Tryptophan hydroxylase 2 (TPH2) linkage disequilibrium (LD) in US whites (A), Finnish whites (B), African Americans (C), and southwestern American Indians (D). Pairwise LD was computed using Haploview.35D′ values of 1 are represented by red squares. The relative positions of the 15 markers are shown at the top. Haplotype block boundaries are defined using a minimum average D′ value of 0.80.
Mean cerebrospinal fluid 5-hydroxyindoleacetic acid (CSF 5-HIAA) levels in controls and suicide attempters predicted by the tryptophan hydroxylase 2 (TPH2) yin haplotype. The CSF 5-HIAA concentrations in individuals who are homozygous 212121/212121 for the yin vulnerability-associated haplotype (+/+), heterozygous 212121/other (+/−), and missing the yin haplotype (−/−) are compared. Asterisk indicates P = .006, +/+ vs −/− in the control group. Error bars represent SD.
Zhou Z, Roy A, Lipsky R, Kuchipudi K, Zhu G, Taubman J, Enoch M, Virkkunen M, Goldman D. Haplotype-Based Linkage of Tryptophan Hydroxylase 2 to Suicide Attempt, Major Depression, and Cerebrospinal Fluid 5-Hydroxyindoleacetic Acid in 4 Populations. Arch Gen Psychiatry. 2005;62(10):1109-1118. doi:10.1001/archpsyc.62.10.1109
Copyright 2005 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2005
Tryptophan hydroxylase 2 (TPH2) encodes the rate-limiting enzyme for brain serotonin biosynthesis. It was recently reported that the TPH2 haplotype was linked to depression in humans.
To determine the association of TPH2 with suicide attempt, major depression, and a neurochemical intermediate phenotype, cerebrospinal fluid 5-hydroxyindoleacetic acid.
We resequenced TPH2 coding, 5′ promoter, 3′-untranslated region, and splice junction regions in 190 individuals selected for ethnic and clinical diversity, determined haplotype structure using 15 single nucleotide polymorphisms spanning 106 kilobases (kb), and performed linkage analysis in 1798 cases and controls representing 4 populations (657 African Americans, including 104 suicide attempters and 135 with major depression; 513 Finnish whites, including 150 suicide attempters; 146 US whites, including 81 with depression, anxiety disorder, or both; and 482 southwestern American Indians, including 123 suicide attempters and 191 with depression, anxiety disorder, or both) and in 94 Finnish whites for cerebrospinal fluid 5-hydroxyindoleacetic acid levels.
Sixteen single nucleotide polymorphisms, including Pro206Ser, were detected. The 15-locus panel defined and maximized information content from 2 haplotype blocks in whites, 3 haplotype blocks in African Americans, and the single haplotype block spanning TPH2 in southwestern American Indians. Among common Block1b haplotypes were 2 in yin and yang (opposite) configuration, indicating ancient origin. The yin haplotype, 212121, was increased in frequency in suicide attempters in both populations tested (Finnish whites and African Americans). It was associated with major depression and anxiety disorders in US whites and with major depression in African Americans. The yin haplotype was moderately predictive of lower cerebrospinal fluid 5-hydroxyindoleacetic acid concentrations in controls but not in cases.
Haplotype linkage of TPH2 to suicide attempt and major depression and to a mediating phenotype, cerebrospinal fluid 5-hydroxyindoleacetic acid, provides preliminary evidence of a functional locus potentially within a haplotype block at least 52 kb in size.
