Inhibition of MALT1 and BCL2 Induces Synergistic Antitumor Activity in Models of B-Cell Lymphoma

Non-Hodgkin lymphoma (NHL), a neoplasm of the lymphoid tissues, is the most common hematologic malignancy worldwide (1) and predominantly arises from mature B cells (2). The most common mature B cell neoplasm is diffuse large B-cell lymphoma (DLBCL) that accounts for approximately 30% of NHL (3). While DLBCL is a heterogenous blood cancer, it can be broadly classified into two main subtypes, germinal center B cell-like (GCB) and activated B cell-like (ABC). GCB and ABC subtypes have quite divergent clinical courses, with ABC-DLBCL being a particularly aggressive subtype associated with an elevated risk of relapse and resistance to treatment (4). In addition, patients with ABC-DLBCL often present with advanced-stage disease and a higher tumor burden, which further contributes to poor prognosis (5, 6). The current standard of care for ABC-DLBCL includes high-dose chemoimmunotherapy, but these treatments are associated with significant toxicity and often fail to achieve long-term remissions (7). Therefore, the development of new therapies that can improve the outcomes of patients with ABC-DLBCL remains a key unmet need.

ABC-DLBCL is characterized by constitutively active NF-κB signaling downstream of autoreactive B-cell receptor (BCR) and toll-like receptor 9 (TLR9) complexes (8). These trigger activation of the NF-κB signaling pathway via BTK and the downstream CARD11/BCL10/MALT1 (CBM) complex. When signal transduction is stimulated, MALT1 proteolytically cleaves negative regulators of the NF-κB resulting in their inactivation and pathway reinforcement (9–11). Autoprocessing of MALT1 has also been reported to contribute to NF-κB activation (12). Activation of the CBM complex is critical for the survival of ABC-DLBCL and can be further potentiated by highly recurrent CD79A/B, MYD88, CARD11, and BCL10 mutations (13–16). Targeting of constitutively active BCR signaling is an attractive prospect for ABC-DLBCL but has had mixed success. While mutations upstream of BTK such as CD79A/B and MYD88 predict sensitivity to BTK inhibitors, downstream mutations in CARD11 or BCL10 circumvent BTK dependency in ABC-DLBCL models. Therefore, catalytic inhibition of MALT1 protease activity has been hypothesized to effectively treat these genetic subtypes unresponsive to BTK inhibitors (17). Over the past decade, there have been numerous efforts to generate MALT1 inhibitors either directly against the enzymatic pocket (18–20) or through an allosteric pocket discovered by a high-throughput drug discovery campaign (21, 22).

As ABC-DLBCL represents a clinically challenging subtype of NHL, new treatment regimens may require combination therapy to provide meaningful patient benefit. Rational combinations of targeted therapies are an appealing option to treat cancers more effectively where pathways and molecular mechanisms of the individual targets are known to intertwine. Here we describe the discovery and characterization of a novel allosteric MALT1 inhibitor, ABBV-MALT1, that displays selective and potent monotherapy activity in ABC-DLBCL preclinical models. We further demonstrate that MALT1 inhibition confers a heightened dependency on the antiapoptotic protein BCL2 for survival of these ABC-DLBCL models. This led to the hypothesis that combination of ABBV-MALT1 and the BCL2 inhibitor venetoclax would provide added benefit, and the data we present demonstrate dramatic antitumor activity with this combination in preclinical ABC-DLBCL models. These results set the stage for a rational and clinically actionable novel chemotherapy-free targeted combination strategy that shows significant potential for the treatment of patients with ABC-DLBCL.

Compounds dissolved in DMSO were predispensed into 384-well ProxiPlates (PE 6008289) using an ECHO. Three microliters of 6 nmol/L MALT1 (expressed and purified in-house) in assay buffer [25 mmol/L HEPES pH 7.5, 1 mmol/L EDTA, 0.8 mol/L sodium citrate, 0.005% bovine serum albumin (BSA), 2 mmol/L DTT] was added to the compound plates. MALT1 was incubated for 40 minutes at room temperature with compound and then the reaction was initiated by the addition of 2 mmol/L peptide substrate (TAMRA-LVSRGAAS-QSY7). The reaction was incubated for 60 minutes at room temperature and then fluorescence measured using a Perkin Elmer EnVision plate reader (Ex.535/Em.590).

Human MALT1-containing residues 334–719 with N-terminal His tag and an incorporated TEV cleavage site was expressed in HEK293 cells. The protein was initially purified using affinity capture of the His tag which was then removed by enzymatic cleavage. It was then further purified by size-exclusion chromatography (SEC) and then concentrated to 15 mg/mL in a final buffer of 20 mmol/L Tris, 300 mmol/L NaCl, 0.5 mmol/L TCEP, pH 8.0. For crystallization, a ternary complex of the protein, covalent peptide (Z-VRPR-fmk) and compound were crystallized by vapor diffusion at 17°C using a reservoir of 16% (w/v) PEG 8000, 0.1 M sodium citrate pH 5.5. Crystals were harvested using the reservoir as the cryoprotectant and cryocooled in liquid nitrogen. Diffraction data was collected at the Advance Photon Source (APS) beamline 17-ID under gaseous nitrogen at 100K.

