The present study provides further evidence, suggesting a possible role for allelic variation of the PER2
gene in alcohol drinking in humans. In agreement with previous exploratory findings by Spanagel et al. 
, this study indicates that carriers of the G allele of a PER2
haplotype tagging single nucleotide polymorphism (SNP rs56013859) were less engaged in alcohol drinking than individuals homozygous for the A allele. As the G allele is the minor allele of this polymorphism, this finding can be interpreted as a protective effect of the G allele on the susceptibility to alcohol abuse. While Spanagel et al. observed an association of this SNP with high vs. low alcohol intake in a group of alcohol-dependent patients, this study confirmed the association in young adults from a high-risk community sample.
The current investigation extends the findings of Spanagel et al. 
by demonstrating that, among experienced alcohol users in a nonclinical sample, variation in the PER2
gene may moderate the impact of stress on hazardous or harmful drinking. Our results revealed that exposure to stressful life events during the past four years was associated with higher AUDIT scores in homozygotes for the A allele of SNP rs56013859, but was unrelated among individuals carrying the G allele of this SNP. The AUDIT test is used to detect the preliminary signs of hazardous drinking and mild dependence, being one of the most accurate alcohol screening tests available 
. Interestingly, this gene-stress interaction was restricted to signs of harmful use and did not refer to patterns of drinking, such as frequency or amount of alcohol use. This finding that the PER2
genotype was associated with patterns of drinking behavior in both experienced and inexperienced alcohol users and with signs of hazardous use following stress in experienced users may be interpreted as suggesting a role of this gene in both the initiation and the progression of use.
Adolescence and early adulthood are periods of life which can become increasingly stressful. Such increases in exposure to stress may account for the rise in prevalence rates of alcohol intake during these age periods, in general, and of stress-reactive drinking, in particular 
. Overall, our findings indicated that the level of stress during the past four years was associated with higher current drinking. This association was most marked regarding the AUDIT score (i.e. signs of harmful drinking), while the number of drinking days appeared to be unrelated to stress.
Interest in PER2
as a candidate gene for alcohol use resulted from animal studies 
. Earlier experiments in rodents have shown that genetic variation in PER2
is associated with a hyperglutamatergic state, which in turn leads to a higher level of alcohol use. The pharmaceutical drug acamprosate is supposed to act on this mechanism in order to reduce stress-induced craving. The present investigation provides no evidence for the exact mechanism by which the specific PER2
genotype studied protects against risk of alcohol problems. In particular, our findings leave open the question as to how exposure to stress influences pathophysiological pathways by which the gene-environment interaction observed in our study is mediated. SNP rs56013859 is localized in the third intron of the Per2 gene and is not in linkage disequilibrium (LD) with common variants analyzed in Hapmap. Hypotheses on the functional role of variation in PER2
suggest that SNP rs56013859 may
alter the binding motives for transcription factors Sp1, c-myb, and nuclear factor B, suggesting a possible regulatory function of this SNP in transcriptional activation of PER2
. However, one limitation of this study is that we did not find any allele-specific expression of Per2
in peripheral blood from IMAGEN participants or on publicly available eQTL browsers of cortical tissue. Since there is distinct temporal variation and pronounced expression differences of Per2
across the life span () indicating temporal and regional-specific regulatory mechanisms, these results do not rule out a role of SNP rs56013859 or perhaps rare or intermediate variants in linkage disequilibrium with this SNP in transcriptional regulation. Recently, Agapito et al. 
demonstrated a PER2
mutation in mice to be associated with the regulation of β-endorphin release to acute and chronic ethanol challenges, providing another clue to the mechanism of different effectiveness of alcohol drinking as a means to cope with life events.
Spatiotemporal mRNA expression patterns of PER2 in humans.
