Serial Thermal Ablation Induces Abscopal Antitumor Immunity and Reveals Targetable CSF1R-Dependent Resistance in Pancreatic Cancer

Abstract

Thermal ablation is increasingly used for local control of pancreatic ductal adenocarcinoma (PDAC), but its capacity to induce systemic antitumor immunity and the mechanisms limiting this response remain incompletely defined. Using a bilateral LSL-Kras^G12D/+; LSL-Trp53^R172H/+; Pdx1-Cre (KPC) flank tumor model, we show that serial radiofrequency ablation (RFA) enhances local tumor control and induces a robust abscopal response. This effect was associated with increased activation of CD8⁺ T cells and natural killer cells, and was abrogated by CD8⁺ T cell depletion. Single-cell RNA sequencing revealed expansion of cytotoxic immune programs alongside induction of a CSF1-driven myeloid response consistent with adaptive immune resistance. Although CSF1R inhibition alone was insufficient to improve tumor control, combinatorial blockade of PD-L1 and CD73 augmented systemic antitumor responses, and the addition of CSF1R inhibition in this context further enhanced both local and distant tumor control. These findings identify a CSF1-dependent myeloid resistance program that constrains ablation-induced systemic immunity and demonstrate that rational combination immunotherapy can potentiate the systemic efficacy of tumor ablation in PDAC.

INTRODUCTION

Radiofrequency ablation (RFA) is an established locoregional therapy that induces tumor destruction through thermal coagulative necrosis and is widely used for the treatment of solid malignancies, including hepatic, renal, and pancreatic tumors^1–9. In the treatment of pancreatic tumors, endoscopic ultrasound–guided radiofrequency ablation (EUS-RFA) has emerged as a minimally invasive approach that delivers targeted thermal energy to pancreatic lesions while minimizing injury to surrounding structures. Early clinical studies have demonstrated the feasibility and safety of EUS-RFA for pancreatic tumors, including pancreatic ductal adenocarcinoma (PDAC) and pancreatic neuroendocrine tumors, suggesting that this strategy may provide local tumor control in patients who are not candidates for surgical resection^10–12. Beyond its direct cytotoxic effects, accumulating evidence indicates that thermal ablation can also modulate the tumor immune microenvironment (TIM)^13–15.

Thermal injury induced by RFA leads to rapid tumor cell death accompanied by the release of tumor-associated antigens, danger-associated molecular patterns, and inflammatory cytokines that can promote immune activation. Preclinical studies have demonstrated that RFA can enhance antigen presentation, stimulate dendritic cell maturation, and promote infiltration of cytotoxic lymphocytes within the tumor microenvironment (TME)^16,17. In some settings, these immune responses extend beyond the site of ablation, generating systemic antitumor immunity and tumor regression at distant untreated sites resembling an abscopal effect^14,18,19. However, the magnitude and durability of these immune responses are often limited, particularly in solid tumors such as PDAC that are characterized by profound immune suppression and resistance to immunotherapy^20,21.

Importantly, most preclinical studies investigating the immunologic consequences of thermal ablation have focused on single-treatment paradigms, even though patients frequently undergo multiple ablation procedures due to incomplete tumor destruction, recurrence, or staged treatment strategies^22,23. Serial tumor injury may amplify inflammatory signaling and antigen release, potentially enhancing immune priming and immune cell recruitment within the TME. At the same time, tissue injury can promote hypoxia, alter the stiffness and subtypes of cancer-associated fibroblasts (CAFs), and enhance the recruitment of immunoregulatory myeloid populations, including tolerogenic or antitumor neutrophils and macrophages, which shape the balance between antitumor immunity and immune suppression^14,24–26. Among these pathways, the CSF1/CSF1R signaling axis plays a central role in regulating macrophage recruitment, differentiation, and polarization within the TME and has been implicated in resistance to immunotherapy across multiple solid tumors, including pancreatic cancer^27–29.

Given the highly immune suppressive nature of PDAC, strategies capable of converting the TME into an immunologically active state may be required to sensitize tumors to immunotherapy³⁰. EUS-guided locoregional therapies such as RFA may provide such an opportunity by inducing immunogenic tumor cell death and promoting immune infiltration. However, the extent to which serial thermal ablation reshapes the tumor immune landscape and generates systemic antitumor immunity remains poorly understood. Moreover, how ablation-induced immune remodeling interacts with key immune suppressive pathways, including macrophage signaling and immune checkpoint regulation, has not been fully defined.

In this study, we investigated how serial thermal ablation using RFA influences tumor progression and the TIM using a bilateral PDAC tumor model that enables simultaneous evaluation of local and systemic immune responses. Through histologic, single-cell transcriptomic, and immunologic analyses, we examined how sequential ablation events remodel immune cell populations and inflammatory signaling networks within the RFA-treated and contralateral tumors. We show serial thermal ablation using RFA in combination with targeting PD-L1, CD73 and CSF1R significantly restrains the growth of treated and contralateral tumors. These data are encouraging given the challenges of generating durable responses to immunotherapy in solid tumors including PDAC.

RESULTS

Serial RFA Remodels the Tumor Immune Microenvironment in Treated and Contralateral Tumors

To investigate the effect of serial RFA sessions, we utilized a bilateral syngeneic tumor bearing KPC model (Fig. 1A). We previously demonstrated significant reductions in the size of RFA-treated tumors at Day 4 post a single ablation session¹⁴ and we began here by comparing the growth rate after 1 RFA session compared to 3 serial RFA treatments at Day 4 after each final ablation (acute response). Serial thermal ablation using RFA every four days significantly suppressed tumor growth compared to both untreated controls and tumors receiving a single ablation (Fig. 1B). Mice treated with three RFA sessions exhibited markedly reduced and comparable growth rates in both the directly treated tumors and the untreated contralateral tumors (Fig. 1B).

Figure 1. Serial thermal radiofrequency ablation significantly alters the immune microenvironment compared to 1 RFA.

