Control, p.14
Control, page 14
This, by the way, was Francis Galton’s idea. In 1874, he invented twin studies when he sent out a questionnaire to various hospitals around the United Kingdom in order to “collect data for estimating the respective shares that ‘Nature’ and ‘Nurture’ ordinarily contribute to the body and mind of adults, meaning by ‘nature’ everything that is inborn, and by ‘nurture,’ every influence subsequent to birth.”
It was a brilliant innovation, typical of Galton.§ He received ninety-four responses, and over the years since, twin studies have become a cornerstone for understanding human genetics and behavior. A further development came in the form of identical twins who had been separated. Identical twins may have near identical genomes, but when they live together in a family, their environments are incredibly similar too. When not raised together, however, the social cues may well be more different than the genetic similarities, so similarities and differences in behaviors can be more readily partitioned into genes or environment. Twin studies have been a core part of genetics for decades now, and I don’t wish to go into much detail for the purposes of this discussion. But it is worth noting that, powerful though they can be, they are not without significant problems. Because twins share the environments in which they are raised, just as brothers and sisters do, extracting the social from the genetic is not clear-cut. As for the relatively few separated twins projects over the years, many have been shown to be deeply methodologically flawed, or in a handful of cases possibly fraudulent.¶
Separated twins are often raised in very similar socioeconomic environments, and always at the same time, so the environment may be less different than might be assumed. In some cases, the children were separated years after birth, so had developed together; in others, they lived near each other, were raised by different members of the same family, and even attended the same schools. Some separated twins had plenty of contact with each other as they were growing up, and others were later reunited and even lived together. In principle, separated twins studies are the perfect way of partitioning nature and nurture. But in the messy reality of lived lives, these studies are often deeply flawed. Hereditarians often lean heavily on twin studies to emphasize the genetic over the environmental, to swing the pendulum toward nature, and away from nurture.
Nevertheless, well-conducted twin studies remain a valid research tool, and historically served to reinforce the genetic component of heredity—we are not born blank slates. Instead, our canvas is already penciled from the moment sperm meets egg.
E PLURIBUS UNUM
By the early 2000s, it was becoming clear that single genes could not explain the wondrous sophistication of human beings, nor the cluttered mess of complex diseases. We began to develop techniques that could unearth the great maelstrom of genetic influence over complex traits. Finding individual genes had been arduous work, and so a new approach was adopted after the completion of the Human Genome Project (HGP). This grand scientific endeavor set out to read the entirety of a single person’s DNA: genes, on-and-off switches, the repeats, the fossils, the junk, the repeats, all of it. It was a magnificent challenge, and came in under budget—about $3 billion, that is one dollar for every letter of DNA—in the summer of 2000 with the first draft sequence. Since then, it has been refined over and over again, until finally, in July 2021, a full annotated complete human reference genome was published, twenty-one years after President Bill Clinton’s portentous announcement that we had finally created “the most wondrous map ever produced by humankind . . . the language in which God created life.”
Mixed metaphors and clumsy religious allusions aside, the finer detail of our genetic makeup was now available to everyone. Galton would have loved it. It is the true metric of human similarities and difference, a colossal data set bulging with the most personal information possible, bearing the keys to unlock the patterns of inheritance and evolution, free for anyone to mine.
One of the many crucial technological leaps that came with the HGP was the ready availability of genetic data. We could sequence genes faster, cheaper and more accurately than ever before. Which meant more data, which meant more answers, which meant more questions. Soon after came the invention of a radical and revolutionary new way of digging around in DNA: genome-wide association studies (GWAS). Instead of hunting down a specific gene, you could instead take a group of people, the more the better, with the same trait or disease. By comparing their whole genomes, it is possible to look for bits within those twenty-three pairs of chromosomes that appear more similar to one another than to a control group who don’t share the trait of interest.
Then you lay out the results on a graph, and see what stands out. We call them Manhattan plots, because they can resemble the New York skyline, with highs and lows, each building representing a location in our genome. The higher a peak, the more likely the DNA sitting at that location is specifically involved in the trait you’re pursuing. If, for example, you did a GWAS for Huntington’s disease, you’d get a large prominent spike in chromosome 4, because that is where the HTT gene dwells, and people with that dreadful disease all have a mutated HTT gene. For more complex diseases and traits, we find dozens or even hundreds of spikes, each indicating that the DNA in that location is probably relevant to the disease and trait. GWAS helped identify thousands of places in the human genome that are potentially important, that vary between people in populations, and that are associated with disease. With this innovation, we could unravel the genetic architecture that underlies every disease, not just the ones that have clear individual genes.
