Advances in targeting 'undruggable' transcription factors with small molecules
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
REVIEwS
Advances in targeting ‘undruggable’
transcription factors with small
molecules
Matthew J. Henley 1,2,3 ✉ and Angela N. Koehler 1,2,3 ✉
Abstract | Transcription factors (TFs) represent key biological players in diseases including
cancer, autoimmunity, diabetes and cardiovascular disease. However, outside nuclear receptors,
TFs have traditionally been considered ‘undruggable’ by small-molecule ligands due to significant
structural disorder and lack of defined small-molecule binding pockets. Renewed interest in the
field has been ignited by significant progress in chemical biology approaches to ligand discovery
and optimization, especially the advent of targeted protein degradation approaches, along with
increasing appreciation of the critical role a limited number of collaborators play in the regulation
of key TF effector genes. Here, we review current understanding of TF-mediated gene regulation,
discuss successful targeting strategies and highlight ongoing challenges and emerging
approaches to address them.
DNA-binding transcription factors (TFs) represent one specificity, such an inhibitor is also likely to be less
of the most essential classes of proteins in the eukaryotic prone to compensatory resistance mechanisms common
proteome1. By binding to specific DNA sites and con- to other pharmacological modalities such as tyrosine
trolling transcriptional output of genes in close spatial kinase inhibitors9. This exceptional potential of ther-
proximity, TFs play foundational roles in the regulation apeutically modulating TF action is illustrated by the
of virtually all of a cell’s genome2. TFs dictate the identity enduring success of myriad nuclear receptor-targeting
and fate of individual cells in multicellular organisms drugs, which represent the standard of care across
by differentially regulating the common genetic code, several different disease areas10.
and are responsible for rapidly coordinating responses Despite the broad therapeutic promise of TF mod-
to internal and external stimuli by serving as end points ulators, there are major roadblocks associated with TFs
in cell signalling networks3,4. It is estimated that there are as a target class that have impeded countless attempts
at least 1,600 TFs in the human genome, around 19% of at drugging TFs outside the nuclear receptor family.
which have been associated with a disease phenotype1. Consequently, of the roughly 300 TFs that have been
Concordantly, given their central importance to bio associated with a disease phenotype, only a handful have
logy, TFs are frequent drivers of disease and represent been successfully targeted by small molecules1. A fun-
tantalizing therapeutic targets3,5,6. damental challenge is that most TFs are predominantly
The significant potential of direct TF modulators intrinsically disordered and lack classical well-formed
1
David H. Koch Institute for was best encapsulated almost two decades ago by James small-molecule binding pockets11. TFs function pri-
Integrative Cancer Research,
Darnell in the context of anticancer therapeutics5. He marily by forming highly dynamic protein–DNA
Massachusetts Institute
of Technology, Cambridge,
highlighted how TFs, more so than upstream signalling interactions and protein–protein interactions (PPIs),
MA, USA. proteins such as GPCRs or kinases, have the capac- and consequently the most critical functional sites also
2
The Broad Institute of MIT ity for highly specific disease modulation given their represent exceptionally challenging regions to directly
and Harvard, Cambridge, foundational role in selective gene regulation. That is, target with small molecules. Beyond just the basic dif-
MA, USA. a hypothetical inhibitor of a dysregulated TF could limit ficulties of TF ligand development, the regulation and
3
Department of Biological toxicity while increasing efficacy by only inhibiting tran- function of individual TF domains is often highly com-
Engineering, Massachusetts scriptional programmes driven by that TF, without the plex or poorly understood, obfuscating the domains that
Institute of Technology,
Cambridge, MA, USA.
collateral damage sometimes associated with inhibit- would actually be fruitful to modulate. This, combined
✉e-mail: mjhenley@mit.edu; ing signalling proteins that are involved with multiple with continually emerging evidence that challenges our
koehler@mit.edu signalling networks unrelated to the disease7,8. Because fundamental understanding of gene regulation and
https://doi.org/10.1038/ individual TFs typically only regulate a limited set of TF mechanisms of action12,13, makes TFs some of the
s41573-021-00199-0 gene targets that are governed by their DNA-binding thorniest targets in the proteome.
Nature Reviews | Drug Discovery
0123456789();:Reviews
Collaborators presence and/or activity of co-repressors19. By this basic
mechanism of recruitment, TFs act as the directors of
Chromatin remodelling Co-activators/ General transcriptional transcriptional output of the genome and play key roles
enzymes co-repressors machinery
across wide-ranging cellular processes18.
The structural and biophysical mechanisms by which
key TF regulatory domains function have been a sub-
ject of intense study for decades. The determinants of
DBD specificity for DNA sequence in vitro have been
Effector domain
extensively dissected with advances in high-throughput
• Recruitment
binding assays and determination of numerous DBD
structures, in both the presence and absence of DNA20.
However, significant challenges in predicting func-
Regulatory domain tional TF binding sites in the genome remain, which is
(optional)
complicated by the complex three-dimensional chro-
• Dimerization
• Nuclear transport matin architecture and an abundance of non-specific TF
• Autoinhibition binding sites that can compete with TF binding to scarcer
specific TF binding sites21.
Conversely, the basic mechanisms by which effec-
DNA-binding domain tor domains function are much less defined. Although
• Sequence-specific there are certainly some instances of well-studied and
recognition
functionally important PPIs made by single effector
domains22–24, the generalizability of these examples to the
Fig. 1 | Anatomy of a TF. All transcription factors (TFs) contain two general protein class as a whole has not been possible25–27. For example,
domains: a DNA-binding domain (DBD) that binds to specific DNA regulatory sequences, although several structures of transactivation domains
and an effector domain that recruits various transcriptional ‘collaborators’ to regulate
bound to co-activators such as CBP/p300 (ref.24) have
chromatin accessibility and transcriptional output. Many TFs also contain one or more
regulatory domains, which typically serve to regulate TF localization and functional
been proposed, these structures do not explain the
activity. repeated observation that roughly 1% of any random
sequence of amino acids — with the only commonal-
ity being the preponderance of acidic and hydrophobic
This Review synthesizes current understanding of TF residues — stitched to a DBD can function as transac-
function and gene regulation with emerging pharmaco- tivation domains25,27. Thus, the general mechanisms by
logical approaches that can or could be used to drug this which effector domains actually effectuate recruitment
target class. We discuss the basic mechanisms by which are still under considerable debate. Current universal
TFs participate in gene regulation and drive myriad dis- models of effector domain function hypothesize that
eases, and then evaluate key lessons from successful and they form non-specific dynamic PPIs with transcrip-
promising examples of TF modulator development. We tional machinery as well as phase-separating with dis-
close by highlighting technologies that could facilitate ordered regions of co-activators/co-repressors to form
progress in drugging even the most recalcitrant TFs and transcriptional condensates13,28–30, although in some indi-
reflect on how emerging medicinal chemistry, biophys- vidual cases there is evidence that other mechanisms are
ics and chemical biology approaches could be adapted more consistent with experimental data31,32. Put together,
Non-specific TF binding to address the unique challenges associated with TFs. there are many remaining questions about the mecha-
sites
Sequences of DNA that do
nisms by which the two key TF domains function that
not contain the consensus Functional domains of TFs may have drastic implications for the success of various
sequence for a transcription The key role of a TF is to recruit transcriptional regula- targeting strategies.
