Sex-specific and sex-independent steroid-related biomarkers in early second trimester maternal serum associated with autism

Autism is a neurodevelopmental condition affecting approximately 2% of the population [1]. Autism is characterized by difficulties with social functioning, repetitive behaviors, restricted interests, insistence on sameness, and atypical sensory response [2]. Several pre-/perinatal autism risk factors associated with metabolic/steroid dysregulation and inflammation have been identified. Among these, maternal conditions are pre-existing/gestational diabetes and hypertension/pre-eclampsia (collectively hereafter referred to as prenatal metabolic syndrome, PNMS), increased pre-pregnancy body mass index (BMI), elevated LDL cholesterol, and gestational weight gain [3‐12].

Changes in maternal metabolism and immune responses foster the health and growth of the fetus [13, 14]. Maternal hyperinsulinemia promotes transfer of nutritional resources to the fetus via the placenta [15, 16]. Maternal hypercortisolemia influences fetal response to stress via hypothalamic–pituitary–adrenal (HPA) programming and contributes to the immune privileged state in utero [14, 17]. The most vulnerable gestational period for disruptions within the maternofetoplacental unit to influence autism risk has yet to be determined, though some studies found an association between increased mid-gestation maternal stress and autism risk [18, 19]. The fetal HPA axis matures during this interval, and its programming can be influenced by maternal metabolic conditions and in utero stress [20‐23]. Prior biomarker studies indicate greater autism risk associated with steroid dysregulation and inflammation during this gestational window [24‐30].

In a pilot study with autistic and comparison groups enriched for PNMS exposure, Bilder et al [30] found higher estradiol and lower dihydroepiandrosterone (DHEA) levels in early second trimester maternal serum from term pregnancies associated with the presence of autism among offspring [30]. In a study of males with autism without consideration of PNMS exposure, Baron-Cohen et al [31] also identified higher estrogen levels (along with increased progesterone) in second trimester amniotic fluid [31]. Bilder et al [30] and Baron-Cohen et al [31] both interpreted their respective estrogen findings as suggestive of increased fetal steroidogenic activity in autism’s prenatal etiologic pathway. Additionally, Bilder et al [30] found lower sex hormone binding globulin (SHBG) levels in pregnancies associated with autism. As SHBG inactivates estradiol through binding [32], Bilder et al [30] proposed that this finding could potentiate the impact of high estradiol levels. Increased placental estradiol activity accelerates fetal HPA axis maturation during this gestational window [33], which may be relevant to autism pathogenesis. Maternal serum SHBG levels are also closely linked to PNMS as a predictor of gestational diabetes and pre-eclampsia [34‐42]. How these prenatal autism biomarkers are influenced by fetal sex and pregnancy duration, both well-recognized autism risk factors, has yet to be investigated.

The current study extends the investigation of autism likelihood associated with early second trimester sex steroid-related serum biomarkers by enlarging autistic and non-affected offspring groups and including preterm pregnancies [30]. Both groups are enriched and matched for PNMS exposure. Study aims are to (1) evaluate autism risk associated with maternal serum SHBG and sex steroid hormone levels and (2) determine whether this relationship differs by fetal sex and term/preterm status.

Approval for this study was obtained from the Utah Registry of Autism and Developmental Disabilities (URADD) Oversight Committee, the Utah State Office of Education, and Institutional Review Boards of the University of Utah (UU), Intermountain Healthcare (IM), Utah Department of Health and Human Services (UDHHS), and Resource for Genetic and Epidemiologic Research Review Committee. The latter is an oversight body that regulates Utah Population Database (UPDB) access. UPDB is a robust, comprehensive medical research resource that accesses many sets of high-quality, population-based, individual-level records [43].

