0.99, df = 4) and in controls (chi2 = 6.29, df = 4, P = 0.18). A new allele with one more GT repeat (176 bp) was found. In a Japanese population, an insertion of.
Molecular Psychiatry (2000) 5, 558–562 2000 Macmillan Publishers Ltd All rights reserved 1359-4184/00 $15.00 www.nature.com/mp
ORIGINAL RESEARCH ARTICLE
Brain Derived Neurotrophic Factor (BDNF) gene variants association with age at onset and therapeutic response in schizophrenia MO Krebs1, O Guillin1,2, MC Bourdel1, JC Schwartz2, JP Olie1, MF Poirier1 and P Sokoloff2 1
Biological Psychiatry Laboratory (UPRES EA 2501, University of Paris V), Department of Mental Health and Therapeutic, Sainte-Anne Hospital, Paris, France; 2Department of Neurobiology and Molecular Pharmacology (U109, INSERM), Paris, France
Keywords: association study; neurodevelopment; abuse; dopamine D3 receptor; neurotrophin
Schizophrenia is a heterogeneous disease involving genetic and environmental factors. The frequency of structural brain abnormalities1,2 or physical anomalies3 supports a neurodevelopmental etiology, especially in early onset schizophrenia.4 Brain-Derived-NeurotrophicFactor (BDNF) is involved in the neurodevelopment of dopaminergic (DA)-related systems5,6 and interacts with the meso-limbic DA systems,7–10 involved in the therapeutic response to antipsychotic drugs11 and substance abuse.12 In addition, BDNF promotes and maintains dopamine D3 receptor (DRD3) expression.13 In a French Caucasian population, we found no statistical difference in allele or genotype distribution of the BDNF gene dinucleotide repeat polymorphism (166–174 bp)14 between the whole group of schizophrenic patients and controls. By contrast, an excess of the 172–176 bp alleles was found in patients with late onset, in neurolepticresponding patients and in non-substance-abusing patients. BDNF gene variants thus appear to be associated with developmental features of schizophrenia. In addition, this association with good treatment responding was independent from the association found with the DRD3 BalI gene polymorphism in the same population.15 These results suggest an independent contribution of each gene to a treatment-sensitive form of schizophrenia. Molecular Psychiatry (2000) 5, 558–562.
Eighty-eight unrelated in- or out-patients with schizophrenia or schizo-affective disorder (DSM-III-R criteria) and fifty-two unrelated age- and gender-matched controls with no personal and/or familial history of psychiatric disease or substance abuse were recruited from a French Caucasian population. The genotype distribution did not show a departure from Hardy–Weinberg equilibrium in either controls (chi2 = 2.90; df = 9; P = 0.97) or patients (chi2 = 7.10; df = 9; P = 0.63). The allelic distribution (Table 1) was similar to that reported16 in patients (chi2 = 1.18, P = 0.99, df = 4) and in controls (chi2 = 6.29, df = 4, P = 0.18). A new allele with one more GT repeat (176 bp) was found. In a Japanese population, an insertion of nucleotides adjacent to the dinucleotide polymorphic
site was found in four subjects.17 Alleles with lesser repeats (164 bp and 154 bp) have also been reported.16 As in previous reports,16,17 we did not find significant differences in BDNF gene allele (WI = 9960; P = 0.14) or genotype distribution (chi2 = 5.81; df = 9; P = 0.75) between the group of schizophrenic patients as a whole and controls, but a false negative result can not be ruled out, due to the limited size of the sample. Negative results were also found in a larger sample (234 controls, 220 patients),16 in which, unfortunately, no additional clinical features were taken into account. Since schizophrenia is a disease with heterogeneous clinical expression and BDNF is a critical factor in neurodevelopment, in particular of the mesolimbic DA system, we analyzed allelic distribution according to additional clinical features, ie age at onset, therapeutic response to antipsychotic drugs and substance abuse. Schizophrenic patients with an age at first contact with a practitioner for psychiatric reason later than 25 years old (n = 20) had a significantly different allelic distribution compared with either patients with first contact at 25 years or below (WI = 3311, P = 0.007) or controls (WI = 1553, P = 0.005). For statistical analysis purposes, alleles were grouped in ‘short’ (166–170 bp) and ‘long’ (172–176 bp) alleles. Patients with late onset had an excess of long alleles as compared to patients with an age at onset before 25 years (chi2 = 5, P = 0.03, odds ratio (OR) = 2.3). Accordingly, patients with the ‘long’ alleles had an age at onset significantly higher than the remaining patients (23.92 ± 5 vs 19.34 ± 7 years old, WI = 641.0, P = 0.024). Allele distribution significantly differed in patients who at least partly responded to neuroleptic treatment (‘treatment-responding’) as compared to patients with no clinical remission (‘refractory’ patients, n = 20; WI = 2106; P = 0.01) and controls (WI = 8061; P = 0.03). Treatment-responding patients had an excess of long alleles (172–176 bp) as compared to refractory schizophrenic patients (chi2 = 4.6, P = 0.04, OR = 2.7). In contrast with previous studies, in our sample, age at onset in treatment-responding patients was not significantly different from that of refractory patients (20.4 ± 7 vs 18.5 ± 7, WI = 757, P = 0.43) and treatment responding patients were equally distributed among
