# News this Week

Science  04 Apr 2003:
Vol. 300, Issue 5616, pp. 28
1. CLINICAL RESEARCH

# A Setback and an Advance on the AIDS Vaccine Front

1. Jon Cohen

The frustrating search for an AIDS vaccine entered a new phase this week, as hopes for one approach all but died while prospects for another got a welcome boost when two of the world's largest pharmaceutical companies working on AIDS vaccines announced that they are joining forces.

First, the bad news. For the past month, AIDS researchers have been eager to see the detailed results of the first full-scale efficacy test of an AIDS vaccine—a 5000-person, 3-year trial conducted by VaxGen of Brisbane, California, testing a genetically engineered version of HIV's surface protein, gp120. The company took a lot of heat when it announced on 24 February that although the vaccine failed to prevent infection in the test population as a whole, it offered some protection in blacks, Asians, and other minority groups (Science, 28 February, p. 1290). In a 31 March talk at a Keystone Symposium in Banff, Canada, VaxGen biochemist Phillip Berman provided the first scientific presentation of the results. He offered few surprises.

Berman reported that the vaccine plainly failed overall; moreover, vaccinated people who became infected did not control the virus any better than those who had received placebo shots. That dashed VaxGen's hopes that the trial might turn up a secondary measure of efficacy. “There was no indication of protection in the overall trial population,” said Berman.

As for the results in racial subgroups, Berman took a cautious stance. When VaxGen first reported the results, it said the vaccine showed a statistically significant 67% efficacy in the relatively small number of black, Asian, and mixed-race trial participants. Berman conceded that the company's statistical analyses relied on the most generous interpretations of the results, but he went on to describe in detail several immunologic and virologic hints that the vaccine might have worked in some people.

Preliminary data show, for example, that women and black men had higher levels of anti-HIV antibodies than the white men in the study. That, Berman said, might help explain their lower infection rates. In addition, molecular analysis of the HIV strains that infected people who had been vaccinated indicated that they differed substantially from the one used to make the vaccine, suggesting that the vaccine did not stop divergent strains but may have protected against similar strains. “The data won't go away,” said Berman. The company hopes to report results later this year from another large efficacy study of its vaccine now underway in Thailand.

Most researchers were not impressed, however. A leading critic of the VaxGen vaccine, John Moore of Cornell University's Weill Medical College in New York City, had a cutting, if understated, reaction to Berman's talk: “It was boring in many respects.”

There was a bright spot on the AIDS vaccine front, though: The Franco-German company Aventis Pasteur and Merck & Co. of Whitehouse Station, New Jersey, announced that they plan to launch studies that combine their AIDS vaccines.

As Merck's Emilio Emini explained at the Banff meeting, tests with monkeys show that combining the two companies' vaccines in a one-two punch may work better than any of the strategies the companies are testing separately in human trials. “We were driven by the data,” says Emini, who heads Merck's HIV vaccine program in West Point, Pennsylvania.

Merck and Aventis Pasteur have both developed AIDS vaccines that stitch HIV genes into harmless viruses. Emini reported that, in monkey studies pitting various combinations of vaccines against each other, Merck's adenovirus-based HIV vaccine chased with a booster dose of the Aventis canarypox/HIV vaccine led to some of the strongest immune responses observed. Human trials of the Merck vaccine followed by the Aventis preparation are awaiting U.S. regulatory approval and could start in the next few months.

AIDS vaccine researcher Norman Letvin of Beth Israel Deaconess Medical Center in Boston says there are “very good data” that pox viruses such as canarypox can give a powerful boost to so-called killer cells, immune warriors that selectively target and destroy cells infected with HIV and other invaders. The proposed clinical trial of the two vaccines is “an experiment that begs to be done,” he says.

2. STRUCTURAL BIOLOGY

# Canada Vaults Into Drug-Oriented Protein Research

1. Robert F. Service

In what is being billed as the largest health research project ever in Canada, the University of Toronto will lead a new $68 million effort over 3 years to map the three-dimensional atomic structure of 350 human health-related proteins. The public-private venture is the latest in the red-hot field of structural genomics, which aims to carry the revolution of high-speed biology beyond the genome to proteins that drive the chemistry of cells. Just under half the funds come from Canadian agencies, and the rest from private sources, including$28.4 million from the Wellcome Trust, a British charity, and $4.75 million from the pharmaceutical giant GlaxoSmithKline. The project, called the Structural Genomics Consortium (SGC), stands out for its large scale and tight focus. The nine structural genomics consortia in a similar U.S. program, for instance, receive a total of$50 million a year and are attempting to catalog very diverse proteins. SGC opted to focus on proteins that could lead to medications for a wide variety of human diseases ranging from cancer to neurological disorders and microbial infections, says Aled Edwards, a structural biologist at the University of Toronto, who will lead SGC.

“It's very exciting,” says Tom Terwilliger, an x-ray crystallographer at Los Alamos National Laboratory in New Mexico and head of the Mycobacterium tuberculosis Structural Genomics Consortium. “It's a very good thing that we have lots of different approaches coming out,” adds John Norvell, who coordinates the structural genomics program for the National Institute of General Medical Sciences in Bethesda, Maryland.

Norvell and others praise SGC's decision to require that all newly acquired protein structures be immediately deposited in public databases and made freely available to all researchers. “This is critical,” says Andrzej Joachimiak, a biophysicist at Argonne National Laboratory in Illinois and head of the Midwest Center for Structural Genomics. “You advance science much faster if you release the data and everyone can mine it.”

Terwilliger and others point out that the goal of solving 350 human protein structures in 3 years is very ambitious. To date, most structural genomics consortia have targeted bacterial proteins, which tend to be easier to express, isolate, and crystallize—all of which must be accomplished before a 3D structure can be determined by the most widely used method, x-ray crystallography. In 2.5 years, the U.S. groups have mapped out a total of 330 proteins.

But setting an ambitious goal was intentional, says Alan Bernstein, president of the Canadian Institutes of Health Research in Ottawa, one of the agencies funding the work. “If you wait until everything is ready, it's too late. You should always be pushing the limits of the technology,” he says.

According to Edwards, SGC leaders will spend the next year setting up their operation, building new lab space, and hiring up to 100 researchers at sites at the University of Toronto and the University of Oxford, U.K. When the staff members are all on board, he says the 3-year countdown clock will start ticking.

3. EVOLUTION

# Colorful Males Flaunt Their Health

1. Elizabeth Pennisi

From flies to birds to people, males tell potential mates, “I'm the best.” Their boasts come in many forms, from colorful body parts to heartfelt serenades. The most flamboyant or melodic suitors are females' favorites, but evolutionary biologists have often wondered exactly why. On pages 103 and 125, two research teams show that in some birds, a female's attraction to a brighter bill ensures her a healthy mate.

In one study, Jonathan Blount, an evolutionary biologist at the University of Glasgow, U.K., and colleagues manipulated the amount of pigments called carotenoids in the diets of zebra finches and then tested the birds' immune responses. Carotenoid supplements meant brighter bills and healthier birds, they report. Conversely, a team led by evolutionary biologist Bruno Faivre of the University of Burgundy in Dijon, France, found that blackbirds whose immune systems are under stress have duller bills.

“The [two] studies are important and complementary,” says Joseph Waas of the University of Waikato, New Zealand. Together they provide experimental evidence that bill color is a true indicator of the male's fitness because carotenoids help boost the immune system.

Scientists have known for almost 30 years that carotenoids are linked to immunity. Parasite infections can lower carotenoid concentrations in the blood, and studies in humans have suggested that eating more carotenoids—carrots and other vegetables are chock-full of them—can reduce the risk of some chronic diseases. At the same time, ecologists have found that these pigments are important for animals' mating displays: They provide the color for vivid feathers and beaks that females find sexy.

In the past decades, researchers have tied these two lines of research together. Marlene Zuk, an evolutionary biologist at the University of California, Riverside, and her colleagues suggested that extraordinary sexual displays—feathers, crests, and so on—might indicate how well a male copes with disease and parasitic infections. And in 1994, George Lozano, a freelance evolutionary biologist living in Ottawa, Canada, proposed that when the immune system is under pressure, those display colors dim, which would make sick males less desirable.

To test this theory, Blount and his colleagues manipulated the amount of carotenoids in 10 pairs of zebra finch brothers. One of each pair received water fortified with carotenoids; the other drank plain water. Within a month, carotenoid supplements turned bills much redder, the researchers report. And females preferred these males to their drab siblings.

Carotenoid supplements also boosted the birds' immunity. When the researchers injected the finches with a protein that causes swelling, those with the reddest beaks had more carotenoids in their blood and the strongest immune response. The implication: When females opt for the colorful male, they are selecting a mate with a strong immune system. “The characteristics that females really pay attention to are things that reflect the [male's] day-to-day well-being,” explains Zuk.

In their experiments, Faivre and his colleagues followed the fate of carotenoids over time. They measured the amount of carotenoids in the bills of about 50 blackbirds. Then they injected sheep red blood cells into all but 15 birds. The treated birds reacted to the foreign cells by mounting a strong immune response. As they did, the amount of carotenoids in the bills—and the intensity of bill color—dropped, Faivre reports. “When the animal gets infected, the carotenoids are mobilized and used to fight off infection,” says Lozano.

Beak color may thus be a truer reflection of a male's current health than feathers, which can't release carotenoids and change color only during molting. When bright-billed males claim they're the best, females are therefore right to listen.

4. ENDOCRINE DISRUPTERS

# Lab Accident Reveals Potential Health Risks of Common Compound

1. Jon Cohen

n August 1998, geneticist Patricia Hunt noticed a bizarre change in the eggs of the female mice she was studying. For some inexplicable reason, the chromosomes in 40% of the eggs looked abnormal—a wild jump from the 1% to 2% abnormality her lab typically observes. Something seemed to have gone terribly wrong with meiosis, the process that separates chromosomes during reproduction so that when egg and sperm come together, they each contribute half the genetic material to an embryo. Hunt, whose lab at Case Western Reserve University in Cleveland, Ohio, specializes in problems with meiosis—which in humans causes more birth defects, mental retardation, and miscarriage than any other factor—had her lab workers redo the study twice, yielding the same baffling results each time.

The first clue came that fall, when Hunt noticed that her mouse cages, made of a plastic called polycarbonate, appeared to be melting. She found that a lab worker had mistakenly washed the cages with a highly alkaline detergent. “He was a temporary worker who made a lasting impression,” Hunt says. Hunt, Terry Hassold, and their colleagues eventually pinned the meiotic abnormalities on a chemical called bisphenol A (BPA) leaching from the damaged plastic. Although several labs have shown that BPA, a compound widely used in plastics manufacturing, can disrupt the reproductive system of rodents, none had previously shown an effect on meiosis.

The detective work, described in a paper in the 1 April issue of Current Biology, has thrust Hunt's lab into the middle of the controversial field of endocrine disrupters. Some researchers and environmentalists have argued that low levels of certain synthetic chemicals in the environment are causing reproductive problems in wildlife and perhaps humans. BPA, which weakly mimics the effects of estrogen has been among the suspects.

Researchers from diverse disciplines say the work deserves serious attention. “It's fascinating,” says Charles Epstein, a developmental biologist at the University of California, San Francisco, who studies chromosomal imbalances in mice. Even Stephen Safe of Texas A&M University in College Station, a leading skeptic of evidence linking endocrine disruptors to health problems, says, “I think their data are very interesting.” Leading proponents of BPA's harmful effects predictably have stronger words: “I look at this as a watershed paper,” says reproductive biologist Frederick vom Saal of the University of Missouri, Columbia, whose lab has published several studies of BPA's impact on mouse reproductive development.

