Indoor fungal composition is geographically
patterned and more diverse in temperate
zones than in the tropics
Anthony S. Amenda,1, Keith A. Seifertb, Robert Samsonc, and Thomas D. Brunsa
a
Department of Plant and Microbial Biology, University of California, Berkeley, CA 94720-3102; bBiodiversity Theme, Agriculture and Agri-Food Canada,
Ottawa, ON, Canada K1A 0C6; and cCBS-KNAW Fungal Biodiversity Centre, 3508 AD, Utrecht, The Netherlands
Edited by Steven E. Lindow, University of California, Berkeley, CA, and approved June 2, 2010 (received for review January 12, 2010)
Fungi are ubiquitous components of indoor human environments,
where most contact between humans and microbes occurs. The
majority of these organisms apparently play a neutral role, but
some are detrimental to human lifestyles and health. Recent
studies that used culture-independent sampling methods demonstrated a high diversity of indoor fungi distinct from that of
outdoor environments. Others have shown temporal fluctuations
of fungal assemblages in human environments and modest
correlations with human activity, but global-scale patterns have
not been examined, despite the manifest significance of biogeography in other microbial systems. Here we present a global
survey of fungi from indoor environments (n = 72), using both
taxonomic and phylogeny-informative molecular markers to determine whether global or local indoor factors determine indoor
fungal composition. Contrary to common ecological patterns, we
show that fungal diversity is significantly higher in temperate
zones than in the tropics, with distance from the equator being
the best predictor of phylogenetic community similarity. Fungal
composition is significantly auto-correlated at the national and
hemispheric spatial scales. Remarkably, building function has no
significant effect on indoor fungal composition, despite stark
contrasts between architecture and materials of some buildings
in close proximity. Distribution of individual taxa is significantly
range- and latitude-limited compared with a null model of randomized distribution. Our results suggest that factors driving fungal composition are primarily global rather than mediated by
building design or function.
|
|
|
fungal biogeography indoor mold latitudinal gradient
community structure rRNA metagenomics
| phylogenetic
I
ndoor environments, where the average person in an industrial
nation spends ≈90% of his or her life (1), represent the most
important interface between humans and microbes. Examples of
well-known fungi include a few human pathogens (2), allergens
(3), agents of structural rot (4, 5), and food spoilers (6, 7). Indoor
fungi’s prominent role in successful litigation around the world
contributes to rising costs for various industries and insurance
companies (8, 9). Increasingly strict standards for indoor sanitation have resulted in regulatory agencies and private industries
seeking to quantify building health. Mould surveys that target the
relatively few visibly apparent fungal species or those readily
cultivable on artificial media are now standard, and a US Environmental Protection Agency-developed set of real-time PCR
probes facilitates their quantification (10). However, the recent
rise in fungal infections caused by species formerly considered
benign but now seen as causing disease in immunosuppressed
humans, and a vastly increased resolution of indoor fungal
composition afforded by culture-independent sampling methods
(11, 12), force us to reconsider what constitutes a normal indoor
environment and the factors that shape it.
Recent efforts to describe the processes shaping indoor and urban fungal composition show temporal effects (13–15) and modest
correlations with human activity (16). The existence of global-scale
13748–13753 | PNAS | August 3, 2010 | vol. 107 | no. 31
patterns in fungal composition, however, is unexamined, despite
evidence of biogeographical patterning in other microbial systems.
As with larger organisms, bacterial and archaeal composition is
determined by both the contemporary environment and historical
processes such as dispersal (17). The relative importance of these
factors and the particular environmental variables involved depend
on the taxa and habitat sampled.
For indoor fungi whose association with highly mobile humans
presents opportunities for global dispersal, we hypothesized that
most taxa would be relatively cosmopolitan on a global scale. We
also hypothesized that the local indoor environment (as determined
by building function, construction material, or circulation system)
would play a relatively large role in shaping composition. Finally, as
a result of the combination of presumably high dispersal rates between indoor habitats and the highly selective indoor environment,
we expected little influence of the outdoor environment on
fungal composition.
We test these hypotheses (i) by examining whether the global
distribution patterns of individual taxa show range limitation, (ii)
by examining whether indoor fungal composition correlates with
dispersal limitation or outdoor environment, and (iii) by testing
an alternative hypothesis that indoor environment determines
fungal composition by comparing nearby buildings with different
architecture and function.
Results
Samples and Processing. We collected settled dust samples from
buildings on six continents (n = 72; Fig. 1 and Table S1) using
sterilized collectors attached to vacuum cleaners. We collected
settled dust to maximize time integration of fungi accumulating
over multiple seasons [rather than the fine-scale temporal partitioning provided by air sampling (18)] and used standardized protocols to collect dust samples pooled from “accessible,” “infrequently accessed,” and “inaccessible” areas (i.e., behind heavy
appliances) to include areas not frequently cleaned. The 61 buildings that were sampled were human dwellings, and the rest were
offices, shops, and a church. Extracted DNA was PCR amplified
from two loci within the nuclear ribosomal coding cistron: a fragment containing the internal transcribed spacer region 2 (ITS) and
a fragment containing the D1 and D2 regions of the large subunit
gene (LSU). These were sequenced in multiplex using 454 GS FLX
titanium technology (454 Life Sciences). The inclusion of two loci
enabled two different types of analysis. The ITS region, an inter-
Author contributions: A.S.A., K.A.S., R.S., and T.D.B. designed research; A.S.A. performed
research; A.S.A. analyzed data; and A.S.A. and T.D.B. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Data deposition: Raw sequence data have been deposited in NCBI’s sequence read archive
under accession nos. SRS010093 and SRS010094 (ITS and LSU sequences, respectively).
1
To whom correspondence should be addressed. E-mail: [email protected].
