Irreversible Electroporation Combined with Checkpoint Blockade and TLR7 Stimulation Induces Antitumor Immunity in a Murine Pancreatic Cancer Model

Over 50,000 patients are diagnosed with pancreatic cancer each year in the United States, and over 40,000 will die from it (1). Most patients present with metastatic disease or develop distant metastatic disease despite treatment of localized disease. Palliative chemotherapy is the only approved treatment option for these patients, and response rates are low (2, 3). Immune-checkpoint inhibitors have demonstrated clinical responses in a number of solid malignancies, but results in pancreatic cancer have been disappointing (4, 5). Pancreatic cancer has only a moderate mutational burden, the factor most correlated with response to immune-checkpoint inhibitor therapy (6). In addition, the pancreatic cancer microenvironment is immunosuppressive. Its desmoplastic stroma contains few effector T cells and a surfeit of immunosuppressive leukocytes, including tumor-associated macrophages, myeloid derived suppressor cells (MDSC), and regulatory T cells (Treg; ref. 7).

Irreversible electroporation (IRE) has been developed as a nonthermal method of inducing tumor cell death without destroying adjacent collagenous structures (8). This technology, marketed as Nanoknife (Angiodynamics), has 510(k) clearance from the FDA and is being used clinically for selected patients with locally advanced pancreatic cancer (9–13). The most common pattern of recurrence in these studies was distant progression, emphasizing the need for better methods to treat micrometastatic disease.

Ablative techniques may induce antitumor immune responses by increasing the availability of tumor-specific antigens in an inflammatory context. Neoantigens released by the tumor are cross-presented by antigen-presenting cells (APC) to activate tumor-specific T-cell responses. Several studies have demonstrated that thermal ablation induces systemic antitumor immune responses in multiple tumor types (14). A limited number of studies have examined the immune effects of IRE. Li and colleagues showed that IRE results in multiple changes in peripheral cytokines and lymphocytes, including increased interferon(IFN)-γ–positive splenocytes, in a rat osteosarcoma model (15). Neal and colleagues demonstrated that IRE generates systemic immune responses in murine subcutaneous (s.c.) renal cell carcinoma models in which the growth of secondary, contralateral tumors was reduced by IRE of a primary tumor 2 weeks earlier (16). In a study by Zhao and colleagues using an orthotopic murine pancreatic cancer model, IRE alone was shown to induce transient softening of the tumor stroma, increased microvascular density, and increased permeability (17). They furthermore showed that IRE reversed resistance to anti-programmed death-1 receptor (PD-1) therapy, inducing CD8⁺ T-cell infiltration and immunogenic cell death more effectively than did the combination of irradiation and anti–PD-1 therapy. It has been theorized that the preservation of acellular collagenous structures, such as blood vessels, by IRE may promote immune cell infiltration to a greater extent than does thermal ablation, but IRE and thermal ablation have not been directly compared.

In our clinical experience, we find that the immune effects of IRE alone are not sufficient to eradicate all distant micrometastatic disease in patients (9–12, 18). The objective of this study was to combine IRE with immunotherapeutic adjuvants to augment antitumor immune responses. Antibodies against PD-1 (e.g., nivolumab and pembrolizumab) are now widely used with FDA approval for several advanced malignancies. Toll-like receptors (TLR) are expressed on APCs and play a role in innate immune responses to inflammation. TLR agonists have the potential to work synergistically with checkpoint inhibitors, which act on adaptive immune cells. This concept was supported by our study in which the effects of systemic PD-1 checkpoint inhibition were enhanced by intratumoral (IT) injections of TLR7 and/or TLR9 agonists in a murine model of head and neck cancer (18). The TLR9 agonist (SD-101, Dynavax) is being studied in several clinical trials, whereas the phospholipid-conjugated TLR7 agonist (1V270) has only been tested in preclinical models. However, the TLR7 agonist was as effective as the TLR9 agonist in this study and induced less systemic cytokine release than did the TLR9 agonist (19, 20). We hypothesize that activating the innate immune system through the TLR pathway will augment the processing of neoantigens released during IRE and that checkpoint inhibition will augment neoantigen-specific adaptive immune responses.

Using an immunocompetent mouse model of pancreatic cancer, we have demonstrated that IRE alone can serve as an “in situ vaccine” and induce systemic immune responses that prevent secondary tumor growth. We also show that combination of IRE with innate immune activators and PD-1 checkpoint blockade can have abscopal effects on established, untreated secondary tumors (therapeutic immunity). The combinatorial approach described in this study may help with overcoming the immunosuppressive pancreatic cancer microenvironment and achieving therapeutic responses in patients.