The binding kinetics of MALT1 inhibitor was evaluated via surface plasmon resonance (SPR) using a Biacore T200 instrument (Cytiva Life Sciences). A CM5 chip was docked and primed in HBS-P+ buffer (10 mmol/L HEPES. 150 mmol/L NaCl, 0.05% v/v Tween-20) followed by injecting a mixture 50 mmol/L NaOH and 1 mol/L NaCl (preconditioning CM5 chip) at a flowrate of 10 µL/minute for 30 seconds. Then, a neutravidin surface was prepared by standard amine coupling procedure by injecting a mixture of 0.4 M N-ethyl-N-(3dimethylaminopropyl) carbodiimide and 0.1 mol/L N-hydroxysuccinimide for 7 minutes followed by injecting neutravidin (50 µg/mL in acetate buffer pH 5.5) for 7 minutes to reach a density of 12,000–15,000 response units (RU). The surface was then deactivated by injection of 1 mol/L ethanolamine for 7 minutes. Biotinylated MALT1 was prepared by incubating a mixture of EZ-Link Sulfo-NHS-LC-LC-Biotin (Thermo Fisher Scientific, catalog no. 21338) and MALT1 (1:2 ratio) on ice for 30 minutes. The reaction mixture was quenched with 50 mmol/L Tris pH 8.0 and then desalted in 10 mmol/L HEPES pH 5.5, 150 mmol/L NaCl and 1 mmol/L TCEP. Biotinylated MALT1 (residues 334–719) was then captured onto neutravidin surface (flow cell 2) to achieve Rmax levels of 20–30 RUs of the analyte (inhibitor) in HBS-P+ buffer at a flowrate of 5 µL/minute. A reference surface (without MALT1) was prepared by injecting biotin solution over neutravidin surface on flow cell 1. Binding kinetics were measured at a flow rate of 90 µL/minute using HBS-P+ containing 1 mmol/L TCEP and 1% DMSO. Compounds were diluted serially fivefold five times with a total of five concentrations. Compounds were titrated using single-cycle kinetics mode (Cytiva Life Sciences) with a contact time of 60 seconds for each analyte concentration and a dissociation time of 2,000 seconds. Sensorgrams were double referenced by subtracting the signal from a reference channel (flow cell 1) alone and a buffer blank. Binding data were processed using Biacore T200 Evaluation Software, and the binding curves were fit to a 1:1 binding kinetic model to determine the binding rate constants K_on (on-rate) and K_off (off-rate) and the K_D (equilibrium dissociation constant).

Cell lines were obtained from the DSMZ-German Collection of Microorganisms and Cell Cultures GmbH (DSMZ). All cell lines were grown and treated in Iscove’s modified Dulbecco’s medium (IMDM; Sigma) supplemented with 20% human serum (HS; Sigma) and maintained in a humidified incubator at 5% CO₂ and 37°C. Cells were treated with ABBV-MALT1 (1.5 nmol/L–30,000 nmol/L) for 120 hours at 37°C, 5% CO₂ humidified incubator and cell proliferation was determined using CellTiterGlo Luminescent assay (Promega) according to the manufacturer’s instructions and the effective concentration (EC₅₀) to inhibit 50% cellular proliferation was determined by nonlinear regression algorithms using Prism 8.0 (GraphPad). Cell line authentication was done by short tandem repeat analysis (STR) using the PowerPlex 1.2 STR system (Promega, #DC2408). All cell lines were verified as Mycoplasma free monthly using MycoAlert detection kit (Lonza, #LT07). All cells were maintained in culture for less than two months or 12 passages.

Mixed lymphocyte reaction was performed as described previously (23). Briefly, monocyte-derived dendritic cells (MoDC) were obtained from peripheral blood mononuclear cells (PBMC) differentiated with GM-CSF and IL4 for 120 hours followed by activation with IL1α and TNFα. On day 7, the MoDCs were harvested and cocultured with allogeneic CD4 T cells at a ratio of 1 MoDC:10 CD4 T cells in a 96-well round bottom plate. The T cells and MoDCs were cocultured for 5 days after which supernatant was harvested. IFNγ was measured using AlphaLISA.

Mouse plasma from in vivo experiments or Jurkat cells cultured in RPMI medium (Gibco) or human PBMCs immune cells cultured in AIM-V medium (Gibco) were pretreated with ABBV-MALT1 for 1 hour and then stimulated with 1× PMA/ionomycin (Sigma) overnight. Cell media containing secreted cytokines were diluted 1:6 and added to a precoated proinflammatory panel multi-spot 96-well plate (Meso Scale Discovery; MSD) and incubated overnight at 4°C with rotation at 900 rpm to capture IL2/IL10. After three washes with 1× wash buffer (MSD), sulfo-tagged anti-IL2/IL10 antibodies (MSD) were added and incubated for 2 hours at room temperature with rotation at 900 rpm. After three washes with 1× wash buffer, 150 µL of 2× MSD read buffer T were added per well and fluorescence was measured using MESO QuickPlex SQ 120.

OCI-LY3 and SUDHL-10 cells cultured in IMDM (Sigma) supplemented with 20% HS (Sigma) were treated with ABBV-MALT1 for 24 hours. Proteins were extracted in CellLytic M lysis buffer (Sigma) containing protease and phosphatase inhibitors. Protein samples (0.5 µg/mL) were prepared following Protein Simple Western protocol and probed for BCL10 (Abcam), RelB (Cell Signaling Technology), Regnase-4 (AbbVie Internal), and GAPDH (Sigma) using Protein-Simple Wes Imager.

