Reward-related ventral striatum reactivity mediates gender-specific effects of a galanin remote enhancer haplotype on problem drinking

Authors


Corresponding author: Y. S. Nikolova, Laboratory of NeuroGenetics, Department of Psychology and Neuroscience, 417 Chapel Drive, Box 90086, Duke University, Durham, NC 27705. E-mail: yuliya.nikolova@duke.edu

Abstract

The neuropeptide galanin has been implicated in the regulation of appetitive and consummatory behaviors. Prior studies have shown that direct injection of galanin into the hypothalamus results in increased release of dopamine (DA) in the nucleus accumbens (NAcc), and parallel increases in food and alcohol consumption. These studies are consistent with a role of hypothalamic galanin in regulating reward system reactivity. In humans, a common functional haplotype (GAL5.1) within a remote enhancer region upstream of the galanin gene (GAL) affects promoter activity and galanin expression in hypothalamic neurons in vitro. Given the effects of hypothalamic galanin on NAcc DA release and the effects of the GAL5.1 haplotype on GAL expression, we examined the impact of this functional genetic variation on human reward-related ventral striatum (VS) reactivity. Using an imaging genetics strategy in Caucasian individuals (N = 138, 72 women) participating in the ongoing Duke Neurogenetics Study, we report a significant gender-by-genotype interaction (right hemisphere: F1,134 = 8.08, P = 0.005; left hemisphere: F1,134 = 5.39, P = 0.022), such that homozygosity for the GG haplotype, which predicts greater GAL expression, is associated with relatively increased VS reactivity in women (n = 50, right hemisphere: P = 0.015; left hemisphere: P = 0.060), but not in men (N = 49, P-values > 0.10). Furthermore, these differences in VS reactivity correlated positively with differences in alcohol use, such that VS reactivity mediated a gender-specific association between GAL5.1 haplotype and problem drinking. Our current results support those in animal models implicating galanin signaling in neural pathways associated with appetitive and consummatory behaviors of relevance for understanding risk for substance use and abuse.

Galanin is a neuropeptide involved in a wide variety of central and peripheral nervous system functions (Gvilia 2010; Rada et al. 1998; Xu et al. 2010). Within the brain, a major population of galanin-synthesizing neurons resides in the paraventricular nucleus (PVN) of the hypothalamus. Studies in rodents have demonstrated that PVN-originating galanin modulates activity in downstream neural regions subserving the regulation of reward-related appetitive and consummatory behaviors. Specifically, microinjection of galanin into the PVN has been shown to result in an increase in dopamine (DA) in the nucleus accumbens (NAcc), a region critically implicated in reward processing and consummatory behavior (Rada et al. 1998). Consistent with these neurochemical effects, galanin infusion into the PVN results in increased food intake (Rada et al. 1998) as well as increased alcohol consumption (Rada et al. 2004).

In humans, relatively increased DA signaling in the ventral striatum (VS), a brain region comprising the NAcc, has been linked to increased impulsivity (Forbes et al. 2009) and delay discounting (Hariri et al. 2006). These behavioral phenotypes have in turn independently been associated with risk for abuse of and/or dependence on alcohol and other substances (Kreek et al. 2005). Despite the demonstrated effects of galanin on DA signaling and alcohol intake in rodents, as well as the established involvement of DA in substance use-related behaviors across species, few studies have investigated the impact of galanin signaling on reward-related brain function and consummatory behavior in humans.

A recent study identified a remote enhancer region upstream of the human galanin (GAL) gene located on chromosome 11. This region, also known as GAL5.1, harbors two single-nucleotide polymorphisms (SNPs), rs2513280 (C/G) and rs2513281 (A/G), the GG haplotype of which has been associated with higher GAL promoter activity in hypothalamic neurons in vitro, compared to all other allele combinations (Davidson et al. 2011). Importantly, the two GAL5.1 SNPs are in complete linkage disequilibrium (R2 = 1.00, D′ = 1.00) in all HapMap populations except YRI (International HapMap Project, http://hapmap.ncbi.nlm.nih.gov/), thus the GG and CA are the only two haplotypes occurring among individuals of European ancestry. While the GG haplotype has been shown to strongly influence promoter activity in vitro, no study has investigated the effects of genetic variation within the GAL5.1 region on neural and behavioral phenotypes in vivo.