The male KPC4580P cell line was established from a spontaneous tumor that developed in a male LSL-Kras^G12D/+; LSL-Trp53^R172H/+; Pdx1^Cre/+; LSL-Rosa26^Luc/+ (KPC-luc) mouse (generously provided by J.J. Yeh, University of North Carolina, in 2015). The generation of this cell line and information regarding growth conditions have previously been described (21). Cells were cultured in DMEM-F12 (Gibco) supplemented with 10% FBS, 2 mmol/L L-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin (Gibco) under standard conditions. The cell line was negative for Mycoplasma and several mouse pathogens by Comprehensive IMPACT II testing (IDEXX Bioresearch). A tumor from the tested stock was sequenced for authentication (see Data Availability). Cells were used within 3 passages of being thawed from frozen aliquots of the tested stock for every experiment but were not reauthenticated in the past year.

All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of University of California, San Diego (UCSD). Wild-type (WT) C57BL/6 mice were purchased from The Jackson Laboratories. Rag-1 knockout mice were obtained from The Jackson Laboratories and bred by the UCSD Animal Care Program. The TLR7 agonist 1V270 was synthesized by our laboratory (22). Rat anti–mouse PD-1 (clone J43, BE0033-2), rat anti-mouse CD8 (clone YTS169.4, BE0117), and rat anti-mouse CD4 (clone GK1.5, BE0003) monoclonal antibodies were purchased from Bio X Cell.

In all IRE experiments, subcutaneous pancreatic tumors were initiated by implanting 5 × 10⁵ KPC4580P cells in the left flank of 6- to 8-week-old male C57BL/6 or Rag-1 knockout mice. Male WT mice were used for gender compatibility with the male KPC4580P cell line. IRE was performed when tumors reached 5–6 mm diameter (8–10 days after implantation) using an ECM 830 square wave pulse electroporator (Harvard apparatus) with a 2-needle probe, separated by 5 mm, to deliver a total of 150 pulses at 1500 V/cm as previously described (23). Mice were anesthetized with 2% isoflurane in 3 L/minute oxygen flow and injected subcutaneously with buprenorphine analgesic (0.1 mg/kg). The skin above the subcutaneous tumor was punctured, and the tumor was bracketed using the 2-needle probe followed by delivery of the electric pulses using a safety foot pedal control. For comparison, surgical resection was performed by excising tumors with a margin of grossly normal tissue using the same anesthetic conditions as IRE. Tumor rechallenge was performed on mice that were tumor-free after IRE or surgical resection with subcutaneous injection of 5 × 10⁵ KPC4580P cells on the contralateral (right) flank. Age-matched C57BL/6 male mice with a single tumor challenge were used as controls for all rechallenge experiments. TLR7 agonist (1V270) was administered as an IT injection of 100 μg/animal every 2 days starting the day of IRE for 3 doses. Systemic anti–PD-1 injections of 200 μg/animal were given intraperitoneally (i.p.) every 2 days starting the day prior to IRE. Vehicle and rat IgG controls were administered in respective control mice. Adoptive cell transfer was performed in Rag-1 knockout mice with T cells isolated from the spleen and lymph nodes of IRE-responsive or tumor-bearing control mice. Lymphocytes (8 × 10⁶) were isolated using density gradient centrifugation on sterile Lympholyte (Cedar Lane) with greater than 90% viability and injected subcutaneously into recipient Rag-1 knockout mice along with 5 × 10⁵ tumor cells.

Mice bearing subcutaneous tumors were euthanized 7 days after IRE, and tumors were harvested and a portion dissociated using a mouse tumor dissociation kit according to the manufacturer's recommendations (Miltenyi Biotec). The cells were then passed through a 70-μm strainer to make single-cell suspensions, and viability was measured using ViCell cell counter (Beckman Coulter). Single-cell suspensions containing 3 × 10⁶ cells/sample were stained on ice for 30 minutes in 50 μL total volume of brilliant staining buffer (BD Biosciences) using appropriate fluorescent antibody cocktails (listed in Supplementary Table S1) after Fc blocking. The samples were washed once and resuspended in 200 μL of cold PBS containing 2% FBS and 1 mmol/L EDTA for analysis using flow cytometry (BD FACSCelesta/Bio-Rad ZE5). Ultracomp compensation beads (Thermo Fisher Scientific) stained with single antibody fluorochrome combinations were used as compensation controls, and the detector voltages were set to achieve best compensation for the spectral overlap of the fluorochromes before acquisition of the study samples. Mouse splenocytes stained with the complete cocktail were used as a positive control, and fluorescence minus one (FMO) stained samples were used as negative controls for each antibody in the cocktail. Cells were fixed and permeabilized using intracellular staining reagents using the manufacturer's instructions (Intracellular Fixation and Permeabilization Buffer Set, eBioscience) for the staining of FoxP3 and IFNγ. Flow cytometry data were then analyzed using Flow Logic software (Miltenyi Biotec; ref. 24).

