Background

Survival outcomes for patients with high-risk neuroblastoma remain dismal despite use of intense multimodal therapies.1 Neuroblastoma tumors express cell surface molecules that are suitable for targeting with immune-based strategies,2 as shown by the clinical efficacy of monoclonal antibodies (mAbs) targeting the membrane disialoganglioside GD2.3 However, GD2 immunotherapy is usually accompanied with off-tumor, on-target toxicities due to GD2 expression in some mature human tissues.4 Furthermore, some tumors progress despite anti-GD2 immunotherapy,5 together highlighting the need to identify additional tumor-restricted targets and develop combinatorial immunotherapeutic approaches with improved clinical efficacy and safety profiles.

Recently, our group identified glypican 2 (GPC2) as a novel immunotherapeutic target differentially expressed in neuroblastomas compared with normal tissues.6 GPC2 is a signaling coreceptor heparan sulfate proteoglycan that promotes neuroblastoma oncogenesis and is transcriptionally regulated by N-myc proto-oncogene protein (MYCN).6 To therapeutically target GPC2, we engineered an antibody–drug conjugate (ADC) by conjugating an anti-GPC2 antibody (D3-GPC2-IgG1) to pyrrolobenzodiazepine (PBD) dimers, obtaining a drug:antibody ratio (DAR) of 2.6 and resulting in robust antitumor effects in human neuroblastoma patient-derived xenograft (PDX) models.6 7 We additionally demonstrated that D3-GPC2-PBD induced apoptosis and bystander cytotoxicity by binding a highly tumor-specific conformational epitope of the core GPC2 protein completely shared by humans and mice.7 The equivalent binding of this ADC to both human and murine GPC2 enables studying its effects on the host immune system in immunocompetent syngeneic murine models, using the identical drug that would be used for human clinical testing.

The concept of immunogenic cell death (ICD) relies on the activation of immune responses initiated as dying cells release damage-associated signals to stimulate innate and adaptive immunity.8 Chemotherapy and radiotherapy have been shown to induce ICD and enhance the antitumor activity of diverse immunotherapies.9 10 ADCs are also an attractive platform to selectively induce ICD and modulate the tumor immune microenvironment (TIME), as they can selectively deliver immunogenic payloads to the tumor bed.11 12 In turn, ADCs bearing DNA-damaging (eg, PBD) or tubulin-targeting (eg, maytansine, mertansine, or monomethyl auristatin F) payloads have previously been shown to induce potent antitumor immune responses and potentiate the efficacy of several immunoregulators.13–17 However, these combination immunotherapies have focused on exploiting the antitumor properties of T cells, whereas therapeutically combining ADCs with immunomodulators enhancing the antitumor properties of macrophages remains poorly explored.18

Here, using both GPC2-expressing murine and human neuroblastoma preclinical models, we demonstrate that the D3-GPC2-PBD ADC induces ICD, modulates the neuroblastoma TIME and enhances the efficacy of macrophage-stimulating therapies including anti-CD40 agonist and anti-CD47 antagonist mAbs, setting the stage for the potential clinical translation of GPC2 ADCs in combination with other immunomodulators.

Materials and methodsCell lines

Human neuroblastoma NB-EbC1, SMS-SAN, NB-SD, and NB69 cell lines were obtained from the Children’s Hospital of Philadelphia (CHOP) repository, and the HEK293T cell line was obtained from the American Type Culture Collection (ATCC, catalog number CRL-3216). Murine neuroblastoma cells 9464D and NXS2 were kindly provided by Dr Paul Sondel and previously characterized in detail.19 All tumor cell lines were cultured in Roswell Park Memorial Institute (RPMI) medium containing 10% fetal bovine serum (FBS), 2 mM L-glutamine and 1% streptomycin/penicillin (RPMI complete; RPMIc) at 37°C under 5% CO2. Cell lines were routinely tested to confirm lack of mycoplasma contamination via the MycoAlert Mycoplasma Detection Kit prior to use in experiments.

Constructs, DNA transfection, lentiviral production, and transduction

Lentiviral plasmid pLEX_307 containing murine Gpc2 cDNA was cloned and characterized as previously described.7 Addgene vectors pLenti PGK V5-LUC Neo (w623-2) and pLenti CMV green fluorescent protein (GFP) Puro (658-5) were a gift from Campeau et al20 and were used to tag neuroblastoma cells with firefly-luciferase and GFP, respectively. DNA transfections, lentiviral production and transduction were performed as previously described.7 21

