EDITORIAL What’s on the Label?

hen you are buying food, are you one of the 30% of shoppers (an estimate in the United Kingdom) who always read the labels, or one of the 20% who rarely or never give them a glance? Do you know what to make of them if you read them? Labels are meant to inform you and to help you to choose. But when you go shopping, how much time do you have to read about the differences between 30 types of chicken soup or 300 varieties of breakfast cereal? Consumers seem to want more and Wmore choice, and consumer pressure groups definitely want more information on food labels. Choice and information are also attractive to regulators, because these options are less likely to be viewed as restricting individual freedom or stifling food industry innovation than the alternative of regulating food content. In the United States, labeling regulations are largely about the material content. In Europe, the method and place of production may also be specified in law, even if they make no material difference to the contents. This difference in approach is evident in the labeling of genetically modified (GM) foods. Whether the plant from which a food is made is GM is irrelevant in the United States, given its emphasis on overall content rather than process. But in Europe, labeling of foods containing DNA or protein from GM plants is mandatory, and legislation has now been extended to include purified derivatives such as glucose syrup and canola oil (but not products from animals fed on GM animal feed or products made with GM technology, such as cheese). Transatlantic differences in food labeling are also apparent when it comes to the biggest current challenge for food policy: obesity. Doing something about obesity is especially difficult for governments and regulators, because diet and lifestyle are in the territory of personal freedom, not state intervention. At the same time, the health care costs are potentially huge, so the pressure for action is on. The blend of action that is emerging, in both Europe and the United States, includes voluntary changes by the food industry, public education, and better labeling. Some countries and U.S. states are going even further; for instance, by restricting what can be sold in school vending machines and restricting television advertising. All of these changes are meant to make it easier for people to choose a healthy diet. The world’s fattest nation, the United States, has what is arguably the best nutrition labeling, with a mandatory nutrition facts panel. So would better labeling help? The largest food retailer in the United Kingdom, Tesco, has said that it plans to test a “traffic light” system, using red, yellow, and green colors to give consumers simple information about the main nutrients. Some object to this because of the potential implication that there are good (green) and bad (red) foods, whereas the traditional mantra from nutritionists is that there are only good and bad diets. But the food/diet distinction has changed as many people rely increasingly on ready-made meals or snacks. Research in the United Kingdom suggests that people would actually favor a simple sign-posting system such as traffic lights. The food industry is responding to public interest in diet and health by making foods that claim to have specific health benefits. These come close to the border between food and medicine. You can buy cholesterol- reducing margarine, eggs that contain long-chain omega-3 unsaturated fatty acids, and yogurts that claim to help you balance your gut flora. The U.S. Food and Drug Administration has a three-tiered system for such health claims, depending on the strength of the evidence for the claim. The European Union does not have specific regulations, but plans to introduce rules within the next 2 years that will require the independent evaluation of health claims by the European Food Safety Authority. The implications of science-based regulation are enormous for the worldwide food industry, both because products that claim to improve your health are generally highly profitable and because, in the science of nutrition, there is often disagreement among experts. Over the next decade, increases in our understanding of the relationship between an individual’s genetic makeup and his or her nutritional needs will open up a whole new area for debate about what goes on the label. The world of choice is not going to get any easier. John Krebs John Krebs is chairman of the Food Standards Agency, UK. CREDIT: RED/GETTY IMAGES PHOTODISC

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1101 Published by AAAS H IGHLIGHTS OF THE RECENT LITERATURE EDITORS’ CHOICE edited by Gilbert Chin

CELL BIOLOGY Endocytosis at the Hub In clathrin-mediated endocytosis, a network of proteins assembles on the cytoplasmic surface of the plasma membrane and promotes the pinching off of a membrane-bounded clathrin-coated vesicle. Together, the proteins select cargoes that are car- ried either inside the vesicle or in its membrane, modify the shape of the membrane, and drive invagination, vesicle scission, and eventual uncoating. A key player in this process is the AP2 clathrin adaptor protein, which is involved in concentrating selected cargo in the newly forming clathrin-coated pits. In protein interaction networks, hubs are proteins that have dispropor- tionately high numbers of interaction partners; in biological processes, hubs provide a temporal or spatial ordering to protein interactions. Praefcke et al. treat clathrin-mediated endocytosis as a module of a net- work and show how the α-appendage part of the AP2 protein works as an interaction hub. Only after being concentrated at sites of endocytosis do the appendages provide a multivalent binding platform (hub) for interac- tion partners (i.e., endocytic cargoes or other cargo adaptors). Thus, the partners will then be represented according to their relative affinities and concentrations in endocytic clathrin-coated pits and vesicles, even though any individual interactions will have been transient. — SMH EMBO J. 10.1038/sj.emboj.7600445 (2004). Interacting partners.

APPLIED PHYSICS those regions that are buried To assess the amount of IMMUNOLOGY Registering under several layers of epitaxially clathrate stored and to Three in One grown semiconductor material. evaluate its stability, Buffett Nanostructures Ruess et al. have used a registra- and Archer developed a Vaccines are designed to The manipulation of atoms tion technique that allows the mechanistic model for generate robust immunity using scanning tunneling alignment of macroscopic clathrate dynamics based on through the coactivation of microscopy (STM) has long electrodes to the nanoscale experimental and theoretical the adaptive and innate arms promised the ability to fabricate device elements buried under- data on its stability and on of the immune system. This is nanometer- and atomic-scale neath. The registration markers factors affecting its formation achieved by steering tripartite electronic device structures. are etched into the substrate and release, such as the responses to antigenic epitopes However, the realization of before the STM manipulation supply of carbon to sediments by helper T (TH) cells, antigen- presenting dendritic cells Registration markers stage and so should be a and its diagenesis, storage,

1969 (2004) Buried nanostructure general method for bottom-up and oxygen content. (DCs), and antibody-producing 4,

. 24nm epitaxial fabrication of other nanoscale Application to the current B cells or the cytotoxic silicon device structures. — ISO ocean basin implies that the lymphocytes (CTLs) that Nano Lett. 4, 1969 (2004). ultimately execute pathogen Contact leads global inventory is on the 18 clearance. However, the poor ET AL.,ET NANO LETT order of 10 g of carbon 25 µm GEOLOGY stored as methane gas and inherent immunogenicity of clathrate. The modeling peptide epitopes favored in Si substrate Taking Inventory results imply that increasing some vaccine formulas dictates Schematic showing An enormous amount of temperature would likely the need for including complex contacts (yellow) to a buried methane, an important green- deplete this inventory and potentially toxic adjuvants 90-nm-wide quantum wire. house gas, is stored in considerably; rebuilding that stimulate the essential sediments in the ocean basins would take several million priming activity of DCs. robust devices has been a as icy methane clathrate and years. The model also reveals Jackson et al. have refined difficult goal to attain simply as gas trapped by this ice and that unless the oxygen this approach by synthesizing because of the engineering by sediments. Catastrophic content of the deep oceans structures containing TH

. RUESS 10.1038/SJ.EMBOJ.7600445 (2004)(BOTTOM) problem of making electrical release of methane from this was lower than at present, epitopes coupled to B cell contact to the fabricated warehouse has been suggested not enough methane would or CTL target epitopes. These structure. The problem is that to have caused abrupt climate have been stored to account were linked via a lipid moiety, ET AL.,ET EMBO J once the sample is removed change (warming) in the past, for the carbon isotope shift which served to activate DCs from the ultrahigh vacuum and there are concerns that a and the abrupt warming at through binding and activation where the STM atomic warmer future climate may the Paleocene-Eocene of the innate signaling receptor manipulation has taken place, destabilize this reservoir, boundary, about 55 million TLR2. With different epitope the actual location of the which would enhance years ago. — BH

CREDITS: PRAEFCKE (TOP) structure is lost, particularly warming further. Earth Planet. Sci. Lett. 227, 185 (2004). CONTINUED ON PAGE 1105

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1103 Published by AAAS CONTINUED FROM 1103 EDITORS’ CHOICE

combinations, strong antibody and CTL conjunction with solid particles, are used responses could be elicited in models of as stabilizers because they reduce the viral and bacterial infection, as well as to surface tension of the liquid, preventing tumors, and were comparable to responses the bubbles from coalescing. to an adjuvant traditionally used in Alargova et al. have found that polymer vaccines. The ability to combine adjuvant microrods made from an epoxy-based and antigenic properties in a single photoresist can stabilize foams so that synthetic formula offers an attractive they resist collapse even when most of approach for future vaccine design. — SJS the liquid is allowed to evaporate. In Proc. Natl. Acad. Sci. U.S.A. 101, 15440 (2004). contrast to foams made with the common household detergent sodium dodecyl PHYSIOLOGY sulfate, which survived for 2 days, the polymer rod foams were stable for more Weight Control: It Takes a Village than 2 weeks. The authors speculate that About 250 million adults worldwide are the greater stability is due to two factors. obese, a condition that puts them at First, the rods induce a much thicker great risk for diabetes, heart disease, and liquid layer between the air bubbles, and other serious health problems. Although these layers sterically repulse each other, remarkable progress has been made in thus preventing coalescence of the understanding the physiological and bubbles. Second, the rods within a layer environmental factors that regulate form an intertwined network, thus body weight in mammals, much remains increasing its overall strength and also to be learned. imparting to the bubbles a spherical A new study in mice points to a shape, which tends to be highly unstable surprising participant in body weight in ordinary foams. — MSL control: the community of bacteria Langmuir 10.1021/la048647a (2004). (microbiota) that colonize the gut. Bäckhed et al. found that when they MICROBIOLOGY introduced the gut microbiota of normal Treasure Trove mice into a special strain of “germ-free” mice, the recipients showed a 60% Many bioactive small molecules were increase in total body fat within 2 weeks, originally identified by screening extracts even though they had eaten less and from microorganisms. These so-called exhibited an increased metabolic rate. natural products, some possessing medicinal The microbiota appeared to promote fat value, then became the targets of structure storage by stimulating the synthesis of determination and total synthesis. triglycerides in the liver and Traditional pro- their deposition in adipocytes duction methods (fat cells). Based on their depended on results, the authors hypothesize being able to pin- that changes in microbial point the source ecology prompted by Western of the metabolite diets or differences in microbial and to cultivate ecology between individuals high-yielding living in Western societies strains of the may affect predisposition isolated organism, toward obesity. — PAK but molecular Proc. Natl. Acad. Sci. U.S.A. 101, 15718 T. swinhoei. biological advances (2004). have made it feasible to look directly for the genetic CHEMISTRY components of the biosynthetic pathways. Keeping One’s Head Piel et al. have extracted from the marine sponge Theonella swinhoei the Foam consists of a mass of bubbles—air gene clusters encoding the enzymes that trapped within thin liquid shells—that make the polyketide onnamide A. Analysis forms on agitation, such as when beer is of the gene structure indicates that their poured into a glass. The stability of a true source is probably an as-yet-uniden- foam depends on the nature of the liquid: tified bacterial symbiont, possibly of evanescent in some cases, hardier in others the Pseudomonas genus, harbored by because of additives that extend their the sponge. — GJC lifetime. Surfactants, occasionally in Proc. Natl. Acad. Sci. U.S.A. 101, 16222 (2004). CREDITS: YOICHI NAKAO, UNIVERSITY OF TOKYO

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 Published by AAAS NETWATCH edited by Mitch Leslie

TOOLS Protein Sorter Proteomic and genomic experiments pour out long lists of proteins. Re- searchers who need help comparing these proteins and figuring out what IMAGES they do can open PANDORA, a protein-clustering tool hosted by the Shooting the Moon Hebrew University of Jerusalem in Is- In the last year, the moon has put on a show for earthly observers, rael. Users enter the proteins from with two eclipses. If the events have whetted your appetite for their experiment, and then PANDORA lunar images, this pair of sites will allow you to explore the moon on gathers descriptions of the entries large and small scales. The gallery* from the Lunar and Planetary In- from other databases and uses them stitute in Houston, Texas, supplies a digital version of NASA’s classic to parcel the molecules into smaller 1971 atlas, a compilation of photos snapped by the Lunar Orbiter groups. The procedure “grabs the big picture,” says co-creator missions. Armchair astronauts can search for the 114-km-across Michal Linial of Hebrew University. For example, proteins that H. G. Wells crater, the pockmarked Mare Australis, and other surface clump together in the analysis may also work together to perform features.You can also browse the text of the original atlas. a specific task or may congregate in the cell. We think of the moon as gray, but under a microscope some of its www.pandora.cs.huji.ac.il rocks are surprisingly colorful. For a sample, check out this primer† on moon rocks and soil from geologist Kurt Hollocher of Union College EDUCATION in Schenectady, New York. The multicolored speckles above come from impact melt breccia, rock that partially melted when a mete- orite or other wandering object slammed into the moon. Chemistry Behind the Headlines *www.lpi.usra.edu/research/lunar_orbiter A researcher who submits a paper to a journal knows it has to † www.union.edu/PUBLIC/GEODEPT/COURSES/petrology/ pass the scrutiny of other scientists.The Web site Chemistry Is in moon_rocks/index.htm the News gives students the chance to put their work through peer review while thinking and writing about science’s role in EDUCATION current issues, from global warming to OxyContin When Genes Go Bad addiction. Run by chemist This primer on genetic diseases from Rainer Glaser of the Uni- the U.S. National Library of Medicine versity of Missouri, Colum- can serve as a reference for students bia, and colleagues, the and help teachers catch up on the latest site provides guidelines to findings.The goal of the Genetics Home help students write reports Reference is to bridge a gap between re- about science-related sto- searchers and genomics newbies, says ries that appear in the project director Alexa McCray: “We press. After exploring, say, were well aware of the wonderful the chemistry of the things that have happened as a result of ozone-depleting pesticide the human genome project, but there methyl bromide and its was no system that translated that in- possible effects on society, formation so that members of the pub- students can then post lic could understand it.” their efforts for evaluation The handbook section explains top- by their classmates or stu- ics such as inheritance, different kinds dents at other universities. of mutations, genetic testing, and gene therapy. (Above, a virus toting modified DNA slips into a In most science courses, says cell.) Users can learn about the genes responsible for illnesses and read up on some 100 conditions, Glaser, “students are not chal- from Alzheimer’s disease (certain forms stem from mutations) to X-linked sideroblastic anemia, in lenged to think in broad terms which patients make too little hemoglobin. You can browse the descriptions by gene, condition, or and write about it.” Teachers chromosome. For readers who want to delve deeper, links lead to technical resources such as can apply to join the four uni- PubMed abstracts and gene reviews written for clinicians. versities already participating. ghr.nlm.nih.gov/ghr/page/Home ciitn.missouri.edu/testsite/www/ ciitn_main.html Send site suggestions to [email protected]. Archive: www.sciencemag.org/netwatch CREDITS (TOP TO BOTTOM):TO COLLEGE;HOLLOCHER/UNION KURT (TOP CREDITS HEBREW UNIVERSITY OF JERUSALEM; OF MEDICINE LIBRARY NATIONAL

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1109 Published by AAAS This Week PAGE 1115 1116 1120 1123 1124 Setback for Spintronics “open access” comes into NEWS focus

U.S. SCIENCE POLICY between the White House and some scien- tists, led by the Boston-based Union of Con- cerned Scientists (UCS), over allegations that Bush Victory Leaves Scars—and the Administration has manipulated or sup- pressed science advice to advance its political Concerns About Funding agenda (Science, 9 April, p. 184). “I don’t think that it was good for science to have U.S. presidential science who handles government done that,” he says. “It was clear from the be- adviser John Marburger affairs for the American ginning because of the sweeping nature of the has some sharp words Physical Society. “It’s one charges that the list of concerns were coming for researchers who pub- more factor in an increas- from the Democrats.” licly opposed President ingly complex situation,” That’s not true, says UCS chair Kurt Got- George W. Bush’s reelec- says David Moore of the tfried. “We’re used to having our advice ig- tion: Wrong message. Association of American nored or our recommendations rejected by Wrong audience. Wrong Medical Colleges, who both parties,” says Gottfried, a physicist candidate. worries that fallout from emeritus at Cornell University. “But we felt Fresh from the elec- the recent campaign could strongly that the quality of the scientific in- tion triumph by his boss formation coming from and the Republican Party, this Administration was Marburger warned last “If we’re not careful, the being compromised by the week in an interview scientific community can way the process was being with Science that criti- managed. And that’s why cism of the Administra- become estranged from we spoke up.” Adds ecolo- tion’s science policies the rest of society.” gist Jane Lubchenco of during the campaign may Oregon State University in be undermining public —John Marburger Corvallis, “I can’t speak support for science. White House Science Adviser for others, but I can assure Offering a vigorous de- you that my own motiva- fense of the Administration’s record, Mar- determine whether the Bush Administration tion was not political.” burger blamed critics for “looking at how “reaches out and engages [the science com- Representative Sherwood Boehlert the sausage is made” rather than at the prod- munity] or goes in its own direction.” (R–NY), chair of the House Science Commit- uct itself, which he characterized as a record If Marburger’s analysis is correct, it’s not tee, thinks that everybody needs to take a deep windfall for science. Such partisan attacks, the Administration but its scientific critics breath. “Shame on both sides,” says the self- he suggested, may make it harder to prevent who have gone their own way, losing touch proclaimed science booster, who this fall science from losing ground in the next with society’s concerns in the process. “Sci- underwent triple bypass surgery but still man- 4 years given the demands of the war in Iraq, ence needs patrons, and our patron is society,” aged to be reelected comfortably to a 12th national security, and economic recovery. said the 63-year-old applied physicist, a for- term. “The rhetoric got a little bit excessive. I Marburger’s remarks came just 1 day after mer university president and head of hope the Administration will demonstrate a Bush described how he planned to “spend the Brookhaven National Laboratory. “But if greater degree of interest in the opinions of the political capital” from a 51% to 48% victory we’re not careful, the scien- scientific community. And I over Democrat Senator John Kerry and the tific community can be- hope that scientists will increased Republican majority in both houses come estranged from the realize that what they have of Congress to reform Social Security, rewrite rest of society and what it been saying [about the Bush the tax code, and achieve other priorities. Sci- cares about.” Administration] hasn’t ence lobbyists are already worried about what Marburger said his re- helped the profession.” will happen when Congress returns next marks were directed at the For most rank-and-file week to finish the 2005 budget for the fiscal 48 Nobel laureates who scientists, the acid test for year that began 1 October. Their level of anxi- publicly endorsed Kerry whichever party is in power ety rises when they speculate about possible last summer and a group— is the flow of federal dollars flat funding for key science agencies in the Scientists and Engineers into research. Not surpris- president’s 2006 budget request this winter. for Change—that spent ingly, there is sharp dis- And they may have to court new chairs of $100,000 to stage about 30 agreement between the legislative panels that set policy and control events on university cam- president’s supporters and budgets after a major reshuffling next year. puses around the nation at his critics about how well “Rightly or not, I think the science com- which researchers criticized science has fared in his first munity is now perceived by this White House Bush’s policies. The get- term even under that seem- as the enemy, and that will make it harder to out-the-vote effort came on Bush whacks. Supporters cheer the ingly objective measure. L

open doors,” says physicist Michael Lubell, the heels of a fierce fight president’s reelection over John Kerry. Throughout the cam- BOTTOM): TO MADDALONI; CHRIS (TOP CREDITS EDMONDS/AP PHOTO RON

1110 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org Published by AAAS Focus 1115 1116 1120 1123 1124 Music and Flu models RNAi: not evolution say vaccinate so specific kids after all?

paign and again last week, Marburger touted ment that science has not fared well in this ton and Bush Administrations. And they say a 44% increase over 4 years in the govern- Administration,” says Marburger. that most of the added defense spending ment’s overall research and development Critics see the numbers differently, how- goes to new weapons systems and fighting budget, from the $95 billion the Bush Admin- ever. In particular, they cite the government’s terrorism, not basic research. istration inherited in 2001 to its 2005 request failure even to begin the process of doubling Marburger and the Administration’s critics for $132 billion. Although defense-related re- NSF’s budget, a promise written into a non- may disagree over the record of the past search has led the way with a 62% hike, the binding 2001 law, and the fact that Congress 4 years, but both sides accept that the next National Institutes of Health and the National has typically bulked up the Administration’s few years will be tough for science budgets. Science Foundation (NSF) have chalked up initial request for NSF each year. They note Science’s share of the overall discretionary gains of 42% and 30%, respectively. “You re- that NIH’s budget rose only 3% this year spending budget (minus the mandatory pay- ally have to work at it to make a counterargu- after a 5-year doubling that spanned the Clin- ments and debt service that consume the L

California’s Proposition 71 Launches viewers from outside the state who are experts in the field. California researchers will be able to put the voters’ largess to Stem Cell Gold Rush good use, the initiative’s backers assert. “Our goal has always been to mimic the NIH structure as much as possible,” says Snyder. “We all California is poised to leap ahead of the federal government as a backer come from the NIH tradition. … We’re not amateurs. We really know of stem cell research after voters last week approved a 10-year, $3 bil- how science should be funded and conducted and administered.” lion plan to invest in the field. But the state is likely to proceed down a Snyder says training will be a priority, and 10% of the budget familiar path: Supporters of Proposition 71, will go to build research which passed by a 59% to 41% margin, say facilities and buy equip- the new California Institute for Regenerative ment. As for the rest, Medicine will be modeled after the National Weissman expects the Institutes of Health (NIH) in allocating its bulk of the research $295 million annual budget. funding to support in- The state bond initiative, which will sup- vestigator-initiated ba- port work involving nuclear transfer (so- sic research rather than called research cloning), was backed by a any top-down priorities staggering array of scientists and high- set by the oversight profile groups and received a last-minute committee. endorsement from Republican Governor One important early Arnold Schwarzenegger. decision will be the se- Jubilant supporters are lection of a full-time di- now moving on to the rector. It “could easily next phase, beginning with be someone who’s al- the selection within 40 Close to home. Hollywood producer Doug Wick and ready an NIH adminis- days of a 29-member In- family celebrate passage of California’s Proposition trator, with California dependent Citizen’s Over- 71, which may one day help his daughter Tessa (far roots,” says Snyder. One sight Committee. The gov- right) and others with diabetes. person who fits that de- ernor and three top state scription is James Bat- officials will each appoint tey, who coordinates NIH’s $214-million-a-year investment in hu- five members, and five University of California campuses will also man stem cell research, some $25 million of which goes to embry- have seats at the table. The likely chair is Donald Klein, the real estate onic stem cells. Battey wouldn’t comment, and an NIH spokesper- magnate who led the campaign. In January the board will set up son says NIH has nothing to say because Proposition 71 is “a state working groups on research funding, facilities, and standards, with the matter.” But earlier this year Battey told The New Yorker that the last being the first order of business, says campaign spokesperson measure “could have a really transforming effect on stem-cell re- Fiona Hutton. The first awards are to be made within 60 days of issu- search … [and] would certainly make California an extremely at- ing interim standards. tractive place to conduct it.” Organizers promise that the new institute, at a site yet to be se- Despite the victory, supporters remain concerned about a bill spon- lected and with a permanent staff of 50, will be a first-class operation sored by U.S. Senator Sam Brownback (R–KS) that would outlaw all both ethically and scientifically. “The burden is upon us to prove that forms of cloning. The new Senate includes four members whose we are above reproach,” says stem cell researcher Evan Snyder of the records suggest they are likely supporters of the ban, which would leave Burnham Institute in La Jolla. Stanford University stem cell researcher Brownback only about five votes short of the 60 needed to move for- Irving Weissman, another leader in the campaign, expects universities ward. At the same time, supporters of an opposing bill that would out- to send either their presidents or medical school deans to represent law reproductive cloning but permit cloning for research remain hope- them on the oversight committee, which he hopes will require re- ful that they, too, will prevail. –CONSTANCE HOLDEN CREDITS: FRAZER HARRISON/GETTY IMAGES

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1111 Published by AAAS major part of the $2 trillion federal budget), key committee chairs. For example, Senator at almost 14%, stands at a 40-year peak, Mar- Ted Stevens (R–AK), who has shown little ScienceScope burger notes, and sustaining such a “local interest in regulating greenhouse gas emis- maximum” is unlikely. sions, is expected to become chair of the Standards Set for Anthrax His co-chair of the President’s Council of Commerce, Science, and Transportation Detection Kits Advisors on Science and Technology, Silicon Committee. He would replace Senator John Valley financial guru Floyd Kvamme, goes McCain (R–AZ), who used the slot to push Only one portable anthrax-detection kit one step further. He argues that science for climate change legislation and criticize of the five now on the market meets new doesn’t need a larger slice of the discretionary the Administration’s climate policies. standards established to help police and other first responders identify the deadly pie. “We’re close to the limit of the amount of Boehlert seems assured of remaining head bacterium, an expert group says. spending that the research enterprise can ef- of the equivalent House panel. AOAC International (formerly the Asso- fectively absorb,” says Kvamme, partner There will certainly be departures within ciation of Official Analytical Chemists) gave emeritus of the venture capital firm Kleiner, the executive branch. The first science-related the word to the Department of Homeland Perkins, Caufield, and Byers. “Of course, post to be vacant is at the Environmental Pro- Security (DHS), which paid for the analysis there are lots of other sectors demanding a tection Agency, where research chief Paul after DHS officials questioned the reliability share of that money. Gilman announced of hand-held kits.AOAC’s new standards re- But I think we’ve that next month he quire a kit to detect the presence of anthrax reached a ceiling. It’s will become head of a in a sample that contains at least 1 million unrealistic to talk about soon-to-be-an- anthrax spores and to distinguish accurately science taking up 15% nounced research between anthrax and related organisms.Ac- or 20% of the discre- consortium of univer- cording to AOAC, only the RAMP Anthrax tionary budget.” sities. NASA Admin- Test, manufactured by Canada’s Response The notion of a istrator Sean O’Keefe Biomedical Corp., meets these criteria. ceiling doesn’t sit well is thought by many to “This is going to have a major impact with research advo- be headed to another on the first-responder market” by im- cates such as Nils Has- federal post, possibly proving reliability, predicts Calvin Chue, a selmo, president of the even before this sum- pathogen-detection expert at the Johns 62-member Associa- mer’s planned shuttle Hopkins Center for Civilian Biodefense tion of American Uni- launch, the first in Co- Strategies. Adds Stephen Morse, director of Columbia University’s Center for Pub- versities, science lob- lumbia’s aftermath. lic Health Preparedness, “This is one nice byists, nor with Boeh- And Health and Hu- step in the right direction.” lert. “I think there’s man Services Secre- –DAVID GRIMM room for growth,” says tary Tommy Thomp- Boehlert, who helped son, long rumored to shepherd the NSF be out the door after Wisconsin Academics Decry reauthorization bill the election, said last Move to Water Down Darwin through Congress. week that “the presi- Wisconsin academics are rallying to re- Still, Boehlert says More space. Bush hopes Congress will fund dent and I haven’t had verse a decision last month by a local he expects “to be his plans to explore the moon and Mars, an- a chance to talk, and school board that would require students among those yelling” nounced earlier this year. it’s clearly up to the to “study various scientific models/theo- that the president’s president.” If he hits ries of origins” rather than stick with Dar- 2006 budget request for NSF and other the road, one frontrunner for the job is Mark winian theory only. science agencies, due out in February, is McClellan, a physician and economist who The Grantsburg school board’s action inadequate. headed the Food and Drug Administration be- spurred Michael Zimmerman, dean of the Next week Congress returns for a lame- fore moving to head Medicare. Lobbyists see College of Letters and Sciences at the Uni- versity of Wisconsin, Oshkosh, to organize a duck session to complete work on the 2005 him as more attuned to the interests of re- flurry of letter writing by hundreds of scien- budget and, perhaps, conduct other business. searchers than the former Wisconsin governor. tists and theologians from universities But the real action won’t begin until the more Despite the harsh language of the past around the state as well as high school sci- heavily Republican 109th Congress convenes year, Marburger says that “nothing has ence teachers.“We want to send as a strong in January. In the Senate, the GOP picked up changed” in the way he will operate: “I’m a message as we can,” says Zimmerman.Al- four seats and holds a de facto margin of 55 to going to continue to try to make sure that though Wisconsin state standards mandate 45. (Senator James Jeffords of Vermont is an scientific issues are addressed.” Harold Var- the teaching of evolution, the board con- independent but usually sides with the De- mus, head of the Memorial Sloan-Kettering tends that the district has a right to make mocrats.) In the 435-member House of Repre- Cancer Center in New York City and one of the standards more “inclusive.” sentatives, the GOP could have a 31-seat ad- the Nobelists who endorsed Kerry, expects Last month, the Dover Area School Board vantage after runoffs next month in Louisiana. to “still have conversations with people in in Pennsylvania approved the teaching of The new Senate is more likely to endorse government who hold opposing views.” And “intelligent design” (Science, 5 November, the president’s massive energy bill and over- NASA climate scientist James Hansen, who p. 971).And a trial over an evolution “dis- haul the Endangered Species Act, but legis- gave a widely publicized speech criticizing claimer” in textbooks is under way in Geor- lators may be more skeptical of his expan- Bush just before Election Day, says “every- gia. Says Eugenie Scott of the National sive—and expensive—moon/Mars explo- body is telling me to watch my back. … But Center for Science Education in Oakland, ration program, particularly at a time when I spoke from the heart, and these are honor- California: “After last Tuesday there are a lot of happy creationists around the private spaceships are capturing the head- able people.” –JEFFREY MERVIS country.” –CONSTANCE HOLDEN lines. The climate for science-related issues With reporting by Jocelyn Kaiser, Andrew Lawler,

CREDIT: J. APPLEWHITE/AP SCOTT PHOTO will also depend, in part, on who fills several and David Malakoff

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1113 Published by AAAS N EWS OF THE WEEK U.S. HIGHER EDUCATION Decline in New Foreign Grad Students Slows The number of international students begin- ceived “not one e-mail or phone call this the overall numbers are still discouraging.” ning graduate studies at U.S. universities has year complaining about a visa.” Another sur- University administrators say their schools declined for the third year in a row. But the vey released this week by five groups, in- still need to combat the perception that U.S. 6% drop is the smallest in 3 years, an im- cluding CGS and NAFSA: Association of campuses are unfriendly toward international provement that some attribute in part to International Educators, reported no change students. Toward that goal, some universities faster handling of visa applications. in undergraduate enrollments but declines in are reimbursing students for the $100 fee the The news, in a survey released last government charges to implement the week by the Council of Graduate Student and Exchange Visitor Infor- Schools (CGS), comes as a relief to mation System, which tracks foreign higher education organizations, which students once they arrive. “It’s not a lot had braced for the worst earlier this of money, and it sends out a welcom- year after a 28% drop in international ing message,” says Patricia Parker, graduate applications and an 18% assistant director of admissions at drop in offers of admission. Enroll- Iowa State University in Ames, which ments, which represent the final step saw a 25% drop in first-time interna- in that progression, were down 10% tional graduate enrollment this fall. in the fall of 2003 following an 8% Overall graduate enrollment is drop the year before. The decline ap- down 1%, according to the CGS sur- peared in the first academic year after vey, and 2% fewer domestic students the 11 September terrorist attacks and are entering graduate school. The life reversed several years of growth in sciences and engineering show the the number of international students. steepest declines in first-time inter- Universities have stepped up their national enrollment within the sci- efforts to assist foreign students, says Showing up. Enrollments are the last step in the process of at- ences, whereas the physical sciences CGS president Debra Stewart, “by tending graduate school, and trends vary by field. are enjoying a 6% rise in first-time streamlining their admissions processes, international students (see graph). enhancing their use of technology, and form- graduate enrollments at two-thirds of major “We made offers to more international ing international partnerships.” The council research institutions. students this year than usual, anticipating that says those measures contributed to a rise this “The good news is that the administration some of those who might accept would have year in the percentage of admitted inter- has become aware of the seriousness of the trouble getting visas,” says Allen Goldman, national students who ended up enrolling. problem and has begun to take steps to ad- chair of the physics department at the Univer- “I am pretty sure that we have gotten dress some of the obstacles that are discour- sity of Minnesota, Twin Cities. “That didn’t over a hump in terms of visa delays,” says aging or preventing legitimate students and materialize.” The result, says Goldman, is a Sherif Barsoum, associate director of the Of- scholars from coming to the United States,” larger entering class—35 rather than 25 stu- fice of International Education at Ohio State says NAFSA’s Marlene M. Johnson. “The dents—that is also more international. University in Columbus, who says he has re- bad news is that, despite some positive signs, –YUDHIJIT BHATTACHARJEE

OCEAN SCIENCE NOAA to Retool Research Programs The National Oceanic and Atmospheric Ad- ternal grants program. The panel also called side” researchers, Mahoney said. He also ministration (NOAA) has embraced sugges- for the agency to develop 5- and 20-year sci- promised to increase the number of adminis- tions to consolidate some laboratories and ence plans and to put the agency’s research trators overseeing key science programs, say- make its funding practices easier to under- programs under the control of a new senior ing that although NOAA has “an abundance stand. But officials have rejected a proposal administrator and an allied advisory board of capable researchers,” the administrative from an outside advisory panel for a new sci- (Science, 11 June, p. 1579). corps “is very thin indeed.” ence czar. On 3 November, NOAA officials told the Perhaps the biggest question mark is the “This looks like a very good start,” said board that they have nixed the last two ideas fate of NOAA’s six Colorado laboratories. Science Advisory Board chair Leonard because of congressional opposition to any Congressional critics have argued for lump- Pietrafesa, of North Carolina State University more bureaucratic layers. Instead, NOAA ing the labs together into fewer and less ex- in Raleigh, reacting last week to NOAA’s plan Deputy Administrator James Mahoney says pensive units, but researchers worry that the to shake up its $350 million research pro- his position has been “restructured” to in- move could hurt science programs. Ma- gram, which includes everything from space- crease his oversight of research. Officials said honey says a task force could issue a consol- based climate and weather studies to fisheries they expect other changes will take place over idation plan as early as this fall, followed by research and deep-sea exploration. The out- the next 18 months, including clarifying both another group looking at NOAA’s ecological side panel, led by climate scientist Berrien the amount of money available and the appli- research programs. Any proposed changes, Moore of the University of New Hampshire, cation process for extramural researchers and however, must survive vetting from the Durham, suggested in August that the agency creating a Web-based grants management White House Office of Management and consolidate half a dozen laboratories in Boul- system. “I don’t think anyone would give Budget and win the support of Congress.

der, Colorado, and revamp a convoluted ex- NOAA an ‘A’ for our involvement with out- –DAVID MALAKOFF SOURCE: CGS

1114 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org Published by AAAS SCIENTIFIC PUBLISHING ScienceScope Mixed Week for Open Access in the U.K. High Demand Leads to CAMBRIDGE, U.K.—Supporters of “open ac- ries for their researchers’ preprints and pub- cess” scientific publishing—in which au- lished papers, which would be posted at Shortage of Malaria Drug thors pay the cost of publication and accept- some agreed point after publication in a The World Health Organization (WHO) ed papers are freely available online—have journal, and called for the government to set cautioned last week that supplies of a po- received a public setback and a private boost up a body to coordinate these archives. tent antimalaria drug may fall up to 4.5 in the United Kingdom in the past few days. The true costs of open-access publishing million doses short of their demand fore- The British government, saying it is “not are still not clear, the government responded, casts. Officials at Novartis, the Swiss- obvious … that the ‘author pays’ business noting that “before fully supporting any new based company that manufactures the drug, blamed the shortfall on agricultural model would give better value for money business model, the Government will need to suppliers failing to keep up with growing than the current one,” rejected recommenda- be convinced that this model is better and demand in developing countries. tions from the House of Commons Science cheaper.” It declined for now to require that Artemether-lumefantrine (brand name and Technology Committee to fund some government-funded researchers deposit their Coartem) is a form of artemisinin-based costs associated with open-access publish- published papers in open repositories or to es- combination therapy (ACT) favored by ing. The committee promptly accused the tablish a fund to pay authors’ publication fees. WHO because of the increasing number of government of buckling under pressure from The House of Commons committee pub- strains of malaria that are resistant to tradi- scientific publishers. On the other hand, the lished the government’s response on 8 No- tional drugs (such as chloroquine). Since Wellcome Trust—the largest funder of basic vember, along with its own commentary* 2001, Novartis has provided Coartem to biomedical research in the United King- accusing the Department of Trade and In- developing countries at cost. dom—threw its considerable weight behind dustry (DTI) of neutralizing dissenting voic- A key ingredient of the drug comes from open-access publishing. It an- es within the govern- the Chinese wormwood plant (Artemisia an- nounced that it will require re- ment. “DTI is apparent- nua). Suppliers of the ingredient have strug- searchers it funds to deposit pa- ly more interested in gled to ramp up production to meet the pers in a public archive “within kowtowing to the pow- growing interest in ACT, says Andrew 6 months of publication,” and it erful publishing lobby Bosman, a medical officer at WHO.The plant than it is in looking af- takes 6 months to cultivate, and the drug re- ter the best interests of quires 3 to 5 months to process, resulting in British science,” says a mismatch between orders and supplies, he committee chair Ian says.To help combat the shortage,WHO will step up its malaria-prevention efforts and Gibson. The govern- plans to develop a prioritization system for ment’s response is “a drug distribution. –SEAN BRUICH defence of the status quo,” adds Jan Velterop, IBM Study Challenges publisher of the author- Cancer Claims pays journal Biomed Central. An IBM-funded study has found that the While the govern- company’s workers face no greater risk than Open up. Parliamentary com- the general population of developing cancer. ment was urging cau- mittee chair Ian Gibson (left) The results, which have not yet been peer and Wellcome Trust chief Mark tion on open-access reviewed, contradict an earlier study paid for Walport back public access. publishing, the Well- by attorneys of former IBM workers suing come Trust was step- the company for causing their cancers is discussing the creation of a European ver- ping up its support. The trust will now fund (Science, 14 May, p. 937). sion of PubMed Central, the open archive of “reasonable” costs for its grantees to publish Researchers at Harvard University’s published papers run by the U.S. National in author-pays online journals; in addition, it School of Public Health and the University Library of Medicine. will require grantees to deposit published of Alabama examined health records of The government’s statements came in a papers in a public archive. Trust director 126,000 workers and compared overall can- detailed response to a report issued this Mark Walport estimates that publication cer death risks between IBM workers and summer by the House of Commons commit- charges will only amount to 1% of the trust’s the general population.The workers had a tee (Science, 23 July, p. 458). The committee research costs. The goal is “to achieve maxi- 16% lower risk of cancer and a 35% lower concluded that open-access publishing is an mum value from our research through maxi- risk of dying than the general population, “attractive” idea but called for further study mum distribution,” says Walport. according to a memo released last week. because of the possible impact on learned Gibson says the committee will continue Joseph LaDou, director of the University societies, which rely financially on journal the fight. Librarians are “gung-ho” about of California, San Francisco’s International subscriptions. The panel also was concerned public access, he says, and he hopes that the Center for Occupational Medicine, who gave that the pharmaceutical industry, which sub- research councils will soon come out in fa- unpaid advice to lawyers for former IBM scribes to many journals but contributes few vor too. In the new year, Gibson says, there workers suing the company, believes IBM bi- papers, would get a free ride. In the mean- will be a debate in the House of Commons: ased the study and says the results could re- flect the fact that manufacturing workers time, the committee recommended that the “We’re going to argue this with them.” tend to be healthier than the population at research councils create a fund to which au- –DANIEL CLERY large. IBM representatives did not return re- thors could apply for the costs of publishing peated calls seeking comment. * in open-access journals. It also called for www.publications.parliament.uk/pa/cm200304/ –DAN FERBER

CREDIT (RIGHT): WELLCOME THE TRUST U.K. universities to set up online reposito- cmselect/cmsctech/1200/1200.pdf

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1115 Published by AAAS N EWS OF THE WEEK PALEOANTHROPOLOGY Skeptics Question Whether Flores Hominid Is a New Species When a research team announced last month quire a long period without interbreeding). that it had found a new species of 18,000- But archaeologist Michael Morwood of year-old tiny human in a cave on the Indone- the University of New England in Armidale, sian island of Flores, it seemed almost too Australia, a leader of the team that discov- amazing to be true (Science, 29 October, ered the skeleton, insists that the skeleton is p. 789). Now a small but vocal group of sci- not a pathological case. “We now have the entists argues that the skeleton dubbed Homo remains of at least seven individuals,” he floresiensis is actually a modern human af- says. “All are tiny, and all can be referred to flicted with microcephaly, a deformity char- as Homo floresiensis.” acterized by a very small brain and head. The team is backed by several outside re- Meanwhile, an Indonesian scientist who also searchers. Anthropologist Leslie Aiello of Surprising skull. A few researchers say the Flo- challenges the skeleton’s status has removed University College London says the skeleton res skull may be a deformed modern human. the skull to his own lab for study. But mem- cannot be that of a modern human because the bers of the original team of Australian and In- postcranial bones indicate a separate species. relation to the legs.” Chris Stringer of the Nat- donesian scientists staunchly defend their “The pelvis is virtually identical to that of an ural History Museum in London sums up analysis, and outside experts familiar with the australopithecine,” much wider than the mod- many researchers’ opinions by saying, “This discovery are unmoved by the critique. ern human pelvis, she says. And compared cannot be a peculiar modern human.” The main challenge comes from paleo- with modern humans, “the arms are long in –MICHAEL BALTER pathologist Maciej Henneberg of the Univer- sity of Adelaide in Australia and anthropolo- CONDENSED MATTER PHYSICS gist Alan Thorne of the Australian National University in Canberra. Neither has seen the Spin Current Sighting Ends 35-Year Hunt specimen itself, and as Science went to press, they had yet to publish their criticisms in a The electron’s charge gets all the glory: It is, for more than a century. But despite decades peer-reviewed journal. But Henneberg pub- after all, responsible for the plethora of elec- of work, the spin-based Hall effect had never lished a letter in the 31 October Adelaide Sun- tronic gizmos that surround us. But the par- been spotted—until now. day Mail arguing that the skull of the Flores ticle’s magnetic behavior—a property In a paper published online this week by hominid is very similar to a 4000-year-old mi- known as spin—has also been tantalizing Science (www.sciencemag.org/cgi/content/ crocephalic modern human skull found on the scientists for decades. Thirty-five years ago, abstract/1105514), researchers led by David island of Crete. And at a press conference on 5 for example, Russian theorists suggested Awschalom, a physicist at the University of November, Indonesian paleoanthropologist that impurity atoms in a semiconductor California, Santa Barbara (UCSB), report Teuku Jacob of Gadjah Mada University in might interact with electrons’ spins to redi- the first experimental sighting of the spin Jakarta claimed that the specimen was a rect currents flowing through it. A related Hall effect. “It is as beautiful as it is a diminutive modern human. Jacob, once de- effect, called the breakthrough experiment,” says Gerrit n (a.u.) scribed as the “king of paleoanthropology” in s Hall effect—in Bauer, a theoretical physicist at the Delft –2 –1 0 1 2 Indonesia (Science, 6 March 1998, p. 1482), which magnetic University of Technology in the Nether- has had the skull transported to his own lab 150 fields push elec- lands. Daniel Loss, a theoretical physicist at

from its original depository at the Center for trons around by the University of Basel, Switzerland, agrees. ET AL. Archaeology in Jakarta, according to center interacting with “The data is very clear,” he says. “It’s very archaeologist Radien Soejono, who is a mem- their charge— impressive.” The new scheme works in stan- ber of the original discovery team. 100 had been known dard semiconductors widely used in indus- In its original paper, the team considered try today. That could be a major boon to the and rejected several possible deformities, in- Sidetracked. Impu- nascent field of spintronics, which promises cluding a condition called primordial micro- 50 rity atoms deflect to create a new class of high-speed, low-

cephalic dwarfism (Nature, 28 October, ) electrons with dif- power electronic devices that manipulate the m µ

p. 1055). But Henneberg claims that the au- ( ferent spins (red and spin of electrons. n

thors failed to consider a related condition o 0 blue) to opposite An American physicist named Edwin i t i

called secondary microcephaly. “They s sides of a semicon- Hall discovered the original Hall effect in o

jumped the gun,” he told Science. Hen- P ductor chip. 1879. The effect occurs when an electric current moves through a metal strip while a neberg, who with Thorne favors a multi- –50 regional model of human origins that some magnetic field is applied top down through say is at odds with the finding of a distinct the metal. The magnetic field interacts with but recent human species on Flores, con- the charge of the moving electrons, deflect- cludes that the skeleton is “a simple Homo –100 E ing some to the left and some to the right sapiens with a pathological growth condi- sides of the strip. In 1971, Mikhail tion.” (Multiregionalism holds that modern Dyakonov and Vladimir Perel of the Ioffe 300 µm humans evolved after 2 million years of –150 Physico-Technical Institute in Leningrad interbreeding among worldwide populations; –40 –20 0 20 40 suggested that electrons’ spins might trigger L

the evolution of a distinct species would re- Position (µm) similar detours. These spins behave like BOTTOM): TO P.(TOP CREDITS OF NEW ENGLAND, BROWN/UNIVERSITY ARMIDALE, AUSTRALIA;Y. K. KATO

1116 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org Published by AAAS N EWS OF THE WEEK tiny bar magnets that point up or down. MEDICINE Dyakonov and Perel proposed that electrical defects in semiconductors could create a lo- Sperm-Targeting Vaccine Blocks calized electromagnetic field that would shunt spin-up and spin-down electrons to Male Fertility in Monkeys opposite sides of the semiconductor, a scheme that came to be known as the “ex- When the 1880s debut of vulcanized rubber antibodies, a problem other immunocontra- trinsic” spin Hall effect. Decades later, oth- and the 1930s advent of latex mark the latest ceptives have encountered. er theorists suggested that such deflection advances, one quickly understands the sorry The monkeys received boosters of vac- might also result from an “intrinsic” effect state of male contraceptives. Researchers in cine every 3 weeks. Because the vaccine due to the strain between atoms in a semi- this beleaguered field have tried to supply didn’t lower sperm count or alter sperm in conductor alloy. In recent years, theorists men with other options but have had little an easily detectable way, scientists resorted have clashed sharply over which effect success. Now, on page 1189 of this issue of to another method of testing its effective- would more likely be spotted. Science, a team in the United States and In- ness: The immunized male monkeys spent Awschalom says his team waded into dia reports preliminary results of its new several days each with three different fe- this battle somewhat by accident. Earlier contraceptive vaccine for males. Although males during the fertile peak of the fe- this year, the Santa Barbara researchers, not without problems, the vaccine prevented males’ menstrual cycle. The upshot: None who included graduate students Yuichiro pregnancy in female partners of the male of the seven vaccinated monkeys managed Kato and Roberto Myers along with electri- monkeys receiving it. to impregnate a female. cal engineer Art Gossard, discovered a “There seems to be Four of the six control scheme for electrically injecting spins into a some real promise,” says monkeys did. semiconductor, another long-sought goal of Ronald Swerdloff, a repro- The contraceptive ef- spintronics (Science, 2 April, p. 42). They ductive endocrinologist at fect of the vaccine was were tracking spins by a technique called the University of Califor- intended to be reversible; scanning Kerr microscopy, in which re- nia, Los Angeles. Still, once the booster shots searchers bounce polarized laser light off a “it’s just early in the were stopped, the re- semiconductor sample. If electrons on the game,” with too few mon- searchers anticipated that surface atoms all have spins in one pre- keys tested, to conclude antibodies to Eppin ferred orientation, the polarization of the whether the approach will would decline and fertil- ricocheting photons will rotate slightly. pan out, he adds. ity would return. But on- When the UCSB researchers used a Kerr Reproductive biologist ly five of the seven vac- microscope lens with 1-square-micrometer Michael O’Rand of the cinated monkeys, some resolution, they saw clear bands of electrons University of North Car- of whom received boost- with opposite spins huddled along the sides olina, Chapel Hill, crafted er shots for nearly 2 of the semiconductor chips. All it took to the vaccine several years years, recovered their fer- create the bands was to push an electric cur- after reporting his discov- tility during the course of rent through the semiconductor chip. ery of a novel male-only the study. “It’s hard to The researchers first detected the bands protein in 2001. The pro- say” what that means, in a semiconductor chip made from gallium tein, called Eppin, has says O’Rand. “Maybe arsenide (GaAs), similar to the chips in cell been found so far on the Sperm stopper. Male contraceptive they recovered 2 weeks phones. Then, in hopes of resolving the bat- surface of sperm cells and vaccine works in monkeys. after we quit” testing tle over intrinsic and extrinsic effects, they elsewhere in the testis and them. Although conced- looked for the effect in a semiconductor the epididymis. Its function isn’t clear. But ing that the vaccine has a long way to go, called indium gallium arsenide (InGaAs). drawing on the general strategy of immuno- O’Rand believes the study offers “a proof Theorists had suggested that the atomic contraception, in which vaccines are designed of principle” for immunocontraception, bonds in InGaAs had the right sort of strain to act as contraceptives, O’Rand reasoned that which has been so relegated to the sidelines to produce a sizable intrinsic spin Hall ef- if a male harbored antibodies to this protein, that the National Institutes of Health no fect even if the impurities didn’t play a role. his sperm might malfunction. longer funds research on it. But the researchers found that impurities O’Rand teamed up with colleagues at the Companies have also been hesitant. could explain virtually all of the buildup of Indian Institute of Science in Bangalore, New Jersey–based Organon studied im- spins they saw. “It looks like [the extrinsic which hosts a large primate research center. munocontraception for females before effect] is the major player” in the semi- There, the group vaccinated six monkeys with backing out, says Willem de Laat, the com- conductors studied so far, Loss says. human recombinant Eppin protein and admin- pany’s medical director. Instead, Organon Using standard techniques for tailoring istered a sham vaccine to six others. The team and another company, Schering in Berlin, the amount of impurity atoms and other “de- hypothesized that the monkeys needed high are testing a combination of oral progestin fects” in semiconductors, “it should be possi- levels of antibodies to Eppin in their blood for and injected testosterone. ble to engineer materials to increase the size the vaccine to work, especially because anti- O’Rand and his colleagues, heartened by of this effect,” Awschalom says. That in turn body levels drop in the epididymal fluid. So what they consider a success, are now trying could point the way for spintronics re- when only four of the six treated monkeys dis- to understand how, exactly, their vaccine dis- searchers to develop an array of spin- played antibody levels that O’Rand’s team rupts fertility. One possibility is that the manipulating devices to switch currents of deemed sufficient, the other two animals were technique leaves sperm sluggish. Whatever THE PICTORIAL GUIDE TO THE LIVING PRIMATES TO GUIDE THE PICTORIAL particular spins on and off, as well as steer, dropped from the study. The team brought in its mechanism, O’Rand and other contracep- filter, and amplify them. That might be three additional monkeys, who also reached tive researchers hope the new vaccine will enough to finally bring the electron’s spin a the desired antibody levels. It’s not clear why provide a shot in the arm for the field.

CREDIT: N. ROWE/ little limelight of its own. –ROBERT F. SERVICE some monkeys did not produce sufficient –JENNIFER COUZIN

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1117 Published by AAAS N EWS OF THE WEEK DRUG TESTING Regulators Talk Up Plans for Drug Biomarkers … New methods of predicting clinical outcomes getting new medicines to patients quickly. risk-free. “As we study biomarkers, we’re are getting serious consideration at the U.S. “We’re putting these molecules in these going to develop evidence that impugns Food and Drug Administration (FDA). Last painfully slow, archaic programs” for testing, their use,” said John Wagner, senior director week the agency invited members of one of says Paul Watkins, a liver specialist at the of clinical pharmacology at Merck Research its advisory committees to a University of North Carolina, Labs, at last week’s meeting. And Watkins, meeting in Washington, D.C., Chapel Hill. who heads a new government-funded con- where attendees explored the Although FDA is enthusi- sortium on drug-induced liver toxicity, notes use of biomarkers to monitor astic, it remains vague about that biomarkers will gain acceptance only if everything from protein levels how it might change its meth- they’re used to flag issues of safety as well to bone density, to gauge a ods. As a first step, says as efficacy. In May, the consortium began drug’s effectiveness, or possi- Lesko, it will set up an internal enrolling the first of dozens of patients who bly to measure the progression working group to discuss what suffered severe liver effects from one of four of a disease. FDA officials at qualifies a biomarker as a sur- drugs, along with matching controls. It aims the meeting said they’re rogate endpoint. It may also to link molecular markers with susceptibility launching a multiyear effort to comb through archival data for to this common side effect. bring biomarkers into the promising biomarkers. “We “Everybody wants to sit down and use mainstream of drug discovery. have to get down to some something that’s predictive” of clinical out- “There are some signifi- more specifics … to make this come, says Charles Grudzinskas, a former cant payoffs if this is success- proposal come to life,” Lesko drug industry executive and founder of NDA ful,” such as speedier drug Architect. FDA’s Lawrence Lesko admits. Working closely with Partners LLC, a Washington, D.C., consulting trials, says Lawrence Lesko, is pushing biomarkers forward. FDA’s acting commissioner firm. He adds that FDA seems ready to lead FDA’s director of clinical for operations Janet Wood- the way with an attitude that, “we’re willing pharmacology and biopharmaceutics. Still, cock, he says he hopes to foster collaborations to go out on skinny branches.” But he and Lesko has encountered hesitancy: “People between industry, academia, and FDA to get others are waiting to see whether FDA’s next are sensitive to past failures” of biomarker ideas moving beyond the basic-research stage. chief will throw the agency’s prestige and use, he says. But pushing biomarkers forward is not funds behind this cause. –JENNIFER COUZIN The new effort began with former FDA commissioner Mark McClellan, who left the CANCER RESEARCH agency in February and has not yet been re- placed. McClellan made biomarkers part of … And NCI Hears a Pitch for Biomarker Studies the agency’s “Critical Path” initiative, a plan released in March to speed drugs to market. Cancer researcher Lee Hartwell this week tember, p. 1885). Both proposals were re- Currently, most drugs are approved based proposed a major new initiative to discover quested by NCI Director Andrew von Es- on so-called clinical endpoints, such as longer biomarkers for the early detection of cancer. chenbach as part of his plan to eliminate survival for cancer drugs or fewer fractures In a white paper presented to the National death and suffering from cancer by 2015. for an osteoporosis drug. Researchers have Cancer Institute’s (NCI’s) Board of Scientific Board members peppered Hartwell with long believed that there are reliable surrogates Advisors, Hartwell, of the Fred Hutchinson questions. Because the same protein can oc- that can be detected earlier for many clinical Cancer Research Center in Seattle, Washing- cur in many forms, for example, “you have to endpoints. For example, the time it takes a ton, outlined how the project would scan realize it’s going to be much more complicat- cancerous tumor to resume growing during or thousands of blood samples from cancer pa- ed than looking for a single protein,” said Su- after a specific type of treatment may indicate tients for proteins and other biological mole- san Horwitz of Albert Einstein School of how long a patient will survive. Biomarkers cules that can indicate incipient tumors. Medicine in New York City. Richard Schilsky tied to a clinical outcome, like this one, could Hartwell said his plan would boost spend- of the University of Chicago suggested that be used as surrogate endpoints in trials. ing on biomarkers, now overshadowed in costs might approach those of drug testing if FDA already approves some drugs based NCI’s budget by new drug development and each new marker had to be validated clini- on surrogates, particularly for life-threatening even prevention trials. Although only a hand- cally. Hartwell disagreed, saying the project diseases like AIDS (for which it has used sur- ful of biomarkers are widely used, the se- would “piggyback” on other large studies rogates since 1992) and cancer. But the his- quencing of the human genome and the debut such as the Women’s Health Initiative by bor- tory of biomarkers is marred by some high- of new, automated mass spectrometry ma- rowing tissue samples. “I don’t see the valida- profile disasters. One was the widespread use chines for detecting proteins leaves the field tion adding a great deal of expense,” he said. of two antiarrhythmic drugs, encainide and ripe for new breakthroughs, Hartwell said: Several members also asked how the ini- flecainide. These drugs were intended to re- “We need to organize our efforts.” His pro- tiative would fit with existing NCI programs duce the likelihood of a second heart attack posed “Coordinated Clinical Proteomics and and the National Institutes of Health’s broader because uncontrolled arrhythmia was consid- Biomarkers Discovery Initiative” would in- Roadmap, which includes proteomics. Von ered a predictor. When a clinical trial actually clude centers for testing biomarker technolo- Eschenbach responded that it would “dove- tested the drugs for this indication in the late gies, a repository of reagents, and a public tail” with them but did not specify how. 1980s, three times as many people died in the proteomics software package. Questions about one such detail—how NCI drug arm as in the placebo group, and the The initiative, for which Hartwell offered would pay for a new biomarker plan with an study was halted. no price tag, would complement a technol- ever-tightening budget—may come up later Advocates nevertheless argue that using ogy plan outlined by the Broad Institute’s this month at a meeting of the National Can-

CREDIT: FDA biomarkers will do more good than harm by Eric Lander in September (Science, 24 Sep- cer Advisory Board. –JOCELYN KAISER

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1119 Published by AAAS News Focus

Why did the ability to carry a tune evolve? At an unusual, high-level meeting, researchers pondered whether music helped our ancestors survive and reproduce or whether it is merely a happy evolutionary accident Seeking the Key to Music

READING,ENGLAND—On a recent fall of Technology, threw down the gauntlet in hardwired into our brains. “A predisposition evening, the lobby of the archaeology build- his book How the Mind Works, when he sug- to engage in musiclike activities seems to be ing at the University of Reading was the gested that music itself played no adaptive part of our biological heritage,” says Ian scene of a strange ritual. Twenty-five re- role in human evolution. Rather, Pinker ar- Cross, a psychologist of music at Cambridge searchers danced in a circle while blowing gued, music was “auditory cheesecake,” a University. He and others point to the work on the ends of differing lengths of rubber byproduct of natural selection that just hap- of University of Montreal neuroscientist tubing. Pedro Espi-Sanchis, a music educa- pened to “tickle the sensitive spots” of other Isabelle Peretz, whose studies of musically tor based in South Africa, had cut the tubing truly adaptive functions, such as the rhyth- challenged neural patients, which suggest such that the notes produced by the pieces mic bodily movements of walking and run- that distinct regions of the brain specialize in spanned two full octaves. Espi- music processing, have made Sanchis encouraged everyone to her a leading opponent of the toot to his or her own inspiration, Pinker viewpoint (Science, but to try not to repeat what oth- 1 June 2001, p. 1636). Indeed, ers were doing. After several min- Cambridge University anthro- utes, to everyone’s delighted pologist Robert Foley argues surprise, the individual notes coa- that the evidence is suggestive lesced into a single pleasing enough that “an adaptive model melody to which the dancers for music should be the default swayed and dipped in rhythm. hypothesis.” This spontaneous musical per- All the same, many re- formance, a highlight of a recent searchers agree that Pinker’s ar- workshop on the evolution of mu- gument represents the key chal- sic and language,* illustrated one lenge to be met: If music is the of the meeting’s key themes: Mu- result of Darwinian natural se- sic, like language, can be a form of lection, how did it evolve, and in communication and coordination what way did it make humans among people. Moreover, music is more fit? At the interdiscipli- an exquisitely powerful way of nary meeting, many talks fo- conveying emotion, a task at which cused on music’s ability to ce- language all too often falls short. ment social bonds. Some re- Yet although few researchers searchers argued that the roots question that human language of music could perhaps be arose by means of natural selec- traced back to “performance tion, presumably because more ac- Scientific bonding. Researchers at a meeting danced and played in step. spaces” created by earlier curate communication helped early species of human. Others see humans survive and reproduce, the evolution- ning, the natural cadences of speech, and the music as a way of getting high with one’s ary significance of music has remained open brain’s ability to make sense of a cacophony peers, again to lubricate human bonding. to debate. The meeting, organized by Read- of sounds. Music, Pinker maintained, is And new studies focus attention on mothers ing archaeologist Steven Mithen and music what the late paleontologist Stephen Jay and infants, suggesting that music might educator Nicholas Bannan, was intended as a Gould called a “spandrel,” after the highly have evolved as a way for parents to soothe first step in setting a research agenda to ex- decorative but nonfunctional spaces left by babies while foraging for food. plore the evolution of music. arches in Gothic buildings. By the end of the meeting, says Peretz, “I In 1997, cognitive scientist Steven But many researchers disagree, arguing felt a consensus around the idea that music Pinker, then of the Massachusetts Institute that music clearly had an evolutionary role. is not only distinct from language but also They point to music’s universality and the has biological foundations.” Yet there was * European Science Foundation Workshop on Mu- sic, Language, and Human Evolution, Reading, ability of very young infants to respond also broad agreement that Pinker’s challenge

U.K., 28 September to 1 October 2004. strongly to it as evidence that music itself is had not been fully answered. CREDITS TO BOTTOM):(TOP GETTY IMAGES; M. BALTER

1120 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org Published by AAAS N EWS FOCUS

