Available online at www.sciencedirect.com

98 (2008) 129–138www.elsevier.com/locate/schres

Schizophrenia Research

Expression of oligodendrocyte-associated genes in dorsolateralprefrontal cortex of patients with schizophrenia

Shruti N. Mitkus, Thomas M. Hyde, Radhakrishna Vakkalanka, Bhaskar Kolachana,Daniel R. Weinberger, Joel E. Kleinman, Barbara K. Lipska ⁎

Clinical Brain Disorders Branch, Section on Neuropathology, DIRP/NIMH/NIH, Bethesda, MD, 20892-1385, USA

Received 22 August 2007; received in revised form 20 September 2007; accepted 24 September 2007Available online 26 October 2007

Abstract

Prior studies have found decreased mRNA expression of oligodendrocyte-associated genes in the dorsolateral prefrontal cortex(DLPFC) of patients with schizophrenia. However, it is unclear which specific genes are affected and whether the changes occur in thecortical white or grey matter. We assessed the mRNA expression levels of four oligodendrocyte-related genes: myelin-associatedbasic protein (MOBP), myelin-associated glycoprotein (MAG), 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNP) andoligodendrocyte-lineage transcription factor 2 (OLIG2) in DLPFC white and grey matter using quantitative-PCR (∼70 controlsand∼30 patients with schizophrenia). We also examined the effects of high-risk polymorphisms in CNP andOLIG2 onmRNA levelsof these genes. We found that genetic polymorphisms in CNP (rs2070106) and OLIG2 (rs1059004 and rs9653711), previouslyassociated with schizophrenia, predicted low expression of these genes. Expression of MAG, CNP and OLIG2 did not differ betweenpatients with schizophrenia and controls in the grey or white matter but MOBP mRNA levels were increased in the DLPFC whitematter in patients with a history of substance abuse. MOBP and CNP protein in the white matter was not altered. Although previouslyreported reductions in the expression of myelin-related genes in the DLPFCwere not detected, we show that individuals carrying risk-associated alleles in oligodendrocyte-related genes had relatively lower transcript levels. These data illustrate the importance ofgenetic background in gene expression studies in schizophrenia.Published by Elsevier B.V.

Keywords: Schizophrenia; Myelin; Oligodendrocytes; MOBP; CNP; OLIG2; Postmortem studies

1. Introduction

Recent studies using various approaches have inves-tigated the hypothesis that oligodendrocyte function andmyelination are disrupted in schizophrenia (Davis andHaroutunian 2003;Davis et al., 2003; Kubicki et al., 2005;McInnes and Lauriat 2006; Dwork et al., 2007). The mostreplicable findings in the post-mortem brain tissue were

⁎ Corresponding author. 10 Center Drive, Room 4N306, Bethesda,MD 20892-1385, USA. Tel.: +1 301 496 9501; fax: +1 301 402 2751.

E-mail address: lipskab@intra.nimh.nih.gov (B.K. Lipska).

0920-9964/$ - see front matter. Published by Elsevier B.V.doi:10.1016/j.schres.2007.09.032

reduced expression ofMAGandCNP (Table 1).However,the same myelin-associated genes were not consistentlyaltered across various studies, although most used tissuefrom the same two brain collections, Stanley Consortiumand Mount Sinai. Two studies assessed expression ofoligodendrocyte-related genes separately in the whitematter and both found a reduction in CNP in the tissue ofpatients with schizophrenia (Prabakaran et al., 2004;McCullumsmith et al., 2007) (Table 1). There are alsoseveral association studies of polymorphisms in oligo-dendrocyte-related genes with schizophrenia. MOG,MAG, PLP1, NOGO and CNP have been found to be

Table 1Summary of postmortem findings of oligodendrocyte-related gene expression abnormalities in schizophrenia

Study Brain regionsstudied

Braincollection

Method Positive findings Comments NumbersS/C

Age

Hakak et al.(2001)

PFC MountSinai

Microarray CNP, MAG, ErbB3, MAL,TF, gelsolin reduced

MOBP, OLIG2 not reported 12/12 N70

Hof et al. (2003) SFG MountSinai

IHC Number of CNP positivecells reduced

MOBP, MAG, OLIG2, ErbB3 nottested

7/7 N75

Flynn et al.(2003)

PFC NYPI ELISA CNP reduced No change for MAG; MOBP, CNP,OLIG2, ErbB3 not tested

13/11 ∼46

Tkachev et al.(2003)

PFC Stanley Microarray/Q-PCR

OLIG2, MAG, ErbB3reduced

No change for MOBP; CNP notreported

15/15 N/A

Prabakaran et al.(2004)

PFC Stanley Proteomics CNP reduced in WM MOBP, MAG, OLIG2, ErbB3 notreported

10/10 N/A

Aston et al.(2004)

TMPGy Stanley Microarray/qPCR

MAG, ErbB3 reduced MOBP, CNP, OLIG2 not reported 12/14 ∼45

Katsel et al.(2005)

Multiple MountSinai

Microarray CNP, MAG, OLIG2, ErbB3reduced

MOBP not reported ∼13/13 N70

Dracheva et al.(2006)

HIPP, CING,CAU, PUT

MountSinai

qPCR,Western

CNP, MAG reduced No change for MOBP; OLIG2,ErbB3 not tested

∼25/20 N75

Aberg et al.(2006)

FC Stanley+Maudsley+Harvard

qPCR MAG reduced MOBP, CNP, OLIG2 not tested 55/55 N/A

Novak andTallerico(2006)

FC Stanley qPCR NOGO C increased MOBP, CNP, MAG, ErbB3, OLIG2not tested

8/14 ∼45

McCullumsmithet al. (2007)

CING MountSinai

ISH MAG, CNP, QKI, TFreduced in WM (not GM)

No change for OLIG2, ErbB3;MOBP not tested

41/34 N75

Abbreviations: CING— cingulate cortex, CAU— caudate, CNP— 2′,3′-cyclic nucleotide 3′-phosphodiesterase, ErbB3— v-erb-b2 erythroblasticleukemia viral oncogene homolog 3 (avian), ISH — in situ hybridization, IHC — immunohistochemistry, FC — frontal cortex, GM-grey matter,HIPP — hippocampus, MAG — myelin-associated glycoprotein, MOBP — myelin-associated basic protein, NYPI — New York PsychiatricInstitute, OLIG2 — oligodendrocyte-lineage transcription factor 2, PFC — prefrontal cortex, PUT — putamen, SFG — superior frontal gyrus,TMPGy — temporal gyrus, WM — white matter, N/A — not available.

