Immunotherapy, including PD-1/PD-L1 agonists, has shown limited efficacy in pancreatic ductal adenocarcinoma (PDAC). We examined the PD-1/PD-L1 expression and immunoarchitectural features by automated morphometric analysis using multiplex immunofluorescence and 118 microsatellite-stable, treatment–naïve, surgically resected PDACs (study cohort). Five microsatellite-instable cases were stained in parallel (MSI cohort). Molecular analysis was additionally performed. An independent PDAC cohort (n = 226) was immunostained for PD-L1 and used as a validation cohort. PD-L1 expression on tumor cells (TC) and/or immune cells (IC) was present in 32% and 30% of the study and validation cohorts, respectively, and assigned into one of four patterns: “adaptive-1” (TC: 0, IC > 1%), “adaptive-2” (TC > 1% to < 25%, IC > 1%), “constitutive” (TC ≥ 25%, IC: 0), and “combined” (TC ≥ 25%, IC > 1%). “Constitutive” tumors were characterized by reduced numbers of all ICs and poor outcome. In contrast, “adaptive-1” tumors exhibited abundant T cells, including high counts of cytotoxic CD3+CD8+ and PD-1+CD3+CD8+ cells, but low counts of PD-L1+CD3+CD8+ cells and associated with the best outcome. “Adaptive-2” tumors displayed higher proportions of PD-L1+CD3+CD8+ T cells and tumor-associated macrophages (CD68+ and CD68+CD206+) compared with “adaptive-1” tumors. In the “combined” pattern, extensive PD-L1 expression on TCs was accompanied by increased numbers of T cells and improved overall survival. ICs were closer to PD-L1 than to PD-L1+ PDAC cells. TP53 and PIK3CA alterations tended to be more frequent in PD-L1+ tumors. The 5 MSI cases were PD-L1. The distinct PD-1/PD-L1–associated immunoarchitectural patterns underpin the heterogeneity of the immunologic responses and might be used to inform patient outcomes and therapeutic decisions in pancreatic cancer.

Immunotherapy, including programmed death-1/programmed death ligand-1 (PD-1/PD-L1) antagonists, has shown so far only very limited clinical activity for patients with pancreatic ductal adenocarcinoma (PDAC; refs. 1, 2). Combining immune-checkpoint blockade with chemotherapy has also not resulted in measurable clinical responses (3). The limited effectiveness of immunotherapy in PDAC has been attributed to its immunologically “cold” tumor microenvironment (TME; ref. 4). Nevertheless, PD-1/PDL-1 antagonists have been more successful in a small subset of PDAC patients with microsatellite-instable (MSI) tumors (5). Results from vaccine strategies suggest that the weak immunogenicity of pancreatic cancer may be reversible (6). These data imply that the key for a successful clinical response of immunotherapy in PDAC may be a more precise patient selection and the unveiling of targets that would shift the PDAC TME from an immunologically “cold” into an immunologically “hot” phenotype (4).

In many solid tumors, the presence of an active immune microenvironment, with a high density of activated T cells, is associated with favorable prognosis and responsiveness to immunotherapy (7, 8). However, the functional spectrum of immune infiltrates in the TME of most malignant neoplasms exhibits a high inter- and intratumoral variation and constitutes a crucial factor for defining “hot” immunogenic tumors (8, 9). The upregulation of proinflammatory signals following recognition of tumor antigens by the T cells, and the subsequent amplification of the inflammatory response, will eventually set in motion counterregulatory mechanisms, leading to compensatory immune resistance (10). The expression of interferon-gamma (IFNγ)-inducible immune regulatory molecules, such as PD-L1, and the activation of the PD-1/PD-L1 axis are of particular interest because tumors often use this pathway to suppress host antitumor immune responses, simultaneously providing us with an important predictive biomarker for response to anti–PD-1/PD-L1 treatment (7).

PD-L1–mediated adaptive immune resistance by tumors was first described in melanoma but has since been extended to a large number of tumor types (7, 11–13). Two types of PD-L1–mediated adaptive immune resistance have been described, including PD-L1 expression on tumor cells (TC) immediately adjacent to PD-L1 expressing immune cells (TC+IC+) and PD-L1 expression on infiltrating immune cells immediately adjacent to PD-L1–negative TCs (TCIC+; ref. 14). Oncogenic alterations have also been implicated in modulating PD-L1 expression, including PTEN loss (15, 16) and activation of the AKT–mTOR or JAK/STAT pathways (7, 17). Some oncogenic events can lead to constitutive or “innate” PD-L1 expression on the surface of TCs in the absence of an activated T-cell infiltrate (14, 18). A combined, oncogenic-driven (constitutive) plus adaptive pattern of PD-L1 expression can also be observed (14).

