The DMEM or MEM were supplemented with 10% fetal bovine serum (FBS; R&D Systems, Minneapolis, MN), 1% glutamax, 1% non-essential amino acids, and 1% penicillin/streptomycin as complete medium. The RPMI1640 was supplemented with 10% heat inactivated FBS, 1% glutamax, 10mM HEPES, and 1% penicillin/streptomycin. All the medium and supplements were obtained from Thermo Fisher Scientific (Waltham, MA) if not separately mentioned.

Selected human pancreatic cancer cell lines (HPAC, Capan2, PANC-1 MiaPaCa-2, CFPAC, HPAFII) were obtained from American Type Culture Collection (ATCC, Manassas, VA 20110, USA) and were cultured as instructed. The retroviral vector packaging line GP2–293 cell line was cultured in complete DMEM.

The tMUC1 expression on human PDA cell lines was assessed by TAB004 staining (1μg/sample; provided by OncoTab Inc., Charlotte, NC) conjugated with HiLyte Fluor 647 (Dojindo Molecular Technologies, Inc., Rockville, MD), named as TAB004-Fluor 647. Dead cells were excluded by 7-AAD staining (BD Biosciences). Data was acquired on BD LSRFortessa flow cytometer (BD Biosciences), and analyzed with FlowJo software (version 10.8.1, Tree Star Inc).

The second-generation tMUC1-specific human CAR construct was synthesized by Dr. John Maher group and the schematic of MUC1 CAR was detailed in (31). The monoclonal antibody TAB004 was developed against tumor-associated form of human MUC1 (14). A Myc epitope-tagged framework was engineered to enable the monitoring of tMUC1-specific CAR expression.

To produce the tMUC1 CAR-encoding retroviral supernatants, GP2–293 cells were transfected with the SFG-retroviral CAR plasmid and VSV-G plasmid via Effectene following the manufacturer’s instruction (Qiagen, Germantown, MD). Virus supernatants were collected 48h and 72h after transfection and stored at -80°C.

Human PBMCs from various healthy donors (purchased from STEMCELL Technologies, Cambridge, MA) were activated with anti-CD3 (5μg/ml, plate-bound) andanti-CD28 (2μg/ml, soluble) antibodies (ThermoFisher Scientific, Waltham, MA). On day 3, activated T cells were transduced with the CAR retroviral supernatants and spinoculated at 32 °C for 1 hour at 2,000g in 6-well non-tissue culture plates coated with Retronectin (Takara, Mountain View, CA). Transduced T cells were maintained in recombinant human IL-2 (10ng/ml, PeproTech, Rocky Hill, NJ). On the next day, the culture supernatants were replaced with fresh complete RPMI1640 plus IL-2. The transduced T cells were passaged, and fresh media was supplied when appropriate. The CAR transduction efficiency was monitored by flow cytometry analysis. Activated but non-retroviral transduced cells were included as Mock T cell control. CAR T cells were used for functional assays on day 7–15 post transduction.

The human tMUC1-CAR expression was assessed via Myc-tag-AF488 staining (Cell Signaling Technologies, Danvers, MA). T cell phenotypes were determined by staining for CD4-PE/Cy7 and CD8-eF450 (BioLegend, San Diego, CA). Dead cells were excluded by 7-AAD staining (BD Biosciences). Data was acquired on BD LSRFortessa flow cytometer, and analyzed with FlowJo software

The antigen-specific tumor cell lysis by tMUC1-CAR T cells was determined using MTT assay. Briefly, a panel of PDA cell lines (10,000 cells/well) were plated in 96-well plates overnight. The next day, culture media from the plate was aspirated and CAR T cells or Mock T cells were added at the E:T ratio of 5:1. After co-culture for 24hr, T cells were washed with phosphate buffered saline (PBS) and removed and replaced with fresh media containing MTT (500μg/ml, Sigma) for 3hr. After the uptake of MTT, supernatants from wells were discarded and 150μl of DMSO was added to wells to dissolve the formazan crystals. The OD value was read at 540nm. When we used the mock T cell lysis data for calculating % lysis, the calculation formula was: [(OD of co-culture with Mock T cells – OD of co-culture with CAR T cells)/OD of co-culture with Mock T cells] ×100. When we used the media control data for calculating % lysis, the calculation formula was: [(OD of culture with media – OD of co-culture with CAR T or Mock T cells)/OD of culture with media] ×100. The exact applied formula was described in the relevant Figure Legends.

