news and views
Cancer gene discovery goes mobile Louise van der Weyden, Marco Ranzani & David J Adams
One clear truth of cancer genomes is that chaos reigns supreme. Although sequencing is revealing the landscapes of somatic mutation, in some genomic regions, such as those that are highly rearranged, the picture remains smudged. Much of what we have learned about cancer-related genes has come from functional studies, and our interpretation of the sequence of cancer genomes rests largely on this wealth of knowledge. In this issue of Nature Genetics, Molyneux et al.1 describe a new approach for cancer gene discovery—human somatic cell mutagenesis using lentiviral vectors that introduce a Sleeping Beauty (SB) transposon2 into the genome. Termed Lentihop, this bipartite mutagenesis system comes in two flavors, with one vector carrying a strong promoter to upregulate gene expression and a second vector carrying a bidirectional loss-of-function cassette. Both elements are flanked by SB repeats within the lentiviral vector backbone. Lentihop viral integrations were shown to be mutagenic, as were the gain-of-function and loss-of-function alleles generated after SB mobilization. Using mesenchymal cells expressing the SB transposase, which were derived from human bone explants and engineered to be poised on the brink of transformation, the authors screened for new drivers of sarcoma formation. Analysis of the primary lentiviral and secondary SB integration events gave intriguing results.
such as CHFR, DOCK4 and CADM1 that are well-established tumor suppressors. This group was followed by a list of ‘smoking gun’ genes, such as UBR4 and HDLBP—genes with a link to cancer-associated processes that are mutated in databases such as the Catalogue of Somatic Mutations in Cancer (COSMIC)3 but that had not previously been tied to sarcoma. The next group constituted a long list of some 80 genes that were mutated at a statistically significant frequency by Lentihop vectors but that had not as yet directly been implicated as drivers by the genomic analysis of human tumors.
Can all these new genes possibly be real cancer genes? Recent advances made in cancer genome sequencing and the analysis of the data sets produced have established the view that a true cancer-related gene is one that is mutated at a statistically significant frequency, often irrespective of a clear mechanism or the ultimate proof of functional validation. The sensitivity and specificity inherent in defining driver genes on the basis of statistical analysis alone are not known, and most sequencing studies use modest numbers of samples and thus have limited power. Somatic variant calling is also difficult, and assigning pathogenicity
HDLBP
UBR4
SPTAN1
BRAF
PTEN
RAS
TP53
Cancer gene road trip The first group of genes described by Molyneux et al. included the ‘known knowns’: genes Louise van der Weyden, Marco Ranzani and David J. Adams are at Experimental Cancer Genetics, Wellcome Trust Sanger Institute, Hinxton, UK. e-mail: [email protected]
928
Figure 1 The cancer gene road trip. Genetic screens in model systems suggest that there are many more cancer genes than have been identified by sequencing. Integrating cancer genome data and data from functional studies offers a new horizon of discovery.
volume 46 | number 9 | SEPTEMBER 2014 | nature genetics
Debbie Maizels/Nature Publishing Group
npg
© 2014 Nature America, Inc. All rights reserved.
A new study describes a tool, Lentihop, for somatic insertional mutagenesis in human cells and uses this system in combination with cancer genome data to define new genes and pathways involved in sarcoma development. Gene discovery in this way suggests that we are far from a complete catalog of cancer drivers.
news and views
npg
© 2014 Nature America, Inc. All rights reserved.