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1000454107/-/DCSupplemental.
www.pnas.org/cgi/doi/10.1073/pnas.1000454107
A
17
3 2
1
1
7
7
7
5
9
32
3
5
68
1
Canada
1
7
1
112
28
35
4
26
2
4
1
17
1
19
26
5
17
57
56
4
22
11
3
2
69
74
23
6
39
3
4
11
4
1
4
2
1
1
1
2
2
4
2
1
3
1
25
13
15
14
2
2
1
2
1
1
1
2
1
1
3
8
6
2
1
21
54
13
8
1
3
34
5
3
8
2
5
15
2
3
3
6
25
73
3
2
1
5
2
4
296
5
3
6
4
4
454
1
7
1
1
4
1
1
21
4
45
3
15
2
4
47
3
4
1
36
3
12
2
6
14
19
4
1
1
19
9
1
12
2
15
1
19
3
1
1
3
1
3
2
9
3
1
1
4
5
1
1
1
3
21
4
4
1
6
2
1
5
3
4
5
2
8
1
1
1
4
2
1
16
1
37
19
1
74
4
17
1
3
32
3
37
2
6
5
3
5
1
12
5
26
8
6
8
1
8
11
1
12
5
29
25
5
22
4
11
5
1
3
7
47
6
25
6
17
11
19
29
21
13
15
9
6
5
5
56
8
5
1
1
2
1
1
1
1
7
4
9
3
5
1
26
3
4
6
2
14
13
1
22
7
9
1
3
1
1
9
1
2
9
2
2
4
4
1
14
9
1
1
2
1
6
8
2
1
1
3
1
1
11
2
2
14
2
5
2
6
3
1
2
1
1
1
9
3
4
1
2
5
1
1
3
4
3
1
7
1
1
5
2
21
1
17
2
6
1
1
2
1
2
2
1
1
96
1
2
3
1
5
1
1
1
1
1
2
1
1
1
13
4
6
5
8
11
7
13
1
4
2
29
72
3
33
19
1
3
7
1
2
26
21
3
2
1
5
13
2
3
7
2
2
5
29
2
28
2
32
2
7
31
7
1
1
7
9
2
1
1
1
5
3
2
2
1
1
1
7
3
35
66
13
26
2
2
2
1
1
3
4
4
2
3
1
2
1
3
5
2
1
3
3
2
1
2
22
2
7
1
2
1
2
1
6
4
1
9
6
12
2
15
1
2
5
1
4
1
2
1
3
96
1
4
2
1
2
3
2
11
3
4
5
2
16
2
11
8
13
3
3
2
12
5
6
1
2
1
9
5
1
6
13
2
7
9
2
1
8
8
1
1
44
29
23
2
3
1
1
56
1
1
14
1
1
35
6
112
5
1
1
2
2
9
36
3
6
2
1
1
939
21
2
1
2
4
1
1
2
1
1
2
1
3
25
3
4
2
1
5
36
2
6
1
1
23
29
27
7
1
2
1
59
25
1
3
3
2
1
5
21
2
12
15
2
5
13
5
2
4
9
4
1
3
13
44
3
8
1
2
11
2
6
5
4
4
22
36
7
2
28
3
16
21
25
65
11
14
3
43
45
1
9
1
1
121
91
444
1
29
12
12
397
3
8
31
65
57
14
11
2
2
65
52
4
1
5
37
12
1
4
1
13
12
21
4
1
1
15
1
3
174
7
7
24
3
1
89
12
7
4
1
3
12
66
11
4
9
7
4
1
8
2
15
3
92
4
9
4
61
16
6
1
11
3
5
1
26
2
16
9
9
1
2
1
4
3
59
1
19
1
6
7
44
9
14
17
5
1
2
12
174
7
4
8
5
1
16
23
1
14
23
35
1
5
26
3
13
9
1
14
9
3
9
2
8
58
37
6
11
6
15
1
49
13
2
9
21
1
1
1
21
1
4
12
57
1
1
3
4
2
11
188
62
7
3
17
4
29
2
1
2
1
1
1
1
3
1
24
7
4
2
74
2
8
1
17
5
7
1
4
4
117
4
3
5
1
8
36
23
1
1
15
1
87
3
14
16
4
62
13
4
39
47
8
5
3
43
24
29
48
1
13
1
12
6
12
15
13
3
2
43
18
27
27
14
1
1
1
2
2
13
1
3
6
86
2
6
6
87
9
112
11
11
22
3
9
1
2
3
3
23
14
7
3
1
1
5
1
2
11
13
12
28
1
35
1
1
6
1
1
37
3
1
2
3
124
9
1
14
6
1
3
2
2
5
5
51
1
6
2
1
12
45
6
3
18
4
1
5
1
11
2
1
2
1
5
2
1
1
11
2
9
3
7
1
7
1
2
11
4
22
4
7
1
513
8
1
2
3
7
4
5
4
4
2
5
1
1
1
1
1
4
1
1
3
1
5
13
2
4
7
2
4
1
5
33
1
3
1
1
2
4
2
1
1
9
4
2
1
16
24
19
44
9
4
2
12
5
1
1
6
14
1
1
14
19
2
4
57
1
37
21
33
6
5
3
11
2
4
1
11
4
1
4
6
1
12
5
25
1
1
1
1
6
1
29
1
2
1
4
1
41
8
1
4
18
15
1
35
7
9
1
3
2
5
1
27
21
2
1
1
5
8
2
1
4
11
1
1
27
6
19
2
9
13
1
15
2
25
1
3
1
1
2
9
2
2
1
1
2
48
5
2
1
27
5
12
1
11
67
2
9
1
1
14
3
1
1
6
1
2
8
3
2
8
1
1
1
1
5
8
3
5
9
8
52
13
17
1
9
2
3
7
7
1
4
1
25
36
11
1
1
8
16
1
1
17
55
23
8
9
1
13
19
2
6
1
7
18
1
22
5
5
25
82
33
16
21
6
1
1
2
19
4
21
3
18
11
13
2
14
15
1
3
2
43
9
35
3
11
4
1
7
277
8
16
8
36
21
18
56
3
1
1
13
2
2
4
2
8
2
5
6
2
3
2
3
11
2
1
7
3
2
12
22
5
1
8
35
16
52
29
3
18
2
4
4
12
14
1
59
12
2
5
5
7
2
6
34
14
11
6
28
4
3
4
13
2
21
1
6
316
1
19
1
1
3
13
12
92
98
2
1
4
29
6
3
2
21
1
1
1
15
2
4
2
4
1
4
5
7
4
2
12
2
1
1
17
2
1
3
2
6
9
3
2
13
24
1
1
2
15
4
2
3
3
47
13
28
3
1
2
6
2
5
13
1
6
3
2
2
1
17
S. Africa
Uruguay
Australia
9
15
1
491
4
2
1
6
2
11
18
3
17
3
Micronesia
Indonesia
1
1
8
5
3
1
1
3
11
18
3
1
19
24
25
18
2
26
2
2
42
1
5
5
14
2
1
7
25
2
4
18
18
2
8
1
4
26
5
6
1
4
4
8
2
3
1
2
42
34
23
11
4
3
57
48
3
1
24
14
44
2
17
2
19
4
2
25
4
4
18
1
11
31
4
2
85
11
1
15
22
25
34
13
62
36
41
1
1
1
13
1
1
4
135
44
52
52
1
5
2
1
2
18
21
1
7
1
1
59
5
59
13
6
1
1
2
4
2
6
55
1
2
22
89
58
91
13
17
6
1
2
2
8
49
35
6
24
2
5
11
1
12
13
9
12
17
2
3
3
1
22
17
14
5
1
1
4
63
35
38
61
53
1
11
1
3
8
5
1
45
13
7
5
6
8
2
127
14
4
1
56
3
2
26
51
15
52
37
7
31
2
4
3
81
17
6
19
2
9
26
61
4
5
42
17
12
3
8
2
3
5
3
1
23
38
4
4
3
3
9
32
31
4
42
9
62
45
85
68
21
2
68
4
4
1
2
2
25
17
17
26
43
9
1
28
49
17
6
5
114
153
45
3
Netherlands
UK
USA
Mexico
15
3
77
56
27
2
2
63
79
1
1
18
1
2
3
1
4
3
8
11
1
1
6
3
1
17
11
2
1
1
1
4
4
1
2
3
1
1
2
5
1
6
