The Possible Role of Postnatal Biphasic Dysregulation of IGF-1 Tone in the Etiology of Idiopathic Autism Spectrum Disorder

1. Introduction

1.1. Background

Autism spectrum disorder (ASD) is a group of pervasive developmental and behavioural disabilities [1,2] with a steadily increasing burden on society and families [3]. ASD is diagnosed based on behavioural symptoms, but its manifestations are notoriously heterogeneous [2,4]. Despite high prevalence and increasing awareness, little is known about the etiology of ASD [5,6] and although numerous genetically defined syndromes display high comorbidity with autism, idiopathic ASD represents the majority of cases [7,8]. Despite clearly demonstrated familial risk and high heritability of idiopathic ASD [9,10], increasingly large-scale genetic studies with a concomitantly increasing number of candidate genes have so far failed to find mono- or oligogenic causes for ASD [8,11].

The disorder is typically diagnosed in toddlers or in early childhood [12], although first manifestations seem to occur around 6-12 months of age [13,14]. The diagnosis of autism is rather stable throughout childhood [15,16] and adolescence [17,18] suggesting a critical role for early brain maldevelopment in the emergence of autism symptoms [19,20]. Therefore, in addition to the heterogeneity in symptoms, studies of ASD are also hampered by timing of symptom onset, namely incidence during a period of intense brain development and maturation of the nervous system [21,22], requiring longitudinal observations and analysis often in at-risk populations placed in a developmental framework [19,23].

Recognizing the proposed early origin of the disorder, environmental, familial, and gestational risk factors have been investigated extensively, and several risk factors of idiopathic ASD have been identified [1,24]. Among familial and environmental factors, an affected sibling [9,25] or male sex [26,27] are widely documented to substantially increase the risk for autism. Furthermore, increased maternal and paternal age have also been demonstrated as risk factors [28,29]. Among gestational and maternal conditions, the most robust and consistently identified risk factors for ASD are conditions related to maternal hypertension, namely preeclampsia and gestational or chronic hypertension [24,30,31]; maternal use of antidepressants, in particular selective serotonin reuptake inhibitors (SSRIs) [24,32,33] and symptoms related to maternal metabolic disturbances like gestational diabetes [34,35,36,37] and pre-gestational overweight or obesity [24,38,39]. For perinatal factors, signs of fetal growth insufficiency, like preterm birth [35,40] or low or very low birth weight [35,41] are associated with ASD risk, but, interestingly, macrosomia (fetal overgrowth) also bears a slight risk for later ASD diagnosis [41].

Although ASD is typically not diagnosed before the age of 18 months, at the population level several growth and developmental deviations have been described to precede or correlate with later ASD diagnosis. Accelerated early growth has been demonstrated for body length [42,43], total brain volume [20,44,45,46], cortical surface area [47] and excess extra-axial fluid volume [48]. Importantly, following a period of overgrowth in toddlers, a slowing or even arrested growth has been reported for many of these parameters in early childhood [46,47,49,50].

In addition to that observed in anthropometric measures, a biphasic developmental trajectory was also found for brain structural connectivity [51]. In particular, two highly valuable longitudinal imaging studies, comparing at-risk infants later diagnosed with ASD to non-affected peers [52] or diagnosed toddlers to typical controls [53]), reported structural hyperconnectivity between brain regions at earlier scans, in contrast to hypoconnectivity at the later time points (24- ca. 36 months, depending on the study). In studies of head circumference direct comparisons found early overgrowth in ASD (and in pervasive developmental disorder not otherwise specified, in earlier studies), which was absent in patients with attention deficit hyperactivity disorder (ADHD) [54,55] or generalized or other developmental delays [43,56].

