Background
The human major histocompatibility complex on chromosome 6 contains many highly polymorphic genes involved in immune function [
1,
2], notably the human leukocyte antigen (HLA) genes. Specific HLA alleles have been associated with more than 100 traits, including striking association with autoimmune diseases [
3,
4]. Given the identification of possible abnormal immune responses observed in participants with autism (for example, antibodies which react to proteins found in the central nervous system [
5]) and similarities between schizophrenia and diseases with established HLA associations [
6], several studies have examined the possible involvement of HLA loci in neurodevelopmental disorders. Such investigations have typically targeted autism and attention deficit hyperactivity disorder (ADHD) as well as psychiatric disorders such as schizophrenia. Specific alleles (please see a note regarding the terminology in the Methods section) of
HLA-DRB1, including DR4 and DR12, have been associated with autism, ADHD, and schizophrenia, although these findings have not always replicated within or between disorders. In addition, association between the A2 allele of the
HLA-A gene and autism has been reported [
7].
Pertinent to the present study is the suggestion that autoimmunity may also be involved in neurodevelopmental disorders that include language deficits: a positive autoimmune reaction to myelin was detected in children with Landau-Kleffner syndrome, a disorder characterized by aphasia and epileptic seizures [
8]. Autoantibodies that react with brain tissue have been found in a relatively large number of children with an atypical variant of Landau-Kleffner syndrome and autism compared with controls. Such antibodies were also found in children affected by other neurological disorders, albeit less frequently [
9]. Another study found antibodies that bound to rodent Purkinje cells in the serum of a mother with a child with specific language disorder, a child with autism, and a child with typical development. When this serum was injected in pregnant mice, the mouse pups had reduced exploratory behavior and impaired motor coordination [
10]. The HLA region has also been implicated in a linkage study of dyslexia [
11].
Specific language impairment (SLI) is diagnosed when a child has problems with the acquisition of language but shows normal development in all other areas [
12]. SLI is a heterogeneous and complex disorder with a strong genetic component [
13]. Several linkage regions and genes have been identified in genetic studies of SLI [
14‐
17]. One gene in particular,
CNTNAP2 (contactin associated protein-like 2), has been implicated in SLI, autism, and ADHD [
16,
18,
19]; in the context of SLI, this gene was associated across three language-related measures: nonword repetition (NWR) scores, Clinical Evaluation of Language Fundamentals (CELF) expressive language scores (ELS), and CELF receptive language scores (RLS) [
16], all of which have been employed in the present study. The product of the
CNTNAP2 gene is involved in forming the gaps in the myelin sheath [
20], which is interesting in the context of the autoimmune reaction to myelin detected in children with Landau-Kleffner syndrome. Though clinically distinct, SLI, autism, and ADHD have some phenotypic similarities; language may be impaired in all three disorders [
21‐
23]. The core linguistic deficit in SLI may not be the same as the one in autism or ADHD. Children with ADHD and children with SLI may differ in their conversational profiles. Whereas children with SLI tend to have more limitations in the areas of lexical diversity, sentence length, and morphosyntactic development, children with ADHD differ from children with normal language in their utterance formulation (that is, in the number of pauses, repetitions, revisions, or other such elements), and children with autism typically have pragmatic impairments [
24‐
26]. Nevertheless, some children with autism exhibit language impairments that are more frequently observed in children with SLI and
vice versa[
22]. Given the genetic and phenotypic overlaps between SLI, autism, and ADHD and the HLA linkage and associations identified in studies of dyslexia, autism, and ADHD, we decided to investigate the possible involvement of HLA loci in SLI. Our initial investigations involved the completion of quantitative and case-control association analyses with single-nucleotide polymorphisms (SNPs) across the HLA region. We went on to impute HLA types from SNP array data and tested for association with language impairment, again within quantitative-trait and case-control association models.
