Allergic women show reduced T helper type 1 alloresponses to fetal human leucocyte antigen mismatch during pregnancy

Abstract

Low-level alloreactivity between mother and fetus may provide stimulation for fetal T helper type 1 (Th1) cell immune maturation. This study explored the effects of human leucocyte antigen (HLA) mismatch on materno–fetal interactions detected as cytokine responses and lymphoproliferation in mixed lymphocyte reactions, and whether this was altered in allergic women (n = 62) who have a Th2 propensity compared with non-allergic women (n = 65). HLA-DRβ1 mismatch was associated with significantly increased Th1 interferon (IFN)-γ, Th2 interleukin (IL)-13 and lymphoproliferative responses by both mothers and fetuses. Allergic women showed significantly lower IFN-γ Th1 production in response to HLA-DRβ1 mismatch. The infants of these women also showed significantly lower IL-10 and lower IFN-γ production relative to IL-13. Both HLA-DRβ1 mismatch and maternal allergy had significant independent effects on maternal IFN-γ Th1 responses. Maternal allergy modifies HLA-mediated alloreactivity between the mother and the fetus, reducing Th1 activation. This may affect the cytokine milieu at the materno–fetal interface and could be implicated in the attenuated Th1 responses observed commonly in infants of atopic mothers.

Introduction

There is growing evidence that immunological interactions that influence the cytokine environment during pregnancy may influence subsequent immune programming of the fetus. The capacity for fetal production of T helper type 1 (Th1) responses has been associated with reduced risk of subsequent allergic disease [1–4], and we have shown previously that fetal Th1 interferon (IFN)-γ production (in response to both maternal alloantigens and mitogens) is associated with maternal human leucocyte antigen (HLA) mismatch [5]. Foreign paternal HLA expressed on fetal cells provide a source of low-grade maternal immune activation. Based on our previous studies in pregnancy [5], we speculate that this may vary with the degree of HLA mismatch, thereby influencing the cytokine milieu at the materno–fetal interface. Thus, alloreactivity between the mother and the fetus may provide a stimulus for fetal Th1 immune maturation.

Unlike most other allografts (i.e. organ transplantation) a number of systemic hormonal and local placental mechanisms have evolved to skew maternal allogeneic responses towards Th2 cytokine production in order to protect the fetus from Th1-mediated maternal rejection. Foreseeably, differences in HLA compatibility could contribute to subtle variations in the balance between Th1 allogeneic reactivity and the success of pro-Th2 immune deviation. We have shown recently that allergic women with an increased Th2 propensity have reduced Th1 IFN-γ alloresponses to their fetus in pregnancy compared with non-allergic women [6], and we speculate further that allergic women react differently to the same degree of HLA mismatch to non-allergic women. As also suggested by our previous studies, prior exposure to paternal HLA antigens in previous pregnancies may also affect these alloresponses [5,6].

Previously, we used mixed lymphocyte reactions (MLR) to measure the level and pattern of alloreactivity between maternal and fetal immune cells, and demonstrated differences in both cytokine responses and lymphoproliferation [5,6]. These effects are mediated most probably by CD4⁺ T helper (Th) cells activated by foreign Class II HLA molecules: HLA-DP, -DQ and -DR. HLA-DR mismatch is known widely to be a source of T cell stimulation in MLR responses between unrelated individuals. HLA-C-mediated interactions between extravillous trophoblasts and uterine natural killer (NK) cells may be an additional source of immune stimulation during pregnancy. Trophoblasts do not express Class II antigens, but rather the Class I antigen HLA-C, which contains a bi-allelic set of ligands (C1 and C2) for the killer-cell immunoglobulin-like receptors (KIR) on NK cells [7,8]. Uterine NK cells are abundant during pregnancy and are thought to regulate the invasion of trophoblasts into maternal tissues through these HLA-C interactions by production of a range of cytokines, including IFN-γ[9].

Whereas our previous study used HLA antibodies as an indirect measure of HLA reactivity, for the first time we have examined here these cytokine relationships using more detailed HLA typing. Specifically, we determined if the degree of HLA mismatch based on HLA-DRβ1 alleles, and HLA-C C1/C2 epitope mismatches resulting in potential NK alloreactivity, predicts MLR responses between mother and fetus and the resulting cytokine production. Secondly, we determined whether maternal allergy and previous exposure to paternal antigens (gravidity) alters maternal response to HLA mismatch. Finally, we examined the effects on developing fetal immune responses.

