Patients with pancreatic cancer were recruited from the Osaka University Hospital (Osaka, Japan) and Osaka International Cancer Institute (Osaka, Japan) from September 2019 to December 2021. All patients received neoadjuvant chemotherapy or chemoradiotherapy with the following regimens: gemcitabine + nab-paclitaxcel (n=9), gemcitabine + nab-paclitaxcel + radiation (n=7), gemcitabine + S-1 (n=20), gemcitabine + S-1 + radiation (n=13), FOLFIRINOX (n=1), gemcitabine + radiation (n=1), and multiple regimens (n=6). This study was approved by the local institutional ethics committees of Osaka University, Osaka International Cancer Center, and the National Institutes of Biomedical Innovation, Health, and Nutrition, Osaka, Japan, and was conducted in accordance with the Declaration of Helsinki (1975). All participants provided written informed consent. PBMCs were isolated from patients with PDAC within 6 h of blood sampling using BD Vacutainer CPT (BD Biosciences, Franklin Lakes, NJ, USA). PBMCs were cryopreserved in fetal bovine serum (FBS) containing 10% dimethyl sulfoxide (DMSO) and stored in liquid nitrogen vapor until analysis.

Pancreatic cancer cell lines were cultured in an appropriate culture medium supplemented with 10% FBS and 1% penicillin–streptomycin (Gibco, Thermo Fisher Scientific, Waltham, MA, USA). Dulbecco’s modified Eagle medium (Sigma-Aldrich, St. Louis, MO, USA) was used for BxPC3, MiaPaCa2, Panc1, and PSN1 cells, and Eagle’s minimal essential medium (Sigma-Aldrich) was used for SUIT-2 cells. TYPK1 cells were cultured in a 1:1 mixture of DMEM and Ham’s F-12 medium supplemented with 5% FBS and 1% penicillin–streptomycin. Cells were cultured at 37°C and 5% CO₂ until reaching 80% confluency, at which point mRNA was extracted using an RNeasy mini kit (Qiagen, Hilden, Germany).

qRT-PCR was performed using SuperScript III Master Mix (Invitrogen, Thermo Fisher Scientific), RT enzyme, ROX, and TaqMan probes for candidate genes (CEACAM5, Hs00944025_m1; CTAG1, Hs00265824_m1; DCT, Hs01098278_m1; MUC1, Hs00159357_m1; Telomerase reverse transcriptase (TERT), Hs00972656_m1; and WT1, Hs00240913_m1). The following thermal cycling conditions were used: 30 min at 45°C, 5 min at 95°C, 50 cycles of 1 s at 95°C, and 60 s at 50°C. qRT-PCR was replicated three times.

The CCLE (14) was used to evaluate the mRNA expression of six TAAs (CEACAM5, NY-ESO-1 [CTAG1A], TRP2 [DCT], MUC1, TERT, and WT1) in the pancreatic cancer cell lines. The normalized transcripts per kilobase million (TPM) dataset were downloaded from the Cancer Dependency Portal (DepMap) on 2022.12.14. The original log2 (TPM+1) values were plotted as a heatmap without scaling.

Overlapping peptides covering four TAAs (all from JPT Peptide Technologies, Berlin, Germany) were used in this study: PepMix™ Human CEA (#PM-CEA) for CEACAM5; PepMix™ Human Mutin-1 (#PM-MUC1) for MUC1; PepMix™ Human TERT (#PM-TERT) for TERT; and PepMix™ Human WT1/WT33 (#PM-WT1) for WT1. The peptides consisted of 15 amino acids spanning the complete amino acid sequence of the indicated protein antigen, with 11 overlapping amino acids between adjacent peptides ( Supplementary Table 1 ).

TAA-specific CD8⁺ T cells were enriched from the PBMCs of patients with PDAC. Briefly, frozen PBMCs were thawed and treated with 1 mL of 50 unit/mL of benzonase (Merck, Rahway, NJ, USA) in R10 medium for 2 h at 37°C in 5% CO₂. Subsequently, 20% of the PBMCs (1.1×10⁵ [0.2×10⁵ – 3.5 ×10⁵] cells) were pulsed with 2 μg/mL of each of the overlapping TAA peptides (CEA, MUC1, TERT, and WT1) for 1 h at 37°C. After washing, the pulsed PBMCs were co-cultured with the remaining 80% un-pulsed PBMCs in R10 in the presence of 20 U/mL of recombinant interleukin (IL)-2 (R&D Systems, Minneapolis, MN, USA) for 10 days; the medium was changed on days 4 and 7. For the ICI experiments, 1 mg/mL of anti-PD-1 (Cat# 621604, RRID: AB_2820105; BioLegend, San Diego, CA, USA), anti-TIGIT (Cat# 16-9500-82, RRID: AB_10718831; Thermo Fisher Scientific), or mouse IgG1 isotype control antibody (Cat# 400102, RRID: AB_2891079; BioLegend) was added to the culture medium at the beginning of the culture, and half of the medium was replaced with a culture medium without antibodies on days 4 and 7.

