Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Article
  • Published:

Purine synthesis promotes maintenance of brain tumor initiating cells in glioma

Nature Neuroscience volume 20, pages 661–673 (2017)Cite this article

Subjects

Abstract

Brain tumor initiating cells (BTICs), also known as cancer stem cells, hijack high-affinity glucose uptake active normally in neurons to maintain energy demands. Here we link metabolic dysregulation in human BTICs to a nexus between MYC and de novo purine synthesis, mediating glucose-sustained anabolic metabolism. Inhibiting purine synthesis abrogated BTIC growth, self-renewal and in vivo tumor formation by depleting intracellular pools of purine nucleotides, supporting purine synthesis as a potential therapeutic point of fragility. In contrast, differentiated glioma cells were unaffected by the targeting of purine biosynthetic enzymes, suggesting selective dependence of BTICs. MYC coordinated the control of purine synthetic enzymes, supporting its role in metabolic reprogramming. Elevated expression of purine synthetic enzymes correlated with poor prognosis in glioblastoma patients. Collectively, our results suggest that stem-like glioma cells reprogram their metabolism to self-renew and fuel the tumor hierarchy, revealing potential BTIC cancer dependencies amenable to targeted therapy.

This is a preview of subscription content, access via your institution

Access options

Buy this article

  • Purchase on Springer Link
  • Instant access to full article PDF
Buy now

Prices may be subject to local taxes which are calculated during checkout

Figure 1: BTICs upregulate de novo purine synthesis.
Figure 2: Genomic profiling reveals specific upregulation of de novo purine synthesis pathway in BTICs.
Figure 3: Glycolytic activity modulates de novo purine synthesis in BTICs.
Figure 4: Transcription factor MYC regulates de novo purine synthesis pathway enzymes in BTICs.
Figure 5: ADSL regulates BTIC growth and self-renewal.
Figure 6: GMPS regulates BTIC growth and self-renewal.
Figure 7: PRPS1 regulates BTIC growth and self-renewal.
Figure 8: Overexpression of purine synthesis genes is associated with poor brain tumor patient survival.

Similar content being viewed by others

Accession codes

Primary accessions

Sequence Read Archive

Referenced accessions

Gene Expression Omnibus

References

  1. Stupp, R. et al. Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised phase III study: 5-year analysis of the EORTC-NCIC trial. Lancet Oncol. 10, 459–466 (2009).

    CAS  PubMed  Google Scholar 

  2. Singh, S.K. et al. Identification of human brain tumour initiating cells. Nature 432, 396–401 (2004).

    CAS  PubMed  Google Scholar 

  3. Gage, F.H. & Temple, S. Neural stem cells: generating and regenerating the brain. Neuron 80, 588–601 (2013).

    CAS  PubMed  Google Scholar 

  4. Bao, S. et al. Glioma stem cells promote radioresistance by preferential activation of the DNA damage response. Nature 444, 756–760 (2006).

    Article  CAS  PubMed  Google Scholar 

  5. Liu, G. et al. Analysis of gene expression and chemoresistance of CD133+ cancer stem cells in glioblastoma. Mol. Cancer 5, 67 (2006).

    PubMed  PubMed Central  Google Scholar 

  6. Bao, S. et al. Stem cell-like glioma cells promote tumor angiogenesis through vascular endothelial growth factor. Cancer Res. 66, 7843–7848 (2006).

    CAS  PubMed  Google Scholar 

  7. Cheng, L. et al. Glioblastoma stem cells generate vascular pericytes to support vessel function and tumor growth. Cell 153, 139–152 (2013).

    CAS  PubMed  PubMed Central  Google Scholar 

  8. Pavlova, N.N. & Thompson, C.B. The emerging hallmarks of cancer metabolism. Cell Metab. 23, 27–47 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  9. Vander Heiden, M.G., Cantley, L.C. & Thompson, C.B. Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science 324, 1029–1033 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  10. Ward, P.S. & Thompson, C.B. Metabolic reprogramming: a cancer hallmark even Warburg did not anticipate. Cancer Cell 21, 297–308 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  11. Yan, H. et al. IDH1 and IDH2 mutations in gliomas. N. Engl. J. Med. 360, 765–773 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  12. Turcan, S. et al. IDH1 mutation is sufficient to establish the glioma hypermethylator phenotype. Nature 483, 479–483 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  13. Flavahan, W.A. et al. Brain tumor initiating cells adapt to restricted nutrition through preferential glucose uptake. Nat. Neurosci. 16, 1373–1382 (2013).

    CAS  PubMed  PubMed Central  Google Scholar 

  14. Li, Z. et al. Hypoxia-inducible factors regulate tumorigenic capacity of glioma stem cells. Cancer Cell 15, 501–513 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  15. Hjelmeland, A.B. et al. Acidic stress promotes a glioma stem cell phenotype. Cell Death Differ. 18, 829–840 (2011).

    CAS  PubMed  Google Scholar 

  16. Panopoulos, A.D. et al. The metabolome of induced pluripotent stem cells reveals metabolic changes occurring in somatic cell reprogramming. Cell Res. 22, 168–177 (2012).

    CAS  PubMed  Google Scholar 

  17. Masui, K. et al. Glucose-dependent acetylation of Rictor promotes targeted cancer therapy resistance. Proc. Natl. Acad. Sci. USA 112, 9406–9411 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

  18. Derr, R.L. et al. Association between hyperglycemia and survival in patients with newly diagnosed glioblastoma. J. Clin. Oncol. 27, 1082–1086 (2009).

    PubMed  PubMed Central  Google Scholar 

  19. Brennan, C.W. et al. The somatic genomic landscape of glioblastoma. Cell 155, 462–477 (2013).

    CAS  PubMed  PubMed Central  Google Scholar 

  20. Cancer Genome Atlas Research Network. Comprehensive genomic characterization defines human glioblastoma genes and core pathways. Nature 455, 1061–1068 (2008).

  21. Son, M.J., Woolard, K., Nam, D.H., Lee, J. & Fine, H.A. SSEA-1 is an enrichment marker for tumor-initiating cells in human glioblastoma. Cell Stem Cell 4, 440–452 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  22. Liu, X. et al. Discrete mechanisms of mTOR and cell cycle regulation by AMPK agonists independent of AMPK. Proc. Natl. Acad. Sci. USA 111, E435–E444 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  23. Guo, D. et al. The AMPK agonist AICAR inhibits the growth of EGFRvIII-expressing glioblastomas by inhibiting lipogenesis. Proc. Natl. Acad. Sci. USA 106, 12932–12937 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  24. Altman, B.J., Stine, Z.E. & Dang, C.V. From Krebs to clinic: glutamine metabolism to cancer therapy. Nat. Rev. Cancer 16, 619–634 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  25. Wang, J. et al. c-Myc is required for maintenance of glioma cancer stem cells. PLoS One 3, e3769 (2008).

    PubMed  PubMed Central  Google Scholar 

  26. Zheng, H. et al. p53 and Pten control neural and glioma stem/progenitor cell renewal and differentiation. Nature 455, 1129–1133 (2008).

    CAS  PubMed  PubMed Central  Google Scholar 

  27. Stine, Z.E., Walton, Z.E., Altman, B.J., Hsieh, A.L. & Dang, C.V. MYC, metabolism, and cancer. Cancer Discov. 5, 1024–1039 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

  28. Suvà, M.L. et al. Reconstructing and reprogramming the tumor-propagating potential of glioblastoma stem-like cells. Cell 157, 580–594 (2014).

    PubMed  PubMed Central  Google Scholar 

  29. Venere, M. et al. Therapeutic targeting of constitutive PARP activation compromises stem cell phenotype and survival of glioblastoma-initiating cells. Cell Death Differ. 21, 258–269 (2014).

    CAS  PubMed  Google Scholar 

  30. Freije, W.A. et al. Gene expression profiling of gliomas strongly predicts survival. Cancer Res. 64, 6503–6510 (2004).

