SEARCH

SEARCH BY CITATION

Keywords:

  • Association;
  • dyslexia;
  • genes;
  • GWAS;
  • language;
  • reading

Abstract

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. References
  7. Acknowledgments
  8. Supporting Information

Candidate genes have been identified for both reading and language, but most of the heritable variance in these traits remains unexplained. Here, we report a genome-wide association meta-analysis of two large cohorts: population samples of Australian twins and siblings aged 12–25 years (n = 1177 from 538 families), and a younger cohort of children of the UK Avon Longitudinal Study of Parents and their Children (aged 8 and 9 years; maximum n = 5472). Suggestive association was indicated for reading measures and non-word repetition (NWR), with the greatest support found for single nucleotide polymorphisms (SNPs) in the pseudogene, ABCC13 (P = 7.34 × 10−8), and the gene, DAZAP1 (P = 1.32 × 10−6). Gene-based analyses showed significant association (P < 2.8 × 10−6) for reading and spelling with genes CD2L1, CDC2L2 and RCAN3 in two loci on chromosome 1. Some support was found for the same SNPs having effects on both reading skill and NWR, which is compatible with behavior genetic evidence for influences of reading acquisition on phonological-task performance. The results implicate novel candidates for study in additional cohorts for reading and language abilities.

Reading and language disorders are heritable, with genetic effects accounting for between 45% and 61% of the phenotypic variance (Bishop et al. 2006; Harlaar et al. 2005; Hawke et al. 2006). Moreover, dyslexia and specific language impairment (SLI) are often comorbid, with evidence of shared genetic etiology (Bishop 2001; Bishop & Hayiou-Thomas 2008; Newbury et al. 2011; Rice et al. 2009). The identification of putative genes influencing reading disability has been relatively successful compared with other complex cognitive traits. Linkage analysis identified a number of chromosomes of interest (e.g. 3, 6p, 15q) and fine mapping of these regions implicated strong candidate genes—including ROBO1 (Hannula-Jouppi et al. 2005), DCDC2 (Meng et al. 2005), KIAA0319 (Cope et al. 2005) and DYX1C1 (Taipale et al. 2003)—both linkage and gene associations have been replicated, although null replications have also been reported (e.g. Skiba et al. 2011; Tran et al. 2013; Zhong et al. 2013; Zou et al. 2012). For SLI, evidence from linkage and targeted association studies has converged on two regions of interest, 7q and 16 (SLI Consortium 2002), and three candidate genes, CNTNAP2, CMIP and ATP2C2 (Newbury et al. 2009; Vernes et al. 2008), with evidence suggesting that these genes exert pleiotropic effects on language and reading (Newbury et al. 2011). The genetic linkage and association findings for dyslexia and language impairment may also generalize to normal variation in reading and language (Bates et al. 2010; Lind et al. 2010; Luciano et al. 2007; Whitehouse et al. 2011). Here, we report a genome-wide association study (GWAS) for these traits in individuals from population-based samples.

One GWAS for reading ability has previously been reported. This used a multistage DNA-pooling design in which the allele frequencies of low and high reading ability groups were compared using a 100K single nucleotide polymorphism (SNP) microarray (Meaburn et al. 2008). A subsequent design stage involved individual genotyping and confirmed support for association with 10 SNPs; although these SNPs were not replicated in an independent cohort (Luciano et al. 2011). Field et al. (2013) carried out genome-wide association (on only 133 165 SNPs) in the context of a family design (718 individuals from 101 families with dyslexia probands) but found no significant associations. More recently, an event-related brain potential phenotype, the mismatch negativity, has been subjected to GWAS in a sample of individuals with dyslexia, and showed replicable association with rs4234898, a marker located on chromosome 4q32.1 and associated with mRNA-expression levels of SLC2A3, a neural glucose transporter gene (Roeske et al. 2011). No GWAS of the non-word repetition (NWR) marker of SLI risk has been reported to date to our knowledge. The present GWAS meta-analysis of reading traits capturing phonological decoding and orthographic skill and of NWR performance—a marker of SLI—is undertaken in a primarily adolescent sample from the general Australian population and a UK population sample of children. It, then, is the first GWAS of these traits using continuously varying reading and language measures from the general population and comprehensive GWAS data, that is, imputed to ∼2.4 million SNPs.

Materials and methods

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. References
  7. Acknowledgments
  8. Supporting Information

Sample

Brisbane Adolescent Twin Sample (BATS)

Twins and their non-twin siblings were recruited from ongoing studies of melanoma risk factors and cognition in a population-based sample (Wright et al. 2001). This including 1177 individuals from 538 families (134 monozygotic twins, 338 dizygotic twins, two families comprising both a monozygotic and dizygotic twin pair, 11 triplets, three dizygotic pairs + one or two siblings, six pairs of non-twin siblings, one set of three siblings and 43 individuals—unpaired twins/non-twins) who had Components of Reading Examination (CORE) reading (described below), intelligence quotient (IQ) and genotyping data. A subset including 1111 individuals from 505 families had language, IQ and genotyping data, and 1057 individuals from 479 families had whole word reading, IQ and genotyping data. For monozygotic twin pairs the mean of their test scores was used to increase power (Miller 1998). The age range of the sample was between 12.3 and 25.1 years [mean = 17.9, standard deviation (SD) = 2.9] at the time of testing reading, and ranged 13.7–26.1 years (mean = 20.1, SD = 3.4) at the time of collecting language data. The sample was 54.5% female, and 98% reported Caucasian ancestry, predominantly Anglo-Celtic (∼82%). Written informed consent was obtained from each participant and their parent/guardian (if younger than 18 years). Ethical approval for this study was received from the Human Research Ethics Committee, Queensland Institute of Medical Research.

Avon Longitudinal Study of Parents and their Children (ALSPAC)

Participants formed part of the ALSPAC, a longitudinal, population-based sample recruited from the county of Avon, UK in the early 1990s (see Golding et al. 2001). In this study, phenotype and genotype data were available for the following number of participants: 5078 (reading at age 9), 5070 (non-word reading at age 9), 5071 (spelling at age 9) and 5472 (NWR at age 8). Their IQ was measured at age 8. Written informed consent was obtained from parents and ethical approval was granted from the ALSPAC Law and Ethics committee and other local ethics committees. Participants in both cohorts were free of neurological conditions and major psychiatric illness.

