Bayesian Taxonomic Assignment for the Next-Generation Metagenomics

“Bayesian Taxonomic Assignment
for the Next-Generation
Metagenomics”
Jonathan A. Eisen
August 7, 2013
DHS Meeting
Wednesday, August 7, 13

Shotgun Metagenomics

DNA

DNA Sequence

DNA Sequence
?????

DNA Sequence
Who is there?
What are they
doing?

• Which communities
are most similar /
different?
• What accounts for the
differences?
• Natural vs. unnatural
• Community level
signatures (of events,
stability,
biogeography, etc)

Our Approach - Phylogeny
Phylogeny of
sequences can reveal
details about history,
taxonomy, function,
and ecology

DNA
extraction
PCR
Sequence
rRNA genes
Sequence alignment = Data
matrix
Phylogenetic tree
PCR
rRNA1
rRNA2
Makes lots of
copies of the
rRNA genes
in sample
rRNA1
5’...ACACACATAGGTGGAGCTA
GCGATCGATCGA... 3’
E. coli
Humans
A
T
T
A
G
A
A
C
A
T
C
A
C
A
A
C
A
G
G
A
G
T
T
C
rRNA1
E. coli Humans
rRNA2
rRNA2
5’..TACAGTATAGGTGGAGCTAG
CGACGATCGA... 3’
rRNA phylotyping
rRNA3
5’...ACGGCAAAATAGGTGGATT
CTAGCGATATAGA... 3’
rRNA4
5’...ACGGCCCGATAGGTGGATT
CTAGCGCCATAGA... 3’
rRNA3 C A C T G T
rRNA4 C A C A G T
Yeast T A C A G T
Yeast
rRNA3
rRNA4
Phylotyping

Uses of Phylogeny in Metagenomics
• Taxonomic assessment
• Phylogenetic OTUs
• Phylogenetic taxonomy assignment
• Phylogenetic binning
• Sample comparisons and hypothesis testing
• Alpha diversity (i.e., PD)
• Beta diversity
• Trait evolution
• Dispersal
• Functional predictions
• Rates of evolution
• Convergence

Venter et al., Science 304: 66. 2004
rRNA Phylotyping - Sargasso Metagenome

RecA Phylotyping - Sargasso Metagenome

0
0.125
0.250
0.375
0.500
Alphaproteobacteria
Betaproteobacteria
G
am
m
aproteobacteria
Epsilonproteobacteria
Deltaproteobacteria
C
yanobacteriaFirm
icutesActinobacteria
C
hlorobi
C
FB
C
hloroflexiSpirochaetesFusobacteria
Deinococcus-Therm
us
EuryarchaeotaC
renarchaeota
Sargasso Phylotypes
Weighted%ofClones
Major Phylogenetic Group
EFG EFTu HSP70 RecA RpoB rRNA
Phylotyping - Sargasso Metagenome

GOS 1
GOS 2
GOS 3
GOS 4
GOS 5
Phylogenetic ID of Novel Lineages
Wu et al PLoS One 2011

Wu et al. 2006 PLoS Biology 4: e188.
Baumannia makes vitamins and cofactors
Sulcia makes amino acids
Phylogenetic Binning

Phylogenetic Functional Prediction

Sequencing Revolution
• More Samples
• Deeper sequencing
• The rare biosphere
• Relative abundance estimates
• More samples (with barcoding)
• Times series
• Spatially diverse sampling
• Fine scale sampling

http://phylosift.wordpress.com
PhyloSift
Supported by DHS Grant

Acknowledgements
Jonathan
Eisen
Students and other staff:
- Eric Lowe, John Zhang, David Coil
Open source community:
- BLAST, LAST, HMMER, Infernal, pplacer, Krona,
metAMOS, Bioperl, Bio::Phylo, JSON, etc. etc.
PhyloSift is open source software:
- Website: http://phylosift.wordpress.org
- Code: http://github.com/gjospin/phylosift
Erick Matsen
FHCRC
Todd Treangen
BNBI, NBACC
Holly
Bik
Tiffanie
Nelson
Mark
Brown
Aaron
Darling
Guillaume
Jospin
Supported by DHS Grant

Analysis &
Summary
PhyloSift

Analysis &
Summary
Analysis &
Summary
•Metagenomic reads
•Contigs
•Genes
PhyloSift

