Unchartered waters: the unintended impacts of residual chlorine on water quality and biofilms

Chlorine residual impacted discolouration

Unexpectedly, flushing of the High-chlorine system (by incrementally increasing the flow rate) produced a significantly greater discolouration response (assessed via turbidity) than the Medium- or Low-chlorine systems across all stages of flushing, of both tests (Fig. 2). Compared to the other regimes, the High-chlorine system also had a greater final concentration of iron (known to be associated with discolouration) at the end of Flush 1 and a greater rate of iron mobilisation during Flush 2 (Fig. 2). Conversely, the Low-chlorine regime consistently resulted in the lowest impact on water quality with the lowest discolouration and metal concentrations. Even after just 28 days of growth, material was mobilised from the High-chlorine regime at sufficient volumes to approach or breach the water quality standards for discolouration and iron concentrations (Fig. 2 and Supplementary Table 1). This contradicts the common perception of residual chlorine impacts on water quality and also studies of cast iron pipes, which suggest increasing oxidant concentration (disinfectant or dissolved oxygen) in drinking water decreases iron release^26,27. Although surprising, High-chlorine repeatedly resulted in the greatest discolouration and Low-chlorine the least; as observed during the flushing of test 1, test 2 (Fig. 2) and preliminary tests (Supplementary Fig. 2).

Irrespective of chlorine regime, the bulk water turbidity, iron and manganese concentrations increased significantly during flushing, as shear stress increased (Fig. 2; other physio-chemical parameters did not differ significantly). Manganese (also associated with discolouration) was consistently mobilised at lower concentrations than iron, but, similarly to iron, greater manganese concentrations were mobilised from High- and Medium-chlorine than Low-chlorine biofilms (Fig. 2). However, manganese concentrations showed the least change between chlorine regimes, compared to turbidity or iron. Critically, the different discolouration responses were not driven by the rate of (in)organic supply because the chlorine regimes had equivalent hydraulics (governing rate of supply and transfer to biofilms) and bulk water quality (e.g. iron, manganese, TOC concentrations), indicating this trend is process driven/limited. These results could be considered to infer that chlorine use should be decreased or eliminated to reduce the possibility of a measurable discolouration response occurring. However, the application of machine learning to historical water quality data highlighted that destabilisation of scale can occur when oxidant concentrations are low²⁸. Also, chlorine plays several critical roles in DWDS and water treatment, namely microbial inactivation and limiting planktonic regrowth. Therefore, it would be naïve to withdraw chlorine without better consideration of the drivers causing the discolouration differences described herein, which requires characterisation and understanding of the (in)organics (i.e. biofilm) at the pipe wall, including the impact that chlorine has upon them.

Biofilm inorganics

Greater accumulation (and subsequent mobilisation) of iron occurred in High-chlorine biofilms compared to Medium- or Low-chlorine biofilms, with the latter having the least (Fig. 3), mirroring the bulk water discolouration responses (Fig. 2). Various elements were detected at the pipe wall (Supplementary Fig. 3), analysis of total elemental fingerprints highlighted iron and chlorine as the main inorganic descriptors for differences between biofilms. Manganese was only detectable in three High-chlorine biofilms (Methods section: cell concentration analysis), where, similarly to the bulk water trends, concentrations were lower than iron. As the biofilms were formed within HDPE pipes no leaching of iron would occur from the pipe wall into the biofilm, rather the input of iron in to the system is as a trace inorganic within the incoming source water. This is a possible limitation of this study, in that cast iron pipes are known to promote different bacterial communities to plastic pipes²⁹, so there may be a need to research the effects of chlorine residual specifically within cast iron pipes. However, the current study is useful in that it facilitates exploration of the specific role of biofilms (in isolation) in concentrating iron and acting as a sink/source of inorganics within DWDS.

