A Functional Genetic Link Between Distinct Developmental Language Disorders

The new england journal of medicine

original article

A Functional Genetic Link between Distinct Developmental Language Disorders

Sonja C. Vernes, D.Phil., Dianne F. Newbury, D.Phil., Brett S. Abrahams, Ph.D., Laura Winchester, B.Sc., Jérôme Nicod, Ph.D., Matthias Groszer, M.D., Maricela Alarcón, Ph.D., Peter L. Oliver, Ph.D., Kay E. Davies, D.Phil., Daniel H. Geschwind, M.D., Ph.D., Anthony P. Monaco, M.D., Ph.D., and Simon E. Fisher, D.Phil.

Abstract

Background Rare mutations affecting the FOXP2 transcription factor cause a monogenic speech From the Wellcome Trust Centre for Hu- and language disorder. We hypothesized that neural pathways downstream of FOXP2 man Genetics (S.C.V., D.F.N., L.W., J.N., M.G., A.P.M., S.E.F.) and the Medical Re- influence more common phenotypes, such as specific language impairment. search Council Functional Genomics Unit (S.C.V., P.L.O., K.E.D.), University of Ox- Methods ford, Oxford, United Kingdom; and the Department of Neurology (B.S.A., M.A., We performed genomic screening for regions bound by FOXP2 using chromatin D.H.G.) and the Semel Institute and the immunoprecipitation, which led us to focus on one particular gene that was a Department of Human Genetics (D.H.G.), strong candidate for involvement in language impairments. We then tested for asso- David Geffen School of Medicine, Uni- versity of California, Los Angeles, Los ciations between single-nucleotide polymorphisms (SNPs) in this gene and language Angeles. Address reprint requests to Dr. deficits in a well-characterized set of 184 families affected with specific language Fisher at the Wellcome Trust Centre for impairment. Human Genetics, University of Oxford, Roosevelt Dr., Headington, Oxford OX3 7BN, United Kingdom, or at simon.fisher@ Results well.ox.ac.uk. We found that FOXP2 binds to and dramatically down-regulates CNTNAP2, a gene This article (10.1056/NEJMoa0802828) was that encodes a neurexin and is expressed in the developing human cortex. On analyz- published at www.nejm.org on November ing CNTNAP2 polymorphisms in children with typical specific language impair- 5, 2008. ment, we detected significant quantitative associations with nonsense-word repeti- −5 N Engl J Med 2008;359. tion, a heritable behavioral marker of this disorder (peak association, P = 5.0×10 Copyright © 2008 Massachusetts Medical Society. at SNP rs17236239). Intriguingly, this region coincides with one associated with language delays in children with autism.

Conclusions The FOXP2–CNTNAP2 pathway provides a mechanistic link between clinically distinct syndromes involving disrupted language.

n engl j med 10.1056/nejmoa0802828 1

evelopmental disorders of speech, specific language impairment had a significant language, and communication account for speech delay.13 Moreover, analyses of FOXP2 in D40% of referrals to pediatric services.1 persons with typical forms of specific language Although many children grow out of early lan- impairment have not detected etiologic mutations guage delay, others have persistent difficulties or evidence of association.14,15 Mutation of FOXP2 with language expression and comprehension, itself is therefore unlikely to be a major risk fac- despite normal nonverbal ability and lack of an tor for common language impairments. Indeed, obvious reason. In some children, developmental to date we know of no report of a gene associated speech or language impairments are part of a with typical specific language impairment.12 broader syndrome such as autism, in which these Because FOXP2 encodes a neurally expressed deficits are accompanied by unusual repetitive transcription factor,16,17 we reasoned that one or behaviors and disturbances in social interaction. more of the genes that it regulates in the brain More commonly, such impairments occur in the might be implicated in common language-related absence of autistic features.2 Longitudinal studies phenotypes. Here we describe the isolation of a have indicated that when language impairments novel FOXP2-regulated target with neural func- persist to school age, they are likely to be asso- tions and provide evidence of its association with ciated with enduring academic and psychiatric language-related deficits in a large set of well- problems.3 characterized families with specific language Developmental speech and language disorders impairment. are highly heritable, with most cases showing complex multifactorial inheritance.4 The isolation Methods of relevant genetic effects will yield new insights into the causes of such impairments, along with Screening for Targets of FOXP2 improved classification, diagnosis, and treatment. We engineered the human neuroblastoma SH-SY5Y One notable success in this area was the discov- cell line to stably express FOXP218 and then, using ery that heterozygous disruptions of the FOXP2 this transfected cell line, carried out unbiased gene cause a rare mendelian speech and language screening for genomic sites bound by FOXP2 pro- disorder.5-9 Point mutations and chromosomal tein. This involved the use of chromatin immuno- abnormalities that affect FOXP2 are associated precipitation with anti-FOXP2 antibodies, followed with difficulties in the learning and production by shotgun sequencing of purified DNA, a pro- of sequences of oral movements, which impair cess of randomly cloning fragments of DNA and speech (also called developmental verbal dyspraxia then determining their sequence (for details, see or childhood apraxia of speech).5-9 The affected the Supplementary Appendix, available with the persons also have variable levels of impairment full text of this article at www.nejm.org). We de- in expressive and receptive language, extending termined the positions of DNA sequences that to problems with production and comprehension were isolated with chromatin immunoprecipita- of grammar.10 However, FOXP2 disruptions are tion, using BLAT on the University of California, rare. It has been estimated that approximately Santa Cruz, Genome Server (http://genome.ucsc. 2% of people with verbal dyspraxia carry etio- edu/), which enabled identification of putative logic point mutations in this gene.6 target genes. Specific language impairment is the most fre- quently diagnosed form of developmental lan- Validation of Binding and Regulation guage disorder, affecting up to 7% of children by FOXP2 who are 5 or 6 years of age.11 Although there is Binding of FOXP2 to target sites was independent- considerable variation in the profile of linguistic ly verified and further localized with the use of deficits observed and in the functions affected semiquantitative polymerase-chain-reaction (PCR) (expressive, receptive, or both),12 specific lan- assay of chromatin isolated from additional guage impairment often occurs without accom- chromatin-immunoprecipitation experiments and panying difficulties in speech articulation. For electrophoretic mobility shift assays (EMSAs), ac- example, an epidemiologic study showed that cording to protocols reported previously.18 Regu- only about 5 to 8% of children with persistent lation of putative target genes was assessed with

