Targeted Protein Degradation: from Chemical Biology to Drug Discovery

Abstract

Traditional pharmaceutical drug discovery is almost exclusively focused on directly controlling protein activity to cure diseases. Modulators of protein activity, especially inhibitors, are developed and applied at high concentration to achieve maximal effects. Thereby, reduced bioavailability and off-target effects can hamper compound efficacy. Nucleic acid-based strategies that control protein function by affecting expression have emerged as an alternative. However, metabolic stability and broad bioavailability represent development hurdles that remain to be overcome for these approaches. More recently, utilizing the cell’s own protein destruction machinery for selective degradation of essential drivers of human disorders has opened up a new and exciting area of drug discovery. Small molecule-induced proteolysis of selected substrates offers the potential of reaching beyond the limitations of the current pharmaceutical paradigm to expand the druggable target space.

eTOC

Small molecule-induced proteolysis has emerged as a powerful and promising strategy, capable of reaching beyond the boundaries presented by traditional drug discovery. Cromm and Crews summarize recent advances in the field and discuss future challenges as well as opportunities.

Introduction

Tight regulation of the cellular proteome is critical for the flawless interplay of different proteins necessary for normal cellular function, survival and proliferation. Part of this regulatory network is the control of protein synthesis and degradation. The Ubiquitin Proteasome System (UPS) plays a central role in protein homeostasis with the proteasome as the major component of the eukaryotic protein degradation machinery (Finley, 2009). Proteins marked for proteasomal degradation are tagged via covalent attachment of ubiquitin to surface lysines (Komander and Rape, 2012; Yau and Rape, 2016). Inherited or acquired diseases are often based on abnormal protein functioning, which is currently targeted using a predominantly occupancy-based pharmaceutical strategy; inhibitors bind to disease-implicated proteins and the longer protein function is blocked by inhibitors, the larger the clinical benefit achieved. Therefore, high local inhibitory concentrations (IC_90–95) need to be maintained at all times to assure therapeutic efficacy (Figure 1). However, this often results in off-target binding and side effects (Adjei, 2006). An alternative is presented by event-driven strategies whereby drug binding triggers an event that reduces the cellular levels of the disease-implicated protein. Many of these approaches e.g., small interfering RNA (siRNA), antisense oligonucleotides or genome editing strategies such as CRISPR–Cas9 (Le Cong et al., 2013) are nucleotide-based and limited by their low in vivo stability and narrow bioavailability (Whitehead et al., 2009; Deng et al., 2014; Conde and Artzi, 2015). Recently, however, small molecules have been used to selectively induce the degradation of a variety of interesting target proteins (Bondeson et al., 2015; Winter et al., 2015). This technique, called PROteolysis-TArgeting Chimeras (PROTACs), provides a highly promising new modality for drug discovery and is capable of reaching beyond the boundaries posed by traditional drug discovery (Toure and Crews, 2016; Bondeson and Crews, 2017; Lai and Crews, 2017; Ottis and Crews, 2017; Salami and Crews, 2017). This review summarizes recent advances in small molecule-induced proteolysis of targeted proteins and provides an outlook on future opportunities and challenges in the field.

Figure 1. Pharmacology models.

Many diseases are caused by abnormal protein function. Occupancy-driven pharmacology blocks malfunctioning proteins via inhibition, i.e., applying high concentrations of inhibitor. In event-driven pharmacology, protein function is controlled by decreasing the cellular protein level. The disease-implicated protein is displayed in dark yellow, the applied inhibitor in red.

