Comparison of bacterial communities and antibiotic resistance genes in oxidation ditches and membrane bioreactors

Oxidation ditches (ODs) and membrane bioreactors (MBRs) are widely used in wastewater treatment plants (WWTPs) with bacteria and antibiotic resistance genes (ARGs) running through the whole system. In this study, metagenomic sequencing was used to compare the bacterial communities and ARGs in the OD and MBR systems, which received the same influent in a WWTP located in Xinjiang, China. The results showed that the removal efficiency of pollutants by the MBR process was better than that by the OD process. The composition and the relative abundance of bacteria in activated sludge were similar at the phylum and genus levels and were not affected by process type. Multidrug, fluoroquinolones and peptides were the main ARG types for the two processes, with macB being the main ARG subtype, and the relative abundance of ARG subtypes in MBR effluent was much higher than that in the OD effluent. The mobile genetic elements (MGEs) in the activated sludge were mainly transposons (tnpA) and insertion sequences (ISs; IS91). These results provide a theoretical basis for process selection and controlling the spread of ARGs.

The activated sludge process is currently the most widely used wastewater treatment technology. Oxidation ditches (ODs) and membrane bioreactors (MBRs), which are the two main activated sludge processes, utilize different "sludge-water" separation methods. Specifically, the secondary sedimentation tank following the OD relies on gravity sedimentation, while the MBR relies on filtration and interception via the membrane 1 . These systems have unique artificial microbial ecosystems that characteristically harbor high microbial diversity and a large number of microorganisms, especially bacteria carrying various types of antibiotic resistance genes (ARGs), which can also drive changes in ARGs 2,3 . As an emerging type of biological pollutant, ARGs can persist, transfer and spread within different environmental media, causing serious harm to human health 4 .
The pollutant removal performance and stable operation of the treatment system depend on the microorganisms, with bacteria playing a key role 5 . At present, many studies have explored the bacterial communities in OD and MBR systems. These studies showed that the bacterial community members at the phylum level in the two systems are similar, with Proteobacteria and Bacteroides generally dominating 6,7 . However, there were some differences at the genus level. The OD system produces alternating anoxic and aerobic conditions in space and time, enabling nitrification and denitrification to be carried out simultaneously 8 . Subsequently, Nitrospira 9,10 , Nitrosococcus 10 , Nitrosomonas 8 , Thauera 8,9 , Arcobacter 8 and Zoogloea 8 were detected as the most abundant nitrogen removal bacteria. Dechloromonas, a key phosphate-accumulating organisms (PAOs), has been detected with high abundance in ODs. Compared with ODs, MBRs have a longer sludge residence time (SRT), which provides favorable living conditions for some slow-growing microorganisms, especially the autotrophic nitrifying bacteria Nitrosomonas 11 and Nitrospira 1,11,12 . Moreover, membrane interception results in richer microbial diversity and higher microbial biomass. Accordingly, Zoogloea 1 , Flavobacterium 7 , Thauera 1,7 , Comamonas 7 , Haliscomenobacter 12 and Rhodobacter 12 are considered dominant, and the PAOs Tetrasphaera and Accumulibacter are frequently detected in MBRs 13,14 . These abundant bacterial genera are widely distributed in the treatment system and are responsible for effectively removing nitrogen, phosphorus and other pollutants, as well as ensuring stable effluent quality 7,12,13 .

Results
Treatment performance of the WWTP. The quality of the incoming and outgoing water from the WWTP and the treatment performance are shown in Table 1. The average ratio of biochemical oxygen demand (BOD 5 ) to chemical oxygen demand (COD) was 0.45 in the influent, indicating that the biochemical performance of the influent was good. The average ratios of BOD 5 to total nitrogen (TN) and total phosphorus (TP) were 3.75 and 30.69, respectively. Thus, both treatment systems can operate efficiently and stably, and the treatment effects meet the design standards. Differences in the systems resulted in higher removal rates of NH 4 + -N and TN in the OD than in the MBR and higher removal rates of COD, BOD 5 , SS, and TP in the MBR than in the OD.