2 n engl j med 10.1056/nejmoa0802828

Downloaded from www.nejm.org on November 6, 2008 . For personal use only. No other uses without permission. Copyright © 2008 Massachusetts Medical Society. All rights reserved. A Genetic Link in Distinct Developmental Language Disorders the use of quantitative reverse-transcriptase PCR and a standard deviation of 15 in the general (RT-PCR) of RNA extracted from SH-SY5Y cells population. Lower scores indicate poorer perfor- expressing different FOXP2 levels, as described mance.) This measure is thought to provide an previously18 (see Table S1 in the Supplementary index of phonologic short-term memory.12 Chil- Appendix for primer sequences). In situ hybrid- dren with specific language impairment perform ization was performed on human fetal brains,19 particularly poorly on nonsense-word repetition, as described in the Methods section in the Sup- and impaired phonological short-term-memory plementary Appendix. has been proposed as a core deficit in the disorder. An impairment in the ability to repeat nonsense Study Subjects words is highly heritable, persists in persons with The study subjects were members of epidemio- historical language problems that have otherwise logically and clinically ascertained families iden- resolved,27 and appears to be relatively unaffected tified by the Specific Language Impairment by environmental factors.28 Additional information Consortium.20,21 These families were recruited on the consortium families has been reported from four sites in the United Kingdom: the New- previously20,21 and is available in Table S2 in the comen Centre at Guy’s Hospital,20,21 the Cam- Supplementary Appendix, which shows means, bridge Language and Speech Project,22 the Child standard deviations, and intertrait correlations for Life and Health Department at Edinburgh Univer language measures used in this study. Written sity,23 and the Manchester Language Study.24 informed consent was obtained from all subjects Families were selected through a proband with or their parents; assent was obtained from chil- specific language impairment whose past or cur- dren of appropriate age. rent language skills were 1.5 SD or more below the normative mean for the child’s age on the Single-Nucleotide-Polymorphism Genotyping Clinical Evaluation of Language Fundamentals– To directly test the hypothesis that variants of the Revised (CELF-R) scale,25 a tool that is routinely identified FOXP2 target (the CNTNAP2 gene) may used for diagnosis and follow-up evaluation of increase susceptibility to common language im- language disorders in school-age children. (Scores pairments, we genotyped single-nucleotide poly- on the scale range from 50 to 150, with a mean morphisms (SNPs) in consortium families, fol- of 100 and a standard deviation of 15 in the gen- lowed by quantitative association analyses of eral population. Lower scores indicate poorer per- measures of expressive and receptive language formance.) We excluded any children with a non- abilities and nonsense-word repetition. We geno- verbal IQ of less than 80, a clinical diagnosis of typed and validated 38 SNPs from the CNTNAP2 an autistic-spectrum disorder, or another known locus on chromosome 7q35 in samples from 847 medical or developmental condition that can im- members of 184 consortium families, using Gold- pair language, such as hearing loss, cleft lip, or en Gate assays on the Illumina platform (for de- cleft palate. Moreover, for clinically ascertained tails, see the Methods section and Table S3 in the samples, children were comprehensively assessed Supplementary Appendix). on scales evaluating language, IQ, and behavior, and those with overt pragmatic difficulties, be- Statistical Analysis havioral characteristics associated with autism, For analyses of differences in gene expression in or a family history indicative of autism were also SH-SY5Y cells, we assessed statistical significance excluded. using unpaired t-tests (two-tailed). For family- We collected quantitative phenotypic data from based association analyses of SNP data from the probands and all available siblings. We then de- consortium series, we used a quantitative trans- termined composite CELF-R scores for expressive mission disequilibrium test (QTDT), adopting an and receptive language abilities. We also used a orthogonal association model that considers only measure of the ability to repeat nonsense words, the within-family variance and is robust to popu- the Children’s Test of Nonword Repetition,26 which lation stratification.29 After identifying significant has been established as a robust endophenotype single SNP associations, we used the Merlin pack of specific language impairment.12 (Scores on the age30 to generate haplotypes for the cluster of nine scale range from 46 to 141, with a mean of 100 associated SNPs, which were similarly analyzed

