Aerobic biodegradation of quinoline under denitrifying conditions in membrane-aerated biofilm reactor

.


Introduction
Quinoline, a representative nitrogen-heterocyclic compound, is used as the raw material and solvent in chemical industry and manufacture of dyes, pesticides and pharmaceuticals (Lam et al., 2012;Rafiee et al., 2004;Zou et al., 2015). Quinoline is abundant in coal tar, mineral oil, petroleum and industry wastewater (e.g., coking, coal gasification and pharmaceutical wastewater) (Jiang et al., 2017;Wu et al., 2020). For instance, quinoline has been reported to account for 27.4% of organic pollutants in coal gasification wastewater (He and Wang, 2015). Quinoline can accumulate in water environments due to it high solubility and low biodegradation, and can cause adverse effects on wildlife and human health because of its toxicity, carcinogenicity and mutagenicity (Felczak et al., 2014). Considering its high toxicity and serious adverse environmental impacts, removal of quinoline from residual waters is of great importance.
Bacillus (Tuo et al., 2012), Comamonas (Cui et al., 2004), etc. Aerobic degradation of quinoline generally begins with the hydroxylation reaction, resulting in the formation of 2-hydroxyquinoline, and follows by the cleavage of pyridine and benzene rings (Cui et al., 2004;Fetzner, 1998;Qiao and Wang, 2010;Shukla, 1989;Zhu et al., 2008). Since denitrifying pathways are often present in these identified quinoline-degrading genera (e.g., Pseudomonas), it is posited that they might be able to utilize quinoline as the sole energy and carbon source, and NO 3 − as electron acceptor and nitrogen source under oxic condition (Bai et al., 2010;Zhu et al., 2008). In a pure culture study of Pseudomonas, both aerobic quinoline-degrading gene of 1H-2-oxoquinoline 8-monooxygenase (oxoO) and denitrifying genes of nitrite reductase (nirS) and nitrous oxide reductase (nosZ) were identified, and simultaneous quinoline degradation and nitrate (NO 3 − ) reduction were obtained (Bai et al., 2010). Hence, these aerobic quinoline-degrading microbes could be enriched and applied to treat quinoline-containing wastewater for simultaneous removals of carbon and nitrogen in a single aerated bioreactor. However, previous studies mainly focused on characteristics and pathways of aerobic quinoline biodegradation in pure cultures. The enrichment and application of aerobic denitrification dependent quinoline degradation in bioreactors have to our knowledge not been investigated. The dominant aerobic quinoline-degrading bacteria, their enzymatic activities under varying operational conditions, and the interaction with other microorganisms in complex community remain unknown.
Membrane aerated biofilm reactor (MABR) is being implemented as an energy and resource-efficient process compared to traditional aerobic wastewater treatment processes (e.g., activated sludge reactor). In MABR, membrane serves as both immobilization carrier and microporous aerator for biofilms. The gaseous substrate (e.g., oxygen (O 2 )) is supplied through the supporting membrane and diffuses into the biofilm from the base, while the dissolved substrate (e.g., quinoline and NO 3 − ) is provided from the bulk liquid and diffuses from the top part of the biofilm (Nerenberg, 2016). Counter-diffusion of MABR results in unique process features, including flexible gas supply control, unique microbial community structures, high sensitivity to biomass accumulation, and low susceptibility to boundary layer resistance Nerenberg, 2016;Syron and Casey, 2008). Owing to the characteristics of bubble-free aeration and highly efficient O 2 transfer, MABR offers improved energy efficiency and effective removals of both carbon and nitrogen in a single step (Aybar et al., 2015;He et al., 2021a). MABR technology has been applied for municipal/high strength industrial wastewater treatment, removals of carbon, nitrogen and refractory pollutants, and retrofitting of existing activated sludge plants (Côté et al., 2015;Lu et al., 2021). Thus, MABR is a promising aerobic technology that can be embedded in current wastewater treatment processes to accomplish simultaneous and efficient removals of quinoline and NO 3 − .
The primary goal of this study was to investigate the feasibility of applying aerobic denitrification in an MABR for treating quinolinecontaining wastewater. The performance and stability of MABR were evaluated under different operational conditions, including quinoline and NO 3 − loadings, HRTs and co-occurrence of other toxic aromatic hydrocarbons, etc. Based on 16 S rRNA and metagenomic analyses, the dominant aerobic quinoline-degrading bacteria were identified, tentative key functional genes for aerobic quinoline degradation and denitrification were quantified, and an aerobic biodegradation pathway of quinoline was proposed. To the best of our knowledge, this is the first study with the purpose of engineering implementations to apply aerobic denitrification dependent quinoline biodegradation process for simultaneous, efficient, and robust removal of recalcitrant organic carbon and nitrogen from quinoline-containing wastewater.

