Towards a probiotic approach for building plumbing – nutrient-based selection during initial biofilm formation on flexible polymeric materials

Upon entering building plumbing systems, drinking water bacteria experience considerable changes in environmental conditions. For example, some flexible polymeric materials leach organic carbon, which increases bacterial growth and reduces diversity. Here we show that the carbon supply by a flexible polymeric material drives nutrient-based selection within establishing biofilm communities. We found that migrating carbon from EPDM coupons resulted in considerable growth for different drinking water communities (0.2 – 3.3 × 108 cells/cm2). All established biofilm communities showed low diversity (29 – 50 taxa/biofilm), with communities dominated by even viewer taxa (e.g., 5 taxa accounting for 94 ± 5 % relative abundance, n = 15). Interestingly, biofilm communities shared some taxa (e.g., Methylobacterium spp.) and families (e.g., Comamonadaceae), despite the difference in starting communities. Moreover, selected biofilm communities performed better than their original communities regarding maximum attachment (91 ± 5 vs. 69 ± 23 %, n = 15) and attachment rate (5.0 ± 1.7 × 104 vs. 2.4 ± 1.2 × 104 cells/cm2/h, n = 15) when exposed to new EPDM coupons. Our results demonstrate nutrient-based selection during initial biofilm formation on a flexible polymeric material and a resulting benefit to selected communities. We anticipate our findings to help connecting observational microbiological findings with their underlying ecological principles. Regarding initial biofilm formation, attachment dynamics, growth, and selection thereof are important for the management of microbial communities. In fact, managing initial colonization by supplying specific carbon and/or introducing consciously chosen/designed communities potentially paves the way for a probiotic approach for building plumbing materials.


Introduction 51 52
Uncontrolled microbial growth in building plumbing systems is generally undesirable 53 as it can lead to operational and/or hygienic problems 1,2 . Such growth is caused by 54 changes in environmental conditions, which is what drinking water bacteria experience 55 as soon as they enter a building plumbing system. For example, water temperature 56 increases and fluctuates spatially and temporally, which was shown to alter community 57 composition 3,4 . Also, pipe diameters are considerably smaller (e.g., < 2 cm) compared 58 to main distribution pipes (e.g., ≥ 10 cm), which provides more surface area per water 59 volume 5 , and increases the impact of biofilms on the water phase. Regarding 60 operation, flow pattern and rates have been shown to impact biofilm structure and 61 community composition 6,7 . Finally, diverse materials are used for pipes and non-pipe 62 components 8 , and some of these support microbial growth by leaching biodegradable 63 substances 9 , which is especially critical under long stagnation times of the water 10 . 64 The bottom line is that building plumbing systems often provide more favorable 65 environmental conditions for bacterial growth than the main distribution network and 66 that it is important to understand and control not only their individual but also their 67 combined impact on the drinking water microbiome. 68 69 Several previous studies investigated the impact of building plumbing conditions on its 70 microbiome. Overall, microbial community compositions tend to change considerably, 71 e.g., (1) during stagnation 11 , (2) while forming biofilms inside flexible shower hoses 12 , 72 or (3) due to the combined impact of material, temperature, and stagnation 13 . 73 Considering one of the above in more detail, studies in our research group that were 74 addressing biofilm formation on flexible polymeric materials revealed (1) high bacterial 75 9

Selection experiment 170
For all five water samples, triplicate microcosms were assembled with the testing 171 material (EPDM) and an additional one containing glass as a control set up (Figure 1, 172 B), as described above. After assembly (t 0 ), the microcosms were incubated (14 d, 173 35°C, 90 rpm) (Figure 1, C). After 14 days (t 1 ), biofilms were removed from the material 174 surface (EPDM and glass; see 2.6.2) and both the biofilm and water phase of each 175 microcosm were sampled for TCC (see 2.6.3) and community composition by 16S 176 rRNA gene sequencing (see 2.6.4). For a second selection step, biofilm samples were 177 re-inoculated into new microcosms. For this, the corresponding drinking water matrix 178 was filtered using sterile bottle top filter units and membrane filters (Whatman® 179 Nucleopore™Track-Etched Membranes, 47 mm, 0.2 µm, Sigma Aldrich). New 180 material was cleaned and stacked, additional nutrients added, and selected biofilm 181 communities were added in a final concentration of 1 x 10 7 cells/microcosm (i.e., 1 x 182 10 5 cells/mL). After another 14d incubation, biofilms and water phases of all 183 microcosms were again sampled (t 2 ), following the same procedure. Regarding 184 terminology, in the course of this study, initial drinking water communities are referred 185 to as original drinking water communities and the biofilm communities of t 1 and t 2 as 186 selected biofilm communities. 187 188