Preparation of antibodies and ADCs

A naïve human Fab phage display library was used for selection of Fabs against purified recombinant GPC2 ectodomain6 or the SARS-CoV-2 spike protein to use as an isogenic control antibody.22 The best binders were selected to be used for conversion to full-length human IgG1 format. CHO or Expi293 stable cell lines were generated for antibody production. Two PBD-payload ADCs were developed, each with a maximum theoretical DAR of 4. The two variants delivered the same PBD warhead; however, they employed different conjugation chemistries, and the antibody conjugation sites were two amino acids apart. In the first ADC preparation, we produced a Y289L mutant β4Gal-T1 enzyme which was used to conjugate GalNaz to terminal biantennary N-acetylglucosamines followed by click-reaction conjugation of a dibenzocyclooctyne (DBCO)-activated PBD-payload (Levena Biopharma). Antibodies with terminal biantennary N-acetylglucosamines were expressed in CHO cells with knocked-down uridine diphosphate galactose transporter UGT1.23 The UGT1(−) CHO cell line was generated using CRISPR/Cas9. In the second ADC variant, antibody glycans were removed using PNGase F and amino-PEG6-bis-PEG5-azide (ConjuProbe) and conjugated to Q295 using bacterial transglutaminase.24 DBCO-activated PBD payload was conjugated in a follow-up click reaction. The efficiency of GlcNaz and amino-PEG6-bis-PEG5-azide conjugation was evaluated with liquid chromatography/mass spectometry (LC/MS) and was more than 95% efficient. Excessive payload conjugation induced antibody aggregation; therefore, payload was added to the reaction mixture in small amounts and the reaction was monitored using LC/MS. The reaction was stopped when the appearance of molecular aggregates was detected. This resulted in a maximal amount of conjugated payload for a given ADC variant. Enzyme-mediated conjugation steps were performed at 37°C in phosphate-buffered saline (PBS). PBD-payload conjugation was performed using a mixture of 60% PBS and 40% propylene glycol at room temperature.

D3-GPC2-IgG internalization and D3-GPC2-PBD cytotoxicity

Internalization studies were performed on an Incucyte ZOOM live-cell monitoring system (Essen Bioscience) using 5 µg/mL of red-labeled D3-GPC2-IgG1 antibody, as previously detailed.21 Cytotoxicity of D3-GPC2-PBD, AB1-SARS-CoV-2-PBD, and free PBD (Levena Biopharma) was measured using Cell Titer-Glo V.2.0 Cell Viability Assay (Promega).21

Evaluation of ICD

ICD markers including calreticulin, HSP70 and HSP90 membrane translocation, and HMGB1 and ATP release were measured after GPC2 ADC treatment (150 ng/mL) in GPC2 and empty-vector NXS2 and 9464D murine cell lines and in human neuroblastoma cell lines NB-EbC1, SMS-SAN, NB69, and NB-SD. ICD was also evaluated after treatment with control ADC (AB1-SARS-CoV-2-PBD, 150 ng/mL), free PBD at ADC equimolar concentrations and doxorubicin (50 and 500 nM for NXS2 and 9464D cell lines, respectively), a well-known ICD inducer and used as a positive control for these experiments.25 At 72 hours, cells were collected, stained with Annexin-V-APC (1:50, 422201; BioLegend), calreticulin-PE (1:50, D3E6; Cell Signaling), HSP70-Alexa Fluor488 (1:20, 648003; BioLegend), HSP90-APC (1:100, NB110-61640APC; Novusbio) and Dapi (1:3000, 422801; BioLegend) and processed by flow cytometry. Percentages of apoptotic cells were defined as Annexin-V+/Dapi− cells. Mean fluorescence intensity (MFI) of calreticulin, HSP70, and HSP90 was determined in Annexin-V+/Dapi− cells. Secretion of HMGB1 in the extracellular media was measured using an IBL International’s HMGB1 ELISA kit (Tecan). Secretion of ATP in the extracellular media was measured using an ENLITEN ATP Assay System (Promega).

In vitro macrophage-mediated phagocytosis

To generate macrophages, human monocytes were obtained from the Immunology Core at University of Pennsylvania and cultured at 1×106 cell/mL density in 75 cm2 flasks with RPMIc with human recombinant macrophage colony-stimulating factor (M-CSF, 25 ng/mL; Peprotech, 300–25) for 7 days.

For phagocytosis assays, GFP-labeled NB-EbC1, SMS-SAN, NB69, and NB-SD cells were plated in 6-well plates overnight (50,000 cells/well) and treated with D3-GPC2-PBD (0.15 ng/mL), AB1-SARS-CoV-2-PBD (0.15 ng/mL), and D3-GPC2-PBD (0.15 ng/mL) added 4 hours after addition of D3-GPC2-Fab (×1000 equimolar excess), free PBD at ADC equimolar concentrations or vehicle for 72 hours. Then, macrophages were added in a tumor:macrophage ratio of 1:1 together with anti-CD47 (MIAP-410, 10 µg/mL) or IgG1 isotype (MOPC-21, 10 µg/mL). After 24 hours, cells were stained with human CD11b-PE (#301306, BioLegend) and CD40-APC (#334309, BioLegend) antibodies and were analyzed on a Beckman CytoFLEX S (Beckman Coulter). Percentage of phagocytosis was determined by quantifying live GFP-FITC and CD11b-PE-positive macrophages. MFI of CD40 was determined in GFP/CD11b-expressing cells. All phagocytosis assays were repeated with macrophage cultures derived from at least two different donors.

