Editor's Note: This story was published in the August issue of Scientific American.

A few heaping piles of scrap metal and a rusty coal shed are all that is left of the power plant that until recently squatted like an immense, smoke-belching dragon in the middle of Tongliang, a gray city of 100,000 in south-central China. As we walk toward the shed, a Belgian Shepherd begins barking furiously, jerking its iron chain and baring sharp teeth. A brown-eyed face peeks out from the open doorway—it belongs to a girl in a stained shirt, holding a tabby cat that jumps away to hide under a slab of concrete as we approach. The girl is no more than six or seven years old and appears to be living in the shed with her father, who watches us warily from within.

The delegation of local officials who are taking us on a tour of the site are embarrassed; they want to hustle us along to a nearby office to show us an elaborate scale model of an extravagant (by Tongliang standards) 900-unit housing development planned for the property. But Frederica Perera is intrigued. She strides toward the girl and gives a friendly “ni hao” and a smile. The girl smiles back before retreating back into the shadows with her father.

Children, after all, are why Perera is here. She is looking for connections between air pollution and disease, especially in children who were exposed to pollutants in the womb. The director of Columbia University’s Center for Children’s Environmental Health, Perera helped to pioneer the field of molecular epidemiology, which applies the tools of molecular analysis to identify genetic and environmental factors that contribute to disease. She and other molecular epidemiologists who focus on environmental links to illness increasingly do much of their work in the developing world, where pollution is so ubiquitous that its complex connections to health can be calibrated even in small study populations. But their conclusions should also apply in places such as the U.S., Europe and Japan, where environmental exposures are subtler and their effects more difficult to measure in small-scale studies.

Wherever they work, what distinguishes the approach of molecular epidemiologists is their search for biological indicators that closely correlate with toxic exposures and illness. Often these markers take the form of chemicals bound to DNA or of changes in gene structure or activity that match up with particular types of contaminants and disease. Now that DNA microarrays and other screening technologies are making it much easier to measure many of those biomarkers, routine use of such tools could, at least theoretically, save lives by identifying populations at risk from specific pollutants.

The science is still controversial, however, because relatively few candidate molecular biomarkers of susceptibility, exposure or early disease have been fully validated—that is, proved to herald future illness—and because it is very difficult to factor out confounding variables such as diet and genetic predisposition that may be at least as important as exposure to pollutants in causing various ailments. What has proved even more difficult is getting a handle on how those disparate risks may be interacting to affect health.

As a result, more than 25 years after Perera’s first paper on the topic, molecular epidemiology has progressed more slowly than its architects had initially hoped, and the consensus among researchers is that no matter how enticing it sounds in theory, real-world complexities have limited its usefulness. Initial enthusiasm for the idea that changes in a few specific biomarkers—the tumor-suppressing protein p53 is a prominent example—could be reliable indicators of early illness has faded as researchers have identified much more complicated etiologies, involving cascades of biochemical changes, for many diseases. “You could say that the reception was a little overenthusiastic early on. There have been some premature claims,” Perera acknowledges. “The promise of using biomarkers for early detection and custom-managed treatment has turned out to be not so easy.”

But now, in Tongliang, Perera believes she has found the best test case yet for environmental epidemiology at the molecular level—and she and her Columbia colleague, Deliang Tang, are getting results to back up her optimism. How they made their discoveries is almost as interesting as what they found.

A “Biological Dosimeter”
Ever since she began studying biomarkers in 1979, the ones that have interested Perera the most are PAH-DNA adducts, which she and Tang are now measuring in the white blood cells of children who were exposed to exhausts from the Tongliang power plant. PAHs, or polycyclic aromatic hydrocarbons, are a large family of compounds formed by the incomplete combustion of organic material—especially coal, but also other fossil fuels, cigarettes and even barbecued meat. They are among the most widespread and harmful air pollutants in the world. What interests Perera most about PAHs is their sticky molecular structure. Many PAHs readily form tight, covalent bonds with DNA. Those fused pollutant-DNA complexes, or adducts, can disrupt replication of the genome during cell division, altering the functions of genes that promote or suppress disease.

The coal-burning power station that loomed over Tongliang was a Vesuvius of PAHs, and its closure in 2004 changed environmental conditions in the city virtually overnight. That makes Tongliang something very rare in the world of epidemiology: a feasible laboratory for measuring the before-and-after health impacts of air pollutants. The city is still far from pristine, but passing cars no longer kick up clouds of black soot from the street and families can hang their wash outside to dry for more than a few minutes without their white shirts turning gray. The air-pollution monitors that the Columbia team installed around Tongliang confirm the improvements: airborne concentrations of one of the most important PAHs, benzo(a)pyrene, or BaP, fell by about 30 percent between 2002 and 2005. Other PAHs declined even more.

