Maternal and Cord Blood 25-Hydroxyvitamin D₃ Are Associated with Increased Cord Blood and Naive and Activated Regulatory T Cells: The Barwon Infant Study

Development of immune phenotypes in utero and during early postnatal life may influence the risk of both acquiring serious infections and later development of immune-mediated diseases, a hypothesis named the Developmental Origins of Health and Disease (1, 2). Identification of modifiable environmental exposures associated with phenotypic imprinting of specific immune cells may have important scientific and clinical implications (3).

Early-life vitamin D status is one modifiable candidate that has been implicated in a number of chronic conditions that present in early childhood or later in life, such as food allergy (4, 5), asthma (6), persistent wheeze (7, 8), and some autoimmune diseases, including multiple sclerosis (9) and type 1 diabetes (10–12), but results have been inconsistent. There is emerging evidence that vitamin D has modulatory effects on immune function outside its classical role in bone metabolism (13); however, studies examining the role of vitamin D metabolites on immune responses have primarily been conducted on cell lines and in animal models (14, 15). Vitamin D (most commonly measured as 25-hydroxyvitamin D [25(OH)D] levels) can be sourced by dietary forms or generated from UV radiation (UVR) exposure. Furthermore, when examining the role of early-life vitamin D exposure, it is essential to differentiate between 25(OH)D and the more inert epimer C3-epi-25(OH)D [epi-25(OH)D], because of a possible large proportion of epi-25(OH)D of total 25(OH)D levels (8.7–61.1%) (16) and differential cell-specific effects and indirect effects by actors in vitamin D metabolism pathways (17). The vitamin D receptor (VDR), a ligand-dependent transcription regulator molecule, has been demonstrated in all immune cells, including human T cells, and its ligand is the active vitamin D metabolite 1,25-dihydroxyvitamin D [1,25(OH)₂D] (18). VDR expression in T cells is upregulated upon Ag-specific triggering, and T cells have the ability to internally convert 25(OH)D to 1,25(OH)₂D via CYP27B1 expression (19). A number of potential mechanisms by which vitamin D may influence T cell frequency and function have been proposed, including direct effects of systemic 1,25(OH)₂D₃ on various subsets of the T cell population (e.g., CD4⁺ Th cells and regulatory T cells [Treg]) (19–21), or indirect mechanisms, such as induction of tolerogenic dendritic cells that favor differentiation of Tregs (19, 22).

Tregs comprise a small but vital subpopulation of CD4⁺ T cells that have important suppressive and homeostatic immune functions and are predominately derived from the thymus (23, 24), but peripheral development is also present (25). Treg can be further divided into naive Treg (nTreg) and activated Treg (aTreg) according to the degree of FOXP3 positivity and the expression of CD45RA (24, 26). Both nTreg and aTreg populations have been shown to suppress proliferation of stimulated naive CD4⁺ T cells (24), and most recently, the Treg populations isolated from cord blood (CB) exhibited a highly activated and suppressive phenotype (27). The aTreg have a memory (CD45RA⁻) phenotype (28) and are more proliferative, as evidenced by higher expression of Ki67 and CTLA4, respectively (29). At birth, the nTreg population predominates (30), and upon Ag stimulation, nTreg differentiate into aTreg, which are capable of suppressing the proliferation of other Tregs as well as non-Tregs in vitro (26). Lower Treg levels at birth have been associated with a higher risk of subsequent childhood allergy (31, 32). Therefore, it is important to understand the influence of modifiable prenatal factors on Tregs at birth, such as prenatal vitamin D levels.

Few studies have investigated the relationship between maternal vitamin D status and Tregs at birth (33, 34), and findings have been inconsistent, due in part to variation in characterization of the Treg phenotypic, failure to account for important confounding factors and differences in vitamin D level distributions across studies. Importantly, studies have not examined the influence of antenatal or birth vitamin D levels on Treg subpopulations, such as nTreg and aTreg (24), nor have they accounted for the potential role of maternal UVR exposure throughout pregnancy in these associations. UVR may exert independent immune effects (35–37), and it is therefore important to assess it separately when examining 25(OH)D and immune function.