The CAR T cell lysis against PDA with Ruxolitinib (JAKi; R&D Systems; Minneapolis, MN) pretreatment or with IFN receptor knockdown were separately described in Figure Legends.

The antigen-specific cytokine release was performed by co-culturing tMUC1-CAR T cells with PDA cells at E:T ratio of 5:1. After 24hr of co-culture, supernatants were collected. The panel of cytokines was detected using Proteome Profiler Human Cytokine Array Kit, Panel A (R&D Systems; Minneapolis, MN) as instructed. The array data were quantitated by ImageJ software. The cytokine release under other treatment conditions was separately described in Figure Legends. The individual cytokine concentration was determined using Duoset Human Cytokine Detection Kit in accordance with the manufacturer’s instructions (R&D Systems).

Western blot (WB) was performed as previously described (16). Briefly, PDA cells were treated as detailed in Figure Legends. Cells were pelleted and suspended in the lysis buffer (20 mM Tris–HCl, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100), supplemented with Halt protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific) for 15min at 4°C. The cell lysates were centrifuged at 13,000 rpm for 5min at 4°C. Protein supernatants were collected and quantified using BCA assay (Thermo Fisher Scientific). The supernatants were boiled in the SDS sample buffer [50 mM Tris–HCl (pH 6.8), 2% SDS, 10% glycerol, 1.2% 2-mercaptoethanol and 0.02% bromphenol blue] for 5 min at 96°C. Equal amounts of proteins (10µg to 30µg per lane) were electrophoresed in 7%, 10%, or 12% polyacrylamide gel and then transferred to PVDF membranes (Bio-Rad, Hercules, CA, USA). The membranes were treated with 10% non-fat milk for 1hr to block non-specific binding, then rinsed and incubated with various antibodies. All primary antibodies purchased from Cell Signaling Technologies were diluted at 1:1000 (Cat# 35114T, Cat # 44902T, Cat# 53883, Cat# 94994S, Cat#9139S), except that the anti-IFNalpha/beta receptor 1 antibody was diluted at 1:2000 (Cat# ab245367; Abcam, Waltham, MA). The membranes were then treated with 1:1000 dilution of horseradish peroxidase-conjugated anti-rabbit or anti-mouse IgG (Cell Signaling Technologies) for 1hr. Immune complexes were detected with a chemiluminescence substrate (Cat# 34076; Thermo Fisher Scientific) and exposed using ChemiDoc MP Imaging System (Bio-Rad). The density of signal was quantified by ImageJ and presented along with the blots.

HPAC, Capan2, MiaPaCa-2 and HPAFII cells were plated at 100, 000 cells/well in 24-well plates and grown to 50-70% confluence. Cells were transfected with human IFN receptor-specific siRNA (Silence select IFNAR1/IFNAR2/IFNGR2) or control siRNA (Silencer Select Negative Control No. 2 siRNA) (Thermo Fisher Scientific) in Lipofectamine 3000 transfection reagent (Thermo Fisher Scientific), in accordance with the manufacturer’s instructions. The knockdown of IFN receptors was verified 48hr and 72hr after siRNA transfection.

On day 1, MiaPaCa-2 and HPAFII cells were plated overnight for attachment. On day 2, CAR T cells or Mock T cells were cultured with these two PDA cells at E:T ratio of 5:1 or cultured with media alone for overnight. On the same day, fresh MiaPaCa-2 and HPAFII cells were plated overnight for attachment. On day 3, the live CAR T cells and Mock T cells were isolated from the day-2 cultures using Lympholyte-M Cell Separation Media (Cedarlane, Burlington, NC). Thus, three pairs of live T cells were prepared: Mock T cells and CAR T cells from 1) culture with media alone 2); from culture with MiaPaCa-2 3); from culture with HPAFII. These three pairs of T cells were added to the fresh MiaPaCa-2 or HPAFII plates (prepared on day-2) at E:T ratio of 2:1 or 5:1 for 24hr. On day 4, CAR T cell lysis against MiaPaCa-2 or HPAFII was determined using MTT assay. The percentage of lysis was calculated using the formula: [(OD of co-culture with Mock T cells– OD of co-culture with CAR T cells)/OD of co-culture with Mock T cells] ×100. And the lysis was calculated using the Mock T cells and CAR T cells from the same pair.

