Removal of antibiotic resistance genes in secondary effluent by slow filtration-NF process

Antibiotics have been widely used in medicine and animal husbandry since their discovery; however, antibiotic resistance genes (ARGs) are observed in humans and animals subjected to long-term use of antibiotics. As emerging environmental contaminants, ARGs have biological characteristics that can be replicated or spread, along with the physical and chemical characteristics that persist in the environment (Pruden et al. 2006). As a result, ARGs can accumulate and remain in the environment. The urban sewage treatment plants represent an important accumulation site of antibiotics and their metabolites produced by humans, thereby leading to the subsequent aggregation, evolution, horizontal transfer and proliferation of ARGs (Baquero et al. 2012). Compared with the influent water, the concentration of ARGs in the effluent of the sewage treatment plants is usually reduced (LaPara et al. 2011), which indicates that the sewage treatment processes lead to a certain reduction in the concentration of ARGs. Even so, the resistant bacteria, ARGs and integron coding genes (int I) of the ‘movable original’ that spread ARGs are frequently detected in the secondary effluent (Gao et al. 2012; Raizzo et al. 2013), thus underlining the need for secondary effluent regeneration.

In recent years, nanofiltration (NF) technology has been widely used in sewage recycling. It is usually used as an advanced treatment process to remove the residual organic matter and inorganic salts in the secondary effluent to ensure the quality of the sewage effluent. Fan et al. (2019) used the photocatalysis/activated carbon/NF combined process to treat the secondary effluent. It was observed that chemical oxygen demand (COD), dissolved organic carbon (DOC) and ultraviolet absorbance value obtained at 254 nm wavelength (UV₂₅₄) exhibited an obvious removal effect, with the removal rates of 45.7, 74.5 and 89.2%, respectively. Fu et al. (2020) used the ultrafiltration (UF)-NF double-membrane process to deeply treat the secondary effluent of a municipal sewage plant. At an NF influent flow rate of 4 L/min, the average removal rate of COD, NH₃-N and total phosphorus (TP) by the double-membrane process was 87, 68 and 96%, respectively. In another study, Shen et al. (2014) used a combined coagulation-UF-NF process to remove dibutyl phthalate, phthalate diester, dimethoate and atrazine from the secondary effluent. At an initial pH of 5, at 5 °C temperature and 0.5 MPa transmembrane pressure, the removal rate of the earlier-mentioned four pollutants was determined to be 86.6, 94.8, 95.1 and 88.6%, respectively. Overall, as an advanced treatment process, NF exhibits a high removal rate of pollutants, and the application of NF to remove ARGs is feasible.

However, NF also suffers from serious membrane pollution during use; thus, it is necessary to set up a pretreatment process. In recent years, the slow filtration process has received widespread attention owing to no use of any chemical reagents, low operating cost, simple operation and low maintenance. In addition, the slow filtration process is suitable for water with a turbidity lower than 10 NTU and can be used as an advanced treatment process for sewage (Haig et al. 2014). The studies have shown that the slow filtration process can grow biofilms on the surface of the filter media at very low filtration rates, thus leading to microbial degradation as well as removal of colloids, organic matter, ammonia nitrogen and trace pollutants (Tufenkji et al. 2002). The slow filtration process has been applied for the removal of the environmental pollutants pharmaceuticals and personal care products (PPCPs). The removal effect is noted to be superior, and the filtration rate has a significant impact on the removal effect (Li et al. 2018); thus, the slow filtration process can be applied for the removal of ARGs and also as a pretreatment process for NF. Although the filtration rate of the slow filtration process is usually 10–30 cm/h, a filtration rate of 5 cm/h was tested in this study to explore whether ARGs can be removed to the greatest extent at a low filtration rate. In addition, the effect of the different types of biofilms on the removal of ARGs has been rarely reported.

