Genes coding for: interleukin 1, alpha (IL1A), interleukin 1, beta (IL1B), tumor necrosis factor (TNF), serotonin receptor 2A (HTR2A), the promoter of the serotonin transporter gene (5-HTTLPR) and tryptophan hydroxylase 2 (TPH2) are candidate genes whose deregulation is thought to be central to psychiatric disorders [1]. In particular, the relevance of the serotonin system has been widely investigated and supported in psychopathology [2,3], and related to the molecular disruptions associated with alcohol related disorders [4,5]. Consistently, the genetic variations located in the promoter of the serotonin transporter has been convincingly found to regulate mood disorders [6] and alcohol related disorders [7–9]. Conflicting results [10,11] have further stimulated research in this direction. HTR2A plays as a key serotonin receptor: it is responsible for the post-synaptic activation upon serotonin transmission, a role that may modulate the psychiatric-related neuronal deregulations [12]. Recent evidence supports its connection with alcohol addiction [13–15]. In particular, HTR2A A-1438G variation, which is in completely LD with the variation we investigated here (reviewed in [12]), was found to be associated with alcohol dependence [15] and tobacco smoking combined or not with alcohol dependence [13]. Furthermore, HTR2A rs6313, which regulates the expression rate of HTR2A (reviewed in [12]), was found to be associated with alcohol abuse in male patients [14]. TPH2 is the brain specific rate limiting enzyme in the synthesis of serotonin [16], a role that candidates it as a modulator of the serotonin system orchestration. Recent reports discouraged the hypothesis of its participation to alcohol related behaviors [17–19]. IL1A and IL1B are mediators of the acute phase response, some genetic variations located in these genes have been associated with depressive symptomatology, antidepressant outcome and alcoholism [20,21]. TNF is a white cell produced pro-inflammatory cytokine likely involved in depressive disorder [16] which has been poorly investigated as a mediator of alcohol related psychiatric disorders. We yet investigated a set of variations located in the IL1A (rs1800587), IL1B (rs3087258), TNF (rs1799724), 5-HTTLPR (5-HTTLPR) and TPH2 (rs1386493; rs1386494; rs1487275; rs1843809; rs4570625; rs2129575) as modulators of alcohol related disruptions, both enzymatic and psychometric in previous published papers [22,23]: despite the relevance of these receptors and enzymes in the modulation of alcohol related phenotypes, we reported a lack of association. The reasons of this may be attributable to: lack of power, incomplete coverage of genetic mutations, incomplete analysis of epigenetic events, incomplete analysis of socio-demographic events and so on. In particular, epistasis between gene may have been played a fundamental role which remained undiscovered during the analysis of single mutations in previous papers we published on this sample. The aim of this paper is to challenge this particular limit of the previous publications on this sample [6,22] through the analysis of the epistasis between the set of variations yet singularly analyzed, also including a not yet investigated mutation (HTR2A rs6313) in the model. Further on, given that HTR2A rs6313 was not investigated as a single mutation in this sample of patients, we investigated the HTR2A rs6313 alone impact as a regulator of the biological and psychometric alcohol related variables.

Sixty four alcohol dependent patients and 47 relatives (non-consanguineous) were included in the study. Patients’ and their relatives’ characteristics are reported in Table 1. Table 2 reports the psychological assessment at admission and discharge. Biochemical reports are detailed in [22]. As a result of the principal analysis, that is, the interaction between epistasis (independent variable) and diagnosis and psychometric tests’ scores (dependent variable), we found a marginal impact of HTR2A rs6313 + TPH2 rs2129575 on Leyton -Trait at admission (p = 0.01; test accuracy = 0.87; test sensitivity = 0.95; test specificity = 0.56) and a similar marginal impact of ILA rs1800587 + 5-HTTLPR on Pilowski test at admission (p = 0.01; test accuracy = 0.80; test sensitivity = 0.73; test specificity = 0.85).

