BDNF variability in opioid addicts and response to methadone treatment: preliminary findings


  • R. De Cid,

    1. Genes and Disease Program, Center for Genomic Regulation (CRG) and CIBER Epidemiología y Salud Pública (CIBERESP), Barcelona, Spain
    2. National Center of Genotyping (CeGen), Center for Genomic Regulation, Barcelona, Spain
    Search for more papers by this author
    • 1

      These authors contributed equally to this work.

  • F. Fonseca,

    1. Drug Addiction Unit, IAPS, Barcelona, Spain
    Search for more papers by this author
    • 1

      These authors contributed equally to this work.

  • M. Gratacòs,

    1. Genes and Disease Program, Center for Genomic Regulation (CRG) and CIBER Epidemiología y Salud Pública (CIBERESP), Barcelona, Spain
    Search for more papers by this author
  • F. Gutierrez,

    1. Psychology Service, Neuroscience Institute, Hospital Clinic, Barcelona, Spain
    Search for more papers by this author
  • R. Martín-Santos,

    1. Pharmacology Unit, Institut Municipal d’Investigació Mèdica’ (IMIM), Barcelona, Spain
    Search for more papers by this author
  • X. Estivill,

    Corresponding author
    1. Genes and Disease Program, Center for Genomic Regulation (CRG) and CIBER Epidemiología y Salud Pública (CIBERESP), Barcelona, Spain
    2. National Center of Genotyping (CeGen), Center for Genomic Regulation, Barcelona, Spain
    3. Experimental and Health Sciences Department, Pompeu Fabra University, Barcelona, Spain
    Search for more papers by this author
  • M. Torrens

    Corresponding author
    1. Drug Addiction Unit, IAPS, Barcelona, Spain
    2. Pharmacology Unit, Institut Municipal d’Investigació Mèdica’ (IMIM), Barcelona, Spain
    3. Psychiatry and Legal Medecine Department, Universitat Autònoma de Barcelona, Barcelona, Spain
    Search for more papers by this author

*X. Estivill, PRBB Building, Plaça Charles Darwin s/n (Dr. Aiguader 88), 08003 Barcelona, Catalonia, Spain. E-mail: and M. Torrens, IAPs-Hospital del Mar, Passeig Marítim 25-29, 08003 Barcelona, Catalonia, Spain. E-mail:


Brain-derived neurotrophic factor (BDNF) signaling pathways have been shown to be essential for opioid-induced plasticity. We conducted an exploratory study to evaluate BDNF variability in opioid addict responders and nonresponders to methadone maintenance treatment (MMT). We analyzed 21 single nucleotide polymorphisms (SNPs) across the BDNF genomic region. Responders and nonresponders were classified by means of illicit opioid consumption detected in random urinalysis. Patients were assessed by a structured interview (Psychiatric Research Interview for Substance and Mental Disorders (PRISM)-DSM-IV) and personality was evaluated by the Cloninger’s Temperament and Character Inventory. No clinical, environmental and treatment characteristics were different between the groups, except for the Cooperativeness dimension (< 0.001). Haplotype block analysis showed a low-frequency (2.7%) haplotype (13 SNPs) in block 1, which was more frequent in the nonresponder group than in the responder group (4/42 vs. 1/135; Pcorrected = 0.023). Fine mapping in block 1 allows us to identify a haplotype subset formed by only six SNPs (rs7127507, rs1967554, rs11030118, rs988748, rs2030324 and rs11030119) associated with differential response to MMT (global P sim = 0.011). Carriers of the CCGCCG haplotype had an increased risk of poorer response, even after adjusting for Cooperativeness score (OR = 20.25 95% CI 1.46–280.50, = 0.025). These preliminary results might suggest the involvement of BDNF as a factor to be taken into account in the response to MMT independently of personality traits, environmental cues, methadone dosage and psychiatric comorbidity.

Opioid dependence disorder is a complex disease. The development of a drug addiction and the tendency to relapse are caused by a combination of both genetic and environmental factors (Kreek et al. 2005). Methadone maintenance treatment (MMT) is the most widely used treatment for opioid dependence and has been shown to be effective in opioid-dependent subjects who stay in treatment (Amato et al. 2005). If retention in treatment and/or illicit opioid use is considered as the main treatment outcomes, between 30% and 80% of treated patients respond poorly to MMT (GAO 1990). The provision of adequate doses of methadone and other psychosocial services are the main factors related to the success of MMT (Amato et al. 2005; Ward et al. 1999).