factor (TF) DNA-binding tory machinery to specific genomic loci to regulate gene As well as the two class-defining TF functional
domain (DBD). Most DBDs
expression14. A minimal TF is thus defined by just the domains, many TFs contain additional layers of regu-
have low affinity for
non-specific sites, but because presence of two key elements: a DNA-binding domain lation that add further complexity to their function and
of the exceptionally high ratio (DBD) that recognizes specific DNA sequences, and regulation (Fig. 1). For example, the STAT family of TFs
of non-specific to specific sites, an effector domain that recruits members of transcrip- contain a SH2 domain that controls homodimerization
TFs often spend significant tional activation or repression machinery14 (Fig. 1). TFs or heterodimerization with other STAT TFs, and thereby
time at non-specific sites.
that act as transcriptional activators use a transactivation regulates the TF localization to the nucleus33. Nuclear
Specific TF binding sites domain to recruit chromatin remodelling enzymes, his- receptors, by far the most druggable family of TFs, con-
Sequences of DNA that contain tone modifying enzymes, transcriptional co-activators tain a ligand-binding domain (LBD) that typically acts in
the consensus sequence for and/or many general TFs to increase the accessibility cooperation with a prototypical disordered transactiva-
a transcription factor (TF)
of target genes, epigenetically mark them as active, and tion domain to recruit transcriptional machinery when
DNA-binding domain.
recruit and activate RNA polymerase II (Pol II)12,14–18. bound to a small-molecule ligand34. Other TFs such as
Transcriptional condensates Conversely, TFs that behave as transcriptional repres- the basic helix–loop–helix family require dimerization
Liquid–liquid phase-separated sors use a transrepression domain to recruit chromatin with other family members to form competent DBDs35.
droplets containing remodelling and epigenetic enzymes to decrease the These diverse regulatory domains and mechanisms
transcription factors,
co-activators, RNA polymerase
accessibility of target genes and mark them as inactive17. have historically served as the most promising entry
II (Pol II) and other In some cases, prototypical transactivation domains points for medicinal chemists to develop effective TF
transcriptional machinery. can have repressive functions that are controlled by the modulators6.
www.nature.com/nrd
0123456789();:Reviews
Pre-initiation complex
Gene regulation by TFs A breakthrough in understanding how enhancers are
A large complex comprising A key lesson emerging over recent years is that eukar- placed into proximity of the target genes has been the
general transcription factors, yotic gene regulation is an exceptionally complex and identification of chromatin neighbourhoods or topo
Mediator and other proteins dynamic process that is often counter-intuitive and facil- logically associated domains (TADs)38,45–47. TADs are
that position and activate RNA
polymerase II (Pol II) at the
itates many surprising behaviours12,30,36,37. Whether a TF essentially extruded chromatin loops that are bound
transcription start site. functions at a specific binding site depends not only on by the proteins cohesin and CTCF, and enable cells
the thermodynamic stability of the TF–DNA complex to dictate the three-dimensional structure of specific
but on a number of interoperating factors, including regions of the genome38 (Fig. 2a). TADs are frequently
multidimensional DNA/chromatin architecture, the conserved within cell types and are thought to place
cooperative action of other TFs and co-activators at key cell-identity genes under the control of multiple
nearby or overlapping sites and the kinetics of the TF enhancers to maintain robust expression48. Accordingly,
binding to DNA itself13,21,38–40. Here, we focus on recent TADs can be restructured upon differentiation of pro-
insights into the mechanisms that regulate the strength genitor cells as a mechanism to remodel cell-identity
of TF-driven transcriptional activation. transcriptional programmes46. Remarkably, not all genes
within a TAD are necessarily dependent on the TAD for
Influence of genome structure on TF action. For dec- function, suggesting additional complexities to genomic
ades, it has been understood that the organization of the structure that could be relevant for selective therapeutic
genome, across several dimensions, is a key determinant targeting of genes within TADs49.
of whether a gene is turned on or off2. TFs control the A particularly noteworthy advance in the basic
expression of most genes by binding to promoter and/or biology of gene regulation has been the discovery
enhancer regions of DNA18. Promoters are characterized and characterization of super-enhancers50,51 (Fig. 2a).
by their inclusion of a transcription start site (TSS) and Super-enhancers are defined as extended clusters of
a TATA-box/Inr DNA sequence, the latter of which ena- enhancers with particularly elevated levels of bound TFs
bles assembly of the pre-initiation complex and subsequent and co-activators as well as epigenetic marks associated
activation of RNA Pol II2. Enhancers, conversely, do not with active transcription (for example, H3K27Ac). Due
contain a functional TSS and can be located up to sev- to their high sustained levels of transcriptional activ-
eral kilobases away from a TSS. Enhancers instead bind ity, super-enhancers often act in concert with TADs to
to TFs and activate transcription when placed in close control expression of key cell-identity genes48,50,51. For
spatial proximity to a promoter41. This reliance on three- example, in development, super-enhancers have been
dimensional proximity for enhancer function has many observed as regulators of core regulatory TFs that con-
remarkable consequences, chiefly that many enhancers trol the process and timeline of cell differentiation52.
only function at genes located at long genomic distances Super-enhancers are also inactivated or repurposed
(>1 kb) instead of at closer genomic loci2. Although over the course of development to initiate changes in
mechanisms of transcriptional activation at promot- cell characteristics; in cancer, these mechanisms enable
ers have been extensively characterized — down to malignancies to use super-enhancers to drive oncogenic
the structures of the pre-initiation complex at different transcriptional programmes50,51,53.
steps of the activation process42–44 — understanding
how genes are regulated by enhancer regions is still Dynamics of TF action. A crucial fact that underlies
an area of intense study and emerging therapeutically our current understanding of eukaryotic transcrip-
relevant insights. tional regulation is that it is a highly dynamic and
a b
Super-enhancer
TF1
TF2
Cohesin
Co-activator
Promoter
CTCF PIC
Enhancer
Gene RNA Pol II
Fig. 2 | overview of the modern model of the transcriptional activation process. a | Depiction of a topologically
associated domain (TAD) bound by cohesin and CTCF containing a super-enhancer that controls a gene. b | Zoomed view
of the phase-separation model of transcriptional activation, where transcription factors (TFs) and co-activators form
transcriptional condensates spanning the enhancer and promoter. PIC, pre-initiation complex; RNA Pol II, RNA
polymerase II.
