From (Tool)Bench to Bedside: The Potential of Necroptosis Inhibitors

Abstract

Necroptosis is a regulated caspase-independent form of necrotic cell death that results in an inflammatory phenotype. This process contributes profoundly to the pathophysiology of numerous neurodegenerative, cardiovascular, infectious, malignant, and inflammatory diseases. Receptor-interacting protein kinase 1 (RIPK1), RIPK3, and the mixed lineage kinase domain-like protein (MLKL) pseudokinase have been identified as the key components of necroptosis signaling and are the most promising targets for therapeutic intervention. Here, we review recent developments in the field of small-molecule inhibitors of necroptosis signaling, provide guidelines for their use as chemical probes to study necroptosis, and assess the therapeutic challenges and opportunities of such inhibitors in the treatment of a range of clinical indications.

1. Introduction

Cell death processes are fundamentally important to organism development and homeostasis. In addition to apoptosis, this century has seen the rise to prominence of several forms of regulated cell death, such as pyroptosis, ferroptosis, and necroptosis.¹ Morphologically, necroptosis involves cellular swelling, the rupture of the plasma membrane, chromatin condensation, cellular lysis, and the loss of intracellular contents.² Necroptosis can be induced by diverse stimuli including inflammatory markers such as tumor necrosis factor (TNF) receptor 1 (TNFR1), factor-associated suicide receptor (Fas), toll-like receptors 3/4 (TLR3/4),^3,4 nucleotide binding and oligomerization domain (NOD)-like receptors (NLRs), interferon-α/β receptor subunit 1 (IFNAR1), and Z-DNA binding protein 1 (ZBP1).⁵ In the absence of pro-survival signals (via the nuclear factor κB (NF-κB) pathway) and caspase activity, a complex, termed the necrosome, is formed between receptor-interacting serine/threonine-protein kinase (RIPK) 1, RIPK3 and the mixed lineage kinase domain-like protein (MLKL) pseudokinase. Ultimately, RIPK3-mediated phosphorylation of MLKL leads to a conformational change,⁶⁻⁸ oligomerization,^9,10 translocation to the plasma membrane, and membrane rupture,⁹⁻¹² resulting in the release of damage-associated molecular patterns (DAMPs),¹³ inflammation, and organ injury.¹⁴

Animal models of disease and genetic experiments have demonstrated that necroptosis prominently contributes to a variety of disease pathologies, including neurodegenerative,¹⁵⁻²¹ cardiovascular,^22,23 infectious,²⁴⁻³¹ malignant,³²⁻³⁶ and inflammatory diseases.^35,37−47 Furthermore, extensive studies have demonstrated that inhibiting the expression or activity of RIPK1, RIPK3, and MLKL is of therapeutic relevance. The identification of small-molecule inhibitors of necroptosis and their use as chemical biology tools has been essential to these discoveries, and their continued development underpins the efforts of researchers and clinicians in both understanding and treating necroptosis-related diseases. Throughout this Perspective, we aim to arm the reader with information vital to selecting appropriate tool compounds for inhibiting necroptosis in vitro and in vivo. For the purpose of this Perspective, in vivo animal models of necroptotic disease generally refers to rodent models easily accessible to most researchers. This distinction is important, as many inhibitors of necroptosis proteins display (in some cases) very high levels of specificity toward primate, in preference to rodent, orthologues of these proteins.

2. Molecular Functions of Necroptosis Regulator Proteins

Among the various necroptosis signaling pathways (Figure 1), the mechanisms downstream of TNFR stimulation have been best characterized. Upon the binding of TNF to TNFR1, cells are directed to either proinflammatory gene induction or the execution of cell death. Activated TNFR1 undergoes a conformational change and forms a complex, termed complex I, together with TNFR-associated death domain (TRADD), TNFR-associated factor 2 (TRAF2), RIPK1, cellular inhibitors of apoptosis 1 or 2 (cIAP1/2) and linear ubiquitin chain assembly complex (LUBAC).⁴⁸ The ubiquitination of RIPK1 and other subunits in complex I by the E3 ubiquitin ligases cIAP1/2 and LUBAC leads to the subsequent recruitment of the transforming growth factor-β-activated kinase 1 (TAK1) complex and the inhibitor of nuclear factor κB kinase (IKK) complex to induce NF-κB and MAPK signaling.^49,50 In scenarios where the activity of cIAP1/2 is compromised, TNF signaling is shifted to the induction of cell death via apoptosis through the formation of complex IIa (or ripoptosome) that comprises RIPK1, TRADD, Fas-associated protein with death domain (FADD), and caspase 8.^48,51 When caspase activity is limited, cell death is redirected through necroptosis via the receptor interacting protein homotypic interaction motif (RHIM) domain-mediated association of RIPK3 to RIPK1 and formation of complex IIb in the cytoplasm.^24,52 The RHIM domain-mediated homotypic interaction plays a crucial role in assembling subunit proteins into a higher-molecular-weight necrosome complex.⁵³ Phosphorylated RIPK3 subsequently induces MLKL activation, which is represented by the phosphorylation of its pseudokinase domain, oligomerization, and translocation to the plasma membrane, where MLKL ultimately executes necroptotic cell death^6,9−11,54 via an incompletely understood mechanism.^55,56 Conversely, necroptosis originating from TLR3 or TLR4 stimulation proceeds through binding of toll/interleukin-1 receptor/resistance domain-containing adaptor-inducing interferon-β (TRIF), followed by the recruitment of RIPK1/RIPK3,^3,4 while ZBP1, which detects cytosolic nucleotides, encodes a RHIM domain itself that allows for its direct interaction with RIPK1/RIPK3.⁵

Figure 1.

Signal transduction in necroptosis. Various pathways, including signals from death receptors, infection, inflammatory stimuli, and other cellular stresses, activate the necroptotic cell-death machinery.

As the core subunits of the necrosome (for further information on this concept, see Samson et al.¹⁴ and Horne et al.⁵⁷), RIPK1, RIPK3, and MLKL are the most attractive targets of pharmacological inhibitors that aim to treat diseases involving necroptosis. RIPK1 is positioned at the intersection of cellular fates and plays a central role in directing the cell toward survival, or cell death via apoptosis or necroptosis. The role of RIPK1 in pro-survival signaling and the induction of inflammatory gene production is dependent on its scaffolding function. This role was demonstrated by studies in mice where RIPK1 knockout resulted in embryonic lethality,⁵⁸ which was rescued by the expression of a kinase-dead RIPK1 mutant.^59,60 Furthermore, cells expressing kinase-dead RIPK1 are capable of activating NF-κB signaling comparably to wild type cells.^58,61 The kinase activity of RIPK1, which can induce autophosphorylation and the subsequently presumed conformational change that is crucial for caspase activation and necrosome formation, is required for the execution of several cell death modalities, including apoptosis, pyroptosis, and necroptosis.^24,48,62

RIPK3 is required for necroptosis mainly in a RHIM domain- and kinase activity-dependent manner.^60,63−65 RIPK3 has also been linked to other cellular pathways, including apoptosis, activation of the inflammasome, and induction of proinflammatory genes.^60,66,67 It has been proposed that specific modifications in the kinase domain of RIPK3 can result in a pro-apoptotic conformation that leads to the formation of a death complex through RHIM domain-mediated interactions.^60,68,69 The role of RIPK3 in promoting cytokine production has been reported in different diseases and genetic models;^67,70,71 however, the underlying mechanisms are not fully understood.

MLKL is currently recognized as the terminal effector protein in necroptosis, although the mechanism of how MLKL is regulated has only been partially elucidated. Critically, the killer ability of MLKL is activated through the phosphorylation of its pseudokinase domain by RIPK3 and subsequent de-repression of its N-terminal executioner domain.^6,9 Other mechanisms, such as those involving the ESCRT-III machinery, TAM kinases, inositol phosphate kinases, and trafficking via Golgi, microtubules, and actin, have also been reported to regulate MLKL in mediating necroptosis.^10,72−74 The details of the post-translational modifications that control necroptosis signaling and other functions of the necroptotic proteins have been comprehensively reviewed elsewhere.⁷⁵ The involvement of MLKL in other signaling pathways, such as inflammasome activation and autophagy, has been proposed but incompletely explored.⁷⁶

In studies of necroptosis and its inhibition, cells are directed through this cell death pathway by the administration of a cocktail of reagents. The composition of this cocktail is dependent on the cell line being used (i.e., whether or not the cell line is intrinsically deficient in caspase activation), the disease model being studied, and which effector (RIPK1, RIPK3, or MLKL) one is interested in studying or inhibiting. Such cocktails typically comprise one to three ingredients: a cell-death initiator, a second mitochondria-derived activator of caspases (SMAC) mimetic, and a caspase inhibitor. While necroptosis is best understood in the context of TNF activation, lipopolysaccharides (LPS), polyinosinic:polycytidylic acid (poly(I:C)), factor-associated suicide receptor (FasL), and interferon are also routinely used. SMAC mimetics are added to inhibit IAPs, thereby excluding RIPK1 from participating in prosurvival signaling and directing cellular fate toward death. Caspase inhibitors prevent cell death via apoptosis and redirect cell death via necroptosis. As with the choice of the cell death stimulus, it is imperative to carefully consider which caspase inhibitor is most appropriate for the system being studied. The most common caspase inhibitors used in studying necroptosis are zVAD.fmk (irreversible pan-caspase inhibitor),⁷⁷ QVD-OPh (irreversible inhibitor of caspase 1, 3, 8, and 9),⁷⁸ and IDN-6556 (emricasan, irreversible pan-caspase inhibitor).⁷⁹ Aside from their differing potency against caspase 8 (the mediating caspase of necroptosis),⁸⁰⁻⁸² it should also be noted that zVAD.fmk is capable of inhibiting other proteases at concentrations that are generally used in necroptosis assays.⁸³

3. Role of Necroptosis in Disease Pathogenesis

Necroptosis has been implicated in the pathogenesis of many diseases spanning across broad disease areas, including inflammation,^35,37−47 oncology,³²⁻³⁶ central nervous system (CNS)/neurodegeneration,¹⁵⁻²¹ and diabetes,⁸⁴ as well as cardiovascular,^22,23 kidney,^43,85,86 liver,^40,87−89 and infectious diseases.²⁴⁻³¹ A range of experimental methods have been used to build evidence for the link between necroptotic proteins RIPK1/RIPK3/MLKL and disease pathogenesis. These methods include genetic experiments (knockout mice, shRNA or siRNA knockdown, and CRISPR Cas9 technology), pharmacological inhibition of RIPK1/RIPK3/MLKL proteins, protein/gene expression analysis of patient tissue samples, and in vivo animal disease models. A comprehensive analysis is outside the scope of this Perspective, and the reader is directed to key references and thorough reviews for specific disease classes. Additionally, an excellent review by Jouan-Lanhouet et al. summarizes tools for the in vivo detection of necroptosis in experimental disease models,⁹⁰ and further information can be found in Samson et al.⁹¹ and Horne et al.⁵⁷

It should be noted that although many studies linking necroptosis to a particular disease pathology have been reported, there is currently significant debate as to the validity of some of these studies (see examples in the context of liver disease,⁹² amyotrophic lateral sclerosis (ALS),^93,94 colitis,⁹⁵ nonalcoholic steatohepatitis (NASH),^96,97 and pancreatic cancer³⁵). A comprehensive investigation of several different mouse disease models by Newton et al., found that RIPK3 was dispensable in models of sepsis, colitis, pancreatitis, cerebral edema, and cerebral artery occlusion stroke.⁹⁸ The authors hypothesized that RIPK1 and RIPK3 exhibit additional roles beyond the regulation of MLKL-dependent necroptosis, such as the promotion of apoptosis in certain contexts. Additionally, there is a possibility of RIPK3 loss leading to the suppression of pro-inflammatory cytokines and chemokines.⁹⁹ Finally, information on murine models investigating the role of MLKL in diseases has been published.¹⁰⁰ Taken together, these findings demonstrate the evolving nature of the field and should be carefully considered by researchers looking at developing novel therapeutics targeting necroptosis.

