Trace metals induce microbial risk and antimicrobial resistance in biofilm in drinking water

Introduction
Hydraulic stagnation is an important factor that affects drinking water safety. Owing to factors such as the structure of the drinking water distribution systems (DWDS) itself (dead ends and water storage tanks for high-rise buildings) or the user’s usage interval, drinking water may stagnate for hours, days, or even weeks before consumption1,2. Zhang et al. investigated the risk of bacteria caused by stagnation of drinking water under winter heating conditions, and the results showed that stagnation led to an increase in water temperature, a significant increase in the number and activity of bacteria, and triggered complex bacterial community interactions3. Later studies showed that when algal blooms occur in stagnant drinking water, water quality deteriorates severely and that the heterotrophic plate counts of planktonic bacteria exceed 104 CFU·mL−1 within 24 h of stagnation4. In addition, Ra et al. believed that due to the rising vacancy rate of buildings, the decline in water consumption and stagnation periods may lead to an increase in bulk water metal concentrations, a decrease in disinfectant residue concentrations, and an increase in pathogenic bacteria such as Legionella pneumophila and Pseudomonas aeruginosa, etc5. Bacterial proliferation can lead to changes in the taste, odor, and color of drinking water, resulting in aesthetic issues6. In addition, safety hazards may arise owing to the potential risk of pathogenic bacterial proliferation. For example, ESKAPE bacteria (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter spp.) have been reported to pose a risk of disease and are prone to developing antibiotic resistance7.
A substantial portion of the bacteria in the drinking water environment exists in the form of biofilms, which adhere to the walls of pipes or containers, causing corrosion of the lining and inducing biochemical reactions between microbes and metals. However, this problem can become even more complex, and trace amounts of released metal substances typically occur when stable corrosion scaling is disturbed by sudden changes in conveyed water quality, particularly when water sources are converted from groundwater to surface water or when new water sources are utilized through long-distance water transport8. Tian et al.‘s study showed that large amounts of metal spillage can occur when pipes contain some limescale, that the concentration of iron detected in their experiments can reach up to 1.0 mg·L−1 9. In addition, because of the residual coagulants, trace metal pollutants may spread with the flow of drinking water10. These trace metals may lead to sensory changes in drinking water and can be transported to consumers’ homes, thereby resulting in health risks11. Previous studies have suggested that metal pollution can generate biological toxicity, leading to a decrease in overall microbial biomass12. However, new research suggests that the presence of trace metals may lead to an increase in biofilm formation in pipelines13.
Owing to the selective pressure of disinfectants and disinfection byproducts, antibiotic-resistant bacteria (ARB) and antibiotic-resistance genes (ARGs) are stored in biofilms, spread, and transferred through multiple complex pathways14. Metal pollution is another important factor that affects ARGs. Li et al. suggested that low-level metal concentrations might contribute to horizontal gene transfer (HGT) of antibiotic-resistant genes15. Although the impact of biofilms on drinking water safety is profound, there is insufficient research on this topic. Therefore, it is necessary to study the impact of trace metal pollutants on the macroscopic and microscopic characteristics of biofilms in stagnant drinking waters.
At present, most of the public’s attention to water quality safety is concentrated on water bodies, but insufficient on the risks hidden in biofilms. The main purpose of this study was to explore the impact of metal contaminants on biofilm development and water quality under conditions of drinking water stagnation. The specific objectives were to (1) determine the quality of stagnant drinking water, (2) investigate the characteristic changes in the biological properties of biofilms, and (3) examine the threat of opportunistic pathogens and ARGs caused by trace metal contaminants in stagnant drinking water. We aim to provide a new perspective and scientific early warning regarding the water quality and evolution of microbial characteristics caused by metal pollution in stagnant drinking water, thus providing a reference and recommendations for broader drinking water safety management strategies and water system policies.
Results
Changes of water quality parameters in stagnant state
Common drinking water quality parameters were monitored during the stagnation experiments. Turbidity differed significantly among all systems, except for the (Fe(III)) and c (Fe(III) + Al(III)) systems (Supplementary Figure 1). Meanwhile, the variations in turbidity were the highest compared to those of all other water quality parameters (Fig. 1). The control system maintained relatively stable fluctuations, indicating that stagnation had no effect on the turbidity. The presence of Fe(III) alone and coexisting systems (a, c) immediately increased the turbidity at a slightly increasing rate. In contrast, the turbidity of the experimental system containing Al(III) (b) gradually decreased. In conclusion, the presence of Fe(III) in stagnant water makes turbidity more unstable (increasing coverage area).

Control (orange square): collected drinking water with no addition of metals; a (blue circle): Fe(III) input system; b (green triangle): Al(III) input system; c (red lozenge): Fe(III) + Al(III) in the biofilm. The line represents the fitted trend. The shaded area represents the 99% control interval. The text in the figure provides a description and true evaluation of the fitting situation. Error bars represent the standard deviation of triplicate experiments.