In this study, we set out to examine the effects of rs2513281 (A/G), a perfect proxy for the GAL5.1 region haplotype, on VS reactivity, as measured in vivo with BOLD fMRI in a sample of young adult volunteers. We then extended our analyses to examine how any differences in VS reactivity attributable to GAL5.1 haplotype predict self-reported problem drinking in the same population. We hypothesized that the higher activity GG haplotype would be associated with increased VS reactivity, which would in turn predict higher levels of problem drinking. Given reports suggesting galanin may have stronger effects on behavior in females than males (Karatayev et al. 2010; Rugarn et al. 1999; Unschuld et al. 2010), we hypothesized that this effect would be stronger in women.

Materials and methods

Participants

A total of 149 Caucasian participants (77 women; mean age 19.77 ± 1.27) who completed the Duke Neurogenetics Study (DNS) and for whom rs2513281 was genotyped comprised the current sample. The DNS assesses a wide range of behavioral and biological traits among non-patient, young adult, student volunteers. We focused on non-Hispanic Caucasian participants because the two GAL5.1 SNPs comprising the functional haplotype are in complete linkage disequilibrium in this group (International HapMap Project, http://hapmap.ncbi.nlm.nih.gov/). In addition, several prior studies investigating the effects of GAL polymorphisms on behavior have focused on Caucasian participants (Gold et al. 2012; Unschuld et al. 2010).

All DNS participants provided informed consent in accord with Duke University guidelines were in good general health, and free of the following study exclusions: (1) medical diagnoses of cancer, stroke, diabetes requiring insulin treatment, chronic kidney or liver disease or lifetime history of psychotic symptoms; (2) use of psychotropic, glucocorticoid or hypolipidemic medication and (3) conditions affecting cerebral blood flow and metabolism (e.g. hypertension). Diagnosis of any current DSM-IV Axis I disorder or select Axis II disorders (Antisocial Personality Disorder and Borderline Personality Disorder), assessed with the electronic Mini International Neuropsychiatric Interview (Sheehan et al. 1998) and Structured Clinical Interview for the DSM-IV subtests (First et al. 1996), respectively, were not an exclusion as the DNS seeks to establish broad variability in multiple behavioral phenotypes related to psychopathology.

Seven participants were excluded for excessive motion during fMRI data acquisition, three for insufficient or missing task response data due to task non-compliance (n = 2) or a scanner button box malfunction (n = 1), and one participant was excluded due to an incidental neurological finding, leaving a final sample of 138 individuals (72 women; mean age 19.81 ± 1.27). No participants met criteria for either Antisocial or Borderline Personality Disorder, and 26 participants from our final sample (N = 138) met criteria for at least one Axis I disorder (Table S1 for specific diagnoses) but all were medication free. We conducted our analyses with and without current Axis I diagnosis (coded as 0 = no diagnosis, 1 = diagnosis) as a covariate.

Ventral striatum reactivity paradigm

As described previously (Forbes et al. 2009), our blocked-design number-guessing paradigm consists of a pseudorandom presentation of three blocks of predominantly positive feedback (80% correct guess), three blocks of predominantly negative feedback (20% correct guess) and three control blocks. There are five trials in each task and control block. During each task trial, participants have 3 seconds to guess, via button press, whether the value of a visually presented card is lower or higher than 5 (index and middle finger, respectively). The numerical value of the card is then presented for 500 milliseconds and followed by appropriate feedback (green upward-facing arrow for positive feedback; red downward-facing arrow for negative feedback) for an additional 500 milliseconds. A crosshair is then presented for 3 seconds, for a total trial length of 7 seconds. During control blocks, participants are instructed to simply make button presses during the presentation of an ‘x’ (3 seconds), which is followed by an asterisk (500 milliseconds) and a yellow circle (500 milliseconds). Each block is preceded by an instruction of ‘Guess Number’ (positive or negative feedback blocks) or ‘Press Button’ (control blocks) for 2 seconds resulting in a total block length of 38 seconds and a total task length of 342 seconds.