4. Development of Necroptosis Inhibitors

Since necroptosis was established as a regulated cell death pathway, the development of small-molecule inhibitors of necroptosis has played a critical role in delineating its underlying choreography and identifying the key proteins involved. Through a combination of classical screening approaches and cutting-edge technologies, we are now at a point where researchers have access to both tool compounds for in vitro and in vivo experiments (with significant species selectivity) and clinical candidates with the potential to offer breakthrough therapies for debilitating diseases.

4.1. Inhibitors of RIPK1

RIPK1 plays a decisive role in directing the cellular response toward either apoptosis or necroptosis. Several factors likely contributed to inhibitors of RIPK1 being the first inhibitors of necroptosis discovered. First, TNF is the most characterized and widely used stimulus to experimentally induce necroptosis, and this pathway operates through a RIPK1-dependent mechanism. Second, RIPK1 is a kinase, a class of proteins that is considered highly druggable due to the presence of a small-molecule binding site (ATP pocket) and an inhibitable catalytic function (phosphate transfer). Consequently, inhibitors of RIPK1 have been used to delineate the necroptosis pathway, identify the key effector proteins, and demonstrate the therapeutic potential of inhibiting necroptosis.

4.1.1. Necrostatins

Necrostatins are the prototypical inhibitors of necroptosis and have been indispensable for developing our current understanding of necroptosis biology. The necrostatins were identified in a phenotypic screen of a 15 000 compound library against human monocytic U937 cells treated with TNF and the pan-caspase inhibitor zVAD.fmk.¹ This effort identified several diverse structural classes of compounds (necrostatin 1, 3, 4, 5, and 7), with the most used of these classes being the necrostatin 1 (Nec-1, 1) family (Figure 2). As none of the other necrostatin classes have achieved the potency^101−105 and subsequent widespread use of the Nec-1 class, they shall not be discussed further in this Perspective, other than to note that Nec-7 contains a known PAINS motif and its use should be avoided.¹⁰⁶ Investigation of the structure–activity relationship (SAR) around the Nec-1 compound class (>200 analogues) in FADD-deficient Jurkat T cells revealed a limited scope for development, with most modifications being detrimental to activity.¹⁰⁷ Of the analogues developed, the inactive Nec-1i (2; EC₅₀ > 10 μM) and improved Nec-1s (3; EC₅₀ = 50 nM) analogues are of note. Nec-1i was originally proposed as a suitable inactive control for experiments utilizing Nec-1; however, Nec-1i has also been shown to retain inhibitory activity against RIPK1 when used at higher concentrations in vivo (>10 μM), thereby limiting its utility.¹⁰⁸ Furthermore, despite being remarkably selective for RIPK1 over other kinases,¹⁰⁹ Nec-1 is only modestly potent (EC₅₀ = 490 nM), unstable in vivo (t_1/2 < 5 min in mouse liver microsomes (MLM)), and toxic at high concentrations.¹⁰⁷ The use of Nec-1 and Nec-1i is further complicated by the observation that both are inhibitors of the immune regulator indoleamine 2,3-dioxygenase (IDO).¹¹⁰ This might confound the interpretation of experimental findings in inflammatory disease settings due to the high concentrations (typically tens of μM) at which Nec-1 is administered in vivo. In comparison to Nec-1, Nec-1s is not only more potent, more stable in vivo (t_1/2 ∼ 60 min in MLM), and less toxic¹⁰⁷ but is also not an inhibitor of IDO.¹⁰⁸ However, the determination of the pharmacokinetic (PK) properties of Nec-1s (performed on the racemate) revealed its low exposure (AUC_8h = 0.27 μg h mL^–1) and high clearance (61 mL min^–1 kg^–1).¹⁰⁷ Despite these limitations on its in vivo use, Nec-1s remains an invaluable tool for cellular studies.

Figure 2.

Chemical structures of the Nec-1 class of necrostatins.

Nec-1 and its derivatives were shown to inhibit RIPK1 in an ATP-competitive manner, thereby preventing its catalytic activity and rescuing cells from necroptosis.¹¹¹ Through mutagenesis studies, it was established that autophosphorylation of the activation loop abrogated the inhibitory activity of Nec-1 analogues, suggesting a critical role for this segment in their mechanism of action, and that Nec-1 compounds bound to and stabilized an inactive DLG-out conformation of RIPK1.¹¹¹ Subsequent disclosure of an X-ray crystal structure with Nec-1s bound to the kinase domain (residues 1–294, C34A, C127A, C233A, and C240A; PDB 4ITH) of RIPK1 revealed that the binding site was in fact a relatively hydrophobic allosteric pocket between the N-lobe and the C-lobe, designating Nec-1 compounds as type III allosteric inhibitors of RIPK1 (Figure 3).¹¹² Binding of Nec-1s locks RIPK1 in an inactive conformation wherein the αC helix is displaced ∼40° relative to the catalytic subunit of protein kinase A (PKA; PDB 2CPK),¹¹³ with the vacated space being partially occupied by Nec-1s and the activation loop helix. Consequently, the salt bridge between E63 of the αC helix and K45 of the β3 strand (equivalent to E91–K72 in PKA, Figure 3B) that is required for the stabilization of ATP binding is also lost, with an E63–K45 separation of ∼15 Å. The inactive state of Nec-1s-bound RIPK1 is further exemplified through the characteristic “DLG-out” conformation of this motif, wherein D156 and S161 of the DLG motif interact with the Nec-1s inhibitor through two H-bonds. Finally, the hydrophobic regulatory (R)-spine typical of an active kinase conformation, which is formed by linear stacking of L78, M67, L157, and H136 in RIPK1, is disrupted, with M67 and L157 oriented away from the spine (Figure 3B). This allosteric binding mode accounts for the RIPK1 selectivity displayed by the Nec-1 class even against family members of the highest homology, namely, RIPK2 and RIPK3,¹¹¹ due to the known benefits of type III inhibitors (e.g., targeting less-conserved regions).¹¹⁴ The observed SAR of the Nec-1 class could also be rationalized through the hydrophobic nature of the pocket, the H-bond networks formed, and the proximity of Nec-1s to the residues within the narrow binding pocket. This X-ray structure represented a breakthrough in the field of RIPK1 inhibitor development, as it enabled pursuit of RIPK1 inhibitor structure-based design for the first time.¹¹⁵

Figure 3.

Selected crystal structures of RIPK1 with small-molecule inhibitors. (A) The apo human PKA in the active conformation (yellow; PDB 2CPK(113)) compared to human RIPK1 bound to Nec-1s, with the compound shown in surface view (green; PDB 4ITH(112)). The Nec-1s structure is representative of the inactive conformation of human RIPK1 bound to a type III inhibitor. (B) A comparison of the compound binding site of human RIPK1 and the equivalent site in active-conformation PKA, highlighting the regulatory spine rearrangement and the DFG motif F185 (PKA) or DLG L157 (human RIPK1) position. Compounds are shown in stick format.

4.1.2. Dihydropyrazoles

In 2015, GSK identified a dihydropyrazole (DHP) series of RIPK1 inhibitors by screening a 2 million compound library using an ADP-Glo enzymatic assay.¹¹⁶ Hybridization of the two screening hits and subsequent chiral resolution demonstrated the enantiospecific nature of their activity, with the (S)-enantiomer (GSK′963, 4) being highly active in a fluorescence polarization (FP) binding assay (IC₅₀ = 29 nM) and the (R)-enantiomer (GSK′962, 5) being inactive at 10 μM (Figure 4).¹¹⁶ GSK′963 was also shown to be highly specific for RIPK1 when assessed against a panel of 339 kinases at 10 μM (S₍₅₀₎ = 0.00, S₍₂₀₎ = 0.01) and did not inhibit IDO.¹¹⁶ Furthermore, GSK′963 is efficacious in murine cells (IC₅₀ = 1 and 3 nM against L929 and BMDM, respectively, both stimulated with TNF and zVAD.fmk) and is able to protect mice from TNF+zVAD.fmk-induced lethal shock (at 2 mg kg^–1 and 0.2 mg kg^–1 i.p.) in vivo.¹¹⁶ The exquisite kinase selectivity and nanomolar potency of GSK′963 make this inhibitor, coupled with its inactive isomer GSK′962, an excellent tool compound for evaluating necroptosis biology in vitro. However, due to its poor oral exposure (undetectable in rats) and short half-life (t_1/2 = 3.5 and 20 min in rat and human microsomes, respectively), its use should be limited to acute models of necroptotic disease.¹¹⁷

Figure 4.

Chemical structures of the DHP class of inhibitors.

A lead optimization campaign was later undertaken around these DHPs to improve the pharmacokinetic (PK) profile and generate in vivo tools. This led to the development of two improved inhibitors: 6 (GSK′547) and 7 (Figure 4).¹¹⁷ GSK′547 is a potential tool for in vivo studies in murine models of chronic disease due to improvements in potency (IC₅₀ = 32 nM in L929 cells stimulated with TNF+QVD-OPh) and stability (t_1/2 = 69 and 198 min in rat and human hepatocytes, respectively) and moderate PK properties in rats (i.v. and p.o.) relative to other analogues developed in the study.¹¹⁷ This was exemplified in an experimental autoimmune encephalomyelitis (EAE) mouse model of human multiple sclerosis and a Rd10 mouse model of human retinitis pigmentosa (RP). A delay in disease onset and reduced clinical severity were demonstrated in the EAE model, and protection of both retinal cell function and survival was observed in the Rd10 model.¹¹⁷

Importantly, an X-ray crystal structure of 7 bound to the RIPK1 kinase domain was obtained (PDB 6R5F) and showed that 7 occupies the allosteric region in the back of the ATP pocket (Figure 3B).¹¹⁷ Serendipitously, these inhibitors were found to occupy the same allosteric pocket as the Nec-1 series of compounds, with the overall structure of the RIPK1 protein being near indistinguishable from the Nec-1s-bound structure. The RIPK1–compound 7 structure enabled a rational explanation of the SAR observed for this chemical class. First, the ATP-competitive binding of 7 was supported by the orientation of the piperidine–oxadiazole portion toward the ATP pocket and the partial occupancy of the space where the ATP phosphates would be positioned. Second, the chair conformation of the piperidine complements the geometry of the narrow, hydrophobic binding pocket and orients the polar oxadiazole headgroup toward the solvent front.

