Neuroplasticity, Neurotransmission and Brain-Related Genes in Major Depression and Bipolar Disorder: Focus on Treatment Outcomes in an Asiatic Sample

The samples considered for primary analysis in this study were comprised of 242 patients diagnosed with MDD and 132 patients diagnosed with BD according to the Diagnostic and Statistical Manual of Mental Disorders IV criteria [28] consecutively admitted to the Department of Psychiatry of Seoul St. Mary’s Hospital for pharmacological treatment. A total of 326 controls were also collected at the same site. Controls were consecutively collected among hospital staff and non-psychiatric hospital patients, who did not satisfy criteria for current or past psychiatric disorder. Non-psychiatric subjects suffering from moderate-severe, unstable medical or neurological conditions potentially affecting their psychological status and psychiatric evaluations were systematically excluded from the control sample. Recruitment details and exclusion criteria for patients have been previously described [29, 30]. Briefly, all patients had to be eligible for pharmacological treatment. MDD patients had to satisfy criteria for a current episode of MDD requiring treatment with antidepressants, while BD patients had to satisfy criteria for a current manic/mixed episode requiring treatment with mood stabilizers (for assessing antimanic effects with mood stabilizers (e.g., [29, 31‐34]). Patients were all undergoing naturalistic treatments with venlafaxine or paroxetine (MDD), lithium, valproate, carbamazepine and lamotrigine (BD). The choice of treatment was taken by the clinician based on international guidelines. Due to intent prior to the present study, only MDD patients treated with venlafaxine and paroxetine were included in this sample. A severe or unstable medical and/or neurological conditions, treatment with a long-acting antipsychotic, current or recent (past 6 months) comorbidity for alcohol/substance use disorder represented exclusion criteria. All the subjects were Koreans, of Korean ancestry.

The local ethical committee approved the study procedures, and all the subjects were included after they had signed an informed consent. Socio-demographic and clinical features of the samples are reported in Table 1. Case and controls were not different in terms of sex distribution (p = 0.39) and only marginally in terms of age (p = 0.08).

Patients and controls were evaluated for psychiatric disorders by the Mini-International Neuropsychiatric Interview [28]. Demographic and clinical variables, including family history for psychiatric disease, age at first illness episode (onset) and history of suicide attempt, were systematically collected by clinical interviews and review of clinical charts. In BD patients, the presence of psychotic features was also systematically collected. Depressive symptoms severity was evaluated by the Hamilton Depression Rating Scale (HDRS) [35] in both MDD and BD patients. Manic symptoms (BD patients only) were evaluated by the Young Mania Rating scale (YMRS) [36].

Response to antidepressant treatment, remission and resistance to treatments (primary outcomes) were defined according to Schosser et al. [37]. Briefly, response to treatment is defined as a ≥ 50% improvement of HDRS scores from baseline to endpoint; remission as a HDRS score of ≤ 7 at the endpoint; and resistance as non-response to at least two adequate consecutive antidepressant trials (including the present one) [37]. The efficacy of mood stabilizers was evaluated in terms of response, defined as a ≥ 50% improvement of YMRS scores from baseline to endpoint, and remission defined as YMRS total score ≤ 12 [36].

Genetic single nucleotide polymorphisms (SNPs) were chosen as previously described [27], but in the context of the Asian population. Briefly, SNPs were selected based on (1) a reported prevalence of at least 5% for the variant allele among the Asian population according to http://​hapmap.​ncbi.​nlm.​nih.​gov; (2) tagging approach [complete linkage disequilibrium (R2 ≥ 0.08) with one or more neighbor SNPs]; (and 3) availability of a validated assay.

High-throughput genotyping using a pyrosequencer (Biotage, Sweden) was used to genotype the genomic desossiribonucleic acid (DNA) from blood. Polymerase chain reaction primers (Bioneer, Daejeon, Korea) and sequencing primers (Bioneer) used for the pyrosequencing assay were designed through Pyrosequencing Assay Design Software v.1 (Biotage), and 1 primer of each primer set was biotinylated. A total of 141 SNPs within 19 genes were initially genotyped for this study. After pruning according to genotyping rate and minor allele frequency (MAF) within the sample under investigation, 132 SNPs remained. Three SNPs were excluded since they were not in Hardy–Weinberg equilibrium (HWE) (PLA2G4A rs10737276, NCAPG2-ESYT2 rs12668837 and RORA rs1403737), leaving 129 SNPs (details in Table 2).

Of the 19 genes analyzed, 12 (PLA2G4A, CREB1, CHL1, GSK3B, SP4, VIPR2, PPP3CC, BDNF, HTR2A, RORA, ST8SIA2 and COMT) were previously analyzed in our Caucasian sample [27]. Two genes analyzed in Caucasians (MAPK1 and ZNF804A) were excluded from the analysis as previously independently analyzed in these samples [30, 38]. Other 7 new genes were also analyzed: HOMER1, NCAPG2, ESYT2, WDR60, SIGMAR1, S100B and TXNRD2.