A) Experimental Design to evaluate the effect of serial thermal ablation using RFA on treated and contralateral tumors. Figure created using BioRender. B) Three RFA treatments significantly reduces the growth rate of treated and contralateral tumors compared to control and 1 RFA-treated tumors, (***p=0.0007), (****p<0.0001). n=18 control untreated, n=17 1 RFA, n=20 3 RFA. C) Representative 10x hematoxylin and eosin images, necrosis outlined in dashed line and representative 20x IHC images of GR1, iNOS and CD163. D) Necrosis is significantly increased in 3 RFA RFA-treated tumors compared to 1 RFA-treated tumors and increased in 3 RFA contralateral tumors compared to 1 RFA contralateral tumors (****p<0.0001). n=64 fields analyzed for 1 RFA tumors and n=88 fields analyzed for 3 RFA tumors. E) GR1 is significantly increased in 3 RFA RFA-treated tumors compared to 1 RFA-treated tumors (*p=0.0484). GR1 is significantly increased in 3 RFA contralateral tumors compared to 1 RFA contralateral tumors (****p<0.0001). GR1 is also increased in 3 RFA contralateral tumors compared to 3 RFA RFA-treated tumors (****p<0.0001). n=87 fields analyzed for 1 RFA tumors and n=109 fields analyzed for 3 RFA tumors. F) iNOS is significantly increased in 3 RFA RFA-treated tumors compared to 1 RFA-treated tumors (***p=0.0002). iNOS is significantly increased in 3 RFA contralateral tumors compared to 1 RFA contralateral tumors (****p<0.0001). iNOS is also increased in 3 RFA contralateral tumors compared to 3 RFA RFA-treated tumors (*p=0.0202). n=90 fields analyzed for 1 RFA tumors and n=82 fields analyzed for 3 RFA tumors. G) CD163 is significantly increased in 3 RFA contralateral tumors compared to 1 RFA contralateral tumors (*p=0.0155). n=86 fields analyzed for 1 RFA tumors and n=111 fields analyzed for 3 RFA tumors. For all statistical comparisons a One-way ANOVA was used. Scale bars 100 μm for all images.

Histopathologic evaluation of tumors using hematoxylin and eosin (H&E) staining revealed substantial structural differences between treatment groups (Fig. 1C). Representative images demonstrated larger areas of necrosis (dashed black lines) in tumors receiving serial ablation (Fig. 1D). Notably, significantly increased necrosis was also observed in contralateral tumors from the 3 RFA group compared with contralateral tumors from the 1 RFA group indicating a more pronounced abscopal response after serial thermal ablation. Additionally, apoptosis (as measured by staining for cleaved caspase 3 (CC3)) was significantly increased in 3 RFA-treated tumors compared to 3 RFA contralateral or 1 RFA-treated tumors (Supplemental Fig. S1A).

Recent studies in preclinical models have demonstrated neutrophils to be a dominant innate immune cell infiltrating RFA-treated tumors^13,14. To further characterize innate immune infiltration, IHC was performed for neutrophils using NIMPR14 (GR1) and myeloperoxidase (MPO) (Fig. 1C). GR1⁺ cells were significantly increased in tumors treated with serial ablation (Fig. 1C and 1E). Contralateral tumors in the 3 RFA treatment group also displayed a marked increase in GR1⁺ cells compared to contralateral tumors in the 1 RFA group. Interestingly, GR1 expression was significantly higher in 3 RFA contralateral tumors than in the corresponding RFA-treated tumors and MPO staining was also significantly increased in 3 RFA-treated contralateral tumors (Supplemental Fig. S1; Fig. 1E).

IHC was also performed to characterize infiltrating macrophage populations using CD68, inducible nitric oxide synthase (iNOS), and CD163 (Supplemental Fig. S1D, Fig. 1C). We also observed a significant reduction in CD68⁺ cells in the 3 RFA-treated tumors compared to 1 RFA-treated and contralateral tumors from both groups (Supplemental Fig. S1D). Analysis of iNOS expression revealed enhanced inflammatory macrophage activity following serial ablation in 3 RFA-treated tumors compared with 1 RFA-treated tumors (Fig. 1F). Contralateral tumors from the 3 RFA treatment group also showed significantly elevated iNOS expression relative to contralateral tumors from the 1 RFA treatment group. In addition, iNOS levels were modestly but significantly higher in contralateral tumors compared with the corresponding RFA-treated tumors in the 3 RFA-treated group (Fig. 1F). Finally, CD163 expression, a marker associated with immune suppressive macrophages, was significantly increased in contralateral tumors from the 3 RFA treatment group compared to contralateral tumors from the 1 RFA treatment group, whereas no significant differences were detected between the directly treated tumors (Fig. 1G). Together, these findings demonstrate that serial RFA not only increases tumor necrosis at the treatment site but also significantly alters the immune landscape both locally and systemically.

Single-Cell RNA Sequencing Reveals Serial RFA-Dependent Tumor Cell States and Enhanced Cytotoxic Immune Signatures

To define how serial thermal ablation reshapes the cellular composition of the tumor microenvironment, we performed single-cell RNA sequencing (scRNA-seq) on tumors following either 1 or 3 RFA treatments. Dimensionality reduction using uniform manifold approximation and projection (UMAP) revealed substantial cellular heterogeneity across the tumors (Fig. 2A). In total, 19 transcriptionally distinct clusters were identified and annotated into 11 major cell types based on canonical marker gene expression (Supplemental Fig. S2). These included six tumor cell clusters (clusters 0, 3, 4, 5, 9, and 10), two macrophage clusters, two neutrophil clusters, two erythrocyte clusters, and single clusters corresponding to endothelial cells, fibroblasts, B cells, classical dendritic cells, mast cells, natural killer (NK) cells, and T cells.

Figure 2. scRNA-seq defines serial RFA-dependent tumor clusters.

A) uMAPs of identified cell clusters and cell types from single-cell RNA-sequencing analysis. There are 19 cell clusters and 11 cell types (6 tumor clusters, 2 macrophage clusters, 2 neutrophil clusters, 2 erythrocyte clusters, and an endothelial, fibroblast, classical dendritic cell, natural killer cell, T cell, B cell, and mast cell cluster. B) Heat map showing top differentially expressed genes in the 6 tumor clusters (0, 3, 4, 5, 9, and 10). C) Violin plots showing tumor cluster gene expression related to a classical, exocrine, or quasimesenchymal signature and nFeature RNA plot. D) Violin plot showing increased Gzmb mRNA levels in T cell and natural killer cell clusters following 3 RFA sessions. E) Violin plot showing increased Il18 mRNA levels in tumor, macrophage, neutrophil, T cell, and classical dendritic cell clusters following 3 RFA sessions. F-G) Ingenuity Pathway Analysis of T cells and NK cells

The tumor clusters displayed heterogeneous phenotypes, characterized with distinct gene expression programs (Fig. 2B). Cluster 0, which was decreased in 3 RFA-treated tumors compared to 1 RFA treatment, showed high expression of mitochondrial genes with gene expression characteristic of elevated energy metabolism. Cluster 9 displayed strong proliferative activity with high expression of cell cycle genes and was also decreased in 3 RFA-treated compared to 1 RFA-treated tumors. Cluster 3 was enriched in vascular endothelial growth factor a (VEGFa) and was increased in 3 RFA compared to 1 RFA-treated tumors. We assessed the expression of gene signatures associated with Classical, Exocrine, and Quasimesenchymal tumor phenotypes³¹. Across all tumor clusters, Quasimesenchymal, Exocrine and Classical signatures were identified; however, the Classical signature genes were most highly expressed after serial RFA treatments (Fig. 2C).