Nature Reviews | Drug Discovery
0123456789();:Reviews
out-of-equilibrium process12. This has led to several and the TSS13,30. Consistent with this notion, multiple
surprising revelations about in vivo mechanisms of TF experimental efforts have demonstrated that TFs and
and co-regulator function that interact intimately with co-activators form highly concentrated puncta at active
emerging insights about genome structure. super-enhancer sites in vivo13,30,37,63. Corresponding
The remarkable dynamics of TF binding to target in vitro experiments using purified TFs and co-activators
sites in vivo has several surprising consequences. The have shown that low-complexity intrinsically disor-
classical view of TF–DNA interactions is that TFs reside dered regions (IDRs) in TFs, co-activators and gen-
on DNA for long periods of time to carry out their eral transcriptional machinery (including RNA Pol II)
function, but modern in vivo imaging studies have esti- have the capacity to co-condense into phase-separated
mated that the lifetimes of TFs with target DNA sites droplets13,37,64,65.
can be as short as a few seconds12,40,54,55. This dynamism To many, LLPS serves as an exceedingly useful frame-
is thought to serve as a regulatory mechanism to keep work to rationalize otherwise puzzling transcriptional
TFs from being trapped at non-specific DNA sites for phenomena. For example, IDRs are highly enriched in
extended time periods. For example, the relative frac- TFs and co-regulators, but because these regions are
tion of binding to specific sites over non-specific sites of disregarded in standard structure–function paradigms,
some nuclear receptors only marginally increases when it was previously challenging to understand how they
activated, but fast turnover and extended lifetimes at could participate in transcriptional regulation. Within
specific sites facilitate a rapid and significant increase in an LLPS framework, it is theoretically possible to iden-
transcriptional output at target genes40. tify functions and mechanisms of TF IDRs by simply
Dynamic TF–DNA binding can also lead to surpris- considering their physicochemical properties, concen-
ing modes of TF cooperativity. For example, whereas tration and the landscape of DNA-binding sites at a
cooperative activity of multiple TFs at a promoter or given enhancer, which together dictate their ability to
enhancer is classically thought to be enacted by different participate in transcriptional condensates66. Transitions
TFs binding to adjacent sites and stabilizing each other’s into and out of a condensate by a single protein can,
binding, rapid TF binding and unbinding has also been consequently, be facilitated by post-t ranslational
found to lead to cooperativity from different TFs bind- modifications that change IDR properties64, and the
ing the same site12,36. The low lifetimes and long periods formation and dissolution of individual condensates
of time between binding of each TF enable unimpeded can be regulated by fluctuations in composition of
exchange between different TFs, acting to keep chro- the proteins and RNA that are active during the pro-
matin in an open conformation and to recruit distinct cess of transcription65. Examples where applications of
members of the transcriptional apparatus. LLPS frameworks to the dissection of IDR functional
Other key examples of unexpected behaviour in mechanisms have given potential answers to other-
transcription have been observed during characteri- wise perplexing experimental observations include
zation of super-enhancer-driven transcriptional acti- rationalizations of transcriptional bursting 65, RNA
vation. Strikingly, super-enhancers display marked Pol II promoter release64, enhancer–promoter contact
increases in both the inter-reliance and the binding restrictions67 and the extraordinary sequence diversity
and unbinding dynamics of TFs and co-activators over of functional transactivation domains13. Phase transi-
typical enhancers51,56. The amplified cooperativity at tions have also been implicated in distinct mechanisms
super-enhancers causes them to be especially sensitive of gene regulation outside transcriptional activation,
to slight changes in TF and co-regulator composition. and are thought to play roles in the function of splicing
This can result, for example, in inhibitors of general condensates64,68, repressive Polycomb repressor com-
co-activators such as BRD4 displaying exquisite selectiv- plex (PRC) bodies69 and heterochromatin/euchromatin
ity for super-enhancer-driven transcription56. Inhibitors transitions70,71.
of general transcriptional regulatory enzymes have even It is also worth noting that the existence of transcrip-
been observed to copy the phenotypes of removing tional condensates is still somewhat contentious, due
core regulatory TFs in some cell types57–59. On the other to the challenge associated with unequivocally demon-
hand, the action of transcriptional co-repressors at strating that puncta containing TFs and transcriptional
super-enhancers can be paradoxically critical for main- machinery in vivo indeed constitute a separated liquid
taining maximum transcriptional output, exposing a phase72,73. Specifically, there have been concerns raised
highly dynamic steady state of chromatin accessibility that much of the experimental data are phenomeno-
and TF/co-regulator binding needed for super-enhancer logical, and that other mechanisms could underlie the
Cooperativity activity60–62. same observations32. Moreover, the difficulty of stud-
In transcription, a phenomenon Liquid–liquid phase separation (LLPS) has emerged ying transcriptional condensates is higher than for
where binding of one as a popular biophysical framework to rationalize the other well-characterized examples of LLPS due to the
transcription factor and/or
exceptionally cooperative and dynamic behaviour of highly dynamic and localized nature of transcriptional
co-regulator at a regulatory
element enhances the binding TFs and co-regulators at super-enhancers13,30,37 (Fig. 2b). activation. Nonetheless, there is broad agreement that
of other factors, and vice versa. Significant levels of TF binding at super-enhancers the formation of dynamic concentrated hubs of TFs
is hypothesized to create high local concentrations and transcriptional apparatus plays a critical role in
Core regulatory TFs of co-activators and other members of the transcrip- transcription, especially at super-enhancers30,37.
(Also known as master TFs).
Self-regulated transcription
tional machinery; at a critical concentration, these Altogether, the often-surprising outcomes of the
factors (TFs) that control cell TFs and other cofactors form phase-separated ‘tran- three-dimensional and dynamic nature of transcrip-
identity and fate. scriptional condensates’ spanning the super-enhancer tion strongly indicate that many general assumptions
www.nature.com/nrd
0123456789();:Reviews
about TF action developed from decades of mecha- dysregulated transcriptional programmes that are key
nistic studies in simplified systems have significant to pathogenesis and, thus, represent some of the most
potential to be misleading12. Concordantly, TF mech- direct targets for disrupting disease5. In this section,
anism of action may be highly variable from cell type we highlight disease areas where TFs are important,
to cell type, gene to gene and binding site to binding while discussing common mechanisms of transcrip-
site. When selecting possible targets to affect the func- tional dysregulation and the roles that TFs play in this
tion of a given TF, including individual domains of the process (Fig. 3).