Hunt and co-workers first noticed problems when they took a routine “snapshot” of the meiotic process in their mice. They found that 40% of the mouse eggs failed to assemble their chromosomes neatly on the spindle apparatus, a step that must occur for separation to take place properly. They also found abnormal numbers of chromosomes, an aberration called aneuploidy, in about 12% of the eggs. “When I read that I said, ‘Zowee, that's really out of sight,’” says John Eppig, a reproduction biologist at the Jackson Laboratory in Bar Harbor, Maine.

Once they suspected BPA, Hunt and co-workers recreated the accident. They intentionally damaged polycarbonate cages and water bottles with detergent and compared the mice in those cages with animals kept in undamaged cages with glass bottles. They found the same levels of meiotic error in the mice kept in the damaged cages. They then showed that BPA was the culprit by adding the chemical to mouse water; it caused chromosomal problems, though not as severe. “It's probably the only convincing demonstration of an environmental effect on the frequency of aneuploidy,” says Dorothy Warburton, a cytogeneticist at Columbia University in New York City who has studied aneuploidy extensively. “It's a little scary.”

Although Hunt stresses that no data clearly link BPA to human aneuploidy, she thinks the question deserves study because the compound is so widely used: Plastics that contain it are made into baby bottles, the liners of food cans, dental sealants, and many other common products. More immediately, she says her finding should cause researchers to consider retiring their polycarbonate cages, which naturally degrade over time. And for her own research, this serendipitous finding provides a powerful new tool to create and study aneuploidy in mice.

5. NUMBER THEORY

# Prime Proof Helps Mathematicians Mind the Gaps

1. Dana Mackenzie*
1. Dana Mackenzie is a writer in Santa Cruz, California.

PALO ALTO, CALIFORNIA—In an unexpected breakthrough, an American and a Turkish mathematician have brought the erratic behavior of enormous prime numbers into dramatically sharper focus. Last week, speaking at the American Institute of Mathematics (AIM) here before an audience of 50 number theorists who had been buzzing about the new result all week, Dan Goldston of San Jose State University, California, described how he and Cem Yildirim of Bogaziçi University in Istanbul, Turkey, had proven that primes get more and more “clumpy” as they get larger.

“This is the biggest excitement that prime number theory has seen since 1965,” says Hugh Montgomery of the University of Michigan, Ann Arbor. Enrico Bombieri of the Institute for Advanced Studies in Princeton, New Jersey, agreed that Goldston had produced a “magnificent proof.”

The distribution of prime numbers—integers that can be divided evenly only by themselves and 1—has vexed mathematicians for centuries. Primes may pop up in clumps, such as the numbers 101, 103, 107, 109, and 113, or at huge intervals. One of number theory's most celebrated open questions, the Twin Prime Conjecture, states that “twin primes”—those that crop up two numbers apart, such as 17 and 19 or 101 and 103—keep appearing forever as numbers get bigger. But like much else about prime gaps, its truth or falsehood remains a mystery. “[Primes] grow like weeds among the natural numbers, seeming to obey no other law than that of chance, and nobody can predict where the next one will sprout,” number theorist Don Zagier wrote in 1977.

Goldston and Yildirim's proof takes a huge step toward understanding how “weedy” the primes are. Earlier mathematicians had shown that primes get sparser as they get larger. If n is a prime number, then the gap to the next prime will (on average) be the natural logarithm of n, or log n. But no one knew how clumpy the spacing is. Can two consecutive primes fit into a much smaller gap than log n? And can many primes fit into a single log-n interval? Over the past century, progress on the first question has been slow, and progress on the second nonexistent. In 1965, Bombieri, then at the University of Pisa, Italy, and Harold Davenport of the University of Cambridge, U.K., showed that gaps of less than half the average size (i.e., 1/2 log n) occur infinitely often. Later mathematicians whittled the 1/2 down to 1/4. But that was like saying weeds grow a little bit less evenly than planted flowers, when every gardener knows that they grow in haphazard clumps.

The new proof answers both of the above questions in a single stroke. It shows that the shortest gaps between primes continue to shrink relative to the average gap (although, unfortunately for twin-prime aficionados, they could still be much longer than 2). What's more, there is no upper limit to the number of primes that can squeeze into the space “allotted” for one.

“I never thought we'd get this result in my lifetime,” says Goldston, who has been working on prime gaps for 20 years and with Yildirim for 3 years. As recently as last fall, the two were still trying to achieve tiny improvements to the Bombieri-Davenport estimate and losing hope that they would succeed even at that. The key breakthrough came after Goldston discussed the problem with Roger Heath-Brown, an Oxford number theorist who was visiting AIM.

“We basically had the right blueprint, but we had the wrong tool,” Goldston says. With a new “approximate” or “truncated” prime-pair detector, the formulas they had slaved over for 3 years suddenly became simpler—and vastly better. Goldston announced the new results at Oberwolfach, Germany, on 13 March. Says Heath-Brown, “This is something we've been wanting to do for 100 years, and it just came out of the blue.”

Number theorists say the new approach may also solve a standing problem on large prime gaps, for which the late Hungarian mathematician Paul Erdös offered a prize of $10,000 (Science, 5 April 2002, p. 39). As for the twin prime conjecture, Goldston doubts his method can be pushed so far. However, other experts say it's conceivable. “We have more mathematicians than ever thinking about these things,” says Brian Conrey, director of AIM. “We've seen that we have the power to find gems that were missed the first time around. Who knows what else we have missed?” 6. DRUG SAFETY # Conflict-of-Interest Allegations Derail Inquiry Into Antidepressant's 'Dark Side' 1. Daniel Bachtold CAMBRIDGE, U.K.—A government review of the safety of antidepressant drugs is in disarray following revelations that some experts involved in the inquiry are shareholders in a company that makes one of the drugs. Last week, the U.K.'s Medicines Control Agency (MCA) disbanded a panel that had been scrutinizing the track record of selective serotonin reuptake inhibitors (SSRIs), a widely prescribed group of drugs that includes the brands Prozac and Seroxat (sold as Paxil in the United States). The panel bust-up came on the heels of a report in The Guardian, a U.K. newspaper, disclosing that two members of the four-person panel hold shares of GlaxoSmithKline, which produces Seroxat. MCA is now seeking to form a new panel but has not said publicly when it expects the inquiry to resume. The government launched the SSRI review last autumn to investigate anecdotal evidence linking SSRIs to violent and suicidal behavior in a handful of patients on the medication. Reports on such unexpected side effects have in recent years received widespread publicity and sparked litigation. For example, a Wyoming court in 2001 ordered SmithKline Beecham—now GlaxoSmithKline—to pay$6.4 million to the surviving family members of Donald Schell, who shot his wife, daughter, and granddaughter before killing himself. Schell started to take Paxil 2 days before his killing spree. And last month, a coroner in Wales called for Seroxat to be taken off the market pending an investigation into the death of a retired schoolteacher who had committed suicide shortly after starting a prescription. There has been no move to ban Seroxat, which GlaxoSmithKline expects to be vindicated. “We believe in the safety profile of Seroxat and will vigorously defend the integrity of our medication,” says a company spokesperson.

Many experts say that the anecdotal reports do not establish a firm link between the drugs and violent behavior, and they profess doubts that one will emerge. Nevertheless, “it looks as though the government has to do something to reassure patients and people who need to be treated,” says psychopharmacologist Philip Cowen, a Medical Research Council clinical scientist at Oxford University. Seroxat is effective in most patients treated for depression, adds Anthony Cleare, a psychiatrist at King's College London. “We risk throwing out the baby with the bath water,” he warns.

Others, however, believe that SSRIs may in fact be dangerous. Last November the MCA's review panel, chaired by Angus Mackay, director of Mental Health Services in Lomond and Argyll, Scotland, heard from David Healy, a psychiatrist at the University of Wales College of Medicine in Bangor, U.K. He presented epidemiological data on withdrawal symptoms and increased suicidal tendencies in patients treated with SSRIs. Healy, who served as an expert witness for the plaintiffs in the Schell case and has combed through GlaxoSmithKline archives not open to the public, puts it bluntly: “SSRIs can cause people to become suicidal or actually commit suicide,” he says.

MCA pulled the plug on the panel after the newspaper reported on the panel members' GlaxoSmithKline investments. In a 25 March statement, MCA said that “in light of issues raised at the meeting [with Healy] and the further work these will require, we are cognizant of the need to carefully consider the appropriate membership of an expert group. … Individuals' interests in the pharmaceutical industry will be taken into account.” The GlaxoSmithKline spokesperson insists that the company would not “under any circumstances seek to unduly influence” the inquiry. Members of the recently disbanded panel declined to speak with Science.

MCA said that a new expert panel will consider all relevant data including a detailed report from the Welsh coroner and patient testimony. The agency had not appointed a new panel before Science went to press.

7. STEM CELL RESEARCH

# Cells Find Destiny Though Merger

1. Constance Holden

Stem cells are heralded for their ability to morph into many types of tissues. But a year ago, researchers were set abuzz by two reports showing that some of these cells in test tubes altered their identity not by reprogramming to fit into their new environment but by fusing with host cells (Science, 15 March 2002, p. 1989). Now two teams have come up with the first evidence in living animals that adult cells can metamorphose by fusion. The finding suggests that it might be more difficult than some hope to develop therapies using adult stem cells: Fused cells may have extra chromosomes, which can spur cancer.

Earlier research showed that bone marrow stem cells that normally form blood cells are capable of generating new cells in the liver. Marrow from healthy mice was injected into mice with a liver disease whose own marrow had been knocked out. The donor cells helped rebuild the damaged liver.

That study, by Eric Lagasse of Stem Cells Inc. in Palo Alto, California, and colleagues, has been cited as among the best evidence for “transdifferentiation”—the ability of one type of cell to turn into another type of cell when put in a different environment. But in new studies analyzing the genomes of regenerated liver cells “we now better understand how those observations have been made,” says Markus Grompe of Oregon Health and Science University in Portland, who participated in the earlier study.

The new liver cells contain chromosomes from both the host and donor cells, as Grompe's team and a separate group led by George Vassilopoulos of the University of Washington (UW), Seattle, report online in Nature this week. So “the answer is very clear, they're made by fusion, not by differentiation,” says Grompe. The nuclei of some blood cells were reprogrammed as they fused with the nuclei of liver cells, leading them to operate as hepatocytes, he says.

The picture “really changes when you consider fusion instead of differentiation,” says David W. Russell of the UW team. “Transdifferentiation of an adult cell requires that it reprogram its nucleus based on extracellular signals,” he says. But with fusion, the reprogramming can be controlled by factors already in the host cell.

Scientists say the clinical potential of adult stem cells is not necessarily decreased if the mechanism is fusion. But it “makes disease correction less hopeful,” in the view of blood stem cell researcher John Dick of the University of Toronto, because it could induce genomic instability and loss of chromosomes (Science, 26 July 2002, p. 543). As Grompe explains, “when you mix and match chromosomes, there is potential to become malignant.”

8. SCIENCE MUSEUMS

# Nebraska Husks Research to Ease Budget Squeeze

When U.S. Fish and Wildlife agents investigate deaths of endangered wildlife in Nebraska, they turn to Tom Labedz to identify the carcass. So do state health officials worried about West Nile virus and researchers looking for DNA samples from 100-year-old remains of whooping cranes. But not for much longer. Last month, Labedz and five other collection managers at the University of Nebraska State Museum in Lincoln got pink slips, along with half of the museum's eight curators, in what some researchers fear could be the beginning of the end for the museum.

The layoffs are part of proposed cuts in the university's budget that will be debated next week within the academic senate before being finalized by the chancellor. Administrators say museum research was put on the chopping block because it ranked below undergraduate education and economically relevant research. Museum scientists retort that the collections are important for land-use decisions and heavily used for teaching and that cutting research will short-circuit the educational mission of the museum. “What will remain within a very short time will be static displays that will be tantamount to a decrepit theme park,” predicts entomologist Brett Ratcliffe, assistant director for research at the museum.