Given the experimental evidence of a more pronounced abscopal effect with serial RFA treatments, we next examined cytotoxic immune signatures within lymphocyte populations in RFA ablated tumors. Analysis of cytotoxic effector genes revealed increased expression of Granzyme B (Gzmb) in both T cell and NK cell clusters following 3 RFA treatments compared to a single ablation (Fig. 2D), suggesting enhanced cytotoxic immune activation after serial RFA therapy. In addition to cytotoxic mediators, we assessed inflammatory cytokine expression across multiple cell types. Violin plot analysis revealed elevated Il18 mRNA expression following serial RFA treatments across several cell populations, including tumor cells, macrophages, neutrophils, T cells, and classical dendritic cells (Fig. 2E). This broad increase in Il18 expression suggests that serial RFA promotes a pro-inflammatory signaling environment and NK cell and T cell cytotoxic phenotypes across both immune and tumor compartments. Ingenuity Pathway Analysis (IPA) of single-cell transcriptional profiles revealed enrichment of immune activation programs (orange bars) within T- and NK-cell populations in tumors treated with three rounds of RFA compared to a single RFA treatment (Fig. 2F–G). These pathways included signatures associated with cytotoxic lymphocyte activation and inflammatory cytokine signaling, consistent with enhanced recruitment and activation of effector immune cells following serial tumor injury. In contrast, for both T cells and NK cells, PD-1/PD-L1 cancer immunotherapy pathway was significantly decreased in the 3 RFA-treated compared to 1 RFA-treated tumors (blue bars).

Serial RFA induced dynamic immune remodeling characterized by early CD8⁺ T cell–dependent tumor control.

To further characterize the TIM, we performed IHC staining for NK cells using NK1.1. Representative images demonstrated increased NK1.1⁺ cell infiltration following serial RFA (Fig. 3A). Quantification revealed that NK1.1⁺ cells were significantly increased in contralateral tumors from mice receiving 3 RFA treatments compared to contralateral tumors from mice receiving a single RFA treatment (Fig. 3B). Furthermore, NK cell infiltration was significantly higher in the contralateral tumors from the 3 RFA-treated group compared with their RFA-treated counterparts.

Figure 3. CD8⁺ T cells mediate early but not sustained tumor control following serial RFA.

A) Representative 20x NK1.1 immunohistochemistry staining images. Scale bars 100 μm. B) NK1.1 staining is significantly increased in 3 RFA contralateral tumors compared to 1 RFA contralateral tumors (***p<0.0001). NK1.1 is significantly increased in 3 RFA-treated tumors compared to 3 RFA contralateral tumors (*p=0.0141). n=10 fields analyzed per tumor. C) Representative 20x immunofluorescence images of CD4 (purple), CD8α (red), GZMB (green), DAPI nuclear stain (blue). D) CD4⁺GZMB⁺ cells are not increased in 3 RFA contralateral tumors. E) CD8α⁺GZMB⁺ cells are significantly increased in the three RFA contralateral tumors compared to 1 RFA contralateral (****p<0.0001) and 3 RFA RFA-treated tumors (**p=0.0046). n=8 fields analyzed per group. F) Schematic of the protocol to evaluate the effects of CD8 depletion on serial RFA-treated and contralateral tumors. Created using BioRender. G) CD8α depletion significantly reduces the %CD8α⁺ cells per live cells in RFA-treated tumors. H) α-CD8α significantly increases the tumor volume of RFA-treated (*p<0.05; ***p<0.001) and I) contralateral tumors after 2 RFA treatments but does not significantly increase the final tumor volumes after the third RFA treatment. Statistics were done using Prism GraphPad software.

To further characterize cytotoxic lymphocyte activity, multiplex immunofluorescence (IF) staining was performed for CD4, CD8α, and GZMB (Fig. 3C). While not significant, quantitative analysis revealed the presence of CD4⁺GZMB⁺ cells in contralateral, but not RFA-treated tumors following serial RFA treatment (Fig. 3D), indicating that CD4⁺ T cells did not substantially contribute to the increased cytotoxic signature observed in the scRNA-seq data (Fig. 2F). In contrast, CD8α⁺GZMB⁺ cytotoxic T cells were significantly enriched in contralateral tumors from the 3 RFA group compared with both contralateral tumors from the 1 RFA-treated group and RFA-treated tumors from the 3 RFA-treated group (Fig. 3C–E; Supplemental Fig. S3). Collectively, these findings demonstrate that serial thermal ablation using RFA promotes a more immunologically active TME in both local and in distant tumors.

To evaluate the functional role of CD8⁺ T cells, we performed a depletion experiment (Fig. 3F). Anti-CD8α efficiently reduced systemic CD8⁺ T cell frequencies in whole blood (Fig. 3G; Supplemental Fig. S4). Anti-CD8α T cell ablation significantly accelerated tumor growth in both treated and contralateral lesions following two RFA treatments but did not affect final tumor volumes after a third RFA (Fig. 3H–I), indicating that CD8⁺ T cells mediate acute tumor control, but the effect is not as sustained after the third ablation.

Serial RFA thermal ablation significantly enriches CSF1/CSF1R myeloid pathway signaling within treated tumors.

Proteomic profiling of tumor lysates further demonstrated that serial ablation induces a robust inflammatory response. Tumors treated with three RFA sessions exhibited significant upregulation of multiple cytokines (Fig. 4A), including G-CSF, GM-CSF, IFN-γ, LIF, CSF1, MDGF, and TNF-α. IFN-γ and TNF-α are key cytokines that promote cytotoxic T cell activity through upregulation of ICAM-1 and enhanced antigen presentation to promote T cell inflammatory cell death and killing efficacy as well as promote the recruitment CD8⁺ T cells and NK cells^32–35. However, many of the significantly increased cytokines are also key regulators of myeloid cell recruitment and differentiation during inflammatory tissue repair, including CSF1, LIF and GM-CSF, suggesting that serial thermal injury also initiates a wound-healing like program within the TME.

Figure 4. Serial thermal ablation using RFA significantly enriches for CSF1R/CSF pathway.

A) Proteome profiler comparing 1 RFA and 3 RFA-treated tumors shows discovery comparisons using a multiple unpaired t-test. G-CSF, GM-CSF, IFN-γ, LIF, M-CSF, MDGF and TNF-^α were all significantly increased immune cytokines. B) Violin plot of Csf1 gene expression mRNA levels. Csf1 is significantly increased in neutrophil clusters, T cells and B cells after 3 RFA treatments. C) Violin plot of Csf1r gene expression. Csf1r mRNA expression is significantly increased in the mitochondrial-low macrophage cluster in three RFA-treated tumors. D) Proteome profiler comparing 1 RFA and 3RFA-treated tumors shows discovery comparisons using a multiple unpaired t-test. Interleukin Family proteins significantly increased in the proteome profiler include IL-1a, IL-1ra, IL-2, IL-3, IL-4, IL-5, IL-7, IL-11, IL-13, IL-15, IL-17a, IL-22, IL-23, IL-27p28, IL-28A/B, IL-33. E) Proteome profiler comparing 1 RFA and 3RFA-treated tumors shows discovery comparisons using a multiple unpaired t-test. M1-like tumor associated macrophage derived cytokine IL-12p40 is significantly increased in 3 RFA-treated tumors, while there is no discovery for IL-1b, IL-6, CXCL9, and CXCL10. F) Proteome profiler comparing 1 RFA and 3 RFA-treated tumors shows discovery comparisons using a multiple unpaired t-test. M2-like tumor associated macrophage derived cytokines CCL17 and CCL22 are significantly increased in 3 RFA-treated tumors, while there is no discovery for IL-10. A student’s t-test or (E) or multiple unpaired t-test was used for statistical comparisons.