    CAS  PubMed  Google Scholar 

  31. Madhavan, S. et al. Rembrandt: helping personalized medicine become a reality through integrative translational research. Mol. Cancer Res. 7, 157–167 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  32. Phillips, H.S. et al. Molecular subclasses of high-grade glioma predict prognosis, delineate a pattern of disease progression, and resemble stages in neurogenesis. Cancer Cell 9, 157–173 (2006).

    CAS  PubMed  Google Scholar 

  33. Lane, A.N. & Fan, T.W. Regulation of mammalian nucleotide metabolism and biosynthesis. Nucleic Acids Res. 43, 2466–2485 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

  34. Wood, T. Physiological functions of the pentose phosphate pathway. Cell Biochem. Funct. 4, 241–247 (1986).

    CAS  PubMed  Google Scholar 

  35. Boros, L.G. et al. Transforming growth factor β2 promotes glucose carbon incorporation into nucleic acid ribose through the nonoxidative pentose cycle in lung epithelial carcinoma cells. Cancer Res. 60, 1183–1185 (2000).

    CAS  PubMed  Google Scholar 

  36. Tong, X., Zhao, F. & Thompson, C.B. The molecular determinants of de novo nucleotide biosynthesis in cancer cells. Curr. Opin. Genet. Dev. 19, 32–37 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  37. Hosios, A.M. et al. Amino acids rather than glucose account for the majority of cell mass in proliferating mammalian cells. Dev. Cell 36, 540–549 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  38. Ben-Sahra, I., Hoxhaj, G., Ricoult, S.J., Asara, J.M. & Manning, B.D. mTORC1 induces purine synthesis through control of the mitochondrial tetrahydrofolate cycle. Science 351, 728–733 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  39. Hensley, C.T. et al. Metabolic heterogeneity in human lung tumors. Cell 164, 681–694 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  40. Wang, W. et al. The phosphatidylinositol 3-kinase/Akt cassette regulates purine nucleotide synthesis. J. Biol. Chem. 284, 3521–3528 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  41. Eyler, C.E. et al. Brain cancer stem cells display preferential sensitivity to Akt inhibition. Stem Cells 26, 3027–3036 (2008).

    CAS  PubMed  PubMed Central  Google Scholar 

  42. Vander Heiden, M.G. Targeting cancer metabolism: a therapeutic window opens. Nat. Rev. Drug Discov. 10, 671–684 (2011).

    CAS  PubMed  Google Scholar 

  43. Brazelton, T.R. & Morris, R.E. Molecular mechanisms of action of new xenobiotic immunosuppressive drugs: tacrolimus (FK506), sirolimus (rapamycin), mycophenolate mofetil and leflunomide. Curr. Opin. Immunol. 8, 710–720 (1996).

    CAS  PubMed  Google Scholar 

  44. D'Cruz, D.P., Khamashta, M.A. & Hughes, G.R. Systemic lupus erythematosus. Lancet 369, 587–596 (2007).

    PubMed  Google Scholar 

  45. Sintchak, M.D. et al. Structure and mechanism of inosine monophosphate dehydrogenase in complex with the immunosuppressant mycophenolic acid. Cell 85, 921–930 (1996).

    PubMed  Google Scholar 

  46. Sidwell, R.W. et al. Broad-spectrum antiviral activity of virazole: 1-β-D-ribofuranosyl-1,2,4-triazole-3-carboxamide. Science 177, 705–706 (1972).

    CAS  PubMed  Google Scholar 

  47. Wray, S.K., Gilbert, B.E., Noall, M.W. & Knight, V. Mode of action of ribavirin: effect of nucleotide pool alterations on influenza virus ribonucleoprotein synthesis. Antiviral Res. 5, 29–37 (1985).

    CAS  PubMed  Google Scholar 

  48. Markland, W., McQuaid, T.J., Jain, J. & Kwong, A.D. Broad-spectrum antiviral activity of the IMP dehydrogenase inhibitor VX-497: a comparison with ribavirin and demonstration of antiviral additivity with alpha interferon. Antimicrob. Agents Chemother. 44, 859–866 (2000).

    CAS  PubMed  PubMed Central  Google Scholar 

  49. 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).

    CAS  PubMed  Google Scholar 

  50. Wen, P.Y., Lee, E.Q., Reardon, D.A., Ligon, K.L. & Alfred Yung, W.K. Current clinical development of PI3K pathway inhibitors in glioblastoma. Neuro-oncol. 14, 819–829 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  51. Fang, X. et al. The zinc finger transcription factor ZFX is required for maintaining the tumorigenic potential of glioblastoma stem cells. Stem Cells 32, 2033–2047 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  52. Xie, Q. et al. Mitochondrial control by DRP1 in brain tumor initiating cells. Nat. Neurosci. 18, 501–510 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

We appreciate mass spectrometry analysis by R. Zhang, flow cytometry assistance by C. Shemo and S. O'Bryant, the glioblastoma tissue provided by M. McGraw and the Cleveland Clinic Tissue Procurement Service, and the IN528 model from I. Nakano at Ohio State University. We thank T. Roberts, J. Suh, G. Narla and members of the Rich laboratory for discussions. This work was supported by funding from National Institutes of Health grants CA197718, CA154130, CA169117, CA171652, NS087913 and NS089272 (J.N.R.), CA184090, NS091080 (S.B.) and CA168997, CA193256, CA201963 (J.W.L.); the James S. McDonnell Foundation (J.N.R.); the Research Programs Committees of the Cleveland Clinic (J.N.R. and K.Y.); Clinical and Translational Science Collaborative of Cleveland grant UL1TR000439 from the National Center for Advancing Translational Sciences (J.N.R. and K.Y.); a P&F grant from NIH Resource Center for Stable Isotope Resolved Metabolomics (RC-SIRM) at University of Kentucky (J.N.R. and K.Y.); and an ENGAGE grant from the National Center for Regenerative Medicine (K.Y.).

Author information

Author notes

  1. Xiuxing Wang and Kailin Yang: These authors contributed equally to this work.

Authors and Affiliations

  1. Department of Stem Cell Biology and Regenerative Medicine, Cleveland Clinic Lerner Research Institute, Cleveland, Ohio, USA

    Xiuxing Wang, Kailin Yang, Qi Xie, Qiulian Wu, Stephen C Mack, Yu Shi, Leo J Y Kim, Briana C Prager, William A Flavahan, Christopher G Hubert, Tyler E Miller, Wenchao Zhou, Zhi Huang, Xiaoguang Fang, Shideng Bao & Jeremy N Rich

  2. Department of Molecular Medicine, Cleveland Clinic Lerner College of Medicine of Case Western Reserve University, Cleveland, Ohio, USA

    Kailin Yang, Briana C Prager & Jeremy N Rich

  3. Institute of Pathology and Southwest Cancer Center, Southwest Hospital, The Third Military Medical University, and The Key Laboratory of Tumor Immunopathology, The Ministry of Education of China, Chongqing, China

    Yu Shi

  4. Department of Pharmacology and Cancer Biology, Duke Cancer Institute, Duke Molecular Physiology Institute, Duke University School of Medicine, Durham, North Carolina, USA

    Xiaojing Liu & Jason W Locasale

  5. Broad Institute of MIT and Harvard, Cambridge, Massachusetts, USA

    Meromit Singer, Aviv Regev & Mario L Suvà

  6. Department of Pathology and Center for Cancer Research, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, USA

    Mario L Suvà

  7. Department of Quantitative Health Sciences, Cleveland Clinic Lerner Research Institute, Cleveland, Ohio, USA

    Tae Hyun Hwang

Authors
  1. Xiuxing Wang
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Kailin Yang
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Qi Xie
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Qiulian Wu
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Stephen C Mack
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Yu Shi
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Leo J Y Kim
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Briana C Prager
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. William A Flavahan
    View author publications

