Functional polymorphisms in the serotonin 1B receptor gene (HTR1B) predict self-reported anger and hostility among young men

Abstract

Objective

To examine associations between haplotypes of the serotonin 1B receptor gene and individual differences in anger and hostility.

Methods

Data were analyzed from a study of 361 university students (47% male). Participants were genotyped at 5 polymorphisms in the HTR1B gene (rs11568817, rs130058, rs6296, rs6297, rs13212041), including promoter and 3′UTR polymorphisms with opposite functional effects on gene expression. Participants reported their emotional states across 30 consecutive days for up to four years. Haplotype pairs were constructed statistically and assigned to a level of HTR1B expression based on the presence of the functional polymorphisms.

Results

Six haplotypes accounted for >97% of chromosomes. Three low expression haplotypes contained the 3′UTR variant (rs13212041 A-allele) that enables a microRNA-mediated reduction in expression. One intermediate expression haplotype contained the 3′UTR A-allele paired with the high-activity promoter. Two high expression haplotypes contained the 3′UTR variant (rs13212041 G-allele) that attenuates microRNA-mediated reduction in expression. Men with low expression haplotypes reported greater anger and hostility than men with one or two high expression haplotypes. Diplotype classification accounted for 8.4% of the variance in men’s anger and hostility, primarily due to the 3′UTR polymorphism (rs13212041), but with some contribution of the functional promoter combination (rs11568817, rs130058). Associations with anger and hostility were not found in women.

Conclusions

These findings extend our understanding of the genetic basis of anger and hostility by showing that newly characterized HTR1B haplotypes, particularly those with rs13212041, which modulates microRNA-mediated regulation of HTR1B expression, may have important implications for aggression-related phenotypes among young men.

Introduction

Individual differences in anger and hostility underlie a variety of adverse or risky health conditions including elevated cardiovascular risk (Miller et al., 1996), metabolic and immune dysfunction (Kiecolt-Glaser et al. 2005; Boyle et al., 2007; Goldbacher & Matthews, 2007; Gouin et al., 2008), intimate partner violence (Norlander & Eckhardt, 2005), maintenance of heavy drinking beyond young adulthood (Costanzo et al. 2007), drug dependence (De Moja & Spielberger, 1997) and suicide (Baud, 2005; McGirr & Turecki, 2007). Because emotional style is partly heritable (Bouchard, 2004; Cates et al., 1993; Miles & Carey, 1997), it is important to identify genes that increase risk for this emotional phenotype.

Animal research on aggression may provide insight into the genetic basis of anger and hostility. Aggression is closely related to, but distinct from, the constructs of anger and hostility (Buss & Perry, 1992; Martin, Watson & Wan, 2000). Whereas aggression refers to observable behaviors of an organism involving approach and attempted harm towards others, anger and hostility refer to internal emotional and cognitive experiences of an organism. Feelings of anger and hostility do not always lead to aggression, but they can precipitate the “hotter,” impulsive form of aggression (Berkowitz, 1983). Aggressive behavior is also used as a proxy for an angry/hostile phenotype in studies of cardiovascular risk among higher order primates (e.g., Manuck, Kaplan, & Clarkson, 1983). Taken together, these links suggest that genes that increase susceptibility to aggression may also increase susceptibility to angry and hostile emotions.

In the animal literature, studies suggest a role of the serotonin 1B receptor (HTR1B) in aggressive behavior (for reviews Clark & Neumaier, 2001; Oliver & van Oorshot, 2005; Popova, 2006). For example, mice lacking serotonin 1B receptors behave more aggressively than their wild-type littermates (Saudou et al., 1994) and drugs that increase the activity of the serotonin 1B receptor have anti-aggressive effects on mice placed in paradigms designed to elicit impulsive aggressive behaviors (Miczek & de Almeida, 2001). Given this evidence, researchers have sought to identify human HTR1B variants that might influence either the activity of the serotonin receptor or its level of expression and in turn, the associated aggressive phenotype.

To date, a select group of HTR1B polymorphisms have been examined in relation to human aggressive phenotypes. The most frequently studied HTR1B variant, rs6296 [a synonomous exonic single nucleotide polymorphism (SNP): G861C], has been associated with aggressive behavior in children (Davidge et al., 2004) and antisocial alcoholism (Lappalainen et al., 1998; cf. Kranzler et al., 2002). However these allelic associations with aggressive phenotypes are not consistently found (New et al., 2001). Inconsistent or null associations have also been reported between rs6296 and other phenotypes related to aggression and impulsivity including suicidality (with the 861C allele overrepresented among individuals with personality disorders in New et al., 2001, but not Huang et al., 1999; Nishiguchi et al., 2001; Rujescu et al., 2003; or Zouk et al., 2007); attention deficit hyperactivity disorder (found in Hawi et al., 2002; Quist et al., 2003, Smoller et al., 2006; but not Ickowicz et al., 2007) and substance abuse (Huang et al., 2003). Although the 861C allele was associated with a 20% lower average number of serotonin 1B receptor binding sites in 96 post-mortem brain samples (Huang et al., 1999), there is no evidence that the G861C SNP itself is functional; rather, the high degree of linkage disequilibrium (LD) among previously studied markers in HTR1B (Cigler et al., 2001; Duan et al., 2003; Proudnikov et al., 2006; Zouk et al., 2007) suggests that this marker may be in LD with other functional HTR1B polymorphisms.

