(TAAR6) gene is associated with schizophrenia in the Irish stu - Nature

2 downloads 0 Views 292KB Size Report
May 15, 2007 - Trace amines (TAs) are endogenous compounds ..... Figure 1 In ISHDSF sample, the LD map represents a region of 108 kb covering the TRAR ...
Molecular Psychiatry (2007) 12, 842–853 & 2007 Nature Publishing Group All rights reserved 1359-4184/07 $30.00 www.nature.com/mp

ORIGINAL ARTICLE

A region of 35 kb containing the trace amine associate receptor 6 (TAAR6) gene is associated with schizophrenia in the Irish study of high-density schizophrenia families V Vladimirov1, DL Thiselton1, P-H Kuo1, J McClay1, A Fanous1,2,3, B Wormley1, J Vittum1, R Ribble1, B Moher1, E van den Oord1, FA O’Neill4, D Walsh5, KS Kendler1 and BP Riley1 1

Department of Psychiatry, Virginia Institute for Psychiatric and Behavioral Genetics, Virginia Commonwealth University, Richmond, VA, USA; 2Mental Health Service Line, Washington VA Medical Center, Washington, DC, USA; 3Georgetown University School of Medicine, Washington, DC, USA; 4Department of Psychiatry, Queens University, Belfast, Northern Ireland and 5Health Research Board, Dublin, Ireland The TAAR6 gene has been previously associated with schizophrenia in 192 pedigrees of European and African ancestry. To replicate these findings we performed an association study of TAAR6 in 265 pedigrees of the Irish Study of High-Density Schizophrenia Families (ISHDSF). Of the 24 genotyped single-nucleotide polymorphisms only rs12189813 and rs9389011 provided single-marker evidence for association (0.0094pPp0.03). Two-marker haplotypes (rs7772821 and rs12189813; 0.0071pPp0.0023) and four-marker haplotypes (rs8192622, rs7772821, rs12189813 and rs9389011; 0.0047pPp0.018) gave strongest evidence for association. The associated high-risk (HR) haplotype in the ISHDSF is defined by the major alleles at rs7772821 and rs12189813 (0.00097pPp0.023). The associated HR remains positive in a case only test of association by Operational Criteria score analysis in which significant association was observed only with the highest threshold for delusions (P < 0.009). When analysis was restricted to affected individuals under the core schizophrenia (D2) diagnosis, the observed associations for HR were most significant in the highest threshold for delusions (P < 0.004) and hallucinations (P < 0.0004), supporting the family-based association with schizophrenia. Molecular Psychiatry (2007) 12, 842–853; doi:10.1038/sj.mp.4001984; published online 15 May 2007 Keywords: schizophrenia; association; trace amine receptor; gene; haplotype; expression

Introduction Schizophrenia (MIM 181500) is a common and debilitating psychiatric disorder with a lifetime prevalence of 0.5–1% worldwide. Symptoms of the disease include delusions and hallucinations (psychotic symptoms), disordered thought, blunted emotion and impairment of memory, attention, cognition and executive function. Family, twin and adoption studies show consistent evidence of a substantial genetic component in schizophrenia, in conjunction with environmental and developmental factors.1–3 Using single-nucleotide polymorphism (SNP)-based association fine mapping, the trace amine-associated receptor 6 (TAAR6/TRAR4) gene on chromosome 6q23.2 was implicated as a schizophrenia susceptibility gene in a sample of 192 pedigrees (144 families Correspondence: Dr V Vladimirov, Virginia Institute for Psychiatric and Behavioral Genetics, Virginia Commonwealth University, Biotech1/Suite110, 800 East Leigh Street, Richmond, VA 232980424, USA. E-mail: [email protected] Received 5 January 2006; revised 4 February 2007; accepted 27 February 2007; published online 15 May 2007

of European and 48 families of African descent).4 After correction for multiple testing, rs4305745 (located in the 30 UTR of TAAR6) was the only marker to remain significantly associated with the disease. Multiple-marker haplotype analyses gave less significant results than the analysis of rs4305745 alone. The authors suggest that rs4305745 is thus the polymorphism most likely to account for the association of TAAR6 with schizophrenia. Trace amines (TAs) are endogenous compounds structurally related to classical biogenic amines such as p-tyramine, b-phenylethylamine, tryptamine and octapamine. TAs have been implicated in highly prevalent psychiatric disorders like depression and schizophrenia.5–8 They are present in the central nervous system (CNS) at low levels and they share significant overlap with classical biogenic amines in their chemical properties, biosynthesis and breakdown. The absence of specialized TA receptor(s) fostered the assumption that TAs would function via classical biogenic amine receptors.9 This view changed with the identification of a novel family of TA-sensitive receptors belonging to G-protein coupled receptors gene superfamily.10,11

TAAR6 association with schizophrenia in ISHDSF V Vladimirov et al

Six functional and three pseudo TA receptor genes have been identified in humans, all localized on chromosome 6q23.2. With the exception of TAAR2, which is encoded by two exons, the coding sequence of TAAR genes is contained in one exon of approximately 1 kb, with very high homology between the mouse, chimpanzee and rat TAAR orthologues.5 The expression of the best studied TAAR1 gene has been detected in different tissues at very low levels. In brain the overall expression of TAAR1 was also low, although some regions showed higher mRNA abundance compared with others. The highest TAAR6 expression was detected in hippocampus, whereas the lowest expression was seen in basal ganglia. Other brain regions such as frontal cortex, substantia nigra or amygdala showed comparable levels of TAAR6 expression and these regions have been implicated in the pathophysiology of schizophrenia.4,12 Here we report the results of a family-based association study performed in the Irish Study of High-Density Schizophrenia Families (ISHDSF) sample and present evidence for association between schizophrenia and a haplotype of 35 kb overlapping TAAR6 in the ISHDSF.

Materials and methods Subjects and phenotypes The ISHDSF sample consists of 265 high-density schizophrenia families with 1408 people available for genotyping.13 All participating individuals gave appropriate informed consent for the study. The sample was assessed using multiple concentric diagnostic categories, ranging from narrow to very broad. Briefly, D2 (core schizophrenia, 625 affected individuals) includes schizophrenia, poor-outcome schizoaffective disorder and simple schizophrenia. D5 (narrow psychosis spectrum, 804 affected individuals) adds to these schizotypal personality disorder and all other nonaffective psychotic disorders. D8 (broad psychosis spectrum, 888 affected individuals) additionally includes mood-incongruent and mood-congruent psychotic affective illness, and paranoid, avoidant and schizoid personality disorders. D9 (very broad, 1172 affected individuals) includes all psychiatric disorder (i.e. all psychosis spectrum disorders plus nonpsychotic affective disorders, anxiety, alcoholism, etc.). All patients with a history of any psychotic episode were assessed for lifetime psychotic symptoms using the Operational Criteria (OPCRIT) checklist of psychotic illness.14 The Stanley Brain Series (SBS) post-mortem collection donated by the Stanley Medical Research Institute consists of DNA and mRNA samples extracted from dorsolateral prefrontal cortex (Brodmann’s area 46) from 35 individuals each with schizophrenia and bipolar affective disorder and 35 controls. The exclusion criteria for collection of the samples included (1) significant structural brain pathology, (2) history of pre-existing CNS disease, (3) poor RNA quality or (4) documented IQ < 70. Additional exclu-

sion criteria for controls were (1) age less than 30 and (2) substance abuse within 1 year of death.15