Personality characteristics of patients have also been related to MMT outcomes (Cacciola et al. 2001). From the dimensional approach, Cloninger’s model posits that personality encompasses partially inherited temperamental traits and acquired character traits (Cloninger et al. 1993). Temperamental dimensions may be correlated to specific brain systems and have been described as genetically independent from each other (Ebstein 2006; Gerra et al. 2005). In recent years, the study of the therapeutic response to MMT according to patients’ genetic backgrounds has become an issue of increasing interest (Barratt et al. 2006; Crettol et al. 2006; Lawford et al. 2000).

Neurotrophins in the brain enhance the growth and maintenance of several neuronal systems, modulate neurotransmission and play a role in plasticity mechanisms such as long-term potentiation (Chao 2003). As a member of the nerve growth factor-related family of neurotrophins, we focused on brain-derived neurotrophic factor (BDNF) (Anderson et al. 1995; Thoenen 1995). Human BDNF is located on chromosome 11p14.1 and encodes a 247 amino acid (aa) preprotein that is proteolytically cleaved to form the 120 aa mature protein, which is 100% conserved between mice, rats, pigs and humans (Maisonpierre et al. 1991). Linkage and association studies with markers in the BDNF genomic region have been associated with personality traits (Lang et al. 2004, 2005) and some diseases, including Parkinson’s disease, schizophrenia, bipolar disorder, obsessive-compulsive disorder and eating disorders. Most studies have focused on a single polymorphism in the pre-domain of BDNF (Val66Met) (Gratacos et al. 2007a). There is also strong evidence showing that the Met66 allele of this functional Val66Met polymorphism is associated with substance abuse (Beuten et al. 2005; Itoh et al. 2005; Matsushita et al. 2004) and specifically with opioid addiction (Cheng et al. 2005).

We present the preliminary results of an exploratory study aimed to explore the role of genetic variability of BDNF in response to MMT among a cohort of opioid-addicted subjects.

Patients and methods


The study recruited participants who met criteria for Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) opioid dependence from the MMT Program at a drug abuse outpatient center in Barcelona (CAS Barceloneta). The characteristics of the MMT provided were as follows: no upper limit on methadone dosages prescribed, no restriction on duration of treatment, forced discharge occurring only as a result of patients’ violent behavior or drug trafficking and clinical management that includes individual counseling and encouraged drug abstinence.

To be eligible for the study, patients had to be Caucasian, enrolled in MMT for at least 6 months and receiving a stable methadone dose for the last 2 months. Exclusion criteria were language-related barriers, severe cognitive impairment or any medical disorder that would interfere with the research assessments.

Clinical assessment

A close-ended questionnaire was used to record patients’ sociodemographic characteristics, serological status (human immunodeficiency virus, hepatitis C virus (HCV), history of substance use and previous psychiatric treatment. Substance and nonsubstance use psychiatric disorders were diagnosed according to DSM-IV criteria, using the Spanish version of the Psychiatric Research Interview for Substance and Mental Disorders IV (PRISM-IV) in axis I and II (borderline and antisocial personality disorders) (Hasin et al. 2006; Torrens et al. 2004). The degree of addiction-related impairment was assessed using the Spanish version of the Addiction Severity Index (ASI) (Gonzalez et al. 2002; McLellan et al. 1980). Personality characteristics were evaluated using the Spanish version of the self-administered Cloninger’s Temperament and Character Inventory (TCI) (Gutierrez et al. 2001). The temperament dimensions included Novelty Seeking, Harm Avoidance, Reward Dependence and Persistence, and the character dimensions included: Self-Directedness, Cooperativeness and Self-Transcendence. Urinalysis to detect heroin use was performed at the center randomly every 1 or 2 weeks, under the supervision of the nursing staff.


According to urinalyses, patients were classified as responders or nonresponders to MMT. Responders to MMT were those patients whose last four urinalysis tests were negative for illicit opioids, and nonresponders were those with two or more of their last four urinalysis tests positive for illicit opioids. Written informed consent was obtained from each subject after they had received a complete description of the study and been given the chance to discuss any questions or issues. The study was approved by the ethical and clinical research committee of the institution.