Despite robust data on well-defined prenatal risk factors and postnatal developmental alterations with substantial comorbidity with ASD, to date no theoretical framework exists for the etiology of this disruptive disorder which convincingly matches these observations. In recognition of the pervasive heterogeneity of the disorder, recently multiple etiologies have been considered to account for the diverse manifestations of the dysfunctions [57]. However, it has yet to be demonstrated that symptoms group according to suggested pathomechanisms or risk factors [4]. Moreover, multiple disjunct etiologies or genetic causation has to account for the increasing prevalence of ASD, with the disorder retaining its heterogeneity of manifestations, stable pattern of risk factors and characteristic developmental trajectory. Alternatively, instead of multiple etiologies or risk genes, a central pathomechanism could be considered in idiopathic ASD, a process that could inherently account for the heterogeneity of symptoms and be associated with known ASD risk factors. In the latter sections such a hypothesis of idiopathic ASD etiology is proposed and discussed in light of scientific observations on ASD.

1.2. Neuropathological Findings in ASD and a Neurodevelopmental Narrative

Despite occasional and specific peripheral comorbidities of ASD it is reasonable to assume that the brain is the most-affected organ in the disorder. Neuropathology studies of autism patients report a diverse variety of microscopic brain alterations [58,59,60,61]. The most consistent microscopic observation, demonstrated in about 75% of cases, is the decreased number of Purkinje cells (PCs) in the cerebellum [58,60] – typically in the form of dispersed and partial loss [62,63]. The origin of this cell loss was believed to be prenatal, suggested by the lack of retrograde cell loss in the inferior olive [64], however, the presence of excess cerebellar basket and stellar cells [65] and the lack of hypoplastic folia [60] suggests a relatively late timing of this event, and therefore precise time of atrophy could not be convincingly established yet (discussed in [60]). Late postnatal loss of PCs in autism has not been considered in the literature, due to the putative dependence of inferior olivary neuron survival on their PC targets [64]. However, two findings support the possibility of postnatal atrophy. First, the dependence of inferior olive neurons on their target PCs seems to weaken during postnatal development [66,67]. Namely, although retrograde inferior olivary numeric atrophy has been demonstrated in various mutant mouse strains with early PC loss, in strains with the latest postnatal onset of PC death (around postanal days 30 and 20 in leaner and nervous mice, respectively), little or no loss of inferior olivary neurons is reported [67,68]. Secondly, although retrograde atrophy of the inferior olive has been observed in human patients following acute cerebellar lesions [69], gradual and partial loss of PCs, in contrast, does not necessarily lead to detectable inferior olivary atrophy in adults [69,70]. Intriguingly, neuropathologic analysis of the youngest ASD specimens (3 to 4 years) revealed no PC loss [58,63,71], in line with the possibility of late postnatal atrophy. In summary, PC loss in ASD in latter neonatal development cannot be excluded as a possibility and might be even in line with pathological findings [65]. If PCs were present throughout regular development of the cerebellum, their loss might report on later changes in internal environment, like sudden loss of neurotrophic or survival factors, as PCs are known for their particular vulnerability to pathophysiological insults [72,73,74].

When searching for survival-promoting agents of PCs, insulin-like growth factor 1 (IGF-1) has been identified as a hormone having a pronounced and reproducible positive effect on PC survival [75,76] and development [77] both in vitro and in vivo [78,79,80]. The marked survival-promoting effect of IGF-1 on PCs seems specific in comparison to numerous other neurotrophic factors like nerve growth factor, brain-derived neurotrophic factor [81,82,83], ciliary neurotrophic factor [84], basic fibroblast growth factor, insulin and insulin-like growth factor 2 [75], or to contradictory effects of neurotrophin-3 [81,83] or glial cell line-derived neurotrophic factor [79,85]. Thus, although other survival-promoting factors might also play a role, PC loss in ASD might be a result of late postnatal attenuation in the neurotrophic tone of IGF-1. Strikingly, in two highly valuable clinical datasets, liquor IGF-1 level of younger ASD patients, ages 1.9 years to 5 years, was found to be extremely low and lower than that of age-matched nonaffected patients [86] or neurotypical controls [87], a difference not found for nerve growth factor [88], insulin-like growth factor 2 or IGF-1 in ASD patients above the age of 5 [87]. Several authors have even suggested the loss of IGF-1 tone as an underlying mechanism in ASD [89,90,91] and treatment with IGF-1 or analogues has been proposed for treatment of ASD or related disorders [91,92,93]. However, the hypothesis of IGF-1 deficit, unfortunately, does not account for the most consistent characteristics and risk factors of idiopathic ASD, such as early growth and connectivity anomalies, or various perinatal or familial risk factors. Furthermore, although IGF-1 regulates PC survival and development, IGF-1 knockout mice are not characterized by PC loss [94], which suggests a more complex dysregulation of IGF-1 in ASD.