2
19
1
9
1
27
1
3
1
6
2
1
4
4
12
41
7
21
1
175
13
22
8
145
7
79
53
43
3
12
8
97
3
16
8
19
3
37
9
16
2
32
3
1
3
53
17
5
2
1
6
3
4
25
7
9
1
15
1
15
1
3
27
7
3
383
6
1
48
3
18
2
33
2
6
3
1
1
8
3
3
5
7
3
2
1
1
2
2
3
5
19
2
3
1
1
1
13
4
4
7
1
8
2
168
1
6
2
1
4
96
1
5
4
4
8
7
4
1
12
1
2
5
2
2
1
1
1
3
1
5
3
Fig. 1. Dust-sampling locations and distributions of the most cosmopolitan OTUs. All samples were collected between December 2008 and March 2009. (A) Solid
lines indicate the Tropic of Cancer and the Tropic of Capricorn. The number of samples analyzed is noted by the open circles (approximate location). (B) The presence
of the 45 OTUs found in the most number of samples (from left to right, most cosmopolitan to least cosmopolitan) is indicated by a solid square (samples used in ITS
analysis are in rows ordered by latitude). Nomenclature is derived from annotations provided for similar sequences deposited in GenBank. Note the following
equivalencies between anamorphic and teleomorphic stages: Cladosporium = Davidiella, Aspergillus = Eurotium, and Gibberella = Fusarium.
genic spacer, is hypervariable and effective for delineating specieslevel taxonomy, but is unsuitable for multiple sequence alignment
and phylogenetic inference at broad levels. Because many environmental sequences derive from unknown or undescribed organisms, operational taxonomic units (OTUs), calculated from pairwise ITS sequence alignment identity, were used instead of species
for diversity estimates. Because it is much better represented in
public databases such as GenBank, the ITS region was used for
taxonomic determinations. The LSU locus, conversely, is phylogenetically informative across fungi, but not divergent enough to
distinguish closely related species (19). Here, we use the LSU locus
for all comparisons of dust sample phylogenetic composition.
Because 454 sequencing of environmental amplicons can lead to
inflated estimates of diversity, partly due to undetected sequence
errors (20, 21), we mitigated the effects of these potential errors by
removing any sequences containing an ambiguous base call and
those where a mis-call occurred in either the priming site or the
multiplex tag. The remaining sequences were screened in the CLC
Genomics Workbench 2 using the modified Mott trimming algorithm (comparable to Phred score processing), which evaluates
sequences on the basis of the 454 quality score. Where base calls
did not meet a cumulative error probability threshold of 0.01, they
were trimmed from the ends of reads (details of the algorithm and
its implementation are available in the CLC Genomics Workbench user manual). Finally, trimmed sequences shorter than 300
bp were removed. The resulting dataset contained 97,557 ITS and
187,668 LSU sequences ranging from 300 to 569 bp (median
length 393 bp). For comparative analyses among samples, only
those with at least 1,200 LSU reads (n = 62; mean 2,764 ± 2,394
SD reads) or 400 ITS reads (n = 65, mean 1,484 ± 1,539 reads)
were included.
To test the effects of artifactual sequence variance on OTU
diversity, we included positive controls consisting of five aliquots
of one dust sample augmented with equal numbers of fungal
spores from individuals of six species along a dilution gradient
(106–10 spores per individual per aliquot; SI Text). Spiked dust
samples were processed and sequenced as described for other
Amend et al.
samples. A total of 1,307 sequences homologous to those of the
added individuals were recovered, and pairwise distances between
sequences within species were calculated at 99.30 ± 0.79%. These
sequences clustered into conspecific OTUs at 80–97% ITS sequence identity, whereas 98%, 99%, and 100% identity thresholds
resulted in 15, 63, and 701 clusters, respectively. Sequence abundance from individuals added in equal spore counts in our positive
control varied from 24 to 633 (Fig. S1). For this reason, we did not
use quantitative metrics weighted by sequence abundance to calculate assemblage dissimilarity.
