Probing the epigenome - Nature

commentary

Probing the epigenome Andrea Huston1, Cheryl H Arrowsmith1–3, Stefan Knapp4–7* & Matthieu Schapira1,8* Epigenetic chemical probes are having a strong impact in biological discovery and target validation. Systematic coverage of emerging epigenetic target classes with these potent, selective, cell-active chemical tools will profoundly influence understanding of the human biology and pathology of chromatintemplated mechanisms.

npg

© 2015 Nature America, Inc. All rights reserved.

A

dvances in genomics and proteomics methodologies in recent years have made it possible to associate thousands of genes and proteins with specific diseases, biological processes, molecular networks and pathways. However, data from these largescale initiatives alone have not translated widely into new studies on these diseaseassociated proteins, and the biomedicalresearch community still tends to focus on proteins that were already known before the sequencing of the human genome1. For instance, the human kinome, a target class of direct relevance to cancer and other disease areas, is a telling example: among research articles indexed in PubMed in 2011, 75% of the research activity focused on only 10% of the 518 human kinases—largely the same kinases that were the focus of research before sequencing of the human genome— whereas 60% of the kinome (comprising approximately 300 enzymes) was virtually ignored by the community2. We believe that the main reason for the resistance of the research community to working on new protein targets is the paucity of research tools to study the function and chemical tractability of previously uncharacterized proteins. Thus, the availability of a wider variety of tools such as antibodies and chemical probes (potent, selective and cell-active inhibitors) could potentially transform biomedical research and enable development of new target areas. Indeed, the chemical tools released, for instance, for nuclear hormone receptors about 20 years ago have had a strong effect on research in this target area: of the 37 nuclear hormone receptors with known disease association, the 18 most studied have been those for which chemical probes are available1. Chemical probes are useful as tools not only for investigation of the cellular function of proteins but also for preclinical target A full list of affiliations appears at the end of the paper. 542

validation in disease-relevant assays3. Thus, evidence is mounting that unbiased chemical coverage of target classes greatly enables basic research and drug-discovery programs. Enzymes and effector proteins involved in epigenetic mechanisms (epigenetic proteins) represent a new frontier in drug discovery, and epigenetic-related research has grown rapidly in recent years (Supplementary Fig. 1). Protein methyltransferases (PMTs) and lysine demethylases respectively deposit and remove methyl marks on histones, whereas lysine acetyltransferases (KATs) and histone deacetylases write and erase acetyl groups. These post-translational modifications can be read and interpreted by dedicated binding modules such as bromodomains (BRDs), which bind target sequence motifs in an acetylation-dependent manner, or by a large diversity of methyllysine (Kme)- and argininedependent binding domains. Enzymes and effectors controlling DNA methylation also regulate chromatin-mediated signaling. This set of 316 proteins (Supplementary Table 1) regulates gene-expression programs and cell fate during development and in response to environmental cues. Dysregulation of these processes can be a driver in disease development, and epigenetic-protein families are therefore emerging target classes in oncology and other disease-related areas4. Chromatin-associated proteins are often composed of multiple domains, each with distinct functions. For instance, the histone methyltransferase MLL contains, in addition to its catalytic domain, a bromodomain, four PHD-finger domains (which mediate a variety of protein-protein interactions) and a CXXC domain (which binds unmethylated CpG dinucleotide sequences). Precise chemical or genetic targeting of each individual domain, through chemical probes or gene-editing technologies such as CRISPR, is necessary to predict domain function in disease. Here we provide an overview of the epigenetic target space and the impact of chemical tools in this biomedical research area.