BDNF gene variants in schizophrenia MO Krebs et al
Table 1 Allele frequencies for the dinucleotide repeat polymorphism at the human BDNF gene for patients with schizophrenia and controls n
Controls Schizophrenic patients (Sz) (all) Late onset Sz Early onset Sz Treatment-responding Sz Treatment-refractory Sz Substance-abusing Sz Sz with no history of substance abuse
52 88 20 68 68 20 36 52
Mann–Whitney (df = 9)
174 bp (A1)
172 bp (A2)
170 bp (A3)
168 bp (A4)
166 bp (A5)
0.96% 0.57% 2.5% 0% 0.74% 0% 0% 0.94%
18.3% 25% 35% 22.1% 27.9% 15% 20% 28.3%
1.9% 2.8% 5% 2.2% 3.7% 0% 1.4% 3.8%
68.3% 64.2% 55% 66.9% 62.5% 70% 68.6% 61.3%
9.6% 6.8% 2.5% 8.1% 4.4% 15% 8.6% 5.7%
0.96% 0.57% 0% 0.7% 0.74% 0% 1.4% 0%
P = 0.14a P = 0.005b P = 0.007c P = 0.01d P = 0.03e P = 0.03f P = 0.07g
Statistical comparisons were: acontrol vs schizophrenic patients; blate onset vs controls; clate vs early onset; dtreatmentresponding Sz vs controls; etreatment-responding vs treatment refractory; fSz patients with no substance abuse vs controls; gSz patients with vs without history of substance abuse.
patients with age at onset before or after 25 years old (Yates corrected chi2 = 0.23). In fact, in the literature, treatment-sensitive ‘late onset’ schizophrenia refers to a particular form of schizophrenia with first prodromal or acute psychotic symptoms later than 40–45 years,18 that was not present in our sample. In the absence of interaction between treatment response and age at onset in our sample, the difference in allelic distribution seen with regard to treatment response can be attributed to an association with treatment responsiveness itself. Patients with combined feature of treatment response and late onset (n = 36) had an excess of the long alleles (172–176 bp) as compared to the remaining patients (P = 0.007, OR = 2.94). This is consistent with the known involvement of BDNF in the development of the DA systems. It should be interesting to further study samples including very late-onset schizophrenia (ie first symptoms later than 45 years old). Allele distribution in ‘non substance abusing’ patients significantly differed from that in controls (WI = 6299; P = 0.03) and tended to differ from that in patients who had at least once abused or were dependent on any kind of psychoactive substance (‘substance abusing patients’, n = 36; WI = 4221; P = 0.07). As a tendency, non substance-abusing patients had an excess of the long alleles (172–176 bp) as compared to controls (chi2 = 3.7, P = 0.06). This was not due to an unequal distribution of substance-abusing patients in treatment-responding or refractory subgroups (chi2 = 0.25, P = 0.62) or in subgroups defined by age at onset before or later than 25 years (chi2 = 1.03, P = 0.44). In addition, age at onset was not significantly different in non-substance-abusing compared to substance-abusing patients (20.4 ± 7 vs 19.3 ± 7 years of age, WI = 989, P = 0.6). This suggests that the BDNF gene variants could influence the phenotypic expression of schizophrenia in relation to predisposition to substance abuse. It is not clear whether this is specific to schizophrenia and it should be further documented in a sample of drug addicts. Interestingly, BDNF is up-regulated during chronic opiate treatment and withdrawal in rats.19
Using the same clinical characteristics in the same population of patients with schizophrenia, we previously reported that homozygosity at the DRD3 BalI polymorphism was associated with substance abuse and responsiveness to neuroleptics. There was no statistical difference with regards to age at onset of schizophrenia (unpublished results). In that case, the combined features of substance abuse and responsiveness to treatment were associated with homozygosity for the DRD3 BalI polymorphism. We thus tested whether there was an interaction of DRD3 and BDNF (long vs short) gene variants in this association. DRD3 heterozygote (1–2) and homozygote (2–2 or 1– 1) genotypes were equally distributed between the BDNF genotypes with at least one long allele (172– 176 bp) vs those with two short alleles, indicating that the association found here with good therapeutic response was not a false