Hum Genet (2004) 115: 128–138 DOI 10.1007/s00439-004-1126-6
ORIGINA L IN VESTI GATION
Karen E. Deffenbacher . Judith B. Kenyon . Denise M. Hoover . Richard K. Olson . Bruce F. Pennington . John C. DeFries . Shelley D. Smith
Refinement of the 6p21.3 quantitative trait locus influencing dyslexia: linkage and association analyses Received: 20 August 2003 / Accepted: 23 March 2004 / Published online: 11 May 2004 # Springer-Verlag 2004
Abstract Reading disability (RD), or dyslexia, is the most common learning disability with a prevalence rate of ~5%–10% in school-age children. RD is highly heritable with evidence of a neurobiological origin. Linkage studies have identified several quantitative trait loci (QTLs) for RD. The QTL on chromosome 6p21.3 has been independently replicated by several groups and spans a 16.4-Mb (13.8 cM) interval from D6S109 to D6S291. In this study, we performed sib-pair linkage analyses with Haseman–Elston and DeFries–Fulker methods to define more accurately the QTL interval. Linkage was assessed by using five quantitative phenotypes, including a composite measure of reading performance and four component phenotypes. When probands were selected for severe scores, single- and multi-point analyses showed significant linkage with all five phenotypes, converging over an interval of ~3.24 Mb spanning D6S1597 to D6S1571. Maximal linkage converged at marker D6S1554 across phenotypes. Out of 12 genes in the linkage interval, ten clustered within ~680 kb and were selected for association analysis based on central nervous system expression and putative function. Marker-trait associations were assessed by using QTDT (a general test of association for quantitative traits) and the familybased association test (FBAT), and haplotype analysis was performed by using FBAT and the GeneHunter TransmisK. E. Deffenbacher (*) . J. B. Kenyon . D. M. Hoover . S. D. Smith Center for Human Molecular Genetics, Munroe Meyer Institute, University of Nebraska Medical Center, 985455, NE Medical Center MMI Rm. 3085, Omaha, NE 68198-5455, USA e-mail:
[email protected] Tel.: +1-402-5592862 Fax: +1-402-5594001 B. F. Pennington University of Denver, Denver, Colo., USA R. K. Olson . J. C. DeFries Institute for Behavioral Genetics, University of Colorado, Boulder, Colo., USA
sion/Disequilibrium Test TDT. Marker associations were detected in five of the ten genes, results that were corroborated by our haplotype TDT analysis. The results of the association study have thereby allowed us to significantly reduce the number of possible candidate genes and to prioritize genes for further mutation screening.
Introduction Reading disability (RD), also known as dyslexia, is defined as a specific and significant impairment in reading ability not explained by deficits in intelligence, learning opportunity or sensory acuity (Lyon 1995). RD accounts for more than 80% of all learning disabilities (Lerner 1989) with typical prevalence rates estimated at 5%–10% in school-age children (Shaywitz et al. 1990). RD is known to be highly heritable with hg2 estimates ranging from 0.4 to 0.6 (Gayán and Olson 2001). Functional neuroimaging studies suggest a neurobiological origin for RD, documenting significant differences in the activation of specific brain regions between RD and control subjects (Habib 2000). RD is identified through the assessment of a battery of psychometric tests that measure reading and spelling ability. Component reading processes, believed to be integral to reading as a whole, have been examined for use in linkage studies and are derived from a set of experimental measures. The measured component processes test both phonological and orthographic skills including phoneme awareness (PA), phonological decoding (PD), single word reading (WR) and orthographic coding (OC; Olson et al. 1989). By targeting processes specific to RD, the use of component phenotypes in linkage studies provides a more precise phenotypic definition, thereby reducing heterogeneity and facilitating the detection of linkage with a complex trait. Quantitative trait loci (QTLs) influencing RD have been reported on chromosomes 1p36 (Rabin et al. 1993) and 2p12–16 (Fagerheim et al. 1999; Fisher et al. 2002), the pericentromeric region of chromosome 3 (Nopola-Hemmi
129
et al. 2001), and chromosomes 6p21.3 (Cardon et al. 1994; Fisher et al. 1999; Gayàn et al. 1999; Grigorenko et al. 1997; Kaplan et al. 2002), 15q21 (Morris et al. 2000; Nöthen et al. 1999; Fulker et al. 1991), 6q (Petryshen et al. 2001), and 18p11.2 (Fisher et al. 2002). Several groups have independently replicated the finding on chromosome 6p over variable yet overlapping intervals, spanning ~16.4 Mb between markers D6S109 and D6S291. Gayán et al. (1999) have reported significant linkage with quantitative measures of component skills over a 5-cM region between markers D6S461 and D6S306, with a peak of linkage between D6S276 and D6S105. Other reports include a 2-cM region surrounding D6S105 (Cardon et al. 1994), a 15.8-Mb interval from D6S422 to D6S291 (Fisher et al. 1999), an interval of ~4-Mb from D6S464 to D6S273 (Grigorenko et al. 2003), and a 4-Mb interval surrounding marker JA04 (Kaplan et al. 2002). One study failed to detect linkage in the 6p21.3 region with both qualitative and quantitative measures (Field and Kaplan 1998; Petryshen et al. 2000); this study used narrow ascertainment criteria, selecting only families with a phonological coding dyslexia phenotype. Further, linkage to this region has previously been shown to be significantly enhanced when the sample is selected for lower scores, defining probands as having scores ≤2 SD below the population mean (Cardon et al. 1994; Gayán et al. 1999). Replication of linkage is difficult to achieve with complex traits because of variable phenotypic definition, phenocopy, heterogeneity, incomplete penetrance, and oligogenicity. Thus, the report of several independent replications for the 6p21.3 RD QTL has made it the most robust linkage report for a complex cognitive trait. Whereas the regions of linkage on chromosome 6p vary, they all lie within the 15.8-Mb interval defined by markers D6S422 to D6S291. There is consensus for a peak of linkage, or most likely position for the QTL, within the 4.2-Mb interval from D6S461 to D6S105. Further resolution of this interval has proven difficult, as linkage analysis of complex traits is confounded by the presence of phenocopy, incomplete penetrance, and heterogeneity, thus decreasing the limits of resolution for this technique and imposing a requirement for much larger sample sizes. Association analysis can facilitate the fine-mapping of linkage regions giving resolution in the order of kilobases (kb). Because of the decay of linkage disequilibrium (LD) over time and distance, closely spaced markers are necessary, providing high resolution of candidate regions. Association studies also have the advantage of greater statistical power for traits for which there may be a number of genes of small effect (Risch and Teng 1998). Familybased association methods, such as the Transmission/ Disequilibrium Test (TDT; Spielman et al. 1993), are useful for linkage fine-mapping since they utilize parent/ proband trios. The TDT compares the frequencies of transmitted versus untransmitted alleles in affected offspring, by using the untransmitted parental alleles as controls. The use of parental alleles as controls obviates population stratification, providing an advantage over epidemiologic case-control study designs.