Male C57BL/6 mice were injected IP with anti-mouse CD8 or CD4 antibody at 3 doses of 250 μg/injection on consecutive days prior to tumor implantation (25), and subcutaneous tumors were implanted on the 5th day. Successful depletion was confirmed by flow cytometry of inguinal lymph nodes (Supplementary Fig. S1). The depletion was maintained by administering the depleting antibody intraperitoneally once a week until the end of the study. IRE was performed on these mice as described above.

Tumors were excised at days 1, 7, and 14 after IRE. Tumor tissues were snap-frozen and homogenized in gentle tissue lysis buffer (Thermo Scientific), and lysates were collected in the presence of protease and phosphatase inhibitors (Pierce Protease and Phosphatase Inhibitor Mini Tablets, Thermo Scientific). Proteome Profiler Mouse Cytokine Expression Array (R&D Systems) was used to capture intratumoral cytokines according to the manufacturer's instructions. An array intensity measurement layout was created using ImageJ, and the area of the positive control spots was defined and used to define the area of test spots. All signal spots falling outside the filter layout were considered as artifacts. Chemiluminescence intensity of the duplicate spots on the membranes was calculated using ImageJ (26). The relative fold changes in chemiluminescence against control tumors at the same time points were compared.

A portion of tumors harvested on day 7 after procedure was fixed in formalin and embedded in paraffin for IHC analyses. For IHC staining, slides containing serial tissue sections of paraffin-embedded tumor tissue blocks were deparaffinized with xylene for 3 changes of 5 minutes each, hydrated using ethanol gradient (100%, 90%, and 75%) for 2 changes of 5 minutes each, and quenched with methanol/hydrogen peroxide for 30 minutes, followed by antigen retrieval with 10 mmol/L citrate buffer (pH = 6) for 14 minutes at 20% power in a microwave oven. The slides were allowed to cool for 15 minutes and washed extensively with distilled water and with PBS for 3 changes of 5 minutes each. Tissue sections were blocked with 2% horse serum at room temperature for 30 minutes and incubated with primary antibodies (listed in Supplementary Table S2) in PBS with 2% horse serum at 4°C overnight in a humidified chamber. After washing with PBS for 3 changes of 5 minutes each, the sections were incubated with biotinylated secondary antibody for 30 minutes at room temperature followed by HRP conjugation with ABC elite universal kit (Vector Laboratories) according to the manufacturer's instructions. DAB (3,3′-diaminobenzidine) solution was used to develop the staining, and counterstained with hematoxylin. The sections were dehydrated using an increasing gradient of ethanol (75%, 90%, and 100%) and then mounted with a coverslip using clear permanent mounting media. Hematoxylin and eosin (H&E) staining was also performed for histologic analysis. The slides were imaged using an Olympus SC100 microscope at 20 × magnification.

CD8⁺ T cells were isolated from the spleen and draining lymph nodes of IRE-responsive mice subjected to tumor rechallenge for 14 days or control tumor-bearing mice for 14 days using CD8-specific magnetic cell sorting kit (Miltenyi Biotec). RNA was extracted, and cDNA-specific to T-cell receptor (TCR) gene sequences was generated using the manufacturer's recommended protocol for the Takara SMARTer Mouse TCR a/b Profiling Kit by MedGenome, Inc. Next-generation TCR sequencing and analysis using MixCR pipeline (MiLabs) was performed at MedGenome Inc. The resulting clone count data were analyzed for TCR clonal expansion and diversity indices according to methods previously described (18). Briefly, clones with under 5 counts were removed, and clonal fraction was calculated for each remaining clone and used as input into heat map and cumulative clonal fraction plots.