Biofilms contained less iron after flushing than prior to flushing (irrespective of chlorine regime; Fig. 3 and Supplementary Fig. 3) providing direct evidence that the iron mobilised into the bulk water originated from the pipe wall. This release of entrapped metals from the pipe wall has been previously inferred based on increases in bulk water metal concentrations during the flushing of pipelines^4,5 (a trend also observed in the current study) but not conclusively determined.

Metal concentrations in the bulk water were the same between chlorine regimes throughout the growth phases of both test 1 and test 2. Therefore, elevated iron and manganese concentrations at the pipe wall must have been driven by processes expedited in the High-chlorine regime. Chlorine is an oxidative agent, therefore higher residual chlorine may have promoted metal precipitation causing greater accumulation. However, the kinetics of free chlorine oxidising metals is generally slow in relation to the typical hydraulic retention times of water systems as described by Knocke et al³⁰. High-chlorine did not have a significantly greater ORP than Medium-chlorine and only differed slightly from Low-chlorine during test 1, which seems to confirm that the differences in metal accumulation were not dominated by chemical oxidation in response to the different regimes. Oxidation of metals such as iron or manganese can occur due to microbial oxidation as well as chemical oxidation; a review of manganese oxidation specifically mentions the uptake of manganese by media support biofilms³¹, and metal oxidation is known to occur due to intracellular uptake by oxidising organisms which have been detected in DWDS³² or via adsorption to EPS molecules³³. Additionally, iron concentrations were greater in Pre-Flush2 biofilms than Pre-Flush1 biofilms, (a trend most pronounced in High-chlorine), despite no significant differences in ORP between the growth phases of test 1 and test 2. This further refutes the oxidation/chemically driven precipitation theory and suggests that additional processes were governing the differential iron accumulation, such as microbially driven processes due to different EPS or microbiome compositions in test 2 compared to test 1. Furthermore, during flushing of test 1, Medium-chlorine had a greater discolouration response than Low-chlorine, yet these regimes had similar Pre-Flush1 biofilm iron concentrations, suggesting factors other than iron concentration are influencing discolouration. Given the association between biofilms, discolouration and water quality⁵, microbiological processes are considered likely to influence the observed differences.

Biofilm cell quantification

Irrespective of chlorine regime, biofilm total and intact cell concentrations (TCC and ICC) reduced during flushing (Fig. 4). Application of flow cytometry to DWDS biofilms and bulk water confirmed the relationship between the microbial phases: greatest biofilm TCC and ICC mobilisation was from Low-chlorine, which had the greatest increase of planktonic TCC and ICC, and least discolouration response (Figs. 2 and 4).

In both test 1 and test 2, Pre-Flush biofilms from each regime had significantly different TCC and ICC (Fig. 4; χ² ≥ 7, p ≤ 0.04) such that High chlorine concentration reduced (but did not prevent) biofilm formation. The cell concentrations observed, particularly in the Low-chlorine biofilms, are higher than previously reported for simulated DWDS with disinfectant residuals¹⁸ and similar to non-flexible premise plumbing materials tested in a non-chlorinated system³⁴. This could be due to variations in methodology, specifically different biofilm removal or homogenisation methods between studies (the analysis herein included singlet-doublet evaluation to ensure homogenisation of ≥98%, see Methods section: cell concentration analysis). Alternatively, this elevated growth may be a consequence of de-chlorinating a previously chlorinated water supply with relatively high organics compared to non-chlorinated DWDS (where maintaining water biostability relies on improved efficiency of organic removal at treatment works), indeed the planktonic cell counts also increased in the Low-chlorine regime (Supplementary Table 1). Previous studies have also shown disinfectant residuals do not prevent biofilm formation^35,36; at best reducing bacterial concentration^16,18, biological activity or growth^22,37,38 crucially, this inhibitory effect does not necessarily apply to other taxa such as fungi¹⁶. Chlorine residuals may indirectly reduce biofilm cell quantities by reducing the planktonic cells available to colonise the pipe wall (Supplementary Table 1). Intriguingly, increasing residual chlorine did not appear to “kill” biofilm cells at a greater rate; the proportion of ICC (as a % of TCC) was similar between regimes (χ² ≥ 1, p ≥ 0.15). Potentially, ICC proportions within DWDS biofilms are governed by a constraint other than chlorine (e.g. hydraulics, nutrients, EPS characteristics).

