Significant, replicable and functional associations between KTN1 variants and alcohol and drug co-dependence

Abstract

Objective:

The gray matter volume (GMV) of the putamen has been reported to be regulated by kinectin 1 gene (KTN1). As a hub of the dopaminergic circuit, the putamen is widely implicated in the etiological processes of substance use disorders (SUD). Here, we aimed to identify robust and reliable associations between KTN1 SNPs and SUD across multiple samples.

Methods:

We examined the associations between SUD and KTN1 SNPs in four independent population-based or family-based samples (n=10,209). The potential regulatory effects of the risk alleles on the putamen GMVs, the effects of alcohol, nicotine, marijuana and cocaine on KTN1 mRNA expression, and the relationship between KTN1 mRNA expression and SUD were explored.

Results:

A total of 23 SNPs were associated with SUD across at least two independent samples (1.4×10⁻⁴≤p≤0.049), including one SNP (rs12895072) across three samples (8.8×10⁻³≤p≤0.049). Four other SNPs were significantly or suggestively associated with SUD only in European-Australians (4.8×10⁻⁴≤p≤0.058). All of the SUD-risk alleles of these 27 SNPs increased (β>0) the putamen GMVs and represented major alleles (f>0.5) in Europeans. Twenty-two SNPs were potentially biologically functional. Alcohol, nicotine and cocaine significantly affected the KTN1 mRNA expression, and the KTN1 mRNA was differentially expressed between nicotine or cocaine dependent and control subjects.

Conclusion:

There was a replicable and robust relationship among the KTN1 variants, KTN1 mRNA expression, putamen GMVs, molecular effects of substances, and SUD, suggesting that some risk KTN1 alleles might increase kinectin 1 expression in the putamen, altering putamen structures and functions, and leading to SUD.

1. Introduction

Alcohol dependence often co-occurs with other substance such as nicotine [1], marijuana [2], and cocaine dependence [3], etc. Alcohol drinking may enhance the motivation to use these other drugs or vice versa by engaging shared brain targets – the dopaminergic circuits – that support the reinforcing effects. The comorbidity may also suggest that alcohol and drug dependence have common genetic basis. For example, alcohol dependence has been reported to share the susceptibility genes CHRNA5-A3-B4 with nicotine dependence and marijuana dependence [4], and the ADH genes with cocaine dependence [3].

Receiving heavy dopaminergic innervations, [5] the putamen represents a target on which alcohol and drugs exert their molecular effects. Alcohol selectively increases dendritic spine density, intrinsic excitability of medium-sized spiny neurons, dopamine release, and glutamatergic transmission, and suppresses GABAeric signaling in the putamen [6,7]. Nicotine [8,9] and cocaine [10] likewise increases dopamine release in the putamen. Putamen gray matter volume (GMV) has been reported to be reduced in alcohol dependent subjects [11], but enlarged in stimulant including cocaine and methamphetamine dependent subjects [12-14].

Putamen GMV is genetically controlled. The most significant gene controlling putamen GMV is kinectin 1 (KTN1). Earlier work demonstrated abundant expression of KTN1 mRNA in the putamen in three independent cohorts [15-17]. So far, four KTN1 SNPs have been reported to significantly regulate putamen GMV, including rs945270 (p=1.1×10⁻³³), rs2181743 (p=4.0×10⁻⁸), rs8017172 (6.7×10⁻³⁴≤p≤3.0×10⁻¹⁴) and rs17253792 (p=3.2×10⁻⁷) [18-21]. KTN1 has also been reported to play important roles in many neuropsychiatric or neurodegenerative diseases/phenotypes, including attention-deficit/hyperactivity disorder (ADHD) [19], Parkinson’s disease (PD) [22-24], heroin dependence [25], marijuana dependence [26], and cognitive dysfunction in the elderly [27]. Specifically, the functional marker rs945270 showed a significant effect on the severity of hyperactivity symptoms of European patients with ADHD and reward processing in the putamen in European girls with ADHD [19]. Two large-scale meta-analyses of genome-wide association studies identified rs8017172 as the most significant risk variant for PD in Europeans [22,23]. A proteomic analysis of the locus ceruleus identified kinectin encoded by KTN1 as a differentially expressed protein between PD patients and controls in Europeans [24]. A population-based study identified rs945270 as an important risk variant for heroin dependence in Han Chinese [25]. A large-scale meta-analysis of genome-wide association studies from the International Cannabis Consortium identified KTN1 as a modest risk factor for marijuana dependence in Europeans [26]. Finally, a genome-wide association study in a Chinese elderly male gout population identified KTN1 as the most significant risk gene for cognitive performance [27]. These neuropsychiatric or neurodegenerative diseases/phenotypes were frequently associated with altered putamen GMVs. For example, reported consistently across numerous independent studies, putamen GMVs were significantly decreased in ADHD and PD patients regardless of medication status. Youths with developmental delays in cognition, developmental dyslexia, or premature responding in a serial reaction time task showed reduced putamen GMVs. On the other hand, individuals dependent on the stimulants (cocaine/methamphetamine) showed higher putamen GMVs, despite diminished GMVs in a wide array of other brain regions, across many studies [28]. A recent study of a very large sample reported that there was no significant difference in the volumes of seven brain regions between nicotine or marijuana dependence and stimulant (methamphetamine) dependence [11].