positive effect due to a bychance stratification in DRD3 homozygosity distribution (chi2 = 0.001, P = 0.98, OR = 1.01). The logistic regression including the effect of the BDNF and DRD3 gene variants and a possible interaction effect of the two showed no significant interaction between the two genes (Wald test = 0.25; P = 0.62). This interaction term was thus dropped in the subsequent analysis. The effect of the DRD3 gene was then significant (adjusted OR = 3.5; Wald test = 4.85, P = 0.028) and the effect of the BDNF gene was almost significant (adjusted OR = 3.0; Wald test = 3.42; P = 0.06). These results indicate that each of the two polymorphisms contributes separately to the therapeutic response. BDNF and DRD3 gene variants appear thus to be associated with distinct schizophrenic phenotypes, of developmental and pharmacological nature respectively. Namely, the associations of the BalI polymorphism DRD3 gene variants with tardive dyskinesia20,21 and with treatment response15 in schizophrenic patients, suggest that DRD3 gene variants influence the pharmacological sensitivity to antipsychotic drugs. On the other hand, excess of the BDNF gene ‘long’ alleles is found in patients with age at onset later than 25 years Molecular Psychiatry
BDNF gene variants in schizophrenia MO Krebs et al
and treatment responsiveness, clinical features that are associated with a ‘less developmentally weighted’ form of schizophrenia. Such an association might be found in other psychiatric diseases in which neurodevelopmental factors have been suggested. Our observation of an excess of the long alleles in non substance-abusing patients is consistent with the involvement of BDNF in the development of the DA mesolimbic ‘reward’ system, involved in drug abuse and dependence. The functional significance of the BDNF gene polymorphism is unclear. Since it is located 1040 bp upstream of the transcription initiation site, an effect on the regulatory components of this gene cannot be excluded but remains to be demonstrated. The analysis of the respective effect of DRD3 BalI polymorphism and BDNF alleles indicates that each of these genetic markers is independently associated with therapeutic response in schizophrenia. Since BDNF presumably controls DRD3 gene expression,13 it is conceivable that alterations in the function of each gene independently contribute to a DRD3 over-expression, which was found in post-mortem brain of patients with schizophrenia.22 The association of variants of each of these genes with the therapeutic response in schizophrenia is indeed consistent with the fact that antipsychotic drugs act at the DRD3,23 and may therefore normalize the transmission through this receptor in the subset of treatment-sensitive patients. This would be the first example of independent effects of two functionally interacting genes in schizophrenia, in agreement with the presumed polygenic inheritance of the susceptibility to this disorder, each gene having a small contribution.24
Materials and methods Subjects and psychiatric assessment Eighty-eight unrelated patients (55 males; 33 females; mean age: 33.3 ± 9.1 years old) with schizophrenia (n = 84) or schizoaffective disorder (n = 4) were examined by a trained psychiatrist with a semi-standardized interview (SADS-LA lifetime version anxiety version – revised,25 French translation) leading to lifetime diagnosis according to DSM-III-R criteria.26 In addition, information on clinical and treatment history was obtained from the medical staff of our hospital or from previous hospitalizations (most of the patients have been treated in our department from the beginning of their disease or less than 2 years after the onset of the disease), ambulatory private psychiatrists, review of case notes and, whenever possible, information from the family. The majority of patients (77%) had a positive familial history of psychiatric disease (second degree), with 50 patients (57%) having a familial history of psychosis (first or second degree). Psychiatric assessment also included evaluation of the course of illness (mean duration of the disease ± SD = 13.3 ± 10 years), current symptomatology with Positive and Negative Syndrome Scale27 (PANSS mean total score: 80.8 ± 18). The age at first contact with medical practitioner for Molecular Psychiatry