BOLD fMRI data acquisition

Each participant was scanned using a research-dedicated GE MR750 3 T scanner equipped with high-power high-duty-cycle 50-mT/m gradients at 200 T/m/s slew rate, and an eight-channel head coil for parallel imaging at high bandwidth up to 1 MHz at the Duke-UNC Brain Imaging and Analysis Center. A semi-automated high-order shimming program was used to ensure global field homogeneity. A series of 34 interleaved axial functional slices aligned with the anterior commissure–posterior commissure plane were acquired for full-brain coverage using an inverse-spiral pulse sequence to reduce susceptibility artifacts (TR/TE/flip angle = 2000 ms/30 ms/60; FOV = 240 mm; 3.75 × 3.75 × 4 mm voxels; interslice skip = 0). Four initial radiofrequency excitations were performed (and discarded) to achieve steady-state equilibrium. To allow for spatial registration of each participant's data to a standard coordinate system, high-resolution three-dimensional structural images were acquired in 34 axial slices co-planar with the functional scans (TR/TE/flip angle = 7.7 s/3.0 ms/12; voxel size = 0.9 × 0.9 × 4 mm; FOV = 240 mm, interslice skip = 0).

BOLD fMRI data preprocessing

Images for each subject were realigned to the first volume in the time series to correct for head motion, spatially normalized into a standard stereotactic space (Montreal Neurological Institute template) using a 12-parameter affine model (final resolution of functional images = 2 mm isotropic voxels), and smoothed to minimize noise and residual difference in gyral anatomy with a Gaussian filter, set at 6-mm full-width at half-maximum. Voxel-wise signal intensities were ratio normalized to the whole-brain global mean.

Variability in single-subject whole-brain functional volumes was determined using the Artifact Recognition Toolbox (http://www.nitrc.org/projects/artifact_detect). Individual whole-brain BOLD fMRI volumes meeting at least one of two criteria were assigned a lower weight in determination of task-specific effects: (1) significant mean-volume signal intensity variation (i.e. within volume mean signal greater or less than 4 SD of mean signal of all volumes in time series) and (2) individual volumes where scan-to-scan movement exceeded 2 mm translation or 2° rotation in any direction.

BOLD fMRI data analysis

The general linear model of SPM8 (http://www.fil.ion.ucl.ac.uk/spm) was used to conduct fMRI data analyses. Following preprocessing, linear contrasts employing canonical hemodynamic response functions were used to estimate differential effects of feedback (i.e. reward) from the contrast of Positive Feedback > Negative Feedback for each individual. Individual contrast images were then used in second-level random effects models accounting for scan-to-scan and participant-to-participant variability to determine mean condition-specific regional responses using one-sample t-tests. A statistical threshold of P < 0.05, family-wise error rate (FWE) corrected across a VS region of interest (ROI, 10 mm sphere around x = ±12, y = 12, z = −10), and ≥10 contiguous voxels was applied to the contrast of Positive > Negative feedback blocks for this analysis.

Mean BOLD values from VS clusters exhibiting a main effect of task were extracted using the Volume of Interest (VOI) tool in SPM8. These extracted values were then entered into statistical models using ibm spss statistics 20.0 (Chicago, IL, USA) and Mplus 6.12 (Muthén & Muthén 1998–2010). Importantly, by extracting VS BOLD parameter estimates from the functional clusters activated by our paradigm rather than clusters specifically correlated with our independent variables of interest (i.e. problem drinking scores, GAL5.1 haplotype etc.), we preclude the possibility of any correlation coefficient inflation that may result when an explanatory covariate is used to select a ROI (Viviani 2010). We have successfully used this conservative strategy in previous reports (Carre et al. 2012; Hyde et al. 2011; Nikolova et al. 2011, 2012).

Genotyping

Genotyping was facilitated through 23andMe (23andMe, Inc., Mountain View, CA, USA). Genomic DNA from all participants was isolated from buccal cells derived from Oragene DNA self-collection kits (DNA Genotek, Inc., Kanata, Canada) customized for 23andMe. DNA extraction and genotyping were performed at the National Genetics Institute, a CLIA-certified clinical laboratory and subsidiary of Laboratory Corporation of America. The Illumina Omni Express Plus chip and a custom array containing an additional ∼300 000 SNPs were used to provide genome-wide data (Do et al. 2011; Eriksson et al. 2010; Tung et al. 2011). The Illumina Omni Express Plus chip included rs2513281, a SNP within the GAL5.1 region that was previously demonstrated to have a functional effect on the transcription of GAL in vitro (Davidson et al. 2011). We consider this SNP representative of the GAL5.1 functional haplotype, as it is in complete linkage disequilibrium with rs2513280 in Caucasian populations (R2 = 1.00, D′ = 1.00; HapMap CEU data, http://hapmap.ncbi.nlm.nih.gov/). The rs2513281 genotypes were extracted from the master database using the ‘list’ command in Plink (Purcell et al. 2007). Participants from our final sample, who had valid genotype data, were separated into one of two, distinct haplotype groups: those homozygous for the GG haplotype (i.e. GG/GG; n = 99) and those carrying the CA haplotype (i.e. GG/CA and CA/CA; n = 39). Carriers of the CA haplotype were treated as a single group to increase statistical power to detect haplotype effects, as there were only two CA homozygotes (Table 1). Genotype distribution for rs2513281 did not deviate from Hardy–Weinberg equilibrium (χ2 = 0.50; P = 0.48).