In summary, this research focuses on the typical ARGs in the secondary effluent, and explores the efficiency and mechanism of the combined process of slow filtration and low pressure NF on the removal of ARGs and organics, in order to provide theoretical support for the advanced treatment of reclaimed water by the combined process of slow filtration and low pressure NF.

In this study, school domestic sewage is diluted 10 times to simulate the secondary effluent (experimental raw water). The water quality indicators are shown in Table 1. The slow filter media are quartz sand and gravel, with an average particle size of 0.6 mm and 2.0 mm, respectively. The NF membrane material is polyamide, with the molecular weight cut-off of 300 Da.

The following equipment has been employed: ultra-micro spectrophotometer (NanoDrop-8000, United States), fluorescence quantitative polymerase chain reaction (PCR) instrument (CFX Manager Biorad, United States), Hach water quality detector (DR-6000, United States), total organic carbon analyzer (TOC-VCPH, Germany) and high-throughput sequencing platform (Illumina, USA).

The schematic diagram of the experimental setup for the combined slow filtration and NF process is shown in Figure 1.

The slow filtration device is composed of four filter columns, 65 cm in height and 8 cm in diameter. The height of gravel in each filter column is 5 cm, whereas the height of the quartz sand is 45 cm. The inlet of the filter column is controlled by a peristaltic pump, where a valve has been installed at the outlet to control the filtration rate. The filtration rates used in the experiment are 5, 10, and 20 cm/h. The cultivated biofilms are mainly composed of aerobic heterotrophic bacteria, mainly comprising nitrifying and denitrifying bacteria. The operating pressure in the NF process is 0.4 MPa, and the water sample after membrane filtration is collected in a beaker to determine the pollutant concentration.

An artificial inoculation of membrane has been used for this purpose (Li 2018). The three kinds of biofilm cultivation methods include the addition of a certain amount of the activated sludge, followed by passing into the synthetic sewage for domestication. The aerobic heterotrophic, nitrifying and denitrifying biofilms use the stable removal rates of COD_Cr, NH₄⁺-N and NO₃⁻-N as the criteria for judging the maturity of the biofilms. The preparation method of the synthetic wastewater to cultivate the different biofilms is as follows.

Cultivation of the aerobic heterotrophic bacterial biofilm (mg/L): C₆H₁₂O₆: 70, NH₄Cl: 14.5, MgSO₄·7H₂O: 5, NaCl: 2.5, FeCl₃·6H₂O: 2.4, CaCl₂H₁₂, POK₂H₁₂: 1.0 and pH: 7.3–8.0. Subsequently, 1 mL micronutrient solution is added. The composition of the micronutrient solution is (mg/L): H₃BO₃: 50, ZnSO₄·7H₂O: 106, CuSO₄·5H₂O: 56, MnSO₄·H₂O: 50, NaMoO₄·2H₂O: 10, AlCl₃: 50, CoCl₂·6H₂O: 50 and NiCl₂: 50.

Cultivation of the nitrifying bacteria biofilm (mg/L): NH₄Cl: 70, NaHCO₃: 300, NaCl: 50, MgSO₄·7H₂O: 40, KH₂PO₄: 37, K₂HPO₄: 20 and nutrient solution 1 mL. The ingredients of the nutrient solution include: FeSO₄·7H₂O: 700, CuSO₄·5H₂O: 1,500, CoCl₂·6H₂O: 200, CaCl₂: 2,000, NaMoO₄·2H₂O: 800, ZnSO₄·7H₂O: 1,100, MnSO₄·H₂O: 500, FeCl₃·6H₂O: 700 and EDTA: 5,000.

Cultivation of the denitrifying bacterial biofilm (mg/L): NaNO₃: 20, MgSO₄.·7H₂O: 20, K₂HPO₄: 10, KH₂PO₄: 10, sodium citrate (C₆H₅Na₃O₇): 115 and nutrient solution 1 mL. The formulation of the nutrient solution is the same as the solution used for culturing the aerobic heterotrophic bacterial biofilm.

