Leading Edge
Previews From Prescription to Transcription: Genome Sequence as Drug Target Cong Guo,1,2,4 Anthony M. D’Ippolito,1,2,4 and Timothy E. Reddy1,3,* 1Center
for Genomic and Computational Biology Program in Genetics and Genomics 3Department of Biostatistics and Bioinformatics Duke University Medical School, Durham, NC 27708, USA 4Co-first author *Correspondence: [email protected] http://dx.doi.org/10.1016/j.cell.2015.06.033 2University
Personalizing treatments to account for genetically mediated differences in drug responses is an exciting opportunity to improve patient outcomes. In this issue, Soccio et al. reveal new mechanisms by which non-coding variants alter the activity of the anti-diabetic drug rosiglitazone. PPARg, a member of the class of ligandresponsive transcription factors (TFs) known as nuclear receptors, is a major driver of lipid uptake and adipogenesis. Small-molecule pharmaceuticals targeting PPARg have proven to be effective anti-diabetic agents. Although some individuals respond well to those treatments without adverse effects, others experience reduced efficacy and substantial side effects. That heterogeneity in treatment response poses a major limitation for treating type 2 diabetes and is representative of the broader challenges facing physicians when prescribing drugs that target other nuclear receptors. Drug responses are highly heritable, and understanding the genetic mechanisms contributing to that heritability will improve the application of current drugs while informing the design of new ones (Roden and George, 2002). Mutations that alter protein structure—coding mutations—are a known cause of that heritability (Hurley et al., 1991) (Figure 1, top). Meanwhile, non-coding mutations that alter steady-state levels of gene expression are emerging as a major contributor to the heterogeneity of human phenotypes, including drug responsiveness (e.g., Gusev et al., 2014). For example, genetically altered expression of drug efflux pumps has been shown to contribute to chemotherapeutic resistance in some individuals (Patch et al., 2015) (Figure 1, middle). Non-coding mutations are typically thought to act by changing the DNA sequence of TF binding sites in a way 16 Cell 162, July 2, 2015 ª2015 Elsevier Inc.
that alters the affinity of the TF to those sites (McDaniell et al., 2010). In this issue, Soccio et al. demonstrate that different alleles of PPARg sites lead to different transcriptional responses to the anti-diabetic drug rosiglitazone (Soccio et al., 2015) (Figure 1, bottom). In doing so, they recast the PPARg binding sites as drug targets and reveal new ways in which non-coding variants alter disease risk and drug responses. Their findings remarkably expand the realm of personalized medicine, particularly for drugs targeting nuclear receptors. Mutations in PPARg binding sites are only part of the story. A variety of additional TFs interact with PPARg and contribute to its genomic occupancy. The co-binding TFs each have their own DNA binding preferences, complicating our understanding of how individual noncoding mutations contribute to PPARgmediated effects of rosiglitazone treatment. Soccio et al. show that non-coding genetic variation not only alters PPARg occupancy, but also has coordinated effects on the binding of the associated TFs C/EBP and the glucocorticoid receptor. Furthermore, disruption of the DNA binding sequences of any of those TFs leads to decreased PPARg occupancy. The observed coordinated effects support a model in which TFs bind as a complex on a given allele (ENCODE Proect Consortium, 2012) and indicate that drug response heterogeneity may also result from mutations in binding sites for yet-unknown cooperating TFs. In that manner, it may be more valuable to consider regula-
tory complexes rather than individual binding sites as the key functional unit for understanding the genetic contributions to PPARg and rosiglitazone responses. Identifying additional members of PPARg regulatory complexes will also likely be valuable for predicting genetic contributions to both type 2 diabetes and its treatment. While Soccio et al. find a strong enrichment of allele-specific binding sites near rosiglitazone-induced genes, they find no evidence of a similar enrichment near rosiglitazone-repressed genes. The relative challenge of identifying the binding sites responsible for gene repression is a recurring theme in nuclear receptor biology (e.g., Reddy et al., 2009). One possible explanation is that repressive sites are inherently difficult to assay due to chromatin structure, epitope availability, or indirect interaction with the genome. Alternatively, repression by nuclear receptors may be secondary to activation of direct target genes (Abraham et al., 2006). Because nearly as many genes are repressed as are activated by nuclear receptors, determining the predominant mechanisms of that repression will be valuable for ongoing efforts to increase the specificity of nuclear receptor drugs. For example, secondary factors may themselves be targets for drugs that are beneficial either alone or in combination with nuclear-receptor-targeting drugs. To realize that potential, a revised understanding of the ways in which nuclear receptors repress gene expression is needed. In particular, determining the
