Proteolysis-Targeting Chimeras Enhance T Cell Bispecific Antibody-Driven T Cell Activation and Effector Function through Increased MHC Class I Antigen Presentation in Cancer Cells

Low levels of Ag presentation on the surface of cancer cells are a major challenge for the efficacy of Ag-targeting cancer immunotherapeutics, such as TCR mimicking Abs and T cell bispecific Abs (TCBs). Strategies to enhance Ag presentation on cancer cells are desirable (1–3). The processes that influence the availability of Ags on the surface of cells include intracellular protein expression, protein turnover, peptide processing, and MHC class I (MHC I) cell surface abundance (2, 4). Proteolytic degradation is the main process contributing to the generation of MHC I–displayed peptides and is followed by either TAP-dependent or -independent processing pathways (5, 6). After proteasomal processing, peptides escaping amino acid catabolism are mostly transported into the endoplasmic reticulum (ER) by TAP, trimmed by endoplasmic reticulum aminopeptidases 1 (ERAP1) and 2 (ERAP2), and loaded onto MHC I proteins for transport to the cell surface, where they are recognized by cytotoxic CD8⁺ T cells via their TCRs (7).

Understanding of the molecular pathways associated with the ubiquitin-proteasome system, as well as the discovery of small molecules able to interact with and recruit ubiquitously expressed E3 ligases, paved the way for targeted degradation of intracellular proteins as a novel alternative to conventional inhibition (8–12). Proteolysis-targeting chimeras (PROTACs), or so-called degraders, are heterobifunctional molecules that consist of an E3 ligase–binding ligand connected via a linker to a specific ligand that binds the target protein. The mode of action of PROTACs entails hijacking an endogenous E3 ligase and bringing it into close proximity to a target protein, thus inducing ubiquitination and subsequent proteasomal degradation of the target. E3 ligases that are commonly considered for PROTAC design include Cereblon (CRBN), which is the substrate receptor of the cullin 4–RING E3 ligase complex and the von Hippel–Lindau (VHL) E3 ligase. An advantage of PROTACs over classical small-molecule inhibitors is substoichiometric activity due to their catalytic mode of action resulting in multiple rounds of target protein degradation (13). Degraders that target proteins with long half-lives induce sustained inhibition of downstream signaling with prolonged pharmacodynamic effect. Furthermore, degraders hold the promise to expand the druggable proteome, because, unlike inhibitors, they require only binding at a site on the target, but not necessarily the active site (14–16). PROTACs have been exploited to accomplish depletion of proteins driving oncogenesis and are currently under clinical evaluation for the treatment of prostate and breast cancers (17) (ClinicalTrials.gov identifiers: NCT03888612 and NCT04072952).

Apart from their effect of depleting a protein target, PROTACs may also be envisaged as tools to increase Ag presentation on MHC I by contributing to the target protein turnover as a source for the antigenic peptides. As a consequence of enforced protein degradation and increased levels of peptide generation, PROTAC-treated tumor cells potentially display neoantigens on their cell surfaces that are otherwise retained intracellularly or that are not displayed in sufficient quantities on the cell surface to be targeted with TCBs or other targeting agents. The activity of a degrader on Ag presentation was first reported by Moser et al. (18) by using an overexpression system for FKBP12^F36V fused to an OVA peptide. The degrader molecule dTAG-7 induced proteasome-dependent degradation of the Ag-FKBP12^F36V fusion protein and increased presentation of an OVA peptide on MHC I in murine cell lines, determined by flow cytometry. Targeted degradation was also analyzed in the context of Ag presentation by Jensen et al. (19), who reported the activity of ARV-825 (JQ1-CRBN), ARV-771 (JQ1-VHL), and JQ1-MDM2 on presentation of bromo- and extraterminal domain (BET) peptides, assessed by mass spectrometry analysis of immunoprecipitated peptide–MHC I complexes. Although this study established a relationship between degradation of an endogenous protein and the abundance of associated peptides presented by MHC I, the immune response to cancer cells subjected to targeted protein degradation has not yet been explored. Specifically, it remains unclear whether the degrader-induced change in peptide abundance on MHC I is sufficient to prime T cells against cancer cells. Furthermore, no combination of a degrader molecule with a matching T cell engager molecule holding specificity for a degrader target-derived antigenic peptide has previously been investigated.

In this article, we show that small-molecule degraders of BET family proteins [dBET1 (20), dBET6 (21), and MZ1 (12)] induce a 3-fold increase in the presentation of five newly identified antigenic peptides on MHC I by using MHC I–associated peptide proteomics (MAPPs) with isotopically labeled standard peptides. Moreover, we extend the evidence for BET degrader activity on Ag presentation to solid tumor–derived cells, in addition to liquid tumor–derived cells. We ascertain the dependency of such activity on E3 ligase target proximity by reporting the Ag-modulatory inactivity of negative control degraders and of the ligands JQ1 and thalidomide. Next, we applied the concept of harnessing targeted protein degradation to Wilms tumor protein 1 (WT1) as a source for tumor-associated Ags (TAAs). For this purpose, we expressed a WT1-FKBP prolyl isomerase 1A (FKBP12)^F36V fusion protein in breast cancer cells and used an FKBP12^F36V degrader molecule recruiting the E3 ligase CRBN (20, 22). Importantly, we address for the first time, to our knowledge, whether the degrader-dependent activity on Ag presentation is sufficient to enhance the immune response against cancer cells. In a coculture model of cancer cells (pretreated with the FKBP12^F36V degrader) and T cells (supplemented with a TCR-like TCB specific for a WT1 peptide), we investigated the biological rationale for combining a degrader and an Ag-directed immunotherapeutic for oncology indications.

