Nuove frontiere in Neurochirurgia Oncologica:

un progetto di ricerca sui tumori cerebraliDonatella Sgubin MD-PhD

S.O.C. di NeurochirurgiaDirettore Dr. Andrea Barbanera

Pre-op Post-op 6 months later

Stupp et al. (2009) Lancet Oncol

Lastly, we explored the relationship of molecular subclasseswith clinical parameters such as treatment response or survival.In the current larger TCGA cohort, the survival advantage of pro-neural subtype GBM (Phillips et al., 2006) was definitively shown

to be conferred by G-CIMP status, with non-G-CIMP proneuralGBMs and not mesenchymal GBM tending to show less favor-able outcomes in the first 12 months following initial diagnosiscompared to other subtypes (p value 0.07; Figure S6A). Although

Figure 5. Molecular Subclasses of GBM and their Genomic Molecular Correlates(A) Genomic alterations and survival associated with five molecular subtypes of GBM. Expression and DNAmethylation profiles were used to classify 332 GBMs

with available (native DNA and whole-genome amplified DNA) exome sequencing and DNA copy-number levels. Themost significant genomic associations were

identified through Chi-square tests, with p values corrected for multiple testing using the Benjamini-Hochberg method.

(B) Genomic alterations and sample features associated with six GBM methylation clusters. Epigenomic consensus clustering was performed on 396 GBM

samples profiled across two different platforms (Infinium HM27 and Infinium HM450). Six DNA methylation clusters were identified (see related Figure S5),

represented as M1 to M6, where M5 is G-CIMP. These DNA methylation signatures are correlated with 27 selected features composed of clinical, somatic, and

copy-number alterations; DM cluster, G-CIMP status, four TCGA GBM gene expression subclasses, two clinical features (Age at diagnosis/overall survival in

months), somatic mutations (IDH1, TP53, ATRX), and 18 selected copy-number alterations.

472 Cell 155, 462–477, October 10, 2013 ª2013 Elsevier Inc.

Brennan CW (2013) CellJohnson et al. (2014) Science

Lastly, we explored the relationship of molecular subclasseswith clinical parameters such as treatment response or survival.In the current larger TCGA cohort, the survival advantage of pro-neural subtype GBM (Phillips et al., 2006) was definitively shown

to be conferred by G-CIMP status, with non-G-CIMP proneuralGBMs and not mesenchymal GBM tending to show less favor-able outcomes in the first 12 months following initial diagnosiscompared to other subtypes (p value 0.07; Figure S6A). Although

Figure 5. Molecular Subclasses of GBM and their Genomic Molecular Correlates(A) Genomic alterations and survival associated with five molecular subtypes of GBM. Expression and DNAmethylation profiles were used to classify 332 GBMs

with available (native DNA and whole-genome amplified DNA) exome sequencing and DNA copy-number levels. Themost significant genomic associations were

identified through Chi-square tests, with p values corrected for multiple testing using the Benjamini-Hochberg method.

(B) Genomic alterations and sample features associated with six GBM methylation clusters. Epigenomic consensus clustering was performed on 396 GBM

samples profiled across two different platforms (Infinium HM27 and Infinium HM450). Six DNA methylation clusters were identified (see related Figure S5),

represented as M1 to M6, where M5 is G-CIMP. These DNA methylation signatures are correlated with 27 selected features composed of clinical, somatic, and

copy-number alterations; DM cluster, G-CIMP status, four TCGA GBM gene expression subclasses, two clinical features (Age at diagnosis/overall survival in

months), somatic mutations (IDH1, TP53, ATRX), and 18 selected copy-number alterations.

472 Cell 155, 462–477, October 10, 2013 ª2013 Elsevier Inc.

progenitors express bnl>GFP. As progenitorscontinue along the DT, DT larval cells activatebnl>GFP expression one segment at a timefrom anterior to posterior, matching progenitormovement.

This dynamic bnl expression along the migra-tion path is required for progenitor outgrowth.Knock-down of bnl expression byRNA interference (RNAi)in larval tracheal cells blocked migration andresulted in diminished or absent PAT (Fig. 3, Band C; fig. S7, A to C; and movie S2). Mosaicexpression of bnl RNAi in small patches along thepath (23) also arrested migration, so long as thepatch encompassed the full DTcircumference (Fig.3D; fig. S7,D andE; andmovies S3 and S4). Thus,Bnl is required all along themigration path, and thesignal does not cross even short gaps.

Ectopic bnl expression in GFP-labeled clonesof larval tracheal cells induced by dfr-FLP (23)redirected progenitor migration. Depending onthe location of the clones, ectopic bnl caused in-correct exit from the niche, premature entry ontotheDT, or wrong turns on theDT (Fig. 4, B toD).Dual clones induced bifurcation with groups ofprogenitorsmoving toward each ectopic bnl source(Fig. 4E). Clones in Tr3 and posterior metamerescaused progenitors in these regions to leave theniche, even though they do not normally do so(Fig. 4, G and H, and fig. S8, D and E). Whenthere was a large clone, progenitors migratedthroughout the clone (Fig. 4F), implying that pro-genitors do not require a gradient and will spreadto cover an entire region of cells expressingbnl at equivalent levels. When bnl-expressingclones failed to induce migration, the clonesappeared to be too far from the progenitors orthere was competition from another clone closeby (fig. S8, A and B). Ectopic bnl expressionwithin the progenitor cluster arrested migration(fig. S8C).

The results show that the embryonic trachealinducer Bnl FGF guides tracheal progenitors outof the niche and into the posterior during trachealmetamorphosis. The source of Bnl is the larvaltracheal branches destined for destruction, whichserve both as the source of the chemoattractantand as the substratum for progenitor migration.Several days earlier in embryos, these larval tra-cheal branches were themselves induced by Bnlprovided by neighboring tissues. But after em-bryonic development, most tracheal cells, includingthose in the decaying larval branches, down-regulate btl FGFR expression (fig. S2A) and thusdo not respond to (or sequester) the Bnl signalthey later express. One of the most notable as-pects of this larval Bnl is its exquisitely specificpattern in decaying larval branches, which pres-ages progenitor outgrowth. It is unclear how Bnlexpression is controlled, though it does not ap-pear to require signals from migrating progenitorsbecause the bnl reporter expression front progressednormallywhen progenitor outgrowthwas stalled bya tracheal break (fig. S6C). Perhaps expression ofBnl involves gradients in the tracheal system orspatial patterning cues established during embry-

onic development in conjunction with temporalsignals mediated by molting hormones.