4.1.3. Benzoxazepinones

The benzoxazepinone series of RIPK1 inhibitors (Figure 5) was identified through a DNA-encoded library (DEL) screen against the RIPK1 kinase domain (residues 1–375).¹¹⁸ An optimization campaign yielded the lead benzoxazepinone 8 (GSK′481), which was active in RIPK1 FP and ADP-Glo biochemical assays (IC₅₀ = 10 and 1.6 nM, respectively), and a U937 TNF+zVAD.fmk cellular assay (IC₅₀ = 10 nM).¹¹⁸ GSK′481 also demonstrated complete kinome selectivity against two separate kinase panels (318 and 456 kinases, respectively) at 10 μM.¹¹⁸ However, the pronounced species selectivity of GSK′481 toward primate RIPK1 over rodent RIPK1 results in significantly reduced biochemical potency (200- and 329-fold higher IC₅₀ values for mouse and rat, respectively, vs that for humans in the RIPK1 FP assay), and cellular efficacy in mouse L929 cells (IC₅₀ = 3.2 μM),¹¹⁸ rendering GSK′481 useful only as an in vitro tool in human cells. This is further reinforced by its pharmacokinetic (PK) profile in rats, where GSK′481 exhibits low oral exposure (AUC_0–∞ = 0.38 μg h mL^–1 at 2 mg kg^–1), high clearance (69 mL min^–1 kg^–1), and a high volume of distribution (8.5 L kg^–1).¹¹⁹

Figure 5.

Chemical structures of the benzoxazepinones developed by GSK.

Extensive SAR exploration around the benzoxazepinone scaffold culminated in the two clinical candidates 9 (GSK′772) and 10 (GSK′095) (Figure 5).^119,120 Compared to GSK′481, GSK′772 exhibited improved efficacy in the ADP-Glo biochemical and U937 cellular assays (IC₅₀ = 1.0 and 6.3 nM, respectively) while maintaining complete kinase selectivity toward RIPK1.¹²⁰ When tested for activity in a Tox panel, GSK′772 displayed a generally clean profile with only weak activity against CYP2C9 (IC₅₀ = 25 μM), weak inhibition of hERG in HEK293 cells (estimated IC₅₀ of 195 μM), and weak activation of the human Pregnane X receptor (hPXR, EC₅₀ = 13 μM).¹²⁰ GSK′772 also displayed pronounced species selectivity toward primate RIPK1 (156- and 125-fold vs mouse and rat, respectively), resulting in significantly reduced cellular efficacy in murine L929 cells treated with TNF and the pan-caspase inhibitor QVD-OPh (IC₅₀ = 3.2 μM).¹²⁰ However, GSK′772 showed favorable PK properties in rats and monkeys, with moderate exposure (AUC_0–∞ = 2.3 μg·h mL^–1 at 2.1 mg/kg and 2.9 μg·h mL^–1 at 1.9 mg/kg, respectively), low clearance (17 and 10 mL min^–1 kg^–1, respectively), a low volume of distribution (2.7 and 2.2 L/kg, respectively), good stability (t_1/2 = 3.9 and 6.5 h, respectively), and ample safety windows over a one month safety assessment.¹²⁰ Furthermore, GSK′772 was efficacious at inhibiting necroptosis (stimulated with TNF, a SMAC mimetic, and either QVD-OPh or zVAD.fmk) in primary human neutrophils, human whole blood, ulcerative colitis explant tissue, and TNF- or TNF and zVAD.fmk-induced mouse models of lethal shock despite the significant reduction in activity observed upon testing in murine cells vs human cells.¹²⁰ GSK′772 has also completed phase IIa trials in patients with moderate to severe rheumatoid arthritis (NCT02858492),¹²¹ ulcerative colitis (NCT02903966),¹²² or plaque-type psoriasis (NCT02776033).¹²³ The results of these trials suggest that GSK′772 is not suitable as a monotherapy for rheumatoid arthritis or ulcerative colitis.^121,122 However, the outcomes of the trial in mild-to-moderate psoriasis suggest that inhibition of RIPK1 might have an impact on disease and that further studies are warranted with higher doses of GSK′772 and in patients with more active disease.¹²³

GSK′095 similarly showed efficacy in ADP-Glo biochemical and U937 cellular assays (IC₅₀ = 6.3 and 10.0 nM, respectively) while maintaining complete kinase selectivity toward RIPK1.¹¹⁹ GSK′095 also demonstrated high species selectivity toward primate RIPK1 and a good PK profile across rat, dog, and monkey models, with low clearance (27, 9.8, and 6.4 mL min^–1 kg^–1, respectively), a moderate volume of distribution (1.8. 1.1, and 1.8 L/kg, respectively), short-to-moderate terminal half-life (2.2, 1.7, and 4.2 h, respectively), and good oral bioavailability (84, 77 and 88%, respectively).¹¹⁹ Furthermore, GSK′095 was efficacious at inhibiting necroptosis (stimulated with TNF, a SMAC mimetic, and either QVD-OPh or zVAD.fmk) in human whole blood and in ex vivo tumor cultures of patient-derived organotypic spheroids (PDOTS) from freshly resected pancreatic, adenocarcinoma, colorectal, breast, and gastric cancer patients.¹¹⁹ GSK′095 recently completed a phase I clinical trial for pancreatic adenocarcinoma and other advanced solid tumors (NCT03681951).¹¹⁹

X-ray crystal structures of GSK′481, GSK′772, and GSK′095 bound to the RIPK1 kinase domain (residues 1–294, C34A, C127A, C233A, and C240A; PDBs 5HX6, 5TX5 and 6RLN, respectively) revealed that the benzoxazepinones are another class of type III inhibitors, which also bind to the same allosteric pocket as the necrostatins and DHPs.^118−120 In fact, these crystal structures were obtained by first reproducing the published cocrystal structure of Nec-4 bound to this RIPK1 construct and then displacing Nec-4 with GSK′772 or GSK′095.¹¹² Overall, the structure of RIPK1 is highly similar between the benzoxazepinones and other type III inhibitor-bound structures (Figure 3B). The benzoxazepinones all form H-bond interactions with the backbone residues of the DLG motif, which is in a classical “DFG out” conformation. In each structure, the benzoxazepinone moiety resides in a tight pocket formed by two β-strands defined by L90–V91–M92 and I43–M44–K45, with the benzyl group occupying an allosteric lipophilic pocket at the back of the ATP binding site. The benzoxazepinone ring also overlaps with space that would be occupied by the ATP α-phosphate, making these compounds ATP-competitive. Furthermore, the benzyl group of the compound takes the place of L157 from the DLG motif and forms an interaction with H136 of the HRD motif, potentially contributing to the stability of the inhibitor-bound conformation.^118−120 Our understanding of the strong primate specificity of many of the type III RIPK1 inhibitors is hampered by the lack of any published nonhuman RIPK1 structures. However, through identification of the binding site and subsequent sequence alignments of primate and nonprimate RIPK1, it was observed that the sequence differences in the activation loop, the C-helix, and the glycine-rich loop might result in nonprimate RIPK1 having less flexibility to undergo the significant conformational reordering required to bind the benzoxazepinones.¹¹⁸ Mutagenesis studies then confirmed the activation loop as the key region in murine RIPK1 that is responsible for the primate species selectivity of the benzoxazepinones.¹¹⁸

Following the disclosure of GSK′481 and its X-ray structure, other teams also began working on the benzoxazepinone scaffold. Takeda developed benzoxazepinone 11 (Figure 6) through the hybridization of a HTS hit and GSK′772 and subsequent optimization with the aid of structure-based drug design.¹²⁴ Benzoxazepinone 11 showed the best balance of potency (K_i = 0.91 nM), RIPK1 residence time (t_1/2 = 210 min by TR-FRET assay), microsomal stability (CL < 1 μL min^–1 mg^–1 in mouse and human models), and P-gp-mediated efflux (ER = 0.7).¹²⁴ When tested against a panel of 406 kinases, 11 displayed high selectivity for RIPK1 at 10 μM (S₍₅₀₎ = 0.00(2), S₍₂₀₎ = 0.03), with the exception of LIMK2 (85% inhibition @ 10 μM).¹²⁴ When tested against the Eurofins Panlabs panel of 106 other targets, 11 showed significant inhibition (>50% @ 10 μM) of only monoamine oxidase MAO-B (53%) and cannabinoid receptor CB2 (54%).¹²⁴ Additionally, 11 exhibited high plasma exposure (AUC = 658 ng h mL^–1), moderate plasma stability (MRT = 3.1 h), a brain exposure equivalent to 50 nM, and a brain-to-plasma ratio of 0.3 in mice.¹²⁴ Moreover, 11 attenuated disease progression in a mouse EAE model of multiple sclerosis, making it a good tool compound for evaluating the inhibition of RIPK1 in in vivo mouse models of neurodegenerative disease.¹²⁴

Figure 6.

Other examples of the benzoxazepinone class.

Genentech developed benzoxazepinone 12 in a campaign to optimize the physicochemical properties of this inhibitor class (Figure 6).¹²⁵ Benzoxazepinone 12 showed the best balance of affinity (K_i^app = 6.0 nM by ADP² FI), potency (IC₅₀ = 54 nM in HT29 TNF/BV6/zVAD-fmk), and selectivity (S₍₅₀₎ = 0.00, S₍₂₀₎ = 0.01) against 219 other kinases at 10 μM.¹²⁵ Furthermore, 12 also exhibited low clearance (7.1 mL min^–1 kg^–1), a moderate volume of distribution (2.8 L kg^–1), a long half-life (t_1/2 = 4.8 h), and excellent oral bioavailability (63%).¹²⁵ Genentech also developed 13 (GNE684, Figure 6), which is reportedly the least primate-specific type III inhibitor of RIPK1 known (9- and 33-fold by K_i vs mouse and rat models, respectively).³⁵ GNE684 was shown to be a potent inhibitor of necroptotic cell death in vitro across multiple human and mouse cell lines and also maintained the high RIPK1 specificity of the benzoxazepinone class (S₍₂₀₎ = 0.00) against a 221 kinase panel at 10 μM.³⁵ Assessment of the PK profile (5 mg kg^–1) revealed that GNE684 has high clearance (CL_p = 49.2 mL min^–1 kg^–1), a moderate volume of distribution (V_d = 1.84 L kg^–1), and a short half-life (t_1/2 = 0.53 h).³⁵ Despite this profile, GNE684 provided protection in several in vivo models of inflammatory disease, including TNF-driven SIRS, colitis-induced by NEMO deficiency in IECs, and collagen antibody-induced arthritis.³⁵ Altogether, GNE684 is proving to be a good tool compound for in vivo models of necroptotic disease. For example, its lack of efficacy in KRAS mutant pancreatic tumor models and a B16 melanoma model suggests that these diseases are not driven by necroptosis.³⁵

A series of thio-substituted benzothiazoles was recently developed based on the observation that the benzoxazepinone carbonyl in GSK′772 did not form any interactions with the hinge region of RIPK1 (Figure 6).¹²⁶ Interestingly, where GSK′772 showed >70-fold better activity with the (S)-enantiomer (3.6 nM vs 279 nM in HT-29 cells stimulated with TNF, a SMAC mimetic, and zVAD.fmk), this substitution reduced this difference in activity to less than 10-fold (2.8 nM for 14 vs 22.6 nM for 15 in HT-29 cells stimulated with TNF, a SMAC mimetic, and zVAD.fmk).¹²⁶ To explain this dramatic reduction in activity between the enantiomers, the authors modeled the three-dimensional conformations of the enantiomers. These calculations suggest that the enantiomers of GSK′772 would occupy vastly different conformations, whereas the conformations of 14 and 15 would be quite similar. The authors propose that the increased bond length of the C=S bond relative to the C=O bond might increase the steric hindrance, resulting in reduced flexibility and restricted conformations.¹²⁶ A preliminary SAR around this scaffold also suggests it could be a promising lead for further development.

