Brentuximab Vedotin–Driven Microtubule Disruption Results in Endoplasmic Reticulum Stress Leading to Immunogenic Cell Death and Antitumor Immunity Open Access

Tumors evade immune surveillance through tolerance of tumor-associated neoantigens or by evolving immunosuppressive tumor microenvironments (TME; refs. 1, 2). Therapeutics that kill tumor cells directly while also promoting tumor-specific T-cell responses may potentiate checkpoint inhibitor therapies, and therefore promote durable clinical responses (3).

Apoptotic cell death is a normal aspect of cellular turnover and is associated with silent or tolerogenic immune responses (4). In contrast, immunogenic cell death (ICD) is an inflammatory cell death process that promotes the activation and recruitment of innate immune cells (1). During the pre-apoptotic stage of ICD, cells present a defined combination of hallmark signals that provoke a localized immune reaction to the dying cells (5). Cardinal hallmarks of ICD include heightened translocation of calreticulin (CRT) from the endoplasmic reticulum (ER) to the cell surface, active secretion of ATP, and passive release of non-histone chromatin-binding protein high-mobility group box 1 (HMGB1; Supplementary Fig. S1; ref. 6). CRT exposure promotes phagocytosis of dying cells by antigen-presenting cells (APC) via interactions with low-density lipoprotein receptor-related protein 1. In addition, ATP release into the extracellular environment promotes APC recruitment via purinergic receptors and activation via the nucleotide-binding domain, leucine-rich–containing family, pyrin domain–containing-3 (NLRP3) inflammasome. Finally, HMGB1 release directly activates APCs via Toll-like receptor 4 (TLR4; refs. 6–11). Together, the spatiotemporal presentation of these damage-associated molecular patterns (DAMP), concomitant with tumor antigen exposure from dying tumor cells, can result in the development of antigen-specific antitumor immunity (5).

Although a variety of anticancer therapeutic approaches with different molecular mechanisms of action have been described previously to elicit ICD (radiotherapy, photodynamic therapy, anthracyclines, and oxaliplatin; refs. 5, 12, 13), a common factor is the dysregulation of ER homeostasis (1, 14–18). Overloading the ER's capacity for unfolded polypeptides, or other disruptions to the protein-folding environment, initiates a canonical ER stress response known as the unfolded protein response (UPR). A strong UPR, coupled with the associated release of ER Ca²⁺ stores, is required to trigger ICD-related hallmark DAMPs (7, 16, 17, 19). Translocation of CRT to the cell surface and active secretion of ATP resulting from cellular autophagy follow severe ER stress (7, 20–22), whereas the release of nuclear HMGB1 is thought to occur as the dying cell loses nuclear and plasma membrane integrity (22–24). Synthetic knockdown of the ER stress response during treatment with ICD-inducing therapeutics abrogates ICD (21, 25). Conversely, combining pharmacological agents that perturb ER function with certain non–ICD-inducing compounds have been shown to increase the immunogenicity of dying tumor cells (15, 26–28).

Although some systemic chemotherapies elicit ICD, they can also be associated with lymphopenia and systemic immune alterations often requiring months for T cells to return to pre-treatment levels (29–32). Antibody–drug conjugates (ADC) are a clinically validated strategy for the selected delivery of cytotoxic molecules to tumors that largely avoid persistent depletion of peripheral lymphocytes (33). The combination of ICD-inducing ADCs with checkpoint inhibitors may therefore drive complementary immune mechanisms. Multiple combinations of ADCs with checkpoint inhibitors are being actively evaluated in the clinic with promising early results (34, 35).

Multiple preclinical studies have described successful combinations of ADCs with checkpoint inhibitors and other immunotherapies. Müller and colleagues (36) were the first to show that the HER2-directed ADC trastuzumab emtansine (T-DM1) drove immune infiltration in breast cancer lesions in the clinic. In addition, T-DM1 combined with cytotoxic T-lymphocyte antigen 4 (CTLA-4) and programmed cell death protein 1 (PD-1) blockade amplified antitumor responses in preclinical models (36). Data from this group supported the direct activation of APCs by dolastatins, from which monomethyl auristatin E (MMAE) was synthetically derived, and also MMAE itself (37, 38). Rios-Doria and colleagues (39) showed that tubulysin and pyrrolobenzodiazepine dimer (PBD)-based ADCs successfully combined with immunotherapies in syngeneic tumor models, supportive of ICD induction. Finally, monomethyl auristatin F, a derivative of MMAE with greatly reduced membrane permeability, was shown to drive ICD with the B-cell maturation antigen-directed ADC belantamab mafodotin and combine with a checkpoint modulator (40). Together, these studies highlight the potential immunomodulatory abilities of ADC payloads, particularly those causing microtubule destabilization. Indeed, results supporting ICD induction from vinca domain-binding microtubule-destabilizing agents were predicted by early ICD-screening platforms (41, 42).

Brentuximab vedotin is a CD30-directed vedotin ADC, currently approved for use in the US for multiple CD30-expressing lymphomas (43). Brentuximab vedotin comprises a CD30-directed monoclonal antibody (mAb) conjugated to a protease cleavable maleimidocaproyl–valine-citrulline (mc-vc) linker and the microtubule-disrupting agent (MMAE; refs. 44, 45). Upon internalization, the protease-cleavable linker enables the preferential release of MMAE within CD30-expressing target cells, resulting in microtubule disruption and cell death (45–48). We previously reported on the ability of brentuximab vedotin and MMAE to induce ICD in target cells via microtubule disruption and ER stress in a series of in vitro and in vivo experiments (49, 50). Furthermore, the potential for brentuximab vedotin or MMAE treatment to activate antitumor immunity in combination with PD-1 inhibition was previously evaluated (51). Here, we expand upon our initial findings of brentuximab vedotin and MMAE-induced ICD and the enhanced immune response observed in combination with PD-1 inhibition.

