The causes of autism are a subject of scientific research, but understanding of the etiology of autism is incomplete. It is influenced by a complex interplay of genetic, epigenetic, prenatal, perinatal, and environmental factors. Genetics play a major role, with heritability estimates ranging from 60–90%. De novo mutations—including copy number variations and gene-disrupting mutations—contribute to approximately 30–40% of cases. However, most autism cases involve complex interactions among multiple inherited genetic variants, many of which are still unknown.
Epigenetic mechanisms, such as DNA methylation and genomic imprinting, also influence gene expression and susceptibility, often interacting with environmental exposures. In rare cases, autism has been associated with agents that cause birth defects.
Prenatal risk factors include advanced parental age, maternal metabolic or autoimmune disorders, infections, and prenatal stress, while perinatal risks involve preterm birth, low birth weight, and birth complications. Postnatal mechanisms have been proposed, including immune dysregulation, gastrointestinal abnormalities, oxidative stress, and neural circuit differences, though these remain largely unproven and are the focus of ongoing research. Some Neanderthal-derived genetic variants may influence susceptibility. Current high-quality evidence shows no causal link between prenatal use of paracetamol and autism.
Overwhelming scientific evidence shows no causal association between thiomersal and vaccines, including the measles-mumps-rubella (MMR) vaccine, and autism. Parental concern about vaccines has led to a decreasing uptake of childhood immunizations and an increasing likelihood of measles outbreaks. The refrigerator mother theory has been conclusively refuted by scientific evidence.
How causes are established The causes of autism have been a subject of scientific research, but understanding of its etiology remains incomplete. The scientific discipline of epidemiology has been the major tool to hypothesize and evaluate genetic, epigenetic, prenatal, perinatal, and environmental factors.
Genetics
Genetic factors may be the most significant cause of autism. Early studies of twins had estimated heritability to be over 90%, meaning that genetics explains over 90% of whether a child will develop autism. This may be an overestimation, as later twin studies estimate the heritability at between 60 and 90%. Evidence so far still suggests a strong genetic component, with one of the largest and most recent studies estimating the heritability at 83%. Many of the non-autistic co-twins had learning or social disabilities. For adult siblings, the probability of having one or more features constitutive of the broader autism phenotype may be as high as 30%.
In spite of the strong heritability, most cases of autism occur sporadically with no recent evidence of family history. It has been hypothesized that spontaneous de novo mutations in the sperm or egg contribute to the likelihood of developing autism. Additionally, mutations of the Fragile X Messenger Ribonucleoprotein 1 (FMR1) which cause fragile X syndrome, one of the most common causes of intellectual disability and autism, have been linked to the early cessation of reproductive functions of female carriers in the gene. This substantiates the notion that those with autism are more likely to be infertile, weakening the heritability of the disorder. Also, the likelihood of having a child develop autism generally increases with advancing parental age, and mutations in sperm gradually accumulate throughout a man's life.
The first genes to be definitively shown to contribute to autism were found in the early 1990s by researchers looking at gender-specific forms of autism caused by mutations on the X chromosome. An expansion of the CGG trinucleotide repeat in the promoter of the gene FMR1 in boys causes fragile X syndrome, and at least 20% of boys with this mutation have behaviors consistent with autism spectrum disorder. Mutations that inactivate the gene MECP2 cause Rett syndrome, which is associated with autistic behaviors in girls, and in boys the mutation is embryonic lethal.
Besides these early examples, the role of de novo mutations in autism first became evident when DNA microarray technologies reached sufficient resolution to allow the detection of copy number variation (CNV) in the human genome. CNVs are the most common type of structural variation in the genome, consisting of deletions and duplications of DNA that range in size from a kilobase to a few megabases. Microarray analysis has shown that de novo CNVs occur at a significantly higher rate in sporadic cases of autism as compared to the rate in their typically developing siblings and unrelated controls. A series of studies have shown that gene disrupting de novo CNVs occur approximately four times more frequently in autism than in controls and contribute to approximately 5–10% of cases. Based on these studies, there are predicted to be 130–234 autism-related CNV loci. The first whole genome sequencing study to comprehensively catalog de novo structural variation at a much higher resolution than DNA microarray studies has shown that the mutation rate is approximately 20% and not elevated in autism compared to sibling controls. Structural variants in individuals with autism are much larger and four times more likely to disrupt genes, mirroring findings from CNV studies.