By appending alkyne substituents to GSK′772, another series of benzoxazepinones that occupy both the allosteric site and the ATP pocket was developed.¹²⁷ Of the compounds prepared, the cyclopropyl analogue ZB-R-55 (16, Figure 6) was the most potent in cellular assays (IC₅₀ = 0.34 nM against U937 cells stimulated with TNF, a SMAC mimetic, and zVAD.fmk) and enzymatic assays (IC₅₀ = 5.7 nM by ADP-Glo and 16 nM by ³³P-radiolabeled assay).¹²⁷ ZB-R-55 retained the exquisite kinase selectivity of the benzoxazepinone class, showing <30% inhibition of kinase activity against the Reaction Biology Corp and Eurofins panels when tested at 1 μM.¹²⁷ The pharmacokinetic profile was favorable in mice (3 mg kg^–1 p.o. and 1 mg kg^–1 i.v.), with high oral exposure (AUC_last = 15 018 ng h mL^–1), a high volume of distribution (V_{ss_obs} = 902 mL kg^–1), low clearance (CL = 3.54 mL min^–1 kg^–1) and excellent oral bioavailability (F = 99%).¹²⁷ ZB-R-55 was then evaluated in the SIRS and LPS-induced models of lethal shock and sepsis, respectively.¹²⁷ When tested against GSK′772 at the same dose (10 mg/kg), ZB-R-55 exhibited improved protection against hypothermia and improved survival (100% vs 80%).¹²⁷ These data suggest ZB-R-55 is a promising candidate for further development.

4.1.4. Other RIPK1 Inhibitors

As the kinase activity of RIPK1 is required for the induction of death receptor-mediated necroptosis, it is unsurprising that numerous efforts have been made to assess the ability of approved kinase inhibitors to block necroptosis through the inhibition of RIPK1. Such endeavors have identified ponatinib and pazopanib,¹²⁸ sorafenib,¹²⁹ and tozasertib¹³⁰ as type I and II kinase inhibitors of RIPK1. While efforts have been made to generate more selective derivatives from these starting points, no useful tool compounds or clinical candidates have emerged.^131,132 Similarly, other type II inhibitors of RIPK1 have been identified from targeted chemical library screens^133−136 or from compounds where off-target activity was observed toward RIPK1 when developing inhibitors for other kinases.¹³⁷ The key advantage of type II RIPK1 inhibitors is their ability to overcome the primate selectivity observed in all type III scaffolds;^118,133 however, this advantage has so far been overshadowed by their poor kinome selectivity and suboptimal physicochemical properties. Additionally, efforts to optimize type I and type II RIPK1 inhibitors are hampered by the current lack of X-ray cocrystal structures of RIPK1 bound to these compounds. All the current structures are for the binding of type III inhibitors, which stabilize the inactive form of RIPK1. While type II inhibitors may bind to a similar RIPK1 conformation (although there is no experimental data to confirm this), type I inhibitors stabilize kinases in their active forms. We would predict the active conformation of RIPK1 to be quite different from the solved structures and to more closely resemble the active structure of PKA (Figure 3A), with the activation loop helix not structured, the αC helix swung inward with the formation of the classic salt bridge interaction between the αC helix and the VAIK motif, the L of the DLG motif oriented inward (resembling the canonical DFG-in conformation), and the regulatory spine intact. Consequently, no type I or type II RIPK1 inhibitor has entered the clinic or provided a tool for delineating necroptosis biology.

Denali Therapeutics has also developed several clinical candidates from their RIPK1 inhibitor program. Their first candidate, DNL104, was developed for treatment of ALS and Alzheimer’s disease.¹³⁸ In a phase I ascending dose study (NTR6257), DNL104 exhibited marked inhibition of RIPK1 phosphorylation and demonstrated CNS-penetrant properties; however, its development was discontinued due to liver abnormalities observed in the multiple ascending dose cohort.¹³⁸ Two molecules from the same company, DNL747 and DNL758, have also undergone phase Ia (NCT03757351 and NCT03757325) and phase Ib (NCT04469621) trials.¹³⁹ More recently, Eli Lilly purchased the rights to Rigel Pharmaceuticals’ RIPK1 inhibitor R552, which has completed a phase I clinical trial and is planned to undergo phase II studies in autoimmune and inflammatory diseases. Additionally, GenFleet Therapeutics is recruiting for a phase I trial (NCT04676711) of their RIPK1 inhibitor GFH312. For more details on the patent literature describing RIPK1 inhibitors, the reader is directed to a recent review on this topic, which details the development of several compound classes discussed in this review as well as other emerging scaffolds that fall outside the scope of this work due to the lack of published data.¹⁴⁰

RIPA-56 (17) was reported to be a highly potent, selective, and metabolically stable drug candidate for RIPK1-mediated disease (Figure 7).¹⁴¹ RIPA-56 was developed through N-hydroxylation of a hit identified through HTS of a 200 000 compound library against HT29 cells treated with a necroptotic stimulus (TNF, a SMAC mimetic, and zVAD.fmk).¹⁴¹ The authors report that RIPA-56 is potent against both human and murine cells in vitro (EC₅₀ = 28 and 27 nM against HT29 and L929 cells, respectively), highly selective for RIPK1 over other kinases, metabolically stable in vitro (t_1/2 = 128 and 35.5 min and Cl_int = 5.40 and 19.5 μL min^–1 mg^–1 in HLM and MLM, respectively), and efficacious in protecting mice from TNF-induced lethal shock and organ damage.¹⁴¹ However, upon in vivo pharmacokinetic (PK) testing (2 mg kg^–1 i.v.), RIPA-56 also displayed low exposure (AUC_0–∞ = 0.32 μg h mL^–1), high clearance (Cl_p = 103 mL min^–1 kg^–1), and a large volume of distribution (V_ss = 27.8 L/kg).¹⁴¹ The authors also propose that RIPA-56 binds to the same allosteric pocket as other type III inhibitors of RIPK1. Furthermore, the fluorinated analogue SIR1-365 (18) was assessed for safety and efficacy in patients with severe COVID-19 (NCT04622332).

Figure 7.

Chemical structures of other RIPK1 inhibitors.

Various other inhibitors of RIPK1 have been identified (structures not shown); however, as many of these compounds were developed as inhibitors of other kinases or are hits from library screens without optimization, these scaffolds primarily serve as a demonstration of the increasing efforts being directed toward the discovery and development of RIPK1 inhibitors as therapeutics. The reader is directed to another review for further details on these recently identified scaffolds.¹⁴²

4.2. Inhibitors of RIPK3

RIPK3 presents an attractive target for the inhibition of necroptosis, as it is downstream of RIPK1 and can potentially circumvent the issues associated with RIPK1’s other cellular functions. Furthermore, RIPK3 directly phosphorylates MLKL, so inhibiting its catalytic activity, oligomerization, or interaction with MLKL could also present unique opportunities to inhibit necroptosis at a late stage of pathway engagement.

4.2.1. GSK Compounds

GSK first developed the RIPK3 inhibitors 19 (GSK′843) and 20 (GSK′872) (Figure 8) through HTS using a fluorescence polarization (FP) assay and subsequent optimization of the hits.³ These two compounds were shown to be potent in FP (IC₅₀ = 8.6 and 1.8 nM, respectively) and ADP-Glo (IC₅₀ = 6.5 and 1.3 nM, respectively) biochemical assays and upon necroptotic stimulation in both human (HT29 + TNF, a SMAC mimetic, and zVAD.fmk) and murine cells (3T3-SA + TNF and zVAD.fmk; or PEC, L929, SVEC and BMDM + TNF, a SMAC mimetic, and zVAD.fmk) in vitro.^3,68 Furthermore, both compounds showed >1000-fold selectivity for RIPK3 over 291 other kinases at 1 μM (S₍₂₀₎ = 0.24 and 0.14, S₍₅₀₎ = 0.11 and 0.07, respectively) and did not inhibit RIPK1 activity, although data for the RIPK1 assays were not published.⁶⁸ GSK′843 and GSK′872 inhibited RIPK2 activity by 84% and 76%, respectively, when assayed at 1 μM.⁶⁸ This RIPK2 inhibitory activity must be taken into consideration when employing these compounds to investigate necroptosis in the context of autoinflammatory diseases.

Figure 8.

Chemical structures of RIPK3 inhibitors developed by GSK.

Recently, GSK′843 was used as a tool compound to enable the determination of the first human RIPK3 crystal structure (PDB 7MX3).¹⁴³ The compound binds in the expected type I inhibitor mode, and human RIPK3 is in the active conformation with the K50–E60 salt bridge intact and much of the activation loop resolved (Figure 9A). The protein conformation is similar to that of apo mouse RIPK3 (PDB 4M66).¹⁴⁴

Figure 9.

Selected crystal structures of RIPK3 with small-molecule inhibitors. (A) Apo mouse RIPK3 (yellow; PDB 4M66)¹⁴⁴ compared to BMS 27-bound mouse RIPK3 (red; PDB 6OKO),¹⁴⁵ human RIPK3 bound to GSK′843 (19, light orange; PDB 7MX3),¹⁴³ and BMS 28 (orange; PDB 7MON);¹⁴³ the compounds are shown in the surface view. (B) A comparison between the active sites of all published RIPK3 structures, highlighting the repositioning of the regulatory spine and DFG motif residue F162. Compounds are shown in stick format.

Later, GSK also identified the inhibitor 21 (GSK′840) through a DEL screen against the RIPK3 kinase domain (residues 2–328).⁶⁸ GSK′840 was shown to bind and inhibit RIPK3 more potently than either GSK′843 or GSK′872 in biochemical assays (IC₅₀ = 0.9 and 0.3 nM by FP and ADP-Glo, respectively), to be more potent in cellular necroptosis assays (HT29 + TNF, a SMAC mimetic, and zVAD.fmk; IC₅₀ values not disclosed), and to be a more selective RIPK3 inhibitor when tested against a panel of 300 kinases at 1 μM (S₍₂₀₎ = 0.07, S₍₅₀₎ = 0.02).⁶⁸ Interestingly, GSK′840 is inactive in murine cells, suggesting that it might bind to RIPK3 in a manner distinct from GSK′843 and GSK′872 that results in conformational changes that are unfavorable (or ineffective) in murine RIPK3. Unfortunately, no crystallographic data have been reported for GSK′840 to clarify this species specificity. It was observed that while these RIPK3 inhibitors inhibit RIPK1-independent necroptosis, in the absence of a caspase inhibitor they also induce RIPK3-dependent apoptosis at high concentrations (>1 μM).⁶⁸ This has raised some toxicity concerns for the development of RIPK3 inhibitors as anti-inflammatory therapies and may explain why no RIPK3 inhibitor has yet entered the clinic.⁶⁰

