Despite the success of genome-wide association studies, many experts rightly point out the need to remember that most common diseases have a significant environmental component. For example, cigarette smoking increases one’s risk of lung cancer 10- to 20-fold, according to the National Cancer Institute. But other studies have found that even after controlling for smoking, having first-degree relatives with lung cancer increases one’s own risk by a modest amount, suggesting at least a small genetic component.

A trio of papers published today in Nature and Nature Genetics has found a SNP on chromosome 15 that is linked to lung cancer risk. Two of the studies, by Hung et al. and Amos et al., matched lung cancer cases and healthy controls for smoking behavior and compared their genotypes across 300,000 SNPs. Both studies found that one copy of the A version of rs1051730 increased subjects’ odds of lung cancer by about 1.3 times compared to those with none; having two copies increased subjects’ odds by 1.8 times. (23andMe users can see their genotype at rs1051730 in the Genome Explorer (now called Browse Raw Data).)

A 1.8-fold increase may not seem like much compared to the 10- to 20-fold increase seen for smokers. But if this genetic effect is independent of the effect of smoking, smokers with the AA genotype might find themselves at 18- to 36-fold increased risk compared to nonsmokers with the GG genotype.

A third study questions whether the SNP’s effect on lung cancer risk is truly independent of smoking. Thorgeirsson et al. found evidence that the same SNP, rs1051730, is actually linked to nicotine dependence. On average, each copy of the A version increased the number of cigarettes subjects smoked by about one cigarette per day. The authors also reported that most of the increased risk of lung cancer conferred by the A version was due to its effect on smoking. Thus, an environmental component itself appears to have a genetic component.

However, the effect of rs1051730 on smoking quantity was seen only if a subject smoked at all. Genotype at rs1051730 wasn’t connected with whether subjects smoke, only with how much they smoke once they start.

So in the nature vs. nurture debate, perhaps there is room for free will after all.

ShareThis

Three studies published Monday in Nature Genetics report new associations between SNPs and common diseases. Two of the studies find new SNP associations for colorectal cancer, while the third addresses Type 2 Diabetes.

All three studies relied upon very large sample sizes, which increased the chances that the researchers would find SNPs that individually have a very modest effect on the risk of disease, but in the right combinations may greatly increase the odds a person would be affected.

Colorectal Cancer
Colorectal cancer is the third most common cancer (excluding skin cancers) and the second leading cause of cancer-related deaths in the United States. The average lifetime risk of developing colorectal cancer is about 5% in Western European and North American populations.

In Tomlinson et al, researchers from several large consortiums used a multi-phase approach to narrow in on SNPs significantly associated with colorectal cancer. Using three phases that examined about 18,800 people with colorectal cancer and 18,400 controls (all of European ancestry) the scientists identified two new SNPs associated with colorectal cancer – rs10795668 and rs16892766.

When the researchers split the individuals with colorectal cancer into more specific groups, they found some evidence that the riskier version of rs10795668 increased the odds of developing rectal cancer more compared to the odds of developing colon cancer. The effect of rs16892766 was significantly stronger in cases of colorectal cancer that arose before the age of 60.

The second study, by Tenesa et al, also took a multi-phase approach for finding SNPs associated with colorectal cancer. In four phases of testing that examined about 17,500 people with colorectal cancer and 16,400 controls (all of European ancestry except for one group from Japan) they found one new SNP associated with colorectal cancer and replicated two other associations that had been found in previous studies.

In people with European ancestry, the C version of SNP rs3802842 increased the odds of colorectal cancer. In the Japanese population, however, the SNP was not significantly related to the odds of developing colorectal cancer.

But if the researchers took into account whether a patient had rectal or colon cancer (instead of combining the two types of cancer into one category), they found that the C version of rs3802842 did increase the risk for developing rectal cancer in the Japanese population.

The A version rs7014346 and the C version rs4939827 were also found to be associated with an increased risk for colorectal cancer. These two SNPs had been found in previous studies. The riskier versions of both of these SNPs increased the risk for colorectal cancer in both the Japanese and European populations.

Tenesa et al calculated that a person with two copies of the riskier version for each of the three SNPs (rs3802842, rs7014346, and rs4939827) – a situation only expected to happen in 0.5% of the population – would have 2.6 times the odds of developing colorectal cancer as someone with two copies of the less risky version at all three SNPS.

Summary of New Colorectal Cancer Data:
23andMe customers can check out their data at each of the SNPs newly associated with colorectal cancer in the Genome Explorer (now called Browse Raw Data).