The pH of stagnant water was monitored, and it was found that stagnation caused an increase in the pH value (from 7.0 to 8.2). The presence of metallic substances had little influence on these changes (Fig. 1 and Supplementary Figure 1). In the early stages of the experiments (0–20 h), we observed a sharp increase in the pH with the addition of metals, whereas the control showed less dramatic changes. In addition, none of these differences was statistically significant between Fe(III) and Al(III), even when both metals were present simultaneously. Organic carbon is the main food source for microorganisms in drinking water. At 24–48 h, we observed a faster declining trend in TOC values, which demonstrates that stagnant drinking water harbors a greater number of microorganisms that primarily rely on carbon as their main energy source (Fig. 1). The TOC concentration in the water showed a decreasing trend with no significant differences in any of the stagnant experiments (Fig. 1 and Supplementary Figure 1), indicating that the presence of metals did not influence the reduction.
The results obtained for the nitrogen cycle and the relevant transformations during the stagnation period are shown in Fig. 1 and Supplementary Figure 1. The concentrations of NO3–-N in the a (Fe(III)) and c (Fe(III) + Al(III)) systems showed minimal increases during the experiment, whereas those in the control and b (Al(III)) systems exhibited slight decreases. The concentration of NO3–-N was notably higher for Fe(III) than for the other systems (Supplementary Figure 1). The concentration of NO2–-N remained stable with a slight decrease in all systems, and metals had no significant influence on NO2–-N during stagnation (Fig. 1 and Supplementary Figure 1). The concentration of NH3-N continued to increase slightly during the stagnation process (Fig. 1). The boxplot shows that the metals have no obvious influence on NH3-N, as shown in Supplementary Figure 1. The transformations of NO3–-N, NO2–-N, and NH3-N are worth investigating. We speculate that the observed changes may be attributed to the succession of the bacterial communities.
Biomass and extracellular polymeric substances (EPS)
As shown in Fig. 2, the cultivable bacteria in all the experimental systems underwent an initial increase, followed by a decrease after 168 h of stagnant drinking water. The bacterial curve was divided into three stages according to the growth rate: S1, exponential growth phase, 0–16 h; S2, stationary phase, 16–60 h; and S3, decay phase, 60–168 h. In the initial stage of S1 (0–4 h), many free-floating bacteria in stagnant water aggregated on the surface of the carrier owing to adsorption, thereby forming a biofilm16. During stage S1, bacteria on the biofilm proliferated rapidly at an exponential rate because of the relatively abundant substrate concentration, whereas metals had little influence on bacterial growth. In stage S2, it is believed that the consumption of organic matter and inter-species competition became the primary factors limiting the rate of bacterial proliferation, which peaked at 60 h (compared to 48 h in the control group). In our experiments, the bacteria in the metal-containing systems (a, b, and c) grew faster than those in the control system. At 60 h, the bacterial concentrations of a (8.47 × 105 CFU·mL−1), b (1.59 × 106 CFU·mL−1), and c (3.25 × 106 CFU·mL−1) were 4.32 × 105 CFU·mL−1, 1.18 × 106 CFU·mL−1, and 2.84 × 106 CFU·mL−1, respectively, higher than that of the control (4.15 × 105 CFU·mL−1). This phenomenon indicates that, under stagnant conditions, metals significantly promote the proliferation of bacterial biofilms. The promoting effect of Al(III) on bacteria in biofilms was greater than that of Fe(III), and the coexistence of both led to rapid bacterial growth in the biofilm. Finally, at stage S3, the TOC of the water reaches its lowest level (Fig. 1), which further restricts the growth of bacteria17, which have to face nutrient-limited conditions and higher competition, and the scales of the equilibrium are now leaning towards dead rather than live cells.

Control (dark blue square): collected drinking water with no addition of metals; a (azure circle): Fe(III) input system; b (yellow triangle): Al(III) input system; c (red lozenge): Fe(III) + Al(III) input system. The experimental equipment was operated continuously for 168 h. S1, S2, and S3 represent the exponential growth, stationary, and decay phases, respectively. Error bars represent the standard deviation of the triplicate experiments.
In addition, we investigated the toxicity of metals in biofilms using EPS. As shown in Fig. 3, there were significant differences in the EPS on the biofilm between the metal-added systems (a, b, and c) and the control system (p < 0.05, p < 0.01, p < 0.01). The presence of Fe(III) and Al(III) increased the bacterial EPS production, with Al(III) showing a more prominent contribution. Fe(III) and Al(III) are both coagulants that help biofilms capture more organic matter during EPS formation10,18. In addition, stimulation by lower metal concentrations may also be one of the reasons for the increase in EPS19. The increase in protein levels is shown in Supplementary Figure 2 and is positively correlated with the addition of metals, which may be because bacteria secrete proteins to form covalent bonds with metals, preventing their entry into cells20. The increase in the EPS of the Al(III)-added systems also explains why the toxic effect of Al(III) has not been exerted in the bacterial growth stages of S2 and S3 (Fig. 3), and why the stagnation of drinking water would help to enhance the tolerance of bacteria to trace metals.