Consistent with this injury-driven response, in addition to increased CSF1 protein expression identified by the proteomic assay, scRNA-seq analysis demonstrated significant induction of Csf1 expression across multiple immune populations in RFA-treated tumors (Fig. 4B). Elevated Csf1 transcripts were detected in neutrophils as well as T and B lymphocytes in tumors treated with three ablation sessions, indicating that both innate and adaptive immune compartments contribute to CSF1 production in response to tumor injury. In parallel, expression of Csf1r was significantly increased within a macrophage subset characterized by low mitochondrial gene expression (Fig. 4C), consistent with expansion or activation of CSF1R⁺ macrophages. Activation of the CSF1/CSF1R signaling axis is a central component of wound-healing programs and promotes the recruitment, survival, and polarization of immune suppressive macrophages³⁶. In tumor contexts, CSF1R⁺ myeloid cells are known to suppress cytotoxic lymphocyte function. Thus, while serial RFA promotes infiltration and activation of cytotoxic NK and T cells, it simultaneously induces a countervailing CSF1-driven myeloid response that may restrain lymphocyte-mediated tumor killing.

In addition to immune remodeling, serial ablation induced stromal and extracellular matrix changes consistent with tissue repair. Proteome profiler analysis identified increased levels of proteins associated with extracellular matrix remodeling and inflammatory tissue repair (Supplemental Fig. S5), including Tissue Factor (CD142), Dickkopf-related protein 1 (DKK-1), and Osteoprotegerin (OPG). Multiplex cytokine analysis further revealed increased levels of several interleukin family cytokines following three RFA sessions (Fig. 4D; Supplemental Fig. S5), including IL-1α, IL-1RA, IL-3, IL-4, IL-5, IL-12p40, IL-13, IL-15, IL-22, IL-23, IL-27p28, and IFN-γ. Collectively, these data demonstrate that serial RFA reshapes the TME through an injury-induced inflammatory program characterized by increased cytotoxic lymphocyte infiltration together with a concurrent activation of CSF1/CSF1R-dependent myeloid pathways. Multiplex cytokine analysis further revealed increased levels of several interleukin family cytokines following three RFA sessions (Fig. 4D; Supplemental Fig. S5), including IL-1α, IL-1RA, IL-3, IL-4, IL-5, IL-13, IL-15, IL-22, IL-23, IL-27p28, and IFN-γ. Additionally, multiplex cytokine analysis revealed serial ablation did not increase levels of IL-1b, IL-6, CXCL9, CXCl10, and only modestly increased levels of IL-12–40, all of which have been shown to be commonly secreted by M1-like tumor associated macrophages (Fig. 4E)^37,38,39. Alternatively, CCL17 and CCL22, which are typically secreted by M2-like tumor associated macrophages^37,39 was dramatically increased in 3 RFA-treated tumors, while IL-10 levels were unchanged (Fig. 4F). Collectively, these data demonstrate that serial RFA reshapes the TME through an injury-induced inflammatory program characterized by increased cytotoxic lymphocyte infiltration together with a concurrent activation of CSF1/CSF1R-dependent myeloid pathways.

Pharmacologic inhibition of CSF1R does not enhance the antitumor effects of serial RFA

Given the observed activation of the CSF1/CSF1R signaling axis following serial RFA treatment, we next investigated whether pharmacologic inhibition of CSF1R could enhance the therapeutic efficacy of serial thermal ablations. For this experiment, we first established the growth kinetics after a single RFA ablation compared to 3 RFA treatments. Mice with bilateral tumors were treated with either one RFA session, three RFA sessions, three sham treatments, or left untreated (Fig. 5A). A single RFA treatment increased the size of tumors compared to controls; however, serial RFA-treated tumors were significantly decreased compared to 3 sham-treated or 1 RFA-treated tumors (Fig. 5B–C). Serial thermal ablation using RFA significantly reduced growth of the contralateral untreated tumors compared to untreated controls, 3 sham-treated tumors, and contralateral tumors from the single RFA-treatment group (Fig. 5C). These data may be more revealing of the systemic antitumor immune effects due to the absence of ablation-induced injury.

Figure 5. Inhibition of CSF1R does not enhance the efficacy of serial thermal ablation.

A) Experimental design created using BioRender. B) Endpoint tumor volume is significantly reduced in 3 RFA-treated tumors compared to 1 RFA-treated tumors (p=0.0015, **). Endpoint tumor volume is significantly reduced in 3 RFA contralateral tumors compared to untreated tumors (p=0.0487, *) and 1 RFA contralateral tumors (p=0.0032, **). n=12 for untreated tumors, n=10 for 1 RFA, and n=23 for 3 RFA. C) On the treated tumor, serial RFA sessions significantly reduce tumor volume at day 25 compared to control untreated tumors (**p=0.0052), 3 sham treatments (****p<0.0001), and to 1 RFA session (****p<0.0001). Control untreated tumors are significantly smaller than 1 RFA tumors (p=0.0115, *). D) On the contralateral tumor, serial RFA sessions significantly reduce tumor volume at day 25 compared to control untreated tumors (p=0.0001, ***), 3 sham treatments (p<0.0001, ****), and to 1 RFA session (p<0.0001, ****). For C and D, n=8 for untreated tumors, n=19 for 1 RFA, n=23 for 3 RFA, and n=10 for 3 Sham. E) Representative 10x hematoxylin and eosin images, necrosis outlined in dashed line. F) Necrosis is significantly decreased in 1 RFA contralateral tumors compared to 1 RFA-treated tumors (**p=0.0073). Necrosis is significantly increased in 3 RFA-treated and contralateral tumors compared to 1 RFA contralateral tumors (***p=0.0002; **p=0.0075). Necrosis is significantly increased in 3 RFA-treated tumors compared to 3 sham contralateral tumors (*p=0.0118). G) Representative 20x CSF1R immunohistochemistry images. H) CSF1R staining is significantly increased in 3 RFA-treated and contralateral tumors compared to control, sham and 1 RFA-treated tumors (****p<0.0001). CSF1R staining is significantly increased in 1 RFA contralateral tumors compared to control untreated (*p=0.0473) and 3 sham-treated tumors (*p=0.0288). I) Experimental design using PLX3397 (CSF1R inhibitor), created using BioRender. J) CSF1R inhibition does not reduce final tumor volume. For B and C, two-way ANOVA was used for statistical analysis, for all other panels a one-way ANOVA was used for statistical analysis. Scale bars 100 μm.