Mouse plasma from in vivo experiments or OCI-LY3 and SUDHL-10 cells cultured in IMDM (Sigma) supplemented with 20% HS (Sigma) were treated with ABBV-MALT1 for 24 hours, and proteins were extracted in CellLyticM lysis buffer (Sigma) containing protease and phosphatase inhibitors. Protein samples were subsequently added to a small spot high-bind 96-well plate (MSD) coated with anti-BCL10 mouse mAbs (Abcam) and incubated overnight at 4°C with rotation at 900 rpm to capture uncleaved BCL10. After three washes with PBS-tween (PBST), anti-BCL10 rabbit mAbs (Abcam) were added and incubated for 1 hour at room temperature with rotation at 900 rpm, followed by an additional three washes with PBST. Goat anti-rabbit SULFO-TAG secondary antibodies (MSD) were added and incubated at room temperature for 45 minutes with rotation at 900 rpm. Finally, after three washes with PBST, 150 μL of 2× MSD read buffer T were added per well and fluorescence was measured using MESO QuickPlex SQ 120.

OCI-LY3 cells were centrifuged, washed with cold PBS, and snap frozen. PreOmics iST-NHS kit was used for the lysis, digestion, and desalting of peptides following manufacturer’s protocol. After lysis, protein concentration was determined using bicinchoninic acid (BCA) assay. Fifty micrograms of protein from each treatment condition was digested with PreOmics Digestion Mix for 2 hours at 37°C. Two-hundred micrograms of TMT10plex reagent (in 50 μL of dry acetonitrile) was added to samples and incubated at 25°C with shaking at 1,000 rpm for 2 hours. Hydroxylamine (50% in water) was added to a final concentration of 0.5%. PreOmics STOP buffer was added to mix and the samples were desalted using the iST cartridge. Eluted, TMT-labeled peptides were mixed in equal ratio and dried. TMT-labeled peptides were then fractionated by high pH (HpH) fractionation and concatenated into 12 fractions (24).

Samples were analyzed using a Thermo Scientific Orbitrap Fusion Lumos tribrid MS instrument. Two microliters of peptide samples were injected using a Thermo Fisher Scientific EASY-nLC 1200 onto a 75 µm × 50 cm 2 µm, 100A EasySpray PepMap RSLC C18 analytic column (Thermo Fisher Scientific, ES903). 0.1% formic acid in water (Optima LC-MS grade) as mobile phase A, and 0.1% formic acid in 80% MeCN (Optima LC-MS grade) as mobile phase B were utilized as solvents. Peptides were separated using a linear gradient from 6% to 32% B over 166 minutes at a flow rate of 250 nL/minute. The Orbitrap Fusion MS was operated in a data-dependent manner in positive ion mode. MS1 full scan was performed at 120,000 resolution in the Orbitrap scanning from m/z 375–1,500. AGC target setting was 4.0e5 and maximum IT was 50 ms. MIPS was set to Peptide, Intensity Threshold was set to 5.0e3, charge state 2–7 were selected and dynamic exclusion was set to 60 seconds with n exclusion after 1 and a high/low mass tolerance of 10 ppm. Data-dependent (Cycle Time: 3 sec) MS2 in the Linear Ion Trap was performed using Turbo scan rate selected. Quadrupole isolation was enabled with an isolation window of m/z 0.7, CID activation set to a collision energy of 35%, with activation time of 10 ms and Q of 0.25. AGC target setting was 7.5e3 and maximum IT was 50 ms. The top 10 fragment ion precursors from each MS2 scan were selected for MS3 analysis (synchronous precursor selection), in which precursors were fragmented by HCD (NCE 65%) prior toanalysis in the Orbitrap with max AGC of 5.0e4, maximum IT of 105 ms, isolation set to 2.0 kDa, and resolution of 50,000. Lock mass for 445.12 m/z was enabled for all analyses (25).

Peptide identification and quantification were performed using MaxQuant (version 2.3.1.0) with its built-in search engine Andromeda. Tandem mass spectra were searched against the human reference proteome. The quantitation type was specified as Reporter Ion MS3 and TMT10plex was selected. PreOmics iST-NHS modification of cysteine (113.084 – C6H11NO) was set as a fixed modification while oxidation of methionine and N-terminal protein acetylation were set as variable modifications. Trypsin/P was specified as the protease with up to two missed cleavage sites allowed. Precursor tolerance was set to ±4.5 ppm, and fragment ion tolerance to ±20 ppm. Results were adjusted to 1% FDR on peptide spectrum match (PSM) level employing a target decoy approach using reversed protein sequences. Resulting data was analyzed using R. Data was filtered to remove contaminant proteins, reverse hits and proteins that had more than three missing values across channels. Protein abundances were log₂ transformed. Remaining missing values in the dataset were imputed by random selection from a gaussian distribution. Abundances were quantile normalized. Significant changes comparing the relative protein abundance between samples were assessed by two-side moderated t test as implemented in the limma package. A protein was considered a “hit” if it met our predetermined “hit” threshold of P < 0.01 and fold change greater than 3.5 times the SD of all log₂ fold changes.

RNA-sequencing (RNA-seq) data was trimmed with Trimmomatic, with quality control performed by FastQC before and after trimming. The trimmed data were aligned to hg38 using STAR (2-pass). Read counting was done with Htseq-count from HTSeq. Prefiltering was applied to the read count matrix to keep only rows that have at least 10 reads total. Differential gene expression analysis was performed using DESeq2 R package between ABBV-MALT1 (6 or 18 hours) and matched DMSO control samples. The upregulated and downregulated differentially expressed genes (FDR < 0.05, fold-change > 1.5) after 6 hours and 18 hours of treatment were compared with identify overlapping genes with increased or decreased expression at each time point, respectively. Hypergeometric gene set enrichment analysis was performed using hypeR R package for genes that were down-regulated at both timepoints, using the SignatureDB gene set database (https://lymphochip.nih.gov/signaturedb/) that contains signatures that are relevant to B cells and B-cell lymphoma.

