PERSPECTIVES ASTRONOMY On 25 April 1998, BeppoSAX yielded another surprise. Optical observations of an The Link Between Supernovae apparently normal GRB showed a young, very energetic exploding star—a type Ic supernova—at a distance of 80 million light and Gamma Ray Bursts years, one-hundredth the distance of typical GRBs (5). The optical brightness of the super- Brian Schmidt nova, named 1998bw, was high for an object at this distance, but its gamma ray, amma ray bursts (GRBs) were first x-ray, optical, and radio bright- discovered in 1967 by American nesses are several orders of magni- military satellites. The discovery tude fainter than for any other GRB G SN 2003jd was declassified in 1973; within months, observed to date. more ideas were published about the origin Type Ic supernovae are thought of these cosmic explosions than there were to occur when stars 10 to 100 detected bursts. Since then, observations of times the mass of the Sun, which thousands of these objects have yielded a have been stripped of their outer deeper understanding of GRBs, but ques- layers of hydrogen and helium, tions remain. For example, it is known that run out of nuclear fuel in their supernovae underlie some GRBs, but how center and collapse into neutron some supernovae produce GRBs remains stars or black holes. If the dying unclear. star that formed supernova On page 1284 of this issue, Mazzali et 1998bw had lost its outer layers al. (1) shed some light on the link between as a result of a merger with GRBs and supernovae. They report that another star, it should have been supernova 2003jd (see the figure) appears A cosmic explosion 250 million light years away. Observa- spinning rapidly at the time of its to have been disrupted by a highly asym- tions of supernova (SN) 2003jd show it to be a highly ener- death. Theorists therefore pro- metric process, similar to the jets that pro- getic explosion, similar to the hypernovae associated with posed that a jet, produced by the duce GRBs. Thus, supernova 2003jd may GRBs. Mazzali et al. (1) argue that this object was ripped apart collapse of a rapidly rotating itself be a GRB, with its jets pointed away by something like a jet, suggesting that SN 2003jd might have massive star into a black hole, from Earth when it exploded. This is the been a GRB with its jets directed away from Earth. created supernova 1998bw and first time that researchers have found a could produce typical GRBs (6). supernova that might be a GRB, rather than gun, theorists could still bend their models The jet produces observable gamma rays if a GRB that might be a supernova. of nearby bursts to fit the observations. it points toward the observer; the resulting The data revolution for GRBs started in In 1996, the Italian-Dutch BeppoSAX expanding debris looks like a very ener- 1991 with the launch of NASA’s Compton satellite was launched. This satellite was getic type Ic supernova—a so-called hyper- Gamma Ray Observatory. This satellite and not as sensitive as BATSE to gamma rays, nova. This model could explain supernova its dedicated GRB instrument, BATSE but—inspired by predictions that lower- 1998bw, but its relevance to the rest of the (Burst and Transient Source Experiment), energy x-rays would persist long after the GRB population was unclear because of the detected more than 8000 bursts until it was burst of gamma rays—it was able to point very different energies of supernova sent into the Pacific Ocean in 2000. Before an x-ray detector onto a GRB within a few 1998bw and other more distant GRBs. BATSE, models for GRBs ranged from col- hours of the burst. Beppo-SAX’s x-ray Since the discovery of 1998bw, much lisions of objects in our solar system to detectors could localize these x-ray emis- effort has been put into studying the connec- mergers of black holes on the other side of sions to much less than 1° uncertainty. This tion between GRBs and hypernovae. The the universe. BATSE observations showed precision was sufficient for optical and HETE-2 (High Energy Transient Explorer that GRBs are distributed uniformly across radio telescopes to follow up individual Mission) satellite, launched in 2000, and the the sky, providing strong evidence that they objects and look for these explosions. Swift Gamma Ray Burst Explorer, launched must be occurring in the distant universe This strategy paid off when a GRB in 2004, are providing accurate GRB posi- (2), because it is only at great distances that detected on 28 February 1997 was localized tions for astronomers to chase on an almost the sky begins to look uniform. in x-rays; astronomers using the William daily basis. The light associated with a GRB However, although BATSE was very Herschel Telescope in the Canary Islands jet fades quickly, whereas the brightness of a sensitive to high-energy photons, it could then found a fading optical afterglow in a supernova tends to brighten for 1 to 3 weeks. not discern the location of a burst to better distant galaxy at this location (3). This dis- Therefore, 2 weeks past an explosion, a than a few degrees uncertainty—an area of covery confirmed the cosmological origin supernova should in most cases outshine a sky too large to pinpoint the location of of GRBs. Another event detected on 8 May GRB’s jet. Observations of nearby GRBs individual explosions. Without a smoking 1997 was determined to be more than 7 bil- have shown that some objects exhibit super- lion light years away—halfway across the nova-like features as their brightness evolves visible universe (4). The huge distance over time. The author is in the Research School of Astronomy implied an enormous energy. GRBs were and Astrophysics, Mount Stromlo Observatory, The GRB-hypernova connection was Australian National University, Weston, ACT 2611, thus proclaimed to be the largest bangs in confirmed directly on 29 March 2003,

