Molecular analysis of single room humidifier bacteriology

Abstract

Portable, single-room humidifiers are commonly used in homes for comfort and health benefits, but also create habitats for microbiology. Currently there is no information on home humidifier microbiology aside from anecdotal evidence of infection with opportunistic pathogens and irritation from endotoxin exposure. To obtain a broader perspective on humidifier microbiology, DNAs were isolated from tap source waters, tank waters, and biofilm samples associated with 26 humidifiers of ultrasonic and boiling modes of operation in the Front Range of Colorado. Humidifiers sampled included units operated by individuals in their homes, display models continuously operated by a retail store, and new humidifiers operated in a controlled laboratory study. The V1V2 region of the rRNA gene was amplified and sequenced to determine the taxonomic composition of humidifier samples. Communities encountered were generally low in richness and diversity and were dominated by Sphingomonadales, Rhizobiales, and Burkholderiales of the Proteobacteria, and MLE1-12, a presumably non-photosynthetic representative of the cyanobacterial phylum. Very few sequences of potential health concern were detected. The bacteriology encountered in source waters sampled here was similar to that encountered in previous studies of municipal drinking waters. Source water bacteriology was found to have the greatest effect on tank water and biofilm bacteriology, an effect confirmed by a controlled study comparing ultrasonic and boiler humidifiers fed with tap vs. treated (deionized, reverse osmosis, 0.2 μm filtered) water over a period of two months.
Copyright © 2014 Elsevier Ltd. All rights reserved.

Full text

Molecular analysis of single room humidifier
bacteriology
Natalie M. Hull
a
, Abigail L. Reens
a
, Charles E. Robertson
a
,
Lee F. Stanish
a
, J. Kirk Harris
b
, Mark J. Stevens
b
, Daniel N. Frank
b
,
Cassandra Kotter
b
, Norman R. Pace
a,*
a
Dept. of Molecular, Cellular, and Developmental Biology, University of Colorado, Boulder, CO 80309-0347, USA
b
Dept. of Pediatrics, University of Colorado School of Medicine, Aurora, CO, USA
article info
Article history:
Received 7 April 2014
Received in revised form
14 November 2014
Accepted 15 November 2014
Available online 26 November 2014
Keywords:
Portable humidifiers
16S ribosomal RNA sequences
V1V2 region
Illumina MiSeq
Phylogenetic analysis
Microbiology of the built
environment
abstract
Portable, single-room humidifiers are commonly used in homes for comfort and health
benefits, but also create habitats for microbiology. Currently there is no information on
home humidifier microbiology aside from anecdotal evidence of infection with opportu-
nistic pathogens and irritation from endotoxin exposure. To obtain a broader perspective
on humidifier microbiology, DNAs were isolated from tap source waters, tank waters, and
biofilm samples associated with 26 humidifiers of ultrasonic and boiling modes of opera-
tion in the Front Range of Colorado. Humidifiers sampled included units operated by in-
dividuals in their homes, display models continuously operated by a retail store, and new
humidifiers operated in a controlled laboratory study. The V1V2 region of the rRNA gene
was amplified and sequenced to determine the taxonomic composition of humidifier
samples. Communities encountered were generally low in richness and diversity and were
dominated by Sphingomonadales,Rhizobiales, and Burkholderiales of the Proteobacteria, and
MLE1-12, a presumably non-photosynthetic representative of the cyanobacterial phylum.
Very few sequences of potential health concern were detected. The bacteriology encoun-
tered in source waters sampled here was similar to that encountered in previous studies of
municipal drinking waters. Source water bacteriology was found to have the greatest effect
on tank water and biofilm bacteriology, an effect confirmed by a controlled study
comparing ultrasonic and boiler humidifiers fed with tap vs. treated (deionized, reverse
osmosis, 0.2 mm filtered) water over a period of two months.
©2014 Elsevier Ltd. All rights reserved.
1. Introduction
Nearly 10 million humidifiers are purchased in the US each
year (ConsumerReports.org 2013), primarily to improve
household comfort and for health purposes. Modes of
operation of household humidifiers include boiler (produces
steam via a heating element), cool mist (produces moisture
through an impeller or evaporation), and ultrasonic (produces
mist through vibration) (USEPA, 1991). Use of humidifiers can
reduce static electricity and ease breathing difficulties such as
runny noses associated with colds and shortness of breath
*Corresponding author. Tel.: þ1 303 735 1864; fax: þ1 303 492 7744.
E-mail address: nrpace@colorado.edu (N.R. Pace).
Available online at www.sciencedirect.com
ScienceDirect
journal homepage: www.elsevier.com/locate/watres
water research 69 (2015) 318e327
http://dx.doi.org/10.1016/j.watres.2014.11.024
0043-1354/©2014 Elsevier Ltd. All rights reserved.

associated with bronchitis (Trenchs Sainz De La Maza et al.
2002). Humidifiers can be used to keep indoor relative hu-
midity in the optimal range of 40e60%, which has been shown
to decrease survival of dust mites, viruses, and microorgan-
isms, and to limit off-gassing of building materials (e.g.
formaldehyde) (Arundel et al. 1986; Mohan et al. 1998; Trenchs
Sainz De La Maza et al. 2002; Myatt et al. 2010).
On the other hand, humidifiers aerosolize their tank wa-
ters, potentially along with any attendant particulates and/or
microbial constituents present in that water or in biofilms that
develop on humidifier tanks, reservoirs, and spouts. Infre-
quent cleaning, as practiced by the 40% of Americans who
rarely or never clean their humidifiers (ConsumerReports.org
2013), has been shown to be associated with increased risk
of adverse respiratory symptoms such as “humidifier fever”
and “humidifier lung”(Muller-Wening et al. 2006), or extrinsic
allergic alveolitis and hypersensitivity pneumonitis
(Nakagawa et al. 1995; Yamamoto et al. 2002; Muller-Wening
et al. 2006; Lacasse et al. 2012). Specific opportunistic patho-
gens that have linked humidifiers with respiratory syndromes
are the yeast Debaryomyces hansenii (Yamamoto et al. 2002),
and bacteria including Elizabethkingia meningosepticum
(Nakagawa et al. 1995), Pseudomonas spp. (Forsgren et al. 1984;
Rylander et al. 1984), Mycobacteria spp. in immune-
compromised individuals (Lacasse et al. 2012) and Legionella
spp. (Moran-Gilad et al. 2012). Endotoxins produced by mi-
crobes can be present in tap waters and in air treated by hu-
midifiers at concentrations that are capable of inducing
respiratory distress (Rylander et al. 1984; Ohnishi et al. 2002;
Anderson et al. 2007). Particulates and mineral dusts can
also contribute to lung irritation and injury resulting from
humidifier exposure (Daftary et al. 2011; Umezawa et al. 2013).