In-house developed and MSD-customized 96-well four spots plates coated with capture antibody of BCL2, BCL-xL, and MCL1 was blocked with 150 μL 3% blocking buffer in washing buffer for 1 hour with shaking at room temperature. Plates were washed three times with 1× Tris wash buffer. Samples of cell or tissue lysate in MSD Tris lysis buffer plus protease inhibitors were loaded to each well in duplicates along with the diluted calibrator standards for both total proteins and the complex. Plates were incubated overnight at 4°C. After three washes, Sulfo-tag-labeled anti-BCL2, BCL-xL, MCL1 antibody, or anti-BIM antibody were added to each corresponding well and incubated for 1 hour in the dark at room temperature with shaking. Plates were washed three times and 150 μL of MSD GoldTM read buffer was added to each well. Electrochemiluminescent signal was measured with MSD MESO QuickPlex SQ120 microplate reader (MSD). Data were presented in BCL2 and MCL1 protein total and BCL2-BIM MCL1-BIM complex, and their ratio.

All animal studies were conducted in accordance with the guidelines established by internal Institutional Animal Care and Use Committees (IACUC) at AbbVie. The experimental designs described herein were approved and monitored by IACUC throughout the research study. OCI-LY3 and U-2932 were obtained from the DSMZ-German Collection of Microorganisms and Cell Cultures GmbH (DSMZ). HBL-1 (Nozawa et. al. 1988), OCI-LY3 and U-2932 cells were maintained in RPMI1640 medium (Invitrogen) supplemented with 10% heat-inactivated FBS (Hyclone). OCI-LY10 cells were maintained in IMDM (Invitrogen) supplemented with 10% FBS (Hyclone).

For OCI-LY3, OCI-LY10, HBL-1, and U-2932 studies, female NSG mice were obtained from Jackson Laboratories. All experiments were conducted in compliance with AbbVie’s Institutional Animal Care and Use Committee and the NIH Guide for Care and Use of Laboratory Animals guidelines in a facility accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care.

Mice bearing OCI-LY3 tumors were derived by inoculating mice subcutaneously with tumor brei material from earlier passaged tumors (which in turn, had been derived by inoculation of cells), to provide a consistency in this model. For OCI-LY10, U-2932 and HBL-1 tumor bearing mice, cells prepared from tissue culture (2–3 passages) were utilized. Mice were injected subcutaneously 5 × 10⁶ cells or the tumor brei material in 0.1 mL of the inoculum preparation in 1:1 Matrigel (BD). For PDX models, mice were implanted subcutaneously by trocar with tumor fragments, derived from serial passaged tumors, which were processed to approximately 30 mm³ in size.

During the initial staging phase, tumor volume was monitored by use of calipers until the calculated volume reached approximately 180–200 mm³. Measurements of the length (L) and width (W) of the tumor were taken via electronic caliper and the volume was calculated according to the equation: V = L × W²/2. At that time, mice were assigned to study groups based on a matched distribution of tumor volumes so that the mean tumor volume within all study groups was similar. Dosing was initiated within 24 hours of reallocation of mice into treatment groups.

All compounds were dosed by oral gavage. ABBV-MALT was formulated in 10% ethanol, 30% PEG400, and 60% Phosal 53MCT. ABBV-MALT was dosed twice daily × 14 days or twice daily × 21 days, as indicated in the figure legends. Venetoclax was formulated in 10% ethanol, 30% PEG400 and 60% Phosal 50 PG, and was dosed every day × 14 days or every day × 21 days, as indicated in the figure legends. The total volume administered was adjusted to 0.2 mL/dosing event. Twice daily doses were administered 12 hours apart.

Throughout the course of the studies, tumors were measured and animal body weights were determined twice per week. Individual animals were removed from the study when their tumor volume reached 2,000 or 3,000 mm³, depending on the model or were removed upon animal health issues dictated by IACUC protocols. For studies utilizing OCI-LY3, OCI-LY10, and HBL-1, the most common health-related reason for removal of mice from the study was the onset of hind limb paresis, a common issue associated with DLBCL flank tumor studies.

The parameter of drug efficacy is tumor growth inhibition, comparing tumor growth of treated mice to that in vehicle-treated mice. It is calculated as %TGI = 1 − (mean tumor volume of treatment group/mean tumor volume of treatment control group) × 100. P values are derived from Student t test comparison of log-transformed data of treatment group versus treatment control group.

To create the patient-derived xenograft (PDX) models, 6- to 8-week-old NSG mice (the Jackson Laboratory) were anesthetized with 5% isoflurane vaporizer and administered with sustained release buprenorphine, followed by being implanted with the primary B-cell lymphoma core biopsy in the renal capsule. Specifically, mouse kidney was exposed by a small nick incision (∼0.5 cm) posterior to the curvature in the spine, and subsequently applied with toothed forceps to make a 1 mm tear in the kidney capsule. A pocket between the capsule and kidney was then made using a glass micro-dissector to accommodate the core biopsy. After implantation, F1 tumor generally required 8–12 weeks to be palpable. MRI was used to monitor engraftment and tumor burden. After successful generation of a PDX model (F1) using renal capsule implantation, these tumors are reimplanted into a subsequent generation of NSG mice to generate F2 tumors. Freshly harvested F2 PDX tumor samples were immediately processed and viably cryopreserved for in vivo efficacy studies.

