Strong Evidence That KIAA0319 on Chromosome 6p Is a Susceptibility Gene for Developmental Dyslexia

Abstract

Linkage between developmental dyslexia (DD) and chromosome 6p has been replicated in a number of independent samples. Recent attempts to identify the gene responsible for the linkage have produced inconsistent evidence for association of DD with a number of genes in a 575-kb region of chromosome 6p22.2, including VMP, DCDC2, KIAA0319, TTRAP, and THEM2. We aimed to identify the specific gene or genes involved by performing a systematic, high-density (∼2–3-kb intervals) linkage disequilibrium screen of these genes in an independent sample, incorporating family-based and case-control designs in which dyslexia was defined as an extreme representation of reading disability. Using DNA pooling, we first observed evidence for association with 17 single-nucleotide polymorphisms (SNPs), 13 of which were located in the KIAA0319 gene (P<.01–.003). After redundant SNPs were excluded, 10 SNPs were individually genotyped in 223 subjects with DD and 273 controls. Those SNPs that were significant at P⩽.05 were next genotyped in a semi-independent sample of 143 trios of probands with DD and their parents, to control for possible population stratification. Six SNPs showed significant evidence of association in both samples (P⩽.04–.002), including a SNP (rs4504469) in exon 4 of the KIAA0319 gene that changes an amino acid (P=.002; odds ratio 1.5). Logistic regression analysis showed that two SNPs (rs4504469 and rs6935076) in the KIAA0319 gene best explained DD status. The haplotype composed of these two markers was significantly associated with DD (global P=.00001 in the case-control sample; P=.02 in trios). This finding was largely driven by underrepresentation of the most common haplotype in cases (P=.00003 in the case-control sample; P=.006 in trios; 1–degree-of-freedom tests). Our data strongly implicate KIAA0319 as a susceptibility gene for dyslexia. The gene product is expressed in brain, but its specific function is currently unknown.

Introduction

Developmental dyslexia (DD [MIM 600202]), or reading disability, is a relatively common, complex cognitive disorder that affects 5%–10% of school-aged children (Shaywitz et al. 1992). The disorder is characterized by an impairment of reading performance despite adequate motivational, educational, and intellectual opportunities and in the absence of sensory or neurological disability. Although the pathophysiology of DD is unknown, there is strong evidence that genes make a substantial contribution to individual variation in risk of DD, with twin studies reporting heritability estimates of up to 0.71 (Fisher 1905; Hinshelwood 1907; DeFries et al. 1987, 1991; Stevenson et al. 1987; Pennington et al. 1991; Schulte-Körne et al. 1996). The mode of transmission is unknown, but DD is almost certainly a complex genetic disorder in which multiple genes play a role (Hohnen and Stevenson 1999).

Genetic linkage and association studies have implicated a number of chromosomal regions that may harbor susceptibility genes for DD. Regions showing replicated evidence for a role in DD include chromosome 1p (Rabin et al. 1993; Grigorenko et al. 2001; Tzenova et al. 2004), 2p (Fagerheim et al. 1999; Francks et al. 2002; Petryshen et al. 2002; Kaminen et al. 2003; Chapman et al. 2004), 6p (Cardon et al. 1994, 1995; Grigorenko et al. 1997, 2000, 2003; Fisher et al. 1999; Gayán et al. 1999; Kaplan et al. 2002; Turic et al. 2003; Chapman et al. 2004), 15q (Grigorenko et al. 1997; Schulte-Körne et al. 1998; Morris et al. 2000), and 18p (Fisher et al. 2002; Marlow et al. 2003; Chapman et al. 2004).

We have sought to identify a gene(s) in the most consistently supported region on chromosome 6p that shows association with DD. The broadest evidence for linkage stretches from marker D6S109 (Grigorenko et al. 1997) to marker D6S291 (Fisher et al. 1999), a distance of ∼16 Mb, with numerous studies implicating regions between these markers (D6S105–TNFB [Cardon et al. 1994, 1995], D6S109–D6S306 [Grigorenko et al. 1997] D6S276–D6S105 [Gayán et al. 1999] D6S464–D6S273 [Grigorenko et al. 2000], D6S109–JA01, D6S299–D6S1621, and D6S105–D6S265 [Grigorenko et al. 2003]). A region between D6S461 and D6S105 shows greatest overlap between studies, spanning ∼4.2 Mb.

Recently, Deffenbacher and colleagues (2004) examined 10 candidate genes (VMP, DCDC2, MRS2L, GPLD1, ALDH5A1, KIAA0319, TTRAP, THEM2, C6orf62, and GMNN) in the ∼680 kb between markers D6S276 and D6S1554, a region that yielded maximal evidence for linkage in their sample. Thirty-one SNPs were selected at a marker density of ∼1 marker per 20 kb; 13 of the SNPs provided some evidence for association with at least one of the phenotypes analyzed (discrepancy score, phoneme awareness, phoneme deletion, word reading, and orthographic coding). These SNPs were located in five genes: two in VMP, eight in DCDC2, and one each in KIAA0319, TTRAP, and THEM2. Francks and colleagues (2004) examined eight genes (ALDH5A1, KIAA0319, TTRAP, THEM2, C6orf32, SCGN, BTN3A1, and BTN2A1) on chromosome 6p as candidates for DD on the basis of known brain expression. Fifty-seven SNPs were analyzed in 89 U.K. families (sample 1). Association was detected in a 77-kb region spanning TTRAP and the first four exons of KIAA0319. In a second U.K. sample of 175 families (sample 2), 20 of the SNPs were analyzed; weaker evidence of association was found. It is noteworthy that the effect was most pronounced in a sample of families with probands representing the lower end of the reading-ability spectrum. Twenty-one SNPs were also analyzed in a U.S. sample of 159 families (sample 3). Again, association was observed in selected families representing the lower end of the reading-ability spectrum. A three-marker haplotype was significantly associated in both the U.S. sample and the combined U.K. sample, with different alleles forming the most significant haplotype in each sample.

