Lung Cancer: A Model of Susceptibility, Biomarkers, and Molecular Signatures
Cancer is a leading killer because most tumors are found too late, when they are large and polyclonal. Many cancers are shrunk by current treatments, but they recur because resistant clones survive. The new "rational" therapies may be less toxic, but none has solved the resistance problem. Combinations of the new wonder drugs may work better, but even a doubling of the "cure" rate would impress only statisticians; the average American would not notice because durable cures of common, adult-onset cancers are so rare. Still, most researchers look for "magic bullets", because incentives and expectations favor frontal assaults in the "War on Cancer".
This pessimistic view of the usual strategies led me to work on early detection and prevention. Unlike current therapies, both of these tactics have nearly eradicated famous killers. For example, cervical cancer was the leading cause of cancer death for American women in 1940 (Fig. 1). But early detection by the Pap smear slashed the toll within a generation. Similarly, gastric cancer was number one for men and number two for women in 1940, but mortality plummeted when home refrigeration reduced nitrates in meats and other foods. These victories show us how to win the "War": We need molecular "Pap tests" for common solid tumors, and we need better tools to find environmental carcinogens.
Lung cancer is a good research model because it is common, the causes are known, and it has been studied for decades. I have developed three projects that address susceptibility, etiology, and early diagnosis. All are bottlenecks in multistage carcinogenesis, the process whereby normal cells evolve stepwise to cancer. For example, in Fig. 2, the lower pathway for susceptibles shows the traditional progression of morphologic changes from normal to abnormal through dysplasia to cancer. The upper pathway for resistant individuals shows limited progression because it represents the half of persistent smokers who live as long as their nonsmoking peers. These hardy souls seem immune to hundreds of toxins in tobacco smoke. All in all, these observations imply that smoking exposes two biological phenotypes: susceptible vs. resistant. If genetic testing could identify high-risk individuals, then early interventions - smoking cessation, screening, and chemoprevention - could be targeted to those at greatest risk.
Project I - Genetic Susceptibility
Although smoking causes most lung cancers, approximately 16,000 lifelong never-smokers died of the disease in 2000. Passive smoking (i.e. second-hand smoke) and residential radon are believed to be the leading causes of these cancers. In a study of nonsmoking women in Missouri, we found that homozygous deletion of glutathione transferase (GSTM1Null) doubled their risks from passive smoking and tripled their risks from radon. Our radon finding is unique, but a Japanese series reproduced our report on passive smoking, and two studies of indoor air pollution from stoves burning wood or coal provide indirect confirmation. Our findings suggest that susceptible individuals accrue unreasonable risks from passive smoking five- or six-fold, compared to the general population (Fig. 3). If our observations are durable, then it may be prudent to further limit encounters with these - and similar - pollutants.
Project II - Molecular Signatures Give Clues to Etiology
Finding the causes of cancer is hard. For example, it took more than a dozen years and millions of subjects to prove that smoking causes lung cancer. Thr problem is that it is hard to prove that an exposure in youth causes a cancer in old age. Accordingly, there was great interest in the discovery that carcinogens leave recognizable patterns of genetic damage in cancer tissues. This suggested that the origins of cancer might be found in DNA from tumor cells, and a distinctive pattern of p53 mutations was discovered in lung tumors from smokers. But this first-generation signature is insensitive, because fewer than 10 in 100 tumors have mutations with all the characteristics of exposure to incinerated tobacco. Ten percent is enough to establish proof of principle, but it is not enough to answer real questions, such as: "Did secondhand smoke cause my lung cancer?" or "Does smoking cause breast cancer?" These are challenges for the next generation.
We aim to complement the current p53 tobacco signature with a new one based on frequencies and patterns of abnormal methylation and allelic deletion. There is evidence that lung tumors from smokers have more methylation and allelic deletion than nonsmokers. So, the simplest approach is to classify tumors by their rates of allelic deletion and methylation. For example, charting methylation and deletion scores on an X-Y axis might produce clusters of smokers and nonsmokers (Fig. 4). If there is a dose-response, then nonsmokers exposed to secondhand smoke might cluster somewhere in between. Over time, a mature index would test only the most informative genes and hotspots.
If successful, this project might reach far beyond tobacco and lung cancer. For example, the causes of colon and breast cancers are largely unknown, but some say that smoking plays a role. Population studies have been unable to make the case (probably because of individual variations in susceptibility), but a robust tobacco signature might finally settle the question. Furthermore, if a "methylation-deletion index" works for tobacco, it could be tried for a long list of environmental toxins, including asbestos, arsenic, acrylamide, aflatoxin, flame retardants (i.e., polybrominated diphenyl ethers [PBDEs]), mercury, pesticides, polychlorinated biphenyls (PCBs), and radiation (i.e., ionizing and ultraviolet).
Project III - Molecular Biomarkers
Early diagnosis and treatment dramatically reduced mortality rates for cervical cancer. But how can we develop molecular "Pap tests" for solid tumors? Recent reports show that sensitive methylation assays can detect abnormal hypermethylation in sputum samples years before lung tumors are found. These findings suggest that quantitative assays could monitor mehtylation levels much as cholesterol concentrations are measured today. Similarly, just as excessive cholesterol levels are treated with statins, so could rising methylation levels be treated with existing demethylation drugs. Much more work is needed, but it is possible that hotspots of methylation and allelic deletion - such as those mapped in Project II - will be useful biomarkers for early detection.