Background
Mother-to-child HIV transmission (MTCT), can occur during pregnancy, delivery or breastfeeding and remains a public health concern especially in sub-Saharan Africa. Despite an overall 48% decrease of newly infected children between 2009 and 2014 due to the implementation of prophylaxis interventions and antiretroviral therapy (ART), there were still 160,000 newly infected children (< 15 years old) in 2016 [
1]. Postnatal HIV transmission through breastfeeding in the presence of ART contributes to a significant part of this residual transmission with a pooled risk estimate of 2.93% at 12 months [
2]. World Health Organization (WHO) guidelines recommend lifelong ART for mothers living with HIV [
3]; exclusive breastfeeding for the first 6 months of life; appropriate complementary foods thereafter and to continue breastfeeding for at least 12 months [
4].
Breast milk, the main and optimal nutrient source of the infant, contains numerous active immune factors [
5,
6]. These immune factors protect the infant from infections [
7], participate in the maturation of the infant’s immune system [
6], down-regulate gut inflammation and promote gut adaptation after birth when the newborn is confronted with the antigens of colonizing commensal bacteria [
8]. Breast milk soluble immune factors consist of a wide array of bioactive agents like cytokines, chemokines, growth factors and acute phase proteins [
6,
9]. These immune factors are involved in multiple immune functions including Th1 response, antimicrobial, anti-inflammatory, pro-inflammatory and immunomodulatory properties [
5,
6,
9,
10] and influence the immune response by attracting, activating or down-regulating different cell effectors [
9,
10]. During lactation, inflammatory processes of the mammary gland like mastitis [
11‐
15] and subclinical mastitis (SCM) [
16‐
19] induce considerable changes in the breast milk immunefactors. In asymptomatic lactating women, SCM is generally defined by an increased breast milk sodium/potassium ratio (Na/K > 1) [
17,
18,
20,
21]. Whereas clinical mastitis occurs in less than 10% of lactating mothers [
22‐
24], SCM (Na/K > 1) is more frequent, especially at start of breastfeeding in the first few days after delivery and again at the time of weaning, with a prevalence ranging from 9 to 45% in mature milk in HIV-uninfected mothers [
17,
18]. Studies have suggested that SCM may be associated with HIV shedding in breast milk and HIV mother-to-child transmission [
25‐
34].
Human immunodeficiency virus is known to induce immune activation in blood [
35], gut [
36] and female genital tract [
37,
38] but its impact on milk immune factors remains only partially defined. Few studies have compared breast milk components in ART-naïve HIV-1-infected versus HIV-uninfected women, and a limited number of parameters were assessed [
39‐
41]. We hypothesized that HIV infection induces modifications of the composition of immune factors in breast milk, and impairs immune response to SCM in breast milk.
In a first step, our study assessed the influence of HIV infection (without ART exposure) on the pattern of breast milk immune factors. We further sought to explore the association between HIV infection and immune response to SCM.
Materials and methods
Study population and clinical features
This study was part of the ANRS 1271 project exploring factors associated with HIV-1 transmission from mother to child through breastfeeding. The study was nested in a South-African non-randomized prospective intervention cohort, the Vertical Transmission Study (VTS), examining the effect of feeding practices on infant HIV infection and survival rates, in a community where HIV prevalence in pregnant women was 23.7% [
42]. A total of 2722 HIV-infected and uninfected pregnant women attending antenatal clinics in KwaZulu-Natal were enrolled after providing informed consent between October 2001 and April 2005 [
43]. Participants were naive to antiretroviral therapy except for single-dose nevirapine provided to all HIV-infected women and their newborns during delivery as per national guidelines at the time. The intervention included a personalized antenatal counseling session on infant feeding choices, and a postnatal home-based breastfeeding counseling intervention for women who chose to breastfeed. Formula feeding women were supported at their clinic visits by study nurses. Daily infant feeding practices and maternal breast health problems or breastfeeding difficulties were recorded. Maternal health and infant growth and morbidity were regularly recorded in monthly scheduled clinic visits until 9 months and every 3 months afterwards until 2 years of age. Mother’s blood samples were taken before delivery and at 6 months post-delivery to measure HIV viral load and CD4 count and child’s HIV status was checked at each study clinic visit on dried blood spots. Breast milk samples were taken separately from right and left breast at each clinic visit if the mother was still breastfeeding [
44]. The VTS study and breast milk analyses were approved by the Biomedical Research Ethics Committee of the University of KwaZulu-Natal and women gave informed consent.