Measures

BATS

Regular-word, irregular-word and non-word reading and spelling were assessed using the CORE (Bates et al. 2004), which was lengthened to a 120-word version (Castles & Coltheart 1993) to include spelling and to increase the difficulty level for an older sample. This test was administered over the telephone by a trained researcher. Test scores for the individual component subtests were calculated as a simple sum of correct items and were Box-Cox transformed to attain normality, and a principal component from these tests derived representing a general factor of reading and spelling ability. The reading and spelling factor was assessed to tap general processes underlying reading and spelling abilities. A measure of whole word reading, the Schonell graded word reading test (Schonell & Schonell 1960), which tested irregular and regular words, was also investigated. The Schonell reading data were negatively skewed and transformed by a logarithmic function of the reverse distribution. The assessment of language ability (representing phonological storage efficiency) was via the Gathercole and Baddeley (Gathercole 1994) and Dollaghan and Campbell (Dollaghan & Campbell 1998) NWR tests; the sum of the standard scores from each measure was used to increase trait reliability. Intelligence [performance IQ from the Multidimensional Aptitude Battery (Jackson 1984)] was entered as a covariate to increase the sensitivity to detect genetic effects for reading ability, as has been previously shown (Luciano et al. 2007), and language skill independent of IQ. This has the effect of adjusting reading and language scores that are low because of low general cognitive ability. The IQ and the Schonell were measured as close as possible to participants' 16th birthday. The CORE measures were collected between 1 month and 8.3 years before IQ in 53.8% of the sample, with the remainder being collected between 1 month and 9 years after IQ administration. The NWR and CORE measures were collected at the same time in 29.6% of the sample, with the remaining sample being assessed on NWR up to 4.1 years after CORE assessment.

ALSPAC sample

Reading was assessed using separate tests of word reading, non-word reading and spelling. Reading items were chosen from a larger selection of words used by Nunes et al. (2003). Test–retest reliability of word reading was 0.80 and had a correlation of 0.85 with the Schonell Graded Word Reading Task and 0.81 with the word spelling test given 4 months later. Test–retest reliability of the non-word reading task was 0.73 and showed a correlation of 0.73 and 0.77 with the reading and spelling tasks, respectively, given 4 months later. Language was assessed by an adapted NWR test comprised of 12 non-words, four each of three, four and five syllables and conforming to phonological rules for sound combinations. Each word was presented via a cassette recording and the participant was required to repeat each item, with a correct score given if there was no phonological deviation from the target item. A composite measure of word reading, non-word reading and spelling was formed, providing a measure comparable to the reading and spelling factor assessed in the BATS cohort. The IQ was measured with items based on the Wechsler tests, in this case a short form of the WISC-III (Wechsler et al. 1992) in which alternate items were used for all subtests, except the coding subtest which was administered in full. The performance IQ measure was used.

Genotyping, quality control and imputation

BATS

DNA was extracted from blood samples and SNP genotyping was performed with the Illumina 610k Quad Bead chip by deCODE Genetics (Reykjavik, Iceland). The samples were part of a larger adolescent cohort (n = 4391) whose genotyping data were subjected to tests of quality control, including checks for pedigree, sex and Mendelian errors, and for ancestry using HapMap3 and GenomeEUTwin individuals as a reference panel. Five individuals were removed because of gender inconsistencies, and 28 individuals (14 twin pairs) because of non-European ancestry. Quality control filters, as previously described (Medland et al. 2009), ensured no samples had a call rate ≤0.95, and that all SNPs included in analyses had the following characteristics: call rate ≥ 0.95, minor allele frequency (MAF) ≥ 0.01 and Hardy-Weinberg equilibrium (HWE) test with P ≥ 1 × 10−6. The HapMap phase II CEU data [NCBI build 36 (UCSC hg18)] was used as the reference sample for imputation using MACH software. The SNPs with low imputation (r2 < 0.30) and low MAF (<0.01) were excluded. A total of 2 373 249 SNPs were examined.

ALSPAC sample

A broader sample of 9912 participants were genotyped using the Illumina HumanHap550 quad genome-wide SNP genotyping platform by 23andMe subcontracting the Wellcome Trust Sanger Institute, Cambridge, UK and the Laboratory Corporation of America, Burlington, NC, USA. Individuals were excluded from analyses on the basis of excessive or minimal heterozygosity, gender mismatch, individual missingness (>3%), cryptic relatedness as measured by identity by decent (genome-wide IBD > 10%) and sample duplication. Individuals were assessed for population stratification using multidimensional scaling modeling seeded with HapMap Phase II release 22 reference populations. Individuals of non-European ancestry were removed from further analysis. Individuals were imputed to HapMap Phase II (Build 36 release 22) using Markov Chain Haplotyping software [MACH v.1.0.16 (Li et al. 2010)], with removal of poorly imputed SNPs (r2 hat < 0.30).

Statistical analyses

In the BATS, each reading and language measure was first residualized for the effects of sex, age, examiner and performance IQ. Genome-wide association analysis, implemented in Merlin, was conducted on standardized scores using a total test of association which corrected for sample relatedness (Chen & Abecasis 2007). In the ALSPAC sample, standardized residuals were derived for each trait controlling for the effects of sex, age and performance IQ and these were regressed on expected allelic dosage scores using the software package MACH2QTL (Li et al. 2009). A weighted inverse variance method in METAL (Willer et al. 2010) was used to meta-analyze the results. The reading and spelling factor from BATS was meta-analyzed with the word reading, non-word reading and spelling composite from ALSPAC (henceforth termed the ‘reading and spelling measure’). The Schonell from BATS was meta-analyzed with word reading from ALSPAC (henceforth termed the ‘word reading measure’), and the NWR measures from each cohort were meta-analyzed together.

