In addition to degradation, neurotransmitters can be removed from the synapse by being taken back up into the axon terminal for recycling.52 Dopamine reuptake is accomplished by the dopamine transporter (DAT). Naturally, the DAT gene comes in different variants, and those that produce higher levels of synaptic dopamine (i.e., transporter variants that are less efficient) in the striatum are associated with people who are more oriented toward social signaling—they’re drawn more than average to happy faces, are more repelled by angry faces, and have more positive parenting styles. How these findings merge with the findings from the DRD4 and COMT studies (i.e., fitting risk taking with a preference for happy faces) is not immediately apparent.
Cool people with certain versions of these dopamine-related genes are more likely to engage in all sorts of interesting behaviors, ranging from the healthy to the pathological. But not so fast:
These findings are not consistent, no doubt reflecting unrecognized gene/environment interactions.
Again, why should the COMT world be related to sensation seeking, while there are the DAT people and their happy faces? Both genes are about ending dopamine signaling. This is probably related to different parts of the brain differing as to whether DAT or COMT plays a bigger role.53
The COMT literature is majorly messy, for the inconvenient reason that the enzyme also degrades norepinephrine. So COMT variants are pertinent to two totally different neurotransmitter systems.
These effects are tiny. For example, knowing which DRD4 variant someone has explains only 3 to 4 percent of the variation in novelty-seeking behavior.
The final piece of confusion seems most important but is least considered in the literature (probably because it would be premature). Suppose that every study shows with whopping clarity and consistency that a DRD4 variant is highly predictive of novelty seeking. That still doesn’t tell us why for some people novelty seeking means frequently switching their openings in chess games, while for others it means looking for a new locale because it’s getting stale being a mercenary in the Congo. No gene or handful of genes that we are aware of will tell us much about that.
The Neuropeptides Oxytocin and Vasopressin
Time for a quick recap from chapter 4. Oxytocin and vasopressin are involved in prosociality, ranging from parent/offspring bonds to monogamous bonds to trust, empathy, generosity, and social intelligence. Recall the caveats: (a) sometimes these neuropeptides are more about sociality than prosociality (in other words, boosting social information gathering, rather than acting prosocially with that information); (b) they most consistently boost prosociality in people who already lean in that direction (e.g., making generous people more generous, while having no effect on ungenerous people); and (c) the prosocial effects are within groups, and these neuropeptides can make people crappier to outsiders—more xenophobic and preemptively aggressive.
Chapter 4 also touched on oxytocin and vasopressin genetics, showing that individuals with genetic variants that result in higher levels of either the hormones or their receptors tend toward more stable monogamous relationships, more actively engaged parenting, better skill at perspective taking, more empathy, and stronger fusiform cortex responses to faces. These are fairly consistent effects of moderate magnitude.
Meanwhile, there are studies showing that one oxytocin receptor gene variant is associated with extreme aggression in kids, as well as a callous, unemotional style that foreshadows adult psychopathy.54 Moreover, another variant is associated with social disconnection in kids and unstable adult relationships. But unfortunately these findings are uninterpretable because no one knows if these variants produce more, less, or the usual amount of oxytocin signaling.
Of course, there are cool gene/environment interactions. For example, having a particular oxytocin receptor gene variant predicts less sensitive mothering—but only when coupled with childhood adversity. Another variant is associated with aggression—but only when people have been drinking. Yet another variant is associated with greater seeking of emotional support during times of stress—among Americans (including first generation Korean Americans) but not Koreans (stay tuned for more in the next chapter).
Genes Related to Steroid Hormones
We start with testosterone. The hormone is not a protein (none of the steroid hormones are), meaning there isn’t a testosterone gene. However, there are genes for the enzymes that construct testosterone, for the enzyme that converts it to estrogen, and for the testosterone (androgen) receptor. The most work has focused on the gene for the receptor, which comes in variants that differ in their responsiveness to testosterone.*
Intriguingly, a few studies have shown that among criminals, having the more potent variant is associated with violent crimes.55 A related finding concerns sex differences in structure of the cortex, and adolescent boys with the more potent variant show more dramatic “masculinization” of the cortex. An interaction between receptor variant and testosterone levels occurs. High basal testosterone levels do not predict elevated levels of aggressive mood or of amygdaloid reactivity to threatening faces in males—except in those with that variant. Interestingly, the equivalent variant predicts aggressiveness in Akita dogs.
How important are these findings? A key theme in chapter 4 was how little individual differences in testosterone levels in the normal range predict individual differences in behavior. How much more predictability is there when combining knowledge of testosterone levels and of receptor sensitivity? Not much. How about hormone levels and receptor sensitivity and number of receptors? Still not much. But definitely an improvement in predictive power.
