Original Article

The Pharmacogenomics Journal (2012) 12, 513–520; doi:10.1038/tpj.2011.32; published online 2 August 2011

Support for association of HSPG2 with tardive dyskinesia in Caucasian populations

L Greenbaum1, A Alkelai1, P Zozulinsky1, Y Kohn1 and B Lerer1

1Biological Psychiatry Laboratory, Department of Psychiatry, Hadassah–Hebrew University Medical Center, Jerusalem, Israel

Correspondence: Professor B Lerer, Biological Psychiatry Laboratory, Department of Psychiatry, Hadassah–Hebrew University Medical Center, Ein Karem, Jerusalem 91120, Israel. E-mail: lerer@cc.huji.ac.il

Received 30 January 2011; Revised 6 June 2011; Accepted 20 June 2011
Advance online publication 2 August 2011



Tardive dyskinesia (TD) is a severe adverse effect of chronic antipsychotic drug treatment. In addition to clinical risk factors, TD susceptibility is influenced by genetic predisposition. Recently, Syu et al. (2010) reported a genome-wide association screening of TD in Japanese schizophrenia patients. The best result was association of single-nucleotide polymorphism (SNP) rs2445142 in the HSPG2 (heparan sulfate proteoglycan 2) gene with TD. In the present study, we report a replication study of the five top Japanese TD-associated SNPs in two Caucasian TD samples. Applying logistic regression and controlling for relevant clinical risk factors, we were able to replicate the association of HSPG2 SNP rs2445142 with TD in a prospective study sample of 179 Americans of European origin by performing a secondary analysis of the CATIE (Clinical Antipsychotic Trials of Intervention Effectiveness) genome-wide association study data set, and using a perfect proxy surrogate marker (rs878949; P=0.039). An association of the ‘G’ risk allele of HSPG2 SNP rs2445142 with TD was also shown in a sample of Jewish Israeli schizophrenia patients (retrospective, cross-sectional design; P=0.03). Although the associations were only nominally significant, the findings provide further support for the possible involvement of HSPG2 in susceptibility to TD.


schizophrenia; tardive dyskinesia; antipsychotics; HSPG2



Tardive dyskinesia (TD) is a chronic adverse effect of prolonged antipsychotic drug treatment, manifesting as abnormal involuntary movements of the face, extremities and trunk.1, 2 TD incidence is estimated as 5% per year for typical, first-generation antipsychotic treatment,3 and its prevalence among chronic schizophrenia patients treated with antipsychotics is 20–25%.1, 4 Atypical, second-generation antipsychotics are generally considered less likely to cause TD than first-generation antipsychotics, approximately 1% annually.5 However, systemic review of ~28000 patients from 12 long-term studies discovered a much higher annualized TD incidence of 3.9% for second-generation antipsychotics compared with 5.5% for first-generation antipsychotics.6 The clinical risk factors for TD reported in the literature are numerous, and include female gender, older age, duration and intensity of antipsychotic treatment, smoking, organic brain abnormalities and affective disorder.7, 8 Nevertheless, recent meta-analysis showed that only early extrapyramidal symptoms and non-white ethnic group qualified as TD risk factor among schizophrenia patients.9 The association with older age was suggestive but inconclusive.9 In addition, genetic susceptibility contributes to individual susceptibility to develop TD following antipsychotic exposure,10 and despite methodological difficulties, family studies support a genetic component.11, 12

The pathogenesis of TD is unknown and may involve supersensitivity of the nigrostriatal pathway dopamine D2 receptors,13 neuronal damage caused by free radical overproduction14, 15 and GABAergic system dysregulation.16, 17 Interestingly, several studies have shown that spontaneous movement disorders, mainly spontaneous dyskinesia, may be present in antipsychotic-naive schizophrenia patients and may represent an intrinsic neuromotor component of this disorder.18, 19, 20 If this concept is valid, it is not possible to differentiate between dyskinesia related to the illness of schizophrenia and drug-induced TD.

Using a candidate gene approach, several genetic risk factors for TD have been identified with variable support. These include variants within COMT, DRD2, MnSOD, DRD3, HTR2A, CYP1A2 and CYP2D6.21, 22, 23, 24, 25, 26 Recently, two genome-wide association studies (GWASs) and a large candidate gene study of TD were published, based on the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) study, using two different definitions for TD phenotype.27, 28, 29 Although none of the top results reached genome-wide significance, the association of single-nucleotide polymorphism (SNP) rs3943552 in the GLI2 gene with TD was observed in the CATIE sample and independently supported in Jewish Israeli schizophrenia patients of Ashkenazi origin.29

