High throughput genomic sequencing of bioaerosols in
broiler chicken production facilities
Kate M. O’Brien,
Michael S. Chimenti,
and Matthew W. Nonnenmann
Department of Occupational and Environmental Health,
University of Iowa, Iowa City, IA, USA.
Iowa Institute of Human Genetics, Bioinformatics
Division, University of Iowa, Iowa City, IA, USA.
Mississippi State University Extension Service,
Mississippi State, MS, USA.
Department of Agriculture, Stephen F. Austin State
University, Nacogdoches, TX, USA.
Chronic inhalation exposure to agricultural dust pro-
motes the development of chronic respiratory dis-
eases among poultry workers. Poultry dust is
composed of dander, chicken feed, litter bedding
and microbes. However, the microbial composition
and abundance has not been fully elucidated. Geno-
mic DNA was extracted from settled dust and per-
sonal inhalable dust collected while performing litter
sampling or mortality collection tasks. DNA libraries
were sequenced using a paired-end sequencing-by-
synthesis approach on an Illumina HiSeq 2500.
Sequencing data showed that poultry dust is pre-
dominantly composed of bacteria (64–67%) with a
small quantity of avian, human and feed DNA (<2%
of total reads). Staphylococcus sp. AL1,Salinicoc-
cus carnicancri and Lactobacillus crispatus were the
most abundant bacterial species in personal expo-
sure samples of inhalable dust. Settled dust had a
moderate relative abundance of these species as
well as Staphylococcus lentus and Lactobacillus
salivarius. There was a statistical difference between
the microbial composition of aerosolized and settled
dust. Unlike settled dust composition, aerosolized
dust composition had little variance between sam-
ples. These data provide an extensive analysis of the
microbial composition and relative abundance in
personal inhalable poultry dust and settled poultry
Agricultural dust is generated in animal production facili-
opez et al., 2010). Inhalation exposure to
agricultural dust induces pulmonary inﬂammation and
can lead to the development of chronic respiratory dis-
eases (Palmberg et al., 1998; Redente and Massengale,
2006; Poole and Romberger, 2012). Several studies
have demonstrated that agricultural workers, particularly
poultry workers, have a reduction in lung function and
higher prevalence of chronic respiratory diseases (e.g.,
chronic bronchitis) (Zuskin et al., 1995; Simpson et al.,
1998; Radon et al., 2001; Rylander and Carvalheiro,
2006; Viegas et al., 2013). Furthermore, poultry dust
induces pulmonary lesions and cardiac morphological
changes in broiler chickens (Oyetunde et al., 1978;
Riddell et al., 1998; Lai et al., 2009). Due to these
human and animal health implications, it is critical to
characterize the composition of poultry dust to deter-
mine the source and potential targets for engineering
Dust generated in poultry production consists of a
complex mixture of chicken and human derived dander,
bedding, chicken feed, and viable and nonviable micro-
bial populations (Lenhart and Olenchock, 1984; Radon
et al., 2002). Previous studies have used more speciﬁc
molecular biology tools such as enzyme-linked
immunosorbent assay and polymerase chain reaction to
characterize poultry dust (Kwon et al., 2000; Oppliger
et al., 2008; Prester and Macan, 2010; Just et al., 2011;
Gerald et al., 2014). These studies have shown that
poultry dust contains inﬂammatory agents such as
lipopolysaccharide (LPS) and peptidoglycan (PGN), con-
stituents of bacterial cell walls in Gram-negative and
Gram-positive bacteria (Thedell et al., 1980; Sonesson
et al., 1988; Gerald et al., 2014). Also, zoonotic viruses
(e.g., avian inﬂuenza) have been detected in dust (Chen
et al., 2010; Spekreijse et al., 2013). Settled agricultural
Received 20 May, 2016; revised 17 June, 2016; accepted 21 June,
2016. *For correspondence. E-mail matthew-nonnenmann@uiowa.
edu; Tel. +1 3193354207; Fax. 319 38 44138.
Microbial Biotechnology (2016) 0(0), 000–000
The authors would like to thank the University of Minnesota
Genomic Core for their excellent work on genomic DNA sequencing
and the usage of the Illumina 2500. We would also like to thank
Sarah Williams Ischer, Courtney Isbell, Marcelo Moreira, and Joey
Bray PhD for their assistance in sample collection and access to
the broiler chicken farms. This research project was supported by
the Southwest Center for Agricultural Health, Injury Prevention and
Education - CDC/NIOSH 2U54OOH007541. The authors do not
have a conﬂict of interest to declare.
ª2016 The Authors. Microbial Biotechnology published by John Wiley & Sons Ltd and Society for Applied Microbiology.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and
reproduction in any medium, provided the original work is properly cited.
dust extracts are predominantly employed to determine
the mechanism underlying pulmonary toxicity (Palmberg
et al., 1998; Redente and Massengale, 2006). Further-
more, it is unknown if there is a microbial difference
between settled dust and aerosolized dust in large ani-
The advancement of genomic sequencing technology
has made the comprehensive analysis of microbes in
poultry dust readily available. Previously, pyrosequenc-
ing was used to characterize bacteria and fungi targeting
ribosomal RNAs (i.e., 16S and 18S), which are ubiqui-
tously expressed in the bacteria and eukaryota domains
(Nonnenmann et al., 2010; Boissy et al., 2014). In one
sample of poultry dust, Staphylococcus cohnii and
Sagenomella sclerotialis, respectively, were the most
abundant bacterial and fungal species (Nonnenmann
et al., 2010). Whole-genome shotgun pyrosequencing
demonstrated that there are differing taxonomic proﬁles
and genus abundance among swine, grain, and house
dust (Boissy et al., 2014). Pyrosequencing technology
generally has a high error rate, high cost per megabase,
and lack of sufﬁcient sequence coverage needed to
assess complex whole genomic samples containing hun-
dreds of low-abundance bacterial, viral and fungal spe-
cies. Newer sequencing technology such as the Illumina
HiSeq 2500 platform has several advantages over other
sequencing technologies. Speciﬁcally, the Illumina HiSeq
2500 platform has lower error rates, better breadth and
coverage depth, and lower cost per megabase than
older technologies. Illumina HiSeq 2500 uses the
sequencing by synthesis approach. Brieﬂy, genomic
DNA is fragmented into 300–600 base pair segments,
and fragments are indexed using adapter sequences.
