# News this Week

Science  05 Apr 2002:
Vol. 296, Issue 5565, pp. 24
1. NIH DIRECTOR-DESIGNATE

# Money, Mission, Management Top Zerhouni's Agenda

1. Jocelyn Kaiser

Elias Zerhouni has spent 27 years at Johns Hopkins University in Baltimore building a reputation as a creative thinker and adept manager. Those skills will be sorely tested as the Algerian-born radiologist, nominated last week by President George W. Bush to be the next director of the National Institutes of Health (NIH), prepares to lead an agency nearing the end of an extraordinary 5-year run, during which its budget will have doubled to a projected $27.3 billion in 2003. “The thing staring the new director in the eye is explaining what has been done with [the budget] and what's to come,” says Harold Varmus, who stepped down as NIH director in December 1999 to lead the Memorial Sloan-Kettering Cancer Center in New York City. The first step for 51-year-old Zerhouni is a confirmation hearing before the U.S. Senate, which is expected to press him on his views about research involving human embryonic stem cells. But because Bush has already laid down the Administration's policy on that issue, the biomedical community is probably more anxious to hear what he thinks about the federal crown jewel with which he is being entrusted. “As someone who's less widely known, he's going to have to really establish what his values are as the leader of NIH,” says Harvard provost Steven Hyman, who was director of the National Institute of Mental Health until last December. (Zerhouni has declined to give interviews until after he is confirmed.) One early test for him will be shaping the newest of NIH's 27 institutes and centers, the National Institute for Biomedical Imaging and Bioengineering (NIBIB). Championed by radiology and bioengineering groups—including the Academy of Radiology Research, on whose executive committee Zerhouni sits—the institute was created by Congress in December 2000 over the objections of Varmus and others, who felt that creating more entities undermined good management practices. Congress allocated$112 million to NIBIB this year, including $67 million in existing extramural grants from other institutes to develop new imaging and bioengineering technologies that cut across organ systems and diseases. But in a bid for a larger pot, lobbyists persuaded Congress to order NIH to set up an advisory group to examine the entire NIH portfolio. Preliminary criteria from the working group have sparked heated debate, however, with some NIH staff and outside researchers saying that many of the grants flagged by the panel lie outside NIBIB's territory. Zerhouni also inherits a proposal for an imaging center at the University of Mississippi Medical Center in Jackson. Research director David Dzielak says that the center would enhance ongoing campus projects as well as the work of scientists at nearby NASA Stennis Space Center. The university has talked with top NIH officials about its plans, Dzielak acknowledges, but he says it is not trying to sidestep peer review. “We don't want to see the NIH [budget] earmarked even if it would benefit us,” Dzielak says. “We just think this is a good idea, and we're testing the waters to see how it could be done.” Although submitting a proposal may be standard practice, Mississippi legislators have upped the stakes by claiming that the peer-review system discriminates against their state and informing NIH that they are keenly interested in the project's fate. The imaging institute isn't the only political hot potato in the NIH budget. Another is the first-ever request by a president for a specific total for cancer research. The amount encompasses more than the recommended funding for the National Cancer Institute. Special-interest groups have pushed Congress to adopt similar language for their particular cause, a step that some feel would limit NIH's ability to pursue the most worthy science. “NIH would have to spend no less than$5.1 billion” on cancer, says Richard Turman of the Association of American Universities. “Our biggest worry is that it would create me-too's” from other disease groups. Several lawmakers have also raised questions about the president's request.

That debate is part of a broader discussion of how NIH will cope with budget increases that are expected to be much smaller than the 14% or 15% received each year since 1999. In particular, lobbyists can't see how NIH can maintain its current annual pace of 9377 new grants and allow for growth in the size of existing grants with budget increases as small as 2.1%, the figure projected for 2004 in the president's budget. NIH acting director Ruth Kirschstein told legislators last month that she is working on various “models.” One key variable is the $770 million in the 2003 request for bioterrorism funding to build facilities and buy a new anthrax vaccine, an NIH official says. Although all the money would go to those two activities in 2003, officials hope that an equivalent amount might be put into the 2004 NIH budget, allowing them to use some of it for regular grants. However, if Zerhouni pushes for increases that outpace inflation—7% to 10% is most often mentioned—he will bump up against intense lobbying by nonbiomedical groups that argue that the physical sciences have been shortchanged during NIH's recent budget spree. “He's going to have to figure out the right balance,” says Varmus. Zerhouni will also have to map out a plan for spending the$1.5 billion for bioterrorism in this year's budget, nearly all of it to the National Institute of Allergy and Infectious Diseases (NIAID). “Many of us were surprised by the large size of the bioterrorism allocation,” says Varmus. “[Zerhouni] will need to work out with [NIAID director Anthony] Fauci where it's going to be invested.”

Another looming task is finding top-quality institute directors. Varmus recruited several talented people, such as Hyman and Gerald Fischbach, head of the National Institute of Neurological Disorders and Stroke, who have since moved on. Six directorships—of the imaging, mental health, neurology, mental health, general medical sciences, and alcohol abuse institutes—are or will soon be vacant. “The recruitment process is hard,” Varmus says.

In the larger picture, the new NIH director will be expected to explain the importance of bench biology to a public that hungers for cures. The race to sequence the draft human genome “made it a little easier,” says MIT molecular biologist Phil Sharp, a member of the National Cancer Institute advisory board. But explaining how those discoveries will help vanquish disease is a huge challenge, he adds.

Edward Benz, president of the Dana-Farber Cancer Institute in Boston and former chair of medicine at Hopkins, thinks that Zerhouni is up to the task. “Varmus is a hard act to follow,” he says. “But I think [Zerhouni] has all the tools to do it. He has the intellectual capability, he's an outstanding manager and consensus builder, he's extremely fair, and he has a lot of integrity.”

2. SMALLPOX VACCINES

# New Cache Eases Shortage Worries

1. Martin Enserink

Since 11 September and the spate of anthrax attacks, the U.S. government has been scrambling to prepare for an even worse scenario: a bioterrorist attack with the variola virus, the cause of smallpox. Now, the nation can breathe a little easier. A new study has found that the 15.4 million doses of aging smallpox vaccine currently in the U.S. stockpile—an amount considered woefully inadequate—can be safely diluted by a factor of 5 or 10 without losing their potency. And, in another unexpected windfall, Aventis Pasteur, the vaccines business of Aventis Pharma with U.S. headquarters in Swiftwater, Pennsylvania, has just announced that it plans to donate to the government about 85 million additional doses of a similar vaccine that have been stored in its freezers for about 40 years.

Although the Aventis vaccine still needs to be tested for safety and efficacy, “this ratchets back our anxiety meter quite a bit,” says Peter Jahrling, a smallpox researcher at the U.S. Army Medical Research Institute of Infectious Diseases in Fort Detrick, Maryland.

The Aventis vaccine may offer the U.S. government an extra “insurance policy” until new vaccines are ready for use, Tommy Thompson, secretary of the Department of Health and Human Services, said last week. Last November, the U.S. government placed a rush order for 155 million doses of a brand-new smallpox vaccine with a British-U.S. company called Acambis—in addition to the 54 million doses it had ordered from Acambis a year before. Clinical trials with the Acambis vaccine started last month; the entire batch is to be delivered before the end of 2002.

