The genetic basis of cerebral palsy

Cerebral palsy (CP) is a major neurodevelopmental disorder, currently estimated to affect approximately 1 in 500 children. As a clinical diagnosis, the aetiology of the syndrome varies and is often unknown.1 Although prematurity, hypoxia–ischaemia, placental insufficiency, and prenatal infection are well-characterized causes of CP, for other patients, particularly those born at term and/or without a clear aetiology identifiable by magnetic resonance imaging (MRI), the cause of the condition has remained obscure.

Although there is a clear role for hypoxic–ischaemic injury in some cases of CP, estimates suggest that acute intrapartum hypoxia–ischaemia accounts for fewer than 10% of cases.2-4 Furthermore, despite international consensus criteria for defining perinatal asphyxia,5-7 many published studies have not applied these rigorous definitions. Despite improved obstetric practice and better antenatal and perinatal care, several studies indicate there has been little reduction in the incidence of CP over the last several decades.7-9 These data and other findings have led some investigators to suggest that ‘unknown pathophysiologic processes’ must be at work to account for a significant proportion of CP.10 We suspect that much of this unknown pathophysiology may be owing to genetic or epigenetic factors. Indeed, current estimates indicate that as many as 30% of CP cases may be genetic in nature.11-13 Regardless of the final numbers, there is a growing consensus (National Institutes of Health Workshop on Basic and Translational Research in Cerebral Palsy; https://videocast.nih.gov/summary.asp?Live=18384&bhcp=1) that genetic contributions to CP are significant and important for understanding the disorder. There are four main types of DNA variation that contribute to CP pathogenesis (Fig. 1), yet the ultimate effect of most mutations is a loss of the normal cellular function of the protein encoded by that gene.

Few studies to date have been published on the genetics of CP, given the historical focus on other aetiologies. Despite the criticism that CP represents a ‘wastebasket diagnosis’, we would argue that once one excludes CP mimics (Table 114) there is no better term to characterize ‘permanent disorders of … movement and posture’ that affect ‘the developing … brain’.15 As our knowledge continues to advance, it will likely be necessary to refine classifications of individual disorders based on their aetiology, but the concept of CP as a unifying and defining term remains useful. Similar to epilepsy or autism, which have also undergone a similar process of ‘lumping and splitting’,16-18 the term CP unifies a cardinal neurodevelopmental disorder in its own right.

As with autism and intellectual disability, genetic insights have the potential to provide a framework for understanding fundamental neurobiological pathways that lead to CP when they go awry. Interestingly, work to date on the autistic spectrum disorders suggest that much information can be gained from the study of individually rare mutations, which collectively lead to a relatively common disorder.16-18 Similarly, work to date has already clearly indicated that there is no single ‘CP gene’, and we suspect that hundreds of CP genes likely await discovery.

The effects of a given genetic mutation may vary depending on the nature of the mutation, the presence or absence of environmental insults, and the individual genomic context in which a mutation occurs. A severely deleterious mutation may lead to a major effect size, while a less damaging mutation that does not disrupt protein function as profoundly may lead to a smaller effect size. Accordingly, a highly damaging mutation may be sufficient in and of itself to cause CP in some individuals, whereas in other cases, the additive effects of a less damaging mutation coupled with an environment insult such as ischaemia may cross the threshold needed to produce neuromotor disability (Fig. 2). Finally, in some instances, a single gene alone will not lead to clinical CP, but the cumulative effect of several less deleterious mutations acting together in a polygenic fashion may lead to the disorder.

Genetic variation may be either protective or deleterious in a given situation. Many clinicians have cared for infants who suffered apparently severe insults yet enjoyed relatively good long-term outcomes. The converse is also true. This scenario is also encountered in animal models of CP. For example, in a hypoxic–ischaemic rabbit model produced by aortic ligation proximal to the uterine arteries, all the kits in a given litter are exposed to the same insult, yet consistently a subset develops permanent motor impairment and a subset does not.19 These findings implicate genetic and/or epigenetic modifiers that may influence the extent of injury and eventual motor outcome.

