Pathogens and Antimicrobial Resistance Genes in Household Environments: A Study of Soil Floors and Cow Dung in Rural Bangladesh

Abstract

In low- and middle-income countries, living in homes with soil floors and animal cohabitation may expose children to fecal organisms, increasing risk of enteric and antimicrobial-resistant infections. Our objective was to understand whether cow cohabitation in homes with soil floors in rural Bangladesh contributed to the presence and diversity of potential pathogens and antimicrobial resistance genes (ARGs) in the home. In 10 randomly selected households in rural Sirajganj District, we sampled floor soil and cow dung, which is commonly used as sealant in soil floors. We extracted DNA and performed shotgun metagenomic sequencing to explore potential pathogens and ARGs in each sample type. We detected 6 potential pathogens in soil only, 49 pathogens in cow dung only, and 167 pathogens in both soil and cow dung. Pathogen species with relative abundances >5% in both soil floors and cow dung from the same households included E. coli (N=8 households), Salmonella enterica (N=6), Klebsiella pneumoniae (N=2), and Pseudomonas aeruginosa (N=1). Cow dung exhibited modestly higher pathogen genus richness compared to soil floors (Wilcoxon signed-rank test p=0.002). Using Bray-Curtis dissimilarity, pathogen species community composition differed between floors and cow dung (PERMANOVA p<0.001). All soil floors and cow dung samples contained ARGs against antibiotic classes including sulfonamides, rifamycin, aminoglycosides, lincosamides, and tetracycline. Paired floor and cow dung samples shared ARGs against rifamycin. Our findings support the development of interventions to reduce soil and animal feces exposure in rural, low-income settings.

Figure A1.

Alpha-diversity indices for potential pathogen genera by sample type. Includes 10 paired cow dung and soil floor samples. Indices were compared between sample types using the Wilcoxon signed-rank test.

Figure A2.

Bray Curtis dissimilarity between communities of potential pathogen species by household membership, sample type, and presence of animal feces on the household floor. Includes 10 household-paired cow dung and soil floor samples. Bray-Curtis dissimilarity was compared between sample types using PERMANOVA.

Figure A3.

Number of antibiotic resistance genes (ARGs) in each human health risk quartile in each sample. Based on classifications in Zhang et al., 2022 (https://doi.org/10.1038/s41467-022-29283-8).

Figure 1.

Sample-level relative abundance of non-host reads for potential pathogens by mNGS analysis at the A) genus-level and B) species-level. Includes the top 30 species by average relative abundance across all samples, with all other genera or species labeled as “Other.” C1-C10 refer to cow dung samples, and S1-S10 refer to floor soil samples; each number corresponds to a different household (e.g., C1 and S1 are from household 1).

Figure 2.

Heatmap of filtered, non-host reads for potential pathogens detected in samples of cow dung (listed as C1-C10) and floor soil (listed as S1-S10). Tile colors indicate the relative abundance of each species within each sample. Gray tiles indicate that a species was not detected. Includes taxa with an average relative abundance across all samples of at least 0.5%. The heatmap displays hierarchical clustering of rows using Euclidean distance and Ward’s minimum variance method.

Figure 3.

Alpha-diversity indices for potential pathogen species by sample type. Includes 10 household-paired cow dung and soil floor samples. Indices were compared between sample types using the Wilcoxon signed-rank test.

Figure 4.

Heatmap of antibiotic resistance genes detected in cow dung and soil samples. Tile colors indicate the read coverage breadth. Includes genes with read coverage breadth > 10% or contig coverage breadth > 10% and > 5 reads mapped. Right annotation indicates the drug class that the ARG confers resistance to. Colors indicate the highest alignment confidence based on contig match quality (blue) or reads (green). “Perfect” contig matches identically matched reference sequences in the Comprehensive Antibiotic Resistance Database. “Strict” contig matches were those that matched previously unknown variants of known ARGs, including secondary screening for key mutations. “Nudged” contig matches had at least 95% identity to known AMR genes and were matched using a percent identity threshold not taking alignment length into account.

Figure 5.

Relative abundance of antibiotic resistance genes detected in cow dung and soil samples by A) drug class and B) antibiotic resistance mechanism. Both panels include genes with read coverage breadth > 10% or contig coverage breadth > 10% and > 5 reads mapped.