Geographic Structure of Indoor Fungal Composition. A minimum
evolution phylogenetic tree was constructed from a multiple
alignment of all LSU sequences and used to compare sample
phylogenetic composition as calculated by the unweighted UniFrac metric (22). This method measures phylogenetic dissimilarity
among samples as the proportion of the phylogenetic branch
length unique to each. To correct for differences in sample size,
1,200 sequences were randomly selected from each sample, and
each analysis was calculated with distance matrices derived from
10 such randomizations (mean statistics and SDs among randomization replicates are reported below).
Phylogenetic similarity of fungal composition correlated most
strongly with geography (Table 1 and Fig. 2). Distance from the
equator (DFE) was the strongest predictor of assemblage similarity and was colinear with all other measured variables. We used
partial Mantel tests to determine the significance of the correlation contributed by each variable in isolation, accounting for the
colinearity of the other partial predictors. No variables were significant when DFE was included in partial Mantel models, and
therefore we removed DFE from subsequent analyses. With DFE
removed, pairwise differences between mean annual rainfall,
mean annual temperature, and geographic distance between
samples each correlated with sample phylogenetic dissimilarity.
Analysis of similarity (ANOSIM) (23), a nonparametric multivariate analog of ANOVA, indicated that fungal composition
of samples in the temperate Northern Hemisphere, temperate
PNAS | August 3, 2010 | vol. 107 | no. 31 | 13749
ECOLOGY
B
8
Cladosporium sp.
Davidiella sp.
Eurotium sp.
Epicoccum sp.
Choiromyces sp.
Cladosporium sp.
Cladosporium sp.
Leptospaerulina sp.
Alternaria sp.
Davidiella sp.
Ampellomyces sp.
Choiromyces sp.
Aspergillus sp.
Phoma sp.
Aureobasidium sp.
Cladosporium sp.
Davidiella sp.
Fusarium sp.
Epicoccum sp.
Leptosphaerulina sp.
Alternaria sp.
Alternaria sp.
Aureobasidium sp.
Phoma sp.
Cryptococcus sp.
Cryptococcus sp.
Leptosphaerulina sp.
Cryptococcus sp.
Pleosporaceae sp.
Gibberella sp.
Davidiella sp.
Wallemia sp.
Eurotium sp.
Davidiellaceae sp.
Leptosphaerulina sp.
Alternaria sp.
Davidiella sp.
Cryptococcus sp.
Coniosporium sp.
Cryptococcus sp.
Penicillium sp.
Aureobasidium sp.
Lewia sp.
Epicoccum sp.
Aureobasidium sp.
4
Table 1. Mantel, partial Mantel, and ANOSIM analyses
performed with the phylogenetic composition dissimilarity
distance matrix generated using UniFrac
SD*
P
None
Rain, temperature
Distance, temperature
Distance, rain
0.572
0.138
0.323
0.298
0.008
0.008
0.015
0.008
0.001
0.005
0.001
0.001
None
Regional group
Regional group
Country
0.529
0.614
0.282
0.282
0.010
0.010
0.010
0.010
Southern Hemisphere, or the tropics (regional grouping) were
significantly more similar to other samples within the same group
compared with other samples (Table 1). When we accounted for
colinearity of regional grouping, samples within the same country
were also significantly more similar to each other than those from
different countries. Surprisingly, no significant difference was found
between human dwellings and other building types when colinearity
of either regional group or country was accounted for, despite stark
contrasts between some buildings in close proximity. Some tropical
offices and shops, for example, contained tightly sealed windows
and air conditioning, whereas some nearby homes in the same location had open ventilation. Thus, the observed compositional
variance is consistent with geography and is not an artifact of the
types of buildings sampled.
Phylogenetic and OTU Diversity Gradients. Contrary to the predominant ecological pattern showing biological diversity increasing with proximity to the equator (24–26), our study showed
OTU diversity significantly increased with latitude. (Pearson’s
product moment R = 0.37, t = 3.111, d.f. = 63, P = <0.005; Fig.
3D). Phylogenetic diversity showed a weaker, although significant, pattern (R = 0.32, t = 2.578, d.f. = 60, P = 0.01; Fig. 3E).
Distribution of OTUs and Taxonomic Groups. OTUs showed distinct
patterns of geographic structure. Of 4,473 OTUs defined at 97%
ITS sequence identity (Fig. 3B), only 31 were found in more than
half of the samples. All but 3 of these belonged to the phylum
Ascomycota, and 25 of 31 were in the class Dothideomycetes; Fig.
3A). The dominant Dothidiomycetes genera found here are also
commonly detected indoors in studies using culturing-based methods (6). These genera are mostly composed of species associated
with plants and are commonly found on the surface of leaves and/or
as saprobes in decaying organic matter. These fungi share several
traits, including desiccation-tolerant reproductive structures and
melanized spores. Potentially, these characters that aid in dispersal
and longevity in outdoor environments confer advantages to fungi
on indoor substrates where they commonly grow such as building
materials, carpeting, and food.
Other fungi had more limited distributions. We use a randomized permutation-based technique to delimit spatial patterns
to reduce bias based on abundance. The mean pairwise geographic distance between samples containing the same OTU was
significantly less than expected from a null model of random
distribution (Fig. 4A), indicating overall spatial limitations to
Temperature
Latitude
0.001
0.001
0.134
0.322
Colinear variables accounted for in predictive model are noted as partial
predictors. Significant P values are shown in bold.
*SD of the R-value was calculated among 10 replicates of 1,200 random
draws from each sample.
†
Distance from the equator.
‡
Geographic distance between samples.
