Endogenous Pancreatic Cancer Cell PD-1 Activates MET and Induces Epithelial-Mesenchymal Transition to Promote Cancer Progression

Simple Summary Here, we describe a previously unreported mechanism of PD-1/MET interaction and PD-1 induction of epithelial-to-mesenchymal transition (EMT), which is observed early in pancreatic cancer development and progression. This oncogenic mechanism is distinct from established immune functions of PD-1 and presents a new paradigm of oncogenic functionality of PD-1 in cancer cells. Our studies demonstrate the foundation and rationale for testing combination therapies targeting PD-1 and MET in pancreatic cancer patients. Abstract We recently demonstrated that immune checkpoint PD-1 was endogenously expressed in pancreatic ductal adenocarcinoma (PDAC) cells. Our data indicated that PD-1 proteins are not exclusive to immune cells and have unrecognized signal transduction cascades intrinsic to cancer cells. Building on this paradigm shift, we sought to further characterize PD-1 expression in PDAC. We utilized a phospho-explorer array to identify pathways upregulated by PD-1 signaling. We discovered PD-1-mediated activation of the proto-oncogene MET in PDAC cells, which was dependent on hepatocyte growth factor (MET ligand) and not secondary to direct protein interaction. We then discovered that the PD-1/MET axis in PDAC cells regulated growth, migration, and invasion. Importantly, the PD-1/MET axis induced epithelial-to-mesenchymal transition (EMT), a well-established early oncogenic process in PDAC. We observed that combined targeting of PDAC cell PD-1 and MET resulted in substantial direct tumor cell cytotoxicity and growth inhibition in PDAC cell lines, patient-derived organoids, and patient-derived xenografts independent of cytotoxic immune responses. This is the first report of PDAC-endogenous PD-1 expression regulating MET signaling, which builds upon our growing body of work showing the oncogenic phenotype of PD-1 expression in PDAC cells is distinct from its immunogenic role. These results highlight a paradigm shift that the tumor-specific PD-1 axis is a novel target to effectively kill PDAC cells by antagonizing previously unrecognized PD-1-dependent oncogenic pathways.


Introduction
Pancreatic ductal adenocarcinoma (PDAC) remains a deadly disease due to a multitude of factors, including frequently late clinical presentation and a limited number of first-line therapeutic regimens that are only moderately effective [1]. Despite clinical activity in

Cell Culture
MIAPaCa-2 and PANC-1 cells were obtained from and validated by American Type Culture Collection (ATCC). Cells were maintained as described [2]. Testing for Mycoplasma contamination, which can cause chromosomal mutations and genetic drift from parental cell lines, is recommended after 10 passages to ensure. All experiments utilized early passage (<P10) cells. However, given the time between receiving the P0 cells from ATCC, cell banking, and subsequent use of early passage cells, we validated the genomic integrity of our cell lines through ATCC. Validation of genomic integrity mitigates the concern for mycoplasma contamination. PD-1 knockdown (KD) PDAC cells were generated as reported [2].

Ethics Approval and Patient Tumor Acquisition
The University of Kentucky (UK) Institutional Review Board approval and written informed patient consent were obtained prior to obtaining tissue samples in accordance with the Declaration of Helsinki. Excess cancer tissues were processed to create patientderived organoids (PDOs) and xenografts (PDXs) as described [2,12,13].

Western Blots
Cell and PDO lysates were collected and analyzed by western blot as described [2,13]. For EMT studies, cells were grown to 60-80% confluence to capture mesenchymal processes. Cells were serum-starved overnight prior to treatment with PD-L1 (1 µg/mL), vehiclecontrol (0.1% BSA), or when evaluating EMT and phosphorylated proteins. When noted, cells were treated with PEM (250 µg/mL) or TIV (0.25 µM) 30 min before PD-L1 stimulation. Blots were probed independently with antibodies listed in Table S1. Blots were visualized with ECL solution. Quantified expression values were normalized to controls from the same blot and from at least two separate experiments. Original western blot images can be found in Supplemental File S2.

Quantitative PCR
Cells were treated with recombinant-human PD-L1 (1 µg/mL) or solvent-control (0.1% BSA) for 2 h. Cells were lysed, total RNA extracted, and reverse transcription performed per the manufacturers' instructions. Quantitative real-time PCR (qRT-PCR) was performed in quadruplicate using TaqMan reagents per the manufacturer's instructions. Data were analyzed using the 2 −∆∆CT method with results normalized to β-actin.

