RIP3 induces apoptosis independent of pro-necrotic kinase activity

Introduction

Necroptosis is an alternative form of programmed cell death that is triggered when apoptosis is inhibited. This death pathway contributes to host defense against infection as well as to inflammation. Apoptosis is characterized by a requirement for caspases. Necroptosis depends on RIP3 kinase activation initiated by diverse innate immune mediators (Kaiser et al., 2013b). RIP3 kinase activity supports the recruitment of the mixed lineage kinase domain-like (MLKL) to trigger membrane leakage (Sun et al., 2012) with the consequent release of proinflammatory intracellular components (Kaczmarek et al., 2013).

Necroptosis has been broadly implicated in developmental failure following germ line disruption of caspase (Casp)8 or Fas-associated protein with death domain (FADD) (Kaiser et al., 2011; Oberst et al., 2011; Zhang et al., 2011). These same studies showed that RIP3-dependent necrotic death is crucial for inflammation, an association that has been extended to infection (Upton et al., 2010, 2012), ischemia reperfusion, sepsis, macular degeneration, neurodegenerative disorders, pancreatitis, gastroenteritis, and dermatitis (Kaczmarek et al., 2013). Death receptor-induced necroptosis requires RIP1 kinase activity (Holler et al., 2000) and is inhibited by necrostatin-1 (Nec-1) (Degterev et al., 2005; Degterev and Yuan, 2008), an inhibitor that stabilizes RIP1 in a catalytically inactive conformation and prevents the elaboration of inflammation in disease models (Xie et al., 2013). The viability of RIP1 kinase-dead knockin mice (Berger et al., 2014; Newton et al., 2014) and demonstration that kinase-dead RIP1 alleviates susceptibility to TNF-associated inflammatory disease (Berger et al., 2014) affirm the therapeutic potential of RIP1 kinase inhibitors. Interest in small molecule RIP3 inhibitors is also supported by accumulating evidence that germ line elimination of RIP3 is safe (Newton et al., 2004) and that elimination of this kinase reverses inflammatory disease (Kaczmarek et al., 2013). RIP3 inhibitors have the potential to protect cells from a broader range of stimuli than RIP1 inhibitors (Kaiser et al., 2014; Kaiser et al., 2013a); however, RIP3 kinase-dead knockin mice die prematurely from Casp8-mediated apoptosis (Newton et al., 2014), raising concern that RIP3 kinase plays an essential pro-survival function in addition to its role in necroptosis.

RIP3 has emerged as the common pro-necrotic protein kinase whether death is triggered by TNF superfamily death receptors, DNA activator of interferon (DAI) (Upton et al., 2012), Toll-like receptor (TLR)3 or TLR4 (He et al., 2011; Kaiser et al., 2013a), or interferon (Dillon et al., 2014; Kaiser et al., 2014; Thapa et al., 2013). Insights from death receptor signaling (Murphy and Silke, 2014) have shown that necroptosis depends on the formation of a cytosolic death complex (Complex II or Ripoptosome complex) composed of Casp8, FADD and RIP1, together with the long form of cellular FLICE-like inhibitory protein (cFLIP_L) (Feoktistova et al., 2011; Tenev et al., 2011; Wang et al., 2008). Once activated, Casp8 homodimerization and self-cleavage drives cell-extrinsic apoptosis. When Casp8 activity is compromised, RIP1 and RIP3 undergo RIP homotypic interaction motif (RHIM)-dependent oligomerization that promotes phosphorylation events driving necroptosis (Cho et al., 2009; He et al., 2009; Zhang et al., 2009). This death pathway is naturally suppressed by a murine cytomegalovirus (MCMV)-encoded RHIM signaling competitor, the viral inhibitor of RIP activation (vIRA) (Upton et al., 2010). RIP3 has been ascribed additional functions, such as inducing apoptosis when overexpressed (Kasof et al., 2000; Pazdernik et al., 1999; Sun et al., 1999; Yu et al., 1999), driving Casp8 activation when cIAP1 and cIAP2 E3 ubiquitin ligases are depleted (Vince et al., 2012), and, most recently, driving Casp8-dependent embryonic lethality of RIP3 D161N mutant mice (Newton et al., 2014). It has been difficult to reconcile the divergent roles of RIP3 in suppressing and activating apoptosis with its crucial role as a pro-necrotic kinase.

In this study, we dissect the contribution of RIP3 kinase activity, kinase domain, and RHIM interactions in apoptotic and necroptotic cell death. We previously identified two small molecule inhibitors (Kaiser et al., 2013a) that directly support the contribution of RIP3 in necroptosis (Kaiser et al., 2013b). Here we identify and characterize a third chemically distinct RIP3 kinase inhibitor (RIP3i). At high concentration each of these compounds triggers apoptosis, unveiling a property of RIP3 that reproduces the striking phenotype of RIP3 D161N kinase-dead knockin mice (Newton et al., 2014). Pharmacologic intervention, mutation, and overexpression all transform RIP3 from a pro-necrotic kinase into a pro-apoptotic adapter that nucleates assembly of a Casp8-activation platform. Importantly, RIP3 kinase does not play a vital prosurvival role (Newton et al., 2014) because the kinase-inactive K51A, D143N and D161G mutants do not induce apoptosis. Furthermore, unlike D161N knockin mice, K51A knockin mice develop into fertile and immunocompetent adults, providing a proof-of-concept guide to future refinement of RIP3-based therapies.

Results

Small Molecule RIP3 Kinase Inhibitors Block Necroptosis

Candidate small molecule RIP3 inhibitors were identified with the target of purified, baculovirus-expressed recombinant human RIP3 kinase domain (amino acids 2-328), using binding and kinase inhibition assays. Encoded Library Technology (ELT) employed a set of over 10 ^{total compounds (}Deng et al., 2012) where candidates were selected with the help of a DNA tag and affinity selection. Representative hits were then synthesized without a tag and tested for ability to bind RIP3 kinase domain in a fluorescence polarization (FP) assay using the ATP-competitive fluorescent probe GSK'657 (Figure S1A). RIP3i GSK'840 (Figure 1A) was identified through this effort to bind at subnanomolar concentrations (IC₅₀ = 0.9 nM; Figure 1B). An ADP-glo assay showed that GSK'840 suppressed recombinant RIP3 kinase activity (IC₅₀ = 0.3 nM) (Figure 1C).

Figure 1

Small molecule RIP3 inhibitors block programmed necrosis

(A) Chemical structure of three distinct RIP3 kinase inhibitor (RIP3i) compounds, GSK'840, GSK'843 and GSK'872. (B) Dose response of RIP3i compound binding to recombinant human RIP3 kinase (aa 1-328) assessed by fluorescence polarization. The IC₅₀ was calculated from % maximum binding (mean +/- standard error of the mean [SEM]. (C) Dose response of RIP3i on recombinant human RIP3 kinase activity assessed by ADP-Glo assay. The IC₅₀ was calculated from the response to increasing concentrations of compound (mean +/-range of 4, 4, and 2 replicates of GSK'840, GSK'843 and GSK'872, respectively). (D) Relative viability of human HT-29 cells 24 h post treatment (hpt) with TNF (10 ng/ml), zVAD-fmk (zVAD; 20 μM) and SMAC007 (100 nM) in the presence of increasing concentrations of RIP3i, assessed by determining ATP levels (mean +/- range is shown) compared to cells treated with vehicle (DMSO) alone. (E) Relative viability of murine peritoneal exudate cells (PECs) treated and graphed as described in panel D. (F) Relative viability of necrosis sensitive murine L929 cells treated for 18 h with TNF (25 ng/ml) plus zVAD (25 μM) (left) or primed with interferon (IFN)β (50 U/ml) for 18 h followed by stimulation with poly(I:C) (50 mg/ml) plus indicated concentrations of GSK'872 +/- zVAD (right) and assessed by determining ATP levels (mean +/- SEM) compared to cells treated with vehicle (DMSO) alone. (G) Relative viability of the murine endothelial cell line SVEC4-10 (SVEC) 18 hpt with TNF (25 ng/ml) plus zVAD (25 μM) in increasing concentrations of GSK'872 (left), or 18 hpi with WT or M45mutRHIM MCMV (MOI=10) +/- zVAD (25 μM) in increasing concentrations of GSK'872 (right) assessed as described in Figure 1F. See also Figure S1 and Table S1.

We previously identified the RIP3i GSK'843 and GSK'872 by screening conventional small molecule libraries (Kaiser et al., 2013a) (Figure 1A). These compounds bound RIP3 kinase domain with high affinity (IC₅₀ = 8.6 nM and 1.8 nM, respectively; Figure 1B), and inhibited kinase activity (IC₅₀ = 6.5 nM and 1.3 nM, respectively; Figure 1C). When assayed individually at 1 μM, all three structurally distinct compounds failed to inhibit most of 300 human protein kinases tested, with GSK'840 showing the best profile (Figure S1B, Table S1). All compounds failed to inhibit RIP1 kinase when tested directly (data not shown). Taken together, this demonstrates that GSK'840, GSK'843, and GSK'872 bind to RIP3 kinase domain and inhibit enzyme activity with minimal cross-reactivity.

When evaluated in cell culture using human HT-29 cells, RIP3i compounds blocked TNF-induced necroptosis in a concentration-dependent manner (Figure 1D). In cell–based assays, there was a 100- to 1000-fold shift in the IC₅₀ compared to the cell-free biochemical assays. All of the compounds blocked necroptosis in primary human neutrophils isolated from whole blood (Figure S1C). As previously reported (Kaiser et al., 2013a), GSK'843 and GSK'872 blocked necroptosis in mouse cells. Mouse bone-marrow-derived macrophages (BMDM) or thioglycolate-elicited peritoneal macrophages (PECs), as well as 3T3SA fibroblasts (Figure 1E and S1D) were also protected by RIP3i concentrations ranging from 0.04 to 1 μM. GSK'840 was active in human but inactive in mouse cells, suggesting that species differences dictated the ability of RIP3 to bind this compound. These results confirm the central contribution of RIP3 kinase activity in necroptosis. The ability of RIP3 kinase inhibitors to block death in a wide variety of human and murine cell types reinforces the prospect of employing such inhibitors for therapeutic intervention in inflammatory diseases.

