Introduction

Water scarcity is a pressing global challenge1. In recent years, the accelerated carbon neutrality process contributed to an increasing installation of renewable energy, which consume more water producing the same electric power2. As a consequence, this worldwide action against climate change will worsen the water shortages problem3. Meanwhile, lots of countries and regions continue to face water contamination4. Water reclamation is a win-win strategy for increasing freshwater supply and shortening the water footprint of human beings5,6,7.

Reverse osmosis (RO) is one of the most applicable and stable units for high-quality reclaimed water production in industrial reuse, potable reuse, and groundwater replenishment8,9,10,11,12. Many large-scale water reclamation plants have been successively operated13,14,15. However, RO system has an inconvenient defect, that is, the complex and rebellious membrane fouling, which leads to the surge of operating cost16,17. Membrane fouling of RO mainly includes scaling, colloidal fouling, organic fouling, and biofouling18. Among these, biofouling is the most complicated and noteworthy one one16,19, although numerous studies have made significant efforts to reduce it20,21,22,23.

Disinfection is a widely applied pretreatment process used to deal with biofouling in RO systems. However, it may lead to undesirable effects22. After reducing the number of bacteria in the feed water, the disinfection process exerts a salient selection effect on the bacterial community and the extracellular polymeric substance (EPS) secreting ability of the bacteria13,22,24,25,26. Some unwanted bacteria, that are resistant to disinfection or can adapt to adverse environments, might survive disinfection processes and become disinfection-residual-bacteria (DRB)27,28. DRB might possess a higher proportion of bacteria with higher EPS secreting ability, leading to a more severe biofouling of RO membranes, especially in the long-term operation of RO systems29. Research in laboratory and full-scale water treatment plants has shown the probability of aggravated biofouling after disinfection24,25,26,30. However, most of these studies are limited to a single type of disinfection process31,32. To date, there is still a lack of systematic and broad comparison of the biofouling-control effects of various disinfection methods.

To provide more constructive and reliable guidance on the prevention of RO biofouling, this study compared the DRB characteristics of seven different disinfection methods, including five widely used disinfection methods (NaClO, NH2Cl, ClO2, UV, and O3) and two recently developed disinfection methods (K2FeO4 and flow-through electrode system (FES)33) via a long-term (32 days) biofilm cultivation experiment on RO membranes. Furthermore, the bacteria and organic matter in the biofilm on the RO membranes were analysed using heterotrophic plate count (HPC), adenosine triphosphate (ATP), dissolved organic matter (DOM), and fluorescence excitation-emission matrix (EEM) to determine the primary reasons for the differences in biofouling in each group. The microbial community structure of DRB biofilms was analysed using high-throughput sequencing. Based on these results, the correlations among the fouling characteristics of DRB biofilms and bacterial density, organic density, biofilm morphology, and the microbial community of DRB were analysed.

Results

Effects of DRB biofilm on the RO membrane performance

The effects of different water disinfection methods on the RO membrane flux of the DRB biofilms were investigated. Not all disinfection methods can control biofouling. These methods were divided into three groups, according to the normalised flux (Fig. 1a). The first group, classified as “alleviated”, included K2FeO4 and O3. Their DRB performed light normalised flux drop (22.4 ± 2.4% and 23.9 ± 1.17%, respectively) compared to the control group (without disinfection pre-treatment), thereby indicating alleviated biofouling potentials. Meanwhile, the DRB of NH2Cl had a similar biofouling degree as the control group (~27%), and was classified as “equal”. Finally, the DRB of the “aggravated” group, which included FES, UV, NaClO, and ClO2, caused more biofouling of the RO membrane than the control group (29.1 ± 0.3%, 33.3 ± 7.8%, 34.6 ± 6.4%, and 35.5 ± 4.0% flux drop, respectively). The two most frequently used disinfection methods, namely UV and NaClO aggravated the biofouling of RO membranes, matching the results of long-term studies in previous research24,26. The initial flux values were shown in Supplementary Table 1.

Fig. 1: Fouling degree of RO membranes with DRB biofilm (32 days) after seven different kinds of disinfection treatment.
figure 1

a The normalised flux of fouled membranes and b the hydraulic resistance of the biofilms. The experiment was conducted in parallel in three sets. Error bars represent the standard deviation for three independent experiments. The p value was generated by t test (*p < 0.05, **p < 0.01).