psychological or psychiatric reason was used as age at onset (mean ± SD: 20.0 ± 6.9 years old). Late or early onset of schizophrenia was defined on the basis of a first contact before or after the age of 25. This cut-off of 25 was chosen because, in most cases, the first symptoms (whether specific or not) occur at or before the age of 25.28 It fitted approximately to the mean + one SD of our sample. Patients with an age at onset before or at the age of 25 years corresponded to 77% of the studied population (n = 68). The same cut-off was chosen in male and female, since the mean age at onset was similar in both genders in our population in agreement of previous reports on the familial form of schizophrenia.18,29 Response to neuroleptic treatment was evaluated on the basis of the rapidity and quality of symptom remission, necessity for prolonged hospitalization, ability for autonomy and social rehabilitation.30 ‘Treatment-responding’ patients (n = 68, score: 1–4) had at least a partial clinical remission under antipsychotic treatment allowing discharge from hospital. Conversely, ‘treatment refractory’ patients (n = 20, score: 5–6) had no clinical remission despite several trials with different antipsychotic drugs for a sufficient duration and required permanent day care. Former or current substance abuse of and/or dependence on any kind of psychoactive substance including alcohol or tranquilizers, but not nicotine, were systematically screened. Patients who had at least once in their life fulfilled the DSM-III-R criteria for abuse or dependence (but not occasional use) were defined as ‘substance-abusing’ (n = 36). ‘Non substance-abusing’ schizophrenic patients (n = 52) had never, in all their life, fulfilled the DSM-IIIR criteria for abuse and/or dependence on any kind of psychoactive substance. In every case, substance-induced psychosis could be eliminated: in only four patients, the interview found that substance abuse preceded psychotic or prodromal symptoms, but psychotic symptoms remained much more than 6 weeks after the end of intoxication and the abuse was of moderate extent and/or limited duration. The fifty-two unrelated controls (26 males, 26 females) were locally recruited (medical and non medical staff of the hospital, medical students) and screened with a semi-standardized interview (SCID-NP) and additional questions about family history (first and second degree relatives), done by a psychologist independent of the study. They were included only when they had no personal and/or familial psychiatric or substance abuse history. Patients did not differ significantly from controls with regard to gender distribution (P = 0.17) or ages (35.8 ± 10 years old in controls, P = 0.36). All subjects but four (two patients and two controls) had at least three French Caucasian grandparents (with the possibility of one grandparent originating from another country of Western Europe, except two patients with one grand-parent of Caribbean origin). The results were unchanged when excluding those six subjects. The psychiatric assessment was completed before genetic analysis and the choice for analysis fac-
BDNF gene variants in schizophrenia MO Krebs et al
tors was done blind to the genotyping results. This study received the agreement of the local Ethical Committee. All subjects gave their written informed consent prior to the study. DNA analysis Venous blood samples were collected in ethylenediaminetetracetic acid-containing tubes. Genomic DNA was extracted from peripheral leukocytes of venous blood samples, collected in EDTA-containing tubes. After amplification by PCR with primers flanking exon 1 (5⬘-GCC-ACT-TTA-TCT-CCT-CCA-GT-3⬘; 5⬘ AGC-ACTAGC-TGC-CTA-TTC-CA-3⬘)14 and thermostable Taq Polymerase (AmpliTaq, Perkin Elmer Cetus, Rockville, MD, USA), products were resolved on 6% DNA sequencing gels and visualized by autoradiography. Genotyping was carried out blind to the clinical status and was at least duplicated. Statistical analysis Genotype distribution was compared to the predictable value from Hardy–Weinberg equilibrium. Chi-Square was used for categorical comparisons. Allele distribution was analyzed with the non-parametric Mann– Whitney test. Corrections for multiple testing (Bonferroni) were not applied because analyses were of exploratory nature. Since the 172-bp, 174-bp and 176-bp alleles varied in the same direction, as compared to 166-bp, 168-bp and 170-bp alleles, alleles were classified in two groups, referred to as ‘long’ and ‘short’ alleles, respectively. The hypothesis of association between clinical variables and long alleles (172– 176 bp) was tested with Pearson chi-square or, when appropriate, Yates corrected chi-square. The non-parametric Mann–Whitney U test was used to compare the distribution of allele sizes in sub-groups of patients and controls. Logistic regression was used to describe the loglinear relationship between the genotypes for BDNF and DRD3 genes and therapeutic response. First, the full model with interaction between BDNF and DRD3 genes was tested. Because the interaction was not significant, the term corresponding to the interaction was dropped. The Wald test was used to test the significance of the coefficient of each component included in the model.
