npg
© 2013 Nature America, Inc. All rights reserved.
news and views of cancer types8–10. Consequently, the complex has been identified as the most frequently mutated chromatin regulatory complex in cancer and one of the most frequently mutated targets overall. The gene encoding SNF5 (SMARCB1) is mutated in pediatric rhabdoid tumors and in a few other rare cancers, but it is not widely mutated across cancer types8. However, mechanisms that indirectly disrupt SNF5 are now being identified in additional types of cancer. For instance, a recent study has shown that the SS18-SSX fusion in synovial sarcoma promotes tumorigenesis by disrupting SNF5 integration into the SWI/SNF complex11. The work of Prensner et al.7 now illustrates how lncRNA-mediated disruption of this SWI/SNF complex subunit can also drive malignancy. The new findings of Prensner et al.7 suggest that SChLAP1 may specifically affect the targeting mechanism of the SWI/SNF complex, a function that remains poorly understood. Although decreased SNF5 binding was observed at some promoters, leading to decreased expression of downstream genes, the specific effect of reduced SNF5 binding
on epigenetic regulation needs to be explored further. The SWI/SNF complex is canonically described as a regulator of gene expression via remodeling of nucleosomes, especially at promoters12,13. Thus, it is possible that SChLAP1 expression leads to disordered nucleosomes or other changes in chromatin structure at proximal promoter sites. Taking these results together with those for EZH2 and BAF57, another SWI/SNF subunit, a theme may be emerging in which aberrantly expressed epigenetic regulators contribute to aggressive prostate cancers14,15. SChLAP1 can be added to a growing list of lncRNAs associated with prostate cancer6. In addition, the ability of SChLAP1 to inhibit SNF5 binding further emphasizes the central role that chromatin regulators have in driving cancer. In essence, SChLAP1 expression in prostate cancer ties together two frontiers in cancer biology: noncoding RNAs and chromatin regulators. By uncovering the interaction between SChLAP1 and SNF5, this new study demonstrates both a mechanism through which lncRNAs can act to drive malignancy
and an additional mechanism by which SWI/ SNF activity can be disrupted in a way that affects cancer growth. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. 1. Siegel, R., Naishadham, D. & Jemal, A. CA Cancer J. Clin. 63, 11–30 (2013). 2. Varambally, S. et al. Nature 419, 624–629 (2002). 3. Prensner, J.R. & Chinnaiyan, A.M. Cancer Discov. 1, 391–407 (2011). 4. Du, Z. et al. Nat. Struct. Mol. Biol. 20, 908–913 (2013). 5. Prensner, J.R. et al. Nat. Biotechnol. 29, 742–749 (2011). 6. Yang, L. et al. Nature 500, 598–602 (2013). 7. Prensner, J.R. et al. Nat. Genet. 45, 1392–1398 (2013). 8. Wilson, B.G. & Roberts, C.W.M. Nat. Rev. Cancer 11, 481–492 (2011). 9. Lee, R.S. et al. J. Clin. Invest. 122, 2983–2988 (2012). 10. Kadoch, C. et al. Nat. Genet. 45, 592–601 (2013). 11. Kadoch, C. & Crabtree, G.R. Cell 153, 71–85 (2013). 12. You, J.S. et al. PLoS Genet. 9, e1003459 (2013). 13. Tolstorukov, M.Y. et al. Proc. Natl. Acad. Sci. USA 110, 10165–10170 (2013). 14. Link, K.A. et al. Cancer Res. 68, 4551–4558 (2008). 15. Balasubramaniam, S. et al. Clin. Cancer Res. 19, 2657–2667 (2013).