843

SNP selection, genotyping and error checking At the beginning of this study, few SNPs in the TAAR6 region had been deposited in HAPMAP. We therefore chose the most significant SNPs reported in the original study (rs8192625, rs4305745, rs6903874 and rs6937506) and additionally selected rs8192622, rs8192624 and rs7772821, which were the only additional TAAR6 polymorphisms available in HAPMAP to extend coverage of the gene. rs8192623 was monomorphic and was excluded. TAAR6 belongs to a gene family of highly homologous members all located in a small genomic region of 108 kb. To exclude the possibility of association signal from another member of this gene cluster, we included 17 additional SNPs from the region either within or flanking additional TAAR loci, based on the HapMap data (http://www.hapmap.org) available at the time. Since the TAGGER option was not implemented in the earlier versions of Haploview, haplotype tagging SNPs were selected within each LD block to discriminate haplotypes with frequency X1% using the default HAPLOVIEW settings. The DNA samples were genotyped by fluorescence polarization template-directed incorporation of dyeterminator using the relevant AcycloPrime FP SNP detection kit (Perkin-Elmer Life Sciences, Boston, MA, USA), according to manufacturer’s instructions. The FP data were converted to genotypes by an automated allele scoring platform.16 Polymerase chain reaction (PCR) primers were designed using PRIMER3.17 PCR and single-base extension primer sequences for markers in this study are given in the Supplementary Table. The average completion rate of the genotype experiments was 94%. Mendelian inconsistencies within families were identified with PEDCHECK v1.1 (20) and unlikely recombinants were identified with MERLIN.18 For all Mendelian inconsistencies and unlikely recombinants, marker genotypes were zeroed for the entire pedigree. Hardy–Weinberg equilibrium was assessed for all markers using HAPLOVIEW v3.2 and no significant deviations were observed.19 The schizophrenic (n = 35) cases and controls (n = 35) from the SBS series were genotyped for four markers in the same methods as for the ISHSDF. Intermarker linkage disequilibrium analysis Linkage disequilibrium (LD) between SNPs was estimated with Haploview from the genotypes of unrelated founders. Haplotype frequencies were estimated from genotype data using the expectation-maximization algorithm.20 The LD blocks based on confidence intervals (D0 ) were determined by the default settings of Haploview.21 Association analysis We analyzed genotypic data using the Family-Based Association Test as implemented in the program Molecular Psychiatry

TAAR6 association with schizophrenia in ISHDSF V Vladimirov et al

844

FBAT version 1.1.5 and the Pedigree Disequilibrium Test as implemented in the program PDTPHASE version 4.0.22–25 Both are able to deal with the transmission of multilocus haplotypes, even when the phase is unknown and parental genotypes may be missing. Although the approaches are broadly similar, we include both for several reasons. First, FBAT was the only analytic tool used in the original study, and second PDTPHASE reports the components of the association from vertical parental transmissions and horizontal sibling genotype concordance or discordance separately. Since the directions of these two components of the test are not always the same, it is important to be able to assess them separately. In FBAT analyses, we did not use the empirical-variance estimator (-e flag) since we had no prior evidence of linkage to this region in our sample and for the analysis we assumed the additive model. Analyses of OPCRIT factor scores were performed using TRANSMIT. The robust variance estimate was used to allow for multiple-affected offspring per pedigree.26,27 Quantitative real-time PCR Expression of the TAAR genes was analyzed by quantitative real-time PCR using SYBR Green I on a Bio-Rad (Hercules, CA, USA) iCycler platform and normalized against two reference genes TBP and GAPDH (used as an internal controls) by the 2DDCT algorithm implemented in the Relative Gene Expression Macro software (Bio-Rad Inc.). The inclusion of multiple housekeeping genes as internal controls has been reported to give more accurate and reliable normalization of gene-expression data.28,29 Reverse transcription of mRNA samples into cDNA was performed using the ready-to-go kit from Amersham Bioscience (Piscataway, NJ, USA) following the manufacturer’s instructions. The primers for TAAR6 TBP and GAPDH were designed using the Beacon Designer software version 4.02. Primer sequences used in this experiment are shown in the Supplementary Table. TAAR2, TAAR5 and TAAR6 primer sets optimized for SYBR Green I-based real-time PCR were obtained from SuperArray (Frederick, MD, USA). To generate the standard curves, the cDNAs for these TAAR genes, TBP and GAPDH were cloned in the pcDNA3.1 expression vector. The concentration of such prepared vectors was determined by measuring their absorbance at 260 nm. The quality of the DNA sample was verified by the A260/A280 ratio, which was between 1.7 and 1.9. The standard points were made using five 1:5 serial dilutions and the quality of the standard curve was judged from the slope and the correlation coefficient (r). A correlation coefficient r > 0.99 (showing that a linear relationship is observed) was obtained for all loci. The PCR efficiency ranged from 98 to 102%. Each point on the standard curve was assayed in triplicate.

Molecular Psychiatry

Statistical analysis of gene expression Differences in TAAR gene expression level between cases and controls were evaluated by the nonparametric Kruskal–Wallis test implemented in Prism v4.0 (GraphPad v4.03 for Windows, San Diego, CA, USA; www.graphpad.com). Effects of potential confounder variables (age, post-mortem interval (PMI), refrigeration interval, brain pH and smoking) were assessed by multiple regression analysis implemented in analysis of variance (ANOVA) using NCSS 2004 (Number Crunching Statistical System, Kaysville, UT, USA; www.ncss.com). Haplotype reconstruction Haplotypes were reconstructed in the SBS samples using the PHASE v2.1.30,31 The program implements a Bayesian statistical method and for haplotype reconstruction PHASE uses an iterative approach. The number of iterations required to obtain accurate answer depends on the complexity and size of the data. For the haplotype reconstruction in the SBS we have used 100 iterations. Bioinformatics tools We used VISTA (http://genome.lbl.gov/vista/index. shtml) to define conserved regions in genomic sequences from different species.32 To predict secondary structure of TAAR6 mRNA we used RNAstructure version 4.2, available through Isis Pharmaceuticals Inc., Carlsbad, CA, USA (http://rna.urmc.rochester. edu/rnastructure.html). These predictions are based on the algorithm for free energy minimization using the nearest neighbor parameters.33,34 Default values set (1) the lowest free energy structure at 100 kcal/mol (DG137) and (2) the maximum energy difference at 10%. The graphical visualization of the RNA secondary structures was performed by XRNA software version 1.1.12Beta available through University of California, Santa Cruz, CA, USA (http://rna.ucsc.edu/ rnacenter/). To find potential target sites for miRNAs in the TAAR6 30 UTR sequence we used the miRanda package (http://www.microrna.org/miranda.html).35 Potential target sites are identified using a two-step strategy. First a dynamic programing local alignment is carried out between the query miRNA sequence and the reference sequence. The miRNAs sequences were retrieved from the miRNA registry.36 This alignment procedure scores is based on sequence complementarity rather than sequence identity. The second phase of the algorithm takes high-scoring alignments (those above a user-defined score threshold) detected in phase 1 and estimates the thermodynamic stability of RNA duplexes based on these alignments. Optionally some rudimentary statistics about each target site can be generated by performing a number of alignments using shuffled reference sequences generated from the input sequence. A distribution is built from these data and statistical parameters from this distribution are used to produce a Z score for a detected target site. A Z score of X5