Acknowledgements This work was promoted by INSERM. OG was supported by a Lundbeck grant and by the Fondation pour la Recherche Medicale. We wish to thank Marylis Corbex for her advice on statistical analysis.
References 1 Kovelman JA, Scheibel AB. A neurohistological correlate of schizophrenia. Biol Psychiatry 1984; 19: 1601–1621. 2 Weinberger DR, Lipska BK. Cortical maldevelopment, anti-psychotic drugs and schizophrenia: a search for common ground. Schizophrenia Res 1995; 16: 87–110. 3 Gualtieri CT, Adams A, Shen D, Loiselle D. Minor physical anomal-
ies in alcoholic and schizophrenic adults and hyperactive and autistic children. Am J Psychiatry 1982; 139: 640–643. Delisi LE, Hoff AL, Schwartz JE, Shields GW, Halthore SN, Gupta SM et al. Brain morphology in first episode schizophrenic-like psychotic patients: a quantitative magnetic resonance imaging study. Biol Psychiatry 1991; 29: 159–175. Hyman C, Hofer M, Barde Y, Juhasz M, Yancopoulos G, SaL Squinto R. BDNF is a neurotrophic factor for dopaminergic neurons of the substantia nigra. Nature 1991; 350: 230–232. Thoenen H. Neurotrophins and neuronal plasticity. Science 1995; 270: 593–598. Altar A, Boylan C, Jackson C, Hershenson S, Miller J, Wiegand S et al. Brain-derived neurotrophic factor augments rotational behavior and nigrostriatal dopamine turnover in vivo. Proc Natl Acad Sci USA 1992; 89: 11347–11351. Blochl A, Sirrenberg C. Neurotrophins stimulate the release of dopamine from rat mesencephalic neurons via the Trk and p75Lntr receptors. J Biol Chem 1996; 271: 21100–21107. Seroogy KB, Lundgren KH, Tran TMD, Guthrie KM, Isackson PJ, Gall C. Dopaminergic neurons in rat ventral midbrain express brain-derived neurotrophin factor and neurotrophin-3 mRNAs. J Comp Neurol 1994; 342: 321–334. Altar CA, Cai N, Bliven T, Juhasz M, Conner JM, Acheson AL et al. Anterograde transport of brain-derived neurotrophic factor and its role in the brain. Nature 1997; 389: 856–860. Carlsson A, Lindqvist M. Effect of chlorpromazine or haloperidol on the formation of 3-methoxytyramine and normetanephrine in mouse brain. Acta Pharmacol Toxicol 1963; 20: 140–144. Di Chiara G, Imperato A. Drugs abused by humans preferentially increase synaptic dopamine concentrations in the mesolimbic system of freely moving rats. Proc Natl Acad Sci USA 1988; 85: 5274–5278. Guillin O, Damier L, Griffon N, Diaz J, Carroll P, Schwartz J et al. Role of brain-derived neurotrophic factor in the control of D3 receptor expression. Eur Neuropsychopharmacol 1999; 9, suppl 5: S184. Pro¨schel M, Saunders A, Roses A, Mu¨ller C. Dinucleotide repeat polymorphism at the human gene for the brain-derived neurotrophic factor (BDNF). Hum Molec Gene 1992; 1: 353. Krebs MO, Sautel F, Bourdel MC, Sokoloff P, Schwartz JC, Olie JP et al. Dopamine D3 receptor gene variants and substance abuse in schizophrenia. Mol Psychiatry 1998; 3: 337–341. Hawi Z, Straub RE, O’Neill A, Kendler KS, Walsh