Sociability versus sex nia, Los Angeles, and her co-workers pub- it needs to prove, namely why music is Like language, most musical behavior lished a study in Science. After 16 years of needed for bonding and where it got its leaves no trace in the archaeological observing wild baboons, they demonstrated group-stimulating powers.” Merker prefers a record. The earliest undisputed instruments that infants of more sociable female ba- hypothesis that “is driven exclusively by the are flutes made from bird bones found at boons had a higher survival rate (Science, individual advantage of sexual selection.” Geissenklösterle cave in Germany and Istu- 14 November 2003, p. 1231). Pinker, who was not at the meeting and is ritz cave in France, created and played by Foley points out that the apparent fitness now at Harvard University, adds that “uni- modern humans a scant 32,000 years ago. benefit of social cohesion is also the current versality and early development don’t show But the first instruments were probably leading hypothesis for why language itself that music is an adaptation. It just shows made of perishable materials such as bark that music is innate. That’s a necessary con- or bamboo and are not preserved, says dition for something being an adaptation but Bannan. And given the universality of mu- not a sufficient one.” sic today, most researchers assume that its origins extend back much further, possibly Music for the masses even before modern humans arose some For social-cohesion theorists, the challenge 150,000 years ago. “If there is a strong is to explain why singing or dancing en- genetic basis to musicality, then for it to be hanced social bonding—and why that in universally present in the human popula- turn fosters greater fitness and survival. tion it must have been in place more than Robin Dunbar, a psychologist at the Univer- 150,000 years ago,” says Foley. sity of Liverpool, U.K., has suggested that In the workshop’s opening talk, Foley music might have put groups of hominids pointed out that Charles Darwin himself was into a collective endorphin high, making hard put to explain how music made hu- them feel more positively disposed toward mans better adapted to their environment. In their fellow hominids—and thus more likely the end, Darwin concluded that music was to cooperate and survive. Researchers have the result of “sexual selection,” the elabora- long known that listening to music can trig- tion of traits—such as the peacock’s tail— ger the production of endorphins, natural designed to attract a mate and thus ensure opiates that are produced in response to pain reproductive success. Just as some song- or other stress. In a frequently cited 1980 birds sing as part of the courtship process, study by Stanford University neuroscientist Darwin proposed that humans evolved the Avram Goldstein, volunteers who received ability to sing to each other to express emo- injections of an endorphin-receptor blocker tions such as love and jealousy. reported getting considerably less pleasure That theory has some leading propo- when they listened to normally moving mu- nents today, including University of New sical pieces. Mexico evolutionary psychologist Geoffrey Dunbar is well known for his “social Miller, author of The Mating Mind. Miller brain” hypothesis of human evolution, notes that in some bird species, such as which holds that larger hominid brain sizes marsh warblers and nightingales, the male and language both evolved as a response to signals his supposed genetic fitness to the increasing group sizes in our primate ances- female by the sheer number of songs he can tors (Science, 14 November 2003, p. 1160). sing and can reach a repertoire of more than He argues that the endorphin release from 1000 numbers. He argues that music might music may enhance the subjective feeling of have evolved as a way for humans to show bonding, creating stronger social cohesion. off their reproductive fitness. But the sexual He told the attendees of the meeting about a selection hypothesis continues to be a pilot study that he and his students recently minority view among music evolution carried out in English churches. In the study, researchers. “If it was sexual selection, First flutes. These 32,000-year-old flutes are which aimed to look at the effects of music [music] would be a lot more restricted,” the oldest undisputed evidence of music. in a social setting, the endorphin levels of says Foley. “We would see it more in churchgoers who attended Anglican services courtship and less in other activities. Musi- evolved. “So it makes sense to extend it to with and without singing were monitored by cal ability and activity are too widespread.” music and indeed most other activities,” he indirect methods that measured tolerance to Foley and others favor another hypothe- says. The evening performance led by Espi- pain. (Measuring endorphins directly re- sis, which holds that in humans, music Sanchis was a good example of music’s quires an invasive lumbar puncture.) After plays an important role in maintaining so- “ability to be used in group bonding,” adds services, parishioners who had sung were cial cohesion—critical to mounting coordi- psychologist Helen Keenoo of the Open able to endure having a fully inflated blood nated actions—which was essential for ho- University in Milton Keynes, U.K. “Many pressure cuff on their arms for significantly minid survival. Experts in primate behavior people seemed to come away from this ex- longer than those who had not sung. have long assumed that cooperation among perience on an emotional high.” Dunbar stressed that although his own members of a group boosted the survival But others, including Pinker, say the so- study is very preliminary, the overall evi- rates of early hominids and their offspring, cial-cohesion hypothesis suffers from circu- dence suggests that group singing and thus selecting for genes that enhance social lar reasoning. Björn Merker, an expert in an- dancing might have helped bridge what he bonding. But direct evidence has been imal vocalizations at Uppsala University in calls the “endorphin gap” between the lacking—until last year, when anthropolo- Sweden who attended the meeting, says that nonverbal grooming activities of our pri-

CREDIT: FRANCESCO D'ERRICO gist Joan Silk of the University of Califor- the hypothesis “takes for granted that which mate ancestors and the later development

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1121 Published by AAAS 1122 N as far ashalfamillionyearsas far by ago, seeking singing anddancingmightbetraceable back versity ofLondonproposedthatgroup talk, archaeologist Clive Gamble Uni- the of has beenhardtocomeby. In an imaginative evolutionary evidence such but past, mans’ evidence interactionsinhu- group such of archaeological hypothesis mightbetofind other way.” to create any difficult seems that longing strongsenseofbondednessandbe- a very context, hesays, surges create “endorphin they like,” Dunbar says. And inasocial thingsthattriggerthesensations at finding levels.raises endorphin “Humansaregood monkeys of glue many and other primates, shownwhich isthesocial grooming, that language. of A numberofstudieshave features thatcan beconsideredmusical,in- speech hasanumberof Maternal infants. how studied mothers talkandsingtotheir that boostedsurvival. allowing motherstoget on withothertasks bonding withandsoothingbabies, aswell as the meetingthatmusicwasboth to crucial chologist Sandra Trehub. Shesuggestedat involved,of TorontoUniversitysayspsy- ship—that ofmothersandinfants—was human relation- ble specific very a that belonging amongearly hominids, it’spossi- evolve did music If asenseof facilitate to Music andmotherese standing of [early humans’]lifestyle.” “givespeculations us a perceptiveunder- fashion,” sayswho Mithen, feelsthatsuch communicating inamusicalanddancelike am surethatthehominidsatBoxgrove were earlygathered humans tosinganddance. axes mayhave representedaspacewhere that this possibly symbolicdepositofhand grove, Sussex, in U.K. Gamble suggested area, atthe500,000-year-old site of Box- from a butchering used andalllocatedfar tration of321handaxes, manyun- them of mans. Healsopointedtoanunusualconcen- areasoftheseearlyperformance and hu- which hesuggestedrepresentedgathering minid siteofBilzingslebeninGermany, atthe400,000-year-oldstone, unearthed ho- diameter and markedby anvils and bone of several to circles those circles,8metersin made by themaledancers.Hecompared Gamble couldseethecircleinsand danced aroundthem. The next morning, rattles on their legssticks, striking and while men, wearingwomen sataroundafire where watched he which in a performance Namibia, the Makuripeopleofnorthern such activities mighthave takenplace. evidencewhere spaces” for“performance EWS Foryears Trehub andher colleagueshave Gamble’s Although evidence isscant,“I He drew onarecentvisittovillageof way One thesocial-cohesion support to F OCUS 2NVME 04VL36SINEwww.sciencemag.org SCIENCE VOL 306 NOVEMBER12 2004 her baby’sattention. Music to hisears. stream to saliva. Mothersthemselvestook passes easilyhormone blood- the from indicator ofhigher arousallevels,the and levels blood Higher cortisol areareliable ter their mothers spoke them. to sang or levels in the saliva ofbabiesbeforeandaf- theybut subjects, didmeasurethecortisol levels endorphin measure intheirinfant with eventouch rule longer gaze times. efforts tocompensatefor theno- mothers’ their to responded for their part, infants, when they couldnottouchtheirinfants. The —that is,madethemmuchmoremusical— markedly increasedthepitchoftheirvoices hub andNakatafoundthatthewomen theywere toldnottotouchtheirbabies. Tre- as much as theywanted; in asecondsession, mothers were allowed totouchtheirbabies minutes atatime.Duringonesession,the 2 for unteer motherstotalktheirinfants recent study, Trehub andNakataaskedvol- speech.Inanother normally melodic maternal during than singing ing episodes of maternal gaze times wereers. Infant even longerdur- as theyfants watched videosoftheirmoth- measured theresponsesof6-month-oldin- UniversityCatholic shin in Japan, Trehub with Takayuki NakataoftheNagasaki Jun- sidered ameasureofattention. baby spendslookinginonedirection,con- gaze,” that monitor“infant how or a long adultspeechinstudies normal to cooing nal have prefer mater- demonstratedthatinfants ated withtenderness. Trehubothers and slowera ness—and tempo, which is associ- which isassociatedemotionallyhappi- with speech— cluding higherpitchthannormal Trehubco-workersher and to did not try In amorerecentstudy,collaboration in P u b l i s A mother’s songcaptures h e d b y

A A A S hypothesis,” recently by proposed anthro- idea, calledthe“puttingdownbaby the be, might Trehubfavors aspeculativenew gin. what for As their adaptivefunction evidence that theyhave evolutionary an ori- versality oflullabies, Trehubsaid, strong is goes upandthetempodown.” The uni- cultures. across They are emotive: pitch The the meeting.“Andthey similar very sound culture in the worldlullabies,”has told she for Trehub, that’s the whole“Every point. who one ever has sung a baby tosleep.But level of the infant,” Trehubconcluded. singing seemstoberegulate the arousal “Thefunctionofmaternal normal. to tisol levels, which thenquicklyrebounded hand, other the causedaninitialdropincor- on speech, the singingstopped.Maternal was maintainedforatleast25minutesafter levelsa marked decreaseincortisol that were sults caused singing striking:Maternal mouthswithacottonroll. The re- infants’ the saliva samples bygently swabbing their logical adaptations.” that make life most worth livingbio- not are value its on today: us to “Someofthe things survival humanshaslittlebearing of modern whether that adds music was essentialtothe children andadultslistentomusic.”he But he says, itdoesn’t“but explainwhy older explainwhy mothers singtotheirbabies,” that music evolved tosoothebabiesmight his challengehasnotbeenmet.“Theidea sic’s adaptive functions, Pinker, says one, for adaptive function. and thus early ofmusic,alsohadan forms singing, tent withthenotionthatmaternal universal—Trehub is says thatitisconsis- everyone sial—not that“motherese” agrees Falk’s although And hypothesis iscontrover- language. true to prelinguistic forerunner vocalizations, Falk concludes,were the while sleep, to foragingforfood. These baby,her reassure and if not actuallyit put tions, or “motherese,” sothatshecouldcalm by gues, developing melodiousvocaliza- male respondedtothissituation,Falk ar- are toohelplesstodoso. The hominid fe- young mothers at a very infants human age, Unlikebaby who chimps, canclingtotheir whogave tomoreimmatureinfants. birth canal—natural selectionfavored females headstopassthroughthebirth infant for creased—thus makingitmoredifficult as thebrainsizeofearlyin- hominids well as asdatafromfossils.Shearguesthat humans tions ofchimpanzeesandmodern comparisons ofthemother-infant interac- Behavioral Brain and Sciences Tallahassee.in sity pologistFalk Dean ofFloridaStateUniver- Of course,thisisratherobviousany- to Despite thisrangeofsuggestionsformu- Falk’shypothesis, inpressatthejournal –M , isbasedon ICHAEL B ALTER

CREDIT: JOHN CARTER/PHOTO RESEARCHERS INC. Vaccine Policy

those diseases began to steadily increase. Biostatisticians Ira Longini and Elizabeth Immunizing Kids Against Flu May Halloran of Emory University in Atlanta, Georgia, have developed a mathematical mod- Prevent Deaths Among the Elderly el to find the optimal way to distribute flu vac- cine. As they describe in a 2000 Vaccine pa- Increasing evidence suggests that vaccinating schoolchildren can create a “herd per, vaccinating just 30% of schoolchildren in immunity” that indirectly benefits the unvaccinated a community reduces the likelihood of epi- demic spread of flu from 90% to 65%. If half The flu vaccine shortage in the United States rates in all age groups were three times lower the children are vaccinated, the likelihood has had one clear benefit: It has forced a de- in Tecumseh than in a largely unvaccinated drops to 36%; if 70% are vaccinated, the prob- bate about the best vaccination strategy. And nearby town. “There’s a lot of benefit to be ability of epidemic spread plummets to 4%. mounting data suggest that there are much gained by targeting school-age kids,” says “Children are highly connected among them- more effective ways to combat the annual Monto, who chairs a subcommittee of ACIP selves, and they’re connected through adults onslaught of this deadly disease than what that focuses on herd immunity. Monto adds to families and neighborhoods,” says Longini, the country does now. that vaccinating the “frail” elderly may be noting that they’re twice as likely to become The Advisory Committee for Immuniza- less effective because they often do not de- infected as adults. “By vaccinating children, tion Practices (ACIP) of the Centers for Dis- velop a robust immune response. you’re tearing the heart out of that web.” ease Control and Prevention recommends the As was found in the Japanese study, vaccine for healthy infants and people 50 or Longini and Halloran’s models suggest that older, as well as people with chronic illnesses, such a strategy could greatly reduce deaths the groups that suffer the most hospitaliza- among the elderly. If, for example, coverage tions and deaths. But a study published on- of schoolchildren increased from the current line 2 November in Vaccine concludes that 5% to 20%, they predict it would reduce vaccinating school-age children could have a more deaths in the over-65 population than greater impact on slowing the spread of in- increasing their vaccination coverage from fluenza virus and reducing disease. the current 68% to 90% (see graph). Vaccinating a high percentage of people in Given the evidence, why hasn’t herd im- a community can decrease the spread of a munity to flu received more attention? “Flu pathogen and create a “herd protection” that until now wasn’t really a sexy topic,” says extends to the unvaccinated. Just such a phe- Halloran. “And it’s sort of a medical thing nomenon has occurred in two neighboring that you look at protecting people directly.” Texas towns, Temple and Belton, report Pedro Although the results hold for both the in- Piedra, W. Paul Glezen, and colleagues from jected vaccine and the recently licensed in- Baylor College of Medicine in Houston. Since haled one, Monto suspects that the latter the 1998–99 flu season, the researchers have works better. The nasal flu vaccine relies on offered a new, nasally administered flu vaccine live but weakened virus, whereas the injected to everyone between 18 months and 18 years version contains killed influenza. In theory, of age, reaching 20% to 25% of the 20,000 eli- the live vaccine can trigger a broader im- gible children each year. The researchers tallied mune response and may better thwart viral serious flu-related disease in all age groups. transmission because it stimulates immunity For comparison, they analyzed three other at the site where the virus typically enters. communities in which less than 1% of the For mass immunization campaigns, says 39,000 children received flu vaccine. Monto, the spray is also easier to deliver. In Temple and Belton, the researchers The Texas herd-immunity study will con- found that each year they vaccinated school- tinue for two more seasons, and MedImmune age children, serious flu cases in adults 35 Inc. in Gaithersburg, Maryland, the maker of and older were 8% to 18% lower than in the Kid you not. Vaccinating even 20% of school-age the nasal vaccine, has four of its own studies comparison communities. “This translates to children may prevent elderly deaths from flu more under way that will attempt to assess the in- a major reduction in illness,” says Piedra. effectively than increasing elderly vaccination rates. direct benefits of vaccinating schoolchildren. “With the current policy, you only try to con- In another novel strategy, two reports pub- trol mortality. If you want to control flu, our Still more evidence of a herd effect lished online 3 November by the New Eng- hypothesis is to focus on kids.” Earlier stud- comes from Japan. As Thomas Reichert, land Journal of Medicine suggest that the ies showed similar herd protection by vacci- Baylor’s Glezen, and co-authors reported in current vaccine supply can be “stretched” by nating school-age children, but they did not the 22 March 2001 New England Journal of injecting smaller doses under the skin, instead persuade policymakers. U.S. flu-control cam- Medicine, Japan vaccinated 50% to 85% of of intramuscularly. This approach offers paigns “have never really taken this approach schoolchildren during the mid-1970s and “great promise,” wrote Anthony Fauci and the seriously,” says epidemiologist Arnold Mon- 1980s, but the elderly rarely received the late John La Montagne of the National Insti- to of the University of Michigan, Ann Arbor. vaccine. During that period, deaths from in- tute of Allergy and Infectious Diseases in an In 1968–69 Monto and colleagues vacci- fluenza and pneumonia—which mainly kill accompanying editorial. “It’s not going to be nated 85% of the schoolchildren in the small the elderly—dropped by at least 10,000 per a practical solution to the shortage problem town of Tecumseh, Michigan. Virtually no year. As of 1987, parents could exempt their we have this year,” stresses Fauci, “but it

CREDITS: (GRAPH) IRA LONGINI; DEMETRIO CARRASCO/GETTYIIMAGES (PHOTO) adults received the vaccine. Even so, disease children from the program; deaths from could be in the future.” –JON COHEN

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1123 Published by AAAS Molecular Biology

by Linsley’s work and in some cases their own, that belief shifted. “We saw more and RNAi Shows Cracks in Its Armor more unexplained phenomena,” says René Bernards, a cancer geneticist at the Nether- RNAi’s tendency to influence genes and proteins that it’s not designed to target is lands Cancer Institute in Amsterdam. Phillip provoking questions and controversy, as scientists labor to solve the problem Zamore, a biochemist at the University of Massachusetts Medical School in Worcester, A promising new approach to manipulating a subsidiary of the drug giant Merck. “We says his thinking evolved “when I couldn’t genes is showing blemishes as it moves from thought it would be cool,” Linsley recalls, to find a way to disprove Peter Linsley.” Like its glamorous early days to a more nuanced use siRNAs to try to design more targeted many of his colleagues, Zamore now believes adolescence. The technique, RNA interference drugs. The plan: Use siRNAs to knock down that RNAi’s limitations should have been ob- (RNAi), shuts down genes; this braking effect expression of a particular gene that an experi- vious and that to presume such specificity helps reveal a gene’s function and could po- mental compound is already designed to tar- was “incredibly unreasonable.” Genetics, says tentially be used to treat a host of diseases. But get. Then add that compound to the mix, and Zamore, is rarely so neat. a growing number of researchers are learning see if it disrupts other genes as well— Why off-target effects occur remains a that RNAi, which was hailed for its laserlike something that might suggest it’s not targeted matter of debate. One possibility is that specificity by scientists and the press (includ- enough for treating patients. introducing foreign siRNAs into a cell’s ex- ing Science, which anointed it 2002’s Break- isting RNAi system—upon which it relies through of the Year), comes with some unin- for a range of functions, from early develop- tended baggage. In particular, it can hijack ment to protecting the genome’s integrity— genes and proteins it wasn’t designed to tar- risks throwing a wrench into the machinery. get—a potential problem for both basic ge- Soon after scientists began experimentally netics studies and RNAi-based therapies, adding siRNAs to mammalian cells, they some of which are just beginning hu- learned that these cells naturally use man testing. hundreds of so-called microRNAs, Even experts concerned about which are similarly sized and help these so-called off-target effects translate RNA molecules into pro- hasten to point out that RNAi’s fu- teins. MicroRNAs are widely con- ture remains bright. But the issue sidered much less specific than is stirring controversy in the field. siRNAs, frequently targeting Biologists are struggling to deter- sequences that only partly match mine—and agree upon—just how their own. widespread off-target effects are, This has left scientists wondering why they occur, and what can be whether mammalian cells, awash in done to avoid them. Some are fever- microRNAs, are mistaking foreign ishly working to circumvent the prob- siRNAs for more of the same, espe- lem, with early hints of success. cially because both microRNAs and “We don’t know all the rules” of the siRNAs need many of the same enzymes RNAi machinery, says Mark Kay, a pediatri- to function. An RNAi study last year showed cian and geneticist at Stanford University, that this mistaken identity could occur. who’s conducting RNAi animal studies to “There’s probably a fine balance between the treat hepatitis B and C viruses. “My philoso- microRNA pathway and what we’re putting phy is that we move forward with caution, but Eyeing RNAi’s potential. With RNAi trials launch- into cells of animals,” says John Rossi, a we move forward.” ing for macular degeneration (above), researchers molecular biologist at the City of Hope In the late 1990s, scientists discovered the are watching closely to see whether the tech- Graduate School of Biological Sciences in potency of small RNA molecules just 21 nu- nique has any unexpected effects on humans. Duarte, California. cleotides or so in length—some lab- Weak sequence matching between produced, others naturally occurring. Inject- But as it turned out, says Linsley, it wasn’t siRNAs and genes has also been traced to a ing these RNAs, often called small interfering the compounds that were poorly targeted. specific part of the siRNA. That bit, called RNAs (siRNAs), into worms and flies si- The siRNAs were turning down expression the 5´ end, helps govern how an siRNA binds lenced only messenger RNA (mRNA) mole- in multiple genes instead of just one. “The to its target. As Linsley found and others such cules containing a complementary sequence. siRNAs were dirtier than our compounds,” as Zamore confirmed, if that particular piece, That, in turn, blunted expression of the gene says Linsley, whose team was taken aback. about seven nucleotides long, matches a se- producing that messenger RNA. In these or- The pattern persisted, and the researchers fi- quence in another gene, there’s a risk of the ganisms, there was no sign that an mRNA nally concluded that siRNAs could “cross- entire siRNA binding to that gene instead. with a slightly mismatched sequence—with, react” with other genetic targets. After some Increasingly, biologists are turning up oth- say, 17 compatible nucleotides out of 21 in struggle convincing reviewers that the paper er seemingly esoteric details that may also de- the siRNA—could also be affected. was accurate, it appeared in Nature Biotech- termine whether an siRNA shuts down unin- But as scientists moved on to studies in nology in June 2003. tended genes. In the fall of 2003, a group led mammals, the picture changed. One of the RNAi enthusiasts responded skeptically. by Anastasia Khvorova at Dharmacon, a first to see irregularities was Peter Linsley, the After all, they’d trusted for several years that company in Lafayette, Colorado, and another executive director of cancer biology for the small RNA molecules they were crafting led by Zamore, reported in Cell that siRNAs

Rosetta Inpharmatics in Seattle, Washington, were undeniably specific. Gradually, prodded with certain sequences and structures unwind RESEARCHERS INC.CREDIT:PARKER/PHOTO PAUL

1124 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org Published by AAAS N EWS FOCUS

slightly differently—and the pat- life trying to see all 10,000 pro- tern in which they unwind can ul- teins in the cell,” says Sharp, “and timately affect how good they are you’ll never get an answer.” at targeting the right gene. Scientists are quick to add a A year ago Bryan Williams, a caveat: Even if off-target effects cancer biologist at the Cleveland occur, they don’t necessarily af- Clinic in Ohio, identified another, fect the phenotype, or how a cell more controversial kind of off- or animal actually functions. If target effect. In fruit fly cells and phenotype isn’t altered, notes human cancer cells, he found that Zamore, the effects rarely make a siRNAs activated the interferon difference. pathway, which is the body’s first The significance of off-target line of defense against viruses. effects also depends on how RNAi How widespread the interfer- is used: to unearth the function of on response is remains uncertain. mystery genes or as a medical Although some unpublished re- therapy. In the first case, scientists ports of interferon response in an- are getting around the problem by imals exist, “by and large, people applying several different siRNAs, who’ve treated animals with each of which corresponds to a siRNAs have not seen significant different sequence in their gene of interferon induction,” says Phillip interest. That way, if one siRNA Sharp, a biologist at the Massa- prompts an off-target effect that chusetts Institute of Technology in changes a cell’s phenotype, it will Cambridge and co-founder of the be more apparent. company Alnylam, which next When it comes to RNAi- year hopes to begin testing RNAi On target? Hepatitis B is one of the diseases RNAi enthusiasts are based treatment, though, the po- in patients with the eye disease working to disable. tential challenges multiply. The macular degeneration. first clinical trial of RNAi thera- In animals generally, the impact of off- sounds. For the most part, scientists are rely- py—for use in macular degeneration—was target effects isn’t clear. Mice with liver dis- ing on microarrays, which show gene- launched last month by the Philadelphia ease treated with RNAi technology do suffer expression levels, to learn whether their company Acuity Pharmaceuticals. Because toxic effects, says Harvard’s Judy Lieberman, siRNAs are hitting unintended genes; in gen- treatments can be restricted to the eye, the but those are considered more a result of the eral they’re finding that a dozen genes may risk of off-target effects is of less concern. way siRNAs are delivered—in this case, un- be affected by a single siRNA. (Linsley has For other diseases, “it’s unclear how der extremely high pressure, to ensure that recorded on average at least 40.) Still, it’s dif- much of an issue this is going to be,” says they infiltrate liver cells. Lieberman says she’s ficult to gauge how big a problem that is. Stanford’s Kay, whose RNAi work focuses seen no visible evidence of off-target effects in Mismatches provoke a less dramatic change on hepatitis. The disease is a popular her mice, but she is planning to examine the in gene expression than complete matches. choice for RNAi therapies because RNAi animals more carefully. Says Linsley, “You Most gene expression varies by less than can disable the virus. Yet it’s also difficult can’t conclude it’s not there until you look.” twofold when the siRNA doesn’t fully to target the liver without affecting other Looking, though, can be trickier than it match—often not enough to have a substan- parts of the body. In Kay’s view, RNAi tial biological impact on how a cell, or an ani- therapies shouldn’t be viewed differently mal, actually functions. from traditional drugs: “If you give some- But using microarrays to look for body aspirin, they’re going to have off-target effects has one big drawback: They changes in gene expression in specific tis- show only gene expression, not protein levels. sues.” He expects that RNAi clinical trials, If siRNAs are imitating microRNAs, that like all others, will need to home in on the means they’re not affecting DNA directly but lowest effective dose and monitor patient rather are altering how RNA is translated into safety carefully. protein. Microarrays thus might not detect To avoid any effects that may cause changes in protein abundance. “What’s really problems, researchers are chemically modi- important is what’s happening at the protein fying siRNAs to try to stop them from level, and we don’t have a lot of data on that,” glomming onto messenger RNAs they says Linsley. should ignore. Modifications can also make Researchers at Dharmacon and elsewhere the key 5´ bit of siRNAs more sluggish in are trying to see whether microarray results its binding, rendering mismatches less like- correlate with changes in protein levels. At a ly. Linsley and some Dharmacon colleagues meeting last week in Titisee, Germany, Sharp have just submitted a paper on the subject. presented preliminary data from his lab show- “We’ve made a few steps,” he says, declin- ing a 10-fold change in protein levels with ing to be more specific. But “I don’t think only a twofold microRNA difference, the level we’ve completely solved it.” Although off- commonly seen from an off-target effect. But target effects may forever linger as a risk of Cautious but upbeat. Stanford’s Mark Kay doing the kind of broad protein screens that RNAi, he and others say, they hope that hopes that off-target effects won’t derail microarrays today accomplish for genes isn’t they’ll become less of a worry, and soon.

CREDITS (TOP TO CREDITS TO BOTTOM):(TOP CDC/PETER ARNOLD INC.; MARK UNIVERSITY KAY/STANFORD RNAi’s extraordinary possibility. yet possible. “You can spend the rest of your –JENNIFER COUZIN

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1125 Published by AAAS Meeting Society for Neuroscience

SAN DIEGO—From 23 to 27 October, this California coastal city hosted the annual Society for Neuro- science meeting, at which more than 30,000 Brain Cells May Pay the Price for researchers presented data on topics such as sleep A Bad Night’s Sleep problems, addictive anesthesia, and baby talk. It’s enough to keep you up at night: Sleep of papers in the last few years that describe “It’s an incredibly important observa- apnea, a condition in which breathing irreg- what happens in the brains of rodents ex- tion,” says Sigrid Veasey, a neuroscientist at ularities occur during sleep, may kill neu- posed to brief periods of reduced oxygen the University of Pennsylvania in Philadel- rons in brain regions crucial for learning and similar in nature to those experienced by a phia. “The injuries can be compounded by memory, according to research on rodents person with sleep apnea. what would be considered an unhealthy diet, described at the meeting. The findings may It’s not a pretty picture. Gozal’s team has but [is] probably a pretty standard diet for a provide a disturbing explanation for the cog- found that intermittent hypoxia kills rodent lot of people in this country.” brain cells in the hippocampus, a –GREG MILLER key memory center, and inter- feres with a process called long- term potentiation, a strengthen- Anesthesia’s ing of neural connections consid- ered crucial for learning and Addiction Problem memory. The same reduced level of oxygen, kept constant, has lit- Warning: Anesthesiology may be haz- tle or no effect. ardous to your health. According to Mark Gozal has also begun to eluci- Gold, who has spent much of his career in- date some of the molecular vestigating the problem of physicians who events underlying the damage. abuse drugs, data from the state of Florida Intermittent hypoxia stresses show that “every year since 1995, anesthe- brain cells, causing them to pro- siologists were the number one [medical] duce molecules called oxygen specialty for substance abuse or depend- free radicals that are notorious ence.” His group found, for example, that for wreaking havoc on cells and in 2003, anesthesiologists represented less driving them to self-destruct. In- than 6% of all physicians in the state but hibiting certain enzymes in- made up almost 25% of the physicians volved in the stress response can monitored for substance-abuse disorders. save neurons and prevent learn- Although Gold, chief of the McKnight ing deficits in rodents subjected Brain Institute at the University of Florida You snooze, you lose. Rodent research suggests that sleep to hypoxic periods, Gozal’s team (UF), isn’t the first to conclude that anes- apnea can kill brain cells. has found. They also reported at thesiologists are especially susceptible to the meeting, for the first time, abusing drugs, particularly opiate-based nitive deficits often seen in people with that regular exercise—the rat equivalent of compounds similar to those used in gener- sleep apnea. And to make matters worse, walking in the park for an hour a day, Gozal al anesthesia, he has a new explanation. new evidence suggests that adding an un- says—cancels the learning deficits caused The problem isn’t simply easy access to healthy diet to the mix greatly compounds by intermittent hypoxia. drugs, he says. Instead, he and his col- the neural harm caused by disordered sleep. That’s the good news. The bad news is leagues propose that the physicians may In the United States, sleep apnea affects that a diet high in fat and refined carbo- become primed for drug abuse because at least 2% of children, 4% of middle-age hydrates appears to magnify the deleterious they chronically inhale small amounts of adults, and 10% of older adults, according to effects of intermittent hypoxia. The diet anesthetics that sensitize the brain’s re- conservative estimates, and it is even more alone caused a mild learning impairment ward pathways. Indeed, at the meeting, common in obese people. The cognitive when rats were tested in a water maze, and it Gold’s team reported finding traces of in- problems that result, including hyperactivity, reduced the level of the activated form of a travenously delivered anesthetics in the air wandering attention, and learning deficits, protein called CREB in their hippocampi. of operating rooms. were long thought to stem solely from the CREB plays an important role in memory When early anesthesiologists depended fatigue that follows a bad night’s sleep, says consolidation and neuron survival, so it is a on gases such as ether and chloroform, Gordon Mitchell, a respiratory neurobiolo- good general marker of hippocampal health, secondhand exposure was a serious prob- gist at the University of Wisconsin, Madi- Gozal says. Pairing the high-fat, refined- lem. But better ventilation and increased son. “The concept that you’re causing spe- carbohydrate diet with intermittent hypoxia use of intravenous drugs reduced such ex- cific damage through cell death in particular had a synergistic negative effect on learning posures. As for the issue of drug access, areas of the brain is pretty new,” he says. and activated CREB. “It’s a major disaster hospitals have gone to great pains to safe- Mitchell and others credit their colleague for the brain,” Gozal says. If the finding guard their medications. Even so, anesthe- David Gozal with much of the work that has holds true for humans, he adds, that would siologists continue to have much higher led to the unanticipated realization. Gozal, a be especially troubling because sleep apnea rates of opiate-related substance abuse pediatric researcher at the University of and unhealthy diets are a common combo than other physicians with similar access.

Louisville in Kentucky, has published a slew for many people with obesity. From his review of records from Flori- CREDIT: CORBIS

1126 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org Published by AAAS N EWS FOCUS Those follow-up studies will be needed break down a stream of speech into individ- to convince anesthesiologists and others ual words. That skill, in turn, is necessary who are intrigued by Gold’s hypothesis. for assigning words to objects, creating “He’s a guy who thinks outside the box,” more complex sentences, and so on. says Robert L. Dupont, president of the In- Recent work in Kuhl’s lab bears this out. stitute for Behavior and Health in Her team has been using electroencephalo- Rockville, Maryland, and former director gram (EEG) electrodes to monitor the brain of the National Institute on Drug Abuse. activity of 7-month-old infants, who are just “But it’s hard for me to imagine that the at the cusp of the change in phoneme per- doses people are getting this way are hav- ception. The babies listened to a recorded ing any biological effect. I’m ready to be voice repeat a single phoneme several times persuaded, but I’m skeptical.” before switching to another phoneme. If the –JOHN TRAVIS baby caught the switch, a blip appeared in the EEG record. This evoked related poten- tial (ERP) is a standard indicator that the Listen, Baby brain has picked up something new. The re- searchers tested the babies’ ability to dis- How quickly babies home in on the sounds criminate both native and non-native of their native language during their first phonemes and then followed up with a bat- year may predict how quickly they learn tery of language tests at 14, 18, 24, and Breathing problem. IV-delivered anesthetics new words, string together complex sen- 30 months of age. may escape into the air and affect physicians. tences, and acquire other language skills as The ERP recordings revealed that in- toddlers. The new research, presented in San fants who at 7 months of age were good at da’s Impaired Professional programs, Gold Diego, helps pin down a milestone in lan- native phoneme discrimination tended to has found that anesthesiologists who abuse guage development and may shed light on be bad at non-native phoneme discrimina- drugs tend to start much later in life than why the ability to pick up a new language tion, and vice versa. This fits with the other addicts, who typically experiment with wanes with age. “neural commitment” theory proposed by drugs during their youth. Anesthesiologists When it comes to language, babies are Kuhl several years ago, says Mirella also tend to relapse unless they change pro- “citizens of the world,” Patricia Kuhl of the Dapretto of the University of California, fessions, says Gold. University of Washington, Seattle, said in Los Angeles. Kuhl’s work suggests that Having conducted research on whether a lecture here. In the early 1990s, her team “the more your brain gets committed to secondhand smoke sensitizes brain reward found that 6-month-old infants naturally picking up what’s relevant in the first lan- pathways—children of smokers are much possess a language skill far beyond the guage you’re exposed to, the more you’re more likely to smoke—Gold wondered reach of adults: They can distin- tuning out distinctions that are relevant in whether secondhand anesthesia might be at guish all the sounds of all the other languages,” Dapretto says. work. He teamed up with several UF anes- world’s languages—about 600 The follow-up studies suggest that the thesiologists to answer that question. As a consonants and 200 vow- brain’s commitment to native first step, they used a mass spectrometer to els. By the end of their speech sounds provides the foun- examine the air in operating rooms during first year, however, dation for later language learn- cardiac bypass surgeries in which fentanyl, babies begin to spe- ing. Although all the children in an opiate many times more potent than cialize. As they be- Kuhl’s study tested in the nor- morphine, and a nonopiate anesthetic, come better at recog- mal range, the ones who did propofol, were administered intravenously nizing the basic ele- best at native phoneme dis- to patients. Both compounds were found in ments, or phonemes, crimination at 7 months the operating room air and at higher con- of their native lan- scored higher at later centrations in the space between the anes- guage in all their times on all language thesiologist and patient, Matt Warren, a acoustic variations- measures, including graduate student with the Florida team, re- —learning to lump number of words pro- ported in San Diego. /o/ as pronounced by duced, duration of utter- In another test, the researchers found mom together with /o/ ances, and sentence that volunteers given fentanyl exhale it. In as pronounced by grand- complexity. lengthy operations such as cardiac bypasses, pa and Bugs Bunny, they “It’s extremely inter- anesthesiologists “could be breathing anal- lose the ability to distin- esting [that] you can look gesics and anesthetics for 8 hours,” specu- guish phonemes in other at infants and learn some- lates Gold. languages. By about thing really important The Florida team is now collecting 11 months, for ex- Baby talk. An infant’s level of speech discrimi- about their future blood samples of anesthesiologists during ample, certain vow- nation predicts language skills. learning,” says April operations to see if fentanyl is present. els that sound dis- Benasich, a cognitive They also intend to expose rodents to the tinct to Swedes start to sound the same for neuroscientist at Rutgers University in same air concentrations of the anesthetics a baby born to an English-speaking family. Newark, New Jersey. Benasich says the as found in the operating rooms and test Kuhl and others have argued that this findings add to evidence, including work whether those animals are more suscepti- change in speech perception is an essential from her own lab, that it may be possible to ble to drug addiction and develop changes step in language learning. They contend that screen young infants to identify those likely in brain regions involved in the rewarding babies need to be adept at identifying native to need extra help with language learning.

CREDITS: GETTY IMAGES aspects of drugs. phonemes, for example, before they can –GREG MILLER

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1127 Published by AAAS RANDOM SAMPLES Edited by Constance Holden

The Definitive Mozart

A filmmaker hopes to use DNA to settle a long-standing mystery: Does an Austrian foundation really have Wolfgang Image not Amadeus Mozart’s skull? Researchers have already discreetly unearthed bones of several Mozart family members for a available for documentary celebrating the composer’s 250th birthday in online use. January 2006. In the coming months, they will attempt to compare DNA samples taken from the bones to a sample from the skull, held by the International Mozarteum Foundation in Salzburg. Tiwanaku goblet The skull was supposedly unearthed by a gravedigger several years after the 35-year-old composer was buried in an unmarked Life in the pauper’s grave in December 1791. It passed through several hands before arriving at the Mozarteum Foundation in 1902. Decades of historical, dental, and even Ancient Andes facial reconstruction studies have been frustratingly inconclusive, says foundation Archaeologists have discovered a 1000- director Stephan Pauly. year-old trove of finely painted and Journalist Burgl Czeitschner, who proposed the project, says the team has beautifully sculpted drinking vessels and unearthed bones thought to be those of Mozart’s maternal grandmother and his figurines on an island in Bolivia’s Lake niece—both of whom would carry the same mitochondrial DNA as the composer. If Titicaca. Some of the jugs, 20 of which the samples match, Czeitschner says, then scientists might be able to use the skull to are unbroken, are shaped like human learn more about how Mozart lived and the unresolved mystery of how he died. (One heads and animals, offering a rare persistent rumor has him being poisoned by a jealous rival composer.) Czeitschner glimpse of the Tiwanaku culture. says producers plan to keep a tight wrap on any results until the film is broadcast. The finds, announced last month by Antti Korpisaari and Martti Pärssinen of the University of Helsinki working with Bolivian extremely valuable,” says Deborah Eileen convention on 2 September through colleagues, also included the shards of Blom of the University of Vermont, Election Day, the odds on President Bush’s hundreds of vessels that priests may have Burlington. The ceramics also show jewelry eventual victory never dropped below 50% ceremonially tossed into a pit. The work- and tattoos. on the commercial Tradesports.com Web manship represents “some of the very site. On election eve, a $1 contract for highest artistic achievements of Tiwanaku Bush to win closed at 53 cents on the potters ever discovered,” says Korpisaari. Betting on Bush dollar, meaning he was believed to have a Because textiles haven’t survived in In a year when pre-election polls 53% chance of winning. the moist Andean highlands, archaeologists fluctuated, election markets (Science, On the Iowa Electronic Market, run by have little information on how the people 30 July, p. 603) managed to send a the University of Iowa, Bush consistently were clothed. “To actually see what people relatively consistent message. led in the betting. A late Kerry surge might have looked like while alive is From the end of the Republican brought the Democrat up to only 49.5% in the “vote share” market—a 1.5% overestimate (as calculated for a two- AIDS Art in Paris way race), which was nevertheless within the market’s historical error rate. And This anti-AIDS there was no need to sweat all night mural photo- over swing states Florida and Ohio: graphed in Election followers had only to log on to South Africa is Tradesports.com to see that they were part of a travel- pegged in the GOP camp. ling photography According to Iowa accounting professor exhibition, Joyce Berg, the Iowa market stayed “Sciences au steadier and closer to the final results Sud,” focusing throughout the year than polls, which on French underestimated Bush’s final vote share research efforts before the Republican convention, then in the Southern overestimated it until the first debate. Hemisphere, Even after the last debate, “the polls were which opened in all over the map” compared with the Paris last month. markets, says Berg. CREDITS (TOP TO CREDITS TO BOTTOM):(TOP ANTTI KORPISAARI; ARCHIVO ICONOGRAFICS, S.A./CORBIS; ELIZABETH ANTHEAUME/IRD DELIRY

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1129 Published by AAAS RANDOM SAMPLES PEOPLE Edited by Yudhijit Bhattacharjee

IN BRIEF IN PRINT GAVI head. British health policy Paleo buffs. Nothing gets between Fred expert Julian Lob-Levyt is the Ransdell and his fossils—at least not when new executive secretary of the he’s posing as Mr. March for a calendar put Global Alliance for Vaccines and out by the Dallas Paleontological Society. Immunization (GAVI), based in “It’s supposed to be educational, attractive, Geneva, Switzerland. He succeeds and funny,” says Ransdell, an amateur pale- Tore Godal, who has held the ontologist and president of the society, post since the alliance was which hopes the proceeds will finance two formed in 2000. GAVI, which is a graduate scholarships awarded each year. partnership of government The calendar features “revealing but agencies and private foundations tasteful” photos of local fossil enthusiasts, worldwide, and its financing men and women, posing in the field with arm—The Vaccine Fund—have nothing but a smile and their favorite so far given out $429 million for specimens. For their part, Ransdell and his vaccinations in more than 70 son Brent are seen toting a cast of the countries. Lob-Levyt has worked femur of a sauropod called Pleurocoelus at the United Kingdom’s Depart- found at a quarry about 150 kilometers ment for International Develop- south of Dallas. In the background lurks a ment since 1998 and is a senior predator from that era, 110 million years policy adviser to the executive ago, which is actually the image of a dino director of the Joint United Na- sculpture superimposed on the photo. tions Programme on HIV/AIDS. To order a copy, visit dallaspaleo.org.

New NPA chair. Keith Micoli, a postdoc in pathology at the University of Alabama, DEATHS City. The 61-year-old micro- Birmingham, has been named biologist was waiting in a chair of the National Post- Microbe expert. John La passport line at the Mexico doctoral Association for 2005, Montagne, deputy director of City airport when he collapsed succeeding founding chair Carol the National Institute of Allergy suddenly. NIAID officials say he Manahan. The 2-year-old and Infectious Diseases (NIAID), had no serious health problems, association has also elected five died on 2 November en route and the cause of death is as executive board members for 2- to a Pan American Health yet unknown. year terms starting 2005. Organization meeting in Mexico Born in Mexico, La Montagne JOBS Tackling big questions. University of Chicago cosmologist Edward Kolb has been chosen to head the new Particle Astrophysics Center at Fermi National Accelerator Laboratory in Batavia, Illinois. The center will bring together some 40 scientists from Fermi’s experimental and theoretical astrophysics groups, along with new worked at NIAID on a wide hires, to focus on the growing overlap between range of infectious diseases astrophysical observations and high-energy including influenza, malaria, particle physics experiments. It will analyze and tuberculosis. “He had an findings from projects such as the Sloan Digital encyclopedic knowledge of Sky Survey and the Pierre Auger Cosmic Ray microbiology,” says Anthony Observatory to investigate questions on dark Fauci, La Montagne’s boss matter and the big bang. and longtime friend. “And “There are so many new ideas and so much he had to be one of the new data,” says Kolb, 53, who came to Fermilab finest human beings anyone in 1981 and helped found its theoretical astro- of us had ever met.” physics group. Kolb hopes to attract private La Montagne is survived by funding in addition to in-house support for the his wife of 35 years. new center, whose debut next month features a lecture by Nobelist Riccardo Giacconi and a Got any tips for this page? E-mail workshop on galaxy clusters. [email protected] CREDITS (TOP TO CREDITS TO BOTTOM):(TOP NIH/NIAID; FERMILAB/VMS

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1131 Published by AAAS Letters to the Editor Letters (~300 words) discuss material published LETTERS in Science in the previous 6 months or issues of general interest. They can be submitted through the Web (www.submit2science.org) interview request that it was a sensitive Large Binocular Telescope (LBT). Those or by regular mail (1200 New York Ave., NW, juncture in the appropriations process and quoted in the article suggest that “RC was Washington, DC 20005, USA). Letters are not to talk about our budget at that time would abandoning its historical mission … of acknowledged upon receipt, nor are authors have been counterproductive. advancing the natural sciences through early- generally consulted before publication. Fortunately, NASA has a brighter future career awards and small grants that foster Whether published in full or in part, letters are than the one portrayed in Lawler’s article. innovative research and teaching.” As current subject to editing for clarity and space. By pursuing the goals of our Vision for and former members of the RC grants advi- Space Exploration, we expect to signifi- sory committee, we strongly disagree with The Future of NASA cantly advance our scientific frontiers. these opinions. AL DIAZ The grant programs at RC are stronger AMONG THIS YEAR’S MOST SIGNIFICANT Assistant Administrator for Science, National than ever and making a vital difference to science findings was the Mars Exploration Aeronautics and Space Administration, 300 E 150 to 200 annual recipients. Total annual Rover Opportunity’s discovery that Street, SW, Washington, DC 20546–0001, USA. expenditures on grants at RC have steadily Meridiani Planum was once subsumed increased each year, from $2.5 million in under an ancient salty sea. Through 1981 (the year before Schaefer took over) missions like Cassini-Huygens and to over $6.2 million in 2003. RC grant Genesis, and the Spitzer Infrared telescope, Research Corporation programs include a long-standing program NASA continues to conduct cutting-edge and John Schaefer of research support at undergraduate insti- science. I don’t feel Science’s readers were tutions (Cottrell College Science Awards), told any part of that story in Andrew I WAS DISAPPOINTED IN JEFFREY MERVIS’S a new program for beginning faculty who Lawler’s article “Rising cost of shuttle and article “Shooting for the stars” (News Focus, excel at both research and teaching Hubble could break NASA budget” (News 27 Aug., p. 1231), about Research (Cottrell Scholar Awards), and a program of the Week, 24 Sept., p. 1882). Corporation (RC) and its retiring president, for established scientists seeking to explore NASA is studying the many options for John P. Schaefer. There are many debatable new research areas or help reestablish a potential robotic servicing mission of the points in Mervis’s piece, but one allegation federally funded research programs Hubble Space Telescope. The cost estimates misleads the scientific community even more (Research Opportunity Awards). mentioned in Lawler’s article are premature. than it might damage Research Corporation. We have watched Schaefer at close We continue to search for the best possible Mervis quotes Joan Valentine as saying “I range as he strongly supported and solution in servicing Hubble. It is too early don’t think that John [Schaefer] ever cared nurtured the RC grant programs. The to place a price tag on this mission. very much about the grants programs; his country’s scientific research community passion was astronomy.” That statement is as has gained consistent, essential grant parochial as it is false. In the 12 years that I support from RC over the past 22 years. have worked closely with Schaefer, he has DON DEARDORFF,1 PETER DORHOUT,2 repeatedly demonstrated his passion for the NANCY HAEGEL,3 BRENT L. IVERSON,4 advancement of all science. MICHAEL MORRISON,5 JACK PLADZIEWICZ,6 Schaefer has always characterized the MATS SELEN,7 THOMAS D. TULLIUS,8 grant programs as core to the foundation’s TIMOTHY S. ZWIER9 mission. From 1992 to 2003, RC spending 1Occidental College, Los Angeles, CA 90041, USA. on these programs had more than doubled 2Colorado State University, Fort Collins, CO 80523, and accounted for 68% of total RC expen- USA. 3Pacific Grove, CA, USA. 4University of Texas ditures. The community we serve needs to at Austin, Austin, TX 78712, USA. 5University of understand that RC is committed to Oklahoma, Norman, OK 73019, USA. 6University academic research and is ready—as of Wisconsin, Eau Claire, Eau Claire, WI 54702, always—to provide funding to faculty USA. 7University of Illinois at Urbana-Champaign, doing cutting-edge research in the physical Urbana, IL 61801, USA. 8Boston University, Boston, sciences through its regular programs. MA 02215, USA. 9Purdue University, West RAY KELLMAN Lafayette, IN 47907, USA. Vice President, Research Corporation, 4703 East Camp Lowell Drive, Suite 201, Tucson, AZ 85712, Hubble Space Telescope USA. E-mail: [email protected] The File-Drawer There were other inaccuracies in the Problem, Revisited article. First, I have not assigned managers WE READ WITH INTEREST THE RECENT PROFILE in the Science Directorate to “find $400 of John P. Schaefer (“Shooting for the stars,” IT HAS BEEN CONTENDED THAT DRUG million in cuts so that the space shuttle J. Mervis, News Focus, 27 Aug., p. 1231), companies have hidden, in the “file could resume flying.” On the contrary, president of the Research Corporation (RC), drawer,” the results of unsuccessful clinical NASA is forging ahead, hopeful that the oldest scientific research foundation in the trials while publishing the results of more Congress will enact the President’s full country. An objective reading of the article successful trials (1). As others have pointed budget request. leaves the distinct impression that the grant out, no one’s interest (drug company, Second, I never declined an interview programs of RC (www.rescorp.org) were doctor, patient) is served by the marketing

CREDIT: NASA for the article. I said at the time of the taking a back seat to large projects such as the of ineffective drugs, but nevertheless, the

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1133 Published by AAAS L ETTERS scientific enterprise is set back when full will be administratively difficult. results are not published. The remedy Epidemiological studies undertaken in proposed is to make all clinical trials data connection with some regulatory proce- available (“The old file-drawer problem,” dures or a court action can be subject to D. Kennedy, Editorial, 23 July, p. 451); required disclosure. Many and probably indeed, some companies are already plan- most, however, are undertaken in pursuit of ning to post their clinical trials data. a problem that interests the investigator. It If there is minor mischief in the drug might be useful, as Young suggests, to have company file-drawer, there is major all that stuff out there. Alas, we don’t have Where can you review new mischief with many epidemiology studies. a process. scientific books, et cetera? Although there are legitimate reservations DONALD KENNEDY (2), why shouldn’t epidemiologists also open their data sets (3)? Arguably, the scientific need is greater than that for clin- ical trials. Whereas clinical trials test Oxide Surface Films on prespecified, well-defined hypotheses, in Metal Crystals epidemiological investigations, hundreds to thousands of questions may be asked, but IN THEIR REPORT “SAMPLE DIMENSIONS only those results that pass the minimal influence strength and crystal plasticity” threshold of “statistically significant, P < (13 Aug., p. 986), M. D. Uchic et al. 0.05” are reported. Statistical analyses are demonstrate that the dimensions of often not adjusted to reflect multiple micrometer-sized metal crystals have testing, even though procedures are readily dramatic effects on the properties of the available (4). If hundreds to thousands of crystals. However, they do not mention a results are locked in the epidemiology file critical factor; namely, that oxide films drawer, even readers who wish to make the were present on their specimens. All metals adjustment for themselves cannot. The oxidize. Therefore, except for short times worst-case scenario is that the vast majority in ultra-high vacua, they have oxide (or of results reported from epidemiology hydroxide) surface films. These films have studies are false positives. negligible effects on the behaviors of most To cite (but not single out) only one macroscopic metals. As sizes decrease, example of epidemiology file-drawer however, they become increasingly impor- cases, antibiotic use was associated with tant. The necessity for shearing through Books et al. at breast cancer (5). The antibiotic–breast these films may well account for the www.sciencemag.org/books cancer association was culled from a very increase in jaggedness of the stress-strain large search process. If analyses were curves observed by the authors with Read Science’s weekly reviews of adjusted to reflect the number of questions, decreasing size. Also, this interpretation is current books in all fields of science there might well be no significant associa- consistent with the lack of a “size effect” and place orders online. Peruse tion. The central point, however, is that observed for the nickel-based superalloy, as past reviews sorted by title, author, keeping these results in the file drawer well as the large glide offsets. reviewer, and date of publication. prevents others from doing alternative JOHN J. GILMAN analyses, and again the loser is the scien- Department of Materials Science and Engineering, If you are looking for tific enterprise. University of California, Los Angeles, 6532 Boelter recommendations on print, S. STANLEY YOUNG1 AND HEEJUN BANG2 Hall, Los Angeles, CA 90095, USA. audiovisual, and electronic learning 1National Institute of Statistical Sciences, Research tools – check out Science Books Triangle Park, NC 27709, USA. 2Department of & Films Online. Biostatistics, University of North Carolina at Response Not a member of AAAS? Sign up Chapel Hill, Chapel Hill, NC 27514, USA. GILMAN POINTS OUT THAT A THIN SURFACE FILM today, for Science et cetera, References (oxide or hydroxide) on a micrometer-sized 1. L. Abboud, “Lilly plans broad access to results on its crystal could affect the sample flow books et al., and other benefits: drug trials,” Wall Street J., 3 Aug. 2004, pp. B1–B2. www.aaas.org/join 2. E. Marshall, Science 290, 2829 (2000). behavior, especially as the sample size 3. Board on Life Sciences, Sharing Publication-Related decreases. He raises an important point. Data and Materials: Responsibilities of Authorship in However, it is not entirely clear to us how a the Life Sciences (National Academies Press, Washington, DC, 2003) (see www.nap.edu/books/ surface film could contribute significantly 0309088593/html/). to the strength under progressively rising 4. P. H. Westfall, S. S. Young, Resampling-Based Multiple stresses, without also leading to severe Testing (Wiley, New York, 1993). strain localization. High-resolution scan- 5. C. M. Velicer et al., JAMA 29, 827 (2004). ning electron micrographs of deformed samples generally show slip traces at many Response points along the sample gage length. One I THINK MOST WOULD DISAGREE WITH YOUNG’S might expect much more slip localization characterization of the unpublished clinical in these samples if oxide films were acting trials problem as “minor mischief.” His as strong barriers to deformation. larger point, about epidemiological trials, is The “jaggedness” of the stress-strain well taken—but forcing these out of hiding response that Gilman mentions may be

1134 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org Published by AAAS L ETTERS simply the result of mechanical testing at The underlying problem is that evolu- the micrometer scale. A slip event in a tion worries about long-term survival of it takes micrometer-sized crystal will be more the species and not about protein stability clearly resolved than if the same event or structural perfection. As stability was both sides of occurred in a millimeter-sized sample. never nature’s goal, it is not difficult for us In addition, Gilman indicates that the to make proteins more stable by mutagen- the brain. lack of a “size effect” in nickel-based esis. And Kuhlman et al. have shown that superalloys is consistent with the interpre- it also is no longer difficult to design a tation that surface films are affecting the stable globular protein. observed mechanical behavior. Although Around 1990, the European Molecular this is open to interpretation, we have Biology Laboratory organized several “ab recently learned through additional testing initio protein design courses” (1, 2). The that sample size effects of the type participants used the same principles as discussed in our Report do occur for a Kuhlman et al. These designs suffered nickel-based superalloy at sample sizes from databases that were too small, primi- that are 5 µm in diameter and smaller (1). tive software, insufficient CPU power, too Finally, significant sample size effects little time, and lack of experience. Later have recently been observed in testing of we sent the sequence of one of the designs 1-µm-diameter single crystals of pure gold to the first CASP modeling competition (2), using the same fabrication and testing (3, 4). All predictors found the correct fold methodology. The magnitude of the size- with such high confidence levels that they dependent strengthening is similar to the worried that this so-called “mystery CALL FOR ENTRIES results for pure nickel. However, gold does protein” might be a joke. We synthesized not form a native oxide. Clearly, more the mystery protein, and CD and NMR Science &Engineering research is needed to positively establish data showed the right amount of alpha- the connection between sample size effects helix. But it surely did not fold properly, Visualization Challenge to either changes in deformation mecha- which shows that even when software tells nisms or extrinsic effects such as surface you that things are perfect, it still can be When the left brain collaborates with case hardening. very wrong. And that makes Top7 very the right brain, science merges with MICHAEL D. UCHIC,1 DENNIS M. DIMIDUK,1 special. art to enhance communication and JEFFREY N. FLORANDO,2 WILLIAM D. NIX3 GERT VRIEND understanding of research results— 1Air Force Research Laboratory, Materials & CMBI, RU, Toernooiveld 1, Nijmegen 6525 ED, illustrating concepts,depicting Manufacturing Directorate, Wright-Patterson Air Netherlands. E-mail: [email protected] phenomena, drawing conclusions. 2 Force Base, OH 45433–7817, USA. Lawrence References The National Science Foundation and Livermore National Laboratory, Livermore, CA 1. C. Sander, G.Vriend, Eds., PRODES90 Protein design on Science, publishedby the American 94550, USA. 3Department of Materials Science computers, EMBL Biocomputing Technical document 6 (European Molecular Biology, Laboratory, Association for the Advancement of and Engineering, Stanford University, Stanford, CA Heidelberg, Germany, 1991). Science, invite you toparticipate in 94305–2205, USA. 2. C. Sander et al., Proteins 12, 105 (1992). the annual Science andEngineering References 3. See http://predictioncenter.llnl.gov/. Visualization Challenge. The competition 1. M. D. Uchic, D. M. Dimiduk, in preparation. 4. Fifth Meeting on the Critical Assessment of recognizes scientists, engineers, visu- 2. J. R. Greer, W. C. Oliver, W. D. Nix, presentation at the Techniques for Protein Structure Prediction, E. E. Lattman, Ed., Proteins 53 (issue 6, suppl.), 333 (2003). alization specialists,andartists for 2004 Fall Minerals, Metal, and Materials Society producing or commissioning innova- Meeting, New Orleans, LA, 26 to 29 Sept. 2004. tive work in visual communications. ENTRY DEADLINE: Protein Design: CORRECTIONS AND CLARIFICATIONS: May 31, 2005 AWARDS CATEGORIES: News of the Week: “Researchers build quantum Photos/Still Images,Illustrations, Quo Vadis? info bank by writing on the clouds” by C. Seife (22 Oct., p. 593). The name of Klaus Mølmer, of the Explanatory Graphics,Interactive IN THEIR LETTER “PREDICTING PROTEIN University of Aarhus in Denmark, was misspelled. Media, Non-interactive media structures accurately” (11 June, p. 1597), COMPLETE INFORMATION: M. von Grotthuss et al. discuss the ab www.nsf.gov/od/lpa/events/sevc initio–designed Top7 protein (“Design of a 2004 Presidential Forum: (1 Oct., p. 46). Some novel globular protein fold with atomic- readers have pointed out that an error in the Awards in each category will bepublished wording of a question on creationism could be in the September 23, 2005issue of level accuracy,” B. Kuhlman et al., Research interpreted to suggest that Science believes intel- Articles, 21 Nov., p. 1364). The discussion Science and Science Online and ligent design and creationism are scientific displayed on the NSF website. in this Letter and in the Response by P. critiques of evolutionary theory. That is not the Bradley et al. (p. 1597) revolves around the case. The question should have read: “Should difference between predicting a “real” and a ‘intelligent design’ or other unscientific critiques designed structure. Obviously, a fold that of evolutionary theory be taught in public schools?” We regret the error. has never been seen before cannot be predicted by software that compares with known folds in the Protein Structure News Focus: “Finding reactions in a haystack: try ‘em all, see what works” by R. F. Service (10 Sept., Database (PDB), nor necessarily the fold of p. 1558). In the final panel of the diagram, the red Accept the challenge. a sequence for which PDB files with dot should have indicated reactants A1 and B4 exactly that fold are available. instead of A3 and B2. Show how you’vemastered the art of understanding. www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 Published by AAAS BOOKS et al. HISTORY OF SCIENCE Solid Tools for Visualizing Science Julie K. Brown

he general public has long been fasci- three-dimensional models: as research tools nated with three-dimensional models, (“epistemic objects”) in the making of knowl- Tfor what they represent as well as for edge, as mediators of information for com- their sheer beauty and ingenious construc- munication, and as visual objects for public tion. However, until recently such objects display. Herbert Mehrtens points out that have received little notice from historians of physical models were research tools for some science and medicine. Now, with increasing German mathematicians such as Felix Klein attention being paid to the visual and materi- but only during a short window of time, in the al culture of science, scholars finally have 1890s. However, their use, Mehrtens notes, as begun to examine three-dimensional models teaching tools, indicators of professional sta- for their part in “knowledge production.” tus, and aesthetic forms continued well into Models: The Third Dimension of the 20th century. In molecular Science, a collection of fascinat- Models biology the story is more intri- Mathematical model, 1986. The four plaster ing essays that grew out of a The Third Dimension cate and layered, as Soraya de models (here, Clebsch’s diagonal surface) reis- 1998 symposium in London of Science Chadarevian makes clear. Three- sued to accompany (2) served as objects of hosted by the Wellcome Institute Soraya de Chadarevian dimensional modeling was al- beauty rather than research tools. for the History of Medicine and and Nick Hopwood, Eds. ready an integral tool and an the Science Museum, stands as Stanford University established research practice for well as the growing rift between scientific the most important contribution Press, Stanford, CA, crystallographers when, in the authority and the demands of popular en- to this topic to date. 2004. 482 pp. $65, 1950s, molecular biologists tertainment in the early 20th century. In bringing together the £45.95. ISBN 0-8047- adapted it for their studies of pro- The processes of crafting and constructing work of 17 international histori- 3971-4. Paper, $24.95, teins. For publication purposes, three-dimensional models provide a focus for ans of science who have previ- £17.50. ISBN 0-8047- however, these models had to be several authors. Nick Hopwood discusses the ously written on their respective 3972-2.Writing Science. transposed back into two dimen- wax embryological models created by Adolf topics, the volume offers an op- sions by specialized illustrators. and Friedrich Ziegler, drawing on his exten- portunity to view both the continuity and The life of these three-dimensional research sive documentation of their studio (1). the diversity in the uses of three-dimension- models was further extended as they ap- Besides offering excellent descriptions of the al models. This finely focused theme and peared in displays at lectures and exhibitions, process for making wax anatomical models the historical specificity of each author’s entered museum collections, and even were and Felice Fontana’s 18th-century collection contribution sustain and propel the reader featured in television programs, each time en- in Florence, Renato Mazzolini also redirects through this rich, provocative collection. tering a new cultural context where their attention to Fontana’s less-known but excep- Unraveling the how, what, where, and why meaning was redefined. tional “demountable models.” These wooden of three-dimensional models that represent Rationalizing three-dimensional models models, which Fontana developed late in life, aspects of archaeology, medicine, chem- as educational tools has been a recurrent epitomized his growing belief that learning istry, biology, mathematics, and economics theme in their history, because of the fluc- by touch and assembling physical compo- does not necessarily produce a consensus. tuating values attributed to tactile and spa- nents superceded the purely visual experi- But it does show, as editors Soraya de tial experiences in the learning process. As ence. Thomas Schnalke, who has written Chadarevian and Nick Hopwood point out, James Secord points out, it was the ration- widely on 19th-century moulages (clinical that models operate on many levels within al education theories of Johann Pestalozzi “pictures of diseases in wax”), gives the social and cultural “networks of production combined with the Victorian commercial clearest and most layered discussion of and communication.” Although the chapters capitalism of the mid-19th century that model-making as a socio-cultural process. are chronologically arranged—with the fo- were responsible for the famous Crystal More than a modeling technique, moulage cal examples drawn from the 18th century Palace (Sydenham) display of prehistoric was a process of interaction between the through the mid-20th century—the individ- animal models. However, Secord finds that “mouleurs,” skilled in taking direct molds of ual contributors move well beyond fixed the enterprise, “the apotheosis of a short- diseased surfaces; the patients, whose partic- historical frameworks. Two final commen- lived conjunction,” ultimately failed in its ipation and presence “never disappeared”; taries provide additional context to the intention of bringing the sciences together and the supervising physician, for whom the topic: James Griesemer offers a philosophi- in a display format popular with the public. moulage served as a clinical record that en- cal perspective on models and Ludmilla Documenting a more complex version of hanced professional status. Jordanova discusses the relation of models this story, Lynn Nyhart discusses the con- The physical disappearance of many of to visual culture in general. trasting pedagogical approaches to three- the examples discussed in the volume high- One key point of the volume concerns the dimensional models and dioramas taken by lights the intrinsic fragility of these objects ), VOL. 1, FIG. VIEWEG VERLAG, 10.2/COURTESY STUTTGART differing yet often overlapping functions of two German natural history museums. The as well as the tenuousness of our connec- 2 use of these “biological groups,” she notes, tion to and understanding of the knowledge The reviewer is the author of Making Culture Visible. reflected the escalating cultural tensions that produced them. Models reconfirms the