Additionally, lung injury and respiratory failure have resulted
from inhalation of commercially available chemical humidi-
fier disinfectants (Hong et al. 2014).
Despite the contribution of microbes present in humidi-
fiers to various health conditions, the microbiology of hu-
midifiers and the sources of any such microbiology are not
known. Here, Illumina sequencing of the V1V2 region of the
SSU rRNA gene and phylogenetic analyses were used to
characterize bacterial microorganisms present in single room
humidifiers and their feed waters in relation to variables such
as mode of operation of humidifier, microenvironment within
the humidifier, and source water type. This is the first culture
independent (sequence-based) survey of the microbial con-
stituents of home humidifiers.
2. Materials and methods
2.1. Sample collection
Samples for analysis were collected over a two-month period
near the end of the high-use winter season within 25 km of the
city Boulder, CO. Samples were taken from three usage cate-
gories: (1) in-use humidifiers owned by volunteer residents in
the Colorado Front Range, (2) in-use humidifiers on display at
a retail store, and (3) new humidifiers purchased for controlled
experiments. Private and display model humidifiers were
sampled once, while humidifiers purchased for the controlled
study were sampled every two weeks for two months, starting
with the day they were first filled. The new humidifiers were of
two modes of operation, two boilers and two ultrasonic. One
of each mode of operation was fed with deionized/reverse
osmosis/0.2 mm filtered water (DI/RO), and the other was fed
with tap water.
For each sampling event, three sample types were
collected: (1) water from the tank and reservoir of the hu-
midifier and (2) source water used to fill the humidifier were
collected into clean 1 L HDPE Nalgene bottles; (3) swabs of
observable biofilm from any component of the humidifier
were collected using sterile culture swabs and stored in 1 mL
of 1X PBS in 2 mL sterile polypropylene tubes at "80 #C. Water
samples were immediately returned to the lab and filtered
onto 0.2 mm pore polycarbonate filters to capture microbial
constituents, which were stored at "80 #C until further pro-
cessing. Water collection bottles were soaked for 30 min in
sodium percarbonate (B-Brite) dissolved in 0.2 mm filtered DI/
RO water, and then rinsed twice with fresh 0.2 mm filtered DI/
RO water and allowed to dry prior to sampling. We were un-
able to collect credible aerosol samples because of dilution of
aerosol output with ambient air. Additionally, we did not
collect information to discern whether microbes encountered
in this study were alive or active. Nonetheless, the function of
humidifiers is to aerosolize tank water, so humidifier output is
expected to carry portions of the microbiology encountered.
2.2. Cell counts
For cell counts, 45 mL of each water sample was added to 5 mL
of 37% formaldehyde and stored at 4 #C. Formaldehyde-fixed
cells were filtered onto black 0.2 mm polycarbonate filters,
stained with propidium iodide, and counted by direct fluo-
rescence microscopy (Matsunaga et al. 1995). Cell counts were
not conducted for biofilm swabs.
2.3. DNA extraction
For purification of genomic DNA from water samples, poly-
carbonate filters were removed from frozen storage tubes and
placed into 2 mL polypropylene DNA extraction tubes con-
taining: 500 mL of 1:1 phenol:chloroform, 500 mL lysis buffer
(71.5 mM NaCl, 71.5 mM TRIS pH 8.0, 7.15 mM EDTA, 2.85% SDS
dissolved in DEPC-treated sterile water (Fisher)), and 250 mL of
a slurry of 0.1 mm silica/zirconium beads (Biospec Products
Inc.) in lysis buffer. For purification of bulk genomic DNA from
biofilm samples, frozen storage tubes containing swabs stored
in PBS were thawed and vortexed vigorously for 1 min. The top
500uL of the aqueous phase was transferred to a DNA
extraction tube. Extraction tubes were subjected to mechani-
cal bead-beating for 2 min in a 16-channel bead-beater (Min-
ibeadbeater 16 # 607, Biospec Products) to lyse cells, during
which time filters completely dissolved. Tubes were centri-
fuged for 5 min at 16,000 $g. The top 450 mL of the aqueous
phase was transferred to a sterile 1.5 mL polypropylene tube.
DNA was precipitated by addition of 10 mL 10 mg/mL glycogen,
200 mL 7.5 M ammonium acetate, and 650 mL isopropanol.
Tubes were centrifuged for 30 min at 16,000 $gto pellet DNA
and the supernatant was decanted. Pellets were washed with
1 mL 70% ethanol and centrifuged 5 min at 16,000 $g, after
water research 69 (2015) 318e327 319

which the supernatant was decanted. Pellets were allowed to
air dry overnight before resuspension in 30 mL TE (10 mM Tris,
pH 8.0, 1 mM EDTA).
2.4. DNA quantification
DNA was stained with Quant-it PicoGreen fluorescent dye
(Invitrogen). Fluorescence of DNA samples was measured
using a fluorospectrometer (NanoDrop ND-3300) and
compared to a standard curve constructed using known con-
centrations of lDNA suspended in TE. DNA was checked for
amplification prior to sequencing via 30 cycle PCR using 1 mL
each of 515F and 1391R primers (1 mg/mL), 10 mL hot mix
(HotMasterMix, 2.5X, 5Prime), 2 mL 10 mg/mL bovine serum
albumin, and 10 mL DEPC-treated sterile water (Fisher) per
25 mL reaction. Amplified samples were loaded on 1% agarose
Tris-Borate-EDTA gels stained with Ethidium Bromide. Elec-
trophoresis was performed at 150 V for 30e60 min. Quantita-
tive PCR was used to determine bacterial loads in DNA
samples as described previously (Nadkarni et al. 2002). DNA
concentrations are not compared between water samples and
biofilm swabs, or between individual biofilm swab samples,
because biofilm collection was not normalized by mass or
volume.
2.5. 16S amplicon library construction and Illumina
sequencing
Bacterial profiles were determined by broad-range amplifica-
tion and sequence analysis of 16S rRNA genes following pre-
viously described methods (Hara et al. 2012; Markle et al.
