Introduction

Gravity-driven membranes (GDMs) rely on the use of gravity rather than suction pumps, thus offering a simplified system design and lower energy consumption for water treatment systems. However, the low stable flux of GDMs (2–4 L/m²/h) makes them unsuitable for the treatment of highly concentrated wastewater or for use in large-scale applications. Consequently, as summarized in Table 1, the flux of GDM is highly sensitive to total organic carbon (TOC) and suspended solid (SS). Generally, higher TOC levels result in lower flux, leading to consistently low flux in wastewater applications. Therefore, GDMs have primarily been employed in small-scale water treatment systems for the treatment of influents with low TOC and chemical oxygen demand (COD) levels (12.5 mg TOC/L and 7.5–15 mg COD/L)1,2 or low SS levels, such as river water and grey water3,4,5. Recent research has focused on addressing the critical limitation of low flux in gravity-driven membrane (GDM) systems. Flux enhancement has been achieved through the implementation of intermittent aeration in gravity-driven membrane bioreactor (GD-MBR) systems to reduce cake layer resistance, or by employing biological pretreatment techniques such as granular activated carbon (GAC), modified fiber ball (MFB), and microfiltration (MF). These initiatives aim to improve the practical applicability of GDM6,7,8. In parallel, quorum quenching (QQ), a strategy aimed at disrupting bacterial quorum sensing (QS), has been extensively studied in conventional membrane bioreactors (MBRs) as an effective approach to mitigate membrane biofouling. The majority of these investigations have been conducted under constant-flux operational conditions employing suction pumps, wherein QQ microorganisms (e.g., bacteria or fungi) are immobilized within various carrier (e.g., beads, sheets, cylinders, or hollow cylinders) and introduced into the membrane tank to inhibit QS-mediated biofilm formation12,13.

Table 1 Summary of the stable flux for GDM systems fed with wastewater sources
Full size table

MBRs integrate membrane filtration with conventional activated sludge (CAS) processes to maintain higher mixed liquor suspended solid (MLSS) concentrations compared to traditional CAS systems. This integration results in enhanced treatment efficiency, with a constant flux of 20 L/m²/h or more, eliminates the need for secondary clarifiers for solid–liquid separation, reduces the operating footprint, and consistently produces high-quality treated water that is robust to seasonal and temporal variation. However, MBRs demand substantial energy input for pump operation, including both purchase and maintenance costs, to produce high-quality treated water. They also have high operating costs for air scouring and chemical cleaning to prevent membrane biofouling9,10,11. To date, no studies have investigated the impact of QQ in gravity-driven, constant-pressure GDM systems. This represents an innovative approach, as integrating QQ with GDM could offer new opportunities for advancing energy-efficient, low-biofouling MBRs. Therefore, this study aimed to reduce the energy required for the pumps in MBRs by integrating a GDM and mitigating membrane biofouling via QQ, which inhibits bacterial QS, a key contributor to biofilm formation12,13,14. The feasibility of the proposed GD-MBR was assessed by combining a GDM with a CAS process characterized by high MLSS levels (4000–7000 mg MLSS/L) for the treatment of high-concentration wastewater (135–270 mg TOC/L, 400–800 mg COD/L). An external chamber was installed between the feed and GD-MBR tanks to control membrane biofouling with the use of QQ beads (Fig. 1c). Most importantly, the use of this external chamber prevented the trapping of the fluidized QQ beads in the hollow-fiber membrane module15,16. According to Kim et al.16, the exo-type QQ strain Pseudomonas sp. 1A1 was used to fabricate the QQ beads in the present study because exo-type strains, which secrete QQ enzymes extracellularly, have been shown to superior in terms of the effect of the external chamber compared to endo-type strains (e.g., Rhodococcus sp. BH4), which preserve QQ enzymes intracellularly. Furthermore, we hypothesized that pre-treating the wastewater with the 1A1-containing QQ beads before feeding it into the GD-MBR would maximize the production of QQ enzymes, which would then suppress QS at the membrane and effectively mitigate biofouling. We assessed the proposed QQ GD-MBR during continuous operation for more than 100 days in terms of the flux level, the extracellular polymeric substance (EPS) concentration, QS signals, the microbial community structure, and the quality of the treated water.