Two family-based association studies with microsatellite markers have been reported for RD. Using the TDT and haplotype relative risk methods, Morris et al. (2000) reported a significant single-marker and a three-marker haplotype association with RD on chromosome 15q. Using QTDT (a general test of association for quantitative traits), Kaplan et al. (2002) recently reported significant association between several RD component phenotypes and short tandem repeat (STR) markers on chromosome 6p21.3. They found a peak of transmission disequilibrium at marker JA04, suggesting a likely QTL position in the 4Mb region surrounding this marker. This study used markers spaced ~300 kb apart on average, with little to no intermarker LD. In order to fine-map this region further, a higher marker density should be used. The high density of single-nucleotide polymorphisms (SNPs) in the genome meets this criterion allowing for candidate gene approaches and haplotype analysis. This study is the first report using high density SNP map to refine the 6p21.3 RD QTL further. The goal of this study was to refine the 6p QTL through linkage and association analysis. This study is an extension of that of Gayán et al. (1999) who, in a study of eight markers, reported linkage in 79 of the families used. Genotyping of 22 STR markers spanning 22.58 Mb (22.52 cM) on chromosome 6p was carried out on 349 nuclear families. An overall measure of reading ability, discriminant score (DISC), and the five component phenotypes examined by Gayán et al. (1999) were analyzed for linkage and association. With our extended sample, sib-pair linkage analyses narrowed the linkage interval to ~3.24 Mb between markers D6S1597 and D6S1571, with convergence of maximal linkage across phenotypes at marker D6S1554. A high density SNP map was constructed over ~680 kb of this interval based on a candidate gene approach. Single-marker and haplotype analyses identified significant associations in five genes. Thus, the results from this study have further refined the 6p QTL, enabling us to prioritize candidate genes for mutation analysis.
Methods Subjects Dizygotic (DZ) and monozygotic (MZ) twin pairs and their families were recruited from Colorado school districts by the Colorado Learning Disabilities Research Center (CLDRC) at the Institute for Behavioral Genetics. Families in which at least one twin had a school history of reading problems were ascertained. Subjects with sensory deficits, neurological, or emotional problems were excluded from the sample. Molecular genetic analyses were performed on DZ twins and any siblings or on one MZ twin and non-twin siblings. The sample used in this study included 1,559 individuals from 349 nuclear families. Some of these families were used in previous studies (Cardon et al. 1994; Gayán et al. 1999; Kaplan et al. 2002), with 64% of the families in this study being new. The Kaplan et al. (2002) study utilized 104 families included in the present analyses, although many of the STR markers were unique between studies, and none of the SNP markers were previously genotyped in this sample. Of the 349 families, there were 218 with two offspring, 100
130 Table 1 Linkage marker map
a Relative physical distance between markers listed in Mb, obtained from November 2002 freeze of UCSC genome browser b Relative genetic distance listed in cM, obtained from deCODE genetics c Average recombination rate between markers expressed in cM/Mb, obtained from deCODE genetics d Marker heterozygosities were computed with our sample
Marker
Distance (Mb)a
Distance (cM)b
Average recombination ratec
Heterozygosityd (%)
D6S1605 D6S274 D6S1567 D6S422 D6S1597 D6S1588 D6S1663 D6S461 D6S276 D6S1554 D6S1571 D6S105 D6S306 D6S258 D6S1683 MOG D6S273 D6S1666 D6S1568 D6S439 D6S291 D6S1019
0 0.34 1.12 4.05 5.35 5.73 6.16 7.24 7.82 8.56 8.71 11.45 11.48 12.61 12.66 13.33 15.41 16.12 17.68 18.82 19.93 22.58
0 1.5 2.71 7.13 8.23 9.16 10.36 11.35 12.23 13.16 13.56 15.05 15.1 15.15 15.2 15.26 15.61 21.92 18.15 18.98 19.81 22.52
1.74 1.77 1.74 1.11 1.39 1.45 1.52 1.46 1.33 1.05 1 0.26 0.26 – 0.21 – 0.46 – 0.62 – 0.77 –
77.5 80.3 77.6 73.9 58.3 66.8 61.1 69.8 75.3 70.7 74.6 81.1 61.4 72.5 69.0 73.1 79.8 70.3 87.1 66.8 70.5 66.7
with three offspring, 30 with four offspring, and one with five offspring, generating a total of 708 sib-pairs or 512 independent pairs (computed as n−1 per family of n offspring). Twin pairs ascertained were between 8 and 19 years of age, with a mean age of 11.5 years. This study was approved by the Human Research Committee of the University of Colorado at Boulder and the Institutional Review Board at the University of Nebraska Medical Center. Informed consent was obtained from participants at the University of Colorado at Boulder.
Phenotypes Subjects were given a battery of psychometric tests at the CLDRC to assess reading, spelling, and cognitive ability including the Wechsler Intelligence Scale (Wechsler 1974, 1981) and the Peabody individual achievement test (Dunn and Markwardt 1970). These tests were measured in a known sample (with and without RD) from which a discriminant function was derived separating normal from disabled readers. The discriminant function was then applied to test results from our sample generating a quantitative DISC score assessing overall reading ability. Additional tests were given to measure component processes of reading, as described in detail in Gayán et al. (1999). Component phenotypes include PA (the ability to isolate and manipulate phonemes in speech), PD (the ability to read aloud nonwords), WR (single word reading), and OC (the ability to rapidly choose target words from pairs that included a phonologically identical foil, such as rain versus rane). Results from these tests were age-regressed and expressed as standard deviations (SD) from the population mean obtained from the twin database at the CLDRC, thereby forming a continuous distribution of z-scores for each phenotype. Group means in our sample, expressed in SD below the population mean, were as follows: PA=−1.00, PD=−1.20, OC=−0.79, WR=−1.37, and DISC=−1.40.
