Genetic variation of the growth hormone secretagogue receptor gene is associated with alcohol use disorders identification test scores and smoking.

The multifaceted gut-brain peptide ghrelin and its receptor (GHSR-1a) are implicated in mechanisms regulating not only the energy balance but also the reward circuitry. In our pre-clinical models, we have shown that ghrelin increases whereas GHSR-1a antagonists decrease alcohol consumption and the motivation to consume alcohol in rodents. Moreover, ghrelin signaling is required for the rewarding properties of addictive drugs including alcohol and nicotine in rodents. Given the hereditary component underlying addictive behaviors and disorders, we sought to investigate whether single nucleotide polymorphisms (SNPs) located in the pre-proghrelin gene (GHRL) and GHSR-1a gene (GHSR) are associated with alcohol use, measured by the alcohol use disorders identification test (AUDIT) and smoking. Two SNPs located in GHRL, rs4684677 (Gln90Leu) and rs696217 (Leu72Met), and one in GHSR, rs2948694, were genotyped in a subset (n = 4161) of a Finnish population-based cohort, the Genetics of Sexuality and Aggression project. The effect of these SNPs on AUDIT scores and smoking was investigated using linear and logistic regressions, respectively. We found that the minor allele of the rs2948694 SNP was nominally associated with higher AUDIT scores (P = 0.0204, recessive model) and smoking (P = 0.0002, dominant model). Furthermore, post hoc analyses showed that this risk allele was also associated with increased likelihood of having high level of alcohol problems as determined by AUDIT scores ≥ 16 (P = 0.0043, recessive model). These convergent findings lend further support for the hypothesized involvement of ghrelin signaling in addictive disorders.

Introduction

Recent pre‐clinical and clinical studies suggest that the orexigenic ghrelin signaling system may be of importance for addictive behaviors such as alcohol use disorder and compulsive over‐eating (for review, see Dickson et al. 2011; Leggio et al. 2011). Pre‐clinical studies have shown that ghrelin, just as addictive drugs (Imperato & Di Chiara 1986; Engel et al. 1988; Larsson & Engel 2004; Larsson et al. 2005) and natural rewards (Hernandez & Hoebel 1988; Rada et al. 2000), activates the cholinergic‐dopaminergic reward link in rodents (Jerlhag et al. 2007). Ghrelin has thus been suggested to increase the incentive salience of motivated behaviors such as those associated with drug taking. This hypothesis has been confirmed in several of our pre‐clinical studies in which pharmacological suppression of the growth hormone secretagogue receptor subtype 1a (GHSR‐1a), also known as the ghrelin receptor, was shown to attenuate the rewarding properties of alcohol and nicotine (Jerlhag et al. 2009; Jerlhag & Engel 2011). Further studies involving rodents have shown that ghrelin increases alcohol consumption while a GHSR‐1a antagonist decreases alcohol intake, operant self‐administration of alcohol and prevents relapse drinking (Jerlhag et al. 2009; Kaur & Ryabinin 2010; Landgren et al. 2012; Bahi et al. 2013; Suchankova et al. 2013b).

Plasma and serum ghrelin levels have been investigated in several studies in alcohol‐dependent individuals with varying outcomes possibly relating to inherent differences between study samples (for reviews, see Wurst et al. 2007; Kenna et al. 2012). Nevertheless, some studies suggest a positive correlation between ghrelin levels and alcohol craving in abstinent alcohol‐dependent individuals (Addolorato et al. 2006; Hillemacher et al. 2007; Koopmann et al. 2012; Leggio et al. 2012), possibly implicating the ghrelin signaling system as a potential target in the pharmacological treatment of alcohol use disorders in humans. A recent study reports a dose‐dependent increased alcohol craving in heavy‐drinking alcohol‐dependent subjects receiving exogenous ghrelin intravenously (Leggio et al. 2014). Furthermore, clinical studies investigating the effect of smoking on ghrelin levels report conflicting results. Some show no difference in baseline levels between smokers and non‐smokers or with craving in acute nicotine withdrawal (Kokkinos et al. 2007; Mutschler et al. 2012), while others report the decrease of ghrelin levels in saliva after acute smoking (Kaabi & Khalifa 2014) as well as, contradictory, increase in plasma ghrelin levels (Fagerberg, Hulten & Hulthe 2003; Bouros et al. 2006). The short‐term effect of smoking has also been shown to cause suppression in ghrelin plasma levels in non‐smoker but not in smokers (Kokkinos et al. 2007).