The plasmid concentration containing the target gene in the 10³, 10⁴, 10⁵, 10⁶ and 10⁷ ng/μL series is diluted, and the corresponding gene copy number is calculated to generate a standard curve. Based on the standard curve, quantitative PCR is used to determine the copy numbers of the different ARGs (tet A, tet W, sul I and sul II), type I integrons (intI 1) and 16S rDNA.

An experimental water sample of 1 L is taken in an Erlenmeyer flask, of which an amount of 500 mL is directly filtered through a 0.22 μm membrane, and the concentration of ARGs retained by the membrane is used as the total concentration; 1 mg DNA degrading enzyme is added to the remaining 500 mL amount, followed by placing in a water bath shaker at a constant temperature of 25 °C. The contents are allowed to react for 2 h, followed by keeping the contents at 80 °C for 10 min. At this stage, the free ARGs in the water sample have basically degraded, and the water sample is filtered through a 0.22 μm membrane to retain the cellular ARGs in the water sample. The concentration of free ARGs denotes the difference between the total concentration of ARGs and concentration of cellular ARGs.

During the experiment, the biofilm samples are collected once the biofilm is matured and after passing through the secondary effluent, followed by storage in a 4 °C constant temperature refrigerator. During sequencing, DNA is extracted from the sample, and the V3 and V4 regions of the bacterial 16S rDNA gene are subsequently amplified using primers 338F and 806R. The PCR amplification process is as follows: preheating at 95 °C for 5 min, denaturation at 95 °C for 30 s, annealing at 55 °C for 30 s and extension at 72 °C for 1.5 min. The denaturation to extension processes are cycled 30 times, followed by the final extension at 72 °C for 7 min. 0.8% (w/v) agarose gel electrophoresis is used to check the PCR products as well as to recover them with a gel extraction kit. The amplified genes are analyzed using the Illumina HiseqV4 PE250 (Illumina, USA) high-throughput sequencing platform, and the gene sequences of the different biofilms are obtained.

The removal of four ARGs (tet A, tet W, sul I and sul II), intI 1 and 16S rDNA on different slow filtration columns as a function of the filtration rate is shown in Figure 2 and supplementary table S1.

The average concentrations of tet A, tet W, sul I, sul II, intI 1 and 16S rDNA in the secondary effluent are determined to be 7.76 × 10⁸, 6.24 × 10⁸, 7.88 × 10⁸, 2.70 × 10⁸, 7.34 × 10⁸ and 1.94 × 10⁸ copies/mL, respectively. It is observed that the wastewater treatment plant cannot completely remove the ARGs in the wastewater, thus resulting in a high concentration of ARGs in the secondary effluent.

As observed from Figure 2(a), in the absence of any biofilm in the filter column, the removal of ARGs, intI 1 and 16S rDNA is not optimal, with the removal amount ranging between 0.26 and 1.22-log. In addition, the filtration rate is observed to have almost no effect on the removal of ARGs, as the retention effect almost solely occurs on the surface of the filter column and is not related to the filtration rate. As the influent water flows through the filter column, the ARGs and resistant bacteria adsorbed on the large-size suspended particles are intercepted, whereas the small particles with a particle size smaller than the gap between the sand particles pass through the filter column and are not effectively removed.

As observed from Figure 2(b)–2(d), in the presence of a mature biofilm in the filter column, the extent of removal of ARGs is increased. This is due to the reason that the sediment covering the quartz sand gradually reduces the opening of the gap during the formation of the biofilm, thereby further improving the interception efficiency. As the biofilm matures, the extracellular polymers secreted by the microorganisms also have a certain viscosity, which enhances the removal of ARGs. At the same filtration rate, the removal of ARGs by the different biofilms is observed to be slightly different. Overall, the slow filtration with the aerobic heterotrophic biofilm exhibits the best ARGs removal efficiency.