    You can also search for this author in PubMed Google Scholar

  10. Xiaojing Liu
    View author publications

    You can also search for this author in PubMed Google Scholar

  11. Meromit Singer
    View author publications

    You can also search for this author in PubMed Google Scholar

  12. Christopher G Hubert
    View author publications

    You can also search for this author in PubMed Google Scholar

  13. Tyler E Miller
    View author publications

    You can also search for this author in PubMed Google Scholar

  14. Wenchao Zhou
    View author publications

    You can also search for this author in PubMed Google Scholar

  15. Zhi Huang
    View author publications

    You can also search for this author in PubMed Google Scholar

  16. Xiaoguang Fang
    View author publications

    You can also search for this author in PubMed Google Scholar

  17. Aviv Regev
    View author publications

    You can also search for this author in PubMed Google Scholar

  18. Mario L Suvà
    View author publications

    You can also search for this author in PubMed Google Scholar

  19. Tae Hyun Hwang
    View author publications

    You can also search for this author in PubMed Google Scholar

  20. Jason W Locasale
    View author publications

    You can also search for this author in PubMed Google Scholar

  21. Shideng Bao
    View author publications

    You can also search for this author in PubMed Google Scholar

  22. Jeremy N Rich
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

X.W., K.Y. and J.N.R. designed the experiments, analyzed the data and wrote the manuscript with contributions from all authors. X.W., K.Y., Q.X., Q.W., S.C.M., C.G.H., T.E.M., W.Z., Z.H. and X.F. performed the experiments. K.Y., Y.S., L.J.Y.K., B.C.P., W.A.F., M.S., A.R., M.L.S. and T.H.H. performed database analyses. X.W., K.Y., Q.X., X.L. and J.W.L. performed metabolic experiments. S.B. provided scientific input and helped edit the manuscript.

Corresponding author

Correspondence to Jeremy N Rich.

Ethics declarations

Competing interests

The authors declare no competing financial interests.

Integrated supplementary information

Supplementary Figure 1 Genomic profiles and epigenetic characterizations comparing TCGA glioma specimens and T4121 BTICs.

(a) Mutations in coding sequences of frequently mutated genes in glioblastoma were identified through whole exome sequencing. Top left legend: Number of genes mutated in the selected gene set per specimen. Right legend: Percentage of specimens harboring a mutation in each gene. (b) Following quantile normalization across samples and median centering of gene expression, the top 10% most variant genes were used to cluster low grade and high grade glioma samples with RNA-seq data from T4121 BTICs. (c) Hierarchical clustering on the Pan-glioma probe set (1,300 probes) of DNA methylation 450K array data from BTICs (n = 3) and TCGA GBM (n = 155). (d) Hierarchical clustering on the IDH wild type probe set (914 probes) from DNA methylation 450K array data from IDH wild type samples among TCGA GBM specimens (n = 149) and BTICs (n = 3).

Supplementary Figure 2 BTICs contain elevated levels of purine degradation products compared to differentiated progeny.

Targeted mass spectrometry was performed to examine the level of a purine degradation product (hypoxanthine) using cell lysates of matched brain tumor initiating cells (BTICs) and differentiated glioma cells (DGCs) from five patient-derived glioblastoma specimens (T3691, T387, T4121, TIN528, and T3565). The relative levels of hypoxanthine were displayed normalized to DGCs. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 4 independent experiments per group. (T3691, P < 0.0001, t(6) = 26.710; T387, P = 0.0001, t(6) = 8.020; T4121, P < 0.0001, t(6) = 15.232; TIN528, P < 0.0001, t(6) = 48.811; T3565, P < 0.0001, t(6) = 13.688.) (b,c) Upregulation of purine synthesis pathway genes in CD15+ BTICs. Surgical specimens from two patients newly diagnosed with glioblastomas were immediately dissociated and prospectively sorted with beads conjugated with an anti-CD15 antibody in the absence of culture. The mRNA levels of six purine synthesis pathway enzymes (PRPS1, PPAT, ADSL, ADSS, GMPS, and IMPDH1) were measured using quantitative RT-PCR in matched CD15+ and CD15- populations derived from these two primary glioblastoma specimens. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (Patient 1: PRPS1, P < 0.0001, t(4) = 34.461; PPAT, P < 0.0001, t(4) = 24.505; ADSL, P < 0.0001, t(4) = 19.579; ADSS, P < 0.0001, t(4) = 50.404; GMPS, P < 0.0001, t(4) = 13.235; IMPDH1, P < 0.0001, t(4) = 227.357. Patient 2: PRPS1, P < 0.0001, t(4) = 34.882; PPAT, P < 0.0001, t(4) = 34.345; ADSL, P < 0.0001, t(4) = 17.519; ADSS, P < 0.0001, t(4) = 19.213; GMPS, P < 0.0001, t(4) = 24.373; IMPDH1, P < 0.0001, t(4) = 39.910.)

Supplementary Figure 3 Effects of purine synthesis inhibitors and AICAR.

(a-c) Sensitivity of four patient-derived DGC models (T387, T3565, T3691, and T4121) to three purine synthesis inhibitors: (a) mycophenolic acid, (b) mycophenolate mofetil, and (c) ribavirin. Data are presented as the mean ± s.e.m. (d-f) Purine synthesis inhibitors attenuate GMP synthesis in BTICs. Two patient-derived BTIC models (T387 and T4121) were treated with each agent at concentrations that induce cellular responses: (d) mycophenolic acid (5 μM), (e) mycophenolate mofetil (5 μM), or (f) ribavirin (50 μM). DMSO was used as vehicle control. GMP levels were determined using targeted mass spectrometry. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 4 independent experiments per group. (Mycophenolic acid: T387, P < 0.0001, t(6) = 26.220; T4121, P < 0.0001, t(6) = 11.383. Mycophenolate mofetil: T387, P < 0.0001, t(6) = 13.002; T4121, P < 0.0001, t(6) = 21.557. Ribavirin: T387, P < 0.0001, t(6) = 33.811; T4121, p = 0.0002, t(6) = 7.257.) (g,h) Cell numbers of T387 and T4121 BTICs treated with 0, 100 μM, 200 μM, or 500 μM AICAR (5-aminoimidazole-4-carboxamide ribonucleotide). Time points were taken at days 1, 3, and 5. All statistical analyses were performed using two-tailed unpaired Student’s t-test; *, P < 0.05; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (T387: day 3: 100 μM, P = 0.0031, t(4) = 4.755; 200 μM, P = 0.0002, t(4) = 7.731; 500 μM, P < 0.0001, t(4) = 9.579; day 5: 100 μM, P < 0.0001, t(4) = 20.229; 200 μM, P < 0.0001, t(4) = 30.923; 500 μM, P < 0.0001, t(4) = 31.516. T4121: day 3: 100 μM, P < 0.0001, t(4) = 13.556; 200 μM, P < 0.0001, t(4) = 26.207; 500 μM, P < 0.0001, t(4) = 27.817; day 5: 100 μM, P < 0.0001, t(4) = 21.848; 200 μM, P < 0.0001, t(4) = 23.165; 500 μM, P < 0.0001, t(4) = 65.922.) (i,j) Cell viability of T387 and T4121 BTICs treated with 0, 100 μM, 200 μM, or 500 μM AICAR (5-aminoimidazole-4-carboxamide ribonucleotide) was assessed by CellTiter-Glo. Time points were taken at days 1, 3, and 5. All statistical analyses were performed using two-tailed unpaired Student’s t-test; *, P < 0.05; **, P < 0.01; n = 6 independent experiments per group. (T387: day 3: 100 μM, P = 0.0022, t(10) = 4.089; 200 μM,, P < 0.0001, t(10) = 10.301; 500 μM, P < 0.0001, t(10) = 11.282; day 5: 100 μM, P < 0.0001, t(10) = 7.399; 200 μM, P < 0.0001, t(10) = 19.776; 500 μM, P < 0.0001, t(10) = 20.225. T4121: day 3: 100 μM, P < 0.0001, t(10) = 7.434; 200 μM, P < 0.0001, t(10) = 31.419; 500 μM, P < 0.0001, t(10) = 41.475; day 5: 100 μM, P < 0.0001, t(10) = 7.521; 200 μM, P < 0.0001, t(10) = 19.140; 500 μM, P < 0.0001, t(10) = 21.404.)

Supplementary Figure 4 Pathway analyses comparing BTICs to DGCs.