Table 1. Gender distribution among the haplotype groups
 GAL5.1 haplotype
 GG/GGGG/CACA/CATotal
  1. No significant differences emerge.

Women5021172
Men4916166
Total99372138

Alcohol use

Self-reported alcohol use, including problem drinking, during the past year was ascertained using the Alcohol Use Disorder Identification Test (AUDIT). The AUDIT is a 10-item self-report checklist specifically developed to identify harmful drinking patterns before explicit dependence and physical health problems have developed (Saunders et al. 1993). Possible scores range from 0 to 41, and individuals with current alcoholism or more severe alcohol-related problems tend to score above 8 (Conigrave et al. 1995; Saunders et al. 1993). Some authors have suggested lower cutoff scores for women (Reinert & Allen 2007). However, we model AUDIT scores continuously and take into account the moderating role of gender, thus any difference in cutoff scores is unlikely to affect our results. The range of AUDIT scores in the current sample was 0–20 with a mean of 5.44 (±4.40 SD). Thus, substantial variability in problem drinking exists in our sample providing an opportunity for mapping onto individual differences in brain function and genotype. Additionally, the AUDIT score distribution was not sufficiently skewed or kurtotic (skewness = 0.873, kurtosis = 0.585) to pose concerns regarding violating assumptions of normality.

Statistical analyses

Three principal analyses were conducted. First, a 2 × 2 (Gender × GAL5.1 haplotype) analysis of variance was conducted on VS reactivity bilaterally. Significant interactions were probed using post hoc two-tailed t-tests (LSD-corrected). Second, linear regression analyses were conducted using VS reactivity as an independent variable predicting problem drinking, as measured by total scores on the AUDIT. Third, a moderated mediation analysis was performed to probe any gender-specific effects of GAL5.1 haplotype on problem drinking mediated by its effects on VS reactivity. The moderated mediation model was tested using path analysis in Mplus 6.12 (Muthén & Muthén 1998–2010), as previously described (Preacher et al. 2007). Bootstrapped 95% confidence intervals (CIs) (bias-corrected) for all indirect effects are reported alongside parameter estimates, consistent with published guidelines (Efron 1987; Preacher et al. 2007). Overall model fit was assessed using the root mean square error of approximation (RMSEA) as well as the comparative fit index (CFI).

Results

Sample demographics

Consistent with the higher prevalence of alcohol-related disorders in men than in women (Harford & Grant 1994), the men in our sample had higher problem drinking scores than women [men: 6.64 ± 4.58; women: 4.35 ± 3.96; t(136) = 3.15, P = 0.002]. Specifically, men scored higher on items related to drinking frequency and quantity, as well as overall hazardous drinking patterns. No gender differences emerged on any other items, including items related to guilt or dependence symptoms (Table S2). These gender differences are consistent with those reported in the literature (Martens et al. 2009).

In addition, age was a positive predictor of self-reported alcohol use (b = 0.172, P = 0.044), such that its effects remained at a trend level even when controlling for gender (ΔR2 = 0.19, b = 0.140, P = 0.094). There were no gender differences in GAL5.1 haplotype distribution (Table 1; χ2 < 0.43, P > 0.53) or between groups with and without DSM-IV Axis I diagnosis (Table 2; χ2 = 2.41, P = 0.13). In addition, the two GAL5.1 haplotype groups did not differ in age (P = 0.807) or diagnostic status (χ2 = 0.028, P = 0.866). Participants with a DSM-IV diagnosis reported more problem drinking than those without [t(136) = −4.76, P < 0.001, corrected for inequality of variances, Levene's test P = 0.048). However, these participants were included in our analyses to better model the dimensional variability in problem drinking. Not surprisingly, these participants also had higher VS reactivity [right: t(136) = −2.62, P = 0.010; left: t(136) = −3.91, P = 0.0002]. Regardless, all analyses were conducted with and without diagnostic status, in addition to gender and age, as a covariate.

