Genetic Causes of Cerebral Palsy

The single most important conceptual shift in this module is that CP is a description, not a diagnosis. The clinical definition — a group of permanent disorders of movement and posture caused by a non-progressive disturbance in the developing brain — says nothing about cause. It is a syndrome defined by what you see at the bedside (early-onset motor impairment that does not get worse) and explicitly leaves etiology open. For decades that openness was filled by a single assumption: that CP meant a perinatal injury, usually hypoxic-ischemia around the time of birth.

Why that assumption took hold — and why it is wrong. The birth-asphyxia model was attractive because it gave families and clinicians a discrete, blameable event, and because the most visible CP cases (term infants with neonatal encephalopathy) genuinely do follow such injuries. But epidemiology never supported it as the dominant story: large cohort studies have repeatedly shown that only a minority of CP is attributable to intrapartum hypoxia, and the historic push to attribute CP to delivery was driven as much by medicolegal pressure as by biology. The brain that is malformed for genetic reasons and the brain that is injured perinatally can look strikingly similar on a snapshot exam — both produce fixed, non-progressive motor disability — which is exactly why etiology was so easily mis-assigned.

Why genetics is now part of the definition of good CP care. When exome sequencing is applied to children carrying a CP label, a causative monogenic variant is found in a substantial fraction. A systematic review and meta-analysis put the pooled exome yield at roughly 23% and chromosomal microarray at ~5% (Srivastava et al. 2022); a large clinical-laboratory cohort reported a yield as high as 32.7%, against 10.5% in an unselected health-system cohort (Moreno-De-Luca et al. 2021). A 2026 meta-analysis put the yield at 19% in unselected cohorts and 44% in cryptogenic CP (Joo et al. 2026). The reason yields vary so much is selection: the higher numbers come from referral labs enriched for atypical cases, the lower numbers from unselected population samples — and that variation is itself the clinical lesson. The genetic contribution concentrates in children without a clear perinatal cause, although a history of perinatal asphyxia does not exclude one: in a Chinese exome cohort of 1,578 children, those labeled asphyxiated had a higher diagnostic rate (Wang et al. 2024), and it spans three tiers of genomic change:

• Chromosomal aneuploidies and structural rearrangements
• Pathogenic copy number variants (sub-microscopic deletions/duplications)
• Single-gene (monogenic) disorders

Reframing CP this way is not academic. A genetic etiology can redirect treatment toward something specific (sometimes curative), recalibrate recurrence risk for the family, and flag organ systems that need surveillance — none of which is possible while 'CP' is treated as a self-explanatory endpoint.

Chromosomal and copy-number causes occupy a particular niche in CP genetics: they are individually rare but collectively common, and they are detected by a test (chromosomal microarray) that is cheap, fast, and already familiar in developmental clinics. In the pooled meta-analysis, microarray identified a pathogenic CNV in about 5% of children with CP (Srivastava et al. 2022), lower than exome; because many laboratories now call CNVs from exome or genome data, a separate array is needed mainly when the sequencing test does not.

Why CNVs cause a CP-like picture at all. A deletion or duplication removes or adds a dose of many contiguous genes at once. When that interval contains genes governing neuronal migration, cortical patterning, or synaptic function, the result is a malformed or mis-wired brain that has been abnormal since fetal life — which presents exactly as early, non-progressive motor disability. The MRI in these children is often the tell: a 17p13.3 deletion (LIS1/PAFAH1B1, YWHAE) produces the smooth brain of lissencephaly/pachygyria in Miller-Dieker syndrome, and seeing that pattern reframes a 'severe spastic CP' as a defined microdeletion syndrome with its own prognosis and recurrence figures.

Why imprinting must be kept separate in your head. The 15q11-q13 region is the classic trap because the same physical lesion produces different diseases depending on parent of origin — a maternal deletion or paternal UPD yields Angelman; a paternal deletion or maternal UPD yields Prader-Willi; a maternal duplication yields the dup15q phenotype. Angelman in particular is one of the great CP mimics: an ataxic, severely speech-impaired, seizure-prone child is easily filed under 'ataxic CP.' This is why methylation testing matters: a SNP microarray shows isodisomic UPD as copy-neutral loss of heterozygosity but misses heterodisomy and imprinting-center defects, which are invisible to dosage-only analysis.

Why recognizing a chromosomal cause changes care. Beyond the precise diagnosis and recurrence-risk reset, many of these syndromes carry non-neurological risks — cardiomyopathy in 1p36 deletion, progressive respiratory disease in MECP2 duplication — so the label switches on a surveillance program that a generic 'CP' diagnosis would never trigger.

Single-gene disorders are where the stakes of the CP label are highest, because several of them are treatable — and the treatment window can close. A monogenic condition earns a 'CP mimic' designation when it produces motor impairment from infancy that looks static on any single visit, even though the underlying biology is anything but neutral.

Why mimics are so easy to miss. Clinical examinations are snapshots. A slowly progressive disorder, observed once, looks fixed. A relapsing-remitting disorder, observed between episodes, looks resolved. A treatable enzyme deficiency, observed before metabolic crisis, looks like ordinary spasticity. The non-progressive appearance that defines CP is therefore exactly the property that hides the dangerous diagnoses — which is why the mimics must be actively excluded rather than passively waited out. At least ten conditions recur often enough to memorize: DRD (GCH1), HSP (SPG4 and 80+ genes), AHC (ATP1A3), the leukodystrophies, Rett (MECP2), arginase deficiency (ARG1), glutaric aciduria type 1 (GCDH), Niemann-Pick C, mitochondrial disease, and treatable spinal cord pathology.