Analysis &
Summary
Searching inputs against
reference family DB
PhyloSift

Analysis &
Summary
Align to reference HMMs
for each family
PhyloSift

Analysis &
Summary
Place reads into reference
phylogeny using pplacer
PhyloSift

Analysis &
Summary
Summarize results &
additional analyses
PhyloSift

Output 1: Taxonomy

Taxonomic
summary
plots in
Krona
(Ondov et al
2011)
Taxonomic Summaries (via Krona)

Tree Reconciliation in PhyloSift

Tree Reconciliation in PhyloSift
Environmental
Sequences
Named
Taxa

Output 2: Phylogenetic Tree of Reads

PhyloSift Tree Browsing
Darling et al Submitted
Placement tree from 2 week old infant gut data

Output 3: Edge PCA
 Edge PCA for exploratory data analysis (Matsen and Evans 2013)
 Given E edges and S samples:
− For each edge, calculate difference in placement mass on either side of edge
− Results in E x S matrix
− Calculate E x E covariance matrix
− Calculate eigenvectors, eigenvalues of covariance matrix
 Eigenvector: each value indicates how “important” an edge is in explaining
differences among the S samples
Example calculating a matrix entry for an edge:
This edge gets 5-2=3
mass=5 mass=2

Edge PCA: Identify
lineages that explain most
variation among samples
Matsen and Evans 2013, Darling et al Submitted.
Edge PCA

QIIME and Edge PCA on
110 fecal metagenomes from
Yatsunenko et al 2012
Nature.
Sequenced with 454, to
about 150Mbp/metagenome
Darling et al
Submitted.
Edge PCA vs. UNIFRAC PCA

Output 4: Forensics

Analysis &
Summary
Analysis &
Summary
•Metagenomic reads
•Contigs
•Genes
PhyloSift
Challenge - Short Non
Overlapping Reads

Markers
• PMPROK – Dongying Wu’s Bac/Arch
markers
• Eukaryotic Orthologs – Parfrey 2011 paper
• 16S/18S rRNA
• Mitochondria - protein-coding genes
• Viral Markers – Markov clustering on
genomes
• Codon Subtrees – finer scale taxonomy
• Extended Markers – plastids, gene families

PMPROK Genes

Analysis &
Summary
PhyloSift
Challenges:
•Limited ref. genomes
•Limited markers, families
Searching inputs against
reference family DB

Improving I: More Markers
Phylogenetic group Genome
Number
Gene
Number
Maker
Candidates
Archaea 62 145415 106
Actinobacteria 63 267783 136
Alphaproteobacteria 94 347287 121
Betaproteobacteria 56 266362 311
Gammaproteobacteria 126 483632 118
Deltaproteobacteria 25 102115 206
Epislonproteobacteria 18 33416 455
Bacteriodes 25 71531 286
Chlamydae 13 13823 560
Chloroflexi 10 33577 323
Cyanobacteria 36 124080 590
Firmicutes 106 312309 87
Spirochaetes 18 38832 176
Thermi 5 14160 974
Thermotogae 9 17037 684
Wu et al. PLOS One 2013. In press.

Representative
Genomes
Extract
Protein
Annotation
All v. All
BLAST
Homology
Clustering
(MCL)
SFams
Align &
Build
HMMs
HMMs
Screen for
Homologs
New
Genomes
Extract
Protein
Annotation
Figure 1
Sharpton et al. 2013
A
B
C
Improving II: More Families

Improving III: Filling in the Tree
Figure from Barton, Eisen et al. “Evolution”, CSHL Press based on Baldauf et al Tree

Genomic Encyclopedia of Bacteria & Archaea
Wu et al. 2009 Nature 462, 1056-1060

Family Diversity vs. PD
Wu et al. 2009 Nature 462, 1056-1060

The Dark Matter of Biology
From Wu et al. 2009 Nature 462, 1056-1060

50
Number of SAGs from Candidate Phyla
OD1
OP11
OP3
SAR406
Site A: Hydrothermal vent 4 1 - -
Site B: Gold Mine 6 13 2 -
Site C: Tropical gyres (Mesopelagic) - - - 2
Site D: Tropical gyres (Photic zone) 1 - - -
Sample collections at 4 additional sites are underway.
Phil Hugenholtz
GEBA Uncultured

