All mouse model procedures are in compliance with UTHealth’s CLAMC Animal Welfare Committee Review and approved on Dr. Bailey’s AWC protocol. 100k KPC cells were prepared in PBS : Matrigel mix (1:1) and injected subcutaneously in the right flank of 2-month-old female C57BL/6 mice. Tumor size was recorded twice per week with 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 pancreatic ductal adenocarcinoma.

RFA was performed when tumors reached 200-500 mm3. Mice were anesthetized with isoflurane via a Patterson Scientific Posi-Vac nose cone. In the right-side tumor a small incision was made in the center of the tumor to insert the Habib EUS RFA probe perpendicular to the skin. At time of enrollment, tumors were 21 days post-implantation. Average power delivered was 1.82 W over 10-20 seconds. Mice were observed for signs of pain or discomfort post-ablation. Results were compared with no RFA treated (Sham) controls which were transferred to the procedure room, but no incision was performed in the tumor.

Tissues were fixed in zinc buffered formalin, processed according to standard protocols, and embedded in paraffin. The unstained sections were baked at 60°C for 45 minutes. The sections were deparaffinized with Histoclear and rehydrated. Antigen retrieval was performed using heat-mediated microwave methods and the antigen unmasking solution (Abcam 100X Tris-EDTA pH 9) was used for the Granzyme B (Thermo Fischer, catalog number 13-8822-82), CD73 (Cell Signaling, catalog number D7F9A), ADA (Abcam, catalog number ab187048), CD39 (Abcam, catalog number ab223842) antibodies, Ki67 (Abcam, catalog number ab16667) and CD8α (Cell Signaling, catalog number 98941T). All sections were blocked for one hour in 10% FBS in PBST and primary antibodies were incubated overnight at 4°C. Secondary antibodies were used at 1:500 and incubated at room temperature for 30 minutes. The Vectastain ABC kit Peroxidase Standard (Vector Laboratories, PK4000) and DAB Peroxidase (HRP) Substrate kit (Vector Laboratories, SK-4100) were used. Slides were mounted in Cytoseal XYL (Epredia 8312-4) mounting media. Analysis of IHC staining and necrotic area in H&E images was performed in 3 to 5 sections per group using ImageJ (http://imagej.nih.gov/ij/) software. For IHC, color intensity threshold was used to identify positive staining and normalize all tissues. A total of 14 to 20 representative fields per group were selected for quantification. Necrosis was calculated as percentage of total tumor area in each field and assessed by morphological appearance. Necrosis was assessed in areas with dissolution of the cellular architecture, with absent or shrinked nuclei; whereas areas with growing tumor were considered when cells presented well-pronounced nuclei. Additionally, freehand selection tool was used to isolate necrosis versus KPC tumor growing areas, and its quantification was performed in a total of 10 to 15 representative fields per group analyzed. Composite images were taken at 4X

To determine the effects of RFA in vivo, we established a syngeneic mouse model of PDA and performed tumor ablation in one flank as described above. In vivo CD73 inhibition was assessed every other day after RFA treatment to study CD73/ADO pathway implication in RFA response. IP injections of AB680 (diluted in 10%DMSO+90% SBE beta cyclodextrin (SBE-b-CD) in 0.9% saline) in a dose of 10mg/kg was administered. Vehicle group received IP injections of 10%DMSO+90% SBE-b-CD in 0.9% saline. Results were compared with RFA alone treated mice.

Aspartate aminotransferase (AST) assay was used for the quantitative determination of aspartate aminotransferase (AST) in mouse serum, following the manufacturer instructions (Teco diagnostics, catalog number A559-150).

Total RNA was extracted with RNeasy RNA isolation kit (Qiagen) and reverse transcribed with a cDNA Reverse Transcription Kit (Bio-Rad). qPCR was performed with SYBR Green Master Mix (Bio-Rad) on a Bio-Rad real-time PCR system. PCR primer sequences for CD39, Nt5E and ADA were obtained from PrimerBank (https://pga.mgh.harvard.edu/primerbank/) and synthesized at Integrated DNA Technologies. Cyclophilin B was used as housekeeping gene and the expression levels of mRNA of interest were normalized to Cyclophilin B.

Mouse serum and tumor tissues were collected at the end of the study. Serum was frozen with adenosine inhibitor cocktail containing 10 μM APCP (CD73 inhibitor; αβ-methylene ADP, Sigma-Aldrich; catalog no. M3763) 10 μM dipyridamole (equilibrative nucleoside transporter inhibitor; Sigma-Aldrich; catalog no. D9766) and 10 μM deoxycoformycin (adenosine deaminase inhibitor; R&D Systems; catalog no. 2033) for preservation of nucleosides. Tumors were flash frozen with liquid nitrogen and stored at -80C. Protein lysates were extracted from serum of tumor tissues with perchloric acid and neutralized with KHCO3/KOH. Samples were acidified with ammonium dihydrogen phosphate and phosphoric acid. Reaction supernatant was collected after centrifugation. Extracted samples were analyzed by reversed phase high performance liquid chromatography (RP-HPLC). Representative AMP, ADO and INO peaks were identified and measured using the respective standard HPLC curve. For tumors metabolite levels were normalized to the protein levels in the tumor lysate.