Marker panel and genotyping Twenty-two dinucleotide repeat markers were selected for genotyping spanning 22 Mb (22 cM) on chromosome 6p21.3 (Table 1). DNA was extracted from blood or buccal samples collected from parents and offspring. Genotyping was carried out with multiplexed fluorescently labeled primers on an ABI 3700 DNA Analyzer (Applied Biosystems). Allele calls were made by using Genotyper software version 3.7 (Applied Biosystems), and inheritance checking was performed with Genetic Analysis System (GAS) software, version 2.0 (A. Young, Oxford University, 1993–1995). The error function in Merlin (Abecasis et al. 2002) was used to flag potential genotyping errors for re-analysis.
Linkage analysis Multipoint identity-by-descent (IBD) values were computed with GeneHunter (GH; Kruglyak and Lander 1995). Single-point linkage analysis was performed with SIBPAL, implemented in S.A.G.E. 4.2 (S.A.G.E. 2002). SIBPAL is a non-parametric linkage program that models sib-pair covariance of a trait as a function of allele sharing IBD. Traditional Haseman–Elston (H-E) analysis (Haseman and Elston 1972) regresses the squared trait difference of sibs on the mean IBD sharing at a locus. S.A.G.E. 4.2 (S.A.G.E. 2002) also implements the revised H-E in which both the squared trait difference and the mean-corrected trait sum are used, weighted by the inverse of their residual variances. This revised method has been shown to have much greater statistical power to detect linkage (Elston et al. 2000). Our analyses utilize the revised H-E algorithm from which we report empirical P-values. Single-point analysis was performed both on our entire sample of 708 sib-pairs and on samples selected for severe scores. Maximal linkage was detected with each phenotype when the sample was selected by severity. At least one sib had to have a phenotypic score at the 20th percentile or lower. By using control group means, SD cut-points were derived for each phenotype by computing the z-score corresponding to the 20th percentile. These were −2 SD for OC, −2.5 SD for PA, −3 SD for
131 Table 2 SNP markers
a Assay ID is from Applied Biosystems Assays-on-Demand; rs IDs are from dbSNP at NCBI b Physical distance is the distance in base pairs between a SNP and the previous marker c Locus is the genic location of an SNP. Minor allele frequency was computed from a Caucasian control panel at ABI. Both locus and allele frequency were obtained from the ABI Assays-onDemand database
SNP no.
Assay IDa
Physical distance (bp)b
Locusc
Minor allele frequencyc
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
C 1809129 10 C 443745 10 C 9373644 10 C 7454570 10 rs 793854 C 7454462 10 C 9486883 10 C 7454704 10 C 9344981 1 C 7454814 10 C 7454810 10 C 7454790 10 C 2100395 10 C 2100415 10 C 2100443 10 C 2100480 10 C 7466744 1 C 2479666 10 C 2479683 10 C 3073694 10 C 3073662 10 rs2076314 C 1691926 10 rs2745334 C 2463871 10 C 2463856 10 C 3248054 10 C 2140732 10 C 16285816 10 C 2140695 10 C 151407 10
0 18,813 20,157 26,795 19,562 9,332 53,860 17,809 12,335 19,441 25,998 18,507 14,365 36,194 25,285 19,332 16,315 24,742 13,551 18,687 44,666 12,238 3,745 33,264 27,937 22,322 31,483 24,093 51,716 15,281 8,857
Intergenic VMP VMP DCDC2 DCDC2 DCDC2 DCDC2 DCDC2 DCDC2 DCDC2 DCDC2 DCDC2 Intergenic MRS2L MRS2L GPLD1 GPLD1 ALDH5A1 ALDH5A1 ALDH5A1 KIAA0319 KIAA0319 KIAA0319 KIAA0319 TTRAP THEM2 THEM2 C6orf62 GMNN GMNN Intergenic
0.46 0.47 0.44 0.17 0.06 0.27 0.25 0.21 0.24 0.17 0.37 0.44 0.34 0.31 0.3 0.33 0.39 0.47 0.49 0.3 0.42 0.11 0.37 0.13 0.42 0.35 0.29 0.36 0.09 0.44 0.48
PD and WR, and −3.5 SD for DISC. The selected sample for each of the phenotypes consisted of the following numbers of sib-pairs: OC, 153; PA, 111; WR, 110; PD, 84; DISC, 106. Multipoint analysis was carried out by using the non-parametric lod (NPL) statistic in GH and both DeFries–Fulker (DF) basic (DeFries and Fulker 1985) and revised H-E methods implemented in the SAS macro, QMS2 (Lessem and Cherny 2001). Double entry of probands was performed for the DF model. Results are reported as zscores for GH analyses and P-values for QMS2 tests. The significance criterion for linkage was P≤0.05 and z≥1.5.
SNP genotyping and association A subset of families was selected for association studies based on at least one child having a phenotypic score at least 2 SD below the population mean in at least one phenotype. The sample included 114 families of which 19 had one offspring, 73 had two offspring, 15 had three offspring, and seven had four offspring. SNPs were selected from the ABI Assays-on-Demand website, which offers optimized TaqMan assays for validated SNPs. SNPs were chosen based on heterozygosity, marker spacing, and a preferential location within candidate genes. Table 2 lists the 31 SNPs that were genotyped. The selected SNPs span ten genes that cluster within ~680 kb of our linkage interval, giving an average marker density of ~21 kb. SNPs were genotyped by TaqMan assay with Fam and Vic labeled MGB probes. Polymerase chain reactions were set up according to the
manufacturer’s protocol and thermocycled in T-Gradient thermocyclers (Biometra). Endpoint fluorescent readings and allele calling were carried out on an ABI 7000 sequence detection system (SDS) with 7000SDS software (Applied Biosystems). Inheritance and error checking of allele calls were carried out as above with GAS and Merlin, respectively. Marker-trait associations were performed with parent-proband trios by using the family-based association test (FBAT; Horvath et al. 2001) and QTDT (Abecasis et al. 2000). For FBAT analysis, both the dominant and additive models were used. Empirical variances were computed to correct for correlation of sibling genotypes attributable to the presence of linkage. QTDT analysis was performed by using the orthogonal (Abecasis et al. 2000), Allison (1997), and Rabinowitz (1997) models. The Rabinowitz (1997) model provides a robust T-test for association but does not allow dominance or variance components modeling. The orthogonal and Allison (1997) tests were carried out in a variance components framework modeling environmental, additive, and polygenic effects. Empirical significance levels are reported for the QTDT tests and were computed from 1,000 Monte Carlo permutations. Bonferroni correction for multiple testing is thought to be too conservative for associated alleles at linked markers, resulting in a loss of power (McIntyre et al. 2000). Computation of empirical P-values should reduce associations attributable to Type I error.