In a cohort of Australian mothers and infants, recruited using an unselected antenatal sampling frame, we investigated the association between maternal and CB 25(OH)D₃ and epi-25(OH)D₃ and the proportion of both nTreg and aTreg populations (30) in CB (38), adjusted for the maternal personal UVR dose in the three trimesters.

The aims and methodology for the Barwon Infant Study (BIS) have been described previously (39). In brief, BIS is a birth cohort study (n = 1064 mother–1074 infant pairs [10 sets of twins]) with antenatal recruitment conducted in the Barwon region in Victoria, Australia. The Barwon region population characteristics are similar to those of the overall Australian population but with a smaller proportion of families from non-English speaking backgrounds (39). Pregnant women were recruited before 28 wk of gestation between years 2010 and 2013. Of the invited families, 33% agreed to participate (39). Nonresponse was because of anticipated infant biosample burden, and brief demographic information was collected on nonresponders (39). Exclusions included birth <32 wk of gestation or serious disease or genetic abnormality (39). Data on vitamin D metabolite measurements were available at 28–32 wk and birth, and Treg proportions in CB (30, 38).

Our study consists of a random subsample of mother–infant’s dyads with both vitamin D metabolite measurements and Treg phenotyping at prespecified time points (Supplemental Fig. 1). No power calculation was performed because of the fact that both exposure and outcome variables were already measured, and we therefore refer to the presented 95% confidence intervals (CI).

Flow cytometry was used to characterize subpopulations of Tregs, as previously described (31). In brief, isolated mononuclear cells were stained with Abs to CD4-PE and CD45RA-PECy5, washed in PBS, and formalin fixed. After overnight fixation, cells were permeabilized (0.5% Tween) and stained with anti-FOXP3–Alexa Fluor 488. All samples were analyzed by the three-channel flow cytometry (FACSCalibur; Becton Dickinson) (30). For information about blood sampling and processing, see Collier et al. (30). Characterization of the thymic-derived Treg subsets was performed in accordance with the method used by Miyara et al. (24, 26), with the CD4⁺CD45RA⁺ FOXP3^low subset equalling nTreg and CD4⁺CD45RA⁻ FOXP3^high subset equalling aTreg. Treg proportions were presented as a percentage of CD4⁺ T cells (gating strategies are illustrated in Fig. 1) and are termed as “levels” for ease of reading. Treg data at 6 and 12 mo of age were also examined to determine the extent that maternal and CB 25(OH)D₃ levels were associated with later postnatal Treg levels (Supplemental Table I).

In addition, data were available for another CD4⁺ T cell subset that was gated according to CD4⁺CD45RA⁻ FOXP3^low and designated as “nonsuppressive FOXP3⁺ T cells” (24) (Fig. 1). Nonsuppressive FOXP3⁺ T cells produce proinflammatory cytokines (e.g., IL 17) and may therefore promote inflammation and autoimmunity (24, 40, 41), but the immunological function of this Treg subpopulation is largely unknown.

The findings for this Treg subpopulation are presented in Supplemental Tables II, III.

The 25(OH)D₃ metabolites and epi-25(OH)D₃, were measured in serum using two-dimensional ultra-performance liquid chromatography separation-coupled tandem mass spectrometry detection (2D LC-MS/MS), which is described in detail elsewhere (42). The lower limit of quantification for 25(OH)D₃ was 2 nmol/l. The interassay coefficient of variation for 25(OH)D₃ and C3-epi-25(OH)D₃ was 0.5 and 5.3% at a 25(OH)D₃ concentration near 70 nmol/l, respectively (42). We have previously reported, in a random BIS subcohort (n = 233), that only 9% of the mothers have 25(OH)D₃ levels below 50 nmol/l, but 45% of the CB 25(OH)D₃ levels were below 50 nmol/l (38).