The absence of splicing (HTR1B is encoded by a single exon) or common (>2% frequency) amino acid coding polymorphisms suggests that any functional variants would exist in regulatory regions outside the protein coding region. Indeed, two such functional variants have been described. Duan et al. (2003) provided the first report of functional HTR1B polymorphisms. Using in vitro reporter gene expression assays, these investigators showed that two SNPs in the promoter region, −261G paired with the −161 A allele [rs11568817 (T-261G) and rs130058 (A-161T)], increase the binding of transcription factors in the HTR1B promoter producing a 2.3-fold increase in gene transcription in transfected cells. They noted that these promoter SNPs were in significant LD with the commonly examined G861C marker (rs6296). Two groups have examined these promoter SNPs with respect to human phenotypes. Zouk et al. (2007) investigated the T-261G, A-161T promoter polymorphisms, plus other previously investigated SNPs (G861C, A1180G, and the lesser studied C129T), and showed that only the lower expression promoter region −161T allele predicted higher scores on informant reports of the Buss–Durkee Hostility Inventory in a sample of suicide completers. The −161T allele was overrepresented among suicide completers as well (Zouk et al., 2007). Proudnikov et al. (2006) used a hybrid molecular and statistical haplotyping procedure to examine HTR1B haplotype associations with heroin dependence. They observed that haplotypes containing a G-allele at rs6297 (A1180G), a SNP in the proximal region of the HTR1B 3′UTR (Sanders et al., 2001), which has not been demonstrated to have direct functional effects, were protective for heroin dependence (Proudnikov et al., 2006). Interestingly, the protective effect of the G-allele at rs6297 was unrelated to genotype at either the promoter T-261G/A-161T, C129T or G861C alleles, suggesting that it may be linked to other more distal functional polymorphisms.

Recently, evidence for a second functional regulatory variant was reported by Jensen et al. (2008) who characterized a SNP occurring in the distal 3′UTR of HTRIB messenger RNA (rs13212041; A1997G) that disrupts a binding site for the microRNA, miR-96. MicroRNAs are 20-21 nucleotide ribonucleic acids that regulate gene expression by binding to complementary sites on messenger RNA triggering mRNA degradation and/or inhibition of translation (Ambros, 2004; Chendrimada et al., 2007; Baek et al., 2008; Hutvágner & Zamore, 2002; Kiriakidou et al., 2007; Mathonnet et al., 2007; Selbach et al., 2008; Yekta et al., 2004; Wakiyama et al., 2007). Jensen et al. used a luciferase reporter gene assay to show that the rs13212041 polymorphism modulates miR-96 regulation of gene expression. The A-allele mRNA was repressed by miR-96, while the G-allele attenuated this miR-96 regulatory function. Importantly, introduction of a compensatory change in mir96 rescued repression of the G allele, confirming a direct interaction between mir96 and the regulatory element in HTR1B.

These molecular differences found in Jensen et al. (2008) suggested that A-allele individuals have reduced HTR1B expression, and greater risk of aggressive behaviors, compared to G-allele individuals. Indeed, Jensen et al. (2008) found that university students with two copies of the microRNA-repressed HTR1B A-allele reported a greater history of aggressive behaviors (e.g., starting fires, damaging property, fighting) than individuals heterozygous or homozygous for the G-allele. Genotype differences were not found in non-aggressive anti-social behaviors (e.g., lying). As yet, it is not known whether this 3′UTR microRNA targeting polymorphism (rs13212041) is in LD with other known polymorphisms (i.e., rs11568817, rs130058, rs6296, rs6297) and whether these polymorphisms act synergistically or, potentially, antagonistically to influence risk for other aggression-related emotional phenotypes such as anger and hostility. For example, it is not known whether the 3′UTR (rs13212041) and 5′-promoter functional alleles (rs11568817, rs130058) are inherited separately or together, and if together, what those inherited combinations are. In addition, linkage with the 3′UTR could explain previously found effects. For example, co-inheritance of the 1180G allele of rs6297 and the 3′UTR G-allele of rs13212041 could explain why rs6297 has been linked to other impulsive disorders (Proudnikov et al., 2006) in the absence of both known functional effects of rs6297 and associations of it with currently known functional SNPs.

In this report, we examined the recently characterized 3′UTR functional SNP (rs13212041; A1997G; Jensen et al., 2008), the two functional promoter region SNPs (rs11568817, rs130058; Duan et al., 2003), and two other commonly studied SNPs, rs6296 and rs6297, and their associations with individual differences in anger and hostility. Because these markers span a relatively short genetic interval of 2,258 base pairs, they are expected to show considerable LD. Thus, we also examined the pattern of LD among markers and identified naturally occurring haplotypes. With this approach, we could determine whether the 5′-promoter and 3′UTR functional alleles are inherited together or separately, as well as their linkage with the G861C and A1180G alleles, which others have examined in relation to human phenotypes. We then classified the haplotypes expected to have lower versus higher levels of gene expression based on prior molecular research showing functional effects of rs11568817, rs130058, and rs13212041. Finally, we tested whether individuals with the predicted lower expression HTR1B diplotypes scored higher on their reports of anger and hostility than those with predicted higher HTR1B expression diplotypes.

Self-reports of anger and hostility were measured using real-time data capture techniques (Stone, Shiffman, Atienza & Nebeling, 2007). Participants completed a daily report of current feelings of anger and hostility for 30 consecutive days, for up to four years. By aggregating multiple emotion state reports over time, we could obtain a highly reliable and stable assessment of anger and hostility over a four-year period of early-adulthood, roughly from ages 18 to 21. This approach also reduced error in the emotion measurement, thereby increasing the statistical power to detect genotype-phenotype associations.

The present study uses the same human sample as in Jensen et al. (2008) but it extends the prior report in three critical ways. First, the present investigation helps to integrate prior literature by genotyping the four most commonly occurring polymorphisms in addition to the one in the 3′UTR and testing their LD. Second, the present investigation uses a haplotype approach to examine the natural pattern of inheritance for these five markers. A haplotype approach provides a biological point of reference for the inheritance of the potentially synergistic or opposing effects of these polymorphisms. Third, the present investigation uses a more extensive phenotype measure. Whereas Jensen et al. (2008) analyzed a single checklist of past aggressive behaviors, the present investigation analyzed real-time reports of anger and hostility measured in a robust and ecologically valid fashion over four years. Moreover, the content of the questions—asking about common everyday emotions (anger and hostility) rather than relatively uncommon aggressive events (setting fires; starting fights; as in Jensen et al.)— may further increase power to detect genetic associations via more refined differentiation among subjects.