During the years of the Third Reich, schizophrenia was a specific diagnosis isolated for purging under the Nazis’ euthanasia and eugenics programs, which began in 1933, and escalated with Aktion T4. The number of patients is not easy to verify, and the diagnoses are similarly opaque, but reasonable estimates from historians who have tried to account for the murderous chaos of Nazi Germany put the number of schizophrenia patients who were either sterilized or murdered at somewhere between 220,000 and 269,500. This was largely down to the work of Ernst Rüdin, who crafted the sterilization laws of the Reich but escaped prosecution with only minor inconvenience. In the 1930s, Rüdin worked alongside Franz Kallmann, a German Jew who fled Nazi Germany in 1935, and became a leading geneticist in the United States.# Together they developed twin studies with a focus on psychiatric illnesses, and just like Charles Davenport and so many others, concluded that schizophrenia was caused by a single gene.
They could not have been more wrong. Around 1 in 200 Americans today suffer with schizophrenia. It’s a mental disorder characterized by episodes of psychosis, disorganized thinking, hallucinations and other deeply troubling symptoms. It’s also highly heritable, meaning that the majority of the differences we see in risk for schizophrenia are genetic in origin—we know this from twin and family studies, which show that for example if one identical twin has schizophrenia then the other has a 40 percent chance of suffering it too. Indeed, the highest risk factor for schizophrenia is having a first-degree family member with it. So this means that there should be identifiable genetic differences in people with and without schizophrenia. The GWAS is a perfect method for finding these differences, and many studies with tens of thousands of schizophrenic patients have been performed over the last few years to track down the underlying genetics. Most recently, in 2018, one study found 145 individual DNA differences (that is, single letter changes—an A instead of a T, a G instead of a C, etc.) across the whole genome, meaning that there are at least 145 genetic variables that positively contribute to the probability of having schizophrenia.
Unlike Rüdin and Kallman’s assertion that schizophrenia was a monogenic disease, schizophrenia is in fact a hugely polygenic disorder—dozens of genes are involved, and none is causative. It is only in aggregation that these variants make up the increased risk for schizophrenia, which accounts for just a fraction of the proportion that is genetic and not environmental.
Regardless of this caveat, knowing that DNA is involved is an important step—thumbtacks planted in the map of the genome. When we discover these genetic variants, some of the questions that logically follow are: What is the function of that bit of DNA? Is it in a gene, or in a part of the genome that switches genes on or off? If so, what does that gene do? Why would the variation predispose someone ever so slightly toward having this terrible disease? These questions remain mostly unanswered, because we just don’t know enough about human genetics and neuroscience yet. Or because we don’t know how genotype and phenotype are related to each other. But as you might expect with a psychological disorder, many of those differences reside in genes involved in the brain—though it’s worth noting that maybe half of all human genes are involved in some way in our big expansive brains.
Recall the first rule of behavioral genetics: “everything is heritable.” Another one is the fourth: “A typical human behavioral trait is associated with very many genetic variants, each of which accounts for a very small percentage of the behavioral variability.” Psychiatric conditions and psychological traits are polygenic, meaning that genes are involved, and they are many, but the different versions of genes play a cumulative but small role in the risk of the outcome. Schizophrenia showcases this well: the 145 bits of DNA found so far in schizophrenia patients represent a fraction of the variants that will be detected, for two reasons. First, these are only the common variants, the ones that can be detected in large groups of pooled patients; the more we look, the more rare variants we will find. But that is work still to be done, and for now, we can say with certainty that we have found only a small percentage of the genetic risk of schizophrenia. The second reason is that schizophrenia affects people all over the world, but different populations will reveal different variants for the same disease, and as yet our samples are somewhat skewed toward people of European descent—we simply haven’t collected the genomes representing most people on Earth.
The proportion of heritability that can be identified is only a small percentage of the overall risk, which is also modulated by rare, still undiscovered genetic variants, and all of the social, cultural and environmental influences as well. It is perfectly possible to have every one of those 145 genetic variants, and never show the slightest sign of schizophrenia.
Alcoholism was another behavior that was targeted specifically for eugenic purification and it exemplifies this point too. Alcohol use disorder or alcohol dependency are contemporary and more precise diagnoses, and we know that they are heritable, because everything is. We even know that certain genes involved in the metabolism of alcohol are associated with addiction to it. In the latest studies (2018), GWAS identified eighteen different genetic risk factors for alcoholism, which accounts for a small proportion of the heritability, as per rule 4. But drinking alcohol is, of course, entirely socially mediated. You could have every single one of the genetic risk factors for alcoholism and never become an addict if you don’t drink alcohol. The inverse is also true: you can have none of the risk factors and still become an alcoholic. These variants are specific to populations and environments. We might look for the same ones in different populations, but may well find that the phenotype of polygenic traits is not the same even though the genotype is.
Schizophrenia and alcoholism were both conditions specifically targeted by the Nazis. But let’s go back further and consider the founding trait of eugenics, the one that Galton championed above all when taking his first baby steps into the field he founded: intelligence.
Few subjects cause as much ire and toxicity as discussions of intelligence and of its inheritance. The reasons for this are myriad, but they most certainly include the fact that intelligence—notably in the form of early IQ testing in the United States—was a significant criterion for eugenic intervention. Nevertheless, and I will be brief, there are certain things we can say about intelligence that should be uncontroversial.
