What Causes Severe Autism? Genetic and Prenatal Factors

Severe autism arises from an accumulation of genetic disruptions and prenatal exposures, with no single cause explaining most cases. In people with the most significant support needs, large-effect genetic mutations play a bigger role than they do in autism broadly, but even then, the picture is rarely one gene, one outcome. Prenatal factors like maternal infection, metabolic conditions, and certain medications can layer onto genetic vulnerability to push development further off course. The interplay between these forces helps explain why “severe” autism varies so much from one person to the next.

What “Profound Autism” Actually Refers To

The term “profound autism” has gained traction as a way to distinguish the subset of autistic people who need around-the-clock support. Researchers generally classify individuals as having profound autism if they have co-occurring intellectual disability and minimal or no spoken language, requiring lifelong 24-hour care.¹ In epidemiological studies, that definition often translates to being nonverbal or minimally verbal, or having an IQ below 50.² This is not a separate diagnosis in official manuals but a descriptor meant to flag how different the daily reality is for this group compared to autistic people who live independently and hold jobs. The genetic and prenatal causes discussed here apply across the autism spectrum, but many of the highest-impact factors cluster in individuals with the most severe presentations.

De Novo Mutations and Their Outsized Role

Some of the strongest genetic contributions to severe autism come from mutations that appear for the first time in the child and are not inherited from either parent. These de novo mutations can be small single-letter changes in DNA or larger disruptions that knock out a gene entirely. A large sequencing study comparing affected children to their unaffected siblings estimated that about a fifth of autism diagnoses could be traced to de novo coding mutations, with gene-disrupting variants accounting for roughly 9% of cases and damaging missense variants contributing another 12%.³ These numbers are averages across all severity levels. In children with intellectual disability and limited language, the proportion attributable to a single de novo hit tends to be higher because a powerful enough mutation can disrupt brain development on its own.

The genes most frequently hit by de novo mutations tend to converge on a few biological processes: how synapses form and communicate, how chromatin (the packaging around DNA) is remodeled, and how early brain cells migrate to the right positions during fetal development. When those processes are disrupted early and severely, the downstream effects on cognition and language can be profound.

Copy Number Variations

Beyond single-letter mutations, chunks of the genome can be deleted or duplicated. These copy number variations, or CNVs, range from small enough that they carry only one gene to large enough to be visible on a standard chromosome test. CNVs are estimated to account for roughly 5 to 10% of autism cases, with certain well-known deletions and duplications (like those at chromosomal regions 16p11.2, 15q11-13, and 22q11.2) recurring across many unrelated families.⁴

A key finding from large CNV surveys is that these structural changes are not autism-specific. In a review of over 1,400 individuals referred for genetic testing because of autistic features, potentially causative CNVs were identified, but at most of these genetic locations, autism was only present in a fraction of carriers. The same deletions and duplications could show up as intellectual disability without autism, language delays, or other psychiatric conditions. Other genetic and environmental factors appear to influence which outcome a person ends up with.⁵ This variability matters for families: finding a CNV in a child does not always predict exactly how that child will develop.

Single-Gene Syndromes That Include Autism

A handful of well-characterized single-gene conditions carry high rates of autism as part of a broader syndrome. Fragile X syndrome, caused by an expansion in the FMR1 gene, is the most common inherited intellectual disability and one of the better-known genetic causes of autism. Tuberous sclerosis complex, Rett syndrome, and Phelan-McDermid syndrome (caused by deletions or mutations affecting the SHANK3 gene) are others. SHANK3 disruptions alone explain at least half a percent of all autism cases and remain underdiagnosed.⁶

These monogenic conditions tend to produce more severe autism phenotypes, including significant intellectual disability and limited speech. They also illustrate how environment can modify even a strong genetic cause. A case study of five brothers carrying the FMR1 mutation who were also exposed to alcohol prenatally found that the combination produced severe behavioral problems, including aggression, self-harm, and anxiety, that went beyond what Fragile X alone typically causes.⁷ Even a “monogenic” case is not purely genetic in practice.

How Common and Rare Variants Combine

For most autistic individuals, the genetic architecture is not one big mutation but a mixture of common inherited variants (each with a tiny effect) and occasional rare variants (each with a larger effect). Research comparing people who carry a known damaging rare variant with those who do not has found that common variant burden still matters on top of the rare hit. In other words, common and rare variants likely combine in an additive way to determine a person’s overall liability for autism.⁸

A separate large genetic analysis confirmed this picture from a different angle: de novo mutations, rare inherited variants, and polygenic scores were each independently associated with various dimensions of symptom severity. The phenotypic spectrum of autism maps onto a spectrum of genetic factors affecting different neurodevelopmental processes.⁹ This is part of why severe autism can emerge in a family with no prior history: a child can accumulate enough genetic risk through a new mutation layered on top of common background variants that neither parent had in sufficient quantity to be affected themselves.