to nucleotide changes is still more of an art than a science. Further, we talk about amplified and deleted regions of the genome4 but fail to appreciate what drivers lie within these regions and how they might contribute to tumorigenesis. The promoters of many genes are epigenetically altered in cancer5 and regulatory mutations contribute to disease development6, but it is still a difficult task to define the complete catalog of cancer drivers. Despite all the sequence data that are now available, we are still not ‘there yet’, and defining the complete catalog of cancer drivers will inevitably be a long journey. So how do we make it home? One important way will be by integrating genome-wide functional studies with the sequencing of cancer genomes. Inducing tumorigenesis with a hop, skip and a jump Genome-wide forward genetic screens using insertional mutagens are well known to be powerful tools for cancer gene discovery in model systems. The mammalian WNT pathway7,8, for example, was discovered using the mouse mammary tumor virus, and other important cancerrelated genes such as BMI1, PIM1 and USP9X were defined in similar ways9,10. Here, with the development of the Lentihop system, Molyneux et al. extend the approach to human cells. This has several advantages. First, although many important pathways in cancer are highly conserved by humans and model organisms such as the mouse, there are some notable differences. For example, it has been suggested that
mouse cells require fewer driver mutations to become transformed than human cells11. This difference potentially means that there are subtleties in how cancer pathways are wired, making screens in human cells advantageous. Second, projects such as the Encyclopedia of DNA Elements (ENCODE)12 are defining in great detail the regulatory landscape of the human genome, which will facilitate the interpretation of the insertional events generated by the Lentihop system in human cells. So little is known about the regulatory landscape of human cancers, and data integration in this way will surely be a powerful tool for discovery. Third, the germline genome contributes significantly at all stages of tumor development13. The Lentihop system will allow the same screen to be performed in cells and cultures from different individuals to explore how germline alleles influence the somatic mutation profile. It may also be possible to explore how different drivers or driver variants influence Lentihop insertion profiles, thereby allowing synergistic and, possibly, mutually exclusive interactions to be defined. In the studies described here, cells were transplanted into nude mice, but, by engineering the host as has been described previously14, it should be possible to ask questions about how the tumor microenvironment influences tumor growth and the profile of mutated genes. It will also be possible to identify drivers of cancer drug resistance and, possibly, to use the Lentihop system to explore mechanisms of drug resensitization. Finally, as described by Molyneux et al., because Lentihop vectors
integrate proximally to cancer genes, it will be possible to define drivers in highly rearranged regions of the genome. In this way, HDLBP was defined as a new driver gene and found to be deleted in around 10% of tumors. Although the current study describes the use of Lentihop in the mesenchyme, the system should be widely applicable across many cell and cancer types. Thus, comprehensively profiling the cancer genome using systems such as Lentihop will allow us to functionalize the cancer genome. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests.
1. Molyneux, S.D. et al. Nat. Genet. 46, 964–972 (2014). 2. Ivics, Z., Hackett, P.B., Plasterk, R.H. & Izsvák, Z. Cell 91, 501–510 (1997). 3. Forbes, S.A. et al. Nucleic Acids Res. 39, D945–D950 (2011). 4. Bignell, G.R. et al. Nature 463, 893–898 (2010). 5. Plass, C. et al. Nat. Rev. Genet. 14, 765–780 (2013). 6. Horn, S. et al. Science 339, 959–961 (2013). 7. Nusse, R., van Ooyen, A., Cox, D., Fung, Y.K. & Varmus, H. Nature 307, 131–136 (1984). 8. van Ooyen, A. & Nusse, R. Cell 39, 233–240 (1984). 9. van Lohuizen, M. et al. Cell 65, 737–752 (1991). 10. Pérez-Mancera, P.A. et al. Nature 486, 266–270 (2012). 11. Rangarajan, A. & Weinberg, R.A. Nat. Rev. Cancer 3, 952–959 (2003). 12. ENCODE Project Consortium. Nature 489, 57–74 (2012). 13. Rahman, N. Nature 505, 302–308 (2014). 14. Thibaudeau, L. et al. Dis. Model. Mech. 7, 299–309 (2014).
RNA switch at enhancers Jeffrey J Quinn & Howard Y Chang Polycomb/Trithorax response elements (PRE/TREs) are genetic elements that can stably silence or activate genes. A new study describes how long noncoding RNAs (lncRNAs) transcribed from opposite strands of the Drosophila melanogaster vestigial PRE/TRE throw the switch between these two opposing epigenetic states. Polycomb (PcG) and Trithorax Group (TrxG) proteins are important regulators of the chromatin landscape and are essential for development in flies and mammals. PcG and TrxG proteins associate with regulatory elements throughout the genome called PRE/TREs where they act in Jeffrey J. Quinn is in the Department of Bioengineering at Stanford University, Stanford, California, USA, and Howard Y. Chang is in the Program in Epithelial Biology and the Howard Hughes Medical Institute at Stanford University, Stanford, California, USA. e-mail: [email protected]
opposition to one another: PcG proteins enforce gene silencing, whereas TrxG proteins promote activation1,2. PRE/TREs can switch between repressive and activating epigenetic states. Many PRE/TREs are sites of extensive lncRNA transcription, which is often correlated with the expression of the gene regulated by the PRE/TRE3,4. In a new study reported on page 973 of this issue, Leonie Ringrose and colleagues examined the PRE/TRE downstream of the vestigial (vg) gene in Drosophila and observed that this element bidirectionally expresses lncRNAs5. Curiously, the forward-strand transcript was correlated with vg repression,
nature genetics | volume 46 | number 9 | SEPTEMBER 2014
whereas the reverse-strand transcript was associated with activation of vg expression. Using cytological, transgenic and biochemical techniques, the authors show that the reversible silencing-to-activating switch is caused by a forward-to-reverse switch in the directionality of transcription at the PRE/TRE and demonstrate that lncRNAs are the functional components of this regulatory element. Opposite RNAs, opposite functions The vg gene is regulated by PcG at a downstream PRE/TRE6. Initially, Herzog et al.5 detected that this element expressed tissue- and developmental 929
Cancer gene discovery goes mobile Louise van der Weyden, Marco Ranzani & David J Adams
One clear truth of cancer genomes is that chaos reigns supreme. Although sequencing is revealing the landscapes of somatic mutation, in some genomic regions, such as those that are highly rearranged, the picture remains smudged. Much of what we have learned about cancer-related genes has come from functional studies, and our interpretation of the sequence of cancer genomes rests largely on this wealth of knowledge. In this issue of Nature Genetics, Molyneux et al.1 describe a new approach for cancer gene discovery—human somatic cell mutagenesis using lentiviral vectors that introduce a Sleeping Beauty (SB) transposon2 into the genome. Termed Lentihop, this bipartite mutagenesis system comes in two flavors, with one vector carrying a strong promoter to upregulate gene expression and a second vector carrying a bidirectional loss-of-function cassette. Both elements are flanked by SB repeats within the lentiviral vector backbone. Lentihop viral integrations were shown to be mutagenic, as were the gain-of-function and loss-of-function alleles generated after SB mobilization. Using mesenchymal cells expressing the SB transposase, which were derived from human bone explants and engineered to be poised on the brink of transformation, the authors screened for new drivers of sarcoma formation. Analysis of the primary lentiviral and secondary SB integration events gave intriguing results.