Mining the epigenetic landscape in ALL Lindsay M LaFave & Ross L Levine The significance of epigenomic aberrations in cancer development has been underscored by the discovery of mutations in key chromatin modifiers, most notably in hematological malignancies. A new study of pediatric acute lymphoblastic leukemia (ALL) demonstrates the usefulness of mapping global epigenetic signatures and applying these data in a framework to identify and characterize underlying somatic genetic alterations in human cancers. Epigenetic dysregulation has become a cardinal feature of human malignancies, warranting the development of innovative approaches to identify epigenetic driver alterations. Numerous large-scale sequencing studies, including targeted exome, transcriptome and whole genome, have led to the identification of mutations or expression changes in genes critical for normal Lindsay M. LaFave and Ross L. Levine are affiliated with the Human Oncology and Pathogenesis Program, Memorial SloanKettering Cancer Center, New York, New York, USA. Lindsay M. LaFave is also affiliated with Gerstner Sloan-Kettering Graduate School, Memorial Sloan-Kettering Cancer Center, New York, New York, USA. Ross L. Levine is also affiliated with the Leukemia Service, Department of Medicine, Memorial SloanKettering Cancer Center, New York, New York, USA. e-mail: [email protected]
epigenetic regulation in a spectrum of cancer types, including in hematological malignancies. For example, oncogenic mutations in histone methyltransferases that modify key lysine residues on histone tails have been identified in leukemias and lymphomas. These findings have led to the development of selective small molecule histone methyltransferase inhibitors for therapeutic applications. Selective inhibitors targeting gain-of-function alterations in EZH2 (refs. 1,2), the catalytic methyltransferase in the PRC2 complex that mediates repressive trimethylation of histone H3 at lysine 27 (H3K27me3), and DOT1L3, the methyltransferase that methylates histone H3 at lysine 79 (H3K79), are being tested in EZH2-mutant B cell lymphoma and MLL-rearranged leukemia, respectively. The successful development of EZH2 and DOT1L inhibitors has driven efforts to assess mutations in histone methyltransferases that drive cancer development and progression and represent tractable therapeutic targets. In this
nature genetics | volume 45 | number 11 | NOVEMBER 2013
issue of Nature Genetics, Frank Stegmeier, Levi Garraway and colleagues describe an innovative approach coupling histone mass spectroscopy analyses with mutational data for the identification of genetic alterations in the NSD2 gene, which encodes a H3K36 methyltransferase4. Global chromatin profiling Stegmeier and colleagues used a tandem mass spectroscopy strategy to measure chromatin marks on extracted histone proteins, a technique they term ‘global chromatin profiling’. Mass spectrometry circumvents the use of antibodies for the analysis of chromatin marks, which is advantageous given that the lack of specific antibodies for many histone marks remains a significant challenge for the field. Whereas a typical study design for the identification of driver mutations involves elucidating genetic alterations by sequencing and then studying the mechanisms by which these alterations contribute to disease pathogenesis, Stegmeier 1269
news and views
H2B
H2A
H4
H3
ALQKRPAK AKT GG RA K TS S HP A KR ATQKT R A P K K VG GTA 36
npg
© 2013 Nature America, Inc. All rights reserved.
Katie Vicari
t(4;14)
27
NSD2 E1099K
H3K27me3 H3K36me2 Unmethylated H3K36
Figure 1 Schematic of the chromatin signature that was studied by Stegmeier and colleagues. Modifications of interest in the signature, H3K36me2 and H3K27me3, are highlighted on the histone H3 tail sequence. Methyl groups are coded on the basis of whether their levels are increased (H3K36me2; reds) or decreased (H3K27me3 and unmethylated H3K36; blues). This histone pattern was unique to cell lines with NSD2 alterations, either the t(4;14) translocation or the NSD2 mutation encoding Glu1099Lys. The t(4;14) translocation is common to multiple myeloma, whereas the NSD2 mutation was predominantly found in ALL cell lines and patient samples. In the representation of clustering output, genomic alterations are identified on the top, with altered chromatin modifications shown on the left (blue indicates depletion and red indicates enrichment in the signature).