E-mail: [email protected] between the “real” and the “artificial” as value of three-dimensional models and the CREDIT: ( FROM

1136 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org Published by AAAS B OOKS ET AL. continued need to understand how the vi- As physicist Isidor Rabi, Oppenheimer’s curately assessed Oppenheimer’s contribu- sual process is integral to both the creation lifelong friend, was fond of observing, being tion to physics. As have others, Cassidy ar- and communication of science. Jewish is a fate, not a faith, and, being fate, gues that had Oppie been better focused, References is inescapable. Oppenheimer’s politicaliza- and had the war not intervened, his work on 1. N. Hopwood, Embryos in Wax. Models from the Ziegler tion in the late 1930s proceeded, in part, stellar evolution might have led him to the Studio (Whipple Museum of the History of Science, Cambridge, 2002); reviewed by J. Uglow, Science 297, from his sympathy for the plight of the Jews discovery of so-called black holes and pos- 1651 (2002). in Europe. Similarly, Oppie would eventual- sibly even to a Nobel Prize. But Cassidy 2. G. Fischer, Ed., Mathematische Modelle / Mathematical ly find in Communism the “City of Light” does not grant Oppenheimer the easy defer- Models (Vieweg, Braunschweig, 1986). that graduates of Felix Adler’s secular acade- ence that prior biographers have. In his my sought; the school’s early motto was “To view, Oppie was not a leading figure in HISTORY OF SCIENCE give the best to the poorest.” quantum physics, but simply adept at Indeed, a speech that Adler gave during “mopping up” after the first wave had One Path Through the First World War would later seem haunt- passed. Oppenheimer even fares poorly in ingly prophetic of Oppenheimer’s own situa- comparison with Heisenberg. Thus, by Los Alamos tion. “The time will come,” Adler warned, 1929—when Oppie was being lured by “when the scientist will be considered and both Caltech and Berkeley—“he was be- Gregg Herken will consider himself a disgrace to the human coming known,” writes Cassidy, “not only hysicist Robert Oppenheimer would race who prostitutes his knowledge of for his ideas but also for his carelessness in have been 100 years old last April— Nature’s forces for the destruction of his fel- calculations.” Elsewhere, Cassidy sees Phad he not died prematurely at age 62, low men.” Adler’s admonition was the result Oppenheimer as “the enabler, but not the a victim of throat cancer. Any centenary is of seeing scientists blindly follow soldiers in- prime mover, of fundamental innovation.” traditionally a celebratory occasion, and to the great war, wearing lab coats rather than Despite his focus more on Oppenheimer “Oppie” has been the subject of several pre- uniforms, but marching in the same serried as scientist than as political figure, Cassidy vious near-hagiographies. David Cassidy, a ranks. Oppenheimer was only twelve when makes an effort to include some of the re- Hofstra University historian of 20th-century Adler spoke, but the precocious Oppie’s own cent evidence and controversies that have physics and the author of a well-received words—especially the poetry he wrote while surfaced regarding Oppie’s protracted dal- biography of Werner Heisenberg, an undergraduate at Harvard in liance with Communism. But, by including has written a book that neither J. Robert the 1920s—suggests that he all possible theories and embracing none, praises Oppenheimer nor buries Oppenheimer and shared Adler’s view. Yet, a gener- the result may be more confusing than en- his reputation but, rather, puts the American ation later, Oppenheimer himself lightening to the reader. some tarnish upon the icon. Century would be at the head of another Cassidy’s research is generally both thor- As Cassidy notes, he meant J. by David C. Cassidy such column, involved in an ough and careful, but he too quickly accepts Robert Oppenheimer and the Pi Press, New York, 2004. even more terrible war, fought some claims without corroborating evidence. American Century to be in keep- 480 pp. $27.95, C$39.95. with still more terrible weapons. The author blithely passes along the gossip ing with the recent trend in biog- ISBN 0-13-147996-2. This irony is what gives that Oppie had sexual designs on not only raphy, which shifts the focus Oppenheimer’s life its magnetic Linus Pauling’s wife but on Pauling himself. from the individual toward his or her times. appeal to biographers, and it provides the nar- The source for this rather remarkable claim To that end, Cassidy borrows the theme of rative push that propels Cassidy’s story. Along is, evidently, speculation by Pauling’s biogra- “scientific militarism” from sociology and the way, however, the author journeys down pher. Cassidy also adds some careless errors the concept of the “American Century” from numerous side roads—including education of his own to the book: Oppie’s brother Frank Life publisher Henry Luce. But neither thread reform movements in the United States, com- was anything but “a very successful rancher,” really becomes the book’s leitmotif, and both parisons of science in Europe and America, and Oppenheimer’s friend Haakon Chevalier, are only spottily referenced in the text. This and U.S. immigration patterns in the early to the best of my knowledge, was never “an may be because Oppenheimer is a difficult 20th century—that, while often interesting adviser to DeGaulle.” subject for such an approach—he was, ar- and relevant to the tale, dilute its impact. While hardly demonology, neither is guably, unique (certainly his like had not Historians of science and practicing sci- Oppenheimer and the American Century a been seen before, nor has it since)—and also entists will debate whether Cassidy has ac- very flattering portrait. It is, rather, a rebut- because the times turned out to be as com- tal to the kind of mid-century hero- plex as the man. worship that tended to portray Oppie Although more a scientific biography as the Galileo of Cold War America. than a cradle-to-grave life, the book begins But it is also a portrait that seems, on with the young Oppenheimer as a way of il- balance, a bit too harshly drawn. One luminating its subject’s seemingly inexplica- reason for Cassidy’s disillusionment ble behavior as an adult. It has long been de- with Oppenheimer may be the disap- bated whether and how Oppenheimer’s pointment that follows unmet expecta- Jewishness and his education at Manhattan’s tions. Ultimately, what Oppenheimer Ethical Culture School affected him in later did in putting his science at the service life, but Cassidy makes a persuasive case of the state made him not so very that each had an effect greater than previous different from Edward Teller, Ernest biographers—or Oppie himself—realized. Lawrence, or any other scientist at Los Alamos. The problem is that, with Oppenheimer, somehow, one expected The reviewer, the author of Brotherhood of the Bomb, more. As Rabi also observed, puck- is a historian at the University of California, Merced, Partial vindication. Oppenheimer received the Fermi Post Office Box 2039, Merced, CA 95344, USA. Award, the U.S. Atomic Energy Commission’s highest ishly: “People like to think that

CREDIT: J. OPPENHEIMER MEMORIAL COMMITTEE ROBERT E-mail: [email protected] honor, in December 1963. Oppenheimer built only good bombs.”

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1137 Published by AAAS POLICY FORUM MOLECULAR BIOLOGY ers of high-quality compound libraries, small molecules can now be obtained on a large scale. At the same time, advances in NIH Molecular robotics and informatics have made screen- ing and analysis of such large compound li- braries possible. Up to a million com- Libraries Initiative pounds can now be screened against a tar- get in a single day, three orders of magni- 1* 2 2 1 Christopher P.Austin, Linda S. Brady, Thomas R. Insel, and Francis S. Collins tude greater than was possible only a decade ago. Together, these developments he purpose of the Molecular Libraries than the gene locus or mRNA, have virtual- make a public-sector small-molecule Initiative (MLI) component of the ly limitless structural diversity, can affect screening and chemistry initiative such as TNIH Roadmap for Medical Research particular target functions for defined peri- the MLI possible. (1, 2) is to expand the availability, flexibili- ods in isolated proteins, cells, or organisms, The MLI was developed over the course ty, and use of small-molecule chemical and can serve as either agonists or antago- of 9 months through consultations with rep- probes for basic research. Because this ini- nists. The characteristics that make this class resentatives of multiple NIH institutes, and tiative is particularly novel and far-reach- of molecule useful as drugs—their potential external consultants from the public and pri- ing, it has been the subject of considerable for selectivity, cell permeability, and subtle vate sectors. The MLI research agenda has discussion (3–5), and sometimes misinter- reversible modulation of important physio- three components focused on screening, pretation (6), in the research community. logical functions—also make them good re- cheminformatics, and technology develop- Two imperatives motivated the develop- search tools for dissecting the functions of ment, and is being carried out via NIH grant ment of the MLI. The first, related to NIH’s novel genes, pathways, and cells. and contract mechanisms (11). mission in basic biomedical research, was The human genome encodes 20,000 to The Molecular Libraries Screening the need for fundamentally new approach- 25,000 genes (8) and perhaps a million Center Network (MLSCN) will be a con- es to determine function and therapeutic proteins, of which only ~500 are targeted sortium of five or six high-throughput potential for all genes in the newly sequenced human Probability of success genome. The second, related to NIH’s mission to improve public health, was the need to accelerate the translation of Cumulative cost basic research discoveries into new therapeutics. Current Future Expense (particularly large capital costs), expertise, and Dedicated cultural divides between public Public biology-chemistry Compound accepted and private sectors have his- Sector project team initiated into clinical development torically kept discovery and Science optimization of small mole- cules largely restricted to phar- maceutical and biotechnology Indefinite companies. Dissemination of Ass Scre H Target it- Lead Phase I Phase II Phase III Regulatory Phase IV to- small-molecule research tools ay eni identification probe development review and V and technologies into the pub- dev ng elo and optimization Clinical trials p lic sector via the MLI is timely m for several reasons. en 1 yr t First, sequencing of the hu- CREDITS: (TOP) man genome has provided an Interface of the MLI and drug development. abundance of new targets for study, and small-molecule research tools by currently available small molecules (9). screening (HTS) centers that will screen will accelerate the translation of genome se- For the most part, pharmaceutical and assays submitted by the research communi- quence into biological and therapeutic in- biotechnology companies prefer to focus ty on a large number of compounds sights (7). Small molecules are complemen- on the “druggable genome” thought to be (>100,000) maintained in a central com- tary to nucleic acid–based translational tools more amenable to drug development (10). pound repository, and perform optimiza- such as knockout mice and siRNAs, in that The majority of the genome that is current- tion chemistry required to produce in vitro they target the protein gene product rather ly considered “undruggable” (i.e., unma- chemical probes of the targets or pheno- nipulable by small molecules) is therefore types studied in the assays (12). All results a major focus of the MLI. will be placed into a new public database 1National Human Genome Research Institute and Large libraries of small molecules have (PubChem, see below), and probe com- 2 National Institute of Mental Health, National traditionally been unavailable to academic pounds will be made available without en- Institutes of Health, Bethesda, MD 20892, USA. *Author for correspondence. E-mail: austinc@ researchers, but with the advent of combi- cumbrance to all researchers, in public and mail.nih.gov natorial chemistry and commercial suppli- private sectors, for their use in studying bi-

1138 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org Published by AAAS P OLICY FORUM ology and disease. The first of the MLSCN goals and deliverables from those of the pared with >$30 billion per year spent by screening centers (the NIH Chemical private sector. For example, the need to de- pharmaceutical and biotechnology compa- Genomics Center) has been established velop compounds for human use as quick- nies on drug development). We estimate within the NIH intramural program. It will ly as possible drives biopharmaceutical that the probes produced by the MLI will begin full-scale screening in early 2005 companies to screen only “druglike” com- entail only 2% of the cost, and 5% of the (13). The other MLSCN centers will be pounds having potential for intellectual time, required to develop a novel drug (19). funded via extramural grants; applications property novelty, on assays for a relatively In exceptional circumstances, the NIH it- to this NIH Roadmap Request for limited group of targets (e.g., GPCRs, ki- self might attempt the subsequent steps of Applications (12) are under review and will nases, nuclear hormone receptors), and to drug development, particularly if a small- be awarded in the late spring of 2005. The not disclose the results of screens, keeping molecule probe shows special potential for contract for the Molecular Libraries Small them as trade secrets or intellectual proper- development into a drug for an orphan in- Molecule Repository, which will house the ty. In contrast, the types of compounds syn- dication or a disease occurring primarily in screening collection, was recently awarded thesized and screened by the MLI will be the developing world, where there is un- (14), and the composition of the compound broader, including metabolic intermedi- likely to be commercial interest. In this collection is being determined by a distin- ates, a range of natural products, and case, NIH Roadmap mechanisms such as guished panel of chemists from the public agents with potential in vivo toxicity. Types the Translational Research Core Services and private sectors. of assays will be broader also, including, Program and the Regional Translational A comprehensive database of chemical for example, protein-protein interactions, Research Centers (20) could be used. But structures and their activities, PubChem splicing events, and diverse cellular and for the most part, we expect that chemists (15), has been developed by the National even organismal phenotypes. Via PubChem in the public and private sectors will use Center for Biotechnology Information at and the compound repository, all data and the MLI probes as proof-of-concept com- the National Library of Medicine/NIH. chemical probes will be available to the en- pounds and as starting points to produce a PubChem links small-molecule informa- tire research community without encum- variety of chemical analogs with improved tion to GenBank, MEDLINE, and the oth- brance. The MLI will thereby enable study properties. For this reason, intellectual er Entrez databases, and will serve as a of the majority of biological and chemical property claims on the MLI probes will be public portal for MLSCN screening results space that is currently unexplored. strongly discouraged, as such claims would and chemistry data. New algorithms and The small-molecule research tools pro- prevent the widest use of these tools and is tools for computational chemistry, virtual duced by the MLI should accelerate valida- contrary to the “community resource” na- screening, and other research aspects of tion of new drug targets (17) and thereby ture of this initiative (21). cheminformatics, will be funded via new enable new drug development, but clinical- The MLI is a bold initiative to catalyze grants (11). ly useful compounds cannot be expected science in the genome era. By providing a The ultimate goal of the MLI is to de- from the MLI itself. Drug development is a new path for discovery, this program aims velop small-molecule modulators of thou- complex, time-consuming, and expensive to accelerate science and its translation sands of cellular targets. To succeed, the process (18, 19), only the first steps of into benefits for the health of the public. MLI will be developing technology in four which will be performed by the MLI (see critical areas (11): the figure on page 1138). The “probe com- References and Notes (i) Chemical diversity: production of pi- pounds” produced by the MLI will have 1. E. Zerhouni, Science 302, 63 (2003). lot-scale compound libraries in novel areas potency and aqueous solubility adequate 2. http://nihroadmap.nih.gov/molecularlibraries/ index.asp. of chemical space (16), and methods devel- for in vitro applications, but chemical mod- 3. J. Kaiser, Science 304, 1728 (2004). opment for natural product isolation, char- ifications will generally be needed to con- 4. D. Janssen, Bio-IT World 3, 1 (2004). acterization, and chemistry. Compounds fer the selectivity, pharmacokinetic, and 5. F. S. Collins, Nature Rev. Drug Discov. 3, 640 (2004). 6. J. Couzin, Science 302, 218 (2003). will be placed into the Small Molecule metabolic properties required for in vivo 7. C. P. Austin, Curr. Opin. Chem. Biol. 7, 511 (2003). Repository and screened by the MLSCN use. From the probe stage, 10 to 12 8. International Human Genome Sequencing Consor- centers. chemist-years are commonly required to tium, Nature 431, 931 (2004). 9. J. Drews, Science 287, 1960 (2000). (ii) Assay diversity: development of inno- develop a “lead compound” with minimal 10. A. L. Hopkins, C. R. Groom, Nature Rev. Drug Discov. vative high-throughput assays for novel pharmaceutical properties, and an addi- 1, 727 (2002). proteins, cellular phenotypes, biological tional 20 to 30 chemist-years to produce a 11. http://nihroadmap.nih.gov/molecularlibraries/ grants.asp. functions, and disease mechanisms. “clinical candidate” compound appropriate 12. http://grants1.nih.gov/grants/guide/rfa-files/ (iii) Instrumentation: development of new for testing in humans. During this time, RFA-RM-04-017.html. technologies to allow HTS of novel assay >3000 different compounds based on the 13. www.genome.gov/12512295 14. http://pubs.acs.org/cen/news/8235/8235notw5.html. formats and to increase throughput and ac- initial probes are typically synthesized and 15. http://pubchem.ncbi.nlm.nih.gov/. curacy of current screening technologies tested. Even after this investment in chem- 16. N. P. Savchuk, K.V. Balakin, S. E. Tkachenko, Curr. Opin. (e.g., methods for highly parallel noncom- ical optimization, >90% of clinical candi- Chem. Biol. 8, 412-417 (2004). petitive detection of target binding, lab-on- dates fail in human testing (19). 17. M. A. Lindsay, Nature Rev. Drug Discov. 2, 831-834 (2003). chip technologies enabling complex multi- Examination of the cost and expected 18. K. H. Bleicher, H. J. Bohm, K. Muller,A. I.Alanine, Nature step assays). success rate at each stage of drug develop- Rev. Drug Discov. 2, 369-78 (2003). (iv) Predictive ADME/Toxicology: devel- ment demonstrates that the assay develop- 19. J. A. DiMasi, R. W. Hansen, H. G. Grabowski, J. Health Econ. 22, 151 (2003). opment of data sets and analysis algorithms ment, HTS, and hit-to-probe chemistry 20. http://nihroadmap.nih.gov/clinicalresearch/overview- for improved prediction of ADME (absorp- steps to be performed by the MLI are inex- translational.asp tion, distribution, metabolism, and excre- pensive and straightforward compared with 21. http://grants1.nih.gov/grants/guide/notice-files/ NOT-RM-04-014.html. tion) and toxicity properties of small the later stages of drug development (see 22. The Molecular Libraries Roadmap is a trans-NIH ini- molecules. the figure on page 1138). This should allow tiative conceptualized and implemented with the Although the MLI will utilize tools and the MLI to produce probes for a broad participation of all Institutes and Centers; see http://nihroadmap.nih.gov/molecularlibraries/mem- technologies found in biopharmaceutical range of targets with its relatively limited bers.asp for the members of the Molecular Libraries

CREDIT: companies, the MLI has quite distinct budget of ~$100 million per year (com- Project Team. The authors are leaders of the initiative.

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1139 Published by AAAS PERSPECTIVES ECOLOGY could be random or losses could be ordered with respect to the sensitivity of species to How Extinction Patterns environmental stress, body size, and abun- dance, traits that in turn reflect different kinds of impact (see the figure). Affect Ecosystems Not surprisingly, both types of scenario reduced sediment bioturbation, but the or- David Raffaelli der in which species disappeared had a greater overall impact on the ecosystem. e live in a world where the accel- those predicted by scenarios that assume One particular species of brittlestar (a rela- erated extinctions of species and species extinctions occur at random. tive of the starfish) was clearly a keystone Wchanges in biodiversity are no Solan et al. (5) combine into an elegant species. Due to its high abundance, large longer disputed issues. Much effort has gone model an extremely well-documented data body size, and high mobility, the loss of this into quantifying biodiversity loss rates for set of invertebrate communities in marine brittlestar species affected bioturbation dis- particular animal and plant groups (1). Less sediments off the coast of Galway, Ireland proportionately. When this species disap- clear, however, is the impact of such losses (7). This fusion, facilitated by the BIO- peared, the ecosystem changed dramatical- on ecosystems, especially when many differ- MERGE initiative (8), enables the authors to ly. In contrast, the loss of other species that ent kinds of species of plants and animals predict what will happen to the cumulative were small or rare had only minor effects are lost simultaneously (2). Yet policy-mak- effects of the small-scale sediment distur- on the ecosystem. Thus, rare or small-bod- ers urgently need guidance on the effects of bances (bioturbation) caused by the move- ied species are unable to compensate for multispecies losses if they are to plan for and ment, feeding, and respiration activities of declines in ecosystem processes when larg- advise on the societal consequences of bio- all 139 species of clams, worms, sea urchins, er, more abundant species are lost, contrary diversity changes. The ecological research brittle stars, and shrimps present in this sys- to earlier suggestions. Clearly, the effects of community has been highly active in at- tem if species are lost through impacts such biodiversity loss on ecosystem integrity tempting to provide such guidance (3, 4), but as overfishing, habitat destruction, and pol- will depend largely on the order in which many challenges remain. Foremost among lution. They scored each species for its body species are lost, which in turn is determined these is that most real extinction events are size, mobility, and mode of sediment mixing by the susceptibilities of ecosystems to dif- nonrandom with respect to species identi- to calculate an index of bioturbation poten- ferent types of stresses (overfishing, habitat ty—some species are more likely to go ex- tial for different species combinations and fragmentation, and pollution). tinct than others—whereas research studies for different degrees of species richness. In In contrast to the modeling strategy often assume that extinctions are random. their model, either extinction scenarios used by Solan et al., Zavaleta and Hulvey Two papers in this issue, by Solan et al. on page 1177 (5) and Zavaleta and Hulvey on Traits that Top increase down page 1175 (6), reporting on work in two very carnivore the food web different types of ecosystem, reveal that the impact of nonrandom species extinctions on ecosystems is markedly different from that Body size Carnivores predicted by scenarios where extinctions are Longevity Vulnerability to stress random. These studies bring us a step nearer Home range area to understanding the impact of nonrandom Ability to adapt Vulnerability to Herbivores (evolve) species losses on ecosystems and should habitat fragmentation Species richness help to provide policy-makers with a firmer (insurance) basis for decision-making. The two studies examine very different Primary habitats (marine versus terrestrial), each Traits that producers increase up with different kinds of organisms (sea-bed the food web invertebrates versus grassland plants), dif- ferent ecosystem processes (sediment bio- Extinctions are rarely random in the real world. In the food web shown, the loss of biodiversity geochemistry versus resistance to invasion has a different impact depending on the trophic level, because the traits of these species that make by exotic species), and different types of them vulnerable to different impacts covary both between and within trophic levels. Thus, the body experimental approaches (data analysis and size (represented by the size of the circles) of top carnivore species tends to be larger than that of species at lower trophic levels (predators are usually larger than their prey). Similarly, there is a range modeling versus controlled experimenta- of body sizes (and other traits) within each trophic level, and there are also more species represent- tion). So it is all the more interesting, for ed at lower trophic levels. Traits such as body size and its covariates such as home range and toler- scientists and policy-makers alike, that ance to stress, together with differences in species richness between trophic levels, will determine both papers arrive at the same conclusion: the impact on ecosystems of different biodiversity loss scenarios. For instance, top predators with Nonrandom extinction events have impacts their large body size, low abundance, and large home range requirements are particularly vulnerable on ecosystems that are quite different from to habitat fragmentation/destruction, but may be less susceptible to contaminant stress, which af- fects smaller species disproportionately. In contrast, because of the higher species richness at low- The author is in the Environment Department, er trophic levels, there is more “insurance” against the effects of species losses, and these species University of York, York YO10 5DD, UK. E-mail: may also have a greater capacity for adaptive change due to their shorter life-spans and faster [email protected] turnover rates. With respect to species traits, extinction is unlikely to occur randomly (5, 6).

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1141 Published by AAAS P ERSPECTIVES (6) adopted a highly controlled experimen- ments in which species richness treatments These two studies illustrate how it is possi- tal approach. They used a relatively simple were assembled randomly. The authors ble to use alternative approaches to tackle (21 species) assemblage of California conclude that the consequences of species this aspect of the biodiversity-ecosystem grassland plants, which they arranged in loss may be greater than that predicted on dynamic more realistically, but much re- treatment plots to provide a range of the basis of studies in which extinctions are mains to be done. Zavaleta and Hulvey fo- species richness. Half of the plots were assumed to occur randomly. cus on a single trophic level (plants). Solan challenged by an exotic invader, the yellow Both the Solan et al. and Zavaleta and et al. concentrate on a community that com- star thistle, to assess the resistance of the Hulvey studies were motivated by the need prises different trophic types, but this com- different communities to invasion (an to inject greater realism into analyses of munity does not capture the larger food web ecosystem characteristic). Their experi- how biodiversity affects ecosystems. Many that includes fish and other consumers. mental design constitutes a major advance of the high-profile studies in this area are Different ecological impacts will affect bio- over previous experiments on grassland concerned with the measurement of a re- diversity differently at various trophic levels systems in that the species richness treat- sponse variable (ecosystem process) under (see the figure). The major challenge facing ments that they assembled represent plant conditions of different numbers of species ecologists is to be able to advise policy- communities at different stages of biodiver- assembled randomly and usually compris- makers about the effects of such combined sity loss and reflect nonrandom extinctions. ing a single trophic level. Although such re- losses on ecosystems throughout a food The authors achieved this by applying nest- search provides a bivariate plot of ecosys- web. These two papers take us an important ed subset analysis to a 4-year data set of tem integrity versus species richness, it is step nearer to achieving that goal. spatial variation in plant communities in uncertain whether such relationships reflect order to reveal those changes in richness what happens when species are lost from References and Notes that occur in a consistent, nested order. ecosystems. Controlled removal of species 1. See www.royalsoc.ac.uk/events. These changes were then reflected in the from species-rich assemblages to produce a 2. See www.millenniumassessment.org/en/index.aspx. setup of the experimental plots. lower level of richness would go a long way 3. M. Loreau et al., Science 294, 804 (2001). 4. M. Loreau, S. Naeem, P. Inchausti, Biodiversity and The results were dramatic: Resistance to toward answering this question. Species re- Ecosystem Functioning (Oxford Univ. Press, Oxford, invasion by the thistle was tightly correlat- moval, however, is fraught with practical 2002). ed with the species richness of the invaded obstacles and difficulties over interpretation 5. M. Solan et al., Science 306, 1177 (2004). 6. E. S. Zavaleta, K. B. Hulvey, Science 306, 1175 (2004). community. These findings are completely of results, nothwithstanding the issue of 7. J. L. Ruesink, D. S. Srivastava, Oikos 93, 221 (2001). at odds with the results of similar experi- random versus nonrandom extinctions. 8. See www.columbia.edu/cu/biomerge.

PHYSICS Early efforts focused on the first ques- tion. It was proposed that pairing could arise Superfluid Helium-3 from the exchange of spin or from fluctua- tions in the magnetization rather than from lattice vibrations. This idea received a boost Has a Metallic Partner when superfluidity was discovered in the 3 Maurice Rice early-1970s in liquid He and was shown to be a pair condensate with p-wave symmetry fter the unexpected discovery of relative motion of the paired electrons can (l = 1). Just like electrons, the 3He isotope is high-temperature superconductors, it be described by an s-wave structure, corre- a fermion (that is, no two 3He atoms can oc- took almost a decade and a new class sponding to an angular momentum l = 0. cupy the same state at the same time). The A 3 of experiments (1) to establish that their Not long after this theory was developed, He condensate is called a superfluid rather electronic nature was fundamentally differ- Kohn and Luttinger speculated that super- than a superconductor simply because, in ent from that of conventional, low-tempera- conductors in which the electrons have fi- contrast to electrons, the helium atoms do ture superconductors. Now, Nelson et al., nite angular momentum (l = 1 or higher) not carry a charge. on page 1151 of this issue, (2) have per- could also occur (4). It was unclear, how- However, the search for a metallic formed similar experiments on strontium ever, how such an unconventional super- counterpart to superfluid 3He remained ruthenate (Sr2RuO4), a metallic supercon- conductor could be stabilized and how its fruitless at first. In the 1980s, supercon- ductor that does not conform to either of unconventional nature could be proved. ductivity was observed in metals that did the two types. In an experimental tour de force, the authors confirm that this ruthen- L ate metal is the long-sought metallic analog Josephson junction Josephson junction 3 S of superfluid helium-3 ( He). L + S = J = 0 L + S = J = 0 In a superconductor, electron pairs move through the material without encoun- J = 0 J = 0 tering any electrical resistance. The elec- trons move both as a pair and relative to s-wave Auln s-wave Auln each other. According to the original p-wave Sr2RuO4 L = 1, S = 1 Bardeen-Cooper-Schrieffer theory (3), the

pairing is caused by interactions between SCIENCE The definitive experiment. The p-wave superconductor Sr2RuO4 is joined at either end to a con- the electrons and the crystal lattice. The ventional s-wave superconductor. The angular momentum vector L of the Cooper electron pairs in Sr2RuO4 is oriented along the c axis of the tetragonal crystal, whereas the spin vector S lies in the The author is at the Institut für Theoretische Physik, ab plane. At the insulating barriers, L rotates in the ab plane, and J = 0 electron pairs are formed. ETH Hönggerberg, 8093 Zürich, Switzerland. E-mail: The formation of these J = 0 pairs allows interconversion between the spin-singlet pairs in the s-

[email protected] wave superconductor and the spin-triplet pairs in the p-wave superconductor. CREDIT: HUEY/ PRESTON

1142 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org Published by AAAS P ERSPECTIVES not fit the Bardeen-Cooper-Schrieffer In their experiments, Nelson et al. (1), References mold. First came the “heavy fermion” met- chose a geometry in which half of the ring 1. C. C.Tsuei, J. R. Kirtley, Rev. Mod. Phys. 72, 969 (2000). 2. K. D. Nelson, Z. Q. Mao, Y. Maeno, Y. Liu, Science 306, als, which contain rare earth or actinide is a conventional s-wave superconductor, 1151 (2004). ions with local magnetic moments that are which is joined to opposite ends of a 3. J. Bardeen, L. N. Cooper, J. R. Schrieffer, Phys. Rev. 108, 1175 (1957). only weakly coupled to the conducting Sr2RuO4 sample. As illustrated in the fig- 4. W. Kohn, J. J. Luttinger, Phys. Rev. Lett. 15, 524 (1965). electrons. There are many reasons to be- ure, tunneling between s- and p-wave su- 5. P. Thalmeier et al., Condens. Matter, in press; preprint lieve that superconductivity in these metals perconductors requires interconversion be- available at http://xxx.lanl.gov/abs/cond-mat/0409363. is unconventional, but their intrinsic com- tween spin-singlet and spin-triplet pairs. 6. P.W.Anderson et al., J. Phys. Condens. Matter 16, R755 (2004). plexity has so far prevented a definitive de- The observation of the added phase shift by 7. Y. Maeno et al., Nature 372, 532 (1994). termination of their electronic nature (5). Nelson et al. is the final piece of evidence 8. T. M. Rice, M. Sigrist, J. Phys. Condens. Matter 7, L643 Next came the high-temperature cuprates, that confirms the close analogy between (1995). 9. A. P. Mackenzie,Y. Maeno, Rev.Mod.Phys. 75, 657 (2003). in which the relative motion of the paired the metallic superconductor Sr2RuO4 and 10. V. B. Geshkenbein, A. I. Larkin, A. Barone, Phys. Rev. B 3 electrons has a d-wave structure (l = 2). superfluid He. 36, 235 (1987). However, cuprates have many additional anomalies that do not fit into a generalized OCEAN SCIENCE Bardeen-Cooper-Schrieffer theory (6). A decade ago, Maeno et al. (7) succeed- ed in growing perfect crystals of Sr2RuO4 and found that they were superconducting at Deep Ocean Overturning— low temperatures. Above the superconduct- ing transition temperature, the material be- haves as a standard metal, but there are clear Then and Now signs that the motion of the electrons is Jess F. Adkins and Claudia Pasquero strongly correlated at all temperatures, sim- ilar to 3He atoms, which move in this way to he deep ocean contains nearly all the ocean sediments from this time record the avoid each other both in the superfluid and mass, thermal inertia, and carbon in chemistry of the water in which they grew. the normal liquid. This observation quickly Tthe ocean-atmosphere system. The But here we run into a problem. We cannot led to speculation that the superconducting rate at which it overturns can therefore use radiocarbon in bottom-dwelling fora- state of Sr2RuO4 would also be p wave (l = have a profound effect on climate. minifera as a tracer of water mass age, 1) (8). This type of relative motion has con- Measurements of the radiocarbon (14C) dis- because their radiocarbon values record the sequences for the spin state of the pairs. tribution in the deep sea suggest that the sum of the ages of the sediment and of the Basic quantum mechanics for fermions re- modern ocean water in which they lived. Without an inde- quires the spins of electrons pairs with odd Enhanced online at overturns once pendent chronometer for sediment age, we angular momentum to be parallel—that is, www.sciencemag.org/cgi/ every ~1000 are stuck with one measurement and two un- to form spin-triplet pairs, rather than the content/full/306/5699/1143 years. Radio- knowns. However, we can still measure the spin-singlet pairs in conventional and carbon is an accurate tracer of this process radiocarbon difference between old sam- cuprate superconductors. Subsequent stud- because it is created in the upper atmosphere ples, as long as we know that they formed at ies of the magnetic properties of Sr2RuO4 by cosmic rays and can only enter the ocean the same time in the past. pointed to triplet pairing, and much support- through absorption of CO2 This trick is used by ing evidence has accumulated since then (9). gas at the ocean surface. Broecker et al., on page The final confirmation requires that the When the surface water 1169 of this issue (1), to esti- internal p-wave structure of the pairs be test- sinks to the deep ocean, it is mate the radiocarbon age of ed in a quantum interference experiment. The isolated from the radiocar- the equatorial deep Pacific at way to do this was outlined by Geshkenbein bon source, and its radiocar- a depth of 2 km during the et al. (10) almost two decades ago. They pro- bon content decays with its Last Glacial Maximum. The posed to modify the standard superconduct- characteristic half-life of authors compare the age dif- ing quantum interference device (SQUID), 5730 years. Today, there is ference between surface- which consists of a ring with weak links in more radiocarbon in the dwelling and bottom- the left and right arms. A weak link (called a deep Atlantic Ocean (as a re- dwelling foraminifera both Josephson junction) can be made by inserting sult of deep-water formation today and at the Last Glacial a narrow insulating barrier. Quantum tunnel- in the North Atlantic) than in Maximum. They find that ing of the electron pairs through the barrier is the Pacific Ocean. We say the surface-to-deep age dif- usually maximized when their phases on ei- that the Pacific is “older” or A foraminiferan of the kind ference was the same, or a ther side are equal. However, Geshkenbein et less well “ventilated” than used in (1) to determine the little larger, in the past. al. (10) pointed out that, in the case of p-wave the Atlantic. age of mid-depth waters. This result is somewhat superconductivity, a suitably chosen geome- But what about the past surprising. In a paper pub- try would add an extra phase shift to the tun- ocean, especially at the time of the Last lished earlier this year in Science, Hughen neling amplitude of one of the two Josephson Glacial Maximum, about 20,000 years ago, et al. came to a different conclusion (2) junctions. The interference pattern of a when large ice sheets covered most of (see the second figure). The authors docu- SQUID is observed by measuring the de- Canada? The calcium carbonate shells of mented the history of the atmosphere’s ra- pendence of the maximum supercurrent foraminifera (see the first figure) in deep- diocarbon content over the past 50,000 through the device as the total magnetic flux years. The atmosphere provides an indirect flowing through the ring is varied. The extra The authors are in the Department of Geology and record of the ocean overturning rate, be- phase shift is easily recognized as a displace- Planetary Sciences, California Institute of Technology, cause its radiocarbon budget is very sensi-

CREDIT:RESEARCHERS, DEE BERGER / PHOTO INC. ment of the interference pattern. Pasadena, CA 91125, USA. E-mail: [email protected] tive to how much CO2 it exchanges with

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1143 Published by AAAS P ERSPECTIVES

800 these sediments remove radio- agrees with recent theoretical arguments 700

C) carbon from the system. Another (6–8). For example, Huang has shown that a

14 600 option is that the waters below 2- modeled increase in the density of high-lat- ∆

500 km depth were much older than itude waters did not directly result in an in- those measured by Broecker et creased rate of deep-water formation (8). In 400 al. There is not much data, but fact, a large meridional surface density gra- 300 two papers have found very ra- dient induces a strong vertical stratification, 200 diocarbon-depleted waters in the which inhibits the return of deep water to deepest parts of the ocean at the the surface and weakens circulation. 100 Last Glacial Maximum (4, 5). Observational data from the paleocli- 0 Clearly we need water column mate record support this theory. The ocean

Atmospheric radiocarbon ( –100 profiles of radiocarbon for the interior was more highly stratified at the Last Glacial Maximum from Last Glacial Maximum than it is in the –200 0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 45, 000 both the Atlantic and the Pacific. modern ocean (9), but the circulation was Calendar age (years) Finally, it is possible that not much stronger, and was possibly slow- Atmospheric radiocarbon content over the past 45,000 19,000 years ago—the age of er, than it is today. More data are needed to years. Data from tree rings [red line (11)] and from foraminifera some of the data used by Broecker determine whether the strength of the over- from the Cariaco Basin [green line (12) and blue points (2)] are et al.—the system was not in a turning circulation depends on winds and compared to modeled production rate variations (2) (black line). steady state. The atmospheric tidal energy (7, 8, 10), rather than on sur- Deviations between the data and the model reflect changes in record in the figure shows that face warming/cooling or evaporation/pre- 14 the exchange of carbon between active reservoirs relative to to- large swings in ∆ C occurred reg- cipitation budgets at the surface. The day.The two most important processes for the largest deviations ularly during the last glacial peri- growing paleoceanographic data set shows are changes in carbonate sedimentation and in the ocean over- od. Such swings did not occur great promise to answer this question. turning rate.The data suggest that, assuming constant carbonate over the past 10,000 years of rela- flux, the glacial ocean overturned more slowly than it does today. tively constant climate, during References 1. W. Broecker et al., Science 306, 1169 (2004). which the radiocarbon cycle was 2. K. Hughen et al., Science 303, 202 (2004). the deep ocean. Land biota and soils also in a steady state. Ice core records in 3. R. Muscheler et al., Earth Planet. Sci. Lett. 219, 325 take up radiocarbon, but export to the deep Greenland show that no equivalent period (2004). 4. N. J. Shackleton et al., Nature 335, 708 (1988). sea is by far the largest sink for modern at- occurred during the last glacial period. 5. L. Keigwin, Paleoceanography, in press. mospheric radiocarbon. If this sink were Moreover, the small but growing num- 6. C. Wunsch, R. Ferrari, Annu. Rev. Fluid Mech. 36, 281 reduced, the atmospheric radiocarbon con- ber of deep radiocarbon values from the (2004). tent would rise, and rise fast. If a new 7. W. Munk, C. Wunsch, Deep-Sea Res. 45, 1977 (1998). Last Glacial Maximum (1, 4, 5) provides in- 8. R. X. Huang, J. Phys. Oceanogr. 29, 727 (1999). steady state were reached, and the produc- sights into what drives the strength of the 9. J. F. Adkins, K. McIntyre, D. P. Schrag, Science 298, tion rate remained constant, the partition- overturning circulation: Contrary to a wide- 1769 (2002). ing of radiocarbon between the atmosphere ly held belief, the circulation rate is not 10. J. R. Toggweiler, B. Samuels, Deep-Sea Res. 42, 477 (1995). and deep ocean would shift toward higher driven by surface density gradients in 11. M. Stuiver et al., Radiocarbon 40, 1041 (1998). 14 12 14 atmospheric C/ C (∆ C) ratios. This is the north-south direction. This conclusion 12. K. Hughen et al., Science 290, 1951 (2000). exactly what Hughen et al. found for 11,500 to 13,000 years ago (see the second EVOLUTION figure, green line) (2). But there are several important other factors to consider. The production rate of Genomic Databases radiocarbon is not constant. Radiocarbon production can be tracked through time by measuring the abundance of other radioac- and the Tree of Life tive isotopes produced by cosmic rays, Keith A. Crandall and Jennifer E. Buhay such as 10Be and 36Cl, in ice cores (3); these isotopes are not affected by carbon sinks. It lthough we have not yet counted the quences sampled from the GenBank and turns out that radiocarbon production was total number of species on our plan- Swiss-Prot genetic databases. From these higher in the past (see the second figure, et, biologists in the field of system- data, they generated “supermatrices” and 14 A black line). But the measured ∆ C at the atics are eagerly assembling the Tree of then supertrees. Supermatrices are extreme- Last Glacial Maximum was even higher Life (1, 2). The Tree of Life aims to define ly large data sets of amino acid or nucleotide (see the second figure, blue symbols). the phylogenetic relationships of all organ- sequences (columns in the matrix) for many Hence, the ocean overturning must have isms on Earth. On page 1172 of this issue, different taxa (rows in the matrix). Driskell been slower. However, Broecker et al. show Driskell et al. (3) propose an intriguing et al. constructed a supermatrix of 185,000 with a simple calculation that the implied computational method for assembling this protein sequences for more than 16,000 increase in the radiocarbon age of the deep phylogenetic tree. green plant taxa and one of 120,000 se- ocean is much larger than their measured These investigators probed the phyloge- quences for nearly 7500 metazoan taxa. This age difference from the foraminifera. netic potential of ~300,000 protein se- compares with a typical systematics study How can these disparate results be rec- of, on a good day, four to six partial gene se- onciled? One possibility is that the carbon The authors are in the Department of Integrative quences for 100 or so taxa. Thus, the poten- cycle was fundamentally different at the Biology, and K.A. Crandall is also at the Monte L. Bean tial data enrichment that comes with care- Last Glacial Maximum. Changes in the Life Science Museum and the Department of Microbiology and Molecular Biology, Brigham Young fully mining genetic databases is terrific. burial rate of carbonate sediments can af- University, Provo, UT 84602, USA. E-mail: keith_ However, this enrichment comes at a cost.

fect atmospheric radiocarbon, because [email protected], [email protected] Traditional phylogenetic studies sequence CREDIT:

1144 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org Published by AAAS P ERSPECTIVES the same gene regions for all the taxa of in- nisms such as horizontal gene transfer and Foundation’s estimate of 1.7 million terest while minimizing the overall amount hybridization. Depicting genealogical rela- species. However, estimates of the total of missing data. With the database superma- tionships as networks might better represent number of species on Earth (not just those trix method, the data overlap is sparse, the true underlying biology (7, 8). known to science) range from 4 million to resulting in many empty cells in the super- Nonetheless, the ability of Driskell et al. 100 million. The computational methodolo- matrix, but the total data set is massive. to estimate apparently robust phylogenetic gy of Driskell and colleagues may be well To solve the problem of sparseness, the estimates from an impressively large and suited to future efforts both in terms of tax- authors built a “supertree” (4). The su- equally impressively sparse data set—all on sampling and gene sampling for maxi- pertree approach estimates phylogenies for collected from existing databases—has im- mizing coverage of the Tree of Life. This subsets of data with good overlap, then represents an extreme departure combines these subtree estimates into a su- from the notion of “barcoding” all of pertree. Driskell and colleagues took indi- life, which emphasizes sequencing vidual gene clusters and assembled them one gene for all species (9). Both ap- into subtrees, and then looked for sufficient proaches rely heavily on having a taxonomic overlap to allow construction of well-populated database, but the su- e t a supertree. For example, using 254 genes a pertree strategy does not have the V r e b r t e r (2777 sequences and 96,584 sites), the au- e t constraint that these data must be r la b e r u thors reduced the green plant supermatrix a v c from the same gene. It may turn out t n s e I a to 69 taxa from 16,000 taxa, with an aver- v r that two different databases are n ula o sc age of 40 genes per taxon and 84% missing Animalia N Va needed for these two distinct pur-

sequences! This represents one of the poses (that is, establishing relation- F Plantae un largest data sets for phylogeny estimation gi ships versus diagnosing species). ta in terms of total nucleotide information; tis If the supertree approach estab- Pro ro teob P act but it is the sparsest in terms of the per- C eria lishes the trunk and thick branches ya no centage of overlapping data. Yet even with ba of the Tree of Life, then perhaps the ct er ia yta such sparseness, the authors are still able to oc barcoding approach is more appro- E a e estimate robust phylogenetic relationships a priate for discerning the twigs and h c r that are congruent with those reported us- a leaves of the tree (see the figure). Ba y ci r lli u ing more traditional methods. Computer E Currently, most attention has been

simulation studies (5) recently showed that, focused on the trunk at the expense contrary to the prevailing view, phylogenet- of the leaves. However, the leaves ic accuracy depends more on having suffi- are dropping quickly. We are losing cient characters (such as amino acids) than 27,000 species each year while only on whether data are missing. Clearly, build- describing 18,000 new species (10). ing a supertree allows for an abundance of The Driskell et al. study provides characters even though there are many hope for combining diverse and missing entries in the resulting matrix. sparse data sets collected from both Several questions remain, however, about leaf and trunk areas of the Tree of this strategy. First, the supertree strategy de- Building the Tree of Life. A current view of the Tree of Life to provide a robust estimate of pends fundamentally on our ability to distin- Life (7). Information is biased toward vertebrate animals this tree, but this should in no way guish between orthologous (derived from a and vascular plants (the thick branches); lesser-known undermine efforts to characterize as speciation event) and paralogous (derived groups such as bacteria, fungi, and protists are largely many leaves as possible before they from a duplication event) gene sequences underrepresented. Also shown are species known to sci- hit the ground. Future applications (6). The methods to draw this distinction are ence (green leaves), extinct species (leaf litter, brown), of Driskell et al.’s computational in their infancy. Little work has been done to endangered species (falling leaves), and species for which method and verification of its per- compare such methods in terms of their ac- “barcode” information is available (red leaves). formance compared with computer- curacy and their robustness with respect to simulated known phylogenetic his- data that do not fit underlying assumptions portant implications for future work on the tories and empirically “known” histories (such as neutral evolution). The distinction Tree of Life. This and other studies demon- will provide further insights into the gener- between the two types of gene sequences strate the importance of computational ap- ality of mining our genetic databases to as- typically relies on a well-populated data- proaches compared with brute-force gene semble the Tree of Life. base. Second, supertree approaches them- sequencing. In addition, Driskell et al.’s ap- References and Notes selves are controversial, in part because the proach allows for a more comprehensive 1. M. Pagel, Nature 401, 877 (1999). methodology results in a degree of discon- sampling of existing data, which enables 2. A new NSF program funds computational approaches for “assembling the Tree of Life” (AToL). Total AToL program nect between the underlying genetic data characterization of evolution across the funding is $13 million for fiscal year 2004. NSF, and the final tree produced. Moreover, this greatest possible diversity of life. Driskell et Assembling the Tree of Life: Program Solicitation NSF 04- strategy has yet to be validated by computer al. report that more than 100,000 species 526 (www.nsf.gov/pubs/2004/ nsf04526/nsf04526.pdf). 3. A. C. Driskell et al., Science 306, 1172 (2004). simulation or well-established phylogenetic have at least one molecular sequence 4. M. J. Sanderson et al., Trends Ecol. Evol. 13, 105 (1998). methods. Third, the supertree approach archived in public databases. They fail, 5. J. Wiens, Syst. Biol. 52, 528 (2003). makes a fundamental assumption: that a bi- however, to mention the extreme sampling 6. J. W. Thornton, R. DeSalle, Annu. Rev. Genomics Hum. Genet. 1, 41 (2000). furcating tree topology represents the ge- bias in these databases toward vertebrate 7. M. C. Rivera, J. A. Lake, Nature 431, 152 (2004). nomic evolutionary history of species. This animals and green plants. They suggest that 8. W. Doolittle, Science 284, 2124 (1999). assumption has been called into question be- the number of these favored species is 6% 9. P. D. N. Herbert et al., Proc. Natl. Acad. Sci. U.S.A. 101, 14812 (2004). cause of the reality of genetic exchange of those known to science, which is rough- 10. E. O. Wilson, The Diversity of Life (Norton, New York,

CREDIT: across species boundaries through mecha- ly equivalent to the National Science 1992).

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1145 Published by AAAS REVIEW Biodiversity Conservation and the Eradication of Poverty William. M. Adams,1 Ros Aveling,2 Dan Brockington,3 Barney Dickson,2 Jo Elliott,5 Jon Hutton,4 Dilys Roe,6 Bhaskar Vira,1 William Wolmer7

ination (21, 22). There is a clear need for It is widely accepted that biodiversity loss and poverty are linked problems and that these to be integrated with national sustain- conservation and poverty reduction should be tackled together. However, success with able development strategies (1, 23). The UN integrated strategies is elusive. There is sharp debate about the social impacts of MDGs are premised on such integration, conservation programs and the success of community-based approaches to conservation. with the area of land protected to maintain Clear conceptual frameworks are needed if policies in these two areas are to be biological diversity being an indicator of combined. We review the links between poverty alleviation and biodiversity conserva- performance against MDG Goal 7 (Bto tion and present a conceptual typology of these relationships. ensure environmental sustainability[). How- ever, the co-listing of poverty elimination iodiversity conservation scientists have substantial negative impacts on local and environmental goals does not mean that face a dilemma. There is increasing people. The eviction of former occupiers or integrated solutions are possible or that B concern that global efforts to maintain right holders in land or resources can cause protected areas can contribute to growth and biodiversity are in conflict with those to the exacerbation of poverty, as well as con- poverty reduction in poor countries. Indeed, reduce poverty (1). The decline of popula- travention of legal or human rights (10–14). the separation by the MDGs of environmen- tions, extinction of species, and habitat Globally, it is recognized that the costs of tal sustainability issues from development transformation demand urgent action (2). biodiversity conservation are not distributed goals alarms some observers (24). It has even The leading response to these threats since in proportion to their benefits (15). Typi- been suggested that the urgent global push the late 19th century has been the creation of cally, many of the costs of protected areas in for poverty reduction has subsumed or protected areas (3). Technical capacity to poor biodiverse countries are paid by local supplanted conservation goals (1). design effective protected-area systems is people (16). The 7th CBD COP called for an increasing (4), allowing the identification of assessment of Bthe economic and socio- Combining Conservation and coverage and remaining gaps in the interna- cultural costs of protected areas (including Development Goals tional protected-area system (5). This, com- the cost of livelihood opportunities foregone), The combination of poverty elimination and bined with positive assessments of the and policies to ensure that they are equitably biodiversity conservation goals has been effectiveness of protected areas is encourag- compensated[ (8). By the start of the 21st approached in various ways. The specific ing the consolidation and expansion of the century, a remarkable international agree- problem of the social impacts of protected network of protected areas (6). The 2004 ment on the urgency of global poverty areas has been recognized by conservation World Database on Protected Areas includes elimination had made the relation between planners for two decades. The principle that over 105,000 sites covering an area of 19.7 biodiversity conservation and poverty reduc- the needs of local people should be system- million km (2, 7). The establishment and tion an important element of debate about atically integrated into protected-area plan- effective management of a global series of conservation policy (1, 13). ning was agreed to at the third World Parks protected areas was a key element of the 7th The meaning of poverty may be intui- Congress in Bali in 1982 (25). In 1992, the Conference of the Parties (COP) to the tively obvious, but its measurement is com- president of IUCN–The World Conserva- Convention on Biological Diversity (CBD) plex. Common definitions are based on tion Union argued that ‘‘if local people do in 2004 (8). monetary (such as per-capita income) or not support protected areas, then protected The problem with this strategy is that its nonmonetary (such as health or mortality) areas cannot last’’ (26). IUCN’s director gener- impacts on poverty are often negative. The criteria, although broader approaches have al now suggests that protected areas should creation of protected areas causes the fore- been suggested (17, 18). In 1999, 1.2 billion be seen as ‘‘islands of biodiversity in an closure of future land use options, with people worldwide had consumption levels ocean of sustainable human development,’’ potentially significant economic opportunity below $1 a day and 2.8 billion lived on less with their benefits extending far beyond their costs (9). The creation of protected areas can than $2 a day (17). Poverty is not a static boundaries (27), but this is still an aspiration. condition, but it is estimated that between Delegates from the human rights and minority 1Department of Geography, University of Cambridge, 300 and 420 million people live in a state of peoples’ movements prominently voiced Cambridge, CB2 3EN, UK. 2Fauna & Flora Interna- chronic poverty (always or usually poor) concern at the persistence of such impacts tional, Great Eastern House, Tenison Road, Cam- (19). The first of the United Nations Millen- at the fifth World Parks Congress in bridge, CB1 2RS, UK. 3School of Geography and the Environment, University of Oxford, Mansfield Road, nium Development Goals (MDGs), agreed September 2003 (28). There are coherent Oxford, OX1 3TB, UK. 4Resource Africa, Post Office on in 2000, was to halve, between 1990 and calls for better understanding of the social Box 198, Cambridge, CB3 0TF, UK. 5Department for 2015, both the proportion of people whose impacts of protected areas (29, 30). International Development, 1 Palace Street, London 6 income is less than $1 a day and the propor- Beyond protected areas, the question of SW1E 5HE, UK. International Institute for Environ- tion of people who suffer from hunger (20). whether it is possible to combine poverty ment and Development, 3 Endsleigh Street, London WC1H 0DD, UK. 7Institute of Development Studies, National poverty reduction strategies are elimination and biodiversity conservation University of Sussex, Falmer, Sussex, BN1 9RE, UK. central to attempts to achieve poverty elim- relates to the more general debate, familiar

1146 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EVIEW to conservation scientists, about the environ- and poverty reduction simultaneously, it may limited conservation resources and compro- mental dimensions of development. In the only be possible under specific institutional, mising biodiversity preservation (37, 38). 20th century, the dominant approach was to ecological, and developmental conditions The key to the success of conservation is push for economic growth first and assume [such as in long-lasting field projects in the establishment and effective management that environmental problems (and indeed small human communities in fragile ecosys- of a complete global network of protected improved social welfare) could be sorted out tems (1)]. The links between biodiversity and areas selected because of scientific criteria later. Economists argue that as economies livelihoods, and between conservation and and owned or legally established by the state grow, they can invest in cleaner technologies poverty reduction, are dynamic and locally or legitimate private owners. Success is and less resource-depleting processes: Argu- specific (34, 45). In most cases, hard choices measured in terms of biodiversity criteria, ably, an ‘‘environmental Kuznets curve’’ can will be necessary between goals, with signif- not of measures of social development (6). be observed in industrialized and newly icant costs to one goal or the other. The 2) Poverty is a critical constraint on industrialized countries, with improvements acceptability of these costs will vary for conservation. This position makes the em- in factors such as air pollution (31). In the different organizations and actors. pirical, pragmatic argument that poverty 1950s and 1960s, development planners paid limits conservation success to a sufficient scant attention to environmental impacts, Diverse Relations Between degree that biodiversity conservation will whether focusing on poverty elimination, the Conservation and Poverty Reduction fail if it does not successfully address pov- creation of high-productivity agriculture, or Clarity over the choices between biodiversity erty elimination. Such a position might be physical infrastructure such as dams or conservation and poverty elimination goals expected in a scenario where poor people industrialization and the associated problems is essential. The desire to package projects as were overharvesting wild species, poaching of pollution (32, 33). delivering win-win solutions plays down the critical species, or colonizing and cultivating Critics of this technocratic top-down de- incompatibilities between goals. Equally, biodiverse land, and if the political or eco- velopment focused on its environmental and exclusive conservation or development goals nomic costs of stopping them (such as by a social failures (33, 34). The need to improve can be blind to alliances that favor both (1). conventional strict protected-area strategy) the environmental record of development We therefore offer a conceptual typology of were prohibitive. Poverty reduction would be gave rise to the second approach to the en- the relationships between poverty reduction undertaken in this instance simply as a vironmental aspects of development, in the andconservationinordertopromotea means to achieve more effective conserva- concept of sustainable development, which clearer understanding of them. The typology tion. This position holds that to achieve its underpinned the 1980 World Conservation presents four different ways of looking at the goal, conservation must provide (and be seen Strategy document (33). As developed since, connections and disconnections between to provide) effective contributions to poverty notably at the World Conference on Environ- poverty reduction and conservation, reflect- reduction, including both net benefits to the ment and Development in Rio de Janeiro in ing positions in the current debate. It includes poor and the avoidance of significant local 1992 and the World Summit on Sustainable both the moral and pragmatic dimensions of costs to any social group. Conservation Development in Johannesburg in 2002, the arguments for the conservation of biodiversity organizations will invest in addressing the concept of sustainable development was and the reduction of poverty. Disentangling poverty of critical protected-area neighbors extended to make explicit reference to justice, these makes for clarity. and actors with the power to disrupt conser- equity, and the elimination of poverty. World 1) Poverty and conservation are separate vation programs. Examples of policy action leaders agreed that biodiversity and resource policy realms. This position sees poverty include classic park outreach strategies (such conservation must be fully integrated into elimination and conservation as quite differ- as service provision to neighboring villages, strategies for economic development and are ent problems comprising distinct sectors of employment for local people, and participa- essential elements of sustainable livelihoods policy concern. Thus, conservation is a tion in park planning processes) and income- at local scales (35). It is widely argued that legitimate objective that can be pursued generating projects (such as sharing revenue biodiversity underpins the livelihood strate- independently of any benefits in poverty from wildlife tourism in protected areas, gies of the rural poor (16). These political and reduction (and vice versa). Under this integrated conservation-development proj- policy insights have been accompanied by the position, conservation strategies would focus ects, or the provision of locally acceptable emergence of new academic subfields that on the establishment of protected areas or alternatives to lost resources) (41, 43). offer integrative transdisciplinary insights into approaches such as direct payments (46). If 3) Conservation should not compromise social-ecological systems (36). poverty is recognized as an important cause poverty reduction. This position recognizes Sceptics point to the large element of of conservation failure, the response is the that conservation agencies have conservation wish fulfillment in arguments about the pos- designation of further critical biodiverse as their primary goal, but it holds that in sibility of win-win solutions in sustainable habitat and the stronger defense of protected pursuing that goal they should, at a mini- development (1, 33, 37, 38). A strong body areas, rather than the dissipation of scarce mum, not increase poverty or undermine the of opinion, however, maintains that poverty conservation resources to maintain diversity livelihoods of the poor. This position was elimination and conservation can happen across landscapes or in poverty alleviation adopted at the Fifth World Parks Congress in together. The term ‘‘pro-poor conservation’’ activities (37, 38). This position sees conser- 2003, but has its critics (27). Examples of has been used to identify conservation vation benefiting poverty reduction indirect- strategies resulting from this position might strategies that are designed to deliver both ly where it secures ecosystem services that include codes of conduct for conservation poverty reduction and biodiversity protection yield economic benefits to society, such as organizations, social impact assessment of (39, 40). But is this confidence in win-win enhanced water yields from forested catch- protected areas (29, 30), and the payment of solutions justified? Lasting positive out- ments (47, 48). There may also be local the full local opportunity costs of conserva- comes of conservation-with-development opportunities for win-win strategies that tion in protected areas (50). Conservation projects are elusive (41, 42). Projects that combine biodiversity and poverty reduction strategies might also seek to generate posi- seek to integrate conservation and develop- [such as protected-area tourism arrangements tive economic benefits for local communities ment have tended to be overambitious and (49)]. However, this position holds that within constraints of biodiversity conserva- underachieving (41–44). Although it is de- trying to combine conservation with poverty tion targets, for example through nonextrac- sirable to satisfy the goals of biodiversity reduction everywhere risks misallocating tive use [such as ecotourism (49)] or