Red box and circle: polysaccharides; blue box and circle: proteins. Control: collected drinking water with no addition of metals; a: Fe(III) input system; b: Al(III) input system; c: Fe(III) + Al(III) input system. Path coefficients are indicated by asterisks (**p < 0.01; *p < 0.05; N. S., p > 0.05).
Differences in functional traits
The significantly different functional traits for each system are shown in Supplementary Figure 3. At level 2, the most significantly altered functions were xenobiotic biodegradation and metabolism, metabolism of other amino acids, translation, aging, and the sensory system (Supplementary Figure 3). It can be observed that metals lead to a slowdown in xenobiotic and amino acid degradation, while translation, aging, and the sensory system are enhanced. At level 3, the effects of fatty acid degradation, butanoate metabolism, porphyrin metabolism, tryptophan metabolism, and cyanoamino acid metabolism were weakened in the presence of metals, and were more pronounced at higher pollution levels (Supplementary Figure 3). Conversely, amino sugar and nucleotide sugar metabolism, aminoacyl-tRNA biosynthesis, ribosomes, and o-antigen nucleotide sugar biosynthesis were also enhanced. Finally, metal stress reinforces the longevity-regulating pathway and antifolate resistance.
Modifications in bacterial communities and pathogens
Taxonomic annotation of the processed metagenomic datasets revealed that Pseudomonadota, Actinomycetota, and Bacteroidota were the dominant populations at the phylum level, with Pseudomonadota having the highest relative abundance (Supplementary Figure 4). This finding is consistent with previous studies on the dominance of bacteria in other freshwater ecosystems21. Compared to the control system (Pseudomonadota accounted for 74.49% and Actinomycetota accounted for 16.56%), the relative abundance of Pseudomonadota gradually increased in treatments a and b (75.51% and 80.45%, respectively), whereas the relative abundance of Actinomycetota decreased (12.51% and 10.06%, respectively). Conversely, in treatment c, the proportion of Pseudomonadota decreased (67.56%), and the proportion of Actinomycetota increased (18.86%). Some bacteria in Pseudomonadota (such as Pseudomonas and Sphingomonas) can switch from an active reproductive state to a dormant state, thereby adapting to low-nutrient conditions and environmental stress22. An increase in Actinomycetota may lead to the production of more geosmin (GSM), thereby exceeding the odor threshold of drinking water23.
In a finer division, we found that Limnobacter, Rhodococcus, Sphingobium, Variovorax, Brevundimonas, Aquabacterium, Sphingopyxis, and Hyphomicrobium comprised the dominant populations in the community (Fig. 4). Compared to the control, metal pollutants resulted in varying degrees of increase in the relative abundance of Limnobacter (control, 13.20%; a, 13.87%; b, 24.15%; c, 19.46%), whereas the presence of Rhodococcus showed an inverse trend (control, 11.13%; a, 5.68%; b, 6.43%; c, 4.91%). Limnobacter, a sulfur-oxidizing bacterium (SOB), contributes to the corrosion of metals24. More species with statistically significant differences (p < 0.05) are shown in Supplementary Figure 5 using the Kruskal-Wallis H test for Brevundimonas, Parviterribacteraceae, Solirubrobacteraceae, Caulobacter, Perlucidibaca. Parviterribacteraceae and Solirubrobacteraceae were found to be significantly increased in the experimental group containing Fe(III) (a and c), which suggests that Fe(III) contamination may alter the composition of the drinking water bacterial community. The results of the Principal Coordinate Analysis (PCoA) confirmed this hypothesis and presented differences in the community structure of the three experimental systems compared with the control system (Supplementary Figure 6). In addition, the results of alpha diversity analysis indicated that the experimental systems showed varying degrees of increase in species abundance and diversity (Fig. 4), which was most pronounced in system a.

a Bacterial community composition at the genus level in the top 20. b, c Alpha diversity at the species level by Chao1 and Simpson indices. d The relative abundance of the top 50 pathogens. e Abundance of ARGs based on annotated gene counts. f PCoA plots showing the differences in ARGs among different treatments based on the Bray-Curtis distance (red circle: control; blue triangle: a; yellow lozenge: b; purple square: c). a: Fe(III) input system; b: Al(III) input system; c: Fe(III) + Al(III) input system; Control: collected drinking water with no added metals. All results represent the mean of the three parallel experiments.
Figure 4 showed the relative abundances of the top 50 most abundant pathogens among the control, a, b, and c treatment groups. Salmonella enterica, Pseudomonas aeruginosa, and Xanthomonas oryzae were the top three pathogens in terms of their relative abundance. Common pathogens, such as Escherichia coli, Legionella pneumophila, Staphylococcus aureus, and Mycobacterium tuberculosis, were also found in stagnant drinking water, constituting a significant proportion. Overall, the relative abundance of pathogens was the highest in system b, whereas systems a and c exhibited lower abundances than the control system. For example, the relative abundance of Salmonella enterica, a well-known pathogen, increased by 3.96% and 4.72% in treatments containing Al(III) compared to the control system (b and c), whereas the abundance of Pseudomonas aeruginosa increased by 5.25% and 0.02%, respectively. In the treatment system where Fe(III) was present alone (a), the abundance of these two pathogens appeared to decrease (by 4.41% and 2.63%, respectively). PCoA revealed differences in pathogenic bacteria (Supplementary Figure 7), and some differences were observed in the distance between the center of the circle composed of samples from different experimental groups, especially the control and metal-containing groups. The above evidence points to the possibility that the presence of metals may cause some changes in the composition of the pathogenic bacterial community, which may induce population dominance of some pathogens.