Discussion
SLI is a complex and heterogeneous neurodevelopmental disorder that affects language development in children. HLA associations have been reported for other neurodevelopmental disorders that show some genetic and phenotypic overlap with SLI. We investigated the possible involvement of HLA loci in SLI using several approaches: quantitative and case-control association analyses both with SNPs in and around the HLA region and with HLA types.
QTDT analyses of HLA-type data identified risk and protective alleles with regard to the three measured traits, and these findings did survive multiple testing. HLA-A A1 was the most highly positively correlated allele (with NWR, nominal P = 0.004, empirical P = 0.017), and HLA-A A3 was the most highly negatively correlated allele (with ELS, nominal P = 0.006, empirical P = 0.016), thus implicating the HLA-A locus in susceptibility to SLI.
We also found interesting evidence of parent-of-origin effects. In particular, the
HLA-B B8 allele and HLA-DQA1*0501 are negatively correlated with RLS when inherited from the mother but positively correlated with the same trait when paternally inherited. Parent-of-origin effects have been reported with regard to the involvement of HLA alleles in neurological disorders such as multiple sclerosis [
48], and in this respect it is interesting to find such an effect in SLI. The associations with B8 in both parent-of-origin effects analyses remained significant following the permutation procedure. Interestingly,
HLA-B has been implicated in schizophrenia; however, the increase in risk was the result of matching
HLA-B genotypes in the mother and the child [
33]. While the association detected in our study is with a specific allele, it is interesting to note that the association is correlated with negative test scores only when the allele is inherited from the mother (so there is a matching of at least one allele between the mother and the child).
Although many of the associations observed in this study were of borderline significance, it is interesting to note that the identified association trends are supported by studies of related neurodevelopmental disorders. In an ADHD study of a Chinese cohort [
49], the DR10 allele of
HLA-DRB1 was found to be significantly more frequent in children with ADHD than in controls, as is the case with the probands with SLI in our study. This result translates to a particularly high relative risk of 2.575, with a 95% confidence interval of 1.773 ≤ relative risk ≤ 3.737, and was supported by FDR analyses. Similarly, Wang and colleagues [
49] reported that the DR12 allele of
HLA-DRB1 was significantly more frequent in controls than in cases with ADHD, a trend which again is replicated in our study. The trend of association found between the DR4 allele of
HLA-DRB1 in our probands with language-impairment matches that described by Wright and colleagues [
50] in a study of schizophrenia but is the opposite of that observed in ADHD by Odell and colleagues [
51]. Note, however, that this ADHD association was not replicated in the Chinese study previously mentioned [
49]. Interestingly, the
HLA-DRB1 case-control associations observed in the present study also seem to be the opposite to previously described association trends of HLA alleles with autism (albeit not always significantly so). The DR4 allele was associated with autism [
52,
53] but is significantly less frequent in probands with SLI than in controls (Table
3, Figure
1). DR13 and DR14 (grouped together) have been negatively associated with autism [
52], which is, again, the opposite trend to the one observed here (Figure
1, frequency in probands with language-impairment: 0.14 and 0.03; frequency in controls: 0.11 and 0.02, for DR13 and DR14, respectively). Although these differences were not even nominally significant in SLI, they may suggest that if HLA alleles play a role in both SLI and autism, the mechanism itself may be different.
Although quantitative and case-control SNP-based analyses consistently identified associations within the LRRC16A gene, none of the SNP-based analyses found strong association trends within HLA loci. Furthermore, the FDR q values obtained for SNP-based associations were all greater than 0.05, indicating a high false-positive rate. Nonetheless, it should be noted that one might not expect to find a direct correlation between SNPs in HLA genes and effects mediated by HLA types, because of the high degree of variation in the HLA region.