(a) Pathway map of enhancer associated gene network (identified with H3K27 acetylation ChIP-seq in T4121) comparing BTICs (red) versus DGCs (green). Pathways shown were identified using GREAT (http://bejerano.stanford.edu/great/public/html/), which was used to query the MSIGdb Broad pathway database (http://software.broadinstitute.org/gsea/msigdb/collections.jsp). Shown are networks with a FDR value of less than 0.05. (b-e) BTICs upregulate genes involved in the purine pathway. T4121 BTICs and DGCs underwent RNA-seq analysis in biologic duplicates with selective analysis of metabolic pathways, including (b) glutathione, (c) glycolysis, (d) pentose phosphate, and (e) purine biosynthesis. Z-scores were calculated from the log2 transformed FPKM values for each sample.

Supplementary Figure 5 Effects of short-term versus long-term glucose restriction and GLUT3 knockdown.

(a-c) Effect of glucose restricted conditions (0.45 g/L) and compared to standard conditions (4.5 g/L) for 48 hours on stemness markers in three BTIC models (a) T3691, (b) T387 and (c) T4121. Quantitative RT-PCR was then performed for stem cell markers, including SOX2, NES, and OLIG2, and differentiation markers, including GFAP and MBPF. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (T3691: SOX2, P = 0.0002, t(4) = 11.262; NES, P < 0.0001, t(4) = 18.175; OLIG2, P < 0.0001, t(4) = 18.728; GFAP, P < 0.0001, t(4) = 34.228; MBPF, P < 0.0001, t(4) = 17.103. T387: SOX2, P < 0.0001, t(4) = 14.251; NES, P = 0.0029, t(4) = 5.385; OLIG2, P = 0.0003, t(4) = 9.799; GFAP, P = 0.0008, t(4) = 7.509; MBPF, P < 0.0001, t(4) = 25.319. T4121: SOX2, P < 0.0001, t(4) = 63.347; NES, P < 0.0001, t(4) = 30.664; OLIG2, P < 0.0001, t(4) = 22.384; GFAP, P = 0.0065, t(4) = 4.266; MBPF, P = 0.0026, t(4) = 5.513.) (d-f) Effect of restricted glucose conditions (0.45 g/L) compared to standard conditions (4.5 g/L) for 7 days on stemness markers in three BTIC models (d) T3691, (e) T387, and (f) T4121. Quantitative RT-PCR was then performed for stem cell markers, including SOX2, NES, and OLIG2, and differentiation markers, including GFAP and MBPF. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 3 independent experiments per group. (T3691: SOX2, P = 0.0002, t(4) = 10.369; NES, P = 0.0003, t(4) = 10.074; OLIG2, P = 0.0015, t(4) = 6.409; GFAP, P = 0.0015, t(4) = 6.436; MBPF, P < 0.0001, t(4) = 14.048. T387: SOX2, P < 0.0001, t(4) = 16.883; NES, P < 0.0001, t(4) = 20.536; OLIG2, P = 0.0003, t(4) = 9.481; GFAP, P = 0.0047, t(4) = 4.699; MBPF, P = 0.0010, t(4) = 7.217. T4121: SOX2, P < 0.0001, t(4) = 35.185; NES, P < 0.0001, t(4) = 57.581; OLIG2, P = 0.0015, t(4) = 9.355; GFAP, P = 0.0017, t(4) = 6.183; MBPF, P = 0.0051, t(4) = 4.585.) (g) Knockdown of GLUT3 using two independent shRNAs decreased mRNA level of genes involved in purine synthesis in three BTIC models, T3691, T387 and T4121. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (T3691: shGLUT3-1, P = 0.0023, t(4) = 11.828; shGLUT3-2, P = 0.0017, t(4) = 13.453; T387: shGLUT3-1, P < 0.0001, t(4) = 17.209; shGLUT3-2, P < 0.0001, t(4) = 16.923. T4121: shGLUT3-1, P = 0.0007, t(4) = 13.700; shGLUT3-2, P = 0.0005, t(4) = 15.455.) (h) Quantitative RT-PCR assessment of PRPS1, PPAT, ADSL, ADSS, GMPS, and IMPDH1 mRNA levels in T3691 BTICs expressing shCont, shGLUT3-1, or shGLUT3-2. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (T3691: PRPS1, shGLUT3-1, P = 0.0036, t(4) = 10.693; shGLUT3-2, P = 0.0028, t(4) = 9.098; PPAT, shGLUT3-1, P = 0.0003, t(4) = 7.292; shGLUT3-2, P = 0.0041, t(4) = 5.102; ADSL, shGLUT3-1, P = 0.0029, t(4) = 8.948; shGLUT3-2, P = 0.0009, t(4) = 13.386; ADSS, shGLUT3-1, P = 0.0056, t(4) = 6.619; shGLUT3-2, P = 0.0100, t(4) = 7.136; GMPS, shGLUT3-1, P = 0.0021, t(4) = 8.058. shGLUT3-2, P = 0.0010, t(4) = 11.836; IMPDH1, shGLUT3-1, P = 0.0064, t(4) = 8.040; shGLUT3-2, P =0.0047, t(4) = 8.966.) (i) Quantitative RT-PCR assessment of PRPS1, PPAT, ADSL, ADSS, GMPS, and IMPDH1 mRNA levels in T387 BTICs expressing shCont, shGLUT3-1, or shGLUT3-2. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (T387: PRPS1, shGLUT3-1, P < 0.0001, t(4) = 19.053; shGLUT3-2, P < 0.0001, t(4) = 16.493; PPAT, shGLUT3-1, P < 0.0001, t(4) = 16.840; shGLUT3-2, P = 0.0001, t(4) = 15.255; ADSL, shGLUT3-1, P = 0.0013, t(4) = 7.966; shGLUT3-2, P = 0.0018, t(4) = 7.371; ADSS, shGLUT3-1, P = 0.0012, t(4) = 8.168, shGLUT3-2, P = 0.0020, t(4) = 7.184; GMPS, shGLUT3-1, P = 0.0002, t(4) = 13.656, shGLUT3-2, P = 0.0003, t(4) = 11.689; IMPDH1, shGLUT3-1, P = 0.0017, t(4) = 7.527; shGLUT3-2, P = 0.0024, t(4) = 6.845.) (j) Quantitative RT-PCR assessment of PRPS1, PPAT, ADSL, ADSS, GMPS, and IMPDH1 mRNA levels in T4121 BTICs expressing shCont, shGLUT3-1, or shGLUT3-2. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (T4121: PRPS1, shGLUT3-1, P = 0.0003, t(4) = 18.993; shGLUT3-2, P = 0.0002, t(4) = 22.626; PPAT, shGLUT3-1, P = 0.0030, t(4) = 8.595; shGLUT3-2, P = 0.0015, t(4) = 8.029; ADSL, shGLUT3-1, P < 0.0001, t(4) = 18.242; shGLUT3-2, P < 0.0001, t(4) = 19.000; ADSS, shGLUT3-1, P = 0.0010, t(4) = 8.626; shGLUT3-2, P = 0.0007, t(4) = 9.225; GMPS, shGLUT3-1, P = 0.0015, t(4) = 10.968; shGLUT3-2, P = 0.0008, t(4) = 14.626; IMPDH1, shGLUT3-1, P < 0.0001, t(4) = 53.575, shGLUT3-2, P < 0.0001, t(4) = 61.720.) (k,l) Glutamine partially rescued defects in cell proliferation induced by glucose restriction. Two BTIC models (k) T3691 and (l) T4121 were treated with restricted glucose conditions (0.45 g/L) compared to standard conditions (4.5 g/L). 2 mM glutamine or 100 μM acetate was used to rescue glucose restriction separately. Cell viability was measured by CellTiter-Glo. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 6 independent experiments per group. (T3691 BTICs: Day 3, glucose restricted, P < 0.0001, t(10) = 25.748; glucose restricted + glutamine, P < 0.0001, t(10) = 9.168; glucose restricted + acetate, P < 0.0001, t(10) = 25.762; Day 5, glucose restricted, P < 0.0001, t(10) = 26.061; glucose restricted + glutamine, P < 0.0001, t(10) = 25.288; glucose restricted + acetate, P < 0.0001, t(10) = 26.288. T4121 BTICs: Day 3, glucose restricted, P < 0.0001, t(10) = 13.688; glucose restricted + glutamine, P = 0.0043, t(10) = 3.676; glucose restricted + acetate, P < 0.0001, t(10) = 11.881; Day 5, glucose restricted, P < 0.0001, t(10) = 42.796; glucose restricted + glutamine, P < 0.0001, t(10) = 25.721; glucose restricted + acetate, P < 0.0001, t(10) = 39.037.)