There is mounting evidence from other neurodevelopmental disorders that disruption to specific molecular pathways, particularly those involved in synaptic function, activity-dependent transcription and translation, cortical cytoarchitecture and circuitry, neuron–glia signalling, and neuro-inflammation can all lead to human disease and are amenable to in-depth study using animal models.18 We expect that in CP many of the same systems will be disrupted and that genetic models will similarly prove essential to vertical progress in the field.

So how can advances in genomics help us to better understand and care for patients with CP? Firstly, published genetic association studies have suggested DNA variants that could alter an individual's susceptibility to insults such as thrombosis or haemorrhage, which may influence motor outcome. These single-nucleotide polymorphisms represent important factors of relatively small effect size that, nevertheless, may contribute to the development of CP in a given individual exposed to an environmental risk. Susceptibility genes are discussed in the next section.

Secondly, evidence to date also suggests that, similar to other major neurodevelopmental disorders, duplications or deletions of portions of a chromosome (genomic copy number variants [CNVs]) may explain CP in about 10% to 20% of cases.20-22 CNV analysis may be diagnostically and prognostically valuable in clinical settings, and CNV studies of CP cohorts have already been instructive by implicating genes relevant to CP neurobiology within the genomic interval that has been duplicated or deleted. CNV contributions to CP are reviewed below.

Thirdly, advances in human-genome sequencing have led to the identification of single gene mutations of major effect size that may lead to CP on their own. Although individually rare, these monogenic contributors may collectively account for a large proportion of CP cases. Single genes that lead to CP when mutated are also discussed below.

Currently, there is no consensus, nor it is obvious which patients with CP are the most suitable for genetic studies. This complicates progress in the field based on studies of relatively small cohorts of patients, as these cohorts will be intrinsically heterogeneous. In addition, although some candidate CP genes have been identified, rigorous follow-up functional validation studies demonstrating that putative mutations actually impair gene function have been performed for only a handful of genes.23, 24 However, several important studies have been completed, others are underway, and interesting patterns are beginning to emerge.

Susceptibility genes

Candidate gene-association studies were initially used to identify genetic variations that were found with significantly higher frequency in CP cohorts compared with controls. Such studies were typically undertaken with the hypothesis that a relatively common single-nucleotide variant may lead to an increased risk for developing CP. However, a meta-analysis of studies assessing polymorphisms in apolipoprotein E (ApoE), methylenetetrahydrofolate reductase, coagulation factor II, coagulation factor V, coagulation factor VII, interleukin-6, endothelial nitric oxide synthase, fibrinogen β-polypeptide, plasminogen activator inhibitor 1, tumour necrosis factor-β lymphotoxin α precursor, adducin 1 (α), β-2 adrenergic receptor, and tumour necrosis factor-α in CP did not identify any variants that were strongly associated with CP.25 Correlations between ApoE alleles, thrombophilias, and CP have received the most study to date, yet these relationships remain tenuous.

CNVs in CP

Three recent studies have begun to delineate the role of genomic CNVs in CP.20-22 Although identifying CNVs can lead to new insights into the neurobiology of disease, significant challenges to interpretation exist. For instance, the rarity of individual CNVs consequently leads to infrequent overlap of the genomic regions that those CNVs span. Thus, it has been challenging to confirm a definite role for a given CNV in CP. Additionally, although de novo CNVs (genomic variants not found in mother or father but occurring in an affected child) are enriched in neurological disease cohorts, CNVs can also be inherited maternally or paternally from unaffected parents. Although this often leads to the conclusion that a given CNV is not relevant to the disease process, well-documented examples exist wherein variable expressivity, incomplete penetrance, higher genetic robustness of females versus males,38 or the presence of genetic/epigenetic modifiers or second hits instead explain why the parent does not manifest symptoms but their child does. Finally, most CNVs by their very nature encompass multiple genes in a genomic interval, making it difficult to draw firm conclusions about the relationship of a single gene product to disease. Nevertheless, studies to date have supported a strong role for CNVs in CP aetiology, and have implicated candidate genes in some cases.

McMichael et al.20 identified putatively deleterious CNVs affecting 20% of an unselected Australian CP cohort. No de novo deleterious CNVs were detected in this study. Both duplications and deletions were identified, and many patients had large or complex CNVs that made it challenging to identify likely ‘culprit’ genes. Nevertheless, candidate genes implicated by genomic variations in this manner included CTNND2, DAAM1, MCPH1 (biallelic mutations lead to primary microcephaly), SPG6/NIPA1 (known to lead to spastic paraplegia) and NIPA2, and MC2R.