Serotonergic dysfunction is one cause of negative mood states, including anxiety and depression.1 In addition, serotonin modulates sleep, food intake, and many other behaviors. Low serotonin turnover leads to behavioral disinhibition, causing impulsive aggression and antisocial behaviors in humans2,3 and irritable aggression in rodents, for example, after inhibition of tryptophan hydroxylase (TPH) with parachlorophenylalanine. Serotonergic hypofunction has also been implicated in suicide.4,5 5-Hydroxyindoleacetic acid (5-HIAA), an oxidative metabolite measurable in cerebrospinal fluid (CSF), indexes serotonin turnover, at least in the frontal cortex,6 and has been repeatedly correlated to impulsive behaviors.2,3 Cerebrospinal fluid 5-HIAA shows substantial heritability in rhesus macaques,7 and there are strong indications that CSF 5-HIAA is also heritable in humans, where the question has been addressed only in a relatively small sample.8 However, CSF 5-HIAA levels are also affected by a variety of environmental factors, including alcohol exposure,9,10 stress,11 diet,12 and selective serotonin reuptake inhibitors.13 Serotonergic dysfunction seems to be a cause and consequence of substance abuse.9,10 Linkage studies of serotonin-related genes to behavior are numerous and include coherent findings that implicate serotonin transporter functional variation in anxiety and dysphoria14- 16 and in obsessive-compulsive disorder17 (Xiangzhang Hu, PhD, R.L., and D.G., unpublished data, 2005).
The first and rate-limiting step in serotonin synthesis is catalyzed by TPH. Studies on associations between sequence variations of the TPH1 gene and psychiatric and behavioral disorders yielded extensive but inconclusive results. Two sequence polymorphisms in intron 7 of TPH1, A779C and A218C,18 were the focus of positive association studies with suicidal behavior,19- 21 mood disorders,22 and aggression, irritability, and anger-related personality traits.23 However, these findings were not always replicated.24- 26 Subsequently, Walther and colleagues27 discovered TPH2. Based on several lines of evidence, TPH2 encodes brain TPH. A TPH2 Pro447Arg missense variant found in several mouse strains altered brain serotonin levels.28 The human TPH2 gene spans 97 kilobases (kb), consists of 11 exons, and is located on chromosome arm 12q15. The TPH1 behavioral linkages remain of interest because of possible behavioral effects of serotonin in the periphery or in the pineal. However, interest has understandably shifted to TPH2.
Evidence of a haplotype-based association of TPH2 with depression was reported in German whites.29 In this initial study linking TPH2 to behavior, 10 single nucleotide polymorphisms (SNPs) were used, defining a 28-kb TPH2 gene region across which linkage disequilibrium (LD) is high and thus indicating the existence of a haplotype block anywhere in which a responsible functional locus might reside. A smaller study30 reported negative findings on TPH2 linkage to suicidality. Recently, a functional Arg441His missense variant was discovered and linked to major depression.31 However, this interesting variant is rare in most populations because our group and 2 collaborating laboratories could not detect it by resequencing the relevant region of TPH2 in approximately 779 individuals, including 403 with major depression.32
In the present study, to investigate the role of TPH2 in interindividual variation in serotonin function and in the pathogenesis of suicide and major depression, we surveyed this gene for variation by resequencing the coding region, splice junctions, and core promoter in 190 individuals. For genetic linkage, we used a 15-locus panel spanning a 106-kb TPH2 region, defining haplotype structure and detecting ancient yin and yang TPH2 haplotypes. We observed congruent haplotype-based associations with behavior in 3 ethnically and culturally distinct case-control data sets and also found a congruent association of the TPH2 risk haplotype with low levels of CSF 5-HIAA, a neurochemical intermediate phenotype for suicidality and depression.
To identify polymorphic DNA markers and potentially functional sequence variations in TPH2, a total of 4.25 kb of the gene was analyzed in 190 individuals. The 11 TPH2 exons, intronic regions flanking each exon, 1 kb of promoter region immediately upstream of the transcription initiation site, and 0.5 kb in the 3′-untranslated region (UTR) were screened using denaturing high-performance liquid chromatography followed by direct sequencing. The resequencing sample was composed of 40 African Americans, 40 Finnish whites, and 110 US whites. The sample was also clinically diverse, being composed of individuals with disorders potentially related to serotonin dysfunction, including anxiety disorders (n = 35), major depression (n = 35), anorexia nervosa (n = 20), obsessive-compulsive disorder (n = 20), antisocial personality disorder (ASPD) (n = 40), suicidal behavior (n = 40), and alcohol dependence (n = 110). Genomic DNA was prepared from lymphoblastoid cell lines derived from blood samples obtained after informed consent was provided and under research protocols approved by institutional review boards at the National Institute on Alcohol Abuse and Alcoholism, the National Institute of Mental Health, and the University of Helsinki.
