A feasibility study characterising tumour microenvironment of longitudinally collected biopsies of pancreatic tumours upon endoscopic ultrasound guided radiofrequency ablation

Most patients with pancreatic ductal adenocarcinoma (PDAC) are metastatic at presentation 27 with dismal prognosis warranting improved systemic therapy options. Longitudinal sampling 28 for assessment of treatment response poses a challenge for validating novel therapies. In this 29 feasibility study, we evaluate the role of endoscopic ultrasound (EUS)-guided serial fine-needle 30 aspiration biopsies (FNABs) to study the molecular changes associated with radiofrequency 31 ablation (RFA). For the first time, this feasibility study validates longitudinal sampling by EUS-FNABs as an appropriate research tool to study tumour microenvironmental changes associated with local 56 PDAC immunomodulatory treatments like RFA.


Abstract 25
Background 26 Most patients with pancreatic ductal adenocarcinoma (PDAC) are metastatic at presentation 27 with dismal prognosis warranting improved systemic therapy options. Longitudinal sampling 28 for assessment of treatment response poses a challenge for validating novel therapies. In this 29 feasibility study, we evaluate the role of endoscopic ultrasound (EUS)-guided serial fine-needle 30 aspiration biopsies (FNABs) to study the molecular changes associated with radiofrequency 31 ablation (RFA). 32

Methods 33
Two stage III inoperable gemcitabine-treated PDAC patients were recruited in the treatment 34 arm of ARDEO (ethically approved Phase-II prospective randomized clinical study). Post 35 examination, targeted RFA was delivered thrice, and sequential FNABs of tumour were taken 36 before and after treatment and analysed using a custom NanoString panel (144 genes) 37 consisting of cancer and cancer-associated fibroblast (CAFs) subtypes and immune changes. 38 CAF culture was established from one FNAB and characterised by immunofluorescence and 39 immunoblotting. 40

Study Design 117
Patients were recruited through the treatment arm of the pilot phase of the ARDEO trial (A 118 phase II prospective Randomised clinical study of enDoscopic ultrasound guided 119 radiofrEquency ablation for inOperable pancreatic ductal adenocarcinoma). This is a 120 prospective randomised clinical study (REC reference: 18/SW/0103) in patients with stage 3 121 inoperable adenocarcinoma of the pancreas starting gemcitabine chemotherapy with a 1:1 122 randomisation. Patients in the treatment group receive up to 3 endoscopic ultrasound 123 radiofrequency ablation procedures, each a month apart and sequential EUS FNABs were taken 124 during each procedure prior to RFA treatment. Written informed consent was taken from the 125 patients before enrolling into the study. Patients in the control group will receive best medical 126 care. A study flow diagram is illustrated in Fig 1a. PDAC patients in the treatment group were 127 recruited over 2 months at the Hammersmith Hospital, Du Cane Road, London. 128

Endoscopic-ultrasound guided radiofrequency ablation 129
Following EUS examination of the pancreas and primary tumour site, an FNAB of the tumour 130 was taken and then targeted RF delivered using 10 Watts for 2 minutes per application using 131 the Habib EUS RFA (Boston Scientific, MA, USA). For the second and third EUS 132 examinations, the ablation zone is examined, and a fine needle biopsy taken prior to RFA. 133 FNABs were snap frozen in liquid nitrogen and transported on dry ice. Routine biochemistry 134 tests including bilirubin and CA19-9 were analysed at Imperial College Healthcare NHS 1 h at 37°C. Following digestion, tubes were centrifuged for 5 min at 1200 rpm and the tissue 174 pellets were resuspended in 2 ml of ACK lysis buffer (Gibco, NY, USA) to deplete red blood 175 cells by incubating on ice for 3 min. 5 ml of complete media was added to neutralize the solution and tubes were centrifuged for 5 min at 1200 rpm. 3 ml of pre-warmed complete media 177 was added to the tissue pellet and contents were placed on 35 mm dishes. After a week, 178 fibroblasts that grew out from the tissue bits were trypsinised and passaged once a week. Cells 179 were maintained at 37°C in a 5% CO2 incubator. Cells were screened for mycoplasma by 180 polymerase chain reaction and found to be negative ( Figure S1). DNA for STR profiling was 181 sent to Eurofins Genomics (Wolverhampton, UK).  Table S3. Odyssey Fc (Cambridge, UK) using enhanced chemiluminescence (ECL, Merck Millipore, 208 Darmstadt, Germany). b actin was used a loading control. Antibody information is listed in 209 Table S3. 210