In the present study, we aimed to identify common genetic factors underlying alcohol and drug co-dependence. Our goal was to identify robust and reliable associations between KTN1 SNPs and substance use disorders (SUD) that could be replicated across multiple independent samples and supported by functional studies. Additionally, although Europeans and Africans have distinct genetic background (e.g., different allele frequencies), they may share common biomarkers for diseases. For example, ADH and APOE genes predispose individuals to alcoholism and Alzheimer’s disease, respectively, in both European and African populations [29,30]. We thus aimed to identify the common risk alleles across these two populations. In sum, we investigated the relationship among KTN1 variants, KTN1 mRNA expression, putamen GMV, and SUD risk, in order to examine the associations between them.

2. Materials and Methods

2.1. Subjects

We examined four independent samples including European-Americans (dbGaP access number: phs000092.v1.p1 and phs000125.v1.p1), European-Australians (phs000181.v1.p1), African-Americans (1^st) (phs000092.v1.p1 and phs000125.v1.p1), and African-Americans (2^nd) (phs000425.v1.p1). The European-American sample included 400 subjects with alcohol and marijuana co-dependence (278 males and 122 females) and 1,481 healthy subjects (441 males and 1,040 females); the European-Australian sample included a total of 6,438 family subjects with 1,645 alcohol and nicotine co-dependence (1,020 males and 625 females); the 1^st African-American sample included 392 subjects with alcohol and cocaine co-dependence (233 males and 159 females) and 475 healthy subjects (171 males and 304 females); and the 2^nd African-American sample included 525 subjects with alcohol and marijuana co-dependence (355 males and 170 females) and 498 healthy subjects (145 males and 353 females).

Subjects from the European-American, European-Australian and 1^st African-American samples were interviewed using the Semi-Structured Assessment for the Genetics of Alcoholism (SSAGA) [31]; and subjects from the 2^nd African-American sample [32] were interviewed using the Semi-Structured Assessment for Drug Dependence and Alcoholism (SSADDA) [33]. Affected subjects met lifetime DSM-IV criteria for alcohol, nicotine, cocaine and/or marijuana dependence [34]. The control subjects were defined as individuals who were exposed to alcohol as well as nicotine, cocaine or marijuana (and possibly to other drugs), but did not have a lifetime diagnosis of SUD. All subjects gave written informed consent to participate in protocols approved by the relevant institutional review boards (IRBs). All subjects were de-identified in this study that was approved by Yale IRB.

2.2. Imputation

The European-American and 1^st African-American samples were genotyped on the Illumina Human 1M microarray (with one million SNPs); the European-Australian sample was genotyped on the Illumina CNV370v1 array (with 370,000 SNPs); and the 2^nd African-American sample was genotyped on the Illumina HumanOmni1_Quad_v1-0_B array (with one million SNPs). To make the genetic marker sets consistent across different samples, we imputed the untyped SNPs across the entire KTN1 region using the same reference panels of 1000 Genome Project and HapMap3 Project data by the program IMPUTE2 [35]. This entire KTN1 region starts from Chr14:54995382 (5’-UTR) and ends with Chr14:55550419 (3’-UTR) (Genome Build 36).