Numerous additional PDX models were also used in these studies (LY-24–0017, LY-24–0051, LY-24–0063, LY-24–0179, LY-24–0236, LY-24–0262 and LY-24–0338) are identified as NHL PDX lineages. These models are the proprietary assets of WuXi AppTec and studies for this report were done by WuXi under contractual agreement in their facilities. These models were maintained by serial passage in mice.

For PDX studies, female NOD-SCID mice were obtained from Vital River Laboratory Animal Technology Company. To generate fresh seed tumors for in vivo efficacy studies, viably cryopreserved F2 tumor fragments were implanted into the renal capsule of NSG mice and expanded for 1 month, or until a palpable tumor was detected. Seed tumors were then harvested and processed into 2 × 2 × 2 mm³ tumor fragments and implanted subcutaneously into the flank of NSG mice and monitored for expansion until the average tumor volume reached 75 mm³. Mice (n > 7 per group) were treated twice daily with: (i) vehicle control, (ii) ABBV-MALT1, (iii) venetoclax, (iv) ABBV-MALT1 + venetoclax. ABBV-MALT1 inhibitor was administered 30 mg/kg twice daily (60 mg/kg/day) and venetoclax administered 50 mg/kg every day by oral gavage, daily for 21 days. Body weight and tumor volume were measured a minimum of two times per week. All mice were humanely euthanized at the end of treatment (day 21) or when tumor volume exceeded 1,500 mm³.

ABBV-MALT1 was run in Eurofins Cerep In Vitro Pharmacology Panel. Data are presented as percent inhibition of control values or percent inhibition of control-specific binding. Assays were performed in duplicate at 10 µmol/L against all enzymes and receptors. Compound binding was calculated as a % inhibition of the binding of a radioactively labeled ligand specific for each target. Compound enzyme inhibition effect was calculated as a % inhibition of control enzyme activity.

The enzymatic reactions were conducted at BPS Bioscience. Assays were run in duplicate at room temperature for 30 minutes in a 50 μL mixture containing 10 mmol/L HEPES buffer, pH 7.4, 10 mmol/L EDTA, 0.05% Chaps, 5 mmol/L DTT, 1 µmol/L substrate, a caspase enzyme and a test compound. Fluorescence intensity was measured at an excitation of 380 nm and an emission of 505 nm using a Tecan Infinite M1000 microplate reader. The fluorescent intensity data were analyzed using the computer software, Graphpad Prism. In the absence of the compound, the fluorescent intensity (Ft) in each data set was defined as 100% activity. In the absence of Caspase, the fluorescent intensity (Fb) in each data set was defined as 0% activity. The percent activity in the presence of each compound was calculated according to the following equation: %activity = (F-Fb)/(Ft-Fb), where F = the fluorescent intensity in the presence of the compound.

The data generated in this study are available within the article and its Supplementary Data files. Coordinates and structure factors for the crystal structure have been deposited at the RCSB PDB under accession number 8V4X.

A medicinal chemistry campaign began with systematic optimization of a diaryl urea hit resulting in the discovery of ABBV-MALT1 (ref. 26; Fig. 1A). To determine the binding orientation, an X-ray cocrystal structure of ABBV-MALT1 bound to MALT1 was solved at 2.49 Å resolution. This structure reveals an allosteric binding pose anchored by key electrostatic interactions between the urea at E397 and triazole moieties with N393 (Fig. 1B; Supplementary Table S1), similar to the binding pose of other related diaryl-urea allosteric MALT1 inhibitors (21). As measured in a SPR assay, ABBV-MALT1-bound purified MALT1 protein with a K_D of 37 nmol/L and displayed a slow off-rate (Fig. 1C; Supplementary Table S2). This tight binding of ABBV-MALT1 to MALT1 translated to potent biochemical inhibition of MALT1 protease activity, EC₅₀ = 349 nmol/L (Fig. 1D). To confirm selectivity of ABBV-MALT1, we tested the compound against a panel of 177 kinases representing each branch of the kinome, 55 pharmacologic receptor binding assays and four caspase activity assays (Supplementary Data S1). These data demonstrate that there is no meaningful off-target activity detected of ABBV-MALT1 at 10 µmol/L in these assays.

The MALT1 protease is the key member of the CBM complex responsible for integrating upstream signals from B- and T-cell receptors through regulation of NF-κB gene expression programs and previous selective MALT1 inhibitors have demonstrated the ability to inhibit lymphocyte activation (refs. 27, 28; Fig. 1E). To first explore the potential of ABBV-MALT1 to inhibit TCR activation, vehicle or ABBV-MALT1-treated Jurkat (T-cell leukemia) cells were stimulated overnight with phorbol 12-myristate 13-acetate (PMA) and robust IL2 secretion (∼100 pg/mL) was observed in the supernatant of vehicle-treated Jurkat cells, while cells pretreated with 3 µmol/L ABBV-MALT1 had no detectable cytokine (Fig. 1F). Using human PBMCs from two different healthy donors, we repeated the ABBV-MALT1 pretreatment with PMA stimulation in a dose-response fashion. ABBV-MALT1 effectively inhibited immune cell activation as measured by secreted IL2 with approximate EC₅₀s of ∼300 nmol/L in both donors (Fig. 1G).

We next performed MLR assays to understand the role of MALT1 in T-cell response to allogenic stimulation through the TCR. ABBV-MALT1 treatment completely inhibited the ability of T cells to express IFNγ after costimulation with 10 µg/mL anti-PD-1 (Fig. 1H) with no concurrent effect from ABBV-MALT1 on T-cell viability (Fig. 1I). These experiments demonstrate the potential for ABBV-MALT1 to effectively inhibit the activity of MALT1 biochemically and in cells.