A conventional genome-wide significance level of P < 5 × 10−8 was adopted (Dudbridge & Gusnanto 2008). Correcting for multiple testing of three variables, the level for genome-wide significance is P < 1.7 × 10−8. For reporting, significance at a suggestive level (P < 6.09 × 10−6 as proposed for the CEU SNP panel of markers adjusted for SNP non-independence) is used (Duggal et al. 2008). Empirical analyses indicate that uncorrected SNP associations of P < 1 × 10−7 typically constitute replicable associations (Panagiotou & Ioannidis 2012), although this may depend on factors such as adequate study power and trait heritability. For clarity, all P-values are reported uncorrected for multiple testing.

Gene-based tests were performed on the meta-analysis results using VEGAS, a program in which SNPs are assigned to genes, and their combined effect within a gene is tested taking into account SNP linkage disequilibrium (LD) (Liu et al. 2010). Roughly 18 000 autosomal genes were tested, annotated to positions on the UCSC Genome Browser (hg18 assembly) and including regulatory regions located ±50 kb of 5′ and 3′ untranslated regions. Bonferroni corrected significance was set at P < 2.8 × 10−6 based on a correction for the number of genes tested.

GWAS results were further used to check for replication of the previously associated candidate genes with dyslexia and SLI. This was also done within each cohort so that any future meta-analyses incorporating these results can test for the presence of sample heterogeneity thereby enabling the specification of random effects models.

Results

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. References
  7. Acknowledgments
  8. Supporting Information

Phenotypic distributions in both cohorts were approximately normal and extreme outliers were removed prior to analysis. The maximum included value was 5.42 SDs from the mean for the Schonell in the BATS and 3.03 SDs from the mean for the word reading score in the ALSPAC sample. All results of the GWAS analysis were tested for evidence of population stratification (see Q–Q plots presented in Fig. S1, Supporting Information), with λ values within an acceptable range of 1.016–1.026. There was some indication of SNP tests deviating from the null distribution for NWR.

SNP-based GWAS results

The SNPs reaching a suggestive significance level are shown in Table 1. The distribution of P-values across the 22 chromosomes is shown in Manhattan plots provided in Fig. S2 and the top 100 SNPs for each trait are provided in Table S1.

Table 1. Description and results (β, SE and P) of the top SNPs (P < 6.09 × 10−6) associated with reading and language measures
SNPChPositionA1A2AFPA1 βSEGeneOther associations P < 0.05: β, SE, P
  1. A1, allele 1; A2, allele 2; AF, allele frequency (BATS, ALSPAC where unequal). Other significant SNPs that were in strong LD (r2 > 0.80) with the SNPs listed in the table: RS: rs17135159 with rs12415036, rs2182195, rs11251863 and rs11251868; rs12351590 with rs10960483, rs7025259, rs10960488 and rs791775; rs13307587 with rs12699504 and rs1357978; rs11666805 with rs3786983. WR: rs764255 with rs8059199, rs716270, rs716271, rs4243136, rs8048857, rs8056327, rs1480566, rs990258, rs4243137 and rs3956052; rs1928007 with rs7989551. NWR: rs2192161 with rs909257; rs2822560 with rs12329784, rs17002266, rs7283184 and rs2822584; rs7187223 with rs7196091, rs956315, rs7202694, rs2161694, rs1424049, rs6565017, rs1025066 and rs3909533; rs6954796 with rs1019248, rs4721865 and rs10246471; rs6954796 with rs2041664, rs2041663, rs7795980 and rs1005499.

Reading and spelling (RS)         
  rs4807927191374199AG0.941.32 × 10−60.2070.043DAZAP1WR: 0.15 ± 0.03, 5.59 × 10−6; NWR: 0.15 ± 0.04, P = 0.0003
  rs17135159103398947AC0.81, 0.802.40 × 10−60.1070.023BC037918WR: 0.06 ± 0.02, P = 0.0009; NWR: 0.07 ± 0.02, P = 0.0006
  rs479526611726153TC0.54, 0.552.71 × 10−60.0930.020 WR: 0.05 ± 0.01, P = 0.0005
  rs32130561062218396AG0.04, 0.033.34 × 10−6−0.3760.081CDC2WR: −0.23 ± 0.06, P = 7.65 × 10−5
  rs10508253103420951TG0.863.77 × 10−60.1210.026BC037918WR: 0.06 ± 0.02, P = 0.001; NWR: 0.10 ± 0.02, P = 9.75 × 10−5
  rs13307587713510254AG0.66, 0.674.94 × 10−6−0.0880.019 WR: −0.06 ± 0.01, 7.09 × 10−5; NWR: −0.05 ± 0.02, P = 0.005
  rs11666805191376448TC0.08, 0.075.71 × 10−6−0.1630.036DAZAP1WR: −0.12 ± 0.03, 1.70 × 10−5; NWR: −0.10 ± 0.03, P = 0.004
Word reading (WR)         
  rs7642551672271184TC0.63, 0.661.80 × 10−7−0.0770.015 RS: −0.07 ± 0.02, P = 0.0004
  rs48395161117000000AC0.313.61 × 10−70.0790.015 RS: 0.07 ± 0.02, P = 0.0004; NWR: 0.04 ± 0.02, P = 0.029
  rs111583451460964798TC0.80, 0.823.29 × 10−60.0840.018PRKCHRS: 0.10 ± 0.02, P = 1.32 × 10−5
  rs19280071369708224AG0.914.77 × 10−6−0.1080.024 RS: −0.13 ± 0.03, P = 2.18 × 10−5; NWR: −0.07 ± 0.03, P = 0.015
  rs11945798412378330TC0.964.94 × 10−6−0.1750.038 RS: −0.16 ± 0.05, P = 0.002
  rs120504121482447756TG0.804.95 × 10−60.0770.017 RS: 0.09 ± 0.02, P = 4.59 × 10−5
  rs115776281161000000AG0.925.00 × 10−6−0.1230.027NOS1APRS: −0.11 ± 0.04, P = 0.0016; NWR: −0.10 ± 0.04, P = 0.004
  rs4807927191374199AG0.945.59 × 10−60.1470.032DAZAP1RS: 0.21 ± 0.04, P = 1.32 × 10−6; NWR: .015 ± 0.04, P = 0.0003
  rs1357978713526746AC0.305.84 × 10−60.0700.016 RS: 0.09 ± 0.02, P = 5.46 × 10−6; NWR: 0.05 ± 0.02, P = 0.004
Non-word repetition (NWR)        
  rs21921612114603865AG0.067.34 × 10−80.1980.037ABCC13 
  rs28225602114596503TC0.058.73 × 10−80.2160.040ABCC13 
  rs71872231681015234AG0.969.90 × 10−80.2510.047  
  rs28226092114639665AT0.964.64 × 10−7−0.2100.042  
  rs124825282114632164TC0.076.04 × 10−70.1890.037ABCC13 
  rs72025751681005610TC0.049.08 × 10−7−0.2270.046  
  rs6954796720053507CG0.131.59 × 10−60.1240.026  
  rs2041660720074148TC0.133.30 × 10−60.1240.027  
  rs7777159720051126AG0.155.81 × 10−60.1120.025  

