City-scale distribution and dispersal routes of mycobiome in residences
Pathogenic and allergenic bacteria and fungi within the indoors can bring detrimental health effects on the occupants. We previously studied the bacterial communities found in households located throughout Hong Kong as well as the skin surfaces of the occupants. As a complementary study, here, we investigated the fungal communities (mycobiome) in the same residences and occupants and identified factors that are important in shaping their diversity, composition, distribution, and dispersal patterns.
We observed that common skin and environmental fungal taxa dominated air, surface, and skin samples. Individual and touch frequency strongly and respectively shaped the fungal community structure on occupant skin and residential surfaces. Cross-domain analysis revealed positive correlations between bacterial and fungal community diversity and composition, especially for skin samples. SourceTracker prediction suggested that some fungi can be transferred bidirectionally between surfaces and skin sites, but bacteria showed a stronger dispersal potential. In addition, we detected a modest but significant association between indoor airborne bacterial composition and geographic distance on a city-wide scale, a pattern not observed for fungi. However, the distance-decay effects were more pronounced at shorter local scale for both communities, and airflow might play a prominent role in driving the spatial variation of the indoor airborne mycobiome.
Our study suggests that occupants exert a weaker influence on surface fungal communities compared to bacterial communities, and local environmental factors, including air currents, appear to be stronger determinants of indoor airborne mycobiome than ventilation strategy, human occupancy, and room type.
KeywordsMycobiome Indoor built environment Dispersal potentials Distance-decay Biogeography
Individuals in urban societies spend most of their time indoors . The indoor environment (also referred to as the built environment, BE) is also the habitat for microorganisms and represents a major interface of contact between humans and microbes . In recent years, indoor microbiome has received much research interest, as organisms detected within BEs can be detrimental to human health [2, 3, 4, 5, 6]. Therefore, understanding factors that shape the indoor microbiome and its dispersal has great importance for occupants’ health and well-being.
Human skin itself is home to a diverse community of commensal microbiota, consisting of bacteria, fungi, archaea, and viruses [7, 8, 9]. Since the advent of high-throughput sequencing, bacteria, as the predominant skin colonizers, have thus far garnered tremendous attention in microbiome work [8, 10, 11, 12]. The diversity and composition of skin bacterial community are highly personalized  and site-specific . Direct contact between human skin and indoor surfaces is a major route for microbial dispersal, and residential surfaces harbor unique skin bacterial signatures [15, 16]. In addition, occupants release particles into the surrounding environment and leave the indoor air with distinctive human-associated microbial fingerprints [17, 18].
While bacterial communities on skin are better understood, cutaneous microbial assemblages are not limited to bacteria, with fungi also occupying a wide range of skin niches . Within fungal communities (mycobiome), Malassezia, Rhodotorula, Debaromyces, Cryptococcus, and Candida are among the most prevalent cutaneous taxa across different individuals and population groups [19, 20]. Direct shedding of fungal microbes from occupant clothing may exert an influence on the indoor mycobiome . Furthermore, human occupancy and behavior appear to affect fungal richness and diversity within the BEs [18, 22]. In agreement with this, common skin fungal taxa, possibly originating from human occupants, are widely detected in indoor dust [2, 23, 24, 25, 26], air [21, 27], or surfaces [22, 28] of different BEs.
In addition to the host-associated factors, indoor mycobiome also tends to be highly influenced by fungi from the outdoors [15, 22, 29, 30]. Outdoor microorganisms rarely cross significant geographic barriers through active dispersal , but the small size of microbes can facilitate their long-distance passive dispersal potentially with the help of air, water, and/or animals [31, 32]. The outdoor spatial variation in microbial diversity with geographic distance generates the distance-decay biogeographic pattern [32, 33], which is also encountered in indoor environments at different spatial scales [3, 21] as a function of outdoor dispersal .
