A population-specific HTR2B stop codon predisposes to severe impulsivity.

Impulsivity, describing action without foresight, is an important feature of several psychiatric diseases, suicidality and violent behaviour. The complex origins of impulsivity hinder identification of the genes influencing it and the diseases with which it is associated. Here we perform exon-focused sequencing of impulsive individuals in a founder population, targeting fourteen genes belonging to the serotonin and dopamine domain. A stop codon in HTR2B was identified that is common (minor allele frequency > 1%) but exclusive to Finnish people. Expression of the gene in the human brain was assessed, as well as the molecular functionality of the stop codon, which was associated with psychiatric diseases marked by impulsivity in both population and family-based analyses. Knockout of Htr2b increased impulsive behaviours in mice, indicative of predictive validity. Our study shows the potential for identifying and tracing effects of rare alleles in complex behavioural phenotypes using founder populations, and indicates a role for HTR2B in impulsivity.

Impulsivity is a broad term describing behavior characterized by action without foresight, decreased inhibitory control and lack of consideration of consequences[1]. Cognitive function, attention and responses to reward are factors that are thought to contribute to the trait of impulsivity. Although impulsivity can be an adaptive dimension of personality, intolerance for delay, disinhibition, and the inappropriate weighting of contingencies are maladaptive[2]. The behavioral manifestations of impulsivity include suicide, addictions, Attention Deficit Hyperactivity Disorder (ADHD), and violent criminality[3] as well as Antisocial Personality Disorder (ASPD), Borderline Personality Disorder (BPD) and Intermittent Explosive Disorder (IED). These behaviors and diagnoses, including impulsivity itself, are moderately heritable[4,5], indicating that it should be feasible to identify genes influencing them. Gene identification would also validate the idea that it is possible to deconstruct the multi-process origins of impulsivity. Still, studies demonstrating that genetic variation predicts impulsivity have been relatively sparse[6-11]. The fact that few genes influencing impulsivity have been discovered could reflect the complexity of the phenotype, the nature of the samples or the methodologies used.

To detect novel alleles that influence impulsivity, we studied severely impulsive Finnish criminal offenders and matched controls. This study had six components (as charted in Supplementary Fig. 1): resequencing and identification of putatively functional variants in severe impulsive probands from a founder population, association and linkage with impulsive behavior, population genetics, evaluation of cognitive effects of the identified variant, gene expression and functionality, and animal studies.

Exon-centric sequencing was performed on fourteen genes involved in dopamine or serotonin function (the genes are listed in Supplementary methods). Dysregulated activity of the monoamine neurotransmitters has been implicated in impulsivity both on a neuropharmacologic basis and on a genetic basis via gene knockouts and/or association studies with common functional variants. In rats, serotonin and dopamine interact in the control of impulsive choice, with differential actions in regions of the prefrontal cortex involved[12]. The spontaneous impulsivity of rats correlates with lower levels of D2 receptors in the nucleus accumbens, predicting liability to compulsive drug seeking and addiction[13], also in humans a reduction in D2 receptors, as well as a decrease in dopamine release, has been described in the ventral tegmental area of cocaine abusers[14]. The serotonin system has long been implicated in impulsivity[15,16] and, in particular, impulsive aggression and suicide. MAOA knockout mice have higher levels of monoamines and increased aggressive behavior[17], and a functional variable number tandem repeat in the MAOA regulatory region (MAOA-LPR) moderates the effect of maltreatment on vulnerability to develop antisocial behavior[8,18]. Brunner et al had shown that a stop codon variant that produces complete deficiency of MAOA activity co-segregated with severe impulsivity[6]. Stress-modified associations have been reported also for a polymorphism in the serotonin transporter (HTTLPR) to suicidality[19,20].

Deep sequencing recently was successfully applied to gene identification in rare Mendelian disorders[21]. In the domain of complex disorders, sequencing revealed putatively functional alleles at a gene previously implicated by GWAS of Type 1 diabetes[22]. Here, we attempted to used sequencing to identify novel loci contributing to a non-Mendelian phenotype.

