WHAT makes us unhealthy? Sometimes the answer is easy. In 1854 an English doctor, , founded modern epidemiology by mapping a cholera outbreak in London and noticing that sick people lived near one public water pump. In the now classic version of the story, he had the pump’s handle removed and the outbreak ended.
Often, though, the true links between sickness and environmental exposures are more subtle. Now, a novel method promises a revolution to rival Snow’s, by using our knowledge of human genes to pinpoint how the environment and our lifestyle choices affect health.
“The true links between sickness and our environmental exposures are very subtle”
Advertisement
Last week, the method burst into the public arena, with the finding that moderate drinking during pregnancy lowers the future IQ of a developing fetus (PLoS One, ).
Previous studies that charted mothers’ drinking and children’s IQs have given conflicting results. That’s because they use classic epidemiology, which involves surveying a large group of women who drink during pregnancy and another similar group who don’t, then comparing their kids’ IQs. Sounds simple, but what if mothers who drink have other traits in common that non-drinkers don’t have, and it is one or more of those traits, rather than the alcohol, that actually influences IQ? In fact, it turns out that women who drink moderately during pregnancy are more likely to have a higher IQ, and this is passed to the child, obscuring any effect of drinking.
Such traits are called confounders. You can correct for them, or try to make them equal in both groups, so the women being compared differ only in their drinking habits – but shared traits that you are unaware of can still cloud the results. “All epidemiological studies are confounded somehow,” says epidemiologist at University College London.
The answer, says Ron Gray at the University of Oxford, a co-author of the new study, is to use genes as a proxy for drinking.
For example, the gene for alcohol dehydrogenase (ADH), an enzyme that breaks down alcohol, can contain small mutations called single-nucleotide polymorphisms (SNPs). These slow the enzyme’s action. Humans carry several dehydrogenase genes and differ in the number of SNPs that they contain – from none to 10.
The more of these mutations that a fetus has, the more slowly it will break down the alcohol that its mother drinks. So if alcohol reduces IQ, the IQ should be lower in children with more mutations, and that in turn should be independent of social or other variables. The team, led by of the University of Bristol, UK, studied these mutations in the children of more than 4000 mothers. Those who had four or more of the SNPs averaged 3.5 IQ points lower on a standard test than children with 2 SNPs or fewer – but that was only the case if their mothers drank 1 to 6 units per week during the pregnancy (see “What do a few IQ points mean?“). Genes made no difference to IQ if the mothers did not drink. “This avoids confounders, and shows it must be alcohol that is making the difference,” says Gray.
Alcohol is just the start. at the University of Bristol, UK, and colleagues proposed in 2003 that epidemiologists might find genes that affect the extent of your exposure to environmental and dietary chemicals, or affect your own fatness or blood-sugar level. If so, those genetic variants could then be used as a proxy for the exposure itself in assessing how it leads to disease.
The method is called Mendelian randomisation (MR). Genes inherited randomly, according to Mendel’s laws, are also inherited independently of other things that affect the disease in question – i.e. confounders.
Davey Smith’s group found that having one or another mutation in 5000 genes that affect the extent of your exposure to potential causes of disease bares no relation to many confounding factors, such as your social class, say, or whether you eat cheese.
In other words, studies of obesity’s links to disease can be confounded by obese individuals’ tendency to also drink and smoke, but using a specific gene mutation linked to obesity instead of obesity itself means the mutation’s effect on a disease should reveal the impact of obesity on that disease, and nothing else. And now, studies of human DNA have revealed useful mutations to focus on. “In 2003 MR was hypothetical,” says Davey Smith. “Now, it’s going mainstream.” For the first time, the American Society for Human Genetics annual conference at its meeting in San Francisco this month.
The technique is starting to clear up all sorts of conflicted epidemiology. For years, people thought raising high-density lipoprotein – commonly known as “good cholesterol” – would protect from heart attack, says Davey Smith. But a large trial of a drug that raises good cholesterol, , saw no such benefit. MR could have saved the expense: a study of 15 genes that affect cholesterol metabolism found that people with SNPs causing low or high levels of good cholesterol have the same heart attack risk. This suggests there was some confounder in the traditional epidemiological studies that concluded good cholesterol lowered the risk of heart attack.
Another big question is also on its way to being answered: how is being fat bad for you? The makes a protein associated with obesity and fat mass. It predisposes a person to obesity and accounts for some of the variations in fatness between people. Along with 30 more genetic proxies for obesity, it could reveal how being overweight influences diseases such as cancer.
The next phase, says Davey Smith, is to use MR to sort through thousands of chemical markers in the “metabolome” – all the chemical intermediates in a metabolic process, which can now be identified. How genetic variants change those, and also affect disease, should show which metabolic processes cause disease.
The point, he says, is not to show how genes cause disease, which was the hope of our early efforts to study human DNA. “It is to use genes to show what parts of our behaviour or environment make us ill.”
“Now we know that moderate alcohol during pregnancy does affect IQ, why risk it?”
What do a few IQ points mean?
“Research is showing that even small differences in IQ correlate with socio-economic achievement,” says Ron Gray at the University of Oxford, who last week showed that moderate drinking in pregnancy can reduce the future IQ of a developing fetus.
The study doesn’t show how much higher IQs would be if no alcohol was consumed, only that faster rather than slower breakdown of alcohol prevented 3.5 points worth of damage. The total damage is unlikely to be very much greater, says Gray.
How might this influence a child’s future? Studies that attempt to tease out the effect of IQ on health struggle with confounders (see main story). But at the University of Edinburgh, UK, says that in studies that account for this, 15 more IQ points made young people 24 per cent less likely to die in the next 20 years, and 14 per cent less likely to develop high blood pressure, obesity and diabetes.
The effect is about the same size as the impact of smoking on the same diseases, and doesn’t appear to be influenced by social class.
So what effect might that glass of champagne during a second trimester have had? “There’s no way we could determine whether such a low dose has an effect,” says Gray. “But now we know moderate alcohol does affect IQ. If you have a choice, why risk it?”