TAAR6 association with schizophrenia in ISHDSF V Vladimirov et al

was used as cutoff value to screen against false positive matches.

Results LD structure of TAAR region and SNP selection In our sample of pedigrees from Ireland, the haplotypic structure of the region is defined by four LD blocks (using the default confidence interval setting in HAPLOVIEW). Haplotype and LD structure with r2 values of the region in our sample are shown in Figure 1. TAAR9 and TAAR8 are located in the first (most centromeric) LD block, TAAR6 between LD blocks 1 and 2, TAAR5 and TAAR2 between LD blocks 2 and 3 and TAAR1 in LD block 4. These results are similar to the regional data from HapMap,

which also identifies these LD blocks interspersed with regions of very low LD.

845

Single-marker association results Results of single-marker analyses are presented in Table 1. The single-marker association study was performed across D2, D5, D8 and D9 diagnostic categories. Of the 24 SNPs screened, only two showed significant evidence for association, rs12189813 (0.0094pPp0.02), and rs9389011 (0.03pPp0.1597) in the narrowest D2 category. These two markers are located 35 and 42 kb telomeric from TAAR6, respectively, and fall between LD block 2 and 3. The observed associations were due to excess transmission and/or sharing of the common allele at both markers. Of the four SNP variants in the TAAR6 gene, none gave significant evidence of association individually in any

Figure 1 In ISHDSF sample, the LD map represents a region of 108 kb covering the TRAR genes on 6q23.2. Four LD blocks (D0 confidence interval (CI)) are defined, with block 2 showing extended LD with markers rs12189813 and rs9389011. Markers rs8192622, rs8192624, rs8192625 and rs7772821 localized in the TAAR6 gene fall between LD block 1 and 2. The LD map color intensity is based on D0 with darker shades representing higher LD. Values in boxes represent the pair-wise r2 measure between markers. Molecular Psychiatry

TAAR6 association with schizophrenia in ISHDSF V Vladimirov et al

846

Table 1 Single-marker analysis provided association evidence for two markers, rs12189813 and rs9389011 (significant P-values below nominal 0.05 are given in bold), using a narrow diagnostic model Marker dbSNP

Position (bp)

Distance SNP MAF (bp) nucleotide

FBAT

PDT Transmissions

Observed Expected P-value Trio-T Trio-NT Affsib Unafsib P-value 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

rs2842899 rs2153174 rs1933988 rs2255071 rs1361280 rs8192622 rs8192624 rs8192625 rs7772821 rs4305745 rs6903874 rs7765655 rs6937506 rs7744878 rs7759367 rs17195227 rs3813354 rs17061477 rs4320395 rs12189813 rs9389011 rs9389015 rs9389020 rs9402439

132 840 179 132 842 919 132 852 223 132 861 218 132 865 906 132 872 108 132 872 823 132 872 902 132 873 076 132 874 282 132 877 480 132 878 428 132 879 969 132 882 090 132 884 315 132 889 521 132 891 204 132 891 840 132 900 518 132 908 072 132 915 149 132 924 006 132 943 623 132 948 172

0 2740 9304 8995 4688 6202 715 79 174 1206 3198 948 1541 2121 2225 5206 1683 636 8678 7554 7077 8857 19 617 4549

T>A G>T A>C C>T T>C C>T G>A G>A T>G A>G T>C G>A G>A G>A G>A T>C G>A A>C G>A G>C T>C C>T T>C C>G

0.280 0.482 0.401 0.340 0.406 0.052 0.073 0.068 0.122 0.455 0.167 0.267 0.177 0.183 0.445 0.355 0.098 0.100 0.460 0.408 0.229 0.466 0.280 0.406

202 184 214 201 182 82 118 95 158 164 122 173 113 146 164 228 77 82 187 184 247 201 247 175

201.53 186.12 209.57 204.77 182.1 79.92 113.48 96.82 152.05 171.33 129.51 178.53 119.04 156.32 159.97 217.65 73.65 79.42 178.56 168.88 233.16 207.53 242.41 176.98

0.94 0.75 0.51 0.56 0.99 0.52 0.24 0.61 0.22 0.22 0.12 0.32 0.16 0.06 0.50 0.13 0.31 0.45 0.21 0.02 0.03 0.35 0.47 0.78

722 498 595 648 506 985 954 939 834 465 787 612 763 783 459 662 895 926 514 532 803 520 760 557

724.6 488.1 575.2 633.2 513.8 976.9 936.4 928.6 825.9 474.8 783.1 638.7 758.6 782.8 476.5 623.7 888.1 921.1 488.5 495.4 789.4 525.4 730 572.6

896 632 758 800 609 1328 1270 1199 1053 480 939 678 914 948 537 866 1148 1203 723 645 1021 716 985 725

933 608 728 794 604 1323 1256 1215 1015 514 963 724 929 963 537 844 1135 1186 686 592 996 729 969 727

0.2852 0.5723 0.2953 0.5342 0.9480 0.4072 0.2503 0.9210 0.0751 0.1562 0.7121 0.1350 0.3771 0.8617 0.2175 0.1253 0.7337 0.8290 0.1566 0.0094 0.1597 0.8803 0.05437 0.2773

Both markers are located 35 and 42 kb telomeric from the TAAR6 gene. Of the four TAAR6 SNP variants (M6, M7, M8 and M9) only rs7772821 showed trend toward association.