The dilution studies were launched last fall. The National Institute of Allergy and Infectious Diseases (NIAID) funded Sharon Frey and her colleagues at St. Louis University to test whether Dryvax, the vaccine currently in stock, could be stretched to protect more people if needed. Wyeth produced the vaccine, an attenuated virus called vaccinia, before routine smallpox vaccination was halted in 1972.

The study of 680 volunteers, released online last week by The New England Journal of Medicine, showed that more than 97% of the participants who received undiluted vaccine developed a “take”—a localized skin infection at the site of the shot, known to reliably indicate protection. So did 100% of those in the group given a vaccine diluted fivefold and 99% of those who got a 10-fold dilution. The slight variation in numbers, the researchers suspect, probably means that some of the volunteers were vaccinated earlier in life.

Few people knew of the cache of old vaccine in Aventis's freezers until The Washington Post reported it last week. But Aventis Pharma CEO Richard Markham denied that the company recently “discovered” the supply, as the Post reported. The company knew about the vaccine, which had been produced at the request of the U.S. Department of Defense, all along, he says, and had in recent years discussed with the government what to do with it. But the issue was never urgent—until last fall, when the company offered the vaccine to the government for free.

The company, which is hoping to be reimbursed for repackaging and other costs, has transferred the vaccine—stored in about 60 2-liter bottles at −20°C—into standard vials containing about 100 doses each. Like Dryvax, it consists of a vaccinia strain dubbed “New York City Board of Health.” Thompson says the government decided to keep the huge cache a secret while awaiting the results of preliminary tests of the vaccine's viability and potency. Those test tube experiments, completed recently, suggest that the vaccine is about as potent as Dryvax, the company says.

NIAID will test the Aventis vaccine in human volunteers over the next 6 to 8 weeks. Researchers will also test whether it can be diluted, like Dryvax. If so, the U.S. government could soon have more than half a billion vaccine doses at its disposal, more than enough for the country's 286 million residents.

With the acute shortage most likely solved, the government also plans to address the needs of millions of infants, the immunocompromised, and eczema patients—groups at high risk of serious side effects from the current vaccines. Several companies, including Aventis, are already working on the next-generation, much more attenuated, vaccines.

3. IMPACT HAZARD

# Celestial Billiards Threaten Hit in 2880

1. Richard A. Kerr

The first master of gravity, Isaac Newton, saw the solar system as a clockwork machine whose workings depend solely on the mass of the machine's parts and thus their gravitational pull. Would that it were so. On page 132 of this issue, a team of researchers reports that the 1-kilometer asteroid 1950 DA has up to a 1-in-300 chance of clobbering Earth with a 10,000-megaton punch on 16 March 2880. For an event 878 years hence, that may seem like a relatively precise prediction. But if basic Newtonian mechanics were the only consideration, there would be little room for doubt. A pivotal uncertainty, it seems, is how the asteroid spins.

As its temporary name implies, astronomers discovered 1950 DA 52 years ago, but they lost track of it when it faded from view before anyone determined its precise orbit. Unknowingly photographed in 1981, it was finally snared for good on New Year's Eve 2000 in an automatic telescopic search for threatening asteroids. Celestial mechanicist Jon D. Giorgini of the Jet Propulsion Laboratory in Pasadena, California, and 13 colleagues then calculated—based on telescopic and subsequent radar observations—that it had up to a 1-in-300 chance of gravitationally caroming off 15 close encounters with Earth and Mars to hit Earth in 2880. That probability is 1000 times greater than the well-determined impact probability of any other sizable known object. It's also 1.5 times the combined impact probability of all other asteroids.

But sources of potentially significant uncertainty remained, notwithstanding orbital observations spanning 51 years, high-precision radar data, and an orbit-stabilizing gravitational interaction between Earth and the asteroid. The tidal pull of the galaxy varies; the sun's gravity weakens as the sun blows off mass; its slightly squashed shape creates an irregular gravity field; estimates of planets' masses are uncertain; and sunlight exerts a gentle push. Combined, says Giorgini, these uncertainties could make 1950 DA cross Earth's orbit as much as a few days before or after the 20-minute window when a collision is possible in 2880.

But the killer uncertainty is the so-called Yarkovsky effect (Science, 13 August 1999, p. 1002). A century ago, the Russian engineer I. O. Yarkovsky recognized that the “afternoon” quadrant of an asteroid—the side most thoroughly heated by the sun—could act like a little rocket as it rotated into dusk and early evening, emitting thermal radiation that could push the asteroid into a different orbit. But astronomers know so little about 1950 DA—especially the direction of its spin axis—that celestial mechanicists can't tell what direction its Yarkovsky “rocket” points. The probability of impact could be higher than calculated—or it could be zero. Gauging the future path of an asteroid “is a lot like shooting billiards,” says Giorgini. Given 1950 DA's 15 close planetary encounters before 2880, “we're trying to line up a 15-bank shot. We can do the first 12 really well. For the last three, we need to know more about the cue ball” and how much the Yarkovsky effect will influence its course.

No one in the planetary science community is panicking. “The work is done about as well as it can be,” says planetary dynamicist William Bottke of the Southwest Research Institute (SRI) in Boulder, Colorado, but “it's a long time away, and the probability is small.”

Small, but the biggest around, notes Bottke's SRI colleague, asteroid specialist Clark Chapman: “It looks to me [like] there's a good chance that for the next 30 years this will be the one to watch.” If the threat continues to grow, one solution could be as low-tech as dusting 1950 DA with soot or powdered chalk to power up or throttle back the Yarkovsky effect (as well as sunlight pressure) and steer the asteroid away from Earth. Or just plastic-wrap it in reflective Mylar by sending a sunlight-driven solar sail on a collision course with the asteroid. Such tinkering with his clockworks would stun Newton.

4. BIOETHICS

# U.S. Questions Harvard Research in China

1. Andrew Lawler

BOSTON—Government investigators last week took Harvard University and two other U.S. institutions to task for their handling of research studies involving rural Chinese subjects. University officials insist they have already tightened their procedures, but the U.S. Department of Health and Human Services has asked for more information on whether researchers failed to obtain informed consent in advance, backdated documents, and misled investigators about the number of people involved in the studies.

The department's Office for Human Research Protections (OHRP) began its inquiry in 1999 after a Harvard School of Public Health researcher filed a complaint alleging that two occupational epidemiologists at the school, Xiping Xu and David Christiani, had taken advantage of Chinese subjects in rural Anhui Province, where they conducted a variety of genetic and environmental studies. The department's first public comment about its continuing probe appeared in the form of letters to Harvard, Brigham and Women's Hospital, and Massachusetts Mental Health Center, dated 28 March, that outline the government's concerns and ask for more information.

The letter to Harvard does not draw any conclusions but questions whether some subjects were enrolled in investigations before they signed informed consent documents. It also notes that “the handwriting for the dates next to the subject's signatures appear to be identical,” indicating that either the subjects—even those who could write—did not do the dating or that the documents may have been backdated. The letter also cites a large discrepancy between a report to OHRP and a journal article on the number of women enrolled in one particular study. Harvard has until 10 May to respond.

OHRP notes that the institutions have made strides in correcting the apparent problems, and Xu says the investigation has “improved our protection of human subjects in China.” A Harvard statement points to the “complex and difficult … ethical and cross-cultural issues” in international research, adding that the university has beefed up its monitoring staff. It has also formally reprimanded Xu and Christiani and placed their work under greater scrutiny.