Variation of ARGs
The metagenomic results displayed all the ARGs with specific annotations for the top 20 ARGs in terms of abundance (Fig. 4). Overall, experimental systems containing metals exhibited higher levels of ARGs, increasing the overall exposure risk of resistance genes by 7.81, 5.92, and 6.98%, respectively (comparison of a, b, and c with the control). The dominant positions in the distribution of ARGs were macB, evgS, and tetA(58), which were the top three. Under metal stress, the exposure risk of macB increased by 6.26%, 6.46%, and 7.49% (compared with the control system) in systems a, b, and c, respectively. The abundances of evgS and tetA(58) increased by 4.26%, 4.21%, and 4.21%, and 5.99%, 4.55%, and 5.86%, respectively. Most of the other ARGs also showed an upward trend in the experimental systems containing metals, indicating that the resistance developed under metal stress was collective. The PCoA results reflect intergroup differences in the distribution of ARGs (Fig. 4).
Co-occurrence networks of pathogens and ARGs
Co-occurrence network analysis demonstrated a significant positive correlation between pathogens and resistance genes in each experimental system (Fig. 5, p < 0.05), which was further strengthened in the iron-containing experimental systems (b and d). The increase in the number of pathogens in system b (increase in the size of the circles) indicates that, in the presence of Al(III), pathogens may be associated with more ARGs, potentially leading to the emergence of more antibiotic-resistant individuals. Units in systems a and c had higher average degrees (Supplementary Table 1), indicating an increased correlation between each pathogen and resistance genes, possibly leading to an increase in the selection of more antibiotic-resistant pathogens. In all experimental systems, adeL, a gene commonly associated with the development of multidrug resistance, consistently showed a high degree of association, which may lead to public health concerns25. The most significant promoting effect in this regard is tetA(58), comparing a and c to the control, which suggests that the presence of Fe(III) may increase the risk level of tetA(58).

Dark orange circle: pathogens; azure circle: ARGs. Red lines indicate a positive correlation (p < 0.05). The size of each node is correlated with its degree, with larger sizes indicating larger degrees. a Drinking water collected without metal addition. b Fe(III) input system. c Al(III) input system; d Fe(III) + Al(III) input system.
Discussion
Stagnation occurs in many areas of DWDS, such as dead-end pipes, storage tanks, and taps. The decay of residual chlorine is inevitable under stagnant conditions, as confirmed by previous studies26. Therefore, we conducted sampling surveys of freshwater bodies at different locations in Hohhot (Supplementary Table 2), which confirmed this viewpoint. Organic matter, bacteria, temperature, and pH are the potential causes of residual chlorine decay4. After the treated water leaves the treatment plant and is transported through the distribution system to consumers’ homes, the original chlorine concentration is likely to have decreased to a lower level. Under hydraulic stagnation, low levels of residual chlorine fail to meet drinking water safety standards, thereby creating an ideal environment for bacterial growth.
Although separate stagnation did not lead to a significant increase in turbidity, free-form Fe(III) still had a significant promoting effect on turbidity (Fig. 1). This can also be clearly observed in the images (Supplementary Figure 8); trace Fe(III) had a significant impact on the “yellow” appearance of the water. After undergoing oxidation-deposition processes, these free iron ions form iron oxides anchored on the biofilm or float in the water as large particulate matter27. Besides, in closed faucets, the presence of iron is unavoidable, and this phenomenon may occur even at concentrations that comply with standards. This may contribute to the aesthetic sensory issues caused by iron. Changes in the quantities of metal pollutants were monitored during the experiment (Supplementary Figure 9). Fe(III) in drinking water is mostly present in the form of oxides anchored to the surface of pipelines, forming scales and biofilms28. The hydrolysis products of aluminum can bind to high-molecular-weight biopolymers, thus serving as a framework for biofilm29. Therefore, during biofilm formation, trace amounts of Al(III) are adsorbed onto the biofilm. When the bacterial community disintegrates, the aluminum originally adsorbed onto the biofilm is released back into drinking water. Furthermore, the presence of other metals can interfere with adsorption and precipitation effects, leading to higher levels of contamination30. This also explains the higher turbidity detected in system c (Fig. 1).
As mentioned in the literature review, the reduction of nitrates and oxidation of iron may generate hydroxide ions, leading to an increase in the pH31. However, this may not be sufficient to fully explain our findings. Another possible explanation for this is primarily due to the release and fixation of CO2, which is also influenced by temperature32. In addition, an increase in pH may be associated with chlorine decay. Because of the addition of disinfectants, alkalinity decreases; in some processes, alkalinity may even be actively adjusted to a lower level to enhance the retention of residual chlorine. However, owing to the rapid decay of the disinfectant under stagnant conditions, the pH value returned to its original level.