The prepared library DNA is added to a ﬂowcell and
undergoes bridge ampliﬁcation to form clonal clusters,
where each cluster represents a single DNA fragment.
Four different ﬂuorescent dye-tagged nucleotides
(dNTPs) are added to the reaction. Once the comple-
mentary dNTPs are incorporated into the DNA strand,
the instrument detects the emission of light at each clus-
ter simultaneously, which allows parallel sequencing of
the library. Using paired-end sequencing, two reads are
generated from the same DNA strand by sequencing
from both the 50and 30end.
The aim of this study was to characterize the composi-
tion of microbial communities in airborne and settled
poultry dust using a whole-genome shotgun sequencing
approach and state-of-the-art analysis tools.
Comprehensive analysis of microbial composition of
airborne and settled poultry dust in broiler chicken
Microbes account for the vast majority of the extracted
DNA from both inhalable and settled poultry dust,
yielding 95% of total reads (Fig. 1). DNA belonging to
Homo sapiens (human) and Gallus gallus (broiler
chicken) contributed less than 2% of reads in all poul-
try dust samples. Zea mays (maize) and Glycine max
(soybean) DNA, components of chicken feed, were
less than 1% of reads from settled and inhalable dust.
Therefore, the vast majority of the DNA reads from
each sample represent genuine microbial diversity pre-
sent in the dust.
Metagenomic Phylogenetic Analysis 2 (MetaPhlan2)
was used as the relative abundance analysis because
it reports relative abundance on a community basis
(i.e., the proportion of the total number of microbes in
the sample belonging to a given species, genus, etc.)
rather than on a read basis (Segata et al., 2012). This
is advantageous because it yields percentages that
have greater biological meaning and interpretability
than simply the proportion of reads sequenced.
MetaPhlan2 analysis showed that inhalable litter sam-
pling (LS) dust, settled (S) dust and inhalable mortality
collection (MC) dust had similar relative abundance at
the domain level with the majority classiﬁed as bacte-
ria (67%, 64% and 64% respectively). Viruses were
the second most abundant domain in LS dust (25%),
settled dust (29%) and MC dust (29%) (Fig. 2A).
Although most viruses in the samples are attributed as
‘no-name’or unclassiﬁed, there were signiﬁcant contri-
butions from the plant viral family Potyviridae and well-
characterized viral orders Caudovirales,Picornavirales,
Tymovirales,Mononegavirales and Herpesvirales (data
Fig. 1. The primary sources of genomic DNA in poultry dust. FastQ
Screen was used to determine the percentage of total sequencing
reads from microbes, chickens, humans, feed particulates (Zea
mays and Glycine max) and ligated adapters in genomic DNA
extracted from poultry dust collected during litter sampling (black),
settled (grey) and mortality collection (white).
ª2016 The Authors. Microbial Biotechnology published by John Wiley & Sons Ltd and Society for Applied Microbiology.
2K. M. O’Brien et al.
At the bacterial phylum level, Proteobacteria (33%,
37%, 35%), Firmicutes (28%, 27%, 28%) and Actinobac-
teria (19%, 15%, 16%) were the most abundant in LS
dust, settled dust and MC dust respectively (Fig. 2B).
The fungal phylum Ascomycota (1.4%, 1.3%, 1.4%) was
a smaller relative abundance in poultry dust than bacte-
rial phyla (Fig. 2B).
The heat map in Fig. 3 shows the 25 most abundant
species in common among the poultry dust samples.
The samples are clustered hierarchically using the Bray-
Curtis dissimilarity method. With the exception of three
locations (LS13-15), the LS dust samples were one dis-
tinct cluster. The MC and settled dust samples formed a
secondary cluster. Staphlyococcus sp AL1,Salinicoccus
carnicancri and Lactobacillus crispatus had a high rela-
tive abundance in personal inhalable LS dust (LS1-15).
In comparison, MC dust (MC1-3) had a reduction in
L. crispatus but similar relative abundance of Staphylo-
coccus sp AL1 and S. carnicancri. Bacterial genera
detected in settled dust (S1-3) had a moderate relative
abundance of Staphylococcus sp AL1,L. crispatus,
S. carnicancri,Staphylococcus lentus and Lactobacillus
salivarius. Settled dust had an elevated abundance of
Enterococcus cecorum compared with LS and MC sam-
ples (Fig. 3).
In addition, the viral species, Lactococcus phage 936
sensu lato, Carrot red leaf luteovirus,Mycobacterium
phage Hamulus,Pseudomonas phage JG024, were pre-
sent at low relative abundance among the poultry dust
samples (Fig. 3).