Amplicons were optimized using a software program (Wavemaker; Transgenomic Co, Omaha, Neb). Fifty nanograms of DNA was amplified in a 20-μL reaction volume containing 0.25μM forward primer and 0.25μM reverse primer (available on request), 50mM potassium chloride, 100mM Tris hydrochloride (pH 8.8), 2.5mM magnesium chloride, 0.2mM deoxynucleotide triphosphates, and 1 U of AmpliTaq Gold (Applied Biosystems Inc, Foster City, Calif). Polymerase chain reactions were performed at 95°C for 10 minutes, followed by 35 cycles each at 95°C for 15 seconds, 56°C to 62°C for 20 seconds, 72°C for 30 seconds, and a final extension step of 72°C for 10 minutes. Polymerase chain reaction products were denatured at 95°C for 6 minutes and were slowly cooled to 60°C for renaturation before being analyzed using denaturing high-performance liquid chromatography for the presence of heteroduplexes, indicating sequence heterozygosity. Melting temperature was also empirically verified for each amplicon. Heteroduplex DNAs were separated from homoduplex DNAs in DNA separation columns on denaturing high-performance liquid chromatography devices (Wave; Transgenomic Co). A 2-buffer gradient was used: buffer A consisted of 0.1M triethylammonium acetate, and buffer B contained 0.1M triethylammonium acetate and 25% acetonitrile.
DNA was resequenced using a cycle-sequencing kit (BigDye Terminator v3.1 kit; Applied Biosystems Inc). Fifty nanograms of polymerase chain reaction product was used in a 10-μL reaction volume with 5μM primer, 1 μL of BigDye Terminator mix, and 1.5 μL of 5× buffer. Sequencing reactions were performed at 96°C for 1 minute, followed by 25 cycles each at 96°C for 10 seconds, 50°C for 5 seconds, and 60°C for 4 minutes. Sequencing reaction products were purified using ethanol precipitation and were analyzed using a sequencer (model 3100; Applied Biosystems Inc).
The Finnish sample consisted of 513 men, including 196 controls and 317 individuals diagnosed as having various psychiatric disorders (Table 1). They were studied after informed consent was obtained using a human research protocol approved by the National Institute of Mental Health and the University of Helsinki institutional review boards. Most cases were alcoholic violent offenders who had undergone forensic psychiatric examination. The Structured Clinical Interview for DSM-III-R was administered to all the participants, and DSM-III-R lifetime psychiatric diagnoses were blind rated. Findings from this population have also been reported previously, and additional details are available elsewhere.20 Individuals were scored as positive for a history of suicide attempt only when medical records documented a severe attempt, a relative or partner confirmed the history, or there was physical evidence (eg, scars) confirming the history. Of the 317 cases (many with multiple diagnoses), 28 also had major depression. However, because of the small number of individuals in this population with major depression, those data were not analyzed. The 196 controls were unrelated, healthy volunteers.
Participants were paid volunteers recruited from the general population of the Washington, DC/Baltimore, Md area. Written informed consent was obtained according to a human research protocol approved by the human research committee of the National Institute on Alcohol Abuse and Alcoholism. Ethnic origin was self-identified. Blind-rated DSM-III-R lifetime psychiatric diagnoses were obtained from Schedule for Affective Disorders and Schizophrenia–Lifetime version interviews. Tridimensional Personality Questionnaire scores and neuroticism subscale scores from the Eysenck Personality Questionnaire were obtained from most participants. All the participants completed the Spielberger State Anxiety Questionnaire. This sample consisted of 146 individuals, including 65 controls free of psychiatric disorders (Table 1). The 81 anxiety/depression cases included 38 with major depression; 26 with panic disorder, phobic disorder, or obsessive-compulsive disorder; and 17 with depression not otherwise specified. In the anxiety/depression group, 22 individuals also had a history of alcohol dependence or abuse.
