Discovery of BMS-986408, a First-In-Class Dual DGKα and DGKζ Inhibitor that Unleashes PD-1 Checkpoint and CAR T-cell Immunotherapies

Abstract

Diacylglycerol kinase α (DGKα) and DGKζ are lipid kinases that negatively regulate T-cell signaling through diacylglycerol metabolism, making them attractive targets for next-generation immunotherapy. In this study, we report the discovery and preclinical characterization of the clinical-stage DGKα and DGKζ lipid kinase inhibitor, BMS-986408. BMS-986408 binds to the accessory subdomain of the catalytic domain and inhibits DGKα/ζ through a mechanism of action that includes competitive inhibition for the diacylglycerol substrate, subcellular translocation to the plasma membrane, and proteosome-dependent degradation. DGKα/ζ inhibition markedly improved the therapeutic benefit of PD-1 therapy by unleashing T-cell responses in the tumor while also amplifying the priming and expansion of tumor-reactive T cells in tumor-draining lymph nodes. Simultaneous inhibition of both DGKα and DGKζ was required to maximize combination benefit with PD-1 therapy. Furthermore, we observed in non-small cell lung cancer (NSCLC) patient samples that DGKα and DGKζ were broadly expressed in tumor-infiltrated T cells and that combination therapy invigorated a robust cytokine response in organotypic tumors derived from patients with NSCLC, supporting the clinical evaluation of this combination in patients with NSCLC. BMS-986408 also markedly improved CD19-targeted CAR T-cell therapy efficacy by overcoming hypofunctionality, insufficient expansion, and lack of costimulatory ligands. BMS-986408 represents a critical step toward evaluating the broad immunotherapy potential of DGKα/ζ inhibitors in patients with cancer.

Figure 1.

BMS-986408 is a potent DGKα and DGKζ lipid kinase inhibitor and degrader. A, Chemical structure of BMS-986408. B, Plots showing the inhibitory dose–response curves for BMS-986408 in recombinant DGKα and DGKζ biochemical lipid kinase assays and corresponding IC₅₀ values. C, Conversion of D4-Oloeyl-DAG to D4-Oleoyl PA in Jurkat cells treated with 0.25 μmol/L of BMS-986408. Data are represented as the mean ± SD; n = 3 per group. D, Schematic of the BMS-986408 NanoBRET target engagement assay in live cells. E, Time lapse of milliBRET (mBRET) ratio with the BMS-986408-NB590 tracer in DGKα-NanoLuc–overexpressing and NanoLuc-DGKζ–overexpressing cells with (■) or without () saturating unlabeled BMS-986408 (20 μmol/L) to normalize for specificity (top) and DGKi-NB590 binding kinetics to DGKα-NanoLuc and NanoLuc-DGKζ (bottom). Binding affinity is presented in K_d; Data are represented as the mean ± SD; n = 2 per group. F, CETSA melting curves of DGKα (top) and DGKζ (bottom) from Jurkat cells treated with () or without (●) 0.5 μmol/L BMS-986408. Data show the percent change from the 37°C baseline. G, Representative images showing the subcellular localization of YFP-tagged DGKα or DGKζ with or without BMS-986408 (0.25 μmol/L). YFP is colored in green, and nuclear staining is colored in blue. H, Quantification of BMS-986408–induced DGKα () and DGKζ (◆) plasma membrane translocation with half-maximal efficacious concentrations (EC₅₀). I, Degradation dose–response for DGKα and DGKζ in human PBMCs treated with BMS-986408 for 24 hours. β-actin is presented as a loading control. J, Rescue of BMS-986408-mediated degradation with proteosome (bortezomib, BZ) and ubiquitination (TAK-243, E1i) inhibitors. K, Schematic of the whole blood DGKi potency assay, highlighting phospho-ERK and IL2 pharmacodynamic biomarkers. L, Flow cytometry quantification of BMS-986408 phospho-ERK induction potency in whole blood T cells. The EC₅₀ value is shown for CD4⁺ (●) and CD8⁺ () T cells. Data are represented as the mean ± SD.; n = 11 per group. M, AlphaLISA quantification of BMS-986408 IL2 production from human whole blood from two donors. The EC₅₀ value is shown for each donor. (D and K, Created with BioRender.com.)

Figure 2.