Likelihood of autism among offspring was associated with low SHBG and increased estradiol levels in early second trimester maternal serum. SHBG levels demonstrated a significant inverse association with the presence of autism among offspring overall and across gestational age and sex categories. SHBG production occurs in the liver and is influenced by insulin sensitivity and body fat composition during pregnancy [36, 52, 55]. Although SHBG’s main function is to bind sex steroids to reduce free sex hormone levels, SHBG has also been found to link inversely to insulin resistance independent of sex hormone levels [56]. Wallace et al [56] provide a review of observational and genetic studies that describe the relationship between low SHBG levels and type 2 diabetes mellitus and support insulin resistance as a mechanistic link underlying this association. SHBG has previously been used as a proxy for gestational insulin sensitivity [41] as estradiol and testosterone play a limited role in SHBG regulation during pregnancy [35]. Current study findings support this prior work as the relationship between SHBG and autism likelihood was unchanged following adjustment for sex hormone levels. Prenatal maternal serum SHBG levels have previously been found to correlate negatively with fasting plasma glucose, insulin, and C-peptide levels [34, 35]. Serum SHBG levels are lower in women with gestational diabetes [36] and have demonstrated predictive value for gestational diabetes and hypertension [34, 35, 37‐42]. Interestingly, the current study found lower SHBG levels in the autism cohorts with and without PNMS exposure. If lower maternal serum SHBG were to reflect insulin resistance, study findings suggest that insulin insensitivity could be occurring in the early second trimester of pregnancies associated with autism even in the absence of subsequent clinical PNMS manifestations.

BMI is recognized as a major determinant of SHBG levels and has been inversely associated with SHBG in prior obstetrical studies [42, 52, 53]. The connection between BMI and SHBG levels has been attributed to higher liver fat content because lipogenesis suppresses hepatic SHBG synthesis [52, 55, 57]. While the post hoc analysis of SHBG and BMI demonstrated a significant association in the overall cohort, this relationship appeared to be driven entirely by the unaffected cohort and was absent in pregnancies from which offspring developed autism. Instead, SHBG levels in the autism group were low across the BMI range. Low SHBG levels in the autism group regardless of BMI suggest that metabolic processes could be occurring during pregnancies associated with autism that supersede SHBG synthesis suppression by hepatic lipogenesis.

A pilot version of the current study found higher maternal serum estradiol levels, in combination with lower DHEA and SHBG levels, to be associated with increased likelihood of autism among term offspring [30]. Collectively, these findings were interpreted as indicating increased steroidogenic activity by the fetus at risk for autism during early second trimester, as DHEA from both maternal and fetal adrenal glands acts as a substrate for placental estradiol synthesis [58, 59]. Additionally, higher placental estradiol activity during this gestational window can accelerate fetal HPA axis maturation [33, 60]. By extending the pilot study with 65 additional participants, 39 of whom were born preterm, current findings indicate that the link between increased maternal serum estradiol and autism likelihood appears specific to male sex and term delivery. In an exclusively male autism cohort, Baron-Cohen et al [31] identified elevated mid-gestation amniotic fluid estrogen levels and attributed this finding to their longstanding theory that autism is caused by the impact of increased fetal steroidogenic activity on sex-specific brain development. Amniotic fluid and maternal serum estradiol levels reflect complementary components of the maternofetoplacental unit as amniotic fluid at this stage in pregnancy is populated by fetal urine output, while maternal serum estradiol is produced primarily by the placenta using fetal and maternal substrates [58, 59, 61]. Further study is needed to understand whether the link between higher amniotic fluid steroid levels, and autism is also limited to male and term offspring.

Unlike the current study’s null findings between estradiol and autism likelihood in the overall model following adjustment for covariates, Windham et al [28] found lower levels of another estrogen (i.e., estriol) in mid-gestation maternal serum significantly linked to autism. Study design differences may account for these discrepancies as the Windham et al [28] study used a diverse, population-based sample rather than a small sample enriched for PNMS exposure; estradiol and estriol also differ significantly in their receptor binding capacity and clearance rate within maternal serum. Further investigation into steroid-related maternal serum biomarkers within a large population-based sample is merited to understand how the mother, pregnancy, and child’s characteristics could influence the relationship between maternal mid-gestation serum estrogen levels and autism.