Fig. 1: GD-MBR configuration.
figure 1

a Cross-sectional images of beads generated via fluorescence microscopy. b Comparative analysis of the QQ activity of beads based on the C8-HSL concentration. c Schematic diagram of two GD-MBRs with external chambers. d Transmembrane pressure (TMP) profiles of the suction membrane (SM) for GD-MBR maintenance in the non-QQ GD-MBR (GD-MBR 1) and the QQ GD-MBR (GD-MBR 2). e GDM flux of the non-QQ GD-MBR (GD-MBR 1) and the QQ GD-MBR (GD-MBR 2). The error bars represent the standard deviation (n = 3).

Full size image

Results

AHL QQ activity of the 1A1-containing beads

The dimensions of the soft hydrogel QQ beads ranged from 4.5 to 5.0 mm, with a spherical morphology. Cross-sectional images of the QQ beads taken with a fluorescence microscope confirmed the successful immobilization of 1A1 (Fig. 1a). Figure 1b presents the results for the C8-HSL degradation activity of the QQ beads in comparison to the control beads without 1A1. After 30 min, the concentration of C8-HSL in the presence of the QQ beads decreased to nearly 0, whereas the concentration with the control beads remained almost unchanged. The rate constant (k, h−1) after 30 min was also much higher for the QQ beads (k = 6.7 h−1) than for the control beads (k = 0 h−1).

Impact of QQ on the GD-MBR: comparative analysis of the TMP rise-up and the flux decay rate

Initially, for the SM in phase 1 without any beads, the time for transmembrane pressure (TMP) to reach 23 kPa was identical for both reactors, indicating readiness for comparative experiments (Fig. 1d). In phase 2, with cell-free and QQ beads applied respectively, the time for TMP to reach 30 kPa in GD-MBR 2 (with QQ beads) was delayed by up to 20% compared to GD-MBR 1 (with cell-free beads) over the 60-day operation period.

For the GDM, the two GD-MBRs were operated over three phases for a combined period of >120 days. In phase 1, the two GD-MBRs were operated for 15 days with an MLSS concentration of around 4000–7000 mg/L to measure the flux of the GDM (Fig. 1e). Both GD-MBRs maintained a stable flux of 6 L/m²/h, which was higher than the stable flux of 2–4 L/m²/h typically observed for GDM processes with low SS levels, such as those involving diluted wastewater without activated sludge1,2. Contrary to expectations, despite the relatively high TOC (135–270 mg/L) of wastewater and high MLSS (4000–7000 mg/L) of activated sludge, a stable flux of 6 L/m²/h was maintained in the GD-MBRs. This effect is attributed to the highly concentrated activated sludge, which rapidly decomposes organic matter in the raw wastewater, thereby maintaining a low TOC concentration (4–12 mg/L) in the mixed liquor of GD-MBR. This higher stable flux suggested that the use of GDMs in activated sludge bioreactors for wastewater treatment had the potential to significantly reduce energy consumption. Subsequently, in phase 2, the variation in the flux of the GDMs in the two reactors was compared to determine the effect of the QQ beads. It took approximately 30 d for a stable flux of 6 L/m²/h to be achieved in the non-QQ GD-MBR, while the QQ GD-MBR required 60 days to reach the same steady flux. Thus, during the 65 days under phase 2, the QQ GD-MBR treated approximately 130 L of water, while the non-QQ GD-MBR treated only about 105 L, representing a 24% improvement in water production efficiency when using the QQ beads (Fig. 2a). In phase 3, operated for approximately 35 d, the same QQ effects observed in phase 2 were reproduced. The QQ GD-MBR treated approximately 73 L of water, while the non-QQ GD-MBR treated only 58 L, indicating a 26% improvement in water production efficiency with QQ beads. To confirm the sustained effectiveness of the QQ activity of strain 1A1 in the QQ GD-MBR, the EPS concentration was measured. In particular, the polysaccharide and protein levels of the bound EPS in both the mixed liquor and the biofilm that formed on the membrane surface were analyzed at the end of phase 2 (i.e., after 85 days of operation). Though the EPS levels in the mixed liquor did not exhibit a substantial difference between the two GD-MBRs (Fig. 2b), a significant difference was observed for the EPSs in the biofilms (Fig. 2c). In the non-QQ GD-MBR, the polysaccharide and protein concentrations were 53.6 mg EPS/g TAB and 6.0 mg EPS/g TAB, respectively. In contrast, the QQ GD-MBR had polysaccharide and protein concentrations of 38.5 mg EPS/g TAB and 4.8 mg EPS/g TAB, respectively, which were approximately 30% and 20% lower. This suggests that the QQ enzymes contributed to biofouling control by reducing the quantity of EPS in the biofilm.

