Evidence is accumulating of close connections between in utero exposure to toxicants and the development of the immune system, the impairment of which is involved in the development of various health conditions 1 . In recent years, there have been a growing number of arguments showing that susceptibility to disease is largely set in utero or early in infancy 2 . Asthma 3 , elevated immunoglobulin E (IgE) 4 and mortality in children 5 and adults 4 6 may be determined in utero and early life. Several population-based studies have associated exposure to urban air pollution during pregnancy with birth outcomes, including impaired fetal growth 7 , low birth weight and premature birth 8 9 10 , postneonatal infant mortality, intrauterine growth restriction and a decreased lung function 11 as well as allergies 12 and asthma 13 in childhood. A link has also been shown between maternal exposure to environmental tobacco smoke and a low birth weight 14 , decreased lung function as well as allergies and asthma in childhood 15 16 .

Furthermore, thymus functions are altered by exposure during pregnancy to environmental immunotoxicants, such as 2, 3, 7, 8-tetrachlorodibenzop-dioxin (TCDD) and polychlorobiphenyls (PCB) in animals 17 and humans 18 . Few studies have related maternal exposure to toxicants during pregnancy and cord blood cells involved in organism defenses. Significant changes in immunological variables in cord blood of babies from smoking mothers compared to those of non-smoking mothers have been shown 15 19 . Similarly, living in a highly polluted urban area was related to a lower percentage of CD4+ T-lymphocytes and a lower CD4+/CD8+ ratio but a higher percentage of NK cells 20 in the cord blood among newborns from the Czech Republic. More recently, maternal exposure to polycyclic aromatic hydrocarbons (PAHs) and fine particulate matter less than 2.5 μm in diameter (PM_2.5) assessed during the 14 days before birth in two districts in the Czech Republic was associated with lower percentages of CD3+, CD4+ and CD8+ T-lymphocytes and a higher percentage CD19+ B-lymphocytes in newborns 21 . However, such investigations have been limited by ecological approaches, exposure misclassification, limited sample size and failure to take potential confounders of the relationship into account.

The general objective of our work was to investigate whether maternal exposure to air pollution before and during pregnancy intervenes in the development of the newborn's immune cells. Our study was conducted in the frame of the French EDEN mother-child cohort.

This study indicates that ambient air pollution to which the mother is exposed before and during pregnancy might alter the relative distribution of cord blood lymphocyte phenotypes in her newborn, thus possibly influencing the future child's health. Fetal blood CD8+ and NK cells increased whereas CD3+, CD4+CD25+ and CD4+/CD8+ ratio decreased as maternal exposure to PM₁₀ and NO₂ the three-month period before and during pregnancy increased. Similarly, CD4+CD25+ T-cells decreased while maternal exposure to benzene during the second trimester of pregnancy increased, as assessed using a personal diffusive air sampler.

Our findings extend previous data from the rare studies having investigated the links between prenatal exposure to air pollution and the immune system. In the first study carried out in the Czech Republic, living in polluted areas was found to be associated with a decrease in CD3+ and CD4+ T-cells and in the CD4+/CD8+ ratio, as well as with a higher proportion of NK cells in fetal blood 20 . Our study also found an inverse relationship between PM₁₀ exposure and the CD4+/CD8+ ratio, although only at the borderline level (p = 0.08). Successively, the same group showed short-term associations between PAHs and PM_2.5 and immune parameters in Czech newborns 21 . Average PAH or PM_2.5 levels during the 14 days before birth were associated with significant decrements in the T-lymphocyte phenotype fractions (i.e. CD3+, CD4+ and CD8+) and a clear increase in the B-lymphocyte (CD19+) fraction in cord blood. In spite of the similarities in the results, Hertz-Picciotto et al. studies 20 and ours differ in the concentrations to which the mothers were exposed that were very high in the Czech area and comparable to current World Health Organization standards in our study and in the estimation of maternal exposure that was both geo-localized and assessed with personal samplers in our investigation. More recently, in agreement with our observations, children exposed to elevated air pollution in Mexico City exhibited significantly increased numbers of CD8+ cells (p = 0.02) 28 . However, these children also exhibited significant decreases in the numbers of NK cells (p = 0.003), which contrasts with our results. However, in the Mexican investigation the children were seen at 10 years of age and no cord blood was available. The Mexican study thus targeted a window of exposure that was different from the one considered in our study. In addition, a recent study carried out in children showed that increased exposure to ambient air pollution was associated with the impairment of Treg cells function and may increase asthma morbidity 29 .

