Behave: The Biology of Humans at Our Best and Worst

by Robert M. Sapolsky

  What did they observe? Some gene variants showed massive gene/environment interactions, with variants having radically different effects in different labs.

  Here’s the sort of data they got: Take a strain called 129/SvEvTac and a test measuring the effects of cocaine on activity. In Oregon cocaine increased activity in these mice by 667 centimeters of movement per fifteen minutes. In Albany, an increase of 701. Those are pretty similar numbers; good. And in Alberta? More than 5,000. That’s like identical triplets pole-vaulting, each in a different location; they’ve all had the same training, equipment, running surface, night’s rest, breakfast, and brand of underwear. The first two vault 18 feet and 18 feet one inch, and the third vaults 108 feet.

  Maybe the scientists didn’t know what they were doing; maybe the labs were chaotic. But variability was small within each lab, showing stable environmental conditions. And crucially, a few variants didn’t show a gene/environment interaction, producing similar effects in the three labs.

  What does this mean? That most of the gene variants were so sensitive to environment that gene/environment interactions occurred even in these obsessively similar lab settings, where incredibly subtle (and still unidentified) environmental differences made huge differences in what the gene did.

  Citing “gene/environment interactions” is a time-honored genetics cliché.36 My students roll their eyes when I mention them. I roll my eyes when I mention them. Eat your vegetables, floss your teeth, remember to say, “It’s difficult to quantitatively assess the relative contributions of genes and environment to a particular trait when they interact.” This suggests a radical conclusion: it’s not meaningful to ask what a gene does, just what it does in a particular environment. This is summarized wonderfully by the neurobiologist Donald Hebb: “It is no more appropriate to say things like characteristic A is more influenced by nature than nurture than . . . to say that the area of a rectangle is more influenced by its length than its width.” It’s appropriate to figure out if lengths or widths explain more of the variability in a population of rectangles. But not in individual ones.

  —

  As we conclude part 2 of this chapter, some key points:

  A gene’s influence on the average value of a trait (i.e., whether it is inherited) differs from its influence on variability of that trait across individuals (its heritability).

  Even in the realm of inherited traits—say, the inheritance of five fingers as the human average—you can’t really say that there is genetic determination in the classically hard-assed sense of the word. This is because the inheritance of a gene’s effect requires not just passing on the gene but also the context that regulates the gene in that manner.

  Heritability scores are relevant only to the environments in which the traits have been studied. The more environments you study a trait in, the lower the heritability is likely to be.

  Gene/environment interactions are ubiquitous and can be dramatic. Thus, you can’t really say what a gene “does,” only what it does in the environments in which it’s been studied.

  Current research actively explores gene/environment interactions.37 How’s this for fascinating: Heritability of various aspects of cognitive development is very high (e.g., around 70 percent for IQ) in kids from high–socioeconomic status (SES) families but is only around 10 percent in low-SES kids. Thus, higher SES allows the full range of genetic influences on cognition to flourish, whereas lower-SES settings restrict them. In other words, genes are nearly irrelevant to cognitive development if you’re growing up in awful poverty—poverty’s adverse effects trump the genetics.* Similarly, heritability of alcohol use is lower among religious than nonreligious subjects—i.e., your genes don’t matter much if you’re in a religious environment that condemns drinking. Domains like these showcase the potential power of classical behavior genetics.

  PART 3: SO WHAT DO GENES ACTUALLY HAVE TO DO WITH BEHAVIORS WE’RE INTERESTED IN?

  The Marriage of Behavior Genetics and Molecular Genetics

  Behavior genetics has gotten a huge boost by incorporating molecular approaches—after examining similarities and differences between twins or adoptees, find the actual genes that underlie those similarities and differences. This powerful approach has identified various genes relevant to our interests. But first, our usual caveats: (a) not all of these findings consistently replicate; (b) effect sizes are typically small (in other words, some gene may be involved, but not in a major way); and (c) the most interesting findings show gene/environment interactions.