MiaPaCa-2 and HPAFII cells were plated overnight for attachment. On the following day, CAR T cells were added to PDA culture at E:T ratio of 5:1 or cultured with media alone for overnight. Then CAR T cells from culture were collected and stained for Myc-tag-AF488, CD45-APC/Cy7, CD4-PE/Cy7, CD8-eF450, CD25-PE, PD-1-PE, TIM-3-PE, LAG-3-PE, TIGIT-APC, and 7-AAD. All the antibodies used here for flow cytometry were purchased from BioLegend, except for Myc-tag-AF488 and 7-AAD.

MiaPaCa-2 and HPAFII cells were plated overnight for attachment. On the following day, PDA cells were treated with CAR T cells at E:T ratio of 5:1 or cultured with media alone for overnight. After gently rinsing off suspension cells with PBS in co-culture, adherent MiaPaCa-2 or HPAFII cells were collected and stained with CD45-APC/Cy7, TAB004-Fluor 647, PD-L1-PE (BioLegend), and 7-AAD. CD45 staining was used to exclude non-tumor cells from analysis.

MiaPaCa-2 and HPAFII cells were plated overnight for attachment. On the next day, PDA cells were treated with JAKi at 25μM for overnight. Then JAKi was removed, and CAR T cells were added in the absence of JAKi. CAR T cells were added at E:T ratio of 5:1, calculated based on the initial number of PDA cells plated. After 24hr treatment with CAR T cells, live adherent PDA cells were collected and stained with CD45-APC/Cy7, PD-L1-PE, and 7-AAD.

HPAFII cells were plated overnight and followed by CAR T cell treatment at E:T ratio of 5:1 for another overnight. After gently rinsing off suspension cells with PBS, adherent HPAFII cells were collected, resuspended in flow cytometry buffer, and aliquoted into 100μl cell suspension per staining tube. PD-L1 blocking antibody (Cat# 329716; BioLegend) were added at 10μg/ml or 100μg/ml in the respective tubes and incubated for 30min at 4 °C. The control tube was added equal volume of PBS. After 30min incubation, cells were washed with flow cytometry buffer to remove unbound PD-L1 blocking antibody. Then cells were resuspended and stained with CD45-APC/Cy7, PD-L1-PE, and 7-AAD.

MiaPaCa-2 and HPAFII cells were plated overnight for attachment. On the following day, PDA cells were pre-incubated with PD-L1 blocking antibody (10μg/ml) or its isotype control antibody (10μg/ml; Cat# 400348; BioLegend), anti-ICAM-1 neutralizing antibody (5μg/ml; Cat# AF720; R&D Systems), anti-CXCL10 neutralizing antibody (5μg/ml; Cat# MAB266-100; R&D Systems), or fresh media alone for 2hr. Then without removing antibodies, CAR T cells were added at E:T ratio of 5:1 or cultured with media alone for 24hr. MiaPaCa-2 or HPAFII cell lysis was determined using MTT assay. The percentage of lysis was calculated using the formula: [(OD of co-culture with Mock T cells– OD of co-culture with CAR T cells)/OD of co-culture with Mock T cells] ×100.

RNA-seq gene count data in the TCGA (The Cancer Genome Atlas) pancreatic adenocarcinoma project were downloaded from the GDC (Genomics Data Commons Portal) (32, 33). A total of 183 pancreatic tumor samples were included in the analysis. Gene expression was calculated as TPM (transcripts per million) from STAR alignment counts (34). For each gene, samples are considered to have low expression if the value was below the median gene expression and high expression if above. Survival curves were generated from the resulting dataset using the survival package in R and significant differences between the two groups were also calculated using the default parameters according to the documentation (35, 36).

Data were analyzed using Prism (version 10; GraphPad Software) and results were presented as mean ± SD or mean ± SEM. For all the cytotoxicity assay, individual values were also included in figures and shown using symbols. Data were representative of two or more independent experiments. The calculation of significance was described in Figure Legends.

