Human Molecular Genetics

doi:10.1093/hmg/ddm343 17:717-723, 2008. First published 30 Nov 2007; Hum. Mol. Genet.

L. Murphy Jens R. Wendland, Pablo R. Moya, Matthew R. Kruse, Renee F. Ren-Patterson, Catherine L. Jensen, Kiara R. Timpano and Dennis

associates with obsessive-compulsive disorderSLC6A4A novel, putative gain-of-function haplotype at

http://hmg.oxfordjournals.org/cgi/content/full/17/5/717The full text of this article, along with updated information and services is available online at

References http://hmg.oxfordjournals.org/cgi/content/full/17/5/717#BIBL

This article cites 28 references, 8 of which can be accessed free at

Reprints http://www.oxfordjournals.org/corporate_services/reprints.html

Reprints of this article can be ordered at

Email and RSS alerting Sign up for email alerts, and subscribe to this journal’s RSS feeds at http://hmg.oxfordjournals.org

image downloadsPowerPoint® Images from this journal can be downloaded with one click as a PowerPoint slide.

Journal informationat http://hmg.oxfordjournals.org Additional information about Human Molecular Genetics, including how to subscribe can be found

Published on behalf ofhttp://www.oxfordjournals.org Oxford University Press

 

at Columbia University Libraries on 22 September 2008 http://hmg.oxfordjournals.orgDownloaded from

A novel, putative gain-of-function haplotypeat SLC6A4 associates withobsessive-compulsive disorder

Jens R. Wendland1,{,�, Pablo R. Moya1,{, Matthew R. Kruse1, Renee F. Ren-Patterson2,

Catherine L. Jensen1, Kiara R. Timpano3 and Dennis L. Murphy1

1Laboratory of Clinical Science and 2Clinical Brain Disorders Branch, National Institute of Mental Health, National

Institutes of Health, Bethesda, MD, USA and 3Department of Psychology, Florida State University, Tallahassee, FL

32306, USA

Received August 2, 2007; Revised and Accepted November 22, 2007

Obsessive-compulsive disorder (OCD) is a disabling neuropsychiatric illness with strong segregation dataindicative of major genetic contributions. Association analyses of common functional variants of theserotonin transporter gene (SLC6A4), a long-standing OCD candidate, have so far been inconsistent. Here,we set out to investigate the role of additional functional SLC6A4 loci in OCD. We describe a common, func-tional C > T single nucleotide polymorphism, rs25532, located less than 150 nucleotides centromeric of theserotonin transporter-linked polymorphic region indel known as 5-HTTLPR. The minor allele of rs25532 sig-nificantly decreased luciferase reporter gene expression levels by 15–80%, depending on 5-HTTLPR allelebackground and cell type. Haplotype-based testing of rs25532 and all other known non-coding functionalSLC6A4 variants revealed a highly significant omnibus association with OCD in a large case–controlsample. Remarkably, the haplotype significantly overrepresented in probands contained the higher-expres-sing allele at each locus, supporting the notion of increased serotonin transporter functioning being patho-genetically involved in OCD. Conditional haplotype analyses with the software WHAP revealed that thisassociation is primarily driven by 5-HTTLPR, rs25532 and rs16965628. Our results contribute to a betterunderstanding of SLC6A4 expression genetics and provide a functional haplotype framework for futureserotonin-related studies.

INTRODUCTION

Altered functioning of the serotonin transporter (SERT, 5-HTT,SLC6A4) has long been postulated to be involved in the patho-genesis of obsessive-compulsive disorder (OCD). Support forthis hypothesis stems from the therapeutic efficacy of serotoninre-uptake inhibitors in OCD (1,2) and from the recent identi-fication of a rare gain-of-function missense mutation ofSLC6A4, I425V (3) (termed ‘OCD1’ in Online MendelianInheritance in Man), which results in a constitutive transporteractivation (4,5) and a primarily OCD-like phenotype (6,7).

Non-coding functional polymorphisms of SLC6A4, in par-ticular the serotonin transporter-linked polymorphic region,

5-HTTLPR, have been extensively investigated in OCD.While two initial studies (8,9) observed an association withthe greater-expressing L allele and LL genotype, a numberof reports from other groups were inconclusive or negative(10–15). Recently, Hu et al. (16) demonstrated that a possibleexplanation for the lack of consistent association was theuntil-then unappreciated presence of a single nucleotide poly-morphism (SNP), rs25531, within the repetitive region thatcomprises the 5-HTTLPR. The minor allele (G) of rs25531is almost always in phase with the long (L) allele of the5-HTTLPR and attenuates the initially reported (17)gain-of-function relative to the short (S) allele. Thus, modu-lation of 5-HTTLPR by rs25531 results in three common

†The authors wish it to be known that, in their opinion, the first two authors should be regarded as joint First Authors.

