Background
Perinatal supplementation with probiotics has been shown to reduce the incidence of atopic dermatitis (AD) in infancy [
1,
2]. Our own study, the Probiotics in the Prevention of Allergies among Children in Trondheim (ProPACT) trial, demonstrated a 40 % reduction in the development of AD following maternal probiotic supplementation [
3]. However, the biological mechanisms behind this effect are incompletely understood. Using samples taken during the ProPACT trial, we have previously reported that the maternal intestinal microbiota is modified by probiotic supplementation, and that children born to mothers who received the probiotics have a higher abundance of
Lactobacillus rhamnosus GG (LGG) which persists up until 3 months of age [
4]. Whilst the transfer of such probiotic bacteria from mother to child may prevent AD, alterations in various components of breast milk have also been postulated as possible mediating factors.
In addition to being a source of nutrition and hydration for newborn infants, breast milk contains a number of immunologically active cells and molecules, such as immunoglobulins, lactoferrins, growth factors and cytokines [
5]. Among the cytokines found in breast milk are thymic stromal lymphopoietin (TSLP) and the transforming growth factor-β (TGF-β) cytokines. TSLP has been implicated in the establishment and maintenance of T helper type 2 (Th2) responses, and thus also in defence against helminthic infections and in the pathogenesis of allergy related diseases [
6]. Genetic variants of TSLP have been associated with AD and asthma [
7] and high levels of epidermal TSLP precede the clinical presentation of childhood AD [
8], suggesting that TSLP may be particularly important in the establishment of AD. Other studies have demonstrated that TSLP activates skin dendritic cells promoting a Th2 response and interacts directly with skin-homing Th2 cells to enhance interleukin-4 (IL-4) production which is thought to contribute to the maintenance of inflammation in chronic AD [
9,
10]. Consistent with these studies, higher concentrations of TSLP have been reported in both acute and chronic AD lesions [
6]. Murine models have shown that over-expression of TSLP in keratinocytes induces an AD-like skin disease and can predispose to allergic airway inflammation after intranasal challenge. Thus, TSLP may also be involved in the progression to other allergy related diseases, a process often referred to as the atopic march [
6]. However, the biological effects of breast milk TSLP for the mother and child are unknown [
11], and this cytokine is previously unstudied within the context of maternal probiotic supplementation. Neither has the effect of maternal atopy on breast milk TSLP been investigated, nor the association between breast milk TSLP and the development of AD. In contrast to TSLP, the human isoforms of TGF-β (TGF-β
1, TGF-β
2 and TGF-β
3) are primarily implicated in inhibition of allergic inflammation through a wide range of immunoregulatory effects [
12]. All three isoforms of TGF-β are found in breast milk. Studies investigating the effect of probiotic supplementation on breast milk TGF-β concentrations and or their association with later allergy related disease in offspring, have produced conflicting results [
13‐
18].
Both TSLP and TGF-βs appear to have important roles in acute and chronic phases of allergy related disease, which makes them interesting as potential mediators of this preventive effect. The aims of the current study were to: (a) determine if perinatal maternal probiotic supplementation alters the concentration of TLSP, TGF-β1, TGF-β2 or TGF-β3 in breast milk at 10 days and 3 months postpartum and (b) investigate if these breast milk cytokines contribute to the preventative effect of maternal probiotic supplementation on the development of AD at 2 years of age through causal mediation analysis.
Discussion
Thymic stromal lymphopoietin was the only cytokine with a tendency to be affected by maternal probiotic supplementation, resulting in higher concentrations of TSLP at 10 days postpartum. Causal mediation analysis indicated that the higher concentrations of TSLP at 10 days did not appear to contribute to the beneficial effect of probiotics on the development of AD in offspring. TSLP concentrations at 3 months postpartum and TGF-βs at both time points were not affected by probiotic supplementation. As such, these breast milk cytokines are unlikely to be mediators of the beneficial effect of probiotics.
This study provides the first report of the influence of a maternal probiotic supplementation on breast milk TLSP concentrations and investigates its contribution to the prevention of AD. In experimental and observational studies, TSLP has been associated with promotion of a Th2 and allergic type inflammation, particularly in the lung and skin [
6,
24]. Therefore, the finding that probiotic supplementation increased the breast milk TSLP concentrations was contrary to what one might have expected. At the same time, high breast milk TSLP may not be as detrimental in the intestinal system where it is described as having a more regulatory function [
24,
25]. TSLP released by intestinal epithelial cells in response to commensal bacteria has been shown to promote tolerogenic properties in dendritic cells and macrophages which then produce IL-10 and retinoic acid and promote regulatory T helper (Treg) cell differentiation [
25]. Thus the increased TSLP seen in the probiotic group may encourage intestinal immune homeostasis, which in turn is theoretically beneficial for the development of the neonatal immune system and prevention of AD. However, the current data does not suggest that breast milk TSLP contributes to the preventative effect of probiotics on AD, nor does it appear to be associated with the development of AD. It is unclear if the observed lack of mediating effect is due to a loss of information and power due to dichotomisation of TSLP concentration, TLSP degradation prior to reaching the intestines, or a more complex interplay between TSLP, the intestinal microbiota and the immune system. Further studies are required to confirm these findings.