The PROTAC molecules dBET1 (catalog no. HY-101838), dBET6 (catalog no. HY112588), and MZ1 (catalog no. HY-107425) were purchased from MedChemExpress. Control compounds JQ1, thalidomide, dBET1-(R) (20), and dBET1-NMe (23) were synthesized at Roche. A PROTAC molecule containing an FKBP12^F36V ligand and a VHL ligand (hereinafter FKBP12^F36V degrader) was synthesized by Roche, following a previously reported protocol (24). For the FKBP12^F36V ligand, we adopted a system developed by Ariad Gene Therapeutics, wherein a mutant FKBP12 presents a specificity pocket (F36V) available to be bound by an orthogonal “bump”-ligand-chloroalkane. This permitted both high target affinity and specificity for the mutant protein FKBP12^F36V over endogenous wild-type FKBP12 (22, 24, 25). Distribution of degrader compounds and control vehicle fluid in 96-well plates was carried out with a D300e Digital Dispenser (Tecan).

A TCB of the 2 + 1 format and selective for WT1 (hereinafter RMF-TCB) served as a tool for detection of the WT1–MHC I complex and for generating a “synthetic” immunity in a coculture model of cancer and T cells (26). RMF-TCB targets WT1 by bivalent recognition of the peptide RMFPNAPYL (hereinafter RMF) in the context of HLA-A02 and holds monovalent complementary binding to CD3ε on T cells. Similarly, a bispecific Ab (hereinafter VLD-TCB) selective for the WT1 peptide VLDFAPPGA (hereinafter VLD) was used for this work. TCBs were kindly provided by Vesna Pulko and Lydia Jasmin Hanisch (Roche Innovation Center Zurich, Schlieren, Switzerland).

Buffy coat from healthy volunteers was diluted 1:1 with Dulbecco’s PBS (DPBS; catalog no. 14190-094; Life Technologies). A total of 15 ml of Ficoll-Paque Premium solution (catalog no. 17-5442-02; Sigma) was added to a Leucosep separation tube (catalog no. 227290; Greiner-Bio) and centrifuged at 1000 × g for 1 min (brake 7). Diluted buffy coat was overlaid on the Leucosep barrier above the Ficoll and centrifuged at 1000 × g for 10 min (brake 0). The layer of PBMCs was harvested and washed three times with 0.22 µm filtered DPBS + 2 mM EDTA, by centrifuging at 250 × g for 10 min. Cell pellets were resuspended in RPMI 1640 medium (catalog no. A10491; Life Technologies) + 10% Premium Grade FBS (catalog no. 97068-085; VWR). Finally, cells were counted by Countess II FL (Thermo Fisher Scientific), centrifuged at 350 × g for 10 min, and frozen in 60% RPMI 1640, 30% FBS, and 10% DMSO. Pan T cells or CD8⁺ T cells were isolated from PBMCs, using a Pan T Cell Isolation Kit, human (catalog no. 130-096-535; Miltenyi Biotec) and a CD8⁺ T Cell Isolation Kit, human (catalog no. 130-096-495; Miltenyi Biotec), respectively; LS Columns (catalog no. 130-042-401; Miltenyi Biotec); and a QuadroMACS separator (catalog no. 130-042-303; Miltenyi Biotec), following the manufacturer’s instructions.

SKM1, CTV-1, A375, and MDA-MB-231 cells were cultured in RPMI 1640 (catalog no. A10491; Life Technologies) + 20% FBS, RPMI 1640 GlutaMAX (catalog no. 6187-010; Life Technologies) + 15% FBS, DMEM High Glucose (catalog no. 41966; Life Technologies) + 10% FBS, and DMEM GlutaMAX (catalog no. 31966; Life Technologies) + 10% FBS, respectively. Coculture assays of cancer and T cells, in the presence of WT1-TCB, were carried out in RPMI 1640 medium (catalog no. A10491; Life Technologies) + 2% FBS.

MDA-MB-231 cells were transduced to express the nuclear-restricted red fluorescent protein mKate2, using the Incucyte NucLight Red Lentivirus Reagent (EF-1α, Puro; catalog no. 4625; Sartorius), selected under 1 µg/ml puromycin and maintained under 0.5 µg/ml puromycin, for stable expression of nuclear red fluorescent protein. This red nuclear protein also served for flow cytometry–based gating of cancer cells.

A WT1 gene was synthesized with human codon optimization by GenScript to encode the human Wilms tumor protein isoform D (NCBI reference: NP_077744.4). WT1 was fused C-terminally or N-terminally via a linker (1×Gly-Ser) to a F36V mutant of human FKBP12 (GenBank: CAG46965.1) and cloned into pLenti6.3/TO/V5-DEST vector, using SpeI and XbaI restriction sites. The starting codon of WT1 sequence was removed from the FKBP12^F36V-N-terminal tagged construct. pLenti6.3/TO/V5-DEST and pLenti3.3/TR were supplied in the ViraPower HiPerform T-REx Gateway Vector Kit (catalog no. A11144; Invitrogen). Lentivirus was produced in HEK293T cells after transfection of pLenti plasmids, using Fugene 6 transfection reagent (catalog no. E2691; Promega) and MISSION Lentiviral Packaging mix (catalog no. SHP001; Sigma). MDA-MB-231 cells with stable expression of red fluorescent protein were lentivirally retransduced both with pLenti3.3/TR and either pLenti6.3/TO/V5-DEST (hereinafter EV), pLenti6.3/TO/WT1/V5-DEST (hereinafter WT1), pLenti6.3/TO/WT1/Gly-Ser/FKBP12^F36V/V5-DEST (hereinafter WT1-FKBP12^F36V), or pLenti6.3/TO/FKBP12^F36V/Gly-Ser/WT1/V5-DEST (hereinafter FKBP12^F36V-WT1). Transduced cells were cultured with 0.5 µg/ml puromycin in the media for stable expression of nuclear red fluorescent protein, 900 µg/ml geneticin for stable expression of TET repressor protein, and 10 µg/ml blasticidin for doxycycline-inducible expression of WT1 or WT1 fusion proteins with FKBP12^F36V.

