Taxonomic assignment and quantification of ARG-carrying MAGs were performed using the classify_wf module of gtdbtk (v1.4.0) and the quant_bins module of MetaWRAP, respectively⁴⁸.

Quantification of ARGs with qPCR method

To evaluate the reduction of ARGs during the purification of swine wastewater, 9 targeted ARGs (i.e., sul1, sul2, tetA, tetC, tetM, tetG, tetW, aadA, and ermF) and 1 MGE (intI1) were quantified using qPCR-based method as described in our previous study⁴⁹. The total reaction systems consisted of 12.5 μl of 2 × TB Green Premix Ex Taq II (Takara, Japan), 1.0 μl of each primer (final concentration 0.2 μM), 1.0 μl of template DNA, and 9.5 μl of DNA-free water. qPCR was conducted on Bio-Rad CFX96 Real-Time PCR Detection System (California, USA) with the condition of 30 s at 95 °C for initial enzyme activation, followed by 40 cycles of denaturation at 95 °C for 5 s, annealing at 60 °C for 30 s and extension at 72 °C for 10 s. The information on primers and standard curves is listed in Supplementary Table 2 and Supplementary Table 3. All reactions were conducted in triplicate.

Extraction and determination of antibiotic residues

To assess the removal effects of M. elatinoides-planting ponds on antibiotics, we extracted and quantified their residues in water samples using the methods established in our previous study⁹. Briefly, water samples were processed using solid-phase extraction method with Oasis HLB cartridges (500 mg, 6 mL, Waters) and eluted with 6.0 mL MeOH. The extracts were then dried under a gentle nitrogen stream to approximately 20 μL, reconstituted in 2 mL of a diluent mixture (15: 85 MeOH: Ultrapure water), and transferred to auto sampling vials for analysis. The residual antibiotics were determined using a Waters Acquity™ ultra-performance liquid chromatography–tandem mass spectrometry (UPLC–MS/MS) system.

Hydroponic experiment

To ascertain the capacity of M. elatinoides to remove antibiotics and ARGs, a controlled laboratory hydroponic experiment was conducted using actual wastewater samples from the aforementioned swine breeding farm. M. elatinoides plants were collected from the ecological ponds and were precultured in a 1/2 Hoagland nutrient solution within a climate chamber under a 12/12 h light/dark cycle at temperatures ranging from 20 to 25 °C, with 40−60% atmospheric humidity, and a light intensity of 4000−5000 lux. Prior to formal experiment, the wastewater was diluted to a concentration suitable for the growth of M. elatinoides. A total of 5 plants (length 20 ± 2 cm) were rinsed with deionized water, and then transplanted into a sterile cylindrical glass jar (3000 mL) containing 1500 mL of the diluted wastewater. The roots of M. elatinoides were immersed just below the surface of the wastewater. The outside walls of the jars and the open areas between the cap and were covered with foil paper to prevent exposure to volatilization of wastewater. Jars excluding plants served as controls. Each treatment was replicated five times and incubated in the artificial climate chamber under the abovementioned conditions. Wastewater samples of 100 mL were collected at 0, 4, 8, and 12 d, with each set of samples being divided into two portions, one stored at 4 °C and the other at −20 °C.

Isolation of ARB in wastewater and plant tissue

To determine the ability of M. elatinoides to remove ARB, a traditional culture-dependent method was employed to characterize the abundance and diversity of cultural ARB in wastewater and plant tissues. For plant tissue, root and shoot samples were washed three times with sterilized distilled water, cut into pieces using a sterile lancet, and ground with a mortar and pestle. Tissue pellets from roots and shoots were then added into 90 ml of 0.85% sterilized sodium chloride solution and incubated on a shaker for 30 min (150 rpm and 30 °C). One hundred microliters of wastewater samples or plant tissue suspensions were spread on LB or Mueller-Hinton agar, each supplemented with one of the seven antibiotics: SMX, SDZ, TET, AMP, KAN, CIP, and CHL. The working concentrations of these antibiotics were set according to the minimum inhibitory concentration for resistant bacteria as per the Clinical and Laboratory Standards Institute guidelines (Supplementary Table 4). All plates were incubated at 30 °C for 24 to 72 h. ARB counts were enumerated with an icount10 Shineso fully automatic colony counter (Hangzhou, China) and expressed as colony forming unit per milliliter (CFU mL⁻¹). Besides, individual colonies from each sample plate were picked and streaked onto fresh plates supplemented with the corresponding antibiotic. Genomic DNA from the selected isolates was extracted and used for amplification of 16S rRNA gene sequence with primer pairs 27 F/1492 R⁵⁰. The PCR products served as templates for sanger sequencing, which was conducted by TSINGKE BioTech (Nanjing, China). The nucleotide sequences of the isolated ARB were searched against the NCBI database using a web-based tool (http://www.ncbi.nlm.nih.gov/BLAST) for taxonomic identification.