The HWE of alleles and the linkage disequilibrium (LD) among SNPs were analyzed using the Haploview software (v.3.2) for Windows [39]. Main statistical analyses were performed using the IBM SPSS package for windows (http://​www.​ibm.​com/​analytics/​us/​en/​technology/​spss/​). Single SNPs associations were tested by the Chi square test, logistic regression models when controlling for potential confounding factors, one-way analysis of variance/covariance or corresponding non-parametric tests. The statistical package “haplo.score” (R-project, http://​cran.​r-project.​org/​) was employed to compute block of alleles in LD (haplotypes) and test their association with the outcome variables.

The primary outcomes were represented by treatment effects in terms of response to antidepressant or mood stabilizers, remission and resistance to antidepressants. In secondary analyses, we tested genetic effects on disease risk (cases vs. controls) and disease-associated features (age of onset, history of suicide attempt, psychotic BD). The effect of potentially confounding factors was systematically checked, including age, sex and type of drug administered. Baseline symptoms severity (baseline HDRS and YMS scores) was used as a covariate when analyzing remission (not when analyzing response because it includes baseline severity by definition).

To limit the risk of false positives due to multiple testing, a Bonferroni correction was applied [40]. The corrected alpha-value was calculated dividing 0.05 for the number of genes analyzed (0.05/19 = 0.0026). Power estimation was calculated through G*Power v.3.1.9.2 [41]. Considering a MAF of 0.30, we had a sufficient power of 0.80 to detect odds ratios (OR) of 1.89 for the BD subsample and of 1.77 for the MDD subsample.

All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1964, as revised in 2013. All the patients were informed in detail about the aims and the procedures of the study, and they signed an informed consent prior to inclusion into the study. The protocol and the informed consent were approved by the local ethical committee (approval number HC10TISI0031).

Single SNPs were not significantly associated with response, remission with or resistance to antidepressant treatment in MDD, and neither were they significantly associated with response to or remission with mood stabilizers treatment in BD patients. However, some trends were observed. In MDD, BDNF (rs1103010, p = 0.006; rs11030104, p = 0.028; rs6265, p = 0.017) and HOMER1 (rs10042665, p = 0.003; rs12514775. p = 0.011) showed trends of association with response; RORA (rs1020729, p = 0.025;, rs11071570, p = 0.003; rs12148149, p = 0.045; rs8041466, p = 0.040) and HTR2A (rs643627, p = 0.008) with remission; CHL1 (rs1516338, p = 0.008) with treatment resistance. Two BDNF variants (s1519480, p = 0.003; rs7124442, p = 0.004) also showed trends of association with response to mood stabilizers in BD, in both single SNP and haplotype analysis. Further details are reported in supplementary Table 1.

Significant associations are reported in Table 3.

Strengths are represented by: focus on genes that previously received substantial evidence of involvement in mood disorders (COMT, HTR2A, BDNF, CHL1, Homer1 in particular); analysis of multiple SNPs within each gene; reanalysis in Asians of a set of genes previously tested for antidepressant response in Caucasians [27]; simultaneous evaluation of genetic effects on multiple disease phenotypes and critical associated features; and relatively large samples of affected subjects and healthy controls.

On the other hand, testing a large number of genetic variants on multiple phenotypes exponentially increases the risk of false positive results, especially on medium-sized samples. In some cases, the low allelic variability might decrease the power of the statistical analysis. However, we used a multiple testing correction that could be considered as overly conservative given the previous results. Further, we analyzed common SNPs only, but not other rare mutations or other polymorphic variants such as copy number variations or rare variants. Therefore, a complete coverage of the genes is not guaranteed. All recruited BD patients were in a manic or mixed phase. This made it possible to specifically evaluate the response to antimanic agents. On the other hand, it prevented the evaluation of antidepressant response in BD patients. Control subjects were recruited among hospital staff and patients. Although we checked for the absence of lifetime psychiatric conditions, as well as the absence of severe/unstable medical or neurological conditions in control hospital patients, we cannot ensure full representativeness of our control sample with a control sample drawn from the normal population. Moreover, controls were not systematically evaluated for family history of psychiatric disease. Both patients and controls were consecutively collected and not matched for age and gender. However, cases and controls resulted to be similar in terms of age and gender. Patients underwent a naturalistic treatment and were not randomized to one or another treatment arm. All MDD patients were treated with paroxetine or venlafaxine for purposes prior to this work. Patients with BD were also treated with different mood stabilizers. There is the possibility of a potential confounding effects of different compounds; however, this has not been commonly observed in the previous studies (e.g., [76]). Furthermore, treatment outcomes were evaluated after the 6th and before the end of the 8th week of treatment in order to identify early predictors of response to treatments. However, this relatively short observation period may have led to an overestimation of the rates of non-response and treatment resistance. When focusing on component phenotypes (disease-associated features), we systematically collected only age at onset, suicidal behavior and psychotic BD. Psychosis in MDD was not systematically collected. The evaluation of other relevant phenotypic aspects could have highlighted a greater number of genetic effects. Finally, we could only detect small genetic effects with moderately sufficient size (OR > 2.00) only within secondary analyses in the BD sample (risk of illness, age of onset and psychotic BD) and history of attempted suicide in the MDD sample.