Reactor configuration and operation
A lab-scale MABR with working volume of 9 L was used (Supplementary Information Fig. S1). A membrane module, consisting of 243 polyvinylidene fluoride hollow fibers (Table S1), was submerged in the reactor tank. Air was supplied into fibers by an air compressor and consumed by biofilms adhered on the membrane. The exhaust gas in the membrane lumen was regulated by a drain valve. Synthetic wastewater was fed and completely mixed by a recirculation pump in reactor tank, and effluent was discharged through overflow.
The seeding sludge, originated from the oxidation ditch of activated sludge stream at Wulongkou wastewater treatment plant (WWTPs) (Zhengzhou, China), was inoculated into the membrane surface. A thin biofilm was formed on the membranes after one week. The operational period (240 days) was divided into nine phases (Table 1). For the fast start-up at P1-P2, glucose (153− 308 mg/L) and ammonium chloride (NH 4 Cl) (110 mg N/L) together with low concentration of sodium nitrate (NaNO 3 ) (3.1− 5.8 mg N/L) were supplied as the organic carbon and inorganic nitrogen source, respectively. During the remaining operational periods, the carbon and nitrogen source were then changed to quinoline (50− 259 mg/L) (and phenol at P8-P9) and NaNO 3 (4.7− 62 mg N/L), respectively. The applied influent concentrations of quinoline and NO 3 − were within the range of concentrations previously detected in real coking and coal gasification wastewater (28− 200 mg/L and 20− 149 mg N/L, respectively) (Bai et al., 2011a;Bai et al., 2011b;Cui et al., 2020;Zhu et al., 2017). To investigate the removal capacity of quinoline and NO 3 − , quinoline and NO 3 − loading rates were continuously increased at P4-P9. Pyridine (150 mg/L) was spiked at P6 and phenol (32− 105 mg/L) was added in influent at P8-P9 (Shi et al., 2019;Wen et al., 2013;Zhang et al., 2019), in order to examine the combined effect of other aromatic pollutants on quinoline biodegradation and denitrification.

Extracellular polymeric substances
At the end of each operational phase, biofilms from five random sampling positions were detached using a sterilized razor, and homogenized for the extraction of extracellular polymeric substances (EPS). While 2 mL of biomass was used for the determination of volatile suspended solids (VSS), the remaining biomass (2 mL) was washed with phosphate buffer saline (PBS) and then centrifuged at 6000 rpm at 4 • C for 5 min. After discarding supernatant, biomass was suspended again in PBS and incubated at 80 • C in a thermostat water bath for 40 min. After cooling at room temperature and centrifugation at 8000 rpm at 4 • C for 15 min, the supernatant was filtered through 0.45 μm pore size filter to obtain loosely bound EPS (LB-EPS). The remained biomass was further treated with 8 mL of a mixture containing 1.5 M NaOH and formaldehyde solution, followed by shake cultivation at 100 rpm at 30 • C for 3 h and successive centrifugation at 12000 rpm at 4 • C for 30 min. The supernatant was filtered using 0.45 μm pore size filter to obtain tightly bound EPS (TB-EPS).
Anthrone-sulphuric acid colorimetry was used to analyze polysaccharides (PS) content in EPS (with glucose as the standard) (Frølund et al., 1996). The protein content (PN) was analyzed with the Coomassie Brilliant Blue method (with bovine serum albumin as the standard) (Pierce and Suelter, 1977). Deoxyribonucleic acid (DNA) was measured via diphenylamine colorimetric method (with calf thymus deoxyribonucleic acid as the standard) (Liu and Fang, 2002). Humic substance was determined using the UV254 method (with humic acids as the standard) (Wang & Fujii, 2011).

DNA extraction and qPCR
Biofilm samples were collected at the end of each phase and stored at − 80 • C until DNA extraction. DNA was extracted using E. Z.N.A.® Soil DNA Kit (Omega Bio-tek, USA) according to the manufacturer's instructions. The quantity and quality of the extracted DNA were measured and checked by its 260/280 ratio via a NanoDrop 2000 UV-vis spectrophotometer (Thermo Scientific, Wilmington, USA).
Real-time PCR technology was employed to quantify the copy numbers of relevant degradation genes: oxoO, anaerobic quinoline transformation (benzoyl-CoA reductase (bcrA)), and denitrifying genes (dissimilatory nitrate reductase (napA), nirS, nitrite reductase (nirK), nitric oxide reductase (norB) and nosZ) as well as the 16 S rRNA gene. Primers and conditions used in various PCR assays were listed in Table S2. All samples, including control reactions without template DNAs, were measured in triplicates.