The changes that interest Perera the most, however, are taking place in the bodies of the city’s youngest residents. Since 2002 she and Tang, along with Tin-yu Li of Chongqing Children’s Hospital, have been studying 450 children who live within two kilometers of the plant site by testing their DNA and measuring their physical and mental development starting at birth. The researchers’ preliminary analysis shows that children born in 2002, when the power plant was still operating, have smaller heads and score worse on developmental tests than those born in 2005, a year after the plant closed. There are also differences at the molecular level: concentrations of BaP-DNA adducts were about 40 percent higher in the white blood cells of newborn babies in Tongliang in 2002 than in those of children born three years later.

Perhaps most meaningful of all, in the children born in 2002, measured concentrations of BaP-DNA adducts closely correlated with head circumference and developmental test scores. In other words, the more damage a child’s DNA suffered in the womb, the more likely he or she was to be born with a smaller head and to score worse on tests of motor skills and overall development as a toddler. The correlations were weaker among babies born in 2005, suggesting that air pollution becomes less of a risk as overall levels drop. Children born in 2005 will probably be slightly less likely to get cancer, according to Perera, whose previous work suggests that adduct counts correlate with cancer risk.

The Tongliang data, along with the results of earlier studies she conducted in Poland and New York City, suggest that measurements of adducts in white blood cells are reliable “biological dosimeters” for estimating the impact of PAHs on neurodevelopment, Perera says. That being the case, testing for adducts could someday become part of a pediatrician’s arsenal for identifying children who are at high risk for developmental problems and thus need early intervention, she observes.

The 2002/2005 comparison is “a very powerful finding,” remarks molecular epidemiologist John D. Groopman of Johns Hopkins University. “One of the critical steps in the validation of biomarkers is to demonstrate that if you modulate the biomarker, you can show an effect on a health outcome.”

“I’m in awe of what Dr. Perera has accomplished,” adds John F. Rosen of Children’s Hospital at Montefiore in Bronx, N.Y., a longtime lead-poisoning researcher who has also worked in China. “Translating her clinical results unquestionably will advance child health in China and the rest of the world.”

Finding Tongliang
A fine mist was falling the morning I arrived in Tongliang with Perera and Tang. The hills surrounding the city on three sides seemed as blurred as they must have looked back in the days when the coal plant was running. Most of the electricity in this part of China comes from hydropower, but because snowmelt does not swell the Yangtze River until the spring, many of the smaller cities have relied on primitive coal-burning plants that lack even basic pollution controls to fill the gap during the winter months. Tongliang’s power plant consumed more than 4,000 tons of coal a month between November and May, and that coal was especially problematic because it contained very high concentrations of sulfur and thus did not burn completely. PAH-laden ash and exhaust gases would pour out of the plant’s 279-foot smokestack and settle over the low-lying city like a thick blanket.

Back in 2000, Perera had been searching for just such a place to build on her two decades of research into the role of PAH-DNA adducts as indicators of disease risk. Her initial studies—the first ever in human subjects—had measured the adducts in the lung tissue of adult cancer patients [see “Uncovering New Clues to Cancer Risk,” by Frederica P. Perera; Scientific American, May 1996], and she went on to measure them in mothers and children living in polluted neighborhoods of New York City and the industrial city of Krakw, Poland. What she found is that people exposed to air pollution had higher adduct levels in their blood and that those high levels in turn correlated with the presence of genetic mutations that were known to be risk factors for cancer and developmental problems in young children. Children in cities where the air was cleaner, Perera determined, had fewer adducts and were less likely to suffer from growth deficits.

Still, her studies were small, and she could not rule out the possibility that unmeasured lifestyle differences, instead of pollution, might explain the differences in health and in adduct levels that she found between mothers and children who lived in polluted cities and those who did not. Now Perera wanted to take the next step by finding one city in which emissions of PAHs were abruptly reduced, allowing a before-and-after comparison in a single, discrete population.

China, with its extreme environmental problems and an authoritarian government capable of shutting down a major polluter in one fell swoop, was the obvious place to look. And Tang was the best person to direct the study Perera envisioned. A physician and Shanghai native, Tang did his doctoral training in public health in Perera’s lab and was now her frequent research collaborator. For his study he would have to train a small army of Chinese doctors and nurses to collect placentas and cord blood from mothers enrolled in the study and to administer cognitive tests to the children as they grew. He would also need to parley with an array of government officials, from Beijing bureaucrats to provincial apparatchiks, to secure the cooperation of hospitals, arrange for lab space, import pollution-control equipment and export blood samples—all very politically sensitive tasks.