n engl j med 10.1056/nejmoa0802828 3

with QTDT. Finally, we investigated the possibil- Figure 1 (facing page). Identification of CNTNAP2 ity of an effect of sex or imprinting within QTDT, as a Direct Neural Target Bound by Human FOXP2. using these nine SNP-tag haplotypes. In Panel A, a 300-bp clone was identified through shotgun cloning of gene fragments identified by FOXP2– Results chromatin immunoprecipitation and localized to intron 1 of the human CNTNAP2 gene in 7q35. Semiquantitative CNTNAP2 as a Target of FOXP2 PCR analysis indicated consistent enrichment of this region in multiple independent experiments in a neuron- To identify candidate genes that might be involved like cell line immunoprecipitated with an N-terminal in typical specific language impairment, we used FOXP2 antibody (lane 2), as compared with a control an unbiased screening method to isolate genomic sample without the antibody (lane 3) and input DNA fragments bound by the FOXP2 protein in chro- samples (lane 1). Lane 4 shows the water control sam- matin of human neuronlike cells. We thereby dis- ple. Two FOXP2 consensus binding sites were identified (highlighted in red). In Panel B, electrophoretic mobility covered a FOXP2-bound fragment that was of par- shift assays (EMSAs) using nuclear extracts from trans- ticular interest, because it was located within fected HEK293T cells assessed the ability of FOXP2 pro- intron 1 of CNTNAP2 (Fig. 1A). This gene encodes tein to bind to the 5′ consensus binding site (highlighted CASPR2, a member of the neurexin superfamily in red). Efficient binding to the CNTNAP probe was ob- of transmembrane proteins, found at the nodes served when FOXP2 was present but not when either un- transfected cells or cells expressing a mutant form of of Ranvier in myelinated nerve fibers. In mice, FOXP2 (R553H) were used.17 Binding to the labeled Caspr2 is important for the regulation of the local- probe was efficiently reduced by competition with an un- ization and maintenance of Shaker-type voltage- labeled probe (CNTNAP) but not by a mutant form of activated potassium channels31 and is implicated the probe (CNTNAP-M) or an irrelevant binding site in neuronal recognition and cell adhesion.32 In hu- (NFK). The arrow shows the position of the shift caused by FOXP2 binding to the CNTNAP probe. mans, it has been suggested that CASPR2 is involved in cortical development, possibly mediating intercellular interactions during neuroblast migra- and significantly reduced in neuronlike cells that tion and laminar organization.33 were stably transfected with FOXP2, as compared We used PCR to amplify sequences spanning with sham-transfected control samples (Fig. 2A). the FOXP2-bound fragment of CNTNAP2 in inde- A recent genomewide analysis of differential pendent FOXP2 chromatin-immunoprecipitation gene expression in the developing human cerebral samples and in control samples in which no anti- cortex independently highlighted CNTNAP2 as a bodies were used and observed evidence of en- gene with substantial enrichment in frontal gray richment only when FOXP2-specific antibodies matter, which is primarily restricted to the region had been used to isolate the chromatin (Fig. 1A). between the orbital gyrus and superior frontal Primers amplifying regions of 1000 bp or more anlage, spanning the inferior and middle frontal away from the bound fragment did not display gyri.34 Because FOXP2 is also expressed in the de FOXP2–chromatin immunoprecipitation enrich- veloping human cortex,16,34,35 we carried out ex- ment. FOXP2 is thought to bind chromatin as a pression analyses of this structure in fetal tissue dimer, and our in silico analyses of the chroma- (18 to 22 weeks’ gestation) through in situ hybrid- tin immunoprecipitation–enriched fragment iden- ization. We observed complementary patterns with tified two adjacent sites, separated by 48 bases, respect to cortical lamination: CNTNAP2 expres- matching a known consensus sequence for FOXP2 sion was lowest in layers that showed the highest binding (CAAATT). EMSA analyses indicated that levels of FOXP2 (Fig. 2B). These in vivo findings FOXP2 could bind both sites (data not shown). At are consistent with our data from neuronal mod- each site, binding could be disrupted by the muta- els, supporting negative regulation of human tion of three core nucleotides of the recognition CNTNAP2 expression by FOXP2. sequences (CAAATT→CGGGTT), with more dra- matic effects observed for the 5′ site. Full compe- Association Analyses of CNTNAP2 tition assays for this site showed highly efficient Several studies underscored CNTNAP2 as a partic- and specific binding by FOXP2 (Fig. 1B). ularly compelling candidate gene to test for as- We then used quantitative RT-PCR to directly sociation with specific language impairment. In test whether modulation of FOXP2 protein levels addition to our identif ication of it as a direct neu- would yield altered CNTNAP2 expression. Indeed, ral target of FOXP2, it has known neuronal CNTNAP2 messenger RNA levels were consistently functions,31,32 and its expression is enriched in hu-