To date, most indoor mycobiome studies have been largely limited to the western world, with only a few exploring the dispersals of mycobiome within the BEs [21, 34]. Furthermore, the reported fungal distance-decay pattern is either on a short geographic distance (400 m) within a housing complex  or in residences located across continents , with no information on a city-wide scale with different levels of urbanization and the associated factors driving the indoor biogeographic pattern. The mechanism(s) by which fungi disperse within BEs, and whether endemic fungal taxa exist in Asian households, are also unknown. In this study, the air, skin, and surface mycobiomes of 19 households distributed throughout Hong Kong (HK) were analyzed and compared with our previous bacterial work in the same households . The objective is to determine whether, like bacteria, fungi can be transferred between surface and skin and, therefore, whether household surfaces harbor the human skin “fungal fingerprint.” Also, we aim to determine whether the indoor airborne mycobiome is strongly influenced by the local outdoor environment and whether households that are nearby share more taxa than those farther apart regardless of ventilation type and building design (i.e., demonstration of distance-decay relationships in the indoor airborne mycobiome).
Sample collection, DNA extraction, and sequencing
Forward and reverse reads were merged, trimmed, and filtered to a uniform length of 300 bp with an error rate of less than 0.5 error per read in USEARCH (version 9.0.2132) . Dereplicated reads were clustered into operational taxonomic units (OTUs) at 97% identity following the UPARSE pipeline . All sequences were first assigned taxonomy with the UNITE database  using the UCLUST algorithm in QIIME (version 1.9.1) , and a curated dataset  was then adopted to provide taxonomic information for reads that were unclassified at the genus rank with UNITE. Chimeras, singletons, and contaminants were discarded, and samples were rarefied to an even depth of 1058 reads per sample prior to the downstream analyses.
Data analyses and statistics
Statistical analysis was implemented using the R framework (version 3.3.0). Differences in the relative abundance of a given genus between groups were determined using the Mann-Whitney (MW)  and Kruskal-Wallis (KW)  tests. Indicator species analysis was performed with the “indval” function in the R package “labdsv” . The significances of different building and location factors on rarefied alpha- and beta-diversity of indoor mycobiome were assessed using analysis of variance (ANOVA) of the linear mixed-effects models in the R package “lme4”  and permutational multivariate analysis of variance (PERMANOVA) by the “adonis” function in the R package “vegan” , respectively. The tested metrics include observed number of OTUs, Chao1 , Shannon , and Simpson  for alpha-diversity and Bray-Curtis dissimilarity  and Binary Jaccard distance  for beta-diversity analyses. The factors considered above contained customized variables such as household occupancy and surface properties (details are provided in Additional file 2: Table S1). Spearman test was used to investigate the cross-domain bacterial and fungal alpha- and beta-diversity correlations. Mantel test  in the R package “ade4”  was applied to calculate the correlation between the geographic distance and the indoor bacterial or fungal community dissimilarity at three spatial scales, with Binary Jaccard as the community distance metric. Dispersals between different sample types within the households were predicted using the SourceTracker algorithm . Indoor and outdoor air source tracking was carried out with HK indoor air samples as sinks and outdoor air samples from Beijing  and Berkeley  as the surrogate sources after performing closed-reference and open-reference OTU picking in QIIME. The sampling and sequencing information of the three studies used in source tracking are summarized in Additional file 3: Table S2.
A total of 1501 fungal OTUs were retained after quality control of the sequencing data. Although some taxa were ubiquitous, the majority of OTUs (1371) were found in less than 10% of the samples, and 444 OTUs appeared only in one sample even after singleton OTU removal. The OTU rank-abundance curve shows a long tail of many low-abundance taxa (Additional file 4: Figure S1), which is also observed in another indoor mycobiome study . Following rarefaction, skin and surface samples shared a larger proportion of their OTUs (145) than air and surface (54) or air and skin (38) (Additional file 5: Figure S2).
