Introduction

The autistic spectrum is a complex array of disorders involving impairment in the ability to socialise and to communicate, as well as the presence of repetitive or stereotyped behaviours.1 Autistic children have accelerated rates of brain growth during infancy, with the size of their brain reverting towards normal during childhood.2 Thus, the underlying pathology of autism includes an atypical developmental trajectory.

Autism appears to arise through multiple influences3 on a large number of genes, none of which invariably cause an autistic condition. Similarly, few if any of the genes associated with autism are common to all autistic individuals.4, 5, 6 Hence, the autistic spectrum is likely to have a diverse molecular pathology that converges to produce outcomes that are similar but not identical between individuals.

There is a significant male bias in the incidence of autism.7 The genes with known linkage to the autistic spectrum are predominantly autosomal,6, 8 with no apparent link to sexual dimorphism. This suggests that the male bias in the autistic spectrum may be due to the modulation of a non-dimorphic pathology by sex-specific factors.

The profile of testicular hormones varies during development, with Müllerian-inhibiting substance (anti-Müllerian hormone, AMH) and inhibin B (InhB) being the main secretions for extended periods (Figure 1). AMH triggers the regression of the uterine precursor9 and InhB contributes to the control of spermatogenesis in men.10 However, the functions of these hormones in male fetuses and boys are only beginning to be examined.

Figure 1
figure 1

The pattern of testicular hormones varies during development. The diversity of autistic symptoms suggests that the autistic spectrum arises from multiple influences on multiple cellular processes within the brain, occurring at various times. The figure illustrates the pattern of testicular hormones during male development. Note that the Sertoli cell hormones (AMH: red line, InhB: dark blue line) have a wider temporal influence than testosterone: green line. Each hormone is dimorphic, with the female levels illustrated in Supplementary Figure S1. The levels of AMH are supra-adult throughout development, as is InhB during part of the first year. The figure is based on multiple sources including (refs. 15, 21, 36, 37, 38, 39, 40, 41).

Full size image

In mice, AMH establishes some of the sex biases in the nervous system and non-reproductive behaviours.11, 12, 13 One of the most sexually dimorphic features in children is their rate of maturation, with the brains of boys taking between 2 and 4 years longer to reach peak size than girls.14 This sex bias may relate to AMH and InhB, as a boy's level of these hormones correlate with his developmental trajectory (stature: AMH and InhB:15 representational drawing: AMH). This raises the possibility that these testicular hormones contribute to the male bias in the autistic spectrum. Consistent with this, we report here that the levels of InhB and AMH in autistic boys correlate with the breadth of their autistic symptoms.

Materials and methods

Study participants

Blood samples from 98 boys were obtained from the Autism Genetic Resource Exchange (AGRE) biobank (Autism Speaks, USA), and included 82 boys from the autistic spectrum and 16 controls. The diagnosis was determined by the autism diagnostic interview—revised (ADI-R).16 Sixty-eight of the autistic spectrum boys had a diagnosis of autism, eight not-quite autism and six broad-spectrum. The AGRE defines not-quite autism as being one point away from the cutoff on one or more domains of the ADI-R. The age range of the autistic participants was 4.4–8.9 years (mean=6.63±1.24 s.d.), whereas the non-autistic boys had a range of 5.1–8.1 years (mean=7.13±1.3). The University of Otago's Human Ethics Committee approved the use of the AGRE samples.

Hormone assays

The levels of hormones were measured using commercial enzyme-linked immunosorbent assay (ELISA), according to the manufacturer's instructions. The assays were conducted in two batches, with four samples being assayed in both batches. The ELISAs were: AMH, Beckman Coulter, Brea, CA, USA; A16507 (batch 1) and A79765 (batch 2); InhB, Beckman Coulter, A81303. Assays were conducted in duplicate using serum or plasma samples, with quadratic equations used to calculate sample values from the standard curves. The correlates between the hormones and ADI-R were similar for both batches. The absolute values for the AMH ELISAs showed good agreement between the batches. The data from second batch of InhB ELISAs were 20% higher in both the four cross-batch samples and the mean, and a correction factor was therefore applied. No significant difference was observed between plasma or serum samples for the mean concentrations of either hormone (data not shown).

