Mutations in PIK3CA, which encodes the p110α subunit of the insulin-activated phosphatidylinositol-3 kinase (PI3K), and loss of function mutations in PTEN, which encodes a phosphatase that degrades the phosphoinositide lipids generated by PI3K, are among the most frequent events in human cancers1,2. However, pharmacological inhibition of PI3K has resulted in variable clinical responses, raising the possibility of an inherent mechanism of resistance to treatment. As p110α mediates virtually all cellular responses to insulin, targeted inhibition of this enzyme disrupts glucose metabolism in multiple tissues. For example, blocking insulin signalling promotes glycogen breakdown in the liver and prevents glucose uptake in the skeletal muscle and adipose tissue, resulting in transient hyperglycaemia within a few hours of PI3K inhibition. The effect is usually transient because compensatory insulin release from the pancreas (insulin feedback) restores normal glucose homeostasis3. However, the hyperglycaemia may be exacerbated or prolonged in patients with any degree of insulin resistance and, in these cases, necessitates discontinuation of therapy3,4,5,6. We hypothesized that insulin feedback induced by PI3K inhibitors may reactivate the PI3K–mTOR signalling axis in tumours, thereby compromising treatment effectiveness7,8. Here we show, in several model tumours in mice, that systemic glucose–insulin feedback caused by targeted inhibition of this pathway is sufficient to activate PI3K signalling, even in the presence of PI3K inhibitors. This insulin feedback can be prevented using dietary or pharmaceutical approaches, which greatly enhance the efficacy/toxicity ratios of PI3K inhibitors. These findings have direct clinical implications for the multiple p110α inhibitors that are in clinical trials and provide a way to increase treatment efficacy for patients with many types of tumour.

Access optionsAccess options

Rent or Buy article

Get time limited or full article access on ReadCube.


All prices are NET prices.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Change history

  • 29 August 2018

    In this Letter, author Xing Du was incorrectly listed as Du Xing; this has been corrected online.


  1. 1.

    Kandoth, C. et al. Mutational landscape and significance across 12 major cancer types. Nature 502, 333–339 (2013).

  2. 2.

    Millis, S. Z., Ikeda, S., Reddy, S., Gatalica, Z. & Kurzrock, R. Landscape of phosphatidylinositol-3-kinase pathway alterations across 19784 diverse solid tumors. JAMA Oncol. 2, 1565–1573 (2016).

  3. 3.

    Bendell, J. C. et al. Phase I, dose-escalation study of BKM120, an oral pan-class I PI3K inhibitor, in patients with advanced solid tumors. J. Clin. Oncol. 30, 282–290 (2012).

  4. 4.

    Juric, D. et al. Phase I dose-escalation study of taselisib, an oral PI3K inhibitor, in patients with advanced solid tumors. Cancer Discov. 7, 704–715 (2017).

  5. 5.

    Patnaik, A. et al. First-in-human phase I study of copanlisib (BAY 80-6946), an intravenous pan-class I phosphatidylinositol 3-kinase inhibitor, in patients with advanced solid tumors and non-Hodgkin’s lymphomas. Ann. Oncol. 27, 1928–1940 (2016).

  6. 6.

    Mayer, I. A. et al. A Phase Ib study of alpelisib (BYL719), a PI3Kα-specific inhibitor, with letrozole in ER+/HER2 metastatic breast cancer. Clin. Cancer Res. 23, 26–34 (2017).

  7. 7.

    Hopkins, B. D., Goncalves, M. D. & Cantley, L. C. Obesity and cancer mechanisms: cancer metabolism. J. Clin. Oncol. 34, 4277–4283 (2016).

  8. 8.

    Fruman, D. A. et al. The PI3K pathway in human disease. Cell 170, 605–635 (2017).

  9. 9.

    Belardi, V., Gallagher, E. J., Novosyadlyy, R. & LeRoith, D. Insulin and IGFs in obesity-related breast cancer. J. Mammary Gland Biol. Neoplasia 18, 277–289 (2013).

