Molecular Psychiatry (2006) 11, 1085–1091 & 2006 Nature Publishing Group All rights reserved 1359-4184/06 $30.00 www.nature.com/mp ORIGINAL ARTICLE Further evidence that the KIAA0319 confers susceptibility to developmental D Harold1,6, S Paracchini2,6, T Scerri2, M Dennis2,3, N Cope1, G Hill1, V Moskvina4, J Walter5, AJ Richardson5, MJ Owen1, JF Stein5, ED Green3, MC O’Donovan1, J Williams1,4 and AP Monaco2 1Department of Psychological Medicine, Cardiff University, Heath Park, Cardiff, UK; 2Wellcome Trust Centre for Human Genetics, University of Oxford, Roosevelt Drive, Oxford, UK; 3Genome Technology Branch, National Research Institute, Bethesda, MD, USA; 4Biostatistics and Bioinformatics Unit, Cardiff University, Heath Park, Cardiff, UK and 5Department of Physiology, University of Oxford, Parks Road, Oxford, UK

The DYX2 on 6p22.2 is the most replicated region of linkage to developmental dyslexia (DD). Two candidate within this region have recently been implicated in the disorder: KIAA0319 and DCDC2. Variants within DCDC2 have shown association with DD in a US and a German sample. However, when we genotyped these specific variants in two large, independent UK samples, we obtained only weak, inconsistent evidence for their involvement in DD. Having previously found evidence that variation in the KIAA0319 gene confers susceptibility to DD, we sought to refine this genetic association by genotyping 36 additional SNPs in the gene. Nine SNPs, predominantly clustered around the first exon, showed the most significant association with DD in one or both UK samples, including rs3212236 in the 50 flanking region (P = 0.00003) and rs761100 in intron 1 (P = 0.0004). We have thus refined the region of association with developmental dyslexia to putative regulatory sequences around the first exon of the KIAA0319 gene, supporting the presence of functional mutations that could affect gene expression. Our data also suggests a possible interaction between KIAA0319 and DCDC2, which requires further testing. Molecular Psychiatry (2006) 11, 1085–1091. doi:10.1038/sj.mp.4001904; published online 10 October 2006 Keywords: disability; susceptibility locus; genetic association; KIAA0319; DCDC2; epistasis