Statistical analysis

Statistical calculations were undertaken with PASWstatistics18.0. (SPSS Inc, IBM, Armonk, NY, USA). The mean differences between the autistic and non-autistic boys were examined by Student's t-test and by a generalised linear model including age and condition. The relationships between AMH and InhB and a boy's ADI-R score were examined by linear regression, with the partial correlates and the Pearson correlates recorded.

The three ADI-R domains (social development and play; communication; repetitive or restricted behaviours) and their associated subdomains were separately analysed. Nineteen of the autistic spectrum boys were non-verbal, and therefore lacked scores for the non-verbal subdomains (COM2T_CS and COM3T_CS). This reduced the sample size to 63 for these subdomains and for the communication domain. The cohort was also stratified by the boy's ADI-R score for the social development and play domain, in order to compare boys with mild and more severe symptoms.

The number of symptoms for each domain was calculated by adding the number of subdomains with scores of 1, 2 or 3. The average severity of a boy's symptoms was the mean score of subdomains with a score of 1, 2 and 3.

Results

Autism has been postulated to be an extreme male phenotype.17 We therefore examined whether autistic boys have abnormal levels of Sertoli cell hormones. Eighty-two boys with diverse autistic symptoms and sixteen age-matched controls were examined using blood samples from the AGRE biobank. The levels of InhB and AMH in the autistic boys were similar to those of the non-autistic boys, as was the ratio of the two hormones (Figure 2).

Figure 2
figure 2

The levels of Sertoli cell hormones are similar in autistic and non-autistic boys. (a) InhB, (b) AMH and (c) the ratio between AMH and InhB, expressed in arbitrary units. The values for non-autistic boys are shown as black squares (n=16), with the data from autistic boys being illustrated in coloured circles (n=82). The correlation of the hormones to the age of the autistic spectrum boys was: InhB r=−0.17, P=0.14; AMH r=−0.19, P=0.09 and AMH/InhB r=−0.09, P=0.41; n=82. The trend for a slight decrease in InhB and AMH with age has been previously observed, and is statistically significant in some studies.15 Age was therefore included in subsequent regression analyses. The levels of hormones were not significantly different (Student's t-test, and general linear model, to account for any effect of the age) between the autistic (InhB 99.8±44.9; AMH, 127.5±59.7) and non-autistic boys (InhB 84.1±30.2; AMH, 113.3±50.0).

Full size image

We then examined whether a boy's level of Sertoli cell hormones affected the severity of his symptoms, and/or the probability of him being diagnosed with an autistic trait. The boy's level of InhB correlated with their ADI-R score for both the social interaction and communication and language domains, but not for the repetitive and stereotypic behaviour domain (Figure 3). There was no clear association between AMH and the ADI-R domains (Supplementary Figure S2).

Figure 3
figure 3

A boy's level of InhB correlates with two of his ADI-R scores. The boy's level of InhB was measured by ELISA and correlated with the scores for each of the three ADI-R domains. The level of InhB positively correlated with the boy's social score (a) R=0.26, P=0.017, n=82 and his communication and language score (b) R=0.27, P=0.035, n=63, but not with his repetitive and stereotypic behaviour score (c) R=0.09, n=82.

Full size image

The three ADI-R domains are each composed of four subdomains. The 12 subdomains are largely but not totally independent, with the average pairwise correlation coefficient between them being 14±2% (range 0–41%, median 11%, Supplementary Table S1). InhB positively correlated (R>0.2) with 8 of the 12 subdomains, 7 of which were in the social interaction and communication and language domains (Table 1). Five of these correlations were statistically significant, with the other three having P values between 0.05 and 0.08 (Table 1). This indicates that InhB has a broad but moderate association across the breadth of the social and communication domains, rather than a strong effect on a few traits. The boy's AMH levels were not significantly associated with any of the subdomains. However, all of the subdomains with stronger associations to InhB exhibited moderate negative correlates to AMH (Table 1).

Table 1 Relationship between the ADI-R domains and the boy's age and levels of Sertoli cell hormones
Full size table

The scores for each subdomain have multiple components and a high score can be due to a greater number of autistic traits and/or a greater severity of traits. InhB positively correlated with the number of symptoms in both the social interaction (R=0.34, P=0.002) and the communication and language domain (R=0.27, P=0.035), but not with the average severity of the traits (Table 2). As with the total score, there was a non-significant trend for AMH to negatively correlate with the number of symptoms whenever InhB was positively correlated with the symptom (Table 2).

