Sexual Differentiation and the Neuroendocrine Hypothesis of Autism



The phenotypic expression of autism spectrum disorders varies widely in severity and characteristics and it is, therefore, likely that a number of etiological factors are involved. However, one finding which has been found consistently is that there is a greater incidence of autism in boys than girls. Recently, attention has been given to the extreme male hypothesis—that is that autism behaviors are an extreme form of typical male behaviors, including lack of empathy and language deficits but an increase in so-called systemizing behaviors, such as attention to detail and collecting. This points to the possibility that an alteration during sexual differentiation of the brain may occur in autism. During sexual differentiation of the brain, two brain regions are highly sexually dimorphic—the amygdala and the hypothalamus. Both of these regions are also implicated in the neuroendocrine hypothesis of autism, wherein a balance between oxytocin and cortisol may contribute to the disorder. We are thus proposing that the extreme male hypothesis and the neuroendocrine hypothesis are in fact compatible in that sexual differentiation of the brain towards an extreme male phenotype would result in the neuroendocrine changes proposed in autism. We have preliminary data, treating developing rat pups with the differentiating hormone 17-β estradiol during a critical time and showing changes in social behaviors and oxytocin, to support this hypothesis. Further studies should be undertaken to confirm the role of extremes of normal sexual differentiation in producing the neuroendocrine changes associated with autism. Anat Rec, 2011. © 2011 Wiley-Liss, Inc.


Simply stated, the hypothesis which we are proposing in this article, and the work currently ongoing in our lab, is that enhanced sexual differentiation of the brain, towards the male phenotype, leads to changes in the neuroendocrine factors, oxytocin and cortisol. These factors are involved in the processing and evaluation of social bonding experiences and any disruption in their typical function can lead to the primary defining characteristic of the autism phenotype—inadequate social bonding.

The Extreme Male Hypothesis of Autism

Autism is a disorder of brain development and thus any factors which regulate brain development and are known to be altered in autism should be considered as possibly contributing to the phenotype. For example, much of our previous work and work of others, has focused on the known role of serotonin in brain development as well as the well known hyperserotonemia which occurs in autism (Whitaker-Azmitia,2005). More recently, researchers have proposed a role for sexual differentiation of the brain in the phenotypic behaviors of autism, in large part because of the finding of a 3.9–9-fold increase in the incidence of autism in boys compared to girls (Bertrand et al.,2001; Chakrabarti and Fombonne,2001; Gurney et al.,2003; Chakrabarti and Fombonne,2005).

The extreme male theory of autism, as proposed in most detail by Baron-Cohen, states that the sexually divergent behaviors which are more closely aligned with men are overemphasized in autism, while those behaviors more typically seen as female are deficient in autism. As further described by Baron-Cohen (2002) women have greater language skills, spend more time looking at the eyes and making eye contact and have a greater ability to understand the feelings of others than men do—and all of these characteristics are deficient in autism. Conversely, men have a greater attention to detail, to collecting and a preference for interactions with strict rules or closed systems such as computers—and all of these are overdeveloped in autism. Men have a greater need for a predictable environment while women have a greater facility for responding to novel social situations. According to the extreme male hypothesis, then, the traits of systemizing far outweigh the traits of empathy and social bonding, such that the phenotypic behaviors of autism result (Baron-Cohen,2009).

Neuroendocrine Hypothesis

Social approach or avoidance is regulated by a number of neuropeptides and neurohormones, in particular oxytocin and cortisol, which are found principally in sexually dimorphic brain regions—the hypothalamus and the amygdala.

The paraventricular nucleus (PVN) of the hypothalamus contains cells expressing a number of hormones and neuropeptides, some of which are considered neurosecretory (i.e., project to the pituitary) and some of which project to other brain regions. The cells projecting to other brain regions consist primarily of two neuropeptides, oxytocin and vasopressin, and the neurohormone corticotrophin releasing factor (CRF). Both CRF and oxytocin have been implicated in autism as part of the neuroendocrine hypothesis of Jay Shulkin (2007).