2013). In brief, amplicons were generated using primers that
target approximately 300 base pairs of the bacterial V1V2
variable region of the 16S rRNA gene. PCR products were
normalized based on agarose gel densitometry, pooled, puri-
fied, and concentrated using a DNA Clean and Concentrator
Kit (Zymo, Irvine, CA). Pooled amplicons were quantified using
a Qubit Fluorometer 2.0 (Invitrogen, Carlsbad, CA). The pool
was diluted to 4 nM and denatured with 0.2 N NaOH at room
temperature. The denatured DNA was diluted to 15pM and
spiked with 25% of Illumina PhiX control DNA prior to loading
the sequencer. Illumina paired-end sequencing was per-
formed on the MiSeq platform with version 2.2.0.2 of the
MiSeq Control Software and version 2.2.29 of MiSeq Reporter,
using a 500 cycle version 2 reagent kit.
2.6. Analysis of Illumina paired-end reads
As previously described (Markle et al. 2013), Illumina MiSeq
paired-end sequences were sorted by sample via barcodes in
the paired reads with a python script. Sorted paired end
sequence data were deposited in the NCBI Short Read Archive
under accession number SRP032423. The sorted paired reads
were assembled using phrap (Ewing et al. 1998; Ewing and
Green, 1998). Pairs that did not assemble were discarded.
Assembled sequence ends were trimmed over a moving
window of 5 nucleotides until average quality met or exceeded
20. Trimmed sequences with more than 1 ambiguity or shorter
than 200 nucleotides (nt) were discarded. Potential chimeras
were identified with Uchime (usearch6.0.203_i86linux32)
(Edgar et al. 2011) using the Schloss (Schloss et al. 2011) Silva
reference sequences and removed from subsequent analyses.
Assembled sequences were aligned and classified with SINA
(1.2.11) (Pruesse et al. 2012) using the 629,125 bacterial se-
quences in Silva 111Ref (Quast et al. 2013) as reference
configured to yield the Silva taxonomy. Operational taxo-
nomic units (OTUs) were produced by clustering sequences
with identical taxonomic assignments.
2.7. Analysis of taxonomic data
After removal of unclassified (7698) and singleton (85) se-
quences, this process generated 1,967,511 sequences for 86
samples (average library size: 22,878 sequences/sample;
minimum: 7199; maximum: 58,465) of average length 288 nt.
In addition to Microsoft Excel (2007), the software package
Explicet (v2.10.2, www.explicet.org)(Robertson et al. 2013) was
used for display, analysis (rarefacted values for species
observed (Sobs), Good's, Shannon diversity, Morisita-Horn
(Magurran, 2004)), and figure generation of results. SigmaPlot
(v11.0, (SystatSoftware, 2008)) was used to perform t-tests,
substituted with ManneWhitney Rank Sum Tests when data
being compared were not normally distributed, to analyze
differences in richness and diversity between usage categories
or sample types. The vegan package (v2.0-9, (Oksanen et al.
2013)) in R (v3.0.3, (RCoreTeam, 2014)) using RStudio
(v0.98.501, (RStudio, 2014)) was used to perform ordinations on
the dataset and subsets of the dataset via Non-Metric Multi-
dimensional Scaling (NMDS) (Shepard, 1962; Clarke, 1993).
Before ordination, the OTU table was rarefied to the minimum
of 7199 sequences per sample and transformed using a Hel-
linger transformation (square root of percent abundance) to
reduce noise, followed by calculation of a BrayeCurtis dis-
tance matrix used in a 3-dimensional NMDS ordination. The
resulting ordination plot represents compositional similarity
between samples by proximity, where more similar samples
are located closer together. Hulls were drawn to enclose
samples of the same sample type or humidifier type, or those
sharing the same source water. The vegan package in R was
used to perform permutational multivariate analysis of vari-
ance (permanova) (Anderson, 2001) to determine the effect of
sample type and humidifier type on microbiology in store and
home humidifier samples using a BrayeCurtis distance matrix
calculated from a rarefied, Hellinger-transformed OTU table.
The vegan package in R was also used to perform a Mantel test
(Legendre et al. 2012) on the controlled study data to deter-
mine the temporal relationship with microbiology using the
longitudinal data. The Mantel test used two distance matrices:
a BrayeCurtis distance matrix was calculated from a rarefied,
Hellinger-transformed OTU table; and a Euclidian distance
matrix calculated based on time since sampling began for
each sample.
3. Results
3.1. Study overview
Three types of humidifier samples: (1) source waters used to
feed humidifiers, (2) humidifier tank waters, and (3)
water research 69 (2015) 318e327320

observable biofilms were collected from four usage categories
of single room humidifiers: (1) property of resident volunteers,
(2) retail store displays; (3) newly purchased humidifiers fed
with tap water in the lab and (4) newly purchased humidifiers
fed with DI/RO water in the lab. Samples from 24 humidifiers
were chosen for sequence analysis, with focus on the most
commonly owned humidifier types encountered in this study,
which were boiler and ultrasonic units (including those pur-
chased for the laboratory controlled study). The sampling
design, including sample numbers for each sample type, hu-
midifier type, and usage category, is shown in Table 1. All
samples were collected and processed as detailed in Materials
and Methods. Information was collected on all units with
regards to cleaning method and frequency, mode and fre-
quency of operation, unit brand, as well as water source and
possible use of pretreatment, for metadata potentially perti-
nent to humidifier bacteriology, which is summarized in
Supplementary Table S1.
3.2. Overall bacteriology of humidifiers is low in richness
and diversity
In the overall dataset, 353 unique taxa were identified. Mea-
sures of microbial richness by Sobs and sampling depth by
Good's coverage are compiled in Supplemental Table S1. All of
the Sobs collector's curves shown in Supplemental Fig. S1 for
the individual samples and for the entire sequence collection
approach a horizontal asymptote, indicating that the main
taxa expected to be in association with single room humidi-
fiers are identified in each sample and the study as a whole.
Additionally, the Good's coverage metric (Supplemental Table
S1) was greater than 99% for all samples, further indicating
that adequate sequencing coverage is available for compari-
son between libraries (Lemos et al. 2011, 2012). An average of
48 ±23 (std. dev.) taxa were detected in each sample at the
rarefaction point (Supplemental Fig. S1a and Table S1). This
average represents 14% of all taxa detected, indicating a high
degree of potential for shared taxa between samples. Addi-
tionally, only the top 48 OTU's represent individual taxa
comprising %0.25% overall relative abundance.