The expression of tMUC1-CAR on activated human T cells was shown in Supplementary Figure 1 . On day 7 post-retrovirus transduction, 32.5% of CD4⁺T cells and 16.3% of CD8⁺T cells were Myc-tag-positive. The whole transduced T cells were used for this study.

First, human PDA cell lines were analyzed for the levels of tMUC1 on their cell surface by flow cytometry. Data were presented as percentage of cells that expressed tMUC1 ( Figure 1A ). PDA cells were categorized into two groups based on their tMUC1 expression levels: tMUC1-high and tMUC1-moderate expressing cells. Next, the efficacy of CAR T cell cytolysis against PDAs was shown as Figure 1B . PDA cells were co-cultured with CAR T cells at a E:T ratio of 5:1 for 24hr. The tumor lysis data of CAR T cells was normalized to the corresponding mock T cell control for each PDA cell line. CAR T cells were able to kill tMUC1-positive PDA cell lines ( Figure 1B ). However, the sensitivity of PDAs to CAR T cell-mediated killing varied and was not strictly correlated with their tMUC1 levels. Among the tMUC1 high-expressing PDAs, Capan2 exhibited significantly lower sensitivity to CAR T cell killing compared to HPAC. Similarly, within the tMUC1 moderate-expressing PDAs, HPAFII displayed strong resistance to CAR T cell killing compared to MiaPaCa-2. Overall, PDA cell lines demonstrated variable resistance to CAR T cell cytolysis.

To investigate the mechanisms underlying the strong resistance observed in HPAFII cells, we co-cultured the highly sensitive MiaPaCa-2 cells and the highly resistant HPAFII cells with CAR T cells at a E:T ratio of 5:1 for 24hr. Following the co-culture, supernatants were analyzed using a human cytokine array to identify differentially released cytokines and chemokines ( Figure 2 ), in which the focus was given to what was higher in the co-culture with HPAFII cells. We observed significantly higher levels of the adhesion molecule ICAM-1 and the chemokines CXCL10 and CXCL11 in the HPAFII-CAR T cell co-culture compared to the MiaPaCa-2-CAR T cell co-culture. ICAM-1, CXCL10, and CXCL11 are predominantly induced by IFN-γ (37–39), though they can also be triggered by type I IFNs (40). The elevated release of these IFN-related factors in the HPAFII-CAR T cell co-culture suggested a potential correlation between higher tumor-associated IFN activation and stronger PDA resistance to CAR T cell killing.

Ruxolitinib, a selective JAK1/2 kinase inhibitor known to suppress JAK-STAT signaling (41), was used to assess the role of tumor-intrinsic IFN signaling in the elevated release of ICAM-1 and CXCL10. The cytotoxic effect of Ruxolitinib on PDA cells was first determined ( Supplementary Figure 2 ). To investigate the impact of IFN signaling within PDA cells, we pre-treated PDA cells with or without Ruxolitinib (briefly as JAKi, at 10μM or 25μM) overnight. Following treatment, JAKi was removed, and CAR T cells were added for co-culture with PDA cells for 24hr. Analysis of the co-culture supernatants revealed a significant reduction in ICAM-1 and CXCL10 levels when PDA IFN signaling was pre-disrupted, with the exception of ICAM-1 in the highly sensitive MiaPaCa-2 cell line ( Figures 3A, B ).

Since T cells are also known to produce ICAM-1 and CXCL10 upon stimulation (42–44), we further examined their contribution to cytokine release. CAR T cells were pre-treated with or without JAKi (10 µM or 25 µM) overnight, thoroughly washed to remove residual JAKi, and then co-cultured with PDA cells for 24hr. As shown in Figure 4 , pre-disrupting IFN signaling in CAR T cells only mildly reduced ICAM-1 levels in co-cultures with three PDA cell lines.

Therefore, data from Figures 3 and 4 indicated that PDA cells were the primary source of ICAM-1 and CXCL10 in the co-cultures. The reduction in these cytokines following JAKi treatment highlighted the critical role of tumor-intrinsic IFN signaling in their production.