Methods

Study population

This was a longitudinal study of 169 healthy pregnant women recruited from the private clinics of obstetricians in the Perth metropolitan area, Western Australia between August 2002 and March 2005, as described previously [6]. Briefly, women aged ≥ 18 and ≤ 44 years of age were recruited at 20 weeks gestation into study groups defined by allergic status and gravidity. Only healthy term pregnancies were included. All women were non-smokers. The women were reviewed regularly through their pregnancy and the post-partum period [6]. Maternal allergy was defined as a doctor-diagnosed clinical history of asthma, eczema or allergic rhinitis plus a positive skin prick test (SPT) to one or more common allergens. Non-allergic mothers had no history of allergic disease and negative SPT to all tested allergens. Women were considered as ‘primigravid’ if they had no known previous pregnancies (including miscarriages and terminations). For the purpose of this study, ‘multigravid’ women were those with one or more previous pregnancies with their current partner (thereby excluding multiparous women having their first pregnancy with a new partner). This study used peripheral blood samples collected at 36 weeks gestation and cord blood collected from the placental vessels via venipuncture (19G needle) into heparinized (preservative-free) RPMI-1640 tissue culture medium. Mononuclear cells (MNC) from all collections were isolated by Ficoll-Hypaque gradient centrifugation and cryopreserved in 7·5% dimethylsulphoxide (DSMO) using established techniques [10].

MLR

Thawed MNC were employed in MLR to measure maternal responses to fetal alloantigens and vice versa, as described previously [6]. Briefly, maternal ‘responder’ populations (125 µl at a concentration of 0·5 × 10⁶ cells/ml) were added to culture wells in AIM-V (Gibco, Life Technology, Paisley, Scotland, UK) supplemented with 2 mercaptoethanol (ME) (4 × 10⁻⁵ M final concentration; Sigma, Castle Hill, Australia) plus 5% heat-inactivated antibody serum (Sigma-Aldrich Inc., St Louis, MI, USA). CB MNC ‘stimulator’ populations were CD3-depleted using CD-3 Dynabeads (Dynal Biotech ASA, Oslo, Norway, as per the manufacturer's instructions) and irradiated at 3000 rad (Gammacell 3000 Elan; MDS Nordion, Ontario, Canada) to inhibit proliferation by this population. Stimulator cells were then added to culture wells (125 µl at 0·5 × 10⁶ cells/ml) containing responders. The same process was repeated to assess fetal responses to maternal alloantigens. All cell cultures were incubated at 37°C in 5% CO₂ for 72 h for cytokine protein measurement or 5 days for lymphoproliferation.

Cytokines [interleukin (IL)-6, IL-10, IL-13, IFN-γ] in MLR supernatants were quantified using a ‘sandwich-type’ time-resolved fluorometry [6]. Data were expressed as concentration in pg/ml. The detection limit of the assay was 5 pg/ml for all cytokines. Samples with concentrations below this value were considered as a non-detectable response. For analysis of lymphoproliferation, 0·5 µCi of tritiated thymidine was added to MLR culture wells as a measure of DNA synthesis [6]. Disintegrations per minute (dpm) were measured on a beta counter and the median of each triplicate was used as a measure of T cell proliferation. A positive response was defined as a delta dpm > 1000 (dpm of test minus dpm of unstimulated control) plus a stimulation index of ≥ 2 (ratio of test to unstimulated control).

HLA typing

Genomic DNA was extracted from the buffy coats of maternal blood at 36 weeks and cord blood using Qiagen DNeasy kits as instructed by the manufacturer. Concentration and purity was checked by spectroscopy. Extracted DNA from mothers and infants was sent to the Department of Clinical Immunology, Royal Perth Hospital, Perth, Western Australia. HLA-DRβ1 and HLA-C typing was performed by DNA sequencing using protocols established previously at the laboratory [11].