Introduction ibility locus. In independent samples from the UK and the US, we previously found evidence that Developmental dyslexia (DD [MIM 600202]), or read- variation in the KIAA0319 gene confers risk of ing disability, is a relatively common, complex dyslexia.17–19 A 77 kb region was found to be asso- cognitive disorder that affects 5–10% of school-aged ciated with dyslexia in two independent samples of children.1 The disorder is characterized by an nuclear families from the UK (hereafter referred to as impairment of reading performance despite adequate the Oxford sample) and the US.17 The region motivational, educational and intellectual opportu- extended from the first four exons of KIAA0319 to nities. Although the pathophysiology of DD is the first exon of THEM2, covering the entire TTRAP unknown, there is strong evidence that genes make gene. A specific three-SNP haplotype across this a substantial contribution to individual variation in region was significantly associated with a range of risk of DD, with twin studies reporting heritability reading-related measures in both the Oxford UK and estimates of up to 0.71.1–8 Linkage studies have US samples, particularly when the samples were identified several genomic regions that may harbor selected for severity of phenotype. In a comple- susceptibility quantitative trait loci (QTL) for DD, and mentary study, Cope et al.18 employed a case/control the most consistently replicated of these is DYX2, sample in which a severe definition of dyslexia was which lies on chromosome 6p22.2.1,9–16 used to ascertain probands (hereafter referred to as the A number of association studies have now been Cardiff sample). By performing a systematic, high- performed attempting to identify the DYX2 suscept- density linkage disequilibrium screen of candidate genes within the DYX2 region, we found that Correspondence: Professor J Williams, Department of Psycholo- variation in the KIAA0319 gene was most strongly gical Medicine, Henry Wellcome Building, Cardiff University, implicated in dyslexia susceptibility. Subsequent Heath Park, Cardiff CF14 4XN, UK. functional analysis showed that the risk haplotype E-mail: [email protected] identified by Francks et al.17 is associated with a 6These authors contributed equally to this paper. Received 13 July 2006; revised 31 August 2006; accepted 5 reduction of expression of KIAA0319, but does not September 2006; published online 10 October 2006 affect either THEM2 or TTRAP.19 KIAA0319, DCDC2 and DD D Harold et al 1086 Two recent studies have indicated that another gene et al.17 As previous evidence had indicated that the within the DYX2 QTL, DCDC2, may also play a role in DYX2 locus influences the most severely affected developmental dyslexia. Meng et al.20 identified two with ,17,21,25 we used the same sub- SNPs within DCDC2 (rs807724 and rs1087266) asso- sample of sibships as in Francks et al.,17 that is ciated with a weighted composite measure of reading scoring below a mean measure calculated on phono- and spelling subtests (discriminant score) in a sample logical decoding ability and orthographic coding, of US families. They also described an intronic which were the two measures contributing most to deletion containing a compound simple tandem the linkage signal. This selection yielded 126 families, repeat (STR). The deletion, in combination with 10 including 313 siblings. For the Cardiff sample, rare STR alleles, showed significant association with children with DD were ascertained in Wales and different reading traits. Using a categorical definition England through contacts with local education autho- of dyslexia based on spelling impairment, Schuma- rities and schools specializing in the education of cher et al.21 also found association with variation children with reading difficulties. The inclusion within the DCDC2 gene in two independent trio criteria for probands were an IQ of X85 and a reading samples of German descent, most significantly with a age X2.5 years behind that expected from chrono two-marker haplotype in intron 7 comprised of the logical age. The accuracy score from the Neale SNPs rs793862 and rs807701. This association ap- analysis of