Comparison of ARGs. There were 10 ARG types with relatively high abundances in all samples (Fig. 4).
Among them, the multidrug type was the most abundant, followed by fluoroquinolones and peptides, the sum of which was as high as 41.09-54.81%.
Overall, 413-573 ARG subtypes were detected in all samples. Sample Inf had the most abundant subtypes (573), while sample O-Eff had the fewest subtypes (413). The total relative abundance was assigned to the top 15 ARG subtypes in all samples, ranging from 791.34 to 3774.37 ppm (Fig. 5), accounting for 50.23-53.23% of all ARGs, of which macB (95.16-500.71 ppm) had the highest abundance, followed by rpoC (100.99-296.52 ppm),    5 . Firmicutes (31.32%) and Bacteroidetes (13.01%) had higher abundances in the Inf sample, and the former decreased to 3.65% and 2.07% in samples O-Eff and M-Eff, respectively, while the latter decreased to 5.80% and 2.02%, respectively. It has been reported that Firmicutes is mainly responsible for the degradation of lignocellulose and hemicellulose 27 . Bacteroidetes is responsible for the decomposition of polymers and complex organic compounds, producing simple molecules that are easy to absorb, transform and utilize by other microorganisms 26 . Furthermore, Actinobacteria,  www.nature.com/scientificreports/ Chloroflexi, and Nitrospirae were enriched in activated sludge and wastewater. Actinobacteria can decompose complex compounds and certain toxic compounds, and Chloroflexi plays an important role in the degradation of carbohydrates and cellular materials 11 . The existence of these bacteria makes a great contribution to the removal of pollutants. The heatmap (Fig. 2) shows the distribution of bacterial genera in all samples. Acidovorax (17.55%) was the predominant bacteria in sample Inf. It is worth noting that Acidovorax is commonly found in WWTPs with high abundance and has been indicated to carry ermB and blaTEM 28 . Other major bacteria were Acinetobacter (11.96%), Bacteroides (6.97%), Lactobacillus (5.13%), Bifidobacterium (4.25%), Escherichia (3.71%), and Prevotella (3.63%). Among these genera, Acinetobacter and Bacteroides are considered potential pathogens, and are also thought to carry a variety of types of ARGs and be resistant to multiple antibiotics 18 . In addition, Escherichia and Prevotella are human conditional pathogens. Notably, Lactobacillus and Bifidobacterium are both beneficial microorganisms in the human body 29 . These results indicated that there were a large number of pathogens and potential hosts of ARGs in the influent water of WWTPs, which need to be removed to reduce or even eliminate the risk of ARG transmission.
The abundance of the main bacteria in the influent decreased to varying degrees in the effluent sample. The relative abundance of Acidovorax in samples O-Eff and M-Eff was reduced to 8.95% and 6.43%, respectively. Similarly, Acinetobacter and Bacteroides were reduced to 5.78% and 2.14% in O-Eff and 1.86% and 0.02% in M-Eff, respectively. Escherichia and Prevotella also clearly decreased in O-Eff. This result implies that the enrichment degree of ARG-carrying bacteria and pathogens in OD effluent was higher than that in MBR effluent, which  www.nature.com/scientificreports/ may be due to the interception of the MBR membrane to prevent microorganisms from being washed out 30 . Although these dominant components in the influent decreased to some extent, they did not disappear completely, and some harmful bacteria even increased. In sample O-Eff, Luteimonas (5.91%), Acinetobacter (5.78%), Mycolicibacterium (4.70%), Arcobacter (4.47%), and Tetrasphaera (3.63%) were more abundant. Comparatively, Luteimonas (20.01%) and Deinococcus (13.62%) were the main bacteria in sample M-Eff, and Tetrasphaera (6.91%) and Mycolicibacterium (4.83%) also had higher abundances. Luteimonas has a strong ability to degrade starch and can be detected in biofilters, soil, sediment, fresh water and sea water 31 . Deinococcus is known for its ability to repair large amounts of DNA damage and is highly tolerant to extreme radiation and dry conditions 32 , and Tetrasphaera is a common phosphorus removal bacterium in WWTPs 33 . Arcobacter and Mycolicibacterium are not only pathogens of humans and animals but also potential hosts of ARGs and MGEs 18 . More specifically, Mycolicibacterium carries the gyrA gene and is resistant to fluoroquinolones 34 . These ARG-carrying pathogens can lead to the spread of disease and drug resistance 21 , which is worthy of attention.