Sequencing Finnish impulsive subjects

Founder populations can increase power to detect effects of rare alleles. At autosomal loci, Finns are equally diverse as other Europeans, yet a restricted number of founders and isolation have molded the Finnish gene pool[23]. Many disease alleles are more abundant, or unique, to Finland and conversely some disease alleles common in other European populations are rare or nonexistent[23]. From the standpoint of identifying rare or uncommon alleles with roles in complex phenotypes, it is perhaps most important that Finnish ancestry appears to have reduced the genetic heterogeneity of various diseases. For seventeen Finnish disease alleles, 70% of disease chromosomes (and as many as 98% for some diseases) were attributable to a single allele[23].

Sequencing was conducted in 96 unrelated Finnish males with impulsive behavior and an equal number of unrelated Finnish males free of psychiatric diagnoses (Supplementary Table 1 and Methods). Probands had ASPD, BPD or IED and were all violent offenders and arsonists who, because of the extreme nature of their crimes, underwent inpatient forensic psychiatric examination at the University of Helsinki at the time of their initial incarceration. ASPD and BPD share genetic risk for impulsive aggression[4], which is a central characteristic of both of these personality disorders. Impulsivity is also key to IED, described in the DSM-IV as a failure to resist aggressive impulses.

The ninety-six cases were selected for resequencing from a larger Finnish violent offenders cohort comprising 228 cases on the basis that they had highest Brown-Goodwin Lifetime Aggression scores, 23.7 (SD ± 4.9) as compared to 8.1 (SD ± 4.9) in controls. Their higher scores were indicators of a life history of aggressive, violent, and impulsive behavior as behavioral manifestations of impulsive temperament. The ninety-six male controls were free of Axis I and II diagnoses and matched for age, and were selected for sequencing for SNP discovery from a larger control cohort comprising 295 individuals. As compared to controls, cases also had significantly higher impulsivity (action on the spur of the moment) scores on the Karolinska Scales of Personality (p < 0.0001)[24]. However, analysis was conducted on a behaviorally based phenotype, rather than a measure of temperament, because behavior has repeatedly shown the strongest relationship to biological predictors including genes. Genetic loci previously implicated in impulsivity include the MAOA stop codon linked to impulsive behavior in one Dutch family[6], HTTLPR at the serotonin transporter which predicts suicidality[19,20] and the dopamine transporter VNTR which has been associated to ADHD[11]. Impulsive behavior also can be predicted by neurotransmitters and endocrine factors, as illustrated by associations of brain serotonin turnover[25], testosterone levels and a gene/testosterone interaction[9]. Animal behavioral pharmacology, gene knockout and strain-difference studies all primarily rely on measured behavior. By selecting the most phenotypically extreme probands for sequencing, we increased the probability that we would detect functional variants altering impulsivity. Clinical and criminal records, including evaluation of premeditation and spontaneity of crimes, were available for all cases.

Exonic and promoter regions (comprising 82 kb) were amplified in pools of 12 genomic DNAs and sequenced simultaneously at 80x coverage on an Illumina Genome Analyzer, as described in Methods. Sequencing allowed us to identify and accurately estimate frequencies of alleles (Supplementary Fig. 8 compares frequencies determined by sequencing and genotyping, r = 0.94). Of 360 SNPs identified, 44% were known (NCBI Build128). Frequencies of novel SNPs ranged as high as 0.2. Within 37 kb of protein-coding DNA, 25 synonymous SNPs, of which 9 were novel, and 26 nonsynonymous SNPs (nsSNPs) were detected. Of a total of 22 nsSNPs confirmed by Sanger sequencing, 10 were novel.