Syu et al.30 performed a genome-wide association screening of treatment-resistant TD in Japanese schizophrenia patients. In the first step, they studied association of 40573 tagging SNPs with TD in a small screening sample of 50 TD and 50 non-TD Japanese schizophrenia subjects.31 Cases were defined as schizophrenia patients treated with antipsychotics for at least a year, in whom TD persisted for more than 1 year and did not improve even after a year of appropriate therapy. Controls had never developed TD despite antipsychotic exposure of more than 10 years.31 Although no association was significant after correction for multiple testing, a confirmation step was performed by genotyping the top 63 SNPs in an independent sample of 36 Japanese TD patients and 136 non-TD patients.30 Four intronic SNPs were nominally associated with TD in both the discovery and confirmation samples: SNP rs2445142 (HSPG2), rs4738269 (KCNB2), rs886292 (ABCC8) and rs2061051 (GABRG3). Focusing on HSPG2, which encodes heparin sulfate proteoglycan 2, an additional 24 tagging SNPs were genotyped in this gene. SNP rs2124368 (not in linkage disequilibrium (LD) with rs2445142) was significantly associated with TD, surviving correction for multiple testing in the validation step. Functional studies provided support for involvement of HSPG2 in TD. The risk allele ‘G’ of rs2445142 was significantly associated with higher expression of HSPG2 in human brain tissue, and Hspg2 heterozygous knockout mice showed lower levels of vacuous chewing movement induced by chronic treatment with a haloperidol–reserpine regiment when compared with wild-type mice.30 The authors concluded that higher expression of HSPG2 may be associated with increased TD susceptibility.

In this work, we performed a replication study of the findings of Syu et al.30 in Caucasian populations. We decided a priori to study only the four SNPs for which TD association in the original discovery sample was supported in the validation step, as well as the additional HSPG2 SNP, rs2124368, which survived correction for multiple testing (five SNPs in all). We used two available TD samples: a cross-sectionally evaluated TD sample of Jewish Israeli schizophrenia patients (equivalent to the probable level of TD diagnostic) and a prospectively evaluated European TD sample derived from the CATIE study.


Materials and methods

Jewish Israeli sample


This sample is essentially the same as reported by us in a previous study,29 although slightly smaller in terms of participant number (DNAs of four patients were unavailable). Briefly, all 166 participants were Jewish Israeli schizophrenia patients (according to DSM-IV (Diagnostic and Statistical Manual of Mental Disorders-fourth edition) criteria), aged>18 years. They were screened for TD according to the RDC-TD (Research Diagnostic Criteria for Tardive Dyskinesia).21, 32, 33 Only patients treated with antipsychotics medication for at least 3 months before TD evaluation were included. Patients with a history of alcohol/substance abuse in the past 5 years or active or past medical/neurologic illness that might confound study evaluations were excluded. The project was approved by the Helsinki Committee (Internal Review Board) of the Hadassah–Hebrew University Medical Center, and participants provided written informed consent. As data for cumulative lifetime antipsychotic exposure were unavailable, we calculated the number of years since first antipsychotic treatment as an indirect indicator for this variable (Table 1).

Diagnostic instruments

Evaluation tools included standardized checklists for schizophrenia diagnosis according to DSM-IV on the basis of clinical examination and case notes, and the Abnormal Involuntary Movement Scale (AIMS)34 for assessment of TD. AIMS measures involuntary movements for seven different body regions (orofacial, extremities and trunk) and scores them on a 0–4 scale. To insure consistency in the rating of TD, all raters achieved inter-rater reliability of >0.9, with the principal investigator bearing overall responsibility for the TD assessments, based on the same series of patients, before commencing the study.32, 33 For the broader TD yes/no phenotype, patients who received a rating of mild dyskinesia (score of 2) in two or more body regions or of moderate or severe dyskinesia (score of 3 to 4) in any one body region were categorized as having TD according to the Schooler–Kane TD-RDC35 (cases, n=75). All other patients were classified as non-TD controls (n=91) (Table 1b). As patients were assessed only once, the TD diagnosis in our sample corresponds to the category of ‘cross-sectional TD’.36, 37, 38

For an additional analysis of extreme distribution of the phenotype, we defined controls as patients whose AIMS score was zero on all seven items evaluated (abnormal involuntary movements were completely absent), whereas cases were the same as described above. This is a much more rigorous definition, resembling conceptually the Japanese TD treatment-resistant phenotype and the extreme phenotype used in the European CATIE TD sample (see below). However, only a small number of patients (32) met this ‘extreme control’ criterion (Table 1).


We genotyped the five SNPs of interest (rs2445142, rs2124368, rs4738269, rs886292 and rs2061051) with the Sequenom MassARRAY system at the Washington University Human Genetics Division Genotyping Core (St Louis, MO, USA). Quality control measures were implemented.

Data analysis

To analyze the association of the genotyped SNPs and HSPG2 dyplotypes with TD (two phenotype definitions) in the Jewish Israeli sample, we used logistic regression (additive model). Because of clinical and theoretical considerations (see Introduction), and consistent with our previous study,29 the following variables were included as covariates: gender, age, years since the inception of antipsychotic treatment and self-reported ancestry (Ashkenazi vs non-Ashkenazi origin). Statistical analysis, including Hardy–Weinberg equilibrium calculation, was performed using PLINK.39 For our five SNPs, we had sample size power of greater than or equal to0.8 to detect a genetic effect size ranging from 1.9 to 2.25 (depending on each SNP minor allele frequency), with an α of 0.05 in the broader phenotype (as calculated with Quanto software (University of Southern California, Los Angeles, CA, USA)).

