Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
8 C o l u m b i a U n i v e r s i t y RECORD June 10, 2003 Columbia Researchers Find Hormone Estrogen Can be a Cancer-Causing Agent BY SUSAN CONOVA I t's disconcerting to think that a natural hormone circulating in significant amounts through the bodies of half the world's population is a carcinogen, but it's now official. In December, the National Institute of Environmental Health Sciences (NIEHS) added estrogen to its list of known cancer-causing agents. For years, estrogen has been a suspected carcinogen, since strong epidemiological evidence associates the hormone to breast, endometrial, and uterine cancers. Women who begin menstruating early, or who start menopause late, produce more estrogen over their lifetimes and have a higher risk of breast cancer. Recently, the clinical trial of estrogen plus progestin treatment therapy was terminated because of an increased risk of breast cancer. Despite estrogen's new status on the NIEHS listing, not all forms of the hormone are carcinogenic. And that has called into question the route cancer-causing estrogens take in stimulating tumors. A new study by Mailman School of Public Health researchers shows that more than one sequence of steps is necessary before estrogen can cause cancer. In addition to a hormone receptor-mediated process, a second process is also required, says the study's lead author, Hari Bhat, assistant professor of environmental health sciences. Their results suggest that blocking the second pathway could prevent estrogeninduced cancers. But they also suggest that even non-carcinogenic estrogens can cause cancer given the right conditions. This research was published in the April 1 Proceedings of the National Academy of Sciences. Estrogen was originally believed to cause cancer by helping cells proliferate. After the hormone binds to its receptors in a cell, it turns on hormone-responsive genes that promote DNA synthesis and cell proliferation. If a cell happens to have cancer-causing mutations, those cells will also proliferate and have a chance to grow into tumors. "But if cell proliferation via receptor-mediated processes is the only mechanism, then all estrogens should cause cancer," Bhat says. "So it is hypothesized that estrogen metabolism may play a key role in estrogen-induced cancers because different estrogens differ in how they're broken down in the cell." The picture above shows renal tumors, labeled “T”, and congested tubules, noted by arrows, of male Syrian hamsters induced by treatment with the natural female sex hormone 17 beta-estradiol. Hari Bhat, assistant professor of environmental health sciences, above, found estrogen only causes cancer when cells also experience oxidative stress. The cell uses a series of reactions to rid itself of estrogen. In metabolizing carcinogenic estrogens, the reactions produce intermediates capable of producing oxygen radicals that can damage the cell's fats, proteins, and DNA. Unrepaired DNA damage can turn into a mutation, which can later promote cancer. To see if cancer-causing estrogens need oxygen radicals to produce tumors, Bhat implanted pellets of the hormone in hamsters that are susceptible to estrogen-induced kidney cancer. This model is widely used as an animal model of hormonal cancer. As expected, when the carcinogenic 17beta-estradiol (E2) was used, nearly all hamsters with the pellets developed cancer within seven months. E2 promotes cell proliferation and produces oxygen radicals when metabolized by the cell. Also, as expected, none of the hamsters developed kidney cancer when a non-carcinogenic estrogen, 17alphaethinylestradiol (EE) was implanted. EE acts through estrogen receptors to create new cells like E2, but unlike E2, is poorly broken down and does not produce oxygen radicals. But when EE was combined with a non-estrogen molecule that generates oxygen radicals, 30 percent of the hamsters developed kidney cancer within seven months. The nonestrogen used, menadione, did not produce tumors when used alone. "That we found tumors in the EE plus menadione treated hamsters clearly suggests that estrogen receptor activity and oxidative stress are both needed for estrogen to produce cancer," Bhat says. In other experiments, Bhat and his colleagues confirmed that the oxidative damage suffered by the cancer- ous kidney cells was caused by the metabolic breakdown of E2. "That's why E2 acts as a complete carcinogen," Bhat says. "It's a potent estrogen and it can also produce oxidative stress." The more complete knowledge of how the estrogen increases the risk of cancer could lead to new anti-oxidant therapies to treat or prevent cancer. But it also suggests that reputedly "safe" estrogens that are touted as replacements for the estrogens in hormone replacement therapy may not be so safe after all. "If we have oxidative stress in cells from other chemicals, then women are at risk for cancer even with estrogens that are considered non-carcinogenic," Bhat says. "The therapy may be safer if taken with antioxidants, but more research is needed to make safe and more effective antioxidants." Computer Scientist Julia Hirschberg Explores Frontiers of Computational Linguistics BY JOSEPH KENNEDY More than three decades ago, the film 2001: A Space Odyssey introduced audiences to a chillingly malevolent supercomputer named HAL. In what was then the far-off future, the film imagined machines that could speak and understand human speech (and even read lips!). HAL became an instant symbol for the dangers of runaway technology. While artificial intelligence researchers have managed thus far to avoid creating monsters like HAL, the idea of humans and computers speaking to each other is no longer the stuff of science fiction. It is instead the driving force behind the grow- ing discipline of computational linguistics, which studies the computational aspects of human language. “Basic speech recognition systems have now become commonplace,” says Julia Hirschberg, who joined the Department of Computer Science in Fall 2002. “Researchers today are moving into some very interesting and complex areas. We’re looking at how to enable computers to recognize speech errors, perform audio browsing and retrieval of email, and recognize and produce emotional speech.” Hirschberg received a Ph.D. in computer science from the University of Pennsylvania in 1985. That same year, she began working in the linguistics research department at Julia Hirschberg AT&T Bell Laboratories. In 1996 she moved to AT&T Labs – Research, specializing in human-computer interface research. Hirschberg is part of a Columbia team researching speech summarization, an area of natural language pro- cessing that enables systems to summarize spoken words. The Columbia Natural Language Processing Group, headed by Kathy McKeown, obtained an NSF and a DARPA grant for research in this area. “We want to learn what speech cues will lead a system to understand what is important to summarize,” Hirschberg says. “Although a lot of research has been done on text summarization, not too many people are trying to produce spoken summaries. It could lead to applications that summarize things like news broadcasts or voice mail messages in spoken form.” Another focus of Hirschberg’s research at Columbia is emotional speech. Computer scientists are studying this process from the perspectives of speech recognition as well as speech generation. “There is an effort underway to enable computers to recognize emotions in human speech,” she said. “Work is also being done to help computers generate emotional speech. Both of these would have enormous potential for commercial applications.” Applications might include a tutorial system that could recognize confidence, or the lack thereof, in a student, and customize its lessons accordingly. On the generation side, text-to-speech systems that speak emotionally could have a number of consumer uses, from books for the blind to to computer games.