The hydraulic resistance of the DRB biofilms (Rm) is shown in Fig. 1b. The hydraulic resistance of the fouling layer is consistent with the flux drop data (R2 = 0.91, p < 0.05). The resistance of the K2FeO4-DRB biofilm (2.48 × 1013 m−1) was the lowest as it had the least impact on the RO membrane flux drop. Also, the resistance of the O3-DRB biofilm (2.71 × 1013 m−1) in the “alleviated” group was lower than that of the control group (3.07 × 1013 m−1). The resistance of the NH2Cl-DRB biofilm (3.48 × 1013 m−1) was slightly higher than that of the control group. The DRB biofilm resistance of all the “aggravated” groups was much higher than that of the control group. Among them, the average resistance of NaClO was the highest, at 5.18 × 1013 m−1. However, it had a relatively high within-group error, corresponding to the membrane flux drop shown in Fig. 1a.

Disinfection effect of seven kinds of disinfection methods

Alleviation of biofouling after disinfection might be associated with lower bacterial concentrations in the feed water after disinfection24. However, the bacterial concentrations in the feed water could not explain the difference of flux drop in this long-term experiment. Variations in the concentrations of HPC and ATP in the water samples during the disinfection processes were tested (Table 1). The five conventional water disinfection methods achieved a bacterial inactivation rate of 3-log, while K2FeO4 and FES exhibited a relatively low inactivation effect (~1-log). Overall, HPC had no monotonous correlation with biofouling potential. The DRB biofilm of K2FeO4 caused only a 22% decrease in membrane flux, although the inactivation effect of K2FeO4 was the lowest (<1-log). NaClO was the most effective method for bacterial inactivation. However, its DRB aggravated biofouling at an average relative flux drop of ~35%.

Table 1 Disinfection effect measured by HPC and ATP content.

Previous studies have reported that ATP can be an effective indicator for predicting biofouling potentials in the short- or medium-term operation of RO systems (less than 16 days)18,34. However, this conclusion is not valid in long-term experiments (Table 1). The intracellular ATP in the control group was 2.90 ng/mL, accounting for 97% of the total ATP. Total and intracellular ATP concentrations increased after UV treatment, partly because of the rapid DNA repair mechanism triggered by UV radiation. No correlation was found between the DRB biofouling potentials (membrane flux drop) and the ATP concentration of DRB (Supplementary Fig. 1a, p > 0.05). Evaluation of the water disinfection effect suggested that the amount and the activity of DRB were not the decisive factors of biofouling potential during long-term operation.

Characteristics of the DRB biofilm on RO membranes

To determine the reasons for varying performances of RO membranes fouled by different DRB biofilms, this study dissected the fouled RO membranes and performed a series of tests. The number of bacteria and organics in the DRB biofilms, the morphological characteristics of DRB biofilms, and the bacterial community structure were analysed.

The live bacterial density of the DRB biofilm was measured using HPC, as shown in Supplementary Fig. 1b. DRB biofilms of NH2Cl, FES, and UV possessed the maximum live bacteria density (approximately 106 CFU/cm2). The live bacterial density of NaClO- and ClO2-DRB biofilms were lower than that of the control group (approximately 105 CFU/cm2), but they led to the worst flux drop in the RO membrane. Hence, the live bacterial density of the DRB biofilm was not a key factor leading to differences in biofouling degrees. This conclusion is consistent with previous studies26.

Unlike bacterial cells, it has been reported that the extracellular polymeric substances (EPS) on RO membranes have a more direct relationship with flux drop35,36. Hence, the total amount of organics in the DRB biofilms was measured by DOM (Fig. 2a), and the different component of DOM was tested using EEM (Supplementary Fig. 2). The amount of DOM did not show a clear correlation with the biofouling degree. The DOM of K2FeO4, which was in the “alleviated” group, was much higher than that of the “aggravated” group. Therefore, the total amount of organics did not play a key factor in the degree of DRB biofilm fouling.

