TIGIT Expression Delineates T-cell Populations with Distinct Functional and Prognostic Impact in Pancreatic Cancer

This article is featured in Highlights of This Issue, p. 2559

Pancreatic ductal adenocarcinoma (PDAC) has a dismal prognosis with a 5-year overall survival (OS) rate of only 10% (1). Surgical resection remains the only potentially curative treatment for PDAC (2). Although immunotherapies targeting the immune checkpoint receptor (ICR) programmed cell death protein 1 (PD-1) and its ligands, as well as CTLA-4 successfully improved the treatment of numerous solid tumors (3), single ICR-inhibition has failed to show benefits over standard chemotherapy in PDAC (4, 5). Poor immunogenicity and a highly immunosuppressive tumor microenvironment must be overcome (6). Several studies demonstrated a predictive value of lymphocyte infiltration in PDAC, highlighting a therapeutic potential (7, 8). High tumor infiltration by CD8⁺ and effector CD4⁺ T cells and proximity of CD8⁺ T cells to tumor cells characterized patients with prolonged survival (9–11). Furthermore, the composition of T-cell infiltrate is associated with patient prognosis, as an increased proportion of T cells with immunosuppressive capabilities, namely regulatory T cells (Treg) and Th2, correlates with poor outcomes (12–14). Although T-cell infiltration has been widely characterized, few studies addressed the relationship between ICR expression, T-cell functionality, and clinical outcome in PDAC. Previously, we showed that the expression of lymphocyte-activation gene 3 (LAG-3) by intratumoral T cells is a prognostic factor for disease-free survival (DFS) in PDAC (15). In addition, high expression of PD-1 on intratumoral CD8⁺ T-cells detected by IHC was associated with poor survival (16). Although the blockade of the PD-1/PD-L1–axis was successful in PDAC mouse models (17), clinical trials have failed to demonstrate benefits in patients with PDAC (18, 19). A detailed understanding of the role of PD-1 expression by intratumoral T cells and an in-depth understanding of the interplay between different ICR are required to identify new targets and develop novel combinational immunotherapeutic strategies in PDAC. Previous studies indicate the involvement of T-cell immunoreceptor with Ig and immunoreceptor tyrosine-based inhibition motif (ITIM) domains (TIGIT) and its ligands in tumor immune escape of PDAC. TIGIT was identified as a central marker of exhausted CD8⁺ T cells in PDAC (20). In a neoantigen-expressing PDAC mouse model, expression of TIGIT on CD8⁺ T cells and its ligand CD155 by tumor cells was shown as a mechanism of PDAC immune evasion (21). In colorectal (22), bladder cancer (23), and melanoma (24), expression of TIGIT by CD8⁺ T cells or Treg was associated with worse clinical outcomes. Therefore, we investigated the expression and prognostic relevance of TIGIT and PD-1 by multicolor flow cytometry–based phenotyping of freshly isolated T cells from 144 blood and 107 matched PDAC samples. In addition, we analyzed cytokine expression, activation and differentiation status to decipher the association of TIGIT and PD-1 expression with T-cell functionality and exhaustion.

Fresh blood and tumor samples were obtained from patients with PDAC who underwent surgery at our institution between 2018 and 2023. The study was conducted under the principles of the Declaration of Helsinki and all patients gave written informed consent to a protocol approved by the Ethics Committee of the TU Dresden (No. EK446112017). Blood was drawn before the surgical incision. A trained pathologist determined the clinical stages of the tumors according to the tumor–node–metastasis TNM classification system (UICC; Edition 8). Summarized clinical characteristics are shown in Supplementary Table S1.