4 n engl j med 10.1056/nejmoa0802828

A FOXP2-bound Fragment in CNTNAP2 CNTNAP2

Chromosome 7

7q35

7q34 7q35 7q36.1 7q36.2

CNTNAP2 Coding Region

Exon 1 23

P 2 P1 P3

12 34 12 34 12 34 Shotgun Clone (~300 bp) 5' 3' TTC AAATTTA...... CC C AAATTGT

B EMSA of FOXP2 Binding

FOXP2 Nuclear Extract HEK293FOXP2-R553H

Competitor – – – CNTNAPCNTNAP-M NFK

EMSA Probes CNTNAP: 5'–AGCTGCTTTCAAATTTAAGCAATCAAGTG–3' CNTNAP-M: 5'–AGCTGCTTTCGGGTTTAAGCAATCAAGTG–3' NFK: 5'–AGCTCCGGGGGTGATTTCACTCCCCG–3' man language-related circuitry.34 Furthermore, with one study showing association with a mea- the gene is disrupted in a family with Tourette’s sure of language delay (the age at the first spoken 36 37 syndrome, and a rare point mutation causes a word) in multiplexRETAKE autism1st families. ICM AUTHOR: Vernes (Fisher) 2nd CNTNAP2 severe recessive disorder involvingREG FcorticalFIGURE: dyspla1 of 3 - We therefore went on to assess in- 3rd sia and focal epilepsy, associatedCASE with language volvement inRevised specific language impairment by 33 regression and autistic characteristics.EMail Recent in- Linegenotyping4-C polymorphismsSIZE across the locus in ARTIST: ts H/T H/T 33p9 dependent studies have implicatedEnon variants at the Combothe large series of consortium families and test- 37-39 CNTNAP2 locus in autistic-spectrum disorders,AUTHOR, PLEASEing forNOTE: marker-trait association, using a family- Figure has been redrawn and type has been reset. Please check carefully. n engl j med 10.1056/nejmoa0802828 5

JOB: 35922 ISSUE: 11-27-08 Downloaded from www.nejm.org on November 6, 2008 . For personal use only. No other uses without permission. Copyright © 2008 Massachusetts Medical Society. All rights reserved. The new england journal of medicine

based association method. Using an approach A CNTNAP2 Regulation by FOXP2 that was consistent with previous studies of this CNTNAP2 series,21,22 we analyzed quantitative phenotypes Primer A P<1.0×10−4 from probands and all available siblings, regard-

Primer B P<1.0×10−4 less of the diagnosis of specific language impairment, and focused on composite diagnostic mea- Primer C P<1.0×10−4 sures of expressive and receptive language abilities, as well as a test of nonsense-word repetition, FOXP2 P<0.001 which was previously established as a robust −5 0 5 10 15 20 endophenotype.12 Mean Log2 Ratio of Normalized mRNA in FOXP2-Positive Cells We observed significant associations (with P values from 0.01 to 5.0×10−5) between nonsense- B Messenger RNA Expression in Fetal Cortex word repetition and nine intronic SNPs (rs851715, CNTNAP2 FOXP2 rs10246256, rs2710102, rs759178, rs1922892, a b rs2538991, rs17236239, rs2538976, and rs2710117), all mapping between exons 13 and 15. The most strongly associated SNP was rs17236239 (P = 5.0×10−5) (Fig. 3A, and Table S3 in the Supple- mentary Appendix). Even after an overly conser- vative Bonferroni correction for testing of multiple SNPs was made, this P value remained Cresyl Violet Schematized Expression significant (P = 0.002). The rs17236239 SNP was c d also the marker showing strongest evidence of association with expressive language abilities (P = 0.008). The exon 13–15 region was similarly implicated in analyses of receptive language abilities, but in this case the strongest association was observed for a different SNP, rs4431523 (P = 0 . 0 0 3 ) . We constructed multimarker haplotypes with the 9 SNPs implicated in the ability to repeat Figure 2. Analyses of the Effects of FOXP2 on Neural Expression of CNTNAP2. RETAKE 1st ICM AUTHOR: Verne (Fisher) nonsense words and observed 11 different com- Panel A shows the regulation of CNTNAP2 expression by FOXP22nd in human REG F FIGURE: 2 of 3 binations. Four haplotypes represented 94% of neuronlike cells. The expression of messenger RNA (mRNA) was3rd assessed with quantitativeCASE reverse-transcriptase PCR in SH-SY5YRevised cells. The cells were subjects (Table S4 in the Supplementary Appen- Line 4-C stably transfectedEMail either with a construct expressing FOXP2SIZE (FOXP2-positive dix). The most common haplotype, ht1, negative ARTIST: ts H/T H/T cells) or with anEnon empty control vector thatCombo does not contain22p3 a gene insert ly inf luenced the ability to repeat nonsense words; (control cells). Levels of CNTNAP2 mRNA in these cells were inversely pro- AUTHOR, PLEASE NOTE: it was more often present in family members portional to that ofFigure FOXP2. has Findingsbeen redrawn were and consistent type has been for threereset. sets of primer pairs recognizing distinct combinationsPlease check of carefully. CNTNAP2 exons (primers A to C). with poor scores (a frequency of 40% among Expression changes are given as the mean log2 expression ratios in FOXP2- those with scores of >2 SD below the population positive cells,JOB: as 35921compared with empty controls, normalizedISSUE: 11-27-08 for equal expres- mean) than in those with good scores (a fre- sion of the internal control, GAPDH. The I bars represent standard errors. quency of 29% among those with scores of >0.5 The P values were calculated with the use of two-tailed unpaired t-tests. SD above the population mean). We classified Panel B shows nonoverlapping mRNA expression for CNTNAP2 and FOXP2 in human fetal cortex. Adjacent sections from human fetal brain (at 18 to probands and siblings according to the number 22 weeks’ gestation) were processed for in situ hybridization, dipped in of copies (none, one, or two) of this putative risk film emulsion, and visualized by dark-field microscopy. Within the cerebral haplotype they had and calculated the mean score cortex, the highest levels of CNTNAP2 mRNA are observed between bands for nonsense-word repetition for each group of FOXP2 expression, putatively within layers II and III of the cortical plate (Fig. 3B). There was a decrease of approximately (subpanel a). In contrast, FOXP2 is present at high levels in the molecular zone, deep layers of the cortical plate, and subplate (subpanel b). Cortical 6 points (0.4 SD) between the mean scores for lamination is highlighted in a bright-field image of a section stained with cre- nonsense-word repetition of children carrying no syl violet (subpanel c). Nonoverlapping expression patterns for CNTNAP2 and copies of ht1 (mean, 95.2) and those carrying FOXP2 are schematized in subpanel d. CP denotes cortical plate, MZ molecu- two copies (mean, 89.4). We observed a difference lar zone, and SP subplate. of similar magnitude in scores between children carrying no copies of ht1 and those carrying one

