at SciVerse ScienceDirect

Journal of Psychiatric Research 46 (2012) 271e278

Contents lists available

Journal of Psychiatric Research

journal homepage: www.elsevier .com/locate/psychires

Genome-wide study identifies PTPRO and WDR72 and FOXQ1-SUMO1P1interaction associated with neurocognitive function

Marissa LeBlanc a,b,*, Bettina Kulle a,b, Kjetil Sundet c,d, Ingrid Agartz d,e, Ingrid Melle d,f, Srdjan Djurovic d,Arnoldo Frigessi b,g, Ole A. Andreassen d,f

a Epi-Gen, Institute of Clinical Medicine, Akershus University Hospital, University of OslobDepartment of Biostatistics, Institute of Basic Medical Sciences, University of Oslo, Oslo, NorwaycDepartment of Psychology, University of Oslo, Oslo, Norwayd Psychosis Research Unit, Division of Mental Health and Addiction, Oslo University Hospital, Oslo, NorwayeDepartment of Research and Development, Diakonhjemmet Hospital, Oslo, NorwayfDivision of Mental Health and Addiction, Institute of Clinical Medicine, University of Oslo, Oslo, Norwayg (sfi)2 Statistics for Innovation, Oslo, Norway

a r t i c l e i n f o

Article history:Received 9 June 2011Received in revised form1 October 2011Accepted 2 November 2011

Keywords:NeurocognitionGWASProtein tyrosine phosphataseReceptor typeO (PTPRO)WD repeat domain 72 (WDR72)Forkhead box Q1 (FOXQ1)SUMO1 pseudogene 1 (SUMO1P1)

* Corresponding author. Institute of Basic MedicBiostatistics, University of Oslo, PO Box 1122 Blindern,22851402; fax: þ47 22851313.

E-mail address: marissa.leblanc@medisin.uio.no (M

0022-3956/$ e see front matter � 2011 Elsevier Ltd.doi:10.1016/j.jpsychires.2011.11.001

a b s t r a c t

Background: Several aspects of neurocognitive function have high heritability, but the molecular geneticmechanisms underlying neurocognition are not known. We performed a genome-wide association study(GWAS) to identify genes associated with neurocognition.Methods: 700 Subjects (schizophrenia spectrum disorder, n¼ 190, bipolar disorder n¼ 157 and healthyindividuals n¼ 353) were tested with an extensive neuropsychological test battery, and genotyped usingthe Affymetrix Genome-Wide Human SNP Array 6.0. After quality control, linear regression analysis ofeach of the 24 cognitive tests on the SNP dosage was performed, including age, gender, education anddisease group as covariates. Additionally, 9 SNPs trending toward genome-wide significance wereconsidered for epistatic interactions.Results: Four SNPs and 2 independent association signals achieving genome-wide significance wereidentified. Three intronic SNPs in PTPRO were associated with learning and memory (CVLT-II LDFR)(rs17222089, p¼ 1.55� 10�8; rs11056571, p¼ 1.68� 10�8; and rs2300290, p¼ 1.09� 10�8). rs719714downstream of WDR72 was associated with executive functioning (CW-3: Inhibition, D-KEFS)(p¼ 4.32� 10�8). A highly significant epistatic interaction was found between rs9378605 upstream ofFOXQ1 and rs11699311 downstream of SUMO1P1 for the Grooved Pegboard test (p¼ 7.6� 10�14).Conclusions: We identified four novel loci associated with neurocognitive function and one novelepistatic interaction. The findings should be replicated in independent samples, but indicate a role ofPTPRO in learning and memory, WDR72 with executive functioning, and an interaction between FOXQ1and SUMO1P1 for psychomotor speed.

� 2011 Elsevier Ltd. All rights reserved.

1. Introduction

Neurocognitive function is essential for most daily life activities,and seems to play a role in evolution of humans (Balter, 2010).Neurocognitive dysfunction is seen in a range of brain disorders,from neurodevelopmental to neurodegenerative disorders (Boraet al., 2009; Lott and Dierssen, 2010; Geschwind, 2009). Cognitiveimpairment, particularly in learning and memory and executive

al Research, Department of0317 Oslo, Norway. Tel.: þ47

. LeBlanc).

All rights reserved.

functioning, has been associated with severe mental disorders,especially schizophrenia but also bipolar disorders (Simonsen et al.,2009; Bora et al., 2009).

Several lines of evidence indicate that neurocognitive functionsare heritable (McClearn et al., 1997; Finkel et al., 1995; Bouchardet al., 1990). A new study has estimated the heritability of intelli-gence to w40e51% (Davies et al., 2011), and for other neuro-cognitive domains there are indications of same range ofheritability (Green et al., 2008). Despite the clear evidence forgenetic factors, the molecular mechanisms underlying cognitivefunctions are not well understood.

Investigations of candidate genes implicate the involvement ofCOMT, BDNF and genes related to severe mental disorders in

M. LeBlanc et al. / Journal of Psychiatric Research 46 (2012) 271e278272

cognitive function (Egan et al., 2001, 2003; Porteous et al., 2006).Animal studies have suggested genetic effects on learning andmemory (Arguello and Gogos, 2010; Crawley, 2008). Recentevidence has indicated overlapping genetic mechanisms in severemental disorders and cognitive functioning (Burdick et al., 2010;Toulopoulou et al., 2007). Given this emerging but incompletepicture of the relationship between genetics, disease and neuro-cognition, it is reasonable to further investigate thegenetic effects onneurocognition in clinical samples, particularly using a hypothesis-free genome-wide study approach. This approach has provenuseful for identification of genetic factors of cognition(Papassotiropoulos et al., 2006; Butcher et al., 2008; Bates et al.,2009) although the findings do not always replicate (Need et al.,2008).

Recent evidence suggests possible epistatic interactionsbetween susceptibility genes for neuropsychiatric disorders. AnfMRI study found an epistatic interaction between COMT and GRM3on the engagement of the prefrontal cortex during workingmemory (Tan et al., 2007). DISC1, CIT and NDEL1 have also beenimplicated in epistatic interactions for both susceptibility toschizophrenia and prefrontal cortex inefficiency (Nicodemus et al.,2010). Thus, investigation of epistatic interactions of genes associ-ated with neurocognitive domains is sensible.