Ubiquitin Ligases

Proteins are marked for proteasomal degradation via post-translationally modification with ubiquitin. However, addition of ubiquitin regulates a multitude of diverse cellular processes and functions (Kerscher et al., 2006; Grabbe and Dikic, 2009; Komander, 2009). These include subcellular localization, protein sorting and protein–protein interactions (PPIs), with the protein substrates themselves ultimately subjected to different actions including protein activation, inhibition as well as proteasomal or lysosomal degradation depending on their ubiquitination pattern (Komander and Rape, 2012). Ubiquitination can occur as monoubiquitination (a single ubiquitin on a single lysine), multi-monoubiquitination (single ubiquitins on multiple lysines) or as polyubiquitination by the formation of ubiquitin chains. Multiple different ubiquitin chains have been identified as ubiquitin itself can be ubiquitinated at any of seven internal lysines or N-terminus. The attachment of ubiquitin is conducted by a sequential cascade of three classes of enzymes: namely, ubiquitin-activating enzymes (E1), ubiquitin-conjugating enzymes (E2) and ubiquitin-protein ligases (E3) (Pickart and Eddins, 2004). In a first step, the C-terminus of ubiquitin is activated in an ATP-dependent reaction forming a thioester with the active-site Cys of an E1 enzyme (Figure 2A). Ubiquitin is subsequently trans-thiolated to an E2 active-site Cys yielding an E2–Ub intermediate. In the final step, an E3 binds simultaneously to the E2–Ub intermediate as well as the substrate protein to mediate the transfer of ubiquitin to the substrate. E3s exert a special role within this cascade as they bring both reaction partners in close spatial proximity and simultaneously enhance the rate of ubiquitin transfer in a catalytic manner (Deshaies and Joazeiro, 2009; Berndsen and Wolberger, 2014). As E3s must be able to interact very specifically with a broad array of substrates, unsurprisingly more than 600 E3s are encoded in the mammalian genome (Li et al., 2008). This large number is in contrast to only two ubiquitin E1s and ~40 E2s identified thus far (Jin et al., 2007; Pelzer et al., 2007). Substrate recognition of E3s is essential for efficient and selective target degradation and is mediated by a variety of well-defined peptide motifs known as degrons (Varshavsky, 1991; Lucas and Ciulli, 2017). Due to their implication in numerous pathophysiological conditions and their ability to regulate protein function and stability, E3s themselves have emerged as drug targets (Kirkin and Dikic, 2011; Lipkowitz and Weissman, 2011; Bielskiene et al., 2015).

Figure 2. The ubiquitination machinery.

(A) ATP dependent ubiquitin activation by ubiquitin-activating enzyme (E1) and transfer to an ubiquitin-conjugating enzyme (E2) results in an E2–Ub intermediate. (B) For RING E3s, ubiquitin is directly linked to the protein of interest (POI) after simultaneous binding to E2–Ub and POI. (C) Simplified schematic blueprint of a Cullin RING E3. Cullin RING E3s are multi-subunit RING ligases comprising an F-box protein for substrate recognition, a RING domain for E2– Ub binding and additional regulatory proteins (omitted for simplicity – more detailed information about Cullin RING ligases can be found elsewhere (Hua and Vierstra, 2011)). (D) HECT- and RBR-type E3s catalyze the ubiquitin transfer in a two-step process. Before ubiquitin is tethered to the POI it is trans-thiolated to the active site of the E3.

E3s can be divided into three different classes depending on their conserved structural domains and mode of action (Metzger et al., 2012; Berndsen and Wolberger, 2014). The predominant class is the Really Interesting New Gene (RING) family E3s (Figure 2B) that bear a RING or RING-like domain responsible for E2 binding and catalyze the direct ubiquitin transfer from the E2 enzyme to the substrate (Lorick et al., 1999; Budhidarmo et al., 2012). RING E3s exist as monomers, dimers or even vast multi-subunit assemblies like the Cullin RING ligase (CRL) superfamily (Petroski and Deshaies, 2005; Hua and Vierstra, 2011; Metzger et al., 2014). E3 CRLs are comprised of a substrate recognition domain (F-box protein) and an E2 binding RING domain, which are connected by the scaffolding protein Cullin (Figure 2C). In contrast to RING E3s, the ubiquitin transfer of the Homologous to the E6AP Carboxyl Terminus (HECT) and the RING-between-RING (RBR) family E3s occurs in a two-step process (Figure 2D). These E3s facilitate the ubiquitin conjugation via a thioester intermediate with the active site Cys of the E3 before it is attached to the substrate. Despite the variety of diverse mechanisms, ubiquitination patterns and functions, the most studied and common ubiquitin assembly is a chain formation (minimum 4 ubiquitins) linked through Lys48, which tags the substrate for proteasomal degradation.