BRD4 gene was synthesized with human codon optimization by ATUM to encode the human bromodomain-containing protein 4, long isoform (NCBI reference: NP_490597.1). BRD4 was fused C-terminally or N-terminally via a linker (1×Ala) to the antigenic WT1 peptide RMFPNAPYL and cloned into pcDNA4/TO/myc-His A (catalog no. K103001; Invitrogen) using BamHI and XbaI restriction sites. MDA-MB-231 cells with stable expression of red fluorescent protein were transfected with pcDNA6/TR (catalog no. V1025-20; Invitrogen) and either pcDNA4/TO/BRD4/RMF/myc-His, pcDNA4/TO/RMF/BRD4/myc-His, or pcDNA4/TO/myc-His for transient doxycycline-inducible expression of BRD4-RMF, RMF-BRD4, and EV, respectively. Lipofectamine 3000 (catalog no. L3000001; Thermo Fisher Scientific) was used as transfection reagent.

A total of 1 × 10⁷ SKM1, CTV-1, or A375 cells was treated with degrader or control compounds for the time indicated in each experiment. After treatment, cell pellets were frozen at −80°C. Cells were lysed in lysis buffer (20 mM Tris-HCl [pH 7.8], 5 mM MgCl₂, 1% Triton X-100, 1× protease inhibitors) for 1 h with agitation at 1100 rpm in a Thermomixer (Eppendorf) at 4°C and centrifuged at 16,000 × g for 10 min. Clarified supernatants were incubated overnight on a rotator at 4°C with NHS Mag Sepharose beads (catalog no. 28-9440-09; GE Healthcare), which had been previously equilibrated in 1 mM HCl, coupled to a pan-anti-HLA class I Ab (catalog no. ab23755; Abcam) for 2 h at 1100 rpm in a Thermomixer, blocked in 100 mM Tris-HCl (pH 8.0), 150 mM NaCl for 2 h at 1100 rpm, and washed four times in PBS. After immunoprecipitation, beads were washed four times in wash buffer (20 mM HEPES [pH 7.9], 150 mM KCl, 1 mM MgCl₂, 0.2 mM CaCl₂, 0.2 mM EDTA, 10% glycerol, 0.1% Tergitol), four times in wash buffer without Tergitol, and four times in milli-Q H₂O. Peptide–MHC I complexes were eluted in 0.1% TFA for 30 min at 1100 rpm at 37°C in a Thermomixer, evaporated using a SpeedVac, and resuspended in 0.5% (v/v) formic acid, 2% (v/v) acetonitrile/water, prior to analysis using tandem mass spectrometry. After identification of two BRD4 peptides (AEALEKLFL, QEFGADVRL) in the MHC I complexes, isotopically labeled peptides (AEALE-[U-[¹³C]₆,¹⁵N₂-Lys]-LFL-acid, QEFGADV-[U-[¹³C]₆,¹⁵N₄-Arg]-l-acid; Cambridge Research Biochemicals) were spiked into the resuspended peptide solution for quantification analyses.

Data-dependent acquisition (DDA) of peptides was carried out using a Q Exactive HF Orbitrap Mass Spectrometer (Thermo Fisher Scientific) equipped with an Ultimate 3000 RSLCnano UPLC (Thermo Fisher Scientific, Waltham, MA) and a nanoelectrospray interface. Mobile phases used for separation of peptides were 0.1% formic acid in 2% v/v acetonitrile and 0.1% formic acid in acetonitrile. Samples (10 μl volume) were loaded for 2 min at 10 μl/min onto an AcclaimPepMap trap column (100-μm internal diameter × 20 mm; catalog no. 164564; Thermo Scientific) using a vented-Tee design before being chromatographed using a self-packed C₁₈ analytical column (75-μm internal diameter × 170 mm, ReproSil-Pur C18-AQ, 3 μm; Dr. Maisch) at a flow rate of 250 nL/min. Peptides were eluted using a nonlinear 39-min gradient of 2–45% buffer B, followed by an 11-min column wash and re-equilibration for 10 min (buffer A: 0.1% [v/v] formic acid in 2% [v/v] acetonitrile/water; buffer B: 0.1% [v/v] formic acid in acetonitrile). The mass spectrometer was operated in data-dependent mode with a survey scan range of 400–1650 m/z at a resolution of 60,000 at 200 m/z with an automated gain control target of 1 × 10⁶ ions and a maximum injection time of 35 ms. Up to a maximum of 10 precursor ions (charge states 2–5) were selected per duty cycle at a 1.8 Th isolation window for higher-energy collisional dissociation at a resolution of 15,000. Fragmentation was performed at stepped normalized collision energies of 23, 25, and 27 eV using an automated gain control target of 1 × 10⁵ ions and a 60 ms maximum injection time. Dynamic exclusion was set to 12 s to avoid repeated fragmentation of abundant ions. Parallel reaction monitoring (PRM) analysis of selected ions was achieved by inserting a PRM analysis step in the DDA workflow to be performed at each instrument duty cycle (mixed DDA/PRM mode). PRM was performed typically during a 10-min window bracketing the peptide’s expected retention time. The ion of interest was isolated at a 1.8 Th isolation window, and the fragmentation pattern was acquired at 30,000 resolution using a 200 ms maximum injection time.

Peptide identifications were performed using the PEAKS Studio software version 8.5 (Bioinformatics Solutions, http://www.bioinfor.com) against the UniProt database (http://www.uniprot.org, release 2018_02) filtered for the taxonomy “human” (71803 TrEMBL and SwissProt entries). Data were searched with a mass tolerance of ± 5 ppm for parent ions and 0.025 Da for fragment ions considering no cleavage specificity and methionine oxidation, asparagine and glutamine deamidation, and pyroglutamic acid as differential modifications. Peptide identifications were filtered at a false discovery rate ≤1%. Label-free quantification was performed using the PEAKS Label-Free module.