Murine models

To generate syngeneic neuroblastoma allografts, GPC2 or empty-vector 9464D or NXS2 cells (2×106) were injected subcutaneously with matrigel (Corning, 100 µL) in the flanks of 6-week-old C57BL/6 or A/J female mice, respectively (Jackson Laboratory, Bar Harbor, Maine, USA). To generate neuroblastoma PDXs, 3×3 mm3 fresh tumor fragments of COG-N-421x tumor (characterized in Bosse et al6) were engrafted subcutaneously in the flanks of 6-week-old immunodeficient female NSG mice.

In vivo vaccination/rechallenge studies

To study ADC-induced ICD in vivo, we performed different vaccination/rechallenge experiments as reviewed in Galluzzi et al.26 First, A/J mice were subcutaneously vaccinated with 2×106 NXS2-GPC2 cells previously treated with ADC at 30 ng/mL for 96 hours (n=7–8). As non-immunogenic controls, a group of seven mice were vaccinated with necrotic cells obtained on three freeze–thaw cycles. After 7 days, all animals were rechallenged subcutaneously in the contralateral flank with 2×106 naïve NXS2-GPC2 cells. To evaluate immunological memory against GPC2 negative cells, a similar study was performed, rechallenging mice with NXS2 empty-vector cells.

Second, A/J mice were vaccinated subcutaneously with NXS2-GPC2 cells treated with ADC (30 ng/mL for 96 hours) or necrotic control cells (n=8–9 per group), and animals were rechallenged intravenously with 1×106 luciferase-tagged NXS2-GPC2 cells 7 days later. Disseminated tumor growth was followed by bioluminescence imaging two times per week using a Xenogen IVIS-Spectrum system. The experimental endpoint was when the total body bioluminescence reached 1×108 total flux (p/s) or when animals died secondary to progressive disease (PD).

Third, C57BL/6 mice were vaccinated subcutaneously with 2×106 9464D-GPC2 cells treated with ADC (30 ng/mL for 96 hours) or necrotic cell controls (n=10 per group) and rechallenged subcutaneously in the contralateral flank 7 days layer with 2×106 naïve 9464D-GPC2 cells.

Evaluation of systemic CD8+ T-cell activation after vaccination

We also studied CD8 T-cell activation during vaccination experiments. First, A/J mice were vaccinated subcutaneously with ADC-treated (30 ng/mL for 96 hours) or necrotic NXS2-GPC2 cells (n=9 per group), rechallenged with naïve cells 7 days later and sacrificed at 1, 3 and 7 days post rechallenge to collect spleens (n=3 per timepoint). Splenocytes were then stained with CD8-APC and CD69-PE antibodies and analyzed by flow cytometry. Percentage of CD69+ cells gated in CD8+ populations were defined as systemically activated cytotoxic T cells.

In a second experiment, A/J mice were vaccinated as described previously (n=3 per group) and sacrificed 7 days later prior to tumor rechallenge. Spleens were harvested, processed to obtain single-cell suspensions, and CD8+ T cells were isolated using the Dynabeads FlowComp Mouse CD8 Kit (Thermo Fisher). NXS2-GPC2 cells were cultured in six-well plates overnight (50,000 cells/well) and cocultured with mouse CD8+ T cells in a T-cell:tumor ratio of 5:1 for 24 hours. All cells were collected, stained with CD8-APC and CD69-PE antibodies, and analyzed by flow cytometry.

In vivo efficacy studies

For single-agent ADC in vivo efficacy studies, A/J mice bearing 0.10–0.50 cm3 NXS2-GPC2 allografts were randomized to four treatment groups and received a single intraperitoneal dose of D3-GPC2-PBD (1 mg/kg×1), AB1-SARS-CoV-2-PBD (1 mg/kg×1), free PBD at ADC equimolar dose or vehicle (PBS×1) (n=5–10 per group). At day 7 post treatment, five animals from the ADC or vehicle groups were sacrificed to perform TIME analyses, and five animals were monitored for tumor growth and total body weights.

We also evaluated ADC efficacy when blocking either cytotoxic T cells (anti-CD8) or phagocytic macrophages (anti-CSF1R plus clodronate27). For this experiment, A/J mice bearing 0.10–0.50 cm3 NXS2-GPC2 allografts were randomized to four treatment groups: group 1, D3-GPC2-PBD (0.1 mg/kg×1, intraperitoneal); group 2, D3-GPC2-PBD (0.1 mg/kg×1, intraperitoneal, day 1) plus anti-CD8 antibody (clone 2.43; BioXCell; 400 µg per mouse, intraperitoneal; days −1, 1, and 7); group 3, D3-GPC2-PBD (0.1 mg/kg×1, intraperitoneal, day 1) plus anti-CSF1R (clone AFS98, BioXCell; 400 µg per mouse, intraperitoneal; days 1 and 7) and clodrosome (Encapsula NanoSciences; 500 µg per mouse intravenous, day −1); and group 4, vehicle (PBS×1) (n=4–8 per group).

