Biological mitigation of antibiotic resistance gene dissemination by antioxidant-producing microorganisms in activated sludge systems

Abstract

Antibiotic resistance is the principal mechanism of an ever-growing bacterial threat. Antibiotic residues in the environment are a major contributor to the spread of antibiotic resistance genes (ARGs). Subinhibitory concentrations of antibiotics cause bacteria to produce reactive oxygen species (ROS), which can lead to mutagenesis and horizontal gene transfer (HGT) of ARGs; however, little is known about the mitigation of ARG dissemination through ROS removal by antioxidants. In this study we examine how antioxidant-producing microorganisms inoculated in replicate activated sludge systems can biologically mitigate the dissemination of ARGs. Through quantitative polymerase chain reaction (qPCR) we showed that antioxidant-producing microorganisms could decrease the persistence of the RP4 plasmid and alleviate enrichment of ARGs (sul1) and class 1 integrons (intl1). Metagenomic sequencing identified the most diverse resistome and the most mutated E. coli ARGs in the reactor that contained antibiotics but no antioxidant-producing microorganisms, suggesting that antioxidant-producing microorganisms mitigated ARG enrichment and mutation. Host classification revealed that antioxidant-producing microorganisms decreased the diversity of ARG hosts by shaping the microbial community through competition and functional pathway changes. Conjugative experiments demonstrated that conjugative transfer of ARGs could be mitigated by co-culture with antioxidant-producing microorganisms. Overall, this is a novel study that shows how ARG enrichment and HGT can be mitigated through bioaugmentation with antioxidant-producing microorganisms.

Graphical Abstract

Introduction

Antibiotic resistance contributes to an ever-growing microbial threat. According to the forecast of the United Nations Interagency Coordination Group (UNICG) on Antimicrobial Resistance, more than 10 million people per year will be killed by drug-resistant disease by 2050.¹ Antibiotic overuse and environmental residuals promote the emergence and spread of antibiotic resistance genes (ARGs) by inducing mutations and accelerating horizontal gene transfer (HGT) in a variety of environments.^2-4 Even before the global Covid-19 pandemic, global antibiotic consumption was already rising at an alarming rate, and this trend is expected to continue without targeted intervention.^{5, 6} How to control the dissemination of ARGs under antibiotic selection pressure is therefore a matter of increasing urgency.

Municipal wastewater treatment plants (WWTPs) are a major sink of antibiotic residues due to the intersection of multiple antibiotic sources, including hospital, pharmaceutical and livestock wastewater.^7-9 Unfortunately, antibiotics cannot be removed effectively by traditional wastewater treatment methods. Low concentrations of antibiotics pose an ecotoxicological risk to receiving environments.^10-12 Moreover, antibiotics at subinhibitory concentrations can induce mutations that confer resistance and HGT of ARGs, both of which drive drug selective pressure and ARG prevalence.^13-17

Antibiotics stimulate bacteria to produce reactive oxygen species (ROS),^18-20 which can disturb normal physiological functions.^{21, 22} Excess ROS induced by high levels of antibiotic is a common mechanism of cellular death.^{19, 23, 24} While ROS-mediated cell death plays an important role in the clinical efficacy of antibiotic drugs, non-lethal stress caused by ROS in the environment is a challenge to effective removal of ARGs. Subinhibitory antibiotic concentrations promote antibiotic resistance via ROS-induced mutagenesis, contributing to the development of ARGs.^{25, 26} In addition, ROS induced by antibiotics and other compounds can enhance HGT of ARGs by increasing cell membrane permeability, pilus generation, plasmid replication and competence.^27-32. Mitigating resistance mutations and HGT of ARGs in the WWTP environment could be achieved theoretically by removing ROS induced by antibiotics.

Non-enzymatic antioxidants, such as carotenoids, scavenge ROS and can thus protect microbial cells from oxidative damage.³³ Carotenoids, acting as a secondary metabolite, are produced and excreted by many microorganisms, including bacteria, algae, molds and yeasts.³⁴ Bacteria in the natural environment that produce carotenoids include Deinococcus, Myxococcus, Streptomyces, Flectobacillus, Exiguobacterium and Sphingomonas.³⁵ Algae, including cyanobacteria, red algae, brown algae and green algae, necessarily synthesize many kinds of carotenoids, such as astaxanthin in microalgae Haematococcus pluvialis, and β-carotene, zeaxanthin, echinenone and myxol pentosides in cyanobacteria.^{36, 37} Specifically, Deinococcus radiodurans is an extremophilic bacterium that harbors biosynthetic pathways for various carotenoids^{38, 39} and produces exopolysaccharide and small-molecule antioxidants that protect Escherichia coli cells from oxidative damage.^{40, 41} Rhodotorula is a fungus known to produce carotenoids.^42-45 While antioxidant-producing microorganisms have been demonstrated to protect neighboring cells from oxidative damage, little is known about if and how they influence the development and HGT of ARGs in the setting of subinhibitory concentrations of antibiotics.