Prior to data analysis, we applied stringent criteria to “clean up” the phenotype and genotype data, as described in detail previously [36]. In brief, subjects with missing diagnosis, missing race, or a missing genotype call rate ≥ 2% across all SNPs were excluded. Furthermore, we excluded SNPs with an overall missing genotype call rate ≥ 2% across all subjects. The SNPs with minor allele frequencies ≤ 0.01 in either affected or unaffected subjects, or in Hardy-Weinberg disequilibrium (p<0.001) in unaffected subjects were excluded.

2.3. Association analysis

The allele frequencies of SNPs were compared between individuals with SUD and controls using the Fisher exact test as implemented in the program PLINK. A p < 0.05 indicates a nominally-significant SNP-SD association. A SNP-SD association with p<0.05 across at least two samples was taken as a replicable association. Replication significantly reduces the false positive rate.

2.4. Bioinformatic analyses

A series of bioinformatic analyses were conducted to predict the potential biological functions of the risk SNPs. Using the bioinformatics software packages including FuncPred [37] and VE!P [38], the relationship of the risk SNPs with long non-coding RNAs (lncRNAs), transcription factor binding sites (TFBS), CCCTC-Binding Factor (CTCF) binding site, and enhancers was explored. The potential regulatory effects of the risk SNPs on the KTN1 mRNA expression in human postmortem putamen in a UK European cohort (n=129) [16] and a European-American cohort (n=111) [15] were analyzed using cis-acting expression quantitative trait locus (cis-eQTL) analysis. Finally, we reviewed the literature for phenotypes associated with these risk SNPs.

2.5. Molecular effects of alcohol, nicotine, marijuana and cocaine on KTN1 mRNA expression

We searched the Gene Expression Omnibus (GEO) database for available KTN1 mRNA expression data that were relevant to the molecular effects of alcohol, nicotine, marijuana or cocaine, or the risk for alcohol, nicotine, marijuana, or cocaine dependence. The first cohort comprised four neural progenitor cell (NPC) lines from differentiated H1 human embryonic stem cell (hESC) lines; half of them were treated with 20mM ethanol (EtOH) and the other half were untreated [39]. The second cohort comprised 11 human subjects; six of them were administered with orange juice with EtOH and five were administered with orange juice without EtOH. The blood alcohol concentration (BAC) of all subjects was monitored by breathalyzer analysis [40]. The third and fourth cohorts respectively comprised six C57BL/6J mice; half of them were intraperitoneally injected with nicotine (1 mg/kg) or cocaine (25 mg/kg), and the other half were intraperitoneally injected with saline. The striatum tissues from these two cohorts were collected 1 hour and 4 hours after injection, respectively [41,42]. The fifth cohort comprised lymphoblast cell lines from 15 human subjects; six subjects were diagnosed as nicotine dependence, and nine had no tobacco abuse [43]. Finally, the sixth cohort comprised midbrain tissues from 20 human subjects, 10 with cocaine dependence and the other 10 subjects without a cocaine use disorder [44].

The KTN1 mRNA expression levels in these cell lines, blood samples or brain tissues from these six cohorts were assessed using microarrays, and then compared between cases and controls by t-test. A p < 0.05 and a p-value close to 0.05 indicated significantly and suggestively different expression, respectively.

2.6. SD-risk alleles and putamen GMVs

After the risk KTN1 alleles for SUD were identified from the afore-described association analyses, the potential regulatory effects of these risk alleles on the putamen GMVs were analyzed in a European sample (n=13,145) [Enhancing Neuro Imaging Genetics through Meta-Analysis (ENIGMA2) consortium – GWAS Meta-Analysis of Subcortical Volumes] [18] using multiple linear regression analysis. These subjects were free of neurodegenerative and neuropsychiatric disorders. The β values, a measure of effect size, and the p values of the regression are reported in Table 1.

Table 1.

