Information3). Another hypothesis, that the modal shift in El Niño is just a manifestation of natural climate variability on decadal to cen-tennial timescales, also needs to be examined, at least in model simulations. One such study13 indicates that both the frequency and the inten-sity of the ENSO change on multi-centennial timescales.

From the perspective of society’s adapta-tion to climate change, predicting the El Niño Modoki is essential because of its distinctive global impacts on various timescales4–11,14,15. However, our recent experience shows that the effectiveness of the current coupled models in predicting the ENSO Modoki index is low compared with their predictions of the conven-tional ENSO indices. Further improvement of methods for assimilating observed data into models, as well as improvement of seasonal

prediction models, is necessary. It is also worth examining whether the increasing frequency of the Modoki is related to the recent decadal shift in another ocean–atmosphere phenomenon, the Indian Ocean Dipole: over the past 50 years the SST of the Indian Ocean has increased by about 0.6 °C. ■

Karumuri Ashok is at the APEC Climate Center,

1463 U-dong, Haeundae-gu, Busan 612-020,

Republic of Korea. Toshio Yamagata is at the

Application Laboratory, JAMSTEC,

3173-25 Showa-machi, Kanazawa-ku,

Yokohama 236-0001, Japan.

e-mails: ashok@apcc21.net (karumuriashok@

hotmail.com);

yamagata@jamstec.go.jp

1. Rasmusson, E. M. & Carpenter, T. H. Mon. Weath. Rev. 110, 354–384 (1982).

CHEMICAL BIOLOGY

Caught in the activationYi Liu

A crystal structure reveals how a protein kinase is activated by the binding of a small molecule at a pocket far from the catalytic site. This opens the door to the design of modulators of protein phosphorylation.

The protein functions that regulate cell behav-iour are themselves tightly regulated by many different inputs. One of these is phosphoryla-tion, the process in which a phosphate group from an ATP molecule is attached to a protein by a protein kinase enzyme1. There is consid-erable interest in identifying small molecules that can modulate protein kinase activity, both as research tools and as potential human thera-peutic agents. Indeed, eleven kinase inhibitors are already used as drugs for treating cancer, and several more are in clinical trials for the treatment of immune and inflammatory diseases2.

Identifying protein kinase inhibitors is rela-tively easy — it’s ‘simply’ a matter of finding molecules that prevent substrates from bind-ing to the catalytic site of the enzyme. It has proved much more difficult to identify small molecules that activate these enzymes, but a few have been discovered, including those described in 2006 by Engel and colleagues3. Reporting in Nature Chemical Biology, the same group (Hindie et al.4) now disclose the crystal structure of the protein kinase PDK1 in a complex with one of these activators. This is the first crystal structure of a small-molecule activator bound to a protein kinase, and it pro-vides a structural explanation of how this bind-ing activates the enzyme.

A protein kinase catalytic domain consists of a small amino-terminal lobe and a large carboxy-terminal lobe (Fig. 1a). ATP binds in the cleft between the two lobes, whereupon its

terminal phosphate group can be transferred to the protein substrate also bound to the enzyme. Three regions in the ATP-binding site must be precisely positioned for efficient catalysis of the phosphate-transfer reaction: a loop rich in gly-cine amino acids (the Gly-rich loop); an α-helix

Figure 1 | Small-molecule activation of PDK1. The protein kinase enzyme PDK1 transfers phosphate groups from ATP to a protein substrate, and consists of a large and a small lobe. a, Three regions around the ATP-binding site of the enzyme are critical for PDK1 activity — a loop rich in the amino acid glycine (the Gly-rich loop), the α-C-helix and an activation loop. In the inactive state, these regions are flexible and in motion (indicated by dotted lines), and are not aligned for catalysis. Part of the activation loop is unstructured (dashed lines). b, Hindie et al.4 report the crystal structure of PDK1 in a complex with a small-molecule activator of the enzyme. The activator binds at a location known as the PIF-binding pocket, which is remote from the ATP-binding site. Activator binding induces local conformational changes around the PIF-binding pocket, which stabilize the structures of the nearby Gly-rich loop and α-C-helix, leading to conformational changes in these two regions. This stabilization acts as a signal to the disordered region of the activation loop, which adopts an ordered structure in response. Thus, binding of a small-molecule activator to the PIF-binding pocket orients the ATP-binding site into a conformation that is efficient for catalysing phosphate-transfer reactions.

Small-molecule

activator

a b

Activation loopLarge lobe

Small lobe

ATP binding site

PIF-binding

pocket

α-C-helix

Gly-rich

loop

ATA

2. Philander, S. G. H. El Niño, La Niña, and the Southern

Oscillation (Academic, 1990).

