S C I E N C E ’ S C O M PA S S

65 PERSPECTIVES: GENETICS studies and verified in mice lacking the sero-

64 tonin transporter gene, the effect of this gene
63 on subjective measures of personality is
62
61
Rethinking Behavior Genetics modest, typically accounting for only a few
percent of total variance (4).
60 Dean Hamer Hariri and colleagues predicted that the
59 gene would have a larger effect on directly
58 he first 100 years spent studying the agree on how to design experiments or to measured brain activity. To test this idea, they
57
56
55
T genetics of behavior were straightfor-
ward. The aim was to determine the
extent to which individual differences in the
interpret the results; their advice has proven
as faddish (and useful) as the Hula-Hoop.
What’s the problem? It’s not the basic
performed functional nuclear magnetic reso-
nance imaging on normal subjects as they per-
formed either an emotional task—matching
54 way people think, feel, and behave are due premise of linkage and candidate gene analy- the affect of angry or afraid faces to a target
53 to variations in their genetic makeup. The sis; these approaches have identified dozens face—or a sensorimotor control task. As ex-
52 basic approaches, first described by Sir of genes involved in inherited diseases. Nor pected, subjects with the poorly transcribed,
51 Francis Galton in the late 1800s, were to is it the lack of DNA sequence information; high-anxiety serotonin transporter genotype
50 compare identical and fraternal twins, other virtually the entire code of the human showed a larger response of the amygdala to
49 family members, and adoptees that had genome is now known. The real culprit is the the emotional task than did the subjects with
48 been raised together or apart. assumption that the rich complexity of hu- the highly transcribed, low-anxiety genotype;
47 The results were consistently striking, man thought and emotion can be reduced to a there was no difference between the two geno-
46 albeit slow to be accepted. Genes were simple, linear relation between individual types on the control task. Importantly, the dif-
45 shown to influence virtually every aspect of
44 human personality, temperament, cognitive
43 style, and psychiatric disorder. The effects A Gene
42 of heredity were substantial, typically rep- Behavior
41 resenting 30 to 70% of total variation, and B Gene networks
40 highly replicable across societies and cul-
39 tures. The long reach of genes extended
38 from a friendly disposition to xenophobia, Brain
37 from bipolar disease to bedwetting, from
36 getting married to keeping a job. About the
35 only characteristics that seemed not to be at
34 least partially heritable were purely learned Behavior
33 traits such as the particular language one Parents Economic
32 spoke or the religion one believed in (1).
Environment
31 The second century of behavior genet-
30 ics has gotten off to a less satisfying start. Siblings Education
29 The current aim is to identify the specific
28 genes that contribute to individual differ- Two views of behavior genetics. (A) A simplified model underlying much behavior genetics re-
27 ences and determine what they do in the search envisages a direct linear relationship between individual genes and behaviors. (B) The reali-
26 brain. The approach is to search for DNA ty is likely to be far more complex with gene networks and multiple environmental factors impact-
25 sequence variations that correlate with be- ing brain development and function, which in turn will influence behavior.
24 havioral and personality traits, either by
23 tracking anonymous markers close to the genes and behaviors (see the figure). This ference between the two genotype groups was
22 genes of interest in family members (link- oversimplified model, which underlies most nearly fivefold, accounting for 20% of total
21 age analysis) or by directly comparing the current research in behavior genetics, ignores variance—an effect size nearly 10-fold higher
20 coding and regulatory sequences of candi- the critical importance of the brain, the envi- than in typical experiments using subjective
19 date genes (association analysis). ronment, and gene expression networks. behavioral or personality measures as the out-
18 The results have been disappointing and Three recent publications in Science come. This is precisely as expected from view-
17 inconsistent. Large and well-funded linkage show how measuring brain activity, environ- ing the brain as the obligatory intermediate be-
16 studies of the major psychiatric disorders mental variables, and subtle alterations in tween genotype and behavior (see the figure).
15 including schizophrenia, alcoholism, gene expression can strengthen behavior ge- Technical advances in noninvasive func-
14 Tourette syndrome, and bipolar disorder netics research. Hariri et al. (2) examined the tional neuroimaging will rapidly make this
13 have come up empty-handed; not a single influence of the serotonin transporter gene in type of analysis both more routine and more
12 new gene has been conclusively identified. the response of the amygdala to a fearful powerful (5, 6). Besides increasing signal
11 Most candidate gene findings have failed stimulus. There is a frequent DNA sequence size and thereby decreasing statistical noise,
10 consistent replication, and even those that variation in the control region of the human the main advantage of directly studying the
9 have been verified account for only a small serotonin transporter gene that influences the brain will be to focus attention on those be-
8 fraction of total variation. Meanwhile, the expression levels of this protein and thereby haviors that have a dedicated brain circuit to
CREDIT: KATHARINE SUTLIFF/SCIENCE