Peripheral blood mononuclear cells (PBMC) were isolated by density centrifugation (800 × g, 25 minutes, 18°C) from blood diluted with PBS (1:2) over Pancoll (PAN-biotech). PDAC tissue, usually between 0.1 and 1.5 g, was identified as tumor by a trained pathologist and freshly obtained right after surgery. It was manually chopped with a surgical scissor for 30 minutes before being digested in PBS containing 1 mg/mL trypsin inhibitor (Merck), 1 mg/mL collagenase type IV (Thermo Fisher Scientific), and 20 IU/mL deoxyribonuclease I (Merck) for 30 minutes at 37°C while constantly shaking. Afterward, the mix was filtered through a 100-μm and a 40-μm mesh and washed with cold PBS. The PBMC and PDAC single-cell suspensions were stained with mAbs listed in Supplementary Table S2 for flow cytometry. For cytokine expression, cells were additionally stimulated with phorbol 12-myristate and 13-acetate (PMA; 50 ng/mL) and ionomycin (1 μg/mL) for 4 hours at 37°C, 5% CO₂ in the presence of 1 mg/mL brefeldin A (BD Biosciences) in RPMI medium (Gibco) containing 10% heat-inactivated (60°C) FCS (Gibco) and 1% penicillin/streptomycin (Gibco) before staining. Cells were fixed and permeabilized for staining of intracellular cytokines and transcription factors with eBioscience FOXP3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific) after extracellular staining and stained with respective antibodies according to the manufacturer's protocol. Flow cytometry was carried out on the LSR Fortessa flow cytometer (BD Biosciences). Data were analyzed using FlowJo v10.7.1 (Treestar). Combination gates were used for the coexpression of two markers (PD-1/TIGIT, CD45RA/CCR7, or CD39/CD103). A minimum number of 200 cells was set for subanalysis. Each patient was analyzed individually.

FlowJo v.10.7.1 (Treestar) was used for t-SNE analysis. Intratumoral T cells from 15 patients were down-sampled to 10,000 cells each with the FlowJo DownSample v3.3 plugin. Cells were concatenated to perform t-SNE analysis on a total of 150,000 T cells on the basis of the expression of CD4, CD8, FOXP3, PD-1, TIGIT, IFNγ, TNFα, IL2, IL4, IL10, and IL17α with 3,000 iterations, 30 perplexities, 5,000 learning rate (eta), exact (vantage point tree) KNN algorithm, and Barnes–Hut gradient algorithm. Manual gating was performed and CD8⁺ T cells, Tconv, and Treg cell subsets were shown in t-SNE plot. In addition, heatmap statistics for PD-1, TIGIT, IFNγ, TNFα, IL2, IL4, IL10, and IL17α were shown to depict individual mean fluorescence intensities (MFI) in t-SNE plots.

HRs for clinicopathologic characteristics, namely T, N, and M stage, resection margin (R), neoadjuvant chemotherapy (neoCTx), age, and sex, were defined by multivariate Cox proportional-hazards regression using R Environment for Statistical Computing (R Core Team; 2022. R: A language and environment for statistical computing. R Foundation for Statistical Computing. http://www.R-project.org/). Each HR of PD-1, TIGIT, or their combinational expression was determined individually in a multivariate Cox proportional hazards regression as a function of all named clinicopathologic characteristics for each T-cell subset. In addition, univariate proportional-hazards Cox regression was performed to compare the survival probability of patients dichotomized according to high and low expression of the respective markers in R0-resected patients. To define the threshold for dichotomization R Environment for Statistical Computing was used by determining the odd's ratio between low versus high expression of every possible threshold in steps of 0.1%. Consecutively the threshold that minimized or maximized the odds ratio (OR) between the patient's with high versus low expression was chosen. Of note, we only chose thresholds within the second or third quartile of the cohort to ensure robust statistics also for smaller cohorts and allow balanced separation of groups. For this threshold, the survival function was plotted. In addition, Kaplan–Meier plots with the log-rank test were used to depict the observed data used to calculate the survival functions. They were generated with GraphPad Prism version 9.3.1 for Windows (GraphPad Software, www.graphpad.com). Patients that died within 30 days after surgery were excluded from the analysis. P ≤ 0.05 was considered statistically significant.

