The Interaction of BDNF and NTRK2 Gene Increases the Susceptibility of Paranoid Schizophrenia

The association between BDNF gene functional Val66Met polymorphism rs6265 and the schizophrenia is far from being consistent. In addition to the heterogeneous in schizophrenia per se leading to the inconsistent results, the interaction among multi-genes is probably playing the main role in the pathogenesis of schizophrenia, but not a single gene. Neurotrophic tyrosine kinase receptor 2 (NTRK2) is the high-affinity receptor of BDNF, and was reported to be associated with mood disorders, though no literature reported the association with schizophrenia. Thus, in the present study, total 402 patients with paranoid schizophrenia (the most common subtype of schizophrenia) and matched 406 healthy controls were recruited to investigate the role of rs6265 in BDNF, three polymorphisms in NTRK2 gene (rs1387923, rs2769605 and rs1565445) and their interaction in the susceptibility to paranoid schizophrenia in a Chinese Han population. We did not observe significant differences in allele and genotype frequencies between patients and healthy controls for all four polymorphisms separately. The haplotype analysis also showed no association between haplotype of NTRK2 genes (rs1387923, rs2769605, and rs1565445) and paranoid schizophrenia. However, we found the association between the interaction of BDNF and NTRK2 with paranoid schizophrenia by using the MDR method followed by conventional statistical analysis. The best gene-gene interaction model was a three-locus model (BDNF rs6265, NTRK2 rs1387923 and NTRK2 rs2769605), in which one low-risk and three high-risk four-locus genotype combinations were identified. Our findings implied that single polymorphism of rs6265 rs1387923, rs2769605, and rs1565445 in BDNF and NTRK2 were not associated with the development of paranoid schizophrenia in a Han population, however, the interaction of BDNF and NTRK2 genes polymorphisms (BDNF-rs6265, NTRK2-rs1387923 and NTRK2-rs2769605) may be involved in the susceptibility to paranoid schizophrenia.


Introduction
Schizophrenia is a chronic, recurrent, disabling mental disease with a high cost to society and individuals across the world [1][2]. Although the underlying etiology of schizophrenia is still poorly understood, lines of evidence suggests that dysfunction in neurodevelopmental processes resulting from genetic and environmental factors, contributes to the development of schizophrenia [3][4]. Brain-derived neurotrophic factor (BDNF), a member of the neurotrophin superfamily, plays important roles in various neurodevelopmental processes of the central nervous system (CNS), including neuronal differentiation and survival, synaptic connections as well as plasticity [5][6]. Mounting evidence has demonstrated that BDNF were involved in the pathophysiology of schizophrenia. Recently, Zhang et al. showed that BDNF levels were significantly lower in drug-free patients with schizophrenia [7]. Lee et al. also demonstrated that BDNF levels decreased significantly in unmedicated schizophrenic patients and elevated after successful antipsychotic treatment which parallel symptom improvement of the patients [8]. Furthermore, increasing postmortem studies have shown that BDNF levels were significantly lower in prefrontal cortex of schizophrenia patients [9][10]. In addition, it is notable that BDNF most likely functions through its high-affinity receptor, neurotrophic tyrosine kinase receptor 2 (NTRK2) [11]; and NTRK2 has also been found decreased in postmortem of schizophrenic subjects [12]. More interestingly, some previous study have also demonstrated that BDNF and NTRK levels were both decreased in the brain tissue of schizophrenic patients [13][14]. All these evidence suggest that dysfunction of BDNF and TrkB may be involved in the pathophysiology underlying schizophrenia.
At the molecular level, position 196 in exon 5 of the BDNF gene contains a G to A transition leading to an amino acid substitution (valine to methionine) at codon 66 in the precursor BDNF peptide sequence (dbSNP: rs6265). This polymorphism can affect activity-dependent secretion of BDNF, hippocampal function and morphology [15][16]. Thus, more and more research was carried out to explore the role of BDNF functional polymorphism rs6265 in development of schizophrenia. Muglia et al. found the association between this functional SNP and schizophrenia in Italian subjects [17]. Neves-Pereira et al. also showed a positive association in Caucasian [18]. The same result was also found in Chinese populations [19]. However, some literatures reported the contrary results, and the association studies between the functional SNP Val66Met and schizophrenia is far from being consistent [20][21][22].
As to NTRK2 gene polymorphisms, previous studies have shown that NTRK2 gene polymorphisms rs2769605, rs1387923, and rs1565445 were associated with mood disorders or antidepressants response [23][24][25][26]. However, no literature reported the association between those three NTRK2 gene polymorphisms and the susceptibility to schizophrenia.
The most possible reason for inconsistent results may be that schizophrenia is characterized by heterogeneous clinical features; and different subtype of schizophrenia may lead to the genetic complexity of the disease [27]. Furthermore, the most important thing is that gene-gene interaction is critical to describe a phenotypic effect, when a specific individual genetic variant has a minor marginal effect in a complex psychiatric disease [28].
Thus, in present study, patients with paranoid schizophrenia (the most common subtype of schizophrenia) were recruited to improve the homogeneity of samples. The aim of the present study is to investigate the role of BDNF functional Val66Met polymorphism (rs6265), three polymorphisms in NTRK2 gene (rs1387923, rs2769605 and rs1565445) and their interaction in pathophysiology of paranoid schizophrenia. To our knowledge this is the first study to investigate the genetic risk factors for the development of paranoid schizophrenia concerning gene-gene interaction.