Untreated KPC4580P subcutaneous tumors were excised from euthanized C57BL/6 mice 14 days after implantation followed by DNA and RNA extraction using DNeasy mini kit (Qiagen) and Purelink RNA mini kit (Thermo Scientific). Whole-exome sequencing and mRNA sequencing were performed using the miSeq platform (Illumina) to identify expressed nonsynonymous genetic variants present in the tumor, cross-referenced against C57BL/6 genome (see Data Availability). Variants were prioritized according to their ability to be presented by MHC molecules using prediction algorithms. Peptides (20 amino acids in length) harboring these potential antigens at positions 6 and 15 were synthesized by A&A Labs and pooled into peptide pools of 10 each (list of KPC peptides in Supplementary Information). Bone marrow–derived dendritic cells (BMDC) were generated from male C57BL/6 mice as described earlier (27). Briefly, tibia from an age-matched male C57BL/6 mouse was excised aseptically, and the bone marrow was collected by rinsing with 10 mL of sterile PBS. Cells were resuspended at a concentration of 1 × 10⁶ cells/mL in BMDC media (RPMI-1640 10% FBS + 15 mmol/L HEPES + 50 μmol/L β-mercaptoethanol) containing 20 ng/mL of IL4 and GM-CSF, grown at standard conditions for 6 days. On day 6, the concentration of IL4 and GM-CSF was increased to 40 ng/mL, and the BMDCs in suspension were collected on day 7. The BMDCs were incubated with 5 μg/mL of mutant peptides pools to facilitate antigen presentation for 1 day at 37°C. A pre-wet multiscreen-IP filter plate (Millipore) was coated with 50 μL/well of 2 μg/mL IFNγ capture antibodies (AN18; Mabtech) diluted in PBS at 4°C overnight. The plate was washed thrice with PBS and blocked with RPMI media containing 10% FBS for 1 hours. Lymphocytes (2 × 10⁵) isolated from IRE-responsive mice were incubated with 5 μg/mL of the different mutant peptide pools in each well of the IFNγ capture filter plates for 30 minutes at 37°C, 5% CO₂. Concanavalin A (5 μg/mL; Sigma-Aldrich) was used as the positive stimulus control, and no peptide wells were used as negative control. Then 20,000 activated BMDCs generated earlier were added to the wells containing the respective peptide pools with lymphocytes, and the coculture was incubated for 20 hours at 37°C 5% CO₂. Lymphocytes from naïve (non-tumor-bearing) mice were used as negative controls along with lymphocytes from untreated tumor-bearing (tumor) mice and mice subcutaneously injected with irradiated 5 × 10⁶ KPC4580P cells followed by live-cell rechallenge (Vaccinated). The cells were discarded from the plate after incubation and washed 6 times with PBS + 0.05% Tween 20 and incubated with 100 μL/well of 1 μg/mL biotinylated anti-mouse IFNγ (R4-6A2; Mabtech) in PBS + 0.5% BSA for 2 hours at 37°C. Avidin peroxidase conjugation was performed using an APC kit (Vector Laboratories), and the plate was developed using 100 μL/well of 3-amino-9-ethylcarbazole (Sigma-Aldrich) for 10 minutes. The plate was washed, dried, and imaged using an ELISPOT reader, and the wells with >100 SFC/10⁶ cells more than the negative control were considered positive. Peptide pools that were positively recognized by T cells (expressed IFNγ) were then deconvoluted using the same procedure to identify the individual antigens recognized by T cells.

Sequencing data for mutational profiling, TCR sequencing, and RNA-sequencing are available under the NCBI BioProject link http://www.ncbi.nlm.nih.gov/bioproject/545738.

All results were expressed as means ± standard error of the mean (SEM). Statistical difference between groups was calculated using either the Student t test or ANOVA with post hoc multiple comparisons depending on the data, using GraphPad Prism 8.0 software. A value of P < 0.05 was considered statistically significant.

To assess the efficacy of IRE as a monotherapy against pancreatic cancer, immunocompetent male C57BL/6 mice were treated with IRE when subcutaneous KPC4580P tumors reached 5 mm in diameter (Fig. 1A). The most effective IRE dose and the immunogenicity of this cell line in C57BL/6 mice were determined in preliminary tumor growth and vaccination experiments, respectively (Supplementary Fig. S1). Monotherapy with IRE significantly inhibited tumor growth compared with control mice (P < 0.01; Fig. 1B). Survival was monitored until tumor diameter reached 15 mm and was significantly prolonged in the IRE group (Fig. 1C). IRE induced complete tumor regression in 3 of 9 mice at 14 days after treatment (Fig. 1D). Over repeated experiments, 20% to 35% of mice had complete responses and did not demonstrate tumor recurrence after 6 months of monitoring.

