Cold and hot potable water samples were obtained from four different multifamily buildings in different geographical regions (A: Kiel, B: Hamburg, C: Hameln, and D: Munich) in Germany. All buildings contained central hot water circulation systems and were equipped with an ultrafiltration module, directly after the entry point of the building (point of entry, POE; existing buildings A, B, and newly constructed building D) or in the hot water circulation system (building C), resulting in reduced cell concentrations due to removal of bacteria. Cold-water sampling locations comprised: point of entry (POE), water supplying the water heater (water heater supply, WHS) and distal taps in the periphery (point of use, POU_Cold). Hot water samples were taken from the central hot water circulation after leaving the hot water tank (water heater exit, WHE), from the hot water circulation (directly before the ultrafiltration unit in building C), from hot water reflow into the heater (water heater return, WHR) and from the taps in the periphery (point of use, POU_Hot). Sampling sites were heat-disinfected with a flame prior to sampling following standard procedures. The first liter served for DNA extraction and subsequent sequencing (in case of selected samples) and the second liter was collected to quantify bacterial concentrations and regrowth potentials. Samples were typically collected in 0.1 L, 0.5 L, or 1 L borosilicate glass bottles with Teflon seals, which were free of assimilable organic carbon (AOC). Samples were sent cold via express shipment to the laboratory and processed within 24 h after sampling. Prior to flow cytometric analysis, samples were aliquoted in three 50 ml AOC-free glass vessels holding 20 ml of sample each. Concentrations of total and intact cells were quantified by flow cytometry on the day of arrival (day 0) to determine the microbiological status at the time point of sampling. Three aliquots of each sample were subsequently incubated either at 22°C, 36°C, or 50°C and flow cytometric analysis was repeated after 7, 14, and 21 days or as indicated.

An alkaline solution of potassium permanganate was made by separately dissolving (A) 30 g of permanganate (cat.nr. 105082; Merck KGaA, Germany) and (B) 100 g sodium hydroxide pellets (cat.nr. 106498; Merck KGaA, Germany) in 0.5 L deionized water each and equilibrated to room temperature. The two solutions were subsequently combined and stored in a 1 L bottle in the dark prior to use. Clean glass vials (50 ml, cat.nr. 7612150; Th. Geyer GmbH & Co.KG) were first washed with deionized water and then filled with this alkaline solution and sealed with appropriate caps (cat. nr. 292401305, DWK Life Sciences, Wertheim, Germany). The vials were filled to enable contact of the entire surface with the oxidant. The permanganate solution was decanted (for recycling purposes) and glassware and caps were rinsed three times with tap water and twice with deionized water. Dried glassware and caps were wrapped in aluminum foil. Glassware was muffled at 550°C (50 ml vials) for at least 8 h or overnight, caps were heated at 180°C during that time. Caps were provided with a Teflon seal (special fabrication by Max Werth GmbH Co. KG, Mülheim an der Ruhr, Germany) prior to use.

Bacterial cell numbers were quantified using flow cytometry. For this purpose cells were stained with the fluorescent dyes SYBR Green I (10,000× stock; cat.nr. S-7567; Life Technologies Ltd.) and propidium iodide (PI; 1 mg ml⁻¹, cat. P3566; Life Technologies Ltd.). SYBR Green I was diluted to a working stock concentration of 100× using dimethylsulfoxide (DMSO; Sigma Aldrich) and stored at −20°C until use. Aliquots (250 μl) of water samples were transferred into a 96-well plates (cat. nr. 601808, HJ-Bioanalytik GmbH). To determine total cell concentrations (TCC), 200 μl sample aliquots from this plate were transferred into the wells of a second 96-well plate with pre-aliquoted 2 μl of the 100× SG working stock solution followed by thorough mixing by pipetting up and down several times with a multichannel pipette. To determine intact cell concentrations, 200 μl sample aliquots from this plate were transferred into the wells of a second 96-well plate with pre-aliquoted 2.4 μl of a mixture of 100× SG I (2 μl) and PI (0.4 μl) followed by thorough mixing through repeated aspiration. Staining was performed at 37°C for 13 min in an air incubator. Data was collected using Novocyte ACEA benchtop instrument equipped with a 488 nm laser (ACEA Biosciences Inc., USA) and an autosampler to enable analysis on a 96-well plate basis. Data were analyzed using the instrument-specific software and a gating procedure similar to the one described by Gatza et al. (2013). In case that the total signals exceeded ~1.000 signals/s water samples were diluted with 0.1 μm filtered mineral water (Evian, Evian-les-Bains, France) to ensure osmotic balance.

