Identification of an antibody‐based immunoassay for measuring direct target binding of RIPK1 inhibitors in cells and tissues

Abstract Therapies that suppress RIPK1 kinase activity are emerging as promising therapeutic agents for the treatment of multiple inflammatory disorders. The ability to directly measure drug binding of a RIPK1 inhibitor to its target is critical for providing insight into pharmacokinetics, pharmacodynamics, safety and clinical efficacy, especially for a first‐in‐class small‐molecule inhibitor where the mechanism has yet to be explored. Here, we report a novel method for measuring drug binding to RIPK1 protein in cells and tissues. This TEAR1 (Target Engagement Assessment for RIPK 1) assay is a pair of immunoassays developed on the principle of competition, whereby a first molecule (ie, drug) prevents the binding of a second molecule (ie, antibody) to the target protein. Using the TEAR1 assay, we have validated the direct binding of specific RIPK1 inhibitors in cells, blood and tissues following treatment with benzoxazepinone (BOAz) RIPK1 inhibitors. The TEAR1 assay is a valuable tool for facilitating the clinical development of the lead RIPK1 clinical candidate compound, GSK2982772, as a first‐in‐class RIPK1 inhibitor for the treatment of inflammatory disease.

necrosis factor (TNF) receptor pathway, which must be tightly regulated to maintain tissue homeostasis. If dysregulated, the TNF receptor pathway may result in spontaneous and robust inflammation. 7,34 Recent work has shown that RIPK1 kinase activity can drive this inflammation through directly regulating necroptosis and proinflammatory cytokine production. 4,5,11,14 The regulatory role of RIPK1 is not limited to only TNF receptor signaling, but is a critical driver of inflammation downstream of several other receptor pathways, including FasL, TRAIL, TLR3, and TLR4. 6,23,25 Recent literature describing the use of selective RIPK1 inhibitors and RIPK1 kinasedead knock-in mice in preclinical models have highlighted the pathogenic role for RIPK1 kinase activity, particularly in diseases associated with aberrant TNF receptor signaling. 3,19,36 Hence, we believe that blocking this pathway with RIPK1 small-molecule inhibitors has the potential to result in a broad therapeutic application in multiple inflammatory diseases including psoriasis, ulcerative colitis, and rheumatoid arthritis.
A series of RIPK1 inhibitors was first discovered using a phenotypic cell screen to identify compounds that block necroptotic cell death. 9, 10 We have recently reported on a number of potent inhibitors of RIPK1 kinase activity 2,20,21 An initial DNA-encoded library screen yielded the benzodiazepinone (BOAz) series of potent and highly selective RIPK1 kinase inhibitors. 20 Optimization of this series identified GSK2982772 which has nanomolar potency, exquisite selectivity and an excellent preclinical pharmacokinetic and developability profile leading to its selection for clinical development as a first-in-class small-molecule inhibitor for the treatment of inflammatory disease. 22 As with any new area of therapeutic intervention, it is not only critical to have a high-quality molecule, but it is also important to have methods to monitor direct target engagement to properly guide dose selection and interpret pharmacodynamic effects and clinical efficacy. 27 One of the main issues in monitoring drug efficacy during early clinical development is that drug binding is difficult to measure in cells and tissues. 26 Traditionally, drug efficacy has been monitored indirectly by assessing cellular responses downstream of the target protein. This can be rather simple if the target protein has a physiological function that is directly tied to alterations in circulating factors such as cytokines or chemokines in the blood. This becomes increasingly difficult, however, when measurement of drug efficacy is required at the tissue site of action where protein functions may not be easily measured.
The evaluation of target engagement at the cellular level is quite challenging as there are relatively few established methods that translate across multiple protein targets. One methodology with potential to translate across multiple discovery programs utilizes an ELISA-based immunoassay, whereby unoccupied target protein is captured and quantified. A recent work describes such an assay whereby target engagement of Bruton's tyrosine kinase (BTK) using a covalent inhibitor, CC-292, was assessed by incubating cell or tissue homogenates with a second covalent BTK inhibitor chemicallylinked to biotin, then capturing the biotinylated probe-BTK complex on a streptavidin plate. 13 Here, we describe a novel immunoassay method for measuring the direct interaction of the The FREE-RIPK1 immunoassay is no longer able to detect RIPK1 protein likely due to structural alterations in the RIPK1 activation loop, the region of epitope recognition by the FREE-RIPK1 antibody.
In this article, we describe the identification of the TEAR1 assay and its potential application for monitoring direct target engagement of BOAz-RIPK1 inhibitors in clinical tissue samples from patients with inflammatory disease.