Data are shown as the median in dot plots or mean in bar graphs. To compare marker expression on the three T-cell subsets between matched blood and PDAC, paired two-sided t test with Holm–Šídák correction was applied. For comparison of four combinational groups, e.g., PD-1⁻TIGIT⁻, PD-1⁺TIGIT⁻, PD-1⁻TIGIT⁺, and PD-1⁺TIGIT⁺, repeated measurement ANOVA was applied. A mixed effects analysis was used instead in case of missing values due to low cell numbers within a certain group. Pearson correlation coefficient was used to analyze linear correlation. GraphPad Prism 9.3.1 (GraphPad Software) was used, and P ≤ 0.05 was considered statistically significant.

The raw flow cytometry data of this study are available from the corresponding author upon reasonable request.

To analyze T-cell composition and PD-1 and TIGIT expression in PDAC, we performed multicolor flow cytometry of freshly isolated immune cells from blood and matched tumor samples. First, we examined the composition of T-cell subsets (Supplementary Table S1). Although conventional CD4⁺ T cells (Tconv, CD3⁺ CD8⁻ CD4⁺ FOXP3⁻) were the most abundant T-cell subtype in peripheral blood, the proportion of CD8⁺ T cells (CD3⁺ CD8⁺ CD4⁻) and Treg (CD3⁺ CD8⁻ CD4⁺ FOXP3⁺) among all CD3⁺ T cells was significantly increased in the tumor (Fig. 1A; Supplementary Fig. S1A and S1B). Of note, 4% of T cells in the blood and 4.6% in PDAC were CD4⁻ CD8⁻ or CD4⁺ CD8⁺ and are not addressed in this study. The ratio of CD8⁺ T cells and Treg, as well as the ratio of Tconv and Treg, were significantly decreased in the tumor, indicating a predominately immunosuppressive T-cell infiltrate (Fig. 1B). Whereas all intratumoral T-cell subsets exhibited highly increased PD-1 expression, TIGIT expression was more heterogeneous among the three subsets (Fig. 1C and D). Notably, the expression level remained unchanged for CD8⁺ T cells, but TIGIT expression was significantly increased by intratumoral Tconv and Treg, with Treg showing the highest expression. Pearson correlation analysis showed a positive correlation of PD-1 and TIGIT expression between the different T-cell subsets in blood (Supplementary Fig. S1C and S1D) and PDAC (Fig. 1E and F). Furthermore, PD-1 and TIGIT expression between blood and tumor-infiltrating T cells showed a positive correlation, indicating an association between peripheral and intratumoral ICR expression (Supplementary Fig. S1E and S1F). Although T-cell composition and ICR expression were not associated with tumor stages, male patients showed significantly higher TIGIT expression by CD8⁺ T cells and Tconv in PDAC. NeoCTx altered the immune infiltrate by significantly reducing the expression of PD-1 on Treg and TIGIT on all three T-cell subsets (Supplementary Tables S3–S5).

Next, we assessed the prognostic value of PD-1 and TIGIT to investigate the association of their expression for patient outcomes. First, the impact of patient characteristics on survival in our cohort was determined by multivariate Cox proportional hazard regression analysis. Here, the incompleteness of resection (R1) and the treatment with neoCTx negatively affected patient outcomes (Supplementary Fig. S2A). However, neoCTx is often used in patients with advanced disease, so the informative value in comparing partial response (PR) and neoCTx is limited. Performing multivariate Cox proportional hazard regression for PD-1 and TIGIT expression that accounted for patient characteristics, we found HRs significantly greater than one for PD-1 expression by Tconv and TIGIT expression by Treg and similar trends for the other T-cell subsets in the blood (Fig. 2A). This translates into a significantly worse clinical outcome with increasing PD-1 or TIGIT expression on blood T cells. In univariate analyses, we stratified the patients into two groups with low or high ICR expression, that differed significantly in their survival functions. Of note, R1-resected patients were not included in the univariate analysis to avoid a superimposing effect of incomplete resection and Kaplan–Meier plots were included in the Supplementary Figures to depict the observed outcomes. PD-1 and TIGIT expression on blood CD8⁺ T cells were strongly associated with patient outcome and enabled patient stratification (Fig. 2B and C; Supplementary Fig. S2B). With an OR of 2.036, patients with more than 24.1% of CD8⁺ T-cells expressing PD-1 were twice as likely to die within the observed time. Similar trends were seen for Tconv and Treg (Supplementary Fig. S2B and S2C). In PDAC, we saw a tendency of high PD-1 expression characterizing patients with beneficial outcomes, although there was a trend for an association of TIGIT expression and reduced survival (Supplementary Fig. S3A and S3B).