Because the signal guiding progenitor migra-tion is provided by tracheae destined for destruc-tion, progenitors become positioned along thelarval branches they replace (Fig. 4I). Perhapsduring tissue repair and homeostasis, recruitmentof adult stem or progenitor cells from the niche issimilarly guided by signals from decaying tissue,thereby ensuring that new tissue is directed to theappropriate sites.

References and Notes1. N. Barker, S. Bartfeld, H. Clevers, Cell Stem Cell 7,

656–670 (2010).2. G. B. Adams, D. T. Scadden, Nat. Immunol. 7, 333–337

(2006).3. A. Alvarez-Buylla, D. A. Lim, Neuron 41, 683–686 (2004).4. C. Blanpain, E. Fuchs, Nat. Rev. Mol. Cell Biol. 10,

207–217 (2009).5. E. Sancho, E. Batlle, H. Clevers, Curr. Opin. Cell Biol. 15,

763–770 (2003).6. G. L. Ming, H. Song, Neuron 70, 687–702 (2011).7. E. Nacu, E. M. Tanaka, Annu. Rev. Cell Dev. Biol. 27,

409–440 (2011).8. K. D. Poss, Nat. Rev. Genet. 11, 710–722 (2010).9. T. Matsuno, Jap. J. Appl. Entomol. Zool. 34, 165–167 (1990).10. G. Manning, M. A. Krasnow, in The Development of

Drosophila melanogaster, M. Bate, A. Martinez-Arias, Eds.(Cold Spring Harbor Laboratory Press, Woodbury, NY,1993), vol. 1, pp. 609–685.

11. M. Weaver, M. A. Krasnow, Science 321, 1496–1499(2008).

12. A. Guha, L. Lin, T. B. Kornberg, Proc. Natl. Acad. Sci. U.S.A.105, 10832–10836 (2008).

13. C. Pitsouli, N. Perrimon, Development 137, 3615–3624(2010).

14. M. Sato, Y. Kitada, T. Tabata, Dev. Biol. 318, 247–257 (2008).15. C. Ribeiro, M. Neumann, M. Affolter, Curr. Biol. 14,

2197–2207 (2004).16. L. Liu, W. A. Johnson, M. J. Welsh, Proc. Natl. Acad. Sci.

U.S.A. 100, 2128–2133 (2003).17. K. Guillemin et al., Development 122, 1353–1362 (1996).18. C. Klämbt, L. Glazer, B. Z. Shilo, Genes Dev. 6, 1668–1678

(1992).19. M. Reichman-Fried, B. Z. Shilo, Mech. Dev. 52, 265–273

(1995).20. D. Sutherland, C. Samakovlis, M. A. Krasnow, Cell 87,

1091–1101 (1996).21. J. Jarecki, E. Johnson, M. A. Krasnow, Cell 99, 211–220

(1999).22. M. Sato, T. B. Kornberg, Dev. Cell 3, 195–207

(2002).23. Materials and methods are available as supporting

material on Science Online.24. S. Hayashi et al., Genesis 34, 58–61 (2002).

Acknowledgments: We thank M. Weaver, M. Metzstein, andother lab members for advice and reagents. This work wassupported by a Genentech Graduate Fellowship and aRuth L. Kirschstein NIH training grant (F.C.) and the HowardHughes Medical Institute.

Supplementary Materialswww.sciencemag.org/content/343/6167/186/suppl/DC1Materials and MethodsFigs. S1 to S10References (25–39)Movies S1 to S4

4 June 2013; accepted 12 November 201310.1126/science.1241442

Mutational Analysis Reveals theOrigin and Therapy-Driven Evolutionof Recurrent GliomaBrett E. Johnson,1* Tali Mazor,1* Chibo Hong,1 Michael Barnes,2 Koki Aihara,3,4Cory Y. McLean,1† Shaun D. Fouse,1 Shogo Yamamoto,3 Hiroki Ueda,3 Kenji Tatsuno,3Saurabh Asthana,5,6 Llewellyn E. Jalbert,7 Sarah J. Nelson,7,8 Andrew W. Bollen,2W. Clay Gustafson,9 Elise Charron,10 William A. Weiss,1,9,10 Ivan V. Smirnov,1 Jun S. Song,11,12Adam B. Olshen,6,11 Soonmee Cha,1 Yongjun Zhao,13 Richard A. Moore,13 Andrew J. Mungall,13Steven J. M. Jones,13 Martin Hirst,13 Marco A. Marra,13 Nobuhito Saito,4 Hiroyuki Aburatani,3Akitake Mukasa,4 Mitchel S. Berger,1 Susan M. Chang,1 Barry S. Taylor,5,6,11‡ Joseph F. Costello1‡

Tumor recurrence is a leading cause of cancer mortality. Therapies for recurrent disease may fail, at leastin part, because the genomic alterations driving the growth of recurrences are distinct from those inthe initial tumor. To explore this hypothesis, we sequenced the exomes of 23 initial low-grade gliomasand recurrent tumors resected from the same patients. In 43% of cases, at least half of the mutationsin the initial tumor were undetected at recurrence, including driver mutations in TP53, ATRX, SMARCA4,and BRAF; this suggests that recurrent tumors are often seeded by cells derived from the initial tumor at avery early stage of their evolution. Notably, tumors from 6 of 10 patients treated with the chemotherapeuticdrug temozolomide (TMZ) followed an alternative evolutionary path to high-grade glioma. At recurrence,these tumors were hypermutated and harbored driver mutations in the RB (retinoblastoma) and Akt-mTOR(mammalian target of rapamycin) pathways that bore the signature of TMZ-induced mutagenesis.

The genetic landscape of tumors is contin-ually evolving, which can be an impedi-ment to the clinical management of cancer

patients with recurrent disease (1, 2). In contrastto the clonal evolution of hematological malig-nancies (3, 4) and solid tumor metastases (5–7),

the local regrowth of solid tumors after surgeryoccurs under a unique set of evolutionary pres-sures, which are further affected by adjuvant ther-apies. Through the acquisition of new mutations,residual tumor cells can progress to a more ag-gressive state. Grade II astrocytic gliomas are

www.sciencemag.org SCIENCE VOL 343 10 JANUARY 2014 189

REPORTS

on

May

12,

201

7ht

tp://

scie

nce.

scie

ncem

ag.o

rg/

Dow

nloa

ded

from

Brennan CW (2013) CellJohnson et al. (2014) Science

Lastly, we explored the relationship of molecular subclasseswith clinical parameters such as treatment response or survival.In the current larger TCGA cohort, the survival advantage of pro-neural subtype GBM (Phillips et al., 2006) was definitively shown

to be conferred by G-CIMP status, with non-G-CIMP proneuralGBMs and not mesenchymal GBM tending to show less favor-able outcomes in the first 12 months following initial diagnosiscompared to other subtypes (p value 0.07; Figure S6A). Although

Figure 5. Molecular Subclasses of GBM and their Genomic Molecular Correlates(A) Genomic alterations and survival associated with five molecular subtypes of GBM. Expression and DNAmethylation profiles were used to classify 332 GBMs

with available (native DNA and whole-genome amplified DNA) exome sequencing and DNA copy-number levels. Themost significant genomic associations were

identified through Chi-square tests, with p values corrected for multiple testing using the Benjamini-Hochberg method.