reading ability26 was used to determine peared strongest in the most severely affected patients reading age, except when probands were aged > 13 of both samples. years, in which case the accuracy score of British Given the evidence implicating both the KIAA0319 Ability Scale (BAS) single-word reading was used.27 and DCDC2 genes in developmental dyslexia, we Children classed as controls were required to have an aimed to study this genomic region in more detail, in IQ of X85 and a reading delay of no more than 6 two large, independent UK samples. We have analyzed months. A total of 350 cases and 273 controls were a new selection of SNPs covering KIAA0319 and included. As probands were ascertained on the basis DCDC2, including the DCDC2 markers that showed of a severe definition of dyslexia, no further pheno- significant association with DD in previous studies.20,21 typic selection was performed for genetic analysis. The READ measure in the Oxford sample (which is based on BAS single-word reading) is most similar to Materials and methods the Cardiff definition of dyslexia. Comparison of Subjects standardized BAS single-word reading scores in the Individuals with DD were ascertained as previously severe Oxford sibships and all Cardiff cases indicated described.17,18 Briefly, the Oxford sample included a similar level of reading disability (data not shown). 264 unrelated nuclear families identified from the All participants in both samples were collected dyslexia clinic at the Royal Berkshire Hospital in within the UK and were of white European descent. Reading, UK. The majority of families were recruited Ethical approval for the Oxford and Cardiff sample on the basis of having at least one proband whose was obtained from the NHS Research Ethics Commit- single-word reading ability was > 2 s.d. below that tee and the Multi-Centre Research Ethics Committee predicted by tests of verbal or nonverbal reasoning22,23 for Wales, respectively. Appropriate and informed and where at least one sibling displayed evidence of written consent was obtained from all participants. DD. The remaining families were recruited through probands that were required to have single-word Genotyping reading ability X1 s.d. below that predicted for their The Cardiff group identified polymorphisms within age, with a minimum IQ of 90. All probands and KIAA0319 through dbSNP or by denaturing high- siblings were administrated a battery of psychometric performance liquid chromatography as previously tests and we age-adjusted and standardized their described,28 while the Oxford group selected new scores against a normative control data set as markers that were located within multi-species con- previously described.22,23 These included the follow- served sequences (MCSs).29 SNPs with a minor ing measures used in the present study for quantita- allele frequency < 5% were included if they were of tive analysis: orthographic coding using irregular putative functional significance (e.g. lying within words (OC-irreg), phonological decoding ability MCSs), and as such could be directly associated with (PD), orthographic coding assessed by a forced word DD. SNP genotyping was performed using either the choice test (OC-choice), single-word reading ability MassARRAY system (Sequenom) or the Amplifluor (READ), spelling ability (SPELL), phonemic aware- SNPs Genotyping System (Serologicals), in accor- ness (PA), and tests of verbal (SIM) and nonverbal dance with the manufacturers’ instructions. For (MAT) reasoning. Multivariate linkage analysis had Sequenom genotyping, PCR primers and primer- previously shown that the DYX2 QTL influenced extension probes were designed with the Spectro- variability shared by all the DD measures but did not DESIGNER software. For Amplifluor genotyping, influence IQ,24 and univariate quantitative linkage primers were designed using Amplifluor AssayArch- analysis using IQ-adjusted traits refined the linkage itect. All primer and probe sequences are available on signal.17 Therefore, all quantitative measures used request. The DCDC2 intronic deletion and STR were in the present study were IQ-adjusted as in Francks genotyped as described by Meng et al.20