Conclusions
The results of this study suggest a potential involvement of HLA loci in SLI. Quantitative association analyses highlighted HLA-A and parent-of-origin effects for HLA-B, whereas case-control analyses implicated HLA-DRB1 alleles. Further, larger-scale studies will be required to replicate these findings. The relatively small sample sizes employed are reflected by nominal P values. Since our analyses used imputed HLA alleles and since no HLA associations have been previously reported for SLI, we did not confine ourselves to testing only a subset of the HLA alleles, and, consequently, we performed a relatively large number of tests. Nonetheless, we believe that association testing and imputation of HLA types provide a cost-effective way of studying the involvement of immune-related genes in neurodevelopmental disorders. The preliminary data presented here provide an intriguing link to those described by previous studies of other neurodevelopmental disorders suggesting a possible role for HLA loci in language disorders.
Acknowledgments
We would like to thank all the families, professionals, and individuals who participated in this research. In particular, we would like to thank Simon Fiddy for his assistance with data transformation and Benjamin Fairfax for the use of the Oxfordshire Control samples. DN is a Medical Research Council (MRC) Career Development Fellow and a Junior Research Fellow at St John’s College, University of Oxford. The work of the DN lab is funded by the MRC (G1000569/1 and MR/J003719/1). RN is funded by a University of Oxford Nuffield Department of Medicine Prize Studentship. The genotyping of samples was funded by the Max Planck Society. The collection of the SLIC samples was supported by the Wellcome Trust (060774 and 076566). Recruitment of controls for the Oxfordshire study of gene expression in primary immune cells was supported by the Wellcome Trust (074318 and 088891), the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007-2013) (281824), and the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre. Recruitment of controls for the POBI study was supported by the Wellcome Trust (072974, 088262). PFB is supported by an NIHR (UK) Senior Investigator award and the Biomedical Research Centre in Mental Health at the South London & Maudsley NHS Trust Hospital, London. The work of the Wellcome Trust Centre in Oxford is supported by the Wellcome Trust (090532/Z/09/Z).
We are very grateful to the other members of the SLIC for their contributions to this work: V. Slonims (Newcomen Centre, Evelina Children’s Hospital, London), A. Clark, J. Watson (Speech and Hearing Sciences, Queen Margaret University, Edinburgh, UK), E. Simonoff, A Pickles (King’s College London, Institute of Psychiatry); A. Everitt (University Child Health and DMDE, University of Aberdeen); J. Seckl (Molecular Medicine Centre, University of Edinburgh); H. Cowie (Department of Speech and Language Therapy, Royal Hospital for Sick Children, Edinburgh); W. Cohen (Psychological Sciences and Health, University of Strathclyde); J. Nasir (Division of Biomedical Sciences, St George’s University of London); D.V.M. Bishop (Department of Experimental Psychology, University of Oxford); Z. Simkin (School of Psychological Sciences, University of Manchester).
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
RN conceived and designed this study, performed the imputation and association analyses, and drafted the manuscript. NHS helped with the quality control for the DNA and generation of the SNP genotype data. GB managed the acquisition of data and DNA from the Guy’s Hospital SLIC cohort. AO managed the acquisition of data and DNA from the Edinburgh SLIC cohort. GC-R managed the acquisition of data and DNA from the Manchester SLIC cohort. PFB managed the acquisition of data and DNA from the Cambridge SLIC cohort. ERH managed the acquisition of data and DNA from the Aberdeen SLIC cohort; the SLIC is a group of individuals who collected the DNA and data for the SLIC cohort and provided vital intellectual input to the study design and management of the SLIC resource. APM is the principal investigator for the SLIC genetic data and assisted in the conceptualization of the study and contributed to the intellectual content of the manuscript. JCK managed the acquisition of data and DNA from the control individuals from the Oxfordshire study of gene expression in primary immune cells and assisted with the interpretation of the data for the manuscript. BW managed the acquisition of data and DNA from the control individuals for the POBI cohort and provided intellectual input into the study design. SEF managed the generation of SNP data for the SLIC individuals and contributed to the intellectual conception of this study. DFN performed quality-control procedures on the SLIC genetic data and helped with the conceptualization and design of the study and the drafting of the manuscript. All authors read and approved the final manuscript.