�To whom correspondence should be addressed at: Laboratory of Clinical Science, National Institute of Mental Health, NIH, Building 10, Room 3D41,10 Center Dr, MSC 1264, Bethesda, MD 20892, USA. Tel: þ1 3015940219; Fax: þ1 3014020188; Email: wendlandj@mail.nih.gov

Published by Oxford University Press 2007

Human Molecular Genetics, 2008, Vol. 17, No. 5 717–723doi:10.1093/hmg/ddm343Advance Access published on November 30, 2007

alleles: LA (highest-expressing), LG and S (both low-expressing). With this refined understanding of 5-HTTLPRfunctionality, Hu et al. (16) found an association in twoindependent samples, both case–control and family-based,of the higher-expressing LA allele and LALA genotype withOCD. In an immediate replication attempt using a largecase–control design, however, we were unable to corroboratethis finding with statistical significance surviving correctionfor multiple testing, although we also observed an increasedfrequency of the LA allele and LALA genotype in OCD pro-bands (18). This prompted us to determine whether additionalfunctional variation of SLC6A4 could have contributed to thislack of replication. Specifically, we pursued screening of theentire 5-HTTLPR for additional common and possibly func-tional variants and also sought to analyze two recently-described SNPs associated with allelic expression imbalanceof SLC6A4 in lymphoblastoid cell lines, rs2020933 andrs16965628 (19). We describe a heretofore-uninvestigatedSNP, rs25532, which we found to modulate functionality ofthe 5-HTTLPR and which is part of a presumedgain-of-function haplotype associated with OCD.

RESULTS

As an initial step, we re-sequenced 40 SS and 40 LL5-HTTLPR individuals from our OCD and control samplesfor the entire repetitive region comprising the 5-HTTLPR.We observed a common C . T polymorphism in 20 out of80 S alleles and in 5 out of 80 L alleles, that is known todbSNP as rs25532. This SNP is located 176 bp (L allele) or133 bp (S allele) centromeric of rs25531 in the third-lastrepetitive element of the 5-HTTLPR which Nakamura et al.(20) termed the m element (Fig. 1). Consistent with their sys-tematic nomenclature, we termed these alleles 14-E (S allelewith A at rs25531 and T at 25532) and 16-G (L allele withA at rs25531 and T at 25532). We submitted the sequencesof these two novel alleles to the National Center for Biotech-nology Information’s GenBank database as accession numbersEU035982 (14-E) and EU035981 (16-G). We did not observethe rs25532 T allele on the SG or LG background. In the fol-lowing, to facilitate readability and in line with the nomencla-ture introduced by Hu et al. (16), we will refer to the alleles orhaplotypes at 5-HTTLPR studied herein as LAC (16-A withNakamura et al.’s nomenclature), LAT (16-G), LG (16-D),SAC (14-A) and SAT (14-E), where the capital letter refers tothe 43 bp indel, the first subscript letter to rs25531 and thesecond to rs25532. Further, we will refer to the molecular hap-lotype of the 43 bp indel and rs25531 only as ‘triallelic5-HTTLPR’ with the common alleles S, LA and LG as our gen-otyping method for these two loci is phase-certain (21).

We generated four reporter gene constructs to determinewhether rs25532 modulates 5-HTTLPR-conferred expressionlevels in vitro. As can be seen in Figure 2, expression of boththe S and L constructs was significantly attenuated by theT allele of rs25532 in RN46A, PC12 and JAR cell lines. Theeffect size of the T allele at rs25532 differed between S and Lalleles and between cell lines: the mean normalized firefly/Renilla ratio was reduced by 15–30% in S constructs (SAT rela-tive to SAC) and by 25–80% in L constructs (LAT relative to LAC).

We then genotyped 295 Caucasian OCD probands and 657ethnically matched control individuals for rs25532 butobserved no significant allelic association of this locus whentested individually (Table 2, bottom line). Next, we used thesoftware WHAP: haplotype-based association analysis (22)to analyze rs25532 as part of a haplotype with the triallelic5-HTTLPR, and detected a strong omnibus association ofrs25532 with OCD (x2 ¼ 19.3, df ¼ 4, P , 0.0007; Table 1).This association was heavily driven by the rare LAT haplotype,which was twice as common in controls compared with OCDprobands. To rule out possibly spurious results related to rarehaplotypes, we re-analyzed our data without LAT (by exclud-ing haplotypes with ,5% frequency), and found that theomnibus association remained significant, although lessstrongly so (x2 ¼ 10.1, df ¼ 3, P , 0.018; Table 1). As canfurther be seen in Table 1 and in line with the hypothesis ofincreased SLC6A4 functioning in OCD, the LAC haplotype(which corresponds to the luciferase construct with thehighest expression) was more frequent in OCD probands thanin controls (52 versus 47.5%), although this did not reach stati-stical significance in the haplotype-specific test (Table 1).

In light of the small effect size of rs25532 and the 5-HTTLPRin general, we then investigated two additional SNPs, rs2020933and rs16965628, that were recently shown to predict allelicgene expression imbalance of SLC6A4 in lymphoblastoid celllines (19). As with rs25532, these SNPs individually were not(rs2020933) or only moderately (rs16965628, P , 0.04) signi-ficantly associated with OCD (Table 2).