In addition, for the same kind of biofilm, as the filtration rate is increased from 5 cm/h to 20 cm/h, the removal of ARGs by slow filtration increases, and the removal effect is noted to be the optimal at 20 cm/h. Further, the filtration rate is noted to have a significant impact on the removal of ARGs. In this respect, the slow filtration with the aerobic heterotrophic biofilm exhibits the highest ARGs removal at a filtration rate of 20 cm/h. Generally, the level of bacterial activity decreases as the sand depth increases; however, it continues to a depth of 400 mm (Ellis 1985), which is almost the same as the thickness of the filter layer (450 mm) used in this study. As the filtration rate increases, more nutrients in the influent are carried to the deep layers of the filter column, which promotes the maturation of the deep biofilms; thus, an increment in the filtration rate is conducive for the removal of ARGs (Langenbach et al. 2009).

The presence of ARGs in the secondary effluent can be classified as cellular and free states (Zhang et al. 2017). As ARGs have the ability to survive longer than their bacterial hosts, the released free ARGs can still survive after the antibiotic-resistant bacteria are completely inactivated (Dodd 2012). These can also be transferred to the other recipient bacteria through the horizontal gene transfer, which is an important factor leading to the spreading of ARGs in the environment, caused by the effluent of the sewage treatment plants. During the experiment, at a filtration rate of 20 cm/h, the cellular and free ARGs in the effluent from the different slow filtration columns have been detected, and the results are shown in Figure 3 and supplementary table S2.

As observed from Figure 3(a), the concentration of the cellular states tet A, tet W, sul I and sul II in the secondary effluent is determined to be 5.40 × 10⁷, 8.52 × 10⁷, 1.58 × 10⁸ and 1.79 × 10⁸ copies/mL, while the concentration of the free state tet A, tet W, sul I and sul II is noted to be 7.22 × 10⁸, 5.39 × 10⁸, 6.30 × 10⁸ and 9.14 × 10⁷ copies/mL, respectively. Except for sul II, the concentration of the free ARGs in the secondary effluent is noted to be higher than the concentration of the cellular ARGs. This is due to the fact that the cellular ARGs settle more easily with the activated sludge in the secondary sedimentation tank than the free ARGs. Thus, once the anti-free ARGs are released after apoptosis and lysis of the bacteria, the concentration of the free ARGs in the secondary effluent is increased.

As can be observed from Figure 3(b), after slow filtration treatment without any biofilm, the concentration of tet A, tet W, sul I and sul II in the cell state is decreased by 0.66–1.54-log, while the concentration of these ARGs in the free state declines by 0.50–1.02-log. It indicates that the physical interception removes a high extent of the cellular ARGs, owing to their large size, leading to swift trapping.

As observed from Figure 3(c)–3(e), after slow filtration with the aerobic heterotrophic biofilm, the removal of the cellular and free tet A, tet W, sul I and sul II is increased by 0.40–1.98-log and 1.23–2.05-log, respectively, as compared with the slow filtration without any biofilm. For the slow filtration with the nitrifying biofilm, the increase is noted to be 0.75–1.50-log and 0.85–1.95-log, respectively. For the slow filtration with the denitrifying biofilm, the increment is determined to be 0.11–2.09-log and 0.87–2.22-log, respectively. Obviously, the presence of the biofilm significantly increases the removal of the free ARGs, as the presence of the biofilm narrows the gap between the sand grains, thus enhancing the retention efficiency. In addition, the adsorption and degradation of microorganisms also increases the removal of the free ARGs.

At a slow filtration rate of 20 cm/h, the removal of ARGs by direct NF, slow filtration without biofilm-NF and slow filtration with the aerobic heterotrophic biofilm-NF is shown in Figure 4 and supplementary table S3.

