Genetic Hyping

Spring is in the air, the new fashion season is upon us, and once again genes are all the rage. A great example is a recent report by a group of scientists at Princeton, led by Joe Tsien, which was published in Nature, one of the two most prestigious and influential general science journals in the world. The scientists did some molecular biology magic with some mice—engineering them so that neurons in one part of their brains had an extra copy of a particular gene. The neurons there made abnormally large amounts of the protein coded for by that gene, a protein that is part of a receptor for a neurotransmitter that appears to play a key role in learning and memory. And remarkably, the animals scored significantly higher than ordinary laboratory mice on an array of memory tests. The mice, it seemed, were genetically engineered to be abnormally smart.

This was great science: important subject, slick techniques, careful documentation. And some good marketing whimsy by the scientists as well, who called the mouse Doogie, after television series wunderkind D. Howser of some years back, who was so precocious that he had graduated med school by age fourteen.

The whole thing made a big splash with the media. Desk editors, who had exhausted every possible pun about Dolly the cloned sheep, had to find clever ways to work Doogie into the headline. Pundits erupted with the obligatory essays about whether parents should want their children to be Doogie mice in time for preschool entrance exams. And Time magazine, which at least showed some restraint in placing a question mark after the “The IQ Gene” headline, made it a cover story.

That’s great. But I’m not here to go on about the Doogie mouse. Instead, I want to focus on another paper about genes and behavior that was published around the same time in the equally prestigious journal Science. That paper, in contrast, attracted little notice from the media, and what attention it did get seemed wrongly directed. In fact, the commentaries managed to completely miss the point.

Genes, of course, have plenty to do with behavior. Genes determine your intelligence and your personality, and certain genetic profiles cause criminality, alcoholism, and a proclivity toward misplacing car keys. Hopefully, three essays into this book, you know this is a total crock, medieval genetic determinism. Genes don’t cause behaviors. Sometimes, they influence them.

With that out of the way, we can flout our sophistication. Genes influence behavior, environment influences behavior, and genes and environment interact—a point that I’m hammering at over and over. What that means is that the effects of a gene on an organism will usually vary with changes in the environment, and the effects of environment will vary with changes in the genetic makeup of the organism.

I say usually because a powerful influence from one side of the interaction can overwhelm the other. In the realm of intellect, for example, even the most salutary environment will not compensate for the catastrophic consequences of, say, the genetic makeup that leads to Tay-Sachs disease. And conversely, some environmental influence can overwhelm the effects of genetics. Even the most impressive genetic pedigree of smarts isn’t going to do you much good if you’re subjected to severe and prolonged protein malnutrition during childhood. But in the less extreme realm, genes and environment happily interact, achieving a balance.

The cleanest way to study gene/environment interactions is to hold one-half of the interaction constant, modify the other half systematically, and then see what happens. Manipulating the environment can be relatively straightforward—all of us have known about that ever since our mothers objected to the friends we were hanging out with back when. But the controlled and selective manipulation of genes is hot stuff, the world of Web site headlines and twentysomething molecular-biology geeks becoming gazillionaires when their biotech companies go public. Newfound techniques of the genetic trade—inserting into an animal a gene from a different species, to create a so-called transgenic animal; replacing one of an animal’s own genes with a nonfunctional version, to make a “knockout” animal; even selectively mutating one of an animal’s genes—are flashy and exciting.

In recent years, molecular biologists have manipulated the genes in mice that code for neurotransmitters (the chemicals that carry messages between brain cells), as well as the genes for neurotransmitter receptors (molecules that reside on the surface of a cell and react to incoming neurotransmitters). Altering those genes, biologists have found, can affect such aspects of mouse behavior as sexuality, aggressiveness, risk-taking, substance abuse, and more. Is it such a jump to infer that the same link between genes and behavior could exist for people?

