Autism spectrum disorder (ASD) is defined by impairments in social interaction and communication and restricted and repetitive patterns of behavior [American Psychiatric Association, 2013]. Typically, symptoms manifest by early childhood and persist throughout an individual’s lifetime [Volkmar, Chawarska, & Klin, 2005]. While the genetic contribution to ASD etiology is well documented in a subset of individuals [Sandin et al., 2014; Gaugler et al., 2014] a growing body of evidence is revealing that nongenetic factors also play a critical role, especially during gestation and the early postnatal period [Hallmayer et al., 2011; Lyall et al., 2017].

There is strong evidence from animal and human studies that activation of the immune system, influenced by both genetic and nongenetic components, is involved in ASD etiology [Meltzer & Van de Water, 2017; Onore, Careaga, & Ashwood, 2012]. Numerous studies have found immune system dysregulation in individuals with ASD and among their family members, including disruptions in levels of immune system molecules and increased rates of clinical immune-mediated conditions. Several studies have provided evidence that disruption of normal levels of immune molecules (e.g., cytokines, chemokines, immunoglobulins, C-reactive protein) during gestation [Goines et al., 2011; Jones et al., 2017; Grether et al., 2016; Zerbo et al., 2016; Brown et al., 2014; Koks et al., 2016; Mahic et al., 2017], at birth [Zerbo et al., 2014; Grether et al., 2016; Grether, Croen, Anderson, Nelson, & Yolken, 2010; Abdallah et al., 2013] or in early childhood [Ashwood et al., 2011b, 2011c; Careaga et al., 2017; Ashwood et al., 2011a] is associated with ASD. Furthermore, associations between maternal and child autoantibodies to fetal brain proteins and risk of ASD has been reported across multiple study populations [Edmiston, Ashwood, & Van de Water, 2017].

Frequently, studies have found that children with ASD have a higher prevalence of allergies, atopic dermatitis, asthma, and autoimmune diseases including psoriasis [Mostafa, Hamza, & El-Shahawi, 2008; Magalhaes et al., 2009; Chaidez, Hansen, & Hertz-Picciotto, 2014; Zerbo et al., 2015; Jyonouchi, Geng, Cushing-Ruby, & Quraishi, 2008] than typically developing children, but the specific conditions and magnitude of the effect have varied across studies. Moreover, many studies have reported associations between increased ASD risk and immune-mediated conditions in the mother [Brown et al., 2015; Andersen, Laurberg, Wu, & Olsen, 2014; Atladottir et al., 2009; Lyall, Ashwood, Van de Water, & Hertz-Picciotto, 2014; Lyall, Pauls, Spiegelman, Ascherio, & Santangelo, 2012; Keil et al., 2010; Croen, Grether, Yoshida, Odouli, & Van de Water, 2005; Comi, Zimmerman, Frye, Law, & Peeden, 1999; Mouridsen, Rich, Isager, & Nedergaard, 2007; Sweeten, Bowyer, Posey, Halberstadt, & McDougle, 2003]. There is, however, considerable variability in the specific conditions found to be associated with ASD risk, and only a handful of studies had information on presence of these conditions during the pregnancy period [Chen et al. 2016]. Finally, some studies have investigated ASD risk in association with history of autoimmune conditions in fathers [Andersen et al., 2014; Atladottir et al., 2009; Keil et al., 2010; Mouridsen et al., 2007] or any family member [Atladottir et al., 2009; Comi et al., 1999; Mostafa & Shehab, 2010; Valicenti-McDermott et al., 2006] again with varying results [Wu et al., 2015].

The goal of this paper was to further examine the role of family history of autoimmune diseases, asthma, and allergies in ASD as well as other developmental disorders (DD) in a large, geographically and demographically diverse US population. Using data from the Study to Explore Early Development (SEED) multi-site case-control study, we investigated maternal immune conditions during the pregnancy period, as well as lifetime history of these conditions in several family members (mother, father, siblings, and study child). The comprehensive and robust data collection in SEED allowed for more refined exposure assignments based on both diagnoses and treatment for condition, and more in-depth analysis of ASD phenotypes, features that were lacking in previous studies. Findings will contribute to a better understanding of the etiology of ASD and may also aid earlier identification, intervention and subsequent prevention of these conditions.