Bacterial species composition is signiﬁcantly different
between aerosolized and settled dust
A statistical analysis of sample relative abundance was
performed using the Linear discriminant analysis Effect
Size (LEfSe) demonstrated ‘features’(i.e., taxonomic rel-
ative abundances) of samples that may be used as a
predictor of either LS, MC or settled dust (Fig. 4). The
features shown in Fig. 4 are statistically different
between dust types using a Kruskal–Wallis test
(P<0.05) and are subsequently found to have the lar-
gest effect score in a linear discriminant analysis model
between the dust types (log LDA score >3). Speciﬁcally,
settled dust samples are characterized by an enrichment
in the Enterococcus,Bacteroides,Vibrio and Clostridium
genera. In contrast, LS dust is characterized by enrich-
ment in the phylum Actinobacteria, the genus Brachy-
bacterium and family Dermabacteraceae. Furthermore,
MC dust is enriched in the Lachnospiricae family and the
Porphyromonadaceae family. These taxonomic features
could potentially provide reliable ‘biomarkers’of LS, MC
or settled dust in poultry houses. However, further stud-
ies are warranted.
Clustering samples on Bray-Curtis dissimilarity
The Bray-Curtis dissimilarity matrix, a statistical test used
to quantify the compositional dissimilarity between differ-
ent ecological sites, was calculated for all samples.
Then, a principal component analysis (PCA) was per-
formed to reduce the higher dimensional matrix to allow
visualization of the amount of variation between sample
proﬁles (Fig. 5). The LS and MC dust samples showed
little variance among the groups, demonstrated by the
formation of two distinct clusters. However, LS13-15
samples are outliers (which could be caused by either
batch effects or genuine biological diversity). In contrast,
settled dust samples show a greater variance among the
In this study, next generation sequencing was used to
make a comprehensive DNA analysis of poultry dust.
Fig. 2. Microbial taxonomic proﬁle of aerosolized or settled poultry
dust. Relative abundance of (A) domain and (B) phylum taxonomic
proﬁles in inhalable and settled poultry dust was determined by
analysis with the MetaPhlAn2 package. Personal inhalable dust
samples were collected during litter sampling (black) or mortality col-
lection (white). Settled dust (grey) samples were obtained from the
side-walls and curtains of the poultry house.
ª2016 The Authors. Microbial Biotechnology published by John Wiley & Sons Ltd and Society for Applied Microbiology.
Microbial composition of poultry dust 3
Metagenomics analysis of inhalable poultry dust demon-
strated that the Gram-positive bacteria species, Staphy-
lococcus sp AL1,L. crispatus and S. carnicancri, are
highly abundant in bioaerosols during LS and MC tasks
(Fig. 3). Staphylococcus sp AL1, L. crispatus and Salin-
icoccus carncancri belong to the Firmicutes phylum, the
second most abundant bacterial phylum in poultry dust
(Fig. 2B). Staphylococcus sp AL1 was ﬁrst isolated from
Chinese soy sauce brine fermentation and encodes an
N-acyl homoserine lactone (AHL) lactonase gene
(Chong et al., 2012). AHL lactonases are quorum-
quenching enzymes that inhibits cell-to-cell communica-
tion of bacteria through the quorum sensing signalling of
AHL (Dong and Zhang, 2005). L. crispatus has been
found to colonize the epithelium of the chicken crop,
mature enterocytes at the villus tip of the small intestine,
and in the follicle-associated epithelium (FAE) and
Peyer’s patches of the ileum (Edelman et al., 2002). In
addition, L. crispatus strain ST1 and 134mi inhibited
E. coli 789 adherence to the epithelium of the crop and
ileal FAE (Edelman et al., 2003). S. carnicancri is a
halophilic bacteria isolated from Korean ‘ganjang-
gejang’, raw crabs preserved in soy sauce, that prefers
a salt rich environment of 5–20% (wt/vol) NaCl (Jung
et al., 2010; Hyun et al., 2013). Lu et al. (2003) demon-
strated that Salinococcus spp was present in broiler
chicken litter collected from farms in Northeast Georgia
using 16S rDNA sequencing. Furthermore, S. lentus,
Salinicoccus spp., Lactobacillus spp., Corynebacterium
terium sp. and Brachybacterium spp. were detected in
broiler chicken dust and broiler chicken litter (Fig. 3) (Lu
et al., 2003; Nonnenmann et al., 2010). Staphylococcus
and E. coli were the only bacteria detected across stud-
ies using culture based and molecular techniques that
corresponded to our ﬁndings (Brooks et al., 2010; Gerald
et al., 2014). Clearly, the phenomenon of culture bias is
limiting the reported compositional diversity of poultry
Fig. 3. Microbial relative abundance in aerosolized or settled poultry dust. Heat map was generated using MetaPhlAn2 and illustrates the 25
most abundant species in inhalable litter sampling dust (LS1-15) (white), settled dust (SD1-3) (grey), inhalable mortality collection dust (MC1-3)
(white) collected on broiler chicken farms. Columns represent the relationship between dust samples, and rows represent the relative abun-
dance of species in the poultry dust.
4K. M. O’Brien et al.
dust in past studies. The results from our study suggest
that poultry production facilities have a diverse and
highly complex microbial ecosystem which varies sub-
stantially between airborne and settled dust.
Both workers and broiler chickens suffer from expo-
sure to dust generated during broiler chicken production.