CATIE TD sample


Our original CATIE TD GWAS sample contained 327 patients: 217 of white ancestry (European or Hispanic origin), 102 African Americans and 8 mixed.29 The analysis was based on phenotypic and genotypic data collected in the CATIE clinical trial and GWAS.40, 41 Of the five SNPs of interest, four were not included in the original CATIE GWAS. As we did not have access to DNA from the CATIE study, new genotyping was not feasible. Therefore, in the present analysis we used surrogate, genotyped SNPs as proxies for the non-genotyped SNPs. For this reason we included in the current analysis only patients of European origin. This European origin sample included 179 schizophrenia patients, 75 affected with TD and 104 unaffected, according to the criteria described below (Table 1a).

CATIE study design and pharmacological interventions

The CATIE study was a randomized, controlled, multiphase study, which took place in the United States between January 2001 and December 2004. Its aim was to assess response of schizophrenia patients to antipsychotic medication.40, 42 A total of 1460 patients diagnosed with schizophrenia (DSM-IV criteria) were included, aged 16–67 years. Our TD GWAS secondary analysis sample included subjects who took part in CATIE phases 1–3 for a maximum of 18 months, with a minimum participation time of 90 days, and also supplied DNA sample. Participants received one of several antipsychotic drugs as monotherapy (first-generation or second-generation antipsychotics) in each of the study stages. Treatment duration and discontinuation were determined by judgment of the CATIE clinicians. Patients were evaluated for extrapyramidal symptoms at baseline, after 1 month, after 3 months and then every 3 months up to 18 months. Further information regarding the pharmacological intervention regiment, including list of agents used in each of the three trail stages, is given in our previous publication.29

TD assessment

CATIE participants were assessed for TD with AIMS.34 In this prospective study design, patients were categorized as affected with TD if they demonstrated involuntary movements of a mild degree in two or more body regions or of a moderate to severe degree in at least one body region, on at least two separate evaluations during the follow-up period (including baseline measurement). For TD unaffected (controls), we used a strict, ‘clean’ phenotype definition—absence of any abnormal involuntary movement (grade 0 on all first seven AIMS items), throughout the whole study period of greater than or equal to90 days. The same approach of extreme definition of phenotypes was implemented by us in our previous CATIE TD GWAS,29 and is parallel to the narrower, extreme phenotype in the Jewish sample.

Surrogate SNP selection

Out of the five SNPs chosen for study, only rs2124368 had been genotyped in the CATIE GWAS. Therefore, we choose surrogates for the other four markers according to CEU HapMap data. In our European ancestry TD sample, rs878949 was chosen as surrogate for rs2445142 (D=1, r2=1); rs2237979 as surrogate for rs886292 (D=1, r2=1); and rs7009604 as surrogate for rs4738269 (D=1, r2=0.87). For rs2061051, no suitable surrogate SNP with r2>0.8 was available. These SNPs are not good surrogates in the African population, as seen in HapMap YRI sample (r2 equals 0.14, 0.05 and 0.29, respectively). As the CATIE sample population included patients of European, African-American and other (including Hispanic) ancestry, we were forced to use only participants of European origin, for our surrogate marker-based analysis.

Data analysis

In total, four SNPs were tested for association (additive model) with the dichotomized TD phenotype, using a logistic regression analysis performed with PLINK. As we had the complete GWAS genotype data (495172 SNPs) of the CATIE European participants, we used principal components analysis to correct for possible effects of population stratification.43, 44 In addition to population stratification covariates (computed by HelixTree software (http://www.goldenhelix.com)), the logistic regression included the following covariates: gender, age, treatment with typical antipsychotics (yes/no) and years since first antipsychotic prescription (as an estimate of the unavailable cumulative lifetime exposure to antipsychotics). The covariate selection is identical to our previous CATIE TD GWAS selection, based on the well-established contribution of these variables to TD susceptibility.29 For our four SNPs, we had sample size power of greater than or equal to0.8 to detect a genetic effect size ranging from 1.85 to 2.2 (depending on each SNP minor allele frequency), with an α of 0.05 (as calculated with Quanto software).



Jewish Israeli sample

None of the SNPs showed deviation from Hardy–Weinberg equilibrium. In the TD yes/no phenotype, the KCNB2 intronic SNP, rs4638269, showed a trend for association with TD at a nominal level of significance (P=0.056; Table 2a). The direction of association (risk allele is G) was opposite to that reported in the Japanese sample (risk allele is A). In the TD extreme phenotype, the HSPG2 SNP, rs2445142, was significantly associated with TD (P=0.03), with G as the risk allele (as in the original Japanese report). In the TD yes/no analysis, only a marginal trend for association was seen (P=0.13) for this SNP. Additionally, nominally significant association was demonstrable for GABRG3 SNP, rs2061051 (P=0.01), but in the opposite direction to the association in the Japanese sample (risk allele is A, but G in the Japanese study). As was the case in the Japanese report, the two HSPG2 SNPs (rs2124368 and rs2445142, separated by ~40kb) were not in LD (D=0.79, r2=0.04). Association analysis of dyplotypes composed of these two SNPs did not yield more significant results than at the single marker level.