Subjects
All procedures including standard informed consent were reviewed and approved by Institutional Review Boards of Shanghai Mental Health Center. This study was conducted in accordance with the Helsinki Declaration as revised 1989. All participants or their guardians (if the paptients had a compromised capacity) read the informed consent and were explained carefully about each item. Written informed consent was obtained from each participant or their guardians (if the paptients had a compromised capacity) before any study-related procedures were performed. All potential participants who declined to participate or otherwise did not participate were eligible for treatment and were not disadvantaged in any other way by not participating in the study.
All the subjects in this study undergo the Mini International Neuropsychiatric Interview (MINI) and were interviewed by two experienced psychiatrists. MINI is a brief structured interview for Axis I diagnosis of major psychiatric disorders in Diagnostic and Age and gender-matched healthy subjects were recruited excluding individuals with any major Axis I disorders and family history of mental disorder. Eventually, total 402 patients with a mean age of 34.965.4 years (230 males and 172 females) and matched 406 healthy controls with a mean age of 35.667.0 years (249 males and 157 females) were involved in this study. There were no significant difference of age and gender between cases and healthy subjects (t = 1.46, df = 806, p = 0.14; x 2 = 1.42, df = 1, p = 0.23, respectively).

DNA Extraction and SNP Genotyping
Peripheral blood samples were collected from the paticipants in 2 ml EDTA vacuum tube. Genomic DNA was extracted from peripheral blood according to standard laboratory procedures (Blood genomic DNA extraction kit, TIANGEN, Beijing) and stored in at 280uC until genotyping. The genotypes of the BDNF (rs6265) and NTRK2 gene (rs1387923, rs2769605 and rs1565445) were identified as reported in our previous study [26]. For quality control, all genotypes were called blind to the case or control status in the genotyping process. Of the samples collected, 10% were repeated for the genotyping assay, and the results were more than 99% concordant.

Statistical Analyses
Differences in age and gender between cases and control subjects were compared by using t-test and chi-square (x2) test respectively. Hardy-Weinberg Equilibrium, genotype and allele frequencies of individual SNPs, pair-wise linkage disequilibrium of all pairs of SNPs and haplotype analysis were calculated by using the Haploview 4.1 software. The extent of linkage disequilibrium (LD) was measured by the standardized D'. Those haplotypes with a frequency under 3% were ignored. For multiple test correction, the Bonferroni correction was applied and the significance level a was set at 0.0125 (0.05/4). The power to detect significant association was estimated using the online software Quanto (Version 1.2.3, available at http://hydra.usc.edu/GxE/) [29]. The statistical power of our study was more than 90% under the assumption of a moderate effect size (Odds Ratio [OR] = 1.5), a log additive model and disease prevalence of 1%.
Gene-gene interactions were investigated by using multifactordimensionality reduction (MDR) method [30]. In brief, MDR combined higher and lower predisposing genotypes into two different groups (high or low-risk). Then the combination model is selected based on the lower misclassification error. Furthermore, 10-fold cross-validation was applied to assess the predictive ability of the each model by calculating the prediction error. Then, the best model with the maximization of cross-validation consistency was selected. P values of prediction accuracy were determined empirically by permuting the case and control labels 1,000 times. Hierarchical interaction graphs and interaction dendrogram of MDR were applied to present the SNPs interaction of the best model [31]. In addition, traditional statistical methods were performed to examine the results from MDR analyses. A p value of less than 0.05 was considered statistically significant.