The most significant association (P = 7.34 × 10−8) was observed between rs2192161—located in ABCC13—and NWR. The minor allele (frequency of 6%) was associated with increased reading skill. Two other SNPs in this gene were also suggestively significant. Figure S3 shows the results for the surrounding region of this SNP in detail including SNP LD and the location of nearby genes. The SNPs in DAZAP1 were suggestively associated with the reading and spelling measure (P = 1.32 × 10−6; see Fig. S4) and the word reading measure (P = 5.59 × 10−6). Correlated traits that showed nominal significance are also highlighted; we were interested especially in whether reading and NWR SNP associations overlapped. Although 10 of the suggestively associated SNPs for the reading measures showed nominal association with NWR, none of the suggestively significant NWR associations were nominally associated with reading traits.

Gene-based tests

Results for the 10 most significant genes according to VEGAS are shown in Table S2. All genes significant at a nominal level (P < 0.05) are shown in Table S3. Four genes (at two independent loci) associated with the reading and spelling measure reached the threshold for Bonferroni significance, including CDC2L1, CDC2L2, LOC728661 and RCAN3. Four genes showing nominal significance for the reading and spelling measure also appeared in the top 10 for the word reading measure, including CDC2L1, CDC2L2, LOC728661 and RPS15.

Single SNP replication of dyslexia and SLI candidate genes from the literature

BATS

Our group has already published on candidate dyslexia genes KIAA0319, DYX1C1 and DCDC2 based on targeted genotyping of regions in a smaller dataset than is used here (Bates et al. 2010; Lind et al. 2010; Luciano et al. 2007, 2011). Since this time, additional subjects have entered the dataset, primarily recruited at age 12. Therefore, we re-examined the results for these SNP associations in the larger, but more age-diverse sample. It should be noted too that slight differences in the covariate adjustments were used between studies.

For KIAA0319, we had originally observed association with two SNPS; of these, rs2143340 was available in the current GWAS and was associated with the Schonell [major allele β =−0.06 (Standard Error (SE) 0.03), P = 0.02] but not the reading and spelling measure [major allele β =−0.09 (SE 0.06), P = 0.13]. For DCDC2, two SNPs had originally passed Bonferroni significance for association with CORE measures, but only one of these was available in the GWAS (rs1419228) and it fell short of significance with the CORE reading and spelling measure [major allele β = 0.106 (SE 0.06), P = 0.08].

We checked evidence of association between NWR and SNPs previously associated with SLI, having previously published on ROBO1 variants from the current GWAS (Bates et al. 2011). The results for CMIP and ATP2C2 genes (Newbury et al. 2009) were as follows: in CMIP: rs12927866, major (C) allele β = 0.071 (SE 0.05), P = 0.13; rs4265801, major (G) allele β =−0.062 (SE 0.05), P = 0.18; rs16955705, major (A) allele β = 0.087 (SE 0.05), P = 0.06, and three SNPs in ATP2C2: rs16973771, major (T) allele β = 0.023 (SE 0.05), P = 0.64; rs2875891: major (C) allele β = 0.042 (SE 0.05), P = 0.40; rs8045507, major (G) allele β = 0.027 (SE 0.05), P = 0.58. Nine SNPS in CNTNAP2 reached corrected significance in the Vernes et al. (2008) study of nonsense word repetition in children with SLI and their non-affected siblings. Six of these were available in our GWAS: rs851715, major (T) allele β = 0.046 (SE 0.05), P = 0.35; rs10246256, major (T) allele β = 0.05 (SE 0.05), P = 0.32; rs2710102, major (G) allele β = 0.038 (SE 0.05), P = 0.41; rs759178, major (C) allele β = 0.038 (SE 0.05), P = 0.41; rs17236239, major (A) allele β = 0.03 (SE 0.05), P = 0.52 and rs2710117, major (A) allele β = 0.058 (SE 0.05), P = 0.21. Finally, of the 11 available SNPs in LD with a haplotype found to be significantly associated with SLI in an isolated Chilean population (three of which were located in CNTNAP2) (Villanueva et al. 2011) none showed association.

ALSPAC sample

Targeted SNP association has previously been reported in ALSPAC for variants in MRPL19/C2ORF3, DCDC2, KIAA0319, CMIP and ATP2C2: only rs2143340 (in KIAA0319) showed Bonferroni-corrected association with reading ability (Scerri et al. 2011). Because these analyses excluded reading traits at age 9 and were conducted in a smaller sample (N ∼ 3700) that had filtered out 1036 individuals with non-white ethnicity, low performance IQ (<85) and likely presence of autistic traits we report the result for rs2143340 here. In the current GWAS dataset, association with rs2143340 was marginally significant (i.e. ∼P = 0.05) with the reading and spelling measure [minor allele β =−0.052 (SE 0.03), P = 0.06] and not associated with word reading [minor allele β =−0.008 (SE 0.03), P = 0.77]. For DYX1C1 marker rs17819126, the reading and spelling measure showed no association [major allele β =−0.002 (SE 0.04), P = 0.96] nor did the word reading measure [major allele β = 0.007 (SE 0.04), P = 0.86]. In the BATS, this marker was only associated with non-word reading (Bates et al. 2010)