Perera and Tang investigated 12 candidate sites before settling on Tongliang, where its sole power plant was scheduled to close as part of a government program to replace inefficient coal-burning plants. Tongliang was a good fit not only because the plant was set to shut down but also because the city had relatively few other important sources of air pollution aside from vehicle traffic. Natural gas had already replaced wood- and coal-burning stoves in the city, which had no large factories. The four local hospitals that agreed to participate were just large enough to collectively generate enough cases to satisfy the statistical requirements of the study Perera and Tang planned: 150 nonsmoking women whose pregnancies coincided with the months the plant operated. Additional mothers and newborns would be tested in later years, after the power plant closed.

But when Perera and Tang arrived in Tongliang in the spring of 2002 to begin recruiting pregnant women for the study, they found themselves in the middle of a controversy. The question of whether to close the power plant had been a simmering issue for years in Tongliang. Some local mothers had even stood outside during government meetings in silent protest against continued operation—an extraordinary step in China. Now city officials, worried about the economic impact, were considering retrofitting the plant or moving it to the edge of the city instead of shuttering it. The Columbia researchers had to wait for months for a resolution. Ultimately, publicity over their planned study helped to tip the balance in favor of closure. The old smokestack spewed its final cloud of gray smoke in May 2004.

Persistent Complications
Three anxious young fathers, all of them smoking cigarettes, slump in the crowded central hallway of Tongliang County Maternal Hospital, not far from two “No Smoking” signs. Smoking is endemic among Chinese men—half of the male population and two thirds of adult men younger than 25 light up regularly—and the nurses and doctors seem to have given up trying to enforce a ban.

In a room at the end of the corridor, medical personnel are lavishing attention on a young boy named Junshan Li, who was born in the spring of 2002, when the power plant was still going strong. Li is one of the original 150 subjects of the Tongliang study, but there are no signs today that he is anything but a healthy, active five-year-old—active enough to have broken his collarbone a week earlier in a playground accident. Li is at the hospital today not because of the collarbone injury but for his yearly developmental assessment for Perera and Tang’s study. At the moment, he is perched on the edge of a chair and excitedly shouting out numbers in Chinese—“ba!” ... “san!” ... “qi!”—in response to simple arithmetic questions posed by a tester, a young pediatrician named Xu Tan. Her goal is to assess Li’s growth both mentally and physically; after the test, Tan will weigh him and measure his height and head circumference.

The scene in the hospital illustrates a key challenge of molecular epidemiology. The growth and developmental deficiencies Perera and Tang are scrutinizing for links to prenatal exposure to PAH pollution are subtle and can have multiple causes—including secondhand tobacco smoke such as that emanating from the fathers smoking at the other end of the hallway. The researchers have tried to deal with the problem of alternative causes by enrolling only nonsmoking mothers with low-risk pregnancies and by asking mothers about their education, family smoking habits, and exposure to PAHs via grilled meats and other foods, as well as some other possibly confounding explanations. The investigators have also taken measurements of neurotoxic metals and antioxidants in blood, because those could also affect child development. So they are reasonably confident that the correlations they have uncovered between adduct counts in Tongliang children and several important measures of growth and learning stem mostly from exposure to air pollution.

Though statistically significant, the differences between Tongliang children born in 2002, when the power plant was still open, and those born in 2005, after it had closed, are small: a few millimeters in head circumference and height, an ounce of body weight, a point or two on a developmental test. According to Perera, the results suggest that the 2002 children will be slightly more likely to be slower learners and to need extra help at school and will develop fine-motor skills later on average than their counterparts born in 2005.

These kinds of subtle health effects have always been a source of contention in biomarker research. In the 1970s and 1980s Herbert L. Needleman, now at the University of Pittsburgh, pioneered the study of lead levels across populations by grinding up baby teeth and measuring trace levels of the metal, eventually finding correlations with diminished learning capacity and delinquent behavior. But the indistinct nature of the neurological problems Needleman was seeing, and the very low levels of lead he was measuring, opened his work to fierce criticism from the lead industry. Ultimately, his findings were replicated and are now widely accepted as a key justification for removing lead from gasoline and paint. Moves to crack down on secondhand smoke, similarly, have cited studies that measured blood and urine levels of cotinine, the most important breakdown product of nicotine. Cotinine levels in pregnant mothers and newborns have become extensively used markers in research into links between environmental tobacco smoke and a broad range of developmental problems in children.

There is an additional complication in drawing conclusions about health effects by measuring molecular biomarkers. Just because concentrations of lead, cotinine or PAH-DNA adducts are relatively easy to measure in the bloodstream does not mean that the same dose is reaching the brain or other organs where meaningful damage can occur. Earlier adduct studies, some of them by Perera, have suggested that white blood cells are a reasonable proxy for target organs such as the brain and lungs, but contaminants are rarely distributed uniformly throughout the body, and metabolic differences can cause huge variations among individuals even if they are part of a cohesive population that breathes the same air or drinks the same water.

