2.1 Bulk water quality is affected by temperature and source variations
A drinking water biofilm was grown for 137 days using a pilot-scale distribution system with 3 identical loops operating at 16°C, 20°C, and 24°C (Fig. 1A). The pilot infrastructure is located at the CAPTURE building in Ghent, Flanders. This building is fed with water from the Farys network. As Farys distributes water from both ground- and surface water, CAPTURE is fed with water from alternating sources (Supplementary Fig. 1). This resulted in the fact that the drinking water fed to the pilot alternated between treated ground- and surface water. Based on the records of Farys and discerning variations in conductivity between treated groundwater and surface water (i.e., > 600 µS/cm, < 600 µS/cm, respectively), we could determine when the pilot was filled with each type of drinking water (Supplementary Fig. 3A). The experiment started with the introduction of treated surface water and changed to treated groundwater by day 7. Subsequently, after 77 days, the water source reverted to treated surface water, which it remained until day 129. Between days 130 and 133, the water type shifted back to treated groundwater, before concluding with treated surface water in the final days of the experiment.
Throughout the experiment, the bulk water quality, including cell densities and community composition, alongside physicochemical content, was followed. Physicochemical parameters (conductivity, pH, pressure, flow, nitrate-nitrogen (NO3-N), nitrite-nitrogen (NO2-N), total nitrogen, orthophosphate (P2O5), iron (Fe), non-purgeable organic carbon (NPOC) were similar across the loops (Supplementary Fig. 2, 3, 4). Temperature had no significant effect on iron (Fe), orthophosphate (P2O5) and nitrogen concentrations (NO3-N, NO2-N, total nitrogen), however, statistical analysis revealed a slight significant effect of temperature on NPOC content (p = 0.01, Kruskal-Wallis test). Additionally, every 8 hours total cell counts were measured using online flow cytometry (FC) (Fig. 1B, Fig. 2A). The average cell density through the experiment was (1.87 ± 0.66) × 105, (3.12 ± 1.16) × 105 and (3.39 ± 1.27) × 105 cells/mL for loop 1 (16°C), 2 (20°C) and 3 (24°C), respectively, which correspond to similar concentrations in literature24,38,39. The temperature exhibited a significant impact on bacterial bulk cell densities (p = 2.20 × 10− 16, Kruskal-Wallis test), with elevated temperatures resulting in higher cell densities (Fig. 2A). Previous literature has reported higher heterotrophic plate counts (HPC) and bacterial cell concentrations in summer (> 18°C) compared to winter (< 11°C)40–43. Francisque et al.40 demonstrated that the augmented HPC levels in warmer seasons are attributed to increased organic matter content, thereby fostering bacterial proliferation. Similarly, Prest et al.42 observed higher bacterial counts during summer despite lower organic carbon content at the treatment plant, suggesting that not only organic carbon is a growth-controlling factor within this distribution system. Contrary to expectations, some studies have indicated that seasonal variations do not consistently promote bacterial growth, nor do they consistently correlate with organic carbon concentrations during summer25,44. This indicates that the effect of increasing temperatures during distribution depend on other factors as well, such as water quality (e.g., disinfectant residual, available nutrients) and hydraulics, which will be further discussed5,42,45. In our research, we observed a positive association between elevated temperatures and bacterial abundances, while a negative correlation with NPOC levels was observed. This observed negative correlation may be attributed to increased oxidation of organic matter at elevated temperatures46. Notably, it is important to highlight that the water supplied to the pilot, with a refreshment of once a week, was similar for alle loops, and in the pilot the water was heated from 10°C – 16°C to 16°C, 20°C or 24°C, effectively mitigating the influence of temperature variation in the water source. By increasing the water temperature of the feed water, specific microbial groups with higher optimal growth temperatures were selectively favored8. Furthermore, the recirculation of water over a 7-day period resulted in an increase in water age, a factor known to amplify the impact of temperature fluctuations4,9,41.