Replicable risk KTN1 SNPs significantly associated with both putamen GMV and alcohol and drug co-dependence in five independent samples

SNP	Genomic position at 14q	Risk allele	European-American			European-Australian			African-American (1st)			African-American (2nd)			ENIGMA2 Data		Bioinfomatics	Associated phenotypes reported In literature
			Alcohol and marijuana			Alcohol and nicotine			Alcohol and cocaine			Alcohol and marijuana			Putamen GMV
			Frequency		(n=1,881)	Frequency		(n=6,438)	Frequency		(n=867)	Frequency		(n=1,023)	(n=13,145)
			Case	Control	p	Observed	Expected	p	Case	Control	p	Case	Control	p	β	p
rs946066	55002625	C	0.93	0.89	1.4×10−4	0.89	0.88	0.021	0.75	0.73	-	0.62	0.57	-	3.5	-	--	--
rs11158044	55004557	C	0.85	0.81	2.5×10−3	0.78	0.77	0.036	0.73	0.73	-	0.73	0.72	-	14.1	0.050	--	--
rs56969685	55004865	C	0.85	0.81	2.5×10−3	0.78	0.77	0.036	0.75	0.75	-	0.79	0.79	-	15.6	0.036	--	--
rs2181740	55052059	A	0.61	0.56	3.5×10−3	0.51	0.51	-	0.32	0.35	-	0.11	0.06	0.012	27.0	4.5×10−6	lncRNA	--
rs2341883	55054765	A	0.69	0.63	1.1×10−3	0.59	0.59	-	0.45	0.45	-	0.13	0.06	0.011	17.9	3.8×10−3	lncRNA	prostate cancer
rs10132888	55063703	A	0.61	0.56	0.011	0.53	0.53	-	0.38	0.42	-	0.28	0.21	6.1×10−3	26.0	1.2×10−5	lncRNA	--
rs10148298	55067916	T	0.92	0.88	6.4×10−3	0.89	0.88	0.037	0.84	0.84	-	0.75	0.88	-	31.5	2.7×10−5	lncRNA	GMV, MjD
rs4901578	55069085	G	0.88	0.86	0.028	0.86	0.85	0.032	0.89	0.89	-	0.70	0.70	-	43.2	4.1×10−8	lncRNA, CTCF	GMV
rs4901580	55069173	G	0.88	0.85	0.025	0.85	0.84	0.033	0.86	0.86	-	0.76	0.76	-	43.2	4.1×10−8	lncRNA, CTCF	--
rs11847588	55095399	G	0.86	0.82	0.016	0.83	0.82	0.032	0.84	0.84	-	0.79	0.79	-	40.6	4.3×10−7	lncRNA	MjD
rs2341884	55095553	C	0.86	0.82	0.016	0.83	0.82	0.032	0.84	0.84	-	0.85	0.85	-	40.5	4.6×10−7	lncRNA	--
rs4143891	55095928	G	0.86	0.82	0.016	0.83	0.82	0.032	0.86	0.86	-	0.79	0.79	-	40.5	4.5×10−7	lncRNA	--
rs6573040	55097568	T	0.83	0.80	0.029	0.80	0.79	0.021	0.80	0.80	-	0.82	0.82	-	44.0	5.5×10−9	lncRNA	--
rs12589867	55115552	G	0.79	0.77	0.040	0.79	0.78	0.036	0.93	0.93	-	--	--	-	26.5	7.0×10−5	lncRNA, TFBS	--
rs60233569	55229977	A	0.83	0.78	0.016	0.88	0.87	0.035	0.75	0.75	-	0.88	0.88	-	25.7	9.7×10−5	cis-eQTL	--
rs12895072	55238252	C	0.88	0.84	8.8×10−3	0.84	0.83	0.025	0.97	0.95	0.049	0.99	0.99	-	41.1	3.6×10−7	enhancer	cognitive dysfunction
rs1953354	55255787	C	0.78	0.75	2.6×10−3	0.77	0.75	4.2×10−3	0.70	0.69	-	0.71	0.68	-	45.2	4.2×10−11	--	--
rs8014725	55256706	A	0.58	0.56	-	0.51	0.49	0.033	0.17	0.19	-	0.29	0.24	0.025	55.6	2.2×10−21	cis-eQTL	GMV, PD
rs2342588	55266792	T	0.57	0.55	-	0.50	0.48	0.025	0.76	0.74	-	0.67	0.61	0.046	56.1	1.9×10−21	cis-eQTL	PD
rs2342589	55266821	G	0.57	0.55	-	0.50	0.48	0.025	0.76	0.74	-	0.67	0.61	0.046	56.1	1.9×10−21	cis-eQTL	PD
rs8017172**	55268801	G	0.60	0.58	-	0.56	0.54	0.023	0.20	0.22	-	0.20	0.20	-	60.4	5.7×10−24	cis-eQTL, CTCF	GMV, PD
rs945270**	55270226	C	0.60	0.58	-	0.61	0.59	6.6×10−3	0.16	0.19	-	0.02	0.02	-	48.9	1.1×10−33	cis-eQTL	GMV, PD, ADHD, heroin
rs1959087*	55270950	T	0.69	0.68	-	0.66	0.64	4.8×10−4	0.79	0.79	-	0.88	0.86	-	59.4	1.1×10−20	--	PD
rs17253792**	55274783	T	0.94	0.93	-	0.94	0.93	0.058	--	--	-	--	--	-	52.8	1.9×10−7	--	GMV
rs7153309	55471330	A	0.89	0.86	0.028	0.93	0.92	0.022	0.45	0.47	-	0.24	0.15	-	7.9	-	lncRNA	--
rs12882780	55512345	G	0.97	0.95	0.042	0.98	0.97	3.3×10−3	0.99	0.99	-	0.97	0.97	-	8.6	-	--	MjD
rs1542577	55516009	A	0.91	0.89	-	0.89	0.89	0.044	0.98	0.96	0.032	0.97	0.95	-	14.5	-	--	--