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1147 R EVIEW harvesting within sustainable limits [such as development of humankind, to have regard larger challenge is to allow human society to safari hunting, medicinal products, or bio- to their demands, or footprint, on the bio- meet its potential and share the fruits of mass products (51, 52)]. This position differs sphere (3, 58, 59). economic growth while sustaining a biosphere from the empirical claim in position 2 that Different agencies (and different individ- that not only sustains full ecological functions poor people, if ignored, will undermine uals) are likely to wish to adopt different po- but retains its living diversity (3, 34). conservation. Rather it reflects independent sitions. For example, differences in thinking moral and political obligations on conserva- about the balance to be struck between pov- References and Notes tion agencies to take account of human erty reduction and biodiversity conservation 1. S. E. Sanderson, K. H. Redford, Oryx 37, 1 (2003). poverty. It is a claim that recognizes that underlie different positions in the ‘‘parks ver- 2. S. Palumbi, Science 293, 1786 (2001). 3. W. M. Adams, Against Extinction: The Story of conservation action can be sustained despite sus sustainable use’’ debate (37, 38, 54, 60). Conservation (Earthscan, London, 2004). negative social impacts (53). It applies even Those advocating strictly enforced protected 4. C. R. Margules, R. L. Pressey, Nature 405, 243 (2000). where it is possible to do conservation areas in poor developing countries to guaran- 5. A. S. Rodrigues et al., Nature 428, 640 (2004). 6. A. G. Brunner, R. E. Gullison, R. E. Rice, G. A. B. da effectively without benefiting poor people. tee the maintenance of populations of vul- Fonseca, Science 291, 125 (2001). 4) Poverty reduction depends on living nerable species (such as forest primates) are 7. http://sea.unep-wcmc.org/wdbpa/ (6 July 2004). resource conservation. This position rests on adopting position 1, treating the problems of 8. www.biodiv.org (22 February 2004). 9. M. Norton-Griffiths, C. Southey, Ecol. Econ. 12, 125 the empirical claim that financially poor and extinction and poverty as separate. Those (1995). socially and politically marginalized people advocating programs to tackle the poverty of 10. D. Brockington, Fortress Conservation: The Preserva- depend on living species in biodiverse eco- people living around such parks in order to tion of the Mkomazi Game Reserve, Tanzania (Currey, Oxford, 2002). systems for livelihoods and ecosystem ser- persuade them not to trespass or hunt are 11. C. L. Fortwangler, in Contested Nature: Promoting vices, and that their livelihoods can be adopting position 2, seeing poverty as a crit- International Biodiversity with Social Justice in the improved through appropriate conservation ical constraint on conservation. Those who Twenty-First Century, S. R. Brechin, P. R. Wilshusen, C. L. Fortwangler, P. C. West, Eds. (State Univ. of activities (33). Conservation is therefore a would seek to increase the flow of revenues New York Press, Albany, NY, 2003), pp. 25–40. tool for achieving poverty reduction, with from such parks to a level that would fully 12. M. Colchester, Salvaging Nature: Indigenous Peoples, the sustainable use of natural resources being compensate all stakeholders for associated op- Protected Areas and Biodiversity Conservation (World a foundation of strategies for achieving portunity costs of the park are adopting posi- Rainforest Movement, Montevideo, 2002). 13. K. Ghimire, M. Pimbert, Social Change and Conser- poverty reduction and social justice. Biodi- tion 3, attempting to ensure that conservation vation (Earthscan, London, 1997). versity benefits not immediately necessary to does not increase poverty in any way. Those 14. C. Geisler, R. de Sousa, Public Adm. Dev. 21, 159 (2001). this goal are a secondary gain. This position who propose conservation strategies building 15. M. Wells, Ambio 21, 237 (1992). 16. D. Roe, J. Elliott, Oryx 38, 137 (2004). might lead to the rejection of a protected-area on the needs of local communities for sustain- 17. World Bank Poverty Net, www.worldbank.org/poverty. strategy because, except under special cir- able harvests of wild species resources, and 18. A. Sen, Development as Freedom (Oxford Univ. Press, cumstances (for example, where shares of not necessarily a formally declared protected Oxford, 2001). 19. www.chronicpoverty.org/chronic_poverty_report_ ecotourism revenues exceeded all other forms area at all, are adopting position 4, seeing 2004.htm (7 July 2004). of land use), protected areas were unlikely to conservation strategies based on sustainable 20. www.developmentgoals.org (10 March 2004). achieve poverty reduction goals. Alternative use primarily as a means to reduce poverty. 21. www.worldbank.org/poverty/strategies/index.htm (19 June 2004). approaches would include the sustainable use Policy that fails to take account of the 22. J. Bojo, R. C. Reddy, Poverty Reduction Strategies and of living resources to optimize economic diverse relationships between conservation Environment (World Bank Environment Department return and/or positive impacts on the rural needs and the demands of poverty reduction, Paper 86, World Bank, Washington, DC, 2002). 23. www.un.org/esa/sustdev/natlinfo/nsds/map2002.htm or urban poor (54). Examples of policy and the related consumptive demands of the (19 June 2004). include conservation programs outside pro- growing world economy, risks failure (1). 24. D. Roe, in The Millennium Development Goals: Hitting tected areas; for example, to promote the Organizations committed to the preservation the Target or Missing the Point? (IIED, London, 2003), local management of common-pool resources of species and those committed to sustain- pp. 55–70. 25. A. Phillips, George Wright Forum 20, 8 (2002). within the constraints of ecological sustain- able rural livelihoods based on natural re- 26. S. Ramphal, in Parks for Life: Report of the IVth ability such as fisheries, wildlife, grazing, or source use are likely to engage with issues of World Congress on National Parks and Protected forestry that are targeted at improving the poverty and biodiversity in very different Areas, J. McNeely, Ed. (IUCN, Gland, Switzerland, 1993), pp. 56–58. livelihoods of the poor (54–56). Conservation ways. Their interactions will be facilitated if 27. A. Steiner, New Sci. 180, 21 (2003). in response to this position tends toward the they can understand their mutual positions. 28. www.iucn.org/themes/wcpa/wpc2003/english/outputs/ maintenance of yields of harvestable species The recognition of different starting points in recommendations.htm (10 March 2004). 29. D. Brockington, K. Schmidt-Soltau, Oryx 38, 140 and ecosystems rather than the preservation of the way in which biodiversity conservation (2004). biodiversity. Outcomes may deviate to a and poverty elimination goals are prioritized 30. C. Geisler, in Contested Nature: Promoting Interna- greater or lesser degree from biodiversity con- is essential if there is to be success in iden- tional Biodiversity Conservation with Social Justice in the Twenty-First Century, S. R. Brechin, P. R. Wilshusen, servation targets. This principle is reflected in tifying common ground and differences C. L. Fortwangler, P. C. West, Eds. (State Univ. of New the ‘‘ecosystem approach’’ adopted by the between biodiversity and development orga- York Press, Albany, NY), pp. 217–229. CBDin2000(57). nizations. Such recognition will facilitate the 31. World Bank, World Development Report 1992: Devel- opment and the Environment, (Oxford Univ. Press for task of those who believe that the goals must the World Bank, New York, 1992). Conclusion be achieved together. 32. R. F. Dasmann, J. P. Milton, P. H. Freeman, Ecological No position outlined here suggests that either It is premature to abandon attempts to Principles for Economic Development (Wiley, Chich- the conservation of biodiversity or the elim- combine conservation and development. The ester, UK, 1973). 33. W. M. Adams, Green Development: Environment and ination of poverty are improper goals. All elimination of poverty and the preservation of Sustainability in the Third World (Routledge, London, positions are consistent with the call for con- biodiversity are two distinct objectives. Each 2001). servation organizations to identify and mon- may be driven by different moral agendas, 34. S. E. Sanderson, K. H. Redford, Oryx 38, 146 (2004). 35. Livestock and Wildlife Advisory Group, Wildlife and itor the social impacts of their work, and to but there is considerable overlap in practice. Poverty Study (Department for International Devel- take corporate responsibility for operating in At the local scale, the policy need is to opment, London, 2002). a socially accountable manner (29). They are reconcile the interests of different stake- 36. F. Berkes, Conserv. Biol. 18, 621 (2004). 37. J. Terborgh, Requiem for Nature (Island Press, also all consistent with the need for poverty holders in the management of the natural Washington, DC, 1999). alleviation efforts, and wider projects for the resources of biodiverse ecosystems (45). The 38. J. F. Oates, Myth and Reality in the Rain Forest: How

1148 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EVIEW

Conservation Strategies Are Failing in West Africa 45. T. Kepe, M. Saruchera, W. Whande, Oryx 38, 143 54. J. Hutton, N. Leader-Williams, Oryx 37, 215 (2003). (Univ. of California Press, Berkeley, CA, 1999). (2004). 55. T. Franks, Issues Nat. Resour. Manage. 1, 2 (2003). 39. IUCN, Beyond Rhetoric: Putting Conservation to Work 46. P. J. Ferraro, Science 298, 1718 (2002). 56. I. Koziell, J. Saunders, Eds., Living Off Biodiversity: for the Poor (IUCN, Gland, Switzerland, 2002). 47. A. Balmford et al., Science 297, 950 (2002). Exploring the Livelihoods and Biodiversity Issues in 40. D. Roe, J. Hutton, J. Elliott, K. Chitepo, M. Saruchera, 48. G. C. Daily, Ed., Nature’s Services: Societal Depen- Natural Resources Management (IIED, London, 2001). Policy Matters 12, 52 (2003). dence on Natural Ecosystems (Island Press, Wash- 57. www.biodiv.org/decisions/default.aspx?m0cop-05 (6 41. D. Hulme, M. W. Murphree, Eds., African Wildlife & ington, DC, 1997). July 2004). Livelihoods. The Promise and Performance of Com- 49. S. Go¨sling, Ecol. Econ. 29, 303 (1999). 58. M. L. Imhoff et al., Nature 429, 870 (2004). munity Conservation (Currey, Oxford, 2001). 50. A. Balmford, T. Whitten, Oryx 37, 238 (2003). 59. M. Wackernagel, W. Rees, Our Ecological Footprint: 42. J. S. Murombedzi, J. Int. Dev. 11, 287 (1999). 51. B. M. Campbell, M. K. Luckert, Uncovering the Hidden Reducing Human Impact on Earth (New Society 43. C. S. Barrett, P. Arcese, World Dev. 23, 1073 (1995). Harvest: Valuation Methods for Woodland and Forest Publishers, Gabriola Island, British Columbia, 44. M. Wells, K. Brandon, People and Parks: Linking Resources (Earthscan, London, 2002). 1996). Protected Areas with Local Communities (World 52. D. S. Wilkie, J. F. Carpenter, Oryx 3, 338 (1999). 60. P. R. Wilshusen, S. R. Brechin, C. L. Fortwangler, P. C. Bank, Washington, DC, 1992). 53. D. Brockington, Policy Matters 12, 22 (2003). West, Soc. Nat. Resources 15, 17 (2002).

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1149 BREVIA These results challenge the hypothesis that southern brown bears are descended Pleistocene Brown Bears in the from Beringian populations dispersing through the IFC È13 to 12 ka, because those Mid-Continent of North America dispersers would have belonged to clade 2 or 3. With clade 4 bears inhabiting central 1 2 3 Paul Matheus, * James Burns, Jaco Weinstock, Alberta È26 ka, a more parsimonious model Michael Hofreiter4 is that clade 4 bears penetrated into southern regions well before the LGM, that they Current biogeographic models hypothesize Museum of Alberta no. P98.5.374) collected in became isolated south of the ice during the that brown bears (Ursus arctos) migrated fluvial gravels near Edmonton, Alberta (Fig. 1, LGM, and that modern bears there are from Asia to North America via east Beringia fig. S1, and table S1). Two accelerator descended from in situ populations. (unglaciated Alaska and Yukon) È100 to 50 radiocarbon dates on collagen returned ages of By implication, northern and southern thousand radiocarbon years ago (ka) but did 25,210 T 560 years (AA48743) and 27,410 T female brown bears in North America have not reach areas south of Beringia until the 200 years (OxA-12902). This find indicates been genetically isolated from each other for opening of a mid-continental ice-free corri- that brown bears reached areas south of at least 35,000 years. Furthermore, the first dor (IFC) È13 to 12 ka (1, 2). This model Beringia well before the postglacial period appearance of brown bears south of the LGM has been problematic because migration to and before the coalescence of the Laurentide ice margin should not be used to date the the mid-continent was blocked by glacial ice and Cordilleran glaciers. earliest availability of a late glacial IFC for only during the relative- human expansion in the ly short period È23 to New World. 12 ka, meaning the mid- continent was largely ice- References and Notes free before the Last Glacial 1. B. Kurte´n, E. Anderson, Pleisto- Maximum (LGM), È24 to cene Mammals of North Amer- ica (Columbia Univ. Press, New 18 ka (3–5). York, 1980). Thirty-one brown 2. I. Barnes, P. Matheus, B. Shapiro, bear fossils have been D. Jensen, A. Cooper, Science 295, 2267 (2002). radiocarbon-dated in east 3. R. R. Young, J. A. Burns, D. G. Beringia (2). Three have Smith, L. D. Arnold, R. B. Rains, infinite ages (953 ka), Geology 22, 683 (1994). G 4. A. S. Dyke et al., Quat. Sci. Rev. the rest are 48 ka, and 21, 9 (2002). there is a hiatus in their 5. J. A. Burns, Quat. Int. 32, 107 record from È35 to 21 (1996). 6.J.A.Leonard,A.K.Wayne, ka, suggesting brown A. Cooper, Proc. Natl. Acad. bears were absent during Sci. U.S.A. 97, 1561 (2000). that interval (2). Ancient 7. Materials and methods are available on Science Online. mitochondrial DNA stud- 8. Primary DNA sequencing was ies have shown that three performed at the Henry Well- clades of brown bears come Ancient Biomolecules inhabited east Beringia Centre, and replication at the Max Planck Institute for Evolu- before 35 ka: clades 2c, tionary Anthropology. Partly 3c, and 4 (2) (Fig. 1). supported by the Bureau of Populations that recolo- Land Management, U.S. De- partment of the Interior. nized east Beringia after 21 ka all belonged to sub- Fig. 1. (A) Brown bear cranial fragment P98.5.374 recovered near Edmonton, Alberta. Scalebar,3cm.(B) The maximum extent of glaciation (gray shading) in northwestern Supporting Online Material clades that were not pres- www.sciencemag.org/cgi/content/ ent before 35 ka (2). Clade North America during the LGM (4) and the present distribution of brown bear clades (6). (C) Brown bear clades in North America over the past 50,000 years (2, 6). Dashed lines full/306/5699/1150/DC1 4 never reappeared in east Materials and Methods indicate that presence is presumed even though fossils are lacking. The box in the upper Fig. S1 Beringia. After 10 ka, the left represents populations in Pleistocene east Beringia. AK, Alaska; YT, Yukon Territory; BC, Table S1 modern phylogeographic British Columbia; NWT, Northwest Territories; AB, Alberta; CA, Canada; US, United States. References and Notes structure of North Ameri- can brown bears became established: clade 3a We extracted mitochondrial DNA from 15 June 2004; accepted 8 September 2004 in western and central Alaska, clade 3b in È0.5 g of the root of the specimen_s extreme eastern Alaska and northwestern second molar, using established ancient- Canada, and clade 4 in southern Canada and DNA techniques (2, 7). Two nonoverlap- 1Alaska Quaternary Center, University of Alaska Fairbanks, Fairbanks, AK, USA. 2Provincial Museum the contiguous United States (6). It has been ping but highly variable fragments (60 and 3 difficult to explain the descendency of bears 135 base pairs) of the control region were of Alberta, Edmonton, Canada. Henry Wellcome Ancient Biomolecules Centre, University of Oxford, in the mid-continent if clade 4 has been amplified, and the sequences obtained Oxford, UK. 4Max Planck Institute for Evolutionary absent from North America since 35 ka. show that P98.5.374 belongs to clade 4 Anthropology, Leipzig, Germany. Here, we report a well-preserved cranial (GenBank accession nos. AY796010 and *To whom correspondence should be addressed. fragment from a brown bear (Provincial AY796011). E-mail: [email protected]

1150 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org REPORTS dependence of the phase of the superconduct- ing order parameter is probed directly through Odd-Parity Superconductivity the Josephson effect. In the presence of the

spin-orbit coupling, a Josephson current js in Sr RuO between an odd- and an even-parity supercon- 2 4 ductor is expected to be K. D. Nelson,1 Z. Q. Mao,1* Y. Maeno,2,3 Y. Liu1. ºb à E à I ^À js Reðc21s21Þ Im Y d ðn  kÞ FS ð1Þ

Phase-sensitive measurements were made on Sr2RuO4 to establish un- (18, 19), where c21 and s21 are the spin-orbit ambiguously the odd-parity pairing in this material. The critical current of and spin-independent parts of the transmis- Au0.5In0.5-Sr2RuO4 superconducting quantum interference devices prepared sion matrix, k is the wave vector, n is the unit on Sr2RuO4 single crystals was found to be a maximum for devices with vector normal to the junction plane, and y junctions on the same side of the crystal and a minimum for devices with and d are functions defining the supercon- junctions on opposite sides, in the limit of zero magnetic flux; these findings 0 ductor order parameters with Deven iy(k)sy indicate that the phase of the superconducting order parameter in Sr2RuO4 0 E I ^ for even parity and Dodd i s d(k) sy for changes by p under inversion. This result verifies the odd-parity pairing 0 odd parity (1). Here, s (sx, sy, sz) denotes symmetry and the formation of spin-triplet Cooper pairs in Sr2RuO4. the Pauli matrices, * represents complex conjugation, Re and Im are the real and

Superconductivity originates from the con- Sr2RuO4 (4), the only layered-perovskite imaginary parts of a complex number, and bÀ densation of Cooper pairs into a macroscopic superconductor containing no Cu (5), was FS denotes an appropriate average over the quantum state described by a single wave predicted to be an odd-parity superconductor Fermi surface. Under the parity (inverting) function, also known as the superconducting (6, 7). A body of experimental evidence for operation k Y –k, y(k) is even but d(k)is order parameter. Because of the Fermi odd-parity superconductivity in Sr2RuO4 has odd. statistics of the electron, quantum mechanics been obtained Ereviewed in (8)^. To date, the Interesting consequences are expected from demands that the parity of the wave function strongest evidence for odd-parity supercon- Eq. 1. For example, if Sr2RuO4 is an odd- of the Cooper pairs be either even or odd. In ductivity in Sr2RuO4 is found in the nuclear parity superconductor with d along the c I the absence of spin-orbit coupling and crystal magnetic resonance (NMR) Knight shift axis, d (n  k), and therefore js,willbe field, both spin and angular momentum of the result. With a magnetic field applied along finite for an in-plane junction (n and k in Cooper pairs are good quantum numbers. The the ab plane, the Knight shift (proportional the ab plane) and zero for a c-axis junction total spin is 0 in even-parity and 1 in odd- to the spin susceptibility after the orbital (with n¬d), a selection rule found previously parity superconductors. Consequently, having contribution is subtracted), measured care- (20). In addition, js through two parallel junc- an even or odd parity in the order parameter fully at both O(1) (9) and Ru (10) sites, was tions prepared on the same odd-parity super- corresponds to spin-singlet or spin-triplet found to be essentially a constant across the conductor but with normal directions n and –n, superconductivity, respectively. In real crys- superconducting transition temperature Tc of respectively (Fig. 1A), will be out of phase tals with finite spin-orbit coupling and a Sr2RuO4, a result confirmed by polarized- by p (so long as the sample configuration is crystal field, strictly speaking, neither spin neutron scattering measurements (11)and such that n is not parallel to d, in which case 0 nor angular momentum is a good quantum interpreted (assuming weak spin-orbit cou- js 0). If such a pair of junctions is in- number. In this case, the parity becomes the pling in Sr2RuO4) as a consequence of spin- corporated into a superconducting quantum parameter more appropriate for characterizing triplet pairing in this material. interference device (SQUID), as suggested the pairing symmetry (1). Nonetheless, these Although this interpretation of the NMR by Geshkenbein, Larkin, and Barone (GLB) two classes of superconductors are still fre- results is recognized, it is also known that (21), the critical current (Ic) of the SQUID quently referred to as spin-singlet and spin- other reasons unrelated to pairing sym- would be expected to be modulated by F, the triplet superconductors, and furthermore, as metry, such as strong spin-orbit coupling flux threading the SQUID, in the form s-ord-wave superconductors for the former (12) or peculiarities of the electronic struc- I º cos EðF=F þ ½Þp^ ð2Þ and p-orf-wave superconductors for the lat- ture (13),canalsoleadtoconstantspin c 0 3 0 –15 ter. Studies of odd-parity superfluid He (2) susceptibility. For example, vanadium, an where F0 2.07  10 Wb is the flux 0 and other systems where odd-parity pairing s-wave superconductor, was found long ago quantum. Therefore, Ic(F 0) corresponds is likely realized, ranging from neutron stars to have a constant Knight shift across Tc to a minimum. In contrast, for a conventional (2) to superconductor-ferromagnet microstruc- (14). YBa2Cu3O7–d,ad-wave superconduc- SQUID—including a SQUID prepared with tures (3), have revealed highly unusual phys- tor, was also found to possess a constant both junctions on the same surface of an 0 ical phenomena, providing strong motivation Knight shift across Tc when measured at odd-parity superconductor—Ic(F 0) corre- for the long-standing quest for an example of the Cu(2) site with the field applied along sponds to a maximum. The resistance of the odd-parity superconductivity. the c axis (15). Therefore, further work was SQUID, measured by applying a current

needed to establish unambiguously the odd- larger than Ic, will exhibit a maximum 1 0 Department of Physics, Pennsylvania State Univer- parity superconductivity in Sr2RuO4 (8). instead of a minimum at F 0. As noted 2 sity, University Park, PA 16802, USA. Department of Work on high-Tc superconductors (16, 17) previously (22), however, the GLB experi- Physics, Kyoto University, Kyoto 606-8502, Japan. 3 has shown that phase-sensitive measure- ment does not work for certain sample International Innovation Center, Kyoto University, ments, which played a key role in settling configurations (e.g., those in which n is Kyoto 606-8501, Japan. the debate on pairing symmetry in the high- parallel to d). *Present address: Department of Physics, Tulane T arena, will enable the most definitive We pursued the GLB experiment in University, New Orleans, LA 70118, USA. c .To whom correspondence should be addressed. determination of the parity of the order param- structures (Fig. 1A) prepared on high-quality 0 E-mail: [email protected] eter. In such an experiment, the orientation single crystals of Sr2RuO4 (Tc 1.4 K)

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1151 R EPORTS A I+ B C V+ SiO 5 I AuIn-SRO A d Au0.5 In0.5 4 ) Ω

La Lb 3 I AuIn-SRO C h λ2 s I I a H b 2 Resistance ( AuIn-SRO B 100 nm λ1 λ 1 c f w H V− 0 b 0 0.5 1 1.5 2 2.5 3 Sr2RuO4 a I− Temperature (K) 0 0 0 0 Fig. 1. (A) Schematic of Au0.5In0.5-Sr2RuO4 GLB SQUID with measurement h 0.5 mm, and d 0.33 mm; for sample B, w 1.15 mm, h 0.4 mm leads. Both junctions in the SQUID are in-plane tunnel junctions to ensure (left) and 0.6 mm (right), and d 0 0.15 mm; for sample C, w 0 0.68 mm, m 0 0 js 0. The side insulated by the SiO layer may cross the ab plane as h 0.3 mm, and d 0.4 mm. The SiO layer thickness is 150 nm. (B) depicted (samples A and B) or cross a third face perpendicular to the ab Equivalent circuit of the SQUID. I and I are the current on the left and the 0 a b plane (sample C). The shaded area indicates the flux penetration at T 0, right side of the SQUID loop, La and Lb are effective inductances, I is total , , l1 3.7 mm, and l2 0.18 mm. For sample C, the flux penetration is to a current, and Is is circulating current. (C) Sample resistance R as a function depth of l2 on all three sides. The value of lf is not known but may be of T for three GLB SQUIDs. A smooth R(T) across the Tc of Sr2RuO4 (around slightly larger than that of pure In, 0.07 mm. For sample A, w 0 1.05 mm, 1.4 K) indicates that R(T) is dominated by the tunnel barrier.

prepared by the floating zone method. This 1.5 15 required us to overcome several technical A B difficulties (23). First, a suitable convention- 1 AuIn-SRO B

al superconductor had to be identified. A)

0.5 µ 10

Obtaining a SQUID with minimal asymme- V) try and self-inductance (see below) requires µ 0.4 K 0 thermal evaporation of the conventional 0.045 K

superconductor onto mechanically polished Voltage ( -0.5 5 AuIn-SRO C

single-crystal surfaces. However, few ele- Critical current ( mental or alloy superconductors were found -1 AuIn-SRO A to form a Josephson junction with Sr2RuO4. AuIn-SRO A Alloy Au1–xInx prepared by interdiffusion -1.5 0 was the only material we found with which -6 -4 -2 0 246 0 0.1 0.2 0.3 0.4 0.5 0.6 Current (µA) Temperature (K) Josephson coupling to bulk Sr2RuO4 could be achieved with the above requirements. Fig. 2. (A) Current-voltage curves of sample A at several temperatures. T 0 0.045 K (triangles), Also, the presence of a minority phase in 0.250 K (diamonds), 0.350 K (squares), and 0.400 K (circles), showing a zero-voltage supercurrent. Sr RuO —the Ru inclusions referred to as The asymmetry with respect to the current reversal is due to the presence of a circulating current. 2 4 (B) The temperature dependence of the critical current, I (T), determined from a particular the 3 K phase of Sr RuO (24)—had to be c 2 4 (positive) current direction for all three samples. dealt with carefully. These inclusions are known to form in all single-crystal rods –5 2 during the crystal growth. Whereas the Tc metric, which suggests that the two junctions ple A and 2.2 Â 10 cm for sample B, which 0 0 for the bulk Ru is 0.5 K, the Tc of the Ru in the SQUID are not identical. The critical is expected to yield DH F0/Aeff 9mG inclusions embedded in the bulk Sr2RuO4 temperature of the SQUID determined from for both junctions. For sample C, Aeff is 1.8 Â 0 0 –6 2 0 crystal can be as high as 3 K. The presence Ic(T Tc) 0 is consistent with that of 10 cm with the expected DH 117 mG. of these Ru inclusions at the junction Au0.5In0.5 (23). Furthermore, quantum inter- We attribute the difference between the ex- interface could potentially lead to unwanted ference patterns in both the critical current pected and measured values of DH to the complications (23). We found that close to and sample resistance were observed (Fig. 3). possible local variations of the penetration the surface of a single-crystal rod, there The period in magnetic field (H) was found depths and the misalignment of the samples existed areas free of Ru inclusions. There- to be 8.5 T 0.4 mG for sample A, 15.7 T with respect to the applied field. fore, SQUID samples were prepared on these 0.4 mG for sample B, and 98 T 2 mG for Two important issues must be addressed 0 areas. The presence of Ru inclusions at the sample C. The direction of the c axis for before we can show whether Ic(F 0) junction interface was checked by examining samples A and B is as indicated in Fig. 1A, corresponds to a minimum or maximum,

resistance as a function of temperature to see whereas for sample C it is pointing out of the and therefore whether Sr2RuO4 is an odd- whether any features were present between page. On the basis of the relevant lengths parity superconductor. The total flux F is 0 1.5 and 3 K (23). associated with these samples, and taking given by F Fapp þ Find þ Ftrap þ Fbkgd, 0 The current-voltage characteristics of a into account that the areas of the two junc- where Find is the induced flux, Fapp ( Aeff H ) Ru-free GLB SQUID are shown in Fig. 2. tions in the SQUID are large and that the is the applied flux, Ftrap is the trapped Josephson coupling marked by a finite, zero- phase change across the two junctions is flux, and Fbkgd is the flux due to any voltage supercurrent is apparent. However, position-dependent, the effective area of the background field. To find out what values –5 2 0 the current-voltage characteristics are asym- SQUID loop (Aeff)is2.1Â 10 cm for sam- of Fapp would result in F 0, it is necessary

1152 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS 4 0.6 we found that applying a large field appeared B 3.5 A AuIn-SRO A 4 µA to lead to an asymmetric Ic(F), a sign that 0.5 Ftrap was present. The applied field was 3 ) 0.30 therefore kept below a relatively small value. A) Ω

µ 0.4 2.5 Finally, Fbkgd is minimized by the good magnetic shield. 2 0.3 µ 3 A These strategies to minimize Find, Ftrap,

1.5 Resistance ( 0.29 and Fbkgd appeared to work fairly well: 0.2 Nearly symmetric patterns of Ic(H), taken at Critical current ( 1 different temperatures, were obtained on 0.1 0.5 three fairly symmetric GLB SQUID samples. AuIn-SRO A The actual residual field contributing to the 0 0.28 0 -80 -60 -40 -20 0 20 40 60 80 residual flux (F F ) could be inferred -120 -80 -40 0 40 80 120 ind þ trap Magnetic field (mG) Magnetic field (mG) from the position of the center of the envelope in Ic(H) (1 to 2 mG for sample A; Fig. 3. Examples of quantum oscillations in Ic(H)(A) and R(H)(B). Values of the current used to see Fig. 4A). For such nearly symmetric measure R(H) are indicated. Both sets of data were taken on sample A at T 0 0.30 K. patterns, the overall Ic(H, T) decreased as the temperature increased to Tc, accompanied by shifts of the patterns that were clearly due to 5 8 the decrease in F (Fig. 4A). As a self- A GLB AuIn-SRO A B SS AuIn-SRO A ind 7 consistency check, we estimated the values c c 4 0.30 K of L from the shift in Ic(H, T) and from the 0.21 K 6 sample dimensions and found a reasonable A) A) µ µ agreement (23). 3 5 The detailed evolution of the Ic(H, T ) 4 shown in Fig. 4A clearly shows that Ic(F)is a minimum at F 0 0. Other GLB SQUIDs, 2 0.33K 3 samples B and C, showed essentially the Critical current (

Critical current ( same behavior (23). On the other hand, 2 1 control samples with both junctions of the 1 0.475 K SQUID on the same side (SS) of the crystal 0.42 K (perpendicular to the ab plane) were found to 0 0 possess an I (F) interference pattern with a -20 -15 -10 -5 0 5 10 15 20 -20 -15 -10 -5 0 5 10 15 20 c maximum at F 0 0 (Fig. 4B), as expected Magnetic field (mG) Magnetic field (mG) given that the phase of the order parameter is Fig. 4. (A) I (H) for GLB sample A (opposite-side GLB SQUID) taken close to T of the SQUID. From the same at both junctions. We also prepared c 0 c top to bottom, T 0.210, 0.250, 0.300, 0.330, 0.360, 0.380, 0.400, and 0.420 K. A dashed line is another type of control sample in a corner- used to highlight the shift of the minimum in I (F )towardF 0 0asT approaches T .(B) I (H) c app app c c junction configuration, which was used in the for same-side (SS) sample A, showing a maximum Ic(Fapp)asT approaches Tc.Fromtopto bottom, T 0 0.300, 0.350, 0.400, 0.425, 0.450, and 0.475 K. For each curve, the field was swept in high-Tc work (16). In the interference pattern both directions to ensure that flux was not trapped during the sweep. for such a corner junction, the critical current at F 0 0 is a maximum for an s-wave, a mi- nimum for a d-wave, and neither a minimum to determine the value of Find and to avoid or than pressing the bulk In wire as done nor a maximum for a p-wave superconductor. minimize Ftrap and Fbkgd (25). The induced previously (20) onto the polished Sr2RuO4 Experimentally, a Au0.5In0.5-Sr2RuO4 corner 0 0 flux Find LIs LaIa – LbIb (Fig. 1B) is surface, to obtain a sample with minimal junction was found to show an interference intrinsic for SQUIDs with large areas and an inductance and asymmetry. Furthermore, we pattern consistent with p-wave pairing (23). unavoidable asymmetry between the two measured systematically the Ic(Fapp, T)of Our observations indicate that the phase junctions in the SQUID, whereas the trapped selected SQUID samples with minimal asym- of the superconducting order parameter in , , flux Ftrap is typically a by-product of the metry (Ia Ib and La Lb)asT was raised to Sr2RuO4 changes by p after inversion. This cooling and measurements (26). As shown in Tc. In this case, Ia and Ib would each decrease establishment of the odd-parity superconduc- 0 Y 0 the high-Tc work (16), the presence of Ftrap to zero, so that LIs LaIa – LbIb 0 and F tivity in Sr2RuO4 by phase-sensitive experi- Y 0 in an ordinary SQUID could mimic the Fapp þ Find Fapp,ifFtrap 0. ments constrains the interpretation of other behavior of an unconventional SQUID. To avoid Ftrap, we carefully shielded the pairing symmetry–related experiments to the The following steps were taken to address SQUID magnetically (23) to minimize the odd-parity scenario, which in turn provides a these issues: We sought to reduce L and background field (G1 mG) and cooled it solid starting point to determine the detailed therefore Find by minimizing the width of the down slowly to initialize a Ftrap-free state, as k-dependence of the order parameter. This B SQUID w, which was unfortunately limited done previously (28). The absence of Ftrap will help in resolving, for example, the p- by the tendency of the crystal to cleave dur- was ensured by verifying that the envelope versus f-wave[ issue (8). The tetragonal ing the mechanical polishing. Using thin of Ic(F) was symmetric (28). We followed a crystal symmetry of Sr2RuO4 allows only films of Sr2RuO4 could, in principle, help; procedure in which the sample, the magnet several odd-parity states (29). The result of however, even the highest quality films of coil, and the magnetic shielding enclosure the selection rule in Josephson coupling (20),

Sr2RuO4 grown so far are not yet super- were warmed to 14 K or higher and cooled which can now be interpreted with consider- conducting (27). The two junctions in the back down slowly (with a uniform rate of able certainty to indicate that d is along the c SQUID were prepared simultaneously by dT/dt 0 2 K/hour) in zero field repeatedly to axis, reduces the number of allowed odd- thermally evaporating Au0.5In0.5 (23), rather obtain symmetric Ic(H) traces. In addition, parity states even further. The establishment

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1153 R EPORTS of an odd-parity superconductor may also 14. R. J. Noer, W. D. Knight, Rev. Mod. Phys. 36, 177 (1964). adjust themselves to accommodate any arbitrary, 15. S. E. Barrett et al., Phys. Rev. B 41, 6283 (1990). not necessarily quantized, amount of F . provide an arena to explore its possible trap 16. D. J. van Harlingen, Rev. Mod. Phys. 67, 515 (1995). 27. M. A. Zurbuchen et al., Appl. Phys. Lett. 78, 2351 applications. For example, all electronic spins 17. C. C. Tsuei, J. R. Kirtley, Rev. Mod. Phys. 72, 969 (2001). (2000). 28. B. Chesca et al., Appl. Phys. Lett. 76, 912 (2000). in superconducting, nonmagnetic Sr2RuO4 are condensed into the ab plane Ebecause d 18. V. B. Geshkenbein, A. I. Larkin, JETP Lett. 43, 395 29. V. P. Mineev, K. V. Samokhin, Introduction to ^ (1986). Unconventional Superconductivity (Gordon and is along the c axis (29) , a fact possibly useful 19. A. Millis, D. Rainer, J. A. Sauls, Phys. Rev. B 38, 4504 Beach, Amsterdam, 1999). for novel devices. (1988). 30. A. Wollman, D. J. van Harlingen, W. C. Lee, D. M. 20. R. Jin, Y. Liu, Z. Mao, Y. Maeno, Europhys. Lett. 51, Ginsberg,A.J.Leggett,Phys. Rev. Lett. 71, 2134 (1993). References and Notes 341 (2000). 31. Supported by NSF grants DMR-9974327 and DMR 1. G. E. Volovik, L. P. Gor’kov, Sov. Phys. JETP 61, 843 21. V. B. Geshkenbein, A. I. Larkin, A. Barone, Phys. Rev. B 0202534, Japan Society for the Promotion of Science (1985). 36, 235 (1987). and Ministry of Education, Culture, Sports, Science, and 2. D. R. Tilley, J. Tilley, Superfluidity and Superconduc- 22. A. J. Leggett, Philos. Mag. B 74, 509 (1996). Technology grants-in-aid, and the 21st Century Center tivity (A. Hilger, Bristol, UK, ed. 2, 1986). 23. See supporting data on Science Online. of Excellence program. We thank V. B. Geshkenbein, J. 3. F. S. Bergeret, K. B. Eftov, A. I. Larkin, Phys. Rev. B 62, 24. Y. Maeno et al., Phys. Rev. Lett. 81, 3765 (1998). Kirtley, C.-C. Tsuei, D. F. Agterberg, J. Banavar, J. K. Jain, 11872 (2000). 25. This issue was encountered in the pioneering work of A. J. Leggett, J. A. Sauls, M. Sigrist, D. J. van Harlingen,

4. S. N. Ruddlesden, P. Popper, Acta Crystallogr. 10, 538 SQUID-based phase-sensitive experiments on high-Tc and S.-K. Yip for useful discussions, and C. Andreou, R. (1957). superconductors (30). The SQUID resistance R(H, I), Jin, H. Wang, D. Schlom, and H. Yaguchi for discussions 9 5. Y. Maeno et al., Nature 372, 532 (1994). measured at I Ic (in which case finite Find was and assistance. 6. T. M. Rice, M. Sigrist, J. Phys. Condens. Matter 7, necessarily included), was extrapolated to I 0 0 and L643 (1995). the minimum position for R(Hmin, 0) was inferred. The Supporting Online Material 7. G. Baskaran, Physica B 223–224, 490 (1996). extrapolation was avoided in the elegant (single) www.sciencemag.org/cgi/content/full/306/5699/1151/ 8. A. P. Mackenzie, Y. Maeno, Rev. Mod. Phys. 75, 657 corner-junction experiment. DC1 (2003). 26. The requirement for the superconducting phase to be Materials and Methods 9. K. Ishida et al., Nature 396, 658 (1998). single-valued in the sample leads to the constraint Figs. S1 to S9 0 10. K. Ishida et al., Phys. Rev. B 63, 060507 (2001). 2pm F1 – F2 þ (2p/F0)(Fext þ Find þ Fbkgd þ References 11. J. A. Duffy et al., Phys. Rev. Lett. 85, 5412 (2000). Ftrap), where m is an integer (or more likely, 0) and 12. P. W. Anderson, Phys. Rev. Lett. 3, 325 (1959). F1 and F2 are the phase differences at the two 13. D. E. MacLaughlin, Solid State Phys. 31, 2 (1976). junctions in the SQUID, respectively. F1 and F2 can 10 August 2004; accepted 21 September 2004

susceptible to the emergence of new ordered Disorder-Sensitive Phase phases. Qualitatively, the high susceptibility to novel phase formation is thought to arise Formation Linked to Metamagnetic because tuning a thermal phase transition toward zero temperature flattens the free Quantum Criticality energy landscape, causing competing low- temperature phases to become nearly degen- S. A. Grigera,1* P. Gegenwart,1,2 R. A. Borzi,1 F. Weickert,2 erate. The characteristics of the quantum A. J. Schofield,3 R. S. Perry,1,4,5 T. Tayama,6 T. Sakakibara,6 critical fluctuations associated with particu- Y. Maeno,4,5 A. G. Green,1 A. P. Mackenzie1* lar quantum critical points (QCPs) can then tip the balance in favor of a particular new Condensed systems of strongly interacting electrons are ideal for the study of phase. A notable example is the formation quantum complexity. It has become possible to promote the formation of new of unconventional superconductivity in the quantum phases by explicitly tuning systems toward special low-temperature vicinity of antiferromagnetic QCPs (6). In quantum critical points. So far, the clearest examples have been appearances this case, the diverging spin fluctuations are of superconductivity near pressure-tuned antiferromagnetic quantum critical thought to provide the effective bosons for points. We present experimental evidence for the formation of a non- the Cooper pair binding. superconducting phase in the vicinity of a magnetic field–tuned quantum Superconductivity is not the only form of low-temperature order that can be adopted critical point in ultrapure crystals of the ruthenate metal Sr3Ru2O7,andwe discuss the possibility that the observed phase is due to a spin-dependent by correlated electron systems; the fractional symmetry-breaking Fermi surface distortion. quantum Hall state is a spectacular example of the subtle quantum self-organization that The field of quantum criticality continues to study of the collective behavior that arises in is possible (7). A particularly interesting attract widespread theoretical and experi- many-body quantum physics in the presence issue is whether forms of magnetically tuned mental attention because of its importance of strong interactions. Much of the interest QCPs can be used to explicitly promote the to the global effort to understand the has focused on the effects that quantum formation of other new types of quantum behavior of correlated electron systems (1–4). critical fluctuations have on itinerant sys- order (8) in the special situation in which the It gives the rare opportunity for controlled tems, notably the link between their strength magnetic field disfavors superconducting and the breakdown of Landau_s Fermi liquid ground states. We demonstrate the formation 1School of Physics and Astronomy, University of St. theory (5). Quantum criticality is now un- of a previously unknown, non-superconducting Andrews, North Haugh, St. Andrews, Fife KY16 9SS, derstood not to be a trivial extension of the phase in an applied magnetic field of 8 T, Scotland. 2Max-Planck Institute for Chemical Physics of Solids, D-01187 Dresden, Germany. 3School of classical case; quantum critical fluctuations near a metamagnetic QCP in Sr3Ru2O7.The Physics and Astronomy, University of Birmingham, can have surprising strength and subtlety, experiments place strong constraints on the Edgbaston, Birmingham B15 2TT, UK. 4International including important mode-mode interaction order parameter of this phase, which appear to Innovation Center, Kyoto University, Kyoto 606- terms (3, 4). be satisfied by invoking a spin-dependent 8501, Japan. 5Department of Physics, Kyoto Univer- 6 Recently several examples have been Fermi-surface instability, the possibility of sity, Kyoto 606-8502, Japan. Institute of Solid-State Physics, University of Tokyo, Kashiwa, Chiba discovered of a phenomenon that is poten- which was first discussed theoretically nearly 2778581, Japan. tially even more exciting, namely the use of half a century ago by Pomeranchuk in phe- _ *To whom correspondence should be addressed. quantum critical points to create regions of nomenological calculations using Landau s E-mail: [email protected]; [email protected] phase space in which systems are highly Fermi liquid theory (9).

1154 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS

Sr3Ru2O7 is the bilayer member of the independent (elastic) scattering mechanism. sample moment and length increase at each Ruddlesden-Popper series of layered perov- It is not, however, compelling thermodynamic transition. skite ruthenates whose single-layer member, evidence of the kind necessary to support the Temperature scans of thermodynamic

Sr2RuO4, is a well-studied unconventional postulate of a phase change. Such thermo- quantities also show well-defined features superconductor (10). The growth of high- dynamic signatures are provided by the at the boundaries of the anomalous region quality single crystals in an image furnace susceptibility, magnetostriction, thermal ex- seen in the resistivity, for example, the revealed it to be a strongly enhanced para- pansion, and magnetization. A sample temperature dependence of the magnetiza- magnet in zero field, displaying an itinerant comparison of alternating current (ac) sus- tion in a fixed field of 7.9 T (Fig. 2, inset). metamagnetic transition in applied fields in ceptibility (c) and linear magnetostriction (l) At temperatures above 1.1 K, the data show the range from 4.9 T (field parallel to ab is shown (Fig. 2). There is a clear correlation upward curvature consistent with divergent plane) to 7.9 T Efield parallel to c (11, 12)^. between the real part of c and l, indicating a fluctuations on the approach to a QCP, but Good evidence has been acquired that, for strong magnetostructural coupling. A low- this is then cut off at a well-defined kink, fields parallel to ab, the metamagnetism is field peak at about 7.5 T, which is probably a followed by a much weaker temperature de- due to a line of first-order phase transitions crossover, is followed by two sharp peaks at pendence. The kink is seen only on entry to terminating in a finite temperature critical fields corresponding precisely to the steep the anomalous region; scans at fields outside point. As the field is rotated away from the walls in the resistivity of Fig. 1. As discussed this range remain smooth to the lowest ab planes, this critical end point is tuned in (13, 16), the appearance of a dissipative temperature reached (50 mK). downward in temperature, becoming quan- peak in cµ indicates that each of these two The data shown in Fig. 2 represent only a tum critical at an angle less than 10- from c peaks corresponds to a first-order phase small subset of the experiments performed. (12–14). The crystals on which the work of boundary at 100 mK (17). All peaks in both A collation of all the relevant information is (12–14) was performed were of sufficiently c¶ and l are positive, showing that both presented (Fig. 3), which also includes points È high purity (residual resistivity rres 3 mW cm) that disorder might not have been expected to play an important role in de- Fig. 1. The resistivity r termining their properties. However, a fur- of very pure single– crystal Sr Ru O as a 2.1 ther improvement of nearly an order of 3 2 7 function of magnetic È 1.3K magnitude to rres 0.4 mW cm (15) revealed field (applied parallel substantial changes in the observed behavior. to c) at a series of Away from the QCP, the properties are temperatures between indeed essentially independent of purity, 0.1 and 1.3 K in steps but in its immediate vicinity a large and of 100 mK. Increasing 1.6 striking peak was observed in the resistivity, the temperature broad- ens and increases the in a region bounded at 100 mK by two first- resistivity outside the order magnetic phase transitions (16). Those field range of the pro- observations opened the possibility of the nounced central fea- 0.1K formation of a non-superconducting phase in ture, but within that 1.2 the vicinity of the metamagnetic QCP in field range and below Sr Ru O . 1.1 K, it locks in to the 3 2 7 peak. The value of r at We show the magnetic field (H ) depen- which this occurs is dence (for H // c) of the resistivity of a high- strongly temperature- 6.5 7.0 7.5 8.0 8.5 9.0 purity single crystal of Sr3Ru2O7 as temper- dependent, but both ature increased from 0.1 to 1.3 K (Fig. 1). At the lock-in field and 0.1 K, the large, steep-sided feature is similar the variation of r through the peak are strikingly independent of temperature. to that reported and discussed in (16). By 1.3 K, the observed resistivity behavior is Fig. 2. Sample ac mag- 60 unsurprising, with a pronounced thermal netic susceptibility (c) 0.300 broadening. The surprising results come at and linear magneto- 200 50 intermediate temperatures. At all temper- striction (l)dataat 0.295 100 mK. The suscepti- atures between 0.1 and 1.1 K, the anticipated 150 bility was measured at 40 broadening is seen outside the field range 0.290 a low frequency of 17 from 7.8 to 8.1 T; but once that field range is Hz to avoid excessive B [ 100 entered, r appears to lock in to the field finite-frequency effects 30 0.285 0.0 0.5 1.0 1.5 2.0 and temperature dependence seen at the (13). Both the main T (K) lowest temperature. The value of r at which field and the small ac j 20 50 this occurs is temperature-dependent, but field of 3 10 5 T there is almost no temperature dependence root mean square were applied parallel to c, 10 0 to the field at which it occurs or to its field and l 0 d(Dc/c)/dB. dependence once it has locked in to the The close similarity anomalous resistivity. between l and c¶ 0 The data of Fig. 1 suggest the presence of demonstrates a strong 6.5 7.0 7.5 8.0 8.5 9.0 a low-temperature phase in which the strong magnetostructural cou- temperature-dependent (inelastic) scattering pling, whereas the peaks in cµ accompanying the two higher-field peaks in l and c¶ identify two first-order phase transitions associated with quantum critical fluctuations [also (17)]. a.u., arbitrary units. (Inset) Sample dc magnetization (M) data. As the temperature is is cut off at well-defined boundaries and lowered at a fixed field (in this case 7.9 T), the incipient divergence of fluctuations seen in M is cut off replaced by a large and essentially temperature- at a clearly identifiable kink at the same temperature as that of the top of the peak seen in Fig. 1.

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1155 R EPORTS obtained from plotting the positions of peaks it can be defined with the clarity shown (Fig. measurements taken at temperature T 9 1K G in the thermal expansion a. Data obtained 3) only in the very best samples with rres 1 on the highest purity crystals. Second, we from five different crystals of similar purity mW cm. A strong purity dependence like this draw attention to the relative orientation of studied independently in four separate labo- is characteristic of some of the best-known the first-order phase boundaries (Fig. 3). ratories yield consistent information. Sur- ordered phases in itinerant interacting elec- These are seen to have opposite slopes and rounding the temperature and field at which tron systems such as the fractional quantum curvatures with respect to field variation, a QCP had been identified in more disor- Hall state (7) and unconventional supercon- similar to the quantum critical contours that dered samples is a region in the (H,T) plane ductivity (10). Several key pieces of exper- can be calculated by using a Hertz-Millis enclosed by well-defined phase boundaries. imental evidence further indicate that the model for the underlying QCP (19). In At sufficiently low temperatures these are previously unknown phase is intrinsically contrast, the naBve expectation for two first-order, but they then change to second- linked with the existence of the previously successive first-order metamagnetic transi- order as indicated (Fig. 3, red arrows) (18). identified metamagnetic QCP. First, as stated tions unrelated to the QCP would be lines of The existence of this new phase is the above, it encloses the characteristic field of qualitatively similar slope and curvature, central experimental result of our paper. The that QCP as estimated either by direct whereas the boundary of a new phase might anomalous transport properties that accompa- measurement on more disordered samples have been expected to be dome-shaped. ny it are strongly purity-dependent (14, 16); or by extrapolation of transport and magnetic The third piece of experimental support for a connection with the quantum critical physics of the underlying metamagnetism Fig. 3. An empirical phase dia- 1.4 comes from an initial study of the angular gram from collating data from the four measurements contrib- dependence of r. Previous work on more uting to Figs. 1 and 2, combined 1.2 disordered samples has shown that rotating with the results of thermal ex- the field from the c axis toward the ab plane pansion measurements. The col- 1.0 raises the characteristic temperature of the ored triangles are the loci of underlying critical point, Bdetuning[ the peaks in ac susceptibility; the quantum criticality (13). It is, therefore, white triangles, the loci of max- 0.8 ima (up) and minima (down); the natural to investigate the effect of changing orange circles are derived from field angle on the formation of the anoma- dc magnetization; the yellow 0.6 lous phase. To do this for all the thermody- diamonds, the loci of maxima in namic quantities contributing to Fig. 3 will be dr/dH; the green diamonds, the a formidable experimental task, but the 2 2 0.4 loci of maxima in d r/dT ; and resistivity data (Fig. 4) show that the anoma- the blue circles, the loci of lous behavior is well bounded in field angle as maxima in the magnetostriction. 0.2 The red arrows mark the temper- well as in field and temperature. Each method atures of critical points deduced of tuning is therefore seen to affect the from the temperatures above 0.0 underlying critical point and the appearance which no peak in cµ is observ- 7.6 7.7 7.8 7.9 8.0 8.1 8.2 8.3 of new phase in a qualitatively similar way. able. We believe that all the data Field (tesla) Although itinerant metamagnets have are consistent with the presence been studied fairly extensively in the past of a phase enclosed with first-order phase boundaries for temperatures below these critical points E ^ and a second-order line linking them at higher T. for example, (5, 20, 21) , novel phase for- mation in the vicinity of the metamagnetism

has been reported only in URu2Si2 (22, 23). Fig. 4. The resistivity, The phenomena seen there may have a 2.1 r, at 100 mK as the similar origin to those reported here, but magnetic field is rotat- there are important differences in the physics ed away from the c axis toward the ab of the two materials. In URu2Si2, f electrons plane at intervals of play an important role, and both supercon- - B [ 5 . The striking peak 1.6 ductivity and a much-studied hidden order that we associate with phase are observed in zero applied field. formation of the phase Also, the phase formation seems relatively for H // c (black trace) insensitive to the presence of disorder, in exists only over a nar- row range of angle. contrast to the present case. Although we For angles G 80-, fea- 1.2 certainly do not rule out relevance to the tures are still seen, but fascinating physics that has been discovered they are much weaker. in URu2Si2, the following discussion will (Insets)Sketchesofthe have its basis purely in the observations kind of spin-dependent reported here on Sr3Ru2O7. Fermi surface distortion 0.8 that we discuss in the The key facts that any model for the new text. They are not meant phase needs to address are (i) the existence to be representative of 5 6 7 8 91011 of a large enhancement in the absolute value the true Fermi surfaces of r accompanied by a suppression of its but are designed to il- temperature dependence, (ii) an itinerant lustrate the kind of symmetry breaking that we discuss. H denotes the field (about 7.8 T) at which c1 rather than a localized picture, as evidenced the symmetry-broken phase is entered. Below this field and above Hc2 (the field of about 8.1 T at which the phase is left), both the spin-up and spin-down Fermi surfaces have the same symmetry (left inset). by the observation of dHvA oscillations at both high and low fields (24)andthe Between Hc1 and Hc2, the spin-up Fermi surface distorts to a lower symmetry. This symmetry- breaking distortion could equally well be rotated by 90-, so domain formation is likely. absenceofalargechangeintheHall

1156 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS coefficient as the anomalous phase is entered understand the large resistive peak in terms of 16. R. S. Perry et al., Phys. Rev. Lett. 92, 166602 (2004). (25), (iii) a magnetization that increases both domain formation between regions in which 17. The dissipative peak in cµ is accompanied both by a peak in the imaginary part in field-modulated on entry to and exit from the new phase as a the symmetry-breaking distortion is rotated resistivity measurements and by hysteresis between function of field, (iv) a magnetization whose by 90-. Alignment of these domains by the up and down sweeps in c¶ (25). divergence is cut off as the phase is entered inplane field component would also likely 18. Our identification of the second-order phase bound- ary in Fig. 3 has its basis in the locus of the kinks in from high temperatures at constant field and be a factor in the rapid disappearance of the magnetization coinciding with that of maxima in whose absolute value shows a kink but no the peak with field angle. d2r/dT2. sharp decrease as the phase boundary is Symmetry-breaking Fermi-surface distor- 19. A. J. Millis, A. J. Schofield, G. G. Lonzarich, S. A. Grigera, Phys. Rev. Lett. 88, 217204 (2002). crossed, and (v) existence of the phase over tions have been discussed theoretically in 20. T. Goto, K. Fukamichi, H. Yamada, Physica B 300, 167 only a narrow range of applied field angle. various contexts since the advent of Landau_s (2001). These facts place some fairly tight con- Fermi liquid theory and the subsequent work 21. J. Flouquet, P. Haen, S. Raymond, D. Aoki, G. Knebel, Physica B 319, 251 (2002). staints on theory. For example, they appear of Pomeranchuk (9, 28–33). However, there 22. M. Jaime, K. H. Kim, G. Jorge, S. McCall, J. A. Mydosh, to rule out a partially gapped state such as have been, to our knowledge, no conclusive Phys. Rev. Lett. 89, 287201 (2002). that recently discussed in relation to the observations in real metals. This raises the 23. N. Harrison, M. Jaime, J. A. Mydosh, Phys. Rev. Lett. 90, 96402 (2003). hidden order of URu2Si2 (26). One scenario, question of what is so special about 24. R. A. Borzi et al., Phys. Rev. Lett. 92, 216403 (2004). however, seems to offer a plausible explana- Sr3Ru2O7. We believe that part of the answer 25. S. A. Grigera et al., unpublished data. tion for all the observations. In a standard lies in the underlying metamagnetic QCP. As 26. P. Chandra, P. Coleman, J. A. Mydosh, V. Tripathi, picture of itinerant metamagnetism, the stressed in (12–14), the fluctuations associ- Nature 417, 831 (2002). 27. B. Binz, M. Sigrist, Europhys. Lett. 65, 816 (2004). system develops the additional moment by ated with an itinerant metamagnetic QCP are 28. It is, perhaps, debatable whether the Fermi-surface a sudden extra exchange polarization of the rather unusual. They are fluctuations of the distortion discussed here should be classed as a Fermi surface, changing the relative volumes Fermi surface itself and so would act to Pomeranchuk instability. In Pomeranchuk’s original work, the presence of a lattice was not necessary for of the spin-up and spin-down parts (Fig. 4, soften the original fourfold Fermi surface, the many-body distortion to take place. Here, it is left inset). The extra exchange splitting at the making it easier for it to lower its total likely that the lattice does play a role by modulating metamagnetic transition is driven by the energy by adopting a symmetry-breaking the k dependence of the density of states and possibly also through the strong magnetostructural Zeeman splitting of the Fermi surface in an distortion. The other important feature of coupling demonstrated in Fig. 2. This does not, applied field, allowing one or both spin Sr3Ru2O7 is the level of purity to which it however, mean that many-body effects are unim- species to access a higher density of states can be grown. Just like anisotropic super- portant to the observed behavior. In fact, they are probably crucial to what takes place. First, they than is available in zero field so that the conducting gaps, anisotropic Fermi-surface renormalize the energy scale of the density of states system satisfies the Stoner criterion. Many- distortions would be averaged away by down to a few kelvin, and secondly they dominate body enhancements also play the role of sufficiently strong disorder scattering (34). the quantum critical fluctuations that we presume to be vital in softening the undistorted Fermi surfaces. reducing the energy scale of the important The fact that we observe a strong depen- In this sense, we believe that it is appropriate to features in the density of states Eto È10 K in dence of the new phase on elastic scattering describe the model that we propose with use of the ^ term ‘‘spin-dependent Pomeranchuk instability.’’ Sr3Ru2O7 (11, 12) . is consistent with this behavior being driven A natural assumption is that in k space, by a Fermi-surface distortion rather than, for 29. L. P. Gor’kov, A. Sokol, Phys. Rev. Lett. 69, 2586 (1992). 30. C. J. Halboth, W. Metzner, Phys. Rev. Lett. 85, 5162 the metamagnetic exchange splitting will example, a structural transition. (2000). 31. V. Oganesyan, S. A. Kivelson, E. Fradkin, Phys. Rev. B respect the fourfold symmetry of the RuO2 We suggest the above scenario as an planes. We postulate, however, that some- intuitively appealing possible explanation for 64, 195109 (2001). 32. I. Khavkine, C.-H. Chung, V. Oganesyan, H.-Y. Kee, thing more unusual can happen in Sr3Ru2O7. the observations that are the core of this paper (2004); available online at http://arxiv.org/abs/ Its quasi-two-dimensional electronic struc- but do not claim it to be the only way to cond-mat/?0402565. ture will incorporate fourfold density of account for the data. Real-space phase sepa- 33. H. Y. Kee, Y. B. Kim (2004); available at http:// arxiv.org/abs/cond-mat/?0408004þ (partly motivat- states maxima because of van Hove singu- ration or valence transitions are two other ed by the results and discussion presented here). larities in the electronic structure. Indeed, the possibilities worthy of investigation (35). 34. A. G. Green, S. A. Grigera, B. D. Simons, unpublished data. importance of the Fermi level lying close to More theoretical and experimental work will 35. A real-space phase separation giving a coexistence region between 7.8 and 8.1 T cannot be completely these has been discussed in the context of the be necessary to tell for certain (36). ruled out on the basis of our data. However, we standard metamagnetism (27). We believe believe that it is unlikely. Both hysteretic dissipation that the main features of our observations References and Notes (measured by cµ) and hysteresis of the dc magneti- would be explained by the onset of a spon- 1. S. L. Sondhi, S. M. Girvin, J. P. Carini, D. Shahar, zation are confined to narrow regions about the Rev. Mod. Phys. 69, 315 (1997). first-order phase boundaries identified in Fig. 3. A taneous, twofold, symmetric Fermi-surface 2. S. Sachdev, Quantum Phase Transitions (Cambridge phase coexistence might be expected to yield distortion in which the Fermi surface of one Univ. Press, Cambridge, 1999). maximum dynamic dissipation where the mixture is spin species elongates along the direction of 3. T. Senthil, A. Vishwanath, L. Balents, S. Sachdev, closest to 50/50, namely in between these two lines. M. P. A. Fisher, Science 303, 1490 (2004). 36. Some experiments that would provide strong sup- two of the four available density of states 4. For example, D. Belitz, T. R. Kirkpatrick, J. Low Temp. porting evidence of ‘‘Pomeranchuck domains’’ are peaks (Fig. 4, right inset). This Fermi Phys. 126, 1107 (2002). study of the resistivity under uniaxial inplane pres- surface, which breaks the original fourfold 5. G. R. Stewart, Rev. Mod. Phys. 73, 797 (2001). sure anisotropy and study of the inplane resistivity 6. N. D. Mathur et al., Nature 394, 39 (1998). tensor as a function of the polar angle in tilted symmetry, persists throughout the anoma- 7.Forexample,H.L.Stormer,T.R.Kirkpatrick, magnetic fields. Further insight might be gained by lous new phase until a second phase transi- Rev. Mod. Phys. 71, S298 (1999). using spatially resolved probes such as scanning tion restores the original symmetry by 8. S. Ikeda, Y. Maeno, S. Nakatsuji, M. Kosaka, Y. Uwatoko, tunnelling spectroscopy. All these experiments are planned. polarizing equally along the orthogonal Phys. Rev. B 62, R6089 (2000). 9. Y. A. Pomeranchuk, Sov. Phys. JETP 8, 361 (1959) 37. We are pleased to acknowledge useful discussions with direction. This scenario has a number of [translation from Zh. Eksp. Teor. Fiz. 35, 524 (1959)]. K. Ishida, G. G. Lonzarich, V. Oganesyan, and B. D. attractive features. It would naturally ac- 10. A. P. Mackenzie, Y. Maeno, Rev. Mod. Phys. 75, 657 Simons. This work was supported by Engineering and count for the facts that the magnetization of (2003). Physical Sciences Research Council, The Leverhulme 11. S. Ikeda, Y. Maeno, S. Nakatsuji, M. Kosaka, Y. Uwatoko, Trust, and the Royal Society (UK), and by the Japanese the system rises at each transition and that Phys. Rev. B 62, R6089 (2000). Society for the Promotion of Science and the Ministry it is not diminished on entering the new 12. R. S. Perry et al., Phys. Rev. Lett. 86, 2661 (2001). of Education, Culture, Sports, Science, and Technology phase as a function of temperature at con- 13. S. A. Grigera et al., Phys. Rev. B 67, 214427 (2003). (Japan). 14. S. A. Grigera et al., Science 294, 329 (2001). stant field. It involves no partial gapping of 15. R. S. Perry, Y. Maeno (2004); available online at the Fermi surface but still gives a way to http://arxiv.org/abs/cond-mat/?0403572. 20 August 2004; accepted 5 October 2004

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1157 R EPORTS electronic state has been central to the de- velopment of transition-state theories, and the nature of the transition state itself has The Roaming Atom: Straying been the focus of many of these studies (4, 5). Formaldehyde has literally become a text- from the Reaction Path in book example (6) with which to study these issues. It is one of the simplest molecules in which to examine the correlation between Formaldehyde Decomposition rotational and vibrational excitation in the D. Townsend,1,2 S. A. Lahankar,3 S. K. Lee,1,2,3 S. D. Chambreau,3 products. Moreover, laser excitation effi- 1,2,3 4 4 A. G. Suits, * X. Zhang, J. Rheinecker, ciently prepares H2CO on the ground elec- L. B. Harding,5 J. M. Bowman4* tronic state with specific amounts of internal energy. We present a combined experimental and theoretical investigation of form- Moore and co-workers have explored the unimolecular reaction dynamics of formal- aldehyde (H2CO) dissociation to H2 and CO at energies just above the threshold for competing H elimination. High-resolution state-resolved imaging measure- dehyde in great depth (7–12). In studies of ments of the CO velocity distributions reveal two dissociation pathways. The both H2 and CO product state distributions first proceeds through a well-established transition state to produce rotationally and state-resolved H2 Doppler profiles, they excited CO and vibrationally cold H . The second dissociation pathway yields found about 65% of the available energy is 2 released in translation; the rest of the energy rotationally cold CO in conjunction with highly vibrationally excited H2.Quasi- classical trajectory calculations performed on a global potential energy surface is partitioned between strong CO rotational excitation and modest H2 vibrational excita- for H2CO suggest that this second channel represents an intramolecular hydrogen abstraction mechanism: One hydrogen atom explores large regions tion. They found little energy in CO vibra- of the potential energy surface before bonding with the second H atom, tional excitation or in rotational excitation of bypassing the saddle point entirely. As chemical kinetic theory has evolved from profound importance for determining the rate Eyring_s work in the 1930s through the Rice- of reaction, so studies of reaction rate theory Ramsperger-Kassel-Marcus (RRKM) and have focused mainly on elucidating these related statistical approaches, the transition- properties. Absent from this paradigm is the state concept has remained paramount. Reac- possibility of two or more distinct pathways to tants are assumed to proceed along the lowest the same products. We present here a com- energy pathway to products, with the config- bined theoretical and experimental study of

uration at the energetic maximum of this formaldehyde (H2CO) dissociation that sup- pathway termed the transition state. The ports two distinct active pathways to H2 and properties of the transition state, in particular CO products. Whereas the dominant pathway its geometry and vibrational frequencies, have involves a conventional transition state, the alternative pathway is an intramolecular hydrogen abstraction that avoids the transition 1Department of Chemistry, Stony Brook University, Stony Brook, NY 11794, USA. 2Chemistry Depart- state region entirely. These results, considered ment, Brookhaven National Laboratory, Upton, NY with recent work on the CH3 þ O(1)andOþ 11973, USA. 3Department of Chemistry, Wayne State C H (2) reactions and other cases in which 4 2 6 University, Detroit, MI 48202, USA. Department of regions of deep potential energy wells are Chemistry and Cherry L. Emerson Center for Scien- tific Computation, Emory University, Atlanta, GA avoided in reaction (3), suggest that transition- 30322, USA. 5Chemistry Division, Argonne National state theories are incomplete in their descrip- Laboratory, Argonne, IL 60439, USA. tion of chemical reactivity. *To whom correspondence should be addressed. The study of unimolecular dissociation of E-mail: [email protected]; [email protected] vibrationally excited molecules in the ground

Fig. 2. (A to C) Solid lines are experimental translational energy distributions obtained from the corresponding images in Fig. 1. Dashed lines are translational energy distributions obtained from the trajectory calculations. Markers indi- 0 cate positions for correlated H2 (v and j 1) vibrational levels for v from 0 to 4 and odd

rotational levels for H2 v from 5 to 7. P(ET)is Fig. 1. DC sliced images of CO (v 0 0) after dissociation of H CO at 30,340.1 cmj1 excess energy the probability of a given translational energy 0 0 2 0 for the CO product rotational levels jCO 40 (A), jCO 28 (B), and jCO 15 (C). (ET) (arbitrary units).