Our study focused on the 575-kb region of chromosome 6p22.2, which was selected to include all genes implicated in recent candidate-gene association studies (Deffenbacher et al. 2004; Francks et al. 2004) and is situated within a region that shows the most consistent evidence of linkage (Deffenbacher et al. 2004; Francks et al. 2004). Our analysis employs a high-density (137 SNPs in seven genes, at 2–3-kb intervals) screen for linkage disequilibrium (LD) and uses both case-control and family-based designs to take account of possible population stratification.

The targeted genes comprise vesicular membrane protein p24 (VMP), doublecortin domain–containing 2 (DCDC2), kidney-associated antigen 1 (KAAG1), magnesium homeostasis factor (MRS2L), KIAA0319, TRAF and TNF receptor associated protein (TTRAP), thioesterase superfamily member 2 (THEM2), and chromosome 6 ORF 62 (C6orf62).

VMP is a neuron-specific vesicular membrane protein that is thought to play a role in vesicular organelle transport and neurotransmission (Cheng et al. 2002). DCDC2 is a ubiquitously expressed gene with a doublecortin-homology domain. Doublecortin itself has been implicated as a cause of X-linked lissencephaly (Gleeson et al. 1998) and is involved in neuronal migration in the CNS, including in the cortex (Gleeson et al. 1999).

KAAG1 is encoded on the strand that is opposite to—and overlaps with—DCDC2. It is a kidney antigen–associated gene, found in numerous tumors and normal testis and kidney (Van Den Eynde et al. 1999). Although it has no known CNS function, it was included in our analysis since it is encoded on the strand that is opposite to the DCDC2 gene and since it was covered by the SNP grid encompassing that gene. MRS2L is a ubiquitously expressed gene that is thought to encode a magnesium-transporter protein (Zsurka et al. 2001).

KIAA0319 is a protein of unknown function that is highly expressed in brain (Londin et al. 2003). The four polycystic kidney disease (PKD) domains found in KIAA0319 show homology to the extracellular domains of the PKD protein PKD1, which are involved in cell-adhesive functions (Streets et al. 2003).

TTRAP encodes a tumor necrosis factor receptor–associated protein. It has been shown to inhibit nuclear factor-κB (NF-κB) activation and subsequent downstream activation of transcription (Pype et al. 2000). NF-κB transcription has been shown to play a role in long-term potentiation and synaptic plasticity associated with learning and memory. In mice, inhibition of NF-κB has been shown to result in neurodegenerative-like phenotypes (Fridmacher et al. 2003). TTRAP can also interact with the cytoplasmic TNF receptor–associated factors (TRAFs) and with cytoplasmic domains of some members of the TNF-receptor superfamily (Pype et al. 2000).

THEM2 encodes an uncharacterized hypothalamus protein and is part of the thioesterase superfamily. The thioesterase superfamily catalyzes the hydrolysis of long-chain fatty acyl-CoA thioesters. It has been suggested that abnormal fatty-acid metabolism plays a role in DD (Richardson and Ross 2000; Richardson et al. 2000; Taylor and Richardson 2000).

C6orf62 is a gene with unknown function that is expressed ubiquitously, including in brain. Although not part of the present study, aldehyde dehydrogenase 5 family, member A1 (ALDH5A1 [succinate-semialdehyde dehydrogenase]), and glycosylphosphatidylinositol-specific phospholipase D1 (GPLD1) on 6p22.2-p22.3, were previously examined using a direct gene-analysis approach that is based on de novo polymorphism discovery and analysis of all detected variants. Extensive analysis of both genes failed to provide evidence for association with DD (authors' unpublished data).

This study thus provides a systematic, high-density LD screen spanning putative functional candidates in a region that shows the most consistent evidence of linkage to DD. Our strategy employed both case-control and family-based designs in which dyslexia was defined, to capture the extreme end of the reading ability/disability continuum.

Material and Methods

Ethical approval was obtained from local ethics committees in the United Kingdom; appropriate and informed written consent was obtained from subjects' parents. All participants were of white U.K. origin. Children with DD—and, if available, their parents and siblings—were ascertained in the United Kingdom through contacts with local education authorities and schools specializing in the education of children with reading difficulties. The inclusion criteria for probands were an IQ of ⩾85 and a reading age ⩾2.5 years behind that expected from chronological age. No age-IQ discrepancy measures were employed. Four subtests from the WISC-III UK were used to provide a prorated, full-scale IQ score: Vocabulary, Similarities, Block Design, and Picture Completion (Wechsler 1992). The accuracy score from the Neale (1989) analysis of reading ability was used to determine reading age, except when probands were aged >13 years, in which case, we used the accuracy score of British Ability Scale (BAS) single-word reading (Elliot 1983).

Initially, controls were adult white U.K. blood donors. Subsequently, control children matched for age and sex were ascertained from the same schools as the children with DD. Children classed as controls were required to have an IQ of ⩾85 and a reading delay (RD) of no more than 6 mo. Control children were assessed using the accuracy score of the Neale (1989) analysis of reading ability and/or BAS single-word reading (Elliot 1983) to calculate reading age; the Vocabulary, Similarities, Block Design, and Picture Completion subsets of WISC-III UK enabled calculation of prorated IQ.