Biofilm cell accumulation and mobilisation were affected by chlorine concentration and inversely correlated with biofilm iron concentration (Figs. 3 and 4; Supplementary Table 2) and water quality degradation (observed as discolouration; Fig. 2, Spearman’s rank correlation could not be calculated reliably for the flushing phases as difference in averages would have had to be used). Note that in test 2 there was an initial and brief lag (<24 h) in reducing chlorine concentrations of the Low-chlorine regime to concentrations comparable to test 1 (Supplementary Fig. 1), this was due to the influx of a fresh water volume after the flushing, which was then dechlorinated. This could have impacted the initial recovery of the Low-chlorine biofilms and led to a decrease in TCC between Post-Flush1 and the Pre-flush2 sample points. Although, other aspects of biofilm behaviour such as succession, community function or the EPS will have likely had a greater impact on biofilm regrowth¹⁶. In all regimes, EPS accounted for the majority of the biofilm volume that accumulated during growth, not cells (Pre-Flush1 and Pre-Flush2 EPS-to-cell ratios >1; Supplementary Table 3). Also, more EPS than cells were mobilised during flushing (EPS-to-cell ratios decrease; Supplementary Table 3), demonstrating clearly that EPS is an essential biofilm component to consider.

EPS characterisation

EPS (predominantly comprising proteins and carbohydrates) is produced at an energetic cost to microorganisms for various vital roles²¹, including concentrating (in)organics from the fluid-phase, mechanical stability and disinfection protection. At Pre-Flush1, Low-chlorine biofilms had the most EPS-per-cell (4.39 Arbitrary Units; AU), the presence of residual chlorine reduced EPS-to-cell ratios although no concentration effect occurred between Medium- (1.14 AU) and High-chlorine (1.24 AU; Supplementary Table 3). By Pre-Flush2, High-chlorine and Low-chlorine biofilms had similar EPS-to-cell-ratios. Possibly, EPS production was accelerated during biofilm regrowth in High-chlorine (compared to the growth phase of test 1) due to chlorine-tolerant EPS or microorganisms remaining post-flush, providing a niche and/or community that promoted recolonisation and EPS synthesis.

EPS matrices were generally protein dominated. The EPS of High-chlorine biofilms sampled at Pre-Flush1 had the greatest carbohydrate proportion of all the chlorine regimes, although the matrix was still predominantly protein (Supplementary Table 3). By the Pre-Flush2 sample point, carbohydrates dominated High-chlorine EPS. Chlorine was documented to degrade the EPS of batch-cultured bacterial biofilms, promoting cell survival and culturability²². Similarly, chloramine has been shown to reduce the biomass (determined via cell and polysaccharide quantification) of drinking water biofilms in reactors³⁸. A study by Xue et al.³⁹, found that EPS from Pseudomonas aeruginosa biofilms had a high reactivity with chlorine but a low reactivity with monochloramine, this suggests that different protective mechanisms may be inferred by different EPS compositions. High-chlorine biofilms may have synthesised greater protein volumes than other regimes with the elevated residual chlorine “eroding” a proportion, altering the matrix compositional ratio (potentially affecting cohesion/adhesion, viscosity and diffusional properties) and, indirectly, protecting biofilm microorganisms (hence similar biofilm ICC proportions between regimes).