  10. 10.

    Gallagher, E. J. & LeRoith, D. Minireview: IGF, insulin, and cancer. Endocrinology 152, 2546–2551 (2011).

  11. 11.

    Klil-Drori, A. J., Azoulay, L. & Pollak, M. N. Cancer, obesity, diabetes, and antidiabetic drugs: is the fog clearing? Nat. Rev. Clin. Oncol. 14, 85–99 (2017).

  12. 12.

    Ma, J. et al. A prospective study of plasma c-peptide and colorectal cancer risk in men. J. Natl. Cancer Inst. 96, 546–553 (2004).

  13. 13.

    Xu, J. et al. Association between markers of glucose metabolism and risk of colorectal cancer. BMJ Open 6, e011430 (2016).

  14. 14.

    Ma, J. et al. Prediagnostic body-mass index, plasma c-peptide concentration, and prostate cancer-specific mortality in men with prostate cancer: a long-term survival analysis. Lancet Oncol. 9, 1039–1047 (2008).

  15. 15.

    Olive, K. P. et al. Inhibition of Hedgehog signaling enhances delivery of chemotherapy in a mouse model of pancreatic cancer. Science 324, 1457–1461 (2009).

  16. 16.

    Pauli, C. et al. Personalized in vitro and in vivo cancer models to guide precision medicine. Cancer Discov. 7, 462–477 (2017).

  17. 17.

    Komoroski, B. et al. Dapagliflozin, a novel, selective SGLT2 inhibitor, improved glycemic control over 2 weeks in patients with type 2 diabetes mellitus. Clin. Pharmacol. Ther. 85, 513–519 (2009).

  18. 18.

    Demin, O., Jr, Yakovleva, T., Kolobkov, D. & Demin, O. Analysis of the efficacy of SGLT2 inhibitors using semi-mechanistic model. Front. Pharmacol. 5, 218 (2014).

  19. 19.

    Pollak, M. Metformin and other biguanides in oncology: advancing the research agenda. Cancer Prev. Res. (Phila.) 3, 1060–1065 (2010).

  20. 20.

    Pollak, M. Potential applications for biguanides in oncology. J. Clin. Invest. 123, 3693–3700 (2013).

  21. 21.

    Saura, C. et al. Phase Ib study of buparlisib plus trastuzumab in patients with HER2-positive advanced or metastatic breast cancer that has progressed on trastuzumab-based therapy. Clin. Cancer Res. 20, 1935–1945 (2014).

  22. 22.

    Juvekar, A. et al. Combining a PI3K inhibitor with a PARP inhibitor provides an effective therapy for BRCA1-related breast.1158/2159–8290.CD-11–0336

  23. 23.

    Puchalska, P. & Crawford, P. A. Multi-dimensional roles of ketone bodies in fuel metabolism, signaling, and therapeutics. Cell Metab. 25, 262–284 (2017).

  24. 24.

    Sampaio, L. P. Ketogenic diet for epilepsy treatment. Arq. Neuropsiquiatr. 74, 842–848 (2016).

  25. 25.

    Pauli, C. et al. An emerging role for cytopathology in precision oncology. Cancer Cytopathol. 124, 167–173 (2016).

  26. 26.

    Lee, C., Kim, J. S. & Waldman, T. PTEN gene targeting reveals a radiation-induced size checkpoint in human cancer cells. Cancer Res. 64, 6906–6914 (2004).

  27. 27.

    Pelossof, R. et al. Prediction of potent shRNAs with a sequential classification algorithm. Nat. Biotechnol. 35, 350–353 (2017).

  28. 28.

    Fellmann, C. et al. An optimized microRNA backbone for effective single-copy RNAi. Cell Reports 5, 1704–1713 (2013).

  29. 29.