TF itself or its co-regulatory binding partners, unbi-
ased functional data are therefore critical for effective Cancer. Dysregulated transcription is a hallmark of
decision-making. cancer, and TFs frequently serve as fundamental driv-
ers of oncogenic transformation, proliferation and
Dysregulated transcription in disease survival74. TFs can be responsible for causing onco-
A principal reason why TFs are considered highly allur- genic phenotypes by a range of diverse mechanisms.
ing therapeutic targets is that transcriptional dysreg- Overactivation and/or overexpression of TFs that control
ulation plays an essential role across a wide variety of growth pathways often drives cancer proliferation75–77.
diseases3 (Table 1). As the fundamental drivers of selec- Conversely, aberrant inactivation of tumour suppressor
tive gene expression, TFs are intimately involved in the TFs enables evasion of apoptosis and cancer survival78,79.
Table 1 | selected examples of TFs that drive disease
TF Associated diseases Dysregulation mechanisms refs
Cancer
MYC Various forms of cancer Amplifies oncogenic transcriptional programmes 89,90
MYB Various forms of cancer Overactivation by gene duplication, overexpression 84
and genetic fusions to other proteins
E2F Various forms of cancer Overactivation by dysregulation of co-repressor pRB 19,287
TAL1 T cell acute lymphoblastic leukaemia Overexpression and overactivation 288
PAX3-FOXO1 Alveolar rhabdomyosarcoma Oncogenic fusion TF, dysregulates muscle 95,289
development transcriptional programmes
p53 Various forms of cancer Downregulation by the ubiquitin–proteasome 141,142
system or loss-of-function mutations
Autoimmune and inflammatory disease
STAT1 Atherosclerosis, infection Overactivation by signalling pathways 290
STAT3 Various forms of autoimmune and Gain-of-function mutations and/or overactivation 101,102,104,114
inflammatory disease, as well as by signalling pathways (cancer, autoimmune
cancer and diabetes disease), or loss-of-function mutations
(hyper IgE syndrome)
STAT6 Asthma and allergy Overactivation by signalling pathways 103
T-bet Multiple sclerosis, systematic lupus TH1 cell master TF, drives and/or increases severity 106
erythematosus of autoimmunity
GATA3 Atopic asthma, allergies TH2 cell master TF, drives and/or increases severity 107,111
of autoimmunity
RORγt Psoriasis TH17 cell master TF, drives and/or increases severity 109
of autoimmunity
FOXP3 IPEX Loss-of-function mutation 113
NF-κB Various forms of autoimmune Overactivation by signalling pathways 77,98,99
and inflammatory disease, cancer
Diabetes
HNF1α Maturity-onset diabetes of the young Loss-of-function mutation 117
HNF4α Maturity-onset diabetes of the young Loss-of-function mutation 117
NEUROD1 Maturity-onset diabetes of the young Loss-of-function mutation 117
Cardiovascular disease
GATA4 Maladaptive cardiac hypertrophy, Overactivation (cardiac hypertrophy) 120,122
congenital heart disease or loss-of-function mutation (congenital
heart disease)
Nkx2-5 Congenital heart disease Loss-of-function mutation 122
Tbx5 Congenital heart disease Loss-of-function mutation 122
TF, transcription factor; TH1 cell, T helper 1 cell; TH2 cell, T helper 2 cell; TH17 cell, T helper 17 cell.
Nature Reviews | Drug Discovery
0123456789();:Reviews
Normal transcription
Transcription
of new gene
↑ Transcription ↓ Transcription
Upregulation Downregulation Changes in target genes
• Amplification • Loss of function • Chromatin architecture shifts
• Gain of function • Overactivation of • Gene translocations
• Pathway overactivation repressors • Fusion transcription factors
Fig. 3 | common mechanisms of transcriptional dysregulation in disease. Various mechanisms by which the normal
transcriptional profile of an example transcription factor (TF) effector gene (top middle) can be dysregulated in disease.
Effector genes that drive a disease state can be upregulated by overactive and/or overabundant TFs, which commonly
occurs by gene amplification, TF gain-of-function mutations and/or signalling pathway overactivation (bottom left).
Effector genes that protect against disease can be aberrantly downregulated by TF loss-of-function mutations and
overactivation of repressor proteins of the TF (bottom middle). Genes that are regulated by a TF in normal conditions
can change due to chromatin architecture shifts, genetic translocation of enhancers to new effector genes and TF fusions
that change or disrupt the DNA specificity of the parent TF(s) (bottom right).
Genetic fusion events that generate fusion TFs are a exhibits this ‘oncogenic addiction’ behaviour across
common cause of paediatric cancers, and typically dys- a variety of cancers is MYC. MYC is a TF in the basic
regulate developmental transcriptional programmes helix–loop–helix family that, along with its binding part-
to initiate transformation and drive proliferation 80. ner MAX, binds to the widespread E-box sequences at
Oncogenic viruses are known to initiate transformation promoters and enhancers across the genome86. MYC
via a combination of the activity of viral TFs in addition primarily functions by recruiting transcriptional elon-
to other viral proteins that co-opt or dysregulate cellular gation machinery to enhancers to increase transcrip-
TFs and transcriptional co-regulators81. There are also tional output87,88. MYC is one of the most frequently
some cases, for example in certain gliomas82, where TFs overexpressed oncogenes and is thought to act as a gen-
drive oncogenic phenotypes simply by rewiring their eral transcriptional ‘amplifier’ to drive a wide variety of
transcriptional programmes to regulate a different set oncogenic transcriptional programmes across diverse
of effector genes83. cancer types3,89,90. In vivo experiments using genetic
The TF MYB serves as an excellent example of a sin- knockdown of MYC and expression of dominant neg-
gle oncogenic TF that can act by several of the mech- ative MYC variants have shown that several distinct
anisms outlined above84. MYB is intimately involved cancers are addicted to MYC’s amplification activity,
in a variety of cancers including leukaemia (especially rapidly dying or differentiating into normal cell types
acute myeloid leukaemia (AML)), adenoid cystic car- upon MYC inhibition91–93. Similarly, TFs as a class rep-
cinoma, colorectal cancer and breast cancer, where it resent a large fraction of hits in cancer genetic depen
generally drives oncogenesis by becoming overactivated. dency databases such as DepMap94, supporting the idea
Most commonly, gene duplications and overexpression that oncogenic addiction to TFs is a shared vulnerabil-
of MYB lead to overactivation of MYB target genes, ity across myriad cancers. Thus, there is exceptionally
but MYB can also become overactivated by genetic trans- high potential for targeting TF activity as a therapeutic
locations (for example, MYB-NFIB) that fuse it to other strategy for cancer.