Method

Participants

The sample consisted of 361 university students (168 men; 193 women) from a mid-sized US state university. It was derived from a sample of 575 participants in a four-year longitudinal study who provided DNA samples (N = 416). To avoid population stratification, we limited the analysis sample to individuals of non-Hispanic Caucasian ancestry. This sample included individuals analyzed elsewhere for unrelated hypotheses (Armeli et al., 2008; Covault et al., 2007; Gacek et al., 2008; Gunthert et al., 2007). The average age of sample participants at the start of this four-year study was 18.7 +/ 0.9 yrs with 61% percent freshmen, 32% sophomores, and 7% upper classmen. All participants gave written informed consent under guidelines approved by the Institutional Review Boards at the University of Connecticut (Storrs) and the University of Connecticut Health Center (Farmington).

Measures and Procedure

In each year, participants completed an initial assessment battery one month following the start of either the fall (59% of participants) or spring (41% of participants) semester by accessing a secure website. This on-line assessment included demographic variables and the history of aggression questionnaire reported on previously (Jensen et al., 2008). Approximately two weeks after the initial assessment, participants logged on to a secure website to complete a 5-minute survey between 2:30 and 7:00 p.m. daily for 30 days. The daily survey contained questions about a range of behaviors and experiences, including current emotional states. Using a 5-point scale from 1 (very slightly/not at all) to 5 (extremely), participants rated how “angry” and “hostile” they felt right now, and these responses were averaged within each daily survey for a measure of hostility (α = .76). Participants also rated how “sad” and “dejected” they felt, averaged for sadness (α = .75); “jittery” and “nervous” for anxiety (α = .72); and “cheerful” and “happy” for happiness (α = .88). Participants repeated this same 30-day survey procedure every fall or spring for up to four years.

We initiated DNA collection in year two of the study. At that time, the 535 participants who were thought to be still living in the area were invited to participate in a genetic sub-study, with eligibility to continue in the larger study not contingent on providing a sample for DNA. Of those invited, 416 (77.8 %) consented to provide genetic data, 69 (12.9 %) declined, 34 (6.4%) did not respond and 16 (3.0%) had moved out of the area. After giving informed consent, participants were asked to “swish” Scope™ mouthwash (Procter and Gamble, Cincinnati, Ohio) for 20-30 seconds, and then spit the mouthwash into a collection tube. The 416 participants who consented to provide DNA samples did not differ from the other 119 participants on any of the primary outcome variables, although consenters were more likely to be female (χ² = 7.74, p < .01), younger (18.7 vs. 19.0 years, p < .05), and to participate in the study longer (3.5 vs. 2.3 years, p < .001).

Genotyping

DNA was extracted from mouthwash-stabilized samples using a commercial DNA isolation protocol (PureGene, Gentra Systems, Minneapolis, MN). The five HTR1B polymorphisms were genotyped using PCR-based TaqMan® allelic discrimination assays. A 10μl PCR reaction was prepared in 1x ABI TaqMan Universal master mix (Applied Biosystems Inc., Foster City, CA) containing 100nM of each Fam and Vic-labeled allele specific TaqMan® MGB probes, and 600nM of each forward and reverse primer (Table I). For the promoter SNPs, rs115688177 and rs130058, primers and probes were as described by Proudnikov et al. (2006). MGB-probes and primers for the remaining 3 SNPs were designed using Primer Express v3.0 software (Applied Biosystems Inc.). The polymerase chain reaction was performed with the following cycling parameters: 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds and 60°C for 60 seconds. End point FAM and VIC fluorescence levels were captured using an ABI 7500 Sequence Detection System (Applied Biosystems Inc.) and genotype calls were made based on the level of fluorescence signal. The genotyping assay was repeated for 13% of samples with no discrepancies in genotype calls. Finally, we carried out molecular 2-SNP analysis of the T-261G and A-161T polymorphisms using the A-161T TaqMan probe and downstream primer paired with T-261G allele-specific forward primers (CCAGCTCTTAGCAACCCAGTT or CCAGCTCTTAGCAACCCAGTG) as described by Proudnikov et al. (2006) with genotype calls made by examining change in fluorescence as a function of cycle number for each sample. The resulting molecular 2-SNP allele calls were in agreement with haplotypes statistically determined using Phase 2.1. The 2-SNP haplotype – 261T/-161T was not observed. (Table I Here)

Table I.

HTR1B SNPs Examined in this University Student Cohort

NCBI SNP (dbSNP build 126)	Location relative to HTR1B1	PCR amplicon primers	TaqMan MGB probes (SNP position in bold)
rs115688177	Promoter	AGCGGAGCAGGGCTCG	Vic- CCCAGGGCTAAGAC
	T-261G	TTCCTCAATTATTCCTCCGCC	6FAM- CCCAGGTCTAAGACC
rs13058	Promoter	GCGGCTTAGCTAGGCGCT	Vic- CACCCTTGACCTCTA
	A-161T	CGCAGGTTTGTCCCCAGTT	6FAM- CACCCATGACCTCT
rs6296	Exon 1	CGTCCTCGGTCACCTCTATTAACT	Vic- CTCCTGTGTATGTGAAC
	G861C	GTCGGAGACTCGCACTTTGACT	6FAM- TCTCCTGTGTATGTCAAC
rs6297	Exon 1	AAACAAGCATTCCATAAACTGATACG	Vic- ACTTGCCATTTGC
	A1180G	GGTCCCCAAAGGTCGCTTAG	6FAM- TGACTTGCCGTTTGCAGT
rs13212041	3′UTR	AAAGTGACAGGTACATGAAATTAAGAGAA	Vic- TGCAGACTTTGGC
	A1997G	CACAACTCTAACAAACAAACCATTATGTG	6FAM- TGCAGACTTCGGC

¹Nucleotide position relative to HTR1B ATG using the March 2006 version of the human genome chromosome 6 annotation (www.genome.ucsc.edu).

