Human- and infrastructure-associated bacteria in greywater

Abstract

Greywater, the wastewater from sinks, showers and laundry, is an understudied environment for bacterial communities. Most greywater studies focus on quantifying pathogens, often via proxies used in other wastewater, like faecal indicator bacteria; there is a need to identify more greywater-appropriate surrogates, like Staphylococcus sp. Sequencing-based studies have revealed distinct communities in different types of greywater as well as in different parts of greywater infrastructure, including biofilms on pipes, holding tanks and filtration systems. The use of metagenomic sequencing provides high resolution on both the taxa and genes present, which may be of interest in cases like identifying pathogens and surrogates relevant to different matrices, monitoring antibiotic resistance genes and understanding metabolic processes occurring in the system. Here, we review what is known about bacterial communities in different types of greywater and its infrastructure. We suggest that wider adoption of environmental sequencing in greywater research is important because it can describe the entire bacterial community along with its metabolic capabilities, including pathways for removal of nutrients and organic materials. We briefly describe a metagenomic dataset comparing different types of greywater samples in a college dormitory building to highlight the type of questions these methods can address. Metagenomic sequencing can help further the understanding of greywater treatment for reuse because it allows for identification of new pathogens or genes of concern.

Introduction

As water resources become increasingly scarce, creative solutions for water reuse are more important than ever. Using recycled water for non-potable purposes allows freshwater resources to be used primarily for drinking water. However, in order to make water reuse a viable and scalable solution, the pathogen load of used water must be mitigated prior to reuse. A widely adopted source for water reuse is greywater, or wastewater from showers, sinks and laundry. Some studies consider kitchen sink effluent to be greywater, while others exclude it. Several studies have found greywater to have lower concentrations of faecal pathogens than blackwater (the water from toilets), making source-separated greywater a promising candidate for wider adoption of on-site water reuse (Jefferson et al. 2004; Li et al. 2009; Eregno et al. 2018). Additionally, greywater represents the largest fraction of household effluent (Larsen et al. 2013), making it more available in general for these uses. On-site treatment may also be considered prior to greywater discharge into the environment if its stream is separate from blackwater, which can remove some load off wastewater treatment plants.

Despite its potential value in water reuse systems, relatively little is known about the microbial composition of greywater. Most studies focus on the quantification of pathogens that are considered to be of concern in blackwater, and in many cases, pathogen concentrations are actually estimated based on measuring quantities of nonpathogenic faecal indicator bacteria. These indicators may not be as useful for assessing the quality of greywater since greywater contains fewer faecal inputs. Furthermore, greywater and its infrastructure provide an entirely different environment for its microbial constituents, driving selection that ultimately leads to very different microbial communities from those in blackwater. Important environmental factors impacting greywater microbial communities include the sources and use of the influent water, the types of infrastructure (including plumbing materials), chemicals used (like detergents and surfactants) and residence time of the water. More research is necessary to determine whether there is a ‘core greywater community’ and to find species that can be used as signatures of different types of influent greywater. In this review, we describe the properties of greywater in its context as a distinct environment for bacterial communities and summarize what is known of the bacterial community composition in different categories of greywater and greywater infrastructure (Fig. 1). We also report some preliminary findings from a metagenomics analysis of greywater and infrastructure samples that were previously characterized using 16S amplicon sequencing in Keely et al. (2015). We suggest that environmental sequencing is key to understanding the ecological dynamics underlying greywater communities. Characterizing these dynamics, along with the specific taxonomic composition of bacterial communities, at different locations within greywater systems and over time will inform the use of recycled greywater by helping identify (ⅰ) biological processes for removal of organics and other chemicals, (ⅱ) the most abundant pathogens and (ⅲ) new pathogens or surrogates.

Figure 1.

Schematic of a typical greywater system and factors affecting the bacterial community in each element of the system.

Distinctive features of greywater

The composition of greywater reflects its unique inputs and is impacted by geography. Most greywater contains elevated levels of detergents, surfactants, pharmaceuticals, pigments and toxic heavy metals, reflecting human use of these chemicals. A recent review of the literature found that the typical temperature range of greywater was 18–35°C, higher than other wastewater sources (Oteng-Peprah et al. 2018). Kitchen greywater contains a wide range of organic compounds, including nitrates and their derivatives, phosphorus derivatives and xenobiotic organic compounds (XOCs). At least 200 XOCs have been found in analyses of shower and handwash water, representing surfactants, emulsifiers, fragrances and antioxidants (among many other categories), and in one study, these corresponded with reports of cosmetic and hygiene products used by residents of the facility measured (Eriksson et al. 2003). High levels of fatty acids have been found from chemical analyses of handwash water, likely from soaps (Eriksson and Ledin 2003). In terms of nutrients, Jefferson et al. (2004) found that greywater was nitrogen-, phosphorus- and trace metal-deficient relative to its chemical oxygen demand, which may have implications for downstream treatments that rely on biological processes for chemical removal. Despite greywater having some distinct and shared characteristics relative to other types of wastewater, there is a large variability in its biological and chemical composition—even within the same category of greywater (Leal et al. 2007; Callewaert et al. 2015).