As mentioned before, the DWDS pilot at CAPTURE was fed with alternating water types (i.e., treated ground- and surface water), this change had a significant effect on orthophosphate as well as NPOC content (p = 4.99 × 10− 5, p = 2.28 × 10− 11, Kruskal-Wallis test) (Supplementary Fig. 4B, 4C). A higher orthophosphate (71.55 ± 38.40 µg P2O5/L compared to 11.04 ± 17.84 µg/L) and NPOC concentration (1.81 ± 0.48 mg/L compared to 1.22 ± 0.33 mg/L) was measured when treated surface water was fed instead of treated groundwater, consistent with findings from earlier studies15,36,47. Furthermore, the water source variations had a statistically notable influence on cell densities of loop 2 (20°C) and 3 (24°C) (p = 2.71 × 10− 11, p = 2.64 × 10− 7, Mann-Whitney test), but no significant influence on cell densities of loop 1 (16°C) (p = 0.9944, Mann-Whitney test). For loop 3 (24°C), this resulted in an average of (3.94 ± 1.41) × 106 cells/mL for drinking water produced from groundwater, while drinking water produced from surface water contained (2.89 ± 0.88) × 106 cells/mL. As mentioned before, previous studies highlighted that the impact of temperature depends on the water quality5,42,45. Our study confirmed the impact of temperature on bacterial bulk cell concentrations. However, we emphasized the crucial role of the water source, indicating a combined effect of environmental factors like substrate composition and availability with temperature on specific microbial populations. Our results highlight temperature's importance alongside water quality in shaping bacterial growth and survival.
Growth curves were fitted based on the FC results, and growth rates, along with carrying capacities were calculated, based on the method of Candry et al.48 (Fig. 2B, 2C, 2D, Supplementary Fig. 5). The carrying capacity, which represents the maximum bacterial concentration sustained by nutrients and environmental conditions, is subject to influence from factors such as nutrient availability and temperature. Higher temperatures in combination with treated groundwater resulted in significant higher growth rates (p = 0.0041, Mann-Whitney test), more specifically the median growth rate for treated surface water and groundwater was 0.0069 h− 1 and 0.017 h− 1, respectively. The measured growth rates align in the same order of magnitude as growth rates from the resident drinking water community as reported in previous studies (i.e., ± 0.007 h− 1, ± 0.075 h− 1)8,29,30. In addition, we observed that treated groundwater at elevated temperatures (20°C, 24°C) led to a statistically noteworthy increase in carrying capacities (ploop2 = 0.0045, ploop3 = 0.027, Mann-Whitney test). For instance, the median carrying capacity of loop 2 (20°C) for surface water was determined to be 3.03 × 105 cells/mL, while the median carrying capacity for groundwater was measured at 4.59 × 105 cells/mL. Producing biostable water aims for bacterial densities near the carrying capacity, reducing net growth16. Our results indicate that the surface water community is approaching its carrying capacity, suggesting that higher temperatures in this context did not result in increased growth rates or carrying capacities, suggesting more biostable water. In the case of treated groundwater, we observed that the environmental factor, temperature in this case, led to increased growth rates and carrying capacities, indicating the presence of available niches and nutrients for growth. This increase can also be attributed to more biofilm detachment due to lower concentrations of orthophosphate and organic carbon, as phosphate and organic carbon are growth limiting nutrients49–52. Moreover, we need to be careful as in our study we refreshed the system twice a week, leading to a residence time of 7 days, which is quite extreme as 95% of the population receives its drinking water maximum 5 days after it has been distributed9. Also, water was recirculated leading to an increase in water age and the two water types were mixed in our system. This mixing and the worst-case recirculation time might have altered the carrying capacities.
In addition, analysis of the cytometric data through fingerprinting revealed significant differences of both temperature and water source on microbial phenotypic traits (p = 9.99 × 10− 4, p = 9.99 × 10− 4, PERMANOVA) (Fig. 3A)53. In a study by Favere et al., FC fingerprinting also demonstrated the ability to distinguish between treated groundwater and surface water within a water tower15. Throughout the experiment, the 16S sequencing results revealed that the dominant phyla observed in the bulk water community were Proteobacteria, more specifically the families Comamonadaceae and Sphingomonadaceae (Fig. 3B), which are often found in chlorinated drinking water28,36,37,45,54. This community composition was not significantly influenced by temperature (p = 0.321, PERMANOVA), although some families (e.g., NS11-12 marine group) were observed in higher abundances at increased temperatures. Additionally, the presence of specific genera within the Sphingomonadaceae family, such as Novosphingobium, was notably prominent at lower temperatures, while others, such as Sphingopyxis, exhibited predominance at higher temperatures (Supplementary Fig. 9). On the other hand, a significant effect of the source water on composition of the bulk water community was detected (p = 9.99 × 10− 4, PERMANOVA) (Fig. 3B). A higher relative abundance of the Chitinophagaceae family, more specifically Sediminibacterium spp., was observed when treated surface water was fed to the pilot. Sediminibacterium spp. are commonly found in treated surface water55. These results suggest that the water communities during distribution are more influenced by the initial water community composition fed to the DWDS and are less likely to be modified by other factors, such as temperature25,37,42. To conclude, our results showed that the water type is mainly shaping the bulk community composition in the DWDS, whereas elevated temperature in combination with treated groundwater can lead to increased growth rates and carrying capacities.