Many human studies have found a role for oxytocin in social memory and memory for faces (Savaskan et al.,2008; Guastella et al., 2009) but not an increased memory for non-social stimuli (Rimmele et al.,2009) increased trust (Kosfield et al.,2005; Zak et al.,2005; Baumgartner et al.,2008; Petrovic et al.,2008), increased security of attachments and partner support (Grewen et al.,2005; Buchheim et al.,2009), increased empathy (Barraza and Zak,2009) and increased ability to read the affective state of others (Theory of mind; Domes et al.,2007; Goldman et al.,2008). Oxytocin levels can be observed to change quickly in response to the social environment, including after warm contact between couples (Grewen et al.,2005; Holt-Lunstad et al.,2008) massage (Carter et al.,2007) or maternal-infant bonding (White-Traut et al.,2009). Oxytocin mediates the effects of social support in dealing with stress and improves communication between couples (Ditzen et al.,2009). The administration of oxytocin has been shown to decrease the fear-induced activity of the amygdala (Kirsch et al.,2005). There are suggestions that some of the therapeutic effects of antidepressants in social anxiety disorders are through oxytocin release in the PVN (Raap and Van de Kar,1999). Many oxytocin-containing cells of the PVN have estradiol receptors in the adult, such that these may be a site of gender based activating (as opposed to organizing) effects of hormones (Bodo and Rissman,2006) and blood levels of oxytocin are higher in women than men (Ozsoy et al.,2009). Oxytocin plays an important role in the sexually dimorphic aspect of human social reciprocity and social brain development (Yamasue et al.,2009).

Uvnas-Moberg has summarized the role of oxytocin as being responsive to changes in the social environment to bring about “calm and connection” (1997,1998). The connection may be due to the direct central effects of oxytocin, and the calm may be produced by oxytocin's ability to decrease levels of the stress-related hormone, cortisol (Heinrichs and Gaab,2007; Ditzen et al.,2009, and below). All of these functions of oxytocin can be seen to be lacking in people with autism and a lower peripheral blood level of oxytocin has been reported (Modahl et al.,1998; Green et al.,2001). Moreover, several studies have shown that administering oxytocin improves symptoms of autism. Intranasal oxytocin improves the ability of young autistics to recognize emotions in the “Reading the Mind in the Eyes” task (Guastella et al.,2010) and increases the positive feelings of interacting socially in a simulated ball toss game and increases preference for reciprocal relationships (Andari et al.,2010).

The amygdala plays a role in interpreting sensory and emotional information and motivating resultant behaviors and is particularly active during fearful events (LeDoux,2000). The amygdala has also been shown to play a role in primates in gaze monitoring, the ability to perceive where someone's eyes are focused and what they are looking at (Perrett et al.,1989; Rolls,1999) and may determine whether or not someone is trustworthy and whether or not social approach should take place (Adolphs,1999). Changes in the function and neuropathology of the amygdala have long been known in autism (Baron-Cohen et al.,2000). Postmortem studies show cell packing changes (Bauman and Kemper,2005) and size changes in MRI studies have been observed (Aylward et al.,1999; Schumann et al.,2009) which can be correlated with the ability to engage in joint attention (Mosconi et al.,2009). fMRI activation when looking at faces is different in autism (Critchley et al.,2000).

Corticotrophin releasing hormone (CRF) cell bodies are found both in the PVN and the amygdala. Cells in the amygdala also project to the hypothalamus, further regulating the release of adrenocorticotrophin hormone (ACTH) and eventually of cortisol, through the HPA axis. CRF is therefore thought to be the principle initiator of endocrine, behavioral and autonomic responses to stress (Shekhar et al.,2005). Specifically, increased CRF in the amygdala leads to heightened vigilance and fear and social withdrawal (Erickson et al.,2005) and the number of corticotrophin-releasing hormone cells are greater in men than women (Bao and Swaab,2007).

There are numerous studies on the role of altered HPA axis in autism. Children with autism do not have the usual morning surge in cortisol but cortisol levels are increased more in children with autism than in typically developing children, when exposed to a novel event (Corbett et al.,2006,2008). This suggests a dysfunction of the HPA axis in autism.