The distinct physicochemical characteristics and source waters of greywater create an environment that can select for very different microbial consortia than those found in other types of wastewater. The unique and diverse organic molecules often present in greywater effluents cannot be utilized as a carbon source by all taxa, leading to selection for taxa with specialized metabolic capabilities. For example, the presence of aromatic carbons like benzene and toluene or their precursors in greywater might promote the persistence of bacterial taxa that can metabolize them, like Tolumonas and Dechloromonas (Keely et al. 2015; Rodríguez-Martínez et al. 2016; Truu et al. 2019). The type of pre-treatment intermediate storage, its residence time and the chemicals and nutrients available can also lead to anaerobic environments, particularly in infrastructures like holding tanks, which also can only support a subset of bacterial taxa and may cause a major shift in community composition based on which taxa can proliferate. Shower/handwash and laundry water contains a strong signature of human skin-associated bacterial genera like Staphylococcus and Corynebacterium (Huttenhower et al. 2012; Zimmerman et al. 2014; Callewaert et al. 2015; Keely et al. 2015). Their survival and persistence in plumbing and storage systems may differ from that of human gut-associated bacteria. The significantly lower faecal input to greywater is responsible for differences in pathogen load and composition. But again, there is a wide range of diversity in concentrations; previous surveys of greywater literature have reported a variety of Escherichia coli, total or faecal coliform and total heterotrophic bacterial counts spanning several orders of magnitude (non-detects to 10⁷ × 100 ml⁻¹) (Friedler et al. 2006; Boyjoo et al. 2013).

Antibiotics are particularly concerning among the many chemicals present in greywater, as they may exert selective pressure on bacterial communities that can ultimately lead to the development of antibiotic resistance. Several studies have detected triclosan, an antibiotic found in many personal care products, in household greywater (Eriksson et al. 2003; Almqvist and Hanaeus 2006). A recent study which used liquid chromatography–mass spectrometry to measure a number of biocides in greywater systems in the West Bank Palestinian Territories found a number of antibiotics including azithromycin, ciprofloxacin, erythromycin, linezolid, oxacillin, oxolinic acid, penicillin G, sulfamethoxazole, tetracycline, triclocarban and vancomycin (Craddock et al. 2020). In treated greywater, over 20 strains of tetracycline-resistant bacteria, including those with multiple resistances, were found in a system where this greywater was being used for crop irrigation (Troiano et al. 2018). While concentrations of antibiotics in greywater would not be expected to reach levels found in blackwater, they must be considered as a part of the chemical environment for greywater bacteria.

Evaluation of greywater for reuse

Commonly proposed end uses of recycled greywater include landscaping and toilet flushing. For any end use, the risk of human exposure must be assessed and is determined based on the probabilities of exposure and subsequent infection (Benami et al. 2016; Jahne et al. 2017; Shi et al. 2018). Safe reuse of greywater requires assessment and elimination, to an acceptable level based on regulatory criteria, of (ⅰ) pathogenic organisms, (ⅱ) chemicals of concern to human health and (ⅲ) organic materials. Each of these categories presents different levels of risk, in multiple temporal increments (chronic vs acute), to several endpoints (human health, environmental health). Pathogen removal is usually conducted using disinfectants, while organic carbon compounds and nutrients are often removed using biological processes. While removal of chemicals is not covered in this review, it is important to note that the treatment methods and the extent of removal can impact the microbial community.

A number of bacterial pathogens or indicator species have been identified and quantified in greywater, including Bacteroidales spp., Clostridium spp., total and faecal coliforms, E. coli enterococci, streptococci, Legionella spp., Pseudomonas spp., Salmonella spp., Shigella spp., Staphylococcus spp., Campylobacter spp. and Vibrio spp. (A concise summary of findings from these can be found in Benami et al. 2015). The concentrations of these pathogens in greywater must be measured in order to determine the appropriate level of reduction for reuse. However, many attempts to quantify pathogens in greywater have been unsuccessful due to low or inconsistent concentrations, motivating the need for basing log reduction targets on other indicators. For example, concentrations and log-reduction targets for the bacterial pathogens Campylobacter and Salmonella have been modelled in greywater based on faecal loadings, household water usage and population size along with probability of infection over time for different-sized households and buildings. The greatest reduction was necessary in large greywater systems compared with either smaller greywater systems or captured stormwater (Jahne et al. 2017; Schoen et al. 2017).

Since many of the pathogens of concern in greywater are present at very low concentrations and sometimes inconsistently (i.e. present only during relatively rare infection events), there is a need to identify appropriate surrogate organisms to demonstrate log reduction of potential treatments (O’toole et al. 2014). The limitations of using quantification of surrogate microbial species to quantify pathogen load and reduction have been well documented (O’toole et al. 2014; Benami et al. 2016). Briefly, the organisms chosen may not be good indicators if their residence time and ability to reproduce in greywater do not correlate with those of the pathogens. Additionally, the pathogens of interest have generally been selected based on pathogens known to be of concern in blackwater, which has a much higher faecal content; greywater-specific pathogens are not always considered. To address this issue, Staphylococcus has been suggested as an alternative surrogate to evaluate the performance of greywater treatments (Shoults and Ashbolt 2018) since skin-associated bacteria are likely to be more abundant than enteric bacteria in most greywater. Other studies use reduction of total heterotrophic bacteria (Friedler et al. 2006; Winward et al. 2008) as a metric, but this assumes that reduction of heterotrophic bacteria can be considered an indicator of pathogen reduction. Given the assumptions associated with the above approaches, there is a need to characterize the bacterial communities specific to greywater in order to determine whether the same pathogens (as those in blackwater) are of concern or whether resources would be better directed at exploring alternative targets. Additionally, partitioning the human-associated and infrastructure-associated components of the bacterial communities in greywater can help inform design of future plumbing and treatment systems based on knowledge of which bacteria are selected for in different types of infrastructure as well as which taxa are specific to the greywater sources that will pass through the system.