Targeted proteasomal degradation

In the past few decades, advances in genetics, genomics and cell biology have facilitated the identification and validation of a myriad of highly interesting drug targets. However, the druggability of these newly identified target proteins is hampered by the available tools in drug discovery, which have not kept pace with the fast-emerging methods of target identification. For the past 20 years, drug development has almost exclusively focused on small molecules (enzymatic inhibitors, GPCR or ion channel ligands) or biologics such as antibodies. These drug modalities render only a small portion of the human proteome (~20%) pharmaceutically accessible, and are unable to target whole classes of proteins known to play key roles in disease, e.g. transcription factors, scaffolding proteins or non-enzymatic proteins inside the cell (Hopkins and Groom, 2002; Wells and McClendon, 2007; Surade and Blundell, 2012; Lazo and Sharlow, 2016). Small molecule induced proteolysis offers an alternative strategy as it combines the drug-like properties of small molecules with an event-driven intracellular control of proteins levels as transient binding can already be sufficient for effective target degradation (Toure and Crews, 2016; Bondeson and Crews, 2017; Lai and Crews, 2017; Ottis and Crews, 2017; Salami and Crews, 2017). This approach offers the potential to reach beyond the limitations of traditional drug discovery and target the “undruggable” proteome, as binding to the protein of interest (POI) does not need to be connected to inhibition of the POI since it induces its removal from the system.

HSP90 inhibition

Early attempts to induce protein degradation were based on blocking the molecular chaperone Heat Shock Protein 90 (HSP90). HSP90 engages more than 200 different client proteins and helps to protect from cellular stress by facilitating their correct folding. HSP90 was identified as a drug target after its upregulation was observed in many cancer cells (Jego et al., 2013). HSP90 inhibition results in degradation of its client proteins, which are in many cases essential for cell proliferation and survival. However, the high expectations of HSP90 inhibitors remain unfulfilled as they have not yet been approved by the FDA due to poor in vivo pharmacological properties and severe side effects (Taldone et al., 2008; Hong et al., 2013; Chatterjee et al., 2016). These shortcomings might be based on the multitude of client proteins affected by HSP90 inhibition rendering this approach unspecific. Consequently, a more selective and predictable approach to induced protein degradation was desired.

Selective Estrogen Receptor Downregulators

The small molecule regulated transcription factor Estrogen Receptor α (ERα) is a well-known driver of oncogenic signaling in cancer and has long been an established drug target (Liang and Shang, 2013). Originally designed as modulators of protein function, Selective Estrogen Receptor Downregulators (SERDs) were the first molecules identified to degrade their target protein selectively (Dauvois et al., 1992). Since fulvestrant (Faslodex, AstraZeneca) gained FDA approval in 2002 additional compounds claiming to work as SERDs have entered clinical trials. The exact mode of action by which SERDs induce ERα degradation is still not fully elucidated and might vary among different compounds. However, it is believed that SERD binding to ERα induces structural changes that result in increased hydrophobic surfaces and subsequent target degradation via mechanisms that safeguard proper protein folding (Wu et al., 2005; Wittmann et al., 2007). Similar effects were observed for Selective Androgen Receptor (AR) Downregulators (SARDs) (Bradbury et al., 2011; Loddick et al., 2013).