Water samples contained in AOC-free vials were incubated at 22°C, 36°C, or 50°C (±1°C) for up to 3 weeks without shaking. At indicated time intervals cell numbers were repeatedly quantified by flow cytometry. In experiments addressing whether the regrowth characteristics could be transplanted from one sample to another, hot water samples (20 ml) from building A were supplemented with a 1/1,000th volume (20 μl) of cold water (POE) from the same building. These spiked samples together with non-spiked control samples were incubated at 22°C for 3 weeks.

For experiments addressing potential changes of the microbiome during incubation at elevated temperatures (50°C, 55°C, 60°C), hot water samples from building C were collected in AOC-free 1 L bottles. At the time point of sampling, the temperature of the central water heater was set to 45°C−50°C and the actual temperature of the sampled water was 45.5°C. Samples for direct filtration were cooled on ice and transported to the laboratory. Samples for regrowth experiments and subsequent sequencing were not placed on ice, but transported at their given temperature in insulated transport boxes to minimize cooling and potential growth inhibition. The samples were incubated in the laboratory at the indicated temperatures without shaking for 3 weeks using standard laboratory incubators.

For microbial community analysis, 1 L of each water sample was filtered through 0.2 μm polycarbonate filters (Pieper Filter GmbH, Germany). Genomic DNA was extracted using the Aquadien™ DNA Extraction and Purification Kit (cat. nr. 3578121, Bio-Rad Laboratories GmbH, Feldkirchen, Germany). Illumina MiSeq sequencing was performed by Eurofins Genomics (Ebersberg, Germany) after amplifying the V3/4 region of the 16S rRNA gene (467 bp ± 20 bp) using universal primers (forward: 5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG-3′ and reverse: 5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3′). The data presented in the study are deposited in the NCBI BioSample database repository, accession numbers SAMN35749896–SAMN35749922.

Paired-end reads were initially processed using cutadapt (version 2.7, Martin, 2011) by the sequencing company Eurofins Genomics (Ebersberg, Germany) to remove primer and adapter sequences. DADA2 (version 1.29.0, Callahan et al., 2016) was subsequently employed to obtain an ASV (amplicon sequence variants) abundance matrix including taxonomical classifications. Firstly, forward and reverse reads were filtered and truncated at lengths 280 bp and 200 bp, respectively using the DADA2 filterAndTrim function. Filtered reads were further dereplicated (derepFastq) before error rate estimation subsampling 100 Mbp for training (learnErrors). Finally, the DADA algorithm was applied to infer ASVs and read pairs were merged using a minimum overlap of 10 bp under the default parameters (DADA2 functions dada, mergePairs) and bimeric sequences were removed via function removeBimeraDenovo. Taxonomy was assigned to ASVs using the SILVA database version 138.1 clustered at 99% similarity (https://www.arb-silva.de/documentation/release-1381/) with assignTaxonomy. An overview of the generated data, e.g., the total raw reads, their numbers across DADA2 workflow steps, and final ASVs are given in Supplementary Table S3.

Metadata, taxonomic classifications and abundance matrix were integrated into a phyloseq object and analyzed using R packages phyloseq (version 1.42.0, McMurdie and Holmes, 2013), tidyverse (version 2.0.0, Wickham et al., 2019) and ggplot2 (version 3.4.2, Wickham, 2016) within R (version 4.1.2, https://intro2r.com/citing-r.html). The phyloseq object was firstly filtered to include only those ASVs affiliated to the kingdom “Bacteria” and following calculation of relative abundances, ASVs representing <1% of each sample were removed. A Principal Components Analysis (PCoA) of Bray-Curtis dissimilarities and Shannon–Wiener (Shannon, 1948), Simpson (Simpson, 1949) and Chao 1 (Chao, 1987) indices were created making use of the phyloseq and vegan R packages (https://cran.r-project.org/web/packages/vegan/index.html). Barplots and heatmaps of class- and genus-level abundances were plotted using ggplot2 following taxonomic rank-specific summarizing of the filtered phyloseq object using tidyverse functions (https://www.tidyverse.org/).