All animal studies were executed in compliance with institutional guidelines and regulations and approved by an Institutional Animal Care and Use Committee (IACUC) and internal review committee under protocols SGE-024 and SGE-029. Mice were age- and sex-matched and randomly assigned to treatment groups in all experiments with females being predominantly used for studies. The number of mice used in each experiment was determined on the basis of previous experience with variability of model performance. Mice were housed in an IACUC-certified housing facility under the supervision of a licensed veterinarian. All animal studies were designed to minimize animal usage and distress.

Cell lines L540 (DSMZ, ACC72; RRID:CVCL_1362; human female), KARPAS 299 (DSMZ, ACC31, RRID:CVCL_1324; human male), HDLM-2 (DSMZ, ACC17; RRID:CVCL_0009; human male), and A20 (ATCC, TIB-208; RRID:CVCL_1940; mouse) were originally purchased from the vendors listed and since maintained by the Seagen cell bank (2016–2023). For each study, cell lines were requested from the cell bank and aliquots were thawed and cultured for periods between 1 and 3 weeks. Cell bank cell lines are routinely checked for contamination, including Mycoplasma, and cell line identity authentication by an STR-based DNA profiling and multiplex PCR, CellCheck 16–Human, or CellCheck 19 Mouse Plus, IDEXX Laboratories, Inc. Following any cell line modifications or cloning, expanded lines were resubmitted for pathogen burden and cell line identity validation before experimental use.

The lymphoma cell line L540cy was grown in RPMI and 20% FBS (Gibco). Cells were treated with IC₅₀ or IC₉₀ concentrations of brentuximab vedotin, doxorubicin, etoposide, or oxaliplatin for 0, 18, 24, or 48 hours. Cells were also treated with controls of a non-binding ADC (IgG-MMAE), chimeric anti-CD30 antibody (cAC10) naked mAb at the IC₅₀ or IC₉₀ concentrations of brentuximab vedotin. Cells were harvested and stained for cell-surface CRT (FMC75, Enzo, RRID:AB_11180758). Staining was concomitant with propidium iodide (PI; Sigma) and all staining was performed as per the manufacturer's recommendations. Surface levels of CRT on PI-negative cells were analyzed by flow cytometry.

L540cy cells were plated at 200,000–3,000,000 cells/well on a 24-well dish and allowed to reach 50% confluence. Cells were then treated with IC₉₀ (1 μg/mL) of brentuximab vedotin or a non-binding ADC (IgG-MMAE), 100 nmol/L (IC₉₀) MMAE, 1 μg/mL oxaliplatin, or 30 μmol/L etoposide for 18 hours. For additional cell lines shown in the Supplementary Figures, approximately 50,000 cells were incubated with IC₂₀–IC₅₀ concentrations of test articles for 48–72 hours. After treatment of cells, 250 μL of media were harvested from each well by tilting the plate and transferred to a 1.5-mL Eppendorf tube. Cells were spun at 1,600 rpm (200 rcf) for 1 minute in a microcentrifuge. Subsequently, 50 μL of media were transferred to wells in a 96-well clear-bottom plate and 50 μL of CellTiter Glo (Promega) was added to each well and shielded from light. Plates were shaken for 1 minute on an orbital shaker, incubated for an additional 10 minutes at room temperature to stabilize luminescent signal, and immediately read on an Envision Plate Reader. ATP levels were normalized to levels from untreated cells.

HMGB1 release was assessed in a similar manner to ATP release. L540cy cells were plated and treated for 18–72 hours. Extracellular HMBG1 was measured in supernatants using an HMGB1 ELISA (Tecan, IBL International).

An Enzo Life Sciences’ ER-ID kit was used for fluorescent labeling of the ER in L540cy cells. Antibodies against phosphorylated inositol–requiring enzyme 1 (pIRE1; NB100–2323, Novus Biologicals, RRID:AB_0145203) and anti-tubulin (YOL1/34, Abcam, RRID:AB_305329), were used for fluorescent imaging. L540cy cell nuclei were stained using Diamond Antifade Mountant with 4′,6-diamidino-2-phenylindole (Thermo Fisher Scientific).

L540cy cells were plated at 500,000 cells/well in a 12-well plate overnight. Cells were then treated with IC₉₀ (1 μg/mL) of brentuximab vedotin or a non-binding ADC (IgG-MMAE), 100 nmol/L (IC₉₀) MMAE, vincristine, or paclitaxel for 18 hours. Treated L540cy cells were collected, lysed in RIPA buffer (Cell Signaling Technology) and centrifuged at 14,000–16,000 rpm for 10 minutes and stored at −20°C. Cell pellets were resuspended in 4X BOLT LDS sample buffer (Thermo Fisher Scientific) and lysed by heating at 95°C for 5–10 minutes to produce cell lysates. Sample lysates were run on a Bis-Tris 4%–12% gradient gel at 140 V for 1 hour 40 minutes in MOPS buffer. The Bis-Tris gel was then transferred onto a nitrocellulose membrane using an iBlot2. Membranes were washed once in 1X TBS and incubated overnight at 4°C with antibodies against total and pIRE1, total and phosphorylated c-Jun N-terminal kinase (pJNK), and activating transcription factor 4 (ATF4) (Cell Signaling Technology and Novus; RRID:AB_10145203, RRID:AB_2058748). Equal loading was confirmed with total actin or tubulin assessment (Cell Signaling Technology, RRID:AB_330288, RRID:AB_2210548). Membranes were washed four times in 1X TBS-T for 5–10 minutes each. Signals were normalized against actin.

Induction of CHOP was measured using a reporter system for CHOP activity according to the manufacturer's instructions (Bright-Glo Luciferase Assay System, Promega). In brief, 100,000 Mia-PaCa-2 cells/well (Signosis; SL-0025-FP) were plated in a 96-well, flat-bottom clear plate (aliquot 150 μL/well). At 24, 48, and 72 hours, plates were removed from the incubator and allowed to come to room temperature. 100 μL of media were removed from the wells and 100 μL of Bright-Glo reagent was added to each well. Plates were shaken for at least 2 minutes before measurement on an Envision plate reader (PerkinElmer).