| Animals
Eight healthy male cynomolgus macaques (Macaca fasicularis), age 8-12 years, previously obtained from Charles River BRF (Houston, TX), Covance (Alice, TX), or Mannheimer (Miami, FL) and housed in an AAALACi-accredited facility at GlaxoSmithKline (Collegeville) were utilized for the study. These monkeys were housed indoors, maintained on a 12:12-hour light-dark cycle, fed a standard primate diet (LabDiet Certified Primate Diet 5043) plus a variety of additional fresh foods and foraging each day, and had access to ad libitum water. They were provided with toys and auditory or visual enrichment daily and acclimated to all conscious study procedures. (GSK'064), for 24 hour at 4°C in lysis buffer. Following incubation, lysates were analyzed by immunoassay using mouse anti-human RIPK1 antibody (Abcam ab72139; 1 lgÁmL À1 ) as capture and various rabbit anti-human RIPK1 antibodies for primary detection (1 lgÁmL À1 final concentration). Raw electrochemiluminescent (ECL) counts were plotted for single data points. Percent target engagement was determined based on calculated concentrations using a recombinant human RIPK1 protein (ab135220; Abcam) as a standard curve.

| TEAR1 immunoassays
Two RIPK1 immunoassays for each TEAR1 assay were performed on lysates produced from cells and tissues using the MULTI-ARRAY 96-

| Western blotting
Cell lysates (10 lg) were separated on 8% bis-tris Bolt gels (Invitrogen) following reduction and denaturation. Following transfer to nitrocellulose membranes, blots were blocked in Protein-Free TBS blocking buffer (ThermoFisher Scientific, Waltham, MA). Primary antibodies were incubated for 2 hours at room temperature in blocking buffer at a final concentration of 1:1000. Blots were washed three times in TBS + 0.05% Tween, followed by incubation with appropriate secondary antibodies. Immunoblots were read on an Odyssey Imager.

| Analytical methods for GSK'253
Analysis of blood samples from study days for GSK'253 was performed using liquid chromatography-tandem mass spectrometric (LC-MS/MS) detection. The samples were thawed, blood proteins were precipitated with 200 lL of 95/5 acetonitrile/0.1% aqueous formic acid, containing 200 ngÁmL À1 of a mass spectral internal standard (ie, Verapamil), and the resulting mixture was vortex-mixed for 2 minutes followed by centrifugation for 30 minutes at >2500g. Analytical calibration standards were prepared in monkey blood; blood proteins were precipitated as described above. A full standard curve consisting of nine different concentrations (ranging from 0.1 to 1000 ngÁmL À1 ) was prepared for GSK'253 in monkey blood. Homogenate of each tissue specimen was prepared in appropriate volume of acetonitrile with a homogenizer. The final weight of each homogenate was determined to calculate the actual homogenate weight: tissue weight ratio. 50 lL aliquots of each homogenate were FINGER ET AL. | 3 of 10 transferred to sample tubes and stored at approximately À80°C until analysis. Control tissue specimens were also collected from a control animal and a homogenate was prepared as described above. The control homogenates were used for preparation of analytical standards. The HPLC system consisted of a Waters Acquity pump, auto sampler, column oven, and online degaser (Waters, Massachusetts, MA). A 2 9 20 mm, 4 l, Synergi Hydro RP analytical column was used (Phenomenex, Torrance, CA). The mobile phase consisted of a gradient that transitioned linearly from 95% aqueous 0.1% formic acid / 5% acetonitrile to 100% acetonitrile over 1.0 minutes at 750 lLÁminutes À1 flow rate.
Unless otherwise specified, data are presented as mean AE SD. Comparisons were performed with a Student's t test whose values are represented in the figures as *P < .05, **P<.01, and ***P<.001. inhibitor-protein complex suggesting that these regions become less ordered than in the GSK'064 complex or are potentially blocked by the hinge-binding GSK'064. These structural changes in RIPK1 protein upon BOAz-RIPK1 inhibitor binding were confirmed with our recently published cocrystal structure data using GSK'481. 21