Therefore, a balance exists between oxytocin and cortisol in interpreting social interactions as being supportive and pleasant (oxytocin) or stressful and to be avoided (cortisol) whereby oxytocin blunts the stress responses. This balance can be seen in several studies. Intranasal oxytocin decreases salivary levels of cortisol during conflict (Ditzen et al.,2009) or when subjects are exposed to a psychosocial stress (Heinrichs et al.,2003). In a study of young adults, higher levels of oxytocin were associated with less psychosocial stress but higher levels of cortisol were associated with higher levels of psychosocial distress (Gordon et al.,2008). Administration of oxytocin can also be shown to inhibit release of ACTH (Legros et al.,1982,1984; Suh et al.,1986). Oxytocin may play a general role in balancing the effects of cortisol to maintain homeostasis, as oxytocin decreases the cortisol elevation induced by experimental bacterial toxins (Clodi et al.,2008).

Thus, oxytocin plays a dual role in social responding by being both prosocial and at the same time anti-stress. The neuroendocrine hypothesis of autism proposes a dysregulation of this important cortisol/oxytocin balance.


Most species show sexually dimorphic behaviors, principally for propagation of the species, with the female gender being responsible for “tending and befriending” and the male for “defending.” These behavioral differences are brought about by the effects of gonadal (sex) hormones, at critical periods, of the developing brain. In many cases, the divergent adult behaviors can be attributed to the effects of the sexually dimorphic bonding hormone, oxytocin.

Human and Other Primates

Although human behaviors occur along a continuum, there are some behaviors which are more prominent in one gender than the other. For the purposes of understanding early brain development, only those behaviors evident in young children (i.e., before the hormonal changes of puberty make further activational changes to behavior) are relevant. Boys prefer toys such as cars and constructing kits to dolls, while girls show no preference. The same is observed in primates (Gerrianne et al.,2002; Wallen,2005; Hassett et al.,2008). Boys engage more often in rough and tumble play while girls are more interested in infants (Campbell et al.,2002)—these preferences are also observed in primates (Lovejoy and Wallen,1988). Infant girls develop language skills earlier than boys (Lung et al.,2009) and infant monkeys also show differences in calling behaviors, whereby female infants use more cooing sounds while male use sharper noisy screams. In either case, the women are more vocal (Tomaszycki et al.,2001). Young girls show more prosocial behaviors than boys (including, sharing and empathetic concern) (Litvack-Miller et al.,1997).

The primary sexually differentiating hormone in the human brain is testosterone (Swaab,2007). Testosterone has an effect in the human brain, both as an organizing factor (that is organizing the developing brain) and an activating factor, once the brain is developed (McCarthy,2009). There are two peaks of testosterone levels associated with the organizational or brain developmental effects—one at midgestation and another in the first few weeks of life (Hines,2008). The testosterone is produced by the testes of the male fetus, beginning at 8 weeks of gestation, and thus lack of testes and testosterone in the female fetus is in effect a default setting and masculinization of the brain does not take place. The effects of testosterone are produced via the androgen receptor which is a nuclear transcription factor. Activation of this receptor can lead to a number of changes in gene transcription, ultimately increasing or decreasing cell survival, as well as determining synaptogenesis and neurotransmitter or neuropeptide content. Sexual dimorphism may not be simply related to testosterone, but regional differences may also be regulated by the presence or absence of the androgen receptor (Lavranos et al.,2006). Those regions of human brain shown to be most sexually dimorphic are also those regions thought to have the highest numbers of sex steroid receptors (Quadros et al.,2002).

Much of our information on the developmental effects of testosterone in humans is based on studies of a genetic condition, congenital adrenal hyperplasia (CAH), in girls. In this condition, testosterone is produced in high amounts in utero due to lack of an enzyme involved in cortisol production. As these children develop, they show more male-typical behavior, including more rough and tumble play (Hines and Kaufman,1994), more preference for male toys and less empathy or interest in infants (Leveroni and Berenbaum,1998). In women who have had CAH, the amygdala responds in a male manner to faces (Ernst et al.,2007). In addition to these studies, more recent work has shown correlations between gestational amniotic fluid levels of testosterone and subsequent behaviors. Again, these studies show that higher levels of midgestation testosterone are associated with more masculine behaviors and lower measures of empathy and vocabulary development (Auyeung et al.,2009). Thus, it is apparently the midgestational levels of testosterone which may most predict the behaviors associated with the extreme male hypothesis of autism.