L540cy cancer cells were plated in 12-well plates and treated with 1 μg/mL of brentuximab vedotin, IgG-MMAE, or 0.1 μmol/L of MMAE for 16 hours. Cells were harvested by moving samples to a 15-mL conical tube and centrifuging for 5 minutes at 400 × g. Tissue culture supernatants were removed and cells resuspended in complete RPMI (10% FCS, 1× sodium pyruvate, 1× MEM NEAA, 1× GlutaMax; Gibco) at a total of 1×10⁷ cells/mL and plated at 250,000 cells/well in a 24-well plate. Human PBMC were thawed and washed three times in complete RPMI and resuspended at a total of 1×10⁷ cells/mL. PBMCs were mixed with treated tumor cells 1:1 at 500,000 cells/well and incubated at 37°C, 5% CO₂, for 24 or 48 hours. Tissue culture supernatants were harvested for analysis of innate and adaptive cell activation using the MILLIPLEX MAP Human Cytokine/Chemokine Magnetic Bead Panel (Sigma).

Five million L-540cy cells were implanted subcutaneously in 25% Matrigel (Corning) in female (RRID:IMSR_TAC:CB17SC) severe combined immune deficient (SCID) mice (Taconic). When tumors reached 200 mm³, mice (n = 5) were randomized into treatment groups and treated intraperitoneally (i.p.) with 3 mg/kg brentuximab vedotin, IgG-MMAE, or cAC10 antibody. Seventy-two hours after treatment, tumors were harvested, weighed, and manually dissociated through a 70-μm cell strainer to generate single-cell suspensions for flow cytometry. Tumor cell suspensions were stained with Live/Dead Viability Dye (Thermo Fisher Scientific) and fluorescent antibodies targeting murine CD45 (RRID:AB_2563542), CD11c (RRID:AB_313779), CD11b (RRID:AB_312791), Ly-6G (RRID:AB_10643269), Siglec-F (RRID:AB_2904295), and MHC II (RRID:AB_493147; BioLegend). Alternatively, portions of each tumor were weighed and lysed in RIPA buffer (Cell Signaling Technology) for cytokine analysis using the MILLIPLEX MAP Human Cytokine/Chemokine Magnetic Bead Panel (Sigma).

A total of 2.5×10⁶ Epstein-Barr virus (EBV)-transformed CD30⁺ LCLs, derived from a healthy donor, were implanted subcutaneously into groups of eight nonobese diabetic (NOD)/SCID/gamma-chain mice (NSG) mice (RRID:IMSR_JAX:005557). For assessing PBMC-mediated rejection of LCL tumors, 0, 2.5, 5, or 10 million thawed autologous donor PBMCs were given to mice intravenously 5 days after tumor implant following which tumor volumes were measured periodically.

For experiments testing the effect of brentuximab vedotin on responding immune cells, mice received a single subtherapeutic dose of brentuximab vedotin or a non-binding ADC (human IgG-MMAE; 1 mg/kg, i.p.) when tumor volumes averaged 250 mm³, then received 2.0×10⁶ autologous PBMCs 3 days post-dose. Four days following the transfer of PBMCs, tumors were harvested from six mice. Tumor RNA was purified using the RNeasy Plus Kit (Qiagen, Germany), and RNA evaluated using the NanoString nCounter Human Immunology panel (NanoString Technologies). Data were processed and analyzed using nSolver software (NanoString Technologies). Eleven days following transfer of PBMCs, tumors were harvested from five mice, weighed, and manually dissociated through a 70-μm cell strainer for FACS analysis. Tumor cell suspensions were stained with Zombie Aqua Viability Dye (BioLegend) followed by incubation with fluorescently labeled antibodies targeting human CD19 (RRID:AB_2564143), CD2 (RRID:AB_2572066), CD8 (RRID:AB_2562790), CD4 (RRID:AB_2572097), CD56 (RRID:AB_2563915), CD45 (RRID:AB_2562057), programmed death-ligand 1 (PD-L1; RRID:AB_2563852), PD-1 (RRID:AB_2922606), and murine CD45.1 (RRID:AB_2562565; BioLegend) in staining buffer (PBS, 2% FCS, 1% normal rabbit serum, 0.05% NaN₃) at 4°C for 30 minutes. Cells were washed and resuspended in 120 μL of staining buffer for plate-based flow cytometry using an Attune NXT flow cytometer. Total events were collected from 80 μL of sample, and flow cytometry-measured cell concentrations were used to calculate numbers of infiltrating immune cells. CD8⁺ T cells were identified as viability dye negative, hCD45⁺, mCD45.1⁻, CD2⁺, and CD8⁺ cells. Natural killer (NK) cells were identified as viability dye negative, hCD45⁺, mCD45.1⁻, CD2⁺, and CD56⁺ cells. Experiments combining brentuximab vedotin with nivolumab treatments were conducted as described above, with mice receiving two doses of nivolumab (10 mg/kg, i.p.) 2 and 7 days after adoptive transfer of PBMCs.

The detection of infiltrating cells in the A20 syngeneic model was performed as follows: mouse tissues were dissected and embedded in OCT (TissueTek; #62550–01) and frozen in isopentane cooled in liquid nitrogen. Sections were cut at 4 μm and stored frozen at −80°C. Slides with frozen tissue sections were thawed, fixed in acetone for 10 minutes at −20°C, and then air-dried for 10 minutes. Immunostaining was performed on the Bond-Max and BondIII autostainers (Leica Biosystems, Inc.) using Bond Polymer Refine Detection (DAB) Kits (Leica; #DS9800). Endogenous peroxidase activity was blocked using PeroxAbolish (BioCare Medical; #PXA969). Protein Block (Dako; #X0909) was used to block non-specific staining. Sections were then incubated with Biotinylated Rat anti-Mouse CD8 clone 4SM15 (eBioscience, RRID:AB_2572771) or Biotinylated Rat IgG2a clone eBR2a (eBioscience, RRID:AB_470084) as isotope control at 3 μg/mL. Bound antibody was detected using the Vectastain Elite ABC HRP Kit (Vector Laboratories) and developed with DAB, from the Bond Polymer Refine Staining Kit (Leica).