JGI Dark Matter Project
environmental
samples (n=9)
isolation of single
cells (n=9,600)
whole genome
amplification (n=3,300)
SSU rRNA gene
based identification
(n=2,000)
genome sequencing,
assembly and QC (n=201)
draft genomes
(n=201)
SAK
HSM ETLTG
HOT
GOM
GBS
EPR
TAETL T
PR
EBS
AK E
SM G TATTG
OM
OT
seawater brackish/freshwater hydrothermal sediment bioreactor
GN04
WS3 (Latescibacteria)
GN01
+Gí
LD1
WS1
Poribacteria
BRC1
Lentisphaerae
Verrucomicrobia
OP3 (Omnitrophica)
Chlamydiae
Planctomycetes
NKB19 (Hydrogenedentes)
WYO
Armatimonadetes
WS4
Actinobacteria
Gemmatimonadetes
NC10
SC4
WS2
Cyanobacteria
:36í2
Deltaproteobacteria
EM19 (Calescamantes)
2FW6SDí )HUYLGLEDFWHULD

GAL35
Aquificae
EM3
Thermotogae
Dictyoglomi
SPAM
GAL15
CD12 (Aerophobetes)
OP8 (Aminicenantes)
AC1
SBR1093
Thermodesulfobacteria
Deferribacteres
Synergistetes
OP9 (Atribacteria)
:36í2
Caldiserica
AD3
Chloroflexi
Acidobacteria
Elusimicrobia
Nitrospirae
49S1 2B
Caldithrix
GOUTA4
6$5 0DULQLPLFURELD

Chlorobi
)LUPLFXWHV
Tenericutes
)XVREDFWHULD
Chrysiogenetes
Proteobacteria
)LEUREDFWHUHV
TG3
Spirochaetes
WWE1 (Cloacamonetes)
70
ZB3
093í
'HLQRFRFFXVí7KHUPXV
OP1 (Acetothermia)
Bacteriodetes
TM7
GN02 (Gracilibacteria)
SR1
BH1
OD1 (Parcubacteria)
:6
OP11 (Microgenomates)
Euryarchaeota
Micrarchaea
DSEG (Aenigmarchaea)
Nanohaloarchaea
Nanoarchaea
Cren MCG
Thaumarchaeota
Cren C2
Aigarchaeota
Cren pISA7
Cren Thermoprotei
Korarchaeota
pMC2A384 (Diapherotrites)
BACTERIA ARCHAEA
archaeal toxins (Nanoarchaea)
lytic murein transglycosylase
stringent response
(Diapherotrites, Nanoarchaea)
ppGpp
limiting
amino acids
SpotT RelA
(GTP or GDP)
+ PPi
GTP or GDP
+ATP
limiting
phosphate,
fatty acids,
carbon, iron
DksA
Expression of components
for stress response
sigma factor (Diapherotrites, Nanoarchaea)
ı4
ȕ ȕ¶
ı2ı3 ı1
-35 -10
Į17'
Į7'
51$ SROPHUDVH
oxidoretucase
+ +e- donor e- acceptor
H
1
Ribo
ADP
+
1+2
O
Reduction
Oxidation
H
1
Ribo
ADP
1+
O
2H
1$' + H 1$'++ + -
HGT from Eukaryotes (Nanoarchaea)
Eukaryota
O
+2+2
OH
1+
2+3
O
O
+2+2
1+
2+3
O
tetra-
peptide
O
+2+2
OH
1+
2+3
O
O
+2+2
1+
2+3
O
tetra-
peptide
murein (peptido-glycan)
archaeal type purine synthesis
(Microgenomates)
PurF
PurD
3XU1
PurL/Q
PurM
PurK
PurE
3XU
PurB
PurP
?
Archaea
adenine guanine
O
+ 12
+
1
1+2
1
1
H
H
1
1
1
H
H
H1 1
H
PRPP )$,$5
IMP
$,$5
A

GUA
G U
G
U
A

G
U
A U
A U
A U
Growing
AA chain
W51$*O

recognizes
UGA
P51$
UGA recoded for Gly (Gracilibacteria)
ribosome
Woyke et al. Nature 2013.

A Genomic Encyclopedia of Microbes (GEM)

Analysis
Summary
for each family
PhyloSift

Analysis
Summary
for each family
PhyloSift
Challenge:
How to align?