RIP3i GSK'843 or GSK'872 inhibit TLR3- or DAI-induced death (Kaiser et al., 2013a), two RIP1-independent pathways of necroptosis. TLR3-induced necroptosis was triggered by poly(I:C) in the presence of zVAD (Figure 1F) and DAI-induced necroptosis was triggered by M45mutRHIM MCMV (Figure 1G). Consistent with previous findings (Kaiser et al., 2013a; Upton et al., 2010, 2012), RIP1 inhibitors Nec-1 and GSK'963 (Weng et al., 2014) were ineffective against either of these pathways but, as expected, blocked TNF-induced necroptosis (Figure S1E and S1F). Thus, RIP3 kinase inhibitors prevent death from a broader range of stimuli than RIP1 inhibitors.

RIP3i-induced Apoptosis

We next evaluated cytotoxicity of RIP3i compounds in different cell types. Treatment with GSK'843 or GSK'872 at either 3 μM or 10 μM uniformly triggered cell death that was reversed by zVAD (Figure 2A). When cell death was monitored microscopically, TNF-induced necroptosis was completely blocked by GSK'872 (Figure 2B right); however, cells treated in the absence of zVAD followed a time course similar to cells undergoing TNF plus CHX-induced apoptosis (Figure 2B left). The level of RIP3i-associated death was also comparable when propidium iodide (PI) uptake was evaluated using flow cytometry (Figure S2A). Cleaved Casp3 accumulated within 1.5 hpt and was strongly detected by 2.5 hpt with GSK'872 (Figure 2C). Flow cytometry showed that 29% of cells had Cl-Casp3 by 2.5 hpt (Figure S2B), in line with the proportion of PI positive cells at 3 hpt (Figure S2A) and increased activity of effector caspases (Figure 2D) as well as membrane blebbing (Supplemental movies S1 and S2) at that time. In contrast, TNF-induced necroptosis was associated with expected cell swelling and loss of membrane integrity without blebbing (Supplemental movie S3). Transmission electron microscopy confirmed apoptotic cell morphology associated with RIP3i cytotoxicity (Figure 2E), establishing that GSK'843 and GSK'872 induce caspase activation followed by apoptotic cell death.

Figure 2

RIP3i-induced apoptosis

(A) Relative viability of SVEC, L929, 3T3SA and mouse embryo fibroblast (MEF) 18 hpt with increasing concentrations of GSK'872 (black bars) or GSK'843 (grey bars) +/-zVAD (25 μM), assessed as described in Figure 1F. (B) Time course cell viability analysis of 3T3SA cells measuring permeability to the nucleic acid detecting stain SYTOX Green (50 nM) using an IncuCyte instrument. Cells were treated with GSK'872 (10 μM) in the +/- zVAD or with TNF (25 ng/ml) plus cyclohexamide (CHX; 5 μg/ml) (left), or with TNF and zVAD +/- GSK'872 (right). (C) Immunoblot (IB) for Casp3 and Casp3 cleavage products (Cl-Casp3) in 3T3SA cells at the indicated hpt with GSK'872 +/- zVAD. (D) Time course of Casp3/Casp7 proteolytic activity (DEVDase) in 3T3SA cells treated with GSK'872 +/-zVAD. (E) Transmission electron microscopy images of 3T3SA cells treated with DMSO or GSK'872 for 2.5 h. See also Figure S2 and Supplemental Movies.

RIP3i-induced On-target Apoptosis

To determine whether apoptosis depended on RIP3, we first assessed susceptibility of cells known to differ in levels of this kinase. Murine NIH3T3 cells, a necroptosis-resistant fibroblast line with low levels of RIP3 (Upton et al., 2010), were less susceptible to GSK'872-induced apoptosis than 3T3SA cells, a necroptosis-sensitive fibroblast line with high levels of this kinase (Figure 3A and B). To further investigate the contribution of RIP3, we employed knock-down strategies in 3T3SA, SVEC, and L929 cells (Figure 3C, 3D, and S2C) and also evaluated Rip3^-/- MEF (Figure 3E). In all of these settings, RIP3i-induced apoptosis required the presence of RIP3. The RIP3 pro-necrotic partner MLKL was subjected to knockdown in 3T3SA cells and shown to be dispensable for apoptosis (Figure 3F and data not shown). Thus, RIP3i promotes the concentration-dependent ability of RIP3 to trigger caspase activation and apoptotic cell death completely independent of necroptosis machinery.

Figure 3

Concentration-dependent apoptosis of GSK'840, GSK'843 and GSK'872 requires RIP3

(A) Relative viability of NIH3T3 cells (left) 18 hpt with increasing concentrations of GSK'872 (black bars) or GSK'843 (grey bars), or 3T3SA cells (right) in 10 mM RIP3i compounds, +/- zVAD, assessed as described in Figure 1F. (B) IB showing RIP3 and β-actin levels in NIH3T3 and 3T3SA cells. (C) Relative viability at 18 hpt with GSK'872 (10 mM) in 3T3SA, SVEC and L929 cells, transduced with non-targeting (NT) shRNA (black bars) or RIP3-specific (grey bars) shRNA. (D) Analysis of Casp3/Casp7 proteolytic activity (DEVDase) in transduced 3T3SA cells at 4 hpt with GSK'843, GSK'872 or TNF plus CHX. (E) Relative viability comparing WT (black bars) and Rip3^-/- (grey bars) MEF at 18 hpt with GSK'872 +/- zVAD. (F) Relative viability comparing 3T3SA cells transfected with NT or MLKL-specific siRNA 18 hpt with GSK'872 +/- zVAD or with TNF plus zVAD +/- GSK'872. An IB inset (right) show the levels of MLKL prior to any other treatment. (G) Relative viability comparing Rip3^-/- MEF alone and after transduction with human hRIP3 and hRIP3mutRHIM 18 hpt post treatment with GSK'840, GSK'843 or GSK'872, or with a combination of TNF (25 ng/ml), zVAD (25 μM) and BV6 (0.5 μM).

To investigate the properties of human-specific RIP3i, GSK'840, we reconstituted Rip3 ^{MEF with human RIP3 and conferred sensitivity to RIP3i GSK'840 as well as GSK'843 and GSK'872 (}Figure 3G). This data demonstrates the capacity of human RIP3 to promote apoptosis in the presence of high concentrations of RIP3 kinase inhibitor. These data also reveal that the chemically distinct RIP3i GSK'840 behaved the same as GSK'843 and GSK'872 in a mouse cell environment. In this setting, human RIP3 cannot confer sensitivity to necroptosis because it cannot interact with mouse MLKL (Sun et al., 2012), reinforcing observations that MLKL is dispensable for RIP3i-induced apoptosis. Cell death was nevertheless dependent on RHIM-signaling such that human RIP3mutRHIM failed to confer sensitivity to apoptosis (Figure 3G). Thus, for apoptosis, the trigger seems to be RIP3 itself.

Identification of Genes Essential for RIP3-induced Apoptosis

The impact of RIP3i seemed reminiscent of RIP3 overexpression-induced apoptosis (Kasof et al., 2000; Pazdernik et al., 1999; Sun et al., 1999; Yu et al., 1999), leading us to perform a genome-wide loss-of-function screen in human haploid HAP1 cells to identify genes that contribute to RIP3-initiated apoptosis. This method has previously identified mediators of cell death (Carette et al., 2009). HAP1 cells, which lack RIP3, were engineered (HAP1-Tet-RIP3) to express RIP3 following doxycycline (DOX) induction (Figure S3A) and subsequently died by apoptosis as RIP3 levels increased (Figure S3B). A retroviral gene-trap vector strategy was used to select death-resistant mutants and genes disrupted by insertion were mapped using deep sequencing. We identified a set of genes significantly enriched for insertions within the DOX-selected population compared to the unselected control, grouped categorically according to predicted function (Figure 4A), where the frequency of insertions within Rip3 itself provided confidence that the screen was specific. Hits within VP16-responsive Mediator (MED) transcription complex genes confirmed reliance of the expression system on this transactivator (Uhlmann et al., 2007). A subset of genes involved in transcription and chromatin remodeling (e.g., RPRD2, SP1, ZCCHC14) were also identified, though the mechanism by which they may contribute to RIP3-mediated apoptosis is currently unclear. RIP1, FADD, cFLIP_L and Casp8, were all implicated, consistent with the contribution of Ripoptosome-like machinery in death. Additionally, the screen also identified the Casp8 substrate Bid, implicating mitochondrial amplification machinery as necessary for RIP3-driven apoptosis.

Figure 4

Haploid genetic screen for genes involved in RIP3-mediated cell death

(A) Haploid screen results depicting each gene as a bubble where size corresponds to the number of independent gene trap insertions (also indicated in parentheses) the significance of enrichment is plotted on the y-axis. The top 25 most significantly enriched genes are labeled, colored and horizontally grouped by function (other genes are grey and in arbitrary position along the x-axis). (B) IB of the indicated components from anti-FADD immunoprecipitation of Triton X-solubilized 3T3SA cell supernatants (IP: FADD; top section), together with total supernatant (middle section) and total pellet (bottom section) fractions. Cells were assessed without treatment (untreated; lane 1), as well as 2.5 hpt with vehicle alone (DMSO; lane 2), zVAD (25 μM) alone (lane 3) and varying concentrations (10, 3, 1 and 0.3 mM) of GSK'872 in the presence (lanes 4 through 7) or absence (lanes 8 through 11) of zVAD. Single asterisk indicates modification of RIP1 in the presence of GSK'872 when caspases are active (lane 8) and double asterisks indicate slower migrating forms of RIP3 in the pellet fraction (lanes 4, 5 and 8). See also Figure S3.