We then studied triallelic 5-HTTLPR, rs25532, rs2020933,rs16965628 and the SLC6A4 intron 2 variable number oftandem repeat polymorphism as 5-locus haplotypes. Usingdefault parameters of WHAP, we observed a strong omnibusassociation of these five loci with OCD (x2 ¼ 25.4, df ¼ 9,P , 0.003; Table 2). Remarkably, the haplotype that was sig-nificantly associated with an increased odds ratio of 1.60 forOCD when tested individually against all other haplotypes(H6 in Table 2) is the haplotype containing the higher-expressing allele at each of the five polymorphisms.

Lastly, in order to dissect this haplotypic association, wetook advantage of the ability to specify different alternateand null models in WHAP for conditional analyses (22) and,again, excluded rare haplotypes of ,5% frequency. Systema-tic testing of all five polymorphisms through individual exclu-sion from the null model as well as controlling for each locusby excluding all other polymorphisms from the null modelrevealed that three loci (triallelic 5-HTTLPR, rs25532 andrs16965628) with five common haplotypes suffice for anoverall strong association with OCD (x2 ¼ 13.8, df ¼ 4,P , 0.008; Table 3). As earlier, the haplotype conferring anincreased odds ratio of 1.63 is the one containing the higher-expressing allele at each of the three loci.

DISCUSSION

The present study investigated the involvement of multiplecommon, non-coding functional variants of SLC6A4 inOCD. We demonstrated functionality in vitro for a previouslyreported SNP, rs25532, which interacts with the 5-HTTLPR.Consistent with the pathogenetic hypothesis of increased

718 Human Molecular Genetics, 2008, Vol. 17, No. 5

serotonin transporter activity in OCD, we showed that a hap-lotype containing the higher-expressing allele at each ofseveral SLC6A4 loci is significantly associated with increasedgenetic susceptibility to OCD. Importantly, the associationwith OCD was observed only when multiple variants wereanalyzed as haplotypes, but not as single loci. Our work con-tributes to a more detailed understanding of SLC6A4expression genetics in OCD and provides a molecular frame-work to explain, at least in part, inconsistent previous associa-tions of 5-HTTLPR and other functional SLC6A4polymorphisms in serotonin-related neuropsychiatric genetics.

The serotonin transporter gene encodes the single moleculethat terminates serotonergic neurotransmission and has there-fore been an intensely studied candidate. Allelic variation inthis transporter’s gene expression, regulated by a 43 bp indelin the upstream region termed 5-HTTLPR, was initiallyassociated with anxiety-related traits (23) and has since beeninvestigated in many fields of serotonin-related human gen-etics (24). Association results have generally been markedby at least some degree of inconsistency, for which anumber of explanations such as underpowered samples,small effect size, epistasis, polygenic contributions and

Figure 1. Schematic overview of the repetitive region containing the 5-HTTLPR indel and two functional SNPs, rs25531 and rs25532. Individual elements of therepetitive region are depicted by Greek letters according to the nomenclature introduced by Nakamura et al. (20). The two previously described functional poly-morphisms depicted here are a 43 bp indel known as the 5-HTTLPR (grey shading) and the rs25531 A/G SNP. Note that, as we have reported previously, thedefinition of the exact beginning and end of the indel varies between authors (21). A previously uncharacterized C . T SNP known as rs25532 is located,150 bp from the indel in the third-last repetitive element. The T allele of this SNP occurred much more frequently on S than on L allele background, andboth alleles of rs25532 were always in phase with A at rs25531 in this study. Three chromatograms from amplified genomic DNA of all three genotypes atrs25532 (arrowheads) are shown in the bottom part.

Figure 2. Luciferase reporter gene assay analyses of rs25532 in RN46A, PC12 and JAR cell lines in the context of the 5-HTTLPR. Columns and error barsrepresent mean normalized luciferase / Renilla luciferase ratios+SEM for four different constructs, SAC, SAT, LAC and LAT, that were used to assess the func-tional impact of the rs25532 in transfected rat medullary raphe (RN46A), adrenal pheochromocytoma (PC12) and human choriocarcinoma (JAR) cell lines. TheT allele significantly decreased expression by 15–30% in S constructs (SAT relative to SAC) and by 25–80% (LAT relative to LAC) in L alleles. N ¼ 5–6 transfec-tions per construct for RN46A and JAR cells, N ¼ 13–15 transfections per construct for PC12 cells. One-way analysis of variance for all four constructs: RN46Acells, F(3,18) ¼ 169.0, P , 0.0001; PC12 cells, F(3,51) ¼ 58.2, P , 0.0001; JAR cells, F(3,20) ¼ 56.3, P , 0.0001; Newman-Keuls post hoc test: RN46A cells,P , 0.001 for every pairwise comparison; PC12 cells, P , 0.001 for every pairwise comparison with the exception of SAC versus SAT, where P , 0.05; JARcells, P , 0.001 for every pairwise comparison with the exception of SAC versus SAT and LAT versus SAT where P , 0.01.