Segel et al.21 studied Israeli patients with pyramidal and/or extrapyramidal forms of CP. They excluded patients with PVL in the context of prematurity (24–34wks’ gestation), hypoxic–ischaemic encephalopathy (defined as requiring resuscitation or Apgar score <7 at 5min), hemiplegia, encephalitis, head injury, or spinal cord lesions (including patients with spastic paraplegia or any spinal cord imaging abnormalities). The investigators identified putatively pathogenic CNVs in 19% of their cohort, and identified CNVs that they considered likely to be pathogenic in another 12%. De novo pathogenic CNVs were found in 13% of patients. The investigators implicated gene dosage alterations in KANK1 (originally identified by Lerer et al.39), SPG4/SPAST, WDR45, SPG34, and FLNA, among others.

Oskouio et al.22 studied an unselected Canadian CP cohort. The investigators found de novo deleterious CNVs in 7% of their patients and, in total, about 10% carried CNVs thought to account for their symptoms. They not only identified copy number abnormalities affecting KANK1, but also found dosage alterations in RAPGEF1, HSPA4, PARK2, and PACRG. Abnormalities of AGAP1 and TENM1 (earlier implicated by whole-exome sequencing40) were also detected.

Monogenic causes of CP

Over the last decade, several single-gene causes of CP have been identified, primarily using next-generation sequencing. Somewhat surprisingly, these monogenic causes of CP have not mapped to pathways controlling inflammation or thrombosis. If one views these genetic mutations as a glimpse into crucial neuronal pathways relevant to CP, an important role for neurodevelopmental processes has instead been highlighted.

Candidate CP genes

Our recent whole-exome sequencing study identified putative disease-causing mutations in 14% of an unselected CP cohort and also identified multiple novel CP candidate genes.40 This work also suggested that de novo mutations may significantly contribute to CP pathogenesis, as has been shown for related neurodevelopmental disorders, including intellectual disability, epilepsy, and autism. This work identified de novo heterozygous mutations in known Online Mendelian Inheritance in Man (OMIM) genes TUBA1A (OMIM #611603), SCN8A (OMIM #614558), and KDM5C (OMIM #300534). De novo heterozygous mutations were also identified in AGAP1, JHDM1D, MAST1, NAA35, RFX2, and WIPI2 as novel candidate CP genes. In addition, hemizygous X-linked variants in known disease-associated genes L1CAM (OMIM #307000) and PAK3 (OMIM #300558), as well as novel genes CD99L2 and TENM1, were inherited from an unaffected mother.

Distinction from hereditary spastic paraplegia

Clinically, it can be difficult to distinguish CP from hereditary spastic paraplegia. In principle, hereditary spastic paraplegia affects the lower limbs and is familial and progressive, while CP may be associated with quadriplegia and is nonprogressive. However, these distinctions are readily blurred in clinical practice, as CP may be relatively limited to the lower extremities and, as a neurodevelopmental disorder, its manifestations change over time. Further complicating the issue is the designation of mutations in AP4M1 as a cause of both CP (OMIM #603513) and hereditary spastic paraplegia (OMIM #614066) (Table 2). In the future, we suspect that distinctions between genetic forms of CP and hereditary spastic paraplegia will become increasingly blurred, particularly as one considers ‘complicated’ forms of hereditary spastic paraplegia. We recommend that clinicians and researchers not be limited by nomenclature, but rather to think broadly and consider cases that highlight ‘grey areas’ as opportunities to revise and refine diagnostic categories.

Developing new models for CP

Current animal models for CP are largely limited to hypoxia–ischaemia or inflammation-based rodent or large-animal models with neuronal development analogous to humans. Such animal models have led to many advances, particularly in perinatal and neonatal care.53-55 However, moving forward, as we define what are likely to be many additional CP-related genes, vertebrate and invertebrate models such as zebrafish, Drosophila, or C. elegans may be valuable in understanding the effects of mutations on the developing nervous system. An advantage of these lower eukaryotes is that they are highly genetically tractable, and models can thus be generated much faster and for less cost than mouse or large-animal models. Such systems can thus be used as supportive ‘proof of pathogenesis’ for a given gene, and many genes can be tested in parallel in a higher throughput than is currently possible for models such as mice.