Fig. 2: DOM density in the DRB biofilms (32 days).
figure 2

a DOM content and b the ratio of the ratio of DOM to HPC in the biofilms. The experiment was conducted in parallel in three sets. Error bars represent the standard deviation for three independent experiments. The p value was generated by t test (*p < 0.05, **p < 0.01).

Considering that the ratio of EPS to bacteria could affect the biofouling degree36, this study calculated the ratio of DOM to HPC in the DRB biofilms (Fig. 2b). The DOM/HPC ratio in the control group was approximately 4 mg/107 live cells. The DRB biofilm of K2FeO4, O3, NH2Cl, FES, and UV possessed a relatively lower DOM/HPC ratio, while those of NaClO and ClO2 (5.4 and 7.6 mg/107 live cell, respectively) were significantly higher than the control group (p < 0.05). This implied that these two disinfectants could aggravate biofouling of the RO membrane by changing the microbial community and changing the EPS production capacity of the residual bacteria. The DOM/HPC ratio could only partly explained the aggravated biofouling of NaClO and ClO2 group, it has no statistical association with the flux drop value overall (R2 < 0.5, p > 0.05).

The proportions of DOM components among all the groups were similar (Supplementary Fig. 2). Tyrosine/tryptophan amide (Zone I) and protein-containing tyrosine/tryptophan (Zone II) were the dominant components of fluorescent organic matter in each DRB biofilm, indicating that amino acids and proteins were predominant in the DRB biofilms, compared with polysaccharides, fulvic acids, or humid acids.

Compared to the amount of bacteria and organics in the biofilm, the degree of biofouling was more closely correlated with their arrangement and accumulation condition, which resulting in the thickness and consecutiveness of the biofilm on RO membranes36. The thickness of the DRB biofilm was tested using z-stack images of the LSCM. The surface image of the DRB biofilm showed that bacteria, proteins, and α-/β-polysaccharides were evenly distributed in the DRB biofilm (Supplementary Fig. 3). The DRB biofilm of K2FeO4 was relatively loose in section view, while the others were consecutive (Supplementary Fig. 4). The average thickness of the DRB biofilm was measured via cross-section (Supplementary Fig. 4) and is shown in Fig. 3. As a result, the DRB biofilm of the UV group (55 ± 1 μm) was significantly thicker than that of the control group (33 ± 1 μm) (p < 0.01). The DRB biofilm thickness of the “alleviated” group was approximately 22 μm, which was the lowest of all the groups. The biofilm thickness partly illustrated the difference in the biofouling degree, supplemented by the DOM/HPC ratio. The two disinfection methods in the “alleviated” group caused a low DOM/HPC ratio and thinner biofilm on the RO membrane, leading to a marginal flux drop of the fouled membrane. The UV DRB developed a thick biofilm and led to severe biofouling of the RO membrane. This association was not statistical significance (p > 0.05, R2 < 0.5). There were also counter examples. For instance, the biofilms of NaClO and ClO2 were not very thick. However, the high DOM/HPC ratio in the biofilm could narrow the water channels between bacterial cells36, leading to the highest flux drop in the RO membrane. Therefore, neither the thickness or the density of biomass in the biofilm was the decisive factor. These two factors jointly affected the degree of RO membrane fouling.

Fig. 3: Thickness of DRB biofilms (32 days).
figure 3

The thickness of biofilms were measured with laser scanning confocal microscope (LSCM). Ten points were randomly taken and averaged. Error bars represent the standard deviation for each independent experiments. The p value was generated by t test (*p < 0.05, **p < 0.01).

In addition, we also use electron microscopy to observe the morphology of biofilms. The FESEM images of the DRB biofilm are shown in Fig. 4. The DRB biofilm fully covered all the RO membranes. However, the compactness of the biofilms was different. The UV and NaClO-DRB biofilms were compact and continuous, partly illustrating their severe biofouling performance. In contrast, there were gaps between the bacteria and the EPS matrix in the DRB biofilm of the remaining groups. Regular crystals containing Fe were observed in the K2FeO4 group (Supplementary Fig. 5), indicating that K2FeO4 could cause scaling of the RO membrane.