  Studying Candidate Genes

  Gene searches can take a “candidate” approach or a genomewide association approach (stay tuned). The former requires a list of plausible suspects—genes already known to be related to some behavior. For example, if you’re interested in a behavior that involves serotonin, obvious candidate genes would include those coding for enzymes that make or degrade serotonin, pumps that remove it from the synapse, or serotonin receptors. Pick one that interests you, and study it in animals using molecular tools to generate “knockout” mice (where you’ve eliminated that gene) or “transgenic” mice (with an extra copy of the gene). Make manipulations like these only in certain brain regions or at certain times. Then examine what’s different about behavior. Once you’re convinced of an effect, ask whether variants of that gene help explain individual differences in human versions of the behavior. I start with the topic that has gotten the most attention, for better or worse, mostly “worse.”

  The Serotonin System

  What do genes related to serotonin have to do with our best and worst behaviors? Plenty.

  Chapter 2 presented a fairly clear picture of low levels of serotonin fostering impulsive antisocial behavior. There are lower-than-average levels of serotonin breakdown products in the bloodstreams of people with that profile, and of serotonin itself in the frontal cortex of such animals. Even more convincingly, drugs that decrease “serotonergic tone” (i.e., decreasing serotonin levels or sensitivity to serotonin) increase impulsive aggression; raising the tone does the opposite.

  This generates some simple predictions—all of the following should be associated with impulsive aggression, as they will produce low serotonin signaling:

  Low-activity variants of the gene for tryptophan hydroxylase (TH), which makes serotonin

  High-activity variants of the gene for monoamine oxidase-A (MAO-A), which degrades serotonin

  High-activity variants of the gene for the serotonin transporter (5HTT), which removes serotonin from the synapse

  Variants of genes for serotonin receptors that are less sensitive to serotonin

  An extensive literature shows that for each of those genes the results are inconsistent and generally go in the opposite direction from “low serotonin = aggression” dogma. Ugh.

  Studies of genes for TH and serotonin receptors are inconsistent messes.38 In contrast, the picture of 5HTT, the serotonin transporter gene, is consistently in the opposite direction from what’s expected. Two variants exist, with one producing less transporter protein, meaning less serotonin removed from the synapse.* And counter to expectations, this variant, producing more serotonin in the synapse, is associated with more impulsive aggression, not less. Thus, according to these findings, “high serotonin = aggression” (recognizing this as simplified shorthand).

  The clearest and most counterintuitive studies concern MAO-A. It burst on the scene in a hugely influential 1993 Science paper reporting a Dutch family with an MAO-A gene mutation that eliminated the protein.39 Thus serotonin isn’t broken down and accumulates in the synapse. And counter to chapter 2’s predictions, the family was characterized by varied antisocial and aggressive behaviors.

  Mouse studies in which the MAO-A gene was “knocked out” (producing the equivalent of the Dutch family’s mutation) produced the same—elevated serotonin levels in the synapse and hyperaggressive animals with enhanced fear responses.40


  This finding, of course, concerned a mutation in MAO-A resulting in the complete absence of the protein. Research soon focused on low-activity MAO-A variants that produced elevated serotonin levels.*41 People with that variant averaged higher levels of aggression and impulsivity and, when looking at angry or fearful faces, more activation of the amygdala and insula and less activation of the prefrontal cortex. This suggests a scenario of more fear reactivity and less frontal capacity to restrain such fear, a perfect storm for reactive aggression. Related studies showed decreased activation of frontal cortical regions during various attentional tasks and enhanced anterior cingulate activity in response to social rejection in such individuals.