The underline in the first column separates SNPs into two locations: 5’ and 3’ flanking regions. Four non-replicable risk SNPs are listed in Table 1:

^*this SNP is the most significant one in the European-Australians

^**these SNPs were reported to be GMV-associated. -, p>0.05; --, missing. ENIGMA2 data, imaging genomic data for subcortical structure; β, effect size for SNPs to control putamen volume. lincRNA, located in a long interspersed ncRNA sequence; TFBS, located in a transcription factor binding site; enhancer, located in an enhancer; CTCF, located in a transcriptional repressor CTCF binding site; cis-eQTL, this SNP significantly regulates KTN1 mRNA expression in putamen (p<0.05). Previously-reported associated phenotypes: GMV, gray matter volume; MjD, Marijuana dependence; PD, Parkinson’s disease; ADHD, attention deficit hyperactive disorder; heroin, heroin dependence. Detailed demographic information is available in these citations: Gelernter et al. Mol Psychiatry. 2014;19(1):41-49; Bierut et al. PNAS. 2010;107(11):5082-5087; Edenberg et al. ACER. 2010;34(5):840-852; Heath et al. Biol psychiatry. 2011;70(6):513-518; Hibar et al. Nature. 2015;520(7546):224-229.

3. Results

3.1. Replicable associations between KTN1 SNPs and SD across independent samples (Table 1)

A total of 2,955 imputed KTN1 SNPs were analyzed in the four samples, including 223 SNPs in the European-American sample, 91 SNPs in the European-Australian sample, 56 SNPs in the 1^st African-American sample, and 31 SNPs in the 2^nd African-American sample that were nominally (p<0.05) associated with SUD. Among them, a total of 23 SNPs were associated with SUD across at least two independent samples (1.4×10⁻⁴≤p≤0.049), including one SNP (rs12895072) across three samples (8.8×10⁻³≤p≤0.049). Three other SNPs (rs8017172, rs945270, and rs17253792) that were previously reported to be associated with putamen GMV were significantly or suggestively associated with SUD in European-Australians (p=0.023, 6.6×10⁻³, and 0.058, respectively). In this European-Australian sample, the most significant SNP was rs1959087 (p=4.8×10⁻⁴). Table 1 lists these 27 risk SNPs, which are located in two haplotype blocks (D’>0.8; Supplementary Figure S1). The first 14 SNPs and the 15^th-24^th SNPs in the first haplotype block are located in the 5’ and 3’ flanking regions close to KTN1, respectively; and the last three SNPs in the second haplotype block are located in the 3’ flanking region far from KTN1.

The risk alleles of all 23 replicable risk SNPs (with p<0.05), whose frequencies were higher in cases than controls, were identical across four independent samples. The risk alleles of all 27 risk SNPs were major alleles (f>0.5) in Europeans, but some of them were minor alleles (f<0.5) in Africans.