Nebraska's woes are a local symptom of economic pain that's affecting science across the country (Science, 21 March, p. 1826). In January, Nebraska Governor Mike Johanns proposed a 10% university cut that amounts to $21 million. In addition to trimming administrative offices and axing the Nebraska Forest Service, among other cuts, officials have proposed saving$1.1 million by firing four tenured museum curators and 19 collections managers, preparators, and other support staff. They would also eliminate the zoology collection and the 11,000 artifacts in the anthropology division.

The museum's high-profile paleontology collection would remain intact, as would entomology, parasitology, and botany. Four curators would be reassigned to academic departments and could continue to work on those collections. But museum scientists say the loss of support staff will cut research productivity and limit access by visiting researchers and could endanger the viability of the specimens. “What I fear most,” says museum paleontologist Mike Voorhies, “is that the maintenance and constant care that a collection has to have is going to go by the boards.”

Chancellor Harvey Perlman says those fears are exaggerated. “I think when the smoke clears, at least five of those collections will be actively generating research.” He notes that a tenured faculty position in public programs would remain and up to three museum specialists would be hired to maintain collections. But those assurances haven't stopped Labedz from digging through his freezers to save key specimens. “I'm afraid that once I'm gone, they're just going to be thrown out,” he says.

9. ECOLOGY

# 'Combat Biology' on the Klamath

1. Robert F. Service

Biologists charged with protecting endangered species are caught in a battle over water rights; a critical National Academy of Sciences report has exposed them to heavy fire

KLAMATH FALLS, OREGON—As a cold February night settles in, Rip Shively wades into the icy waters of Upper Klamath Lake near the Oregon-California border and hauls ashore a squirming, meter-long fish. The fish, netted as it prepared to spawn, is an endangered male Lost River sucker. Shively, a fisheries biologist at the U.S. Geological Survey (USGS), scans the fish with a wand. Similar tests on about a dozen earlier catches produced no response, but this time the wand beeps, indicating that the fish had been caught previously and tagged. Based on its size, Shively judges the sucker to be more than 15 years old, and from the tag's location on the fish's back, he surmises that it was tagged in 1995. That means it lived through three massive fish die-offs that hit the lake in 1995, 1996, and 1997. “She's beautiful,” he says. “A real survivor.”

Shively and colleagues at USGS and other government agencies, universities, and Indian tribes are racing to study the suckers and endangered coho salmon that swim the Klamath River below the lake. Their work guides federal plans to prevent the fishes' extinction. Federal wildlife managers used the scientists' preliminary research to recommend limiting the withdrawal of irrigation water from the lake in 2001 to minimize the impact of a regional drought on the endangered fish. But a report issued last year by the National Academy of Sciences (NAS) has cast a cloud over much of the fisheries research in the Klamath Basin. The report concluded that there was “no sound scientific basis” to justify turning off the irrigation spigot from the lake to farmers dependent on its water for crops.

The report's conclusion sparked an outcry in this small farming community that federal agencies are supporting “junk science,” and it bolstered calls for reforming or scrapping the Endangered Species Act (ESA). But over the past year, it has also sparked another, more muted outcry, this one among fisheries biologists. They contend that the report's analyses were simplistic, its conclusions overdrawn, and—perhaps worst of all—that the report has undermined the credibility of much of the science being done in the region if not fueled an outright antiscience sentiment.

“The opinions of [NAS's National Research Council] committee pretty much run counter to [those of] all the people who work in the region,” claims Mike Rode, a fisheries biologist at the California Department of Fish and Game (DFG) in Mount Shasta, California. “It was very offensive to many folks here,” adds Larry Dunsmoor, a research biologist working for the Klamath Tribes in Chiloquin, Oregon, who has studied the endangered suckers for the last 15 years. “It has been a very painful thing to see everything we have worked for over the past decade [described] as useless.”

Biologists here are caught in a classic western water fight, one that pits two of the region's major occupations—farming and fishing—against each other. At stake is the future not just of the suckers but of the salmon downstream—and the needs of the two fish populations are sometimes also in conflict. Instead of defusing these tensions, the biologists say the report has only made matters worse, ratcheting up an already hostile environment for many of the researchers working in the area. “Some people refer to it as combat biology,” says Ron Larson, a fisheries biologist at the U.S. Fish and Wildlife Service (USFWS) in Klamath Falls. “It's perhaps an exaggeration. But not by much,” he says.

Now all sides are girding for another major battle, and not just over the academy's final report, which is due out this summer. As of late March, the region's snowpack was a little over half the normal level. Because snow feeds the region's streams through typically dry summers, this year is shaping up to be nearly as parched as 2001. This month, the U.S. Department of Reclamation is expected to make its call on how dry a summer it foresees and therefore how much lake water it expects to release for irrigation. Court rulings expected this spring could tighten water supplies further if judges rule that additional water must be kept in area lakes and rivers to protect endangered fish.

How these events play out could set a new precedent for how much scientific proof is needed to take action to protect endangered wildlife. The ongoing NAS review of Klamath Basin water distribution evaluates whether wildlife managers have solid evidence that the actions they take will benefit species. This standard, some researchers say, is almost impossible to apply universally and could derail other protection efforts.

## Historic battles

A quiet, high desert landscape of sagebrush and juniper, the Klamath Basin seems dominated more by solitude than acrimony. The upper basin sits on the eastern flank of Oregon's southern Cascade Mountains and is one of North America's busiest way stations for migrating waterfowl. Before the arrival of the first white settlers in the 1820s, the basin was home to members of the Klamath, Modoc, and Snake Indians. A treaty with the U.S. government in 1864 guaranteed those tribes—by then collectively referred to as the Klamath Indians—abundant fish stocks in perpetuity. But those stocks were soon to face pressures they'd never seen before.

In 1902, Congress passed the Reclamation Act in an effort to promote settlement in the arid west. One of the effort's first undertakings was the Klamath Irrigation Project to support the establishment of farms in the basin. Its target was water flowing in and around Upper Klamath Lake, 32 kilometers long but, with an average winter depth of just 3 meters, practically a pond. Homesteaders diked and drained 16,000 hectares of marshland along the lake's northern reach. To the south, an 830-km network of canals carried lake water to hundreds of farms. Seven dams were added to lakes and streams in the region to provide additional irrigation water. In 2001, the Klamath Irrigation Project encompassed 97,000 hectares of irrigable land. In addition to the farms, water from the lake also feeds a series of wildlife refuges.

In a typical year, about 62,000 hectare-meters (500,000 acre-feet) of water is diverted from Upper Klamath Lake and surrounding waterways to irrigate nearby farms. Additional water is diverted from upstream tributaries before it reaches the lake. By the mid-1980s, the lake's fish had begun to show the stress of the annual drawdowns in water and the altered habitat. Phosphorus-rich runoff from farms and ranches prompted massive algal blooms every summer, turning the lake into a vast cauldron of pea soup. The blooms triggered wild swings in the lake's acidity level and dangerous drops in the amount of oxygen dissolved in the water.

These factors, together with chronic overfishing, caused a steady decline in the lake's two populations of suckers, the shortnose and Lost River suckers. By 1988, both species were on the endangered species list. The Klamath River coho salmon was listed as threatened in 1997. The listings required USFWS and the National Marine Fisheries Service (NMFS) to come up with recovery plans—known as biological opinions, or “BiOps”—for the fish and specify how much irrigation water the Bureau of Reclamation was allowed to divert to Klamath Irrigation Project farmers.

In its April 2001 BiOp for the suckers, USFWS biologists stated that, for the safety of the fish, the lake should not be drained below 4140 feet (1262 meters) above sea level, just below its historic level. Meanwhile, NMFS's opinion for the oceangoing coho salmon stated that the flow of water in the lower Klamath River had to stay above a minimum of 1000 cubic feet (28 cubic meters) per second.

But 2001 was a bad year for water. That winter, the Cascades tallied less than half the usual snowpack. Managers at the Bureau of Reclamation were in a bind. With so little water in the system and the need to fulfill the NMFS and USFWS recommendations, they announced in April that there would be no water diversions for irrigation. The head gates of the Klamath Irrigation Project were locked.

Farmers and many others in the surrounding community revolted. That summer, they staged continual demonstrations at the head gates calling for water to be released, and they even forced open the head gates briefly in an act of defiance. Angry signs sprouted throughout the community: “Some sucker stole my water,” read one common refrain. A notice at a local restaurant stated that U.S. government employees were not welcome. Hoping to stay out of the line of fire, USFWS and USGS biologists went so far as to remove the government license plates from their vehicles, a practice some still follow today.

In October 2001, after much of the fervor had died down, Interior Department Secretary Gale Norton asked NAS to determine whether the water cutoff was scientifically justified. The academy's scientific arm, the National Research Council (NRC), hastily organized a 12-member panel made up primarily of academic fisheries biologists and led by William M. Lewis Jr. of the University of Colorado, Boulder. The panel was given a deadline of 3 months to turn in a preliminary report, which was published in draft form in February 2002. The final version of the interim report appeared that September.

The NRC panel concluded that most of the recommendations in the USFWS and NMFS biological opinions were scientifically justified. But it balked at the two most important ones: the minimum water level for Upper Klamath Lake and the downstream flow for the coho. “A substantial data-collection and analytical effort by multiple agencies, tribes, and other parties has not shown a clear connection between water levels in Upper Klamath Lake and conditions that are adverse to the welfare of the suckers,” the report said. As a result, “there is presently no sound scientific basis” for the mandated lake levels. As for the coho, it added that there was equally little justification for increased minimum water flows down the main stem of the Klamath River.

Opponents of the BiOps seized on the panel's conclusions. “A handful of U.S. Fish and Wildlife Service bureaucrats withheld desperately needed water from farmers in the Klamath Basin last summer. Now we find out that that decision was based on sloppy science and apparent guesswork. … This latest travesty in the enforcement of the Endangered Species Act should be one more nail in the coffin of that broken law,” said Representative James Hansen (R-UT), chair of the House Committee on Resources.

Congressional representatives and farmers weren't the only ones to draw on NRC's conclusions. In February 2002, the Bureau of Reclamation came out with a revised management plan for the Klamath Irrigation Project designed to govern operations for 10 years. The bureau recommended dropping summer water flows in the Klamath River below NMFS's recommended 1000 cubic feet per second to provide extra water for irrigation.