Richness and diversity by sample type are described in
Fig. 1, with samples pooled into the four usage categories:
Homes, Store, Lab Tap and Lab DI/RO. Biofilm samples pre-
sented higher richness (m¼53, p¼0.015, ManneWhitney test)
and diversity (m¼2.8, p¼0.024, t-test) than other sample
types. Store samples presented the highest observed richness
(m¼75, p<0.001, ManneWhitney test) and diversity (m¼3.3,
p<0.001, t-test) compared to other usage categories. The
collection of microbes obtained from Lab humidifiers fed with
DI/RO water were observed to have lower richness (m¼36,
p¼0.057, ManneWhitney test) and diversity (m¼2.2, p¼0.107,
t-test) than other usage categories.
3.3. Humidifier bacteriology is dominated by a- and b-
proteobacteria
The large-scale taxonomic assignments within the overall
sequence dataset are summarized in Fig. 2. Only five phyla
comprised >95% overall relative abundance of the bacterial
diversity detected. Representatives of Proteobacteria domi-
nated (83.60%) humidifier bacteriology and were comprised
mainly of Sphingomonadales (31.94%), Rhizobiales (20.03%), and
Burkholderiales (19.41%). Sphingomonadales were dominated by
Sphingomonadaceae (17.08%), Rhizobiales were dominated by
Methylobacteria (11.99%), and Burkholderiales were dominated
by Comamonadaceae (9.62%) followed by fairly equal repre-
sentation of Burkholderiaceae (5.25%) and Oxalobacteraceae
(4.44%). A putative non-photosynthetic relative of Cyanobac-
teria, MLE1-12, which commonly occurs in drinking water
(LaPara et al. 2000; Ley et al. 2005; Di Rienzi et al. 2013), was the
second most abundant taxon (6.30%), followed by low levels of
Bacteriodetes (4.96%), Firmicutes (1.86%), and Actinobacteria
(1.64%).
3.4. Humidifier bacteriology is consistent across tap-fed
usage Categories
Fig. 3 shows a more detailed taxonomic breakdown of se-
quences encountered in the study, indicating the average
relative abundances of the top 30 taxa observed in the cate-
gorically grouped samples, with taxonomic identifications
binned to the family level or higher where family level clas-
sifications were not available. This collection includes all taxa
with >0.36% overall relative abundance in these categories.
For reference, Supplemental Fig. S2 compiles the top 50 taxa
encountered in all individual samples at the highest taxo-
nomic resolution available, which includes individual taxa
with >0.23% overall relative abundance in all samples. Sam-
ples from humidifiers fed with tap water, as seen in the
Homes, Store, and Lab Tap categories, were generally similar
in microbial composition to each other and to their source
waters. In general, these tap-fed samples tended to be
Table 1 eHumidifier Sampling Campaign Study Design. Row 1) Humidifier Sampling Campaign Design, with number of
humidifiers sampled in total and for each mode of operation. Row 2) Number of Sampling Events per Humidifier. Row 3)
Usage Category, with number of humidifiers sampled of each mode of operation. Row 4) Sample Type. Row 5) Humidifier
Type (B¼Boiler, U¼Ultrasonic). Row 6) Number of Samples.
24 Humidifiers (14 ultrasonic, 10 boiler)
Single sampling event Longitudinal sampling events
Homes
(8 ultrasonic, 4 boiler)
Store
(4 ultrasonic, 4 boiler)
Lab tap
(1 ultrasonic, 1 boiler)
Lab DI/RO
(1 ultrasonic, 1 boiler)
Source
water
Tank water Biofilm Source
water
Tank water Biofilm Source
water
Tank water Biofilm Source
water
Tank water Biofilm
BUBU BUBU BUBU BUBU
9 4 7 5 10 1 3 3 4 5 2 5 4 6 8 0 3 2 3 2
water research 69 (2015) 318e327 321

particularly enriched in sphingomonads, methylobacteria,
and comamonads. Samples from laboratory humidifiers fed
with tap water followed this trend, with the exception of
enrichment of Acidovorax of the Comamonadaceae, presum-
ably due to increased presence of this taxon in lab tap water in
comparison to other tap source waters. In contrast, the labo-
ratory humidifiers fed DI/RO water almost completely lacked
the sphingomonads found in all other samples and instead
were particularly enriched in Ralstonia of the Burkholderiaceae.
Humidifiers fed DI/RO water also had lower bacterial loads, as
indicated by a lower quantitative PCR average copy number of
3.6 $10
6
for samples from humidifiers fed by DI/RO water
versus 2.3 $10
7
for all samples from humidifiers fed by tap
water (Supplemental Table S1). The general similarities of the
bacteriology of humidifiers fed with tap water and that of their
source tap waters indicates that humidifier bacteriology can
be attributed largely to source water introduction (see also
Section 3.7).
3.5. Increased richness and diversity in biofilms is
idiosyncratic
Although the microbial contents of biofilms belonged to taxa
that were abundant in the overall dataset, some low overall
abundance taxa were dominant in biofilm samples, as indi-
cated by high relative abundance of ‘Other’taxa (Fig. 3). To
further examine the bacteriology of biofilm samples beyond
the 50 most abundant taxa presented in Supplemental Fig. S2,
Supplemental Figure S3 compiles the next 50 most abundant
taxa. The kinds of taxa that deviated from the dominant taxa
observed in tank and source waters varied unpredictably, and
were usually due to a few abundant taxa that were rare or
absent in other samples, for example, the enrichment of
Arthrobacter in one Home boiling humidifier biofilm sample
and the enrichment of Pseudoxanthomonas in one Store ultra-
sonic humidifier biofilm sample (Supplemental Fig. S3).
3.6. Very few taxa of potential concern were encountered
Because of the relevance to human health, samples were
scrutinized for enrichment of taxa that might indicate
opportunistic pathogens including specifically the clades of
Elizabethkingia spp., Legionella spp., Mycobacterium spp., and
Pseudomonas spp. Because of short sequence lengths and high
degrees of similarity, taxonomic resolution to the species level
Fig. 1 ea) Richness and b) Diversity in Usage Categories. Box plots show the mean (A), minimum, maximum, and the 25th,
50th, and 75th percentiles of a) richness (Sobs) and b) diversity (ShannonH) within each usage category. Metrics were
calculated as the mean for each sample at the rarefaction point of 7199 sequences per sample with 25 replicates.