A strong relationship between stagnation and NO–3-N has been reported in the literature4. Nitrospira was more abundant in these systems. Nitrospira members are well-known nitrite-oxidizing bacteria, while Pseudomonadota and Actinomycetota are better known as denitrifiers. This result may be related to the activities of the nitrifying bacteria, denitrifying bacteria, and ammonifiers. Previous studies have found that metal stress affects the N cycling process, mainly by affecting nitrification33. NO3–-N, NO2–-N, and NH3-N are transformed into each other in the water body, forming a balanced cycle4,34. Therefore, competition and natural selection within a bacterial community can cause fluctuations in this cycle. Although stagnation has some influence on these indicators, all the values are within the limits of the relevant standards.
Under stagnant conditions, the dissolved organic matter provided favorable conditions for the overgrowth of bacteria in the biofilm (Fig. 1). The increasing decline in organic matter during phase S2 (stationary phase; 16–60 h) also reflected a larger overall bacterial population (Fig. 2). It has been confirmed that overnight stagnation of drinking water could increase to 6.99 × 104 cells·mL−1 over 12 h35. In our study, we found explosive growth of bacteria on biofilms in stagnant drinking water, with a peak value of 4.15 × 105 CFU·mL−1. Similar to other studies, stagnation can lead to bacterial multiplication in drinking water6,35. However, our study showed that biofilms contain a higher number of bacteria and present a much higher microbial risk than water.
Furthermore, Li et al. found that the input of Mn(II) resulted in one order of magnitude higher biomass in the biofilm in the water supply pipe13. This phenomenon was significantly amplified by more than seven times in the presence of metal pollutants and was even more pronounced in the experimental system containing Fe(III)-Al(III), up to 3.25 × 106 CFU·mL−1 in our study. Hu et al. found that metal Fe-Mn stress enhanced microbial adhesion to the carrier and promoted microbial metabolic activity within 30 days, leading more viable but non-culturable (VBNC) bacteria growing within 70 days36. In our study, we found that even under a short period of metal stress (48 h), the biofilm increased significantly, which deserves high vigilance.
Many metal components are used in water distribution systems, particularly at the end of the distribution network, significantly increasing the risk of metal pollutant leaching. We believe that the ionization, adsorption, complexation, and anchoring of these metallic substances, as well as the larger individual particulate matter produced by oxidation, aggregate the bacteria in drinking water, thereby increasing the formation of biofilms. In addition, these metals have been shown to enhance EPS production and are therefore an indication of large biofilm expansion (Fig. 3). One can imagine that when stagnant conditions prevail, just as when one leaves their house, they collide with the released metal pollutant. As such, bacteria have a grand feast on the corroded faucets and pipes in their home, while EPS have “decorated” this “palace” to their liking; thanks to the assistance of trace Fe(III) and Al(III), the faucet is now become an all-you-can-eat buffet for the bacteria.
The differences in functional genes explain the response mechanisms of the bacteria (Supplementary Figure 3). The toxicity of Fe(III) may arise from the oxidation of iron in drinking water to iron-based nanoparticles and cause higher levels of reactive oxygen species (ROS), which may stimulate oxidative stress. Fe(III) may be reduced to Fe(II) owing to the reduction of bacteria, and Fe(II) is a catalyst for ROS, which can produce hydroxyl radicals that lead to lipid peroxidation, DNA damage, and even cell death37. The promoting effect of Al(III) on bacterial numbers is attributed to hormesis and polymerization38. Metal oxide surfaces are positively charged in the pH range of natural environments, while bacterial surfaces dominantly exhibit an overall negative charge; consequently, an attraction between the opposite surface charges is expected, which promotes biofilm formation39,40. The neutralization of charge, compression of the double layer, and reduction in the energy barrier of repulsion may also explain why these metal cations promote biofilm formation41.
Furthermore, according to the results described by level 2 functional genes, the differences in xenobiotic biodegradation and metabolism further reflect the indirect toxicity of metal pollutants (Supplementary Figure 3). The reduction in amino acid metabolism and the strengthening of translation clearly reflect bacterial resistance mechanisms, which help in the interception, transportation, and efflux of metals by producing more proteins20,38,42. Thus, higher protein levels were observed (Fig. 3 and Supplementary Figure 2). The increase in functional genes associated with aging and the sensory system reflects the stimulating effect of metals (Supplementary Figure 3). This suggests that, under higher metal stress, although the emergency response functions are enhanced, the aging mechanisms also affect their life cycle, which is also reflected in Stage 3 of Supplementary Figure 3. Level 3 reflects a more detailed functional pathway than Level 2, and the results are similar to those of Level 2 in Supplementary Figure 3, which reflects the bacterial response to metal stress. These functional changes further strengthen the fact that biofilm multiplication occurs after exposure to metal stress. Stimulated by metal stress, several metabolic and protein secretion functions of bacteria are upregulated to produce organic acids, EPS, and mucus, which play crucial roles in microbial aggregation, adsorption, and biofilm adhesion43. In addition, these compounds can bind and sequester specific metals, reduce their toxicity, and provide temporary resistance to metal stress, thus providing a safer habitat for bacteria36.