Fig. 2: Impact of QQ on the GD-MBRs.
figure 2

a Water production (L) of the non-QQ GD-MBR (GD-MBR 1) and the QQ GD-MBR (GD-MBR 2). EPS concentration in (b) MLSS and (c) the biofilm (i.e., total attached biomass; TAB) of the GD-MBRs during phase 2. d Concentration of C8-HSL equivalent in the permeate of the GDM during phase 3. The error bars represent the standard deviation (n = 3).

Full size image

Impact of QQ on the AHL levels of the GD-MBRs

To confirm that 1A1 can effectively control AHL levels in the GD-MBR, the AHL concentration was measured in the permeate from each GD-MBR on days 106 and 119 during phase 3. Figure 2d shows that the AHL concentration was significantly lower in the QQ GD-MBR than in the non-QQ GD-MBR (0.34 ± 0.03 pM and 0.02 ± 0.01 pM on day 106 and 0.28 ± 0.02 pM and 0.08 ± 0.01 pM on day 119, respectively). These findings indicate that the QQ strain 1A1 can successfully inhibit bacterial QS signaling within the reactor. The presence of 1A1 led to a lower AHL concentration, which reduced the EPS levels. This reduction suppressed biofilm formation on the membrane surface, thus mitigating biofouling.

Impact of QQ on the quality of the treated water from the GD-MBRs

To assess the impact of QQ on the quality of the treated water resulting from GD-MBR operation, we monitored the TOC concentration in the permeate (Fig. 3a). There was a 96% reduction in the TOC levels in the synthetic wastewater, decreasing from 270 mg/L to approximately 10 mg/L in both reactors, with no significant difference between the two during phase 2–3. The MLSS in the activated sludge was maintained at 4,000 mg/L during phase 1 and increased to 7000 mg/L in phase 2. This increase was attributed to the rise in TOC of the synthetic wastewater to 270 mg/L (from 135 mg/L in phase 1) to provide sufficient carbon source for 1A1. In phase 3, the TOC was lowered back to 135 mg/L, matching phase 1, resulting in the MLSS of both GD-MBRs converging to 4000 mg/L, indicating that QQ did not have a significant impact (Fig. 3b). Furthermore, apart from an increase in the floc size for the same reason, no notable differences between the two GD-MBRs were observed during phases 1 and 2 (Fig. 3c). Overall, while the administration of QQ did not affect the GD-MBRs or the purity of the treated water, it led to the production of a greater volume of treated water.

Fig. 3: Adverse effects of QQ on the GD-MBRs.
figure 3

a TOC concentration (feed and permeate), (b) MLSS concentration, and (c) floc size of the two GD-MBRs. The error bars represent the standard deviation (n = 2 or 3).