Subsequently, we wanted to analyze the coexpression of PD-1 and TIGIT. Although PD-1⁻ TIGIT⁻ cells were prevalent among CD8⁺ T cells and Tconv in blood, PD-1⁺ TIGIT⁻ and PD-1⁺ TIGIT⁺ cells were significantly increased in the tumor (Fig. 3A). PD-1⁻ TIGIT⁺ CD8⁺ T cells were reduced in the tumor compared with blood, whereas PD-1⁻ TIGIT⁺ expression was rare but increased among Tconv. Most intratumoral Treg coexpressed both receptors. Of note, there were substantial interpatient differences in the distribution of the combination subsets, e.g., intratumoral PD-1⁻ TIGIT⁻ of Tconv ranged from 6.7% up to 86.6% with a mean of 36.2% in PDAC. Considering the prevalence of the subsets among all T cells, PD-1⁻ TIGIT⁻ Tconv was the most prominent T-cell subset in blood, averaging 45.7% (Fig. 3B). The composition within the PDAC was more balanced between the subsets. Antitumor immunity depends on T-cell activity and fate, which is regulated by differentiation. CD45RA and CCR7 serve to distinguish naïve (N, CD45RA⁺ CCR7⁺), effector memory (EM, CD45RA⁻ CCR7⁻), central memory (CM, CD45RA⁻ CCR7⁺) and EM reexpressing CD45RA (EMRA, CD45RA⁺ CCR7⁻) cells (25). We analyzed T-cell differentiation and association with ICR expression to unravel the relevance of the PD-1 and TIGIT combinational subsets. Although blood CD8⁺ T-cells predominantly exhibited an EMRA phenotype, EM cells were significantly increased in PDAC (Fig. 3C). Naïve and CM CD8⁺ T cells were rare. Tconv and Treg showed a significant decrease in the naïve phenotype but an increase in the EM phenotype. Overall, EM was the prominent phenotype of all PDAC-infiltrating T cells. Next, we performed a subanalysis of PD-1 and TIGIT expression as a function of the differentiation status that showed distinct distributions of both ICR for each phenotype in PDAC (Fig. 3D). As expected, naïve T-cells showed low expression of both receptors. Interestingly, the proportion of PD-1⁺ TIGIT⁻ was highest in EM CD8⁺ T cells and Tconv, which were low in PD-1⁻ TIGIT⁺ cells. In turn, PD-1⁻ TIGIT⁺ expression was dominant in the EMRA phenotype. Intratumoral CM CD8⁺ T cells showed the highest percentage of PD-1⁺ TIGIT⁺ cells. In the blood, a similar heterogeneous distribution of PD-1 and TIGIT expression with generally more PD-1⁻ TIGIT⁻ and less PD-1⁺ TIGIT⁺ cells among all differentiation status was identified, highlighting an association of differentiation and ICR expression independent of T-cell environment (Supplementary Fig. S4).

