Radiofrequency Ablation Remodels the Tumor Microenvironment and Promotes Neutrophil-Mediated Abscopal Immunomodulation in Pancreatic Cancer

Current therapeutic approaches, including chemotherapy and radiation alone, or combined with immune checkpoint blockade (ICB) have shown little success in pancreatic ductal adenocarcinoma (PDAC), placing an imperative need for the discovery and application of innovative techniques (1, 2). Endoscopic ultrasound-guided radiofrequency ablation (EUS-RFA) is an FDA-approved technique currently available for the treatment of several gastrointestinal malignancies including pancreatic cancer and has the advantage of being minimally invasive and safe (3). Studies in both animal models and patients with cancer show that RFA induces not only local burning/disruption of the tumor by heat but also generates coagulative necrosis, which releases large amounts of cellular debris that represent a source of tumor antigens that can ultimately trigger a host adaptive immune response against local and distant tumors, an effect referred to as the abscopal effect (4, 5).

All animal procedures were performed in compliance with UTHealth's Center for Laboratory Animal Medicine and Care Animal Welfare Committee Review and approved on Dr. Bailey's Animal Welfare Committee protocol. Three independent experiments were performed. 100,000 KPC cells in PBS:Matrigel (Corning, cat. CB-40230) mix (1:1) were injected in each of the left and right flanks of female C57BL/6mice (The Jackson Laboratory, cat. 000664). KPC cells were originally derived in the Tuveson Lab from Kras^LSL-G12D/+;Trp53^LSL-R172H/+;Pdx1-Cre mice, which develop PDAC and obtained from Chin-Yo Lin from the University of Houston in 2020. Cells were maintained at low passage and have not been reauthenticated. Cell culture was performed in DMEM media (Gibco, cat. 10566–016), with 10% FBS (Corning, cat. 35–010-CV) and 1% penicillin streptomycin (Gibco, cat. 15140–122) and trypsinized with Trypsin 0.25% in HBSS (Gen Clone, cat. 25–510). Cells were passaged approximately 3 times before use. No Mycoplasma test was performed immediately before the experiment and no cell line manipulations were performed for this study. Tumor size was assessed twice per week with a Vernier caliper and calculated as (length × width × width)/2 in mm³. For the RFA-treated cohort, ablated tumors were analyzed by hematoxylin and eosin (H&E) composite staining and the presence of a necrotic core, as indicative of ablation success, determined inclusion of the mice for further studies. Non-RFA–treated tumors with positive abscopal effect were included for further analysis.

RFA treatment was initiated when tumors reached 200 to 500 mm³ using a previously described protocol (6). The non-RFA side contralateral tumor did not receive RFA treatment. Mice were observed for signs of pain or discomfort post ablation.

Dry tissue slides from control, RFA and non-RFA–treated tumor tissues from tumor-bearing mice and resected samples from PDAC patients were soaked in Histoclear (National Diagnostics, cat. HS-200) for 8 minutes and gradually rehydrated in decreasing concentrations of alcohol (Fisher Bioreagents, cat. BP28184) 100% and 95% for 4 minutes each. Slides were then rinsed in the following solutions for 3 minutes each: tap water, Millipore Sigma 65067–75 HARLECO Hematoxylin (cat. EM-65067–75), tap water, Epredia Signature Series Clarifier (Thermo Scientific, cat. 22–050–117), tap water, Scott's Bluing Reagent (Polysciences, cat. 24605–1) and tap water. A counterstain was performed by 95% EtOH for 1 minute followed by Eosin-Y stain (Richard-Allan Scientific, catalog no. 71311) and a final dehydration step was performed in increasing concentrations of alcohol 95% for 1 minute, followed by 100% for 4 minutes and Histoclear for 4 minutes. Slides were mounted in Cytoseal XYL (Epredia, cat. 8312–4) mounting media.