Molecular Psychiatry KIAA0319, DCDC2 and DD D Harold et al 1087 Data analysis German and US studies20,21 that showed the most In the Oxford sample, marker-trait association was significant association either as single markers, or as evaluated using the total association model within part of a haplotype. Markers genotyped in both the the QTDT package30 and haplotypes were estimated Oxford and Cardiff samples were the two SNPs using MERLIN.31 Standard contingency tables were (rs807724 and rs1087266) and the intronic deletion used for single-marker case/control analysis in the identified by Meng et al.,20 and the two SNPs forming Cardiff sample; haplotypes were analyzed using the haplotype implicated by Schumacher et al.21 UNPHASED.32 Bonferroni corrected P-values in the (rs793862 and rs807701). The compound STR Oxford sample are based on 57 tests (Supplementary described by Meng et al.20 was also genotyped in the Tables 1 and 2); corrected P-values in the Cardiff Cardiff sample. Nominally significant associations sample are based on 42 tests (Supplementary Table 3). were detected with several traits in the complete Where markers showed significant association with Oxford sample of 264 families with the common dyslexia in both the Oxford and Cardiff sample alleles of rs793862, rs807724 and rs1087266 (Supple- (Pp0.05), P-values were combined using Fisher’s mentary Table 1). The strongest association was 33 combined probability test. Logistic regression mod- observed between rs1087266 and PA (P = 0.005). The els were used to test for pairwise locus–locus two-marker haplotype described by Schumacher interactions between genes using Genetic Association 21 34 et al. did not show any significant association, even Interaction Analysis software. The GoldSurfer pack- though marker rs793862 showed modest association age was used to simultaneously evaluate pairs of in single marker analysis (P = 0.04, PD; P = 0.04, linkage disequilibrium features with a color-coded READ, P = 0.02, SPELL). However, in the subset of three-dimensional graphical interface.35 severe Oxford families, these associations were no longer significant and the only marker that showed a nominally significant association was the intronic Results deletion (Table 1, Supplementary Table 2). No DCDC2 analysis association with DD was observed for any of the To attempt replication of the association observed in DCDC2 variants analyzed in the Cardiff sample DCDC2, we selected those polymorphisms from the (Table 1), in agreement with our previous study in

Table 1 Association results for markers genotyped both in the severe Oxford sample and in the Cardiff sample

Marker Gene Location Selected sibships from the Oxford sample Cardiff cases Combined N=313 siblings, 126 families and controls P-valuea N=350/N=273