In fact, biofilm formation and the adhesion of microorganisms are influenced by several factors, such as pipe characteristics (e.g., material, roughness), water physicochemical parameters (pH, temperature, load in organic compounds, existence of minerals), and characteristics of microorganisms (e.g., ability to produce extracellular polymeric substances, cell hydrophobicity, and motility)44. Two studies by Hu et al. revealed that Fe(III) and Al(III) influence biofilm development and activity not only by directly impacting microbial physiology but also by indirectly affecting EPS constituents45,46. The first barrier to bacterial resistance to metals is the bacterial wall. The presence of the S layer, particularly in Gram+, allows the adsorption of metal ions and therefore indirectly leads to a decrease in the metal concentration and, therefore, its toxicity towards the most sensitive bacterial strains. Phenotypic tolerance can occur within the biofilms.
Stress from Fe(III) and Al(III) induces adaptive changes in the bacterial community. Limnobacter is a denitrifying bacterium, and an increase in its dominance in the Al(III) experimental group may lead to an increase in intermediate products such as NO2- 47, which is also a typical thiosulfate oxidizing bacterium responsible for the change and circulation of ions in drinking water48. The increase in species diversity in biofilms also leads to more complex drinking water safety hazards as well as pathogenic bacteria. We identified various pathogens in the biofilms after stagnation (Fig. 4). Salmonella enterica was found to have the highest abundance and is considered a significant cause of foodborne and severe systemic infections49. Pseudomonas aeruginosa is a typical pathogen in drinking water and is considered to have high resistance owing to its low outer membrane permeability50. Other well-known pathogens such as Escherichia coli, Legionella pneumophila, Staphylococcus aureus, and Mycobacterium tuberculosis are also present in stagnant drinking water. Most pathogens are already inactivated during the raw-water treatment stage; however, some survivors may still persist. The appearance of persistent bacteria could be linked to the ‘phenotypic switching’ between growth and dormancy. Genetic mechanisms or environmental factors can trigger or increase the frequency of this switch51. The presence of disinfectants can induce them to enter the VBNC state, enabling them to easily withstand environmental pressures easily52. Under stagnant conditions, the decay of disinfectants allows dormant pathogens to regrow and proliferate in large numbers, which may allow them to retain their pathogenicity, posing serious health risks to residents52. In summary, the increase in the number of bacteria, the development of biofilms due to short-term stagnation, and the effects of trace metals further amplify the threat of pathogens. However, these threatening biofilms are generally difficult to detect through direct monitoring of water bodies. When the biofilm sheds, danger spreads, leading to public health threats.
Previous studies have reported that low concentrations of metals (low-toxicity metals such as Fe, Zn, Cu, Mn, etc., around 5–30 mg·L−1) have a significant promoting effect on ARGs15. Our study found that even trace amounts of metals (lower than 0.3 mg·L−1) in stagnant drinking water could promote ARGs production (Fig. 4). It is reasonable to predict that higher levels of metal stress may be experienced during long-term biofilm development owing to the accumulation effect of metals in the biofilm, which may induce more ARGs. it is also known that co-resistance to metals and antibiotics exists, it is all the more frightening since it has been reported recently that this co-resistance appears to be synergistic in bacterial isolates via similar mechanisms, and this synergy has the potential to amplify ARG in the environment, which can be transferred clinically53. Another recent study showed that the presence of metal resistance genes (HMRG) is strongly believed to play a major role in the proliferation of ARGs. In fact, cross-resistance between antibiotics and metals has been observed in Pseudomonas aeruginosa strains. In addition, the majority of ARGs and HMRG are borne on mobile genetic elements54. Interestingly, this enhancing effect of metals applies to most resistance genes and is collective in nature. MacB (macrolide antibiotic), evgS (tetracycline antibiotic, penam, fluoroquinolone antibiotic, macrolide antibiotic) and tetA(58) (tetracycline antibiotic) occupied the top three in abundance. The high abundance of these ARGs deserves our vigilance of biofilms in drinking water that humans may contact directly in daily life, and special attention is needed for tetracyclic and macrolide antibiotics. The current monitoring of drinking water is often inadequate for ARGs, as these risks are hidden in biofilms that can burst.
The level of co-occurrence of the abundance of the top 20 pathogens and ARGs was analyzed to explore potential pathogen hosts for ARGs in biofilms of stagnant drinking water (Fig. 5). These results indicate that, in stagnant drinking water, Fe(III) stress strengthens the positive correlation between ARGs and a greater number of pathogens, which suggests that more potential antibiotic-resistant pathogens may emerge. Table 1 shows the top 3 abundant ARGs and their pathogens that may be transmitted in each system. Bombaywala et al. showed that when challenged by oxidative stress, microbial communities tend to accumulate larger amounts of ARGs by triggering non-lethal mutagenesis to increase the potential for transmission, including pathogens55. Our study further demonstrated and refined this result, showing that metal stress effectively increases the level of coexistence between ARGs and pathogens. When one or more ARGs spread to pathogens through HGT, they become stronger and survive more easily56. These multidrug-resistant pathogens may pose a higher risk of transmission and greater threat to human health57. Multidrug-resistant pathogens have been reported in drinking water that led to infections in clinical and community settings, posing a public health threat to susceptible subjects58,59. According to our study, we believe that these risks most likely arise from biofilms in drinking water.