In addition, there were other pathogenic bacteria with low abundance (0.09-1.30%) in both effluent samples, such as Enterobacter, Escherichia, and Streptococcus 18 . After processing treatment, there were still many kinds of pathogenic bacteria and microorganisms carrying ARGs in the effluent. A previous study on the removal efficiency of ARGs in wastewater by five processes (activated sludge, MBR, SBR, upflow anaerobic sludge blanket (UASB), and biological filter) showed that all of these processes can remove pathogens to some extent although there are still residues in the effluent 35 , which is similar to the results of this study. More efforts need to be made to study the inactivation of pathogens and ARGs in the effluent after ultraviolet disinfection.
For activated sludge, samples O1 and O2 contained mainly Acidovorax (16.03% and 9.23%), Mycolicibacterium (4.02% and 3.92%), Candidatus_Promineofilum (3.87% and 3.66%), Micropruina (3.75% and 3.59%), Thauera (3.23% and 3.89%), and Nitrospira (2.72% and 4.26%). Similarly, in both samples O2 and M2, the main bacteria were Mycolicibacterium (10.02% and 13.12%), Tetrasphaera (9.06% and 9.06%), Acidovorax (6.73% and 6.73%), Candidatus_Promineofilum (5.31% and 5.50%), and Micropruina (4.37% and 2.54%). This implies that the distribution of bacteria in the OD and MBR activated sludge was similar, and the clustering result of the heatmap (Fig. 2) also confirms this point. In addition, there was no significant difference in the relative abundance of bacteria between the two kinds of activated sludge (O1 and M1, and O2 and M2), by ANOVA (p > 0.05). A recent study indicated that similar wastewater treated by different processes (ODs, A 2 O and cyclic activated sludge system (CASS)) led to a similar microbial community distribution 18 . Another study on the treatment of municipal wastewater with AO and A 2 O also showed that the microbial communities in activated sludge were similar 36 . These results are consistent with the results of this study that under the same influent conditions, the microorganisms in activated sludge are not affected by the type of process.
Further analysis of the identified functional bacteria for nitrogen and phosphorus removal in activated sludge showed that Nitrosomonas (0.04-0.90%) and Nitrospira (1.48-3.73%) were the main AOB and NOB, respectively, which is in line with other WWTPs 18 . These two nitrifying bacteria grow slowly, and as typical "K-strategists", they have the advantage of using resources efficiently and growing at the longer SRT employed by MBRs 30 . This is similar to the results reported by Ma et al., in that Nitrosomonas (0.33%) and Nitrospira (0.17%) had higher abundances in the MBR process than in a conventional activated sludge process after treatment of the same wastewater 7 . Both Tetrasphaera and Accumulibacter are widely regarded as the main PAOs in activated sludge 37 . Tetrasphaera can rely on fermentation metabolism for cell maintenance and proliferation, and its fermentation products can even supply Accumulibacter with synergistic phosphorus removal 33 . In this study, Tetrasphaera was the most abundant phosphorus-accumulating bacteria in the two kinds of activated sludge, with a relative abundance of more than 2.79%, while Accumulibacter was less than 0.6%. A previous study suggested that Tetrasphaera can absorb hydrophilic substances under anaerobic, anoxic and aerobic conditions and accumulate hydrophobic substances, such as long-chain fatty acids, for long-term survival. In contrast, Accumulibacter can hardly survive under a long-term hypoxia/aerobic alternate environment 38,39 , which explains why the proportion of Tetrasphaera in most WWTPs is much higher than that of Accumulibacter 33 . The existence of these functional microorganisms provides a reasonable explanation for the good nitrogen and phosphorus removal efficiency of the OD and MBR processes. Surprisingly, these functional bacteria were confirmed to carry many types of ARGs, such as Nitrosomonas and Nitrosospira carrying EF-TU 16 ; Nitrosomonas may also carry penA, oqxBgb, vanHAc2, vanR-F, and dfrK 17 ; and Accumulibacter is a host of peb-EC, SFO-1, vanR-B, and vanR-C 17 . Although the relative abundance of these functional bacteria in the influent was very low (< 0.1%), it was detected in activated sludge and effluent and would pose a potential threat to the receiving environment.