16 S rRNA amplicon sequencing and metagenomic sequencing
Amplification of V3-V4 hypervariable fragments (338 F and 806 R) of 16 S rRNA gene was performed on an Illumina MiSeq instrument (Illumina, USA) at Majorbio Corporation (Shanghai, China). Filtration and cluster of high-quality sequences and taxonomy annotation were performed as described in the previous study (Tian et al., 2020).
The homogenized total genomic DNA of biofilms was sequenced on an Illumina HiSeq 4000 platform (Illumina Inc., CA, USA) at Majorbio Corporation (Shanghai, China) by using HiSeq 3000/4000 PE Cluster Kit and HiSeq 3000/4000 SBS Kit, according to the manufacturer's instructions (www.illumina.com). Detailed descriptions of genome assembly, assembled sequences prediction, taxonomy assignment and functional gene annotation were listed in Supplementary materials Section S2. Functional annotation of open reading frames was conducted using evolutionary genealogy of genes: Non-supervised Orthologous Groups (eggNOG) (http://eggnog.embl.de/) and BLASTP search (Version 2.2.28+) against the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (http://www.genome.jp/kegg/), with an e _ value cutoff of 10 − 5 . Functional contributions of microbial taxa (at genus level) to the functional pathways were examined using customized R scripts. The relative contribution of taxon to a metabolic function was calculated by dividing the abundance of taxon participating in the function to the total abundance of all taxa involved in the function. Raw sequencing data of metagenomes of biofilm samples can be found in NCBI Sequence Read Archive with accession numbers of PRJNA836146 and PRJNA836157.

Analytical methods
Bulk samples were collected and filtered through 0.45 μm pore size filters for NH 4 + , NO 2 − and NO 3 − measurements. Concentrations of nitrogen species were determined spectrophotometrically using a multiparameter bench photometer (DR900, Hach Instruments Inc., USA). COD was analyzed using potassium dichromate method (APHA, 2005). Liquid samples were further filtered through cellulose ester membranes with a pore size of 0.22 μm before quinoline and phenol determination. Reverse phase high-performance liquid chromatography (RP-HPLC) (1260 Infinity II, Agilent, USA) equipped with C18 column (250 mm × 4.6 mm, 5 μm) and UV detector was applied to monitor quinoline and phenol concentrations. The mobile phases of RP-HPLC were consisted of water and methanol with the ratio of 50:50 or 40:60 and the flow rate of 0.8 or 1.0 mL/min. Transformation products were analyzed using an ultra-high performance liquid chromatography (UHPLC) coupled to a QE mass spectrometer (1290 Infinity, Agilent, USA). The details of sample preparation and LC-MS/MS analysis were described in Supplementary materials Section S3. DO and pH were measured via a multiparameter water quality analyzer (HQ30D, Hach Instruments Inc., USA). VSS concentrations were quantified following standard methods (APHA, 2005). Surface morphologies of biofilms were characterized by a scanning electron microscope (SEM, FEI Quanta 250, USA).

Table 1
Overview of reactor operational conditions. At P3, the air outlet was closed to enhance oxygen transfer efficiency for energy conservation, and exhaust gas was periodically vented.
c Inner recirculation was closed during P4 to evaluate the effect of recirculation pattern on reactor performance, and the feeding pattern was changed from batch to continuous inflow at P4-P9.
d Pyridine (150 mg/L) was spiked at Day 103-105 to investigate the combined effect of other aromatic pollutants on quinoline degradation.

Statistical analysis
The 16 S rRNA and metagenomic data analyses were carried out using the free online platform of Majorbio I-Sanger Cloud Platform (http: //www.i-sanger.com/), including principal coordinates analysis (PCoA), redundancy analysis (RDA), Spearman correlation analysis, functional contribution analysis and network analysis. All sequencing results were shown in relative abundance calculated by reads per kilobase per million reads (RPKM) (Tang, 2020). Correlations of functional genes and KEGG modules with operational conditions and removal performance, and correlation networks of the key genera related to quinoline biodegradation and denitrification were performed in R using the package "psych" and visualized using the package "corrplot" or "pheatmap". The statistical difference was confirmed by analysis of variance (ANOVA), and a p value < 0.05 represented significant.