Data analysis

The survey data were screened prior to analyses to exclude participants who provided fewer than the minimum of 15 of 30 daily surveys for a given year following our standard protocol (Covault et al., 2007). Five of the 361 participants failed to meet this minimum standard in any of the years and were eliminated from all survey analyses. We also limited our analyses to years in which participants were undergraduate students to ensure that data were from a similar population. This screening did not compromise the sample, as the majority of participants were undergraduate students throughout the study. Thus, we retained data across an average of 3.1 years (+/− 0.9), reflecting an average of 79 reports per person (+/−27) (range 16 – 120 reports). The Hierarchical Linear Modeling statistical software program (HLM; 6.04 (Raudenbush, Bryk, & Congdon, 2007) was used to examine associations between the survey data and the genetic information. This approach took into account the unbalanced (e.g., varying number of daily and yearly reports) and nested nature of the data. We ran hierarchical linear models to test (i) whether average hostility scores differed across each genotype group; (ii) whether hostility scores differed across the common haplotype configurations; and (iii) whether hostility scores differed across diplotypes. Men and women were analyzed separately to provide separate effect size estimates. More detailed descriptions of the hierarchical linear models are found in Supplementary Materials.

Linkage disequilibrium (LD) plots, haplotype blocks, and haplotype frequencies for the sample of Caucasian students were generated using the software program Haploview v3.2.2 (Barrett et al., 2005). Best-estimate haplotype pairs for each individual subject were generated using Phase v2.1 software, which incorporates a Bayesian statistical method for reconstructing estimated haplotypes from population data (Stephens et al., 2001; Stephens & Donnelly, 2003).

Results

Genotype frequencies for the 361 Caucasian participants are presented in Table II. All frequencies were in Hardy-Weinberg Equilibrium. Allele frequencies for the 5 SNPs were consistent with prior reports in the literature.

Table II.

Marker Frequency Data for Caucasian University Student Participants

Marker	Position relative to AUG	Minor Allele	Minor Allele Frequency	Genotype Counts			Total N*	HWE (χ2, p)
rs115688177	−261	G	.469	TT (95)	TG (191)	GG (73)	(359)	1.66, 0.20
rs130058	−161	T	.306	AA (165)	AT (168)	TT (26)	(359)	3.66, 0.06
rs6296	861	C	.234	GG (206)	GC (138)	CC (15)	(359)	1.88, 0.17
rs6297	1180	G	.142	AA (261)	AG (89)	GG (6)	(356)	0.26, 0.61
rs13212041	1997	G	.199	AA (232)	AG (114)	GG (15)	(361)	0.04, 0.83

Note. Frequencies and Hardy-Weinberg Equilibrium statistics (HWE) computed for the 361 Caucasian participants.

^*Varying N for each marker reflects differences in the % of individuals for whom genotypes were obtained in this Caucasian DNA sample.

Genotype Results

Genotype differences in hostility for men and women are shown in Table III. Only men showed significant differences in hostility as a function of genotypes. Of the five genotypes tested, the 3′UTR variant (rs13212041) showed the largest differences in hostility scores among the men, with the AA vs. G carrier difference accounting for 7.1% of the variance in hostility scores. Rs6297 also showed significant differences, with the AA vs. G carrier difference accounting for 4% of the variance. Analysis of rs130058 and rs6296 both yielded non-significant trends, and each accounted for approximately 2% of the variance in hostility ratings. Rs115688177 was unrelated to hostility. Genotypes were also related to sadness and anxiety for men only, although effect sizes were generally smaller for these measures than for hostility (see Supplementary Materials, Table S-1). For example, the 3′UTR variant (rs13212041) accounted for 4.0% and 4.9% of the variance in anxiety and sadness, compared to 7.1% of the variance in hostility. Genotypes were unrelated to happiness either for men or women.

Table III.

HTR1B Genotype Differences in Hostility for Caucasian Young Men and Women

	Genotypes				Significance Tests			Effect Size
Men				Coding	b	SE	p	% var
rs115688177	TT 1.31 (.31)	TG 1.35 (.46)	GG 1.39 (.36)	0,1,2	.037	.039	.335	0.5%
rs130058	AA 1.28 (.28)	AT 1.41 (.49)	TT 1.37 (.35)	0,1,1	.113	.059	.057	2.2%
rs6296	GG 1.30 (.36)	GC 1.40 (.46)	CC 1.49 (.49)	0,1,1	.114	.066	.084	2.5%
rs6297	AA 1.39 (.43)	AG 1.25 (.29)	GG 1.05 (.04)	0,1,1	−.158	.057	.007	4.0%
rs13212041	AA 1.42 (.46)	AG 1.23 (.25)	GG 1.22 (.31)	0,1,1	−.198	.054	.001	7.1%
Women				Coding	b	SE	p	% var
rs115688177	TT 1.32 (0.34)	TG 1.24 (.23)	GG 1.26 (0.36)	0,1,2	−.034	.036	.357	1.0%
rs130058	AA 1.28 (0.27)	AT 1.26 (0.32)	TT 1.18 (0.18)	0,1,1	−.025	.042	.553	0.2%
rs6296	GG 1.24 (0.28)	GC 1.29 (0.30)	CC 1.35 (0.28)	0,1,1	.058	.043	.182	1.3%
rs6297	AA 1.26 (0.29)	AG 1.28 (0.31)	GG 1.24 (0.12)	0,1,1	.017	.049	.732	0.1%
rs13212041	AA 1.25 (0.27)	AG 1.31 (0.33)	GG 1.22 (0.21)	0,1,1	.049	.047	.292	1.0%

Note. Hostility scores reflect the averaged self-reports of anger and hostility rated on a 1 – 5 scale with higher numbers indicating greater intensity of emotion. Means (standard deviations) computed on raw data by averaging the hostility scores over 120 (max) days of reporting and then averaging across participants with identical genotypes. Significance tests and effect sizes used Hierarchical Linear Modeling. b = unstandardized coefficient testing for genotype differences using the corresponding coding scheme (linear = 0,1,2; dominant = 0,1,1); SE = standard error of thecoefficient; p = chance probability value (two-tailed). Effect Size, % var = the percentage of variance in hostility accounted for by genotype.

Haplotype Results

The 5 markers examined were in a single haplotype block with D”values of 0.87 to 1.0 (see Figure 1). Six haplotypes comprised >97% of chromosomes (See Table IV). The frequency of haplotypes as estimated for the entire sample by Haploview was in close agreement with the frequency based on individual subject best-estimate haplotype pairs generated using the Phase software application. The observed haplotype frequencies were comparable with frequencies reported using a subset of these markers (Duan et al., 2003; Proudnikov et al., 2006).