As a first pass, SNPs were analyzed in DNA pools. Because of sample availability, early analyses (n=56 SNPs) used case pools containing 140 unrelated probands with DD (mean age [± SD] 13.22±2.3 years; mean RD [± SD] -5.07±1.76; mean IQ [± SD] 100.13±11.03; 116 males, 24 females) and 550 adult blood-donor controls (mean age 41.39±12.5 years; 391 male, 159 female) (fig. 1, Start Point A). Later, pooled analyses were based on an extension of the original sample of subjects with DD (n=240 unrelated probands with DD; mean age 13.17±2.18 years; mean IQ 98.88±18.38; mean RD -4.93±1.87; 204 males, 36 females) and pools containing 312 age-matched and screened controls (mean age 11.98±2.39 years; mean IQ 103.35±11.97; mean RD +1.14±1.45; 178 males, 134 females) (fig. 1, Start Point B).

Figure 1.

Flow diagram of the samples used at each stage of the analysis

DNA was extracted from venous blood or 25 ml saline mouthwashes by use of standard procedures (Morris et al. 2000). SNPs were selected from Ensembl or SNPper (CHIP Bioinformatics Tools) for each of the eight positional/functional candidate genes, at intervals of ∼2–3 kb between each SNP. SNPs were also chosen for analysis if they showed association in the study by either Deffenbacher et al. (2004) or Francks et al. (2004), excluding SNPs in LD (r²⩾0.80) with those we had already typed on the basis of HapMap data. A total of 137 SNPs were analyzed across the region. With the exception of DCDC2, SNP grids extended from 3 kb upstream to the predicted start of transcription across all exons and introns. Introns 2, 7, and 8 of DCDC2 would have required 60–90 SNPs. Therefore, for pragmatic reasons, we restricted our analysis at these introns to 3 kb of flanking sequence on either side of each of the exons.

Genotyping of DNA pools was undertaken using the SNaPshot (Applied Biosystems) primer-extension method described by Norton and colleagues (2002). Pools were carefully constructed by a serial dilution method, each stage accompanied by quantitation of DNA concentration by use of the PicoGreen method, as we have described in detail elsewhere (Norton et al. 2004). Markers were selected for individual genotyping if pooled analysis revealed evidence for association with DD at an estimated level of P⩽.05 in the pooled samples of 240 cases and 312 controls. However, to allow for the smaller sample of cases (n=140) in the early analysis, those markers showing only a trend for association (P⩽.1) were regenotyped in the larger case pools and were selected for individual genotyping if the trend was confirmed at P⩽.05 (see fig. 1). The pooling data presented in our tables reflect the analysis that is based on the larger of the samples in which it was conducted. SNPs selected for individual genotyping were examined, and those subjects for whom we had sufficient DNA (223 DD cases and 273 controls) were included in the pooling experiments. Those markers for which individual genotyping in this case-control sample confirmed the pooling data (P⩽.05) were then genotyped in a semi-independent sample of 143 parent-proband trios to ensure that the results were not attributable to population stratification. Mean age of probands in the trio sample was 13.17±2.08 years; mean IQ was 104.01±11.88; mean RD was -5.06±1.76. All but 25 of the probands in the parent-proband trios were also included in the case-control sample. When genotypes were available for these 25 individuals, they were included in the final analysis of case-control association.

Individual genotyping was undertaken using a proprietary Amplifluor (Serologicals) genotyping method. Amplifluor reactions were performed in 5-μl reactions containing 50 ng DNA, in accordance with manufacturer instructions. Primers were designed using Amplifluor AssayArchitect and were obtained from Sigma-Genosys. PCR reactions were performed under standard conditions, with an initial denaturation stage of 96°C for 4 min, then 19 cycles at 96°C for 10 s, at 58°C or 60°C for 5 s, and at 72°C for 10 s; followed by 22 or 27 cycles at 96°C for 10 s, 20 s at 55°C, and 40 s at 72°C; and a final extension step at 72°C for 3 min. Genotypes were read on an LJL Biosystems Analyst. When we were unable to optimize Amplifluor, RFLP analysis of PCR products was undertaken. PCR reactions were performed under standard conditions in 12-μl reaction volumes with 32 ng of genomic DNA. Digests were undertaken in 17-μl reactions by use of the appropriate restriction enzyme (New England Biolabs), in accordance with manufacturer instructions. Products were visualized on 3% agarose gels stained with ethidium bromide.

All genotypes were tested for Hardy-Weinberg equilibrium with a χ² goodness-of-fit test (see the Simple Interactive Statistical Analysis Web site). Analysis of LD between markers (r² and D′) was performed using Haploview. Standard contingency tables were used for single-marker case-control analysis. Trios were analyzed using UNPHASED (Dudbridge 2003) (see the Rosalind Franklin Centre for Genomics Research Web site). Haplotypes were analyzed using EHPlus (Zhao et al. 2000) and UNPHASED (Dudbridge 2003). Logistic and conditional logistic regression analyses were performed on case-control and trio data, respectively.

Results

Of the 137 SNPs analyzed in pools, 17 yielded evidence for association at P⩽.05, and 13 of these were located within KIAA0319 (see table 1 for results of the pooled genotyping). Of these SNPs, 15 were then typed for 42 subjects with DD and 48 controls, to identify redundant markers on the basis of marker-marker LD (the two remaining SNPs, rs1555090 and rs926529, were known to be in LD with other SNPs analyzed on the basis of HapMap data). In an attempt to replicate the most significant haplotype of Francks and colleagues (2004), we also genotyped rs2143340, although this SNP did not show significant evidence for association in DNA pools. Perfect LD (r²=1) was noted for several of the markers in KIAA0319 (see table 2). Genotyped markers showing LD, as assessed by r², of at least 0.8 were dropped. On this basis, a minimal set of 10 markers was chosen for individual genotyping in the case-control sample (see table 3).

Table 1.

