Bio-ori-DBPFP in pipelines
Figure 1 presents the bio-ori-DBPFP in pipelines which were cultured with water disinfected by Cl2 and UV/Cl2. Bio-ori-DBPFP tests, using biofilm from the two pipelines, found that three THMs (DBCM, BDCM, and TCM) and four HAAs (BCAA, TCAA, CDBAA, and BDCAA) generated. Generally, the concentration of bio-ori-HAAFP was higher than that of bio-ori-THMFP for the two pipelines. As to pipeline with Cl2 disinfection ahead, the concentration of specific bio-ori-THMFP and bio-ori-HAAFP followed the order of TCMFP > DBCMFP > BDCMFP and BDCAAFP > TCAAFP > CDBAAFP > BCAAFP, respectively. Bio-ori-THMFP and bio-ori-HAAFP for the pipeline after UV/Cl2 disinfection showed a similar trend to the case of Cl2 disinfection. With the increase in UV dose (40–120 mJ/cm2), the median value of total bio-ori-THMFP decreased from 9.98 to 7.01 µg/mg-C, whereas total bio-ori-HAAFP showed a tendency of the first decrease and then increase (from 80.13 to 33.66 then to 60.87 µg/mg-C). The results indicated that UV dosage exhibited different effects towards bio-ori-THMFP and bio-ori-HAAFP of pipeline following UV/Cl2 disinfection. Notably, the concentrations of total bio-ori-THMFP and bio-ori-HAAFP were the highest at UV dosage 40 mJ/cm2. This is consistent with the observation of several previous works (Buchanan et al. 2006; Choi and Choi 2010), which found an increase of THMFP in the DWDS at a low UV dose. Such increase of total bio-ori-THMFP and bio-ori-HAAFP can be reduced by further UV dose elevation. This phenomenon may be explained by the fact that different oxidation exposure caused the difference in water matrix (e.g., carbon/nitrogen source, types of survived bacterial) induced (Choi and Choi 2010; Z. Wang et al. 2012), which influences the property of biofilm such as EPS composition (DBPs precursor).
To estimate the formation potential of other DBPs, bio-ori-TOXFP was also tested. As shown in Fig. 1c, the concentration of bio-ori-TOXFP for pipeline following Cl2 disinfection was 1.10 mg-Cl/mg-C, which was much higher than bio-ori-THMFP or bio-ori-HAAFP. Interestingly, introducing 40 mJ/cm2 UV irradiation before Cl2 disinfection hardly changed the concentration of bio-ori-TOClFP (1.10 mg-Cl/mg-C). This phenomenon is different from that behaving by bio-ori-HAAFP. Because there is usually upregulation of genes involved in defense mechanisms and biofilm formation under low disinfectant pressure (Wicaksono et al. 2022), fine-tuning of EPS is believed to take place despite of hardly changed bio-ori-TOClFP. The increase of SCD from 0.56 to 0.60 mg-Cl2/mg-C verified this variation. Considering the difference in concentration levels for bio-ori-HAAFP (µg/L) and bio-ori-TOClFP (mg/L), it was reasonable to consider bio-ori-HAAFP behaved more sensitive to EPS tunning than bio-ori-TOClFP. With UV dose further increasing to 80 mJ/cm2, bio-ori-TOClFP decreased to a much lower level (0.12 mg-Cl/mg-C). UV dose increase from 80 to 120 mJ/cm2 raised bio-ori-TOClFP concentration to around 0.46 mg-Cl/mg-C. The phenomena may be attributed to 80 mJ/cm2 UV irradiation enhanced photolysis of DOM and broke chlorine attack sites such as the C = C bond in a locally trimmed way (Goslan et al. 2006), which will decrease the DBPFP of DOM absorbed by the biofilm. The high UV dose (120 mJ/cm2) can exert more powerful destruction to DOM and increase the low molecule fraction with high DBPFP (Choi and Choi 2010), which well explained the SCD increase and corresponding TOClFP increase when UV dose increased from 80 to 120 mJ/cm2.