The aim of the current study was to identify genetic variantsassociated with tests in neurocognitive domains. We used genome-wide genotyping technology together with an extensive neuro-psychological test battery in a genetically homogenous sample ofhealthy individuals and psychiatric patients.

2. Methods and materials

2.1. Sample

The sample consisted of 758 north-western Europeans (88%ethnic Norwegians, 12% with one parent born elsewhere in north-western Europe) including healthy individuals (n¼ 377) andpatients with schizophrenia spectrum disorders (n¼ 204) andbipolar disorders (n¼ 177) with DNA and neurocognitive data. Atotal of 700 individuals (age 35�10.4 years; 190 with schizo-phrenia, 157 with bipolar disorder and 353 healthy individuals, 357male and 343 female) remained for analysis after quality control(Table 1).

The subjects are participants in a large ongoing study onschizophrenia and bipolar disorders, the Thematically OrganizedPsychosis (TOP) Study (Birkenaes et al., 2007). The patients wererecruited from outpatient and inpatient psychiatric units at fourUniversity Hospitals in Oslo, Norway, from May 2003 through July2008. The healthy individuals were randomly selected fromnational records of people from the same catchment areas as thepatients in the same time period.

The health care system is catchment area based, free of charge,and no other psychiatric health care provider exists. The patientswere invited to participate in the study by the clinician responsible

Table 1Demographic description of cohort. For continuous variables, data is presented as mean

Bipolar disorder (n¼ 157) Schizop

Age (years) 37.26 (11.74) 33.21 (Education (years) 13.87 (2.49) 12.52 (Sex (%male) 42 54GAF (symptom)a 58.67 (11.12) 41.93 (GAF (function)a 55.80 (12.13) 43.29 (Duration of diseaseb 29.17 (11.50) 7.93 (7

a Missing data from n¼ 7.b Missing data from n¼ 9.

for their treatment. In order to participate in the current study theyneeded to be 17e65 years old and meet the Diagnostic and Statis-tical Manual of Mental Disorders (DSM) -IV (Amer.Psych.Assoc.,2010) criteria for schizophrenia, schizoaffective or schizophreni-form disorder or bipolar disorder I, bipolar disorder II or bipolardisorder NOS, and bewilling and able to give informed consent. Thestudy was approved by the Regional Committee for MedicalResearch Ethnics and the Norwegian Data Inspectorate, and wascarried out in accordance with the latest version of the Declarationof Helsinki. The biobank was approved by the Norwegian Direc-torate of Health.

2.2. Clinical assessment

Clinical assessment was carried out by trained psychiatrists andclinical psychologists. Diagnosis was established using the Struc-tured Clinical Interview for DSM-IV-TR-axis I disorders (SCID-I)(Spitzer et al., 1992). All interviewers underwent training in SCIDassessment based on a UCLA training program (Ventura et al.,1998), and participated in frequent diagnostic evaluation meet-ings led by an experienced clinical researcher in the field of diag-nostics in severe mental disorder. To assess reliability for the actualstudy interviews, a stratified random sample was drawn consistingof cases from each of the raters. Anonymous vignettes describingsymptoms and development of the illness were then rated by twoexperts blind to the study ratings. For the 28 vignettes the overallagreement for the DSM-IV diagnostic categories was 82% and theoverall Kappa 0.77 (95% CI:0.60e0.94).

The healthy participants were randomly selected from nationalstatistical records from the same catchment area and contacted byletter inviting them to participate. The healthy sample wasscreened with interview and with the Primary Care Evaluation ofMental Disorders (PRIME-MD) (Spitzer et al., 1994) and subjectswere excluded if they or any close relatives had a history of a severepsychiatric disorder (schizophrenia, bipolar disorder and majordepression), or substance abuse or dependency in the last threemonths.

Exclusion criteria for all groups were: IQ score below 70,hospitalized head injury, neurological disorder, unstable oruncontrolled medical condition that interferes with brain function(including hypothyroidism, uncontrolled hypertension and dia-betes), outside the age range 17e65 years. To assure valid neuro-cognitive test performance all participants had to have Norwegianas their first language or have received their compulsory schoolingin Norway, and had to score� 15 on the forced recognition trial inthe California Verbal Learning Test (CVLT-II) (Delis et al., 2004).

2.3. Neurocognitive assessment

Neurocognitive assessment was carried out by psychologiststrained in standardized neuropsychological testing. A 3 h testbattery (including measures of estimated premorbid IQ and

(standard deviation).

hrenia (n¼ 190) Controls (n¼ 353) All (n¼ 700)

9.66) 34.89 (10.02) 34.96 (10.42)2.45) 14.08 (2.21) 13.61 (2.43)

49 4910.95)10.38).37)

M. LeBlanc et al. / Journal of Psychiatric Research 46 (2012) 271e278 273

adequate test effort) was administered in a fixed order with twobreaks with refreshments (Supplementary Table 1).

Attention was tested with the Digit Span Backward, Digit SpanForward and Letter Number components of the Wechsler AdultIntelligence Scale e Third Edition (WAIS-III) (Wechsler, 2007a) andthe simple RT and d-prime components of the Bergen n-back test(Haatveit et al., 2010). Executive functioning was tested with theLetter Fluency, Category Fluency and Category Switching compo-nents of the Delis-Kaplan Executive Function System (D-KEFS)Word Fluency Test (Delis et al., 2005) and Condition 3:Inhibition(CW-3) and Condition 4:Inhibition/Switching components (CW-4)of the D-KEFS ColoreWord Interference Test (Delis et al., 2005).Psychomotor speedwas tested with Condition 1:Color Naming (CW-1) and Condition 2:Word Reading components (CW-2) of the D-KEFS ColoreWord Interference Test (Delis et al., 2005), the DigitSymbol Coding component of the WAIS-III test (Wechsler, 2003)and with the left and right hand average for the Grooved PegboardTest (Kløve, 1963). Learning and Memory was tested with the LongDelay Free Recall component of the California Verbal Learning Teste Second Edition (CVLT-II) (Delis et al., 2004), the Logical Memory IRecall Total Score and Logical Memory I Thematic Total Scorecomponents of the Wechsler Memory Scale e Third Edition (WMS-III) (Wechsler, 2007b) and the Long Term Memory (RCFT LTM)component of the Rey Complex Figure Test (RCFT) (Meyers andMeyers, 1995). Intelligence was tested with the Block Design,Matrix Reasoning, Similarities and Vocabulary components of theWechsler Abbreviated Scale of Intelligence (WASI) (Wechsler,2007a) and the National Adult Reading Test (NART) (Sundet andVaskinn, 2008).