But often, on closer examination, it turns out that the evidence supporting the asserted links between genes and behavior is slight. For example, as mentioned in the previous essay, starting around 1996, a series of studies was published linking a particular gene in humans to novelty-seeking behavior, and the media went wild over this. However, these studies showed, collectively, that this gene accounted for only about 5 percent of the variability in the data.

Now, people tend to crave—and consequently overvalue—virtually anything new. The result is a pretty widespread impression among the lay public, who (through no fault of their own) learn their science in ten-second sound bites, that it takes dramatic and extreme environments to blunt the influence of genes.

This is where that study published in Science comes in. No Time cover story, no catchy mouse nicknames. The study was a collaboration among three behavioral geneticists: John Crabbe, of the Veterans Affairs Medical Center and Oregon Health Sciences University, both in Portland; Douglas Wahlsten, at the University of Alberta in Edmonton; and Bruce Dudek, at the State University of New York in Albany. Crabbe and colleagues had a modest goal: they wanted to standardize the various tests that have been devised to measure the effects of genes on such mouse behaviors as alcohol craving and anxiety. The investigators’ aim was to identify tests that would measure the effects accurately enough to give results that were highly reproducible from one lab to the next.

To do so, the team created uniform conditions in their three labs. First, each investigator used groups of mice from the same eight strains (a strain is a pedigree of mice in which close relatives are mated with each other for umpteen generations, until eventually the animals are about as alike as identical twins). Some were control strains; others had undergone some kind of fancy genetic manipulation, such as having a gene knocked out. The key point is that these strains had already been studied. It was common knowledge, for instance, that Strain X was your basic, off-the-rack strain used in many labs, Strain Y was more prone than other mice to drinking alcohol when it was offered, Strain Z tended to be anxious, and so forth.

Once the experimenters were sure they had acquired identical strains of mice, they took steps to make sure the mice were raised in standardized conditions. No unnoticed advantage or disadvantage—a more delicious brand of food, say, or a particularly dirty cage—was to be allowed that might cause the mice to act differently from one another for reasons that had nothing to do with genes. Finally, the experimenters chose six standardized behavioral tests—tests that trapped the mice in mazes, forced them to swim to safety, or imposed some other task whose success or failure is readily measurable.

That was the game plan. But the execution was an obsessive’s heaven. Crabbe, Wahlsten, and Dudek did cartwheels to make sure that these animals were tested in identical environments in all three labs. They standardized every element of the process—from the way the animals were raised to the way the tests were conducted to the equipment that was deployed. For example, because some of the mice were born in the lab but others came from commercial breeders, the homegrowns were taken for a bouncy van ride to simulate the jostling that commercially bred mice undergo during shipping, just in case that sort of stressor had an effect.

The team tested animals of exactly the same age (to the day) on the same date at the same local time. Animals had been weaned at the same age, all their mothers had been weighed at the same time. They all lived in the same type of cage, with the same brand and thickness of sawdust bedding, which was changed on the same day of the week. When handled, it was at the same time and by human hands in the same type of surgical glove. They were fed the same food, kept in the same lighting environment, at the same temperature. And when their tails were marked for identification, it was always with a Sharpie pen. The environments of these animals could hardly have been more similar if Crabbe, Wahlsten, and Dudek had been identical triplets separated at birth.

What the three geneticists created was a world of genetically indistinguishable mice raised in virtually identical environments. If genes were all-powerful and deterministic, one might expect that there would be complete replicability of scores within and between labs. All the animals of Strain X would have gotten six points on test one, twelve points on test two, eight points on test three, and so on, regardless of which lab they were tested in. The mice from Strain Y would also perform in a uniform manner, getting, say, nine points on test one, fifteen points on test two, and so on. Such a result would constitute convincing proof that genes are massively deterministic of behavior…at least for the genes in question…in these mice…on those particular tests.