Frequently, proteins were more readily mobilised than carbohydrates: carbohydrate-to-protein ratios increased between biofilms from Pre-/Post-Flush1 and Pre-/Post-Flush2 (Supplementary Table 3). Significant decreases in protein volume (after flushing) were detected where discolouration was the greatest: Medium-chlorine (W = 201, p < 0.001) and High-chlorine (W = 169, p < 0.020). Proteins may be more reactive than carbohydrates, with greater influence in concentrating (in)organics (or binding/deactivating chlorine residuals). Subsequently, greater protein mobilisation could release greater metal concentrations, causing elevated discolouration. However, protein mobilisation was not detected from Low-chlorine biofilms, yet a discolouration response was observed. This could be attributed to limits of detection, data variability or scale differences: EPS analysis is based on fields of view, discolouration assesses material mobilised from the entire pipe surface.

Overall, High-chlorine biofilms required the most disinfection protection, potentially conveyed by a different EPS composition, which, indirectly, influenced the concentration of discolouration material. The EPS response to residual chlorine was complex and likely impacted by biofilm development rate, cell growth/replication, nutrient availability, flushing and the ecology of the microbiome synthesising the EPS matrix.

Biofilm microbiome

Bacterial and fungal communities were distinct between the three chlorine regimes prior to flushing (Fig. 5; ANOSIM, bacteria, global-R = 0.383, p < 0.001; fungi, global-R = 0.444, p < 0.001) and remained distinct after flushing (Supplementary Fig. 4; ANOSIM, bacteria, global-R = 0.513, p < 0.001; fungi, global-R = 0.319, p < 0.001). Although, ecological indices (Supplementary Table 4) were unaffected by flushing (or chlorine regime) and had substantial variance, suggesting heterogenic biofilm communities developed. Contrasting trends regarding disinfectant and ecological indices are reported in the literature with some studies reporting decreased biofilm diversity where a chloramine residual was present^13,14, others an increase in diversity with increased residual concentration⁴⁰. The differing trends suggest that constraints other than oxidative stress are governing diversity; the chloramine studies in particular compared DWDS of different countries with different water sources, treatment and quality, which will have an impact on the microbiome. However, ecological indices are reductive and can only evidence major differences between community structures, the findings in the current study (which benefitted from using the same water source between experiments) are indicative that changes in the biofilm microbiome between chlorine regimes were more subtle. Disinfection is known to alter planktonic bacterial community compositions^10,11,12,19 and has previously been demonstrated to also shape the succession of biofilm bacterial communities during growth (there was less of an impact of chlorine on fungi) within the DWDS test facility used in this study, even when the inoculum was pre-conditioned by being chlorinated water¹⁶. In the current study, pre-flush and post-flush data were analysed (Fig. 5 and Supplementary Fig. 5) to determine any differences in response to flushing between regimes, comparing changes in the biofilm microbiome (structure and composition) that remains adhered, whilst considering the microbial consortium that was mobilised and its impacts on microbial water quality. Consideration is also given to the any compositional differences between regimes at the pre-flush sample points, which could provide insight into the differences in the previously discussed distinct biofilm characteristics (iron concentration and EPS).

A significant shift in bacterial communities was observed during flushing of Medium-chlorine (global-R ≥ 0.384, p ≤ 0.016; Fig. 5; and Supplementary Fig. 5) and High-chlorine biofilms (global-R ≥ 0.600, p ≤ 0.008; Fig. 5; and Supplementary Fig. 5), where greater discolouration was observed. This indicates the preferential loss of some bacteria over others, such that relative abundances were significantly affected. Conversely, the Low-chlorine pre- and post-flush bacterial communities did not cluster independently and their composition was not significantly affected by flushing (global-R ≤ 0.016, p ≥ 0.413; Fig. 5; and Supplementary Fig. 5), perhaps suggesting that there was no-preferential loss of certain bacteria and all taxa were equally likely to be mobilised. The bacterial genera that were highlighted as responsible for driving the differences between pre- and post-flush biofilms were inconsistent between the Medium- and High-chlorine regimes, demonstrating that there was no one key organism that was associated with weakly adhered biofilm and always removed entirely, and rather, a mixed bacterial consortium was mobilised. The mechanical resistance of a biofilm has been described to be governed by the EPS adhesive/cohesive forces, which is perhaps more influenced by environmental parameters than the microbiome, as has been suggested with regard to the impact of hydraulics on DWDS biofilms⁴¹.