Questionnaire data quantifying exposure to direct sunlight daily were recorded during all three trimesters. The ambient UVR was estimated using monthly averages of daily total ambient UVR, provided by the Australian Radiation Protection and Nuclear Safety Agency, in standard erythemal doses (43) for Melbourne from 2010 to 2014. The total UVR exposure dose (maternal personal UVR dose) was calculated as the product of time in direct sun and the average of the daily ambient UVR exposure. Maternal personal UVR dose data were calculated for three time points; 12-wk gestation (end of first trimester), 27-wk gestation (end of second trimester), and date of birth (end of third trimester) (38).

Our group has recently reported a number of antenatal and perinatal factors that influence birth Tregs, including exposure to labor (compared with a planned Caesarean section) (44), duration of labor (45), sex (32), birth weight z-score (32), and preeclampsia (46). We additionally evaluated each of the following covariates, one at a time, as possible confounders or intermediate variables (mediators): season at birth, antibiotics given during labor, gestational diabetes, maternal smoking during pregnancy, parental age, ethnicity, number of siblings in the household during pregnancy, family history of asthma, livestock, family pets during pregnancy, and socioeconomic status. These data were collected from participant questionnaires and hospital records.

Descriptive data were expressed as median (quantile 1 to quantile 3 [Q1–Q3]) for continuous variables and frequency/percentage of total for categorical variables. Spearman rank correlation test was used for Fig. 2A–D. The primary plan of analysis was determined a priori. The primary analysis was a multiple regression analysis with the CB Treg subset levels as outcomes and maternal and CB 25(OH)D₃ measurements as our exposures of primary interest. Treg data were log transformed. After running the models, regression coefficients were back transformed using the antilogarithm; results are therefore interpreted as relative mean differences in the expected geometric mean of Treg subset levels per 25-U (nmol/l) increase in vitamin D metabolites [maternal (28- to 32-wk gestation) or CB 25(OH)D₃]. Our models only included participants with information on all covariates [Supplemental Fig. 1 depicts data availability regarding Tregs and 25(OH)D₃].

Preselected covariates, including those based on prior knowledge (47), were tested if they were confounders or mediators. If a potential confounder was identified, it was included in the model if it was associated with a >10% change to the regression coefficients for 25(OH)D₃ (48, 49). Sex and gestational age were routinely included.

In a series of sensitivity analyses, we examined the following: 1) overall functional misspecification by including 25(OH)D₃ and maternal personal UVR dose as continuous variables. (Specifically, we included a quadratic term of 25(OH)D₃ and maternal personal UVR dose [fractional polynomials (50)] and tested whether the alternative model fitted the data better than the linear model (after log2 transformation) using a likelihood ratio test. Furthermore, 25(OH)D₃ quintiles (data derived) were also examined to assess possible nonlinearity); 2) the association between total 25(OH)D₃ levels [25(OH)D₃ plus epi-25(OH)D₃] and the Treg subset in our adjusted analyses; 3) a threshold analysis was performed modeling maternal and CB 25(OH)D₃ as dichotomous (below or above 50 nmol/l, deficient/insufficient versus replete) based on recommendations from the Institute of Medicine (14, 51); 4) interaction between maternal personal UVR dose and 25(OH)D₃ on Tregs; 5) whether interaction between the mode of delivery and duration of labor might “unmask” associations between 25(OH)D₃ and Tregs (44); 6) excluding twins from the analysis; and 7) whether selection bias due to nonresponse influenced our main results using an inverse probability weighting (IPW) approach (52, 53). Briefly, first we calculated the probability for responding in the BIS cohort for each mother with respect to family history of asthma, socioeconomic status, number of residents in the household, and maternal age. Second, the weight, which is the inverse of the probability of responding, was computed and included in the main analyses. Each BIS mother that responded is now not only accounting for itself but also for those with similar characteristics who did not respond when invited for study participation (53).

All p values are evaluated at a 5% significance level, and no multiple testing adjustments were made; however, all results are reported so that readers can interpret findings accordingly (49, 54). This should enable the reader to reach a reasonable conclusion on their own while having in mind the risk of a type 1 error (false positive) (54).