For NWR, SNPs in CMIP and ATP2C2 have been previously tested for association in ALSPAC and had not been found to be significant (Newbury et al. 2009). Two SNPs (rs6803202 and rs4535189) in high LD in ROBO1 and associated with NWR in the BATS sample (Bates et al. 2011) were not significantly associated with NWR in ALSPAC (P = 0.88). The SNPS in CNTNAP2 previously reported as associated with nonsense word repetition (Vernes et al. 2008) showed the following results: rs851715, major (T) allele β = 0.012 (SE 0.02), P = 0.56; rs10246256, major (T) allele β = 0.013 (SE 0.02), P = 0.55; rs2710102, major (G) allele β =−0.028 (SE 0.02), P = 0.15; rs759178, major (C) allele β =−0.028 (SE 0.02), P = 0.14; rs17236239, major (A) allele β = 0.037 (SE 0.02), P = 0.06 and rs2710117, major (A) allele β = 0.01 (SE 0.02), P = 0.70. Of the 11 SNPs in LD with a haplotype reported as associated with SLI (Villanueva et al. 2011) we found association approaching significance with one: rs1371463 [minor allele β =−0.04 (SE 0.02), P = 0.08] in SVOPL.

Meta-analysis sample

Meta-analysis results of the replication SNPs detailed above showed nominally significant results for: rs17236239 (in CNTNAP2) where major (A) allele β = 0.04 (SE 0.02), P = 0.05 and for rs1371463 (in SVOPL) where major (C) allele β = 0.04 (SE 0.02), P = 0.03. All other results were non-significant.

Discussion

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. References
  7. Acknowledgments
  8. Supporting Information

Twenty-five independent SNPs were shown to be suggestively associated with three measures of reading and language in cohorts of British children and primarily adolescent Australians. The most significant association (P = 7.34 × 10−8) was observed between rs2192161—located in the ABCC13 [ATP-binding cassette, subfamily C (CFTR/MRP), member 13] pseudogene on 21q11.2—and NWR, the measure of language. Given that there was no nominal association of this SNP to the reading measures, this might be an SLI-specific locus. It is not located near previously reported dyslexia loci and given its low MAF (6%) this finding might represent type 1 error, but if not, this variant confers a large effect, explaining 3.9% of the variance. ABCC13 is an ABC gene that differs functionally between monkeys and apes. The gene codes for a functional protein in the rhesus macaque, where it plays a role in transport of ATP across membranes (Annilo & Dean 2004). In humans, it is a pseudogene with no known role in transporting activity but instead encodes a truncated protein that is expressed in fetal human liver, similar to ABCC2 (Zhou et al. 2008).

All the suggestively significant SNP associations for the reading and spelling measure showed nominal significance for the word reading measure and vice versa. One SNP, rs4807927 [located in DAZ-associated protein 1 (DAZAP1)], was associated with both the reading and spelling measure and the word reading measure at a suggestive significance level, so it may represent a more robust finding than the others given that these traits are moderately correlated. Most of these SNPs were also nominally associated with the language measure. The overlap in SNP findings across traits might suggest their effect on general reading (and even language) processes. A multivariate analysis of more distinct individual reading component measures is needed to confirm the magnitude of any shared effects. If the SNP effect is at the ‘general processes’ level, then such analysis might even increase statistical power (O'Reilly et al. 2012).

Four of the gene-based tests reached Bonferroni-corrected significance for the reading and spelling measure, including CDC2L1 (Cell Division Cycle 2-Like 1), CDC2L2 (Cell Division Cycle 2-Like 2), LOC728661 and RCAN3 (RCAN family member 3) located on chromosome 1. The first three genes represented one locus because their chromosomal region overlapped, but RCAN3 was an independent locus. Four genes appeared in the top 10 most significant results for both the reading and spelling measure and the word reading measure: CDC2L1, CDC2L2, LOC728661 and RPS15 (Ribosomal Protein S15). These genes have not previously been reported in relation to reading traits and their biological relevance as candidate genes is unclear.

Previously implicated SNPs/genes associated with reading or SLI showed somewhat inconsistent, albeit potentially informative, results across our two cohorts. As previously reported by Luciano et al. (2007), rs2143340 in KIAA0314 was associated with word reading in BATS but was only marginally significant in ALSPAC. In a subsample of ALSPAC, which excluded non-white, very low IQ and potentially autistic children this SNP was, however, found to be significant (Scerri et al. 2011). Our finding suggests, then, that background population structure as discussed by Paracchini et al. (2008) can be a barrier to replication for genetic association. The exclusion of individuals with comorbidities (like autism and Attention Deficit Hyperactivity Disorder) may also be advisable, reducing non-reading specific variation in reading ability attributable instead to comorbid disorders. Of course, for a complete understanding of genetic influences on reading ability, it will be important also to understand and incorporate these alternate influences on reading ability.

Skiba et al. (2011) have discussed at length the problems that phenotype heterogeneity present for replication studies of reading ability/dyslexia. The use of comorbid reading and language impaired samples has been shown to produce a particularly sensitive genetic design, perhaps reflecting severe affection (e.g. see Scerri et al. 2011). By contrast, if genes have specific functionality, such comorbidity may diffuse genetic signals. In the present case, for instance, DYX1C1 SNP, rs1781926, was not associated with the reading measures in the ALSPAC. However, this SNP was associated in BATS with non-word reading (Bates et al. 2010). This may be interpreted as an example of gene specificity, in this case for phonological decoding. For NWR, replication results approached significance for SNPs in CMIP (in BATS but not ALSPAC) and in CNTNAP2 and SVOPL (for both genes in ALSPAC but not BATS). Again, differences between cohorts due to ethnic composition and age might explain these discrepancies. Because these replication results were not the focus of our investigation, we only presented results for the most relevant previous SNP associations; obviously, a more thorough replication analysis should focus on all SNPs in the gene with adjustments made for within-gene SNP LD.