Figure 1.

/I> Haploview generated LD plot illustration for Caucasian participants. Values shown represent 100 × ∣D”∣, empty boxes represent a value of 100 (i.e. D”=1). The location of 5 SNPs across the 2,258 bp HTR1B gene are schematically noted at the top beginning with the T–261G promoter region SNP rs115688177 at the far left and the 3′UTR microRNA site SNP rs13212041 at mRNA position 1,997 at the far right.

Table IV.

HTR1B Haplotype Frequencies for Caucasian Samples and Anticipated HTR1B Expression Levels for the Six Most Frequent Haplotypes

Haplotype1	HTR1B mRNA Activity2	Haploview v3.2.2 Frequency	Phase v2.1 Best- Estimate Frequency	Haplotype Counts	Frequency Observed by Duan et al.3	Frequency Observed by Proudnikov et al.4
GTGAA	lower	0.295	0.302	218	0.225 (GTGxx)	0.345 (GTGAx)
TACAA	lower	0.215	0.231	167	0.240 (TACxx)	0.282 (TACAx)
GAGAA	higher	0.160	0.162	117	0.214 (GAGxx)	0.106 (GAGAx)
TAGGG	higher	0.137	0.139	100	0.225 (TAGxx)	0.141 (TAGGx)
TAGAA	lower	0.112	0.098	71	-	0.127 (TAGAx)
TAGAG	higher	0.054	0.055	40	-	-
Total		0.973	0.987	713	-	-

¹Note5 SNP haplotype using markers rs11568817, rs130058, rs6296, rs6297, rs13212041 at positions −261, −161, 861, 1180, and 1999, respectively. Haplotypes shown are those with frequencies >0.02. SNPs associated with higher gene expression activity in vitro are underlined.

²Predicted higher mRNA activity haplotypes included either the promoter −261G/−161A, or 3′UTR microRNA site 1997G.

³Duan et al. did not examine markers rs6297 or rs13212041.

⁴Proudnikov et al. did not examine marker rs13212041.

For the 6 common haplotypes, we classified three as having lower HTR1B gene expression (GTGAA, TACAA, and TAGAA) and three haplotypes as having higher HTR1B gene expression (GAGAA, TAGGG, and TAGAG). We based this classification on in vitro functional assays by Duan et al. (2003) who showed that the GA pairing at T-261G and A-161T had a 2.3-fold higher transcriptional activity and Jensen et al. (2008) who showed that the rs13212041 G-allele attenuated microRNA-mediated down regulation of gene expression relative to the A-allele. This classification combined the putative transcriptional and microRNA-mediated mRNA translation/stability effects contributed by the two separate functional polymorphisms.

With these haplotypes (see Table IV), it is interesting to note that the 861C allele, which had previously been reported to be associated with reduced HTR1B binding sites in post-mortem brain (Huang et al., 1999), was observed only in the lower expression TACAA haplotype containing the rs13212041 A-allele. Similarly, the −161T allele, which had previously been associated with increased hostility (Zouk et al., 2007), was observed only in the lower expression GTGAA haplotype containing the rs13212041 A-allele. Lastly, the rs6297 1180G allele, which had previously been shown to have a protective effect for heroin dependence (Proudnikov et al. 2006), was observed only in the higher expression TAGGG haplotype containing the rs13212041 G-allele.

We then examined the relation between the six common haplotypes and the hostility ratings from the daily survey data (see Table V). As predicted, but for men only, haplotypes expected to have lower HTR1B mRNA expression (TACAA, GTGAA, TAGAA) were associated with higher levels of self-reported hostility, whereas haplotypes expected to yield high HTR1B levels (TAGAG, TAGGG) were associated with lower levels of hostility. The GAGAA haplotype containing the high-activity GA promoter with the low-activity 3′UTR A-allele was associated with intermediate levels of hostility. It is notable that we did not observe the haplotype GAxxG containing the more inducible promoter (GA at T–261G/A–161T) paired with the microRNA binding site disrupted by the G-allele. Instead, we observed only the haplotype GAGAA, containing the more inducible promoter paired with the active microRNA binding site (A-allele). This pattern of inheritance coupled with the intermediate score for the hostile mood phenotype measure suggests that the functional increase in activity of the GA promoter may be ameliorated by microRNA-mediated down regulation of gene expression by the rs13212041 A-allele. As such, the haplotype GAGAA was classified as having intermediate HTR1B expression for consideration of diplotype associations with phenotype.

Table V.

The Six Frequent Haplotypes and Hostility Ratings for Men and Women

Haplotypes	Hostility Mean(SD)	Hostility Ranking	Predicted mRNA activity1	N Haplotypes	N Data2	Post hoc Reference - Highest Ranking Hostility Haplotype
Men
TACAA	1.41(0.51)	1	lower	76	228	-
GTGAA	1.38(0.50)	2	lower	105	317	ns
TAGAA	1.33(0.40)	3	lower	34	108	ns
GAGAA	1.31(0.40)	4	higher	53	151	p < .10
TAGAG	1.23(0.23)	5	higher	20	61	p < .001
TAGGG	1.20(0.29)	6	higher	45	125	p < .05
Total				333	990
Women
TAGAG	1.41(0.51)	1	higher	20	71	-
TACAA	1.31(0.35)	2	lower	91	278	ns
TAGGG	1.26(0.30)	3	higher	55	169	ns
GAGAA	1.25(0.32)	4	higher	64	202	ns
GTGAA	1.23(0.29)	5	lower	113	364	ns
TAGAA	1.22(0.24)	6	lower	37	112	ns
Total				380	1196

¹Note. Predicted higher activity mRNA haplotypes included either the promoter −261G/−161A, or 3′UTR microRNA site 1997G (underlined).

²N Data = the number of data points analyzed for each haplotype (each participant contributed up to four hostility means- one per year of data collection- for each haplotype). Means (Standard Deviations) computed on raw data. Statistical comparisons used Hierarchical Linear Modeling.