After studying the proposal, NMFS biologists concluded that the bureau's plan was inadequate to protect the coho and recommended bringing the flows back up. In the end, the agencies settled on dropping summertime flows to as little as half the minimum recommended in the 2001 BiOp. The bureau's plan would eventually restore the flows by establishing a “water bank” and taking land out of production: The bureau would “lease” water from Klamath Irrigation Project farmers, paying to keep it in the lake and streams rather than diverting it for irrigation. (Last month, the Bureau of Reclamation announced that it would spend $4 million this year on water leases, which it estimates will idle 5000 hectares of farmland during the summer.) But fish-friendly critics cried foul, pointing out that flows would drop immediately and that the plan would restore full minimum flows only after 9 years. As if on cue, 33,000 fish went belly-up in the lower Klamath River in September 2002—reportedly the largest fish kill in North American history. Most of the fish were Chinook, although some were endangered coho and oceangoing steelhead trout. According to a preliminary report from DFG, the fish died when low water levels forced spawners into cramped quarters, spreading naturally occurring infections. If true, it would seem to validate recommendations in NMFS's 2001 BiOp. But some have questioned DFG's objectivity, saying that its scientists blamed federal policy for the fish die-off before their study was even begun. The NRC panel is now reviewing the causes and will include its findings in its final report. After the fish kill, it was the environmentalists and fishers who went on the offensive against biologists for caving in to the Bureau of Reclamation. “The current federal water plan ignores science and instead relies on guesswork, wishful thinking, and voluntary measures,” said Glen Spain of the Pacific Coast Federation of Fishermen's Associations in Eugene, Oregon. “This is a water plan for killing fish. Why should farmers have all the water they need while coastal fishing-dependent communities and fishing families wind up with dead fish and dry rivers?” In late September 2002, a coalition of fisheries groups, environmental organizations, and Representative Mike Thompson (D-CA) filed suit, seeking an injunction against the NMFS BiOp that accepted reduced flows for 9 years and asking a judge to require higher summer water flows in the lower Klamath. A hearing is scheduled for 29 April. Bureau of Reclamation spokesperson Jeff McCracken defends his agency's handling of the water distribution plan. He says the bureau took its lead from the NRC report, which he calls “the best available science.” ## The best science? But many fisheries biologists in the Klamath Basin disagree. “To see [the NRC report] held up as some great science proving the ESA has run amok hit us the wrong way,” says Douglas Markle, a fisheries biologist at Oregon State University (OSU) in Corvallis, who co-wrote an extended critique of the NRC interim report in the March 2003 issue of Fisheries. For Upper Klamath Lake, the NRC panel found that poor water quality conditions that are harmful to fish do not coincide with years with low water levels. And the best years for young fish aren't clearly associated with high water levels. As a result, panel scientists concluded that there was no clear link between lake levels and the health of fish. For the Klamath River, they found that water added in dry years to bolster flows was small “and probably insignificant.” It could even make matters worse, because sun-warmed lake water might harm cold-water coho. Few biologists claim that there is an ironclad case that higher water levels in the lake and river will always help the fish. The 2001 USFWS BiOp, they point out, didn't argue that low lake levels are always associated with poor water quality, rather that higher lake levels carry numerous benefits to water quality and fish habitat. But the NRC panel, critics charge, didn't look beyond the lack of a clear link between water levels and fish health for indications that—all other factors being equal—the fish would do better with higher water levels. The panel “pursued an unnecessarily simple view of a complex ecosystem which, combined with several clear errors in their assessment of existing data, led them to a flawed conclusion,” wrote the Klamath Tribes' Dunsmoor and Jacob Kann, an aquatic ecologist at Aquatic Ecosystem Sciences in Ashland, Oregon, in another detailed critique sent to the NRC committee last year. Markle and others contend that numerous examples show the importance of taking a more complex view of the Klamath ecosystem. In the summers of 1995, 1996, and 1997, for instance, lake levels were intermediate or high compared with the rest of the 1990s; nevertheless, the 3 years saw successive fish kills. Algae in the lake experienced massive blooms and crashes, causing swings in pH and depleting oxygen, which can kill fish or make them more susceptible to infection. The NRC panel noted that “… lake level fails to show any quantifiable association with extremes of dissolved oxygen or pH.” But Dunsmoor and Kann argue that the panel overlooked another important factor: wind. It aerates and mixes the water, driving much of the algae below the level of light penetration and reducing their growth rate. Without the wind, as in the relatively calm summers of 1995 to 1997, the water stagnates, the algae explode, and water quality plummets. Wind of course can't be predicted. But higher water levels, they argue, can soften the blow by diluting nutrients to slow the algae bloom. In 1991 something of the reverse happened. The population of young suckers boomed, despite a lake level at its lowest since 1950. But Markle and Cooperman note that in June, one of the most important months for the emergent fish fry, the lake level was fairly high and dropped considerably only in October. As well, the OSU authors point out that 1991 was a cool, windy year, which forestalled the algae bloom and led to relatively good water quality. That information was ignored by the NRC panel, say Dunsmoor and Kann. In a rebuttal to the Markle and Cooperman article in the same issue of Fisheries, NRC panel chair Lewis fires back that “variations of weather conditions from year to year do seem to underlie variations in mass mortality of adult suckers from year to year, but there is no hint of any connection with water level.” And even though the notion that a higher water level could benefit the lake fish is a plausible theory and potential justification for keeping more water in the lake, he points out that it's a decision based not on scientific evidence but on professional judgment. The panel, he noted, “unanimously reached several strong conclusions because it was confident that the evidence presented to it supported those conclusions.” The scientific brawling isn't limited to the suckers. Critics charge the committee with oversimplifying matters with regard to river-based coho as well. The NMFS 2001 BiOp recommended releasing additional water from Upper Klamath Lake in the summer months, in part to increase the amount of habitat available to the juvenile coho before they migrate to the ocean. But the NRC panel concluded that additional water sent down the main stem of the Klamath River would likely have little impact on the tributaries where the coho linger. The panel concluded that coho use the main stem of the river chiefly to migrate to and from the ocean. But DFG's Rode points out that some of the fish feed in the main stem for part of the day and return to the cooler tributaries while building strength for their migration. Excess water—and the habitat improvement it would bring—is of critical importance to the young fish, he says. Proponents of low river flows have used the NRC report to “try to use science to justify the low flows,” Rode says. Lewis responds that the NRC panel's job was simply to see whether flow rates were justified by documented science. “That doesn't preclude the agencies from recommending [higher flow rates] anyway,” he says. ## Asking too much? Perhaps the most fundamental objection to the NRC's interim report is that the panel was asked the wrong question. The committee's charge, settled upon after negotiations with its sponsors, the Departments of Interior and Commerce, was to determine whether there was scientific proof that the policies embraced by USFWS and NMFS would accomplish what they set out to do. But critics note that this isn't the standard set for the wildlife agencies. In carrying out the ESA, USFWS and NMFS are charged with using the best available science to protect the species. Where the science is questionable, they are supposed to err on the side of conservation to protect species already on the brink. In some cases, that means taking steps to preserve habitat or living conditions even if the steps haven't been proven to work. Farmers, Markle points out, can tell you precisely how they will use a given volume of water and its value for their crops. “But with fish data, there is no certainty of the benefit you get from an added acre-foot of water or the cost of removing it,” he says. By asking for scientific proof that those actions would benefit fish, the NRC panel was setting the bar too high, he says. The trouble, adds Dunsmoor, is that the NRC report has put pressure on agencies to mandate only those recovery actions that are scientifically well established. In essence, Dunsmoor says, that puts the burden of proof on the conservation agency to show that particular management actions will help the fish: “It's a paradigm shift. It would reset how these decisions are made.” To prove that a particular action will have consequences, agencies would be forced to wait until they see that harm is done by not carrying out the action, which some would say is exactly what happened with the fish kill in the lower Klamath. “All conservation goes out the window if you have to wait for fish to die to say there is an effect,” Dunsmoor says. In his Fisheries rebuttal, Lewis readily agrees that the NRC panel's purpose was different from that of the agencies. But he writes, “Where the economic stakes are high … it is useful for all parties to recognize which components of Biological Opinions are indeed scientifically solid and which are to varying degrees based on informed speculation.” Raising the bar on how much proof wildlife agencies must have before they take action would doom long-term restoration efforts, says Mark Buettner, a fisheries biologist at the Bureau of Reclamation. One example, he says, is the ongoing effort to prevent phosphorus-rich farm and ranch runoff from reaching Upper Klamath Lake. Even if vast strides are made in reducing the amount of phosphorus that reaches the lake, such an effort may not have an impact on water quality for years or decades to come. That's because the lake's muddy bottom is chock-full of phosphorus and other nutrients that leach back into the water, Buettner points out. If wildlife agencies were required to show a rapid effect of their actions, reducing nutrient inflows—which virtually all fisheries experts agree is important—would never get off the ground. It may take decades of research to demonstrate any link between lake level and the health of the endangered fish. But many researchers worry that public reaction to the NRC interim report could undermine the research efforts needed to unravel the basin's complex ecology. “It has led many nonscientists to the conclusion that the question [of proper management] has been answered,” Larson says. “It's frustrating,” says one biologist who asked not to be identified. “What we do has instantly become junk science.” The NRC panel may soften its tone in the final report due out this summer. But many on biology's front lines here fear that the damage has already been done. 10. SCIENTIFIC MISCONDUCT # The Multiple Repercussions of a Fudged Grant Application 1. David Malakoff Connecticut microbiologist admits to falsifying data on grant proposal but says the university's conduct is no less reprehensible When the University of Connecticut hired Justin Radolf 4 years ago to lead its new Center for Microbial Pathogenesis, officials gave him a$175,000 annual salary and a spanking new laboratory and touted his “industry, intellect, and enthusiasm.” Radolf used those talents to build up the center, which included helping to snare a $2.5 million congressional earmark to develop a vaccine against tick-borne diseases. But that promising partnership has since fallen apart in a welter of recriminations. Last week the federal government announced in the Federal Register that Radolf had falsified preliminary data in two grant applications. Although Radolf has admitted the misconduct, calling it “inexcusable,” he has sued university officials for removing him as head of the research center and alleges that a colleague “misappropriated” his data in garnering the earmark. The university won't comment on the charges in Radolf's suit, which was amended last week in a Connecticut federal court. The public falling out is a painful lesson in how an “unacceptable but scientifically minor fabrication,” as one researcher calls it, can have multiple repercussions. Says entomologist Thomas Mather of the University of Rhode Island, Kingston, “I knew there were hard feelings [at Radolf's center], but they seemed to have been settled internally. … Then it went way beyond [the university].” Radolf is an expert on spirochete bacteria, the class of organisms that cause Lyme disease and syphilis. He moved to the University of Connecticut's Health Center in Farmington in 1999 after 12 years at the University of Texas Southwestern Medical Center in Dallas. Among his collaborators was Stephen Wikel, a tick specialist at Oklahoma State University, Stillwater. The two were studying tick saliva proteins in hopes of developing a vaccine that would prevent ticks from feeding or passing disease-causing spirochetes to their hosts. In January 2000, Radolf and Wikel—whom Radolf was in the process of recruiting to Connecticut—were preparing to submit funding proposals to the U.S. Department of Agriculture and a Connecticut agency. Shortly before submitting the proposals, Radolf told Science, he altered a figure showing preliminary gene-expression data and tweaked some other language so that it better supported the proposal. “[M]y actions would have misled the reviewers … into thinking that we were closer to the development of an anti-tick vaccine that we actually were,” Radolf wrote in an admission released last week by the Department of Health and Human Services' Office of Research Integrity (ORI). ORI has ordered the university to closely oversee Radolf's work for the next 5 years. University officials first heard about Radolf's actions in February 2000 from Marianne Hebenstreit, an administrative assistant in Radolf's lab, according to university documents provided to Science. Once a university investigation was under way, Radolf admitted the deception and asked Wikel—who was the principal investigator on the proposals—to withdraw the applications. In August 2001, the university formally reprimanded Radolf and put him on 3 years' probation, according to Radolf's suit. Radolf remained head of his center, and he says health center officials assured him that “we will get through this.” Sometime later, the federal government began its own investigation after a third party informed ORI about the university's findings, according to a September 2002 story in The Hartford Courant. During the university's inquiry, Radolf says, he and Wikel were also helping school officials seek Department of Defense (DOD) funding for tick vaccine research. Radolf says that health center officials became less supportive after they believed their efforts, aided by Representative Nancy Johnson (R-CT), had borne fruit. In January 2002, according to Radolf's suit, health center officials urged Radolf to temporarily step down as head of the center. He also claims that university officials told him he would not be a lead investigator on the DOD grant but could participate in the research. Shortly after Radolf agreed to step aside, however, he says medical school dean Peter Deckers informed him that his removal as head of the center was permanent. A few days later, Johnson announced the$2.5 million DOD grant. Standing beside her was Wikel, who was subsequently named head of the center. “The university turned on me,” says Radolf.

In his lawsuit, Radolf protests Wikel's appointment as head of the center. He also says that Wikel didn't have his permission to use his data in winning the DOD funds. A January 2003 report from a university review panel, obtained by Science, appears to support Radolf's claim. “I've been frozen out” of the research, he adds. Wikel declined to comment on Radolf's allegations.