Full size image

Microbial community profiling: PCoA, PICRUSt, and co-occurrence network

Past research has reported conflicting findings on the influence of QQ on the composition and dynamics of microbial communities. Some studies have suggested that QQ alters the structure of microbial communities17, while others attribute these changes to physical factors such as the passage of time and the characteristics of the influent water18. Thus, this study investigated the relationship between the QQ-induced reduction in biofilm formation and changes in the microbial community of the GD-MBRs using 16S rDNA sequencing. PCoA was used to visualize the temporal dynamics in the species composition in the two reactors (Fig. 4a). The first and second axes of the PCoA explained 39.7% and 15.8% of the variation, respectively. The QQ GD-MBR exhibited a more pronounced divergence compared to the non-QQ GD-MBR based on the increasing distance between the points representing the two reactors over time. This indicates that the microbial composition in the QQ GD-MBR underwent a more rapid transformation due to the presence of the QQ beads. PICRUSt analysis was conducted to investigate changes in the functional gene composition of the two GD-MBRs, with a particular focus on genes associated with biofilm formation (Fig. 4b). Previous studies have reported that biofilm formation is inhibited by antitoxin HigA, which promotes the formation of antitoxin HigB and the HigB/HigA TA system19. In the present study, antitoxin HigA-1 was found to be significantly more abundant in the QQ GD-MBR than in the non-QQ GD-MBR (p < 0.05). The Fis family transcriptional regulator, which modulates the Fis protein involved in QS in Vibrio cholerae20, was also significantly more abundant in the non-QQ GD-MBR (p < 0.05). Similarly, the LuxR family transcriptional regulator, which governs the AHL QS-related protein LuxR21,22, was also found to be more prevalent in the non-QQ GD-MBR (p < 0.05). These findings indicate that QQ had a negative impact on genes associated with biofilm formation. Co-occurrence network analysis was conducted to elucidate the interactions between the microorganisms within the mixed liquor and biofilm in the reactors based on their relative abundance (Fig. 4c, d). These interactions were represented by positive edges (blue lines) when the relative abundance of two taxa at the class level both increased simultaneously over time and negative edges (red lines) when their relative abundance at the class level decreased together. The analysis revealed that the QQ GD-MBR had a higher number of edges (positive: 14; negative: 18) compared to the non-QQ GD-MBR (positive: 11, negative: 16), indicating a greater level of interaction within the microbial community23. Notably, Blastocatellia, known for its role in biofilm development24, and Planctomycetia, a subgroup of Planctomycetes widely found in the biofilm of kelp25, exhibited a stronger adverse interaction in the QQ GD-MBR than in the non-QQ GD-MBR. Additionally, Sphingobacteriia, which contributes to biofilm resistance26, also exhibited a higher frequency of adverse interactions in the QQ GD-MBR. These findings suggest that QQ enhances microbial interactions and significantly influences the composition of the biofilm-forming microbial community. These results indicate that QQ not only inhibits biofilm formation but also influences the microbial community structure through functional gene regulation and inter-species interactions. QQ’s regulation of microbial ecology has potential applications in developing more stable and efficient GD-MBRs by guiding community dynamics in desired directions. Future research on gene expression and microbial activity could further elucidate how QQ influences microbial populations and its practical implementation in wastewater treatment processes.

Fig. 4: Microbial community profiling.
figure 4

a Principal coordinate analysis (PCoA) based on a Bray-Curtis matrix at the species level. b Analysis of functional genes related to biofilm formation and QS using PICRUSt based on the KEGG database. Co-occurrence network for (c) the non-QQ GD-MBR (GD-MBR 1) and (d) the QQ GD-MBR (GD-MBR 2). The size of each yellow node indicates the number of connections it has, while the nodes themselves reflect the class level. Positive and negative correlations are shown by blue and red lines, respectively (p < 0.05). Each correlation has a strong Spearman’s coefficient of > 0.75 or < –0.75.