The distribution of demographic covariates is shown for each study group in Table 1. As expected, the proportion of males was significantly higher in the ASD group compared to the DD and POP groups. Mothers of ASD cases were significantly more likely to be non-White, and to have a lower educational attainment at time of birth of index child and lower current household income compared with mothers of POP controls. Most ASD cases were classified as mild/moderate, simplex, having ID, and not experiencing developmental regression (Table 2).

In this large and diverse study population with robust exposure and outcome assessment, the maternal immune conditions assessed during pregnancy were relatively common, occurring in 17%–50% of all women. Pregnancy and lifetime maternal eczema/psoriasis and asthma were associated with a 20%–40% increased odds of both ASD and DD. There was some indication that risk estimates varied by specific ASD phenotypes in association with these exposures. In addition, children with ASD were more likely to have a history of eczema/ psoriasis or allergies than POP controls.

In this study we found an association with maternal autoimmune conditions and ASD risk which is in line with previous literature [Andersen et al., 2014; Atladottir et al., 2009; Brown et al., 2015; Lyall et al., 2014; Lyall et al., 2012; Keil et al., 2010; Croen et al., 2005; Comi et al., 1999; Mouridsen et al., 2007; Sweeten et al., 2003]. Several specific autoimmune diseases have been reported to be associated with elevated risk of ASD in past studies (e.g., rheumatoid arthritis, thyroid disease, inflammatory bowel disease, systemic lupus erythematosus); however, these conditions occurred very infrequently in our study population, precluding further analyses. Women with eczema/psoriasis diagnosed by delivery, the most commonly reported autoimmune disease during pregnancy in this study, were significantly more likely to have children with ASD, consistent with some [Croen et al., 2005] but not other [Mouridsen et al., 2007; Keil et al., 2010] previous studies. Although not yet defined as a specific autoimmune condition, several studies have reported associations between ASD and the presence of maternal autoantibodies, where an as yet unknown auto-inflammatory process in the mother led to the increased presence of antibodies reactive to fetal brain proteins.

Asthma is one of the most common chronic diseases among pregnant women, [Meakin, Saif, Jones, Aviles, & Clifton, 2017] with up to 45% of pregnant women seeking medical help and at least 6% being hospitalized [Murphy 2015]. Our finding of an association between maternal asthma during pregnancy and ASD and DD is consistent with previous studies showing links between maternal asthma and risk for neurodevelopmental disorders including ASD, intellectual disability and attention-deficit/hyperactivity disorder (ADHD) [Croen et al., 2005; Lyall et al., 2014; Theoharides, Tsilioni, Patel, & Doyle, 2016; May-Benson, Koomar, & Teasdale, 2009; Langridge et al. 2013; Leonard, de Klerk, Bourke, & Bower, 2006]. Furthermore, a recent mouse model showed that induction of maternal asthma during pregnancy led to offspring with ASD-like behaviors including deficits in social interactions and repetitive behaviors as well as significantly longer body length and higher body weight than controls throughout neonatal development [Schwartzer et al. 2017; Schwartzer, Careaga, Chang, Onore, & Ashwood, 2015].