Potential anti-maternal and anti-infant alloreactivity were defined separately. Potential anti-maternal T cell alloreactivity was said to be present if either of the mother's DRB1 alleles was not present in the fetus. According to the ‘missing self’ hypothesis [12,13] for NK cell alloreactivity, potential anti-maternal NK alloreactivity was said to be present if the mother lacked an HLA-C epitope (either C1 or C2) which was present in the fetus. Potential anti-fetal T cell alloreactivity was said to be present if either of the fetal DRB1 alleles were not present in the mother. Potential anti-fetal NK alloreactivity was said to be present if the fetus lacked an HLA-C epitope that was present in the mother.

Statistical analysis

All statistical analysis was performed using the Statistical Package for Social Sciences (SPSS version 13·0 for Macintosh). MLR cytokine responses were analysed as both binary (detected or non-detected response) and continuous variables. Continuous cytokine data were normalized by logarithmic transformation (assessed by Kolmogorov–Smirnov test). Groups were compared with Student's t-test and data expressed as geometric mean and 95% confidence intervals. χ² tests were used for comparisons involving binary variables, including if any HLA-DRβ1 or HLA-C allele were more frequent in mothers with allergic disease, asthma or house dust mite sensitivity compared to non-allergic mothers. Logistic regression was used to determine if potential alloreactivity predicted the detection of maternal responses to the fetus (with the detection of a response as a dichotomous dependent variable). Separate models were run for each cytokine (IL-6, IL-10, IL-13 and IFN-γ) and proliferative response. Similarly, linear regression was used to assess if potential alloreactivity predicted the magnitude of maternal responses to the fetus, where the dependent variable (magnitude of a response) was normalized by logarithmic transformation (assessed by Kolmogorov–Smirnov test) prior to inclusion in the linear model. P-values < 0·05 were considered significant. A correction factor for multiple testing (such as a Bonferroni factor) was not applied due to the exploratory nature of this study [14]; however, appropriate caution was taken interpreting the results.

Results

Of the mothers who were recruited to this study, 127 (65 non-allergic women and 62 allergic women) had sufficient samples for both MLR studies and HLA typing, including equal numbers of primigravid and multigravid women. In the final study population 45% of the allergic group and 43% of the non-allergic group were primigravid. There were no differences in mode of delivery, maternal age or birth measures (gestational age, length, height, head circumference and Apgar score) between allergic and non-allergic groups. More detailed characteristics of this study population have been described previously [6]. All women were HLA-C typed. Two women in the allergic group did not have sufficient samples for HLA-DRβ1 typing. A total of 116 infants had DNA extracted for HLA typing, and only one could not be typed for HLA-DRβ1 due to the limited concentration of DNA extracted. In this population, the DRβ1*1101 allele was associated positively with allergic disease (P = 0·013; 2·4% non-allergic, 9·6% allergic), although this was not significant after correcting for the number of HLA alleles tested. HLA-C was not associated significantly with maternal allergy status (P = 0·180).

Associations between potential alloreaoctivity and maternal responses to the fetus

There were 90 of 109 (82·6%) mothers with a potential anti-fetal T cell alloreactivity and this proportion was similar, regardless of maternal allergy or gravidity. Potential anti-fetal T cell alloreactivity was associated with increased chance of detecting maternal IL-13 [odds ratio (OR) 3·09, 95% confidence interval (CI) 1·09–8·76, P = 0·034], IFN-γ (5·10, 1·55–16·79, P = 0·007) and proliferation (4·14, 1·26–13·67, P = 0·020) in response to the fetus. Similarly, the magnitude of maternal cytokine and proliferative responses to the fetus were higher when potential anti-fetal T cell alloreactivity was present. Specifically, potential anti-fetal T cell alloreactivity was associated with higher maternal IL-13 (P = 0·020), IFN-γ (P = 0·053) and proliferation (P = 0·012), as shown in Fig. 1c–e. Notably, non-allergic women showed greater differences in lymphoproliferation between the HLA-DR matched and mismatched subgroups (P = 0·002) compared with the allergic women (P = 0·079). Furthermore, the differences in IFN-γ production between HLA-DR matched and mismatched women was also more apparent in the non-allergic mothers (P = 0·046) than the non-allergic mothers (P = 0·378), suggesting that allergic women respond differently to HLA mismatch compared with non-allergic women (as investigated further below). There were no differences in the mean maternal IL-6 or IL-10 responses (Fig. 1a, b) between those with and without potential anti-fetal T cell alloreactivity (or when analysed separately according to allergic disease).

Fig. 1.