Biochemical properties of biofilm
Chemical composition of EPS
The water quality and the hydraulic conditions (such as water age) have been reported to show a significant effect on the composition and spatial structure of EPS of pipeline biofilm (Srinivasan et al. 2008). Given that EPS of biofilm closely correlates with DBPs formation, the main components of EPS, proteins (PN) and polysaccharides (PS), were directly quantified (Fig. 2a). Meanwhile, other organic fractions were semiquantified by EEM-FRI (Fig. 2b).
Generally, the proportion of PN was higher than that of PS for the two pipelines. The pipeline with front Cl2 disinfection obtained biofilm EPS with 5.46 mg-PN + PS/mg-C. When 40–80 mJ/cm2 UV irradiation was dosed before Cl2 disinfection, the total concentration of PN and PS in biofilm EPS showed a tendency of the first decrease and then increase (from 5.00 to 4.81 then to 6.97 mg/mg-C). As UV pre-treatment can change the chlorine consumption kinetics of bulk water (H. Wang et al. 2014), biofilm at the same water age undoubtedly was immersed by water with different residual chlorine although the residual chlorine in influent of pipeline was the same (Fig. S2 in Supporting Information).
To draw a more detailed picture about EPS, EEM characterization was also performed. According to the EEM-FRI method, the quantity of fluorescent component (indicated with Фi,n and P value) in biofilm EPS matrix was calculated (Fig. 2b). There were three types of fractions observed: protein-like component (C1, including tyrosine-like and tryptophan-like), humus-like component (C2, including fulvic-like and humic-like), and soluble microbial product (SMP)-like component (C3). The total amount of fluorescent components (C1 + C2 + C3) obviously increased when UV pre-treatment was added on the basis of Cl2 disinfection. Because UV pre-treatment can reduce the SCD of pipeline bulk water, higher disinfectant pressure will exert to biofilm from pipeline disinfected with UV/Cl2. It was believed to happened that genes involved in defense mechanisms and biofilm formation was upregulated under the disinfectant pressure (Wicaksono et al. 2022). The increase of C1 + C2 + C3 was supposed to be a response of biofilm under such pressure. Interestingly, 80 mJ/cm2 UV dose yielded the lowest amount of protein-like (C1) and SMP-like (C3) components in biofilm EPS in the UV dose range of 40–120 mJ/cm2. These results suggested that the biofilm EPS matrix dynamically adjusted with the variety of UV doses, which macroscopically behaved as an adaptation to the environment for the biofilm.
Correlations between EPS components and bio-ori-DBPFP
Pearson correlation analysis was conducted to determine the correlation between EPS components (C1, C2, C3, PS, and PN) and the bio-ori-DBPFP (Fig. 3). The bio-ori-THMFP negatively correlated with C1, C2, C3, and PN, while bio-ori-HAAFP showed weak positive correlations with C1 and C3. Similar to bio-ori-HAAFP, bio-ori-TOClFP also positively correlated with C1 and C3. Obviously, protein-like (represented by C1 or PN) and SMP-like components in biofilm EPS were main contributors to bio-ori-DBPFP. Considering the fact that most of DOM in finished water of waterworks was medium-size molecular weight and dominated by negative-charge natural organic matters (Deng et al. 2019; Zhang et al. 2020), neutral and hydrophobic fractions of DOM were supposed to be the main force attaching to biofilm EPS based on the physicochemical characteristics of EPS (Z. Wang et al. 2012). This explained the inferior contribution to bio-ori-DBPFP for C2.
Microbial community and metabolic capabilities of biofilms
Microbial community
The above results indicated that UV pre-treatment before Cl2 disinfection indeed changed EPS composition in pipeline biofilm. Such regulation of EPS was actually a direct effect of UV irradiation on feeding water matrix of pipeline, which then influenced the growth and metabolism of biofilm. Thus, the pipeline biofilm showed different community structures and metabolic viability. It is necessary to figure out the response of biofilm community structures and metabolic behaviors to UV irradiation before Cl2 disinfection, which is helpful to understand how UV/Cl2 disinfection affects bio-ori-DBPFP.