132 Table 3 Significant single-point linkage values. Empirical P-values are reported for markers showing significant linkage. The significance criterion for linkage is P≤0.05. Single-point linkage was computed by using the revised H-E algorithm implemented in S.A.G.E. 4.2 (S.A.G.E. 2002). Definition of probands was as follows: −2 SD, OC; −2.5 SD, PA; −3 SD, WR and PD; −3.5 SD, DISC Marker
Position (Mb) P-value DISC
D6S1597 D6S1588 D6S1663 D6S461 D6S276 D6S1554 D6S1571 D6S105 D6S306 D6S258 D6S1683 MOG D6S273
5.35 5.73 6.16 7.24 7.82 8.56 8.71 11.45 11.48 12.61 12.66 13.33 15.41
OC
PA
PD
WR
0.0250 0.047 0.0142 0.046 0.0089 0.0403 0.048 0.0071 0.0317 0.025 0.0079 0.0230 0.01 0.039 0.0504 0.0051 0.011 0.042 0.0422 0.031 0.034 0.036 0.039 0.024 0.046
Intermarker LD and haplotype analysis Intermarker LD was evaluated by using the graphical output of LD (GOLD) application of Abecasis and Cookson (2000). Founder haplotypes and recombinations were generated by Simwalk2 software. Genotypic data from any available siblings were included with the parent-proband trios to increase the information used in the haplotype calculations. GOLD uses output from Simwalk2 to calculate several disequilibrium statistics, including Lewontin’s disequilibrium coefficient (D′). D′ values were summarized graphically for all marker pairings across the 31 SNP markers. Haplotypes were analyzed for association with the phenotypic data by using the GH TDT option in which two, three, or four adjacent marker haplotypes can be examined by the TDT for association with the trait. Analysis was carried out in a four-marker sliding window across the 31 SNPs. TDT significance was determined by 1,000 permutations with the perm1 option in GH. GH TDT does not control for the presence of linkage in the region; thus, four-marker haplotypes showing association were subsequently analyzed by FBAT. Empirical variances were computed in FBAT, which provides a pure test of association, ruling out associations detected by GH attributable to linkage. FBAT analysis was carried out on markers within each gene in isolation, emphasizing SNPs showing single-marker association.
Results Linkage analysis Single-point analysis of our entire sample failed to detect linkage with any of the phenotypes. In accordance with previous findings (Cardon et al. 1994; Gayán et al. 1999) analysis of samples selected for severity detected significant linkage of 6p markers with all five phenotypes. The markers for which significant P-values were obtained are summarized in Table 3. Defining probands as ≤2 SD below the mean, the OC phenotype showed a broad region
Fig. 1A, B GeneHunter multipoint analysis. Results of multipoint analysis with the GeneHunter NPL statistic are plotted for each phenotype as the z-score versus the relative position (cM). Results shown are from an analysis of the selected samples (−2 SD, OC; −2.5 SD, PA; −3 SD, WR and PD; −3.5 SD, DISC)
of linkage spanning ~10 Mb from D6S1588 to D6S273. Significant linkage with WR spanned the distal end of the OC interval between markers D6S1597 and D6S1571 with a proband definition of ≤−3 SD. Linkage with DISC, PA, and PD was significant for markers within the WR interval. These results were obtained by using proband definitions of ≤−3.5 SD for DISC, ≤−3 SD for PD, and ≤−2.5 SD for PA. Maximal linkage with all five phenotypes occurred at marker D6S1554. The WR phenotype showed the most significant evidence of linkage at D6S1554 (P=0.005). Multipoint results obtained with the GH NPL statistic are depicted graphically in Fig. 1. Figure 1A shows a significant overlap of linkage with the PA, PD, OC, and WR phenotypes. Significant linkage spanned markers D6S461 to D6S1554 with these four phenotypes. A convergent peak of maximal linkage was detected between markers D6S276 and D6S1554 (zmax=2.055, PA; zmax=1.91, PD; zmax=1.569, WR; zmax=1.75, OC). Figure 1B shows the GH results obtained for DISC, demonstrating a broad region of linkage spanning D6S1597 to D6S1571. Maximal linkage with DISC was between markers D6S1597 and D6S1588 (z=2.073), although significance remained comparable across the entire interval, resulting in a flat peak of linkage. A less significant peak of linkage proximal to this interval was also detected with all phenotypes except OC near marker D6S1666. Maximal linkage in this region converged between markers D6S1666 and D6S1568 (z=1.347, DISC; z=1.662, PA; z=1.662, PD; z=1.177, WR).