All analyses were made using the statistical software package R, version 3.5.1 (the R Foundation for Statistical Computing, Vienna, Austria) and the add-on package “survey.”

All mothers provided written informed consent. Ethics approval was obtained from the Barwon Health Human Research Ethics Committee (10/24).

The characteristics of the study participants and measurements are presented in Table I. The median (Q1–Q3) 25(OH)D₃ was 85.4 (66.7–97.6) and 50.7 (37.8–64.2) nmol/l for maternal and CB samples, respectively. Of these 164 maternal blood samples, 158 were paired with a matched CB sample, resulting in 158 mother–infant pairs with both Treg and 25(OH)D₃ measurements (Supplemental Fig. 1). Pearson correlations between maternal and CB 25(OH)D₃ and epi-25(OH)D₃ were both moderate (r = 0.47; p < 0.0001) and (r = 0.52; p < 0.0001), respectively. The epi-25(OH)D₃/25(OH)D₃ percentage was 5.7 and 9.8% for maternal and CB measurements, respectively, which is comparable with results from other studies (55, 56).

Thirteen percent of the mothers and forty-nine percent of the neonates had 25(OH)D₃ below 50 nmol/l, respectively.

Fig. 2A, 2B shows a positive correlation between maternal 25(OH)D₃ levels and CB subset Treg levels.

The multivariate regression analysis also demonstrated a positive association between maternal 25(OH)D₃ levels and CB nTreg (relative adjusted mean difference [AMD] per 25 nmol/l increase: 1.05; 95% CI: 1.01–1.09), and aTreg (AMD per 25 nmol/l increase: 1.17; 95% CI: 1.06–1.28) (Table II). This equated to a 5 and 17% increase in the expected geometric mean of nTreg and aTreg levels, respectively, for a 25 nmol/l increase in maternal circulating 25(OH)D₃ levels. Further adjustment for a number of possible confounders did not alter the findings in Table II (e.g., ethnicity, maternal smoking during pregnancy, preeclampsia, gestational diabetes, pets during pregnancy, number of siblings in the household during pregnancy, and antibiotics given during labor, etc.; all covariates examined as possible confounders are listed in the 2Materials and Methods section). Results persisted after adjustment for season at birth and maternal personal UVR dose. Additionally, no changes in estimates were found when adding epi-25(OH)D₃ levels to 25(OH)D₃ levels.

Last, analyses demonstrated that maternal 25(OH)D₃ levels were not associated with nTreg and aTreg at 6 and 12 mo of age (Supplemental Table I).

Fig. 2D shows a positive correlation between CB 25(OH)D₃ and CB aTreg.

The multivariate regression analysis also demonstrated an association between CB 25(OH)D₃ levels and CB aTreg (AMD per 25 nmol/l increase: 1.29; 95% CI: 1.13–1.48) (Table II). Further adjustment for a number of possible confounders did not alter the findings in Table II (all covariates examined as possible confounders are listed in the 2Materials and Methods section). The association between CB 25(OH)D₃ levels and CB aTreg persisted after adjustment for season at birth and maternal personal UVR dose. Additionally, no changes in estimates were found when adding epi-25(OH)D₃ levels to 25(OH)D₃ levels.

Last, analyses demonstrated that CB 25(OH)D₃ levels were not associated with nTreg at 6 and 12 mo of age and aTreg at 12 mo but were positively associated with aTreg at 6 mo (AMD per 25 nmol/l increase: 1.20; 95% CI: 1.08–1.33) (Supplemental Table I).

Between 396 and 459 mother–infant pairs had complete maternal personal UVR dose data. We found a minor inverse association between higher maternal personal UVR dose in the end of the first trimester and CB aTreg (Table III). This association persisted after adjustment for sex, gestational age, and season at birth but did not persist when including maternal 25(OH)D₃ into the model. That is, the association between personal UVR exposure in the first trimester and CB aTreg was not independent of 25(OH)D₃. The distribution of Tregs was not influenced by season (Supplemental Table IV).