Recently, two cyclized analogues of GSK′872 have been reported, namely, Zharp-99 (22)¹⁴⁶ and compound 23(147) (Figure 10). Zharp-99 demonstrated improved efficacy when compared to GSK′872 in cell lines treated with necroptotic stimuli: human HT29 and mouse embryonic fibroblasts (MEF) stimulated with TNF, a SMAC mimetic, and zVAD.fmk; mouse and rat BMDMs stimulated with LPS and zVAD.fmk; and HSV-1-infected murine L929 cells.¹⁴⁶ Zharp-99 was also shown to inhibit RIPK3 and MLKL phosphorylation in both human HT29 and mouse L929 cells and to inhibit death arising from forced RIPK3 dimerization in NIH3T3 cells expressing FKBP-tagged mRIPK3.¹⁴⁶ However, Zharp-99 was unable to inhibit the formation of the RIPK1–RIPK3 necrosome complex in HT-29 cells stably expressing Flag–RIPK3 or MLKL polymerization in HeLa cells expressing a construct encoding an inducibly dimerizable form of the killer N-terminal region of MLKL.¹⁴⁶ Furthermore, Zharp-99 is able to directly bind RIPK3 (K_D = 1.35 nM) and block its kinase activity (IC₅₀ < 1 μM by ADP-Glo) and has no impact on RIPK1 kinase activity at 10 μM; however, its wider kinome selectivity has not yet been determined.¹⁴⁶ The toxicity profile of Zharp-99 was also evaluated, where it showed minimal (<10%) inhibition of cytochrome P450 isozymes (CYP3A4, CYP2D6, CYP1A2, CYP2C9, and CYP2C19) at 10 μM and low inhibition of hERG (IC₅₀ > 10 μM).¹⁴⁶ However, Zharp-99 was found to induce on-target apoptosis to a greater extent than GSK′872 in MEF cells, suggesting that this molecule is more relevant as a tool compound than as a lead for clinical development.¹⁴⁶ Zharp-99 also demonstrated moderate clearance (Cl_int = 212 and 67 mL^–1 min^–1 kg in MLM and RLM, respectively) and low stability (t_1/2 = 26 and 36 min in MLM and RLM, respectively) in vitro, which translated into good exposure (AUC = 8.2 μg h mL^–1 at 10 mg kg^–1 dosed p.o.), a moderate clearance (33 mL min^–1 kg^–1) and volume of distribution (4.4 L/kg), and a short half-life (t_1/2 = 1.5 h) in vivo.¹⁴⁶ Importantly, administration of Zharp-99 (5 mg/kg, i.p.) also provided strong protection against lethal shock and hypothermia in a TNF-induced in vivo SIRS model.¹⁴⁶

Figure 10.

Recently developed cyclized analogues of GSK′872.

Compared to GSK′872, 23 (Figure 10) demonstrated improved cellular potency in both human HT29 and murine MEF cells (IC₅₀ = 0.42 and 0.54 μM vs 1.51 and 2.51 μM, respectively) and no toxicity at 20 μM.¹⁴⁷ While 23 was shown to be 1000-fold more selective for RIPK3 over RIPK1 in a direct-binding assay (K_D = 7.1 and 7200 nM, respectively), the broader aspects of its selectivity have not been fully explored, with the inhibition of only seven other kinases being investigated.¹⁴⁷ Furthermore, 23 displayed moderate inhibition of CYP1A2 and 2C19 (37% and 20% at 10 μM, respectively) and minimal inhibition of hERG (IC₅₀ > 30 μM).¹⁴⁷ The DMPK profile of 23 also revealed high levels of plasma protein binding in human, rat, and mouse models (>97% in all cases), a low efflux ratio (0.72), and poor metabolic stability in human and mouse liver microsomes, with short half-lives (t_1/2 = 28.75 and 20.26 h, respectively) and high clearance (Cl_int = 60 and 270 mL min^–1 kg^–1, respectively).¹⁴⁷ Interestingly, the ketal functionality present in 23 was shown to be stable in simulated gastric fluid (pH 1.2) over 24 h, and administration of 23 (5 mg/kg, i.p.) also provided strong protection against lethal shock and hypothermia in a TNF-induced in vivo SIRS model.¹⁴⁷

4.2.2. Aminobenzothiazole Compounds

A screen of 500 fluorinated compounds in an MTT assay against human HT29 cells treated with a necroptotic stimulus (TNF, a SMAC mimetic, and zVAD.fmk) identified aminobenzothiazole 24 (TAK-632) as a novel scaffold for RIPK3 inhibitors (Figure 11).¹⁴⁸ TAK-632 was able to inhibit necroptosis in human THP-1 and U937 cells (treated with TNF, a SMAC mimetic, and zVAD.fmk), murine L929 and J774A.1 cells (treated with TNF and zVAD.fmk), and HT29 cells treated with Fas in addition to RIPK1-independent necroptosis induced by Poly(I:C) plus zVAD.fmk in RIPK1-deficient MEF cells.¹⁴⁸ Furthermore, TAK-632 was shown to inhibit RIPK3 kinase activity in the KINOMEscan and [γ-³²P] ATP radiometric assays (K_D = 326 nM and IC₅₀ = 90 nM, respectively).¹⁴⁸ However, TAK-632 was originally developed as a pan-RAF inhibitor and shows potent inhibition of not only RAF kinases (IC₅₀ = 8.3 and 1.4 nM against B-RAF and C-RAF, respectively)¹⁴⁹ but also RIPK1 (K_D = 105 and 480 nM by ADP-Glo and KINOMEscan, respectively) and the wider kinome (S₍₃₅₎ = 0.14 against 90 kinases at 1 μM).¹⁴⁸ Subsequent interrogation of the SAR around TAK-632 culminated in the development of benzothiazole 25 (SZM594, Figure 11).^148,150 Compared to TAK-632, 25 was shown to be more potent in the HT29 cellular necroptosis assay (EC₅₀ = 0.44 μM vs 1.44 μM), displayed reduced cytotoxicity (CC₅₀ > 50 μM vs 36.5 μM), had greater affinity for RIPK3 (K_D = 81 nM) and selectivity over RIPK1 (K_D > 5,000 nM) in the KINOMEscan assay, and conferred better protection in the in vivo TNF and zVAD.fmk-induced SIRS mouse model (100% at 25 mg kg^–1 vs 20% at 25 mg kg^–1).¹⁵⁰ Benzothiazole 25 still maintains significant activity against B-RAF (96.4% inhibition at 1 μM) and other kinases.¹⁵⁰ While B-RAF is not involved in necroptosis, its inhibition might still complicate the interpretation of experimental results, especially in vivo. Further development of this scaffold delivered SZM679 (26), which is selective for RIPK1 over RIPK3 (K_D = 8.6 nM and >5 μM, respectively, by KINOMEscan; selectivity over other kinases not reported).¹⁵¹

Figure 11.

Chemical structures of aminobenzothiazole-based inhibitors.

4.2.3. BMS Compounds

BMS undertook a HTS campaign with a homogeneous time-resolved fluorescence (HTRF) assay, followed by time-dependent inhibition studies, to specifically identify type II inhibitors of RIPK3.¹⁴⁵ A search of the BMS internal library for compounds that were structurally related to the type II inhibitors identified in the HTS campaign then produced pyrrolopyridine 27 (Figure 12). Pyrrolopyridine 27 was shown to be selective for RIPK3 over RIPK2 and RIPK1 (IC₅₀ = 0.18, >15, and 1.5 μM by HTRF, respectively); however, 27 was originally developed as an inhibitor of c-Met (IC₅₀ = 5.7 and 1.9 nM by HTRF and [γ-³²P] ATP radiometric assays, respectively) and also exhibits strong inhibition of FLT-3 and VEGFR-2 (IC₅₀ = 2 and 27 nM, respectively, by [γ-³²P] ATP radiometric assay).^145,152

Figure 12.

Chemical structures of RIPK3 inhibitors developed by BMS.

Importantly, pyrrolopyridine 27 enabled access to an inhibitor-bound mouse RIPK3 X-ray crystal structure (residues 1–313, C111A, C-terminal His₁₀ tag; PDB 6OKO, Figure 9A).¹⁴⁵ Compared to the apo mouse RIPK3 structure (PDB 4M66),¹⁴⁴ the DFG motif has flipped into a classical DFG-out conformation, confirming that pyrrolopyridine 27 adopts the targeted type II binding mode (Figure 9B). The terminal aromatic group binds deeply into the hydrophobic allosteric site of RIPK3 and replaces the F162 of the DFG, exposing the D161 backbone and facilitating the formation of H-bond interactions with the pyridone oxygen. The positioning of this terminal aromatic group also stabilizes the R-spine, with residues H141, M65, and L76 remaining in similar positions compared to those in the apo structure. Interestingly, 27 makes a H-bond with K51 of the VAIK motif and stabilizes the catalytically essential E61–K51 interaction; however, the position of the DFG motif would still prevent kinase activity.

More recently, the closely related compound 28 (Figure 12) was used to determine the structure of the human MLKL pseudokinase and human RIPK3 kinase domain complex (PDB 7MON, Figure 9A).¹⁴³ Interestingly, unlike that in the mouse RIPK3 27-bound structure (PDB 6OKO),¹⁴⁵ the R-spine of human RIPK3 remains intact upon compound binding due to the alternate positioning of the fluorophenyl tail group in the human structure. The tail group sits adjacent to the αC helix, which is pushed away from the center of the N-lobe so that the K50–E60 salt bridge cannot form (side chains separated by 7.4 Å, Figure 9B). Therefore, the kinase is in an inactive conformation even though F161 of the DFG motif is not flipped out as it is in the inactive mouse RIPK3 structure bound to compound 27. Despite the differences in the DFG motif conformation and the presence or absence of the K–E salt bridge in the inactive human and mouse RIPK3 structures (PDB 7MON and 6OKO, respectively) both show a translated αC helix and a marked shortening of the resolved section of their activation loop in comparison with their active conformation counterpart (PDB 7MX3 and 4M66, respectively; Figure 9A).

Through comparison of the crystal structures of 27 bound to mouse RIPK3 and c-MET kinase (PDB 3CE3), the structure-based optimization of this scaffold was performed, leading to the development of aminopyridine 29 (Figure 12).¹⁴⁵ Aminopyridine 29 is a more potent inhibitor of RIPK3 (IC₅₀ = 9.1 nM) with improved selectivity over RIPK1 and c-MET (IC₅₀ = 5.5 and 1.1 μM, respectively). However, screening against an undisclosed number of other kinases revealed that 29 inhibits 42 other kinases with IC₅₀ values of <1 μM, potentially limiting its utility as a tool to interrogate RIPK3 biology.¹⁴⁵ Furthermore, the anti-necroptotic activity of these compounds has not yet been demonstrated either in vitro or in vivo.

4.2.4. Other RIPK3 Inhibitors

Numerous efforts have been pursued to investigate approved kinase inhibitors in the context of inhibiting RIPK3. These studies have identified ponatinib,¹²⁸ dabrafenib, regorafenib, vemurafenib, and sorafenib as inhibitors of RIPK3.¹⁵³ However, these inhibitors display a lack of kinome selectivity. This is highlighted by the fact that both ponatinib and sorafenib are also inhibitors of RIPK1,^128,129 limiting their development as therapies for necroptotic diseases.