Requirement for RIP1, FADD, cFLIP_L and Casp8 in RIP3i-induced Apoptosis

To determine whether the Casp8-activation platform plays a direct role in apoptosis induced by RIP3i, we employed inhibitors with different specificities. zVAD as well as the Casp8-specific inhibitor, zIETD, blocked death; whereas, neither Casp1- nor Casp9-specific inhibitors (zYVAD and zLEHD, respectively) had any impact (Figure 5A). Neither the reactive oxygen species scavenger BHA nor the autophagy inhibitor 3-MA blocked apoptosis. Consistent with this pattern, Cl-Casp8 accumulated in parallel with IETDase activity (Figure 5B and C) and knock-down of Casp8 prevented death and eliminated the accumulation of cleaved Casp3 forms (Figure 5D and S4A). Similarly, FADD (Figure 5E) and RIP1 (Figure 5F and S4B) were both necessary, but with an important distinction from Complex II formation in TNF signaling (Wang et al., 2008) where RIP1 kinase-mediated phosphorylation events predominate. Here, RIP1 kinase activity was completely dispensable. Neither Nec-1 (Figure S4C and data not shown) nor kinase-dead RIP1 (Rip1^K45A/K45A) MEF (Berger et al., 2014; Kaiser et al., 2014) prevented apoptosis (Figure 5G) even though necroptosis was blocked. Consistent with these findings, MEF from mice with RIP1, FADD, cFLIP_L or Casp8 deficiency resisted apoptosis (Figure 5H, 5I, 5J, and S4D). The requirement of cFLIP_L for RIP3i-induced apoptosis was particularly surprising given the acknowledged protective role of cFLIP_L in extrinsic apoptosis. To confirm the contribution of cFLIP_L to RIP3i-induced apoptosis, reconstitution of cFLIP_L-null cells restored sensitivity to RIP3i and, as expected, protected cells from TNF-induced apoptosis (Figure 5J, 5K, and S4E). This is consistent with the requirement of cFLIP_L for RIP3-initiated apoptosis (Figure 4A). Furthermore, cFLIP_L was essential for RIP3i-induced recruitment of RIP1 to FADD (Figure S4F). Thus, RIP3 recruits components and acts as a scaffold or adaptor to drive assembly of a Ripoptosome death complex dependent on cFLIP_L but independent of RIP1 or RIP3 pro-necrotic kinase activities.

Figure 5

Requirement for Casp8, FADD and RIP1 in RIP3-initiated apoptosis

(A) Relative viability of 3T3SA cells 18 hpt with GSK'872 in the absence (DMSO) or presence of caspase inhibitors (zVAD, zYVAD zLEHD, zIETD; 25 μM), an inhibitor of autophagy, 3-methyladenine (3-MA; 5mM), or an inhibitor of reactive oxygen species, butylated hydroxyanisole (BHA; 25 μM), assessed as described in Figure 1F. (B) Analysis of Casp8 proteolytic activity (IETDase) in 3T3SA cells treated with GSK'872 for the indicated times +/- zVAD. (C) IB of Cl-Casp8 in 3T3SA cells treated with GSK'872 for the indicated times +/- zVAD. (D) Relative viability of 3T3SA cells transfected with NT (black bars) or Casp8-specific (grey bars) siRNA and treated as described in panel C. An IB inset on the right shows the level of Casp8 knockdown. (E) Relative viability of 3T3SA cells transduced with EV (black bars) or FADD-DN-expressing retrovirus vector (grey bars) 18 hpt with GSK'872, zVAD and/or TNF, as indicated. (F) Relative viability of 3T3SA cells transfected with NT (black bars) or RIP1-specific siRNA (grey bars) 18 hpt with GSK'872, zVAD and/or TNF. An IB inset on the right shows level of RIP1 knockdown. (G) Relative viability of WT (black bars) or Rip1^K45A/K45A mutant MEF (grey bars) 18 hpt with GSK'872 +/- zVAD. (H) Relative viability of different MEF genotypes 18 hpt with GSK'872. (I) Relative viability of primary WT or cFLIP^-/- MEF treated with GSK'872, TNF, zVAD and/or BV6. (J) Relative viability of cFLIP^-/- MEF transduced with EV or cFLIP_L-expressing retrovirus 18 hpt with GSK'872 or TNF. (K) Analysis of Casp8 proteolytic activity (IETDase) in cFLIP_L-deficient MEFs 4 hpt with GSK'872 and TNF. (L) Relative viability in SVEC cells transduced with EV (white bar) or M45-expressing retrovirus (black bars), or M45mutRHIM-expressing retrovirus (grey bars) 18 hpt with GSK'872 alone or TNF plus zVAD. See also Figure S4.

To determine the extent to which RHIM signaling contributes to RIP3-initiated apoptosis, we showed that MCMV M45-encoded vIRA (Upton et al., 2010) suppressed; whereas, a nonfunctional RHIM mutant (M45mutRHIM) failed to suppress RIP3i-induced apoptosis (Figure 5L). WT-MCMV, expressing both a functional M45 (vIRA) and the Casp8 inhibitor M36 (vICA), completely prevented GSK'872-induced apoptosis (Figure S4G), so long as cells were infected for at least 2 h before RIP3i compound was added. RIP3i induced RHIM-dependent NF-κB activation (Figure S4H and S4I) similar to RIP3 overexpression settings (Kaiser et al., 2008; Rebsamen et al., 2009); however, NF-κB inhibition had no impact on death (Figure S4J). Furthermore, TNF-deficient MEF remained sensitive to RIP3i (Figure S4K). Apoptosis therefore occurs independent of TNF and NF-κB, but relies on RIP3 RHIM-mediated recruitment of RIP1, resulting in Casp8 activation.

RIP3 Kinase Domain Mutants in RIP3-Initiated Apoptosis

To directly address the recently proposed (Newton et al., 2014) prosurvival role of RIP3 kinase, we reconstituted Rip3^-/- cells with a series of nonfunctional RIP3 mutants that were either non-toxic when stably expressed or toxic due to the conversion of RIP3 into an autoinducer of apoptosis (Figure 6A). Neither the kinase ATP binding pocket mutant K51A nor phosphorylation site mutant T231A/S232A or S204A/S207A/S211A induced apoptosis when introduced into cells, a pattern similar to the impact of MEF transduced with WT RIP3 (Figure 6B). Cells expressing the kinase domain mutant D161N could only be cultured when caspase inhibitor was included in the medium (Figure 6A). As expected, cells expressing kinase-inactive mutant, K51A, were unable to support necroptosis (Figure 6B, 6C and data not shown). Thus, the pro-apoptotic behavior of kinase-inactive mutant D161N was unique, seemingly paralleling high concentration RIP3i or RIP3 overexpression.

Figure 6

Characterization of RIP3 kinase domain mutants

(A) Table summarizing microscopic assessment of Rip3^-/- MEF survival 3 days after transduction with WT RIP3 or indicated mutants +/-zVAD. (B) Relative viability of transduced cells 18 h post treatment with TNF, zVAD and GSK'872 as indicated and assessed as described in Figure 1F. (C) IB for FLAG-RIP3 levels in transduced cells. (D) Relative viability of RIP3shRNA transduced 3T3SA cells following transfection with shRNA-resistant RIP3 or indicated RIP3 mutants for 16 h and treated with GSK'872, zVAD, and TNF. (E) IB of the indicated components from RIP3shRNA transduced 3T3SA cells showing anti-FLAG IP, supernatant, and pellet fractions 8 h post transfection with RIP3 or mutants (K51A, D161N) +/- GSK'872 for 2 h prior to harvesting. (F) IB of the indicated components in unmanipulated supernatants and pellets of RIP3shRNA transduced 3T3SA cells 16 h after transfection with shRNA RIP3 or mutants (K51A, D161N) +/- GSK'872 without or with zVAD for 3 h prior to harvesting. See also Figure S5.

RIP3i compound triggered apoptosis in K51A mutant-transduced cells (Figure 6B), comparable to WT RIP3, indicating that compound binding rather than kinase inhibition promoted cell death. Mutant T231A/S232A, which prevents pro-necrotic MLKL binding (Sun et al., 2012), as well as a triple mutant (S204A/S207A/S211A) in the activation loop, also supported apoptosis. As expected, necrotoposis required MLKL interaction and both apoptosis and necroptosis required RHIM-signaling (Figure 6B). Overall, these data preclude a pro-survival role of RIP3 kinase activity but are consistent with a kinase-independent ability of RIP3 to recruit RIP1 and drive FADD-cFLIP_L-Casp8 death complex activation.

The striking difference in viability of cells stably transduced with these two RIP3 kinase-inactive mutants prompted further investigation. RIP3 shRNA knock-down 3T3SA cells were transiently transfected with shRNA-resistant WT RIP3, RIP3-kinase-inactive mutants (K51A, D161N, D161G and D143N), phosphorylation mutant (T257A), and a D161N/mutRHIM double-mutant. Of the mutant RIP3 forms evaluated, only the previously characterized (Newton et al., 2014) charge swap mutant D161N promoted apoptosis (Figure 6D). All RIP3 forms were expressed at similar levels (Figure S5A), so it was striking to observe cells survived when a neutral aa substitution (D161G) was made at this same position (Figure 6D). These data implicate a dramatic impact of charge reversal in producing the toxic effects. Cells also survived when a different kinase-inactive charge swap mutant (D143N) was evaluated, suggesting a particular consequence of D161N and not a general alteration within the active site. As expected, cells expressing any of the nontoxic RIP3 kinase-inactive mutants were resistant to TNF-induced necroptosis. Only WT and T257A RIP3 conferred sensitivity to either TNF+zVAD or zVAD alone, the latter a consequence of high RIP3 levels produced by transfection, and all cells survived when treated with a combination of GSK'872 and zVAD. Similar to WT RIP3, cells expressing mutants K51A, D161G, D143N or T257A were sensitive to high concentration GSK'872 but cells expressing mutRHIM or EV were not. The inherent toxicity of D161N was neither blocked nor enhanced by addition of RIP3i compound and a D161N/mutRHIM double mutant showed expected viability (Figure S5B), confirming the role of RHIM signal transduction in death. Furthermore, D161N- or RIP3i-induced apoptosis was prevented when either Casp8-inhibitor vICA or RHIM antagonist vIRA was present (Figure S5C). As expected, neither vICA nor M45mutRHIM suppressed necroptosis (Kaiser et al., 2014; Kaiser et al., 2011). These data challenge the notion that RIP3 kinase activity has any inherent role promoting survival, but rather demonstrate the conversion of RIP3 into a pro-apoptotic adapter by specific mutation (D161N) through a mechanism that requires RHIM signaling and leads to Casp8 activation, analogous to that triggered by high concentration RIP3i or overexpression.

Next, we evaluated the role of RIP1 in D161N-induced apoptosis. D161N associated with RIP1 and Casp8 spontaneously in cell extracts (Figure 6E), in contrast to the pattern of WT or K51A mutant RIP3 where association was dependent on RIP3i. Doxycycline-induced expression of D161N RIP3 in RIP3-deficient MEF triggered RIP1 association with FADD and RIP3, a RHIM-dependent pattern that was only observed with WT RIP3 and K51A in the presence of RIP3i (Figure S5D). D161N enhanced the accumulation of RIP3 in the pellet fraction as early as 12 h post transfection (Figure S5E); whereas, WT, K51A, D143N, and D161G RIP3 failed to accumulate in this fraction unless treated with RIP3i and zVAD (Figure 6F and S5E). Consistent with a role for RHIM-signaling in oligomerization, aggregated forms of RIP3 did not appear in the pellet when RHIM signaling was blocked (data not shown). D161N mutation or RIP3i-binding therefore alters the conformation of RIP3 and triggers RHIM-dependent oligomerization reminiscent of aggregate, “amyloid-like” RIP1-RIP3 complexes (Li et al., 2012).

