Zhou et al. reply

Abstract

Replying to: C. H. Cotton, J. Flint & T. G. Campbell Nature 458, 10.1038/nature07927 (2009)

The inability of Cotton et al.¹ to detect an effect of a functional haplotype (and locus) of neuropeptide Y (NPY), a stress regulatory neuropeptide, on neuroticism is interesting. Although it is important to measure effects of functional loci on complex behaviours, the strength of our study², and primary basis of its conclusions, was the larger and convergent effects of NPY on intermediate phenotypes, including regional brain responses to emotional stimuli and pain, and brain NPY messenger RNA and plasma NPY levels. Eysenck Neuroticism is a trait that we did not directly investigate. We reported modest association of NPY with two Harm Avoidance subscales from the Tridimensional Personality Questionnaire. Association of NPY with the complex trait of anxiety, especially when measured differently, is not the first place we would look to validate our results.

Main

Concerning their advocacy of genome-wide approaches, if we follow the conclusions of their genome-wide association study with the same data set³ then no loci contribute >1% of the variance in neuroticism. This is plausible, and could explain why they found no effect of NPY. However, Cotton et al.¹ genotyped the extremes of a large but relatively uncharacterized sample. Theoretically powerful, this approach may in practice be problematic. At the extremes of the distribution various confounds such as severe environmental stresses, rare functional alleles and measurement errors are more likely to be over-represented. Their study did not identify new functional loci for anxiety nor confirm functional loci for which there is independent evidence, as mentioned later. It is reasonable to request evidence that a tool works before using it to ‘weed the garden’.

There is indeed debate as to how to proceed in gene discovery for behaviour. However, candidate gene and genome-wide approaches are not at war. The goal of genome-wide studies is to identify locations of functional polymorphisms. Studies using intermediate phenotypes, on which alleles exert larger effects than complex behaviours, may be better able to expand our understanding of mechanism. Consistent and convergent effects of several functional alleles on intermediate phenotypes have demonstrated the validity of this approach. Recent discoveries relating common alleles to behaviour have primarily relied on brain imaging tools. Examples include the serotonin-transporter-linked polymorphic region (5-HTTLPR) that has a weak effect on depression and anxiety—an association that was indeed obscured when only the extremes of the distribution were compared⁴—but strong effects on brain metabolic responses to emotional stimuli⁵ and the uncoupling of limbic feedback circuitry (accounting for 30% of the variance in anxious temperament⁶). Brain imaging studies have also shown that a functional missense variant (Val158Met) of COMT alters brain activity during cognition⁷, pain⁸ and response to emotional stimuli (accounting for 38% of the variance in emotionality⁹), while having much more modest effects on complex behaviours, including anxiety. If allele effects on crudely measured behavioural phenotypes are undetectable in very large data sets, this may suggest that genome-wide genetic methods should be applied to data sets of more modest size, in which intermediate phenotypes have been measured that are more robust in detecting genetic influences on behaviour.