Radolf is also asking the court to block a university investigation into a claim that he filed timesheets improperly. He says university administrators involved in the case have attempted to protect themselves by filing their allegations under “whistleblower” rules intended for less senior workers. A federal judge could rule as early as this month on whether that investigation will move forward, but the rest of the legal battle could drag on for years.

In the meantime, Radolf says he has sufficient funding to keep his 10-member lab busy. Other scientists in the infant field of tick vaccine research say the episode should not put a kink in their research, because the altered material was never published. Indeed, university investigators concluded that Radolf's doctored data “were ultimately shown to be true” in later experiments.

“That's the irony,” says Radolf. “This wouldn't have happened if I'd waited a little longer [for the data].”

11. BRUCE BUDOWLE PROFILE

# FBI's Top Scientist Takes the Lead in Forensic Biology

1. Martin Enserink

He's a veteran of the wars over human genetic fingerprinting. Now, the geneticist is focusing on microbial DNA and threats of bioterrorism

QUANTICO, VIRGINIA—It's eerily quiet in the FBI's brand-new forensic science laboratory. A few workers are putting the final touches on the building, which smells of fresh paint. With their blue-gray countertops and wood-veneer cabinets, the rooms could be in an Ikea catalog. But they're totally empty; not a microscope or pipette in sight. This is where Bruce Budowle, the FBI's premier biological scientist, is setting up his new home. Thin, red-haired, and wearing an anthrax-themed tie that his daughter ordered on the Internet, Budowle, 49, was unpacking boxes in his office on a recent Wednesday morning as he explained what lies ahead for him and his colleagues.

In a few months, the place should be bustling. The FBI hopes the new lab, besides doing everyday forensic science work, will dramatically improve its ability to investigate bioterrorist crimes—an area that was barely on its radar screen until the anthrax attacks of October 2001. In the panicky months that followed, the government recognized that it had to beef up technical resources at the FBI, and the FBI turned to Budowle. Now he's assembling a broad coalition of academic scientists and other outsiders to help out.

If anyone has the expertise in genetic, forensic, and legal issues to bring the agency up to speed, colleagues say, it's Budowle, a geneticist who has worked for the FBI for 20 years. He helped put DNA fingerprinting of crime suspects on a solid scientific footing and get it accepted during a divisive battle in the early 1990s. “I think he is going to be pivotal in organizing this,” says Randall Murch, a former FBI colleague who is now at the Institute for Defense Analyses in Alexandria, Virginia. Others praise the way he has brought scientific rigor into an agency that has not traditionally embraced it. “He's really a scientist, and a good one—not a cop,” says Paul Keim, a microbial geneticist at Northern Arizona University in Flagstaff.

Budowle is happy with the new lab, a far cry from the current, cramped space at the FBI headquarters in downtown Washington, D.C. It is located in a new, 50,000-square-meter building, close to his home, at a vast Marine Corps base that also houses the FBI Academy. “Hogan's alley,” a mockup town for practicing real-life crime-fighting, is nearby, as is a driving range for high-speed car chases.

## Beyond street smarts

In the battle against bioterrorism, one thing is certain: The FBI will need a lot more than street-smart detectives and dexterous drivers. After the anthrax letters killed five people and sickened 17 others, the bureau came under withering attacks from Congress, the media, and bioterrorism experts for its alleged clumsiness. Clueless field agents asked all the wrong questions, critics said, and the bureau even gave Iowa State University in Ames the go-ahead to destroy a collection of historic anthrax samples that may have contained valuable information. Most important, the oddly named “Amerithrax” investigation has not identified a suspect, although the bureau did focus media attention on what Attorney General John Ashcroft called a “person of interest,” microbiologist Steven Hatfill, who has not been charged with wrongdoing.

Budowle says he cannot discuss specifics of the case, but he shrugs off most of the criticism. Although few people understand what really goes on in such an investigation, the FBI will always be held to a high standard, he says, adding that it should be. But he concedes that the bureau has much work to do if it wants to deter future bioterrorists—or nab them. Until recently, bioterrorism ranked relatively low on the agenda, he says. The Hazardous Materials Response Unit was started in 1996, but until the anthrax mailings, every suspected bioterror attack turned out to be a hoax or a false alarm.

As Budowle points out, however, the anthrax investigation did show the potential power of microbiological forensics. Keim, for example, had spent years developing a genetic fingerprinting system for Bacillus anthracis, which helped establish that all anthrax letters contained the same strain, called Ames. That helped focus the search on a limited number of labs and people.

But much more remains to be done, a panel assembled by the American Academy of Microbiology concluded in a report issued in February (Science, 21 February, p. 1164). To make sure microbial DNA evidence holds up in court, for instance, a “chain of custody” will have to be preserved from every incident, meaning that everyone from first responders to lab technicians will have to follow standardized protocols. To say with confidence that microbes found at a bioterror crime scene are identical to those found in, say, a suspect's garage, researchers need to know much more about microbial genetic diversity—which means sequencing many more genomes. “We have to prepare for the fact that every piece of evidence is going to be challenged,” Budowle says. “That's the way the system works.”

## DNA war survivor

Budowle should know. When human DNA fingerprinting made its debut in the late 1980s, critics challenged both its scientific basis and the quality control in fingerprinting labs. Sloppiness could put innocent people behind bars, they said. Budowle, who helped pioneer the technique, came to its defense. The “DNA wars” became extremely heated and at times personal; Budowle says critics angered him when they suggested that he was selling out his scientific integrity to the prosecutors. “I'm not just trying to get convictions,” he says.

As lab quality standards were enforced and studies yielded a clearer understanding of the odds of having a DNA match occur by chance, the controversy waned. In 1994, Budowle and one prominent critic, Eric Lander, who now heads the Massachusetts Institute of Technology's genome center, concluded in a joint Nature paper that “the scientific issues have all been resolved.” Although some declared the peace premature—colleagues even chastised Budowle for teaming up with the enemy—the paper helped put the issue to rest. Today, DNA evidence is seldom contested. During the move into his new digs, Budowle threw out heaps of documents that he had hoarded for the DNA wars.

Scientifically, microbial DNA evidence will be different. For one, because bacteria reproduce asexually, they don't produce unique genomes, as humans do, that can be used to tell them apart. Moreover, there are dozens and dozens of potential bioterrorist weapons. (The government's current lists don't even include food-borne agents such as Escherichia coli or Salmonella, which Budowle thinks are actually just as likely to be used.)

But the past 10 to 15 years offer valuable lessons, he says. To help him sort out the problems with human DNA, Budowle put together a scientific working group of outside experts in 1988, then a radical departure for the traditionally introverted bureau. The approach has now taken hold. “The best way to get feedback is to expose yourself,” he says. “It works much better than a dictatorial approach.” For microbial forensics, a new 30-member panel includes Ronald Atlas, president of the American Society for Microbiology; Timothy Read of The Institute for Genomic Research in Rockville, Maryland; Keim; and other scientists at the CIA, the military, the national labs, the U.S. Department of Agriculture, and public health departments. Its first report is expected shortly.

Ideally, Budowle says he would like to do most of the forensic work in-house, but the threats are so numerous that this would be impossible. Even if the FBI acquired the expertise to fingerprint dozens of different pathogens, the next attack could be with an unheard-of, exotic virus. That's why he's currently building a network of labs that can help.

Budowle and his panel are also thinking of creating a vast library of different strains of many pathogenic bugs, so that investigators could quickly compare a microbe used in an attack to every known isolate. But such a collection needn't sit in one place, Budowle says; each lab could keep its own strain collections—Keim, for instance, has many hundreds of anthrax isolates—as long as the FBI could get access. Budowle says many scientists are interested in forensic science and are willing to help.

It doesn't hurt that the general public is equally fascinated, he says. The popularity of CSI: Crime Scene Investigation, a slick TV series set in a Las Vegas crime lab, indicates that people care about this work. Budowle says his wife enjoys the show. But he says he can't make sense of it himself: Real forensic science is not that glamorous—and it's a lot harder.

12. NEUROIMAGING

# Still Debated, Brain Image Archives Are Catching On

1. Marcia Barinaga

Basic and clinical researchers are finding ways to overcome the challenges to storing, analyzing, and sharing large collections of data

Techniques for capturing images of the human brain in action fill the pages of neuroscience journals with colorful pictures. And they fill researchers' labs with huge quantities of data required to generate the images. These data are too voluminous to include in publications, and some researchers say their absence makes it impossible for a reader to fully understand a study or repeat it. The countless variations in data processing make it difficult for researchers to compare their results directly with those of other labs or even to keep track of their own data. “Every lab has had the experience that they can't even reanalyze their own data, because … they don't have good records,” says cognitive neuroscientist Mark D'Esposito, who directs the brain-imaging center at the University of California, Berkeley.

There is a good side to this information tangle, though: It has prompted some neuroscientists to begin to store, organize, and share their brain-image data in easily accessible archives. Some of the archives are up and running, but even as they expand, researchers continue to debate the fundamentals: Should raw data be made available for others to exploit? Is it meaningful to compare findings from differently designed studies? And are the benefits of such databases worth the effort it takes to organize and deposit the data?

Proponents of standardized databases say they will make it easier to share data from lab to lab, something that is becoming more important as brain imagers take on questions that require comparing large numbers of brain scans. And free access to raw imaging data will enable researchers to use existing data to address new research questions: “Data sets are expensive to produce and may have utility far beyond that originally envisioned by the investigators who collected the data,” says Randy Buckner of Washington University in St. Louis, Missouri.

Databases can also serve as supplements to published papers. Imaging experiments generate three-dimensional representations of activity throughout the brain, but authors usually publish just a few cross-sectional brain views of the areas most relevant to their work. Access to the complete set of images generated by an experiment would allow those interested to select the views they wish to see and even change parameters that affect how the data are displayed.

A well-designed and well-stocked database, says Washington University neuroscientist David Van Essen, ideally could transform a project like the one that just took him the better part of a week—pulling together published information on the location of the brain's language centers—to a mere few minutes' work. With the right queries and a few mouse clicks, he says, “you could see not just 10 or 20, but 100 studies, all stacked up on each other, see the patterns, and tie into the methodological differences among the studies.”

Some neuroscientists expect the power of databases to catapult their field to new levels of research and discovery, much as GenBank revolutionized the research fields that depend on DNA sequences. But others doubt that any brain-image database could be the neurological equivalent of GenBank. A DNA sequence “is a physical property that can be tested and confirmed,” says Bradley Postle of the University of Wisconsin, Madison. “Cognitive neuroscience is different in that we are testing hypotheses about abstract constructs.” It is much harder to catalog the results from such experiments in a way that will make them useful to others, he argues.

There are sociological hurdles as well. Some researchers are unwilling to relinquish control of their data. “Why should I give my data to other people to analyze?” says Peter Fox of the University of Texas Health Science Center in San Antonio.

Fox's solution has been to pour his energy into a database that describes but does not contain data, what he refers to as a “low-impact” alternative to a full data archive. Others are working on different variations, from private specialized archives to help researchers handle data sets in their own labs to a database that serves as a brain atlas to provide a better standard to which brains can be compared.

In the next few months, several of these projects will unveil their newest Web sites, software, and demonstrations and invite neuroscientists to give them a test drive. If those tests prove a database to be “a powerful scientific tool,” predicts Washington University neuroscientist Marcus Raichle, researchers are likely to flock to it and not look back.

## The Dartmouth experience

The notion of a public repository for brain-image data gained widespread attention 3 years ago when Michael Gazzaniga of Dartmouth College in Hanover, New Hampshire, editor of the Journal of Cognitive Neuroscience (JOCN), caused a stir by announcing that the authors of brain-imaging papers published in the journal would be required to deposit their raw data sets into the newly formed public fMRI Data Center at Dartmouth (Science, 1 September 2000, p. 1458; fMRI stands for functional magnetic resonance imaging).