(B) Genomic alterations and sample features associated with six GBM methylation clusters. Epigenomic consensus clustering was performed on 396 GBM

samples profiled across two different platforms (Infinium HM27 and Infinium HM450). Six DNA methylation clusters were identified (see related Figure S5),

represented as M1 to M6, where M5 is G-CIMP. These DNA methylation signatures are correlated with 27 selected features composed of clinical, somatic, and

copy-number alterations; DM cluster, G-CIMP status, four TCGA GBM gene expression subclasses, two clinical features (Age at diagnosis/overall survival in

months), somatic mutations (IDH1, TP53, ATRX), and 18 selected copy-number alterations.

472 Cell 155, 462–477, October 10, 2013 ª2013 Elsevier Inc.

progenitors express bnl>GFP. As progenitorscontinue along the DT, DT larval cells activatebnl>GFP expression one segment at a timefrom anterior to posterior, matching progenitormovement.

This dynamic bnl expression along the migra-tion path is required for progenitor outgrowth.Knock-down of bnl expression byRNA interference (RNAi)in larval tracheal cells blocked migration andresulted in diminished or absent PAT (Fig. 3, Band C; fig. S7, A to C; and movie S2). Mosaicexpression of bnl RNAi in small patches along thepath (23) also arrested migration, so long as thepatch encompassed the full DTcircumference (Fig.3D; fig. S7,D andE; andmovies S3 and S4). Thus,Bnl is required all along themigration path, and thesignal does not cross even short gaps.

Ectopic bnl expression in GFP-labeled clonesof larval tracheal cells induced by dfr-FLP (23)redirected progenitor migration. Depending onthe location of the clones, ectopic bnl caused in-correct exit from the niche, premature entry ontotheDT, or wrong turns on theDT (Fig. 4, B toD).Dual clones induced bifurcation with groups ofprogenitorsmoving toward each ectopic bnl source(Fig. 4E). Clones in Tr3 and posterior metamerescaused progenitors in these regions to leave theniche, even though they do not normally do so(Fig. 4, G and H, and fig. S8, D and E). Whenthere was a large clone, progenitors migratedthroughout the clone (Fig. 4F), implying that pro-genitors do not require a gradient and will spreadto cover an entire region of cells expressingbnl at equivalent levels. When bnl-expressingclones failed to induce migration, the clonesappeared to be too far from the progenitors orthere was competition from another clone closeby (fig. S8, A and B). Ectopic bnl expressionwithin the progenitor cluster arrested migration(fig. S8C).

The results show that the embryonic trachealinducer Bnl FGF guides tracheal progenitors outof the niche and into the posterior during trachealmetamorphosis. The source of Bnl is the larvaltracheal branches destined for destruction, whichserve both as the source of the chemoattractantand as the substratum for progenitor migration.Several days earlier in embryos, these larval tra-cheal branches were themselves induced by Bnlprovided by neighboring tissues. But after em-bryonic development, most tracheal cells, includingthose in the decaying larval branches, down-regulate btl FGFR expression (fig. S2A) and thusdo not respond to (or sequester) the Bnl signalthey later express. One of the most notable as-pects of this larval Bnl is its exquisitely specificpattern in decaying larval branches, which pres-ages progenitor outgrowth. It is unclear how Bnlexpression is controlled, though it does not ap-pear to require signals from migrating progenitorsbecause the bnl reporter expression front progressednormallywhen progenitor outgrowthwas stalled bya tracheal break (fig. S6C). Perhaps expression ofBnl involves gradients in the tracheal system orspatial patterning cues established during embry-

onic development in conjunction with temporalsignals mediated by molting hormones.

Because the signal guiding progenitor migra-tion is provided by tracheae destined for destruc-tion, progenitors become positioned along thelarval branches they replace (Fig. 4I). Perhapsduring tissue repair and homeostasis, recruitmentof adult stem or progenitor cells from the niche issimilarly guided by signals from decaying tissue,thereby ensuring that new tissue is directed to theappropriate sites.

References and Notes1. N. Barker, S. Bartfeld, H. Clevers, Cell Stem Cell 7,

656–670 (2010).2. G. B. Adams, D. T. Scadden, Nat. Immunol. 7, 333–337

(2006).3. A. Alvarez-Buylla, D. A. Lim, Neuron 41, 683–686 (2004).4. C. Blanpain, E. Fuchs, Nat. Rev. Mol. Cell Biol. 10,

207–217 (2009).5. E. Sancho, E. Batlle, H. Clevers, Curr. Opin. Cell Biol. 15,

763–770 (2003).6. G. L. Ming, H. Song, Neuron 70, 687–702 (2011).7. E. Nacu, E. M. Tanaka, Annu. Rev. Cell Dev. Biol. 27,

409–440 (2011).8. K. D. Poss, Nat. Rev. Genet. 11, 710–722 (2010).9. T. Matsuno, Jap. J. Appl. Entomol. Zool. 34, 165–167 (1990).10. G. Manning, M. A. Krasnow, in The Development of

Drosophila melanogaster, M. Bate, A. Martinez-Arias, Eds.(Cold Spring Harbor Laboratory Press, Woodbury, NY,1993), vol. 1, pp. 609–685.

11. M. Weaver, M. A. Krasnow, Science 321, 1496–1499(2008).

12. A. Guha, L. Lin, T. B. Kornberg, Proc. Natl. Acad. Sci. U.S.A.105, 10832–10836 (2008).

13. C. Pitsouli, N. Perrimon, Development 137, 3615–3624(2010).

14. M. Sato, Y. Kitada, T. Tabata, Dev. Biol. 318, 247–257 (2008).15. C. Ribeiro, M. Neumann, M. Affolter, Curr. Biol. 14,

2197–2207 (2004).16. L. Liu, W. A. Johnson, M. J. Welsh, Proc. Natl. Acad. Sci.