and colleagues used an alternate approach. They identified cell lines with similar epigenetic patterns and then searched for somatic alterations that would be predicted to cause the specific epigenetic signature. This methodology provides the advantage of using a phenotypic output that can then be traced back to a common genetic lesion. The authors also leveraged the Cancer Cell Line Encyclopedia (CCLE), a resource of well-characterized cell lines with extensive genetic and pharmacological data5. Chromatin profiling was conducted on a panel of 115 CCLE cell lines; subsequent clustering identified 6 discrete chromatin patterns. Two of these signatures were linked to known gainof-function and loss-of-function mutations in EZH2, consistent with the known duality of EZH2 activity in cancer pathogenesis6,7. The authors then focused their attention on a novel cluster characterized by increased dimethylation of histone H3 at lysine 36 (H3K36me2), which is thought to mark active chromatin and antagonize the repressive H3K27me3 mark, as has been shown functionally in multiple myeloma8 (Fig. 1). Approximately half of the cell lines in this cluster had a t(4;14) translocation, which is known to induce overexpression of NSD2 (refs. 9,10). After mining CCLE data, the authors identified a new NSD2 mutation in the remaining cell lines in the cluster, encoding a Glu1099Lys alteration, and this mutation was validated in primary leukemia samples. The inclusion of global histone profiling data sets in publically accessible cell line and patient sample databases will permit the 1270
identification of relevant dysregulated epigenetic modifiers. Integrating mutational and histone modification data sets will make it possible for users to search for specific alterations in their disease of interest and to determine whether specific mutations have a global chromatin signature. Subsequent histone profiling studies in the CCLE collection and in annotated patient cohorts will likely lead to the identification of new somatic alterations that induce changes in global chromatin state, although not all histone changes are necessarily disease relevant. Further, there is partial redundancy in histone-modifying enzymes, for example, between EZH1 and EZH2 (ref. 11), such that global chromatin profiling may miss more subtle or locus-specific alterations driven by specific somatic mutations. In addition, disruption of the stoichiometry of epigenetic complexes through gain- and loss-of-function mutations may result in the formation of atypical epigenetic complexes that can result in differential epigenetic signatures and changes in gene expression. NSD2 mutations in B-ALL B cell acute lymphoblastic leukemia (B-ALL) is the most common pediatric cancer, with about 20% of children with B-ALL presenting with the ETV6-RUNX1 translocation12,13. Most pediatric B-ALL cases have a favorable response to induction chemotherapy and radiation. There is a risk of relapse, which tends to result in unfavorable responses to future therapies, as well as the risk of developing secondary
malignancies as a result of the toxicities of induction therapy; therefore, there is a need to develop more targeted therapies. Given the prevalence of the NSD2 mutation encoding Glu1099Lys in ALL cell lines in the CCLE profiling cohort, Stegmeier and colleagues sequenced NSD2 in an extensive pediatric cancer sample set and found prevalent NSD2 mutations in pediatric B-ALLs positive for ETV6-RUNX1 and TCF-PBX1 translocations. The NSD2 protein is a SET-domain containing histone methyltransferase that can monoor dimethylate H3K36 (ref. 10). The NSD2 Glu1099Lys alteration causes increased histone methyltransferase activity, which is indicative of hyperactivation of the NSD2 enzyme. The authors expressed NSD2 Glu1099Lys in a multiple myeloma cell line in which the t(4:14) NSD2 translocation was deleted, with expression of mutant NSD2 resulting in increased H3K36me2 levels. Knockdown of NSD2 protein levels in ALL cell lines encoding the NSD2 Glu1099Lys mutant resulted in decreased colony formation and tumor xenograft growth, suggesting a requirement for NSD2 function in NSD2-mutant ALL. These data suggest that NSD2 is an excellent candidate for future biological and therapeutic studies. Of note, NSD1 and NSD3 translocations, in which these enzymes are fused to NUP98, are observed in myelodysplastic syndromes and acute myeloid leukemia14,15. Developing selective inhibitors for this family of methyltransferases may be of therapeutic import in patients with mutant or translocated NSD1, NSD2 or NSD3. As such, the discoveries described here provide a framework for the identification and characterization of new mutations in epigenetic modifiers in human cancers, underscore the role of altered histone methyltransferase activity in B-ALL and provide a therapeutic rationale for targeting H3K36 methyltransferases in human cancers. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. 1. Knutson, S.K. et al. Nat. Chem. Biol. 8, 890–896 (2012). 2. McCabe, M.T. et al. Nature 492, 108–112 (2012). 3. Bernt, K.M. et al. Cancer Cell 20, 66–78 (2011). 4. Jaffe, J.D. et al. Nat. Genet. 45, 1386–1391 (2013). 5. Barretina, J. et al. Nature 483, 603–607 (2012). 6. Yap, D.B. et al. Blood 117, 2451–2459 (2011). 7. Ernst, T. et al. Nat. Genet. 42, 722–726 (2010). 8. Yuan, W. et al. J. Biol. Chem. 286, 7983–7989 (2011). 9. Martinez-Garcia, E. et al. Blood 117, 211–220 (2011). 10. Kuo, A.J. et al. Mol. Cell 44, 609–620 (2011). 11. Shen, X. et al. Mol. Cell 32, 491–502 (2008). 12. Loh, M.L. & Mullighan, C.G. Clin. Cancer Res. 18, 2754–2767 (2012). 13. Shurtleff, S.A. et al. Leukemia 9, 1985–1989 (1995). 14. Jaju, R.J. et al. Blood 98, 1264–1267 (2001). 15. Rosati, R. et al. Blood 99, 3857–3860 (2002).
volume 45 | number 11 | NOVEMBER 2013 | nature genetics
© 2013 Nature America, Inc. All rights reserved.