3. Yeh, S.-W. et al. Nature 461, 511–514 (2009).

4. Ashok, K., Behera, S. K., Rao, S. A., Weng, H. & Yamagata, T.

J. Geophys. Res. 112, doi:10.1029/2006JC003798 (2007).

5. Trenberth, K. E. & Stepaniak, D. P. J. Clim. 14, 1697–1701

(2001).

6. Larkin, N. K. & Harrison, D. E. Geophys. Res. Lett. 32, doi:10.1029/2005GL022738 (2005).

7. Kumar, K. K. et al. Science 314, 115–119 (2006).

8. Wang, G. & Hendon, H. H. J. Clim. 20, 4211–4226 (2007).

9. Weng, H., Ashok, K., Behera, S. K., Rao, S. A. & Yamagata,

T. Clim. Dynam. 29, doi:10.1007/s00382-008-0394-6

(2007).

10. Weng, H., Behera, S. K. & Yamagata, T. Clim. Dynam. 32, doi:10.1007/s00382-008-0394-6 (2009).

11. Kao, H.-Y. & Yu, J.-Y. J. Clim. 22, 615–632 (2009).

12. Kug, J.-S., Jin, F.-F. & An, S.-I. J. Clim. 22, 1499–1515 (2009).

13. Wittenberg, A. T. Geophys. Res. Lett. 36, doi:10.1029/

2009GL038710 (2009).

14. Kim, H.-M., Webster, P. J. & Curry, J. A. Science 325, 77–80

(2009).

15. Ashok, K., Tam, C.-Y. & Lee, W.-J. Geophys. Res. Lett. 36, doi:10.1029/2009GL038847 (2009).

from the small lobe (known as the α-C-helix); and the ‘activation’ loop from the large lobe. To be activated, most protein kinases must first be phosphorylated in the activation loop5, but for full activation some also require a protein to bind to a site remote from the ATP site6. In PDK1, for example, a binding site known as the PIF-binding pocket in the small lobe recog-nizes a phosphorylated amino-acid sequence that activates the kinase7.

The authors’ PDK1 activators3 bind at the PIF-binding pocket, thus offering the team a golden opportunity to investigate the molecular mechanism of kinase activation mediated by this site. By solving the crystal structure4 of a PDK1–activator complex, they found that activator binding causes a change in the conformation

484

NATURE|Vol 461|24 September 2009NEWS & VIEWS

479-488 News and Views MH IF.indd 484479-488 News and Views MH IF.indd 484 18/9/09 17:41:5218/9/09 17:41:52

© 2009 Macmillan Publishers Limited. All rights reserved

mutated specific amino acids in PDK1 to identify an amino-acid residue that is impor-tant in transducing activator binding at the PIF-binding pocket into the conformational change that activates the enzyme. The crucial residue was identified as a threonine resi-due located in the activation loop; when the threonine is replaced by a tryptophan, the mutant enzyme still allows small-molecule activators to bind to its PIF-binding pocket, but no activation occurs.

The potency of the PDK1 activators used in this study4 is relatively weak, and the carboxylate group present in the molecules may prevent them from entering cells, both of which could prevent these compounds from being developed as drugs. Nevertheless, by providing a molecular basis for protein kinase activation, Hindie and colleagues’ findings are a major advance in the field. They also raise some intriguing questions. For example, could small molecules be designed that induce a dif-ferent set of conformational changes in the

PIF-binding pocket and inhibit PDK1 activity? Time will tell, but for now this work opens the door to the rational design of small-molecule modulators that target the PIF-binding pocket in PDK1, and in other kinases that have similar regulatory pockets in their small lobes. Such molecules might serve as tools for studying the functions of protein kinases, and as potential new entry points for drug discovery. ■

Yi Liu is in the Department of Research and

Development, Intellikine, Inc., 10931 North Torrey

Pines Road, La Jolla, California 92037, USA.

e-mail: yi@intellikine.com

1. Pawson, T. & Scott, J. D. Trends Biochem. Sci. 30, 286–290

(2005).

2. Zhang, J., Yang, P. L. & Gray, N. S. Nature Rev. Cancer 9, 28–39 (2009).

3. Engel, M. et al. EMBO J. 25, 5469–5480 (2006).

4. Hindie, V. et al. Nature Chem. Biol. 5, 758–764

(2009).

5. Huse, M. & Kuriyan, J. Cell 109, 275–282 (2002).

6. Pellicena, P. & Kuriyan, J. Curr. Opin. Struct. Biol. 16, 702–709 (2006).

7. Biondi, R. M. et al. EMBO J. 19, 979–988 (2000).

of a key amino-acid residue at the base of the PIF-binding pocket. This conformational change deepens the pocket to better accom-modate the small molecule. The activator con-tains a charged carboxylate group (CO2 –) that mimics a phosphate group; the interactions of the carboxylate with the surrounding charged and polar amino-acid residues stabilize a con-formation of the PIF-binding pocket that is similar to the conformation induced by the binding of a phosphorylated peptide.