Ten days after culture, amplified cells were stimulated with 2 μg/mL of the TAA-derived overlapping peptides (CEA, MUC1, TERT, and WT1) for 30 min at 37°C with CD107A antibody (BD Biosciences). After 30 min of stimulation, 1 μL/mL BD GolgiPlug (containing Brefeldin A) (BD Biosciences) and 0.7 μL/mL of BD GolgiStop (containing Monensin) (BD Biosciences) were added to the media, and the cells were cultured for another 5.5 h. After incubation, the cells were washed with phosphate-buffered saline (PBS) and stained using a Live/Dead Fixable Aqua Dead Cell Stain Kit (L34957; Thermo Fisher Scientific) at 18-25°C for 5 min. Subsequently, the samples were probed with antibodies ( Supplementary Table 2A ) against cell surface markers for 15 min at 18-25°C. After washing, the cells were thoroughly resuspended in 200 µL BD Cytofix/Cytoperm solution per well and incubated for 20 min at 18-25°C. The cells were washed twice with BD Perm/Wash buffer and probed with intracellular staining antibodies for 25 min at 18-25°C. After staining, the cells were washed twice with BD Perm/Wash buffer and fixed with 1% paraformaldehyde. The data were collected using FACSymphony A5 (BD Biosciences).

Cryopreserved PBMCs were thawed, and genomic DNA was extracted from a portion of cells (approximately 100,000 cells) using a QIAamp DNA mini kit (Qiagen) and stored at −30°C until use. Isolated genomic DNA was used as a template to prepare cDNA libraries for HLA typing using the commercially available kit AlloSeq™ Tx 17.1 (CareDx) or WAKFlow® HLA DNA Typing (Wakunaga Pharmaceutical Co. Ltd., Osaka, Japan). The combination of each HLA was visualized using the “circlize” package (version 0.4.15) in the R software language (version 4.2.1). The frequency of HLA-A types among the donors was calculated based on the cumulative total number of alleles.

Cryopreserved PBMCs were thawed, washed with PBS, and stained with Fixable Viable Stain UV440 (BD Biosciences) at 18-25°C for 5 min. CC-chemokine receptor 7 (CCR7) was stained at 37°C for 10 min and probed using antibodies against the remaining markers ( Supplementary Table 2B ) at 18-25°C for 15 min. After washing, the cells were thoroughly resuspended in 500 µL Fix/Perm solution per tube and incubated for 40 min at 4°C. The cells were washed twice with BD Perm/Wash solution and probed with anti-Ki67 antibodies for 40 min at 4°C. After staining, the cells were washed twice with BD Perm/Wash solution and fixed with 1% paraformaldehyde, and the data were collected using FACSymphony A5 (BD Biosciences).

Flow cytometry FCS files were analyzed using FlowJo software (version 10.8.1; RRID: SCR_008520; BD Biosciences). The gating setting for settings for each population are described in result section and Figures. TAA-specific CD8⁺ T cell responses were determined by subtracting the value obtained by peptide-free stimulation (DMSO; background) from that obtained by TAA stimulation. After background subtraction, values less than 0.01% were considered negative (no response). For the IFNγ⁺4-1BB⁺ criteria, values greater than 0.01% after background subtraction corresponded with responders, while those with values less than 0.01% corresponded with non-responders. For the IFNγ⁺ and/or 4-1BB⁺ criteria, if the sum of the values after the background subtraction of IFNγ⁺4-1BB⁺, IFNγ⁺4-1BB^-, and IFNγ^-4-1BB⁺ was greater than 0.03%, it corresponded with a responder, while if it was less than 0.03%, it corresponded with a non-responder. For the IFNγ⁺ and 4-1BB⁺ criteria, if the sum of the values after the background subtraction of two gates in each marker positive cells was greater than 0.02%, it corresponded with a responder, while if it was less than 0.02%, it corresponded with a non-responder.

Statistical analyses were performed using R/Bioconductor (R version 4.2.1) or GraphPad Prism (version 9.0.0; GraphPad Software, RRID: SCR_002798). Experiments and data analysis were performed by individuals blinded to the collection of blood samples and clinical information. For the HLA analysis, statistical significance of all combinations of four gating sets in each HLA type were obtained using Fisher’s exact test, and are displayed in tile format.