The impact on water quality of the biofilm microbiota being mobilised depends upon the presence and concentration of deleterious microorganisms (aesthetic or potentially pathogenic) in the community that develops and is subsequently detached. At higher taxonomic levels (phyla/class/order) the same bacterial constituents were present in all biofilms prior to flushing (Supplementary Fig. 6), although the dominant taxa varied. However, chlorine regime significantly impacted the presence/absence (global-R ≥ 0.136, p ≤ 0.007) and relative abundance (global-R ≥ 0.227, p ≤ 0.001) of bacterial families (Supplementary Fig. 4) and genera (Fig. 5). Alphaproteobacteria abundance increased with increasing residual chlorine concentration, whilst Betaproteobacteria (recently reclassified as the order Betaproteobacteriales within Gammaproteobacteria⁴²) dominated at low chlorine, supporting trends reported at low chloramine concentrations⁴³ and those highlighted during detailed analysis of community development in our previous paper¹⁶. Some phyla such as Firmicutes and Actinobacteria were absent from Low-chlorine, and some taxa such as Clostridiaceae (a family within Firmicutes, which includes pathogens) were unique to High-chlorine (Supplementary Fig. 6). This is consistent with trends reported in comparison of chlorinated operational DWDS, where an enrichment of Firmicutes was associated with higher residual chlorine concentrations⁴⁰. Actinobacteria have been associated with discolouration in field-studies²⁹ and reported to dominate under high chloramine⁴³, suggesting disinfection/oxidant resilience. Chloramines are reportedly less reactive than chlorine, penetrate biofilms more deeply and persist throughout DWDS⁴⁴. Consequently, the selective pressure of chloramines upon biofilms may be greater than those of free chlorine and the effects observed here perhaps accentuated. Indeed, several studies of chloraminated systems have reported significant enrichment of bacteria that are associated with disinfection tolerance and ammonia oxidation (driven by the input of ammonia into systems due to chloramination)^13,18,38.

At the genus level, Pseudomonas, Sphingobium and Acidovorax are commonly detected in water^10,45,46 and exhibited inconsistent trends with residual chlorine and flushing. This is consistent with previous disinfectant studies, which showed that while there was a selective pressure of disinfectant residual for specific bacteria or genes^13,20,40,38, there were also shared taxa between biofilms from DWDS with different disinfectant types²⁰ or disinfectant concentrations¹³. However, the abundance of some genera (e.g. Achromobacter, Bosea and Rhizobium) decreased with increasing chlorine concentration, while others such as Novosphingobium, Sphingopyxis and Methylobacterium increased in abundance. Prior to flushing, some bacteria were only detected in one regime, for instance Rhodobacter, which at pre-flush was unique to High-chlorine biofilms but then absent from the post-flush High-chlorine samples (Fig. 5). Methylobacteria are commonly isolated from drinking water; they are reported to exhibit resistance to disinfectant/cleaning agents^13,40,47,48, promote aggregation^46,49 and impact microbially-influenced-corrosion⁵⁰. Possibly the genetic components implicating Methylobacteria in microbially influenced corrosion could increase the tendency of High-chlorine bacterial communities to oxidise iron (especially if shared via horizontal gene transfer). Similarly, the exclusivity of Rhodobacter (known iron oxidisers^51,52) to High-chlorine biofilms could help explain the increased iron concentration observed, despite all the regimes being supplied with the same bulk water iron concentrations. Ultimately, despite a greater mobilisation of cells from the Low-chlorine regime (Fig. 4), the absence or reduced abundance of certain bacteria within these communities prior to flushing suggests that the mobilised cells were less associated with iron (and therefore discolouration) than in the Medium- or High-chlorine regimes.