Most ARGs entering the wastewater treatment system are carried by microorganisms 21 . Multidrug (27.31%), fluoroquinolones (12.97%) and polypeptides (12.91%) were the top ARG types identified in sample Inf. A study found that macrolides, tetracyclines, aminoglycosides, beta-lactams, and sulfonamides were the most abundant in untreated wastewater in 60 countries around the world 40 , which is significantly different from the results of this study and may be influenced by socioeconomic, health and environmental factors 40 . The ARGs in the activated sludge samples were consistent with those in the influent, namely, multidrug (24.91-25.57%), fluoroquinolones (13.14-14.40%) and polypeptides (12.46-13.10%), which is different from the other regions of China (Beijing 17 , Shanghai 16 and Hong Kong 20 ). It is worth noting that all major ARG types were consistent in type and similar in relative abundance in the OD and MBR. However, it can be seen from the effluent samples that the removal proportion of OD on multidrug is significantly higher than that of MBR, while the removal proportion of these two processes on other drug resistance categories were relatively similar. The results show that the ARG removal effect of the OD process is better than that of the MBR process. This result is similar to the findings of Yuan et al., who noted that the activated sludge method greatly reduced ARGs, while MBR did not 35 . This result is surprising as most previous results have shown that MBR is the most effective way to remove ARGs from wastewater due to ARGs been trapped by the membrane. A large number of microorganisms were trapped in the MBR membrane pool; under the conditions of high aeration and stirring, the cells were lysed, and intracellular DNA (iDNA) leaked out, which may lead to the persistence of extracellular DNA (eDNA) in the effluent after MBR treatment 35 . In addition, the higher biomass concentration in sample M1 (expressed by suspended solids in mixed liquid (MLSS), 10,380 mg/L) than in sample O1 (4834 mg/L) also indicates that there will be more DNA leakage. The migration of eDNA is an important reason for the migration and transformation of ARGs. Current evidence shows that high activated sludge concentrations and biofilm-and antibiotic-resistant bacteria in the MBR process provide favorable conditions for horizontal transfer of ARGs (Rp4 plasmid) 42 . The good settling performance of activated sludge may lead to the transfer of most bacteria to biological solids, which is helpful to reduce eDNA. Compared with sample O1 (94 mg/L), sample M1 (156 mg/L) has a higher sludge volume index (SVI) value, so the MBR process cannot effectively prevent the leakage and dissociation of eDNA. Previous studies have confirmed that a decrease in ARGs is positively correlated with a decrease in the microbial community (p < 0.01) 43 . Further study of eDNA is more convincing to the speculated results in the future. In addition, the pore size of the membrane is also one of the important factors affecting ARG removal 44 . The MBR process with an average pore size of 0.1-0.4 μm was used in a WWTP in Wuxi (China) and showed a notable removal effect on three tetracycline resistance genes (tetG, tetW and tetX), one sulfonamide resistance gene (sul1) and one class 1 integron gene (intI1) 43 .