Single nucleotide polymorphisms selection and genotyping

Genetic variability in the BDNF genomic region (GenBank accession number NC_000011) was assessed by selecting single nucleotide polymorphisms (SNPs) in a 63.8 kb region including complete coding sequence, 3′ untranslated region (UTR) and 5′UTR regions. SNPs were selected from available databases at the time of experimental design (NCBI database, dbSNP 120; and from Celera databases ( A main criterion for inclusion markers was the availability of validation data and uniform distribution of SNPs along BDNF reference sequence. Based on the NCBI database (NCBI Build 36.1;, the common complete coding region and most of the 3′UTR region, common for preprotein isoforms a, b and c, are covered, as well the 5′UTR region for most of the described transcripts (Table 1).

Table 1.  Description of SNP markers genotyped in the BDNF genomic region
SNPSNP codeReference IDChromosome positionGene positionTranscript positionProtein positionAllelesMAFHWE*Missing (%)
  • ND, no Reference ID exists in NCBI; NA, no data because of lack of rare homozygote; MAF, Minor Allele Frequency.

  • Chromosome position is referred to NCBI build 36.1. Gene position is refereed to exon–intron structure boundaries reported in Pruunsild et al. (2007). Transcript position is referred to all different transcripts described for BDNF; NCBI build 36.1.

  • *

    HWE was calculated in all group, with a threshold P value of <0.001 accounting for multiple testing of 21 polymorphic SNPs.

1rs7124442rs712444227633617Exon IX3′UTR C/T27.60.1270
2rs11030099rs1103009927634159Exon IX3′UTR C/A21.41.0002
3rs2353512rs235351227636238Exon IXCodingAla150AlaG0 0
4rs6265rs626527636492Exon IXCodingMet66ValG/A19.11.0001
5rs3750934rs375093427636771Exon IX5′UTR A0 0
6hcv9278624rs1181980827637964Intron VIIIh /promoter IXIntronic C/T0.5ND0
7rs11030102rs1103010227638172Intron VIIIh /promoter IXIntronic C/G24.50.4170
8rs11030104rs1103010427641093Intron VIIIntronic A/G22.70.7731
9hcv26878171NA27645480Intron VIIIntronic T0 1
10rs7940188rs794018827650315Intron VIIIntronic G/C1.5ND0
11rs2049045rs204904527650817Intron VIIIntronic G/C19.11.0001
12hcv1751795rs1083521027652486Intron VIIIntronic A/C44.10.6784.1
13rs11030109rs1103010927653527Intron VIIIntronic G/A3.60.1751
14rs11030110rs1103011027655823Intron VIIIntronic G0 21.4
15rs7103411rs710341127656701Intron VIIIntronic T/C23.20.7733.1
16rs7103873rs710387327656893Intron VIIIntronic C/G480.6850
17rs10835211rs1083521127657941Intron VIIIntronic A/G240.5780
18hcv1751799rs1182608727658097Intron VIIIntronic G0 0
19hcv1751800NA27667224Intron VIIIntronic A0 0
20hcv1751802NA27667789Intron VIIIntronic T/A22.90.7732
21rs7127507rs712750727671460Intron VIIIntronic C/T28.60.0800
22rs1967554rs196755427676135Intron VIIIntronic C0 0
23rs11030118rs1103011827679639Exon IV5′UTR G0 0
24rs988748rs98874827681321Intron IIIIntronic C/G23.50.5780
25rs2030324rs203032427683491Intron IIIIntronic C/T490.4180
26rs11030119rs1103011927684678Intron IIIIntronic A/G250.0320
27rs7937405rs793740527685895Intron IIIIntronic A0 0
28rs7934165rs793416527688559Intron IIIIntronic A/G490.5402
29rs10767665rs1076766527690434Intron IIIIntronic A/G49.50.4142
30rs962369rs96236927690996Intron IIIIntronic C/T25.30.0311

Blood samples were collected from all subjects and genomic DNA was obtained from peripheral blood using standard procedures. SNPs were genotyped using the SNPlex™ platform (Applied Biosystems, Foster City, CA, USA) according to the manufacturer’s instructions and analyzed on an Applied Biosystems 3730xl DNA Analyzer. Allele calling was performed by clustering analysis using GeneMapper v.4.0 software. The genotype call rate was >95%. Genotyping quality was controlled in two ways. First, internal positive and negative controls provided by the manufacturer ABI (Applied Biosystems) were included in the reaction plates. Second, six duplicated samples of two HapMap reference trios were incorporated in the genotyping process. Both genotype concordance and correct Mendelian inheritance were verified.