Diplotype Results

We computed a diplotype score for each participant based on the predicted HTR1B activity value of both haplotypes. Each haplotype was assigned a value: “0” for lower-activity haplotypes (TACAA, GTGAA and TAGAA), “0.5′for the intermediate-activity haplotype (GAGAA), and “1′for higher-activity haplotypes (TAGAG and TAGGG). The two haplotype values were added, yielding five HTR1B diplotype activity levels (0, 0.5, 1, 1.5 or 2). Frequencies were as follows: 0 = 141 (40%); 0.5 = 81 (23%), 1 = 99 (28%); 1.5 = 21 (6%); 2 = 13 (3%). Six participants could not be assigned a diplotype due to missing haplotype information for one chromosome. The majority (92 of 99) of those with a diplotype score of 1 had one high and one low-activity haplotype; the remainder (7 of 99) had two intermediate activity GAGAA haplotypes containing both the −261G/A-161 promoter alleles and 3′UTR miR-96 binding site A-allele. Because of the low frequency of participants with two intermediate activity haplotypes (2 men; 5 women), we grouped them with the other 92 participants with a diplotype score of 1 in subsequent statistical analyses. Diplotype frequencies did not vary by gender, χ² = .70, p = .951 [men: 0 = 67 (40%); 0.5 = 37 (22%), 1 = 46 (28%), 1.5 = 11 (7%), 2 = 5 (3%), plus two unassigned; women 0 = 74 (39%); 0.5 = 44 (23%); 1 = 53 (28%), 1.5 = 10 (5%); 2 = 8 (4%), plus four unassigned].

Mean hostility and other emotion scores as a function of diplotypes are shown in Table VI. The pattern of means showed that men (top rows) with no high-activity haplotypes (0s) reported the highest levels of hostility, compared to men with one or more higher-activity haplotypes (1s, 1.5s, 2s). Men with one intermediate activity haplotype containing the GAxxA sequence (0.5) reported intermediate levels of hostility. The similar hostility scores among men with one or more high activity haplotypes (1s, 1.5s, 2s), suggested a dominant effect of higher-activity haplotypes containing the G-allele of rs13212041. ¹ In fact, the presence of one or more higher-activity haplotypes (vs. one intermediate vs. none) accounted for 8.4% of the variance in hostility corresponding to a Cohen”s d of 0.62, a medium effect size (Cohen, 1988). This effect size is stronger, although not substantively so, than the effect size found when testing rs13212041 alone (7.1%; d = .57).

Table VI.

Average Hostility, Anxiety, Sadness, and Happiness Scores for the HTR1B Receptor Diplotypes

	Number of Higher Activity HTR1B Haplotypes					Significance Tests			Effect Size
Men	0	.5	1	1.5	2	b	SE	p	% var
Hostility	1.47 (0.51)	1.31 (0.31)	1.25 (0.28)	1.19 (0.18)	1.22 (0.31)	−.235	.069	.001	8.4%
Anxiety	1.61 (0.52)	1.38 (0.41)	1.42 (0.36)	1.25 (0.14)	1.30 (0.24)	−.223	.075	.004	5.8%
Sadness	1.47 (0.51)	1.30 (0.33)	1.27 (0.26)	1.19 (0.12)	1.29 (0.29)	−.206	.068	.003	6.3%
Happiness	2.61 (0.64)	2.76 (0.65)	2.66 (0.79)	2.60 (0.49)	2.76 (1.11)	.049	.124	.697	0.1%
	n = 67	n = 35	n = 45	n = 11	n = 5
Women	0	.5	1	1.5	2	b	SE	p	% var
Hostility	1.23 (0.22)	1.27 (0.33)	1.32 (0.36)	1.24 (0.22)	1.22 (0.21)	.070	.047	.135	1.4%
Anxiety	1.44 (0.33)	1.47 (0.39)	1.61 (0.55)	1.40 (0.24)	1.74 (0.44)	.148	.071	.038	2.9%
Sadness	1.33 (0.27)	1.37 (0.43)	1.40 (0.39)	1.25 (0.29)	1.50 (0.40)	.069	.056	.225	1.0%
Happiness	2.74 (0.72)	2.81 (0.63)	2.74 (0.61)	3.08 (0.65)	2.39 (0.80)	.007	.113	.949	0.0%
	n = 73	n = 43	n = 53	n = 10	n = 8

Note. Means (standard deviations) computed on raw data by averaging the emotion reports over 120 (max) days of reporting and then averaging across participants with identical diplotypes. Emotion reports were made on a 1 - 5 scale. Ns for diplotype groups were reduced by three men and two women due to insufficient survey data. Significance tests and effect sizes used Hierarchical Linear Modeling. b = unstandardized coefficient testing a three-group linear model (0s vs. 0.5s vs. 1 – 2s, recoded 0,0.5,1). SE = standard error of the genotype coefficient; p = chance probability value (two-tailed). Effect Size, % var = the percentage of variance in the outcome accounted for by genotype.

Men with no higher expression chromosomes also reported significantly more anxiety and sadness (see Table IV); however effect sizes were smaller than those observed for hostility (5.8% and 6.3% of the variance, ds = .50 and .52, respectively). Furthermore, when controlling for hostility², the diplotype associations were substantially attenuated for anxiety and sadness (Anxiety % var = 4.0, d = .42; b(SE) = −.149(.060), p = .015; Sadness % var = 2.0, d = .32; b(SE) = −.072(.039), p = .067). By contrast, the diplotype association with hostility was minimally attenuated when controlling for anxiety (% var 7.9, d = .61; b(SE) = −.184 (.056), p = .002), sadness (% var 7.6, d = .60; b(SE) = −.121(.040), p = .003), or both sadness and anxiety together (% var 7.1%, d = .57; b(SE) = −.106(.037), p = .005). These patterns suggest that diplotype was associated most strongly with hostility, which partially accounted for the associations with anxiety and sadness, although not completely. Diplotype associations were not observed in reports of happiness for men.