To further decipher the importance of PD-1 and TIGIT expression for T-cell functionality, we investigated the association of PD-1 and TIGIT coexpression with pro (IFNγ, TNFα, IL2, and IL17γ) and antiinflammatory (IL4 and IL10) cytokine expression as central mediators of T-cell function. To depict the coexpression of PD-1, TIGIT, and all cytokines for all T cells, we uniformly concatenated intratumoral T cells from 15 patients and performed t-SNE analysis for dimensionality reduction (Fig. 4A). The t-SNE analysis imposingly showed that Treg highly expressed TIGIT (Fig. 4B and C). Whereas IL17α, IL4, and IL10 were expressed by only a small proportion of T cells and rarely coexpressed, IFNγ, TNFα, and IL2 were frequently coexpressed (Fig. 4D). IFNγ expression was increased considerably by Tconv, indicating potential antitumor activity (Supplementary Figs. S5 and S6). Intratumoral Treg had a significantly reduced expression of TNFα and IL2, and IL2 was also reduced by Tconv but increased by CD8⁺ T cells (Supplementary Fig. S6B and S6C). A trend toward increased expression of IL4 and IL17a by CD8⁺ T cells and Tconv was observed, highlighting their functional heterogeneity (Supplementary Fig. S6D and S6E). Although Tregs showed the highest expression of IL10, the expression was significantly increased by Tconv (Supplementary Fig. S6F). Highly divergent patterns were identified when analyzing cytokine expression as a function of PD-1 and TIGIT expression. PD1⁺ TIGIT⁻ CD8⁺ T-cells showed the highest expression of IFNγ, TNFα, IL2, and IL17α compared with all other combinational CD8⁺ T-cell subsets (Fig. 4E–H). TIGIT⁺ CD8⁺ T cells highly expressed antiinflammatory cytokines (Fig. 4I and J) but showed a low level of IL2 expression. We did not detect a difference in TNFα, IL2, IL4, or IL17α expression between PD-1⁺ TIGIT⁺ and PD-1⁻ TIGIT⁺ CD8⁺ T cells. This indicates that TIGIT expression labels antiinflammatory or exhausted CD8⁺ T-cells independently of PD-1 expression, whereas the phenotype of PD-1-expressing CD8⁺ T-cells highly depends on the coexpression of TIGIT. Similarly, PD1⁺ TIGIT⁻ Tconv showed the highest expression of IFNγ, TNFα, IL2, and IL17α, while PD-1⁻ TIGIT⁺ showed the lowest expression. TIGIT expression labeled Treg with deficient levels of IFNγ and IL2 expression. PD1⁺ TIGIT⁻ Tconv and Tregs were the primary producers of IL17α. Also, in blood, similar cytokine expression patterns as a function of PD-1 and TIGIT expression were identified (Supplementary Fig. S7).

We next investigated the coexpression of CD103 and CD39 as surrogate markers for tumor-reactive T cells (26, 27). CD39 was upregulated by all intratumoral T-cell subsets, with the highest expression on Tregs (Supplementary Fig. S8A–S8C). Whereas CD103 expression on blood T cells was low, it was significantly increased on intratumoral T cells (Supplementary Fig. S8D). CD39⁺ CD103⁺ T cells were barely detected in the peripheral blood, but especially intratumoral CD8⁺ T-cells coexpressed CD39 and CD103 (Supplementary Fig. S8E). PD1⁺ TIGIT⁻ CD8⁺ T cells showed the highest expression of CD103 and a high level of CD39/CD103 coexpression (Fig. 5A–C). Interestingly, TIGIT-expressing Tconv and Treg showed the highest expression of CD39. Also in the blood, PD-1 and TIGIT were associated with divergent expressions of CD39 and CD103 (Supplementary Fig. S8F–S8H). The transcription factor Eomesodermin (EOMES) is important for the antitumor function of CD8⁺ T cells, but also drives their exhaustion (28). In CD4⁺ T cells, EOMES may drive cells into a cytotoxic phenotype (29). Therefore, we assessed EOMES expression in relation to PD-1 and TIGIT. Although EOMES expression was significantly decreased by CD8⁺ T cells, it was increased by Tconv and Treg in PDAC (Supplementary Fig. S9A and S9B). Independent of PD-1 expression, intratumoral TIGIT⁺ CD8⁺ T-cells showed the highest expression of EOMES, indicating an exhausted phenotype (Fig. 5D). Interestingly, EOMES expression by intratumoral Tconv was mainly found in PD-1–expressing cells, especially by PD-1⁺ TIGIT⁻ Tconv indicating a cytotoxic differentiation in line with the proinflammatory cytokine profile. In the blood, EOMES expression showed a similar relationship to ICR expression, except that PD-1⁺ TIGIT⁻ CD8⁺ T cells also showed high EOMES expression (Supplementary Fig. S9C).