MHC I peptide quantification using PRM data was carried out using Skyline version 4.1 (https://skyline.ms/project/home/software/Skyline/begin.view?) using the heavy-labeled peptide standard to normalize for response.

At day 1, 1 × 10⁴ MDA-MB-231 cells with nuclear-restricted red fluorescence were seeded in 96-well plates. At day 2, doxycycline was added in medium supplemented with 10% TET-reduced FBS (catalog no. P30-3602; Pan Biotech). In the BRD4-degradation model, the expression of BRD4-RMF or RMF-BRD4 was induced in MDA-MB-231 cells with doxycycline for 24 h; MDA-MB-231 cells were pretreated with dBET6 for 24 h to induce BRD4 degradation and cocultured with primary T cells freshly purified from human PBMCs (E:T ratio, 1:1) in the presence or absence of WT1-TCB in 2% FBS-RPMI 1640 medium. In the WT1-FKBP12^F36V degradation model, the expression of WT1-FKBP12^F36V was induced with doxycycline for 3.5 h prior to addition of degrader compound. Twenty-four hours after degrader treatment, pan T cells or CD8⁺ T cells and WT1-TCB were added to the coculture. Cells were incubated for the indicated times in the Incucyte S3 imaging system (Essen Bioscience) with time-lapse bright-field and red fluorescence microscopy. Four representative images were captured per well every 4 h. Data were reported as red object count per well, which is indicative of the number of cancer cell nuclei. Alternatively, viability of cancer cells in the coculture was determined by flow cytometry.

Flow cytometry was performed on T cells that had been cocultured with degrader-primed cancer cells and exposed to WT1-TCB or vehicle, following the same protocol as described in the cancer cell viability assays. Where indicated in the text, cancer cells were pretreated with recombinant human IFN-γ (catalog no. 285-IF; R&D). For analyses of T cell proliferation, purified human CD8⁺ T cells were stained using a CellTrace Violet Cell Proliferation Kit (catalog no. C34557; Thermo Fisher Scientific) prior to addition of T cells to the coculture. Supernatant (containing T cells, but not adherent cancer cells) was harvested at the indicated times after addition of T cells and TCB to cancer cells and centrifuged at 350 × g for 5 min at 4°C. Cell pellets were washed with DPBS and stained for 30 min at 4°C with 50 µl of staining solution in DPBS, including Zombie NIR (catalog no. 423105) and Abs for detection of early/late T cell activation (Ab panel 1 on samples harvested from a 24-h coculture: TCR-α/β-Alexa Fluor 647 [catalog no. 306714], CD8-PE [catalog no. 344706], CD4-BV421 [catalog no. 300532], CD69-Alexa Fluor 488 [catalog no. 310916]; Ab panel 2 on samples harvested from a 4-d coculture: TCR-α/β-Alexa Fluor 647, CD8-PE, CD25-PECy7 [catalog no. 302612]). For analyses of cancer cell killing and T cell PD1 expression by flow cytometry, CD8⁺ T cells were transferred in a 96-well U-bottom plate (catalog no. 353077; Falcon), and the remaining cancer cells were treated with Trypsin 0.25% (catalog no. 25200056; Life Technologies) for 5 min at 37°C. After detachment, cancer cells were pooled with CD8⁺ T cells and centrifuged at 350 × g for 5 min. Supernatant was discarded, and cell pellets were washed with DPBS and stained for 30 min at 4°C with 50 µl of staining solution in DPBS, including Zombie NIR and the BioLegend Abs for Alexa Fluor 488-TCR (catalog no. 306712), allophycocyanin-CD8 (catalog no. 344722), and BV-421-PD1 (catalog no. 329920). Stained cells were blocked, washed, and resuspended in MACS buffer (0.5% BSA, 2 mM EDTA, DPBS), prior to analysis, using CytoFLEX Flow Cytometer (Beckman Coulter). For determination of cell surface expression of HLA-A2-treated cells in dBET1-, MZ1-, or DMSO-treated SKM1 cells, cell pellets were washed with DPBS, stained with Zombie NIR and anti-HLA-A2-Pacific Blue Ab (catalog no. 343312; BioLegend), blocked in MACS buffer, equilibrated in Annexin-binding buffer, and restained in Annexin V (catalog no. V13421; Thermo Fisher Scientific), following the manufacturer’s instructions. For determination of HLA-A, B, C surface expression in MDA-MB-231 cells, we used an anti-HLA-A, B, C-allophycocyanin Ab (catalog no. 311410; BioLegend).

A Cytometric Bead Array assay was performed to quantitatively measure IFN-γ, TNF-α, and granzyme β (GZMB) in the supernatant harvested 24 h after addition of T cells and TCB to the cancer cells. The following kits were used: BD Cytometric Bead Array Human IFNg Flex Set (catalog no. 558269), BD Cytometric Bead Array Human Granzyme B Flex Set (catalog no. 560304), and BD Cytometric Bead Array Human TNFa Flex Set (catalog no. 560112). Standards and capture beads mix were prepared following the manufacturer’s instructions. The supernatant was mixed on a plate shaker for 5 min at 1200 rpm, and 10 μl was transferred in a 96-well V-bottom plate (catalog no. 353263; Falcon). In parallel, a 96-well V-bottom plate (catalog no. 353263; Falcon) was prepared with 10 μl standards per well. The capture beads mix was vortexed, and 10 μl per well was transferred in both sample and standards plates. Plates were incubated for 1 h at room temperature on a plate shaker at 300 rpm protected from light. After 1 h, 10 μl of PE Detection Reagent was added to each well, and plates were incubated for 2 h. A total of 150 μl of wash buffer (catalog no. 560105; BD Biosciences) was added to each well of sample and standards 96-well V-bottom plates. Standards and samples were then transferred to prewashed MultiScreen HTS-BV 1.2-µm clear nonsterile filter plates (catalog no. MSBVN1210; Millipore Sigma), and wells were drained by vacuum filtration for 5 s. A total of 50 μl of wash buffer (catalog no. 560105; BD Biosciences) was added to each well, and plates were measured at IntelliCyt IQUE Plus flow cytometer (Sartorius).