U.S.A. 100, 2128–2133 (2003).17. K. Guillemin et al., Development 122, 1353–1362 (1996).18. C. Klämbt, L. Glazer, B. Z. Shilo, Genes Dev. 6, 1668–1678

(1992).19. M. Reichman-Fried, B. Z. Shilo, Mech. Dev. 52, 265–273

(1995).20. D. Sutherland, C. Samakovlis, M. A. Krasnow, Cell 87,

1091–1101 (1996).21. J. Jarecki, E. Johnson, M. A. Krasnow, Cell 99, 211–220

(1999).22. M. Sato, T. B. Kornberg, Dev. Cell 3, 195–207

(2002).23. Materials and methods are available as supporting

material on Science Online.24. S. Hayashi et al., Genesis 34, 58–61 (2002).

Acknowledgments: We thank M. Weaver, M. Metzstein, andother lab members for advice and reagents. This work wassupported by a Genentech Graduate Fellowship and aRuth L. Kirschstein NIH training grant (F.C.) and the HowardHughes Medical Institute.

Supplementary Materialswww.sciencemag.org/content/343/6167/186/suppl/DC1Materials and MethodsFigs. S1 to S10References (25–39)Movies S1 to S4

4 June 2013; accepted 12 November 201310.1126/science.1241442

Mutational Analysis Reveals theOrigin and Therapy-Driven Evolutionof Recurrent GliomaBrett E. Johnson,1* Tali Mazor,1* Chibo Hong,1 Michael Barnes,2 Koki Aihara,3,4Cory Y. McLean,1† Shaun D. Fouse,1 Shogo Yamamoto,3 Hiroki Ueda,3 Kenji Tatsuno,3Saurabh Asthana,5,6 Llewellyn E. Jalbert,7 Sarah J. Nelson,7,8 Andrew W. Bollen,2W. Clay Gustafson,9 Elise Charron,10 William A. Weiss,1,9,10 Ivan V. Smirnov,1 Jun S. Song,11,12Adam B. Olshen,6,11 Soonmee Cha,1 Yongjun Zhao,13 Richard A. Moore,13 Andrew J. Mungall,13Steven J. M. Jones,13 Martin Hirst,13 Marco A. Marra,13 Nobuhito Saito,4 Hiroyuki Aburatani,3Akitake Mukasa,4 Mitchel S. Berger,1 Susan M. Chang,1 Barry S. Taylor,5,6,11‡ Joseph F. Costello1‡

Tumor recurrence is a leading cause of cancer mortality. Therapies for recurrent disease may fail, at leastin part, because the genomic alterations driving the growth of recurrences are distinct from those inthe initial tumor. To explore this hypothesis, we sequenced the exomes of 23 initial low-grade gliomasand recurrent tumors resected from the same patients. In 43% of cases, at least half of the mutationsin the initial tumor were undetected at recurrence, including driver mutations in TP53, ATRX, SMARCA4,and BRAF; this suggests that recurrent tumors are often seeded by cells derived from the initial tumor at avery early stage of their evolution. Notably, tumors from 6 of 10 patients treated with the chemotherapeuticdrug temozolomide (TMZ) followed an alternative evolutionary path to high-grade glioma. At recurrence,these tumors were hypermutated and harbored driver mutations in the RB (retinoblastoma) and Akt-mTOR(mammalian target of rapamycin) pathways that bore the signature of TMZ-induced mutagenesis.

The genetic landscape of tumors is contin-ually evolving, which can be an impedi-ment to the clinical management of cancer

patients with recurrent disease (1, 2). In contrastto the clonal evolution of hematological malig-nancies (3, 4) and solid tumor metastases (5–7),

the local regrowth of solid tumors after surgeryoccurs under a unique set of evolutionary pres-sures, which are further affected by adjuvant ther-apies. Through the acquisition of new mutations,residual tumor cells can progress to a more ag-gressive state. Grade II astrocytic gliomas are

www.sciencemag.org SCIENCE VOL 343 10 JANUARY 2014 189

REPORTS

on

May

12,

201

7ht

tp://

scie

nce.

scie

ncem

ag.o

rg/

Dow

nloa

ded

from

Hadjipanayis CG & Van Meir mod.

GSC

GSC

GSC X

XX

X

X

X

GSC

XX

X

GSC

GBM Chemoradiation Recurrence

3-6 months

Brennan CW (2013) CellJohnson et al. (2014) Science

Lastly, we explored the relationship of molecular subclasseswith clinical parameters such as treatment response or survival.In the current larger TCGA cohort, the survival advantage of pro-neural subtype GBM (Phillips et al., 2006) was definitively shown

to be conferred by G-CIMP status, with non-G-CIMP proneuralGBMs and not mesenchymal GBM tending to show less favor-able outcomes in the first 12 months following initial diagnosiscompared to other subtypes (p value 0.07; Figure S6A). Although

Figure 5. Molecular Subclasses of GBM and their Genomic Molecular Correlates(A) Genomic alterations and survival associated with five molecular subtypes of GBM. Expression and DNAmethylation profiles were used to classify 332 GBMs

with available (native DNA and whole-genome amplified DNA) exome sequencing and DNA copy-number levels. Themost significant genomic associations were

identified through Chi-square tests, with p values corrected for multiple testing using the Benjamini-Hochberg method.

(B) Genomic alterations and sample features associated with six GBM methylation clusters. Epigenomic consensus clustering was performed on 396 GBM

samples profiled across two different platforms (Infinium HM27 and Infinium HM450). Six DNA methylation clusters were identified (see related Figure S5),

represented as M1 to M6, where M5 is G-CIMP. These DNA methylation signatures are correlated with 27 selected features composed of clinical, somatic, and

copy-number alterations; DM cluster, G-CIMP status, four TCGA GBM gene expression subclasses, two clinical features (Age at diagnosis/overall survival in

months), somatic mutations (IDH1, TP53, ATRX), and 18 selected copy-number alterations.

472 Cell 155, 462–477, October 10, 2013 ª2013 Elsevier Inc.

progenitors express bnl>GFP. As progenitorscontinue along the DT, DT larval cells activatebnl>GFP expression one segment at a timefrom anterior to posterior, matching progenitormovement.

This dynamic bnl expression along the migra-tion path is required for progenitor outgrowth.Knock-down of bnl expression byRNA interference (RNAi)in larval tracheal cells blocked migration andresulted in diminished or absent PAT (Fig. 3, Band C; fig. S7, A to C; and movie S2). Mosaicexpression of bnl RNAi in small patches along thepath (23) also arrested migration, so long as thepatch encompassed the full DTcircumference (Fig.3D; fig. S7,D andE; andmovies S3 and S4). Thus,Bnl is required all along themigration path, and thesignal does not cross even short gaps.