Given the opposing phenotypes of PD⁺ TIGIT⁻ and PD-1⁻ TIGIT⁺ T cells, we consecutively analyzed the association of survival with PD-1 and TIGIT coexpression. In the multivariate Cox proportionalhazard regression analyses, the proportion of PD-1⁻ TIGIT⁻ Tconv and Treg in blood was significantly associated with favorable survival (HR < 1; Fig. 6A). When comparing groups of R0-resected patients with low and high combinational PD-1 and TIGIT expression by univariate analysis, a high percentage of PD-1⁻ TIGIT⁻ Tconv or a low percentage of PD-1⁻ TIGIT⁺ Treg in the blood stratified patients with an improved prognosis (Supplementary Figs. S10 and S11). Surprisingly, a high proportion PD-1⁺ TIGIT⁻ CD8⁺ T cells in patients characterized with PDAC with bad OS (Fig. 6B; Supplementary Fig. S12A). In turn, patients displayed significantly favorable outcomes with an OR of 0.304 if more than 34.8% of Tconv showed the proinflammatory and potentially cytotoxic PD-1⁺ TIGIT⁻ phenotype. A similar situation was found for Treg in line with the multivariate analysis. In contrast, there was a strong trend for an association of high PD-1⁻ TIGIT⁺ Tconv and Treg with reduced survival (Fig. 6C). The percentage of PD-1⁻ TIGIT⁻ or PD-1⁺ TIGIT⁺ T cells was not found to be significantly associated with patient survival (Supplementary Fig. S12A and S12B).

Our study provides new insights into the relevance of PD-1 and TIGIT coexpression on PDAC-infiltrating T cells. By analyzing differentiation, cytokine expression, and markers of tumor reactivity as a function of combinational PD-1 and TIGIT expression, we deciphered distinct T-cell phenotypes and linked them to patient survival. In addition, we prove a substantial prognostic value of ICR expression on blood T cells. Our analysis of freshly isolated immune cells from blood and PDAC enabled the detection and characterization of rare phenotypes that might not be possible after cryopreservation (30, 31), and our large cohort with an extensive patient follow-up supports their translational relevance. Several studies focus on CD8⁺ T cells, often called cytotoxic T cells, as the primary cells in T-cell–mediated tumor cell killing but omit CD4⁺ T cells (32). Here, we analyzed CD8⁺ T cells, Tconv, and Treg and gained new insights into their function in PDAC immunity and clinical relevance.

In PDAC, the CD8⁺ T-cell phenotype is dominated by an exhausted, dysfunctional state that was recently linked to TIGIT expression on the basis of single-cell RNA-sequencing and CyTOF data (20, 33). Our data confirm TIGIT as a marker of dysfunctional CD8⁺ T cells. Although the proportion of TIGIT-expressing cells among CD8⁺ T cells was very heterogenous among patients, these cells had reduced proinflammatory, but increased antiinflammatory capacities and highly expressed the transcription factor EOMES that regulates CD8⁺ T-cell exhaustion (28). Interestingly, PD1⁺ TIGIT⁻ CD8⁺ T cells had a contrary phenotype toward TIGIT-expressing cells characterized by high proinflammatory, and low antiinflammatory cytokine expression and were dominant in the EM phenotype. In addition, PD1⁺ TIGIT⁻ CD8⁺ T cells highly expressed CD103 alone or coexpressed with CD39. This coexpression was found to mark tumor-reactive T cells that underwent a clonotypic expansion and are associated with improved prognosis in other cancers (27, 34, 35). Our study uncovers TIGIT as superior to PD-1 in defining the CD8⁺ T-cell phenotype, as there were only minor differences between PD1⁻ TIGIT⁺ and PD1⁺ TIGIT⁺ cells, but an opposing phenotype of PD1⁺ TIGIT⁻ CD8⁺ T cells. Of note, we contradictory saw that a low proportion of PD-1⁺ TIGIT⁻ cells among CD8⁺ T cells characterized a small group of patients with better survival, but overall there was no trend in the multivariate analysis supporting this finding. The PD-1 and TIGIT combinational subsets associated with significantly different survival functions between high and low expressors are summarized in Supplementary Fig. S13.