Quinoline degradation performance
The MABR was continuously operated in a constant temperature room at 24 ± 2 • C for 240 days under different operational conditions (Table 1), and displayed stable quinoline and NO 3 − removal at the end of each stage (Fig. 1). With stepwise increases in loading from 2.4 to 6.9 g/ (m 2 ⋅d) at P2, the quinoline degradation efficiency increased from 72.3% at Day 29-91% at Day 54 ( Fig. 1a). Efficient quinoline degradation was accompanied with intensive O 2 consumption, with bulk DO concentrations decreased from 4.6 mg O 2 /L at Day 29 to 1.1 mg O 2 /L at Day 54 ( Fig. S2b). Although the adjustment of HRT from 11 h to 6 h at P3 resulted in a slight decrease in quinoline degradation efficiency to 87.3 ± 2.0%, the quinoline degradation rate doubled from 5.7 g/(m 2 ⋅d) at Day 53 to 10.9 g/(m 2 ⋅d) at Day 66 ( Fig. 1b). Quinoline degradation performance was strongly affected by inner liquid recirculation: degradation efficiency and rate decreased to 64.4 ± 8.7% and 8.0 ± 0.9 g/(m 2 ⋅d), respectively, after closing recirculation at P4; when switching on recirculation at P5, degradation performance rapidly recovered within 4 days and stabilized at 94.8 ± 0.8% and 11.9 ± 0.2 g/(m 2 ⋅d). Quinoline degradation rate decreased by two times (i.e., 5.9 g/(m 2 ⋅d) at Day 144) after the spike of pyridine at P6, and gradually recovered and maintained at 12.8 ± 0.2 g/(m 2 ⋅d) at the end of P7 (Fig. 1b). Despite of substantial decline in quinoline degradation at P8 and beginning of P9 due to the addition of phenol, quinoline degradation efficiency and rate gradually increased to 78.2% and 10.7 g/(m 2 ⋅d) at the end of P9, respectively. Meanwhile, efficient phenol degradation was obtained during P8-P9, with degradation efficiency and rate of 86.0 ± 11.4% and 3.1 ± 1.2 g/(m 2 ⋅d), respectively.

NO 3 − removal performance
High NO 3 − removal rates, up to 2.9 g N/(m 2 ⋅d), were obtained after increasing NO 3 − loadings from 0.4 ± 0.06 g N/(m 2 ⋅d) to 3.2 ± 0.2 g N/ (m 2 ⋅d) at P6 (Fig. 1b). Spike of pyridine at the beginning of P6, doubled quinoline concentrations in influent at P7, and addition of phenol in influent from P8 led to temporary decreases in NO 3 − removal rates, which all recovered to a similar level of 2.1 ± 0.6 g N/(m 2 ⋅d) at the end of each phase (Fig. 1b). After switching NO 3 − as the sole nitrogen source from P3, NH 4 + was detected at concentrations of 9.3 ± 2.7 mg N/L in effluent (Fig. S2a). NO 2 − was not detected in the effluent (below the analytical detection limits of 0.1 mg N/L, data not shown).

Microbial community structure and function
3.3.1. Microbial community composition dynamics 16 S rRNA analysis revealed abundance of Actinobacteria, Gammaproteobacteria, Anaerolineae, Alphaproteobacteria and Bacteroidia, accounting for 73.1 ± 8.1% of the whole community throughout the operational period (Fig. 3a). Some other classes were important at specific phases, such as Saccharimonadia, which appeared at P2 and P9 with relative abundances of 34.3% ± 3.6 and 25.9% ± 2.3%, respectively (Fig. 3a). At the genus level, gradual increasing quinoline loadings from P2 caused a clear increase in the relative abundance of Rhodococcus, a bacterial genus known containing aerobic quinoline degraders, from 0.43 ± 0.08% at P1 to 26.9 ± 3.7% at P9 (Fig. 3b). While the abundance of Rhodococcus was substantially decreased after pyridine spike at P6, Rhodococcus gradually recovered under increasing NO 3 − loadings from P7 and became the dominant genus at the end of operational period. Other potential aerobic quinoline-degrading genus Pseudomonas and Comamonas were also detected yet with low abundances of 1.70 ± 1.2% and 0.94 ± 0.9%, respectively. The addition of phenol from P8 also shifted the overall microbial community, where the growth of Raineyella, norank_f__Saccharimonadaceae, Paraclostridium and Hydrogenophaga was selectively stimulated. Particularly, the abundance of Raineyella increased by 25 times under the availability of phenol from P8. The bacteria potentially capable of denitrification identified in community included Rhodococcus, Raineyella, unclassified_p_Chloroflexi, Hyphomicrobium, Tessaracoccus, etc (Fig. 3b). Compared to Rhodococcus and Raineyella, the abundance of other non-quinoline/phenol degrading denitrifiers was 1-3 order of magnitudes lower. Unclassified_p_Chloroflexi was likely one of the dominant denitrifiers, despite of substantial declines in the abundance from 19 ± 8.5% at P1-P2 to 2.7 ± 1.7% at P3-P9 after quinoline addition. The potential denitrifiers Hyphomicrobium and Tessaracoccus were less abundant in microbial community (0.22 ± 0.038% and 0.057 ± 0.012% at P9, respectively). The PCoA based on OTU similarities confirmed significant changes in microbial community at P2, P3 and P8 due to quinoline addition, decreasing HRT and phenol spike, respectively, together with decreasing community diversity (Fig. 3c− d). While the increase in NO 3 − loadings and spike of pyridine at P6 did not induce significant differences in community compositions compared to P5 (Fig. 3c− d), the community became more diverse at P6, with 1.2 times higher Shannon index (Fig. S4).