But others say the problems with the research involve more than lax record-keeping. “This is a blatant and massive institutional violation of the human rights of subjects,” says Vera Hassner Sharav, director of the New York City-based Alliance of Human Research Protection, a private activist group that has followed the issue closely.

Harvard officials insist that OHRP has accepted its action plan and that the matter is largely laid to rest. But OHRP officials say the case remains open. Former Harvard researcher Gwendolyn Zahner, who brought the original complaint, says she is happy that OHRP has examined her allegations but chides it for not looking “beyond [the] paperwork.” Investigators need to visit China, she says, to find out what really happened.

5. PROTEIN STRUCTURE

# Harmless Proteins Twist Into Troublemakers

1. Jennifer Couzin

A select group of proteins are known contortionists; they bend and fold and stick together, contributing to neurodegenerative diseases such as Alzheimer's. Now researchers report the startling news that all sorts of proteins can perform the same devilish tricks in the lab. Furthermore, initial tests reveal that at least two of these normally benign proteins are toxic to cells when misfolded. The findings, discussed at a colloquium* in Washington, D.C., in late March and reported this week in Nature, hint that a certain type of misfolding may be common to all proteins, prompting theories that cells have evolved techniques to guard against such behavior. Scientists caution, however, that the work has been limited to test tubes and that the proteins were exposed to extreme conditions.

About 20 proteins share the ability to misfold and clump together to form distinctive “amyloid fibrils” found in Alzheimer's, Creutzfeldt-Jakob disease, and a variety of lesser-known disorders. The most famous of these proteins is β amyloid, which has been implicated in Alzheimer's.

Chris Dobson, a chemist and structural biologist at the University of Cambridge, U.K., suspected that a much broader range of proteins could form amyloid fibrils in test tubes. But the thesis sounded so implausible that Fabrizio Chiti, then a graduate student at Oxford and now at Cambridge, recalls trying to flee the project. Chiti quickly switched from skeptic to believer as protein after protein proved capable of forming fibrils when heated or immersed in a solution containing acid or a form of alcohol. Thus far, Dobson and a team at the University of Florence in Italy, led by biochemist Massimo Stefani, have experimented with roughly a dozen proteins with a variety of structures and functions. “We haven't yet got a protein we haven't been able to convert into fibrils,” says Dobson.

The proteins the team tested are easily found in humans, plants, or yeast. Even the common oxygen transport protein myoglobin underwent a complete structural overhaul and adopted a fibrillar form. The amyloidlike fibrils the team induced had parallel structures and similar binding properties to those malformed proteins found in diseased brains. The research suggests that even in healthy organisms, “forming misfolded aggregates is an essential part of what proteins do,” says Martin Karplus, a chemist at Harvard University and Louis Pasteur University in Strasbourg, France.

The lab-induced transformations may be as deadly as the misfolding of disease-related proteins. The researchers tested the effects of two of their manipulated proteins on mouse and rat cells. Products from an early stage of fibril formation were far more toxic to cells than the final, fibrillar form—just as some researchers suspect is the case for early stages of protein misfolding in Alzheimer's disease.

What this all means is yet to be determined. In people, as opposed to the test tube, “this doesn't happen to every protein,” says Jeffery Kelly, a chemist at the Scripps Research Institute in La Jolla, California. No one knows what qualities, if any, are unique to the 20 proteins known to form amyloid fibrils in humans, or whether most proteins have evolved properties to prevent certain kinds of misfolding.

A small but growing cohort of scientists suspects that if this style of misfolding is a generic property of proteins, it's likely to play a still-hidden but useful role in normal biology. “My view is that there are some cases where these kinds of transitions are beneficial,” says Susan Lindquist, a molecular biologist and director of the Whitehead Institute for Biomedical Research in Cambridge, Massachusetts. Her work suggests that this applies to yeast, sometimes moderating gene expression in helpful ways; other researchers are examining whether this is also true in other organisms, including humans.

• *“Self-Perpetuating Structural States in Biology, Disease and Genetics,” sponsored by the National Academy of Sciences, 22–24 March.

6. ANIMAL BEHAVIOR

# Last Year's Food Guides This Year's Brood

1. Jay Withgott*
1. Jay Withgott writes from San Francisco.

Timing is everything for some breeding birds. They must hatch their young in time to exploit a brief springtime abundance of food. Seasonal cues such as day length help birds calibrate their breeding. But now a study on page 136 shows that some birds adjust their efforts according to lessons they learned the previous year. The finding implies that such birds might be able to accommodate some environmental changes spurred by global warming, but scientists caution that such adjustments may be limited.

As oak trees leaf out in European woodlands each spring, caterpillars hatch and devour fresh young foliage before the trees pump too many noxious tannins into the leaves. The 2-week burst of caterpillars provides blue tits, small birds akin to chickadees, the food they need to satisfy a nest full of clamoring little mouths.

Birds likely use a host of cues to sense that spring is in the air, such as temperature, young leaves, or hatching caterpillars. But some researchers have suggested that birds breeding too late or too early one year might learn from their mistake and adjust their timing the next. Indeed, a few studies have suggested that past experience can guide other reproductive decisions. For example, collared flycatchers adjust their clutch size, and great tits decide whether to stick with the same mate, based on past breeding success. Now, Fabrizio Grieco, Arie J. van Noordwijk, and Marcel E. Visser of the Center for Terrestrial Ecology in Heteren, the Netherlands, provide the first experimental evidence that experience can influence reproductive timing.

The group monitored pairs of blue tits as they bred in nest boxes for two consecutive years. In the first year, the team supplied half the pairs with caterpillars and mealworms as they tended their broods. The researchers took advantage of the birds' tendency to nest later than the natural caterpillar peak during their first year of breeding, due to inexperience, unfamiliarity with their territories, and the challenges of finding a mate. Thus, most unfed control birds bred late in the first year, then advanced their breeding to match the caterpillar peak in the second. But the pairs given food did not move up their breeding time in the second year. In fact, they delayed it, apparently because the first year's supplemental feeding led them to expect that food abundance would peak later.

The results show that timing reproduction “is more complex than we previously thought,” says ornithologist Ruedi Nager of the University of Glasgow, U.K. Although past experience may be only one among many cues, others say, it is probably important for a species that returns to the same territory year after year and would benefit from learning the idiosyncrasies of its real estate.

But in another way, the blue tit seems to be an unlikely candidate for long-term learning: The species is among the shortest lived of all birds. More than half the population dies each year, making blue tits “just about the closest thing you get to an annual bird,” quips University of Oxford behavioral ecologist Ben Sheldon. Long-lived species are thought to be more likely to evolve the capacity for learning, says Oxford ornithologist Christopher Perrins. Thus, if the blue tit shows this ability, Sheldon and Perrins reason, then long-lived species might well possess it in spades.

The birds' ability to adjust their reproductive timing implies a certain degree of resistance to the ill effects of climate change, note Sheldon, Nager, and others. Recent data have shown that global warming can lead to mismatches between phenomena that once were synchronous, such as leaf emergence, caterpillar hatching, and bird breeding. Although the study suggests that birds can deal with some environmental variation, ecophysiologist Donald Thomas of the University of Sherbrooke, Quebec, points out that it merely “provides the mechanism for fine-tuning.” But “major climate change over decades,” he says, “probably will overcome the birds' ability to learn.”