All Pooled Genotyping Data^[Note]

				Minor-Allele Frequencya
Position	Distance to next SNP(bases)	SNP ID	Gene and Position	Cases	Controls	P
24220516	12,172	rs1419229	Intergenic	.51	.48	.3751
24232688	1,379	rs9393529	VMP, 5′ flank	.10	.09	.4068
24234067	5,247	rs2876666	VMP, 5′ flank	.35	.31	.1364
24239314	2,367	rs9356928	VMP, intron	.51	.47	.1813
24241681	2,142	rs7455023	VMP, intron	.46	.42	.2094
24243823	1,994	rs12208318	VMP, intron	.27	.25	.4334
24245817	2,104	rs4357122	VMP, intron	Rare	Rare
24247921	1,978	rs12202381	VMP, intron	.12	.12	.8114
24249899	2,947	rs10946675	VMP, intron	.47	.51	.1972
24252846	2,616	rs10946676	VMP, intron	.13	.14	.8292
24255462	4,038	rs1053047	VMP, 3′ UTR	.51	.48	.2568
24259500	21,441	C_9373644_10b	Intergenic	.43	.48	.1541
24280941	2,162	rs1832709	DCDC2, 3′ UTR	.15	.13	.2020
24283103	3,182	rs3789219	DCDC2, intron	.18	.17	.6900
24286285	2,060	rs1419228	DCDC2, intron	.11	.13	.4000
24288345	20,043	rs2996452	DCDC2, intron	.14	.14	.9800
24308388	3,315	rs1277192	DCDC2, intron	.37	.35	.5300
24311703	2,892	rs1277194	DCDC2, intron	.41	.40	.6500
24314595	584	rs793861	DCDC2, intron	.26	.28	.6600
24315179	53,860	rs793862	DCDC2, intron	.26	.29	.3670
24369039	15,193	rs870601	DCDC2, intron	.34	.35	.6964
24384232	1,747	rs807698	DCDC2, intron	.20	.24	.2300
24385979	2,101	rs807726	DCDC2, intron	.24	.27	.3700
24388080	7,428	rs3789224	DCDC2, intron	.15	.13	.3840
24395508	1,923	rs3789227	DCDC2, intron	.16	.18	.4900
24397431	1,751	rs2296539	DCDC2, intron	.12	.14	.4500
24399182	2,121	rs2274305	DCDC2, exon	.44	.44	.8100
24401303	3,979	rs6907864	DCDC2, intron	.09	.13	.0660
24405282	3,543	rs807709	DCDC2, intron	.15	.15	.8000
24408825	1,241	rs807704	DCDC2, intron	.08	.07	.5590
24410066	446	rs807703	DCDC2, intron	.03	.01	.1000
24410512	2,218	rs3857541	DCDC2, intron	.05	.05	.6500
24412730	5,893	rs707862	DCDC2, intron	.07	.07	.9300
24418623	26,000	rs807685	DCDC2, intron	.20	.18	.3666
24444623	15,719	rs793704	DCDC2, intron	.44	.43	.7694
24460342	917	rs793722	DCDC2, intron	.32	.37	.0560
24461259	1,870	rs793720	DCDC2, intron	.31	.37	.0730
24463129	3,333	rs1277350	DCDC2, intron	.10	.09	.3180
24466462	2,706	rs1277349	DCDC2, 5′ flank	.04	.05	.2070
24469168	8,326	rs2792666	DCDC2, 5′ flank	.43	.47	.2020
24477494	32,139	rs793663	DCDC2, 5′ flank	.32	.36	.1742
24509633	1,517	rs9393553	MRS2L, 5′ flank	.28	.25	.3049
24511150	2,186	rs2295650	MRS2L, 5′ flank	Rare	Rare	NA
24513336	2,176	rs2273606	MRS2L, intron	.15	.16	.5679
24515512	2,831	rs1277347	MRS2L, intron	Rare	Rare	NA
24518343	1,946	rs7769012	MRS2L, intron	.34	.36	.5622
24520289	3,070	rs1298764	MRS2L, intron	Rare	Rare	NA
24523359	2,968	rs3761788	MRS2L, exon	.10	.08	.3293
24526327	2,196	rs2793422	MRS2L, exon	.33	.39	.0429
24528523	1,999	rs1056285	MRS2L, intron	.16	.12	.0564
24530522	1,081	rs7738943	MRS2L, intron	.12	.08	.0798
24531603	81,597	rs13735	MRS2L, intron	.39	.35	.1817
24613200	37,042	rs2817220	ALDH5A1, intron	.05	.05	.9017
24650242	1,457	rs2817241	KIAA0319, 3′ flank	.46	.45	.8754
24651699	1,183	rs807526	KIAA0319, 3′ UTR	.44	.43	.7332
24652882	693	rs699463	KIAA0319, 3′ UTR	.27	.28	.9479
24653575	2,010	rs2817243	KIAA0319, 3′ UTR	.11	.08	.0780
24655585	2,062	rs2817245	KIAA0319, intron	.42	.39	.2328
24657647	1,956	rs807532	KIAA0319, intron	.12	.12	.9383
24659603	2,650	rs2076313	KIAA0319, intron	.27	.24	.3305
24662253	2,028	rs807536	KIAA0319, intron	.38	.40	.5181
24664281	2,217	rs4083411	KIAA0319, intron	.26	.25	.8430
24666498	510	rs2760167	KIAA0319, intron	.35	.31	.1660
24667008	1,304	rs807540	KIAA0319, intron	.16	.13	.0755
24668312	1,969	rs807542	KIAA0319, intron	.29	.28	.6635
24670281	1,250	rs807544	KIAA0319, intron	.10	.10	.9190
24671531	1,858	rs2744549	KIAA0319, intron	.04	.04	.9637
24673389	4,934	rs2760161	KIAA0319, intron	.17	.19	.2876