In this study, we sought to demonstrate the effectiveness of bioaugmentation with antioxidant-producing microorganisms to decrease ARG dissemination. To that end, we conducted bioaugmentation of activated sludge in a set of replicate sequencing batch reactors (SBRs) with antioxidant-producing microorganisms (Deinococcus radiodurans R1 or Rhodotorula sp.) for ARG dissemination inhibition. We used quantitative PCR to monitor the abundance of ARGs, and metagenomics to distinguish the resistomes and their hosts in each reactor. Conjugative transfer experiments were performed to further corroborate the role of the antioxidant-producing microorganisms in HGT of ARGs.

Material and methods

2.1. Strains and culture media

Escherichia coli DH5α, carrying the RP4 plasmid with a compatible host range, was chosen as the donor to spread ARGs through conjugative transfer. A tetracycline efflux pump tetA (carried by Gram positive and negative pathogens), an aminoglycoside phosphotransferase aphA (mostly carried by Gram positives but can also impact Pseudomonas and Enterobacteriaceae), and a broad-spectrum beta-lactamase TEM-2 are located on plasmid RP4. Deinococcus radiodurans R1 and Rhodotorula sp. were chosen as the antioxidant-producing microorganisms due to their ability to produce carotenoids. All the three strains were purchased from the China General Microbiological Culture Collection Center, CGMCC. Culture conditions for growth of the donor and antioxidant-producing strains are described in Supporting Information (SI) Text S1.

2.2. Reactor setup and operation

Four parallel automatically operated reactors (r0, r1, r2 and r3) with a liquid working volume of 1 L (60 mm internal diameter and 450 mm height, total volume 1.27 L) were run in sequencing batch mode. An air pump was used to aerate at a rate of 0.8 L min⁻¹. Temperature was maintained at 17 °C−25 °C. The hydraulic retention time (HRT) was set at 480 mins, including 20 min filling, 430 min aerating, 25 min settling, 1 min draining and 4 min idling. Seed sludge used in this study was collected from the aeration tank of a municipal wastewater treatment plant of the Chengxi wastewater treatment plant (Hangzhou, China). The chemical oxygen demand (COD) of influent was about 2000 mg/L⁻¹, and the ratio of COD, N (NH₄⁺-N), and P (PO₄³⁻-P) was kept at 100:5:1. Synthetic wastewater was prepared according to Van Loosdrecht et al. with slight modification; the detailed components are described in Supporting Information (SI) Text S2.⁴⁶ Influent of all four reactors did not include any antibiotics for the first 32 days (Stage Ⅰ), and then four kinds of antibiotics (ampicillin, kanamycin sulfate, tetracycline and sulfamethoxazole) at a concentration of 500 μg/L (each) were added into the influent of r1, r2 and r3 for the following 58 days (Stage Ⅱ). At the 33rd day, E. coli DH5α was inoculated into r1, r2 and r3 at a final cell density of 1-5×10⁸ colony-forming units mL⁻¹(CFU mL⁻¹). At the 33rd, 47th, 61st and 75th day, D. radiodurans R1 was inoculated into r2 at a final cell density of 5×10⁷ −1×10⁸ CFU mL⁻¹ and Rhodotorula sp. was inoculated into r3 at a final cell density of 2-5×10⁷ CFU mL⁻¹. COD was measured using a water quality analyzer (GL-200, gelinkairui Co., Ltd, China). The effects of antibiotics and bioaugmentation on ROS production were judged based on the fold changes of fluorescence intensity.^{47, 48} Measurement of fluorescence intensity are described in Supporting Information (SI) Text S3.