ABC-DLBCL is characterized by hyperactivation of the BCR pathway CBM complex, MALT1 protease activity, and upregulation of NF-κB gene expression programs making these types of lymphomas addicted to MALT1 signaling (17, 19, 20). As such, we hypothesized that ABBV-MALT1 would be effective in shutting down the BCR signaling pathway in ABC-DLBCL tumor models.

To test the hypothesis that ABBV-MALT1 would show antitumor activity specifically in ABC-DLBCL cell models, we carried out proliferation assays in a panel of 20 lymphoma cell lines from both the ABC and GCB subtypes (14). ABBV-MALT1 selectively inhibited the growth of ABC-DLBCL models to varying maxima and showed no activity in GCB-DLBCL cells (Fig. 2A). A dose response of ABBV-MALT1 confirmed these observations as treatment of the OCI-LY3 tumor cell line (EC₅₀ = ∼300 nmol/L) affected proliferation but had no effect on the GCB-DLBCL SUDHL-10 cell line (EC₅₀ > 30 µmol/L) at five days (Fig. 2B). Strikingly, a PRISM proliferation assay demonstrated a strong selectivity of ABBV-MALT1, with minimal antiproliferative activity in a broader panel of 500 cell lines (Supplementary Fig. S1). It should be noted that this included minimal antiproliferative activity across other lymphoma in vitro models at five days. However, this observation is a pooled screen that requires further evaluation in dedicated studies, particularly in mantle cell lymphoma, as this tumor type has previously been suggested to have dependency on MALT1 (29, 30). Next, we assessed the mechanism of action in which ABBV-MALT1 inhibited ABC-DLBCL cell line growth. Because of the reported inhibition of the NF-κB pathway and its role in proliferation, we hypothesized that ABBV-MALT1 may be affecting the cell cycle. To test this, we monitored OCI-LY3 cells treated with ABBV-MALT1 during cell-cycle progression and observed significant accumulation of cells in Gi (Fig. 2C), suggesting cell-cycle arrest. Together, these data support the hypothesis that ABBV-MALT1 shows selective antiproliferative activity in ABC-DLBCL cell lines, while also demonstrating high general selectivity for the MALT1 protease (31).

Next, the ability for ABBV-MALT1 to inhibit MALT1 function was assessed by measuring the upregulation of the proteolytic substrates of MALT1, including BCL10 (32), RELB (9), and ZC3H12D protein levels (33). Like the antiproliferative growth effects, ABBV-MALT1 treatment of OCI-LY3 but not SUDHL-10 resulted in a dose-dependent upregulation of all three MALT1 substrates in a dose-dependent manner (Fig. 2D and E; Supplementary Fig. S2).

As ABBV-MALT1 was active in vitro, we next evaluated in vivo efficacy in an OCI-LY3 cell line-derived xenograft (CDX) model. Daily administration of ABBV-MALT1 for 14 days afforded a tumor growth inhibition (TGI) of 92% with persistent tumor stasis past the end of dosing (Fig. 2F). This tumor growth inhibition was concurrent with a significant upregulation in tumor BCL10 levels (Fig. 2G) and a significant decrease in circulating human IL10 (Fig. 2H). In addition, ABBV-MALT1 was efficacious in the OCI-LY10 CDX model (TGI of 73%), which harbors an activating CD79A mutation (Supplementary Fig. S3), and also resulted in tumor regression (TGI of 96%) in the LY-24–0063 CARD11-activating mutant PDX model (Fig. 2I). In contrast, while highly responsive to ABBV-MALT1, LY-24–0063 was refractory to the covalent BTK inhibitor acalabrutinib. These preliminary in vivo results support the hypothesis that ABBV-MALT1 would have antitumor activity in BTK inhibitor sensitive and refractory models with mutations downstream of BTK. To further explore ABBV-MALT1 activity, we performed studies in five unique non-GCB PDX models with positive NF-κB signatures of known genotypes or unknown origin (Supplementary Fig. S4). In three of the models tested (LY-24–0262, LY-24–0051, and LY-24–0338), ABBV-MALT1 monotherapy resulted in tumor regression (TGI of 99%, 100%, and 99%, respectively) during the treatment period that demonstrated good durability through approximately ten days after cessation of treatment. ABBV-MALT1 administration in the LY-24–0236 and LY-24–0017 PDX models resulted in modest TGI (49%) and no efficacy, respectively. While the cause of the lower sensitivity of LY-24–0236 to ABBV-MALT1 is unclear, the LY-24–0017 model harbors an inactivating mutation in TNFAIP3 downstream of MALT1 and therefore was hypothesized to be resistant to MALT1 inhibitor treatment. Taken together, ABBV-MALT1 is a potent inhibitor of MALT1, NF-κB signaling, and growth in ABC-DLBCL cell lines and xenografts in vitro and in vivo demonstrating the potential for monotherapy activity in these tumors.

Single-agent kinase inhibitors have historically shown limited activity in patients with DLBCL, leading us to explore rational combination strategies for ABBV-MALT1. We approached this using transcriptional profiling of ABBV-MALT1 following treatment for 6 and 18 hours in OCI-LY10 cells. A total of 570 and 426 genes were significantly (FDR < 0.05, fold-change >1.1) down- or upregulated at both time points, respectively (Supplementary Data S2). As MALT1 signaling integrates through the NF-κB pathway, we observed robust inhibition of NF-κB signaling, where downregulated genes were highly enriched for NF-κB signaling pathways and included the canonical NF-κB target genes BCL2L1 (BCL-xL) and BCL2A1 (Fig. 3), suggesting that ABBV-MALT1 may tilt the balance of BCL2 family member proteins.