Epigenetic targets are underexplored

A parallel can be made between the current state of epigenetic research and the kinase field of 30 years ago, when great expectations for this target class were confounded by the lack of tools for prioritizing and characterizing potential therapeutic targets. In the 1980s, even the feasibility of targeting any kinase was questioned, owing to the high cellular ATP concentrations potentially competing with inhibitor binding. Natural products with promiscuous ligand binding, such as staurosporine, mitigated these concerns and served as a basis for development of more-selective inhibitors that were then studied clinically; thus, reasonable selectivity can be achieved by targeting the kinase catalytic site. Approval of the first kinase inhibitors in the early 1990s opened the floodgates, and focus on this target area intensified in pharmaceutical and academic research. Thus, chemical tools not only enable research but also demonstrate druggability of new binding sites and establish lead scaffolds for further exploitation. Consequently, a systematic chemical biology approach toward the development of high-quality chemical tools should also translate into more efficient exploration of new areas of drug discovery such as epigenetics. Given the strong and growing interest in epigenetic proteins, we analyzed recent research activity among these protein families. The number of articles indexed in PubMed and the number of patents filed for a given protein have been used as indicators of research activity. Genomewide association studies (GWAS)5, which link genetic variations in a protein with diseases, and studies of susceptibility of cancer cell lines to the genetic knockdown of a protein by short hairpin RNA (shRNA)6, serve as metrics of relevance to translational research. In light of these unbiased metrics, and as previously seen for human kinases2, epigenetic target classes suffer a strong research bias toward previously well-studied

nature chemical biology | VOL 11 | AUGUST 2015 | www.nature.com/naturechemicalbiology

commentary a 43 human bromodomain-containing proteins

Patents

BRDs 20

15

10

5

0

Chemical probes

EP300 (p300) CREBBP (CBP)

>5,000

BRPF3 0

200

400

0 800

600

1,000

Number of publications Chemical probes

EZH2 MLL

63 human protein methyltransferases


npg

OMIM

PubMed

b

Patents

shRNA

GWAS

OMIM

PubMed PRMT7

PMTs 30

GWAS

BRD4

Number of patents

40

shRNA

SETD9 20

10

0

0

200

Number of patents

400

600

800

1,000

Number of publications

Figure 1 | Leveling the landscape of epigenetics research (a,b). Despite the rapid growth in epigenetics research activity, the majority of epigenetic targets are still largely unexplored: the number of patents filed and number of articles indexed in PubMed are shown for each bromodomain-containing protein (a) and protein methyltransferase (b). This biased research activity on a small subset of epigenetic targets is unjustified, on the basis of unbiased disease-association studies such as GWAS or studies of susceptibility of cancer cell lines to shRNA knockdown (black bars). The growing availability of chemical probes should enable research on underexplored protein targets. Data used to generate the figure are provided in Supplementary Table 1 and Supplementary Figure 1.

targets. Although the two metrics used here as indicators of disease association are not comprehensive across all diseases, the lack of correlation between these unbiased signals and the volume of research activity on each gene or protein is striking, as shown here for bromodomain-containing proteins, protein methyltransferases and other epigenetictarget classes (Fig. 1, Supplementary Fig. 2 and Supplementary Table 1). For example, the number of articles indexed in PubMed exceeds 9,000 for the transcriptional activator EP300 (p300) (which has been identified in shRNA studies) and 5,000 for its close homolog CREBBP (CBP) (which has been identified in GWAS and is a frequently mutated target in various diseases). However, there are only 36 articles for the arginine methyltransferase PRMT7, a target that has been associated with diseases in both shRNA studies and GWAS: silencing PRMT7 significantly decreased breast cancer cell invasion in vitro and metastasis in vivo7,8, and single-nucleotide-polymorphism mapping at the PRMT7 locus showed associations with variations in magnesium

concentration9. Thus, we believe that a potent selective inhibitor of PRMT7 would enable mechanistic investigation of the function of PRMT7 in diverse cellular models and could possibly have direct implications in drug discovery. Interestingly, the most-studied genes seem to be enriched in diseaseassociated phenotypes, according to data in the manually curated catalog Online Mendelian Inheritance in Man (OMIM; http://omim.org/): the more research activity on a gene, the greater the chances of finding links to a disease.

Impact of epigenetic chemical probes

According to our hypothesis, the increasing availability of chemical probes during the past 5 years for both well- and under-studied epigenetic proteins should have greatly enabled biomedical research in this target area. Indeed, we find that, when available, these chemical tools have had a strong scientific influence. The research article describing the first chemical probe against a given epigenetic target is consistently the most cited article published on the targeted