REFERENCES relative contribution of each mechanism of repression to Abraham, S.M., Lawrence, T., Kleithe total repressive effects man, A., Warden, P., Medghalchi, will prioritize the most prevaM., Tuckermann, J., Saklatvala, J., lent mechanisms for suband Clark, A.R. (2006). J. Exp. sequent pharmaceutical tarMed. 203, 1883–1889. geting. Crowley, J.J., Zhabotynsky, V., Sun, Identifying non-coding variW., Huang, S., Pakatci, I.K., Kim, Y., Wang, J.R., Morgan, A.P., Calaway, ants that alter rosiglitazone’s J.D., Aylor, D.L., et al. (2015). Nat. effects in human cells creates Genet. 47, 353–360. a new challenge: how do we ENCODE Proect Consortium (2012). determine the resulting organNature 489, 57–74. ismal effects? The genetic diGusev, A., Lee, S.H., Trynka, G., versity of recombinant inbred Finucane, H., Vilhja´lmsson, B.J., mouse strains makes them a Xu, H., Zang, C., Ripke, S., Bulikpowerful resource for linking Sullivan, B., Stahl, E., et al.; mouse loci with phenotypes. Schizophrenia Working Group of Protein sequence and functhe Psychiatric Genomics ConFigure 1. Genetic Variation Affects Drug Responses tion are often conserved besortium; SWE-SCZ Consortium; (Top) Genetic variants in protein-coding exons can alter protein structure and tween mouse and human, Schizophrenia Working Group of subsequent drug responses. In the example of familial glucocorticoid resisthe Psychiatric Genomics Conand therefore it is reasonable tance, a mutation in the ligand binding domain of the glucocorticoid receptor sortium; SWE-SCZ Consortium reduces binding of glucocorticoid drugs and subsequent regulation of target to hypothesize that mutations gene expression. (Middle) Regulatory element rearrangements can change the (2014). Am. J. Hum. Genet. 95, in orthologous genes will expression of genes that govern drug responses. For example, a promoter 535–552. have similar phenotypic effusion increases the expression of ABCB1, which encodes the drug efflux Hurley, D.M., Accili, D., Stratakis, pump MDR1. Increased MDR1 is associated with chemotherapy resistance. fects. There is now growing C.A., Karl, M., Vamvakopoulos, N., (Bottom) Genetic variation in regulatory elements can disrupt TF binding and evidence that the phenotypic drug responses by altering the affinity of a drug-targeted TF. Soccio et al. Rorer, E., Constantine, K., Taylor, effects of regulatory variation demonstrate that phenomenon by revealing that genetic variants within PPARg S.I., and Chrousos, G.P. (1991). binding sites alter PPARg and cofactor occupancy and alter the response to may also be conserved beJ. Clin. Invest. 87, 680–686. rosiglitazone. tween species (Crowley et al., McDaniell, R., Lee, B.K., Song, L., 2015). Here, for example, SocLiu, Z., Boyle, A.P., Erdos, M.R., cio and colleagues use natural Scott, L.J., Morken, M.A., Kucera, variation in metabolic phenotypes beTogether, the findings of Soccio et al. K.S., Battenhouse, A., et al. (2010). Science 328, tween disparate mouse strains to implicate reveal new mechanisms by which non- 235–239. non-coding variation in PPARg binding coding variation leads to differences in Patch, A.M., Christie, E.L., Etemadmoghadam, D., sites as contributing to diabetic pheno- drug responses. Because the genetic var- Garsed, D.W., George, J., Fereday, S., Nones, K., types and rosiglitazone responses. They iants alter the genomic targets of the drug Cowin, P., Alsop, K., Bailey, P.J., et al.; Australian Ovarian Cancer Study Group (2015). Nature 521, then leverage the conserved structure itself, their findings pave the way to ulti489–494. and function of PPARg between species mately predict individual transcriptional Reddy, T.E., Pauli, F., Sprouse, R.O., Neff, N.F., to identify genetic variants with similar mo- responses to a prescribed medication. Newberry, K.M., Garabedian, M.J., and Myers, lecular effects in human tissues. Their reli- When combined with evidence of which R.M. (2009). Genome Res. 19, 2163–2171. ance on conserved TF functions between target genes contribute to downstream Roden, D.M., and George, A.L., Jr. (2002). Nat. organisms is an appealing strategy to un- drug effects, the transcriptional re- Rev. Drug Discov. 1, 37–44. cover regulatory mechanisms contributing sponses may form the basis for new apSoccio, R.E., Chen, E.R., Rajapurkar, S.R., Safato type 2 diabetes and one that may gener- proaches to identify patients who are bakhsh, P., Marinis, J.M., Dispirito, J.R., Emmett, alize broadly to other complex human most or least likely to respond favorably M.J., Briggs, E.R., Fang, B., Everett, L.J., et al. (2015). Cell 162, this issue, 33–44. traits. to treatment.