Owing to water consumption habits and limitations of the water supply structure design, stagnation of drinking water is an unavoidable phenomenon. Stagnation of drinking water leads to the decay of residual chlorine and excess biochemical indicators in water, which may threaten the health of water users, leading to complaints and public opinion pressure on government departments and water-supply units. In addition, under stagnant conditions, biofilms in drinking water proliferate on a large scale, amplifying many microbial risks, such as the propagation and spread of ARGs and pathogens, which mainly occur within the biofilm on the surface of the contact medium. When users turn the tap on, the risk of pollutant exposure occurs because of the change in hydraulic conditions60. Moreover, owing to the corrosion of metal materials and coagulant residues, even compliant levels of trace metals can further trigger the aforementioned series of hazards and render them riskier10. It can be seen that many potential risks in drinking water are quietly occurring, and biofilm is a ‘key culprit’, which poses a serious threat to drinking water health and public health.
The issue of drinking water safety has always been an important concern for the public and government. In some parts of China, it is necessary to clean the water tanks of high-rise buildings at least twice a year to reduce the accumulation of dirt and biofilms inside the tanks. Some countries and regions have already begun to reduce the use of metal-based water pipes; however, in households, components or fixtures made of metal remain the preferred choice for decorations61. Designing new materials and coatings for drinking water supplies and contacts to replace or cover the metal media is an option. Some water departments regularly inspect water quality and pipes in DWDS. However, at the end of the distribution system, these measures are difficult to implement in the community or users’ homes because of private or ownership issues. At the end of the drinking water supply, many citizens installed water purifiers. However, recent research has shown that these public or household water purifiers still pose a risk of releasing microorganisms and organic pollutants62.
To reduce the occurrence and spread of these hidden risks in drinking water, we recommend: 1. More online water monitoring devices and safety checks should be effectively implemented at the end of the drinking water supply and final faucets, and relevant connectivity systems and real-time monitoring should be in place. 2. Appropriate and scientific adjustment of the limits of pollutants in water standards should be upgraded, considering not only their own health risks to human beings but also their hidden risks in the water environment. 3. Reduce the use of metal components in water distribution systems; consider more plastic tubes, but be wary of their spillage from organics. 4. Optimize the water supply structure to reduce the water retention time. 5. For sudden water pollution accidents, additional safety rehearsals and simulation exercises should be performed. 6. Hidden risks in drinking water should be further reported and disseminated in the media so that users are genuinely aware of these phenomena, thereby promoting proactive prevention and protection.
Finally, most research and reports have focused on drinking water bodies, but less on the problem of biofilm formation in drinking water. However, we believe that biofilms generally pose a potential risk to drinking water safety, which should not be overlooked. Moreover, there has been a lack of effective solutions to deal with this problem (in lieu of regular replacement of pipes or pipe flushing), and more technical approaches could be further investigated to solve the problem of biofilm formation in drinking water.
Methods
Set-up of drinking water stagnation experiments
We collected 2 L of tap water for each sample from Hohhot (Latitude 40.48 °N, longitude 111.41 °E), which is in northern China, covers 3.5 million of the urban population, as a freshwater sample for the experiments. Tap water was run for 20 min to avoid the possibility that the drinking water had been stagnant for a long time before the experiment. Based on previously published studies, we set up a plastic container for each experimental system (2 L tap water) and placed it in a dark environment (Supplementary Figure 10), with small glass balls (each surface area was consistent and 1 cm in diameter) used for the growth and collection of biofilms63. All the experimental containers and balls were cleaned and autoclaved prior to the start of the experiment to ensure that there were no interferences. Fe(III) and Al(III) are common metal pollutants found in drinking water systems, and were selected as the invading metals for our research. The input concentration of Fe(III) (add as FeCl3·6H2O, AR) and Al(III) (add as poly aluminum chloride, of which the aluminum content is ~29%) were set as 0.2 mg·L−1 and 0.3 mg·L−1, that reference was made to published literatures as well as standards9,38,64,65. After the start of the drinking water stagnation experiment, we collected water and biofilm samples at time intervals of 4, 8, 12, 16, 20, 24, 30, 36, 42, 48, 60, 72, 84, 96, 120, 144, and 168 h, followed by immediate analysis. The characteristics of the tap water used in the experiments are listed in Supplementary Table 3. All experiments were conducted in triplicates at room temperature (22 ± 2 °C).
Collection and determination of biological parameters: bacteria and EPS
For the biological part of the study, a single glass bead was selected (one set of three portions) at different times during the experiment, and the biofilms that formed on the glass beads were collected for subsequent tests.