BMS-986408 binds with the accessory region of DGKα and DGKζ lipid kinase domain. A, Schematic of the CRISPR base editing screen to select for DGKA or DGKZ mutations that conferred resistance to BMS-986408–mediated degradation of eGFP-DGKα or mNeonGreen-DGKζ expressed in Jurkat cells. B and C, Scatterplot of Log₂ fold change (LFC) sgRNA enrichment in DGKA and DGKZ adenine base editor scanning screens. The dotted line indicates LFC = 1.5, and validated hits are highlighted in green. AlphaFold models of DGKα and DGKζ are shown as ribbons, with C-alpha atoms of enriched residues shown as spheres. The surfaces of the docked ligands (see “Materials and Methods”) are shown to highlight the proposed binding sites. D, Validation of the base editing CRISPR screen using KI cell clones: Jurkat eGFP-DGKα cells harboring the S532P, L556P, or H606R mutations and mNeonGreen-DGKζ cells harboring the F463S, S490P, or C534R mutations were treated with BMS-986408 (0.75 μmol/L), and fluorescence signal was quantified as the mean fluorescence intensity (MFI). Each point represents a cell clone. E, BMS-986408 CETSA dose–response at 41.5°C showing that HiBit-tagged DGKα harboring the S532P, L556P, or H606R mutations was resistant to BMS-986408–mediated thermal destabilization. F, BMS-986408 CETSA dose–response at 43.1°C showing that HiBit-tagged DGKζ harboring the F463S, S490P, or C534R mutations was resistant to BMS-986408–mediated thermal destabilization. G, Closer view of docked poses with enriched residues’ side chains shown as orange sticks and validated residues’ side chains shown as green sticks. H, Electrostatic surface representation of the proposed binding sites, with the surfaces of validated residues shown in green. WT, wild-type. (A, Created with BioRender.com.)

Figure 3.

Dual DGKα/ζ inhibitor BMS-986408 unleashes PD-1 T-cell checkpoint therapy. A, Schematic of TCR signaling cascade, with TCR and CD28 providing positive signals and PD-1 and DGKα/ζ providing negative signals. B, Therapeutic efficacy of anti–PD-1, BMS-986408, or the combination therapy in SA1N, MC38, and CT26 tumor models. Each line represents tumor volume of one individual animal. n = 10 per group. The percentage of animals achieving CR is noted on each plot. C, Heatmap of differentially expressed genes from RNA-seq data from MC38 tumors at day 7 after treatment. The black and grey barcodes indicates whether the expression change is statistically different between the vehicle and combination treatment group. D, Volcano plots of RNA-seq data from the same analysis. Upregulated genes are highlighted in purple, and downregulated genes in blue; a subset of upregulated T-cell effector genes are labeled in each plot. E, Flow cytometry quantification of granzyme B+ and Ki67+ effector CD8⁺ populations in the MC38 tumors. Data were collected at day 7 after treatment initiation. F, Flow cytometry quantification of naïve (CD44⁻ CD62L⁺), effector/effector memory (E/EM; CD44⁺ CD62L⁻), central memory (CM; CD44⁺ CD62L⁺), and activated (CD69⁺; PD-1+ or Ki67+) CD8⁺ T-cell subsets in MC38 TdLNs. Data were collected at day 7 after treatment initiation. G, Flow cytometric quantification of GFP+ CD8⁺ T cells in the TdLN of MC38 tumors implanted into Nur77-GFP transgenic mice. Data were collected 24 hours after treatment with anti–PD-1, BMS-986408, or the combination. H, In vivo priming of tumor antigen–specific T cells. TRP1^high or TRP1^low transgenic CD8⁺ T cells were labeled with CTV and adoptively transferred into mice implanted with C2VTrp1 tumors. Mice were dosed with anti–PD-1, BMS-986408, or the combination treatment. Representative flow cytometry analysis of CTV dilution in adoptively transferred cells is shown. Gates delineate different generations of proliferated cells. I, Calculated proliferation index (see “Materials and Methods”) of adoptively transferred TRP1^High and TRP1^Low CD8⁺ T cells in the TdLN 5 days after treatment with either anti–PD-1, BMS-986408, or the combination therapy; n = 5 per group. Error bars represent the SD. Statistical analysis was performed using an ordinary one-way ANOVA. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Combo, combination; Ctrls, controls; SSC, side scatter.

Figure 4.