Poultry workers have a higher prevalence of acute and
chronic respiratory diseases (e.g., occupational asthma
and chronic bronchitis) than any other agricultural occu-
pation, which has been attributed to the induction of the
inﬂammatory response by endotoxin, a constituent of
Gram-negative bacterial cell wall, in organic dust (Clark
et al., 1983; Simpson et al., 1998; Donham et al., 2000;
Malireddy et al., 2013; Viegas et al., 2013). For instance,
inhalation of LPS, a marker of endotoxin, induced poly-
morphonuclear neutrophil and myeloperoxidase in the
blood and sputum of healthy subjects 24 h post-expo-
sure (Thorn and Rylander, 1998). In addition, endotoxin
was shown to be a predictor of chronic phlegm in poultry
workers (Kirychuk et al., 2006). For these reasons, an
occupational exposure limit of 614 endotoxin units
has been recommended for poultry workers by
(Donham et al. 2000). However, the inhibition of endo-
toxin activity in agricultural dust did not attenuate the
inﬂammatory response in lung epithelial cells nor human
peripheral blood monocytes (Romberger et al., 2002;
Poole et al., 2010; Gottipati et al., 2015). Muramic acid,
a marker of the bacterial cell wall component PGN and
Gram-positive bacteria, has a higher concentration in
large animal agricultural dust (i.e., swine and dairy) than
human house dust (Poole et al., 2010). Our results
Fig. 4. Indicators of either aerosolized or settled dust based on bacterial taxonomical proﬁle. Linear discriminant analysis Effect Size (LEfSe)
greater than 3 (P<0.05) was used to illustrate the statistical differences in relative abundances between inhalable litter sampling dust (black),
settled dust (grey) and inhalable mortality collection dust (white). The analysis can discriminate based on ‘features’of the abundance proﬁle
including phyla (p), class (c), order (o), family (f), genus (g), species (s) and strain (t).
Fig. 5. Relationship between the sources of poultry dust. Ordination
graph was derived from the Bray-Curtis dissimilarity matrix calcu-
lated in Python. Principal component analysis shows the distances
between microbial compositions in inhalable litter sampling dust
(black), settled dust (grey) and inhalable mortality collection dust
(white). The distance between samples is proportional to their simi-
larity of microbial composition.
Microbial composition of poultry dust 5
further suggest that Gram-positive bacteria should be
used for future mechanistic studies. Furthermore, the
vast majority of studies elucidating the agricultural dust
induced inﬂammatory mechanism in the respiratory tract
have focused on aqueous or organic settled dust
extracts (Poole et al., 2010; Gottipati et al., 2015). Our
results demonstrate a signiﬁcant difference of microbial
diversity and abundance between airborne and settled
poultry dust (Figs 3–5). These results suggest that
future in vitro and in vivo toxicological studies should
be performed to differentiate the immunomodulatory
effects between bacteria in either airborne dust or set-
tled dust. In addition, this study provides direction for
selecting bacteria species as potential targets in future
exposure characterization and respiratory toxicological
There are potential limitations that may affect a study
of this magnitude and scope. First, batch effects can
occur when undertaking a study involving multiple sites
spread throughout the Southeast region of the United
States. Different environmental and housing conditions
may impact the microbial populations present in broiler
production in ways that are presently not well under-
stood. For example, variation between samples may be
partially due to differences in litter treatment and/or bed-
ding, age of birds, and environmental conditions (e.g.,
temperature, relative humidity and poultry facility).
Although whole-genome shotgun sequencing is an extre-
mely powerful technique to characterize the ecological
diversity of any environment, it cannot distinguish
between viable and non-viable organisms in the sample.
Therefore, the relative abundance of viable microorgan-
isms in poultry dust needs to be further evaluated by
speciﬁc cell culture assays.
Despite these potential sources of variability, we note
that there is a robust consistency in the taxonomic pro-
ﬁles reported by the analysis that provides strong sup-
port for the general taxonomic proﬁle of poultry
production environments as reported. More studies will
be needed to assess the role of environmental sources
of variability on the taxonomic composition in poultry pro-
In conclusion, our ﬁndings demonstrate that poultry
dust is predominately composed of bacteriophage
viruses and Gram-positive bacteria with Staphylococcus
sp. AL1,L. crispatus and S. carnicancri having the high-
est relative abundance among the airborne samples. In
addition, these results show that the microbial composi-
tion of poultry dust varies between airborne and settled
dust as well as working conditions. Our results provide
an in-depth analysis of different types of poultry dust
which can be used to elucidate mediators of the inﬂam-
matory response in poultry workers’respiratory tract and
provide guidance for the development of engineering
controls designed for microorganisms present in broiler
chicken production facilities.
In this study, we collected either inhalable dust or settled
dust samples on commercial broiler chicken farms in the
southern region of the United States. Speciﬁcally, per-
sonal exposure of inhalable dust samples was collected
among research staff performing LS (LS1-15) or MC
(MC1-3) at 16 commercial broiler chicken farms (15
farms:LS and 1:MC). Litter sampling inhalable dust sam-
ples were collected among poultry farms during April
and May. Seventy-three percent (11/15) of LS inhalable
dust samples and 100% of MC inhalable dust samples
had litter bedding treated with sodium bisulphate (PLT)
(Table S1). Settled dust (S1-3) was collected into a
50 ml conical tube (Fischer Scientiﬁc, Pittsburgh, PA,
USA) from curtains and side-walls of the chicken house
on a university commercial broiler chicken farm during
the month of March. This farm had litter bedding treated
with liquid aluminium sulphate (Al
Clear A7) (Table S1).
Two inhalable dust samplers were used for this study,
Institute of Occupational Medicine (IOM; SKC, Eighty
Four, PA, USA) or Button sampler (SKC). Each sampler
contained a polyvinyl chloride ﬁlter with diameter of
25 mm and pore size 5.0 lm (SKC). The samplers were
located in the personal breathing zone of a research per-
sonnel for times raging between 30 and 90 min depend-
ing upon the time needed to complete either LS or MC.