    Douris, N. et al. Adaptive changes in amino acid metabolism permit normal longevity in mice consuming a low-carbohydrate ketogenic diet. Biochim. Biophys. Acta 1852, 2056–2065 (2015).

Download references


This work was supported by NIH grant R35 CA197588 (L.C.C.), R01 GM041890 (L.C.C.), U54 U54CA210184 (L.C.C.), Breast Cancer Research Foundation (L.C.C.) and the Jon and Mindy Gray Foundation (L.C.C.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. We appreciate the help of the small animal imaging core at MSKCC for assistance with FDG-PET imaging and the Columbia Irving Cancer Center Flow Core Facility, funded in part through Center Grant P30CA013696.

Reviewer information

Nature thanks V. Longo, M. Pollak, C. Rask-Madsen and the other anonymous reviewer(s) for their contribution to the peer review of this work.

Author information


  1. Meyer Cancer Center, Weill Cornell Medicine, New York, NY, USA

    • Benjamin D. Hopkins
    • , Diana G. Wang
    • , David Wu
    • , Solomon C. Amadiume
    • , Marcus D. Goncalves
    • , Cindy Hodakoski
    • , Mark R. Lundquist
    • , Rohan Bareja
    • , Andrea Sboner
    • , Himisha Beltran
    •  & Lewis C. Cantley
  2. Institute of Pathology and Molecular Pathology, University Hospital Zurich, Zurich, Switzerland

    • Chantal Pauli
  3. Englander Institute for Precision Medicine, Weill Cornell Medicine-New York Presbyterian Hospital, New York, NY, USA

    • Chantal Pauli
    • , Rohan Bareja
    • , Andrea Sboner
    • , Himisha Beltran
    •  & Mark A. Rubin
  4. Department of Medicine, Division of Hematology and Oncology, Columbia University Medical Center and New York Presbyterian Hospital, New York, NY, USA

    • Xing Du
    • , Yan Ma
    • , Emily M. Harris
    •  & Siddhartha Mukherjee
  5. Weill Cornell Medicine/Rockefeller University/Sloan Kettering Tri-Institutional MD-PhD Program, New York, NY, USA

    • Diana G. Wang
  6. Weill Cornell Graduate School of Medical Sciences, New York, NY, USA

    • Xiang Li
  7. Division of Endocrinology, Department of Medicine, Weill Cornell Medicine, New York, NY, USA

    • Marcus D. Goncalves
  8. Institute for Computational Biomedicine, Weill Cornell Medicine, New York, NY, USA

    • Rohan Bareja
    •  & Andrea Sboner
  9. Department of Pathology, Weill Cornell Medicine, New York, NY, USA

    • Andrea Sboner
  10. Department of Medicine, Division of Hematology and Medical Oncology, Weill Cornell Medicine, New York, NY, USA

    • Himisha Beltran
  11. Department of Biomedical Research and the Center for Precision Medicine, University of Bern and the Inselspital, Bern, Switzerland

    • Mark A. Rubin


  1. Search for Benjamin D. Hopkins in:

  2. Search for Chantal Pauli in:

  3. Search for Xing Du in:

  4. Search for Diana G. Wang in:

  5. Search for Xiang Li in:

  6. Search for David Wu in:

  7. Search for Solomon C. Amadiume in:

  8. Search for Marcus D. Goncalves in:

  9. Search for Cindy Hodakoski in:

  10. Search for Mark R. Lundquist in:

  11. Search for Rohan Bareja in:

  12. Search for Yan Ma in:

  13. Search for Emily M. Harris in:

  14. Search for Andrea Sboner in:

  15. Search for Himisha Beltran in:

  16. Search for Mark A. Rubin in:

  17. Search for Siddhartha Mukherjee in:

  18. Search for Lewis C. Cantley in:


B.D.H., S.M. and L.C.C. conceived of the project. B.D.H., C.P., X.D., Y.M. and D.W. performed the mouse experiments. S.C.A., C.P., B.D.H., E.M.H. and X.L. did the culture assays. S.C.A. performed immunoblotting. B.D.H., C.H. and M.D.G. assessed the impact of treatments on cellular and systemic metabolism. D.G.W. and S.C.A. cloned and validated the IR knockdowns. M.R.L. and R.B. performed the data analysis. C.P., A.S., H.B., M.A.R., L.C.C., S.M., B.D.H. and R.B. assisted with implementation of patient-derived models. All authors assisted with data interpretation and contributed to the writing and editing of the manuscript.