7 statisticians who are supposed to be guiding the amount of synaptic serotonin, a potent begin with. It should be obvious that the her-
6 and evaluating the research are unable to modulator of emotional responses. When the itability of some complex traits, like sexual
5 polymorphism was discovered 6 years ago, it orientation (7) and language acquisition (8),
4 was found to be associated with abnormal is probably the direct result of evolutionarily
The author is at the Laboratory of Biochemistry, Na-
3 tional Cancer Institute, National Institutes of Health,
levels of anxiety as assessed by self-report selected genetic programs whereas others,
2 Bethesda, MD 20892, USA. E-mail: hamerd@ questionnaires (3). Although this finding has like getting divorced or cigarette smoking,
1 dc37a.nci.nih.gov been replicated in numerous subsequent are more likely to involve an unrelated set of

www.sciencemag.org SCIENCE VOL 298 4 OCTOBER 2002 71

 S C I E N C E ’ S C O M PA S S
65 characteristics that happen to be amalgamat- and to be diagnosed with conduct disorder. the study of complex traits can be im-
64 ed by culture. All too often, though, the deci- Over 85% of the males who had both “bad proved, they are still at the primitive stage
63 sion of what to study is driven by socio-med- genes” and a “bad environment” developed of examining single genes. This isn’t how
62 ical politics rather than biological logic. some form of antisocial behavior by the time the brain works. Human behaviors, and the
61 The study by Caspi et al. (9) also analyzed they were 26. It will now be crucial to repeat brain circuits that produce them, are un-
60 a promoter region polymorphism, in this case this intriguing finding on other populations doubtedly the product of intricate networks
59 for the gene encoding monoamine oxidase A with documented developmental histories. involving hundreds to thousands of genes
58 (MAOA), an enzyme that breaks down the The serotonin transporter and MAOA sto- working in concert with multiple develop-
57 neurotransmitters serotonin, dopamine, and ries nicely illustrate how changes in regulatory mental and environmental events. Further
56 norepinephrine. Although the MAOA gene rather than coding sequences can influence advances in the field will require the devel-
55 had previously been implicated in aggression brain function and behavior. Such variations in opment of techniques, such as microarray
54 and impulse control in both humans and ro- gene expression probably play a predominant analysis, that measure the activity of many
53 dents (10), this transcriptional variant had not role in many types of individual differences, different genes simultaneously. Only then
52 been associated with personality traits (11). but this has been difficult to prove in humans will the gene hunters have a shot at achiev-
51 Caspi et al. hypothesized that the effect of the because we are so genetically outbred. Yan and ing the promises held out by the past centu-
50 gene would be more readily revealed if the en- colleagues (12) devised an elegant solution to ry of classical behavior genetics research.
49 vironment were explicitly taken into account. this problem. They measured the expression of
48 Their study group was a large birth co- different alleles in a single person who was References
47 hort, representative of the male population of heterozygous for the locus in question, thus 1. D. H. Hamer, P. Copeland, Living With Our Genes
(Doubleday, New York, 1998).
46 New Zealand, whose development had been avoiding the problems of extraneous differ- 2. A. R. Hariri et al., Science 297, 400 (2002).
45 carefully followed for 26 years. The environ- ences in genetic background or other factors. 3. K. P. Lesch et al., Science 274, 1527 (1996).
44 mental variable of interest was childhood Remarkably, even though most of the varia- 4. D. L. Murphy et al., Brain Res. Bull. 56, 487 (2001).
5. S. Y. Bookheimer et al. , N. Engl. J. Med. 343, 450
43 maltreatment, and the outcome was a com- tions they studied were random single-nu- (2000).