After separation via SDS-PAGE, protein samples were transferred to 0.2-μm nitrocellulose membranes using a Trans-Blot Turbo Transfer System (Bio-Rad). After blocking with Intercept (PBS) Blocking Buffer (catalog no. 927-70001; LI-COR), the membrane was incubated in primary Ab (anti-BRD4, catalog no. 13440 [Cell Signaling]; anti-ACTIN, catalog no. 5441 [Sigma]; anti-MYC, catalog no. 950-25 [Invitrogen]; anti-WT1, ab89901 [Abcam]) at 4°C overnight, following the manufacturer’s instructions. After a 3 × 10 min wash with TBS supplemented with 0.05% Tween 20, membranes were incubated for 1 h at room temperature in secondary Ab (donkey anti-rabbit, catalog no. 926-32213; donkey anti-mouse, catalog no. 926-68072; LI-COR), following the manufacturer’s instructions. After another washing, protein signals were detected using a LI-COR Odyssey imaging system.

Statistical significance was determined by unpaired t test with Welch correction using GraphPad software, version 7. Two-sided p values (p < 0.05) were considered significant.

Physiological proteasomal degradation of an intracellular protein contributes to the generation of peptides that bind to the MHC I receptor for Ag presentation. To substantiate the hypothesis that drug-enhanced degradation of an endogenous intracellular target also impacts Ag presentation, we set out to identify and quantify the abundance of BRD4 peptides that bind to MHC I on BET degrader treatment in leukemic cells, using MAPPs (Fig. 1A). We first determined the concentrations of the BET protein degraders dBET1 (recruiting the E3 ligase CRBN) and MZ1 (recruiting the E3 ligase VHL) that were required to efficiently degrade BRD4 in SKM1 cells (Fig. 1B). Treatment with 1 nM dBET1 for 8 h was sufficient to yield a 50% reduction in BRD4 protein levels in SKM1 cells. Then we isolated MHC I–binding peptides from 10 × 10⁶ SKM1 cells that were treated with dBET1, MZ1, or DMSO for 8 h. We identified five novel MHC I–binding peptides (QEFGADVRL, AEALEKLFL, AVPPPTKVV, YEEKRQLSL, ATPHPFPAV) by liquid chromatography-tandem mass spectrometry that matched uniquely to BRD4 (long and short isoforms) and not to other BET family members (BRD1, BRD2, and BRD3). Peptide elutions from cellular MHC I complexes were mixed with the isotopically labeled peptides AEALE-[U-[¹³C]₆,¹⁵N₂-Lys]-LFL-acid and QEFGADV-[U-[¹³C]₆,¹⁵N₄-Arg]-l-acid to allow for precise peptide quantification of the two most abundant identified peptides. Relative peptide abundance was determined from the ratio of light (experimental sample) over heavy (isotopically labeled standard) signal. dBET1 and MZ1 treatment resulted in an up to 3.2-fold and 2.6-fold increase in the peptide abundance of QEFGADVRL and in an up to 3.9-fold and 2.4-fold increase in the abundance of AEALEKLFL, respectively (Fig. 1C). A concentration of 1 nM dBET1 or MZ1 was sufficient to show a significant increase in BRD4 peptide abundance on MHC I.

To gain insight about how fast targeted protein degradation enhances peptide abundance on MHC I, we assessed the degrader-induced depletion of BRD4 protein over time (Fig. 1D) and quantified BRD4 peptides eluted from MHC I in SKM1 cells that were treated with dBET1 for 2 or 8 h. A treatment with 100 nM dBET1 decreased the BRD4 protein by 50% and increased BRD4 peptide abundance on MHC I already after 2 h (Fig. 1E).

BRD4 protein is implicated in a number of hematological and solid tumors. BRD4 modulates transcription elongation of essential genes that are involved in cell cycle and apoptosis, such as c-MYC and BCL2 (27). BRD4 also promotes antitumor immune responses through transcriptional repression of the immune checkpoint ligand PD-L1 (28). In view of the immunomodulatory activity of BRD4, we sought to rule out any potential direct contribution of depletion of the biological activity of BRD4 on Ag presentation in our model, because we were interested in investigating BRD4 degradation as a source for Ags. Furthermore, we pursued excluding any off-target effects on Ag presentation potentially arising from hijacking E3 ligases through degraders. For these purposes, we treated SKM1 cells with dBET1-(R) or dBET1-NMe (Fig. 2A), two inactive control degrader molecules. These compounds retain most similar physicochemical properties to the functional dBET1, because they contain only minor modifications in the target- or E3 ligase–binding arm. These modifications lead to the inability to bind to the protein target and to the E3 ligase CRBN, respectively. We also included treatments with JQ1 and thalidomide alone (Fig. 2A), the respective ligands for BRD4 and the E3 ligase CRBN, which are structural components of dBET1. Although dBET1 treatment resulted in a dose-dependent depletion in BRD4 protein and an increase in abundance of both BRD4 peptides, QEFGADVRL and AEALEKLFL, all inactive controls and single-arm binders did not reduce BRD4 protein levels [except for very high concentrations of thalidomide, which might regulate protein translation independent of degradation (29, 30)] (Fig. 2B) and did not affect the abundance of BRD4 peptides on MHC I (Fig. 2C). These data suggest that proximity of the E3 ligase and the target that leads to subsequent functional BRD4 degradation (induced by dBET1), but not BRD4 inhibition (JQ1, dBET1-NMe) or CRBN binding alone [thalidomide, dBET1-(R)], are required for the described effect of degrader activity on Ag presentation. We also inferred quantitative information from mass spectrometry peak area for three other newly identified BRD4 peptides, AVPPPTKVV, YEEKRQLSL, and ATPHPFPAV (Supplemental Fig. 1A). We concluded that dBET1, but not the inactive controls, increases the abundance of all detectable BRD4 peptides on MHC I.