diagnostic category. SNP rs7772821 showed a trend toward association only in D2 (0.075pPp0.22). Haplotype analyses We initially analyzed haplotypes composed of SNPs within the four haplotype blocks (Block 1: M1-M5; Block 2: M11-M16; Block 3 and 4: M21-M24), those defining specific genes not located within blocks (TAAR6: M6-M9; TAAR5: M16-M18; TAAR2: M21M22) and those defining one small unit of high LD not identified as a block by default settings in HAPLOVIEW (M19-M20). These analyses were conducted in sliding windows using 2–5 contiguous markers across all categories. No contiguous-marker haplotype combination was significantly associated with schizophrenia in any diagnostic category (data not shown), with the exception of the two-marker haplotype composed of the two individually associated markers, rs12189813 (M20) and rs9389011 (M21) (0.0094 pGlobal Pp0.01, Table 2). Consistent with the single-marker analysis, the specific haplotype overrepresented in affected family members was composed of the common alleles at both markers (0.012pPp0.049). Since rs12189813 and rs9389011 were the only polymorphisms to show evidence for association at Molecular Psychiatry

the single-marker level, we used either one or both as anchor markers in combination with other specific SNPs (Table 2). All SNPs within TAAR6 in two-marker combinations with rs12189813 (M20) gave evidence of association. The most significant of these, and the most significant two-marker haplotype experimentwide was for rs7772821 (M9) and rs12189813 (M20) (0.0023pGlobal Pp0.0071). Again, the specific overrepresented haplotype was composed of the common alleles at both markers (0.00097pPp0.023). A similar pattern was observed in a three-marker haplotype test of rs8192622 (M6), rs7772821 (M9) and rs12189813 (M20), and a four-marker haplotype test of rs8192622 (M6), rs7772821 (M9), rs12189813 (M20) and rs9389011 (M21), which gave the second most significant association experiment-wide at the global (0.0047pGlobal Pp0.018) level with the overrepresented haplotype composed of the common alleles (0.0016pPp0.0045). The associated haplotypes were always composed of the major allele at each marker. These associations were all observed in the core schizophrenia (D2) diagnostic group. When these haplotypes were analyzed for the broader D5, D8 and D9 categories, there was no evidence for association. These results are consistent with the observed LD structure of the region (Figure 1), showing moderate

TAAR6 association with schizophrenia in ISHDSF V Vladimirov et al

Table 2 Haplotype analysis of the core markers with rs12189813 and rs9389011 revealed positive association with schizophrenia in the ISHDSF FBAT v1.5.5

847

PDTPHASE v5.1 D2

Window size

Marker

SM

M20 M21 M6/M20 M7/M20 M8/M20 M9/M20 M9/M21 M20/M21 M6/M9/M20 M6/M9/M21 M7/M9/M20 M8/M9/M20 M9/M20/M21 M6/M7/M8/M9 M6/M7/M9/M20 M7/M8/M9/M20 M6/M9/M20/M21 M6/M7/M8/ M9/M20 M6/M7/M8/M9/ M20/21

2

3

4

5 6

M6 M7 M8 M9 M20

rs8192622 rs8192624 rs8192625 rs7772821 rs12189813

M21

rs9389011

HT

HT Freq

HT Global P-value P

1 1 1–1 1–1 1–1 1–1 1–1 1–1 1–1–1 1–1–1 1–1–1 1–1–1 1–1–1 1–1–1–1 1–1–1–1 1–1–1–1 1–1–1–1 1–1–1–1–1

0.58 0.80 0.54 0.52 0.54 0.46 0.68 0.46 0.45 0.67 0.44 0.45 0.35 0.81 0.44 0.43 0.35 0.44

0.016 0.028 0.022 0.018 0.1 0.023 0.013 0.049 0.014 0.021 0.013 0.017 0.011 0.242 0.016 0.017 0.0016 0.023

0.016 0.028 0.058 0.032 0.061 0.0071 0.021 0.010 0.064 0.008 0.060 0.034 0.006 0.660 0.049 0.044 0.018 0.061

1–1–1–1–1–1

0.37

0.006

0.031

Trio-T

Trio-NT

AffSib

UnafSib

HT P-value

Global P

532 803 591 574 584 509 722 496 509 712 491 504 388 869 493 491 380 492

495 789 548 530 545 467 705 465 465 696 446 461 358 847 450 442 354 445

645 1021 843 805 805 721 967 667 743 963 706 730 549 1197 710 711 547 711

592 966 790 760 770 645 908 637 671 904 644 669 493 1173 648 647 490 649

0.0094 0.1597 0.0052 0.010 0.020 0.00097 0.02735 0.012 0.00090 0.04513 0.0023 0.0016 0.0020 0.2331 0.0020 0.0012 0.0045 0.0013

0.0094 0.1597 0.010 0.0080 0.024 0.0023 0.05722 0.0084 0.0054 0.163 0.032 0.010 0.0023 0.5023 0.014 0.0091 0.0047 0.0070

TAAR6 TAAR6 TAAR6 TAAR6 30 UTR 35 kb from the TAAR6 gene 42 kb from the TAAR6 gene

Abbreviations: ISHDSF, Irish study of High–Density Schizophrenia Families. The strongest association is represented by the most abundant 1–1 alleles for two-marker haplotype defined by rs7772821 and rs12189813. The second most significant association was found for the 1–1–1–1 four-marker haplotype, defined by rs8192622 and rs8192624 in the TAAR6 gene and rs12189813 and rs9389011 markers. The P-values for two and four marker haplotypes are in typed bold.

levels of LD between rs7772821 (M9) and either rs12189813 (M20) or rs9389011 (M21). The sequential addition of further markers to the sets yielding the results described above did not improve the evidence for association (data not shown). This region also contains the gene for a second-trace amine receptor, TAAR5, and is just centromeric to a third, TAAR2. None of the single-marker or haplotype analyses anchored to either TAAR5 or TAAR2, showed any evidence for association (data not shown), suggesting that effects in these genes are unlikely to account for the observed association with schizophrenia. Gene expression analysis We compared the mRNA levels for TAAR6, TAAR5 and TAAR2 between subjects with schizophrenia, bipolar disorder and unaffected controls by performing a quantitative real-time PCR. GAPDH and TBP

were also quantified in each sample to control for variation in mRNA extraction and reverse transcription. Expression of TAAR6, TAAR5 and TAAR2 were normalized against these two endogenous controls. When analyzed by one-way ANOVA, no significant difference in the mRNA levels between diagnostic groups was observed for TAAR6 (P = 0.4673), TAAR5 (P = 0.2831) or TAAR2 (P = 0.9731; Figure 2). The ANOVA analysis of potential confounder variables did not show any significant effect of age, pH, PMI, refrigerator interval, or smoking on the expression of TAAR6 (F = 1.919; d.f. = 5, P = 0.1008), TAAR2 (F = 0.700; d.f. = 5, P = 0.6249) or TAAR5 (F = 0.589; d.f. = 5, P = 0.7084). Haplotype expression analyses We stratified TAAR6 expression based on the highrisk (HR) haplotype identified in ISDFSH sample. Molecular Psychiatry

TAAR6 association with schizophrenia in ISHDSF V Vladimirov et al

848

HR haplotype stratification regardless of disease status. The stratified analysis did not show any significant difference between HR þ / þ , HR þ / and HR/ (P = 0.1984, Figure 3a). In the second analysis each diagnostic group was analyzed separately. Again, no significant differences were observed for schizophrenic and bipolar cases and controls (P = 0.4548, 0.2013, 0.6056, respectively; Figure 3b). In both analyses no significant interaction between diagnosis and genotypes for rs7772821 and rs12189813 was detected (F = 0.575; d.f. = 30; P = 0.9506).