Our findings add to the growing knowledge implicating genotypic variation in the moderation of the individual’s response to stress exposure 
. In particular, they parallel evidence from our study demonstrating a moderating effect of the CRHR1
genotype on the relationship between stressful life events and drinking behavior in adolescents 
. In 15-year-olds, the number of negative life events during the past three years was found to be related to increasing rates of heavy drinking only among individuals carrying a specific genotype of a haplotype tagging SNP of this gene. While activation of brain circuits involved in stress regulation is considered as the biological basis of the latter gene-environment interaction, a different physiological mechanism has been suggested to underlie the PER2
x stress interaction. The PER
genes (PER1, PER2
) in general, together with other clock genes, are part of the central circadian rhythm organization system in the suprachiasmatic nucleus (SCN). PER1
has recently been shown to be associated with alcohol use, indicating that variation in PER1 gene
mediated stress-induced drinking in animals and humans 
. Molecular experiments in human Per1
gene demonstrated a genotype specific increase in the transcription of Per1
according to the concentration of glucocorticoids. This may point to a gene-stress interaction effect on the transcription of Per1
, possibly resulting in different circadian organization under stressful conditions, which in turn may lead to self-medication with alcohol in those individuals.
Several limitations of the present study have to be considered. First, caution must be exercised in the interpretation of life stress, as it is difficult to separate the effect of environmental factors from genetic liability. Studies using genetically sensitive designs have indicated that many supposed environmental effects actually in part reflect genetic factors 
. Thus, exposure to life events may be genetically mediated and the gene-environment interaction observed in this study might well be due to interactions between the PER2
gene and other anonymous genes that were not identified (gene-gene interaction). Also in our study, the genotype groups differed slightly in their load of negative life events (see ). However, as these differences are not statistically significant, and as the main effect of genotype is included in the regression model revealing the interaction effect, we do not assume the GxE effect to reflect merely a gene-environment-correlation. Second, another point of criticism may be the fact that the design of the present study is not completely longitudinal, with the interval of the life events between 15 and 19 years overlapping to a minor degree with the alcohol use assessed for the last 45 days. Even though we can assume that the majority of negative life events did not occur during this time period, but rather in the preceding four years, the direction of causality between life events, their interaction, and the genetic vulnerability on alcohol consumption at age 19 remains unclear.
Thirdly, the present findings have to be viewed in the light of a number of difficulties inherent in the detection of “true” gene-environment interactions. Major issues of criticism relate to the potential for multiple testing, low statistical power, and the lack of criteria for replication. Multiple testing has long been a serious problem in genetic research. The availability of datasets which afford large numbers of subdivisions (due to different ways of defining genotype and environmental characteristics) multiplies the potential of multiple testing by offering numerous additional possibilities for data mining 
. Another difficulty in genetic association research is that many studies lack sufficient statistical power. Since statistical tests for examining interaction effects are less powerful than tests of main effects, this problem applies particularly to studies of GxE. The power to detect an interaction depends on a number of conditions, including the sample size as well as the distribution of genotypes and environmental exposures in the sample. Given the likely small effects of any single GxE and the associated risk of false positive results, this argues for the critical need for replication. However, differences in the measurement instruments, in assessing genotype, phenotype and environmental variables, between studies complicate to find comparable studies 
. However, recent meta-analyses, e.g. by Karg et al. 
, have demonstrated that replication of GxE effects is actually possible, and that problems, such as low statistical power, could be overcome by using an at-risk group approach. In this line, the results reported here should be considered with caution regarding the sample size and the number of statistical tests performed. Hence, we do not exclude the possibility that the reported associations in the present study may reflect false-positive results (type 1 error), given that the discovery set is rather small and no replication or functional work is available. Also, it should be emphasized that the associations we observed are small in effect and would not hold up to stringent correction for multiple testing and thus should be regarded as preliminary.
In conclusion, in this study, we demonstrated an association of allelic variation in PER2
gene with alcohol use in a sample of young adults from the community, replicating the findings of Spanagel et al. 
in a patient sample. Carriers of the protective variant drank less frequently in total, and displayed fewer alcohol-related problems following exposure to stressful life events. Future research will have to disentangle the actual pathways by which stressful life events affect alcohol use in order to contribute to a better understanding of the underlying mechanisms. However, in light of the likely burden of multiple tests, the nature of the measures used and the nominal evidence of interaction, the GxE interaction reported here needs replication in independent samples, before firm conclusions can be drawn.