Table 2. Gender distribution between participant groups that do and do not meet criteria for a DSM-IV Axis I diagnosis
 No diagnosisDiagnosisTotal
  1. No significant differences emerge.

Women621072
Men501666
Total11226138

GAL5.1 haplotype, gender and VS reactivity

Consistent with previous studies (Forbes et al. 2009; Hariri et al. 2006), our paradigm elicited significant reward-related (i.e. Positive > Negative feedback) bilateral VS reactivity (Fig. 1a). There was no direct effect of GAL5.1 haplotype on problem drinking or VS reactivity in either hemisphere (all P-values > 0.46). However, consistent with prior research showing gender-specific effects of galanin-related polymorphisms on behavior (Belfer et al. 2006; Unschuld et al. 2010), we found a significant gender-by-genotype interaction predicting VS reactivity bilaterally (right hemisphere: F1,134 = 8.08, P = 0.005, Fig. 1b; left hemisphere: F1,134 = 5.39, P = 0.022, Fig. 1c). Post hoc tests (LSD-corrected) revealed that within the CA haplotype group, male participants had higher VS reactivity than women bilaterally (right hemisphere: P = 0.001, left hemisphere: P = 0.039). VS reactivity between men and women did not differ significantly in the GG haplotype group (P-values > 0.24). In addition, women with the CA haplotype had lower right VS reactivity than women homozygous for the GG haplotype (p = 0.015). A similar effect, albeit marginally significant, emerged in the left hemisphere in women (p = 0.060). While the opposite pattern was observed in men, the effect did not reach statistical significance (right hemisphere: P = 0.11; left hemisphere: P = 0.16). The interaction effects for both hemispheres remained significant when controlling for age and current Axis I diagnosis (right hemisphere: F1,132 = 11.75, P = 0.001; left hemisphere: F1,132 = 10.16, P = 0.002).

Figure 1.

Gender x GAL5.1 haplotype interaction effect on bilateral VS reactivity. (a) Statistical parametric map illustrating mean bilateral reward-related VS reactivity (left: x = −12, y = 6, z = −10, t = 8.69, P = 1.77 × 10−12, kE = 313; right: x = 12, y = 10, z = −8, t = 8.63, P = 1.90 × 10−12, kE = 250). Activation clusters are overlaid onto canonical structural brain images in the coronal plane (Y = 8). Color bar represents t values. GAL5.1 haplotype interacted with gender to predict VS reactivity on the right (b) and left (c) side, such that men carrying the CA haplotype had significantly higher VS reactivity than women carriers. In addition, women CA haplotype carriers had lower VS reactivity than women homozygous for the GG haplotype, particularly in the right hemisphere. Error bars represent SEM. *P < 0.05; #P = 0.060. a.u., arbitrary units; lVS, left VS; rVS, right VS; SEM, standard error of the mean.

VS reactivity and alcohol use

Consistent with studies showing an association between VS reactivity and self-reported impulsivity (Forbes et al. 2009), as well as reports linking impulsivity to drug use (Kreek et al. 2005), we found a positive correlation between VS reactivity and problem drinking (right hemisphere: Adj. R2 = 0.062, b = 0.248, P = 0.003, Fig. 2a; left hemisphere: Adj. R2 = 0.082, b = 0.286, P = 0.001, Fig. 2b). This effect remained significant when controlling for gender, age, current Axis I diagnosis and GAL5.1 haplotype (right hemisphere: ΔR2 = 0.024, b = 0.159, P = 0.036; left hemisphere: ΔR2 = 0.023, b = 0.162, P = 0.038). Problem drinking was not significantly associated with GAL5.1 haplotype or its interaction with gender (P-values > 0.27). Likewise, interactions between VS reactivity and either gender or GAL5.1 haplotype did not significantly predict problem drinking (P-values > 0.44).

Figure 2.

Linear correlations between bilateral VS reactivity and problem drinking scores. VS reactivity in the right (a) and left (b) hemisphere correlated positively with self-reported problem drinking (total scores from the AUDIT). The correlations remain significant when the participant with an rVS value of 0.91 and lVS value of 0.99 is removed from analyses (right hemisphere: Adj. R2 = 0.039, b = 0.197, P = 0.021; left hemisphere: Adj. R2 = 0.057, b = 0.239, P = 0.005). a.u., arbitrary units; rVS, right VS; lVS, left VS.