In Vivo Drug Testing
The UK Institutional Animal Care and Use Committee approved the described murine studies. ARRIVE guidelines were followed (Supplementary File S1, Table S2). PDX mice were randomly divided into the following treatment groups (N = 6 per group) and treated for four weeks: control, TIV (150 mg/kg, oral gavage every 5 d + 2 d holiday), PEM (30 mg/kg, intraperitoneal injection twice weekly), TIV + PEM [15]. Researchers were not blinded to treatment groups.

Immunohistochemical (IHC) Staining
Excised PDX tumors were fixed in 10% formalin and submitted to the UK Biospecimen Procurement and Translational Pathology (BPTP) Shared Resource Facility (SRF), which performed all histological sectioning and staining (Supplementary File S1).

Statistical Analysis
Experiments were performed at least twice. Statistical analyses were performed using GraphPad Prism software (v9). For two-group analysis, two-tailed Student's t-tests were used. One-and two-way ANOVA with post-hoc Tukey tests were used for multi-group comparisons. Differences were considered significant at p < 0.05. Results were graphed as mean ± standard error. For multi-treatment studies, percent and fold-changes were calculated compared to controls to enable easier comparison across all data.

Identification of PD-1/PD-L1 Activated Pathways
We identified proteins and signaling pathways enhanced by PD-1/PD-L1 signaling. After PANC-1 cells were exposed to PD-L1, our phospho-protein array demonstrated upregulated expression changes (fold-change > 2) in histone deacetylase (HDAC)-1 (28.98-fold), MET (13.65-fold) [16], and EGFR (6.4-fold) ( Figure S1). Since anti-HDAC therapies are nonselective and target epigenetic alterations [17] and anti-EGFR therapies such as erlotinib have failed to show clinically meaningful benefit in PDAC [18,19], we chose to investigate MET, which is a promising therapeutic target for various cancers and is overexpressed, upregulated, and inversely associated with prognosis in PDAC [9][10][11]. Interestingly, MET therapies have been shown to overcome anti-EGFR monotherapy and gemcitabine resistance in non-small cell lung cancer (NSCLC) and PDAC [9]. Furthermore, MET regulates EMT, tumor growth, metastases, and interacts with numerous oncogenic pathways in PDAC [9]. These features provided a strong rationale to focus on MET for our studies.

PD-1 and MET Expression in PDAC Cells and PDOs
We detected the expression of PD-1, PD-L1, and MET in PDAC cells and PDO lines by western blot assays. Consistent with the Cancer Cell Line Encyclopedia (CCLE) data [20], our results showed that MIAPaCa-2 and PANC-1 cell lines expressed all three proteins  Figure 1A). However, the highest levels of PD-1 and total-MET expression were observed in PDO lines hPT1 and hPT4, respectively. Following exposure to recombinant-human PD-L1, MIAPaCa-2 and PANC-1 cells both demonstrated time-dependent alterations in MET activation, which peaked at 1 h with 481% and 399% increases, respectively, in phosphorylated-MET (pMET) compared to baseline (p < 0.01 vs. 0 min for all) ( Figure 1B). In blocking assays, PD-L1 treatment alone resulted in 177% and 314% increased pMET levels for MIAPaCa-2 and PANC-1, respectively. Pretreatment with PEM and TIV abrogated PD-L1-induced pMET levels by 83% and 103%, respectively, for MIA-PaCa-2 and 117% and 219%, respectively, for PANC-1 ( Figure 1C). These results demonstrated that PD-L1-induced MET activation is dependent on both MET and PD-1. pathways in PDAC [9]. These features provided a strong rationale to focus on MET for our studies.