Fig. 2 eMost Abundant Taxa. Overall relative abundances
of the most abundant taxa encountered in 12 source water,
31 tank water, and 43 biofilm samples associated with 14
ultrasonic and 10 boiler single room humidifiers.
water research 69 (2015) 318e327322

is not possible for organisms in these clades using sequences
in this survey. Relative abundances for these clades were al-
ways <1% and usually <<1% in any usage category, with the
exception of Store boiler units that contained ~1% Mycobacte-
rium spp. and ~5% Pseudomonas spp. (Supplemental Table S2).
3.7. Source water seeds humidifier bacteriology
To assess the effect of source water, sample type, or humidi-
fier type on humidifier bacteriology, ordinations were per-
formed on the dataset using non-metric multidimensional
scaling (NMDS). This statistical method extracts and displays
similarities in bacterial composition between samples, as
represented by the proximity of points in the ordination plot.
The ordination including all samples resulted in a 3-D stress
value of 0.18, which approaches the limit of reliability of 0.2
(Clarke, 1993) and should be regarded with caution. This
dataset was comprised of relatively dissimilar samples, as
shown by the low average MorisitaeHorn similarity index of
0.16 ±0.21 (std. dev.), which could contribute to model stress.
Samples sharing the same source water overlapped in
ordination space, but did not overlap well with samples from
other source waters, suggesting a differentiating effect of
source water on humidifier bacteriology (Supplementary
Fig. 4).
In an ordination including only samples from home and
store humidifiers, the 3-D stress of the modeled solution was
improved to 0.16. In this ordination, source water samples
formed a small cluster that is nested within tank water and
biofilm sample types (Fig. 4a), indicating an expansion in
bacteriology in these habitats beyond but encompassing that
encountered in source waters. No significant difference was
observed between tank water and biofilm samples (perma-
nova, p¼0.053). The cluster of source water samples was also
nested within all samples taken from ultrasonic humidifiers.
Boiler humidifier samples, however, do not cluster with
source water samples and are significantly different than ul-
trasonic humidifier samples (permanova, p¼0.001), indi-
cating that a selective pressure such as temperature may alter
bacteriology in boiler humidifiers.
An ordination was performed on the subset of the data
from the laboratory study samples, and the 3-D stress of the
Fig. 3 eTop 30 OTUs in Usage Categories. Heatmap displays the percent relative abundance of the 30 overall most dominant
taxa, binned to the family level of taxonomic resolution (or higher when classification was not to the family level), in each
usage category.
water research 69 (2015) 318e327 323

modeled solution, 0.09, indicates a well-fitting model that can
be used with confidence to draw inferences about relation-
ships between samples. In this ordination, hulls drawn to
enclose samples sharing the same water source (tap versus
DI/RO water) do not overlap (Fig. 4b), confirming the signifi-
cance of source water in determining humidifier bacteriology.
Additionally, Mantel test results showed that time (or days
since sampling began) was significantly correlated with
bacteriology in the samples from the controlled study, after
accounting for the significance of source water (p¼0.001,
r¼0.27).
4. Discussion
In this study we determined, with adequate sequencing depth
(Supplemental Fig. S1 and Supplemental Table S1), the nature
of the bacteriology associated with single room ultrasonic and
boiler humidifier source waters, tank waters, and biofilms
across usage categories including a retail store, homes, and
the laboratory.
Richness encountered in this survey of humidifier bacte-
riology was low (Fig. 1 and Supplementary Fig. S1) in com-
parison to previous studies of drinking water microbiology.
Our survey identified 353 taxa (excluding 85 singleton se-
quences, which were removed prior to further analyses), with
an average of 48 taxa encountered at rarefaction. In contrast,
recent studies of drinking waters have reported greater rich-
ness estimates ranging from 400 to 1541 taxa detected overall
(Lautenschlager et al. 2013; Holinger et al. 2014; Kelly et al.
2014; Wang et al. 2014), or 147 to 198 taxa detected at rare-
faction (Shaw et al. 2014). Diversity was also lower than re-
ported for drinking waters, by several criteria (Fig. 1 and
Supplementary Table S1). Humidifier study samples consti-
tuted a Simpson index of 0.3 ±0.2 (std. dev.) at rarefaction,
while Wang et al. (2014) encountered a greater Simpson index,
ranging from 3.2 to 11.6 at rarefaction. Humidifier study
samples exhibited Chao1 values ranging from 19 to 145 at
rarefaction, while Lautenschlager et al. (2013) and Hwang et al.
(2012) encountered a greater Chao1, ranging from 1200 to 1800
and 40 to 500, respectively. Humidifier study samples consti-
tuted a Shannon index of 2.6 ±0.9 at rarefaction, while Shaw
et al. (2014) reported a greater Shannon index, ranging from 3
to 7. We believe that humidifiers select for a subset of drinking
water taxa introduced over time, resulting in lower richness
and diversity. This selection is particularly evident in boiler
humidifiers, where increased temperature may enhance the
selective process (Fig. 4a). One complicating factor could be
the potential for seasonal effects, as reported by Pinto et al.
(2014).
The taxa encountered in this assessment of humidifier
bacteriology generally were similar to taxa commonly
encountered in municipal drinking water microbiology in the
US (Eichler et al. 2006; Gomez-Alvarez et al. 2012; Holinger
et al. 2014). The dominant microbes in all samples tended to
be of oligotrophic biofilm-forming kinds including sphingo-
monads, methylobacteria, comamonads and other a- and b-
group proteobacteria (Figs. 2 and 3). The dominance of the
proteobacteria phylum in this study of humidifiers and their
source waters reflects the microbiology encountered in
several recent sequence-based studies of drinking water
(Gomez-Alvarez et al. 2012; Hwang et al. 2012; Lautenschlager
et al. 2013; Holinger et al. 2014; Hong et al. 2014; Liu et al. 2014;
Pinto et al. 2014; Shaw et al. 2014; Wang et al. 2014). At the
class level, domination by a-proteobacteria in humidifiers and
their source waters is in agreement with the results of Wang
et al. (2014), Hollinger et al. (2014), the winter samples taken
by Pinto et al. (2014), and the chlorinated system studied by
Gomez-Alvarez et al. (2012), but contrasts b-proteobacteria
dominated samples studied by Hong et al. (2014) and the
summer samples taken by Pinto et al. (2014), and also con-
trasts the g-proteobacteria dominated systems studied by
Hwang et al. (2012) and Kelly et al. (2014). In accordance with
humidifier bacteriology, Shaw et al. (2014) found a-
Fig. 4 eOrdination Plots of bacterial communities for a) samples from homes and store humidifiers and b) samples from the
laboratory study humidifiers. For both ordinations, the OTU table (not including unclassified or singleton sequences) was
rarefied to 7199 sequences per sample and Hellinger-transformed. BrayeCurtis distance matrices were used, and 3
dimensions were chosen for ordination using non-metric multidimensional scaling (NMDS). Symbols distinguish samples
by habitats including sample type (Biofilm, Tank Water, or Source Water) and humidifier type (Boiler or Ultrasonic). Hulls in
a) enclose samples from the same habitat. Hulls in b) enclose samples sharing the same source water.