The length of the index finger to the ring finger (the 2D:4D ratio) has been proposed to be indicative of the amount of testosterone in utero, with lower ratios indicating more testosterone, although this measurement is not without controversy (for an excellent review, see McIntyre,2006). Children with autism have a lower ratio of the length of their second finger to their fourth finger (Manning et al.,2001; Noipayak,2009). A study of androgen blood levels, showed significant increases in boys with autism (Geier and Geier,2007) and men with autism typically reach puberty earlier than normal men (Tordjman et al.,1997).

Clearly humans and non-human primates show that manipulations of testosterone during gestation can cause differences in sexually dimorphic behaviors, many of which may be related to the extreme male phenotype of autism, Research on autism could be greatly advanced by an animal model and in particular in a rodent species. The current work discusses preliminary findings and the development of such a model.

Preliminary Findings in an Animal Model

Before developing an animal model to test the extreme male hypothesis and its effect on neuroendocrine systems, two factors must be considered—critical periods and hormonal factors—which differ between humans and rodents.

Critical periods.

Although there are a number of schema for adapting rodent development to model human development, for the purposes of early brain development, we use the proposals by Bayer et al. (1993) based on timetables of neurogenesis. Using this scheme, midgestation of human development corresponds to approximately postnatal day (PND) 1 and 2 in the rat.

Hormonal factors.

Although in humans, testosterone is the principal hormone directing sexual differentiation, with little effects of estrogens, it appears to be opposite in rodent species, where the conversion of testosterone to estrogen (specifically 17-β estradiol) by the enzyme P450 aromatase is largely responsible for sexual differentiation (Swaab,2007). Thus, any brain region in a developing male rat fetus which has both aromatase and estrogen receptors would be sexually differentiated resulting in a sexually dimorphic brain region and sexually dimorphic behaviors. (McCarthy,2008). In our model, we have chosen to use the estrogen, 17-β estradiol and the doses chosen are comparable to those used in other studies of sexual differentiation (Bartholomeusz et al.,1999; Patisaul et al.,2006).

Using this approach, we have treated rat pups with 17-β estradiol at a dose of 5 or 50 μg/per pup; twice daily on PND 1 and 2. In these preliminary studies, we have looked at two early social behaviors, huddling (Fig. 1) and proximity to dam (Fig. 2) as well as examining the number and morphology of oxytocin-containing cells of the PVN (Fig. 3) to determine whether or not a model of extreme male differentiation can induce the social and neuroendocrine changes associated with autism. Our results are quite promising.

Figure 1.

This figure represents Huddling behavior on PND 13. Each time point represents the huddling area of the litter at that time. The 5 μg, 50 μg, and vehicle control groups contained 11, 12, and 12 pups, respectively. Huddling area was recorded over a 15 min time interval, during which huddling area was measured every 60 sec. Temperature was sustained at ∼ 32°C to negate the effects of temperature control on huddling behavior. The huddling area was significantly different between groups. F (2, 42) 9.29, *P < 0.01.

Figure 2.

Proximity to Dam on PND 18. The columns represent the mean time pups from each group spent within a body length of the divider that separated them from their mother. All animals from the 5 μg (seven male and four female), 50 μg (six male and six female), and vehicle control (six male and six female) groups were tested. The analysis of variance showed that doses of estradiol on PND 1 and 2 had a significant effect on the amount of time male pups spent near their mother but only the high dose affected the women on PND 18. Overall, F (5, 29) = 3.774, P* < 0.01.

Figure 3.

Number of oxytocin-IR cells in the paraventricular nucleus: Sections were analyzed in groups (numbered 1–5) based on the rostal-caudal scheme of Anderson et al. (1995) with at least six sections counted. All animals from the 5 μg (seven men and four women), 50 μg (six men and six women), and vehicle control (six men and six women) groups were analyzed. Statistical analysis of male oxytocin-IR cells across the sections are significantly different overall F (13, 96) = 61.198, P* < 0.01. Statistical analysis of female oxytocin-IR cells across the sections are also significantly different overall F (13, 63) = 61.198, P* < 0.01. The asterisks denote significant differences between treated animals and controls within that section. P < 0.05.

Social behaviors.