7. PATENT LAW

# Judge Casts Doubt on Scientist's Account

1. David Malakoff

Materials scientist Shuji Nakamura is getting an unpleasant lesson in the take-no-prisoners style of the U.S. legal system. A federal judge last month accused the prominent University of California, Santa Barbara, researcher of lying in a high-stakes patent lawsuit and recommended that he be prosecuted for perjury, a charge that he and his lawyer strongly contest. The accusation comes just weeks before Nakamura is to receive one of his field's highest honors.

The Japanese-born Nakamura won acclaim in the 1990s as the inventor of a blue light-emitting diode (LED) that could lead to cheaper, more efficient lighting. But in 1999 he left Nichia Corp. and moved to the United States, saying that Japanese companies don't do enough to reward their inventors. Last year, Nakamura sued Nichia in a Japanese court, seeking a $16 million share of the firm's profits from his discoveries (Science, 31 August 2001, p. 1575). He now works as a consultant to a U.S.-based LED maker, Cree Inc. in Durham, North Carolina, which is embroiled in a patent fight with Nichia over the multibillion-dollar LED market. As part of Cree's legal struggle, Nakamura last November answered questions from attorneys about the history of the firm's patents. At least one of his answers caught the attention of the judge hearing the case, James Fox of the U.S. District Court for the Eastern District of North Carolina. In a 15 March letter to federal prosecutors, Fox said that Nakamura admitted that he had “intentionally submitted false data in conjunction with the applications for [Nichia's U.S.] patents.” Nakamura's words are still sealed in court documents. But Fox said in his letter that Nakamura either broke laws against submitting false information to the U.S. Patent and Trademark Office or lied about the accuracy of Nichia's patent filings as part of Cree's bid to invalidate them. Either way, Fox recommended that the government prosecute Nakamura for perjury—a crime that can be punished by a jail term. Prosecutors have several years to decide whether they will follow the judge's recommendation. Nakamura was stunned by the letter, which was publicized last week by the electronic newsletter Internet Patent News Service. “Perjury? I don't understand,” he told Science after being informed of the judge's complaint. Nakamura's attorney, William McLean of Thoits, Love, Hershberger, and McLean in Palo Alto, California, says Nakamura is not responsible for any allegedly false statements in Nichia's patent applications. “The judge has just a glimmer of all the pertinent information,” McLean says. Attorneys for Nichia and Cree declined to comment on the letter, and prosecutors and the judge did not return calls. But lawyers familiar with the case say the letter may reflect the judge's unhappiness about courtroom behavior by both sides. “The judge is signaling that he'll be tough on anyone who misbehaves,” says one, who asked to remain anonymous. In the meantime, Nakamura expects to be in Philadelphia on 25 April to receive the Benjamin Franklin Medal in Engineering. Some winners have gone on to receive a Nobel Prize. 8. MIDDLE EAST # Science Foundation Sets Priority Areas 1. Jeffrey Mervis* 1. With reporting by Adam Bostanci in Cambridge, U.K. A unique grantmaking foundation for Arab scientists, modeled on the U.S. National Science Foundation (NSF), hopes to award its first research grants next year. The private 2-year-old Arab Science and Technology Foundation, which held its first international conference last week in Abu Dhabi and Sharjah, United Arab Emirates, is expecting a contribution of almost$20 million from a private donor in the next few months, says Farouk El-Baz, director of the Center for Remote Sensing at Boston University and one of the conference leaders. The foundation currently has about $1 million in cash—a gift from Sharjah's ruler. El-Baz says the foundation needs at least$10 million before it can begin to make grants.

The 900 participants at last week's meeting combined research presentations with discussions of how to run a scientific enterprise based on open peer review rather than top-down directives. The foundation has already recruited four Arab-born, U.S.-based scientists to lead panels that will manage grants competitions in the fields of water and energy, biotechnology, new materials, and information technology.

El-Baz says the foundation will be breaking new ground simply by following NSF-style procedures: “The whole concept of submitting proposals, and then being held accountable for how the money is spent, is alien to most Arab scientists, who are used to getting a budget from the government and then just spending it.” The foundation would also welcome financial help from NSF, he adds.

That's unlikely to happen, says Osman Shinaishin, who oversees NSF's research programs in the Middle East. NSF “can't make that type of commitment” to a private organization, explains Shinaishin. But he says Arab scientists would do well to seek foreign collaborators, including those with NSF grants, to ensure that the research is high quality.

9. THE RICE GENOME

# Rice: Boiled Down to Bare Essentials

1. Dennis Normile,
2. Elizabeth Pennisi

The publication of draft genome sequences for the two major subspecies of rice is a milestone for agricultural research; it could also be critical for an international project hoping to produce a finished sequence

BEIJING, TOKYO, AND WASHINGTON, D.C.—When two groups simultaneously published rough drafts of the human genome sequence just over a year ago, the achievement was hailed as the “beginning of a new era of biology.” This issue of Science contains two research articles that herald a similar transformation for the agricultural sciences.

On page 79, Yang Huanming of the Beijing Genomics Institute (BGI) and colleagues describe a draft sequence of the indica subspecies of rice, the most widely cultivated subspecies in China and most of the rest of Asia. And on page 92, Stephen Goff of the Switzerland-based agrobiotechnology giant Syngenta and his team report a similar achievement for the japonica rice subspecies, which is favored in Japan and other countries with temperate climates. Both sequences are works in progress: They contain many gaps and errors. But they provide the first detailed look at the genetic blueprint of a crop that is a staple for more than half the world's population. These sequences, moreover, will produce key insights into the genetics of other major cereal crops, including maize, barley, and wheat, and they will help researchers interpret the sequence of the only other plant whose genetic code has been spelled out: Arabidopsis, a favorite organism of plant scientists. Unraveling the rice genome, says Michael Freeling, a plant geneticist at the University of California, Berkeley, “is a big deal.”

The publication of these two drafts is only part of the story. With so much at stake, many other groups around the world have been working to sequence rice. And, as with the sequencing of the human genome, these efforts have been marked by rivalries, fears of commercial control over basic genetic data, and a controversy over the conditions under which Science is publishing one of the draft sequences (see sidebar, p. 34). Unlike the bitter disputes that characterized the human genome sequencing, however, there has been a high level of cooperation between the public and private groups.

Paving the way for these developments has been the International Rice Genome Sequencing Project (IRGSP), an international consortium of publicly funded labs, somewhat similar to the Human Genome Project. Led by Japanese researchers, it has been plugging away at the sequence of japonica for nearly 5 years. The consortium had the field largely to itself until 2 years ago, when the U.S.-based agrobiotech company Monsanto announced that it had completed a rough draft of the japonica sequence (Science, 14 April 2000, p. 239). Now, the Syngenta and Beijing groups have stolen a march by taking a bold approach to sequencing, called whole-genome shotgun, that enabled them to complete rough drafts at lightning speed.

Their achievement has, however, sown some nervousness among members of the public consortium, who worry that their effort could be jeopardized if funders believe that the sequence is now essentially in hand (see Letters, p. 45). “It needs to be understood that this is not the end,” warns Rod Wing, a molecular biologist at Clemson University in South Carolina. He emphasizes that even though commercial benefits and research insights are already being wrung from the data, the drafts are just that: drafts, not finished sequences. The consortium itself hopes to complete a much higher quality draft by the end of the year and a finished sequence, with few gaps and errors, by 2005 if funding for the work continues. “For the benefit of the world, it's important that we get the [complete] genome [sequence] out to everyone as soon as possible,” says Wing.