Statistical analysis

Differences in sociodemographic and clinical characteristics between groups were examined using chi-squared and one-way analysis of variance tests, with Bonferroni post hoc analysis, using the SPSS software analysis package version 10.0. TCI scales values were converted into T scores (Population Norms in Cloninger et al. 1994).

Genotyped SNPs with a call rate of fewer than 80% were not considered for the association analysis. We tested each polymorphism in the whole group to ensure the fit with Hardy Weinberg equilibrium (HWE). Because of multiple testing, we used a threshold of = 0.001.

Multivariate logistic regression was used to assess the genetic effect of the SNPs. To better determine the real effect of SNPs on MMT outcome, we first performed a logistic regression analysis including the following variables: sex, age, TCI scores, psychiatric comorbidity and methadone dosage; factors that have a priori a high probability to be related to MMT response. Using a stepwise procedure, we selected the statistically significant variables that were included in the final model as covariates. Intergroup comparisons of genotype frequency differences were performed by regression analysis for dominant, recessive, overdominant and log-additive models of inheritance. Unadjusted crude odds ratios (OR) and 95% confidence intervals (95% CI) were calculated. We then calculated the OR adjusted for the clinical variables that were selected in the stepwise logistic regression procedure. The best genetic model was selected using the Akaike’s information criteria. Analysis was carried out using the SNPassoc R library, from the Comprehensive R Archive Network.

As a high correlation exists between the assayed markers because of tight linkage disequilibrium in the region, the problem of multiple testing was solved using the Spectral Decomposition approach (Li & Ji 2005). Allelewise experiment-wide significance threshold to control type I error rate at 5% resulted in a P value of 0.002. An additional correction factor was introduced to account for the four inheritance models tested (dominant, recessive, overdominant and log-additive models of inheritance) in the genotype analysis.

Haplotype association analysis was performed using a two-step method; first, we used a structured approach to test for association based on block structure and then we performed fine mapping of specific haplotype subsets inside the block using a sliding window approach. All SNPs with a minor allele frequency (MAF) >0.001 and inferred haplotypes with frequencies higher than 1% were included to define the haploblock structure. The block definition was based on the Linkage Disequilibrium (LD) measure D′ confidence interval block partition algorithm (Gabriel et al. 2002). Haplotype association analysis, as implemented in haploview v.4.0 software (Barrett et al. 2005), was based on block structure definition and the P value was based on simulation procedures derived after 10 000 simulation steps. A P value of <0.05 was considered statistically significant.

The sliding window approach was undertaken to better define specific haplotype subsets inside the block. In order to provide a comprehensive assessment of the haplotype subsets within the gene, the size of the window varied from 3, 6 and 10 SNPs. Only haplotypes with frequencies higher than 1% were considered. Global significance of the associated haplotype was estimated using a permutation test implemented in the haplo.stats v.1.3.1 package (Lake et al. 2003). A P value based on simulation procedures was derived (P global simulation) and a P value of <0.05 was considered to be statistically significant.

Crude and estimated OR and 95% CI for associated haplotypes were calculated using the function haplo.glm implemented in haplo.stats v.1.3.1 in the R programming language (Lake et al. 2003).


Sociodemographic and clinical characteristics of subjects

The clinical sample included 91 patients [66 (73%) male, age range 25–66 years, mean age 38 ± 8 years]. A total of 68 patients were classified as responders (49 males, mean age 38 ± 7 years) and 23 patients as nonresponders (17 males, mean age 37 ± 9 years). There were no differences between responder and nonresponder groups in terms of prescribed daily methadone dosage (106.27 ± 70.96 vs. 90.00 ± 49.86 mg/day, = 1.014, df = 88, = 0.313) and length of time in MMT (40 ± 43 vs. 29 ± 41 months, = 1.057, df = 88, = 0.293). The main sociodemographic, medical and psychopathological characteristics of patients are summarized in Table 2.