In the chart below, each SNP’s rsid number is followed by the version of the SNP that is less prevalent in the population. The two columns labeled “effect” show the change in odds for developing colorectal cancer when either one or two copies of the less prevalent version of the SNP are present compared with the odds of developing the disease when two copies of the other version of the SNP are present. The data in this chart only apply to people with European ancestry.

(Note that for rs10795668, the less prevalent version actually protects against colorectal cancer – that’s why the change in odds is less than one. For rs4939827 the two versions are about equal in European populations.)

Type 2 Diabetes
Type 2 Diabetes occurs when chronically high blood sugar levels cause a breakdown of the body’s natural response to sugar. It can result in kidney failure, blindness, and circulatory problems that increase the risk of heart attack or stroke. In the United States, almost 21 million children and adults have diabetes, but the rate of new diagnoses is increasing.

Several SNPs have already been associated with Type 2 Diabetes (check out the Gene Journal (now called Health and Traits) article), but the established associations only explain a small proportion of the genetic component of the disease. Studies done to date may have been too small to find SNPs with very modest effects. SNPs with small effects, however, may be key to capturing the contribution that genes make to Type 2 Diabetes and many other common diseases.

Zeggini et al pooled the data from three large studies that previously examined SNPs and Type 2 Diabetes (more than 10,000 people total). The researchers then followed up on their findings in another 57,000 people (about 14,000 with Type 2 Diabetes).

The research replicated several SNPs that were found to be associated with Type 2 Diabetes in previous, smaller studies. But this new study also found six new associations. Two of these are available to 23andMe customers through the Genome Explorer (Browse Raw Data): rs7578597 and rs10923931, which is not included in the 23andMe Personal Genome Service but is equivalent to rs2793831, which is.

In the chart, each SNP’s rsid number is followed by the version of the SNP that is less prevalent in the population. The two columns labeled “effect” show the change in odds for developing Type 2 Diabetes when either one or two copies of the less prevalent version of the SNP are present compared with the odds of developing the disease when two copies of the other version of the SNP are present. So far these data apply only to people of European ancestry.

ShareThis

There are two simple ways to train animals to perform a new behavior: the carrot and the stick. The carrot rewards desired, “correct” behaviors, while the stick punishes “errors.” Over time, this scheme results in preference for positively reinforced behaviors, and avoidance of negatively reinforced ones. In the December 7 issue of Science, German scientists reported that the SNP rs1800497 is associated with how well subjects learn to avoid errors in a simple reward/punishment scheme. (What the paper calls the “A1-allele” of the “DRD2-TAQ-IA polymorphism” is actually what we store as the A version of the rs1800497 SNP, according to OMIM and and MutDB.)

Though we do not yet consider this finding solid enough to include in the Gene Journal (now called Health and Traits), 23andMe users can still examine their genotype at the rs1800497 using the Genome Explorer (now called Browse Raw Data). In the study, people with the GG genotype appeared to avoid choices for which they had received negative feedback, while those with the AG or AA genotypes did not seem to avoid those choices.

(There are currently no genetic associations for whether people prefer to be rewarded with actual carrots, although someone is clearly thinking about this question.)

Caveats

1) Study size. The authors only tested 12 subjects with the GG or AG genotype and 14 subjects with the AA genotype. This makes it more likely that the result could be a fluke (though the authors also present functional data to support their conclusions). At the very least, the training/testing portion of the study should be repeated in a much larger group. The same goes for other DRD2 association studies. 2) Real-world relevance. The test was highly abstract and the reinforcement simplistic and binary—this type of learning may not be relevant to real-world situations. 3) Multiple hypothesis testing. Other researchers have looked at whether this SNP is associated with behavioral traits from alcohol dependence to creativity, usually in undersized studies. If you run enough tests, something will eventually look like a positive hit just by chance.

(After the jump: detailed methods and smiley/scary faces.)

You Have Chosen…Poorly

How do you find out how well people learn? Give them a test where they have no idea what the correct answers are—and then reward or punish their choices!

During the training phase of this experiment, subjects took a series of trials. In each trial, they were presented with a pair of abstract symbols and asked to choose one. The subjects received immediate feedback about their choices: either a smiling face as a positive reinforcement (reward), or a frowning face as negative reinforcement (punishment).