Inhibiting both DGKα and DGKζ maximizes anti–PD-1 combination benefit. A, Therapeutic efficacy of anti–PD-1 vs. the combination of anti–PD-1 with either BMS-986408, DGKα-i, DGKζ-i, or DGKα-i + DGKζ-i in the MC38 tumor model. Each line represents the tumor volume curve from one individual animal. The percentage of animals achieving CR is noted on each plot. B, Cytotoxicity evaluation of NY-ESO-1–specific effector T cells in the presence of DGKα-i, DGKζ-i, or BMS-408. All compounds were dosed at 0.1 μmol/L; n = 6 per group. C, Proliferation of human PBMCs (left) and mouse TRP1^high T cells (right) in the presence of DGKα-i, DGKζ-i, or BMS-408. All compounds were dosed at 0.1 μmol/L; n = 5 per group. D, In vivo proliferation indices of adoptively transferred TRP1^high T cells in recipient mice dosed with DGKα-i, DGKζ-i, or BMS-408; n = 5 per group. E and F, Human PBMC proliferation and IFNγ production in a matrixed combination dose–response of DGKα-i or DGKζ-i (left) with corresponding highest single agent (HSA) synergy analysis (right); n = 6 per group. G, Heatmap of phosphopeptides significantly changed in human T cells treated with dose titrations of DGKα-i, DGKζ-i, or BMS-986408 from 0.001 to 1 μmol/L. Values represent the signed effect size of the dose–response curves (see “Materials and Methods”), with purple showing increased phosphorylation and blue showing decreased phosphorylation. H, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of significantly increased phosphoproteins from (G). Top 10 pathways with −log₁₀(FDR) is presented. I, Dose effect sizes of selected phosphopeptides from the NF-κB pathway and MAPK pathway as in (G). Error bars represent the SD. Statistical analysis was performed using an ordinary one-way ANOVA. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Ctrls, controls.

Figure 5.

Translational data supporting the combination of DGKα/ζ and PD-1 inhibitors in NSCLC. A, Schematic of the translational research strategy to evaluate DGKα/ζ expression and inhibition in NSCLC patient tumor biopsies. B, Uniform Manifold Approximation and Projection (UMAP) plot of scRNA-seq data from patients with NSCLC (see “Materials and Methods”). Immune cell populations were plotted and color-coded by their corresponding signature gene expression. Right shows the expression overlay of genes of interest. C, Representative multiplexed immunofluorescence images showing the expression of CD3, CD4, CD8, PD-1, DGKα, and DGKζ in a NSCLC patient tumor biopsy. D, Dot plot summary of DGKα, DGKζ, and several additional immune checkpoint expression in NSCLC TIL subsets from 78 patients with NSCLC. Dot sizes represent log₂ cell count, and dot colors represent log₂ mIF intensity. E, Cytokine quantification in the PDOTS cultures with anti–PD-1, BMS-408, or combination treatment. F, Absolute quantification of IFNγ release in PDOTS cultures, grouped by each individual patients and further divided into responders/nonresponders based on whether anti–PD-1 and BMS-408 combination induced significant increase of IFNγ release. Statistical analysis was performed using an ordinary one-way ANOVA. Error bars represent the SD. G, Tumor mutation burden of PDOTS tumors grouped by IFNγ responder status. Student t test was performed between the two groups. In all studies, data were collected 3 days after treatment, and BMS-986408 was dosed at 0.3 μmol/L. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.001. Ctrls, controls; mIF, multiplexed immunofluorescence; Pt, patient. (A, Created with BioRender.com.)

Figure 6.

Dual DGKα/ζ inhibitor BMS-986408 unleashes CAR T-cell therapy. A, Growth curves of Raji transduced with red-shifted firefly luciferase (Raji-rFluc) tumor in NOD/SCID gamma mice over time. Mice were given a suboptimal dose of 1 × 10⁶ CAR-T cells of different genotypes and dosed with or without 0.3 mpk BMS-986408. B and C, Modified tumor control index (see “Materials and Methods”) and CAR-T cells per μL blood from each group. Nonparametric Kruskal–Wallis test was performed followed by the Benjamini, Krieger, and Yekutieli FDR correction for multiple comparisons. D and E, Chronically stimulated CAR T cells (CAR-T) were removed from plate-bound stimulus and plated with A549.CD19 or Granta-519 3D spheroids with varying treatment levels of BMS-986408. Normalized tumor area (RCU μm²) was assessed on day 9. Friedman test was performed with Dunn post hoc test for multiple comparisons. *, P < 0.05; **, P < 0.01. F, Growth curves of Nalm6 transduced with red-shifted firefly luciferase (Nalm6-rFluc) tumor in NOD/SCID gamma mice over time. Mice were dosed with BMS-986408 (0.3 mpk), 1 × 10⁶ CAR-T, or the combination of both. G, Nalm6-rFluc tumor growth curves were analyzed calculated as modified tumor control index. Student t test was performed between the two groups. H, Blood circulating CAR-T were quantified by flow cytometry on days 8, 16, 23, and 30. For all plots, error bars represent the SD. *, P < 0.05; **, P < 0.01; ****, P < 0.0001.