Sensitivity analyses were performed as described above. We found no evidence for nonlinear associations between 25(OH)D₃ or maternal personal UVR dose on Tregs. For example, the association between 25(OH)D₃ in a linear form and Treg subpopulations fitted the data as well as one in which five separate 25(OH)D₃ categories were fitted (data-derived quintiles). No changes in results were seen when modeling Treg subsets by total 25(OH)D₃. When 25(OH)D₃ levels were split into two groups (above or below 50 nmol/l), a 41 and 24% increase in the expected geometric mean of CB aTreg levels was found when maternal and neonatal 25(OH)D₃ levels were above 50 nmol/l, respectively. No difference in nTregs proportions were found. We also found no interactions between maternal personal UVR dose (all three trimesters) and maternal or CB 25(OH)D₃ levels in relation to the levels of Tregs at birth, and there was no change in our results when including interaction terms between exposure to labor and duration of labor in our models. In addition, our results did not differ when twin pairs (n = 2) were excluded. Last, we evaluated selection bias using an IPW approach and found little influence on the findings, with one exception. Following IPW, a positive association between maternal 25(OH)D₃ and nonsuppressive FOXP3⁺ T cells was also evident (Supplemental Table V).

To our knowledge, this is the first study to examine the association between both maternal (28-week gestation) and CB 25(OH)D₃ and a number of CB FOXP3⁺ CD4⁺ populations. We found positive associations between maternal (28-week gestation) 25(OH)D₃ levels and CB nTreg and aTreg. There was also a positive association between CB 25(OH)D₃ levels and CB aTreg. The results persisted after adjustment for maternal personal UVR dose throughout pregnancy and other potential confounding factors. It is noteworthy that the findings were of considerable magnitude, ranging from a 5 to 29% change in Treg subset proportions for a 25 nmol/l increase in 25(OH)D₃. Furthermore, CB Treg levels varied by maternal (28-week gestation) and CB 25(OH)D₃ levels, even in this vitamin D–replete population. In this study, our median 25(OH)D₃ levels (85.4 and 50.7 nmol/l for maternal and CB samples) were higher than previous studies, which indicates that the influence of vitamin D on Treg levels is not limited to deficient/insufficient populations.

It has been shown in vitro that both nTregs and aTregs are actively suppressive (24), with evidence that aTregs originate from proliferating nTregs (24, 26). Activated but not naive (CD45RA⁺) human CD4⁺ T cells express VDR, and VDR mediates the biological effects of 1,25(OH)₂D by up- and downregulating hundreds to thousands of genes in CD4⁺ cells (57, 58). Gene regulation by 1,25(OH)₂D–VDR genomic binding has been reported to be positively and highly correlated (r = 0.92) with 25(OH)D levels in CD4⁺ cells in healthy adults (which are also specifically expressed in Tregs), and this may result in altered T cell activation and differentiation (58). Furthermore, activated CD4⁺ T cells are able to convert 25(OH)D to 1,25(OH)₂D (57), which supports our finding that the aTreg proportion is more strongly associated with 25(OH)D₃ than the nTreg population. nTreg differentiation may also be influenced by 1,25(OH)₂D, for example, through indirect immunological pathways (23, 24). It is unclear whether the aTreg subpopulation contains a memory-type long-living Tregs (24).

Human studies examining the link between 25(OH)D and overall Tregs are sparse. In line with our results, an Indian study (n = 153 mother–infant pairs) found a positive association between maternal (third trimester blood sampling) and CB 25(OH)D₃ status and maternal and CB Tregs, as well as a positive association between maternal 25(OH)D₃ and placental FOXP3 gene expression (59) (n = 24); however, no FOXP3 subset categorization was performed in that study (59). A Turkish study using CB from full-term neonates (n = 101) found no correlation between 25(OH)D and the overall Treg percentage of CD4⁺ or the number per microliters (60). However, this study has some limitations; the main analyses were stratified by 25(OH)D levels above or below 30 nmol/l, resulting in reduced power, and possible confounders were not considered. Also, like the previous study, Tregs were not differentiated into FOXP3⁺ subpopulations.