We next addressed whether the observed increase in Ag presentation is selective for BRD4 or is involving the entire immunopeptidome pool by regulating MHC I expression itself. For that purpose, we determined MHC I protein levels on treatment with dBET1 or MZ1 for 8 h in SKM1 cells by flow cytometry. Median fluorescence intensity for HLA-A2⁺ cells (pregated on living nonapoptotic cells) did not increase on treatment with BET degraders, but rather modestly decreased (Supplemental Fig. 1B). We cannot exclude that the observed modest decrease in the immunopeptidome might result in rearrangements of MHC I peptide abundance and altered immune responses by altering immunodominance, similarly to the previously reported case of modulation of ERAP1 (31, 32). Overall, these data exclude a direct role of BET proteins in positive regulation of MHC I cell surface abundance and therefore rule out that a nonspecific increase in the immunopeptidome pool underlies the PROTAC-dependent increase in BRD4 Ag presentation.

We examined BRD4 degradation in other cell lines than SKM1. Treatment with dBET1 was applied to the acute myeloid leukemia cell line CTV1 and melanoma cell line A375. dBET1 reduced BRD4 protein levels (Fig. 3A) and increased the relative abundance of both peptides, QEFGADVRL and AEALEKLFL, in a dose-dependent manner in both cell lines (Fig. 3B).

The observed activity of BET degraders on Ag presentation prompted us to test whether BRD4 degradation can be used as a model to investigate the functional impact of targeted protein degradation on the anticancer immune response that is synthetically induced by a TCR-like bispecific Ab. Because we did not have access to TCR-like Abs recognizing BRD4-derived peptides in the context of HLA-A02, we chose to generate a BRD4 fusion protein with a peptide derived from WT1, for which a TCR-like Ab recognizing the WT1-derived RMF peptide in the context of HLA-A02 has been described (26). Similar to a previously reported WT1 bispecific T cell engager (33), this bispecific Ab (RMF-TCB) engages T cells in close proximity to the WT1 peptide RMF displayed in the MHC I context on the surface of cancer cells and thereby triggers an immune response against cancer cells. We chose MDA-MB-231 as target cells because these breast cancer cells express endogenous levels of WT1 and can be killed by T cells in the presence of RMF-TCB, and we set out to detect additional TCB-driven killing in the context of targeted protein degradation. We first transduced MDA-MB-231 HLA-A*0201⁺ cells to express a red fluorescent marker in the nucleus, which can be tracked using the Incucyte technology. We then transfected MDA-MB-231 to express a myc-tagged BRD4 fused C- or N-terminally to the RMF peptide in a doxycycline-inducible manner. Although endogenous BRD4 could be efficiently degraded already at low concentrations of dBET6, a further optimized version of dBET1 (Supplemental Fig. 1C; see EV-transfected cell lysates probed with anti-BRD4 Ab), degradation of the overexpressed BRD4-RMF-myc and RMF-BRD4-myc turned out to require 10 µM dBET6 (Supplemental Fig. 1C; see respective transfected cell lysates probed with anti-BRD4 Ab and anti-myc). Finally, we set out a coculture of cancer cells and purified human pan T cells to determine the activity of dBET6 on BRD4 degradation and on T cell–mediated killing of cancer cells (Supplemental Fig. 1D). As expected, cancer cell viability of all MDA-MB-231 cells (untransfected or EV, BRD4-RMF-myc, or RMF-BRD4-myc transfected) decreased over time with 1 µg/ml RMF-TCB treatment alone (Supplemental Fig. 1E), suggesting the occurrence of T cell–mediated killing of cancer cells as a result of endogenous and exogenous WT1 expression. A 10 µM dBET6 pretreatment of cancer cells resulted in decreased viability in all tested MDA-MB-231 cell lines already at the beginning of the coculture with T cells and did not enhance further but rather attenuated over time TCB-dependent killing of BRD4-RMF/RMF-BRD4-transfected cancer cells. The degradation of endogenous BRD4, which is an essential protein for cell survival, may account for this attenuation. In fact, depletion of endogenous BRD4 results in cancer cell toxicity per se, thereby reducing the potential cell source for Ags and limiting TCB-mediated activity. Subsequently, we concluded that the use of a degrader targeting both an overexpressed target-antigen fusion protein (BRD4-RMF) and the respective endogenous essential target protein (BRD4) is not suitable for investigating the contribution of targeted protein degradation on Ag-specific immune response and subsequent cancer cell killing.