Hydrophobic Tagging

The observation that hydrophobic patches on the surface of proteins can be recognized by the endogenous protein quality control machinery and result in subsequent protein degradation inspired the idea of Hydrophobic Tagging (HyT). Adding a strongly hydrophobic moiety such as adamantyl or Boc₃Arg to ligands mimics, upon binding to their cognate POI, a partially unfolded protein and triggers an unfolded protein response to remove the faulty protein (Figure 3A) (Neklesa and Crews, 2012). This strategy translates the SERD mode of degradation into small molecule probes capable of targeting various proteins for destruction. To explore the general applicability of HyT, both cytosolic and transmembrane HaloTag fusion proteins were degraded by means of adamantyl–chloroalkane probes (Neklesa et al., 2011). POI degradation could not only be monitored under cell culture conditions but also in live animals such as zebrafish and mice. Although this study was limited to the expression of a POI–HaloTag fusion protein, it demonstrated the selective degradation of the target protein by non-toxic, low molecular weight adamantyl probes. Moving away from genetic fusion proteins, another study reported the degradation of glutathione-S-transferase and dihydrofolate via HyT using a Boc₃Arg fused ligand (Long et al., 2012). Interestingly, proteasomal degradation of Boc₃Arg-tagged proteins occurs in an ATP- and ubiquitin-independent mechanism that requires further investigation. Furthermore, Boc₃Arg was recently reported to inhibit the translational machinery by engaging the mammalian Target Of Rapamycin Complex 1 (mTORC1), which may result in severe off-target effects and reduce its applicability in in vivo systems (Coffey et al., 2016). The enormous potential of the HyT technology was demonstrated by reaching beyond the limits of conventional drug discovery and targeting the undruggable pseudokinase Human epidermal growth factor receptor 3 (Her3), which primary functions via a scaffolding role (Xie et al., 2014). Overexpression of Her3 is associated with various forms of cancer and its minimal residual kinase activity hampers the use of classic kinase inhibitors (Yarden and Pines, 2012). Nevertheless, linking a low nanomolar covalent Her3 binder with an adamantyl moiety allowed partial degradation of Her3 at submicromolar concentrations (Xie et al., 2014; Lim et al., 2015). Her3 depletion resulted in reduced signaling and proliferation of Her3-dependent cell lines. As demonstrated in a different study, HyT is capable of turning a receptor agonist into a degrader (Gustafson et al., 2015). The AR, an established drug target in prostate cancer, was selectively degraded using an adamantyl-linked, non-covalent high-affinity AR agonist. The rationally designed SARD reduced AR levels at a concentration of 1 μM and was even effective in cell lines resistant to current standard-of-care drugs for Castration-Resistant Prostate Cancer (CRPC). While HyT technology is a promising approach for small-molecule induced target degradation there is concern that the hydrophobic moiety might be preferentially bound by serum proteins, which could reduce bioavailability. Thus far, HyT presents a valuable tool for chemical biology, e.g. allowing a detailed study of induced stress response (Raina et al., 2014) (Figure 3A).

Figure 3. Small molecule protein degradation techniques.

(A) HyTs degrade their protein target via a not fully elucidated mechanism following one of two possible pathways. (a) HyT binding destabilizes the POI which recruits a chaperone that induces POI proteasomal degradation. (b) The chaperone recognizes the HyT directly and marks the tagged POI for destruction. (B) PROTAC mode of action. The bifunctional PROTAC binds simultaneously to the POI and an E3 bringing both proteins in spatial proximity and inducing ubiquitination. The ubiquitinated POI is subsequently degraded by the proteasome releasing the PROTAC. Both approaches, HyT and PROTACs, can traverse multiple circles allowing for substochiometric usage.

Proteolysis Targeting Chimeras

HyT is based on the addition of a hydrophobic moiety to a known ligand of a POI, which upon binding triggers its degradation utilizing the endogenous protein quality control machinery. However, the potency of HyT is limited and the exact nature of degradation is thus far not fully elucidated and can occur via at least two possible modes of action (Figure 3A). In order to evolve this idea into a more efficient and controllable approach, the recruited degradation machinery needs to be more specific and its mode of action better understood. To overcome these limitations, PROTACs hijack the endogenous protein degradation machinery to specifically target and degrade a defined POI (Figure 3B). Therefore, PROTACs induce a novel and distinct PPI between the target protein and an endogenous E3. Proof-of-concept (POC) studies were first published in 2001 showing that rationally engineered molecules are capable of inducing UPS-mediated degradation of Methionine AminoPeptidase 2 (MetAP2) (Sakamoto et al., 2001). A bifunctional molecule comprised of the covalent MetAP2 inhibitor ovalicin and a ten amino acid phosphorylated peptide fragment of IκBα that binds to the F-box protein β-Transducin Repeat-Containing Protein (β-TRCP) that belongs to the S-phase Kinase-associated Protein 1 (SKP1)-CRL1-F-box E3 complex (CRL1^βTRCP) triggers MetAP2 ubiquitination and subsequent proteasomal degradation in Xenopus egg cell extracts. Applicability of this approach in HEK293 cells was demonstrated by the degradation of ERα and AR after PROTAC microinjection using the same phosphopeptide recruiting element for CRL1^βTRCP (Sakamoto et al., 2003). Despite the success of the POC studies, the effectiveness of these early PROTACs was hampered by low cell permeability, micromolar potency and sensitivity of the E3 recruiting phosphorylated degron towards endogenous phosphatases.