With the expanding use of immunotherapy, there is a clear need for improved biomarker development to help identify patients and tumor subgroups most likely to respond to therapy. As the number of patients treated with checkpoint inhibition increases, the number of treatment-related adverse reactions also increases (7, 19). Therefore, selecting the suitable patients who have a high probability to respond is of paramount importance. To improve the power of PD-L1 as a predictive biomarker for immunotherapy in pancreatic cancer, we need to have a more comprehensive approach for discerning PD-1/PD-L1–mediated immune resistance, including the innate and adaptive antitumor immune responses. In this study, we correlated PD-1/PD-L1–associated responses with the different immunoarchitectural features of the TME in order to improve our understanding of the key mechanisms needing to be addressed to leverage the immunotherapeutic response in PDAC.

The study design is outlined in Supplementary Fig. S1. The study cohort was stained immunohistochemically for the MMR proteins (MLH1, MSH2, MSH6, and PMS2), and two MSI cases were excluded. The remaining cases [microsatellite stable (MSS), n = 118) were stained with PD-L1, and 38 PD-L1+ cases were identified. Morphometric and proximity analysis of all PD L1+ cases of the study cohort by multiplex immunofluorescence (mIF) for CD3, CD4, CD8, FOXP3, CD68, CD206, PD-1, PD-L1, and pancytokeratin revealed four distinct PD-L1–associated immune patterns as depicted in the pie chart and images. One hundred twelve cases of the study cohort underwent next-generation sequencing (NGS) using the Oncomine Comprehensive Assay v3 panel. Similarly, the 229 PDAC cases of the validation cohort were stained immunohistochemically for MMR proteins (MLH1, MSH2, MSH6, and PMS2), and three MSI cases were excluded. The remaining cases (MSS, n = 226) were stained with PD-L1, and 67 PD-L1+ cases were identified. The five MSI cases also underwent morphometric analysis, similar to the MSS PD-L1+ cases of the study cohort, and the results were compared with the four PD-L1 immune resistance patterns. All data were integrated, and statistical analysis was performed. Methods used for this analysis are described below.

Patient characteristics

From 349 consecutive PDAC patients, who underwent oncologic resection between 2003 and 2018 at the Department of Visceral Surgery and Medicine, Insel University Hospital, Bern, and who fulfilled the inclusion criteria such as full clinical information and enough available tumor tissue to perform the analyses, 120 were randomly selected to build the study cohort, and the remaining 229 constituted the validation cohort. Representative paraffin blocks from each tumor were selected by an experienced pancreatobiliary pathologist (EK) for further analysis. All tissues were processed according to international protocols, and paraffin blocks were stored at the archives of the Institute of Pathology, University of Bern under optimal conditions. Clinical data were available for all patients, including gender, age, comorbidities, tumor location, postoperative chemotherapy and radiotherapy, TNM stage, postoperative morbidity and mortality, type of surgery, resection margins status, disease-free survival, and OS. The study was approved by the Ethics Commission of the Canton of Bern (KEK 2019-02212) and was carried out in accordance with the principles expressed in the Declaration of Helsinki. Written informed consent was available for the patients, and their records were deidentified prior to the analysis. No neoadjuvant chemotherapy or immunotherapy was administered to these patients. Most patients (335 of 349, 96%) received adjuvant chemotherapy after oncologic resection. From these patients, 228 (68%) received gemcitabine, 37 (11%) gemcitabine + xeloda, and the remaining 70 patients (21%) received FOLFIRINOX. After IHC analysis for MLH1, MSH2, MSH6, and PMS2 (as described below), five MSI tumors (1.4%) were identified, whereas the remaining 344 PDACs were MSS. From the 344 MSS cases, 182 (52.9%) were men and 162 (47.1%) were women. Age ranged from 32 to 87 (median 66). After excluding the MSI cases, the study cohort comprised 118 MSS cases (n = 120 − 2 MSI cases), and the validation cohort 226 MSS cases (n = 229 − 3 MSI cases). The 5 MSI cases, comprising 3 men and 2 women, age range, 60–76 (median 70), were also analyzed for comparison. The clinicopathologic characteristics of the patients are summarized in Supplementary Table S1 (study cohort), Supplementary Table S2 (validation cohort), and Supplementary Table S3 (MSI cohort).

