Background
Autism spectrum disorder (ASD) varies significantly in presentation among those affected [
1]. It is therefore unsurprising that the etiology of ASD is thought to be similarly heterogeneous and multifaceted in nature. At present, much remains unknown about the factors that may underlie the disorder [
2]. Research involving the identification of early markers is integral to improving our understanding of the mechanisms underlying the disorder and is likely to significantly impact intervention and outcomes for affected individuals [
3].
Males are approximately four times more likely than females to be diagnosed with ASD [
4,
5]. The extreme male brain theory [
6] posits that this may be because characteristics of ASD are on the extreme end of a spectrum of male-typical characteristics. A growing literature has investigated the factors that may trigger brain masculinization, in particular, by examining possible prenatal biological influences affecting brain development and later behavior. The organizational hypothesis proposes that prenatal sex steroid exposure affects fetal brain structure and function and consequently influences postnatal behavior [
2]. It is proposed that greater exposure to testosterone (one of the most biologically active androgens) during a sensitive developmental period (weeks 8–24 of gestation) may have masculinizing effects on the fetal brain [
7] and hence may be a precursor to autistic-like traits. Circulating testosterone measured in blood samples has been found to be significantly higher in male fetuses than in female fetuses in utero [
8].
Prenatal exposure to androgens has been demonstrated to affect a wide range of developmental factors [
9]. There is a plethora of research in animal studies that has revealed the effects of androgens as a masculinizing agent in both primate and non-primate species [
10‐
13]. Testosterone has primarily been the focus of research due to its biological potency (its ability to exert receptor-mediated changes in cellular or physiological function); however, there is also preliminary evidence that other androgens such as androstenedione (A4) may have masculinizing effects [
14]. Dehydroepiandrosterone (DHEA) has also been suggested to be involved with weak masculinization [
15]. In humans, evidence on the masculinizing role of androgens has come in part from research on individuals with congenital adrenal hyperplasia (CAH). CAH occurs as a result of a genetic abnormality that triggers an overproduction of androgens in utero. As a result, female fetuses are exposed to elevated androgen levels that fall within or above the male-typical range. These, and other studies, support the idea that elevated levels of androgens can result in brain and hence behavioral masculinization [
16], though it has been argued that this research may not take into account social and sample characteristics that might influence these findings [
17]. Often, the effects of testosterone on autistic-like traits have been investigated using indirect methods, such as hand digit ratios. The 2D:4D ratio refers to the ratio of the relative lengths of the second and fourth digits. This ratio has been used for over a century as an indirect measure of prenatal exposure to testosterone and estradiol, the primary biologically active sex steroid in females [
18]. There is some evidence of this ratio being related to prenatal sex steroid levels in amniotic fluid [
19]. It is posited to be related to adult endocrine system regulation traits such as aggression and sexual encounters [
20]. Overall, indirect measures such as the 2D:4D ratio do not show consistent associations with autistic-like traits [
7,
21]. Furthermore, the validity of this marker as a representation of sex steroid concentrations is still debatable [
22]. The digit ratio measure can be considered an insufficiently sensitive proxy for examining prenatal hormonal effects [
23].
In contrast, other studies have attempted to examine prenatal testosterone exposure via direct measurement of sex steroids in amniotic fluid, collected by amniocentesis during the second trimester. This procedure typically occurs during a proposed sensitive time period when sex steroids are considered to have organizational effects on brain development [
24]. Amniotic fluid testosterone concentrations have been found to be significantly greater in males compared to females [
25].