3.2. The risk KTN1 SNPs may be biologically functional (Table 1)

Bioinformatic analysis showed that the variants within the same regions were not only highly linked but also shared similar biological functions. Twelve variants are located in lncRNAs, with eleven in the 5’ flanking region; one variant is located in a transcription factor binding site (TFBS); three variants are located in the CTCF binding sites; and one variant is located in an enhancer. Furthermore, the major alleles (f>0.5) of six variants (rs60233569, rs8014725, rs2342588, rs2342589, rs8017172, and rs945270) significantly increased the KTN1 mRNA expression in the putamen (p=0.030, 0.018, 0.020, 0.020, 0.049 and 0.049, respectively). These significant cis-eQTL signals are all located in the 3’ flanking region.

These risk SNPs have been reported to be associated with some phenotypes in the literature. Six SNPs have been reported to be associated with putamen GMVs, including rs10148298 (p=4.0×10⁻⁸), rs4901578 (p=4.0×10⁻⁸), rs8014725 (p=2.8×10⁻³²), rs8017172 (p=6.1×10⁻³⁴), rs945270 (p=1.1×10⁻³³) and rs17253792 (p=3.2×10⁻⁷) [18,20,21]. Six SNPs have been reported to be associated with PD [22,23], including rs8017172 (p=2.0×10⁻⁷), rs8014725 (p=2.8×10⁻⁵), rs2342588 (p=1.9×10⁻⁵), rs2342589 (p=2.4×10⁻⁵), rs1959087 (p=4.2×10⁻⁵), and rs945270 (p<0.05). Four SNPs have been reported to be associated with marijuana dependence [rs10148298 (p=0.017), rs11847588 (p=0.021) and rs12882780 (p=0.045)] [26] or heroin dependence [rs945270 (p=0.007)] [25]. Other phenotypes included ADHD [rs945270 (p=1.1×10⁻¹²)] [19], cognitive dysfunction [rs12895072 (p=4.2×10⁻⁹) and rs12434554 (p=4.7×10⁻⁹)] [27], and prostate cancer [rs2341883 (p=1.5×10⁻⁴)] [45].

3.3. Alcohol, nicotine and cocaine significantly affected the KTN1 mRNA expression, and the KTN1 mRNA was differentially expressed between nicotine or cocaine dependent and control subjects (Table 2)

Table 2.

Molecular effects of alcohol or drug exposure on the KTN1 mRNA expression in six independent cohorts

	Cohort 1	Cohort 2	Cohort 3	Cohort 4	Cohort 5	Cohort 6
Organism	Human	Human	C57BL/6J mice	C57BL/6J mice	Human	Human
Dataset names	GEO	GEO	GEO	GEO	GEO	GEO
Accession number	GSE56906	GSE20489	GSE15774	GSE15774	GSE6264	GSE54839
References	Kim et al., 2016	Kupfer et al. 2013	Korostynski et al. 2013; Piechota et al. 2010	Piechota et al. 2010; Korostynski et al. 2013	Philibert et al. 2007	Bannon et al. 2014
Experiment methods	Affymetrix Human Genome U133A Array	Affymetrix Human Genome U133A Array	Illumina MouseWG-6 Microarray	Illumina MouseWG-6 Microarray	ABI 1700 Human Genome Expression Microarray	Illumina HT-12 Microarray
Measurement of expression	Log2(normalized intensity)	Log2(normalized intensity)	Log2(normalized intensity)	Log2(normalized intensity)	Log2(normalized intensity)	Log2(normalized intensity)
Control subjects:
Treatment/Phenotype	without EtOH	without EtOH	Saline (1h)	Saline (4h)	No tobaco abuse	No cocaine abuse
Tissue types	NPCs from differentiated H1 hESC	Blood	Striatum	Striatum	Lymphoblast cell	Midbrain
Sample sizes	2	5	3	3	9	10
Expression levels	12.8±0.01	9.3±1.5	6.53±0.05	6.44±0.01	6.6±0.5	10.6±0.5
Case subjects:
Treatment/Phenotype	with 20mM EtOH	alcohol depletion to <0.02% wt/vol	Nicotine (1 mg/kg) (1h after injection)	Cocaine (25 mg/kg) (4h after injection)	Nicotine dependence	Cocaine dependence
Tissue types	NPCs from differentiated H1 hESC	Blood	Striatum	Striatum	Lymphoblast cell	Midbrain
Sample sizes	2	6	3	3	6	10
Expression levels	12.7±0.02	6.4±0.5	6.29±0.13	6.74±0.16	7.2±0.4	10.8±0.4
p-values for t-test	0.015	0.001	0.040	0.034	0.034	0.088