Rip3^K51A/K51A Kinase-dead Knockin Mice Are Viable

The behavior of RIP3 kinase-dead mutants supported the striking midgestational lethality observed in D161N mutant knockin mice (Newton et al., 2014) and predicted the opposite outcome would occur with a nontoxic mutant. When generated, Rip3^/K51A kinase-dead knockin mice were clearly viable and fertile (Figure 7A and B). This mutant strain did not show any susceptibility to midgestational or perinatal death. To determine whether the viable Rip3^/K51A mutant, like the lethal Rip3^D161N/D161N mutant (Newton et al., 2014), rescues embryonic lethality of Casp8^-/- embryos (Sakamaki et al., 2002; Varfolomeev et al., 1998), we performed a cross and rescued viable and fertile Casp8^{Rip3K51A/K51A} mice at the expected Mendelian frequency (Figure 7B and S6A). This extends previous rescue of Casp8^Rip3-/-mice (Kaiser et al., 2011; Oberst et al., 2011; Zhang et al., 2011) to clearly show the contribution of pro-necrotic RIP3 enzymatic activity in midgestational death of Casp8-deficient embryos without the complications of the Rip3^D161N/D161N mutant (Newton et al., 2014).

Figure 7

Characterization of RIP3 kinase inactive mice

(A) Schematic representation of the WT, recombined, Cre-mediated excised, Flp-mediated excised RIP3 alleles with the relevant restriction sites for the DNA blot analysis. (B) Photograph of age-matched WT, Rip3^K51A/K51A and Rip3^Casp8-/-mice. (C) Relative viability of WT and Rip3^K51A/K51A MEF 18 hpt with GSK'872, TNF, zVAD and BV6 as indicated and as assessed in 1F. (D) Relative viability of WT and Rip3^K51A/K51A mutant BMDM 18 hpt with GSK'872, TNF, zVAD, BV6, poly (I:C)/zVAD, or LPS/zVAD for 18 h. (E) Relative viability of WT (left) and Rip3^K51A/K51A (right) MEF transfected with either NT siRNA, RIP1 siRNA, or a combination of RIP1 and RIP3 siRNA for 48 h and then 48 hpt with IFNβ (100 U/ml), zVAD, and GSK'872 as indicated. (F) MCMV titers in spleen (left) and liver (right) from mice of indicated genotypes 3 days post infection. (G) Recall response to MCMV infection. Total numbers (left) or percentages (right) of splenic CD8+ T cells producing IFNγ or IFNγ and TNFα following stimulation with M45-specific peptide 4 days post challenge with RM427 virus 14 days post infection with MCMV-M45mutRHIM in mice of indicated genotypes. See also Figure S6.

RIP3 K51A mutant protein was detected in tissues and cells from adult mice, albeit at levels that were lower than WT RIP3 (Figure S6B). When levels of mutant RIP3 in splenic cells from Casp8^{Rip3K51A/K51A} were directly compared to Casp8^{Rip3D161N/D161N} mice, RIP3-K51A and RIP3-D161N expression was comparable (Figure S6C) although lower than a Rip3^+/- spleen. Even though Rip3^D161N/D161N mice have an obvious phenotype, these low levels raise the possibility that Rip3^K51A/K51A kinase-dead mutants simply phenocopied Rip3 ^{mice. To address the biologic potential of K51A mutant protein in these mice, we evaluated MEF and BMDM for the ability to sustain RIP3i-induced apoptosis. Consistent with the behavior of transduced and transfected cells,} Rip3^K51A/K51A mutant cells remained as sensitive as WT cells to RIP3i compound-induced apoptosis (Figure 7C, D, and E). To address the ability of these cells to support inflammasome activation (Vince et al., 2012), Rip3^K51A/K51A BMDMs treated with GSK'872 and IAP agonist SMAC007 produced IL-1β at levels similar to WT and much higher than Rip3^-/-cells (Figure S6D). Thus, K51A mutant RIP3 adapter function is sufficient for inflammasome activation, a result that demonstrates this pathway does not involve kinase activity. As expected, WT MEF and BMDM generated from Rip3^K51A/K51A mice were resistant to necroptosis induced by TNFR1, TLR3, and TLR4 ligands and IFNβ (Figure 7C, 7D, 7E, S6E, and S6F). These data dismiss any vital contribution of RIP3 kinase activity in vivo (Newton et al., 2014) and, despite low expression levels, reveal that Rip3^K51A/K51A mice sustain adapter-dependent processes such as apoptosis and inflammasome activation, consistent with a proposed adaptor function underlying midgestational apoptosis in Rip3^D161N/D161N mouse development.

In addition to controlling inflammation, RIP3 plays an innate immune role in limiting viral infection. To determine the ability of Rip3^K51A/K51A mice to mount an immune response to MCMV, a natural mouse pathogen, we compared WT virus to MCMV-M45mutRHIM, a mutant that fails to replicate in WT mice due to elimination of virally infected cells by the DAI-RIP3 pathway (Upton et al., 2010, 2012). Infection with MCMV-M45mutRHIM was completely normalized in Rip3^K51A/K51A mice as in Rip3^-/- mice, demonstrating the contribution of pro-necrotic RIP3 kinase activity to death of mutant virus-infected cells in vivo (Figure 7F). M45mutRHIM infection of BMDMs from WT mice died by necroptosis; whereas, Rip3^K51A/K51A BMDMs survived (data not shown). These data directly demonstrate the importance of RIP3 kinase in the DAI-driven necroptosis pathway. Our previous investigation showed that M45mutRHIM kills cultured cells independent of RIP1 (Upton et al., 2012). Here, we extended observations in both Rip1^K45A/K45A and Rip1^Rip3Casp8-/- mice (Figure 7F and S6G), providing definitive evidence that neither RIP1 nor its kinase activity is involved in DAI-RIP3 necroptosis that severely attenuates mutant virus replication.

We previously established that WT, Rip3^-/-, Casp8^Rip3-/-, and Casp8^Rip3Rip1-/- mice all control WT MCMV infection, mounting a robust virus-specific T cell response despite the absence of extrinsic death pathways (Kaiser et al., 2014; Kaiser et al., 2011). To study the adaptive response to MCMV, Rip3^K51A/K51A mice were infected with M45mutRHIM for 14 days and boosted by secondary infection with a lacZ-expressing MCMV, conditions that drive a maximal T cell response. Four days after boosting, infected WT, Rip3^-/-, and Rip3^K51A/K51A mice all exhibited comparable numbers of total splenic T cells as well as total CD4 and CD8 T cells (Figure S6G). Importantly, the MCMV M45 epitope-specific CD8 T cell response remained very robust across the genotypes (Figure S6G), rising more than 1,000-fold above levels in naïve mice. This pattern was similar to WT virus infection (Kaiser et al., 2014; Kaiser et al., 2011), reinforcing the fact that extrinsic cell death pathways are dispensable for a robust adaptive immune response. A single exposure to MCMV for four days failed to induce the dramatic response seen in previously infected mice (data not shown). Figure 7F and G show levels of virus-specific CD8 T cells in settings where mutant virus replicates to high levels (Rip3^-/- or Rip3^K51A/K51A mice) were comparable to settings where this virus drives necroptosis and replicates poorly (WT mice). Thus, the induction of immunity assessed as M45 peptide-specific IFNγ ^{and INFγTNF CD8 T cell frequencies did not require sustained infection so long as necroptosis was triggered by the virus (}Figure 7G). RIP3 and its kinase activity are dispensable for robust antiviral immunity in settings where virus replication proceeds, undoubtedly because antigen load compensates when cell death is blocked. Thus, Rip3^K51A/K51A mice are viable and fertile, as well as immunocompetent. Collectively, these data demonstrate that RIP3 kinase activity can be eliminated in vivo without the dire consequence of triggering lethal apoptosis or even compromising immune competence.

Small Molecule RIP3 Kinase Inhibitors Block Necroptosis

Candidate small molecule RIP3 inhibitors were identified with the target of purified, baculovirus-expressed recombinant human RIP3 kinase domain (amino acids 2-328), using binding and kinase inhibition assays. Encoded Library Technology (ELT) employed a set of over 10 ^{total compounds (}Deng et al., 2012) where candidates were selected with the help of a DNA tag and affinity selection. Representative hits were then synthesized without a tag and tested for ability to bind RIP3 kinase domain in a fluorescence polarization (FP) assay using the ATP-competitive fluorescent probe GSK'657 (Figure S1A). RIP3i GSK'840 (Figure 1A) was identified through this effort to bind at subnanomolar concentrations (IC₅₀ = 0.9 nM; Figure 1B). An ADP-glo assay showed that GSK'840 suppressed recombinant RIP3 kinase activity (IC₅₀ = 0.3 nM) (Figure 1C).

Figure 1

Small molecule RIP3 inhibitors block programmed necrosis

(A) Chemical structure of three distinct RIP3 kinase inhibitor (RIP3i) compounds, GSK'840, GSK'843 and GSK'872. (B) Dose response of RIP3i compound binding to recombinant human RIP3 kinase (aa 1-328) assessed by fluorescence polarization. The IC₅₀ was calculated from % maximum binding (mean +/- standard error of the mean [SEM]. (C) Dose response of RIP3i on recombinant human RIP3 kinase activity assessed by ADP-Glo assay. The IC₅₀ was calculated from the response to increasing concentrations of compound (mean +/-range of 4, 4, and 2 replicates of GSK'840, GSK'843 and GSK'872, respectively). (D) Relative viability of human HT-29 cells 24 h post treatment (hpt) with TNF (10 ng/ml), zVAD-fmk (zVAD; 20 μM) and SMAC007 (100 nM) in the presence of increasing concentrations of RIP3i, assessed by determining ATP levels (mean +/- range is shown) compared to cells treated with vehicle (DMSO) alone. (E) Relative viability of murine peritoneal exudate cells (PECs) treated and graphed as described in panel D. (F) Relative viability of necrosis sensitive murine L929 cells treated for 18 h with TNF (25 ng/ml) plus zVAD (25 μM) (left) or primed with interferon (IFN)β (50 U/ml) for 18 h followed by stimulation with poly(I:C) (50 mg/ml) plus indicated concentrations of GSK'872 +/- zVAD (right) and assessed by determining ATP levels (mean +/- SEM) compared to cells treated with vehicle (DMSO) alone. (G) Relative viability of the murine endothelial cell line SVEC4-10 (SVEC) 18 hpt with TNF (25 ng/ml) plus zVAD (25 μM) in increasing concentrations of GSK'872 (left), or 18 hpi with WT or M45mutRHIM MCMV (MOI=10) +/- zVAD (25 μM) in increasing concentrations of GSK'872 (right) assessed as described in Figure 1F. See also Figure S1 and Table S1.