Histopathologic evaluation revealed increased necrotic regions following serial RFA (Fig. 5E). Quantitative analysis showed that a single RFA treatment induced a comparable degree of necrosis in the treated tumor compared with 3 RFA treatments. However, the contralateral tumor necrosis was significantly higher in the 3 RFA treatment group compared to the 1 RFA treatment group (Fig. 5E–F).

Immunohistochemical staining for CSF1R revealed significantly increased expression in 3 RFA-ablated tumors and contralateral tumors as well as single RFA-treated tumors compared to untreated tumors and tumors receiving three sham treatments (Fig. 5G–H). We next tested whether pharmacologic inhibition of CSF1R could enhance RFA-mediated tumor control. Mice were treated with the CSF1R inhibitor PLX3397 in combination with serial RFA treatment (Fig. 5I). However, adding CSF1R inhibition did not significantly reduce final tumor volume compared to serial RFA treatment alone (Fig. 5J), indicating that blockade of CSF1R signaling was insufficient to further enhance the antitumor effects of serial thermal ablation in this model. Together, these findings demonstrate that although serial RFA activates the CSF1/CSF1R myeloid signaling axis, pharmacologic inhibition of this pathway does not significantly improve tumor control when combined with serial thermal ablation.

Enhanced systemic tumor control following serial thermal ablation combined with CSF1R inhibition, PD-L1 blockade, and CD73 targeting

Preclinical studies have evaluated whether anti-PD-1/PD-L1 or targeting CD73 with Quemliclustat (Quemli) can enhance the tumor growth restraining capability of a single RFA ablation^13–15. We previously found the combination of Quemli with a single RFA ablation significantly increased cytotoxic CD8⁺ T cell recruitment into ablated tumors¹⁵. We also analyzed immune cell populations from our published tumors treated with RFA and anti-PD-L1¹⁴. These data show a significant increase in GZMB⁺, CD8⁺ T cells, NK cells and CD68⁺ cells with combination treatment compared to RFA treatment alone, indicating enhanced local and abscopal tumor control (Supplemental Fig. S6). We next investigated whether serial thermal ablation could be leveraged to enhance systemic antitumor immunity when combined with immunomodulatory therapies. Mice bearing bilateral tumors were treated with serial RFA of the primary tumor in combination with anti-PD-L1 and Quemli (Fig. 6A). Tumor growth was monitored in both the ablated tumors and the untreated contralateral tumors.

Figure 6. Serial thermal ablation in combination with CSF1R inhibition, anti-PD-L1 and Quemliclustat significantly reduces the volume of ablated and contralateral tumors.

A) Experimental design setup. B) Serial thermal ablation using RFA in combination with anti-PD-L1 and Quemli does not significantly reduce the volume of treated tumors. C) However, serial RFA in combination with anti-PD-L1 and Quemli significantly reduces the tumor volume of contralateral tumors compared to serial RFA + IgG+Vehicle (*p<0.05) and compared to serial RFA + Quemli treated alone (****p<0.0001). D) Representative 20x CSF1R immunohistochemistry images. E) Quantification of CSF1R⁺ cells showed an increase in CSF1R⁺ area per field in serial RFA tumors + anti-PD-L1 with or without Quemli compared to serial RFA + vehicle in both RFA-treated and F) contralateral tumors. A two-way Anova was used in Prism GraphPad for statistical comparisons. G) Experimental design setup. H) Serial thermal ablation using RFA in combination with CSF1R inhibition, anti-PD-L1 and Quemli significantly reduces the volume of ablated tumors (**p<0.01; ***p<0.001) and I) contralateral tumors (**p<0.01; ***p<0.001). J) Combination treatment using 3 RFA + anti-PD-L1, Quemli and CSF1R inhibition significantly increases the infiltration of CD8⁺ T cells in treated (**p<0.01; *p<0.05) and K) contralateral tumors (**p<0.01; *p<0.05). L) Combination treatment using 3 RFA + anti-PD-L1, Quemli and CSF1R inhibition significantly increases GZMB staining in treated (***p<0.001; **p<0.01) and M) contralateral tumors (*p<0.05; **p<0.01; ***p<0.001). A two-way ANOVA in Prism GraphPad was used for statistical analysis. Scale bars 50 μm.

Serial RFA combined with anti-PD-L1 and Quemli did not reduce the volume of the ablated tumors compared with control conditions (Fig. 6B); however, this therapeutic combination did produce a pronounced effect on distal tumor growth. Specifically, mice receiving serial RFA treatments together with anti-PD-L1 and Quemli exhibited a significant reduction in the growth of contralateral tumors relative to animals treated with serial RFA plus IgG and vehicle, serial RFA and anti-PD-L1, as well as those treated with serial RFA and Quemli alone (Fig. 6C), but did not significantly decrease the growth of contralateral tumors in the serial RFA treated mice with anti-PD-L1.

We next examined the abundance of CSF1R⁺ cells within the TME of mice treated with serial RFA in combination with anti-PD-L1 and Quemli to determine if CSF1R⁺ cells increase in response to immune checkpoint blockade and serial RFA. Quantification of CSF1R⁺ staining revealed significantly increased CSF1R⁺ area per field in tumors treated with serial RFA in combination with anti-PD-L1, compared to tumors treated with 3 RFA and Quemli (Fig. 6D–F), indicating enhanced infiltration or expansion of CSF1R⁺ myeloid cells following these treatments. A more pronounced significant increase in CSF1R⁺ cell area was detected in the contralateral tumors where the 3 RFA-treated with Quemli and anti-PD-L1 had significantly more CSF1r+ cells per tumor volume that all other groups (Fig. 6F), suggesting that systemic immune modulation induced by serial RFA treatment plus combination immunotherapy is associated with recruitment or accumulation of these myeloid populations at distant tumor sites.

To determine whether targeting this myeloid compartment could further enhance therapeutic efficacy, we performed an additional experiment incorporating pharmacologic inhibition of CSF1R alongside anti-PD-L1 and Quemli during serial RFA treatment (Fig. 6G). The addition of CSF1R inhibition in the treatment regimen resulted in a pronounced reduction in tumor growth at both the ablated and distant tumor sites compared to mice that received serial RFA with Quemli and anti-PD-L1 (Fig. 6H); however, in the RFA-treated tumors this did not result in a significant decrease compared to 3 RFA + vehicle or 3 RFA + CSF1Ri alone. In the contralateral tumors, serial RFA treatment in combination with Quemli, anti-PD-L1 and CSF1Ri significantly reduced the size of tumors compared to 3 RFA + vehicle and 3 RFA + Quemli + anti-PD-L1 (Fig. 6I). Immune profiling of tumors from these mice revealed a significant increase in CD8⁺ and GZMB⁺ cells in ablated and contralateral tumors from mice that received 3 RFA treatments plus anti-PD-L1, Quemli and CSF1R inhibition compared to 3 RFA + vehicle, 3 RFA + CSF1Ri and 3 RFA + anti-PD-L1 and Quemli (Fig. 6J–M). Together, these results demonstrate that while serial thermal ablation combined with PD-L1 blockade and CD73 inhibition can promote systemic tumor control, the addition of CSF1R inhibition further improves therapeutic efficacy with a correlative increase in CD8⁺ and GZMB⁺ cells in both ablated and contralateral tumors.