2.4. Genotyping

The subset of the TOP study meeting the above clinical pheno-type definitions and having neurocognitive information available(n¼ 758) was genotyped at Expression Analysis Inc (Durham, USA)using the Affymetrix Genome-Wide Human SNP Array 6.0 (Affy-metrix Inc, Santa Clara, CA, USA). Affymetrix Power Tools (APT),v1.10.0 and the embedded birdseed-v2 algorithm was used forgenotyping. Subsequent output files were modified and combinedwith marker annotation and transposed. ped and .fam files werecreated as input files for PLINK v1.06 (Purcell et al., 2007) whichwas then used for quality control. Samples were excluded if the callratewas<97%, if theywere duplicates, if they had a sex determinedby X chromosome marker homozygosity different from their re-ported sex, were heterozygosity outliers (>3 SD from meanheterozygosity) or if they had non-European ancestry (Athanasiuet al., 2010). Of the 909,622 SNP markers on the Affymetrix 6.0Array, SNPs were excluded if successfully genotyped in <97% ofindividuals, if monomorphic or with minor allele frequencies<0.05; 550,291 markers remained for analysis. The total genotyp-ing rate in remaining individuals was 99.5%.

2.5. Association analysis

Using Plink v1.06, each SNP-neurocognitive outcome combina-tion was analyzed using a linear regression model including clini-cally important covariates but no interaction terms. Covariateswere age at neurocognitive testing, number of years completededucation, gender and disease status (schizophrenia, bipolardisorder or healthy). Disease status was modeled as a categoricalvariable using “healthy” as the base. We decided to includeeducation as a potential confounder in line with previous studies(Need et al., 2009; Cirulli et al., 2010). Genotype was modeled usingthe additive genetic model. The Wald test for the regression

coefficient for the SNP term was used to assess the strength ofassociation.

To further screen the results and to look for patterns of resultsacross both SNPs and neurocognitive tests, SNPs trending towardsignificance were identified. Trending SNPs were considered sinceinteractions are possible without the presence of a strong marginaleffect (Cordell, 2009). The p-values for trending SNPs were plottedagainst all neurocognitive tests using the levelplot() function fromthe lattice package in R (v2.11.). For any pairs of independent SNPsshowing trends toward significance for a given neurocognitive test,epistatic analysis was carried out using a linear regression modelincluding both SNPs, a SNP-by-SNP interaction term and thecovariates used in the main association analysis. The Wald test forthe interaction termwas used to test for epistasis and AIC was usedto assess the model fit.

2.6. Multiple testing

We accounted for multiple testing of SNPs using a Bonferronicorrection (ignoring LD) over 550,291 tests with an overall a¼ 0.05for significance testing (p< 9.1�10�8) and an a¼ 0.2 for identi-fying trends. We used this conservative Bonferroni correction tocorrect for multiple tests within each phenotype. A correction overall 24 neurocognitive tests treating each test as independent wouldbe too conservative because there are known dependenciesbetween the phenotypes. However, several of the neuro-psychological tests measure aspects of the same domain. Based onclinical categorization of the 24 tests, we identified 5 neuro-cognitive domains, and corrected for these domains to control formultiple phenotypes.

3. Results

We carried out a genome-wide association study for SNPsassociated with neurocognitive variables in a sample of caseshaving schizophrenia (n¼ 190) or bipolar disorder (n¼ 157) and353 healthy individuals that met our inclusion criteria and passedgenetic quality control (see Table 1 for cohort description). A total of550,291 autosomal SNPs on the Affymetrix Array 6.0 wereanalyzed. Based on the p-value the Wald test for the SNP term ina linear regression model including disease status and clinicallyimportant covariates, 4 SNPs reach genome-wide significance fora within phenotype Bonferroni correction for 550,291 tests(p< 9.1�10�8); Table 2. We did not see evidence of genomicinflation (see Table S3 and Fig. S1 for genomic inflation values andQQ plots).

Three of these associations are for intronic SNPs in PTPROwith the CVLT-II Long Delay Free Recall cognitive test (learningand memory): rs17222089 (p¼ 1.6�10�8), rs11056571(p¼ 1.68� 10�8), rs23002990 (p¼ 1.09�10�8), and 1 association isfor a SNP downstream of WDR72 with the CW-3: Inhibitioncognitive test (executive functioning): rs719714 (p¼ 4.32�10�8).The 3 significant PTPRO SNPs are in near complete linkagedisequilibrium (LD) (r2> 0.98). See Fig. 1 for Manhattan plots forCVLT-II Long Delay Free Recall and CW-3: inhibition and Fig. 2 forplots of neurocognitive score versus SNP by clinical diagnosis.Fig. 2a and b shows that for both the PTPRO SNPs and the WDR72SNP the trend in neurocognitive score by genotype (for CVLT-II LongDelay Free Recall and the CW-3: Inhibition respectively) isapproximately the same for each of the clinical groups, indicatingthe SNP effect is similar in all 3 clinical groups. For the PTPRO SNPsthe control group appears to be different for minor allele homo-zygotes (AA), however, there are too few minor allele homozygotesto draw this conclusion. For the PTPRO SNPs (only rs2300290 isshown in Fig. 2) there is an increase in the number of words

Table 2SNP annotation and regression results for SNPs with p< 9.1�10�8. MAF, minor allele frequency; Coef, regression coefficient estimate; SE, standard error; HWE, Har-dyeWeinberg equilibrium; bpd, bipolar disorders.