Formalin-fixed, paraffin-embedded LCL tumor tissues were sectioned onto adhesive slides, deparaffinized, and rehydrated to distilled water. Slides were steamed for antigen retrieval in Diva solution (Biocare). Sections were quenched with 3% H₂O₂ and blocked with 0.5% Casein and 5% Goat Serum in TBS. Mouse anti-human CD8 (Abcam RRID:AB_443686) was applied for 1 hour, washed, and detected with Biotinylated Goat anti-Mouse IgG (The Jackson Laboratory, RRID:AB_2338569). Sections were then developed with Vectastain Elite ABC HRP Kit (#PK-6100) and Vector ImmPact DAB (#SK-4103).

To generate human CD30-expressing A20 mouse lymphoma cells, cells were transfected with plasmid constructs encoding human TNFRSF8 (NM_001243.4) and sgRNA/Cas9 for control under the endogenous Rosa26 site for constitutive expression. FACS yielded a clonal population of A20 cells that stably expressed human CD30, called A20^hCD30.

A20^hCD30 cells were cultured in RPMI-1640 with 10% FBS, 10 mmol/L HEPES, 1 mmol/L sodium pyruvate, penicillin (100 U/mL), and streptomycin (100 μg/mL). A20^hCD30 cells were treated with 1 μg/mL brentuximab vedotin or 100 nmol/L mc–vc–MMAE for 4 days. To prepare dying cells for immunization, treated A20^hCD30 cells were overlaid atop Histopaque and centrifuged at 2,000 × g for 30 minutes. Dead and dying cells were pelleted underneath the Histopaque layer and viability was assessed to be <20% live cells by trypan blue exclusion. Dead and dying A20^hCD30 cells were resuspended in PBS and a total of 2 × 10⁶ cells were injected into the peritoneum of immune-competent BALB/c mice (n = 8). Flash-frozen A20^hCD30 cells were prepared by submerging cells in liquid nitrogen for 10 seconds, followed by immersion in 37°C water until completely thawed. The liquid nitrogen freeze-thaw process was repeated five times. Fourteen days later, mice received a second immunization with dead and dying cells prepared in the same manner.

Twenty-one days after initial immunization, mice were subcutaneously implanted with a total of 5 × 10⁶ wild-type (WT) A20 cells and monitored for tumor growth. In some experiments, mice were treated with four administrations of anti–PD-1 over 2 weeks (1 mg/kg; BioLegend, RRID:AB_2783090).

Immune-competent BALB/c mice (n = 4) were immunized with brentuximab vedotin-killed or flash-frozen–killed A20^hCD30 cells as previously described. Sixteen weeks after initial immunization, spleens were harvested from immunized mice or naive BALB/c mice and manually homogenized. Splenic homogenates were combined from four mice per immunization group and CD3⁺ T cells were isolated using the EasySep Mouse T-cell Enrichment Kit (Stem Cell Technologies). One million CD3⁺ T cells were administered intravenously into WT A20 tumor-bearing NSG mice when tumors averaged 100 mm³ (n = 5–8). Tumors were subsequently monitored for growth and at study takedown for CD8⁺ T-cell content by flow cytometry.

Cells were harvested 48 hours following treatment with test articles. Supernatants were transferred to round-bottom plates to collect non-adherent cells. Versene (200 μL; Thermo Fisher Scientific) was then added to each well, and upon detachment, cells were transferred to the corresponding round-bottom plates used to collect non-adherent cells. Cells were pelleted by spinning at 1,500 rpm for 5 minutes. Plates were decanted, washed with PBS and, following an additional spin, cells were stained with Live/Dead Yellow cell viability dye (Thermo Fisher Scientific) per the manufacturer's protocol. Surface staining for calreticulin and Annexin V was then performed in FACS buffer (2% FBS/PBS) with the addition of Annexin V binding buffer (BD Biosciences). Cells were then washed with FACS buffer + Annexin V binding buffer before fixation and permeabilization using the True Nuclear Kit, 96-well plate protocol (BioLegend). Overnight intracellular staining at 4°C was then performed upon the addition of 100 μL of permeabilization buffer containing antibodies to each well. A list of antibodies with corresponding RRIDs used for intracellular staining is provided in Supplementary Table S1. Cells were then washed three times in permeabilization buffer, resuspended in PBS, and analyzed on an Attune Flow Cytometer.

Descriptive statistics were used throughout. Where applicable, significance was determined by an ANOVA with Sidak or Dunnett's multiple comparisons tests, calculated using GraphPad Prism 8.1 (RRID:SCR_000306).

All available data are presented.

Cell surface CRT expression, the first hallmark of ICD, was evaluated on CD30⁺ L540cy cells to determine whether cell death induced by brentuximab vedotin is consistent with ICD induction. The L540cy cell line is a derivative of the CD30⁺ HL L540 line adapted for growth as a xenograft. L540cy cells were treated with brentuximab vedotin, IgG-MMAE, oxaliplatin (a known ICD inducer; ref. 12), or etoposide (which does not induce ICD; ref. 27). After 18 hours of treatment, cells were harvested, stained for cell-surface CRT expression, and levels were analyzed by flow cytometry. The mean fluorescence intensity of surface CRT at 18 hours was increased in live cells (Annexin V⁻/PI⁻) after brentuximab vedotin treatment, whereas IgG-MMAE or etoposide-treated cells exhibited minimal changes in surface CRT (Fig. 1A). Furthermore, the proportion of brentuximab vedotin–treated cells expressing surface CRT was increased at 48 hours (Fig. 1B). Importantly, similar increases in expression of cell surface CRT were observed in three separate human lymphoma cell lines and the murine A20 tumor cell line engineered to express human CD30 (A20^hCD30; Supplementary Fig. S2A), demonstrating a common effect across multiple target cell lines.