## Sequencers' beginnings

It has been a long and difficult road to get to this point. Independently, researchers in China and Japan began probing the rice genome almost 20 years ago, and Cornell University geneticists Steven Tanksley and Susan McCouch began mapping the genome in the early 1980s. But at that time, aside from a few academic labs, there was little interest in the United States and Europe, where rice is not a major crop. That began to change in the early 1990s, when researchers realized that rice is “the Rosetta stone” of cereals, says Stanford University molecular biologist Chris Somerville.

Although there is a huge disparity in the size of their genomes—rice has 430 million bases, corn 3 billion, and wheat a whopping 16 billion—cereals tend to have the same genes in the same order. That synteny, as this matchup is called, helped spark interest by U.S. and U.K. researchers, who began to view the more tractable rice genome as a tool for unlocking genetic secrets in other cereals. For the same reason, companies began to study rice more closely. “Nobody makes money on rice seeds,” says Somerville. Companies are interested in rice because of the potential payoff in the sizable markets for maize, barley, sorghum, and wheat seeds, he adds.

Japan, fearing it might be scooped on something so basic to the national diet and culture, in 1991 reorganized an existing mapping project into the Rice Genome Research Program (RGP). It quickly emerged as the world leader in rice genome work, thanks to a steady flow of funds from the Japan Racing Association, which by law must donate some profits from horse racing to agricultural research. Takuji Sasaki, who took over as RGP director in 1994, proved to be a staunch supporter of an international sequencing effort.

China, meanwhile, was moving on a parallel course. In 1993, the National Center for Gene Research in Shanghai, part of the Chinese Academy of Sciences, received $3.8 million for a 5-year effort involving five labs to develop physical maps for indica and to prepare a library of clones for sequencing. Center director Hong Guofan said he hoped to move on eventually to actual sequencing. The idea of an international consortium to sequence rice emerged from informal discussions in early 1997. That September, with support from the Rockefeller Foundation, researchers got together at an international plant molecular biology meeting to map out a strategy. Even though more people eat indica, the group agreed to focus on the Nipponbare cultivar of the japonica subspecies, because RGP had already done much of the preparatory work. The first official IRGSP meeting was held just 5 months later in Tsukuba, Japan. “Everybody just seemed to be ready for this,” says Ben Burr, a plant geneticist at Brookhaven National Laboratory in Upton, New York. Representatives from five participating countries—Japan, the United States, the United Kingdom, South Korea, and China—hammered out guidelines and divided up rice's 12 chromosomes. Participants agreed to release all data to public databases and to sequence at least 1 megabase a year. In the United States, the Rockefeller Foundation funded preliminary work by several academic researchers, including Wing, who received a grant to develop a library of bacterial artificial chromosomes (BACs), each one carrying a small bit of rice DNA. Subsequently, Novartis, which later became part of Syngenta, awarded Wing$3 million to ready those BACs for sequencing. The BACs and their mapping information were all made public.

But getting money for sequencing was another matter. With a supportive government, RGP was in the strongest position, receiving $10 million in 1998 for the first year of a projected 10-year effort. But no government sequencing funds were forthcoming in the United States until 1999, when the U.S. Department of Agriculture (USDA), the National Science Foundation (NSF), and the Department of Energy finally came up with$12.5 million to fund two groups to work on chromosomes 10 and 3. “The thing that slowed us down the most was the late entry of the U.S. funding agencies,” says Burr. Other countries fared even worse. The U.K. and Canadian groups never won funding. And Thai researchers joined the consortium, contributed a small amount of sequence data, but then withdrew to concentrate their scarce funding on gene discovery.

Japan, however, picked up some of the slack, and France, Taiwan, South Korea, India, and Brazil stepped in to share the burden. Finally, by the beginning of 2000, IRGSP seemed on its way toward its goal of completing the japonica rice genome sequence by 2008, possibly earlier. But the international consortium soon had company.

## The tortoise and the hare

In April 2000, Monsanto announced that it had sequenced the japonica genome. Working with researchers at the University of Washington and the Institute for Systems Biology in Seattle, Monsanto had produced an incomplete, but very informative, version that the company promised to share with individual academic researchers and with IRGSP. The news shocked, then worried, but ultimately delighted the community, because the data promised to speed up the IRGSP effort.

“The Monsanto data has been very helpful and very valuable,” says Machi Dilworth, who oversees plant genomic programs at NSF. The French group sequencing chromosome 12 turned to Monsanto for 75 of the 109 clones it is now sequencing. And the Shanghai group set aside its indica project, picked up the Monsanto materials, and finished a draft of chromosome 4. The Monsanto material “is now proving to be of value in speeding up the IRGSP sequencing and the cost effectiveness of the overall project,” says Ed Kaleikau, a plant biologist at USDA's Cooperative State Research, Education, and Extension Service.