Taxonomy and indicator species analysis
Indicator species analysis identified some taxa that were noticeably more abundant in certain types of samples. For example, three OTUs of the mold genera Cladosporium and Paecilomyces served as unique indicator taxa of air samples. A small number of OTUs deriving from the common skin fungal members of Malassezia, Penicillium, and Pichia as well as an environmental species Aspergillus niger were the indicator taxa for skin samples. However, no indicator species was found for surface samples (Additional file 7: Table S3).
Factors influencing fungal community diversity and composition
ANOVA was adopted to infer the influence of different factors on community alpha-diversity (Additional file 8: Table S4). After rarefaction, community richness (observed number of OTUs and Chao1) and diversity (Shannon and Simpson) metrics differed significantly by sample types, with skin showing the highest alpha-diversity among the three types of samples (p < 0.001 for all metrics). Specifically, air samples were significantly affected by household and relative humidity for all metrics except for Simpson when considering humidity as a factor, while room only showed influence on community richness. Factors (area, household, and individual) considered for skin samples all displayed significant effects regardless of the metric used. Area and room differed significantly for all metrics in surface samples except for Simpson when area was considered, and human occupancy did not affect the richness of surface community.
Bray-Curtis-based PERMANOVA test on factors affecting the indoor mycobiome
p = 0.003, t = 1.781
p = 0.001, t = 2.227
p = 0.054, t = 1.654
p = 0.001, t = 1.727
p = 0.001, t = 3.845
p = 0.001, t = 1.552
p = 0.001, t = 4.298
p = 0.287, t = 1.104
p = 0.081, t = 1.453
p = 0.396, t = 1.055
p = 0.001, t = 2.489
p = 0.001, t = 5.374
p = 0.001, t = 12.790
p = 0.138, t = 1.448
Cross-domain comparison of bacterial and fungal communities
To explore the correlation between bacterial and fungal communities within each sample type, we combined the current fungal data with our previous work that analyzed indoor bacterial communities  of the same households and occupants. After rarefying to the respective sequencing depths, skin bacterial and fungal alpha-diversity correlated significantly and positively for both observed number of OTUs and Chao1 (Spearman, p < 0.001, rho = 0.290 for observed number of OTUs; p < 0.001, rho = 0.289 for Chao1, Additional file 10: Table S6), indicating that samples with a higher bacterial richness also tended to have a higher fungal richness. For surface samples, only the estimated total number of bacterial and fungal OTUs was significantly and positively correlated (Spearman, p < 0.001, rho = 0.301 for Chao1). However, no significant cross-domain alpha-diversity correlation was observed for air samples (Spearman, p = 0.252 for observed number of OTUs, p = 0.849 for Chao1).
The cross-domain beta-diversity correlation was performed based on the rarefied Bray-Curtis dissimilarities and Binary Jaccard distances, respectively (Additional file 11: Table S7). For both metrics, significant and positive bacterial-fungal community composition correlations were identified for all three sample types. Remarkably, skin samples (Spearman, p < 0.001, rho = 0.315, effect-size r = 0.291 for Bray-Curtis dissimilarity) showed the greatest positive correlations between bacterial and fungal community composition dissimilarities, and the effect size even doubled (Spearman, p < 0.001, rho = 0.237, effect-size r = 0.601 for Binary Jaccard distance) when only considering OTU presence/absence. However, regardless of the metric used, the effect size remained low for air and surface samples where the bacterial and fungal communities were moderately correlated.
Our previous bacterial study  and current fungal work jointly revealed that household was the strongest factor that significantly shaped the composition and structure of the indoor airborne communities. Given that the households are distributed spatially across the city, here, we studied the correlation between the community dissimilarity (Binary Jaccard distance metric) and geographic distance for the two domains. On an overall city-wide scale without considering directionality (the longest point-to-point straight-line distance between households is 30 km), a weak but significant correlation was observed for the bacterial community (Additional file 12: Figure S4A), suggesting that households within close proximity of each other tended to share more bacterial taxa than households located further apart (p = 0.001, r = 0.033, Mantel test). However, this pattern was not observed for fungi (Additional file 12: Figure S4B), such that households closer together in geographical distance did not necessarily have a more similar indoor air mycobiome (p = 0.400, r = 0.002, Mantel test).