water research 69 (2015) 318e327324

proteobacteria to be dominated by Sphingomonodales and Rhi-
zobiales. Additionally, Liu et al. (2014) found a-proteobacteria
and more specifically Sphingomonodales to be more dominant
in the attached fractions of distribution samples versus the
bulk water fraction. Considering that most humidifier users
do not flush premise plumbing water from their pipes prior to
filling their humidifier, it is likely that the enrichment of
Sphingomonodales in our study suggests that source water
bacteriology and the resulting humidifier bacteriology is
affected by premise and/or distribution system biofilm and
solids. Stagnation in premise plumbing is known to affect
water microbiology (Lautenschlager et al. 2010). This influence
of water stagnation was observed in laboratory tap-fed hu-
midifier samples whose source was an unused sink on the 4th
floor with a likely greater water age than other source waters.
Laboratory tap-fed humidifiers and their source waters were
enriched in Comamonadaceae, in agreement with
Lautenschlager et al. (2013) who observed increased abun-
dance of Comamonadaceae with increasing water age.
The kinds of microbes detected in the humidifiers were
mainly predicated by respective source waters, as indicated by
shared taxa in tap-fed samples (Fig. 3) and lack of overlap in
bacterial assemblages based on source water (Fig. 4b and
Supplementary Fig. S4). Bacterial assemblages were observed
to cluster based on sample type (biofilm, tank water, or source
water) and source water bacteriology was nested within that
encountered in these humidifier habitats (Fig. 4a). Addition-
ally, time (or days since sampling began) was found to have a
significant correlation with taxonomy encountered in the
longitudinal controlled study. The nestedness of source water
samples within other sample types, in combination with the
correlative effect of time in the controlled study, is consistent
with the scenario that source waters form the basis of hu-
midifier bacteriology, which then expands in tank waters and
biofilms. This expansion is possibly due to longitudinal vari-
ation in source water microbiology (due to changes in premise
plumbing and/or municipal water microbiology), exposure to
ambient microbes in the space occupied by the humidifier,
and/or human handling. In the case of boiler humidifiers,
expansion could occur due to a selective pressure such as
temperature applied by the heating elements. The sources of
the bacteriology that developed in the laboratory humidifiers
fed with DI/RO water eparticularly the enrichment of Ral-
stonia spp., of the Burkholderiales ecould not be determined
due to an inability to obtain PCR products from the DI/RO
source water, presumably due to lower bacterial loads in DI/
RO water compared to tap water (indicated by quantitative
PCR (Supplemental Fig. S2).
Expansion of humidifier taxonomic richness and diversity
was observed most obviously in biofilms (Fig. 2,S2 and S3).
Yet, biofilm phylotypes that were more abundant than tank
water phylotypes were idiosyncratic in occurrence and tended
to occur uniquely in different humidifiers. This contrasts with
showerhead biofilms, for instance, which were shown with
sequence studies to have increased occurrence of mycobac-
teria, presumably because of the hydrophobic and biofilm-
forming qualities of such organisms (Feazel et al. 2009).
The usage category with the highest observed richness and
diversity in this survey was Store humidifiers (Fig. 1). Store
humidifiers also contained somewhat increased cell
concentrations in water (Supplementary Table S1). Store hu-
midifiers were run continuously during operating hours,
necessitating frequent refills and thereby providing more op-
portunities for introduction of source water microbiota. This
is in contrast to humidifiers operated by residents, which were
run and refilled intermittently. Additionally, store humidifiers
were not known by store employees to be cleaned, which is in
contrast to residents who reported cleaning with varying de-
grees of frequency (Supplementary Table S1). These factors,
along with the increased relative abundances of potential
opportunistic pathogens encountered in store humidifiers
(Supplementary Table S2), brings to question whether it is
appropriate for stores to operate display models, since they
may present a potential route of exposure to opportunistic
pathogens. For home users, especially those who operate and
refill their humidifiers frequently, this reinforces (Mohan et al.
1998) the importance of cleaning to prevent buildup of
biomass.
5. Conclusions
We have found that single room humidifier bacteriology is
generally low in richness and diversity, with highest richness
and diversity observed in biofilm samples, and within store
humidifiers as a sampling category. Bacteriology encountered
was reminiscent of that seen in previous studies of municipal
drinking water, being dominated by Sphingomonadales,Rhizo-
biales, and Burkholderiales of the Proteobacteria, and MLE1-12
of the Cyanobacteria, with very few sequences of potential
health concern detected. Source water was found to have a
differentiating effect on humidifier bacteriology, with bacte-
rial richness and diversity likely expanding over time due to
external factors including human handling, cleaning prac-
tices, longitudinal variation in feed water microbiology, and
exposure to microbes from other sources in the home.
Acknowledgements
The authors thank the local retail store and residents who
allowed sampling of their humidifiers, Kimberly Ross for
assistance with planning and laboratory techniques, and
Paula Olsiewski of the Alfred P. Sloan Foundation. This work
was funded by grant 2011-10-02 from the Microbiology of the
Built Environment Program of the Alfred P. Sloan Foundation.
Appendix A. Supplementary data
Supplementary data related to this article can be found at
http://dx.doi.org/10.1016/j.watres.2014.11.024.
references
Anderson, M.J., 2001. A new method for non-parametric
multivariate analysis of variance. Austral Ecol. 26, 32e46.
water research 69 (2015) 318e327 325

Anderson, W.B., George Dixon, D., Mayfield, C.I., 2007. Estimation
of endotoxin inhalation from shower and humidifier exposure
reveals potential risk to human health. J. Water Health 5 (4),
553e572.
Arundel, A.V., Sterling, E.M., Biggin, J.H., Sterling, T.D., 1986.
Indirect health effects of relative humidity in indoor
environments. Environ. Health Perspect. 65, 351e361.
Clarke, K.R., 1993. Non-parametric multivariate analyses of
changes in community structure. Austral Ecol. 18 (1), 117e143.