P-value for trait

Risk OC-irreg PD OC- READ SPELL PA Risk P-value alleleb choice alleleb rs793862_ DCDC2 Intron 7 rs807701 rs793862 DCDC2 Intron 7 rs807701 DCDC2 Intron 7 rs807724 DCDC2 Intron 6 DCDC2_ DCDC2 Intron 3 2 0.0298 0.0428 0.0478 deletion rs1087266 DCDC2 Intron 2 rs4504469 KIAA0319 Exon 4 1 0.0011 0.0082 0.004 0.01 1 0.0051 0.00024 rs2179515 KIAA0319 Intron 1 1 0.0012 0.0131 0.0004 0.0232 1 0.0215 0.00011 rs761100 KIAA0319 Intron 1 1 0.0027 0.0405 0.0107 0.0078 0.0306 1 0.0004 0.00004 rs7766230 KIAA0319 Intron 1 2 0.0049 0.0002 0.0017 0.0096 rs6935076 KIAA0319 Intron 1 2 0.0073 rs17491230c KIAA0319 Intron 1 2 0.0023 0.0009 0.0041 rs2038137 KIAA0319 Intron 1 1 0.0013 0.0026 0.0002 0.0061 1 0.0052 0.00002 rs1555090 Intergene 1 0.001 0.0029 0.0003 0.0131 1 0.0419 0.00015 rs3212236 Intergene 2 0.0006 0.00003 0.0005 0.0028 rs2143340 TTRAP Intron 2 2 0.01 0.0003 0.0115 aCombination of Cardiff P-value and Oxford READ P-value using Fisher’s combined probability test.33 bOnly alleles with nominally significant marker-trait association (P < 0.05) are shown. cEquivalent to rs2235676 in Francks et al.17

Molecular Psychiatry KIAA0319, DCDC2 and DD D Harold et al 1088 Table 2 DCDC2–KIAA0319 interaction analysis in 419 Oxford/Cardiff cases and 273 Cardiff controls

P-valuesa for interaction terms over and above main effects of markers

Marker rs4504469 rs2179515 rs761100 rs2038137 rs1555090

rs793862 0.039 0.152 0.007 0.190 0.111 rs807701 0.033 0.486 0.427 0.546 0.306 rs807724 0.612 0.516 0.122 0.764 0.628 DCDC2_deletionb 0.465 0.523 0.446 0.356 0.427 rs1087266 0.812 0.579 0.236 0.331 0.485

aP-values p0.05 are highlighted in bold. bAdditive only model.