6 n engl j med 10.1056/nejmoa0802828

Downloaded from www.nejm.org on November 6, 2008 . For personal use only. No other uses without permission. Copyright © 2008 Massachusetts Medical Society. All rights reserved. A Genetic Link in Distinct Developmental Language Disorders copy of ht1 (mean, 89.7), a finding that suggests A Association Analyses of CNTNAP2 a dominant effect. Although only 49 children NWR ELS RLS Exon carried two copies of ht1, as compared with 176 with no copy and 174 with one copy, we ob- 0.00001 served very similar results in an augmented data 0.0001 set incorporating all available parental scores (Fig. 3B). Moreover, family-based quantitative 0.001 association analyses of the nine-marker haplo- P Value 0.01 types and scores on nonsense-word repetition yielded a P value of 6.0×10−4 for ht1, again indi- 0.1 cating that this haplotype is significantly associ- 1 ated with impaired language performance (Table 5 0 5 0 5 0 5 0 5 0 145.2 145.5 145.7 146.0 146.2 146.5 146.7 147.0 147.2 147.5 S4 in the Supplementary Appendix). Position on Chromosome 7 (Mb) Investigations of CNTNAP2 in patients with autism showed an increased association in fam- B Risk Alleles and Nonsense-Word Repetition ilies with affected males37 and also suggested 98 38 96 the possibility of imprinting. We therefore re- Children 94 peated our QTDT analysis of the haplotype as- 92 sociated with impaired nonsense-word repetition 90

Score 88 in the families with specific language impair- All subjects 86 ment, using sex as a covariate and testing for 84 differences in transmission of paternal and ma-

Nonsense-Word Repetition 0 ternal alleles. We found no evidence of a role of 012 sex (P = 6.0×10−4 with no adjustment for sex, No. of Risk Alleles P = 8 . 0 ×1 0 −4 with adjustment for sex) or of imprinting (P = 0.27) at this locus. Figure 3. Association between CNTNAP2 and Language Deficits in Families with Specific Language Impairment. Panel A shows the results of analyses with a quantitative transmission dis- Discussion equilibrium test (QTDT) of SNPs from the CNTNAP2RETAKE locus and 1sttheir associa- ICM AUTHOR: Verne (Fisher) tion with measures of an ability to repeat nonsense words (orange),2nd to ex- REG F FIGURE: 3 of 3 We have shown that FOXP2, which is mutant in press language (green), and to understand language (blue). The3rd circles CASE Revised show the positions of individual SNPs, and the black diamonds at the top of people with a rare speech and language disorder, EMail Line 4-C SIZE directly regulates expression of the CNTNAP2 the graph indicate theARTIST: relative tspositionsH/T of exons,H/T according22p3 to the National Center for BiotechnologyEnon Information Combogenetic sequence (build 35). The or- gene. We went on to demonstrate that variants of ange square denotes the positionAUTHOR, of PLEASEthe FOXP2-bound NOTE: fragment from the CNTNAP2 are associated with def icits in common shotgun cloning ofFigure gene has fragments been redrawn isolated and type by chromatin-immunoprecipihas been reset. - forms of language impairment. In so doing, we tation screening. (Details regardingPlease check SNP carefully.locations, allele frequencies, QTDT results, and effect sizes are available in Table S3 in the Supplementa- provide an example of how knowledge of the ge- JOB: 35921 ISSUE: 11-20-08 netic cause of a rare single-gene disorder provides ry Appendix.) Panel B shows the effect of the multimarker haplotype ht1 on scores evaluating the ability to repeat nonsense words on the Children’s an entry point into the causes of a more complex Test of Nonword Repetition.26 Scores on the scale range from 46 to 141, phenotype. Further analyses of the relevant regu- with a mean of 100 and a standard deviation of 15 in the general popula- latory networks — including the FOXP2–CNTNAP2 tion. Lower scores indicate poorer performance. Children (gray) were divid- pathway identified here — may lead to a better ed into three groups on the basis of the numbers of copies of the putative understanding of neurogenetic mechanisms in- risk allele ht1 that they carried. The mean score on nonsense-word repetition dropped by about 6 points as a consequence of carrying one or more volved in typical language disorders. copies of ht1. Similar results were seen with a larger sample that also in- The FOXP2-bound fragment of CNTNAP2 lies cluded parents with available phenotypic data (black). The I bars represent outside the classically defined regulatory regions standard errors. QTDT analyses of the multimarker haplotypes yielded a of the genome represented on available promoter- P value of 6.0×10−4 for the association between ht1 and measures of non- based microarrays. It therefore escaped detection sense-word repetition (Table S4 in the Supplementary Appendix). in recently published efforts in which chromatin immunoprecipitation with FOXP2 antibodies was with many of such sequences mapping within coupled to screening of such microarrays.18,40 introns.41 Indeed, large-scale surveys of transcription-factor Thus far, CNTNAP2 is the only FOXP2 target binding have indicated that functional regulatory that we have tested for association in specific lan- sequences often lie far from known promoters, guage impairment. Of all FOXP2 targets identi-