Attachment experiment 189
Here we compared attachment dynamics of selected biofilm communities with the 190 original drinking water communities. The same microcosm set up was used as 191 described above, with triplicate experimental microcosms (EPDM coupons) and single 192 control microcosms (glass slides). The starting concentration of bacteria in the water 193 phase (TCC, t 0 ) was adjusted to be the same by diluting the biofilm communities 194 relative to the drinking water TCC. The microcosms were incubated (35°C, 90 rpm) 195 and the TCC in the water phase was measured for all at 30 min intervals over the 196 course of 5 h (t 1 -t 10 ). Total organic carbon (TOC) was measured using a TOC-V CPH Analyzer (Shimadzu 202 Schweiz GmbH, Reinach, Switzerland). The minimum detection limit of the instrument 203 is 0.1 mg/L. For total phosphorous, samples were chemically digested with potassium 204 peroxodisulfate at 121°C, followed by a reaction to a phosphorous-molybdenum-blue 205 complex and the determination of ortho-phosphate with spectrophotometry. The 206 minimum detection limit of this method was 3.0 µg/L. Total nitrogen concentrations 207 were measured via chemiluminescence using a Shimadzu TOC-L CSH instrument. The 208 minimum detection was 0.5 mg/L. 209 210 2.6.2 Biofilm removal 211 All biofilms were removed with an electrical toothbrush (Oral-B ® , Advanced Power) 212 and toothbrush heads were replaced after each use to prevent cross-contamination. 213 In short: EPDM or glass coupons were placed into muffled glass petri dishes and 214 covered with filtered (0.2 µm) water. The water volume was always 25 mL per coupon 215 (i.e., a total of 100 mL per microcosm). The coupons were brushed one by one, for 216 approximately 90 sec each (including both coupon sides and the edges). During 217 biofilm removal, 10 mL were saved and after the biofilm removal from all four coupons 218 of a microcosm. This volume was ultimately used to recover biofilm residuals in the 219 12

16S rRNA gene-based community analysis 245
For sequencing, samples of (1) all original drinking waters (t 0 ), (2) all selected biofilms 246 (t 1 , t 2 ), and the water phase of the microcosms (t 1 , t 2 ) were concentrated onto 0. were normalized to a concordant concentration followed by the pooling of 5 µL per 280 sample. This pool was adapted to a final concentration of 2 mM and the base-pair (bp) 281 length of the product determined with the Tape Station (627 bp). 282 Sequencing was performed using the MiSeq platform. For this, 0.1M NaOH was added 283 to the pool, centrifuged (300 g, 60 s) and incubated for 5 min (room temperature) prior 284 to the addition of the hybridization buffer HT1. This step was to (1) generate single 285 stranded DNA and to (2) prevent unspecific bindings to the flow cell during 286 sequencing. As a final step, 10% PhiX was added as a sequencing run control 287 Data processing followed a distinct pipeline. First, data quality was controlled using 293 FastQC (Table S2, A). Then, read ends were trimmed and merged (Table S2, B), 294 which was followed by an in silico-PCR and the trimming of the primer sites (Table S2, 295 C). Finally, sequences were filtered (based on quality and size range) (Table S2, D) 296 and amplicon sequence variants (ASV) were generated and taxonomically assigned. 297 The clustering of sequences was performed as presented in a previous study 14 . It is 298 based on an amplicon sequence variant (ASV) approach using UNOISE3, proposed 299 by Edgar and colleagues 21 , and includes a correction for sequencing errors and a 300 chimaeral removal. Clustered sequences are called zero-radius operational taxonomic 301 units (ZOTUs). Due to a potential overestimation of the actual number of ZOTUs, an 302 additionally clustering was performed at different identity levels of 99, 98, and 97%. 303 For predictions on taxonomic assignments, the Silva 16S database (v128) and the 304 SINTAX classifier were used (cut-off 0.9). See supplementary information for details 305 on data analysis using R (Version 3.3.0) and RStudio (Version 1.1.477). drinking water (e.g., in general 14,22 , or specifically for EPDM 23,24 ). A separate growth 320 potential assay at 30 °C showed that 2.3 ± 0.09 x 10 7 cells/mL (n = 3) were able to 321 grow on migrating carbon from EPDM coupons during 14 days, which is 30x more 322 compared to growth in the absence of EPDM ( Figure 2B). Given that the carbon-323 source for growth was the EPDM coupons, this translated to the growth of 2.3 x 10 7 324 cells/cm 2 coupon. Of these cells, 57 % (i.e., 1.3 x 10 7 cells/cm 2 ) were recovered 325 directly from the surface of the EPDM coupons. To summarize, results show that the 326 EPDM coupons favor biofilm formation by (1) providing a surface for colonization and 327 by (2) adding biodegradable organic carbon to the water. Therefore, this material was 328 deemed suitable for the further experiments on biofilm growth and the selection within 329 growing communities (below). 330

