Domestic shower hose biofilms contain fungal species capable of causing opportunistic infection

Water supplies in domestic environments have been the focus of studies into transmission of opportunistic infections with environmental organisms, usually in people who have a compromised immune system. It is known that environmental organisms, including Legionella can become established in shower hoses (Tobin et al. 1980; Bauer et al. 2008; Whiley et al. 2015), subsequently forming aerosols and being inhaled leading to infection (Pedro-Botet et al. 2002). A number of studies, using culture-based and culture-independent methods, have attempted to characterise the organisms present in these aerosols and the surrounding environment, demonstrating the presence of potentially pathogenic biofilm forming Gram-negative bacteria in hospital water systems (Decker & Palmore 2013). Culture-independent methods have been used to demonstrate the presence of potentially pathogenic bacteria on domestic shower curtains (Kelley et al. 2004) and showerheads (Feazel et al. 2009). Likewise, recent molecular and culture-based work has demonstrated the presence of non-tuberculous Mycobacteria (NTM) in domestic water (Thomson et al. 2013a, 2013b) and demonstrated infection acquired from aerosols of this water (Falkinham et al. 2008; Thomson et al. 2013a, 2013b). O'Brien et al. (2000) have suggested that our increased tendency to shower has contributed to a rise in NTM disease.

In contrast to the significant body of knowledge surrounding the bacterial components of domestic water system biofilms, there is currently a relatively limited understanding of the fungal members of these communities and none of the previous studies of shower systems have included detailed analysis of fungal species.

We have used culture-independent methods to investigate the bacterial and fungal diversity of biofilms recovered from domestic shower hoses, revealing the presence of several genera or species that represent a risk to immuno-compromised hosts.

A total of four hose samples were examined: one hose was recovered from a hotel in Northwest England (NW; latitude 53.414 °N, longitude −2.124 °W) and a further three from domestic settings in Scotland (Fife; 56.135 °N, −3.376 °W) and in the Southeast of England (SE1; 51.167 °N, 1.289 °W & SE2; 51.424 °N, 0.560 °W). The SE1 hose was recovered from a shower that had been installed less than a month before removal of the hose. Other hoses had been in place for an undetermined, but extended time. All samples were obtained with the full consent of the property owners, without restriction.

All wet hoses were filled with tap water at source, their ends being sealed with sterile screw caps, before being transported to Manchester at ambient temperature. On arrival, hoses were stored at 4°C and were analysed within 48 hours.

Hoses were surface sterilised with ethanol, aseptically sectioned into 7 cm lengths and the biofilm physically removed using a sterile scraper and recovered into sterile phosphate buffered saline. Biofilm material was recovered from the sections of hose closest to each end of the hose. Samples were mechanically disrupted by vigorous mixing for 2 minutes in the presence of 2 g of sterile coarse sand and 3 mm diameter glass beads. Total DNA was recovered using the MoBio Soil extraction kit (Cambio Ltd, UK) and stored at −20°C.

Polymerase chain reaction (PCR) of the V1-V2 hypervariable region of the bacterial 16S rRNA gene was performed using tagged fusion universal bacterial primers 27F (Lane 1991) and 338R (Hamady et al. 2008), synthesised by IDTdna (Integrated DNA Technologies, Belgium). The fusion forward primer (5′ CCATCTCATCCCTGCGTGTCTCCGACTCAGNNNNNNNNNNAGAGTTTGATGMTGGCTCAG 3′) contained the 454 Life Sciences ‘Lib-L Primer A’, a 4 base ‘key’ sequence (TCAG), a unique ten-base barcode ‘MID’ sequence for each sample (N), and bacterial primer 27F (underlined). The reverse fusion primer (5′ CCTATCCCCTGTGTGCCTTGGCAGTCTCAGTGCTGCCTCCCGTAGGAGT 3′) contained the 454 Life Sciences ‘Lib-L Primer B’, a 4 base ‘key’ sequence (TCAG), and bacterial primer 338R (underlined). PCR of the fungal 18S rRNA small subunit was carried out using tagged forward fusion primer FU18S1 (5′-GGAAACTCACCAGGTCCAGA-3′) and reverse fusion primer Nu-SSU-1536 (5′-ATTGCAATGCYCTATCCCCA-3′) (Gangneux et al. 2011).