Fig. 4: Scanning electron microscope (SEM) images of DRB biofilms.
figure 4

SEM images of DRB biofilms developed on the surface of RO membranes was taken by field-emission scanning electron microscopy. a control group, b K2FeO4 group, c O3 group, d NH2Cl group, e FES group, f UV group, g NaClO group, and h ClO2 group.

Microbial community analysis of DRB biofilm on RO membranes

Disinfection processes can exert three levels of change on the bacteria: metabolic change of a single bacteria, shift in the microbial community, and variation of nutrient conditions27. Among them, a shift in the microbial community is most likely to affect the biofouling degree during long-term operations37. Therefore, we analysed the alpha and beta diversities of the DRB microbial community.

The alpha diversity indices of the observed species as well as the ACE, Chao1, and Shannon indices of bacteria in the DRB biofilm are listed in Supplementary Table 2. The community richness and evenness of the control group were the highest, followed by K2FeO4 (“alleviated” group), which has been reported to have minimal impact on the bacterial community38. The community evenness of the ClO2-DRB biofilm (“aggravated” group) was the lowest, indicating a significant selection effect of ClO2. Overall, community richness and evenness did not show a monotonous correlation with the biofouling behaviour of DRB biofilms.

A Venn diagram of the OTUs that appeared in each group is shown in Supplementary Figure 6. Only 36 OTUs were shared by all the groups. The control group had the highest OTUs, indicating that each water disinfection method had a selection effect. Besides the control group, the number of OTUs in the K2FeO4-DRB biofilm was the highest. High numbers of OTUs caused fierce competition among the species, and inhibited biofilm growth and EPS secreting, resulting in the least biofouling of the K2FeO4-DRB biofilm. The DRB biofilm with ClO2 and O3 disinfection possessed the lowest OTUs, indicating that the two aforementioned oxidising disinfection methods had strong selectivity for the bacteria39. However, as their biofouling performance differed, ClO2 selected more biofilm formation and EPS-secreting species than O3.

The microbial community structure at the phylum level is shown in Supplementary Fig. 7a, a heat map at the genus level is shown in Supplementary Fig. 7b. The microbial community structure at class and genus levels are shown in Supplementary Fig. 8. α-Proteobacteria and γ-Proteobacteria were the dominant classes in all the DRB biofilms. Actinobacteria and Bacteroidetes were the second and third most abundant phyla, respectively. The top three phyla accounted for 90% of all bacteria.

Previously reported biofouling-related genera had significantly higher relative abundances in the “aggravated” group, namely, ClO2, NaClO, UV, and FES. For instance, Methylobacterium, which is a typical disinfection-resistant and biofouling-related bacteria genus24,36, was dominant in the DRB biofilm of FES and ClO2 with a relative abundance of 54.8 ± 2.3% and 28.6 ± 1.7%, respectively. Sphingobium, a highly secretory genus26,40, was found to dominate the DRB biofilm under UV (30.0 ± 0.8%). In addition, the relative abundance of Pseudomonas was significantly higher in the NaClO-DRB biofilm than in the other groups (45.9 ± 1.8%) (p < 0.05). Pseudomonas, which is a typical DRB of chlorine27 causes biofouling of RO membranes24,41,42,43. In contrast, the relative abundance of these genera was significantly lower in the K2FeO4-and O3-DRB biofilms (<10%) (p < 0.05). Thus, the relative abundance of highly secretory or biofouling-related genera plays a decisive role in the biofouling potential of DRB during long-term operation.

A community structure similarity analysis was performed using the PCA algorithm (Fig. 5). The community structures of all the disinfection groups were highly different from those of the control group. The bacterial community structures of UV-, O3-, NH2Cl-, NaClO-, and K2FeO4-DRB biofilms were similar, whereas the community structures of ClO2 and FES DRB biofilms were similar. This may be the cause of the inability of these two disinfectants to control the highly secretory genus Methylobacterium.

Fig. 5: Principal component analysis (PCA) of microbial communities.
figure 5

Microbial communities of DRB biofilms were displayed in a two-dimensional chart by PCA algorithm. n = 3 for all the experiment groups and the control group.