With regard to ASD, a group of studies conducted by Baron-Cohen and colleagues found that high levels of fetal testosterone in amniotic fluid were related to autistic-like traits in children aged 18 to 24 months for boys [
26], decreased eye contact in 1-year-old children [
27], increased restricted interests in 4-year-old boys [
28], reduced empathy in 6- to 8-year-old children for boys [
29], and greater autistic-like traits in children aged 6 to 10 [
30]. Furthermore, a more recent study in a large pregnancy cohort found a latent steroidogenic factor that included cortisol as well as all hormones in the Δ4 pathway (progesterone, 17α-hydroxy-progesterone, androstenedione, and testosterone) was significantly elevated in the amniotic fluid of males who later received diagnoses of ASD. However, testosterone alone was not found to be related to a later diagnosis of ASD [
31]. However, amniocentesis is often carried out in older women or those with other risk factors and so the results may not be representative of the normal population [
32].
The literature suggests that there may be multiple sensitive periods during which sex steroids affect brain functioning and structure [
33]. This position is further supported by research in animal models, which has found that the effects of sex steroids are not restricted to the first two trimesters [
33‐
35]. Recently, studies from the Western Australian Pregnancy Cohort (the ‘Raine’ cohort) have measured sex steroid levels in umbilical cord blood at birth [
36,
37]. Measuring sex steroids in cord blood is useful because it can provide an indication of how sex steroids may play a role in brain development in late gestation. Furthermore, this method may enable more representative assessment of antenatal steroid exposure on a population level [
23]. The cord blood analysis also permits direct assessment of circulating fetal steroid concentrations as opposed to amniotic fluid which primarily reflects fetal urinary testosterone excretion [
32]. The third trimester is a period where dramatic brain growth occurs and is an important period for neuronal maturation, gyrification, and rapid axonal growth [
38]. Umbilical cord sex steroids are known to relate to a range of informative childhood development markers, such as language development [
39,
40], internalizing and externalizing behaviors [
41], and spatial abilities [
42], and are therefore worth examining in relation to autistic-like traits.
Two studies conducted by our group using the Raine cohort cord blood samples showed that high levels of cord blood testosterone were associated with language delay in boys but not girls [
40,
43]. In contrast, another study found no relationship between testosterone concentrations and autistic-like traits measured at age 19 or 20 years via the autism-spectrum quotient AQ [
44].
Previous studies have focused in general on androgens and testosterone in particular. However, there is evidence to suggest that estrogens, in particular estradiol, may promote masculinization, as has been demonstrated in rodent models [
2]. It has been proposed that estradiol promotes cell survival, death, and proliferation in separate brain regions. It may facilitate new dendritic spine synapses in some brain regions but suppress them in others. The effects of estradiol may be mediated in part via epigenetic changes to the DNA and chromatin in processes that are region-specific but are still poorly understood [
34].
Though empirical evidence for the effects of estrogens on human neural development is lacking, androgen and estrogen exposure has been postulated to be responsible for a variety of sexually dimorphic neurodevelopmental and behavioral characteristics including reproductive function [
35] and immune function [
45]. Sex steroids interact with and may enhance or limit the effects of each other [
46,
47]. The nature of potential interactions between androgens and estrogens remains to be investigated. Previous studies have investigated the relative amounts of androgens and estrogens in relation to the 2D:4D ratio. One study involved the ratio between testosterone and estradiol [
19] while the other looked at the ratio between androgen and estrogen composites [
48]. The relative amounts of androgens to estrogens therefore may be worth examining.
The present research therefore sought to extend previous studies by expanding the number of sex steroids assessed through deriving androgen and estrogen composites using the novel method developed by Hollier et al. [
23]. Sex steroid composites and their ratio were then used as possible predictors of autistic-like traits in both the general population and a subset of individuals with ASD. Sex steroids were measured in umbilical cord blood collected at birth, and A and E composites and their ratio were calculated. The androgen composite weights the concentrations of testosterone, A4, and DHEA according to their relative biological potencies and binding affinities to sex-hormone binding globulin (SHBG). The estrogen composite weights the concentrations of estrone (E1), estradiol (E2), estriol (E3), and estetrol (E4) in a similar manner. Based on the free hormone hypothesis, bioavailable sex steroid concentrations were calculated for testosterone, estradiol, and estrone. While this method has been criticized on the basis of methodological inconsistency and unproved assumptions [
49], it remains the traditional method for examining bioactive sex steroids and is widely accepted [
50]. For the present study, ASD diagnostic measures had been collected in childhood and the AQ had been administered in early adulthood.