We previously identified the RIP3i GSK'843 and GSK'872 by screening conventional small molecule libraries (Kaiser et al., 2013a) (Figure 1A). These compounds bound RIP3 kinase domain with high affinity (IC₅₀ = 8.6 nM and 1.8 nM, respectively; Figure 1B), and inhibited kinase activity (IC₅₀ = 6.5 nM and 1.3 nM, respectively; Figure 1C). When assayed individually at 1 μM, all three structurally distinct compounds failed to inhibit most of 300 human protein kinases tested, with GSK'840 showing the best profile (Figure S1B, Table S1). All compounds failed to inhibit RIP1 kinase when tested directly (data not shown). Taken together, this demonstrates that GSK'840, GSK'843, and GSK'872 bind to RIP3 kinase domain and inhibit enzyme activity with minimal cross-reactivity.

When evaluated in cell culture using human HT-29 cells, RIP3i compounds blocked TNF-induced necroptosis in a concentration-dependent manner (Figure 1D). In cell–based assays, there was a 100- to 1000-fold shift in the IC₅₀ compared to the cell-free biochemical assays. All of the compounds blocked necroptosis in primary human neutrophils isolated from whole blood (Figure S1C). As previously reported (Kaiser et al., 2013a), GSK'843 and GSK'872 blocked necroptosis in mouse cells. Mouse bone-marrow-derived macrophages (BMDM) or thioglycolate-elicited peritoneal macrophages (PECs), as well as 3T3SA fibroblasts (Figure 1E and S1D) were also protected by RIP3i concentrations ranging from 0.04 to 1 μM. GSK'840 was active in human but inactive in mouse cells, suggesting that species differences dictated the ability of RIP3 to bind this compound. These results confirm the central contribution of RIP3 kinase activity in necroptosis. The ability of RIP3 kinase inhibitors to block death in a wide variety of human and murine cell types reinforces the prospect of employing such inhibitors for therapeutic intervention in inflammatory diseases.

RIP3i GSK'843 or GSK'872 inhibit TLR3- or DAI-induced death (Kaiser et al., 2013a), two RIP1-independent pathways of necroptosis. TLR3-induced necroptosis was triggered by poly(I:C) in the presence of zVAD (Figure 1F) and DAI-induced necroptosis was triggered by M45mutRHIM MCMV (Figure 1G). Consistent with previous findings (Kaiser et al., 2013a; Upton et al., 2010, 2012), RIP1 inhibitors Nec-1 and GSK'963 (Weng et al., 2014) were ineffective against either of these pathways but, as expected, blocked TNF-induced necroptosis (Figure S1E and S1F). Thus, RIP3 kinase inhibitors prevent death from a broader range of stimuli than RIP1 inhibitors.

Cytotoxicity of RIP3 Kinase Inhibitors

In the absence of zVAD, inhibition of necroptosis with these compounds produced a starkly different outcome. TLR3- or DAI-dependent necroptosis proceeds efficiently even in the absence of Casp8 inhibition (Kaiser et al., 2013a). Although TLR3-induced necroptosis was modestly blocked by 1 μM GSK'872, there was no additional protection at higher concentrations (Figure 1F), contrasting the result when caspases were inhibited. In combination, poly(I:C) and zVAD restored cell viability, indicating a contribution of caspase activity to RIP3i compound toxicity. Likewise, the viability of cells infected with either WT or pro-necrotic mutant virus declined dramatically in the presence of 10 μM GSK'872 and was reversed by addition of zVAD (Figure 1G). These data reveal that RIP3 kinase inhibitors unleash an unexpected caspase-dependent cytotoxicity.

RIP3i-induced Apoptosis

We next evaluated cytotoxicity of RIP3i compounds in different cell types. Treatment with GSK'843 or GSK'872 at either 3 μM or 10 μM uniformly triggered cell death that was reversed by zVAD (Figure 2A). When cell death was monitored microscopically, TNF-induced necroptosis was completely blocked by GSK'872 (Figure 2B right); however, cells treated in the absence of zVAD followed a time course similar to cells undergoing TNF plus CHX-induced apoptosis (Figure 2B left). The level of RIP3i-associated death was also comparable when propidium iodide (PI) uptake was evaluated using flow cytometry (Figure S2A). Cleaved Casp3 accumulated within 1.5 hpt and was strongly detected by 2.5 hpt with GSK'872 (Figure 2C). Flow cytometry showed that 29% of cells had Cl-Casp3 by 2.5 hpt (Figure S2B), in line with the proportion of PI positive cells at 3 hpt (Figure S2A) and increased activity of effector caspases (Figure 2D) as well as membrane blebbing (Supplemental movies S1 and S2) at that time. In contrast, TNF-induced necroptosis was associated with expected cell swelling and loss of membrane integrity without blebbing (Supplemental movie S3). Transmission electron microscopy confirmed apoptotic cell morphology associated with RIP3i cytotoxicity (Figure 2E), establishing that GSK'843 and GSK'872 induce caspase activation followed by apoptotic cell death.

Figure 2

RIP3i-induced apoptosis

(A) Relative viability of SVEC, L929, 3T3SA and mouse embryo fibroblast (MEF) 18 hpt with increasing concentrations of GSK'872 (black bars) or GSK'843 (grey bars) +/-zVAD (25 μM), assessed as described in Figure 1F. (B) Time course cell viability analysis of 3T3SA cells measuring permeability to the nucleic acid detecting stain SYTOX Green (50 nM) using an IncuCyte instrument. Cells were treated with GSK'872 (10 μM) in the +/- zVAD or with TNF (25 ng/ml) plus cyclohexamide (CHX; 5 μg/ml) (left), or with TNF and zVAD +/- GSK'872 (right). (C) Immunoblot (IB) for Casp3 and Casp3 cleavage products (Cl-Casp3) in 3T3SA cells at the indicated hpt with GSK'872 +/- zVAD. (D) Time course of Casp3/Casp7 proteolytic activity (DEVDase) in 3T3SA cells treated with GSK'872 +/-zVAD. (E) Transmission electron microscopy images of 3T3SA cells treated with DMSO or GSK'872 for 2.5 h. See also Figure S2 and Supplemental Movies.

RIP3i-induced On-target Apoptosis

To determine whether apoptosis depended on RIP3, we first assessed susceptibility of cells known to differ in levels of this kinase. Murine NIH3T3 cells, a necroptosis-resistant fibroblast line with low levels of RIP3 (Upton et al., 2010), were less susceptible to GSK'872-induced apoptosis than 3T3SA cells, a necroptosis-sensitive fibroblast line with high levels of this kinase (Figure 3A and B). To further investigate the contribution of RIP3, we employed knock-down strategies in 3T3SA, SVEC, and L929 cells (Figure 3C, 3D, and S2C) and also evaluated Rip3^-/- MEF (Figure 3E). In all of these settings, RIP3i-induced apoptosis required the presence of RIP3. The RIP3 pro-necrotic partner MLKL was subjected to knockdown in 3T3SA cells and shown to be dispensable for apoptosis (Figure 3F and data not shown). Thus, RIP3i promotes the concentration-dependent ability of RIP3 to trigger caspase activation and apoptotic cell death completely independent of necroptosis machinery.

Figure 3

Concentration-dependent apoptosis of GSK'840, GSK'843 and GSK'872 requires RIP3

(A) Relative viability of NIH3T3 cells (left) 18 hpt with increasing concentrations of GSK'872 (black bars) or GSK'843 (grey bars), or 3T3SA cells (right) in 10 mM RIP3i compounds, +/- zVAD, assessed as described in Figure 1F. (B) IB showing RIP3 and β-actin levels in NIH3T3 and 3T3SA cells. (C) Relative viability at 18 hpt with GSK'872 (10 mM) in 3T3SA, SVEC and L929 cells, transduced with non-targeting (NT) shRNA (black bars) or RIP3-specific (grey bars) shRNA. (D) Analysis of Casp3/Casp7 proteolytic activity (DEVDase) in transduced 3T3SA cells at 4 hpt with GSK'843, GSK'872 or TNF plus CHX. (E) Relative viability comparing WT (black bars) and Rip3^-/- (grey bars) MEF at 18 hpt with GSK'872 +/- zVAD. (F) Relative viability comparing 3T3SA cells transfected with NT or MLKL-specific siRNA 18 hpt with GSK'872 +/- zVAD or with TNF plus zVAD +/- GSK'872. An IB inset (right) show the levels of MLKL prior to any other treatment. (G) Relative viability comparing Rip3^-/- MEF alone and after transduction with human hRIP3 and hRIP3mutRHIM 18 hpt post treatment with GSK'840, GSK'843 or GSK'872, or with a combination of TNF (25 ng/ml), zVAD (25 μM) and BV6 (0.5 μM).

To investigate the properties of human-specific RIP3i, GSK'840, we reconstituted Rip3 ^{MEF with human RIP3 and conferred sensitivity to RIP3i GSK'840 as well as GSK'843 and GSK'872 (}Figure 3G). This data demonstrates the capacity of human RIP3 to promote apoptosis in the presence of high concentrations of RIP3 kinase inhibitor. These data also reveal that the chemically distinct RIP3i GSK'840 behaved the same as GSK'843 and GSK'872 in a mouse cell environment. In this setting, human RIP3 cannot confer sensitivity to necroptosis because it cannot interact with mouse MLKL (Sun et al., 2012), reinforcing observations that MLKL is dispensable for RIP3i-induced apoptosis. Cell death was nevertheless dependent on RHIM-signaling such that human RIP3mutRHIM failed to confer sensitivity to apoptosis (Figure 3G). Thus, for apoptosis, the trigger seems to be RIP3 itself.