Approaches to studying greywater bacteria

Most studies of microbial constituents in greywater focus on counts of individual species, particularly pathogens (e.g. Salmonella spp., Campylobacter spp., pathogenic E. coli, Legionella spp.) or other genera that can be considered indicators of pathogen presence (e.g. faecal coliforms, Enterococcus spp.). Culture-based assays like colony-forming unit (CFU) counts are limited to organisms that can grow in selective media but are valuable in their ability to detect viable cells, which DNA sequencing methods cannot effectively discern (Li et al. 2017). Quantitative PCR (qPCR) is a culture-independent method that allows for higher throughput and quantification of targeted genes of interest. These genes can be taxon specific or indicate the presence of functional groups, such as virulence factors or markers for antimicrobial resistance or specific metabolic processes.

Amplicon sequencing has increasingly been adopted to gain a more comprehensive understanding of microbial communities in different environments. This technique has been used extensively with sewage (Shanks et al. 2013) and drinking water infrastructure (Ji et al. 2015), but only more recently used to characterize greywater (Callewaert et al. 2015; Macedo et al. 2017; Troiano et al. 2018; Truu et al. 2019) and associated biofilms (Feazel et al. 2009; Dalahmeh et al. 2014; Rodríguez-Martínez et al. 2016; Gebert et al. 2018). The 16S ribosomal RNA subunit gene is commonly used to characterize prokaryotic taxonomic diversity based on the variability present across ribosomal subunit prokaryotic genomes. The use of 16S sequencing allows for broad comparisons between communities, although the sample processing methodology, specific primers used, sequencing technology and taxonomic-assignment method chosen can affect the results (Cai et al. 2013; Guo et al. 2013; Fouhy et al. 2016). Within an individual study, samples can be compared between different locations and time points in order to understand spatial and temporal variability (Keely et al. 2015). This method could serve as an important supplement to the measurement of indicator organisms by validating whether there are relatively abundant members of the community (though it is important to note that relative abundance is not the same as absolute abundance, which can be measured using complementary methods like qPCR). It could also be used to identify additional, perhaps more appropriate indicator organisms for evaluating treatment efficacy.

Characterizing the taxa present can offer important insights into community function based on the metabolic and ecological roles of its members. Truu et al. (2019) found associations between taxa and nutrient removal efficiency in artificial wetland filters used for greywater treatment. For example, the genera Acidovorax and Terrimonas were correlated with removal of total organic carbon (TOC) and total nitrogen, as was Tolumonas to removal of total and ammonia-associated nitrogen. Furthermore, in vertical filters, community diversity was strongly linked to treatment efficiency. Because many treatment systems seek to utilize bacterial communities for carbon and nutrient removal, it is important to identify the key taxa—or consortia—responsible for these processes. While inferences can be made based on knowledge of the taxa present, 16S sequencing is not sufficient to describe the functional potential of a bacterial community.

Metagenomics, the shotgun sequencing of whole genomes from environmental samples, allows for a fuller picture of all the genes (along with taxa) present in a community. Very few studies have used metagenomics with greywater samples. Delforno et al. (2017) compared the functional profiles of different types of bioreactors used for treating laundry effluent and found differences in the types of metabolic activity (particularly aerobic vs anaerobic pathways), with a general predominance of energy, carbohydrate and amino acid metabolism genes. Analyses like these demonstrate the enzymatic activities occurring within a water or biofilm sample. Another feature of metagenomic sequencing is that it is not specific to bacteria; depending on sample processing techniques used, these sequence data also capture archaeal, eukaryotic and viral genes. This is very important for gaining a comprehensive understanding of the ecological dynamics in a greywater or infrastructure sample. Finally, even when it comes to characterizing taxonomic diversity, metagenomics can often provide greater taxonomic resolution for bacterial species. In our previous study (Keely et al. 2015), we described the genus-level bacterial communities at several different greywater sampling points using 454 sequencing of the 16S rRNA subunit gene. Later in this review, we report preliminary findings from metagenomic shotgun sequencing (using the Illumina MiSeq platform) of selected samples from the same study.

Bacterial communities in different types of greywater and infrastructure

In this section, we describe what is known about the bacterial constituents in different types of greywater and greywater infrastructure. Most available information focuses on a small number of pathogenic or indicator species; however, sequencing-based studies consistently find hundreds to thousands of bacterial species present in samples. Because we believe that there is a great need to understand the immense complexity found in greywater systems, we have chosen to emphasize results from studies that use environmental sequencing to characterize entire bacterial communities. A comparison of the most abundant genera found in different types of greywater by studies that used amplicon or metagenomic sequencing (and reported genus-level results) is provided in Fig. 2. Proteobacteria were the dominant phylum in almost all samples. Laundry samples had a greater representation of Actinobacteria and Firmicutes, including the skin-associated genera Staphylococcus, Corynebacterium and Propionibacterium.

