Introduction
Alcohol misuse is responsible for 5.1% of the global burden of disease [
63] and can lead to alcohol use disorder (AUD), which is characterized by prolonged, compulsive and detrimental alcohol-drinking patterns, constant preoccupation with alcohol acquisition/drinking, tolerance and/or withdrawal symptoms [
4]. Twin studies show that heritability of alcohol addiction ranges between 40‒60% [
28]. Different developmental patterns and genes may be involved in the development of AUD during adolescence and adulthood [
18]. Adolescence is a critical period in the development of AUD, as first use of alcohol commonly occurs during this period. After this early experimentation phase, individuals display more stable patterns of drinking [
18].
Alcohol-related phenotypes (i.e. response to acute or chronic alcohol, withdrawal symptoms, loss of control over alcohol drinking/seeking, relapse) have consistently been associated with a dysfunctional glutamatergic system [
32,
34]. The glutamatergic system mediates the reinforcing effects of alcohol through various mechanisms, one of which is the interaction with the dopaminergic system in the mesolimbic circuit [
32]. Vesicular Glutamate Transporters (VGLUTs), 1‒3, package glutamate in the presynaptic vesicles [
3] (referring to
Vglut/VGLUT for mRNA/gene and protein, respectively, in rodents and
VGLUT/VGLUT in humans). Thus, any
Vglut/
VGLUT-expressing neuron has the ability to package and release glutamate, rendering
VGLUT genes optimal markers for the glutamatergic phenotype.
One of these three markers, VGLUT2, is broadly expressed in brain areas of relevance to addiction, e.g. the cerebral cortex, hippocampus, thalamus, amygdala and medulla [
26,
64]. Within the midbrain, VGLUT2 is expressed in both glutamatergic and dopaminergic neurons of the ventral tegmental area (VTA), a key area mediating reward [
47,
66]. Preclinical studies provide an association of VGLUT2 mRNA and protein expression with alcohol exposure pre- and post-natally [
68,
69]. In rats having free access to alcohol, we recently observed lower
Vglut2 expression in the medial prefrontal cortex, a key region involved in executive functions such as decision making and processing environmental cues [
65]. Further, studies of rodents have demonstrated the involvement of a VGLUT2 co-phenotype in behaviours of relevance to addiction. For example, mice lacking
Vglut2 in midbrain dopamine neurons show reduced locomotor response to acute injections of amphetamine [
10] and cocaine [
30], and higher cocaine self-administration and cue-induced drug seeking [
2]. VGLUT2 has been shown to contribute to increased dopamine vesicular content by regulating vesicle acidification upon depolarization, further highlighting its relevance to dopaminergic neurotransmission [
1]. Recently, it was demonstrated that VGLUT2 in dopamine neurons contributes to baseline AMPA/NMDA ratio in target neurons of the nucleus accumbens, a finding which suggests a role in synaptic plasticity of relevance to aspects of addiction [
49]. Moreover, we have previously found that the expression profile of
Vglut2 in the VTA differed in rats who voluntary drank alcohol if they had been exposed to early-life stress [
65]. This was the first evidence that the association of
Vglut2 and alcohol consumption was modified by stress.
In humans, an association of
VGLUT2 genotype with alcohol dependence was found by an exploratory, haplotype-tag Single Nucleotide Polymorphisms (SNPs) study of the three
VGLUT genes, such that the minor allele of the SNP rs2290045 in
VGLUT2 was overrepresented (OR 1.660) in a sample of 191 women with alcohol dependence as compared to 184 healthy women [
17]. This association could reflect a gene-by-environment interaction (GxE), since individuals presenting alcohol dependence typically have experienced more negative, and fewer positive, environmental factors, than their healthy peers [
20]. Importantly, not only negative, but also positive environmental experiences may interact with
VGLUT2 to influence individual susceptibility to AUD [
67].
Both supportive and aversive psychosocial factors have indeed been associated with risk for AUD. Adoption studies have shown that a positive rearing environment leads to lower risk of almost 50% in drug misuse among individuals with high genetic risk for addiction [
41]. Furthermore, parental monitoring has been associated with later onset of alcohol misuse, lower rate of alcohol drinking escalation across time, and less-frequent intoxication among adolescents [
7,
44]. On the other hand, many studies have shown that negative environmental factors are associated with AUD. For example, maltreatment in childhood, including neglect [
24,
27,
56], and witnessing physical/verbal abuse between parents, [
15,
60] have been associated with higher alcohol consumption in adolescents.
The “diathesis-stress” hypothesis [
70] proposes that carriers of risk alleles show increased vulnerability to negative environmental factors. The “vantage sensitivity” framework suggests that responses to positive environmental factors depend on inherent characteristics [
51]. The more recent ‘differential susceptibility’ theory [
8], integrates both approaches, and postulates that depending on the genotype some individuals are more, and some less, susceptible to both negative and positive environmental factors. Similarly, the ‘biological sensitivity to context’ theory [
11] postulates that GxE shape individuals’ environmental sensitivity over time, with some individuals having high biological reactivity to both highly stressful and highly protective environments [
11]. Thus, both the differential susceptibility theory and the biological sensitivity to context theory propose that individuals differ in their sensitivity to negative and positive environmental factors [
23]. GxE studies that include both stressful and enriching environmental factors are, therefore, needed to test these theories [
23,
50].