Table 2 Relationship between the number and severity of autistic symptoms and a boy's levels of hormones
Full size table

The association of InhB with the number of symptoms indicates that a boy's level of hormone relates to the clinical discrimination between the absence and presence of an autistic trait. This suggests that the Sertoli cell hormones may have a greater effect when the underlying pathology is less severe. Consistent with this, InhB correlated more strongly with the milder cases than with the boys who had more severe symptoms (Table 2). The correlate was with the number of autistic traits and the total score for each domain, but not with the average severity of the traits (Table 2).

When the mild and severe groups were separately analysed, the negative association between AMH and the ADI-R scores was strong, consistent and statistically significant, whenever InhB positively correlated to a trait (Table 2). For the social interaction domain, the relationship between AMH and the total score of the mild group was R=−0.44, P=0.018. In the mild group, boys with higher levels of AMH exhibited fewer symptoms (R=−0.46, P=0.012).

A semi-quantitative estimate of the potential importance of the hormones can be obtained from the regression curves. With the autistic spectrum disorders as a whole, the effect of reducing a boy's level of InhB from the maxima to the minima of the natural range would on average decrease his social interaction score by 9.1 and his communication and language score by 7.3 (Table 3).

Table 3 Calculated effect of Inhibin B on the autistic scores of boys
Full size table

Discussion

The autistic spectrum disorders have a strong male bias, but are not male-specific. This suggests that the initiation of the underlying pathology involves a non-dimorphic mechanism, with the male bias resulting from modulation of the pathogenic pathway by sex-specific signals. Consistent with this, the levels of Sertoli cell hormones in boys with an autistic spectrum disorder correlated with the breadth of their autistic traits, but not with a uniform direction. The positive correlation between InhB and autistic traits was consistently stronger than the negative correlation between AMH and autistic traits. If these correlates indicate causality, then the net effect of the Sertoli cell hormones would be to exacerbate autistic pathology in most males, but with the extent of the exacerbation being affected by the balance between AMH and InhB.

Neither InhB nor AMH correlated with the boy's repetitive and stereotypic behaviours, even though the incidence of these traits has a male bias. Consequently, the Sertoli cell hormones are unlikely to be the sole male influence on the autistic spectrum. This is broadly consistent with the emerging view that the male phenotype is generated by multiple hormones and various genetic mechanisms, arising from the sex chromosomes.18

The levels of AMH and InhB in the boys with an autism spectrum disorder were similar to the control group (Figure 2), and to previous studies of boys recruited from the community15 (see also Lee et al.19, Andersson et al.20 and Aksglaede et al.21). Consequently, a boy's level of AMH and InhB is insufficient to cause an autistic condition, in the absence of other influences. The autism spectrum disorders are postulated to be an extreme variant of the male brain, secondary to extreme levels of hormones.17 In this study, the autistic boys did not have extreme levels of Sertoli cell hormones.

Both AMH and InhB are plausible modulators of autistic pathology. Brain development is very extended, and unlike testosterone, AMH and InhB are present in male blood throughout the entire period of brain development (Figure 1). Furthermore, as noted in the Introduction, children with an autism spectrum disorder have an altered developmental trajectory,2 with AMH and InhB being recently identified as putative male-specific determinants of the speed of development.15

Complex traits emerge through the interaction of large number of genetic and other determinants. For this to occur, the various influences must converge onto a small number of final pathways. For example, the growth hormone (GH)/insulin-like growth factor (IGF) axis has a central role in growth, even though the GH and IGF genes are not significant genetic determinants of height.22 The GH/IGF axis is important because many of the determinants of height exert their influence by altering the levels of GH and/or IGF1. The influence of GH/IGF on height is therefore large compared with any of the primary determinants of height, because it integrates multiple influences.

