Two Recent Discoveries in Psychiatric Mental Genetics

Two recent discoveries in the genetics of mental phenomena, one from the genetics of schizophrenia and the other from a genetic discovery involving a personality trait known as "novelty seeking," illustrate the problems of a purely genetic approach to mental illnesses. The following sections discuss some of the methodological problems with the research in these areas and argue that the genetics of mental health are best understood through a study of behavioral genetics. A review of recent advances in this area suggests potential problems as well as possible future successes in explaining and treating mental disorders.


Schizophrenia is a major mental disorder that occurs in approximately one out of every hundred individuals, a startlingly high prevalence for such a serious illness. It has been known for many years that the disorder runs in families, and a number of family, twin, and adoption studies have indicated that genetic factors are an important component. The 1980s saw several purported advances in the genetics both of schizophrenia and manic-depressive disorder, notably the localization of the former disease to chromosome 5 and the latter disorder to chromosome 11. Unfortunately, those results could not be confirmed, and these (probably) false positives have fueled criticisms of the entire effort to seek any precise genetic contribution to mental disorders. These false positive reports, however, have also engendered a much more sophisticated and critical approach to the methodology of studying complex traits, of which mental disorders are paradigms.

The November 1995 issue of Nature Genetics published three papers and three letters, four of which provide some fairly sound evidence for a schizophrenia vulnerability or susceptibility locus on the short arm of chromosome 6. These studies followed an earlier investigation by Straub and Kendler and their colleagues (Straub et al. 1995) that informally reported that preliminary evidence for a susceptibility locus for schizophrenia had been found in a sample of Irish schizophrenic families. In the main article in Nature Genetics, Straub and Kendler's group reported on an enlarged sample using 265 pedigrees and stated that "with linkage analysis we find evidence for a vulnerability locus for schizophrenia in the region 6p24-22. The greatest lod score, assuming locus heterogeneity, is 3.51 (P = 0.0002) with D6S296" (Straub et al. 1995, 287). (A lod score is a measure of the evidence in favor of a specific location for a trait-causing gene in comparison with a null hypothesis postulating no linkage to a trait-causing gene in the region [Lander and Schork 1994, 2039]. It is reported as a logarithm to the base 10 of the likelihood ratio, with a score of 3 typically taken as significant, although there are subtleties about this that are not detailed here.6)

A paper by Moises et al. (1995) reports on a three-stage genome-wide search for a schizophrenia locus. In their first stage, they investigated five family lines in Iceland and found 26 loci that were suggestive of linkage. Ten of these loci were then followed up in stage II of their inquiry using a larger international set of 65 families. This stage provided potential linkages to chromosomes 6p, 9, and 20. The combination of the families from these two stages, and the addition of a third sample of 113 schizophrenics and their unaffected parents from China, yielded significant linkage to chromosome 6p. A study by Schwab et al. (1995) on affected sib pairs in 43 German families and 11 Israeli families, with a total of 78 sib pairs, reported a maximum lod score of 2.2 (indicating suggestive evidence) for a schizophrenia linkage to a region in chromosome 6. However, as Schwab et al. (1995, 326) note in comparing their results with Straub's investigation, there are clear differences in the patterns observed (my emphasis). A report by Antonarakis et al. (1995) also found a maximum lod score of 1.17, indicating at best suggestive evidence for chromosome 6. Two other studies by Mowry et al. (1995) and Gurling et al. (1995) appearing in the November 1995 issue of Nature Genetics, however, did not support the 6p localization.

An optimistic interpretation by Peltonen appeared in Nature in 1995. Pooling these four positive studies, Peltonen (1995, 665) concluded that they provide evidence that chromosome 6p does indeed carry a locus that predisposes to schizophrenia (my emphasis). However, like the authors of the original articles reporting a linkage of schizophrenia to chromosome 6p, Peltonen is quick to note that all is not clear methodologically. One major problem has to do with the variable diagnostic definitions for schizophrenia used by the dif ferent research teams. Generally, each of the investigating teams used somewhat different diagnostic inclusion criteria, though several employed two or three different categories simultaneously, seeking the strongest results (in terms of lod scores). All utilize The Diagnostic and Statistical Manual of Mental Disorders, third edition revised (DSM-III-R) (or roughly equivalent) diagnoses and related instruments. What most of the teams call a "narrow" definition of schizophrenia includes DSM-III-R schizophrenia and also schizoaffective disorder (Antonarakis et al. 1995, 236). An "intermediate" definition is typically more inclusive and adds to the narrow definition schizotypal personality disorder and also (in the case of Straub et al. 1995, 287-288) all other nonaffec-tive psychotic disorders (that is, schizophreniform disorder, delusional disorder, atypical psychosis, and good-outcome SAD). The "broad" category of Straub et al. adds to this intermediate category mood incongruent and mood congruent psychotic affective illness; and paranoid, avoidant, and schizoid personality disorder (Straub et al. 1995, 288). Straub et al. found their strongest links using this broad definition of the complex trait of schizophrenia and much weaker links using stricter criteria. Other research teams, however, utilized the narrow definition and obtained their positive linkage results with this stricter definition. One of these stricter groups noted critically that the broad diagnostic model includes milder psychiatric disorders with questionable genetic relatedness to schizophrenia (Schwab et al. 1995, 326) (my emphasis).7