A comparison of maternal responses to fetal alloantigens between women whose fetus was human leucocyte antigen D-related (HLA-DR) compatible or mismatched: maternal cytokine and lymphoproliferation responses to mixed lymphocyte reaction (MLR) activation (as per the Methods) are shown for mothers whose fetus was either compatible (open bars; n = 29) or mismatched (solid bars; n = 90) for HLA-DRβ1. Levels are displayed as geometric means (GM) and 95% confidence intervals (95% CI) derived for log-transformed data.

An HLA-C mismatch with potential anti-fetal NK alloreactivity was seen in 27 of 119 (22·7%) of pregnancies. Potential anti-fetal NK alloreactivity was present in a significantly lower proportion of pregnancies of allergic mothers compared to non-allergic mothers (P = 0·024). Potential anti-fetal NK alloreactivity was associated with a significantly reduced likelihood of detecting a positive maternal IFN-γ response compared to HLA-C compatible pregnancies (OR 0·26, 95% CI 0·10–0·70, P = 0·008) after adjusting for maternal allergy. Potential anti-fetal NK alloreactivity did not affect the magnitude of IFN-γ responses or any other aspects of maternal responsiveness.

In summary, anti-fetal T cell alloreactivity was associated with higher IL-13, IFN-γ and proliferative responses of mothers to the fetus. In contrast, potential anti-fetal NK alloreactivity was associated with less likelihood of detecting maternal IFN-γ responses to fetal alloantigens. The latter observation is difficult to explain in terms of inhibitory KIR, but could be due possibly to the absence of fetal ligands for maternal activating KIR.

The effect of maternal allergy on alloresponses in HLA mismatched mother–infant pairs

We noted previously that pregnant allergic women were more likely to show reduced production of the Th1 cytokine IFN-γ in response to their fetus compared with non-allergic women [6]. Here, we compared the magnitude of responses of allergic women to non-allergic women in the 90 cases in which the infant was HLA mismatched. We observed that among those pregnancies with potential anti-fetal T cell alloreactivity, allergic women (n = 44) showed significantly lower IFN-γ production in response to their infant (Fig. 2d) and lower IL-6 responses (P < 0·01) (Fig. 2a) compared with non-allergic women (n = 46) (P = 0·01). There were no differences in lymphoproliferation (Fig. 2e), IL-10 (Fig. 2b) or IL-13 Th-2 responses (Fig. 2c) in allergic and non-allergic mothers.

Fig. 2.

A comparison of maternal responses to fetal alloantigens between allergic and non-allergic women whose infant was human leucocyte antigen D-related (HLA-DR)-mismatched: maternal cytokine and lymphoproliferation responses to mixed lymphocyte reaction (MLR) activation are shown for non-allergic (open bars; n = 46) and allergic (solid bars; n = 44) mothers whose infant was mismatched for HLA-DRβ1. Levels are displayed as geometric means (GM) and 95% confidence intervals (95% CI) derived for log-transformed data.

The relative production of Th1 : Th2 cytokines (i.e. IFN-γ/IL-13) was significantly lower in allergic mothers compared with non-allergic mothers, as shown in Fig. 3 (P < 0·001), supporting the hypothesis that maternal allergy is associated with altered patterns of cytokine response in response to the fetus. We also saw the same pattern in HLA-DRβ1 compatible women, with lower IFN-γ/IL-13 in allergic mothers (P = 0·047), although the magnitude of responses were smaller than in mismatched women.

Fig. 3.

Pattern of relative T helper type 1 (Th1)/Th2 cytokine production by allergic and non-allergic mothers in response to human leucocyte antigen D-related B1 (HLA-DRβ1)-mismatched fetal alloantigens: the ratio of maternal interferon (IFN)-γ/interleukin (IL-13) cytokine responses to mixed lymphocyte reaction (MLR) activation are shown for non-allergic (circle; n = 46) and allergic (square; n = 44) mothers whose infant was mismatched for HLA-DRβ1. Levels are displayed as geometric means (GM) and 95% confidence intervals (95% CI) derived for log-transformed data.

Only one allergic mother with potential anti-fetal NK alloreactivity had detectable IL-10 and IFN-γ responses, so statistical differences associated with potential anti-fetal NK alloreactivity between allergic and non-allergic mothers could not be determined. There were no differences in other cytokine or proliferation responses between the allergic and non-allergic mothers with potential anti-fetal NK alloreactivity (data not shown).