Dry tissue slides from control, RFA- and non-RFA–treated tumor tissues from tumor-bearing mice and resected samples from patients with PDAC, when applicable, were processed and stained for IHC as previously described (6) using antibodies specific for cleaved caspase 3 (Cell Signaling Technology, catalog no. 9664), Granzyme B (Abcam, catalog no. ab255598), NIMPR14 (Abcam, catalog no. ab2557), myeloperoxidase (MPO; Abcam, catalog no. ab208670), αSMA (Abcam, catalog no. ab5694), CD31 (Abcam, catalog no. ab281583) and CD8α (Cell Signaling Technology, catalog no. 98941T). For immunofluorescence staining, the IHC protocol was followed without antigen retrieval, ABC reagent, DAB reagent, or counterstaining steps. Primary antibodies specific for MPO (Abcam, catalog no. ab208670) and NIMPR14 (Abcam, catalog no. ab2557) were used followed by anti-rat (Jackson immunology, catalog no. AB2340683) and anti-rabbit (Invitrogen, catalog no. 31635) secondary antibodies. For collagen quantification in pancreatic sections, Masson's trichrome staining was performed by using Abcam Trichrome Stain (Abcam, catalog no. ab150686) kit where the sections were deparaffinized with Histoclear and rehydrated as mentioned above. Bouin's Fluid was preheated to 60°C, and sections were incubated for 60 minutes and then stained with Weigert's Iron Hematoxylin for 5 minutes, Biebrich Scarlet for 15 minutes, Phosphomolybdic Acid for 12 minutes, Aniline Blue Solution for 20 minutes and Acetic Acid Solution for 5 minutes. Sections were dehydrated with Histoclear and mounted in Cytoseal XYL (Epredia, catalog no. 8312–4) mounting media. For all experiments described in this report, 15 to 20 fields per group were quantified. Image analysis was performed using ImageJ (http://imagej.nih.gov/ij/) software [Version ImageJ 2.9.0, Java 1.8.0_172 (64-bit)].

Human serum samples and post resection images had been deidentified to all research personnel and were used in compliance with UT Health human subject's research under IRB EUS-RFA Prospective study HSC-MS-18–0192 (PI: N.C. Thosani). Written informed consent was obtained from all patients enrolled in the study. Serum was collected from 1 patient with Stage III PDAC post–EUS-RFA and stored at 4°C until further processing. Two Stage I and one Stage III resected tumor samples from patients with PDAC post–EUS-RFA were fixed in formalin, processed, embedded, and sectioned by the Memorial Hermann TMC Pathology Core. This study was conducted in accordance with UTHealth IRB protocols abided by Helsinki Declaration.

Metal-labeled antibodies were prepared according to the MaxPar antibody conjugation kit protocol (Fluidigm, catalog no. 201300). In brief, antibodies were obtained in carrier/protein-free buffer and then prepared according to the kit instructions. After determining the percent yield by absorbance measurement at 280 nm, the metal-labeled antibodies were diluted in Candor PBS Antibody Stabilization solution (Candor Bioscience, catalog no. 131050) for long-term storage at 4°C. Tumor sections were baked at 60°C overnight, then dehydrated in xylene and rehydrated in a graded series of alcohol for imaging mass cytometry (IMC). Heat-induced epitope retrieval was conducted in a water bath at 95°C in Tris buffer with Tween20 at pH 9 for 20 minutes. After immediate cooling for 20 minutes, the sections were blocked with 3% BSA in TBS for 1 hour. For staining, the sections were incubated overnight at 4°C with an antibody master mix. Samples were then washed 4 times with TBS/0.1% Tween20. For nuclear staining, the sections were stained with Fluidigm Corp Cell-ID Intercalator (Fluidigm Corp, catalog no. NC1605684) for 5 minutes and washed twice with TBS/0.1% Tween20. Slides were air-dried and stored at 4°C for ablation.

The sections were ablated with Fluidigm Hyperion Imaging System (IMC, Fluidigm) for data acquisition. IMC data were segmented by ilastik (www.ilastik.org) and CellProfiler (http://www.cellprofiler.org). Histology topography cytometry analysis toolbox (HistoCAT, https://github.com/BodenmillerGroup/histoCAT) and R (Version 4.2.1, www.r-project.org) scripts developed in house were used to quantify cell number, generate tSNE plots, and perform neighborhood analysis. For all samples, tumor and cellular densities were averaged across 3 images per tumor, with n = 3 per group. Antibodies used in this study are listed in Supplementary Table S1 (7).