The Monsanto boost was not decisive, it turns out. In January 2001, Syngenta reported that it, too, had sequenced japonica. It had contracted Myriad Genetics in Salt Lake City, Utah, to work on rice and other cereals for $30 million. Steven Briggs, head of Syngenta's Torrey Mesa Research Institute in San Diego, California, which managed the effort, says that IRGSP's target completion date of 2008 was too far off. “We needed the data immediately,” he says. The Beijing group, meanwhile, was eager to move ahead quickly as well, although for different reasons. In May 2000, BGI director Yang announced plans to sequence the entire genome of indica. Yang says that with most of the other efforts focused on japonica, “there was a feeling that China should sequence its own rice.” The team, which was separate from the group in Shanghai tackling chromosome 4, promised to deliver a draft of indica within 2 years (see p. 36). Whereas Monsanto and the public consortium took the traditional, orderly route of mapping the japonica genome and sequencing it piece by piece, the Syngenta and Beijing groups turned to the whole-genome shotgun method. They chopped up the entire genome into fragments, sequenced each fragment, and put the data in order with the help of powerful computers. Their successes “demonstrate how powerful a tool [the whole-genome shotgun] is,” says Jeffrey Bennetzen, a plant geneticist at Purdue University in West Lafayette, Indiana. Syngenta did enough sequencing to cover the genome six times, yielding 99% coverage and accuracy. That still leaves thousands of gaps in the sequence coverage, however. The Beijing group achieved four times coverage, yielding a less complete sequence. From their analysis, the Syngenta team estimates that the rice genome contains 32,000 to 50,000 genes, depending on how the genes are picked out. The Beijing team has picked up about 46,022 to 55,615 genes. As Bennetzen points out in a Perspective on page 60, the disparity between the two estimates probably largely reflects differences in the way the groups identified genes, and he expects that both estimates will shrink as more analysis is done. As expected, more than 80% of the Arabidopsis genes have counterparts in the rice genome, which should help researchers identify the functions of those shared genes. From an analysis of coding and noncoding regions within the genes, the Beijing team concludes that each gene codes for just a single protein—unlike human genes, which often code for multiple proteins. Rice may get its diversity of proteins from an abundance of genes that acquired diverse functions after the whole genome was duplicated, an event that Goff and his colleagues estimate occurred about 45 million years ago. The Beijing group has made all its data public, depositing them in GenBank. Already, some 350 researchers have used the data, says Yang. Syngenta is making its sequence available through its own Web site and on a CD-ROM, but it is also discussing sharing its data with IRGSP, says Syngenta's Briggs. “Details are still being negotiated,” says RGP head Sasaki, “but it is likely that an agreement will be reached [that is] very similar to the agreement with Monsanto, allowing IRGSP members to use the Syngenta data to complete their work on the phase II draft.” That would mean all the Syngenta data would end up in GenBank through the consortium's work. In the meantime, company researchers are collaborating with interested researchers; about 65 labs in 11 countries have made use of the information, says Goff. ## What next As news of these two drafts began circulating in the community, IRGSP was forced to rethink its goal of painstakingly closing almost all the gaps in the sequence before publishing a complete sequence. Japan's program, in particular, came “under a lot of pressure from the government to accelerate [its work],” says Joachim Messing, a molecular geneticist at Rutgers University's Waksman Institute of Microbiology in Piscataway, New Jersey. So in December 2001, IRGSP shifted gears and announced that instead of finishing the genome sequence on its original schedule, it would produce its own draft by December 2002. “We started with a certain road map, and then we had to adjust,” says Robin Buell, who heads the rice sequencing effort at The Institute for Genomic Research in Rockville, Maryland. IRGSP members have already placed the data for over 230 megabases, or more than half of the genome, in public databases, and three chromosomes are nearly finished. “I am sure we will complete our [draft] rice genome sequence by the end of 2002,” says Sasaki. On average, each base will have been sequenced 10 times. This is more than enough to guarantee high accuracy, with each clone placed correctly and in the right orientation along the genome. Sequencers will then fill in remaining gaps as they finish the genome. For many groups, the rough drafts are already providing good data. Fragmented though they are, these drafts capture 99% of the genes, notes Bennetzen. Both Monsanto and Syngenta say they are identifying genes that are important not only for rice but also for other cereals. Briggs says they have developed a microarray to study gene expression and are already moving into proteomics. “We're deep into the discovery of a lot of functional aspects of rice and other cereals,” he says. Gerard Barry, a molecular biologist at Monsanto, confirms that his company, too, has used its sequence as a basic research tool to do a wide range of plant studies, to breed new varieties at an accelerated rate, and to look for genes involved in productivity and stress tolerance. And Cornell's McCouch reports, “Breeders are mining the [public] sequence data for improving traits such as grain quality and pathogen resistance.” The rough drafts are being used for more basic research as well. Susan Wessler, a molecular geneticist at the University of Georgia in Athens, is studying how plant species diverge, focusing on the role of transposable elements: blocks of DNA that have been copied from one location to another in the genome. Having the BGI data on indica “is absolutely fantastic,” she says, as it allows comparisons of where transposable elements occur and don't occur between two closely related subspecies. “To my knowledge, it's the highest organism where there are two sequences of subspecies; it saved us literally a year of work,” she says. But for other groups, the drafts may not hold the answers they seek. Masahiro Yano, a rice geneticist at RGP, wants more precise data. His research seeks to link important agronomic traits, such as pest resistance and flowering time, to particular regions of chromosomes. Yano says that because the IRGSP sequence data are accurately tied to maps of chromosomes, researchers can use computer programs to quickly home in on candidate genes. He has used this approach to find several genes that control flowering time in relation to length of day, and he hopes eventually to have a whole collection of genes so that researchers could control the flowering time of plants. This would allow breeders to take varieties with desirable traits and move them from northern to southern latitudes, or vice versa. “This is an ideal example of the promise of having the rice genome [sequence],” Yano says. He adds that for the rough drafts, where the sequence data are not tied to a map, “you can't do this isolation and identification so easily.” Such requirements make it imperative that the IRGSP's work be completed, argue Wing, Sasaki, and others. “I think Monsanto and Syngenta can get enough information out of the drafts to patent genes,” explains Wing, “but we need to know more about the regulatory elements.” But Wing and others are concerned that the public and the funding agencies will get caught up in the excitement over the rough drafts and that funds for finishing the job will dry up. In the United States, for example, there are no funds specifically set aside for finishing the rice genome sequence. Already, companies and researchers interested in maize are pressing to begin pilot sequencing of that genome. “It will be up to the community to decide” what's more important, says NSF's Dilworth. Failing to complete the genome sequencing would be a big mistake, says Wing, not just for basic research but for anyone interested in any of the cereals. ### Rice Genomes Timeline 1991 Japan reorganizes a rice genome mapping project into the Rice Genome Research Program, a projected 7-year,$25 million effort largely funded by horse-racing proceeds funneled through the Japan Racing Association.

1991

Colinearity among cereal genomes is established; Michael Gale of the John Innes Centre in Norwich, U.K., concludes that “wheat is rice.”

May 1993

## Shotgun success

Yang says his “addiction to sequencing” has helped the institute achieve that economy of scale. With work on the human genome sequence well under way, Yang landed two major sequencing jobs in 2000 that broadened the center's scope. In April 2000, he convinced CAS to support sequencing indica (Science, 5 May 2000, p. 795). Then in October, BGI reached an agreement with the Danish Pig Genome Consortium, a joint public-private endeavor, to sequence the pig (Science, 3 November 2000, p. 913).

For both projects, BGI chose a sequencing technique known as the whole-genome shotgun. Traditionally, sequencing has required mapping the genome, then developing a library of clones, or relatively short strings of DNA, tied to a known location, which are then sequenced. In the shotgun approach, the entire genome is broken into pieces and sequenced. Then powerful computers and sophisticated software sort the pieces into the proper order.

Shotgunning appealed to BGI officials because the maps and clones needed for the traditional method were not available for indica and would have taken years to prepare. Just as importantly, “we felt that mastering the shotgun technique would be the next step in our development,” says Liu Bin, BGI's chief of research and collaborations.

The BGI scientists faced a big challenge in developing the computer expertise to pull off the whole-genome shotgun. Initially barred from acquiring U.S. supercomputers because of security-related export restrictions, BGI became one of the best customers of Dawning, a homegrown supercomputer maker. Particularly critical was the so-called assembler, which pieces together the data from the sequenced fragments. Whereas the pig project can use the human genome as something of a template, rice required its own assembler. “The assembler was our biggest worry,” Yang says.

Rather than starting completely from scratch, BGI decided to modify the assembly program used in the public Human Genome Project, called phrap. A key step was developing a subroutine that would identify and temporarily mask repetitive strings, which make up 40% of the rice genome. By temporarily ignoring such repetitive sequences, the BGI assembler reduced the chance of making mistakes in stringing together sequence data and dramatically cut the computing time required. Li Songgang, a professor of bioinformatics at Beijing University and senior adviser to BGI, developed an algorithm to search for such strings, and the BGI group tested it against virtually all the sequence data for all organisms publicly available. The effort took nearly a year.

Once the group was convinced that it would work, BGI pulled out all the stops to beat the competition. In particular, Yang was taking aim at Syngenta Corp., which had announced that it had completed a draft of the rice genome sequence in early 2001 but had never made the data public. Starting last July, the sequencing team was split into two, 12-hour shifts. That allowed BGI to keep the machines running 24 hours a day for the 74 days needed to complete the actual sequencing. Then came the analysis team, working practically around the clock. “It was like going into battle,” says bioinformatics head Wang.

Most of the Beijing staff lives in a housing development barely a kilometer down the road from the center. But to save commuting time, mattresses and sleeping bags were spread out in the halls. And some people didn't even make it that far. “I just slept in my chair,” says Han Yujun. Ping-Pong was the only diversion from the data crunching. On 8 October the paper was sent off to Science.

The completion of the draft is certain to put BGI on the sequencing map. Olson calls it “a major accomplishment.” Adds plant geneticist Jeff Bennetzen of Purdue University in West Lafayette, Indiana, “the Chinese showed how quickly you can do this if you take a modern approach.”