It is hypothesized that, based on previous masculinization studies, a high A composite (and possibly A to E ratio) would be associated with increased autistic-like traits in the whole sample and that those individuals with a clinical diagnosis of ASD would have particularly high A and possibly high A to E ratios. Furthermore, it was expected that these associations would occur for outcome variables that demonstrate a sex difference.
Discussion
The current study explored the relationship between perinatal sex steroids and autistic-like traits in young adulthood. To assess perinatal steroid exposure, composites of biologically active androgens and estrogens in umbilical cord blood were calculated, along with their ratio. This study is therefore unique in both its biological methodology (composite sex steroid calculation and umbilical cord sampling) and the nature of the longitudinal data collected (sample size and period over which data were collected).
Males demonstrated significantly higher androgen and higher A to E composite ratios than females, though these effects were small. In contrast, estrogen levels did not differ significantly between sexes. The significant sex differences in levels of androgens and A to E composite ratio are consistent with previous studies suggesting that males have elevated androgen levels perinatally compared to females [
37] .
In regard to sex differences on the AQ, males scored significantly higher than females on the details/patterns subscale. This contrasts with two previous studies that involved Dutch parent and student samples and a Scottish student sample. These studies showed significantly higher scores for males in the total AQ score and in the social skills subscale but not in the details/patterns subscale [
57,
61]. The contrasting results found here may reflect cultural differences between samples. It may also be that the Raine cohort represents a more general population sample as opposed to the student and parent samples involved in the two previous studies.
In contrast to our hypotheses, correlational and regression analyses of continuous variables revealed no significant relationships between sex steroid variables and AQ scores. In other words, composite indices of concentrations of androgens and estrogens that took into account a wider range of biologically active sex steroids, as well as their ratio, did not relate to the total AQ score or any of the three subscales. Interestingly, when the data were analyzed categorically, inclusion of covariates in the logistic regression resulted in a small effect being found for females, whereby those whose levels of androgens were in the second lowest quartile scored significantly higher than those who scored in the other quartiles on the details/patterns subscale. This result, however, is likely to be spurious given (a) the lack of sex differences found in the outcome measure, (b) the large number of analyses conducted, and (c) that animal models generally demonstrate linear relationships between androgens within the normal range and behavior. Therefore, the overall pattern of results indicate that the perinatal composites of androgens and estrogens as well as their ratio are not significantly related to autistic-like traits at age 19 and 20 in the general population. This further supports the null results found by Whitehouse et al. (2012) where testosterone concentrations were found to not significantly predict AQ scores in the Raine cohort.
It was hypothesized that a high A composite (and possibly A to E ratio) would predict higher scores on the AQ and that those individuals with a clinical diagnosis would have particularly elevated A composites (and perhaps A to E ratios). In the current cohort, five participants had a known diagnosis of ASD. Contrary to expectations, those participants with an ASD diagnosis did not, overall, display atypical sex steroid values with all but one falling within one standard deviation of the mean for the wider sample. Hence, our findings did not support our primary hypothesis.
There are several possible explanations for the lack of relationship between cord blood sex steroid measures and self-reported autistic-like traits on the AQ. First, it may be that prenatal sex steroids exert organizational effects predominantly during the theorized
sensitive period that occurs in early gestation (weeks 8–24). Research conducted by Baron-Cohen and colleagues as part of the Cambridge Fetal Testosterone Project has so far demonstrated several relationships between early social markers and prenatal testosterone levels obtained via amniocentesis [
27‐
30]. However, organizational effects may occur during several periods, and furthermore, these periods may change depending on the area of development in which the effects are exerted.