A German study with 25(OH)D₃ and Treg data on 378 mother–child pairs found no correlation between maternal (34 weeks of gestation) 25(OH)D₃ and Tregs but did find a small negative correlation (r = −0.17) between CB 25(OH)D₃ and Tregs (61) quantified using a FOXP3 demethylation method (62). The median 25(OH)D₃, however, was notably higher in our study population compared with the German study population (median maternal levels 85.4 versus 55.4 nmol/l and CB levels 50.7 versus 27.5 nmol/l). Chi et al. (33) found weak negative correlations (r = −0.29 to 0.20) between CB 25(OH)D levels (median CB levels 50.5 nmol/l) and CB CD25⁺, CD25^bright/high, and CD25⁺ FOXP3 cells in their subcohort (n = 55–106), but they found no correlation between CB 25(OH)D levels and the Treg-suppressive index (n = 69). As seen in Fig. 2C, 2D our Spearman rank correlations coefficients (r) shows no (r = −0.01) or a positive (r = 0.25) relationship between CB 25(OH)D₃ and nTreg and aTreg subsets, respectively. Our results do not converge with those mentioned above by Weisse et al. (61) and Chi et al. (33) Differences in the study participants, confounding bias, differing distribution of 25(OH)D levels across populations, and the measurement/differentiation of Tregs may contribute to the observed differences.

A small randomized controlled trial (RCT) (n = 26) found no difference in proportion (percentage of CD4⁺) of CB CD25⁺ FOXP3 cells between mothers that had been supplemented with 400 IU/d versus 4400 IU/d of vitamin D₃ from 10- to 18-weeks gestation until birth, with mean CB 25(OH)D₃ of 40.5 versus 63.5 nmol/l in each group, respectively (34). An RCT is a powerful design, but this trial was underpowered (34).

We find no evidence for an independent association between maternal personal UVR dose throughout the entire pregnancy and the CB Treg subsets (n = 396–459 mother–infant pairs, depending on trimester) (Table III). In a United States study using an adult population (n = 350), they also found no association between UVR and overall proportion of circulating Tregs (36). Unfortunately vitamin D measures were not available in that study, so vitamin D–independent effects of UVR on Treg subsets could not be examined (63). Studies examining the role of phototherapy (narrow-band UVB) in adult populations as therapy for various dermatological diseases have shown conflicting results concerning the association between narrow-band UVB phototherapy and circulating Treg proportion (64, 65). Further work on this issue is required.

Treg immune regulatory functioning serves as an essential mechanism of negative regulation of immune-mediated responses and features prominently in autoimmune and autoinflammatory disorders, allergies, acute and chronic infections, cancer, and metabolic inflammation (23). Enhancing immunological homeostasis via a Treg pathway early in life through modifiable perinatal exposures (i.e., vitamin D exposure) may help, in part, to prevent some of the abovementioned adverse conditions.

Noteworthy, a previous study in BIS found that higher nTreg proportions at birth were sustained over the entire first year of life, which indicates long-lasting effects of the perinatal environmental exposures on nTregs (44), in line with the Developmental Origins of Health and Disease hypothesis (1). Consistent with this, CB 25(OH)D₃ levels were associated with aTreg levels at six postnatal months as well as birth. In regard to allergic disease, an inverse association between the CB nTreg proportion and later risk of IgE-mediated food allergy has been reported in this birth cohort (44), but pre- and perinatal 25(OH)D₃ levels were not associated with IgE-mediated food allergy at 1 year of age (38), despite being associated with both CB nTreg and aTreg in the current study. It is likely that this partly reflects that birth Treg levels are determined by multiple factors, not only 25(OH)D₃ levels. Prenatal exposure to farm animals and raw farm milk (66, 67) and family pets (68) may also impact neonatal Treg subpopulations as well as perinatal factors (44) and parental allergy history (31, 69). Thus, higher 25(OH)D₃ levels are likely to be only one of several factors boosting Tregs at birth.