But that’s absurd—no one would have expected anything as extreme as the precise same results on a test from each animal. Instead, the expectation would have been something close: perhaps all the animals of Strain X would get roughly similar scores on test number one in all three labs—a statistical dead heat. And that’s precisely what occurred for some of the strains, when they were administered some of the tests. In one test (the most impressive example), nearly 80 percent of the variability in the data across all three labs could be explained by genetics alone. But the truly critical finding was that for some of the tests, the results gave no support to the assertion that genes make mice what they are, let alone make us who we are. In fact, the results on those tests were sheer chaos—the same strain differed radically from lab to lab (though the results within labs were mostly uniform).

Just to give you an example of the sorts of numbers these guys got in some of these cases, take a strain with the uncuddly name of 129/SvEvTac, and a test in which the effects of cocaine on a mouse’s level of activity is measured. In Portland, cocaine caused these mice to increase their activity an average of 667 centimeters of movement per fifteen minutes. In Albany, an increase of 701. Pretty good—similar result. And in Edmonton? More than 5,000 centimeters of activity, by genetically identical mice in a meticulously similar environment. That’s like a set of triplets competing in a pole-vaulting competition. They’ve all had the same amount of training, all have the same night’s rest, the same breakfast, all are wearing the same brand of underwear. The first two manage 18-feet and 18-feet-1-inch pole vaults, and then the third one launches himself 108 feet into the air.

Now, there might be some ways to explain these discrepancies. One might breathe a sigh of relief, for instance, if all the data were utterly random—if the results for any given test within a given strain within a given lab were so variable from mouse to mouse that no pattern could be detected. Then you could be reasonably sure that the tests must be lousy and poorly defined, or there weren’t enough animals tested to begin to discern patterns, or maybe Crabbe and buddies don’t actually know squat about the arcana of mouse behavioral testing. But some of the data, as I noted, were quite similar within tests, within strains, and within labs. These guys knew what they were doing with their mice.

Another possibility is that some of the results differed from site to site because of the nature of the places themselves. Maybe the mice in Albany differed from the mice at the other two labs because they were dispirited by the architecture of the hideous state capitol (on account of early environmental influences as a native of New York City, I’m obliged to consider Albany a dive). Maybe proximity to those amber Canadian waves of grain in Edmonton would do something systematic to those mice. But no, that couldn’t be it either, because the discrepancies in the data across all tests were not systematically attributable to any of the labs.

And a third possible explanation: perhaps the difference in behavior within strains of mice was merely a matter of degree. Suppose some mouse strain is known to exhibit an atypically large amount of Behavior X. Maybe the problem was that at sites one and two, those mice showed vastly more of Behavior X than did the control mice, where at site three, they showed only a little bit more than the controls did. But no, the data were far more chaotic than that: for certain tests, the strain in question showed more of Behavior X than the controls did at one test site, the same amount as the controls at the next site, and less than the controls at the third.

Or, a fourth possibility: perhaps the environmental conditions were not, as some of the critics suggested, as perfectly synchronized as they seemed to be. A group of scientists wrote to Science to suggest that the size and texture of the mouse-chow pellets might have been at fault. Another group argued that the key uncontrolled variable was that the graduate student who oversaw the testing in Edmonton was allergic to mice and so wore a space-suit-like protective helmet—they went on to advance a rather exuberant hypothesis about the possible interactions between behavioral genetics and the ultrasound emitted by the motor on the helmet’s air filter. And, yes, it turns out that there was, indeed, a crucial slipup in all the careful controls: the colors of the Sharpies used for marking the animals were inconsistent—some were black, others were red. Could that have been the extreme environmental influence that skewed the results?

Excuse my facetiousness, but I am troubled by the fact that all too frequently, investigators are reluctant to reject their dearly held preconceptions and allow their expectations to impose blinders. When the Crabbe team’s paper was published, it was accompanied by a commentary written by one of the journal’s staff writers, under the title of “Fickle Mice Highlight Test Problems.” In it, the writer bemoans how hard it will be to deal with the problem of tests that don’t give the expected result.