Competing interests

L.C.C. is a founder and member of the board of directors of Agios Pharmaceuticals and is a founder and receives research support from Petra Pharmaceuticals. S.M. is a founder and on the board of Vor Pharmaceuticals. These companies are developing novel therapies for cancer. All other authors declare no competing interests.

Corresponding authors

Correspondence to Siddhartha Mukherjee or Lewis C. Cantley.

Extended data figures and tables

  1. Extended Data Fig. 1 Blood glucose and c-peptide levels after treatment with agents that target the PI3K pathway.

    a, b, Mean ± s.d. blood glucose levels over time where time 0 is the time of treatment with the indicated inhibitor. n = 5 and 3 mice per arm for a and b, respectively. c, d, Mean ± s.d. c-peptide levels from mice in a and b taken 240 and 180 min after treatment with indicated inhibitors. c, n = 5 for vehicle, BKM120, GDC-0941, and GDC-0980; n = 4 for BEZ235; n = 3 for RAD001. D, n = 3 mice per arm. As a surrogate for total insulin release, c-peptide levels show that the PI3K and IGFR/INSR inhibitors markedly increase insulin release. In all cases, compounds that caused acute increases in blood glucose also increased serum insulin. Source data

  2. Extended Data Fig. 2 Effect of feedback levels of insulin observed in Fig. 1 on BKM120 efficacy in vitro.

    a, Proliferation in minimal growth medium of cells whose growth is partially rescued by the addition of the observed feedback levels of insulin (10 ng ml–1) induced by BKM120 in mice. n = 3 biologically independent samples per arm, mean ± s.d. number of cells. b, Cell viability assay demonstrating the effects of feedback levels of insulin on two patient-derived organoid cultures (PtA and PtB) being treated in a dose response with BKM120 as measured by cell titre-glo at 96 h. n = 3 biologically independent samples per treatment. c, Proliferation in minimal growth medium of mouse TNBC cells treated with PI3K inhibitors partially rescued by the addition of the observed feedback levels of insulin induced by BKM120 in mice (Fig. 1). n = 6 biologically independent samples per treatment, mean ± s.d. d, e, Proliferation of HCT116-neo cells (d) and HCT116 PTEN knockout (KO) cells (e) with and without treatment with physiologically observed levels of insulin (10 ng ml–1) and treatment with the clinically relevant PI3K inhibitors GDC-0032 and BYL-719. n = 4 biologically independent samples per treatment, mean ± s.d. confluence. f, g, Proliferation of DLD1-Neo cells (f) and DLD-1 PTEN knockout cells (g) under the same treatment conditions as in d and e. Of note, the loss of PTEN in these isogenic sets of colon cancer lines does not uniformly alter the response to insulin in the setting of PI3K inhibition. In the context of PTEN loss, physiological levels of insulin can restore normal proliferation in HCT116 cells despite the presence of PI3K inhibitors. n = 4 biologically independent samples per treatment, mean ± s.d. confluence.