Drinking water samples from four buildings (A, B, C, and D) with hot water circulation systems were taken at different locations of the corresponding DWDS. Samples included cold water from the point of entry (POE), cold water entering the water heater (water heater supply, WHS), hot water at the water heater exit (WHE), hot water from the circulation pipe, hot water re-entering the water heater (water heater return, WHR) as well as cold and hot water at the point of use (POU) in the periphery. Regrowth assays were performed by incubating samples either at 22°C, 36°C, or 50°C. Intact cell concentrations (ICC) were monitored by flow cytometry in weekly intervals over 3 weeks.

For buildings A, B and C, cold-water samples from POE and WHS sampling points tended to show strongest bacterial regrowth when samples were incubated at 22°C and in part also at 36°C (Figures 1–3). No regrowth was observed when incubating cold-water samples at 50°C. Hot water from WHE and WHR sampling points showed the opposite behavior with regrowth being strongest when incubating samples at 50 and 36°C, while no regrowth was observed at 22°C. The most diverse regrowth behaviors were measured for POU samples from distal taps which were used to retrieve both cold and hot water. Regrowth for those samples was building-specific. In buildings A and C, hot water POU samples exhibited regrowth at all three incubation temperatures, whereas regrowth of cold-water POU samples was limited to 36 and 22°C (Figures 1, 3). In building B, all POU samples showed regrowth at all three temperatures (Figure 2).

Water samples from building D represented an exception as no regrowth was obtained at 50°C (Figure 4). Neither hot water samples from sampling points WHE nor from WHR showed regrowth at this elevated temperature. The already low concentration of intact cells initially present in the first measurement (representing the time point of sampling) declined within seven days to values near the detection limit of ~200 cells/ml. At the time point of sampling the domestic plumbing system of the building had been in use for only 11 months including a weekly thermal disinfection at 60°C for 12 h. These conditions were seen as the reason for the lack of regrowth at 50°C. Regrowth was obtained on the other hand at 36 and 22°C (Figure 4) although cell concentrations were substantially lower than the ones in comparable samples from buildings A–C (Figures 1–3).

As water samples from different locations had different permissive temperatures for bacterial regrowth, we hypothesized that the temperature dependency could be altered by spiking water of a specific regrowth characteristics with a small volume of water of a different regrowth characteristic. For that purpose, a hot water sample from the water heater exit (WHE) of building A was supplemented with a 1/1,000 v/v ratio of cold water from the point of entry. While water from sampling point WHE displayed no significant regrowth at 22°C for up to 3 weeks, the spiking with cold water resulted in measurable regrowth beginning at day 7 (Figure 5). Cell numbers continued to increase when extending incubation up to 3 weeks.

Regrowth at 22°C was also achieved for hot water sampled at location WHR (Figure 5) when supplemented with cold water from POE whereas no significant regrowth was measured for the unsupplemented WHR hot water sample. Overall data suggested that the regrowth characteristics of the cold-water sample (and the bacteria contained herein) could be “transplanted” to the hot water sample. This was not the case when spiking hot water sampled from the POU_Hot tap in the periphery. This water showed per se regrowth at 22°C and the addition of cold water did not change the regrowth characteristics.

The different regrowth characteristics of the drinking water samples were likely due to different bacterial communities present at different locations in the DWDS. In order to gain more insight into the associated microbiomes, selected samples were sequenced. Using the example of building C, PCoA analysis discriminated dissimilarities in the structure of bacterial communities in cold water and hot water (Figure 6). For the cold-water samples, the bacterial community at the point of entry was distinct from those obtained from taps. Also the hot water communities were not uniform with pronounced divergence between the bacterial communities in the central hot water circulation system (WHE, the circulation pipe and WHR) and those obtained from distal taps.

Amplicon sequence variants (ASVs) recovered from samples collected at building C were assigned to a total of 25 taxonomic classes. Gammaproteobacteria comprised on average the largest fraction of all samples (Figures 7A, B). Their relative abundance was hereby higher in hot water samples (average: 76%) than in cold-water samples (average: 54%), with relative proportions in hot water ranging between 61 and 86% (Supplementary Figure S1). Gammaproteobacteria showed the lowest relative abundance (22%) in cold water at the point of entry, whereas POU_Cold samples contained proportions ranging from 46% to 78%.