For ADC plus CD40 agonist combination studies, A/J mice bearing NXS2-GPC2 allografts were randomized to four treatment groups (n=12–15 per group) and treated with D3-GPC2-PBD (intraperitoneal, 0.1 mg/kg×1) or vehicle. Seven days later, one dose of anti-mouse CD40 agonist antibody (100 µg, intraperitoneal; FKG4.5) or IgG2 isotype antibody (100 µg, intraperitoneal; 2A3) was administered. Animals that displayed maintained complete response (MCR) were rechallenged with NXS2 empty-vector transduced cells (2×106 in matrigel) at day 60 in the contralateral flank. On the day of rechallenge, 10 naive A/J mice were also injected with NXS2 empty-vector transduced cells as controls. Mice were monitored for an additional 40 days after tumor rechallenge. At day 4 post-IgG2 or CD40 dosing, five animals were sacrificed to perform TIME analyses.

For ADC plus CD47 antagonist combination studies, we used both syngeneic and PDX models of neuroblastoma. First, A/J mice bearing NXS2-GPC2 allografts were randomized to four groups (n=5–7 per group) and treated with D3-GPC2-PBD (intraperitoneal, 0.1 mg/kg x1) or vehicle. Anti-mouse CD47 (100 µg, intraperitoneal; MIAP-410) or IgG1 isotype (100 µg, intraperitoneal; MOPC-21) antibodies were administered at days 7, 9, and 11 to both ADC and vehicle groups. In a second study, NSG mice bearing 0.2–0.5 cm3 human COG-N-421x PDXs were randomized to four groups (n=8–9 per group) and treated with D3-GPC2-PBD (intraperitoneal, 0.25 mg/kg×1) or vehicle. Human anti-CD47 (400 µg, intraperitoneal; B6.H12) or human IgG1 isotype (400 µg, intraperitoneal; BE0297) antibodies were administered using the same schedule as in the syngeneic models.

Tumor growth responses were evaluated 7 days after the last dose of mAbs. In all combination studies, tumors that reached 2 cm3 before receiving any dose of antibody injection (day 7) were excluded from the study. Anti-CD40 and anti-CD47 antibodies used in vivo, and the corresponding IgG isotypes were obtained from BioXCell (NH, USA).

Immunohistochemistry (IHC)

Tumors and/or normal murine tissues were collected from euthanized animals, fixed with 10% formalin for a minimum of 24 hours and embedded in paraffin. Five-micrometre slides were stained on a Bond Max automated staining system using the Bond Refine polymer staining kit (Leica Biosystems) according to the standard protocol. Rabbit primary antibodies reacting against human and mouse histone H3 (phospho S10, ab5176, 1:500), Cleaved PARP1 (ab32064, 1:100) and HMGB1 (ab18256, 1:1000) were obtained from Abcam. Stained slides were then digitally scanned at ×20 magnification on an Aperio CS-O slide scanner (Leica Biosystems) and representative pictures were taken with the Aperio ImageScope program V.12.2.2.5015.

RNA sequencing

Total RNA from excised tumors was isolated using TRIzol (Life Technologies) and RNA quality was determined using the TapeStation 2200 (Agilent Technologies). Samples of optimal quality were sequenced using Chemistry Development Kit V.2, 2×150 bp, and run on one high-output flow cell of an Illumina NextSeq 500 instrument. Libraries were demultiplexed; Illumina adapters were trimmed; and FASTQ file was generated using the Illumina NextSeq Control Software V.2.02. The quality of the raw fastq data (average sequencing depth of 70 million reads) was assessed using FastQ Screen and FastQC. The files were aligned to GENCODE V.GRCm38.p4 M10 mouse reference using STAR 2-pass mode, and reads were mapped to the reference genome. Differential expression analysis was completed in R using the DESeq2 package. A gene was considered differentially expressed with an adjusted p value of <0.01 and fold change of >4 between ADC-treated and control groups. The list of differentially expressed genes was fed into PathfindR to define upregulated and downregulated pathways. RNA-sequencing data for SMS-SAN and NB-EbC1 cells were obtained from Harenza et al.28 Genes involved in phagocytosis regulation (n=4829) and with an Fragments Per Kilobase Million (FPKM) of >10 were included in the final analysis.

Flow cytometry

Single-cell suspensions from murine and human tumor cell lines, PDXs, or mouse allografts were stained with LIVE/DEAD Fixable Violet Dead Cell Stain (Invitrogen), incubated with the indicated primary antibody, and run on a Beckman CytoFLEX S cytometer (Beckman Coulter). PE mouse antihuman CD47 (B6H12, BD Bioscience) and PE rat anti-mouse CD47 (miap301, BioLegend) antibodies were used to measure CD47 surface expression in human and mouse neuroblastoma cells, respectively. Cell surface molecules of GPC2 and CD47 were determined using the BD Quantibrite Beads PE Fluorescence Quantitation Kit (BD Bioscience). Intratumoral immune infiltrates from tumor allografts were characterized using CD3-APC (100235), CD8-APC (100431), CD69-PE (104507), PD1-FITC (135213), CD11b-PerCP (101229), CD11c-APC (107339), MHCII-PE (107607), and F4-80-APC (123115) antibodies obtained from BioLegend. Results were analyzed using FlowJo software.