Cell 162, July 2, 2015 ª2015 Elsevier Inc. 17
Previews From Prescription to Transcription: Genome Sequence as Drug Target Cong Guo,1,2,4 Anthony M. D’Ippolito,1,2,4 and Timothy E. Reddy1,3,* 1Center
for Genomic and Computational Biology Program in Genetics and Genomics 3Department of Biostatistics and Bioinformatics Duke University Medical School, Durham, NC 27708, USA 4Co-first author *Correspondence: [email protected] http://dx.doi.org/10.1016/j.cell.2015.06.033 2University
Personalizing treatments to account for genetically mediated differences in drug responses is an exciting opportunity to improve patient outcomes. In this issue, Soccio et al. reveal new mechanisms by which non-coding variants alter the activity of the anti-diabetic drug rosiglitazone. PPARg, a member of the class of ligandresponsive transcription factors (TFs) known as nuclear receptors, is a major driver of lipid uptake and adipogenesis. Small-molecule pharmaceuticals targeting PPARg have proven to be effective anti-diabetic agents. Although some individuals respond well to those treatments without adverse effects, others experience reduced efficacy and substantial side effects. That heterogeneity in treatment response poses a major limitation for treating type 2 diabetes and is representative of the broader challenges facing physicians when prescribing drugs that target other nuclear receptors. Drug responses are highly heritable, and understanding the genetic mechanisms contributing to that heritability will improve the application of current drugs while informing the design of new ones (Roden and George, 2002). Mutations that alter protein structure—coding mutations—are a known cause of that heritability (Hurley et al., 1991) (Figure 1, top). Meanwhile, non-coding mutations that alter steady-state levels of gene expression are emerging as a major contributor to the heterogeneity of human phenotypes, including drug responsiveness (e.g., Gusev et al., 2014). For example, genetically altered expression of drug efflux pumps has been shown to contribute to chemotherapeutic resistance in some individuals (Patch et al., 2015) (Figure 1, middle). Non-coding mutations are typically thought to act by changing the DNA sequence of TF binding sites in a way 16 Cell 162, July 2, 2015 ª2015 Elsevier Inc.
that alters the affinity of the TF to those sites (McDaniell et al., 2010). In this issue, Soccio et al. demonstrate that different alleles of PPARg sites lead to different transcriptional responses to the anti-diabetic drug rosiglitazone (Soccio et al., 2015) (Figure 1, bottom). In doing so, they recast the PPARg binding sites as drug targets and reveal new ways in which non-coding variants alter disease risk and drug responses. Their findings remarkably expand the realm of personalized medicine, particularly for drugs targeting nuclear receptors. Mutations in PPARg binding sites are only part of the story. A variety of additional TFs interact with PPARg and contribute to its genomic occupancy. The co-binding TFs each have their own DNA binding preferences, complicating our understanding of how individual noncoding mutations contribute to PPARgmediated effects of rosiglitazone treatment. Soccio et al. show that non-coding genetic variation not only alters PPARg occupancy, but also has coordinated effects on the binding of the associated TFs C/EBP and the glucocorticoid receptor. Furthermore, disruption of the DNA binding sequences of any of those TFs leads to decreased PPARg occupancy. The observed coordinated effects support a model in which TFs bind as a complex on a given allele (ENCODE Proect Consortium, 2012) and indicate that drug response heterogeneity may also result from mutations in binding sites for yet-unknown cooperating TFs. In that manner, it may be more valuable to consider regula-
tory complexes rather than individual binding sites as the key functional unit for understanding the genetic contributions to PPARg and rosiglitazone responses. Identifying additional members of PPARg regulatory complexes will also likely be valuable for predicting genetic contributions to both type 2 diabetes and its treatment. While Soccio et al. find a strong enrichment of allele-specific binding sites near rosiglitazone-induced genes, they find no evidence of a similar enrichment near rosiglitazone-repressed genes. The relative challenge of identifying the binding sites responsible for gene repression is a recurring theme in nuclear receptor biology (e.g., Reddy et al., 2009). One possible explanation is that repressive sites are inherently difficult to assay due to chromatin structure, epitope availability, or indirect interaction with the genome. Alternatively, repression by nuclear receptors may be secondary to activation of direct target genes (Abraham et al., 2006). Because nearly as many genes are repressed as are activated by nuclear receptors, determining the predominant mechanisms of that repression will be valuable for ongoing efforts to increase the specificity of nuclear receptor drugs. For example, secondary factors may themselves be targets for drugs that are beneficial either alone or in combination with nuclear-receptor-targeting drugs. To realize that potential, a revised understanding of the ways in which nuclear receptors repress gene expression is needed. In particular, determining the