Indirect effects of GAL5.1 haplotype on problem drinking

In light of the significant interaction effect of GAL5.1 haplotype and gender on VS reactivity, as well as the correlation between VS reactivity and problem drinking, we examined possible gender-specific indirect (i.e. mediated) effects of GAL5.1 haplotype on problem drinking using a moderated mediation model (Fig. 3a). Given the stronger effects of GAL5.1 haplotype on VS reactivity in the right hemisphere, we focused our analyses on this region. As current DMS-IV Axis I diagnosis affected both VS reactivity and AUDIT scores, and age affected AUDIT scores, we included diagnosis and age as covariates in the model (Fig. 3b). The model fit the data very well (RMSEA = 0.000; CFI = 1.00; χ2 = 1.048, df = 2, P = 0.59). In addition, we found a significant mediation effect, whereby relatively reduced right VS reactivity mediated an association between the CA haplotype and lesser problem drinking in women (b = 0.491, 95% CI: 0.075–1.185). In contrast, increased VS reactivity mediated an association between the CA haplotype and greater problem drinking in men (b = −0.391, 95% CI: −1.005 to −0.044; Fig. 3b). Even though the overall model fit decreased (RMSEA = 0.141; CFI = 0.89; χ2 = 3.751, df = 1, P = 0.053), the significance of these indirect effects remained unchanged when the covariates were removed from the model (women: b = 0.661, 95% CI; 0.166–1.457; men: b = −0.477, 95% CI: −1.193 to −0.059). A similar pattern was observed for left VS reactivity; however, the significance of the paths was lower overall, with the indirect effect of GAL5.1 haplotype on problem drinking in men becoming non-significant in the model without covariates (Table S3), further strengthening the notion that the effects of GAL5.1 haplotype on VS reactivity and problem drinking are stronger in women. Removing participants meeting criteria for cannabis abuse (n = 1) and dependence (n = 1), or controlling for smoking status (0 = non-smoker, 1 = smoker) did not alter the results significantly (Tables S4 and S5). Demonstrating the specificity of our findings to alcohol use, the same model did not predict substantial variability in depression scores, as measured by the Center for Epidemiological Studies-Depression (CES-D) scale (Weissman et al. 1977) (Table S6).

Figure 3.

Moderated mediation model testing the gender-specific effects of GAL5.1 haplotype on problem drinking mediated by VS reactivity in the right hemisphere. (a) Conceptual depiction of the moderated mediation model tested. Gender is hypothesized to moderate the path between GAL5.1 haplotype and VS reactivity. Thus, VS reactivity is hypothesized to mediate the relationship between GAL5.1 haplotype and problem drinking (total scores for the AUDIT) in a gender-specific manner. (b) Results from a path analysis testing the moderated mediation model presented in (a). VS reactivity mediated the relationship between GAL5.1 haplotype and problem drinking differentially for women and for men. Specifically, among women, carriers of the minor CA haplotype had significantly lower VS reactivity and AUDIT scores, relative to GG haplotype homozygotes. The opposite relationship emerged for men. Numbers represent unstandardized parameter estimates with bootstrapped standard errors presented in parentheses. The circles labeled e1 and e2 denote the variance in VS reactivity and AUDIT scores unaccounted for by the model. *P < 0.05; **P < 0.001.

Discussion

In this study, we report that women homozygous for the GAL5.1 GG haplotype have relatively increased reward-related VS reactivity, which in turn predicts increased problem drinking. In men, however, it is the CA haplotype that predicts relatively increased VS reactivity and problem drinking. These effects are generally consistent with the demonstrated role of galanin in modulating neural processes associated with variability in reward-related appetitive and consummatory behaviors. In light of the divergent directional effects of the GAL5.1 haplotype between genders, we first consider the observed findings in women and then those in men.