We have previously shown in human genetic studies that variants of the pre‐proghrelin gene (GHRL) and GHSR‐1a gene (GHSR) are associated with various factors related to alcohol, nicotine and amphetamine dependence (Landgren et al. 2008, 2010, 2011a; Suchankova et al. 2013a). On the basis of these pre‐clinical and clinical observations, we sought to investigate whether single nucleotide polymorphisms (SNP) located in the GHRL and GHSR gene are associated with alcohol use assessed by the alcohol use disorders identification test (AUDIT) as well as smoking in a large Finnish cohort.

Materials and Methods

Subjects

The Genetics of Sexuality and Aggression sample is a population‐based cohort of Finnish twins and siblings of twins. The subsample used in the present study stems from the second data collection completed in 2006 (for data collection procedure, see Johansson et al. 2013). DNA was extracted from saliva samples (for characteristics, see Table 1). Some genotyping failure was seen for each of the analyzed SNPs resulting in a slight variation in n. The genotypes were imputed in cases where only one of the monozygotic twins was genotyped. A number of subjects with at least one successfully genotyped SNP in the resulting sample were 4161 subjects from a total of 1756 families (581 monozygotic and 380 dizygotic twin‐pairs, 910 siblings and 1329 singletons).

Table 1

Sample characteristics for genotyped individuals

	Overall	Females	Males
n a	4161 (100 percent)	2463 (59.2 percent)	1698 (40.8 percent)
Age	26.39 (5.07)	26.40 (5.25)	26.38 (4.80)
AUDIT
n	4133 (100 percent)	2451 (59.3 percent)	1682 (40.7 percent)
AUDIT sum items 1–10	6.63 (4.99)	5.48 (4.37)	8.31 (5.36)
AUDIT ≥ 16	256 (6.2 percent)	89 (3.6 percent)	167 (9.9 percent)
AUDIT ≤ 1	563 (13.6 percent)	405 (16.5 percent)	158 (9.4 percent)
Cigarettes per day
n	4112 (100 percent)	2441 (59.4 percent)	1671 (40.6)
none	3085 (75.0 percent)	1903 (77.6 percent)	1196 (71.3 percent)
≤ 10	556 (13.5 percent)	339 (13.8 percent)	218 (13.0 percent)
11–20	394 (9.6 percent)	185 (7.5 percent)	210 (12.6 percent)
21–30	75 (1.8 percent)	25 (1.0 percent)	50 (3.0 percent)
> 30	2 (0.05 percent)	0	2 (0.1 percent)

Number of subjects with information on genotype for at least one of the analyzed single nucleotide polymorphisms. Data presented as number of cases (percentage) or mean (standard deviation). AUDIT = Alcohol Use Disorder Identification Test.

Alcohol use assessment

The participants were asked to fill out the AUDIT, a self‐report questionnaire used in the health care to identify individuals with hazardous and harmful patterns of alcohol consumption (Babor et al. 2001). It consists of 10 questions with a 5‐ and 3‐point response format. The response is scored to give a total AUDIT score ranging from 0 to 40. According to the World Health Organization, an AUDIT score of ≥ 16 or above represents a high level of alcohol problems, whereas an AUDIT score of ≤ 1 is reported by alcohol abstainers (Babor et al. 2001).

Smoking

The present study also aimed to investigate potential association between the studied SNPs and smoking status reported using questions about current smoking habits. The participants were asked how many cigarettes they smoke on an average day with a 5‐point response format: not at all; ≤ 10; 11–20; 21–30; > 30 (Table 1). This outcome was significantly correlated to AUDIT scores in the whole sample (partial correlation controlling for gender and age, r = 0.32, n = 4071, P < 0.001). According to the Fagerström Test for Nicotine Dependence (Heatherton et al. 1991), smoking fewer than 11 cigarettes per day does not add to the final score of nicotine dependence. Subjects in this group were thus excluded from all smoking‐related analyses while individuals that reported to smoke more than 10 cigarettes per day (n = 472) were grouped together in order to capture individuals that were more likely to be nicotine‐dependent.