Gazzaniga says that the ensuing controversy was just part of the “growing pains” that many fields experience at the first idea of open access to data. Raichle, who sits on the board of the Dartmouth center, agrees: “In crystallography and genetics, it wasn't all good feelings at the beginning,” he says. “Now it's a way of life.”

The center at Dartmouth is working toward making it a way of life in neuroscience also. It holds supporting data for brain-imaging papers—ranging from the raw output of the brain scanner to the full set of processed images of the brain—and distributes the data sets on CD to researchers who request them. Gazzaniga and Raichle envision a future in which virtually all researchers who publish brain-imaging papers in any journal would deposit their data there. They still have a long way to go: No other journals have adopted a data-deposition requirement. But Postle notes that, as “the flagship journal” in the brain-imaging field, JOCN is well positioned “to effect fundamental change.”

Two years after JOCN announced its policy, the number of studies available from the Dartmouth center is approaching 50. Most come from JOCN, Gazzaniga says, with a handful from other journals. “I haven't seen a slowdown of papers arriving [at JOCN] because of this policy, which is what people had originally predicted,” says D'Esposito, who edits JOCN's imaging papers, adding that no author has refused to deposit data. Some labs, however, have chosen not to publish in the journal.

To illustrate the center's capabilities, JOCN will publish a paper this year from D'Esposito's lab, making the complete set of whole-brain images accessible on the center's Web site (http://www.fmridc.org/). Visitors to the site will be able to manipulate the images, choose the brain section they would like to see, or change the statistical thresholds and observe the effect. Operations director Jack Van Horn says the center hopes to make such access available for all of its data sets. To promote the mining of data for new results, JOCN held a contest last year for the best original research paper using archived data sets. The winner, published in August 2002, was an investigation of the neural basis of consciousness by Dan Lloyd, a philosopher at Trinity College in Hartford, Connecticut.

The contest pointed out some limitations of the database, says Postle, who generated one of the data sets Lloyd used. Postle was invited to comment on Lloyd's paper after it was accepted for publication; he found that Lloyd had made an incorrect assumption about how Postle and his colleagues had processed the data, which affected the results. “If I had seen the manuscript or had had a discussion with him ahead of time, I could have called his attention to that,” Postle says. He suggests that journals ensure that the researcher who generated the primary data check for errors in secondary papers, because “nobody else has the same insider knowledge about how those data were created.”

Postle and others have noted that it is currently difficult and time-consuming to deposit data in the center. The center is addressing this problem; it is about to release a free database software package that will help labs manage their data and facilitate depositing the data in a database. D'Esposito says the software could be the data center's most important contribution so far if it creates an “industry standard” for storing brain-imaging data.

## A low-impact alternative

Not everyone has been won over to the notion of unfettered data sharing, and San Antonio's Fox is one researcher who won't be publishing in JOCN if it means turning over his data. “There is a lot of intellectual property there,” he says. “I personally have a data set I acquired 9 years ago, and I'm writing the fourth paper from it. Why—since I designed the experiment, and I had a pretty good idea I would be doing these analyses—would I give that to somebody else to do first, just because they are not as busy as I am?”

Fox's alternative, a repository called BrainMap, contains no raw data or brain images, but it summarizes papers using a code to describe each experiment and the brain activations found. “We didn't want to be responsible for distributing raw data,” says Fox. “A much simpler strategy that has stood the test of time is that if people want your data, they collaborate with you.”

BrainMap uses “much more searchable, well-structured descriptors” than those found in the MEDLINE or PubMed databases of published papers, Fox says, and it gives a summary of the paper's results. “The entire database is pretty much interpretation free,” he adds. “We don't want your theory, your bias, or your discussion. If people want that, that is a great reason to read the paper.”

Fox says 25 labs have agreed to deposit all their papers in BrainMap, with his team doing the first round of coding. He admits that the laborious process may deter some people from entering papers into the system, but he notes that BrainMap contains useful tools for analyzing and comparing papers and hopes that those tools will entice researchers to encode sets of papers they wish to analyze.

## Made to fit

Even if they don't intend to give results out publicly, some researchers find that they need customized databases to organize and analyze their own research. This is especially true for clinical teams that gather brain images from many patients. Neuroinformatics specialist Edward Herskovits and his colleagues at the University of Pennsylvania's Brain Image Database (BRAID) project are creating specialized databases for half a dozen or so clinical teams whose research ranges from large-scale studies of brain changes during aging to the psychiatric consequences of traumatic brain injury and stroke. “Instead of a lab notebook, you have a huge database that stores images, clinical data, perhaps predisposing factors if you are talking about a certain disease, clinical findings, and outcomes,” says Herskovits.

The BRAID group collaborated with child psychiatrist Joan Gerring of Johns Hopkins University School of Medicine in Baltimore and uncovered an association between injury to a particular brain area in 76 children and subsequent development of attention deficit hyperactivity disorder. The finding, says Herskovits, “was visually striking. You could see a concentration of lesions in right basal ganglia and right putamen.” The researchers then used the database to statistically confirm the visual finding.

Herskovits notes one downside of BRAID's approach: Its custom databases are not suited to sharing or comparison across studies, he says, because “everybody has their own preferences” for how their database is constructed.

## A new standard for brain structure

A central challenge for all brain imagers is how to pinpoint on a map of the brain the neural activity they detect inside their subject's skulls. Human brains are the same basic shape, with the same major folds and functional areas such as those devoted to language or vision. But the exact location and shape of the folds vary from brain to brain, and from one period of life to another—and so do the boundaries of the functional areas relative to the folds.

To deal with this variation, researchers use brain atlases, which are 3D maps of an “average” brain, derived from one or a few individuals. By finding the location of key landmarks in the brain they are studying, researchers can “warp” their subject's brain to match the average brain, and most structures will line up. Then the researchers describe the area of activation using standardized 3D coordinates.

But the atlases are not good enough to deal with the increasingly fine resolution of functional image data, says neurologist John Mazziotta of the University of California, Los Angeles. “We can warp any brain to look like any other brain, structurally, but that may not [give us] better insight into how the brain is organized,” he says, because structural landmarks don't bear a consistent relation to functional landmarks. That means that researchers may sometimes assign brain activation to the wrong functional areas.

Mazziotta and colleagues from eight laboratories around the world set out to make a more realistic and flexible map, in a project called the Probabilistic Atlas of the Human Brain. They have cataloged high-resolution magnetic resonance images of the brains of 7000 healthy individuals aged 18 to 90. In a subset of the subjects, the team also has done functional imaging, using tests designed to activate areas devoted to specific brain functions to correlate them with brain anatomy.

In addition to brain images, the atlas contains detailed information for each subject, including race, ethnicity, diet, education, occupation, handedness, and the results of a battery of behavioral, neurological, and psychiatric tests. The project also has obtained DNA from most subjects. What's more, the team has done postmortem analyses on some of the brains, recording the kinds of neurotransmitters and neurons found in different brain areas.

The atlas provides a wealth of information that can be used by researchers to test hypotheses, says Mazziotta, such as whether a certain variation in brain anatomy corresponds to a particular genotype or other trait. It will also enable researchers to order up a custom atlas for their brain-mapping experiments. “You can say, ‘I need some subjects between these ages, left-handed, Asian, with 2 years of college, who smoke cigarettes,’ and we will say, ‘We have 18 of those,’” Mazziotta says.

The database has been 9 years in the making, and the research community outside the participating labs hasn't gotten a peek yet at how it works. But in July, the archive will be opened to the public, and researchers will be invited not only to use the data, but to add data sets of their own.

Mazziotta envisions the Probabilistic Atlas eventually being tied into other repositories that hold other data, including electrophysiological records and the types of neurotransmitters, receptors, and other proteins found in certain neural cell types. Such links could form what Yale University neuroscientist Gordon Shepherd calls “a grand overriding database, with representation of data at all levels from cell circuits to brain scans.”

Whatever specific types of databases the future holds, Washington University's Van Essen suggests that “neuroscience will look back with a chuckle” to the days before databases “and say, ‘My goodness, how did those guys get anything done?’”

13. # Play-by-Play Imaging Rewrites Cells' Rules

1. Mary Beckman*
1. Mary Beckman is a writer in southeast Idaho.

Watching cells as they grow within an organism is changing the way devopmental biologists think about life's earliest events

Finding a trove of videotapes of your great-grandparents might cause you to revise long-accepted family history based on fading sepia pictures. Similarly, methods that create movielike images of key events as cells grow and connect into tissues are causing biologists to rethink how embryos develop. Three recent efforts using improved imaging techniques are overturning old models of some developmental fundamentals: how tissues form in chicken embryos, how neurons find one another in a zebrafish, and how neurons compete for purchase on muscles in newborn mice.

Static images—until now, the source of most data in developmental biology—give an incomplete view. Neurobiologist Jeff Lichtman of Washington University in St. Louis, Missouri, and developmental biologist Scott Fraser of the California Institute of Technology (Caltech) in Pasadena liken it to learning about a sport by studying a scrapbook. Imagine trying to determine the rules for American football, they say, by examining 1000 snapshots taken at different times during 1000 games. The importance of the football would be evident, but not why players are obsessed with its position. And it would be tough to make any sense of the halftime marching band. The rules of the game would probably remain utterly obscure. Getting away from this approach, teams led by Fraser, Stanford neurobiologist Stephen Smith, and Lichtman have worked out equivalents of instant replay to deduce the rules that guide cells as they construct an organism.

Lichtman says that imaging allows scientists to take advantage of the world's fastest computer processors: their own eyes and brains. Humans can take in lots of visual information at once and extract patterns from it; complex images and movies provide such information. And understanding whole patterns rather than summing parts is becoming important to get the details of development right. Developing cells may respond to hundreds of chemicals at once, a complexity that's lost by reducing the process to discrete steps in test tubes. Fraser says, “You need to image the cells while they're doing their normal behaviors.”

Advances over the last few years, says Fraser, “make experiments that would have been hellishly difficult 3 to 5 years ago still difficult, but doable.” For example, the “green revolution” brought about by the relatively nontoxic, bright green fluorescent protein (GFP) used to mark cells and individual proteins has given way to a rainbow of radiant colors that allow researchers to follow multiple groups of cells or molecules at one time. Indeed, every aspect of imaging—dyes that outline living cells in tissues, proteins that glow from the center of cells, and microscopes—has improved by 25% to 50% in the past 5 years, Fraser estimates. And researchers are eager to tackle what used to be arduous experiments.

## Free spirits

New filming techniques have already overturned the three leading hypotheses about how primordial tissues form in chicken embryos (Science, 1 November 2002, p. 991). Decades of research established the basic outlines: Two-day-old chick embryos contain three layers of cells. Beginning at the head, cells in the middle layer pinch off in groups to form primitive structures called somites. The process repeats at 90-minute intervals, traveling from the head to the rear of the embryo. Each somite segment will develop into particular muscles, bones, or parts of the skin.

Researchers proposed three possible models to explain how somites pinch off at specific times and locations. One held that a molecular control marches through the inner layer of cells like a conductor, periodically waving its baton to mark a boundary and stimulate somite-edge cells to nip and tuck. Another model suggested that a master cell gives a group of synchronized disciples a command that causes them to turn into a somite at a prearranged time. The third model posited that some undiscovered program within cells causes them to link destinies and spontaneously form a somite when they grow up.

An assumption that runs through all these models is that the organization of somite cells must be orchestrated or prearranged. However, it now appears that cells are not locked into a rigid choreography but behave more like football players with a set of rules and options at every play.

Fraser and colleagues injected lipid dyes, which illuminate the whole cell, into groups of cells in three-layered embryos and watched as cells weaved and organized themselves. A confocal microscope scanned through the embryo's layers, and the images were reconstructed into a three-dimensional, time-lapse film.