SNP Position chromosome(band) base pairs

Major/minor(MAF)

Type Gene Neurocognitivephenotype (domain)

n Linear regression HWE

Variable Coef (SE) Stat p-Value p-Value

rs17222089 12 (p12.3) G/A (0.17) Intronic PTPRO CVLT-II long delayfree recall(learning and memory)

698 rs17222089 �1.08 (0.19) �5.72 1.553 10L08 0.8515,618,164 Age �0.05 (0.01) �5.24 2.11� 10�07

Education 0.18 (0.04) 4.19 3.14� 10�05

Sex �1.21 (0.2) �6.01 2.98� 10�09

Schizophrenia �2.06 (0.25) �8.31 4.92� 10�16

bpd �0.83 (0.26) �3.23 0.001321

rs11056571 12 (p12.3) A/G (0.17) Intronic PTPRO CVLT-II long delayfree recall(learning and memory)

696 rs11056571 �1.08 (0.19) �5.71 1.683 10L08 0.8615,621,687 Age �0.05 (0.01) �5.15 3.36� 10�07

Education 0.18 (0.04) 4.15 3.79� 10�05

Sex �1.21 (0.2) �6.02 2.84� 10�09

Schizophrenia �2.05 (0.25) �8.28 6.37� 10�16

bpd �0.82 (0.26) �3.22 0.00136

rs2300290 12 (p12.3) A/G (0.17) Intronic PTPRO CVLT-II long delayfree recall(learning and memory)

698 rs2300290 �1.1 (0.19) �5.79 1.093 10L08 115,626,709 Age �0.05 (0.01) �5.25 2.03� 10�07

Education 0.18 (0.04) 4.2 3.03� 10�05

Sex �1.2 (0.2) �5.97 3.83� 10�09

Schizophrenia �2.06 (0.25) �8.33 4.42� 10�16

bpd �0.83 (0.26) �3.24 0.001245

rs719714 15 (q21.2) G/A (0.06) Downstream WDR72 CW-3: inhibition(executive functioning)

682 rs719714 9.45 (1.71) 5.54 4.323 10L08 151,066,670 Age 0.17 (0.06) 3.03 0.00255

Education �0.85 (0.25) �3.41 0.000691Sex 4.57 (1.16) 3.94 9.01� 10�05

Schizophrenia 12.83 (1.43) 9 2.36� 10�18

bpd 8.66 (1.47) 5.89 6.13� 10�09

M. LeBlanc et al. / Journal of Psychiatric Research 46 (2012) 271e278274

recalled with increasing number of major alleles (G). For theWDR72 SNP (rs179714) there is a decrease in the time required tocomplete the CW-3: Inhibition test as the number of major allelesincreases (A).

We used the conservative Bonferroni correction (ignoring LD) tocorrect for multiple tests over each phenotype. Additionally, the 3PTPRO SNPs remained significant (p< 1.8� 10�8), after correctionfor 5 neurocognitive domains.

In addition to looking for genome-wide significance, we con-ducted an analysis looking for patterns of associations acrossneurocognitive phenotypes and across SNPs. We chose to look at allneurocognitive tests and the subset of SNPs having a trend towardsignificance for at least one neurocognitive test (p< 3.6�10�7)after Bonferroni correction for 550,291 comparisons and a¼ 0.2;Fig. 3. Generally, we did not see any patterns of trending SNPs thatgroup according to the clinically-defined neurocognitive testdomains, with the exception of similar patterns of p-values forCVLT-II Total Learning and CVLT-II Long Delay Free Recall. This is notsurprising as the 2 tests are highly correlated (Pearsoncorrelation¼ 0.85). As expected, SNPs in high LD (i.e. the chromo-some 12 SNPs on Fig. 3) show similar patterns (r2> 0.7 for theseSNPs).

As clearly seen in Fig. 3, the Grooved Pegboard test is the onlyneurocognitive test having 2 independent SNPs trending towardsignificance (rs9378605 upstream of FOXQ1 on chromosome 6(p¼ 1.22�10�7) and rs11699311 downstream of SUMO1P1 onchromosome 20 (p¼ 1.22�10�7)). Given these 2 independenttrends for the Grooved Pegboard test, we tested for epistasisbetween these 2 SNPs. In a linear regression model including bothSNPs, disease status, the clinical covariates and a SNP-SNP inter-action term, a significant interaction was found (p¼ 7.6�10�14).Furthermore, the marginal SNP effects were stronger than for thesingle-SNP models (p< 2.0�10�16 for both rs9378605 andrs11699311). The model with the interaction term has markedlybetter fit compared to either of the single-SNP models, indicatingstrong support for the inclusion of the interaction term. Both

rs11699311 and rs9378605 have relatively low minor allelefrequencies (0.09 and 0.06 respectively) and consequently very fewminor allele homozygotes in our sample. The interaction effect islargely due to the double heterozygotes having much higherGrooved Pegboard scores compared to the other genotypecombinations.

4. Discussion

Four SNPs and two independent association signals exceedingthe levels of statistical significance accepted as genome-widewithin a phenotype were identified. These findings suggest a roleof PTPRO in learning and memory, and WDR72 in executive func-tioning. We also found a highly significant interaction betweenSNPs in FOXQ1 and SUMO1P1 for psychomotor speed.

Previous GWAS of cognition have identified other geneticmarkers associated with other cognitive phenotypes than thoseidentified in the present study. The first GWAS identified commonKIBRA alleles to be associated with visual and verbal memory ina population sample (Papassotiropoulos et al., 2006). This wasreplicated by Bates et al. (2009) but failed to replicate in smallersamples (Need et al., 2008). CAMTA1 was also found to be associ-ated with memory (Huentelman et al., 2007), but was not repli-cated by Need et al. (2009). Seshadri et al. (2007) found strongevidence for association between SORL1 and abstract reasoning.Butcher et al. (2008) reported an association between generalcognitive ability and rs1378810 in a large pediatric study. However,no specific genetic variants seem to be robustly associated withgeneral intelligence, apart from APOE in the elderly.