IHC

Whole-tissue sections (WTS; n = 349) were sectioned at 3 μm, dewaxed, rehydrated, and stained immunohistochemically for PD-L1 (SP263, Ventana Roche, cat. #790-4905, RRID:AB_2819099, prediluted), MLH1 (Abnova, cat. #H00004292-M02, RRID:AB_464158, 1:1000), PMS2 (Abnova, cat. #PAB14829, RRID:AB_10676300, 1:50), MSH2 (Abnova, cat. #MAB1956, RRID:AB_1678867,1:25), and MSH6 (Abnova, cat. #MAB10045, RRID:AB_10903400, 1:500). Antigen retrieval was performed with Tris-HCl, pH 9 for 30 minutes at 95°C. Staining was performed by an automated BenchMark ULTRA System (Ventana, Roche F. Hoffmann-La Roche AG) with the OptiView DAB IHC Detection Kit (Ventana Roche). PD-L1 staining was assessed as a percentage of viable TCs with any membrane staining above background (TC+) and/or by the percentage of tumor-associated immune cells with staining (IC+) at any intensity above background. For the MMR proteins (MLH1, MSH2, MSH6, and PMS2), loss of protein expression was accepted if unequivocal loss of nuclear staining or focal weak equivocal nuclear staining was present in viable TCs in the presence of internal positive controls (nuclear staining in lymphocytes, fibroblasts, or normal epithelium in the vicinity of the tumor). All stained slides were digitalized (3DHISTECH Panoramic 250) and evaluated by virtual microscopy using the Case Viewer software (Case Viewer 3DHISTECH_Ltd Version 2.3.0.99276). Evaluation was performed blinded to clinical parameters.

mIF

WTSs from all PD-L1+ PDACs of the study cohort, as well as the 5 MSI cases, were further stained for mIF using an automated protocol (BOND, Leica Biosystems) as described by Vasaturo and Galon (20). Primary antibodies conjugated to DNA barcodes (Leica Biosystems) were used in order to identify the protein targets of interest in the formalin-fixed, paraffin-embedded (FFPE) PDAC samples. Incubation was performed with a mixture of antibody conjugates (dilution 1:100): CD3 (Thermo Fisher Scientific, RRID:AB_10979571), CD4 (Abnova, RRID:AB_10550833), CD8 (Thermo Fisher Scientific, RRID:AB_11000353), CD68 (Thermo Fisher Scientific, RRID:AB_10987212), FOXP3 (Thermo Fisher Scientific, RRID:AB_2573609), CD206 (Abcam, RRID:AB_1523910, ab64693), PD-1 (Thermo Fisher Scientific, RRID:AB_11152225, 7A11B1), PD-L1 (Thermo Fisher Scientific, RRID: AB_10986627), and pancytokeratin (Thermo Fisher Scientific, RRID:AB_1834350). The DNA barcodes of each target were amplified using a preamplification solution (Leica Biosystems) followed by amplification solution (Leica Biosystems) containing an amplification enzyme and amplification buffer. Amplification was carried out simultaneously for all targets. Fluorescent probes (Leica Biosystems) that were complementary to the barcodes were then added to the samples to bind and label each round of targets. Each target was labeled with a spectrally distinct fluorophore to enable multiplexed whole slide imaging. A removal step was used to exchange fluorescent probes between imaging rounds for multiplexing.