1.3. IGF-1 and Its Dysregulation in Perinatal Complications

IGF-1 is a trophic factor, a mediator of growth via its major molecular target, the IGF-1 receptor, ubiquitously expressed in most tissues. Circulating IGF-1 is predominantly secreted from the liver but can be taken up by the brain, enabling modulation of central levels by circulating IGF-1 of liver origin [95,96]. IGF-1 is widely expressed also in other tissues [97] including the brain [76] and while paracrine action contributes to postnatal growth [98], liver-produced systemic IGF-1 also exerts substantial somatic growth-promoting effects [99,100], indicating that multiple sources of IGF-1 act jointly on tissue development and that endocrine IGF-1 also affects growth. Secretion of IGF-1 from the liver is predominantly regulated via the growth hormone releasing hormone (GHRH) – growth hormone (GH) – IGF-1 axis [101] but peculiarly, perinatal regulation of endocrine IGF-1 is different, creating a vulnerability through a specific constellation of perinatal factors associated with later ASD risk. Finally, unlike the closely related insulin, IGF-1 is bound to a family of carrier proteins (insulin-like growth factor-binding proteins, IGFBP), which can regulate IGF-1 availability and thus exert further control on IGF-1 function in vivo.

In the fetus, IGF-1 release from the liver is regulated by circulating insulin with little effect of GH [102,103,104], also mirrored in the close to normal birth weight of GH-deficient infants [105]. In newborns, circulating IGF-1 is a determinant of somatic growth [106,107,108] and is associated with insulin level [109,110] with no or negative correlation to GH or GH binding protein [111,112,113], in line with typically no signs of growth deficit in congenital GH deficiency prior to 6-12 months [105]. Regulation of circulating IGF-1 in infants thus gradually changes from insulin-dependent to GH-dependent throughout the first year of life [113,114].

As IGF-1 is a major regulator of fetal growth, two groups of gestational maternal complications associated with metabolic and growth alterations in the offspring can lead to its dysregulation. Fetal growth restriction or intrauterine growth restriction (IUGR) often leads to a neonate with birth weight lower than 10th percentile. These small for gestational age (SGA) infants are characterized by short stature, decreased serum or cord IGF-1 level [112,115,116], and elevated insulin sensitivity at birth [116,117]. Their majority undergo a period of compensatory accelerated early postnatal growth (catch-up growth) [118,119] and reach a more insulin-resistant metabolic state [117,120,121]. An opposite alteration from typical perinatal growth trajectory and metabolism is observed after macrosomia or fetal overgrowth. Macrosomic, or large for gestational age, newborns display hyperinsulinemia [122,123] and higher IGF-1 levels [124,125] in a more insulin resistant state [123,126,127] and often suffer decelerated early postnatal growth – catch-down growth [128,129,130]. Catch-up and catch-down growth take place mainly within the first 6-9 months [128,129], the time window of insulin-dependent IGF-1 secretion in the newborn. SGA infants experiencing catch-up growth have elevated circulating IGF-1 levels compared to those remaining short or light [131,132].

These specific perinatal growth anomalies are highlighted because their maternal risk factors show a striking overlap with those of idiopathic ASD discussed earlier. In one group of conditions, SGA offspring and low birth weight are associated with maternal SSRI and antidepressant use [133,134], and IUGR is highly comorbid with preeclampsia [135,136] and gestational hypertension [137,138]. Conversely, in a second group of gestational conditions, macrosomia is strongly associated with maternal metabolic disturbances like gestational diabetes [139], obesity [140] or higher than recommended gestational weight gain [141], driven by insulin resistance [142]. Importantly, as discussed above, both groups of maternal conditions are strongly connected with ASD risk in the offspring [24,31,37,39].