CATIE TD sample

SNP rs878949, a surrogate with r2=1 in the CEU HapMap sample for Syu et al.30 marker rs2445142 in HSPG2, was associated with TD in the European origin CATIE TD sample at a nominal level of significance. The major allele T was the TD risk allele (P=0.039; Table 2b). As rs878949-T allele represents rs2445142-G allele in Europeans, the direction of association is the same in the European, Israeli and Japanese samples (where TD risk allele is G). This SNP was not in LD with the other HSPG2 SNP, rs2124368 (D=0.64, r2=0.04). Again, dyplotype analysis did not improve significance level compared with single marker level. No significant associations were found for the other three SNPs studied. None of the SNPs showed deviation from Hardy–Weinberg equilibrium.



The main finding of this study is support for the association of SNP rs2445142 in the first intron of HSPG2 with susceptibility to TD. Following a two-step TD genome-wide association screening in Japanese schizophrenia patients,30 we performed the first reported replication test in two independent samples of Caucasian origin. We found association of the rs878949 SNP (surrogate marker for rs2445142) with TD in the European origin CATIE TD sample. Moreover, association of rs2445142 was nominally significant in the Jewish sample, when applying the extreme phenotype definition, whereas a trend was observed when using the TD yes/no classification. The direction of association in the two Caucasian samples was identical to the original direction reported in the Japanese population (G is risk allele). Interestingly, the rs2445142 risk allele was associated with higher level of HSPG2 expression in post-mortem prefrontal cortex in both Australian and Japanese samples.30 Daily injection of haloperidol for 50 weeks reduced HSPG2 expression in the mouse brain, interpreted as an adaptive or compensatory response to antipsychotics. Moreover, adult Hspg2 heterozygous knockout mice demonstrated significant reduction of vacuous chewing movement when compared with wild-type mice, following chronic treatment with a haloperidol–reserpine combination.30 Therefore, Syu et al.30 hypothesized that decreased HSPG2 expression is protective for TD, whereas increased HSPG2 expression level may induce susceptibility to TD, by a mechanism unknown at present. We were not able to support association of the second HSPG2 risk variant reported in the Japanese sample (rs2124368), which is not in LD with rs2445142 in both Japanese and Caucasian populations.

Inclusion of important clinical variables as covariates in the regression model of the two analyzed samples is a clear advantage of the present study. In addition to gender, older age, and second-generation antipsychotic treatment that lower TD rates,45, 46 several studies have described ethnic differences in TD susceptibility.47, 48 These possible confounders of the contribution of genetic variables to TD were controlled by the study of genetically homogenous populations and by our logistic regression. As in both samples data on cumulative lifetime exposure to antipsychotics were not available, we subtracted the number of years since first treatment with antipsychotic drugs from the age at TD assessment. This proxy variable for cumulative lifetime exposure to antipsychotics was also included in the regression model. Further advantage is the fact that we analyzed two independent replication samples, each of them larger in terms of participant numbers than the original Japanese sample used for the discovery phase (100 patients).

In addition, we observed association of the GABRG3 SNP, rs2061051, and trend for association of KCNB2 SNP, rs4638269, with TD in the Jewish population (extreme and broad phenotype, respectively), but not in the European CATIE TD sample (rs2061051 was not studied there because of lack for appropriate surrogate marker). However, the association for both SNPs was in the opposite direction to the direction reported in the Japanese population. These findings might be false positive, or represent ‘flip-flop’ association.49 This term refers to a situation where the genetic effect direction of a SNP in the validation sample is opposite to the direction originally reported in the discovery sample.50 Replicating association of a noncausal allele in LD with causative variants in two different populations (in our case, Japanese and Jewish) may lead to inconsistent directions because of differences in LD patterns.49, 50 Another possible explanation for the ‘flip-flop’ may be existence of gene–gene or gene–environment interactions not taken into account, as suggested by Lin et al.49 Future research is required to determine if rs4638269 and rs2061051 associations with TD are indeed a confirmatory of a true locus effect, or a spurious finding.