The strengths of this study include the study design, a prospective population-derived antenatal sampling frame, multiple time points of 25(OH)D₃ measurement, valid information on UVR exposure in all three trimesters, deep phenotyping of CB Treg subsets that enable us to differentiate nTreg from aTregs, 25(OH)D₃ were able to explain up to 8% (adjusted R²) of the variation in neonatal aTregs (this may be of clinical importance), we were able to address the role of both maternal and CB epi-25(OH)D₃ levels on CB Treg proportions, and the BIS population is mainly of European ancestry (39), which reduces the likelihood of population stratification issues. As in any observational study, we cannot exclude the possibility that unknown confounding factors may have influenced our results, but we have examined multiple possible confounders and adjusted our analyses accordingly. Limitations of this study include the lack of data on Treg in vitro suppressive function. Also as the half-life of 25(OH)D₃ is around 2–3 weeks (70), our conclusions regarding the association between 25(OH)D₃ and Tregs are limited to the third trimester, although moderate correlations (r = 0.36–0.53) have been found between season-corrected trimester-specific 25(OH)D levels (71, 72). Unfortunately, we have no data on maternal 25(OH)D₃ during preconception, first, and second trimester, which may be crucial periods for immunological development/health in both the fetus and the mother (22, 73, 74). Mothers in the BIS cohort was generally vitamin D replete, but variation in 25(OH)D₃ levels were still associated with CB Treg profiles. The study lacked the power to investigate the role of maternal vitamin D deficiency on Tregs in detail, although we found that maternal 25(OH)D₃ levels above 50 nmol/l were associated with higher aTreg proportions of 41%.

Our results are based on circulating systemic Treg proportions, and we cannot therefore draw conclusions regarding possible tissue-specific Treg proportions (i.e., compartmentalization of Treg) (75), but these are not in high numbers at birth, and examining the latter would require more invasive procedures than a blood sample (76). We had no prenatal maternal Treg data, but CD4⁺FOXP3 Treg subsets have shown moderate correlation (r = 0.61) between maternal (2–5 hours before delivery), and CB (77), which suggests that the possible effects of 25(OH)D₃ on CB Tregs may be extrapolated to the maternal circulation during pregnancy.

In this study, we found no sign of selection bias concerning our main results, and our findings remained evident and were of higher magnitude after reweighting using IPW methods (Supplemental Table V). IPW does not address unknown or unmeasured factors that influence selection, but this method creates a pseudo-population that accounts for bias introduced by initial nonparticipation or exclusion from the sample with both vitamin D and Treg measures (53).

In summary, our results suggest that fetal and early-life 25(OH)D₃ may be a modulator of the neonatal immune system through a FOXP3⁺ Treg pathway, and effects seem to last throughout the first 6 months of life. Furthermore, the positive association between 25(OH)D₃ and Tregs persists after adjusting for maternal personal UVR dose, providing reassurance that vitamin D is not acting merely as a marker of UVR exposure and UVR direct immune effects.

The biological and clinical significance of these findings, however, remains uncertain, and results should be confirmed using other approaches, such as sufficiently powered RCT using different 25(OH)D₃ doses and with the same or extended Treg subset identification and functional assessment (78). Such a study may be informative, given the considerable molecular evidence that supports 1,25(OH)₂D₃ as a regulator (directly and indirectly) of FOXP3⁺ Tregs (19, 79, 80). This investigation could be incorporated into the design of new RCTs assessing prenatal vitamin D supplementation. It is noteworthy that because CB Tregs predict later child allergic disease (31, 32), these findings highlight the importance of the prenatal period in immune programming.

We thank the BIS participants for the generous contribution they have made to this study. The establishment work and infrastructure for the BIS was provided by the Murdoch Children’s Research Institute, Deakin University, and Barwon Health.

John Carlin, Mimi Tang, Fiona Collier, Terry Dwyer, Sarath Ranganathan, Peter Sly, Len Harrison, and David Burgner.

The authors have no financial conflicts of interest.