Figure 2 Expression of TAAR6, TAAR5 and TAAR2 in schizophrenic and bipolar cases and controls. The y-axis values represent the ratio of the normalized TAAR6, TAAR5 and TAAR2 gene expression to endogenous controls. The error bars reflects the SE with 95% CI.

Cases and controls in SBS sample were genotyped for rs7772821 and rs12189813 together with two additional markers to provide sufficient information for haplotype reconstruction using PHASE. Reconstructed haplotypes with common alleles at rs7772821 and rs12189813 were defined as high-risk positive (HR þ ), and all others defined as high-risk negative (HR). Both, cases and control samples were then assigned to HR homozygous (HR þ / þ ), heterozygous (HR þ /) and non-HR (HR/) diplotype groups. We performed two analyses. In the first analysis the TAAR6 expression was evaluated based on Molecular Psychiatry

Operational criteria checklist for psychotic illness We obtain results similar to the family-based association studies (in which association was observed only in the narrow D2 diagnostic group) when analyzing the SNP data in a case-only test of association with subgroups defined by a priori thresholds of factor-derived scores derived from symptom factors extracted from the OPCRIT (Table 3).37 In our previous work, we showed that the best fit was obtained with a five factor solution composed of individual factors representing negative symptoms, delusions, hallucinations, as well as manic and depressive symptoms. For each factor-derived scale, subjects scoring in the upper 20%, 30% and 40% were selected and analyzed against the remaining 80, 70 and 60% for association testing. In the OPCRIT factor score analyses we focused on the two-marker haplotype defined by the overtransmited major alleles at rs7772821 and rs12189813 since this haplotype showed the strongest observed association. In the first analysis we analyzed all cases with OPCRIT data. We observe only one significant association, with the highest (top 20%) threshold for delusions (P < 0.009). There were trends approaching significance for hallucinations and mania. No significant association was observed in the broader 30 and 40% thresholds for any factor scale. The second analysis was restricted to subjects affected under the core schizophrenia (D2) diagnosis. Again, the observed associations are most significant when using the highest (top 20%) threshold for delusions (P < 0.004) and hallucinations (P < 0.0004). These associations are also observed at the broader top 30% and top 40% thresholds, but are of lesser magnitude. We also observe some evidence for association with the depression factor scale, although this is difficult to interpret as it is less significant for the top 20% threshold than for the top 30% threshold. Nevertheless, these analyses are consistent in showing evidence for association with schizophrenia only in our narrowest and most severely affected diagnostic group. Control for multiple testing While unexpected behavior of the test statistic due to characteristics of our data (patterns of missing data, etc.) is accounted for by permutation tests within the FBAT program, we sought to address the issues of multiple testing by performing a separate set of Monte

TAAR6 association with schizophrenia in ISHDSF V Vladimirov et al

genotype data are maintained as part of this simulation. In each replicate of simulated data, all of the FBAT haplotype analyses performed in the observed data were repeated. These results were used to construct a null distribution of the smallest P-values generated for all global tests from each of the replicates. We believe that the values resulting from Monte Carlo simulations of global tests are more reliable because they exclude P-values generated from very rare haplotypes, which are more likely to generate spuriously small P-values. Since the global tests tend to be more stable, we focused on the question, what is the probability that we would find as the smallest global test P-value one equivalent to or smaller than the one we observed in our data when conducting this number of tests (NB1200 from 24 SNPs in four diagnostic categories with both sliding window and marker-anchored haplotype analyses). We generated 1000 replicate data sets for the purpose of assessing the significance of our most significant global test (P = 0.0071). For each set of permutation tests, significance is calculated by (r þ 1)/ (n þ 1), where n is the total number of Monte Carlo procedure tests conducted and r is the number of those more significant than the most significant observed test from the actual data. As a diagnostic assessment of the performance of the test statistic as a whole, we compared it to the entire null distribution of all global tests conducted over all 1000 replicates. We find an empirical P-value of 0.0073, indicating that P-values less than or equal to our observed P-value of 0.0071 occur with a frequency of 0.0073 in our permutations. This shows that the empirical P-values from the Monte Carlo procedure were similar to those generated from the permutation test in the FBAT package was expected and not the primary focus of the Monte Carlo procedure. In the comparison of our best-observed global P-value to the distribution of per replicate best global test P-values, we generated an empirical P-value of 0.224. Thus, within the assessment of global tests, we have a 22.4% chance of randomly generating a most significant global result of the magnitude of our best-observed global test when performing this total number of tests.

Figure 3 (a) TAAR6 expression evaluated based on haplotype stratification regardless of disease status. (b) The TAAR6 gene expression for three haplotypes (HR þ / þ , HR þ / and HR/) is evaluated in each diagnostic group. The error bars reflect the s.e. with 95% CI.

Carlo simulations. The simulations were first performed in the pedigrees using Merlin to simulate haplotype data and perform gene dropping in the pedigrees.18 Importantly, the LD structure, phenotypes, family structures and pattern of missing

849

In silico analysis Within the HR defined by rs7772821 and rs12189813 we defined the conserved non-coding sequence by comparative genomic analysis of TAAR6 between human, mouse and rat using VISTA. We found that the homology between the coding regions of the TAAR6 orthologs is about 80%. rs7772821 (M9) is located in the 30 UTR of TAAR6 eight base pairs from the stop codon. Homology is generally lower in the 30 UTR, dropping below 50% at the 30 end, but is similar to that for the coding sequence in the specific area containing this polymorphism. The closest conserved region (with homology of over 70% between human, mouse and rat genomic sequences) is located 0.4 kb downstream from rs7772821 (M9) in the Molecular Psychiatry

TAAR6 association with schizophrenia in ISHDSF V Vladimirov et al

850

Table 3 The two-marker haplotype 1–1 represented by rs7772821 and rs12189813 was further analyzed for an association with the five OPCRIT symptom dimensions Threshold (%)

Negative symptoms

Delusions

Hallucinations

Mania

Depression

All individuals with psychosis 20 0.567 (141) 30 0.221 (181) 40 0.071* (227)

0.009*** (119) 0.122 (176) 0.285 (208)

0.075*a (187) 0.075*a (187) 0.135 (261)

0.089* (110) 0.456 (169) 0.819 (224)

0.932 (103) 0.995 (144) 0.960 (190)

Individuals with D2 narrow diagnostic 20 0.491a (139) 30 0.491a (139) 40 0.166 (173)

0.004*** (113) 0.037** (141) 0.047** (166)

0.0004*** (107) 0.012**a (177) 0.012**a (177)

0.733 (99) 0.865 (136) 0.184 (188)

0.035** (93) 0.007** (131) 0.226 (167)

Abbreviation: OPCRIT, operational criteria. The vertical transmissions of the haplotype from parents to affected offspring are given for each psychosis symptom scale. Number of families with transmission to affected offspring is shown in parenthesis. Significant P-values are highlighted in bold using * < 0.1, ** < 0.05 and *** < 0.001. Affected individuals with psychosis (n = 749) and individuals limited to the core schizophrenia diagnosis (n = 614) were included in the analysis. a Same cutoff threshold for each level.

