Antimicrobial Chemicals Associate with Microbial Function and Antibiotic Resistance Indoors

Abstract

Humans purposefully and inadvertently introduce antimicrobial chemicals into buildings, resulting in widespread compounds, including triclosan, triclocarban, and parabens, in indoor dust. Meanwhile, drug-resistant infections continue to increase, raising concerns that buildings function as reservoirs of, or even select for, resistant microorganisms. Support for these hypotheses is limited largely since data describing relationships between antimicrobials and indoor microbial communities are scant. We combined liquid chromatography-isotope dilution tandem mass spectrometry with metagenomic shotgun sequencing of dust collected from athletic facilities to characterize relationships between indoor antimicrobial chemicals and microbial communities. Elevated levels of triclosan and triclocarban, but not parabens, were associated with distinct indoor microbiomes. Dust of high triclosan content contained increased Gram-positive species with diverse drug resistance capabilities, whose pangenomes were enriched for genes encoding osmotic stress responses, efflux pump regulation, lipid metabolism, and material transport across cell membranes; such triclosan-associated functional shifts have been documented in laboratory cultures but not yet from buildings. Antibiotic-resistant bacterial isolates were cultured from all but one facility, and resistance often increased in buildings with very high triclosan levels, suggesting links between human encounters with viable drug-resistant bacteria and local biocide conditions. This characterization uncovers complex relationships between antimicrobials and indoor microbiomes: some chemicals elicit effects, whereas others may not, and no single functional or resistance factor explained chemical-microbe associations. These results suggest that anthropogenic chemicals impact microbial systems in or around buildings and their occupants, highlighting an emergent need to identify the most important indoor, outdoor, and host-associated sources of antimicrobial chemical-resistome interactions. IMPORTANCE The ubiquitous use of antimicrobial chemicals may have undesired consequences, particularly on microbes in buildings. This study shows that the taxonomy and function of microbes in indoor dust are strongly associated with antimicrobial chemicals-more so than any other feature of the buildings. Moreover, we identified links between antimicrobial chemical concentrations in dust and culturable bacteria that are cross-resistant to three clinically relevant antibiotics. These findings suggest that humans may be influencing the microbial species and genes that are found indoors through the addition and removal of particular antimicrobial chemicals.

Keywords: antibiotic resistance; microbiome; triclosan.

FIG 1

Relationships between antimicrobial chemicals, features of the built environment, and microbial communities. (a) Distributions of triclosan (TCS), triclocarban (TCC), benzylparaben (BePB), propylparaben (PrPB), methylparaben (MePB), butylparaben (BuPB), and ethylparaben (EtPB) concentrations in dust (ng g⁻¹) across all sampled athletic facilities (n = 116 rooms). (b) Linear correlations between antimicrobial chemicals and building features (results of ANOVA) (Table S1); well-powered building features were retained for analysis using the entropy filter described in Materials and Methods. (c) Chemical profile distance-decay relationship for microbial Bray-Curtis dissimilarities (β_Bray) and chemical Gower dissimilarities (β_Gower) between sample pairs. The red line indicates fit from a linear model to raw data. (d) Principal-coordinate analysis (PCoA) visualization of pairwise Bray-Curtis dissimilarities, calculated using Hellinger-transformed species’ relative abundances. Points represent microbial communities from individual rooms, colored and sized by the corresponding triclosan concentration (ng g⁻¹ dust). Contour lines show a surface fitted to triclosan values associated with PCoA point coordinates, using generalized additive models as implemented in the R package vegan.

FIG 2

Relationships between biocide concentrations and microbial species. (a) Spearman correlations between microbial species and triclosan (TCS) or triclocarban (TCC) concentrations (ng g⁻¹ dust), with significance as determined by HAllA (see Materials and Methods). The margin shows species’ occurrence frequencies for subjects in the Expanded Human Microbiome Project (35). (b) Number of resistance modules annotated in the pangenomes of species in the rows of panel a. Modules are members of the “Drug resistance” and “Drug efflux transporter/pump” KEGG (38) categories. The size of each bubble is scaled proportionally to the fraction of rooms in which both species-specific marker genes (i.e., the results of MetaPhlAn2 [50]) and the drug resistance gene were detected.

FIG 3

Enrichment of microbial functions with elevated triclosan. (a) Overrepresented functional capabilities among triclosan-related species (results of GSEA [42, 43]). Significantly overrepresented modules are grouped based on KEGG (38) functional categories. (b) Positive relationships between log₂ 1 + x-transformed gene copies per million (CPM) and triclosan levels (ng g⁻¹ dust) for the most enriched module, the mtrAB transcriptional regulators. Bars represent individual rooms, stratified based on the proportions of species-specific annotations. Marker colors for triclosan levels are the same as in Fig. 1d. (c and d) Bar plots as in panel b for the two functions with the highest positive Spearman rank correlation coefficients with triclosan (ρ = 0.29 and 0.23, respectively).

FIG 4

Building groupings based on culture density, diversity, and drug resistance phenotypes. (a) Mean triclosan concentrations (error bars ± 2 SEM) per building level, with colors as in Fig. 1d. Median and quartiles are shown for a large significant cluster of similar buildings (identified in panels b and c). (b) Colony-forming unit (CFU) densities g⁻¹ dust and the fractions of CFUs resistant to clarithromycin, ampicillin, and tetracycline. (c) Cluster dendrogram showing Gower dissimilarities between buildings based on features of their culturable communities. Blocks mark clusters with significant support (P < 0.01), based on 10⁴ multiscale bootstrap resamples (45) of normalized feature values. The only cluster with significant support that consisted of more than three buildings is marked by a thick blue block and a colored star.