Figure 4a shows the class level classification of microbial communities for biofilm from pipelines feeded with Cl2 and UV/Cl2 disinfected water. As to the pipeline with Cl2 disinfection ahead, microorganisms were largely dominated by Alphaproteobacteria, Gammaproteobacteria, and Bacteroidia, which accounted for above 96% of total bacteria. A similar trend was observed for the biofilm from the pipeline after UV/Cl2 disinfection. This is in line with the findings of previous studies (Becerra-Castro et al. 2016; Lautenschlager et al. 2013), which concluded that Proteobacteria, Bacteroidetes, and Actinobacteria had advantages over other microbial species in the DWDS biofilm. The UV dose showed a significant effect on microbial communities. When the UV dose increased from 40 to 120 mJ/cm2, the relative abundance of Alphaproteobacteria increased from 56.64–82.60%. Unlike Alphaproteobacteria, the relative abundance of Gammaproteobacteria showed a tendency to decrease first and then rebounded and mildly increased (from 36.8–8.2% then to 12.3%), whereas Bacteroidia showed the opposite trend. The results indicated that the dominant bacteria behaved differently under UV dosage change. Especially, we found that Bacteroidia accounted for 18.27% in the Cl2 disinfection, and the relative abundance of Bacteroidia (1.44%) was the lowest at UV dosage 120 mJ/cm2. The phenomena can be explained by the fact that Actinobacteria was resistant to UV irradiation but Bacteroidetes was sensitive to UV (Pullerits et al. 2020). The differences in repair ability to UV damage may also cause the variation of dominant bacteria.
The heatmap, comparing the dominant genera of biofilm bacteria in pipelines following Cl2 disinfection and UV/Cl2 disinfection, is shown in Fig. 4b. Sphingobacterium, Ralstonia and Flavobacterium were dominant in the biofilm from pipeline following Cl2 disinfection. When 40 mJ/cm2 UV irradiation was introduced to assist Cl2 disinfection, the difference of microbial communities on the genus level was mainly reflected in the relative abundance change of Stenotrophomonas, Ideonella, Pseudomonas. Further increase of UV dose (80 and 120 mJ/cm2) caused completely different dominant genera. The results indicated that the general pattern of the microbial community was more sensitive to UV introduction than class pattern.
Metabolic capabilities
The shift of microbial communities for biofilm affected bio-ori-DBPFP essentially through regulation of interspecies interactions and metabolism. Thus, the effects of disinfection on protein synthesis and polysaccharide synthesis in bacterial communities were further studied. Analysis of function and species using PICRUSt2 tool was performed and the results are presented in Fig. 5.
The existence of metabolic pathways was inferred by MinPath (Ye and Doak 2009), and then the abundance data of metabolic pathways in each sample were obtained. Additionally, the MetaCyc was referenced in this work as a result of which it is the largest metabolic reference database in the life sciences (Caspi et al. 2016). The results of the metabolic pathway analysis revealed that there were 60 metabolic pathways in the secondary pathway abundance analysis, among which the biofilm bacterial functions were highest in the amino acid biosynthesis, the cofactors, auxiliary groups, electron carriers, vitamin biosynthesis, the nucleoside, and nucleotide biosynthesis, indicating that the bacterial communities in the biofilms actively participated in the foundational metabolic process (Fig. S3 in Supporting Information). The differences in species composition associated with metabolic pathways between samples engaged in protein synthesis and polysaccharide metabolism were found in view of the annotated findings of the MetaCyc database. Utilizing a stratified sample metabolic route abundance table based on substantially different metabolic pathways, species composition analysis of differential pathways was carried out (Fig. 5). The investigation revealed that the Sphingobium, Methylobacterium, and Sphingomonas were principally responsible for the differences in the relative abundance of the synthesis of aromatic amino acids and rhamnose. Additionally, the former was linked to Stenotrophomonas whereas the latter was associated with Dyadobacter.