To validate this, we measured the abundance of BCL2 family member proteins in OCI-LY3 xenografts harvested after 72-hour treatment with ABBV-MALT1 using MSD technology (Fig. 4A). Consistent with NF-κB pathway inhibition, we observed a significant downregulation in BCL-xL protein upon ABBV-MALT1 treatment of OCI-LY3. Interestingly, we also observed significant upregulation in both BCL2 and the proapoptotic BIM protein (Fig. 4B). The ability of BIM to activate BAX and/or BAK and initiate the apoptotic cascade is typically restrained via high affinity protein-protein interactions with the various prosurvival BCL2 family proteins (34). In cells that are challenged with various stressors, including chemotherapy (35), death signaling causes prosurvival family members to sequester significant quantities of proapoptotic proteins including BIM; cells in this state are “primed for death,” as disruption of these complexes leads to release of “free” BIM and cell death (36). We therefore quantified the presence of BIM:BCL2 family protein complexes, including BIM:BCL2, BIM:BCL-xL and BIM:MCL1. Following treatment with ABBV-MALT1, we observed a significant upregulation of BCL2:BIM complex formation (Fig. 4C) and a nonsignificant change in both BCL-xL:BIM and MCL1:BIM complexes (Supplementary Fig. S5A). In addition to OCI-LY3, BCL2 and the BCL2:BIM complexes were also upregulated in the U2932 ABC-DLBCL xenograft (Supplementary Fig. S5B and S5C). It is worth noting that levels of BCL2 protein are upregulated upon BIM complex formation in these cell lines following ABBV-MALT1 treatment. This has not previously been referenced in the literature for BCL2, but is reminiscent of BIM complex formation elevating protein levels of the antiapoptotic family member MCL-1 (37). Collectively, these data support the role of ABBV-MALT1 inhibition in priming ABC-DLBCL models to treatment with BCL2 inhibitors.

Following up on the significant increase of BCL2:BIM complexes following ABBV-MALT1 treatment, we then evaluated the impact of concomitant MALT1 and BCL2 inhibitor treatment on ABC-DLBCL models. In vitro, ABC-DLBCL cell lines with wild-type TNFAIP3 (OCI-LY3, HBL-1, and U-2932) demonstrated additional antiproliferative activity when the BCL2 inhibitor venetoclax was added to ABBV-MALT1 (Supplementary Fig. S6A). Next, we tested the in vivo combination of ABBV-MALT1 and venetoclax in the OCI-LY3 and OCI-LY10 xenograft models where monotherapy activity of ABBV-MALT1 was previously observed. Remarkably, the combination therapy in both OCI-LY3 and OCI-LY10 CDX models was greater than either agent as monotherapy and resulted in a considerable number of complete regressions during and extending past end of dosing periods (Fig. 5A and B). Similar combination activity between ABBV-MALT1 and venetoclax was also observed in the HBL-1 and U-2932 xenograft models, with sustained regressions through the end of dosing (Supplementary Fig. S6B and S6C).

To extend this hypothesis further, we then evaluated these observations in various PDX models. Reliably, treatment with ABBV-MALT1 and venetoclax in PDX models from relapsed/refractory ABC-DLBCL continued to provide combination and significant tumor reductions in both the LY-24–0179 and LPMIC003 PDX models (Fig. 5C and D). We also evaluated activity in highly refractory ABC-DLBCL PDX models derived from post-CAR T-cell failure biopsies, LPMI020 and LPMI322 (Supplementary Fig. S7). Single-agent ABBV-MALT1 showed significant activity in both highly aggressive post-CAR T-cell relapse PDX models. Although these models did not demonstrate the same degree of synergy between MALT1 and BCL2 inhibitors, with LPMI020 being largely insensitive to single agent or combination venetoclax and LPMI322 being highly sensitive to venetoclax as a single agent, we observed no antagonism between the two inhibitors when combined. This suggests that the combination of ABBV-MALT1 and venetoclax will not affect the ability of either drug to inhibit MALT1- or BCL2-dependent models. Compiling both the CDX and PDX combination studies, we observed significant activity of single agent ABBV-MALT1 in 7 of 8 (87.5%) ABC-DLBCL models tested, and further additive benefit of venetoclax in 6 of 8 (75%) of these models.

Worldwide incidence of NHL has increased steadily over the past 30 years and those trends are likely to continue (38). While the incidence of NHL is rising, the advent of targeted therapy has increased survival rates significantly in several NHL subtypes over the same time frame (39). Patients with the DLBCL subset of NHL have a disparate response to treatment that is confounded by pronounced genetic and clinical heterogeneity. Following diagnosis, eligible DLBCL patients receive the standard treatment of rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP). R-CHOP is a highly effective treatment for roughly 50% of patients with DLBCL; however, 30% of the patients that achieve complete remission eventually succumb to relapse. New strategies, including CAR T-cell therapy, bispecific antibodies, and antibody-drug conjugates are currently being explored (40); however, the identification of novel and effective treatment options for patients with relapsed/refractory (R/R) B cell lymphoma remains a key unmet medical need.