The PCR amplification was performed in 50 μl volume reactions using 0.5 μl (2.5 units) FastStart High Fidelity DNA polymerase (Roche Diagnostics GmbH, Mannheim, Germany), 1.8 mM MgCl₂, 200 μM of each dNTP, 0.4 μM of each forward and reverse fusion primers. The PCR conditions included an initial denaturing step at 95°C for 2 min, followed by 35 cycles of 95°C for 30 sec, 55°C for 30 sec, 72°C for 45 sec, and a final elongation step at 72°C for 5 min.

The 16S and 18S rRNA gene fragments were loaded onto an agarose gel and, following gel electrophoresis, bands of the correct fragment size (approximately 410 bp) were excised, purified using a QIAquick gel extraction kit (Qiagen, GmbH, Hilden, Germany) and eluted in 30 μl of DNAse free H₂O. The purified PCR products were quantified using a Life Technologies Qubit 2.0 fluorimeter and the fragment size confirmed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Inc., Santa Clara, CA, USA). PCR products were normalised and pooled at equimolar concentrations. The emulsion PCR and the pyrosequencing runs were performed at the University of Manchester DNA Sequencing Facility, using a Roche 454 Life Sciences GS Junior system (454 Life Sciences, Branford, CT, USA).

The 454 pyrosequencing reads were analysed using QIIME release 1.8.0 (Caporaso et al. 2010). De-noising and chimera removal was performed in QIIME during operational taxonomic unit (OTU) picking (at 97% sequence similarity) with usearch (Edgar 2010). Taxonomic classification of the 16S rRNA bacterial reads was performed in QIIME using the uclust method (Edgar 2010) against the Greengenes database, release 08/2013 (McDonald et al. 2012). Taxonomic classification of the 18S rRNA fungal reads was carried out in QIIME using uclust against the Silva release 111 database (Yilmaz et al. 2014). In addition, the closest GenBank match for the OTUs (a representative sequence for each OTU was used) that contained the highest number of reads was identified by Blastn nucleotide search (Altschul et al. 1990). OTU clustering was carried out on the log-transformed abundances of the identified OTUs, using the Bray Curtis similarity method. Cluster analysis was carried out using the PRIMER v6 software (PRIMER-E, Plymouth, UK) (Clarke 1993).

All raw sequence data relating to this work have been deposited at the Sequence Read Archive under BioProject number PRJNA267447.

Samples NW and Fife were characterised by the presence of sequences closely related to two Mycobacterium gordonae strains isolated from gastric lavage (25.2% and 16.2% of all reads for each sample, respectively; Table 2). In contrast, 11.6% and 1% of the reads from the SE2 and SE1 samples, respectively, were more closely related to M. mucogenicum N248 (Table 2).

Bacteroidetes related sequences were not observed in samples SE1 or SE2. In the NW sample, these were mainly related to Hydrotalea flava, recently described in Swedish water samples (Kämpfer et al. 2011), and in the Fife sample were affiliated to Lacibacter cauensis (Table 2), another recently described aquatic bacterium (Qu et al. 2009).

The fungal populations of samples SE1 and SE2 displayed similar compositions (Table 3). Approximately 21% of the reads in these samples were closely related to Malassezia restricta CBS 7877 (EU192367). Additional sequences (9%) were related to Malassezia sympodialis (EU192369).

Plant pathogen Leptosphaeria maculans JN3 (NW_003533867) was represented by 17% of sequences. In sample SE2, 29% of the reads were affiliated to various Penicillium species (Table 3). Other fungal species detected in one or more of the samples of this study included airborne Cryptococcus vishniacii, Aspergillus glaucus, Cladosporium uredinicola CPC 5390, Trametes sanguinea and wood-rotting Schizophyllum commune, an emerging cause of sinusitis (Lorentz et al. 2012) (Table 3).

The environment around us hosts a significant number of microbes and, in domestic settings, it is suggested that the majority are of human origin (Hospodsky et al. 2012). Previous studies have highlighted the importance of waterborne Legionella species in domestic and healthcare settings (Fields et al. 2002; Leoni et al. 2005; Decker & Palmore 2013), but there is now growing interest in other bacteria that may infect immuno-compromised hosts. In the current study, we have used pyrosequencing approaches to characterise the established bacterial and fungal inhabitants of domestic shower hoses from four sites in England and Scotland, demonstrating the presence of bacteria and fungi that may represent a threat to immuno-compromised hosts, and certain fungi that can be more overtly pathogenic. Although limited in sample numbers, this work was carried out to give an indication of diversity over a geographically dispersed area, supplied by different water sources, rather than to generate a comprehensive catalogue of microbial diversity in this environment.