The second hallmark signature of ICD induction is active secretion of ATP through pannexin channels. Release of ATP serves as a potent chemotactic signal recruiting immune cells to the inflamed site (52). L540cy cells were treated with brentuximab vedotin, a non-binding ADC, free MMAE, oxaliplatin, or etoposide for 18 hours. Cell-free supernatant was collected and analyzed for ATP levels that were normalized to levels from untreated cells. Following treatment with brentuximab vedotin or free MMAE, extracellular ATP levels approximated or were greater than those obtained following treatment with oxaliplatin (Fig. 1C). Treatment with etoposide, a cytotoxic agent that does not induce ICD, did not induce robust ATP secretion from dying L504cy cells. Heightened ATP secretion following brentuximab vedotin and MMAE treatment was also observed with additional lymphoma lines KARPAS 299 and the murine A20^hCD30 (Supplementary Fig. S2A).

Finally, extracellular HMGB1 elicits inflammatory responses as a ligand for TLR4, TLR2, and advanced glycosylation end product-specific receptor (AGER, RAGE), triggering potent NF-κB responses in immune cells (1, 6, 10). After 18 hours of treatment with brentuximab vedotin, free MMAE, etoposide, or oxaliplatin, HMGB1 was elevated in the cell-free supernatant of L540cy cells (Fig. 1D). Increased release of HMGB1 was also observed for KARPAS 299, HDLM-2, and A20^hCD30 cell lines following brentuximab vedotin or MMAE treatment at 72 hours (Supplementary Fig. S2A). Elevation of HMGB1, a late-stage marker of ICD, was seen upon etoposide treatment, but occurred in the absence of robust CRT exposure and ATP release, indicating that molecules like MMAE are unique in their ability to drive all three markers of ICD. Overall, these results demonstrate that treatment of CD30⁺ lymphoma cell lines with brentuximab vedotin or MMAE was associated with the presentation of three hallmarks of ICD.

The ER is a dynamic cellular organelle that expands and contracts depending on the functional needs of the cell (53, 54). Expansion and contraction occur along the tubulin highway (55) and catastrophic disruption of that network can inhibit normal ER function (56). As such, the relationship between MMAE-mediated disruption of the microtubule network and the resultant effects on ER structure and stress response in treated cells was investigated. Compared with control (IgG-MMAE), treatment of L540cy cells with brentuximab vedotin or free MMAE led to a striking qualitative disorganization of the perinuclear ER network (Fig. 2A). Disruption of protein folding in the ER results in a well-defined stress response, referred to as the UPR, purposed for re-establishing ER homeostasis. To determine whether the observed disorganization of the ER network occurs concomitant with the UPR response, L540cy cells were treated for 18 hours with brentuximab vedotin or free MMAE. After 18 hours, there was a rapid induction and activation of multiple arms of the UPR, including increased pIRE1, downstream pJNK, and ATF4 expression; pIRE1 and pJNK were minimally expressed with controls (no treatment and IgG-MMAE; Fig. 2B and D). Furthermore, microscopy showed pIRE1 to be spatially localized with the structurally disrupted ER in MMAE-treated L540cy cells compared with control, corroborating a potent UPR response (Fig. 2C). Further evidence for activation of the three arms of the UPR was provided by staining of intracellular XBP1, ATF6, and phosphorylated protein kinase RNA-activated-like ER kinase by flow cytometry for CD30-expressing lymphoma lines 48 hours after treatment with brentuximab vedotin or MMAE (Fig. 2E). Significant increases in UPR stress markers were observed for each cell line treated with brentuximab vedotin compared with untreated controls. Compared with human lymphoma lines, the engineered A20^hCD30 syngeneic lymphoma expressed much lower levels of CD30 (Supplementary Fig. S2B) and showed stronger UPR induction from free MMAE treatment compared with brentuximab vedotin.

The effects of using microtubule-destabilizing agents (MMAE and vincristine) versus a stabilizing agent (paclitaxel) were investigated to determine whether the underlying mechanism behind ER stress induction was associated with collapse of the microtubule network. Although IRE1 phosphorylation occurred after treatment with all agents, indicating acute ER stress and initial activation of stress responses, only the microtubule-destabilizing agents MMAE and vincristine elicited downstream JNK phosphorylation (Fig. 2F). Pixel densities for all western blot analysis images are provided in Supplementary Fig. S3A–S3D. Robust pIRE1 and pJNK upregulation was also observed following treatment with brentuximab vedotin or MMAE in additional lymphoma lines by intracellular flow cytometry (Supplementary Fig. S4A). Sustained activation of the UPR response ultimately leads to the initiation of ICD through the induction of CHOP. To assess CHOP induction from microtubule-stabilizing versus -destabilizing compounds, an ER stress reporter system of Mia-PaCa-2 cells transduced with a CHOP-driven luciferase was employed. In this system, paclitaxel-driven microtubule stabilization did not drive equivalent CHOP induction as the microtubule-destabilizing compounds MMAE and vincristine, despite visible cell death (Fig. 2G). CHOP induction was confirmed in three additional lymphoma lines treated with brentuximab vedotin or free MMAE by intracellular flow cytometry (Supplementary Fig. S4B).