The major strength of this study is that the randomised placebo controlled, double blind design of the ProPACT trial allows an unconfounded assessment of the effect of the probiotic regime on the selected breast milk cytokines. Furthermore, through causal mediation analysis techniques, we are able to estimate if and to what extent the prevention of AD is due to changes in TSLP. The randomised design also means that two of four assumptions required for this mediation analysis [
21] are automatically satisfied, namely that there is no residual confounding of the effect of probiotics supplementation on either AD or breast milk TSLP. The other two required assumptions are that there is no unmeasured confounding of the relationship between breast milk TSLP and AD and that there is no confounder of the mediator-outcome relationship that is affected by treatment allocation. Whilst these cannot be excluded, we believe we have measured and included all variables which are likely to affect both breast milk cytokines and AD and that probiotics do not have a strong effect on any other confounders of the relationship between these variables. Another strength of this study is that it is the largest to investigate TSLP concentrations in breast milk and the only one to consider infant outcomes and maternal atopy status. The only other study to report breast milk TSLP concentrations involved 44 women who submitted samples at one of two time points [
11]. Whilst they also found that mature milk tended to have lower TSLP concentrations, they reported concentrations which all fell within a relatively narrow range. Our results suggest that there may be a much greater variability between individuals than this small study managed to detect.
One of the limitations of this study is that the breast milk samples were stored between 6 and 8 years before analysis. This has potentially affected the measured concentrations of TSLP and TGF-βs, as has been previously reported for TGF-β
1 [
26]. However, there was no difference in the average length of storage between the treatment groups and adjusting for length of storage did not significantly alter the estimated effect of probiotic supplementation on any of the measured cytokines (data not shown). Another limitation is that the time of sample collection was not standardised with respect to time of day and whether fore- or hindmilk was collected. Although diurnal and fore-/hindmilk variation has not been specifically investigated for TSLP and TGF-βs, previous studies suggest that diurnal variation exists in some breast milk cytokines [
27] and that the quantity of total protein in the aqueous portion of fore- and hindmilk is reasonably constant [
28]. Whilst these factors may have influenced the cytokine concentrations, samples from the probiotic and placebo groups are presumably equally affected and we do not believe they have significantly affected the conclusions of this study. An underlying uncertainty surrounding the measure of breast milk cytokines is another limitation to this study. Previous studies of breast milk TGF-β and other cytokines have produced widely varying concentration measurements. Whilst some of this variability may be explained by factors such as diurnal and fore-/hindmilk variations, ethnicity, maternal atopy status and time point postpartum, it is likely that sample storage, preparation, analysis methods and inter-laboratory differences play a major role in the measured concentrations. As such, it is difficult to compare results between studies, particularly when these variables are inconsistently reported. In studies comparing two exposure groups, be that probiotic versus placebo or maternal atopy versus no atopy, it is reasonable to assume that technical variability between the groups are minimised. Thus, comparing groups within each study is valid and more generally the relative relationship between these groups can be roughly compared across studies.
Five previous studies have reported breast milk TGF-β
1 and or TGF-β
2 concentrations from similar randomised trials of probiotic supplementation. Each of these studies found a statistically significant effect of probiotics on at least one TGF-β isoform in colostrum samples and no effect on concentrations in mature milk. However, they provide conflicting evidence as to whether probiotics increase [
14,
15] or decrease [
16‐
18] colostrum TGF-β concentrations. The current study also found no effect of probiotic supplementation on TGF-β concentrations in mature milk, and provides no evidence to suggest either increased or decreased concentrations in breast milk at 10 days postpartum, a time when the milk is sometimes referred as transition milk before colostrum become mature milk. In addition to the previously mentioned sources of variability, the differing results may also reflect a probiotic strain specific effect as none of these studies used the same probiotic species in the same dose or combinations. In terms of methodological differences, it is interesting to note that the two studies reporting reduced TGF-β
2 after probiotic supplementation also reported the manufacturer recommended acid activation of the breast milk samples prior to measurement which means that the reported concentrations represent the entire pool of TGF-β which is naturally found in both a latent and active form. On the other hand, all three studies reporting higher TGF-β
1 or TGF-β
2 do not mention this activation step in their methods. One of these studies, however, used an alternate activation procedure which is thought to additionally release lipid bound TGF-β [
17,
29]. All other studies have removed the lipid fraction prior to cytokine analysis. If the conflicting reports of positive or negative associations with probiotics are a reflection of a common effect of probiotics and sample preparation method, rather than strain specific effects, it may imply that probiotics affect the ratio between active and latent TGF-β in breast milk or the proportion of lipid bound TGF-β. Other factors are known to affect the proportion of active versus latent TGF-β, for example, mothers of premature infants having a higher proportion of latent TGF-β [
30].
Authors’ contributions
TØ, OS and RJ designed and directed the implementation of the ProPACT trial during which these breast milk samples were collected. TØ and ØG conceived and designed the breast milk analysis experiments. MS and ADBR analysed the data and wrote the initial manuscript draft. All authors have contributed to and approved the final draft of the manuscript.