Most of the ARGs in the influent are transferred to the activated sludge 19 . The distribution of ARGs in sludge samples was similar to that in Inf samples; macB was the most abundant, followed by patA, mfd, rpoC, and oleC. ANOVA showed that there was no significant difference in the abundance of ARGs between the two kinds of activated sludge (p > 0.05). Thus, it was inferred that the ARGs in activated sludge are not affected by the process when treating the same influent. Recently, a study showed that changes in microbial communities directly lead to changes in ARGs 17 . Relevant studies have also shown that bacterial abundance is an important factor affecting ARGs; the higher the abundance is, the greater the associated ARG abundance 2,18 . Direct evidence is provided by the fact that OD and MBR activated sludge have similar bacterial community distributions under the same influent conditions.
Most of the ARGs are located on MGEs, which makes it possible for microorganisms in the distant taxonomic lineage to obtain ARGs 45 . Plasmids are considered to be the main carriers of ARGs. Che et al. found that in addition to multidrug resistance, all other ARGs detected in three WWTPs in China included aminoglycosides, macrolide-lincosamide-streptogramin (MLS), β-lactam, tetracycline, chloramphenicol, quinolone and trimethoprim, which are mostly carried by plasmids and provide convenience for ARGs migration and transformation 46 . In this study, tetracycline (tetQ, tetO and tetM), beta-lactam (CfxA6, CfxA3), and MLS (ErmF, ErmG) were carried by integrative and conjugative elements (ICEs), which can lead to multiple resistance to a variety of gram-positive and gram-negative pathogens 47 . The ARGs on the chromosome cannot move themselves, and horizontal transfer can be achieved if they are integrated into MGEs 46 . Some ARGs, such as macrolide (macB), quinolone (gyrA), sulfonamide (sul1), tetracycline (tetA, tetW), OXA-347, and MLS (ErmF), were confirmed to be located on chromosomes. These ARGs detected in this study run through the whole WWTP, and they bring hidden dangers to human health and ecosystems with the migration and transformation of MGEs.
Furthermore, we identified the existence of MGEs, such as plasmids, transposons, integrons and ISs, which, as carriers of ARGs, play an important role in the horizontal transfer of ARGs 17 . Transposons and integrons are generally believed to be the main acquisition drivers of ARGs by MGEs 53 . Among them, tnpA and intI1 are often regarded as important markers of MGEs and have frequently been found in different environmental media 48,49 . Quantitative analysis of detected MGEs found that the transposon tnpA was the most abundant, with 20.60 and 25.66 copies of MGEs per cell in sample O2 and sample M2, respectively. Li et al. suggested that tnpA is positively correlated with Bacteroidetes and MLSS 49 . In this study, Bacteroidetes in sample M2 (7.01%) was more abundant than that in sample O2 (4.01%), while the MLSS in sample M2 (9381 mg/L) was notably higher than that in sample O2 (4356 mg/L), which may be the reason that the abundance of tnpA in the MBR was higher. intI1, a common typical integron, also exhibits an intense positive correlation with MLSS 49 , which results in a slightly higher abundance in sample M2 (1.00 copies of MGEs per cell) than in sample O2 (0.80 copies of MGEs per cell). Other detected MGEs, such as IS91, ISCrsp1, istA and istB, had relatively high abundance. All of these are categorized as ISs, which are DNA fragments that can be transferred between different positions on the same chromosome or between different chromosomes and are highly diverse types of MGEs that are often detected in the activated sludge of WWTPs 17 . A significant positive correlation exists between ARGs and MGEs 50 , and these MGEs detected in the OD and MBR may lead to the propagation and transformation of ARGs.

Materials and methods
WWTP overview and sludge sample collection. The WWTP was located in Urumqi, Xinjiang, China.