Human Molecular Genetics, 2008, Vol. 17, No. 5 719

population stratification have been suggested. From a molecu-lar perspective, unknown additional functional variationwithin SLC6A4 could further contribute to inconsistentresults. Hu et al. (16) have recently shown that a significantfraction of 10–25% (depending on ethnicity) of L alleles inthe 5-HTTLPR are in fact low-expressing due to contributionsfrom an interacting SNP, rs25531. We were able to furtherrefine the functionality of this locus by confirming and analyz-ing another SNP, rs25532, located ,150 bp from the indel.Our luciferase expression results are in line with previousstudies of this region that have shown an up to 2-fold differ-ence between S and L alleles (16,17). Our data and the workby Hu et al. (16) demonstrate that the classic attribution ofS and L alleles as low- and high-expressing, respectively,can no longer be viewed as valid without appreciation of themodulating effects of rs25531 and rs25532. More importantly,however, it will be crucial to determine (1) how this functionallocus interacts with the allelic expression imbalance markersdownstream (and which the actual locus conferring allelicexpression imbalance is, see in what follows) (2), what therelative contribution to a net effect on serotonin transportergene expression is in vivo and (3) how (if at all) this translatesinto protein abundance and actual genetically determineddifferential functioning of the serotonergic synapse. To thisend, it will be essential to investigate the role of these novelSLC6A4 variants in mRNA levels and protein abundance indorsal raphe nuclei as well as in peripheral cells known toexpress the serotonin transporter, such as lymphocytes, mega-karyocytes and lymphoblastoid cell lines.

The conditional haplotype analyses conducted hereinunveiled that the LA alleles at 5-HTTLPR can be further subdi-vided at least 2-fold (1): by interaction with rs25532 into LAC

and LAT; and (2) on the basis of different rs16965628 alleles.While the LAT haplotype was too rare to be included in our stat-istical tests, appreciation of rs16965628 revealed that it is thepresumably higher-expressing LAC haplotype with C atrs16965628 (H4 in Table 3) that is significantly more frequentin OCD than in controls. In contrast, the more common LAC

haplotype with G at rs16965628 (H1 in Table 3) was not signifi-cantly different. This subdivision of LA alleles on the basis ofrs16965628 is a plausible explanation for our previous non-replication of association of LA with OCD (18), and it illustratesthe usefulness of haplotypic approaches with multiple func-tional markers to complex genetic disorders. It will be of highinterest to see whether our haplotypic association can be

replicated in OCD and in neighboring fields such as OCDspectrum disorders, autism or serotonin re-uptake inhibitorresponse, especially given the recent report of an associationof coding gain-of-function SLC6A4 variants with rigid-compulsive behaviors in autism (25).

Three limitations in our study pose caveats to the interpret-ation and warrant further analyses. First, our case–controlassociation study design is generally susceptible to populationstratification, and although we addressed this issue by match-ing ethnicities between cases and controls, spurious associa-tion can only be ruled out definitively through family-basedstudies or through analysis of a large number ofancestry-informative markers, which was not feasible in thepresent work. Second, although the haplotype associatedwith OCD in this study contains the allele associated withhigher-expression at each of its loci, it remains to be function-ally demonstrated that all of these loci interact additively in away that renders this haplotype as highest-expressing andothers as lower-expressing. These analyses can currently notreadily be undertaken given the very large distance betweenthese loci, but they will be crucial to determine the relativecontribution of each of these polymorphisms in their broadergenomic context. Lastly, the two recently-identified SNPsassociated with allelic expression imbalance, rs16965628and rs2020933, could of course merely be markers inlinkage disequilibrium not only with each other [as reflectedin Table 2 and as noted previously by Martin et al. (19)] butalso with the actual functional locus or loci since they havenot yet been independently analyzed in reporter gene orother in vitro assays. In summary, our work shows that allow-ing for multiple putatively or refined functional variants andtheir analysis as haplotypes can be crucial for the detectionof association, and we hope it contributes to a better under-standing of SLC6A4 expression functionality.

MATERIALS AND METHODS

Subjects

We studied a total of 295 Caucasian OCD probands and 657Caucasian control individuals. Both samples have recentlybeen described in detail (18,26). Briefly, probands wererecruited through an adult outpatient OCD program at theNational Institute of Mental Health (NIMH) in Bethesda,MD, USA. All probands were at least 18 years old atinclusion, have a primary OCD diagnosis on the basis of theStructured Clinical Interview for DSM-IV (SCID) and gavewritten informed consent. Control samples originated fromthree independent sources: (i) Coriell cell repository(Camden, NJ, USA; N ¼ 200 self-declared healthy US Cauca-sians); (ii) European Collection of Cell Cultures(Sigma-Aldrich, St Louis, MO, USA; N ¼ 192 apparentlyhealthy UK Caucasian blood donors); and (iii) undergraduatestudents (N ¼ 265 self-declared healthy Caucasians) from alarge Southeastern university who participated in a separatestudy of genes and personality in return for partial coursecredit. All studies were conducted under a protocol approvedby the Institutional Review Board of the NIMH Division ofIntramural Research Programs in Bethesda, MD, and by theHuman Subjects Committee at Florida State University. As

Table 1. Haplotype analysis of 5-HTTLPR and rs25532 in OCD

Triallelic5-HTTLPR

rs25532 Haplotypefrequency

Haplotype-specifictests (df ¼ 1)

OCD Controls P-value Odds ratio

H1 LA C 0.510 0.466 ns 1.21H2 S C 0.302 0.328 ns 0.89H3 S T 0.128 0.106 ns 1.24H4 LG C 0.046 0.072 0.024 0.61H5 LA T 0.013 0.028 0.007 0.22

Omnibus X2 ¼ 19.3 (df ¼ 4); P , 0.0007.Omnibus test excluding H5: X2 ¼10.1 (df ¼ 3); P , 0.018.N ¼ 295 OCD probands and N ¼ 657 control individuals.