In summary, allergic women had lower IFN-γ and IL-6 production in response to a fetal HLA-DRβ1 mismatch (potential anti-fetal T cell alloreactivity). The infants of these women also showed a trend for lower IFN-γ production relative to IL-13. It is more difficult to draw conclusions about the similar trends seen with HLA-C mismatches (potential anti-fetal NK alloreactivity) because of the small numbers of observations.

Relative effects of maternal allergy and alloreactivity on maternal anti-fetal responses

Although women who lacked potential anti-fetal T cell alloreactivity (n = 29) had significantly lower lymphoproliferative responses to the fetus (as noted above; Fig. 1e) we observed in this subgroup that allergic women had increased lymphoproliferative responses compared with the non-allergic women (P = 0·034, data not shown). Among the women lacking potential anti-fetal T cell alloreactivity, none of the allergic women had any detectable IFN-γ responses, whereas six of 13 of the non-allergic mothers had detectable responses (χ²P = 0·015), also suggesting that maternal allergy has independent effects on the capacity to generate IFN-γ responses.

We therefore performed multivariate linear regression to determine the main predictors of maternal responses to the fetus (with the maternal response as a continuous dependent variable). As shown in Table 1, maternal allergy [beta coefficient (β) = −1·00, 95% CI: −1·65 to 0·36, P = 0·003) and potential anti-fetal T cell alloreactivity (β = 1·73, 95% CI: 0·52–2·95, P = 0·006) were significant independent predictors of maternal IFN-γ responses. Potential anti-fetal T cell alloreactivity was the strongest predictor of maternal proliferation (β = 1·84, 95% CI: 0·32–3·36, P = 0·018). In this population neither maternal allergy nor potential anti-fetal T cell alloreactivity predicted maternal IL-6, IL-10 or IL-13 responses.

Table 1.

Independent effects of maternal allergy and fetal human leucocyte antigen (HLA) mismatch on maternal responses to fetal allo-antigens

Maternal response (dependent variable)	Predictor variables	β (95% CI)	P-value
IL-6	Maternal allergy	−0·580 (−1·68 to 0·52)	0·293
	HLA-DRβ1 mismatch	0·233 (−1·22 to 1·68)	0·748
	HLA-C mismatch	−0·693 (−1·90 to 0·52)	0·235
IL-10	Maternal allergy	−0·490 (−1·06 to 0·09)	0·095
	HLA-DRβ1 mismatch	0·111 (−0·70 to 0·92)	0·782
	HLA-C mismatch	−0·165 (−0·85 to 0·52)	0·628
IL-13	Maternal allergy	0·048 (−0·45 to 0·54)	0·847
	HLA-DRβ1 mismatch	0·613 (−0·21 to 1·44)	0·142
	HLA-C mismatch	−0·256 (−0·84–0·33)	0·386
IFN-γ	Maternal allergy	−1·003 (−1·65 to 0·36)	0·003*
	HLA-DRβ1 mismatch	1·735 (0·52 to 2·95)	0·006*
	HLA-C mismatch	0·235 (−0·62 to 1·09)	0·584
Proliferation	Maternal allergy	0·724 (−0·33 to 1·78)	0·175
	HLA-DRβ1 mismatch	1·845 (0·32 to 3·36)	0·018*
	HLA-C mismatch	0·746 (−0·50 to 1·99)	0·237

^*P-values < 0·05 were considered significant.

Multivariate linear regression analysis for maternal responses to fetal alloantigens. CI: confidence interval; IFN: interferon; IL: interleukin.

Effects of previous pregnancies (gravidity) on responses due to potential anti-fetal alloreactivity

We compared the MLR responses of primigravid (n = 46) and multigravid (n = 51) women with potential anti-fetal alloreactivity. In this population there were no significant differences in the responses of primigravid and multigravid mothers with potential anti-fetal T cell or NK alloreactivity after allowing for effects of maternal allergy (data not shown). There was also no effect of gravidity on any aspect of maternal responses in the women lacking potential anti-fetal T cell or NK alloreactivity.