The air samplers were connected to an AirChek XR5000
air pump (SKC) using an air ﬂow of 2 (IOM) or 4 (Button)
litres per minute depending upon the inhalable sampler
used. Airﬂow was pre- and post-calibrated using a
rotameter (VFB-65; Dwyer, Michigan City, IN, USA) cali-
brated to a primary airﬂow standard (Bios DryCal Defen-
der 510; Mesa Labs, Butler, NJ, USA). After sampling,
ﬁlters were stored in 50 ml polystyrene conical tube
(Fisher Scientiﬁc) and stored at 20°C until further pro-
cessing. Inhalable dust concentrations were determined
Genomic DNA extraction
The inhalable dust and settled dust were allowed to
equilibrate to room temperature. Genomic DNA was
extracted from samples having a minimum of 0.500 mg
of dust on the ﬁlters (Table S2). The samples were
added to the tubes of a commercial DNA extraction kit,
PowerSoil DNA Isolation Kit (MoBio Laboratories,
6K. M. O’Brien et al.
Carlsbad, CA, USA). Genomic DNA was extracted per
manufacturer’s instructions. Contaminants were removed
using Agencourt Ampure XT Beads (Beckman Coulter,
Pasadena, CA, USA). Genomic DNA was quantiﬁed
using Quant-iT PicoGreen dsDNA assay kit (Life Tech-
nologies, Carlsbad, CA, USA). All dust samples had
genomic DNA quantity above the required 1 ng for Illu-
mina 2500 library preparation (Table S2).
Whole genome sequencing
Twenty-one dual-indexed libraries were prepared using
Nextera XT DNA library prep kit (Illumina, San Diego,
CA, USA) per manufacturers’instructions. Libraries were
combined into a single pool sequenced using a 125 bp
paired-end run on a HiSeq 2500 instrument (Illumina)
using version 4 chemistry located at the University of
Minnesota Genomics Center (Minneapolis, MN, USA).
Lane yields were greater than 220 M reads/lane. All
expected barcodes were detected in the demultiplexing
report and were well balanced. Library inserts averaged
500 bp. Sequence data used in this analysis have been
deposited to the NCBI Sequence Read Archive (Study
Accession SRP075218) (Table S3).
Sample genomic composition analysis
FastQ Screen (Babraham Bioinformatics; http://
www.bioinformatics.babraham.ac.uk) was used to identify
the sequence composition of the raw data using the bow-
tie2 aligner (Langmead and Salzberg, 2012) with a
500 000 sequence subset and default parameters. Refer-
ence genomes for human (GRCh37), chicken (galGal4),
corn (AGPv3), soybean (GM01) and phage PhiX (1993-
04-28) were downloaded from iGenomes (https://sup
me.html). The rRNA reference database was downloaded
from greengenes (http://greengenes.lbl.gov), and adapter
sequences were collated from publicly available Illumina
protocols for MiSeq and HiSeq experiments (Illumina).
Although FastQ Screen identiﬁed non-microbial, non-adap-
ter sequences present at low levels in the samples, these
were not speciﬁcally ﬁltered as the probability of aligning to
the microbial clade-speciﬁc reference databases used for
downstream analysis was negligible (Segata et al., 2012).
WGS read quality control
The FastQC tool (Babraham Bioinformatics; http://
www.bioinformatics.babraham.ac.uk) was used to assess
sequence quality for Illumina paired-end dataset prior to
downstream analysis. Data for each sample consisted of
forward and reverse reads of length 126 nt and average
GC content of 46.2%. Read quality was generally very
high at all positions within the read. The introduction of
sequencing errors towards the end of Illumina short
reads can be a concern for accurate taxonomic classiﬁ-
cation (Camarinha-Silva et al., 2014). The marker gene
approach applied in this work relies on many dozens of
markers to identify each species, making the introduction
of random sequencing errors near the end of reads less
of a concern as compared with methods that rely on one
region to assign taxonomic classiﬁcation. Thus, we
decided not to truncate reads unless quality scores were
below a threshold value (described below). FastQC iden-
tiﬁed adapter contamination from the Nextera XT trans-
posase kit (Illumina). Forward and reverse reads were
left unpaired, but were concatenated together into a sin-
gle FASTQ ﬁle for quality control processing and down-
stream analysis. Trimmomatic (Bolger et al., 2014) was
used to remove adapter contamination from the reads as
well as further improve the mean quality scores across
all reads prior to taxonomic proﬁling. The software set-
tings were ‘ILLUMINACLIP:./adapters.fa:2:30:10’to
remove the Nextera adapter sequences and ‘LEADING:3
TRAILING:3 SLIDINGWINDOW:4:20’to improve mean
sequence quality by trimming leading and lagging bases
with Q <3 and any sequences with a four-base sliding
window mean below Q20.
Taxonomic abundance proﬁling of WGS reads
Microbial relative abundance proﬁles were generated with
MetaPhlAn2 (Segata et al., 2012) using default parame-
ters and bowtie2 alignment on a single 10-core node of an
Intel(R) Xeon(R) E7-4850 (2.0 GHz) server with 120GB of
physical (RAM) memory. A single lane of forward and
reverse reads (~13 million read pairs) was provided as
input for each sample to MetaPhlan2. The MetaPhlAn2
reference database consists of clade-speciﬁc marker
genes (i.e., genes that unambiguously characterize a tax-
onomic clade as they are always present in the
sequenced isolates of that clade and never present in any
other sequenced organism) from ~17 000 reference gen-
omes (79% bacteria/archea, 20.4% viral and 0.6% eukary-
otic). Proﬁles for all samples were merged with the script
‘merge_metaphlan_tables.py’included with the MetaPh-
lAn2 distribution and heatmaps were generated with
‘metaphlan_hclust_heatmap.py’script using default
options and the ‘-d braycurtis’,‘-minv 0.01’,‘-c bbcyr’
ﬂags. Pie plots were generated with the script
‘metaphlan2krona.py’and the KronaTools tool suite
(Ondov et al., 2011).