Women showed no significant differences in daily self-reported hostility, sadness or happiness as a function of diplotype (see Table VI). Women with higher expression chromosomes did have higher anxiety than their peers. This finding was unexpected and warrants future replication.

Additional Moderators

We conducted exploratory analyses to examine whether diplotype effects varied with age (See Supplementary Analyses). Indeed, diplotype differences in hostility were strongest among 18-year-old men (b(SE) = −.314(.077), p < .001), followed by 19-year-old men (b(SE) = −.272(.069), p < .001), 20-year-old men (b (SE) = −.230 (.069), p = .001) and 21-year-old men (b(SE) = −.188(.075), p = .014). This pattern suggests that HTR1B genetic associations with hostility are stronger in younger men.

Discussion

These findings extend our understanding of the genetic basis of anger and hostility and suggest that common functional haplotypes in the serotonin 1B receptor gene are risk factors in the experience of hostile emotional states. Male university students with two low-activity HTR1B chromosomes reported feeling significantly more anger and hostility in their daily lives than their peers with at least one high-activity HTR1B chromosome. Male students with one low and one intermediate HTR1B activity chromosome reported intermediate levels of hostility. This three-group diplotype configuration accounted for 8.4% of the variance in men”s reports of hostility (Cohen”s d = .62)—a medium effect size that is substantially higher than many other single-gene association studies of a behavioral phenotype, which more typically account for 2– 4% of variance in phenotype (e.g., Covault et al., 2007; Lesch et al., 1996).

Much of the variance accounted for by the diplotype was attributable to the microRNA binding site polymorphism (rs13212041), which explained 7.1% of the variance in men”s reports of hostility (d = .57). Indeed, the G-allele of rs13212041 appeared to drive the dominant protective effect of high-activity diplotypes on men”s hostility. This pattern suggests that the microRNA binding site polymorphism has greater behavioral effects than the promoter variant. Yet, this microRNA site polymorphism alone did not completely account for HTR1B associated phenotypic differences. The addition of haplotype information, namely the inclusion of the GA-promoter to distinguish intermediate diplotypes, explained an additional 1.3% of the variance. Consequently, we believe that the GA-promoter information is useful to include, particularly given the frequency of the intermediate haplotype (GAGAA, which represented 16% of haplotypes).

The results of our haplotype examination inform prior studies examining other HTR1B markers linked to the two functional alleles examined here. The rs6296 861C allele previously associated with 20% fewer HTR1B binding sites (Huang et al., 1999) was only observed in the haplotype containing a low activity promoter together with the microRNA repressed rs13212041 A-allele (TACAA). In a study of heroin addiction (Proudnikov et al., 2006), the protective HTR1B rs6297 allele(1180G) was observed only in association with the inactive microRNA binding site 1997G allele (TAGGG). Lastly, the –161T allele at rs130058 reported to be associated with higher scores on the Buss–Durkee Hostility Inventory and overrepresented among suicide completers (Zouk et al., 2007) was only carried on a haplotype containing the microRNA-sensitive rs13212041 A-allele (GTGAA). In each case, the association of other SNPs with increased psychopathology or decreased HTR1B protein expression could be explained by the presence of the microRNA active A-allele (leading to less HTR1B activity), while the protective effect of the 1180G allele could be explained by the presence of the microRNA inactive G-allele (leading to more HTR1B activity).

We also found an association between HTR1B diplotype and both anxiety and sadness. Although not predicted, this association is consistent with several lines of research. First, both anxiety and depression have been consistently linked to hostility (Kennedy, Morris, Pedley, & Schwab, 2001; Swan, Carmelli & Roseman, 1989; Suls & Bunde, 2005). Second, the pharmacological treatment literature reveals that anxiolytics and serotonergic antidepressants are effective not only in the treatment of anxiety and depression, respectively, but also in the treatment of aggressive behaviors (reviewed by Lavine, 1997). The psychological treatment literature also demonstrates that interventions directed toward anger reduction also reduce anxiety and depression (Deffenbacher, Dahlen, Lynch, Morris & Gowensmith, 2000). That the associations to anxiety and depression emerged among men is consistent with the notion that men are vulnerable to masked or hidden depression in the form of aggressive behavior (Fischer & Good, 1997; see also Hankin & Abramson, 2001) and that gender norms lead some depressed men to exhibit externalizing symptoms (but see Addis, 2008 for a critique).

The present findings have several implications. They highlight the importance of a comprehensive molecular genetic approach linking functional variants in candidate genes to complex phenotypes. In this way, our findings build on previous research by Duan et al. (2003) and Jensen et al. (2008) suggesting that there are multiple functional HTR1B polymorphisms and several common haplotypes in the population. Our results also demonstrate an important role for microRNA directed repression of HTR1B in the modulation of hostile emotions. The A-allele at rs13212041 appears to predispose individuals to a more hostile emotional style via microRNA-directed repression of the HTR1B mRNA, consistent with our previous study (Jensen et al., 2008). By contrast, the G-allele appears to predispose young men to a less hostile emotional style by the blocking microRNA-directed repression. Whereas two A-alleles are necessary for increased hostility risk, one G-allele is enough to reduce such risk. We also found that there is linkage between the GA promoter of Duan et al. and the microRNA polymorphism of Jensen et al. Interestingly, the −261G/A-161 promoter variant does not appear to act together with the microRNA G-allele to create a “super high expressing” chromosome. Instead, the GA promoter occurred only with the low expression microRNA A-allele (GAGAA). These findings suggest that when inherited together, the high-activity GA promoter and low-activity A-allele temper one another, resulting in an intermediate HTR1B behavioral phenotype. Additional research is necessary to validate these patterns and demonstrate their presence in other populations. Findings may also implicate these functional haplotypes in other health and psychological conditions linked to hostility. Hostility has been independently related to increased cardiovascular reactivity (e.g., Räikkönen et al., 1999), impaired immune function (e.g., Kiecolt-Glaser et al., 2005), substance dependence (e.g., Costanzo et al., 2007; De Moja & Spielberger, 1997) and suicide (e.g., Baud, 2005; McGirr & Turecki, 2007). Genetic research in these areas may profit from investigating the 3”UTR microRNA binding site variant (rs13212041) and associated HTR1B haplotypes.