  3. Extended Data Fig. 3 KPC K8484 allografts treated with PI3K inhibitors with or without supplementary approaches to target systemic insulin feedback.

    a, Mean ± s.d. blood glucose of mice from Fig. 3e–g treated with control diet, ketogenic diet, metformin (250 mg kg–1), or canagliflozin (SGLT2i; 6 mg kg–1), after the first dose of BYL-719 (45 mg kg). n = 5 animals per arm. b, Volumes of tumours treated with the metabolic modifying agents shown in Fig. 3 without PI3K inhibitors. n = 10 tumours per arm for vehicle, metformin, and ketogenic diet; n = 8 tumours per arm for SGLT2i. c, Mean tumour volumes (lines) with scatter (points) for each of these treatment cohorts. n = 10 tumours per arm for BYL-719, BYL-719 + metformin, and BYL-719 + ketogenic diet; n = 9 tumours for BYL-719 + SGLT2i. d, Mean ± s.d. tumour volumes from an independent experiment with mice (n = 4 mice per arm) treated daily with BKM120 with or without 6 mg kg–1 canagliflozin administered 60 min before PI3K treatment, so that peak SGLT2 inhibition is aligned with peak blood glucose levels after PI3K inhibitor treatment. e, f, Mean ± s.d. blood ketones (e) and triglycerides (f) as determined by calorimetric assay of mice shown in Fig. 3a–d after a single treatment with BKM120 with or without pretreatment with metformin, canagliflozin, or the ketogenic diet. n = 5 mice per arm. Source data

  4. Extended Data Fig. 4 Role of insulin receptor inhibition in the observed changes in tumour response.

    a, Western blot of cell lysates from K8484 cells used to generate xenografts in Fig. 4a after 36 treatments with doxycycline to induce the shRenilla and shIR hairpins as indicated. Similar results were observed in two independent experiments. b, Tumour volumes of individual mice allografted with KPC-K8484 tumours as measured by calipers over time. n = 4, 5, 4, 4, 5, 5 and 5 for vehicle, BKM120, BKM120 + ketogenic diet, BKM120 + OSI-906, OSI-906, OSI-906 + ketogenic diet, and ketogenic diet, respectively. c, Survival curves of mice in b. d, e, Mean ± s.d. blood glucose (d) and c-peptide (e) from these mice 240 min after respective treatments. Two of the glucose measurements in the OSI-906 and BKM120 were beyond the range of the detector (>600 mg dl–1). f, Masses of individual mice over the course of treatment. As has been previously published, mice lose 10–20% of their mass upon initiation of a ketogenic diet29. g, Similar to the data for the tumours in a, both OSI-906, a INSR/IGFR inhibitor, and GDC-0032 showed greater anti-tumour efficacy in PIK3CA + MYC mutant mouse breast tumour allografts, ES-278, grown in wild-type c57/bl6 mice fed a ketogenic diet. n = 5 tumours per arm. Points depict mean ± s.d. tumour volume. h, Mean ± s.d. tumour volumes of wild-type c57/bl6 mice bearing KPC allografted tumours as measured by calipers over time. Mice were treated as indicated with combinations of BYL-719, the ketogenic diet, or insulin as in Fig. 4b. Mice in the ketogenic + BYL719 + insulin cohort lost >20% of their body mass over the 1 week of treatment so the experiment was terminated at day 7. n = 6, 4, 4, 6, 6 for vehicle, BYL-719, ketogenic diet, BYL-719 + ketogenic diet, and BYL-719 + ketogenic diet + insulin, respectively. Source data

  5. Extended Data Fig. 5 Effect of PI3K inhibitor treatments on PDX model of bladder cancer and syngeneic allograft models of PIK3CA mutant breast cancer.

    a, Graph of tumour growth over time of a PDX from a patient with bladder cancer (Patient C) treated with the pan-PI3K inhibitor GDC-0941 or the β-sparing inhibitor GDC-0032, alone or with a ketogenic diet. Lines indicate mean tumour volume of each treatment group, points indicate individual tumour volumes over time. n = 5 tumours per arm. b, Mean ± s.d. tumour mass at the time of removal on day 12. c, d, Mean ± s.d. tumour growth over time (c) and tumour mass at time of removal (d) from mice with orthotopic allografts of a PIK3CA (H1047R) mutant mouse breast cancer, ES272, treated as indicated with BKM120 alone or in combination with a ketogenic diet. n = 4, 5, 5 tumours per arm for vehicle, BKM120 and BKM120 + ketogenic diet, respectively. e, Mean ± s.d. mass of mice over time. Source data