Influencing mechanism of UV/Cl2 disinfection on bio-ori-DBPFP
In conclusion, UV/Cl2 disinfection may affect bio-ori-DBPFP through the following routes: 1) UV treatment induced water matrix change such as chemical structure trim of DOM and inactivation of bacterial; 2) the UV/Cl2 disinfected water entered into pipeline and retrofitted the biofilm in the pipeline (e.g., change the amount of microorganisms and population structure), which essentially changed the properties of biofilm EPS; 3) the above two points resulted in two water-biofilm interfaces with same water age but different chlorine residual, chemical environment, and EPS for Cl2 disinfection- prepositioned and UV/Cl2 disinfection-prepositioned pipes.
Based on the three influencing pathways, one can find that bio-ori-DBPFP may directly correlate with biofilm EPS components (e.g., PN, PS, dominant bacterial species) and indirectly correlate with selection pressure (e.g., residual chlorine, ORP), nutrients (C/N/P), and other organic or inorganic constituents. To figure out the relationship between bio-ori-DBPFP and these direct or indirect influencing factors, the water quality parameters (EC, ORP, NH + 4-N, NO- 3-N, TN, phosphate, DOC, and SUVA) of pipeline bulk water (collected before biofilm sampling) were monitored (Fig. S4 in Supporting Information) and correlation analysis was performed (Fig. 6).
According to Fig. 3, bio-ori-HAA/TOXFP showed positive correlations with protein-like components (PN, C1) and SMP-like (C3) in biofilm EPS, implying that they possessed a higher propensity for DBPs production than other fractions in the biofilm matrix. The investigation revealed that the aromatic amino acid synthesis was mostly carried out by Sphingobium, Methylobacterium, Sphingomonas, and Stenotrophomonas. Correlation analysis revealed that the EC, TN, and DOC were all positively correlated to the Sphingomonas in biofilm (R2 min ≥ 0.57). However, the EC, NO- 3-N, and DOC were adversely correlated with the Sphingobium (R2 = -0.54 to -0.39) and Methylobacterium (R2 = -0.43 to -0.31), and the phosphate was also negatively correlated to the Sphingobium (R2 = -0.41). ORP and NH + 4-N showed poor correlation levels with Sphingobium, Methylobacterium, and Sphingomonas, but better correlations with Stenotrophomonas. To conclude, it can be surmised that EPS components (protein-like, SMP-like components), as well as the dominant bacterial species (Sphingobium, Methylobacterium, Sphingomonas, and Stenotrophomonas), may be the direct causes for the bio-ori-DBPFP, while water quality (EC, C/N/P) may indirectly affect the bio-ori-DBPFP by affecting the above direct indicators.
As to the pipelines with UV/Cl2 disinfection ahead, residual Cl2 showed a negative correlation with Stenotrophomona. However, UV dose was poorly positively correlated with the dominant bacteria species, while residual Cl2 exhibited better correlations with them, which suggested that these bacteria were disinfection-resistant. The Cl2 and UV resistance of both Sphingobium and Methylobacterium was observed, which belong to the Proteobacteria phylum. The increasing relative presence of the Proteobacteria phylum after disinfection may be supported by the regrowth (Becerra-Castro et al. 2016). Meanwhile, the UV dose and residual Cl2 were positively correlated to the nutrients (N/P) in water matrix, although were poorly correlated with EC and DOC.
To conclude, on the one hand, UV disinfection directly triggered water matrix change such as chemical structure modification/destruction of DOM and inactivation of UV-sensitive planktonic bacteria, thus affecting the decay rate of residual chlorine in the pipelines. On the other hand, as the direct environment of pipeline biofilm formation, the change in water matrix (Cl2, EC, C/N/P) varied the community composition of dominant microorganisms in biofilm, resulting in a variation in EPS chemical components (PN/C1/C3) through metabolism. Consequently, these changes further impacted the differences in bio-ori-DBPFP. The proposed mechanism of UV/Cl2 disinfection on the bio-ori-DBPFP within the DWDS was showed in Fig. 7.