Very few studies have tested three-way gene-by-environment interactions (GxExE) including both negative and positive environmental factors. One of the first such studies showed that a positive environmental factor (i.e. social support) moderated a genetic effect on depression among maltreated children [
37]. Most other studies have examined associations of genes with various psychopathologies according to the differential susceptibility approach [
8], including only one environmental variable at a time, either negative or positive, while recent results of meta-analyses provided evidence that genotypes increase sensitivity to both negative and positive environmental factors [
6,
62].
To date,
VGLUT2 genotypes have been investigated in relation to neuropsychiatric outcomes, such as schizophrenia [
55] and Parkinson disease [
45]. To our knowledge, the exploratory haplotype-SNPs study, previously conducted by our group, is the only study investigating
VGLUT2 genotype in relation to AUD [
17], however, the interaction of this genotype with environmental factors remains to be studied. Hence, to further understanding of the role of
VGLUT2, and specifically of the rs2290045 genotype, in alcohol misuse, the present study sought to determine whether alcohol-related problems (i.e. hazardous alcohol use, dependence symptoms and harmful alcohol use [
5]) were associated with interactions of
VGLUT2 SNP rs2290045 and positive and negative environmental factors. Considering the strong associations between smoking and alcohol misuse [
28], and that higher
VGLUT2 gene expression has been found post-mortem in the VTA of alcoholic smokers compared to controls, and to alcoholic non-smokers [
25], the potential confounding effects of nicotine use were estimated. The study focused on adolescence/young adulthood, a transitional period characterized by dramatic physical and emotional changes, novelty-seeking and risk-taking behaviors, in an attempt to identify susceptible individuals early in time.
One clinical sample and two general population samples of adolescents and young adults were studied. Guided by the environmental sensitivity framework [
50], we hypothesized that individuals carrying the T allele who were exposed to stressful life events (SLE) would present more alcohol-related problems if they received non-optimal parenting, and fewer alcohol-related problems if they experienced warm, positive parenting. Among T carriers, those not exposed to SLE were expected to display fewer alcohol-related problems than those who experienced SLE, and fewer alcohol-related problems were expected with increasing quality of parenting. By contrast, it was hypothesized that alcohol-related problems would not be associated with the interaction between positive or negative environmental factors among individuals carrying CC genotypes.
Results
Descriptive characteristics
Characteristics of the three samples are presented in Table
1, and by sex in Table S3. Among GP-Adolescents, but not in the CS, alcohol consumption differed over time (
Z = − 25.304,
p < 0.0001). Only 12% (
N = 201) of GP-Adolescents were consuming alcohol when they were, on average, 14 years old. Differences in AUDIT/AUDIT-C scores depending on sex, genotype, SLE and parenting are presented in Table S3. Aggregation of the three samples showed no main effect of genotype on AUDIT-C scores. Weak correlations were observed between AUDIT/AUDIT-C scores and environmental variables (Table S3), whereas no gene-environment correlation was found.
Table 1
Descriptive characteristics in the clinical sample (CS), general population (GP)-adults and -adolescents
Age (years) | 16.5 ± 1.85 12–20 | 22.2 ± 1.8 19–26 | Age (years) | 22.15 ± 1.4 20–24 | Age (years) | 14.4 ± 1.04 13–16 | 17.3 ± 1.04 16–19 |
Sex (females) | 76 (58) | 72 (58) | Sex (females) | 927 (52.8) | Sex (females) | 949 (56.3) | 846 (58.9) |
rs2290045 MAF | 16.4 | 16 | rs2290045 MAF | 14.9 | rs2290045 MAF | 17.1 | 17 |
AUDIT | 10.76 ± 8.16 0–40 (94) | 10.02 ± 6.94 0–35 (93) | AUDIT | 6.77 ± 4.8 0–30 (91) | AUDIT-C | 0.42 ± 1.39 0–11 (12) | 3.3 ± 3.3 0–14 (63) |
SLE (types) | | | SLE (types) | | SLE | | |
None | 26.7% | 29.6% | None | 30.6% | 0.78 ± 1.65 | 1.31 ± 1.94 |
One | 35.1% | 35.2% | One | 29.4% | 0–13 (30) | 0–12 (48) |
Two | 27.5% | 24% | Two | 21.4% | | |
Three | 10.7% | 11.2% | Three | 11.6% | | |
| | | Four | 5.2% | | |
| | | Five | 1.8% | | |
Child-parent openness | 11.32 ± 5.72 0–24 | | Parent–child relationship | 4.32 ± 1.59 0–6 | PASCQ positive | | 28.3 ± 5.3 3–36 |
Parent–child affect | 11.78 ± 5.9 0–24 | | | | | | |
Parent–child support | 10.25 ± 3.72 2.7–16 | | | | | | |
Associations of alcohol-related problems and interactions of rs2290045, maltreatment, and parenting
In the CS, AUDIT scores at follow-up were associated with an interaction of rs2290045, SLE, and child–parent openness (AUDIT: GLM: (F(1,107) = 7.018, η
p
2
= 0.062, p = 0.009; adj. R2 = 0.205; NB: Wald χ2 = 17.246, p = 0.00003). T carriers who had experienced higher levels of SLE reported higher AUDIT scores than CC carriers if they had also experienced poor child–parent openness, and lower AUDIT scores if they had enjoyed a supportive, open, relationship with parents. The opposite pattern was seen in the absence of SLE. ROS analysis showed that the interaction was significant when parenting was higher than 13.6 (range; mean ± SD: 0–24; 11.32 ± 5.72).