  6. Extended Data Fig. 6 Effect of copanilisib with or without ketogenic diet on growth of KPC tumour model K8082 grown in the flank of wild-type c57/bl6 mice.

    a, Survival curves for mice with KPC K8082 allografts grown in the flank and treated as indicated with BAY 80-6946 alone or combined with pretreatment with a ketogenic diet as indicated (P value comparing BAY 80-6946 with the combination of BAY 80-6946 with ketogenic diet was 0.0019 by Mantel–Cox log-rank test). n = 5 mice per arm for vehicle, BAY 80-6946, and BAY 80-6946 + ketogenic diet; n = 4 for ketogenic diet alone. b, Volume of each tumour in this cohort plotted individually. c, d, Mean ± s.d. blood glucose (c) and c-peptide (d) in mice in b, c, 240 min after treatment. e, Mass of these mice over time on treatment. Tumours were allowed to grow until their diameters were >0.6 cm before the initiation of treatment. Source data

  7. Extended Data Fig. 7 Effect of BKM120 + ketogenic diet on a syngeneic model of AML.

    a, IVIS images of AML burden (as reported by DS-red) in over time. The group labelled BKM120 plus ketogenic diet were pre-treated with a ketogenic diet. n = 7 mice per arm. b, Survival curves of mice from a with additional mice to evaluate pre-treatment versus co-treatment with the ketogenic diet in the syngeneic model of AML treated with BKM120 alone or in combination with a ketogenic diet. Individual lines are shown for initiation of ketogenic diet before (pre) or at the same time as the initiation of BKM120 treatment (co); in both cases, BKM120 efficacy is significantly enhanced by the addition of the ketogenic diet (P = 0.0142 and 0.0316 by Gehan–Breslow Wilcoxon test for pre and co compared to BKM alone, respectively). *Mice that were euthanized owing to paralysis resulting from AML infiltrating the CNS, rather than deaths typically seen in these mice due to tumour burden. Of note, the mice in the BKM + ketogenic diet group were frequently euthanized due to paralysis, but this was not frequently a cause of mortality in the other treatment groups. n = 6, 6, 5, 7, 5, 7 mice per arm for vehicle, pre-ketogenic diet, co-ketogenic diet, BKM120, co-ketogenic diet + BKM120, and pre-ketogenic diet + BKM120, respectively. c, d, Disease burden of AML as measured by per cent DS-red positive AML cells in bone marrow (c) and spleen weight across the treatment groups (D) (pre-treated with ketogenic diet). Data are mean ± s.d. n = 5, 4, 4, 4 mice per arm for vehicle, pre-ketogenic diet, BKM120, and pre-ketogenic diet + BKM120, respectively. e, Measurement of AML burden in mice that were pretreated with BKM120 and/or a ketogenic diet to demonstrate that the effects observed in the AML studies are not the result of implantation issues related to the pretreatment. Data are mean ± s.d. n = 4, 4, 4, 5 mice per arm for vehicle, pre-ketogenic diet, BKM120, and pre-ketogenic diet + BKM120, respectively. f, Images of mice treated as indicated with BKM120 and ketogenic diet where the diet and BKM120 therapy were initiated on the same day (co-treatment). n = 5 mice per arm. Source data

Supplementary information

  1. Supplementary Figures

    This file contains Supplementary Tables 2-4 and Supplementary Figures 1-2

  2. Reporting Summary

  3. Supplementary Tables

    This file contains Supplementary Tables 1-3. Supplementary Table 1 contains the nutritional content of normal and ketogenic diets used in these studies, Supplementary Table 2 contains the cell lines used and Supplementary Table 3 contains the targeted inhibitors used

Source data

About this article

Publication history







By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.