1158 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS

H2. All of these observations could be channel and thus a distinct new pathway to These images represent the velocity distri- accounted for qualitatively by a skewed formation of molecular products via intra- butions for specific indicated quantum states transition-state structure in which both hy- molecular hydrogen abstraction. of the CO product, and the structure in the drogen atoms are on the same side of the CO To resolve this issue, we performed high- images reflects the internal energy distribu- molecule. These results have also been resolution direct current (DC) slice imaging tion of the H2 co-product for the particular accounted for semiquantitatively by quasi- (17) measurements of specific CO rovibra- CO level being probed because the total classical trajectory calculations which were tional levels after dissociation of formalde- energy is fixed by the photolysis laser. 0 initiated at this transition state (13–16). hyde above the threshold for the radical In the image for jCO 40 (Fig. 1A), the However, one subtle feature of the mea- channel. These images provide the correlated rings represent formation of the H2 co- surements did not fit well with this conven- H2 vibrational distributions with vibrational product in vibrational states (v) from 0 to 3 tional understanding of the dynamics. At and partial rotational resolution. The exper- with substantial translational energy release. j1 energies above 30,300 cm ,H2CO can imental measurements are combined with Integration of the data in the images and undergo an alternative dissociation to H þ quasiclassical trajectory calculations per- conversion from velocity to total energy HCO (the radical channel). In a prescient formed on a new, high-level, global potential yields the translational energy distribution paper, van Zee et al. noted (12) that, at energy surface for H2CO (18). (Fig. 2A, solid line). This resulting H2 energies below this threshold, the CO rota- The ion imaging experiment (19, 20) vibrational distribution peaks at v 0 2 with tional state ( jCO) distributions were simply involves laser excitation of formaldehyde in the bulk of the available energy appearing in 0 _ Gaussian-shaped, peaking near jCO 45, and a molecular beam under collisionless con- translation, consistent with Moore s mea- 0 were well understood to arise from energy ditions to the first electronically excited surements. The image obtained for jCO 28 and angular momentum conservation given singlet state at 30,340.1 cmj1 (21), one of (Fig. 1B) reveals a bimodal internal energy the nature of the transition state and the exit several dissociation energies used by van distribution in the H2 co-product, strongly channel interactions. However, above the Zee et al. This excitation includes one suggesting a distinct dissociation mecha- threshold for the H þ HCO dissociation quantum in the CO stretch and three in the nism. This rotational level is intermediate channel, the CO rotational distributions out-of-plane bend. Internal conversion then between the peak of the rotational distri- exhibited a shoulder toward lower rotational produces a distribution of vibrationally bution and the low-jCO shoulder reported levels. van Zee and co-workers proposed two excited ground-state molecules that dissoci- by van Zee et al. The corresponding possible explanations for the low-jCO com- ate some picoseconds later (22). The prod- translational energy distribution (Fig. 2B), ponent. One explanation was that at higher uct, CO, is probed in the dominant ground showing partial rotational resolution of the energies the transition-state region may vibrational state on specific rotational levels correlated H2, reveals that the slower CO sample additional geometries that lead to by using DC slice imaging. By using weak product is formed in conjunction with H2 in reduced exit impact parameters, hence lower DC electric focusing fields and resonant vibrational levels up to v 0 7. Lastly, the 0 rotational excitation in the CO and higher multiphoton ionization, we are able to image image for jCO 15 (Fig. 1C) in the region of overall translational and vibrational energy. the central slice through the recoiling prod- the low-jCO shoulder reported by van Zee et al. The second possibility was that the low jCO uct spheres. This yields high-resolution state- shows only the slow component, and the was related to the opening of the radical resolved velocity distributions (Fig. 1). translational energy distribution (Fig. 2C) is

0 Fig. 3. State correlations obtained from the trajectory calculations: (A) CO (all v and j) compared with H2 (v) and (B)CO(v 0 and j) compared with H2 (v).

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1159 R EPORTS quite similar to the slow part of the nearly zero relative to the first peak at low

distribution in Fig. 2B: In this case, we see vH2. The branching ratio for the molecular only formation of highly internally excited channel decreases monotonically from rough- 0 0 H2 up to v 7 and rotational state j 13. ly 1.0 to 0.2 (24). Thus, whereas the high-v The accompanying trajectory calculations H2 and low-jCO component of the molecular were performed on a new H2CO potential channel increases with energy, the branching energy surface (18) constructed from least ratio of the molecular products decreases. squares fits to roughly 60,000 high-level ab These results are consistent with the dissoci-

initio points, followed by a smooth joining ation energy dependence of the low-jCO of these fits. The surface contains the molec- shoulder reported by van Zee et al.

ular (H2 þ CO) and radical (H þ HCO) One compelling feature of these cal- dissociation channels as well as the cis and culations is that trajectories may be exam- trans isomers of HCOH. The relevant ener- ined directly for insight into the associated 0 getics of this surface, including corrections dynamics. A typical trajectory leading to jCO 0 for harmonic zero-point energy, are 28,633 41 and vH2 1 (movie S1) proceeds via the cmj1 for the barrier to the molecular channel skewed transition state, and the correspond- and 30,325 cmj1 for the radical channel, in ing reaction path is closely followed. In con- 0 0 very good agreement with experiment. trast, a trajectory leading to jCO 7andvHH Quasi-classical trajectories on this surface 6 (movie S2) involves one H atom nearly were calculated at a total energy corre- detaching via the H þ HCO channel, al- sponding to the experiment based on a though lacking sufficient energy for com- harmonic estimate of the zero-point energy plete dissociation. The H atom meanders in of 5844 cmj1. Two sets of trajectories were the broad attractive space of the H þ HCO performed: In one the initial molecular surface, far from the reaction path, until it ab- geometry was set at the formaldehyde stracts the other hydrogen atom to yield high-

minimum, and in the other one CH bond ly vibrationally excited H2 and rotationally was initially stretched to a distance of 2.1 ), cold CO. Figure 4 shows six frames of movie about twice the equilibrium value (23). S2, illustrating the Broaming[ mechanism. Roughly 40,000 trajectories were performed These two trajectories illustrate the dra- for each initial geometry with zero total matically distinct mechanisms underlying angular momentum. The generation of initial the bimodal correlated state distribution seen conditions for these trajectories followed the in the experimental images and trajectory standard method (13–16), and the usual calculations. In one case, we have what may histogram binning procedure was used to be described as the conventional dissociation obtain final vibration/rotation (v/j) distribu- pathway involving the well-characterized

tions of the H2 and CO products. The results transition state. For this dominant mecha- from the two sets of calculations are nism, transition-state theories are appropriate identical within the statistical uncertainties and may be used to predict the dissociation of each, confirming the robustness of the rates and product state distributions. For the results. The translational energy distributions second case, we find compelling evidence in obtained from the calculations (Fig. 2, support of van Zee_s conjecture invoking the dashed lines) are in near-quantitative agree- H þ HCO channel. This pathway is analo- ment with the experimental data for each CO gous to recently calculated trajectories for

rotational level. The calculations also yield the O þ CH3 reaction (1) involving near H probabilities for dissociation to given levels elimination followed by intramolecular H

of CO rotational and H2 vibrational excita- abstraction, which accounted for CO yields tion, both summed over CO vibrational seen experimentally. 0 levels (Fig. 3A) and restricted to vCO 0 We suspect this roaming atom mecha- (Fig. 3B). In both cases, there is a bimodal nism is a common pathway. A key question 0 distribution in vH2, with a major peak at vH2 remaining is whether diatomic products 0 1 and a smaller peak at vH2 6. Further in besides H2, or even polyatomic products, accord with experiment, the vibrationally may be formed by such a mechanism. It may

excited H2 correlates with a CO rotational be more common for the roaming species to 0 distribution peaked near jCO 10; the vi- be a hydrogen atom that can rapidly explore brationally cold H2 is formed with rotation- the accessible regions of the surface. Per- 0 ally excited CO (peaked near jCO 45). haps, however, an atom or group besides H Calculations have also been done at five could be the abstraction target. energies spanning a 3000-cmj1 energy range centered at the energy of the experiment. At References and Notes the lowest energy, the low-j shoulder in the 1. T. P. Marcy et al., J. Phys. Chem. A 105, 8361 (2001). CO 2. T. Yan, C. Doubleday, W. L. Hase, J. Phys. Chem. A, in CO rotational distribution is near zero rela- press (ASAP Web release 16 July 2004, DOI 10.1021/ 0 jp048150). tive to the peak value ( jCO 42), and this contribution grows monotonically as the 3. L. Sun, K. Song, W. L. Hase, Science 296, 875 (2002). Fig. 4. Six frames of a sample trajectory 4. T. Baer, W. L. Hase, Unimolecular Reaction Dynamics: 0 0 energy increases. Correspondingly, the sec- leading to CO ( jCO 7) and H2 (vH2 6) at Theory and Experiment (Oxford Univ. Press, Oxford, the indicated times after initiation. ond peak in the bimodal vH2 distribution is 1996).

1160 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS

5. C. B. Moore, Faraday Discuss. 102, 1 (1995). the axis of an imaging time-of-flight mass spectrom- range was discretized in bins of 1000 cmj1 width 0 6. R. Schinke, Photodissociation Dynamics (Cambridge eter. The beam is crossed at right angles by counter- except for the case jCO 15, where the width was 500 Univ. Press, Cambridge, 1993). propagating photolysis and probe lasers. The cmj1. Variation of these bin widths by T 20% had no 7. P. Ho, D. J. Bamford, R. J. Buss, Y. T. Lee, C. B. Moore, photolysis laser is the frequency-doubled output of significant effect on the plotted distributions. J. Chem. Phys. 76, 3630 (1982). a yttrium-aluminum-garnet–Nd (Nd:YAG) pumped 24. In quasi-classical trajectory calculations, the product 8. D. J. Bamford, S. V. Filseth, M. F. Foltz, J. W. Hepburn, dye laser tunable around 330 nm. The probe laser is molecules can be formed with less than zero-point C. B. Moore, J. Chem. Phys. 82, 3032 (1985). the frequency-tripled output of a second Nd:YAG energy, which violates a strict quantum condition. In 9. D. Debarre et al., J. Chem. Phys. 83, 4476 (1985). pumped dye laser, tunable around 230 nm. The the present calculations, we see this effect by 10. T. J. Butenhoff, K. L. Carleton, M. C. Chuang, C. B. Moore, probe uses the well-known 2-plus-1 resonance- observing a small amount of H þ HCO products J. Chem. Soc. Faraday Trans. 85, 1155 (1989). enhanced ionization of CO via the (B-X) Q-branch below the strict threshold energy of 30,300 cmj1.

11. T. J. Butenhoff, K. L. Carleton, C. B. Moore, J. Chem. transition, affording, in this case, detection of single Thus, the branching ratio to form H2 þ CO is not Phys. 92, 377 (1990). CO rotational levels for j 0 10 and higher. Data exact unity as it should be; however, it is near unity. 12. R. D. van Zee, M. F. Foltz, C. B. Moore, J. Chem. Phys. images are recorded with centroiding without One difficulty in establishing a more reliable value at 99, 1664 (1993). filtering or smoothing and then fourfold symme- these lower energies is that a significant number of 13. Y.-T. Chang, C. Minichino, W. H. Miller, J. Chem. trized to improve signal to noise. trajectories became very long-lived and had to be Phys. 96, 4341 (1992). 22. C. B. Moore, J. C. Weisshaar, Annu. Rev. Phys. Chem. terminated before dissociation occurred. 14. X. Li, J. M. Millam, H. B. Schlegel, J. Chem. Phys. 113 34, 525 (1983). 25. This work was supported by the director, Office of (2000). 23. A normal-mode analysis was done at these two Science, Office of Basic Energy Sciences, Division of 15. W. Chen, W. L. Hase, H. B. Schlegel, Chem. Phys. Lett. initial configurations, and normal modes associated Chemical Sciences, Geosciences, and Biosciences, 228, 436 (1994). with CH motion were clearly identified. The sum of U.S. Department of Energy, under contracts DE- 16. J. L. Rheinecker, X. Zhang, J. M. Bowman, in preparation. harmonic zero point energies for all other modes was AC02-9810886 and DE-FG02-04ER15593 (A.G.S.), 17. D. Townsend, M. P. Minitti, A. G. Suits, Rev. Sci. subtracted at the outset. The excess energy was then DE-FG02-97ER14782 (J.M.B.), and W-31-109-ENG- Instruments 74, 2530 (2003). assigned as kinetic energy to the CH normal modes. 38 (L.B.H.). 18. X. Zhang, S. Zou, L. B. Harding, J. M. Bowman, J. Phys. For an initial configuration at the H CO minimum, 2 Supporting Online Material Chem. A 108, 8980 (2004). this excess energy was equally partitioned between www.sciencemag.org/cgi/content/full/1104386/DC1 19. D. W. Chandler, P. L. Houston, J. Chem. Phys. 87, the symmetric and antisymmetric CH stretches. For Movies S1 and S2 1445 (1987). a stretched H2CO initial configuration, the excess 20. A. Eppink, D. H. Parker, Rev. Sci. Instrum. 68, 3477 energy was given to the one real-frequency CH- 23 August 2004; accepted 7 October 2004 (1997). stretch normal mode. To obtain good statistics for the Published online 21 October 2004; 21. The experiment features a pulsed molecular beam of quasiclassical distributions, we collected those results 10.1126/science.1104386 T formaldehyde directed into a vacuum chamber along over a jCO range of 2. Also, the translational energy Include this information when citing this paper.

can be molded into such convoluted shapes, when inorganically formed calcite crystals Sea Urchin Spine Calcite are rhombohedra with flat crystal faces. An- other perplexing issue is how the cells can Forms via a Transient Amorphous efficiently provide the ions and, at the same time, efficiently remove water, during and af- Calcium Carbonate Phase ter crystal deposition. This question becomes critical if it is assumed that the crystal grows Yael Politi,1 Talmon Arad,2 Eugenia Klein,2 out of an aqueous solution saturated with Steve Weiner,1 Lia Addadi1* calcium carbonate (11). Insights into these questions have been The skeletons of adult echinoderms comprise large single crystals of calcite gained from studies of calcitic spicule for- with smooth convoluted fenestrated morphologies, raising many questions mation in sea urchin larvae (12). The larval about how they form. By using water etching, infrared spectroscopy, electron spicules grow on a single calcite crystal seed diffraction, and environmental scanning electron microscopy, we show that by transformation of a transient amorphous sea urchin spine regeneration proceeds via the initial deposition of amorphous calcium carbonate (ACC) phase (13). ACC calcium carbonate. Because most echinoderms produce the same type of is apparently fed into the syncytium by cells skeletal material, they probably all use this same mechanism. Deposition of in the form of ACC-containing vesicles (14). transient amorphous phases as a strategy for producing single crystals with Thus, packages of ACC are delivered to the complex morphology may have interesting implications for the development crystal deposition site and then transform in of sophisticated materials. a controlled manner into calcite single crys- tals. There is also no discernible aqueous Many organisms—including mollusks, echi- form a three-dimensional fenestrated mineral phase around the growing spicule (14). Mol- noderms, calcisponges, corals, certain algae, network surrounding micrometer-sized lusk larvae also form their shells via an ACC and others—form their hard parts out of spaces that are occupied by living cellular precursor phase, which then transforms into calcium carbonate minerals (1, 2). The echi- tissue. This so-called stereom appears to be aragonite (15). Because this strategy may be a noderms, which include sea urchins, sea stars, an Ball purpose material,[ used by members unique feature of larvae adapted to a swim- and brittle stars, among others, are among the of the whole phylum (5). It is molded into ming mode of life in the oceans, it was im- few groups that form skeletal parts consisting different shapes and sizes and is thus adapted portant to determine whether or not adult of large single crystals of Mg-bearing calcite, to fulfill different functions. echinoderms or mollusks also use transient some of which are several centimeters long The strategy evolved by echinoderms to ACC for forming their hard parts. We show that (3, 4). These single crystals usually have build their spines, test plates, and ossicles adult sea urchins do build their regenerating smooth, continuously curved surfaces that has been the subject of study for more than a spines via the deposition of an amorphous ACC hundred years (6, 7). The crystals form in- precursor phase.

1 2 side a syncytium (8), a membrane envelope Sea urchins are able to regenerate spines Department of Structural Biology, Chemical Re- search Support Unit, Weizmann Institute of Science, produced by many cells. The cells provide that break. Spine regeneration first involves a 76100 Rehovot, Israel. the raw materials necessary for constructing wound-healing phase during which the epi- *To whom correspondence should be addressed. the growing single crystals (9, 10). A key dermis is reconstituted around the broken E-mail: [email protected] question is how a single crystal of calcite spine (16, 17). Within this space, the skeleton-

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1161 R EPORTS forming sklerocytes build a new syncytium, tracted spectrum, relative to calcite. This may produced a diffraction pattern typical of cal- which is in contact with the stump of the old be due to water molecules that stabilize the cite (Fig. 3B). spine. The new crystal is presumably nucle- amorphous phase and are intimately associ- Etching, IR spectroscopy, and transmis- ated epitaxially on the old spine (18), ated with it. The presence of some water also sion electron microscopy thus provide inde- because the regenerated and old spines to- in the mature crystalline part of the spine pendent evidence that the regenerated gether diffract x-rays as one single crystal. indicates, however, that this may be intersti- microspines are covered by a thin layer of The regeneration process is thus considered tial water or water associated with residual ACC that, with time, transforms into crystal- to be very similar to the original spine growth. organic matrix. line calcite. The crystallization process occurs Regeneration starts in the form of small Particles removed from regenerating spines both in situ and in particles removed from the projections emerging from the broken stere- by the same procedure as described above spine. om surface. These microspines initially were dispersed on an electron microscope We were also able to directly observe grow parallel to the mature spine long axis, grid and examined by electron imaging and the crystallization of the amorphous sheath which is also the c crystallographic axis of diffraction (Fig. 3) (22). The particles are and tip of the microspines into calcite in a calcite, and then form the complex three- È50 nm in size. Particles obtained from wet atmosphere, using ESEM (Fig. 4) (22). dimensional fenestrated structure (19)(Fig.1, fresh regenerated spines did not produce an The electron beam provides the energy re- A and B). As the process is initiated with the electron diffraction pattern (Fig. 3A). The quired to induce crystallization. The tip of break, a real-time investigation of rapid skel- same particles, examined after 3 weeks on the microspine first undergoes a reduction eton deposition is possible. We used etching the grid and kept in air, were crystalline and in volume (Fig. 4A; compare with Fig. 4, B in water, infrared spectroscopy, electron dif- fraction, and environmental scanning elec- tron microscopy (ESEM) to investigate the possible involvement of ACC in spine re- generation of the sea urchin Paracentrotus lividus. The solubility of ACC (200 mg/liter) is 30 times greater than the solubility of calcite (6.7 mg/liter) (20, 21). Etching in water thus reveals the presence of ACC, which is se- lectively dissolved out of a mixture of ACC and calcite. Four-day-old regenerated micro- spines were etched in water immediately after severing from the spine (22). Extensive etch- ing occurred in a thin (100 to 200 nm thick) surface layer (Fig. 1D). These etched surfaces were compared to regenerated microspines that were left at room temperature in air for 1 month and then etched. The latter were not affected by the treatment (Fig. 1E) and closely resembled the untreated spines (Fig. 1C). Fourier transform infrared (FTIR) was Fig. 1. Scanning electron micrographs of regenerating spines. (A) Five-day-old regenerated spine used to further investigate the nature of the growing on the original broken spine. (B) Higher magnification view of the tip of the new growth, thin surface layer (Fig. 2, A and C). The showing the typical stereom structure and the protruding newly formed microspines. (C) One characteristic broadening of the in-plane car- microspine formed after 4 days of regeneration, observed fresh. (D) Four-day-old microspine, j1 etched in water while fresh. (E) Four-day-old microspine, etched in water 1 month after re- bonate bending peak at 713 cm , relative to generation. All the spines were observed after removal of the exposed organic material by a 3% j the out-of-plane bending peak at 876 cm 1, NaOCl solution. Etching was performed by immersing cleaned spines in double distilled water for is well documented for ACC-containing 12 hours. sea urchin larval spicules (13). Regenerated spines, cleaned of organic matrix, were dried in ethanol and sonicated into acetone Fig. 2. FTIR spectra of particles removed from fresh regenerated after rapid freeze shock treatment in liquid spines by freeze shock and soni- nitrogen. Sonication detached small parti- cation. The intensities are normal- cles from the surface of the microspines. ized to the carbonate stretching The infrared spectrum of these particles peak at 1420 cmj1.(A) IR spec- shows a sharp peak at 876 cmj1, whereas trum of the freshly removed par- the 713 cmj1 peak comprises a sharp peak ticles. The spectrum corresponds to that of a mixture of ACC and superimposed on a very broad peak (Fig. 2A, 0 calcite (ratio of intensities I876/I713 inset). The spectrum of crystalline calcite 4.25). (B) IR spectrum of material taken from the mature part of the same removed from the mature part of spines (Fig. 2B) was then subtracted from the the spines. The spectrum corre- latter. The residual spectrum still shows a sponds to that of crystalline cal- j cite (ratio of intensities I /I 0 sharp out-of-plane bending peak at 876 cm 1, 876 713 but the 713 cmj1 peak consists only of a 3). (C) IR spectrum of the par- ticles in (A), after subtraction of broad hump (Fig. 2C), clearly indicative the spectrum of crystalline calcite of the presence of ACC. A broad peak at from the old part of the spine È3500 cmj1 is also enhanced in the sub- [taken from (B)]. The spectrum corresponds to that of ACC.

1162 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS and C). A single crystal is then observed to ACC in sea urchin larval spicules contains than if the crystal were deposited directly form as thin plates with well-defined edges little or no water (24). If the same mecha- from solution. The subsequent transfor- (Fig. 4, D and E). The delimiting faces sub- nism is also valid for the growth of adult mation of the amorphous phase into a com- tend, between them and with the axis of the spines, this would imply that calcium car- posite crystalline solid with much better microspine, dihedral angles matching those bonate is first deposited as hydrated ACC, mechanical properties (25) leads to a func- expected for the (001), (102), and (010) faces and then dehydrates prior to or concomitant tional skeleton. A similar strategy could be of a calcite crystal oriented with the c axis with crystallization. The shrinkage might, used to form synthetic materials with mini- parallel to the long axis of the spine. The however, also be associated with reorga- mal porosity and maximal shape flexibility crystal then grows at the expense of the nization of the material into a crystalline and has been explored in vitro (26–29). amorphous material that is consumed in the lattice with or without concomitant expul- Our investigation of the regeneration of electron-irradiated region (Fig. 4, F and G), sion of interstitial water. Irrespective of the the adult sea urchin spine shows that the whereas the nonirradiated region remains un- mechanism of ACC-to-calcite transition, skeletal hard part forms through an amor- modified (Fig. 4, C and H). When the same the introduction and deposition of calcium phous precursor phase. Most members of procedure is performed on old regenerated carbonate as ACC may well provide some this phylum form the same type of skele- spines, or on an old part of the same spine, answers to the perplexing questions raised tal material (2), the stereom, with similar no such changes are observed, and the spine above; namely, how the spine is shaped, properties: large single crystals of magnesium- preserves its dimensions and appearance. how the ions are efficiently introduced, and containing calcite, which break with con- The reduction in spine volume upon how water is expelled. ACC is introduced choidal fracture and are reinforced by in- crystallization may indicate expulsion of into the syncytium as an isotropic noncrys- tracrystalline organic matrix. We therefore water, which would be in agreement with talline solid and can thus be molded into suggest that all echinoderms probably use the IR data (Fig. 2). This is interesting any shape. The solid itself is, of course, a this mineral-formation process. It is conceiv- given that stable ACC contains 15 weight % very concentrated source of ions. Water has able that many other animal phyla also use water, corresponding to the molar composi- to be expelled in only small (at most 15%) the same strategy. Isolated examples have I tion CaCO3 H2O(23). In contrast, transient amounts. This is orders of magnitude lower been demonstrated involving, other than ACC, amorphous ferric oxide and amorphous Fig. 3. Transmission calcium phosphate Etable 3.2 in (2)^. With electron micrographs the discovery that this process is used by (TEM) and electron dif- adult echinoderms, the possibility that it is a fraction patterns of par- ticles removed from widespread strategy seems likely. Further- fresh regenerated spines. more, this strategy of molding macroscopic (A) Electron diffraction elements into any desired morphology and pattern of the particles endowing them with the strength and the de- observed immediately gree of perfection of a single crystal, but after removal from the with the reduced brittleness and mechanical spine; the diffuse rings indicate the presence properties of a glass, may have interesting of an amorphous ma- implications for the development of sophis- terial. (Inset) TEM of ticated materials. the particles. Bar, 50 nm. (B) Electron diffraction pattern of the same particles as in (A) after 3 weeks on the grid. The most prevalent diffraction peaks correspond to a d-spacing of 3.04 A˚, characteristic of the calcite plane References and Notes 1. K. Simkiss, K. Wilbur, Biomineralization. Cell Biology {104}. They are detected up to fourth order (arrows, enhanced contrast in window). (Inset) TEM of and Mineral Deposition (Academic Press, San Diego, the particles. Bar, 50 nm. Fresh regenerated spines were subjected to freeze shock and sonication. 1989). The separated material was mounted and examined on a marked TEM grid, and the position of the 2. H. A. Lowenstam, S. Weiner, On Biomineralization particles was recorded. The same particles were examined again after 3 weeks. (Oxford Univ. Press, New York, 1989).

Fig. 4. Environmental scanning electron micro- graphs (ESEM) of a fresh, newly formed microspine taken in succession over a period of È15 min. (A) Microspine at the begin- ning of the observation. (B to H) Pictures taken in succession every 2 to 5 min at the same magni- fication. Between (B) and (C) and between (G) and (H), the viewing frame was moved along the axis of the microspine to compare the condition of a nonirradiated region of the microspine relative to the irradiated region. Lines in (C) and (H) show the boundaries of the previous frames. Arrows in (D) show the growing crystal faces. Regenerated fresh spines were cleaned from epidermal tissue by 0 NaOCl treatment, washed with water, and immediately transferred wet to an ESEM stub, where they were observed under a humid atmosphere (pH2O 6.6 torr, 5-C, 10 kV).

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1163 R EPORTS

3. K. Towe, Science 157, 1048 (1967). 16. K. Ma¨rkel, U. Ro¨ser, Zoomorphology 103, 43 (1983). 28. L. B. Gower, D. J. Odom, J. Cryst. Growth 4, 719 (2000). 4. D. M. Raup, J. Geol. 67, 661 (1959). 17. B. M. Heatfield, J. Exp. Zool. 178, 233 (1971). 29. E. Loste, F. C. Meldrum, Chem. Commun., 901 (2001). 5. S. Weiner, L. Addadi, H. Wagner, Mater. Sci. Eng. 11, 18. P. Dubois, C. P. Chen, Echinoderm Stud. 3, 109 (1989). 30. We thank M. Cohen for help with the ESEM mea- 1 (2000). 19.P.Dubois,L.Ameye,Microsc. Res. Tech. 55, 427 surements. We also thank B. Scharfstein, I. Ivry, and 6. S. Loven, Sven. Vetensk. Akad. Handl. 18, 1 (1892). (2001). Y. Raz (SeaOr Marine Enterprises Ltd.) for providing 7. W. B. Carpenter, Brit. Assoc. Adv. Sci. London Rep. 20. L. Brecevic, N. A. J. Cryst. Growth 98, 504 (1989). the animals and facilities for spine regeneration. L.A. 17, 93 (1847). 21. F. Lippmann, Sedimentary Carbonate Minerals (Springer, is the incumbent of the Dorothy and Patrick Gorman 8. K. Okazaki, Embryologia (Nagoya) 5, 283 (1960). Berlin, 1973). Professorial Chair of Biological Ultrastructure, and 9. J. B. Pilkington, J. Mar. Biol. Assoc. UK 49, 857 (1969). 22. Materials and methods are available as supporting S.W. is the incumbent of the Dr. Trude Burchardt 10. K. Ma¨rkel, U. Ro¨ser, Zoomorphology 103, 25 (1983). material on Science Online. Professorial Chair of Structural Biology. Supported by 11. K. Ma¨rkel, U. Ro¨ser, M. Stauber, Zoomorphology 109, 23. L. Addadi, S. Raz, S. Weiner, Adv. Mat. 15, 959 (2003). a grant from the Philip Klutznick Research Fund. 79 (1989). 24. S. Raz, P. C. Hamilton, F. H. Wilt, S. Weiner, L. Addadi, 12. F. H. Wilt, Zool. Sci. 19, 253 (2002). Adv. Funct. Mater.13, 480 (2003). Supporting Online Material 13. E. Beniash, J. Aizenberg, L. Addadi, S. Weiner, Proc. R. 25.A.Berman,L.Addadi,S.Weiner,Nature 331, 546 www.sciencemag.org/cgi/content/full/306/5699/1161/ Soc. London B Biol. Sci. 264, 461 (1997). (1988). DC1 14. E. Beniash, L. Addadi, S. Weiner, J. Struct. Biol. 125, 26. G. F. Xu, N. Yao, I. A. Aksay, J. T. Groves, J. Am. Materials and Methods 50 (1999). Chem. Soc. 120, 11977 (1998). 15. I. M. Weiss, N. Tuross, L. Addadi, S. Weiner, J. Exp. 27. J. Aizenberg, J. L. Grazul, D. A. Muller, D. R. Hamann, Zool. 293, 478 (2002). Science 299, 1205 (2003). 2 July 2004; accepted 6 October 2004

0.4, or 0.6). For each event, we calculated Earth Tides Can Trigger Shallow the tidal phase angle (q) between –180- and 180- (11); 0- phase is defined to be at the Thrust Fault Earthquakes time of maximum stress that can promote failure, which is extensional for normal Elizabeth S. Cochran,1* John E. Vidale,1 Sachiko Tanaka2. stress and in the direction of slip for shear stress (Fig. 1). In addition, we defined the We show a correlation between the occurrence of shallow thrust earthquakes average of the tidal stress amplitudes at the and the occurrence of the strongest tides. The rate of earthquakes varies from peaks just before and after each earthquake the background rate by a factor of 3 with the tidal stress. The highest (tpb and tpa, respectively) to be the peak tidal correlation is found when we assume a coefficient of friction of m 0 0.4 for the stress tp. crust, although we see good correlation for m between 0.2 and 0.6. Our results We focused on a subset of shallow thrust quantify the effect of applied stress on earthquake triggering, a key factor in earthquakes with depths of 0 to 40 km because understanding earthquake nucleation and cascades whereby one earthquake these earthquakes are in regions with the triggers others.

For more than a century, researchers have and an ocean-loading component (7, 10, 11). sought to detect the effect on the timing of Solid-Earth tides induce stresses only up to earthquakes of the gravitational perturbations 5 103 Pa (0.05 bar), whereas in ocean ba- on Earth from the Moon and Sun (1). However, sins, water loading builds stresses up to nearly the tidal stresses in most locations are small, 5 104 Pa (0.5 bar). Both components must and usually it is difficult to ascertain the be accurately determined to fully resolve tidal orientation of the fault plane, which is critical influences on the initiation of earthquakes when calculating the effect of the stress globally. We resolved tidal stresses into variations. Earthquake-tide correlations have normal and shear stress acting on each of been observed to be small or nonexistent in the two possible fault planes of the CMT normal crust (2–4); however, correlations earthquake focal mechanism. Shear failure have been shown in shallow, possibly hy- under compressive stress can be described by drothermal or magma-related areas (5, 6). the Coulomb criterion, in which a fault fails Fig. 1. Schematic diagram of the tidal stress Here, we take advantage of accurate account- under a combination of shear and normal time series spanning 1 day for a hypothetical 0 earthquake (asterisk). Tidal phase is marked, ing of ocean tides (7) and a large data set stress: tc t þ msn,wheret and sn are the of earthquake focal mechanisms with fairly shear and normal stresses, respectively, and m with the maximum Coulomb stress promoting failure defined at 0- phase. The earthquake well known fault planes (8) to look for a is the coefficient of friction. In addition to occurs at q 0 45-. Peak stress amplitudes correlation. examining shear and normal stress indepen- before and after an event (tpb and tpa, We used global earthquakes in the Harvard dently, we tested different values of m (0.2, respectively) are averaged to determine tp. Centroid Moment Tensor (CMT) catalog (9). For each event, we calculated a tidal-stress Table 1. Comparison of coefficients of friction. Data are shown for the 250 events with the highest time series that includes the solid Earth tide calculated tidal stress (tp) given different values of the coefficient of friction (m). Binomial is approximated by a Gaussian distribution; P values are determined using Schuster’s statistical test of 1 Department of Earth and Space Sciences and data distribution (values below 5% are often considered significantly nonrandom). qpeak is the phase of Institute of Geophysics and Planetary Physics, Uni- the peak of a sinusoidal fit to the data. See text for definition of Nex. versity of California, Los Angeles, CA 90095, USA. 2 Department of Geophysics, Graduate School of Events Binomial (%) P value (%) N (%) q (degrees) Science, Tohoku University, Sendai, Miyagi 980- ex peak 8578, Japan. m 0 0 (shear) 4.38 10.36 5.6 –22.2 *To whom correspondence should be addressed. m 0 0.2 0.1439 0.6253 9.6 –1.2 E-mail: [email protected] m 0 0.4 0.0032 0.0157 12.8 0.2 .Present address: National Research Institute for Earth m 0 0.6 0.0942 0.3265 10.0 6.0 Science and Disaster Prevention (NIED), Tsukuba-shi, m 0 V (normal) 0.4688 3.677 8.4 15.8 Ibaraki-ken 305, Japan.

1164 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS largest tidal stresses; the true fault plane is 0.2, 0.4, or 0.6 (Fig. 3). Schuster_s test, used to probability of 999.99% that earthquakes better known for these events; and shallow find statistical significance of periodicity, correlate with tides when tidal stress ampli- events are under lower confining pressures, so indicates less than a 1% probability that the tudes are high (90.01 MPa) (Fig. 3D). The tidal stress may be proportionally more in- distribution is random for tp above about 0.01 binomial statistic is calculated assuming that fluential. In addition, thrust and normal faults MPa (0.1 bar). More events occur during the just one choice of parameters was tested; the have relatively larger tidal stress amplitudes period of encouraging stress (–90-Gq G 90-) true odds of finding the correlation at than strike-slip faults because of a larger than occur during times of discouraging random are expected to be somewhat lower influence of the ocean loading component. As stress (–180-Gq G –90- or 90-Gq G than those obtained here. Tidal correlation the ocean moves back and forth in an ocean 180-); the sinusoidal fit to the nonrandom with earthquake timing at the highest stress basin, the additional weight of the water acts distribution of events peaks at a tidal phase levels is apparent for most choices of to clamp and unclamp dipping faults. In sub- near q 0 0-, which is the phase of tidal stress coefficient of friction, but the most signifi- duction zones, ocean loading tends to be expected to promote failure (Fig. 4A). cant correlation is for m 0 0.4 (Table 1 and largest and low-angle thrust events are most We also performed a binomial test of Fig. 3D). Assuming m 0 0.4, a minimum common. The focal mechanisms for shallow significance because Schuster_s tests may binomial probability of 0.0027% of fortu- thrust faults show a bimodal distribution of overestimate significance. We estimated the itous correlation (with a corresponding P fault dips for fault planes 1 and 2 (Fig. 2A). probability of seeing at random the larger value of 7.6 10–5 is observed for the 255 9 The shallower-dipping fault planes (plane 1) fraction of earthquakes we observed to occur events that occur at tp 0.003 MPa (11). for the 19 events with tp greater than 0.02 during the half of the tidal phase period with We see fluctuations in the binomial MPa (0.2 bar) are aligned with the local encouraging stress, –90-Gq G 90-.Ifwe statistics (Fig. 3D). The most highly corre- geometry of the subduction zone (Fig. 2B), assume that events occur at random, the lated 45 events are the most highly stressed, so for this study we assume that to be the number of events that occur with a tidal phase and most are shallow, with 41 events at or true fault plane. between –90- and 90- should be equal to the above 15 km depth. The next 100 events are Correlation with the tides is found for number that occur in the other half of the tidal not as highly correlated and tend to be shallow-dipping thrust events, assuming m 0 phase range. This simple binomial test gives a deeper, with only 45% of the earthquakes at

A 0.04

0.02

Peak Tidal Stress (MPa) Peak 0.0

60 B

40

20 Excess Events (#) Excess Events

0 30 C

20

10 Excess Events (%) Excess Events

0 10 2 D 10 1

10 0

10 -1

10 -2 Binomial Probability (%)

10 -3 0 300 600 900 1200 1500 1800 Number of Earthquakes

Fig. 2. (A) Plot of frequency versus fault dip for 2027 reverse-type earthquakes with hypocenters Fig. 3. (A) Plot of peak tidal stress tp versus shallower than 40 km depth. Fault plane 1 (black) and fault plane 2 (gray) dips are from the total number of events Ntot.(B) Plot of the Harvard CMT catalog focal mechanism solutions. (B) Global distribution of maximum t assuming number of excess events N during times of 0 - - p ex m 0.4 in 5 5 grids for all M Q 5.5 global shallow thrust earthquakes. Continents are shown in higher stress versus Ntot.(C) Plot of percentage light gray; plate boundaries are shown by solid black lines. Note higher peak tidal stresses (red and of N versus N .(D) Binomial probability ex tot0 orange grids) at continent-ocean margins. Lower-hemisphere focal mechanisms (circular, half- versus Ntot for m 0.2 (black line), 0.4 (gray filled symbols) shown for the 19 thrust events with tp Q 0.02 MPa agree with tectonic regime for line), and 0.6 (gray line). Dashed line indicates the shallower-dipping choice of the fault plane. 5% probability level for reference.

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1165 R EPORTS

20 correlation in subregions, with the best Fig. 4. (A) Histogram of A frequency of events ver- correlations found in regions that experi- sus q for the 255 events enced a large earthquake (15). Our study with the highest tp, as- shows a statistically significant increase in suming m 0 0.4. Solid line 10 triggering of globally distributed large tec- is the sinusoidal least- tonic events (M 9 5.5) with increasing tidal squares fit to the data; dashed line is the Cou- (%) Frequency stress amplitudes (Fig. 4). lomb stress amplitude as Tidal stress amplitudes required to trigger a function of tidal phase. 0 earthquakes are similar to thresholds sug- -180 -90 0 90 180 (B) Percentage of Nex ver- gested for static and dynamic triggering of Tidal Phase (degrees) sus tp. Values are given in aftershocks of 0.003 MPa (16–18). A study Table 2. Points are lo- of triggering after the M7.4 Landers event, cated at the mean t ; 50 p B for example, showed that aftershocks were range is indicated by sol- triggered by stress increases greater than id horizontal lines. Gray 40 circles are global thrust 0.01 MPa (16). In addition, the observed data; solid triangle at trend of increased triggering with higher lower left denotes Califor- 30 imposed tidal stress can be well fit to friction nia strike-slip results. Es- theories of rate- and state-dependent friction timated SEs are shown by and stress corrosion (11, 19, 20). vertical bars. Crosses and 20 diamonds show the least- Excess Events (%) Excess Events squares fit of the data to References and Notes rate- and state-dependent 10 1. D. Emter, in Tidal Phenomena,H.Wilhelm,W.Zurn, friction and stress corro- H.-G. Wenzel, Eds., vol. 66 of Lecture Notes in Earth sion, respectively. Sciences (Springer-Verlag, Berlin, 1997), pp. 293 –310. 0 2. S. Hartzell, T. Heaton, Bull. Seismol. Soc. Am. 79, 0.0 0.01 0.02 0.03 0.04 0.05 1282 (1989). τp (MPa) 3. J. E. Vidale, D. C. Agnew, M. J. S. Johnston, D. H. Oppenheimer, J. Geophys. Res. 103, 24567 (1998). 4. J. E. Vidale, D. Agnew, D. Oppenheimer, C. Rodriquez, H. Houston, Eos 79 (suppl.), F641 (1998). Table 2. Data summary. Both global thrust events and California strike-slip events (4) are shown for 5. M. Tolstoy, F. L. Vernon, J. A. Orcutt, F. K. Wyatt, 0 various peak Coulomb stress ranges (tp) given a coefficient of friction m 0.4. N is the number of events Geology 30, 503 (2002). 6. M. O. Saar, M. Manga, Earth Planet. Sci. Lett. 214, in each tp bin. Nex includes SE. 605 (2003). 7. S. Tanaka, M. Ohtake, H. Sato, J. Geophys. Res. 107, Data set tp (MPa) NNex (%) 2211 (2002). 8. A. M. Dziewonski, T.-A. Chou, J. H. Woodhouse, J. Global thrust 90.02 19 23.7 T 10.10 Geophys. Res. 86, 2825 (1981). Global thrust 0.01 to 0.02 41 13.4 T 7.52 9. The entire data set consists of the 9350 global Global thrust 0.004 to 0.01 155 8.7 T 3.95 earthquakes of M5.5 or greater from 1977 to 2000 in Global thrust G0.004 1,813 1.2 T 1.17 the Harvard CMT catalog. These events include 2823 California strike-slip G0.004 27,464 0.60 T 0.30 reverse, 1040 normal, 3597 strike-slip, and 1890 oblique-type faulting events (7). 10. K. Matsumoto, T. Takanezawa, M. Ooe, J. Oceanogr. 56, 567 (2000). -G G - 11. See supporting data on Science Online. or above 15 km depth. We observe a recov- (–90 q 90 ). Successively lower peak 12. W. S. D. Wilcock, Geophys. Res. Lett. 28, 3999 ery of the tidal correlation as more events tidal stresses show lower tidal correlations, (2001). are included. This trend suggests a depth indicating less effective earthquake trigger- 13. S. Tanaka, M. Ohtake, H. Sato, Geophys. Res. Lett. 29, 10.1029/2002GL015386 (2002). dependence to the tidal correlation, but this ing (Table 2 and Fig. 4B). Standard errors, 14.S.I.S.Custodio,J.F.B.D.Fonseca,N.F.d’Oreye, trend is not statistically significant. Event calculated using data variance, are large be- B. V. E. Faria, Z. Bandomo, Geophys. Res. Lett. 30, depth may influence correlation with the cause of small numbers of earthquakes in 1816 (2003). 15. S. Tanaka, M. Ohtake, H. Sato, Earth Planets Space tides because confining pressures increase each tp bin. In addition to the global data 56, 511 (2004). with depth, so there is a comparatively larger set, 27,464 strike-slip events from Califor- 16. R. S. Stein, Nature 402, 605 (1999). tidal influence at shallow depths. However, nia were found to have roughly 1 to 2% more 17. J. L. Hardebeck, J. J. Nazareth, E. Hauksson, J. Geophys. Res. 103, 24427 (1998). tidal stress amplitudes decay with depth, so events during times of encouraging tidal stress 18. R. S. Stein, Sci. Am. 288, 72 (January 2003). event depth and tidal stress amplitude are not of G0.004 MPa (Fig. 4B). 19. J. H. Dieterich, Tectonophysics 144, 127 (1987). independent. A previous study using global data hinted 20. I. Main, Geophys. J. Int. 139, F1 (1999). On the basis of the earthquake-tide corre- at an earthquake-tide correlation, suggesting 21. We thank P. Bird, E. Brodsky, J. Dieterich, H. Houston, D. Jackson, C. Scholz, R. Stein, and three anonymous lation observed, we have tried to estimate the that reverse and normal earthquakes corre- reviewers for helpful comments and suggestions and tidal stress amplitude required to trigger an late either with the shear stress or the trace of K. Matsumoto for making available the ocean tide earthquake. We define the percent of excess the stress tensor J1 (7). In addition, a few model NAO.99b. Supported by NSF grant 0125732, an NSF graduate fellowship (E.S.C.), and grant-in-aid events, Nex, to be the number of additional or regional studies (5, 12–14) in areas with a for JSPS fellows 07136 (S.T.). excess events in the tidal phase period with large ocean-loading component of tidal encouraging stress, N 0EN –(N /2)^/ stress have observed triggering of earth- Supporting Online Material ex enc tot www.sciencemag.org/cgi/content/full/1103961/DC1 Ntot, where Nenc is the number of events with quakes or volcanic tremor. Most often, -G G - Materials and Methods –90 q 90 and Ntot is the total number of triggered events have been shallow, normal- References events. For tp greater than 0.02 MPa (0.2 bar), type faulting along axial ridges, such as Juan N is 24%. This corresponds to 74% of the de Fuca (5, 12), perhaps associated with 11 August 2004; accepted 7 October 2004 ex Published online 21 October 2004; events (14 of 19 earthquakes) occurring in hydrothermal circulation. A recent study of 10.1126/science.1103961 50% of the phase time of encouraging stress seismicity in Japan showed earthquake-tide Include this information when citing this paper.

1166 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS less time available for 222Rn to decay, the degassed volume needs to be larger. Geochemical Precursors to Additional constraints on the time scales of degassing can be obtained from pla- Volcanic Activity at Mount gioclase phenocrysts, which record changes in the magmatic environment by changing St. Helens, USA composition. Although the plagioclase pheno- crysts in Mount St. Helens dacite are var- Kim Berlo,1* Jon Blundy,1 Simon Turner,1,2 Kathy Cashman,3 iable in texture and composition within any Chris Hawkesworth,1 Stuart Black4 one sample, all samples display the same textures and compositional range. This is The importance of the interplay between degassing and crystallization before reflected in the similar Ba, Sr, Ti, and Mg and after the eruption of Mount St. Helens (Washington, USA) in 1980 is well contents of plagioclase phenocrysts from established. Here, we show that degassing occurred over a period of decades different eruptions. The striking exception to days before eruptions and that the manner of degassing, as deduced from is Li (Fig. 2). The Li concentration of geochemical signatures within the magma, was characteristic of the eruptive plagioclase phenocrysts in 18 May pumices style. Trace element (lithium) and short-lived radioactive isotope (lead-210 is generally low, whereas those erupted and radium-226) data show that ascending magma stalled within the conduit, before and after 18 May have higher Li leading to the accumulation of volatiles and the formation of lead-210 concentrations (Fig. 1B and Fig. 2). Li excesses, which signals the presence of degassing magma at depth. diffusion in melt (9) and plagioclase (10)is extremely rapid. As Li partitions into vapor On 18 May 1980, an earthquake-triggered degassing, because the key intermediate relative to melt at depth (11), low Li con- 222 0 landslide exposed a subsurface lava ‘‘cryp- isotope Rn (t1/2 3.8 days) is volatile centrations in 18 May plagioclase are con- 218 214 214 214 todome[ at Mount St. Helens that led to ð226Ra Y 222Rn PoY PbY BiY PoY 210PbÞ. sistent with degassing, as inferred from 210Pb its eruption, first in the form of a violent During degassing of major volatile species deficits in the associated whole rocks. 222 lateral blast and subsequently as a 20-km-high such as H2O, SO2, and CO2, Rn will Plinian column and associated pyroclastic diffuse into the gas bubbles, thereby remov- 1.6 flows (1). The eruption followed 2 months of ing 222Rn and leading over time to a 210Pb A precursory activity in the form of seismic deficit in the magma from which volatiles 1.4 2 SD events, phreatomagmatic explosions, and de- escape. Whole-rock samples from the 18 Ra) 1.2 formation, which are attributed to the em- May Plinian phase have a minimum 226 210 226 1.0

placement of the cryptodome. Cryptodome ( Pb/ Ra) of 0.84 (parentheses denote Pb/ 210

samples are typically dense, with a highly activity), which indicates a minimum of 5 ( 0.8 crystalline groundmass of rapidly grown pla- years of continuous pre-eruptive degassing 0.6 gioclase microlites, crystals that provide evi- (7). However, samples from the cryptodome B dence of substantial crystallization produced and several pumices erupted in the summer 14 by decompression and degassing (2, 3). The of 1980 have 210Pb excesses, that is, (ppm) 12 18 May Plinian eruption was followed by a (210Pb/226Ra) 9 1 (Fig. 1A), which have been 10 18 May 25 May series of smaller explosive eruptions during reported in very few studies (7, 8). Contrary to 8 the summer of 1980 and by 6 years of inter- previous suggestions based on an earlier study 6

mittent effusive activity, a pattern interpreted (8), we suggest that there is a trend of plagioclase cores 4 13 Apr 12 Jun 22 Jul 7 Aug 17 Oct to reflect gradually decreasing rates of magma increasing (210Pb/226Ra)withtimeduringthe Li 0 50 100 150 200 250 supply (4, 5). A pronounced increase in summer of 1980 (Fig. 1A). In contrast to the Days since 16 March 1980 groundmass crystallinity of the erupted prod- formation of 210Pb deficits where exsolved Fig. 1. Whole-rock variation of (210Pb/226Ra) and ucts during this time period testifies to the gases leave the system, resulting in the loss of Li concentration in plagioclase during summer increasing importance of degassing-induced 222Rn, 210Pb excesses require gas accumula- 1980. (A) The earliest sign of a pending eruption crystallization (3, 4, 6). Where, when, and tion in the conduit, allowing for the decay of was recorded on 16 March (1); hence, this is how rapidly this crystallization occurred, 222Rn to 210Pb while in contact with the melt, taken as t 0 0. Secular equilibrium is indicated and the physical effects of degassing on the but remote from the site of degassing. Thus, by the dashed line (210Pb/226Ra) at unity. Black evolution of the eruption, are more poorly 210Pb excesses suggest the presence of a squares indicate pumice samples, open squares are dome samples, and dots are data from constrained. deeper degassing reservoir supplying volatiles Bennett et al.(8). (210Pb/226Ra) is measured by a 226 0 The short-lived isotopes Ra (t1/2 to shallow stalled magma. While the creation combination of gamma and alpha spectrometry 210 0 210 1599 years) and Pb (t1/2 22.6 years) of significant Pb deficits requires years (7), (23). The vertical lines with dates indicate can be used to constrain the time scale of 210Pb excesses may be established on much eruptions. A maximum error bar (2 SD) is in shorter time scales. The magnitude of these the top right corner. (B) The lithium concen- 1 Department of Earth Sciences, University of Bristol, excesses depends on the flux of radon relative tration of plagioclase phenocrysts increases Wills Memorial Building, Queens Rd, Bristol BS8 1RJ, 222 with time and follows a pattern similar to United Kingdom. 2GEMOC, Department of Earth and to the rate of Rn decay. A simple (210Pb/226Ra). This figure shows the variation in Planetary Sciences, Macquarie University, Sydney, calculation assuming secular equilibrium be- Li concentration in the cores of phenocrysts NSW 2109, Australia. 3Department of Geological tween (226Ra) and (222Rn) in the deeper [defined here as having Ba concentrations Sciences, 1272 University of Oregon, Eugene, OR between 25 and 80 parts per million (ppm); 4 magma reservoir, complete degassing of 97403–1272, USA. Department of Archaeology, radon, and sufficient time (È1 month) for all see Fig. 2]. After detailed scanning electron School of Human and Environmental Sciences, 222Rn in the gas to decay to 210Pb shows that microscopy study, two to five phenocrysts of University of Reading, Whiteknights, PO Box 227, each eruption were selected to be analyzed by Reading RG6 6AB, UK. the volume of degassing magma needs to be secondary ion mass spectrometry; see (23) for *To whom correspondence should be addressed. roughly 10 times that of the shallow reservoir details. Plagioclase compositions for all sam- 210 226 E-mail: [email protected] to produce a ( Pb/ Ra) of 1.4. If there is ples lie within the range An38 to An82.