such as CHFR, DOCK4 and CADM1 that are well-established tumor suppressors. This group was followed by a list of ‘smoking gun’ genes, such as UBR4 and HDLBP—genes with a link to cancer-associated processes that are mutated in databases such as the Catalogue of Somatic Mutations in Cancer (COSMIC)3 but that had not previously been tied to sarcoma. The next group constituted a long list of some 80 genes that were mutated at a statistically significant frequency by Lentihop vectors but that had not as yet directly been implicated as drivers by the genomic analysis of human tumors.
Can all these new genes possibly be real cancer genes? Recent advances made in cancer genome sequencing and the analysis of the data sets produced have established the view that a true cancer-related gene is one that is mutated at a statistically significant frequency, often irrespective of a clear mechanism or the ultimate proof of functional validation. The sensitivity and specificity inherent in defining driver genes on the basis of statistical analysis alone are not known, and most sequencing studies use modest numbers of samples and thus have limited power. Somatic variant calling is also difficult, and assigning pathogenicity
HDLBP
UBR4
SPTAN1
BRAF
PTEN
RAS
TP53
Cancer gene road trip The first group of genes described by Molyneux et al. included the ‘known knowns’: genes Louise van der Weyden, Marco Ranzani and David J. Adams are at Experimental Cancer Genetics, Wellcome Trust Sanger Institute, Hinxton, UK. e-mail: [email protected]
928
Figure 1 The cancer gene road trip. Genetic screens in model systems suggest that there are many more cancer genes than have been identified by sequencing. Integrating cancer genome data and data from functional studies offers a new horizon of discovery.
volume 46 | number 9 | SEPTEMBER 2014 | nature genetics
Debbie Maizels/Nature Publishing Group
npg
© 2014 Nature America, Inc. All rights reserved.
A new study describes a tool, Lentihop, for somatic insertional mutagenesis in human cells and uses this system in combination with cancer genome data to define new genes and pathways involved in sarcoma development. Gene discovery in this way suggests that we are far from a complete catalog of cancer drivers.
news and views
npg
© 2014 Nature America, Inc. All rights reserved.