Apart from biofouling of RO membranes, the DRB biofilm can act as a shelter for pathogenic bacteria, leading to health risks of RO concentrate44. Thus, the cumulative relative abundance of pathogens and opportunistic pathogens in the DRB biofilm was analysed and shown in Supplementary Fig. 9. The relative abundance of pathogenic bacteria in the DRB biofilm increased significantly after disinfection with NaClO and ClO2 (p < 0.01), indicating that these two chlorine-containing oxidative disinfectants selected (opportunistic) pathogens. K2FeO4, FES, and NH2Cl partially reduced the relative abundance of (opportunistic) pathogens. Furthermore, their abundance in DRB biofilms of O3 and UV were the lowest, demonstrating that they controlled the spread of (opportunistic) pathogens in biofilms. Total read count of OTUs belonging to pathogens and opportunistic pathogens were listed in Supplementary Table 3, showing similar results to the relative abundance. The rising abundance of pathogens and opportunistic pathogens in the NaClO and ClO2 raised the concerns about potential health risks after these water disinfection processes. As the salient health concern and high social impact, further studies shall pay attention to the absolute abundance of potential pathogens, such as the cell number, gene copy number, and DNA concentration in each type of DRB biofilms.

Correlation between microbial community structure and RO membrane flux

To identify the key genera in the microbial community of DRB affecting the RO membrane flux, correlation coefficients between the normalised flux and relative abundance of the top 50 genera in all the experimental groups and the control group are shown in Fig. 6. The relative abundance of Methylobacterium (a typical disinfection-resistance and biofouling-related genus24,36) was negatively correlated with the normalised flux of the fouled RO membrane (p < 0.05). The relative abundances of two kinds of high EPS-secreting bacteria, Microbacterium and Pseudomonas, were also negatively correlated with the normalised flux. These two genera consist of typical chlorine-resistant and highly secretory bacteria. Hence, Methylobacterium, Microbacterium, and Pseudomonas deserve special attention as they play an essential role in the aggravated biofouling potential after disinfection.

Fig. 6: The correlation between the normalised flux of and relative abundance of the top 50 genera.
figure 6

Heat maps shows the correlation coefficient between normalised flux and relative abundance of the top 50 genera. The p value was generated by Pearson correlation analysis (*p < 0.1, **p < 0.05, ***p < 0.01). The colour indicated the correlation coefficient (dark red: 1, dark blue: −1).

Additionally, we calculated the accumulative relative abundance of typical highly secretory or biofouling-related genera reported in the literature, including Pseudomonas, Sphingomonas, Acinetobacter, Methylobacterium, Sphingobium, and Ralstonia24,25,26,40,41,45 (Fig. 7). The cumulative relative abundance of these bacteria in the “aggravated” group, namely the FES, UV-, NaClO-, and ClO2- DRB biofilms was significantly higher than that in the control group (~5%). Remarkably, the highly secretory bacteria accounted for over half of the total bacteria in the DRB biofilms of FES and NaClO, that is 56.4% and 53.0%, respectively.

Fig. 7: The relationship between normalised flux drop and relative abundance of typical highly secretory.
figure 7

a Accumulative relative abundance of typical highly secretory or biofouling-related genera reported by previous research and b its correlation with normalised flux drop of RO membranes. The correlation coefficient and p value were generated by Pearson correlation analysis (R2 = 0.64, p < 0.05).

In contrast, the DRB biofilm of the “alleviated” group possessed a similar proportion of highly secretory bacteria as that of the control group. The decrease in bacterial numbers after disinfection could explain biofouling alleviation. Therefore, variation in the relative abundance of typical highly secretory and biofouling-related genera was the main reason for the change in biofouling potentials after different disinfection processes.

Discussion

In the long-term experiment (32 days), DRB biofilms of seven types of water disinfection technics caused various flux dropped, which showed no significant correlation with the bacterial concentration in the feed water. DRB biofilm in the K2FeO4 and O3 group resulted in alleviated biofouling compared to the control group. Biofouling degree of the NH2Cl-DRB biofilm was similar to that of the control group, whereas the other four types of disinfection aggravated membrane biofouling. Furthermore, as ferrate introduced iron flocs and aggravated inorganic scaling in RO systems, O3 was recommended as a practical approach to prevent biofouling in RO systems.