| The selective RIPK1 inhibitors, GSK2882481
and GSK3011253, compete with antibody binding to RIPK1 protein On the basis of the changes observed during our HDX-MS studies and co-crystallization efforts, we hypothesized that these changes in RIPK1 structure might be leveraged to monitor compound binding to RIPK1 protein using an immunoassay format with similarity to an assay that was developed for Bruton's tyrosine kinase. 13 To identify antibodies whereby the antibody binding is potentially altered when a RIPK1 inhibitor is present, we screened commercially available antibodies ( Figure 1A; Table S1) in a sandwich immunoassay format using a mouse anti-human RIPK1 antibody to immunoprecipitate RIPK1 protein. Surprisingly, the detection of RIPK1 protein by the CS3493 antibody (Ab5) following immunoprecipitation was decreased in a concentration-dependent manner when HT29 cell lysates were incubated with GSK'481 or GSK3011253 (GSK'253), but not with the traditional hinge-binding GSK'064 ( Figure 1B; Fig-ure S1). In comparison, the detection of RIPK1 protein by the ab125072 antibody (Ab3) was unaffected following incubation with all three compounds ( Figure 1C). Depicted in Figure 1D, based on these initial findings, antibody CS3493 recognizes the RIPK1 protein within the RIPK1 activation loop in the unbound state, when RIPK1 is "FREE" of the RIPK1 inhibitor. When a BOAz-RIPK1 inhibitor is F I G U R E 1 Identification of a target engagement antibody for RIPK1. (A) Comparison of RIPK1 antibodies (Ab1-Ab7) with changes observed in HDX-MS experiment between GSK'481-bound protein and GSK'064-bound protein. Ab3 is abcam 125072. Ab5 is Cell Signaling 3493. Ab3, Ab4, Ab6, and Ab7 are shown, though they do not overlap with the RIPK1 (1-324) truncated protein used in HDX-MS. These antibodies recognize a region at the RIPK1 C-terminus from AA325-671. The exact epitopes of these antibodies remain unknown. (B) HT-29 cell lysates (20 lg) were incubated with GSK'481, GSK'253, or GSK'064 at concentrations ranging from 1 lmolÁL À1 to 1 pmolÁL À1 for 24 hours. Cell lysates were analyzed by immunoassay using Abcam ab72139 mouse monoclonal antibody for RIPK1 capture and Ab5 (Cell Signaling 3493) at 1 lgÁmL À1 final concentration. Raw ECL counts were plotted against log molar concentrations of RIPK1 inhibitor. Samples were screened in single well format. (C) HT-29 cell lysates (20 lg) were incubated with GSK'481, GSK'253, or GSK'064 at concentrations ranging from 1 lmolÁL À1 to 1 pmolÁL À1 for 24 hours. Cell lysates were analyzed by immunoassay using Abcam ab72139 mouse monoclonal antibody for RIPK1 capture and Ab3 (Abcam ab125072) at 1 lgÁmL À1 final concentration. Raw ECL counts were plotted against log molar concentrations of RIPK1 inhibitor. Samples were screened in single well format. (D) Schematic representation of target engagement model demonstrating "competition" of FREE-RIPK1 antibody by RIPK1 inhibitors of the BOAZ chemical class. ECL, Electrochemiluminescent FINGER ET AL. | 5 of 10 bound to RIPK1 protein, the epitope for the CS3493 antibody is no longer accessible for binding. On the other hand, antibody ab125072 (TOTAL) recognizes RIPK1 protein irrespective of inhibitor binding. This assay has been termed the TEAR1 assay.

| Engagement of RIPK1 can be detected in cells using the TEAR1 assay
We sought to perform a similar set of compound experiments in HT29 cells using the TEAR1 assay. We focused on the GSK'253 RIPK1 inhibitor as it has better oral pharmacokinetic properties than GSK'481. As shown in Figure 2A