Aerobic quinoline degrading and denitrifying functional genes
Decrease in gene copy number of total bacteria was observed along with the increase of quinoline loading and the spike of pyridine (Fig. S5). While the gene responsible for anaerobic quinoline degradation, i.e., bcrA, was not detected, the abundance of aerobic quinoline degradation gene oxoO increased by 1-3 orders of magnitude under rising quinoline (at P3) and phenol loadings (at P8 and P9) (p < 0.05) (Fig. 4a). Besides, increasing quinoline loadings from P1-P5 enhanced abundances of denitrifying genes of napA, nirS and nirK, but continuously decreased that of norB and nosZ (Fig. 4a). A significant positive correlation between oxoO and genes of nirS and nirK, was observed (p < 0.05) (Fig. 4b). The spike of pyridine at P6 strongly inhibited both oxoO and denitrifying functional genes (except napA) (Fig. 4a). There were 3.6-34 times higher abundances in nirS, norB and nosZ after adding phenol from P8 (Fig. 4a).

Major aromatic degradation and denitrification metabolic modules
Based on metagenomic analysis, major metabolic modules associated with aromatic degradation and denitrification as well as their correlations with operational conditions and removal performance were identified (Fig. 5). Among these metabolic modules, M00878 (phenylacetate degradation) and M00545 (trans-cinnamate degradation) were two of the highest abundant modules; yet their abundances did not significantly correlate to operational conditions and performance, which might involve in secondary degradation processes. M00569 (catechol meta-cleavage) was found to strongly positively correlated with loading and degradation rates of quinoline (p < 0.05) and NO 3 − , with abundances increased by up to 3− 4 times at the end operational period (Fig. 5a). Besides, the abundance of M00568 (catechol ortho-cleavage) were 3-7 times lower than that of M00569 (catechol meta-cleavage), indicating that the quinoline was mainly degraded through catechol meta-cleavage instead of catechol ortho-cleavage (Fig. 5a). M00529 was assembled to denitrification process, which showed positive correlation with loadings and removals of quinoline and NO 3 − (Fig. 5b).

Major quinoline degrading and denitrifying microbes
Considering the importance of M00569 and M00529 in aromatic degradation and denitrification, the relative contribution of different microorganisms to these two modules was further interpreted (Fig. 6a) The results suggested that aerobic aromatic degradation and denitrification could be performed by one microbe (e.g., Rhodococcus and Raineyella) or over different microbes in the microbial community (e.g., unclassified_p_Chloroflexi for denitrification and Sphingomonas for aromatic degradation). Overall, Rhodococcus, Raineyella and Unclassified_p__Chloroflexi were major contributors to napA and nirK, while Unclassified_p__Chloroflexi and unclassified_d__Bacteria contributed mostly for nirS and norZ (Fig. 6b). Unclassified_p__Chloroflexi was one of the dominant genera contributing to aforementioned four denitrifying genes. Compared to other functional denitrifying genes, norB was distributed in more diverse microorganisms, such as unclassified_f__Chitinophagaceae, unclassi-fied_p__Planctomycetes and unclassified_p__Acidobacteria (Fig. 6b). Again, varying operational conditions caused obvious changes in contribution compositions of denitrifying genes. For instance, the dominant contribution of Micropruina to napA (18.5− 24.9%) and nirK (18− 25.7%) at P1-P2 were overtaken by Rhodococcus (10.1% for napA and 13.3% for nirK) at P7 after gradual increases of quinoline loadings, while the addition of phenol from P8 significant stimulated these two genes in Raineyella (from 0.5% to 1.4% at P7 to 11.3% and 35.0% at P9 for napA and nirK, respectively). Although the existence of oxoO was confirmed by qPCR analysis, functional annotation of oxoO was not available in KEGG and eggNOG databases, likely due to the limited genetic studies on biodegradation of quinoline and its derivatives. Hence, potential metabolic pathways of quinoline degradation as well as the involved microbes were interpreted based on aromatic degradation modules (e.g., M00569), which have been widely investigated with relatively complete genetic information.