In addition to disagreeing over which cases of the complex traits of schizophrenia should be counted, the research teams diverged in the type of inheritance model they found in their populations (i.e., whether the gene was dominant, recessive, etc.). Several different models were tested by each of the teams, with Straub et al. as well as Schwab et al. obtaining their best results assuming a dominant gene model. In contrast, Antonarakis et al. found that a recessive gene model produced the most significant lod scores. The implications of these different diagnostic approaches and different genetic models are addressed later, but readers should keep in mind that there are many points where the interpretation of the data can be manipulated to achieve more statistically significant results. This is not meant to hint at any fraud or data manipulation in these studies, but merely to point out that there is much that is methodologically less than well constrained in psychiatric genetics and also more generally in complex-trait genetics. Leading geneticists have recognized this as a methodological problem and have begun to advance new criteria for research design and the reporting of "significant" results. For example, Lander and Kruglyak's (1995) article also proposes that the observed P values (which can unambiguously be related to lod scores) should be multiplied by the number of inheritance models (thus decreasing the significance level), but this assumes independence—a condition not often satisfied. Correlated phenotypes also require a correction to decrease statistical significance, but it is difficult to articulate appropriate guidelines for this correction. These authors argue strongly for computationally intensive simulation studies and for importance sampling to assess these effects.

A Novelty-Seeking Gene

Another recent discovery suggests the existence of a novelty-seeking gene in a quite diverse set of populations. Though novelty-seeking behavior is prima facie quite distinct from concerns about a serious disorder such as schizophrenia, Cloninger and his colleagues (1996) speculated that a better understanding of such personality genes may give us the best purchase on disorders such as schizophrenia.

Several research programs in human behavioral genetics have proposed that fairly general emotional and cognitive features have a substantial genetic component. Jerome Kagan (1994) recently summarized his investigation into temperament in his book Galen's Prophecy, in which he traces back to Galen of Pergamon the idea that much of the variation in human behavior is due to difference in temperament types. Bouchard, in his brief review article (1994), cites Darwin as the source of this view. In Behavioral Genetics: A Primer, Robert Plomin and his co-authors cite studies supporting the view that two "superfactor" personality traits—extraversion and neuroticism—with herita-bilities of 0.46 and 0.51, respectively, cut across many personality traits (Plomin et al. 1990, 1994).8

One of the major contributors to this area, Robert Cloninger, designed what is termed a tridimensional personality questionnaire (TPQ) to measure four domains of temperament (Cloninger et al. 1993). These four domains are novelty seeking, harm avoidance, reward dependance, and persistence. These dimensions of temperament were hypothesized to be biologically based on distinct chemical and genetic elements, with novelty seeking, for example, positively related to the neurotransmitter dopamine. Those scoring higher on the TPQ novelty-seeking scale were characterized as impulsive, exploratory, fickle, excitable, quick-tempered, and extravagant, whereas those who scored lower than average tended to be reflective, rigid, loyal, stoic, slow-tempered, and frugal (Ebstein et al. 1996, 78).

The January 1996 issue of Nature Genetics carried two articles reporting a confirmation of Cloninger's prediction regarding novelty seeking. The first article by a group based in Israel and led by Ebstein, reports a statistically significant association between the personality trait of novelty seeking as measured by Cloninger's TPQ and a long allele (L) of the human D4 dopamine receptor gene known as D4DR. A second group involving Benjamin at the National Institute of Mental Health (NIMH) as well as Dean Hamer from the National Cancer Institute replicated the Ebstein et al. study. To assess personality traits, Benjamin et al. (1996) used a different but related measuring instrument known as the NEO personality inventory. (The original model was a three-factor model containing neuroticism, extraversion, and openness to experience, the origin of the acronym NEO; see McCrae and Costa [1990, v-vi].) This study also found a statistically significant relation of novelty seeking to the long allele of D4DR using a mapping between the NEO instrument and the TPQ questionnaire.

Whereas all the studies on schizophrenia discussed here employed a genetic linkage approach to finding a schizophrenia gene, these personality gene studies used a method termed allelic association. Such an approach, as Ebstein et al. note, works best when it employs candidate genes that a priori make "biological sense" and that have a functional significance in the determination of the trait (Ebstein et al. 1996, 79; see also Lander and Schork 1994, 2041, as well as Risch and Merikangas 1996). The source of Ebstein et al.'s "a priori" hypothesis was, of course, Cloninger's personality theory and his TPQ.