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1167 R EPORTS Melt inclusions in phenocrysts can be details the evolution of the melt while The combination of Li enrichment in used to study variation in the concentrations undergoing degassing and crystallization. plagioclase and melt inclusions for all erup- of volatile elements during magma ascent. The melt inclusions are crystal free, and ted rocks except 18 May pumices suggests

H2O concentration depends on the partial their MgO contents indicate minimal post- that Li is added at 4- to 5-km depth beneath pressure of H2O, pH2O(12), and thus, in entrapment crystallization (13). The variation the volcano. The combined data set indi- , 210 222 volatile-saturated CO2-poor magma, it can of Li with H2O (Fig. 3) shows that at pH2O cates that Li and Pb (or Rn) are be used as a monitor of the pressure at which 125 MPa, the melt inclusions record Li transferred from a deeper reservoir (as the melt inclusion was last in contact with enrichment. For all explosive eruptions after erupted during the Plinian phase of 18

the magma. The variable H2O concentrations 18 May, Li enrichment is confined to this May) to shallow-level magma erupted sub- in melt inclusions from Mount St. Helens depth (Fig. 3, scenario 3), whereas for ef- sequently. As Rn and Li are expected to indicate that they have been sealed off from fusive eruptions the magnitude of Li en- partition into the vapor relative to the melt the magma at different depths (Fig. 3). The richment gradually decreases during ascent at high pressure, transfer by volatiles is the compositional variation of melt inclusions (Fig. 3, scenario 1). most likely mechanism for both 210Pb excesses and Li enrichment. Magma stalling 200 at shallow levels where gases from the deeper magma are accumulated can thus acquire an elevated Li concentration and 160 210Pb excess. Time scales of 210Pb and Li enrichment differ. Because of its extremely high diffusivity in melt and plagioclase 120 (9, 10), Li enrichment depends mainly on the gas flux. Conversely, 210Pb enrichment also relies on the decay of 222Rn. Thus, eruptions Ba (ppm) 80 with high Li concentrations and 210Pb deficits (e.g., June eruption, Fig. 1) imply that degassed magma has been stored shal- 40 lowly, but not long enough to develop 210Pb excess. An additional constraint for forma- tion of 210Pb excesses is presented by the 0 seismic and SO2 records, which suggest 2 6 10 14 18 22 26 magma ascent 9 days prior to the 12 June Li (ppm) eruption (1, 14), suggesting that 9 days is in- Fig. 2. Ba and Li concentrations of plagioclase phenocrysts. Ba and Li variation in plagioclase sufficient to create 210Pb excesses. Figure 3 phenocrysts erupted on 13 April (orange triangles), 18 May cryptodome (blue circles), 18 May shows that during subsequent ascent and climactic eruption (black circles), 12 June (purple diamonds), 7 August (yellow squares), and 17 eruption of shallow magma, Li in the melt October dome (green diamonds). For details on the analyses and for data, see (23). is lost to the volatile phase (post–18 May explosive eruptions) or buffered by the gas Li (ppm) flow (15) (effusive eruptions). In contrast, 10 20 30 40 50 60 70 80 90 100 the Plinian 18 May pumices show different 0 behavior with respect to Li and 210Pb: Their low-Li plagioclase cores and 210Pb deficits 50 indicate that the magma has lost volatiles 1 earlier and did not stall during ascent. hours 210 100 The Pb and Li data require gas days/hours

weeks/months accumulation and transfer of volatiles from O) MPa 2 3 vapor to melt at low pressure. Vapor 150 weeks transport in magma may occur by bubble P (H ascent, convection (16), and/or shearing and 200 2

> 5 years fracturing (15). Thus, volcanic gases can accumulate in permeable foam (17–19) 250 underneath an impermeable cap or simply 1 2 3 in a region where magma stalls because of

Fig. 3. H2O and Li concentrations of melt inclusions in phenocrysts from various eruptions during conduit geometry or favorable crystallization the summer of 1980. H2O concentration is taken as an indication of the minimum pressure at conditions. The cause of transfer of Li and which the melt inclusions last equilibrated with the melt. H2O saturation pressures were 210Pb from the vapor to the melt is unclear, calculated using VolatileCalc (12). Melt inclusions from the 18 May pumice were mainly trapped at but, because both form chloride complexes, high pressures and low Li concentrations (circles, scenario 2). Melt inclusions from explosive eruptions after 18 May record high Li concentrations at intermediate depths (tripods, scenario 1), it may be related to the exsolution of an followed by a rapid decrease in Li concentration during ascent. The cryptodome and October dome immiscible brine phase at shallow pressure (open circles and diamonds, respectively) have melt inclusions recording the same Li enrichment (20), as widely proposed for ore-forming event as seen in the pumices, but during their ascent the Li concentration in the melt was buffered hydrothermal systems (21). Whatever the by a continuous gas flow. Li enrichment is inferred to result from gas fluxing, the same process transfer mechanism, the only records of 210 responsible for the formation of Pb excesses (Fig. 1). Deep degassing magma supplies gas with these changes to survive are the melt Rn and Li to shallow conduit magma. If magma stalls, gases can accumulate in the conduit. At this point, the impermeable cap is breached either suddenly (scenario 3), followed by rapid magma inclusions that are sealed off progressively ascent, or more slowly (scenario 1) by leaking gas while ascending slowly as a semipermeable plug. during ascent, the Li concentration in pla- Magma erupted during the 18 May Plinian phase did not stall, but ascended rapidly. gioclase (if ascent is fast enough to prevent

1168 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS diffusive reequilibration), and the disequilib- 3. K. Cashman, Contrib. Mineral. Petrol. 109, 431 (1992). 19. C. Klug, K. C. Cashman, Bull. Volcanol. 58, 87 (1996). rium between 226Ra and 210Pb. Once 210Pb 4. C.-H. Geschwind, M. J. Rutherford, Bull. Volcanol. 57, 20. S. Signorelli, M. R. Carroll, Geochim. Cosmochim. 356 (1995). Acta 64, 2851 (2000). excess is formed, reestablishing secular 5. R. Scandone, S. D. Malone, J. Volcanol. Geotherm. 21. H. L. Barnes, Ed., Geochemistry of Hydrothermal Ore equilibrium between 226Ra and 210Pb would Res. 23, 239 (1985). Deposits (Wiley, New York, ed. 3, 1997). require about 100 years; thus, the 210Pb ex- 6. J. Blundy, K. Cashman, Contrib. Mineral. Petrol. 140, 22. M. J. Rutherford, P. M. Hill, J. Geophys. Res. 98, 631 (2001). 19,667 (1993). cesses formed by accumulation of gases will 7. P.-J. Gauthier, M. Condomines, Earth Planet. Sci. Lett. 23. Methods and data tables are available as supporting persist. Plagioclase grows during decompres- 172, 111 (1999). material on Science Online. sion (under H O-saturated conditions), thus 8. J. T. Bennett, S. Krishnaswami, K. K. Turekian, W. G. 24. During this research, K.B. was supported by a Uni- 2 Melson, C. A. Hopson, Earth Planet. Sci. Lett. 60,61 versity of Bristol scholarship; J.B. and S.T. acknowl- sealing off the melt inclusions shortly after (1982). edge support from the Royal Society. W. Melson at the gas is lost. The available experimental data 9. F. M. Richter, A. M. Davis, D. J. DePaolo, E. B. Watson, Smithsonian Institution provided some of the samples show that Li diffuses more rapidly than all Geochim. Cosmochim. Acta 67, 3905 (2003). used for this study. Other samples were collected with 10. B. J. Giletti, T. M. Shanahan, Chem. Geol. 139,3 the generous support of J. Pallister and M. Clynne other trace elements in melt (9) and plagio- (1997). during field work. We also thank S. Kasemann (Natural clase (10), which suggests that Li can be 11. J. D. Webster, J. R. Holloway, R. L. Hervig, Econ. Geol. Environment Research Council Ion Microprobe Facility, diffusively homogenized in both phases on 84, 116 (1989). Edinburgh) for assistance with secondary ion mass 12. S. Newman, J. B. Lowenstern, Comput. Geosci. 28, spectrometry analysis. time scales of hours. 597 (2002). In the context of Mount St. Helens, our 13. J. Blundy, K. Cashman, in preparation. Supporting Online Material results provide a detailed picture of magma 14. T. Casadevall et al., Science 221, 1383 (1983). www.sciencemag.org/cgi/content/full/1103869/DC1 and gas movement during 1980. Specifically, 15. A. C. Rust, K. V. Cashman, P. J. Wallace, Geology 32, Materials and Methods 349 (2004). Tables S1 to S3 the cryptodome followed a two-stage de- 16. D. S. Stevenson, S. Blake, Bull. Volcanol. 60, 307 (1998). compression path interrupted at a depth of 17. B. E. Taylor, J. C. Eichelberger, H. R. Westrich, Nature 10 August 2004; accepted 1 October 2004 4- to 5-km where magma stalled and gas 306, 541 (1983). Published online 14 October 2004; 18. J. C. Eichelberger, C. R. Carrigan, H. R. Westrich, R. H. 10.1126/science.1103869 accumulated before further ascent. During Price, Nature 323, 598 (1986). Include this information when citing this paper. magma ascent, continued gas fluxing through a semipermeable magmatic foam buffered the Li concentration of the melt and sup- plied Rn (Fig. 3). Evidence for breaching of Ventilation of the Glacial Deep the impermeable cap and gas flow comes from phreatomagmatic eruptions and steam Pacific Ocean venting in March and April (2). Magma 1 1 1 erupted during the Plinian phase of the Wallace Broecker, * Stephen Barker, Elizabeth Clark, eruption ascended rapidly from a depth of Irka Hajdas,2 Georges Bonani,2 Lowell Stott3 more than about 7 km without stalling (22). Before the post–18 May explosive eruptions, Measurements of the age difference between coexisting benthic and planktic magmas stalled at 4- to 5-km before the foraminifera from western equatorial Pacific deep-sea cores suggest that impermeable cap was breached and magma during peak glacial time the radiocarbon age of water at 2-kilometers depth ascended rapidly. A dome sample from Octo- was no greater than that of today. These results make unlikely suggestions ber 1980 also has high Li melt inclusions at that a slowdown in deep-ocean ventilation was responsible for a sizable G fraction of the increase of the ratio of carbon-14 (14C) to carbon in the pH2O 125 MPa, which suggests that magma ascended more slowly than during atmosphere and surface ocean during glacial time. Comparison of 14C ages for the preceding explosive eruptions and, as coexisting wood and planktic foraminifera from the same site suggests that with the cryptodome, Li concentrations in the atmosphere to surface ocean 14CtoCratiodifferencewasnot the melt were buffered. The increase of substantially different from today’s. (210Pb/226Ra) during the summer of 1980 correlates with a decrease in magma ascent Hughen et al.(1) present a strong case that authors call on a sizable reduction in deep- rate (4, 5). This suggests that the trend of during the last glacial maximum (LGM) (i.e., sea ventilation rate and a possible reduction increasing (210Pb/226Ra) with time reflects 22,000 to 16,000 calendar years ago) the 14C in shallow marine carbonate deposition (i.e., prolonged stalling of magma at 4- to 5-km to C ratio in the atmosphere and surface reef growth). Muscheler et al.(2) reach a depth as the eruption intensity waned. ocean was 375 T 25° higher than that for similar conclusion by another route. They Only recently has attention been turned preindustrial time. The obvious explanation make use of the 10Be measurements on the to the complexity of shallow conduit pro- for this increase is that Earth_s magnetic Summit Greenland ice core to reconstruct cesses and their link to eruption style. Our field was on the average weaker, allowing past production rates of cosmogenic isotopes data shed light on processes occurring on more cosmic rays to reach our atmosphere. and conclude that the excess 14C present in time scales of years to hours before an Hughen et al.(1) contend, however, that the atmosphere and upper ocean cannot be eruption and, as such, may provide an aid although the field was weaker, the conse- accounted for by higher production alone. in interpreting observations from established quent increase in 14C production was insuf- They also call on reduced deep-ocean ven- monitoring techniques, for example, gas emis- ficient to explain the entire observed 14C tilation in an attempt to account for at least sions and seismic surveys. 210Pb excesses increase. To explain the remainder, these part of the remaining radiocarbon increase. are coupled to 210Pb deficits deeper in the Were the deep sea to have been ventilated system. Thus, magma erupting with 210Pb 1Lamont-Doherty Earth Observatory of Columbia Uni- at a slower rate during glacial time, then the excess requires the presence of degassing versity, 61 Route 9W/Post Office Box 1000, Palisades, 14C to C ratio difference between upper- NY 10964, USA. 2AMS 14C lab, ETH Hoenggerberg magma at depth. 3 ocean and deep-ocean carbon must have been HPK H27 and H30, CH-8093 Zurich, Switzerland. Uni- _ versity of Southern California, 3651 Trousdale Parkway, larger than today s. As shown in Table 1, References and Notes Los Angeles, CA 90089, USA. raising the atmosphere and surface ocean 14C 1. P. W. Lipman, D. R. Mullineaux, Eds., U.S. Geol. Surv. Prof. Paper 1250, 93 (1981). *To whom correspondence should be addressed. to C ratio by 200 per mil requires a large 2. K. V. Cashman, R. P. Hoblitt, Geology 32, 141 (2004). E-mail: [email protected] increase in the difference between the 14Cto

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1169 R EPORTS C ratio for surface and deep-sea inorganic from below the original one, yielded consis- would be that during the LGM the age carbon. As this difference is potentially re- tent results. difference would have been 630 years. corded by the 14C age difference between As a bonus, in the first sample from Water-column profiles of %14C in the coexisting benthic and planktic foraminifera MD98-2181, we were surprised to find a western equatorial Pacific yield a value of shells in glacial-age sediments, it is possible number of small pieces of wood (5). The two about –220° at 2000-m water depth (10, 11). to check the prediction by Hughen et al.(1) largest pieces were included in the second Using the coral-based value of –58° for and Muscheler et al.(2) that deep-sea run. The age difference of 560 T 150 years preindustrial surface water in that region, ventilation rates were slower during glacial between the wood and planktic foraminif- this corresponds to an age of 1520 years. In time. era is consistent with expectation. Measure- today_s ocean below 1500 m, there is little In a recently published paper (3), a rather ments on corals from the western equatorial change in %14C with water depth. bleak summary was presented of published Pacific indicate that before the onset of nu- The benthic-planktic 14C age differences attempts to establish the age difference be- clear testing the %14C for surface waters in obtained for the Morotai core are 1325, 1170, tween coexisting benthic and planktic fo- that region was –58 T 2° (6, 7), which and 1325 years (Figs. 1 and 2). For the Ad- raminifera in glacial-age sediments from the corresponds to a reservoir age of 480 years miralty core, the differences are, respectively, deep Pacific Ocean. The published results for (8). As pointed out by Bard (9), other things 1954 and 1640 years. Because atmospheric age difference based on individual benthic- being equal, this difference should be inverse- _ planktic pairs range from as low as a few ly related to the atmosphere sCO2 content. 6 1 hundred to as high as 3500 years. New re- Taking the LGM value to be 195 atm and GEOSECS sults from the eastern equatorial Pacific pre- the preindustrial to be 280, the prediction Station 241 sented by Broecker et al.(3) also yield a 4.5 N 179 E wide range of ages for coexisting planktic foraminifera. Further, as acid leaching de- Calendar age (kyr) creased the age of the residual calcite in 17.5 18.5 19.5 20.5 21.5 2 these samples, the suggestion is that sec- 2.2 ondary contaminant calcite must be present. Water depth (km) Water Taken together, the average age difference as 2.0 a function of water depth for the entire ensemble of benthic-planktic pairs is con- 1.8 C age (kyr) 3 14 sistent with that for preindustrial time. How- 0.5 1.0 1.5 2.0 2.5 ever, this average has meaning only if it can 1.6 Age benthic - age planktic (kyrs) be assumed that the biases responsible for the very large range in results are random 1.4 Fig. 2. Summary of LGM ventilation times based (i.e., on the average not different for benthics on benthic-planktic age differences for cores 1.2 from the western Pacific. The open squares are than for planktics). The important point is for cores from the Sea of Okhotsk (16), the solid that only a few of these published age dif- Benthic - planktic triangles are for a core from off Japan (17), the ferences are large enough to be consistent 1.0 open diamonds are from a core at a depth of 2.7 km in the South China Sea (sill depth 2.0 km) with the requirement that the extra atmo- 15 16 17 18 19 sphere and upper ocean 14C was the result of Planktic radiocarbon age (kyr) (15), the solid circles are for the Admiralty Island a slowdown in deep-sea ventilation. core, and the open circles are for the Morotai Fig. 1. Benthic-planktic age differences as a Basin core. The solid diamond is for a 230Th- Having found that foraminifera from function of planktic age for three sites in the dated benthic coral from the Drake Passage (18). glacial-age sediments in the eastern equato- western tropical Pacific. The circles represent Not shown are results obtained on cores stored rial Pacific appear to be contaminated with the results for the South China Sea (15), the dry in the Lamont-Doherty repository (21)and secondary calcite, we turned to sediments triangles those for the Morotai Basin, and the those from off New Zealand (22). The solid line from the western equatorial Pacific in hopes diamonds those for Admiralty Island. The up- is based on measurements of water samples of avoiding this problem. We obtained 50-g per scale gives the approximate calendar age collected as part of the Geochemical Ocean for the planktics. Sections Study (GEOSECS) survey. samples from core MD98-2181 from the Morotai Basin at 6-N and 126-E, a location just to the south of the island of Mindanao Table 1. Calculation based on a simple box ocean model of the decrease in the 14C to C ratio (relative to at a depth of 2.1 km, and from core MD97- that for the preindustrial atmosphere) for deep inorganic carbon required to raise the upper ocean and %14 14 2138 from the open Pacific at 1-S and 146-E atmosphere C with no change in C production. Also given is the increase in the radiocarbon age difference between coexisting benthic and planktic foraminifera. Case 1: Two-box ocean; distribution of just to the north of the Admiralty Islands at inorganic carbon between atmosphere, surface, and deep reservoirs 0 0.75:1:38. Surface ocean reservoir a depth of 1.9 km (4). The planktics in these age is held at 500 years. Case 2: Three-box ocean; as before, except the deep ocean is now divided into samples yielded 14C ages that respectively intermediate and deep boxes. The 14C age difference between the surface and intermediate boxes is held correspond to calendar ages of about 19,000 at 1500 years. and 21,500 years. The results are listed in Table 2. Because of the problems previous- Atmospheric Surface Intermediate Deep Total Benthic-planktic ly encountered for the eastern equatorial 14C:C 14C:C 14C:C 14C:C 14C:C 14C age (years) Pacific samples, the first set of runs was Case 1. Two-box ocean; surface:deep 0 1:38 limited to planktics. Excellent agreement was 1.00 0.940 0.780 0.788 1500 achieved in both cores for four planktic 1.10 1.034 0.775 0.788 2312 species. Based on this success, we conducted 1.20 1.128 0.771 0.788 3057 a second run in which one of the planktic Case 2. Three-box ocean; surface:intermediate:deep 0 1:19:19 species was repeated along with the mixed 1.00 0.940 0.780 0.780 0.788 1500 1.10 1.034 0.858 0.693 0.788 3214 benthics. Subsequent measurements on two 1.20 1.128 0.936 0.606 0.788 4989 additional samples, one from above and one

1170 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS and surface ocean %14C was relatively stable and Muscheler et al.(2)istobeexplained ing water depths of 2.0 T 0.1 km in the west- during the period 25,000 to 19,000 calendar by a slower deep-sea ventilation rate during ern Pacific) suggest a dip in ventilation age years ago, it is not necessary to account for glacial time, the age of this deep carbon in centered at 19,000 calendar years (see Fig. 1). changes in atmospheric %14C when inter- deeper waters would have to have been at However, verification of this dip will require preting these benthic-planktic offsets (12). least 4000 years. further measurements. Further, as the deep sea is continuously ven- Although no reliable 14C data are avail- LGM benthic–planktic age differences tilated, the 14C to C ratio for the deep sea able for these deeper waters, 13C measure- from the Sea of Okhotsk (16) and from off reflects an average over a time interval of on ments of glacial-age benthic foraminifera Japan (17), as well as the estimate based on the order of a millennium rather than a single from equatorial zone cores from 1.6 to 4.5 a 230Th-dated benthic coral from the Drake point in time. The results suggest that waters km depth show no particular trend (14). If Passage (18), are consistent with those pre- at 2000 m during the LGM in the western the deep waters were ventilated much more sented here (see Fig. 2). Taken together they equatorial Pacific were no older with respect slowly, one would expect that the benthic suggest that, if anything, LGM ventilation to surface waters than today and could ac- 13C would have been more negative than that rates were on the average somewhat less tually have been slightly better ventilated at at 2 km. If anything, the data suggest slightly than today_s. around 19,000 calendar years ago. As these better ventilated waters below 3 km in the Clearly, many more measurements will results represent only the mid-depth of the glacial Pacific. be needed to create an adequate radiocarbon Pacific Ocean, it is possible that the deep The benthic-planktic age differences inventory for the LGM ocean. As it is now high-salinity layer documented for glacial presented here are consistent with those pub- clear that the formation of secondary calcite time by Adkins et al.(13) had a much lower lished for the South China Sea where concor- is a problem, it will be essential to obtain 14C to C ratio. However, as more than half dant ages between multiple planktic species cross-checks by analyzing several coexisting the ocean lies above 2000 m, if the dis- were also obtained (15). Taken together, the planktic species. Further, as bioturbation, cou- crepancy pointed out by Hughen et al.(1) results from these three locales (represent- pled with opposing gradients in the abun- dances of planktic and benthic foraminifera Table 2. Radiocarbon ages for two western equatorial Pacific deep-sea cores. The uncertainty quoted for can give rise to sizable biases in the re- mean planktic ages is the standard deviation about the mean age. constructions of surface to deep radiocarbon differences (19, 20), cores with high ac- Depth (cm) Species 14C age (years) 1 sigma error (years) cumulation rates must be sought. This latter restriction eliminates most open ocean sites Morotai Basin MD98-2181 6-N 126-E 2.1 km and focuses attention on sites located along 1262–1268 continental margins. Another problem is that Pulleniatina obliquiloculata 16,650 110 Neogloboquadrina dutertrei 16,800 95 calcite dissolution limits the water depth at Mix benthics 18,050 130 which such measurements can be conducted. Mean planktics 16,725 150 In the Pacific and Indian Oceans, few cal- % B-P 1,325 200 citic shells are preserved at depths greater 1270–1276 than 4400 m. Wood 1 15,950 120 If it turns out that the reduced storage of Wood 2 15,970 120 14 Globigerinoides sacculifer 16,480 120 C in the deep sea is not the answer, then P. obliquiloculata 16,330 100 what are the alternatives? The Muscheler et al. P. obliquiloculata 16,760 110 (2) argument rests on the assumption that the N. dutertrei 16,740 110 ratio of the accumulation of 10Be to accu- Globorotalia tumida 16,290 110 mulation of snow on the Greenland ice cap Mix benthics 17,690 130 was not substantially different during the Mean planktics 16,520 150 % B-P 1,170 220 LGM than during the Holocene. Considering 10 1279–1285 that the accumulation rate of Be on Green- P. obliquiloculata 16,900 130 land is only one-third of the mean global N. dutertrei 17,150 130 production rate, this assumption rests on Mix benthics 18,350 120 shaky ground. If, for example, the accumu- Mean planktics 17,025 150 lation of 10Be relative to snow was lower % B-P 1,325 200 during glacial time, the discrepancy between Admiralty Island MD97-2138 1-S 146-E 1.9 km the reconstructed 14C and 10Be production 207–210 rates would disappear. Globigerinoides sacculifer 18,780 120 Hughen et al._s(1) contention that vari- P. obliquiloculata 18,620 120 _ N. dutertrei 18,270 120 ations in Earth s magnetic field strength are Globorotalia tumida 18,890 120 not large enough to explain the high LGM N. dutertrei 18,620 120 14C to C ratios for the atmosphere and sur- Mix benthics 20,590 140 face ocean present a more serious challenge. Mean planktics 18,636 150 % These authors point out, however, that un- B-P 1,954 220 explained differences between simulated 14C 211–215 Globigerinoides sacculifer 19,130 130 and their observations may be the result P. obliquiloculata 18,970 130 of uncertainties in the relation between 14C N. dutertrei 18,730 140 production rate and geomagnetic intensity. Globorotalia tumida 18,960 120 Increasing the database of reliable paleo deep- Mix benthics 20,590 150 ocean ventilation ages will provide important Mean planktics 18,950 200 constraints for such reconstructions in the % B-P 1,640 250 future (21, 22).

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1171 R EPORTS References and Notes 11. R. M. Key et al., Radiocarbon 44, 239 (2002). planktic 14C age 25.6 ky); 3.48 ky (depth 2.7 km, 1. K. Hughen et al., Science 303, 202 (2004). 12. J. F. Adkins, E. A. Boyle, Paleoceanography 12, 337 planktic 14C age 24.1 ky); and 0.75 ky (depth 1.3 km, 2. R. Muscheler et al., Earth Planet. Sci. Lett. 219, 325 (1997). planktic 14C age 24.8 ky). We choose not to include (2004). 13. J. F. Adkins, K. McIntyre, D. P. Schrag, Science 298, them because they represent a different age range 3. W. S. Broecker, E. Clark, I. Hajdas, G. Bonani, Paleocean- 1769 (2002). from those that we studied. ography 19, doi: 2003PA000974 (2004). 14. K. Matsumoto, T. Oba, J. Lynch-Stieglitz, H. Yamamoto, 23. We thank L. Beaufort for providing the Admiralty 4. L. Beaufort, T. de Garidel-Thoron, A. C. Mix, N. G. Pisias, Quat. Sci. Rev. 21, 1693 (2002). core sample and T. Guilderson for providing a survey Science 293, 2440 (2001). 15. W. S. Broecker, T.-H. Peng, S. Trumbore, G. Bonani, of coral-derived prebomb %14C values. This paper is 5. Alex Wiedenhoeft of the Forest Products Laboratory W. Wolfli, Global Biogeochem. Cycles 4, 103 (1990). funded in part by a grant/cooperative agreement from in Madison, Wisconsin, advised us that it was not pos- 16. L. D. Keigwin, J. Oceanogr. 58, 421 (2002). the National Oceanic and Atmospheric Administration sible to identify milligram-sized fragments of tropical 17. K. Ohkushi, M. Uchida, N. Ahagon, T. Mishima, (NOAA). The views expressed herein are those of the woods; rather, a hand-sized piece of wood would be T. Kanematsu, Nucl. Instrum. Methods Phys. Res. authors and do not necessarily reflect the views of required. He also commented that, once waterlogged, Sec. B, 406, 223 (2004). NOAA or any of its subagencies. Financial support was any wood would be dense enough to sink in seawater. 18. S. J. Goldstein, D. W. Lea, S. Chakraborty, M. Kashgarian, provided by NSF grant OCE 02-21979 and by NOAA 6. T. P. Guilderson, D. P. Schrag, M. A. Cane, J. Clim. 17, M. T. Murrell, Earth Planet. Sci. Lett. 193, 167 (2001). Consortium on the Ocean’s Role in Climate (CORC) 1147 (2004). 19. M. Andree, Radiocarbon 29, 169 (1987). grant UCSIO P.O. 10196097-003/NA17RJ1231. This 7. T. Guilderson, personal communication, 2004. 20.W.S.Broecker,K.Matsumoto,E.Clark,I.Hajdas, study was funded in part by a Lamont-Doherty post- 8. Because conventional radiocarbon ages are calcu- G. Bonani, Paleoceanography 14, 431 (1999). doctoral fellowship and a fellowship grant (award lated using the Libby half life of 5568 years, we have 21.E.L.Sikes,C.R.Samson,T.P.Guilderson,W.R.Howard, number CSEF CC3B) from the Comer Science and used this value to calculate all age differences in this Nature 405, 555 (2000). Education Foundation to S.B., Lamont-Doherty Earth study for consistency. 22. Four benthic-planktic age differences for LGM sam- Observatory contribution no. 6663. 9. E. Bard, Paleoceanography 3, 635 (1988). ples from cores off New Zealand are reported in 10.W.S.Broecker,W.C.Patzert,J.R.Toggweiler,M.Stuiver, (21): 2.12 thousand years (ky) (depth 2.07 km, J. Geophys. Res. 91, 14345 (1986). planktic 14C age 24.4 ky); 0.84 ky (depth 2.07 km, 2 July 2004; accepted 8 October 2004

spectively (Table 1)^, they retained about one- Prospects for Building the third of the original sequences and a substan- tial fraction (74% and 95%) of the taxonomic Tree of Life from Large diversity originally contained in the sample. We screened minimal phylogenetic clus- Sequence Databases ters for the presence of paralogs with a phylogenetic test of orthology (5). A species Amy C. Driskell,1,2* Ce´cile Ane´,1.- J. Gordon Burleigh,1. tree cannot be easily deduced from a cluster Michelle M. McMahon,1. Brian C. O’Meara,2. containing both orthologs and paralogs, al- Michael J. Sanderson1 though methods for this have been proposed (6–8). Screening reduced the candidate mini- We assess the phylogenetic potential of È300,000 protein sequences sampled mal clusters to smaller sets of orthologous from Swiss-Prot and GenBank. Although only a small subset of these data was Bsingle-copy[ clusters retaining only 24% potentially phylogenetically informative, this subset retained a substantial and 21% of the original sequences, but still fraction of the original taxonomic diversity. Sampling biases in the databases covering 59% and 89% of the original taxa necessitate building phylogenetic data sets that have large numbers of missing in Swiss-Prot and GenBank, respectively entries. However, an analysis of two ‘‘supermatrices’’ suggests that even data (Table 1). These sequences are very sparsely sets with as much as 92% missing data can provide insights into broad distributed among taxa as measured by their sections of the tree of life. Bdensities[ (Table 1) (9). Further assessment of the phylogenetic More than 100,000 species—about 6% of all We examined the phylogenetic informa- utility of these data requires consideration of those known to science—have at least one tion content of the Swiss-Prot database of how the data should be parsed for phyloge- molecular sequence archived in public data- 120,000 sequences for nearly 7500 taxa and netic analyses. One approach is to build gene bases, but what fraction of these sequences is a Btaxonomically enriched[ subset of GenBank, trees from individual clusters and to as- phylogenetically informative? Here, we ex- which consisted of 185,000 amino acid se- semble these trees using supertree methods amine two large samples of proteins and quences for more than 16,000 green plant (3, 10). Supertree methods require at least show how the answer depends on the pattern taxa (1). Clusters of putative homologs were partial taxonomic overlap between trees. A of homology among sequences and the identified via N Â N BLAST searches (2). set of trees (each inferred from a cluster) distribution of sequences among taxa. We Clustering procedures involve trade-offs with enough taxonomic overlap to allow then parse these databases into phylogenetic among the reliability of homology assess- supertree construction is a Bgrove[ (1). The supermatrices for metazoans and green ment, the taxonomic breadth, and the accu- minimum number of groves in a database is plants. Although the databases have sam- racy of tree inference. The trade-offs are a lower bound on the number of supertrees pling biases that cause these matrices to be controlled by the stringency of homology required to encompass all its sequence data. very sparse, they can still provide useful searches and can be adjusted to maximize The single-copy green plant proteins form at information for building the tree of life. the phylogenetic utility of resulting clusters, least 15 groves (Table 1). The largest of on the basis of the depth and breadth of the these groves minimally includes trees from 1Section of Evolution and Ecology, 2Center for Popu- phylogenetic question to be addressed (3, 4). 814 clusters and contains more than 14,000 lation Biology, University of California, One Shields Clusters containing at least four taxa are taxa—87% of all the green plant taxa in the Avenue, Davis, CA 95616, USA. termed Bminimal phylogenetic clusters,[ be- GenBank database. Swiss-Prot has at least *To whom correspondence should be addressed. cause unrooted trees with fewer than four eight times as many groves, which reflects its E-mail: [email protected] taxa contain no information about relations. greater taxonomic breadth but higher frag- .These authors contributed equally to this work. - Although minimal phylogenetic clusters were mentation (Table 1). Both data sets also Present address: Department of Statistics, University E of Wisconsin, Medical Science Center, 1300 University a small fraction of all clusters found 6.5% contain a small number of Borphans,[ clusters Avenue, Madison, WI 53706, USA. and 2.3% for Swiss-Prot and GenBank, re- with no taxonomic overlap with other clusters.

1172 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS A second and more widely used strat- bootstrap values (figs. S1 and S2). However, lations in the final tree. The level of conflict egy is to concatenate clusters into multi- some of the Bbackbone[ of the green plant in green plants is remarkably high: More gene (protein) data matrices. Increasing the tree is weakly supported, and some uncon- individual protein trees conflict with any sequence data per taxon should improve ac- ventional and likely incorrect relations are given final clade than support it. Nonetheless, curacy (11, 12), a theoretical result support- depicted in both topologies. Many of these many nodes on this tree are still well ed by several recent phylogenomic studies unconventional relations have been seen in supported by bootstrap values. Surprisingly (13–15). However, as with supertree con- previous molecular studies (19–22), includ- little relation was seen between the number struction, the taxonomic structure of the set ing even the nonmonophyly of monocots of proteins supporting a node and its boot- of clusters limits the size of data matrices (23). To characterize the sources of these strap score. For example, placement of the that can be assembled from them. To explore signals, we compared (Fig. 3) every clade in lepidopteran Spodoptera frugiperda within these limits, we used an exact algorithm each individual protein tree against the final the Diptera receives 96% bootstrap support, derived in the context of a well-known graph trees (figs. S1 and S2). Compared with the but only one protein cluster of the 1131 problem on Bbicliques[ (5, 16) to enumerate green plant supermatrix, fewer of the proteins sampled contains sequence for both Spodop- all possible concatenated matrices that are in the Swiss-Prot data conflict with the re- tera and other dipterans (none contain both maximal (not contained in larger matri- ces) and complete (no missing sequences). The largest complete multiprotein Swiss-Prot A 100 B 250 Fig. 1. Size distribution of and GenBank matrices have either many taxa 1 x 472 1 x 6492 the maximal complete su- 90 2 x 975 permatrices (maximal and few proteins or the reverse; none has 200 80 bicliques) for two data large numbers of proteins and taxa simulta- sets. Each point represents neously (Fig. 1), and these matrices retain 70 150 the biclique with the larg- only a small fraction of the original taxo- 60 est number of taxa for a nomic diversity in the databases (Table 1). 100 particular number of clus- The numbers of taxa and genes in a matrix 50 ters. (A) Swiss-Prot. (B) 40 50 Green plants from GenBank. can be greatly increased by allowing missing Bicliques on the tails are sequences, or Bholes,[ in the matrix (17, 18). 30 0 not shown, but their sizes Such Bsupermatrices[ are also affected by 0 100 200 300 400 500 600 are indicated in the inset Maximum number of taxa number Maximum 20 the grove structure of the database, because boxes. the missing entries induce patterns of partial 10 overlap among taxa. 0 Statistics on clusters, groves, and bicliques 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 offer a glimpse of the potential phylogenetic Number of clusters information content of the sequence data- bases. To explore the feasibility of realizing Table 1. Summary statistics on the phylogenetic information content of proteins from Swiss-Prot and this potential, we assembled supermatrices GenBank (green plant only) databases. ‘‘Minimal phylogenetic clusters’’ contain at least four taxa. spanning a small (but broad) taxonomic sam- ‘‘Nontrivial’’ bicliques contain at least two genes and four taxa. Density 0 (number of cells in the taxon  ple from each database. We identified the cluster matrix containing Q 1 sequence)/(number of taxa  number of clusters) (1). See text for other largest grove and constructed a super- definitions. matrix from complete matrices having at Swiss-Prot GenBank least 10 clusters and four taxa (1). The green Database statistics release 40.29 release 137 (green plant) plant matrix retained representatives across the group, but Swiss-Prot was culled to Summary statistics metazoans (plus fungal outgroups) to avoid Number of sequences in release 121,218 185,418 conflicting gene histories of nuclear, mito- Total number of sequences clustered 121,218 185,089 chondrial, and chloroplast genomes. The Total number of taxa 7449 16,348 resulting two supermatrices had compa- Total number of clusters 64,712 59,144 Minimal phylogenetic clusters rable numbers of taxa and shared no se- Number of clusters 4214 (6.5%) 1365 (2.3%) quences. The Swiss-Prot supermatrix had Sequence coverage 41,812 (34%) 65,113 (35%) 70 taxa  1131 genes (6623 sequences and Taxon coverage 5538 (74%) 15,599 (95%) 469,497 characters), an average of 95 genes Single-copy clusters per taxon. The green plant supermatrix con- Number of clusters 3592 (5.6%) 853 (1.4%) Sequence coverage 28,742 (24%) 39,443 (21%) tained 69 taxa 254 genes (2777 sequences  Taxon coverage 4404 (59%) 14,502 (89%) and 96,698 characters), an average of 40 genes Density 0.0018 0.0021 per taxon. These are among the largest super- Groves matrices yet analyzed for phylogenetic infer- Minimum number of groves 123 15 ence, as well as the sparsest, with 92% and Number of orphan clusters 67 7 84% missing entries, respectively (Fig. 2). Minimum number of clusters in largest grove 3183 814 Yet, although sparse, they are 46 and 75 times Minimum number of sequences in largest grove 25,272 (21%) 38,700 (49%) Minimum number of taxa in largest grove 2695 (36%) 14,169 (87%) as dense, respectively, as the single-copy Bicliques minimal sequence collections from which Number of nontrivial maximal bicliques 43,576 5587 they were derived (Table 1). Sequence coverage 23,855 (20%) 15,092 (8.2%) Trees from these matrices (1) broadly Taxon coverage 1449 (19%) 4230 (26%) agree with conventional views on phyloge- Number of clusters in biclique set 3187 (4.9%) 645 (1.1%) netic relations, and many nodes, particularly Largest biclique (in terms of taxa) 2  76 2  975 Largest biclique (in terms of clusters) 352  470 4 in the Swiss-Prot tree, are well supported by

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1173 R EPORTS sequence from Spodoptera and both Anoph- contain more taxa (two or more times as added twist that genes need not be sampled eles species) and can therefore shed light on many), but one-fourth or fewer proteins per evenly across taxa and a surprising amount the final topology. Other relations, such as taxon, with a substantially lower overall of missing data is tolerable (18, 25). Never- the placement of Spathiphyllum outside of the density. In two studies with G15 taxa, boot- theless, even our threshold of 10 proteins per main clade of monocots, are not supported by strap support was uniformly very high across taxon was not enough to prevent recovering a single protein and must therefore be an the tree (14, 15) and could be obtained even some unconventional results, as our surpris- emergent property of weak signals buried with a subset of sequences (14). In two other ing findings regarding monocots document. within several proteins. studies (13, 18) with larger taxon sets (30 to Such results, if they stem from genomewide These supermatrices differ from other 36 taxa), many clades were strongly sup- long branch–attraction artifacts (26), will recent phylogenomic analyses of diverse ported but some were not—an expected also occur in phylogenomic studies unless taxa (13–15, 18) in two important respects. consequence of increased taxon sampling problematic taxa are deliberately or acci- First, the taxa and proteins in our super- (24). Our results support the general conclu- dentally excluded and will only be overcome matrices were determined largely by the sion of these other studies that combining by additional taxon sampling or intensive structure of the databases rather than by many genes can provide strong support for analytical efforts. Therefore, we conclude that decisions of the investigators. Second, they nodes in a large and complex tree, with the exploitation of existing databases, taking into account the inherent sample biases of the data, provides a cost-effective comple- ment to intensive genomewide sequencing efforts, especially if we wish to include large numbers of taxa or remote corners of the tree of life.

References and Notes 1. Materials and methods are available as supporting material on Science Online. 2. I. Dondoshansky, BLASTCLUST vers. 6.1, (National Cen- ter for Biotechnology Information, Bethesda, MD, 2002). 3. O. R. P. Bininda-Emonds, S. G. Brady, J. Kim, M. J. Sanderson, Pac. Symp. Biocomp. 6, 547 (2001). 4. Z. Yang, Syst. Biol. 47, 125 (1998). 5. M. J. Sanderson, A. C. Driskell, R. H. Ree, O. Eulenstein, S. Langley, Mol. Biol. Evol. 20, 1036 (2003). 6. R. D. M. Page, Mol. Phylogenet. Evol. 14, 89 (2000). 7. J. Kim, B. Salisbury, Pac. Symp. Biocomp. 6, 571 (2001). 8. L. Arvestad, A.-C. Berglund, J. Lagergren, B. Sennblad, Fig. 2. Visualization of the distribution of sequences among taxa for the supermatrices. (A) Swiss- Bioinformatics 19 (suppl. 1), i7 (2003). Prot metazoan supermatrix. (B) GenBank green plant supermatrix. Columns correspond to clusters 9. M. J. Sanderson, A. C. Driskell, Trends Plant Sci. 8, (proteins); rows to taxa, ordered so that density increases to upper right. Black indicates a 374 (2003). sequence is present in that cluster for that taxon. 10. M. J. Sanderson, A. Purvis, C. Henze, Trends Ecol. Evol. 13, 105 (1998). 11. P. L. Erdo¨s, M. A. Steel, L. A. Sze´kely, T. J. Warnow, 0.05 Rand. Struct. Algorithms 14, 153 (1999). 12. D. M. Hillis, Nature 383, 130 (1996). 13. E. Bapteste et al., Proc. Natl. Acad. Sci. U.S.A. 99, 1414 (2002). 14. A. Rokas, B. Williams, N. King, S. Carroll, Nature 425, nodes from the metazoan tree 798 (2003). nodes from the green plant tree 15. E. Lerat, V. Daubin, N. Moran, PLoS Biol. 1, 101 (2003). 16. G. Alexe et al., DIMACS Tech. Rep. 2002-52 (2002). 17. C. Yan, J. G. Burleigh, O. Eulenstein, in preparation. 18. H. Philippe et al., Mol. Biol. Evol. 21, 1740 (2004). 19. A. M. D’Erchia, C. Gissi, G. Pesole, C. Saccone, U. Arnason, Nature 381, 597 (1996). 20. M. J. Phillips, D. Penny, Mol. Phylogenet. Evol. 28,

Density estimate 171 (2003). 21. R. Zardoya, Y. Cao, M. Hasegawa, A. Meyer, Mol. Biol. Evol. 15, 506 (1998). 22. M. R. Duvall, A. B. Ervin, Mol. Phylogenet. Evol. 30, 97 (2004). 23. D. E. Soltis et al., Ann. Mo. Bot. Gard. 84, 1 (1997). 24. J. Kim, Syst. Biol. 45, 363 (1996). 0 25. J. J. Wiens, Syst. Biol. 52, 528 (2003). Ð40 Ð20 0 20 40 60 80 26. J. Felsenstein, Biol. J. Linn. Soc. 16, 183 (1978). Difference between the number of nonconflicting and conflicting quartet-informative genes 27. See www.r-project.org/. 28. This research was supported by the National Science Fig. 3. Protein-by-protein distribution of support for clades in supermatrix trees (figs. S1 and S2). A Foundation. Thanks to R. Piaggio and O. Eulenstein for protein tree is ‘‘quartet informative’’ for a given branch (bipartition) of the supermatrix tree if its insights on the properties of groves and Wen-Chieh Chang for biclique code. We also thank D. Ferna´ndez- data set had sequence for at least one taxon in each of the four groups attached to this branch. It Baca and J. Kim. is considered ‘‘nonconflicting’’ if it has at least one most-parsimonious (MP) tree displaying the bipartition (pruned to the taxon set for the protein) or if all of its MP trees display a polytomy in Supporting Online Material place of the branch. Otherwise, the protein is classified as ‘‘conflicting.’’ Plots are smoothed www.sciencemag.org/cgi/content/full/306/5699/1172/ DC1 histograms (density estimates) of the number of quartet-informative proteins that are Materials and Methods nonconflicting minus the number that conflict. Values to the right of the origin are therefore Figs. S1 and S2 branches in which the majority of proteins do not conflict. Density estimation was done with R statistical software (27), using a rectangular kernel and bandwidth of 3 for both trees. 28 June 2004; accepted 14 September 2004

1174 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS to be associated with species vulnerability to local extinction (20, 24). Nested subset analy- Realistic Species Losses sis can thus provide an empirical species loss order through either space or time at a specified Disproportionately Reduce Grassland scale and for a particular community type. We observed 4 years of spatial variation in Resistance to Biological Invaders the plant species richness of plots in a California grassland ecosystem to quantify Erika S. Zavaleta* and Kristin B. Hulvey the degree to which changes in richness occurred in a consistent, nested order (25). Consequences of progressive biodiversity declines depend on the functional We treated nested patterns of species occur- roles of individual species and the order in which species are lost. Most studies rences through space as indicators of a of the biodiversity–ecosystem functioning relation tackle only the first of representative order of local extinction, be- these factors. We used observed variation in grassland diversity to design an cause similar mechanisms appear to affect experimental test of how realistic species losses affect invasion resistance. richness through space and time in our study Because entire plant functional groups disappeared faster than expected by system. Variation through space in species chance, resistance declined dramatically with progressive species losses. richness at our level, treeless site appears to Realistic biodiversity losses, even of rare species, can thus affect ecosystem reflect heterogeneous soil conditions resulting processes far more than indicated by randomized-loss experiments. from gopher activity (up to present) and past mechanical soil disturbance and vehicular Rapid biodiversity losses at both local and quantifies the degree to which species dis- compaction (as recently as 30 years ago) global scales disproportionately involve spe- appearances are ordered, most often across a (26). The resulting heterogeneity likely affects cies with particular values of traits such as series of habitat fragments of decreasing size species richness both directly and indirectly size, trophic position, rarity, distribution, and or increasing isolation (22, 23). Although through effects on productivity (27–29)and degree of specialization (1–5). While the evo- seldom extended in this way, nested subset invasibility (30, 31). These mechanisms re- lutionary consequences of the nonrandomness analysis can also be used to quantify the order semble important drivers of grassland biodi- of biodiversity loss have been explored re- in which species are lost through space or time. versity change in California and elsewhere, cently (6), our understanding of the functional It bases this order solely on observations of including soil disturbances associated with consequences of nonrandom loss remains species occurrences across sites or dates rather agriculture and livestock production, biologi- limited mainly to studies of the declines of than on assumptions about specific drivers of cal invasions, and increased productivity due to individual species (7, 8) Ebut see (9)^. These diversity loss or species characteristics thought anthropogenic nitrogen fertilization (32, 33), studies focus on keystone species or on species that were once widespread and abundant. How- ever, most species in ecosystems—and most Table 1. Species compositions and traits comprising each experimental diversity level. Values are target numbers of stems for each species in a given treatment. Life history traits: early annual (E), indeterminate at-risk species—are less dominant and exert annual/late annual (L), biennial/perennial (P). Functional types were defined as early grasses, early forbs, what influence they have on ecosystem func- nitrogen fixers, late forbs, and perennials based on groupings from previous California grassland studies tioning in relatively small numbers (10–12). (13, 38). Abundance ranks are means based on field observations of neighborhood relative abundances The prevailing experimental approach to across plots of varying richness; 1 indicates most locally abundant species. testing the functional consequences of biodi- Life Abundance versity change isolates the effects of species Species Diversity level Type N fixer or functional group richness by (i) random- history rank izing species composition and (ii) equalizing No. species 369121521 species abundances within each richness No. functional types 2 3 3445 treatment (13–16). This approach provides little information about the importance of spe- Avena barbata 206 149 110 98 74 24 Grass E 1 cies loss order EBcommunity disassembly[ (5)^ Bromus hordeaceus 41 41 41 41 41 41 Grass E 2 Geranium dissectum 21 20 21 20 21 21 Forb E 6 or relative species abundances for ecosystem Bromus diandrus 13 13 12 12 12 Grass E 7 processes. Recent studies generating nonran- Lolium multiflorum 35 28 24 18 7 Grass E 5 dom species losses through removals of rare Vicia sativa 10 10 10 10 10 Forb L yes 8 and uncommon species from natural com- Avena fatua 34 33 34 33 Grass E 4 munities provide more insight about the Erodium botrys 9 9 9 9 Forb E 9 effects of nonrandom loss on ecosystem Vulpia microstachys 2 2 2 2 Grass E 17 Anagallis arvensis 10 27 63 Forb L 3 functioning (17, 18), but local abundance or Briza minor 7 10 15 Grass E 10 rarity alone is an imperfect predictor of Epilobium 2 3 5 Forb L 14 species loss order (19) (Table 1). brachycarpum Patterns of nonrandom change and varia- Crepis vesicaria 2 2 Forb P 19 tion in diversity can be better assessed with Torilis arvensis 3 3 Forb L 13 tools specifically designed for the measure- Trifolium hirtum 2 2 Forb L yes 18 Danthonia 5 Grass P 12 ment of extinction order, such as nested subset californica analysis (20, 21). Nested subset analysis Hemizonia congesta 2 Forb L 16 Hordeum murinum 2 Grass E 15 Medicago 2 Forb L yes 21 Environmental Studies Department, University of California, Santa Cruz, CA 95064, USA. polymorpha Phalaris aquatica 2 Grass P 20 *To whom correspondence should be addressed. Rumex acetosella 6 Forb P 11 E-mail: [email protected]

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1175 R EPORTS although they might capture less well species rare as observed in our field plots (Fig. 1). diversity treatments with 91 species. Function- loss orders likely to be associated with other Constructed communities differed in the al group number was held constant across drivers of grassland biodiversity change, such degree of dominance (dominance was higher diversity levels with 91 species in the earlier as woody encroachment and climate change. in more species-poor treatments) but not the study. Our nested subset analyses show, Species-by-site occurrence matrices at identity of the most abundant species. This however, that in our study system, progressive our study location were significantly nested allowed us to distinguish the effects of diversity declines are not evenly distributed in all years (T 0 19.4- to 23.4-, P ¡ 0.001) progressive changes in species richness from across species varying in key functional traits (34). Local rarity was an inconsistent pre- effects associated with the identity of the (Table 1). In our system, species-poor assem- dictor of loss order, with some abundant dominant species (17, 36). We invaded half blages contain mainly early annuals, with no species absent from low-richness plots and of the communities at each richness level or few indeterminate, late-season, nitrogen- some uncommon species present in plots of with Centaurea soltitialis L. (yellow star- fixing, or perennial species that overlap strong- both high and low richness (table S1). The thistle), an expanding California grassland ly with starthistle in the location or timing of most abundant species in our richest plots, exotic that has already caused considerable resource uptake. Progressively richer com- Anagallis arvensis (Primulaceae), seldom ecological and economic damage (37). munities gain late-season annual forbs and occurred at all in plots below the median Starthistle biomass increased with pro- perennial species, the former of which were richness level (Table 1). All of the 38 species gressive species loss by more than 100% shown to compete effectively against star- at our site maintained their approximate from the most diverse to the least diverse thistle (which is also a late-season annual ranked positions in the nested order of communities (F 0 5.58, P 0 0.002, R2 0 0.59) forb) in monoculture (38). The effects of diversity changes across years as individual (Fig. 2). Starthistle production and flower reducing species richness while holding plots increased or declined in species rich- number were strongly correlated (Pearson functional group number constant, while ness (table S1). This indicated that the order coefficient 0 0.790, Bartlett_s c2 0 33.8, P G theoretically important, thus do not repre- of species losses and gains through space 0.001), such that starthistle reproduction also sent the effects of real biodiversity variation. from grassland patches is robust to interan- increased with declining richness. Lost spe- Consequently, the incorporation of realistic nual variability and change at our site. We cies effects on invader performance were species loss order into our experimental used this observed nested order of species highly disproportionate to their abundance in design profoundly altered the observed rela- loss and gain to design a test of how these the community. For example, mean starthistle tion between diversity and invasibility. ordered changes in species richness influ- biomass was 960% lower in 20-species than The decline in functional diversity that ence grassland resistance to invasion, an in 15-species communities, even though the accompanies realistic species losses can also ecosystem function of growing conservation additional 5 species in the 20-species com- influence the degree to which invasions relevance as invasions accelerate (35). munities together comprised G4% of resi- affect ecosystem processes. Starthistle in- We constructed outdoor microcosm com- dent production and G3% of the total creased total biomass (resident þ starthistle) munities by planting locally collected seeds at number of stems. Similarly, mean starthistle more at low than at high resident species 9 0 0 2 0 six levels of species richness to reflect (i) the biomass was 70% lower in 20-species than richness levels (F 6.77, P 0.01, Radj observed, nested order of local-scale variation in 12-species communities, while the 8 addi- 0.17), reflecting greater total starthistle in species richness and (ii) variation in tional species made up G5% of resident biomass in the more species-poor communi- relative species abundances that accompanied production and G7% of total stems. ties. The per-unit impact of starthistle on variation in species richness at our site (Table Our findings differ from the results of resident biomass, in contrast, increased with 1). At all species richness levels, species studies at the same site testing the effects of species richness, from G0.3 to nearly 1 g abundance ranks followed a log series–like randomized changes in species richness on resident biomass lost per g starthistle present distribution (11), with abundances varying starthistle (38). In this earlier work, no dif- (F 0 4.79, P 0 0.04) (Fig. 3). Most likely, by two orders of magnitude and most species ferences in invader production occurred among

1 100.0 440

340 0

10.0 240

Abundance (%) Abundance 140 -1

Starthistle ANPP (g/m2)

1.0 40 8 10 12 14 16 18 20 22 0 5 10 15 20 25 Realized species richness -2 Species rank 0 5 10 15 20 25 Fig. 2. The relation between realized species Target richness Fig. 1. Target rank-abundance curves for 3- richness and starthistle aboveground net pri- (closed circles), 6- (open squares), 12- (closed mary production (ANPP). Realized richness Fig. 3. Effect of species richness on per-unit triangles), and 21-species (open triangles) values reflect effects of negative (species that invader impact (change in resident biomass/gram treatments. Target treatment compositions failed to establish) and positive (unplanted of starthistle biomass) T 1 SE. This metric 0 0 display log series–like patterns of relative volunteers) deviations from treatment richness when resident biomass is unaffected by star- abundance, with most species relatively un- levels. Target diversity F 0 5.58, P 0 0.002, R2 0 thistle invasion and –1 when 1 g of resident common and a single dominant species across 0.59; realized deviations F 0 7.85, P 0 0.01. biomass is displaced by each g of establishing 0 0 0 2 0 all diversity levels. 95% confidence limits are shown. starthistle. N 5, F 4.79, P 0.04, R adj 0.12.

1176 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS this reflects more complete use of certain S. W. Pacala, D. Tilman, Eds. (Princeton Univ. Press, 27. E. Weiher, S. Forbes, T. Schauwecker, J. B. Grace, resources by the more species-rich assem- Princeton, NJ, 2001), vol. 33, pp. 294–313. Oikos 106, 151 (2004). 2. K. Henle, K. F. Davies, M. Kleyer, C. Margules, J. Settele, 28. B. L. Foster, K. L. Gross, Ecology 79, 2593 (1998). blages. As a result, starthistle added substan- Biodivers. Conserv. 13, 207 (2004). 29. D. E. Goldberg, T. E. Miller, Ecology 71, 213 (1990). tial biomass to species-poor communities 3. L. R. Belyea, J. Lancaster, Oikos 86, 402 (1999). 30. L. F. Huenneke, S. P. Hamburg, R. Koide, H. A. while mainly displacing resident biomass in 4. K. A. McDonald, J. H. Brown, Conserv. Biol. 6, 409 (1992). Mooney, P. M. Vitousek, Ecology 71, 478 (1990). 5. B. Fox, Evol. Ecol. 1, 201 (1987). 31. E. I. Newman, Nature 244, 310 (1973). species-rich communities. Invasibility can 6. A. Purvis, P. M. Agapow, J. L. Gittleman, G. M. Mace, 32. A. R. Watkinson, S. J. Ormerod, J. Appl. Ecol. 38, 233 thus decline while per-unit invader impact Science 288, 328 (2000). (2001). on the resident community increases, under- 7. J. A. Estes, M. T. Tinker, T. M. Williams, D. F. Doak, 33. C. J. Stevens, N. B. Dise, J. O. Mountford, D. J. Science 282, 473 (1998). Gowing, Science 303, 1876 (2004). scoring the importance of measuring both. 8. D. E. Blockstein, Science 279, 1831c (1998). 34. W. Atmar, B. D. Patterson, Nestedness Tempera- This study helps bridge the gap between 9. R. S. Ostfeld, K. LoGiudice, Ecology 84, 1421 (2003). ture Calculator (AICS Research, Inc., University our understanding of general biodiversity- 10. F. W. Preston, Ecology 43, 185 (1962). Park, NM, and The Field Museum, Chicago, IL, 1995); 11. A. E. Magurran, Ecological Diversity and Its Measure- www.aics-research.com/nestedness/tempcalc.html. function relations and the role of extinction ment (Princeton Univ. Press, Princeton, NJ, 1988). 35. P.M.Vitousek,C.M.D’Antonio,L.L.Loope,M.Rejmanek, order in determining the consequences of 12. International Union for the Conservation of Nature R. Westbrooks, N. Z. J. Ecol. 21, 1 (1997). biodiversity loss. Additional experiments are (IUCN), (IUCN, Gland, Switzerland, 2003), vol. 2004. 36. L. W. Aarssen, Oikos 80, 183 (1997). 13. D. Hooper, P. Vitousek, Science 277, 1302 (1997). 37. F. S. Chapin et al., Nature 405, 234 (2000). needed to assess the consequences of ordered 14. D. Tilman et al., Science 277, 1300 (1997). 38. J. S. Dukes, Oecologia 126, 563 (2001). species losses for other ecosystems and eco- 15. S. Naeem, S. Li, Nature 390, 507 (1997). 39. We thank N. Chiariello, C. Field, T. Tobeck, E. Cleland, system functions, as well as to expand re- 16. M. Loreau, S. Naeem, P. Inchausti, Eds., Biodiversity E. Hadly, J. Kriewall, D. Croll, R. Shaw, the Carnegie search designs to incorporate species losses and Ecosystem Functioning: Synthesis and Perspectives Institution of Washington, and the Jasper Ridge Bio- (Oxford Univ. Press, New York, 2002). logical Preserve for their valuable contributions and occurring through time at larger spatial 17. M. D. Smith, A. K. Knapp, Ecol. Lett 6, 509 (2003). assistance. K. Andonian, D. Doak, J. Dukes, G. Gilbert, scales. If, as we found, important function- 18. K. G. Lyons, M. W. Schwartz, Ecol. Lett. 4, 358 (2001). P. Holloran, D. Hooper, K. Honey, D. Letourneau, al traits disappear more rapidly than ex- 19. D. Rabinowitz, in The Biological Aspects of Rare Plant J. Levine, B. Tershy, and two anonymous reviewers Conservation, H. Synge, Ed. (Wiley, London, 1981), improved the manuscript. This project was generously pected by chance in other communities, the pp. 205–217. supported by a David H. Smith Conservation Re- ecosystem consequences of real biodiversity 20. D. H. Wright, J. H. Reeves, Oecologia 92, 416 (1992). search Fellowship through The Nature Conservancy. losses—even of rare species—will often ex- 21. D. H. Wright, B. D. Patterson, G. M. Mikkelson, A. Cutler, W. Atmar, Oecologia 113, 1 (1998). Supporting Online Materials ceed expectations based on randomized di- 22. B. D. Patterson, W. Atmar, Biol. J. Linn. Soc. 28,65 www.sciencemag.org/cgi/content/full/306/5699/1175/ versity studies. (1986). DC1 23. R. Kadmon, Ecology 76, 458 (1995). Materials and Methods 24. E. A. Hadly, B. A. Maurer, Evol. Ecol. Res. 3, 477 (2001). Table S1 References and Notes 25. Materials and methods are available as supporting References 1.S.P.Lawler,J.J.Armesto,P.Kareiva,inThe material on Science Online. Functional Consequences of Biodiversity,A.P.Kinzig, 26. R. A. Hobbs, H. A. Mooney, Ecology 72, 59 (1991). 12 July 2004; accepted 23 September 2004

dom process (14) with risk determined by Extinction and Ecosystem life-history traits such as rarity, body size, and sensitivity to environmental stressors Function in the Marine Benthos like pollution (15–18). Interspecific differ- ences in extinction risk have implications for 1 2 3 Martin Solan, * Bradley J. Cardinale, Amy L. Downing, the ensuing changes in trophic interactions Katharina A. M. Engelhardt,4 Jennifer L. Ruesink,5 and community structure (18, 19), such that Diane S. Srivastava6. the ecosystem-level consequences of random versus ordered extinctions are likely to be Rapid changes in biodiversity are occurring globally, yet the ecological im- fundamentally different (14, 20–22). pacts of diversity loss are poorly understood. Here we use data from marine Here we explore how various scenarios of invertebrate communities to parameterize models that predict how extinc- extinction for marine benthic invertebrates tions will affect sediment bioturbation, a process vital to the persistence of are likely to influence bioturbation (the bio- aquatic communities. We show that species extinction is generally expected genic mixing of sediment)—a primary deter- to reduce bioturbation, but the magnitude of reduction depends on how the minant of sediment oxygen concentrations functional traits of individual species covary with their risk of extinction. As a which, in turn, influences the biomass of or- result, the particular cause of extinction and the order in which species are ganisms, the rate of organic matter decom- lost ultimately govern the ecosystem-level consequences of biodiversity loss. position, and the regeneration of nutrients vital for primary productivity (23, 24). Marine coastal ecosystems are among the diversity should present trends in human ac- 1Oceanlab, University of Aberdeen, Main Street, most productive and diverse communities on tivity continue (6–8). Newburgh, Aberdeenshire, Scotland AB41 6AA. 2De- Earth (1) and are of global importance to cli- Given these prospects, researchers have partment of Ecology, Evolution and Marine Biology, mate, nutrient budgets, and primary produc- recently asked how the loss of biodiversity University of California, Santa Barbara, CA 93106, 3 tivity (2). Yet, the contributions that coastal might alter the functioning of marine coastal USA. Department of Zoology, Ohio Wesleyan University, Delaware, OH 43015, USA. 4University of ecosystems make to these ecological pro- ecosystems. Like most studies to date, these Maryland Center for Environmental Science, Appala- cesses are compromised by human-induced experiments have manipulated diversity by chian Laboratory, 301 Braddock Road, Frostburg, MD stresses, including overfishing, habitat de- assembling random subsets of species drawn 21532–2307, USA. 5Department of Biology, Univer- sity of Washington, Box 351800, Seattle, WA 98195, struction, and pollution (3–5). These stressors from a common pool of taxa (9–11). This ap- 6 particularly impact benthic (bottom-living) in- proach (12, 13) may be useful for understand- USA. Department of Zoology, University of British Columbia, 6270 University Boulevard, Vancouver, vertebrate communities because many species ing the theoretical consequences of diversity British Columbia, Canada V6T 1Z4. are sedentary and cannot avoid disturbance. loss but is unrealistic in the sense that it *To whom correspondence should be addressed. Thus, marine coastal ecosystems are likely to assumes species can go extinct in any order. E-mail: [email protected] experience a large proportional change in bio- Extinction, however, is generally a nonran- .All authors contributed equally to this work.

