Cotrimoxazole Prophylaxis Increases Resistance Gene Prevalence and α-Diversity but Decreases β-Diversity in the Gut Microbiome of Human Immunodeficiency Virus-Exposed, Uninfected Infants

Abstract

Background: Prophylactic cotrimoxazole treatment is recommended in human immunodeficiency virus (HIV)-exposed, uninfected (HEU) infants, but the effects of this treatment on developing HEU infant gut microbiotas and resistomes are largely undefined.

Methods: We analyzed whole-metagenome sequencing data from 163 longitudinally collected stool samples from 63 HEU infants randomized to receive (n = 34; CTX-T) or to not receive (n = 29; CTX-N) prophylactic cotrimoxazole treatment. We generated taxonomic, functional pathway, and resistance gene profiles for each sample and compared microbiome signatures between the CTX-T and CTX-N infants.

Results: Metagenomic analysis did not reveal significant differences in taxonomic or functional pathway α-diversity between CTX-T and CTX-N infants. In contrast, resistance gene prevalence (P = .00719) and α-diversity (P = .0045) increased in CTX-T infants. These differences increased over time for both resistance gene prevalence measured by log-normalized abundance (4-month mean, 0.71 [95% confidence interval {CI}, .2-1.2] and 6-month mean, 0.85 [95% CI, .1-1.7]) and α-diversity (P = .0045). Unlike α-diversity, interindividual gut microbiome taxonomic (mean, -0.11 [95% CI, -.15 to -.077]), functional taxonomic (mean, -0.050 [95% CI, -.084 to -.017]), and resistance gene (mean, -0.13 [95% CI, -.17 to -.099]) β-diversity decreased in CTX-T infants compared with CTX-N infants. These results are consistent with persistent antibiotic selection pressure.

Conclusions: Cotrimoxazole prophylaxis in HEU infants decreased gut microbiome β-diversity and increased antibiotic resistance gene α-diversity and prevalence. Antibiotic resistance is a growing threat, especially in low- and middle-income countries where the higher perinatal HIV exposure rates result in cotrimoxazole prophylaxis. Understanding effects from current HEU infant antibiotic prophylaxis guidelines will inform guideline revisions and efforts to reduce increasing antibiotic resistance.

Keywords: HIV-exposed; antibiotic resistance; cotrimoxazole prophylaxis; microbiome; uninfected infant.

Figure 1.

Stool from human immunodeficiency virus (HIV)–exposed, uninfected infants randomized to receive or not receive cotrimoxazole prophylaxis was collected, sequenced, and analyzed. A, HIV-positive mothers with HIV-uninfected infants were enrolled in the study. Their HIV-negative infants were randomized to either receive cotrimoxazole prophylaxis according to World Health Organization guidelines (red, CTX-T) or to not receive cotrimoxazole prophylaxis (blue, CTX-N). B, Three stool samples were collected per infant at 6 weeks (timepoint A), 4 months (timepoint B), and 6 months (timepoint C). Spread in collection time is due to variability in patient visit time. C, Infant stool samples were then shotgun whole-metagenome sequenced.

Figure 2.

Maternal CD4 cell count, physical traits, and blood testing results are not different between cotrimoxazole-treated human immunodeficiency virus–exposed, uninfected (HEU) infants (CTX-T; red) and HEU infants not treated with cotrimoxazole (CTX-N; blue). Points represent individual measurements for infants. A, Comparison (unpaired Wilcoxon rank-sum test) of maternal CD4 count between CTX-T infants and CTX-N infants. Loess regression (bold line) with 95% confidence interval (shaded area) for CTX-T infants and CTX-N infants for weight (B), height (C), and mid-upper arm circumference (MUAC) (D). Boxplots (E–H) show median values (dark middle line) and first and third quartiles (lower and upper lines). Between-group comparisons for E–H are made using unpaired Wilcoxon tests (rank-sum) with Benjamini-Hochberg correction. Comparison between CTX-T infants and CTX-N infants at 3 separate collection times for alanine aminotransferase (ALT) (E), hemoglobin (Hb) (F), platelets (G), and white blood cell (WBC) count (H).

Figure 3.