The significant association we observed with asthma among women treated for asthma could indicate that more severe asthma increases ASD risk or that asthma must be active during the pregnancy period to impact neurodevelopment. During mid-gestation, cytokines associated with allergic responses (e.g., interleukin-4) have been shown to be elevated in maternal blood of mothers whose child was diagnosed with ASD [Goines et al., 2011; Jones et al., 2017], suggesting that an active asthma process in pregnancy may be responsible. Interestingly, elevated levels of IL-4 in amniotic fluid and newborn blood spots have also been associated with ASD risk [Krakowiak et al., 2017], again suggesting that a prolonged or sustained immune response is associated with ASD risk. Although treatment for asthma in the mothers during pregnancy could reflect severity of or persistent maternal asthma, it is possible that the treatment per se negatively impacted neurodevelopment. Recent studies of B2AR agonists, medications commonly used to treat asthma, have identified increased risk of ASD with exposure during pregnancy [Croen et al., 2011; Gidaya et al., 2016]. Where we had medication history we saw no difference in the frequency of reported B2AR use among mothers of ASD and POP children. Further studies are necessary to disentangle the effects of treatment and immunological effects of asthma and risk for ASD.

The observation of no association between ASD and maternal allergy during pregnancy contrasts with an earlier study conducted in northern California that found a significantly elevated risk of ASD among women with an allergy diagnosis recorded in prenatal medical records [Croen et al., 2005]. In that study, the prevalence of maternal allergy was 25%, roughly half the rate reported in the present study. The high prevalence of allergy reported in this study suggests that a more heterogeneous group of allergic conditions was included than in the previous report, with potentially different mechanisms of action (e.g., IgE mediated and non-IgE mediated allergies). Therefore, the null findings should be interpreted with caution, and future studies that delineate the specific types of allergies are warranted.

Our results suggest that different maternal immune conditions during pregnancy may be associated with different ASD phenotypes. We found that maternal eczema/psoriasis was associated with increased risk of ASD among children without ID, children from multiplex families, and children with mild/moderate ASD severity. In contrast, maternal asthma was associated with increased risk of ASD among children with ID and children who experienced regression. While ours is the first study to examine ASD phenotypes in relation to these maternal conditions, a strong inflammatory profile marked by increased levels of cytokines associated with asthma has been previously reported for mothers of children with ASD+ID [Jones et al., 2017], consistent with our results. The exact role of specific cytokines or immune mediators in the development of specific behavioral or comorbid phenotypes in unclear. Future studies with sufficient sample size could further explore phenotypic variability in relation to these exposures and to specific immune responses elicited.

Our finding that children with ASD were more likely to have a history of autoimmune diseases or allergies than POP controls is consistent with findings from previous studies [Gurney, McPheeters, & Davis, 2006; Magalhaes et al., 2009; Bakkaloglu et al. 2008; Jyonouchi 2010; Angelidou et al. 2011; Zerbo et al., 2015; Chang et al. 2013; Tsai, Chang, Mou, Sung, & Lue, 2013; Yaghmaie, Koudelka, & Simpson, 2013; Chen et al. 2014; Billeci et al. 2015; Mostafa et al., 2008]. ASD-relevant behaviors have also been observed in models of early life exposure to food allergies [de Theije et al. 2014]. Immune activation in cells taken from children with ASD also exhibit an activated immune profile, one which was associated with worse behavioral symptoms [Careaga et al., 2017; Onore et al., 2012]. Immune conditions may be considered co-morbid features of ASD and may reflect potential treatment targets that could alleviate or reduce behavioral impairments. Interestingly, no significant associations were shown with immune conditions in children and DD status. This is in line with previous studies that suggest immune profiles in ASD are unique when compared to DD [Onore et al., 2012; Ashwood et al. 2008]. We also did not observe any association between paternal history of autoimmune diseases, asthma, or allergies and ASD or DD, contrary to findings from a recent meta-analysis of four studies that reported a significant 27% increase in risk of ASD associated with paternal history of autoimmune disease [Wu et al., 2015]. The inconsistency in findings across studies could be due to differences in how paternal history of these conditions were ascertained and age of the fathers.