Maternal Immune Activation During Pregnancy

Infections during pregnancy have been linked to increased autism risk in offspring, and the link appears to run through the mother’s immune response rather than any particular virus or bacterium. Animal models across mice, rats, and primates have consistently shown that triggering a maternal immune response during pregnancy alters fetal brain development, and the timing and intensity of the immune activation influences which neurodevelopmental outcome results.¹⁰

In humans, a study of autistic children found that those whose mothers experienced immune activation during pregnancy had measurably worse social-adaptive skills, independent of their overall autism severity or any pregnancy complications. The effect was specific to social functioning rather than communication or daily living skills, and structural equation modeling suggested a direct causal pathway between maternal immune activation and poorer social outcomes in the child.¹¹ This specificity is striking: it suggests that prenatal immune disruption does not just raise autism risk in general but may shape which aspects of the condition are most affected.

Maternal Metabolic Conditions

Pre-existing diabetes and obesity during pregnancy have both been associated with autism risk in the child, but the association is strongest when the two conditions overlap. A large study found that mothers with both obesity and pre-existing diabetes had nearly four times the risk of having a child with autism compared to mothers without either condition. Mothers with obesity and gestational diabetes had about three times the risk.¹² Separately, gestational diabetes diagnosed before 26 weeks of pregnancy was associated with roughly 50% higher autism risk in a large cohort, even after accounting for the mother’s weight and smoking status.¹³

The mechanisms likely involve chronic inflammation, elevated blood sugar crossing the placenta, and disrupted insulin signaling in the developing brain. One nuance: a Norwegian study found that maternal obesity alone was only weakly associated with autism risk, while paternal obesity showed an independent association with autistic disorder, suggesting the metabolic picture is not exclusively about the prenatal environment.¹⁴

Prenatal Medication Exposure

Valproate, a drug used to treat epilepsy and bipolar disorder, is the best-established medication-related prenatal risk factor for autism. A Danish population study of over 500 children exposed to valproate during pregnancy found an absolute autism risk of about 4.4%, with nearly three times the risk compared to unexposed children after adjusting for the mother’s epilepsy.¹⁵ A recent systematic review and meta-analysis confirmed the finding: valproate carried the largest risk among antiseizure medications, with nearly a threefold increase. Some other antiseizure drugs showed smaller but still statistically elevated risks, including oxcarbazepine (about 60% increase) and carbamazepine (about 20% increase).¹⁶

These findings have already changed clinical practice. Medical guidelines in most countries now recommend avoiding valproate in women of childbearing age unless no alternative is available, and pregnancy registries track outcomes closely. For women who took valproate during pregnancy without knowing the risks, the absolute numbers provide some reassurance: the vast majority of exposed children do not develop autism, even though the relative risk is significantly elevated.

Air Pollution and Prenatal Exposure

A growing body of research links prenatal exposure to fine particulate matter (PM2.5) with modestly increased autism risk. A systematic review and meta-analysis of early-life air pollution studies found a pooled odds ratio of about 1.22 for PM2.5 exposure and autism, meaning roughly a 22% increase in risk per standard increase in pollution levels.¹⁷ A Swedish cohort study found similar effect sizes and was able to break down the risk by pollution source, identifying vehicle exhaust and tire-and-brake wear as components with the most robust associations.¹⁸

More recent work has tried to identify which chemical components of PM2.5 drive the association. One large study found that after adjusting for specific pollutant components, particularly ammonium and sulfate particles, the overall PM2.5 association became non-significant, suggesting those components may be doing much of the work.¹⁹ The effect sizes for air pollution are small compared to, say, valproate or a de novo gene-disrupting mutation. But because exposure is so widespread, the population-level impact could still be meaningful.

Parental Age

Advanced paternal age has been consistently linked with autism risk, and one intuitive explanation is that older fathers accumulate more de novo mutations in their sperm over time. But the data suggest the story is more complicated. A study modeling the expected increase in autism risk from paternal-age-related new mutations found it could account for only about a 10% increase in risk for fathers aged 45 versus 25. Yet the epidemiologically observed risk increase was nearly an order of magnitude larger than what de novo mutations alone could explain.²⁰ Something beyond new mutations is driving most of the paternal age effect. Candidates include age-related epigenetic changes in sperm, shared genetic traits that lead some men to both have children later and carry autism-related variants, and social factors that correlate with older parenthood.