42 posite measure of antisocial behavior. Al- cleotide substitutions far from the promoter re- 6. M. F. Egan et al., Proc. Natl. Acad. Sci. U.S.A. 98, 6917
41 though the MAOA genotype by itself failed gion, almost half of them were associated with (2001).
7. D. H. Hamer et al., Science 261, 321 (1993).
40 to predict antisocial behavior, there was a sig- detectable changes in messenger RNA levels. 8. C. S. Lai et al., Nature 413, 519 (2001).
39 nificant interaction with childhood history; A few of the genes they studied are expressed 9. A. Caspi et al., Science 297, 851 (2002).
38 individuals with both a low-activity genotype in the brain, and many more will soon follow. 10. J. C. Shih, R. F. Thompson, Am. J. Hum. Genet. 65, 593
(1999).
37 and previous maltreatment were by far the Although the Hariri, Caspi, and Yan re- 11. S. Z. Sabol et al., Hum. Genet. 103, 273 (1998).
36 most likely to have committed a violent crime ports provide tantalizing glimpses of how 12. H. Yan et al., Science 297, 1143 (2002).
35
34 PERSPECTIVES: COSMOLOGY
33 verse. This gas is either neutral or moder-
32 ately ionized, with a temperature below
31
30
The Cosmic Web of Baryons ~20,000 K. A veritable forest of absorption
lines from hydrogen, the most common el-
29 Joel N. Bregman ement, tells us that at this time, Ωb = 0.04,
28 the same as shortly after the big bang.
27 bout 80% of the mass in the uni- A few billion years later, when the uni- In the present-day universe, this “forest”
26
25
24
A verse is composed of “dark matter,” verse was about one-quarter of its current of gas absorption lines has nearly vanished.
which can only be detected through age, active galactic nuclei (enormously lu- At first, the gas was believed to have been
its effects on gravity. The nature of this minous objects powered by massive black incorporated into the galaxies and stars that
23 material is entirely unknown. The remain- holes in the centers of galaxies) formed. In we see today. However, a mass census of
22 ing 20% is the kind of matter that we are their spectra, we see absorption lines from the local universe shows that the baryons in
21 all familiar with. Known as baryons, this very distant gas distributed across the uni- galaxies and cool gas amount to only Ωb =
20 matter forms the stars and galaxies in the 0.004. Hence, 90% of the baryons
19 local universe. But all the stars contain must be located elsewhere.
18 less than a tenth of the baryons that existed Gas is still found between
17 when the universe was young (1, 2). Four galaxies. For example, galaxy-
16 recent papers shed light on the where- clusters contain a stable hot atmo-
15 abouts of the missing baryons. sphere of gas with temperatures of
14 A few minutes after the big bang, the up to 108 K and masses as great
13 first elements—helium, deuterium, and as, or greater than, those of the
12 trace amounts of other light elements— galaxies. Yet even including the
11 were produced. The relative amounts of gas in these clusters (and in less
10 these isotopes and elements are sensitive massive groups of galaxies), most
9 to Ωb, the ratio of the baryon density to of the baryons are still missing.
8 the critical density of the universe (3). Theoretical calculations sug-
7 Abundance measurements of primordial gest a solution that features “fila-
CREDIT: A. HOBART/SAO/CXC

6 deuterium and other isotopes show that ments” of matter formed by grav-
5 Ωb = 0.04. Absorption by intergalactic gas. This artist’s impression itational collapse. The filaments,
4 shows how x-ray emission from a distant active galactic nu- which are much larger than
3 The author is in the Department of Astronomy, Uni- cleus reaches the Chandra x-ray observatory. Some x-rays galaxy clusters and not nearly as
2 versity of Michigan, Ann Arbor, MI 48109, USA. E- are absorbed by gas filaments in the intergalactic space. See dense, connect the many galaxy
1 mail: [email protected] chandra.harvard.edu/photo/2002/igm/index.html. clusters and groups in a cosmic

72 4 OCTOBER 2002 VOL 298 SCIENCE www.sciencemag.org