130 S.N. Mitkus et al. / Schizophrenia Research 98 (2008) 129–138

associated with the disease in certain populations, but notin others (Table 2). Interestingly, a risk-associated allelealso predicted low expression of CNP mRNA in the braintissue of normal controls (Peirce et al., 2006).

To investigate the hypothesis that oligodendrocytefunction is altered in schizophrenia, we examined mRNAexpression ofMOBP, CNP,MAGandOLIG2 genes in thepost-mortem DLPFC from the CBDB/NIMH braincollection, a cohort of relatively young patients andnormal controls (Lipska et al., 2006) separately in thewhite and greymatter. Since recent reports have identifiedallelic variations in CNP and OLIG2, showing associa-tions with the disease (Peirce et al., 2006; Georgieva et al.,2006), we investigated whether the expression of CNPand OLIG2 is predicted by these polymorphic variants.

2. Materials and methods

2.1. Human post-mortem tissue

Postmortem dorsolateral prefrontal cortex (DLPFC)grey and white matter tissue samples were collected at the

Clinical Brain Disorders Branch, NIMH (Table 3). Greymatter was removed from a 1.5 cm thick coronal slab ofthe frontal cortex anterior to the corpus callosum and theDLPFC dissected using a dental drill according to theguidelines described by Rajkowska and Goldman-Rakic(1995). White matter was dissected from the region justbelow the superior and middle frontal gyri, adjacent to thecortical grey matter ribbon. A smaller subgroup of 47controls and 26 patients was used for protein studies in thewhite matter.

All brain tissue used in this study was obtained withinformed consent from the legal next of kin underNIMH protocol #90-M-0142 (Lipska et al., 2006).Normal controls had no known history of psychiatricsymptoms or substance abuse and negative toxicolog-ical results. Diagnoses of patients with schizophreniawere determined using DSM-IV criteria. Total daily,lifetime and last dose of neuroleptic medication wascalculated for each schizophrenic subject and convertedto chlorpromazine equivalents (as described in Lipskaet al., 2006). Smoking and substance abuse history wasalso recorded. Toxicological analysis of blood, vitreous

Table 2Summary of association findings of oligodendrocyte-related genes and schizophrenia

Study Population n (S/C) Findings

Wan et al. (2005) Han Chinese 470/470 Association of SNPs in MAG with schizophreniaYang et al. (2005) Han Chinese 413 patients and parents Association of MAG haplotype with schizophreniaQin et al. (2005) Han Chinese 487 patients and parents Association of SNP in PLP1 with male

patients with schizophreniaZai et al. (2005) Caucasians/Asian/African

American111 patients and families No association of SNPs in MOG with schizophrenia

Liu et al. (2005) Han Chinese 532 patients and families Association of MOG haplotype with schizophreniaCovault et al. (2004) Caucasian/African American 77/243 No association of NOGO polymorphisms and schizophreniaTan et al. (2005) Chinese 363/253 Association of polymorphisms in NOGO in female patients

with schizophreniaGregorio et al. (2005) Brazil 181/427 No association of NOGO polymorphisms and schizophreniaXiong et al. (2005) Caucasian European 462/153 No association of NOGO polymorphisms and schizophreniaPeirce et al. (2006) Caucasian/UK 708/711 Exonic SNP predicting lower CNP expression is associated

with schizophreniaGeorgieva et al. (2006) Caucasian/UK 673/716 Association of SNPs in OLIG2 with schizophreniaUsui et al. (2006) Japanese 759/757 No association of SNPs in OLIG2 and CNPKanazawa et al. (2007) Japanese 193/121 No association of SNPs in ErbB3 and schizophreniaWatanabe et al. (2007) Japanese 399/438 No association of SNPs in ErbB3 and schizophreniaTang et al. (2007) Han Chinese 426/439 No association of SNPs in CNP

Abbreviations: CNP— 2′,3′-cyclic nucleotide 3′-phosphodiesterase, ErbB3— v-erb-b2 erythroblastic leukemia viral oncogene homolog 3 (avian),MAG — myelin-associated glycoprotein, MOBP — myelin-associated basic protein, MOG — myelin oligodendrocyte glycoprotein, OLIG2 —oligodendrocyte-lineage transcription factor 2, NOGO (RTN4R) — reticulon 4 receptor, PLP1 — proteolipid protein 1.

131S.N. Mitkus et al. / Schizophrenia Research 98 (2008) 129–138

humor fluid, occipital pole and/or urine was conductedfor every control and schizophrenia case. Non-psychi-atric cases with toxicology screenings positive forethanol (blood levels greater than 0.05 g/dl; AmericanMedical Laboratory Inc.) or positive for any over-thecounter or prescription medication above therapeuticlevels or any detectable level of illicit drugs wereexcluded from the control group. Positive toxicologywas not an exclusion criterion for schizophrenic cases.