2.2 Mature drinking biofilms are not influenced by increasing temperatures
To sample and analyze the biofilm cell density and community composition, a coupon system, consisting of removable inserts made out of the same material as the pipes, was implemented in the drinking water distribution pilot (Fig. 1C). An increase in biofilm cell density (± 1 log10 cells/cm²) and a change in the community composition were observed over time, which was expected as the distribution pilot was not used before (Fig. 4). Temperature did not significantly influence biofilm cell density and community composition (p = 0.727, Kruskal-Wallis test, p = 0.667, PERMANOVA). Similar findings were observed in a long-term study (232 days) performed by Ahmad et al.25 where temperatures up to 30°C did not lead to higher biomass density and differences in community diversity. On the other hand, water source variations had a significant influence on cell densities of loop 2 (20°C) and 3 (24°C) (p = 4.08 × 10− 2, p = 4.88 × 10− 3, Kruskal-Wallis test), but no significant influence on cell densities of loop 1 (16°C) (p = 0.159, Kruskal-Wallis test), which was similar to the results for the bulk water cell densities. It is important to note that this increase may also be attributed to a time effect, as the biofilm was not mature yet (discussed further). Regarding the biofilm community composition, no significant influence of source was revealed (p = 0.242, p = 0.133, p = 0.219 for each loop respectively, PERMANOVA).
In addition, consistent cell concentrations (i.e., (3.99 ± 0.64) × 106 cells/cm²) and community composition were observed from day 70 onwards, indicating a mature state of the drinking water biofilm24,33,33. In a study by Boe-Hansen et al.56, this phase is referred to as a quasi-stationary state, characterized by stable biofilm cell counts, EPS formation and maintaining an equilibrium between growth, attachment and detachment. The microbial diversity of mature biofilms is primarily influenced by the quality and composition of the feed water and environmental conditions (such as temperature) only play a significant role during early primary colonization and growth of microbes, which confirms the findings of this research25,37,57. For instance, several studies have demonstrated that switching water sources can impact the biofilm community and its diversity. This impact occurs because new genera are introduced, occupying specific niches provided by the biofilms13,14,58,59. Following such a switch, the biofilm community typically undergoes restoration to a new stable state within a month13,58,59. In our case, we followed biofilm development at different temperatures, while drinking water fed to the pilot alternated between treated ground- and surface water (both chlorinated). A higher biofilm cell density was observed when treated surface water was fed to the loops operating at higher temperatures, however, no significant influence on community composition was revealed (Fig. 4). The influence of exchange of taxa can stay limited as it is hypothesized that the low-abundance bacteria from the bulk water function as a seed bank to the mature biofilms24. Furthermore, to assess the impact of water source switching on biofilm communities and densities over time, it is recommended that the biofilm have already reached a stable state and a comprehensive approach involving increased sampling frequency and additional timepoints is needed. In summary, our results indicate that water type primarily influences the composition of the biofilm community. Elevated temperatures did not result in increased biofilm cell densities. However, elevated temperatures combined with a change in water type can lead to increased cell densities while maintaining similar core biofilm microbes.
2.3 The core biofilm microbiome and bulk-biofilm interactions
To elucidate bulk-biofilm interactions during the experiment, a non-metric multidimensional scaling (NMDS) was constructed based on the relative bacterial community composition of the bulk and biofilm data (Fig. 5, Supplementary Table 2). First, K-means clustering was performed and a corresponding average distance was determined, showing higher similarities between mature biofilm samples (0.50) compared to the bulk samples (0.88). We observed a distinct community composition of the mature biofilm samples compared to the bulk water samples, however, there are unique amplicon sequence variants (ASVs) (e.g., Rhizobacter spp. (ASV4) Methyloversatilis spp. (ASV6) Hydrogenophaga spp. (ASV18)) that are found in both groups, suggesting that water is seeding the biofilm and that biofilm cells are dispersed into the bulk water24,29. The abundance of these species is lower in the bulk water (0.69 ± 0.52%, 0.15 ± 0.16%, 2.56 ± 5.13% for ASV 4, 8, 18, respectively) compared to the abundance in the biofilm samples (7.39 ± 5.95%, 4.04 ± 3.14%, 0.6 ± 0.45% for ASV 4, 8, 18, respectively). This validates previous findings which indicate that low-abundance bacteria from the bulk water serve as a seed for the biofilm24. Next to genera from the Comamonadaceae family, the bulk samples primarily consisted of Sediminibacterium spp. (ASV1) from the Chitinophagaceae family, with an average abundance of 31.93 ± 22.29% across the samples. Besides, a higher community diversity was obtained compared to the bulk water samples (Supplementary Fig. 8), which is different from other studies where Shannon diversities where slightly higher for bulk water samples compared to biofilm samples25,60. This observation can be attributed to the recirculation mode, which could result in growth of high-abundance groups in the bulk water.