To assess the role of the immune system in the efficacy of IRE, Rag-1 knockout mice, which lack mature T and B lymphocytes, were subjected to the same IRE treatment on subcutaneous KPC4580P tumors. Tumors in both the treatment and control groups grew faster than tumors in immunocompetent mice, with no apparent difference between groups (Fig. 1E). This emphasizes the need for an active adaptive immune system for IRE to be effective. Depletion of different subsets of immune cells independently allowed us to identify CD8⁺ and CD4⁺ T cells as players in eliciting the antitumor response following IRE (Fig. 1F; Supplementary Fig. S1). CD8⁺ T-cell depletion prevented antitumor response to IRE with a growth curve similar to the control group. CD4⁺ T-cell–depleted mice did not respond to IRE, and tumors grew more rapidly than in the untreated immunocompetent mice, suggesting CD4⁺ T helper cells are the most important. The innate immune components, NK cells and macrophages, also contribute to antitumor immunity, because their depletion reduced IRE efficacy compared with IRE in an immunocompetent mouse (Supplementary Fig. S1).

Having shown that an intact immune system is necessary for IRE to be effective, we assessed the effects of IRE on the tumor microenvironment (Fig. 2). IRE-treated tumors displayed evidence of cell death consistent with the ablative effects of IRE (Fig. 2A). IHC staining with anti-CD45 showed areas of increased immune cell infiltration into the tumor after IRE. Although there was no visible increase in the infiltrating CD4⁺ T-cell population by IHC (Fig. 2A), there was an apparent increase in the infiltration of CD8⁺ T cells (Fig. 2A). By flow cytometry, there was greater variability in the immune profiles of IRE-treated than in control tumors (Fig. 2B), likely due to variability in response. Most of the differences between IRE-treated and control tumors, including CD45⁺ cells, CD8⁺ T cells, effector CD8⁺ T cells staining positively for intracellular IFNγ, and the ratio of proinflammatory M1-type macrophages (CD11b⁺F4/80⁺MHCII^hiCD206^lo) to M2-type macrophages (CD11b⁺F4/80⁺MHCII^loCD206^hi; ref. 28) did not achieve significance. The only significant difference was the immunosuppressive population of MDSCs that comprised 9.9% of total CD45⁺ cells in the control group and was reduced to 3.2% after IRE (P < 0.05).

Cytokines in the tumor microenvironment recruit or prevent the infiltration of immune cells. We examined intratumoral cytokines at 1 week in mice without complete responses to IRE and detected increases in IFNγ, I-309 (CCL1), and IL2, suggesting Th1 response (Fig. 2C). There was an immediate and sustained decrease in tumor-associated CXCL1, a driver for MDSC recruitment and CD8⁺ T-cell exclusion (29). An increase in CXCL3, a monocyte chemoattractant induced by IFNγ was also observed. The Th2 markers IL4 and IL6 were decreased after IRE, whereas an increase in IL10, a Th2 wound-healing response expected after ablation, was not statistically significant (Supplementary Fig. S2). We also examined plasma cytokines 1 week after IRE, comparing mice without complete response (nonresponders) to mice with complete response (responders), who did not have tumors for analysis of intratumoral cytokines (Supplementary Fig. S3). We observed a trend toward an increase in IFNγ after IRE, with differential effects seen between complete responders and nonresponders on IFNγ and CXCL1, but these differences did not achieve statistical significance due to the relatively small number of complete responders.

Because surgical resection is considered the optimal treatment for localized tumors, we compared the outcomes of mice harboring tumors treated with IRE versus radical resection (Fig. 3A). Surgical resection was more effective at controlling the primary tumor in this model with no evidence of tumor in 6 of 10 mice at 21 days as compared with 3 of 10 mice in the IRE group (Fig. 3B). This local recurrence rate after radical resection is a testament to the aggressive nature of these tumors. However, when tumor-free mice were rechallenged 14 days later with a subcutaneous injection of live KPC4580P cells on the contralateral flank, no tumors grew in the tumor-free mice from the IRE group. In contrast, secondary tumors grew in 3 of 5 tumor-free mice from the surgical resection group (Fig. 3C and D). Rechallenged IRE mice were monitored for 6 months and showed no evidence of recurrence. This demonstrates that successful ablation with IRE alone can induce prophylactic immunity, acting as an “in situ” vaccine against subsequent tumor rechallenge.