Figure 2.

Bacterial genera found in different types of greywater and greywater infrastructure in studies that use amplicon or metagenomic sequencing and report genus-level data. For each study (column), filled black squares represent the five most abundant genera (rows). Note that this summary includes a variety of sample processing, sequencing and data analysis methods. Studies cited: [1] Keely et al. (2015), [2] This study, [3] Callewaert et al. (2015) [4] Jacksch et al. (2020), [5] Truu et al. (2019), [6] Rodríguez-Martínez et al. (2016), [7] Li et al. (2020), [8] Feazel et al. (2009).

Infrastructure-associated bacteria are present as biofilms, which are formed when bacteria attach to a surface and establish a community connected by an extracellular matrix (López et al. 2010). They are found on all types of natural and human-made surfaces, including premise plumbing, water treatment equipment and bathroom surfaces. A number of studies have explored bacterial growth in premise plumbing (Rogers et al. 1994b; Moritz et al. 2010; Ji et al. 2015), typically with the goal of identifying growths of known drinking water pathogens such as Legionella pneumophila and Pseudomonas aeruginosa. These studies have indicated that bacterial biofilms form on most plumbing regardless of the material, though elastomeric substances often support greater colonization and biofilm formation. This may be because they leach more TOC which can provide a food source for heterotrophic bacteria (Rogers et al. 1994b). Moritz et al. (2010) estimated that most bacterial cells present in drinking water form biofilms on pipe surfaces. In the context of greywater, biofilms are an important consideration because they represent a bacterial community that has persisted in the water infrastructure and may actively contribute to the pretreatment effluent water through cell sloughing. The extent of their contribution, as well as how they differ from other types of greywater, may be useful to understand via community sequencing analysis.

Kitchen water

Greywater from kitchen sources is distinct from other types of greywater in its chemical composition and microbial community, which is unsurprising given its wider range of source inputs. The composition of kitchen greywater is more variable than other types of greywater, as it depends on the specific diets and food-preparation habits (including what is considered acceptable to send down the sink, whether the sink has a garbage disposal, etc.) of every individual household. Certain enteric pathogens, such as Salmonella spp. and Campylobacter spp., may be introduced through washing of contaminated foods. In a chemical analysis of pollutants, Bodnar et al. (2014) found that kitchen greywater had a greater abundance of total solids, organic matter, salts and microelements than shower or laundry water. Absent from this water source were faecal coliforms. However, an earlier study found kitchen water to contain higher levels of faecal coliform bacteria relative to washing machine and shower water (Casanova et al. 2001), and its use for non-potable reuse has been discouraged compared with other greywater sources (Christovaboal et al. 1996). A quantitative microbial risk assessment (QMRA) model based on previously reported E. coli abundances found that the exposure risk of kitchen water reuse (even when treated via microfiltration) made it unsuitable to reuse for crop irrigation, but acceptable for toilet flushing (Shi et al. 2018). In comparison, microfiltered sink, shower and laundry water were deemed suitable for both irrigation and toilet flushing.

Dishwashers represent a type of kitchen infrastructure that has been studied for the presence of opportunistic pathogens. Pseudomonas aeruginosa, Ochrobactrum pseudintermedium, Klebsiella oxytoca and Acinetobacter junii were among the opportunistic pathogens found in the rubber seals of dishwashers (Zupančič et al. 2018). A comparison of different ages of dishwashers suggested Proteobacteria, Actinobacteria and Firmicutes to be ‘early colonizers’ due to their predominance on recently purchased dishwashers (Raghupathi et al. 2018); older dishwashers had a decrease in Proteobacteria and an increase in Actinobacteria. Along with age, frequency of use and water hardness contributed to variation in biofilm community structure between different dishwasher samples.

Shower, sink and handwash water

Many on-site water reuse systems collect shower, sink and handwash water together for treatment, excluding laundry and kitchen waters. Shower, sink and handwash water have a lower pathogen load relative to kitchen water. No Salmonella spp. or Campylobacter spp. were detected in source-separated greywater samples from four Australian homes by Christova-Boal et al. (1996). Another study found high levels of indicator organisms (coliforms and enterococci) in greywater from college dorms, but this did not correspond to detection of the pathogens E. coli O157: H7 and Legionella spp. (Birks and Hills 2007). Faecal coliforms, E. coli, Enterococcus spp. and the pathogens Salmonella enterica, Staphylococcus aureus and P. aeruginosa were all detected in raw greywater from domestic residences by Benami et al. (2015). However, according to a QMRA model based on E. coli abundances, treated bathroom greywater (by microfiltration) was acceptable for both toilet flushing and crop irrigation (Shi et al. 2018).