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1177 R EPORTS

Using a comprehensive study of 139 ben- potential (BPi, Equation S1) that accounts for is expected to change when species go ex- thic invertebrate species that inhabit Inner each species_ body size, abundance, mobility, tinct at random versus ordered by their sen- Galway Bay, Ireland (25), we parameterized and mode of sediment mixing. We used data sitivities to environmental stress, body size, models that predict how species extinction is from monthly samples (over 1 year) of the or population size (25). As the functional likely to affect the biogenic mixing depth benthic community to empirically derive a consequences of extinction are known to de- (BMD), an indicator of bioturbation that can relation (Equation S2) between the BMD and pend on the response of surviving species be measured from sediment profile images the bioturbation potential of the community (19, 20, 26), we simulated two different types

(Fig. 1). To estimate species contributions to (BPc). Using this relation, we performed nu- of community interactions (8). First, we used the BMD, we used an index of bioturbation merical simulations to explore how the BMD a model in which species do not interact with one another; thus, surviving species do not exhibit compensatory responses (changes in Fig. 1. The biogenic mixing population size) after extinction. This sce- depth (BMD, white arrows) of sediments [(A), site 1; (B), nario leads to complete loss of bioturbation site 2] in Inner Galway Bay, performed by an extinct species and repre- Ireland. BMD was related to the sents a Bworst-case[ scenario. Second, we bioturbation potential of a used an interactive model of community as- community (BPc), an index that sembly in which species_ abundances are accounts for each species’ limited by competition with other members population size and life-history traits (body size, mobility, of their functional guild (i.e., species with mode of bioturbation) to es- similar bioturbation modes but not necessar- timate the capacity of a com- ily similar extinction risks). This represents a munity to mix sediments (25). Bbest-case[ scenario that assumes compensa- tion is additive and substitutions of abun- dance maintain total community density Ei.e., full numerical compensation (25)^. Our models predict that loss of species diversity leads to a decline in mean BMD, regardless of extinction scenario (Fig. 2). Note, however, that Fig. 2, A to H, depict Fig. 2. Predicted changes in the BMD strikingly different patterns, suggesting that following benthic invertebrate ex- changes in the BMD depend on extinction tinctions. Each panel shows the re- scenario. Indeed, the rate of change, the sults of 20 simulations per level of species richness at which the BMD first de- species richness, constrained by a clines, the variance surrounding the relation probabilistic order of species extinc- (i.e., predictability of change), and the range tion (indicated on the right). Simu- lations (A), (B), (C), and (D) are for a of potential values all depend on how species noninteractive model of community go extinct (Table 1). These divergent patterns assembly assuming no numerical are best explained by examining the covari- compensation by surviving species. ance between each species extinction risk Simulations (E), (F), (G), and (H) are and the biological traits that influence bio- for an interactive model that as- turbation (Fig. 3). To illustrate these patterns, sumes full numerical compensation following extinction of competitors. we first focus on scenarios of extinction that involve no compensatory responses (i.e., the noninteractive model; Fig. 2, A, B, C, and D). Random extinction (Fig. 2A) produces a clear bifurcation of the BMD, with values determined by the presence (94.0 cm) versus absence (G4.0 cm) of a single species—the burrowing brittlestar, Amphiura filiformis.The strong impact of A. filiformis on bioturbation is well documented (27). In this study, A. filiformis has a disproportionate impact (Fig. 3A) on bioturbation because it is consistently one of the most abundant species in Galway Bay (Fig. 3B) and has a high per capita ef- fect that results from it being a large (Fig. 3C), highly mobile species. Consequently, changes in the BMD following extinction largely depend on whether A. filiformis is among the survivors. When extinctions are ordered by species sensitivity to stress (Fig. 2B), estimated as the relative change in the abundance of spe- cies along a gradient of disturbance (25), the risk of extinction among species varies by a

1178 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS factor of 215; yet, stress sensitivity for A. ciable loss of ecological function until a large tion and cannot offset functions performed by filiformis (j0.99, Fig. 3D) is near the me- proportion of species are lost. larger species. When species are lost in order of dian value for the community as a whole Many studies suggest that when species go rarity, even full compensation has no notable (j0.98), which explains why changes in the extinct from communities characterized by effect on the BMD because the proportional BMD are comparable to the scenario of ran- strong interactions, increases in the popula- change in bioturbation is small. Thus, compen- dom extinction (compare Fig. 2, A and B). tion size of species released from competition satory responses of surviving species do not This conclusion is confirmed by statistical can compensate for loss of ecological func- necessarily stabilize ecological processes when comparisons of the mean and range of values tion (20, 31, 32). Our models suggest that the traits required for maintaining function (minimum and maximum) of the BMD, this is only true when the risk of extinction is simultaneously increase extinction risk. which show an identical change with species not correlated with species functional traits. We have used numerical models parame- loss for both scenarios; and a comparison of This is evident because compensatory re- terized by data from a marine benthic com- the variability in BMD, which reveals only a sponses only changed the probabilistic dis- munity to show that species extinction is marginal difference between scenarios (" 0 tribution of the BMD when species were lost generally expected to reduce the depth of 0.0125; P 0 0.01, Table 1). at random (Fig. 2E) or in order of their bioturbated sediments. Such changes might For extinctions ordered by body size sensitivity to stress (compare Fig. 2, A and be expected to alter the fluxes of energy and (Fig. 2C), probabilities of extinction were E, and Fig. 2, B and F) (Table 1). However, matter that are vital to the global persistence assumed to be proportional to mean species when a species_ risk of extinction covaried of marine communities (23), a conclusion biomass to mimic the higher extinction risk with its body size or abundance, compensa- that corresponds to evolutionary patterns in generally faced by large-bodied organisms tory responses did not alter the consequences the fossil record showing a close association that often have small population sizes, have of diversity loss (compare Fig. 2, C and G between the frequency of anoxia and the longer generation times, or are found at and Fig. 2, D and H) (Table 1). This is be- diversification of marine soft-bottom com- higher trophic levels (17, 28). Body size cause when loss is ordered by body size, munities (33). We have also shown that varied by a factor of 9500,000 among species small species have little impact on bioturba- crucial details (mean, range, and predictabil- and was positively correlated with per capita effects on bioturbation (r 0 0.98, P G 0.01) but not abundance (r 0 j0.05, P 0 0.56, even Fig. 3. The relation excluding A. filiformis, r 0 j0.08, P 0 0.33). between per capita bio- turbation, BP, and mean In this scenario, larger species (high per i species abundance (A) capita effects) tended to be lost before reveals that at the pop- smaller species (low per capita effects), ulation level (diagonal leading to a faster decline in the mean dashed lines, each an BMD compared with random extinction order of magnitude dif- (compare Fig. 2, A and C; Table 1). The ference in bioturbation), range of values of the BMD (minimum and most species contribute little to bioturbation maximum) and total variation (CV) also (left of short-dashed changed with species richness more quickly line). Bioturbation is dis- than for random extinctions (Table 1). This proportionately affect- was not due to the loss of entire functional ed by one large and guilds composed of large species because highly active species, there was considerable overlap in species Amphiura filiformis (brittlestar, open cir- body size, and thus extinction risk, among cle). Population level functional guilds (25). Rather, patterns were bioturbation, BPp,is generally a consequence of the early extinc- proportional to spe- tion of A. filiformis, the 19th largest species, cies abundance (B)(r 0 0.83, P G 0.001), body which produced a step change in the BMD at 0 G a species richness of ,100. size (C)(r 0.39, P 0.001), and sensitivity Extinction risk is typically high for rare to stress (D)(r 0 j0.2, species, defined here as those with low local P G 0.05). Arrows in- abundances, because small populations are dicate order of extinctions. more vulnerable to environmental and demo- graphic stochasticity (17, 28). They also of- Table 1. Comparisons of how bioturbation changes with species loss for each extinction scenario (stress, ten have narrow geographic ranges and/or size, rarity) relative to a random model of extinction, and between the interactive and noninteractive high specialization, further compounding models of community assembly. The asterisk (*) denotes significant differences, P G 0.0125 [set extinction risk (28–30). When we assumed conservatively to correct for the number of comparisons (25)]. CV, coefficient of variation. extinction probability was inversely propor- tional to species density, rare species were Mean CV Minimum Maximum 96000 times more likely to be lost than the Comparison of random extinction to extinctions ordered byI most common species. Yet, because small 2 0 0 2 0 2 0 Sensitivity to stress c 4 0.73 F4, 1094 3.38* c 4 1.63 c 4 0.23 populations typically contribute little to bio- 2 0 0 2 0 2 0 Body size c 4 53.8* F4, 1094 42.8* c 4 15.1* c 4 15.1* turbation (Fig. 3B), extinctions of rare species 2 0 0 2 0 2 0 Rarity c 4 28.2* F4, 1094 250* c 4 97.6* c 4 3.8 had little impact on the BMD, and ecosystem Comparison of interactive to noninteractive model for extinctions that areI 2 0 0 2 0 2 0 functioning was maintained until the loss of Random c 2 35.07* F2, 274 629* c 2 30.94* c 2 10.37* Ordered by sensitivity to stress c2 0 25.76* F 0 307* c2 0 20.94* c2 0 10.19* more abundant species, such as A. filiformis 2 2, 274 2 2 Ordered by body size c2 0 7.42 F 0 166* c2 0 10.71* c2 0 5.56 (lower bifurcation, Fig. 2D). Hence, some 2 2, 274 2 2 Ordered by rarity c2 0 1.38 F 0 13.9* c2 0 0.69 c2 0 0.50 scenarios of extinction do not lead to appre- 2 2, 274 2 2

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1179 R EPORTS ity of change) of how bioturbation changes 7. D. Malakoff, Science 277, 486 (1997). 26. J. L. Ruesink, D. S. Srivastava, Oikos 93, 221 (2001). following extinction depend on the order in 8. O. E. Sala et al., Science 287, 1170 (2000). 27. M. Solan, R. Kennedy, Mar. Ecol. Prog. Ser. 228, 179 9. M. C. Emmerson, M. Solan, C. Emes, D. M. Paterson, (2002). which species are lost, because extinction D. Raffaelli, Nature 411, 73 (2001). 28. J. H. Lawton, in Population Dynamic Principles,J.H. risk is frequently correlated with life-history 10. C. L. Biles et al., J. Exp. Mar. Biol. Ecol. 285, 165 Lawton, R. M. May, Eds. (Oxford Univ. Press, Oxford, traits that determine the intensity of bio- (2003). 1995), pp. 147–163. 11. S. G. Bolam, T. F. Fernandes, M. Huxham, Ecol. 29. K. F. Davies, C. F. Margules, J. F. Lawrence, Ecology turbation. This finding is important because Monogr. 72, 599 (2002). 85, 265 (2004). it argues that the particular cause of extinc- 12. D. Raffaelli, M. Emmerson, M. Solan, C. Biles, 30. C. N. Johnson, Nature 394, 272 (1998). tion ultimately governs the ecosystem-level D. Paterson, J. Sea Res. 49, 133 (2003). 31. D. D. Doak et al., Am. Nat. 151, 264 (1998). 13. B. Schmid et al.,inBiodiversity and Ecosystem 32. J. M. Fischer, T. M. Frost, A. R. Ives, Ecol. Appl. 11, consequences of biodiversity loss. Therefore, Functioning,M.Loreau,S.Naeem,P.Inchausti,Eds. 1060 (2001). if we are to predict the ecological impacts of (Oxford Univ. Press, Oxford, 2002), pp. 61–75. 33. D. K. Jacobs, D. R. Lindberg, Proc. Natl. Acad. Sci. extinction and if we hope to protect coastal 14. D. S. Srivastava, Oikos 98, 351 (2002). U.S.A. 95, 9396 (1998). 15. C. R. Tracy, T. L. George, Am. Nat. 139, 102 (1992). 34. We thank J. E. Duffy, J. D. Fridley, A. Hector, A. R. environments from human activities that 16. S. L. Pimm, H. L. Jones, J. Diamond, Am. Nat. 132, Ives, S. Naeem, O. L. Petchey, K. J. Tilmon, D. A. disrupt the ecological functions species 757 (1988). Wardle, and J. P. Wright for comments and the perform, we will need to better understand 17. M. L. McKinney, Annu. Rev. Ecol. Syst. 28, 495 BIOMERGE Second Adaptive Synthesis Workshop for (1997). insightful discussion. Supported by BIOMERGE (Biotic why species are at risk and how this risk 18. D. Pauly, V. Christensen, J. Dalsgaard, R. Froese, Mechanisms of Ecosystem Regulation in the Global covaries with their functional traits. F. Torres Jr., Science 279, 860 (1998). Environment)—an NSF-funded research coordina- 19. J. E. Duffy, Ecol. Lett. 6, 680 (2003). tion network (to S. Naeem). References and Notes 20. A. R. Ives, B. J. Cardinale, Nature 429, 174 (2004). 1. G. C. B. Poore, G. D. F. Wilson, Nature 361, 597 (1993). 21. M. D. Smith, A. K. Knapp, Ecol. Lett. 6, 509 (2003). Supporting Online Material 2. P. G. Falkowski et al., Science 281, 200 (1998). 22. M. Jonsson, O. Dangles, B. Malmqvist, F. Gue´rold, www.sciencemag.org/cgi/content/full/306/5699/1177/ 3. P. Vitousek, H. Mooney, J. Lubchenco, J. Melillo, Proc. R. Soc. London B Biol. Sci. 269, 1047 (2002). DC1 Science 277, 494 (1997). 23. U. Witte et al., Nature 424, 763 (2003). Materials and Methods 4. R. E. Turner, N. N. Rabalais, Nature 368, 619 (1994). 24. K. S. Johnson et al., Nature 398, 697 (1999). Equations S1 and S2 5. J. B. C. Jackson et al., Science 293, 629 (2001). 25. Materials and methods are available as supporting 6. M. Jenkins, Science 302, 1175 (2003). material on Science Online. 23 July 2004; accepted 23 September 2004

case across the tropics, wild terrestrial mammals are used as a secondary source of Bushmeat Hunting, Wildlife animal protein in Ghana, and they comprise the chief commodities in a regional bush- Declines, and Fish Supply in meat trade estimated conservatively at 400,000 tons per year (8). Marine and West Africa freshwater fish are the primary source of animal protein consumed in West Africa, Justin S. Brashares,1,2* Peter Arcese,3 Moses K. Sam,4 and the fisheries sector directly and indirect- Peter B. Coppolillo,5 A. R. E. Sinclair,6 Andrew Balmford1,7 ly accounts for up to one quarter of the workforce in the region (9, 10). From 1965 to The multibillion-dollar trade in bushmeat is among the most immediate 1998, the supply of harvested fish in Ghana threats to the persistence of tropical vertebrates, but our understanding of its (Fig. 1A) ranged from 230,000 to 480,000 underlying drivers and effects on human welfare is limited by a lack of tons annually and varied by as much as 24% empirical data. We used 30 years of data from Ghana to link mammal declines between consecutive years (11). Here, we to the bushmeat trade and to spatial and temporal changes in the availability test a prediction of the protein limitation hy- of fish. We show that years of poor fish supply coincided with increased pothesis that years with low fish supply will hunting in nature reserves and sharp declines in biomass of 41 wildlife species. show larger declines in biomass of terrestrial Local market data provide evidence of a direct link between fish supply and mammals, suggesting a transfer of harvest subsequent bushmeat demand in villages and show bushmeat’s role as a pressure and consumption between these dietary staple in the region. Our results emphasize the urgent need to develop resources. We also test for evidence of a cheap protein alternatives to bushmeat and to improve fisheries management mechanism underpinning such a transfer by by foreign and domestic fleets to avert extinctions of tropical wildlife. examining (i) rates of hunting in nature reserves, (ii) sales and price data from local The trade in bushmeat for human consump- graphic scales (1, 5–7). Furthermore, contrary markets, and (iii) spatial trends in correla- tion is a key contributor to local economies to predictions of the Bprotein limitation[ tions of fish supply and wildlife declines. throughout the developing world (1, 2), but it hypothesis, unsustainable consumption of is also among the greatest threats to the wildlife remains a problem even in many 1Conservation Biology Group, Department of Zool- persistence of tropical wildlife (1–4). Efforts relatively prosperous countries (1). Identifying ogy, University of Cambridge, Cambridge CB2 3EJ, to manage the bushmeat trade are built on the bushmeat_s value as a dietary staple versus a UK. 2Department of Environmental Science, Policy premise that bushmeat consumption is driven nonessential good is vital for targeting con- and Management, University of California, Berkeley, CA 94720, USA. 3Centre for Applied Conservation by protein limitation. Thus, it is assumed that servation interventions and, equally important, Research, University of British Columbia, Vancouver, increases in livestock and agricultural produc- for predicting the impacts of wildlife declines BC V6T 1Z4, Canada. 4Ghana Wildlife Division, Accra, tion will reduce human reliance on wild on human livelihoods. Ghana. 5Wildlife Conservation Society, Bronx, NY 6 sources of food (5–7). Although it makes We evaluated the protein limitation hy- 10460, USA. Centre for Biodiversity Research, University of British Columbia, Vancouver, BC V6T intuitive and economic sense that consump- pothesis by comparing annual rates of 7 1Z4, Canada. Percy Fitz Patrick Institute of African tion of wild meat would be linked to the decline for 41 species of wild carnivores, Ornithology, University of Cape Town, Rondebosch availability of alternative sources of protein, primates, and herbivores (table S1) in six 7701, Cape Town, South Africa. there is little empirical evidence to support nature reserves in Ghana with supply of fish *To whom correspondence should be addressed. this assumption, particularly at large geo- in the region from 1970 to 1998. As is the E-mail: [email protected]

1180 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS In support of the prediction that annual increases in hunters were linked to acceler- Third, more than half of Ghana_s human standing biomass of large mammals would ated declines of wildlife. population of 20 million resides within 100 be linked positively with the annual supply Second, if annual variation in fish supply km of the coast, where the majority of of marine and freshwater fish, we found that and bushmeat hunting are linked causally, employment and dietary protein are derived changes in annual biomass of terrestrial we would expect that the availability of from fishing (10). Poor fish harvests result in mammals from 1970 to 1998 were closely bushmeat in local markets would be related reduced income and food for coastal com- related to annual per capita fish supply. negatively to the supply of fish (5). In munities and reduce the availability of fish Years with a lower-than-average supply of support of this prediction, we found that throughout the region (9, 14). The wide- fish had higher-than-average declines in monthly supply of fish in 12 local markets in spread loss of jobs and income associated mammal biomass, and vice versa (Fig. 1B) northern, central, and eastern Ghana from with poor fish harvests also may lead some (12). In contrast, fish supply and wildlife 1999 to 2003 (13) was related negatively to portion of households to rely on bushmeat declines were unrelated to other potential the volume of bushmeat sold in these mar- hunting both for income and sustenance. If explanatory factors, including annual rain- kets (Fig. 2B). In addition, the price of fish fish supply and bushmeat consumption are fall, land and water temperatures, political sold in markets was closely and negatively linked causally, it follows that the transfer of cycles, oil prices, and gross national prod- related to local fish supply (R 0 0.73, n 0 52, harvest pressure between aquatic and terres- uct (P Q 0.19 for each term in multiple- P G 0.01) and positively related to the trial resources would be most evident in regression models) (13). This correlative volume of bushmeat sold (R 0 0.48, n 0 52, support for the protein limitation hypoth- P G 0.01). The strong negative correlation esis is further supported by three additional between fish price and quantity sold, com- analyses. bined with the positive correlation between First, our working hypothesis suggests fish price and bushmeat sales, is consistent that the observed link between fish supply with the idea that variation in fish supply and wildlife decline occurs because bush- drove bushmeat sales. Comparing monthly meat hunting and consumption increased fish price in markets with the bushmeat sales when fish became scarce. In support of this in the following month yielded even stronger suggestion, we found that annual counts of correlations, again suggesting that bushmeat hunters observed by wildlife rangers in five sales were driven by fish availability and nature reserves in Ghana (13) were related price more so than the reverse case (fig. S1). negatively to per capita fish supply from These results show a substitution of wildlife 1976 to 1992 (Fig. 2A). Annual counts of for fish at the local scale. Taken together hunters were also closely related to annual with the observation of increased bushmeat rates of wildlife decline in these same nature hunting during periods of fish scarcity, these reserves (R 0 0.76, n 0 17, P G 0.01). Thus, results also support our suggestion of a hunters were more common in reserves in causal, macroscale link between fish supply years when fish supply was low, and these and wildlife declines (Fig. 1).

Fig. 1. Year-to-year change in es- timated biomass of 41 large mam- mal species was linked closely to annual harvest of marine and freshwater fish in Ghana (R 0 0.73, n 0 28 years, P G 0.001). (A) Time series plots of annual fish supply and change in esti- mated mammal biomass. (B) Conventional plot of data shown in (A). The trend line describes the equation y 0 0.0058x þ 0.81. Values of annual fish supply [from (11)] represent landings plus imports and minus exports. Biomass of large mammals was calculated for each year by mul- tiplying the number of animals observed in È700 walking counts of 10 to 15 km each (17)by species-specific body weights. The products of these calcula- Fig. 2. Links between fish supply and bushmeat tions were then summed across hunting and consumption are evident in ob- all species. servations that (A) annual counts of hunters in five terrestrial reserves in Ghana from 1976 to 1992 were related negatively to supply of fish in the region (R 0 –0.52, n 0 17, P 0 0.03); (B) monthly sales of bushmeat in 12 rural markets in Ghana were related negatively to local fish supply (R 0 –0.61, n 0 52, P G 0.01); and (C) fish supply and wildlife declines were related most closely in reserves occurring nearest to the coast (R 0 0.81, n 0 6, P 0 0.05).

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1181 R EPORTS coastal areas where reliance on fish for both mammal abundance and marine and fresh- foreign fleets to fish off West Africa (18–24). income and animal protein is greatest. We water fish stocks documented in the region over Declines of fish stocks in nearshore and tested this last prediction by repeating the the past 30 years now suggest that this buffer offshore waters of West Africa have coin- analysis in Fig. 1 separately for each of six system can no longer be sustained (14, 17–20). cided with more than 10-fold increases in nature reserves in Ghana. We found the From 1970 to 1998, the biomass of 41 regional fish harvests by foreign and domes- strongest link between annual variation in species of mammals in nature reserves in tic fleets since 1950 (11). The European marine and freshwater fish supply and annual Ghana declined by 76% (Fig. 3), and 16 to Union (EU) has consistently had the largest change in mammal biomass in reserves near 45% of these species became locally extinct foreign presence off West Africa, with EU the coast and weaker, though still significant, (17). Similarly, trawl surveys conducted in fish harvests there increasing by a factor of linkages for reserves farther inland (Fig. 2C). the Gulf of Guinea since 1977 and other 20 from 1950 to 2001 (fig. S3). Furthermore, These three lines of evidence indicate regional stock assessments estimate that fish EU financial support of its foreign fleet that fish supply is linked negatively to the biomass in nearshore and offshore waters has increased from about $6 million in 1981 to price of fish, the number of wildlife hunters, declined by at least 50% (Fig. 3). At the more than $350 million in 2001 (fig. S3), and the sales and supply of bushmeat in local same time, a threefold increase in human with the effect of artificially increasing the markets. Our results also show that the sub- populations in the region since 1970 has profitability of fishing in African waters for stitution of fish for bushmeat occurs most resulted in per capita declines in fish supply, EU boats, despite declining fish stocks (22). intensively close to the coast, where fish are despite steady increases in regional fish West African commercial fleets also have more important as sources of food and harvests (11, 14). These sharp declines in expanded considerably since 1950 (fig. S3) income. All of these findings are consistent terrestrial wildlife and marine fish suggest and there is no guarantee that reductions of with the protein limitation hypothesis and that stocks in this region may face imminent foreign catches would not be taken up by inconsistent with the notion that bushmeat in collapse (9, 18). The consequences of col- increased domestic fishing. However, even Ghana is primarily a nonessential good lapse of either fish or terrestrial wildlife are short-term increases in the domestic supply (summarized in fig. S2). daunting and may be felt immediately as of fish both for commercial export and local Our results provide clear evidence to widespread human poverty and food inse- consumption may enhance regional econo- suggest that the outcomes of programs aimed curity in the region (14). Reduced fish stocks mies (14) and ease exploitation of terrestrial at promoting economic development, food have already severely damaged the region_s wildlife resources. Over the longer term, security, and the conservation of biological artisanal fisheries sector (14, 21), and recent intensive management to enhance fish stocks diversity in Ghana, and perhaps elsewhere in collapses of mammal populations in some and stabilize harvests must become a region- Africa, will be closely linked. First, the close areas of West Africa have been linked to geo- al conservation and economic priority. correlation between hunting pressure, mar- graphic patterns of poverty and malnourish- A second route to increase the sustainabil- kets, and long-term trends in wildlife abun- ment (8, 17). Agricultural production is a ity of fish and wildlife harvests could come by dance suggests strongly that the persistence third potentially critical, though poorly un- enhancing the protection of harvested marine of the more than 400 species of terrestrial derstood, factor linking human food supply to and terrestrial resources. Pirate fishing ves- vertebrates that supply the bushmeat trade in biodiversity conservation in the region (16). sels from foreign ports are abundant in West West Africa will depend ultimately on the One management response to the poten- African waters and illegally extract fish of the availability of affordable alternative protein tial collapse of fish and terrestrial wildlife highest commercial value while, like many sources for the region_s growing human stocks in West Africa is to build up region- commercial fleets, dumping 70 to 90% of population. Second, our failure to conserve al livestock and agriculture sufficiently to their haul as by-catch (9, 18). Increased existing wildlife populations as core sources alleviate pressure on overexploited wild policing of exclusive fishing zones and for managed, sustainable harvests could have resources (7). However, such efforts could enforcement of existing quotas and tariffs serious deleterious effects on the stability of take decades to implement and face enor- for commercial fleets should reduce exploi- the long-term human food supply and the mous economic, regulatory, and political tation and provide an immediate boost to livelihoods of bushmeat hunters and sellers. hurdles. Thus, more immediate plans to marine resources available to local fisheries Our findings and those of others suggest that enhance the sustainability of wild protein (14, 19). On land, wildlife has persisted at the harvest of terrestrial wildlife can buffer sources are required. One immediate route near historic levels in inaccessible and well- the impact of environmental or other shocks to increasing production and sustainability of protected areas of West Africa_snature by providing animal protein and income in domestic fisheries, and thereby reducing reserves (4, 17). Increasing the size, number, times of economic hardship or food scarcity pressure on terrestrial wildlife, is to limit and protection of wildlife reserves in the (2, 15, 16). However, marked declines in large the access of large and heavily subsidized region may not offer a long-term solution to concerns over human livelihoods and protein supply, but it is likely to offer the most Fig. 3. Estimates of marine fish immediate prospects for slowing the region_s biomass in the Gulf of Guinea catastrophic wildlife decline. (gray circles) and large mammal biomass in Ghana (black circles). References and Notes Estimates of fish biomass are 1. J. G. Robinson, E. L. Bennett, Hunting for Sustain- from trawl surveys (24, 25). Ana- ability in Tropical Forests (Columbia Univ. Press, New lyses of fisheries catch data with York, 2000). ecosystem models indicate that 2. E. J. Milner-Gulland et al., Trends Ecol. Evol. 18, 351 fish biomass in coastal West and (2003). Northwest Africa has declined by 3. J. G. Robinson, K. H. Redford, E. L. Bennett, Science a factor of 13 since 1960 (20). 284, 595 (1999). Estimates of mammal biomass 4. World Conservation Union, International Union for Conservation of Nature and Natural Resources are based on abundances of 41 (IUCN) Red List of Threatened Animals (IUCN, Gland, species observed in È700 wildlife Switzerland, 2000). counts per year in six nature 5. D. S. Wilkie, R. A. Godoy, Science 287, 975 (2000). reserves (17)(seemap,fig.S4). 6. E. L. Bennett, Conserv. Biol. 16, 588 (2002).

1182 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS

7. Organisation for Economic Co-operation and Devel- 14. J. Atta-Mills, J. Alder, U. R. Sumaila, Nat. Resour. 25. Ghan Marine Fisheries Research Division (MFRD), opment (OECD), Shaping the 21st Century: The Forum 28, 13 (2004). Oceangraphic Data Center, Marine database;avilable Contribution of Development Cooperation (OECD, 15. C. B. Barrett, P. Arcese, Land Econ. 74, 449 (1998). at www.ioc.unesco.org/odinafrica/contnets.php?0217. Paris, 1996). 16. G. J. S. Dei, Ecol. Food Nutr. 22, 225 (1989). 26. We thank the Ghana Wildlife Division for permis- 8. Y. Ntiamoa-Baidu, Wildlife Development Plan: 1998– 17. J. S. Brashares, P. Arcese, M. K. Sam, Proc. R. Soc. sion to work in reserves and access to data, and we 2003 (Wildlife Department, Accra, Ghana, 1998). Lond. Ser. B. 268, 2473 (2001). thank J. Atta-Mills for discussion. C. Kresge, P. Kresge, 9. United Nations Environment Programme (UNEP), Africa 18. Food and Agriculture Organization (FAO), The State J. Mason, B. Volta, N. Ankudey, D. Boateng, L. Lanto, Environment Outlook; available at www.unep.org/aeo. of World Fisheries and Aquaculture 2002; available at and G. Agbango provided assistance in Ghana, and 10. Food and Agriculture Organization (FAO), Country www.fao.org/docrep/005/y7300e/y7300e00.htm. C. Barrett, J. Hellmann, D. Pauly, I. Watson, J. Smith, Profiles: Ghana; available at www.fao.org/country- 19. D. Pauly et al., Nature 418, 689 (2002). V. Christensen, E. J. Milner-Gulland and four profiles/index.asp?iso30GHA. 20. V. Christensen et al.,inPeˆcheries Maritimes, Ecosys- anonymous reviewers gave many helpful sugges- 11. Food and Agriculture Organization (FAO), Fisheries Data- te`mes et Socie´te´s: Un Demi-Sie`cledeChangement,B.A. tions. Supported by NSF INT-0301935 (J.S.B.). bases; available at www.fao.org/fi/statist/statist.asp. Moctar, P. Chavance, D. Gascuel, M. Vakily, D. Pauly, 12. Statistics are based on a linear regression of annual Eds. (Institut de Recherche pour le Developpement, Supporting Online Material change in mammal biomass [calculated as (kg )/ Paris, 2004). www.sciencemag.org/cgi/content/full/306/5699/1180/ t þ 1 DC1 kgt] against per capita fish harvest. Regressing per 21. World Wide Fund for Nature (WWF), West Africa Puts capita change in mammal biomass [i.e., (kg EU To Shame (WWF European Policy Office, Brussels, Materials and Methods t þ 1– Figs. S1 to S5 kgt)/human population t þ 1] against per capita fish 2001). catch gave a similar result (adjusted R2 0 0.52, P G 22. V. M. Kaczynski, D. L. Fluharty, Mar. Policy 26, 75 (2002). Tables S1 and S2 0.001). 23. S. L. Pimm et al., Science 293, 2207 (2001). References 13. Materials and methods are available as supporting 24. Sea Around Us Project (SAUP), Web Products: Marine material on Science Online. Database; available at www.seaaroundus.org. 7 July 2004; accepted 7 October 2004

behaved like wild type when kept under non- The Genetic Basis of Singlet permissive light-dark conditions (7) Egroup III (fig. S1, B to D)^. Allelism tests and mapping Oxygen–Induced Stress Responses revealed that they were allelic, representing a single locus that was named EXECUTER1.In of Arabidopsis thaliana contrast to wild-type plants but like flu,the executer1/flu double mutant accumulated free Daniela Wagner,1*. Dominika Przybyla,1* Roel op den Camp,1 protochlorophyllide (Pchlide) in the dark (Fig. Chanhong Kim,1 Frank Landgraf,1 Keun Pyo Lee,1 Marco Wu¨rsch,1 1, A to C, and fig. S1B). After transfer to the Christophe Laloi,1 Mena Nater,1 Eva Hideg,2 Klaus Apel1- light, executer1/flu generated singlet oxygen in amounts similar to those of flu (Fig.1,Fto Plants under oxidative stress suffer from damages that have been interpreted H) but grew like wild type when kept under as unavoidable consequences of injuries inflicted upon plants by toxic levels nonpermissive light-dark cycles (Fig. 1, A to of reactive oxygen species (ROS). However, this paradigm needs to be C). The second stress reaction of flu to the modified. Inactivation of a single gene, EXECUTER1, is sufficient to abrogate release of singlet oxygen is an inhibition of stress responses of Arabidopsis thaliana caused by the release of singlet growth. In flu plants, the growth rate was oxygen: External conditions under which these stress responses are observed reduced immediately after the beginning of and the amounts of ROS that accumulate in plants exposed to these envi- reillumination (Fig. 1D). The executer1/flu ronmental conditions do not directly cause damages. Instead, seedling le- plants, however, grew like wild-type plants thality and growth inhibition of mature plants result from genetic programs (Fig. 1D). Growth inhibition of flu plants was that are activated after the release of singlet oxygen has been perceived by particularly striking when plants were trans- the plant. ferred to repeated light-dark cycles, whereas executer1/flu continued to grow like wild-type Abiotic stress conditions limit the ability of compartment, and which also triggers a plants (Fig. 1E). All three plant lines grew plants to use light energy for photosynthesis, visible stress response that is easy to score. equally well under continuous light (fig. S2). often reducing their growth and productivity Recently, we have isolated the condition- As a first step toward the functional char- and causing photooxidative damages (1–3). al flu mutant of Arabidopsis thaliana that acterization of EXECUTER1, we used a The emergence of these stress symptoms has fulfills these requirements (6). The mutant map-based cloning strategy to isolate the been closely associated with the enhanced generates singlet oxygen in plastids in a con- EXECUTER1 gene. EXECUTER1 was ge- production of several ROS (4, 5). Because trolled and noninvasive manner. Immediate- netically mapped on chromosome IV on a different ROS are generated simultaneously, it ly after the release of singlet oxygen, mature genomic fragment of about 90 kb (Fig. 2A). is difficult to determine the biological activity flu plants stop growing, whereas seedlings A contig consisting of 11 cosmid clones that and mode of action for each of these ROS bleach and die (6). Here, we demonstrate that encompassed this chromosomal region was separately. In order to address this problem, the two stress responses, growth inhibition and generated (Fig. 2A), and the ability to com- one would need to find conditions under seedling lethality, do not result from physico- plement the executer1 mutation was tested which only one specific ROS is generated at chemical damage caused by singlet oxygen (7). Seedlings of the double mutant trans- a given time, within a well-defined subcellular during oxidative stress but are caused by the formed with the genomic DNA of the cosmid activation of a genetically determined stress clone 44 that contained a wild-type copy of 1Institute of Plant Sciences, Plant Genetics, Swiss Fed- response program. EXECUTER1 died like flu seedlings when eral Institute of Technology (ETH), CH-8092 Zurich, We set out to identify such a genetic pro- grown under nonpermissive dark-light con- 2 Switzerland. Institute of Plant Biology, Biological Re- gram by identifying second-site mutations ditions, whereas seedlings of plants trans- search Center, Hungarian Academy of Sciences, H-6701 that abrogate either one or both of the two formed with genomic DNA of other cosmid Szeged, Hungary. stress responses of the flu mutant. Three clones grew like seedlings of the original *These authors contributed equally to this work. different groups of second-site mutations executer1/flu parental line (Fig. 2B). .Deceased 24 February 2004. -To whom correspondence should be addressed. E- could be distinguished (7) (fig. S1A). One of The second test was done with mature T2 mail: [email protected] these groups contained 15 mutants that plants transformed with DNA of cosmid

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1183 R EPORTS clone 44. The original executer1/flu mutant dark light/dark

plants were not visibly affected by a shift to ] A D light-dark cycles and continued to grow like m 15 m

[ ex1/flu wild-type plants, but executer1/flu plants h wt t

complemented with the cosmid 44 DNA w 10 o r

were indistinguishable from the flu mutant g light

n flu

in that they ceased growth and their leaves o

i dark

s 5

developed necrotic lesions (Fig. 2C). Both n e t

assays demonstrate that the rapid bleaching x of flu seedlings and the inhibition of growth E 0 B after the release of singlet oxygen are not 0 400 800 1200 1600 Time [min] because of the toxicity of this ROS and do not reflect photooxidative damage and inju- ry, but instead result from the activation of genetically controlled responses that require the activity of the EXECUTER1 gene. E ex1/flu The complementing genomic fragment of ] 300 flu m wt m

cosmid clone 44 has a size of about 14,000 [ 200

C h base pairs (bp) and contains three open reading t w

frames (ORFs). One of them, At4g33630, was o r 100 identified as EXECUTER1 (7) (fig. S3A). G The ORF of the EXECUTER1 cDNA predicts 0 0102030 a protein of 684 amino acids with a molecular Time [days] mass of 76,534.9 daltons. It is unrelated to known proteins, but its N-terminal part re- sembles import signal sequences of nuclear-

encoded plastid proteins (fig. S3B). This ] e 0 c

= FGH t

prediction was confirmed experimentally n

e o t c (fig. S4, A to C). In all higher plants for wt 100 100 ex1/flu s 100 flu d e e r z

which expressed sequence tag sequence data o i l u

l 80

a 80 80 were available, including major crop plants, f m y r P EXECUTER1 homologs could be found o e n 60 60 60

n ,

(fig. S3C). Thus, EXECUTER1 represents a %

D [ 0246 0246 0246 a highly conserved plastid protein that seems to Time of illumination [min] enable higher plants to perceive singlet oxygen as a stress signal that activates a genetically Fig. 1. A comparison of singlet oxygen production, cell death, and growth inhibition in executer1/ determined stress response program. flu, flu, and wild type (WT). Etiolated seedlings of flu (B) and executer1/flu (A) overaccumulated similar amounts of free Pchlide, as indicated by the bright red fluorescence, in contrast to etiolated The physiological role of the EXECUTER1 WT seedlings (C). Cell death of flu seedlings (B) and growth inhibition of mature flu plants (0) gene in wild-type plants was assessed by first grown under nonpermissive 16 hours light–8 hours dark conditions were blocked by the executer1 isolating executer1 mutant plants that no mutation (h)(D and E). Results from WT control plants (r) are also shown in (D) and (E). (F to H) longer carried copies of the mutated flu gene Generation of singlet oxygen. WT (F), flu (G), and executer1/flu (H) were grown under continuous (7). Wild-type and executer1 plants were light until they were ready to bolt. At this stage, plants were either transferred to the dark for then exposed to higher light intensities in 8 hours or kept under light. Cut leaves were infiltrated with dansyl-2,2,5,5-tetramethyl-2,5- dehydro-1H-pyrrole and subsequently illuminated with white light (100 6mol photons mj2 sj1). the presence of 3-(3, 4-dichlorphenyl)-1,1- Singlet oxygen trapping was measured as relative quenching of DanePy fluorescence (6). In WT dimethylurea (DCMU), which is known to controls, no difference in singlet oxygen production between plants exposed to continuous light stimulate the release of singlet oxygen (8). and a dark-light shift could be found (H), whereas in leaves of flu and executer1/flu generation of The effect of the executer1 mutation on singlet oxygen was enhanced in plants that had been kept in the dark before reillumination (&) singlet oxygen–mediated cell death was but not in plants exposed to continuous light ()). ex1, executer1. assessed by floating cut leaves on solutions with increasing concentrations of DCMU, 3B). In the presence of 25 nM and 250 nM In most of the second-site mutants of flu rangingfrom25nMto256M, first in the DCMU, ion leakage in leaves of wild type that have been identified during our suppres- dark for 30 min and then under light (950 increased strongly to 40 and 55%, respec- sor screen, only one of the two major singlet 6mol photons mj2 sj1) for 24 hours. tively, whereas in leaves of the executer1 oxygen–mediated stress reactions was abol- Inhibition of photosystem II (PSII) by mutant ion leakage remained at a low level, ished, i.e., either growth inhibition or seed- DCMU was similar in wild types and similar to that in the water control (Fig. 3B). ling lethality (fig. S1). In the executer1/flu executer1 mutants, as indicated by the Only when the concentration of DCMU was mutant line, however, both visible stress concurrent decline of the maximum effi- further increased to 25 6M were leaves of responses were abrogated. One could argue ciency of PSII at increasing DCMU con- both wild type and executer1 almost com- that a mutated form of the EXECUTER1 centrations in both leaf samples (Fig. 3A). pletely damaged, with ion leakage being protein may act as a scavenger of singlet Cell death and membrane damage in cut close to 100% after 24 hours of illumination oxygen and in this way may eliminate de- leaves were estimated by measuring ion (Fig. 3B). However, even under these harsh- trimental toxic effects of this ROS. However, leakage. When leaves of wild type and er conditions, the increase of cell damage in such a case, singlet oxygen should no executer1 were tested in the absence of over time in DCMU-treated leaves of wild longer be detectable in the executer1/flu mu- DCMU, ion leakage reached about 10% of type occurred much faster than the damage tant line. Furthermore, suppression of stress the maximum after boiling the leaves (Fig. in leaves of executer1 (9). responses in flu by the executer1 mutation

1184 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS

Fig. 2. Identification of the EXECUTER1 gene. (A) Genetic and physical marked by stars, contained genomic DNA fragments that largely over- map of the DNA region on chromosome IV of A. thaliana that contains lapped. The DNA fragment of cosmid 44 restored the cell death of the EXECUTER1 gene (arrow). The region between markers F17M5 and seedlings (B) and the growth inhibition of mature plants of the parental T16L1 is encompassed by 11 cosmid clones that were used to com- flu line (C) when grown under nonpermissive light-dark cycles. Similar plement the executer1-7/flu double mutant. Cosmid clones 44 and 76, results were obtained with cosmid 76 (15).

A B lower DCMU concentrations, ion leakage and 1.0 100 membrane damage seem to result from the 0.8 WT 80 wt activation of the EXECUTER1-dependent

m 0.6 ex1 60 ex1 cell-death program. At higher DCMU con- F / v 0.4 40 centrations, this genetically controlled cell- F 0.2 20 death reaction is gradually masked by an EXECUTER1-independent cell-death reac- 0.0 0

% of electrolyte leakage tion that seems primarily caused by the 0 0.025 0.25 2.5 25 0 0.025 0.25 2.5 25 toxicity of elevated levels of singlet oxygen. Concentration of DCMU [µM] Concentration of DCMU [µM] In the past, it was not possible to distinguish between these two cell death reactions, Fig. 3. Suppression of cell death in wild-type plants by the inactivation of EXECUTER1.(A)The effects of different concentrations of DCMU on the maximum efficiency of PSII in WT and because the genetically determined part executer1-48 plants grown for 3 weeks under 16 hours light–8 hours dark conditions. (B) remained unnoticed. With the identification Selective suppression of cell death by executer1 in DCMU-treated leaves kept at 950 6mol of the executer1 mutation, it is now possible photons mj2 sj1. The progression of cell death was determined by measuring the electrolyte to define conditions under which the geneti- leakage of cut leaves at different lengths of illumination. The ion leakage of leaves boiled for 25 cally controlled cell death prevails. min was taken as 100%. At concentrations of up to 0.25 6M DCMU, cell death was suppressed in executer1, whereas in WT plants it steadily increased with increasing DCMU concentrations. In References and Notes leaves of the flu mutant kept in the dark for 8 hours before the experiment, the extent of ion 1. J. Barber, B. Andersson, Trends Biochem. Sci. 17,61 leakage after 24 hours of reillumination at 100 6mol photons mj2 sj1 reached about 50% (9). (1992). 2. K. Apel, H. Hirt, Annu. Rev. Plant Biol. 55, 373 (2004). 3. K. K. Niyogi, Annu. Rev. Plant Physiol. Plant Mol. Biol. 50, 333 (1999). was not confined to allelic lines that syn- Carotenoids of light-harvesting complexes 4. M. J. Fryer, K. Oxborough, P. M. Mullineaux, N. R. thesized modified EXECUTER1 proteins effectively quench triplet Chl and singlet Baker, J. Exp. Bot. 53, 1249 (2002). $ 5. E. Hideg et al., Plant Cell Physiol. 43, 1154 (2002). with single amino acid exchanges that could oxygen (11), but the -carotenes bound to 6. R. G. op den Camp et al., Plant Cell 15, 2320 (2003). turn the mutated protein into a scavenger of PSII reaction center fail to do so because they 7. Materials and methods are available as supporting singlet oxygen but occurred also in executer1 are localized too far away from the P680 Chl material on Science Online. 1 8. C. Fufezan, A. W. Rutherford, A. Krieger-Liszkay, FEBS allelic lines that were no longer able to syn- (12). Thus, continuous O2 production seems Lett. 532, 407 (2002). thesize this protein. Lastly, a mutation that to be an inherent property of PSII even under 9. D. Przybyla, unpublished results. confers an increased scavenging capacity to low-light conditions. The quenching of this 10. B. A. Diner, F. Rappoport, Annu. Rev. Plant Biol. 53, 551 (2002). the executer1/flu mutant should be transmit- singlet oxygen has been linked to the turnover 11. R. Cogdell, H. A. Frank, Photochem. Photobiol. 63, ted as a dominant trait, whereas all executer1 of the D1 protein that is oxidized by singlet 257 (1996). alleles represent recessive mutations. oxygen and apparently serves as a scavenger 12. N. Kamiya, J.-R. Shen, Proc. Natl. Acad. Sci. U.S.A. Inactivation of the EXECUTER1 protein of this ROS (13, 14). Excess amounts of 100, 98 (2003). 13. J. Sharma, M. Panico, J. Barber, H. R. Morris, J. Biol. did not only suppress the induction of death in singlet oxygen that cannot be quenched by the Chem. 272, 3935 (1997). flu seedlings grown under light-dark cycles D1 protein and that interact with other targets 14. A. Trebst, Z. Naturforsch. C58, 609 (2003). but also in wild-type plants that were treated within the vicinity of PSII may be the trigger 15. D. Wagner, unpublished results. 16. We are indebted to T. Fitzpatrick for critical reading, with DCMU. DCMU is known to stimulate that initiates singlet oxygen–mediated stress D. Rubli for art work, and M. Geier-Ba¨chtold for the release of singlet oxygen in chloroplasts. It responses in wild-type plants (14). So far, editorial work. This study was supported by the inhibits PSII and mimics photoinhibition by these stress responses have been attributed to Swiss Federal Institute of Technology and the Swiss NSF. binding to the secondary quinone electron the toxicity of this ROS (1, 2). However, as acceptor of PSII, Q , and inhibiting forward shown in our present work, the intensity and Supporting Online Material B www.sciencemag.org/cgi/content/full/306/5699/1183/ electron transport. Charge recombination in quality of these responses to light stress may DC1 PSII favors the formation of a chlorophyll range from necrotic reactions resulting from Materials and Methods (Chl) triplet state that reacts with ground-state severe photooxidative damage to the activa- Figs. S1 to S4 1 triplet oxygen to form O2 (10). tion of a genetically controlled cell death. At 23 July 2004; accepted 21 September 2004

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1185 R EPORTS epithelial fields comparable to those induced by the intact symbiont (Fig. 2, D to F). Cell Microbial Factor-Mediated surface fractions from the nonsymbiotic, marine, Gram-negative bacterium Pseu- Development in a doalteromonas luteoviolacea were also ac- tive (9). Host-Bacterial Mutualism V. fischeri PGN signaled levels of hemocyte infiltration comparable to those Tanya A. Koropatnick,1 Jacquelyn T. Engle,2 Michael A. Apicella,3 initiated by the intact symbiont, whereas V. Eric V. Stabb,4 William E. Goldman,2 Margaret J. McFall-Ngai1,5* fischeri LPS did not induce this cellular reaction (Fig. 2G). Alone, PGN did not Tracheal cytotoxin (TCT), a fragment of the bacterial surface molecule pep- trigger apoptosis, although LPS did stimu- tidoglycan (PGN), is the factor responsible for the extensive tissue damage late it at low levels; however, together PGN characteristic of whooping cough and gonorrhea infections. Here, we report and LPS acted synergistically to induce that Vibrio fischeri also releases TCT, which acts in synergy with lipopolysac- apoptosis at levels characteristic of those charide (LPS) to trigger tissue development in its mutualistic symbiosis with resulting from colonization by the intact the squid Euprymna scolopes. As components of PGN and LPS have commonly symbiont (Fig. 2H). PGN, both alone and in been linked with pathogenesis in animals, these findings demonstrate that synergy with LPS, also induced significant host interpretation of these bacterial signal molecules is context dependent. levels of epithelial regression (Fig. 2I). These Therefore, such differences in interpretation can lead to either inflammation findings are similar to those reported for and disease or to the establishment of a mutually beneficial animal-microbe certain pathogenic associations in which association. PGN and PGN fragments can induce macrophage activation (10) and infiltration To date, molecules conserved among mi- the colonization process through ciliary mo- (11) into inflamed host tissues. Likewise, crobes, such as LPS and PGN, have been col- tion and mucus shedding, resulting in the purified PGN and LPS work in synergy to lectively described as Bpathogen[-associated aggregation of symbiont cells above pores molecular patterns (PAMPs) (1). However, on the surface of the organ. Once aggre- the majority of animal-microbe interactions gated, the symbionts migrate through the are benign or mutualistic, raising the ques- pores, down ducts, and into crypts that are tion: What role might such factors play in located 100 to 200 6m from the surface other types of host-microbe associations? epithelia (Fig. 1). Upon colonization of the The reciprocal dialogue between part- crypts, some of the first processes trig- ners in benign host-symbiont associations gered by the symbiont include the infil- has been shown to be important for host tration of macrophage-like hemocytes tissue maturation (2–4), although the identi- (blood cells) into the sinuses of the ciliated fication of the bacterial signals involved has fields (Fig. 2A) and the induction of proven elusive. The symbiosis between the widespread apoptosis of the epithelial cells Hawaiian bobtail squid E. scolopes and the that compose these fields (5) (Fig. 2B). The luminous, Gram-negative bacterium V. most conspicuous response to light organ ci fischeri offers the opportunity to decipher colonization is the extensive morphogenesis d experimentally the precise dialogue between of the organ_s surface, which culminates host and microbe partners. In this system, the in the complete loss of the ciliated field 4 cr bacterium colonizes epithelium-lined crypts days after the initial colonization by the aa _ is within the host s light-emitting organ as a symbiont (Fig. 2C). In nature, only V. pa monospecific, extracellular symbiont (Fig. 1) fischeri is capable of colonizing the crypts (5). and signaling this morphogenesis, which Shortly after the juvenile squid emerges serves to transform the organ from a morphol- Fig. 1. E. scolopes possesses a light-emitting from the egg, the bacterial inoculum is ogy associated with the colonization process to organ. (Top) A ventral view of a juvenile squid gathered from the environment as seawater one characteristic of the mature, functional shows a window through the mantle to il- passes through the mantle cavity (6). Two organ. PGN in ambient seawater is known to lustrate the position of the organ within the mantle cavity. Scale bar, 500 6m. (Bottom)An prominent fields of ciliated epithelia on the trigger the shedding of mucus from the enlargement shows the details of the juvenile surface of the squid_s light organ facilitate epithelial fields of the light organ (7), and light organ morphology. The left half illustrates LPS can induce low levels of early-stage the surface with its ciliated epithelial field, apoptosis (8). However, the symbiont-derived which is composed of an anterior (aa) and pos- 1Pacific Biomedical Research Center, Kewalo Marine factor(s) capable of triggering the full se- terior (pa) appendage and a base with three Laboratory, University of Hawaii, 41 Ahui Street, quence of light organ morphogenesis have yet pores (circled) that lead to internal epithelium- Honolulu, HI 96813, USA. 2Department of Molecular lined crypts. The right half is a frontal section Microbiology, Washington University School of Medi- to be identified. through the organ showing these epithelium- cine, St. Louis, MO 63110, USA. 3Department of Micro- V. fischeri continuously sheds fragments lined crypts (cr) containing the bacteria (stip- biology, University of Iowa, Iowa City, IA 52242, USA. of its surface in culture, and the crypt pled) within crypt diverticula and the relation of 4Department of Microbiology, University of Georgia, the crypts and ducts to the pores (circled), 5 epithelial cells that interface closely with Athens, GA 30602, USA. Department of Medical the symbiont cross-react with a monoclonal which open onto the surface. This section also Microbiology and Immunology, University of Wis- antibody to V. fischeri LPS (9). Cell surface shows the sinuses (arrows) within the append- consin, 1300 University Avenue, Madison, WI 53706, ages. The sinus spaces are continuous with the USA. fractions isolated from V. fischeri were circulatory system and separate from the *To whom correspondence should be addressed. sufficient to induce levels of hemocyte in- bacteria-containing crypts. ci, cilia; d, ducts; is, E-mail: [email protected] filtration, apoptosis, and regression of the ink sac. Scale bar, 150 6m.

1186 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS trigger inflammatory cytokine release, the passive release due to cell lysis. Specific alanine) (Fig. 3A), a disaccharide-tetrapeptide nitric oxide production, and organ injury release of peptidoglycan by growing cells has monomer of PGN, which causes the epithe- in a rat model of bacterial sepsis (12). only been observed in cultures of Bordetella lial cytopathology of pertussis (15) and gono- Because mixed fragments of PGN with pertussis (13)andNeisseria gonorrhoeae (14), coccal infections (16). Fractionation of culture LPS could not consistently induce regression both of which release large amounts of pep- supernatants of V. fischeri by reversed-phase to the extent elicited by the intact symbiont tidoglycan monomers. The best studied of HPLC revealed a peak with an elution time (Fig. 2I), we reasoned that V. fischeri might these fractions is tracheal cytotoxin (TCT; identical to that of TCT. When purified, this have an active mechanism to release specif- N-acetylglucosaminyl-1,6-anhydro-N-acetyl- fraction contained three amino acids: ala- ic PGN fragments to the host, as opposed to muramylalanyl-,-glutamyldiaminopimelyl- nine, glutamic acid, and diaminopimelic acid,

Fig. 2. Bacterial components ABHemocyte infiltration Apoptosis C Epithelial regression induce light organ morpho- genesis. (A and B) Confocal aa s micrographs of nonsymbi- aa pa aa otic [non-sym, i.e., uninfected s (25)] and symbiotic (sym) or- gan epithelial fields stained with acridine orange (green) and Lysotracker (red). (A) He- pa mocytes (arrowheads) within the appendage sinuses (s). (B) non-sym symnon-sym sym stage:0 stage:4 Apoptotic cells, yellow foci (arrows). (C) SEMs of epithe- D E F lial fields before (stage 0) 20 * 40 * * 4 and after (stage 4) regres- * * * sion. Scale bar, 50 6m. (D to 15 30 3 F) Effects of V. fischeri cell surface fractions (csf) on 10 20 2 hemocytes infiltration (D),

Hemocytes † induction of apoptosis (E), 5 † 10 1 and epithelial regression (F). Apoptotic cells † Stage of regression Animals were exposed to 0 0 0 non- csf sym non- csf sym non- csf sym surface fractions at a protein sym sym sym concentration of 100 6g/ml of seawater. (G to I) Effects G H I of V. fischeri surface compo- 30 * * 50 4 nents, LPS (10 6g/ml) and * * 40 * PGN (50 6g/ml), on hemo- 3 † 20 * cyte infiltration (G), induc- 30 † * tion of apoptosis (H), and † 2 * epithelial regression (I). Data † 20 † 10 † are means T SEM for one of Hemocytes † 1 † † three replicate experiments Apoptotic cells 10 Stage of regression (n 0 8 to 12 per treatment). 0 0 0 G non- lps pgn lps+ sym non- lps pgnlps+ sym non- lps pgnlps+ sym (*) indicates significant (P sym pgn sym pgn sym pgn 0.001) difference compared with non-sym. (.) indicates significant (P G 0.001) difference compared with sym.

Fig. 3. TCT, a mono- A B 1.0 500 CH OH CH O mer of PGN, is re- 2 2 0.9 400 leased by growing OO 0.8 300 cells of V. fischeri. (A) OH O Structure of TCT. V. HO 0.7 200 fischeri strain ES114 NH C CH NH C CH 3 3 was cultured at room O O 0.6 temperature in de- 600 fined medium HM O O CH H H O CH 3 3 OD 0.5 100

(26) supplemented CH CH C N CH C N CH CH CH C N CHC N CH C OH 80 TCT (nM) with 2% glucose. A 3 2 2 H CHO CH H released fragment of O O 2 O 0.4 60 CH V. fischeri PGN was O 2 CH determined to be 2 40 NH CH identical to TCT by 2 0.3 subjecting culture su- CHO 17 18 19 20 21 pernatants to solid O Time (h) phase extraction and two reverse-phase HPLC steps, as described for B. pertussis (13). A peak phase extraction (13), and derivatized with phenylisothiocyanate. The with an elution time corresponding to B. pertussis TCT was collected and resulting phenylthiocarbamyl (PTC) derivatives were separated by further characterized by amino acid analysis and by matrix-assisted laser reversed-phase high-performance liquid chromatography using a C8 desorption mass spectrometry (25). (B) Production of TCT during a 4 hour column and detected at 254 nm. The amount of V. fischeri PTC-TCT in period of log-phase growth. For sensitive quantification of TCT pro- each sample was determined by comparing the peak area and elution 0 duction, broth cultures were inoculated at OD600 0.04, after which log- time with an identically processed TCT standard. Results are phase V. fischeri culture supernatants were collected, subjected to solid representative of two experiments.

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1187 R EPORTS in molar proportions of 2:1:1. Mass spec- LPS in the natural seawater, which typically (24), demonstrate that BPAMPs[ may be trometry revealed a single species with a ranges from 0.01 to 1.5 ng/ml (22), was too narrow an acronym for eukaryotic- mass of 921 daltons. These data correspond sufficient to act together with the more prokaryotic signaling associated with mi- precisely to TCT (17). Unlike most Gram- potent TCT to induce the apoptotic effect crobial molecules, such as LPS and PGN. negative bacteria, V. fischeri, like B. pertussis (Fig. 4B). However, as cells of V. fischeri The data suggest that a more general term, (18), actively released appreciable amounts naturally colonize the crypts in high such as microbe associated molecular pat- of TCT during log-phase growth (Fig. 3B). numbers and subsequently trigger mor- terns (MAMPs), would be more appropriate TCT alone triggered hemocyte infil- phogenesis from within these spaces (19), to describe factors conserved and essential to tration and regression of the epithelial fields it is likely to be symbiont-specific LPS the biology of microbes, which mediate re- at levels similar to those induced by intact which works in concert with TCT to induce cognition and response during host-microbe V. fischeri (Fig. 4, A and C). TCT induced morphogenesis. interactions. epithelial regression 4 days after exposures as These results show that growing cells brief as 14 hours (9), a time course similar of V. fischeri release TCT, which acts as a References and Notes to the intact symbiosis (19). The concen- potent morphogen to induce normal light 1. R. Medzhitov, C. A. Janeway Jr., Science 296, 298 (2002). tration of TCT required to trigger a detect- organ morphogenesis in the squid host. Sim- 2. L. V. Hooper, Trends Microbiol. 12, 129 (2004). able morphogenic response was as little as ilar morphogenic effects of TCT have been 3. J. Xu, J. I. Gordon, Proc. Natl. Acad. Sci. U.S.A. 100, 10452 (2003). 10 nM, and the response was saturated at reported for both B. pertussis and N. 4. M. J. McFall-Ngai, Dev. Biol. 242, 1 (2002). concentrations as low as 1 6M(0.96g/ml) gonorrhoeae, which induce the loss of 5. S. V. Nyholm, M. J. McFall-Ngai, Nature Rev. Micro- (9). Comparatively, PGN, a multimer of ciliated cells from mammalian respiratory biol., 2, 632 (2004). 6. S. V. Nyholm, E. V. Stabb, E. G. Ruby, M. J. McFall- TCT subunits, was generally not active be- and fallopian tube epithelia, respectively Ngai, Proc. Natl. Acad. Sci. U.S.A. 97, 10231 (2000). low 50 6g/ml, and constant exposure was (15, 16). In addition, TCT works in syn- 7.S.V.Nyholm,B.Deplancke,H.R.Gaskins,M.A.Apicella, necessary to trigger epithelial regression by ergy with LPS to stimulate the production M. J. McFall-Ngai, Appl. Environ. Microbiol. 68, 5113 4 days. To determine how specific the mor- of inflammatory cytokines, nitric oxide, (2002). 8. J. S. Foster, M. A. Apicella, M. J. McFall-Ngai, Dev. Biol. phogenic responses were to TCT, we also and the inhibition of DNA synthesis in 226, 242 (2000). tested muramyl dipeptide and glucosyl- hamster tracheal epithelium (23). In these 9. M. J. McFall-Ngai et al., unpublished observations. muramyl dipeptide, which are two smaller studies, TCT-induced epithelial morpho- 10. M. J. Pabst, S. Beranova-Giorgianni, J. M. Krueger, Neuroimmunomodulation 6, 261 (1999). components of PGN known to signal host genesis resulted from a direct interaction 11. Z. Q. Liu, G. M. Deng, S. Foster, A. Tarkowski, Arthritis cell responses in various mammalian model with the target epithelium. Note that, in Res. 3, 375 (2001). systems (20, 21). No morphogenic activity the squid light organ, morphogenesis is 12. G. M. Wray, S. J. Foster, C. J. Hinds, C. Thiemermann, Shock 15, 135 (2001). was detected in squid light organs when triggered from within the crypt spaces, sev- 13. B. T. Cookson, H. L. Cho, L. A. Herwaldt, W. E. Goldman, exposed to these components, either alone eral cell layers away from the target epi- Infect. Immun. 57, 2223 (1989). or in combination with LPS (9). thelium (19). Thus, it will be interesting 14. R. S. Rosenthal, Infect. Immun. 24, 869 (1979). 15. W. E. Goldman, D. G. Klapper, J. B. Baseman, Infect. Under the conditions of the assay, TCT to identify and localize the TCT receptor(s) Immun. 36, 782 (1982). also triggered levels of apoptosis in the within the light organ and to decipher the 16. M. A. Melly, Z. A. McGee, R. S. Rosenthal, J. Infect. Dis. epithelial fields similar to those observed host pathways that mediate these remote 149, 378 (1984). 17. B. T. Cookson, A. N. Tyler, W. E. Goldman, Biochemistry in the intact symbiosis (Fig. 4B), whereas events. 28, 1744 (1989). LPS-free PGN did not induce apoptosis These findings, as well as recent studies 18.R.S.Rosenthal,W.Nogami,B.T.Cookson,W.E. (Fig. 2H). Thus, the background level of of mechanisms of tolerance to gut microbiota Goldman, W. J. Folkening, Infect. Immun. 55,2117 (1987). 19. J. A. Doino, M. J. McFall-Ngai, Biol. Bull. 189, 347 (1995). A * B * 20. N. Inohara et al., J. Biol. Chem. 278, 5509 (2003). 20 * 60 21. S. Traub et al., J. Biol. Chem. 279, 8694 (2004). * * 22. D. M. Karl, F. C. Dobbs, in Molecular Approaches to the Study of the Ocean, K. E. Cooksey, Ed. (Chapman 15 † and Hall, London, 1998), pp. 29–89. 40 23. T. A. Flak, L. N. Heiss, J. T. Engle, W. E. Goldman, †† * Infect. Immun. 68, 1235 (2000). 10 24. S. Rakoff-Nahoum, J. Paglino, F. Eslami-Varzaneh, S. Edberg, R. Medzhitov, Cell 118, 229 (2004).