Although Tconv are the most abundant T-cell subset in the blood, the percentage among intratumoral T cells was significantly reduced. Interestingly, Tconv showed substantially increased IFNγ expression indicating potential antitumoral activity in PDAC. Identical to intratumoral CD8⁺ T cells, PD-1⁺ TIGIT⁻ Tconv showed the highest and PD-1⁻ TIGIT⁺ the lowest expression of proinflammatory cytokines. A recent multiplex IHC study indicated that PD-1⁺ CD4⁺ T cells inhibit immunosuppressive myeloid cells via IFNγ in PDAC long-term survivors (36). In lung cancer, CD103⁺ CD4⁺ were the most potent producers of proinflammatory cytokines (37). It is consistent with our data, in which the intratumoral and proinflammatory PD-1⁺ TIGIT⁻ Tconv also showed the highest expression of CD103. In addition, PD-1⁺ TIGIT⁻ Tconv highly expressed EOMES. In CD4⁺ T cells, EOMES mediates differentiation toward cytotoxic functionality (29). This suggests a highly proinflammatory and potentially cytotoxic phenotype of PD-1⁺ TIGIT⁻ Tconv, and a high abundance of this subset might result from antigen-driven clonotypic expansion. Most interestingly, we prove an association of this phenotype with an improved prognosis, whereas patients with an exhausted PD-1⁻ TIGIT⁺ Tconv phenotype had a near significantly worse clinical outcome. This is consistent with the trends seen in our multivariate analyses. The balance between PD-1 and TIGIT expression among the Tconv appears critical for function and antitumor activity.

The proportion of Tregs was highly increased compared with the peripheral blood, in line with other studies on PDAC that found high intratumoral Treg occurrence (33, 38). Whereas several studies found an association of high Treg infiltration with bad clinical outcomes (12, 39), Treg depletion promoted tumorigenesis in a spontaneous PDAC mouse model (40). We detected a robust immunosuppressive phenotype and increased TIGIT expression on intratumoral Treg. Other studies indicated a high immunosuppressive capacity of TIGIT⁺ Tregs (24, 41). As only TIGIT⁺ CD4⁺ FOXP3⁺ T cells showed very low expression of IFNγ and IL2, we support the concept of TIGIT as a definition marker of actual Tregs (20). These TIGIT⁺ Tregs also showed the highest expression of the cell surface enzyme CD39, which converts adenosine ATP, ADP to AMP and was found to suppress antitumor immunity (42). In mice, CD39 was shown to cooperate with CD73 to promote tumor progression in subcutaneously transplanted PDAC and high expression of CD39 and CD73, in human, PDAC was associated with worse survival (43). Targeting CD39 in PDAC and other solid tumors entered first clinical trials (44). PD-1⁺ TIGIT⁻ CD4⁺ FOXP3⁺ cells, in turn, exhibited a highly activated phenotype associated with improved patient outcomes in our multivariate analysis and should not be stamped as immunosuppressive Treg. Those cells also showed the highest IL17α expression of all investigated T-cell subsets. The Induction of Th17 was beneficial in a murine model of PDAC (45). In our study, Th17 were mainly included in the PD-1⁺ TIGIT⁻ subsets of Tconv and Tregs that were highly beneficial for patient outcomes, providing further evidence that Th17 cells are valuable in PDAC antitumor immunity.

In addition to deciphering intratumoral T-cell phenotypes, we also detected a striking association of ICR expression on circulating T cells with patient survival. Although PD-1 expression on circulating CD8⁺ T cells was already shown to be correlated with nodal and distant metastasis (16) and short OS (46), here we uncover a significant association for PD-1 expression on Tconv and TIGIT expression on Treg in blood with the patient outcome in multivariate analyses. Also patients with high PD-1 or high TIGIT expression by CD8⁺ T cells were characterized by a significantly worse OS. Therefore, these findings could help stratify patients according to their survival prognosis to adjust treatment regimens or intensify their follow-up, as flow cytometric immunophenotyping of blood specimens represents a simple tool. In addition, blood immunophenotyping might help to assess ICR expression in the tumor immune microenvironment (TIME), as we identified a positive correlation between blood and intratumoral ICR expression, as recently shown for TIGIT expression on CD8⁺ T cells (20).