Showerheads have been considered an interesting environment for studying the microbiome of the built environment because they offer a surface on which biofilms can grow that fluctuates drastically between wet and dry states throughout the day. Feazel et al. (2009) found that mycobacteria, opportunistic pathogens of human health concern, were highly enriched on showerhead biofilms (while they comprised a much smaller proportion of the influent water). Methylobacterium spp., E. coli and Pseudomonas spp. were also found on the showerhead biofilms. Legionella spp., a common pathogen of concern in potable water, was rare in this particular study, although it has been found in infrastructure biofilms in other studies (Rogers et al. 1994a; Murga et al. 2001). Mycobacteria were particularly prevalent in showerheads from American households (compared with European) receiving chlorinated municipal water (compared with well water), indicating that resistance to chlorination might be a selecting factor (Gebert et al. 2018). This group also found that the material of the showerhead made a difference in the composition of its bacterial biofilm, with plastic showerheads being less conducive to the growth of mycobacteria. The mycobacteria finding may present a significant health concern as they have been demonstrated to aerosolize when showers are running and exposure to certain species can be harmful for immunocompromised individuals. In Keely et al. (2015), we found both mycobacteria and methylobacteria to be abundant in the potable water samples, which would have the strongest signature of showerhead biofilm and influent plumbing as they were collected prior to encountering human skin.

Laundry water

Laundry water is enriched in anionic surfactants. In a survey of four sites, Christova-Boal et al. (1996) found laundry water to have a higher pH than shower/handwash water, likely due to greater amounts of detergents present. Laundry water also contained several orders of magnitude fewer total coliforms than shower/handwash water, although the abundances of faecal coliforms and faecal streptococci were in a similar range to the shower/ handwash water. Higher faecal inputs might be expected intermittently due to washing of diapers or soiled garments (Nolde 1999). Ultrafiltered laundry water was deemed acceptable for both toilet flushing and crop irrigation by a QMRA-based on pathogenic E. coli concentrations found in literature (Shi et al. 2018).

Sequencing-based studies, including Keely et al. (2015), have found that laundry water is influenced by skin bacterial communities as well as those endogenous to the clothes laundered. A comparison of the bacterial community in laundry before and after washing in a washing machine found that communities clustered based on sample type (influent water, including reuse of greywater, and on the fabric itself), but there was microbial exchange during the washing cycle (Callewaert et al. 2015). Typical skin and clothes-associated genera like Enhydrobacter, Acinetobacter, Corynebacterium, Staphylococcus and biofilm-forming Pseudomonas were found on an untouched cotton shirt that was laundered with used clothing. The Proteobacteria and Bacteroidetes groups dominated both the pre- and post-wash water, and effluent greywater had comparable or higher numbers of intact bacteria. Staphylococcus and Corynebacterium along with Propionibacterium were the most abundant genera in laundry microbiota from a university athletic facility (Zimmerman et al. 2014; Keely et al. 2015).

A comparison of washing machine biofilms in the detergent drawers and rubber seals found there to be more bacteria in the detergent drawers and more fungi in the rubber seals (Nix et al. 2015). Proteobacteria dominated both bacterial communities, but different genera dominated the biofilms in the detergent drawers (Rhizobium, Ochrobacterium, Methylobacterium, Rhodotorula) and the rubber seals (Acinetobacter, Pseudomonas, Cladosporium). Another study found the detergent drawer to have the greatest species richness of several locations sampled, including the rubber seal and sump (Jacksch et al. 2020). Pseudomonas, Acinetobacter and Enhydrobacter were found to be present in all the washing machine-associated samples; we also found these genera to be abundant in laundry samples in Keely et al. (2015).

Greywater treatment systems

Another consideration in the evaluation of greywater reuse is the communities that may form in the treatment systems themselves. This review describes findings from studies of bacterial communities within various unit processes that take advantage of microbiological processes to metabolize greywater chemical constituents, including excess nutrients and pollutants. Common unit processes with biological activity include filtration units, constructed wetlands, rotating biological contactors, sequencing batch reactors and upflow anaerobic sludge blankets (Oteng-Peprah et al. 2018). Community composition in the final effluent will be strongly influenced by incorporation of pathogen disinfection steps (e.g. chlorination, UV radiation or other advanced oxidation) required to minimize risks for certain end uses (Benami et al. 2015; Jahne et al. 2017). Further study is warranted to better understand both the selection for resistance when disinfection is used and the potential impacts of diverse communities in effluents when advanced treatment is not required (e.g. influence of subsurface drip irrigation on the soil and rhizosphere microbiome).

The bacterial communities present can be indicative of the biological processes occurring within the treatment systems, which include removal of organics and chemicals of concern. Since proper treatment of greywater for reuse also requires removal of pathogens, the growth of potentially harmful species within treatment systems is important to monitor as well. Notably, the use of surrogate or indicator species can be confounding if their growth is differentially impacted by the treatment system compared with the pathogens themselves. The different parts of the treatment system provide different physical environments for bacteria to colonize. For example, a study of a fluid bed reactor designed to remove anionic surfactants from laundry water found a number of laundry water-associated bacterial phyla present in different parts of the treatment system. Interestingly, the various phyla formed distinct communities based on the sampling location within the treatment system (the resin support material, the inner wall and the phase separator). It was clear that the laundry influent impacted the biofilm communities formed when compared with a non-human-influenced control (Macedo et al. 2017). Conversely, another study found that in bark filters, the initial bacterial community in the filters was not strongly impacted by the introduction of artificial greywater (made with a mixture of detergents, oils and dilute sewage), even when sampled over time. Greater community shifts were observed when artificial greywater was introduced to sand or charcoal filters (Dalahmeh et al. 2014).