proteins, typically eliminating the MYB transrepression In transcription, TFs do not function alone: the
domain in the process83. In other cases, genetic changes fundamental role of TFs is to recruit the requisite
can generate new MYB binding sites that enhance other machinery to do the work required for transcriptional
oncogenic drivers. For example, in some forms of T cell regulation14. Accordingly, much of the apparatus that
acute lymphoblastic leukaemias, novel MYB binding facilitates TF-driven activation/repression can also
sites can form in the enhancer for the driver TF TAL1 be critical for maintaining oncogenic transcriptional
and increase its expression85. programmes. This is especially relevant when consid-
The central role of TFs in driving oncogenesis fre- ering the important roles that super-enhancers have
quently leads to reliance of malignancies on the activ- been shown to play in cancer, given the heightened
ity of individual TFs74. A classic example of a TF that levels of cooperativity between TFs and cofactors at
www.nature.com/nrd
0123456789();:Reviews
Chromatin readers these regulatory elements. As discussed previously, in immunodeficiency and predisposition to various types
Proteins, such as normal cells, super-enhancers often form around key of infection76. For example, inactivating mutations in
bromodomains, that bind to cell-identity genes50,51,53; it has similarly been observed STAT3 underlie many cases of hyper IgE syndrome104.
post-translationally modified that malignancies frequently generate or repurpose T cells are intimately involved in the development,
histones.
super-enhancers around key oncogenic identity and progression and severity of myriad autoimmune
effector genes53,95. Super-enhancers thereby grant cancer diseases105. Although the molecular mechanisms by
cells increased and sustained activation of these genes, which individual T cell types affect autoimmunity are
and are consequently key to maintaining an undiffer- quite complex, the master TFs that define T cell identity
entiated cell state and enabling rapid and continuous are thought to serve as general orchestrators of many
growth. autoimmune diseases. Overactivation/overexpression
Because of the exceptionally cooperative nature of master TFs in T helper 1 (TH1) cells (T-bet), TH2
of super-enhancer function, super-enhancers are fre- cells (GATA3) and TH17 cells (RORγt), for example,
quently dependent on the action of select members of is linked to several T cell-driven autoimmune diseases
the transcriptional apparatus (for example, chromatin such as multiple sclerosis, systematic lupus erythema-
readers, histone modifying enzymes, transcriptional tosus, atopic asthma and psoriasis106–110. T cell master
kinases) in addition to TFs56–60,95,96. Without even one TFs even, in some cases, protect against autoimmune
of these co-regulators and/or its associated enzymatic disease; for example, overexpression of the master TF
activity, super-enhancers can be rapidly depleted of TFs, GATA3 (TH2 cells) displayed reduced symptom severity
active chromatin marks and the transcriptional appa- of multiple sclerosis in murine models111. Similarly, the
ratus. Unique super-enhancers can also have distinct master TF FOXP3 of regulatory T cells is known to drive
cofactor dependency profiles, which can enable selec- the immunosuppressive effects of regulatory T cells and
tive inhibition of super-enhancer-driven oncogenic its expression is correlated to decreased severity of auto-
transcription57,59. For example, inhibitors of general tran- immune disease112. Inactivation of FOXP3, on the other
scriptional enzymes such as CDK9 have been shown to hand, is highly deleterious and can lead to X-linked
display strikingly selective inhibition of the oncogenic congenital immunodeficiency syndromes113.
transcription programmes of androgen receptor ∆LBD Altogether, immune response and T cell master
splice variants by this mechanism59. Inhibiting the activ- regulatory TFs make enticing targets for the numerous
ity of TF collaborators at super-enhancers therefore has diseases caused by aberrant immune responses, espe-
significant therapeutic potential for treating cancer and cially given their rich regulatory networks that provide
serves as an alternative to targeting the oncogenic TF several possible intervention points76,77,114. We also note
itself, especially in cases where the TF proves recalcitrant that many of these TFs often have a direct relationship
to small-molecule discovery efforts. to cancer as well, where dysregulated immune response
TFs have been shown to play critical roles in enabling
Autoimmune/inflammatory disease. TFs are com- transformation, invasion and metastasis33,77.
mon end points of signalling pathways that medi-
ate the immune response to infection or injury76,77. Diabetes. Diabetes mellitus, characterized by an inabil-
Consequently, dysregulation of TFs involved in immune ity to properly secrete or utilize insulin, is, in general,
response plays a significant role in the pathogenesis of a polygenic disease linked to changes in several genes
autoimmune and inflammatory diseases. For example, simultaneously115. However, there are some forms of
the TF NF-κB is a master regulator of both innate and monogenic diabetes that have been directly linked to
adaptive immunity: among other functions, it controls mutations in single TFs. For example, as previously men-
both the expression of pro-inflammatory cytokines in tioned, activating STAT3 mutations have been linked
macrophages as well as the activation and differentiation to early-onset type 1 diabetes101. A significant form of
of naive CD4+ T helper cells77. Several diverse signalling monogenic diabetes is maturity-onset diabetes of the
pathways regulate activation of NF-κB and its transit to young (MODY), which accounts for around 2% of all
the nucleus. Overactivation of NF-κB activity is strongly diabetes cases in patients younger than 20 years old116.
linked to myriad inflammatory and/or autoimmune dis- Five of the six genes that have been directly linked to
eases, such as rheumatoid arthritis, inflammatory bowel MODY are TFs, which include the hepatic nuclear fac-
disease and multiple sclerosis77,97–100. tors HNF1α, HNF1β and HNF4α, the insulin promoter
Directly downstream of NF-κ B in mediating factor IPF1 and NEUROD1 (ref.117). In all cases, loss-
immune response lies the STAT family of TFs, which of-function mutations lead to MODY. Interestingly, the
regulate the expression of cytokine-inducible genes vast majority of MODY cases are caused by mutations
such as interferons76. Individual STAT family mem- in one of the three hepatic nuclear factors that are pri-
bers are similarly implicated in numerous autoimmune marily associated with liver function and previously had
and inflammatory diseases. Overactivation of STAT no obvious connection to β-cells117. Conversely, only a
activity is, in general, linked to autoimmune disease, small fraction of MODY cases are caused by mutations
for example activating mutations of family member in IPF1 or NEUROD1, even though they both directly
STAT3 have been linked to early-onset type 1 diabetes, regulate insulin expression. HNF1α, HNF1β and HNF4α
Crohn’s disease, psoriasis and multiple sclerosis101,102. are also known to cooperate directly to regulate target
Similarly, overactivation of STAT6 is known to play sig- gene expression118, and thus possible therapeutic strate-
nificant roles in allergy and asthma103. Inactivation of gies may include the development of agonists against one
STAT family members, on the other hand, often leads to of these TFs to restore overall function.