No previous study has related regulatory T-cells in early life to exposure to air pollution in spite of the fact that they have been involved in the development of allergies 30 . We targeted CD4+CD25+ T-cells as physiologically generated CD4+CD25+ T-cells, which are the most widely studied type of regulatory T-cells (Tregs), especially in newborns where most cord blood CD4+CD25+ T-cells correspond to regulatory T-cells 31 . In our birth cohort study, decreased CD4+CD25+ T-cell levels were significantly correlated to PM₁₀, benzene and moderately to NO₂ indicating that by decreasing Treg cells, pollutant exposure might increase the risk of allergy in newborns. Indeed, regulatory T-cells can inhibit the development of allergic Th2 responses and could thus play an important role in the development of allergy 32 . As further confirmation, it has recently been shown that Treg cells are defective in allergic subjects 32 33 including asthmatics 30 34 35 , and impaired regulatory T-cells were found in cord blood of newborns at hereditary risk of allergy 35 36 . This situation is consistent with the involvement of air pollution in the so-called "allergy epidemics" observed in past decades as traffic-related pollution increased. The fact that the relationship was stronger for particulate matter further supports this hypothesis on the basis of experimental evidence having directly related particulate matter to the development of allergy 37 . The observed CD4+CD25+ T-cells deficiency could result from the effect of exposure to air pollution on the Treg compartment or as a consequence of pro-inflammatory mechanisms driven by pollutant exposure and leading to an increase in CD8+ T-cells and NK cells. Latent viral infections could also be involved in this process 38 .

Besides, some studies have suggested the probable contribution to asthma pathogenesis of CD8+ T-cells producing IL4 and IL5 (type 2 CD8+ T-cells) in the lungs and blood of asthmatic individuals 39 . CD8+ T-cells producing IFN-γ (type 1 CD8+ T-cells) may also contribute to asthma symptoms and to asthma exacerbation in individuals suffering already from asthma 40 , and this in spite of the fact that in early life they may protect against asthma development, by eliminating allergen-specific Th2 cells. In addition, virus-specific CD8+ T-cells producing type 2 cytokines may develop during certain viral infections, suggesting a mechanism for virus-induced asthma exacerbation 41 42 . Recently, Natural Killer T (NKT) cells comprising a subpopulation of lymphocytes that express features of NK cells and conventional T-cells have been suggested to play an important pathogenic role in asthma 43 . A role for subsets of NKT cells in asthma has been suggested by extensive studies in animal models of asthma induced with allergen, viral infection, ozone exposure, or bacterial components, suggesting that NKT cells function in causing airway hyperreactivity. The clinical relevance of NKT cells in human asthma is supported by the observation that NKT cells are present in the lungs of some patients with asthma, particularly patients with severe, poorly controlled asthma. However, the validation of the concept that NKT cells may contribute to asthma still requires additional human studies. Thereafter our findings in a sample of mother-child pairs are consistent with previous experimental studies.

Regarding maternal exposure to organic volatile compounds (VOCs) that was objectively assessed in our study, our results are consistent with the findings of one study that showed associations between maternal exposure to VOCs and an altered cytokine secretion profile of cord blood T-cells 44 . In particular, the authors found a diminished capacity of cord blood T-cells to produce the Th1 cytokine IFNγ in infants from exposed mothers, which is a risk factor for the development of allergic disease in infancy 45 46 .

An originality of our work was the systematic assessment, along with the whole CD4+ compartment, of CD4+CD25+ T-cells that have been described as a "proxy" of regulatory T-cells (Tregs) 47 . However, we have to be cautious in interpreting our findings because we did not have at the time of the survey at our disposal the assessment of forkhead/winged-helix transcription factor FoxP3 48 49 which would have been more accurate in order to identify regulatory T-cells. In addition, we did not estimate functional immune parameters that were deemed not feasible in our project because the quality of these measurements degrades with storage time and transportation. Another strength of our study is that we geo-localized the child's address, using Geographic Information System (GIS) technology, in order to attribute to each child representative values of background air pollution 7 and thus reduce exposure misclassification 50 . In addition, for the assessments of exposure to benzene, the mothers carried personal samplers, that allow taking into account both indoor and outdoor exposures to benzene including microenvironments such as vehicle cabins 51 . An additional gist of our investigation consists in the robustness of the results as children were drawn from the general population and their parents could provide information on possible confounders, thus diminishing the probability of having obtained random results. However, considering background air pollution assessment, although performed by monitoring stations located in close proximity of the mothers' residence, could have led to underestimation of exposure in our investigation. In addition, some cells were privileged compared to others, which opens the way to other studies. Although the distributions of some variables were significantly different in the study sample (n = 370) compared with the full cohort (e.g., the proportion of low birth weight newborns was 2.43% in the study total sample and 10.61% in the full cohort), these differences are not likely to have affected either the internal or external validity of the results, because the variables were controlled in the analysis through adjustment in the multivariate models. Lastly, as an additional positive element, we avoided the multiple testing problem by using a second-order linear orthogonal polynomial regression that yielded similar results, strongly significant in the case of PM₁₀ exposure and CD4+CD25+ T-cells. This had never been done previously for the period of pregnancy.