The hydraulic resistance of the fouling layer was consistent with the flux drop value. The amount and metabolic of bacteria and the amount of organic matter in the biofilm did not explain the difference in fouling. Morphological analysis combined with the DOM/HPC ratio explained the difference in the fouling behaviour of the DRB biofilms. A high DOM/HPC ratio along with denser and thicker DRB biofilms led to severe biofouling.

Community analysis revealed that the selection effect of disinfection on typical highly secretory and biofouling-related genera was the primary reason leading to the ascending of the EPS secreting capacity of the total bacteria and resulting the high DOM/HPC ratio in the biofilm. The increased relative abundance of these typical bacteria was the radical mechanism for biofouling aggravation. Typical genera included Pseudomonas, Sphingomonas, Acinetobacter, Methylobacterium, Sphingobium, and Ralstonia.

All disinfection techniques except for NaClO and ClO2 effectively reduced the proportion of (opportunistic) pathogens in the DRB biofilm. The absolutely abundance of (opportunistic) pathogens need to be focused in further studies.

Methods

Water samples

Reclaimed water was sampled from a large-scale water reclaimed plant in Beijing, China. A schematic of the advanced treatment process is shown in Supplementary Fig. 10. The effluent from the denitrification filter was chosen as the sample because subsequent treatment units had a bacterial removal effect46. The reclaimed water samples were transported to the laboratory within 1 h, then filtered by filter papers to remove particles, and kept at 4 °C before disinfection. Water quality parameters of the sampled water were measured as soon as they arrived at the laboratory, and are listed in Supplementary Table 4.

Disinfection and biofilm culture

Five commonly used water disinfection methods, including NaClO, NH2Cl, ClO2, UV, and O3, and two recently developed water disinfection technologies, namely K2FeO4 and a flow-through electrode system (FES), were compared in this study. The steps followed in the experiment are shown in Supplementary Fig. 11. Briefly, water samples were filtered using filter paper (medium speed, Newstar, Hangzhou, China) to remove large flocs before disinfection. Seven types of disinfection processes were conducted following the technical parameters described in preliminary experiments (Table 2) to achieve similar bacterial log removal, except for two recently developed water disinfection methods as they could not achieve a high disinfection effect in actual wastewater. Square-wave alternating pulse current FES devices were set up based on a previous study33. The voltage amplitude and hydraulic retention time were set at 4 V and 27.7 s, respectively, to achieve the best disinfection performance of the system. UV irradiation was performed using a laboratory-scale collimated light-beam apparatus26,47. Other oxidising disinfection processes were performed in sterilised glass bottles at 25 °C and 150 rpm. The reaction was quenched with Na2S2O3 solution (10 mg/L, 10 min) to avoid oxidative damage to the RO membrane36,48,49.

Table 2 Dosage of each disinfection method.

Aromatic polyamide composite LP100 RO membrane (Vorton, China) was cut into round coupons (d = 32 mm, S = 804 mm2), and the pretreatment procedure was conducted based on a previous study50. For biofilm culturing, the membrane coupon was soaked in 18 mL of disinfected water or control samples in a 5 cm sterilised round plastic Petri dish at 25 °C. Water samples were refreshed daily. After 32 days of culture, the membrane was gently rinsed twice with phosphate buffer saline (PBS) to remove suspending bacteria for sequencing analysis. The experiments were conducted in triplicates for each group.

Evaluation of the water disinfection effect

The water disinfection effect was evaluated based on the concentration of culturable bacteria; ATP was tested before the biofilm culture experiments with triplicate experiments immediately after the disinfection processes were completed. Culturable bacteria were measured by HPC via colony-forming unit (CFU) counting46. ATP was tested using luminescence analysis. For total ATP measurement, 100 μL of the bacterial suspension and 100 μL of CellTiter-Glo Luminescent Cell Viability Assay (Promega, USA) were mixed in 96-well plates. After incubation at 25 °C and at 150 rpm in the dark for 1 min, luminescence intensity was measured using a microplate reader (SpectraMax M5, Molecular Devices, USA). For the extracellular ATP test, water samples were filtered through a 0.1 μm membrane (Millipore) to remove bacteria; the subsequent steps were the same as those for total ATP. Intracellular ATP concentration was calculated by subtracting the extracellular ATP concentration from the total ATP concentration.