Table 3Characteristics of subject cohorts used for mRNA expression and protein m

N Race Sex

White matter-mRNAControls 73 43AA/24C/3A/3H 23F/50MSchizophrenia patients 33 20AA/11C/2H 13F/20M

Grey matter-mRNAControls 68 40AA/21C/3A/4H 21F/47MSchizophrenia patients 31 17AA/12C/2H 14F/17M

White matter-proteinControls 47 31AA/13C/2A/1H 13F/34MSchizophrenia patients 26 13AA/11C/2H 9F/17M

AA— African American; A— Asian; H— Hispanic; C— Caucasian; F—RNA integrity number (on a scale 1-10). For age, PMI, pH and RIN the dat⁎ pb0.05.

2.2. RNA extraction and reverse transcription

Tissue was pulverized and stored at −80 °C. TotalRNA was extracted from 300 mg of tissue usingTRIZOL Reagent (Life Technologies Inc., Grand Island,NY, USA). RNA from the DLPFC white matter wasisolated with TRIZOL and purified using RNeasy spincolumns (Qiagen Inc.). The yield of total RNA wasdetermined by absorbance at 260 nm. RNA quality was

easurements

Age PMI pH RIN

41.4±15.1 32.1±14.4 6.6±0.28 6.7±148.8±15.9 ⁎ 38.5±18.5 6.4±0.32 ⁎ 6.2±1.1 ⁎

41.2±14.7 32.3±14.1 6.6±0.25 6.8±1.1946.8±15.3 37.4±17.3 6.4±0.28 ⁎ 6.8±1.13

41.5±13.5 32.4±14.1 6.58±0.25 6.6±144.7±16.3 36.3±13.4 6.51±0.28 6.2±1.3

female; M—male; PMI— postmortem interval (hrs); RIN— Agilenta re presented as mean±SD.

132 S.N. Mitkus et al. / Schizophrenia Research 98 (2008) 129–138

assessed with a high resolution capillary electrophoresis(Agilent Technologies, Palo Alto, CA, USA). TotalRNA (4 μg) was used in 50 μl of reverse transcriptasereaction to synthesize cDNA, by using a SuperScriptFirst-Strand Synthesis System for RT-PCR (Invitrogen,Carlsbad, CA, USA).

2.3. Q-PCR probes and primers

Taqman probes/primer sets from Applied Biosystemswere used for MOBP (Hs00379220_m1), CNP(Hs00263981_m1, which recognizes both CNP isoforms),MAG (Hs00159000_m1) and OLIG2 (Hs00377820_m1)as well as for three endogenous housekeeping genes, beta-actin (Hs99999903_m1), β2-microglobulin (B2M)(Hs99999907_m1) and β-glucuronidase (GUSB)(Hs99999908_m1).

2.4. Quantitative real-time PCR

Expression levels of mRNAs were measured byquantitative real-time PCR, using ABI Assays-on-Demand (see above) and an ABI Prism 7900 sequencedetection system with 384-well format (Applied Bio-systems, Foster City, CA, USA). Each 10 μl reactioncontained 900 nM of primer, 250 nM of probe andTaqman Universal PCR Mastermix (Applied Biosys-tems) containing Hot Goldstar DNA Polymerase,dNTPs with dUTP, uracil-N-glycosylase, passive refer-ence and 100 ng of cDNA template. PCR cycleparameters were 50 °C for 2 min, 95 °C for 10 min,40 cycles of 95 °C for 15 s, and 59 °C or 60 °C for 1 min.PCR data were acquired from the Sequence DetectorSoftware (SDS version 2.0, Applied Biosystems) andquantified by a standard curve method using serialdilutions of pooled cDNA derived from RNA obtainedfrom DLPFC grey or white matter of 10–12 normalcontrol subjects. In each experiment the R2 value of thecurve was more than 0.99, the amplification efficiency96–101% and no-template cDNA controls gave nodetectable signal. All measurements were performed intriplicate and the gene expression levels calculated as anaverage of the three samples. Data analysis was basedon normalization of target mRNAs to three endogenouscontrol genes (a geometric mean of three genes=nor-malizing factor, NF).

2.5. Immunoblotting

Human DLPFC white matter tissue samples (100–130 mg; 1 g tissue: 10 ml buffer) were homogenized in aprotease inhibitor–Tris-glycerol extraction buffer (AEBSF

0.024%, aprotinin 0.005%, leupeptin 0.001%, pepstatin A0.001%, glycerol 50%, Tris 0.6%). Protein concentrationwas determined using the Bradford assay. Homogenates(4 μg of total protein/10 μl) were diluted with water, XTsample buffer and 1X antioxidant (Bio-Rad LaboratoriesInc, Hercules, CA), and heated to 80 °C for 5min. Tenμl ofsample was loaded onto precast 10% Bis–Tris polyacryl-amide gels (Bio-Rad Laboratories Inc, Hercules, CA), andproteins were separated by electrophoresis at 200 V for onehour. Each gel contained a molecular weight marker ladderSeeBlue Plus 2 (Invitrogen, Carlsbad, CA), a pooledsample from 10 normal controls at three concentrations (2,4, 6μg of total protein content per 10 μl), and samples frompatients with schizophrenia and controls. A total of fourseparate gels were used in one experiment (n=73 samples).Gels were transferred onto nitrocellulose membranes at85 V for thirty minutes, membranes blocked for an hour in10% goat serum in Tris buffered saline (TBS) with 0.1%Tween-20 (TBS-T), and incubatedwith primary antibodies.Anti-CNPantibody (Chemicon International Inc, TemeculaCA, 1:10,000 dilution), anti-MOBP (Santa Cruz Biotech-nology Inc., Santa Cruz, CA, 1: 2000 dilution) and anti-GAPDH (Chemicon International Inc, TemeculaCA,1:1,000 dilution) and anti-actin (Chemicon Interna-tional Inc, Temecula CA, 1:10,000 dilution) antibodieswere used. Blots were rinsed in TBS-T, incubated in aperoxidase-conjugated mouse secondary antibody(1:20,000 dilution, Chemicon International Inc, Temecula,CA) for 2 h in 10% normal goat serum in TBS-T, rinsed inTBS-T and developed in ECL-plus (Amersham). Thevalues were expressed as ratios to housekeeping proteins.