720 Human Molecular Genetics, 2008, Vol. 17, No. 5

only the third control group was evaluated by self-report andstandard scales, we cannot rule out the occurrence of OCDwithin the other groups, although frequency above thegeneral population prevalence of 2–3% would seem unlikelyand thus should not have significantly impacted our results.Allelic and genotypic frequencies did not significantly differbetween these three control groups; moreover, none of thepolymorphisms significantly deviated from Hardy–Weinbergequilibrium as determined by contingency table statistics(data not shown).

Genotyping

Deoxyribonucleic acid was extracted from whole bloodobtained through peripheral venipuncture or from salivasamples (Oragene discs, DNA Genotek, Ottawa, ON).

Genotyping procedures and genotype results for 5-HTTLPR/rs25531 and STin2 have been published previously by us(18) and were used here only as part of the haplotype analyses.The three SNPs rs25532, rs2020933 and rs16965628 weregenotyped using 5’-exonuclease assays (TaqMan SNP geno-typing assays-by-design; Applied Biosystems, Foster City,CA) under standard conditions described previously (27) andwith the following primer and probe sequences: rs25532,forward primer CTG CAC CCC CCA GCA T, reverseprimer GGT AGG GTG CAA GGA GAA TGC, VIC probeCCG GcA TCC CCC CT, FAM probe CCC GGt ATC CCCCCT; rs2020933, forward primer TGT ATG TAT TTT TACCAT CAG TTT TGT CCA GAA, reverse primer GAG AGTTAG CTA GCA GGC TCA TAA AT, VIC probe CATTGA CCa GGT TCA C, FAM probe CAT TGA CCt GGTTCA C; and rs16965628, forward primer GGC CTC AGTTTC CCT GCT A, reverse primer GTT GAT GTC ACTATC ACC ACC ATA CA, VIC probe AAC CCA TTg TGCCTT T, FAM probe AAC CCA TTc TGC CTT T. Theoverall genotype completion rate exceeded 94% for eachassay, samples analyzed in duplicate and no-template-controlsconsistently yielded expected results.

Sequencing of the 5-HTTLPR

A total of 30–50 ng of genomic DNA were amplified in a totalreaction volume of 20 ml using final concentrations of 1�multiplex master mix (Qiagen, Valencia, CA) and 500 nM

each of oligonucleotide primers GCC AGC ACC TAA CCCCTA AT and GAG GGA CTG AGC TGG ACA AC(Operon, Huntsville, AL). Thermocycling consisted of15 min initial hot start activation at 958C followed by 38cycles of denaturation at 968C (10 s), annealing at 628C(25 s) and extension at 728C (50 s); final extension was

Table 2. Single locus and haplotype association analyses of multiple functional polymorphisms of SLC6A4 in OCD

Triallelic5-HTTLPR

rs25532 rs2020933 rs16965628 STin2 Haplotype frequency Haplotype-specifictests (df ¼ 1)

OCD Controls P-value Oddsratio

H1 S C A G 12 0.251 0.274 ns 0.89H2 LA C A G 10 0.270 0.240 ns 1.19H3 LA C A G 12 0.166 0.184 ns 0.85H4 S T A G 12 0.074 0.068 ns 1.12H5 S T A G 10 0.063 0.046 ns 1.48H6 LA C T C 12 0.074 0.047 0.026 1.60H7 LG C A G 12 0.036 0.049 ns 1.45H8 S C A G 10 0.048 0.047 ns 1.02H9 LG C A G 10 0.012 0.025 0.019 028H10 LA T A G 10 0.007 0.020 0.005 0.13

Functionalitya LA . LG . S C . T T . A C . G 12 . 10 Omnibus X2 ¼25.3 (df¼ 9),MAF OCD LG, 0.046 T, 0.142 T, 0.086 C, 0.099 10, 0.405 P , 0.003MAF controls 0.072 0.134 0.064 0.070 0.378Single locus testb ns ns ns P , 0.040 ns

N ¼ 295 OCD probands and N ¼ 657 control individuals. A haplotype significantly associated with increased OCD risk, H6 (bold-face values), containsthe higher-expressing allele at each locus. MAF, minor allelic frequency.aFunctionality classification is based on the works by Heils et al. (17) and Hu et al. (16) for triallelic 5-HTTLPR, MacKenzie and Quinn (29) for STin2, Martinet al. (19) for rs2020933 and rs16965628 and the present work for rs25532.bSingle locus association analyses for triallelic 5-HTTLPR and STin2 have been published previously (18).