RO cross-flow unit and membrane performance tests

A laboratory-scale cross-flow RO system was used to evaluate the performance of RO membranes before and after biofilm growth35. Briefly, the membrane compaction phase was performed with a 30 min rinse of ultra-pure water at 1.2 MPa until the permeate flux was stable. The flow rate of the influent was set at 1.0 mL/min via a constant flow pump (NPL-100). The crossflow velocity was set as 6.23 cm/s, which was relatively low value within the actual operating range, to protect the biofilm from cracking51,52,53,54. The feed water was then switched to a 500 mg/L NaCl solution (conductivity of approximately 1000 μs/cm). The permeate flux was recorded after reaching a constant value. The normalised flux was calculated by dividing the flux of the fouled membrane by the clean membrane before biofilm growth.

The resistance of the biofilm on the RO membrane was calculated, as follows:

$$R_b = R_T - R_m - R_P = \frac{{{\Delta}P}}{{\sigma J}} - R_m - R_P$$
(1)

Where, ΔP is the transmembrane pressure, σ is the kinetic viscosity of water, and J is the flux. The resistance of the biofilm Rb is calculated by subtracting the resistance of the virgin membrane Rm and the resistance of the concentration polarisation Rp from the resistance of the fouled membrane RT.

Biofilm analysis

Microbial amount and activity were determined using HPC and ATP concentrations. A piece of 10 mm × 10 mm fouled membrane was cut off and vortexed in 0.5 mL normal saline for 30 s. HPC and ATP concentrations were tested using the same procedure, as described in Section “Evaluation of the water disinfection effect”.

DOM and EEM were applied to reflect the characteristics of organic matter in DRB biofilms55. Briefly, a piece of 10 mm × 10 mm fouled membrane was cut and shaken in 5 mL NaOH solution (pH 12) for 24 h at 25 °C and 150 rpm. Then, HCl solution (pH 2) was added to adjust the pH of the solution to 7.0 ± 0.2. After neutralisation, the volume of the solution was adjusted to 15 mL by adding ultrapure water. The solution was filtered through a 0.45-μm nylon membrane (Whatman, England) before total organic carbon (TOC) measurement on a TOC-5000A analyser. EEM spectra were recorded using a fluorescence spectrophotometre (F-7100, Hitachi, Japan). The EEM spectrum was divided into six zones for integration. The types of fluorescent substances in each zone are shown in Supplementary Table 526,56.

A laser scanning confocal microscope (LSCM, LSM710META, Zeiss, Germany) was used to measure the thickness of the biofilm on the RO membrane. A 5 × 10 mm membrane with the DRB biofilm of each sample was cut for LSCM observation. The staining groups of the fluorescent dyes and their targets are listed in Supplementary Table 6. The thickness of biofilm was measured via the section view photos. Ten points were randomly taken and averaged.

The surface morphology of the DRB biofilm was examined using field-emission scanning electron microscopy (FESEM, SU8220, Hitachi, Japan) and the accelerating voltage was set to 5 kV. A piece of 5 × 5 mm membrane coupon was cut for observation.

Three pieces of 40 mm2 membrane coupons were cut for microbial community analysis. Microbial community analysis was conducted as previously described24. Briefly, DNA was extracted using the E.Z.N.A.® soil DNA Kit (Omega Bio-Tek, Norcross, GA, U.S.). Furthermore, bacterial specific primer 338F (5’-ACTCCTACGGGAGGCAGCAG-3’) and 806R (5’-GGACTACHVGGGTWTCTAAT-3’) were used as primer pairs for the amplification of hypervariable region V3-V4 of the bacterial 16S rRNA gene. Sequencing was performed using an Illumina MiSeq PE300 platform. The raw 16S rRNA gene data were demultiplexed, quality-filtered by Fastp version 0.20.0. All analysed sequences were submitted to the NCBI SRA database under accession number PRJNA803872. Data analysis and figure drawing was accomplished with the Bioinformatics Cloud platform of Majorbio (Shanghai, China). Pathogen identification was based on the Virulence Factor Database (VFDB) from the Institute of Pathogen Biology, CAMS & PUMC57.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.