The Proteome Profiler Mouse Cytokine Array Panel A (ARY006, R&D Systems, Minneapolis, MN) was used to measure the protein levels of cytokines and chemokines in serum and tumor tissues. For serum measurements, 50 μL were used. In tumor tissues, protein concentrations were quantitated using Pierce BCA Protein Assay Kit (Thermo Scientific, catalog no. 23225) following manufacturer instructions. The array was inoculated with 200 μg of protein, and samples were treated according to the product specification. Similarly, pre- and post-RFA human serum samples were used for analysis with the Proteome Profiler Human Cytokine Array Kit (ARY005B, R&D Systems, Minneapolis, MN), which detects human cytokines, chemokines, and acute phase proteins. Assays of human samples were run with 50 μL of serum according to the manufacturer's instructions. Proteins were detected with ChemiDoc MP Imaging System (Bio-Rad, catalog no. 12003154) and analyzed in Image Lab Software (Bio-Rad Laboratories, version 6.1.0 build 7). Only detected proteins were included in graphs.

Mice were intraperitoneally administered 400 μg anti-Ly6G (clone 1A8, BioXCell, West Lebanon, NH) at two and one day prior and three more times after RFA treatment. Depletion of Ly6G⁺ cells was confirmed by IHC staining for NIMPR14 in both RFA-treated and non-RFA–treated tumors. IgG2a isotype (clone 2A3, BioXCell, West Lebanon, NH) was used as control.

Mouse CXCL13/BLC/BCA-1 Quantikine ELISA Kit assay was used to determine CXCL13 levels in mouse serum, splenocytes and tumors homogenates, following the manufacturer's instructions (R&D Systems, catalog no. MCX130).

Mice were intraperitoneally administered a total of 5 injections of 100 μg α-programmed death-ligand 1 (PD-L1) antibody (BIO X CELL, InVivoPlus anti-mouse PD-L1, clone B7-H1, catalog no. BP0101) diluted in buffer pH 7 or vehicle (BIO X CELL, InVivoPlus rat IgG2b, catalog no. BP0090) diluted in buffer pH 6.5 at the day of RFA treatment and every other day up to 10 days after RFA treatment.

Throughout the manuscript, two-way ANOVA was used when comparing two variables (treatment x time); One-way ANOVA was used when comparing one variable (treatment); and unpaired Student t test was used when comparing two groups, except for the analysis of changes in tumor size pre- and post-RFA, which were analyzed as correlated data with paired Student t test to assess differences in each tumor before and after treatment. Multiple t tests were used to analyze IMC and proteome data.

The data generated in this study are available within the article and its Supplementary Data files or from the corresponding author upon reasonable request.

To determine the effects of RFA in vivo, we established a syngeneic mouse model of pancreatic cancer and performed tumor RFA in one flank (Fig. 1A). Assessment of tumor growth in successfully ablated mice revealed significant inhibition of tumor growth in RFA-treated tumors when compared with sham-treated controls (Fig. 1B) with an associated abscopal effect in 87.5% of non-RFA–treated tumors (Supplementary Fig. S1A and S1B; Fig. 1C). RFA ablation did not affect body weight (Supplementary Fig. S1C).

Composite images of H&E staining revealed RFA-treated and non-RFA–treated tumors had significantly increased necrotic area compared with control tumors (Fig. 1D and E). IHC staining of cleaved caspase 3 and granzyme B revealed significantly increased staining in the RFA-treated side when compared with control and non-RFA–treated tumors (Fig. 1F,–I), suggesting differential antitumor mechanisms of RFA response when comparing local and abscopal responses. We performed a membrane-based antibody array in tumor homogenates and found that RFA significantly increased C5/C5a, IL23, and CXCL12 expression in RFA-treated tumors compared with control tumors (Fig. 1J) and chemotactic chemokines CXCL10, CXCL12, and CXCL13 were increased in serum from RFA-treated mice compared with sham-treated mice (Fig. 1K). CXCL13 recruits T and B cells to promote the initiation and maintenance of antitumor responses and is secreted by dendritic cells and myofibroblasts (8–10). Thus, we used a CXCL13 ELISA, which revealed RFA significantly increased CXCL13 in serum, RFA-treated tumors, non-RFA tumors and splenocytes from RFA-treated mice compared with controls (Supplementary Fig. S1D–S1F). These findings are consistent with the role of CXCL13 in the activation of the germinal center and adaptive immunity (11).