PD-1 and MET Regulate Cell Motility and Migration
Given that MET induces EMT, we sought to determine if the PD-1/PD-L1-axis regulates this process to promote motility and migration. To assess cell motility, we performed wound healing assays in PDAC cells. We observed that PD-L1 treatment significantly in-creased motility in MIAPaCa-2 and PANC-1 cells compared to negative-controls (132% and 201%, respectively; p < 0.01) (Figure 2A,B). These motility increases were consistent with HGF positive-control results (237% and 242% for MIAPaCa-2 and PANC-1, respectively), demonstrating that the PD-1/PD-L1-axis induces PDAC cell motility similar to direct MET activation. Additionally, both PD-1 and MET antagonism completely blocked PD-L1induced cell motility in MIAPaCa-2 cells. In PANC-1 cells, compared to negative-control, PD-L1-induced cell motility decreased by 106% and 98% following PD-1 or MET antagonism, respectively (all p < 0.0001 vs. control). This further supported that PD-L1-induced cell motility is dependent on both PD-1 and MET.
To evaluate PD-1/PD-L1/MET activity on cell migration, we next performed transwell migration assays. In both MIAPaCa-2 and PANC-1 cells, PD-L1 treatment alone increased cell migration (138% and 470%, respectively, vs. negative-control). These results were comparable to the increased migration observed with HGF positive-controls (153% and 881% for MIAPaCa-2 and PANC-1, respectively, vs. negative-control). Conversely, compared to the negative control, when PD-L1 treatment was combined with PD-1 or MET inhibition, cell migration was nearly completely blocked for MIAPaCa-2, and decreased by 366% and 220%, respectively, for PANC-1 (all p < 0.01 vs. no treatment) ( Figure 2C,D). The above assays showed that PD-L1-induced motility/growth and migration are both dependent on PD-1 and MET.

The PD-1/PD-L1 Axis Induces EMT
We hypothesized that the PD-1/PD-L1 axis regulates EMT. We used PANC-1 cells for these assays, given their epithelial phenotype compared to the baseline mesenchymal phenotype of MIAPaCa-2 cells. After 48 h of PD-L1 stimulation, PANC-1 cells demonstrated EMT features with increased expression of mesenchymal markers MMP9 (49%), N-cadherin (85%), vimentin (16%), and Snai2 (44%) compared to untreated controls ( Figure 3). Ecadherin increased following treatment with PD-L1 (35%), which is likely due to the increased formation of cell-to-cell adhesions in the interior of cell colonies observed during rapid cell growth [21][22][23]. These interior cells appear to express EMT factors that are unrelated to cell-to-cell interactions. In comparison, N-cadherin is expressed only on cells along the periphery of colonies without cell-to-cell interactions, thus representing the minority of cells.
To corroborate that PD-L1 stimulation increased EMT marker expression through PD-1, we utilized PD-1 KD cells. With PD-1 knockdown, we observed E-cadherin/Ncadherin pattern switching, where E-cadherin increased 30% and N-cadherin decreased 40% vs. untreated controls. Furthermore, mesenchymal-related proteins MMP9, vimentin, and Snai2 decreased in PD-1 KD cells (41%, 26%, and 32%, respectively). These results demonstrate that induction of mesenchymal protein expression is dependent on PD-1, supporting our results showing that PD-L1-induced migration and motility are dependent on PD-1. Taken together, these data demonstrate a novel mechanism for the regulation of EMT by the PD-1/PD-L1 axis.

PD-1 and PD-L1 Do Not Directly Interact with MET
We next sought to investigate if the mechanism for PD-L1-induced MET activation involved a direct PD-1-MET protein interaction. We performed co-IP assays utilizing PD-1 as the "bait" in PDAC cells. Co-IP protein complex samples were analyzed by western blot with anti-MET and anti-PD-1 mAbs ("prey"). We observed no difference in MET nor PD-1 expression between IgG control and co-IP samples ( Figure S2A), indicating there is likely no direct protein interaction between MET and PD-1 in PDAC cells.
To evaluate if PD-L1-induced MET activation is due to exogenous PD-L1 directly binding to and stimulating MET, we performed additional co-IP assays with PD-L1 as the "bait." Cells were treated with PD-L1 for 1 h prior to lysate collection. PD-L1 pull-down did not show increased levels of MET compared to IgG controls, demonstrating there is likely no direct PD-L1-MET interaction and PD-L1 likely does not act as a ligand for MET herin pattern switching, where E-cadherin increased 30% and N-cadherin decreased 40% vs. untreated controls. Furthermore, mesenchymal-related proteins MMP9, vimentin, and Snai2 decreased in PD-1 KD cells (41%, 26%, and 32%, respectively). These results demonstrate that induction of mesenchymal protein expression is dependent on PD-1, supporting our results showing that PD-L1-induced migration and motility are dependent on PD-1. Taken together, these data demonstrate a novel mechanism for the regulation of EMT by the PD-1/PD-L1 axis.