133 Table 4 Relative position (cM) of maximal linkage for each phenotype. Multipoint results computed by using the revised Haseman–Elston (NHE) and DF statistics in QMS2 Method Phenotype Position (cM) P-value Region DF
NHE
OC PA PD WR DISC OC PA PD WR DISC
13.417 12.845 11.745 10.258 12.845 12.845 12.309 12.684 10.058
0.0004 0.0141 0.0166 NS 0.0319 0.0011 0.0343 0.0334 0.0452 0.0134
D6S1554–D6S1571 D6S276–D6S1554 D6S461–D6S276 D6S1588–D6S1663 D6S276–D6S1554 D6S276–D6S1554 D6S276–D6S1554 D6S276–D6S1554 D6S1588–D6S1663
Multipoint results were also computed by using the revised Haseman–Elston (NHE) and DF statistics in QMS2. Table 4 summarizes the relative position (cM) of maximal linkage for each phenotype under these two methods. Significant linkage across all phenotypes, Table 5 Association analysis. The orthogonal and Allison models in QTDT were analyzed in a variance components framework that modeled environmental, polygenic, and additive effects. One thousand Monte Carlo permutations were computed generating empiric Pvalues
Phenotype SNP no. Gene
AOa (P) AAb (P) ARc (P) FBAT domd (P) FBAT adde (P)
OC
0.0251 0.0376 0.0343 0.012
0.004 0.0226 0.0129 0.0195
0.035
0.0345
0.0486
0.0345
PA DISC
PD
a AO, orthogonal model in QTDT b AA, Allison model in QTDT c AR, Rabinowitz model in QTDT d FBAT analysis modeling dominance e FBAT analysis modeling additive gene effects
defined as P≤0.05, spanned the interval D6S1597 to D6S1571 under both methods. Results from the DF analysis were slightly more variable in defining the position of maximal linkage across phenotypes, in comparison with the NHE method. WR failed to show linkage by DF analysis, yet showed suggestive evidence (P=0.045) at 12.684 cM under NHE analysis. Maximal linkage for the OC, PA, and PD phenotypes under the DF method spanned markers D6S461 to D6S1571, as in the GH analysis. A narrow interval between markers D6S276 and D6S1554 (12.309–12.684 cM) showed maximal linkage with OC, PA, PD, and WR under the NHE method. Linkage with DISC was comparable across all three multipoint methods, with maximal linkage between markers D6S1588 and D6S1663 for both the NHE and DF methods. All three methods also detected a proximal region of linkage between markers D6S1666 and D6S439. The H-E method showed linkage of PD and OC in this region (P=0.0234 at 17.54 cM, PD; P=0.017 at 18.69 cM, OC). Similarly, the DF method detected evidence of linkage for PD, OC, and PA near marker D6S1666 (P=0.0345 at 17.221 cM, PD; P=0.002719 at 16.904 cM,
WR
1 3 4 6 7 9 10 13 25 1 3 6 11 13 23 1 4 6 7 11 12 13 25 26 1 3 7 10 12 13 25
VMP VMP DCDC2 DCDC2 DCDC2 DCDC2 DCDC2 DCDC2 TTRAP VMP VMP DCDC2 DCDC2 DCDC2 KIAA0319 VMP DCDC2 DCDC2 DCDC2 DCDC2 DCDC2 DCDC2 TTRAP THEM2 VMP VMP DCDC2 DCDC2 DCDC2 DCDC2 TTRAP
0.0371 0.0529 0.0246 0.0377
0.0233 0.0421
0.051
0.0168 0.0406 0.0496
0.055 0.0081 0.0081
0.0449 0.0041
0.0034
0.041
0.0025
0.023
0.029 0.008
0.0122 0.0167
0.025 0.01 0.033 0.036
0.0278 0.0455 0.0298 0.0383 0.0172 0.039 0.0058 0.0143 0.0351
0.0267 0.045
0.0427
0.001 0.002
0.0344 0.0078 0.0191
0.035 0.0281 0.0249 0.0298
0.0379 0.0247 0.0209
134 Fig. 2 SNP marker associations. The most significant Pvalue is plotted for each marker showing association with one of the phenotypes. The significance criterion is P≤0.05
THEM2. Association with markers in the VMP and DCDC2 genes was detected with all five phenotypes. Intermarker LD and haplotype analysis
Fig. 3 GOLD analysis of SNP markers. Graphical depiction of intermarker LD output from GOLD (Abecasis and Cookson 2000). Lewontin’s disequilibrium coefficient (D′) is plotted for all marker pairs
OC; P = 0.03696 at 16.27 cM, PA). Both single-point and multi-point analyses showed significant linkage with all five phenotypes over the interval D6S1597 to D6S1571, with maximal linkage converging between markers D6S276 and D6S1554. Association analysis Results from the single-marker association analyses are summarized in Table 5 and Fig. 2. Table 5 lists the markers for which a significant association was detected (P≤0.05). Significant P-values are listed under each analysis model used. Figure 2 depicts the markers showing association with at least one phenotype; only the most significant result is plotted. Thirteen of the 31 SNPs showed significant association with at least one of the phenotypes. These 13 SNPs cluster within five genes: two in VMP, eight in DCDC2, and one each in KIAA0319, TTRAP, and
Output from the GOLD package graphically depicts the intermarker LD in our sample. Figure 3 shows pair-wise D′ values plotted for the 31 SNPs. Significant intermarker LD was detected across the ~680 kb interval. Intermarker LD was low at SNPs 5, 13–14, 22–23, and 25–26. The remaining marker pairs had D′ values greater than 0.4, with 70% of pairs being greater than 0.7. Results of haplotype analysis by using FBAT are presented in Table 6. Haplotypes showing significant over-transmission to affected probands were detected across phenotypes. Probands were defined as having a phenotype score ≤2 SD below the mean. Associated haplotypes were detected in the five genes showing singlemarker association: VMP, DCDC2, KIAA0319, TTRAP, and THEM2. Both DISC and WR also showed association with a three-marker haplotype spanning the ALDH5A1 gene. In corroboration with single-marker associations, results from the haplotype analysis strongly implicated two clusters of five genes in our interval as being strong candidates for mutation screening.