TAAR6

TAAR5 100%

intergene intron coding UTR repeat

rs7772821

/Sequence/rn - vs /Sequence/hg 50%

0

9.6kb

19.1kb

TAAR6

28.6kb

38.2kb

TAAR5 100%

intergene intron coding UTR repeat

rs12189813

/Sequence/rn - vs /Sequence/hg 50%

0

9.6kb

19.1kb

28.6kb

38.2kb

-100% intergene intron coding UTR repeat

/Sequence/rn - vs /Sequence/hg 50% 38.2kb

Figure 4 The conserved non-coding regions are defined by VISTA,32 and rs7772821 and rs12189813 positions are shown by the arrows. The conserved regions are generated by comparing human genome with mouse and rat. The region with sequence similarity reaching 70% is defined as a conserved region, and such peak areas are shaded. The darkest shaded areas represent the TAAR6 and TAAR5 genes.

TAAR6 30 UTR. Marker rs12189813 (M20) did not lie in a close proximity to an evolutionarily conserved region (Figure 4). To evaluate any potential effect rs7772821 could have, we analyzed the secondary structure of the TAAR6 mRNA in silico by the algorithm for free energy minimization. The T and G alleles of rs7772821 are predicted to have a pronounced effect on the folding of the mRNA affecting its conformation and stability38,39 (Figure 5). The secondary structure of the TAAR6 mRNA 30 UTR is also important since Molecular Psychiatry

sequences that are not folded could be available to pair with other regulatory factors (e.g. miRNAs). Indeed, after scanning the evolutionarily conserved region within the 30 UTR, four overlapping target sequences for four miRNAs (hsa-miR-92, hsa-miR125a, hsa-miR-302b and hsa-miR-483) were predicted. The observation of overlapping sequences recognized by several miRNAs is not surprising since it has been suggested that multiple miRNAs can act in combination by binding the same mRNA in concentration-dependent manner.40

TAAR6 association with schizophrenia in ISHDSF V Vladimirov et al

851

Figure 5 The RNA secondary structure was predicted with RNAstructure 4.233 for free energy minimization algorithm using the nearest neighbor parameters. A recursive algorithm is used that generates an optimal structure and a series of structures that are called suboptimal structures (structures with free energy similar to the lowest free energy structure). The number of suboptimal structures generated is controlled by the maximum number of structures (N = 20) and maximum % of energy difference (10 kcal), which is the percent difference from the lowest free energy allowed for the structures output. The arrows indicate the position of the rs7772821 polymorphism within circle.

Discussion TAAR genes have been proposed as candidates for schizophrenia based on the pathophysiological evidence as well as linkage mapping and association data.4,41,42 The original association study reported one SNP variant in the 30 UTR of the TAAR6 gene to be significantly associated with schizophrenia after correction for multiple testing. During the course of this work four independent studies attempted to replicate the original finding for association between TAAR6 and schizophrenia, of which three were negative and one positive. The first two negative replications were performed (in chronological order of their publications) in 235 family trios from the Han Chinese population and in two independent sets of Japanese case–control samples.43,44 The third study used 53 Arab Israeli families with previous highly significant evidence for linkage to 6q13–q26 in a putative susceptibility region of 3.90 Mb containing the TAAR6 gene. This study also failed to replicate the original association.45 All negative studies claimed to be adequately powered to detect association. So far, the only positive report found was a common two-marker haplotype (rs6907909, rs9373026, P < 0.001) in the promoter region of TAAR6 to be associated with schizophrenia only in Caucasian cases, and an additional study found TAAR6 to be associated with bipolar disorder in two independently analyzed samples from the German population.46,47 Since these studies were performed in different ethnic groups, a possible explanation for the non-

replication is differences either in the presence or the frequency of a liability allele between different populations. It is known that differences in the allele and haplotype frequencies can affect both the LD structure and strength of association. Thus, the highrisk allele might have a very different pattern of both LD and association across different populations and marker sets useful in one study may not be as useful in another.48,49 Across various analyses performed, the haplotype of rs7772821 (M9) and rs12189813 (M20) gave the strongest evidence for association with schizophrenia in the ISHDSF sample. Although separated by 35 kb, the two markers are in moderate LD with each other (D0 = 0.63). A haplotype of this size is not unusual, as reported the average haplotype size in the European population is about 22 kb.21 The location of rs7772821 in the 30 UTR and in the vicinity of an evolutionarily conserved region between mouse and human makes this an interesting candidate for further study. The predicted RNA secondary structure for the overtransmited rs7772821T allele differed substantially from the undertransmited G allele. Such polymorphisms may have functional effects on mRNA conformation or stability, and hence gene expression level.50 However, we observed no differences in mRNA abundance between clinical groups or haplotypes. Conversely, miRNAs interfere with translation, so their regulatory effect will only be detectable at the protein level. Using the miRanda package, target sequences for four miRNAs were identified in this region. One of these, hsa-miR-125a, Molecular Psychiatry