news and views of cancer types8–10. Consequently, the complex has been identified as the most frequently mutated chromatin regulatory complex in cancer and one of the most frequently mutated targets overall. The gene encoding SNF5 (SMARCB1) is mutated in pediatric rhabdoid tumors and in a few other rare cancers, but it is not widely mutated across cancer types8. However, mechanisms that indirectly disrupt SNF5 are now being identified in additional types of cancer. For instance, a recent study has shown that the SS18-SSX fusion in synovial sarcoma promotes tumorigenesis by disrupting SNF5 integration into the SWI/SNF complex11. The work of Prensner et al.7 now illustrates how lncRNA-mediated disruption of this SWI/SNF complex subunit can also drive malignancy. The new findings of Prensner et al.7 suggest that SChLAP1 may specifically affect the targeting mechanism of the SWI/SNF complex, a function that remains poorly understood. Although decreased SNF5 binding was observed at some promoters, leading to decreased expression of downstream genes, the specific effect of reduced SNF5 binding
on epigenetic regulation needs to be explored further. The SWI/SNF complex is canonically described as a regulator of gene expression via remodeling of nucleosomes, especially at promoters12,13. Thus, it is possible that SChLAP1 expression leads to disordered nucleosomes or other changes in chromatin structure at proximal promoter sites. Taking these results together with those for EZH2 and BAF57, another SWI/SNF subunit, a theme may be emerging in which aberrantly expressed epigenetic regulators contribute to aggressive prostate cancers14,15. SChLAP1 can be added to a growing list of lncRNAs associated with prostate cancer6. In addition, the ability of SChLAP1 to inhibit SNF5 binding further emphasizes the central role that chromatin regulators have in driving cancer. In essence, SChLAP1 expression in prostate cancer ties together two frontiers in cancer biology: noncoding RNAs and chromatin regulators. By uncovering the interaction between SChLAP1 and SNF5, this new study demonstrates both a mechanism through which lncRNAs can act to drive malignancy
and an additional mechanism by which SWI/ SNF activity can be disrupted in a way that affects cancer growth. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. 1. Siegel, R., Naishadham, D. & Jemal, A. CA Cancer J. Clin. 63, 11–30 (2013). 2. Varambally, S. et al. Nature 419, 624–629 (2002). 3. Prensner, J.R. & Chinnaiyan, A.M. Cancer Discov. 1, 391–407 (2011). 4. Du, Z. et al. Nat. Struct. Mol. Biol. 20, 908–913 (2013). 5. Prensner, J.R. et al. Nat. Biotechnol. 29, 742–749 (2011). 6. Yang, L. et al. Nature 500, 598–602 (2013). 7. Prensner, J.R. et al. Nat. Genet. 45, 1392–1398 (2013). 8. Wilson, B.G. & Roberts, C.W.M. Nat. Rev. Cancer 11, 481–492 (2011). 9. Lee, R.S. et al. J. Clin. Invest. 122, 2983–2988 (2012). 10. Kadoch, C. et al. Nat. Genet. 45, 592–601 (2013). 11. Kadoch, C. & Crabtree, G.R. Cell 153, 71–85 (2013). 12. You, J.S. et al. PLoS Genet. 9, e1003459 (2013). 13. Tolstorukov, M.Y. et al. Proc. Natl. Acad. Sci. USA 110, 10165–10170 (2013). 14. Link, K.A. et al. Cancer Res. 68, 4551–4558 (2008). 15. Balasubramaniam, S. et al. Clin. Cancer Res. 19, 2657–2667 (2013).