Throughout the manuscript, Student’s t test was used when comparing only 2 groups. One-way and two-way ANOVA were performed when one variable (treatment) and two variables (treatment x timepoint) were studied, respectively.

To investigate RFA effects on PDA tumor progression, a RFA preclinical model of PDA was designed (see Methods and Experimental Design in Figure 1A ). Kras^LSL-G12D/+;Trp53^LSL-R172H/+;Pdx1-Cre (KPC) cell line was derived from a pure C57BL/6 host; thus, implantation occurred in an immunocompetent C57BL/6 mouse. In RFA treated mice, tumor growth was assessed 4 days before (Pre), immediately before (RFA), 4 (Post 1) and 7 (Post 2) days after treatment and results were compared to a Sham treated control group of KPC tumors which received no RFA. Tumor volume analysis revealed Sham treated tumors continued to grow during the study ( Figure 1B , left panel, black lines), significantly increasing their tumor volume at 4 (Post 1, p<0.05) and 7 days (Post 2, p<0.0001) after Sham treatment. Contrarily, tumors that were initially growing (p<0.05) up to RFA treatment day ( Figure 1B , right panel, red lines) presented no significant difference in tumor size Day 4 after treatment (Post1, p=ns) and showed a slight increase by Day 7 after RFA (Post 2, p<0.05). Similarly, fold change (FC) assessment as evaluated by tumor size ratio between Day 4 and Day 0 (FC 4D, Post 1) and Day 7 and Day 4 (FC 7D, Post 2) revealed FC increase in Sham treated tumors (p<0.05) and no significant change in RFA treated tumors ( Figure 1C ). Thus, as previously shown by other studies (16, 17), in this work we show RFA promotes an acute restraint in tumor growth when compared to Sham treated tumors; however, the effect is transient and tumor growth rates are elevated 4 days after treatment.

Since the PDA TME is highly immunosuppressive, we hypothesized RFA may be triggering resistant mechanisms which could be driving tumor growth right after the initial acute response observed at day 4. To test our hypothesis, we repeated the experiment and euthanized the mice 10 days after Sham/RFA treatment and focused our studies on the ADO pathway, as it has been described to play important roles in PDA progression and immunosuppression. We measured ADO, AMP and INO levels by HPLC in serum and tumors from Sham and RFA treated mice at 4 (Post 4D) and 10 days (Post 10D) after treatment. Tumor AMP content was acutely increased in RFA treated mice at 4 days after treatment ( Figure 1D ; Post 4D; p<0.001). However, AMP decreased to basal levels by day 10 after treatment (Post 10D; p<0.01). Serum AMP levels were increased 10 days after treatment in both Sham (p<0.001) and RFA treated (p<0.0001) mice with no differences between groups ( Figure 1E ). Tumor ADO content was found significantly increased in RFA treated mice at both 4 (Post 4D; p<0.01) and 10 (Post 10D; p<0.05) days after treatment with no variation between timepoints ( Figure 1F ). Similarly, serum ADO levels increased in RFA treated mice but only at 4 days after treatment (Post 4D; p=0.05). This trend was lost by day 10 where only Sham treated mice showed a marked increase in ADO levels when compared to day 4 (p<0.01) ( Figure 1G ). Tumor INO content was found increased (p<0.01) in both Sham and RFA treated mice 4 days (Post 4D) after treatment when compared to levels observed at 10 days (Post 10D), with no significant differences between groups ( Figure 1H ). Serum INO levels ( Figure 1I ) were decreased in RFA treated mice at 4 days after treatment when compared to Sham mice (Post 4D; p<0.01). However, while Sham mice showed a decrease in INO levels by day 10 after treatment (Post 10D; p<0.05), INO levels were elevated in circulation of RFA treated mice (Post 10D; p<0.01) when compared to the effects observed during the acute response. These results suggest the accumulation of both ADO and INO is induced progressively after RFA treatment, which may mediate tumor response by day 7 after treatment.

