Candidate genes that might affect anxiety in humans are many: dozens of genes when knocked-out show an increase in levels of anxiety in mice [20]. Most normal variation is, however, not likely to be due to null alleles. Thus, analysis of quantitative trait loci (QTLs), to study the variations in anxiety between normal mouse strains, may be a better model of normal human population variation. Many mouse QTLs related to anxiety have been published and reviewed recently [21]. But QTLs are often broad and cannot easily be narrowed down to small intervals or genes. Flint and co-workers have recently used heterogeneous stock (HS) mice to overcome this problem. HS mice are akin to a population - mice are mated to maximize heterogeneity, and each mouse is a genetically unique individual. But HS mice are derived from a small number of inbred strains, typically eight, so the number of different alleles is limited. Using two sets of HS mice, Talbot et al. [18] recently narrowed down several different mouse anxiety traits (conditioned fear, contextual fear and open-field exploration) to small, non-overlapping intervals on mouse chromosome 1, in a region previously implicated by other studies. These results suggest that there are several anxiety genes in close proximity, rather than one gene with pleiotropic effects.

Testing specific candidate genes in humans can be done by simple association tests that ask the following: does the average N score depend significantly on the genotype? The first such study implicating a specific gene was by Lesch et al. [22], who found a common functional promoter variant in the serotonin transporter associated with neuroticism. Similarly, we recently identified a coding variant in brain-derived neurotrophic factor (BDNF) [23] associated with N - and this variant is also found associated with bipolar disorder [24], confirming the link between N and depression. Each of these genetic variants has only a small effect size, however, influencing just a few percent of the total variance.

A more systematic way to identify genetic variants affecting N is a whole-genome analysis. Recently Fullerton et al. [12] reported the first large-scale systematic linkage analysis of N. Over 80,000 individuals in over 20,000 sibships of at least two individuals were identified, contacted and administered the NEO-PI. Of these, over 600 sibships scoring at the extremes of the distribution - either both extremely high, both extremely low, or extremely discordant high-low - were selected for the genome scan. Five loci were identified with logarithm of odds (LOD) scores above 3.8, which corresponds to a genome-wide significance p < 0.05. Some loci acted only in females and others only in males, suggesting that gender needs to be considered in studies of anxiety. Interestingly, one of these loci maps to a region of 1q homologous to the region containing the QTLs found in mice on chromosome 1.

Many chromosomal regions show evidence for linkage to panic disorder and related phenotypes, and some of them have been confirmed in more than one study (Table 1). Three loci (1q, 7p, and 13q) are linked to both panic disorder and neuroticism but not agoraphobia or simple phobias. Three different studies found linkage of panic disorder to chromosome 1q [11, 13, 15] - but all map panic disorder distal to the 1q locus related to neuroticism and to the murine anxiety QTL discussed above. Two different study designs implicated the same region of chromosome 13q: an HLOD score (a LOD score under conditions of heterogeneity) of 3.57 was corroborated by multipoint analyses in 60 families segregating panic disorder and bladder/kidney conditions [15], suggesting that this locus includes genes involved in a pleiotropic syndrome. In addition, this region was also implicated by a large sibling study of neuroticism, at least in females [12]. Two studies also found linkage to 7p in panic disorder [11, 16], which is also associated with neuroticism [12].

In a more targeted approach, focusing on a single extended pedigree and only on regions syntenic to suggested QTLs in mice, linkage was obtained for panic disorder combined with agoraphobia to 12q13 [17]. In addition, an HLOD score peak on 12q22 - possibly another locus - was obtained for families segregating panic disorder and bladder/kidney conditions [15]. Fullerton and colleagues' [12] study of discordant and concordant siblings found their highest LOD score of 4.7 for a QTL for neuroticism in this region. In a linkage study of 25 Icelandic extended pedigrees, linkage to chromosome 9q31 was found with a highly significant allele-scanning LOD score of 4.18 [19]. If the phenotype is broadened to include anxiety and an additional 37 pedigrees are included, the LOD score goes down to 2.0, suggesting that 9q31 harbors a locus that is fairly specific for panic disorder. Two independent studies in non-Icelandic Caucasians [11, 15] showed LOD scores between 1 and 2 for panic disorder in a region about 30 centiMorgans distal to 9q31.