| Measuring RIPK1 Target Engagement in
cynomolgus monkeys using the TEAR1 assay As the BOAz class of RIPK1 inhibitors has a consistent species selectivity for inhibition of RIPK1 in primates over nonprimates, 21  Based on the measured drug concentrations in the blood, RIPK1 target engagement was predicted using the IC 50 of 3.1 ngÁmL À1 for GSK'253, as determined in previously established cell-based assays (data not shown). Immediately after infusion, blood-drug levels had an observed Cmax of 197 AE 72 ngÁmL À1 ( Figure 3B). Maximal target engagement was predicted to reach 98.3 AE 0.6% based on the observed Cmax with predicted levels decreasing to 14.6 AE 8.1% by 48 hours postinfusion ( Figure 3C). Observed RIPK1 target engagement closely matched the predicted target engagement based on drug PK ( Figure 3C; Table S2). GSK'253 demonstrated an IC 50 of 3.1 ngÁmL À1 matching that which was predicted from cell-based assays ( Figure 3D).
To better define the utility of the TEAR1 assay in a clinically relevant target tissue associated with psoriasis, we analyzed skin biopsy samples and compared the observed target engagement to predicted target engagement based on tissue PK. Immediately after infusion, drug levels in the skin had an observed Cmax of 108 AE 39 ngÁmL À1 ( Figure 4A, Table S3). Maximal target engagement was predicted to reach 96.9 AE 1.0% based on the observed Cmax with predicted levels decreasing to 25.8 AE 17.9% by 48 hours postinfusion (Figure 4B). Observed target engagement in skin biopsies correlated well with predicted target engagement based on tissue PK (Table S3) To expand our tissue target engagement analysis beyond skin samples, we evaluated RIPK1 target engagement in colon and synovial tissues, as these tissues are the major sites of inflammation in TNF-dependent diseases such as ulcerative colitis and rheumatoid arthritis, respectively. Colon and synovial tissue samples were isolated from each cohort (n = 2/cohort) and analyzed using the TEAR1 F I G U R E 2 RIPK1 target engagement in HT-29 cells using the TEAR1 assay. (A) HT29 cells were incubated with either GSK'253 or GSK'064 for 24 hours. Cells lysates (20 lg) were analyzed for RIPK1 target engagement using both FREE-RIPK1 and TOTAL-RIPK1 immunoassays. FREE-RIPK1 levels were normalized to TOTAL-RIPK1 levels. Data are represented as the percent target engagement AESD; n = 3 replicates per group. *P < .05, **P < .01, and ***P < .001. (B) HT29 cell lysates (10 lg) was analyzed by western blot for RIPK1 (CS3493, ab125072, and ab72139) and normalized to actin assay. After infusion, observed drug levels in the colon ( Figure 4C and Table S4) and synovium ( Figure 4D and While our experiments to date have shown good correlation between predicted (ie, PK drug levels) and measured (ie, TEAR1 assay) target engagement, we have yet to explore whether this correlation holds in disease settings. It is known that RIPK1 can become extensively modified by phosphorylation and ubiquitination following activation of upstream signaling pathways 8 ; 30 and these modifications could change either the affinity of inhibitors for RIPK1 or the ability of the TEAR1 assay antibodies to detect RIPK1.
Moving forward, the TEAR1 assay will be used in early clinical development to validate its utility to monitor target engagement in blood during phase 1 studies and to monitor target engagement in disease-relevant target tissues in future phase 2/3 proof-of mechanism and proof-of-concept studies. Having the ability to quantify target engagement in these studies using the TEAR1 assay will be invaluable in interpreting clinical studies with these novel RIPK1 inhibitors and understanding the relationship between target engagement, pathway inhibition, and clinical response.
F I G U R E 4 Target engagement of GSK'253 in cynomolgus monkey tissues. (A) Skin biopsy-drug concentrations at each time point following IV administration of 0.12 mgÁkg À1 GSK'253. Data are represented mean drug concentration AE SD; n = 2 monkeys per group. *P < .05, **P < .01, and ***P < .001. (B) Comparison of predicted and observed RIPK1 target engagement in skin biopsies following IV administration of GSK'253. Predicted target engagement was calculated using the known IC50 of 3.1 ngÁmL À1 in a monkey whole-blood challenge assay, assuming a hill slope of 1. Data are represented as the percent target engagement AE SD; n = 2-8 animals per group. *P < .05, **P < .01, and ***P < .001. (C) Comparison of predicted and observed RIPK1 target engagement in terminal colon tissue following IV administration of GSK'253. Predicted target engagement was calculated using the known IC50 of 3.1 ngÁmL À1 in a monkey whole-blood challenge assay, assuming a hill slope of 1. Data are represented as the percent target engagement AESD; n = 2 animals per group. (D) Comparison of predicted and observed RIPK1 target engagement in terminal synovium tissue from knee joints following IV administration of GSK'253. Predicted target engagement was calculated using the known IC50 of 3.1 ngÁmL À1 in a monkey whole-blood challenge assay, assuming a hill slope of 1. Data are represented as the percent target engagement AESD; n = 2-8 animals per group