2.6. CNP and OLIG2 genotype determination

DNA for genotyping was extracted from cerebellartissue using standard methods. Based on previousreports of association, four SNPs in CNP (rs2070106,rs11079028, rs11296, rs4796751) and four in OLIG2(rs2834070, rs762178, rs1059004, rs9653711) weregenotyped in all subjects. Only the exonic CNP SNPrs2070106 has been associated with schizophrenia andwith reduced expression of CNP mRNA in a Caucasianpopulation (Peirce et al., 2006). All four OLIG2 SNPschosen for genotyping have been previously associatedwith schizophrenia (Georgieva et al., 2006). Each PCRreaction contained 10 ng of DNA, 1 μM of each primer,100 nM of each probe and 5 μl of Taqman UniversalPCR Mastermix (Applied Biosystems) containingAmpliTaq Gold DNA polymerase, AmpErase UNG,dNTPs, with dUTP passive reference and optimizedbuffer components. The PCR cycling conditions were at50 °C for 2 min, 95 °C for 10 min, 40 cycles of 95 °C for

133S.N. Mitkus et al. / Schizophrenia Research 98 (2008) 129–138

15 s, and 60 °C for 1 min. Genotype reproducibility wasroutinely assessed by re-genotyping all samples forselected SNPs and was generally N99%.

2.7. Statistical analyses

Statistical analyses were conducted using Statisticaversion 7.1. www.statsoft.com. Multiple regressionanalyses were used for determining the contribution ofdemographic, tissue- and disease-related variables togene expression levels. Comparisons between diagnos-tic groups were made using ANCOVA for each mRNAwith diagnosis as an independent variable and poten-tially confounding variables as covariates. The samethree genes (B2M, β-actin and GUSB) were used fornormalization of the data in the white and grey matter ofthe DLPFC. For immunoreactivity, the data werenormalized to actin or GAPDH. Effects of genotypeon gene expression were examined using ANCOVAwith genotype and race or diagnosis as independentvariables. Effects of race were restricted to analyses ofAfrican American and Caucasian individuals due to thesmall sample size in other ethnic groups. For all SNPs

Table 4Contribution of demographic and tissue-related variables to non-normalizeddetermined using forward stepwise multiple regression analysis

DX Age Sex

Adj R2 β β β

White matterMOBP 0.229 n.s. −0.25⁎⁎CNP 0.256 n.s. −0.29⁎⁎MAG 0.197 n.s. −0.26⁎⁎OLIG2 0.116 n.s. −0.23⁎NF 0.234 n.s. n.s. n.s.MOBP/NF 0.127 0.225⁎ n.s. n.s.CNP/NF 0.044 n.s. n.s. n.s.MAG/NF 0.029 n.s. n.s. n.s.OLIG2/NF 0.001 n.s. n.s. n.s.

Grey matterMOBP 0.038 n.s. n.s. n.s.CNP 0.058 n.s. n.s. n.s.MAG 0.061 n.s. n.s. n.s.OLIG2 0.235 n.s. n.s. n.s.NF 0.145 n.s. n.s. n.s.MOBP/NF 0.009 n.s. n.s. n.s.CNP/NF 0.007 n.s. n.s. n.s.MAG/NF 0.04 n.s. n.s. n.s.OLIG2/NF n.s. n.s. n.s.

Abbreviations: adj R2 — adjusted coefficient of determination; β — standardDx— diagnosis (1— controls, 2— schizophrenia patients); Sex— 1: femaleSm — smoking history (1: No, 2: Yes); NF — normalizing factor.Other abbreviations as in Tables 1 and 2.⁎pb0.05, ⁎⁎pb0.01, ⁎⁎⁎pb0.001, ⁎⁎⁎⁎pb0.0001; n.s. — not significant.

examined in this study, the observed genotype frequen-cies were within expected distribution according toHardy–Weinberg equilibrium (evaluated by the chi-squared test, p values N0.05). Because for some DNAsamples, an unequivocal assignment of genotype wasnot possible, the subjects with missing genotypes werenot included in the genotypic analysis. In the entiresample, we eliminated outliers (pointsN±2 SD from themean) for each expression variable (n=0–3).

3. Results

3.1. Analysis of mRNA expression

RNA quality (RIN) significantly influenced the non-normalized expression levels of all four genes, MOBP,CNP, MAG and OLIG2 in the white matter as well as thenormalizing factors (NF) in the white and grey matter ofthe DLPFC (all p values b0.01, Table 4). Importantly,normalization of the data appeared to correct for RNAquality as there was no significant effect of RIN onnormalized levels of any of the target genes. PMI and pHhad a lesser impact, weakly affecting only raw expression

and normalized gene expression levels in schizophrenics and controls

pH PMI RIN Sm AS

β β β β β

n.s. n.s. 0.432⁎⁎⁎⁎ n.s. n.s.n.s. n.s. 0.405⁎⁎⁎⁎ n.s. n.s.n.s. n.s. 0.379⁎⁎⁎⁎ n.s. n.s.n.s. n.s. 0.279⁎⁎ n.s. n.s.n.s. n.s. 0.445⁎⁎⁎⁎ n.s. n.s.n.s. 0.224⁎ n.s. n.s. n.s.n.s. n.s. n.s. n.s. n.s.n.s. n.s. n.s. n.s. n.s.n.s. n.s. n.s. n.s. n.s.

n.s. n.s. n.s. n.s. n.s.n.s. n.s. n.s. n.s. n.s.n.s. −0.21⁎ n.s. n.s. n.s.0.287⁎⁎ −0.186⁎ n.s. n.s. n.s.n.s. −0.227⁎ 0.273⁎⁎ n.s. n.s.n.s. n.s. n.s. n.s. n.s.n.s. n.s. n.s. n.s. n.s.n.s. n.s. n.s. n.s. n.s.n.s. n.s. n.s. n.s. n.s.

ized regression coefficient; AS— agonal state (1=b10 min; 2=N1 h), 2: male; PMI— postmortem interval; RIN— RNA integrity number;

134 S.N. Mitkus et al. / Schizophrenia Research 98 (2008) 129–138

levels of MAG and OLIG2 in the grey matter. AlthoughPMI did not affect expression of non-normalized MOBPlevels or the NF, it was found to be a significant variablein the normalized MOBP expression in the white matter(p=0.02). Increasing age predicted lower expression ofall four genes in the DLPFC white matter (Table 4). Noother tissue-related or demographic parameters had asignificant effect on the expression of myelin-relatedgenes in the DLPFC.