Table 2.  Mean characteristics of study groups
 Responders (= 68)Nonresponders (= 23)P*
  • SD, standard deviation.

  • *

    Values in bold represent significant P–values, after Bonferroni correction.

Male (%)49 (72)17 (74)1.000
Age in years, mean (SD)38 (7)37 (9)0.395
Years at school, mean (SD)9 (2)9 (4)0.522
Single (%)32 (48)10 (44)0.389
Legal background (%)38 (57)11 (52)0.804
Live with family (%)48 (72)16 (76)0.541
Offspring, mean (SD)0.9 (1.1)0.8 (1.0)0.774
Employed (%)19 (28)10 (48)0.187
Human immunodeficiency virus + (%)25 (37)5 (22)0.207
HCV + (%)49 (73)15 (65)0.595
Illicit opioid consumption in months, mean (SD)148 (86)89 (56)<0.001
Other substances dependence disorder – lifetime prevalence (%)
 Alcohol19 (29)6 (26)1.000
 Sedatives17 (26)7 (30)0.786
 Stimulants3 (5)1 (4)1.000
 Cannabis13 (20)3 (13)0.753
 Cocaine44 (67)12 (52)0.223
Days of heroin consumption in the last 30 days, mean (SD)0.4 (1.26)10.0 (11.8)<0.001
Days of cocaine consumption in the last 30 days, mean (SD)1.7 (5.0)8.5 (11.)0.012
Psychiatric comorbidity (lifetime prevalence) (%)50 (75)15 (65)0.424
Months in MMT, mean (SD)40 (43)29 (41)0.293
Methadone dosage, mean (SD)106.27 (70.96)90.00 (49.86)0.313
ASI scores, mean (SD)
 General health status3.1 (2.3)3.3 (2.3)0.719
 Working problems4.3 (2.8)3.3 (2.8)0.144
 Alcohol use1.3 (1.7)1.4 (1.4)0.789
 Substance use4.9 (2.5)6.6 (1.8)0.001
 Legal problems1.4 (2.0)2.0 (2.4)0.219
 Social relationships3.3 (2.5)3.0 (2.1)0.629
 Psychological status3.0 (2.5)2.3 (1.9)0.242
TCI temperament scales, mean (SD)
 Harm Avoidance59.1 (9.9)56.2 (7.9)0.203
 Novelty Seeking52.6 (7.9)55.0 (9.6)0.244
 Reward Dependence45.0 (9.6)47.1 (7.7)0.351
 Persistence42.5 (8.8)43.9 (9.8)0.513
TCI character scales, mean (SD)
 Self-Directness41.3 (12.0)45.3 (8.3)0.085
 Cooperativeness41.4 (8.6)47.4 (5.1)<0.001
 Self-Transcendence41.6 (10.3)40.7 (8.2)0.682

No differences in the proportions of responders and nonresponders in terms of lifetime and current prevalence of other substance use dependence disorders (cocaine, cannabis, alcohol, sedatives and stimulants) and lifetime psychiatric comorbidity in axes I and II (75% in responders vs. 65% in nonresponders, χ2 = 0.456, df = 1, = 0.424) were found. Regarding the severity of addiction, there was a tendency in the ‘Other Substance Use’ dimension of the ASI for the nonresponder group to score higher. However, when substance use in the last 30 days was compared for responders and nonresponders, the only difference found was that as expected, nonresponders reported more days of heroin consumption in last 30 days. The mean number of days of cocaine use during the last 30 days was not significantly different between both groups (Table 2).

There was a significant difference in TCI scores between groups only in the Cooperativeness dimension, with the nonresponder group scoring higher (Table 2).

Block structure of the BDNF region

A detailed description of assayed SNPs is summarized in Table 1. All selected SNPs were in HWE (> 0.01). Nine out of 30 SNPs were not polymorphic and were not informative to block structure definition. Twenty-one SNPs were used for infer block structure. Extended LD defined one block over the entire genomic region in the whole patient sample. Block structure was then independently derived separately in both patient samples: responders (= 68) and nonresponders (= 23). The block definition was slightly different between both groups (Fig. 1). Two blocks were defined in the nonresponder group (47 and 1 kb) and only one in the responder group (57 kb). In the nonresponder group, the major block was conserved in the coding and 3′UTR region but was broken upstream of the 5′UTR region of transcript NM_170733, which codes for BDNF isoform. Detailed composition of haploblocks for each group is summarized in Fig. 1.