CNV studies were closely followed by exome sequencing studies, which sequence the 1–2% of the genome that codes for proteins (the "exome"). These studies found that de novo gene inactivating mutations were observed in approximately 20% of individuals with autism, compared to 10% of unaffected siblings, suggesting the etiology of autism is driven by these mutations in around 10% of cases. There are predicted to be 350-450 genes that significantly increase susceptibility to autism when impacted by inactivating de novo mutations. A further 12% of cases are predicted to be caused by protein altering missense mutations that change an amino acid but do not inactivate a gene. Therefore, approximately 30% of individuals with autism have a spontaneous de novo large CNV that deletes or duplicates genes, or mutation that changes the amino acid code of an individual gene. A further 5–10% of cases have inherited structural variation at loci known to be associated with autism, and these known structural variants may arise de novo in the parents of affected children.
Tens of genes and CNVs have been definitively identified based on the observation of recurrent mutations in different individuals, and suggestive evidence has been found for over 100 others. The Simons Foundation Autism Research Initiative (SFARI) details the evidence for each genetic locus associated with autism.
These early gene and CNV findings have shown that the cognitive and behavioral features associated with each of the underlying mutations is variable. Each mutation is itself associated with a variety of clinical diagnoses, and can also be found in a small percentage of individuals with no clinical diagnosis. Thus the genetic disorders that comprise autism are not autism-specific. The mutations themselves are characterized by considerable variability in clinical outcome and typically only a subset of mutation carriers meet criteria for autism. This variable expressivity results in different individuals with the same mutation varying considerably in the severity of their observed particular trait.
The conclusion of these recent studies of de novo mutation is that the spectrum of autism is breaking up into quanta of individual disorders defined by genetics.
One gene that has been linked to autism is SHANK2. Mutations in this gene act in a dominant fashion and appear to cause hyperconnectivity between the neurons.
A study conducted on 42,607 autism cases has identified 60 new genes, five of which had a more moderate impact on autistic symptoms. The related gene variants were often inherited from the participant's parents.
Epigenetics Epigenetic mechanisms may increase the risk of autism. Epigenetic changes occur as a result not of DNA sequence changes but of chromosomal histone modification or modification of the DNA bases. Such modifications are known to be affected by environmental factors, including nutrition, drugs, and mental stress. Interest has been expressed in imprinted regions on chromosomes 15q and 7q.
Most data supports a polygenic, epistatic model, meaning that the disorder is caused by two or more genes and that those genes are interacting in a complex manner. Several genes, between two and fifteen in number, have been identified and could potentially contribute to disease susceptibility. An exact determination of the cause of ASD has yet to be discovered and there probably is not one single genetic cause of any particular set of disorders, leading many researchers to believe that epigenetic mechanisms, such as genomic imprinting or epimutations, may play a major role.
Epigenetic mechanisms can contribute to disease phenotypes. Epigenetic modifications include DNA cytosine methylation and post-translational modifications to histones. These mechanisms contribute to regulating gene expression without changing the sequence of the DNA and may be influenced by exposure to environmental factors and may be heritable from parents. Rett syndrome and fragile X syndrome (FXS) are single gene disorders related to autism with overlapping symptoms that include deficient neurological development, impaired language and communication, difficulties in social interactions, and stereotyped hand gestures. It is not uncommon for a patient to be diagnosed with both autism and Rett syndrome or FXS. Epigenetic regulatory mechanisms play the central role in pathogenesis of these two disorders.
Genomic imprinting may also contribute to the development of autism. Genomic imprinting is another example of epigenetic regulation of gene expression. In this instance, the epigenetic modification(s) causes the offspring to express the maternal copy of a gene or the paternal copy of a gene, but not both. The imprinted gene is silenced through epigenetic mechanisms. Candidate genes and susceptibility alleles for autism are identified using a combination of techniques, including genome-wide and targeted analyses of allele sharing in sib-pairs, using association studies and transmission disequilibrium testing (TDT) of functional or positional candidate genes and examination of novel and recurrent cytogenetic aberrations. Results from numerous studies have identified several genomic regions known to be subject to imprinting, candidate genes, and gene-environment interactions. Particularly, chromosomes 15q and 7q appear to be epigenetic hotspots in contributing to autism. Also, genes on the X chromosome may play an important role, as in Rett syndrome.
An important basis for autism causation is also the over- or underproduction of brain permanent cells (neurons, oligodendrocytes, and astrocytes) by the neural precursor cells during fetal development.
Prenatal environment The development of autism is associated with several prenatal risk factors, including advanced age in either parent, diabetes, bleeding, and maternal use of antibiotics and psychiatric drugs during pregnancy. Autism has been linked to birth defect agents acting during the first eight weeks from conception, though these cases are rare. If the mother of the child is dealing with autoimmune conditions or disorders while pregnant, it may have an effect on the child's development of autism. All of these factors can cause inflammation or impair immune signaling in one way or another.
Obstructive sleep apnea in pregnancy