A total of 342 cases with various psychiatric disorders were ascertained through the inpatient rehabilitation ward of the Substance Abuse Treatment Program at the Department of Veterans Affairs (East Orange). Owing to the study site and the male-female ratio of addictions, most cases are men. The 315 controls were recruited through the Diabetes Clinic (East Orange) or the churches in the same geographic area, which resulted in a sample with 42.5% men and 57.2% women. Participants gave informed consent, and the study was performed using a human research protocol approved by the institutional review board of the Department of Veterans Affairs. All participants were interviewed using the depression section of the Structured Clinical Interview for DSM-IV to determine whether there was a lifetime history of a major depressive episode according to DSM-IV criteria. A suicide attempt was defined as a self-destructive act with the intent to end one’s life and was not self-mutilatory. Many of the 342 cases had multiple diagnoses. Controls were free of psychiatric disorders (Table 1).
All the participants were members of a southwestern American Indian tribe. Additional description of the study sample is also found elsewhere.33 In summary, participants were recruited from 3 large multigenerational pedigrees that were selected without reference to clinical diagnosis. The calculated average coefficient of relationship was 1.2%, compatible with the restricted size of this population, its low level of outbreeding, and its overall average coefficient of relationship. This average degree of relationship is equivalent to the second cousin once-removed or third cousin level. All the participants gave informed consent, and the study was conducted using a human research protocol approved by the National Institute on Alcohol Abuse and Alcoholism. Semistructured interviews were conducted using a modified version of the Schedule for Affective Disorders and Schizophrenia–Lifetime version, and diagnoses were made by blind rating and using Research Diagnostic Criteria for a Selected Group of Functional Disorders and DSM-III-R criteria. Provision was made to limit diagnostic errors due to culture-specific responses to questions relating to depression, schizophrenia, and ASPD. A total of 62 controls and 420 cases with various psychiatric disorders were included (Table 1). Controls were healthy individuals free of psychiatric disorders.
Genotyping was conducted using 5′-exonuclease (TaqMan) assays. Five SNP markers were Assays-on-Demand (Applied Biosystems Inc). For the other 10 SNPs, pairs of flanking primers and fluorescence-labeled allele-specific probes (available on request) were designed using a software program (Primer Express; Applied Biosystems Inc). Reactions were performed in 384-well plates in a 5-μL volume containing 10 ng of genomic DNA, 0.5μM primers, 0.2μM probes, and 2.5 μL of Master Mix (Applied Biosystems Inc). The amplification protocol consisted of 50°C for 2 minutes and 95°C for 10 minutes, followed by 40 cycles each at 95°C for 15 seconds and 59°C or 60°C for 1 minute. The results were analyzed using a sequence detector (model 7900; Applied Biosystems Inc) and software (Sequence Detection System 2.0; Applied Biosystems Inc). Genotyping accuracy was verified by replicate genotyping of 10% of the samples, randomly selected; genotyping accuracy was greater than 0.99. The genotype completion rate was greater than 0.95.