The heritability of cognitive functions is relatively high(McClearn et al., 1997; Finkel et al., 1995; Bouchard et al., 1990), andcommon variants explain a large proportion of the variance ofgeneral intelligence (Davies et al., 2011). However, it has beendifficult to identify genetic variants in GWAS of healthy individualsthat replicate (Cirulli et al., 2010; Need et al., 2009). This could bedue to the relatively small variation in cognitive function in healthy

Fig. 2. Covariate-adjusted neurocognitive test score versus genotype rs2300290 and rs719714. Note that rs17222089 and rs11056571 are not displayed because they are in nearcomplete linkage disequilibrium with rs2300290, thus have nearly identical plots. Data is displayed as mean and 95% confidence interval. Confidence intervals are excluded forgroups having n� 2. The minor allele homozygotes are very infrequent for both SNPs (between 1 and 10 for each clinical group).

Fig. 1. Manhattan plots for neurocognitive tests having genome-wide significant SNPs. The horizontal line indicates the cut-off for genome-wide significance.

M. LeBlanc et al. / Journal of Psychiatric Research 46 (2012) 271e278 275

Fig. 3. Log p-value for SNPs having at least a trend toward significance for at least one neurocognitive test. All cognitive tests are displayed.

M. LeBlanc et al. / Journal of Psychiatric Research 46 (2012) 271e278276

individuals, which requires large sample sizes to detect a signal.Patient groups have on average reduced cognitive performance sowe see a greater variability in test scores than we would if we hadan identical sample size including only healthy individuals. Toaddress that, we controlled for disease group in our analysis byincluding it as a covariate.

For the SNPs having significance in this study, we reran theanalysis excluding disease status as a covariate and the resultsremain stable. The SNP effect on neurocognition is approximatelythe same for all clinical groups. The concept of using disease groupsto study genetic factors of cognition is supported by Vassos et al.(2010), who replicated the KIBRA association to cognitive functionin a sample including psychosis. Further, we used a geneticallyhomogenous sample of Norwegians, which increases the chance offinding gene effects, as shown earlier in brain phenotypes (Rimolet al., 2010).

There is ongoing debate about the inclusion of education as a co-variatewhen lookingatmeasures of intelligence.Other recent studieshave included years of education in their study (e.g. Need et al., 2009;Cirulli et al., 2010) while others have not (e.g. Davies et al., 2011).Given this uncertainty, we tested our findings for robustness con-cerning the education covariate and found that they remain signifi-cant upon the exclusion of education. Although we found significantresults for two neurocognitive tests, we did not see any significantmain results formostof theneurocognitive phenotypes. This is in linewith previous studies, and could be due to several factors, mostprobably the limited sample size. There is also the problem thatcurrent GWAS are only able to identify genetic factors that explaina small fraction of heritable phenotypes (the missing heritability).

The mechanisms underlying learning and memory have beenextensive studied. The hippocampus and connected brain regionsplay an essential role (Silva et al., 2009; Thompson, 1986), and LongTerm Potentiation and other aspects of neuronal plasticity areinvolved (Hubener and Bonhoeffer, 2010). The molecular mecha-nisms underlying these processes have been studied in details in

animal models, and current evidence makes is likely that PTPROcould be involved. PTPRO is most strongly expressed in the devel-oping nervous system (Bodden and Bixby,1996; Tagawa et al., 1997;Thomas et al., 1994; Beltran et al., 2003) and has been suggested toregulate axon guidance during development (Stepanek et al., 2005;Shintani et al., 2006). PTPRO, which is located at 12p13-p12,encodes a receptor-type protein tyrosine phosphatase containinga single intracellular catalytic domain with a characteristic signa-ture motif. The gene product, which has a transmembrane domain,is an integral membrane protein. Several alternatively splicedtranscript variants, some of which encode different isoforms of theprotein, have been described. These variants exhibit tissue-specificexpression, and is expressed in neuronal tissue (Kotani et al., 2010;Gonzalez-Brito and Bixby, 2009; Hower et al., 2009).

Executive functioning is another critical cognitive domain,essential for human functioning, such as employment ability(Kalechstein et al., 2003), and is suggested to play a key role inhuman evolution (Balter, 2010). Dysfunctional working memoryhas been implicated in several brain disorders (Bora et al., 2009;Robinson and Ferrier, 2006; Levy and Chelune, 2007). There isnow strong evidence from lesion and functional brain imagingstudies to implicate frontal cortex, especially dorsolateralprefrontal cortex, as a key region in executive function (Minzenberget al., 2009). Studies of underlying molecular mechanisms haveindicated some candidate genes related to neurotransmission(COMT, AKT,) and neuronal plasticity (BDNF) in executive function(Greene et al., 2008; Eisenberg and Berman, 2010; Tan et al., 2009).A role of COMT has been supported by both brain imaging andbehavior (neuropsychological) studies.WDR72 is located at 15q21.3and consists of 19 coding exons spanning w250 kb, encodinga protein of 1102 amino acids about which little is known. Severallines of evidence suggest that WDR72 could have a role in theneuronal mechanisms related to executive function. Its closesthuman homolog and a possible paralog is WDR7, also known asRabconnectin-3b or TGF-b resistance-associated protein (TRAG)

M. LeBlanc et al. / Journal of Psychiatric Research 46 (2012) 271e278 277

(Sanders et al., 2000). Rabconnectin contributes to activation anddeactivation of RAB3A, a small GTP-binding protein that is a positiveregulator of vesicle mobilization and Ca2þ dependent exocytosis ofneurotransmitter release at synaptic vesicles (Sakisaka and Takai,2005; Coleman and Bykhovskaia, 2009).

We found an epistatic interaction between FOXQ1 and SUMO1P1for psychomotor speed. While there are several possibleapproaches for looking for epistasis (Cordell, 2009), we decided tofocus on SNPs trending toward significance in the single-SNPmodel. This is reasonable to do since SNPs having interactioneffects may not have a strong marginal effect. The FOXQ1-SUMO1P1interaction remains highly significant upon the exclusion of the 1individual having a large residual (but clinically valid score).Control of movements is a highly studied phenotype, and thedistinct neuronal mechanisms are well described (Scott, 2008;Kurtzer and Scott, 2007), but less is known about the underlyingmolecular genetic architecture. FOXQ1 is a member of the FOX genefamily, which is characterized by a conserved 110-amino acid DNA-binding motif called the forkhead/winged helix domain. Thesegenes are involved in embryonic development, cell cycle regulation,tissue-specific gene expression, cell signaling, and tumorigenesis(Katoh and Katoh, 2004; Bieller et al., 2001). FOXQ1 is a transcrip-tion factor, and loss of FOXQ1 in mice leads embryonic lethalityduring organogenesis. There is some evidence for FOXQ1 expres-sion in human brain tissue (Safran et al., 2010). SUMO1P1 isa polymorphic pseudogene, one of eight human SUMO1 pseudo-genes. The function is largely unknown, and there is little infor-mation about expression. Another member of the FOX family oftranscription factors, FOXP2, is important for language develop-ment (Enard et al., 2002). The protein product is an evolutionarilyconserved transcription factor, which may bind directly toapproximately 300e400 gene promoters in the human genome toregulate the expression of a variety of genes. Thus, similar mech-anisms could be involved with FOXQ1. There could be a potentialmolecular mechanism for interaction between these genes, butgiven our limited knowledge about SUMO1P1, it is difficult discern.