Dispersal pattern predictions
Comparison of HK indoor air mycobiome to outdoor air mycobiome of selected locations
To test the hypothesis that outdoor airborne fungal community was the major source for the indoor air mycobiome, outdoor air samples from Beijing  and Berkeley  were used as surrogate sources (Additional file 3: Table S2). Sequences in the three datasets were clustered into OTUs using two different approaches (closed-reference and open-reference workflows). SourceTracker was used to determine the proportion of outdoor mycobiome that made up the HK residential mycobiome. Although biases could be introduced as OTUs were more likely to cluster within its own dataset, the contribution of the surrogate sources on HK indoor air mycobiome remained high regardless of the OTU picking strategy (especially Beijing outdoor air samples with an average of ~ 50% contribution) (Additional file 16: Figure S7), supporting the hypothesis that the fungal communities in the outdoor air dominated the indoor airborne mycobiome.
Dispersal between occupant skin and residential surfaces
Human skin is widely colonized with mycobionts including Candida, Cryptococcus, Debaryomyces, Malassezia, Penicillium, and Rhodotorula . The interaction between skin and residential surfaces leads to close resemblance between the two respective microbial communities, and the surfaces can also be imprinted with human-associated microbiome signatures [15, 16, 57, 58]. In this study, skin fungal commensals dominated the skin samples, with many being unique indicator species of skin (Additional file 7: Table S3). Skin and surface samples exclusively shared a larger number of Malassezia-associated OTUs than those that were specific to surface (Additional file 5: Figure S2). In SourceTracker analysis (Fig. 4), the surface communities can be correctly matched to the corresponding occupant skin in over 40% of the households. Malassezia, comprising over one third of contributor OTUs (39/94), was not only a major source for surface communities in these correctly matched households, but also contributed to shaping the surface communities in households where predictions were incorrect (Fig. 5a), suggesting that the dispersal of skin fungi occurred in majority of the households. On the other hand, the surfaces in household KT appeared to be the vector rather than the recipient with Candida- and Cladosporium-associated OTUs on skin being identified as sourced exclusively from the surface community (Fig. 5 and Additional file 14: Figure S6). In addition, a small number of contributor OTUs identified on surfaces were the unique sources for skin communities (Additional file 15: Table S8). Therefore, fungi on occupant skin and household surfaces can be transferred bidirectionally. Given the relatively high abundance of Malassezia on frequently touched surfaces (Additional file 6: Figure S3), as well as the relatively high accuracy in predicting the dispersal from frequently touched surfaces to the occupant skin (Fig. 4), we contend that human contact is a major route for fungal dispersal between skin and household surfaces.
Origin of the skin-associated fungi in residential air
Similar to surfaces, a high abundance of skin-associated taxa in the indoor air seems to imply strong dispersals from occupant skin. However, SourceTracker was less accurate when attempting to match the indoor airborne fungal communities to skin or surface samples (Fig. 4). In addition, only a few air samples contained taxa derived from skin or surface and no dispersal was observed in over half of the households (Additional file 13: Figure S5). Therefore, there is only weak evidence of dispersal from either skin or surface to air. This observation could be partly attributed to the fact that the sampling was conducted during winter, and the thicker clothes worn by the occupants may reduce the shedding of microbial particles from skin directly into indoor air . However, given that a large proportion of skin-associated taxa, mostly Malassezia, were found in air, there are likely other sources that cause the indoor air to resemble the occupant skin community. In our previous bacterial work [15, 60], skin-associated taxa in the outdoor air have been identified as a major source of the indoor air. Here, SourceTracker results suggested that outdoor air similarly has a high potential to be the source of the indoor airborne mycobiome (Additional file 16: Figure S7). Therefore, it is possible that the majority of the skin-associated taxa in the indoor airborne mycobiome are sourced from the outdoor environment of HK (which is densely populated). Given the localization and endemism of the microbiome in the outdoor environment, having comparable local outdoor air samples from HK would be much more effective for source tracking of the indoor mycobiome , and this will be a promising avenue to pursue for future work.