MOBP mRNA expression in the white matter of theDLPFC was modestly increased in patients withschizophrenia as compared with controls (by 16%, t-value=2.42, p=0.017, Fig. 1). This difference remainedsignificant after the data were co-varied by age, sex,smoking, agonal state, pH, PMI andRNAquality (F(1,93)=3.84, p=0.03). To check whether an increase in MOBPmRNA in patients with schizophrenia was related to otherfactors than the disease itself, such as antipsychoticmedication or drug abuse, we performed multiple regres-sion analysis, which revealed that a history of substanceabuse significantly predicted MOBP expression levels inthe DLPFC white matter of schizophrenia patients(standardized regression beta coefficient = 0.497,p=0.025, Fig. 2). Further analysis revealed that toxicolo-gy-positive (n=8) and toxicology-negative (n=7) schizo-phrenia patients, who all had history of substance abuse, didnot differ in MOBP mRNA levels (p=0.42, data notshown), suggesting that history-based drug abuse was themain determinant of high MOBP expression in this study.Other parameters included in the analysis (age at the onsetof the disease, age at first hospitalization, duration of illness,estimated daily, life and last dose of neuroleptics) had no

Fig. 1. Expression of normalized MOBP mRNA levels in schizo-phrenic patients and controls in the dorsolateral prefrontal cortex(DLPFC) white matter (WM) in 70 controls and 32 patients withschizophrenia and in the DLPFC grey matter (GM) in 64 controls and30 patients with schizophrenia. Values are expressed as percentage ofcontrols (MOBP % Controls, mean±SE). ⁎ statistically significantfrom controls, pb0.05.

effect onMOBP expression in patients. In the DLPFC greymatter, there was no difference in MOBP mRNA levelsbetween patients with schizophrenia and normal controls(Fig. 1) and, in contrast to the white matter, there was noeffect of substance abuse. Expression of other genes, CNP,MAG andOLIG2, in the DLPFCwhite and greymatter didnot show significant differences between patients withschizophrenia and unaffected controls either without (tvalues b1.5, p values N0.1) or with co-varying forpotentially confounding factors, including substanceabuse (F values b1.2, p values N0.2).

3.2. Protein analysis

We performed protein analysis only in DLPFC whitematter. Immunoblotting produced a prominent 12 kDband for MOBP protein (Fig. 3) and a 46–48 kD doubleband for CNP, corresponding to twoCNP isoforms, CNP1(46 kD) and CNP2 (48 kD) (Fig. 3). B-Actin antibodywhich produced a 43 kD band and GAPDH antibody,which gave a 36 kD band corresponding to the reducedmonomer of GAPDH, were used as loading controls forMOBP and CNP, respectively (Fig. 3). No statisticallysignificant differences in MOBP (t-value=0.01, p=0.7)or CNP (t-value=0.01, p=0.6) immunoreactivities werefound between patients with schizophrenia and controls(Fig. 3). Co-varying for potentially confounding factorsalso did not reveal differences between the diagnosticgroups (F values b1.0, p values N0.5). Furthermore,substance abuse did not have a significant effect onMOBP or CNP protein expression in patients (F valuesb1.0, p values N0.5).

3.3. Genotype effects

For CNP, ANCOVAwith genotype and race/ethnicityas independent variables and RIN, pH and age ascontinuous predictors showed a significant effect of asynonymous exonic SNP rs2070106 on CNP geneexpression in the DLPFC grey matter (F=4.7, p=0.03).There was no genotype by race interaction (F=1.7,p=0.2) and no race effect (F=0.2, p=0.6). However,consistent with the finding of Peirce et al. (2006),Caucasian individuals carrying the risk-associated allele(A) had significantly lower levels of CNP mRNAcompared with Caucasian subjects homozygous for theG allele (by 25%, p=0.02). There was no difference inexpression between these two alleles in AfricanAmericansubjects (p=0.5, Fig. 4). A-carriers (i.e., A/G and A/Asubjects) were combined into one group, as there wereonly two subjects homozygous for anA allele (one in eachracial group), and compared with G/G individuals. There

Fig. 2. Effect of a history of substance abuse on normalized MOBP mRNA expression in the dorsolateral prefrontal cortex A. white matter (WM) in70 controls (Controls), 17 patients with schizophrenia without history of substance abuse (Sch without SubAb), and 15 patients with schizophreniawith history of substance abuse (Sch with SubAb) and in B. grey matter (GM), controls n=64, schizophrenia without substance abuse n=16,schizophrenia with substance abuse n=14. None of the controls had a history of substance abuse. ⁎ statistically significant from schizophreniapatients without substance abuse, pb0.05.

135S.N. Mitkus et al. / Schizophrenia Research 98 (2008) 129–138

was no effect of genotype on CNP expression in whitematter. No other CNP control SNP examined here(negative in previous association studies) showed aneffect on expression in either grey or white matter of the

DLPFC (all F values b1.5, all p values N0.3). Moreover,there was no effect of any genotype, including the high-risk SNP rs2070106, on CNP immunoreactivity (all Fvalues b1.5, all p values N0.3).