Morphometric image and proximity analysis

For the morphometric analysis, the immune cell module and the HighPlexFL cell analysis module of the HALO image analysis platform (Indica Labs) were used. The modules identify the nucleus based on the DAPI stain and algorithm optimization. The cytoplasm is identified by growing the nucleus by a specific number of microns to estimate the cytoplasm compartment (radius = 2.5 μm). The cell positivity for each channel was classified as negative/positive based on an intensity threshold within the appropriate cell compartment by using and adjusting a threshold slider until all cells with nuclei and intensity > 1+ (on a scale of 1–3; i.e., weak, moderate, and strong) were captured. The density (per mm2) and the percentage of each immune cell population, including CD3+, CD4+, and CD8+ tumor-infiltrating lymphocytes (TIL), CD68+ tumor-associated macrophages (TAM), and FOXP3+ T regulatory cells (Treg), as well as their coexpressing phenotypic subpopulations, were estimated in both the intraepithelial (i.e., ICs found interspersed within the tumor epithelial cells) and stromal compartments: CD3+CD4+, CD3+CD8+, CD3+CD4+CD8+, CD3+CD4+FOXP3+, CD3+CD8+FOXP3+, CD68+CD206+ (M2-polarized TAMs), PD-1+CD3+CD8+, PD-L1+CD3+CD8+, and PD-L1+CD68+ cells. Proximity analysis (HALO image analysis platform, Indica Labs) was used to measure the counts of each immune cell type in the proximity area of PD-L1+ and PD-L1 TCs. The data were binned at set distances (concentric circles of 50 μm, radius of 25 μm), and histogram plots for each immune cell population were created by using the same platform.

NGS

DNA from 112 tumor samples of the study cohort and the Oncomine Comprehensive Assay v3 panel (Thermo Fisher Scientific), designed to allow amplification-based capture and sequencing of single-nucleotide variants (SNV), copy-number variations, gene fusions, and indels from 161 cancer-associated genes (Ion Torrent platform), was used. The remaining six cases of the study cohort were excluded due to technical reasons (four cases due to insufficient sample and two cases due to poor DNA quality). Construction of amplicon libraries was performed following the manufacturer's instructions. This was followed by targeted NGS and data analyses.

Raw data (FASTQ files) for each sample were processed for the alignment of sequencing reads with the human genome reference (hg19) using the Torrent Suite software v5.4 (Thermo Fisher Scientific; https://github.com/iontorrent/TS). Sequencing data were uploaded in BAM format to the Ion Reporter Analysis Server for variant calling and annotation. Variant calling was performed on Ion Reporter Analysis Software v5.0.3. Local realignment and quality base score recalibration were also carried out using the Genome Analysis Toolkit (GATK). Coverage metrics for each amplicon were obtained by running the Coverage Analysis Plugin software v5.2.1 (Thermo Fisher Scientific). Identified variants were only considered if the variant had a molecular coverage of at least 3, indicating that the variant was detected in three independent template molecules. Nontumoral pancreatic tissues were used as matched germline samples. Mutations supported by 10 times that of the matched nontumoral variant allele frequency were retained to ensure the somatic nature of the variants. Additionally, MuTect4 and VarScan25 algorithms were used to call somatic SNVs, and VarScan25, Strelka6, and Scalpel7 were used for insertions and deletions (indels). Finally, all mutations were manually inspected by using the Integrative Genomics Viewer version 2.3.69 (https://software.broadinstitute.org/software/igv).

Statistical analysis

Pearson correlation coefficients were calculated to determine correlation of PD-1/PD-L1 pattern classifications between study and validation cohorts. Association between categorical variables were calculated by means Fisher exact test or chi-square test. Nonparametric Kruskal–Wallis test was performed to compare across multiple immune categories, and false discovery rate (FDR)–corrected P values were generated. No correction for multiple comparisons was used. Principal component analysis (PCA), a dimensionality reduction technique, was used for the separation of intratumoral and stromal immune cell profiles across the different PD-L1 staining patterns. Kaplan–Meier survival analysis and log-rank tests were performed to determine the associations of PD-L1 patterns with OS. Analyses were conducted on SAS (V9.3). All tests were two sided, and P values were considered significant at P < 0.05. Multivariate Cox regression was performed using survival package in R (https://CRAN.R-project.org/view=Survival).

Clinicopathologic and molecular profiles

The study cohort (n = 118) included 62 men (52.5%) and 56 women (47.5%), with a median age of 65 years (range, 34–84) and median OS of13.5 months (range, 1–196; Supplementary Table S1). Of the 226 patients constituting the validation cohort, 120 (53%) were men and 106 (47%) women, median age was 66.5 years (range, 32–87), and median OS was 13 months (range, 1–190; Supplementary Table S2). The small MSI cohort (n = 5) included three men (60%) and two women (40%), with a median age of 70 years (range, 62–76; Supplementary Table S3). A comparison between PD-L1+ and PD-L1 cases of the study cohort for frequency of genomic alterations including mutations, homozygous deletions (HD), and loss of heterozygosity showed an increased frequency of TP53 and PIK3CA alterations in PD-L1+ tumors (82.5% and 11.4%, respectively) compared with PD-L1 tumors (61.5% and 3.8%, respectively; Supplementary Table S4). However, these differences did not achieve statistical significance (P = 0.4 and P = 0.2, respectively).