It is important to understand that the long allele of D4DR that is believed to be "causative" of novelty seeking accounts for only a small proportion of the trait in the populations studied. Benjamin and Hamer's group is more emphatic on this point than Ebstein et al. In Benjamin et al. we find that "although the mean score for the L (long allele) subjects is greater than the S (short allele) subjects by 0.4 standard deviations (a moderate effect size), the distributions are highly overlapping and D4DR accounts for only 3 to 4% of the total variance." They estimated from twin studies that the broad heritabil-ity of novelty seeking is 41 percent and in the families they studied there was a correlation of 0.23 for estimated TPQ-novelty seeking scores in siblings." Thus D4DR accounts for roughly 10% of the genetic variance, as might be expected if there are 10 or so genes for this complex, normally distributed trait. These results indicate that Novelty Seeking is partially but not completely mediated by genes, and that the D4DR polymorphism accounts for some but not all ofthe genetic effects" (1996, 83) (my emphasis). Figure 3 depicts the subtle difference between individuals' scores related to the two alleles.

Cloninger and his colleagues (1996) wrote a commentary on the Ebstein and Benjamin papers that appeared in the same January 1996 issue of Nature Genetics. They make several points that are worth citing. Cloninger et al. argue first that the TPQ approach was designed to be genetically homogeneous, in contrast to the NEO personality questionnaire, and that this is confirmed by the two studies discussed. Second, these authors suggest that personality development is a complex, dynamic process that has many influences on susceptibility to pathopsychology. More specifically, they state that a novelty seeker is likely to develop into an extravert with a mature creative character if he or she is also low in harm avoidance (optimistic), high in reward dependence (sociable), and high in persistence. In contrast, they find that a novelty seeker is more likely to become disorganized or schizotypal if he or she is also aloof (low in reward dependence and average in other temperament dimensions). They add "In contrast to the quick and clear replication of the D4DR association with Novelty Seeking by Ebstein et al. and Benjamin et al., replication of specific genetic contributions to genetically complex disorders like schizophrenia have

Figure 3. Distributions of estimated novelty-seeking scores. The x axis shows the estimated TPQ-novelty-seeking scores separated into eight groups with the indicated median T scores. The y axis shows the distribution in each of the eight groups of subjects with short D4DR exon III alleles (group S, n = 217, stippled bars) and subjects with longD4DR exon III alleles (group L, n = 98, solid bars). (Source; Benjamin et al. 1996).

been elusive. The exponential increase in risk of schizophrenia with increasing degree of genetic relationship indicates the importance of non-linear interactions among multiple genetic factors. When a disease is caused by interactions among multiple susceptibility dimensions, each of which may be oligogenic, then replication of particular genes is unlikely in samples of practical size" (1996,4).

As already indicated, replication studies have been the Achilles heel of psychiatric genetics, and Cloninger et al. are right to emphasize these difficulties. In fact, although Benjamin et al. (1996) provided a quick and clear replication of the D4DR association with novelty seeking, several other studies have failed to confirm this association. However, an additional confirmation has appeared.9 Replications or confirmations, particularly of complex traits, are difficult to obtain, partly because of weak gene effects (there are a number of contributory genes), biological variation, and subtle statistical reasons. Lander and Kruglyak make these points eloquently:

Failure to replicate does not necessarily disprove a hypothesis. Linkages will often involve weak effects, which may turn out to be weaker in a second study. Indeed there is a subtle but systematic reason for this: positive linkage results are somewhat biased because they include those weak effects that random fluctuations helped push above threshold [of statistical significance], but exclude slightly stronger effects that random fluctuations happened to push below [the] threshold. Initial positive reports will thus tend to overestimate effects, while subsequent studies will regress to the true value Replication studies should always state their power to detect the proposed effect with the given sample size. Negative results are meaningful only if the [statistical] power is high. Regrettably, many reports neglect this issue entirely.

When several replication studies are carried out, the results may con-flict—with some studies replicating the original findings and others failing to do so. This may reflect population heterogeneity, diagnostic differences, or simply statistical fluctuation. Careful meta-analysis of all studies may be useful to assess whether the overall evidence is convincing (Lander and Kruglyak 1995,245).

Cloninger et al. (1996,4) offer a strategy that may assist with the replication problem in connection with psychiatric disorders such as schizophrenia. They write that it may be more fruitful to map genes contributing to temperament, which has a relatively simple genetic architecture, and can be quantified easily and reliably by questionnaires. They state that later susceptibility to complex disorders such as schizophrenia and alcoholism can be evaluated in terms of the risk from heritable personality traits and possibly disease-specific factors. In this way, they believe, success in mapping genes for a normal personality may indicate a fruitful way to map genes for pathopsychology as well.

Thus Cloninger et al. suggest that there are some simplifications that can be found in human behaviors by using personality genetics that might help to unravel the complexities of less tractable disorders, such as schizophrenia. One question that naturally arises is what will happen as neurobiology advances and identifies all the genes involved in an organism, as well as the neural circuits to which they give rise. Will these accomplishments result in simple and powerful tools that can be used to explicate quite complex behaviors of both normal and pathological forms? We are not even close to this type of arch-reductionist result in the area of human studies, but we are rapidly approaching it in the study of simpler systems. Two of these systems are discussed here to indicate the possible results of complete genetic specifications and their implications for behavior and for mental health.

0 0

Post a comment