Fig. 3. MOBP and CNP protein immunoreactivity in DLPFC whitematter in controls (n=47) and in patients with schizophrenia (n=26).Immunobloting was performed using antibodies against MOBP and aloading control Actin as well as CNP and a loading control GAPDH.MOBP and CNP immunoreactivity in controls and patients withschizophrenia were normalized to control proteins (Actin andGAPDH, respectively) and expressed as a percentage of controls(mean±SE). There were no differences in immunoreactivity betweenthe diagnostic groups.

Fig. 5. Effect of OLIG2 genotype (SNP rs1059004) on OLIG2 mRNAexpression in the dorsolateral prefrontal cortex white matter. Individualshomozygous and for an A allele and heterozygotes had significantlylower OLIG2 mRNA expression than subjects homozygous for theC allele, “C/C” n=37, “A/C” n=40, “A/A” n=14. (mean±SE),⁎pb0.01.

136 S.N. Mitkus et al. / Schizophrenia Research 98 (2008) 129–138

Analysis of the effects of OLIG2 genotypes on geneexpression revealed that two OLIG2 SNPs, rs1059004and rs9653711, both previously associated with schizo-phrenia (Georgieva et al., 2006), predicted lower OLIG2mRNA expression levels in the DLPFC white matter(main effects of genotype: F(2, 85)=5.19, p=0.007 andF(2,86)=4.33, p=0.01, respectively). Post-hoc compar-isons showed that individuals carrying the minor Aallele (enriched in cases according to Georgieva et al.,2006) at SNP rs1059004 had lower OLIG2 mRNAlevels than subjects homozygous for the C allele (A/A

Fig. 4. Effect of race and CNP genotype (rs2070106) on normalizedCNP mRNA expression in the dorsolateral prefrontal cortex greymatter (“African American G/G” n=39, “African American A-carriers” n=11, “Caucasian G/G” n=15, “Caucasian A-carriers”n=13). Caucasian individuals carrying a high-risk A allele showedsignificantly lower levels of CNP expression compared to Caucasiansubjects homozygous for the G allele. (mean±SE) ⁎ statisticallysignificant from Caucasian G/G, pb0.05.

and A/C groups different from C/C, pb0.01) (Fig. 5).Also, the minor C allele at SNP rs9653711, which wasshown to be enriched in cases in the same study, wasfound to be associated with decreased OLIG2 expres-sion as compared with the G allele (C/C and C/G groupsdifferent from G/G, pb0.01, data not shown). This SNPwas in perfect linkage disequilibrium (R2 =1) with SNPrs1059004 in Caucasians, so these results, beingrepetitions of the same genetic analysis, are notindependent. Even so, the data for either SNP remainedsignificant after correcting for multiple comparisons ofall typed SNPs within this gene (i.e., Bonferroni-corrected p values b0.0125). In two-way ANOVAswith genotype (rs1059004 and rs9653711) and race ordiagnosis, there were no effects of race or diagnosis andno significant interactions between genotype and race orgenotype and diagnosis (all F values b1.0, all p valuesN0.5). Furthermore, there were no effects of two otherpreviously associated OLIG2 SNPs in either the grey orwhite matter of the DLPFC.

4. Discussion

We investigated the expression of myelin-associatedgenes in post-mortem human DLPFC. Based on theresults of prior studies, we hypothesized that reducedlevels of CNP,MAGandOLIG2 (but perhaps notMOBP)will be detected in our large cohort of patients. Contrary toour hypothesis, we did not find changes in CNP, MAG orOLIG2 expression in patients with schizophrenia but asmall increase in MOBP mRNA levels in the DLPFCwhite matter, which appeared to be related to co-morbidsubstance abuse in patients. The inconsistency betweenour study and previous reports may be related to the effectsize of myelin gene expression within the population ordifferences in cohort characteristics. The results from our

137S.N. Mitkus et al. / Schizophrenia Research 98 (2008) 129–138

genotype analyses suggest that cohort effects may indeedbe important; in particular, genetic background appears tobe a significant factor contributing to the expression levelsof at least two such genes, CNP and OLIG2. Otherpossibilities include differences between cohorts in age atdeath (e.g., the Mount Sinai collection contains mucholder subjects than our collection), race and the exposureto environmental factors, including substance abuse andmedication, as all these factors are likely to affectoligodendrocyte-associated gene expression.

Substance abuse in our group of patients consistedprimarily of alcohol, THC, PCP, heroin and/or cocaineand in most cases was a combination of these substances.Dysregulation of myelin-related genes has been previ-ously observed in cocaine abusers and alcoholics(Albertson et al., 2004; Liu et al., 2004). In an imagingstudy, significant increases in temporal and occipitalwhite matter volume have been found in methamphet-amine users, suggesting possible increased myelinationin substance abusers (Thompson et al., 2004). Impor-tantly in this context, our normal controls werethoroughly screened for history of substance abuse andthose subjects with a history of drug abuse or positivetoxicological results were not included in the collection.

Although a study inmice suggested that antipsychoticscan reduce expression of myelin-related genes (Narayanet al., 2007), we did not find a reduction in the expressionalthough all our patients were medicated. Moreover, wehave not detected correlations between the expressionlevels and the doses of neuroleptics (with a caveat thatneuroleptic dose estimates are based on medical recordsand thus may not necessarily reflect actual drug intake).

