RIPK2 kinase domain (used for crystallography and assays)
Human RIPK2 (residues 3–317 or residues 8-317) was cloned into pFB-LIC-Bse and prepared by baculoviral expression. RIPK2 protein was purified sequentially by nickel affinity and size-exclusion chromatography. The kinase domain displayed heterogeneous phosphorylation at up to 5 sites which could be dephosphorylated by lambda phosphatase. Note, DNA sequencing of the RIPK2 construct spanning residues 3-317 revealed the expected wild-type sequence, whereas the shorter construct spanning residues 8-317 was found to have a mutation R171C in the kinase activation loop.
XIAP BIR2 domain (used for assays)
Human XIAP BIR2 domain (residues 124-242) was cloned into the pGTVL2 vector in frame with N-terminal 6xHIS and GST tags. Protein was expressed in E. coli (BL21DE3) cells and purified either by batch binding to glutathione sepharose or by sequential nickel affinity and size-exclusion chromatography.
XIAP BIR2 D214S mutant domain (used for assays)
The D214S mutant was prepared similarly to the wild-type XIAP BIR2 domain.
- Dephosphorylated RIPK2 kinase domain a.a. 8-317 bound to ponatinib (4C8B, 2.75 Å)
- Phosphorylated RIPK2 kinase domain a.a. 3-317 bound to CSR35 (6ES0, 2.38 Å)
- Phosphorylated RIPK2 kinase domain a.a. 3-317 bound to CLSP18 (6FU5, 3.26 Å)
Fig. 2. Dimeric structure of RIPK2. (A) Structural features of the RIPK2 kinase domain. (B) Dimeric packing of the RIPK2 kinase domain.
The RIPK2-ponatinib complex was the first structure solved for this kinase. RIPK2 exhibits the canonical bilobal kinase fold followed by a 16-residue αJ helix that packs alongside the loop connecting the αD and αE helices (Fig. 2). Interestingly, RIPK2 contains several unusual sequence changes in its catalytic motifs that are not conserved in other RIPKs. The typical HRD triad in the catalytic loop is changed to HHD, while the activation loop APE motif is changed to PPE. The RIPK2 structure reveals a homodimeric packing arrangement consistent with the dimeric state observed in solution by analytical gel filtration (Fig. 3). The protein interface is highly symmetrical with the two active sites facing in opposite directions and rotated approximately 90° relative to one another. Binding is supported by the αJ helices, which pack against each other in an antiparallel fashion, and form both hydrophobic and hydrogen bonding interactions.
Fig. 3. RIPK2 is dimeric in solution. Analytical gel filtration trace of RIPK2 using a superdex 200 10/300 GL column. The retention volume of RIPK2 was compared to MW standards.
RIPK2 structures with type II inhibitors ponatinib and CSR35 showed a “DFG-Asp out, αC-Glu in” conformation, whereas type I inhibitor CSLP18 was bound to a “DFG-Asp in, αC-Glu in” conformation. Electron density for much of the activation loop was missing in all structures irrespective of whether the protein used for crystallisation was dephosphorylated with lambda phosphatase (ponatinib complex) or left with heterogeneous phosphorylation (CSR35 and CSLP18 complexes). Structural details of the inhibitor binding modes are provided under chemical matter.
Assays of inhibitor potency
Thermal shift assay
A fluorescence-based thermal shift assay (differential scanning fluorimetry (DSF)) was performed as an initial screen to identify potential RIPK2 inhibitors. In this assay, the previously reported RIPK2 inhibitor gefitinib yielded a thermal shift (ΔTm) value of 9.5°C. However, we found the most potent inhibitor was ponatinib with a ΔTm value of 23.1°C.
ADP-Glo in vitro kinase assays
An ADP-Glo (Promega) in vitro kinase assay was performed to determine inhibitor IC50 values. In this assay, the ADP produced from the kinase reaction is converted to a luminescent signal (Fig. 4A-B).
NOD2-HEK-Blue activation assay
To monitor RIPK2 inhibition in cells, we measured the downstream activation of NFĸB in HEKBlue cells, which expressed NOD2 and a NFĸB-SEAP reporter. In this assay, secreted embryonic alkaline phosphatase (SEAP) is used as a reporter of gene expression which acts on QUIATI-Blue detection media to produce a blue/purple colour detectable using a spectrophotometer (Fig. 4C).