D, Gill M. No linkage or linkage disequilibrium between brain-derived neurotrophic factor (BDNF) dinucleotide repeat polymorphism and schizophrenia in Irish families. Psychiatry Res 1998; 81: 111–116. Sasaki T, Dai XY, Kuwata S, Fukuda R, Kunugi H, Hattori M et al. Brain-derived neurotrophic factor gene and schizophrenia in Japanese subjects. Am J Med Genet (Neuropsychiatr Genet) 1997; 74: 443–444. Gorwood P, Leboyer M, Jay M, Payan C, Feingold J. Gender and age at onset in schizophrenia: impact of family history. Am J Psychiatry 1995; 152: 298–212. Numan S, Lane-Ladd S, Zhang L, Lundgren K, Russel D, Seroogy K et al. Differential regulation of neurotrophin and trk receptor mRNAs in catecholaminergic nuclei during chronic opiate treatment and withdrawal. J Neurosci 1998; 18: 10700–10708. Segman R, Neeman T, Heresco-Levy U, Finkel B, Karagichey L, Schlafman M et al. Genotypic association between the dopamine D3 receptor and tardive dyskinesia in chronic schizophrenia. Mol Psychiatry 1999; 4: 247–253. Steen VM, Lovelie R, MacEwan T, McCreadue RG. Dopamine D3receptor gene variant and susceptibility to tardive dyskinesia in schizophrenic patients. Mol Psychiatry 1997; 2: 139–145. Gurevich E, Bordelon Y, Shapiro R, Arnold S, Gur R, Joyce J. Mesolimbic dopamine D3 receptors and use of antipsychotics in patients with schizophrenia. Arch Gen Psychiatry 1997; 54: 225–232. Sokoloff P, Giros B, Martres M, Bouthenet M, Schwartz J. Molecular cloning and characterization of a novel dopamine receptor (D3) as a target for neuroleptics. Nature 1990; 347: 146–151. Karayiorgou M, Gogos J. Dissecting the genetic complexity of schizophrenia. Mol Psychiatry 1997; 2: 211–223. Endicott J, Spitzer RL. A diagnostic interview: a schedule for affective disorders and schizophrenia. Arch Gen Psychiatry 1978; 35: 837–844.
BDNF gene variants in schizophrenia MO Krebs et al
26 APA. Diagnostic and Statistical Manual of Mental Disorders. Revised, 3rd edn. APA: Washington DC, 1987. 27 Kay SR, Fiszbein A, Opler LA. The positive and negative symptom scale (PANSS). Schizophr Bull 1987; 13: 261–276. 28 Ha¨fner H, Maurer K, Lo¨ffler W, Bustamante S, van der Heiden W, Riecher-Ro¨ssler A et al. Onset and early course of schizophrenia. In: Ha¨fner H, Gattaz (eds). Search for the Causes of schizophrenia. Springer Verlag: Berlin-Heidelberg-New York, 1995, pp 43–66. 29 Beratis S, Gabriel J, Holdas S. Age at onset in subtypes of schizophrenic disorders. Schiz Bull 1994; 20: 287–296.
30 May PR, Dencker SJ, Hubbard JW, Midha KK, Liberman RP. A systematic approach to treatment resistance in schizophrenic disorders. In: Dencker S, Kulhanel F (eds). Treatment Resistance in Schizophrenia. Vieweg Verlag: Wiesbaden, 1988.
Correspondence: Dr MO Krebs, Service Hospitalo-Universitaire, Hoˆpital Sainte-Anne, 7, rue Cabanis, 75014 Paris, France. E-mail: krebs얀chsa.broca.inserm.fr Received 16 December 1999; revised and accepted 18 March 2000