Patterns of PD-1/PD-L1 expression

PD-L1 positivity in TCs and/or ICs was present in 38 (32%) PDACs of the study cohort and in 67 (30%) PDACs of the validation cohort, respectively. Based on previous observations (14), we identified four distinct patterns of PD-L1 expression, including both adaptive- and innate-type reactions. In more detail, two patterns of PD-L1–mediated adaptive immune resistance were observed. The first pattern, defined as “adaptive-1,” showed PD-L1 expression on infiltrating immune cells immediately adjacent to TCs, but tumor cells were PD-L1 negative (IC > 1%; TC = 0%). The second pattern, defined as “adaptive-2,” showed focal PD-L1 expression on TCs immediately adjacent to PD-L1–expressing ICs (IC > 1%, TC > 1% to < 25%). Thus, the two adaptive patterns were characterized by the (always present) PD-L1 expression on the ICs but could be differentiated based on the expression of PD-L1 on the TCs.

Two patterns with innate-type PD-L1 expression on the TCs were recognized: the pattern defined as “constitutive” revealed extensive PD-L1 expression on the TCs without concomitant PD-L1 expression on the few existing ICs (TC ≥ 25%, IC = 0%), whereas the pattern defined as “combined” showed extensive PD-L1 expression on the TCs that associated with PD-L1 expression on the adjacent ICs (TC ≥ 25%; IC > 1%). Thus, the two “innate” patterns were characterized by the (always present) extensive PD-L1 expression on the TCs but could be differentiated based on the expression of PD-L1 on the ICs: the “constitutive” pattern showed no PD-L1 expression on the ICs, whereas the “combined” pattern showed PD-L1 expression on ICs.

Of the 38 PD-L1+ cases of the study cohort, 13 (34.2%) were assigned into the “adaptive-1” pattern, 18 (47.3%) into the “adaptive-2” pattern, 4 (10.5%) into the “constitutive” pattern, and 3 (8%) into the “combined” pattern. Of the 67 PD-L1+ PDACs of the validation cohort, 21 cases (31.3%) were assigned into the “adaptive-1” pattern, 34 (50.7%) into the “adaptive-2” pattern, 6 (9%) into the “constitutive” pattern, and 6 (9%) into the “combined” pattern. Our findings demonstrated a concordance between study and validation cohorts regarding the PD-1/PD-L1 staining patterns (P = 0.003). The different PD-1/PD-L1 staining patterns are presented in Supplementary Fig. S2. No PD-1 or PD-L1 expression in either tumor or ICs was observed in the 5 MSI cases.

Immunoarchitectural features and patterns of PD-1/PD-L1 expression

The above-described PD-1/PD-L1 staining patterns associated with distinct immunoarchitectural profiles of the TME. Compared with all other patterns, tumors from “adaptive-1” cases exhibited a T-cell–inflamed TME, rich in CD3+, CD4+, and CD8+ TILs and their coexpressing phenotypes, including CD3+CD8+ and CD3+CD4+ T-cell populations, in both the intraepithelial and stromal compartments, along with diminished populations of CD68+ macrophages, including M2-polarized CD68+CD206+ macrophages. “Adaptive-1” cases showed increased numbers of PD-1+ T cells compared with all other patterns (Figs. 1 and 2; Supplementary Figs. S3 and S4). PD-1–expressing T cells comprised up to 16%, 5%, 0%, and 1% of the intratumoral T-cell population in the “adaptive-1”, “adaptive-2”, “constitutive,” and “combined” patterns, respectively (P = 0.001; Supplementary Fig. S3). The stromal PD-1+ T-cell counts were less numerous in all patterns (0%–3%; Supplementary Fig. S3). PD-1 positivity exhibited a range between weak, intermediate, and strong staining intensity. The strong staining was classified as PD-1high and the weak and intermediate staining as PD-1low. PD-1low phenotypes were more abundant, whereas the PD-1high phenotypes constituted < 10% of the PD-1+ T-cell population across all patterns. The PD-1high T-cell population often (50%) exhibited a double-positive CD4+CD8+ phenotype within CD3+ T cells, whereas the remaining 50% of the PD-1high T cells were CD3+CD4CD8+. In contrast, PD-1low T cells were mostly (90%) CD3+CD4CD8+.