In tumor homogenates, IL23, commonly secreted by neutrophils (12, 13), was significantly elevated (Fig. 1J). We assessed neutrophil abundance by IHC staining for NIMPR14 (Fig. 2A and B) and MPO (Fig. 2C and D, black arrows) and found they were significantly elevated in both RFA and non-RFA–treated tumors compared with controls, especially in necrotic core regions of the RFA and non-RFA tumors. We observed colocalization between MPO and NIMPR14 indicating a number of neutrophils in RFA-treated tumors are pro-inflammatory (Supplementary Fig. S1G).

The immune system orchestrates key roles in neovascularization, collagen deposition, tissue remodeling and is important for immunotherapy response (14). Non-RFA–treated tumors revealed the highest expression of αSMA (Fig. 2E and F), collagen and CD31⁺ cells (Fig. 2G–J) in both RFA- and non-RFA–treated tumors compared with controls, indicating RFA remodels the PDAC tumor microenvironment (TME) in locally treated and distant nontreated sites.

We used IMC to evaluate three regions from distant non-RFA–treated tumors with a focus on the regions in and adjacent to the necrotic core (Fig. 3A). IMC cluster value data revealed increased abundance of Ly6G⁺CD11b⁺CD44⁺ neutrophils (Fig. 3B) and Neighborhood analysis showed neutrophils strongly colocalized with αSMA, PanCK⁺ tumor cells and PanCK⁺CD44⁺ tumor stem cells, as well as with CD11c⁺CD44⁺ dendritic cells, F480⁺CD44⁺ macrophages and CD86⁺ marker (Fig. 3C and D).

To assess whether neutrophils are important in the abscopal effect, we performed an in vivo neutrophil depletion (ND) experiment in combination with RFA ablation (Fig. 3E). After tumors were established, mice were intraperitoneally administered 400 μg anti-Ly6G or placebo 2 times pre- and 3 times post-RFA. Depletion of Ly6G⁺ cells was confirmed by IHC staining (Supplementary Fig. S2A). Group comparison evidenced a significant increase in tumor size only in non-RFA–treated tumors from RFA-treated, neutrophil-depleted mice, indicating neutrophils are critical for the abscopal response (Fig. 3F). RFA-treated, neutrophil-depleted mice revealed increased αSMA expression only in distant non-treated tumors, indicating neutrophils may play a key role in mediating abscopal fibrosis through direct or indirect modulation of myofibroblasts (Fig. 3G; Supplementary Fig. S2B). CD31 IHC staining showed no differences between groups (Fig. 3H; Supplementary Fig. S2C). We analyzed CXCL13 levels in the serum and tumors from these mice. We observed a significant reduction in CXCL13 expression in RFA-treated, neutrophil-depleted tumors (Fig. 3I) and ND reduced systemic CXCL13 levels (Fig. 3J), suggesting its decrease may contribute to reduced tumor growth.

To determine longer-term outcomes of RFA on growth rates of RFA-treated and non-RFA tumors, we examined post-RFA tumor sizes for 7 days compared with sham-treated mice. Although we consistently observed significant differences in tumor size post-RFA in RFA-treated and abscopal tumors at post-RFA Day 4 (Post 4D), when we measured tumors on Day 7 post-RFA (Post 7D), the growth restraining capacity of RFA was not as prominent (Fig. 4A). To comprehensively evaluate the immune response at Post 4D, when the antitumor response was most prominent, we used IMC and selected n = 6 regions from sham, RFA and non-RFA tumors (Fig. 4B). Dendritic cells (Cluster 3; CD11b⁺CD11c⁺) were significantly elevated in RFA-treated tumors compared with either sham or non-RFA tumors (Fig. 4C and D). Myeloid-derived suppressor cells (MDSC; Cluster 2; CD11b⁺Ly6C⁺) were significantly decreased in both RFA and non-RFA tumors compared with sham. CD4⁺ and CD8α⁺ T cells were both significantly increased in non-RFA tumors compared with RFA tumors (Cluster 10; CD4⁺) or RFA and sham tumors (Cluster 13; CD8a⁺). These data indicate RFA promotes a systemic antitumor abscopal response driven by CD4⁺ and CD8α⁺ T cells against the non-RFA tumor.

