Letters to the Editor 240 Several strong candidate and regions have been Convergence of linkage, implicated in bipolar disorder (BP) and association and GWAS (SCZ) through linkage and association studies. These disorders have also recently been studied in genome- findings for a candidate wide association studies (GWAS), identifying further putative candidate loci, albeit with lower levels of region for bipolar disorder significance and reproducibility than in GWAS of other complex disorders.1 Our study focuses on a and schizophrenia on well-established candidate region for psychiatric ill- ness and independently implicates one of the top chromosome 4p candidate genes to emerge from GWAS of BP. The chromosome 4p15–p16 linkage region was first identified in a large Scottish family multiply affected Molecular Psychiatry (2011) 16, 240–242; doi:10.1038/ with major affective disorder.2,3 Subsequently, it has mp.2010.25; published online 30 March 2010 been repeatedly implicated in BP, SCZ and related

Molecular Psychiatry Letters to the Editor 241 phenotypes by linkage, association (reviewed in and Alzheimer’s disease, owing to the variable inter- Le Hellard et al.3)andGWAS.4 We sought to refine locus correlation between the GAPDH and APOE e4 the 20-Mb linkage region (delineated into four sub- alleles in two different samples.8 regions A–D3) in search of a susceptibility (s) by (1) SORCS2 plays a regulatory role in neurons, making it a performing a joint analysis of previously identified strong functional candidate gene for psychiatric illness. It putative regions of association in regions B and D5 in a is thought to be involved in both internalizing surface- combined sample of 600 BP patients, 681 SCZ patients bound ligands and in intracellular trafficking processes. and 765 controls from the Scottish population and (2) It is highly expressed during development and in the screening, for the first time, the remainder of the linkage mature central nervous system, in a transient and region (regions A and C) for association in a sample of dynamic manner, in regions of proliferation and differ- 506 BP patients, 576 SCZ patients and 607 controls entiation. In the mature , expression is mainly from the Scottish population. We tested single markers neuronal in the cerebral cortex, hippocampus and and sliding windows of haplotypes in the BP, SCZ and cerebellum, where it is believed to be involved in the combined samples, and stratified by gender, using 633 regulation of protein transport and is predicted to have a haplotype-tagging single-nucleotide polymorphisms critical impact on neuronal viability and function.9 (SNPs), and refined the region to 16 putative regions In conclusion, we refine the chromosome 4p linkage of association (B3 Mb), which met pre-determined, region for BP and related phenotypes, identifying region-wide significance thresholds, following a meth- candidate genes and regions worthy of further investi- od previously described and validated in Christoforou gation through meta- or mega-analyses10 of available et al.5 (Figure 1, Supplementary Material). Eight genes association data, additional replication, further fine- map directly to these regions, including, in particular, mapping, deep sequencing and/or functional studies to the SORCS2 gene, which maps within, or close to, four help rule out the false-positives and identify the true of the region A putatively associated regions and susceptibility gene(s). Most importantly, this study contains the most significant single-marker result in provides independent evidence of association for regions A and C (rs766008: P = 0.00030; ORC/A =1.4, SORCS2, one of the top candidate genes to emerge 95% confidence interval: 1.2–1.8). from GWAS of BP, illustrating an instance of conver- The SORCS2 gene was first identified in a GWAS of gence of the findings of a hypothesis-based study of a BP,4 in which it was the second best hit. Three well-supported susceptibility region with those of a SORCS2 SNPs (rs4411993, rs7683874 and rs10937823) hypothesis-free genome-wide approach. showed evidence of association in two independent samples, with the same alleles conferring the increase Conflict of interest in risk in both samples (best P = 0.000014 in the combined sample).4 SNP rs4411993 was subsequently The authors declare no conflict of interest. analyzed as part of a meta-analysis, in which it showed a trend towards significance (P = 0.054).6 We also tested these three SNPs and identified nominal A Christoforou1,5, KA McGhee1,2, SW Morris1, evidence of association to BP with rs7683874 PA Thomson1,2, S Anderson1, A McLean1, (P = 0.024) and rs10937823 (P = 0.031), but with the HS Torrance1, S Le Hellard1,5, BS Pickard1, opposite alleles conferring the increase in risk. In a D StClair3, WJ Muir1,4, DH Blackwood1,4, replication study of the 26 SNPs with strongest DJ Porteous1,2 and KL Evans1,2 association from the three GWAS of BP, Ollila et al.7 1Medical Section, Centre for Molecular also reported nominal association to these three Medicine, The University of Edinburgh and SORCS2 SNPs (best P = 0.0042) in a Finnish family- The Institute of Genetics and Molecular Medicine, based cohort, with the same alleles as in our study Molecular Medicine Centre, Western General conferring the increase in risk. Although this ‘flip–flop’ Hospital, Edinburgh, UK; 2Department of Psychology, association may suggest spurious findings, it could Centre for Cognitive Ageing and Cognitive also be a result of allelic heterogeneity, multi-locus Epidemiology, The University of Edinburgh, effects or variable LD patterns between the samples, Edinburgh, UK; 3Institute of Medical Sciences, as, for example, observed between the GAPDH gene University of Aberdeen, Aberdeen, UK and

Figure 1 Regions of association. UCSC gene browser track (March 2006) (http://genome.ucsc.edu) of the chromosome 4p linkage region, delineated into the four subregions (a–d). For each of the regions (a)to(d), two custom tracks are shown. First, the haplotype tagging (ht) SNPs genotyped and tested for association in this study. Second, the putative regions of association (and region-wide ‘Nyholt-corrected’ significance thresholds). Each putative region of association is indicated by a label (ai, aii, etc.), which corresponds to those in the supplementory online material (SOM) tables and text. The best P-value for each region of association is also provided. For region (a), there is an additional track, showing the location of the three SNPs reported by Baum et al. (2008) and their P-values in our study. *The only signle marker to meet its region-wide significance threshold in the second part of the study. **The two Baum et al. SORCS2 SNPs identified as nominally significant in our study.