PD-1/PD-L1 Axis Upregulates HGF mRNA Expression
We hypothesized that PD-1/PD-L1 indirectly activates MET through HGF expression. We assessed HGF mRNA expression following PD-L1 stimulation in PD-1 KD PDAC cells by qPCR (Figure 4). Following exposure to PD-L1, MIAPaCa-2 and PANC-1 PD-1 control cells demonstrated 64% and 84% increased HGF mRNA, respectively (p < 0.05 vs. KD control for all). To determine if PD-1 is a requisite for PD-L1-mediated HGF expression, we evaluated HGF mRNA levels in PD-1 KD cells. We detected minimal expression of HGF mRNA in PD-1 KD cells compared to control cells. Furthermore, PD-L1 exposure in PD-1 KD cells did not alter HGF expression levels, indicating that HGF expression is dependent on both PD-1 and PD-L1. These findings support our earlier data showing that PD-L1-induced MET activation is dependent on PD-1.
control cells demonstrated 64% and 84% increased HGF mRNA, respectively (p < 0.05 vs. KD control for all). To determine if PD-1 is a requisite for PD-L1-mediated HGF expression, we evaluated HGF mRNA levels in PD-1 KD cells. We detected minimal expression of HGF mRNA in PD-1 KD cells compared to control cells. Furthermore, PD-L1 exposure in PD-1 KD cells did not alter HGF expression levels, indicating that HGF expression is dependent on both PD-1 and PD-L1. These findings support our earlier data showing that PD-L1-induced MET activation is dependent on PD-1. Knockdown of PD-1 decreased baseline HGF mRNA expression by 499% and 1532% compared to KD controls, respectively. Treatment with PD-L1 was unable to overcome this reduction in HGF in PD-1 KD cells, signaling PD-1 is required for PD-L1-induced HGF mRNA production. * p < 0.05, ** p < 0.01 vs. untreated KD controls. Knockdown of PD-1 decreased baseline HGF mRNA expression by 499% and 1532% compared to KD controls, respectively. Treatment with PD-L1 was unable to overcome this reduction in HGF in PD-1 KD cells, signaling PD-1 is required for PD-L1-induced HGF mRNA production. * p < 0.05, ** p < 0.01 vs. untreated KD controls.

PD-1 and MET Inhibition Effectively Slows Tumor Growth in PDXs
We next sought to test our mechanism-based regimen in vivo. We established PDX tumors generated from hPT1. PDX mice were given vehicle-control, TIV, PEM, or TIV + PEM as described earlier. We utilized TIV for these studies given its selective MET antagonism, thereby enabling evaluation of the in vivo PD-1/MET axis mechanism without the influence of off-target effects. We observed the slowest tumor growth with combined TIV + PEM treatment ( Figure 6A). Furthermore, IHC staining of cell proliferation marker protein Ki67 ( Figure 6B) and cell apoptosis marker cleaved caspase-3 ( Figure 6C) showed that TIV and PEM monotherapy cohorts inhibited tumor cell proliferation (decreased 0.65-fold and 0.51-fold vs. control, respectively) and promoted tumor cell apoptosis (increased 14.1-fold and 13.3-fold vs. control, respectively). However, this effect was greatest with combined PD-1/MET antagonism (decreased proliferation 0.82% and increased apoptosis 25.3-fold vs. control) ( Figure 6D,E, all p < 0.05). These data demonstrate in vivo synergistic efficacy with combined PD-1/MET antagonism that inhibits tumor growth, decreases proliferation, and promotes direct tumor cytotoxicity. Importantly, our in vivo studies show direct tumor cell targeting, similar to that observed in our in vitro results, in the presence of a complex tumor microenvironment. Furthermore, these results demonstrate a novel mechanism of action for anti-cancer IO therapies apart from immune reactivation.