The trials were repeated many times, using three different training pairs comprising six symbols—AB, CD, and EF. Of the possible choices, the most-rewarded choice was “A”, which received a happy face 80% of the time and a frowning face 20% of the time. The least-rewarded choice (and thus most punished) was “B”, which received a happy face only 20% of the time, and a frowning face 80% of the time. The reward schedule and the actual symbols used are in the figure to the right.

In the second phase, subjects were again asked to choose between two symbols. Although the same six symbols were used, in this phase the pairs did not include the three from the training set (i.e. BE and FD were shown, but not AB). The researchers measured preference for reward by recording the number of times the subjects chose “A” when it was offered. They measured avoidance of punishment by recording the number of times the subjects chose the non-”B” symbol when “B” was offered.

Those with the GG genotype at this SNP—avoiders—chose the non-”B” symbol 70% of the time. But those with the AA or AG genotypes—non-avoiders—chose the non-”B” symbol only 50% of the time, which is no different from choosing at random. The difference between avoiders and non-avoiders was of moderate statistical significance.Interestingly, there was no statistically significant difference between the two groups on preference for “A”—only for avoiding “B”.

Mechanism

The SNP occurs inside a gene called ANKK1, but scientists believe that the SNP acts on a neighboring gene, DRD2. The DRD2 gene encodes a receptor in the brain that responds to the neurotransmitter dopamine. At least two previous studies have shown that people with the AG and AA genotypes have lower levels of the dopamine D2 receptor in their brains. It’s possible that variation at the SNP affects how DRD2 is turned on and off, or that it is tightly linked to another SNP that does.

Klein et al. (2007) also used fMRI images of the brain to show that avoiders and non-avoiders had different activity levels in a zone of the brain thought to be involved in learning from errors. They suggest that “reduced dopamine D2 receptor density is associated with reduced capacity to learn negative characteristics of a stimulus from negative feedback.” In other words, if it’s true that people with the AG or AA genotypes make less of the dopamine D2 receptor, this might explain the reduction in activity they observed.

The authors also cite studies showing that lower dopamine D2 receptor density is linked to addiction, obesity, and novelty-seeking behaviors such as compulsive gambling, which suggests that avoiders in this study might also be prone to these behaviors. Interestingly, another study suggests that people with the same rs1800497 genotype as the non-avoiders (AA or AG) have higher verbal creativity.

ShareThis

Most people who have resolved to give up the bottle after a bout of New Year’s Eve overindulgence went cold turkey for a few weeks with no ill effects (even if by now they have gone back to their old habits). But for some people who are alcohol-dependent, quitting booze in the first place can result in severe symptoms, such as withdrawal seizures and delirium tremens. A French study in the January issue of Alcoholism: Clinical and Experimental Research found evidence that two SNPs in a single gene were linked with how likely alcohol-dependent people were to have withdrawal seizures. Because this finding is of modest statistical significance, and because it only affects a small fraction of people (about 3% of alcohol-dependent people), we do not currently include this report in the Gene Journal (now called Health and Traits).

However, 23andMe users can look at their genotypes in the Genome Explorer (now called Browse Raw Data). The two SNPs identified in the study are rs27072 and rs27048. The study found that alcohol-dependent people with the CC genotype at either SNP have just under twice the odds of having withdrawal seizures as those with CT or TT.

Caveats

1) Small study size. As with many studies of behavioral traits, small sample sizes can lead to evidence for a conclusion that are actually statistical flukes. 2) Multiple hypothesis testing. The authors performed at least three types of analyses on eight markers. There is reason to wonder if these results were due to a fluke—run enough tests and you’ll eventually get a hit. However, seeing statistical significance in more than one type of analysis, along with previous studies showing association between withdrawal symptoms and the DAT1 gene (though not with the SNPs reported here), lend support to the hypothesis that there is a real association, albeit possibly a small one.

(After the jump: detailed background and a dated cultural reference.)

Quitting the Sauce

Alcohol dependence is thought to be heritable because it clusters in families. Yet there have not been any conclusive links between the disease and genetic markers. Some of this lack of progress is because past studies have suffered from statistical biases and small sample size. For a rather extreme example of the problem with the latter, think about flipping a coin twice and coming up with heads twice. Would you be justified in concluding that the coin is double-headed? (The answer is no.)

Another issue is that alcohol dependence is a heterogeneous disorder whose definition is in flux. Most people have heard the term alcoholism, which the American Medical Association defines as “a primary, chronic disease characterized by impaired control over drinking, preoccupation with the drug alcohol, use of alcohol despite adverse consequences, and distortions in thinking.”