Ectopic bnl expression in GFP-labeled clonesof larval tracheal cells induced by dfr-FLP (23)redirected progenitor migration. Depending onthe location of the clones, ectopic bnl caused in-correct exit from the niche, premature entry ontotheDT, or wrong turns on theDT (Fig. 4, B toD).Dual clones induced bifurcation with groups ofprogenitorsmoving toward each ectopic bnl source(Fig. 4E). Clones in Tr3 and posterior metamerescaused progenitors in these regions to leave theniche, even though they do not normally do so(Fig. 4, G and H, and fig. S8, D and E). Whenthere was a large clone, progenitors migratedthroughout the clone (Fig. 4F), implying that pro-genitors do not require a gradient and will spreadto cover an entire region of cells expressingbnl at equivalent levels. When bnl-expressingclones failed to induce migration, the clonesappeared to be too far from the progenitors orthere was competition from another clone closeby (fig. S8, A and B). Ectopic bnl expressionwithin the progenitor cluster arrested migration(fig. S8C).

The results show that the embryonic trachealinducer Bnl FGF guides tracheal progenitors outof the niche and into the posterior during trachealmetamorphosis. The source of Bnl is the larvaltracheal branches destined for destruction, whichserve both as the source of the chemoattractantand as the substratum for progenitor migration.Several days earlier in embryos, these larval tra-cheal branches were themselves induced by Bnlprovided by neighboring tissues. But after em-bryonic development, most tracheal cells, includingthose in the decaying larval branches, down-regulate btl FGFR expression (fig. S2A) and thusdo not respond to (or sequester) the Bnl signalthey later express. One of the most notable as-pects of this larval Bnl is its exquisitely specificpattern in decaying larval branches, which pres-ages progenitor outgrowth. It is unclear how Bnlexpression is controlled, though it does not ap-pear to require signals from migrating progenitorsbecause the bnl reporter expression front progressednormallywhen progenitor outgrowthwas stalled bya tracheal break (fig. S6C). Perhaps expression ofBnl involves gradients in the tracheal system orspatial patterning cues established during embry-

onic development in conjunction with temporalsignals mediated by molting hormones.

Because the signal guiding progenitor migra-tion is provided by tracheae destined for destruc-tion, progenitors become positioned along thelarval branches they replace (Fig. 4I). Perhapsduring tissue repair and homeostasis, recruitmentof adult stem or progenitor cells from the niche issimilarly guided by signals from decaying tissue,thereby ensuring that new tissue is directed to theappropriate sites.

References and Notes1. N. Barker, S. Bartfeld, H. Clevers, Cell Stem Cell 7,

656–670 (2010).2. G. B. Adams, D. T. Scadden, Nat. Immunol. 7, 333–337

(2006).3. A. Alvarez-Buylla, D. A. Lim, Neuron 41, 683–686 (2004).4. C. Blanpain, E. Fuchs, Nat. Rev. Mol. Cell Biol. 10,

207–217 (2009).5. E. Sancho, E. Batlle, H. Clevers, Curr. Opin. Cell Biol. 15,

763–770 (2003).6. G. L. Ming, H. Song, Neuron 70, 687–702 (2011).7. E. Nacu, E. M. Tanaka, Annu. Rev. Cell Dev. Biol. 27,

409–440 (2011).8. K. D. Poss, Nat. Rev. Genet. 11, 710–722 (2010).9. T. Matsuno, Jap. J. Appl. Entomol. Zool. 34, 165–167 (1990).10. G. Manning, M. A. Krasnow, in The Development of

Drosophila melanogaster, M. Bate, A. Martinez-Arias, Eds.(Cold Spring Harbor Laboratory Press, Woodbury, NY,1993), vol. 1, pp. 609–685.

11. M. Weaver, M. A. Krasnow, Science 321, 1496–1499(2008).

12. A. Guha, L. Lin, T. B. Kornberg, Proc. Natl. Acad. Sci. U.S.A.105, 10832–10836 (2008).

13. C. Pitsouli, N. Perrimon, Development 137, 3615–3624(2010).

14. M. Sato, Y. Kitada, T. Tabata, Dev. Biol. 318, 247–257 (2008).15. C. Ribeiro, M. Neumann, M. Affolter, Curr. Biol. 14,

2197–2207 (2004).16. L. Liu, W. A. Johnson, M. J. Welsh, Proc. Natl. Acad. Sci.

U.S.A. 100, 2128–2133 (2003).17. K. Guillemin et al., Development 122, 1353–1362 (1996).18. C. Klämbt, L. Glazer, B. Z. Shilo, Genes Dev. 6, 1668–1678

(1992).19. M. Reichman-Fried, B. Z. Shilo, Mech. Dev. 52, 265–273

(1995).20. D. Sutherland, C. Samakovlis, M. A. Krasnow, Cell 87,

1091–1101 (1996).21. J. Jarecki, E. Johnson, M. A. Krasnow, Cell 99, 211–220

(1999).22. M. Sato, T. B. Kornberg, Dev. Cell 3, 195–207

(2002).23. Materials and methods are available as supporting

material on Science Online.24. S. Hayashi et al., Genesis 34, 58–61 (2002).

Acknowledgments: We thank M. Weaver, M. Metzstein, andother lab members for advice and reagents. This work wassupported by a Genentech Graduate Fellowship and aRuth L. Kirschstein NIH training grant (F.C.) and the HowardHughes Medical Institute.

Supplementary Materialswww.sciencemag.org/content/343/6167/186/suppl/DC1Materials and MethodsFigs. S1 to S10References (25–39)Movies S1 to S4

4 June 2013; accepted 12 November 201310.1126/science.1241442

Mutational Analysis Reveals theOrigin and Therapy-Driven Evolutionof Recurrent GliomaBrett E. Johnson,1* Tali Mazor,1* Chibo Hong,1 Michael Barnes,2 Koki Aihara,3,4Cory Y. McLean,1† Shaun D. Fouse,1 Shogo Yamamoto,3 Hiroki Ueda,3 Kenji Tatsuno,3Saurabh Asthana,5,6 Llewellyn E. Jalbert,7 Sarah J. Nelson,7,8 Andrew W. Bollen,2W. Clay Gustafson,9 Elise Charron,10 William A. Weiss,1,9,10 Ivan V. Smirnov,1 Jun S. Song,11,12Adam B. Olshen,6,11 Soonmee Cha,1 Yongjun Zhao,13 Richard A. Moore,13 Andrew J. Mungall,13Steven J. M. Jones,13 Martin Hirst,13 Marco A. Marra,13 Nobuhito Saito,4 Hiroyuki Aburatani,3Akitake Mukasa,4 Mitchel S. Berger,1 Susan M. Chang,1 Barry S. Taylor,5,6,11‡ Joseph F. Costello1‡