Worms, flies, and zebrafish are widely used to study other human neurodevelopmental and neurodegenerative diseases, and a large tool kit of methods has been developed to exploit the unique advantages of these models.56-58 About 70% of all human genes have orthologues in zebrafish, while 65% of human proteins linked to diseases have similar counterparts in Drosophila.57, 59 Genetic conservation is especially true for proteins required for essential cellular functions like the regulation of actin dynamics and maintaining cytoskeleton integrity60, 61; several such proteins have already been linked to CP. In particular, an intact cytoskeleton is crucial to integrate sensory input with motor output, and mutations that disrupt the actin cytoskeleton interfere with learning and memory, locomotion, and neuronal integrity in zebrafish and flies.62, 63 We have previously shown that disruption of Drosophila Hts, the orthologue of human adducin, leads to altered brain development and locomotor abnormalities.23

These comparatively simple model organisms can thus provide a powerful strategy for investigating how disease-associated changes in human proteins interfere with normal function and cause pathology. For example, knocking down Kank expression in flies leads to motor dysfunction (unpublished results). Identifying how mutations in CP candidate genes interfere with normal function in vivo can thus increase our understanding of CP pathology and may pinpoint targets for the development of new therapeutics which can then be tested in higher animals.

Conclusions

Available evidence indicates that genetic mutations may be responsible for a substantial proportion of CP cases. Clinically, microarray analysis in individuals with unexplained forms of CP may yield a unifying diagnosis, and, in select cases, whole-exome sequencing may be informative. Nevertheless, our understanding of genes that lead to CP is in its early stages, and the existing nomenclature is inadequate. As progress continues, partnerships between physicians, basic scientists, and families and advocacy networks will become increasingly important. Clinicians will need to be meticulous in phenotyping patients, while researchers will need to be similarly thorough in characterizing genotype–phenotype relationships as novel genes are discovered in order to inform clinical care.

Eventually, specific molecular subtypes of CP may be found to respond better or worse to specific treatments, facilitating a personalized medicine approach to CP. Similarly, it may be worthwhile targeting those with a genetic susceptibility to CP who suffer an environmental ‘second hit’; these children might be appropriately targeted for intensive early intervention to prevent poor outcomes. Finally, clinical studies of CP interventions could benefit from incorporating genetic analysis into their study design, allowing clinical investigators to compare, potentially, gene variants in ‘responders’ with ‘nonresponders’.

The discovery of monogenic contributors to CP may have substantial near- and long-term benefits. The identification of rare Mendelian forms of CP may provide a window into CP neurobiology, as has occurred for other neurological diseases such as Alzheimer and Parkinson diseases.64, 65 Of potential immediate impact would be the diagnosis of a genetic form of CP in a given patient. Such insight has the potential to provide a sense of closure for families. A genetic diagnosis may also be invaluable for guiding preventative health care, treating disease-specific clinical manifestations, and for accurate counselling regarding recurrence risk.

Confirming rare genomic variants as contributors to CP will require diligent validation studies. Most immediately, validations may be possible by sequencing large cohorts of individuals with CP; if recurrent mutations in a given gene are found in a cohort of patients with CP but not seen in healthy individuals, this provides an important genomic validation. Subsequent studies may utilize a combination of transcriptomic, proteomic, and/or metabolomic analysis of patient samples to characterize evidence for disruption of important pathways at multiple levels. Both established cell lines and patient-derived cell lines, including induced pluripotent stem cell-derived neurons, will be useful to demonstrate that a given genetic variant is deleterious. Developing and employing small-animal models, such as Drosophila, zebrafish, and mice will be crucial as the first steps to translate findings to and from humans. Large-animal models will be a crucial part of therapeutic development. In sum, we believe that genomic discoveries will set the stage for follow-up in vitro and in vivo studies needed to characterize important molecular and cellular mechanisms that lead to CP when they fail. When such mechanisms are understood, this knowledge can be used to develop effective new therapies seeking not to merely control symptoms, but also to treat the fundamental pathophysiology of CP.