Bacteria which form biofilms in water use infrastructure may be a major component of the ultimate greywater effluent to be treated. Truu et al. (2019) used 16S sequencing to track the colonization and variations over time in the composition of the biofilm bacterial community of a vertical and horizontal flow filter employing a ‘constructed wetland’ treatment system, which uses microbial processes to treat wastewater. Different bacterial communities were established in the two types of filters, suggesting that the filter material (clay wetland substrate and shale ash sediment in both cases, but with different properties and more highly alkaline in the horizontal filters) was an important factor in selection. The filters had different communities at the initial time points as well, but large shifts occurred in both once greywater was introduced. The vertical filter was dominated by Gammaproteobacteria, Betaproteobacteria and Bacteroides, while the Firmicutes were highly abundant in the horizontal filter (along with Gamma- and Betaproteobacteria). Conversely, Alphaproteobacteria were found to be more abundant in bark, sand and charcoal filters and Betaproteobacteria were relatively less abundant (Dalahmeh et al. 2014). It should be noted that comparing the communities in different studies is difficult as it is known that the influent water has a strong impact on the community that eventually develops on filter biofilms.

Importantly, many treatment systems leverage the metabolic activities of resident bacterial communities to degrade organics from the influent water. In wetland filters, Bernardes et al. (2019) found a correlation between community richness and redox potential and therefore pollution removal. Important metabolic reactions occurring in treatment systems include nitrification, oxidation of organic carbon and removal or immobilization of other nutrients. For example, bacterial communities played a critical role in the degradation of aromatic compounds in laundry wastewater (Macedo et al. 2017). Rodriguez-Martinez et al. (2016) sequenced biofilms at different vertical positions on an unsaturated-flow bioreactor used to treat laundry, handwash and bath water and found a number of genera similar to those found in the equalization tank from Keely et al. (2015): Acinetobacter, Pseudomonas, Propionovibrio, Dechloromonas and Zooglea. The results suggest vertical community stratification, although some genera, including the pathogenic Legionella, appeared in all biofilms. Dalahmeh et al. (2014) found evidence of nitrite oxidation in bark, sand and charcoal filters due to the presence of Nitrobacter spp. and Nitrospira spp. and suggested that Pseudomonas, Acidovorax, Aquabacterium and Rhizobium species may also be responsible for denitrification.

Metagenomics case study: bacterial communities in greywater and greywater infrastructure in a college dormitory

Metagenomics has the potential to help describe communities at greater taxonomic resolution compared with 16S approaches. In Keely et al. (2015), we used 16S sequencing to describe, at the genus level, the bacterial communities present in different locations within the greywater infrastructure in a college dormitory greywater recycling system, where water from the sinks and showers of 14 rooms flows into a larger equalization tank in the basement, where it remains for up to 24 h before treatment. We have since conducted metagenomic sequencing of the shower/handwash, building control, equalization tank and potable water samples (sample information in Table 1; Materials S1). This provided a species-level overview of the bacterial communities. When sequence data are described only at the genus level (as can sometimes be a limitation when 16S sequencing is used), important information about the bacterial community can be lost. Here, we found that some genera contained a wider range of species than others; for example, 140 species or strains of Pseudomonas were detected, but only one species of Tolumonas. In terms of community comparisons, we were able to use the taxonomic information from the metagenomic sequences to arrive at a similar finding as before: samples from different locations in the greywater infrastructure, and in the case of the equalization tank, in different years, were distinct from one another in their community structure, and replicates from the same site clustered together across two axes of variation (Fig. 3). The one exception was the potable water sample, which clustered with the shower/handwash.

Table 1.

Description of greywater samples used for reporting metagenomic sequence data

Name	Description
Potable water/PW	Water collected directly from running showerhead (no contact with humans or greywater infrastructure) in dormitory
Building control/BC	Water collected from simulated showering event in dormitory
Shower/handwash/SH	Effluent water collected from university dormitory restrooms
Equalization tank/ET	Water collected from transient storage tank prior to greywater treatment

All water was collected prior to treatment.

Figure 3.

Principal component analysis of samples from Colorado State University based on taxonomic information in the metagenomics sequence data.

The pathogens and surrogates typically measured in blackwater were present in our samples, along with some less ‘typical’ pathogens. Escherichia coli, P. aeruginosa, S. aureus, Klebsiella pneumoniae, S. enterica Shigella spp., Clostridium perfringens, Vibrio spp., Mycobacterium spp., Enterobacter cloacae, Legionella spp. and Campylobacter jejuni sequences were found in every single sample, though many were at very low abundances. Enterobacter cloacae was the most abundant species in the equalization tank in 2012 and the second-most abundant in 2013 (average 43 and 14% of all sequences, respectively), while its relative abundance remained below 4% for all other samples. This indicates that the equalization tank environment may be conducive to E. cloacae growth, or inhibitory to the growth of the more abundant species from the shower/handwash samples. Conversely, K. pneumoniae was more abundant in the shower/handwash effluent samples than any other sample type, indicating a human origin but perhaps low survival in other environments.

The potable water sample, representing water captured directly from the showerhead without encountering human or greywater plumbing influences, contained multiple species of the opportunistic pathogens mycobacteria and methylobacteria, consistent with the findings of Feazel et al. (2009). This represents an important exposure point for humans to pathogenic bacteria. These two taxa (including multiple species within the Mycobacterium avium complex), known to be abundant on showerhead biofilms, strongly influenced the community of water exiting the showerhead, but the human and plumbing-associated bacterial communities clearly overwhelmed this signal by the time water passed through the greywater infrastructure.