protein in the same year (Supplementary Table 2). Chemical probes are not only well cited but also are extensively used: in 2014, the majority of articles published on BET bromodomains and the methyltransferase G9a included experiments using inhibitors of BET and G9a, respectively (Fig. 2a–c). An area of research in which epigenetic chemical probes have had a profound effect is oncology drug discovery. For example, the pan–BET bromodomain inhibitor JQ1 was first used to selectively kill NUT midline carcinoma (NMC) in which nuclear protein in testis (NUT) was aberrantly fused to the bromodomains of either BRD4 or BRD3 (both members of the BET bromodomain family). These gene fusions are strong drivers of tumorigenesis, giving rise to NMC, an aggressive untreatable class of squamous-cell carcinoma10. A related pan-BET inhibitor, I-BET, developed by GlaxoSmithKline, has revealed that selective targeting of BET bromodomains also modulates proinflammatory-cytokine expression and offers protection against lipopolysaccharide-induced endotoxic shock and bacteria-induced sepsis in mouse models11. The now-widespread availability of BET inhibitors has strongly stimulated research activity, and broad efficacy of BET inhibitors has been demonstrated in a variety of cancer models12. Moreover, BET inhibitors have been found to modulate other diseases and biological processes, and this may lead to the development of new drugs outside the cancer area. For instance, BET inhibition suppresses cardiomyocyte hypertrophy in vitro and pathologic cardiac remodeling in vivo and reactivates HIV from latency; it also blocks the development of mature sperm cells, thus suggesting potential applications as male contraceptive agents12. Remarkably, only a few years after their discovery, BET chemical probes have validated BET proteins as attractive therapeutic targets, and there are 18 clinical trials currently under way for BET inhibitors in oncology. Another attractive use of epigenetic chemical probes has been demonstrated for cases in which disease association had already been established for the protein target, and a chemical probe was used to test a ‘therapeutic hypothesis’ in which pharmacological inhibition of a catalytic activity, or of a specific protein-protein interaction, could replicate the effects of genetic deletions. Such examples include inhibitors of methyltransferases with strong clinical potential, such as DOT1L, EZH2 and PRMT5. Chemical probes for the catalytic activity of these enzymes mimic the therapeutic effects of shRNAs directed against these targets, and they have been shown to be efficacious in xenograft models of 543

interaction of Menin (a component of the MLL complex) with lens epithelium–derived growth factor (LEDGF), or the interaction of WD repeat–containing protein 5 (WDR5) with MLL, effectively disrupted the MLL complex and demonstrated profound ontarget activity in leukemia and prostatecancer models19–22. Interestingly, several of the recently reported epigenetic chemical probes target subunits of either wild-type or mutant MLL complexes—very large multiprotein complexes implicated in several types of leukemias and normal development (Fig. 3a–d). Thus, these tools can be used to further strengthen the therapeutic hypothesis of targeting the MLL complex in disease and to dissect its role in developmental pathways. Epigenetic chemical probes have also been used in more exploratory settings to discover new biology. For example, a phenotypic screen of 2,000 compounds revealed that BIX-01294, an inhibitor of the

mixed-lineage leukemia and non-Hodgkin’s and mantle-cell lymphomas, thereby further validating the chemical tractability of the methyltransferase target class13,14. EZH2 chemical probes have also been used to demonstrate dependence of SWI/SNFmutated cells on the catalytic activity of the Polycomb repressive complex 2 (PRC2) and the consequent sensitivity of rhabdoid and ovarian tumors to EZH2 inhibition14,15. Similarly, ependymomas with a CpG island– methylator phenotype have been shown to be exquisitely sensitive to EZH2 inhibitors16. Pharmacological activation of the histone deacetylase SIRT1 sensitized mixed-lineage leukemia to a DOT1L chemical probe, thus raising the possibility of combination therapy17, and resistance of metastatic breast cancer cells to the phosphoinositide 3-kinase inhibitor GDC-0941 was overcome when it was combined with BET inhibitors18. As a final example, chemical probes targeting the a