The cohorts in this study differed significantly in age: in the Australian sample, the mean age was 17.9 years, a period when reading acquisition is largely stable; in the British sample, aged 9 years, reading skill is still actively being acquired, and environmental and developmental timing differences may play a larger role in the phenotype. It may even be the case that different genetic substrates are involved in this earlier phase of reading acquisition. Our GWAS results therefore apply to gene effects on reading ability, which are present across both childhood and adolescence and into young adulthood. Previous research on the increase of heritability across development, and the continuity of genetic influences suggests increasing heritability or genetic penetrance with age, and has supported both substantial genetic continuity and significant genetic innovation. For instance, a longitudinal adoption study (Wadsworth et al. 2001) showed that between the ages of 7 and 16 years common genetic effects explain 88% of the stability in reading performance. Moreover, no new genetic effects emerged at age 12 or 16, suggesting that the genetic effects at age 7 were simply amplified later on. However, a much larger longitudinal twin study of teacher-assessed reading achievement across the ages of 7, 9 and 10 years showed considerable genetic continuity (68% and 77% of stability) and also evidence for novel genetic contributions at both the latter two ages (Harlaar et al. 2007). Clearly, more GWAS samples are needed to obtain the increased power required to investigate age-specific genome-wide effects on reading ability. The loci identified in this study (including the top 100 SNP associations and gene-based tests) will serve as an important repository for future GWAS of reading and language abilities and disorders to confirm overlap with their own findings.