Adducts do have one important advantage over blood measurements of lead and cotinine: they are indicators not just that a contaminant is present in the body but that it is having a discernible effect, in this case by altering DNA molecules. Here, too, however, there are some big uncertainties because no consensus exists on how the formation of adducts could lead to developmental problems in children. In cancer research, where Perera did her initial work with adducts, the connection is somewhat clearer because an adduct’s ability to disrupt accurate genomic copying during cell division could trigger mutations and other genetic changes that give rise to malignant cells. But for developmental deficits in children, the theories are more nebulous.

One of the leading ideas about how PAHs may affect the brain involves the “neural pruning,” or controlled nerve cell death, that occurs naturally as the developing brain adjusts to its environment and becomes more efficient by discarding synapses it does not need. The presence of PAHs may alter brain function by extending this process, called apoptosis, to essential synapses as well—both in the womb and in the child’s early life. Another prominent theory is that PAHs disrupt a fetus’s ability to obtain nutrients and oxygen by occupying molecular receptor sites in the placenta. The pollutants may also trigger the release of metabolism-altering enzymes in fetuses and young children and may also alter levels of growth-regulating hormones. The likeliest answer, some experts suggest, is that most or perhaps even all these mechanisms are taking a toll on the brains of children exposed to high levels of air pollution.

“One of the things we’ve come to learn over the past 10 years is how many pathways there are in many common diseases,” says Groopman, whose studies in China and elsewhere have focused on the interaction of aflatoxins with human hepatitis B virus in inducing liver cancer. Aflatoxins, which are produced by Aspergillus fungi, are a ubiquitous food contaminant in China and Africa. “With aflatoxin and liver cancer, you’re dealing with just one compound and with a specific disease that is very common in countries like China. There are no other confounding sources,” he points out. PAHs and childhood development, he adds, “are a much more complicated story.”

Next Steps
The next steps, Perera and Groopman agree, will be to scale up and drill down. Larger studies with greater statistical power, involving perhaps thousands of children, may be able to find stronger correlations between PAH exposures and specific developmental problems. Meanwhile investigators will also be delving into the intricacies of neural chemistry in the search for new biomarkers that will allow them to better measure the precise changes in the brain that give rise to growth and learning problems.

That search is already taking scientists into the realm of epigenetics, which is the study of heritable, environmentally induced changes that alter the functions of genes without changing their underlying sequence of DNA codes, as PAH adducts do. According to Perera, some evidence already exists that PAHs can lead to epigenetic changes in gene activity that cannot be identified by looking for breaks or disruptions in the genetic code itself. Because adducts are not a proxy for those epigenetic changes, Perera and other molecular epidemiologists are starting to turn to other biomarkers, such as DNA methylation, which is the attachment of methyl groups (CH3) to genetic material. Methylation typically silences genes, and some evidence suggests that methylation increases with PAH exposure and prevents genes from giving rise to proteins involved in suppressing some diseases, including cancer.

If epigenetic changes turn out to be crucial in altering a child’s capacity to grow and learn, then counting methyl groups may end up being a better dosimeter for pollution’s impact on the brain than counting PAH-DNA adducts. The Tongliang study may help settle that question, because Perera and Tang plan to analyze all their Tongliang blood samples for DNA methylation and perhaps for other epigenetic biomarkers as well.

Someday doctors may be able to test for a battery of biomarkers—both genetic and epigenetic—to assess a child’s overall risk for a wide range of health and developmental problems. For now, though, children such as the girl in the abandoned coal shed in Tongliang will have to rely on their governments to protect them from PAHs and other harmful air pollutants, and Perera never misses a chance to prod public officials to do so. “Tongliang is important because it shows that governments can take specific actions to reduce exposures,” she asserts. “Instead of just guessing, we can measure the improvements.” The ability to measure cause and effect is especially important in China and other fast-growing countries that are making hard choices about whether to continue to rely heavily on coal or move to cleaner but more expensive energy sources.

The complexities and uncertainties of molecular epidemiology, Perera says, should not obscure the overall message of the data she has accumulated during 29 years spent studying the health effects of air pollution in Finland, Poland, China and her own neighborhood near Columbia in upper Manhattan. That message is as relevant in developed nations as it is in heavily polluted China, she explains, because PAHs are a ubiquitous pollutant capable of affecting children even at relatively low concentrations. In New York City, where airborne PAH levels are more than 10 times lower than in Tongliang, “we have been able to measure the effects in reduced fetal growth and neurodevelopmental impacts,” Perera notes. “This is not some vague, theoretical threat that may or may not turn out to be valid with more study. We have all the evidence we need to reduce these exposures right now.”

Even as Perera and Tang continue to analyze their Tongliang data, they are on the prowl for opportunities to launch a much larger study based on the same before-and-after premise as the work in Tongliang. “It would require the magic of finding just the right kind of place, in just the right setting,” she says. “We’ve got our eyes peeled for that.”

Note: This article was originally published with the title, "China's Children of Smoke".