The Female Protective Effect

Autism is diagnosed three to four times more often in males than females, and genetic research has uncovered a biological explanation that goes beyond diagnostic bias. When autistic females are compared to autistic males, females consistently carry a higher burden of damaging mutations, including about 30% more loss-of-function and harmful missense de novo mutations.²¹ Females with neurodevelopmental disorders also carry roughly twice the number of large CNVs compared to males with similar conditions.²²

This pattern, called the female protective effect, means that female biology provides some buffer against autism-causing genetic variation. To reach the same clinical threshold, a girl needs to accumulate more genetic disruptions than a boy. The protection also appears to operate through common inherited variants: mothers of autistic children carry higher polygenic risk scores for autism than fathers do, suggesting mothers can silently transmit more common-variant risk without being affected themselves.²³ For severe autism specifically, this means that when girls do present with profound features, the underlying genetic burden is often especially heavy.

Placental Epigenetics as an Early Window

The placenta is both a barrier and a messenger between mother and fetus, and its epigenetic state may offer early clues about autism risk. Studies using whole-genome methylation analysis of placental tissue have identified regions where DNA methylation differs between children who later develop autism and those who do not. These differentially methylated regions cluster around genes involved in synapse formation, brain cell growth, and the structural scaffolding of neurons.²⁴ One study discovered a previously uncharacterized gene, NHIP, at a placental methylation hotspot on chromosome 22 that showed high expression in brain tissue and was underexpressed in both the placenta and brains of people with autism.²⁵

Epigenetic marks in the father’s sperm may also contribute. Researchers have identified hundreds of regions in sperm DNA where methylation patterns differ between fathers whose children later developed autism and those whose children did not. Some of these sperm methylation patterns were directionally consistent with methylation changes found in the cerebellum of autistic individuals, hinting at an epigenetic signal that persists from conception into postnatal brain tissue.²⁶ A follow-up study identified over 800 such regions in sperm and proposed them as a potential biomarker for paternal offspring susceptibility.²⁷ This work is still early-stage, but it points to a mechanism through which environmental exposures in a father’s lifetime could influence autism risk in his children without changing the DNA sequence itself.

Perinatal Complications and Oxygen Deprivation

Complications around birth, including preterm delivery, low birth weight, and oxygen deprivation, have been linked to autism risk, though disentangling cause from correlation is difficult. Being small for gestational age has been associated with higher autism risk in preterm infants, while being large for gestational age appears to confer some protection.²⁸ Separately, exposure to conditions that restrict blood flow through the placenta (uteroplacental hypoxia) has been associated with enlarged brain ventricles in autistic children, a structural marker that may reflect early injury or altered fluid dynamics in the developing brain.²⁹

Many of the same genetic variants that raise autism risk also increase the likelihood of pregnancy and birth complications, so it can be hard to know whether the complication contributed to the autism or both were downstream of the same genetic factor. For severe autism specifically, perinatal distress likely acts as an additional stressor on a brain already genetically vulnerable rather than a primary cause on its own.

What Genetic Testing Can and Cannot Reveal

When a child receives a diagnosis of severe autism, genetic testing is increasingly recommended. The diagnostic yield depends heavily on the type of test and the clinical features of the child. Chromosomal microarray analysis, which detects CNVs, identifies a cause in roughly 4 to 10% of children with autism overall, but the rate jumps to about 25% in children who also have unusual physical features or other medical conditions.³⁰ Whole exome sequencing, which reads the protein-coding portion of the genome, has yielded positive results in over 40% of children tested in some clinical series, though yield varies across centers and populations.³¹

Finding a genetic cause does not change the autism diagnosis itself, but it can open doors: it may identify a treatable co-occurring condition, connect the family with syndrome-specific support networks, clarify the likelihood of recurrence in future pregnancies, and sometimes guide medication choices. For the majority of children in whom no single genetic cause is found, the autism is presumed to arise from the combined effects of many common variants and prenatal exposures, a picture that current testing is not yet equipped to resolve into a clean answer.

The Maternal Gut and Fetal Brain Development

An emerging line of research focuses on the maternal gut microbiome as a potential intermediary between prenatal environment and autism risk. The hypothesis is that disruptions to the mother’s gut bacteria, whether from diet, stress, antibiotics, or metabolic illness, alter the chemical signals that cross the placenta and reach the developing fetal brain. These signals include short-chain fatty acids, neurotransmitter precursors, and inflammatory molecules. Animal studies have shown that transplanting gut bacteria from mothers with disrupted microbiomes into germ-free pregnant mice produces offspring with altered social behavior and brain chemistry.³² Whether this translates meaningfully to human autism risk remains an open question, but it represents one more layer of prenatal influence that researchers are actively tracking.