The CSF 5-HIAA level was measured in 94 Finnish whites, and they were all free of medications, including selective serotonin reuptake inhibitors. The CSF samples were obtained by lumbar puncture between 8 and 9 AM after an overnight rest. Concentrations of 5-HIAA were measured in the first 12 mL of CSF collected. Measurement was accomplished using high-performance liquid chromatography and electrochemical detection.34
Genotyping results were tested for Hardy-Weinberg equilibrium. Individual SNP genotype and allele frequencies were compared between cases and controls by using the χ2 test. D′ values for all marker pairs were computed and LD block structures were evaluated using Haploview.35 A minimum average D′ value of 0.80 was used to define block boundaries (however, average D′ values in most blocks were substantially higher). Haplotype frequencies were determined using PHASE v2.02.36 Individual haplotype associations were analyzed using the χ2 test, and permutation comparisons of the case and control groups were performed using PHASE v2.02. For haplotypes encompassed by a single LD block, no Bonferroni corrections were made for the χ2 test. Diplotype odds ratios were calculated using nominal logistic regression and a software program (JMP version 5.0; SAS Institute Inc, Cary, NC).
Resequencing TPH2 in a clinically and ethnically diverse sample of 190 individuals revealed 16 SNPs (Table 2). Five SNPs were detected in the coding region, including 1 at nucleotide 33681 in exon 6 that results in the previously unreported amino acid substitution Pro206Ser. As noted previously herein, a recently described functional Arg441His variant31 was not observed. Six of the SNPs detected by resequencing were located in the promoter or 5′-UTR, 3 were in the 3′-UTR, and 2 were in flanking intronic sequences. Six of the 16 SNPs had previously been reported in the National Center for Biotechnology Information Single Nucleotide Polymorphism Database (dbSNP) or the Celera Discovery Systems databases, although the allele frequencies of these SNPs were unknown. The preliminary allele frequencies of all the SNPs, based on denaturing high-performance liquid chromatography heterozygosity, are reported in Table 2.
Haplotype structure was determined in controls from 4 populations by genotyping 15 SNPs across 93 kb of the TPH2 gene and 13 kb of the flanking sequence (Table 3). This panel of markers included 6 SNPs either identified or confirmed by resequencing in this study and 9 otherwise available. Significant interpopulation differences in allele frequencies were observed for most markers (Table 4). Three SNPs (SNPs 6, 9, and 11) were monomorphic in southwestern American Indians, and an intron 4 SNP (SNP 5) with a frequency of 0.065 in African Americans was at low frequency (0.02) in US whites and was essentially absent in Finnish whites and southwestern American Indians, 2 populations that are thought to have undergone very low rates of admixture.
In southwestern American Indians, the entire TPH2 gene may be encompassed by 1 haplotype block, as shown in Figure 1, which displays pairwise D′ values across all 15 loci in the 4 populations. In southwestern American Indians, most markers from SNP 2 to SNP 15 are in very strong LD and can be selected for the purpose of defining linkage within the TPH2 region on the basis of information content rather than their physical position in this interval. Both white populations show evidence of 2 LD blocks: Block1, defined by SNP 2 to SNP 11 and encompassing the promoter region to intron 8, and Block2, defined by SNP 12 to SNP 15 and encompassing exon 9 past the 3′-UTR. In African Americans, there is evidence of 3 haplotype blocks: Block1a, SNP 2 to SNP 4 (encompassing the promoter region, intron 4); Block1b, SNP 6 to SNP 11 (introns 5-8); and Block2.
To investigate associations between TPH2 and psychiatric phenotypes, case-control comparisons were performed in 4 populations. The interrelated psychiatric disorders evaluated were anxiety/depression in US whites; ASPD, suicidal behavior, impulsive behavior, and alcohol dependence in Finnish whites; major depression, suicidal behavior, and alcoholism in African Americans; and anxiety/depression and suicidal behavior in southwestern American Indians. Because we do not know whether any of the 15 TPH2 SNPs that we tested are functional, the linkage analysis was haplotype based to reduce testing and to maximize information content within the haplotype blocks. However, the individual results for the 15 SNPs are, for completeness, listed in Table 4, where significant (uncorrected for multiple testing) allele frequency differences for several SNPs in the central region (introns 5-8) of TPH2 are visible. In US whites, SNPs 6 and 9 are associated with anxiety/depression, with trends for association with adjacent markers (SNPs 4-11). In Finnish whites, SNPs 10 and 11 are associated with suicidal-impulsive behavior. In African Americans, SNPs 4, 7, and 12 are associated with suicide attempt or major depression. There were no individual locus associations with anxiety/depression or suicidality in southwestern American Indians. There is no evidence to individually implicate any of these SNPs. All of the SNPs displaying evidence of association are located in intronic regions except for SNP 8, which is a synonymous variant in exon 7.