Bacterial labeling and microscope observation

To track the migration of ARB in the wastewater-plant continuum, an enhanced green fluorescent protein (EGFP) gene labeling approach was employed to visualize translocation. A wastewater-derived antibiotic-resistant strain, Microbacterium schleiferi S07, was prepared as a competent cell and transformed with a plasmid, pTR-EGFP, carrying the EGFP marker. The successful labeling of the M. schleiferi S07 strain with EGFP was confirmed using a Leica DMi8 inversed fluorescent microscope (Germany). The migration experiment of M. schleiferi S07::EGFP in the wastewater-plant continuum was carried out in a hydroponic setup as previously described. Unlabeled M. schleiferi S07 and labeled M. schleiferi S07::EGFP were cultured overnight in LB medium, transferred to sterile tubes, and centrifuged at 8000 × g for 5 min at 4 °C. The supernatants were discarded, and bacterial pellets were washed three times with PBS before being resuspended in PBS. The concentrations of the tested strains were adjusted to 1.0 × 10⁸ CFU mL⁻¹. The ARB migration assay included four treatments: (1) M. elatinoides-planted hydroponic systems with unlabeled strain M. schleiferi S07 (1.0 × 10⁷ CFU mL⁻¹); (2) M. elatinoides-planted hydroponic systems with a high concentration of labeled strain M. schleiferi S07::EGFP (1.0 × 10⁷ CFU mL⁻¹); (3) M. elatinoides-planted hydroponic systems with a low concentration of labeled strain M. schleiferi S07::EGFP (1.0 × 10⁶ CFU mL⁻¹) and; (4) M. elatinoides-free hydroponic systems with a high concentration of labeled strain M. schleiferi S07::EGFP (1.0 × 10⁷ CFU mL⁻¹). To prevent contamination and potential transmission of M. schleiferi S07 through water evaporation, the hydroponic solutions were covered with plastic wrap throughout the cultivation period. The distribution of M. schleiferi S07::EGFP in the collected hydroponic solutions and the plant tissues were visualized at 0, 4, and 12 d, respectively, using a Leica DMi8 inversed fluorescent microscope. All observations were performed at the excitation of 488 nm and emission of 500−550 nm for green fluorescence.

Dispersal risk of ARGs from treated wastewater to the receiving water and soils

To determine the transmission risk of ARGs in wastewater, soil and water microcosms were respectively simulated to compare the influences of treated versus untreated wastewater on the ARGs in the receiving environments. The tested receiving soil and water samples were collected from the agricultural park and the campus pond affiliated to Anhui Agricultural University (Hefei, China) in February 2023. Soil samples were air-dried and filtered with 2-mm sieve to remove the debris and stones. Approximately 500 g of soil samples (equivalent to dry weight) mixed with 40 mL of wastewater treated by M. elatinoides was served as the treatment group, soils mixed with the same volume of wastewater untreated by M. elatinoides was regarded as the control group, and soils mixed with the same volume of sterile water was served as the blank group. Water microcosms were similar with soil microcosms, with the exception that 240 mL of water samples were used instead of soil samples. All microcosms were replicated three times and incubated in the artificial climate chamber at a temperature of 25 ± 1 °C. Soil and water samples were collected at 8 d and used for DNA isolation. The abundance of the targeted ARGs was detected using qPCR methods as aforementioned.

Statistical analysis

Statistical comparisons of the diversity and relative abundances of bacterial taxa, ARGs, and MGEs were performed using one-way analysis of variance followed by a post hoc Tukey HSD test. Significant differences between treatments are denoted by lowercase letters (p < 0.05). PCoA was employed to visualize the differences in bacterial community and ARG compositions based on the Bray-Curtis distances. Mantel test and Procrustes analysis were performed to calculate the similarity between the composition of ARGs and bacterial community. The measure of fit M² (the sum of squared distances between matched sample pairs) and p-values were determined from 9999 labeled permutations. All analyses and visualization were performed in R environment with the packages of vegan, ggplot2, pheatmap, ggnewscale, ggtree, and ggtreeExtra.