The findings on the effects of maternal exposure to air pollution during pregnancy on lymphocyte immunophenotype distributions in cord blood, even at low air pollutants levels, suggest that the fetus might be highly sensitive to maternal exposures and its variations during pregnancy. Moreover, it is frequently surmised that prenatal and early life exposures to air pollution are of greater significance than later exposure due to the susceptibility of target organs and systems during developmental periods of life 52 . By assessing background pollutant exposure levels three months before and during each trimester of pregnancy and then by testing associations between corresponding exposure variables and lymphocyte immunophenotype distributions in cord blood, our study contributes to identify windows of exposure to air pollutants before and during pregnancy. So far only two studies have indicated that in utero exposure to urban air pollution is associated with immunodysregulation in newborns 20 21 , and in these studies the levels of air pollution were higher than the levels encountered nowadays in most industrialized countries.

Our investigation is the first to consider the three-month period immediately preceding pregnancy, which involves the participation of epigenetic mechanisms in the development of the immune system. Recent epidemiological studies have provided evidence suggesting that environmental factors may cause epigenetic alterations affecting both the mother and her child 53 . Heritable epigenetic changes may disrupt the normal functioning of genes by affecting their expression without changes in the DNA sequence. Several epigenetic mechanisms have been suggested 54 , including DNA methylation and histone modifications, which could determine or give a clear prediction of the future transcriptional state of the genome under the influence of environmental factors. Animal 55 and human 56 57 studies have indicated that in utero and early life environmental exposures may generate effects than can be inherited transgenerationally along with epigenetic modifications. An inverse association has been identified between prenatal lead exposure and genomic leukocyte DNA methylation (linked to transcriptional repression) in human cord blood 56 . Similarly, transplacental exposure to high levels of airborne PAHs has been related to aberrant DNA methylation changes, leading to dysregulation of gene expression and possibly childhood asthma 57 . Overall, these results suggest that maternal exposure to air pollutants during critical periods of prenatal development might lead to inappropriate gene expression and disease pathogenesis in later life by inducing epigenetic alterations in the fetus 58 .

Several potential mechanisms may explain the links between in utero exposure and subsequent changes in cord blood cells. Several experimental studies have supported the hypothesis that antigenic cell stimulation by air pollutants and their metabolic by-products could be at the origin of the rise in some lymphocyte subsets by highlighting the disturbance of lymphocyte immunophenotype distributions by xenobiotics including tobacco smoke 59 and chemicals 60 . However, air pollutants could also act as adjuvants by amplifying specific responses towards other antigens such as viruses or allergens. Indeed it has been shown that diesel exhaust particles can amplify an allergen-specific Th2 response 61 but also enhance Th2 and Th1 cytokine production of primed T-cells recovered from asthmatics during exacerbation 62 . Such adjuvant effects could be involved in the increase of CD8+ and NK cells. Also, the depressed maternal immune system could allow for latent viral infections, potentially transmitted to the fetuses. This effect would indeed trigger an increase of NK cells and CD8+ T-cells, and hence a decreased CD4/CD8 ratio amplified by pollutants 38 . Lastly, animal experimental studies, investigating impacts of maternal exposure to diverse chemicals on the neonatal immune system 63 , have indicated that the thymus, through its cellularity and functioning, may also play a role. Maternal exposure to certain chemicals during pregnancy has been associated with a reduction in thymic cellularity, thymic atrophy and an altered T-lymphocyte differentiation. During fetal immune system development, the thymus is the organ responsible for T-lymphocyte maturation. Lymphoid colonization starts from the tenth week of gestation and after twelve weeks, mature CD3+ T-lymphocytes are detectable in fetal blood. Therefore, a functional disturbance of the thymus may have serious consequences on lymphocyte immunophenotypes. Underneath these mechanisms there is the assumption that the fine and ultrafine particles can cross the alveolocapillary membrane into the systemic circulation 64 and also the placenta 65 .

Our data indicate that air pollution before and during pregnancy can be at the origin of disturbances in the functional balance between immune response mechanisms in the newborn. Although the biologic relevance of this finding is not entirely clear, the main observation is that for the first time at the general population level, exposure to major ambient air pollutants before and during pregnancy has been associated with a statistically significant alteration of the relative distribution of cord blood NK and T-lymphocyte phenotypes including a decrease in CD4+CD25+ regulatory T-cells. Whether alterations in developmental patterns of lymphocytes due to air pollution contribute to explain the rise in asthma and allergies - the so-called "allergy epidemics" observed in past decades - needs to be further investigated. Follow-up of the children in the EDEN cohort might shed light on this issue and is currently under way.

VOCs: volatile organic compounds; BETX: benzene: ethylbenzene: toluene: xylenes; CI: confidence interval; SD: standard deviation; NK: natural killer cells; CD19+: B-lymphocytes; BMI: body mass index; bf: before.

The authors declare that they have no competing interests.

NB and IAM conceived and planned the study. NB performed statistical analyses and data interpretation and wrote the manuscript. RS provided the air pollution data. MCB and AM contributed to the immunophenotyping and provided data on the immunological material. MAC was responsible for the human resource organization and contributed to the design of questionnaires. MNKS participated in the cord blood analyses and immunophenotyping. OT participated in the following up of the women during pregnancy and after delivery. GF supervised the group performing immunophenotyping. IAM supervised the project and assisted with writing the manuscript. All authors discussed the results and implications and commented on the manuscript. All authors read and approved the final manuscript.