When PDK1 is in an inactive conformation, the Gly-rich loop, the activation loop and the α-C-helix are flexible and move around, and part of the activation loop is often disordered. Hindie et al.4 observed that binding of their activator in the PIF-binding pocket affects all three regions (Fig. 1b). First, local changes in the PIF-binding pocket are transmitted to the nearby Gly-rich loop and to the α-C-helix, causing some residues in these two regions to rearrange. The induced conformational changes in the small lobe then feed through to the activation loop in the large lobe. Finally, the disordered part of the activation loop is ordered and stabilized by interactions with the rearranged Gly-rich loop and α-C-helix. The crystal structure thus provides molecular details of how local changes at the PIF-binding pocket are transduced to the distant ATP-bind-ing site, and how they reorient the ATP site into an active conformation.

The authors found that the conformational changes they observed in the crystal structure of PDK1 upon activator binding also occur in the enzyme in solution. They monitored the structural changes at the ATP site using a fluorescent ATP analogue (TNP-ATP) whose fluorescence intensity increases when it binds to PDK1 at the ATP site. When the authors added an activator to the mix, however, the fluorescence intensity of TNP-ATP decreased, indicating that a conformational change had been produced at the ATP site that prevented TNP-ATP from binding.

In another experiment in solution, Hindie and colleagues4 monitored changes at the ATP site using hydrogen–deuterium exchange — an effect in which certain hydrogens in the ‘backbone’ of a protein are replaced with deu-terium (D) atoms upon exposure to D2O. When the authors studied a solution of a fragment of PDK1, they found that less hydrogen–deu-terium exchange occurred in the presence of an activator. Apart from the PIF-binding pocket, the sites most protected from exchange included the α-C-helix, the Gly-rich loop and the region just preceding the activation loop. A reduction in hydrogen–deuterium exchange indicates that a conformational change has occurred, so this result tallies well with the changes observed in the crystal structure, further demonstrating that the binding of a small-molecule activator at the PIF-binding pocket induces conformational changes in the ATP site of PDK1 and enhances the activity of the enzyme.

As the pièce de résistance, the authors

ASTROPHYSICS

Inner workings of a starSung-Chul Yoon

By borrowing a technique used by seismologists to investigate Earth’s interior, astronomers have probed the hitherto-unknown interior rotation profile of a white-dwarf star.

Just as the sound of a musical instrument is determined by the material from which it is made, so the period with which the bright-ness of a pulsating star oscillates depends on the specific structure of its interior. Using asteroseismology, astronomers can measure the temporal frequency spectra of pulsating stars — the seismograms of astronomers — to infer information about the stars’ otherwise unobservable interior. For example, rotating pulsating stars ‘sound’ different from non-rotating stars: because rotating stars cannot preserve their spherically symmetric shape during pulsation, their temporal frequency spectrum is marked by non-radial modes of pulsation. This property has allowed astrono-mers to measure the spin rates of pulsating white dwarfs1 — stellar remnants of relatively low-mass stars. But probing the internal rota-tion profiles of these stars is a different matter, and one that astronomers haven’t thus far got to grips with.

On page 501 of this issue, Charpinet et al.2

present the first evidence that a newly born white dwarf, dubbed PG 1159–035, rotates at the same rate for almost the entire depth of its body. The authors exploited non-radial-pul-sation observational data for this star using an innovative method. They constructed a white-dwarf theoretical model that mimics

PG 1159–035 and calculated the resulting pulsation frequencies considering several rota-tional profiles. By comparing the theoretical model with the data, they concluded that the observed non-radial oscillation frequencies of PG 1159-035 are in fine accord with the inter-pretation that the star rotates as a solid body. A newly born white dwarf (Fig. 1, overleaf) such as PG 1159–035 is thought to undergo rapid contraction as it loses thermal energy through the emission of neutrinos. Differential rota-tion, had it been detected by Charpinet and colleagues, would have indicated that the star’s dense inner core and outer envelope were con-tracting at different rates without rapid transfer of angular momentum between the two.

A typical white dwarf has about 60% of the Sun’s mass, but it is as small as Earth. Assuming that stars maintain their angular momentum throughout their evolution, a reduction in their size as they age would lead to a short spin period — between about 10 seconds and a few minutes — at the white-dwarf stage. But the rotation periods of all single white dwarfs observed to date1–3, including PG 1159–035, are longer than that, having periods of a few days. Together with Charpinet and colleagues’ finding that PG 1159–035 rotates as a solid body, this suggests that single white dwarfs have slow rotation periods because, in the

485

NATURE|Vol 461|24 September 2009 NEWS & VIEWS

479-488 News and Views MH IF.indd 485479-488 News and Views MH IF.indd 485 18/9/09 17:41:5518/9/09 17:41:55

© 2009 Macmillan Publishers Limited. All rights reserved