Contribution of the outdoor environment to the indoor mycobiome
In addition to the cutaneous fungi, indoor mycobiome contained relatively high abundance of environmental taxa. Cladosporium, identified as an indicator species of indoor air in this study, also accounted for a proportion of skin and surface communities. Despite this, the environmental taxa in indoor air community showed minimal dispersal tendency (Additional file 14: Figure S6), suggesting that indoor air is unlikely to be the major source of environmental taxa found in skin and surface communities. Furthermore, regionally endemic species such as Auricularia polytricha and Cordyceps militaris were only observed in surface and skin samples. Lentinula edodes, which was scarce in the air community according to our findings, was significantly enriched (p = 0.03, MW test) on surfaces with high deposition potentials. One reason could be that the relatively large spore size of these mushroom-forming fungi facilitated their deposition . In addition, some fungal spores can adhere to abiotic surfaces without external influence , and the adhesion effect increases with the increasing surface roughness . Therefore, it is possible that human skin and residential surfaces are major collectors of large spores that originated from the outdoors.
The distance-decay biogeographic pattern
Architectural design is important for controlling the influence of outdoor air on the diversity and composition of the BE microbiome, with different ventilation strategies and characteristics shown to drive the variation in the indoor airborne community [65, 66, 67]. Here, the ventilation type (natural/mechanical) had no significant effect on either richness or composition of the indoor airborne mycobiome. This contrasts a recent study by Irga and Torpy  that showed naturally ventilated offices have a higher fungal diversity compared to those with mechanical systems. Households in the current study are chosen from a wide range of buildings spatially distributed on a city-wide scale, where differences in the nearby outdoor air have been reported to drive the variation in the indoor mycobiome of the BEs . Therefore, the local environment could exert stronger influences on the indoor air mycobiome than building design when a broader scale is considered, as demonstrated by previous continental-  and global-scale  studies, where local environmental selection is the strongest determinant of household dust-associated mycobiome.
In this study, the geographic location of the households is the sole factor affecting the indoor mycobiome structure and composition. However, the distance-decay pattern of indoor airborne mycobiome was only observed in households located on the North-South transect (Fig. 3c). In contrast, the indoor bacterial community showed a weak but significant city-wide distance-decay pattern (Additional file 12: Figure S4A). These observations raise three questions: (1) What is driving the distance-decay pattern of the airborne bacterial and fungal communities? (2) Why do fungi and bacteria present different dispersal limitations? (3) If there is a correlation between fungal community and geographic distance in the North-South transect, why this pattern is not being observed in the East-West transect or even on a larger city-wide scale?
Fungi can disperse through sporulation , and a large number of propagules increase the chance for long-distance passive dispersal . In addition, fungal spores are usually aerosolized as dry spores, while bacteria are generally trapped in liquid droplets before releasing into the air . The relatively cooler and drier conditions in HK during the winter season  when the study was conducted may favor the dispersal of fungal spores relative to bacteria. Furthermore, given that particle detachment is proportional to the exposed particle surface area, and the removal force increases more rapidly than adhesion forces as particle size increases , the generally smaller size of bacteria compared to fungal propagules  is consistent with the theory that bacteria will detach under a higher airflow speed compared to fungi. Therefore, the lack of a city-wide scale distance-decay pattern in the indoor airborne mycobiome may be attributed to the physical capability of fungi in dispersing farther than bacteria during winter. Given that air is an important vector for the passive dispersal of small organisms , and variation in the landscape might cause spatial heterogeneity in the outdoor airborne microbial community , we suggest that air currents can promote microbial dispersal in the outdoor environment and further drive the biogeographic distribution of microbes in the adjacent BEs. The air mass passing through HK in winter is mostly sourced from the northwest (based on HYSPLIT air trajectory model) , and it likely plays a prominent role in driving the distance-decay pattern in the North-South transect, which also spans a range of geographic landscapes. On the other hand, the more homogenized mycobiome in the East-West transect might be due to the more uniform landscape in the air flow path (e.g., vegetation conditions, land-use type, and population density) (Fig. 1).