2.3. DNA extraction and absolute quantification of conjugal transfer gene (traG), class 1 integron-integrase gene (intI1) and ARGs

To investigate the responses of the RP4 plasmid and ARGs under bioaugmentation and chronic exposure of antibiotics, quantitative polymerase chain reaction (qPCR) was conducted to track abundances of genes related to conjugal transfer (traG), class 1 integron-integrase (intI1), plasmid RP4 (tetA, aphA and TEM-2) and sulfonamide resistance (sul1). Sludge samples were collected from seed sludge and reactors at days 1, 32, 33, 52 and 83, and DNA from 0.2 g solid wet sludge (collected by centrifuging at 8000 rpm for 5 min) was extracted using the DNeasy PowerSoil Kit (Qiagen) following the manufacturer’s instructions. The quality and concentration of extracted DNA were determined using a Nano-300 Micro-Spectrophotometer (ALL FOR LIFE SCIENCE).

DNA amplification was conducted using the Bioer 9600 FQD-96A Real-Time PCR System. Briefly for qPCR reactions, the volume was 20 μL, including 10 μL SYBR Green PCR Master Mix (Biomed, Beijing, China), 0.5 μL 10 μM primer, 8 μL sterile ddH₂O and 1 μL template DNA. Primer sets and PCR conditions are shown in Table S1 and Text S4, respectively. Standard curves of threshold cycle (Ct) values versus the number of gene copies for each gene were generated through serial decimal dilutions of plasmid DNA (Text S4). Standard curves were linear (R² > 0.99) for all targets, with an efficiency of 90−110%. Absolute copy numbers of target DNA fragments were calculated using the Ct values based on standard curves. All qPCR assays were run in triplicate, and data were expressed as average gene copy numbers per ng DNA ± standard deviations.

2.4. Metagenomic Sequencing

Sludge samples from each of the four reactors were collected at days 30 and 80, as detailed in section 2.3. Fifty micrograms of DNA from each reactor at day 30 were mixed together, and used as a control (named ck) before reactors were dosed with antibiotics and underwent bioaugmentation. Library construction was conducted using the Illumina TruSeq DNA Nano Library Prep Kit with high-quality (260/280 > 1.8) DNA samples. The library was then sequenced on an Illumina HiSeq X Ten (2 × 150 bp). More than 240 million reads for the five samples were generated. The percentages of bases that have a Q score above 30 were between 92.3 and 93.3%, demonstrating good quality of the metagenomic sequencing performed in this study. Nonpareil was used to assess the coverage in metagenomic data sets.⁴⁹ The metagenomic sequencing data were deposited into the Sequence Read Archive (SRA) with BioProject accession number PRJNA736480.

2.5. Identification of ARGs

The quality of reads was checked using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Reads were trimmed using Trimmomatic with the default parameters, and reads with low quality scores were removed.⁵⁰ The filtered reads were co-assembled into contigs using metaSPAdes.⁵¹ Contigs of length above 500 bp were binned using Metabat to produce metagenome-assembled genomes (MAGs) and CheckM was used to assess the quality of MAGs.^{52, 53} MetaErg was used to preliminarily annotate the open reading frames (ORFs), and the assembled ORFs were further searched against the Comprehensive Antibiotic Resistance Database (CARD, v 3.1.1) using DIAMOND.^{54, 55} The ORFs with E-value <10⁻¹⁰ and similarity >80% were designated ARG-like sequences.⁵⁶ Plasmid-borne or plasmid-associated ARGs were predicted using the ACLAME database using BLASTP with the same cutoff values as described above.⁵⁷

2.6. Classification of ARG Hosts

To predict the hosts of ARGs, high-quality MAGs with completeness >50% and contamination <10% were selected, and the taxonomy of these MAGs was used to represent the ARG hosts.⁵⁸ Hosts of ARGs not binned into MAGs were classified based on the method of Ma et al.⁵⁶ Briefly, the hosts were defined as the taxon assigned to contigs on which more than 50% of the ORFs fell into the same phylum/class/order/family/genus. All the ARGs affiliated with E. coli were collected to detect mutation by alignment in National Center for Biotechnology Information nucleotide collection (nr/nt) database.

2.7. Microbial community analysis

Considering that the relative abundance of high-quality MAGs was under 70% of the total microbial community, 16S rRNA genes were used to analyze the microbial community structure. Alpha diversity of the microbial community was judged by Shannon index calculated using the vegan package in R. Beta diversity analysis of microbial community was performed using principal coordinate analysis (PCoA) based on a Bray-Curtis similarity matrix calculated using a vegdist function in vegan and visualized with ggplot2.