CD73 expression is correlated with GZMB staining in short versus long term survivors treated with EUS-RFA prior to surgical resection.

The studies in Fig. 6E–F revealed Quemli treatment alone with serial RFA did not increase intratumoral CSF1R levels; however anti-PD-L1 treatment did significantly increase CSF1R levels. We have previously shown inhibition of CD73 with Quemli in established PDAC models and in the setting of immunoprevention significantly alters macrophage polarization to restrain suppressive macrophage populations^40,41. To study the effects of serial thermal ablation on CD73, CD39 and GZMB expression, we stained four resected samples from patients that received EUS-RFA and systemic chemotherapy prior to surgery. Long-term survivors had distinct patterns of staining consistent with minimal CD73, CD39 and increased GZMB staining (Fig. 7A) compared to short-term survivors that had strong staining for CD73 or CD39 and minimal GZMB (Fig. 7B).

Figure 7. Long-term survivors after EUS-RFA show minimal CD73 expression and increased GZMB⁺ cells.

A) Long-term PDAC survivors (greater than 4 years) treated with at least 3 EUS-RFA procedures prior to surgery have minimal CD73 and CD39 expression by IHC and increased GZMB staining compared to B) Short-term survivors (less than 13 months survival) who have high CD73, CD39 and minimal GZMB expression. Scale bars 50uM.

Serial thermal ablation plus CSF1R inhibition, anti-PD-L1, and Quemliclustat yields the greatest reduction in ablated and contralateral tumor volumes.

Serial RFA significantly reduced tumor burden compared to a single ablation, and this effect was further enhanced by combination therapies (Fig. 8). In ablated tumors, the addition of CSF1R inhibition, anti-PD-L1, and Quemli to serial RFA produced the greatest reduction in endpoint tumor volume, outperforming all other treatment groups, including RFA alone and partial combinations. Notably, serial RFA treatment with CSF1R inhibition also significantly decreased tumor size relative to multiple control conditions (Fig. 8A). In contralateral tumors, serial RFA treatment alone was sufficient to reduce tumor volume compared to sham and single RFA, but the most pronounced systemic effect was observed with the full combination therapy (RFA+αPD-L1+Quemli+CSF1Ri), which significantly suppressed tumor growth relative to all controls and other treatment arms (Fig. 8B). Collectively, these data demonstrate that serial thermal ablation synergizes with immune modulation and CD73 inhibition to maximize both local and systemic antitumor responses.

Figure 8. Serial thermal ablation in combination with CSF1R inhibition, anti-PD-L1 and Quemliclustat significantly reduces the volume of ablated and contralateral tumors more than any other treatment.

A) Endpoint tumor volume of RFA-treated tumors. 3 RFA+αPD-L1+Quemli+CSF1Ri tumors are significantly smaller than 3 Sham treatments (*p=0.0359), 1 RFA (***p=0.0001), 3 RFA+IgG+Vehicle (**p=0.0076), 3 RFA+IgG+Quemli (**p=0.0060), and 3 RFA+αPD-L1+Quemli (*p=0.0142). 3 RFA+αPD-L1+Quemli+CSF1Ri vehicle tumors are significantly smaller than 1 RFA tumors (**p=0.0025). 3 RFA+CSF1Ri tumors are significantly smaller than 1 RFA (***p=0.0004), 3 RFA+IgG+Vehicle (*p=0.0177), 3 RFA+IgG+Quemli (*p=0.0148), and 3 RFA+αPD-L1+Quemli (*p=0.0335). 3 RFA tumors are significantly smaller than 1 RFA tumors (**p=0.0043). 3 RFA+CSF1Ri Vehicle tumors are significantly smaller than 1 RFA tumors (*p=0.0189). B) 3 RFA tumors are significantly smaller than 3 Sham treatments (*p=0.0106), 1 RFA (***p=0.0002,), 3 RFA+IgG+Vehicle (***p=0.0003), and 3 RFA+IgG+Quemli (****p<0.0001). 3 RFA+CSF1Ri Vehicle tumors are significantly smaller than 3 Sham treatments (*p=0.0231), 1 RFA (***p=0.0007), 3 RFA+IgG+Vehicle (**p=0.0012), and 3 RFA+IgG+Quemli (****p<0.0001). 3 RFA+CSF1Ri tumors are significantly smaller than 3 Sham treatments (p=0.0113, *), 1 RFA (***p=0.0003), 3 RFA+IgG+Vehicle (***p=0.0005), and 3 RFA+IgG+Quemli (****p<0.0001). 3 RFA+αPD-L1+Quemli+CSF1Ri Vehicle tumors are significantly smaller than 3 Sham treatments (*p=0.016), 1 RFA (***p=0.0005), 3 RFA+IgG+Vehicle (***p=0.0007), and 3 RFA+IgG+Quemli (****p<0.0001). 3 RFA+αPD-L1+Quemli+CSF1Ri tumors are significantly smaller than control untreated tumors (*p=0.0346), 3 Sham treatments (***p=0.001), 1 RFA (****p<0.0001), 3 RFA+IgG+Vehicle (****p<0.0001), 3 RFA+IgG+Quemli (****p<0.0001), 3 RFA+ αPD-L1+Vehicle (*p=0.0252), and 3 RFA+ αPD-L1+Quemli (*p=0.0131).

DISCUSSION

Targeted thermal ablation is increasingly recognized not only as a local cytoreductive therapy but also as a stimulus for systemic immune activation within the tumor microenvironment. In this study, we demonstrate that serial RFA treatment produces substantially greater antitumor effects than a single ablation, enhancing necrosis, cytotoxic immune responses, and growth inhibition in both treated and contralateral tumors. Serial ablation reshapes the immune landscape, including expansion of neutrophils and dynamic alterations in macrophage populations, highlighting the role of innate immunity in mediating both local and systemic responses. Single-cell transcriptomic analysis further identifies a CSF1-driven myeloid program that may limit the durability of these responses, revealing a targetable mechanism of adaptive resistance.

Emerging clinical data reinforce the translational relevance of immunomodulatory combinations in PDAC. The phase 1b ARC-8 trial evaluated Quemliclustat combined with anti-PD-1 therapy and chemotherapy in patients with advanced PDAC. Post-treatment analyses linked maximal suppression of adenosine-associated transcripts, including the NR4A gene family, with enhanced T-cell activation and survival benefit⁴². Similarly, a phase 2 study combining focal radiation with dual checkpoint blockade in microsatellite-stable colorectal and pancreatic cancer demonstrated that local irradiation can sensitize tumors otherwise resistant to immunotherapy, producing disease control in a subset of patients⁴³. Retrospective and prospective analyses of stereotactic body radiotherapy, irreversible electroporation, RFA and other targeted ablative modalities combined with systemic therapy across solid tumors have further suggested that timing, sequencing, and frequency of local interventions critically influence immune priming and clinical outcomes^44,45. These clinical observations support the concept that immune responses to tumor-directed injury are dynamic and potentially programmable.