A first major improvement in PROTAC technology was achieved by switching from the IκBα phosphodegron (phosphatase sensitive) to an oxygen-dependent degron for E3 recognition. The Von Hippel-Lindau (VHL) protein comprises the F-box protein of the CRL2^VHL E3 and specifically recognizes a hydroxylated proline within a seven amino acid binding stretch of the Hypoxia-Inducible Factor 1α (HIF1α) (Hon et al., 2002; Min et al., 2002). Fused with an octa D-Arg stretch for cellular uptake, this peptidic VHL recruiting motif was used successfully to degrade AR- and FKBP12-GFP fusions in whole cell experiments (Schneekloth, JR et al., 2004). Furthermore, the hydroxyproline peptide sequence has been proven active in the degradation of ERα (Zhang et al., 2004; Bargagna-Mohan et al., 2005; Rodriguez-Gonzalez et al., 2008; Cyrus et al., 2010a; Cyrus et al., 2010b), the aryl hydrocarbon receptor (Lee et al., 2007; Puppala et al., 2008), the X-protein of the hepatitis B virus (Montrose and Krissansen, 2014), Akt (Henning et al., 2016), Smad3 (Wang et al., 2016) and Tau (Chu et al., 2016). The first PROTAC successfully studied in live animals was a dipeptidic phosphorylation-dependent PROTAC comprised of a VHL-engaging octa D-Arg/hydroxylproline sequence coupled to a tyrosine phosphorylation sequence of Her3 (Hines et al., 2013). An analogous phosphorylation-dependent PROTAC based on a sequence of the nerve growth factor receptor Tropomyosin receptor kinase A (TrkA) was also created in this study. These phosphorylation-dependent “PhosphoPROTACs” are administered as prodrugs and are activated following their phosphorylation by Her2 or TrkA, respectively. Upon activation, the PROTACs are recognized by their downstream targets Phosphatidylinositide 3-Kinase (PI3K) and Fibroblast growth factor Receptor Substrate 2α(FRS2α), respectively, targeting them for degradation and blocking the corresponding signaling pathways. Despite the success based on degron exchange and subsequent VHL recruitment, peptide-based PROTACs still suffer from reduced potencies in the micromolar range. This drawback is most likely based on low cell permeability and high protease susceptibly commonly associated with peptide-based therapeutics, and thus creates a need for more drug-like PROTACs.

The first small molecule PROTACs were designed utilizing the Mouse Double Minute 2 homologue (MDM2) binder nutlin-3a (Figure 4A) replacing the previously-used peptidic E3 ligands (Schneekloth et al., 2008). MDM2 is a heterodimeric RING-type E3 that targets the tumor suppressor p53 for degradation and has therefore emerged as a highly interesting target in drug discovery. Tethering nutlin-3a to an AR antagonist allowed for AR depletion via MDM2 ubiquitination and subsequent proteasomal degradation (Schneekloth et al., 2008). However, reduction of AR levels was only observed at micromolar concentration and overall AR depletion was reduced compared to VHL induced degradation, suggesting MDM2 to be a less promising E3 for PROTAC applications. The second generation of small molecule PROTACs, also known as Specific Non-genetic Inhibitor-of-Apoptosis-Proteins (IAPs)-dependent Protein ERasers (SNIPERs), recruit the homodimeric RING-type E3 cellular IAP1 (cIAP1) using the small molecule ligand bestatin for POI degradation (Itoh et al., 2010). However, the utility of SNIPER-based protein degradation is limited due to off-target binding of bestatin (Umezawa et al., 1976; Orning et al., 1991) and the induction of autoubiquitination and subsequent degradation of the cIAP1 E3 itself (Sekine et al., 2008). Ligand optimization of the cIAP1 recruiting moiety bestatin has reduced cIAP1 autoubiquitination (Itoh et al., 2011a), but their efficacy remained in the micromolar range. Thus far, SNIPERs have successfully degraded multiple cellular targets, e.g., Cellular Retinoic Acid-Binding Protein I (CRABPI) and CRABPII (Itoh et al., 2010; Itoh et al., 2012; Okuhira et al., 2017), ERα (Itoh et al., 2011b; Demizu et al., 2012; Okuhira et al., 2013), the spindle regulatory protein Transforming Acidic Coiled-Coil-3 (TACC3) (Ohoka et al., 2014), the Breakpoint Cluster Region-Abelson tyrosine kinase (BCR-ABL) (Demizu et al., 2016) and various HaloTag fusion proteins (Tomoshige et al., 2016). Recently, the introduction of the IAP antagonist LCL161 has boosted SNIPER activity to the low nanomolar range, and allowed SNIPER evaluation in mouse xenograft models for the first time (Ohoka et al., 2017). Nevertheless, simultaneous IAP degradation and incomplete target degradation remain issues of this technology and further studies are necessary to evaluate its therapeutic potential.