Because the model of BRD4 degradation was unsuitable for investigation of the functional impact of targeted protein degradation on immune response against cancer cells, we modified our experimental design to use a degrader molecule able to selectively target an exogenously expressed protein without affecting any endogenous essential protein. For this purpose, we chose the FKBP12^F36V degrader, a molecule with high affinity and specificity for the mutant FKBP12^F36V over endogenous wild-type FKBP12 (22, 24, 25). WT1 protein was fused to the degradable tag FKBP12^F36V to generate a fusion protein that can be targeted by the FKBP12^F36V degrader. The full-length sequence of WT1 protein was included here (instead of the RMF peptide only) to reduce any potential risk of interference in the proteasomal processing and MHC I loading of the RMF peptide (Fig. 4A). We stably overexpressed a doxycycline-inducible WT1-FKBP12^F36V fusion protein in MDA-MB-231 cells, and we assessed the degradation of the fusion protein by Western blot–based probing against WT1 (Fig. 4B). We set out to determine whether pretreatment of cancer cells with FKBP12^F36V degrader enhances the activation of T cells in a coculture system (Fig. 4C). As expected, RMF-TCB treatment dose-dependently increased the percentage of CD8⁺ and CD4⁺ T cells that express the early activation marker CD69 in the cocultures of purified human pan T cells and MDA-MB-231 cells (Fig. 4D; see Supplemental Fig. 2A for gating information). Pretreatment of MDA-MB-231 cells expressing WT1-FKBP12^F36V with 0.1 µM FKBP12^F36V degrader resulted in an additional increase in the percentage of CD8⁺CD69⁺ and CD4⁺CD69⁺ T cells when suboptimal RMF-TCB concentrations were tested (Fig. 4E). This observation suggests that targeted degradation of the WT1-FKBP12^F36V fusion protein in cancer cells enhances the early activation of cocultured T cells. Conversely, the percentage of CD69⁺ T cells did not increase on pretreatment of cancer control cells (MDA-MB-231 expressing WT1 alone) with 0.1 µM FKBP12^F36V degrader, suggesting that targeted degradation of exogenous WT1-FKBP12^F36V fusion protein is required for the effect of FKBP12^F36V degrader on early T cell activation (Supplemental Fig. 2B–D). We next addressed whether the expression of MHC I on cancer cells might be a limiting factor for the activity of FKBP12^F36V degrader on early T cell activation. For this purpose, we pretreated WT1-FKBP12^F36V-expressing MDA-MB-231 cells with IFN-γ, a cytokine known to upregulate expression of MHC I. As expected, IFN-γ dose-dependently increased HLA-A, B, C cell surface expression and the percentage of CD8⁺CD69⁺ cocultured T cells (Fig. 4F). However, IFN-γ treatment did not result in a synergistic effect with FKBP12^F36V degrader; instead, the fold increase in the percentage of CD8⁺CD69⁺ T cells on treatment with the FKBP12^F36V degrader remained constant at increasing doses of IFN-γ (Fig. 4F). These data suggest that a higher expression of MHC I does not specifically enhance the effect of targeted protein degradation on early T cell activation but rather contributes to enhancing, in general, early T cell activation irrespective of targeted protein degradation in cocultured cancer cells.

After investigating early T cell activation, we analyzed late T cell activation using CD25 as a marker and T cell proliferation by CellTrace Violet (see Supplemental Fig. 3A for gating information). MDA-MB-231 cells expressing WT1-FKBP12^F36V were pretreated with the FKBP12^F36V degrader for 24 h and later cocultured with purified CD8⁺ T cells in the presence of RMF-TCB for 4 d prior to harvesting of T cells for flow cytometry analysis (Fig. 5A). As expected, RMF-TCB treatment dose-dependently increased the percentage of both CD25⁺ and dividing CD8⁺ T cells, which were cocultured with the MDA-MB-231 cells expressing WT1-FKBP12^F36V (Fig. 5B). Pretreatment of MDA-MB-231 cells expressing WT1-FKBP12^F36V with 0.01, 0.1, and 1 µM FKBP12^F36V degrader resulted in an increased percentage of both CD25⁺ and dividing CD8⁺ T cells compared with DMSO (Fig. 5C, 5D). Notably, pretreatment with a low concentration of IFN-γ (0.01 ng/ml) enriched the percentage of dividing CD8⁺ T cells both in the presence and absence of degrader (Fig. 5D). Similar to CD69⁺ T cells, neither the percentage of CD25⁺ T cells nor that of dividing CD8⁺ T cells increased on pretreatment of control cells (MDA-MB-231 cells expressing WT1 alone) with FKBP12^F36V degrader, while both T cell activation and proliferation could be increased on RMF-TCB treatment also in this control cell line (Supplemental Fig. 3B–E). Taken together, these results support the hypothesis that targeted degradation of the exogenous WT1-FKBP12^F36V fusion protein enhances TCB-driven late activation of CD8⁺ T cells and increases their proliferative potential.

We set out to examine the effector response of T cells in the presence of two TCBs that are selective for either WT1 peptide RMF or VLD (Fig. 6A; see also Supplemental Fig. 4A, 4B for gating strategies). Coculture of FKBP12^F36V degrader-primed cancer cells and CD8⁺ T cells in the presence of RMF-TCB resulted in an enhancement of T cytotoxic cytokine secretion (GZMB, IFN-γ, and TNF-α) (Fig. 6B). Accordingly, FKBP12^F36V prepriming of MDA-MB-231 cells expressing WT1-FKBP12^F36V enhanced both RMF- and VLD-TCB-driven tumor cell killing by in vitro-generated effector CD8⁺ T cells in a 5-d coculture (Fig. 6C; see also 24-h coculture data in Supplemental Fig. 4C). It is noteworthy that FKBP12^F36V degrader treatment did not affect the viability of control cells MDA-MB-231 expressing WT1 only (Supplemental Fig. 4D), thereby suggesting that targeted degradation of the WT1-FKBP12^F36V fusion protein is required for this increase in T cell–mediated tumor cell killing. The expression of the immunosuppressive molecule PD-1 slightly increased on RMF- or VLD-TCB treatment in CD8⁺ T cells in our coculture model, although it was not affected by FKBP12^F36V degrader (Supplemental Fig. 4E). Finally, we explored the mechanism for the higher sensitivity to FKBP12^F36V degrader of VLD- compared with the RMF-TCB-treated coculture, as observed in (Fig. 6C, by MAPP-based determination of the levels for RMF and VLD Ag presentation onto MHC I. The VLD peptide binding to MHC I appeared more abundantly than the RMF peptide, already at basal level. FKBP12^F36V degrader increased the relative abundance of the VLD peptide in MDA-MB-231 cells in a dose-response manner when WT1-FKBP12^F36 protein expression was induced on treatment with 1 or 10 ng/ml doxycycline. We could also detect a slight degrader-dependent increase in the relative abundance of the RMF peptide with 10 ng/ml doxycycline. These data indicate that the VLD peptide is occupying more MHC I pockets on the cancer cell surface compared with RMF and might therefore more effectively sustain T cell–mediated killing of cancer cells. Taken together, these observations support the hypothesis that degradation of WT1-FKBP12^F36V in tumor cells promotes CD8⁺ T cell activation and effector function, indicating that targeted protein degradation is a promising strategy (e.g., in immuno-oncology) to enhance an immune response against specific Ags.