Our preclinical data provide mechanistic insight into these clinical findings. Serial RFA amplifies systemic cytotoxic immunity but also induces compensatory expansion of CCL3⁺ neutrophils, recently defined as protumor⁴⁶ and CSF1R⁺ regulatory myeloid populations. These results suggest that the interval between ablation events and the timing of combination immunotherapy will determine the balance between immune activation and adaptive suppression. Optimizing the temporal relationship between tumor injury and systemic therapy could enhance the durability of antitumor responses and improve clinical outcomes.

In addition to innate immune remodeling, serial RFA induces a broad inflammatory transcriptional program characterized by increased Il18 expression across multiple cell types. IL-18 is a cytokine that promotes natural killer cell activation, enhances T-cell–mediated immunity, and can augment immunotherapy responses, indicating that serial ablation generates a cytokine environment favorable for sustained cytotoxic activity. Pathway analysis revealed activation of T-cell receptor, NK cell, and Th1-associated signaling within infiltrating lymphocytes following serial RFA, accompanied by reduced PD-1/PD-L1 pathway activity. These findings highlight how iterative tumor injury can prime the immune system while also eliciting regulatory mechanisms that may constrain the response.

Finally, evaluation of patient samples treated with serial EUS-RFA in combination with systemic chemotherapy revealed minimal expression of CD73 and CD39 and increased GZMB expression in long-term survivors compared with short-term survivors. These observations align with the clinical rationale demonstrated in the ARC-8 trial, suggesting that interventions targeting immune suppressive pathways such as CD73 inhibition could further enhance ablation-induced antitumor immunity.

Together, these findings define targetable resistance mechanisms and provide a rationale for combinatorial strategies integrating serial tumor ablation with systemic immunotherapy. Rational scheduling of ablation events, consideration of immunosuppressive feedback, and integration of checkpoint or adenosine-targeted therapies represent actionable strategies to convert immunologically cold tumors into responsive disease. Future studies should prospectively evaluate how timing, dose, and sequence of ablation influence immune priming, as well as the functional contributions of neutrophils and CSF1R-positive myeloid cells, to fully translate these insights into effective clinical interventions in pancreatic cancer.

METHODS

Preclinical KPC Subcutaneous Tumor Model

All mouse model procedures followed UNMC’s IACUC protocol. 500,000 KPC cells were prepared in a PBS:Matrigel mix (1:1) and injected subcutaneously in both flanks of 8-week-old male C57BL/6 mice. Sex as a biological variable: male mice were used for these experiments as the KPC line was derived from male mice and the tumor growth rates are more variable in female mice. Tumor size was measured and recorded twice per week with a vernier caliper. Tumor volume was calculated as ((length × width × width)/2) in mm³. KPC cells were derived in the Tuveson Lab from Kras^LSL-G12D/+;Trp53^LSL-R172H/+;Pdx1-Cre mice, which develop PDAC.

Growth Rate Calculations

Growth rate was calculated by dividing the measurements from each day with the measurements from the previous day. Statistical analysis was performed using a one-way ANOVA or student’s t test in Prism GraphPad software.

Radiofrequency Ablation

RFA was performed 14 days after implantation when the tumors were between 200–500 mm³. Mice were anesthetized with isoflurane via a nose cone. Ablation was performed on the flank with the larger tumor. The skin around the ablation site was cleaned with 3 alternating rounds of 70% ethanol and iodine. At the ablation site, a 2.5 mg/kg dose of bupivacaine (NDC 0409-1162-18, Hospira Inc.) and 5 mg/kg dose of meloxicam (NDC 13965-559-20, Vet One) was administered through subcutaneous injection prior to ablation. At the ablation site, a small incision was made in the center of the tumor, and the Habib EUS RFA probe was inserted. Ablation was performed for 10 seconds and the average power delivered was 0.5 W (watts). Mice were observed for signs of pain post-ablation. The sham control mice were transferred to the procedure room, placed under anesthesia, received bupivacaine and meloxicam, and an incision was made, and the probe was inserted, but no ablation was performed. This protocol was repeated for the sham control mice and 3 RFA mice for the second RFA session 18 days after tumor implantation and the third RFA session 21 days after tumor implantation.

Hematoxylin and Eosin Staining and Quantification

Hematoxylin and eosin (H&E) staining was performed as previously described⁵. Necrosis quantification was scored by a board-certified pathologist. Statistical analysis was performed using a student’s t test in Prism GraphPad software.

Immunohistochemistry and Quantification

Immunohistochemistry (IHC) staining and quantification was performed as previously described⁵. Statistical analysis was performed using a one-way ANOVA in Prism GraphPad software. The following primary antibodies were used: NIMPR14/GR1 (1:100, ab2557, abcam), iNOS (1:100, ab15323, abcam), CD163 (1:500, ab182422, abcam), Csf1r (1:150, ab254357, abcam), NK1.1 (1:50, E6Y9G, Cell Signaling Technology). The following secondary antibodies were used: anti-rat (1:500, BA-9400–1.5, Vector Laboratories), anti-rabbit (1:500, BA-1000–1.5, Vector Laboratories)

Immunofluorescence

Slides were baked at 60°C for 30 minutes, then deparaffinized in histoclear for 9 minutes. Rehydration was performed by 2-minute washes in graded alcohols at 100%, 95%, then 70%. Slides were placed in 1x PBS for 6 minutes, then in PBST for 15 minutes, then 1x PBS for 10 minutes. Antigen retrieval was performed based on antibody recommendations. Slides were cooled completely before 10 minutes in 1x PBS. Slides were blocked in 10% FBS in 1x PBST for 60 minutes before primary antibody incubation overnight in a humidified chamber at 4°C. On the second day, slides were washed 3 times with PBST for 10 minutes each. The secondary antibody was diluted in blocking solution and incubated for 90 minutes, with the antibodies protected from light. Quenching (103710–194, Novus Biologicals) was performed for 5 minutes. Hoechst staining (MSPP-62249, Invitrogen) was performed for 5 minutes before slides were mounted with VECTASHIELD vibrance antifade mounting media (H-1000–10, Vector Laboratories) and imaged. The following primary antibodies were used: CD4 (1:35, ab288724, abcam), CD8α (1:50, MAB116–100, R&D Systems), GZMB (1:100, AF1865, R&D Systems). The following secondary antibodies were used: FITC (1:1000, A16000, ThermoFisher), Cy5 (1:500, A10525, ThermoFisher), Texas Red (1:350, T-2767, ThermoFisher). Statistical analysis was performed using a one-way ANOVA in Prism GraphPad software.