24678323	87	rs2817195	KIAA0319, intron	.03	.03	.7629
24678410	1,900	rs807521	KIAA0319, intron	.07	.06	.5813
24680310	1,786	rs3846835	KIAA0319, intron	.20	.17	.2170
24682096	2,131	rs2744556	KIAA0319, intron	.46	.47	.6307
24684227	2,024	rs2817199	KIAA0319, intron	.27	.24	.4261
24686251	1,595	rs2760157	KIAA0319, intron	.26	.24	.4519
24687846	2,165	rs807507	KIAA0319, intron	.30	.27	.3142
24690011	2,334	rs807509	KIAA0319, intron	.34	.36	.6233
24692345	848	rs2817200	KIAA0319, intron	.36	.40	.1134
24693193	3,670	rs2817201	KIAA0319, intron	.18	.15	.2602
24696863	1,663	rs4504469	KIAA0319, exon	.36	.43	.0090
24698526	5,931	rs5026394	KIAA0319, intron	.34	.38	.1650
24704457	7,590	rs4576240	KIAA0319, exon	.11	.09	.3400
24712047	3,243	rs4352670	KIAA0319, intron	.08	.06	.4490
24715290	5,984	rs6911855	KIAA0319, intron	.22	.32	.0005
24721274	4,235	rs6939068	KIAA0319, intron	.19	.29	.0004
24725509	2,872	rs2745334	KIAA0319, intron	.20	.23	.1900
24728381	4,996	rs2817206	KIAA0319, intron	.08	.12	.0540
24733377	2,805	rs7751357	KIAA0319, intron	.31	.38	.0120
24736182	903	rs2179515	KIAA0319, intron	.34	.42	.0061
24737085	2,452	6917660	KIAA0319, intron	.32	.37	.0820
24739537	4,712	rs6456622	KIAA0319, intron	.32	.39	.0160
24744249	43	rs9358783	KIAA0319, intron	.34	.40	.0489
24744292	590	rs9358784	KIAA0319, intron	.33	.39	.0401
24744882	1,837	rs2206525	KIAA0319, intron	.31	.36	.0750
24746719	483	rs7755579	KIAA0319, intron	.33	.42	.0010
24747202	5,099	rs6456624	KIAA0319, intron	.34	.40	.0259
24752301	829	rs6935076	KIAA0319, intron	.44	.36	.0060
24753130	459	rs4363021	KIAA0319, intron	.03	.03	.8868
24753589	87	rs2235677	KIAA0319, intron	.31	.35	.1474
24753676	246	rs2235676	KIAA0319, intron	.18	.16	.5597
24753922	878	rs2038137	KIAA0319, intron	.39	.46	.0150
24754800	398	rs3756821	KIAA0319, 5′ flank	.49	.49	.8516
24755198	888	rs9467247	KIAA0319, 5′ flank	.20	.19	.5685
24756086	299	rs1555090	KIAA0319, 5′ flank	.33	.40	.0106
24756385	49	rs1555089	KIAA0319, 5′ flank	.35	.31	.1908
24756434	2,306	rs3212236	TTRAP, 3′ UTR	.19	.17	.5090
24758740	2,082	rs3087943	TTRAP, exon	.17	.18	.6450
24760822	1,600	rs2294691	TTRAP, intron	.12	.09	.1140
24762422	1,508	rs3181238	TTRAP, intron	.30	.30	.9886
24763930	1,907	rs3212234	TTRAP, intron	.08	.07	.6760
24765837	39	rs3033236	TTRAP, intron	.16	.17	.7397
24765876	1,174	rs3212232	TTRAP, intron	.25	.24	.5390
24767050	2,440	rs2143340	TTRAP, intron	.06	.06	.8210
24769490	3,829	rs2056999	TTRAP, intron	.24	.26	.4842
24773319	667	rs3756819	TTRAP, intron	.41	.38	.3560
24773986	187	rs1061925	TTRAP, intron	.12	.12	.9938
24774173	1,468	rs3756815	TTRAP, intron	.04	.03	.3650
24775641	137	rs3181228	THEM2, intron	.51	.49	.5934
24775778	2,146	rs3181227	THEM2, intron	.25	.23	.3180
24777924	3,675	rs9393576	THEM2, intron	.40	.41	.7906
24781599	4,368	rs7765052	THEM2, intron	.09	.08	.4890
24785967	5,293	rs7451561	THEM2, intron	.21	.18	.3230
24791260	1,740	rs2143338	THEM2, intron	.48	.52	.1734
24793000	2,744	rs1555086	THEM2, intron	.44	.47	.3236
24795744	1,193	rs926529	THEM2, intron	.27	.34	.0200
24796937	3,221	rs1885209	THEM2, intron	.10	.11	.7449
24800158	1,667	rs1885211	THEM2, intron	.42	.43	.9603
24801825	3,542	rs3777664	THEM2, intron	.28	.35	.0170
24805367	2,847	rs2092404	THEM2, intron	.30	.29	.7716
24808214	2,811	rs3777663	THEM2, intron	.34	.30	.185
24811025	1,492	rs1056319	THEM2, 3′ flank	.03	.03	.8151
24812517	1,297	rs1053598	THEM2, 3′ flank	.27	.32	.0460
24813814	3,367	rs3756814	C6orf62, 3′ UTR	.32	.34	.5702
24817181	3,942	rs2294686	C6orf62, intron	.04	.03	.3300
24821123	6,834	rs6913673	C6orf62, intron	.04	.04	.9180
24827957	1,319	rs3813687	C6orf62, 5′ flank	.22	.19	.1580
24829276	370	rs6456632	C6orf62, 5′ flank	.28	.25	.3027
24829646	NA	rs1923187	C6orf62, 5′ flank	.36	3	.7481

Note.— NA = not applicable.