Full size image

Discussion

This study demonstrated the potential of a GD-MBR system that utilizes gravity-driven membranes for low-energy wastewater treatment. Despite operating under relatively high concentrations (135–270 mg TOC/L, 400–800 mg COD/L) and high MLSS levels (4000–7000 mg/L), the GD-MBR maintained a stable flux of 6 L/m²/h—higher than the 2–4 L/m²/h typically reported for other GDMs treating low-TOC and low-SS wastewater. These findings highlight the potential for significant energy savings by integrating GDMs and QQ with activated sludge systems. To enhance system performance and mitigate membrane biofouling, QQ beads containing the exo-type of the QQ strain (Pseudomonas sp. 1A1) were introduced:

  1. 1.

    The QQ GD-MBR system doubled the time required to reach a stable flux and improved water production efficiency by up to 26% compared to the non-QQ system, without compromising treated water quality.

  2. 2.

    Mechanistically, QQ effectively reduced membrane biofouling. While mixed liquor EPS concentrations remained similar, the biofilm EPS on the membrane surface showed significantly lower polysaccharide (30%) and protein (20%) levels in the QQ GD-MBR. This reduction in biofilm EPS was associated with significantly lower AHL (C8-HSL) concentrations in the permeate, indicating successful quorum sensing (QS) inhibition.

  3. 3.

    Microbial community analysis revealed that QQ induced a more complex microbial network, and gene expression analysis suggested downregulation of biofilm- and QS-related genes (e.g., HigA-1, Fis, LuxR family).

Overall, coupling GDMs with activated sludge enables stable, high-flux operation even under high-load conditions, and integrating QQ presents an effective strategy for biofouling control. This approach offers a novel, energy-efficient solution and broadens the applicability of GDM processes to high-strength wastewater such as that from industrial or decentralized sources.

Methods

Setup of the GD-MBR and operating variables

Two GD-MBRs were configured in parallel for comparative experiments (Fig. 1c). Each reactor had an operating volume of 4 L and was equipped with a GDM powered solely by gravity. The operating pressure was maintained at 25 mbar by positioning the valve that allowed treated water to pass through the GDM 25 cm below the water surface. The GDM module used hollow-fiber membranes made of polyvinylidene fluoride (PVDF) with a pore size of 0.1 µm (KOLON, Republic of Korea). These membranes were installed vertically, covering an area of 0.009 m². The flux of the GDM was determined by converting the weight of treated water collected over a 5 min period into the volume of treated water produced per hour (L/h). The membrane flux (L/m²/h) was then calculated by dividing this value by the membrane area (m²), which was used to assess the degree of membrane fouling. In a typical GDM process, flux differences under constant pressure alter the feed inflow, leading to changes in the hydraulic retention time (HRT) and subsequently, the activated sludge community, complicating comparative experiments. To address this, in addition to the two GDMs, each suction membrane (SM and 0.018 m²) with constant flux was installed to maintain consistent feed inflow into the reactor. The primary function of SM is to ensure a constant overall hydraulic retention time (HRT, 16 h) in the reactor and to assess the QQ effect on membrane biofouling in conjunction with GDM. Thus, the permeate from the GDM was recirculated back into the reactor to maintain a consistent HRT in both reactors. The SM for system maintenance operated at a constant flux of 17 L/m²/h, with transmembrane pressure (TMP) collected in real-time to evaluate membrane fouling.

To test the effect of QQ on membrane biofouling, an external chamber with an operating volume of 2 L was installed between the feed and the membrane tanks. In the external chamber for one of the GD-MBRs (hereafter referred to as the QQ GD-MBR), QQ beads containing Pseudomonas sp. 1A1 were added at a concentration of 3.0% v/v to the operating volume of the GD-MBR16. In the other GD-MBR (referred to as the non-QQ GD-MBR), cell-free (empty) beads were added at the same concentration. The HRT in the external chamber was set at 8 h to ensure sufficient time for the QQ beads to secrete QQ enzymes. The aerobic activated sludge utilized in the GD-MBRs was collected from a wastewater treatment plant run by Gwangju Environmental Corporation (Republic of Korea) and maintained under a supply of synthetic wastewater during the stabilization phase. The operating conditions are detailed in Table 2, and the composition of the synthetic wastewater is shown in Table S1.