Other bacterial classes with higher abundances in hot water included Deinococci (2.88%−16.3%), Kryptonia (0.85%−8.28%), Nitrospiria (1.57%−5.89%), and to a minor extent Gemmatimonadetes (0.37%−3.02%) and Ignavibacteria (0.32%−2.25%; Supplementary Figure S1). Within the hot water samples, Deinococci, Kryptonia, and Nitrospiria had highest relative abundances in the central circulation system (WHE, WHR) and somewhat lesser abundances in the POU_Hot samples. The most abundant genera among the Gammaproteobacteria in hot water samples comprised unclassified Hydrogenophilaceae, unclassified Comamonadaceae, Extensimonas, Methyloversatilis, and MND1 (Figure 8, Supplementary Figure S2, Supplementary Table S4). Whereas the genera Hydrogenophilaceae and Methyloversatilis were dominant in the hot water circulation system (WHE, WHR, and the circulation pipe), unclassified Comamonadaceae and Extensimonas were dominant in the POU_Hot samples (Figure 8, Supplementary Figure S2, Supplementary Table S4).

Cold-water samples on the other hand contained apart from Gammaproteobacteria high relative abundances of Alphaproteobacteria, especially in POU_Cold samples with relative proportions between 5 and 34% (Supplementary Figure S1). Cold water from POE in comparison contained <2% of Alphaproteobacteria. POU_Cold samples also showed higher relative abundances of Bacteroidia (up to 12%) than POE water (<1%). Among the Gammaproteobacteria, the genus Acidovorax (belonging to the Comamonadaceae) was dominant in POU_Cold samples, whereas this genus was present with only 0.2% in cold water at the POE (Figure 8, Supplementary Figure S2, Supplementary Table S4).

Differences between cold and hot water samples were also reflected in the diversity indices with higher Simpson and especially higher Shannon–Wiener indices for cold water samples (Figure 7C). The diversity indices were highest for cold POE water (Figure 7C).

As the described bacterial communities reflected the microbiological status of the time point directly after sampling, we addressed the question how stagnation would affect the microbiome of hot water. For that purpose hot water from building C (initial temperature during sampling: 45.5°C) was either processed directly or was incubated for 3 weeks at 50°C, 55°C, or 60°C. Cold water processed directly after sampling served as a reference.

Independent of stagnation, all hot water samples showed a different bacterial community composition and structure and much less diversity than the cold water (Simpson index: 1, Shannon–Wiener index: 6.6; Figure 9). As seen before, Gammaproteobacteria were the dominant bacterial class in hot water samples (Figure 9). Stagnation at elevated temperatures resulted in a strong increase of this bacterial class and led to a pronounced difference to the directly processed hot water. Other abundant bacterial classes comprised as before Kryptonia and Deinococci and to a minor extent unclassified MBNT15 and Gemmatimonadetes (Supplementary Figure S3). The abundance of Ignavibacteria that were initially present in the hot water samples at the time point of sampling with ~7%, decreased during the 3-week incubation in a temperature-depended fashion to 1.28%, 0,75%, and <0.5% at 50°C, 55°C, and 60°C, respectively.

A bacterial class that on the other hand only became abundant during stagnation was the one of Acidimicrobia. Stagnation at 50 and 55°C resulted in proportions up to 12%, whereas this class was hardly detected after stagnation at 60°C.

Among the Gammaproteobacteria, the genera MND1 and unclassified Hydrogenophilaceae were dominant in hot water samples (Supplementary Figure S4). Whereas the relative abundance of MND1 tended to strongly increase during stagnation, the relative abundance of Hydrogenophilaceae varied between samples, with a tendency to higher abundances at higher stagnation temperatures. Unclassified OPS 17 (belonging to the Bacteroidia) showed only strong presence (~29%−32%) in directly processed hot water, their relative abundance strongly declined during stagnation.

The incubation temperature caused differences in the bacterial community compositions. Whereas 50 and 55°C seemed permissive for Gemmatimonadetes, Ignavibacteria, and Acidimicrobia, these bacterial classes showed much lower relative abundances after incubation at 60°C (Supplementary Figure S3). The relative abundance of Kryptonia, on the other hand, seemed to increase when rising the incubation temperature. On genus level, unclassified Actinomarinales were only found in samples incubated at 50 and 55°C, but not after incubation at 60°C (Supplementary Figure S4).

In summary, stagnation of hot water for 3 weeks resulted in pronounced differences in the bacterial community compositions. On class level, Gammaproteobacteria strongly increased in relative abundance during stagnation, whereas the relative proportions of Bacteroidia and Ignavibacteria strongly decreased. On genus level, NMD1 Gammaproteobacteria and Acidimicrobia strongly increased during stagnation, whereas the relative abundance of unclassified OPS 17 (Bacteroidia) strongly decreased. Different incubation temperatures thus shaped the bacterial community compositions in different ways.