The redirection of T cell activity toward cancer cells via targeting of TAAs is one of the most promising cancer immunotherapy strategies. Examples of T cell–redirecting strategies against tumors include bispecific Abs, membrane-anchored chimeric Ag receptors, and delivery of viral epitopes via tumor-targeting Abs (34–36). However, the applicability of Ag-specific therapy to a wide range of tumor types is limited by the availability of Ags with high tumor specificity or with sufficient epitope density on cancer cell surfaces (2). In line with previous observations (18, 19), we show that drug-induced degradation of BRD4 can increase presentation of antigenic BRD4 peptides displayed on MHC I. In addition, we substantiate the essential role of proximity of the E3 ligase to the target protein in degrader activity on Ag presentation, and we extend the evidence for this activity to solid, in addition to liquid, tumor–derived cell lines. In this study, we are furthermore providing the first evidence, to our knowledge, that degrader molecules can act as enhancers of an antitumor immune response. Our data show that degradation of the fusion protein WT1-FKBP12^F36V with a FKBP12^F36V degrader in engineered MDA-MB-231 cells increases the percentage of T cells expressing the early activation marker CD69 after 24 h of coculture and the percentage of CD8⁺ T cells expressing the late activation marker CD25 after 4 d of coculture in the presence of RMF-TCB. Moreover, CD8⁺ T cells had an increased proliferative potential on pretreatment of cocultured MDA-MB-231 cells expressing WT1-FKBP12^F36V with a FKBP12^F36V degrader. FKBP12^F36V degrader treatment also increased cytokine secretion and enhanced the tumor-killing activity of effector CD8⁺ T cells, generated in vitro, in a 5-d coculture model in the presence of two WT1-directed TCBs (RMF- and VLD-TCB). It is noteworthy that in our model the combination of FKBP12^F36V degrader and TCB treatment did not result in complete killing of cancer cells. The incomplete killing may be a result of the relative timing of doxycycline induction of WT1-FKBP12^F36V fusion protein expression, protein degradation, Ag presentation, and T cell effector function. It is plausible that the increase in WT1 peptide presentation is too transient to support complete TCB-driven tumor cell killing. Nevertheless, our WT1-FKBP12^F36V-expressing cellular model system demonstrates that a contribution of drug-induced targeted protein degradation in cancer cells to T cell activation and effector function is achievable and has the potential to be exploited for future modulation of Ag-specific immune responses.

The research field of targeted protein degradation has proven to be able to develop cell-permeable small-molecule degraders (PROTACs and molecular glues) that target proteins intracellularly and reach in vivo exposure compatible with an efficacious dose. As a future therapeutic strategy, we envisage a “priming” therapy with a degrader compound, followed by treatment with a TCB, specific for the corresponding peptide–MHC I complex. Two critical aspects have to be taken into consideration for such an approach. On one hand, the design of a degrader that is selective for a mutated or overexpressed protein in cancer cells will be required to shift the balance in Ag presentation in favor of mounting an immune response only in tumor cells and not in healthy cells. On the other hand, identification of a degradation-derived Ag, with the relevant abundance and affinity features that are sufficient to induce an immune response, will be crucial in the choice of bispecific Ab to be used in combination with the degrader therapy. Our study bridges the gap between proof-of-concept evidence for degrader induction of BRD4 Ag presentation and translation of this approach to a TAA (RMF and VLD peptides of WT1), and anticipates some of the potential challenges in identifying a therapeutic window for optimal interplay of degradation, Ag presentation, and T cell effector activity. For example, our data in the BRD4-RMF model suggest that, for a strategy envisaging the combination of a degrader and a TCR-like TCB, nonessential and constitutively expressed tumor biomarker proteins are likely to be more suitable degrader targets compared with essential proteins, such as oncogenic drivers. Indeed, depletion of an essential protein results in cell toxicity and subsequent reduction of the cell source for Ags, whereas degradation of a nonessential protein is expected to sustain a fully Ag-expressing tumor cell population that might be efficiently recognized by the immune system. In addition, our data in the WT1-FKBP12^F36V model trigger the consideration that, when designing degrader molecules for combination with a TCB, one should take into account that slow target degradation is expected to be advantageous by sustaining a prolonged supply of source peptides for MHC I processing and recognition by bispecific Abs. Finally, careful evaluation of the peptides increased on degradation of a target protein (e.g., using quantitative mass spectrometry), and identification of the immunodominant peptide should be performed initially to develop a tailor-made bispecific Ab for combination with the degrader.

In summary, this study investigated the potential of combining a small-molecule degrader with a cancer immunotherapeutic agent for the first time, to our knowledge, and provided insights into the contribution of targeted protein degradation to the modulation of the immune response against cancer cells. Our data encourage research efforts toward the application of degrader molecules as modulators of immune responses in in vivo models and in the context of disease.

We acknowledge Jeremy Beauchamp for expert review of the manuscript and text editing; Vesna Pulko and Lydia Jasmin Hanisch (Duerner) for providing the RMF- and VLD-TCB; Eva-Maria Grossjohann for expert technical assistance; Philipp Ottis for suggestions about FKBP12^F36V degraders; Céline Marban-Doran and Sabine Kux van Geijtenbeek for providing protocols for the MHC I–MAPPs experiment and the mass spectrometry analysis support; Tina Zimmerman for expert Incucyte technical assistance; Remy Hallet for expert flow cytometry technical assistance; Thomas Lübbers and Antonio Ricci for chemical synthesis expertise; and Hana Janouskova, Alberto Toso, and Piergiorgio Pettazzoni for fruitful scientific discussions.

All authors, except R.E.K., have been employed at F. Hoffmann-La Roche Ltd. C.K. and Y.A.N. declare patents and stock ownership with Roche.