As observed from Figure 4, the extent of removal of tet A, tet W, sul I, sul II, intI 1 and 16S rDNA by direct NF in the secondary effluent is 4.75-log and 4.79-log, 4.57-log, 3.58-log, 4.18-log and 4.21-log, respectively. Further, it can be observed that NF exhibits an optimal effect on the removal of ARGs. For slow filtration without any biofilm combined with NF, the removal amount of ARGs, intI 1 and 16S rDNA is slightly increased, reaching 4.81-log, 4.73-log, 4.73-log, 4.22-log, 4.19-log and 4.23-log, respectively. If the aerobic heterotrophic biofilm grows on the surface of the filter material, the extent of removal of tet A, tet W, sul I, sul II, intI 1 and 16S rDNA by the combined process is 5.52-log, 4.83-log, 5.67-log, 5.27-log, 5.12-log and 4.30-log, respectively. Moreover, the removal effect is observed to be significantly higher as compared to the slow filtration without biofilm-NF, thus indicating that the biofilm plays an important role.

Compared with UF and microfiltration (MF), NF has a smaller membrane pore size. The molecular weight of ARGs is usually in the range of 100–9,000,000 Da (Rizzo et al. 2013), and ARGs with a molecular weight greater than the pore size of the membrane are easily trapped by the NF membrane. Combining the physical retention of slow filtration and adsorption of biofilms enables the combined process to achieve a highly-efficient removal of ARGs.

During the experiment, the biofilm on the surface of the filter material has been examined for the microbial population structure before and after the secondary effluent treatment. Further, the microbial population structure has been compared at the phylum and genus levels, as shown in Tables 2 and 3.

From Table 2, at the phylum level, for the dominant bacterial phyla, the proportion of Proteobacteria increases from 55.77% to 58.08% after the aerobic heterotrophic biofilm is passed through the secondary effluent. In the nitrifying biofilms, this proportion is observed to increase from 31.95 to 36.54%. In addition, in the denitrifying biofilms, the observed proportion increases from 57.12 to 68.12%. The bacteria included in the phylum Proteobacteria (the largest phylum of bacteria) are the gram-negative bacteria, while sul I and sul II usually use the gram-negative bacteria as the host bacteria. As a result, an increment in the proportion of the phylum Proteobacteria is beneficial to absorb sul I and sul II in the influent water, and the filter column reduces the sulfa resistance genes by removing the bacteria.

The gram-positive bacteria are the host bacteria of tet A and tet W. Among the gram-positive bacteria, both can exist stably and are transferable. Actinobacteria, which are gram-positive bacteria, are noted to increase from 1.08 to 1.94% after the aerobic heterotrophic biofilm is passed through the secondary effluent, and the percentage of Firmicutes increases from 0.06 to 0.57%. Similarly, in the nitrifying biofilms, the ratio of the two types of bacteria increases from 1.31 to 1.77%, and from 0.31 to 0.98%, respectively. However, in the denitrifying biofilms, the proportion of the two types of bacteria is noted to decrease. The enhancement in the proportion of the Lans-positive bacteria is conducive for the removal of the tetracycline resistance genes (Roberts 2005).

The studies have shown that Pseudomonas is a potential host of tetracycline ARGs and intI 1 (Xu et al. 2021). Ma et al. (2019) indicated that Acinetobacter is a potential host of tet W and intI 1, and sulfa ARGs are generally easily combined with intI 1. As observed from Table 3, at the genus level, the proportion of Pseudomonas in the aerobic heterotrophic biofilm after passing through the secondary effluent increases from 0.26 to 0.31%, whereas the proportion of Acinetobacter increases from 0.02 to 0.16%. The proportion of Pseudomonas in the nitrifying biofilms improves from 0.07 to 0.13%, while the proportion of Pseudomonas in the denitrifying biofilms is observed to decrease. Therefore, an increment in the number of Pseudomonas and Acinetobacter is conducive for the removal of tet and sul ARGs. In summary, the slow filtration with the aerobic heterotrophic biofilm exhibits the optimal removal of ARGs.