13750 | www.pnas.org/cgi/doi/10.1073/pnas.1000454107
Rainfall
USA
Netherlands
Australia
Canada
Uruguay
Stress = 14.81
Dimension 1
B 0.3
0.3
0.2
0.1
0
-0.1
-0.2
-0.3
-0.4
UK
South Africa
Micronesia
Mexico
Indonesia
C
Mantel correlation (R)
ANOSIM
Regional group
Country
Building type
Building type
R
Dimension 2
Mantel
DFE†
Distance‡
Rain
Temperature
Partial predictor
Mantel correlation (R)
Variable of interest
A
0.2
0.1
0
-0.1
-0.2
-0.3
5 10 15 20 25 30 35 40 45
Pairwise distance
(°from Equator)
2 4 6 8 10 12 14 16
Pairwise distance
(km x 104)
Fig. 2. Sample phylogenetic composition similarity is explained by environmental variables. (A) Nonmetric multidimensional scaling plot demonstrates
goodness-of-fit correlations between the first two dimensions and distance
from the equator, mean annual temperature, and mean annual rainfall (all
correlations were significant at P ≤ 0.001). The ordination was calculated using
pairwise unweighted UniFrac phylogenetic distance. Open symbols are indoor
environments that are not dwellings (e.g., shops, hospitals, a church); the broken
line circles tropical samples. Mantel correlograms show assemblage autocorrelation as (B) a factor of pairwise differences in latitude and (C) pairwise geographic distance between samples. Error bars are 95% confidence intervals.
Mean SD of R-values from all distance classes among 10 random draws of 1,200
sequences was 0.005 for both latitude and geographic distance measurements.
indoor fungal ranges. This trend is driven by OTUs whose
detected ranges are less than 2,000 km, the smallest distance
class measured. Of the nonsingleton OTUs, 19.8% were restricted to one of the three regional groups. Conversely, OTUs
occurring in both hemispheres were significantly more likely to
be found at complementary latitudes (on opposite sides of the
equator) than expected (Fig. 4C), indicating that the distribution
of some dispersal-unlimited species is limited by environmental
factors. Similar results were obtained with 90% identity OTU
thresholds (Fig. 4B and D).
We detected indoor fungi that were likely growing within indoor
environments as well as those introduced from outdoors. Fungi that
were clearly “outdoor” additions to the indoor environment included ecto- and arbuscular-mycorrhizal genera such as Amanita
and Glomus, which are obligate associates of host plants. Genera
known only from aquatic environments, such as Rozella, were also
encountered in indoor dust. All of the obvious outdoor additions
were, however, minor components. The presence of mammalian
mycobionts such as Candida albicans, Malasezzia spp., and Pneumocystis spp. showed that humans and other mammals also contribute directly to fungi in indoor dust.
Discussion
This first global survey of indoor fungi demonstrates a previously
undetected high diversity of indoor fungi. Although some of the
most cosmopolitan taxa in this study (such as Alternaria, CladoAmend et al.
A
C
D
E
ECOLOGY
B
Fig. 3. OTU distribution patterns showed little evidence of differentiation at the class level (A). The most cosmopolitan OTUs, occurring in more than half of the
samples (>50%) or in 9 or 10 of the countries sampled, were dominated by taxa in the class Dothideomycetes. Observed OTU rarefactions (B) were calculated at ITS
percentage identity thresholds labeled as 100%, 99%, 97%, 95%, and 90%. A 100% identity rarefaction curve is not contained in the figure, but is calculated to be
54,815. OTU frequency distribution (C) has a long tail with 2,541 (56%) OTUs occurring in a single sample only. Within-sample OTU richness (D; Pearson’s product
moment, R = 0.37, t = 3.57, df = 63, P = <0.001; rarefied to 400 sequences per sample) and phylogenetic diversity, measured as cumulative branch length (E; R = 0.32, t =
2.58, df = 60, P = 0.01; rarefied to 1,200 sequences per sample), were positively correlated with distance from the equator.
sporium, Epicoccum) are also common in culture-based studies,
fast-growing Eurotiomycetes (i.e., Penicillium, Aspergillus, Paecilomyces, and their teleomorphs), Sordariomycetes (Fusarium,
Stachybotrys), and Zygomycota (Mucorales) tend to dominate on
cultivation media. The significance of our results is not merely
the high taxonomic richness of fungi detected indoors, but also
the phylogenetic diversity that includes every major lineage of
fungi. These results force us to broaden our outlook on measuring and, perhaps, regulating normal indoor environments.
Microbial biogeographical theory (17) suggests that biogeographic patterns of species may result from historical contingencies
including dispersal limitation, contemporary environmental selection, or some combination of the above. Our results do not support
any one unified process over the others and instead suggest that
different OTUs are subject to different limitations. We found many
OTUs limited to ranges less than 2,000 km (suggesting historical
contingencies or contemporary environmental selection), others
with wide ranges but limited by latitude (suggesting contemporary
environmental selection but not historical contingencies), and
others with clearly “cosmopolitan” distributions suggesting neither
historical contingencies nor contemporary environmental selection
as constraints. Determining whether fungal distribution patterns
correlate with, e.g., substrate preference, propagule size, trophic
status, etc., may offer a starting point for future hypothesis testing
of factors driving biogeography.
The degree to which these biogeographic patterns detected in
indoor environments apply to outdoor fungal systems remains
untested, although global, disjunct distributions occur in other
microbial taxa, such as psychrophilic cyanobacteria species in
polar regions (27). We might expect indoor fungi to be dispersed
by human-mediated global travel and commerce, whereas outdoor fungi in natural systems are more limited by natural disAmend et al.
persal mechanisms such as wind or splash. Therefore, the strong
geographic patterns that we find in indoor fungal OTUs point
very strongly toward environmental selection rather than dispersal limitation.