Tumor recurrence is a leading cause of cancer mortality. Therapies for recurrent disease may fail, at leastin part, because the genomic alterations driving the growth of recurrences are distinct from those inthe initial tumor. To explore this hypothesis, we sequenced the exomes of 23 initial low-grade gliomasand recurrent tumors resected from the same patients. In 43% of cases, at least half of the mutationsin the initial tumor were undetected at recurrence, including driver mutations in TP53, ATRX, SMARCA4,and BRAF; this suggests that recurrent tumors are often seeded by cells derived from the initial tumor at avery early stage of their evolution. Notably, tumors from 6 of 10 patients treated with the chemotherapeuticdrug temozolomide (TMZ) followed an alternative evolutionary path to high-grade glioma. At recurrence,these tumors were hypermutated and harbored driver mutations in the RB (retinoblastoma) and Akt-mTOR(mammalian target of rapamycin) pathways that bore the signature of TMZ-induced mutagenesis.

The genetic landscape of tumors is contin-ually evolving, which can be an impedi-ment to the clinical management of cancer

patients with recurrent disease (1, 2). In contrastto the clonal evolution of hematological malig-nancies (3, 4) and solid tumor metastases (5–7),

the local regrowth of solid tumors after surgeryoccurs under a unique set of evolutionary pres-sures, which are further affected by adjuvant ther-apies. Through the acquisition of new mutations,residual tumor cells can progress to a more ag-gressive state. Grade II astrocytic gliomas are

www.sciencemag.org SCIENCE VOL 343 10 JANUARY 2014 189

REPORTS

on

May

12,

201

7ht

tp://

scie

nce.

scie

ncem

ag.o

rg/

Dow

nloa

ded

from

Hadjipanayis CG & Van Meir mod.

GSC

GSC

GSC X

XX

X

X

X

GSC

XX

X

GSC

GBM Chemoradiation Recurrence

3-6 months

Brennan CW (2013) CellJohnson et al. (2014) Science

et al., 1994), breast (Al-Hajj et al., 2003), brain (Galli et al., 2004;Lee et al., 2006; Singh et al., 2004), melanoma (Schatton et al.,2008), colon (O’Brien et al., 2007), and pancreatic cancer(Hermann et al., 2007), and the presence of these tumor subsetscorrelate strongly with tumor recurrence and treatment failure.Elucidation of the CSC niche and regulators of their biologyhas been a subject of intense research. Initial reports on CSCsargued for an intimate relationship with tumor vasculature, sug-gesting a perivascular niche like that seen in NSCs (Figure 2)(Calabrese et al., 2007). This observation has gained wideracceptance, especially with the recent identification of integrina6 (Lathia et al., 2010) and nitric oxide (NO) (Charles et al.,2010) as potent regulators of glioma CSCs. Holland andcolleagues have demonstrated that endothelial nitric oxidesynthase (eNOS) is highly expressed in tumor vasculature, andthe elaboration of NO by vascular endothelial cells can regulateadjacent Nestin- and Notch-expressing glioma cells as evi-denced by an increase in neurosphere-forming capacity in vitroand enhancement of tumorigenic capacity in vivo (Charleset al., 2010). Interestingly, eNOS-deficient mice demonstratedimproved survival in a PDGF-induced glioma model (Charleset al., 2010).Despite a well-established perivascular niche, CSCs have

recently been proposed to exist in a secondary niche withincancers that, ironically, is further away from vasculature and,as a consequence, is more hypoxic (Li et al., 2009). Hypoxicmicroenvironments within tumors have long been appreciatedto be a product of aberrant vasculature and due to a rapidlydividing tumor mass that outstrips its vascular supply (Figure 2)(Carmeliet and Jain, 2000; Pouyssegur et al., 2006). Zones ofnecrosis can readily be seen in rapidly dividing cancers, andhypoxia has been associated with treatment resistance, localinvasion, and poor clinical outcome (Gordan and Simon, 2007;Keith and Simon, 2007; Schindl et al., 2002; Zhong et al.,1999). In addition, the undifferentiated phenotype of solid tumorsseen often in neuroblastoma, breast, and cervical cancersstrongly correlates with tumor hypoxia (Axelson et al., 2005).Since HIFs, and in particular HIF-2a, have been shown to regu-

late signaling pathways that regulate stem cell self-renewal andmultipotency, it was hypothesized that tumor hypoxia may serve

as another stimulus to regulate CSCs and ultimately influencetumor progression by the direct activation of this transcriptionfactor (Keith and Simon, 2007). Direct evidence for this modelwas recently provided by Jeremy Rich’s group, who demon-strated that CD133+ CSCs from gliomas preferentially expressedHIF-2a and multiple HIF-regulated genes when compared tonon-stem tumor cells and normal neural progenitors (Heddlestonet al., 2009; Li et al., 2009). HIF-2a is a critical and uniquetranscriptional regulator in brain tumor stem cells that, whensilenced, leads to the inhibition of self-renewal, proliferation,and survival in vivo with dramatic attenuation of tumor initiationpotential of glioma CSCs (Li et al., 2009).The discovery of HIF-2a regulation of Oct4, as well as several

downstream targets that maintain stem cell self renewal, has ledto a considerable amount of interest in the role of this transcrip-tion factor in stem cells (Covello et al., 2006). Recently, conver-gence of several observations supports a more prominentoncogenic role of HIF-2a in several genetically diverse cancersthat may exploit this signaling cascade (Franovic et al., 2009;Li et al., 2009). Franovic et al. demonstrated that inhibitionof HIF-2a in highly aggressive glioblastoma, colorectal, andnon-small lung carcinomas, cancers that have been shown toharbor CSC populations, prevented in vivo growth and tumori-genesis regardless of their mutational status and tissue of origin(Franovic et al., 2009). NO has been shown to activate HIF inseveral known cell lines in vitro (Berchner-Pfannschmidt et al.,2007); thus, it will be interesting to see if HIF is indeed activatedby NO in the perivascular niche, despite adequate levels ofoxygen in this region. HIF-2a has already been shown to induceeNOS (Coulet et al., 2003), which makes the hypothesis that anautocrine loop that regulates the CSC phenotype in hypoxicniches both plausible and testable.Notch signaling has emerged as a critical regulator of stem

cells (Gustafsson et al., 2005). This relationship has led investiga-tors to examine its role in CSCs and, more specifically, in epithe-lial-to-mesenchymal transition (EMT) within tumors (Sahlgrenet al., 2008). For example, blockade of Notch signaling withgamma-secretase inhibitors depletes cancer stem-like cells ingliomas and inhibits growth of xenografts (Fan et al., 2010).In addition, it has been proposed that the initial steps in local