ConsumerReports. org, 2013. Humidifiers. From:http://www.
consumerreports.org/cro/humidifiers.htm.
Daftary, A.S., Deterding, R.R., 2011. Inhalational lung injury
associated with humidifier “white dust”. Pediatrics 127 (2),
e509e512.
Di Rienzi, S.C., Sharon, I., Wrighton, K.C., Koren, O., Hug, L.A.,
Thomas, B.C., Goodrich, J.K., Bell, J.T., Spector, T.D.,
Banfield, J.F., Ley, R.E., 2013. The human gut and groundwater
harbor non-photosynthetic bacteria belonging to a new
candidate phylum sibling to Cyanobacteria. eLlife 2, e01102.
Edgar, R.C., Haas, B.J., Clemente, J.C., Quince, C., Knight, R., 2011.
UCHIME improves sensitivity and speed of chimera detection.
Bioinformatics 27 (16), 2194e2200.
Eichler, S., Christen, R., H€
oltje, C., Westphal, P., B€
otel, J., Brettar, I.,
Mehling, A., H€
ofle, M.G., 2006. Composition and dynamics of
bacterial communities of a drinking water supply system as
assessed by RNA- and DNA-Based 16S rRNA gene
fingerprinting. Appl. Environ. Microbiol. 72 (3), 1858e1872.
Ewing, B., Green, P., 1998. Base-calling of automated sequencer
traces using phred. II. Error probabilities. Genome Res. 8 (3),
186e194.
Ewing, B., Hillier, L., Wendl, M.C., Green, P., 1998. Base-calling of
automated sequencer traces using phred. I. Accuracy
assessment. Genome Res. 8 (3), 175e185.
Feazel, L.M., Baumgartner, L.K., Peterson, K.L., Frank, D.N.,
Harris, J.K., Pace, N.R., 2009. Opportunistic pathogens enriched
in showerhead biofilms. Proc. Natl. Acad. Sci. U. S. A. 106 (38),
16393e16399.
Forsgren, A., Persson, K., Ursing, J., Walder, M., Borg, I., 1984.
Immunological aspects of humidifier fever. Eur. J. Clin.
Microbiol. 3 (5), 411e418.
Gomez-Alvarez, V., Revetta, R.P., Santo Domingo, J.W., 2012.
Metagenomic analyses of drinking water receiving different
disinfection treatments. Appl. Environ. Microbiol. 78 (17),
6095e6102.
Hara, N., Alkanani, A.K., Ir, D., Robertson, C.E., Wagner, B.D.,
Frank, D.N., Zipris, D., 2012. Prevention of virus-induced type 1
diabetes with antibiotic therapy. J. Immunol. 189 (8),
3805e3814.
Holinger, E.P., Ross, K.A., Robertson, C.E., Stevens, M.J.,
Harris, J.K., Pace, N.R., 2014. Molecular analysis of point-of-use
municipal drinking water microbiology. Water Res. 49,
225e235.
Hong, S.B., Kim, H.J., Huh, J.W., Do, K.H., Jang, S.J., Song, J.S.,
Choi, S.J., Heo, Y., Kim, Y.B., Lim, C.M., Chae, E.J., Lee, H.,
Jung, M., Lee, K., Lee, M.S., Koh, Y., 2014. A cluster of lung
injury associated with home humidifier use: clinical,
radiological and pathological description of a new syndrome.
Thorax.
Hwang, C., Ling, F., Andersen, G.L., LeChevallier, M.W., Liu, W.T.,
2012. Microbial community dynamics of an urban drinking
water distribution system subjected to phases of
chloramination and chlorination treatments. Appl. Environ.
Microbiol. 78 (22), 7856e7865.
Kelly, J.J., Minalt, N., Culotti, A., Pryor, M., Packman, A., 2014.
Temporal variations in the abundance and composition of
biofilm communities colonizing drinking water distribution
pipes. PLoS One 9 (5), e98542.
Lacasse, Y., Girard, M., Cormier, Y., 2012. Recent advances in
hypersensitivity pneumonitis. Chest 142 (1), 208e217.
LaPara, T.M., Nakatsu, C.H., Pantea, L., Alleman, J.E., 2000.
Phylogenetic analysis of bacterial communities in mesophilic
and thermophilic bioreactors treating pharmaceutical
wastewater. Appl. Environ. Microbiol. 66 (9), 3951e3959.
Lautenschlager, K., Boon, N., Wang, Y., Egli, T., Hammes, F., 2010.
Overnight stagnation of drinking water in household taps
induces microbial growth and changes in community
composition. Water Res. 44 (17), 4868e4877.
Lautenschlager, K., Hwang, C., Liu, W.T., Boon, N., Koster, O.,
Vrouwenvelder, H., Egli, T., Hammes, F., 2013. A microbiology-
based multi-parametric approach towards assessing
biological stability in drinking water distribution networks.
Water Res. 47 (9), 3015e3025.
Legendre, P., Legendre, R., 2012. Numerical Ecology. Elsevier,
Amsterdam, NL.
Lemos, L.N., Fulthorpe, R.R., Roesch, L.F., 2012. Low sequencing
efforts bias analyses of shared taxa in microbial communities.
Folia Microbiol. 57 (5), 409e413.
Lemos, L.N., Fulthorpe, R.R., Triplett, E.W., Roesch, L.F., 2011.
Rethinking microbial diversity analysis in the high throughput
sequencing era. J. Microbiol. Methods 86 (1), 42e51.
Ley, R.E., Backhed, F., Turnbaugh, P., Lozupone, C.A., Knight, R.D.,
Gordon, J.I., 2005. Obesity alters gut microbial ecology. Proc.
Natl. Acad. Sci. U. S. A. 102 (31), 11070e11075.
Liu, G., Bakker, G.L., Li, S., Vreeburg, J.H., Verberk, J.Q.,
Medema, G.J., Liu, W.T., Van Dijk, J.C., 2014. Pyrosequencing
reveals bacterial communities in unchlorinated drinking
water distribution system: an integral study of bulk water,
suspended solids, loose deposits, and pipe wall biofilm.
Environ. Sci. Technol. 48 (10), 5467e5476.
Magurran, A.E., 2004. Measuring Biological Diversity. Blackwell
Publishing Company, Malden, MA.