Figure 4. Structural elucidation of E3 ligand binding.

(A) Crystal structure of nutlin 3a bound to its target protein MDM2 (PDB: 4HG7). (B) The VHL ligand 1 provides the essential hydroxyproline to engage the F-box protein VHL (PDB: 4W9H). (C) Pomalidomide bound to its molecular target the F-box protein CRBN (PDB: 4TZU). Sites of linker attachment are highlighted by a black arrow.

Since the most efficient target degradation up to that point had been observed using CRL2^VHL-recruiting PROTACs, strong efforts were invested in identifying a small molecule ligand to replace the hydroxyproline HIFα peptide fragment. A major breakthrough was achieved by the development of the peptidomimetic VHL ligand 1, which binds to VHL with submicromolar affinity (K_d = 185 nM, Figure 4B) (Buckley et al., 2012a; Buckley et al., 2012b; van Molle et al., 2012; Galdeano et al., 2014). VHL ligand 1 was identified pursuing a combination of in silico and fragment-based screening, guided by various crystal structures of intermediates bound to the substrate recognition domain of VHL. The R-hydroxyproline core of VHL ligand 1 is essential for VHL engagement and inversion of this stereo center abolishes binding completely. The potency of VHL ligand 1 as an efficient VHL recruiting element for PROTACs was proven by the degradation of HaloTag-GFP fusion proteins (Buckley et al., 2015). Cell treatment with various VHL ligand 1-based haloPROTACs resulted in dose-dependent GFP depletion with a half maximal degradation concentration (DC₅₀) of 19 nM and a maximum degradation efficacy (D_max) of 90%. Another milestone for this technology was achieved by conclusively proving the ability of PROTACs to function catalytically (Bondeson et al., 2015). Substochiometric amounts of a VHL based Receptor-Interacting serine/threonine Protein Kinase2 (RIP2K) PROTAC were sufficient for RIP2K ubiquitination. Displaying a DC₅₀ of ~100 nM and a D_max of 86%, an analogously designed PROTAC for the Estrogen-Related Receptor α(ERRα) proved active in vivo and showed approximately 40% target degradation in mouse xenograft models.