Induction of ER stress with microtubule-disrupting agents, including MMAE, was validated in vivo by engrafting CHOP-luciferase Mia-PaCa-2 cells into NSG. Xenografts were treated intratumorally with MMAE, vincristine, or paclitaxel and monitored daily for luciferase activity. Soon after administration, CHOP induction, as monitored by luciferase activity, was observed from tumors treated with MMAE or vincristine, peaking at 72 to 96 hours after treatment (Fig. 2H). Notably, luciferase was detectable 24 hours after treatment, indicating that the response happened quickly and increased over time. Treatment with microtubule-stabilizing paclitaxel did not elicit luciferase activity, but inhibited tumor growth, indicating that antitumor activity of paclitaxel did not rely on ER stress induction.

After confirming that cells treated with brentuximab vedotin exhibit signs of severe ER stress and express the key ICD hallmarks, we questioned whether the dying cells were able to activate innate and adaptive immune responses. To this end, brentuximab vedotin–killed L540cy cells were fed to human monocyte-derived CD11c⁺ dendritic cells (DC) in vitro and proinflammatory chemokine and cytokine production were measured in supernatants. Notably, macrophage inflammatory protein-1α (MIP-1α) and IFN gamma–induced protein 10/C-X-C motif chemokine ligand 10 (IP10/CXCL10) were increased in supernatants from DCs that were fed cells treated with brentuximab vedotin, MMAE, etoposide, or oxaliplatin compared with IgG-MMAE (Fig. 3A). Treatment with the non-targeting ADC did not significantly promote proinflammatory mediators, showing that delivery of MMAE to tumor cells was required for driving proinflammatory cytokines from human DCs. Oxaliplatin and etoposide, which were both shown to induce HMGB1 release from L540Cy cells (Fig. 1D), also drove proinflammatory cytokine secretion consistent with this component of ICD. In addition, when brentuximab vedotin- and MMAE-treated cells were co-cultured with DCs and autologous T cells, increased production of T-cell cytokines IFNγ and TNFα was observed, compared with IgG-MMAE and a trending increase compared with etoposide-treated cells (Fig. 3A). Together, the increase in DC- and T-cell–derived proinflammatory cytokines demonstrates the ability of MMAE- and brentuximab vedotin–killed tumor cells to promote APC activity.

The host innate immune response to brentuximab vedotin–treated tumors was investigated in vivo in a human L540cy xenograft model in SCID mice. SCID mice lack T and B cells but retain innate immune cells enabling evaluation of APC activation and recruitment following ADC killing of xenograft tumors. Although brentuximab vedotin treatment did not significantly influence the scale of immune infiltration into the tumor by 72 hours, it did begin to reshape the immune landscape as shown by increased proportions of antigen-presenting CD11c⁺ DCs in tumors relative to Ly6g⁺ neutrophils (Fig. 3B). Skewing of the available innate immune compartment toward mature DCs (CD11c⁺ DCs) may reflect recruitment of DCs into the tumor or differentiation of infiltrating monocytes following tumor cell killing by brentuximab vedotin. Furthermore, consistent with innate immune cell activation and recruitment following ICD induction, brentuximab vedotin treatment, compared with IgG-MMAE and cAC10, led to significant or trending increases in intratumoral and serum murine proinflammatory cytokines TNFα, IP10, and MIP-1α (Fig. 3C and D). Furthermore, tumor-derived cytokine concentrations were notably higher than those detected in circulation, consistent with tumor-localized production. In vitro activation of human immune cells and recruitment of murine DCs in the L540Cy xenograft model following treatment with brentuximab vedotin are consistent with ICD induction and highlight the potential for driving focused immune activation in the TME following treatment with MMAE-based ADCs.

The ability of brentuximab vedotin treatment to promote proinflammatory cytokine production in xenografts is consistent with ICD-driven activation of innate immunity. To further investigate the impact of this inflammatory response on the recruitment and activation of antigen-specific cytotoxic T cells to the TME, we used a humanized LCL/PBMC model. This model pairs EBV-transformed CD30⁺ LCL tumors with adoptively transferred EBV-reactive autologous PBMCs to evaluate antigen-specific T-cell responses in vivo. In this model, intravenous transfer of autologous PBMCs results in a titratable, immune-mediated elimination of the tumor as measured by tumor volume reduction (Fig. 4A). To evaluate whether treatment of LCL tumors with brentuximab vedotin can influence the innate and adaptive arms of the immune system, mice with established LCL tumors (200 mm³) were given a subtherapeutic dose of brentuximab vedotin (the highest dose without consistent tumor reduction; 1 mg/kg) followed by adoptive transfer of autologous PBMCs 3 days later. Tumors from brentuximab vedotin–treated and control groups were harvested for analysis 4 and 11 days after PBMC transfer (Fig. 4B). NanoString analysis of brentuximab vedotin-treated LCL tumors harvested 4 days after PBMC transfer showed increased lysosome-associated membrane glycoprotein 3 transcript consistent with UPR-induced autophagy in tumor cells (Fig. 4C; ref. 57). UPR-induced autophagy in tumor cells undergoing ICD drives the hallmark release of extracellular ATP that signals via purinergic receptors to activate the NLRP3 inflammasome in neighboring macrophages (22, 58, 59). Consistent with increased extracellular ATP, LCL tumors treated with brentuximab vedotin showed increased NLRP3 transcript (Fig. 4C). Finally, signatures reflecting early innate (CD11b, CXCL9) and cytotoxic immune cell (IFNγ, Granzyme B) activity and recruitment were strongly upregulated in brentuximab vedotin-treated samples compared with non-targeted ADC and antibody alone controls (Fig. 4C). Eleven days after PBMC transfer, flow cytometry of processed tumors showed significantly increased CD8⁺ T cells and CD56⁺ NK cells in brentuximab vedotin–treated mice, demonstrating enhanced recruitment, and/or proliferation of innate and adaptive cytotoxic cells (Fig. 4D; Supplementary Fig. S5). Results from the LCL/PBMC xenograft model are consistent with upregulation of inflammatory murine chemokines in L540cy xenografts treated with brentuximab vedotin (Fig. 3) and substantiate enhancement of both innate cell and antigen-specific T-cell responses following brentuximab vedotin–driven ICD induction.