DNA extraction, SNP selection and genotyping

The DNA collection and extraction procedure has been described in detail previously (Jern et al. 2012). Briefly, Oragene™ DNA self‐collection kits (DNA Genotek, Inc., Kanata, ON, Canada) were used when collecting saliva samples from participants. The kits were posted to the participants along with the manufacturer's instructions on collecting the samples (DNA Genotek, Inc.) and participants were instructed to deposit approximately 2 ml of saliva into the collection cup before returning it by mail. DNA was stabilized in the cup upon closure and stored at −20°C until used.

Two missense SNPs in the GHRL (rs4684677 or Gln90Leu and rs696217 or Leu72Met) and one SNPs in the GHSR (rs2948694) were selected on the basis of previous results (Landgren et al. 2011a, 2010, 2008, 2011c). For closer description of the initial selection of these SNPs, see Landgren et al. (2008). SNPs with a previously reported minor allele frequency of < 5 percent were not included in the present study.

Genotyping of the SNPs was made by the LGC Genomics in the UK (http://www.lgcgenomics.com) using the KASPar chemistry—a competitive allele‐specific polymerase chain reaction (PCR) SNP genotyping system performed with fluorescence resonance energy transfer (FRET) quencher cassette oligos. For details on the studied SNPs, see Table 2.

Table 2

Studied single nucleotide polymorphisms (SNPs)

Gene	SNP	Positiona	Alleles	SNP location	SNP type
GHRL	rs4684677	10328453	A/T	Exon 3	Missense (Gln90Leu)
	rs696217	10331457	G/T	Exon 2	Missense (Leu72Met)
GHSR	rs2948694	172165163	A/G	Intron 1	Intron

Position on chromosome 3 for the studied SNPs in GHRL and GHSR. GHRL = pre‐proghrelin gene; GHSR = growth hormone secretagogue receptor gene.

Statistical analyses

Deviation from Hardy‐Weinberg equilibrium (HWE) (P < 0.01) was assessed for all genotyped SNPs using Haploview (Barrett et al. 2005).

Primary analysis involved genotypic effects on the main outcome variables AUDIT and smoking. Additive, dominant and recessive models of inheritance for the minor allele were all evaluated. Age and sex were included as covariates. Total sum of AUDIT items 1–10 was investigated using linear regression. Individuals smoking more than 10 cigarettes per day were grouped together and compared against the group that reported to be non‐smokers using logistic regression. Due to the fact that a large part of the present cohort consists of twins and siblings, gene‐dropping was used to maintain the between‐subjects dependence when evaluating the significance of the effects by repeated simulations (MacCluer et al. 1986).

We also investigated the effect of genotype on AUDIT scores and smoking in models adjusting for not only age and gender but also smoking and AUDIT scores, respectively. Finally, a logistic regression analysis was performed for the associated SNP to estimate odds ratios (ORs) for individuals with high level of alcohol problems, i.e. AUDIT ≥ 16 (n = 256) compared with alcohol abstainers, i.e. AUDIT ≤ 1 (n = 563).

A significance level of 0.05 was used and all presented P‐values are empirical estimates unless otherwise specified. P‐values for main outcome variables (i.e. AUDIT and smoking) were Bonferroni corrected for three models, three SNPs and two primary endpoints (i.e. 18) and designated pc. The statistical analysis was carried out using SPSS for Mac (Version 19.0.0.1, SPSS, Chicago, IL, USA) and R (Team RDC 2011).

Results

All studied SNPs had a HWE P‐value > 0.01 and a minor allele frequency of > 5 percent (Table 3). The minor allele of the rs2948694 SNP (G) located in intron 1 of the GHSR gene was associated with increased AUDIT scores (ΔAUDIT = 1.2; P = 0.0204; pc = 0.3672) in a recessive model as well as with smoking in an dominant model (OR = 1.48; P = 0.0002; pc = 0.0036; Table 4). See the Supporting Information for results on the primary outcomes using the other models (Supporting Information Table S1). When a subset of the subjects where selected and classified according to abstainers (AUDIT ≤ 1) and high consumers (AUDIT ≥ 16), the homozygotes for the risk allele had OR = 3.4 to be high consumers (P = 0.0043). Smokers (reporting to smoke > 10 cigarettes per day) had higher AUDIT scores [Mean(standard deviation, SD) AUDIT score in non‐smokers = 5.63(4.35), Mean(SD) AUDIT score in smokers = 10.18(5.76)] and 61.6 percent of the subjects with AUDIT ≥ 16 were smokers compared with 9.2 percent in the AUDIT ≤ 1 group. Nevertheless, when the primary analyses involving AUDIT scores or smoking were adjusted for smoking or AUDIT scores, respectively, the association between the outcome variables and the rs2948694 SNP remained significant (P = 0.0489, P = 0.0008, respectively; Table 4). Neither of the investigated GHRL SNPs was associated with AUDIT scores or smoking in any of the genetic models (Gln90Leu: P ≥ 0.12; Leu72Met: P ≥ 0.17; Supporting Information Table S1).