The sex-specific association between maternal serum estradiol and the likelihood of autism among offspring may reflect sex differences in fetal response to prenatal adversity that place male newborns at greater risk for complications [62‐65]. Estradiol has a bimodal effect on inflammation with lower levels enhancing T-cell and pro-inflammatory cytokine responses and higher levels (i.e., during pregnancy) reducing these immune responses while promoting B-cell antibody production [66]. Subsequently, cell-mediated autoimmune diseases (e.g. rheumatoid arthritis) improve during pregnancy while autoantibody-mediated conditions worsen (e.g., systemic lupus erythematosus). Maternal anti-fetal brain autoantibodies have been implicated in autism, and mouse models demonstrate that anti-fetal brain antibodies produce behavior changes (e.g., increased repetitive behaviors, reduced social interest) in males only [67, 68]. If extended exposure to high estradiol levels influences inflammatory processes, then, affected pregnancies carried to term may increase offspring’s susceptibility to subsequent adverse effects. Interestingly, Windham et al. also suggested that the inverse relationship they identified between mid-gestation maternal serum estrogen levels and autism likelihood could be attributed to hormonal influences on the immune system [28].

While outside the context of pregnancy, SHBG has also been directly linked to inflammation. SHBG has been inversely associated with the inflammatory marker C-reactive protein and reduces inflammatory processes in vitro [69, 70]. Introducing SHBG to adipocytes and macrophages suppresses the pro-inflammatory response induced by lipopolysaccharide [70]. If these properties extend to pregnancy, low SHBG levels could herald a pro-inflammatory state in the maternal compartment that could adversely affect fetal development. Prior maternal serum biomarker studies have demonstrated higher autism likelihood associated with inflammation during the gestational window in which current study samples were collected [24‐27].

This study harnessed a unique collection of resources available in Utah to investigate maternal metabolic conditions and the in utero steroid environment associated with ASD. Because serum SHBG levels do not require timed sample collection, a proxy for insulin sensitivity can be evaluated within banked serum from an obstetrical cohort large enough to examine risk associated with a childhood outcome affecting 2% of the population [36, 41]. Study findings justify investigation into early second trimester phenomena that may suppress maternal SHBG production and influence fetal neurodevelopment even in the absence of subsequent clinical PNMS manifestations.

The relative racial and ethnic homogeneity of Utah’s population limits the generalizability of study results. Although significant differences by autism status were identified in concentrations of SHBG overall and of estradiol in participant subgroups, differences by PNMS exposure failed to reach statistical significance, which may reflect insufficient statistical power. To eliminate potential batch effects on the comparison between preterm and term autism+/PNMS + offspring, all autism+/PNMS+ offspring were analyzed within the first batch which limits the interpretation of results related to PNMS exposure between the autism subgroups across batches. Compared to steroid hormones, inflammatory serum biomarkers are less stable over time and across freeze–thaw cycles [71, 72 . The absence of accompanying inflammatory biomarkers also limits this study’s ability to inform how findings relate to co-occurring inflammatory processes. Further work is needed to understand how lower SHBG and elevated estradiol levels may coincide with in utero processes that could collectively contribute to autism’s development.

Study findings suggest that maternal SHBG production is suppressed during the early second trimester of pregnancies associated with autism among offspring independent of fetal sex, gestational duration, BMI, sex hormone levels, and subsequent PNMS emergence. Autism likelihood associated with higher maternal estradiol levels appears limited to male offspring and term birth. Understanding how prenatal maternal SHBG and estradiol levels are linked to autism among offspring could facilitate early detection screening strategies and foster the development of therapeutic interventions to optimize fetal health during this critical period of neurodevelopment.