Fig. 4. Comparison of enzymatic and cellular assays for RIPK2. (A) ADP-Glo in vitro kinase assay testing ponatinib inhibition of Abl and RIPK2 kinases. (B) ADP-Glo assay for RIPK2 inhibition by other clinically-used kinase inhibitors. (C) NOD2-HEK-Blue activation assay testing selected inhibitors in HEK293 cells. Gefitinib activity is notably less potent against RIPK2 in cells compared to the in vitro ADP-Glo assay.
Comparison of the in vitro kinase assay and cellular reporter assay revealed remarkable differences (Fig. 4). In the kinase assay, the IC50 value of gefitinib (50.7 nM) was between those of regorafenib (41.0 nM) and sorafenib (75.4 nM). By contrast the cellular EC50 value for gefitinib (7.4 µM) was significantly weaker than those of regorafenib (3.8 nM) and sorafenib (16.9 nM). All of these compounds are FDA approved drugs and therefore have been optimised for cellular and in vivo activity. A striking lack of correlation between in vitro kinase and cellular data for RIPK2 inhibitors was also observed for a set of structurally similar CSLP analogs (Fig. 5).
Fig. 5. Lack of correlation for inhibitor potencies measured using the ADP-Glo and cellular HEKBlue reporter assays. (A) Correlation plot for different CSLP series compounds. Selected compounds illustrating lack of correlation are highlighted red and labelled. (B) General scaffold for the CSLP inhibitor series.
RIPK2 nanoBRET target engagement assay
A nanobret (nano-bioluminescence resonance energy transfer) target engagement assay for RIPK2 was developed through SGC collaboration with Matt Robers at Promega (12). In this assay, cells were transfected with a RIPK2 transgene fused to NanoLuc luciferase and a BRET signal induced by addition of a cell-permeable fluorescent tracer compound SGC-590001 derived from ponatinib (Fig. 6). The binding potencies of other test compounds were determined in these cells by their ability to compete with the tracer leading to a reduction in the BRET signal. An overview of the assay is provided in (12).
Fig. 6. Nanobret assay for RIPK2 in HEK293 cells. (A) Cells transfected with NanoLuc-RIPK2 were treated with serial dilutions of the SGC-590001 fluorescent probe to establish a window for Nanobret measurement. Signal was inhibited by 1 µM ponatinib (grey box indicates range of ponatinib concentrations explored). Resistance to ponatinib could be achieved by introduction of the RIPK2 T95W ‘gatekeeper’ mutation in the kinase hinge region. (B) Example IC50 measurement for RIPK2 inhibition by the GAK1 kinase chemical probe SGC-GAK1.
Assays of RIPK2 signalling
Inhibition of RIPK2 activation upon NOD2 stimulation with bacterial peptidoglycans
Stimulation of NOD2-expressing HEK293 cells with L18-MDP (a lipidated muramyl dipeptide with enhanced potency) caused a rapid increase in endogenous pSer176-RIPK2 (autophosphorylation) that could be blocked by small molecule inhibitors. Blocking RIPK2 activity prevented the subsequent degradation of IĸBα, which is required for activation of NFĸB (Fig. 7).
Fig. 7. Inhibition of RIPK2 activation in HEKBlue cells. Cells pretreated with inhibitor were stimulated with 1 µg/mL L18-MDP. Changes in RIPK2 phosphorylation and IĸBα abundance after 30 mins were analyzed by Western blotting. Levels of tubulin and total RIPK2 were used as loading controls.
Small molecule inhibition of RIPK2 ubiquitination
Stimulation of NOD2 by MDP leads to rapid ubiquitination of RIPK2 by XIAP within 45 mins, a process required for downstream signalling. We analyzed this step in human monocytic THP-1 cells. Pre-treatment with ponatinib interfered with L18-MDP-induced RIPK2 ubiquitination in a dose-dependent manner. Concentrations as low as 5-10 nM reduced the extent and length of Ub-modified RIPK2, while RIPK2 ubiquitination was completely blocked at concentrations of 25 nM or higher (Fig. 8).
Fig. 8. Inhibition of RIPK2 ubiquitination in THP-1 cells. Cells were pretreated with kinase inhibitor or DMSO for 30 minutes and stimulated with 200 ng/mL L18-MDP. Ubiquitinated proteins were isolated using TUBE reagent and analysed by immunoblotting.