Discussion The primary goal of this study was to fine-map the 6p21.3 QTL for RD through association analysis by using a high density SNP map. We first refined the QTL boundaries through linkage analysis of a large cohort of twin pairs. This study is an extension of that of Gayán et al. (1999) who used 126 sib-pairs analyzed on eight markers. We included these original sib-pairs in our 708 total pairs genotyped for 22 STR markers and utilized the same component phenotypes. Our linkage results place the QTL in a 3.24-Mb (5.33 cM) interval between markers D6S1597 and D6S1571, with a peak of linkage between markers D6S276 and D6S1554. These results are consistent with those of Kaplan et al. (2002) who have
135 Table 6 Haplotype TDT analysis in FBAT. Haplotypes are listed with their corresponding SNP marker numbers and the gene(s) spanned by those SNPs (Phenotype phenotype showing association or over-transmission to probands for a particular haplotype)
Markers
Haplotype
Gene(s)
P
Phenotype
1–5 21–25 3–6 6–9 23–26 1–3 9–12 1–3 7/12/13 18–20 21–26 1–6 18–20 23/25
12111 12212 1211 2121 2112 121 2211 121 212 222 122122 121112 222 21
VMP/DCDC2 KIAA0319/TTRAP VMP/DCDC2 DCDC2 KIAA0319/TTRAP/THEM2 VMP DCDC2 VMP DCDC2 ALDH5A1 KIAA0319/TTRAP/THEM2 VMP/DCDC2 ALDH5A1 KIAA0319/TTRAP
0.0438 0.0206 0.0413 0.0105 0.0452 0.0316 0.0431 0.0107 0.0105 0.0323 0.0236 0.0176 0.0326 0.0327
OC OC PA PA PA PD PD WR WR WR WR DISC DISC DISC
reported a peak of linkage at marker D6S461, and those of Fisher et al. (1999) and Gayán et al. (1999) who have detected maximal linkage between D6S276 and D6S105. D6S461 was the most telomeric marker used by Gayán et al. (1999). By extending our marker panel, we were able to resolve a telomeric boundary for our linkage interval. Both orthographic and phonologic skills are influenced by the 6p QTL; however, linkage has only been detected in the sample selected for severe scores and not in the entire sample. Whereas increased sample size should significantly increase the power to detect linkage, the linkage on 6p21 appears to be specific to the low tail of the distribution. This suggests a QTL specific for severe RD rather than influencing normal reading variation. Recent behavioral genetic analyses by N. Harlaar, F.M. Spinath, P. S. Dale, and R. Plomin (in preparation) have shown similar levels of heritability for the group deficit in WR efficiency (hg2=0.60 below the tenth percentile) and individual differences across their entire unselected sample of 3,496 twin pairs (hg2=0.73) and consider that their results imply that the same genetic mechanisms may influence group deficits and individual differences across the normal range. In contrast, our results indicate that for the 6p21 locus the genetic mechanisms may differ, as only the selected sample shows linkage. These results suggest that RD is a distinct quantitative trait, whereby some loci contribute specifically to RD and not to normal reading variation. Some of the genetic variation in RD will necessarily be contributed by genes that also play a role in normal reading variation leading to a bimodal distribution, or an overlap, between these two traits. For example, the 1p36 locus shows significant linkage in our entire unselected sample, yet no linkage in the selected samples, possibly indicating a locus influencing normal reading variation (unpublished data). These data underscore the utility of analyzing both selected and unselected samples in RD QTL mapping studies. When appropriately selected for severity, all five quantitative phenotypes show linkage to the 6p QTL. Initial analyses defined probands as scoring in the lowest
20th percentile corresponding to the following z-scores: −2 SD for OC, −2.4 SD for PA, and −2.6 SD for DISC, PD, and WR. By using this selection criterion, linkage was only significant with the OC and PA phenotypes. The other phenotypes only showed significant linkage upon further selection (−3 SD for WR and PD, and −3.5 SD for DISC). This variability could explain the different findings for these phenotypes across studies. Gayán et al. (1999) have failed to find linkage with DISC and WR in a subset of our sample; however, their lowest selection criterion was −2.5 SD. The inability to detect linkage by using a 20th percentile selection criterion indicates that a significant number of probands in the low tail of the DISC and WR distributions are unlinked because of locus heterogeneity or phenocopy. These more general assessments of reading ability may not target RD as specifically as other component phenotypes, selecting for reading deficits not specific to RD. Linkage analysis also revealed the presence of a second peak of linkage proximal to our primary interval between markers D6S1666 and D6S439. Single-point analysis of the OC phenotype shows a broad region of significance spanning markers D6S1597–D6S273 (~10 Mb) and extending into this centromeric region. Single-point analysis was not significant for other phenotypes in this region, despite significant findings by multipoint analysis. Grigorenko et al. (2000) report an interval that lies just centromeric to ours, between markers D6S105 and D6S273. This peak might represent a second locus within the 6p QTL. Alternatively, it could represent a “shadow peak” resulting as an artifact of multipoint analysis in which marker information from our primary peak of linkage influences the proximal markers. A possible contributory factor to this could be the marked suppression of recombination occurring at marker D6S105 and extending ~5 Mb centromeric, as reported by Malfroy et al. (1997) and detected in the deCODE data (see Table 1). The centromeric boundary of our linkage interval occurs adjacent to this recombination suppression such that increased allele sharing could be extended across this
136
region because of low recombination. If this were the case, it would increase the limits of resolution for linkage in this region and could explain the difficulties in defining a precise location for the QTL within the 17 Mb defined by various groups. Twelve genes lie within our primary linkage interval. Two of the genes, prolactin (PRL) and PWWP1 representing the hepatoma-derived growth-factor-like gene, are found within ~300 kb of the telomeric end of this interval. The remaining ten genes cluster within ~680 kb of the proximal end of the interval, between markers D6S276 and D6S1554, where maximal linkage converges across phenotypes. Based on the putative function and expression of the former two genes and their separation from the remaining ten genes by ~1.6 Mb, the latter cluster of ten genes was selected for the focus of our association study. All ten genes are expressed in the central nervous system (CNS) and have functions plausible for a causal role in RD. Kaplan et al. (2002) recently reported an association study of the 6p QTL. Their study utilized microsatellite markers spanning a region of ~9 Mb. Positive associations were reported for 11 markers spanning ~8.5 Mb. With