Association of putatively functional SNPs

Four nsSNPs were predicted to be functional according to both SIFT (Sorting Intolerant From Tolerant) and PolyPhen (Polymorphism Phenotyping): TPH2 Pro206Ser (rs17110563), DRD1 Ser259Tyr, HTR2B Arg388Trp and HTR2B Q20*, a stop codon (Supplementary Table 5). These four nsSNPs were genotyped in male Finnish cases and controls. In a global test of association to an aggregate of potential susceptibility variants, these four putatively functional variants were twice as common in cases (13.0%) compared to controls (6.5%, χ2 = 6.76, p = 0.009; Supplementary Table 6). However, this global association was driven by HTR2B Q20*. 17/228 cases were heterozygous for HTR2B *20 compared to 7/295 controls (χ2 = 7.26, p = 0.007) (Supplementary Table 6), with an allele frequency in controls of 0.012. Eighty-nine pedigrees comprising family members of the violent offenders were collected and all were genotyped without pre-selection for phenotype or genotype, identifying eight HTR2B *20 carrier families (Fig. 1; Methods). Affected status was defined as presence of ASPD, BPD, or IED. The Transmission Disequilibrium Test detected over-transmission of *20 to affected offspring (McNemar χ2 = 5.0, p = 0.025). Similarly, among affected individuals, 6/7 had *20 transmitted, and among unaffected individuals 10/14 did not have *20 transmitted (Supplementary Table 7). From the cumulative binomial distribution, previously proposed for linkage of functional loci in families[26], the likelihood of 16/21 or more linked outcomes was 0.013.

Fig.1

HTR2B *20 co-segregates with impulsivity

Co-segregation of HTR2B *20 with ASPD and AUD in eight informative families.

The HTR2B gene is on 2q36.3-q37.1, a location implicated in early onset obsessive compulsive disorder[27] and illicit substance abuse[28]. However, resequencing of HTR2B in these two studies yielded no functional variants[27,28]. 5-HT2B receptor function in the brain is mainly unknown, however it has been shown that MDMA (Ecstasy) selectively binds and activates 5-HT2B receptors, inducing serotonin release, in mouse raphe nuclei leading to dopamine release in the nucleus accumbens and ventral tegmentum[29], and HT2B agonists increase serotonin transporter phosphorylation[30].

HTR2B Q20* in humans

We assessed molecular functionality of HTR2B Q20* by using RNA and proteins extracted from lymphoblastoid cell lines, and in addition HTR2B expression was measured in multiple brain regions, including frontal cortex, by qPCR (Methods). *20 led to variable nonsense-mediated RNA decay and blocked expression of the 5-HT2B receptor protein (Fig. 2, Methods). HTR2B is widely expressed in the adult human brain, and the frontal lobe is one of the regions where it is most highly expressed (Methods, and Supplementary Fig. 13).

Fig. 2

HTR2B *20 blocks protein expression

a) cDNA and b) protein locations of HTR2B Q20*. c) Variable stop codon-mediated RNA decay determined by cDNA sequencing of twelve *20 heterozygotes. d) *20-mediated blockade of 5-HT2B protein expression in Western blots (validated with three anti-5-HT2B antibodies – described in Methods). The 5-HT2B protein ratio was 1.93: 1 in fourteen Q20/Q20 homozygotes (mean 1.78, SD 2.24) compared to 14 Q20/*20 heterozygotes (mean 0.92, SD 1.14) (p = 0.03) (Methods).

HTR2B *20 is apparently exclusive to Finns. In >3100 individuals representative of worldwide diversity, including the Human Genome Diversity Panel (Supplementary Table 8) one additional *20 carrier was observed: a female with a Finnish surname, and with alcoholism. Indicative of a common origin and founder population effect, *20 was found on a single haplotype background (Supplementary Fig. 9), and in Finns who were likely to be non-admixed (Supplementary Fig. 2). Genetic sub-isolates have been identified within Finland, including families in Eastern Finland. Also, the Finnish population apparently was founded by two waves of migration: Eastern Uralic founders arrived 4000 years ago, followed by Indo-European speakers 2000 years later[23]. However, it is unlikely that the *20 association is an occult admixture artifact because *20 carriers are common across Finland (in Middle, Western and Eastern regions) (Supplementary Fig. 3), and cases and controls did not differ in Finnish ancestry (Supplementary Fig. 4, Methods).