2.8. Plasmid RP4 transfer experiment

To verify whether antioxidant-producing microorganisms could inhibit conjugative transfer of ARGs under antibiotic selective pressure, co-culture experiments with bacterial isolates were conducted. E. coli DH5α carrying RP4 plasmid was used as the donor of ARGs. To isolate a recipient with streptomycin resistance, 1 mL activated sludge was gradiently diluted and spread onto a Luria-Bertani (LB) agar plate containing 50 mg/L streptomycin to obtain single colonies. The streptomycin-resistant candidate strains were streaked onto LB agar containing 50 mg/L of ampicillin, tetracycline and kanamycin sulfate, and strains that could grow were discarded. The taxonomy of the remaining strains was determined by 16S rRNA gene amplification from the same colony and sequencing analysis (primer and PCR conditions are shown in Table S1 and Text S3, respectively). Finally, an Aeromonas sp. was selected as the recipient of conjugative transfer experiments. Culture conditions for growth of the recipient strain are described in Supporting Information (SI) Text S1.

After culturing, donor, recipient, D. radiodurans R1 and Rhodotorula sp. were harvested by centrifuging and washed with phosphate-buffered saline (PBS, pH=7.2) twice to eliminate the possible influence of culture media. Afterwards, all the strains were re-suspended separately in PBS to obtain initial concentrations of 10⁸ CFU/mL based on OD600 values (the conversion relationships between OD600 and CFU were predetermined by plate count; results not shown). Then, all the strains were mixed at a specified ratio. Extracts of D. radiodurans R1 and Rhodotorula sp. were also prepared as described in Supporting Information (SI) Text S5, and used as protective agent for conjugative systems replacing live organisms. Conjugative transfer experiments were conducted in quadruplicate, and the experimental setup is described in Supporting Information (SI) Table S2. After mating for 8 h at 30 °C, the diluted bacterial suspensions were spread onto LB agar plates containing 50 mg/L of ampicillin, 50 mg/L of kanamycin sulfate, 50 mg/L of streptomycin, and 40 mg/L of tetracycline, to count transconjugants. PCR and DNA Sanger sequencing were used to verify whether plasmid RP4 was transferred into the transconjugants. Primers and PCR conditions were the same as for qPCR (shown in Table S1 and Text S3, respectively). The recipients were counted by spreading the diluted bacterial suspension onto LB agar plates containing 50 mg/L streptomycin. The transfer frequency was calculated by dividing transconjugants over recipients.

2.9. Statistical analysis

Differences in COD of each reactor effluent, copies of each genes quantified by qPCR, fold changes of fluorescence intensity on ROS levels and conjugative transfer frequency were tested by two-tailed Student’s t-test using SPSS version 22.0. Differences at the p < 0.05 level (95% confidence interval) between samples were considered statistically significant.

Results and Discussion

3.1. Reactor performance was affected by antibiotics and antioxidant-producing microorganisms

During reactor operation, COD was monitored to investigate the performance of reactors under antibiotic selection pressure and bioaugmentation protection. As shown in Figure 1, there was no significant difference for COD removal in any reactor from day 1 to 32 (stage Ⅰ), which indicated that the four reactors achieved similar performances in carbon metabolism. However, effluent COD of r1 in stage Ⅱ (68.75±22.45 mg/L, n=8) was significantly higher than that of r1 at stage Ⅰ (123.69±30.06 mg/L, n=13) (p<0.01), indicating that addition of antibiotics harmed the metabolic capability of indigenous microorganisms. Generally, antibiotic concentrations in wastewater treatment systems are on the order of nanogram to microgram per liter range.^{59, 60} Even at low concentrations, however, antibiotics can act as an ecotoxicological risk factor through effects on selective pressure, biomass production and nutrient transformation.^61-63 For example, the addition of sulfadiazine to manure reduced microbial activity and nitrogen turnover processes in an investigation of the interactions between manure and soil biota.⁶² Thiele-Bruhn and Beck found that antibiotics can exert a temporary selective pressure and significantly reduced numbers of soil bacteria even at environmentally relevant concentrations.⁶³ Similarly, antibiotics with environmentally relevant concentrations (500 μg/L) in the influent reduced the microbial COD removal in r1 in this study. The largest decrease of total bacteria within 24 h after antibiotic addtion in r1 reactor quantified by qPCR (Figure S1) also validated the toxicity of antibiotics. In contrast, effluent COD of r2 and r3 was significantly lower than that of r1 (p<0.01), indicating that bioaugmentation with D. radiodurans R1 or Rhodotorula sp. effectively protected the reactors, even though the same concentration of antibiotics were contained in their influent. The direct measurement of intracellular ROS production showed a significant increase in ROS production of 1.30 fold (p<0.01) after exposure to antibiotics with environmentally relevant concentration (Figure S2), which further validated the stress effect of antibiotics. The ROS production increase of bioaugmented groups only 1.16 and 1.11 folds, significantly lower than that of non-bioaugmented group (p<0.01), which confirmed the protective effects of antioxidant-producing microorganisms on the microbial community.