Amplicon pyrosequencing revealed a maximum of 79 bacterial and 36 fungal OTUs per sample, with most of the sequences clustered within a small number of OTUs. These relatively low levels of diversity in the samples examined is similar to that reported for showerhead communities in the USA and probably results from the low nutrient environment that is present in the water supplying the hoses (Feazel et al. 2009).

Cluster analysis and taxonomic classification of all the microbial communities revealed that two geographically distant samples (NW and Fife) shared some similarities and a number of bacterial and fungal species were present across all samples (in some cases in abundance). Environmental Alphaproteobacteria, Actinobacteria and Bacteroidetes dominated the bacterial communities. Several examples of the Sphingomonadaceae were observed (e.g. Blastomonas, Novosphingobium, Sphingopyxis, Sphingomonas and Porphyrobacter). Fungal sequence reads were mainly related to the Eurotiomycetes, Exobasidiomycetes, Dothideomycetes, Sordariomycetes and Agaricomycetes. This may indicate that specific organisms preferentially colonise and establish within the biofilm communities that develop over time within domestic shower hoses.

The differences in the bacterial and fungal communities observed in the samples could be attributed to a range of factors, such as water composition in the different regions; water in Northwest England and Fife is soft, whereas it is generally medium-hard to hard in Southeast England (waterwise.org.uk). A larger and more detailed analysis of the impact of different physicochemical parameters of water on shower biofilm populations is warranted, though the similarities seen between the samples examined here suggest that some genera are widely dispersed, irrespective of water characteristics.

Members of the above-mentioned bacterial phyla have been described in shower associated biofilms (Kelley et al. 2004; Feazel et al. 2009) and are only very rarely associated with infections. However, some form extensive biofilms and it is suggested that waterborne members of the Sphingomonadaceae from hospital and domestic environments may be a reservoir for antibiotic resistance genes (Narciso-da-Rocha et al. 2014). Porphyrobacter donghaensis has been associated with environmental and hospital water in previous studies (Yoon et al. 2004; Furuhata et al. 2007), but not with infection.

Samples contained sequences related to Mycobacterium gordonae and M. mucogenicum. M. gordonae has been reported in showerhead populations (Feazel et al. 2009). Members of the NTM are ubiquitous in the environment and have recently been associated with shower aerosol related infection in Australia where strains of NTM from patients were linked to their domestic environments (Thomson et al. 2013a, 2013b). In a different study of showerheads and hose samples, Rhodes et al. (2014) demonstrated the widespread occurrence of Mycobacterium avium subspecies paratuberculosis (Map), which is significantly associated with development of Crohn's disease. Their findings supported the suggestion that shower aerosols may have an under-recognized role in the acquisition of Crohn's disease.

Although other molecular- or culture-based studies have identified Legionella species in domestic water systems (Leoni et al. 2005; Bauer et al. 2008; Whiley et al. 2015), in the current study, no such sequences were detected. Interestingly, a similar study reported only low-level occurrence of Legionella species 16S rRNA sequences (0.1% of total sequences), while qPCR was unable to detect L. pneumophila specific sequences in any samples (16 from water and 20 swabs) recovered from domestic showerheads (Feazel et al. 2009). Taken together, this may indicate that (pathogenic or non-pathogenic) Legionella species are not expected to be universally present in all domestic water systems.

This study reports, for the first time, the fungal flora of shower associated biofilms. In some cases, there was a predominance of Exophiala mesophila, an organism associated with cutaneous and subcutaneous infections (Zeng et al. 2007), was first isolated from silicone seals in a hospital shower room (Listemann & Freiesleben 1996) and has been reported in dental unit water lines (Porteous et al. 2003). Exophiala species have been recovered from drinking water in Germany (Göttlich et al. 2002) and Slovenia, where the occurrence of these organisms has been studied in detail (Novak Babič et al. 2016). Exophiala species are known for biofilm formation (Heinrichs et al. 2013) and they have been observed in high numbers in sauna facilities (Matos et al. 2002), domestic dishwashers (Gümral et al. 2015) and washing machines (Novak Babič et al. 2015). In the latter environments, conditions like temperature and pH will be different to those found in showers and there will be a plentiful supply of nutrients in the form of detergents, which may support dense biofilm development. Shower hoses are low nutrient environments, so Exophiala phenotypes from these sources would be expected to differ and the role of other flora in succession and biofilm development may be key in shower hoses. Although Exophiala species rarely cause infections, these can be difficult to treat (Rimawi et al. 2013) and their true prevalence may be underestimated given their slow growth and difficulties in identification (Porteous et al. 2003). In studies of domestic washing machines, which can be heavily colonised with these fungi, it has been suggested that detergent drawers should be regularly cleaned with bleach, in line with manufacturers instructions to reduce the possibility of opportunistic infection (Novak Babič et al. 2015).