Mining the epigenetic landscape in ALL Lindsay M LaFave & Ross L Levine The significance of epigenomic aberrations in cancer development has been underscored by the discovery of mutations in key chromatin modifiers, most notably in hematological malignancies. A new study of pediatric acute lymphoblastic leukemia (ALL) demonstrates the usefulness of mapping global epigenetic signatures and applying these data in a framework to identify and characterize underlying somatic genetic alterations in human cancers. Epigenetic dysregulation has become a cardinal feature of human malignancies, warranting the development of innovative approaches to identify epigenetic driver alterations. Numerous large-scale sequencing studies, including targeted exome, transcriptome and whole genome, have led to the identification of mutations or expression changes in genes critical for normal Lindsay M. LaFave and Ross L. Levine are affiliated with the Human Oncology and Pathogenesis Program, Memorial SloanKettering Cancer Center, New York, New York, USA. Lindsay M. LaFave is also affiliated with Gerstner Sloan-Kettering Graduate School, Memorial Sloan-Kettering Cancer Center, New York, New York, USA. Ross L. Levine is also affiliated with the Leukemia Service, Department of Medicine, Memorial SloanKettering Cancer Center, New York, New York, USA. e-mail: [email protected]
epigenetic regulation in a spectrum of cancer types, including in hematological malignancies. For example, oncogenic mutations in histone methyltransferases that modify key lysine residues on histone tails have been identified in leukemias and lymphomas. These findings have led to the development of selective small molecule histone methyltransferase inhibitors for therapeutic applications. Selective inhibitors targeting gain-of-function alterations in EZH2 (refs. 1,2), the catalytic methyltransferase in the PRC2 complex that mediates repressive trimethylation of histone H3 at lysine 27 (H3K27me3), and DOT1L3, the methyltransferase that methylates histone H3 at lysine 79 (H3K79), are being tested in EZH2-mutant B cell lymphoma and MLL-rearranged leukemia, respectively. The successful development of EZH2 and DOT1L inhibitors has driven efforts to assess mutations in histone methyltransferases that drive cancer development and progression and represent tractable therapeutic targets. In this
nature genetics | volume 45 | number 11 | NOVEMBER 2013
issue of Nature Genetics, Frank Stegmeier, Levi Garraway and colleagues describe an innovative approach coupling histone mass spectroscopy analyses with mutational data for the identification of genetic alterations in the NSD2 gene, which encodes a H3K36 methyltransferase4. Global chromatin profiling Stegmeier and colleagues used a tandem mass spectroscopy strategy to measure chromatin marks on extracted histone proteins, a technique they term ‘global chromatin profiling’. Mass spectrometry circumvents the use of antibodies for the analysis of chromatin marks, which is advantageous given that the lack of specific antibodies for many histone marks remains a significant challenge for the field. Whereas a typical study design for the identification of driver mutations involves elucidating genetic alterations by sequencing and then studying the mechanisms by which these alterations contribute to disease pathogenesis, Stegmeier 1269
news and views
H2B
H2A
H4
H3
ALQKRPAK AKT GG RA K TS S HP A KR ATQKT R A P K K VG GTA 36
npg
© 2013 Nature America, Inc. All rights reserved.
Katie Vicari
t(4;14)
27
NSD2 E1099K
H3K27me3 H3K36me2 Unmethylated H3K36
Figure 1 Schematic of the chromatin signature that was studied by Stegmeier and colleagues. Modifications of interest in the signature, H3K36me2 and H3K27me3, are highlighted on the histone H3 tail sequence. Methyl groups are coded on the basis of whether their levels are increased (H3K36me2; reds) or decreased (H3K27me3 and unmethylated H3K36; blues). This histone pattern was unique to cell lines with NSD2 alterations, either the t(4;14) translocation or the NSD2 mutation encoding Glu1099Lys. The t(4;14) translocation is common to multiple myeloma, whereas the NSD2 mutation was predominantly found in ALL cell lines and patient samples. In the representation of clustering output, genomic alterations are identified on the top, with altered chromatin modifications shown on the left (blue indicates depletion and red indicates enrichment in the signature).