CREDIT: UNIVERSITY NATIONAL BRIAN SCHMIDT/AUSTRALIAN Australia. E-mail: [email protected] the universe since the big one. when HETE-2 discovered an object one-

www.sciencemag.org SCIENCE VOL 308 27 MAY 2005 1265 Published by AAAS P ERSPECTIVES third as distant as any other cosmological verse, one can gauge the fraction of reported in (1) provide substantial evidence GRB to that date. As the object faded, hypernovae that are misdirected GRBs. that the engine that powered supernova spectra taken with optical telescopes Current data suggest that this fraction 2003jd to become a hypernova is the same revealed an underlying supernova—an should be substantial (10). To date, rela- that creates GRBs in the distant universe. object almost identical to supernova tively few hypernovae have been discov- With two dedicated GRB telescopes and 1998bw (7, 8), but with a gamma ray ered in the nearby universe, but none many programs scanning the sky for super- energy a thousand times higher. seem to show the intense radio emission novae, the next few years promise many The connection between GRBs and expected for misdirected GRBs. more GRB-hypernova connections. These hypernovae thus seems secure, but impor- Mazzali et al.’s observations of supernova results should help astronomers to further tant questions remain. The GRB jet of 2003jd, a hypernova at a distance of 250 mil- untangle the mysteries of GRBs. supernova 1998bw is much less energetic lion light years (see the figure), may resolve than that associated with any other GRB to the latter problem. The authors provide evi- References date, and no consensus has emerged as to dence for the gross asymmetries that are 1. P. A. Mazzali et al., Science 308, 1284 (2005). 2. C.A. Meegan et al., Nature 355, 143 (1992). how a single mechanism could produce expected for a GRB-hypernova. After the 3. J. van Paradijs et al., Nature 386, 686 (1997). supernova 1998bw and the much more discovery of 2003jd, observers used the Very 4. M. R. Metzger et al., Nature 387, 878 (1997). powerful bursts in the distant universe. Large Array radio telescope to look at it, but 5. T. J. Galama et al. Nature 395, 670 (1998). 6. A. I. MacFadyen, S. E. Woosley, Astrophys. J. 524, 262 In addition, the jets of most GRB-hyper- did not detect it at radio wavelengths. (1999). novae should aim away from Earth; Mazzali et al. argue that better radio or x-ray 7. K. Z. Stanek et al., Astrophys. J. 591, L17 (2003). 8. J. Hjorth et al., Nature 423, 847 (2003). although not visible in gamma rays, these observations are needed to reveal whether 9. E. Berger et al., Astrophys. J. 599, 408 (2003). “misdirected” GRBs should be seen at radio this hypernova had a GRB jet. Even if subse- 10. P. Podsiadlowski et al., Astrophys. J. 607, L17 (2004). wavelengths (9). By comparing the num- quent observations fail to find jets associated bers of hypernovae and GRBs in the uni- with supernova 2003jd, the observations 10.1126/science.1112928