Cotrimoxazole-treated human immunodeficiency virus–exposed, uninfected (HEU) infants (CTX-T; red) and HEU infants not treated with cotrimoxazole (CTX-N; blue) significantly separate by resistance gene annotation profiles, but not by microbial taxonomic or functional pathway annotation profiles. Points are individual samples and labels are centers of gravity for CTX-T infants and CTX-N infants from canonical analysis of principal coordinates on dissimilarity matrices calculated by Bray-Curtis. Permutational multivariate analysis of variance was used to determine if group separations were significant, and P values are reported above each plot. Input data was generated by MetaPhlAn2 (microbial taxonomic profiles) for A–C, by HUMAnN2 (functional pathway annotations) for D–F, and by ShortBRED (resistance gene annotations) for G–I. Abbreviations: CAP1, Canonical Analysis of Principal coordinates; MDS1, multidimensional scaling.

Figure 4.

Total resistance gene abundance is higher in cotrimoxazole-treated human immunodeficiency virus–exposed, uninfected (HEU) infants (CTX-T) than in HEU infants not treated with cotrimoxazole (CTX-N). A, Points represent individual HEU infant samples colored by treatment group (red for CTX-T infants and blue for CTX-N infants), and lines represent predictions of simple linear regression models for the 2 groups. The x-axis for each plot is the day of life for each infant calculated from their day of birth, and the y-axis is log-transformed total resistance gene abundance. The formula for the linear mixed-effects model is reported above the plot. The model was compared to a null model without the cotrimoxazole treatment variable (CTX) included using a likelihood ratio test. The P value for this comparison is reported at bottom left. B, Points are individual HEU infant samples binned by treatment group (red for CTX-T and blue for CTX-N) and by time. The y-axis is log-transformed total resistance gene abundance. Unpaired Wilcoxon tests were used to compare the treated to untreated samples within collection times. C, Gray distributions show the difference at each collection time between the CTX-T infants and CTX-N infants log-transformed total resistance gene abundance for 5000 bootstrapped subsamples. The black points are the mean value for this distribution, and black lines are 95% confidence intervals.

Figure 5.

Resistance richness (α-diversity) significantly increases over time in cotrimoxazole-treated human immunodeficiency virus–exposed, uninfected (HEU) infants (CTX-T) for total resistance genes and for dfr/sul resistance genes. Points represent richness values for individual samples, and boxes show median values (dark middle line) and first and third quartiles (lower and upper lines). The x-axis groups for each plot are the times of collection, and the y-axis for each plot is richness. Paired Wilcoxon tests (signed rank) were used to compare the latter 2 collections (timepoints B and C) to the first collection (timepoint A), and P values are reported above the graph with black lines depicting the comparisons. Graphs show HEU infants not treated with cotrimoxazole (CTX-N; blue) on the left and CTX-T infants (red) on the right. Richness was calculated for microbial taxa (A), functional pathways (B), resistance genes (C), and trimethoprim- and sulfonamide-resistance (dfr/sul) genes (D).

Figure 6.

Resistance gene richness (α-diversity) is significantly higher in cotrimoxazole-treated human immunodeficiency virus–exposed, uninfected (HEU) infants (CTX-T) compared to HEU infants not treated with cotrimoxazole (CTX-N). Points represent individual patient samples colored by treatment group (red for CTX-T infants and blue for CTX-N infants), and lines represent predictions of simple linear regression models for the 2 groups. The x-axis for each plot is the day of life for each infant calculated from their day of birth, and the y-axis is richness. Models were made for microbial taxa (A), functional pathways (B), resistance genes (C), and trimethoprim- and sulfonamide-resistance (dfr/sul) genes (D). Formulas for each linear mixed-effects model are reported above the plots, and these models were compared using likelihood-ratio tests to null models made without the cotrimoxazole treatment variable (CTX) included. The P values for these comparisons of linear mixed-effects models are reported at top left of each graph.

Figure 7.

Microbial taxa, functional pathway, and resistance gene β-diversity are lower in cotrimoxazole-treated human immunodeficiency virus–exposed, uninfected (HEU) infants (CTX-T) than in HEU infants not treated with cotrimoxazole (CTX-N). Data are shown for β-diversity calculated from microbial taxonomic profiles, functional metabolic pathways, and resistance genes. Distributions show the difference at each collection timepoint between the CTX-T cohort’s Bray-Curtis dissimilarities and the CTX-N cohort’s Bray-Curtis dissimilarities for 5000 bootstrapped subsamples. The black points are the mean value for this distribution, and black lines are 95% confidence intervals.