We find that latitude has the greatest influence on indoor
fungal compositional variation. This likely reflects correlations
with local outdoor environmental variables such as rainfall and
temperature (and other variables not measured here). These
results are consistent with those from other studies showing
strong correlations between environmental conditions and microbial composition. A global study of Acidobacteria communities in soil, for example, found that community OTU (28) and
phylogenetic (29) similarity was predicted best by pH, whereas
other studies found strong correlations with environmental
parameters such as salinity and nutrient gradients (30–32). Our
study found that distance between samples explains only a minor
component of fungal composition dissimilarity. If dispersal limitation were a more critical factor in shaping indoor fungal
composition, we might expect a stronger correlation as species
turnover results in dissimilarity by distance patterns (33, 34). In
our case, however, where shared phylogenetic branch length
rather than shared species was used to calculate compositional
similarity, dissimilarity by distance may not be a good indicator
of dispersal limitation per se.
One of the most surprising results of our analysis of OTU
distributions is a higher diversity in temperate compared with
tropical zones. Although recent work proposes that similar trends
may exist among ectomycorrhizal fungi (35, 36), this pattern runs
contrary to almost all measured patterns of global biodiversity,
including those of other microbial systems. Pianka’s seminal review of latitudinal diversity gradients in macroflora/fauna (24)
suggests numerous processes that potentially drive a lower diPNAS | August 3, 2010 | vol. 107 | no. 31 | 13751
A0.7
0.6
0.5
0.4
0.3
0.2
0.1
97%
0.8
Obs: 5,593 km
Exp: 9,392 km
P < 0.01
2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000
Pairwise Geographic Distance (km)
Distance Between OTU Frequency
1
90%
0.8
Obs: 6,092 km
Exp: 9,438 km
P < 0.01
0.3
0.2
0.1
C
Pairwise Geographic Distance (km)
1
97%
0.8
Obs:12.62°
Exp:17.38°
P < 0.01
0.05
5
10
15
20
25
30
35
Pairwise Distance (° from Equator)
40
0.6
0.4
0.2
45
1
0.35
0.25
0.4
2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000
0.15
D
0.6
0.2
0.35
0.25
0.4
0.2
0.5
0.4
0.6
Distance Between OTU Cumulative Frequency
B
1
90%
0.8
Obs: 13.10°
Exp: 16.06°
P < 0.01
0.15
0.05
5
10
15
20
25
30
35
Pairwise Distance (° from Equator)
Expected
Observed
Expected Mean
Observed Mean
40
0.6
0.4
0.2
45
Expected
Observed
Fig. 4. Mean pairwise distance between samples sharing nonsingleton OTUs
is significantly less than would be expected under a null model of random
distribution when OTUs are defined at 97% sequence identity (A; n = 1,932;)
and at 90% sequence identity (B; n = 1,258). OTUs shared among buildings
located within 2,000 km appear to drive this geographic structure, indicating
relatively small ranges of some OTUs. Pairwise comparisons of samples from
opposite sides of the equator (within-hemisphere pairwise comparisons excluded) showed that OTUs were more likely to be shared among samples
taken from similar distances from the equator than would be expected under
a null model of random distribution with OTUs defined at 97% sequence
identity (C) and 90% sequence identity (D). Histograms represent the frequency of mean pairwise distance between locations of each OTU in distance
classes. Expected means and significance value were calculated with 100 randomized permutations of the dataset constrained by sequence abundance at
each location. Vertical lines are the mean of all individual OTU calculations.
Triangles and circles are the cumulative frequency of OTUs at increasing distance. P value is the likelihood that the observed mean is greater than the
expected.
versity in temperate zones, including diminished spatial heterogeneity, decrease in competition and predation, a less stable climate, and lower productivity. How these factors may apply to
microbial communities is unclear, but plausible contradictory hypotheses would not be difficult to devise. Because our samples
were time-integrated, we may have measured the diversity of
seasonally admixed fungal composition that may not all be metabolically active at the same time. Temporal heterogeneity associated with more pronounced seasonality in temperate zones may
be one difference between microbes with short reproductive cycles
and their longer-lived, larger-bodied counterparts.
Although correlations between human socio-economics and
latitude are well documented (37), our results indicate that factors
such as building materials, content, or use have no significant effect
on fungal composition. Fungal composition in vastly different types
13752 | www.pnas.org/cgi/doi/10.1073/pnas.1000454107
of buildings was most similar to other structures in close proximity.
Despite humankind’s best efforts to homogenize indoor climates,
local environmental selection outside the buildings themselves
appears to be the strongest determinant of indoor fungal composition. Therefore, we suggest that indoor habitats should not be
thought of as microcosms isolated by weatherstripping and HVAC
filters, but rather as compositional subsets of a larger biome.
Materials and Methods
Dust Collection. Protocols were adapted and modified from the US Environmental Protection Agency (www.epa.gov/wtc/panel/pdfs/6-att-4aindoor_dust_sampling_protocols_061705.pdf). To maximize time integration of settled dust, samples were collected from (i) accessible, (ii) infrequently accessed, and (iii) inaccessible areas in replication numbers scaled
to unit size. Dustream collectors (Indoor Biotechnologies) were sterilized
before use by soaking in 10% bleach and dried under UV radiation. These
samplers were attached to domestic vacuum cleaners for collection. Samples
were filtered through a 2-mm sieve and refrigerated at 4 °C until further
processing. Samples collected from Regina, Saskatchewan, Canada, were
collected by Health Canada in conjunction with a cohort study of asthmatic
and control households, following similar collection methods. The sample
from Stittsville, ON, Canada, was extracted from the storage container of
a domestic central vacuum system. All samples were collected between
December 2008 and March 2009.
DNA Extraction. Duplicate, 100-mg sample aliquots were processed using
previously published methods (38) with the following modifications: tubes
were beat on a Biospec Mini 8 bead mill (Biospec) set to homogenize for 1 min,
and DNA was eluted with 50 μL of 1× TE (10 mM Tris pH 8, 1 mM EDTA). DNA
concentration was quantified on a NanoDrop spectrophotometer (ThermoScientific).
PCR Amplification. Approximately 1.25 ng of each duplicate DNA extraction
was added to a PCR mixture containing 1.2 units of HotStarTaq polymerase
(Qiagen), 1× PCR buffer supplied by the manufacturer, 200 μM of each DNTP,
0.5 μM of each primer, and H2O to a final concentration of 25 μL. ITS amplification used concatemers (Table S2) containing either the primer pair
ITS1f and ITS4 (39) or the primer pair LROR_F (excluding animals; Table S3)
and LR5-F [(40) excluding plants], with LROR_F as the forward primer (LSU).