^aMinor allele frequencies of 137 SNPs genotyped in DNA pools of 240 subjects with DD and 312 controls.

^bThis SNP ID refers to an ABI assay-on-demand SNP; all other IDs refer to dbSNP IDs.

Table 2.

LD between SNPs Showing Significance (P<.05) in DNA Pools^[Note]

	LD Value for SNP Pair
Gene and SNPa	rs2793422	rs4504469	rs6911855	rs6939068	rs7751357	rs2179515	rs6456622	rs9358783	rs9358784	rs7755579	rs6456624	rs6935076	rs2038137	rs2143340	rs3777664	rs1053598
MRS2L:
rs2793422		.2	.01	.06	.22	.18	.24	.24	.21	.03	.03	.04	.06	0	.15	.07
KIAA0319:
rs4504469	0		1	1	.82	.83	.85	.85	.86	.62	.68	.59	.75	.15	.51	.57
rs6911855	.01	.02		1	1	1	1	1	1	1	1	1	1	.52	1	.02
rs6939068	0	.02	.79		1	1	1	1	1	1	1	1	1	.63	1	.22
rs7751357	.01	.51	.02	.02		1	1	1	1	1	1	1	1	.63	.68	.66
rs2179515	.01	.55	.02	.02	1		1	1	1	1	1	1	1	.68	.69	.68
rs6456622	.02	.54	.02	.02	1	1		1	1	1	1	1	1	.64	.67	.69
rs9358783	.02	.54	.02	.02	1	1	1		1	1	1	1	1	.66	.69	.67
rs9358784	.01	.54	.02	.02	1	1	1	1		1	1	1	1	.65	.67	.67
rs7755579	0	.38	.02	.02	.80	.79	.79	.80	.78		1	1	1	.58	.65	.71
rs6456624	0	.42	.02	.02	.85	.83	.84	.84	.82	.95		1	1	.74	.64	.68
rs6935076	.02	.14	.02	.03	.35	.3	.33	.33	.33	.43	.41		1	.76	.69	.93
rs2038137	.01	.52	.02	.02	.89	.89	.91	.91	.89	.90	.92	.33		.73	.65	.7
TTRAP:
rs2143340	.01	.05	.01	0	.03	.05	.03	.03	.03	.03	.05	.09	.08		1	1
THEM2:
rs3777664	0	.21	0	0	.43	.37	.42	.42	.41	.36	.36	.11	.37	.08		.94
Intergenic:
rs1053598	.01	.23	.03	.01	.41	.41	.45	.45	.42	.38	.35	.22	.41	.07	.77

Note.— Intermarker LD values, as measured by D′ (above the diagonal) and r² (below the diagonal). Values ⩾.8 are set in bold italics.

^aUnderlined SNPs were individually genotyped in the case-control sample.

Table 3.

Genotype and Allele Counts for Selected SNPs in the Case-Control Sample^[Note]

	No. of Subjectswith Genotype				No. (%) of Alleles
Gene, SNP, and Sample	1-1	1-2	2-2	P	1	2	P	OR (95% CI)
MRS2La:
rs2793422:
Cases	99	94	22		304 (69)	138 (31)
Controls	91	117	43	.03	299 (60)	203 (40)	.003	1.50 (1.14–1.96)
KIAA0319:
rs4504469a:
Cases	101	117	22		319 (66)	161 (34)
Controls	88	124	52	.002	300 (57)	228 (43)	.002	1.51 (1.17–1.95)
rs6911855:
Cases	200	17	1		417 (96)	19 (04)
Controls	253	12	0	.17	518 (98)	12 (02)	.07	.51 (.24–1.06)
rs6939068:
Cases	180	19	1		379 (95)	21 (05)
Controls	234	14	0	.16	482 (97)	14 (03)	.06	.52 (.26–1.04)
rs2179515a:
Cases	116	100	16		332 (72)	132 (28)
Controls	109	108	40	.008	326 (63)	88 (37)	.007	1.45 (1.11–1.90)
rs6935076a:
Cases	65	131	35		261 (56)	201 (44)
Controls	107	118	30	.006	332 (65)	178 (35)	.006	.70 (.54–.90)
rs2038137a:
Cases	112	104	13		328 (72)	130 (28)
Controlsb	106	105	46	.0001	317 (62)	197 (38)	.001	1.57 (1.20–2.05)
TTRAP:
rs2143340:
Cases	140	56	7		336 (83)	70 (17)
Controls	179	68	3	.26	426 (85)	74 (15)	.32	.83 (.58–1.19)
THEM2a:
rs3777664:
Cases	119	92	13		330 (74)	118 (26)
Controls	112	113	31	.02	337 (66)	175 (34)	.008	1.45 (1.10–1.92)
Intergenica:
rs1053598:
Cases	124	92	9		340 (76)	110 (24)
Controls	123	112	23	.05	358 (69)	158 (31)	.03	1.36 (1.03–1.81)

Note.— Genotypic and allelic P values are given, as are ORs and 95% CIs. P values ⩽.05 are indicated in bold italics.

^aCase-control analysis includes 25 extra probands from the proband-parent trios.

^bGenotypes deviate from Hardy-Weinberg equilibrium.