Figure 1.

Haploblock structure obtained from genotyped SNPs was determined and visualized using the genetic analysis program haploview.( The transcripts structure is shown at the top. The lower part of the figure shows the LD block identified using haploview. Representation of the extension of the identified haplotypes are depicted as solid red bars under the transcripts map, and upper the block diagram. All genotyped SNPs are represented: 1–30. Linkage disequilibrium is calculated by D′, and each square represents a pairwise value of D′with the standard color coding represented in a GOLD heatmap diagram. Stronger LD is represented by red and weaker by blue as indicated in the color key. The boxes in black indicate block structure for responder and nonresponder groups, respectively. Depicted haploblocks were inferred including all SNPs and haplotypes with a frequency >1% using the D′confidence interval algorithm. The block structure was slightly different between groups. One block was defined in the responder group, but two blocks were present in the nonresponder group. *SNP 4 (rs6265) corresponds to Val66Met BDNF variant.

BDNF variability between responders and nonresponders to MMT

Multivariate logistic regression analysis was performed to identify confounding variables in a predictive model of response to MMT. We included TCI dimension scores, psychiatric comorbidity (in the last 12 months), methadone dosage, age and gender as independent variables in the model. After a stepwise procedure, only Cooperativeness scores remained significant in the model, as also found in the univariate analysis.

The genotype distribution in responders and nonresponders under different models of genetic action did not yield statistically significant differences in the crude analysis (P > 0.05) and only was nominal for rs7940188 after adjusted analysis for Cooperativeness, Puncorrected = 0.045.

Haplotype analysis revealed a low-frequency haplotype (2.7%) in block 1 associated with a poorer response to MMT (Table 3). AGCTGATCCCGAA haplotype (hcv1751792, rs2049045, hcv1751795, hcv1751796, rs7103873, rs10835211, hcv1751802, rs7127507, rs988748, rs2030324, rs11030119, rs7934165 and rs10767665) in block 1 was more frequent in the nonresponder group than in the responder group (4/42 vs. 1/135, Pcorrected = 0.023). In addition, we observed a haplotype more frequent in responders than in nonresponders (34/102 vs. 5/41, Puncorrected = 0.0435), but this result was no longer significant after the permutation validation procedure. The sliding window approach identified a haplotype subset associated with a differential response to MMT, in the 10 and 6 window analyses. The smaller multimarker with statistical significance was the haplotype formed by six SNPs (rs7127507, rs1967554, rs11030118, rs988748, rs2030324 and rs11030119). This haplotype was significantly associated with a differential response (global P sim = 0.011). Haplotype analysis showed that carriers of the CCGCCG haplotype (both homozygotes and heterozygotes) had an increased risk of poorer response (OR = 11.99, 95% CI 1.24–116.33, = 0.032), even after adjusting for Cooperativeness, although in this case, a wider CI was obtained (OR = 20.25, 95% CI 1.46–280.50), = 0.025).

Table 3.  Haplotype block analysis showed a low-frequency haplotype associated with nonresponse to MMT. Haplotype in block 1 was formed by SNPs: hcv1751792, rs2049045, hcv1751795, hcv1751796, rs7103873, rs10835211, hcv1751802, rs7127507, rs988748, rs2030324, rs11030119, rs7934165 and rs10767665
Haplotype*FrequencyNonrespondersRespondersP valueP value_ 10 000
  • P value_10 000; corrected P value after 10 000 step permutation procedure as implemented in haploview.

  • *

    Infered haplotype in block 1 defined by confidence interval block partition algorithm. Bold letters represent informative haplotype tag SNPs (htSNPs).

  • Fractional likelihoods of each individual for each haplotype.



This exploratory analysis suggests that BDNF variability confers a differential susceptibility to MMT response in opioid-dependent patients. Although association of BDNF with opioid dependence disorder has been previously reported (Cheng et al. 2005), as far as we are aware, this is the first time that BDNF variability has been linked to opioid therapeutic response. The relevance of BDNF variability to therapeutic response has been previously suggested with regard to the efficacy of other psychiatric treatments such as prophylaxis with lithium carbonate for bipolar mood disorders (Rybakowski et al. 2005) and antidepressants for unipolar depression (Gratacos et al. 2007b).