To detect major TAA-specific PBMC CD8⁺ T cell responses in patients with PDAC, we selected TAAs based on their expression in PDAC cells. Of the 403 TAAs registered in the TAA database (TANTIGEN 2.0) (15), we selected four candidates that have been utilized in peptide vaccine clinical trials: CEACAM5 [carcinoembryonic antigen (CEA)] (16), MUC1 (17), TERT (18), and WT1 (19). Additionally, we selected two candidates used in peptide vaccine clinical trials for other cancers, NY-ESO-1 (CTAG1A) and TRP2 (DCT).

Using the RNAseq data of 1404 cell lines from the CCLE, we assessed the mRNA expression of six TAAs. Of these, CEACAM5, MUC1, TERT, and WT1 were expressed in multiple PDAC cell lines, whereas CTAG1A and DCT were not expressed ( Figure 1A and Supplementary Figure 1 ). Moreover, mRNA expression was further confirmed through qRT-PCR in six PDAC cell lines (four from primary tumors, one from liver metastasis, and one from lymph node metastasis), revealing upregulation of the expression of CEACAM5, MUC1, TERT, and WT1 in several cell lines ( Figures 1B, C ).

We then detected TAA-specific CD8⁺ T cell responses in PBMCs derived from patients with PDAC at treatment initiation. As these cells are known to have a low frequency (20), to increase detection sensitivity, we stimulated PBMCs with overlapping peptides covering the full length of the four TAAs ( Supplementary Table 1 ) and then cultured them in the presence of IL-2 for 10 days to amplify TAA-specific cells (21). Moreover, we aimed to increase detection sensitivity by stimulation with a mixture of four different TAAs. After culture, we re-stimulated the cells with the peptide pool and detected TAA-specific CD8⁺ T cell responses using a flow cytometer. To evaluate antigen-specific responses, we measured the expression levels of interferon-gamma (IFNγ), a widely used marker for antigen-specific CD8⁺ T cell responses, and 4-1BB, which is known as an activation-induced marker. IFNγ and 4-1BB have each been used as markers of antigen-specific responses (22, 23), and the antigen-specific response of the two in combination has been evaluated (24). However, it was unclear whether the and/or case, single positive, or double positive is more useful in controlling cancer, so we compared them in this study based on four criteria: IFNγ⁺ and/or 4-1BB⁺, IFNγ⁺, 4-1BB⁺, and double-positive (IFNγ⁺4-1BB⁺) ( Figures 2A, B ). Although there was a certain amount of bulk T cell amplification due to culture in the presence of IL-2, there was no increase in TAA-specific CD8⁺ T cell responses due to TAA stimulation in healthy donors ( Supplementary Figure 2 ).

To investigate the relationship between TAA-specific CD8⁺ T cell responses and clinical features of patients with PDAC (n=57), we stimulated PBMCs with a peptide pool containing four TAAs and determined the percentage of responders for TAA-specific CD8⁺ T cell response using the four criteria described above ( Figure 2C and Supplementary Figure 3 ).

We subsequently investigated if these cell responses were affected by specific HLA class 1 types. Among the 57 donors, 12 HLA-A types were detected with 25 allelic combinations ( Supplementary Figures 4A, B ), and TAA-specific CD8⁺ T cell responses were observed for all HLA-A types except A*30:01 ( Supplementary Figure 4C ). We then compared the HLA distribution of responders in each criterion to that of all donors to determine whether TAA-specific CD8⁺ T cell responses were more prevalent in specific HLA-A types. However, we detected no differences in our cohort ( Supplementary Figures 4D, E ). Larger cohorts should be employed to investigate the association between HLA and TAA-specific responses.

We investigated the correlation between TAA-specific CD8⁺ T cell responses and patient characteristics at the initiation of neoadjuvant therapy. The analysis revealed a significant association between TAA-specific IFNγ⁺4-1BB⁺ CD8⁺ T cell response and pancreatic head and pancreatic body tail cancer (p = 0.0307; Supplementary Table 3 ). However, no significant correlations were observed for basic patient characteristics, such as age, sex, blood counts, and tumor factors ( Supplementary Table 3 ). Moreover, there were no significant differences in the presence or absence of TAA-specific responses by treatment content ( Supplementary Table 3 ).

To more comprehensively assess the relationship between clinical background information and total TAA-specific CD8⁺ T cell responses, we generated a new index that may reflect the effects of neoadjuvant therapy. The index comprised ratios and differences of three tumor markers (CA19-9, CEA, and DUPAN-2), along with the tumor diameter on computed tomography (CT) images at the initiation of treatment, after neoadjuvant therapy, and after surgery. We then analyzed the parameters that were correlated with the frequency of TAA-specific CD8⁺ T cell responses based on the four criteria.