Langenbach et al. (2009) investigated the removal of Escherichia coli and Enterococcus by using the slow sand filter column at a filtration rate of 5, 10, and 20 cm/h. It was observed that the removal rate of both Escherichia coli and Enterococcus was higher at 20 cm/h as compared to the other rates. Malzer (2005) also observed that on increasing the hydraulic load from 1 m/d to 5 m/d, the removal of E. coli increased from 2.3-log to 3.7-log. During the experiment, in order to explore the effect of the high filtration rate on the removal of ARGs and organic matter, the microbial population structure at 10 and 40 cm was determined after the aerobic heterotrophic filter column was passed through the secondary effluent. The corresponding results are shown in Tables 4 and 5.

As observed from Tables 2 and 4, at the phylum level, the microbial populations at different heights are dominated by Proteobacteria, accounting for 58.08% (0 cm), 39.62% (10 cm) and 48.04% (40 cm). Further, as noted from Tables 3 and 5, at the genus level, the microbial population structures at different heights exhibit a slight difference, while the proportions of the different genera are different. However, the aerobic genera are noted to still dominate, such as Spirulina at 0 cm (Nitrospira, 9.58%), Acinetobacter (5.31%) at 10 cm and Bdellovibrio (9.92%) at 40 cm. The presence of these bacterial genera indicates that the deep layer of the filter column is rich in oxygen. Therefore, as the filtration rate increases, the organic matter in the influent will be carried to the deep layers of the filter column, which is conducive for the degradation of the organic matter by the aerobic heterotrophic bacteria, thereby promoting their own growth and reproduction, along with enhancing the removal of ARGs.

The removal of DOC by using different slow filtration columns as a function of the filtration rate has been investigated, as shown in Figure 5 and supplementary table S4. Under the condition of a slow filtration rate of 20 cm/h, the DOC removal by employing the combined process is shown in Table 6.

From Figure 5, the concentration of DOC in the secondary effluent is determined to be 11.4 mg/L. After slow filtration without any biofilm, the average DOC removal rate at the filtration rates of 5, 10, and 20 cm/h is observed to be 31.6, 28.1 and 27.2%, respectively. The filtration rate has an insignificant effect on the removal of DOC, as the physical retention only occurs on the filter surface and is not related to the filtration rate.

As the nitrifying bacteria are the autotrophic microorganisms, the DOC removal by the nitrifying filter column basically depends on the retention of quartz sand and adsorption of microorganisms. Among these, the heterotrophic microorganisms also make a certain contribution to the removal of DOC. For the aerobic heterotrophic and denitrifying bacteria, as the growth and reproduction of both require organic matter, DOC can be consumed through catabolism. Thus, the DOC removal rate is higher than that of the nitrifying bacteria. In the filter column with the biofilm, as the filtration rate increases, the removal rate of DOC also increases gradually. The slow filtration with the aerobic heterotrophic, nitrification and denitrification biofilms at a filtration rate of 20 cm/h exhibits the highest DOC removal rates, with the removal rates of 68.4, 40.4 and 65.8%, respectively. This is due to the reason that as the filtration rate increases, more nutrients in the influent will be carried to the deep layers of the filter column, which promotes the growth and reproduction of the deep-layer microorganisms.

As observed from Table 6, the DOC concentrations of the secondary effluent, water with the slow filtration without any biofilm, and water with the slow filtration with the aerobic heterotrophic biofilm after NF treatment are 2.2, 2.1 and 2.0 mg/L, respectively, with the removal rates of 80.7, 81.6 and 82.5%. As NF mainly relies on size exclusion to remove the organics, the macromolecular organics with a size larger than the membrane pore size in the influent will be retained. On the other hand, the small molecular organics will pass through the membrane pores, so that the effluent will still contain a certain concentration of DOC (Chao 2013).

The studies have shown (Zhang et al. 2018) that there is an interaction between the reduction of ARGs in the secondary effluent and removal of total microorganisms, integrons and organic matter. In this study, through the fitting analysis, the correlation between the total amount of 16S rDNA, type I integron intI 1, DOC and ARG removal was quantitatively described (the basis for determining the significance of the linear correlation is: P < 0.05), as shown in Figure 6 and supplementary table S5.