  So studies where serotonin breakdown products are measured in the body, or where serotonin levels are manipulated with drugs, say that low serotonin = aggression.42 And the genetic studies, particularly of MAO-A, say high serotonin = aggression. What explains this discrepancy? The key probably is that a drug manipulation lasts for a few hours or days, while genetic variants have their effects on serotonin for a lifetime. Possible explanations: (a) The low-activity MAO-A variants don’t produce higher synaptic levels of serotonin all that consistently because the 5HTT serotonin reuptake pump works harder at removing serotonin from the synapse, compensating, and maybe even overcompensating. There is evidence for this, just to make life really complicated. (b) Those variants do produce chronically elevated serotonin levels in the synapse, but the postsynaptic neurons compensate or overcompensate by decreasing serotonin receptor numbers, thereby reducing sensitivity to all that serotonin; there is evidence for that too. (c) The lifelong consequences of differences in serotonin signaling due to gene variants (versus transient differences due to drugs) produce structural changes in the developing brain. There is evidence there as well, and in accordance with that, while temporarily inhibiting MAO-A activity with a drug in an adult rodent decreases impulsive aggression, doing the same in fetal rodents produces adults with increased impulsive aggression.

  Yikes, this is complicated. Why go through the agony of all these explanatory twists and turns? Because this obscure corner of neurogenetics has caught the public’s fancy, with—I kid you not—the low-activity MAO-A variant being referred to as the “warrior gene” by both scientists and in the media.*43 And that warrior hoo-hah is worsened by the MAO-A gene being X linked and its variants being more consequential in males than females. Amazingly, prison sentences for murderers have now been lessened in at least two cases because it was argued that the criminal, having the “warrior gene” variant of MAO-A, was inevitably fated to be uncontrollably violent. OMG.

  Responsible people in the field have recoiled in horror at this sort of unfounded genetic determinism seeping into the courtroom. The effects of MAO-A variants are tiny. There is nonspecificity in the sense that MAO-A degrades not only serotonin but norepinephrine as well. Most of all, there is nonspecificity in the behavioral effects of the variants. For example, while nearly everyone seems to remember that the landmark MAO-A paper that started all the excitement was about aggression (one authoritative review referred to the Dutch family with the mutation as “notorious for the persistent and extreme reactive aggression demonstrated by some of its males”), in actuality members of the family with the mutation had borderline mental retardation. Moreover, while some individuals with the mutation were quite violent, the antisocial behavior of others consisted of arson and exhibitionism. So maybe the gene has something to do with the extreme reactive aggression of some family members. But it is just as responsible for explaining why other family members, rather than being aggressive, were flashers. In other words, there is as much rationale for going on about the “drop your pants gene” as the “warrior gene.”

  Probably the biggest reason to reject warrior-gene determinism nonsense is something that should be utterly predictable by now: MAO-A effects on behavior show strong gene/environment interactions.

  This brings us to a hugely important 2002 study, one of my favorites, by Avshalom Caspi and colleagues at Duke University.44 The authors followed a large cohort of children from birth to age twenty-six, studying their genetics, upbringing, and adult behavior. Did MAO-A variant status predict antisocial behavior in twenty-six-year-olds (as measured by a composite of standard psychological assessments and convictions for violent crimes)? No. But MAO-A status coupled with something else powerfully did. Having the low-activity version of MAO-A tripled the likelihood . . . but only in people with a history of severe childhood abuse. And if there was no such history, the variant was not predictive of anything. This is the essence of gene/environment interaction. What does having a particular variant of the MAO-A gene have to do with antisocial behavior? It depends on the environment. “Warrior gene” my ass.

  This study is important not just for its demonstration of a powerful gene/environment interaction but for what the interaction is, namely the ability of an abusive childhood environment to collaborate with a particular genetic constitution. To quote a major review on the subject, “In a healthy environment, increased threat sensitivity, poor emotion control and enhanced fear memory in MAOA-L [i.e., the “warrior” variant] men might only manifest as variation in temperament within a ‘normal’ or subclinical range. However, these same characteristics in an abusive childhood environment—one typified by persistent uncertainty, unpredictable threat, poor behavioral modeling and social referencing, and inconsistent reinforcement for prosocial decision making—might predispose toward frank aggression and impulsive violence in the adult.” In a similar vein, the low-activity variant of the serotonin transporter gene was reported to be associated with adult aggressiveness . . . but only when coupled with childhood adversity.45 This is straight out of the lessons of the previous chapter.