In all 4 populations, there are common yin and yang (opposite configuration) haplotypes representing Block1b, defined by SNP 6 to SNP 11, and spanning the 52-kb region from intron 5 to intron 8 (Table 5). The yin haplotype, 212121, is the most abundant haplotype in US (frequency, 0.53) and Finnish (frequency, 0.47) whites, and it has a frequency of 0.15 in African Americans and 0.38 in southwestern American Indians. The yang haplotype, 121212, has a similar frequency across 3 populations (0.20 in US whites, 0.18 in Finnish whites, and 0.17 in African Americans) but is absent in southwestern American Indians, in whom SNPs 6, 9, and 11 are monomorphic. The existence of yin and yang haplotypes across human populations is evidence of the effects of selection to maintain ancient chromosomes.37
The TPH2 linkage signal to suicidality and dysphoric phenotypes in US whites, Finnish whites, and African Americans derived from Block1b (Table 5), the same region that showed single-marker associations (Table 4). The yin haplotype, 212121, seems to be implicated in risk for anxiety, depression, and suicidal behavior in these 3 populations (Table 5). This haplotype is more abundant among African Americans and Finnish whites with a history of suicidal and suicidal/impulsive behavior compared with controls from these populations, who were free of psychiatric disorders. This haplotype is also associated with major depression in African Americans. A similar but nonsignificant trend for association of the yin haplotype with anxiety or depression was also seen in US whites. In both white populations, the yang haplotype, 121212, seems to be protective. Southwestern American Indians, who lack the yang haplotype, showed no association and no trend for association of haplotypes at Block1b. The implications of TPH2 Block1b haplotypes for risk were evaluated by calculating diplotype odds ratios (Table 6). In African Americans, homozygosity for 212121 leads to odds ratios of approximately 10:1 for suicidality and also approximately 10:1 for lifetime major depression. In Finnish and US white populations, the odds ratios associated with 212121 homozygosity were 2.15 for impulsive suicidality in Finnish whites and 1.75 for anxiety/depression in US whites, and the latter value was nonsignificant.
As noted, haplotype 121212 (the yang haplotype) may be protective against suicide and depression in white populations. It was significantly lower in frequency among US whites with anxiety/depression and among Finnish whites with impulsive/suicidal behavior and impulsive behavior. However, in African Americans, another Block1b haplotype, 222121, seems to be protective. This haplotype displayed significantly lower frequencies in individuals with a history of suicidal behavior or lifetime major depression (Table 5). The protective nature of these 2 population-specific haplotypes was also shown by the negative odds ratios associated with individuals who have either 1 or 2 copies of these haplotypes (Table 6). No significant haplotype association was found for ASPD in Finnish whites or for alcohol dependence in the Finnish and African American populations (data not shown).
Haplotype 212121, implicated in risk for suicidal behavior and anxiety/depression, was associated with lower CSF 5-HIAA levels in controls free of psychiatric disorders and medications (Figure 2). In particular, individuals who are homozygous for haplotype 212121 display significantly lower levels of CSF 5-HIAA than those who do not have haplotype 212121. Suicide attempters showed no significant relationship between TPH2 haplotypes and CSF 5-HIAA levels. As noted previously herein, suicide attempters also had multiple other psychiatric diagnoses.
Detection of linkage of TPH2 haplotypes to major depression by Zill and colleagues29 provided evidence of a functional locus somewhere within TPH2. Herein we add positive evidence for TPH2 haplotype linkage to anxiety/depression phenotypes and suicide attempts in 2 populations. We also find evidence of linkage of TPH2 to an intermediate phenotype: serotonin turnover as indexed by CSF 5-HIAA concentration.