an average marker spacing of ~300 kb and little to no intermarker LD, there is a possible significant Type I error with multiallelic markers. This is the first report of an association study for the 6p21.3 RD QTL to date that utilizes a high density SNP marker map. Our markers lie within genes or are sufficiently close to genes to be able to detect marker-trait associations attributable to LD. Interestingly, Kaplan et al. (2002) report their most significant association at marker JA04, which lies in the KIAA0319 gene. This gene falls within our peak of linkage and is included in our SNP interval. Marker-trait analysis detected associations with markers in two clusters of five genes: VMP, DCDC2, KIAA0319, TTRAP, and THEM2 (Table 5). VMP and DCDC2 are adjacent to one another, separated by ~27 kb. Similarly, KIAA0319, TTRAP, and THEM2 are adjacent to one another, separated by only 3.8 kb and 180 bp, respectively. Because of the small distances separating these genes, the association within one gene probably influences the detection of association in flanking genes, because of LD between markers across these genes. GOLD analysis indicated strong intermarker LD across genes in this region, making it difficult to resolve an association signal within this cluster of closely spaced genes. Markers in VMP showed association with all five phenotypes, and several markers in DCDC2 exhibited association across phenotypes, except for PA. Associations were also detected in the KIAA0319, TTRAP, and THEM2 genes, with one marker in KIAA0319 showing association with DISC, one marker in TTRAP showing association with OC, PD and WR, and one marker in THEM2 showing association with PD. Only one SNP was genotyped for TTRAP and two markers for THEM2. Further, two of the four markers spanning KIAA0319 had low heterozygosities and were relatively uninformative. Given the potential associations spanning these genes, additional markers
should be genotyped in this region to resolve the association signal further. Intermarker LD was assessed in our SNPs independently of phenotypic data. Strong intermarker LD was detected in adjacent marker pairs across our entire SNP map. The level of LD between markers 1–2, 13–14, and 25–26 was low, although these markers are in intergenic regions and so could be sites of recombination. LD was absent at markers 5 and 24, either because of low heterozygosities and lack of informativeness or because of undetectable genotyping error. D′ values for the majority of marker pairs were ≥0.7. Given the strong intermarker LD, haplotypic association was also examined in the region (Table 6). FBAT analysis detected haplotypes that were over-transmitted to probands across all phenotypes. Corroborating single-marker results, haplotypes within the VMP and DCDC2 genes showed association with all phenotypes. Haplotypes spanning the KIAA0319, TTRAP, and THEM2 genes were associated with all phenotypes except PD, which showed single-marker association with TTRAP and THEM2. DISC and WR also showed association with a common three-marker haplotype in the ALDH5A1 gene, probably influenced by LD with markers in the adjacent KIAA0319 gene. Across the region, haplotype analysis corroborated our single-marker results showing robust haplotype transmission disequilibrium with markers in the VMP, DCDC2, KIAA0319, TTRAP, and THEM2 genes. All five of these genes are expressed in the CNS and would be prime candidates for mutation screening. VMP is a neuronspecific vesicular membrane protein thought to play a role in vesicular organelle transport and neurotransmission (Cheng et al. 2002). DCDC2 is expressed ubiquitously and contains a doublecortin homology domain. Doublecortin plays a role in cortical neuron migration in embryonic development. The KIAA0319 gene is highly expressed in brain and codes for a novel protein of unknown function. This gene has a PKD homology domain with an Ig-like fold indicating a possible role in cell–matrix or cell–cell adhesion. TTRAP encodes a tumor necrosis factor receptor-associated protein known to inhibit activation of nuclear factor-kappa B (NF-κB) and subsequent downstream activation of transcription (Pype et al. 2000). Activation of NF-κB transcription has been shown to play a role in long-term potentiation and synaptic plasticity associated with learning and memory. THEM2 encodes a protein belonging to a thioesterase superfamily that catalyzes the hydrolysis of long-chain fatty acyl-CoA thioesters. Abnormal fatty acid metabolism has been suggested to play a role in a spectrum of neurodevelopmental disorders, including dyslexia. In conclusion, our linkage findings indicate a QTL for RD over an interval of ~3.24 Mb spanning markers D6S1597 to D6S1571, overlapping significantly with the interval reported by Kaplan et al. (2002) centered at JA04. Maximal linkage converged between markers D6S276 and D6S1554, flanking the ten genes examined for association. Although our study contained a much larger number of sib-pairs, we failed to refine the locus further through
137
linkage. Linkage to the 6p locus was only detected with a sample selected for severity. Thus, this QTL appeared to influence the low tail of the distribution, and the addition of samples that were less severely affected did not increase the power to detect linkage. Association analysis effectively narrowed our candidate gene interval for the 6p21.3 QTL, aiding in the prioritization of genes for mutation screening. Single marker and haplotype associations were detected with markers in five genes: VMP, DCDC2, KIAA0319, TTRAP, and THEM2, all of which are expressed in the CNS. Associations in VMP and DCDC2 were robust across analysis platforms and phenotypes; mutation screening of these two genes will be the focus of future studies. Associations with markers in the KIAA0319, TTRAP, and THEM2 genes were also compelling, and additional SNPs should be genotyped within these genes to further resolve the association signal. Acknowledgments We are grateful to all of the families who participated in this study, which was supported in part by National Institutes of Health (NIH) NICHD grant 5P50 HD27802–12 and NIH grant 5-R01 HD34812. Some of the results presented here were obtained by use of the program package S.A.G.E. from the Department of Epidemiology and Biostatistics, Case Western Reserve University, Cleveland, which is supported by US Public Health Service resource grant 1 P41 RR03655 from the National Center for Research Resources.