The most interesting finding of this study is that allelicvariation in CNP and OLIG2, both implicated assusceptibility genes in schizophrenia in previous studies,predicted lower expression of these genes in the DLPFCgrey and white matter, respectively. For CNP, we found asimilar effect of SNP rs2070106 on CNP expression aspreviously reported (Peirce et al., 2006) (24% vs 25%reduction in our study), although a different method wasused. There are several caveats, however. First, all threeSNPs showing an effect on mRNA expression do notaffect amino acid sequence and are not localized toknown promoter domains (CNP rs2070106 was asynonymous exonic SNP and two OLIG2 SNPs wereintronic), and thus it is unclear how they might affectgene expression levels. Synonymous SNPs can, however,alter expression by modulating mRNA secondarystructure and thus its stability (Nackley et al., 2006),and polymorphisms in introns can affect gene function byaffecting regulatory motifs within introns or splicing.Nevertheless, we have no evidence yet that any of the

SNPs showing association with mRNA levels act throughthese mechanisms. It is also possible that these SNPs maymonitor other undiscovered SNPs involved with regula-tion of expression or splicing (i.e., they may tag nearbyfunctional SNPs). Second, these SNPs had similar effectsin both diagnostic groups (as there was no main effect ofdiagnosis and no overall reduction of expression inpatients). We could not confirm association of theseSNPs with schizophrenia, but as we lacked power toconduct such a study in the postmortem collection, this isnot surprising. It is possible, however, that if the geneticcontribution to the expression of these genes is small andcohort-specific, it may not be readily detectable,especially in the face of multiple antemortem andpostmortem confounding factors, which are difficult toavoid in human molecular studies. Finally, it is unclearwhy CNP expression was associated with a risk SNP inthe DLPFC grey matter whereas the effects of SNPs onOLIG2 expression were detected in white matter.

Role of funding sourceFunding for this studywas provided by Intramural Research Program

(IRP), NIH, NIMH. The NIMHhad no further role in study design, in thecollection, analysis and interpretation of data, in the writing of the reportand in the decision to submit the paper for publication.

ContributorsEach of the below contributors has made a substantial contribution

to the experimental study and the preparation of the manuscript andhave read and approved the manuscript for submission.

Drs. Lipska, Mitkus, Weinberger and Kleinman were involvedwith the study design, analysis and manuscript preparations.

Drs. Hyde and Kleinman were involved with procurement of braintissue, postmortem clinical characterization and manuscript revisions.

Drs. Vakkalanka and Kolachana provided genotype data.

Conflict of interestAll authors declare that they have no conflicts of interest with

regard to the work presented in this report.

AcknowledgmentsWe would like to thank Dr Llewellyn L. Bigelow and Mrs. Amy

Deep-Soboslay for their contribution to procurement of brain tissueand postmortem clinical characterization of subjects and Mrs. YevaSnitkovsky, and Mr. E. Michael Saylor for their excellent technicalassistance and Mr. Tianzhang Ye for the generation and managementof the database. This research was supported by the IntramuralResearch Program of the National Institutes of Health, NIMH.

References

Aberg, K., Saetre, P., Jareborg, N., Jazin, E., 2006. Human QKI, apotential regulator of mRNA expression of human oligodendro-cyte-related genes involved in schizophrenia. Proc. Natl. Acad.Sci. U.S.A. 103, 7482–7487.

Albertson, D.N., Pruetz, B., Schmidt, C.J., Kuhn, D.M., Kapatos, G.,Bannon, M.J., 2004. Gene expression profile of the nucleusaccumbens of human cocaine abusers: evidence for dysregulationof myelin. J. Neurochem. 88, 1211–1219.

138 S.N. Mitkus et al. / Schizophrenia Research 98 (2008) 129–138

Aston, C., Jiang, L., Sokolov, B.P., 2004. Microarray analysis ofpostmortem temporal cortex from patients with schizophrenia.J. Neurosci. Res. 77, 858–866.

Covault, J., Lee, J., Jensen, K., Kranzler, H., 2004. Nogo 3′-untranslatedregion CAA insertion: failure to replicate association with schizo-phrenia and demonstration of marked population difference infrequency of the insertion. Brain Res. Mol. Brain Res. 120, 197–200.

Davis, K.L., Haroutunian, V., 2003. Global expression-profilingstudies and oligodendrocyte dysfunction in schizophrenia andbipolar disorder. Lancet 362, 758.

Davis, K.L., Stewart, D.G., Friedman, J.I., et al., 2003. White matterchanges in schizophrenia: evidence for myelin-related dysfunction.Arch. Gen. Psychiatry 60, 443–456.

Dracheva, S., Davis, K.L., Chin, B., Woo, D.A., Schmeidler, J.,Haroutunian, V., 2006. Myelin-associated mRNA and proteinexpression deficits in the anterior cingulate cortex and hippocam-pus in elderly schizophrenia patients. Neurobiol. Dis. 21, 531–540.

Dwork, A.J., Mancevski, B., Rosoklija, G., 2007. White matter andcognitive function in schizophrenia. Int. J. Neuropsychopharma-col. 1–24.

Flynn, S.W., Lang, D.J., Mackay, A.L., et al., 2003. Abnormalities ofmyelination in schizophrenia detected in vivo with MRI, and post-mortem with analysis of oligodendrocyte proteins. Mol. Psychiatry8, 811–820.

Georgieva, L., Moskvina, V., Peirce, T., et al., 2006. Convergentevidence that oligodendrocyte lineage transcription factor 2(OLIG2) and interacting genes influence susceptibility to schizo-phrenia. Proc. Natl. Acad. Sci. U. S. A. 103, 12469–12474.

Gregorio, S.P., Mury, F.B., Ojopi, E.B., et al., 2005. Nogo CAA 3′UTR Insertion polymorphism is not associated with Schizophrenianor with bipolar disorder. Schizophr. Res. 75, 5–9.

Hakak, Y., Walker, J.R., Li, C., et al., 2001. Genome-wide expressionanalysis reveals dysregulation of myelination-related genes inchronic schizophrenia. Proc. Natl. Acad. Sci. U.S.A. 98, 4746–4751.

Hof, P.R., Haroutunian, V., Friedrich Jr., V.L., et al., 2003. Loss andaltered spatial distribution of oligodendrocytes in the superiorfrontal gyrus in schizophrenia. Biol. Psychiatry 53, 1075–1085.