Molecular Psychiatry Letters to the Editor 242 4Division of Psychiatry, University of Edinburgh, is supported anecdotally by the tendency for PD Royal Edinburgh Hospital, Edinburgh, UK sufferers to feel a strong urge to flee from the location E-mail: [email protected] where a panic attack occurs.2 Here we provide the 5Current address: Dr Einar Martens’ Research Group first human empirical evidence for a PD–flight link, for Biological Psychiatry, Centre for Medical Genetics showing that flight behaviour is significantly more and Molecular Medicine, University of Bergen, intense in carriers of a candidate genetic risk factor for Haukeland University Hospital, Helse Bergen HF, PD than in non-carriers. 5021 Bergen, Norway Human flight behaviour was measured using a computerized translation of a rodent runway task (Figure 1a) designed to index fear proneness behav- References iourally, as the intensity of flight effort in response to 5 1 WTCCC. Nature 2007; 447: 661–678. a pursuing threat stimulus. The genetic risk factor for 2 Blackwood DH, He L, Morris SW, McLean A, Whitton C, Thomson PD used in this study was the C allele of the 102T/C M et al. Nat Genet 1996; 12: 427–430 (Letter). single-nucleotide polymorphism (rs6313) within the 3 Le Hellard S, Lee AJ, Underwood S, Thomson PA, Morris SW, Torrance HS et al. Biol Psychiatry 2007; 61: serotonin 2a receptor gene (HTR2A) on chromosome 797–805. 13q14.2; the C allele in this SNP is known to be 4 Baum AE, Akula N, Cabanero M, Cardona I, Corona W, Klemens B associated with increased susceptibility to pure but et al. Mol Psychiatry 2008; 13: 197–207. not co-morbid PD,6 as well as to increased intensity of 5 Christoforou A, Le Hellard S, Thomson PA, Morris SW, Tenesa A, panic symptoms.7 All 200 participants (107 of whom Pickard BS et al. Mol Psychiatry 2007; 12: 1011–1025. 6 Baum AE, Hamshere M, Green E, Cichon S, Rietschel M, were male) gave informed consent and self-identified Noethen MM et al. Mol Psychiatry 2008; 13: 466–467 as healthy Caucasians. Buccal cells were collected (Letter). and DNA extracted using established methods (see 7 Ollila HM, Soronen P, Silander K, Palo OM, Kieseppa T, Kaunisto Supplementary Information). MA et al. Mol Psychiatry 2009; 14: 351–353 (Letter). 8 Lin PI, Vance JM, Pericak-Vance MA, Martin ER. Am J Hum Genet The genotype distribution of rs6313 SNP in HTR2A 2 2007; 80: 531–538. was in Hardy–Weinberg equilibrium (w = 0.632, 9 Willnow TE, Petersen CM, Nykjaer A. Nat Rev Neurosci 2008; 9: d.f. = 2, P = 0.73). There were no significant genotype 899–909. effects on flight intensity (F(1, 192) = 2.69, P = 0.070); 10 PGC. Mol Psychiatry 2009; 14: 10–17. however, carriers of the C risk allele showed sig- nificantly greater flight intensity than TT individuals Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp) (F(1, 194) = 4.90, P = 0.033; Figure 1b). The construct validity of flight intensity as a specific measure of fear proneness was supported by its significant positive association with scores on tissue damage fear (mea- sured by the Fear Survey Schedule) (F(1, 194) = 5.92, Flight behaviour in P = 0.022). The construct validity of flight intensity was also supported by the absence of a significant humans is intensified by association with scores on Spielberger trait anxiety (F(1, 194) = 0.01, P = 0.998), which is a widely ac- a candidate genetic risk cepted questionnaire measure of anxiety proneness.8 Sex did not affect flight intensity in this model factor for panic disorder: (F(1, 194) = 1.50, P = 0.222), nor was there a significant allele  sex interaction (F(1, 194) = 1.56, P = 0.213). evidence from a There was no significant interaction between tissue translational model of damage fear and rs6313 carrier status (C carrier or TT homozygote) (F(1, 191) = 0.44, P = 0.511). fear and anxiety In overview, therefore, our study is the first molecular genetic investigation of human defensive behaviour and the first study empirically to support Molecular Psychiatry (2011) 16, 242–244; doi:10.1038/ in humans the hypothesis that PD stems from mp.2010.2; published online 23 February 2010 alterations in the brain systems governing flight behaviour. However, although the HTR2A gene on chromosome 13q4-21 has previously been asso- Panic disorder (PD) is a serious and common ciated with PD, as a caveat it should be noted psychiatric condition1 characterized mainly by recur- that the rs6313 SNP in exon 1 of the coding sequence rent episodes of intense, uncontrollable fear known as of the HTR2A gene is a synonymous (or silent) panic attacks.2 The underlying causal mechanism for polymorphism. Therefore, its previously observed PD is unknown;3 however, the discovery that drugs effects at the phenotypic level may be mediated with clinical effectiveness against PD preferentially not by changes in protein structure but by other alter rodent flight behaviour suggests that PD reflects mechanisms such as gene expression. As rs6313 is alterations in the brain systems that govern flight.4 part of a four-SNP haplotype (, rs1328674, An association between PD and flight in humans rs6313, ), future attempts at understanding

Molecular Psychiatry