Introduction
Biological research in autism has attempted to improve our understanding of the neurobiological mechanisms possibly involved in autistic disorder [AD]; studies have been conducted in domains as diverse as genetics, neurochemistry, neuropharmacology, neuroendocrinology, neuroanatomy, brain imaging, and neuroimmunology. For example, structural and functional imaging and neuropathological techniques applied to autism spectrum disorders [ASD] brains have revealed developmental macroscopic and microscopic abnormalities suggesting neuroinflammation in frontal cortex and cerebellar regions, including cytokine production and activation of microglia and astrocytes [1]. Studies stress increasingly that AD cannot be summed up or explained by a single biological factor, but rather by a multifactorial etiology. A multidisciplinary biological approach allows us to compare different fields and methodological processes, thus to understand better the neurobiology of autism. However, in spite of the numerous studies conducted on AD during the last decades, it appears that no etiological model, no biological or behavioral marker, and no specific psychopathological process have been clearly identified [negative or contradictory results, associations not replicated]. Although the genetic factors and the mode of transmission of AD are not yet fully determined, the underlying genetic architecture, such as known chromosomal rearrangements or single gene disorders, are being identified through, for instance, more and more routine chromosome microarray analysis [CMA] [2]. Thus, more than 200 autism susceptibility genes have been identified to date, and complex patterns of inheritance, such as oligogenic heterozygosity, appear to contribute to the etiopathogenesis of autism. Similarly, cytogenetic abnormalities have been reported for almost every chromosome [for a review, see Ref. [3–7] and //projects.tcag.ca/autism/]. Because of the lack of conclusive results and concensus, it is probably more appropriate to use the concept of syndrome to characterize autism. Autism is defined in the ICD-10 and DSM-5 as a delay or abnormal functioning with onset prior 3 years in social communication, and manifestation of restricted, repetitive and stereotyped patterns of behavior, interests, and activities.
Several authors support the hypothesis that the mechanism underlying autism etiology is most likely polygenic and potentially epistatic, and that environmental factors may interact with genetic factors to increase risk [8, 9]. Arguments for an environmental contribution to AD come from the growing number of studies on environmental factors in ASD, but also from the current lack of conclusive results on an etiopathological genetic model of autism. It seems important to reframe autism in a multifactorial context. Autism could be considered as a psychopathological organization that would result from the effects of diverse biological factors and/or psychological factors, including genetic factors, environmental factors, and gene × environment interactions. The environmental factors could be post- or prenatal [psychosocial environment but also cytoplasmic and uterine environment, with placental exchanges and hormonal effects].
First, we will examine arguments for a genetic contribution to AD based on updated family and twin studies. Then, after reviewing the possible prenatal, perinatal, and postnatal environmental risk factors for AD, we will discuss the hypotheses concerning the underlying mechanisms related to their role in the development of AD in association with genetic factors. In particular, the possible role of epigenetic mechanisms reported in genetic disorders associated with autism will be considered.
Genetic Architecture of Autism Risk
Several recent literature reviews underline the important role of genetics in the etiology of AD [10–13]. Much of the data come from family and twin studies. The concordance rate among monozygotic [MZ] twins ranges on average from 60 to 90%, and from 0 to 20% among dizygotic [DZ] twins. These rates depend on the diagnosis and on the subtype of autism considered. In addition, they are not sufficient to explain by themselves the autistic syndrome. Autism could be considered as a multifactorial hereditary disorder, in other words a disorder that depends on numerous genes [polygenic heredity] and environmental factors. Although genetic studies have identified hundreds of genes associated with ASD, the exact number remains unknown [10, 14]. The wide phenotypic variability of autism may reflect the interaction between genes and environment but also the interaction of multiple genes within an individual’s genome and the existence of distinct genes and gene combinations among those affected.