The dependence of ABC-DLBCL on chronically active BCR signaling led to the hypothesis that a BTK inhibitor would be effective in treating this subset of patients (20). BTK inhibitors work by abrogating the BCR signaling pathway and are approved in chronic lymphocytic leukemia/small lymphocytic leukemia and several NHL indications including Waldenstrom macroglobulinemia and previously treated mantle cell and marginal zone lymphomas. To evaluate the efficacy of BTK inhibition in ABC-DLBCL, a phase III trial was conducted in newly diagnosed non-GCB DLBCL patients with R-CHOP with or without the covalent BTK inhibitor ibrutinib (“Phoenix” trial; ClinicalTrials.gov: NCT01855750; ref. 41). Although this trial failed to meet its primary endpoint, follow-up analyses of the data set uncovered that the patients who benefited most from this treatment were younger in age and in specific genetic subtypes of ABC-DLBCL that are associated with BCR signaling dependence (42).

As MALT1 is downstream of BTK, we hypothesized that ABBV-MALT1 would impede the growth of cell line and primary patient NHL models that are responsive to BTK inhibitors, as well as models that are resistant to BTK inhibitors via downstream gain-of-function mutations. Our data demonstrate that ABBV-MALT1 is highly active against non-GCB models of DLBCL in vitro and in vivo, including those sensitive to and refractory to BTK inhibitors or relapsed from prior lines of therapy including R-CHOP and CAR T. Moreover, we established that the downregulation of NF-κB signaling by ABBV-MALT1 led to a rebalancing of pro- and antiapoptotic proteins, resulting in an increase in BCL2:BIM execution complex levels. The reduction of the canonical NF-κB target proteins BCL-xL and BCL2A1 and increase in BCL2:BIM complex translated to significant combinatorial activity between ABBV-MALT1 and the BCL2 inhibitor venetoclax in xenograft models, suggesting a mechanism-based, rational combination opportunity in ABC-DLBCL. Importantly, even where the individual agents do not provide meaningful efficacy on their own, combination elicits striking synergy in the CDX and PDX models tested. These data provide additional evidence highlighting the critical roles of NF-κB and antiapoptotic machinery in maintaining ABC-DLBCL proliferation and survival.

From a translational medicine perspective, our in vivo pharmacology studies demonstrate that the combination of ABBV-MALT1 and venetoclax affords compelling activity in B-cell lymphoma xenografts which is highlighted by complete regressions in some models. These data complement and may expand potential clinical indications beyond the previously established effective oral combination of ibrutinib plus venetoclax in patients with high-risk and older CLL (43). Toward that end, we are currently investigating a related MALT1 inhibitor, ABBV-525, in a phase I clinical study (NCT05618028) to determine the safety/tolerability, pharmacokinetic, pharmacodynamic, and preliminary efficacy in patients with mature B-cell malignancies. On the basis of our preclinical data presented herein, future clinical studies will include combination of this MALT1 inhibitor with venetoclax for an all-oral chemotherapy-free treatment regimen in these same tumor types.

JPP, AER, VB, CD, CW, PJ, HZ, SKP, SCP, FJK, RAM, CAR, MDG, MJ, NLE, WQ, RAJ, CP, AM, JAM and WNP are employees of AbbVie. RSM was an employee of AbbVie at the time of the study. The design, study conduct, and financial support for this research were provided by AbbVie. AbbVie participated in the interpretation of data, review, and approval of the publication. R.A. Judge reports that the paper describes the discovery and development of a compound that inhibits MALT1. It also discusses combining this MALT1 inhibitor with venetoclax (which targets BCL2) to provide an effective combination therapy. Venetoclax was developed in collaboration with Genentech and approved by the FDA in 2016. Genentech, however, was not involved in this study. The article was written in collaboration with the University of Texas MD Anderson Cancer Center. They contributed to the study design, performed experiments and helped prepare the manuscript. They are listed as authors on the paper. J. Henderson reports grants from AbbVie during the conduct of the study. M.R. Green reports grants from Sanofi, Kite/Gilead, grants from Abbvie during the conduct of the study; and grants from Allogene outside the submitted work. No disclosures were reported by the other authors.

J.P. Plotnik: Conceptualization, methodology, writing-original draft, writing-review and editing. A.E. Richardson: Data curation, methodology. H. Yang: Data curation, validation. E. Rojas: Data curation, software. V. Bontcheva: Data curation. C. Dowell: Data curation. S. Parsons: Data curation. A. Wilson: Data curation. V. Ravanmehr: Data curation. C. Will: Data curation. P. Jung: Data curation. H. Zhu: Data curation. S.K. Partha: Data curation. S.C. Panchal: Data curation. R.S. Mali: Data curation. F.J. Kohlhapp: Data curation. R.A. McClure: Data curation. C.Y. Ramathal: Data curation. M.D. George: Data curation. M. Jhala: Data curation. N.L. Elsen: Data curation. W. Qiu: Data curation. R.A. Judge: Data curation. C. Pan: Data curation. A. Mastracchio: Data curation. J. Henderson: Project administration. J.A. Meulbroek: Conceptualization, resources, writing-original draft, writing-review and editing. M.R. Green: Conceptualization, resources, writing-original draft, project administration, writing-review and editing. W.N. Pappano: Conceptualization, formal analysis, supervision, validation, investigation, methodology, writing-original draft, project administration, writing-review and editing.

Use of the IMCA-CAT beamline 17-ID at the Advanced Photon Source was supported by the companies of the Industrial Macromolecular Crystallography Association through a contract with Hauptman-Woodward Medical Research Institute. Use of the Advanced Photon Source was supported by the US Department of Energy, Office of Science, Office of Basic Energy Sciences, under contract no. DE-AC02–06CH11357. AbbVie is a member of IMCA and as such helps fund the operation of this beamline.