Zorro - Automated Masking
cetoTrueTree
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
200 400 800 1600 3200
DistancetoTrueTree
Sequence Length
200
no masking
zorro
gblocks
Wu M, Chatterji S, Eisen JA (2012) Accounting For Alignment Uncertainty
in Phylogenomics. PLoS ONE 7(1): e30288. doi:10.1371/journal.pone.
0030288

Analysis
Summary
PhyloSift

Analysis
Summary
PhyloSift
Challenges:
•Trees from short reads
•Probabilistic methods

Improving IV: Better Reference Tree
Lang et al.
2013

Analysis
Summary
Summarize results
additional analyses
PhyloSift

Phylosift DB Update
Amino Acid
Tree
Run PhyloSift
(search + align)
Execute'dbupdate'mode'
A'taxa'set'is'selected'with'a'
maxPD'cutoﬀ'of'0.02'and'a'new'
tree'is'inferred'
EBI'
Genomes'
Infer Updated
Tree
Add'new'sequences'to'marker'packages'
JGI'
Genomes'
Private'
Genomes'
NCBI'
Genomes'
Nucleotide
Tree
Prune Tree
Update reference
sequences with
new data
New'sequences'added'at'0.25'PD'for'amino'
acid'tree;'higher'PD'threshold'enables'
more'aggressive'searches'of'reference'
database,'since'LAST'searching'is'faster'
with'fewer'sequences.'
Reconcile'NCBI'taxonomy'IDs'with'
phylogeneOc'topologies,'for'both'
amino'acid'tree'and'codon'subtrees'
Tree
Reconciliation
Package
Markers
Users’'local'marker'databases'are'automaOcally'
scanned'each'Ome'PhyloSiR'is'run'and'any'new'
updates'are'automaOcally'downloaded'if'available'
Automated
Download to
PhyloSift Users
Prune Tree
tree'is'inferred'

Improving VI: Other Methods
• PhylOTU
• Kembel all markers
• Kembel copy # correction

Kembel Correction
Kembel SW, Wu M, Eisen JA, Green JL (2012) Incorporating 16S Gene Copy Number Information Improves Estimates
of Microbial Diversity and Abundance. PLoS Comput Biol 8(10): e1002743. doi:10.1371/journal.pcbi.1002743

alignment used to build the profile, resulting in a multiple
sequence alignment of full-length reference sequences and
metagenomic reads. The final step of the alignment process is a
PD versus PID clustering, 2) to explore overlap betw
clusters and recognized taxonomic designations, and
the accuracy of PhylOTU clusters from shotgun re
Figure 1. PhylOTU Workflow. Computational processes are represented as squares and databases are represented as cylinders in
workflow of PhylOTU. See Results section for details.
doi:10.1371/journal.pcbi.1001061.g001
Finding Meta
Sharpton TJ, Riesenfeld SJ, Kembel SW, Ladau J, O'Dwyer JP, Green JL, Eisen JA, Pollard KS. (2011)
PhylOTU: A High-Throughput Procedure Quantifies Microbial Community Diversity and Resolves Novel
Taxa from Metagenomic Data. PLoS Comput Biol 7(1): e1001061. doi:10.1371/journal.pcbi.1001061
PhylOTU

Kembel Combiner
typically used as a qualitative measure because duplicate s
quences are usually removed from the tree. However, the
test may be used in a semiquantitative manner if all clone
even those with identical or near-identical sequences, are i
cluded in the tree (13).
Here we describe a quantitative version of UniFrac that w
call “weighted UniFrac.” We show that weighted UniFrac b
haves similarly to the FST test in situations where both a
FIG. 1. Calculation of the unweighted and the weighted UniFr
measures. Squares and circles represent sequences from two differe
environments. (a) In unweighted UniFrac, the distance between t
circle and square communities is calculated as the fraction of t
branch length that has descendants from either the square or the circ
environment (black) but not both (gray). (b) In weighted UniFra
branch lengths are weighted by the relative abundance of sequences
the square and circle communities; square sequences are weight
twice as much as circle sequences because there are twice as many tot
circle sequences in the data set. The width of branches is proportion
to the degree to which each branch is weighted in the calculations, an
gray branches have no weight. Branches 1 and 2 have heavy weigh
since the descendants are biased toward the square and circles, respe
tively. Branch 3 contributes no value since it has an equal contributio
from circle and square sequences after normalization.
Kembel SW, Eisen JA, Pollard KS, Green JL (2011) The Phylogenetic Diversity of Metagenomes. PLoS
ONE 6(8): e23214. doi:10.1371/journal.pone.0023214