Fungal community compositions generally showed no significant differences due to flushing (global-R ≤ 0.350, p ≥ 0.071), apart from between Pre-Flush1 and Post-Flush1 of the Low-chlorine regime (global-R = 0.436, p = 0.016; Fig. 5). Nor were there clear trends with chlorine concentration (Fig. 5 and Supplementary Fig. 7), although the communities were distinct between regimes, a finding first reported in Fish and Boxall¹⁶, which concludes that this is possibly because fungi have greater chlorine tolerance due to their robust morphology¹⁹. Variation in fungal communities may be stochastic, impacted by the seeding population dictated by upstream environmental pressures, or indicative of an ecological pressure other than chlorine governing taxonomic composition. Despite the contribution that fungi make to the DWDS microbiome they are rarely sampled or monitored in operational DWDS. However, the greater resilience to disinfection and mechanical stress of fungal communities, compared to bacteria, suggests an integral role in biofilm maturation, possibly promoting biofilms that are resistant to cleaning and management practices.

Taxonomic variation in the biofilm communities suggests the cells mobilised from each regime presented different impacts on microbial and aesthetic water quality (and potentially public health). For example, High-chlorine biofilms accumulated fewer cells and lost fewer during flushing, yet they were more likely the source of iron-associated and chlorine-resistant microorganisms, which would be less affected by bulk water chlorine residuals, thus more likely to reach a customer intact. Additionally, studies have reported associations between chlorine or chloramine tolerance and potentially deleterious traits such as antimicrobial resistance^{10,17,18,38,53} or (potentially) pathogenic taxa^14,20,38, even when biofilm biomass (evaluated by quantifying cells and polysaccharides) was observed to be suppressed by the presence of a residual³⁸. If this association is confirmed to be due to the specific selective pressure of disinfection concentration then biofilms such as the High-chlorine biofilms could be an upstream source of antimicrobial resistance genes. For instance, the family Sphingomonadaceae have been associated with antibiotic resistance⁵⁴. In the current study, Novosphingobium, Sphingopyxis (members of this family) were more abundant in Medium- and High-chlorine regimes than the Low-regime, consistent with trends reported in chlorinated¹⁷ and chloraminated³⁸ systems where Sphingomonadaceae taxa increased in abundance with increased disinfectant concentration. With respect to potential pathogens, Mycobacterium (a genus which includes opportunistic pathogens) were more abundant in the biofilms of a system with high chloramine concentration (3.8 ± 0.1 mgL⁻¹) than a system with no residual (0.08 ± 0.01 mgL⁻¹), despite being present at similar concentrations in the bulk water of each system¹⁴. Mycobacterium were not detected in any of the biofilms from the present study, perhaps because of the high abundance of Methylobacteria (the two taxa have been reported to occupy a similar ecological niche¹³.) Conversely, planktonic and biofilm Legionella concentrations were reduced by the presence of a disinfectant residual within water mains¹⁵. These findings, and possible implications, could suggest that an alternative residual disinfection strategy to continuous dosing, such as pulses at high concentrations or cycling concentrations, might be more efficient at harnessing the benefits of disinfection, limiting the impacts of biofilm on water quality whilst reducing the likelihood of selection pressures conditioning for biofilms that are more difficult to manage.

Metal oxides may convey protection to microorganisms by reacting with chlorine residuals and forming deposits⁵⁵. Greater iron accumulation in High-chlorine biofilms may have occurred due to different EPS bio-chemical compositions and the increased occurrence of iron-associated bacteria. Simultaneous increases in chlorine residual protection and discolouration potential (if mobilised) are consequences of this elevated iron. Conversely, Low-chlorine biofilms did not require disinfectant-resistant adaptations, retained greater mechanical stability (with an EPS with a lower propensity to concentrate iron) and caused less discolouration. Considering the biofilm analyses holistically, it is possible that a trade-off (or progression) exists between a biofilm being “chemically” or “mechanically” stable, with adaptations to survive higher chlorine residuals coming at the expense of the ability to resist hydraulic shear.

Source: Ecology - nature.com