Table 3

Marker statistics for the genotyped SNPs in and genes

Gene	SNP	MAF	HWE p	dd	Dd	DD	Call ratea
GHRL	rs4684677	13.1 percent	0.25	71 (1.8 percent)	867 (22.5 percent)	2923 (75.7 percent)	93.3 percent
	rs696217	12.1 percent	0.68	69 (1.6 percent)	891 (21.0 percent)	3282 (77.4 percent)	95.3 percent
GHSR	rs2948694	20.4 percent	0.02	192 (4.8 percent)	1241 (31.1 percent)	2558 (64.1 percent)	96.2 percent

Not including imputed genotypes for monozygotic twins. rs4684677 and rs2948694 were genotyped in a samples of 4039 subjects, rs696217 was genotyped in a sample of 4325. GHRL = pre‐proghrelin gene; GHSR = growth hormone secretagogue receptor gene; HWE = Hardy‐Weinberg equilibrium; MAF = minor allele frequency; SNP = single nucleotide polymorphism; minor homozygotic, heterozygotic and major homozygotic genotype groups dd/Dd/DD.

Table 4

Associations between rs2948694, AUDIT scores and smoking

	n	b	OR	P	pc d
AUDIT a	3966	0.96		0.0204	0.3672
AUDIT ≤ 1 versus ≥ 16 b	789		3.37	0.0043
Smoking b	3411		1.48	0.0002	0.0036
Smoking b adjusted for AUDIT	3410		1.46	0.0008
AUDIT a in subset c	3410	0.81		0.0475
AUDIT a adjusted for smoking	3410	0.78		0.0489

Data presented as empirical estimates obtained by permutation (50 000 gene‐dropping simulations) of either alinear regression or blogistic regression. A recessive model was used for all analyses concerning AUDIT and a dominant model was used for smoking. The b represents increased AUDIT and the odds ratio (OR) represents increased risk of smoking or having an AUDIT ≥ 16. All analyses were performed by controlling for age and gender. AUDIT was reanalyzed in a matched csubset for comparison when controlling for smoking. Smoking variable is coded as 0 (non‐smokers), 1 (smokers > 10 cigarettes/day). dBonferroni corrected. AUDIT = Alcohol Use Disorder Identification Test.

Using available data from 1000GENOMES:phase_3_FIN confirmed that the rs2948694 has a higher minor allele frequency in the Finnish in Finland subpopulation (22 percent) compared with the European (11 percent) and was further found to capture 11 percent of alleles across the GHSR gene at r2 ≥ 0.8 using the Haploview Tagger function. Submitting the GHSR gene to the online alternative splice site predictor tool (http://wangcomputing.com/assp) predicted 39 sites as confident (≥ 0.9) putative splice sites and one of these sites included the rs2948694 as a cryptic donor alternative splice site. The G‐allele of the SNP was found to increase the confidence score of this splice site prediction. We further explored the possible association between rs2948694 and GHSR expression using the publicly available BrainCloud (Colantuoni et al. 2011), an application consisting of post‐mortem human brains collected at ages from fetal development to senescence (n = 269). Expression of GHSR in prefrontal cortex was analyzed against a proxy SNP for rs2948694 (not genotyped in this sample), namely rs16845548 (1000GENOMES:phase_3_FIN; distance = 20137 bp, r2 = 1.0, D′ = 1.0), in a linear regression controlling for age, gender, postmortem interval, pH, RNA integrity number and ancestral descent. Using a dominant model [heterozygotes (n = 49) pooled with the rare homozygous group (n = 2)], we found that this SNP was significantly associated with GHSR expression with lower expression in the pooled groups compared with the more common homozygous group (β = −0.110, P = 0.020).