Cytometry by time of flight (CyTOF)

A total of 3 million cells from tumor allografts were resuspended in Maxpar Cell Staining Buffer (Fluidigm), incubated with Fc-receptor block (Miltenyi Biotec), and stained with 24 metal-conjugated antibodies. Maxpar antibodies CD45-89Y, Gr1-141, CD11c-142, CD69-143Nd, CD115-144Nd, CD4-145, F4-80-146, CD11b-148, CD19-149, CD25-151, CD3-152-Sm, CD14-156Gd, PD1-159Tb, CD40-161Dy, Tim3-162Dy, CD49b-164Dy, CD8a-168, CD206-169Tm, Nk1.1–170, CD80-171Yb, CD86-172Yb, and IA/IE-174Yb were used to detect and cluster immune cells (Fluidigm). Tumor-targeting antibodies D3-GPC2-IgG (GPC2) and dinutuximab (GD2) were conjugated with 153Eu and 154Sm using metal conjugation kits (Abcam). Cell-ID Intercalator-Ir (125 nM) was used to distinguish live versus dead cells and events were acquired using CyTOF2 mass cytometer (Fluidigm). Results were analyzed using FlowJo software.

Cytokine profiling array

Tumor protein lysates were obtained after tissue homogenization with PBS, protease inhibitors, and 1% Triton X-100 (Sigma-Aldrich). Samples were then examined for the presence of 115 mouse cytokines using a Mouse XL Cytokine Array Kit (R&D). Array membranes were scanned by conventional chemiluminescence and spot pixel density measured using HLImage++ software (Western Vision). Differential cytokine expression (fold-change cytokine expression >3, ADC vs vehicle) was determined using the mean of those cytokines with a spot pixel density higher than 500 arbitrary units.

Statistics

All statistical analyses were performed using GraphPad software except for the RNA-sequencing analyses, which were performed using R software. Student’s t-test and Mann-Whitney t-test (for non-normally distributed data) were used for non-paired comparisons of two groups. Wilcoxon signed-rank test was used to compare paired groups. One-way analyses of variance (ANOVAs) and Tukey’s, Dunn’s or Holm-Šídák’s multiple comparisons test were applied to determine differences between multiple groups with one independent variable. Two-way ANOVA and Šídák’s multiple comparison test were applied to determine differences between multiple groups with two or more independent variables. χ2 tests was used to compare proportions. Median survival rates were calculated using the Kaplan-Meier method, and curves were compared using the log-rank test. The threshold for significance (α) was set at 0.05 unless otherwise specified.

ResultsD3-GPC2-PBD ADC induces ICD in murine neuroblastoma cells

Syngeneic models of neuroblastoma resemble immunological properties of human neuroblastomas and thus are suitable for evaluating the immune modulating effects of the D3-GPC2-PBD ADC.19 30 Since murine neuroblastoma cells NXS2 and 9464D express low endogenous GPC2,7 we stably overexpressed GPC2 (NXS2-GPC2 and 9464D-GPC2) and obtained high cell surface expression by flow cytometry (figure 1A). Overexpression was maintained after NXS2-GPC2 and 9464D-GPC2 in vivo tumor engraftment in immunocompetent syngeneic A/J and C57BL/6 mice, respectively (figure 1B), with GPC2 cell surface levels comparable to those in human neuroblastoma cell lines and PDXs (figure 1C),21 recapitulating the antigen expression of human disease. We next evaluated whether NXS2-GPC2 and 9464D-GPC2 cells would be suitable for ADC functional studies. We first demonstrated that the anti-GPC2 D3-GPC2-IgG antibody conjugated to a pH-sensitive red dye showed a time-dependent internalization selectively in GPC2-transduced murine cells (figure 1D). Treatment with GPC2 ADC decreased cell viability in a concentration-dependent manner in GPC2-transduced murine neuroblastoma cells but not in those transduced with empty-vector (mean half-maximal inhibitory concentration (IC50) values of 7.12 for GPC2 vs >150 ng/mL for empty-vector transduced cells; figure 1E). This cytotoxicity was specific, as these murine neuroblastoma cell lines were not efficiently killed by a control non-targeting ADC (AB1-SARS-CoV-2-PBD) (figure 1F), and cytotoxicity of free PBD was independent of antigen expression (figure 1G). In GPC2-expressing cells, GPC2 ADC was 82.7±10.3-fold more potent than free PBD at equimolar PBD concentrations (figure 1H), correlating with the robust D3-GPC2-IgG1 antibody internalization we observed in GPC2-expressing cells (figure 1D) and suggesting increased, target-dependent intracellular payload accumulation. Further, GPC2 ADC treatment increased percentage of Annexin-V+/Dapi− cells in GPC2-expressing but not in empty-vector murine neuroblastoma cells (online supplemental figure 1). Finally, we observed that treatment with 150 ng/mL of GPC2 ADC (but not control ADC or free PBD at equimolar concentrations) increased membrane expression of calreticulin (figure 1I), HSP70 (figure 1J) and HSP90 (figure 1K), and secretion of HMGB1 (figure 1L) and ATP (figure 1M) in GPC2-expressing murine neuroblastoma cells, whereas this effect was not observed in empty-vector transduced cells. Doxorubicin given at concentrations that induced similar levels of apoptosis to 150 ng/mL of GPC2 ADC (50 and 500 nM for NXS2 and 9464D, respectively; online supplemental figure 1) increased markers of ICD independently of GPC2 expression (figure 1I–M).