References Abecasis GR, Cookson WO (2000) GOLD—graphical overview of linkage disequilibrium. Bioinformatics 16:182–183 Abecasis GR, Cardon LR, Cookson WOC (2000) A general test of association for quantitative traits in nuclear families. Am J Hum Genet 66:279–292 Abecasis GR, Cherny SS, Cookson WO, Cardon LR (2002) Merlin —rapid analysis of dense genetic maps using sparse gene flow trees. Nat Genet 30:97–101 Allison DB (1997) Transmission-disequilibrium tests for quantitative traits. Am J Hum Genet 60:676–690 Cardon LR, Smith SD, Fulker DW, Kimberling WJ, Pennington BF, DeFries JC (1994) Quantitative trait locus for reading disability on chromosome 6. Science 266:276–279 Cheng C, Xu J, Ye X, Dai J, Wu Q, Zeng L, Wang L, Zhao W, Ji C, Gu S, Xie Y, Mao Y (2002) Cloning, expression and characterization of a novel human VMP gene. Mol Biol Rep 29:281–286 DeFries JC, Fulker DW (1985) Multiple regression analysis of twin data. Behav Genet 15:467–473 Dunn LM, Markwardt FC (1970) Peabody individual achievement test examiner’s manual. American Guidance Service, Circle Pines Elston RC, Buxbaum S, Jacobs KB, Olson JM (2000) Haseman and Elston revisited. Genet Epidemiol 19:1–17 Fagerheim T, Raeymaekers P, Tonnessen FE, Pedersen M, Tranebjaerg L, Lubs HA (1999) A new gene (DYX3) for dyslexia is located on chromosome 2. J Med Genet 36:664–669 Field LL, Kaplan BJ (1998) Absence of linkage of phonological coding dyslexia to chromosome 6p23–p21.3 in a large family data set. Am J Hum Genet 63:1448–1456 Fisher SE, Marlow A, Lamb J, Maestrini E, Williams DF, Richardson AJ, Weeks DE, Stein JF, Monaco AP (1999) A quantitative-trait locus on chromosome 6p influences different aspects of developmental dyslexia. Am J Hum Genet 64:146– 156
Fisher SE, Francks C, Marlow AJ, MacPhie IL, Newbury DF, Cardon LR, Ishikawa-Brush Y, Richardson AJ, Talcott JB, Gayán J, Olson RK, Pennington BF, Smith SD, DeFries JC, Stein JF, Monaco AP (2002) Independent genome-wide scans identify a chromosome 18 quantitative-trait locus influencing dyslexia. Nat Genet 30:86–91 Fulker D, Cardon L, DeFries J, Kimberling W, Pennington B, Smith S (1991) Multiple regression analysis of sib-pair data on reading to detect quantitative trait loci. Reading Writing Interdis J 3:299–313 Gayán J, Olson RK (2001) Genetic and environmental influences on orthographic and phonological skills in children with reading disabilities. Dev Neuropsychol 20:483–507 Gayán J, Smith SD, Cherny SS, Cardon LR, Fulker DW, Brower AM, Olson RK, Pennington BF, DeFries JC (1999) Quantitative-trait locus for specific language and reading deficits on chromosome 6p. Am J Hum Genet 64:157–164 Grigorenko EL, Wood FB, Meyer MS, Hart LA, Speed WC, Shuster A, Pauls DL (1997) Susceptibility loci for distinct components of developmental dyslexia on chromosomes 6 and 15. Am J Hum Genet 60:27–39 Grigorenko EL, Wood FB, Meyer MS, Pauls DL (2000) Chromosome 6p influences on different dyslexia-related cognitive processes: further confirmation. Am J Hum Genet 66:715–723 Grigorenko EL, Wood FB, Golovyan L, Meyer M, Romano C, Pauls D (2003) Continuing the search for dyslexia genes on 6p. Am J Med Genet 118B:89–98 Habib M (2000) The neurological basis of developmental dyslexia: an overview and working hypothesis. Brain 123:2373–2399 Haseman JK, Elston RC (1972) The investigation of linkage between a quantitative trait and a marker locus. Behav Genet 2:3–19 Horvath S, Xu X, Laird N (2001) The family based association test method: strategies for studying general genotype-phenotype associations. Euro J Hum Gen 9:301–306 Kaplan DE, Gayán J, Ahn J, Won T-W, Pauls D, Olson RK, DeFries JC, Wood F, Pennington BF, Page GP, Smith SD, Gruen JR (2002) Evidence for linkage and association with reading disability, on 6p21.3–22. Am J Hum Genet 70:1287–1298 Kruglyak L, Lander ES (1995) Complete multipoint sib-pair analysis of qualitative and quantitative traits. Am J Hum Genet 57:439–454 Lerner JW (1989) Educational interventions in learning disabilities. J Am Acad Child Adolesc Psychiatry 28:326–331 Lessem JL, Cherny SS (2001) DeFries–Fulker multiple regression analysis of sibship QTL data: a SAS macro. Bioinformatics 17:371–372 Lyon GR (1995) Toward a definition of dyslexia. Ann Dyslexia 45:3–27 Malfroy L, Roth MP, Carrington M, Borot N, Volz A, Ziegler A, Coppin H (1997) Heterogeneity in rates of recombination in the 6-Mb region telomeric to the human major histocompatibility complex. Genomics 43:226–231 McIntyre LM, Martin ER, Simonsen KL, Kaplan NL (2000) Circumventing multiple testing: a multilocus Monte Carlo approach to testing for association. Genet Epidemiol 19:18–29 Morris DW, Robinson L, Turic D, Duke M, Webb V, Milham C, Hopkin E, Pound K, Fernando S, Easton M, Hamshere M, Williams N, McGuffin P, Stevenson J, Krawczak M, Owen MJ, O’Donovan MC, Williams J (2000) Family-based association mapping provides evidence for a gene for reading disability on chromosome 15q. Hum Mol Genet 9:843–848 Nopola-Hemmi J, Myllyluoma B, Haltia T, Taipale M, Ollikainen V, Ahonen T, Voutilainen A, Kere J, Widen E (2001) A dominant gene for developmental dyslexia on chromosome 3. J Med Genet 38:658–664 Nöthen MM, Schulte-Körne G, Grimm T, Cichon S, Vogt IR, Müller-Myhsok B, Propping P, Remschmidt H (1999) Genetic linkage analysis with dyslexia: evidence for linkage of spelling disability to chromosome 15. Eur Child Adolesc Psychiatry 8 (Suppl 3):56–59
138 Olson R, Wise B, Conners F, Rack J, Fulker D (1989) Specific deficits in component reading and language skills: Genetic and environmental influences. J Learn Disabil 22:339–348 Petryshen TL, Kaplan BJ, Liu MF, Field LL (2000) Absence of significant linkage between phonological coding dyslexia and chromosome 6p23–21.3, as determined by use of quantitativetrait methods: confirmation of qualitative analyses. Am J Hum Genet 66:708–714 Petryshen TL, Kaplan BJ, Lui MF, French NS de, Tobias R, Hughes ML, Field LL (2001) Evidence for a susceptibility locus on chromosome 6q influencing phonological coding dyslexia. Am J Med Genet 105:507–517 Pype S, Declercq W, Ibrahimi A, Michiels C, Van Rietschoten JG, Dewu N, Boer M de, Vandenabeele P, Huylebroeck D, Remacle JE (2000) TTRAP, a novel protein that associates with CD40, tumor necrosis factor (TNF) receptor-75 and TNF receptorassociated factors (TRAFs), and that inhibits nuclear factorkappa B activation. J Biol Chem 275:18586–18593 Rabin M, Wen XL, Hepburn M, Lubs HA, Feldman E, Duara R (1993) Suggestive linkage of developmental dyslexia to chromosome 1p34–p36. Lancet 342:178
Rabinowitz D (1997) A transmission disequilibrium test for quantitative trait loci. Hum Hered 47:342–350 Risch N, Teng J (1998) The relative power of family-based and case-control designs for linkage linkage disequilibrium studies of complex human diseases I. DNA pooling. Gen Res 8:1273– 1288 S.A.G.E. (2002) Statistical analysis for genetic epidemiology. Computer program package available from Statistical Solutions, Cork Shaywitz SE, Shaywitz BA, Fletcher JM, Escobar MD (1990) Prevalence of reading disability in boys and girls. JAMA 264:998–1002 Spielman RS, McGinnis RE, Ewens WJ (1993) Transmission test for linkage disequilibrium: the insulin gene region and insulindependent diabetes mellitus (IDDM). Am J Hum Genet 52:506–516 Wechsler D (1974) Wechsler’s intelligence scale for children, revised examiner’s manual. The Psychological Corporation, New York Wechsler D (1981) Wechsler adult intelligence scale, revised examiner’s manual. The Psychological Corporation, New York