Also observed were sequences related to Fusarium fujikuroi. Members of the Fusarium fujikuroi species complex are predominantly plant pathogens, but they are being increasingly recognized as aetiological agents of infection in immunocompromised and immunocompetent hosts (Kebabci et al. 2014; Tortorano et al. 2014) and infection with some Fusarium species has been linked to colonization of hospital water systems (Anaissie et al. 2001). It has been demonstrated that the concentration of airborne of filamentous fungi, including Fusarium and Aspergillus species, was significantly reduced by washing water-related surfaces in hospital bathrooms immediately prior to showering (Anaissie et al. 2002). These authors also suggested that the findings reported would be applicable to domestic shower environments, where discharged immunocompromised patients may be at risk of fungal infections.

An additional potential risk with Fusarium species is the generation of trichothecenes, protein synthesis inhibitors that are documented to cause intoxications in humans following consumption of food products colonised with Fusarium species (Murphy et al. 2006) and may lead to an increase in respiratory tract infections (Bhat et al. 1989). Given the toxicity of trichothecenes, concentrations in water emerging from showers should be investigated.

In samples from the Southeast of England, over 20% of the reads were related to Malassezia restricta, an organism associated with severe opportunistic infections (Arendrup et al. 2014), although they are also members of skin flora. However, given recent evidence for an association between Malassezia restricta and dandruff and seborrheic dermatitis (Gemmer et al. 2002; De Angelis et al. 2005; Clavaud et al. 2013), the implications of significant numbers of these organisms in the shower hose biofilms is clear. Overall, members of the Malassezia genus are the predominant flora of many parts of the human skin, where they have a complex interaction with other microorganisms and the host (Gaitanis et al. 2012). When these interactions are disturbed, Malassezia species can cause a wide range of conditions; they are among the most common causes of superficial mycoses (Crespo-Erchiga & Florencio 2006) and dermatoses (Jagielski et al. 2014) and are also associated with atopic eczema and allergic reactions (Gaitanis et al. 2012).

Although sample numbers are small, there are suggestions of geographical and temporal associations for some flora. Agaricomycetes are only seen in the older southern sample, SE2, and the Sordariomycetes are only seen in high numbers in Fife, while the Dothideomycetes and Exobasidiomycetes were predominant in the South. Similarly, Actinobacteria are only present in small numbers in sample SE1 and Bacteroidetes are not seen in the southern samples at all. The findings warrant more detailed and extensive studies into these associations, particularly with a focus on fungal flora and exploration of the dynamics of community succession in the biofilm.

The UK Department of Health suggests that flexible hoses should not be used in high-risk situations (Best Practice Guidance; HTM 04-01–Addendum: Pseudomonas aeruginosa–Advice for Augmented Care Units, March 2013). Such guidelines are implemented to reduce infection with Pseudomonas aeruginosa, but the data presented here suggests that effective management of biofilms in flexible hoses may control several additional opportunistic bacterial and fungal pathogens.

Previous studies have extensively catalogued the bacterial populations associated with showerheads and shower curtains, and our data corroborate these findings. Ours is the first study to describe in detail the fungal community members, some of which give cause for concern. Although our findings from a small collection of samples are preliminary, more detailed investigations are justified. The role of different shower water supplies (e.g. tank vs mains feed) and shower hose materials should be established. The impact of frequency of shower use on biofilm populations and the rate of biofilm accumulation should be investigated. We support the suggestion in previous studies that potential pathogens residing in shower hose biofilms will periodically slough off the inner surface, being released and distributed in shower aerosols (Falkinham et al. 2008; Rhodes et al. 2014). A recent study with a model shower system demonstrated that drying the hose between uses did not reduce the load of bacterial pathogens including Legionella species and members of NTM (Whiley et al. 2015). It may be prudent to routinely decolonise or regularly replace shower hoses in domestic and health-care settings, where immuno-compromised individuals are present.

This work was funded in part by Omnia-Klenz Ltd, who also supplied the hose samples for analysis.

All DNA sequence analysis was carried out at the University of Manchester DNA Sequencing Facility.