Figure 2. Cancer Stem Cell NicheNestin- and Notch-expressing glioma-initiatingcells (cancer stem cells) were initially describedto reside in a perivascular niche around tumorvasculature. Recent evidence suggests that NOproduced by endothelial cells maintains thecancer stem cell phenotype. Inhibition of thissignaling pathway results in loss of neurosphereforming capacity and attenuation of tumorigenicforming capacity in vivo. A secondary niche moredistal from the vasculature exhibits lower oxygentension and has also been shown to regulate thecancer stem cell phenotype. Adjacent to the rimof necrotic cells, this hypoxic niche containscancer stem cells whose activity is modulated bymultiple HIF-regulated genes, such as HIF-2aand HIF-1a, and other signaling molecules (Oct4,VEGF, Notch, and c-myc). Recent evidencesuggests that cancer stem cells differentiallyrespond to hypoxia with distinct HIF inductionpatterns. Targeted inhibition of HIF-2a inhibitsself-renewal, proliferation, and survival in vitroand attenuates tumor initiation potential.

Cell Stem Cell 7, August 6, 2010 ª2010 Elsevier Inc. 155

Cell Stem Cell

Review

World Health Organization Classification of Tumours

International Agency for Research on Cancer (IARC)

Revised 4th Edition

WHO Classification of Tumours ofthe Central Nervous System

Edited by

David N. Louis Hiroko

Ohgaki Otmar D.

Wiestler Webster K.

Cavenee

International Agency for Research on Cancer

Lyon, 2016

WHO OMS

Christians et al. (2019) Acta Neuropathol

Glioblastoma, IDH-mutant Ohgaki H.Kleihues P.von Deimling A.Louis D.N.

Reifenberger G.Yan H.Weller M.

Fig. 1.50 Genetic pathways to IDH-wildtype and IDH-mutant glioblastoma. This chart is based on the hypothesis

that IDH-mutant glioblastomas share common glial progenitor cells not only with diffuse and anaplastic astrocytomas,

but also with oligodendrogliomas and anaplastic oligodendrogliomas. Adapted from Ohgaki H and Kleihues P {1830}.

DefinitionA high-grade glioma with predominantlyastrocytic differentiation; featuring nucle-ar atypia, cellular pleomorphism (in mostcases), mitotic activity, and typically a dif-fuse growth pattern, as well as microvas-cular proliferation and/or necrosis; with amutation in either the IDH1 or IDH2 gene.IDH-mutant glioblastomas account forapproximately 10% of all glioblastomas.Glioblastomas that develop throughmalignant progression from diffuse as-trocytoma (WHO grade II) or anaplasticastrocytoma (WHO grade III) are almostalways associated with IDH mutation andtherefore carry the synonym “secondaryglioblastoma, IDH-mutant”. IDH-mutantglioblastoma is morphologically indistin-guishable from IDH-wildtype glioblas-toma, except for a lesser extent of necro-sis. IDH-mutant glioblastomas manifestin younger patients (with a mean patientage at diagnosis of 45 years), are pref-erentially located in the frontal lobe, andcarry a significantly better prognosis thanIDH-wildtype glioblastomas {1797,2810}.

ICD-0 code 9445/3

GradingIDH-mutant glioblastoma correspondshistologically to WHO grade IV.The prognosis of IDH-mutant glioblasto-ma is considerably better than that of IDH-wildtype glioblastoma; however, WHO

grading reflects the natural course of thedisease rather than response to therapy.Therefore, because most patients even-tually succumb to high-grade disease,IDH-mutant glioblastomas are designat-ed as WHO grade IV.Synonym

Secondary glioblastoma, IDH-mutant

EpidemiologyIncidenceUntil the discovery of IDH1 mutation asa molecular marker, the diagnosis ofsecondary glioblastomas was based onclinical observations (i.e. neuroimagingand/or histological evidence of a preced-ing low-grade or anaplastic astrocytoma){1823,1827}. In a population-based studyin Switzerland, using clinical criteria andhistopathological evidence, only 5% of allglioblastomas diagnosed were second-ary {1823,1826}. Similarly, another studyshowed that 19 of 392 cases {5%) of

glioblastoma had a histologically provenprior low-grade glioma {605}. When IDH1mutations were used as a genetic mark-er, secondary glioblastomas accountedfor 9% of all glioblastomas at the popula-tion level {1797} and for 6-13% of casesin hospital-based studies {118,1078,1417,2810}.Age distributionAt the population level, secondary glio-blastomas develop in patients signifi-cantly younger (mean: 45 years) than doprimary glioblastomas (mean: 62 years){1823,1826}. Correspondingly, the meanage of patients with /DHI-mutant glio-blastoma is 48 years, significantly youn-ger than that of patients with glioblasto-mas that lack IDH1 mutations {61 years){1797}. Several hospital-based studieshave also shown that patients with IDH-mutant glioblastoma were significantlyyounger than those with IDH-wildtypeglioblastoma {214,1078,2810}.

52 Diffuse gliomas

Genetic hallmarksThe defining genetic hallmark is thepresence of IDH mutations {1895},which are associated with a hyper-methylation phenotype {1810}. Thesemutations are the earliest detectablegenetic alteration in precursor low-grade diffuse astrocytoma, indicatingthat these tumours are derived fromcommon neural precursor cells thatdiffer from those of IDH-wildtype glio-blastoma. Additional typical geneticalterations are TP53 and ATRX muta-tions and loss of chromosome arm10q {1822,1830}.

World Health Organization Classification of Tumours

International Agency for Research on Cancer (IARC)

Revised 4th Edition

WHO Classification of Tumours ofthe Central Nervous System

Edited by

David N. Louis Hiroko

Ohgaki Otmar D.

Wiestler Webster K.

Cavenee

International Agency for Research on Cancer

Lyon, 2016

WHO OMS

Christians et al. (2019) Acta Neuropathol

World Health Organization Classification of Tumours

International Agency for Research on Cancer (IARC)

Revised 4th Edition

WHO Classification of Tumours ofthe Central Nervous System

Edited by

David N. Louis Hiroko

Ohgaki Otmar D.

