Maternal HIV infection and the milk microbiome

Abstract

Background: Children born to women with HIV but who do not become HIV infected experience increased morbidity and mortality compared with children born to women without HIV. The basis of this increased vulnerability is unknown. The microbiome, specifically the infant gut microbiome, likely plays an important role in infant immune development. The human milk microbiome is thought to have an important role in the development of the infant gut and therefore, if perturbed, may contribute to this increased vulnerability. We investigated the effects of HIV and its therapies on the milk microbiome and possible changes in the milk microbiome before or after infant HIV infection.

Results: Seven-hundred fifty-six human milk samples were selected from three separate studies conducted over a 15-year period to investigate the role of HIV and its therapies on the human milk microbiome. Our data reveal that the milk microbiome is modulated by parity (R² = 0.006, p = 0.041), region/country (R² = 0.014, p = 0.007), and duration of lactation (R² = 0.027-0.038, all p < 0.001). There is no evidence, however, using 16S rRNA V4 amplicon sequencing, that the human milk microbiome is altered by HIV infection (R² = 0.003, p = 0.896), by combination antiretroviral therapy (R² = 0.0009, p = 0.909), by advanced maternal disease (R² = 0.003, p = 0.263), or in cases of infant infection either through isolated early mucosal (R² = 0.003, p = 0.197) or early mucosal and breast milk transmission (R² = 0.002, p = 0.587).

Conclusions: The milk microbiome varies by stage of lactation, by parity, and by region; however, we found no evidence that the human milk microbiome is altered by maternal HIV infection, disease severity, or antiretroviral therapy. Additionally, we found no association between the milk microbiome and transmission of HIV to the infant. Investigations including higher resolution microbiome approaches or into other potential mechanisms to understand why the approximately one million children born annually to women with HIV escape infection, but do not escape harm, are urgently needed. Video Abstract.

Keywords: Breast milk transmission; CHEU; Children; HEU; HIV; HIV-1; HIV-exposed uninfected; Human breast milk microbiome; Infants; Lactation.

Fig. 1

Study schematic of participants and samples included in the final analyses. Overall study design and distribution of samples from each cohort. Each rectangular box indicates a sub-study, with the number of participants included in the final analyses indicated in parentheses next to each grouping variable. The open circles denote the timing and number of samples included in the final analysis. The dotted vertical lines show the best estimate for timing of mother-to-child transmission of HIV in the transmitters and timing of pre- and post-samples for comparison in the non-transmitters. The total number of samples is the number of samples included in the final analyses. *ZEBS samples were analyzed to assess advanced maternal disease as well as cases of transmission. ^Parentheses in total sample number (N) column for PROMISE transmission cohort is the number of unique samples, 9 samples from 4 participants were also part of the ART analysis. WWoH women without HIV, WWH women with HIV, ART antiretroviral therapy

Fig. 2

Comparison of human milk microbiome profiles between WWoH and WWH on ART. a Principal coordinates analysis of human milk microbiome profiles using Jenson-Shannon divergences. Large points denote centroids and ellipses show 95% confidence areas for the groups as marked. Numbers in brackets denote percent of overall variation explained by each component. b Relative abundances at the genus level. c Boxplots of alpha diversity using the Shannon metric. d Coefficients from linear regression analysis of taxa. Taxa with positive estimates are increased in WWH and taxa with negative estimates are increased in WWoH. Error bars denote 95% confidence intervals

Fig. 3

Antiretroviral therapy does not modulate the human milk microbiome. a Principal coordinates analysis of human milk microbiome profiles using Jenson-Shannon divergences demonstrates that maternal ART was not a driver of overall microbiome variation (p = 0.909). Large points denote centroids, and ellipses show 95% confidence areas for the groups as marked. Numbers in brackets denote percent of overall variation explained by each component. b Relative abundances at the genus level stratified by visit. c Boxplots of alpha diversity using the Shannon metric. d Coefficients from linear regression analysis of taxa stratified by visit. Taxa with positive estimates are increased in mothers on ART, and taxa with negative estimates are decreased in mothers on ART. Error bars denote 95% confidence intervals

Fig. 4

Comparison of human milk microbiome profiles between mothers with high (> 200) and low (≤ 200) CD4 counts. a Principal coordinates analysis of human milk microbiome profiles using Jenson-Shannon divergences. Large points denote centroids, and ellipses show 95% confidence areas for the groups as marked. Numbers in brackets denote percent of overall variation explained by each component. b Relative abundances at the genus level stratified by visit. c Boxplots of alpha diversity using the Shannon metric. d Coefficients from linear regression analysis of taxa stratified by visit. Taxa with positive estimates are increased in mothers with low CD4 counts and taxa with negative estimates are decreased in mothers with low CD4 counts. Error bars denote 95% confidence intervals

Fig. 5

Comparison of human milk microbiome profiles between non-transmitters and early mucosal transmitters. a Principal coordinates analysis of human milk microbiome profiles using Jenson-Shannon divergences. Large points denote centroids and ellipses show 95% confidence areas for the groups as marked. Numbers in brackets denote percent of overall variation explained by each component. b Relative abundances at the genus level stratified by visit. c Boxplots of alpha diversity using the Shannon metric. d Coefficients from linear regression analysis of taxa stratified by visit. Taxa with positive estimates are increased in transmitters and taxa with negative estimates are increased in non-transmitters. Error bars denote 95% confidence intervals

Fig. 6

Comparison of human milk microbiome profiles between non-transmitters and transmitters. a Schematic of samples collected by postpartum age. Each row represents a single participant. Light and dark purple points denote samples collected prior to (pre) and after (post) transmission, respectively. The two numbers in each row show the age of the infant at the last negative and first positive HIV PCR test (light and dark purple digits, respectively). Dashed boxes mark the early mucosal versus human milk transmitters as indicated. Non-transmitters are not shown in this diagram. b Principal coordinates analysis of human milk microbiome profiles using Jenson-Shannon divergences. Large points denote centroids and ellipses show 95% confidence areas for the groups as marked. Numbers in brackets denote percent of overall variation explained by each component. c Relative abundances at the genus level stratified by whether the samples were collected pre- or post-transmission. d Boxplots of alpha diversity using the Shannon metric. e Coefficients from linear regression analysis of taxa stratified by visit. Taxa with positive estimates are increased in transmitters and taxa with negative estimates are increased in non-transmitters. Error bars denote 95% confidence intervals