Although the evaluation tool in all the samples was the AIMS, the TD phenotype definitions were different between samples. In the original Japanese report,30 the phenotype was rigorously defined as treatment-resistant TD, a severe form of TD. Cases had TD persisting for more than a year, which did not improve after drug therapy, whereas controls had never developed TD despite more than 10 years of antipsychotic exposure.31 This phenotype is conceptually similar to our definition of TD in the prospective European CATIE TD sample, where we applied the extreme distribution of phenotype approach. Because of the prolonged clinical surveillance (453 days on average), we defined cases as those who met TD criteria on at least two separate assessments. Controls were completely devoid of any abnormal involuntary movements during the entire CATIE participation period (research phases 1–3). It has been suggested that focusing on the extremes of a sample distribution (as done in both the Japanese sample and the CATIE TD sample) is one of the most advantageous strategies in conducting pharmacogenetic GWASs.51

To be consistent with the extreme phenotype distribution approach on one hand, but not to lose large number of participants on the other hand, we analyzed the Jewish sample using two different definitions of controls: those who showed some dyskinetic movement but did not reach the RDC-TD level, and those who were devoid of any involuntary abnormal movements, representing a ‘cleaner’ phenotype (extreme controls). Defining cases and controls as the extremes of a larger distribution is based on the assumption that the majority of patients manifest only partial therapeutic response or adverse effects (in our case, minimal involuntary movements manifestations, not reaching the required TD cutoff).51 The accuracy of phenotyping patients in the extremes of response distribution is likely to be the highest. Therefore, contrasting extremely good responders vs extremely poor responders maximizes statistical power and was used, for example, by Turner et al.52 in their pharmacogenetic study of hypertension. Moreover, genotyping of individuals in the extremes of distribution may provide nearly equivalent power to a complete sample.51

The CATIE TD sample was based on prospective design with multiple evaluations per patients, whereas the retrospective Israeli study employed a cross-sectional design and patients were assessed for TD only once.21, 22, 32, 33 Therefore, a further consideration is the use of the potentially highly valuable extreme phenotype approach in this cohort sample. Because of the fluctuating nature of TD, and the fact that TD may be suppressed by increasing antipsychotics doses, it is possible that for some patients the classification may not be accurate. This is a valid criticism and should be noted. Although classification based on more than a single measurement (as in the longitudinal CATIE sample) would have been preferable, these data were not available. Thus, the diagnosis of TD is at a probable level according to the RDC-TD criteria.33

An additional limitation is that in the European CATIE TD association analysis, the risk markers were not genotyped directly (as DNA was not available for us), and the study was performed by analyzing surrogate SNPs. Nevertheless, rs878949 is a perfect proxy for rs2445142 (D=1, r2=1) and the rs2445142-G allele is equivalent to rs878949-T allele according to CEU HapMap sample. The correlation between rs4738269 and it surrogate rs7009604 in the CEU sample was also very high (D=1, r2=0.87), although not complete. The necessity to use surrogates forced us to include only patients of European origin (excluding Hispanic) in the CATIE-TD sample, because of substantial differences in r2 values between European, African and Asian populations. We acknowledge that the CEU HapMap sample may not fully and accurately represent this specific CATIE TD European sample, which was recruited from multiple sites in the United States, but this strategy is currently the widely acceptable one for choosing surrogate markers. Last, we could not control for inter-rater reliability between raters of the CATIE TD sample, and the raters of the Jewish Israeli population. This could introduce a bias because raters may differ in the way they assess the severity of dyskinetic movements. This limitation is partially balanced by focusing on phenotypes from extremes of distribution, intuitively easier to rate and classify than middle range cases with partial response.

All the reported P-values were not corrected for multiple testing. If we had implemented Bonferroni correction, the required P-value would have been 0.003. Nevertheless, we consider at least the rs2445142 association as a true positive. According to the criteria of Sullivan53 and van den Oord et al.54 in a replication trial, an uncorrected standard P-value of <0.05 may be used, if association is for the same SNP and phenotype and the direction of effect is the same as in the original report.

The HSPG2 gene, encoding the perlecan proteoglycan, is an emerging candidate for TD. HSPG2 is expressed in basement membranes in several tissues, including brain.55 It has been previously associated with Alzheimer's disease56 and intracranial aneurysms.57 As suggested by Syu et al.,30 its involvement in TD pathophysiology may be related to its influence on cholinergic transmission. Cholinergic medication is used for TD symptom amelioration.58 Perlecan is involved in localization of the acetylcholinesterase enzyme, which degrades acetylcholine, to the neuromuscular junction.59 Therefore, perlecan may affect the levels of acetylcholinesterase in the neuromuscular synapse, indirectly modulate cholinergic neurotransmission and contribute to TD susceptibility.

In conclusion, we have presented the results of a replication study of a two-tier Japanese genome-wide association screening for TD, in two Caucasian TD case–control samples. Support for association of the HSPG2 intronic SNP, rs2445142, with TD susceptibility was demonstrated, but further studies are required.


Conflict of interest

The authors declare no conflict of interest.