Family Studies
The prevalence of autism in the general population has been estimated in various ways that depend mainly on sampling methods and diagnostic criteria, as noted already many years ago in the report by Agence Nationale pour le Développement de l’Evaluation Médicale [ANDEM] [15]. Thus, the prevalence of autism varies according to the diagnostic criteria of Kanner, DSM-III, and DSM-IV classifications: from 1 to 5/10,000 according to Kanner or DSM-III criteria up to 20/10,000 according to DSM-IV-TR criteria [16]. Prevalence of parent-reported diagnosis of ASD among 3- to 17-year-old children in the USA reaches the very high rate of 1/91 [17]. Similar results are expected using the DSM-5 criteria given that the diagnosis of autism is only based on ASD in this classification [according to DSM-5 criteria, ASD includes two main domains of autistic behavioral impairments: social communication impairments and stereotyped behaviors or interests]. Broadening of the diagnostic criteria for autism and better recognition of the autism behavioral phenotype may explain this rising prevalence, but a true increase in incidence cannot be ruled out [Autism and Developmental Disabilities Monitoring Network, 2008; [18]]. However, Fisch [19, 20] showed clearly that this rising prevalence is related to the use of different diagnostic criteria. He concluded by saying “There is no autism epidemic but a research epidemic on autism.” Still, the reasons of such an increased interest in autism remain to be understood [for example, a better organization of association of parents, more funding contributing to an increase in the number of researchers and studies in autism, a growing interest for social communication impairments in a society promoting social communication networks].
It is noteworthy that there is a male prevalence in autism [about three to four times higher in males than in females [21]], which might also fit with greater social communication difficulties observed in males compared to females with typical development. Studies on the prevalence of autism in families with autistic children show a higher rate than in the general population. The concordance rate for siblings of individuals with autism of unknown cause ranges from 5 to 10% and approaches 35% in families with two or more affected children [22–25]. Taken together, the rates of AD in siblings of children with autism are on average 50–150 times higher than the rate of autism in the general population, which suggests that autism has a family feature [family meaning here environmental as much as genetic]. Carlier and Roubertoux [26] emphasized that in evaluating the risk, the degree of genetic proximity and the degree of environmental similarity were correlated. Only two studies have attempted to assess the presence of parental pathology at the same time as sibling pathology in the families of autistic individuals. The first was the Utah epidemiological survey [1989] and the second was Gillberg et al. [27] study. Ritvo et al. [25] reported that of the 214 parents seen in the Utah survey, 7 were autistic, the majority being fathers. In the epidemiologically based, case–control study by Gillberg et al. [27] four fathers of the 33 autistic probands were considered to have Asperger’s syndrome. This gives an overall prevalence of autism in parents between the two studies of 2.3%. As underscored by Todd and Hudziak [28] the presence of affected father–son pairs is not compatible with simple X-linked transmission.
Twin Studies
In the field of genetic research on AD, which compares MZ with DZ twins, three interesting results can be presented [see Table 1 for a summary of the results from the updated studies]:
• In each study, the concordance rate for MZ twins is higher than for DZ twins.
• The concordance rate of AD in MZ twins is incomplete, suggesting a contribution of environmental factors. Hallmayer et al. [9] underline in their twin studies the involvement of both genetic and environmental factors in the development of ASD.
• In the Folstein and Rutter [29], Bailey et al. [30], and Hallmayer et al. [9] studies, concordance rates vary according to the diagnosis: the concordance rates are higher for the broader autism phenotype than for AD [full diagnostic criteria].
Table 1. Pairwise concordance rates for autism in monozygotic twins [MZ] and dizygotic twins [DZ].
These results point to a possible etiological heterogeneity of autism. The etiology could be different according to the subtype of autism considered, a subtype that could be clinical as much as biological. This may help us to better understand why none of the genetics inheritance models proposed for autism, including the polygenic model, can fully explain the autism phenotype in the family and twin studies presented above. One of the current issues in the field of genetic research on AD is to work on different subtypes in order to identify the relevant genes. There are three main approaches to identifying genetic hotspots or chromosomal regions likely to contain relevant genes: [1] cytogenetic studies, [2] whole genome screens, and [3] evaluation of a priori selected candidate genes known to affect brain development or possibly involved in the pathogenesis of autism.