Supplementary Figure 6 Nitrogen tracing in BTICs using [15N2]glutamine.

(a-f) Targeted mass spectrometry was used to examine the levels of (a,b) IMP, (c,d) AMP, and (e,f) GMP using cell lysates of BTICs and matched DGCs from patient-derived glioblastoma specimen T387 treated with media containing either 2 mM regular glutamine or 2 mM 15N2-glutamine, respectively, for 24 hours before harvesting.

Supplementary Figure 7 Transcription factor MYC regulates mRNA levels of purine synthesis pathway genes in BTICs.

(a-g) Knockdown of MYC using two independent shRNAs decreased mRNA levels of genes involved in purine synthesis in BTICs. Quantitative RT-PCR assessment of MYC, PRPS1, PPAT, ADSL, ADSS, GMPS, and IMPDH1 mRNA levels in T387 and T4121 BTICs expressing shCont, shMYC-1, or shMYC-2. All statistical analyses were performed using two-tailed unpaired Student’s t-test; *, P < 0.05; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (a) MYC levels: T387: shMYC-1, P = 0.0005, t(4) = 14.853; shMYC-2, P = 0.0003, t(4) = 17.277. T4121: shMYC-1, P = 0.0004, t(4) = 17.929; shMYC-2, P = 0.0002, t(4) = 22.822. (b) PRPS1 levels: T387: shMYC-1, P < 0.0001, t(4) = 26.097; shMYC-2, P = 0.0002, t(4) = 11.178. T4121: shMYC-1, P = 0.0063, t(4) = 9.693; shMYC-2, P = 0.0021, t(4) = 7.116. (c) PPAT levels: T387: shMYC-1, P = 0.0002, t(4) = 13.522; shMYC-2, P < 0.0001, t(4) = 21.005. T4121: shMYC-1, P = 0.0011, t(4) = 9.169; shMYC-2, P = 0.0014, t(4) = 8.675. (d) ADSL levels: T387: shMYC-1, P < 0.0001, t(4) = 23.201; shMYC-2, P < 0.0001, t(4) = 17.491. T4121: shMYC-1, P < 0.0001, t(4) = 20.244; shMYC-2, P = 0.0002, t(4) = 13.751. (e) ADSS levels: T387: shMYC-1, P < 0.0001, t(4) = 17.853; shMYC-2, P < 0.0001, t(4) = 17.784. T4121: shMYC-1, P = 0.0002, t(4) = 7.265; shMYC-2, P = 0.0003, t(4) = 6.901. (f) GMPS levels: T387: shMYC-1, P < 0.0001, t(4) = 17.462; shMYC-2, P < 0.0001, t(4) = 16.212. T4121: shMYC-1, P = 0.0035, t(4) = 8.809; shMYC-2, P = 0.0234, t(4) = 3.256. (g) IMPDH1 levels: T387: shMYC-1, P = 0.0005, t(4) = 11.061; shMYC-2, P = 0.0007, t(4) = 8.687. T4121: shMYC-1, P < 0.0001, t(4) = 50.570; shMYC-2, P < 0.0001, t(4) = 39.171. (h-n) Overexpression of exogenous MYC upregulates mRNA levels of genes involved in purine synthesis pathway in BTICs. Quantitative RT-PCR assessment of PRPS1, PPAT, ADSL, ADSS, GMPS, and IMPDH1 mRNA levels in T387 and T4121 BTICs expressing empty vector (Vector) or MYC overexpression vector (Myc). All statistical analyses were performed using two-tailed unpaired Student’s t-test; *, P < 0.05; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (h) MYC levels: T387: p = 4.76e-05., t(4) = 18.750. T4121: p = 1.85e-06. t(4) = 42.371. (i) PRPS1 levels: T387: P = 0.0072, t(4) = 16.118. T4121: P = 0.0054, t(4) = 7.577. (j) PPAT levels: T387: P < 0.0001, t(4) = 17.038. T4121: P < 0.0001, t(4) = 17.734. (k) ADSL levels: T387: P = 0.0010, t(4) = 10.814. T4121: P = 0.0082, t(4) = 5.112. (l) ADSS levels: T387: P < 0.0001, t(4) = 19.064. T4121: P = 0.0120, t(4) = 4.923. (m) GMPS levels: T387: P = 0.0054, t(4) = 5.487. T4121: P < 0.0001, t(4) = 31.383. (n) IMPDH1 levels: T387: P = 0.0004, t(4) = 14.926. T4121: P = 0.0027, t(4) = 6.588. (o,p) MYC regulates stemness markers in BTICs. Expression of MYC was knocked down using two non-overlapping shRNAs in (o) T387 and (p) T4121 BTICs. Quantitative RT-PCR was performed for precursor markers, including SOX2, NES, and OLIG2, and differentiation markers, including GFAP and MBPF. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (o) T387: SOX2, shRNA1, P = 0.0035, t(4) = 5.077; shRNA2, P = 0.0003, t(4) = 9.675. NES, shRNA1, P = 0.0044, t(4) = 4.775; shRNA2, P = 0.0006, t(4) = 8.081. OLIG2, shRNA1, P < 0.0001, t(4) = 30.207; shRNA2, P < 0.0001, t(4) = 51.140. GFAP, shRNA1, P = 0.0008, t(4) = 7.698; shRNA2, P < 0.0001, t(4) = 30.218. MBPF, shRNA1, P = 0.0007, t(4) = 7.748; shRNA2, P < 0.0001, t(4) = 13.316. (p) T4121: SOX2, shRNA1, P = 0.0005, t(4) = 8.549; shRNA2, P < 0.0001, t(4) = 14.691. NES, shRNA1, P = 0.0010, t(4) = 7.237; shRNA2, P < 0.0001, t(4) = 13.479. OLIG2, shRNA1, P = 0.0001, t(4) = 12.169; shRNA2, P < 0.0001, t(4) = 15.191. GFAP, shRNA1, P = 0.0003, t(4) = 9.957; shRNA2, P = 0.0010, t(4) = 7.318. MBPF, shRNA1, P = 0.0021, t(4) = 5.865; shRNA2, P = 0.0044, t(4) = 4.786. (q,r) MYC induces stemness markers in DGCs. MYC was transduced into in (o) T387 and (p) T4121 DGCs using a lentiviral construct. An empty lentiviral vector was used as a negative control. Quantitative RT-PCR was performed for precursor markers (SOX2, NES, and OLIG2), and differentiation markers (GFAP and MBPF). All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (q) T387: SOX2, P = 0.0002, t(4) = 11.187; NES, P < 0.0001, t(4) = 15.998; OLIG2, P < 0.0001, t(4) = 15.315; GFAP, P = 0.0004, t(4) = 9.214; MBPF, P = 0.0002, t(4) = 10.481. (r) T4121: SOX2, P < 0.0001, t(4) = 19.782; NES, P = 0.0006, t(4) = 8.348; OLIG2, P < 0.0001, t(4) = 23.285; GFAP, P < 0.0001, t(4) = 14.848; MBPF, P < 0.0001, t(4) = 23.382.

Supplementary Figure 8 Purine pathway gene expression analysis by subtype from TCGA datasets.