IMC analyses also revealed RFA-treated tumors significantly increased cluster density levels of PD-L1 compared with sham tumors (Cluster 16). PD-L1 was elevated in non-RFA tumors, although the trend was not significant compared with sham-treated tumors (Fig. 4E). Spatial analysis in RFA-treated tumors revealed PD-L1 localizes closely with PanCK⁺ and PanCK⁺αSMA⁺ cells (Fig. 4F; Clusters 22 and 23, respectively).

To test the hypothesis that PD-L1 expression post-RFA was important for relapse of the post-RFA response, we repeated the experiment and began treating mice with anti–PD-L1 compared with isotype control antibodies at the time of RFA treatment (Fig. 4G). Treatment with anti–PD-L1 restrained tumor growth on both the RFA and non-RFA tumors on Day 7 and Day 10 post-RFA. This effect was not observed when correlated tumor growth was analyzed within each tumor in mice treated with an isotype control antibody (Fig. 4H).

We examined the impact of EUS-RFA in 3 patients with PDAC who had received RFA ablation as previously described (15). Stage I and locally advanced Stage III pancreatic cancer resected tumors presented a necrotic area with limited epithelial cells in the ablated region, residual tumor foci and normal pancreas (Supplementary Fig. S3A and S3B). We analyzed MPO, CD31⁺ cells and granzyme B in the locally advanced Stage III resected tumor and observed a significant presence of MPO (Supplementary Fig. S3C, top), CD31⁺ cells (middle) and Granzyme B (bottom) in ablated and distant areas. We performed a proteome profile array using PDAC patient serum pre– and post–EUS-RFA ablation and found increased levels of CCL5, CD40, C5/C5a, ICAM, MIF, and SERPIN (Supplementary Fig. S3D). The results in human pancreatic tumors after EUS-RFA show a strong concordance with our preclinical findings.

Together, the studies presented herein, provide evidence that RFA-induced local and abscopal effects are capable of restraining tumor progression and inducing immune and stromal modulations that can be leveraged to improve therapeutic response. Our findings show ICB therapy targeting RFA-induced PD-L1 in combination with RFA sustained inhibition of tumor progression in both local and abscopal tumors. Prospective clinical trials are needed to better evaluate RFA effects in metastatic disease and assess whether RFA improves the therapeutic outcome of chemotherapy or ICB at distant sites. Our data are in line with previous studies that have shown radiation therapy modifies the abscopal action of tumor-infiltrating neutrophils by polarizing them to an antitumor phenotype (16). Future preclinical experiments evaluating RFA responses using multiple treatments including ICB will determine if these combination strategies can be further considered for clinical stages based on their ability to potentially restrain tumor growth and promote survival for patients with PDAC, especially in patients with locally advanced or metastatic disease where innovative therapeutic options are needed to improve survival.

V. Mota reports grants from UT Health Science Center Houston during the conduct of the study. N.C. Thosani reports other support from Boston Scientific Corp., Pentax America, UpToDate, AbbVie; and other support from ROSEAid outside the submitted work. No disclosures were reported by the other authors.

E.Y. Faraoni: Conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft. B.J. O'Brien: Data curation, formal analysis, investigation, visualization, writing–review and editing. L.N. Strickland: Data curation, validation, investigation, visualization, writing–review and editing. B.K. Osborn: Data curation. V. Mota: Data curation, investigation. J. Chaney: Data curation. C.L. Atkins: Data curation, investigation. P. Cen: Methodology. J. Rowe: Methodology. J. Cardenas: Investigation, methodology. K.L. Poulsen: Data curation, investigation, methodology. C.J. Wray: Data curation, formal analysis, methodology, writing–review and editing. N.C. Thosani: Data curation, methodology. J.M. Bailey-Lundberg: Conceptualization, resources, formal analysis, supervision, funding acquisition, visualization, writing–original draft, writing–review and editing.

J.M. Bailey-Lundberg received funding from the Texas Medical Center Digestive Disease Center Pilot Award 2P30DK056338–16, R21CA249924, and a grant from the National Pancreas Foundation. K.L. Poulsen is funded by R00AA026648.