GEO, Gene Expression Omnibus database;EtOH, ethanol; NPC, Neural progenitor cells; hESC, human embryonic stem cells.

The expression levels of KTN1 mRNA in NPCs from H1 hESCs were significantly different between EtOH treatments (20Mm) and non-treatments in Cohort 1 (p=0.015). The expression level of KTN1 mRNA in human blood samples was significantly different between alcohol treatment (depleting from 0.08% to 0.02% wt/vol BAC) and non-treatment in Cohort 2 (p=0.001). The expression levels of KTN1 mRNA in mouse striatum were significantly different between saline and nicotine or cocaine treatment in Cohort 3 (p=0.040) and Cohort 4 (p=0.034), respectively. The expression level of KTN1 mRNA in lymphoblast cell lines was significantly different between nicotine dependence and control in Cohort 5 (p=0.034), and suggestively different in the midbrain between cocaine dependent and control subjects in Cohort 6 (p=0.088). In summary, alcohol decreased KTN1 mRNA expression in humans in both Cohorts 1 (12.8±0.01 in controls vs. 12.7±0.02 in cases) and 2 (9.3±1.5 in controls vs. 6.4±0.5 in cases); cocaine increased KTN1 mRNA expression in mice in Cohort 4 (6.44±0.01 in controls vs. 6.74±0.16 in cases) and cocaine dependence was associated with higher KTN1 mRNA expression in humans in Cohort 6 (10.6±0.5 in controls vs. 10.8±0.4 in cases); and nicotine decreased KTN1 mRNA expression in mice in Cohort 3 (6.53±0.05 in controls vs. 6.29±0.13 in cases) but nicotine dependence was associated with higher KTN1 mRNA expression in human in Cohort 5 (6.6±0.5 in controls vs. 7.2±0.4 in cases).

3.4. The SUD-risk alleles potentially increased the putamen GMVs (Table 1)

The alleles with significantly higher frequencies in the SUD groups than the controls were identified as the risk alleles from the afore-mentioned association analyses. All of the 27 SUD-risk alleles were associated with higher putamen volumes (β>0) in the ENIGMA2 sample, including 23 statistically significant ones (14.1≤β≤60.4; 1.1×10⁻³³≤p≤0.05) and four non-significant ones (3.5≤β≤14.5; p>0.05); and all of them were major alleles in the ENIGMA2 sample (f>0.5).

4. Discussion

We found 23 KTN1 variants that were associated with SUD across at least two independent samples (primarily Europeans), and four other GMV-associated variants that were associated with SUD in one sample (Europeans). Many of these risk variants may be biologically functional. The expression of KTN1 mRNAs in NPCs, embryoid bodies (EBs), blood, and striatum were significantly or suggestively affected by alcohol, nicotine or cocaine in four independent cohorts, and were differentially expressed in the lymphoblast cell lines and midbrain between SUD and controls in two cohorts. Together, these results suggested that KTN1 played a functional role in the development of SUD.

Twelve risk variants located in lncRNAs, including one in TFBS, might regulate KTN1 mRNA expression. One variant located in a CTCF binding site might affect the CTCF's activity and influence gene expression [46]. One variant located in an enhancer might affect the transcription too. Fifteen variants have been reported to play roles in putamen GMVs, PD, marijuana dependence, heroin dependence, ADHD, cognitive dysfunction and/or prostate cancer in the published studies, primarily in European populations. The potential biological functions of SUD-risk SNPs, along with the molecular effects of substances on the KTN1 mRNA expression, the differential KTN1 mRNA expression between SUD cases and controls, and the regulatory effects of KTN1 variants on putamen GMVs, again suggested a functional role of KTN1 in the development of SUD.