Similar themes concern the genetics of the estrogen receptor.56 For example, different receptor variants are associated with higher rates of anxiety among women, but not men, and higher rates of antisocial behavior and conduct disorder in men, but not women. Meanwhile, in genetically manipulated mice, the presence or absence of the receptor gene influences aggression in females . . . depending on how many brothers there were in the litter in utero—gene/environment again. Once again, the magnitude of these genetic influences is tiny.
Finally, there is work on genes related to glucocorticoids, particularly regarding gene/environment interactions.57 For example, there is an interaction between one variant of the gene for a type of receptor for glucocorticoids (for mavens: it’s the MR receptor) and childhood abuse in producing an amygdala that is hyperreactive to threat. Then there is a protein called FKBP5, which modifies the activity of another type of receptor for glucocorticoids (the GR receptor); one FKBP5 variant is associated with aggression, hostility, PTSD, and hyperreactivity of the amygdala to threat—but only when coupled with childhood abuse.
Buoyed by these findings, some researchers have examined two candidate genes simultaneously. For example, having both “risk” variants of 5HTT and DRD4 synergistically increases the risk of disruptive behavior in kids—an effect exacerbated by low socioeconomic status.58
Phew; all these pages and we’ve only gotten to thinking about two genes and one environmental variable simultaneously. And despite this, things still aren’t great:
The usual—results aren’t terribly consistent from one study to the next.
The usual—effect sizes are small. Knowing what variant of a candidate gene someone has (or even what variants of a collection of genes) doesn’t help much in predicting their behavior.
A major reason is that, after getting a handle on 5HTT and DRD4 interactions, there are still roughly 19,998 more human genes and a gazillion more environments to study. Time to switch to the other main approach—looking at all those 20,000 genes at once.
Fishing Expeditions, Instead of Looking Where the Light Is
The small effect sizes reflect a limitation in the candidate gene approach; in scientific lingo, the problem is that one is only looking where the light is. The cliché harks back to a joke: You discover someone at night, searching the ground under a street l
amp. “What’s wrong?” “I dropped my ring; I’m looking for it.” Trying to be helpful, you ask, “Were you standing on this side or that side of the lamp when you dropped it?” “Oh, no, I was over by those trees when I dropped it.” “Then why are you searching here?” “This is where the light is.” With candidate gene approaches, you look only where the light is, examine only genes that you already know are involved. And with twenty thousand or so genes, it’s pretty safe to assume there are still some interesting genes that you don’t know about yet. The challenge is to find them.
The most common way of trying to find them all is with genomewide association studies (GWAS).59 Examine, say, the gene for hemoglobin and look at the eleventh nucleotide in the sequence; everyone will pretty much have the same DNA letter in that spot. However, there are little hot spots of variability, single nucleotides where, say, two different DNA letters occur, each in about 50 percent of the population (and where this typically doesn’t change the amino acid being specified, because of DNA redundancy). There are more than a million of such “SNPs” (single-nucleotide polymorphisms) scattered throughout the genome—in stretches of DNA coding for genes, for promoters, for mysterious DNA junk. Collect DNA from a huge number of people, and examine whether particular SNPs associate with particular traits. If an SNP that’s implicated occurs in a gene, you’ve just gotten a hint that the gene may be involved in that trait.*
A GWAS study might implicate scads of genes as being associated with a trait. Hopefully, some will be candidate genes already known to be related to the trait. But other identified genes may be mysterious. Now go check out what they do.
In a related approach, suppose you have two populations, one with and one without a degenerative muscle disease. Take a muscle biopsy from everyone, and see which of the ~20,000 genes are transcriptionally active in the muscle cells. With this “microarray” or “gene chip” approach, you look for genes that are transcriptionally active only in diseased or in healthy muscle, not in both. Identify them, and you have some new candidate genes to explore.*
These fishing expeditions* show why we’re so ignorant about the genetics of behavior.60 Consider a classic GWAS that looked for genes related to height. This was a crazy difficult study involving examining the genomes of 183,727 people. 183,727. It must have taken an army of scientists just to label the test tubes. And reflecting that, the paper reporting the findings in Nature had approximately 280 authors.
And the results? Hundreds of genetic variants were implicated in regulating height. A handful of genes identified were known to be involved in skeletal growth, but the rest was terra incognita. The single genetic variant identified that most powerfully predicted height explained all of 0.4 percent—four tenths of one percent—of the variation in height, and all those hundreds of variants put together explained only about 10 percent of the variation.
Meanwhile, an equally acclaimed study did a GWAS regarding body mass index (BMI). Similar amazingness—almost a quarter million genomes examined, even more authors than the height study. And in this case the single most explanatory genetic variant identified accounted for only 0.3 percent of the variation in BMI. Thus both height and BMI are highly “polygenic” traits. Same for age of menarche (when girls menstruate for the first time). Moreover, additional genes are being missed because their variants are too rare to be picked up by current GWAS techniques. Thus these traits are probably influenced by hundreds of genes.61
What about behavior? A superb 2013 GWAS study examined the genetic variants associated with educational attainment.62 The usual over-the-top numbers—126,559 study subjects, about 180 authors. And the most predictive genetic variant accounted for 0.02 percent—two hundredths of one percent—of the variation. All the identified variants together accounted for about 2 percent of the variation. A commentary accompanying the paper contained this landmark of understatement: “In short, educational attainment looks to be a very polygenic trait.”