Table 3. Three-locus haplotype association of SLC6A4 with OCD afterconditional effect tests

Triallelic5-HTTLPR

rs25532 rs16965628

Haplotypefrequency

Haplotype-specifictests (df ¼ 1)

OCD Controls P-value Odds ratio

H1 LA C G 0.433 0.422 ns 1.04H2 S C G 0.287 0.326 ns 0.84H3 S T G 0.140 0.118 ns 1.21H4 LA C C 0.093 0.059 0.010 1.63H5 LG C G 0.047 0.073 0.027 0.62

Omnibus X2 ¼ 13.8 (df ¼ 4); P , 0.0008.N ¼ 295 OCD probands and N ¼ 657 control individuals. As in Table 2, thehaplotype with the higher-expressing allele at each locus is significantlyassociated with increased susceptibility to OCD. Note also that rs16965628distinguishes between a common (H1) and a rare (H4, bold-face values) LAC

haplotype, and that only the rare H4 haplotype with the higher-expressing Callele at rs16965628 significantly differs between OCD and controls.

Human Molecular Genetics, 2008, Vol. 17, No. 5 721

performed at 728C for 5 min. Specific amplification of a468 bp (SS) or 511 bp (LL) product was confirmed byagarose gel electrophoresis and ethidium bromide staining(Sigma). Amplicons were purified using MultiScreen HTSfilter plates (Millipore, Bedford, MA) attached to a vacuummanifold (Millipore), and the concentration was determinedspectrophotometrically (NanoDrop, Wilmington, DE). Atotal of 60 ng of purified amplicon DNA was subsequentlysequenced bidirectionally with amplification primers usingBigDye Terminator chemistry on an ABI 3100 automatedsequence analyzer (Applied Biosystems) at the National Insti-tute of Neurological Disorders and Stroke DNA sequencingcore facility. Raw chromatograms were analyzed separatelyfor SS and LL samples with the phred/phrap/consed suiteand PolyPhred version 6.02 beta (28).

Reporter gene assays

We created 5-HTTLPR reporter gene constructs for LAC, LAT,SAC and SAT. Genomic DNA was amplified with primers andunder conditions described above (Sequencing of the5-HTTLPR). Amplicons were cloned into pCRII-TOPOvector (Invitrogen, Carlsbad, CA) and then subcloned intopGL4.10 (Promega, Madison, WI). All pGL constructs werebidirectionally sequenced with primers AGT GCA GGTGCC AGA ACA TT and TCT TCC ATG GTG GCT TTACC to check for proper orientation and to confirm sequencespecificity and absence of artificial mutations. Undifferen-tiated rat raphe medullary raphe cells (RN46A, a kind giftfrom Dr. Scott R. Whittemore, University of Louisville), aswell as rat adrenal pheochromocytoma (PC12) and humanchoriocarcinoma (JAR) cells grown under standard conditionswere cotransfected with the pGL4.10 constructs and pRL(Renilla luciferase, Promega) using lipofectamine 2000reagent (Invitrogen). Twenty-four hours after transfection,cells were harvested and lysed, and luciferase activity wasmeasured using Dual Luciferase Assay (Promega) followingthe manufacturer’s protocol in a 20/20n luminometer(Turner Biosystems, Sunnyvale, CA). Renilla luminescencewas used to correct for variable transfection efficiency. Eachof the four constructs was transfected 3–5 times in each offour experiments for a total of 13–15 independent transfec-tions for PC12 cells; for RN46A and JAR cells, each of theconstructs was transfected 5–6 times; the empty pGL4.10vector was transfected thrice for control purposes (data notshown). Firefly/Renilla ratios were normalized to the geo-metric mean of SAC ratios. Normalized ratios for the four con-structs were then analyzed by one-way analysis of variancefollowed by Newman–Keuls test to compare all pairwisegroup means with Prism 4 for Windows (GraphPad Software,San Diego, CA).

Statistical genetics

All analyses were carried out using the software WHAP:haplotype-based association analysis version 2.09 (22). Weanalyzed 5-HTTLPR/rs25531 as one triallelic locus with LA,LG and SA alleles (two SG alleles (21) were treated as SA)using the -usat command. Nine-repeat alleles (38 out of1904 chromosomes) at the STin2 polymorphism were treated

as 10-repeat alleles. After detection of a significant maineffect, we performed conditional analyses in WHAP as perthe author’s website instructions (http://pngu.mgh.harvard.edu/~purcell/whap/condtut.shtml). Each of the five loci wasboth dropped from the null model (to detect whether it hasan independent effect after controlling for everything else)and specified as the sole null model (to control for it andtest whether it can explain the total association). For these con-ditional analyses, the haplotype frequency threshold wasraised to 5% with the -at 5 flag. Haplotype-specific testswith df ¼ 1 and single locus tests with df ¼ 2 for triallelic5-HTTLPR/rs25531 and df ¼ 1 for all other loci werecarried out with the -hs and -alt options, respectively. Oddsratios were calculated by raising Euler’s number to thepower of the regression coefficients as returned by WHAP.A P-value of 0.05 was considered significant.