Summary of the dispersal potentials for bacteria and fungi in households
In summary, our study provides evidence that occupants can transfer fungi to household surfaces via contact. Although almost half of the residential surfaces harbor occupants’ skin fungal fingerprint, fungi seem to be less readily transferred than their bacterial counterparts. Our study also shows that fungi appear to disperse farther in distance compared to bacteria during winter, and bacteria show dispersal limitation at a local scale. In addition, air currents, rather than building factors, appear to drive the spatial variation in the indoor airborne mycobiome in winter. This study provides important insights into the indoor transmission route of potential fungal allergens and pathogens, and the foundation to further study the complex relationships between indoor fungal exposure, occupant health, and the influence of outdoor environment.
We thank Kelvin Chan, Kevin Distor, and Karen Yip of SeqMatic for their technical assistance on sequencing. In addition, we express special thanks to Ellen Li, Fred Kong, Wai Shan Chow, Catherine Chung, Ka Yan Ng, Yuet Ying Wong, and Flora Yeh for their assistance with sampling. We are grateful for the participation of the cohort in this study.
This research was supported by the Research Grants Council of Hong Kong through project 11276116. The funder had no role in the design, data collection, interpretation, and submission of the work presented here.
Availability of data and materials
Sequencing reads generated for this study have been deposited at the NCBI Sequence Read Archive (SRA) under the accession numbers SRP059432 (for skin samples) and SRP099248 (for residential air and surface samples). Custom scripts and workflows used in this study are available on FigShare (https://figshare.com/s/c7e50fa1537abaf43762).
XT carried out the bioinformatics and sequence analysis, and manuscript writing. MHYL contributed to the conception and design of the study, sample collection and preparation, and manuscript writing. DW helped in bioinformatics and data analysis. PKHL conceived and designed the study, and contributed to data analysis and manuscript writing. All authors have read and approved the final manuscript.
Ethics approval and consent to participate
Ethics approval for subject sampling and publication of data originating from subjects included in this study was granted by the City University of Hong Kong Ethics Committee (reference number 3-2-201312 (H000334)). All subjects provided written consent to release personal and related data for publication as needed.
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 41.Roberts D. Labdsv: ordination and multivariate analysis for ecology. R package version 1.7-0. (2015).Google Scholar
- 42.Bates D, Mächler M, Bolker B, Walker S. Fitting linear mixed-effects models using lme4. J Stat Softw. 2014;67:1–48.Google Scholar
- 43.Oksanen J, Blanchet F, Friendly M, Kindt R, Legendre P, Mcglinn D et al. vegan: community ecology package. R package version 2.4-2. 2017.Google Scholar
- 45.Shannon CE, Weaver W. The mathematical theory of communication. Urbana: University of Illinois press; 1964.Google Scholar
- 48.Jaccard P. Contribution au problème de l; immigration post-glacière de la flore alpine: etude comparative de la flore alpine du massif du Wildhorn, du haut bassin du Trient et de la haute vallée de Bagnes. 1900.Google Scholar
- 50.Chessel D, Dufour A-B, Dray S. Analysis of ecological data: exploratory and methods in environmental sciences. R package version 1.4-17. (2011).Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.