References

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. References
  7. Acknowledgments
  8. Supporting Information
  • Annilo, T. & Dean, M. (2004) Degeneration of an ATP-binding cassette transporter gene, ABCC13, in different mammalian lineages. Genomics 84, 3446.
  • Bates, T.C., Castles, A., Coltheart, M., Gillespie, N., Wright, M.J. & Martin, N.G. (2004) Behaviour genetic analyses of reading and spelling: a component processes approach. Aust J Psychol 56, 115126.
  • Bates, T.C., Lind, P.A., Luciano, M., Montgomery, G.W., Martin, N.G. & Wright, M.J. (2010) Dyslexia and DYX1C1: deficits in reading and spelling associated with a missense mutation. Mol Psychiatry 15, 11901196.
  • Bates, T.C., Luciano, M., Medland, S.E., Montgomery, G.W., Wright, M.J. & Martin, N.G. (2011) Genetic variance in a component of the language acquisition device: ROBO1 polymorphisms associated with phonological buffer deficits. Behav Genet 41, 5057.
  • Bishop, D.V. (2001) Genetic influences on language impairment and literacy problems in children: same or different? J Child Psychol Psychiatry 42, 189198.
  • Bishop, D.V.M. & Hayiou-Thomas, M.E. (2008) Heritability of specific language impairment depends on diagnostic criteria. Genes Brain Behav 7, 365372.
  • Bishop, D.V., Laws, G., Adams, C. & Norbury, C.F. (2006) High heritability of speech and language impairments in 6-year-old twins demonstrated using parent and teacher report. Behav Genet 36, 173184.
  • Castles, A. & Coltheart, M. (1993) Varieties of developmental dyslexia. Cognition 47, 149180.
  • Chen, W.M. & Abecasis, G.R. (2007) Family-based association tests for genomewide association scans. Am J Hum Genet 81, 913926.
  • Cope, N., Harold, D., Hill, G., Moskvina, V., Stevenson, J., Holmans, P., Owen, M.J., O'Donovan, M.C. & Williams, J. (2005) Strong evidence that KIAA0319 on chromosome 6p is a susceptibility gene for developmental dyslexia. Am J Hum Genet 76, 581591.
  • Dollaghan, C. & Campbell, T.F. (1998) Nonword repetition and child language impairment. J Speech Lang Hear Res 41, 11361146.
  • Dudbridge, F. & Gusnanto, A. (2008) Estimation of significance thresholds for genomewide association scans. Genet Epidemiol 32, 227234.
  • Duggal, P., Gillanders, E.M., Holmes, T.N. & Bailey-Wilson, J.E. (2008) Establishing an adjusted p-value threshold to control the family-wide type 1 error in genome wide association studies. BMC Genomics 9, 516.
  • Field, L.L., Shumansky, K., Ryan, J., Truong, D., Swiergala, E. & Kaplan, B.J. (2013) Dense-map genome scan for dyslexia supports loci at 4q13, 16p12, 17q22; suggests novel locus at 7q36. Genes Brain Behav 12, 5669.
  • Gathercole, S.E. (1994) Neuropsychology and working memory: a review. Neuropsychology 8, 494505.
  • Golding, J., Pembrey, M. & Jones, R. (2001) ALSPAC--the Avon Longitudinal Study of Parents and Children. I. Study methodology. Paediatr Perinat Epidemiol 15, 7487.
  • Hannula-Jouppi, K., Kaminen-Ahola, N., Taipale, M., Eklund, R., Nopola-Hemmi, J., Kaariainen, H. & Kere, J. (2005) The axon guidance receptor gene ROBO1 is a candidate gene for developmental dyslexia. PLoS Genet 1, e50.
  • Harlaar, N., Spinath, F.M., Dale, P.S. & Plomin, R. (2005) Genetic influences on early word recognition abilities and disabilities: a study of 7-year-old twins. J Child Psychol Psychiatry 46, 373384.
  • Harlaar, N., Dale, P.S. & Plomin, R. (2007) From learning to read to reading to learn: substantial and stable genetic influence. Child Dev 78, 116131.
  • Hawke, J.L., Wadsworth, S.J. & DeFries, J.C. (2006) Genetic influences on reading difficulties in boys and girls: the Colorado twin study. Dyslexia 12, 2129.
  • Jackson, D.N. (1984) Manual for the Multidimensional Aptitude Battery. Research Psychologists Press, Port Huron, MI.
  • Li, Y., Willer, C., Sanna, S. & Abecasis, G. (2009) Genotype imputation. Annu Rev Genomics Hum Genet 10, 387406.
  • Li, Y., Willer, C.J., Ding, J., Scheet, P. & Abecasis, G.R. (2010) MaCH: using sequence and genotype data to estimate haplotypes and unobserved genotypes. Genet Epidemiol 34, 816834.
  • Lind, P.A., Luciano, M., Wright, M.J., Montgomery, G.W., Martin, N.G. & Bates, T.C. (2010) Dyslexia and DCDC2: normal variation in reading and spelling is associated with DCDC2 polymorphisms in an Australian population sample. Eur J Hum Genet 18, 668673.
  • Liu, J.Z., McRae, A.F., Nyholt, D.R., Medland, S.E., Wray, N.R., Brown, K.M., Hayward, N.K., Montgomery, G.W., Visscher, P.M., Martin, N.G. & Macgregor, S. (2010) A versatile gene-based test for genome-wide association studies. Am J Hum Genet 87, 139145.
  • Luciano, M., Lind, P.A., Duffy, D.L., Castles, A., Wright, M.J., Montgomery, G.W., Martin, N.G. & Bates, T.C. (2007) A haplotype spanning KIAA0319 and TTRAP is associated with normal variation in reading and spelling ability. Biol Psychiatry 62, 811817.
  • Luciano, M., Montgomery, G.W., Martin, N.G., Wright, M.J. & Bates, T.C. (2011) SNP sets and reading ability: testing confirmation of a 10-SNP set in a population sample. Twin Res Hum Genet 14, 228232.
  • Meaburn, E.L., Harlaar, N., Craig, I.W., Schalkwyk, L.C. & Plomin, R. (2008) Quantitative trait locus association scan of early reading disability and ability using pooled DNA and 100K SNP microarrays in a sample of 5760 children. Mol Psychiatry 13, 729740.
  • Medland, S.E., Nyholt, D.R., Painter, J.N., McEvoy, B.P., McRae, A.F., Zhu, G., Gordon, S.D., Ferreira, M.A., Wright, M.J., Henders, A.K., Campbell, M.J., Duffy, D.L., Hansell, N.K., Macgregor, S., Slutske, W.S., Heath, A.C., Montgomery, G.W. & Martin, N.G. (2009) Common variants in the trichohyalin gene are associated with straight hair in Europeans. Am J Hum Genet 85, 750755.
  • Meng, H., Smith, S.D., Hager, K., Held, M., Liu, J., Olson, R.K., Pennington, B.F., Defries, J.C., Gelernter, J., O'Reilly-Pol, T., Somlo, S., Skudlarski, P., Shaywitz, S.E., Shaywitz, B.A., Marchione, K., Wang, Y., Paramasivam, M., Loturco, J.J., Page, G.P. & Gruen, J.R. (2005) DCDC2 is associated with reading disability and modulates neuronal development in the brain. Proc Natl Acad Sci U S A 102, 1705317058.
  • Miller, M.B. (1998) Monozygotic twins increase the power of genetic association studies of complex phenotypes. Behav Genet 28, 476.
  • Newbury, D.F., Winchester, L., Addis, L. et al (2009) CMIP and ATP2C2 modulate phonological short-term memory in language impairment. Am J Hum Genet 85, 264272.
  • Newbury, D.F., Paracchini, S., Scerri, T.S., Winchester, L., Addis, L., Richardson, A.J., Walter, J., Stein, J.F., Talcott, J.B. & Monaco, A.P. (2011) Investigation of dyslexia and SLI risk variants in reading- and language-impaired subjects. Behav Genet 41, 90104.
  • Nunes, T., Bryant, P. & Olsson, J. (2003) Learning morphological and phonological spelling rules: an intervention study. Sci Stud Read 7, 289307.
  • O'Reilly, P.F., Hoggart, C.J., Pomyen, Y., Calboli, F.C., Elliott, P., Jarvelin, M.R. & Coin, L.J. (2012) MultiPhen: joint model of multiple phenotypes can increase discovery in GWAS. PLoS One 7, e34861.
  • Panagiotou, O.A. & Ioannidis, J.P. (2012) What should the genome-wide significance threshold be? Empirical replication of borderline genetic associations. Int J Epidemiol 41, 273286.
  • Paracchini, S., Steer, C.D., Buckingham, L.L., Morris, A.P., Ring, S., Scerri, T., Stein, J., Pembrey, M.E., Ragoussis, J., Golding, J. & Monaco, A.P. (2008) Association of the KIAA0319 dyslexia susceptibility gene with reading skills in the general population. Am J Psychiatry 165, 15761584.
  • Rice, M.L., Smith, S.D. & Gayan, J. (2009) Convergent genetic linkage and associations to language, speech and reading measures in families of probands with specific language impairment. J Neurodev Disord 1, 264282.
  • Roeske, D., Ludwig, K.U., Neuhoff, N., Becker, J., Bartling, J., Bruder, J., Brockschmidt, F.F., Warnke, A., Remschmidt, H., Hoffmann, P., Muller-Myhsok, B., Nothen, M.M. & Schulte-Korne, G. (2011) First genome-wide association scan on neurophysiological endophenotypes points to trans-regulation effects on SLC2A3 in dyslexic children. Mol Psychiatry 16, 97107.
  • Scerri, T.S., Morris, A.P., Buckingham, L.L., Newbury, D.F., Miller, L.L., Monaco, A.P., Bishop, D.V. & Paracchini, S. (2011) DCDC2, KIAA0319 and CMIP are associated with reading-related traits. Biol Psychiatry 70, 237245.
  • Schonell, F.J. & Schonell, P.E. (1960) Diagnostic and Attainment Testing. Oliver & Boyd, Edinburgh.
  • Skiba, T., Landi, N., Wagner, R. & Grigorenko, E.L. (2011) In search of the perfect phenotype: an analysis of linkage and association studies of reading and reading-related processes. Behav Genet 41, 630.
  • SLI Consortium (2002) A genomewide scan identifies two novel loci involved in specific language impairment. Am J Hum Genet 70, 384398.
  • Taipale, M., Kaminen, N., Nopola-Hemmi, J., Haltia, T., Myllyluoma, B., Lyytinen, H., Muller, K., Kaaranen, M., Lindsberg, P.J., Hannula-Jouppi, K. & Kere, J. (2003) A candidate gene for developmental dyslexia encodes a nuclear tetratricopeptide repeat domain protein dynamically regulated in brain. Proc Natl Acad Sci U S A 100, 1155311558.
  • Tran, C., Gagnon, F., Wigg, K.G., Feng, Y., Gomez, L., Cate-Carter, T.D., Kerr, E.N., Field, L.L., Kaplan, B.J., Lovett, M.W. & Barr, C.L. (2013) A family-based association analysis and meta-analysis of the reading disabilities candidate gene DYX1C1. Am J Med Genet B Neuropsychiatr Genet 162, 146156.
  • Vernes, S.C., Newbury, D.F., Abrahams, B.S., Winchester, L., Nicod, J., Groszer, M., Alarcon, M., Oliver, P.L., Davies, K.E., Geschwind, D.H., Monaco, A.P. & Fisher, S.E. (2008) A functional genetic link between distinct developmental language disorders. N Engl J Med 359, 23372345.
  • Villanueva, P., Newbury, D.F., Jara, L., De Barbieri, Z., Mirza, G., Palomino, H.M., Fernandez, M.A., Cazier, J.B., Monaco, A.P. & Palomino, H. (2011) Genome-wide analysis of genetic susceptibility to language impairment in an isolated Chilean population. Eur J Hum Genet 19, 687695.
  • Wadsworth, S.J., Corley, R.P., Hewitt, J.K. & Defries, J.C. (2001) Stability of genetic and environmental influences on reading performance at 7, 12, and 16 years of age in the Colorado Adoption Project. Behav Genet 31, 353359.
  • Wechsler, D., Golombok, S. & Rust, J. (1992) WISC-IIIUK Wechsler Intelligence Scale for Children: UK Manual, 3rd edn. Psychological Corporation, Sidcup.
  • Whitehouse, A.J., Bishop, D.V., Ang, Q.W., Pennell, C.E. & Fisher, S.E. (2011) CNTNAP2 variants affect early language development in the general population. Genes Brain Behav 10, 451456.
  • Willer, C.J., Li, Y. & Abecasis, G.R. (2010) METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics 26, 21902191.
  • Wright, M., De Geus, E., Ando, J., Luciano, M., Posthuma, D., Ono, Y., Hansell, N., Van Baal, C., Hiraishi, K., Hasegawa, T., Smith, G., Geffen, G., Geffen, L., Kanba, S., Miyake, A., Martin, N. & Boomsma, D. (2001) Genetics of cognition: outline of a collaborative twin study. Twin Res 4, 4856.
  • Zhong, R., Yang, B., Tang, H., Zou, L., Song, R., Zhu, L.Q. & Miao, X. (2013) Meta-analysis of the association between DCDC2 polymorphisms and risk of dyslexia. Mol Neurobiol 47, 435442.
  • Zhou, S.F., Wang, L.L., Di, Y.M., Xue, C.C., Duan, W., Li, C.G. & Li, Y. (2008) Substrates and inhibitors of human multidrug resistance associated proteins and the implications in drug development. Curr Med Chem 15, 19812039.
  • Zou, L., Chen, W., Shao, S., Sun, Z., Zhong, R., Shi, J., Miao, X. & Song, R. (2012) Genetic variant in KIAA0319, but not in DYX1C1, is associated with risk of dyslexia: an integrated meta-analysis. Am J Med Genet B Neuropsychiatr Genet 159B, 970976.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. References
  7. Acknowledgments
  8. Supporting Information