Selection experiment 342
The basic concept and design of the growth potential assay was used to test the 343 growth of five different drinking water communities on identical EPDM coupons ( Figure  344 1). All communities showed (1) intensive growth and (2)

Comparatively low taxa diversity detected in biofilm communities 375
A comparison of the original drinking water communities with the biofilm communities 376 at the conclusion of the experiment revealed a notable loss in diversity (apart from B2, 377 see below). Richness, i.e., the number of different taxa, decreased and became more 378 comparable between the different waters. Figure 4A shows the richness values for the 379 five original drinking water communities. Interestingly, tapped waters showed 380 considerably more taxa (2'178 ± 131, n = 9) than the bottled waters, with the original 381 water B2 comprising strikingly few taxa (54 ± 0, n = 3) compared to B1 (270 ± 15, n = 382 3). Overall, biofilm communities comprised comparatively few taxa (29 -50 taxa), 383 which corresponded to a diversity loss of 46 -98 % from the original waters. This 384 impressive loss of up to 2'000 individual taxa (tap waters) highlights the relevance of 385 nutrient-based selection within establishing biofilm communities. As a consequence to 386 this loss in diversity (through growth and selection), the similarity between the biofilm 387 communities increased (with respect to diversity), with only 21 % variation in richness 388 between biofilms as opposed to 73 % between the original drinking water 389 communities. Shannon diversity followed a similar trend. Figure 4B shows a 390 comparable dissimilarity between the original tapped water communities (5.9 ± 0.7, n 391 = 9) and bottled water communities (2.3 ± 0.03, n = 3 for B1; 1.1 ± 0.01, n = 3 for B2). 392 The relation between the Shannon Index (H') and its maximum value (H' max ) is 393 important for drawing conclusions on diversity, i.e., the closer H' to H' max , the higher 394 the diversity within the community. Here, the relation was 1:1.3 for the tapped waters, 395 1:2.5 for B1, and 1:3.7 for B2 respectively, indicating a higher diversity in the tapped 396 waters. This difference decreased with biofilm formation, resulting in a comparable 397 degree of diversity. Here, the ratio between H' and H' max is close to 1:3 for all samples. 398 This shows that (1) diversity decreased for (almost) all communities and (2) that biofilm 399 communities are more similar to each other compared to the original drinking water 400 communities. As indicated above, bottled water B2 was the misfit amongst the original 401 water communities with a particularly low initial richness and diversity. Interestingly, 402 this community also grew the least during the selection experiment (Figure 3). This 403 suggests that the initially low diversity did not allow the community to metabolize as 404

Biofilm growth alters community composition 433
The decrease in taxa diversity came along with a change in community composition 434 from the original drinking water to the selected biofilm communities. Figure 5 illustrates 435 the dissimilarities between the bacterial communities of the original drinking waters (t 0 , 436 triangles) and their corresponding biofilm communities that grew on EPDM coupons 437 (t 2 , circles). The distance between original and biofilm communities was large for the 438 tapped waters (e.g., highlighted for T3, Figure 5). were also detected in the original water community (Table S5)