TAAR6 association with schizophrenia in ISHDSF V Vladimirov et al

852

was shown to be expressed in mouse cortex and cerebellum.51 It is therefore possible that protein-level expression differences mediated by miRNA might be observed. This is an area of growing importance, as recent reports predict that over one-third of all human genes might be regulated by miRNAs.52 The rapid decline in the strength of association as we moved away from the core schizophrenia diagnostic category suggests that in our sample, TAAR6 might be implicated as a liability factor only for schizophrenia although in a recent study from Germany, TAAR6 was implicated as a liability locus for bipolar disorder.46 Because of very limited number of bipolar cases both in our sample and in SBS we had insufficient power to test meaningfully for association between TAAR6 and this disease. Our analyses of clinical subgroups defined by thresholds on OPCRIT factors suggest that genetic variation in TAAR6 predisposes to a form of psychotic illness marked by high levels of delusions, hallucinations, and possibly depressive symptoms. As such, we would classify TAAR6 as a mixed susceptibilitymodifier gene, as described previously.53 To date, there have been two reports of susceptibility genes influencing clinical features of illness: DTNBP1 and negative symptoms of schizophrenia, as well as DAOA and persecutory delusions in bipolar disorder. Hallucinations and delusions are correlated with each other and have been traditionally categorized together as ‘positive’ symptoms. Although delusions are significantly heritable in this sample, hallucinations are not.53 These results are, of course, preliminary and need to be confirmed in other samples, but nevertheless they suggest that TAAR6 might explain part of the variance in both hallucinations and delusions.54 The comparison between human and chimpanzee genomic sequence reveals an interesting feature of the locus and the markers defining the associated haplotype. The ancestral haplotype for rs7772821 and rs12189813 appears to be G–C, but this is not the most common haplotype of these two markers observed in our sample. In our sample, the frequency of the associated T–G haplotype is 46%, whereas the frequency of the ancestral G–C haplotype is 1.6%. A similar shift in allele and haplotype frequency has been reported for other genes, where the ancestral haplotype has been supplanted by evolutionarily more recent haplotypes.55,56 Finally, the issue of multiple testing must also be considered. In our association analyses, multiple testing results from the assessment of multiple markers and multiple haplotypes in conjunction with multiple diagnostic categories. Both (1) diagnostic categories and (2) genotypes or haplotypes are highly correlated, the former because of the concentric nature of the multiple diagnoses, the latter because of LD. Since they are correlated, an appropriate specific correction is very difficult to define. A possible approach to address the issue of multiple testing is by performing a set of Monte Carlo simulations followed by a permutation test to assess distri-

Molecular Psychiatry

bution of the test statistics under the null hypothesis. We observed very good agreement between the empirical and actual P-values strengthening the robustness of our findings. Additionally, there is a higher prior probability of association because we are studying a region previously associated with the trait of interest. While we cannot be certain, we would suggest that these lines of evidence (prior association and our association) make it unlikely that our results reflect false positives. As always, replication of these findings will be necessary. In summary, our results show an associated haplotype overlapping the previously described haplotype and supporting the original study, and thus providing additional evidence of a role for TAAR6 as a susceptibility locus for schizophrenia.

Acknowledgments We thank the patients and families for their participation. This work was supported by NIH grants MH041953-13 and MH068881-02.

References 1 Jablensky A, Sartorius N, Ernberg G, Anker M, Korten A, Cooper JE et al. Schizophrenia: manifestations, incidence and course in different countries – a World Health Organization Ten Country Study. Psychol Med Monogr Suppl 1992; 20: 1–97. 2 Kendler KS, Gruenberg AM, Kinney DK. Independent diagnosis of adoptees and relatives as defined by DSM-III in the provincial and national samples of the Danish Adoption Study of Schizophrenia. Arch Gen Psychiatry 1994; 51: 456–468. 3 Kendler KS, Diehl SR. The genetics of schizophrenia: a current, genetic–epidemiologic perspective. Schizophr Bull 1993; 19: 261–285. 4 Duan J, Martinez M, Sanders AR, Hou C, Saitou N, Kitano T et al. Polymorphisms in the trace amine receptor 4 (TRAR4) gene on chromosome 6q23.2 are associated with susceptibility to schizophrenia. Am J Hum Genet 2004; 75: 624–638. 5 Lindemann L, Ebeling M, Kratochwil NA, Bunzow JR, Grandy DK, Hoener MC. Trace amine-associated receptors form structurally and functionally distinct subfamilies of novel G protein-coupled receptors. Genomics 2005; 85: 372–385. 6 Davis BA, Boulton BA. The trace amines and their acidic metabolites in depression – an overview. Prog Neuropsychopharmacol Biol Psychiatry 1994; 18: 17–45. 7 Sandler M, Ruthven CRJ, Goodwin BL, Coppen A. Decreased cerebrospinal fluid concentration of free phenylacetic acid in depressive illness. Clin Chim Acta 1979; 93: 169–171. 8 Potkin SG, Karoum F, Chuang LW, Cannon-Spoor HE, Phillips I, Wyatt RJ. Phenylethylamine in paranoid chronic schizophrenia. Science 1979; 206: 470–471. 9 Premont RT, Gainetdinov RR, Caron MG. Following the trace of elusive amines. Proc Natl Acad Sci USA 2001; 98: 9474–9475. 10 Borowsky B, Adham N, Jones KA, Raddatz R, Artymyshyn R, Ogozalek KL et al. Trace amines: identification of a family of mammalian G protein-coupled receptors. Proc Natl Acad Sci USA 2001; 98: 8966–8971. 11 Bunzow JR, Sonders MS, Arttamangkul S, Harrison LM, Zhang G, Quigley DI et al. Amphetamine, 3,4-methylenedioxymethamphetamine, lysergic acid diethylamide, and metabolites of the catecholamine neurotransmitters are agonists of a rat trace amine receptor. Mol Pharmacol 2001; 60: 1181–1188. 12 Grossberg S. The imbalanced brain: from normal behavior to schizophrenia. Biol Psychiatry 2000; 48: 81–98. 13 Kendler KS, O’Neill FA, Burke J, Murphy B, Duke F, Straub RE et al. Irish study on high-density schizophrenia families: field methods and power to detect linkage. Am J Med Genet 1996; 67: 179–190.