IRE's ability to induce protective immunity against future tumor rechallenge prompted us to study if IRE generates tumor antigen-specific T-cell responses. RNA and DNA were isolated from KPC4580P subcutaneous tumors in order to identify tumor-specific (against WT C57BL/6 background) expressed nonsynonymous variants using whole-exome sequencing and RNA-sequencing (RNA-seq). Identified variants are depicted in a Circos plot (Supplementary Fig. S4), and 44 variants were prioritized based on their high RNA expression and sequencing depth. Most of the identified variants were single-nucleotide polymorphisms resulting from the mixed background of the KPC4580P model. However, these model-inherent alloantigens were specific to the tumor, relative to the C57BL/6 host, acting as surrogates for tumor neoantigens. Peptides representing potential antigens were tested for their ability to induce IFNγ secretion in T cells isolated from naïve, tumor-bearing, vaccinated or IRE-treated and rechallenged mice. Representative ELISPOT results (Fig. 4A) show that both IRE and vaccinated T cells show reactivity against peptides from pool 2 and 6 peptides. IRE alone showed reactivity to pool 9 and 1 peptides, and vaccination group alone showed reactivity to pool 3. The reactivities were consistent only against pools 2, 6 and, 9 across three independent experiments and were the only pools deconvoluted (Fig. 4B). Although pool 9 collectively induced IFNγ secretion, none of the individual peptides had a significant signal. From pool 2, Car12 and Cdk12 peptides were recognized by T cells from IRE mice but not by T cells from tumor-bearing mice. The Hook2 peptide in pool 6 was recognized by both IRE-treated and vaccinated mice, whereas HPS1 was recognized only by the tumor-bearing and IRE-treated mice. The fact that 4 of the 5 peptides with reactivity were recognized significantly by T cells from IRE-treated mice, versus only 1 in tumor-bearing mice, emphasizes that the presence of the antigen alone is not sufficient for the recognition of most antigens. It also suggests that IRE increased recognition of weaker antigens by the immune system. The recognition patterns of T cells from IRE-treated mice were similar to, if not better than, those from mice vaccinated with irradiated tumor cells, supporting the hypothesis that IRE induces an “in situ” vaccination effect.

To further investigate the systemic immune effects of the IRE model, TCR sequencing was performed on the CD8⁺ T cells isolated from tumor-bearing control mice and IRE-responsive mice. The differences in relative abundance of the individual T-cell clones between representative individual mice are shown (Fig. 4C). None of the clones exceeded 0.2% of total TCR clonotypes in the control mouse, indicating a lack of active systemic CD8 T-cell activation. In the IRE mouse, in contrast, several clones constituted more than 0.4% of the total clones with the top clone constituting 0.8%, suggesting an enrichment of specific CD8⁺ T-cell clones. Despite the lack of a significant difference in the diversity indices, clonal expansion within the top 100 clones was more apparent in the IRE mice compared with control mice (Fig. 4D). It cannot be concluded that the enriched clones are tumor specific. However, from the combination of increased tumor mutant antigen recognition (Fig. 4A and B) with elevated T-cell clonal expansion (Fig. 4E) after IRE, it can be postulated that IRE can induce systemic T-cell activation. Adoptive transfer of these systemic post-IRE T cells into Rag-1 knockout mice, along with live KPC4580P cells s.c. prevented the growth of this allogeneic tumor in immunocompromised mice (P < 0.001) without prior exposure (Fig. 4E). In contrast, T cells from tumor-bearing donor WT mice were not able to prevent tumor growth in 4 of 5 recipient Rag-1 knockout mice. These results show that tumor alone is not sufficient to induce an effective systemic adaptive immune response. IRE not only reduces tumor burden but also helps to protect from future recurrences by inducing T-cell–mediated protective immunity.

To test the ability of IRE to induce a systemic immune response against a concomitant distant untreated tumor, C57BL/6 mice were injected with KPC4580P cells s.c. in both flanks simultaneously. The larger of the two tumors was treated with IRE 7 days after implantation. IRE failed to elicit antitumor responses against either the primary or contralateral tumor, in contrast to the robust antitumor responses observed when only a single tumor was present (Supplementary Fig. S5A and S5B). The loss of efficacy in the setting of increased secondary tumor burden demonstrates the sensitivity of IRE to immunosuppression and emphasizes the need to improve immune responses using combination immunotherapeutic strategies.

Consistent with clinical studies in humans (5, 30), checkpoint inhibition alone was not effective against subcutaneous KPC tumors (Fig. 5A). The combination of anti–PD-1 and IRE was significantly more effective than anti–PD-1 alone but not better than IRE alone in improving tumor growth or overall survival (Fig. 5B). Complete responders to either IRE or the combination of IRE and PD-1 were protected from subsequent tumor rechallenge (Fig. 5C). Tumor analysis by flow cytometry did not reveal significant increases in CD8⁺ T-cell infiltration (Fig. 5D) or M1/M2 macrophage ratio (Fig. 5E), but there was a significant reduction in MDSCs with IRE and anti–PD-1 therapy (Fig. 5F).

