LETTER doi:10.1038/nature14904 Mediator kinase inhibition further activates super-enhancer-associated genes in AML Henry E. Pelish1*, Brian B. Liau1*, Ioana I. Nitulescu1, Anupong Tangpeerachaikul1, Zachary C. Poss2, Diogo H. Da Silva1, Brittany T. Caruso1, Alexander Arefolov1, Olugbeminiyi Fadeyi1, Amanda L. Christie3, Karrie Du1, Deepti Banka4, Elisabeth V. Schneider5,6, Anja Jestel5,GeZou1, Chong Si1, Christopher C. Ebmeier2, Roderick T. Bronson7, Andrei V. Krivtsov8, Andrew G. Myers1, Nancy E. Kohl3, Andrew L. Kung9, Scott A. Armstrong8, Madeleine E. Lemieux10, Dylan J. Taatjes2 & Matthew D. Shair1 Super-enhancers (SEs), which are composed of large clusters of terms pertinent to haematopoiesis, cellular differentiation and tran- enhancers densely loaded with the Mediator complex, transcrip- scription, supporting the notion that SEs regulate cellular identity tion factors and chromatin regulators, drive high expression of (Supplementary Table 1). genes implicated in cell identity and disease, such as lineage- To determine whether pharmacological inhibition of Mediator controlling transcription factors and oncogenes1,2. BRD4 and kinases regulates SE function and inhibits AML proliferation, in ana- CDK7 are positive regulators of SE-mediated transcription3–5.By logy to BRD4, we characterized CA (Fig. 2a) as an inhibitor of CDK8 contrast, negative regulators of SE-associated genes have not been and its paralogue CDK19 (77% identical overall and 94% in the cata- well described. Here we show that the Mediator-associated kinases lytic domain). CA was reported to bind CDK8 and CDK19, as well as cyclin-dependent kinase 8 (CDK8) and CDK19 restrain increased ROCK1 and ROCK2, as individual proteins in vitro10. We synthesized activation of key SE-associated genes in acute myeloid leukaemia CA11,12 and determined that it potently inhibited the kinase activity of (AML) cells. We report that the natural product cortistatin A (CA) the CDK8 module in vitro (half-maximum inhibitory concentration selectively inhibits Mediator kinases, has anti-leukaemic activity (IC50) 5 12 nM; Fig. 2b and Extended Data Fig. 2a). By contrast, CA in vitro and in vivo, and disproportionately induces upregulation did not inhibit other transcriptional cyclin-dependent kinases CDK7 of SE-associated genes in CA-sensitive AML cell lines but not in CA- (TFIIH), CDK9 (P-TEFb), CDK12 or CDK13 in vitro, nor did it bind insensitive cell lines. In AML cells, CA upregulated SE-associated CDK9, CDK12, CDK13, ROCK1 or ROCK2 up to 2,500 nM in genes with tumour suppressor and lineage-controlling functions, including the transcription factors CEBPA, IRF8, IRF1 and ETV6 abRegion length (kb) MED1 SE (refs 6–8). The BRD4 inhibitor I-BET151 downregulated these SE- 0150 Group 130 RNA pol II BRD4 MED1 CDK8 H3K4me1 associated genes, yet also has anti-leukaemic activity. Individually H3K27ac increasing or decreasing the expression of these transcription factors 69 suppressed AML cell growth, providing evidence that leukaemia 41 175 BRD4 SE cells are sensitive to the dosage of SE-associated genes. Our results 17 74 CDK8 SE demonstrate that Mediator kinases can negatively regulate SE-assoc- 19 iated gene expression in specific cell types, and can be pharmacolo- gically targeted as a therapeutic approach to AML. c 20 kb 5,000 CDK8 associates with CCNC (cyclin C), MED12 and MED13 to CDK8 form a CDK8 module that can reversibly associate with the 26-subunit 10,000 MED1 9 Mediator complex . Because SEs are disproportionately loaded with 20,000 Mediator2, we examined whether CDK8, as a Mediator-associated BRD4 kinase, might regulate SE function. Using chromatin immunoprecipi- 10,000 RNA pol II 16,203 CDK8-positive regions tation followed by sequencing (ChIP-seq), we mapped the genome- 5,000 H3K27ac wide occupancy of CDK8, along with known SE-associated factors Scaled reads per million Scaled reads 1,000 and histone modifications, in the AML cell line MOLM-14. Semi- H3K4me1 supervised hierarchical clustering revealed that CDK8 most closely CEBPA associated with MED1, followed by BRD4 and histone 3 Lys27 acet- Regular Super-enhancer enhancer ylation (H3K27ac), at putative enhancer elements marked with –0.050 0.03 H3K4me1 (red bar, Fig. 1a and Extended Data Fig. 1a–c). A fraction of these regions was particularly large and loaded with CDK8, MED1 Figure 1 | CDK8 is asymmetrically loaded at SEs in MOLM-14 cells. a, Clustering of total ChIP-seq signal of CDK8, MED1, BRD4, H3K27ac, RNA and BRD4, suggesting that they may represent SEs. Consistent with pol II and H3K4me1 on CDK8-positive regions. Each respective cluster is this notion, most of the CDK8, MED1, BRD4 and H3K27ac ChIP-seq ordered by CDK8 signal. The red bar indicates the cluster most highly enriched signal was disproportionately located on a small number of SEs iden- for the factors listed above. b, Overlap between SEs independently identified tified by each factor separately (Extended Data Fig. 1d–f). These SEs by ChIP-seq signal for CDK8, MED1 and BRD4 based on the collapsed superset significantly overlapped (Fig. 1b, c and Supplementary Table 1). Genes of regions identified by any one factor. c, ChIP-seq binding profiles at the associated with these SEs were enriched with Gene Ontology (GO) CEBPA locus. 1Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts 02138, USA. 2Department of Chemistry and Biochemistry, University of Colorado, Campus Box 596, Boulder, Colorado 80303, USA. 3Lurie Family Imaging Center, Dana-Farber Cancer Institute, Boston, Massachusetts 02215, USA. 4Division of Hematology/Oncology, Children’s Hospital, Boston, Massachusetts 02215, USA. 5Proteros Biostructures GmbH, Bunsenstrasse 7a, D-82152 Martinsried, Germany. 6Max-Planck-Institut fu¨ r Biochemie, Am Kloperspitz 18, D-82152 Martinsried, Germany. 7Dana-Farber Cancer Institute, Boston, Massachusetts 02215, USA. 8Cancer Biology and Genetics Program and Department of Pediatrics, Memorial Sloan Kettering Cancer Center, New York, New York 10065, USA. 9Department of Pediatrics, Columbia University Medical Center, New York, New York 10032, USA. 10Bioinfo, Plantagenet, Ontario K0B 1L0, Canada. *These authors contributed equally to this work. 8 OCTOBER 2015 | VOL 526 | NATURE | 273 G2015 Macmillan Publishers Limited. All rights reserved RESEARCH LETTER abc 100 d CDK8/19 common 100 peptide H106 75 W105 A155 D103 1 μM CA OH 75 HO 50 100 nM CA A100 TFIIH Me 35 Me O 13 N 50 P-TEFb N 3 5 8 H CDK8 module Me H Inhibition (%) 0 Kinase activity (%) 25 Cortistatin A (CA) –25 0 –40 10,000 217 peptides representing 1 10 100 1,000 Dimethylamine CA (nM) up to 220 kinases e 125 f 125 MOLM-14, 7-day treatment g Flag–CDK8 WT W105M CA (nM) 0 40 20 10 1 0 40 20 10 1 100 HCT116 100 Flag–CDK8 STAT1-pS727 HEL (W105M) 75 K562 75 Flag–CDK8 β-actin TF-1 Flag–CDK19 50 50 GI50 ≤ 10 nM (W105M) Flag–CDK19 WT W105M Growth (%) Growth (%) MEG-01 MOLM-14 Flag–CDK19 CA (nM) 0 40 20 10 1 0 40 20 10 1 25 UKE-1 RS4;11 25 SET-2 SKNO-1 STAT1-pS727 0 MV4;11 0 0.1 1 10 100 1,000 110100400 β-actin CA (nM) CA (nM) Figure 2 | CA suppresses AML cell proliferation by inhibiting Mediator gold, CDK8 in grey) with contact residues and CA in stick representation. kinases. a, CA structure with N,N-dimethylamine in red, C5–C8 ethano bridge Dotted red lines indicate hydrogen bonds. e, Effect of CA on growth of in magenta, C13-methyl in green and isoquinoline in blue. b, Phosphorylation indicated cell lines (mean 6 s.e.m., n 5 3 biological replicates, one of two of the RNA pol II CTD (mean 6 s.e.m., n 5 3 biological replicates, one of experiments shown). f, Sensitivity of MOLM-14 cells to CA after expression of two experiments shown, autorad in Supplementary Fig. 1). c, Kinome profiling indicated kinases (mean 6 s.e.m., n 5 3 biological replicates, one of two in MOLM-14 lysate (mean, n 5 2 biological replicates, experiment performed experiments shown). g, Immunoblot showing IFN-c-stimulated STAT1-S727 once, values , 35% indicate no change). d, CA-binding pocket of CDK8 phosphorylation in MOLM-14 cells expressing indicated kinases and treated from CA–CDK8–CCNC crystal structure (semi-transparent surface; CA in with CA (one of two experiments shown, full scan in Supplementary Fig. 1). MOLM-14 cell lysate (Fig. 2b and Extended Data Fig. 2b, c). In cells, (Extended Data Fig. 4b, c). Although SET-2 and HEL cell lines contain CA dose-dependently inhibited the phosphorylation of known CDK8 the JAK2(V617F) mutation, and MEG-01 and K562 contain the BCR- substrates STAT1-S727 (ref. 13; IC50 , 10 nM), Smad2-T220 and ABL translocation, megakaryoblastic cell lines SET-2 and MEG-01 Smad3-T179 (ref. 14; IC50 , 100 nM) (Extended Data Fig. 2d). No cells were sensitive to CA whereas erythroleukaemia-derived cell lines kinase substrates have been reported for CDK19. HEL and K562 were not, suggesting that cell lineage may be a con- We more broadly evaluated CA selectivity in cell lysate (using tributing determinant for CA sensitivity18. The phenotypic effects of KiNativ, see Methods)15 and in vitro, which collectively tested 387 CA were cell-line-dependent. CA treatment increased megakaryocyte kinases. At 100-times its CDK8 IC50 value, CA was fully selective in markers CD41 and CD61 on SET-2 cells, whereas CA treatment of MOLM-14 cell lysate for CDK8 and CDK19, and in vitro only inhib- MOLM-14, MV4;11 and SKNO-1 cells increased cleaved PARP levels, ited the CDK8–CCNC complex and GSG2, the latter of which we annexin V staining and the sub-G1 cell population, consistent with disqualified as a cellular target of CA (Fig. 2c, Extended Data Fig. 2c, apoptosis (Extended Data Fig. 4d–f). e–h, Supplementary Table 2 and Supplementary Information).