  1. Blanchet PJ. Antipsychotic drug-induced movement disorders. Can J Neurol Sci 2003; 30: S101–S107. | PubMed |
  2. Haddad PM, Dursun SM. Neurological complications of psychiatric drugs: clinical features and management. Hum Psychopharmacol 2008; 23(Suppl): 15–26. | Article | PubMed |
  3. Glazer WM, Morgenstern H, Doucette JT. Predicting the long-term risk of tardive dyskinesia in outpatients maintained on neuroleptic medications. J Clin Psychiatry 1993; 54: 133–139. | PubMed | ISI |
  4. de Leon J. The effect of atypical versus typical antipsychotics on tardive dyskinesia: a Naturalistic Study. Eur Arch Psychiatry Clin Neurosci 2007; 257: 169–172. | Article | PubMed | ISI |
  5. Remington G. Tardive dyskinesia: eliminated, forgotten, or overshadowed? Curr Opin Psychiatry 2007; 20: 131–137. | Article | PubMed | ISI |
  6. Correll CU, Schenk EM. Tardive dyskinesia and new antipsychotics. Curr Opin Psychiatry 2008; 21: 151–156. | Article | PubMed | ISI |
  7. Yassa R, Jeste DV. Gender differences in tardive dyskinesia: a critical review of the literature. Schizophrenia Bull 1992; 18: 701–715.
  8. Kane JM. Tardive dyskinesia: epidemiological and clinical presentation. In: Bloom FE, Kupfer DJ (eds). Psychopharmacology: The 4th Generation of Progress. Raven Press: New York, 1995.
  9. Tenback DE, van Harten PN, van Os J. Non-therapeutic risk factors for onset of tardive dyskinesia in schizophrenia: a meta-analysis. Mov Disord 2009; 24: 2309–2315. | Article | PubMed |
  10. Lerer B, Segman RH. Pharmacogenetics of antipsychotic therapy: pivotal research issues and the prospects for clinical implementation. Dialogues Clin Neurosci 2006; 8: 85–94. | PubMed |
  11. Muller DJ, Schulze TG, Knapp M, Held T, Krauss H, Weber T et al. Familial occurrence of tardive dyskinesia. Acta Psychiatr Scand 2001; 104: 375–379. | Article | PubMed | ISI | CAS |
  12. Ismail B, Cantor-Graae E, McNeil TF. Neurodevelopmental origins of tardive like dyskinesia in schizophrenia patients and their siblings. Schizophr Bull 2001; 27: 629–641. | Article | PubMed |
  13. Egan MF, Apud J, Wyatt RJ. Treatment of tardive dyskinesia. Schizophr Bull 1997; 23: 583–609. | Article | PubMed | CAS |
  14. Sagara Y. Induction of reactive oxygen species in neurons by haloperidol. J Neurochem 1998; 71: 1002–1012. | Article | PubMed | ISI | CAS |
  15. Naidu PS, Singh A, Kulkarni SK. Carvedilol attenuates neuroleptic-induced orofacial dyskinesia: possible antioxidant mechanisms. Br J Pharmacol 2002; 136: 193–200. | Article | PubMed |
  16. Delfs JM, Ellison GD, Mercugliano M, Chesselet MF. Expression of glutamic acid decarboxylase mRNA in striatum and pallidum in an animal model of tardive dyskinesia. Exp Neurol 1995; 133: 175–188. | Article | PubMed |
  17. Sakai K, Gao XM, Hashimoto T, Tamminga CA. Traditional and new antipsychotic drugs differentially alter neurotransmission markers in basal ganglia-thalamocortical neural pathways. Synapse 2001; 39: 152–160. | Article | PubMed |
  18. Fenton WS. Prevalence of spontaneous dyskinesia in schizophrenia. J Clin Psychiatry 2000; 61(Suppl 4): 10–14. | PubMed | ISI |
  19. Pappa S, Dazzan P. Spontaneous movement disorders in antipsychotic-naive patients with first-episode psychoses: a systematic review. Psychol Med 2009; 39: 1065–1076. | Article | PubMed |
  20. Koning JP, Tenback DE, van Os J, Aleman A, Kahn RS, van Harten PN. Dyskinesia and parkinsonism in antipsychotic-naive patients with schizophrenia, first-degree relatives and healthy controls: a meta-analysis. Schizophr Bull 2010; 36: 723–731. | Article | PubMed |
  21. Lerer B, Segman RH, Fangerau H, Daly AK, Basile VS, Cavallaro R et al. Pharmacogenetics of tardive dyskinesia: combined analysis of 780 patients supports association with dopamine D3 receptor gene Ser9Gly polymorphism. Neuropsychopharmacology 2002; 27: 105–119. | Article | PubMed | ISI | CAS |
  22. Lerer B, Segman RH, Tan EC, Basile VS, Cavallaro R, Aschauer HN et al. Combined analysis of 635 patients confirms an age-related association of the serotonin receptor gene with tardive dyskinesia and specificity for the non-orofacial subtype. Int J Neuropsychopharmacol 2005; 8: 411–425. | Article | PubMed | ISI | CAS |