Identification of Genes Essential for RIP3-induced Apoptosis

The impact of RIP3i seemed reminiscent of RIP3 overexpression-induced apoptosis (Kasof et al., 2000; Pazdernik et al., 1999; Sun et al., 1999; Yu et al., 1999), leading us to perform a genome-wide loss-of-function screen in human haploid HAP1 cells to identify genes that contribute to RIP3-initiated apoptosis. This method has previously identified mediators of cell death (Carette et al., 2009). HAP1 cells, which lack RIP3, were engineered (HAP1-Tet-RIP3) to express RIP3 following doxycycline (DOX) induction (Figure S3A) and subsequently died by apoptosis as RIP3 levels increased (Figure S3B). A retroviral gene-trap vector strategy was used to select death-resistant mutants and genes disrupted by insertion were mapped using deep sequencing. We identified a set of genes significantly enriched for insertions within the DOX-selected population compared to the unselected control, grouped categorically according to predicted function (Figure 4A), where the frequency of insertions within Rip3 itself provided confidence that the screen was specific. Hits within VP16-responsive Mediator (MED) transcription complex genes confirmed reliance of the expression system on this transactivator (Uhlmann et al., 2007). A subset of genes involved in transcription and chromatin remodeling (e.g., RPRD2, SP1, ZCCHC14) were also identified, though the mechanism by which they may contribute to RIP3-mediated apoptosis is currently unclear. RIP1, FADD, cFLIP_L and Casp8, were all implicated, consistent with the contribution of Ripoptosome-like machinery in death. Additionally, the screen also identified the Casp8 substrate Bid, implicating mitochondrial amplification machinery as necessary for RIP3-driven apoptosis.

Figure 4

Haploid genetic screen for genes involved in RIP3-mediated cell death

(A) Haploid screen results depicting each gene as a bubble where size corresponds to the number of independent gene trap insertions (also indicated in parentheses) the significance of enrichment is plotted on the y-axis. The top 25 most significantly enriched genes are labeled, colored and horizontally grouped by function (other genes are grey and in arbitrary position along the x-axis). (B) IB of the indicated components from anti-FADD immunoprecipitation of Triton X-solubilized 3T3SA cell supernatants (IP: FADD; top section), together with total supernatant (middle section) and total pellet (bottom section) fractions. Cells were assessed without treatment (untreated; lane 1), as well as 2.5 hpt with vehicle alone (DMSO; lane 2), zVAD (25 μM) alone (lane 3) and varying concentrations (10, 3, 1 and 0.3 mM) of GSK'872 in the presence (lanes 4 through 7) or absence (lanes 8 through 11) of zVAD. Single asterisk indicates modification of RIP1 in the presence of GSK'872 when caspases are active (lane 8) and double asterisks indicate slower migrating forms of RIP3 in the pellet fraction (lanes 4, 5 and 8). See also Figure S3.

RIP3-driven Assembly of a FADD-associated Casp8 Complex at High Concentrations of RIP3i Compound

We directly assessed the association of Ripoptosome components identified in the HAP1 screen during compound-induced apoptosis. Treatment with either GSK'843 or GSK'872 (Figure 4B, S3C, S3D, and S3E) resulted in a RIP3i concentration-dependent association of RIP3 as well as RIP1 with FADD (Figure 4D). In the presence of zVAD, this complex was stabilized and, in addition, Casp8 and cFLIP_L were present (Figure 4B, S3C, S3D, and S3E). These data directly implicate a Ripoptosome-like RIP1-FADD-cFLIP_L-Casp8 complex (Feoktistova et al., 2011; Tenev et al., 2011) in compound-induced apoptosis. We verified the RIP3i compound-driven association of Casp8, RIP1 and FADD with RIP3 in MEF (Figure S3D), an association that was absent in Rip3^-/- MEF (data not shown). Thus, RIP3 initiates assembly of a Casp8-activation platform at concentrations of RIP3i compound (3 and 10 μM) that trigger apoptosis (see Figure 2A). Under these conditions, RIP1 was partially processed, and, consistent with Casp8 activation, Casp3 and Casp8 matured into their pro-apoptotic forms (Figure 4B and S3C). These observations demonstrate that RIP3 drives assembly of a RIP1-FADD-cFLIP_L-Casp8 complex during apoptosis. These observations indicated that the assembly of this complex was more dramatically enhanced by RIP3i treatment, particularly when zVAD was present, compared to TNF plus CHX in the presence or absence of zVAD (Figure S3C and data not shown). Treatment also drove the accumulation of unmodified as well as more slowly migrating modified forms of RIP3 in the pellet fraction (Figure 4B and S3E). Such modifications characterize necrotic (Li et al., 2012) as well as apoptotic conditions. Thus, when triggered by RIP3i compound, RIP3 is a powerful recruiter of components known to drive extrinsic apoptosis.

Requirement for RIP1, FADD, cFLIP_L and Casp8 in RIP3i-induced Apoptosis

To determine whether the Casp8-activation platform plays a direct role in apoptosis induced by RIP3i, we employed inhibitors with different specificities. zVAD as well as the Casp8-specific inhibitor, zIETD, blocked death; whereas, neither Casp1- nor Casp9-specific inhibitors (zYVAD and zLEHD, respectively) had any impact (Figure 5A). Neither the reactive oxygen species scavenger BHA nor the autophagy inhibitor 3-MA blocked apoptosis. Consistent with this pattern, Cl-Casp8 accumulated in parallel with IETDase activity (Figure 5B and C) and knock-down of Casp8 prevented death and eliminated the accumulation of cleaved Casp3 forms (Figure 5D and S4A). Similarly, FADD (Figure 5E) and RIP1 (Figure 5F and S4B) were both necessary, but with an important distinction from Complex II formation in TNF signaling (Wang et al., 2008) where RIP1 kinase-mediated phosphorylation events predominate. Here, RIP1 kinase activity was completely dispensable. Neither Nec-1 (Figure S4C and data not shown) nor kinase-dead RIP1 (Rip1^K45A/K45A) MEF (Berger et al., 2014; Kaiser et al., 2014) prevented apoptosis (Figure 5G) even though necroptosis was blocked. Consistent with these findings, MEF from mice with RIP1, FADD, cFLIP_L or Casp8 deficiency resisted apoptosis (Figure 5H, 5I, 5J, and S4D). The requirement of cFLIP_L for RIP3i-induced apoptosis was particularly surprising given the acknowledged protective role of cFLIP_L in extrinsic apoptosis. To confirm the contribution of cFLIP_L to RIP3i-induced apoptosis, reconstitution of cFLIP_L-null cells restored sensitivity to RIP3i and, as expected, protected cells from TNF-induced apoptosis (Figure 5J, 5K, and S4E). This is consistent with the requirement of cFLIP_L for RIP3-initiated apoptosis (Figure 4A). Furthermore, cFLIP_L was essential for RIP3i-induced recruitment of RIP1 to FADD (Figure S4F). Thus, RIP3 recruits components and acts as a scaffold or adaptor to drive assembly of a Ripoptosome death complex dependent on cFLIP_L but independent of RIP1 or RIP3 pro-necrotic kinase activities.

Figure 5

Requirement for Casp8, FADD and RIP1 in RIP3-initiated apoptosis

(A) Relative viability of 3T3SA cells 18 hpt with GSK'872 in the absence (DMSO) or presence of caspase inhibitors (zVAD, zYVAD zLEHD, zIETD; 25 μM), an inhibitor of autophagy, 3-methyladenine (3-MA; 5mM), or an inhibitor of reactive oxygen species, butylated hydroxyanisole (BHA; 25 μM), assessed as described in Figure 1F. (B) Analysis of Casp8 proteolytic activity (IETDase) in 3T3SA cells treated with GSK'872 for the indicated times +/- zVAD. (C) IB of Cl-Casp8 in 3T3SA cells treated with GSK'872 for the indicated times +/- zVAD. (D) Relative viability of 3T3SA cells transfected with NT (black bars) or Casp8-specific (grey bars) siRNA and treated as described in panel C. An IB inset on the right shows the level of Casp8 knockdown. (E) Relative viability of 3T3SA cells transduced with EV (black bars) or FADD-DN-expressing retrovirus vector (grey bars) 18 hpt with GSK'872, zVAD and/or TNF, as indicated. (F) Relative viability of 3T3SA cells transfected with NT (black bars) or RIP1-specific siRNA (grey bars) 18 hpt with GSK'872, zVAD and/or TNF. An IB inset on the right shows level of RIP1 knockdown. (G) Relative viability of WT (black bars) or Rip1^K45A/K45A mutant MEF (grey bars) 18 hpt with GSK'872 +/- zVAD. (H) Relative viability of different MEF genotypes 18 hpt with GSK'872. (I) Relative viability of primary WT or cFLIP^-/- MEF treated with GSK'872, TNF, zVAD and/or BV6. (J) Relative viability of cFLIP^-/- MEF transduced with EV or cFLIP_L-expressing retrovirus 18 hpt with GSK'872 or TNF. (K) Analysis of Casp8 proteolytic activity (IETDase) in cFLIP_L-deficient MEFs 4 hpt with GSK'872 and TNF. (L) Relative viability in SVEC cells transduced with EV (white bar) or M45-expressing retrovirus (black bars), or M45mutRHIM-expressing retrovirus (grey bars) 18 hpt with GSK'872 alone or TNF plus zVAD. See also Figure S4.

To determine the extent to which RHIM signaling contributes to RIP3-initiated apoptosis, we showed that MCMV M45-encoded vIRA (Upton et al., 2010) suppressed; whereas, a nonfunctional RHIM mutant (M45mutRHIM) failed to suppress RIP3i-induced apoptosis (Figure 5L). WT-MCMV, expressing both a functional M45 (vIRA) and the Casp8 inhibitor M36 (vICA), completely prevented GSK'872-induced apoptosis (Figure S4G), so long as cells were infected for at least 2 h before RIP3i compound was added. RIP3i induced RHIM-dependent NF-κB activation (Figure S4H and S4I) similar to RIP3 overexpression settings (Kaiser et al., 2008; Rebsamen et al., 2009); however, NF-κB inhibition had no impact on death (Figure S4J). Furthermore, TNF-deficient MEF remained sensitive to RIP3i (Figure S4K). Apoptosis therefore occurs independent of TNF and NF-κB, but relies on RIP3 RHIM-mediated recruitment of RIP1, resulting in Casp8 activation.