The results are shown in Figure 1: the frequencies of combined genotypes which have been found to be significantly associated with the scores of the dependent variables are in grey boxes; positive scores upon the right column in every box represent the frequencies of combined genotypes that have been associated with higher scores at the investigated test, negative scores upon each left column in each box represent the frequencies of the combined genotypes which have been associated with lower scores at the investigated test. Higher and lower scores are defined accordingly to the normal distribution obtained from the analysis of input data, “0” corresponding to the mean value of dependent variable, positive scores corresponding to scores upon the mean and negative scores corresponding to scores below the mean in the sample. Light boxes represent not significant associations. These results did not survive the Bonferroni correction. In the secondary analysis of the study, rs6313 did not impact the psychopathologic dimension and did not separate patients from their relatives (data not shown). In the pictures the grey shadowed boxes designate the combinations of genotypes associated with a statistically significant different distribution of the scores of the test under investigation. The light shadowed boxes designate the combinations of genotypes associated with a non statistically significant different distribution of the scores of the test under investigation.

To be implemented in the GMDR software the scores were normalized with “0” as the mean and threshold value. The GMDR software computes the frequencies of subjects who score higher (positive values on the left column on each box) or lower (negative values on the right column on each box) than “0” which is representative of the mean of the scores at the test under investigation. The scores are differentiated in positive and negative: the mean score given as a threshold, the positive scores stand for subjects who ranked higher than the mean, the negative stand for subjects who ranked lower than the mean. The difference in the frequencies of positive and negative scoring subjects is labeled as significant in the grey shadowed boxes. Light shadowed boxes stand for not significant difference in distribution of positive and negative scores. For example, in the first picture it is shown that IL1A rs1800587 and 5-HTTLPR combined variations are associated with a significant different score at Pilowski test at admission only in some specific combination of genotypes: the genotype CC at rs1800587 is associated with higher frequencies of subjects who ranked higher than the mean of the Pilowski test, only when it is found to be in combination with the LL or SS genotypes at the 5-HTTLPR. On the other hand, the combination with heterozygosis at the 5-HTTLPR with the same genotype at rs1800587 resulted in a not significant different distribution of subjects who ranked higher or lower than the mean of the Pilowski test. In the second picture, the combination of the CT genotype at rs6313 and the GG genotype at rs2129575 results into statistically significant higher scores at the Leyton – Trait test at admission, along with the other combinations of genotypes which are grey shadowed.

We found a marginal impact of the epistatic effect between the investigated variations and the obsessive-compulsive traits and hypochondriac symptoms. The low level of significance which did not stand the Bonferroni correction, is likely mainly related to the small sample size we investigated here: as formerly reported, the power was sufficient to detect a large effect size. Moreover, the genetic analysis we performed here did not cover the genes that are under investigation, which results in a possible genetic bias: some variations remain hidden in the present investigation that could impact the functionality of the investigated mutations. This exposes our study to a biological stratification bias which is quite common in literature [31]. Finally, the most part of the investigated variations belong to the TPH2 gene which has been quite consistently reported to be not associated with alcohol dependence. In all facts, even though TPH2 is the rate limiting enzyme for the synthesis of serotonin in central nervous system [32], a molecular role that candidates it as a central regulator of psychiatric disorders such as major depressive disorder [33–35] and bipolar disorder [35–37], recent reports rejected the hypothesis that variations located in the TPH2 gene could modulate alcohol dependence [17–19]. This is in contrast with the inhibitory role played by the serotonin system toward the mesolimbic dopaminergic system [38,39] which is thought to be primarily involved in the disruptions that lead to alcohol dependence [5]. This is the biological evidence that steered our attention on the possible epistasis impact of some relevant and well replicated variants (5-HTTLPR and HTR2A rs6313) over a putative good candidate as the TPH2 is. Some recent reports on the significant epistasis interaction between genetic variables that did not cast informative results when singularly investigated [40,41], convinced us to re-analyze our previously published results. Unfortunately, we could only find not conclusive data in this specific case. The question arises as whether the marginal significant associations we report are false positive findings, which are a common concern in genetic association studies [42], or the first partial evidence of a significant association.