Yang has drawn up a long list of organisms he thinks should be sequenced, and the institute has already dipped its toe into proteomics and drug discovery, including a project to isolate the active compounds in the herbs used in traditional Chinese medicine. “Diversification is a challenge all of the sequencing centers must face,” says Olson.

So is finding money to take the next step. Chen Zhu, CAS vice president for life sciences, says BGI is free to compete for project grants but that “we think the big genome sequencing projects should depend on international cooperation.” Yang isn't saying where he plans to get the institute's next round of funding. But after going from an empty building to a paper in Science in 40 months, Yang is confident that he hasn't run out of miracles.

12. MATHEMATICS

# Erdös's Hard-to-Win Prizes Still Draw Bounty Hunters

1. Charles Seife

Half a decade after his death, a colorful mathematician continues to spur his colleagues with challenges—and checks

It takes more than death to keep a good mathematician down. Among many other things, Paul Erdös proved that. In life, the world's premier number theorist supported himself by wandering from colleague to colleague, freeloading off friends and strangers, and, in return, sharing his vast mathematical insight with all comers. His 6-decade frenzy of near-nonstop work resulted in more than 1500 papers that link him to almost every academic mathematician in the world. Erdös's death in 1996 has slowed, but not stopped, his publication rate: Over the past 5 years, journals have published some 62 new papers bearing his name. And he is still guiding the research of mathematicians with the problems he left behind.

Problems, problems, problems. Erdös hurled them out relentlessly, in lectures, papers, and informal talks: problems in number theory, logic, graph theory, geometry, combinatorics, and other disciplines. Every mathematician agrees on their importance. Yet nobody has listed them all; nobody even knows how many there are.

“I'd say it's in the small number of thousands,” says Ronald Graham, a mathematician and computer scientist at the University of California (UC), San Diego, who managed many of Erdös's affairs in the last few years of Erdös's life. András Hajnal, a set theorist at Rutgers University in New Brunswick, New Jersey, says he can name at least 100 in set theory alone.

As remarkable as their variety are the incentives Erdös attached to them. Early in his career, Erdös began to offer small prizes for solutions to his problems. “They were $10,$25, $100. It was a way to stimulate people and calibrate how difficult he thought the problem was,” Graham says. “Some prizes were larger: There is a famous$3000 one, and there's kind of a $10,000 one, but it's not well stated. There are hundreds and hundreds of such problems.” Alert listeners jotted the figures down in lecture notes; compliant journal editors let Erdös publish them in his papers. During Erdös's lifetime, the mounting most-wanted list could reap unexpected windfalls for his colleagues. “Some years ago, he was visiting in Athens, Georgia. He was going back into town, and my colleague Helmut Maier was giving him a ride,” says Carl Pomerance, a mathematician at Lucent Technologies' Bell Labs in Murray Hill, New Jersey. The conversation naturally turned to mathematics and to a theorem that Maier had just proven. “‘Maybe I offered a prize for that,’ said Erdös, and instead of going to town, they went to the library,” says Pomerance. Sure enough, Erdös had offered$100 for that particular problem in a math journal, and he paid up. “I said to Erdös, ‘That's a pretty expensive taxi ride,’ and he found that hilarious,” adds Pomerance.

Collecting a bounty could be tricky. Erdös often changed his mind about a problem's value, stating one sum in a lecture and another in print, and he could alter the figure on a whim based on his own sense of aesthetics. “I solved a $250 problem, but I only got$50 because Erdös didn't like the proof,” says Hajnal. Hajnal's proof, which had to do with partitioning the real numbers into different sets, hinged upon a logical trick rather than any special properties of the numbers. Erdös also abhorred proofs related to Gödel's incompleteness theorem, which states, in essence, that certain propositions are unprovable and unfalsifiable. As a result, there are statements that one can simply declare to be true or false. In the early 1960s, mathematician Paul Cohen of Stanford University proved that the answer to a crucial question about set theory was both yes and no. “Erdös didn't like that,” says Hajnal. “He stopped offering prizes for that kind of question.”

Erdös's cashless, bank-free lifestyle also played havoc with his incentive system. In 1993, Graham tried to iron out the snags by having Erdös sign some of Graham's checks to be filled out later, “suitable for framing.” Graham himself provided checks suitable for cashing. Since Erdös's death, Graham says, his informal foundation has cost him about $3000 in award money. He estimates that the outstanding bounties on unsolved problems total about$25,000.

The arrangement has worked smoothly on the whole, Graham says, but there have been hiccups. One occurred in 1999, when Ernie Croot, currently a postdoc at UC Berkeley, solved a $750 Erdös problem related to the ancient Egyptian style of writing fractions. Instead of writing a fractional number as a ratio of two numbers, like 7/8, Egyptian mathematicians expressed it as a sum of “unit fractions” whose numerator is always one, such as 1/2 + 1/4 + 1/8. Erdös challenged his peers to find out, given certain restrictions, how large the denominators have to get to represent a given number. When Croot announced his solution, Graham duly prepared one of the Erdös-signed checks. “Graham presented it to Ernie at a meeting,” says Pomerance. “But Ernie, being a poor graduate student at the time, didn't realize that it was only suitable for framing—he went and cashed it.” Graham still marvels that his bank honored Erdös's posthumous check. Graham's banking skills, of course, don't explain the staying power of Erdös's problems. That rests in the mathematical power of the problems themselves, some of which have turned out to be incredibly subtle and important. “One of the attractions of a lot of these problems is that they are of unknown difficulty,” says Graham. “It might be a marshmallow, tasty and light, or it might be an acorn and grow into something huge.” At the top end, those acorns can grow into huge oaks indeed. The record-holding$10,000 problem may never yield a check, Graham says, because it's for a “significant improvement” upon a previous result: a subjective judgment. But the next few in line have proven Herculean in their difficulty. In the 1930s, for example, Erdös challenged number theorists to prove a conjecture having to do with a collection of whole numbers: If the collection has a certain type of “density,” then there must be an arithmetic progression (a set of evenly spaced numbers, such as 4, 8, 12) of arbitrary length. In 1958, Klaus Roth of University College London won mathematics' top honor, the Fields Medal, in part for solving a special case of Erdös's problem. The problem itself will net $1000 for the person who eventually solves it in full. The famous$3000 problem is a variation on the same theme: to prove that there is an arithmetic progression of arbitrary length if the reciprocals of the elements in a collection goes off to infinity. If true, it will imply several important results in number theory. For instance, it will mean that there are arbitrarily long arithmetic progressions of prime numbers. So far, however, the proof has resisted all comers. “Nobody can touch it,” says Graham.

Most of the other problems are more accessible. “Erdös had an unparalleled innate ability to form problems just out of current reach,” says Graham. “You can get it if you stand on your tiptoes and jump a little bit. Once you're there, it's one more piton into the cliff.”

How rare that knack for problem-framing is can be seen in the fate of some who have dared to emulate it. In 1989, for example, John Conway, a mathematician at Princeton University, offered $10,000 for the solution to a problem about a peculiar sequence of numbers. Colin Mallows, a statistician currently at Avaya Labs in Basking Ridge, New Jersey, solved it within a few weeks. Conway paid up. “I don't know what kind of maneuvers he went through to get$10,000, but he sent it,” says Mallows. “I was embarrassed and sent it back.” Mallows says that Conway had misspoken and meant the prize to be $1000 rather than$10,000. “Most importantly, the problem was not of Erdös caliber, so that both John and I would be embarrassed to have it enshrined as the world-record prize.” The two settled on \$1000 as the bounty.

