tionally analogous to control of feeding on carbon sources in budding yeast. In yeast, switching growth onto nonfermentable su- gars activates the AMPK homolog Snf1 through phosphorylation by upstream ki- nases that are homologous to LKB1. Activated Snf1 then phosphorylates the transcriptional regulator Mig1, which causes its trans- location from the nucleus to the cytoplasm, altering gene expression that allows sur- vival under the nutrient-poor environment. Metformin has been used clinically for decades (43). Our data provide genetic proof that AMPK activation is absolutely required for the glucose-lowering action of metformin in intact animals. The dele- tion of LKB1 in the liver did not impair AMPK activation in muscle, yet it elimi- nated the effect of metformin on serum glucose levels. This result suggests that in mice, metformin primarily decreases blood glucose concentrations by decreasing he- patic gluconeogenesis. LKB1 also acts as a tumor suppressor. Thus, increased CREB-dependent or SREBP- 1–dependent transcription could have a role in LKB1-dependent tumorigenesis. Constitutive activation of CREB transcription contributes to oncogenesis in human salivary tumors and clear cell sarcomas (46–48). These findings reinforce the emerging intimate relationship that exists between phys- iological control of metabolism and cancer. The mTOR, insulin, and LKB1 pathways rep- resent a fundamental eukaryotic network governing cell growth in response to environ- mental nutrients. Dysregulation of each con- tributes to both diabetes and cancer. References and Notes 1. B. B. Kahn, T. Alquier, D. Carling, D. G. Hardie, Cell Metab. 1, 15 (2005). 2. Y. Minokoshi et al., Nature 428, 569 (2004). 3. U. Andersson et al., J. Biol. Chem. 279, 12005 (2004). 4. E. K. Kim et al., J. Biol. Chem. 279, 19970 (2004). 5. S. A. Hawley et al., J. Biol. 2, 28 (2003). 6. R. J. Shaw et al., Proc. Natl. Acad. Sci. U.S.A. 101, 3329 (2004). 7. A. Woods et al., Curr. Biol. 13, 2004 (2003). 8. T. Yamauchi et al., Nat. Med. 8, 1288 (2002). 9. R. R. Banerjee et al., Science 303, 1195 (2004). 10. T. Hayashi, M. F. Hirshman, E. J. Kurth, W. W. Winder, L. J. Goodyear, Diabetes 47, 1369 (1998). 11. G. Zhou et al., J. Clin. Invest. 108, 1167 (2001). 12. L. G. Fryer, A. Parbu-Patel, D. Carling, J. Biol. Chem. 277, 25226 (2002). 13. H. Park et al., J. Biol. Chem. 277, 32571 (2002). 14. A. K. Saha et al., Biochem. Biophys. Res. Commun. 314, 580 (2004). 15. M. E. Cleasby et al., Diabetes 53, 3258 (2004). 16. R. Pold et al., Diabetes 54, 928 (2005). 17. S. A. Hawley et al., Cell Metab. 2, 9 (2005). 18. R. L. Hurley et al., J. Biol. Chem. 280, 29060 (2005). 19. A. Woods et al., Cell Metab. 2, 21 (2005). 20. M. J. Birnbaum, Mol. Cell 19, 289 (2005). 21. S. P. Hong, F. C. Leiper, A. Woods, D. Carling, M. Carlson, Proc. Natl. Acad. Sci. U.S.A. 100, 8839 (2003). 22. C. M. Sutherland et al., Curr. Biol. 13, 1299 (2003). 23. S. P. Hong, M. Momcilovic, M. Carlson, J. Biol. Chem. 280, 21804 (2005). 24. R. R. McCartney, E. M. Rubenstein, M. C. Schmidt, Curr. Genet. 47, 335 (2005). 25. N. Bardeesy et al., Nature 419, 162 (2002). 26. Materials and methods are available as supporting materials on Science Online. 27. J. Huard et al., Gene Ther. 2, 107 (1995). 28. K. Inoki, T. Zhu, K. L. Guan, Cell 115, 577 (2003). 29. R. J. Shaw et al., Cancer Cell 6, 91 (2004). 30. M. N. Corradetti, K. Inoki, N. Bardeesy, R. A. DePinho, K. L. Guan, Genes Dev. 18, 1533 (2004). 31. M. Foretz et al., Diabetes 54, 1331 (2005). 32. S. Herzig et al., Nature 413, 179 (2001). 33. S. H. Koo et al., Nat. Med. 10, 530 (2004). 34. J. C. Yoon et al., Nature 413, 131 (2001). 35. J. Lin, C. Handschin, B. M. Spiegelman, Cell Metab. 1, 361 (2005). 36. J. Rhee et al., Proc. Natl. Acad. Sci. U.S.A. 100, 4012 (2003). 37. P. Puigserver et al., Nature 423, 550 (2003). 38. L. Zhang, N. E. Rubins, R. S. Ahima, L. E. Greenbaum, K. H. Kaestner, Cell Metab. 2, 141 (2005). 39. J. Nakae, T. Kitamura, D. L. Silver, D. Accili, J. Clin. Invest. 108, 1359 (2001). 