Statistics 211
To assess RFA-mediated transcriptome modulation in serial FNABs, log2 transformed gene 212 expression values at T2 (after 1 st RFA) and T3 (after 2 nd RFA) were subtracted from T1 213 (baseline) and denoted as difference. 2^ (difference) was then performed to calculate fold 214 change in gene expression over the course of RFA treatment. complications or morbidity relating to the EUS or RFA treatment. However, she did suffer with 232 Grade 1 nausea related to gemcitabine treatment that was fully responsive to the anti-nausea 233 drug, domperidone. The patient elected to take a treatment break after cycle 6 of gemcitabine 234 as she was thinking of relocating and wanted to go away on holiday. At the time of stopping 235 treatment, she had stable disease radiologically and her CA19-9 had decreased from 4887 when enrolled in the trial to 392 on the 25 th of April, which was after her final EUS treatment (cycle 237 3). After cycle 3, the CA19-9 increased to 1324 at the end of cycle 6 and at the start of the 238 treatment break. She re-presented in October with liver metastases, portal vein thrombosis and 239 a CA19-9 of 503,293 and underwent one cycle of CAPOX (oxaliplatin and capecitabine) 240 chemotherapy but sadly passed away on the 15 th of November. 2S. Updated imaging on 28 th of August 2020 shows stable disease (Fig. 1b). 256

Feasibility of evaluating whether EUS-RFA treatment elicits an anti-tumour immune response 257 in PDAC 258
To obtain insight into RFA mediated modulation of the TME, we performed transcriptomic 259 profiling of FNABs collected longitudinally over three courses of RFA for the two patients 260 enrolled in our proof-of-principle study. A schematic representation of sample collection and 261 approaches is outlined in Fig. 2a. A low-cost custom NanoString panel was designed to include 262 genes for PDA and CAF subtypes identified previously (Collisson et al. 2011;Neuzillet et al. 263 2019), along with markers to assess changes in the TME as a result of RFA treatment (Table  264 S2). Initial unsupervised hierarchical clustering of the six tumour samples grouped T3 FNABs 265 (after 2 nd RFA) of both the patients together and showed enrichment of T cell markers, immune 266 checkpoint genes and increased fibrosis observed as upregulated CAF markers. T3 samples 267 were most similar to T2 (after 1 st RFA) profile but only in patient 2. T1 (baseline) and T2 (after 268 1 st RFA) from patient 1 clustered together and shared similar expression profiles to T1 269 (baseline) from patient 2, except the latter was highly enriched for T and B cell genes along 270 with cytolytic function markers, Fig. 2b. This suggests that the baseline sample from patient 2 271 was already immune-enriched, which is not the case with patient 1. 272 We first assessed RFA-induced immunological changes resulting in an increased adaptive 273 immune infiltration into the tumours. A high T (7.3-fold) and B (2.2-fold) cell score at T3 in 274 patient 1 compared to baseline and T2 was observed. There was no change in T and B cell 275 score in patient 2 at T2 and T3 compared to baseline ( assessed this ratio using CD8A and FOXP3 gene expression and found greater than 3-fold 280 upregulation in patient 1 at both T2 and T3 compared to baseline. This ratio, however, was 281 lower in patient 2 after RFA treatment (Fig. 2d). at T3 was also found to be 3.9-fold greater than at baseline but a decreasing trend in HLADQA1 288 was observed (Fig. 2e). However, we warrant further validation of MHC-I and -II using 289 additional gene sets representing the complexes in the future. baseline levels of CD4 and CD8 were, however, higher in patient 2 compared to patient 1, 297 suggesting that patient 2 was already inflamed with CD4 and CD8 T cells before RFA. 298 A reduction in transcript levels of genes representing macrophages (2.6-fold) and mast cells 299 (3.7-fold) and no change in NK cell genes was observed in patient 2 post-RFA (at T3) 300 compared to baseline (Fig. 2f). Further, apart from upregulated T cell markers, a 6.4-fold and 301 8-fold increase in cytotoxic T cell function score and inflammation score, respectively was 302 observed at T3 compared to baseline in patient 1. Decrease in cytolytic function score (14.17-303 fold at T3) and a 2-fold increase in inflammation score was observed after 2 courses of RFA 304 in patient 2 (Fig. 2g). 305 T cell infiltration after RFA treatment has been shown to be accompanied by an exhausted T 306 cell phenotype marked by increased expression of immune checkpoint genes (Fei et al. 2020). contrast, RFA treated patient 2 tumour (T3) showed a 3.6-fold, 9-fold and 5-fold upregulation 326 of pCAF subtypes A, B and D, respectively (Fig. 3c). We then selected the genes that have 327 been previously validated by immunohistochemistry to exclusively represent pCAF subtypes 328 and found an enrichment of POSTN (subtype A), MYH11 (subtype B) and PDPN (subtype C) 329 in both patients after RFA, however, the increase was pronounced particularly in patient 2 330 tumour at T2 and T3, Fig. 3d. In order to validate CAF-specific expression of these markers, 331 we established cultures from patient FNABs and successfully obtained CAFs from patient 2 332 tumour at T2. A strong expression of pan-CAF markers -PDGFRa, aSMA and VIM was 333 observed along with expression of subtype A (POSTN) and subtype B (MYH11) in the cultured 334 CAFs (Fig. 3e). 335

