Maternal steroid levels and the autistic traits of the mother and infant

Autism is a neurodevelopmental condition characterised by difficulties in social communication alongside unusually restricted interests, repetitive behaviour and frequent sensory hypersensitivity [4]. Co-occurring clinical conditions are common and include learning difficulties, sleep disorders and nutritional intolerances [19]. Diagnosis of autism is possible as early as 18 months of age [27]. Autistic traits exist along a spectrum in the wider population [11]. This can be measured in infancy, as demonstrated by the Quantitative Checklist for Autism in Toddlers (Q-CHAT), a novel, dimensional measure of autistic traits shown to predict later autism diagnosis in several validation studies [2, 3, 43].

Autism is diagnosed more often in males than in females, despite increasing awareness of its differential presentation in females [33]. The reasons for this gender ratio likely reflect both prenatal biology (genetics and sex steroids) and limitations in current diagnostic methods [32]. In the general population, there are on average, sex differences in specific psychological traits (e.g. systemising and empathising), which are shifted towards a “male” profile in autistic individuals of both genders [8, 23].

Several lines of evidence indicate that prenatal sex steroid hormones may be mediating autism likelihood. First, a study of neuroanatomical differences between autistic and neurotypical brains revealed autistic females have atypical structure in regions that substantially overlap with sexually dimorphic regions in neurotypical controls. This suggests that autism affects female brains in regions related to sexual differentiation, which is in turn regulated by prenatal sex steroid hormones [31]. This male-shift is also evident in childhood, as indicated by studies showing that facial features of autistic boys, girls and their siblings are masculinised compared to neurotypical controls [48, 49]. Autistic males also have higher levels of steroid hormones in amniotic fluid, when these were assessed in a multivariate analysis, as well as higher levels of estrogens in particular in univariate analyses of the same amniotic fluid samples [9, 10]. Additionally, estradiol levels are elevated in maternal serum of pregnancies linked to a later autism diagnosis in the child [13].

Traits and developmental profiles related to autism have also been linked to prenatal sex steroids. For example, prenatal levels of testosterone measured during amniocentesis are negatively correlated with the frequency of eye-contact at 12 months [34] and, vocabulary size at 18 and 24 months [35], and are positively correlated with restricted interests, attention to detail [29] and with autistic traits at 18 and 48 months [5, 6].

Epidemiological studies have revealed that maternal polycystic ovary syndrome (PCOS) increases the likelihood of autism in their children [16, 17, 30]. This effect is not primarily driven by shared genetics and is further modulated by the sex of the offspring [15].

PCOS is a complex syndrome affecting between 4 and 20% of women of reproductive age (depending on the diagnostic criteria), impacting their fertility, metabolism and endocrine regulation [44]. It is particularly associated with higher baseline levels of androgens, as well as with a wider endocrine dysregulation during pregnancy (e.g. on placental function) [37]. Steroid-related conditions such as PCOS and placental complications are also more common in autistic people [17, 36, 40], and autistic adults demonstrate signs of steroid imbalance in various tissues [18, 28, 42]. Further research is needed to understand the nature and timing of this endocrine imbalance in autism, as well as to study whether it extends to the wider spectrum of autistic traits in the general population.

The prenatal environment shows baseline sex differences in steroid production via the placenta [21], as well as in markers of placental formation and function, such as human choriogonadotropin (hCG) [1, 20, 38]. Atypical levels in both have been found in maternal serum of pregnancies that resulted in an autistic child, with both higher (for estradiol) and lower levels (for estriol) reported compared to controls [13, 53]. However, these prenatal factors have not been studied together with fetal sex and in relation to both maternal and infant autistic traits.

To evaluate all these prenatal factors, we assessed both the mothers and their infants for autistic traits [via the Autism Spectrum Quotient (AQ) and Q-CHAT, respectively] in a longitudinal cohort. We then tested whether steroid hormone levels (estradiol, testosterone, DHEAS, progesterone) and placental markers (e.g. hCG) in maternal serum collected during the mother’s first ultrasound appointment predicted the autistic traits measured in mothers during pregnancy and in infants at the 18-month follow-up.

This is the first clinical longitudinal study to report on the endocrine profile of pregnant women and how this relates to their own autistic traits and to the autistic traits of their infants. First, we found that the fraction of free circulating estradiol (FEI) during pregnancy correlated positively with the autistic traits of pregnant neurotypical women. Second, we found that maternal estradiol was associated with infant autistic traits (as measured by the Q-CHAT) in a sex-dependent way, with a positive correlation in males but not females. Third, we found that the opposite was true for standardised hCG MoM values, where a negative correlation with Q-CHAT scores in males was noted. These associations with early neurodevelopment were independent of the mother’s AQ score or PCOS status, as well as other infant characteristics such as birth weight or age at the time of assessment. Finally, this is the first longitudinal study to show a significant positive correlation between maternal and infant autistic traits, as measured by the AQ and Q-CHAT, respectively.

The association between free circulating estradiol and maternal AQ was independent of maternal age or a diagnosis of PCOS. In addition, women with clinical history of excess body hair growth in more than one area of their body also had significantly higher autistic traits scores, as well as higher levels of circulating estradiol and FEI (Fig. 2). This is consistent with previous studies that reported higher rates of steroid-exposure related symptoms, such as hirsutism in autistic women [40]. The correlation between FEI and AQ was independent of PCOS diagnostic status, and women diagnosed with PCOS did not have higher AQ scores (Table 1). This may be due to reduced statistical power (n = 13 with PCOS diagnosis) or the fact that all the women in this cohort had no autism diagnosis themselves and were without significant challenges in fertility, contrary to previous studies that found a link between PCOS and autism [17].