We determined the TPH2 haplotype structure using 15 SNP markers spanning the entirety of the TPH2 gene and its flanking sequences (a total of 106 kb). Southwestern American Indians showed the strongest LD, the lowest SNP marker and haplotype diversity, and a single haplotype block across the entire TPH2 gene. Southwestern American Indians lacked the yang haplotype, implicated as a protective haplotype in the present study, and showed no TPH2 linkages to behavior, in contrast to the other 3 populations we studied. Two haplotype blocks were observed in both white populations we studied. In African Americans, LD was weakest and 3 haplotype blocks were observed. Because of population-specific LD patterns and haplotype block structures, the African American population was most useful in narrowing the implicated region. For African Americans and both white populations, the major linkage signals to suicidality and anxiety/depression derive from Block1b, which spans 52 kb, from intron 5 to intron 8, and encompasses the region implicated by Zill et al29 except for exon 5.
Direct haplotype comparisons between this study and that by Zill et al are complicated by the fact that different SNP sets were used. However, the yin haplotype, 212121, which we implicate in the risk for anxiety, depression, suicidal behavior, and lower CSF 5-HIAA levels, is similar in frequency to haplotype 1 of Zill et al (0.53 in US whites and 0.47 in Finnish whites vs 0.51 in the German whites of Zill et al29). None of the other haplotypes representing Block1b has a frequency greater than 0.21, so whereas the cross-study identity of other Block1b haplotypes is questionable, there is little (or no) question that our yin haplotype is haplotype 1 of Zill et al.29 Haplotype 1 also displayed higher frequency in patients with major depression compared with controls (0.57 vs 0.51).29 The protective yang haplotype in our study is equivalent in frequency to haplotype 3 of Zill et al. Identity is again not completely certain, but haplotype 3 of Zill et al29 was significantly less frequent in cases with major depression than in controls (0.08 vs 0.16), and haplotypes 1 and 3 of Zill et al are also in yin and yang (opposite) configurations.
Although they are statistically significant, the yin and yang haplotype associations with vulnerability and protection found by Zill et al29 and in the present study are of modest effect size. This finding could imply that the functional locus on the haplotype is itself less common or that the functional locus has a modest effect on vulnerability. De Luca et al30 reported negative findings for linkage of TPH2 to suicidality, but, as the researchers pointed out, small sample size might have made their results inconclusive. Also, only 1 of the 3 markers they genotyped (hCV8376173) is located in haplotype Block1b. We needed 6 SNPs to define the risk and vulnerability haplotypes in this region. Another potential problem is that rare and uncommon haplotypes estimated using any method (eg, E-M methods) are subject to higher error because of increased sampling variance and genotyping errors that produce rare haplotypes that actually do not exist. However, the yin and yang risk and vulnerability haplotypes we identified are fairly common, ranging in frequency from 0.15 to 0.53 (yin) and 0.17 to 0.20 (yang), except in southwestern American Indians, in whom the yang haplotype is absent. It is likely that there is some inaccuracy in the estimation of haplotype frequencies; however, the level of inaccuracy is likely to be low (PHASE v2.02, which implements a bayesian algorithm, estimates the SE for these haplotype frequencies to be <0.01). The consistency of haplotype associations across populations and with an intermediate phenotype, the accuracy of the genotyping, and the opposite configuration nature of the TPH2 risk and vulnerability haplotypes indicate that these haplotype linkages are not attributable to errors in haplotype estimations.
The TPH2 associations are with phenotypes in the behavioral domain of anxiety dysphoria (internalizing disorders) and the closely related behavior of suicidality. These data sets include a variety of psychiatric disorders, including ASPD (an externalizing disorder) and alcohol dependence. The Finnish population included 196 individuals with alcohol dependence and 100 with ASPD, but linkage to TPH2 was not observed (data not shown) despite the fact that serotonin dysfunction is implicated in both disorders. It is possible that linkage might be detected in larger samples of patients with externalizing disorders or in clinically refined subsamples of patients with addictions.