The animals were tested in two early measures of bonding, one to littermates (huddling at PND 13—Fig. 1) and one to the dam (proximity at PND 18—Fig. 2). Littermates, introduced into an open space together in one confined group, stay huddled together as closely as possible, and the total area of huddles can be used as a measure of this behavior, wherein, the less huddling occurring, the greater the area. In the paradigm used here, as proposed by Alberts et al. (2007), the treated animals dispersed more than control litters, and thus had a significantly greater area. This paradigm has been shown to be highly dependent on oxytocin, whereby the more oxytocin, the greater the litter huddling (i.e., the lower the area of huddles). Our results indicate that the pups are less bonded to each other and predict there will be oxytocin changes as well. Social bonding behaviors are considered to be plastic and imprintable—that is later behaviors build on earlier behaviors. It is therefore not surprising that at a later age (PND 18) the animals also show less bonding to their dams, as evidenced by the amount of time spent close to her (Fig. 2). In this case, the pups could be studied by gender, and significantly different results were found by gender. At both doses, the male pups spent significantly less time in close proximity to the dam, whereas the female pups only showed this effect at the high dose. Again, this suggests that the developing female brain is less vulnerable to conditions which would set in motion changes leading to the behaviors associated with the extreme male phenotype of autism.

Oxytocin immunochemistry.

Using an antibody specific for oxytocin, we have visualized and counted the oxytocin cells of the PVN at PND 32 (approximating a young adult). As in the human, in the rat, oxytocin is a sexually dimorphic neuropeptide associated with social recognition and affiliative (or pro-social) behaviors (Winslow and Insel,2002; Winslow and Insel,2004). In animal studies, the brain regions with the most sexual differentiation showed the greatest differences in oxytocin receptors (Uhl-Bronner et al.,2005) suggesting that oxytocin is highly important in regulating sexually dimorphic behaviors. The number of oxytocin-containing cells in the PVN were indeed found to be sexually dimorphic in response to the 17-β estradiol treatment, with both region and dose effects (Fig. 3).

In male rats, the region of the PVN from which oxytocin-containing cell projects primarily to brain regions such as the amygdala and brainstem nuclei (designated section 3) shows a decrease in oxytocin-ir cells in response to the high dose of 17-β estradiol, while the female rats showed a significant increase in cell number to the low dose, and then a loss of cells in response to the high dose. Dose effects such as these have been reported before, after developmental estradiol treatment (McCarthy,2008). More interestingly, the effects of treatment in the male rats seem to indicate a re-distribution of oxytocin-ir cells, wherein the cells are distributed in greater numbers in more caudal regions. This effect was not observed in female rats. This observation is strengthened by the report of Tobet et al. (2009) which shows sexual differentiation includes migrational changes within the PVN. A representative photomicrograph is given in Fig. 4.

Figure 4.

Oxytocin-IR cells in the paraventricular nucleus of the hypothalamus (PVN). Picture A represents the control group. Picture B represents the high dose group which received 50 μg 17-B estradiol two times per day on post natal days 1 and 2. As seen in the pictures and in the previous graph, the high dose group has less cells than the control group in a particular region of the PVN called the cap. The measurement bar represents 50 μm.


Studies reviewed in this article support the hypothesis that altered sexual differentiation of the brain (into an “extreme male” phenotype) leads to the symptoms of autism. Other studies reviewed support a neuroendocrine hypothesis for autism functioning. Our hypothesis is that these two apparently distinct hypotheses are in fact compatible—extreme male differentiation of the brain causes the neuroendocrine changes, whereby there is a dysfunction of the oxytocin/cortisol balance.

We are currently examining this hypothesis in an animal model. By using 17-β estradiol administration during a critical time, we have essentially produced extreme male differentiation of the brain. As predicted by our hypothesis, this resulted in lowered oxytocin content and poorer social bonding with the dam and with littermates. In ongoing studies, we are examining the effects of this treatment on CRF expression and on morphology of the amygdala. We are expanding the behavioral tests, to include social bonding at older ages and other behaviors associated with the autism spectrum, including “need for sameness” and anxiety and we will use various critical periods for treatment, to see if the spectrum of autistic behaviors can be dependent on extreme sexual differentiation at different stages of development. Finally, to more fully test an animal model, we will use treatments with oxytocin to reduce the behavioral deficits.