Nature Reviews | Drug Discovery
0123456789();:Reviews
Cardiovascular disease. Cardiovascular disease, like dia- such as oestrogen, androgen and glucocorticoids) to a
betes, is a group of diseases that are linked to multiple LBD typically leads to activation of the TF by a variety
interrelated genetic risk factors and are not necessarily of mechanisms including localization to the nucleus,
driven by single proteins. However, TFs are critical to the homo- o ligomerization or hetero- o ligomerization
development and maintenance of the cardiovascular sys- and recruitment of co-activators10,125. Dysregulation
tem, and thus can play significant roles in certain forms of nuclear receptors is a feature of several cancers
of cardiovascular disease119,120. Congenital heart defects, and other diseases, and LBDs can serve as intrinsi-
for example, are commonly linked to loss-of-function cally ligandable control points for modulating tran-
mutations in master TFs that control development of scriptional activity. Accordingly, the drug discovery
the cardiovascular system119,121,122. Core regulatory car- community has exploited this fact to develop many
diovascular TFs such as GATA, HAND, MEF2 and SRF nuclear receptor drugs and chemical probes (for exam-
also play critical roles in directing the response of the ple, the FDA-approved androgen receptor antagonist
cardiovascular system to stress, and overactivation of enzalutamide) (Fig. 4).
these TFs to stimuli such as pressure and volume over- One important realization from these efforts is that
load can lead to maladaptive cardiac hypertrophy120. protein conformational flexibility underlies many of the
Therapeutic modulation of key cardiovascular TFs regulatory mechanisms controlled by LBDs10,126. Ligand
therefore has the potential for treating several forms of binding to the LBD typically activates the receptor by
cardiovascular disease. exposing a hydrophobic co-activator binding groove
as well as, in some cases, enhancing binding to nuclear
Advances in targeting TFs localization factors and other nuclear receptor mole-
Modulation of TFs by small molecules is an alluring cules. Practically speaking, this has enabled multiple
therapeutic objective given their importance across forms of modulation to be pursued (that is, agonism,
numerous diseases3,5. However, outside nuclear hormone antagonism and inverse agonism) for individual recep-
receptors10, few drugs or even well-validated chemical tors, giving drug discovery efforts a wealth of approaches
probes are known to directly target TFs. Further, many to modulate aberrant transcriptional programmes.
consider TFs to be predominantly ‘undruggable’ because Conformation flexibility is a common, if not central,
they have significant structural disorder and lack clas- feature of TFs11, which suggests that tuning of TF activ-
sical small-molecule binding pockets 5,6. The basic ity by controlling conformation may be achievable for
mechanisms by which TFs function also contribute to TFs as a class.
this image: most known effector domains are disor- Although nuclear receptors are by far the most drug-
dered when unbound to partner proteins11, and whereas gable TF family, several challenges still hamper efforts
DBDs are typically more structured, DNA-binding sur- to target all family members. For example, there are
faces tend to be highly charged and convex in shape6. many orphan receptors where the endogenous ligands
Together, these qualities can make TFs hostile enviro are unknown or where the apparent LBD does not have
nments for the development of potent and selective a ligand binding pocket127. Further, in some diseases,
drug-like small molecules. Perhaps unsurprisingly, many such as castration-resistant prostate cancer, there can
molecules that have been reported as direct TF inhibi- be expression of functional receptor splice variants that
tors have questionable structural properties and poorly lack the LBD (∆LBD), rendering LBD-targeting drugs
defined mechanisms of action123,124. ineffective128,129. Thus, due to the lack of well-defined
However, advances in structural characterization, and functional small-molecule binding pockets, the
basic biological insights and ligand design strategies challenges associated with orphan receptors and ∆LBD
have enabled the identification of several examples of receptor variants are more in line with the challenges of
drugs and high-quality chemical probes that target TFs. targeting other classes of TFs.
Below, we review the lessons learned from examples of Finally, it is also becoming increasingly recognized
successful TF targeting and discuss how these insights that nuclear receptors are not the only class of TFs that
can be applied to currently unliganded TFs. contain effector domains that bind to small-molecule
ligands. One noteworthy example is the TEAD family
Modulating TFs with ligand-binding domains. One of of TFs, which use a folded Yap-binding domain (YBD)
the most successful areas of drug discovery, in general, to recruit co-a ctivators in a mechanism remini
has been targeting TFs containing well-folded LBDs, scent of the LBDs of nuclear receptors130. TEAD TFs
namely nuclear hormone receptors. Although the spe- are an end point of the Hippo signalling pathway,
cifics of targeting this class of proteins have been exten- and are thus attractive therapeutic targets for cancer
sively discussed in the literature10, here we highlight
key concepts and lessons learned from decades of drug
discovery with nuclear receptors. Fig. 4 | examples of molecules that target TFs by various ▶
Similar to numerous other TFs, nuclear receptors mechanisms. Affinity (Kd/Ki), inhibitory activity (IC50) or
degradation activity (DC50) are included as reported in the
contain a DBD and a prototypical intrinsically dis-
literature. For indirect inhibitors of transcription factor (TF)
ordered transactivation domain (known as activat- protein–protein interactions (PPIs), the molecular target
ing function 1 (AF1))125. The class-defining feature of is in bold. DC50, half-maximal degradation concentration;
nuclear receptors is a well-folded LBD that acts as a HAT, histone acetyltransferase; IC50, half-maximal inhibitory
second tunable effector domain (AF2)125. Binding of concentration; Kd, dissociation constant; Ki, inhibitory
specific signalling molecules (for example, hormones constant; PROTAC, proteolysis targeting chimaera.