Figure 1.

Stromal immune densities across the PD-1/PD-L1 staining patterns for the study cohort. A, The heatmap of the stromal immune cell densities (per mm2) shows the up- and downregulation of the immune cell populations across the different PD-1/PD-L1 staining patterns of the study cohort (“adaptive-1,” n = 13; “adaptive-2,” n = 18; “constitutive,” n = 4; and “combined,” n = 3), including the 5 MSI cases. Differences in immune cell densities between groups were analyzed using the Kruskal–Wallis test B, The PCA performed using the stromal immune cell densities shows clustering of “adaptive-1,” “adaptive-2,” “constitutive,” and “combined” PD-L1 staining patterns. Each symbol denotes a patient, and 95% confidence ellipses are drawn. The percentage variation is explained by the first principal component (PC1; x-axis) and the second principal component (PC2; y-axis).

Figure 1.

Stromal immune densities across the PD-1/PD-L1 staining patterns for the study cohort. A, The heatmap of the stromal immune cell densities (per mm2) shows the up- and downregulation of the immune cell populations across the different PD-1/PD-L1 staining patterns of the study cohort (“adaptive-1,” n = 13; “adaptive-2,” n = 18; “constitutive,” n = 4; and “combined,” n = 3), including the 5 MSI cases. Differences in immune cell densities between groups were analyzed using the Kruskal–Wallis test B, The PCA performed using the stromal immune cell densities shows clustering of “adaptive-1,” “adaptive-2,” “constitutive,” and “combined” PD-L1 staining patterns. Each symbol denotes a patient, and 95% confidence ellipses are drawn. The percentage variation is explained by the first principal component (PC1; x-axis) and the second principal component (PC2; y-axis).

Close modal
Figure 2.

Representative mIF images of the TME showing the “adaptive-1” pattern. PD-L1+ patient PDAC samples from the study cohort (n = 38) were assessed for TCs (pancytokeratin; cyan), PD-L1 (magenta), CD3+CD8+ T cells (green), PD-1 (red), CD68+ TAMs (orange), and FOXP3+CD4+ Tregs (white). Nuclear counterstain (DAPI, blue) also shown. Magnification, 200×.

Figure 2.

Representative mIF images of the TME showing the “adaptive-1” pattern. PD-L1+ patient PDAC samples from the study cohort (n = 38) were assessed for TCs (pancytokeratin; cyan), PD-L1 (magenta), CD3+CD8+ T cells (green), PD-1 (red), CD68+ TAMs (orange), and FOXP3+CD4+ Tregs (white). Nuclear counterstain (DAPI, blue) also shown. Magnification, 200×.

Close modal

Double-positive CD4+CD8+ T cells (within the CD3+ T-cell population) were significantly expanded in the “adaptive-1” pattern (P < 0.0001; Table 1 and Fig. 1). Although CD3+CD4+FOXP3+ cells (CD4+ Tregs) were reduced, CD3+CD8+FOXP3+ cells (CD8+ Tregs) were increased in “adaptive-1” cases, especially in the stromal compartment (Fig. 1). The “adaptive-2” pattern was characterized by moderate TIL counts, with increased PD-L1+CD3+CD8+ T cells, as well as increased counts of CD4+ Tregs and CD68+ TAMs, including M2-polarized CD68+CD206+ TAMs, compared with the “adaptive-1” pattern (Figs. 1 and 3; Supplementary Figs. S4 and S5). No double-positive CD4+CD8+ T cells were recognized among the PD-L1+ T cells. These findings suggest that both innate and adaptive immune cells act together to promote disease progression in cases with an “adaptive-2” pattern. “Constitutive” cases, on the other hand, were characterized by extensive tumor cell PD-L1 staining but extremely reduced counts of all immune cell populations, apart from a few CD68+ TAMs (Fig. 1; Supplementary Figs. S4 and S6). The few “combined” cases showed extensive PD-L1 expression on the TCs, similar to that observed in the “constitutive” pattern, with increased TIL counts, along with the presence of numerous PD-L1+CD3+CD8+ T cells, especially in the intraepithelial compartment (Figs. 1 and 4; Supplementary Fig. S4). The median counts (range) of all immune cell populations across the different PD-1/PD-L1 patterns are presented in Table 1.