Single cell RNA-seq and Analysis

For the 1 RFA group, tissue from two mice was pooled into a single sample. For the 3 RFA group, tissue from four mice was pooled into two samples (two mice per sample). Consequently, a total of three processed samples were subjected to paired scRNA-seq and scTCR-seq. Processing was performed as previously described¹⁴. Following tissue processing, the two tumors from each respective group were combined into one sample per group and submitted to the Cancer Genomics Center at UTHealth. The raw data was processed from paired scRNA library and single-cell TCR library Cell Ranger (v7.0.0) multi pipeline. The RNA library was aligned to mm10–2020-A reference genome, and the TCR libraries were aligned to GRCm38-alts-ensembl-7.0.0 provided by 10× Genomics. Due to low numbers of T cells, TCR output was not included in the analysis. Subsequently, detailed QC metrics were computed and assessed using Seurat package (v5.1.0)⁴⁷. Cells with less than 200 genes, or more than 30% mitochondrial gene counts were filtered out to ensure data quality. The filtered data were then normalized by library size and log₂-transformed. Inferred doublets were then removed using scDBlFinder (v.1.8.0)⁴⁸. Subsequently, 2,000 highly variable genes were selected for dimension reduction and clustering analysis. The FindNeighbors function was used with the first 30 PCs to build the 17 nearest neighbor graph and the FindClusters function was used with a 0.4 resolution to identify cell clusters. The UMAP method was employed for dimensionality reduction and 2D visualization of the cell clusters. Differential expression analysis was performed using the FindMarkers function with Wilcoxon ranked-sum test. Cell types were annotated by examining the expression of canonical cell markers in cell clusters. The difference in cell proportion between two conditions of a cell type was tested using Fisher exact test. P-values were adjusted using the Benjamini-Hochberg procedure.

Cytokine Array Proteome Profiler

The cytokine array proteome profiler was performed as previously described¹⁴. Statistical analysis was performed using multiple unpaired t-tests in Prism GraphPad software.

PLX3397 Administration

PLX3397 (HY-16749, MedChem Express) was dissolved in 10% DMSO (D8418, Sigma-Aldrich) and 90% 20% SBE-β-CD (HY-17031/CS-0731, MedChem Express) in saline. PLX3397 was administered through oral gavage at a 50 mg/kg dose daily beginning 13 days after tumor implantation on the day before the first RFA treatment until the experiment endpoint. The vehicle control was 10% DMSO in 20% SBE-β-CD in saline.

Anti-PD-L1 Administration

Anti-PD-L1 antibody (B7-H1, BioXCell) was diluted in InVivo Pure pH 7.0 dilution buffer (IP0070, BioXCell). Anti-PD-L1 was administered through intraperitoneal injections at a 200 μg dose beginning 14 days after tumor implantation, on the day of the first RFA treatment and was given every other day until the experiment endpoint. The anti-IgG2b isotype control (BE0090, BioXCell) was diluted in InVivo Pure pH 7.0 dilution buffer (IP0070, BioXCell) and administered at the same dose and schedule as anti-PD-L1.

Quemliclustat Administration

Quemliclustat (HY-125286, MedChemExpress) was dissolved in 20% DMSO (D8418, Sigma-Aldrich) and 90% 20% SBE-β-CD (HY-17031/CS-0731, MedChem Express) in saline. Quemliclustat was administered through oral gavage at a 10 mg/kg dose beginning 14 days after tumor implantation, on the day of the first RFA treatment and was given every other day until the experiment endpoint. The vehicle control was 10% DMSO in 20% SBE-β-CD in saline.

Anti-CD8α Administration

Anti-CD8α antibody (BP0061, BioXCell) was diluted in InVivo Pure pH 7.0 dilution buffer (IP0070, BioXCell). Anti- CD8α was administered through intraperitoneal injections at a 200 μg dose beginning 13 days after tumor implantation, on the day before the first RFA treatment and was given every other day until the experiment endpoint. The anti-IgG2b isotype control (BP0090, BioXCell) was diluted in InVivo Pure pH 7.0 dilution buffer (IP0070, BioXCell) and administered at the same dose and schedule as anti-CD8α.

Flow Cytometry Validation of CD8α Depletion

Whole blood was collected at the time of euthanasia and centrifuged for 5 minutes at 500 g. Serum was collected. 1 mL of 1x PBS was added and samples were centrifuged for 5 minutes at 500 g. Remaining serum was collected. Samples were incubated in 1 mL RBC lysis buffer (420301, BioLegend) for 10 minutes, then 1 mL 1x PBS was added and samples were centrifuged for 5 minutes at 500 g. Samples were incubated in Live/Dead Fix Blue (Thermo Fisher Scientific, L34962) for 15 minutes then washed with 1 mL 1x PBS and centrifuged for 5 minutes at 500 g. Fc receptor block (BD Biosciences, 553142) was added for 5 minutes on ice, then anti-CD8α antibody (100714, BioLegend) was incubated for 15 minutes at room temperature. Samples were washed with 1x PBS and centrifuged for 5 minutes at 500 g. Samples were resuspended in .5% BSA/PBS. Data was acquired using a BD LSRFortessa and analyzed with FlowJo v10.10.0.

Data Availability Statement

All the raw and processed data supporting the findings of this study are available through GEO, accession number GSE325036. The R scripts supporting the findings of this paper are available upon request.

Clinical Samples

This study represents an institutional review of a prospective longitudinal cohort involving EUS (protocol HSC-MS-18–0192). All procedures involving human participants were conducted in compliance with the ethical standards set by the UTHealth Houston Committee for the Protection of Human Subjects and adhered to the principles outlined in the 1961 Declaration of Helsinki. Eligible participants were required to have histologically confirmed PDAC. Based on physician’s choice, patients received chemotherapy consisting of modified FOLFIRINOX, gemcitabine and nab-paclitaxel with or without cisplatin. In a subset of patients, EUS-RFA was performed after one month of chemotherapy. Under EUS guidance, an electrode needle probe was advanced into the pancreatic ductal adenocarcinoma via either a transduodenal or transgastric approach. Prophylactic antibiotics and nonsteroidal anti-inflammatory drugs were not routinely administered. Color Doppler imaging was used to minimize the risk of vascular injury. Ablation was continued until electrical impedance reached 200 Ohms or demonstrated a rapid increase. Treatment effect was confirmed by real-time sonographic changes in tumor appearance, including alterations in echogenicity and evidence of tissue liquefaction. Patients typically continued with RFA sessions until the interventional gastroenterologist evaluated that pancreatic tumor could no longer be amenable to further RFA. Clinical, demographic, and perioperative data if patient underwent surgical resection of PDAC tumor were systematically collected. Radiologic response was assessed by comparing pre- and post-RFA axial computed tomography or magnetic resonance imaging using RECIST 1.1 criteria. Additionally, tumor characteristics were evaluated through changes in Hounsfield units before and after treatment. Size of the pancreatic tumor was also measured and recorded for each RFA session.