In the 17 violent offenders (from the case/control study) who carried *20, impulsivity played a strong role in their crimes. Although convicted for a variety of offenses including homicide, attempted homicide, arson, battery and assault, 94% of their crimes were committed under the influence of alcohol. The crimes of the *20 carrier probands occurred as disproportionate reactions to minor irritations and were unpremeditated, without potential for financial gain, and recurrent. From court records up to an average age of 43, *20 carriers had committed an average of 5 violent crimes (range 2-12). The *20 cases tended to fulfill criteria for ASPD (82%) and IED (57% meeting 3 of 4 IED criteria), except that alcoholism, ASPD and BPD are exclusionary for IED. Overall, *20 carriers were cognitively normal (mean IQ: 98, SD: 14.9, range 75-124, two with IQ <87, Wechsler Adult Intelligence Scale).

In temperament, measured with the Tridimensional Personality Questionnaire, *20 carriers are, like others with ASPD, higher in Novelty Seeking and Harm Avoidance, but are otherwise more socially attached, empathic and dependent than the other violent offenders within the study group (Supplementary data). Extrapolating from the *20 frequency of 0.012 (and with 174 *20 carriers directly genotyped), 53,000 Finnish males (and as many females) are heterozygous. However, while few *20 carriers are criminals, violent criminals with *20 appear to represent some of the most impulsive individuals within our violent offender cohort. Among 100-155 homicides annually in the Finnish population of 5.3 million, there are few instances of multiple homicide. In our sample, only three individuals were convicted of multiple homicide, and all carried the *20 allele.

In our sample, the influence of *20 was not due to interaction with MAOA or serotonin transporter genotypes (data not shown). However, it was not possible to rule out other gene interactions, or a modifying role of stress. CSF monoamine metabolite levels, another potential confound, did not differ in *20 carriers (Supplementary data). Therefore, it is unlikely that their impulsivity was due to low turnover of serotonin, dopamine or norepinephrine or that *20 substantially affects monoamine metabolism, as does the MAOA stop codon[6].

Risk conferred by *20 appears to be modulated by sex and alcohol. Worldwide, suicide accounts for 1.5% of deaths, and Finland has a very high suicide rate, especially among men[32]. In our study, 70% of the *20 male cases displayed impulsive suicidal behavior (e.g. slashings, hanging attempts, drug overdoses) usually while intoxicated, for an average of 3.2 suicide attempts. At age 33.5 (SD ± 11), 66% had at least one life-threatening suicide attempt. It is unknown if suicide risk conferred by *20 extends to the general population, whose members are at lower risk. Males are more likely to commit suicide[32] and to have ASPD and aggression, with a 10-fold higher preponderance for the early onset-life course persistent variant of ASPD[33]. Moreover, alcohol-related violence is known to be higher among males, and the serotonin system is thought to contribute to individual differences in alcohol-facilitated impulsive aggression[34].

In the violent offender cohort, *20 carriers were cognitively normal and in almost every instance acted out on their impulsivity only when inebriated. Having found the association of *20 to impulsivity in a phenotypically extreme sample, it was important to define *20 frequency and relationship to behavior in the wider population, even though the only possible follow-up was in Finland. In >6000 Finns ascertained epidemiologically (rather from the criminal population), the *20 allele frequency was 0.012 (the same as the frequency in controls) (Supplementary Table 9). We identified one *20 homozygote, a young male adult with no major medical illness, but with a history of violent behavior whilst under the influence of alcohol (Supplementary methods).