As the GG haplotype has been linked to increased GAL promoter activity in hypothalamic neurons in vitro (Davidson et al. 2011), the heightened VS reactivity observed in women homozygous for this haplotype may reflect increased galanin production in the hypothalamus and heightened galanin signaling, and subsequently increased striatal DA release (Rada et al. 1998). While the precise molecular mechanisms for our findings have yet to be established, prior studies suggest that the positive relationship between galanin and alcohol consumption may be mediated through direct projections from galanin-synthesizing PVN neurons to dopaminergic neurons in the ventral tegmental area (VTA) and their projection targets in the NAcc (Hawes & Picciotto 2004). Such direct hypothalamic modulation of reward-related brain regions may also explain the involvement of galanin in regulating food consumption (Barson et al. 2011). Alternatively, these effects may be mediated by an indirect path from the PVN to the VTA via the parabrachial nucleus or the solitary tract nucleus to midbrain cholinergic cell groups 5 (pedunculopontine nucleus) and 6 (laterodorsal tegmental nucleus), which in turn project to DA cell bodies in the VTA (Yeomans et al. 1993).

Several prior studies have suggested galanin may exert a host of its biochemical and behavioral effects in a gender-specific manner. Evidence in support of this proposition comes from studies showing that estradiol treatment of ovariectomized rats results in higher levels of immunoreactive galanin in both reproductive and non-reproductive areas of the brain (Rugarn et al. 1999). The existence of sex-specific regulation of GAL expression is further corroborated by the presence of estrogen response elements in the human GAL gene (Howard et al. 1997). Thus, the effects of GAL5.1 haplotype may be more pronounced in the women in our sample due to their overall higher levels of circulating estrogen and, hence, heightened estrogen-mediated regulation of GAL expression and all galanin-regulated processes downstream.

Interestingly, the opposite pattern of GAL5.1 haplotype effects on VS reactivity and problem drinking was observed in men. Even though these effects fell just short of significance in some of our analyses, the opposite patterns we observe in men and women suggest the gender-specificity of the effects we report may be the product of more complex interactions among hormones and neurotransmitters. Prior studies in rats showing that galanin increase in the PVN results in DA release and heightened alcohol consumption used only males (Rada et al. 1998, 2004). However, one study explored the effects of a galanin gene knockout (GALKO) on both male and female mice and revealed significant gender specificity of nearly all observed knockout effects (Karatayev et al. 2010). Most notably, this study reported that female GALKO mice exhibited a reduction in alcohol consumption compared to their wild-type counterparts, while no such effect occurred in male mice. These results are consistent with the finding that women with the CA haplotype in our sample show reduced VS reactivity and, hence also, decreased problem drinking, compared to either women homozygous for the GG haplotype or men with the CA haplotype.

The opposite genotype effect on VS reactivity and drinking behavior we see in men from our sample may reflect GAL effects that extend beyond the PVN, as suggested by gender-specific neurochemical findings in GALKO mice. Specifically, while both male and female GALKO mice show a galanin decrease in the PVN, female GALKO mice exhibit a concomitant decrease in orexin (ORX) and melanin-concentrating hormone (MCH) in the perifornical lateral hypothalamus. Interestingly, male GALKO mice show a marked increase in these same peptides (Karatayev et al. 2010). Critically, another study has shown that pharmacologic administration of both ORX and MCH is associated with increased alcohol consumption in male rats (Duncan et al. 2005; Schneider et al. 2007). The increase in those peptides in male GALKO mice may thus compensate for any decrease in alcohol consumption resulting from reductions in galanin levels. These results suggest that reduction of Gal expression may have opposite downstream neurochemical effects in male and female mice, which may in turn account for the gender-specificity of the neural and behavioral changes associated with genetically driven variability in galanin signaling. A similar mechanism may exist in humans, wherein the neural and behavioral effects of any reductions in GAL expression associated with the GAL5.1 haplotype vary between men and women due to gender-specific divergent downstream neurochemical changes.

The gender specificity of our findings raises the intriguing possibility that galanin may be implicated in commonly observed gender differences in the prevalence of alcoholism subtypes. Epidemiological studies have found an excess of men among patients diagnosed with externalizing subtypes of alcohol dependence (i.e. subtypes frequently comorbid with antisocial personality disorder and additional substance abuse problems), while women predominate among internalizing subtypes (i.e. subtypes frequently comorbid with mood and anxiety disorders) (Epstein et al. 2002; Moss et al. 2007; but see Dawson et al. 2010). In this study, we found that gender moderates the path between GAL5.1 haplotype and VS reactivity, but the path linking VS reactivity to problem drinking scores did not show gender specificity. As relatively increased VS reactivity has been implicated in heightened impulsivity (Forbes et al. 2009) and steeper delay discounting (Hariri et al. 2006), which are in turn associated with externalizing psychopathology (Krueger et al. 2007), we believe our findings reflect primarily the externalizing pathway to alcoholism in both genders. Given prior links between galanin and depression (Unschuld et al. 2010), however, future studies could attempt to identify specific pathways via which genetic variation in galanin signaling may affect additional neural phenotypes (e.g. threat-related amygdala reactivity) to increase propensity for internalizing psychopathologies, including relief drinking.

