Genetic Architecture of Temperament

To the Editor: With great interest we read the article by Jeffrey H. Herbst, Ph.D., and colleagues (1), who, in a sample of up to 587 elderly Baltimore community residents, failed to find support for the previously described associations of a 48-base-pair repeat in the dopamine D₄ receptor gene (DRD4) with the personality trait of novelty seeking and of the 5-HTTLPR polymorphism of the serotonin transporter gene with harm avoidance. Although in this study on DRD4, as in most previous ones, the sample seemed to cover a broad range of trait values, a prior report on the same cohort (2) compared 188 individuals selected from the extremes of the novelty-seeking distribution. Although the authors were unaware of the actual degree of overlap in their samples, it is reassuring to see that both strategies, when applied to the same cohort, may well lead to identical results.

Dr. Herbst and co-workers discussed the ambiguous factor structure of Cloninger’s Temperament and Character Inventory as a possible source of negative findings. Given the less than 4% contribution of the DRD4 polymorphism to the overall variance in the novelty-seeking trait, as described in the original studies, blurred factor structure could indeed mask an existing but considerably weak genetic association even in sizable samples. We can partly confirm the results of the authors’ factor analysis with data obtained from the 4,753 participants of the Northern Finland 1966 Birth Cohort Study who fulfilled the temperament items from the Temperament and Character Inventory at age 31. Novelty seeking has been found to be associated with the DRD4 polymorphism in 190 subjects from this large, unselected cohort from the general population (3) who scored in the extreme range. Similar to the results in Table 3 of the article by Dr. Herbst et al., all of the items from the four subscales for harm avoidance loaded strongly (0.75–0.79) on a single factor in a promax-rotated principal-components analysis that forced four factors to be extracted from the 11 subscales that measured novelty seeking (subscales 1–4), harm avoidance (subscales 1–4), reward dependence (subscales 1, 3, and 4), and persistence (formerly reward dependence subscale 2). Subscale 1 for novelty seeking loaded negatively on the same factor I (–0.47). Of the four novelty seeking subscale loadings greater than 0.66 in their analysis, two each were confined to one of two separate factors (IV and V). The loadings of the novelty seeking subscales were more homogeneously distributed in our sample (novelty seeking subscales 2–4 loaded on a common factor with loadings of 0.61–0.80); this may have allowed detection of a DRD4 association in our sample but not in theirs. Less dispersion than in their analysis was also seen in major loadings for reward dependence, which fell on two separate factors (reward dependence subscales 3 and 4: loadings=0.72 and 0.76; reward dependence subscale 1: loading=0.81; persistence loaded at 0.63 on the same factor). However, we agree that according to its most stringent phenotypic structure, harm avoidance should offer an easier target than novelty seeking for the identification of the genetic factors behind temperament measures.