Genome-wide association studies [GWAS] examine associations between disease and genetic variants such as single-nucleotide polymorphisms [SNPs] or copy number variations [CNVs]. Genetic variants can be either inherited or caused [which is often the case] by de novo mutations. CNVs and SNPs have both been reported to play a major role in autism incidence [36–41]. Common SNPs acting additively have been reported as a major source of risk for ASD [42] with heritability exceeding 60% for ASD individuals from multiplex families and approximately 40% for simplex families. CNVs, including insertions, deletions, and repeated sequences, can be highly disruptive to developmentally regulated genes. Several CNV studies [36, 43, 44] identified also structural changes in DNA, which contribute to the risk for ASD. Recent findings suggest the possibility that not only single, but also aggregate molecular genetic risk factors, linked in particular to alterations in calcium-channel signaling, are shared between autism and four other psychiatric disorders [schizophrenia, attention-deficit hyperactivity disorder, bipolar disorder, and major depressive disorder] [45, 46]. However, the mechanisms underlying the role of these mutations in the development of ASD phenotypes remain to be ascertained. More generally, children with neurodevelopmental problems, including ASD, are often affected in more than one area of functioning of mental health to the extent that hierarchies of mutually excluding categorical diagnoses have to be considered as conflicting with scientific evidence [47]. It suggests, according to Anckarsäter [48], that genetic susceptibilities behind mental health problems have to be sought both in relation to specific problem types and to general dysfunction, using multivariate analyses with measures of all types of mental disorders.
Concerning candidate genes, several of them have been studied at chromosome regions 7q22–q33 or 15q11–q13, and variant alleles of the serotonin transporter gene at 17q11–q12 are more frequent in individuals with autism [see Ref. [12], for a review]. Linkage data from genome screens and animal models suggest also a possible role of the oxytocin receptor gene at 3p25–p26 [49]. Interestingly, the majority of the genes reported to be associated with autism is involved in various physiological processes, such as chromatin remodeling, metabolism, translation, and synaptogenesis. These genes may converge into pathways affecting distinct neuronal functions such as synaptic homeostasis. Such a genetic basis of synaptic and neuronal signaling dysfunction in ASD has been confirmed by recent findings [50] demonstrating differences in transcriptome organization between autistic and normal brain through gene co-expression network analysis.
Finally, it should be highlighted that the polygenic model does not exclude a role of environment. It is noteworthy that heritability [h2] is defined as h2 = GV/[GV + EV] where GV is the cumulative genetic variance and EV, the environmental variance [51]. The possible prenatal, perinatal, and postnatal environmental risk factors for ASD are presented below.
Prenatal, Perinatal, and Postnatal Environmental Risk Factors for ASD
The prenatal factors associated with autism risk in the meta-analysis provided by Gardener et al. [52] were advanced paternal and maternal age at birth, gestational diabetes, gestational bleeding, multiple birth, being first born compared to being third or after, and maternal birth abroad. In fact, several recent studies suggest that parental immigration, especially maternal immigration but also paternal immigration, is a risk factor for ASD [53–59]. This association between migration and autism is more particularly observed in male children of immigrant parents living in urban areas compared to rural areas [60]. In addition, concerning the prenatal risk factors for AD, a rare consistent association with AD is in utero exposure to two known teratogenic medications, thalidomide, and valproate [valproate is a broad-spectrum anticonvulsant drug used in seizures, bipolar disorder, or migraine headache], or the abortifactant misoprostol [7, 61–63]. Thus, children exposed to valproate in utero were seven times more likely to develop autism than those not exposed to antiepileptic drugs [61, 62]. A large population-based cohort study of all children born alive in Denmark from 1996 to 2006 was conducted on 655,615 children, including 508 prenatally exposed to valproate and 5437 identified with autism spectrum disorder [2067 with AD]. Children of women who used a high valproate dose [>750 mg/day] or a low valproate dose [