Correlations of quinoline degrading and denitrifying microbes with operational conditions and removal performance
Correlation networks of the key genera related to quinoline biodegradation and denitrification were shown in Fig. 6c. Genera with significantly positive correlation (p < 0.05) indicated that they could collaborate or have similar niches in biofilm, which were clustered into one group. Thus, five groups were identified, including Group I (Tessaracoccus, Candidatus_Promineofilum, unclassified_c__Anaerolineae, unclassi-fied_c__Actinobacteria, unclassified_p__Chloroflexi, Propionibacterium and Micropruina), Group II (Rhodococcus, unclassified_o__Flavobacteriales, unclassified_f__Comamonadaceae, unclassified_o__Burkholderiales, unclassi-fied_d__Bacteria and Raineyella), Group III (Pseudomonas, Hyphomicrobium, Tardibacter, Sphingomonas, unclassified_c__Betaproteobacteria), Group IV (Hyphomicrobium, Hydrogenophaga, unclassified_c__Alphaproteobacteria) and Group V (Cellulomonas, Georgenia). Clearly, the increasing quinoline and NO 3 − loadings shifted the relative abundance of different groups in microbial community. The group dominant in quinoline degradation and denitrification was Group I at P2, Group II, III and IV at P3-P7, and Group II at P8-P9.

Occurrence of aerobic denitrification dependent quinoline biodegradation
MABR was operated toward high degradation quinoline efficiencies and rates of 91.5 ± 5.2% and 12.6 ± 0.8 g/(m 2 ⋅d), respectively, which are in the high range of that reported in previous aerobic (with NH 4 + released from quinoline degradation as nitrogen source) and anaerobic bioreactors (e.g., sequencing batch reactor (SBR) and upflow anaerobic sludge blanket reactor (UASB)) ( Table S3). Considering the relatively low Henry's law coefficient (1.5 × 10 − 6 atm m 3 /mol) and solid− liquid partition coefficient (0.06− 0.12 L/kg TSS) of quinoline, removal through volatilization and sorption was believed to be limited (Jeswani and Mukherji, 2015;Yaws, 1999). Furthermore, previous studies on quinoline removal in biofilm reactors suggested negligible sorption of quinoline onto the biofilm (Pribyl et al., 1997;Yan et al., 2015). Therefore, biotransformation was likely the main removal process for quinoline in our MABR. Along with efficient quinoline degradations, high NO 3 − removal rates up to 2.9 g N/(m 2 ⋅d) were obtained in MABR under oxic conditions (1.3 ± 1.0 mg O 2 /L), indicating the simultaneous occurrence of aerobic denitrification and quinoline biodegradation. This was further confirmed by strong positive associations between quinoline degradation rates/oxoO gene and NO 3 − removal rates/denitrifying genes (i.e., napA, nirK and nirS) (Figs. 1 and 4). While NO 2 − accumulation has been reported during anaerobic denitrifying quinoline biodegradation in a SBR with quinoline as the sole carbon source (Wang et al., 2015a), NO 2 − was not accumulated in this study. The accumulation of NO 2 − was likely caused by the competition for electron donors between NO 2 − and NO 3 − reductases under limited carbon source (COD/NO 3 − -N ratio of 2.5-5.1) in Wang et al. (2015a). Here, considering the high COD/NO 3 − -N ratio (7− 18) maintained and relatively higher abundance of nirK and nirS ((1.4 ± 1.1) × 10 10 copies/g VSS) than napA ((6.3 ± 8.3) × 10 9 copies/g VSS) obtained, the inhibition on NO 2 − reductases and their electron competition with NO 3 − reductases unlikely occurred. In contrast to napA, nirK and nirS, norB and nosZ were negatively corelated with increasing quinoline loadings. The inhibition of quinoline on norB and nosZ would imbalance enzymatic reaction steps of denitrification, leading to the accumulation of nitric oxide (NO) and nitrous oxide (N 2 O) (Blum et al., 2018;Wang et al., 2015a). As nosZ is generally more sensitive to changes of environmental conditions compared to other denitrifying genes, nosZ has been reported to more severely inhibited by aromatic compounds (Gao et al., 2019;Li et al., 2021). The addition of phenol from P8 relieved the inhibitory effect of quinoline on norB and nosZ, likely due to the enhanced denitrifying activities by utilizing phenol as carbon source, since phenol was more easily degradable compared to quinoline (Wang et al., 2015b). Furthermore, the increasing phenol loading resulted in over one-third reduction in quinoline degradation rates from 12.9 g/(m 2 ⋅d) (Day 184) to 8.6 g/(m 2 ⋅d) (Day 210). The decrease in quinoline degradation could be attributed to the competition for O 2 required in initial steps of aerobic biodegradations of quinoline and phenol, especially hydroxylation catalyzed by monooxygenases (Cui et al., 2004;Li et al., 2005;Zou et al., 2018). After 30 days of acclimation, efficient degradations of both quinoline and phenol were achieved at the end of P9, probably linked to the remarkable changes in microbial community composition and adjustments of functional genes (Fig. 3− 4). During aerobic quinoline biotransformation, nitrogen in quinoline is released as NH 4 + (Yan et al., 2015), as indicated by 9.3 ± 2.7 mg N/L of NH 4 + detected in effluent (P3-P9) (Fig. S2a). In this study, 48 ± 10% of nitrogen in quinoline was transformed into NH 4 + , which was in agreement with 40-48% reported in previous quinoline biodegradation studies with Pseudomonas sp. strain (Bai et al., 2010;Sun, 2009). Considering the theoretical stoichiometry of NH 4 + production from quinoline biodegradation and high bulk DO concentrations at 1.3 ± 1.0 mg O 2 /L (Fig. S2b), the activity of dissimilatory nitrate reduction to ammonium (DNRA) in MABR was assumed to be limited, and NO 3 − removal was mainly through heterotrophic aerobic denitrification. The accumulation of NH 4 + in effluent indicated the inhibition of nitrification by quinoline, phenol and/or their biotransformation products. Despite of co-metabolic degradation ability of certain organic molecules (including various hydrocarbons) through ammonia monooxygenase (AMO), autotrophic ammonia oxidizers are often sensitive to the toxicity of hydrocarbons caused by noncompetitive and competitive inhibitions on AMO (Su et al., 2021). Compared to typical heterotrophs, ammonia oxidizing bacteria and archaea were 100 and 1000 times, respectively, more sensitive to hydrocarbon toxicity (Urakawa et al., 2012(Urakawa et al., , 2019. A polishing step needs to be further developed for advanced nitrogen removal, such as combining MABR with partial nitritation-anammox process.