Markle, J.G., Frank, D.N., Mortin-Toth, S., Robertson, C.E.,
Feazel, L.M., Rolle-Kampczyk, U., von Bergen, M., McCoy, K.D.,
Macpherson, A.J., Danska, J.S., 2013. Sex differences in the gut
microbiome drive hormone-dependent regulation of
autoimmunity. Science 339 (6123), 1084e1088.
Matsunaga, T., Okochi, M., Nakasono, S., 1995. Direct count of
bacteria using fluorescent dyes: application to assessment of
electrochemical disinfection. Anal. Chem. 67 (24), 4487e4490.
Mohan, A.K., Feigley, C.E., Macera, C.A., 1998. Humidifier use in
the home environment and its effects on respiratory health.
Appl. Occup. Environ. Hyg. 13 (11), 782e787.
Moran-Gilad, J., Lazarovitch, T., Mentasti, M., Harrison, T.,
Weinberger, M., Mordish, Y., Mor, Z., Stocki, T., Anis, E.,
Sadik, C., Amitai, Z., Grotto, I., 2012. Humidifier-associated
paediatric Legionnaires'disease, Israel, February 2012.
Eurosurveillance 17 (41), 20293.
Muller-Wening, D., Koschel, D., Stark, W., Sennekamp, H.J., 2006.
[Humidifier-associated disease in the general population].
Dtsch. Med. Wochenschr. 131 (10), 491e496.
Myatt, T.A., Kaufman, M.H., Allen, J.G., MacIntosh, D.L.,
Fabian, M.P., McDevitt, J.J., 2010. Modeling the airborne
survival of influenza virus in a residential setting: the impacts
of home humidification. Environ. Health 9, 55.
Nadkarni, M.A., Martin, F.E., Jacques, N.A., Hunter, N., 2002.
Determination of bacterial load by real-time PCR using a
broad-range (universal) probe and primers set. Microbiology
148 (Pt 1), 257e266.
Nakagawa, A., Yamaguchi, T., Takao, T., Amano, H., 1995. [A case
of hypersensitivity pneumonitis probably caused by a
humidifier in winter]. Nihon Kyobu Shikkan Gakkai Zasshi 33
(2), 202e208.
Ohnishi, H., Yokoyama, A., Hamada, H., Manabe, S., Ito, R.,
Watanabe, A., Katayama, H., Yasuhara, Y., Ikezoe, J., Higaki, J.,
water research 69 (2015) 318e327326

2002. Humidifier lung: possible contribution of endotoxin-
induced lung injury. Intern Med. 41 (12), 1179e1182.
Oksanen, J., Guillaume Blanchet, F., Kindt, R., Legendre, P.,
Minchin, P.R., O'Hara, R.B., Simpson, G.L., Solymos, P.,
Henry, M., Stevens, H., Wagner, H., 2013. Vegan: Community
Ecology Package. http://CRAN.R-project.org/package¼vegan.
Pinto, A.J., Schroeder, J., Lunn, M., Sloan, W., Raskin, L., 2014.
Spatial-temporal survey and occupancy-abundance modeling
to predict bacterial community dynamics in the drinking
water microbiome. MBio 5 (3), e01135e01114.
Pruesse, E., Peplies, J., Glockner, F.O., 2012. SINA: accurate high-
throughput multiple sequence alignment of ribosomal RNA
genes. Bioinformatics 28 (14), 1823e1829.
Quast, C., Pruesse, E., Yilmaz, P., Gerken, J., Schweer, T., Yarza, P.,
Peplies, J., Glockner, F.O., 2013. The SILVA ribosomal RNA gene
database project: improved data processing and web-based
tools. Nucleic Acids Res. 41 (Database issue), D590eD596.
RCoreTeam, 2014. R: a Language and Environment for Statistical
Computing. R Foundation for Statistical Computing, Vienna,
Austria. http://www.R-project.org/.
Robertson, C.E., Harris, J.K., Wagner, B.D., Granger, D., Browne, K.,
Tatem, B., Feazel, L.M., Park, K., Pace, N.R., Frank, D.N., 2013.
Explicet: graphical user interface software for metadata-
driven management, analysis and visualization of
microbiome data. Bioinformatics 29 (23), 3100e3101.
RStudio, 2014. RStudio: Integrated Development Environment for
R. Boston, MA. http://www.rstudio.org/.
Rylander, R., Haglind, P., 1984. Airborne endotoxins and
humidifier disease. Clin. Allergy 14 (1), 109e112.
Schloss, P.D., Westcott, S.L., 2011. Assessing and improving
methods used in operational taxonomic unit-based
approaches for 16S rRNA gene sequence analysis. Appl.
Environ. Microbiol. 77 (10), 3219e3226.
Shaw, J.L., Monis, P., Fabris, R., Ho, L., Braun, K., Drikas, M.,
Cooper, A., 2014. Assessing the impact of water treatment on
bacterial biofilms in drinking water distribution systems using
high-throughput DNA sequencing. Chemosphere 117C,
185e192.
Shepard, R.N., 1962. The analysis of proximities:
multidimensional scaling with an unknown dstance function,
I. Psychometrika 27 (2), 125e140.
SystatSoftware, 2008. SigmaPlot for Windows,.
Trenchs Sainz De La Maza, V., Domingo Garau, A., Garcia-Tornel
Florensa, S., Gaspa Marti, J., 2002. [Domestic humidifiers. What
do we know about them?]. An. Espanoles Pediatr. 57 (3),
231e237.
Umezawa, M., Sekita, K., Suzuki, K., Kubo-Irie, M., Niki, R.,
Ihara, T., Sugamata, M., Takeda, K., 2013. Effect of aerosol
particles generated by ultrasonic humidifiers on the lung in
mouse. Part. Fibre Toxicol. 10 (1), 64.
USEPA, 1991. Use and Care of Home Humidifiers. Indoor Air Facts
No. 8, from: http://www.epa.gov/iaq/pdfs/humidifier_
factsheet.pdf.
Wang, H., Masters, S., Edwards, M.A., Falkinham 3rd, J.O.,
Pruden, A., 2014. Effect of disinfectant, water age, and pipe
materials on bacterial and eukaryotic community structure in
drinking water biofilm. Environ. Sci. Technol. 48 (3),
1426e1435.
Yamamoto, Y., Osanai, S., Fujiuchi, S., Yamazaki, K., Nakano, H.,
Ohsaki, Y., Kikuchi, K., 2002. Extrinsic allergic alveolitis
induced by the yeast Debaryomyces hansenii. Eur. Respir. J. 20
(5), 1351e1353.
water research 69 (2015) 318e327 327