www.nature.com/nrd
0123456789();:Reviews
and regenerative medicine131,132. Notably, it has recently function remains controversial134,135, multiple efforts
been determined that TEAD family members are pal- have demonstrated that the palmitate pocket can be
mitoylated at a conserved Cys residue in the YBD, liganded by both drug-like covalent and non-covalent
which contains a deep hydrophobic pocket to bury and small molecules 136–138 (Fig. 4) . For this specific TF
stabilize this typically transient PTM133,134. Although family, significant effort is still required to develop
the specific role of the palmitoylation PTM in TEAD molecules that are selective for single family members
Effector domain modulators
Direct Indirect
O
O
O O O
O F N O
O O O NH2 OH
HN N N
N S
F3C N O
N HN N O
O H O
O NH2 H
S F
NC O O
F
Enzalutamide, androgen receptor inhibitor ‘Compound 2’, TEAD inhibitor OHM1, HIFα–p300 inhibitor Lobaric acid, MLL–CBP inhibitor
IC50 = 21 nM Kd = 230 nM Kd = 530 nM IC50 = 17 μM
Regulatory domain modulators
Direct Indirect
O O
OH O
CN
N Cl NH
O NH2
N
O N FO CN
F O NH
N H
HN H O N Cl N
O N F
H NH CF3
O F
HN
O S O OH O S
HN
F
F PO3H2
N N
SI-109, STAT3 inhibitor PT2385, HIF2α inhibitor MI-77301, p53–MDM2 inhibitor MI-1481, MLL–menin inhibitor
Ki = 21 nM Ki = 49 nM Ki = 0.88 nM Ki = 3.6 nM
Degraders
Monomeric PROTAC
Cl
H
N O
O NH2
O
N N N N
O N
N N
O O H
N O O N
NH NH O
NH O N
O O H
O O O
NH2
NH
N O
H
H2O3P
Pomalidomide, IZKF1/3 degrader BI-3802, BCL6 degrader SD-36, STAT3 degrader
Kd = 260 nM (CRBN) DC50 = 20 nM F F IC50 = 10 nM
Collaborator modulators
Cl
N F
O
H HN N
N NH NH2
N
HN
NH O
N CF3
O N
N N
O NH O
O
N O O
O O
F
HO N N N
N H H
H
ABBV-744, BRD4 BD2 inhibitor THZ1, CDK7/12/13 inhibitor KB-0742, CDK9 inhibitor A-485, CBP/p300 HAT inhibitor
Kd = 2.1 nM Kd = 3.2 nM IC50 = 6 nM IC50 = 60 nM
Nature Reviews | Drug Discovery
0123456789();:Reviews
Ubiquitin–proteasome
(due to significant risks of on-target toxicity of pan- Whereas it would be exceptionally challenging to
system TEAD inhibitors)132,139. However, the results from these develop a small molecule that binds to p53 and stabilizes
A system of intracellular initial efforts indicate exciting potential for using folded it from this mechanism of proteasomal degradation, as
protein degradation that effector domains with conserved small-molecule pockets p53 is a highly disordered TF143, an alternative approach
is mediated by transfer of
ubiquitin to target proteins
as handles for expanding the druggability of TFs. is to develop antagonists of p53 binding to MDM2. The
by ubiquitin E3 ligases to mark p53 binding site of MDM2 is, fortunately, highly drug-
them for degradation by Inhibiting TF protein–protein complexes. PPIs are a key gable: it is a relatively small and well-defined hydro-
the proteasome. feature of TF regulation and function. Most TFs are phobic pocket, which has enabled several highly potent
tightly regulated by PPIs with regulatory proteins in the peptide and small-molecule antagonists (for example,
cytosol and/or nucleus, and similarly the basic mecha- spiro-oxindole MI-77301) (Fig. 4) of this interaction to
nism of TF function (via effector domains) is to form be developed144,145. These inhibitors have been shown
PPIs with members of the transcriptional apparatus. to effectively induce apoptotic cell death in a variety of
Consequently, inhibiting specific TF PPIs is a valuable cancers by increasing p53 levels, and several are being
means for modulating TF transcriptional activity. investigated in current clinical trials140 (Table 2). Given
Inhibition of the p53–MDM2 interaction by small that many other important TFs are known to be regu-
molecules serves as an excellent illustration for the poten- lated by specific E3 ligases, such as the hypoxia-inducible
tial of drugging TF PPIs140. In cancer, the PPI between TF HIF1α by the Von Hippel–Lindau E3 ligase146,147,
the TF p53 — the ‘guardian of the genome’ — and the there is significant potential for this approach with TFs
ubiquitin E3 ligase MDM2 often functions as a mecha- that are aberrantly downregulated in disease.
nism for cancer cells to evade apoptosis by downregu- In addition to targeting E3 ligases, developing inhib-
lating p53 levels via the ubiquitin–proteasome system141,142. itors of other TF regulatory machinery is frequently
Table 2 | selected examples of TF modulators in clinical and preclinical development
Ligand Target Mechanism of action indication status clinical trials
Direct binding to TF
ARV-110 (Arvinas) Androgen PROTAC Metastatic castration-resistant prostate Ongoing NCT03888612
receptor cancer phase I/II trial
ARV-471 (Arvinas) ER PROTAC Advanced or metastatic ER+/HER2– Ongoing NCT04072952
breast cancer phase I/II trial
TTI-101 (Tvardi STAT3 Inhibitor of SH2 domain Advanced cancers including breast Ongoing NCT03195699
Therapeutics) cancer, head and neck squamous cell phase I trial
carcinoma, non-small-cell lung cancer,
hepatocellular cancer, colorectal cancer,
gastric adenocarcinoma, melanoma
JPX-1188 (Janpix, STAT3/5 Monomeric degrader Acute myeloid leukaemia Preclinical
Centessa development
Pharmaceuticals)
PT2977 (Peloton HIF2α Inhibitor of PASB Renal cell carcinoma, advanced solid Ongoing NCT03445169,
Therapeutics, Merck) dimerization domain tumours phase I and II NCT02974738,
trials NCT03634540,
NCT03401788
CB-103 (Cellestia CSL/RBPJ Inhibits formation of Advanced breast cancer, advanced/ Ongoing NCT03422679,
Biotech) NOTCH transcriptional metastatic solid tumours and phase I and II NCT04714619
complex haemotological malignancies trials
Indirect modulation of TF
AMG 232 (Amgen) MDM2 Inhibitor of p53–MDM2 PPI Acute myeloid leukaemia, advanced Completed NCT02016729,
solid tumours, glioblastoma/ and ongoing NCT01723020,
gliosarcoma, metastatic melanoma, soft phase I and II NCT04190550,
tissue sarcoma trials NCT03107780,
NCT02110355,
NCT03217266,
NCT03031730
KO-539 (Kura Menin Inhibitor of menin–MLL PPI Relapsed/refractory acute myeloid Ongoing NCT04067336
Oncology) leukaemia phase I/II trial
SY-5609 (Syros) CDK7 Inhibitor of transcriptional Advanced solid tumours Ongoing NCT04247126
kinase CDK7 activity phase I trial
KB-0742 (Kronos Bio) CDK9 Inhibitor of transcriptional Relapsed/refractory solid tumours or Phase I/II trial NCT04718675
kinase CDK9 activity non-Hodgkin lymphoma starting 2021
ABBV-744 (AbbVie) BRD4 Inhibitor of transcriptional Relapsed/refractory acute myeloid Ongoing NCT03360006,
(bromodomain 2 co-regulator BRD4 binding leukaemia, myelofibrosis phase I trials NCT04454658
selective) to acetylated histones
ER, oestrogen receptor; PPI, protein–protein interaction; PROTAC, proteolysis targeting chimaera; TF, transcription factor.