The first phase adopts an OD process followed by a deep bed denitrification filter with a treatment volume of 8 × 10 4 m 3 /day. The second phase adopts an A 2 O + MBR process with a treatment volume of 12 × 10 4 m 3 /day. The two processes treat the same influent, which is mainly domestic sewage. The inlet and outlet water parameters were measured for BOD 5 (280 and 6 mg/L, respectively), COD (620 and 30 mg/L, respectively), SS (350 and 10 mg/L, respectively), NH 4 + -N (5 and 1.5 mg/L, respectively), TN (65 and 15 mg/L, respectively) and TP (7 and 0.3 mg/L, respectively). In order to strengthen the phosphorus removal effect, chemical agents were added at the back end of the OD process and the front end of the MBR process. The influent water was collected from the distribution well, the effluent water of the OD and MBR were collected from the outlet of the secondary www.nature.com/scientificreports/ sedimentation tank and the membrane tank, respectively, and activated sludge samples were collected from the aerobic sections of the OD and the MBR membrane tanks. Table 2 provides the particular sample acquisition information. Kit (Omega Bio-Tek, Norcross, GA, USA) was used to extract DNA from all samples in accordance with the operational manual. The DNA integrity was then determined by 1% agarose gel electrophoresis. The concentration and purity of the DNA were detected using TBS-380 and NanoDrop200 fluorometers, respectively. The DNA was broken into 400 bp fragments with the ultrasonic breaker Covaris M220 (Gene Company, China), and then the paired-end library was constructed by NEXTFLEX Rapid DNA-Seq (Bioo Scientific, Austin, TX, USA). First, the splice of the fragment was linked, and then the splice self-attachment fragment was removed by magnetic bead screening. Next, the library template was enriched by PCR amplification, and finally, the PCR product was recovered by magnetic beads to obtain the final library.

Measurement of pollution indicators
Metagenomic sequencing was performed on the Illumina NovaSeq/HiSeq Xten sequencing platform by Majorbio Bio-Pharmaceutical Technology Co., Ltd. (Shanghai, China). The original sequences obtained in this study were deposited in the Sequence Read Archive (SRA) database of National Center for Biotechnology Information (NCBI) under accession number SRP311506.
Bioinformatic analysis. First, fastp 51 was used to cut the 3′ and 5′ ends of the original adapter sequences.
After removing the splicing, the length of the adapter sequence was less than 50 bp, the average base quality value was less than 20, and the retained reads, containing N bases, were high-quality paired-end reads and single-end reads. Next, MEGAHIT 52 was used to assemble the optimized sequences, and contigs ≥ 300 bp were selected for final assembly. The assembled reads were used for further gene prediction and annotation. Open reading frame (ORF) prediction was performed with MetaGene 53 , and motifs ≥ 100 bp were selected and translated into amino acid sequences. A nonredundant gene catalog was constructed using CD-HIT 54 , and the identity and coverage were both 90%. Reads after quality control were mapped to the nonredundant gene catalog with 95% identity using SOAPaligner 55 , and the abundance information of the gene was counted in the corresponding sample. Taxonomic classification of metagenomes was analyzed by Kraken2 56 , and the relative abundance was estimated by Bracken 57 . The amino acid sequence of the nonredundant gene set was compared with the CARD database using DIAMOND (the BLASTP alignment parameter e-value was set to 1e −5 ), and then the corresponding annotation information of the antibiotic resistance function was obtained. Finally, the relative abundances of the ARGs were represented as ppm (i.e., sequences annotated as ARGs per one million sequences).
Quantification of MGEs. The original sequences were processed for quality control by fastp (v0.12.1) using the default parameters. The resulting filtered reads were then used as the inputs for pipeline ARGs-OAP (v2.0) 58 , which consists of a SARG database and uses a reference MGE database to replace SARG (v2.0) for quantitative analysis of MGEs in the samples 60 . The reference MGE database 59 contains 278 different annotated genes and more than 2000 unique sequences and is available at https:// github. com/ Katar iinaP arnan en/ MobileGenet-icElementDatabase. The number of MGE copies per cell was calculated by normalizing the abundance relative to the number of MGEs with MGE copies as the unit 58 .