NMF in MetagenomesCharacterizing the niche-space distributions of components
Sites
North American East Coast_GS005_Embayment
North American East Coast_GS002_Coastal
North American East Coast_GS011_Estuary
Eastern Tropical Pacific_GS021_Coastal
North American East Coast_GS006_Estuary
Polynesia Archipelagos_GS051_Coral Reef Atoll
Galapagos Islands_GS036_Coastal
Indian Ocean_GS117a_Coastal sample
Galapagos Islands_GS031_Coastal upwelling
Galapagos Islands_GS030_Warm Seep
Sargasso Sea_GS001c_Open Ocean
Eastern Tropical Pacific_GS022_Open Ocean
Indian Ocean_GS149_Harbor
Indian Ocean_GS123_Open Ocean
Caribbean Sea_GS016_Coastal Sea
Indian Ocean_GS148_Fringing Reef
Indian Ocean_GS112a_Open Ocean
Caribbean Sea_GS017_Open Ocean
Caribbean Sea_GS018_Open Ocean
Indian Ocean_GS108a_Lagoon Reef
Eastern Tropical Pacific_GS023_Open Ocean
Caribbean Sea_GS019_Coastal
Caribbean Sea_GS015_Coastal
Galapagos Islands_GS026_Open Ocean
Polynesia Archipelagos_GS049_Coastal
Polynesia Archipelagos_GS048a_Coral Reef
Component 1
Component 2
Component 3
Component 4
Component 5
0.1 0.2 0.3 0.4 0.5 0.6 0.2 0.4 0.6 0.8 1.0
Salinity
SampleDepth
Chlorophyll
Temperature
Insolation
WaterDepth
General
High
M edium
Low
NA
High
M edium
Low
NA
Water depth
4000m
2000!4000m
900!2000m
100!200m
20!100m
0!20m
4000m
2000!4000m
900!2000m
100!200m
20!100m
0!20m
(a) (b) (c)
Figure 3: a) Niche-space distributions for our ﬁve components (HT
); b) the site-
similarity matrix ( ˆHT ˆH); c) environmental variables for the sites. The matrices are
aligned so that the same row corresponds to the same site in each matrix. Sites are
ordered by applying spectral reordering to the similarity matrix (see Materials and
Methods). Rows are aligned across the three matrices.
Functional biogeography of ocean microbes
revealed through non-negative matrix
factorization Jiang et al. In press PLoS
One. Comes out 9/18.
w/ Weitz, Dushoff,
Langille, Neches,
Levin, etc

Other Uses of PhyloSift
• Integration with other tools (e.g., QIIME)
• LGT detection
• Contamination screening
• Synthetic Biology Orders

w
68
Amino Acid
Tree
Run PhyloSift
(search + align)
Execute'dbupdate'mode'
tree'is'inferred'
EBI'
Genomes'
Infer Updated
Tree
Add'new'sequences'to'marker'packages'
JGI'
Genomes'
Private'
Genomes'
NCBI'
Genomes'
Nucleotide
Tree
Prune Tree
Update reference
sequences with
new data
New'sequences'added'at'0.25'PD'for'amino'
acid'tree;'higher'PD'threshold'enables'
more'aggressive'searches'of'reference'
database,'since'LAST'searching'is'faster'
with'fewer'sequences.'
Reconcile'NCBI'taxonomy'IDs'with'
phylogeneOc'topologies,'for'both'
amino'acid'tree'and'codon'subtrees'
Tree
Reconciliation
Package
Markers
Users’'local'marker'databases'are'automaOcally'
scanned'each'Ome'PhyloSiR'is'run'and'any'new'
updates'are'automaOcally'downloaded'if'available'
Automated
Download to
PhyloSift Users
Prune Tree
tree'is'inferred'