2. Hypothesis

2.1. A Possible Etiological Role of IGF-1 Dysregulation in Idiopathic ASD

The gestational risk factors of ASD described above thus form two groups associated with opposing dysregulation of fetal growth, neonatal insulin sensitivity, circulating neonatal or cord IGF-1 level, and infant growth trajectory. More importantly, timing of their postanatal consequences also overlap with the earliest developmental alterations preceding ASD in infancy. Based on these overlaps the novel hypothesis for idiopathic ASD presented here is that simultaneous occurrence of these counteracting conditions during gestation could lead to a complex dysregulation of perinatal insulin homeostasis and insulin-mediated IGF-1 action in the affected offspring. Prenatally, these co-occurring factors might partially neutralize each other despite misalignment of fetal insulin sensitivity, however, after birth, isolated from maternal circulation, asynchronous compensatory processes could lead to biphasic dysregulation of neonatal insulin sensitivity and IGF-1 tone. In the first phase, separated from the placental unit, prevailing elevated insulin action in tissues could result in higher IGF-1 secretion, leading to accelerated growth in a sensitive period of brain development in infancy. Next, by the age of 12-15 months, maturation of the GHRH-GH-IGF-1 axis and sharp negative feedback due to elevated IGF-1 levels could lead to an abrupt drop in GHRH-GH-IGF-1 tone, and concomitantly to arrest of IGF-1-dependent developmental processes and slowed maturation of the central nervous system. According to this hypothesis, thus, prenatal gestational anabolic disturbances result in a complexly dysregulated neonatal insulin-IGF-1 axis, leading first to elevated and later to decreased IGF-1 levels in affected infants leading to somatic and brain overgrowth and subsequent growth arrest and delayed maturation (summarized in Figure 1). This biphasic postnatal growth differs from that of catch-up growth in SGA or low birth weight infants in that the latter is characterized by in utero growth restriction or insufficiency leading to delayed growth up to birth, while cooccurrence of maternal insulin resistance would partly compensate IUGR, resulting in a generally well-developed newborn with high capacity for overdevelopment in a postnatal compensatory growth spurt.

Since somatic growth anomalies, observed only in a subset of ASD patients, supposedly do not form the etiologic factors of ASD, core symptoms should depend on developmental alterations in affected brain regions. Brain development and connectivity of ASD brains seem to show peculiar dynamics with early age [19,20,51]. Early connectivity abnormalities – mixed functional over- and underconnectivity – at 6 months were used to predict later ASD diagnosis with 100% specificity [143] and the alterations seem specific to ASD [144]. Additionally, brain overgrowth predominantly in cortical surface area by 6 and 12 months identified affected infants with high specificity (95%) [20]. As these developmental trajectories are distinct from those in global developmental delay [145] and ADHD [146] and as diagnosis stability of ASD is relatively high, it can be reasoned that this early period of brain overgrowth and probably overconnectivity is specific to autism and might form the pathomechanistic basis of idiopathic ASD [147].

A potential role of IGF-1 in ASD has been suggested in the literature and even IGF-1 administration has been considered as a treatment of ASD [89,90,91,148,149], but the hypothesis of early postnatal elevated central IGF-1 tone or its biphasic dysregulation has not been raised before. The above hypothesis thus represents a novel viewpoint of the possible neurodevelopmental dysregulation by IGF-1 in the appearence of idiopathic ASD.