Hardy-Weinberg Equilibrium (HWE) and Linkage Disequilibrium (LD)
The distributions of genotypes in cases and control subjects were consistent with Hardy-Weinberg equilibrium (P.0.05) respectively. Analysis of pair wised LD was conducted for three SNPs in NTRK2 gene. The D9 and r 2 of three SNPs in NTRK2 gene are shown in Table 1.

Association Analysis of Four Polymorphisms with Paranoid Schizophrenia
As shown in Table 2, no statistically significant differences were found in allele or genotype frequencies between cases and control subjects for four individual SNPs (rs1387923, rs1565445 and rs2769605 in NTRK2 gene, rs6265 in BDNF gene). Haplotypebased analysis showed no significant association between the haplotypes from rs1387923, rs1565445 and rs2769605 in NTRK2 and paranoid schizophrenia (detailed in Table 3).

MDR Analysis of Gene-gene Interaction
Cross-validation consistency and the prediction error obtained from MDR analysis for each number of loci were shown in Table 4. One three-locus model BDNF (rs6265)-NTRK2 (rs1387923, rs2769605) had a maximum testing accuracy of 57.01% and a maximum cross-validation consistency (10/10) that was significant at p,0.0001 level, after determined empirically by permutation testing. Figure 1 summarizes the three-locus genotype combinations associated with high risk or low risk for each multilocus-genotype combination. The varied patterns of high-risk and low-risk cells across each of the different multi-locus dimensions provide evidence of epitasis, or gene-gene interaction; that is, the influence that each genotype at a particular locus has on disease risk is dependent on the genotypes at each of the other two loci (30).
Traditional statistical methods were applied to this three-locus model to aid in interpretation, which identified three high-risk genotype combinations and one low-risk genotype combination from all possible genotype combinations. In this three-locus (rs1387923-rs2769605-rs6265) model, the OR for the low-risk genotype combination (CC)-(GA)-(GG) was 0.33 (95% CI: 0. 18

Hierarchical Interaction Graphs
After identifying a high-risk combination of SNPs using MDR, we used the theory of information gain to interpret the relationship between these four SNPs and drew the hierarchical interaction graphs. As shown in Figure 2A, we found a positive interaction effect of rs6265 in BDNF gene and rs1565445 in NTRK2 with interaction entropy of 0.23%; rs6265 in BDNF gene and rs2769605 in NTRK2 with interaction entropy of 0.22%; rs6265 in BDNF gene and rs1387923 in NTRK2 with interaction entropy of 0.13%; rs1565445 and rs1387923 in NTRK2 with interaction entropy of 0.13%; a negative interaction effect of rs1565445 and rs2769605 in NTRK2 with interaction entropy of 20.14%.

Interaction Dendrogram
Interaction dendrogram demonstrated that rs6265 in BDNF gene and its receptor's one polymorphism rs1565445 in NTRK2 gene located on the same branch ( Figure 2B). These two SNPs were estimated to have the strongest synergy interaction, as indicated visually by the red line. The rs1387923 in NTRK2 gene was on a different branch, demonstrating a synergy interaction with its two polymorphisms in NTRK2 gene (rs1565445 and rs2769605) and rs6265 in BDNF gene as indicated visually by the orange line.