As observed in the current study, the major alleles of all 27 risk SNPs significantly increased risk for SUD in Europeans, KTN1 mRNA expression levels in the putamen, and the putamen GMVs. Further, in multiple tissues from six independent cohorts, we found that alcohol (in humans) and nicotine (in mice) decreased KTN1 mRNA expression, but cocaine (in mice) increased KTN1 mRNA expression; and both cocaine and nicotine dependence (in humans) were associated with higher KTN1 mRNA expression, consistent with previous findings [11-14]. Together with the literature, these findings overall suggested a consistent, replicable, robust and positive relationship among the KTN1 variants, KTN1 expression, putamen GMVs, molecular effects of substances and SUD risk. The findings supported the hypothesis that some risk KTN1 alleles might increase kinectin 1 expression in the putamen, altering putamen structure and functions, and lead to SUD.

A highly interesting finding was that the same alleles increased KTN1 mRNA expression in the putamen, putamen GMVs, and risk for SUD across multiple independent samples and studies. A possible mechanism might account for this finding: the major alleles of KTN1 variants increased the KTN1 mRNA and kinectin expression in the putamen; neurons with more kinectin had larger cell bodies [47] that might increase putamen GMVs and alter dopaminergic signaling. On the other hand, the major alleles of KTN1 variants also increased the risk for SUD that was usually linked to dopaminergic dysfunction. In this model, both SUD risk and larger putamen GMVs were the products of KTN1 gene expression, and thus we could postulate an association between larger putamen GMVs and SUD risk.

Psychoactive substances increase synaptic dopamine and lead to euphoria, by stimulating dopamine release from presynaptic neurons (e.g., alcohol) or blocking the presynaptic transporters from recycling dopamine (e.g., cocaine) [12,48,49]. However, chronic use of substances results in a hypodopaminergic state, which through a compensatory process may lead to putaminal hypertrophy. Thus, in this alternative model, the enlargement of putamen GMVs represents a consequence of polysubstance abuse. We demonstrated significant or suggestive molecular effects of substances on KTN1 mRNA expression in the mouse striatum and other tissues or cell lines, potentially supporting this model. However, this evidence should be considered indirect indirect, because it was not derived from human putamen samples. Further, the latter hypothesis should also be considered along with the findings of diminished putamen GMVs in alcohol dependent individuals [11]. Altogether, one is tempted to speculate that the KTN1 variants predispose individuals to higher putamen volumes and SUDs; however, whether the putamen continues to be enlarged (as in stimulant dependence) or not (as in alcohol dependence) after chronic substance use may vary according to the molecular mechanisms underlying the effects of the substance on dopaminergic signaling in the putamen.

A major limitation of this study is that the primary analyses on the associations of SUD, KTN1 DNA variants and putamen GMVs in humans were retrospective; therefore, the findings reveal a statistical but not causal relationship. The analyses of the effects of substance exposure on KTN1 mRNA expression in cell lines or blood tissues were prospective; however, these analyses were not conducted in humans. Therefore, our findings supported an association between putamen GMV and SUD risk, but did not clarify whether larger putamen GMV leads to SUD or substance use contributes to larger putamen GMV. Prospective studies in humans are required to address this important issue in the future. The second limitation is that the postmortem putamen samples of humans with SUD were not available to this study, which would have allowed us to directly examine the associations of SUD, KTN1 variants, KTN1 mRNA expression, and putamen GMVs in the same individuals. Currently, these associations were examined separately in different samples. The third limitation of this study is that some samples with certain substance use were not available to this study. For example, the samples with single substance use were not available to this study, and thus the specific effects of each single substance were unable to be examined. In this study, polysubstance including alcohol and stimulant use was associated with larger putamen GMVs. In the literature, alcohol and stimulant use disorder was associated with reduced and enlarged putamen GMVs, respectively. A sample with alcohol or stimulant use alone may help to evaluate each substance in relation to GMVs, although, in practice, such samples would not be characteristic of substance using populations and very difficult to identify. Further, KTN1 variants were reported to be associated with heroin dependence and marijuana dependence. We investigated marijuana but not heroin dependence in this study, and thus had no way to replicate previous reports on heroin dependence.

Supplementary Material

Figure S1

Supplementary Figure S1. Haplotype blocks of risk KTN1 variants