Following an initial denaturation at 95 °C for 10 min, PCR was cycled 34
times at 95 °C for 1 min, 51 °C (ITS) or 54 °C (LSU) for 1 min, 72 °C for 1 min,
and a final extension at 72 °C for 7 min. Periodic negative controls were run
on samplers for every 24 DNA extractions and on every PCR.
PCR products were cleaned using the QIAquick PCR purification kit (Qiagen)
following the manufacturer’s instructions, quantified using a Qubit Fluorometer (Invitrogen) pooled by locus into equimolar concentrations, and sequenced using approximately five-eighths of a 454 Titanium sequencing run.
LSU and ITS amplicons were run in separate regions of a gasketed plate.
Sequence Processing and Analyses. Reads were sorted by multiplex tags using
the Ribosomal Database Project (RDP) pyrosequencing pipeline (41). ITS
sequences were clustered into OTUs using the program CD-HIT-EST (42) using
the “accurate but slow” iteration algorithm. Resulting clusters were input into
the software package MOTHUR (43) for rarefaction analyses. OTUs clustered
at 97% were compared using BlastN against all identified fungal sequences in
GenBank and determined to the last common ancestor (LCA) using the program MEGAN [(44); LCA parameters: minimum support—1, minimum score—
200, score/length ratio—1.97, top percentage—1] and cross-referenced to
class-level designation with the Dictionary of Fungi (45) as an authority in an
SQL database.
LSU sequences were aligned using MAFFT [(46); parameters: retree 1, maxiterate 0, nofft, parttree). Alignment columns containing more than 80% gaps
were trimmed using trimAL (47), and FastTree (48) was used for phylogenetic
tree construction. The command-line version of the unweighted UniFrac algorithm was used for phylogenetic diversity rarefaction, and the phylogenetic
dissimilarity matrix was used in multivariate analyses. All multivariate statistics
were calculated using the Vegan (49) and Ecodist (50) packages in the R programming environment (51). Confidence intervals for Mantel and ANOSIM
analyses were calculated with 999 resamples. Mean geographic distance between OTU locations was calculated using a modification of the program SAShA
(52). Null distributions and significance values were calculated by randomizing
OTU locations while maintaining distance classes and the number of reads per
sample constant. We ran SAShA with 100 random permutations in MATLAB
v. R2008b (MathWorks).
Amend et al.
Hynson, and two anonymous reviewers for helpful suggestions on an earlier
version of the manuscript; and Thomas A. Oliver for his assistance with OTU
distribution analyses. This research was supported by a grant to the Indoor
Mycobiota Barcode of Life project from the Alfred P. Sloan Foundation.
1. Höppe P, Martinac I (1998) Indoor climate and air quality. Review of current and
future topics in the field of ISB study group 10. Int J Biometeorol 42:1–7.
2. Garber G (2001) An overview of fungal infections. Drugs 61 (Suppl 1):1–12.
3. Aimanianda V, et al. (2009) Surface hydrophobin prevents immune recognition of
airborne fungal spores. Nature 460:1117–1121.
4. Kauserud H, et al. (2007) Asian origin and rapid global spread of the destructive dry
rot fungus Serpula lacrymans. Mol Ecol 16:3350–3360.
5. Schmidt O, Czeschlik D (2006) Wood and Tree Fungi Biology, Damage, Protection, and
Use (Springer-Verlag, Berlin).
6. Pitt J, Hocking A (2009) Fungi and Food Spoilage (Springer-Verlag, Berlin).
7. Samson R, Hoekstra E, Frisvad J, Filtenborg O, Schimmelcultures C (2004) Introduction
to Food and Airborne Fungi (CBS Fungal Diversity Centre, Utrecht, The Netherlands),
7th Ed.
8. Chang C, Gershwin ME (2005) Mold hysteria: Origin of the hoax. Clin Dev Immunol 12:
151–158.
9. Schmidt O (2007) Indoor wood-decay basidiomycetes: Damage, causal fungi,
physiology, identification and characterization, prevention and control. Mycol Prog 6:
261–279.
10. Haugland R, Vesper S (2002) US Patent 6,387,652.
11. Tringe SG, et al. (2008) The airborne metagenome in an indoor urban environment.
PLoS One 3:e1862.
12. Pitkäranta M, et al. (2008) Analysis of fungal flora in indoor dust by ribosomal DNA
sequence analysis, quantitative PCR, and culture. Appl Environ Microbiol 74:233–244.
13. Shelton BG, Kirkland KH, Flanders WD, Morris GK (2002) Profiles of airborne fungi in
buildings and outdoor environments in the United States. Appl Environ Microbiol 68:
1743–1753.
14. Sautour M, et al. (2009) Profiles and seasonal distribution of airborne fungi in indoor
and outdoor environments at a French hospital. Sci Total Environ 407:3766–3771.
15. Fierer N, et al. (2008) Short-term temporal variability in airborne bacterial and fungal
populations. Appl Environ Microbiol 74:200–207.
16. Ren P, Jankun TM, Belanger K, Bracken MB, Leaderer BP (2001) The relation between
fungal propagules in indoor air and home characteristics. Allergy 56:419–424.
17. Martiny JB, et al. (2006) Microbial biogeography: Putting microorganisms on the map.
Nat Rev Microbiol 4:102–112.
18. Portnoy JM, Barnes CS, Kennedy K (2004) Sampling for indoor fungi. J Allergy Clin
Immunol 113:189–198; quiz 199.
19. Kõljalg U, et al. (2005) UNITE: A database providing web-based methods for the
molecular identification of ectomycorrhizal fungi. New Phytol 166:1063–1068.
20. Huse SM, Huber JA, Morrison HG, Sogin ML, Welch DM (2007) Accuracy and quality of
massively parallel DNA pyrosequencing. Genome Biol 8:R143.