Kanazawa, T., Glatt, S.J., Tsutsumi, A., et al., 2007. Schizophrenia isnot associated with the functional candidate gene ERBB3: resultsfrom a case-control study. Am. J. Med. Genet. B. Neuropsychiatr.Genet. 144, 113–116.

Katsel, P., Davis, K.L., Haroutunian, V., 2005. Variations in myelinand oligodendrocyte-related gene expression across multiple brainregions in schizophrenia: a gene ontology study. Schizophr. Res.79, 157–173.

Kubicki, M., McCarley, R.W., Shenton, M.E., 2005. Evidence forwhite matter abnormalities in schizophrenia. Curr. Opin. Psychi-atry 18, 121–134.

Lipska, B.K., Deep-Soboslay, A., Weickert, C.S., et al., 2006. Criticalfactors in gene expression in postmortem human brain: focus onstudies in schizophrenia. Biol. Psychiatry 60, 650–658.

Liu, J., Lewohl, J.M., Dodd, P.R., Randall, P.K., Harris, R.A.,Mayfield, R.D., 2004. Gene expression profiling of individualcases reveals consistent transcriptional changes in alcoholic humanbrain. J. Neurochem. 90, 1050–1058.

Liu, X., Qin, W., He, G., et al., 2005. A family-based association studyof the MOG gene with schizophrenia in the Chinese population.Schizophr. Res. 73, 275–280.

McCullumsmith, R.E., Gupta, D., Beneyto,M., et al., 2007. Expressionof transcripts for myelination-related genes in the anterior cingulatecortex in schizophrenia. Schizophr. Res. 90, 15–27.

McInnes, L.A., Lauriat, T.L., 2006. RNA metabolism and dysmyeli-nation in schizophrenia. Neurosci. Biobehav. Rev. 30, 551–561.

Nackley, A.G., Shabalina, S.A., Tchivileva, I.E., et al., 2006. Humancatechol-O-methyltransferase haplotypes modulate protein expres-sion by altering mRNA secondary structure. Science 314,1930–1933.

Narayan, S., Kass, K.E., Thomas, E.A., 2007. Chronic haloperidoltreatment results in a decrease in the expression of myelin/oligodendrocyte-related genes in the mouse brain. J. Neurosci.Res. 85, 757–765.

Novak, G., Tallerico, T., 2006. Nogo A, B and C expression inschizophrenia, depression and bipolar frontal cortex, and correla-tion of Nogo expression with CAA/TATC polymorphism in 3′-UTR. Brain. Res. 1120, 161–171.

Peirce, T.R., Bray, N.J., Williams, N.M., et al., 2006. Convergentevidence for 2′,3′-cyclic nucleotide 3′-phosphodiesterase as apossible susceptibility gene for schizophrenia. Arch. Gen.Psychiatry 63, 18–24.

Prabakaran S., Swatton J.E., Ryan M.M., et al., (2004): Mitochondrialdysfunction in schizophrenia: evidence for compromised brainmetabolism and oxidative stress. Mol Psychiatry 9:684–697, 643.

Qin, W., Gao, J., Xing, Q., et al., 2005. A family-based associationstudy of PLP1 and schizophrenia. Neurosci. Lett. 375, 207–210.

Rajkowska, G., Goldman-Rakic, P.S., 1995. Cytoarchitectonic defini-tion of prefrontal areas in the normal human cortex: II. Variabilityin locations of areas 9 and 46 and relationship to the TalairachCoordinate System. Cereb. Cortex 5, 323–337.

Tan, E.C., Chong, S.A., Wang, H., Chew-Ping Lim, E., Teo, Y.Y.,2005. Gender-specific association of insertion/deletion poly-morphisms in the nogo gene and chronic schizophrenia. BrainRes. Mol. Brain Res. 139, 212–216.

Tang, F., Qu, M., Wang, L., et al., 2007. Case-control association studyof the 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNP) geneand schizophrenia in the Han Chinese population. Neurosci. Lett.416, 113–116.

Thompson, P.M., Hayashi, K.M., Simon, S.L., et al., 2004. Structuralabnormalities in the brains of human subjects who use metham-phetamine. J. Neurosci. 24, 6028–6036.

Tkachev, D., Mimmack, M.L., Ryan, M.M., et al., 2003. Oligoden-drocyte dysfunction in schizophrenia and bipolar disorder. Lancet362, 798–805.

Usui, H., Takahashi, N., Saito, S., et al., 2006. The 2′,3′-cyclic nucleotide3′-phosphodiesterase and oligodendrocyte lineage transcriptionfactor 2 genes do not appear to be associated with schizophrenia inthe Japanese population. Schizophr. Res. 88, 245–250.

Wan, C., Yang, Y., Feng, G., et al., 2005. Polymorphisms of myelin-associated glycoprotein gene are associated with schizophrenia inthe Chinese Han population. Neurosci. Lett. 388, 126–131.

Watanabe, Y., Fukui, N., Nunokawa, A., et al., 2007. No associationbetween the ERBB3 gene and schizophrenia in a Japanesepopulation. Neurosci. Res. 57, 574–578.

Xiong, L., Rouleau, G.A., Delisi, L.E., et al., 2005. CAA insertionpolymorphism in the 3′UTR of Nogo gene on 2p14 is notassociated with schizophrenia. Brain Res. Mol. Brain Res. 133,153–156.

Yang, Y.F., Qin, W., Shugart, Y.Y., et al., 2005. Possible association ofthe MAG locus with schizophrenia in a Chinese Han cohort offamily trios. Schizophr. Res. 75, 11–19.

Zai, G., King, N., Wigg, K., et al., 2005. Genetic study of the myelinoligodendrocyte glycoprotein (MOG) gene in schizophrenia.Genes Brain Behav. 4, 2–9.