n engl j med 10.1056/nejmoa0802828 7

fied to date,18,40 we prioritized CNTNAP2 for as- pairment, but when they occur in association with sociation testing in common language disorders other social and behavioral deficits, such impair- because it represented the most compelling of ments can result in a diagnosis of autism.45 Thus, candidates, with converging support from multi- altered CNTNAP2 function or regulation could rep- ple lines of independent investigation. That these resent a shared mechanism contributing to lan- first association analyses were positive illustrates guage-related endophenotypes in both specific the promise of our function-based approach. In language impairment and autism. These findings future work, we will go on to assess other neural illustrate the value of using endophenotypes for targets of FOXP2 in a similar manner. the genetic dissection of such disorders. There is considerable debate over the existence In conclusion, by integrating functional ge- of shared causes of neurodevelopmental syn- nomics and quantitative trait analyses, we have dromes that are treated as diagnostically distinct, identified a shared neurogenetic pathway that is such as autism and specific language impair disturbed in distinct forms of language impairment.2,42,43 In this study, we observed an asso- ment. This work represents a move away from ciation between the endophenotype of nonsense- isolated studies of individual genes and toward word repetition and polymorphisms in the exon an understanding of molecular networks that may 13–15 region of CNTNAP2 in children with spe- go awry in neurodevelopmental disorders affect- cific language impairment. A study of the gene ing language. in children with autism37 showed an association Supported by project and program grants from the Wellcome Trust (to Drs. Fisher and Monaco, respectively), a pilot grant between polymorphisms in the exon 13–15 region from Autism Speaks (to Dr. Fisher), awards (MH75028-R21, (similarly centered on tagging-SNP rs2710102) and MH081754-RO1, MH60233-R37, and HD055784-P50) from the the age at the first spoken word. The same SNP National Institute of Mental Health (to Dr. Geschwind), a Chris- topher Welch Biological Sciences Scholarship from the Univer- alleles were associated with susceptibility in both sit y of Oxford (to Dr. Vernes), a postdoctoral fellowship from the studies (Table S3 in the Supplementary Appendix). Tourette Syndrome Association (to Dr. Abrahams), and a Marie Therefore, similar CNTNAP2 variants may repre- Curie Intra-European Fellowship (to Dr. Nicod). Dr. Monaco is a Wellcome Trust Principal Research Fellow, and Dr. Fisher is a sent susceptibility factors for language-related Royal Society Research Fellow. Drs. Newbury, Monaco, and Fisher deficits in both specific language impairment and and Ms. Winchester are members of the Specific Language Im- autism. pairment Consortium. No potential conflict of interest relevant to this article was The CNTNAP2 associations we describe here reported. for specific language impairment are not simply We thank all the families who participated in this study a replication of those reported previously for pa- through the Specific Language Impairment Consortium and other consortium members for their contributions to this work: tients with autistic-spectrum disorders. We made L. Addis, Wellcome Trust Centre for Human Genetics, University a rigorous effort to exclude persons with autistic- of Oxford; I.M. Goodyer, Cambridge Language and Speech Proj- spectrum disorders from our analysis. Although ect; E. Simonoff and P.F. Bolton, Child and Adolescent Psychia- try Department and Medical Research Council Centre for Social, diagnostic boundaries are not always clear, it is Developmental and Genetic Psychiatry, Institute of Psychiatry; unlikely that persons who received a misdiagno- V. Slonims and G. Baird, Newcomen Centre, Guy’s Hospital; sis of autistic-spectrum disorder remained in such A. Everitt, E. Hennessy, M. Main, D. Shaw, and P.J. Helms, Depart- ment of Child Health, University of Aberdeen; A.D. Kindley, numbers that they accounted for the strength of Raeden Centre and Grampian University Hospitals Trust; A. Clark the association that we observed. Moreover, we and J. Watson, Speech and Hearing Sciences, Queen Margaret used a quantitatively defined endophenotype pre- Universit y; A. O’Hare, Department of Reproductive and Develop- mental Sciences, University of Edinburgh; J. Seckl, Molecular viously proposed to underlie typical specific lan- Medicine Centre, University of Edinburgh; H. Cowie, Department guage impairment, rather than a categorical des- of Speech and Language Therapy, Royal Hospital for Sick Chil- dren, Edinburgh; W. Cohen, Department of Educational and ignation of affected status. Professional Studies, University of Strathclyde; J. Nasir, Clinical Instead, our findings are compatible with the Developmental Sciences, St. George’s University of London; idea that different components of autistic-spec- D.V.M. Bishop, Department of Experimental Psychology, Univer- sity of Oxford; Z. Simkin and G. Conti-Ramsden, Human Com- trum disorders (communication def icits, impaired munication and Deafness, School of Psychological Sciences, social interaction, and rigid or repetitive behav- University of Manchester; and M. Falcaro and A. Pickles, Biosta- iors) may be under different genetic influences.44 tistics Group, School of Epidemiology and Health Science, Uni- versity of Manchester; as well as D.V.M. Bishop, G. Baird, G. Conti- In this view, language impairments are observed R a m sden, a nd A . O’H a r e f or t he i r a s si st a nc e w it h t he pr e p a r at ion in relatively pure form in specific language im- of the manuscript.