Discussion
Our current studies demonstrated endogenous PD-1 expression in PDAC cells and tissues and revealed that PD-1 signaling upregulated oncogenic MET activity to subsequently induce EMT. The dependence of PDAC growth on the PD-1/MET axis was substantiated by high levels of cytotoxicity when both PD-1 and MET were therapeutically targeted. These data, along with our prior report [2], continue to challenge the paradigm that PD-1 expression is exclusive to T-cells and cytotoxic immune responses. Instead, our results involving the PD-1/MET axis in PDAC cells underscore the growing body of work showing that the tumor-endogenous immune checkpoint PD-1 has oncogenic properties supporting cancer progression. Examples of such dual functionality of proteins are evident even from our prior studies on chemokine receptors [29][30][31]. G-protein coupled receptor expression in immune cells has physiologic roles in organogenesis, development, and inflammation [32]. In cancer cells, however, chemokine receptors (e.g., CXCR4) have been shown to support an invasive phenotype [29][30][31]33].
PD-1 regulation of MET, which is an important proto-oncogene for many cancers, underlies how this axis supports PDAC progression. MET is well-known to activate numerous downstream oncogenic pathways, induce EMT signals, and promote cancer growth, invasion and metastases [9,10]. In fact, MET overexpression has been detected in nearly two-thirds of PDACs and has been associated with worse clinical outcomes [11]. Interestingly, other recent studies have shown that activated MET and EMT pathways increase tumor-endogenous PD-L1 production [34,35]. These studies, along with our results, identify a potential positive feedback loop and will be explored in future investigations. Nevertheless, in this report, we are the first to demonstrate that tumor-endogenous PD-1 promotes oncogenic MET signaling to induce EMT and promote PDAC progression.
Given that our data demonstrated that the PD-1/MET axis supported PDAC growth, we sought to test whether therapeutic targeting would substantiate PDAC growth dependency on this pathway. Indeed, our results revealed that combined anti-PD-1/-MET antagonism resulted in high levels of PDAC cytotoxicity in both in vitro and in vivo cancer models. There are important implications from these results. First, our assays revealed that single-agent regimens were suboptimal and that combination regimens yielded higher synergistic cytotoxic responses. These results may suggest that alternate or escape signaling pathways are not attenuated with single-agent regimens. Second, our drug treatment assays were performed in immune-deficient PDAC models, and the high levels of cytotoxicity challenge the concept that IO anti-cancer efficacy is derived only from immune reactivation. Notwithstanding, it is reasonable to deduce that in the clinical management of PDAC patients, therapeutic regimens that incorporate IO drugs combined with targeted therapies may produce optimal cytotoxicity from both harnessing adaptive immune responses as well as from direct cytotoxic killing. Our group is developing immune-competent PDAC models to examine both avenues of cytotoxicity in our future studies.
Our study results are of particular clinical importance given that there has been slow progress on the development of novel effective therapeutic regimens for metastatic PDAC since FDA approval of first-line regimens FOLFIRINOX and Gemcitabine/nab-Paclitaxel in 2010 and 2013, respectively, and second-line regimen NALFIRIFOX in 2015 [36][37][38]. Regimens that provide durable improvements in the long-term survival of this disease are lacking, creating an urgent need for innovative therapeutics. The CheckMate 9ER trial was a landmark study combining an IO drug (nivolumab, anti-PD-1 mAb) and multi-kinase inhibitor with anti-MET properties (CABO) and has resulted in a new standard-of-care firstline therapy for advanced renal cell carcinoma (RCC) [39]. Trials such as CheckMate 9ER provide foundational support to investigate combined anti-PD-1/anti-MET therapies in PDAC patients. Consequently, we have organized a clinical trial evaluating CABO + PEM in advanced PDAC (NCT05052723). This is the first clinical evaluation of combined PD-1/MET antagonism in PDAC.

Conclusions
Our data challenges the current paradigm of immune checkpoints and immune tolerance by showing that tumor-intrinsic PD-1 supports an oncogenic phenotype in PDAC apart from its role in the inhibition of cytotoxic immune responses. Our studies revealed a novel mechanism wherein tumor-endogenous PD-1 increased MET activation and induced EMT in PDAC. This oncogenic signaling axis was targeted to effectively promote direct tumor cytotoxicity in PDAC cells and tissues. These study results served as the basis for a current clinical trial testing the combination of PEM and CABO in metastatic PDAC (NCT05052723).