CELL BIOLOGY On page 1318 in this issue, Cohen et al. (3) describe such an approach to designing Lessons in Rational Drug Design small-molecule inhibitors of p90 ribosomal S6 kinases (RSKs) by selectively targeting two determinants, or “selectivity filters,” in for Protein Kinases the ATP binding pocket. One selectivity fil- ter, the “gatekeeper,” is a residue that flanks Natalie G.Ahn and Katheryn A. Resing a highly variable hydrophobic pocket at the rear of the ATP binding site (4). When the ow do you design a drug that selec- that selectively inhibits a single enzyme. side chain of the gatekeeper residue occu- tively recognizes one enzyme from This complicates strategies for developing pies a small volume, the hydrophobic Hamong hundreds, all of which share protein kinase inhibitors. It is difficult to pocket is empty. This allows bulky groups the same substrate? This is the challenge identify small-molecule compounds with on adenine ring analogs to fit into the confronting biochemists when developing high specificity for a single kinase. One pocket. On the other hand, a gatekeeper small-molecule inhibitors of protein must further verify that residue with a large side kinases, enzymes that regulate cellular cellular responses to drug Hydrophobic pocket chain precludes binding of growth, homeostasis, and signal transduc- treatment are caused by bulky substituents. This tion. Fifteen years ago, intracellular kinases inhibition of the targeted mechanistic understanding were considered too ubiquitous to be useful enzyme. of selectivity enables candi- drug targets. But with mounting evidence Enter rational drug HO date inhibitors to be identi- for highly specific function in many design, combining struc- fied by comparing aligned kinases, many pharmaceutical endeavors tural determination and NH2 HS sequences at the gatekeeper today have screening programs for computational modeling to F position. inhibitors of this class of enzymes. identify small molecules N The second selectivity Since the mid-1990s, three kinase that complement amino N O filter is a reactive cysteine inhibitors have been approved by the U.S. acid residues within the N residue within the active Food and Drug Administration (FDA) for active site of a target site. The strategy of target- treatment of chronic myelogenous leukemia, enzyme. So far, only a few HO ing active-site cysteine gastrointestinal stromal tumors, and lung successful lead compounds residues was previously tumors (1, 2), and many others are in clinical have been developed based Proposed binding mode of a successful in developing rationally designed inhibitor of trials. Each drug works by competitively dis- on principles of molecular irreversible inhibitors for p90 RSK. Two “selectivity filters” placing adenosine 5´-triphosphate (ATP), recognition, as opposed to in the ATP binding site are required the epidermal growth fac- the nucleotide that binds to the active site of random screening. How- for potent inhibition. A threonine tor (EGF) receptor (5). protein kinases. However, because all ever, bioinformatics strate- residue (red) in the gatekeeper However, cysteine residues kinases share molecular recognition deter- gies have added new tools position allows the inhibitor within active sites of kinases minants in their conserved nucleotide bind- to the arsenal, enabling re- (black) to access a hydrophobic are relatively rare. Using ing pocket, no ATP analog has been found searchers to predict varia- pocket. A poorly conserved cys- bioinformatics analysis, tions in nucleotide and pro- teine (blue; valine in many kinases) Cohen et al. showed that The authors are in the Department of Chemistry and tein sequence that may is positioned for attack by the elec- only 11 kinases among a set Biochemistry, University of Colorado, Boulder, CO serve as specificity deter- trophilic fluoromethylketone sub- of 491 had cysteine residues

80309, USA. E-mail: [email protected] minants. stituent of the inhibitor (F; green). within the conserved gly- CREDIT: M. COHEN AND R. BATEMAN

1266 27 MAY 2005 VOL 308 SCIENCE www.sciencemag.org Published by AAAS