TAAR6 association with schizophrenia in ISHDSF V Vladimirov et al 14 McGuffin P, Farmer A, Harvey I. A polydiagnostic application of operational criteria in studies of psychotic illness. Development and reliability of the OPCRIT system. Arch Gen Psychiatry 1991; 48: 764–770. 15 Torrey EF, Webster M, Knable M, Johnston N, Yolken RH. The Stanley Foundation brain collection and Neuropathology Consortium. Schizoph Res 2000; 44: 151–155. 16 van den Oord EJCG, Jiang Y, Riley BP, Kendler KS, Chen X. FP-TDI SNP scoring by manual and statistical procedures: a study of error rates and types. Biotechniques 2003; 34: 610–624. 17 Rozen S, Skaletsky HJ. Primer3 on the WWW for general users and for biologist programmers. In: Krawetz S, Misener S (eds). Bioinformatics Methods and Protocols: Methods in Molecular Biology. Humana Press: Totowa, NJ, 2000, pp 365–386. 18 Abecasis GR, Cherny SS, Cookson WO, Cardon LR. Merlin – rapid analysis of dense genetic maps using sparse gene flow trees. Nat Genet 2002; 30: 97–101. 19 Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 2005; 21: 263–265. 20 Excoffier L, Slatkin M. Maximum-likelihood estimation of molecular haplotype frequencies in a diploid population. Mol Biol Evol 1995; 12: 921–927. 21 Gabriel SB, Schaffner SF, Nguyen H, Moore JM, Roy J, Blumenstiel B et al. The structure of haplotype blocks in the human genome. Science 2002; 296: 2225–2229. 22 Laird NM, Horvath S, Xu X. Implementing a unified approach to family-based tests of association. Genet Epidemiol Suppl 2000; 19: S36–S42. 23 Rabinowitz D, Laird N. An unified approach to adjusting association tests for population admixture with arbitrary pedigree structure and arbitrary missing marker information. Hum Hered 2000; 50: 211–223. 24 Martin ER, Monks SA, Warren LL, Kaplan NL. A test for linkage and association in general pedigrees: the pedigree disequilibrium test. Am J Hum Genet 2000; 67: 146–154. 25 Dudbridge F. Pedigree disequilibrium tests for multilocus haplotypes. Gen Epidemiology 2003; 25: 115–121. 26 Clayton D. A generalization of the transmission/disequilibrium test for uncertain haplotype transmission. Am J Hum Genet 1999; 65: 1170–1177. 27 Clayton D, Jones H. Transmission/disequlibrium tests for extended marker haplotypes. Am J Hum Genet 1999; 65: 1161–1169. 28 Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quanti. Methods 2001; 25: 402–408. 29 Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, De Paepe A et al. Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biol 2002; 3: RESEARCH0034. 30 Stephens M, Smith NJ, Donnelly P. A new statistical method for haplotype reconstruction from population data. Am J Hum Genet 2001; 68: 978–989. 31 Stephens M, Donnelly PA. Comparison of Bayesian methods for haplotype reconstruction from population genotype data. Am J Hum Genet 2003; 73: 1162–1169. 32 Couronne O, Poliakov A, Bray N, Ishkhanov T, Ryaboy D, Rubin E et al. Strategies and tools for whole-genome alignments. Genome Res 2003; 13: 73–80. 33 Mathews DH, Disney MD, Childs JL, Schroeder SJ, Zuker M, Turner DH. Incorporating chemical modification constraints into a dynamic programming algorithm for prediction of RNA secondary structure. Proc Natl Acad Sci USA 2004; 101: 7287–7292. 34 Mathews DH, Sabina J, Zuker M, Turner DH. Expanded sequence dependence of thermodynamic parameters improves prediction of RNA secondary structure. J Mol Biology 1999; 288: 911–940. 35 Enright AJ, John B, Gaul U, Tuschl T, Sander C, Marks DS. MicroRNA targets in Drosophila. Genome Biol 2003; 5: R1. 36 Griffiths-Jones S. The microRNA registry. Nucleic Acids Res 2004; 32: D109–D111.

37 Fanous AH, van den Oord EJ, Riley BP, Aggen SH, Neale MC, O’Neill FA. Relationship between a high-risk haplotype in the DTNBP1 (Dysbindin) gene and clinical features of schizophrenia. Am J Psychiatry 2005; 162: 1824–1832. 38 Puga I, Lainez B, Fernandez-Real JM, Buxade M, Broch M, Vendrell J et al. A polymorphism in the 30 untranslated region of the gene for tumor necrosis factor receptor 2 modulates reporter gene expression. Endocrinology 2005; 146: 2210–2220. 39 Conne B, Stutz A, Vassalli JD. The 30 untranslated region of messenger RNA: a molecular ‘hotspot’ for pathology? Nat Med 2000; 6: 637–641. 40 Gru¨n D, Wang Y-L, Langenberger D, Gunsalus KC, Rajewsky N. microRNA target predictions across seven Drosophila species and comparison to mammalian targets. PLoS Comput Biol 2005; 1: e13. 41 Branchek TA, Blackburn TP. Trace amine receptors as targets for novel therapeutics: legend, myth and fact. Curr Opin Pharmacol 2003; 3: 90–97. 42 Cao Q, Martinez M, Zhang J, Sanders AR, Badner JA, Cravchik A et al. Suggestive evidence for a schizophrenia susceptibility locus on chromosome 6q and a confirmation in an independent series of pedigrees. Genomics 1997; 43: 1–8. 43 Duan S, Du J, Xu Y, Xing Q, Wang H, Wu S et al. Failure to find association between TAAR6 and schizophrenia in the Chinese Han population. J Neural Transm 2005 [Advance online publication, August 3]. 44 Ikeda M, Iwata N, Suzuki T, Kitajima T, Yamanouchi Y, Kinoshita Y et al. No association of haplotype-tagging SNPs in TAAR6 with schizophrenia in Japanese patients. Schizophr Res 2005; 78: 127–130. 45 Amann D, Avidan N, Kanyas K, Kohn Y, Hamdan A, Ben-Asher E et al. The trace amine receptor 4 gene is not associated with schizophrenia in a sample linked to chromosome 6q23. Mol Psychiatry 2005 Adv online publication, September 27. 46 Paciga SA, Fan Y-T, Walton KM, King DP. Identification of a haplotype in TAAR6 (TRAR4) associated with susceptibility to schizophrenia in a Caucasian population. The 55th Annual Meeting of the American Sociaty of Human Genetics. Salt Lake City, Utah, USA, October 25th–29th, 2005. Abstract book, p 316. 47 Jamra RA, Sircar I, Becker T, Freudenberg-Hua Y, Ohlraun S, Freudenberg J et al. A family-based and case–control association study of trace amine receptor genes on chromosome 6q23 in bipolar affective disorder. Mol Psychiatry 2005; 10: 618–620. 48 Neale BN, Sham PC. The future of association studies: gene-based analysis and replication. Am J Hum Genet 2004; 75: 353–362. 49 Nakajima T, Jorde LB, Ishigami T, Umemura S, Emi M, Lalouel J-M et al. Nucleotide diversity and haplotype structure of the human angiotensinogen gene in two populations. Am J Hum Genet 2002; 70: 108–123. 50 Hilgers V, Pourquie´ O, Dubrulle J. In vivo analysis of mRNA using the Tet-Off system in chicken embryo. Dev Biol 2005; 284: 292–300. 51 Lagos-Quintana M, Rauhut R, Yalcin A, Meyer J, Lendeckel W, Tuschl T. Identification of tissue-specific microRNAs from mouse. Curr Biol 2002; 12: 735–739. 52 Lewis BP, Burge CB, Bartel DP. Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microRNA targets. Cell 2005; 120: 15–20. 53 Fanous AH, Kendler KS. Genetic heterogeneity, modifier genes, and quantitative phenotypes in psychiatric illness: searching for a framework. Mol Psychiatry 2005; 10: 6–13. 54 Finch CE, Sapolsky RM. The evolution of Alzheimer disease, the reproductive schedule, and apoE isoforms. Neurobiol Aging 1999; 20: 407–428. 55 Conrad C, Vianna C, Schultz C, Thal DR, Ghebremedhin E, Lenz J et al. Molecular evolution and genetics of the Saitohin gene and tau haplotype in Alzheimer’s disease and argyrophilic grain disease. J Neurochem 2004; 89: 179–188. 56 Gregory SG, Barlow KF, McLay KE, Kaul R, Swarbreck D, Dunham A et al. The DNA sequence and biological annotation of human chromosome1. Nature 2006; 441: 315–321.

853

Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)

Molecular Psychiatry