2.2. Explanation of Observations and Risk Factors of ASD

IGF-1 is a pleiotropic modulator of growth [97] and brain development [150], therefore, dysregulation of IGF-1 tone affects a broad range of neurodevelopmental processes and therefore could account for the most pervasive observation in ASD: the puzzling diversity of brain developmental, pathophysiological, and pathological findings, as well as the resulting variety of behavioural or neurological symptoms. According to this view, it is not a single site, but most of the organism and the central nervous system affected by abnormal biphasic early overgrowth, and genetic and other factors could further shape the appearance of ASD. Since the highly arbored PCs of the cerebellum are known to display elevated vulnerability to pathophysiological conditions in vivo [72,74,151], their survival and development could be the most affected by altered central IGF-1 availability. Unprogrammed postnatal elevation of IGF-1 tone followed by an abrupt drop could lead to late and selective postnatal PC loss, without substantial atrophy of cerebellar basket cells, cerebellar stellar cells, and inferior olivary neurons. As cited above, the lack of PC loss in the youngest ASD cases subjected to neuropathological analysis is noteworthy, not contradicting the supposed PC loss throughout early childhood. This assumption is in line with the extremely low liquor levels of IGF-1 between 1.9 to 5 years in ASD patients [86,87]. Similarly, although peripheral and central IGF-1 do not necessarily correlate, lower serum level of IGF-1 was found in ASD patients of 2-3 years age compared to age-matched controls [152], and lower urinary secretion of IGF-1 has been reported in ASD in a very similar age range (2-5 years) [153]. Interestingly, in older pediatric subjects cerebrospinal fluid (CSF) levels of IGF-1 did not differ between affected and control subjects [87], while serum levels were reported varied in older children with ASD [152,154,155,156].

The timing of the suggested biphasic alteration in circulating IGF-1 level coincides well with the ASD-specific switch in growth rate compared to controls [51,157] in light of the observations that neonatal somatic growth correlated to circulating IGF-1 levels [107,158]. In infants with later ASD diagnosis, excess extraaxial volume [48] and brain volume [47] seem to stabilize after 12-18 months. Similarly, although imaging studies reported structural hyperconnectivity prior to 20 months, hypoconnectivity was observed in ASD patients by the age of 2-3 years [52,53]. Additionally, a particularly interesting analysis has found that white matter overgrowth in ASD is specific to myelinization processes starting within the first 4 months after birth [159]. Based on the well-described effect of IGF-1 on brain development [150], the timing of these brain myelination periods overlaps well with the proposed timing of early postnatal developmental dysregulations in ASD.

The most consistent non-gestational risk factors of ASD could also be explained by the concept of perinatal IGF-1 dysregulation. Boys bear approximately 3-fold risk of ASD incidence compared to girls [27]. Male infants are known to be born larger [160,161], have higher brain volume [162,163] and undergo slightly faster postnatal brain growth than females [21,164]. The IGF-1 level of newborn boys, in contrast, is reported as lower or similar to that of girls, both in serum [107,165,166,167] and CSF [168], and female infants are more insulin-resistant than boys [161,169]. Therefore, male infants, since they display higher growth rate with lower or similar IGF-1 levels and concomitant higher insulin sensitivity than females, might be more sensitive to insulin-regulated hypertrophy mediated by excess IGF-1 and its abrupt drop, and this amplification of IGF-1-mediated effects might be the cause of their susceptibility to ASD.

The single highest risk for idiopathic ASD is an affected sibling [1,25] with clear familial connection [10] despite lack of high penetrance single risk gene variants [170]. Genetic architecture clearly modifies the penetrance of neurodevelopmental disturbances and thus definitely exerts a contribution to appearence of ASD that should not be underestimated. Still, according to the hypothesis above, combined incidence of two groups of gestational conditions could be the specific etiologic trigger for idiopathic ASD, and therefore careful separation of genetic influence from the influence of the intrauterine environment is required. Contribution of genetic and environmental factors is most often assessed by analyzing concordance in monozygotic and dizygotic twins [10]. However as intrauterine growth restriction, as presented above, can be suspected in the etiology of ASD, such a heritability analysis has to take into account the difference in placental organization of dizygotic and monozygotic twins. Dizygotic twins namely are obligatory dichorionic, they possess separate placentas, while the majority of monozygotic twin pairs are monochorionic, sharing the same placenta. Placental dysfunction is a major contributor to fetal growth insufficiency and IUGR [171,172], therefore, monochorionicity might lead to overestimation of heritability as impaired placental function might not be separated from genetic overlap. It has to be noted that preeclampsia, the strongest maternal risk factor of ASD and a risk factor for IUGR, parallels idiopathic ASD in its familial aggregation with yet unclear genetic etiology [173].