Discussion

The present study is the first to report associations between the SNP rs2948694 in the GHSR gene and AUDIT scores as well as smoking. The association to smoking was not mediated by the association to AUDIT and vice versa. The data are in line with our previous studies in which the minor allele of rs2948694 was associated with increased risk of amphetamine dependence (Suchankova et al. 2013a) as well as increased body mass index in heavy alcohol‐consuming individuals (Landgren et al. 2008). The present study further reports a significant association between a proxy SNP for rs2948694 and GHSR expression in the prefrontal cortex of the human brain, providing the first evidence for a possible functional role of this SNP. The previous association between the rs4684677 SNP and composite score of drug use from the addiction severity index (Suchankova et al. 2013a) was not backed up by the current study. Nevertheless, current and previous association studies report that the rs2948694 may be of importance for reward‐ and dependence‐related phenotypes.

Genetic variation of the ghrelin signaling system has previously been associated with paternal alcohol dependence, reported withdrawal symptoms, smoking and type 2 alcohol dependence (i.e. early‐onset and more heredity driven type of alcohol dependence) (Landgren et al. 2010). There is also a report of associations with personality traits such as self‐directedness and self‐transcendence in a sample consisting of controls and type 1 alcohol‐dependent individuals (Landgren et al. 2011a). Interestingly, low self‐directedness characterizes various kinds of addictive behaviors, raising the possibility that the system is involved in normal mental functions, which in turn may render a person susceptible to addictive behaviors.

Preliminary data on the possible functional value of the rs2948694 SNP reveal that it may affect GHSR expression in the prefrontal cortex of the human brain. Moreover, its location in intron 1 of GHSR leads the authors to speculate on a possible effect of this variant on the splicing of the gene especially given the prediction of a cryptic donor alternative splice site at the SNPs location. The GHSR gene is in fact subject to alternative splicing leading to two different mRNAs, GHSR‐1a containing two exons forming the G‐protein coupled receptor with high affinity for acylated‐ghrelin and GHSR‐1b encoded by exon 1 and the first 74 bases of intron 1. The physiological role of GHSR‐1b is still unknown (Liu, Garcia & Korbonits 2011). An in vitro study reported heterodimerization between GHSR‐1a and GHSR‐1b resulting in a decrease in the constitutive signaling of GHSR‐1a possibly due to the receptors translocation from the membrane to the nucleus (Leung et al. 2007). The rs2948694 SNP, although located in a non‐coding region of either GHSR transcripts, could tentatively affect the splicing of the gene and cause a shift in GHSR‐1a and GHSR‐1b transcription, however, this remains to be confirmed. Furthermore, the SNP is in close proximity to a previously studied two‐base pair deletion (Gueorguiev et al. 2009), rs10618418, and although we were not able to find any data on the linkage disequilibrium (LD) between these two SNPs, the previous study report strikingly similar genotype frequencies as seen for rs2948694 in the present study. It is thus possible that the markers are in strong LD and it is rather the effect of the deletion that is being studied here. Nevertheless, the functional value of this deletion remains to be determined.

The present findings showing associations between AUDIT scores, smoking and a SNP in GHSR gene are supported by pre‐clinical data showing role for the GHSR‐1a in drug reward. Pre‐clinical studies have shown that pharmacological or genetic suppression of the GHSR‐1a in mice attenuated the ability of alcohol to induce locomotor stimulation, increase accumbal dopamine release and induce a conditioned place preference (Jerlhag et al. 2009). In rodents, GHSR‐1a antagonism reduces alcohol intake and operant self‐administration of alcohol and prevents relapse drinking (Jerlhag et al. 2009; Kaur & Ryabinin 2010; Landgren et al. 2012; Bahi et al. 2013; Suchankova et al. 2013b). Similarly, the rewarding properties of several psychostimulant drugs, including nicotine, cocaine as well as amphetamine, are attenuated by GHSR‐1a antagonists in rodents (Jerlhag et al. 2010; Jerlhag & Engel 2011; Wellman et al. 2011; Wellman, Clifford & Rodriguez 2013), which further supports our hypothesis that ghrelin signaling system regulates reward in general.