Enhanced formation and function of EPS under increasing quinoline loadings
EPS plays an important role in developing a protective shield and adsorbing pollutants under harsh environments (Laspidou and Rittmann, 2002). More EPS was formed under increasing quinoline loadings (Fig. 2), which agreed with previous observations of increasing EPS generations by activated sludge and biofilm in the presence of toxic compounds (Bao et al., 2016;Chen et al., 2017;Zhang et al., 2015). The increased EPS may provide a protective barrier for microorganisms in  biofilms and relieve the toxicity of quinoline at the early stage, by avoiding direct contact of quinoline molecules to bacterial cell surfaces conferring protection from hydrophobicity damage (Chen et al., 2017;Laspidou and Rittmann, 2002;Zhang et al., 2015). With the growth and enrichment of aerobic quinoline degrading microbes, the enhanced formation of EPS that often contains large quantities of aromatic structures and unsaturated fatty chains with three dimensional networks, may facilitate an efficient interact of aerobic quinoline degrading microbes with quinoline (Sheng et al., 2010;Zhang et al., 2015). This was further confirmed by significantly positive correlations of EPS characteristics with rates and metabolic pathways of quinoline degradation and NO 3 − reduction (Fig. 2c− l). Furthermore, EPS may enhance adhesion and aggregation of cells in biofilms, and thus promote the stability of biofilm structure (Flemming and Wingender, 2010;Sheng et al., 2010), as indicated by more compact biofilm structure at the later stage of operation period (Fig. S3). Thus, the possible role of EPS involved in the biofilm formation and the subsequent facilitation of quinoline biodegradation.
Compared to humic substances and other EPS components, concentrations and characteristics of PS and PN were strongly affected by various operational parameters, such as C/N ratio (Chen et al., 2017), HRT (Iorhemen et al., 2019), availability of different carbon sources (Johnsen and Karlson, 2004). In this study, enhanced PN and PS formations were obtained under increasing quinoline and phenol loadings, which might facilitate stronger cell aggregations by changing the surface charge and/or hydrophobicity of microbial cells and forming polymeric matrix (More et al., 2014). Similar observations have been reported in previous biofilm reactors (Zheng et al., 2016;He et al., 2021b), such as 2− 5 times higher concentrations of PS and PN when increasing phenol concentrations from 10 to 100 mg/L in an aerobic granular reactor (He et al., 2021b). Furthermore, the predominance of PN in LB-EPS (43.0 ± 6.9%) than in TB-EPS (24.9 ± 9.9%) (Fig. 2a) would promote cell adhesion in the outer region and efficient biofilm formation by increasing cell surface hydrophobicity and charge (Wingender et al., 1999). The significant increase in TB-PS under additional spike of pyridine and phenol during P6-P9 would strengthen structure and stability of biofilms due to its viscous and hydrophilic properties (He et al., 2021b;Hou et al., 2015). To confirm the respective contribution of EPS content to cell adhesion and aggregation within biofilms, additional adhesion and aggregation tests combined with surface characteristics and in-depth analysis of EPS compositions using nuclear magnetic resonance technique and three-dimensional excitation-emission matrix fluorescence spectroscopy are needed in future studies.
Pseudomonas and Comamonas were identified as the most abundant genera responsible for aerobic quinoline biodegradation in a nitritation granular reactor (Ramos et al., 2016). The difference could be due to different reactor types (biofilm or granular reactors), nitrogen sources (NH 4 + or NO 3 − /NO 2 − ) and operational conditions (e.g., continuous or batch operation and DO concentration), which requires for further investigations. The significant phenol degradation obtained during P8-P9 might be attributed to Raineyella, whose growth was selectively stimulated after the phenol addition (Fig. 3b). Raineyella is aerobic heterotroph, affiliating to Actinobacteria (Kim et al., 2020;Pikuta et al., 2016); yet there is very limited information on its metabolism. Here, we found that Raineyella may be capable of NO 3 − dependent phenol biodegradation as indicated by its high contribution to both denitrification (12.5%) and aromatic degradation (20.3%) based on metagenomic analysis ( Fig. 6a− b). As the dominant denitrifier, Unclassified_p_Chloroflexi accounted for 2.7 ± 1.7% of microbial community, and contributed importantly to both genomic potential for aromatic degradation (2.6 ± 1.3%) and denitrification (5.4 ± 1.6%) (Figs. 3 and 6a− b). Chloroflexi may have played a beneficial role in microbial community by degrading complex organic compounds to low molecular weight substrates to support their growth as well as other microbial populations (Speirs, 2019). This assumption was further indicated by significantly positive correlation (p < 0.05) of Unclassified_p_Chloroflexi with other microorganisms in the community (Fig. 6c). Thus, removal of quinoline and NO 3 − within biofilm was teamwork community activity, requiring the collaboration of different functional groups in microbial community.