Sociodemographics, medical status and psychiatric comorbidity were not associated with a differential response to MMT, as reported by others (Kellogg et al. 2006). Psychiatric disorders such as mood disorders, anxiety disorders, eating disorders and schizophrenia have been associated with both opioid dependence disorder (Brooner et al. 1997) and BDNF variability (Gratacos et al. 2007a). However, our results suggest that the effect of BDNF on MMT response is independent of the presence of psychiatric comorbidity. Similar to pharmacological treatment studies on eating disorders (Klump et al. 2004), lower Cooperativeness scores were observed in responder group. However, other studies on major depression and panic disorder showed opposite results (Hirano et al. 2002; Marchesi et al. 2006). Although a sample size bias cannot be discarded, these differences could be diagnosis specific or might be related to the TCI assessment time (Marchesi et al. 2006).

Genetic variability in BDNF plays a role in patients undergoing MMT, as demonstrated by our findings. From the adjusted analysis, we identified one haplotype with a low frequency that confers a poorer response in patients undergoing MMT. Nevertheless, given the low frequency of this haplotype, it seems clear that no single effect is present and that other factors should be considered. Because BDNF has been linked to both the dopaminergic system (Seroogy et al. 1994) and the noradrenergic system (Akbarian et al. 2002) that play a relevant role in opioid dependence disorder, it is not unlikely that BDNF could affect the response to MMT. Differential response to MMT could be a consequence of a reduction of brain plasticity arising from an altered expression and functionality of BDNF.

From our data, we cannot infer any biological effect on BDNF protein levels. To date, the most widely reported BNDF variant that has been functionally tested is Val66Met (rs6265), which has been shown to affect intracellular trafficking and activity-dependent secretion of BDNF protein (Egan et al. 2003). Strong evidence has been reported of the role of Val66Met in substance abuse (Cheng et al. 2005; Itoh et al. 2005; Matsushita et al. 2004), but we were not able to find evidence of the effect of this single polymorphism in terms of response to MMT in our analyses. Moreover, in accordance with observed haploblock structure differences between groups, informative SNPs that define the associated haplotype are located in the 5′ region of BDNF. Other undetermined functional or regulatory unknown variants could be carried for the associated haplotype. Promoter variants not included in this study have been associated with BDNF levels, in addition to Val66Met (Jiang et al. 2005). Furthermore, there may be an alteration of silencing mechanisms regulated by fine control of messenger RNA expression. Specific patterns of expression in different brain regions and peripheral tissues have been reported in rats (Liu et al. 2006; Pattabiraman et al. 2005) and mice (Dennis & Levitt 2005), and a complex genomic structure has been reported in the human region (Liu et al. 2005; Pruunsild et al. 2007). Liu et al. have characterized a new gene (BDNFos) in the antisense complementary chain (Liu et al. 2005, 2006). BDNFos is transcribed to produce alternatively spliced natural antisense transcripts, and its fifth exon overlaps with the coding exon of human BDNF, suggesting a role as regulatory RNA. None of the associated SNPs in our study is located in the coding exon, but the risk haplotype could carry some of these functional variants, not directly related to the expression or function of BDNF but perhaps to a regulatory effect by BDNFos.

While the small sample size of our study may limit the generalizability of our results, the careful phenotypic assessment is a point of strength of the study. Results of the present exploratory study suggest involvement of BDNF as a factor to be considered in the response to MMT independently of personality traits, environmental cues, methadone dosage and the presence of medical and psychiatric comorbidity. Moreover, taking into account preclinical and clinical studies (Graham et al. 2007; Williamson et al. 2007), the role of BDNF and cocaine use in MMT response has yet to be determined. Future studies should confirm our results in a larger sample and in addition should account for pharmacokinetic factors that influence the response to methadone treatment.


We would like to thank the patients for taking part in the study and the CAS Barceloneta nursing team and Laura Diaz for their valuable help with collecting the data. We would also like to thank Marta Ribasés for accurate BDNF SNP selection and Antonio Verdejo for helpful comments. We thank Marta Pulido, MD, for editing the manuscript and editorial assistance. Financial support was received from TV3 Marató (01/810), FIS G03/005, G03/184, PI06/0940 and CIBERESP (Spanish Ministry of Health).