Assays showing RIPK2 catalytic activity is dispensable for NOD2-dependent inflammatory signalling
Ubiquitination and NF-ĸB signalling of kinase-dead RIPK2 mutants
U2OS/NOD2 RIPK2 KO cells were used to test the requirement for RIPK2 catalytic activity. Cells were reconstituted with either WT or kinase-dead RIPK2 mutants. Upon NOD2 stimulation with L18-MDP (200 ng/ml, 1 h) all variants of RIPK2 were ubiquitinated to a similar extent showing that RIPK2 catalytic activity was dispensable for this activation step. Ubiquitination of kinase-dead RIPK2 was also accompanied by the downstream destruction of IĸBα which allows for NF-ĸB signalling (Fig. 9). Importantly, the ubiquitination of kinase-dead RIPK2 was still inhibited by ponatinib. These results suggested that efficacious RIPK2 kinase inhibitors might work through blocking XIAP binding and ubiquitination of RIPK2 rather than by blocking RIPK2 catalytic activity.
To quantify NF-ĸB signalling we used a dual luciferase NF-ĸB reporter assay in HEK293FT cells. NF-ĸB signalling was observed for both WT and kinase-dead RIPK2 variants and was inhibited by ponatinib consistent with the RIPK2 ubiquitination assay results. As ponatinib is a highly promiscuous kinase inhibitor we engineered a T95W gatekeeper mutation into all RIPK2 constructs to make them drug resistant (evidenced in Fig. 6). Kinase-dead RIPK2 mutants with the T95W mutation were still active in the NF-ĸB reporter assay, but were now unaffected by ponatinib confirmating that the signalling was specific to RIPK2 (Fig. 9).
Fig. 9. Kinase inhibitors block NOD2-dependent inflammatory signalling from kinase-dead RIPK2 mutants. (A) U2OS/NOD2 RIPK2 KO cells were reconstituted with RIPK2 variants or empty vector. Stimulation with L18-MDP led to ubiquitination of all RIPK2 variants that was inhibited by ponatinib. (B) Signalling from WT and kinase-dead HA-RIPK2 constructs was measured in HEK293FT cells using a dual luciferase NF-ĸB reporter assay. Cells were treated with DMSO or ponatinib (200 nM, 24 h) as indicated. Relative luciferase activity in ponatinib-treated samples is shown relative to the activity in the corresponding HA-RIPK2 transfected sample not treated with inhibitor. Data represent mean ±SEM (n =4 independent experiments).
Assays of XIAP binding to RIPK2
Fig. 10. Mapping of the XIAP-interacting surface on RIPK2. (A) SPOT synthesis 15-mer peptide array spanning the RIPK2 sequence and probed with 6xHis-GST-BIR2 domain protein. Bound protein by immunoblotting with anti-His antibody. (B) Quantification performed using ImageJ software for selected peptides reprinted in triplicate and probed with either WT or D214S mutant XIAP BIR2 domain. Peptide A13 (red labelling) showed specific binding to WT XIAP.
SPOT synthesis peptide array mapping of the XIAP-binding site on RIPK2
The XIAP BIR2 domain mutation D214S causes the immunodeficiency syndrome XLP2 (X-linked lymphoproliferative syndrome type-2) and has been shown to abolish XIAP binding to RIPK2. To map the corresponding XIAP-interaction surface on RIPK2, a SPOT synthesis peptide array was prepared spanning the entire sequence of RIPK2 with overlapping 15 a.a. peptides. Probing this array with 6xHis-GST-BIR2 protein identified seven putative interacting peptides in the kinase domain of RIPK2 (Fig. 10A). These peptides were reprinted in triplicate and further probed with either WT or D214S mutant BIR2 domain protein. A WT-specific interaction was observed for peptide A13 (RIPK2 residues 28-42) corresponding to the β2-β3 region of the kinase N-lobe (Fig. 10B).
RIPK2 kinase domain β2-β3 loop residues R36 and R41 appear critical for XIAP interaction
Alanine or leucine-scanning mutagenesis of RIPK2 a.a. 28-42 using the SPOT arrays identified R36 and R41 (β2-β3 loop) as the most critical residues for XIAP interaction (data not shown). To test this, we reconstituted RIPK2 KO U2OS/NOD2 cells with RIPK2 mutants R36L and/or R41L and performed a GST pull down with recombinant GST-BIR2-XIAP. Indeed, mutation of these residues impaired the interaction of RIPK2 with the XIAP BIR2 domain consistent with the SPOT array data (Fig. 11A).