Attachment experiment 518
The selected biofilm communities attached more and much faster to new surfaces 519 compared to the original water communities (Figure 6). 520  30 % more than for the original water communities (69 ± 23 %, n = 15). Between the 530 original communities, strong differences were measured in maximum attachment. For 531 example, cells from original water B1 attached to 80 ± 1 % (n = 3) within 5 hours as 532 opposed to T1 with only 45 ± 2 % (n = 3). The direct comparison between original and 533 selected communities showed a clear advantage for the selected cells. For example, 534 after 5 hours of incubation one tapped water (T3) showed a relative attachment of 94 535 ± 1 % (n = 3) for the selected community, but only 22 ± 4 % (n = 3) for the original 536 community ( Figure 6E). In absolute numbers, this percentage translates to a maximum 537 attachment of 6.7 ± 0.1 x 10 4 cells/cm 2 (n = 3) for the selected community and 1.4 ± 538 0.3 x 10 4 cells/cm 2 (n = 3) for the original community (Table S6). In addition to the high 539 maximum attachment, maximum attachment rates were on average 5.0 ± 1.7 x 10 4 540 cells/cm 2 /h (n = 15) in selected communities and 2.4 ± 1.2 x 10 4 cells/cm 2 /h (n = 15) 541 in original communities. Regarding T3, maximum attachment rate for selected cells 542 was almost 10-fold higher with 3.9 ± 0.4 x 10 4 cells/cm 2 /h (n = 3) as opposed to 4.6 ± 543 0.9 x 10 3 cells/cm 2 /h (n = 3). The comparison of the relative attachment between 544 selected and original communities of all waters showed a 1-to 4-times higher 545 maximum attachment and a 1-to 7-times higher maximum attachment rate for the 546 selected communities. Interestingly, the attachment dynamics were similar with glass 547 coupons as surface ( Figure S1). Maximum attachment rates were almost identical 548 between EPDM and glass coupons. Maximum attachment after 5 hours was, however, 549 6-fold higher on EPDM coupons (Table S7) We analyzed biofilm growth on flexible EPDM coupons for five different drinking water 558 communities (Figure 1). The purpose was to study the amount of growth due to 559 biodegradable carbon migrating from the EPDM (Figure 2) and to assess selection 560 within the developing biofilm communities due to this carbon. In the course of biofilm 561 growth (Figure 3), all samples showed a significant loss in species diversity ( Figure 4)  colleagues 32 who correlated the ability to outcompete others to a species' growth rate 592 and yield. Our data (e.g., Figure 5) does not allow separation between selection 593 caused by metabolic capabilities and growth physiology. However, this may be an 594 explanation why some taxa dominated in the microcosms. 595 596 The arguments above explain selection during growth, but may lead to an erroneous 597 conclusion that growth on the EPDM coupons should by default result in similar 598 communities being selected. Here, our data clearly showed that all five original water 599 samples resulted in completely different final communities following growth ( Figure 5).    (Figures 3, 4, 5). We demonstrated selection, but also showed 642 that selection differed based on the source water ( Figure 5). While there is an obvious 643 need and scope for larger observational studies on drinking water microbiomes 46 , 644 there is also a clear need for basic laboratory-scale ecological studies that can help to 645 inform on interpretations from complex building plumbing data. Moreover, 646 understanding the basic ecology of building plumbing systems will provide a basis for 647 proactive management of the microbiomes in these systems. 648 649

Managing colonization of building plumbing materials 650
Better knowledge on growth-dependent selection within biofilm communities can be 651 used to design building plumbing systems where the microbiology is controlled or even 652 specifically tailored to the system. Microbial colonization and growth on building 653 plumbing materials is currently not (properly) controlled. Upon commissioning of a 654 building, all new plumbing material is exposed to complex drinking water communities 655 during the first use. In fact, there is essentially no control over the identity and 656 composition of bacteria that attach and proliferate in the new system, irrespective of 657 the location, source water, disinfectant use, building type, or plumbing materials. To 658 date, there are surprisingly few studies looking at this initial colonization of building 659 plumbing materials, both full-and pilot-scale. A notable exception is the study by 660 Salehi and colleagues 34 , where they monitored changes in water chemistry and 661 bacterial growth during the first days/weeks of building occupation. Also, a study by 662 Douterelo and colleagues 47 , showed that specific bacteria are dominant during the 663 initial colonization (7 -28 d) of distribution pipe materials. The fact is, in current 664 practice the owners/operators have effectively no control over the communities that 665 colonize their building plumbing systems. 666 667 Smart use of material properties can control microbial growth (and thus biofilm 668 communities). For example, the use of high quality materials and the avoidance of low 669 quality ones (e.g., flexible hoses) reduces the potential of bacteria to actually grow. 670 For this, standards for material quality requirements have already been implemented 671 in Europe (e.g., 48 ) and official tests on carbon migration and corresponding growth 672 potentials have been established (e.g., 17,49 ) (e.g., Figure 2). Using such tests to qualify 673 the use of individual materials in new buildings should be a must for the industry. A 674 more expensive but sensible approach is to use materials that do not leach any carbon 675 (e.g., stainless steel plumbing). For example, Van der Kooij and colleagues 50 showed 676 that biofilm growth and the incorporation of Legionella spp. was less on stainless steel 677 compared to polymeric PE-X pipes. It is, however, important to take into account that 678 high quality polymeric materials can perform as good as metal piping with regard to effect on the establishment of Legionella spp. in biofilms 57-59 , which can potentially be 704 exploited as probiotic communities against Legionella pneumophila. 705

706
Here we propose a combination of the approaches above. We argue for the use of 707 plumbing materials that provide specific substrates and for the targeted colonization 708 of these materials of a benign microbial community. The approach foresees the use 709 of materials that migrate organic carbon in such a quality and quantity that it allows 710 bacteria to grow and to sustain their existence in the developing biofilm. We 711 furthermore propose colonizing these materials with bacteria from a safe source (e.g., 712 bottled water), pre-selected on the substances migrating from the material (e.g., 713 Figure 5). This adaption to the nutrients ultimately allows for a rapid colonization (e.g., 714