After following the formation of two somites, the team identified regions where certain genes known to set up somite boundaries turned on. Correlating the action with gene expression allowed the team to “know what was going to happen in those tissues,” says developmental biologist Mark Cooper of the University of Washington, Seattle. “That's the advance. It eliminates the ambiguity” that comes from combining data from many embryos. In other words, the team watched the whole football game in one beer-and-potato chip sitting.

As expected, two genes lay down a somite pattern; the product of one gene outlines the top boundary, and the other defines the lower edge that crimps off. But individual cells move in and out of presomites at will. “The pattern is more organized than the cells themselves,” says Fraser. The once-popular idea that cells stay put in one somite is “an appealing mechanism because it's simple,” says Fraser. But the new finding “means that simple models are wrong, and we need to apply our molecular toolbox to this complicated world.”

Although it might seem as though developmental biologists are back at square one in trying to figure out just what does determine somite formation, Fraser says the results move the field forward in a different way. “It's especially cool when you're thinking about evolution,” he says. If the mechanisms by which development occur are “pristine and precise, imagine how many things have to change at once” to get the variation that can lead to new species. But with the “sloppy” mechanisms that cells apparently use to find their place in the organism, “you have a lot more machinery making the variation.”

## Grabby dendrites

You won't find the zebrafish Jell-O cooked up by Smith and fellow Stanford neurobiologist James Jontes at any picnics this summer. But immobilizing zebrafish embryos in thin, cool gelatin molds is giving the researchers access to something they have long wanted to study: developing nerves that have not been stunned by an anesthetic. After 3 years of improving a recipe calling for fluorescent neurons and two-photon microscopy, they're finding how neurons establish long-term relationships.

Scientists once assumed that neurons connect with one another when one of them extends an output conduit called an axon. The theory proposed that the axon feels around for another neuron's dendrite, a projection that receives signals. When the axon bumps into a dendrite, the two make a commitment to each other—or so the axon-centric story went. But Smith and Jontes's work illuminated the growing dendrite and thereby knocked that hypothesis out.

“They showed that the dendrite is an active partner in synapse development,” says neurobiologist Rachel Wong of Washington University; it reaches out and grabs an axon. The finding “split the field,” she says: Instead of focusing on the assertiveness of the axon, many people have turned to dendrites to study synapse formation.

Now Smith and his team have lit up the dendritic side of the developing synapse. They hooked up a protein commonly found at the receiving end of the synapse to GFP, allowing them to watch the budding relationship between the axon and dendrite where the two cells interact. “What we brought to the party”—aside from that Jell-O—“was labeling both sides of the synapse,” says Smith. Two-photon microscopy, which uses low-energy lasers to light up tissue bearing fluorescent markers only where the two lasers intersect, allows the team to zoom in three-dimensionally on the junctions without damaging tissue in the immobile but alert zebrafish, he says.

The researchers have been focusing on how an axon from the developing zebrafish's retina contacts dendrites in a visual area of the brain. The team found that bunches of optic nerve axons form a mass in the visual region—and wait. Then brain-based dendrites “grow into the mess and start synaptogenesis,” says Smith. Over the week in which the zebrafish optic system develops, “we see a whole lot of trial and error.” As in somite formation, cells have a certain amount of free will in how they shape their fates: “The wiring is far more discretionary” than previously thought,” he concludes.

## Dueling axons

Imbuing lab animals with a rainbow of fluorescent colors is allowing researchers to see how cells compete during development. Neurobiologists Joshua Sanes, Mark Walsh, and Lichtman of Washington University and colleagues ventured into the peripheral nervous system to watch neurons attach to muscles in neonatal mice engineered to sport glowing neurons.

The team took time-lapse microscopic pictures of surgically exposed junctions between neurons and breast muscle cells for an hour. Sewing the mice up and opening them again every 24 to 48 hours allowed the researchers to film the forming synapses over 1 to 2 weeks. They reported in 2001 that neurons send out many projections to muscle fibers, most of which eventually thin and retract. The work made the scientists wonder if thin axons were always the ones to shrivel.

The researchers created four strains of mice in which some neurons glowed in one of four different colors: red, green, yellow, or blue. Then they mixed and matched the animals. By mating a blue-neuroned mouse with a yellow one, for example, they could watch differently colored neurons competing for the muscle surface in the progeny. Each neuron, they reported in the January issue of Neuron, sends out multiple axons to many muscle units, although eventually only one neuron will remain connected to each unit. Time-lapse movies show a wave of brilliant, differently colored axons crashing into the muscle, then all filaments but one withering away.

Lichtman found it impossible to predict which neuron would win control of the muscle piece in the end. Some axons waned a bit but came back full force to overwhelm a rival. And axons thinned not just before receding but also when losing the fight. The monumental amount of activity surprised the researchers. “We had no sense of the dynamism,” Lichtman says, before seeing the neurons in action.

“It is very elegant what they are doing,” says neurobiologist Susana Cohen-Cory of the University of California, Irvine. Before, scientists thought the muscles were the ones that picked the winning axon—by removing contacts to the competitors, she says. “But now, we see that the axons are very active. It's interesting how [ideas] change once you have more tools.”

With the current state of imaging, Caltech's Fraser says scientists can ask a lot of questions about developmental processes, such as how cells create patterns in embryos or how scaffolding proteins shape and reshape cells. He and others say the imaging field is on its way to more precise work, like lighting up individual molecules (see p. 80). Lichtman adds that imaging often provides unexpected data: “Rarely a week goes by when we don't see something new.” Fans should stay tuned for the complete story while developmental biologists shoot more film of the game life is playing.

14. # Spying on the Brain, One Neuron at a Time

1. Greg Miller

A clever combination of genetics and imaging may unravel the secret of how mature brains learn

Luckily for us, the cliché about old dogs not learning new tricks isn't true. Our minds may be more nimble when we're young, but we can still learn from our experiences well into our golden years.

However, researchers know relatively little about how the adult brain learns. Part of the mystery is that the mature brain lacks much of the young brain's ability to tinker with its own wiring. At birth, the brain is a thicket of branching extensions that connect neurons and allow them to talk to one another. Early in life, experience works on this tangle like a bonsai master, pruning away connections that don't play a part in working neural circuits and reinforcing those that do (see p. 76). By adulthood, this process is finished, and much of the capacity for major rewiring is lost.

Still, grown-ups manage to learn—which means they must have some way to adjust their neural circuits. The mechanisms that enable the adult brain to learn may be subtler versions of those that wire the developing brain. For example, learning could occur by adding or removing synapses—the connections between neurons—without changing the overall shape of the neural branches themselves. Alternatively, learning may depend entirely upon modifying the synapses that already exist, with no structural changes at all.

Although researchers have gained valuable clues about the changes that accompany learning from experiments with cultured neurons and slices of brain tissue kept alive in a petri dish, these studies have built-in limitations for understanding what goes on under more normal circumstances. And until recently, few tools for visualizing neurons in living animals existed.

That is changing. In December 2002, two teams reported in Nature that they had repeatedly taken detailed images of individual neurons in the cerebral cortex of mice over a period of weeks or even months, a remarkable feat made possible by a clever combination of tools from genetic engineering and optical imaging. Although the new studies come to different conclusions about whether structural changes are needed for learning in the adult brain, researchers say they mark the beginning of a new era of studying neural plasticity in exquisite detail in living animals.

Both of these new studies took advantage of unusual mice with glowing brain cells developed by Joshua Sanes and colleagues at Washington University in St. Louis, Missouri. The animals produce a fluorescent protein in some cerebral cortex neurons, making visual tracking possible. The two research teams—one led by Karel Svoboda of Cold Spring Harbor Laboratory in New York and the other by Wenbiao Gan of New York University—focused on the cortex using a technique called two-photon microscopy.

This increasingly popular method sends laser pulses of infrared light into the area of interest. Infrared photons have much less energy than the photons of visible light used in conventional fluorescent microscopy. As a result, it takes the combined impact of two photons to set a fluorescent marker molecule aglow. The lower energy photons cause far less damage to living tissue. And because infrared light isn't scattered and absorbed by the tissue as much as visible light is, it penetrates farther and enables researchers to peer much deeper. In fact, Gan's group imaged cortical neurons without completely removing the skull, leaving a paper-thin layer intact.

Svoboda and colleagues homed in on an area that responds to stimulation of the whiskers. This region is named the barrel cortex for the cylindrical clusters of neurons that represent each whisker. The team took pictures of the fluorescent neurons in adult mice once a day for 8 to 10 days and then less frequently for up to a month. These images focused on the dendrites, the treelike structures where neurons receive inputs from other neurons. As expected, the shape of these dendritic arbors was very stable.

The team then zoomed in for a look at tiny protrusions, so-called dendritic spines, which contain synapses. About half of the 1224 spines examined were stable for at least 8 days, and some for up to a month. The rest, however, vanished within a few days, only to be replaced by a similar number of new spines. Indeed, the team found that about 20% of the spines were replaced from one day to the next. Time-lapse movies of the images show the tiny spines popping up and disappearing like lights on a badly wired Christmas tree.

Losing one out of every five synapses each day may sound like bad news. But re-assuringly, the newly formed replacement spines appeared to make good connections: When the researchers killed the mice and took a look at the spines with an electron microscope, they found that they made normal-looking synapses with neighboring neurons.

It's a “technically masterful” and provocative study, says Lawrence Katz, a neuroscientist at Duke University in Durham, North Carolina. “The prevailing idea in [adult] cortex would have been that things are quite stable,” he says. “If it's really true that synapses are turning over at the rate they suspect, it's remarkable.”

Svoboda's group suggests that synaptic turnover serves an important purpose. When the researchers trimmed every other whisker on some mice—a manipulation shown previously to change which neurons in the barrel cortex respond to which whiskers—the daily turnover rate jumped to about 30%, which suggests that spine turnover is important for this rewiring.

The finding adds to evidence that plasticity in the adult brain may require rerunning processes such as synapse formation that are common during development, says Kevin Fox, a neuroscientist at Cardiff University in the United Kingdom. “What you do in adults may be on a slower scale or lower key, but when you want to change the circuit, you basically repeat the program.”

However, the Gan team's report suggests that the brain is far more stable. The researchers used a similar approach to study neurons in the visual cortex of young and old mice. In 1-month-old juveniles, about 73% of dendritic spines stayed put over a 1-month period. In 4-month-old adults, spines were even more stable: 96% remained in place for a month or more.

The two groups' strikingly different conclusions—20% daily turnover vs. 4% monthly turnover—may be partly the result of differences in their methods, says Stephen Smith, a neuroscientist at Stanford University. For example, barrel cortex and visual cortex may not be equally malleable in adults. Because a mouse is probably more likely to have a whisker chewed off than to suffer damage to its retina, one might expect to see more plasticity in barrel cortex, Smith says.

Another key difference may be the way the two groups defined a dendritic spine. Svoboda's group was fairly liberal, classifying all protrusions from the dendrite as spines. In doing so, the researchers included thin, transient structures called filopodia that lack the fully formed “head” of mature spines and are thought to represent tentative or immature connections between neurons. Gan's group excluded filopodia from its analysis. It's not clear that one method is better than the other, Smith says: “It's not a no-brainer to decide how you're going to classify things for the purpose of this analysis.”

Counting only protrusions that form functional synapses might be a reasonable solution, he says, and one that should be feasible in the not-too-distant future with the help of new fluorescent markers (see sidebar). In the meantime, both groups are working on additional experiments that they hope will settle the issue of how stable adult neurons really are.

15. # A Wish List for Learning About Learning

1. Greg Miller

Given a magic lamp and an obliging genie, neuroscientists interested in the mechanisms of learning might ask for something like this: a way to visualize individual neurons—any ones they chose—in ultrafine detail, tracking minute changes in the cells' structure and electrical activity. Oh, and they'd want to do this in real time in unanesthetized animals—preferably humans.