Associations between potential anti-maternal alloreactivity and fetal responses to the mother

Consistent with other reports, infants of allergic women showed differences in cytokine production, although to our knowledge this is the first study to examine alloresponses in this context. Specifically, fetuses of allergic women showed lower IL-10 production (P = 0·03) (Fig. 4a) and a trend for lower IFN-γ (Fig. 4c; P = 0·07) and higher IL-13 production (Fig. 4a; P = 0·15) in response to the mother.

Fig. 4.

Fetal responses to maternal alloantigens are different for fetuses of allergic and non-allergic women: fetal cytokine and lymphoproliferation responses to mixed lymphocyte reaction (MLR) activation are shown for those of non-allergic (open bars; n = 43) and allergic (solid bars; n = 43) mothers. Levels are displayed as geometric means (GM) and 95% confidence intervals (95% CI) derived for log-transformed data.

A further aim of this study was to examine the relationship between potential anti-maternal alloreactivity and the pattern and magnitude of fetal responses to the mother. Of 110 infants assessed, n = 86 (78·2%) had potential anti-maternal T cell alloreactivity, and this proportion was similar regardless of maternal allergy. Potential anti-maternal T cell alloreactivity was associated significantly with higher odds of detecting fetal IL-13 (OR = 2·79, 95% CI: 1·08–7·19, P = 0·033) and IFN-γ (β OR = 4·09, 95% CI: 1·39–12·02, P = 0·010) production in response to the mother. There was also a tendency towards higher odds of detecting fetal proliferation (OR = 3·42, 95% CI: 0·86–13·62, P = 0·081) but not fetal IL-6 or IL-10 production. The magnitude of fetal proliferation in response to the mother was significantly higher when potential anti-maternal T cell alloreactivity was present (P = 0·033) (Fig. 5), but there were no differences in magnitude of any cytokine responses (data not shown). There were 21 of 125 (16·8%) infants with potential anti-maternal NK alloreactivity. Potential anti-maternal NK alloreactivity was not associated with detection or the magnitude of any fetal responses to the mother, even after adjusting for maternal allergy, delivery method, infant gender and other confounding factors (data not shown). There was a significant correlation between the relative pattern of maternal Th1/Th2 responses (i.e. IFN-γ/IL-13) and corresponding fetal responses (r = 0·23, P = 0·03). This remained statistically significant after allowing for maternal allergy and the presence of potential anti-maternal alloreactivity.

Fig. 5.

Fetal lymphoproliferation responses to maternal alloantigens are influenced by maternal human leucocyte antigen D-related (HLA-DR) matching: fetal lymphoproliferation responses to mixed lymphocyte reaction (MLR) activation are shown for infants' mothers who were either compatible (open bars; n = 24) or mismatched (solid bars; n = 86) for HLA-DRβ1. Levels are displayed as geometric means (GM) and 95% confidence intervals (95% CI) derived for log-transformed data.

In summary, potential anti-maternal T cell alloreactivity was associated with higher fetal proliferative responses to maternal alloantigens with increased detection of fetal IL-13 and IFN-γ, but there were no associations with potential anti-maternal NK alloreactivity.

Discussion

The study confirms that HLA Class II mismatch is a significant factor influencing materno–fetal interactions. Potential T cell alloreactivity (DRB1 mismatch) was associated with higher IFN-γ, IL-13 production and proliferation by both mothers and neonates. This is consistent with data on MLR responses between unrelated individuals [15], and suggests that allogeneic responses, including Th1 reactivity, are not suppressed completely in pregnancy. The novel finding of this study was that allergic women, who are prone to Th2 skewed responses, have significantly lower IFN-γ responses to fetal HLA-DRβ1 mismatch during pregnancy, and lower relative Th1/Th2 production compared with non-allergic women. Furthermore, patterns of maternal cytokine responses (i.e. production of Th1 cytokines relative to Th2 cytokines) were correlated significantly with fetal responses, suggesting that patterns of maternal reactivity can influence fetal immune development.

We also found that HLA-DRβ1*1101 was more frequent in allergic than non-allergic mothers. Others have associated this allele with asthma and pollen sensitization in European cohorts [16,17]. The HLA genes have shown consistent linkage to asthma and atopy-associated phenotypes in several studies (reviewed in [18]) and may be a major locus influencing allergic diseases. In vitro studies suggest that individual HLA-DR alleles restrict the ability of T cells to respond to particular allergens [19]. We speculate that HLA alleles could also influence infant allergic predisposition by modulating maternal effects on fetal immune programming in utero. Unfortunately, the sample size was too small to explore this idea further in this population.