Feasibility of evaluating whether RFA remodels the cancer subtypes of PDAC 336
In order to study how the crosstalk between immune cells and CAFs in the TME regulates 337 cancer cells, we performed PDA subtyping of tumours pre-and post-RFA by NTP using 338 Nevertheless, it is important to note that patient 2 had increased anti-tumour immune infiltrate 406 at baseline itself, potentially responsible for prolonged overall survival. RFA courses led to no 407 particular increase in T and B cells or cytolytic activity. Interestingly, immune checkpoint 408 genes were upregulated upon RFA in patient 2. The overall immune milieu supports that patient 409 2 may have been a contender for immune checkpoint therapy. Further, response to RFA in the 410 two patients was found to be heterogeneous providing an opportunity to obtain deeper insights 411 into immune signaling in pre-and post-RFA samples. In addition, how these changes are 412 associated with prior treatment, like gemcitabine, may be addressed in the larger sample set 413 which is part of the ARDEO trial. 414 Both the patients at baseline showed enrichment of classical PDA subtype genes which were 415 dramatically downregulated over the RFA courses. However, there was no drastic switch in 416 the subtype to QM or exocrine-like, suggesting that RFA may be controlling cancer cells from 417 switching to poor prognostic cancer type. 418 The results from this feasibility study need to be validated in the larger ARDEO cohort and RFA treated FNABs measured as cytolytic function and inflammatory score, h. and i. 606 Upregulation of immune checkpoint genes as a result of RFA in patient 1 and 2 respectively. 607 # represents greater than equal to two-fold change in gene expression between T1 and T3. 608 Colour key represents T1 in red, T2 in blue and T3 in cyan. 609 PDGFRa and ACTA2 (aSMA) as a response to RFA in patient 1 and 2 respectively, b. and c. 611 Heatmaps representing a switch in the CAF subtypes over the course of RFA in patient 1 and 612 2 respectively; separate scale bars for the two patients depicting averaged log2 normalised 613