which 30 SNPs across DCDC2 were examined using a (P = 0.0004; Bonferroni corrected P = 0.02), which is DNA pooling strategy.18 The intronic deletion showed located 13 kb downstream of the first KIAA0319 exon. no evidence of association in the Cardiff sample, rs761100 was also significantly associated with five having a frequency of 9% both in cases and controls, reading traits in the severe sub-group from the Oxford which is identical to that seen in the Oxford severe sample, in particular with OC-irreg (P = 0.0027). In dyslexic probands (data not shown). Nor was any total, five SNPs were significantly associated with DD association with DD observed with the deletion/STR in both samples; in each case the major allele is polymorphism either globally (P = 0.4236) or with the associated with poorer reading skills. P-values for specific risk ‘allele’ comprised of the deletion and association with the READ trait in the severe Oxford rare STR alleles as described in Meng et al.20 sample and association with DD in the Cardiff sample (P = 0.7468; see Supplementary Table 3). Thus, our were combined for these five SNPs using Fisher’s study finds only weak, inconsistent evidence impli- combined probability test.33 Consequently, the most cating DCDC2 in DD susceptibility. Although we significant association with DD in the combined cannot rule out the involvement of this gene in analysis is with the intron 1 SNP rs2038137 developmental dyslexia, as a single locus it is (P = 0.00002). A further four SNPs in the 50 flanking unlikely to play a major role in the disorder in the region or intron 1 showed association with the minor UK population. allele, but only in one of the two UK samples. This may be a result of the differing ascertainment criteria, KIAA0319 analysis phenotypic tests or statistical analyses used by the We also undertook a detailed genetic study of the two groups. Another explanation could be that, since KIAA0319 gene. Employing a combination of SNPs the association with minor alleles was due to a selected from databases, identified through mutation smaller number of individuals, a few missing geno- screening or falling within MCSs, the Cardiff and types or missing phenotype information could affect Oxford groups independently genotyped different association results. sets of SNPs. In total, 12 new polymorphisms in or flanking the KIAA0319 gene were genotyped in the KIAA0319–DCDC2 interaction analysis Oxford sample and 24 new polymorphisms were We tested for statistical interactions between markers genotyped in the Cardiff sample. Polymorphisms in DCDC2 with the five SNPs in KIAA0319 showing showing most evidence for association in the Oxford association with DD in both the Oxford and Cardiff sample were also genotyped in the Cardiff sample, sample. This analysis was performed using a case/ and vice versa (Table 1). We found that variation in control design. In order to incorporate both the the 20 kb of sequence around the first exon is Oxford and Cardiff samples, unrelated probands from predominantly responsible for the association we the Oxford or Cardiff sample scoring X1 s.d. below observe between KIAA0319 and DD (Figure 1, Table the population mean on the BAS single word reading 1). In the complete Oxford sample, the strongest test were selected as cases (N = 419) and were association was detected with the marker rs3212236 compared to the 273 Cardiff controls. Where possible, (P = 0.002, OC-choice; Supplementary Table 1), lo- marker–marker interactions were tested using cated approximately 2 kb upstream of the 50 end of interaction models allowing for both additive and KIAA0319. After selecting for severity, rs3212236 still dominance effects; for markers with small minor showed the most significant association with OC- allele frequencies it was not possible to fit all the choice (P = 0.00003; Bonferroni corrected P = 0.002) interaction terms and therefore an additive only and was also highly significantly associated with model was tested. Significant interactions were OC-irreg (P = 0.0006), READ (P = 0.0005) and SPELL observed between variants in the two genes (Table 2, (P = 0.0028) (Table 1, Supplementary Table 2). In Supplementary Table 4), most significantly between the Cardiff sample, the strongest association was rs793862 in DCDC2 and rs761100 in KIAA0319 observed with the rs761100 intronic marker (P = 0.007).

Molecular Psychiatry KIAA0319, DCDC2 and DD D Harold et al 1089 a Oxford pter cen DCDC2 KIAA0319 Intron 1 − 5' flanking

P = 0.05

b Cardiff

pter cen

DCDC2 KIAA0319 Intron 1 − 5' flanking

P = 0.05

Figure 1 LD evaluation in the Oxford and Cardiff samples in a three-dimensional color-coded plot. Colors represent D0 values: green indicates high LD and blue indicates low LD. The height of the peaks represents r2; high peaks correspond to high LD. The red ribbon represents the significance of SNP associations, according to –ln (P-value). (a) The Oxford plot is derived from the analysis of 56 markers and the red ribbon shows the P-value for association with the READ trait in the severe sub-group of families (Supplementary Table 2). (b) The Cardiff plot is derived from the analysis of 41 markers (Supplementary Table 3).