ACKNOWLEDGEMENTS

We gratefully thank all the participants in this study. We areindebted to Diane Kazuba, Brenda Justement and MichaelWheaton for conducting patient interviews; to Teresa Tolliverfor technical assistance in whole blood DNA extraction; andto Prof Dr K.-P. Lesch, from the University of Wurzburg, forhelpful discussion and comments on the manuscript. We alsothank James Nagle and Debbie Kauffman, from the NINDSDNA sequencing core facility, for excellent technical assist-ance with sequencing the serotonin transporter gene promoteras part of a ‘large sequencing project’ granted to J.R.W.; andDr Scott R. Whittemore, University of Louisville, for providingus with the RN46A cell line.

Conflict of Interest statement. None declared.

FUNDING

This research was supported by the Intramural ResearchProgram of the National Institute of Mental Health, NationalInstitutes of Health.

REFERENCES

1. Pigott, T.A., Pato, M.T., Bernstein, S.E., Grover, G.N., Hill, J.L., Tolliver, T.J.and Murphy, D.L. (1990) Controlled comparisons of clomipramine andfluoxetine in the treatment of obsessive-compulsive disorder. Behavioraland biological results. Arch. Gen. Psychiatry, 47, 926–932.

2. Greist, J.H., Jefferson, J.W., Kobak, K.A., Katzelnick, D.J. and Serlin, R.C.(1995) Efficacy and tolerability of serotonin transport inhibitors inobsessive-compulsive disorder. A meta-analysis. Arch. Gen. Psychiatry,52, 53–60.

3. Ozaki, N., Goldman, D., Kaye, W.H., Plotnicov, K., Greenberg, B.D.,Lappalainen, J., Rudnick, G. and Murphy, D.L. (2003) Serotonintransporter missense mutation associated with a complex neuropsychiatricphenotype. Mol. Psychiatry, 8, 933–936.

4. Kilic, F., Murphy, D.L. and Rudnick, G. (2003) A human serotonintransporter mutation causes constitutive activation of transport activity.Mol. Pharmacol., 64, 440–446.

5. Prasad, H.C., Zhu, C.B., McCauley, J.L., Samuvel, D.J., Ramamoorthy, S.,Shelton, R.C., Hewlett, W.A., Sutcliffe, J.S. and Blakely, R.D. (2005)Human serotonin transporter variants display altered sensitivity to proteinkinase G and p38 mitogen-activated protein kinase. Proc. Natl Acad. Sci.USA, 102, 11545–11550.

722 Human Molecular Genetics, 2008, Vol. 17, No. 5

6. Delorme, R., Betancur, C., Wagner, M., Krebs, M.O., Gorwood, P., Pearl, P.,Nygren, G., Durand, C.M., Buhtz, F., Pickering, P. et al. (2005)Support for the association between the rare functional variant I425V ofthe serotonin transporter gene and susceptibility to obsessive compulsivedisorder. Mol. Psychiatry, 10, 1059–1061.

7. Wendland, J.R., DeGuzman, T.B., McMahon, F., Rudnick, G.,Detera-Wadleigh, S. and Murphy, D.L. SERT Ileu425Val (OCD1) andother uncommon gene variants and chromosomal anomalies associatedwith an obsessive-compulsive disorder phenotype. Psychiatr. Genetics,in press.

8. McDougle, C.J., Epperson, C.N., Price, L.H. and Gelernter, J. (1998)Evidence for linkage disequilibrium between serotonin transporter proteingene (SLC6A4) and obsessive compulsive disorder. Mol. Psychiatry, 3,270–273.

9. Bengel, D., Greenberg, B.D., Cora-Locatelli, G., Altemus, M., Heils, A.,Li, Q. and Murphy, D.L. (1999) Association of the serotonin transporterpromoter regulatory region polymorphism and obsessive-compulsivedisorder. Mol. Psychiatry, 4, 463–466.

10. Frisch, A., Michaelovsky, E., Rockah, R., Amir, I., Hermesh, H., Laor, N.,Fuchs, C., Zohar, J., Lerer, B., Buniak, S.F. et al. (2000) Associationbetween obsessive-compulsive disorder and polymorphisms of genesencoding components of the serotonergic and dopaminergic pathways.Eur. Neuropsychopharmacol., 10, 205–209.

11. Kinnear, C.J., Niehaus, D.J., Moolman-Smook, J.C., du Toit, P.L., vanKradenberg, J., Weyers, J.B., Potgieter, A., Marais, V., Emsley, R.A.,Knowles, J.A. et al. (2000) Obsessive-compulsive disorder and thepromoter region polymorphism (5-HTTLPR) in the serotonin transportergene (SLC6A4): a negative association study in the Afrikaner population.Int. J. Neuropsychopharmacol., 3, 327–331.

12. Camarena, B., Rinetti, G., Cruz, C., Hernandez, S., de la Fuente, J.R. andNicolini, H. (2001) Association study of the serotonin transporter genepolymorphism in obsessive-compulsive disorder. Int. J. Neuropsychophar

macol., 4, 269–272.