Principal component analysis
Principal component analysis of the sample abundance
proﬁles was carried out by ﬁrst merging MetaPhlan2
Microbial composition of poultry dust 7
results into a single BIOM format table. A Qiime (Capo-
raso et al., 2010) utility script for calculating the Bray-Cur-
tis (Bray and Curtis, 1957) dissimilarity matrix on BIOM-
formatted taxonomic proﬁles was then used. The resulting
dissimilarity matrix was imported into Python; calculation of
PCA was performed with the scikit-learn ‘decomposition’
library (Pedregosa et al.,2011).Theﬁrst principal compo-
nent contained 65% of the variation and the second princi-
pal component contained 9% of the variation in the data.
Graphical analysis was performed in Python.
Statistical biomarker analysis
A statistical analysis of differentially abundant features
between classes of samples was accomplished with the
LEfSe statistical analysis package. A table of species
abundances for all samples was ﬁrst prepared for analy-
sis using the ‘format_input.py’script with the ‘-o
1000000’option to normalize all values to one million as
recommended by the package authors. Three classes
were deﬁned as LS (15 samples), MC (3 samples) and
S (3 samples). Running the analysis was accomplished
with the ‘run_lefse.py’script using a default P-value of
0.05 and the ‘-y l’option to specify the ‘all-versus-all’
comparison for the initial class-level Kruskal–Wallis test
prior to model building with LDA (Segata et al., 2011).
The internal Wilcoxon test on subclasses was not per-
formed because no subclasses were deﬁned in the data.
The cut-off value of three or greater for the log LDA
effect size is arbitrary, selected to limit the features iden-
tiﬁed by the analysis to a reasonable number of the most
signiﬁcant features for plotting.
Boissy, R.J., Romberger, D.J., Roughead, W.A., Weis-
senburger-Moser, L., Poole, J.A., and LeVan, T.D. (2014)
Shotgun pyrosequencing metagenomic analyses of dusts
from swine conﬁnement and grain facilities. PLoS ONE 9:
Bolger, A.M., Lohse, M., and Usadel, B. (2014) Trimmo-
matic: a ﬂexible trimmer for Illumina sequence data. Bioin-
formatics 30: 2114–2120.
Bray, J.R., and Curtis, J.T. (1957) An ordination of the
upland forest communities of southern wisconsin. Ecol
Monogr 27: 325–349.
Brooks, J.P., McLaughlin, M.R., Schefﬂer, B., and Miles,
D.M. (2010) Microbial and antibiotic resistant constituents
associated with biological aerosols and poultry litter within
a commercial poultry house. Sci Total Environ 408:
Camarinha-Silva, A., Jauregui, R., Chaves-Moreno, D.,
Oxley, A.P., Schaumburg, F., Becker, K., et al. (2014)
Comparing the anterior nare bacterial community of two
discrete human populations using Illumina amplicon
sequencing. Environ Microbiol 16: 2939–2952.
opez, M., Aarnink, A.J.A., Zhao, Y., Calvet, S.,
and Torres, A.G. (2010) Airborne particulate matter from
livestock production systems: a review of an air pollution
problem. Environ Pollut 158: 1–17.
Caporaso, J.G., Kuczynski, J., Stombaugh, J., Bittinger, K.,
Bushman, F.D., Costello, E.K., et al. (2010) QIIME allows
analysis of high-throughput community sequencing data.
Nat Methods 7: 335–336.
Chen, P.S., Tsai, F.T., Lin, C.K., Yang, C.Y., Chan, C.C.,
Young, C.Y., and Lee, C.H. (2010) Ambient inﬂuenza and
avian inﬂuenza virus during dust storm days and back-
ground days. Environ Health Perspect 118: 1211–1216.
Chong, T.M., Tung, H.J., Yin, W.F., and Chan, K.G. (2012)
Insights from the genome sequence of quorum-quenching
Staphylococcus sp. strain AL1, isolated from traditional
Chinese soy sauce brine fermentation. J Bacteriol 194:
Clark, S., Rylander, R., and Larsson, L. (1983) Airborne
bacteria, endotoxin and fungi in dust in poultry and
swine conﬁnement buildings. Am Ind Hyg Assoc J 44:
Dong, Y.H., and Zhang, L.H. (2005) Quorum sensing and
quorum-quenching enzymes. J Microbiol 43: 101–109.
Donham, K.J., Cumro, D., Reynolds, S.J., and Merchant,
J.A. (2000) Dose-response relationships between occupa-
tional aerosol exposures and cross-shift declines of lung
function in poultry workers: recommendations for expo-
sure limits. J Occup Environ Med 42: 260–269.
Edelman, S., Westerlund-Wikstrom, B., Leskela, S., Ket-
tunen, H., Rautonen, N., Apajalahti, J., and Korhonen,
T.K. (2002) In vitro adhesion speciﬁcity of indigenous Lac-
tobacilli within the avian intestinal tract. Appl Environ
Microbiol 68: 5155–5159.