Figure 1.

Performance of COD removal. The arrow represents that antibiotics were added into reactor r1, r2 and r3 at day 33, at which time D. radiodurans R1 and Rhodotorula sp. were inoculated into reactors r1 and r2, respectively.

3.2. Bioaugmentation decreased persistence of plasmid RP4 and alleviated enrichment of sul1 and intl1

To investigate the effects of bioaugmentation on the diffusion of a mobile exogenous resistance plasmid, we inoculated E. coli DH5α (RP4) into reactors r1, r2 and r3 at day 33 (the start of Stage Ⅱ). During reactor operation, ARGs conferring resistance against aminoglycosides (aphA), beta-lactams (TEM-2) and tetracyclines (tetA), as well as a conjugal transfer gene (traG), located on plasmid RP4 were quantified by qPCR. The copy numbers of aphA, TEM-2 and tetA showed a downward trend in Stage Ⅰ (days 1 and 32) compared with those in the seed sludge for all reactors (Figure 2). These ARGs thus decreased in systems fed with synthetic wastewater and no selective pressure.

Figure 2.

Absolute quantification of ARGs (aphA, TEM-2, tetA, and sul1), conjugal transfer gene traG, and class 1 integrase gene intl1 by qPCR. Names with an asterisk denote genes located on plasmid RP4. Error bar represents three technical replicates.

In light of resistant bacterial infections causing human disease, the effects of antibiotic selective pressure on the emergence and HGT of ARGs is of major concern.^{3, 64-66} Despite the competitive advantages conferred by plasmids carrying ARGs, they produce a burden (fitness cost) in hosts due to the costs of replication, repair and expression of plasmid genes.^{67, 68} In other words, bacteria may lose plasmids following the energy minimization principle in the absence of selection pressure, which is congruent with our observations for aphA, TEM-2 and tetA. Moreover, the copy number of traG in the sample of both seed sludge and reactors in Stage Ⅰ was less than 2.0×10³ copies/ng DNA (Figure 2), which suggested that plasmid RP4 is not prevalent among indigenous microorganisms, and comparable to the level of IncP plasmid in the natural environment.⁶⁹ There was no traG detected in the control sample ck by metagenomic data (Figure S3).

The fitness costs can be counterbalanced by fitness benefits carried by plasmids to the host cells under selective pressure.⁷⁰ In our system, we would thus not expect plasmid RP4 to be maintained in the absence of a selective pressure. Indeed, even in the reactors to which we added antibiotics, the effective concentrations were low. After E. coli DH5α (RP4) inoculation, the copy number of the four RP4-encoded genes increased by at least an order of magnitude in r1, r2 and r3 at day 33 compared to their respective counts prior to inoculation (the start of Stage Ⅱ; Figure 2). As expected, plasmid RP4 decreased from when E. coli DH5α (RP4) was inoculated in all reactors, and the decrease was more pronounced in reactors bioaugmented with antioxidant producers (r2 and r3) to levels comparable with r0 (Figure 2). In contrast, the abundances remained high at day 83 in r1 (Figure 2). The relative abundance of these four genes estimated based on metagenomic data also confirmed this trend (Figure S3). Thus, pressure was exerted by the antibiotics, and that pressure was diminished in the presence of bioaugmentation with antioxidant-producing microorganisms.