In case-control association studies, population stratification can produce false-positive results due to ethnic stratification of diagnostic groups and a difference in allele (haplotype) frequencies between populations. This could be a special concern for the African Americans we studied because of their genetic diversity and history of admixture. However, several findings make it unlikely that the associations with the TPH2 haplotype are stratification induced. First, the same yin vulnerability haplotype associations were observed in Finnish whites, US whites, and African Americans and very likely also in the German whites studied by Zill et al.29 The haplotype associations are thus directionally congruent across multiple populations, including Finnish whites, in whom population stratification is unlikely. We recently verified the low genetic admixture of our Finnish sample using a genomic control panel composed of 200 SNP loci specifically selected for large cross-population variation (data not shown). Second, the associations are specific to phenotypes in the anxiety and dysphoria domain, whereas the frequencies of various other diagnoses (eg, alcoholism) actually vary more across populations. Third, the significant findings are not driven by TPH2 SNPs that show interpopulation differences. Nine of the 15 SNP markers differ substantially in allele frequencies between the African Americans and the 2 white populations we studied (Table 4). However, in African Americans, where 3 TPH2 SNPs were associated with major depression and suicidal behavior, 6 other TPH2 SNPs showed no association with diagnosis but nevertheless had large, or larger, cross-population differences. This strongly indicates that TPH2 associations in African Americans were not primarily caused by admixture. Finally, low levels of CSF 5-HIAA, a relevant but clinically occult intermediate phenotype, were predicted by the TPH2 yin haplotype.
The CSF 5-HIAA level has been correlated to depression, behavior disinhibition, and suicide in many studies.2- 4 These behaviors have multiple origins, making it possible to parse out the effect of TPH2 on particular dimensions of behavior determined by serotonin neurochemistry. The availability of CSF 5-HIAA, with its limitations as a measure of brain serotonin turnover,6 takes us part of the way toward this goal. In controls, the lowest CSF 5-HIAA concentrations were seen in homozygotes for the risk yin haplotype. This provides evidence of genetic determination of serotonin turnover in healthy individuals whose serotonin function is not greatly perturbed by environmental factors. In controls, each copy of the vulnerability haplotype may lower the CSF 5-HIAA concentration by an average of 16% to 17% (Figure 2). As with association with clinical phenotype, this is only a modest effect size, which could ultimately mean that the functional allele represented by the haplotype is uncommon or itself has only a modest effect. The lack of association of the THP2 haplotype with 5-HIAA levels in suicidal individuals is consistent with the effect of various environmental factors on serotonin function and 5-HIAA levels. These environmental effects include alcohol exposure,9,10 stress,11 diet,12 and, in contexts other than the present study, selective serotonin reuptake inhibitor drugs.13
A TPH2 functional locus is unknown. It is not the Pro206Ser variant we found because Ser206 was observed only in 2 siblings. The linkage signal also does not derive from the functional Arg441His variant reported by Zhang et al31 because, as mentioned at the beginning of this article, that variant is also rare, being absent in approximately 779 sequenced individuals, including 403 patients with depression.32 To identify the functional locus, it will be important to closely evaluate the significance of sequence variants in noncoding regions that could alter gene expression at the DNA or RNA levels. A guide to this discovery process can be the study of alleles specifically associated with the yin vulnerability and the yang protective haplotypes reported herein.
Correspondence: David Goldman, MD, Laboratory of Neurogenetics, National Institute on Alcohol Abuse and Alcoholism, National Institutes of Health, 5625 Fishers Ln, Room 3S32, Rockville, MD 20852 (email@example.com).
Submitted for Publication: July 21, 2004; final revision received March 10, 2005; accepted April 30, 2005.