Table 2 Summary of operating conditions for the GD-MBR systems utilized in this study
Full size table

Preparation and internal visualization of 1A1-containing QQ beads

The QQ beads used in this study contained Pseudomonas sp. 1A1, which secretes N-acyl-homoserine lactone (AHL)-degrading enzymes extracellularly16,27. The 1A1 strain was cultured in Luria–Bertani (LB) medium at 30 °C for 18 h. The cells were then harvested using centrifugation at 9000 g and 4 °C for 10 min, and the resulting pellet was resuspended in sterile distilled water. This suspension was then mixed with a polymer solution containing 10.0% w/w polyvinyl alcohol (Sigma-Aldrich, USA) and 1.0% w/w sodium alginate (Junsei, Japan) to achieve a final concentration of 10.0 mg of 1A1 per 1.0 g of the polymer solution16,27,28. The mixture was dropped into a primary crosslinking solution comprising 7.0% w/w boric acid (Duksan, Republic of Korea) and 4.0% w/w calcium chloride (Daejung, Republic of Korea) to form spherical beads, which were then agitated for 1 h. To replace the boron with sulfur, the beads were treated in a secondary crosslinking solution of 0.5 M sodium sulfate (Daejung, Republic of Korea) for 18 h. The beads were then washed three times with distilled water for 60 min each and stored before use in the experiments. The control beads lacking cells were prepared using the same process, except that the polymer solution was mixed with sterile distilled water rather than a cell suspension at the same ratio. The immobilization of 1A1 was confirmed by staining the beads with LIVE/DEAD BacLight Bacterial Viability reagent (L7012, Thermo Fisher Scientific, USA) and observing them under a fluorescence microscope (Eclipse Ts2, Nikon, Japan) at 100 × magnification. This reagent contains SYTO 9, which selectively stains both intact and damaged cells, producing green fluorescence, and propidium iodide (PI), which stains damaged cells, resulting in red fluorescence.

Evaluation of AHL QQ activity

To quantify the AHL-degradation activity of the QQ beads, 200 nM of an AHL solution was prepared using N-octanoyl-L-homoserine lactone (C8-HSL, Sigma-Aldrich, USA), which is the dominant AHL in activated sludge29, in a 50 mM Tris-HCl buffer solution (pH 7.00, Sigma-Aldrich, USA). Fifty QQ beads and control beads were added respectively to 20 mL of the prepared C8-HSL solution. Samples were collected at 0, 1, 30, 90, and 180 min. To prevent further degradation of C8-HSL by the QQ enzymes secreted by the QQ beads, the samples were heat-treated in a 90 °C water bath for 3 min. The samples were then diluted 20-fold with Tris-HCl buffer (50 mM, pH 7.00) for quantitative C8-HSL bioassays. In these bioassays, 5 µL of each sample and C8-HSL standards (0, 1.25, 2.5, 5.0, and 10.0 nM C8-HSL) were added respectively to a 96-well microplate and mixed with 95 µL of Agrobacterium tumefaciens A136 (OD600 of 0.1), a reporter strain for AHL, followed by incubation at 30 °C for 90 min29. After incubation, 30 µL of Beta-Glo® reagent (Promega, USA) was added and allowed to react at room temperature for 40 min16,28. Luminescence was then measured using a microplate reader (Synergy HTX, Biotech, USA).