To investigate the effects of bioaugmentation on ARGs and mobile gene elements (MGEs) in indigenous microorganisms, sulfonamide resistance (sul1) and the class 1 integrase gene (intl1) were monitored using qPCR. Like ARGs on plasmid RP4, the abundance of sul1 also decreased in Stage Ⅰ with no selective pressure, from 1.02×10⁷ copies/ng DNA in seed sludge to lower than 6.77×10⁶ copies/ng DNA in samples at day 32 (Figure 2). However, sul1 began to increase after antibiotic addition to more than 2.91×10⁷ copies/ng DNA in r1, with only slight fluctuations in r2 and r3, which were bioaugmented with antioxidant-producing microorganisms. Although intl1 did not show an obvious decline in Stage Ⅰ, it increased gradually to 3.77×10⁷ copies/ng DNA in r1 at day 83, much higher than in r2 and r3. The relative abundance of these two genes estimated based on metagenomic data also showed the highest abundance in r1 at day 80 (Figure S3). These results indicate that bioaugmentation alleviated the enrichment of sul1 and intl1 from indigenous microorganisms under the pressure effects of antibiotics, which could be also validated by different levels of ROS production (Figure S2).

3.3. Bioaugmentation decreased diversity of ARGs

Nonpareil assessment showed that metagenomic data from sludge samples (except r1) was close to or exceeded 95% average coverage (Table S3), a criterion that can be regarded as “nearly complete coverage”.⁴⁹ Therefore, while r1 (with 93.27% average coverage) likely underestimated the number of low-abundance genes to some extent, metagenomic sequences of the other samples included almost all genes. There were 406, 401, 567, 479 and 404 open-reading frames (ORFs) predicted as ARGs from the ck, r0, r1, r2, and r3 metagenomes, respectively, of which 127, 126, 153, 138 and 137 ORFs were affiliated with plasmid-borne or plasmid-associated ARGs (Figure 3A). Reactor r1 had both higher abundance of ARGs (total and plasmid-associated) and resistome diversity (55 antimicrobial resistance (AMR) gene families, compared to 47 for ck, 37 for r0, 49 for r2 and 42 for r3) (Figure 3A). Note that the ARGs of r1 were likely underestimated (lower average coverage). The number of different AMR gene families in r0 was 21.28% less than that of ck, indicating that richness of ARGs decreased with no selective pressure. The richness of AMR gene families in r1 was greater than that of ck by 17.02%, indicating that ARG diversity increased, driven by antibiotics. However, compared to r1, AMR gene family richness only slightly increased in r2 (+4.25%) and even decreased in r3 (−10.64%). Therefore, bioaugmentation with antioxidant-producing microorganisms could decrease the richness of ARGs under antibiotic pressure effects.

Figure 3.

ARGs in the microbial community of each sludge sample. A: Total and plasmid-borne ARGs, and antimicrobial resistance gene family. B: Venn network of ARGs unique to each reactor compared with ck. The nodes represent ORFs predicted as resistance genes. C: Composition of ARGs unique to each reactor compared with ck.

There were 190 ORFs shared by all five samples, accounting for only 24.93% of all ARGs (Figure S4). Thus the resistome changed dramatically under the influences of antibiotic addition and bioaugmentation. There were 149, 274, 184 and 144 ARGs detected in sludge samples of the four reactors but not in ck (Figure 3B). These newly emerging ARGs may have been present at low abundance in the former stage of operation and enriched under antibiotic selective pressure. Manifestly, ARG abundance has been increasing since antibiotics were widely used and discharged,⁷¹ and even environmental concentrations of antibiotics are high enough to exert a selective pressure on clinically relevant bacteria.⁷² Among the four reactors, the most new ARGs were found in r1, accounting for 35.96% of all the ARGs, while there were only 24.15% and 18.90% in r2 and r3, respectively. Specifically, the numbers of multi-drug resistance (MDR), aminoglycoside, fluoroquinolone, tetracycline and peptide antibiotic ARGs were higher in r1 than that in r0, r2 and r3 (Figure 3C). New tetracycline antibiotic ARGs accounted for 61.54% of all the tetracycline antibiotic ARGs in r1; however, the values were 46.97%, 48.65% and 50.88% in r0, r2 and r3. These results indicate that bioaugmentation with antioxidant-producing microorganisms could effectively reduce the enrichment of antibiotics on ARGs.

In addition to ARG enrichment, antibiotics can stimulate the production of reactive oxygen species (ROS), which can lead to ROS-induced mutagenesis.^{19, 25, 26} In this study, 17 mutated ARGs (including single-base substitution and insertion-deletion mutations) affiliated with E. coli were detected in r1 (Dataset S1). In comparison, there were 0, 1 and 2 mutated E. coli ARGs detected in r0, r2 and r3. Thus, bioaugmentation may reduce ARG mutation by scavenging ROS.