The “gold-standard” experiment for assessing a compound's ability to elicit functional ICD involves vaccination of mice with treated syngeneic tumor cells followed later by tumor rechallenge. Bona fide ICD-inducing compounds drive lasting antigen-specific memory responses that prevent tumor growth upon subsequent implant with the same cell line (6, 12, 60). To validate ICD induction and tumor-specific adaptive immune responses following brentuximab vedotin treatment, the human CD30-engineered syngeneic A20 B-cell lymphoma model (A20^hCD30) was used. A20^hCD30 cells were killed in vitro with brentuximab vedotin or MMAE and injected into WT BALB/c mice. As a negative control, A20^hCD30 cells were flash-frozen in liquid nitrogen providing tumor-associated antigens in the absence of ICD induction. Mice were immunized two times, and 21 days after the initial immunization mice were implanted with WT A20 lymphoma cells and monitored for tumor growth. The majority of mice immunized with brentuximab vedotin- or MMAE-killed cells began to reject the tumors around 7 days after implant, resulting in 5/8 and 6/8 complete responses (CR)s, respectively, consistent with a primed adaptive immune response. In contrast, unimmunized naive mice and mice immunized with control freeze-thawed cells showed accelerated tumor growth and fewer CRs, 1/5 and 3/8 CRs, respectively (Fig. 5A). Consistent with memory T-cell generation following vaccination, mice that had previously received brentuximab vedotin- or MMAE-killed cells showed rapid infiltration and increased murine CD8⁺ T cells within A20 tumors shortly after implant (Fig. 5B).

Clinical response rates to PD-1/PD-L1 therapies correlate with the presence of CD8⁺ T cells in the TME (61). Given that T cells from mice immunized with brentuximab vedotin–killed tumor cells could confer anamnestic antitumor protection, the combination with PD-1 inhibition was investigated. BALB/c mice were immunized with brentuximab vedotin- or MMAE-killed A20^hCD30 cells, implanted with WT A20 tumor cells 14 days later, and then given anti–PD-1 7 days after tumor implantation. Although immunization with brentuximab vedotin- or MMAE-killed A20 cells alone again provided antitumor immunity, the addition of anti–PD-1 robustly increased the proportion of CRs (Fig. 5C). These results demonstrate that MMAE-induced ICD can be combined with PD-1 blockade to improve antitumor immunity.

To further demonstrate the impact of MMAE-induced ICD on the long-lived cytotoxic T-cell response, CD3⁺ T cells were isolated from the spleens of tumor cell-immunized BALB/c mice 5 months after vaccination and adoptively transferred to NSG mice harboring existing A20 tumors (approximately 100 mm³). Use of NSG mice precludes host antitumor immunity, enabling the assessment of adoptively transferred T-cell activity. NSG mice receiving no T cells, T cells transferred from naive mice, or T cells from mice immunized with control freeze-thawed–killed tumor cells showed minimal tumor control and were sacrificed at study take down (day 19), when all control tumors exceeded 1,000 mm³. In contrast, mice receiving T cells from mice immunized with brentuximab vedotin–killed A20^hCD30 tumor cells showed significantly slowed tumor growth up to day 19 as determined by the tumor volume area under the curve (Fig. 5D and E). Furthermore, mice receiving T cells from brentuximab vedotin–immunized mice showed a trend toward increased CD8⁺ T-cell infiltration in A20 tumors at take down (Fig. 5F). Together, these results are consistent with enhanced T-cell immunity elicited by known ICD inducers in similar models alone and in combination with anti-PD-1 (21, 27, 39) and provide evidence that MMAE is an ICD-inducing agent capable of promoting long-lived T-cell responses.

The combination of brentuximab vedotin and PD-1 inhibition was also evaluated in the humanized autologous LCL/PBMC xenograft model. In this model, LCL tumor clearance is mediated by autologous EBV-specific cytotoxic T-cell responses. As previously described, treatment of LCL tumors with a subtherapeutic dose of brentuximab vedotin augmented infiltration of cytotoxic T and NK cells and increased production of IFNγ (Fig. 4C and D). Inhibitory PD-L1 expression can be driven though IFNγ and the JAK/STAT pathway following activation of cytotoxic T cells and may limit antitumor immunity (62). Flow cytometric evaluation of on-study LCL tumor cells and intra-tumoral CD8⁺ T cells in this model showed high expression of PD-L1 and PD-1, respectively, supporting the presence of this inhibitory axis in LCL tumors (Fig. 6A).

PBMC-humanized mice receiving a subtherapeutic dose of brentuximab vedotin showed accelerated LCL tumor rejection relative to untreated humanized mice, consistent with enhanced T-cell activity (Fig. 6B and C). Although blocking of human PD-1 with nivolumab did not significantly accelerate tumor rejection as a single agent, the combination of brentuximab vedotin and nivolumab showed superior tumor rejection compared with either treatment alone (Fig. 6B and C). Results from the humanized LCL tumor model are consistent with the murine A20 vaccination model and further substantiate the complementarity of MMAE-driven ICD and PD-1 inhibition. In addition to direct tumor cell killing by brentuximab vedotin, priming and recruitment of T cells following ICD induction may help potentiate therapeutic blockade of PD-1/PD-L1 in CD30-expressing lymphomas.

Failures in immunosurveillance can lead to cancer progression (1, 2). Specifically, immunological ignorance of tumor-associated antigens, immune exclusion, or active immunosuppression within the TME are key barriers to protective cancer immunity (1). Overcoming these barriers may be possible by eliciting tumor cell death and providing tumor-associated antigens in an inflammatory context (6). In contrast with normal cellular apoptosis, ICD elicits the release of inflammatory DAMPs that promote APC activation and productive T-cell priming against tumor antigens (6). Therefore, induction of ICD by anticancer treatments is highly desirable to improve patient immune responses to cancer.