All genotypes were in Hardy-Weinberg equilibrium for cases and controls and probands and parents, except for SNP rs2038137) (located near the exon 1/intron 1 boundary of KIAA0319), which showed slight distortion in the controls (P=.03). Our most significant results (P⩽.01) were found in KIAA0319, MRS2L, and THEM2 in the case-control sample (see table 3). To ensure that these data did not arise from population stratification, the seven markers that were significant in the case-control sample at P⩽.05 were then individually genotyped in a sample of 143 parent-proband trios (table 4). This also provides for a degree of independence, since the control (nontransmitted) alleles are independent of the controls in the case-control study. After this step, six SNPs remained significant—three in the KIAA0319 gene (family-based, P=.04–.002; case-control, P=.007–.001), one in MRS2L (family-based, P=.04; case-control, P=.003), and two in or flanking the THEM2 gene (family-based, P=.03–.01; case-control, P=.03–.008). Two of these SNPs (rs4504469 and rs2179515) were reported in the study by Francks et al. (2004) to be associated with a number of componential measures of DD in the combined U.K. sample. Both are located in the KIAA0319 gene (r²=0.55).

Table 4.

Transmission of Alleles at Selected SNPs in the Parent-Proband Trios^[Note]

	No. (%) of Alleles
Gene, SNP, and Transmission	1	2	P	OR (95% CI)a
MRS2L:
rs2793422:
Transmitted	191 (71)	79 (29)
Nontransmitted	168 (62)	102 (38)	.04	1.47 (1.02–2.1)
KIAA0319:
rs4504469:
Transmitted	166 (68)	78 (32)
Nontransmitted	144 (59)	100 (41)	.04	1.48 (1.02–2.14)
rs2179515:
Transmitted	184 (71)	77 (30)
Nontransmitted	162 (62)	99 (38)	.04	1.46 (1.01–2.10)
rs6935076:
Transmitted	154 (56)	120 (44)
Nontransmitted	189 (69)	85 (31)	.002	.57 (.41–.82)
rs2038137:
Transmitted	169 (69)	77 (31)
Nontransmitted	152 (62)	94 (38)	.11	1.36 (.94–1.97)
THEM2:
rs3777664:
Transmitted	165 (74)	59 (26)
Nontransmitted	144 (64)	80 (36)	.03	1.55 (1.04–2.33)
Intergenic:
rs1053598:
Transmitted	186 (76)	59 (24)
Nontransmitted	160 (65)	85 (35)	.01	1.67 (1.13–2.48)

Note.— P values for association were calculated using UNPHASED (Rosalind Franklin Centre for Genomics Research). P values ⩽.05 are indicated in bold italics.

^aORs and 95% CIs refer to allele 1.

To determine which SNPs accounted for the association, stepwise logistic regression analyses were performed on the case-control and trio data on the basis of all SNPs for which we had individual genotyping data (see tables 3 and 4). For the case-control sample, the SNPs—rs2793422 (MRS2L); rs4504469, rs6911855, rs6939068, rs2179515, rs6935076, and rs2038137 (KIAA0319); rs2143340 (TTRAP); rs3777664 (THEM2); and Intergenic rs1053598—were initially submitted into the logistic regression model (P=.029; 10 df). The stepwise procedure reduced the number of SNPs to three—rs2793422 (MRS2L) and rs4504469 and rs6935076 (KIAA0319)—that showed a highly significant fit (P=.00002; 3 df). For the proband-parent trios, the probands were considered as cases, and nontransmitted alleles were employed to create pseudocontrols (Cordell and Clayton 2002). These data were submitted into conditional logistic regression analyses (P=.347; 7 df). The best model was again identified by use of a stepwise procedure (P=.02; 2 df) that removed every SNP except SNPs rs4504469 and rs6935076. The addition of rs2793422 did not significantly improve the model (P=.10 [log-likelihood ratio test]). rs4504469 is a nonsynonymous SNP in exon 4 (Ala→Thr), and rs6935076 is located in intron 1 of the KIAA0319 gene (see fig. 2).

Figure 2.

Location of candidate genes on chromosome 6p. The location of SNPs found to be significant (P⩽.05) in our case-control sample are shown relative to nearby markers. The direction of transcription is shown for each gene. LD blocks across the region are based on data from HapMap. The P value refers to the most significant haplotype (2-1) comprising the two SNPs indicated. An asterisk (*) indicates the amino acid–changing SNP in exon 4.

On the basis of the results of the regression analysis, we analyzed the two-marker haplotype that consisted of the KIAA0319 SNPs rs4504469 and rs6935076 in the case-control and trio samples (see tables 5 and 6). Significant evidence for association was obtained on the basis of the global test (P=.0001 in the case-control sample; P=.02 in trios). In each sample, the 1-2 haplotype was associated with DD, but more striking is the significant underrepresentation of haplotype 2-1 in the cases based on the case-control (P=.00003; odds ratio [OR] 0.53; 95% CI 0.40–0.70) and family-based (P=.006; OR 0.57; 95% CI 0.39–0.84) analyses. (Fig. 2 summarizes our results.)

Table 5.

Analysis of Haplotypes in KIAA0319 Comprising SNPs rs4504469 and rs6935076 in 248 Subjects with DD and 273 Controls

Allele at SNP		Frequency in
rs4504469	rs6935076	Cases	Controls	Haplotype Pa
1	1	.31	.27	.26
1	2	.35	.30	.02
2	1	.25	.39	.00003
2	2	.09	.05	.17

^aGlobal P=.0001.

Table 6.

Analysis of Haplotypes in KIAA0319 Comprising SNPs rs4504469 and rs6935076 in a Sample of 143 Families with DD

Allele at SNP		Frequency
rs4504469	rs6935076	Transmitted	Nontransmitted	Haplotype Pa
1	1	.32	.32	.95
1	2	.35	.27	.03
2	1	.24	.36	.006
2	2	.09	.05	.22

^aGlobal P=.02.