Although they'll be kept waiting a while on some of these counts (the last one in particular), several techniques now in the pipeline could greatly expand scientists' ability to image neurons in living animals—particularly in rats and mice, whose brains are organized much like our own.

The degree to which learning in the mature brain depends on physically remodeling the contact points, or synapses, between neurons is a major question in neuroscience (see main text). Researchers have recently found anatomical signs of synapse formation in live mice, but it's difficult to tell whether these correspond to working synapses. Dyes that change color in response to voltage changes or changes in the concentration of calcium ions, both markers of neural activity, could help identify active synapses. Karel Svoboda of Cold Spring Harbor Laboratory in New York has been using such dyes in fruit flies. He and his team would like to use the dyes to follow up on their finding that neurons in the mouse cortex form new synapses with surprising ease. “We'd like to see synapses come online as they're made and see the synaptic strength,” he says. The biggest obstacle is that many of the currently used dyes are toxic or work best at cool temperatures, forming useless clumps at mouse body temperature.

Kinks like this should be worked out in the next few years, but other techniques appear to be farther off. Take selective cell imaging: Researchers have engineered mice that express a fluorescent protein in some neurons, but so far they've had little control over which cells light up. Being able to pick and choose which neurons glow would help them focus on how different components of a given neural circuit respond as an animal learns. Alas, says Svoboda, “there are fundamental genetics problems” in doing this in mice, such as identifying promoters that will drive expression of fluorescent proteins in particular types of cells.

Another item on the wish list is a method that provides a deeper look into the brain, says Wenbiao Gan of New York University. Like Svoboda's group, his team has been using two-photon microscopy, a powerful technique that can peer about half a millimeter into the brain. That's enough to see most of the sheetlike cortex. But researchers would like to be able to trace important changes in deep brain structures, including the hippocampus, a region known to play a crucial role in memory formation.

Help may be on the way. Mark Schnitzer of Lucent Technologies' Bell Laboratories in Murray Hill, New Jersey, has recently developed a fiber-optic “microendoscope” for two-photon imaging that can reach anywhere in a rodent's brain.

One of researchers' deepest desires is to see into the unanesthetized brain. “The real question here is, ‘How does the brain work?’” says Stephen Smith of Stanford University in California. And “anesthesia means the brain is not working, or at least it's not working normally.” Hoping to break this barrier, Winfried Denk, a co-inventor of two-photon microscopy who is now at the Max Planck Institute for Medical Research in Heidelberg, Germany, has developed a miniature two-photon microscope that can be mounted on the head of an unanesthetized rat. Ultimately, Denk hopes to image activity in the cortical neurons of behaving animals, but he says the system isn't quite there yet.

As all these techniques evolve, they may bring about an optical revolution in neuroscience. And at the very least, they should generate loads of pretty pictures.

16. # Quantum Dots Get Wet

1. Caroline Seydel*
1. Caroline Seydel is a freelance science writer in Los Angeles.

New coatings protect nanocrystals from the watery environment of the cell and pave the way for widespread use of the versatile fluorescent clusters in biology

Decades ago, physicists first observed that semiconductor particles smaller than about 10 nanometers in diameter behaved in some ways like a single atom, despite being made up of thousands of atoms. These so-called quantum dots appealed to electrical engineers as potentially useful in computer chips and lasers. Quantum dots, also called nanocrystals, further captured the imagination because laser light can make them glow in a rainbow of colors depending on their size.

Biologists have recently begun to borrow these nanobaubles. First used to stain cells in 1998, the Technicolor specks showed great promise for biological studies, but they needed some fine-tuning. They tended to clump together, for one, and they couldn't be targeted to specific molecules. Since then, biologists and materials scientists have tinkered with different coatings to make the dots more palatable to the cell and tag certain cellular molecules without interfering with normal biology.

In the last few months, several papers have shown that quantum dots can stand up to the rigors of biological processes, bringing quantum dots closer to replacing traditional organic dyes for some types of imaging studies. And companies springing up to produce the tiny fluorescent particles commercially have made the technology more easily obtainable.

The recent work has shown that biological molecules, such as antibodies, can be attached sturdily to quantum dots. The antibodies can then ferry these glowing crystals to specific molecules either on the cell surface, in its sloshy interior, or even inside the nucleus. The dots have no apparent adverse effects on the cells' health or development. Other investigators have used quantum dots to tag molecules in live frogs or mice without harming the animals.

“The first experiments showed it was possible and that there might be advantages,” says chemist A. Paul Alivisatos of the University of California, Berkeley, whose team published one of those early papers in 1998. “At this point … biologists are seeing experiments where new biology is being discovered,” he says. “Five years from now, I think [quantum dots] will be a very commonly used label.”

## A quantum leap

Unlike commonly used organic dyes, which fade within minutes, quantum dots keep shining for weeks or longer, allowing researchers to watch cellular processes unfold. Quantum dots also allow researchers to use multiple colors simultaneously, perhaps to track several molecules as they mingle. That's difficult to do with organic dyes. Because organic dyes are chemically distinct, each one requires a different laser wavelength to activate its fluorescence. The orange dye rhodamine 6G, for example, glows when zapped with green light, whereas DAPI appears blue and absorbs violet. But because one material—cadmium selenide is a common choice—can generate many colors of quantum dots, all those colors can be activated by one wavelength.

Small quantum dots emit photons of a short wavelength, which means they look blue; the larger the dot, the longer the wavelength and the redder the light given off. This color scheme arises because the semiconductor's electrons exist at distinct energy levels, called bands. A photon gives the electron a burst of energy, moving it up to a higher band. As the electron drops back into its original position, it releases a photon with the same energy as the gap between the bands. The smaller the particle, the more energetic the photons—and therefore the shorter their wavelength and the bluer the color.

Still, quantum dots might not be right for every experiment. Despite their diminutive moniker, nanocrystals are about 10 times bigger than most organic dyes. These more nimble dyes could potentially get into tightly regulated areas of the cell off-limits to bulkier quantum dots. The comparatively large dots could interfere with molecular interactions or spawn new ones by clinging to one another. Despite efforts to reduce the dots' affinity for one another, they still tend to cluster together once they enter a cell.

Although some drawbacks can't be helped, researchers have recently overcome a number of practical hurdles to adapting quantum dots for use in cells. First, the dots are typically synthesized in a bath of organic solvents; to become suitable for use in cells, they had to be stable in water. “In the water environment, the quantum dots died,” recalls cell biologist Sanford Simon of the Rockefeller University in New York City. The outer coatings of the dots had to be improved as well, so that they would stick to their designated target without too much background binding. Finally, biologists wanted assurance that the cells they study wouldn't be poisoned by the nanoparticles.

Several groups have now demonstrated long-term, specific labeling of living cells without diminishing the cells' vigor. Using dots designed by Hedi Mattoussi and colleagues at the Naval Research Laboratory in Washington, D.C., Simon and his colleagues tested the dots' ability to label live human and slime mold cells over time. Mattoussi's team had developed a way to replace the usual organic shell around quantum dots with a more water-friendly coating, allowing the dots to withstand the cell's watery milieu. In addition, the researchers improved techniques for affixing proteins to that coating, which makes the dots more useful for labeling proteins within the cell.

Mixing the dots into a dish of growing cells allowed the cells to gobble the dots, Simon and collaborators reported in the January issue of Nature Biotechnology. The cells remained labeled until they outgrew their dish, or for about a week. Slime mold cells, which gang up when starved to form a sluglike creature, behaved normally even when full of the glowing dots, indicating that the semiconductors didn't interfere with the cells' ability to develop normally or to interact with one another.

The team also labeled mammalian cells using quantum dots linked to an antibody. When they added the dots to the cells, the antibody connected with its target, a protein on the cell surface called Pgp. Only cells known to express Pgp glowed, demonstrating the quantum dots' specificity.

Similarly, Xingyong Wu of Quantum Dot Corp. in Hayward, California, and his colleagues used quantum dots to tag a molecule called Her2 on the surface of live breast cancer cells. Pathologists use Her2 to identify tumors that are likely to respond to an anticancer drug, but current testing procedures miss some tumors that are sensitive to the drug. Wu says that quantum dots produce a more sensitive, stronger signal than the chemical tests now in use. The researchers reported in January that they successfully labeled Her2 and molecules embedded in the nucleus simultaneously, which showed that different colored dots could be used together to distinguish different parts of a single cell.

“We're very excited about [quantum dots] as a way to do multiple color imaging,” says Scott Fraser, a cellular dynamics researcher at the California Institute of Technology in Pasadena. Still, the technology is in its infancy, he says. “We certainly don't have the whole tool kit that people have developed” for widely used dyes.

Alivisatos and his team have applied the glimmering nanocrystals to cancer diagnosis in a different way. To test tumor cells for their propensity to roam, pathologists currently grow the cells in a chamber covered by a membrane. If the cancer cells crawl through the membrane, that signals that the tumor is likely to spread. The test doesn't always work, though, says Alivisatos. Enter quantum dots.

Growing the cells on a layer of quantum dots reveals whether the cells move, because the cells ingest the dots near them. Cells that travel leave a dot-free swath of darkness in their path, Alivisatos says. Among cancer cell lines known to move around the body, even those that fool the membrane test get caught by the quantum dots.

## It's alive

Looking at cells in a dish reveals only so much about biological systems. To answer some questions, scientists must study live animals. And quantum dots could prove just as useful in vivo as in the test tube.

One way to make a semiconductor friendly to a cell is to hide it in a lipid envelope, as physicist Benoit Dubertret, now at the Optical Physics Laboratory in Paris, and his colleagues at Rockefeller University have done. Tucked inside cell-friendly lipids, quantum dots had no effect on the frog embryos the team injected the dots into. The embryos developed normally into tadpoles (Science, 29 November 2002, p. 1759).

All of the offspring of the original dot-containing cell inherited a portion of the fluorescent grains, allowing researchers to track cell lineages over time. “There's no other way to do this,” says co-auther David Norris, a chemical engineer at the University of Minnesota, Twin Cities, because other dyes and fluorescent tags fade too quickly or aren't passed on when the cell divides.

Another group, led by biomedical engineer Sangeeta Bhatia of the University of California, San Diego, coated quantum dots with a molecule that binds to lung blood vessel cells, then injected the dots into mice. As hoped, the dots appeared in the lungs and almost nowhere else. Dots similarly targeted to tumor blood vessels also homed in on their quarry, lighting up the expected regions inside the mice. The researchers found, however, that the dots didn't accumulate in the designated tissues as much as organic dye did, possibly because the dots' larger size prevented them from getting beyond the periphery of the tissue.

Shuming Nie of Emory University in Atlanta is also developing quantum dots to target human tumor cells growing in mice. “We can conjugate the dots to a peptide or antibody to recognize a specific cancer cell in the body,” he says. The dots don't need much light to become energized and start glowing, which is a bonus when the light has to travel through a whole mouse. Tuning the dots to radiate in the infrared, he says, eliminates the problem of tissue damage from the dots' energy emissions. He hopes to eventually develop a quantum dot drug-delivery system targeted to the cancer cells. The dot, bound to both a molecule that recognizes the cancer cell and a drug, could be rigged to release the drug only when hit with laser light, allowing control of which cells receive the toxin and minimizing side effects.

The new results have made quantum dots more accessible to researchers who haven't worked with them before, says biomedical engineer Warren Chan of the University of Toronto. “The old way was like witchcraft magic,” he says. “People knew the tricks, because they'd been working on [quantum dots] for several years. … The new procedure is more standardized.” Simon concurs: “I do think they've come of age to the point where most people can just pick them up and use them,” he says.

Quantum dots might never completely replace organic dyes, says Wu; some applications will require the dyes' smaller size. And because dyes have been around longer, he adds, they're more predictable. “It's hard when people have been doing something for 10 or 15 years to try something new,” adds Chan. “But people will be amazed at the different advantages of quantum dots for their microscopy work.”