Forty HIV-infected women who did not transmit HIV to their infant during follow-up and 43 HIV-uninfected women were enrolled for this study. Subjects were selected based on availability of mature breast milk samples (breast milk collected > 15 days after delivery) and in a random draw from the VTS cohort. Breast milk samples from HIV-infected women who transmitted HIV to their child were excluded in our study to focus on the impact of HIV infection on breast milk immune factors and on response to SCM. A total of 156 breast milk samples were tested. Paired left and right breast samples were available for 73 (88.0%) women.
SCM was defined by a Na/K ratio in breast milk > 1, as described elsewhere [
17,
18,
45,
46]. Maternal breast health problems (such as mastitis, engorgement, cracked/bleeding nipple, blocked duct, breast thrush, breast/nipple oozing pus and abscess [
24]) were considered if they occurred during the week before the sample collection date because mastitis markers are normalized within 1 week after symptom resolution [
13]. Breastfeeding practices were defined as in the VTS cohort [
43], but were only included on the 15 days before the sample collection date, in order to detect a recent breastfeeding practice change that may be indicative of a breast health problem [
47].
We explored four groups of breast milk samples: samples without SCM from HIV-uninfected women, samples with SCM from HIV-uninfected women, samples without SCM from HIV-infected women and samples with SCM from HIV-infected women.
Immune factors and biochemical assays in breast milk
Whole breast milk samples were stored at − 80 °C until processing at the Montpellier University Teaching Hospital, France. All parameters were measured in lactoserum after centrifugation at 1200g for 15 min. Lactoserum sodium and potassium concentrations were measured with ion selective electrode (AU640 analyzer, Beckman Coulter, Fullerton, CA). β2microglobuline (B2M)and C-reactive protein (CRP) were determined by immunoturbidimetric methods (AU640, Beckman Coulter, Fullerton, CA). Lactoserum erythropoietin (EPO) (IMMULITE2000 EPO assay, Diagnostic Products Corporation, Los Angeles, CA), α-defensin (Hycult Biotech, Uden, The Netherlands), lactoferrin (Calbiochem, Dramstadt, Germany), secretory leukocyte peptidase inhibitor (SLPI) (R&D Systems, Minneapolis, MN), lipopolysaccharide-binding protein (LBP) (Hycult Biotech), soluble CD14 (sCD14) (Hycult Biotech) and S100A9 protein (PS100A9) (CycLex, Nagano, Japan) were quantified by commercial enzyme immunoassays as recommended by the manufacturer. Interleukins (IL) 1β, 2, 4, 5, 6, 7, 8 (IL-8 or CXCL8), 10, 13, 15 and 17, interleukin 12p40/70, receptor antagonist of interleukin 1β (IL-1RA), interleukin 2 receptor (IL-2R), granulocyte and macrophage growth factor (GM-CSF), tumor necrosis factor- α (TNF-α), γ interferon (IFN-γ), α interferon (IFN-α), macrophage inflammatory protein 1α and 1β (MIP-1α and MIP-1β), inflammatory protein 10 (IP-10 or CXCL10), monokine induced by gamma interferon (MIG or CXCL9), eotaxin, regulated upon activation normal T-cell expressed and secreted (RANTES or CCL5), and monocyte chemotactic protein-1 (MCP-1 or CCL2) were quantified using a multiplex microbeads assay (Invitrogen Human Cytokine 25-Plex Panel, Marne-La-Vallée, France) and a Luminex 100 apparatus (Luminex, Oosterhout, The Netherlands) following the manufacturer’s instructions.