To our knowledge, this is the first study to demonstrate an effect of a functional polymorphism affecting GAL expression on reward system reactivity and problem drinking in humans. The few human studies that have investigated the effects of galanin on alcohol use have focused on polymorphisms other than the GAL5.1 haplotype. One study has linked a seven SNP GAL haplotype to alcoholism in two ethnically distinct male populations (Belfer et al. 2006). Another study by the same group has associated a SNP in the galanin receptor 3 gene (GALR3) with alcoholism in Caucasians (Belfer et al. 2007). However, as the authors point out in both studies, there is no evidence of functionality for those SNPs. Thus, our findings not only provide further support for the involvement of the GAL gene and galanin signaling more broadly in the regulation of alcohol consumption, but also extend this line of research to include a polymorphism of known functionality.

It is worth noting that, while we observed similar patterns for the left and right VS, our results were stronger in the right hemisphere. These laterality effects are consistent with results from prior reports using the same VS reactivity paradigm in at least two independent samples (Forbes et al. 2009; Nikolova et al. 2011, 2012). The consistency with which these laterality effects occur may suggest that the number-guessing fMRI paradigm used here and in these past studies may activate the right VS more strongly due to its visuospatial component. Alternatively, it is possible that the hemisphere-specific results reflect natural asymmetries in neurotransmitter action (Besson & Louilot 1995; Merali et al. 2004).

This study is not without limitations. First, while our sample (N = 138) is comparable to most other imaging genetics studies, future studies may benefit from employing larger samples to afford better powered analyses that can clarify some of the weaker or trend-level effects we observed within our moderated mediation framework. Relatedly, we analyzed GAL5.1 CA homozygotes and heterozygotes together, as there were only two participants in the former group. Thus, we were unable to examine any additive effects of the minor CA haplotype. Such effects would only be testable in samples large enough to allow for the presence of a substantial number of homozygotes for both minor alleles. Second, even though several participants from our sample met criteria for alcohol and substance use related disorders (Table S1), the sample consisted primarily of high-functioning young adults attending competitive higher education institutions. This may limit the clinical significance of our results. However, college drinking patterns have been shown to predict substance use problems later in life (Merline et al. 2008; Zucker 2008). Thus, our findings may have predictive value for understanding downstream risk for substance use disorders including alcoholism. Finally, while the number-guessing paradigm we used here has been shown to reliably engage reward-related neural circuitry in several prior studies (Carre et al. 2013; Forbes et al. 2009; Hariri et al. 2006; Nikolova et al. 2012), it does so in a way that does not allow for the parsing of different components of reward processing (i.e. ‘wanting’, ‘liking’ and ‘learning’). Future studies using tasks specifically designed to distinguish between these components may allow for a more precise dissection of the specific phases of reward processing modulated by genetically driven variability in galanin signaling.

In summary, we provide novel evidence that a functional haplotype in a remote enhancer region upstream of the human GAL gene is associated with differences in problem drinking via its gender-specific modulatory effects on reward-related reactivity of the VS. These results are consistent with non-human animal studies which have implicated hypothalamic galanin in alcohol consumption (Rada et al. 2004), as well as human genetic association studies linking variants in galanin-related genes to various forms of substance abuse and dependence (Belfer et al. 2006, 2007; Gold et al. 2012; Jackson et al. 2011). Thus, future research could benefit from encompassing variants affecting hypothalamic peptides in the search for pathways modulating appetitive behaviors in humans, and it may be worthwhile to further consider galanin as a potential therapeutic target for substance use disorder treatment and, possibly, prevention.

Acknowledgments

Y. S. N. is a Howard Hughes Medical Institute International Student Research Fellow. E. M. D. is a full-time employee of 23andMe, Inc. (Mountain View, CA). The Duke Neurogenetics Study is supported by Duke University. E. K. S. and A. R. H. declare no potential conflict of interest.

Ancillary