21. Quince C, et al. (2009) Accurate determination of microbial diversity from 454
pyrosequencing data. Nat Methods 6:639–641.
22. Lozupone C, Knight R (2005) UniFrac: A new phylogenetic method for comparing
microbial communities. Appl Environ Microbiol 71:8228–8235.
23. Clarke K (1993) Non-parametric multivariate analyses of changes in community
structure. Austral Ecol 18:117–143.
24. Pianka ER (1966) Latitudinal gradients in species diversity: A review of concepts. Am
Nat 100:33.
25. Hillebrand H (2004) On the generality of the latitudinal diversity gradient. Am Nat
163:192–211.
26. Arnold AE, Lutzoni F (2007) Diversity and host range of foliar fungal endophytes: Are
tropical leaves biodiversity hotspots? Ecology 88:541–549.
27. Jungblut AD, Lovejoy C, Vincent WF (2010) Global distribution of cyanobacterial
ecotypes in the cold biosphere. ISME J 4:191–202.
28. Fierer N, Jackson RB (2006) The diversity and biogeography of soil bacterial
communities. Proc Natl Acad Sci USA 103:626–631.
29. Jones RT, et al. (2009) A comprehensive survey of soil acidobacterial diversity using
pyrosequencing and clone library analyses. ISME J 3:442–453.
30. Lozupone CA, Knight R (2007) Global patterns in bacterial diversity. Proc Natl Acad Sci
USA 104:11436–11440.
31. Fuhrman JA, et al. (2006) Annually reoccurring bacterial communities are predictable
from ocean conditions. Proc Natl Acad Sci USA 103:13104–13109.
32. Hollister EB, et al. (2010) Shifts in microbial community structure along an ecological
gradient of hypersaline soils and sediments. ISME J 4:829–838.
33. Morlon H, et al. (2008) A general framework for the distance-decay of similarity in
ecological communities. Ecol Lett 11:904–917.
34. Reche I, Pulido-Villena E, Morales-Baquero R, Casamayor EO (2005) Does ecosystem
size determine aquatic bacterial richness? Ecology 86:1715–1722.
35. Peay KG, Kennedy PG, Davies SJ, Tan S, Bruns TD (2010) Potential link between plant
and fungal distributions in a dipterocarp rainforest: Community and phylogenetic
structure of tropical ectomycorrhizal fungi across a plant and soil ecotone. New
Phytol 185:529–542.
36. Tedersoo L, Nara K (2010) General latitudinal gradient of biodiversity is reversed in
ectomycorrhizal fungi. New Phytol 185:351–354.
37. Hall RE, Jones CI (1999) Why do some countries produce so much more output per
worker than others? Q J Econ 114:83–116.
38. DeSantis TZ, Stone CE, Murray SR, Moberg JP, Andersen GL (2005) Rapid
quantification and taxonomic classification of environmental DNA from both
prokaryotic and eukaryotic origins using a microarray. FEMS Microbiol Lett 245:
271–278.
39. White T, Bruns T, Lee S, Taylor J (1990) Amplification and direct sequencing of
fungal ribosomal RNA genes for phylogenetics. PCR: Protocols and Applications—A
Laboratory Manual, eds Innis N, Gelfand D, Sninsky J, White T (Academic Press, New
York), pp 315–322.
40. Tedersoo L, et al. (2008) Strong host preference of ectomycorrhizal fungi in
a Tasmanian wet sclerophyll forest as revealed by DNA barcoding and taxon-specific
primers. New Phytol 180:479–490.
41. Cole JR, et al. (2009) The Ribosomal Database Project: Improved alignments and new
tools for rRNA analysis. Nucleic Acids Res 37 (Database issue):D141–D145.
42. Li W, Godzik A (2006) Cd-hit: A fast program for clustering and comparing large sets
of protein or nucleotide sequences. Bioinformatics 22:1658–1659.
43. Schloss PD, Handelsman J (2005) Introducing DOTUR, a computer program for
defining operational taxonomic units and estimating species richness. Appl Environ
Microbiol 71:1501–1506.
44. Huson DH, Auch AF, Qi J, Schuster SC (2007) MEGAN analysis of metagenomic data.
Genome Res 17:377–386.
45. Kirk P, Cannon P, Minter D, Stalpers J (2008) Dictionary of the Fungi (CAB
International, Oxford, UK)10th Ed.
46. Katoh K, Asimenos G, Toh H (2009) Multiple alignment of DNA sequences with
MAFFT. Methods Mol Biol 537:39–64.
47. Capella-Gutiérrez S, Silla-Martínez JM, Gabaldón T (2009) trimAl: A tool for
automated alignment trimming in large-scale phylogenetic analyses. Bioinformatics
25:1972–1973.
48. Price MN, Dehal PS, Arkin AP (2009) FastTree: Computing large minimum evolution
trees with profiles instead of a distance matrix. Mol Biol Evol 26:1641–1650.
49. Dixon P (2003) VEGAN, a package of R functions for community ecology. J Veg Sci 14:
927–930.
50. Goslee SC, Urban DL (2007) The ecodist package for dissimilarity-based analysis of
ecological data. J Stat Softw 22:1–19.
51. Team RDC (2005) R: A Language and Environment for Statistical Computing (R
Foundation for Statistical Computing, Vienna).
52. Kelly RP, Oliver TA, Sivasundar A, Palumbi SR (2010) A method for detecting population genetic structure in diverse, high gene-flow species. J Hered 101:423–436.
Amend et al.
PNAS | August 3, 2010 | vol. 107 | no. 31 | 13753
ECOLOGY
ACKNOWLEDGMENTS. We thank Toni Atkinson, Martin Bidartondo, Bryony
Horton, Karin Jacobs, Wayne Law, Martin Meijer, Endang Rehayu, Health
Canada, and many others who assisted with dust collection; Noah Fierer, David
Hibbett, Timothy James, Jennifer Martiny, Kabir Peay, John Taylor, Nicole