Data from Figure 3 indicated a potential role for IFN signaling in PDA cytokine release. Building on this observation, we further investigated the presence of IFN signaling cascades within PDA cells and their regulation by IFN blockade in response to CAR T cell cytolysis. The four PDA cell lines were pretreated with or without JAKi (at 25μM) overnight. The following day, JAKi was removed, and PDA cells were co-cultured with CAR T cells for 3 hours. Adherent live PDA cells were then collected for Western blot analysis, and data was summarized in Figure 5 . The sensitivity difference of PDA to CAR T lysis was restated in Figure 5A for signaling reference. Notably, MiaPaCa-2 cells displayed the weakest IFN signaling cascade expression. Treatment with JAKi effectively reduced baseline levels and/or CAR T cell-induced activation of type I and type II IFN signaling components, including phosphorylated STAT1 (p-STAT1 at Ser 727), phosphorylated STAT2 (p-STAT2 at Tyr690), and interferon regulatory factor 9 (IRF-9). Interestingly, IRF-9 and p-STAT2/STAT2 levels were endogenously higher in the resistant HPAFII and Capan2 cell lines, as well as in the less sensitive HPAC cells, compared to MiaPaCa-2 cells ( Figure 5B ). IRF-9, a transcription factor that partners with activated STAT1/2 to mediate type I IFN responses and downstream IFN-stimulated genes expression (45), together with p-STAT2, was dramatically suppressed by JAKi in PDA cells upon CAR T cell challenge.

Phosphorylated Jak1 (p-Jak1) and phosphorylated Jak2 (p-Jak2) were undetectable at this time point ( Supplementary Figure 3 ). Interestingly, an increase in phosphorylated Tyk2 (p-Tyk2) was observed following JAKi treatment. This may result from JAKi’s inhibition of Jak1 phosphorylation, leading to the accumulation of unpaired p-Tyk2, which, without its typical pairing with p-Jak1, may hinder further cycling and dephosphorylation ( Supplementary Figure 3 ).

Building on data from Figures 3 and 5 , we further investigated the functional role of PDA-intrinsic type I/II IFN signaling in modulating PDA sensitivity to CAR T cell cytolysis. To assess this, PDA cells were pretreated with JAKi (10 µM or 25 µM) overnight to suppress IFN signaling, followed by co-culture with CAR T cells. As shown in Figure 6 , JAKi pretreatment at 25 µM significantly enhanced the sensitivity of all four PDA cell lines to tMUC1-targeted CAR T cell killing. Interestingly, we also observed an increase in tumor lysis by non-targeting Mock T cells in JAKi-pretreated PDAs. JAKi alone (at both 10 µM and 25 µM) had minimal to no impact on PDA cell survival in this assay (data not shown). To evaluate the potential contribution of CAR T cell-intrinsic IFN signaling, CAR T cells were pretreated with JAKi overnight before being used to lyse PDA cells. However, this treatment did not significantly enhance PDA killing ( Supplementary Figure 4 ). Together, these findings suggested that PDA-intrinsic IFN signaling might play an essential role in PDA resistance to CAR T cell-mediated cytolysis. Suppressing this signaling pathway effectively enhanced PDA susceptibility to both targeted and non-targeted T cell cytotoxicity.

To further validate the findings from Figure 6 , where pharmacological IFN blockade improved PDA sensitivity, we used siRNA to specifically knock down type I (IFNAR1 and IFNAR2) or type II (IFNGR2) IFN receptors in four PDA cell lines. The efficiency of siRNA knockdown was evaluated, as shown in Supplementary Figure 5 . Following 48 hours of siRNA treatment, the siRNAs were removed from the PDA cultures, and CAR T cells were subsequently added for 24 hours. Consistent with the results observed using JAKi, siRNA-mediated downregulation of IFN receptors significantly enhanced PDA sensitivity to CAR T cell killing ( Figure 7 ). These findings further supported the critical role of PDA-intrinsic IFN signaling in promoting resistance to CAR T cell-mediated cytotoxicity and highlighted IFN receptor targeting as a potential strategy to improve CAR T cell therapy.

We then used the sensitive MiaPaCa-2 cells and resistant HPAFII cells to further look into the downstream molecules post IFN signaling that could be responsible for PDA resistance. We first looked at CAR T cell changes in their function and phenotype. CAR T cells were retrieved from their 1^st round co-culture with PDA cells and then the live CAR T cells were used in the 2^nd round co-culture with PDA cells to evaluate their cytotoxicity function. Compared to the cytotoxicity of CAR T cells which were maintained in culture media, the lysis activity of CAR T cells recovered from their prior co-culture with MiaPaCa-2 or with HPAFII was significantly compromised, especially at lower E:T ratio of 2:1 ( Figure 8A ). Notably, the cytotoxicity of CAR T cell recovered from their co-culture with HPAFII was impaired the most.