(a-c) The expression levels of (a) individual precursor marker mRNAs (OLIG2, MYC, and SOX2), (b) individual purine synthesis pathway mRNAs (IMPDH1, PRPS1, PPAT, GMPS, ADSL, and ADSS), and (c) purine pathway expression signature based on KEGG definition, from the TCGA dataset were plotted by four subtypes (Classical, Mesenchymal, Neural, and Proneural) as designated by the Verhaak 2010 paradigm. (d-f) PPAT, ADSS, and IMPDH1 are upregulated in BTICs. H3K27ac ChIP-seq enrichment plots centered at the gene locus for (d) ADSS, (e) IMPDH1, and (f) PPAT were displayed. Enrichment is shown for three matched pairs of DGCs and BTICs from patient-derived glioblastoma specimens. H3K27ac ChIP-seq data were downloaded from NCBI Gene Expression Omnibus (GEO) GSE54047. (g-i) qRT-PCR quantification of ADSS, IMPDH1, or PPAT mRNA levels in three patient-derived BTIC models during differentiation. BTICs derived from three primary human glioblastoma specimens (T3691, T387, T4121) were treated with serum to induce differentiation, with time points taken at 2, 4, and 6 days. All statistical analyses were performed using two-tailed unpaired Student’s t-test; *, P < 0.05; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (ADSS: T3691, 2d, P = 0.7400, t(4) = 0.353; 4d, P = 0.0005, t(4) = 10.524; 6d, P = 0.0006, t(4) = 10.049; T387, 2d, P < 0.0001, t(4) = 24.171; 4d, P < 0.0001, t(4) = 28.101; 6d, P < 0.0001, t(4) = 29.636; T4121, 2d, P < 0.0001, t(4) = 15.710; 4d, P < 0.0001, t(4) = 15.975; 6d, P < 0.0001, t(4) = 24.548. IMPDH1: T3691, 2d, P = 0.0133, t(4) = 3.996; 4d, P = 0.0024, t(4) = 6.522; 6d, P = 0.0029, t(4) = 6.237; T387, 2d, P < 0.0001, t(4) = 19.792; 4d, P < 0.0001, t(4) = 22.904; 6d, P < 0.0001, t(4) = 22.964; T4121, 2d, P = 0.0105, t(4) = 4.553; 4d, P = 0.0049, t(4) = 8.337; 6d, P = 0.0017, t(4) = 6.913. PPAT: T3691, 2d, P = 0.0200, t(4) = 4.784; 4d, P = 0.0033, t(4) = 7.025; 6d, P = 0.0035, t(4) = 7.681; T387, 2d, P < 0.0001, t(4) = 21.370; 4d, P < 0.0001, t(4) = 24.175; 6d, P < 0.0001, t(4) = 25.004; T4121, 2d, P < 0.0001, t(4) = 18.839; 4d, P < 0.0001, t(4) = 15.865; 6d, P < 0.0001, t(4) = 18.083.)

Supplementary Figure 9 Confirmation and reconstitution of ADSL, GMPS or PRPS1 knockdown.