We also analyzed the three-marker haplotype that consisted of rs4504469, rs2038137, and rs2143340, which was reported as significantly associated with DD by Francks and colleagues (2004). This haplotype did not yield global evidence for association in our sample (see table 7). Two individual haplotypes did, however, show evidence for association with DD. The 1-1-1 haplotype was more frequent in subjects with DD than in control individuals (P=.03). The 2-2-1 haplotype, which was significantly associated with the READ phenotype in the combined U.K. sample of the study by Francks et al. (2004), also displayed evidence of association with DD in our case-control sample (P=.01) and showed the same direction of effect (in their study, the 2-2-1 haplotype was associated with better performance; in our sample, this haplotype was more frequent in control individuals). In our sample, this haplotype is, in fact, perfectly defined by the first two SNPs (since there was no observation of the 2-2-2 haplotype). We therefore excluded rs2143340 and looked at the two-marker haplotype (2-2) that consisted of the other two SNPs in our family-based sample, but, although it was undertransmitted to the probands, this was not significant (P=.10). It should be noted that, since rs4504469 shows more significance individually than does the 2-2-1 haplotype, no extra information was obtained from this haplotype in our sample. Moreover, the 1-1-2 haplotype that was reported to show association with componential measures of DD in both the U.S. sample and the combined U.K. sample in the study by Francks et al. (2004) was not significantly associated with DD in our sample (P=.21).

Table 7.

Haplotype Analysis Spanning KIAA0319 and TTRAP^[Note]

Allele at SNP			Frequency in
rs4504469	rs2038137	rs2143340	Cases	Controls	Haplotype Pa
1	1	1	.47	.41	.03
1	1	2	.15	.12	.21
1	2	1	.04	.05	.66
2	1	1	.07	.08	.67
2	1	2	.02	.02	.67
2	2	1	.25	.33	.01

Note.— Analysis of haplotypes comprising SNPs rs4504469, rs2038137, and rs2143340 in 223 subjects with DD and 273 controls. The 1-1-2 haplotype was observed by Francks et al. (2004) to be significantly associated with a number of reading-related measures but is not significant in our sample.

^aGlobal P=.10.

Discussion

Previous linkage and association studies of DD and chromosome 6p have implicated a region between markers D6S461 and D6S105. More recently, following other positional candidate-gene studies, VMP, DCDC2, KIAA0319, TTRAP, and THEM2 have been suggested as possible susceptibility genes within this region (Deffenbacher et al. 2004; Francks et al. 2004). Our study tested for association with each of these genes (VMP, DCDC2, KAAG1, MRS2L, KIAA0319, TTRAP, THEM2, and C6orf62) by use of a high-density SNP map and an independent sample. Initially, we genotyped DNA pools from subjects and controls and followed up those findings with individual genotyping in a case-control sample and a nested family-based association sample. In both samples, we observed evidence for association with three SNPs in KIAA0319 (rs4504469, P=.002; rs2179515, P=.007; and rs6935076, P=.006), with one SNP in MRS2L (rs2793422, P=.003) and in THEM2 (rs3777664, P=.008), and with an intergenic SNP (rs1053598, P=.02). Two of these SNPs, rs4504469 and rs2179515 (both located in KIAA0319), have been reported elsewhere to display significant association with a number of componential measures of DD in a U.K. sample (Francks et al. 2004). Our results support existing data (Deffenbacher et al. 2004; Francks et al. 2004) that implicate genes in this region in DD, and our results extend the previous findings by demonstrating that the source of the signal is likely to be variation in KIAA0319. The study by Francks and colleagues (2004) implicated a region containing KIAA0319, TTRAP, and THEM2, whereas that of Deffenbacher and colleagues (2004) implicated KIAA0319, DCDC2, VMP, TTRAP, and THEM2. Combining their data with our own produces a pattern of evidence that implicates KIAA0319 as a susceptibility gene for DD.

This is compatible with the logistic regression and conditional logistic regression analyses in our case-control and proband-parent trio samples, respectively. Case-control data analyses with the use of a stepwise procedure revealed three SNPs that account for the association observed: rs2793422 in MRS2L and rs4504469 and rs6935076 in KIAA0319 (P=.00002). Analysis of the proband-parent trios identified two SNPs that account for the association observed: rs4504469 and rs6935076 (P=.03). The results, therefore, are consistent with crude inspection of the genes showing overlap between the studies and provide strong evidence that KIAA0319 SNPs rs4504469 and rs6935076 are responsible for the association with DD observed in this study. A haplotype comprising these two SNPs is highly significantly associated with DD in both the case-control sample (P=.00003) and the trio sample (P=.006). This effect is largely driven by haplotype 2-1 as a “protective” haplotype.

Although the logistic regression analyses imply that, of the markers tested, rs4504469 and rs6935076 can account for the association signal, this does not imply that they are the direct susceptibility alleles, per se. However, interestingly, rs4504469 (one of the two SNPs that makes up our most significant haplotype) is a nonsynonymous SNP (Ala→Thr), which suggests the possibility that this might, in part, contribute directly to the association. However, in our own sample, the threonine at this locus that is present on the protective rs4504469/rs6935076 2-1 haplotype is also present on the 2-2 haplotype (tables 5 and 6), which is more common in cases, albeit not significantly more so. This suggests that, if the nonsynonymous change at rs4504469 can influence risk of DD directly, then its effects can be modified by a second susceptibility allele in the gene. Given that the SNP showed the same pattern of allelic association in the U.K. sample of Francks and colleagues (2004) but not their U.S. sample, perhaps a more likely explanation is that this SNP does not directly influence susceptibility to DD. Thus, although our study provides strong evidence that variation in the KIAA0319 gene is associated with increased risk of developing dyslexia, the true susceptibility alleles remain to be identified.