  23. Bakker PR, van Harten PN, van Os J. Antipsychotic-induced tardive dyskinesia and the Ser9Gly polymorphism in the DRD3 gene: a meta analysis. Schizophr Res 2006; 83: 185–192. | Article | PubMed | ISI |
  24. Bakker PR, van Harten PN, van Os J. Antipsychotic-induced tardive dyskinesia and polymorphic variations in COMT, DRD2, CYP1A2 and MnSOD genes: a meta-analysis of pharmacogenetic interactions. Mol Psychiatry 2008; 13: 544–556. | Article | PubMed | ISI | CAS |
  25. Patsopoulos NA, Ntzani EE, Zintzaras E, Ioannidis JP. CYP2D6 polymorphisms and the risk of tardive dyskinesia in schizophrenia: a meta-analysis. Pharmacogenet Genomics 2005; 15: 151–158. | Article | PubMed | CAS |
  26. Tiwari AK, Deshpande SN, Lerer B, Nimgaonkar VL, Thelma BK. Genetic susceptibility to Tardive Dyskinesia in chronic schizophrenia subjects: V. Association of CYP1A2 1545 C>T polymorphism. Pharmacogenomics J 2007; 7: 305–311. | Article | PubMed |
  27. Tsai HT, Caroff SN, Miller DD, McEvoy J, Lieberman JA, North KE et al. A candidate gene study of tardive dyskinesia in the CATIE schizophrenia trial. Am J Med Genet B Neuropsychiatr Genet 2010; 153B: 336–340.
  28. Aberg K, Adkins DE, Bukszár J, Webb BT, Caroff SN, Miller del D et al. Genomewide association study of movement-related adverse antipsychotic effects. Biol Psychiatry 2010; 67: 279–282. | Article | PubMed | ISI |
  29. Greenbaum L, Alkelai A, Rigbi A, Kohn Y, Lerer B. Evidence for association of the GLI2 gene with tardive dyskinesia in chronic schizophrenia patients. Mov Disord 2010; 25: 2809–2817. | Article | PubMed | ISI |
  30. Syu A, Ishiguro H, Inada T, Horiuchi Y, Tanaka S, Ishikawa M. Association of the HSPG2 gene with neuroleptic-induced tardive dyskinesia. Neuropsychopharmacology 2010; 35: 1155–1164. | Article | PubMed | ISI |
  31. Inada T, Koga M, Ishiguro H, Horiuchi Y, Syu A, Yoshio T et al. Pathway-based association analysis of genome-wide screening data suggest that genes associated with the gamma-aminobutyric acid receptor signaling pathway are involved in neuroleptic-induced, treatment-resistant tardive dyskinesia. Pharmacogenet Genomics 2008; 18: 317–323. | Article | PubMed | ISI | CAS |
  32. Segman R, Neeman T, Heresco-Levy U, Finkel B, Karagichev L, Schlafman M et al. Genotypic association between the dopamine D3 receptor and tardive dyskinesia in chronic schizophrenia. Mol Psychiatry 1999; 4: 247–253. | Article | PubMed | ISI | CAS |
  33. Segman RH, Heresco-Levy U, Finkel B, Goltser T, Shalem R, Schlafman M et al. Association between the serotonin 2A receptor gene and tardive dyskinesia in chronic schizophrenia. Mol Psychiatry 2001; 6: 225–229. | Article | PubMed | ISI | CAS |
  34. Guy W (1976). ECDEU Assessment Manual for Psychopharmacology, Revised edn. Department of Health, Education and Welfare: Washington DC.
  35. Schooler NR, Kane JM. Research diagnoses for tardive dyskinesia. Arch Gen Psychiatry 1982; 39: 486–487. | Article | PubMed | ISI | CAS |
  36. Steen VM, Llie R, McEwan T, McCreadie RG. Dopamine D3 receptor gene variant and susceptibility to tardive dyskinesia in schizophrenic patients. Mol Psychiatry 1997; 2: 139–145. | Article | PubMed | ISI | CAS |
  37. Andreassen OA, MacEwan T, Gulbrandsen AK, McCreadie RG, Steen VM. Non-functional CYP2D6 alleles and risk for neuroleptic-induced movement disorders in schizophrenic patients. Psychopharmacology (Berl) 1997; 131: 174–179. | Article | PubMed | CAS |
  38. Løvlie R, Daly AK, Blennerhassett R, Ferrier N, Steen VM. Homozygosity for the Gly-9 variant of the dopamine D3 receptor and risk for tardive dyskinesia in schizophrenic patients. Int J Neuropsychopharmacol 2000; 3: 61–65. | Article | PubMed | ISI | CAS |
  39. 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. Am J Hum Genet 2007; 81: 559–575. | Article | PubMed | ISI | CAS |
  40. Lieberman JA, Stroup TS, McEvoy JP, Swartz MS, Rosenheck RA, Perkins DO et al. Effectiveness of antipsychotic drugs in patients with chronic schizophrenia. N Engl J Med 2005; 353: 1209–1223. | Article | PubMed | ISI | CAS |
  41. Sullivan PF, Lin D, Tzeng JY, van den Oord E, Perkins D, Stroup TS et al. Genomewide association for schizophrenia in the CATIE study: results of stage 1. Mol Psychiatry 2008; 13: 570–584. | Article | PubMed | ISI | CAS |