Discussion relatively small, and any one of the associated polymorphisms alone is unlikely to fully explain the We have examined KIAA0319 and DCDC2, the two consistent evidence for linkage at the DYX2 locus. putative susceptibility genes for developmental dys- However, the observed linkages may reflect variation lexia at the DYX2 locus on chromosome 6p22.2, and at multiple sites within KIAA0319 (or in multiple our findings provide strong support for a role of the genes in the linkage region). There is a high level of KIAA0319 gene in DD. Five SNPs in this gene show LD between all markers within the region of strong significant association with DD in two independent association with DD in both the Oxford and Cardiff UK samples from Oxford and Cardiff and a further samples, implying that the functional genetic var- four SNPs show association in one of the samples. iant(s) may be difficult to identify by further genetic The effect sizes associated with these SNPs are mapping. Mutation screening of the exons and the 50

Molecular Psychiatry KIAA0319, DCDC2 and DD D Harold et al 1090 flanking region of KIAA0319 failed to reveal any gene expression. Preliminary evidence also suggests a obvious disruptive mutations.17,18 It is possible that possible interaction between KIAA0319 and DCDC2, the true susceptibility variants have a more subtle although this requires confirmation in independent effect. The association is clustered in sequences samples. around the first exon that could be involved in gene regulation, suggesting the presence of variants affect- Acknowledgments ing control of gene expression. This observation is in agreement with our recent functional analysis show- We thank all the parents and children who took part ing that the risk haplotype previously described for in this study. We also thank Fredrik Pettersson for DD17 is likely to carry genetic variants that influence helping with GoldSurfer. This work was supported KIAA0319 expression.19 The KIAA0319 transcript by a Wellcome Trust Principal Research Fellowship generated from the risk haplotype was observed to to APM, grants to JWilliams, MOD and MO from be at a significantly lower level in comparison to the Health Foundation and the Medical Research transcripts derived from other haplotypes. Council, and funding to MD in part by the Intramural In our examination of the DCDC2 gene, we Research Program of the National Human Genome genotyped the specific variants that have shown Research Institute of the National Institutes of Health. association with DD in previous studies.20,21 Although we only find weak, inconsistent evidence implicating Electronic Database Information DCDC2 in the UK population, we do not exclude it as a DD susceptibility gene. Interestingly KIAA0319 and The URLs for data presented herein are as follows: DCDC2 seem to share functional similarities: the Amplifluor AssayArchitect, https://apps.serologicals. produced by both genes appear to play a com/AAA/ role in neuronal migration, which is a key step in the dbSNP Home Page, http://www.ncbi.nlm.nih.gov/ development of the neocortex.19,20 During this pro- SNP/ cess, neurons migrate to reach their final destination Genetic Association Interaction Analysis, http:// in the highly structured organization of the cerebral www.bbu.cf.ac.uk/html/research/biostats.htm cortex. Neuronal migration has been proposed as a Online Mendelian Inheritance in Man (OMIM), neurological mechanism implicated in the develop- http://www.ncbi.nlm.nih.gov/Omim/ ment of DD. Galaburda et al.36,37 showed that post- Multi-species conserved sequences (MCSs), http:// mortem brains of dyslexic individuals were charac- hgwdev-elliott.cse.ucsc.edu/cgi-bin/hgGateway and terized by the presence of subtle cortical anomalies, http://genome.lbl.gov/vista/index.shtml. which resulted from defective neuronal migration. Given the proximity of the genes and the functional similarity of the proteins, it is possible that both References KIAA0319 and DCDC2 are involved in the develop- ment of dyslexia. The DYX2 locus has been identified 1 Williams J, O’Donovan MC. The genetics of developmental dyslexia. Eur J Hum Genet 2006; 14: 681–689. by different linkage studies, and as previously 2 Fisher JH. Case of congenital word blindness (inability to learn to mentioned, there may be more than one gene read). Ophthal Rev 1905; 24: 315. responsible for the QTL effect. It is possible, for 3 Hinshelwood J. Four cases of hereditary word-blindness occurring example, that the two genes contribute to different in the same family. Br Med J 1907; 2: 1229–1232. 4 DeFries JC, Fulker DW, LaBuda MC. Evidence for a genetic subsets of individuals with dyslexia and that specific aetiology in reading disability of twins. Nature 1987; 329: ascertainment criteria or phenotypic characterization 537–539. might drive genetic associations towards one gene 5 Stevenson J, Graham P, Fredman G, McLoughlin V. A twin study of rather than the other. Moreover, epistatic interactions genetic influences on reading and spelling ability and disability. between variants in the two genes may occur that J Child Psychol Psychiatry 1987; 28: 229–247. 6 DeFries JC, Olson R, Pennington BF, Smith SD. Colorado reading contribute to disease risk. Indeed, we observed project: past, present, and future. Learn Disabil 1991; 2: 37–46. significant statistical interactions between variants 7 Pennington BF, Gilger JW, Pauls D, Smith SA, Smith SD, DeFries in the two genes. Although the P-values do not JC. Evidence for major gene transmission of developmental withstand correction for multiple testing, given the dyslexia. JAMA 1991; 266: 1527–1534. 8 Schulte-Korne G, Deimel W, Muller K, Gutenbrunner C, functional similarity of the proteins, this is a Remschmidt H. Familial aggregation of spelling disability. potentially important result. It does, however, require J Child Psychol Psychiatry 1996; 37: 817–822. replication in an independent sample. A more robust 9 Cardon LR, Smith SD, Fulker DW, Kimberling WJ, Pennington BF, interaction may be observed between the functional DeFries JC. Quantitative trait locus for reading disability on susceptibility variants in the two genes. . Science 1994; 266: 276–279. 10 Cardon LR, Smith SD, Fulker DW, Kimberling WJ, Pennington BF, In conclusion, two candidate genes within the DeFries JC. Quantitative trait locus for reading disability: correc- DYX2 locus have been reported to confer suscept- tion. Science 1995; 268: 1553. ibility to DD; KIAA0319 and DCDC2. Through a 11 Grigorenko EL, Wood FB, Meyer MS, Hart LA, Speed WC, Shuster collaborative effort we have provided compelling A et al. Susceptibility loci for distinct components of develop- mental dyslexia on 6 and 15. Am J Hum Genet 1997; evidence for a role of the KIAA0319 gene in DD. In 60: 27–39. particular, our data support the presence of functional 12 Fisher SE, Marlow AJ, Lamb J, Maestrini E, Williams DF, mutations that could have an impact on KIAA0319 Richardson AJ et al. A quantitative-trait locus on chromosome