HTS efforts have also identified other putative type I and II RIPK3 inhibitor scaffolds^154−156 (Figure 13). The quinoline GW′39B (30) was identified by screening against NIH-3T3 cells expressing a RIPK3 construct that kills cells following inducible dimerization.¹⁵⁴ GW′39B was able to inhibit RIPK3 in this assay (EC₅₀ = 73.6 nM), protect both human and murine cells from diverse necroptotic stimuli, and inhibit MLKL phosphorylation and oligomerization.¹⁵⁴ However, because GW′39B was originally developed as a RET kinase inhibitor,¹⁵⁷ its kinase specificity and potential to activate RIPK3-dependent apoptosis remain of outstanding interest and preclude its use as a tool compound at this stage. HS-1371 (31), another quinoline, was also identified as a potent RIPK3 inhibitor (EC₅₀ = 20.8 nM) in another HTS of known kinase inhibitor chemotypes.¹⁵⁵ While HS-1371 inhibited RIPK3 autophosphorylation and rescued both human and murine cell lines from various necroptotic stimuli in vitro, it also showed apoptosis-related cytotoxicity similar to GSK′872.¹⁵⁵ Additionally, the kinase specificity of HS-1371 has yet to be evaluated, which is an important consideration due to its original development as an ALK inhibitor.¹⁵⁸ Two thienopyridines, 32 (GSK′067) and 33 (GSK′074, Figure 13), were identified by screening known kinase inhibitors against the MOVAS murine cell line treated with a necroptotic stimulus (TNF and zVAD.fmk).¹⁵⁶ Despite the identified compounds being potent RIPK3 inhibitors in both human and murine cells in vitro and not driving RIPK3-mediated apoptosis, they are extremely nonspecific (GSK′074 S₍₃₅₎ = 0.05 against 403 kinases at 100 nM) and have higher affinities for RIPK1 than RIPK3.¹⁵⁶

Figure 13.

Chemical structures of other identified RIPK3 inhibitors.

4.3. Small Molecules Targeting MLKL

Because MLKL was only identified as the terminal effector of necroptosis as recently as 2012, the field of small-molecule targeting of MLKL is still in its infancy. Being a pseudokinase and therefore lacking any catalytic activity, the impact of targeting the pseudoactive ATP-binding site with ligands remains unclear. Fortunately, the development of small molecules that target other pseudokinase domains (e.g., TYK2¹⁵⁹ and JAK2¹⁶⁰) suggests this could be a viable strategy.

4.3.1. Necrosulfonamide

The initial discovery of MLKL binders was somewhat serendipitous. In a screening campaign of a 200,000 compound library against HT29 cells treated with a necroptotic stimulus (TNF, a SMAC mimetic, and zVAD.fmk), a hit was identified that inhibited necroptosis with IC₅₀ < 1 μM.⁵⁴ Through investigation of the SAR around this hit, the optimized necrosulfonamide (NSA, 34) was developed (Figure 14).¹⁶¹ NSA inhibited necroptosis in HT29 cells treated with TNF, a SMAC mimetic, and zVAD.fmk (IC₅₀ = 124 nM); however, it had no effect in murine L929 or 3T3 cells treated with a necroptotic stimulus.⁵⁴ The molecular target of NSA was determined to be downstream of RIPK3 and was identified as MLKL.⁵⁴ NSA covalently binds to C86 of human MLKL; however, this C86 is a tryptophan in mouse MLKL, which explains the species selectivity of NSA.⁵⁴

Figure 14.

Chemical structure of the MLKL inhibitor necrosulfonamide (NSA).

Recently, studies employing NMR experiments have indicated that upon covalent binding to human MLKL, NSA only forms weak interactions with the helical bundle and the first brace helix.¹⁶² This supports earlier molecular dynamics simulations suggesting that NSA forms a critical π–cation interaction with K157 of the second brace helix, leading to a reduction in the α-helical content of the brace helices and the subsequent formation of several new interactions between the pseudokinase domain, the second brace helices, and the helical bundle.¹⁶³ NSA is therefore proposed to lock MLKL in an inactive conformation wherein it inhibits the release of the killer 4HB domain and subsequent oligomerization and translocation to the membrane.¹⁰ Unfortunately, further development of NSA has been limited due to its narrow SAR and the moderate potency and selectivity associated with targeting a surface cysteine residue rather than a specific binding pocket. Therefore, it should only be used as a tool compound for in vitro research and strictly only in studies of primate cells. In addition, other studies have shown that NSA can also impact gasdermin D (GSDMD) processing, either directly or upstream via caspase 1, and therefore can inhibit pyroptosis, another form of inflammatory cell death.^164,165 This cross-reactivity with other cysteine-containing proteins is perhaps not surprising, especially when NSA is used at high concentrations.

4.3.2. Xanthine-Based Ligands

In a screening campaign of a 200 000 compound library against HT29 cells treated with a necroptotic stimulus (TNF, a SMAC mimetic, zVAD.fmk), a xanthine-based hit was identified that inhibited necroptosis with EC₅₀ = 390 nM.¹⁶⁶ This hit was then developed into TC13172 (35, Figure 15), which inhibited necroptosis in the same cellular assay with EC₅₀ = 2.0 nM.¹⁶⁶ Xanthine 35 was also shown to have no inhibitory effect on the kinase activity of RIPK1 or RIPK3 at 10 μM. Furthermore, 35 was shown to covalently bind to C86 of the MLKL 4HB, suggesting it has a similar mechanism of action to NSA. This was further demonstrated through the inability of 35 to inhibit the phosphorylation of MLKL by RIPK3 and its ability to prevent MLKL oligomerization and membrane translocation.¹⁶⁶ Interestingly, 3D NOESY NMR experiments of 35 binding to the MLKL helical bundle (residues 2–154) suggest that the phenyl ring did not strongly interact with the helical bundle and was likely oriented toward the C-terminal end of the first brace helix.¹⁶² This suggested that this moiety might interact with the second brace helix or the pseudokinase domain in the context of full-length MLKL. Like NSA, 35 relies on the presence of C86 in human MLKL and primate orthologs and is therefore useful only as a tool for studies of MLKL in cell lines from primates.

Figure 15.

Chemical structures of xanthine-based MLKL ligands.

To understand the mechanism of this class of xanthine-based compounds, Boehringer Ingelheim developed xanthine 36 and determined its impact on the structure of the MLKL N-terminal executioner domain.¹⁶² Using a combination of 3D NOESY NMR experiments (residues 2–154) and X-ray crystallography (residues 2–150, PDB 6ZZ1, Figure 16) the authors demonstrated that upon the covalent modification of C86 by 36, the xanthine ring forms a π-stacking interaction with F148 of the first brace helix that stabilizes packing of the first brace helix against the helical bundle.¹⁶² Stabilization of this packing was proposed to lock the helical bundle and adjacent brace helices of MLKL in an inactive conformation, thereby impeding MLKL oligomerization and its necroptotic activity. The authors were also able to solve the apo structure (residues 2–150, PDB 6ZVO) and compare this to the xanthine-bound structure.¹⁶² This comparison revealed that the binding of xanthine 36 induces a rearrangement of only the side chains involved in or directly adjacent to compound binding (C86, F148, and R82). No impact on the overall architecture of the helical bundle was observed, and only minimal repositioning of the first brace helix upon compound binding was evident. Both the apo (PDB 6ZVO) and compound-bound (PDB 6ZZ1) structures closely resemble a previously described crystal structure of the human MLKL N-terminal domain cocrystallized with a monobody protein (PDB 6UX8).¹² This monobody binds to an alternative site on the α4 helix but also inhibits MLKL translocation to the membrane and subsequent cell death.¹²

Figure 16.

Selected crystal structures of MLKL with covalent small molecules interacting with the helical bundle domain. (A) Crystal structure of full-length mouse MLKL (PDB 4BTF),⁶ with the domain architecture highlighted (note: the domain architecture is drawn with the C-terminal to the left, i.e., the reverse of the sequence direction, to correspond to the orientation of the structure). The inset shows that MLKL is catalytically inactive due to sequence divergences at key kinase catalytic motifs. (B) An overlay of apo (PDB 6ZVO),¹⁶² xanthine ligand-bound 36 (PDB 6ZZ1),¹⁶² and benzimidazole 41-bound (PDB 7MN2)¹⁶⁷ human MLKL helical bundle domain structures, with compounds labeled. (C) Comparison of apo and compound-bound structures at the compound binding site. (D) The binding pocket that benzimidazole 41 induces is shown in surface view. Compounds are shown in stick format.

4.3.3. Other Small Molecules Targeting the N-Terminal Domain of MLKL

A series of uracil compounds was recently developed (Figure 17) to overcome the liabilities of the xanthine-based small molecules targeting MLKL, namely, their high levels of reactivity toward nucleophiles under physiological conditions and the potential for off-target toxicity.¹⁶⁸ A scaffold-morphing strategy identified the hit compound 37, with EC₅₀ = 3380 nM in HT29 cells treated with necroptotic stimulus (TNF, a SMAC mimetic, and zVAD.fmk). Elaboration of the SAR around this scaffold resulted in the development of uracils 38 and 39 (Figure 17) with markedly improved potencies (EC₅₀ = 82 and 31 nM, respectively).¹⁶⁸ Importantly, these optimized uracils showed little-to-no inhibition of RIPK1 (18 and 0% at 10 μM, respectively) and no inhibition of RIPK3 in ADP-Glo enzymatic assays, as well as no effects on the level of MLKL phosphorylation. Furthermore, no inhibition of cell proliferation or cell survival was observed when HT29 cells were treated with 5 μM 38 or 39, and the reactive 6-chloro group showed reduced reactivity toward glutathione (GSH) compared to xanthine 35 (t_1/2 = 48 or 160 h, respectively, compared to 20 min) when incubated together in DPBS buffer, suggesting a reduced propensity for off-target toxicity.¹⁶⁸

Figure 17.

Chemical structures of uracil-based MLKL ligands.

It was observed that these uracils did not inhibit necroptosis in MEF cells (stimulated with TNF, a SMAC mimetic, and zVAD.fmk), suggesting that they targeted the Cys86 in human MLKL that is not present in mouse MLKL. Furthermore, the binding of both uracil 38 and 39 to MLKL was shown to outcompete that of TC13172 35 in a pull-down assay.¹⁶⁸ The binding of 38 to Cys86 was then confirmed by incubation with the MLKL protein and analysis via MS/MS, as well as mutagenesis studies wherein 38 and 39 failed to inhibit necroptosis in HT29 cells transfected to express a C86S mutant.¹⁶⁸ Immunoblotting assays indicated that the uracils were partly able to reduce MLKL oligomerization, while immunofluorescence experiments showed that they inhibited the translocation of MLKL to the plasma membrane.¹⁶⁸ However, without available structures of 38 or 39 bound to full-length MLKL or structures of MLKL oligomers, the authors can only speculate that these compounds disrupt an unidentified amino acid that has no impact on MLKL oligomerization but is integral for its membrane translocation.

Recently, a series of fragments that bind noncovalently to the MLKL executioner domain have been reported (Figure 18).¹⁶⁷ Indole 40 was identified through an NMR-based fragment screening against the MLKL executioner domain (residues 2–154) and was shown to have a binding affinity (K_D) of 933 μM by NMR titration.¹⁶⁷ Optimization of this scaffold led to the development of benzimidazole 41 (Figure 18), with an improved K_D of 50 μM.¹⁶⁷ Importantly, the switch from an indole to benzimidazole scaffold afforded increased chemical stability upon chiral separation, facilitating the determination of the (S)-enantiomer as the active species. Furthermore, alkylation of the chiral alcohol fills an induced pocket in the binding site, while the carboxylate group was appended to improve compound solubility.¹⁶⁷

Figure 18.

Chemical structures of indole 40 and optimized benzimidazole 41.