Such cautionary tales are few; for most mathematicians, Erdös's beyond-the-grave sweepstakes remains the only game in town. Graham still has a few Erdös-signed checks left and will pay for many of the Erdös problems. So will Andrew Beal, a Texas banker and amateur mathematician who made a splash in the mathematical world in 1997 when he offered a prize to anyone who could solve a problem resembling Fermat's Last Theorem (Science, 21 November 1997, p. 1396). Graham says he keeps a little extra padding in his bank account to ensure that he can pay up the odd prize here or there—although he's no more concerned about being bankrupted by prize problems than he is about a run on the bank. Not that it would matter if the money ran out; becoming a piece of the Erdös legend is enough of a reward for most mathematicians.

Deciding which claims are bona fide may pose more of a challenge. Unfortunately, the problems are in disarray, because Erdös talked at so many universities and wrote papers in so many journals and languages. (After hearing a limerick about “a paper by Erdös/Written in Kurdish,” he allegedly tried, unsuccessfully, to publish a paper in that language.) Graham and his wife, UC San Diego mathematician Fan Chung, collected more than 125 Erdös problems in graph theory and published them in 1998. Hajnal says he hopes someday to compile all the problems into a single volume. “It's still being planned, but not much is happening,” he sighs.

Despite the disorder, most mathematicians would agree that their field has benefited from Erdös's gentle goading. His hit list spurs them to approach problems they wouldn't normally tackle, says Jerrold Grossman, a graph theorist at Oakland University in Rochester, Michigan. “It's sort of like Fermat's Last Theorem, which guided research in number theory for the good,” Grossman says. Graham reaches farther back for an analogy. “The problems serve as Socratic gadflies,” he says, “reminding us of how much we have to learn.”

13. ASTROPHYSICS

# Closing In on the Cause of the Cosmos's Biggest Blasts

1. Govert Schilling*
1. Govert Schilling's book Flash! The Hunt for the Biggest Explosions in the Universe will be published this month by Cambridge University Press.

A flurry of new observations brings astronomers ever closer to understanding the cataclysmic explosions known as gamma ray bursts

When a massive earthquake strikes, no sane seismologist would try to trace the rumble of a passing truck amid the noise of the aftershocks. But astronomers may have pulled off an equally challenging feat: detecting the glimmer of a supernova explosion in the fading afterglow of a titanic gamma ray burst (GRB). The new results support the popular idea that GRBs—the most energetic explosions in the universe—are the birth cries of black holes, the collapsed remnants of dying stars. “It is definitely strong evidence,” says theorist Martin Rees of Cambridge University, U.K. A second team studying supernova debris near another GRB has strengthened the black hole connection, but its observations challenge the conventional wisdom on the sequence of events.

GRBs are brief, incredibly powerful flashes of high-energy radiation that astronomers detect in distant galaxies. According to the popular “collapsar” theory, a burst occurs when a star at least 10 or 20 times as massive as the sun collapses into a black hole, spewing jets of matter into space at close to the speed of light. At about the same time, the dying giant star blows away its outer layers in a more modest, though still enormous, supernova explosion. Because they are mainly powered by the slow decay of radioactive nickel, supernovas take days or weeks to reach their maximum brightness. Thus, the collapsar model predicts that astronomers should see a GRB first, followed by a temporary brightening of the fading afterglow of the burst, at the time the supernova reaches its peak brightness.

Ever since April 1998, when Dutch astronomers discovered a supernova at the exact spot where a GRB had been detected (Science, 19 June 1998, p. 1836), astronomers have been checking other bursts for such telltale supernova signatures. In a handful of cases, tentative “bumps” have been found in the light curves (brightness plots) of GRB afterglows, but the measurements have been too imprecise to be truly convincing or to rule out alternative explanations.

Recent observations with the Hubble Space Telescope, however, leave little doubt that the “bump” in the light curve is real. A team of astronomers led by Joshua Bloom of the California Institute of Technology in Pasadena used Hubble to study the afterglow of GRB 011121, which the Italian-Dutch x-ray satellite BeppoSAX detected on 21 November 2001. In an online paper submitted to The Astrophysical Journal Letters, Bloom and colleagues write that the brightness and the reddish color of the bump in the afterglow's light curve are “remarkably well described” by a supernova. The team concludes that the results serve as “compelling evidence for a massive star origin of … gamma ray bursts.”

Other astronomers say they are impressed but want more. “This is by far the most detailed observation yet of a bump in an afterglow light curve,” says Paul Vreeswijk of the University of Amsterdam, the Netherlands, “but I'd like to see a spectrum before I'm fully convinced that it was produced by a supernova.” Stan Woosley of the University of California, Santa Cruz, the father of the collapsar theory, agrees. “We need a spectrum,” he says. “This [bump] smells like gunpowder, but it is no smoking gun [yet].” And Rees adds that “it would be more important if we could infer the nature of the precursor star and check whether it was, for instance, especially massive.”

A team led by James Reeves of the University of Leicester, U.K., has taken a step in that direction. Using the European Space Agency's orbiting XMM-Newton x-ray observatory, the researchers studied the afterglow of another GRB, GRB 011211, which BeppoSAX spotted on 11 December 2001. In this week's issue of Nature, Reeves and colleagues report that they have found the spectral signatures of silicon, sulfur, argon, magnesium, and calcium—heavy elements that massive stars produce by nuclear fusion and then eject into space during supernova explosions. From the x-ray spectra, the team concludes that the supernova debris forms a shell billions of kilometers in diameter, expanding at some 26,000 kilometers per second—87% of the speed of light. Ions detected in the debris peg its temperature at a sizzling 50 million degrees. Apparently, a massive giant star exploded as a supernova, and then a few days later the ejected shell was heated by the energetic radiation of a subsequent GRB.

This surprising order of events—first the supernova, then the GRB—is predicted by a two-stage theory of GRBs called the “supranova” model, which a minority of theorists prefer to the collapsar model. In the supranova model, the core of the exploded star first collapses into a dense neutron star, triggering a supernova. The GRB occurs later, when the neutron star further collapses into a black hole (Science, 3 November 2000, p. 926). In the original supranova model, the GRB occurs weeks or even months after the supernova explosion. The Hubble observations of GRB 011121 rule out that timing, Bloom and colleagues say, but they could accommodate a much shorter delay suggested by the XMM-Newton observations.

But Mario Vietri of the Third University of Rome, one of the original authors of the supranova model, says the long-delay model is still viable. Vietri points to another online paper, in which the Hubble team cites radio and infrared observations of GRB 011121 as evidence that the doomed star ejected large amounts of gas and dust in the final stages of its life. In that case, Vietri says, the suggestive bump in the afterglow light curve might be due to radiation from the GRB heating this dusty environment or reflecting off the dust. And if the bump is not due to a supernova, it's conceivable that the supernova explosion happened long before the GRB.

“More work needs to be done before they can dismiss the effects of dust and light echoes,” says Vietri. Obtaining a high-resolution spectrum will do the job, says Woosley, “but this has been difficult to come by observationally.” It may well take the help of a dedicated GRB satellite such as NASA's Swift, due to be launched in 2003, to solve the many mysteries that still litter the field of GRB research.