The most abundant species in all three of the building controls was Acidipropionibacterium acidipropionici (Fig. 4). In the original 16S analysis, the genus Propionibacterium was abundant, and we suggested that this was indicative of the contribution of human skin bacteria due to members of the Propionibacterium being associated with the human skin microbiome (Brandwein et al. 2016). However, more recently proposed phylogenetic changes in the Propionibacterium designate a new genus, Acidipropionibacterium, distinct from the skin-associated Cutibacterium (Scholz and Kilian 2016). The majority of the metagenomics sequences from the building control samples were assigned to the Acidipropionibacterium, which contain known fermenters of dairy and have actually been used for fermentation of sewage sludge (Li et al. 2013). Additionally, A. acidipropionici has been reported to produce extracellular polymers (Van Schalkwyk et al. 2003) and can be induced to form biofilms under stress (Cavero-Olguin et al. 2019). Pseudomonas aeruginosa is also known to form biofilms, and indeed, sequences for alginate synthesis and regulatory genes, which are associated with biofilm formation (Hentzer et al. 2001), were found only in building control samples (S. Keely, unpublished data). The building control samples can be used to identify organisms that are forming biofilms on the building infrastructure without having to directly sample plumbing biofilm; there is also the potential to detect biofilm-associated genes. Importantly, sequencing-based descriptions of microbial communities only describe relative abundances of taxa, so while Acidipropionibacterium spp. comprised the largest portion of sequences, this does not correspond to actual abundance relative to any of the other samples. In fact, it might be expected that the building control would contain far fewer cells in general. Cell quantification or qPCR would be necessary to address this question.

Figure 4.

Ten most abundant taxa from each sample collected from Colorado State University. Sizes of circles represent relative abundances. On the x-axis, PW is potable water, BC is building control, SH is shower/handwash and ET 12 and ET 13 are the equalization tank from 2012 and 2013, respectively.

The sequences found in the shower/handwash samples help differentiate the human-associated greywater signal from the non-human-impacted greywater signal associated with the building control samples, which presumably represent the infrastructure. Notably, the building control samples share far more abundant taxa with the shower/handwash samples than with the potable water samples, indicating that the building infrastructure community might be shaped more by human influence than by the plumbing itself. While the shower/handwash samples shared some of their most abundant taxa with the building control (like Aquaspirillum and Acidipropionibacterium), other genera like Acinetobacter and Pseudomonas were present at far greater relative abundances in the shower/handwash than in the building control. Notably, skin-associated genera like Staphylococcus and Micrococcus, which were previously found in laundry and some shower/handwash samples from our site, were not abundant in any shower/handwash samples reported here. These disparate results from two different types of sequencing (16S vs shotgun) illustrate the importance of further understanding biases introduced in sequencing studies.

The equalization tank samples from 2012 and 2013 had very distinct bacterial communities from all the other sampling locations but also between years (Figs 2 and 3). Enterobacter cloacae was the most abundant member of all the 2012 samples, comprising approximately 40% of sequences in all three time-points. This species is known to be facultatively anaerobic (Kumar and Das 2000) and is currently of concern from an antibiotic resistance perspective (Amador et al. 2015). Strains of E. cloacae have demonstrated both nitrification and denitrification capabilities under aerobic conditions (Padhi et al. 2017; Wan et al. 2017). Conversely, the dominant member of the equalization tank in 2013 was Pseudomonas sp. HLS-6, another strain with known antibiotic resistance (Hu et al. 2019). The relative abundance of antibiotic-resistant bacteria, albeit different species, in the equalization tank in both years indicates that this environment is readily amenable to the growth of pathogenic species which have the potential to spread antibiotic resistance through horizontal gene transfer. And indeed, in a preliminary analysis of the metagenomic sequence data among all samples, we found at least 95 different antibiotic resistance genes (ARGs) associated with resistance to at least 24 different antibiotics (S. Keely, unpublished data).

These broad and preliminary findings from a metagenomic sequencing effort of greywater samples illustrate a small fraction of the information that this type of data can provide. With a more in-depth analysis, we will be able to understand entire metabolic pathways, make associations between viruses and bacteria and suggest useful gene targets for understanding treatment.

Future challenges and recommendations for study design

As further research is conducted to explore bacterial communities in greywater, there are certain sampling and methodological considerations that might allow for a more accurate representation of what a ‘greywater bacterial community’ comprises. Firstly, as has been amply demonstrated in the studies discussed, bacterial communities differ widely between different parts of greywater infrastructure, so selection of multiple sampling sites will allow for some partitioning of contributions from different sources. Because of studies demonstrating that bacteria in water can be influenced by biofilms from plumbing, the inclusion of biofilm sampling should be incorporated in studies as much as possible. Another important variable is residence time, which can have an impact on bacterial growth and community composition (Friedler et al. 2006).