Neurocognitive outcomes present challenges in terms ofreplication. While other samples having measures of cognitiveperformance and genome-wide genotypes do exist, there is vari-ation in the specific neuropsychological tests used. Furthermore,neurocognitive testing is usually carried out as secondary pheno-typing and there may be large differences in the inclusion criteriabetween each study. Thus, we were not able to identify anysamples with the same neurocognitive test battery and inclusioncriteria. However, we expect this will change in the future, whenmore samples will become available. In this study we chose to useraw test scores while other groups have taken the approach ofusing principal components analysis to calculate a g factor, whichdoes help reduce the dimension of the neurocognitive data but hasthe drawback of not capturing all of the variability in the data. Thecomparability of this g factor across studies with different neu-rocognitive tests is still uncertain and is an area of research thatcurrently is being explored.

The current study has some limitations. We lack a replicationsample, but this has been difficult to obtain due to the variation inneuropsychological tests and inclusion criteria. However, our genesignals reached the threshold usually recognized as genome-widesignificant. We did not correct for all cognitive phenotypes due tothe known dependencies between the cognitive phenotypes.However, even after correcting for the main cognitive domains thethree PTPRO SNPs remained genome-wide significant. We useda sample including both healthy individuals and mental disorders,but we controlled for the diagnostic categories in the analysis andsimilar SNP affects in all diagnostic groups were observed for thereported results.

To conclude, the current results indicate a role of PTPRO inlearning and memory, WDR72 with executive functioning anda joint role of FOXQ1-SUMO1P1 for psychomotor speed. Thesefindings should be replicated in independent samples.

Role of funding source

This workwas supported by Oslo University Hospital, Universityof Oslo, South-Eastern Norway Health Authority (grant # 2004-123), and the Research Council of Norway (#167153/V50, #163070/V50). Eli Lilly supported parts of the genotyping costs.

Contributors

ML and OAA wrote the first draft of the paper. IA, KS, IM, OAArecruited, diagnosed and gathered phenotypes, and SD collectedgenotype data. ML, BK, AF, analyzed the data. BK, AF, SD, KS, OAAplanned and coordinated the work. All authors contributed to thecurrent version of the paper.

Conflict of interestNone of the authors have any competing interests in relation to

the topic of this paper.

Acknowledgments

We are thankful to the study participants and the members ofthe TOP study group involved in data collection.

Appendix. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.jpsychires.2011.11.001.

References

Amer.Psych.Assoc.. Diagnostic and statistical manual of mental disorders. Wash-ington, DC: APA; 2010.

Arguello PA, Gogos JA. Cognition in mouse models of schizophrenia susceptibilitygenes. Schizophrenia Bulletin 2010;36:289e300.

Athanasiu L, Mattingsdal M, Kahler AK, Brown A, Gustafsson O, Agartz I, et al. Genevariants associated with schizophrenia in a Norwegian genome-wide study arereplicated in a large European cohort. Journal of Psychiatric Research 2010;44:748e53.

Balter M. Evolution of behavior. Did working memory spark creative culture?Science 2010;328:160e3.

Bates TC, Price JF, Harris SE, Marioni RE, Fowkes FG, Stewart MC, et al. Association ofKIBRA and memory. Neuroscience Letters 2009;458:140e3.

Beltran PJ, Bixby JL, Masters BA. Expression of PTPRO during mouse developmentsuggests involvement in axonogenesis and differentiation of NT-3 and NGF-dependent neurons. Journal of Comparative Neurology 2003;456:384e95.

Bieller A, Pasche B, Frank S, Glaser B, Kunz J, Witt K, et al. Isolation and charac-terization of the human forkhead gene FOXQ1. DNA and Cell Biology 2001;20:555e61.

Birkenaes AB, Opjordsmoen S, Brunborg C, Engh JA, Jonsdottir H, Ringen PA, et al.The level of cardiovascular risk factors in bipolar disorder equals that ofschizophrenia: a comparative study. Journal of Clinical Psychiatry 2007;68:917e23.

Bodden K, Bixby JL. CRYP-2: a receptor-type tyrosine phosphatase selectivelyexpressed by developing vertebrate neurons. Journal of Neurobiology 1996;31:309e24.

Bora E, Yucel M, Pantelis C. Cognitive functioning in schizophrenia, schizoaffectivedisorder and affective psychoses: meta-analytic study. British Journal ofPsychiatry 2009;195:475e82.

Bouchard Jr TJ, Lykken DT, McGue M, Segal NL, Tellegen A. Sources of humanpsychological differences: the Minnesota study of twins reared apart. Science1990;250:223e8.

Burdick KE, DeRosse P, Kane JM, Lencz T, Malhotra AK. Association of geneticvariation in the MET proto-oncogene with schizophrenia and general cognitiveability. American Journal of Psychiatry 2010;167:436e43.

Butcher LM, Davis OS, Craig IW, Plomin R. Genome-wide quantitative trait locusassociation scan of general cognitive ability using pooled DNA and 500K singlenucleotide polymorphism microarrays. Genes, Brain and Behavior 2008;7:435e46.

M. LeBlanc et al. / Journal of Psychiatric Research 46 (2012) 271e278278

Cirulli ET, Kasperaviciute D, Attix DK, Need AC, Ge D, Gibson G, et al. Commongenetic variation and performance on standardized cognitive tests. EuropeanJournal of Human Genetics 2010;18:815e20.