8 n engl j med 10.1056/nejmoa0802828

References 1. Harel S, Greenstein Y, Kramer U, et al. 16. Lai CS, Gerrelli D, Monaco AP, Fisher Juxtaparanodal clustering of Shaker-like Clinical characteristics of children referred SE, Copp AJ. FOXP2 expression during brain K+ channels in myelinated axons depends to a child development center for evaluation development coincides with adult sites of on Caspr2 and TAG-1. J Cell Biol 2003; of speech, language, and communication pathology in a severe speech and language 162:1149-60. disorders. Pediatr Neurol 1996;15:305-11. disorder. Brain 2003;126:2455-62. 32. Rasband MN. It’s “juxta” potassium 2. Conti-Ramsden G, Simkin Z, Botting 17. Vernes SC, Nicod J, Elahi FM, et al. channel! J Neurosci Res 2004;76:749-57. N. The prevalence of autistic spectrum Functional genetic analysis of mutations 33. Strauss KA, Puffenberger EG, Huentel- disorders in adolescents with a history of implicated in a human speech and language man MJ, et al. Recessive symptomatic focal specif ic language impairment (SLI). J Child disorder. Hum Mol Genet 2006;15:3154-67. epilepsy and mutant contactin-associated Psychol Psychiatry 2006;47:621-8. 18. Vernes SC, Spiteri E, Nicod J, et al. protein-like 2. N Engl J Med 2006;354: 3. Cohen NJ. Language impairment and High-throughput analysis of promoter oc- 1370-7. psychopathology in infants, children, and ad- cupancy reveals direct neural targets of 34. Abrahams BS, Tentler D, Perederiy JV, olescents. Thousand Oaks, CA: Sage, 2001. FOXP2, a gene mutated in speech and lan- Oldham MC, Coppola G, Geschwind DH. 4. Fisher SE, Lai CS, Monaco AP. Deci- guage disorders. Am J Hum Genet 2007; Genome-wide analyses of human perisyl- phering the genetic basis of speech and 81:1232-50. vian cerebral cortical patterning. Proc Natl language disorders. Annu Rev Neurosci 19. Abu-Khalil A, Fu L, Grove EA, Zecevic N, Acad Sci U S A 2007;104:17849-54. 2003;26:57-80. Geschwind DH. Wnt genes define distinct 35. Teramitsu I, Kudo LC, London SE, 5. Lai CS, Fisher SE, Hurst JA, Vargha- boundaries in the developing human brain: Geschwind DH, White SA. Parallel FoxP1 Khadem F, Monaco AP. A forkhead-domain implications for human forebrain pattern- and FoxP2 expression in songbird and hu- gene is mutated in a severe speech and lan- ing. J Comp Neurol 2004;474:276-88. man brain predicts functional interaction. guage disorder. Nature 2001;413:519-23. 20. The SLI Consortium. A genomewide J Neurosci 2004;24:3152-63. 6. MacDermot KD, Bonora E, Sykes N, scan identifies two novel loci involved in 36. Verkerk AJ, Mathews CA, Joosse M, et al. Identification of FOXP2 truncation as specific language impairment. Am J Hum Eussen BH, Heutink P, Oostra BA. a novel cause of developmental speech and Genet 2002;70:384-98. CNTNAP2 is disrupted in a family with language deficits. Am J Hum Genet 2005; 21. Idem. Highly signif icant linkage to the Gilles de la Tourette syndrome and obses- 76:1074-80. SLI1 locus in an expanded sample of indi- sive compulsive disorder. Genomics 2003; 7. Feuk L, Kalervo A, Lipsanen-Nyman M, viduals affected by specific language im- 82:1-9. et al. Absence of a paternally inherited pairment. Am J Hum Genet 2004;74:1225- 37. Alarcón M, Abrahams BS, Stone JL, et FOXP2 gene in developmental verbal dys- 38. al. Linkage, association, and gene-expres- praxia. Am J Hum Genet 2006;79:965-72. 22. Burden V, Stott CM, Forge J, Goodyer sion analyses identify CNTNAP2 as an 8. Shriberg LD, Ballard KJ, Tomblin JB, I. The Cambridge Language and Speech autism-susceptibility gene. Am J Hum Duff y JR, Odell KH, Williams CA. Speech, Project (CLASP). I. Detection of language Genet 2008;82:150-9. prosody, and voice characteristics of a difficulties at 36 to 39 months. Dev Med 38. Arking DE, Cutler DJ, Brune CW, et al. mother and daughter with a 7;13 translo- Child Neurol 1996;38:613-31. A common genetic variant in the neurexin cation affecting FOXP2. J Speech Lang 23. Clark A, O’Hare A, Watson J, et al. Se- superfamily member CNTNAP2 increases Hear Res 2006;49:500-25. vere receptive language disorder in child- familial risk of autism. Am J Hum Genet 9. Zeesman S, Nowaczyk MJ, Teshima I, hood — familial aspects and long-term 2008;82:160-4. et al. Speech and language impairment and outcomes: results from a Scottish study. 