Temporal sampling is also imperative to truly partition human- vs infrastructure-associated bacterial communities. Biofilm communities might be expected to maintain greater temporal consistency, whereas human influences might lead to seasonal signatures or differences over shorter time scales in greywater. In fact, the relative consistency between the bacterial consortia associated with the infrastructure alone (including biofilms) and the consortia in greywater may serve as a useful metric for monitoring. Key in this effort is isolating the impacts of different parts of the infrastructure. For example, in the work presented here, a ‘building control’ was measured where water was collected from sinks in the absence of actual human impact (i.e. by simply running water). Furthermore, a different sample type (‘potable water’) was collected prior to water going down drains, representing the influent water and plumbing alone. Greywater filtration systems have been studied temporally from their initial ‘colonization’ by bacterial communities to well into their operations. In a study of horizontal and vertical artificial wetland filtration systems, shifts in the bacterial community were seen when there were changes in circulation or other disruptive events, but these eventually stabilized over periods of weeks or months. A similar trend was seen in different types of filters tested with artificial greywater (Dalahmeh et al. 2014).

Towards an ecological perspective on the greywater microbiome

One reason to better characterize the greywater microbiome is that in some cases, treated effluent is released into the environment, either directly into local surface waters or on land if used for irrigation. This transports the post-treatment microbial community to a new ecosystem. Troiano et al. (2018) found over 20 species of tetracycline-resistant bacteria in greywater, some of which could be detected in greywater-irrigated soil, even when greywater was treated. In another study of bacterial communities in soil that was irrigated using greywater effluent, the community composition and structure was different in the detergent-contaminated soil, dominated by Proteobacteria more than in the untreated soil, which was dominated by Actinobacteria and Acidobacteria (Rojas-Herrera et al. 2015).

Greywater bacteria are part of a broader community of organisms, shaped by their biotic and abiotic environment. In greywater and its infrastructure, organisms compete for carbon, nutrients and (in the case of biofilms) space. Ecological communities, including microbial communities, that exhibit temporal stability can experience gradual or drastic compositional changes when exposed to environmental perturbations. In the case of water infrastructure, changes in operational conditions can lead to shifts in bacterial community structure (Dalahmeh et al. 2014; Tian et al. 2015), which often stabilize if the new conditions persist. These alternative stable states can be reversible or irreversible (Faust et al. 2015). Regardless, this phenomenon indicates a strong need for temporal sampling, as well as further study to determine an appropriate sampling frequency that might reasonably capture changes. One hypothesis in microbial ecology, which could be explored using metagenomics, is that the functional profile of microbial communities in specific environments remains consistent even when the taxonomic structure is highly dissimilar (Burke et al. 2011). Studying communities in this context, like distinct equalization tank communities from 2012 and 2013 in our samples, may reveal greater temporal stability.

In order to truly describe a system, it is important to understand all ecological interactions, including competition and predator–prey interactions. The use of metagenomic sequencing informs this because its results are not limited to bacteria. In greywater environments, bacteria may be in competition with archaea and eukaryotes for organic resources and space (Crone et al. 2020) and are susceptible to predation by protists and infection by phages. For example, on different dishwasher surfaces, Raghupathi et al. (2018) found some biofilms to be dominated by bacteria, others by fungi and others still containing complex communities of both bacteria and fungi. Both the interactions within such a shared community, as well as factors that might lead to the complete dominance of bacteria over fungi (or vice versa), warrant further study, particularly given that many of the involved species were opportunistic pathogens. Studies of viruses in greywater or other types of recycled water demonstrate the presence of both human pathogens, including adenoviruses and noroviruses, and bacteriophages (Jahne et al. 2020; Rusinol et al. 2020).

Wang et al. (2013) have proposed the potential for leveraging knowledge of the microbial ecology within a plumbing system to control resident pathogens. Specifically, they suggest a ‘probiotic approach’ utilizing ecological principles such as competition, antagonism and removal of keystone species in order to drive down populations of undesirable bacteria. A similar approach can be used in treatment systems where specific metabolic processes are desirable to remove organics. Importantly, this perspective embodies an understanding that not all bacteria are harmful and requires a knowledge of the entire microbial community present. Such a picture of the total community is not available for most water systems at this time. As on-site greywater recycling becomes more widely adopted, it will be important to understand these types of dynamics in order to monitor its suitability for reuse and adjust treatment as needed.

Conclusions

1. There is no one ‘greywater’—there are very distinct communities in different types of greywater (Figs 1–3) and between the water and the infrastructure within a single greywater system. Further knowledge is required on the bacterial communities present in all the different types of source-separated greywater.

2. Bacterial communities in greywater may experience variability over time, including large compositional changes, particularly as environmental conditions shift with transport, storage and treatment. While many studies have found that greywater communities, including colonized biofilms or filter systems, stabilize over time, temporal monitoring is still necessary. Metagenomics will allow us to think about greywater communities beyond just their taxonomic makeup and rather in terms of ‘functional guilds’ which may inform design of treatment systems.

3. Moving forward, it is important to supplement culture-based methods with approaches that describe the entire greywater microbial community. This includes other microbial constituents (archaea, viruses, protozoa), of which the latter two represent significant risks that must be accounted for in models. Amplicon sequencing is useful for understanding the extent of taxonomic diversity, and metagenomics is the most comprehensive method to gain information about both the taxa present and the processes occurring. Sequencing techniques can be used to identify and validate new surrogates (like Staphylococcus or other skin-associated genera) for assessing treatment, monitor other ‘contaminants’ like virulence factors and ARGs and better understand genes or taxa which may be responsible for certain metabolic or chemical and pollutant remediation processes.