The steeply increasing prevalence of ASD can be partly related to the increased awereness and diagnostic tolerance of the behavioral spectrum, however, a true increase in prevalence is nevertheless suggested [174,175]. According to the etiologic hypothesis presented, cooccurrence of maternal insulin resistance and risk factors of IUGR like chronic or gestational hypertonia are prerequisites for biphasic neonatal metabolic dysregulation. While prevalence trends in gestational hypertensive disorders in the last decades are reported as varied [176,177,178], prevalence of pregestational obesity and pregestational/gestational diabetes has been rising significantly since decades [179,180,181,182] in line with the steadily increasing ASD prevalence.

Finally, increasing maternal age, too, is unambiguously a risk factor for metabolic syndrome – by multiple definitions – in women of childbearing age in developed countries [183,184], and also for its various manifestations like gestational diabetes [179,185], overweight, or obesity [186]. Increasing maternal age as a risk factor for ASD could therefore indirectly transmit the effect of these gestational conditions.

2.3. Prospective Testing of the Hypothesis

The etiologic hypothesis presented here attempts to add points for consideration in order to improve our understanding of this pervasive disorder. Numerous epidemiologic studies have investigated the contribution of environmental, gestational, and maternal conditions to ASD risk. The principle proposed above is that of a combination of gestational conditions leading to opposing postnatal growth dysregulation (e.g. maternal metabolic disturbances in combination with preeclampsia, gestational hypertension or SSRI use). Therefore, investigation of combinations of these specific conditions in regard to ASD incidence risk would be interesting, suggested in general in [30]. Additionally, as signs of maternal metabolic syndrome and insulin resistance are risk factors of ASD [37] and as gestational diabetes bears a significant risk of later type 2 diabetes of the mother, the apparent risk posed by having an ASD offspring on later maternal type 2 diabetes is worth investigating. Moreover, as discussed above, heritability studies on ASD prevalence in monozygotic twins taking chorionicity into account could improve dissection of genetic and intrauterine environmental effects, further increasing our understanding of role of genetic influence. Higher concordance in monochorionic compared dichorionic twins would support a role of placental dysfunction in the latter incidence of ASD.

Finally, and perhaps most importantly, CSF IGF-1 level determination could be extended to at-risk neonates 3-6 months old in order to test the hypothesis of early elevated level of this neurotrophic factor as such an analysis has not been reported yet. The lower concentration of arginine vasopressin, but not oxytocin, in the CSF of 0-3 months old neonates in high concordance with idiopathic ASD diagnosis in a quasi-prospective re-evaluated random population sample [187] supports that neonatal liquor composition could show clear alteration in ASD from typical development at such an early age.

Limitations of the above hypothesis lie in the lack in incorporation of the paracrine effects of IGF-1 or the complex contribution of IGFBPs, and the role of genetic predisposition in affecting the incidence and the phenotype of emerging ASD, factors that could serve as subjects of further research.

3. Concluding Remarks

The most consistent maternal and gestational risk factors of idiopathic ASD overlap with two groups of gestational conditions affecting fetal growth through the insulin-IGF-1 axis. Asynchronous termination of these drivers after birth would lead to a biphasic postnatal dysregulation of IGF-1 tone in the infant, with the early insulin resistance-driven overgrowth in the first life year coinciding with the well-documented and specific accelerated development and growth in ASD. This excess IGF-1 spurt might lead to irreversible structural connectivity abnormalities specific to ASD followed by slowed maturation of neurocircuits, resulting in lasting network dysfunction. If this concept wins confirmation in at-risk infants, in addition to providing new research avenues, it could support effective early prevention initiatives for this burdensome condition.