Figure 1

The D3-GPC2-PBD ADC is cytotoxic and induces biomarkers of immunogenic cell death in GPC2-expressing murine neuroblastoma cells. (A) GPC2 flow cytometry histograms of GPC2 isogenic murine neuroblastoma cell lines 9464D and NXS2 versus paired empty-vector transduced cells. (B) GPC2 flow cytometry histograms of GPC2 isogenic 9464D and NXS2 allografts in C57BL/6 and A/J mice, respectively. (C) GPC2 flow cytometry histograms (left) and quantification of cell surface molecules (right) of NXS2-GPC2 allografts compared with human neuroblastoma cell lines and PDXs (one-way analysis of variance plus Holm-Šídák’s multiple comparisons test). (D) Internalization kinetics of D3-GPC2-IgG antibody in GPC2 isogenic murine neuroblastoma cell lines 9464D and NXS2 versus paired empty-vector transduced cells (left panel) and corresponding AUC plot (right panel). Means and SDs are represented (n=3). *P=0.0021, **P=0.0007 (unpaired Student t-test). (E) Concentration-dependent cytotoxicity of D3-GPC2-PBD ADC (concentration range of 0.0019–150 ng/mL) in GPC2 isogenic murine cell lines 9464D and NXS2 versus paired empty-vector transduced cells. Means and SDs are represented (n=4 from two different D3-GPC2-PBD batches). (F) Concentration-dependent cytotoxicity of AB1-SARS-CoV-2-PBD ADC (concentration range of 0.0019–150 ng/mL) in GPC2 isogenic murine neuroblastoma cell lines 9464D and NXS2 versus paired empty-vector transduced cells. Means and SDs are represented (n=2). (G) Concentration-dependent cytotoxicity of free PBD (concentration range of 9×10-6-30.62 ng/mL) in GPC2 isogenic murine cell lines 9464D and NXS2 versus paired empty-vector transduced cells. Means and SDs are represented. (H) Comparison of ADC versus free PBD cytotoxicity against GPC2 isogenic murine neuroblastoma cells at equimolar concentrations. (E–H) Means and SDs of IC50 values. (I–M) Cell surface translocation of (I) calreticulin, (J) HSP70 and (K) HSP90, and extracellular release of (L) HMGB1 and (M) ATP in GPC2 isogenic and paired empty-vector transduced 9464D and NXS2 cells after 72 hours of GPC2 or control ADC treatment (150 ng/mL), or equimolar free PBD or doxorubicin (50 nM for NXS2 and 500 nM for 9464D, respectively) treatment compared with vehicle-treated controls. Each heatmap spot represents the mean fold change from vehicle-treated cells of two to four replicates. Data in E, H and I–M represent the mean±SD of data from two to four experiments performed with different batches of the D3-GPC2-PBD ADC. ADC, antibody–drug conjugate; AUC, area under the curve; GPC2, glypican 2; IC50, half-maximal inhibitory concentration; n.s. not signficant; PBD, pyrrolobenzodiazepine; PDX, patient-derived xenograft; ROA, red objective area; uns., unstained.

Vaccination with D3-GPC2-PBD-treated cells induces endogenous T cell-dependent immune responses

We next studied GPC2 ADC-mediated ICD in vivo using vaccination/rechallenge experiments,26 summarized in figure 2A. Prophylactic subcutaneous vaccination of A/J mice with NXS2-GPC2 cells treated with ADC prevented growth of subsequent subcutaneous tumor rechallenge with naïve NXS2-GPC2 cells in 72% of mice and significantly increased survival compared with controls (figure 2B, left). Notably, such immunological memory was also observed after rechallenging with GPC2-low NXS2 empty-vector cells, where vaccination with ADC-treated GPC2-expressing cells prevented subcutaneous NXS2 empty-vector cell growth in 70% of mice, again significantly extending animal survival (figure 2B, right). No changes in GPC2 expression were observed in GPC2 or empty-vector rechallenged NXS2 tumors that relapsed after vaccination (online supplemental figure 2A). In addition, vaccination with ADC-treated cells decreased disseminated tumor growth and increased murine survival on intravenous rechallenge with naïve luciferase-tagged NXS2-GPC2 cells (figure 2C and online supplemental figure 2B,C), in part by likely preventing metastatic lesions, which were observed in the livers of mice vaccinated with control freeze–thawed NXS2 cells (figure 2D).