Edelman, S., Leskel€
a, S., Ron, E., Apajalahti, J., and Korho-
nen, T.K. (2003) In vitro adhesion of an avian pathogenic
Escherichia coli O78 strain to surfaces of the chicken
intestinal tract and to ileal mucus. Vet Microbiol 91: 41–56.
Gerald, C., McPherson, C., Xu, Z., Holmes, B., Williams, L.,
Whitley, N., and Waterman, J.T. (2014) A biophysiochemi-
cal analysis of settled livestock and poultry housing dusts.
American Journal of Agricultural and Biological Sciences
Gottipati, K.R., Bandari, S.K., Nonnenmann, M.W., Levin,
J.L., Dooley, G.P., Reynolds, S.J., and Boggaram, V.
(2015) Transcriptional mechanisms and protein kinase
signaling mediate organic dust induction of IL-8 expres-
sion in lung epithelial and THP-1 cells. Am J Physiol Lung
Hyun, D.W., Whon, T.W., Cho, Y.J., Chun, J., Kim, M.S.,
Jung, M.J., et al. (2013) Genome sequence of the moder-
ately halophilic bacterium Salinicoccus carnicancri type
strain Crm(T) (=DSM 23852(T)). Stand Genomic Sci 8:
Jung, M.J., Kim, M.S., Roh, S.W., Shin, K.S., and Bae, J.W.
(2010) Salinicoccus carnicancri sp. nov., a halophilic bac-
terium isolated from a Korean fermented seafood. Int J
Syst Evol Microbiol 60: 653–658.
Just, N., Kirychuk, S., Gilbert, Y., Letourneau, V., Veillette, M.,
Singh, B., and Duchaine, C. (2011) Bacterial diversity char-
acterization of bioaerosols from cage-housed and ﬂoor-
housed poultry operations. Environ Res 111: 492–498.
8K. M. O’Brien et al.
Kirychuk, S.P., Dosman, J.A., Reynolds, S.J., Willson, P.,
Senthilselvan, A., Feddes, J.J., et al. (2006) Total dust
and endotoxin in poultry operations: comparison between
cage and ﬂoor housing and respiratory effects in workers.
J Occup Environ Med 48: 741–748.
Kwon, Y.M., Woodward, C.L., Pillai, S.D., Pe~
na, J., Corrier,
D.E., Byrd, J.A., and Ricke, S.C. (2000) Litter and aerosol
sampling of chicken houses for rapid detection of Sal-
monella typhimurium contamination using gene ampliﬁca-
tion. J Ind Microbiol Biot 24: 379–382.
Lai, H.T., Nieuwland, M.G., Kemp, B., Aarnink, A.J., and Par-
mentier, H.K. (2009) Effects of dust and airborne dust com-
ponents on antibody responses, body weight gain, and
heart morphology of broilers. Poult Sci 88: 1838–1849.
Langmead, B., and Salzberg, S.L. (2012) Fast gapped-read
alignment with Bowtie 2. Nat Meth 9: 357–359.
Lenhart, S.W., and Olenchock, S.A. (1984) Sources of res-
piratory insult in the poultry processing industry. Am J Ind
Med 6: 89–96.
Lu, J., Sanchez, S., Hofacre, C., Maurer, J.J., Harmon,
B.G., and Lee, M.D. (2003) Evaluation of broiler litter with
reference to the microbial composition as assessed by
using 16S rRNA and functional gene markers. Appl Envi-
ron Microbiol 69: 901–908.
Malireddy, S., Lawson, C., Steinhour, E., Hart, J., Kotha,
S.R., Patel, R.B., et al. (2013) Airborne agricultural partic-
ulate matter induces inﬂammatory cytokine secretion by
respiratory epithelial cells: mechanisms of regulation by
eicosanoid lipid signal mediators. Indian J Biochem Bio-
phys 50: 387–401.
Nonnenmann, M.W., Bextine, B., Dowd, S.E., Gilmore, K.,
and Levin, J.L. (2010) Culture-independent characteriza-
tion of bacteria and fungi in a poultry bioaerosol using
pyrosequencing: a new approach. J Occup Environ Hyg
Ondov, B.D., Bergman, N.H., and Phillippy, A.M. (2011)
Interactive metagenomic visualization in a Web browser.
BMC Bioinformatics 12: 385.
Oppliger, A., Charriere, N., Droz, P.O., and Rinsoz, T.
(2008) Exposure to bioaerosols in poultry houses at differ-
ent stages of fattening; use of real-time PCR for airborne
bacterial quantiﬁcation. Ann Occup Hyg 52: 405–412.
Oyetunde, O.O., Thomson, R.G., and Carlson, H.C. (1978)
Aerosol exposure of ammonia, dust and Escherichia coli
in broiler chickens. Can Vet J 19: 187–193.
Palmberg, L., Larsson, B.M., Malmberg, P., and Larsson, K.
(1998) Induction of IL-8 production in human alveolar
macrophages and human bronchial epithelial cells in vitro
by swine dust. Thorax 53: 260–264.
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V.,
Thirion, B., Grisel, O., et al. (2011) Scikit-learn: machine
learning in Python. J Mach Learn Res 12: 2825–2830.
Poole, J.A., and Romberger, D.J. (2012) Immunological and
inﬂammatory responses to organic dust in agriculture.
Curr Opin Allergy Clin Immunol 12: 126–132.