to nucleotide changes is still more of an art than a science. Further, we talk about amplified and deleted regions of the genome4 but fail to appreciate what drivers lie within these regions and how they might contribute to tumorigenesis. The promoters of many genes are epigenetically altered in cancer5 and regulatory mutations contribute to disease development6, but it is still a difficult task to define the complete catalog of cancer drivers. Despite all the sequence data that are now available, we are still not ‘there yet’, and defining the complete catalog of cancer drivers will inevitably be a long journey. So how do we make it home? One important way will be by integrating genome-wide functional studies with the sequencing of cancer genomes. Inducing tumorigenesis with a hop, skip and a jump Genome-wide forward genetic screens using insertional mutagens are well known to be powerful tools for cancer gene discovery in model systems. The mammalian WNT pathway7,8, for example, was discovered using the mouse mammary tumor virus, and other important cancerrelated genes such as BMI1, PIM1 and USP9X were defined in similar ways9,10. Here, with the development of the Lentihop system, Molyneux et al. extend the approach to human cells. This has several advantages. First, although many important pathways in cancer are highly conserved by humans and model organisms such as the mouse, there are some notable differences. For example, it has been suggested that
mouse cells require fewer driver mutations to become transformed than human cells11. This difference potentially means that there are subtleties in how cancer pathways are wired, making screens in human cells advantageous. Second, projects such as the Encyclopedia of DNA Elements (ENCODE)12 are defining in great detail the regulatory landscape of the human genome, which will facilitate the interpretation of the insertional events generated by the Lentihop system in human cells. So little is known about the regulatory landscape of human cancers, and data integration in this way will surely be a powerful tool for discovery. Third, the germline genome contributes significantly at all stages of tumor development13. The Lentihop system will allow the same screen to be performed in cells and cultures from different individuals to explore how germline alleles influence the somatic mutation profile. It may also be possible to explore how different drivers or driver variants influence Lentihop insertion profiles, thereby allowing synergistic and, possibly, mutually exclusive interactions to be defined. In the studies described here, cells were transplanted into nude mice, but, by engineering the host as has been described previously14, it should be possible to ask questions about how the tumor microenvironment influences tumor growth and the profile of mutated genes. It will also be possible to identify drivers of cancer drug resistance and, possibly, to use the Lentihop system to explore mechanisms of drug resensitization. Finally, as described by Molyneux et al., because Lentihop vectors
integrate proximally to cancer genes, it will be possible to define drivers in highly rearranged regions of the genome. In this way, HDLBP was defined as a new driver gene and found to be deleted in around 10% of tumors. Although the current study describes the use of Lentihop in the mesenchyme, the system should be widely applicable across many cell and cancer types. Thus, comprehensively profiling the cancer genome using systems such as Lentihop will allow us to functionalize the cancer genome. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests.
1. Molyneux, S.D. et al. Nat. Genet. 46, 964–972 (2014). 2. Ivics, Z., Hackett, P.B., Plasterk, R.H. & Izsvák, Z. Cell 91, 501–510 (1997). 3. Forbes, S.A. et al. Nucleic Acids Res. 39, D945–D950 (2011). 4. Bignell, G.R. et al. Nature 463, 893–898 (2010). 5. Plass, C. et al. Nat. Rev. Genet. 14, 765–780 (2013). 6. Horn, S. et al. Science 339, 959–961 (2013). 7. Nusse, R., van Ooyen, A., Cox, D., Fung, Y.K. & Varmus, H. Nature 307, 131–136 (1984). 8. van Ooyen, A. & Nusse, R. Cell 39, 233–240 (1984). 9. van Lohuizen, M. et al. Cell 65, 737–752 (1991). 10. Pérez-Mancera, P.A. et al. Nature 486, 266–270 (2012). 11. Rangarajan, A. & Weinberg, R.A. Nat. Rev. Cancer 3, 952–959 (2003). 12. ENCODE Project Consortium. Nature 489, 57–74 (2012). 13. Rahman, N. Nature 505, 302–308 (2014). 14. Thibaudeau, L. et al. Dis. Model. Mech. 7, 299–309 (2014).
RNA switch at enhancers Jeffrey J Quinn & Howard Y Chang Polycomb/Trithorax response elements (PRE/TREs) are genetic elements that can stably silence or activate genes. A new study describes how long noncoding RNAs (lncRNAs) transcribed from opposite strands of the Drosophila melanogaster vestigial PRE/TRE throw the switch between these two opposing epigenetic states. Polycomb (PcG) and Trithorax Group (TrxG) proteins are important regulators of the chromatin landscape and are essential for development in flies and mammals. PcG and TrxG proteins associate with regulatory elements throughout the genome called PRE/TREs where they act in Jeffrey J. Quinn is in the Department of Bioengineering at Stanford University, Stanford, California, USA, and Howard Y. Chang is in the Program in Epithelial Biology and the Howard Hughes Medical Institute at Stanford University, Stanford, California, USA. e-mail: [email protected]
opposition to one another: PcG proteins enforce gene silencing, whereas TrxG proteins promote activation1,2. PRE/TREs can switch between repressive and activating epigenetic states. Many PRE/TREs are sites of extensive lncRNA transcription, which is often correlated with the expression of the gene regulated by the PRE/TRE3,4. In a new study reported on page 973 of this issue, Leonie Ringrose and colleagues examined the PRE/TRE downstream of the vestigial (vg) gene in Drosophila and observed that this element bidirectionally expresses lncRNAs5. Curiously, the forward-strand transcript was correlated with vg repression,
nature genetics | volume 46 | number 9 | SEPTEMBER 2014
whereas the reverse-strand transcript was associated with activation of vg expression. Using cytological, transgenic and biochemical techniques, the authors show that the reversible silencing-to-activating switch is caused by a forward-to-reverse switch in the directionality of transcription at the PRE/TRE and demonstrate that lncRNAs are the functional components of this regulatory element. Opposite RNAs, opposite functions The vg gene is regulated by PcG at a downstream PRE/TRE6. Initially, Herzog et al.5 detected that this element expressed tissue- and developmental 929