Coleman WL, Bykhovskaia M. Rab3a-mediated vesicle recruitment regulates short-term plasticity at the mouse diaphragm synapse. Molecular and CellularNeuroscience 2009;41:286e96.

Cordell HJ. Detecting geneegene interactions that underlie human diseases. NatureReviews Genetics 2009;10:392e404.

Crawley JN. Behavioral phenotyping strategies for mutant mice. Neuron 2008;57:809e18.

Davies G, Tenesa A, Payton A, Yang J, Harris SE, Liewald D, et al. Genome-wideassociation studies establish that human intelligence is highly heritable andpolygenic. Molecular Psychiatry 2011;16:996e1005.

Delis DC, Kramer JH, Kaplan E, Ober BA. California verbal learning test e secondedition (CVLT-II). Norwegian manual supplement. Stockholm: Pearson Assess-ment; 2004.

Delis DC, Kaplan E, Kramer JH. Delis e Kaplan executive function system (D-KEFS)Norwegian manual. Stockholm: Pearson Assessment; 2005.

Egan MF, Goldberg TE, Kolachana BS, Callicott JH, Mazzanti CM, Straub RE, et al.Effect of COMT Val108/158 Met genotype on frontal lobe function and risk forschizophrenia. Proceedings of the National Academy of Sciences of the UnitedStates of America 2001;98:6917e22.

Egan MF, Kojima M, Callicott JH, Goldberg TE, Kolachana BS, Bertolino A, et al. TheBDNF val66met polymorphism affects activity-dependent secretion of BDNFand human memory and hippocampal function. Cell 2003;112:257e69.

Eisenberg DP, Berman KF. Executive function, neural circuitry, and genetic mecha-nisms in schizophrenia. Neuropsychopharmacology 2010;35:258e77.

Enard W, Przeworski M, Fisher SE, Lai CS, Wiebe V, Kitano T, et al. Molecularevolution of FOXP2, a gene involved in speech and language. Nature 2002;418:869e72.

Finkel D, Pedersen NL, McGue M, McClearn GE. Heritability of cognitive abilities inadult twins: comparison of Minnesota and Swedish data. Behavior Genetics1995;25:421e31.

Geschwind DH. Advances in autism. Annual Review of Medicine 2009;60:367e80.Gonzalez-Brito MR, Bixby JL. Protein tyrosine phosphatase receptor type O regu-

lates development and function of the sensory nervous system. Molecular andCellular Neuroscience 2009;42:458e65.

Green AE, Munafo MR, DeYoung CG, Fossella JA, Fan J, Gray JR. Using genetic data incognitive neuroscience: from growing pains to genuine insights. NatureReviews Neuroscience 2008;9:710e20.

Greene CM, Braet W, Johnson KA, Bellgrove MA. Imaging the genetics of executivefunction. Biological Psychology 2008;79:30e42.

Haatveit BC, Sundet K, Hugdahl K, Ueland T, Melle I, Andreassen OA. The validity ofd prime as a working memory index: results from the “Bergen n-back task”.Journal of Clinical and Experimental Neuropsychology 2010;32:871e80.

Hower AE, Beltran PJ, Bixby JL. Dimerization of tyrosine phosphatase PTPROdecreases its activity and ability to inactivate TrkC. Journal of Neurochemistry2009;110:1635e47.

Hubener M, Bonhoeffer T. Searching for engrams. Neuron 2010;67:363e71.Huentelman MJ, Papassotiropoulos A, Craig DW, Hoerndli FJ, Pearson JV, Huynh KD,

et al. Calmodulin-binding transcription activator 1 (CAMTA1) alleles predisposehuman episodic memory performance. Human Molecular Genetics 2007;16:1469e77.

Kalechstein AD, Newton TF, van Gorp WG. Neurocognitive functioning is associatedwith employment status: a quantitative review. Journal of Clinical and Exper-imental Neuropsychology 2003;25:1186e91.

Katoh M, Katoh M. Human FOX gene family. International Journal of Oncology2004;25:1495e500 [Review].

Kløve H. Clinical neuropsychology. In: Forster FM, editor. The medical clinics ofNorth America. New York: Saunders; 1963.

Kotani T, Murata Y, Ohnishi H, Mori M, Kusakari S, Saito Y, et al. Expression of PTPROin the interneurons of adult mouse olfactory bulb. Journal of ComparativeNeurology 2010;518:119e36.

Kurtzer I, Scott SH. A multi-level approach to understanding upper limb function.Progress in Brain Research 2007;165:347e62.

Levy JA, Chelune GJ. Cognitive-behavioral profiles of neurodegenerative dementias:beyond Alzheimer’s disease. Journal of Geriatric Psychiatry and Neurology2007;20:227e38.

Lott IT, Dierssen M. Cognitive deficits and associated neurological complications inindividuals with Down’s syndrome. Lancet Neurology 2010;9:623e33.

McClearn GE, Johansson B, Berg S, Pedersen NL, Ahern F, Petrill SA, et al. Substantialgenetic influence on cognitive abilities in twins 80 or more years old. Science1997;276:1560e3.

Meyers JE, Meyers KR. Rey complex figure test and recognition trial. Odessa, FL:Psychological Assessment Resources, Inc.; 1995.

Minzenberg MJ, Laird AR, Thelen S, Carter CS, Glahn DC. Meta-analysis of 41functional neuroimaging studies of executive function in schizophrenia.Archives of General Psychiatry 2009;66:811e22.

Need AC, Attix DK, McEvoy JM, Cirulli ET, Linney KL, Hunt P, et al. A genome-widestudy of common SNPs and CNVs in cognitive performance in the CANTAB.Human Molecular Genetics 2009;18:4650e61.

Need AC, Attix DK, McEvoy JM, Cirulli ET, Linney KN, Wagoner AP, et al. Failure toreplicate effect of Kibra on human memory in two large cohorts of Europeanorigin. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics2008;147B:667e8.

Nicodemus KK, Callicott JH, Higier RG, Luna A, Nixon DC, Lipska BK, et al. Evidenceof statistical epistasis between DISC1, CIT and NDEL1 impacting risk forschizophrenia: biological validation with functional neuroimaging. HumanGenetics 2010;127:453e4.