39. Bakkaloglu B, O’Roak BJ, Louvi A, et oromotor dyspraxia due to deletion of 7q31 Arch Dis Child 2007;92:614-9. al. Molecular cytogenetic analysis and re- that involves FOXP2. Am J Med Genet A 24. Conti-Ramsden G, Botting N. Charac- sequencing of contactin associated protein- 2006;140:509-14. teristics of children attending language like 2 in autism spectrum disorders. Am J 10. Watkins KE, Dronkers NF, Vargha- units in England: a national study of Hum Genet 2008;82:165-73. Khadem F. Behavioural analysis of an in- 7-year-olds. Int J Lang Commun Disord 40. Spiteri E, Konopka G, Coppola G, et al. herited speech and language disorder: 1999;34:359-66. Identification of the transcriptional tar- comparison with acquired aphasia. Brain 25. Semel EM, Wiig EH, Secord W. Clini- gets of FOXP2, a gene linked to speech 2002;125:452-64. cal evaluation of language fundamentals and language, in developing human brain. 11. Tomblin JB, Records NL, Buckwalter P, — revised. San Antonio, TX: Psychological Am J Hum Genet 2007;81:1144-57. Zhang X, Smith E, O’Brien M. Prevalence Corporation, 1992. 41. Cawley S, Bekiranov S, Ng HH, et al. of specific language impairment in kinder- 26. Gathercole SE, Willis CS, Baddeley Unbiased mapping of transcription factor garten children. J Speech Lang Hear Res AD, Emslie H. The Children’s Test of Non- binding sites along human chromosomes 1997;40:1245-60. word Repetition: a test of phonological 21 and 22 points to widespread regulation 12. Newbury DF, Bishop DV, Monaco AP. working memory. Memory 1994;2:103-27. of noncoding RNAs. Cell 2004;116:499-509. Genetic influences on language impair- 27. Bishop DV, North T, Donlan C. Non- 42. De Fossé L, Hodge SM, Makris N, et ment and phonological short-term memory. word repetition as a behavioural marker al. Language-association cortex asymme- Trends Cogn Sci 2005;9:528-34. for inherited language impairment: evi- try in autism and specific language im- 13. Shriberg LD, Tomblin JB, McSweeny JL. dence from a twin study. J Child Psychol pairment. Ann Neurol 2004;56:757-66. Prevalence of speech delay in 6-year-old Psychiatry 1996;37:391-403. 43. Whitehouse AJ, Barry JG, Bishop DV. children and comorbidity with language 28. Gathercole SE. Nonword repetition and The broader language phenotype of autism: impairment. J Speech Lang Hear Res 1999; word learning: the nature of the relation- a comparison with specific language im- 42:1461-81. ship. App Psycholinguist 2006;27:513-43. pairment. J Child Psychol Psychiatry 2007; 14. Newbury DF, Bonora E, Lamb JA, et al. 29. Abecasis GR, Cardon LR, Cookson 48:822-30. FOXP2 is not a major susceptibility gene WO. A general test of association for 44. Happé F, Ronald A, Plomin R. Time to for autism or specific language impair- quantitative traits in nuclear families. Am give up on a single explanation for autism. ment. Am J Hum Genet 2002;70:1318-27. J Hum Genet 2000;66:279-92. Nat Neurosci 2006;9:1218-20. 15. O’Brien EK, Zhang X, Nishimura C, 30. Abecasis GR, Cherny SS, Cookson WO, 45. Bishop DV. Autism and specific lan- Tomblin JB, Murray JC. Association of spe- Cardon LR. Merlin — rapid analysis of guage impairment: categorical distinction cific language impairment (SLI) to the dense genetic maps using sparse gene f low or continuum? Novartis Found Symp 2003; region of 7q31. Am J Hum Genet 2003;72: trees. Nat Genet 2002;30:97-101. 251:213-26. 1536-43. 31. Poliak S, Salomon D, Elhanany H, et al. Copyright © 2008 Massachusetts Medical Society.

n engl j med 10.1056/nejmoa0802828 9