REFERENCES relative contribution of each mechanism of repression to Abraham, S.M., Lawrence, T., Kleithe total repressive effects man, A., Warden, P., Medghalchi, will prioritize the most prevaM., Tuckermann, J., Saklatvala, J., lent mechanisms for suband Clark, A.R. (2006). J. Exp. sequent pharmaceutical tarMed. 203, 1883–1889. geting. Crowley, J.J., Zhabotynsky, V., Sun, Identifying non-coding variW., Huang, S., Pakatci, I.K., Kim, Y., Wang, J.R., Morgan, A.P., Calaway, ants that alter rosiglitazone’s J.D., Aylor, D.L., et al. (2015). Nat. effects in human cells creates Genet. 47, 353–360. a new challenge: how do we ENCODE Proect Consortium (2012). determine the resulting organNature 489, 57–74. ismal effects? The genetic diGusev, A., Lee, S.H., Trynka, G., versity of recombinant inbred Finucane, H., Vilhja´lmsson, B.J., mouse strains makes them a Xu, H., Zang, C., Ripke, S., Bulikpowerful resource for linking Sullivan, B., Stahl, E., et al.; mouse loci with phenotypes. Schizophrenia Working Group of Protein sequence and functhe Psychiatric Genomics ConFigure 1. Genetic Variation Affects Drug Responses tion are often conserved besortium; SWE-SCZ Consortium; (Top) Genetic variants in protein-coding exons can alter protein structure and tween mouse and human, Schizophrenia Working Group of subsequent drug responses. In the example of familial glucocorticoid resisthe Psychiatric Genomics Conand therefore it is reasonable tance, a mutation in the ligand binding domain of the glucocorticoid receptor sortium; SWE-SCZ Consortium reduces binding of glucocorticoid drugs and subsequent regulation of target to hypothesize that mutations gene expression. (Middle) Regulatory element rearrangements can change the (2014). Am. J. Hum. Genet. 95, in orthologous genes will expression of genes that govern drug responses. For example, a promoter 535–552. have similar phenotypic effusion increases the expression of ABCB1, which encodes the drug efflux Hurley, D.M., Accili, D., Stratakis, pump MDR1. Increased MDR1 is associated with chemotherapy resistance. fects. There is now growing C.A., Karl, M., Vamvakopoulos, N., (Bottom) Genetic variation in regulatory elements can disrupt TF binding and evidence that the phenotypic drug responses by altering the affinity of a drug-targeted TF. Soccio et al. Rorer, E., Constantine, K., Taylor, effects of regulatory variation demonstrate that phenomenon by revealing that genetic variants within PPARg S.I., and Chrousos, G.P. (1991). binding sites alter PPARg and cofactor occupancy and alter the response to may also be conserved beJ. Clin. Invest. 87, 680–686. rosiglitazone. tween species (Crowley et al., McDaniell, R., Lee, B.K., Song, L., 2015). Here, for example, SocLiu, Z., Boyle, A.P., Erdos, M.R., cio and colleagues use natural Scott, L.J., Morken, M.A., Kucera, variation in metabolic phenotypes beTogether, the findings of Soccio et al. K.S., Battenhouse, A., et al. (2010). Science 328, tween disparate mouse strains to implicate reveal new mechanisms by which non- 235–239. non-coding variation in PPARg binding coding variation leads to differences in Patch, A.M., Christie, E.L., Etemadmoghadam, D., sites as contributing to diabetic pheno- drug responses. Because the genetic var- Garsed, D.W., George, J., Fereday, S., Nones, K., types and rosiglitazone responses. They iants alter the genomic targets of the drug Cowin, P., Alsop, K., Bailey, P.J., et al.; Australian Ovarian Cancer Study Group (2015). Nature 521, then leverage the conserved structure itself, their findings pave the way to ulti489–494. and function of PPARg between species mately predict individual transcriptional Reddy, T.E., Pauli, F., Sprouse, R.O., Neff, N.F., to identify genetic variants with similar mo- responses to a prescribed medication. Newberry, K.M., Garabedian, M.J., and Myers, lecular effects in human tissues. Their reli- When combined with evidence of which R.M. (2009). Genome Res. 19, 2163–2171. ance on conserved TF functions between target genes contribute to downstream Roden, D.M., and George, A.L., Jr. (2002). Nat. organisms is an appealing strategy to un- drug effects, the transcriptional re- Rev. Drug Discov. 1, 37–44. cover regulatory mechanisms contributing sponses may form the basis for new apSoccio, R.E., Chen, E.R., Rajapurkar, S.R., Safato type 2 diabetes and one that may gener- proaches to identify patients who are bakhsh, P., Marinis, J.M., Dispirito, J.R., Emmett, alize broadly to other complex human most or least likely to respond favorably M.J., Briggs, E.R., Fang, B., Everett, L.J., et al. (2015). Cell 162, this issue, 33–44. traits. to treatment.
Cell 162, July 2, 2015 ª2015 Elsevier Inc. 17