RIP3 Kinase Domain Mutants in RIP3-Initiated Apoptosis

To directly address the recently proposed (Newton et al., 2014) prosurvival role of RIP3 kinase, we reconstituted Rip3^-/- cells with a series of nonfunctional RIP3 mutants that were either non-toxic when stably expressed or toxic due to the conversion of RIP3 into an autoinducer of apoptosis (Figure 6A). Neither the kinase ATP binding pocket mutant K51A nor phosphorylation site mutant T231A/S232A or S204A/S207A/S211A induced apoptosis when introduced into cells, a pattern similar to the impact of MEF transduced with WT RIP3 (Figure 6B). Cells expressing the kinase domain mutant D161N could only be cultured when caspase inhibitor was included in the medium (Figure 6A). As expected, cells expressing kinase-inactive mutant, K51A, were unable to support necroptosis (Figure 6B, 6C and data not shown). Thus, the pro-apoptotic behavior of kinase-inactive mutant D161N was unique, seemingly paralleling high concentration RIP3i or RIP3 overexpression.

Figure 6

Characterization of RIP3 kinase domain mutants

(A) Table summarizing microscopic assessment of Rip3^-/- MEF survival 3 days after transduction with WT RIP3 or indicated mutants +/-zVAD. (B) Relative viability of transduced cells 18 h post treatment with TNF, zVAD and GSK'872 as indicated and assessed as described in Figure 1F. (C) IB for FLAG-RIP3 levels in transduced cells. (D) Relative viability of RIP3shRNA transduced 3T3SA cells following transfection with shRNA-resistant RIP3 or indicated RIP3 mutants for 16 h and treated with GSK'872, zVAD, and TNF. (E) IB of the indicated components from RIP3shRNA transduced 3T3SA cells showing anti-FLAG IP, supernatant, and pellet fractions 8 h post transfection with RIP3 or mutants (K51A, D161N) +/- GSK'872 for 2 h prior to harvesting. (F) IB of the indicated components in unmanipulated supernatants and pellets of RIP3shRNA transduced 3T3SA cells 16 h after transfection with shRNA RIP3 or mutants (K51A, D161N) +/- GSK'872 without or with zVAD for 3 h prior to harvesting. See also Figure S5.

RIP3i compound triggered apoptosis in K51A mutant-transduced cells (Figure 6B), comparable to WT RIP3, indicating that compound binding rather than kinase inhibition promoted cell death. Mutant T231A/S232A, which prevents pro-necrotic MLKL binding (Sun et al., 2012), as well as a triple mutant (S204A/S207A/S211A) in the activation loop, also supported apoptosis. As expected, necrotoposis required MLKL interaction and both apoptosis and necroptosis required RHIM-signaling (Figure 6B). Overall, these data preclude a pro-survival role of RIP3 kinase activity but are consistent with a kinase-independent ability of RIP3 to recruit RIP1 and drive FADD-cFLIP_L-Casp8 death complex activation.

The striking difference in viability of cells stably transduced with these two RIP3 kinase-inactive mutants prompted further investigation. RIP3 shRNA knock-down 3T3SA cells were transiently transfected with shRNA-resistant WT RIP3, RIP3-kinase-inactive mutants (K51A, D161N, D161G and D143N), phosphorylation mutant (T257A), and a D161N/mutRHIM double-mutant. Of the mutant RIP3 forms evaluated, only the previously characterized (Newton et al., 2014) charge swap mutant D161N promoted apoptosis (Figure 6D). All RIP3 forms were expressed at similar levels (Figure S5A), so it was striking to observe cells survived when a neutral aa substitution (D161G) was made at this same position (Figure 6D). These data implicate a dramatic impact of charge reversal in producing the toxic effects. Cells also survived when a different kinase-inactive charge swap mutant (D143N) was evaluated, suggesting a particular consequence of D161N and not a general alteration within the active site. As expected, cells expressing any of the nontoxic RIP3 kinase-inactive mutants were resistant to TNF-induced necroptosis. Only WT and T257A RIP3 conferred sensitivity to either TNF+zVAD or zVAD alone, the latter a consequence of high RIP3 levels produced by transfection, and all cells survived when treated with a combination of GSK'872 and zVAD. Similar to WT RIP3, cells expressing mutants K51A, D161G, D143N or T257A were sensitive to high concentration GSK'872 but cells expressing mutRHIM or EV were not. The inherent toxicity of D161N was neither blocked nor enhanced by addition of RIP3i compound and a D161N/mutRHIM double mutant showed expected viability (Figure S5B), confirming the role of RHIM signal transduction in death. Furthermore, D161N- or RIP3i-induced apoptosis was prevented when either Casp8-inhibitor vICA or RHIM antagonist vIRA was present (Figure S5C). As expected, neither vICA nor M45mutRHIM suppressed necroptosis (Kaiser et al., 2014; Kaiser et al., 2011). These data challenge the notion that RIP3 kinase activity has any inherent role promoting survival, but rather demonstrate the conversion of RIP3 into a pro-apoptotic adapter by specific mutation (D161N) through a mechanism that requires RHIM signaling and leads to Casp8 activation, analogous to that triggered by high concentration RIP3i or overexpression.

Next, we evaluated the role of RIP1 in D161N-induced apoptosis. D161N associated with RIP1 and Casp8 spontaneously in cell extracts (Figure 6E), in contrast to the pattern of WT or K51A mutant RIP3 where association was dependent on RIP3i. Doxycycline-induced expression of D161N RIP3 in RIP3-deficient MEF triggered RIP1 association with FADD and RIP3, a RHIM-dependent pattern that was only observed with WT RIP3 and K51A in the presence of RIP3i (Figure S5D). D161N enhanced the accumulation of RIP3 in the pellet fraction as early as 12 h post transfection (Figure S5E); whereas, WT, K51A, D143N, and D161G RIP3 failed to accumulate in this fraction unless treated with RIP3i and zVAD (Figure 6F and S5E). Consistent with a role for RHIM-signaling in oligomerization, aggregated forms of RIP3 did not appear in the pellet when RHIM signaling was blocked (data not shown). D161N mutation or RIP3i-binding therefore alters the conformation of RIP3 and triggers RHIM-dependent oligomerization reminiscent of aggregate, “amyloid-like” RIP1-RIP3 complexes (Li et al., 2012).

Rip3^K51A/K51A Kinase-dead Knockin Mice Are Viable

The behavior of RIP3 kinase-dead mutants supported the striking midgestational lethality observed in D161N mutant knockin mice (Newton et al., 2014) and predicted the opposite outcome would occur with a nontoxic mutant. When generated, Rip3^/K51A kinase-dead knockin mice were clearly viable and fertile (Figure 7A and B). This mutant strain did not show any susceptibility to midgestational or perinatal death. To determine whether the viable Rip3^/K51A mutant, like the lethal Rip3^D161N/D161N mutant (Newton et al., 2014), rescues embryonic lethality of Casp8^-/- embryos (Sakamaki et al., 2002; Varfolomeev et al., 1998), we performed a cross and rescued viable and fertile Casp8^{Rip3K51A/K51A} mice at the expected Mendelian frequency (Figure 7B and S6A). This extends previous rescue of Casp8^Rip3-/-mice (Kaiser et al., 2011; Oberst et al., 2011; Zhang et al., 2011) to clearly show the contribution of pro-necrotic RIP3 enzymatic activity in midgestational death of Casp8-deficient embryos without the complications of the Rip3^D161N/D161N mutant (Newton et al., 2014).

Figure 7

Characterization of RIP3 kinase inactive mice

(A) Schematic representation of the WT, recombined, Cre-mediated excised, Flp-mediated excised RIP3 alleles with the relevant restriction sites for the DNA blot analysis. (B) Photograph of age-matched WT, Rip3^K51A/K51A and Rip3^Casp8-/-mice. (C) Relative viability of WT and Rip3^K51A/K51A MEF 18 hpt with GSK'872, TNF, zVAD and BV6 as indicated and as assessed in 1F. (D) Relative viability of WT and Rip3^K51A/K51A mutant BMDM 18 hpt with GSK'872, TNF, zVAD, BV6, poly (I:C)/zVAD, or LPS/zVAD for 18 h. (E) Relative viability of WT (left) and Rip3^K51A/K51A (right) MEF transfected with either NT siRNA, RIP1 siRNA, or a combination of RIP1 and RIP3 siRNA for 48 h and then 48 hpt with IFNβ (100 U/ml), zVAD, and GSK'872 as indicated. (F) MCMV titers in spleen (left) and liver (right) from mice of indicated genotypes 3 days post infection. (G) Recall response to MCMV infection. Total numbers (left) or percentages (right) of splenic CD8+ T cells producing IFNγ or IFNγ and TNFα following stimulation with M45-specific peptide 4 days post challenge with RM427 virus 14 days post infection with MCMV-M45mutRHIM in mice of indicated genotypes. See also Figure S6.

RIP3 K51A mutant protein was detected in tissues and cells from adult mice, albeit at levels that were lower than WT RIP3 (Figure S6B). When levels of mutant RIP3 in splenic cells from Casp8^{Rip3K51A/K51A} were directly compared to Casp8^{Rip3D161N/D161N} mice, RIP3-K51A and RIP3-D161N expression was comparable (Figure S6C) although lower than a Rip3^+/- spleen. Even though Rip3^D161N/D161N mice have an obvious phenotype, these low levels raise the possibility that Rip3^K51A/K51A kinase-dead mutants simply phenocopied Rip3 ^{mice. To address the biologic potential of K51A mutant protein in these mice, we evaluated MEF and BMDM for the ability to sustain RIP3i-induced apoptosis. Consistent with the behavior of transduced and transfected cells,} Rip3^K51A/K51A mutant cells remained as sensitive as WT cells to RIP3i compound-induced apoptosis (Figure 7C, D, and E). To address the ability of these cells to support inflammasome activation (Vince et al., 2012), Rip3^K51A/K51A BMDMs treated with GSK'872 and IAP agonist SMAC007 produced IL-1β at levels similar to WT and much higher than Rip3^-/-cells (Figure S6D). Thus, K51A mutant RIP3 adapter function is sufficient for inflammasome activation, a result that demonstrates this pathway does not involve kinase activity. As expected, WT MEF and BMDM generated from Rip3^K51A/K51A mice were resistant to necroptosis induced by TNFR1, TLR3, and TLR4 ligands and IFNβ (Figure 7C, 7D, 7E, S6E, and S6F). These data dismiss any vital contribution of RIP3 kinase activity in vivo (Newton et al., 2014) and, despite low expression levels, reveal that Rip3^K51A/K51A mice sustain adapter-dependent processes such as apoptosis and inflammasome activation, consistent with a proposed adaptor function underlying midgestational apoptosis in Rip3^D161N/D161N mouse development.