But the psychiatrists’ bible, the DSM-IV, recently split its diagnoses into alcohol abuse and alcohol dependence. The former is defined as repeated drinking despite negative consequences (for example, on work or relationships), while the latter is defined as alcohol abuse in addition to tolerance and withdrawal: the classic signs of addiction. However, not all withdrawal symptoms—cravings, hallucinations, the shakes, seizures, and delirium tremens—need to occur to justify a diagnosis of alcohol dependence. Furthermore, the nature of addiction itself could be heterogeneous—physical, psychological, or both—meaning that the environmental and biochemical (and thus potentially genetic) events that lead to alcohol dependence can differ from person to person.

For example, take people who wear watches. Some do because they are obsessive about being on time, others do because they like stylish accessories, while a third group might do it to cover up the tan line on their wrist. Add to this heterogeneity an expansion of the category of “wearing a watch” to “knowing what time it is”—which could include carrying a pocketwatch, a cell phone, or standing next to Flavor Flav—and it becomes a nightmare to find a common factor.

Le Strat et al. (2008) try to reduce heterogeneity by looking for a genetic link to a specific symptom in alcohol-dependent patients recruited from French treatment clinics—in our example, whittling down the definition to just those people who wear digital watches. Withdrawal seizures are a potentially life-threatening side effect that occur when someone who is alcohol dependent quits drinking. Yet not all alcohol dependents have seizures. Withdrawal seizures are thus a sufficient, but not necessary, symptom of alcohol dependence.

The authors examined a single candidate gene, DAT1, based on previous (though inconclusive) evidence that different DAT1 variations are linked to alcohol dependence and withdrawal symptoms. DAT1 encodes a protein that plays a role in the signaling of the neurotransmitter dopamine. Dopamine receives a lot of attention in this field because of its well-accepted role in rewarding behavior. In particular, mice engineered to lack the DAT1 gene avoid alcohol, although it is not clear whether dopamine is playing a role in rewarding alcohol intake, the negative effects of withdrawal, or both.

Three different types of analyses were performed for each SNP (and six other markers), and the authors found that having two copies of the C version of the SNPs rs27072 and rs27048 increased one’s chance of having withdrawal seizures. The SNPs mentioned here reached statistical significance in two of the analyses performed, but only moderately.

ShareThis

One of the much-heralded claims of the post-genomic era is that personal genetic information may allow medical treatments to be tailored to an individual’s genetic makeup. While we are not there yet, a trio of new studies on heart disease may just have brought us one step closer. The studies were the product of a collaboration between researchers at Harvard Medical School and Celera (the company that helped spur completion of the public Human Genome Project).

In the first study of over 25,000 women, those with the GG or AG genotypes at the SNP rs20455 were found to have a slightly increased chance of having a heart attack—about 34% higher than those with the AA genotype. (This study confirmed at least two other preliminary findings that this SNP, which is in the gene KIF6, is associated with heart disease.)

A second study of about 3,000 subjects further confirmed the first finding, and also examined whether the SNP was associated with the outcome of treatment with the cholesterol-lowering drug pravastatin. As other studies have reported, those with the GG or AG genotypes were found to be at higher risk of having a heart attack than those with the AA genotype. But when treated with pravastatin, the risk of heart attack for people with GG or AG was substantially reduced—approximately equal to that of the lower-risk AA group. (Pravastatin treatment had little or no effect on those with the AA genotype.)

Lastly, a third study of about 4,000 subjects compared how rs20455 affected differences in outcomes for people given the standard-dose pravastatin therapy versus high-dose treatment with atorvastatin, another anti-cholesterol drug. Those with the AA genotype saw no additional benefit of receiving the more intense atorvastatin treatment compared to the standard pravastatin treatment. But people in the study with GG or AG did see a significant reduction in their risk of having a heart attack, cutting their risk by about 40% over two years.

23andMe customers may examine their genotype at rs20455 in the Genome Explorer (now called Browse Raw Data) and compare their genetic data to that of the studies in this article. Keep in mind, however, the points made in the introduction to this SNPwatch article.

In addition to the SNPwatch reminder, remember that genetics are only one factor that physicians may consider when prescribing preventive medications for heart disease or any condition. If you have concerns or questions about what you learn through 23andMe, you should contact your physician or other appropriate professional.

Update [2008/01/24]: Although our Gene Journal article on heart attacks (now called Health and Traits) does not currently cover this SNP, we believe that this association has been replicated sufficiently enough to include in the article–we’ll get to it as soon as we can!

ShareThis