Microbial community analysis

Mixed liquor and biofilm samples from the two GD-MBRs were collected after 60 days of operation and 16S rDNA sequencing was carried out at CJ Bioscience, Inc. (Republic of Korea) with an Illumina MiSeq Sequencing system (Illumina, USA) according to the manufacturer’s instructions. DNA was extracted from the collected samples, and its concentration was determined using Quat-iT PicoGreen dsDNA assay reagent (Invitrogen, USA). Polymerase chain reaction (PCR) was then performed using the 341 F and 805 R fusion primers, targeting the V3-V4 region of the bacterial 16S rDNA gene. The PCR protocol included initial denaturation at 95 °C for 3 min, followed by 25 cycles of denaturation (95 °C for 30 s), annealing (55 °C for 30 s), and extension (72 °C for 30 s), with a final extension at 72 °C for 5 min. Gel electrophoresis was used to assess the size of the PCR products, and AMpure bead XP (Beckman Coulter, USA) was used to remove non-target products. The quality of the library was evaluated using an Agilent DNA 7500 kit and Agilent 2100 Bioanalyzer (Agilent Technology, USA), and the purified products were quantified using a Quat-iT PicoGreen dsDNA assay kit. The prepared library was sequenced using the Illumina iSeq 100 system, and the data were classified using the EzBioCloud database. This classified data formed the basis for secondary analysis. Data obtained from LDA effect size (LEfSe) analysis were classified according to protein function using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt)30. Principal coordinate analysis (PCoA) was performed using the Bray–Curtis method to visualize differences between samples31. Co-occurrence network analysis was also conducted using the Gephi 0.10.1 program to identify interactions between microorganisms.

Monitoring AHL QS signals in the permeate of the GD-MBRs

AHL has a molecular weight ranging from 150 to 300 Da, meaning that it could readily pass through the pores (0.1 µm) of the membrane used in the present study and enter the permeate tank. Thus, the AHL concentration in the permeate was assumed to reflect its concentration in the mixed liquor of the GD-MBR32. However, due to the low concentration of AHL in the GD-MBR, a concentration process was required to allow for a meaningful comparison. To isolate the AHL, 2 L of permeate from each GD-MBR was collected and subjected to solid-phase extraction (SPE) with a Supelclean™ LC-18 SPE tube (Supelco, USA)28,32. The extracted AHL was dissolved in 10 mL of acetonitrile (ACN), after which the ACN was completely evaporated using a SpeedVac (Labogene, Lynge, Denmark). The residue was then reconstituted in 250 µL of 50 mM Tris-HCl buffer, resulting in an approximately 8000-fold increase in the concentration. The concentrated AHL was evaluated using luminescence-based bioassays, as described in the method section ‘Evaluation of AHL QQ activity’, and dual-layer agar bioassays. In the dual-layered agar bioassays, a supporting layer composed only of LB agar was combined with a functional second layer produced by combining A136 with LB agar (containing tetracycline, spectinomycin, and X-gal) at a 1:9 ratio. Two circular wells were created on the plate, and each well was filled with the concentrated AHL. Following incubation at 30 °C, the development of a blue pigment was monitored, indicating AHL activity33.

EPS analysis

Heat treatment was used to extract EPS from the mixed liquor of the GD-MBRs and the biofilm on the surface of the GDMs16,28,34. The EPS concentration was compared between the two MBRs based on the quantity of polysaccharides and proteins, the primary components of EPS. The protein content was quantified using Bradford protein assays35, while the polysaccharide content was quantified using the phenol–sulfuric acid method36. The EPS concentrations in the mixed liquor and biofilm samples from the two GD-MBRs were normalized by dividing them by the MLSS and total attached biomass (TAB) on the membrane, respectively. TAB was calculated using the same method as MLSS, by mixing the biofilm detached from the membrane with a constant volume of distilled water after the phase ended and the GDM was removed. The concentration of TOC was measured using the non-purgeable organic carbon (NPOC) method with a TOC analyzer (TOC-L CPH, Shimadzu, Japan). The floc size was measured using a particle size analyzer (Partica Mini LA-350, Horiba, Japan). Each analytical experiment was carried out three times (n = 3), and two-tailed Student’s t-tests were used to test for significant differences between treatment groups.