(a) Confirmation of ADSL knockdown effect in BTICs (T387 and T4121) (b) Confirmation of GMPS knockdown effect in BTICs (T387 and T4121). (c) Confirmation of PRPS1 knockdown effect in BTICs (T387 and T4121). (d-f) Reconstitution of shRNA-mediated knockdown of purine synthesis genes in BTICs. Expression constructs for human ADSL, GMPS< and PRPS1, which were resistant to the respective shRNAs, were used to reconstitute the expression of corresponding genes in two BTIC models, T387 and T4121. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (d) ADSL: T387, shADSL-1, P < 0.0001, t(4) = 15.560, shADSL-2, P < 0.0001, t(4) = 14.701; shADSL-1 + ADSL, P < 0.0001, t(4) = 19.428; shADSL-2 + ADSL, P < 0.0001, t(4) = 13.765; T4121, shADSL-1, P = 0.0001, t(4) = 11.904, shADSL-2, P = 0.0003, t(4) = 11.571; shADSL-1 + ADSL, P = 0.0002, t(4) = 9.588; shADSL-2 + ADSL, P < 0.0001, t(4) = 29.885. (e) GMPS: T387, shGMPS-1, P = 0.0001, t(4) = 12.301, shGMPS-2, P = 0.0002, t(4) = 11.375; shGMPS-1 + GMPS, P < 0.0001, t(4) = 22.867; shGMPS-2 + GMPS, P < 0.0001, t(4) = 58.042; T4121, shGMPS-1, P < 0.0001, t(4) = 17.121, shGMPS-2, P < 0.0001, t(4) = 13.955; shGMPS-1 + GMPS, P < 0.0001, t(4) = 4.795; shGMPS-2 + GMPS, P < 0.0001, t(4) = 15.223. (f) PRPS1: T387, shPRPS1-1, P = 0.0003, t(4) = 10.059, shPRPS1-2, P = 0.0003, t(4) = 9.448; shPRPS1-1 + PRPS1, P < 0.0001, t(4) = 32.778; shPRPS1-2 + PRPS1, P < 0.0001, t(4) = 84.438; T4121, shPRPS1-1, P < 0.0001, t(4) = 20.042, shPRPS1-2, P < 0.0001, t(4) = 18.371; shPRPS1-1 + PRPS1, P < 0.0001, t(4) = 14.407; shPRPS1-2 + PRPS1, P < 0.0001, t(4) = 29.308. (g-i) Cell viability was measured on T387 BTICs with reconstitution of (g) ADSL, (h) GMPS, or (i) PRPS1 upon targeting each gene. BTICs expressing shCont were used as positive controls, and BTICs expressing shRNA targeting each gene without reconstitution were used as negative controls. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 6 independent experiments per group. (g) ADSL: 3d, shCont vs. shADSL-1, P < 0.0001, t(10) = 11.702; shCont vs. shADSL-2, P < 0.0001, t(10) = 12.057; shADSL-1 vs. shADSL-1 + ADSL, P < 0.0001, t(10) = 12.419; shADSL-2 vs. shADSL-2 + ADSL, P < 0.0001, t(10) = 23.708; 5d, shCont vs. shADSL-1, P < 0.0001, t(10) = 11.940; shCont vs. shADSL-2, P < 0.0001, t(10) = 12.030; shADSL-1 vs. shADSL-1 + ADSL, P < 0.0001, t(10) = 50.707; shADSL-2 vs. shADSL-2 + ADSL, P < 0.0001, t(10) = 12.524. (h) GMPS: 3d, shCont vs. shGMPS-1, P < 0.0001, t(10) = 21.323; shCont vs. shGMPS-2, P < 0.0001, t(10) = 20.689; shGMPS-1 vs. shGMPS-1 + GMPS, P < 0.0001, t(10) = 19.455; shGMPS-2 vs. shGMPS-2 + GMPS, P < 0.0001, t(10) = 27.714; 5d, shCont vs. shGMPS-1, P < 0.0001, t(10) = 21.273; shCont vs. shGMPS-2, P < 0.0001, t(10) = 21.327; shGMPS-1 vs. shGMPS-1 + GMPS, P < 0.0001, t(10) = 45.018; shGMPS-2 vs. shGMPS-2 + GMPS, P < 0.0001, t(10) = 18.834. (i) PRPS1: 3d, shCont vs. shPRPS1-1, P < 0.0001, t(10) = 21.930; shCont vs. shPRPS1-2, P < 0.0001, t(10) = 22.394; shPRPS1-1 vs. shPRPS1-1 + PRPS1, P < 0.0001, t(10) = 16.273; shPRPS1-2 vs. shPRPS1-2 + PRPS1, P < 0.0001, t(10) = 27.214; 5d, shCont vs. shPRPS1-1, P < 0.0001, t(10) = 24.397; shCont vs. shPRPS1-2, P < 0.0001, t(10) = 25.813; shPRPS1-1 vs. shPRPS1-1 + PRPS1, P < 0.0001, t(10) = 17.896; shPRPS1-2 vs. shPRPS1-2 + PRPS1, P < 0.0001, t(10) = 16.070. (j-l) Cell viability was measured on T4121 BTICs with reconstitution of (j) ADSL, (k) GMPS, or (l) PRPS1 upon targeting each gene. BTICs expressing shCont were used as positive controls, and BTICs expressing shRNA targeting each gene without reconstitution were used as negative controls. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 6 independent experiments per group. (j) ADSL: 3d, shCont vs. shADSL-1, P < 0.0001, t(10) = 21.756; shCont vs. shADSL-2, P < 0.0001, t(10) = 20.821; shADSL-1 vs. shADSL-1 + ADSL, P < 0.0001, t(10) = 20.792 shADSL-2 vs. shADSL-2 + ADSL, P < 0.0001, t(10) = 24.815; 5d, shCont vs. shADSL-1, P < 0.0001, t(10) = 17.779; shCont vs. shADSL-2, P < 0.0001, t(10) = 17.833; shADSL-1 vs. shADSL-1 + ADSL, P < 0.0001, t(10) = 14.576; shADSL-2 vs. shADSL-2 + ADSL, P < 0.0001, t(10) = 9.860. (k) GMPS: 3d, shCont vs. shGMPS-1, P < 0.0001, t(10) = 24.485; shCont vs. shGMPS-2, P < 0.0001, t(10) = 24.741; shGMPS-1 vs. shGMPS-1 + GMPS, P < 0.0001, t(10) = 20.847; shGMPS-2 vs shGMPS-2 + GMPS, P < 0.0001, t(10) = 26.123; 5d, shCont vs. shGMPS-1, P < 0.0001, t(10) = 19.968; shCont vs. shGMPS-2, P < 0.0001, t(10) = 20.034; shGMPS-1 vs. shGMPS-1 + GMPS, P < 0.0001, t(10) = 13.139; shGMPS-2 vs. shGMPS-2 + GMPS, P < 0.0001, t(10) = 16.020. (l) PRPS1: 3d, shCont vs. shPRPS1-1, P < 0.0001, t(10) = 14.668; shCont vs. shPRPS1-2, P < 0.0001, t(10) = 13.888; shPRPS1-1 vs. shPRPS1-1 + PRPS1, P < 0.0001, t(10) = 25.521; shPRPS1-2 vs. shPRPS1-2 + PRPS1, P < 0.0001, t(10) = 32.771; 5d, shCont vs. shPRPS1-1, P < 0.0001, t(10) = 15.793; shCont vs. shPRPS1-2, P < 0.0001, t(10) = 15.749; shPRPS1-1 vs shPRPS1-1 + PRPS1, P < 0.0001, t(10) = 15.179; shPRPS1-2 vs. shPRPS1-2 + PRPS1, P < 0.0001, t(10) = 16.676. (m-r) Attenuation of purine synthesis pathway genes decreases levels of precursor markers. The expression levels of three purine synthesis pathway genes (ADSL, GMPS, and PRPS1) were knocked down using shRNAs in (m-o) T387 and (p-r) T4121 BTICs. Two independent shRNAs were used for each gene, and shCont was used as a control. Quantitative RT-PCR assessment was performed for precursor markers (SOX2, NES, and OLIG2), and differentiation markers (GFAP and MBPF). All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (m) shADSL/T387: SOX2, shADSL-1, P < 0.0001, t(4) = 14.673; shADSL-2, P < 0.0001, t(10) = 17.606. NES, shADSL-1, P = 0.0002, t(4) = 11.146; shADSL-2, P < 0.0001, t(4) = 16.397. OLIG2, shADSL-1, P < 0.0001, t(4) = 58.604; shADSL-2, P < 0.0001, t(4) = 79.446. GFAP, shADSL-1, P = 0.0006, t(4) = 8.312; shADSL-2, P = 0.0001, t(4) = 12.230. MBPF, shADSL-1, P < 0.0001, t(4) = 28.772; shADSL-2, P = 0.0001, t(4) = 12.936. (n) shGMPS/T387: SOX2, shGMPS-1, p = 0.0017 P < 0.0001, t(4) = 6.174; shGMPS-2, P = 0.0022, t(4) = 5.770. NES, shGMPS-1, P = 0.0006, t(4) = 8.178; shGMPS-2, P = 0.0004, t(4) = 9.273. OLIG2, shGMPS-1, P < 0.0001, t(4) = 21.918; shGMPS-2, P < 0.0001, t(4) = 20.266. GFAP, shGMPS-1, P = 0.0006, t(4) = 8.254; shGMPS-2, P = 0.0001, t(4) = 12.016. MBPF, shGMPS-1, P < 0.0001, t(4) = 19.169; shGMPS-2, P < 0.0001, t(4) = 17.426. (o) shPRPS1/T387: SOX2, shPRPS1-1, P < 0.0001, t(4) = 15.687; shPRPS1-2, P < 0.0001, t(4) = 14.160. NES, shPRPS1-1, P < 0.0001, t(4) = 13.845; shPRPS1-2, P < 0.0001, t(4) = 16.453. OLIG2, shPRPS1-1, P = 0.0008, t(4) = 7.737; shPRPS1-2, P = 0.0005, t(4) = 8.555. GFAP, shPRPS1-1, P < 0.0001, t(4) = 15.062; shPRPS1-2, P = 0.0010, t(4) = 7.229. MBPF, shPRPS1-1, P < 0.0001, t(4) = 18.115; shPRPS1-2, P = 0.0002, t(4) = 11.449. (p) shADSL/T4121: SOX2, shADSL-1, P < 0.0001, t(4) = 21.903; shADSL-2, P < 0.0001, t(4) = 21.559. NES, shADSL-1, P < 0.0001, t(4) = 13.181; sh-2, P = 0.0001, t(4) = 12.622. OLIG2, shADSL-1, P < 0.0001, t(4) = 35.077; shADSL-2, P < 0.0001, t(4) = 33.981. GFAP, shADSL-1, P = 0.0017, t(4) = 6.204; shADSL-2, P = 0.0067, t(4) = 4.227. MBPF, shADSL-1, P = 0.0005, t(4) = 8.531; shADSL-2, P = 0.0002, t(4) = 10.723. (q) shGMPS/T4121: SOX2, shGMPS-1, P < 0.0001, t(4) = 24.236; shGMPS-2, P < 0.0001, t(4) = 35.825. NES, shGMPS-1, P = 0.0003, t(4) = 9.719; shGMPS-2, P = 0.0001, t(4) = 12.040. OLIG2, shGMPS-1, P < 0.0001, t(4) = 15.920; shGMPS-2, P < 0.0001, t(4) = 17.743. GFAP, shGMPS-1, P = 0.0003, t(4) = 10.097; shGMPS-2, P = 0.0011, t(4) = 7.052. MBPF, shGMPS-1, P < 0.0001, t(4) = 14.167; shGMPS-2, P = 0.0006, t(4) = 8.057. (r) shPRPS1/T4121: SOX2, shPRPS1-1, P = 0.0003, t(4) = 9.883; shPRPS1-2, P = 0.0010, t(4) = 7.120. NES, shPRPS1-1, P < 0.0001, t(4) = 13.532; shPRPS1-2, P = 0.0008, t(4) = 7.601. OLIG2, shPRPS1-1, P = 0.0002, t(4) = 10.928; shPRPS1-2, P = 0.0003, t(4) = 10.153. GFAP, shPRPS1-1, P < 0.0001, t(4) = 20.580; shPRPS1-2, P = 0.0001, t(4) = 12.366. MBPF, shPRPS1-1, P = 0.0108, t(4) = 3.658; shPRPS1-2, P = 0.0024, t(4) = 5.659.

Supplementary Figure 10 Purine synthesis enzymes are dispensable in DGCs.