The molecular toolbox for the synthesis of highly potent PROTACs has recently benefitted from the addition of yet another E3 recruiting small molecule. Based on the ability of phthalimide ImunoModulatory Drugs (IMiDs) to bind the CRL4^CRBN E3 complex and recruit the transcription factors Ikaros and Aiolos, as well as the Casein Kinase 1α(CK1α) for ubiquitination and subsequent degradation ((Ito et al., 2010; Chamberlain et al., 2014; Fischer et al., 2014; Lu et al., 2014; Kronke et al., 2015), highly efficient PROTACs for Bromodomains and Extra Terminal domain (BET) proteins have been reported that use IMiDs as CRBN recruiting elements (Lu et al., 2015; Winter et al., 2015). Connecting the bromodomain-containing protein 4 (BRD4) inhibitor OTX015 with CRBN recruiting pomalidomide resulted in a highly potent degrader (ARV-825) of the epigenetic regulator BRD4 (Figure 4C, Figure 5)(Lu et al., 2015). The resulting PROTAC ARV-825 reduced BRD4 levels in various Burkitt’s lymphoma cell lines even at picomolar concentration and provided significantly enhanced c-MYC suppression and increased apoptosis when compared to the parent inhibitor. Follow-up studies confirmed superior efficacy compared to OTX015 in Acute Myeloid Leukemia (AML) xenografts (Saenz et al., 2017). Analogous BET-targeting PROTACs were designed based on 4-hydroxy thalidomide and the BRD inhibitor JQ1 (Winter et al., 2015). The most efficient BRD PROTAC, dBET1, reduced the protein levels of BRD2, BRD3 and BRD4 with low nanomolar potency and outperformed JQ1 in inducing apoptosis in AML and lymphoma cell lines (Figure 5). Furthermore, this PROTAC is effective in mouse xenograft models of human leukemia. Tethering the pan-BET inhibitor JQ1 to VHL ligand 1 yielded the BET PROTAC ARV-771, which is highly active against cellular models of CRPC at picomolar concentrations (Figure 5) (Raina et al., 2016). This potency translates into mouse models of CRPC leading to tumor regression of enzalutamide-resistant 22Rv1 xenografts and advancing the scope of PROTAC technology to solid tumors. Superior efficacy compared to BET inhibition was also observed for a pomalidomide based PROTAC comprising a second generation BET inhibitor which induces growth inhibition and apoptosis in triple-negative breast cancers (TNBC) (Bai et al., 2017). Tethering this second generation BET inhibitor to lenalidomide resulted in PROTACs degrading BET proteins at picomolar concentration which caused >90% tumor regression in acute leukemia mouse xenograft models (Zhou et al., 2017). These studies highlight the therapeutic advantage of event-driven pharmacology given that BRD4 inhibition is often counteracted by compensatory BRD4 overexpression to hamper inhibitor efficacy (Shimamura et al., 2013). Furthermore, PROTAC technology has successfully been used for target validation (Erb et al., 2017). Using a 4-hydroxy thalidomide based FKB12 degrader and a FKBP12–POI fusion construct the ENL YEATS domain was identified as a novel target in acute leukemia.

Figure 5.

Small molecule PROTACs targeting BET proteins.

Interestingly, tethering JQ1 with VHL ligand 1 using a different linker resulted in PROTAC MZ1 that favors the degradation of BRD4 over BRD2 and BRD3 (Figure 5) (Zengerle et al., 2015). While BRD4 was depleted at low nanomolar concentrations of MZ1, micromolar concentrations were required to degrade BRD2 and BRD3. A crystal structure of the ternary complex between VHL, BRD4 and MZ1 revealed that the PROTAC folds into itself when it binds into a cavity formed between BRD4 and VHL (Figure 6A) (Gadd et al., 2017). This study reports the first crystal structure of a PROTAC in complex with its target proteins and allows groundbreaking insight into PROTAC mode of action. As evident in the crystal structure, BRD4 and VHL form extensive PPIs upon PROTAC binding resulting in high stability and cooperativity of the ternary complex (Figure 6B). ITC measurements as well as mutation studies confirmed the isoform-specific and highly cooperative nature of the ternary complex. Based on these results, the rationally designed PROTAC AT1 was identified as a highly selective degrader of BRD4 showing negligible activity against BRD2 and BRD3 (Figure 6C). Likewise, in another study different selectivity profiles were obtained for VHL- and CRBN-based PROTACs that target BCR-ABL and ABL using the tyrosine kinase inhibitors bosutinib and dasatinib (Lai et al., 2016). Depending on the warhead and the E3 recruiting ligand incorporated into these PROTACs, preferences in target degradation were observed which were independent of kinase binding. As a result, varying the warhead (if applicable) as well as the E3 recruiting moiety constitutes an important task for PROTAC development as it can have a dramatic impact of PROTAC potency and selectivity.

Figure 6. Structural insight into PROTAC mode of action.

(A) Binding interface between BRD4 (dark grey) and VHL (grey) with MZ1 (light orange) embedded in a bowl-shaped cavity formed by both proteins. The warhead JQ1 as well as E3 recruiting VHL ligand 1 form contacts with both proteins BRD4 and VHL, respectively. (B) Upon formation of the ternary complex BRD4 and VHL engage in extensive PPIs resulting in high cooperativity. (C) Rationally designed BRD4 selective PROTAC AT1. (PDB: 3T35).