The NMR costructure of benzimidazole 41 bound to the MLKL executioner domain indicated that these compounds occupy an induced pocket near the N-terminus of the protein, between the pairs of α-helices that form the four-helix bundle (Figure 16D).¹⁶⁷ The iso-propyl-phenyl group projects into the hydrophobic core of the four-helix bundle domain, with the carboxylate group oriented toward the solvent. Interestingly, the binding of 41 to the four-helical bundle domain does not markedly modify the structure compared to the unbound protein.¹⁶² Furthermore, the binding site of these compounds was shown to also be the binding site of the detergent monomer nonyl-maltoside, which when added with inositol phosphates can be used to induce the activation of the MLKL executioner domain in vitro.¹⁶⁹ Due to the shared nature of this binding site, the authors postulate that competition to detergent binding is a potential mode of action for MLKL inhibition. Unfortunately, benzimidazole 41 was unable to demonstrate efficacy in any cellular assays due to poor membrane permeability (3.1 × 10^–8 cm s^–1 in a PAMPA assay) or in a liposome leakage assay, presumably because it was out-competed by detergents under the assay conditions as a result of its low binding affinity for MLKL (K_D = 50 μM).¹⁶⁷ The further optimization of compounds to probe this newly identified binding site is of great interest, as it might illuminate possibilities to inhibit necroptosis at a late stage, such as MLKL multimerization or membrane interaction, and avoid the aforementioned issues of targeting the upstream kinases RIPK1 and RIPK3 or other regions of MLKL.

4.3.4. Aminopyrimidines

A screen of known kinase inhibitors against recombinant mouse MLKL using thermal shift assays identified aminopyrimidine 42 (compound 1, Figure 19) as a MLKL binder.⁹ It was demonstrated through surface plasmon resonance (SPR, K_D = 9.3 μM) and saturation transfer difference NMR (STD-NMR) studies that 39 bound to the nucleotide-binding site of MLKL. Furthermore, 42 was able to inhibit necroptosis in MDFs (IC₅₀ < 50 nM, treated with TNF, a SMAC mimetic, and QVD-OPh) and retarded the translocation of MLKL to the membrane.⁹ The inhibitory properties of 42 against human MLKL were later examined, where it demonstrated a moderate affinity for MLKL (K_D = 530 nM), poor kinome selectivity (inhibited 56 out of 403 kinases at 1 μM) including the inhibition of RIPK1 and RIPK3 (K_D = 64 and 680 nM, respectively), moderate cellular potency in TNF-stimulated FADD-deficient Jurkat cells (EC₅₀ = 1.85 μM), and toxicity at higher concentrations (∼2 μM).^170,171 As 42 was developed as a VEGFR2 inhibitor (IC₅₀ = 2 nM),¹⁷² this lack of selectivity for MLKL is unsurprising.

Figure 19.

Chemical structures of aminopyrimidine-based MLKL ligands.

More recently, an analogous compound (43, Compound 2, Figure 19) was described as a more potent inhibitor of necroptosis.¹⁷¹ Compared to 42, aminopyrimidine 43 showed a reduced affinity for MLKL and RIPK3 (K_D = 1800 and 1900 nM, respectively) and an improved affinity for RIPK1 (K_D = 19 nM) in a competitive binding assay despite also showing relatively poor selectivity over the wider kinome (S₍₃₅₎ = 0.31).¹⁷¹ Furthermore, the binding of 43 to all three key necroptosis effector proteins was confirmed through cellular thermal shift assays (CETSA) in both human U937 and murine MDF cells and in vitro photoaffinity labeling experiments. Interestingly, aminopyridine 43 was shown to be a more potent inhibitor of necroptotic cell death in U937 and HT29 cells (treated with TNF, a SMAC mimetic, and QVD-OPh or IDN-6556) relative to 42, with similar levels of toxicity.¹⁷¹ By employing a series of knockout cell lines, the authors also demonstrated that this cell death was mediated through caspase-dependent and BAX/BAK-independent apoptosis, which was also independent of RIPK1, RIPK3, and MLKL, suggesting that the cell death was mediated through the complex pharmacology of this scaffold.¹⁷¹ Finally, aminopyrimidine 43 was evaluated in vivo using a TNF-induced systemic inflammatory response syndrome (SIRS) model, where treatment corresponded to a modest delay (3 h) in hypothermia and a reduced incidence of death (one death in the treatment arm vs three deaths in the vehicle control).¹⁷¹

X-ray crystal structures of the human MLKL pseudokinase domain bound to 42 (residues 191–471, E366A and K367A; PDB 5KNJ)¹⁷⁰ and 43 (residues 190–471; PDB 6O5Z)¹⁷¹ confirmed that these small molecules occupy the ATP-binding site and extend into the allosteric pocket in a type II conformation (Figure 20). The aminopyrimidine motif forms H-bonds with the hinge region, while the urea interacts with E250 of the αC helix, as is typical of a type II binding conformation. Binding of the terminal aryl group in the allosteric pocket modestly displaces the αC helix relative to the apo structure (PDB 4MWI),¹⁷³ however, the αC E250-K230 H-bond is maintained. The trifluoromethoxy group of 42 displaces F350 of the GFE motif and, although the side chain is truncated, F350 appears to become positioned in a conformation analogous to a classical DFG-out motif. Despite these conformational changes, however, 42 does not prevent MLKL phosphorylation and therefore does not prevent RIPK3 binding.

Figure 20.

Selected crystal structures of the MLKL pseudokinase domain with and without small-molecule ligands. (A) Human MLKL bound to aminopyrazine 44 (PDB 5KO1)¹⁷⁰ has a similar conformation to the apo human MLKL pseudokinase domain (PDB 4MWI).¹⁷³ (B) Human MLKL bound to aminopyrimidine 42 (PDB 5KNJ)¹⁷⁰ has a similar conformation to human MLKL bound to aminopyrimidine 43 (PDB 6O5Z)¹⁷¹ and differs only slightly from apo human MLKL (panel A). (C) Comparison of the MLKL pseudoactive site in the apo and compound-bound structures, highlighting repositioning of F350, from the GFE motif in aminopyrimidine 42- and 43-bound human MLKL. Compounds are shown in surface format (panels A and B) or in stick format (panel C).

4.3.5. Other Small Molecules Targeting MLKL

Biogen screened a library of 5000 compounds using an ATP-competitive probe displacement assay to identify compounds that specifically bind the active site of MLKL but not RIPK1 or RIPK3.¹⁷⁰ One hit this screen identified was crizotinib, a type I kinase inhibitor. Crizotinib had a higher affinity for MLKL than either RIPK1 or RIPK3 (K_D = 217, 1300, and 6700 nM, respectively) but showed no cellular activity against TNF-stimulated FADD^–/– Jurkat cells, even at 40 μM.¹⁷⁰ Furthermore, crizotinib was developed as an inhibitor of c-MET and ALK (enzymatic IC₅₀ < 1.0 nM) and inhibits other kinases with similar potency,¹⁷⁴ making it unsuitable for specifically targeting MLKL in cellular or in vivo settings. This screen also identified aminopyrazine 44 (Figure 21), which was found to modestly bind MLKL (K_D = 230 nM) while being completely selective over RIPK1 and RIPK3 (K_D > 30 μM) and the broader kinome (inhibited only MLKL versus 403 kinases at 1 μM).¹⁷⁰ Interestingly, the enantiomer had a markedly lower affinity for MLKL (K_D = 7.6 μM), suggesting it could potentially be a useful negative control. Unfortunately, aminopyrazine 44 showed no cellular activity against TNF-stimulated FADD^–/– Jurkat cells at 40 μM and was unable to inhibit MLKL phosphorylation by RIPK3.¹⁷⁰

Figure 21.

Chemical structure of aminopyrazine 44.

The X-ray crystal structure of aminopyrazine 44 bound to the MLKL pseudokinase domain (residues 191–471, E366A and K367A; PDB 5KO1) has also been obtained, which confirms that it occupies the ATP site in a type I binding mode (Figure 20A and C).¹⁷⁰ The binding of 44 does not markedly change the conformation of MLKL compared to the apo structure (PDB 4MWI),¹⁷³ except for the glycine-rich loop being raised slightly (3.9 Å at E213, Cα to Cα). However, this does not appear to translate to other conformational changes. As our understanding of the conformational changes that MLKL must undergo to be in an activated state are limited, one can only speculate that type I inhibitors (such as crizotinib and 44) do not induce sufficient repositioning of key MLKL activity regulating elements (namely the αC helix or activation loop) to inhibit activity.^6,8,175 The cocrystal structure of the mouse MLKL pseudokinase domain in complex with mouse RIPK3 (PDB 4M69)¹⁴⁴ shows the repositioning of the MLKL αC helix to form an interface with mouse RIPK3 and a movement of the activation loop helix into the cleft between the N- and C-lobes. Interestingly, this conformation of the activation loop helix would induce minor clashes with both aminopyrazine 44 and aminopyrimidine 42. A recently published structure of the human MLKL and RIPK3 complex (PDB 7MON) suggests that unlike in the mouse counterpart aminopyrazine 44 binding could be accommodated within the human MLKL RIPK3 complex.¹⁴³ The binding of the type II inhibitor aminopyrimidine 42 would clash with F350 of the GFE (MLKL equivalent to DFG) motif, however, which is in the “in” conformation in the structure.¹⁴³ Whether the complex could tolerate conversion to the “DFG out” conformation is unknown.

The claims and conclusions of the paper disclosing aminopyrazine 44 leave several outstanding questions. First, while the cellular assay described in this study did not show antinecroptotic activity, it remains of outstanding interest whether a more conventional HT29, L929, MDF, or MEF cell line treated with necroptotic stimuli would be impacted by the compounds studied. Further, the assertion that targeting the pseudokinase domain of MLKL has no functional impact on necroptosis is based on only a handful of compounds, none of which can be considered “optimized” inhibitors of MLKL. A larger number and greater diversity of compounds with more potent binding affinities need to be evaluated before such a conclusion can be made. More broadly, this conclusion definitively highlights the challenge associated with targeting pseudokinase pseudoactive sites because of their ancestral origins shared with paralogous kinases. MLKL functions as a protein switch, with the necroptotic function regulated by interconversion between an inactive “open” state and an active “closed” state.^6,7,9 The authors’ own crystallographic studies demonstrate that aminopyrazine 44 binds to the closed state of MLKL, which is more suggestive that type I inhibitors do not stabilize a conformation of the pseudoactive site that inhibits the killer activity of the MLKL helical bundle. Additionally, the authors’ suggestion that aminopyrimidine 42, a type II binder of MLKL, exerts its antinecroptotic activity primarily through inhibition of RIPK1 and not its binding to MLKL further weakens their claim. Overall, small molecule targeting MLKL with increased potency and selectivity are needed to better understand the effect that compounds engaging MLKL have on necroptotic signaling and to remove any confounding effects due to simultaneous RIPK1 or RIPK3 inhibition.

5. The Right Tool for the Job

Throughout this Perspective, we aimed to arm the reader with data-encompassing selectivity, affinity, PK/PD properties, and species specificity, which are critical for selecting the best tool compound for inhibiting necroptosis either in vitro or in vivo. Table 1 summarizes our suggestions for first-choice tool compounds, along with their limitations. Again, we remind the reader that for the purpose of this Perspective, when we refer to in vivo animal models of necroptotic disease, we are generally referring to the rodent models easily accessible to most researchers.