Together these data suggest that the basic patch formed by RIPK2 R36/R41 in the kinase N-lobe β2-β3 loop mediates XIAP binding via a direct, electrostatic interaction with the acidic patch formed by E211, D214, E219 in the BIR2 IBM (IAP-binding motif) groove (Fig. 11B). The acidic pocket on the XIAP BIR2 domain is also the site of interaction of smac-mimetic small molecules that antagonise XIAP function.
Fig. 11. RIPK2 residues R36 and R41 are critical for XIAP binding. (A) Pulldown of RIPK2 variants from U2OS/NOD2 cell lysates using recombinant GST-BIR2-XIAP protein. (B) Putative interaction surfaces (dashed borders) on XIAP BIR2 domain (left, PDB 1C9Q) and RIPK2 kinase domain (right, PDB 5AR2).
- FDA-approved inhibitors ponatinib, regorafenib, sorafenib
- Type II inhibitor CSR35 and fragments
- Type I inhibitor CSLP series
Inhibitor co-crystal structures
Co-crystal structures were solved for three inhibitors: ponatinib, CSR35 and CSLP18 (Fig. 12).
Findings from the ponatinib complex
We identified ponatinib as a RIPK2 inhibitor with an EC50 of 0.8 nM in HEKBlue cells. Ponatinib is a highly promiscuous type II inhibitor developed against drug-resistant mutants of the Abl kinase. It is approved by the US FDA for late stage leukaemia. Due to its severe adverse effects, ponatinib is not suitable for use in chronic indications, such as those potentially involving RIPK2. Nonetheless, the co-crystal structure of ponatinib revealed that the back pocket created by the DFG-out configuration of RIPK2 was increased in size relative to other kinases due to the presence of RIPK2 Ala73 (Fig. 13).
Fig. 12. Crystal structures of RIPK2-inhibitor complexes. (A) Ponatinib complex (4C8B, 2.75 Å). (B) CSR35 complex (6ES0, 2.38 Å). (C) CSLP18 complex (6FU5, 3.26 Å).
Fig. 13. Extended back pocket in RIPK2. The trifluoromethyl group of ponatinib occupies the hydrophobic pocket vacated by the inverted DFG motif. Here the pocket is greatly enlarged in RIPK2 due to the presence of Ala73 (αC), whereas nearly all kinases contain a bulky side chain at this position, such as Leu70 in RIPK1.
Thus, there may be opportunities for the development of larger RIPK2-selective molecules that would be sterically restricted from binding to the wider kinome.
Findings from the CSR35 complex
The type II inhibitor CSR35 was modified from the regorafenib scaffold to find a new interaction with the activation loop of RIPK2 as a strategy to explore novel kinase regions for improved inhibitor selectivity. Our co-crystal structure of CSR35 with RIPK2 revealed a new ionic interaction between the inhibitor’s carboxylic acid and the side-chain of Lys169. Unfortunately, the inhibitory potency was reduced some 50-fold suggesting further chemistry would be required to optimise this approach (Fig. 14). Notably, fragments CSR35F1 and CSR35F2 targeting the allosteric back pocket also inhibited RIPK2 suggesting potential for investigation of type III inhibitors that engage the activation loop (Fig. 14).
Fig. 14. Chemical structures and IC50 values for RIPK2 inhibitors derived from regorafenib.
Findings from the CSLP inhibitor series
The CSLP inhibitor series derives from our previous work on ALK2 kinase inhibitors (Fig. 15)(13). Compounds such as LDN-214117 showed potent inhibition of RIPK2 (IC50 = 50 nM) as well as promising selectivity against the human kinome.
Fig. 15. Discovery of pyridine derivatives as RIPK2 inhibitors. (A) Chemical scaffolds of early derivatives. (B) Kinome selectivity profile of LDN-214117 (Nanosyn, >200 kinases).
As shown previously in Fig.5, a significant number of inhibitors in the CSLP series displayed potent inhibition of RIPK2 enzymatic activity, but only some of these were potent in cells, notably CSLP37 and CSLP43 (Fig. 16).
Fig. 16. SAR of pyridine derivatives in the CSLP series comparing in vitro kinase (ADP-Glo) and cellular assays (HEKBlue NF-ĸB reporter and RIPK2 nanoBRET).