Figure 2

Vaccination with D3-GPC2-PBD-treated neuroblastoma cells induces T cell-dependent immunological memory. (A) Schematic representation of the different vaccination/rechallenge approaches. (B) PFS in animals subcutaneously vaccinated with ADC-treated (30 ng/mL for 96 hours) (maroon) or freeze–thawed NXS2-GPC2 cells (gray) followed by subcutaneous rechallenge with NXS2-GPC2 (left panel) or NXS2-empty-vector (right panel) naïve cells 7 days after vaccination (n=7–10 mice per group). Number of engrafted tumors in each group is indicated. *P=0.009, **P=0.042 (log-rank test). (C) PFS in animals subcutaneously vaccinated with ADC-treated (30 ng/mL for 96 hours) or freeze–thawed NXS2-GPC2 cells and intravenously rechallenged with luciferase-expressing NXS2-GPC2 cells 7 days after vaccination (n=9–10 per group). *P=0.007 (log-rank test). (D) H&E staining of a tumor-infiltrated liver from a control cell (freeze–thawed NXS2-GPC2) vaccinated mouse rechallenged intravenously with luciferase-expressing NXS2-GPC2 cells. Scale bars represent 500 µm (top) and 100 µm (bottom). (E) Schematic representation of the evaluation of systemic T-cell activation kinetics after vaccination (freeze–thaw Ctrl vs ADC (30 ng/mL), n=3 for each timepoint). (F) Time-dependent systemic activation of cytotoxic T cells (splenic CD8+/CD69+ cells) after tumor rechallenge in vaccinated mice (control vs ADC (30 ng/mL)). Means and SDs are represented. *P=0.046, **P=0.015 (unpaired Student t-test). (G) Schematic representation of the ex-vivo naïve NXS2-GPC2 cells/isolated CD8+ T-cell coculture assay from vaccinated mice (control vs ADC (30 ng/mL), n=3 each group). (H) Representative flow cytometry plots showing CD8+ and CD69+ T-cell populations in cocultures (left) from G. Quantified percentage of CD8+ and CD69+ coexpressing T cells in cocultures (right) from G. *P=0.038 (unpaired Student t-test). Means and SDs are represented. ADC, antibody–drug conjugate; Ctrl, control; GPC2, glypican 2; PBD, pyrrolobenzodiazepine; PFS, progression-free survival.

To elucidate whether these antitumor effects were regulated in part by T cells, we first determined the total amount of systemic activated CD8+ T cells at different timepoints after tumor rechallenge with naïve NXS2-GPC2 cells (figure 2E). Activated splenic T cells (coexpressing CD8 and CD69) increased at days 1 and 3 after tumor rechallenge in ADC-vaccinated mice compared with controls (figure 2F). Second, we studied whether isolated CD8+ T cells from vaccinated animals could undergo potent reactivation on ex vivo priming with naïve NXS2-GPC2 cells (figure 2G). We observed a significantly higher percentage of CD8+ and CD69+ T cells in cocultures derived from ADC-vaccinated animals compared with controls (figure 2H). Finally, we also performed a vaccination experiment using the less immunogenic 9464D-GPC2 cell line.19 Vaccination with 9464D-GPC2-ADC-treated cells neither prevented subcutaneous rechallenge nor improved survival compared with controls (online supplemental figure 2D), correlating with prior evidence of 9464D cells producing immunologically cold neuroblastoma tumors.31

D3-GPC2-PBD reprograms the neuroblastoma immune microenvironment to a proinflammatory state

We next studied whether direct GPC2 ADC treatment of NXS2-GPC2-bearing mice could result in tumor regressions and modify the neuroblastoma TIME. First, we observed that GPC2 ADC alone given at 1 mg/kg led to MCRs in 100% of NXS2-GPC2-bearing animals (figure 3A and online supplemental figure 3A), whereas mice treated with vehicle, control ADC (AB1-SARS-CoV-2-PBD) (figure 3A), or free PBD at equimolar concentrations (online supplemental figure 3A) displayed rapidly PD, confirming target-mediated ADC antitumor effects. Histologically, ADC treatment induced extensive areas of necrosis and differentiation and diminished tumor proliferation rates measured by pH3 IHC (figure 3B and online supplemental figure 3B), along with increased levels of cleaved PARP1 and HMGB1 (figure 3C and online supplemental figure 3C), confirming robust levels of apoptosis and ICD in vivo, respectively. By bulk RNA sequencing, ADC treatment resulted in differential expression of 1950 genes, most of which were upregulated (93%, 1816/1950; figure 3D). Genes related to monocyte recruitment (eg, Cxcl16, Ccl7, and Ccl2), innate immune cell activation (eg, Cd40, Cd80, and Siglec-1), or T-cell priming and activation (eg, Batf, Il2ra, and Cd69) were significantly upregulated, suggesting an ADC-mediated induction of a proinflammatory TIME. In contrast, genes related to neuroblastoma oncogenesis (eg, Foxr2 and Ascl1) were downregulated.32 33 Further, gene ontology analysis showed that 5 of the top-10 upregulated pathways after ADC treatment were directly associated with the immune system (figure 3E). Next, in paired cytokine arrays, ADC-treated tumors showed increased expression of proteins associated with myeloid cell recruitment and/or activation (CXCL16, ICAM1, CD40, and CCL12; figure 3F and online supplemental figure 3D).34–36