3.4. Bioaugmentation decreased richness of ARG hosts

A growing body of research has indicated that the resistome is mainly determined by microbial community composition,^73-75 although the resistome was once thought not to correlate with phylogeny or taxonomy due to the highly mobility of ARGs.^76-78 To investigate the relationship between the microbial community and ARG hosts in our reactors, the microbial community structure was analyzed based on 16S rRNA gene presence in metagenomic data, and the ARG hosts of both binned and unbinned ARGs were predicted and counted. There were 115, 118, 130, 119 and 101 genera identified as ARG hosts from ck, r0, r1, r2, and r3, respectively (Dataset S2). Alpha diversity and community structure of ARG hosts were analyzed (Figure 4), the richness of ARG hosts mirrored that of the overall community in the four reactors. Reactor r0 without antibiotics had the highest alpha diversity of both the overall community and ARG hosts (Figure 4A), which indicated that antibiotic ecotoxicity decreased the diversity of microbial community in the other reactors. Although the alpha diversity of the overall community in r1 was lower than that of ck, the alpha diversity of ARG hosts in r1 was greater. If the toxic effects of antibiotics reduce community diversity, one may wonder why the alpha diversity of the overall community in r1 would be higher than that of r2 and r3, given that antibiotics were present in all three reactors. The protective effects of plasmid RP4 on the community may be one of the reasons. Based on the report of Blaustein et al., resistance conferred via antibiotic transformation could provide a protective effect on the community.⁷⁹ Plasmid RP4, a mobile plasmid with compatible host range, was introduced into reactors by E. coli DH5α (RP4) inoculation. Of the three ARGs encoded on RP4, aphA and TEM-2 both confer the ability to detoxify their target antibiotics, which could potentially extend a protective effect not only to the host but also its neighbors. Moreover, indigenous bacteria could also acquire plasmid RP4 via HGT, thus surviving and further decreasing the ecotoxicity of the added antibiotics. The protective effects of RP4 was diminished by bioaugmentation, which lessened the stimulus for HGT of RP4 in r2 and r3. Thus the alpha diversity of overall community in r2 and r3 declined, and then diversity of ARG hosts decreased.

Figure 4.

Alpha diversity and community structure of overall microbial community and ARG-hosts. A: Alpha diversity shannon index of the microbial community and ARG host community. B: Relative abundance of overall microbial community at the genus level. C: Relative abundance of ARG hosts at the genus level.

Furthermore, bioaugmentation by exogenous microorganisms could also reduce community evenness but increase richness due to ecological competition.⁸⁰ That may explain why reactors r2 and r3 harbored less diverse and less even overall communities, and higher enrichment of Thauera sp. The relative abundance of Thauera sp. was greater than 58% and 45% in r2 and r3, respectively, much higher than that in ck (29%), r0 (19%) and r1 (33%) (Figure 4B).

On the other hand, the addition of exogenous species may have led to rearrangement of the microbial community structure by changing the catabolic traits and energy flux.⁸¹ Changes in abundance of functional pathways with relative abundance > 0.1% (Figure S5) were calculated in different treated reactors (r1, r2 and r3) relative to the control reactor r0 (Figure 5). In this study, carbon, nitrogen, sulfur metabolisms and oxidative phosphorylation processes necessary for life increased in bioaugmented reactors. The MAPK and FoxO signaling pathways that play key roles in the regulation of proliferation, differentiation, the stress response, oxidative stress resistance, motility, growth, survival, and death, were also higher in bioaugmented reactors r2 and r3 than in r1. Thus, once again antioxidant-producing microorganisms improved the ability of the community to resist oxidative stress, which could be also validated by the direct measurement of intracellular ROS production (Figure S2).

Figure 5.

Changes in the abundance of functional pathways after antibiotic addition and bioaugmentation. Circles represent the fold change of functional pathways in different treated reactors relative to the control reactor r0 (values below 1 denote a decrease and values above 1 denote an increase).

Beta diversity might also be impacted by multiple effects of antibiotic selection, RP4 protection, competition and functional pathway changes induced by bioaugmentation. Bray-Curtis similarity matrix-based PCoA analysis (Figure S6) revealed that the microbial communities of r0, r1, r2 and r3 differ from ck mainly on axis 1, and microbial community of r0, r1, r2 and r3 differ from each other mainly on axis 2. The relatively large Bray-Curtis distance between r0 and r1 suggests that antibiotics led to dramatic microbial community shifts. Reactors r2 and r3 had a similar microbial community structure, different from r0 and r1, which suggested that D. radiodurans R1 and Rhodotorula sp. played a similar role in microbial community changes.