Fig. 17. CSLP37 and CSLP43 inhibit XIAP binding and HEKBlue NF-ĸB reporter. (A) Mini-SAR highlighting the importance of a larger substituent at the R1 position to achieve potency in HEKBlue cells. (B) Effective RIPK2 inhibitors such as CSLP37 and CSLP43 block binding of XIAP.
Specific comparison of CSLP18, CSLP37 and CSLP43 revealed a striking dependence of the cellular activity of these inhibitors on the size of the substituent at the R1 position (Fig. 17A). CSLP37 and CSLP43 blocked XIAP binding, whereas CSLP48 and CSLP55, which lacked cellular potency, failed to block XIAP binding (Fig. 17B). Structural modelling suggested that the critical R1 substituent fills the deep back pocket besides the Thr95 gatekeeper residue in the kinase hinge region (Fig. 18). This pocket lies adjacent to and below the mapped interaction surface for XIAP. Other efficacious RIPK2 inhibitors such as ponatinib, fill the same pocket. Thus, one can envision that inhibitor interactions at this site may interfere with the preferred β2-β3 loop conformation for XIAP interaction.
Fig. 18. CSLP43 fills a deep pocket by the gatekeeper Thr95. (A) Model of the RIPK2-CSLP43 complex. (B) Spacefill models of CSLP series inhibitors highlighting their occupation of the deep back pocket.
Fig. 19. Inhibitory activities of CSLP37 and CSLP43. Inhibition of the NF-κB SEAP reporter by (A) CSLP37 and (B) CSLP43 in THP1-Blue™ NF-κB cells (Invivogen) treated with either MDP (10 µg/mL), TriDAP (10 µg/mL), E. coli LPS (10 ng/mL), HKLM (1x107 cells/mL or Pam3CSK4 (10 ng/mL). Data represent mean ± SEM (n = 3). (C) ELISA measurement of TNF release from RAW264.7 cells treated with MDP (10 µg/mL, 24 h) and CSLP compounds. (D) Inhibition of MDP-elicited TNF release in vivo. Mice (n=6 per group) were administered i.p. with 10 mg/kg of WEHI-345 or CSLP36 30 min prior to ip injections of 100 µg/mouse of MDP. After 4 h, blood was collected by cardiac puncture and circulating levels of TNF were analyzed by ELISA.
CSLP37 and CSLP43 selectively inhibit NOD responses in cells and display potent activity in vivo
The inhibitory properties of CSLP37 and CSLP43 experiments are summarised in Fig. 19. In human THP1 monocytes containing an NF-ĸB-SEAP reporter, CSLP37 and CSLP43 selectively blocked RIPK2 signalling from NOD1 (TriDAP stimulation) and NOD2 (MDP stimulation), with no detected effect on TLR1, TLR2 and TLR4 responses (LPS, HKLM, or Pam3CSK4 stimulation, Figure 4A-B). CSLP37 and CSLP43 also blocked the downstream release of TNF in mouse RAW264.7 macrophages with similar IC50 values to those in the previous NOD2/HEKBlue assay. Finally, CSLP37 could effectively block TNF release in a MDP-challenge model in vivo.
The following antibodies from were used from Santa Cruz Biotechnology: anti-RIPK2 mouse monoclonal (clone A-10, sc-166765), goat polyclonal (sc-8610, discontinued) and rabbit polyclonal (H-300, sc-22763, discontinued). The phospho-specific antibody rabbit anti-phospho-Ser176-RIPK2 was obtained from Cell Signaling.
RIPK2 KO U2OS/NOD2 cells were prepared by our collaborator Mads Gyrd-Hansen using the CRISPR/Cas9 KO plasmids (containing gRNA, Cas9, and EGFP reporter) from Santa Cruz Biotechnology (Hrdinka et al. manuscript under review). Additionally, CRISPR/Cas9 reagents targeting human RIPK2 have been published by others (14) and are available from Addgene (plasmids #76912, #76911, #76910).
A manuscript (Hrdinka et al.) reporting that RIPK2 catalytic activity is dispensible for NOD2-RIPK2 pro-inflammatory signalling is currently under review. Further development of the CSLP inhibitor series is in progress in collaboration with Alexei Degterev and Greg Cuny with potential to generate a highly selective and potent chemical probe.
Chemistry partner: Gregory D. Cuny (University of Houston)
Kinase assays: Alexei Degterev (Tufts University)
XIAP and ubiquitination assays: Mads Gyrd-Hansen (University of Oxford)
FACS assays: Holm H. Uhlig (University of Oxford)