We tested biofilm samples collected at each time interval for heterotrophic plate counts (HPC) The collected biofilms were transferred into 1.5 mL centrifuge tubes containing 1 mL of sterile saline solution (0.8% sodium chloride, 0.08% potassium chloride). The entire centrifuge tube was sonicated in an ultra-sound (KQ-600DV, Kunshan Ultrasonic Instrument Co., Ltd., China) for 10 min (40 kHz, 600 W) and vortexed using a mixer (MX-S, DLAB, China) for 30 s. Thus, we obtained a bacterial suspension from the biofilm. R2A solid medium was used to cultivate biofilm-forming bacteria in a thermostatic incubator (HuYue Instrument Equipment Co., Ltd., China) at 25 °C for 7 days.
Owing to the limitations of pre-experiments and minimum detection, we chose to perform EPS content determination at 48 h for samples. The biofilms were placed in a centrifuge tube filled with 50 mL saline solution. The solution was sonicated at 40 kHz for 1 min, heated in a water bath at 80 °C for 30 min, and centrifuged at 8000 rpm at 20 °C for 4 min66. The supernatant was collected in a tube and the EPS solution was successfully obtained. The protein fraction of EPS was determined using the Bradford assay (Coomassie Brilliant Blue staining and spectrophotometry at A = 595 nm)67. The polysaccharides of EPS are chromogenic using concentrated sulfuric acid and phenol, which are then determined spectrophotometrically68.
Determination of the chemical parameters of drinking water
We measured water quality parameters during the stagnant experiment, including turbidity, pH, total organic carbon (TOC), nitrate nitrogen (NO3–-N), nitrite nitrogen (NO2–-N), and ammonia nitrogen (NH3-N). In addition, the remaining Fe(III) and Al(III) content was measured. The details of the (common) detection methods are listed in Supplementary Table 4. Residual chlorine in water from municipal taps was sampled in the city and detected using previously published methods26.
Metagenomics analysis
The processed biofilm samples (48 h time interval) were further enriched by passing through a 0.22-nm sterile filter membrane and collected in sterile centrifuge tubes for storage at −80 °C. One set of three portions was used for all the samples. All the samples were sent to Shanghai Meiji Biomedical Technology Co., Ltd. (Shanghai, China) for metagenomic sequencing. DNA extraction was performed using E.Z.N.A. software®. Soil DNA Kit (Omega Bio-tek, USA). DNA purity and concentration were measured using a NanoDrop2000 (Thermo Fisher Scientific, USA) and Quantus Fluorometer (Promega Biotech Co., Ltd., China). DNA integrity was assessed by agarose gel electrophoresis. After the DNA was fragmented to 350 bp using a Covaris M220 (Covaris, USA), a NEXTFLEX Rapid DNA-Seq Kit (Bioo Scientific, USA) was used to construct the PE library. After bridge PCR amplification, the DNA was sequenced using the Illumina NovaSeq/Hiseq Xten (Illumina, USA) sequencing platform.
Filtering and quality control were performed for the processed data. Fastp (https://github.com/OpenGene/fastp, version 0.20.0) was used to control the data quality, and the low-quality and presence of N bases reads in the data were cut out accordingly69, that we obtained approximately 109.1 Gb of metagenomic clean reads, with data ranging from 7.4 to 10.7 Gb per sample (Supplementary Table 5). Megahit was applied to assemble the optimized sequence (https://github.com/voutcn/megahit, version 1.1.2)70. ORF prediction for contigs in the assembled results was performed using Prodigal (https://github.com/hyattpd/Prodigal, version 2.6.3)71 and genes with nucleotide lengths greater than or equal to 100 bp were selected and translated into amino acid sequences. The predicted gene sequences from all samples were clustered using the CD-HIT (http://www.bioinformatics.org/cd-hit/, version 4.6.1) software to create non-redundant gene sets, with 90% sequence identity and 90% coverage72. The high-quality reads from each sample were aligned against the non-redundant gene set separately using the SOAPaligner (https://github.com/ShujiaHuang/SOAPaligner, version 2.21) software73. The abundance of genes in the corresponding samples was calculated.
The non-redundant gene set was compared with the non-redundant (NR) database using BLASTP (version 2.2.28+) to obtain microbial community annotations through the corresponding classification information database in the NR library (all e-values were set to 10−5 as the cut-off threshold). Functional genes were aligned against the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. Pathogen annotations were obtained by aligning the target amino acid sequences against the Pathogen Host Interactions Database (http://www.phi-base.org/index.jsp). ARGs were aligned against the Comprehensive Antibiotic Resistance Database (http://arpcard.Mcmaster.ca) with “Strict” alignment parameters.
Statistical analysis
Data processing was performed using Microsoft 365. R software (version 4.3.2) was used for differential calculations, fitting, data visualization, and other analyses. Most of the metagenomic analyses were performed using the Majorbio Cloud platform (https://edu.majorbio.com/). Co-occurrence networks were generated using R and visualized using Gephi (version 0.9.2).
Responses