70

BET BET inhibitors BET minus inhibitors

120

Number of publications/year


commentary

100

G9a

UNC0638

G9a inhibitors

60

G9a minus inhibitors 50

80

40

JQ1/iBE JQ1 or I-BET

60

Bix-01294

30 40 20 20 10 0

1996

0

1998 2000 2002 2004 2006 2008 2010 2012 2014

2000

2002

2004

2006

Year

b

npg

2010

2012

2014

Landmark BET chemical-probe publications First BETinhibitor patent

cMYC regulation

JQ1 NMC

2009

AML inhibition

Isoxazoles Male contraception

2011

2012

I-BET

I-BET-151

Suppression of inflammation

c

2008

Year

Dual kinase-BET inhibition

PFI-1

2013

2014 Heart failure

HIV latency activation

Chemical probe published

Superenhancer regulation

First clinical trials in cancer

2015

Chemical probe used

Landmark G9 and GLP chemical-probe publications

BIX01294

UNC0638, less toxic

Myogenic differentiation

Reprogramming to ESC

2007

2009

Polycomb gene regulation

A-366 Ototoxicity and hair-cell death

2011

Response to hypoxia Reprogramming to iPS

Early differentiation

H3K9me2 patterning at CpG islands

2013 Cholesterol biosynthesis

Hypoxia-mediated signaling

2015

UNC0642, in vivo

Hematopoietic-lineage commitment

HoxA9-dependent AML

Figure 2 | Epigenetic chemical probes are extensively used. (a) A surge in research activity followed the publication and wide availability of inhibitors of BET bromodomain (left) and G9a (right). (Red, articles in which chemical probes were used; gray, articles in which chemical probes were not used; black, all articles.) (b,c) Chemical probes have been instrumental in revealing the clinical relevance of the BET bromodomains (b) and in exploring the biology of the methyltransferases G9a and GLP (c). AML, acute myeloid leukemia; ESC, embryonic stem cells; iPS, induced pluripotent stem cells; H3K9me2, dimethylated histone H3 Lys9. 544

methyltransferases G9a and GLP, was able to reprogram mouse embryonic fibroblasts transduced with only two transcription factors into embryonic stem cells. The effect was further improved when the calciumchannel agonist BayK8644 was combined with BIX-01294, whereas BIX-01294 alone could efficiently reprogram neural progenitor cells into induced pluripotent stem cells23,24. These results suggest that G9a (and/or GLP) is an important regulator of cellular pluripotency. UNC0638, a second-generation, less-toxic chemical probe for G9a and GLP, was also found to alter patterning of dimethylated histone H3 Lys9 and chromatin structure at CpG islands in hematopoietic progenitor cells25. In another example, the G9a inhibitors BIX-01204 and UNC0638 protected hair cells from neomycin-induced damage in vitro, and from aminoglycoside-induced damage in vivo, thus suggesting a potential avenue for the treatment of sensorineural hearing loss26. Finally, prolonged exposure of lymphoma cells to a chemical inhibitor of EZH2 revealed resistance mechanisms that should be addressed by next-generation inhibitors27. Thus, when chemical probes are comprehensively validated28 (so that target engagement is clearly established in cells at low concentration, and selectivity is evaluated on a diverse array of enzymes and receptors) and are properly applied3 (so that doses do not exceed ten times the cellular halfmaximal inhibitory concentration (IC50), and inactive derivatives of the probe are used as ‘control’ compounds), they can be profoundly influential in target discovery and in broader understanding of human biology and pathology.

Outlook

The chemical-probe coverage of BRDs and PMTs is now progressing rapidly (Fig. 1), and these new chemical tools will probably be instrumental in probing target function and relevance for therapy. Lysine demethylases have so far proven more challenging, owing to the poor cell penetrance of metal-chelating inhibitors. Repeated efforts to develop chemical probes against KATs have been largely unsuccessful, and new screening strategies, for instance those incorporating other components of KAT–chromatin complexes to increase druggability, target allosteric sites or inhibit protein interactions, should be explored. Kme reader domains represent a barely explored landscape consisting of almost 200 proteins. The recently reported chemical probe for the MBT-domain protein L3MBTL3 (ref. 29) provides proof of principle that Kme-binding domains may be chemically tractable, yet further studies are required to reveal whether this target class will be fertile ground for drug discovery.


commentary a

b

EP300 H3K4me3

CBP MLL N

Oncogenic fusion

MLL N

MLL C SET

WDR5

Menin

BRD4

Kac DO T1L

LE DG F

LE DG F

Menin

H3K79me3

Oncogenic reprogramming

Homeostasis

c

d

EP300 MLL N

MLL PTD

MLL C

N

WDR5

Oncogenic reprogramming

Poor treatment outcome

O

CF3 HN

Cl

N

N N

N

N

N N

MI-503 Menin Kd 9.3 nM

CBP-30 CBP–EP300 Kd