13. Chabane, N., Millet, B., Delorme, R., Lichtermann, D., Mathieu, F.,Laplanche, J.L., Roy, I., Mouren, M.C., Hankard, R., Maier, W. et al.

(2004) Lack of evidence for association between serotonin transportergene (5-HTTLPR) and obsessive-compulsive disorder by case control andfamily association study in humans. Neurosci. Lett., 363, 154–156.

14. Meira-Lima, I., Shavitt, R.G., Miguita, K., Ikenaga, E., Miguel, E.C. andVallada, H. (2004) Association analysis of the catechol-o-methyltransfer-ase (COMT), serotonin transporter (5-HTT) and serotonin 2A receptor(5HT2A) gene polymorphisms with obsessive-compulsive disorder.Genes Brain Behav., 3, 75–79.

15. Walitza, S., Wewetzer, C., Gerlach, M., Klampfl, K., Geller, F., Barth, N.,Hahn, F., Herpertz-Dahlmann, B., Gossler, M., Fleischhaker, C. et al.

(2004) Transmission disequilibrium studies in children and adolescentswith obsessive-compulsive disorders pertaining to polymorphisms ofgenes of the serotonergic pathway. J. Neural. Transm., 111, 817–825.

16. Hu, X.Z., Lipsky, R.H., Zhu, G., Akhtar, L.A., Taubman, J., Greenberg, B.D.,Xu, K., Arnold, P.D., Richter, M.A., Kennedy, J.L. et al. (2006)Serotonin transporter promoter gain-of-function genotypes are linked toobsessive-compulsive disorder. Am. J. Hum. Genet., 78, 815–826.

17. Heils, A., Teufel, A., Petri, S., Stober, G., Riederer, P., Bengel, D. andLesch, K.P. (1996) Allelic variation of human serotonin transporter geneexpression. J. Neurochem., 66, 2621–2624.

18. Wendland, J.R., Kruse, M.R., Cromer, K.C. and Murphy, D.L. (2007)A large case-control study of common functional SLC6A4 and BDNFvariants in obsessive-compulsive disorder. Neuropsychopharmacology,32, 2543–2551.

19. Martin, J., Cleak, J., Willis-Owen, S.A., Flint, J. and Shifman, S. (2007)Mapping regulatory variants for the serotonin transporter gene based onallelic expression imbalance. Mol. Psychiatry, 12, 421–422.

20. Nakamura, M., Ueno, S., Sano, A. and Tanabe, H. (2000) The humanserotonin transporter gene linked polymorphism (5-HTTLPR) shows tennovel allelic variants. Mol. Psychiatry, 5, 32–38.

21. Wendland, J.R., Martin, B.J., Kruse, M.R., Lesch, K.P. and Murphy, D.L.(2006) Simultaneous genotyping of four functional loci of humanSLC6A4, with a reappraisal of 5-HTTLPR and rs25531. Mol. Psychiatry,11, 224–226.

22. Purcell, S., Daly, M.J. and Sham, P.C. (2007) WHAP: haplotype-basedassociation analysis. Bioinformatics, 23, 255–256.

23. Lesch, K.P., Bengel, D., Heils, A., Sabol, S.Z., Greenberg, B.D., Petri, S.,Benjamin, J., Muller, C.R., Hamer, D.H. and Murphy, D.L. (1996)Association of anxiety-related traits with a polymorphism in the serotonintransporter gene regulatory region. Science, 274, 1527–1531.

24. Murphy, D.L., Lerner, A., Rudnick, G. and Lesch, K.P. (2004) Serotonintransporter: gene, genetic disorders, and pharmacogenetics. Mol. Interv.,4, 109–123.

25. Sutcliffe, J.S., Delahanty, R.J., Prasad, H.C., McCauley, J.L., Han, Q.,Jiang, L., Li, C., Folstein, S.E. and Blakely, R.D. (2005) Allelicheterogeneity at the serotonin transporter locus (SLC6A4) conferssusceptibility to autism and rigid-compulsive behaviors. Am. J. Hum.Genet., 77, 265–279.

26. LaSalle, V.H., Cromer, K.R., Nelson, K.N., Kazuba, D., Justement, L. andMurphy, D.L. (2004) Diagnostic interview assessed neuropsychiatricdisorder comorbidity in 334 individuals with obsessive-compulsivedisorder. Depress. Anxiety, 19, 163–173.

27. Wendland, J.R., Kruse, M.R. and Murphy, D.L. (2006) FunctionalSLITRK1 var321, varCDfs and SLC6A4 G56A variants and susceptibilityto obsessive-compulsive disorder. Mol. Psychiatry, 11, 802–804.

28. Bhangale, T.R., Stephens, M. and Nickerson, D.A. (2006) Automatingresequencing-based detection of insertion-deletion polymorphisms.Nat. Genet., 38, 1457–1462.

29. MacKenzie, A. and Quinn, J. (1999) A serotonin transporter gene intron 2polymorphic region, correlated with affective disorders, hasallele-dependent differential enhancer-like properties in the mouseembryo. Proc. Natl Acad. Sci. USA, 96, 15251–11525.

Human Molecular Genetics, 2008, Vol. 17, No. 5 723