Educational attainment—how many years of high school or college one completes—is relatively easy to measure. How about the subtler, messier behaviors that fill this book’s pages? A handful of studies have tackled that, and the findings are much the same—at the end, you have a list of scores of genes implicated and can then go figure out what they do (logically, starting with the ones that showed the strongest statistical associations). Hard, hard approaches that are still in their infancy. Made worse by a GWAS missing more subtly variable spots,* meaning even more genes are likely involved.63
—
As we conclude this section, some key points:64
This review of candidate genes barely scratches even the surface of the surface. Go on PubMed (a major search engine of the biomedical literature) and search “MAO gene/behavior”—up come more than 500 research papers. “Serotonin transporter gene/behavior”—1,250 papers. “Dopamine receptor gene/behavior”—nearly 2,000.
The candidate gene approaches show that the effect of a single gene on a behavior is typically tiny. In other words, having the “warrior gene” variant of MAO probably has less effect on your behavior than does believing that you have it.
Genomewide survey approaches show that these behaviors are influenced by huge numbers of genes, each one playing only a tiny role.
What this translates into is nonspecificity. For example, serotonin transporter gene variants have been linked to risk of depression, but also anxiety, obsessive-compulsive disorder, schizophrenia, bipolar disorder, Tourette’s syndrome, and borderline personality disorder. In other words, that gene is part of a network of hundreds of genes pertinent to depression, but also part of another equally large and partially overlapping network relevant to anxiety, another relevant to OCD, and so on. And meanwhile, we’re plugging away, trying to understand interactions of two genes at a time.
And, of course, gene and environment, gene and environment.
CONCLUSIONS
At long last, you (and I!) have gotten to the end of this excruciatingly but necessarily long chapter. Amid all these tiny effects and technical limitations, it’s important to not throw out the genetic baby with the bathwater, as has been an agitated sociopolitical goal at times (during my intellectual youth in the 1970s, sandwiched between the geologic periods of Cranberry Bell-bottoms and of John Travolta White Suits was the Genes-Have-Nothing-to-Do-with-Behavior Ice Age).
Genes have plenty to do with behavior. Even more appropriately, all behavioral traits are affected to some degree by genetic variability.65 They have to be, given that they specify the structure of all the proteins pertinent to every neurotransmitter, hormone, receptor, etc. that there is. And they have plenty to do with individual differences in behavior, given the large percentage of genes that are polymorphic, coming in different flavors. But their effects are supremely context dependent. Ask not what a gene does. Ask what it does in a particular environment and when expressed in a particular network of other genes (i.e., gene/gene/gene/gene . . . /environment).
Thus, for our purposes, genes aren’t about inevitability. Instead they’re about context-dependent tendencies, propensities, potentials, and vulnerabilities. All embedded in the fabric of the other factors, biological and otherwise, that fill these pages.
Now that this chapter’s done, why don’t we all take a bathroom break and then see what’s in the refrigerator.
Nine
Centuries to Millennia Before
Let’s start with a seeming digression. Parts of chapters 4 and 7 have debunked some supposed sex differences concerning the brain, hormones, and behavior. One difference, however, is persistent. It’s far from issues that concern this book, but bear with me.
A remarkably consistent finding, starting with elementary school students, is that males are better at math than females. While the difference is minor when it comes to considering average scores, there is a huge difference when it comes to math stars at the upper extreme of the distribution. For example, i
n 1983, for every girl scoring in the highest percentile on the math SAT, there were eleven boys.
Why the difference? There have always been suggestions that testosterone is central. During development, testosterone fuels the growth of a brain region involved in mathematical thinking, and giving adults testosterone enhances some math skills. Oh, okay, it’s biological.
But consider a paper published in Science in 2008.1 The authors examined the relationship between math scores and sexual equality in forty countries (based on economic, educational, and political indices of gender equality; the worst was Turkey, the United States was middling, and, naturally, the Scandinavians were tops). Lo and behold, the more gender equal the country, the less of a discrepancy in math scores. By the time you get to the Scandinavian countries, it’s statistically insignificant. And by the time you examine the most gender-equal country on earth at the time, Iceland, girls are better at math than boys.*
L. Guiso et al., “Culture, Gender, and Math,” Sci 320 (2008): 1164.
Visit bit.ly/2o88s4O for a larger version of this graph.
In other words, while you can never be certain, the Afghan girl pictured on top, on the next page, seated next to her husband, is less likely than the Swedish girl pictured below her to solve the Erdös-Hajnal conjecture in graph theory.
Behave: The Biology of Humans at Our Best and Worst Page 27