40. S. H. Koo et al., Nature 437, 1109 (2005). 41. R. A. Screaton et al., Cell 119, 61 (2004). 42. J. M. Lizcano et al., EMBO J. 23, 833 (2004). 43. L. A. Witters, J. Clin. Invest. 108, 1105 (2001). 44. M. Zang et al., J. Biol. Chem. 279, 47898 (2004). 45. K. Sakamoto et al., EMBO J. 24, 1810 (2005). 46. L. Wu et al., EMBO J. 24, 2391 (2005). 47. J. Zucman et al., Nat. Genet. 4, 341 (1993). 48. M. D. Conkright, M. Montminy, Trends Cell Biol. 15, 457 (2005). 49. R. J. Shaw, K. A. Lamia, data not shown. 50. We thank M. Loda for the FAS antibody; J. Luo, A. Shaywitz, and O. Peroni for technical advice; K. Cichowski for help with the manuscript; D. Gwinn and C. Mealmaker for technical assistance; and the Uni- versity of Iowa Gene Transfer Vector Core, supported in part by NIH and the Roy J. Carver Foundation, for adenoviral Cre preparations. This work was supported by grants GM056203, GM37828, and CA84313 from the NIH to L.C.C., M.M., and R.A.D., respectively. M.M. also was supported in part by the Hillblom Foundation. Supporting Online Material www.sciencemag.org/cgi/content/full/1120781/DC1 Materials and Methods Figs. S1 to S6 References 30 September 2005; accepted 9 November 2005 Published online 24 November 2005; 10.1126/science.1120781 Include this information when citing this paper. A Systems Model of Signaling Identifies a Molecular Basis Set for Cytokine-Induced Apoptosis Kevin A. Janes, 1,2 * John G. Albeck, 2,3 * Suzanne Gaudet, 2,3 Peter K. Sorger, 1,2,3 Douglas A. Lauffenburger, 1,2,3 Michael B. Yaffe 1,2,3 . Signal transduction pathways control cellular responses to stimuli, but it is un- clear how molecular information is processed as a network. We constructed a systems model of 7980 intracellular signaling events that directly links measure- ments to 1440 response outputs associated with apoptosis. The model accurately predicted multiple time-dependent apoptotic responses induced by a combina- tion of the death-inducing cytokine tumor necrosis factor with the prosurvival factors epidermal growth factor and insulin. By capturing the role of unsuspected autocrine circuits activated by transforming growth factor–a and interleukin-1a, the model revealed new molecular mechanisms connecting signaling to apoptosis. The model derived two groupings of intracellular signals that constitute fundamental dimensions (molecular ‘‘basis axes’’) within the apoptotic signaling network. Projection along these axes captures the entire measured apoptotic network, suggesting that cell survival is determined by signaling through this canonical basis set. Despite extensive molecular-level information on how external stimuli affect cell fate, there is minimal understanding of how such intra- cellular processing occurs at a systemwide level. Most extracellular Binputs[ initiate complex signaling patterns that propagate through an intracellular network to change the response Boutputs[ that determine a cell_s phenotype (1). Molecular signaling through this network is branched (fig. S1) and dy- namically interconnected to the molecular history of the previous inputs, signals, and outputs (2). Thus, we sought to develop a mathematical formalism to connect signals and outputs in such a way that cellular re- sponses could be predicted from molecular signaling patterns alone. To reduce the biological complexity of cellular signal processing, experimental sys- tems usually monitor changes in one defined extracellular stimulus (for example, a sol- uble cytokine) and one output response (such as apoptosis). However, physiological input stimuli are not processed in isolation, be- cause signaling networks constantly receive 1 Biological Engineering Division, 2 Center for Cell De- cision Processes, 3 Center for Cancer Research, De- partment of Biology, Massachusetts Institute of Technology, Cambridge, MA 02139, USA. *These authors contributed equally to this work. .To whom correspondence should be addressed. E-mail: myaffe@mit.edu R ESEARCH A RTICLES 9 DECEMBER 2005 VOL 310 SCIENCE www.sciencemag.org 1646