Taken together, we show that an exhausted or dysfunctional phenotype of CD8⁺ T cells is linked to TIGIT expression and that TIGIT marks highly suppressive Tregs. These findings highlight TIGIT as a potential target for immunotherapeutic approaches. PD-1⁺ TIGIT⁻ Tconv and Tregs were found highly proinflammatory and potentially cytotoxic and are associated with improved patient outcomes. Immunophenotyping of patient blood may stratify patients with PDAC and serve as a prognostic tool. The immunosuppressive action of ICR often depends on interaction with their respective ligands. Although the expression of the PD-1 ligand PD-L1 was intensively studied and found to be rare in PDAC (47), the expression of TIGIT ligands is poorly investigated with first studies indicating clinical significance and functional relevance of CD155 for immune evasion in PDAC (21, 48). In addition, very recent transcriptomic studies identified specific gene expression programs of tumor cells and fibroblasts that define distinct states of the PDAC microenvironment and thereby may be related to the differential PD-1 and TIGIT coexpression we observed (49, 50).

Although different mechanisms targeting TIGIT are discussed (51), our data suggest that depletion of TIGIT⁺ cells with specific antibodies might be a promising approach because mainly immunosuppressive or exhausted, but not proinflammatory cells would be affected. Of note, PD-1 and TIGIT converge in suppressing costimulatory signaling via CD226 as recently found in vitro and tested for functional relevance by a coblockade in a subcutaneous non–small cell lung carcinoma (NSCLC) mouse model (52). The coblockade was superior to single blockade in increasing IFNγ production by T cells from patient with PDAC blood activated by artificial antigen-presenting cells (APC) expressing PD-1 and TIGIT ligands (38). Unfortunately, TIGIT-blockade by Tiragolumab in combination with atezolizumab (anti–PD-L1) and chemotherapy was unsuccessful in a phase III trial in small cell lung cancer (SCLC; ref. 53). For PDAC, this combination is currently tested in the Morpheus-Pancreatic Cancer trial (NCT03193190) in parallel to other chemo and immunotherapeutic combinational approaches. Notably, blocking TIGIT might be more effective before or with chemotherapy, as neoCTx was found to reduce TIGIT/CD155 interaction (49). Hence, adding immunotherapy to neoCTx may be promising for PDAC as it alters the TIME towards a proinflammatory situation (54–56). Combinational approaches incorporating TIGIT-blockade may improve the treatment and prognosis of patients with PDAC.

L. Seifert reports grants from German Research Fundation (DFG), Jung-Stiftung, Monika Kutzner Stiftung, German Cancer Consortium (DKTK), Medical Faculty Dresden, and Else Kröner Fresenius Stiftung during the conduct of the study. No disclosures were reported by the other authors.

M. Heiduk: Conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. A. Klimova: Formal analysis, validation, writing–review and editing. C. Reiche: Writing–review and editing, acquisition of patient follow-up. D. Digomann: Writing–review and editing. C. Beer: Investigation. D.E. Aust: Resources. M. Distler: Resources. J. Weitz: Resources. A.M. Seifert: Conceptualization, resources, supervision, funding acquisition, methodology, project administration, writing–review and editing. L. Seifert: Conceptualization, resources, supervision, funding acquisition, methodology, project administration, writing–review and editing.

We thank Jung-Stiftung (L. Seifert), Monika Kutzner Stiftung (A.M. Seifert), German Research Foundation (DFG; SE2980/5–1; L. Seifert), the German Cancer Consortium (DKTK; A.M. Seifert), Medical Faculty Carl Gustav Carus Technische Universität Dresden (M. Heiduk, A.M. Seifert), Else Kröner-Fresenius-Stiftung (L. Seifert – Else Kröner Clinician Scientist Professorship) for the funding. Furthermore, we are grateful to Heike Polster for the assistance in the acquisition of human blood and to members of the Institute of Pathology in Dresden for the aid in the acquisition of tumor specimens. In addition, we thank Cindy Fuchs for her help with the graphical abstract.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.