  Since then, this MAO-A variant/childhood abuse interaction has been frequently replicated, and even demonstrated with respect to aggressive behavior in rhesus monkeys.46 There have also been hints as to how this interaction works—the MAO-A gene promoter is regulated by stress and glucocorticoids.

  MAO-A variants show other important gene/environment interactions. For example, in one study the low-activity MAO-A variant predicts criminality, but only if coupled with high testosterone levels (consistent with that, the MAO-A gene also has a promoter responsive to androgens). In another study low-activity MAO-A participants in an economic game were more likely than high-activity ones to retaliate aggressively when exploited by the other player—but only if that exploitation produced a large economic loss; if the loss was small, there was no difference. In another study low-activity individuals were more aggressive than others—but only in circumstances of social exclusion. Thus the effects of this genetic variant can be understood only by considering other, nongenetic factors in individuals’ lives, such as childhood adversity and adult provocation.47

  The Dopamine System

  Chapter 2 introduced the role of dopamine in the anticipation of reward and in goal-directed behavior. Lots of work has examined the genes involved, most broadly showing that variants that produce lowered dopamine signaling (less dopamine in the synapse, fewer dopamine receptors, or lower responsiveness of these receptors) are associated with sensation seeking, risk taking, attentional problems, and extroversion. Such individuals have to seek experiences of greater intensity to compensate for the blunted dopamine signaling.

  Much of the research has focused on one particular dopamine receptor; there are at least five kinds (found in different parts of the brain, binding dopamine with differing strengths and duration), each coded for by a gene.48 Work has focused on the gene for the D4 dopamine receptor (the gene is called DRD4), which mostly occurs in the neurons in the cortex and nucleus accumbens. The DRD4 gene is super variable, coming in at least ten different flavors in humans. One stretch of the gene is repeated a variable number of times, and the version with seven repeats (the “7R” form) produces a receptor protein that is sparse in the cortex and relatively u
nresponsive to dopamine. This is the variant associated with a host of related traits—sensation and novelty seeking, extroversion, alcoholism, promiscuity, less sensitive parenting, financial risk taking, impulsivity, and, probably most consistently, ADHD (attention-deficit/hyperactivity disorder).

  The implications cut both ways—the 7R could make you more likely to impulsively steal the old lady’s kidney dialysis machine, or to impulsively give the deed of your house to a homeless family. In come gene/environment interactions. For example, kids with the 7R variant are less generous than average. But only if they show insecure attachment to their parents. Secure-attachment 7Rs show more generosity than average. Thus 7R has something to do with generosity—but its effect is entirely context dependent. In another study 7R students expressed the least interest in organizations advocating prosocial causes, unless they were given a religious prime,* in which case they were the most prosocial. One more—7Rs are worse at gratification-postponement tasks, but only if they grew up poor. Repeat the mantra: don’t ask what a gene does; ask what it does in a particular context.49

  Interestingly, the next chapter considers the extremely varied frequency of the 7R variant in different populations. As we’ll see, it tells you a lot about the history of human migration, as well as about differences between collectivist and individualist cultures.50

  We shift now to other parts of the dopamine system. As introduced in chapter 2, after dopamine binds to receptors, it floats off and must be removed from the synapse.51 One route involves its being degraded by the enzyme catechol-O-methyltransferase (COMT). Among the variants of the COMT gene is one associated with a more efficient enzyme. “More efficient” = better at degrading dopamine = less dopamine in the synapse = less dopamine signaling. The highly efficient COMT variant is associated with higher rates of extroversion, aggression, criminality, and conduct disorder. Moreover, in a gene/environment interaction straight out of the MAO-A playbook, that COMT variant is associated with anger traits, but only when coupled with childhood sexual abuse. Intriguingly, the variants seem pertinent to frontal regulation of behavior and cognition, especially during stress.