AMH and InhB may be points of conversion for the modulation of autistic traits. Neither hormone is a known genetic determinant of autism spectrum disorders. The coefficients of determination (R2) from the observed correlates indicate that the combined effect of AMH and InhB accounted for approximately 20% of the variation in the number of autistic traits in mild cases of male autism and about 7% of the variance across the autism spectrum as a whole. This is similar to the influence of GH/IGF on the variation in height between similar-aged boys15 and is very large relative to the influence of various genetic determinants of autism.4, 5, 6

InhB and AMH are members of the large TGFβ superfamily (Figure 4), whose actions are predominantly channelled through two intracellular pathways, SMAD1,5,8 and SMAD2,3.23, 24, 25 InhB and AMH have opposing effects on autistic symptoms (Table 3), and on the activation of the SMAD1,5,8 pathway (Figure 4). This implicates the SMAD1,5,8 pathway as a potential site of convergence for the autistic spectrum, with low activation of this pathway being predicted to lead to a poor prognosis. The SMAD1,5,8 pathway has multiple sexually dimorphic and non-dimorphic activators (Figure 4). Consequently, this putative point of convergence may be relevant to both males and females with autism spectrum disorder.

Figure 4
figure 4

Signaling by the TGFβ superfamily. In humans, the superfamily has 33 members, consisting of 3 TGFβs, 21 BMP/GDFs, 8 Activins/inhibins and AMH. The family share 5 type II receptors and 7 type I receptors, which lead to the activation of either the Smad 1/5/8 or the Smad 2/3 pathway. The activated Smads are transported to the nucleus, where they regulate transcription. Various binding proteins, such as betaglycan, modulate the specificity of the signaling (reviewed in David et al.23, Lowery and de Caestecker24 and Santibanez et al.25).

Full size image

The various autistic traits are not strongly concordant (Supplementary Table S1), indicating that the autistic spectrum has multiple points of convergence. InhB, but not AMH, also affects activation of the SMAD2,3 pathway,23, 24, 25 and we do not discount the possibility that this pathway contributes to autistic pathology.

All stages of brain development are regulated by the TGFβ superfamily, as is the mature function of the brain. The SMAD pathways are thus plausible determinants of autistic pathogenesis. The receptors for AMH are present on most neurons, and AMH contributes to the sex biases in the brain.11, 12, 13, 15 InhB is a negative regulator of signalling (Figure 4), and its effects on the brain have been rarely studied. This is surprising, given that the receptors it modulates are broadly expressed in the brain26, 27, 28 and regulate the proliferation, differentiation, migration, survival and mature29, 30 function28 of neurons.31 Similarly, astrocyte and oligodendrocyte glial lineages are also regulated through the SMAD pathways.32, 33

The current study is a cross-sectional analysis of hormone levels in boys whose autistic symptoms have already developed. The validity and statistical power of this approach is dependant on two characteristics of the hormones. First, the levels of AMH and InhB in boys are highly variable compared with other hormones, this variation being essential for the generation of statistical power.15, 21 Second, the differences between boys are relatively stable: neither hormone varies with time of day, from day-to-day or from season-to-season, in contrast to the regulation of these hormones in adults.15, 21, 34, 35 Consequently, the rank order of boys based on their hormone levels varies little from year to year15 (see also Aksglaede et al.21). Variation in rank order for hormones during development would be expected to degrade the correlation between hormones and traits. The current study may therefore be a conservative estimate of the influence of InhB and AMH on the breadth of autistic symptoms.

This study has also several inherent limitations; the most obvious being that it does not establish causality, or identify when the Sertoli cell hormones may act during development. If the relationship is causal, the data also does not establish whether InhB and AMH effect the development of autistic traits or vice versa. Furthermore, the complexity of the autistic spectrum means that correlative analyses have more power to detect some influences than others. These limitations are discussed in greater detail in the Supplementary Information. We emphasise that the significance of the reported data is that it gives rise to a novel hypothesis that needs to be tested by interventional studies and various other correlative investigations, including longitudinal studies of at-risk male babies to determine whether their levels of AMH and InhB early in development is predictive of the breadth of autistic traits later in life.

In summary, the reported observations suggest that the male bias in autism arises from multiple factors, with both InhB and AMH being putative influences. They also point to the TGFβ superfamily as being points of convergence for autistic influences. In addition, the data support the notion that sex differences in disorders can arise through the interaction between sex-specific biology and a non-dimorphic pathology. When this is occurring, the influence of the sex-specific factor should be evident in a correlative study, but may not be evident when the mean levels of the factor are compared between the study and control group.