Modest LOD scores for panic disorder on chromosome 3q were reported in two studies [11, 13]. One of these studies, a genome scan of families segregating panic disorder, agoraphobia, specific phobias and other anxiety disorders, also identified a region on chromosome 14p as linked to panic disorder and agoraphobia [13] as well as to specific phobias [14]. If a narrow phenotype of exclusively agoraphobia was considered, the linkage peak on 3q moved to a more distal location, but this result was dominated by a single family that showed no linkage to 14p. Removal of this family from the genome-scan analysis for specific phobias increased the LOD score for 14p in the remaining families. Thus, by careful attention to phenotypes, the differential contribution of 14p, predominantly to phobias, and 3q, predominantly to agoraphobia but also panic disorder, could be dissected in spite of some overlap. These linkage studies suggest that agoraphobia and specific phobias may have different etiologies, a conclusion that is consistent with epidemiological data [6]. Moreover, in mouse one of five QTLs for open-field activity, an anxiety-related trait, mapped to a region homologous to 14P [25].

Finally, linkage of a chromosome 22q marker, but not multipoint linkage, was reported in the same families segregating panic disorder and bladder/kidney conditions that showed linkage to 13q [15]. In addition, association was reported between panic disorder and a haplotype around a candidate gene on 22q12, catechol-O-methyltransferase (COMT), which may have a biological role in anxiety [26]. A mouse QTL for transition from dark to light, an anxiety-related phenotype, has been mapped to a mouse chromosome 15 region homologous in part to human 22q13 [27], but the map position is too broad to conclude with certainty that the mouse QTL is syntenic with human 22q.

The high comorbidity of multiple anxiety disorders within the same individuals and between individuals in the same pedigree presents a particular challenge for genetics. Are some of these disorders extremes of a dimensional trait, do they share some susceptibility genes, or are they completely different phenotypes with some degree of overlap? There is some agreement between epidemiological and genetic linkage data to suggest that panic disorder and agoraphobia are related to each other but distinct from specific phobias. But there seem to be genes that increase risk only for specific disorders, as well as genes that broadly increase the risk for anxiety disorders. Some (see [6]) argue strongly that dimensional traits such as neuroticism should be incorporated into the diagnostic criteria for genetic studies. The fact that all five QTLs for neuroticism identified in the large population study [12] also map near anxiety disorder loci strongly supports this approach. The focus on extremes in this QTL study may have introduced a bias towards anxiety disorders, however. On the other hand, some studies have been quite successful by focusing on rare subtypes such as panic disorder associated with bladder and renal abnormalities [15]. Estivill and colleagues [28] used a similar narrow subtype approach, and found many patients with panic disorder and hyperextendable joints to have a somatic chromosome 15 abnormality.

A final question to consider is whether linkage to the same region always implies the same gene. Although a number of related phenotypes may map to the same chromosomal interval, these regions are often large and may harbor more than one susceptibility locus. This is exemplified by the mouse study in which several anxiety-related QTLs mapped to the same region of chromosome 1 but were later shown by fine mapping to map to more than one locus [18]. Thus, while it is tempting to conclude that some loci are associated with more generalized anxiety whereas others are more specific and linked to only one disorder, we have to be careful: some of the 'reproducible' linkage regions may contain more than one gene, and others will be false positives. Nevertheless, enormous progress has been made in the past year. Careful ascertainment and recording of all relevant phenotypes will be necessary to allow linkage and association analysis with both narrowly and more broadly defined traits and disorders, and will probably be essential if we are to succeed with the next step, the identification of the genes involved in these complex and disabling disorders.