The other strategy we tested was to target the innate immune system using the TLR7 agonist 1V270. Similar to anti–PD-1 therapy, IT 1V270 injections showed no additive improvement over IRE alone in tumor growth reduction, survival, or protection against rechallenge (Fig. 5G–I). Although ineffective against KPC tumors as a monotherapy (Supplementary Fig. S5C), i.t. injections of 1V270 with IRE significantly improved the M1/M2 ratio (Fig. 5J) without improvements in CD8⁺ T-cell infiltrates or MDSCs over IRE alone (Fig. 5K and L). These results suggested that the incremental improvements offered by either anti–PD-1 or 1V270 alone are not sufficient to overcome the aggressive nature of these tumors, but perhaps can be combined to achieve a more robust immune response.

A second study was performed in which the secondary tumor was implanted 7 days after the primary tumor on the contralateral flank, then the primary tumor was treated 3 days later, modeling micrometastatic spread from a primary tumor. Under these conditions, primary tumor growth was significantly reduced after IRE (P < 0.05) and the combination of IRE + 1V270 + anti–PD-1 (P < 0.001; Fig. 6A and B). After IRE alone, the untreated secondary tumors grew more slowly after IRE than in the control mice (Fig. 6C and D), but only one tumor completely regressed. However, the combination of IRE + 1V270 + anti–PD-1 was more effective at preventing growth of tumors on the untreated site, with 5 of 9 mice demonstrating complete regression (P< 0.001; Fig. 6D).

Finally, we examined the immune profiles of distant tumors from the previous abscopal experiment. By IHC (Fig. 7A), immune cell infiltration could be seen approaching from the periphery toward the center in the combination group. CD8⁺ cytotoxic T-cell staining was elevated in tumors treated with the combination compared with control tumors or tumors treated with IRE alone. By flow cytometry, the greater than 4-fold increase in IFNγ-secreting CD8⁺ T cells compared with IRE alone further demonstrates the efficacy of this combination (Fig. 7B). Whereas we observed a decrease in the MDSC population in primary tumors treated with IRE (Fig. 2B), a difference in MDSCs was not observed in distant tumors responding to abscopal effects of IRE. The combination had an effect on other myeloid populations, as it increased the M1/M2 ratio without having any effect on total tumor-associated macrophages. Finally, the increase in CD8⁺ DCs, which mediate tumor antigen cross-presentation to the CD8⁺ T cells, supports the hypothesis that combination therapy of IRE with 1V270 and anti–PD-1 enhances antigen presentation.

IRE is an ablative therapy that is being used in selected patients with locally advanced pancreatic cancer who are not candidates for resection. Although experienced centers have reported local control rates greater than 90%, most patients develop distant recurrence (9–12). Using an immunocompetent pancreatic cancer mouse model, we have demonstrated that IRE requires an intact immune system for efficacy and that IRE is capable of inducing systemic adaptive immune responses and prophylactic immunity to tumor rechallenge. This study extends work on IRE by Neal et al., which showed that IRE is more effective in immunocompetent than in nude mice and that IRE induces robust local CD3⁺ T-cell infiltration (16). In our study, surgical resection was superior to IRE in achieving local tumor control, but surgical resection did not effectively induce prophylactic immunity, demonstrating that the mere presence of tumor is not sufficient to induce adaptive immune responses. IRE can act as an “in situ vaccine,” generating neoantigen-specific T-cell responses that confer protection against tumor growth by adoptive cell transfer into treatment-naïve immunocompromised mice. We have established the role of the adaptive immune system, particularly T cells, in IRE mediated antitumor activity.

The study by Zhao and colleagues modeled incomplete IRE ablation (17), with approximately 7-mm orthotopic pancreatic tumors treated with the same 2-needle probes, separated by 5 mm, used in our study for 5- to 6-mm subcutaneous tumors. They observed no significant therapeutic effect on local tumor growth after IRE alone and no complete responses. However, even incomplete IRE helped to overcome resistance to checkpoint blockade with a prolongation of survival. In our study, IRE alone had effects on local tumor growth. The addition of anti–PD-1 antibody or local innate immune stimulation with a TLR7 agonist had at least additive effects on local tumor immune infiltrates but no effects on local tumor growth over IRE alone. The combination of IRE with checkpoint inhibition and TLR7 agonist, however, not only improved the local effects of IRE but also generated therapeutic abscopal effects against small secondary tumors, modeling the potential eradication of distant micrometastatic disease.