HLA-DRβ1 and HLA-C mismatches consistent with potential anti-fetal alloreactivity had opposing influences on maternal IFN-γ responses to fetal antigen. As indicated above, HLA-DRβ1 mismatch correlated with higher maternal IFN-γ response, while there was a lower odds of detecting maternal IFN-γ in the HLA-C epitope mismatched pairs with potential anti-fetal NK alloreactivity. Although others have reported suppression of IFN-γ production by peripheral NK cells during pregnancy [20], it is not clear why HLA-C mismatched stimulators, which have fewer ligands for inhibitory KIR, should inhibit IFN-γ secretion. However, while uterine NK cells in direct contact with HLA-C on fetal trophoblasts are abundant at the maternal–fetal interface [21], NK cells represent only 5–20% of lymphocytes in adult peripheral blood and may be recruited from the periphery to the decidua during pregnancy [22]. Therefore, the contribution of NK–HLA-C interactions on IFN-γ production in maternal peripheral blood may be very small. This could explain why HLA-C epitope mismatch was not a significant predictor for maternal IFN-γ production after adjusting for maternal allergy and HLA-DRβ1 mismatch.

Although HLA typing was limited to HLA-DRβ1 and HLA-C it is recognized that other contributory factors can affect these Th1 allogeneic responses during pregnancy, including mismatch of other Class I and Class II HLA and minor histocompatibility antigens [23,24]. Other gene polymorphisms could also be important, as the inheritance of certain genes can influence cytokine responses to infectious agents [25,26], and these genes may also regulate allogeneic responses.

In our study population previous pregnancies did not result in any significant differences in the patterns or magnitude of cytokine responses in women with HLA mismatch. Although this is in apparent contradiction to (albeit limited) evidence that the immune status of women changes after multiple births [27,28], this study may not have been powered appropriately to address this specific issue.

As maternal allergy is a stronger risk factor than paternal allergy, it is possible that rising rates of maternal allergy could be having compounding effects in the ongoing epidemic rise infant allergic diseases such as food allergy and eczema. Differences in neonatal immune function of children who go on to develop allergic disease suggest that factors driving this rise in disease are likely to begin to exert their effects in utero (reviewed in [29]). While these changes are environmental (and cannot be explained by genetic changes in this time-frame), this may be the result of effects in both the exogenous and the endogenous environment in pregnancy. Our data may support speculation that maternal immune propensity could influence the endogenous environmental provided for the fetus. It is already clear that the ‘Th2 skewed’ pattern of neonatal immune response is influenced heavily by the maternal Th2 immune modification of pregnancy [30]. Here we propose that variations in maternal of Th1/Th2 modulation that are conferred by her allergic propensity may influence the fetus similarly. There is emerging evidence of how in utero exposures can modify fetal gene expression through heritable epigenetic effects, which can be amplified across generations (also reviewed in [29]). Thus, it is important to consider the possibility that ‘allergy may beget more allergy’ with subsequent generations. While we acknowledge that this is speculative, it is notable that the more recent epidemic of infant allergy (food allergy and eczema) has lagged 20–30 years behind the original epidemic of inhalant allergies (asthma and allergic rhinitis), which could suggest a ‘second-generation’ effect.

In conclusion, this study indicates that HLA mismatch has an influence on the immune interactions between mother and fetus. We have also shown that maternal allergy modifies these interactions, and reduces Th1 activation to fetal alloantigens. This may affect the cytokine milieu at the materno–fetal interface and could be implicated in the attenuated Th1 responses observed commonly in infants of atopic mothers [1–4]. It is possible that this could be a mechanism by which maternal allergy could have direct effects on the fetus and compound or amplify the environmental pressures promoting allergy with successive generations.

Disclosure

Professor Prescott has been a speaker at meetings sponsored by SHS/Nutricia and Nestlé. She has been a member of the independent Scientific Advisory Board of Nestlé Nutrition Institute Oceania, an expert panel on Cows Milk Allergy for Nutricia Australia, and an expert panel on Omega-3 fatty acids for Mead Johnson. She has received travel assistance and speaker fees from these companies to present at or attend scientific meetings. None of these relationships is relevant to the present study.