Changes in microbial community structure consequently led to the difference in ARG host distribution. In all five samples, Thauera was the dominant genus and increased with the addition of antibiotics in r1, r2 and r3 (Figure 4B), and a similar situation also ocurred in ARG hosts (Figure 4C). Likewise, Flavobacterium sp. (in ck), Amaricoccus sp. (in r0), Arenimonas sp. (in r1) and Acidovorax sp. (in r2 and r3) also had a relative high abundance in their respective communities, and they happened to be the hosts of ARGs. However, the relationship between overall and ARG host abundance is not always one-to-one, e.g., OLB12 sp. had a relative high abundance in r0, r1 and r3, but they were not identified as ARG hosts. Furthermore, more than 64% of the ARG hosts were Thauera sp. in bioaugmented reactors (r2 and r3), and other ARG hosts accounted for very little (less than 7%), while several genera identified as ARG hosts accounts for high proportion (more than 11%) in ck, r0 and r1. These community structure changes resulting from exogenous antioxidant-producing microorganisms did not destroy the original ecological balance, and even improved the performance for COD removal under antibiotic selection pressure.

Consequently, it can be concluded bioaugmentation with antioxidant-producing microorganisms may shape microbial community, and further change the resistome host profile.

3.6. Conjugative transfer of ARGs was mitigated by co-culture with antioxidant-producing microorganisms

ROS induced by subinhibitory concentrations of antibiotics enhances HGT of ARGs.^{30, 31, 82} However, addition of thiourea, an ROS scavenger, decreased the production of ROS, and further prevented the enhancement of HGT frequency.^{30, 31} Based on the decrease in ARG host diversity in our bioaugmented reactors, we hypothesize that the presence of carotenoids reduced oxidative stress and thus HGT. To corroborate that hypothesis, we performed bacterial co-culture conjugation experiments between the E. coli donor and a recipient bacteria isolated from our seed sludge. Through the conjugative experiments, both D. radiodurans R1 and Rhodotorula sp. decreased the conjugative transfer frequency of plasmid RP4 with 0.5, 1.0, 2.5 or 5.0 mg/L antibiotics after 8 hours conjugative transfer (Figure 6). The frequency of conjugative transfer decreased more than 3.8 and 3.3 times with D. radiodurans R1 and Rhodotorula sp. co-culture, respectively. Moreover, conjugative transfer frequency of plasmid RP4 under 5.0 mg/L antibiotics selective pressure also declined more than 9.1 and 6.5 times with addition of extracts of these two microorganisms, respectively (Figure S7). This suggests that antioxidant-producing microorganisms may play the same role as thiourea and protect donor and recipient bacteria by scavenging ROS induced by antibiotics.

Figure 6.

Effect of antioxidant-producing microorganism co-culture on the conjugative transfer frequency of ARGs. Significant differences between co-culture groups and the control were detected using a t-test; * denotes P < 0.05, and ** denotes P< 0.01.

3.7. Outlook of application potential

In this era of antibiotic overuse and antibiotic resistance prevelence, and there is no effective method to remove ARGs and antibiotic resistance at present. Most of the current research has focused on the promotion and spread of ARGs by all kinds of pollutants. Antimicrobial stewardship programs aim to prevent the introduction of antibiotics, and thus ARGs, by reducing the overall use of these drugs. However, no interventions exist to remediate ARGs in antibiotic-laden environments. This study provides the first evidence that bioaugmentation with antioxidant-producing microorganisms can mitigate the dissemination of ARGs. The persistence of ARGs, the diversity of both resistomes and ARG hosts, and HGT all decreased in the presence of antioxidant-producing microorganisms. Antioxidant-producing microorganisms are ubiquitous in natural environment and widely available for future applications in wastewater treatment. Our findings thus provide a practical and sustainable approach to control the spread of ARGs based on bioaugmentation against a background of antibiotic selection. Future studies are needed to investigate the mechanisms underlying successful bioaugmentation and feasibility on a larger scale.

Synopsis.

Antioxidant-producing microorganisms can mitigate dissemination of ARGs by shaping community structure and decreasing HGT under environmentally relevant antibiotic stress.