As the ADO pathway was found elevated after RFA treatment, we hypothesized this elevation could be, at least, partially responsible for the relapse in tumor growth observed at day 7 after treatment. To test our hypothesis, we designed an in vivo treatment to inhibit CD73 activity and test whether its enzymatic inhibition could prevent tumor growth ( Figure 2A ). We intraperitoneally injected AB680, a CD73 small molecule inhibitor, every other day from day 4 (Post 4D) to day 10 (Post 10D) after RFA treatment and compared the results with a group injected with vehicle (RFA + VEH) and a group of KPC tumors treated with RFA alone (RFA). Serum AST levels were analyzed, and reduced toxicity was observed in all RFA treated mice with or without AB680 when compared to Sham treated controls ( Figure S1 ). Inhibition of CD73 with AB680 resulted in sustained prevention of tumor enlargement 10 days after RFA treatment (Post 10D; n=9) when compared to initial tumor volume measurements immediately before RFA (p=ns) ( Figure 2B ). Contrarily, both RFA treated (RFA, p<0.01; n=7) and vehicle treated (RFA + VEH, p<0.0001; n=7) groups presented increased tumor volume 10 days after treatment (Post 10D), after initial acute response (Post 4D) in restraining tumor growth.

To investigate whether the restraint in tumor growth was accompanied by increased necrosis, we analyzed the histopathology of several Hematoxylin and Eosin staining of endpoint (Post 10D) RFA, RFA+VEH and RFA+AB680 treated tumors and compared the results with a standard KPC subcutaneous tumor grown up to 10 days (Control) which received no treatment ( Figures 2C, E ). Increased necrotic area was observed when RFA was used in combination with AB680 compared to KPC control tumors subcutaneously grown up to 10 days (p<0.0001), RFA (p<0.05) and RFA+VEH (p<0.05) treated mice. Additionally, we observed a decrease in Ki67 positive cells in all RFA treated tumors with and without combination therapy ( Figures 2D, F ), suggesting RFA significantly reduces proliferation (p<0.0001).

Previous studies indicate RFA treatment acutely induces anti-tumor immunity in PDA (16, 17). Thus, we next determined by IHC whether Granzyme B was being affected by RFA treatment alone or in combination with AB680 in the long term (Post 10D) and compared the results with a standard KPC subcutaneous tumor grown up to 10 days (Control) which received no treatment ( Figures 3A, C ). Analysis and quantification revealed RFA treatment increased abundance of Granzyme B + cells in RFA+AB680 treated tumors when compared to KPC control tumors subcutaneously grown up to 10 days (p<0.0001), RFA (p<0.01) and RFA+VEH (p<0.01) tumors. Moreover, CD8α positive cells presented a similar pattern ( Figure 3B, D ), suggesting they may be the source of Granzyme B+ secretion (p<0.0001). These results suggest the prevention in tumor enlargement observed at day 10 after treatment is due to increased necrosis and anti-tumor immunity and decreased tumor cell proliferation.

Based on these data, we studied by IHC protein expression of CD73 and Granzyme B + cells to analyze their potential spatial correlation. Composite pictures after RFA treatment revealed a negative correlation between CD73 protein expression and abundance of Granzyme B + cells in RFA treated KPC subcutaneous tumors 10 days after treatment ( Figure 3E ). Additionally, areas with decreased CD73 expression had increased staining for Granzyme B+ cells ( Figure 3F , left panels -a and b-), whereas areas with increased CD73 expression showed decreased staining for Granzyme B+ cells ( Figure 3F , right panels -c and d), suggesting CD73 activity in PDA induces local immunosuppressive effects.

Generation of ADO and INO is well known for immunosuppressive actions in the TME. Hence, to test whether RFA in combination with AB680 modified tumor metabolite generation, we next analyzed by HPLC the tumor content of AMP, ADO and INO at 10 days after RFA treatment. Our results revealed tumor AMP ( Figure 4A ), ADO ( Figure 4B ) and INO ( Figure 4C ) content were significantly reduced only in tumors which received combination treatment (RFA+AB680) when compared to RFA alone (p<0.05). No differences were found between combination treatment and RFA+VEH group in any of the metabolites analyzed.

To study whether VEH and/or AB680 administration may be affecting the expression of the enzymes responsible for production of these metabolites, we studied the protein and mRNA expression of CD39, Nt5E (CD73) and Ada. Expression levels from both CD39 ( Figures 4D, E ) and CD73 ( Figures 4F, G ) did not show significant differences between treatments. Similarly, no differences were found in ADA protein expression by IHC ( Figure 4H ). However, Ada mRNA levels ( Figure 4I ) were found decreased in combination treatment (RFA+AB680) when compared to both RFA (p<0.05) and RFA+VEH (p<0.01), suggesting CD73 inhibition with AB680 effectively reduced the capability of the tumor to produce INO when ADO generation is being inhibited by CD73 activity blockade.

In summary, these results suggest that an in vivo inhibition of CD73 in combination with RFA may prevent the onset of ADO pathway as a resistant mechanism by depriving tumors from the capability to produce immunosuppressive ADO and INO, leading to improved response and increased antitumor immunity.