Among adult females in the general population sample, AUDIT scores were associated with an interaction between rs2290045, SLE, and quality of the parent–child relationship (GLM: (F(1,919) = 9.404, η
p
2
= 0.01, p = 0.002; adj. R2 = 0.030; NB: Wald χ2 = 9.121, p = 0.003). T carriers who had experienced higher levels of SLE reported higher AUDIT scores than the CC group if they had also experienced poor parenting, but lower AUDIT scores if they had enjoyed supportive parenting. The opposite pattern was seen in the presence of lower levels of SLE. ROS analysis showed that the interaction was significant when parenting was lower than 1 and higher than 4.4 (range; mean ± SD: 0–6; 4.32 ± 1.59).
Among the adolescents in the general population sample, follow-up AUDIT-C scores were borderline significantly associated with a four-way interaction between rs2290045, SLE (follow-up), parenting style and sex (GLM: F(1,1415) = 3.063, η
p
2
= 0.002, p = 0.08, adj. R2 = 0.030; NB: Wald χ2 = 3.152, p = 0.076). The model was re-run separately among males and females. Among males, AUDIT-C scores were associated with a three-way interaction of genotype, SLE, and parenting style (GLM: F(1,581) = 3.754, η
p
2
= 0.006, p = 0.053, adj. R2 = 0.009; NB: Wald χ2 = 4.485, p = 0.034). T carriers who had been exposed to higher levels of SLE (follow-up), reported higher AUDIT-C scores, than the CC group, if they experienced poor parenting, but lower AUDIT-C scores if they experienced positive parenting. The opposite pattern was seen in T carriers exposed to lower levels of SLE. ROS analysis showed that the interaction was significant when parenting was between 3 and 17.5 (range; mean ± SD: 3–36; 28.3 ± 5.3).
As illustrated in Fig.
1, consistent with the environmental sensitivity framework [
50], in all three samples the association of
VGLUT2 SNP rs2290045 with AUDIT/AUDIT-C scores was modified by negative and positive environmental factors. T carriers who had experienced SLE reported more alcohol-related problems if they had a poor relationship with parents, and fewer alcohol-related problems if they enjoyed a positive relationship with parents. T carriers who had not experienced SLE, reported few alcohol-related problems if they had a poor relationship with parents, and more alcohol-related problems if they had a positive relationship with parents. By contrast, among CC carriers, levels of alcohol-related problems did not differ as a function of SLE or quality of the parent–child relationship. The results were virtually similar when considering TT carriers as a separate group (data not shown).
In a separate analysis, nicotine use was considered as potential confounding factor (Table S2) and was included in the model following the abovementioned approach by
Keller [
39]. In GP-Adults, the association of AUDIT scores with the interaction of rs2290045, SLE and parent–child relationship was strengthened when nicotine use was entered in model (GLM:
F(1,912) = 12.144,
η
p
2
= 0.013,
p = 0.001; adj.
R2 = 0.119; NB: Wald
χ2 = 11.783,
p = 0.001). ROS analysis showed that the interaction was significant when parenting was lower than 1.4 and higher than 4.6 (range; mean ± SD: 0–6; 4.32 ± 1.59). In GP-Adolescents, adding nicotine use to the model, weakened the association of AUDIT scores with the interaction of genotype, SLE, and parenting style (rs2290045 × SLE (follow-up) × PASCQ positive on AUDIT scores (follow-up): GLM:
F(1,574) = 2.823,
η
p
2
= 0.005,
p = 0.093, adj.
R2 = 0.267; NB: Wald
χ2 = 2.345,
p = 0.126), as well as in the CS, (rs2290045 × SLE (follow-up) × child-parent openness on AUDIT scores (follow-up): GLM:
F(1,94) = 6.907,
η
p
2
= 0.068,
p = 0.01; adj.
R2 = 0.174; NB: Wald
χ2 = 13.645,
p = 0.00002). ROS analysis showed that the interaction was significant when parenting was between 3 and 20.5 (range; mean ± SD: 3–36; 28.3 ± 5.3) for GP-Adolescents, and when it was lower than 3 and higher than 15.1 (range; mean ± SD: 0–24; 11.32 ± 5.72) for the CS.