and colleagues used an alternate approach. They identified cell lines with similar epigenetic patterns and then searched for somatic alterations that would be predicted to cause the specific epigenetic signature. This methodology provides the advantage of using a phenotypic output that can then be traced back to a common genetic lesion. The authors also leveraged the Cancer Cell Line Encyclopedia (CCLE), a resource of well-characterized cell lines with extensive genetic and pharmacological data5. Chromatin profiling was conducted on a panel of 115 CCLE cell lines; subsequent clustering identified 6 discrete chromatin patterns. Two of these signatures were linked to known gainof-function and loss-of-function mutations in EZH2, consistent with the known duality of EZH2 activity in cancer pathogenesis6,7. The authors then focused their attention on a novel cluster characterized by increased dimethylation of histone H3 at lysine 36 (H3K36me2), which is thought to mark active chromatin and antagonize the repressive H3K27me3 mark, as has been shown functionally in multiple myeloma8 (Fig. 1). Approximately half of the cell lines in this cluster had a t(4;14) translocation, which is known to induce overexpression of NSD2 (refs. 9,10). After mining CCLE data, the authors identified a new NSD2 mutation in the remaining cell lines in the cluster, encoding a Glu1099Lys alteration, and this mutation was validated in primary leukemia samples. The inclusion of global histone profiling data sets in publically accessible cell line and patient sample databases will permit the 1270
identification of relevant dysregulated epigenetic modifiers. Integrating mutational and histone modification data sets will make it possible for users to search for specific alterations in their disease of interest and to determine whether specific mutations have a global chromatin signature. Subsequent histone profiling studies in the CCLE collection and in annotated patient cohorts will likely lead to the identification of new somatic alterations that induce changes in global chromatin state, although not all histone changes are necessarily disease relevant. Further, there is partial redundancy in histone-modifying enzymes, for example, between EZH1 and EZH2 (ref. 11), such that global chromatin profiling may miss more subtle or locus-specific alterations driven by specific somatic mutations. In addition, disruption of the stoichiometry of epigenetic complexes through gain- and loss-of-function mutations may result in the formation of atypical epigenetic complexes that can result in differential epigenetic signatures and changes in gene expression. NSD2 mutations in B-ALL B cell acute lymphoblastic leukemia (B-ALL) is the most common pediatric cancer, with about 20% of children with B-ALL presenting with the ETV6-RUNX1 translocation12,13. Most pediatric B-ALL cases have a favorable response to induction chemotherapy and radiation. There is a risk of relapse, which tends to result in unfavorable responses to future therapies, as well as the risk of developing secondary
malignancies as a result of the toxicities of induction therapy; therefore, there is a need to develop more targeted therapies. Given the prevalence of the NSD2 mutation encoding Glu1099Lys in ALL cell lines in the CCLE profiling cohort, Stegmeier and colleagues sequenced NSD2 in an extensive pediatric cancer sample set and found prevalent NSD2 mutations in pediatric B-ALLs positive for ETV6-RUNX1 and TCF-PBX1 translocations. The NSD2 protein is a SET-domain containing histone methyltransferase that can monoor dimethylate H3K36 (ref. 10). The NSD2 Glu1099Lys alteration causes increased histone methyltransferase activity, which is indicative of hyperactivation of the NSD2 enzyme. The authors expressed NSD2 Glu1099Lys in a multiple myeloma cell line in which the t(4:14) NSD2 translocation was deleted, with expression of mutant NSD2 resulting in increased H3K36me2 levels. Knockdown of NSD2 protein levels in ALL cell lines encoding the NSD2 Glu1099Lys mutant resulted in decreased colony formation and tumor xenograft growth, suggesting a requirement for NSD2 function in NSD2-mutant ALL. These data suggest that NSD2 is an excellent candidate for future biological and therapeutic studies. Of note, NSD1 and NSD3 translocations, in which these enzymes are fused to NUP98, are observed in myelodysplastic syndromes and acute myeloid leukemia14,15. Developing selective inhibitors for this family of methyltransferases may be of therapeutic import in patients with mutant or translocated NSD1, NSD2 or NSD3. As such, the discoveries described here provide a framework for the identification and characterization of new mutations in epigenetic modifiers in human cancers, underscore the role of altered histone methyltransferase activity in B-ALL and provide a therapeutic rationale for targeting H3K36 methyltransferases in human cancers. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. 1. Knutson, S.K. et al. Nat. Chem. Biol. 8, 890–896 (2012). 2. McCabe, M.T. et al. Nature 492, 108–112 (2012). 3. Bernt, K.M. et al. Cancer Cell 20, 66–78 (2011). 4. Jaffe, J.D. et al. Nat. Genet. 45, 1386–1391 (2013). 5. Barretina, J. et al. Nature 483, 603–607 (2012). 6. Yap, D.B. et al. Blood 117, 2451–2459 (2011). 7. Ernst, T. et al. Nat. Genet. 42, 722–726 (2010). 8. Yuan, W. et al. J. Biol. Chem. 286, 7983–7989 (2011). 9. Martinez-Garcia, E. et al. Blood 117, 211–220 (2011). 10. Kuo, A.J. et al. Mol. Cell 44, 609–620 (2011). 11. Shen, X. et al. Mol. Cell 32, 491–502 (2008). 12. Loh, M.L. & Mullighan, C.G. Clin. Cancer Res. 18, 2754–2767 (2012). 13. Shurtleff, S.A. et al. Leukemia 9, 1985–1989 (1995). 14. Jaju, R.J. et al. Blood 98, 1264–1267 (2001). 15. Rosati, R. et al. Blood 99, 3857–3860 (2002).
volume 45 | number 11 | NOVEMBER 2013 | nature genetics