Conclusion

Controlling protein function by controlling cellular protein levels represents a promising new therapeutic approach in modern drug discovery. These event-driven strategies offer a novel pharmaceutical modality compared to traditional inhibitors. In contrast to traditional small molecule therapeutics that function via an occupancy-driven paradigm and struggle in many cases to achieve and maintain the high local drug concentrations essential for effective POI inhibition, target protein degraders such as PROTACs act in an event-driven manner and remove the POI from the system. Their mode of action is not dependent on blocking any function – rather their purpose is solely to mediate a transient interaction with an E3 and allow for proximity induced ubiquitination. As such, protein degraders can expand the drug target space that limits current medicinal research. Since PROTAC efficacy is based on target engagement, even agonists can function as recruiting moieties and successfully prevent downstream signaling. As demonstrated for BRD4, inhibition can sometimes result in target upregulation feedback thus requiring higher inhibitor dosing. This “pharmacological arms race” is lost at the point where the increased inhibitor concentrations needed to maintain maximal target engagement can no longer be tolerated. In contrast, by degrading BRD4, PROTACs remain active even at low concentrations and provide a prime example of differential biology highlighting the difference between target inhibition and target degradation (Lu et al., 2015; Winter et al., 2015; Raina et al., 2016). Target degradation might also be effective in overcoming other feedback mechanisms, but further studies are needed to elucidate any potential resistance responses to PROTAC treatment. As transient target engagement is adequate for POI degradation, it is likely that PROTACs are less vulnerable to mutations that reduce binding affinity of the warhead. However, depletion of the recruited E3 or efflux mechanisms are likely to hamper PROTAC efficacy.

Studies have shown that promiscuous inhibitors can be converted into more selective degraders upon incorporation into PROTAC molecules and that different combinations with various E3s can result in unexpected selectivity profiles (Zengerle et al., 2015; Lai et al., 2016). Based on the first crystal structure of a PROTAC in a ternary complex, a highly selective degrader of BRD4 was designed using a promiscuous warhead (Gadd et al., 2017). Therefore, it is important to test a variety of warheads (if applicable), linkers and E3s when optimizing a specific PROTAC. In particular, the linker connecting the two functional moieties has a significant influence on PROTAC activity (Cyrus et al., 2011; Buckley et al., 2015). Its composition impacts cell permeability and an optimized linker length is essential for efficient target ubiquitination. Intracellular PROTAC assembly might offer some advantages in this respect, but a more drug-like character of the components rand higher activities are needed to pursue this idea further (Lebraud et al., 2016).

The modular assembly of PROTACs allows for a “plug-and-play” approach to this technology where any identified ligand can be linked to an E3 recruiting ligand. Importantly, the linker attachment point must to be identified. However, ligand SAR studies or X-ray analysis of the binding mode usually allows for the identification of suitable positions. The advantages arising from the event-driven mode of action open new possibilities for the resurrection of “near drugs”; compounds that failed in clinical trials due to the inability to achieve/maintain sufficiently high in vivo systemic concentrations are perfect candidates for PROTAC development. These compounds are already optimized binders, having been successful in preclinical development and might serve as excellent recruiting elements in successful PROTACs since this approach does not need the excess concentrations required for inhibitor-based strategies. The efficacies of PROTAC-induced protein degradation observed in tissue culture and tumor xenografts are highly encouraging and clinical trials for PROTACs are planned for the near future.

Although PROTACs might revive some shelved small molecule inhibitor-based drug candidates, the main potential of this technology is in its application to the expansion of the druggable proteome. Together with other upcoming techniques advanced within the last decades, such as next generation peptides (Craik et al., 2013; Cromm et al., 2015a) and peptidomimetics (Pelay-Gimeno et al., 2015), PROTACs are part of the “new modalities” in drug discovery (Waldmann et al., 2017). All of these techniques attempt to target the undruggable proteome and it will be interesting to see whether oncogenic drivers like c-Myc or K-Ras (Spiegel et al., 2014; Cromm et al., 2015b; Papke and Der, 2017), transcription factors and pseudokinases can finally be dragged into the druggable space. The major challenge will be identifying suitable ligands for these undruggable proteins. However, since binding to the POI at any site is sufficient for PROTAC–target interaction, there are more chances for POI modulation by PROTACs than there would be for traditional inhibitors. Thus, PROTACs have the potential to play a leading role in addressing the undruggable proteome, since they combine the drug-like properties of conventional small molecules with the event-driven benefits of nucleotide-based strategies. Looking ahead, the PROTAC technology might hold the key for developing catalytic small molecule drugs that work like enzymes.