BATS: We especially thank the twins and their families for participation; the project managers who coordinated the reading and language projects; the research nurses for phenotype and blood collection; staff of the Molecular Epidemiology Laboratory for DNA sample processing and preparation; Genetic Epidemiology Computing staff for IT support and Drs Dale R. Nyholt and Scott Gordon for their substantial efforts involving the QC and preparation of the GWAS datasets. The research was supported by the Australian Research Council (A7960034, A79906588, A79801419, DP0212016 and DP0343921), with genotyping funded by the National Health and Medical Research Council (Medical Bioinformatics Genomics Proteomics Program, 389891). Genotype imputation was carried out on the GenEpi Cluster, which is financially supported by contributions from grants from the NHMRC (389892, 496682, 496688, 496739 and 613672) and ARC (FT0991022 and FT0991360). ALSPAC: We are extremely grateful to all the individuals who took part in this study, the midwives for their help in recruiting them and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses. We thank the Sanger Centre, Centre National de Geánotypage and 23andMe for generating the ALSPAC GWA data. The publication is the work of the authors and D.M.E. will serve as guarantor for the contents of this article. The UK Medical Research Council, the Wellcome Trust (grant ref: 092731) and the University of Bristol provide core support for the Avon Longitudinal Study of Parents and their Children. The authors declare no conflict of interest.

Supporting Information

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. References
  7. Acknowledgments
  8. Supporting Information
FilenameFormatSizeDescription
gbb12053-sup-0001-TableS1.xlsExcel spreadsheet77KTable S1: GWA results for the 100 most significant associations for measures of reading and spelling, word reading and non-word repetition.
gbb12053-sup-0002-TableS2.docWord document50KTable S2: Top 10 most significant genes for reading and language traits as evaluated by VEGAS. Significant genes appear in bold and genes associated with multiple traits are superscripted.
gbb12053-sup-0003-TableS3.xlsExcel spreadsheet558KTable S3: Nominally significant (P < 0.05) results for gene-based tests of association with reading and spelling, word reading and non-word repetition.
gbb12053-sup-0004-FigureS1.pptPowerPoint presentation130KFigure S1: Q–Q plots of the GWAS association meta-analysis results for reading and spelling, word reading and non-word repetition.
gbb12053-sup-0005-FigureS2.pptPowerPoint presentation171KFigure S2: Manhattan plots depicting the GWAS association meta-analysis results for reading and spelling, word reading and non-word repetition.
gbb12053-sup-0006-FigureS3.pptPowerPoint presentation189KFigure S3: Annotation for NWR association in region surrounding rs2192161. Linkage disequilibrium is represented by r2. No non-synonymous coding SNPs can be observed.
gbb12053-sup-0007-FigureS4.pptPowerPoint presentation189KFigure S4: Annotation for the reading and spelling association in the region surrounding rs4807927. Linkage disequilibrium is represented by r2. Non-synonymous (image) and coding (image) SNPs can be observed.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.