We followed up cognitive effects of *20 in 933 individuals in the FT12/FT16 Twin Studies (22 with the stop codon) (Supplementary methods). Overall, *20 carriers were again cognitively normal. However, male (but not female) *20 carriers had significantly lower Digit Span Forward (p = 0.002), and Backward (p < 0.001) scores possibly indicating selective impairment in working memory (Supplementary Fig. 12), a specific measure of frontal cortical function.

Htr2b−/− mice

Although severe developmental consequences have been observed in Htr2b knockout mice, approximately 50% of the mice that survive the first postnatal week are apparently normal as adults[35]. These mice were reported to be impulsive in an open field novelty test[29]. We assessed Htr2b knockout mice for five separate measures of impulsivity and novelty seeking: delay discounting, activity in a novel environment, exposure to a novel object, motor activity after a dopamine D1 receptor agonist, and decreased latency to eat in the novelty suppressed feeding test (hyponeophagia). The Htr2b−/− mice were more impulsive and more responsive to novelty in all of these tests (Fig. 3). In rats, both impulsivity and response to novelty are predictors for the development of addiction-like behaviors[36]. In addition to their differences in behavior, Htr2b−/− males had a three-fold elevation in plasma testosterone (Fig. 3) (Supplementary methods). Testosterone (measured in the CSF of nine heterozygous violent offenders) also appeared to be higher in human males carrying *20 (Supplementary Fig. 11). This raises the possibility of interaction between *20 and testosterone to contribute to impulsive behaviors, as was reported between MAOA and testosterone in the same population of Finns that we studied here[9].

Fig.3

Increased impulsivity and novelty seeking in Htr2b−/− mice

a) Increased locomotor response of Htr2b−/− mice to environmental novelty and b) to dopamine D1 receptor agonists. c) Increased number of contacts of Htr2b−/− mice with a novel object. d) Increased delay-discounting of Htr2b−/− mice. e) Reduced hyponeophagia in 18-hr starved Htr2b−/− mice. f) Male Htr2b−/− mice have three-fold higher plasma testosterone levels as compared to control mice. * p<.05, ** p<.01, *** p<.001.

Discussion

The aim of this study was to identify genetic variation associated with impulsivity, an intermediate phenotype thought to contribute to several psychiatric disorders, including addictions. [12] The goal is to track shared genetic factors in these diseases and contribute to their reconceptualization on a neurobiological basis. Another purpose of identifying genes influencing impulsivity is to determine which of the potential etiologies and types of impulsivity, for example novelty seeking versus executive dysfunction[36] are important in human populations. The discovery of genes influencing impulsive behavior would validate the idea that it is possible to deconstruct the multi-process origins of impulsive behavior.

HTR2B *20 is associated and co-segregates with disorders characterized by impulsivity, reflected in severe crimes committed on the spur of the moment, as documented by criminal and clinical records, and under alcohol intoxication, a condition where impulse control is impaired. Thus the *20 allele can be regarded as one determinant of behavioral variation. However, presence of *20 was not itself sufficient: male sex, testosterone level, the decision to drink alcohol, and probably other factors such as stress exposure, have important roles. Although relatively common in Finland, HTR2B *20 is unlikely to explain a large fraction of the overall variance in impulsive behaviors. There are likely to be many pathways to impulsivity in its various manifestations, and the genetic association may be present only in the most phenotypically extreme.

It is un-surprising that a stop codon variant discovered by sequencing within a founder population is common only in it, and even restricted to it. However, this observation is also in line with the significance of *20 as a complete loss of function variant, and with the behavioral consequences in some heterozygous carriers. The relatively high frequency of *20 in Finns would thus reflect its status as a founder mutation, in contrast to MAOA, COMT and HTT alleles that are common worldwide, that more modestly affect molecular function, and that may have counterbalancing selective advantages. However, it is highly unlikely that Finns are unique in possessing a severe genetic variation leading to impulsivity. There is the previous example of the MAOA stop codon, found in one Dutch family. On average, ten or more heterozygous stop codons reside in the genomes of each individual of European ancestry[21] but perhaps because the source populations from which the probands sequenced did not have founder characteristics, no common stop codon had yet been reported for a neurotransmitter gene. Although rare variants identified in founder populations are more likely to be confined to those populations, analyses of the relationship between gene variation and phenotype can be conducted within the founder population, identifying new candidate genes and pathways influencing behavior or other etiologically complex phenotypes.