Wiestler Webster K.

Cavenee

International Agency for Research on Cancer

Lyon, 2016

WHO OMS

Christians et al. (2019) Acta Neuropathol

World Health Organization Classification of Tumours

International Agency for Research on Cancer (IARC)

Revised 4th Edition

WHO Classification of Tumours ofthe Central Nervous System

Edited by

David N. Louis Hiroko

Ohgaki Otmar D.

Wiestler Webster K.

Cavenee

International Agency for Research on Cancer

Lyon, 2016

WHO OMS

Christians et al. (2019) Acta Neuropathol

et al., 1994), breast (Al-Hajj et al., 2003), brain (Galli et al., 2004;Lee et al., 2006; Singh et al., 2004), melanoma (Schatton et al.,2008), colon (O’Brien et al., 2007), and pancreatic cancer(Hermann et al., 2007), and the presence of these tumor subsetscorrelate strongly with tumor recurrence and treatment failure.Elucidation of the CSC niche and regulators of their biologyhas been a subject of intense research. Initial reports on CSCsargued for an intimate relationship with tumor vasculature, sug-gesting a perivascular niche like that seen in NSCs (Figure 2)(Calabrese et al., 2007). This observation has gained wideracceptance, especially with the recent identification of integrina6 (Lathia et al., 2010) and nitric oxide (NO) (Charles et al.,2010) as potent regulators of glioma CSCs. Holland andcolleagues have demonstrated that endothelial nitric oxidesynthase (eNOS) is highly expressed in tumor vasculature, andthe elaboration of NO by vascular endothelial cells can regulateadjacent Nestin- and Notch-expressing glioma cells as evi-denced by an increase in neurosphere-forming capacity in vitroand enhancement of tumorigenic capacity in vivo (Charleset al., 2010). Interestingly, eNOS-deficient mice demonstratedimproved survival in a PDGF-induced glioma model (Charleset al., 2010).Despite a well-established perivascular niche, CSCs have

recently been proposed to exist in a secondary niche withincancers that, ironically, is further away from vasculature and,as a consequence, is more hypoxic (Li et al., 2009). Hypoxicmicroenvironments within tumors have long been appreciatedto be a product of aberrant vasculature and due to a rapidlydividing tumor mass that outstrips its vascular supply (Figure 2)(Carmeliet and Jain, 2000; Pouyssegur et al., 2006). Zones ofnecrosis can readily be seen in rapidly dividing cancers, andhypoxia has been associated with treatment resistance, localinvasion, and poor clinical outcome (Gordan and Simon, 2007;Keith and Simon, 2007; Schindl et al., 2002; Zhong et al.,1999). In addition, the undifferentiated phenotype of solid tumorsseen often in neuroblastoma, breast, and cervical cancersstrongly correlates with tumor hypoxia (Axelson et al., 2005).Since HIFs, and in particular HIF-2a, have been shown to regu-

late signaling pathways that regulate stem cell self-renewal andmultipotency, it was hypothesized that tumor hypoxia may serve

as another stimulus to regulate CSCs and ultimately influencetumor progression by the direct activation of this transcriptionfactor (Keith and Simon, 2007). Direct evidence for this modelwas recently provided by Jeremy Rich’s group, who demon-strated that CD133+ CSCs from gliomas preferentially expressedHIF-2a and multiple HIF-regulated genes when compared tonon-stem tumor cells and normal neural progenitors (Heddlestonet al., 2009; Li et al., 2009). HIF-2a is a critical and uniquetranscriptional regulator in brain tumor stem cells that, whensilenced, leads to the inhibition of self-renewal, proliferation,and survival in vivo with dramatic attenuation of tumor initiationpotential of glioma CSCs (Li et al., 2009).The discovery of HIF-2a regulation of Oct4, as well as several

downstream targets that maintain stem cell self renewal, has ledto a considerable amount of interest in the role of this transcrip-tion factor in stem cells (Covello et al., 2006). Recently, conver-gence of several observations supports a more prominentoncogenic role of HIF-2a in several genetically diverse cancersthat may exploit this signaling cascade (Franovic et al., 2009;Li et al., 2009). Franovic et al. demonstrated that inhibitionof HIF-2a in highly aggressive glioblastoma, colorectal, andnon-small lung carcinomas, cancers that have been shown toharbor CSC populations, prevented in vivo growth and tumori-genesis regardless of their mutational status and tissue of origin(Franovic et al., 2009). NO has been shown to activate HIF inseveral known cell lines in vitro (Berchner-Pfannschmidt et al.,2007); thus, it will be interesting to see if HIF is indeed activatedby NO in the perivascular niche, despite adequate levels ofoxygen in this region. HIF-2a has already been shown to induceeNOS (Coulet et al., 2003), which makes the hypothesis that anautocrine loop that regulates the CSC phenotype in hypoxicniches both plausible and testable.Notch signaling has emerged as a critical regulator of stem

cells (Gustafsson et al., 2005). This relationship has led investiga-tors to examine its role in CSCs and, more specifically, in epithe-lial-to-mesenchymal transition (EMT) within tumors (Sahlgrenet al., 2008). For example, blockade of Notch signaling withgamma-secretase inhibitors depletes cancer stem-like cells ingliomas and inhibits growth of xenografts (Fan et al., 2010).In addition, it has been proposed that the initial steps in local

Figure 2. Cancer Stem Cell NicheNestin- and Notch-expressing glioma-initiatingcells (cancer stem cells) were initially describedto reside in a perivascular niche around tumorvasculature. Recent evidence suggests that NOproduced by endothelial cells maintains thecancer stem cell phenotype. Inhibition of thissignaling pathway results in loss of neurosphereforming capacity and attenuation of tumorigenicforming capacity in vivo. A secondary niche moredistal from the vasculature exhibits lower oxygentension and has also been shown to regulate thecancer stem cell phenotype. Adjacent to the rimof necrotic cells, this hypoxic niche containscancer stem cells whose activity is modulated bymultiple HIF-regulated genes, such as HIF-2aand HIF-1a, and other signaling molecules (Oct4,VEGF, Notch, and c-myc). Recent evidencesuggests that cancer stem cells differentiallyrespond to hypoxia with distinct HIF inductionpatterns. Targeted inhibition of HIF-2a inhibitsself-renewal, proliferation, and survival in vitroand attenuates tumor initiation potential.

Cell Stem Cell 7, August 6, 2010 ª2010 Elsevier Inc. 155

Cell Stem Cell

Review

Hypoxia markersʼ expression and IDH status: a possible new prognostic profile in GBM