Concentrations below the lower limit of quantification were assigned half the value of the lower limit, as described elsewhere [
39,
48]. Conversely, concentrations above the upper limit of quantification were assigned the value of the upper limit [
17]. Immunologic factors for which 50% or more of the samples in a group were below the lower limit of quantification of the test weren’t quantitatively analyzed.
Statistical analyses
In the first step of the study, breast milk samples were tested for SCM based on Na/K ratio. Then, the association between ART-naïve HIV infection and breast milk environment in the absence of SCM were analyzed by comparing concentrations of immune markers in samples from ART-naïve HIV-infected versus HIV-uninfected women. Next, changes induced by SCM were explored by comparing levels of immune factors in SCM samples to breast milk samples without SCM in ART-naïve HIV-1 infected and uninfected women.
Demographic and clinical characteristics of women, immunologic factors detection rates and concentrations were compared between HIV positive and negative women using the Chi-square or Fisher exact test for qualitative variables and the Student or Wilcoxon rank-sum test for quantitative variables according to the variable’s distribution. The associations between HIV infection and SCM on immune factor’s concentration were tested separately by Wilcoxon rank-sum test. Spearman’s non parametric test was used to assess correlation between immunologic factors by group. When immunologic factors were compared in sets of bivariate analyses, p-values were corrected for false discovery rate (FDR;
p < 0.05) to correct for multiple testing [
49].
To control for potential confounder effects, multivariate mixed linear models were performed for each immunologic factor. Models were done with the SAS PROC MIXED procedure, adjusted on child age at the time of sampling [
48,
50‐
52] and taking into account the clustering of left and right breast milk samples of a woman by introducing a random effect. All statistical analyses were conducted using SAS statistical software version 9.2 (SAS Institute, Cary, NC).
Discussion
We provided the first comparative analysis of a large number of soluble immune factors in the breast milk of ART-naive HIV-infected women and uninfected women and highlighted the association between HIV infection and immune response to SCM. Our study was performed on samples collected years ago, before the era of universal, life-long maternal ART, offering a unique opportunity to better understand the interactions between HIV infection and inflammation in the mammary gland without interference of therapy. Although pregnant and lactating women have today increasing access to ART, HIV replication during lactation still occurs in cases of virological failure or poor adherence, and in HIV-infected women unaware of their status. Furthermore, data collected before the era of ART are necessary to further explore the immune response to SCM in HIV-infected women undetectable for HIV on ART, in whom SCM may still occur, especially during weaning.
Our results showed that HIV infection moderately alters the soluble immune factor environment in breast milk samples without SCM or any other breast health problem. Hence, in the absence of SCM, the concentrations of immune factors in breast milk appeared different between HIV-uninfected women and HIV-infected women who didn’t transmit the infection by breastfeeding, regarding MIG and IP-10 that were significantly higher in the breast milk of HIV-infected women. These two CXC-chemokines are induced by IFN-γ and belong to a Th1 response and a cascade that is critical in antiviral defense. In addition, the level of the anti-inflammatory IL-1RA cytokine was significantly lower. Furthermore, the network of soluble immune factors appeared only slightly impaired, with correlations between the Th1 cytokines themselves and with the inflammatory markers that were slightly higher in the samples of HIV-positive women.
Three other studies who did not exclude women having transmitted HIV to their infants have assessed the impact of HIV infection on breast milk immune profile regardless of the presence of a SCM. Bosire et al. who compared MIP-1α, MIP-1β, RANTES and Stromal cell-Derived Factor-1α (SDF-1α) in the breast milk of Kenyan women several times after childbirth, found that MIP-1β was significantly higher at 10 days and RANTES at 1 month among HIV-infected versus HIV-uninfected women [
39]. Shapiro et al. found significantly higher levels of total IgM, IgG, IgA and SLPI in HIV-infected women compared to HIV-uninfected women in Botswana [
40]. Henrick et al. found significantly higher levels of soluble toll-like receptor 2 (sTLR2) in HIV-infected women [
41].