In addition to controlling inflammation, RIP3 plays an innate immune role in limiting viral infection. To determine the ability of Rip3^K51A/K51A mice to mount an immune response to MCMV, a natural mouse pathogen, we compared WT virus to MCMV-M45mutRHIM, a mutant that fails to replicate in WT mice due to elimination of virally infected cells by the DAI-RIP3 pathway (Upton et al., 2010, 2012). Infection with MCMV-M45mutRHIM was completely normalized in Rip3^K51A/K51A mice as in Rip3^-/- mice, demonstrating the contribution of pro-necrotic RIP3 kinase activity to death of mutant virus-infected cells in vivo (Figure 7F). M45mutRHIM infection of BMDMs from WT mice died by necroptosis; whereas, Rip3^K51A/K51A BMDMs survived (data not shown). These data directly demonstrate the importance of RIP3 kinase in the DAI-driven necroptosis pathway. Our previous investigation showed that M45mutRHIM kills cultured cells independent of RIP1 (Upton et al., 2012). Here, we extended observations in both Rip1^K45A/K45A and Rip1^Rip3Casp8-/- mice (Figure 7F and S6G), providing definitive evidence that neither RIP1 nor its kinase activity is involved in DAI-RIP3 necroptosis that severely attenuates mutant virus replication.

We previously established that WT, Rip3^-/-, Casp8^Rip3-/-, and Casp8^Rip3Rip1-/- mice all control WT MCMV infection, mounting a robust virus-specific T cell response despite the absence of extrinsic death pathways (Kaiser et al., 2014; Kaiser et al., 2011). To study the adaptive response to MCMV, Rip3^K51A/K51A mice were infected with M45mutRHIM for 14 days and boosted by secondary infection with a lacZ-expressing MCMV, conditions that drive a maximal T cell response. Four days after boosting, infected WT, Rip3^-/-, and Rip3^K51A/K51A mice all exhibited comparable numbers of total splenic T cells as well as total CD4 and CD8 T cells (Figure S6G). Importantly, the MCMV M45 epitope-specific CD8 T cell response remained very robust across the genotypes (Figure S6G), rising more than 1,000-fold above levels in naïve mice. This pattern was similar to WT virus infection (Kaiser et al., 2014; Kaiser et al., 2011), reinforcing the fact that extrinsic cell death pathways are dispensable for a robust adaptive immune response. A single exposure to MCMV for four days failed to induce the dramatic response seen in previously infected mice (data not shown). Figure 7F and G show levels of virus-specific CD8 T cells in settings where mutant virus replicates to high levels (Rip3^-/- or Rip3^K51A/K51A mice) were comparable to settings where this virus drives necroptosis and replicates poorly (WT mice). Thus, the induction of immunity assessed as M45 peptide-specific IFNγ ^{and INFγTNF CD8 T cell frequencies did not require sustained infection so long as necroptosis was triggered by the virus (}Figure 7G). RIP3 and its kinase activity are dispensable for robust antiviral immunity in settings where virus replication proceeds, undoubtedly because antigen load compensates when cell death is blocked. Thus, Rip3^K51A/K51A mice are viable and fertile, as well as immunocompetent. Collectively, these data demonstrate that RIP3 kinase activity can be eliminated in vivo without the dire consequence of triggering lethal apoptosis or even compromising immune competence.

Discussion

The ability of RIP3 deficiency to prevent disease has heightened interest in the therapeutic potential of small molecule inhibitors that target RIP3 kinase activity. In this study, we investigated the consequence of ablating RIP3 kinase activity with such inhibitors. GSK'840, GSK'843 and GSK'872 bind the kinase domain and inhibit kinase activity with high specificity, targeting a broader range of pro-necrotic stimuli than can be achieved with RIP1 kinase inhibitors. Characterization of RIP3 inhibitors demonstrated the crucial role of RIP3 kinase activity in necroptosis but unveiled a surprising conformation-dependent capacity of RIP3 to recruit RIP1 via RHIM binding and trigger rapid Casp8-dependent apoptosis. Mutagenesis of the RIP3 kinase domain to D161N produced an impact similar to chemical inhibitors. The propensity of kinase domain to drive apoptosis was not absolute as cells expressing K51A, D161G and D143N mutants retained viability. At one level, this process seems reminiscent of the paradoxical activation of RAF kinases by inhibitors (Holderfield et al., 2013). Here, though, high concentration RIP3i has an impact on conformation that unleashes RHIM-dependent oligomerization independent of the kinase activity itself. Although the aggressive induction of apoptosis by inhibitors we describe here may have specific utility in cancer chemotherapy, RIP3i-induced apoptosis represents a significant challenge to the development of anti-inflammatory therapies targeting RIP3.

Our work contrasts the Casp8-dependent embryonic lethality of the kinase-dead D161N mutant (Newton et al., 2014) and consequent implication of RIP3 phosphorylation events in the suppression of a Ripoptosome-like complex (Feoktistova et al., 2011; Tenev et al., 2011). We describe several RIP3 kinase domain mutants that do not convert RIP3 into an autoinducer of apoptosis, particularly a neutral substitution at this same residue (D161G) that destroys the kinase activity without causing dire consequences. Something about the dramatic charge swap (D161N) causes a gain-of-function resulting in RHIM oligomerization. Other kinase-dead mutants K51A, D161G and D143N only drive this oligomerization when high concentrations of RIP3i compound are present. WT RIP3 has a propensity to drive this same process when overexpressed (Kasof et al., 2000; Pazdernik et al., 1999; Sun et al., 1999; Yu et al., 1999). The ability to trigger RHIM-signaling that leads to Casp8-dependent apoptosis is therefore distinct from the pro-necrotic role of RIP3 kinase.

We used these findings to derive viable and fertile Rip3^K51A/K51A kinase inactive knockin mice that cannot support necroptosis, a result that demonstrates RIP3 kinase activity is dispensable for life. Rip3^K51A/K51A mice are fully immunocompetent and retain the ability to generate a robust, virus-specific T cell response as well as control infection by a natural mouse pathogen. We anticipate these mice will reveal additional insights into the contribution of necroptosis in inflammatory disease beyond those provided by Rip1^K45A/K45A mice (Berger et al., 2014; Kaiser et al., 2014). Although RIP1 kinase-inactive mutants and necrostatins are nontoxic, they do not prevent all forms of RIP3 necroptosis. The derivation of Rip3^K51A/K51A mice is the best evidence that nontoxic small molecule RIP3 inhibitors are feasible. Going forward, therapeutic strategies that directly target RIP3 kinase activity certainly must avoid the conformational changes that unleash RHIM-signaling and nucleate Casp8-mediated apoptosis.

The dialogue between RIP3 and Casp8 represents a détente where the enzymatic activity of each sits in control cell fate, expanding the concept of pathogen supersensor as a trap set to eliminate infection (Kaiser et al., 2013b). It has long been appreciated that suppression of Casp8 opens the necroptotic trap door leading to RIP3-MLKL oligomerization (Kaiser et al., 2013b). This study highlights not only Casp8 suppression of RIP3 kinase activity but also the remarkable capacity for RIP3 to function upstream of Casp8 to drive apoptosis. RIP3 sits in control of a Ripoptosome-like RIP1-FADD-cFLIP_L-Casp8 complex (Feoktistova et al., 2011; Tenev et al., 2011). In this pathway cFLIP_L plays some direct role that is reminiscent of settings where cFLIP_L was implicated in stabilizing Casp8 activity (Micheau et al., 2002). The mechanistic understanding described here may extend to other settings where RIP3 has been described as an apical adapter upstream of Casp8 activation, such as IFN stimulation where RIP3 promotes Casp8 activation (Dillon et al., 2014; Kaiser et al., 2014; Thapa et al., 2013) and IAP-depletion in macrophages where RIP3 promotes Casp8-dependent inflammasome activation (Vince et al., 2012). Based on the evidence presented here, these processes are likely to rely on RHIM-dependent signal transduction, but unlikely to require the kinase activity of RIP3. The kinase domain of RIP3 now emerges as a crucial brake on RHIM-signaling that can be released following perturbation of RIP3 by mutation (D161N) or high concentration kinase inhibitors. RIP3-based therapies will harness a better understanding of the intricate interplay of RIP3 kinase domain function, kinase activity, and RHIM interactions in apoptotic and necrotic cell death.

Experimental Procedures

Cell culture and reagents

HT-29, L929, NIH3T3, 3T3SA, SVEC, and MEF were maintained in DMEM containing 4.5 g/ml glucose, 10% fetal calf serum (Atlanta Biologicals), 2 mM L-glutamine with 100 units penicillin/ml and 100 units streptomycin/ml (Invitrogen). Primary neutrophils were maintained in RPMI containing 4.5 g/ml glucose, 10% fetal calf serum (SAFC Bioscience), 2 mM L-glutamine with 100 units penicillin/ml and 100 units streptomycin/ml. For BMDM cultures, pooled bone marrow cells from flushed tibias and femurs were differentiated for 5 to 7 days in DMEM containing 20% fetal calf serum and 20% filtered L929-conditioned medium (to provide macrophage colony-stimulating factor). The reagents and compounds used were IFNβ (PBL), TNFα (Peprotech), qVD-OPH, zYVAD-fmk, zIETD-fmk, zLEHD-fmk, 3-Methyadenine (3-MA) and Necrostatin (Nec)-1 (Calbiochem), non-targeting, RIP1, Casp8, and, MLKL siRNA ON-TARGET SMARTpools (Thermo Scientific), cycloheximide, butylated hydroxyanisole (BHA) and BMS-345541 (Sigma), DMSO (Corning), zVAD-fmk (Enzo Life Sciences and Promega), SYTOX Green (Invitrogen), poly(I:C) (Amersham Pharmacia), BV6 (Genentech) and SMAC007 (GlaxoSmithKline).

Transfections and transductions

Retrovirus and lentivirus transductions were performed as described (Upton et al., 2010). siRNAs were transfected using Lipofectamine RNAi Maxx (Invitrogen) according to manufacturer instructions. For transient transfections, 3T3SA cells stably expressing RIP3-shRNA from pLKO.1-RIP3 shRNA (TRCN0000022535) (Open Biosystems), were transfected with empty or RIP3 construct-expressing vector and pMAX-GFP (Amaxa) in a 20:1 ratio using Superfect (Qiagen). For transfection of dual luciferase plasmids, 3T3SA cells were transfected with NFκB-luc (Stratagene) and phRL-TK (Promega) in a 4:1 ratio using Superfect and treated with GSK'872 at 24 h post transfection. Luciferase was measured using Dual-Glo Luciferase Assay system (Promega).

For further experimental procedures see the extended experimental procedures in Supplemental Information.

Supplementary Material

1

2

3

4

5

Acknowledgments