(a) Cell viability of T387 and T4121 DGCs expressing non-targeting control shRNA (shCont), shADSL-1, or shADSL-2. All statistical analyses were performed using two-tailed unpaired Student’s t-test; n = 6 independent experiments per group. (T387: shADSL-1, 3d, P = 0.0898, t(10) = 1.901; 5d, P = 0. 0.6953, t(10) = 0.405. shADSL-2, 3d, P = 0. 0.2007, t(10) = 1.381; 5d, P = 0.0.3285, t(10) = 1.033. T4121: shADSL-1, 3d, P = 0. 0.7437, t(10) = 0.336; 5d, P = 0.5498, t(10) = 0.619. shADSL-2, 3d, P = 0.0.4010, t(10) = 0.877; 5d, P = 0. 0.1005, t(10) = 1.857.) (b) Cell viability of T387 and T4121 DGCs expressing shCont, shGMPS-1, or shGMPS-2. All statistical analyses were performed using two-tailed unpaired Student’s t-test; n = 6 independent experiments per group. (T387: shGMPS-1, 3d, P = 0. 0.5411, t(10) = 0.648; 5d, P = 0. 5498, t(10) = 0.619. shGMPS-2, 3d, P = 0. 5308, t(10) = 0.665; 5d, P = 0. 1243, t(10) = 1.786. T4121: shGMPS-1, 3d, P = 0. 1611, t(10) = 1.598; 5d, P = 0. 2542, t(10) = 1.261. shGMPS-2, 3d, P = 0. 1765, t(10) = 1.532; 5d, P = 0. 1334, t(10) = 1.735.) (c) Cell viability of T387 and T4121 DGCs expressing shCont, shPRPS1-1, or shPRPS1-2. All statistical analyses were performed using two-tailed unpaired Student’s t-test; n = 6 independent experiments per group. (T387: shPRPS1-1, 3d, P = 0. 6128, t(10) = 0.509; 5d, P = 0. 6473 t(10) = 0.433. shPRPS1-2, 3d, P = 0. 1689, t(10) = 1.483; 5d, P = 0. 7176, t(10) = 0.372. T4121: shPRPS1-1, 3d, P = 0. 9627, t(10) = 0.048; 5d, P = 0. 9848, t(10) = 0.019. shPRPS1-2, 3d, P = 0. 7391, t(10) = 0.342; 5d, P = 0. 1960, t(10) = 1.385.)

Supplementary Figure 11 In vivo impact of targeting purine synthesis enzymes on proliferation, apoptosis and stemness.

(a) Representative images of immunohistochemical (IHC) staining of cross-sections of mouse brains harvested on day 18 after transplantation of T387 BTICs expressing shCont, shADSL-1, or shADSL-2. Ki-67, cleaved Caspase-3, and SOX2 were employed as markers of proliferation, apoptosis, and stemness respectively. Each image was representative for at least 3 similar experiments. (b) Representative images of IHCstaining of cross-sections of mouse brains harvested on day 18 after transplantation of T387 BTICs expressing shCont, shGMPS-1, or shGMPS-2. Each image was representative for at least 3 similar experiments. (c) Representative images of IHC staining of cross-sections of mouse brains harvested on day 18 after transplantation of T387 BTICs expressing shCont, shPRPS1-1, or shPRPS1-2. Each image was representative for at least 3 similar experiments.

Supplementary Figure 12 Purine synthesis regulates in vivo BTIC tumorigenesis in an immunocompetent mouse model.

(a-c) Successful targeting of three purine synthetic genes: (a) Adsl, (b) Gmps, and Prps1, in the murine GL261 glioma line was confirmed by RT-PCR. All statistical analyses were performed using two-tailed unpaired Student’s t-test; **, P < 0.01; n = 3 independent experiments per group. Data were presented as median ± s.e.m. (shAdsl-1, P < 0.0001, t(4) = 22.059; shAdsl-2, P < 0.0001, t(4) = 20.764. shGmps-1, P < 0.0001, t(4) = 20.040; shGmps-2, P < 0.0001, t(4) = 17.503. shPrps1-1, P < 0.0001, t(4) = 61.767; shPrps1-2, P < 0.0001, t(4) = 30.503.) (d-f) Kaplan-Meier survival curves of MHC-matched B6 immunocompetent mice bearing intracranial glioma 261 (GL261) BTICs, transduced with shCont, shAdsl, shGmps, or shPrps1 are displayed. For each target, two, non-overlapping shRNAs were used. n = 5 animals per group. Statistical analysis was performed using log-rank test. **, p < 0.01. (shAdsl: P = 0.0002; shGmps: P < 0.0001; shPrps1: P = 0.0001.) (g-i) Representative images of hematoxylin and eosin stained cross-sections of B6 mouse brains harvested on day 18 after transplantation of GL216 BTICs expressing shCont, shAdsl-1, shAdsl-2, shGmps-1, shGmps-2, shPrps1-1, or shPrps1-2 are shown. Scale bar, 2 mm. Each image was representative for at least 3 similar experiments.

Supplementary Figure 13 Purine synthesis enzymes are expressed in non-G-CIMP gliomas.

(a) Investigation of purine pathway gene expression in the pan-glioma TCGA cohort demonstrates enrichment in high grade glioma. (b) Gene set enrichment analysis reveals significant positive correlation of the purine metabolic signature with regulation of cell cycle and transcriptional activity. (c-f) Survival analysis of the Freije high-grade glioma dataset for PRPS1, ADSL, ADSS, and GMPS. All statistical analyses were performed using log-rank test. (PRPS1low, n = 44; PRPS1high, n = 41; P = 0.0854. ADSLlow, n = 43; ADSLhigh, n = 42; P = 0.2701. ADSSlow, n = 43; ADSShigh, n = 42; P = 0.9828. GMPSlow, n = 51; GMPShigh, n = 34; P = 0.0145.) (g-j) Survival analysis of the Phillips high-grade glioma dataset for PRPS1, ADSL, GMPS, and IMPDH. All statistical analyses were performed using log-rank test. (PRPS1low, n = 39; PRPS1high, n = 38; P = 0.3408. ADSLlow, n = 37; ADSLhigh, n = 38; P = 0.2416. GMPSlow, n = 52; GMPShigh, n = 25; P < 0.0001. IMPDH1low, n = 39; IMPDH1high, n = 38; P = 0.8485.) (k,l) Survival analysis of the TCGA glioblastoma for PRPS1 and GMPS. All statistical analyses were performed using log-rank test. (PRPS1low, n = 244; PRPS1high, n = 237; P = 0.098. GMPSlow, n = 230; GMPShigh, n = 251; P = 0.3408.) (m-r) Survival analysis of the REMBRANDT dataset for 6 enzymes in the purine synthesis pathway. All statistical analyses were performed using log-rank test. (PRPS1low, n = 254; PRPS1high, n = 255; P = 0.2029. PPATlow, n = 254; PPAThigh, n = 255; P = 0.0002. ADSLlow, n = 254; ADSLhigh, n = 255; P = 0.0063. ADSSlow, n = 254; ADSShigh, n = 255; P < 0.0001. GMPSlow, n = 254; GMPShigh, n = 255; P < 0.0001. IMPDH1low, n = 254; IMPDH1high, n = 255; P < 0.0001.)

Supplementary Figure 14 Proposed model of purine biosynthesis in BTICs.

BTICs upregulate GLUT3 to hijack the neuronal survival mechanism to compete for glucose in a dynamic microenvironment. The increased carbon influx is channeled to the synthesis of purine nucleotides, mediated by the transcriptional factor MYC. Elevated levels of purine nucleotides promote BTIC maintenance through numerous mechanisms, including DNA synthesis, self-renewal and tumorigenecity, and signaling pathways.

Supplementary Figure 15 Full-length Western blots.

Full-length Western blots for cropped images in Fig. 1g, 3d, 3i, 3k, 4a, 4b, 4k, 4l, and Supplementary Fig. 9a-c.

Supplementary information

Supplementary Text and Figures

Supplementary Figures 1–15 and Supplementary Table 1 (PDF 3879 kb)

Supplementary Methods Checklist (PDF 1401 kb)

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Wang, X., Yang, K., Xie, Q. et al. Purine synthesis promotes maintenance of brain tumor initiating cells in glioma. Nat Neurosci 20, 661–673 (2017). https://doi.org/10.1038/nn.4537

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/nn.4537

This article is cited by

Search

Advanced search

Quick links

Nature Briefing: Cancer

Sign up for the Nature Briefing: Cancer newsletter — what matters in cancer research, free to your inbox weekly.

Get what matters in cancer research, free to your inbox weekly. Sign up for Nature Briefing: Cancer