This seems all turned around to me. If the behavioral tests fail to show a reliable genetic effect, the first conclusion that jumps to mind shouldn’t be that the tests need some fixing. If environmental variables that are too subtle to be detected in a study as thorough as this can markedly disrupt a genetic effect on a behavior, then there’s not much of a genetic influence going on here. Or maybe none at all.

The moral is that one should not get too excited about some new genetic component of behavior until the effect has been replicated in a number of different places and with a broad array of tests—something that is seldom done. Instead, what happens is this: A team of scientists do some fancy molecular tinkering in a batch of mice. They manipulate a gene relevant to the brain, and, well, after all that impressive work, something must be different about the animals. Test them, and lo and behold, some behavior does turn out to vary in a statistically significant way on one test. Aha, an effect, a splashy publication, and when the next lab can’t replicate it, the onus of proof can easily shift to identifying their “test problem.” That scenario has been played out for many of the wonder genes. The conclusion must be that many published accounts linking groups of genes to specific behaviors could well be off base.

Don’t get me wrong and overestimate how much I’m trying to bash genes. Genetics influences neurobiology, behavior, every facet of biology, and to extraordinary extents in some cases. The data in this study demonstrate it pretty convincingly for some of the strains and behaviors. There’s just the danger of expectations running away with you, even among the supposedly hard-nosed science community. It is most certainly not the case that this new genetic emperor has no clothes.

But amid our current near-feverish interest in genes, especially among the lay public, it’s worth noting that the emperor is a bit less accessorized than usually assumed. The environment, even a subtle one, can still more than hold its own in the biological interactions that shape who we are.

NOTES AND FURTHER READING

The development of the Doogie mouse was reported in Tang Y, Shimizu E, Dube G, Rampon C, Kerchner G, Zhuo M, Liu G, and Tsien J, “Genetic enhancement of learning and memory in mice,” Nature 410 (1999): 63. The paper by Crabbe and colleagues was “Genetics of mouse behavior: interactions with laboratory environment,” Science 284 (1999): 1670. The “Fickle Mice Highlight Test Problems” commentary, by Enserink M, can be found in the same issue, page 1,599. The Crabbe paper documents some of the remarkably detailed efforts they made at standardizing conditions in the three different labs. Further information was provided in a Web site that they set up (www.albany.edu/psy/obssr). The letters published in response to the Crabbe paper can be found in Science (1999): 285, 2,067–70.

The demonstration that the gene related to novelty-seeking behavior accounts for only about 5 percent of the variability in the data in humans can be found in Ebstein R and Belmaker R, “Saga of an adventure gene: novelty seeking, substance abuse and the dopamine D4 receptor 9D4DR) exon III repeat polymorphism,” Molecular Psychiatry 2 (1997): 361.

Demoralizing postscript: A few years after the Crabbe et al. paper was published, I found myself sitting in the office of a Nobel laureate, whose work should have made him cognizant of this science. This was one huge alpha male of a baboon, and I was terrified of him, wildly intimidated. No doubt the stress of the situation was gravely impairing the executive functioning of my frontal cortex (stay tuned for what that’s about) and my ability to make a prudent decision, because I decided to bring up the Crabbe paper. “So what did you think of that Crabbe paper in Science?” I ventured enthusiastically. A blank look. “You know, the paper where they were testing the different mouse strains at the three different labs…?” I offered. A cold, blank look. I was dumbfounded—he seemed not to have heard of the study, which hadn’t, after all, been published in some biology newsletter in Estonian. I launched into a description of the methods and results of the study. He made a snarfly sort of exhaling sound through his nostrils and said something to the effect of, “It sounds like they don’t know the first thing about how to do basic behavioral testing.” Thank God, my time was soon up with him and I was allowed to scuttle out of his office before my imprudent frontal cortex had allowed me to say what I thought of his attitude about an inconvenient scientific finding.