Our study has several limitations. First, complete responses were achieved in only 20% to 35% of mice with IRE alone. This is likely attributable to the technical difficulty of achieving complete ablation in small tumors using fixed 2-needle probes. In humans, probes can be positioned individually to precisely bracket the tumor, and change in resistance can be monitored in real time to confirm successful ablation (31). Our model, however, arguably simulates the common situation in human patients in which occult local invasion results in incomplete ablation. Second, the abscopal effects observed with the combination of IRE with anti–PD-1 and TLR7 agonist therapy were modest and only evident when the secondary tumor was smaller than the primary tumor. This models the clinical scenario in which IRE is typically performed, with a locally advanced primary tumor with micrometastatic disease, but we anticipate achieving better responses with optimized combinations of immune adjuvants. Third, most of the genetic variants identified in this tumor model were single-nucleotide polymorphisms resulting from inadequate backcrossing of the genetically engineered mouse model from which the cell line was generated (21). Mutations in homologous human proteins are therefore unlikely to be identified as neoantigens in human patients with pancreatic cancer. Although most of these genetic variants did not arise from somatic mutations within the tumor, the mutational burden in this cell line is comparable with the mutational burden seen in most human pancreatic cancers. The identified model-inherent alloantigens may be useful to other researchers using KPC cell lines, and the demonstrated reactivity to these antigens demonstrates proof of concept. Finally, subcutaneous models do not fully recapitulate the pancreatic microenvironment. Ongoing studies will examine IRE and combination immunotherapy in orthotopic models that better represent the dense stroma and complex microenvironment of human tumors.

A phase II clinical trial is assessing the safety and efficacy of IRE with postoperative nivolumab in patients with locally advanced pancreatic cancer at the University of Louisville (NCT03080974). Local delivery of 1V270 or other innate immune stimulators at the time of IRE (and/or perioperatively by endoscopic ultrasound-guided injection) would be feasible and also warrants clinical investigation as a way to decrease both local and distant recurrence. “Abscopal” effects were originally described in reference to distant effects of tumor irradiation. Numerous preclinical studies have demonstrated synergistic effects of radiation with immunotherapy in other tumor types, but only a few have focused on pancreatic cancer, which is relatively radioresistant (32–35). The role of chemoradiation in pancreatic cancer has been threatened by negative phase III clinical trials (36, 37), but multiple early-phase clinical trials are evaluating the combination of irradiation with various immunotherapeutic strategies (38). Zhao and colleagues found the combination of IRE and checkpoint inhibition to be more effective than the combination of irradiation and checkpoint inhibition in their model (17). However, similar to IRE, the advanced radiation techniques in use today (such as stereotactic body radiotherapy) are difficult to model in mice (39), and direct comparison in humans will be necessary. Ultimately, a clinical strategy combining local ablation with agents that enhance both the adaptive and innate immune systems will likely be necessary to prevent recurrence in this recalcitrant disease.

J.S. Shankara Narayanan reports receiving speakers bureau honoraria from Angiodynamics. No potential conflicts of interest were disclosed by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Conception and design: J.S. Shankara Narayanan, P. Ray, T. Hayashi, D. Vicente, D.A. Carson, A.M. Miller, S.P. Schoenberger, R.R. White

Development of methodology: J.S. Shankara Narayanan, P. Ray, D.A. Carson, A.M. Miller, S.P. Schoenberger, R.R. White

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.S. Shankara Narayanan, P. Ray, R.R. White

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.S. Shankara Narayanan, P. Ray, T. Hayashi, T.C. Whisenant, D. Vicente, D.A. Carson, S.P. Schoenberger, R.R. White

Writing, review, and/or revision of the manuscript: J.S. Shankara Narayanan, P. Ray, T. Hayashi, T.C. Whisenant, D. Vicente, A.M. Miller, R.R. White

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.S. Shankara Narayanan, T. Hayashi, D. Vicente

Study supervision: D.A. Carson, R.R. White

This work was supported by the Padres Pedal the Cause (#PTC2017). Bioinformatic analyses were partially supported by a voucher from the Altman Clinical and Translational Research Institute at UCSD, which is funded by the NIH (CTSAUL1TR001442). Immunohistochemistry was performed by the UCSD Moores Cancer Center Biorepository and Tissue Technology Shared Resource, which is funded by the NCI (NCI P30CA23100).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.