Papassotiropoulos A, Stephan DA, Huentelman MJ, Hoerndli FJ, Craig DW,Pearson JV, et al. Common Kibra alleles are associated with human memoryperformance. Science 2006;314:475e8.

Porteous DJ, Thomson P, Brandon NJ, Millar JK. The genetics and biology of DISC1 e

an emerging role in psychosis and cognition. Biological Psychiatry 2006;60:123e31.

Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, et al. PLINK:a tool set for whole-genome association and population-based linkage analyses.American Journal of Human Genetics 2007;81:559e75.

Rimol LM, Agartz I, Djurovic S, Brown AA, Roddey JC, Kahler AK, et al. Sex-depen-dent association of common variants of microcephaly genes with brain struc-ture. Proceedings of the National Academy of Sciences of the United States ofAmerica 2010;107:384e8.

Robinson LJ, Ferrier IN. Evolution of cognitive impairment in bipolar disorder:a systematic reviewof cross-sectional evidence. BipolarDisorders2006;8:103e16.

Safran M, Dalah I, Alexander J, Rosen N, Iny ST, Shmoish M, et al. GeneCards version3: the human gene integrator. Database. Oxford; 2010.

Sakisaka T, Takai Y. Purification and properties of rabconnectin-3. Methods inEnzymology 2005;403:401e7.

Sanders S, Keck-Waggoner CL, Zimonjic DB, Popescu NC, Thorgeirsson SS. Assign-ment of WDR7 (alias TRAG, TGF-beta resistance associated gene) to orthologousregions of human chromosome 18q21.1/q22 and mouse chromosome 18D.1-E.3 by fluorescence in situ hybridization. Cytogenetics and Cell Genetics 2000;88:324e5.

Scott SH. Inconvenient truths about neural processing in primary motor cortex.Journal of Physiology 2008;586:1217e24.

Seshadri S, DeStefano AL, Au R, Massaro JM, Beiser AS, Kelly-Hayes M, et al. Geneticcorrelates of brain aging on MRI and cognitive test measures: a genome-wideassociation and linkage analysis in the Framingham study. BMC MedicalGenetics 2007;8(Suppl. 1):S15.

Shintani T, Ihara M, Sakuta H, Takahashi H, Watakabe I, Noda M. Eph receptors arenegatively controlled by protein tyrosine phosphatase receptor type O. NatureNeuroscience 2006;9:761e9.

Silva AJ, Zhou Y, Rogerson T, Shobe J, Balaji J. Molecular and cellular approaches tomemory allocation in neural circuits. Science 2009;326:391e5.

Simonsen C, Sundet K, Vaskinn A, Birkenaes AB, Engh JA, Faerden A, et al. Neuro-cognitive dysfunction in bipolar and schizophrenia spectrum disorders dependson history of psychosis rather than diagnostic group. Schizophrenia Bulletin;2009.

Spitzer RL, Williams JB, Gibbon M, First MB. The structured clinical interview forDSM-III-R (SCID). I: history, rationale, and description. Archives of GeneralPsychiatry 1992;49:624e9.

Spitzer RL, Williams JB, Kroenke K, Linzer M, deGruy III FV, Hahn SR, et al. Utility ofa new procedure for diagnosing mental disorders in primary care. The PRIME-MD 1000 study. JAMA 1994;272:1749e56.

Stepanek L, Stoker AW, Stoeckli E, Bixby JL. Receptor tyrosine phosphatases guidevertebrate motor axons during development. Journal of Neuroscience 2005;25:3813e23.

Sundet K, Vaskinn A. Estimating premorbid IQ. Journal of the Norwegian Psycho-logical Association 2008;45:1108e15 [in Norwegian with English abstract].

Tagawa M, Shirasawa T, Yahagi Y, Tomoda T, Kuroyanagi H, Fujimura S, et al.Identification of a receptor-type protein tyrosine phosphatase expressed inpostmitotic maturing neurons: its structure and expression in the centralnervous system. Biochemical Journal 1997;321:865e71.

Tan HY, Callicott JH, Weinberger DR. Prefrontal cognitive systems in schizophrenia:towards human genetic brain mechanisms. Cognitive Neuropsychiatry 2009;14:277e98.

Tan HY, Chen Q, Sust S, Buckholtz JW, Meyers JD, Egan MF, et al. Epistasis betweencatechol-O-methyltransferase and type II metabotropic glutamate receptor 3genes onworking memory brain function. Proceedings of the National Academyof Sciences of the United States of America 2007;104:12536e41.

Thomas PE, Wharram BL, Goyal M, Wiggins JE, Holzman LB, Wiggins RC. GLEPP1,a renal glomerular epithelial cell (podocyte) membrane protein-tyrosinephosphatase. Identification, molecular cloning, and characterization in rabbit.Journal of Biological Chemistry 1994;269:19953e62.

Thompson RF. The neurobiology of learning and memory. Science 1986;233:941e7.Toulopoulou T, Picchioni M, Rijsdijk F, Hua-Hall M, Ettinger U, Sham P, et al.

Substantial genetic overlap between neurocognition and schizophrenia: geneticmodeling in twin samples. Archives of General Psychiatry 2007;64:1348e55.

Vassos E, Bramon E, Picchioni M, Walshe M, Filbey FM, Kravariti E, et al. Evidence ofassociation of KIBRA genotype with episodic memory in families of psychoticpatients and controls. Journal of Psychiatric Research 2010;44:795e8.

Ventura J, Liberman RP, Green MF, Shaner A, Mintz J. Training and quality assurancewith the structured clinical interview for DSM-IV (SCID-I/P). PsychiatryResearch 1998;79:163e73.

Wechsler D. Wechsler adult intelligence scale e third edition (WAIS-III). Norwegianmanual. Stockholm: Pearson Assessment; 2003.

Wechsler D. Wechsler abbreviated scale of intelligence (WASI). Norwegian manualsupplement. Stockholm: Pearson Assessment; 2007a.

Wechsler D. Wechsler memory scale e third edition (WMS-III). Norwegian manual.Stockholm: Pearson Assessment; 2007b.