The first is that it is measurable. Intelligence may be a difficult thing to define and, more broadly, cognitive abilities include many aspects of our behaviors, including reasoning skills, problem solving, abstract thought and learning capability. The IQ test, as originally conceived by Binet and Simon in France, and then translated and modified in America by Henry Goddard (the author of the 1912 Kallikak feeblemindedness folly), are nowadays standardized and designed to test reasoning, knowledge, mental processing speed and spatial awareness. The tests are culturally biased; this is well understood, and these days responsible psychologists try to account for that. As with all human traits, cognitive abilities are not evenly distributed. IQ across a group of people falls into a pattern known as a normal distribution, aka a bell curve, with the population average being set at 100 points, and around two-thirds of people being within 15 IQ points in either direction; 1 in 40 people is above 130 or below 70. There are other measures of cognitive ability, such as educational attainment—that is, how many years you stay in formal education. All the different metrics tend to correlate pretty well with one another.
It is a flawed system, but we do understand how it’s flawed, and it’s the best we have. IQ is a reflection of current abilities, and can change during life, including if you practice doing IQ tests. It is not immutable, nor is it absolute. It is a single metric for a complex range of abilities, but so is a drivers license, or a university degree classification. Like all behaviors, IQ is heritable. We are not blank slates for any of our behaviors, and intelligence is most certainly not exempt from that rule. One of the strengths of IQ as a metric is that it has been tested so many times over the last hundred years that we have masses of data on it, and when it comes to heritability, what we find is that about half of the variation we see in a population is down to genetic differences between people. That is not the same as saying that intelligence is half genetic and half environmental; it is that in any given population, there will be a range between the top scorers and the bottom, and we can sensibly attribute half of that difference to DNA, meaning that it is encoded in our differing genomes. This is scientifically uncontroversial.
And so, as with schizophrenia, heart disease, height and any other trait you might be interested in, nowadays we can find the actual bits of DNA that are influencing that difference. The genome-wide association studies for cognitive abilities started big, became huge, and are now gargantuan. The first major GWAS on cognitive abilities (in 2013, this time via the metric of educational attainment) featured 126,559 people and it uncovered 3 single-letter genetic changes of significance. Three years later, the sample size had doubled, but the genetic landmarks of interest had gone up to 74. Or there was the landmark 2018 study that had 269,867 participants and found genomic locations of note in 1,016 genes. Or the other 2018 landmark paper that had 300,486 individuals and found 148 genetic markers and 709 genes. Or maybe the big daddy, also in 2018, when the number was 1.1 million people and 1,271 places in the genome that were associated with cognitive abilities.
Finally, after a hundred years of searching, we had found the 709 genes associated with general intelligence. Or the 1,016 genes. Or whatever the correct number turns out to be. I am not disparaging this research; they are all terrific studies involving huge amounts of work, and powerfully demonstrate several things very clearly: (1) a lot of genes are involved in cognitive abilities; (2) we don’t know what almost all of them do, but they do many things in many different tissues; (3) the differences we have found so far account for only a small amount of the variance in cognitive abilities, which means (4) we still have a lot of work to do if we really want to pin down the genetic architecture behind our minds.
Genome-wide associations are a fabulous tool for analyzing our DNA, but they are just a tool, and all tools can be misused. GWAS also provided much of the perpetuation of the myths of genetic determinism, where a single gene is misattributed as the cause of something messy and complex. The thousands of news reports I mentioned in Part 1, where headlines claimed, “Scientists have found the gene for . . . ,” were almost always prompted by this type of study. Even that most august journal Scientific American, even in the era when polygenic traits were well-known and widely discussed, published a headline in 2016 “ ‘Schizophrenia Gene’ Discovery Sheds Light on Possible Cause.”
At least Scientific American had the good grace to put it in scare quotes. The A in GWAS stands for “association,” meaning that the piece of DNA that sits at the bottom of a skyscraper in those Manhattan plots is associated with the trait. It doesn’t say that it causes it, nor does it say how it works, what it does or even what that bit of the genome is doing.
Furthermore (and this is really getting buried in the haystack of human genetics), although we have sequenced millions of people’s genomes, they do not represent a complete read. Instead, we look at the bits that are important and the variations that are common. These common variants are useful because they are informative when looking across populations. So, when we account for the influence of genetics for a trait in your sample, we’re almost certainly looking only at the common variants, the ones we know many people have. For many traits and diseases, some, much or most of the real genetic variance has not yet been identified because it is harbored in rare genetic variants that we are yet to discover. Every human genome is absolutely unique, and statistically, that remains true for the entire history and future of humankind. The possibilities are endless.
This is why genomics is a growth field, and why the completion of the Human Genome Project was only the beginning of the era of the genome. This is why when identifying genes or gene variants that influence traits such as cognition, we’re accounting for a small proportion of the genetic influence, each variant of which plays a minuscule role, which is associated with the trait, and measurable only at a population level.