Molecular Psychiatry KIAA0319, DCDC2 and DD D Harold et al 1091 6p influences different aspects of developmental dyslexia. Am J 24 Marlow AJ, Fisher SE, Francks C, MacPhie IL, Cherny SS, Hum Genet 1999; 64: 146–156. Richardson AJ et al. Use of multivariate linkage analysis for 13 Gayan J, Smith SD, Cherny SS, Cardon LR, Fulker DW, Brower AM dissection of a complex cognitive trait. Am J Hum Genet 2003; 72: et al. Quantitative-trait locus for specific language and reading 561–570. deficits on chromosome 6p. Am J Hum Genet 1999; 64: 157–164. 25 Deffenbacher KE, Kenyon JB, Hoover DM, Olson RK, Pennington 14 Grigorenko EL, Wood FB, Meyer MS, Pauls DL. Chromosome 6p BF, DeFries JC et al. Refinement of the 6p21.3 quantitative trait influences on different dyslexia-related cognitive processes: locus influencing dyslexia: linkage and association analyses. Hum further confirmation. Am J Hum Genet 2000; 66: 715–723. Genet 2004; 115: 128–138. 15 Kaplan DE, Gayan J, Ahn J, Won TW, Pauls D, Olson RK et al. 26 Neale MD. Analysis of Reading Ability, revised british edition Evidence for linkage and association with reading disability on (British adaptation and standardisation by Una Christophers and 6p21.3–22. Am J Hum Genet 2002; 70: 1287–1298. Chris Whetton). NFER-Nelson: Windsor, United Kingdom, 1989. 16 Grigorenko EL, Wood FB, Golovyan L, Meyer M, Romano C, Pauls 27 Elliot CD. British Ability Scales. NFER-Nelson: Windsor, United D. Continuing the search for dyslexia genes on 6p. Am J Med Genet Kingdom, 1993. B Neuropsychiatr Genet 2003; 118: 89–98. 28 Harold D, Peirce T, Moskvina V, Myers A, Jones S, Hollingworth P 17 Francks C, Paracchini S, Smith SD, Richardson AJ, Scerri TS, et al. Sequence variation in the CHAT locus shows no association Cardon LR et al. A 77-kilobase region of chromosome 6p22.2 is with late-onset Alzheimer’s disease. Hum Genet 2003; 113: associated with dyslexia in families from the United Kingdom and 258–267. from the United States. Am J Hum Genet 2004; 75: 1046–1058. 29 Margulies EH, Green ED. Detecting highly conserved regions of the 18 Cope N, Harold D, Hill G, Moskvina V, Stevenson J, Holmans P human genome by multispecies sequence comparisons. Cold et al. Strong evidence that KIAA0319 on chromosome 6p is a Spring Harb Symp Quant Biol 2003; 68: 255–263. susceptibility gene for developmental dyslexia. Am J Hum Genet 30 Abecasis GR, Cardon LR, Cookson WO. A general test of 2005; 76: 581–591. association for quantitative traits in nuclear families. Am J Hum 19 Paracchini S, Thomas A, Castro S, Lai C, Paramasivam M, Wang Y Genet 2000; 66: 279–292. et al. The chromosome 6p22 haplotype associated with dyslexia 31 Abecasis GR, Cherny SS, Cookson WO, Cardon LR. Merlin–rapid reduces the expression of KIAA0319, a novel gene involved in analysis of dense genetic maps using sparse gene flow trees. Nat neuronal migration. Hum Mol Genet 2006; 15: 1659–1666. Genet 2002; 30: 97–101. 20 Meng H, Smith SD, Hager K, Held M, Liu J, Olson RK et al. DCDC2 32 Dudbridge F. Pedigree disequilibrium tests for multilocus haplo- is associated with reading disability and modulates neuronal types. Genet Epidemiol 2003; 25: 115–121. development in the brain. Proc Natl Acad Sci USA 2005; 102: 33 Fisher RA. Statistical Methods for Research Workers. Oliver and 17053–17058. Boyd: Edinburgh, 1932. 21 Schumacher J, Anthoni H, Dahdouh F, Konig IR, Hillmer AM, 34 Macgregor S, Khan IA. GAIA: an easy-to-use web-based applica- Kluck N et al. Strong genetic evidence of DCDC2 as a susceptibility tion for interaction analysis of case-control data. BMC Med Genet gene for dyslexia. Am J Hum Genet 2006; 78: 52–62. 2006; 7:34. 22 Marlow AJ, Fisher SE, Richardson AJ, Francks C, Talcott JB, 35 Pettersson F, Jonsson O, Cardon LR. GOLDsurfer: three dimen- Monaco AP et al. Investigation of quantitative measures related to sional display of linkage disequilibrium. Bioinformatics 2004; 20: reading disability in a large sample of sib-pairs from the UK. Behav 3241–3243. Genet 2001; 31: 219–230. 36 Galaburda AM, Kemper TL. Cytoarchitectonic abnormalities in 23 Fisher SE, Francks C, Marlow AJ, MacPhie IL, Newbury DF, developmental dyslexia: a case study. Ann Neurol 1979; 6: 94–100. Cardon LR et al. Independent genome-wide scans identify a 37 Galaburda AM, Sherman GF, Rosen GD, Aboitiz F, Geschwind N. chromosome 18 quantitative-trait locus influencing dyslexia. Nat Developmental dyslexia: four consecutive patients with cortical Genet 2002; 30: 86–91. anomalies. Ann Neurol 1985; 18: 222–233.

Supplementary Information accompanies the paper on the Molecular Psychiatry website (http:// www.nature.com/mp)

Molecular Psychiatry