Why the levodopa trial is non-negotiable. Dopa-responsive dystonia is the archetype of a catastrophe-of-omission: a child with GCH1 deficiency has a near-normal brain starved of dopamine, and a low dose of levodopa can convert a wheelchair-bound 'dystonic CP' into a walking child. Because the trial is cheap, low-risk, and produces a response within days-to-weeks, the cost-benefit math is lopsided — the downside of trialing is trivial, the downside of not trialing is a lifetime of preventable disability. That asymmetry, not diagnostic certainty, is why an empiric levodopa trial is mandatory in any child with dystonia and a normal MRI.

Why metabolic mimics reward a simple blood draw. Arginase deficiency presents as a progressive spastic diplegia with markedly elevated plasma arginine and — unlike most urea cycle disorders — little or no hyperammonemia, so it slips past the usual 'sick neonate' filter and masquerades as worsening CP for years. Plasma amino acids cost little and can convert an untreatable label into a treatable diet.

The red-flag rule that ties it together: a course that is genuinely progressive or regressive is, by definition, not CP. The moment a clinician documents loss of skills or relentless worsening, the CP hypothesis must be abandoned and metabolic plus genomic workup pursued urgently — that single observation is the highest-value piece of data in the whole evaluation.

The workup for CP is really an exercise in conditional probability: each test is chosen not because it might find something but because the pre-test probability of a genetic cause is high enough to justify it. That is why the workup is tiered and why it begins with imaging rather than sequencing.

Why MRI comes first. A brain MRI does two jobs at once. It is abnormal in more than 80% of children with CP and shows the pattern — periventricular leukomalacia of prematurity, a middle-cerebral-artery infarct, a malformation — and, just as importantly, it risk-stratifies the genetic workup. A lesion that fully and plausibly explains the phenotype (classic PVL in a 28-weeker) lowers the genetic pre-test probability; a normal or non-lesional MRI raises it sharply. In CP genetics the normal MRI is not reassurance — it is a red flag that should escalate, not end, the investigation.

Why the workup is sequenced the way it is. After imaging, the order of testing tracks treatability and cost. Metabolic screening (plasma amino acids, urine organic acids, lactate, ammonia, acylcarnitines) comes early because it is cheap and can catch a treatable mimic before the genome report returns weeks later. A levodopa trial and, where indicated, CSF neurotransmitter studies are layered in for any dystonic or fluctuating presentation. Sequencing then follows: exome or genome (ideally trio — child plus both parents) with CNV calling, which covers the copy-number tier in one test; a separate chromosomal microarray is needed mainly where sequencing with CNV calling is not available. Trio analysis is what makes de novo variants — the dominant mechanism in sporadic CP — interpretable by showing they are absent in both parents.

Who to test hardest. The features that drive genetic yield up are consistent across studies: term birth, no documented hypoxic-ischemic event, a normal or malformation-type MRI, a positive family history, and any feature beyond pure motor impairment — epilepsy, regression, intellectual disability, or a movement disorder. A child accumulating these features approaches the higher diagnostic yields reported in selected cohorts. The frontier is whole-genome sequencing as a first-tier test, which adds structural variants, deep-intronic changes, and some repeat expansions that exome misses. Its yield, like exome's, depends on selection: 24.7% in an unselected Australian cohort, counting variants that cause or increase the risk of CP (van Eyk et al. 2021), and 12.8% in a UK trio pilot (Ratnaike et al. 2026). Because white matter injury can be invisible before myelination matures, a normal infant MRI should be repeated after about age 2 before it is read as a red flag.

A common objection is: if the motor disability is fixed and the rehab plan is the same, why does the genetic cause matter? The answer is that a genetic diagnosis acts on four things rehabilitation cannot touch — the recurrence risk, the comorbidity map, eligibility for cause-specific therapy, and the family's narrative — and it does all of this without removing a single rehabilitation service. The label adds; it does not subtract.

Why recurrence risk is the most concrete payoff. Families almost always ask 'will it happen again?', and only the mechanism can answer. A confirmed de novo variant carries a recurrence risk under ~1% (with a germline-mosaicism caveat that keeps it non-zero); an autosomal recessive cause carries 25% per pregnancy; a dominant variant inherited from an affected parent, 50%; X-linked risk depends on sex and carrier status. 'CP, cause unknown' offers families a vague empiric figure; a molecular diagnosis replaces it with a real number and the option of prenatal testing.

Why the diagnosis redraws the surveillance map. Generic CP follow-up watches the obvious things — tone, hips, feeding. A specific diagnosis adds organ-specific vigilance that would otherwise be missed: cardiac and thyroid screening in Down syndrome, epilepsy and scoliosis monitoring in Angelman, and crucially the progressive respiratory failure of MECP2 duplication, which changes how aggressively respiratory infections are managed. These are surveillance programs you cannot run without the name.

Why this is increasingly about therapy, not just labels. A growing subset of mimics are not merely explained but treated — levodopa for DRD, ketogenic diet for GLUT1 deficiency, gene therapy (eladocagene exuparvovec, FDA accelerated approval November 2024) for AADC deficiency, lysine-restricted diet to prevent the striatal crisis of GA1, biotin for biotinidase deficiency (which prevents further injury, though established hearing loss or optic atrophy may persist). Each is a case where the genetic diagnosis is the prerequisite for the cure.

A counseling caveat to hold honestly. Exome and genome testing also generate variants of uncertain significance in a meaningful fraction of cases; a VUS is not a diagnosis and must be counseled as provisional, re-reviewed as databases grow, and not allowed to either over-reassure or over-alarm a family. Treatment approaches for the motor disorder itself (baclofen, trihexyphenidyl, BoNT-A, ITB pump, SDR) are covered in the Genetic Dystonias module.