It is interesting that in our four-year longitudinal study, the strongest genetic differences for males occurred at age 18, and decreased somewhat thereafter. This pattern suggests the existence of developmental effects; however, given the small age range in this sample (18 – 21 years), we could examine these age effects only in an exploratory fashion. Future research needs to examine age effects in a sample with a larger age range (e.g., 18 – 50+). If age-related effects were found, it might suggest that phenotypic manifestation of these genetic variants depend on developmental factors, i.e. prefrontal cortical function, hormone levels, etc. Future research combining genotyping with fMRI, hormone assays, and/or additional psychosocial questionnaires would help to disambiguate key moderators. Moreover, examination of the effects of HTR1B haplotype differences on hostility in individuals outside college-age individuals—for example, in adolescent boys or older adult men—could help determine the broader implications of these genetic variants. If differences were observed among adolescent boys, the HTR1B haplotypes could be considered a risk factor for teen aggression and its correlates (fights, arrests, drug and alcohol use, etc.). If genetic differences persisted among adult men, HTR1B haplotypes could be confirmed as a potential cardiovascular risk factor (e.g., Type 1 personality: Miller et al., 1996).

No genotype differences were observed among women. This finding could be interpreted biologically, for example, that only men with low-expression haplotypes are especially prone to aggression-related emotional states due to hormonal or other Y chromosome-linked neural circuitry differences. However, this interpretation is tempered by the fact that these same women showed significant an association of the rs13212041 SNP with self-reported earlier lifetime behaviors (Jensen et al., 2008). A-allele homozygous women at rs13212041 reported a history of more aggression-like behaviors (starting fires, damaging property, and fighting) than G-allele carrier women. Therefore, a more parsimonious explanation is that women with low HTR1B expressing chromosomes may manifest the aggressive phenotype, but they do so earlier in their development. Research has shown that genetic influences on aggressive behaviors are stronger earlier in development among female than male children (e.g., for antisocial behaviors: Jacobson et al., 2002). Indeed our own data indicate that the genotype difference in hostility among males was highest at age 18. If so, then a retrospective history checklist covering the period including childhood to late adolescence could reveal further genetic associations of these HTR1B haplotypes with hostile emotional states among women. Yet, other work has shown that genetic influences on aggression tend to increase for both boys and girls from childhood to young adulthood (Jacobson et al., 2002; Miles & Carey, 1997). Additional research is needed to replicate and determine the mechanism for this gender-specific effect.

There are several methodological strengths in the present study. First, design considerations reduced the likelihood of Type 1 error (moderate sample size; robust measurement of emotional outcomes; demonstrated consistency in findings longitudinally). Second, our choice of HTR1B as a candidate gene was based on studies in animals and humans that implicate this gene in aggressive behavior. Third, three of the five SNPs we examined have previously been shown to modulate two key steps in gene expression: transcription and mRNA translation (Duan et al., 2003; Jensen et al., 2008). Nonetheless, because of the risk for Type 1 error in genetic association studies, we strongly advocate efforts to replicate these findings.

There are also limitations to this study that should be considered. As a genetic association study, our study was correlational in design. Although Duan et al. (2003) and Jensen et al. (2008) established functional effects of three of the SNPs studied, these polymorphisms could also be in LD with other yet to be identified genetic variants that are functional. Our use of haplotype analysis partially addressed this issue by examining linkage among previously identified and commonly studied SNPs. Also, we assumed a static effect of these functional variants on HTR1B gene expression, yet it is possible that these polymorphic regulatory elements have a dynamic effect on the phenotype. For instance, the emotional effects of the HTR1B 3′UTR A/G polymorphism should depend on the presence and activity level of the complementary microRNA (miR-96), yet microRNAs are not continuously active—their amount and activity are regulated by a variety factors still being discovered (Davis et al., 2008, Newman et al., 2008; Viswanathan et al., 2008; Zhou et al., 2008). As a result, it is possible that miR-96 does not exert its effect in all individuals with the rs132120411041 A/A genotype because of other factors that regulate the production and activation of miR-96. Future studies may profit from identifying factors that control the expression and activity of miR-96 and ascertain their potential interaction with genotype to predict hostility. Similarly, the functional effects of the GA promoter haplotype may be conditional on the presence of Ap1-mediated transcription inducers and its effect may be most evident under specific developmental, environmental or life experience conditions. Because our analysis was limited to non-Hispanic Caucasians, it is important to examine these associations in more diverse samples.

Finally, future studies of the association of this SNP with psychological phenotypes might also include standard trait measures that target anger and hostility. However, we anticipate that the effect sizes may be attenuated with standard trait measures as compared with real-time data capture techniques. Although we have assessed emotional states repeatedly over four years, our aggregation of these repeated state assessments essentially created a trait indicator of anger/hostility. We believe that this aggregated indicator of emotional states may be more reliable and less vulnerable to recall error and bias than traditional, once-administered, questionnaire-based trait measures. At the same time, our two-item indicator lacked the breadth of standard trait measures.

Conclusion

Our study broadens the genetic bases of emotionality and suggests a link between common haplotypes in the serotonin 1B receptor gene and angry and hostile emotions in daily life among young men. Phenotypes were best explained by haplotypes that varied in the microRNA binding polymorphism (rs13212041) and, to a lesser extent, the promoter variants (rs11568817 and rs130058). These findings add to a growing literature showing that polymorphisms that influence gene expression related to the serotonin 1B receptor gene play an important role in phenotypic variation—particularly that which is part of a broad aggression-related cluster, which includes anger and hostile emotions. Additional research is needed on the correlates of these haplotypes and the mechanisms by which they exert effects on phenotype, particularly the brain-based processing differences and affective mechanisms that may underlie gene-based emotional differences, e.g., using neuroimaging (e.g., Canli et al., 2006; Hariri et al., 2002). Finally, the apparent gender-specific and potentially developmental manifestations of the association should be examined further and taken into account in subsequent studies of the SNPs examined in this study. Use of real-time data capture techniques may help to increase effect sizes in future genetic association studies.