  42. Stroup TS, McEvoy JP, Swartz MS, Byerly MJ, Glick ID, Canive JM et al. The National Institute of Mental Health Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) project: schizophrenia trial design and protocol development. Schizophr Bull 2003; 29: 15–31. | Article | PubMed | ISI |
  43. Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, Reich D. Principal components analysis corrects for stratification in genome-wide association studies. Nat Genet 2006; 38: 904–909. | Article | PubMed | ISI | CAS |
  44. Reich D, Price AL, Patterson N. Principal component analysis of genetic data. Nat Genet 2008; 40: 646–649. | Article | PubMed | ISI | CAS |
  45. Nasrallah HA. Focus on lower risk of tardive dyskinesia with atypical antipsychotics. Ann Clin Psychiatry 2006; 18: 57–62. | Article | PubMed |
  46. Pierre JM. Extrapyramidal symptoms with atypical antipsychotics: incidence, prevention and management. Drug Saf 2005; 28: 191–208. | Article | PubMed | CAS |
  47. Chong SA, Mahendran R, Machin D, Chua HC, Parker G, Kane J. Tardive dyskinesia among Chinese and Malay patients with schizophrenia. J Clin Psychopharmacol 2002; 22: 26–30. | Article | PubMed |
  48. Bhatia T, Sabeeha MR, Shriharsh V, Garg K, Segman RH, Uriel HL et al. Clinical and familial correlates of tardive dyskinesia in India and Israel. J Postgrad Med 2004; 50: 167–172. | PubMed |
  49. Lin PI, Vance JM, Pericak-Vance MA, Martin ER. No gene is an island: the flip-flop phenomenon. Am J Hum Genet 2007; 80: 531–538. | Article | PubMed | ISI | CAS |
  50. Zaykin V, Shibata K. Genetic flip-flop without an accompanying change in linkage disequilibrium. Am J Hum Genet 2008; 82: 794–796. | Article | PubMed |
  51. Crowley JJ, Sullivan PF, McLeod HL. Pharmacogenomic genome-wide association studies: lessons learned thus far. Pharmacogenomics 2009; 10: 161–163. | Article | PubMed |
  52. Turner ST, Bailey KR, Fridley BL, Chapman AB, Schwartz GL, Chai HS et al. Genomic association analysis suggests chromosome 12 locus influencing antihypertensive response to thiazide diuretic. Hypertension 2008; 52: 359–365. | Article | PubMed | CAS |
  53. Sullivan PF. Spurious genetic associations. Biol Psychiatry 2007; 61: 1121–1126. | Article | PubMed | ISI | CAS |
  54. van den Oord EJ, Kuo PH, Hartmann AM, Webb BT, Möller HJ, Hettema JM et al. Genomewide association analysis followed by a replication study implicates a novel candidate gene for neuroticism. Arch Gen Psychiatry 2008; 65: 1062–1071. | Article | PubMed | ISI |
  55. Snow AD, Sekiguchi R, Nochlin D, Fraser P, Kimata K, Mizutani A et al. An important role of heparan sulfate proteoglycan (Perlecan) in a model system for the deposition and persistence of fibrillar A beta-amyloid in rat brain. Neuron 1994; 12: 219–234. | Article | PubMed | ISI | CAS |
  56. Rosenmann H, Meiner Z, Kahana E, Aladjem Z, Friedman G, Ben-Yehuda A et al. An association study of a polymorphism in the heparan sulfate proteoglycan gene (perlecan, HSPG2) and Alzheimer's disease. Am J Med Genet B Neuropsychiatr Genet 2004; 128B: 123–125.
  57. Ruigrok YM, Rinkel GJ, Wijmenga C, Kasuya H, Tajima A, Takahashi T et al. Association analysis of genes involved in the maintenance of the integrity of the extracellular matrix with intracranial aneurysms in a Japanese cohort. Cerebrovasc Dis 2009; 28: 131–134. | Article | PubMed |
  58. Tammenmaa IA, McGrath JJ, Sailas E, Soares-Weiser K. Cholinergic medication for neuroleptic-induced tardive dyskinesia. Cochrane Database Syst Rev 2002; 3: CD000207. | PubMed |
  59. Rotundo RL, Rossi SG, Kimbell LM, Ruiz C, Marrero E. Targeting acetylcholinesterase to the neuromuscular synapse. Chem Biol Interact 2005; 157–158: 15–21.


This study was supported in part by a grant from the Michael J Fox Foundation (to BL). The principal investigators of the CATIE (Clinical Antipsychotic Trials of Intervention Effectiveness) trial were Jeffrey A Lieberman, T Scott Stroup, and Joseph P McEvoy. The CATIE trial was funded by a grant from the National Institute of Mental Health (N01 MH900001) along with MH074027 (PI PF Sullivan). Genotyping was funded by Eli Lilly and Company.