Furthermore, a core biofilm microbiome (unique ASVs present in all mature biofilm samples) was characterized, predominantly comprising Rhodocyclaceae, specifically Zoogloea spp. (ASV3), Methyloversatilis spp. (ASV6), and Zoogloea spp. (ASV14), with average abundances of 13.11 ± 17.40%, 4.04 ± 3.14%, and 2.33 ± 2.99% in each sample, respectively. Additionally, Comamonadaceae, particularly Rhizobacter spp. (ASV4) and Hydrogenophaga spp. (ASV18), exhibited average abundances of 7.39 ± 5.95% and 0.6 ± 0.45%, respectively and Xanthobacteraceae, specifically Xanthobacter autotrophicus (ASV13) and Bradyrhizobium spp. (ASV23), were present with average abundances of 1.97 ± 1.43% and 1.62 ± 1.73%, respectively. Finally, Sphingomonadaceae, including Plot4-2H12 spp. (ASV20) and Sphingomonas spp. (ASV21), exhibited average abundances of 0.85 ± 0.53% and 2.20 ± 3.02% in each sample, respectively. Xanthobacer autotrophicus (ASV13) and Caulobacter spp. (ASV19) were also present in both bulk and biofilm samples from day 1 and 14, with average abundances of 1.92 ± 1.84% and 3.21 ± 4.65% in each sample, respectively. Similar bacterial groups were observed as core members of biofilms resulting from chlorinated as well as from non-chlorinated treated surface water24,25,36,61. They are all known to degrade a wide range of carbon sources62. Xanthobacteraceae species are known for their ability to fix nitrogen, Zoogloea species are typical floc formers, and members of the Sphingomonadacea known to form biofilms and produce EPS within DWDS25,57,60,62,63. This bacterial core community was not affected by increasing temperatures and densities of the bulk water, indicating a high resilience of this biofilm community during distribution.
In conclusion, increasing temperatures had an influence on the bulk cell density, but not on biofilm development and cell densities. Elevated temperatures in combination with treated groundwater resulted in increased growth rates and carrying capacities of the bulk water. Water source variations had an influence on bulk and biofilm cell densities for the loops operating at 20°C and 24°C, however, it is important to note that the biofilm was still developing and was defined mature from day 70. Increasing temperatures and water source variations did not change the biofilm community composition, whereas the bulk community was mainly shaped and influenced by the drinking water community fed to the DWDS pilot. A combined NMDS plot showed higher similarity across the mature biofilm samples compared to the bulk water samples, indicating bigger influences of temperature and water type on the bulk water quality. A core biofilm microbiome was identified, dominated by Alpharoteobacteria, more specifically, Rhodocyclaceae (ASV3, ASV6, ASV14), Xanthobacteraceae (ASV13, ASV23), and Sphingomonadaceae (ASV20, ASV21), and Betaproteobacteria, more specifically Comamonadaceae (ASV4, ASV18), with average abundances ranging from 0.6–13.11% across the samples. The bulk water was primarily characterized by Chitinophagaceae (ASV1) and Comamonadaceae (ASV4, 5, 18), with average abundances varying from 0.15–31%. This holistic approach of investigating the drinking water microbiome on a unique drinking water distribution pilot offers a comprehensive understanding of microbial responses to changing environmental conditions, which is crucial for predicting and managing microbial community behavior in diverse ecosystems. In practice, when a new pipe will be installed in the DWDS, colonization will not be more pronounced when temperatures are higher, for example when climate change causes further increase soil temperatures. Once a mature biofilm is formed, its composition remains stable and unaffected by changing water temperatures and source water quality. However, when source water quality changes because of climate change, the bulk water community and density will be affected, possibly resulting in biological instability and potential (unwanted) quality changes at the customer’s tap after the DWDS. This study demonstrates how microbial ecology can contribute to the understanding of microbial dynamics during distribution of drinking water, and can help the drinking water sector to meet the Sustainable Development Goal 6: access to safe water.