SCM can be viewed as an initial stage of infection and inflammation that carries a risk of subsequent progression to a more severe mastitis [
19]. Regarding the low frequency of mastitis, the immune response in the mammary gland is most of the time able to prevent the adverse evolution of SCM into symptomatic mastitis. Hence, only a few women reported clinical mastitis (1% in HIV-infected and 0.5% in HIV-uninfected women) or other breast health problems in the VTS cohort [
24,
53]. In our study, approximately a quarter of HIV-uninfected women and a third of HIV-infected women had SCM, which is consistent with other studies conducted in both HIV-infected and uninfected women [
16,
17,
19,
54].
Previous studies have explored breast milk immune factors during SCM (Na/K ratio > 1). Increased IL-8, lactoferrin, SLPI and RANTES were observed in HIV-uninfected women with SCM [
16,
17,
54]. We recently explored breast milk environment during SCM in HIV-uninfected mothers from the VTS cohort. Our findings indicated that SCM is associated with higher levels of B2M, PS100A9, TNF-α, IL-6, IL-8, IL-17, RANTES, IL-2R, IL-12p40/70, IFN-α, IFN-γ, MIG and IP-10 [
19]. All these results highlighted a robust, prompt and predominant Th1 and pro-inflammatory response to SCM in HIV-uninfected women. Our data confirmed that, in the presence of SCM, breast milk immune environment of HIV-uninfected women was characterized by a robust immune response involving a broad panel of Th1 and inflammatory related immune factors, as well as anti-bacterial response. By comparison, the breast milk immune factor environment appeared severely impaired during SCM in HIV-infected women who didn’t transmit HIV to their child by breastfeeding, with only five immune factors that were significantly increased compared to fourteen parameters in HIV-uninfected women. Furthermore, the magnitude of immune factors concentration differences in the presence of SCM wa slower in HIV-infected women compared to HIV-uninfected women.
These findings suggest that HIV infection is associated with a chronic stimulation of the Th1-related cytokines cascade in breast milk and an impaired ability to respond to SCM. The local inflammation of the mammary gland during SCM modulates both breast milk cell-free and cell-associated HIV levels and was found to increase HIV transmission through breastfeeding in several studies [
25‐
30]. SCM is probably involved in mechanisms fuelling local viral replication and traffic of infected cells in the mammary gland from the vascular compartment [
31,
32]. Cytomegalovirus and Epstein–Barr virus that are part of the normal environment of the mammary gland can also facilitate breast milk HIV shedding and were found associated with HIV-1 transmission by breastfeeding in the same cohort [
33]. Furthermore, in a recent study we reported that impaired capacity to secret IL-8 in breast milk during SCM was associated with detection of Epstein–Barr virus in breast milk from HIV-infected Zambian women which may in turn fuel HIV shedding [
34].
As a cross-sectional study the absence of follow-up is one of the limitations of our study. Infant age at sampling varied, but all specimens were mature breast milk and multivariate analysis took into account this possible confounding factor known to influence immunologic environment in the mammary gland. In addition, the relative homogeneity of our study population was a strength to address the issue of the multiple environmental factors that influence the immune composition of breast milk [
55]. Exclusion of mother having transmitted HIV by breastfeeding may be viewed as a selection bias regarding the global population of HIV infected women. Based on the Vertical Transmission Study, the overall risk of postnatal HIV infection has been estimated at 3.9% among children breastfed for less than 6 months, each additional month of breastfeeding beyond 6 months of age being associated with a 1% risk of acquisition of HIV [
56]. To be representative of the global population of HIV-infected mothers at risk of postnatal transmission, our study should have include an estimated number of two to three mothers having transmitted HIV infection by breastfeeding. However, we did not include women who transmitted HIV to their infant postnatally, assuming that the soluble immunologic pattern in breast milk from HIV-transmitters could form a peculiar type of pattern, as mentioned by several authors [
46,
57‐
59].
Authors’ contributions
MLN and RB contributed to the design, implementation and analysis of the VTS. PVdP, NN, JV and ET conceived and designed the ANRS 1271 study. SB and KB carried out laboratory measures. RS and PD analyzed the data. RS and ET interpreted the results and wrote the manuscript. All authors read and approved the final manuscript.