Discussion
Recently, BDNF and its high-affinity receptor, NTRK2, widely expressed in the adult brain, were both reported decreased in human postmortem of schizophrenia suggesting they were involved in pathophysiology of schizophrenia [4,9,14]. However, genetic associations between BDNF and schizophrenia had a contradicting result. Single nucleotide polymorphism rs6265 is the most common and critical functional genetic polymorphisms of the BDNF gene. Meanwhile, BDNF polymorphism rs6265 can influence activity-dependent secretion of BDNF [15]. Some previous studies have showed significant association between this polymorphism of BDNF gene and schizophrenia, but some showed no association between them. The most essential reason for the inconsistent results is that the heterogeneity of schizophrenia may contribute to the genetic complexity of the disease, like varied ethnic of samples and different subtype of schizophrenia [27]. Currently, Sun et al. reported the significant differences in the genotype distribution and allelic frequencies of the BDNF polymorphism rs6265 between schizophrenic patients in a Chinese Han population (n = 456) and controls (n = 483) [19]. However, most research and meta-analysis reported the negative association between BDNF polymorphism rs6265 and schizophrenia in a Chinese population. Xu et al. showed that no significant differences were found in allele or genotype or haplotype frequencies of BDNF polymorphism rs6265 between Chinese schizophrenic patient and controls; and their meta-analysis demonstrated that the this polymorphism did not contribute to the susceptibility to schizophrenia [22]. Wang et al. [32] and Sun et al. [33] sequentially reported no association between BDNF variants rs6265 and schizophrenia in a Chinese population. Recently, Zhang et al. also found no association between BDNF polymorphism rs6265 and the susceptibility to schizophrenia [34]. Yi et al. also demonstrated that there was no significant differences of genotype or allele distribution between early onset schizophrenic patients (onset before age 18) (n = 353) and controls (n = 394) in a Chinese Han population [35].
Paranoid schizophrenia is the most common type of schizophrenia, and little research studied the association between specific schizophrenia subtype and polymorphism rs6265 in BDNF gene. Recently, Suchanek et al. found no association between BDNF polymorphism rs6265 and the development of paranoid schizophrenia [36]. Until now, no literature reported the relationship between BDNF polymorphism rs6265 and paranoid schizophrenia in a Chinese Han population.
NTRK2, a high-affinity receptor of BDNF play a critical role in BDNF/NTRK2 pathway; and it has also been found decreased in postmortem of schizophrenic subjects [12]. Previous studies have demonstrated that three polymorphisms in NTRK2 gene (rs2769605, rs1387923, and rs1565445) were associated with mood disorders [23][24][25][26]. However, up to date, no literature reported the correlation between NTRK2 gene (rs2769605, rs1387923, and rs1565445) and schizophrenia.
In the present study, we recruited 402 patients with paranoid schizophrenia and 406 control subjects to examine the putative association between paranoid schizophrenia and polymorphisms in BDNF (rs6265) and NTRK2 genes (rs1387923, rs2769605, and rs1565445) in a Chinese Han population. No statistically significant differences in allele and genotype frequencies were observed between cases and control participants for all these four polymorphisms separately. The haplotype analysis showed no association between haplotype of NTRK2 genes (rs1387923, rs2769605, and rs1565445) and paranoid schizophrenia. However, we found the association between the interaction of BDNF and NTRK2 with paranoid schizophrenia by using the MDR method followed by conventional statistical analysis. The best gene-gene interaction model identified was a three-locus model (BDNF rs6265, NTRK2 rs1387923 and NTRK2 rs2769605). In this model, one low-risk and three high-risk four-locus genotype combinations were identified. We speculate that from our findings, since schizophrenia is a complex psychiatric disease, the individual genetic variants may just display minor marginal effects on its pathogenesis, and are hard detected; or some of the components, such as BDNF and its receptor NTRK2, in development of schizophrenia may act synergistically in ways we don't understand.    To the best of our knowledge, this is the first study to explore the correlation between BDNF (rs6265) and three NTRK2 gene polymorphisms and paranoid schizophrenia in a Chinese Han population; and the first time reporting the association between the interaction of BDNF and NTRK2 polymorphisms and the development of paranoid schizophrenia. However, several concerns or limitations still need to be addressed. Firstly, it is noteworthy that the moderate sample size and lack of independent replication, the current results should be interpreted with caution and further studies of independent, multiple-center, large-scale samples should be conducted to validate our results. Secondly, other crucial genes in BDNF/NTRK2 signaling pathway were not examined in the present study. For example, Kawanishi et al. found that two genetic variants (2933T-.C and 2413G-.A) in the promoter region of the cyclic adenosine monophosphate response element binding (CREB) gene were found only in schizophrenics, not in controls [37]. Thus, the possible role of other genes in the BDNF/NTRK2 signaling pathway and their interactions on susceptibility to schizophrenia should be further examined.
In conclusion, our findings implied that single polymorphism of rs6265 rs1387923, rs2769605, and rs1565445 in BDNF and NTRK2 did not demonstrate the association with the development of paranoid schizophrenia in a Han population, however, our finding suggested statistically significant role of interaction of BDNF and NTRK2 genes polymorphisms (BDNF-rs6265, NTRK2-rs1387923 and NTRK2-rs2769605) in schizophrenia susceptibility.