Improving VII: More Samples

The Built Environment
ORIGINAL ARTICLE
Architectural design influences the diversity and
structure of the built environment microbiome
Steven W Kembel1
, Evan Jones1
, Jeff Kline1,2
, Dale Northcutt1,2
, Jason Stenson1,2
,
Ann M Womack1
, Brendan JM Bohannan1
, G Z Brown1,2
and Jessica L Green1,3
1
Biology and the Built Environment Center, Institute of Ecology and Evolution, Department of
Biology, University of Oregon, Eugene, OR, USA; 2
Energy Studies in Buildings Laboratory,
Department of Architecture, University of Oregon, Eugene, OR, USA and 3
Santa Fe Institute,
Santa Fe, NM, USA
Buildings are complex ecosystems that house trillions of microorganisms interacting with each
other, with humans and with their environment. Understanding the ecological and evolutionary
processes that determine the diversity and composition of the built environment microbiome—the
community of microorganisms that live indoors—is important for understanding the relationship
between building design, biodiversity and human health. In this study, we used high-throughput
sequencing of the bacterial 16S rRNA gene to quantify relationships between building attributes and
airborne bacterial communities at a health-care facility. We quantified airborne bacterial community
structure and environmental conditions in patient rooms exposed to mechanical or window
ventilation and in outdoor air. The phylogenetic diversity of airborne bacterial communities was
lower indoors than outdoors, and mechanically ventilated rooms contained less diverse microbial
communities than did window-ventilated rooms. Bacterial communities in indoor environments
contained many taxa that are absent or rare outdoors, including taxa closely related to potential
human pathogens. Building attributes, specifically the source of ventilation air, airflow rates, relative
humidity and temperature, were correlated with the diversity and composition of indoor bacterial
communities. The relative abundance of bacteria closely related to human pathogens was higher
indoors than outdoors, and higher in rooms with lower airflow rates and lower relative humidity.
The observed relationship between building design and airborne bacterial diversity suggests that
we can manage indoor environments, altering through building design and operation the community
of microbial species that potentially colonize the human microbiome during our time indoors.
The ISME Journal advance online publication, 26 January 2012; doi:10.1038/ismej.2011.211
Subject Category: microbial population and community ecology
Keywords: aeromicrobiology; bacteria; built environment microbiome; community ecology; dispersal;
environmental filtering
Introduction
Humans spend up to 90% of their lives indoors
(Klepeis et al., 2001). Consequently, the way we
microbiome—includes human pathogens and com-
mensals interacting with each other and with their
environment (Eames et al., 2009). There have been
few attempts to comprehensively survey the built
The ISME Journal (2012), 1–11
2012 International Society for Microbial Ecology All rights reserved 1751-7362/12
www.nature.com/ismej
Microbial Biogeography of Public Restroom Surfaces
Gilberto E. Flores1
, Scott T. Bates1
, Dan Knights2
, Christian L. Lauber1
, Jesse Stombaugh3
, Rob Knight3,4
,
Noah Fierer1,5
*
1 Cooperative Institute for Research in Environmental Science, University of Colorado, Boulder, Colorado, United States of America, 2 Department of Computer Science,
University of Colorado, Boulder, Colorado, United States of America, 3 Department of Chemistry and Biochemistry, University of Colorado, Boulder, Colorado, United
States of America, 4 Howard Hughes Medical Institute, University of Colorado, Boulder, Colorado, United States of America, 5 Department of Ecology and Evolutionary
Biology, University of Colorado, Boulder, Colorado, United States of America
Abstract
We spend the majority of our lives indoors where we are constantly exposed to bacteria residing on surfaces. However, the
diversity of these surface-associated communities is largely unknown. We explored the biogeographical patterns exhibited
by bacteria across ten surfaces within each of twelve public restrooms. Using high-throughput barcoded pyrosequencing of
the 16 S rRNA gene, we identified 19 bacterial phyla across all surfaces. Most sequences belonged to four phyla:
Actinobacteria, Bacteriodetes, Firmicutes and Proteobacteria. The communities clustered into three general categories: those
found on surfaces associated with toilets, those on the restroom floor, and those found on surfaces routinely touched with
hands. On toilet surfaces, gut-associated taxa were more prevalent, suggesting fecal contamination of these surfaces. Floor
surfaces were the most diverse of all communities and contained several taxa commonly found in soils. Skin-associated