As has often been illustrated, the availability of mouse genetic models, including gene knockouts, offers an opportunity to test the predictive validity of genetic discoveries and to define effects in contexts where genetic background and environment are better controlled. The Htr2b mouse knockout reveals more general effects of HTR2B deficiency on behavior, including effects on novelty seeking. This could be explained by pleiotropic actions of the serotonin 2B receptor. On the other hand, the effect of the Htr2b knockout on delay discounting appears to validate the effect of the *20 stop codon on impulsivity in people. In people, we observed a significant association between the HTR2B *20 variant and impairment in working memory, a neurocognitive process contributing or predictive of executive cognitive function. The ability to store and integrate knowledge about possible choices with current context enables the individual to select appropriate cognitive strategies and generate optimal reactions. This is coherent with the impulsivity observed in HTR2B *20 cases who seemed deficient in ability to weigh the consequences of their acts.

The use of deep sequencing to detect a stop codon associated with impulsivity in a founder population reveals a role for the HTR2B gene in behavior. It also suggests that this approach may be applicable to other complex behavioral traits, including those diseases for which impulsivity is itself an intermediate phenotype.

METHODS SUMMARY

Fourteen serotonergic and dopaminergic genes were resequenced (Solexa GA2) in 96 Finnish Caucasian male violent offenders and 96 matched controls free of psychiatric diagnosis. Exon-centric sequencing was performed by amplifying 108 regions, for a total of 82 kb, in pools of 12 subjects. HTR2B Q20* was genotyped in a Finnish sample of 228 cases and 295 controls, in 89 Finnish families, and in 5684 individuals belonging to either a Finnish family dataset (N = 1885), the Older Finnish Twin cohort (N = 2388) and FinnTwin16 and FinnTwin12 studies (N = 1411) as described in detail in supplementary methods, and in >3100 samples representing worldwide diversity. Genotyping was performed with a custom 5′ exonuclease assay (Applied Biosystems 7900) using these primers and probes: 186 ancestry informative markers were genotyped on 1536-SNP arrays (Illumina). Quantitative PCR for HTR2B expression in thirteen human brain regions was determined by ABI Taqman gene expression assays (Hs01118766 and Hs00168362). B-actin was the internal control. Total protein and total RNA were extracted from lymphoblastoid cell lines using the TRIzol LS reagent protocol (Invitrogen). Nonsense-mediated RNA decay was detected by sequencing on a 3700ABI capillary sequencer cDNA from HTR2B Q20/*20 heterozygotes. HT2B protein was measured in 12 Finnish Q20/Q20 homozygotes and 14 Finnish Q20/*20 heterozygotes. Blots were probed with antisera raised against the N-terminal (mouse monoclonal Ab, Novus Biologicals), internal (goat polyclonal Ab, Santa Cruz Biotechnology) or C-terminal (rabbit polyclonal Ab, Santa Cruz Biotechnology) regions of the HT2B receptor, and GAPDH Ab (Millipore). Densitometry performed using NIH ImageJ. Htr2b−/− knockout mice were made in a pure 129Sv/PAS background and compared to 129/SvPAS control mice (8-10 week old) for four measures of response to novelty and for delay discounting.

forward primer: 5′-AGAGTGTCTGAACTTCAAAGCACAA-3′

reverse primer: 5′-TCCAGACCAGTTAGAAGAGATAACGT-3′

probe 1: 5′–AGGTGCTCTGCAAAAT-3′

probe 2: 5′-AGGTGCTCTACAAAAT-3′