Background
Cerebral palsy (CP) is a group of nonprogressive motor impairment syndromes caused by lesions of the brain arising early in development. The motor disorders of CP are often accompanied by disturbances of sensation, perception, cognition, communication, and behavior; by epilepsy; and by secondary musculoskeletal problems [
1]. A number of risk factors for the condition have been identified, including preterm birth, intrauterine infection, intrauterine growth restriction, perinatal stroke, and a sibling with CP [
2‐
7]. How the fetus responds in the presence of these risk factors is influenced by its genetic makeup, with some genotypes creating susceptibility to cerebral damage [
8,
9]. Several rare genetic variants that cause Mendelian forms of CP have been described [
10]; however, these do not account for the vast majority of CP cases.
Interleukin-6 (IL6) is a cell signaling molecule that has been associated with over 20 diseases, including inflammatory, neurological, vascular, and malignant processes (Additional file
1: Table S1). The most studied single nucleotide polymorphism (SNP) in the IL6 gene is rs1800795 (−174G < C). This SNP lies within the promoter region and the C allele has been associated with either increases or decreases inIL6 levels in several tissues [
11‐
14].
Both the IL-6 gene and IL-6 protein levels have been linked with increased risk of CP. Within the IL6 gene, the C allele at rs1800795 has been associated with increased risk of CP in large population studies [
15‐
17]. Elevated cerebrospinal fluid IL-6levels have been associated with worse neurologic outcomes following hypoxic-ischemic encephalopathy [
18]. Elevated IL-6 levels in umbilical cord plasma and amniotic fluid and within brain lesions have been associated with periventricular leukomalacia, a major risk factor for CP in preterm infants [
19‐
22]. Higher cord blood IL-6 protein levels also predict cerebral lesions on MRI in preterm infants [
23].
To our knowledge, no IL6 sequencing or haplotype studies have been performed to further evaluate the relationship between IL6 genotype and CP. Therefore, we performed Sanger sequencing of the IL6 gene in a nested case control study of CP among term infants to identifyIL6 haplotypes and mutations that predispose individuals to high risk of CP.
Conclusions
In this study, we found strong association of CP with an extended 7-SNP haplotype containing the functional promoter SNP rs1800795 in a recessive model of disease inheritance. We observed greater risk for homozygotes at the entire haplotype than for homozygotes at a smaller 4-SNP haplotype identified in African Americans, or for homozygotes at any individual locus. This suggests that there may be more than one functional locus within the 7-SNP haplotype, and that at least one functional locus may be outside of the smaller 4-SNP haplotype. In addition to the promoter function ofrs1800795 that is within the smaller 4-SNP haplotype, disruption of a methylation site at position rs2069845 [
27] could be part of the mechanism conferring additional risk.
The variant rs1800795 has been implicated in various neurological, vascular, and malignant processes (Additional file
7). In these disease processes, it has often been implicated as part of a haplotype with (or implicated simultaneously with) other SNPs including many of the SNPs identified in this study (Additional file
1: Table S1). Two SNPs which have been previously found on haplotypes with rs1800795 include a rare missense SNP rs13306435(4220 T < A) [
28] and an intron SNP rs2069840 [
29]; neither showed evidence of association with CP in this study, although analysis of the rare missense SNP was likely underpowered. Other SNPs that have been previously found on haplotypes with rs1800795 but were outside the genotyping bounds of this study include possibly functional promoter SNPs(rs1800796 and rs1800797) [
13] and less likely functional SNPs from the 5′ untranslated region (rs12700386, rs7802307, rs7802308, rs2069827) and the 3′ untranslated region (rs11766273, rs2069861, rs1818879). A study published after our analysis was complete [
30] suggests that rs1800796-rs2069837 may be associated with risk in a subset of CP patients: Chinese males with spastic paraplegia. We found no association between rs2069837 and CP. However, our analysis was underpowered as this variant is rare in whites, Hispanics, and African Americans. The effect of the rs1800795 variant or the entire haplotype variant on the expression level of IL6 is not fully understood, with evidence for the CC genotype being associated with high or low expression in different tissues [
11‐
14].
We found a trend for association of CP with a likely damaging rare frameshift mutation at position 316. Due to the rarity of this allele (frequency ~ 1%), a larger study would be needed to confirm this finding. Larger studies or meta-analyses can assess potential association with the other uncommon variants. However, such variants would be uncommon, even in CP patients, and therefore explain a small fraction of population risk.
There are a variety of common approaches to haplotype definition which have been used to describe IL6. Among other common methods (Additional file
1: Table S1), the method presented here is appealing because it achieves unequivocal haplotype boundaries (i.e. intra-haplotype r
2 < 0.3, inter-haplotype r
2 < 0.3 for any pair-wise comparison of SNPs). This method is not impaired by contiguous SNP inclusion criteria (i.e. it allows defined haplotypes to include non-contiguous SNPs), an impairment which would have made the discovery of this haplotype impossible in a sequencing dataset due to the high number of variants found. This method minimizes probability of winner’s curse by reducing the number of hypotheses tested to less than n (where n is the number of SNPs found), while other methods increase the number of tests to several fold higher than n. Other haplotype methods that seek independence rather than correlation can also achieve the same criteria, see for example [
31]. A combination of both approaches (i.e. seeking an effect of a second haplotype while conditioning on the lead haplotype) is also possible, but did not yield significant results and was not shown. The method used in this current study is more likely to prove useful in a low-quality (high missing-ness) dataset where increasing signal-to-noise (by selecting multiple correlated measures) is more important.
Previous genetic studies of IL6 in CP have been limited to a handful of SNPs; sequencing presented several advantages. We have been able to fully describe the haplotype block of rs1800795 (to within the bounds of genotyping). Sequencing has also revealed a frameshift variant that causes early protein termination and absence of the four largest protein domains. Finally, sequencing allows us to rule out any other high-risk common variants (other than those discussed) within the boundaries of genotyping. A limitation of this study is the boundaries of genotyping. Functional promoter elements could be part of the mechanism conferring additional risk to CP. Another limitation was the low quality of the sequencing due in large part to low quantity of blood available in each sample, as this study “recycled” unused portions of blood samples after they had been used for routine newborn screening. To address this, we used manual calls made by 2 independent operators to confirm significant findings, discarding all calls that were not concordant. Thus, we sacrificed call rate for accuracy. Another limitation is the lack of confirmation of genotyping and lack of external replication, although it is reassuring that the main haplotype findings were significant independently in whites and Hispanics, as well as in males and females. The haplotype identified in this study, plus additional SNPs in the promoter region, may play a role in other diseases. Targeted genotyping of all haplotype SNPs in CP and other diseases using independent replication populations would be of interest. A genome-wide association study of CP may reveal additional genetic loci which predispose to this heterogeneous group of diseases.
YWW is a professor in the Departments of Neurology and Pediatrics at the University of California, San Francisco (UCSF), and is supported by the National Institute of Health (NINDS K02 NS46688), and the United Cerebral Palsy Foundation (EH-005-03). SEB is a professor in the Department of Neurology at the UCSF and is a Harry Weaver Neuroscience Scholar of the National Multiple Sclerosis Society. LAC is a senior research scientist and directory of the Autism Research Program at Kaiser Permanente. The funding bodies played no role in the design, collection, analysis, interpretation of data, writing of the manuscript, or decision to publish the manuscript for presentation.
Competing interests
The authors have no competing interests to declare.
Authors’ contributions
YWW, LAC, SEB, and PK conceived and designed the experiment. YWW and LAC performed chart review and selected samples. YWW, PK, BJ, LM, and DN performed data analysis. All authors read and approved the manuscript.