Future perspectives
Our study demonstrated the feasibility of coupling aerobic quinoline biodegradation with denitrification in MABR for treating quinolinecontaining wastewater, and achieved stable and efficient quinoline degradations under low HRT (6− 9 h) and high loadings (up to 13.7 ± 0.3 g quinoline/(m 2 ⋅d) and 3.2 ± 0.21 g N/(m 2 ⋅d)). Aerobic biodegradation of complex organics is often faster than anaerobic biodegradation (Kristensen et al., 1995;Su et al., 2022), as indicated by up to 4.3 times higher quinoline degradation rates obtained in this study compared to previously anaerobic bioreactors (Table S3). Besides, the presence of O 2 has been suggested to facilitate a more complete biodegradation process (Kristensen et al., 1995;Su et al., 2022). Compared to conventional aerobic bioreactors (e.g., activated sludge reactor), applying aerobic biodegradation in MABR would offer an energy-and resource-efficient option for simultaneous removals of quinoline and NO 3 − , due to the bubble-free aeration and highly efficient O 2 transfer. Furthermore, we observed the fast formation of biofilm with highly enriched aerobic denitrification dependent quinoline-degrading bacteria, different functional microorganisms in microbial community and robust biofilm structure. These would allow for strong resistance towards varying operational conditions (including HRTs, loadings, co-occurrence of other toxic aromatic hydrocarbons, etc.) for our MABR, which would be beneficial in practical applications of this process. Compared to conventional membrane bioreactors using hydrophilic membranes for solids retention and water permeation, membranes in MABRs are less vulnerable to fouling, as biofilm is formed on the surface of gas permeable membranes (Lu et al., 2021;Pellicer-Nàcher et al., 2013). Yet, the challenge for MABRs is the ability to form stable biofilm with an optimal biofilm thickness and desired biofilm properties that can enhance gas permeability and substrate transfer and biodegradation, which is a controlled fouling phenomena and is essential for the MABR to be an effective technology (Lu et al., 2021). Considering the coexistence of physiologically diverse microorganisms in complex community, the exact collaboration and competitiveness between aerobic quinoline-degrading bacteria with other functional microorganisms, such as heterotrophic aerobic bacteria, in terms of organic carbon degradation, NO 3 − reduction and O 2 utilization, remain to be unraveled in future studies. Besides, the link between carbon, nitrogen and oxygen metabolisms within aerobic quinoline-degrading bacteria as well as microbial community also requires for further investigations. The accumulation of NH 4 + in effluent suggested the need of a polishing step for advanced nitrogen removal, such as combining MABR with partial nitritation-anammox process. Before full-scale implementation, MABR should be fed with real quinoline-containing wastewater (e.g., coking, coal gasification and pharmaceutical wastewater) over a long period of time in future studies, in order to optimize and validate the design of operational parameters, and evaluate adaptation and proliferation of the microbial community.