To determine whether the suppression of cytotoxicity in CAR T cells after their co-culture with PDA was due to the lack of CAR T cell activation, CAR T cells from co-culture were analyzed for the level of CD25, a T cell activation marker. Compared to the CAR T cells cultured in media only, the CD25 expression was increased in CAR T cells after their culture with either MiaPaCa-2 or HPAFII cells, both in CAR-positive or CAR-negative populations in CD4 T cells or in CD8 T cells ( Figure 8B ).

We then analyzed the levels of immune checkpoint (IC) molecules on CAR T cells, including PD-1, TIM-3, LAG-3, and TIGIT. The presence of MiaPaCa-2 or HPAFII in culture did not significantly affect PD-1 and TIGIT expressions, but their presence increased the TIM-3 and LAG-3 expressions on CAR T cell surface when compared to CAR T cell cultured in media only ( Figure 8C ). From the same co-culture, we also analyzed the expression of tMUC1 and the IC molecule PD-L1 on PDA cells. There was no clear loss in tMUC1expression on cell surface of remaining live MiaPaCa-2 or HPAFII after their treatment with CAR T cells. However, there was dramatic up-regulation of PD-L1 expression on PDA cell surface after engagement with CAR T cells ( Figure 8D ). As we mentioned above in Supplementary Figure 4 , pre-disruption of IFN signaling in CAR T cells did not significantly affect their cytotoxicity against PDA cells. Therefore, even though there was an increase in TIM-3 and LAG-3 on CAR T cells after their engagement with PDA cells, we preferred to manipulate the ICs from the PDA tumor side, in another word, on PD-L1 molecule.

To dissect the association of CAR T cell-triggered PD-L1 up-regulation with tumor intrinsic IFN signaling-induced PDA resistance, PD-L1 was analyzed in PDA cells which were pretreated with JAKi for overnight before they were challenged with CAR T cells. As shown in Figure 8E , CAR T cell-induced PD-L1 increase was greatly revoked in JAKi-pretreated MiaPaCa-2 and HPAFII cells. We then determined the functional contribution of PD-L1 increase to PDA resistance. The efficiency of PD-L1 blocking antibody was confirmed using a competition binding assay. PD-L1 blocking antibody at 10μg/ml or 100μg/ml were able to effectively block the binding sites of tumor cell surface PD-L1 ( Figure 8F ). In the cytotoxic function assay, PD-L1 blocking antibody partially rescued the CAR T cell killing in HPAFII cells ( Figure 8G , bottom panel). PD-L1 blocking antibody did not further enhance CAR T lysis against MiaPaCa-2 since CAR T cells at E:T ratio of 5:1 already exerted a near maximal killing in MiaPaCa-2.

As shown in the above figures, the levels of released ICAM-1 and CXCL10 were much higher in resistant HPAFII cells. JAKi pretreatment strongly inhibited their production by HPAFII and other 3 PDA cell lines and at the same time enhanced those PDA cell response to CAR T cell lysis. We neutralized the ICAM-1 and CXCL10 during the PDA cell co-culture with CAR T cells. There was a trend of increased response of HPAFII to CAR T cell killing, with CXCL10 neutralization being significant ( Figure 8H , bottom panel). Similar to PD-L1 blocking assay, neutralizing ICAM-1 or CXCL10 did not affect CAR T lysis against MiaPaCa-2 at the given condition.

Finally, to assess the clinical relevance of our findings, we chose several essential signaling cascade molecules along the IFN pathway and analyzed the levels of their gene expression in predicting the survival of patients with PDA. The RNA-seq gene count data in TCGA pancreatic adenocarcinoma project was utilized and gene expression was calculated and grouped as low expression group and high expression group. As shown in Figure 9 , compared to the patients with lower gene expression, patients with higher expression of those selected genes were likely to have poorer survival probability, with JAK1 and STAT1 showing the statistically significant roles.