bacteria, especially the Propionibacteriaceae, dominated surfaces routinely touched with our hands. Certain taxa were more
common in female than in male restrooms as vagina-associated Lactobacillaceae were widely distributed in female
restrooms, likely from urine contamination. Use of the SourceTracker algorithm confirmed many of our taxonomic
observations as human skin was the primary source of bacteria on restroom surfaces. Overall, these results demonstrate that
restroom surfaces host relatively diverse microbial communities dominated by human-associated bacteria with clear
linkages between communities on or in different body sites and those communities found on restroom surfaces. More
generally, this work is relevant to the public health field as we show that human-associated microbes are commonly found
on restroom surfaces suggesting that bacterial pathogens could readily be transmitted between individuals by the touching
of surfaces. Furthermore, we demonstrate that we can use high-throughput analyses of bacterial communities to determine
sources of bacteria on indoor surfaces, an approach which could be used to track pathogen transmission and test the
efficacy of hygiene practices.
Citation: Flores GE, Bates ST, Knights D, Lauber CL, Stombaugh J, et al. (2011) Microbial Biogeography of Public Restroom Surfaces. PLoS ONE 6(11): e28132.
doi:10.1371/journal.pone.0028132
Editor: Mark R. Liles, Auburn University, United States of America
Received September 12, 2011; Accepted November 1, 2011; Published November 23, 2011
Copyright: ß 2011 Flores et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported with funding from the Alfred P. Sloan Foundation and their Indoor Environment program, and in part by the National
Institutes of Health and the Howard Hughes Medical Institute. The funders had no role in study design, data collection and analysis, decision to publish, or
preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: noah.fierer@colorado.edu
Introduction
More than ever, individuals across the globe spend a large
portion of their lives indoors, yet relatively little is known about the
microbial diversity of indoor environments. Of the studies that
have examined microorganisms associated with indoor environ-
ments, most have relied upon cultivation-based techniques to
detect organisms residing on a variety of household surfaces [1–5].
Not surprisingly, these studies have identified surfaces in kitchens
and restrooms as being hot spots of bacterial contamination.
Because several pathogenic bacteria are known to survive on
surfaces for extended periods of time [6–8], these studies are of
obvious importance in preventing the spread of human disease.
However, it is now widely recognized that the majority of
microorganisms cannot be readily cultivated [9] and thus, the
overall diversity of microorganisms associated with indoor
communities and revealed a greater diversity of bacteria on
indoor surfaces than captured using cultivation-based techniques
[10–13]. Most of the organisms identified in these studies are
related to human commensals suggesting that the organisms are
not actively growing on the surfaces but rather were deposited
directly (i.e. touching) or indirectly (e.g. shedding of skin cells) by
humans. Despite these efforts, we still have an incomplete
understanding of bacterial communities associated with indoor
environments because limitations of traditional 16 S rRNA gene
cloning and sequencing techniques have made replicate sampling
and in-depth characterizations of the communities prohibitive.
With the advent of high-throughput sequencing techniques, we
can now investigate indoor microbial communities at an
unprecedented depth and begin to understand the relationship
between humans, microbes and the built environment.
In order to begin to comprehensively describe the microbial
the stall in), they were likely dispersed manually after women used
the toilet. Coupling these observations with those of the
distribution of gut-associated bacteria indicate that routine use of
toilets results in the dispersal of urine- and fecal-associated bacteria
throughout the restroom. While these results are not unexpected,
they do highlight the importance of hand-hygiene when using
public restrooms since these surfaces could also be potential
vehicles for the transmission of human pathogens. Unfortunately,
previous studies have documented that college students (who are
likely the most frequent users of the studied restrooms) are not
always the most diligent of hand-washers [42,43].
Results of SourceTracker analysis support the taxonomic
patterns highlighted above, indicating that human skin was the
primary source of bacteria on all public restroom surfaces
examined, while the human gut was an important source on or
around the toilet, and urine was an important source in women’s
restrooms (Figure 4, Table S4). Contrary to expectations (see
