Household triclosan and triclocarban effects on the infant and maternal microbiome

Abstract

In 2016, the US Food and Drug Administration banned the use of specific microbicides in some household and personal wash products due to concerns that these chemicals might induce antibiotic resistance or disrupt human microbial communities. Triclosan and triclocarban (referred to as TCs) are the most common antimicrobials in household and personal care products, but the extent to which TC exposure perturbs microbial communities in humans, particularly during infant development, was unknown. We conducted a randomized intervention of TC-containing household and personal care products during the first year following birth to characterize whether TC exposure from wash products perturbs microbial communities in mothers and their infants. Longitudinal survey of the gut microbiota using 16S ribosomal RNA amplicon sequencing showed that TC exposure from wash products did not induce global reconstruction or loss of microbial diversity of either infant or maternal gut microbiotas. Broadly antibiotic-resistant species from the phylum Proteobacteria, however, were enriched in stool samples from mothers in TC households after the introduction of triclosan-containing toothpaste. When compared by urinary triclosan level, agnostic to treatment arm, infants with higher triclosan levels also showed an enrichment of Proteobacteria species. Despite the minimal effects of TC exposure from wash products on the gut microbial community of infants and adults, detected taxonomic differences highlight the need for consumer safety testing of antimicrobial self-care products on the human microbiome and on antibiotic resistance.

Keywords: antibiotic; microbiome; resistance; triclosan.

© 2017 The Authors. Published under the terms of the CC BY 4.0 license.

Figures

Figure 1. Urinary triclosan levels are elevated in TC mothers following 6 months of exposure

The concentrations of triclosan in picograms/microliter (pg/μl) are plotted as a log10 transformation of the absolute concentration +1, for visual clarity. The median value is represented as a black line, the interquartile range is represented by the box plot. The violin plot represents the full range of values obtained from the source data, where the width of the blue or red colored region represents the probability density of sample values at that level. The black dots represent outliers. Urinary triclosan measurements are available for 38 mothers (17 TC, 21 nTC) and 33 infants (15 TC, 18 nTC). The ranges for the violin plots are found in the source data. The P‐value for the differences in urinary triclosan levels between treatment arms in the mothers (represented by ***) is P = 5.66e‐5 (Mann–Whitney U‐test).Source data are available online for this figure.

Figure 2. Mother and infants have distinct microbiome compositions not driven by household TC exposure

1. PCoA of Bray–Curtis dissimilarity for all (n = 221) samples shows that gut communities cluster by mothers and infants.

2. PCoA separated by time and treatment.

Source data are available online for this figure.

Figure EV1. Individuals in mother and infant groups have similar relative microbiome compositions at the phylum level throughout the first year of life independent of TC exposure

TC households (n = 17) and nTC (n = 22) households have stable relative abundance of phyla at 2‐, 6‐, and 10‐month visits. Phyla present in < 2% abundance are condensed for visual clarity.

Figure EV2. Maternal samples within households are more similar than between households

1. Hierarchical clustering of Canberra distances shows that maternal samples from a given individual throughout the first year of life are more self‐similar than other household mothers at the same visit. The colors are representative of a single household, labeled by household and visit (B1 = 2 months, B2 = 6 months, B3 = 10 months), and node colors reflect TC grouping (red = nTC, blue = TC).

2. PCoA of Canberra distances shows 9.3% of the variance is explained by the first two principal components and the absence of TC treatment clustering for any visit among mothers.

Figure EV3. Infant intestinal microbiome variability is minimally driven by some known external factors by 2 months of age (n = 34)

1. A–D

   PCoA with Bray–Curtis dissimilarity of the infants at 2 months of age suggests factors known to impact the microbiome, such as (A) birth method (PERMANOVA R2 = 0.0683, P = 0.011) and (B) formula (R2 = 0.128, P = 0.041) drive minimal variation. However, (C) maternal ethnicity (R2 = 0.0274, P = 0.505) and (D) pets (R2 = 0.0310, P = 0.393) in the household are not sufficient to distinguish the infants at 2 months of age.

Figure 3. TC randomization does not decrease gut microbial diversity in infants or mothers

Shannon diversity measures are plotted as the interquartile range with median for each TC exposure class and time‐point grouping for infants and mothers. The upper and lower whisker for box plots indicate ± 1.5× interquartile range. The points plotted beyond these whiskers are considered outliers and are plotted individually. The P‐value represented for the increase in infant gut microbiota diversity over time is ***P = 4.4e‐4 (Mann–Whitney U‐test).Source data are available online for this figure.

Figure EV4. Microbial diversity does not correlate with urinary triclosan levels at 6 months for mothers or infants

Correlations were determined using a linear regression model (Shannon diversity ~ Urinary triclosan levels) on mother and infant samples separately.

Figure 4. Enrichment of Proteobacteria is observed in the mothers of TC households

1. A

   Differentially abundant taxa between nTC and TC households. Values left of the gray line indicate an enrichment in nTC household and values to the right indicate an enrichment in TC households. Analyses are separated by mothers and infants for all samples across the three time points (FDR‐adjusted P‐value < 0.01).

2. B, C

   Differentially abundant taxa are displayed for (B) infants and (C) mothers at per visit (FDR‐adjusted P‐value < 0.05).

Source data are available online for this figure.

Figure EV5. Triclosan resistance gene abundances do not distinguish TC and nTC maternal samples following 6 months of exposure

Euclidean distance was calculated between samples of whole shotgun sequencing reads that aligned to CARD (approximately 0.03% or 6,000 reads for each sample in both intervention arms), then clustered with a hierarchical agglomeration method.