Table 1. Existing Tool Compounds and Their Applicability.

compound	target protein	when to use	rationale	refs
GSK′772	RIPK1	√ cellular (primate) necroptosis assays	GSK′772 is highly potent and selective for RIPK1, with extensive characterization in PK/PD and preclinical models. It is unsuitable for use in nonprimate cells and in vivo models due to excessive primate selectivity	(120)
		√ biochemical assays
		√ in vitro primary cell (primate) assays
GNE684	RIPK1	√ cellular necroptosis assays	GNE684 is highly potent and selective for RIPK1, with similar levels of activity against primate and rodent orthologues. Due to some limitations with the PK/PD properties, GNE684 is more suitable for acute models of disease.	(35)
		√ biochemical assays
		√ in vitro primary cell (primate) assays
		√ in vivo models of necroptotic disease
GSK′840	RIPK3	√ cellular necroptosis assays	GSK′840 is more potent and selective for RIPK3 than GSK′872 or GSK′843 but is only active against primate RIPK3.	(68)
		√ biochemical assays
		√ in vitro primary cell (primate) assays
GSK′872	RIPK3	√ cellular necroptosis assays	GSK′872 is active against both primate and rodent RIPK3. Caution must be taken when used in vivo, as toxicity can occur at concentrations >1 μM.	(3)
		√ biochemical assays
		√ in vitro primary cell (primate) assays
		√ in vivo models of necroptotic disease
NSA	MLKL	√ cellular (primate) necroptosis assays	NSA can only bind and inhibit primate MLKL and not rodent orthologues. Its use should be limited to in vitro assays. It is important to note that additional studies have also linked NSA to pyroptosis.	(54, 164, 165)
		√ biochemical assays
		√ in vitro primary cell (primate) assays

6. Final Remarks

Continued efforts are still required to fully characterize the necroptosis pathway. For this, the ongoing development and correct use of chemical probes targeting RIPK1, RIPK3, and MLKL will be essential. Novel chemotypes and greater creativity are required to develop next-generation inhibitors of necroptosis that circumvent issues such as off-target activity, species specificity, and interference with the non-necroptosis functions of RIPK1 and RIPK3. Conversely, the potential of small-molecule activators of necroptosis remains unclear. These endeavors also require improved structural details of the necroptosis proteins, especially RIPK3 and MLKL, and their interactions.

Efforts to inhibit the necroptosis pathway have thus far focused on RIPK1, presumably because of its apical role and its implication in the pathway 16 years ago. Further research aimed at delineating the various kinase and scaffolding functions of RIPK1, as well as its roles in cell survival and apoptosis, will facilitate the further development of RIPK1 inhibitors as therapeutics for human diseases. An improved understanding of RIPK1 biology might provide further options for targeting RIPK1, such as altering its post-translational modifications other than the phosphorylation state (e.g., ubiquitination), modulating the conformation of the kinase domain, disrupting protein–protein interactions, and possibly developing activators for certain applications.

The development of RIPK3 inhibitors has been hampered by a lack of publicly available crystallographic data to support structure-based design and the potential toxicity issues arising from the ability of RIPK3 inhibitors to induce RIPK3-dependent apoptosis. However, opportunities still exist for potentially targeting RIPK3. It might be possible to develop type III inhibitors that bind to the distal pocket created by the DFG-out conformation, as identified in the pyrrolopyridine 27–mouse RIPK3 structure. While it is typically more difficult to target protein–protein interactions, especially without detailed structural knowledge of the necrosome, it might be possible to inhibit the interaction of RIPK3 with either RIPK1 or MLKL or indeed its own oligomerization.

Direct inhibition of MLKL represents an attractive option for generating antinecroptosis therapies because MLKL is the final effector of necroptosis, is expressed in a greater number of cell types than RIPK3, and might circumvent the issues observed with inhibition of RIPK1 (scaffolding function) and RIPK3 (activating apoptosis). Unfortunately, our understanding of the structural requirements for MLKL activation/inactivation is still lacking, as are our chemical tools to interrogate this. While type I binders of the pseudokinase domain might not be effective at reorganizing the αC helix and activation loop to prevent phosphorylation by RIPK3, allosteric binders (equivalent to type III kinase inhibitors) might have greater utility in this area, should they be developed. Such MLKL ligands might be able to inhibit release of the killer 4HB domain through noncovalent interactions, allowing their potential use in vivo. As our understanding of MLKL biology increases, it might also offer insight into ways we could design small molecules to target the oligomerization, translocation, or membrane interactions of MLKL.

Glossary

Abbreviations Used

BAK

BCL-2 homologous antagonist

BAX

BCL-2 associated X-protein

BMDM

bone marrow-derived macrophage

CB2

cannabinoid receptor type 2

CETSA

cellular thermal shift assay

CLint

intrinsic clearance

CNS

central nervous system

CRISPR

clustered regularly interspaced short palindromic repeats

Cas

CRISPR-associated protein

DAI

DNA-dependent activator of IFN regulatory factors

DAMPs

damage-associated molecular patterns

DPBS

Dulbecco’s phosphate-buffered saline

DEL

DNA-encoded library

DHP

dihydropyrazole

ds

doubled stranded

EAE

experimental autoimmune encephalomyelitis

ER

efflux ratio

ESCRT

endosomal sorting complexes required for transport

FADD

factor-associated suicide receptor-associated protein with death domain

FAS

factor-associated suicide receptor

FasL

factor-associated suicide receptor

FKBP

FK506 binding protein

FLT-3

fms-like tyrosine kinase

FP

fluorescence polarization

HB

helical bundle

GSDMD

gasdermin D

HLM

human liver microsome

hPXR

human pregnane X receptor

HTRF

homogeneous time-resolved fluorescence

cIAP1/2

cellular inhibitors of apoptosis 1 or 2

IDN

Idun

IEC

immune effector cell

IFN

interferon

IDO

indoleamine 2,3-dioxygenase

IFNAR1

interferon-α/β receptor subunit 1

IKK

inhibitor of nuclear factor κB kinase

IPMK

inositol polyphosphate multikinase

JAK2

janus kinase 2

KRAS

Kirsten rat sarcoma

LIMK2

LIM domain kinase 2

LPS

lipopolysaccharides

LUBAC

linear ubiquitin chain assembly complex

MEF

mouse embryonic fobroblast

MLKL

mixed lineage kinase domain-like protein

MLM

mouse liver microsomes

MRT

mean residence time

MAO-B

monoamine oxidase

NASH

nonalcoholic steatohepatitis

Nec

necrostatin

NF-κB

nuclear factor κB

NEMO

NF-κB essential modulator

NOD

nucleotide binding and oligomerization domain

NLRs

nucleotide binding and oligomerization domain-like receptors

NSA

necrosulfonamide

PAINS

pan-assay interference compounds

PDB

Protein Data Bank

PDOTS

patient-derived organotypic spheroids

PEC

parietal epithelial cells

poly(I:C)

polyinosinic:polycytidylic acid

PsKD

pseudokinase domain

Rd10

retinal degeneration 10

RHIM

receptor-interacting protein homotypic interaction motif

RLM

rat liver microsome

RIPK1

receptor-interacting protein kinase 1

RIPK2

receptor-interacting serine/threonine kinase 2

RIPK3

receptor-interacting protein kinase 3

RP

retinitis pigmentosa

shRNA

short hairpin ribonucleic acid

siRNA

small interfering ribonucleic acid

SIRS

systemic inflammatory response syndrome

SMAC

second mitochondria-derived activator of caspases

STD

saturation transfer difference

TAK1

transforming growth factor-β-activated kinase 1

TAK

tat-associated kinase

TNF

tumor necrosis factor

TNFR1

tumor necrosis factor receptor 1

TLR3

toll-like receptors 3

TLR4

toll-like receptors 4

TRADD

tumor necrosis factor receptor-associated death domain

TRAF2

tumor necrosis factor receptor-associated factor 2

TR-FRET

time-resolved fluorescence energy transfer

TRIF

toll/interleukin-1 receptor/resistance protein-domain-containing adaptor-inducing interferon-β

ZBP1

Z-DNA binding protein 1

Biographies

Christopher R. Gardner received his B.Sc.(Hons) and Ph.D. from the University of New South Wales (UNSW Sydney), where he worked in collaboration with the Children’s Cancer Institute (CIA). He completed his postdoctoral training at Bayer Cropscience in Frankfurt (Germany) before joining the Lessene lab at the Walter and Eliza Hall Institute of Medical Research (WEHI) in Melbourne (Australia). He is also a member of the medicinal chemistry team at Anaxis Pharma.

Katherine A. Davies is a postdoctoral researcher in the Czabotar laboratory at WEHI and a structural biologist for Anaxis Pharma. She received her B.Sc.(Hons) and Ph.D. from the University of Melbourne (Melbourne, Australia) for research performed at WEHI. Her research focuses on the understanding, at the structural level, of the proteins controlling necroptosis. She also drives the structural biology efforts at Anaxis Pharma.

Ying Zhang is a postdoctoral researcher in the Lessene laboratory at WEHI. She completed her B.Sc.(Hons) and M.Phil. at the Ocean University of China (Qingdao, China), followed by PhD studies at the University of Melbourne (Melbourne, Australia) for research performed at The Peter Doherty Institute for Infection and Immunity. Her research aims at understanding the events underpinning necroptosis using molecular biology and cutting-edge imaging strategies. She is also a member of biology team at Anaxis Pharma.

Martin Brzozowski is the Director of Medicinal Chemistry at SYNthesis med chem, a global drug discovery research provider. He previously held the position of senior postdoctoral fellow at WEHI, contributing to translational research covering early stage start-up company creation through to late-stage lead optimization programs. Martin obtained his Ph.D. from La Trobe University (Melbourne) studying protein kinase inhibitors and later completed postdoctoral training in continuous flow chemistry and photoredox catalysis at the University of Melbourne.

Peter E. Czabotar is a joint head of the Structural Biology Division at WEHI. He obtained his B.Sc.(Hons) and Ph.D. from Curtin University (Perth, Australia) for research performed at the Telethon Institute for Child Health Research (Perth, Australia) and undertook postdoctoral training at the National Institute for Medical Research (Mill Hill, UK) and at WEHI. The Czabotar Laboratory uses structural biology to understand cell death pathways at the molecular level in order to target these with therapeutics. He is also the Lead Structural Biologist for Anaxis Pharma.

James M. Murphy heads the Inflammation Division at WEHI. He completed his B.Sc.(Hons) at the University of Canterbury (New Zealand) before undertaking his Ph.D. studies at the Australian National University. He completed postdoctoral training with Tony Pawson and Frank Sicheri at the Lunenfeld-Tanenbaum Research Institute in Toronto (Canada) and at WEHI. The Murphy laboratory uses molecular, cellular, and physiological approaches to understand the roles of pseudokinases and kinases in cell signaling. He is also the Director of Biology at Anaxis Pharma.

Guillaume Lessene is a Laboratory Head and Theme Leader (New Medicines and Advanced Technologies) at WEHI. He graduated from the Ecole Nationale Superieure de Chimie de Paris and received a B.Sc.(Hons) from Paris 6 Pierre et Marie Curie (both in 1995). He received his Ph.D. from the University of Bordeaux 1 and, as a postdoctoral student. joined the laboratory of Prof Ken Feldman at the Pennsylvania State University working on the total synthesis of diazonamide A. He then joined medicinal chemistry team at WEHI in 2001. The Lessene laboratory works at the crossroad of medicinal chemistry and chemical biology of cell death pathways. He is also the Chief Scientific Officer of Anaxis Pharma.

Author Present Address

^# Bio21 Institute, 30 Flemington Road, Parkville, VIC 3052, Australia

The authors declare the following competing financial interest(s): All authors contribute or have contributed to a project developing necroptosis inhibitors with Anaxis Pharma Pty Ltd.