Hemocytes 25. Materials and methods are available on Science Online.

Apoptotic cells 20 5 † † 26. E. G. Ruby, K. H. Nealson, Appl. Environ. Microbiol. 34, 164 (1977). 27. We are grateful to C. Chun, W. Crookes, M. Goodson, 0 0 J. Graber, D. Millikan, E. Ruby, and A. Schaefer for non- non- tcttct sym non- non- tct tct sym critical reading of this manuscript; M. McMahon for sym sym +ab sym sym +ab technical assistance; and C. Yap for Fig. 1 illustra- +ab +ab tions. We also thank M. Hadfield for the provision C 4 of P. luteoviolacea. Supported by an NIH grant to Fig. 4. TCT induces light organ morphogenesis. (A and M.M.-N., M.A.A., and E.V.S. (grant no. R01-AI50661), B) Animals were exposed to TCT (10 6M) alone or TCT an NSF grant to M.M.-N. (grant no. IBN0211673), an 3 † * and a monoclonal antibody to LPS (ab, 1:500) to bind NIH grant to E. G. Ruby and M.M.-N. (grant no. * and sequester exogenous LPS and scored for hemocyte NCRR12294), a W. M. Keck Foundation grant to * M.M.-N. and M.A.A., and a Canadian NSERC infiltration (A) and induction of apoptosis (B). (C) scholarship to T.A.K. 2 Animals exposed to TCT were scored for epithelial regression. Data are means T SEM for one of three Supporting Online Material replicate experiments (n 0 8 to 12 per treatment). www.sciencemag.org/cgi/content/full/306/5699/1186/ 1 † Stage of regression sym, symbiotic. (*) indicates significant (P G 0.001) DC1 difference compared with nonsymbiotic (non-sym). Materials and Methods 0 Bracket indicates significant (P G 0.001) difference References non- 1 10 sym . sym between indicated treatments. ( ) indicates significant µM tct (P G 0.001) difference compared with sym. 1 July 2004; accepted 21 September 2004

1188 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS recombinant Eppin could influence the titer of antibodies to Eppin in their semen. These Reversible Immunocontraception results are similar to a previous study in female macaques demonstrating that dif- in Male Monkeys ferent serum titers resulted in correspond- ingly different antibody titers in oviductal Immunized with Eppin fluid (11). Consequently, in the fertility study described below, two male monkeys in the M. G. O’Rand,1,2* E. E. Widgren,1,2 P. Sivashanmugam,1,2. initial immune group were dropped because R. T. Richardson,1,2 S. H. Hall,1,3 F. S. French,1,3 C. A. VandeVoort,4 they could not sustain a high serum titer and S. G. Ramachandra,5 V. Ramesh,5 A. Jagannadha Rao5 were unlikely to have a high semen titer. The lack of a strong immune response to an im- Various forms of birth control have been developed for women; however, munogen in a particular individual animal is there are currently few options for men. The development of male contra- a reflection of the major histocompatibility ceptives that are effective, safe, and reversible is desired for family planning complex and T cell response (12) as well as _ throughout the world. We now report contraception of male nonhuman pri- the antigen s availability to regulate the im- mates (Macaca radiata) immunized with Eppin, a testis/epididymis-specific mune response (13). Such responses are protein. Seven out of nine males (78%) developed high titers to Eppin, and all found in heterozygous populations and of these high-titer monkeys were infertile. Five out of seven (71%) high–anti- would need further study before proceeding Eppin titer males recovered fertility when immunization was stopped. This with additional Eppin fertility trials, which study demonstrates that effective and reversible male immunocontraception might include a linear B cell epitope (fig. S7). is an attainable goal. This method of immunocontraception may be extended We tested the effect of Eppin immuniza- to humans. tion on male fertility at the Indian Institute of Science. Six adult male monkeys (M. radiata) Although several different choices and ap- online material text (10). The studies on were immunized with human recombinant proaches are available for contraception in normal Macaca monkeys at UC-Davis Eppin and six controls received adjuvant only. women, the choices for men are currently allowed us to determine that IgG is present High–anti-Eppin titers were detected in four limited to condoms and vasectomy (1, 2). in the normal epididymal tract. Moreover, of the six monkeys immunized with Eppin in Male hormonal contraceptives (3, 4) devel- immunization of two monkeys at UC-Davis squalene; two monkeys were low responders oped over the past several years have now indicated that the immune response to with titers G1:400 on postimmunization day advanced to clinical trials, and the outcome of these studies may determine whether the Table 1. Fertility test of Eppin-immunized male monkeys (M. radiata). The seven males were 7 to 12 suppression of sperm production through years old and had each sired 1 to 3 offspring in the previous 1 to 2 years. No pregnancies resulted from immunized males. Each exposed female had an estradiol 17$ surge, indicating that an ovulation probably androgen regulation can become a realistic occurred in that cycle. ELISA O.D., enzyme-linked immunosorbent assay O.D. at 1:1000 on nearest day of product. Immunocontraception, an alterna- cohabitation. tive nonhormonal method, has been studied for many years (1, 5), with the major em- Days after Male monkey no. Cycle no. Female monkey no. ELISA O.D. phasis on immunization of females to pre- first immunization vent pregnancy (6) or fertilization (7). In the present study, we report the successful Group 1 Eppin (squalene) 602 contraception of male nonhuman primates I 55 0.39 476 (M. radiata) immunized with Eppin, a testis/ II 88 1.19 522 epididymis-specific protein (8, 9). This rep- III 107 1.19 528 resents a non–hormonally disruptive male 619 immunocontraceptive for primates. I 84 0.67 475 Before using the monkey as a model for II 64 0.56 523 III 23 0.56 530 the test of a male immunocontraceptive, we 625 determined the presence of both Eppin and I 60 1.48 397 immunoglobulin (IgG) in the male reproduc- II 97 1.00 525 tive tract, the immunogenicity of Eppin, and III 89 1.15 568 the effects of immunization on sperm mo- 679 tility at the University of California, Davis I 19 0.70 475 II 47 0.62 534 (UC-Davis). Results from these studies are III 82 0.62 542 shown in figs. S1 to S3 and the supporting Group 2 Eppin (CFA) 610 1Laboratories for Reproductive Biology, 2Department I 29 0.91 147 3 of Cell and Developmental Biology, and Department II 73 0.59 225 of Pediatrics, University of North Carolina at Chapel 4 III 97 1.19 250 Hill, Chapel Hill, NC 27599, USA. California National 656 Primate Research Center, University of California, 5 I 107 1.86 147 Davis, CA 95616, USA. Department of Biochemistry, II 47 1.93 225 Primate Research Laboratory, Indian Institute of III 55 2.14 250 Science, Bangalore 560012, India. 657 *To whom correspondence should be addressed. I 79 1.27 147 E-mail: [email protected] II 6 1.53 276 . Present address: Department of Urology, Duke III 82 1.00 323 University, Durham, NC 27710, USA.

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1189 R EPORTS Table 2. Fertility test of male monkeys (M. After the completion of fertility testing, Eppin interaction with the sperm surface and radiata) in adjuvant control group. The six males immunizations of monkeys stopped on day with semenogelin. were 10 to 11 years old and had each sired 1 to 3 691 (day 448 of immunization for group 2). offspring in the previous 1 to 3 years. Male 609 Groups 1 and 2 were maintained without References and Notes impregnated two different females. Four out of 1. S. J. Nass, J. F. Strauss, Eds., New Frontiers in Contra- six males (67%) impregnated females. further immunizations for 450 days (group 1; ceptive Research (National Academies Press, Washing- Eppin/squalene) and 451 days (group 2; Eppin/ ton, DC, 2004). CFA), respectively, to test their ability to re- 2. C. Holden, Science 296, 2172 (2002). Male Female Days after first 3. A. Kamischke, E. Nieschlag, Trends Pharmacol. Sci. 25, monkey monkey immunization cover fertility after immunization. During this 49 (2004). no. no. when conceived recovery time period, three of four monkeys 4. J. K. Amory, W. J. Bremner, Trends Endocrinol. Metab. in the Eppin/squalene group and two of three 11, 61 (2000). 569 37, 107, 47 No pregnancy monkeys in the Eppin/CFA group recovered 5. M.G.O’Rand,I.A.Lea,J. Reprod. Immunol. 36, 51 (1997). 604 35 650 6. G. P. Talwar et al., Proc. Natl. Acad. Sci. U.S.A. 91, 607 87 650 their fertility for a total recovery of 71% (5 8532 (1994). 609 86; 6 502; 594 out of 7; table S3). The males exhibited no 7. P. Primakoff, W. Lathrop, L. Woolman, A. Cowan, 688 96 502 symptoms of autoimmune disease and had D. Myles, Nature 335, 543 (1988). 8. R. T. Richardson et al., Gene 270, 93 (2001). 690 6, 84, 11 No pregnancy no detectable serum titer of antibody to 9. P. Sivashanmugam et al., Gene 312, 125 (2003). Eppin at 1:1000 dilutions. 10. Materials and methods are available as supporting This study demonstrates that effective and material on Science Online. 11. I. A. Lea, B. Kurth, M. G. O’Rand, Biol. Reprod. 58, 126. Because the purpose of this study was reversible male immunocontraception in pri- 794 (1998). to test the efficacy of antibodies to Eppin mates is an attainable goal. We found that 12. I. A. Lea et al., Biol. Reprod. 59, 527 (1998). on fertility, the two low-titer monkeys were a high serum titer (91:1000), sustained over 13. R. M. Zinkernagel, H. Hengartner, Science 293, 251 (2001). dropped from the study group and their fer- several months, achieves an effective level 14. R. T. Richardson, E. Widgren, Z. Wang, P. Sivashanmugam, tility was not tested. Three additional males of contraception. Seven out of nine males M. G. O’Rand, Biol. Reprod. Suppl. 70, 98 (abstr.) (2004). were added, which were immunized (pri- (78%) developed high titers to Eppin, and 15. Supported by grant CIG-96-06 from the Consortium for Industrial Collaboration in Contraceptive Research mary immunization) with recombinant human all these high-titer monkeys were infertile. Program of Contraception Research and Development Eppin in complete Freund_s adjuvant (CFA) Five out of seven (71%) high–anti-Eppin (CONRAD). to boost immunogenicity. The original four titer males recovered fertility when immu- Supporting Online Material monkeys immunized with Eppin in squalene nization was stopped. www.sciencemag.org/cgi/content/full/306/5699/1189/ were designated group 1, and the three Eppin on the surface of spermatozoa and DC1 Materials and Methods additional males immunized with Eppin in in semen is bound to semenogelin (14), SOM Text CFA were designated group 2. which is involved in coagulum formation in Figs. S1 to S7 Antibody titers in all of the monkeys were the ejaculate. We can speculate that one Tables S1 to S3 References 91:10,000 at the time the matings started and mechanism to explain the infertility is that remained elevated throughout the mating pe- antibodies to Eppin interfere with normal 29 April 2004; accepted 17 September 2004 riod. Figure S4, A (group 1) and B (group 2), shows the mean optical density (O.D.) value at 450 nm for the monkeys at a 1:1000 dilution A Cluster of Metabolic Defects of serum. A titer of 91:1000 was sustained for 775 days in group 1 (fig. S4A) and for 481 days in group 2 (fig. S4B). There was no effect Caused by Mutation in a on serum testosterone levels in the immunized malesineithergroup1orgroup2compared Mitochondrial tRNA with control values (fig. S5) and no effect on 1,2,3 1,4 1,2 sperm counts in either group (fig. S6). Frederick H. Wilson, * Ali Hariri, * Anita Farhi, 2,5 4 1,2 Each male monkey in the immune and Hongyu Zhao, Kitt Falk Petersen, Hakan R. Toka, 1,2 8 control groups was subjected to fertility test- Carol Nelson-Williams, Khalid M. Raja, ing by cohabiting with a proven fertile fe- Michael Kashgarian,6 Gerald I. Shulman,1,4,7 male between days 9 and 14 of her menstrual Steven J. Scheinman,8 Richard P. Lifton1,2,3,4. cycle. Each male was exposed to three ovu- latory cycles of three different females to test Hypertension and dyslipidemia are risk factors for atherosclerosis and occur their fertility. Immune and control groups together more often than expected by chance. Although this clustering sug- began fertility testing on days 390 to 397 gests shared causation, unifying factors remain unknown. We describe a large (Table 1 and table S2; for group 2, day 390 kindred with a syndrome including hypertension, hypercholesterolemia, and is 147 days after their first day of immuni- hypomagnesemia. Each phenotype is transmitted on the maternal lineage with zation). Group 1 completed testing on day a pattern indicating mitochondrial inheritance. Analysis of the mitochondrial 568, group 2 completed testing on day 566 genome of the maternal lineage identified a homoplasmic mutation sub- (323 days after their first day of immuniza- stituting cytidine for uridine immediately 5¶ to the mitochondrial transfer tion), and the control group completed RNAIle anticodon. Uridine at this position is nearly invariate among transfer testing on day 691 (table S2). None of the RNAs because of its role in stabilizing the anticodon loop. Given the known immunized monkeys was able to impregnate loss of mitochondrial function with aging, these findings may have im- females, indicating that males with sustained plications for the common clustering of these metabolic disorders. high–anti-Eppin titers were infertile (Table 1). Four monkeys in the adjuvant control group Hypertension and dyslipidemia are important These traits are concordant in individual pa- impregnated 5 females (4 out of 6, 67%, risk factors for many common cardiovascu- tients more often than expected by chance Table 2). All the monkeys used for breeding lar diseases, including myocardial infarction, (3, 4). Large epidemiologic studies have dem- exhibited ovulatory cycles (table S1). stroke, and congestive heart failure (1, 2). onstrated that subjects with hypertension

1190 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS have a marked increase in the prevalence magnesemic subjects being on the maternal again predominantly among hypomagne- of hypercholesterolemia, hypertriglyceridemia, lineage by chance is extremely small (22 0 semic subjects (P 0 4 10–6) despite normal hypomagnesemia, diabetes, insulin resistance, 49, P G 10–11), strongly supporting mito- serum calcium levels. Hypomagnesemia and obesity (5–9). Various combinations of chondrial transmission. Autosomal domi- with reduced urinary calcium is characteris- these abnormalities affect up to a quarter of nant transmission with imprinting was much tic of a primary defect in the renal distal the U.S. adult population and are referred to less likely from the observed distribution convoluted tubule (DCT) (16). In addition, as the metabolic syndrome, syndrome X, or (the odds favoring mitochondrial transmis- hypokalemia due to inappropriate renal loss dyslipidemic hypertension. The factors ac- sion were 9106:1) (18). A genome-wide anal- was seen more frequently on the maternal counting for this phenotypic clustering are ysis of linkage was performed and found lineage (Fig. 2D) (22 0 11.6, P 0 0.0007), unknown, although obesity, insulin resist- no evidence for a shared segment of the nu- predominantly among hypomagnesemic sub- ance, and increased local glucocorticoid tone clear genome among hypomagnesemic sub- jects. There was no difference in 24-hour have been suggested to play a role (4, 10). jects (18). urinary sodium excretion between maternal Although rare mutations with large effects Quantitative serum Mg2þ levels were low- and nonmaternal lineages. Electrolyte values on blood pressure (11), lipids (12), insulin er in individuals from the maternal lineage are summarized in table S1. resistance (13, 14), obesity (15), and magne- compared with relatives in the nonmaternal Hypertension also segregated with the ma- sium (16) have established critical pathways lineage (Fig. 2A) (P 0 2 10–9). Members ternal lineage. Thirty of 53 adults on the for homeostasis of each of these traits, they of the maternal lineage had a marked in- maternal lineage had blood pressure greater have typically affected only one of these crease in the urinary fractional excretion of than 140/90 mm Hg or were being treated phenotypes and therefore have not provided Mg2þ (Fig. 2B) (P 0 0.0001); this effect with antihypertensive medication versus 8 of an explanation for their clustering (17). was most pronounced among subjects with 53 on the nonmaternal lineages (22 0 19.9, A Caucasian kindred (K129) was ascer- hypomagnesemia (Fig. 2B) (P 0 5 10–6 P G 0.00001). The prevalence of hyper- tained through a proband with hypomag- comparing hypomagnesemic subjects ver- tension on the maternal lineage showed a nesemia. Evaluation of her extended kindred sus subjects in the nonmaternal lineage), es- marked age dependence, increasing from 5% revealed a high prevalence of hypomag- tablishing impaired renal Mg2þ reabsorption in subjects under age 30 (1 of 20 subjects), to nesemia, hypertension, and hypercholester- as the cause of hypomagnesemia in K129. 44% in those from age 30 to 50 (10 of 23 olemia. We ultimately performed a detailed Evaluation of other urinary electrolytes was subjects), and to 95% in those over age 50 clinical evaluation of 142 blood relatives in notable for reduced urinary calcium on the (19 of 20 subjects). Because the oldest gen- the kindred (Fig. 1). Including the index case, maternal lineage (Fig. 2C) (P 0 0.0005), erations of K129 are enriched for the ma- 38 members had hypertension (with blood pressure 9 140/90 mm Hg or on treatment for hypertension), 33 had hypercholesterol- emia (with total cholesterol 9 200 mg/dl or on treatment for hypercholesterolemia), and 32 had clinically significant hypomagne- semia (range 0.8 to 1.7 mg/dl, normal 1.8 to 2.5 mg/dl). Because it is the least common of these traits in the general population, we initially focused on the distribution of hypomag- nesemia in the kindred. Hypomagnesemic individuals are distributed through four gen- erations and 16 sibships, and both genders are affected (Fig. 1); there was no significant effect of age on Mg2þ levels and no hypo- magnesemic subjects were taking Mg2þ- altering medications. All 32 members with hypomagnesemia are on the same maternal lineage (Figs. 1 and 2A). Affected fathers never transmitted the trait to their offspring (0 of 17 offspring), whereas affected mothers transmitted the trait to a high fraction of their offspring (16 of 21). These features are hallmarks of inheritance via the mitochon- drial genome. The probability of all 32 hypo-

1Howard Hughes Medical Institute, 2Department of Genetics, 3Department of Molecular Biophysics and Biochemistry, 4Department of Internal Medicine, 5Department of Biostatistics, 6Department of Pathol- ogy, 7Department of Cell and Molecular Physiology, Yale University School of Medicine, New Haven, CT 06510, USA. 8Department of Medicine, State Uni- Fig. 1. The structure of Kindred 129. Individuals with serum Mg2þ G 1.8 mg/dl are indicated by versity of New York Upstate Medical University, black symbols. Family members taking antihypertensive medications or having blood pressures Syracuse, NY 13210, USA. over 140/90 mm Hg are indicated by an H. Members with hypercholesterolemia (serum *These authors contributed equally to this manuscript. cholesterol 9 200 mg/dl or taking lipid-lowering agents) are denoted by C. Blood relatives who .To whom correspondence should be addressed. did not have electrolyte values measured are indicated by gray symbols. The index case is E-mail: [email protected] indicated by an arrow.

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1191 R EPORTS ternal lineage (Fig. 1), we reanalyzed the mitochondrial deletion, we performed direct previously undescribed thymidine-to-cytidine data, excluding subjects over age 60; the re- sequencing and single-strand conformational transition at nucleotide 4291, which lies with- sults remain highly significant (P 0 0.0005) polymorphism analysis of the entire mito- in the mitochondrial tRNAIle gene (GenBank (supporting online text). The prevalence of chondrial genome to search for sequence var- accession no. NC_001807) (Fig. 4, A and B). hypertension on the maternal lineage is also iants. Fourteen variants were identified on This mutation is found only on the maternal high compared with the general population the maternal lineage; 13 are previously iden- lineage in K129, does not appear among the (supporting online text). tified polymorphisms of no known conse- thousands of mitochondrial genomes pre- Quantitative assessment confirmed the ef- quence (table S3). One variant, however, is a viously sequenced (19), and was absent fect of maternal lineage on blood pressure (Fig. 3, A and B, and Table 1). After ad- justment of blood pressure for the major co- variates age, sex, and body mass index (BMI), adults on the maternal lineage had highly significant increases in systolic and diastolic blood pressures compared with their nonmaternal relatives. Among adults age 18 to 60, maternal lineage increased systolic blood pressure by an average of 13 mm Hg (P 0 0.00007) and diastolic blood pressure by 5 mm Hg (P 0 0.002). Similar results are seen in analysis of all adults. Estimates of these quantitative effects are conservative because of the higher use of antihypertensive medication among members of the maternal lineage. Plasma renin and aldosterone levels were no different between members of the maternal and nonmaternal lineages (table S2). Hypercholesterolemia also segregated with the maternal lineage. Twenty-four of 46 adults on the maternal lineage had fasting total cholesterol of 9200 mg/dl or were being treated with cholesterol-lowering medication versus 9 of 49 on the nonmaternal lineage (22 0 12.4, P 0 0.0004). The relationship remained highly significant (P 0 0.0008) when the analysis was restricted to adults age 18 to 60. Similar results were obtained for elevated fasting low-density lipoprotein (LDL) cholesterol (LDL 9 130 mg/dl; 22 0 11.6, P 0 0.0007). Quantitative analysis of total cholesterol among adults age 18 to 60 after adjustment for age, sex, and BMI Fig. 2. Renal hypomagnesemia, hypocalciuria, and hypokalemia in the maternal lineage of K129. revealed that maternal lineage increased total (A) Serum Mg2þ values for individuals in maternal and nonmaternal lineages of K129 are shown cholesterol by an average of 26 mg/dl (Fig. and are significantly different (P 0 2 10–9). (B) Fractional renal Mg2þ excretion (FEMg2þ) on the 3C and Table 1). This increase is attributable maternal and nonmaternal lineages is shown; on the maternal lineage, individuals with normal and to elevations in LDL and very low-density low Mg2þ levels are separated. Hypomagnesemic subjects have significantly elevated fractional ex- 2þ 0 0 lipoprotein (VLDL), with no effect on cretion of Mg , indicating a renal defect (P 0.0001 comparing maternal to nonmaternal; P 5 10–6 comparing hypomagnesemic subjects versus those not in the maternal lineage). (C)Urinary fasting high-density lipoprotein (HDL) or calcium to creatinine ratios (UCa/cr) are shown grouped as in (B); maternal subjects have significantly triglycerides (Fig. 3D, Table 1, and fig. S1). reduced urinary calcium levels (P 0 0.0005). (D)SerumKþ levels. Hypokalemia is seen predominantly Similar results are seen among all adult on the maternal lineage among hypomagnesemic subjects. subjects. The magnitude of these effects is likely an underestimate because of the Table 1. Age, sex, and BMI-adjusted traits in adults age 18 to 60 in maternal and nonmaternal lineages increased use of cholesterol-lowering agents of K129. Values are mean T SEM. Total cholesterol, LDL, VLDL, HDL, triglyceride, glucose, and insulin among maternal relatives. sensitivity (18) were measured after an overnight fast. SBP, systolic blood pressure; DBP, diastolic blood In sum, of 45 adults on the maternal lineage pressure; HOMA, homeostasis model assessment. who had all three traits measured, 38 had one or more of hypertension, hypercholes- Nonmaternal Maternal P terolemia, or hypomagnesemia, 26 had two or SBP (mm Hg) 122 T 2 135 T 3 0.00007 more, and 7 had all three (fig. S2). The DBP (mm Hg) 77 T 182T 1 0.002 maternal lineage accounts for virtually all of Total cholesterol (mg/dl) 173 T 4 199 T 7 0.002 the clustering of these traits in K129 (fig. S2). LDL þ VLDL (mg/dl) 124 T 5 150 T 8 0.004 Collectively, these data provide strong HDL (mg/dl) 50 T 249T 3 0.78 T T evidence for a mitochondrial mutation as the Triglyceride (mg/dl) 129 14 148 22 0.46 Glucose (mg/dl) 84 T 195T 11 0.28 cause of the syndrome in K129. Because HOMA 3.5 T 0.3 4.0 T 0.3 0.22 Southern blotting revealed no evidence of

1192 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS among 170 unrelated control individuals. Poly- exceptions are eukaryotic initiator tRNAMet types commonly associated with mitochondri- merase chain reaction–restriction fragment genes (20). The extreme conservation of uri- al dysfunction. The prevalence of migraine length polymorphism analysis revealed that dine at this position is explained by the headache, sensorineural hearing loss, and this mutation is apparently homoplasmic in structure of tRNAs. The anticodon loop re- hypertrophic cardiomyopathy were increased leukocytes of all members of the maternal sults from a sharp turn in the phosphodiester on the maternal lineage (supporting online lineage regardless of phenotype, with the as- backbone, allowing presentation of the anti- text). Measures of fasting HDL, triglycerides, say sufficiently sensitive to detect 1% hetero- codon to its cognate mRNA codon in the insulin resistance, BMI, and diabetes mellitus plasmy (fig. S3) (18). ribosome (21, 22). This turn is stabilized by were not significantly different between the The thymidine-to-cytidine mutation in a hydrogen bond between the amino group two lineages (Table 1 and figs. S1 and S4). K129 occurs immediately 5¶ to the tRNAIle of the conserved uridine and the phosphate Immunohistochemistry of a skeletal mus- anticodon (Fig. 4C). Uridine at this position backbone of the third base of the anticodon cle biopsy from a member of the maternal is one of the most extraordinarily conserved (22, 23). Cytidine lacks this amino group and lineage revealed an increase in ragged red bases in the biological world. It is conserved cannot form this hydrogen bond. Biochemi- fibers and subsarcolemmal succinate dehy- in every sequenced isoleucine tRNA, includ- cal studies with anticodon stem-loop analogs drogenase staining, characteristic features of ing 242 different species of archaebacteria, of tRNAs have been performed and indicate individuals carrying mitochondrial mutations eubacteria, unicellular and multicellular eu- that substitution of cytidine for uridine at this (fig. S5, A and B) (24). Electron microscopy karyotes, animals, plants, chloroplasts, and position markedly impairs ribosome binding of the biopsy demonstrated cytoplasmic lip- mitochondria (20). Moreover, uridine is con- (23), providing evidence of the functional id accumulation, increased glycogen stores, served at this position in virtually all se- importance of this mutation. and dysmorphic mitochondrial cristae, further quenced tRNAs of all specificities (96% of Members of K129 were carefully evaluated signs of mitochondrial dysfunction (fig. S5C). 4300 tRNAs among all species); nearly all for the presence of additional clinical pheno- Finally, in vivo nuclear magnetic resonance

Fig. 4. Mitochondrial tRNAIle mutation in K129. Mitochondrial DNA from both blood leuko- cytes and renal epithelial cells was analyzed and yielded identical results. (A) A fragment of mtDNA containing the tRNAIle gene was am- plified from members of K129 and normal con- trols and was fractionated by nondenaturing gel electrophoresis (18). A thymidine-to-cytidine var- iant (indicated by arrow) is present in individuals from the maternal lineage (M) but absent in offspring of affected males (paternal lineage, P) and unrelated controls. (B) The sequence of a Fig. 3. Quantitative blood pressure and cholesterol values in K129. Values in maternal and portion of the mitochondrial tRNAIle gene from nonmaternal K129 members between the ages of 18 and 60 are shown. All values represent the amplicon in (A) from a wild-type control difference from the mean value after adjustment for age, sex, and BMI. For blood pressure and (left) and a member of the maternal lineage of lipids, units are mm Hg and mg/dl, respectively. Mean and SEM values are indicated for maternal K129 (right). A single base substitution (asterisk) and nonmaternal groups. Values are significantly elevated in members of the maternal lineage. (A) changes the wild-type thymidine to cytidine. (C) Systolic blood pressure (P 0 0.00007). (B) Diastolic blood pressure (P 0 0.002). (C) Fasting total The T Y C transition alters the nucleotide im- cholesterol (P 0 0.002). (D) Fasting LDL þ VLDL cholesterol (P 0 0.004). mediately 5¶ to the tRNAIle anticodon.

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1193 R EPORTS (NMR) spectroscopy of skeletal muscle in (26, 28) could contribute to the characteristic as the few patients reported have both insulin re- this patient demonstrated normal tricarboxylic age-related increase in blood pressure (34) sistance and hypertension. 18. Materials and methods are available as supporting acid cycle flux but reduced adenosine tri- and to its clustering with hypocholesterol- material on Science Online. phosphate (ATP) production, suggesting im- emia in the general population. The mutation 19. MITOMAP: A Human Mitochondrial Genome Data- paired coupling of these processes (fig. S5, D in K129 results in a complex pattern of phe- base, available at www.mitomap.org. 20. M. Sprinzl, C. Horn, M. Brown, A. Ioudovitch, S. Steinberg, and E). Additional studies of other kindred notypic clustering that is reminiscent of the Nucleic Acids Res. 26, 148 (1998). members will be required to establish the fre- frequent but not obligatory clustering seen 21. S. H. Kim et al., Science 179, 285 (1973). quency and severity of these manifestations. in the general population. This highlights 22. G. J. Quigley, A. Rich, Science 194, 796 (1976). 23. S. S. Ashraf et al., RNA 5, 188 (1999). These findings establish a causal relation- the complexity that can arise from a single 24. D. C. Wallace, Science 283, 1482 (1999). ship between a mitochondrial mutation and mutation because of the combined effects of 25. P. Maechler, C. B. Wollheim, Nature 414, 807 (2001). hypertension, hypercholesterolemia, and hy- reduced penetrance and pleiotropy and un- 26. K. F. Petersen et al., Science 300, 1140 (2003). 27. V. K. Mootha et al., Nature Genet. 34, 267 (2003). pomagnesemia. The mitochondrial origin of derscores the value of studying very large 28. A. Trifunovic et al., Nature 429, 417 (2004). this disorder is of particular interest given kindreds. The present findings motivate fur- 29. V. Carelli, C. Giordano, G. d’Amati, Trends Genet. 19, recent evidence implicating mitochondrial ther investigation of a potential role for mito- 257 (2003). 30. R. F. Reilly, D. H. Ellison, Physiol. Rev. 80, 277 (2000). dysfunction in type 2 diabetes mellitus and chondrial dysfunction in common forms of 31. D. B. Simon et al., Nature Genet. 12, 24 (1996). insulin resistance, other components of the hypertension and hypercholesterolemia. 32. R. A. Zager, A. C. Johnson, S. Y. Hanson, Am. J. Physiol. metabolic syndrome. Rare mitochondrial mu- Renal Physiol. 285, F1092 (2003). tations cause diabetes with deafness (25). In 33. A. Atlante et al., Int. J. Mol. Med. 1, 709 (1998). References and Notes 34. R. S. Vasan et al., JAMA 287, 1003 (2002). vivo NMR of skeletal muscle has linked 1. J. Stamler, D. Wentworth, J. D. Neaton, JAMA 256, 35. We thank the members of K129 for their generous loss of mitochondrial function to insulin re- 2823 (1986). participation in this project; I. Beerman, C. Mendenhall, sistance (26). Finally, expression of genes 2. A. Mosterd et al., N. Engl. J. Med. 340, 1221 (1999). and F. Niazi for assistance with patient evaluation; 3. D. L. Wingard, E. Barrett-Connor, M. H. Criqui, L. Suarez, D. Befroy and S. Dufour for assistance with spec- involved in oxidative phosphorylation is re- Am.J.Epidemiol.117, 19 (1983). troscopy; C. Ariyan and J. Kim for help with muscle duced among patients with type 2 diabetes 4. G. M. Reaven, Diabetes 37, 1595 (1988). biopsy; C. Garganta for measurement of urinary mellitus and insulin resistance (27). Thus, al- 5. M. H. Criqui et al., Circulation 73, I40 (1986). amino acids and organic acids; the staff of the Yale 6. R. R. Williams et al., JAMA 259, 3579 (1988). General Clinical Research Center; and A. Gharavi for though insulin resistance, obesity, and hy- 7. S. Mizushima, F. P. Cappuccio, R. Nichols, P. Elliott, helpful discussions. Supported by NIH grant nos. pertriglyceridemia are absent in K129, these J. Hum. Hypertens. 12, 447 (1998). MO1 RR-00125, P50 HL-55007, and R01 DK-49230. traits have been previously linked to loss 8. J. M. Peacock, A. R. Folsom, D. K. Arnett, J. H. Eckfeldt, A.H. is the recipient of an American Heart Associa- M. Szklo, Ann. Epidemiol. 9, 159 (1999). tion Fellowship (no. 0475003N). of mitochondrial function. These observa- 9. F. Guerrero-Romero, M. Rodriguez-Moran, Acta Dia- tions raise the possibility that all the fea- betol. 39, 209 (2002). Supporting Online Material tures of the metabolic syndrome can result 10. H. Masuzaki et al., Science 294, 2166 (2001). www.sciencemag.org/cgi/content/full/1102521/DC1 11. R. P. Lifton, A. G. Gharavi, D. S. Geller, Cell 104, 545 Materials and Methods from pleiotropic effects of impaired mito- (2001). SOM Text chondrial function; we speculate that the loss 12. J. L. Goldstein, M. S. Brown, Science 292, 1310 (2001). Figs. S1 to S5 of mitochondrial function with aging (26, 28) 13. G. I. Bell, K. S. Polonsky, Nature 414, 788 (2001). Tables S1 to S3 might commonly contribute to all compo- 14. S. George et al., Science 304, 1325 (2004). References and Notes 15. S. O’Rahilly, I. S. Farooqi, G. S. Yeo, B. G. Challis, nents of the metabolic syndrome. Endocrinology 144, 3757 (2003). 8 July 2004; accepted 6 September 2004 The variation in the phenotypic conse- 16. M. Konrad, K. P. Schlingmann, T. Gudermann, Am. J. Published online 21 October 2004; quences of this homoplasmic mitochondrial Physiol. Renal Physiol. 286, F599 (2004). 10.1126/science.1102521 17. Mutations in PPAR, and Akt2 may be an exception, Include this information when citing this paper. mutation is notable. Hypomagnesemia, hyper- tension, and hypercholesterolemia each show È50% penetrance among adults on the ma- ternal lineage. Incomplete penetrance arising Multidimensional Drug Profiling from homoplasmic mutations is well de- scribed and has been attributed to nuclear By Automated Microscopy genome and/or environmental modifiers (29). The nearly stochastic distributions of these Zachary E. Perlman,1,2* Michael D. Slack,3*. Yan Feng,1*- traits on the maternal lineage (fig. S2) and Timothy J. Mitchison,1,2 Lani F. Wu,3` the nonsignificant correlations among their Steven J. Altschuler3` quantitative values on the maternal lineage suggests that these are independent, pleio- We present a method for high-throughput cytological profiling by microscopy. tropic effects of the mitochondrial mutation. Our system provides quantitative multidimensional measures of individual cell Prior studies suggest potential mecha- states over wide ranges of perturbations. We profile dose-dependent phe- nisms linking each trait to impaired mito- notypic effects of drugs in human cell culture with a titration-invariant chondrial function. Cells of the DCT have the similarity score (TISS). This method successfully categorized blinded drugs and highest energy consumption of the nephron suggested targets for drugs of uncertain mechanism. Multivariate single-cell (30), and Mg2þ reabsorption in the DCT re- analysis is a starting point for identifying relationships among drug effects at a quires ATP-dependent Naþ reabsorption (31). systems level and a step toward phenotypic profiling at the single-cell level. Our Inhibitors of mitochondrial ATP production methods will be useful for discovering the mechanism and predicting the increase cholesterol biosynthesis while inhib- toxicity of new drugs. iting clearance in vitro (32). Finally, reduced ATP production has been reported in animal High-throughput methods for describing cell tial to complement these profiling approaches models of hypertension (33). Further work phenotype such as transcriptional and pro- by allowing fast and cheap collection of data will be required to elucidate the molecular teomic profiling allow broad, quantitative, describing protein behaviors and biological mechanisms linking genotype and phenotype. and machine-readable measures of the re- pathways within individual cells (5–9). Ac- The results of this study suggest that sponses of cell populations to perturbation cessing these data to produce useful profiles the loss of mitochondrial function with age (1–4). Automated microscopy has the poten- of cell phenotype will require new image

1194 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS analysis methods, the development of which but unrelated structures. We analyzed 13 were approximated as an annulus surrounding has so far lagged behind the adoption of high- threefold dilutions of each drug, covering a each identified nucleus (Fig. 1A and SOM throughput imaging technologies. final concentration range on cells from text B). For each cell, region, and probe, a In the context of drug discovery, profiling micromolar to picomolar Etable S2 and set of descriptors was measured. These technologies are useful in measuring both supporting online material (SOM) text A^. included measures of size, shape, and inten- drug action on a desired target in the cellular HeLa (human cancer) cells were cultured in sity, as well as ratios of intensities between milieu and drug action on other targets. 384-well plates to near confluence, treated regions (93 descriptors total, table S3). In all, Ideally, such profiling should be performed with drugs for 20 hours, fixed, and stained with È7 107 individual cells were identified as a function of drug concentration, because fluorescent probes for various cell components from 9600,000 images, yielding È109 data several factors make the effects of drugs and processes. We chose 11 distinct probes points. highly dose dependent. For example, the that covered a range of cell biology, multiplex- We can examine the population response degree to which a primary target is perturbed ing a DNA stain and two antibodies per well of each descriptor to increasing concentra- may affect different downstream pathways Ethe probe sets are SC35, anillin; "-tubulin, tions of a given drug, which we show with differently, and drugs can bind to multiple actin; phospho-p38, phospho–extracellular the genotoxic compound camptothecin (24) targets with different affinities. In some signal–regulated kinase (ERK); p53, c-Fos; (Fig. 1B). At low concentrations, the histo- cases, the therapeutic mechanism may in- phospho–adenosine 3¶,5¶-monophosphate re- gram for the total DNA content has the volve binding to more than one target with sponse element–binding protein (CREB), characteristic bimodal shape reflecting a ^ differing affinity (10, 11). To date, drug calmodulin . Using automated fluorescence mixture of G1, S, and G2/M cell populations. effects have been broadly profiled with microscopy, we collected images of up to G2 and M populations may be distinguished transcript analysis, proteomics, and measure- È8000 cells from each well. On each plate, 26 by 2D display of total DNA signal against ment of cell line dependence of toxicity (11– wells were treated only with dimethyl sulfox- nuclear area (25). As drug concentration

21). In these studies, multidimensional profil- ide (DMSO) to generate a control population increases, the cells arrest with S/G2 DNA ing methods were only applied at a single- (SOM text A). The experiment was per- content (24). The measured DNA content drug concentration. The only studies in which formed twice in parallel to provide a replicate distribution shifts leftward as dose increases, drug dose has been explicitly considered as a data set. Image segmentation procedures were and at the highest concentrations apoptosis is variable used the degree of cell proliferation, used to automatically identify nuclei and widely induced. Anillin, a cytokinesis protein an essentially one-dimensional (1D) readout of nuclear organelles, and cytoplasmic regions whose levels reflect cell cycle progression phenotype (12, 13). Two recent reviews have highlighted the possibility of using combina- tions of targeted phenotypic imaging screens to generate profiles of drug activity (6, 22). Here, we suggest that large sets of unbiased measure- ments might serve as high-dimensional cytol- ogical profiles analogous to transcriptional profiles. We present a method based on hypothesis-free molecular cytology that pro- vides multidimensional single-cell phenotypic information yet is simple and inexpensive enough to allow extensive dose-response profiles for many drugs. We assembled a test set of 100 com- pounds (table S1). Of these, 90 were drugs of known mechanism of action, six were blinded alternate titrations from this set of known drugs, one (didemnin B) was a toxin reported to have multiple biological targets (23), and three were drugs of unknown mechanism. The known drug set was chosen to cover common mechanisms of toxicity or therapeutic action in cancer and other dis- eases and to include several groups with a common target (macromolecule or pathway) Fig. 1. Key steps in algorithm for reducing image data to compound profile. (A) Image segmentation. For each image [examples show DNA (blue), SC35 (red), and anillin (green)], we 1Institute of Chemistry and Cell Biology, Harvard generated a nuclear region (blue) and a set of associated regions [shown here are cytoplasmic Medical School, Boston, MA 02115, USA. 2Depart- annulus (yellow) and SC35 speckles (red)]. For each defined nuclear region, we measure multiple ment of Systems Biology, Harvard Medical School, descriptors. (B) Quantification of population response. For a given compound, titration, and Boston, MA 02115, USA. 3Bauer Center for Genomics descriptor, we generated a population histogram and related cumulative distribution function (cdf) Research, Harvard University, Cambridge, MA 02138, (black) to be compared with the control population (blue). Shown is a threefold dilution series USA. ranging from 65 pM to 35 6M camptothecin. We reduced each experimental cdf to a single *These authors contributed equally to this work. dependent variable through comparison with a control population with the nonparametric KS .Present address: Alphatech, Inc., San Diego, CA statistic against a control population (SOM text C). Each vertical red or green line indicates the 92123, USA. position and sign of the maximal height difference between the curves; this height is the KS -Present address: Novartis Institutes for BioMedical statistic. (C) Heat map of compound profile. A z score is calculated for each KS statistic (SOM text Research, Cambridge, MA 02139, USA. C), and the vector of z scores for all descriptors and all titrations is displayed for rapid visual `To whom correspondence should be addressed. assessment. Increased scores are represented in red and decreased in green, with intensity E-mail: [email protected] (S.J.A.); lwu@ encoding magnitude. Arrowheads to the right indicate descriptors shown in (B), and the arrowhead cgr.harvard.edu (L.F.W.) at the bottom indicates the dose shown in (A).

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1195 R EPORTS (26), shows marked nuclear accumulation in ferent affinities. For example, camptothecin concentrations (24). Other drugs in our test

the G2 arrested state (Fig. 1A). p53, a trans- binds primarily to DNA complexes with set are known to have multiple targets, such cription factor that is part of the genotoxic topoisomerase I, promoting DNA strand as histone deacetylase inhibitors (27) and the response pathway, is strongly induced at high breaks and S-phase arrest at low concen- general kinase inhibitor staurosporine (28) camptothecin concentrations, but much less so trations, but it also blocks transcription and a and were thus expected to show complex

at concentrations sufficient to promote G2 number of other cellular processes at higher dose-response behavior. Such phenotypic arrest (Fig. 1B). For profiling studies, it is useful to reduce each population of descriptor values to a single number. Our study made several demands of this reduction: It must be able to compare distributions of arbitrary shape (Fig. 1B); it must be robust to variation in dynamic range and noise levels among different descriptors; it must convert differ- ent types of measurement into a common unit for comparison; it must be descriptor parameterization independent (e.g., an inten- sity ratio should behave the same as its reciprocal); and it must be insensitive to the precise quantitative relationship between antibody-staining intensity and antigen den- sity. We devised a measure based on the Kolmogorov-Smirnov (KS) statistic, allow- ing nonparametric comparison of experimen- tal and control distributions from the same plate (Fig. 1B, fig. S1, and SOM text C). Dividing by a measure of the variability within the control population yielded a z score, which can be displayed as a function of descriptor and drug concentration in a heat plot to allow rapid visual comparison of compound response profiles (Fig. 1C). These plots represent a family of dose-response curves for a single drug but differ from tra- ditional curves reflecting changes in a bio- Fig. 2. Comparison of compound profiles. As in Fig. 1C, the x axis shows increasing dose and the y chemical measurement. In particular, the axis encodes descriptors. Dose ranges are shown from 65 pM to 35 6M for all drugs except 6 relationship between z score and the original epothilone B, which is shown from .65 pM to .35 M. Color scale is as in Fig. 1C. For ease of physical measure may be nonlinear. For visualization, descriptors in all profiles are sorted in decreasing order of camptothecin response. (A) Compounds of similar mechanism show similar profiles. Shown are representative compound example, the statistically significant res- profiles. HDAC, histone deacetylase; ALLN, N-acetyl-Leu-Leu-norleucinal. (B) Compound profiles ponses of p53 to low doses of camptothecin can distinguish differences between drugs with similar mechanisms. Wells with too few cells for (Fig. 1C) reflect subtle effects not easily analysis are represented in white. discerned by eye in the source images. The heat plots typically have a sharp transition, reflecting a concentration at Table 1. Assessment of TISS by literature categories. For each category that has more than two which many descriptors become different compounds, we computed two sets of TISS scores: pairwise TISS comparisons between members of the category (intrapair) and comparisons in which only one element of the pair is in the category (interpair). from control values. We will refer to this as As a crude in silico comparison to other cell-based assays such as fluorescence-activated cell sorting the primary effective concentration (PEC) (single-cell based) and cytoblots (whole-population based), we repeated this procedure with a descriptor for the drug. The isolated responses observed set consisting of only total intensity measures and compared it with either our KS-based TISS values or a at some low concentrations represent noise mean-based TISS values (SOM text C). P values (columns 2 to 4) describe the probability that the rank that could be reduced by increasing repli- ordering of the two sets of TISS values would have been seen by random draws from the same cates, improving experimental procedures, distribution (SOM text C). KS, KS-based TISS (P value); mean, mean-based TISS (P value). and normalizing for local variation in cell density. For 39 drugs, we saw no strong Total intensity No. pairwise TISS All descriptors descriptors comparisons effect, leaving a heat plot dominated by Category noise. Those drugs either lack a target in KS KS Mean Intrapair Interpair HeLa cells, were used at inactive dosages, or Actin 0.025 0.776 0.327 6 218 effected changes not detectable with our DNA replication 0.011 0.057 0.007 3 168 antibody set. For almost all of the 61 drugs Histone deacetylase 0.001 0.024 0.489 10 265 that showed a strong response, some descrip- Kinase 0.223 0.746 0.902 3 168 tors responded at concentrations other than Kinase CDK 0.057 0.221 0.050 6 218 the PEC (Fig. 2). This may reflect varying Microtubule 3.86 10j20 9.81 10j6 0.295 55 484 j5 biological consequences of low and high Protein synthesis 6.02 10 0.004 0.180 15 309 Topoisomerase 0.005 0.011 0.693 3 168 saturation of a single target, or it may reflect Vesicle trafficking 0.206 0.314 0.514 3 168 interactions with multiple targets with dif-

1196 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org R EPORTS complexity may help explain why toxicity at developed a titration-invariant similarity Of the blinded alternate titrations of known high doses is common even for therapeutic score (TISS) to allow comparison between drugs, scriptaid, hydroxyurea, emetine, and drugs that are apparently highly selective at dose-response profiles independent of start- two alternate series of nocodazole showed sig- the level of target binding. ing dose (SOM text C). TISS values were nificant responses. These clustered closely with Drugs with common reported targets but generated for the 61 compounds that showed their unblinded counterparts and compounds of diverse chemical structures often showed significant signal, and these were used for similar reported mechanism. Didemnin B, for similar profiles readily distinguished from unsupervised clustering (Fig. 3). TISS was which the reported range of activities includes those of drugs of different mechanism (Fig. successful at grouping compounds with sim- inhibition of protein synthesis (23), clustered 2A). In other cases, markedly different ilar reported targets (Table 1). As expected, with ribosome inhibitors (Fig. 2B). Two of the profiles were evident within a family, most clustering reflected biological mechanism three poorly characterized compounds showed notably the protein synthesis inhibitors (Fig. rather than chemical similarity. For example, strong responses. One, concentramide, is 2B). This may reflect different cell responses kinase inhibitors, most of which are adeno- difficult to interpret. The other, austocystin, to alternative biochemical mechanisms of sine 5¶-triphosphate–mimetic compounds, did clusters with transcription and translation poisoning ribosomes (29) or perhaps the not cluster as a group. Clustering was poor inhibitors. Preliminary experiments suggest existence of alternate targets (23). even within a set of kinase inhibitors with that this compound inhibits transcription in When comparing drug mechanism, changes overlapping targets Ecyclin-dependent kinase vitro (25). Thus, our methods can group com- in specificity (and thus phenotype) are rele- (CDK) inhibitors^, perhaps reflecting variable pounds of like mechanism and thereby vant, but changes in affinity (and thus PEC) inhibition of other kinases. The CDK inhib- suggest mechanism for new drugs. are not. Two different dosage series of the itors related by structure and reported target Extensions of cytological profiling to same drug should result in similar heat plots (purvalanol, roscovitine, and olomoucine) did reflect dependencies among descriptors will shifted along the concentration axis. We cluster. allow more sophisticated analysis of drug

Fig. 3. Hierarchical clustering of the 61 most responsive compound profiles ture. In blue are compounds that were blinded or are of unknown mech- by TISS values. Compound stock concentrations (6M) are in parentheses anism. Middle panel shows matrix of P values derived from pairwise TISS (fig. S3). Left panel shows mechanism of compound as described in litera- values (SOM text C). Dendrogram at top shows degree of association.

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1197 R EPORTS Fig. 4. Single-cell analysis shows dif- References and Notes fering patterns of dose-dependent 1. M. B. Eisen, P. T. Spellman, P. O. Brown, D. Botstein, p53 and c-Fos responses to differ- Proc. Natl. Acad. Sci. U.S.A. 95, 14863 (1998). ent drugs. (A) Scatter plot of 2. A. C. Gavin et al., Nature 415, 141 (2002). 3. Y. Ho et al., Nature 415, 180 (2002). average nuclear p53 intensity 4. P. Uetz et al., Nature 403, 623 (2000). versus average c-Fos intensity in 5. R. F. Murphy, M. Velliste, G. Porreca, J. VLSI Signal a typical control well and repre- Process. 35, 311 (2003). sentative image. The bright cells 6. J. H. Price et al., J. Cell. Biochem. Suppl. 39, 194 atthetopoftheimagearein (2002). mitosis. (B) Dose-dependent in- 7. W. K. Huh et al., Nature 425, 686 (2003). creases in response to MG132 8. T. L. Saito et al., Nucleic Acids Res. 32, D319 (2004). shown in heat maps are correlat- 9. C. Conrad et al., Genome Res. 14, 1130 (2004). 10. J. G. Hardman, L. E. Limbird, A. G. Gilman, Eds., The ed in scatter plots and images Pharmacological Basis of Therapeutics (McGraw-Hill, (orange nuclei), shown for the four New York, ed. 10, 2001), pp. 39–42. highest concentrations. (C)Dose- 11. M. J. Marton et al., Nature Med. 4, 1293 (1998). dependent increases in response to campto- 12. J. N. Weinstein et al., Science 275, 343 (1997). thecin shown in heat maps are anticorre- 13. K. D. Paull, C. M. Lin, L. Malspeis, E. Hamel, Cancer Res. lated in scatter plots and images. The black 52, 3892 (1992). (c-Fos) and green (p53) heat map values 14. U. Scherf et al., Nature Genet. 24, 236 (2000). for the highest dose reflect the contribu- 15. E. C. Gunther, D. J. Stone, R. W. Gerwien, P. Bento, M. P. Heyes, Proc. Natl. Acad. Sci. U.S.A. 100, 9608 tion of apoptotic cells with negligible p53 (2003). and c-Fos nuclear staining. 16. D. Leung, C. Hardouin, D. L. Boger, B. F. Cravatt, Nature Biotechnol. 21, 687 (2003). 17. M. A. Lindsay, Nature Rev. Drug Discov. 2, 831 (2003). 18. P. Y. Lum et al., Cell 116, 121 (2004). 19. G. Giaever et al., Proc. Natl. Acad. Sci. U.S.A. 101, 793 (2004). 20. S. J. Haggarty, P. A. Clemons, S. L. Schreiber, J. Am. Chem. Soc. 125, 10543 (2003). responses at a systems level. For example, terized drugs and to suggest systems-level 21. D. E. Root, S. P. Flaherty, B. P. Kelley, B. R. Stockwell, Chem. Biol. 10, 881 (2003). both p53 and c-Fos, a transcription factor relationships between signaling pathways. 22. V. C. Abraham, D. L. Taylor, J. R. Haskins, Trends involved in mitogen-activated protein kinase The complex dose-response curves and large Biotechnol. 22, 15 (2004). (MAPK) signaling, are involved in cell stress cell-to-cell variability we frequently observed 23. M. D. Vera, M. M. Joullie, Med. Res. Rev. 22, 102 (2002). responses, but the interrelationship of the reinforce the utility of unbiased multidimen- 24. C. J. Thomas, N. J. Rahier, S. M. Hecht, Bioorg. Med. p53 and MAPK pathways is poorly under- sional characterization of drug effects over Chem. 12, 1585 (2004). stood (30). Single-cell profiling reveals that wide ranges of doses. 25. Z. E. Perlman et al., data not shown. 26. C. M. Field, B. M. Alberts, J. Cell Biol. 131, 165 (1995). different drug mechanisms induce different Many improvements and extensions of 27. M. Yoshida et al., Cancer Chemother. Pharmacol. 48 relative patterns of response by these two this work are possible. These include better (suppl. 1), S20 (2001). pathways (Fig. 4). The proteasome inhibitor lab automation, broader drug reference sets, 28. M. E. Noble, J. A. Endicott, L. N. Johnson, Science 303, MG132 causes increased correlated induc- different types of perturbation (such as RNA 1800 (2004). 29. J. D. Laskin, D. E. Heck, D. L. Laskin, Toxicol. Sci. 69, tion in these pathways, whereas responses to interference), improved strategies for cell 289 (2002). camptothecin are anticorrelated. Anticorre- segmentation, more sophisticated feature ex- 30. B. Kaina, Biochem. Pharmacol. 66, 1547 (2003). lated responses observed in fixed-time traction (5, 9), different sets of antibody 31. G. Lahav et al., Nature Genet. 36, 147 (2004). 32. We thank A. Daneau, M. Ethier, and B. Mantenuto at images may reflect switching of mutually probes and cells, the inclusion of more time the Bauer Center for Genomics Research for their exclusive cell states in response to different points and live cell imaging, and the inte- assistance with the use of the Bauer Center compu- degrees of stress or might reflect a dynamic gration of complementary profiling strategies. tational cluster and K. Maciag, A. Murray, O. Rando, A. Salic, and A. Yonetani for helpful discussions. temporal response, such as oscillation, that is Additionally, our methods may be extended to Supported in part by National Cancer Institute PO1 not synchronized among cells (31). These allow the characterization of responses by CA078048. Z.E.P. is a Howard Hughes Medical Institute data help establish a concentration and time subpopulations defined by such variables as Predoctoral Fellow. window, but live imaging will be required to cell cycle state, cell density, or neighboring Supporting Online Material distinguish between these hypotheses. environment. This analysis, extended to work www.sciencemag.org/cgi/content/full/306/5699/1194/ DC1 Cytometric dose-response profiling is a in tissues or clinical samples, offers the po- SOM Text fast and cheap method for quantitatively tential to speed the identification of toxic Figs. S1 to S5 surveying broad ranges of individual cell compounds during therapeutic drug develop- Tables S1 to S4 Database S1 responses. We have used our methods to ment and the targeting of drug effects to assign mechanism to blinded and uncharac- specific subtypes of cells. 25 May 2004; accepted 8 September 2004

1198 12 NOVEMBER 2004 VOL 306 SCIENCE www.sciencemag.org N EW PRODUCTS

Millipore KINASE ASSAY PLATES surement is accomplished with an in situ fiber-optic probe to insure For more information MultiScreenHTS-PH filter plates for fast, accurate, and convenient reaction feedback control. 800-MILLIPORE high-throughput kinase screening as- www.millipore.com says are automation-compatible 96- Greiner Bio-One LOW BIREFRINGENCE http://science.labvelocity.com well plates that follow a prevalidated For more information MICROPLATES protocol for high sensitivity, high 800-884-4703 As a key technology in structural ge- specificity, and reliability. For optimized kinase assay performance, the www.gbo.com nomics and structure-based drug de- plates incorporate PH paper, which retains peptide substrates to re- http://science.labvelocity.com velopment, protein crystallization re- move unreacted, unlabeled adenosine triphosphate. The PH paper also quires increasingly efficient analytical provides lower background and improved discrimination between procedures. Polarized light is now used with imaging systems to better drug target hits and non-hits. In addition, the paper is not subject to distinguish protein crystal structures from precipitate artifacts. naturally fluorescent or quenching compounds. For this reason, the Polystyrene, the material widely used for microplate manufacture, ex- plates are especially compatible with aromatic-based libraries. Kinase hibits a wide spectrum of color to interfere with polarized detection. assays are a critical step in both lead identification and characteriza- To reduce this kind of background interference, the CrystalQuick-LBR tion. Compared with most homogeneous assays kinase assays, com- Plates for sitting-drop vapor diffusions are manufactured of a low bire- pleted on MultiScreenHTS-PH filter plates are easy to set up. fringent resin with superior optical properties for crystal scoring. Char- acterized by exceptional transparency, high chemical resistance, and Applied Kilovolts HIGH-VOLTAGE POWER SUPPLIES low water absorption, the plates feature three crystallization wells per For more information These precision high-voltage power each 96-well reservoir. Available with “square” flat bottom or “round” 410-308-0069 supplies offer performance enhance- curved bottom crystallization wells, the plates facilitate high-through- www.appliedkilovoltsms.com ment options for mass spectrometer put screening in manual or fully automated systems. http://science.labvelocity.com producers. Developed to minimize both direct and injected ripple, the Textco BioSoftware GENE INSPECTOR SOFTWARE HP Series power supplies can ensure maximum sensitivity and res- For more information Gene Inspector 1.6 for Macintosh OS X olution while maintaining extreme-temperature stability and long- 603-643-1471 integrates a comprehensive sequence term reliability. They are available in a variety of configurations and www.textco.com analysis package and powerful illustra- power supply ranges to meet specific engineering requirements. http://science.labvelocity.com tion tools into a versatile electronic re- From 500V to 30kV, the HP, HPR, HPF, and HPRD Series for mass search notebook. The intuitive interface spectrometers and floating detector applications offer precision allows users to organize large amounts of diverse yet interrelated infor- unipolar and bipolar high-voltage supplies with low output ripple. mation, enabling researchers to plan, annotate, and execute experiments and analyses in a single application. The notebook features help users Niles Scientific MICROARRAY IMAGE easily produce documents to summarize and manage complex experi- For more information PROCESSING SOFTWARE ments and produce publication-quality illustrations. The analysis setups 415-346-6461 SpotReader 1.0 is new microarray allow parameters to be set and stored so consistent results are easy to www.spotreader.com image processing software for spot- obtain, even among multiple users. The program features more than 60 http://science.labvelocity.com ted (robot printed) images. Major standard analyses for both nucleic acids and peptides, including multiple features include hands-off creation sequence alignments, BLAST searching, dot matrix comparisons, and of data analysis grids, a mini-database to organize data, and a prediction of the physical properties of translation-product peptides. significant advance in a grid editing user interface. Automation features allow walk-away processing of microarray images: if you Axxora LITERATURE have your scanner save images into a folder, grids are made auto- For more information Fluorescent Proteins is a 16-page guide matically. The true Windows interface provides live-update auto- 858-550-8830 describing the use of novel fluorescent matic flagging, a pop-up magnifying glass, modeless editing, and www.axxora.com proteins in various biological activities. a feature-pinning function for extremely hard-to-grid arrays. http://science.labvelocity.com Among the proteins featured are KFP- Red, a photoactivatable Kindling-Red CEM Corp MICROWAVE SYSTEM FOR fluorescent protein; CopGFP, a bright and fast-maturing green fluores- For more information LOWER TEMPERATURES cent protein; PhyYFP, a true yellow fluorescent protein; and t-HcRed, a 704-821-7015 The Discover CoolMate microwave non-oligomeric far-red fluorescent protein. This reference details a vari- www.cem.com system is designed to perform sub- ety of in vivo applications, including gene expression analysis, protein http://science.labvelocity.com ambient temperature chemistries. localization, cell labeling, and protein–protein interaction and co-local- Discover CoolMate performs tem- ization studies. It discusses the general properties, advantages, limita- perature-sensitive reactions including carbohydrate chemistry and tions, and recommended uses of these innovative fluorescent proteins. the formation of carbanions and other highly reactive intermediates easily and controllably. The system also allows the safe use of highly Newly offered instrumentation, apparatus, and laboratory materials of interest to re- reactive species, which would normally be excluded when using a searchers in all disciplines in academic, industrial, and government organizations are microwave system. The system’s proprietary, jacketed, low-tempera- featured in this space. Emphasis is given to purpose, chief characteristics, and availabil- ity of products and materials. Endorsement by Science or AAAS of any products or ture vessel and cooling technology keep the bulk temperature of the materials mentioned is not implied. Additional information may be obtained from the reactants low, preventing thermal degradation of the product while manufacturer or supplier by visiting http://science.labvelocity.com on the Web, where microwave energy drives the reaction kinetically. Temperature mea- you can request that the information be sent to you by e-mail, fax, mail, or telephone.

www.sciencemag.org SCIENCE VOL 306 12 NOVEMBER 2004 1199 Published by AAAS