Since different maternal immune conditions were associated with risk of ASD and DD, these data suggest a potential biological framework whereby increased immune activation during gestation rather than a specific disease process drives neurodevelopmental changes. This notion is supported by evidence from animal models that demonstrate associations between increased maternal immune activation during pregnancy and behavioral and brain abnormalities in the offspring [Bauman et al. 2014; Shi, Fatemi, Sidwell, & Patterson, 2003; Shi, 2009; Giulivi, Napoli, Schwartzer, Careaga, & Ashwood, 2013; Onore, Schwartzer, Careaga, Berman, & Ashwood, 2014; Rose et al. 2017; Schwartzer, Careaga, Onore, Rushakoff, & Berman, 2013].

Several studies have shown genetic links within the major histocompatibility complex (MHC) region and neuropsychiatric disorders such as schizophrenia, autism and bipolar disorder [Torres et al. 2016; Mokhtari & Lachman, 2016] The MHC region encodes both immune molecules such as human leukocyte antigen (HLA), complement proteins and cytokines, as well as many non-immune molecules. In addition, studies have linked various HLA molecules with autoimmune diseases and asthma [Kontakioti, Domvri, Papakosta, & Daniilidis, 2014; Prinz 2017]. It is possible that a genetic association with asthma and autoimmune diseases may also explain some of the risk for autism. An investigation studying the connections between HLA genetic risk factors in the MHC region with autoimmunity/asthma and autism in the same subjects would be warranted. As all immune responses are governed by genetic mechanisms the role of immune genes in neurodevelopmental risk in the context of maternal immune activation needs further investigation.

Pregnancy is a time when epigenetic changes help the genome adapt to the maternal environment. Activation of the maternal immune system may alter the regulation of gene expression in the developing fetal brain or immune system. For example, perinatal exposure to maternal asthma has been shown to alter DNA methylation of immune-related genes in human infants, suggesting that maternal asthma has long-lasting effects on the offspring’s immune function that may make them more susceptible to developing further allergies [Gunawardhana et al. 2014]. In nonhuman primates, immune activation during pregnancy led to offspring with altered behaviors and immune responses [Rose et al., 2017]. In this study, immune conditions in the mother were linked to increased risk of ASD in the child. Moreover, we observed increased risk of immune conditions in children with ASD. The exact epigenetic changes as a consequence of early life immune activation requires further analysis.

In this large and diverse study population drawn from several geographic areas across the United States, all ASD and DD status was validated by a comprehensive in-person developmental assessment protocol using gold-standard diagnostic instruments. Our case-control design, including a DD group, allowed us to look at the specificity of our findings to ASD. The comprehensive data collection battery allowed us to conduct analyses by specific ASD subgroups. The immune conditions were ascertained from several different sources, including maternal interview, self-reported questionnaires, and medical records, likely resulting in more complete ascertainment than previous studies that relied on only one source. Finally, we controlled for several potential confounders.

Despite these strengths, several study limitations deserve mention. While this is a large study population, the prevalence of most autoimmune diseases is quite low among individuals of reproductive age, resulting in very small numbers of children exposed to specific conditions. Thus, we were not able to conduct adjusted or stratified analyses of several autoimmune conditions that have been reported to increase risk of ASD in previous studies. We had no information on the status (active, flaring-up, or dormant) of the autoimmune conditions and asthma during the pregnancy period, and no indication of severity, other than whether women were being treated with medication for the disorder during pregnancy. Information on specific type of medications taken to treat asthma was not available for the majority of our study sample, and we were thus unable to fully disentangle treatment and disease indication effects. Observed associations were modest and given the multiple analyses that were performed, significant findings could be due to chance. Finally, while several families of potentially-eligible children did not respond to the SEED invitation letter, analyses of data from one SEED site with the most complete data available to assess nonresponse indicated that maternal age, education, and race-ethnicity were associated with nonresponse but other pregnancy variables were not (unpublished analysis). All multivariable analyses were thus adjusted for all three aforementioned demographic factors.

In conclusion, these data support a link between maternal and child immune conditions and ASD, and further suggest that associations may be influenced by disease severity in the mother and ASD phenotype of the child.