The number of lymphocytes in the peripheral blood was positively correlated with TAA-specific 4-1BB⁺ CD8⁺ T cell and TAA-specific IFNγ⁺4-1BB⁺CD8⁺ T cell frequencies ( Figure 2D , columns 3 and 4). TAA-specific CD8⁺ T cell frequency, characterized by IFNγ⁺ and/or 4-1BB⁺, was inversely correlated with both clinical progression (cStage) and tumor diameter on CT before treatment initiation. However, no inverse correlation was observed between the pathological progression (pStage) of the resected specimen and the difference in tumor diameter on CT after neoadjuvant therapy ( Figure 2D , column 1). In contrast, the frequency of TAA-specific IFNγ⁺4-1BB⁺ CD8⁺ T cell responses showed an inverse correlation with the rate of CA19-9 change during neoadjuvant therapy: the lower the rate, the better the therapeutic effect (R = −0.56, p = 0.011; Figure 2D , column 4 and Figure 2E ). CA19-9 is a widely used serum biomarker in PDAC, and changes in its levels during neoadjuvant therapy are considered to be prognostic (25). The results suggest that patients with TAA-specific IFNγ⁺4-1BB⁺ CD8⁺ T cell response observed prior to treatment initiation had a higher response to neoadjuvant therapy. Moreover, this criterion resulted in the lowest background; therefore, we used it for subsequent analyses.

To investigate the prognostic impact of TAA-specific IFNγ⁺4-1BB⁺ CD8⁺ T cells at different time points (before treatment [TP1], after neoadjuvant therapy but before resection [TP2], and after resection [TP3]) using PBMCs collected from the same patients ( Figure 3A ), we compared the postoperative disease-free survival in the two groups according to the presence or absence of TAA-specific IFNγ⁺4-1BB⁺ CD8⁺ T cells. There was no difference in the number of responders at each time point ( Figure 3B ). No difference was observed before treatment [TP1] or before surgery [TP2], whereas a significant difference was observed after surgery [TP3] ( Figure 3C ). The modulation of TAA-specific IFNγ⁺4-1BB⁺ responses over time did not exhibit a consistent trend, with some patients exhibiting a loss of response and others becoming new responders before and after treatment ( Supplementary Figure 4 ). Notably, the four patients who had responded before treatment and maintained this response after surgery remained recurrence-free, whereas the three patients exhibiting loss of TAA-specific IFNγ⁺4-1BB⁺ responses experienced recurrence (data not shown). There were no significant differences in clinicopathologic factors (including treatment regimen) between the presence or absence of TAA-specific IFNγ⁺4-1BB⁺ reactions in TP3 ( Supplementary Table 4 ).

We investigated the impact of ICIs on TAA-specific IFNγ⁺4-1BB⁺ CD8⁺ T cell response. To identify candidate immune checkpoint molecules as potential therapeutic targets, we compared the ex vivo expression profiles of PD-1, TIGIT, Tim-3, CD160, and BTLA on CD8⁺ Tm cells in PBMCs derived from patients with PDAC via flow cytometry ( Figure 4A ; Supplementary Figure 6 ). Among these, only PD-1 was significantly more expressed in samples derived from patients with PDAC than in those from healthy participants ( Figure 4B ). Therefore, we selected PD-1 as the target molecule. We stimulated PBMCs with TAA peptide pools in the presence of ICIs and cultured them to analyze TAA-specific IFNγ⁺4-1BB⁺ CD8⁺ T cells. We identified more responders in the anti-PD-1 antibody group than in the no-antibody and isotype groups, although this observation was not significant ( Figure 4C ; Supplementary Figure 7 ), suggesting that anti-PD-1 treatment detected new TAA-specific IFNγ⁺4-1BB⁺ CD8⁺ T cells in some non-responders. Furthermore, anti-PD-1, but not anti-TIGIT, treatment significantly increased TAA-specific IFNγ⁺4-1BB⁺ CD8⁺ T cell frequency ( Figures 4C–E ).

To date, clinical trials of ICIs in patients with PDAC have included patients treated with cytotoxic anticancer drugs. Thus, we next investigated whether chemotherapy, radiation therapy, or surgery would affect the efficacy of ICI in patients with PDAC. PD-1 inhibition was assessed at the TP1, TP2, and TP3 time points using PBMCs collected from the same patients. The efficacy of ICI in donors increased as follows: 50% at TP1, 25% at TP2, and 35% at TP3 ( Figure 5A ). PD-1 inhibition significantly increased the frequency of TAA-specific IFNγ⁺4-1BB⁺ CD8⁺ T cells only at TP1 ( Figure 5B ), whereas TIGIT inhibition did not result in significant changes at any time point ( Figure 5C ). These results suggest that anti-PD-1 ICIs may be less effective after chemotherapy, either alone or in combination with radiotherapy, which could explain the poor outcomes reported in previous clinical trials.