Clinical studies involving alcohol‐dependent subjects have shown a reversible increase in plasma levels of ghrelin during the initial phase of abstinence (Wurst et al. 2007), as well as correlations between elevated ghrelin levels and craving during alcohol withdrawal (Addolorato et al. 2006; Leggio et al. 2012). However, the circuitry and mechanism by which ghrelin modulates these cravings in humans remain to be elucidated. In light of published pre‐clinical results, a possible route could involve direct actions on the cholinergic‐dopaminergic reward link. Firstly, Ghsr are expressed in nodes of this link including the nucleus accumbens (NAc) (Landgren et al. 2011b), on dopaminergic neurons in the ventral tegmental area (VTA) (Abizaid et al. 2006) and on cholinergic neurons in the laterodorsal tegmental area (LDTg) (Dickson et al. 2010). Secondly, local administration of ghrelin into either the VTA or LDTg increases the locomotor activity, releases accumbal dopamine and increases alcohol consumption in mice (Jerlhag et al. 2007, 2009). Another possible route by which the ghrelin signaling system may alter the sensitivity of the mesolimbic dopamine system and the ability of addictive drugs to activate this system is by interacting with dopamine receptors. This was demonstrated recently in vivo and in vitro studies where GHSR‐1a were found to regulate the activity of tegmental dopamine neurons by heterodimerizing with dopamine D1‐like (Jiang, Betancourt & Smith 2006) and D2 receptors (DRD2) (Kern et al. 2012). Interestingly, GHSR‐1a has been hypothesized to act as an allosteric modulator of dopamine‐DRD2 signaling as effects are seen even in the absence of ligand (i.e. ghrelin) (Kern et al. 2012). This further suggests that GHSR‐1a may function in areas with no ghrelin production and affect neurobiological processes involved in reward. This was further supported by a recently published study showing that when peripherally circulating endogenous ghrelin is pharmacologically hindered from entering the brain (via the high affinity compound Spiegelmer NOX‐B11‐2), there is no subsequent attenuation of the rewarding properties of alcohol in rodents (Jerlhag et al. 2014). Taken together with previous experiments, the importance of central GHSR‐1a receptors in drug‐induced reward was raised. It is thus possible that the minor allele of the rs2948694 alters the sensitivity of the cholinergic‐dopaminergic reward link, which in turn may lead to a predisposition to smoking and increased alcohol consumption, reflected by the higher AUDIT scores in the present study.

Previous studies show that alcohol and nicotine dependence share a genetic background (True et al. 1999; Funk, Marinelli & Le 2006; von der Pahlen et al. 2008). This is further supported by the well‐documented co‐use of alcohol and nicotine—also observed in the present study—and the neurobiological interaction between alcohol and nicotine within the mesolimbic dopamine system (for review, see Larsson & Engel 2004; Soderpalm, Lof & Ericson 2009). Given the present results, it is possible that variations in GHSR contribute to the shared genetic background seen for these addictions.

The limitations of the current study include the low number of SNPs investigated and the fact that the genetic variation across the whole GHRL and GHSR genes are not taken into account. The best genetic models were recessive for AUDIT scores and dominant for smoking. A possible explanation to this discrepancy may be a threshold effect of the SNP on the various parameters i.e. while heterozygosity is enough to infer an increased risk in smoking, one needs to be homozygote for the risk allele in order to have an increased probability of having higher AUDIT scores. As nicotine is more addictive than alcohol (van Amsterdam et al. 2010), it is not farfetched to suggest that fewer genetic risk factors are needed to become addicted to nicotine. The poor fit with a recessive model for smoking explains why effect sizes for AUDIT are unchanged when controlling for smoking and a similar argument holds when controlling for AUDIT.

In summary, the minor allele of the rs2948694 in the GHSR gene was associated with increased AUDIT scores and smoking possibly reflecting an association with alcohol use disorder and nicotine dependence, respectively. These disorders share common reward mechanisms, involving the mesolimbic dopamine system. The current study suggests that the ghrelin signaling system may be a common denominator and thus a possible target in the development of new treatment strategies for addictive behaviors and disorders.

Table S1 Associations between investigated SNPs, AUDIT scores and smoking.

Click here for additional data file.