www.nature.com/nrd
0123456789();:Reviews
considered a promising avenue for TF modulation. For Critical TF PPIs can also be modulated by stabilizing
example, there is significant interest in developing inhib- or destabilizing repressed forms of the TF in the nucleus.
itors of deubiquitinating enzymes, which act in direct For example, the fusion protein CBFβ–SMMHC drives
opposition to ubiquitin E3 ligases. Deubiquitinating some forms of AML by homodimerizing and seques-
enzyme inhibitors thus have the potential to destabilize tering the RUNX1 TF from target DNA sites164, and
overactive TFs that evade the ubiquitin–proteasome effective dimeric inhibitors have been developed to
system6,148. The function of latent cytoplasmic TFs is often selectively inhibit CBFβ–SMMHC dimers and restore
tightly regulated by PPIs with cytosolic repressors and active RUNX1 (ref.165). Conversely, it has recently been
import proteins5, and as the methodology for drugging shown that inhibition of MYC activity can be achieved
PPIs has advanced, these targets have become exciting by stabilizing transcriptionally incompetent homodi-
avenues for developing TF modulators. Multiple modu- mers of its requisite binding partner MAX with small
lators of the TF NF-κB, for example, have been developed molecules, leaving MYC unable to form a functional
by targeting PPIs involved in its activation pathways149–151. DBD and causing it to be rapidly degraded166.
TFs also commonly require stable PPIs to become In addition to stable PPIs with cofactors, TF tran-
transcriptionally active. A prominent example of this scriptional activity is largely dictated by recruiting
is the STAT family of TFs, which generally require co-activators to specific genomic loci, making TF–co-
homodimerization or heterodimerization with other activator PPIs intriguing targets for controlling TF
STAT proteins to translocate to the nucleus and acti- activity. Whereas many co-activators function as gen-
vate transcription33. STAT dimerization is intrinsically eral transcriptional hubs, which could raise doubts
regulated by a SH2 domain and a tyrosine residue that about the level of selectivity achievable with this strategy,
is phosphorylated by JAK kinases upon cytokine recep- many co-activators such as CBP/p300 or Mediator con-
tor stimulation; the SH2 domain of one STAT molecule tain multiple distinct and usually well-folded activator
binds to the phosphotyrosine of the other, and vice binding domains (ABDs) that recognize specific sub-
versa. Antagonists of the phosphotyrosine–SH2 inter- sets of TFs via their transactivation domains15,24. Thus,
action therefore represent a means for direct inhibition targeting individual ABDs may be an effective avenue
of STAT activity. Several efforts have demonstrated for selective inhibition of TF activity. Major challenges
the ligandability of the SH2 domain, which has led with this approach, however, are that these PPIs tend to
to the development of potent chemical probes derived be considerably more dynamic and transient than PPIs
from phosphotyrosine mimetics as well as other between TFs and cofactors or regulatory proteins, and
non-peptidic scaffolds152 (Fig. 4). Although no STAT inhib- the functional binding surfaces of the ABDs are rela-
itor has yet successfully advanced through clinical trials, tively large and shallow. However, advances in peptid-
several inhibitors are in varying stages of clinical and omimetic strategies167,168 and increasing data indicating
preclinical development33 (Table 2). the highly allosteric nature of ABDs169,170 have enabled
There are several other noteworthy examples where some progress against these targets. For example, mod-
blocking stable PPIs required for TF activity has shown erately potent oligooxopiperazine α-helix mimetics have
significant promise. The hypoxia-inducible TF HIF2α been developed for the TAZ1 domain of the co-activator
is a well-validated target for renal cell carcinoma, and CBP/p300 (ref.167) (Fig. 4), and natural products such as
potent inhibitors — including the clinical candidates lobaric acid (Fig. 4) have been discovered to allosterically
PT2385 and PT2977 (Fig. 4; Table 2) — have been devel- inhibit the CBP/p300 KIX domain171,172. Although out-
oped that block the dimerization of HIF2α with its side the scope of this review, peptide-based strategies
obligatory cofactor ARNT by targeting the HIF2α PASB have also shown promise for targeting TF–co-activator
heterodimerization domain153–157. A similar approach PPIs173,174. The future will hopefully see the continued
has also been applied to the oncogenic TF MLL and development of more potent and selective chemical
various MLL fusion proteins, which are common driv- probes of TF–co-activator PPIs.
ers of AML. Molecules that bind the MLL cofactor
menin and inhibit its association with MLL effectively Modulating stability with molecular glues and mono-
abrogate oncogenic MLL transcriptional activity in cell meric degraders. One exciting avenue for targeting TFs
and animal models of AML158–162 (example structure in that is currently making clinical impact is the develop-
Fig. 4). Clinical trials are currently underway to inves- ment of molecular glues and/or monomeric degraders
tigate MLL–menin inhibitors as treatments for refrac- that directly control TF stability. Molecular glues func-
tory and relapsed AML (for example, KO-539) (Table 2). tion by inducing non-native PPIs between proteins and
Finally, dysregulated NOTCH signalling is implicated in have been described in the literature for decades, but
a wide variety of cancers163, and it was recently shown until recently were thought of as rare quirks of natural
that NOTCH transcriptional activity can be effectively products175. However, it has been increasingly observed
abrogated by a small molecule (now clinical candi- that molecular glues are a relatively frequent mechanism
date CB-103) (Table 2) that binds to the TF CSL/RBPJ of action for natural and synthetic bioactive molecules175.
and inhibits association with the NOTCH intracellu- A watershed moment in the field was the discovery that
lar domain8. Importantly, preclinical data highlight a the clinically approved thalidomide-based antican-
therapeutic advantage of directly targeting CSL/RBPJ, cer immunomodulatory imide drugs (IMiDs) (Fig. 4)
Molecular glues
Small molecules that directly
as the gastrointestinal toxicity commonly associated function by inducing non-native PPIs between Ikaros
mediate a non-native with upstream NOTCH inhibitors is not observed with zinc-finger (IKZF) TFs and the E3 ligase CRBN, leading
protein–protein interaction. CB-103 (ref.8). to degradation of IKZFs by the ubiquitin–proteasome
Nature Reviews | Drug Discovery
0123456789();:You can also read