above), soil was not identified by the SourceTracker algorithm as
being a major source of bacteria on any of the surfaces, including
floors (Figure 4). Although the floor samples contained family-level
taxa that are common in soil, the SourceTracker algorithm
probably underestimates the relative importance of sources, like
Figure 3. Cartoon illustrations of the relative abundance of discriminating taxa on public restroom surfaces. Light blue indicates low
abundance while dark blue indicates high abundance of taxa. (A) Although skin-associated taxa (Propionibacteriaceae, Corynebacteriaceae,
Staphylococcaceae and Streptococcaceae) were abundant on all surfaces, they were relatively more abundant on surfaces routinely touched with
hands. (B) Gut-associated taxa (Clostridiales, Clostridiales group XI, Ruminococcaceae, Lachnospiraceae, Prevotellaceae and Bacteroidaceae) were most
abundant on toilet surfaces. (C) Although soil-associated taxa (Rhodobacteraceae, Rhizobiales, Microbacteriaceae and Nocardioidaceae) were in low
abundance on all restroom surfaces, they were relatively more abundant on the floor of the restrooms we surveyed. Figure not drawn to scale.
doi:10.1371/journal.pone.0028132.g003
Bacteria of Public Restrooms
high diversity of floor communities is likely due to the frequency of
contact with the bottom of shoes, which would track in a diversity
of microorganisms from a variety of sources including soil, which is
known to be a highly-diverse microbial habitat [27,39]. Indeed,
bacteria commonly associated with soil (e.g. Rhodobacteraceae,
Rhizobiales, Microbacteriaceae and Nocardioidaceae) were, on average,
more abundant on floor surfaces (Figure 3C, Table S2).
Interestingly, some of the toilet flush handles harbored bacterial
related differences in the relative abundances of s
some surfaces (Figure 1B, Table S2). Most notably
were clearly more abundant on certain surfaces
restrooms than male restrooms (Figure 1B). Some
family are the most common, and often most abun
found in the vagina of healthy reproductive age w
and are relatively less abundant in male urine
analysis of female urine samples collected as part
Figure 2. Relationship between bacterial communities associated with ten public restroom surfaces. Communities were
PCoA of the unweighted UniFrac distance matrix. Each point represents a single sample. Note that the floor (triangles) and toilet (as
form clusters distinct from surfaces touched with hands.
doi:10.1371/journal.pone.0028132.g002
Bacteria of P
time, the
un to take
of outside
om plants
ours after
ere shut
ortion of
e human
ck to pre-
which
26 Janu-
Journal,
hanically
had lower
y than ones with open win-
ility of fresh air translated
tions of microbes associ-
an body, and consequently,
pathogens. Although this
hat having natural airflow
Green says answering that
clinical data; she’s hoping
ital to participate in a study
ence of hospital-acquired
they move around. But to quantify those con-
tributions, Peccia’s team has had to develop
new methods to collect airborne bacteria and
extract their DNA, as the microbes are much
less abundant in air than on surfaces.
In one recent study, they used air filters
to sample airborne particles and microbes
in a classroom during 4 days during which
students were present and 4 days during
which the room was vacant. They measured
pant in indoor microbial
ecology research, Peccia
thinks that the field has
yet to gel. And the Sloan
Foundation’s Olsiewski
shares some of his con-
cern. “Everybody’s gen-
erating vast amounts of
data,” she says, but looking across data sets
can be difficult because groups choose dif-
ferent analytical tools. With Sloan support,
though, a data archive and integrated analyt-
ical tools are in the works.
To foster collaborations between micro-
biologists, architects, and building scientists,
the foundation also sponsored a symposium
on the microbiome of the built environment
at the 2011 Indoor Air conference in Austin,
100
80
60
40
20
0
Averagecontribution(%)
DoorinDoorout
StallinStallout
Faucethandles
SoapdispenserToiletseat
ToiletflushhandleToiletfloorSinkfloor
SOURCES
Soil
Water
Mouth
Urine
Gut
Skin
Bathroom biogeography. By
swabbing different surfaces in
public restrooms, researchers
determinedthatmicrobesvaryin
where they come from depend-
ing on the surface (chart).
onFebruary9,2012

Bayesian Taxonomic Assignment for the Next-Generation Metagenomics

Empfohlen

Empfohlen

Weitere ähnliche Inhalte

Was ist angesagt?

Was ist angesagt? (20)

Andere mochten auch

Andere mochten auch (20)

Ähnlich wie Bayesian Taxonomic Assignment for the Next-Generation Metagenomics

Ähnlich wie Bayesian Taxonomic Assignment for the Next-Generation Metagenomics (20)

Mehr von Jonathan Eisen

Mehr von Jonathan Eisen (20)

Kürzlich hochgeladen

Kürzlich hochgeladen (20)

Bayesian Taxonomic Assignment for the Next-Generation Metagenomics