Perhaps the sex steroid levels and interactions that affect aspects of brain development leading to certain autistic-like traits (e.g., those such as early social behaviors) occur early in gestation whereas sex steroids may influence other neurodevelopmental characteristics (e.g., language development) later in gestation. For example, previous studies have found relationships between perinatal testosterone in cord blood and delay in early language development in boys [
40,
43]. There is neural evidence to support the development of brain regions in the third trimester of pregnancy that are central to language. During this period, the appearance and growth of the gyri associated with the posterior and frontal areas, including perisylvian structures, occurs [
62,
63]. These structures, including the planum temporal, pars triangularis, and the inferior frontal gyrus, are suggested to relate to language function [
64,
65]. Furthermore, postmortem [
66‐
68] and neuroimaging [
69,
70] studies have found atypical perisylvian structures among children with developmental language difficulties. Therefore, the neurodevelopmental influence of steroids in the perinatal period may be informative for language development.
Future investigation could further delineate the periods of gestation in which sex steroids may play a part in different aspects of brain development and determine how we might go about measuring the factors that contribute to this development in a non-intrusive and safe manner in order to be able to relate these factors to behavior later in life.
Another potential explanation for these results is that the extent to which sex steroids are biologically active is not solely determined by potency and amount. The distribution and functionality of androgen receptors present in the fetus may also be important as receptors are responsible for the binding and translocation of androgens to the nucleus of cells. Furthermore, the functionality of these receptors is determined by the androgen receptor gene which affects the sensitivity of the fetus to androgens [
71]. Studies have found that defects in this gene can result in complete or partial androgen insensitivity whereby complete insensitivity results in genotypic males presenting as phenotypic females [
72,
73]. The cysteine, adenine, and guanine (CAG) repeat sequence in the androgen receptor (AR) gene, located at Xq11-12, has been known to affect serum testosterone levels. More repeats have been shown to be related to diminished androgen sensitivity and result in a negative feedback loop that increases androgen production [
74]. Studies have demonstrated that AR-CAGn and serum testosterone levels are positively correlated in males [
75] and inversely associated in females [
76]. Similarly, estrogen uptake relies on the dispersion of estrogen-specific receptors. In addition, other sex steroids such as progesterone and xenoestrogens may have possible prenatal influences on the developing brain [
77,
78]. Examining sex steroid receptors, their genetic counterparts, and other possible influencing sex steroids is another important direction for research in understanding the complexities of prenatal brain development.
Those individuals whose mother was living below the poverty line as well as children of mothers who did not complete secondary education had significantly higher AQ scores. This is consistent with a recent population-based study in Sweden where low socioeconomic status was found to be a risk factor for ASD [
79]. It may be that socioeconomic factors are a risk factor for genetic and or environmental vulnerabilities related to the broad autism phenotype. It is important to note that the direction of causation is unknown, and it may be that parents with autistic-like traits are less likely to complete secondary education and more likely to live below the poverty line. These parents may have children who are more likely to have autistic-like traits.
When examining sample attrition characteristics, it was found that the mothers of the individuals who did not complete the AQ were socioeconomically disadvantaged and were more likely to be younger, less educated to a secondary level, have an income below the poverty line, smoke cigarettes, and drink alcohol while pregnant, relative to the mothers of those who did complete the AQ. However, sex steroid levels did not differ significantly between males who did and did not complete the AQ. In females, those who completed the AQ did have significantly lower A to E ratios compared to the women who did not complete the questionnaire, but this effect was only small. These results suggest socioeconomic status in the current study may have influenced drop out such that those with lower income were more likely to drop out before completing the AQ.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
JAK undertook the analysis of the sex steroid data and worked in conjunction with LPH to develop the calculation of the composite sex steroid measures. ESLJ, AJOW, and MTM developed the hypotheses for this study, while ESLJ conducted the statistical analyses, drafted the manuscript, and was responsible for correspondence and coordinating the study. All authors contributed to the interpretation and discussion of the results and have read and approved the final version of the manuscript.