As observed from Figure 6(a), the concentration of the ARGs has a significant correlation with the concentration of 16S rDNA, indicating that the bacterial content in the secondary effluent affects the transfer and spread of ARGs in the environment. The previous literature studies have shown that the presence of ARGs in sewage can be classified as the cellular and free states (Zhang et al. 2017). Therefore, the concentration of ARGs can be reduced by removing the resistant bacteria (cellular ARGs). The free ARGs can be passed to other bacteria through HGT, followed by their removal.

As a mobile genetic element, intI 1 plays an important role in the transfer or spread of ARGs in the horizontal direction (Yang et al. 2013). As shown in Figure 6(b), there is a significant correlation between the concentration of ARGs and intI 1, among which the correlation between sul I and intI 1 is noted to be the strongest. This indicates that the ARGs may be combined to result in the type I integration in order to carry out the horizontal transfer in the secondary effluent.

The fitting results of DOC and ARGs (Figure 6(c)) reveal a significant correlation between the concentration of ARGs and DOC, indicating that the concentration of ARGs in water decreases with the removal of DOC. Riquelme Breazeal et al. (2013) reported that the total organic carbon, protein and polysaccharide concentrations in the membrane effluent were significantly correlated with the concentration of ARGs during the removal of ARGs by membranes, indicating the co-removal of DOC and ARGs. As DOC is an essential nutrient for the survival of heterotrophic bacteria, a decline in its concentration will inevitably inhibit bacterial growth and reproduction. It will subsequently also inhibit the spread of ARGs, thereby resulting in a reduction in the concentration of ARGs in the effluent.

1. The optimal filtration rate for the slow filtration with the aerobic heterotrophic biofilm, slow filtration with the nitrification biofilm, and slow filtration with the denitrification biofilm to remove different ARGs (tet A, tet W, sul I and sul II) is 20 cm/h. Among these, the slow filtration effect of the aerobic heterotrophic biofilm is noted to be best, with the removal amounts of 2.50-log, 2.96-log, 1.92-log and 2.33-log, respectively. The slow filtration without any biofilm can remove a high extent of the cellular ARGs, while the presence of the biofilm significantly enhances the removal of the free ARGs. Compared with the direct NF and biofilm-free slow filtration-NF, the aerobic heterotrophic biofilm slow filtration-NF exhibits the highest removal of ARGs, with removal amounts determined to be 5.52-log, 4.83-log, 5.67-log and 5.27-log, respectively.

2. Proteobacteria, Actinomycetes, Firmicutes, Pseudomonas and Acinetobacter species are noted to be increased the most in the aerobic heterotrophic biofilms, thus enabling significant ARGs removal. The microbial population structure of the aerobic heterotrophic biofilm filter column at different heights indicates that the high filtration rates promote the maturation of the deep biofilm, which is beneficial for the reduction of ARGs.

3. The optimal filtration rate for the biofilm-based slow filtration to remove DOC is noted to be 20 cm/h. Among these, the slow filtration with the aerobic heterotrophic biofilm exhibits the best removal effect, with a removal rate of 68.4%. As compared with the direct NF of the secondary effluent and slow filtration without biofilm-NF, the slow filtration with the aerobic heterotrophic biofilm-NF demonstrates the highest removal rate of DOC, with a removal rate of 82.5%.

4. 16S rDNA, intI 1 and DOC in water have a significantly positive correlation with the different ARGs; thus, the removal of 16S rDNA, intI 1 and DOC is conducive for reducing the ARGs amounts.

The research was supported by the National Natural Science Foundation of China (Grant No. 52070011 and No. 51678027) and the Fundamental Research Funds for Beijing University of Civil Engineering and Architecture (Grant No. X18025).

All relevant data are included in the paper or its Supplementary Information.

The table in the supplementary file is the specific data information in Figure 2 to Figure 6 in the article.

The file “Revision Notes” is a file used to respond to reviewers' comments, and is not a subsidiary file of the main article.