Brentuximab vedotin is an ADC that selectively delivers the highly potent microtubule inhibitor MMAE to CD30-expressing target cells (44). MMAE is a synthetic derivative of the dolostatin 10 family of microtubule inhibitors with mechanistic similarities to vinca alkaloid antineoplastic agents, vincristine and vinorelbine (45). Here, treatment of CD30⁺ lymphoma cells with brentuximab vedotin or MMAE was associated with induction of the three hallmarks of ICD: expression of CRT on the cell surface, secretion of ATP, and extracellular release of HMGB1 to a similar degree as the known ICD-inducing agent oxaliplatin. Etoposide, a compound described not to induce ICD, was unable to fulfill all three of these characteristics (a schematic of the proposed mechanism is shown in Supplementary Fig. S6C). Unlike other known ICD-inducing agents that affect DNA replication/integrity (18, 12), these results demonstrate that MMAE-mediated ICD is induced through microtubule disruption and subsequent disorganization of the ER, leading to a potent ER stress response. Phagocytic uptake of brentuximab vedotin–killed cells by APCs led to the production of proinflammatory cytokines and efficiently primed the host immune system to mount a long-lived antitumor T-cell response. Overall, our results expand upon our understanding of brentuximab vedotin's multifaceted anticancer mechanism of action (Supplementary Fig. S6).

Brentuximab vedotin–induced innate cell activation was observed in an in vivo murine model by inducing high expression of hallmark ICD cytokines. Blood biomarker analyses from a phase I/II study in patients with relapsed or refractory HL demonstrated a marked increase in similar proinflammatory cytokines and chemokines following treatment with brentuximab vedotin (33). These in-human results validate the findings from the preclinical experiments investigating the potential for brentuximab vedotin to induce ICD.

Antitumor immunity imparted by brentuximab vedotin–killed cells led to increased CD8⁺ T-cell infiltration and improved tumor clearance in both murine and humanized tumor models. Given that clinical response rates to checkpoint inhibitors correlate with the presence of CD8⁺ T cells in the TME (61), it was hypothesized that brentuximab vedotin–mediated ICD and the downstream effects on T cells could be augmented by checkpoint inhibition. Our models confirmed that brentuximab vedotin–mediated ICD and resultant immune-mediated tumor rejection were enhanced by PD-1 inhibition.

These results provide a mechanistic rationale for investigating the antitumor response of brentuximab vedotin and checkpoint inhibition as a novel regimen in CD30⁺ lymphomas. Previously published results of treatment with brentuximab vedotin with a PD-1 inhibitor (nivolumab) in patients have demonstrated promising antitumor responses. In a phase I/II study in which 91 patients with relapsed or refractory HL were treated with brentuximab vedotin plus nivolumab, the CR rate was 67% and the objective response rate was 85%. Both the CR and objective response rates with brentuximab vedotin plus nivolumab were higher than previously observed with brentuximab vedotin or nivolumab alone (63). A similar study of brentuximab vedotin plus nivolumab in patients with relapsed or refractory non-Hodgkin lymphomas demonstrated an objective response rate of 70% and a 50% CR rate in patients with mediastinal gray zone lymphoma (CHECKMATE 436; ref. 64). Ongoing studies are investigating the combination in different populations; interim analyses of these studies have demonstrated durable response rates in patients aged ≥60-years with classical HL (65) and high progression-free survival rates in patients aged 9–30 years with relapsed or refractory classical HL (66).

The results of this study further demonstrate that the small-molecule MMAE itself is a potent driver of ICD irrespective of the anti-CD30 antibody backbone. Therefore, the ability to induce targeted immune activation is likely intrinsic to other vedotin ADCs. Indeed, preclinical studies with other vedotin ADCs (tisotumab vedotin and enfortumab vedotin) have demonstrated the presence of ICD hallmarks in ADC-treated cell lines (67, 68). Collectively, these findings support the hypothesis that induction of ICD is a vedotin ADC class effect and provide rationale for the clinical investigation of vedotin ADCs in combination with checkpoint inhibitors and other immunotherapies.

These studies were funded by Seagen Inc. R.A. Heiser, W. Zeng, M. Ulrich, P. Younan, and R. Thurman are employees of Seagen and hold stocks and shares in Seagen. A.T. Cao, M.E. Anderson, M. Jonas, C.-L. Law, E.S. Trueblood, and S.J. Gardai were employees of Seagen at the time of the studies. R.A. Heiser reports a patent for US11299543B2 issued; as well as employment and being a stockholder of Seagen Inc. A.T. Cao reports personal fees from Seagen during the conduct of the study. P. Younan reports employment with Seagen. S.J. Gardai reports other support from Seagen during the conduct of the study; as well as a patent for U.S. 201862668104 pending and U.S. 20200239585 pending. No other disclosures were reported by the authors.

R.A. Heiser: Conceptualization, investigation, writing–original draft, writing–review and editing. A.T. Cao: Conceptualization, data curation, formal analysis, investigation. W. Zeng: Investigation. M. Ulrich: Investigation. P. Younan: Investigation, methodology. M.E. Anderson: Investigation. E.S. Trueblood: Investigation. M. Jonas: Validation. R. Thurman: Writing–review and editing. C.-L. Law: Writing–review and editing. S.J. Gardai: Conceptualization, investigation, writing–original draft, writing–review and editing.

Medical writing support was provided by Moamen Hammad, and editorial support, including formatting, proofreading, and submission, was provided by George Chappell and Travis Taylor, all of Scion, London, UK, supported by Seagen according to Good Publication Practice guidelines (https://www.acpjournals.org/doi/10.7326/M22-1460). Some of these results were presented at the 2015, 2016, and 2017 American Association for Cancer Research (AACR) annual meetings (49–51).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.