There was no association between testosterone levels or the FTI and maternal autistic traits in this study. Testosterone correlated positively with estradiol (Additional file 1: Table S3) but did not differ between women with and without PCOS, or correlate with a clinical history of hirsutism. Previous studies have also shown that circulating testosterone levels measured at a single time-point do not always correlate closely with their associated clinical parameters, such as hair growth [39]. Particularly during pregnancy, estradiol levels may be more clinically informative, as testosterone is rapidly aromatised into estradiol by the placenta. Estradiol could then be interpreted as the end-product of wider steroidogenesis and may be a better biomarker of the ‘steroidopathy’ previously indicated by epidemiological studies of autistic women [40].

This is also the first study to investigate circulating hormones in maternal serum in relation to autistic traits in their infants measured between 18 and 20 months of age via the Q-CHAT. This questionnaire was developed to enable parents to quantify autistic traits, moving away from a restrictive binary format of items toward a Likert format. To date, the Q-CHAT has been validated in several studies. Interestingly, the Q-CHAT also shows sex differences in early autistic traits, with males scoring significantly higher than females in larger cohorts [2, 3, 41, 43].

The current study found significant moderation effects of infant sex on the associations between hormones and neurodevelopment. Specifically, increased maternal estradiol was more predictive of Q-CHAT scores of males than females (Fig. 4; Table 3). This was independent of potential confounding variables, such as maternal age, diagnosis of PCOS and maternal autistic traits. Contrary to the results with the AQ, free estradiol levels (FEI), as estimated via serum SHBG, were not predictive of Q-CHAT scores. SHBG is a peptide, and it does not cross the placenta as easily as steroids, as shown in rare cases of partial deficiency in mothers but not their fetus [26]. Maternal SHBG may therefore not accurately capture the bioavailability of steroids in the fetal circulation nor accurately predict potential effects on infant neurodevelopment.

Further research is needed to understand the interaction between maternal estradiol levels and fetal sex in predicting Q-CHAT scores. As with the observed male bias in autism diagnoses, the lack of an association between hormone levels in females may reflect behavioural differences that are not adequately captured by this particular instrument [32]. Alternatively, this interaction may represent evidence of differential liability to prenatal hormone exposure, whereby males are affected more than females by the same maternal endocrine factors. Since males are undergoing an additional increase in steroid levels during mid-pregnancy, due to the activation of the testes, this interaction could also be attributed to the added effect of elevated steroids in the fetal circulation of males [51]. While fetal steroids were not measured directly in this study, previous comparisons in humans showed that estradiol levels correlated between maternal serum and amniotic fluid, in both the second and third trimesters [50]. The same was found for hCG levels in the second trimester [47]. In contrast, increases in androgen levels of male pregnancies were not detectable in maternal serum [50]. Therefore, the interaction of sex with estradiol levels in predicting Q-CHAT scores, in this study, could potentially be attributed to the additive effect of elevated androgens in the fetal circulation of males, which may not have been captured by assaying maternal serum, rather than amniotic fluid [5].

The findings with hCG mirror those for estradiol, showing a negative association with Q-CHAT scores in females but not males. hCG is produced by the developing trophoblast cells in the placenta and regulates early implantation as well as steroid production [14]. hCG levels also show baseline sex differences in typical cohorts as early as the first half of pregnancy [1]. In the current cohort, hCG was measured late in the first trimester, during the first ultrasound visit, as part of the screening programme for Down’s syndrome. This is part of routine prenatal screening in the UK to test for placental dysfunction that can often be indicative of genomic instability due to aneuploidies [54]. In cases of clinically diagnosed autism, both very low and very high levels of hCG have been found in maternal serum, leading to a “U-shaped” association when studying both males and females [53]. Furthermore, autistic traits in the children have been associated with the severity of nausea and ‘morning sickness’ during pregnancy, symptoms that have also been linked to high hCG levels [22, 52]. Further studies into placental functionality could offer insight into these observations, the role of sex, and more specifically, whether the observed interaction effect for Q-CHAT score is part of an adaptive response that is more pronounced in females than males.

Limitations of the current study include the relatively small sample size, as well as potential ascertainment bias given the voluntary process of recruitment. In addition, AQ and Q-CHAT scores were both rated by the mother herself. This was addressed by controlling for maternal AQ when studying associations between hormones and infant Q-CHAT scores. The items of the AQ-Adult and Q-CHAT are also substantially different, with the latter dealing with behavioural and developmental milestones that are specific to infants rather than to interests and personality traits that are more evident in adulthood. Furthermore, the findings of the current study may have been inflated by Type I errors, as p value thresholds were not corrected for multiple testing. However, the findings on estradiol were consistent for both maternal and infant outcomes (Fig. 4), were reflected in clinical hirsutism differences (Fig. 2), and are in accordance with those observed in other studies of clinical autism [10, 13]. The high degree of correlation between many of the assessed hormones (Fig. 3) also indicates a common functional and regulatory framework. The association tests for individual hormones may therefore not be entirely independent, but instead affected by a common steroidogenic factor as previously reported [9]. Replication of these findings in a larger, independent cohort is warranted to confirm their validity.

This is the first longitudinal study to report associations between maternal steroidogenic factors and the autistic traits of both the mother and her infant, with significant moderating effects of sex being noted for the latter. Additional research is needed to replicate these findings, to establish how maternal steroidogenesis may affect fetal neurodevelopment, and to determine how these processes interact with genetics to disproportionately increase the liability for autism in males.