Figure 3

The D3-GPC2-PBD ADC reshapes the neuroblastoma immune microenvironment to a proinflammatory antitumor state. (A) NXS2-GPC2 tumor growth curves after treatment with D3-GPC2-PBD or control AB1-SARS-CoV-2-PBD ADC at 1 mg/kg, or vehicle control. Means and SEMs are represented (n=4–5 mice per group). (B) H&E and pH3 IHC staining in NXS2-GPC2 tumors 7 days after ADC (1 mg/kg) or vehicle treatments. Scale bars represent 200 µm. (C) Cleaved-PARP1 and HMGB1 IHC staining in NXS2-GPC2 tumors 7 days after ADC (1 mg/kg) or vehicle treatments. Scale bars represent 200 µm. (D) Volcano plot of differentially expressed genes (log2 fold-change >2 in ADC (1 mg/kg) versus vehicle-treated tumors shown in maroon) 7 days after ADC treatment (n=3 per group). Upregulated genes involved in innate immune cell activation as well as T-cell recruitment, and downregulated genes involved in neuroblastoma tumorigenesis are labeled and highlighted in red. (E) Top 10 upregulated Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways after ADC treatment. Immune-related pathways (n=5) are highlighted in maroon. (F) Summary of differential cytokine expression (fold-change ADC vs vehicle) in NXS2-GPC2 tumor homogenates. Black dot indicates reference spots (controls). Dotted line represents fold change of 1. (G) t-SNE plots generated from cytometry by time of flight data. Left panels show CD45+ (immune) versus GD2+/GPC2+ (tumor) populations in vehicle versus ADC-treated animals. Right panels show paired clustered immune populations gated on CD45+ cells, expanded for vehicle treated cells. (H) Intratumoral immune infiltrates in vehicle versus ADC-treated animals (n=4–5 tumors per group) quantified by flow cytometry. *P=0.007, **P=0.015 (Mann-Whitney t-test). (I) Quantification of cytotoxic T cells in an activated (CD69+) or exhausted (PD1+) state. *P=0.007 (Mann-Whitney t-test). (H,I) Individual tumor data, along with means and SDs, are represented for each immune cell subset. (J) Individual NXS2-GPC2 tumor growth curves after treatment with (1) the D3-GPC2-PBD ADC (0.1 mg/kg) alone or (2) combined with either anti-CD8 antibodies (T-cell inhibition) or (3) clodronate liposomes plus anti-CSF1R antibodies (macrophage inhibition) (n=4–8 mice per group). (K) Tumor volumes on day 10 after ADC treatment in groups indicated in J. *P=0.043, **P=0.020 (one-way analysis of variance plus Tukey’s comparisons test). ADC, antibody–drug conjugate; DC, dendritic cell; GPC2, glypican 2; MDSC, myeloid-derived suppressor cell; NK, natural killer; PBD, pyrrolobenzodiazepine; pH3, phosphohistone 3; t-SNE, t-distributed stochastic neighbor embedding.

We next studied ADC-mediated TIME modifications at a single-cell level using CyTOF to generate a comprehensive map of the ADC-induced intratumoral immune cell infiltrates (figure 3G). After 7 days of ADC treatment, we observed a significant enrichment of CD45+ cells and depletion of neuroblastoma cells in ADC-treated tumors, most notable for the robust infiltration of lymphocytes (CD4 and CD8 T, B, and natural killer (NK) cells), monocytes, proinflammatory CD40+ MHCII+ macrophages and myeloid-derived suppressor cells (online supplemental figure 3E). These data were validated with paired-flow cytometry studies where we also found a significant increase in macrophages and both total (CD3+) and cytotoxic (CD8+) T cells in ADC-treated tumors (figure 3H). Notably, cytotoxic CD8+ T cells from ADC-treated tumors were more activated (CD69+) and less exhausted (PD1+) than T cells isolated from vehicle-treated tumors (figure 3I). Finally, to further validate the functional relevance of this robust macrophage and T-cell tumor infiltration to ADC efficacy, we blocked the effects of either phagocytic macrophages (with anti-CSF1R plus clodronate27) or cytotoxic T cells (with anti-CD8 antibodies), respectively, both of which resulted in significantly impaired in vivo ADC efficacy (figure 3J,K).

D3-GPC2-PBD ADC enhances efficacy of macrophage-targeted immunoregulators

Considering the contribution of macrophage infiltration to the efficacy of this GPC2 ADC, we next asked whether lower ADC doses (0.1 mg/kg) would enhance antitumor activity of macrophage-associated immunoregulators in these syngeneic neuroblastoma models. We first selected CD40 agonism for ADC combination studies because (1) CD40 expression was upregulated after ADC treatment (figure 3D,F); (2) ADC treatment increased intratumoral infiltration of CD40-expressing immune cells (online supplemental figure 3E,F); and (3) ADC promoted T cell-mediated responses (