Poole, J.A., Dooley, G.P., Saito, R., Burrell, A.M., Bailey,
K.L., Romberger, D.J., et al. (2010) Muramic acid, endo-
toxin, 3-hydroxy fatty acids, and ergosterol content
explain monocyte and epithelial cell inﬂammatory
responses to agricultural dusts. J Toxicol Environ Health
Prester, L., and Macan, J. (2010) Determination of Alt a 1
(Alternaria alternata) in poultry farms and a sawmill using
ELISA. Med Mycol 48: 298–302.
Radon, K., Weber, C., Iversen, M., Danuser, B., Pedersen,
S., and Nowak, D. (2001) Exposure assessment and lung
function in pig and poultry farmers. Occup Environ Med
Radon, K., Danuser, B., Iversen, M., Monso, E., Weber, C.,
Hartung, J., et al. (2002) Air contaminants in different
European farming environments. Ann Agric Environ Med
Redente, E.F., and Massengale, R.D. (2006) A systematic
analysis of the effect of corn, wheat, and poultry dusts on
interleukin-8 production by human respiratory epithelial
cells. J Immunotoxicol 3: 31–37.
Riddell, C., Schwean, K., and Classen, H.L. (1998) Inﬂam-
mation of the bronchi in broiler chickens, associated with
barn dust and the inﬂuence of barn temperature. Avian
Dis 42: 225–229.
Romberger, D.J., Bodlak, V., Von Essen, S.G., Mathisen, T.,
and Wyatt, T.A. (2002) Hog barn dust extract stimulates IL-
8 and IL-6 release in human bronchial epithelial cells via
PKC activation. J Appl Physiol (1985) 93: 289–296.
Rylander, R., and Carvalheiro, M.F. (2006) Airways inﬂam-
mation among workers in poultry houses. Int Arch Occup
Environ Health 79: 487–490.
Segata, N., Izard, J., Waldron, L., Gevers, D., Miropolsky,
L., Garrett, W.S., and Huttenhower, C. (2011) Metage-
nomic biomarker discovery and explanation. Genome Biol
Segata, N., Waldron, L., Ballarini, A., Narasimhan, V., Jous-
son, O., and Huttenhower, C. (2012) Metagenomic micro-
bial community proﬁling using unique clade-speciﬁc
marker genes. Nat Methods 9: 811–814.
Simpson, J.C., Niven, R.M., Pickering, C.A., Fletcher, A.M.,
Oldham, L.A., and Francis, H.M. (1998) Prevalence and
predictors of work related respiratory symptoms in work-
ers exposed to organic dusts. Occup Environ Med 55:
Sonesson, A., Larsson, L., Fox, A., Westerdahl, G.,
and Odham, G. (1988) Determination of environmental
levels of peptidoglycan and lipopolysaccharide using
gas chromatography with negative-ion chemical-ioniza-
tion mass spectrometry utilizing bacterial amino acids
and hydroxy fatty acids as biomarkers. J Chromatogr
Spekreijse, D., Bouma, A., Koch, G., and Stegeman, A.
(2013) Quantiﬁcation of dust-borne transmission of highly
pathogenic avian inﬂuenza virus between chickens. Inﬂu-
enza Other Respir Viruses 7: 132–138.
Thedell, T.D., Mull, J.C., and Olenchock, S.A. (1980) A brief
report of Gram-negative bacterial endotoxin levels in air-
borne and settled dusts in animal conﬁnement buildings.
Am J Ind Med 1: 3–7.
Thorn, J., and Rylander, R. (1998) Inﬂammatory response
after inhalation of bacterial endotoxin assessed by the
induced sputum technique. Thorax 53: 1047–1052.
Viegas, S., Faisca, V.M., Dias, H., Clerigo, A., Carolino, E.,
and Viegas, C. (2013) Occupational exposure to poultry
dust and effects on the respiratory system in workers. J
Toxicol Environ Health A 76: 230–239.
Microbial composition of poultry dust 9
Zuskin, E., Mustajbegovic, J., Schachter, E.N., Kern, J.,
Rienzi, N., Goswami, S., et al. (1995) Respiratory function
in poultry workers and pharmacologic characterization of
poultry dust extract. Environ Res 70: 11–19.
Additional Supporting Information may be found online in
the supporting information tab for this article:
Table S1 Environmental and Chicken House Conditions at
Time of Sampling. During litter sampling and mortality col-
lection, personal exposure to inhalable dust was collected
using a personal inhalable dust sampler at a ﬂow rate of 4
L/min. Environmental and house conditions were recorded
at time of sampling.
Table S2 Genomic DNA quantity and total bases
sequenced by Illumina 2500 in Poultry Dust Samples. Dust
concentrations of personal inhalable dust collected during
litter sampling (LS) and mortality collection (MC) were deter-
mined through gravimetrically analysis of polyvinyl chloride
ﬁlters. Genomic DNA was extracted from LS, MC, and set-
tled (S) poultry dust using PowerSoil DNA Isolation Kit.
Genomic DNA was quantiﬁed using the Quant-iT PicoGreen
dsDNA assay kit. Total reads and total gigabases (Gb)
sequenced from the poultry dust samples were obtained
through de novo synthesis of genomic DNA using 125 bp
paired-end Illumina 2500 instrument. *Only one lane out of
ﬁve lanes were used for data analysis. Therefore, approxi-
mately 13 million reads were used for assembly.
Table S3 Biosample Accessions Numbers of Sequencing
Data uploaded into the NCBI Biosample Database.
Sequence data used in this analysis have been deposited
to the NCBI Sequence Read Archive (Study Accession
SRP075218). The sequencing data was uploaded under the
sample number, which corresponds to the following Sample
ID in the manuscript.
10 K. M. O’Brien et al.