Table 1.

Median (range) percentage values of stromal immune cell infiltrates across PD-1/PD-L1–mediated immune resistance patterns.

% Immune cellsAdaptive-1Adaptive-2ConstitutiveCombinedP value
CD3+ 54.7 (41.28–77.8) 14.155 (7.03–31.03) 3.63 (2.73–3.99) 10.59 (6.12–15.07) <0.00001 
CD4+ 38.96 (23.65–68.93) 13.16 (4.17–22.26) 0.55 (0.2–2.04) 8.2 (7.05–9.35) <0.00001 
CD8+ 21.85 (14.93–36.34) 5.98 (3.34–12.96) 1.75 (1.6–1.96) 4.59 (4.07–5.54) <0.00001 
CD68+ 3.76 (2.47–6.93) 9.35 (4.87–24.64) 8.45 (6.86–12.35) 11.72 (11.12–12.31) 0.000062615 
CD68+CD206+ 2.34 (1.56–4.87) 6.47 (3.25–18.94) 5.39 (3.26–8.17) 8.57 (6.07–11.28) 0.00001 
FOXP3+ 1.44 (0.63–2.53) 2.69 (1.35–4.95) 1.36 (0.78–1.83) 1.38 (1.1–1.74) 0.00011 
CD3+CD4+ 34.76 (21.22–64.47) 6.93 (3.067–16.79) 1.25 (0.1–1.59) 5.51 (4.15–6.86) <0.00001 
CD3+CD8+ 18.92 (11.38–35.87) 5.24 (3.94–11.11) 1.27 (1.11–1.53) 4.83 (3.51–5.14) <0.00001 
CD3+CD4+CD8+ 9.67 (5.41–12.14) 4.55 (2.6–10.72) 1.2 (1.01–1.91) 2.47 (1.49–4.45) 0.00001533 
CD3+CD4+FOXP3+ 1.37 (0.2–2.53) 2.52 (1.13–4.95) 0.6 (0.17–0.76) 0.89 (0.7–1.3) 0.000085517 
CD3+CD8+FOXP3+ 0.61 (0.07–1.26) 0.14 (0–0.82) 0.1 (0.02–0.45) 0.1 (0–0.2) 0.0019902 
PD1+CD3+CD8+ 1.25 (0.17–5.98) 0.16 (0–0.86) 0.045 (0–0.13) 0.51 (0.2–0.84) 0.000062615 
PD-L1+CD3+CD8+ 0.04 (0–0.82) 2.17 (1.08–3.76) 0 (0–0) 0.64 (0.47–0.81) <0.00001 
PD-L1+CD68+ 2.35 (0.2–5.67) 4.19 (1.55–12.8) 0.8 (0.37–0.93) 1.08 (1.02–1.99) 0.00042488 
% Immune cellsAdaptive-1Adaptive-2ConstitutiveCombinedP value
CD3+ 54.7 (41.28–77.8) 14.155 (7.03–31.03) 3.63 (2.73–3.99) 10.59 (6.12–15.07) <0.00001 
CD4+ 38.96 (23.65–68.93) 13.16 (4.17–22.26) 0.55 (0.2–2.04) 8.2 (7.05–9.35) <0.00001 
CD8+ 21.85 (14.93–36.34) 5.98 (3.34–12.96) 1.75 (1.6–1.96) 4.59 (4.07–5.54) <0.00001 
CD68+ 3.76 (2.47–6.93) 9.35 (4.87–24.64) 8.45 (6.86–12.35) 11.72 (11.12–12.31) 0.000062615 
CD68+CD206+ 2.34 (1.56–4.87) 6.47 (3.25–18.94) 5.39 (3.26–8.17) 8.57 (6.07–11.28) 0.00001 
FOXP3+ 1.44 (0.63–2.53) 2.69 (1.35–4.95) 1.36 (0.78–1.83) 1.38 (1.1–1.74) 0.00011 
CD3+CD4+ 34.76 (21.22–64.47) 6.93 (3.067–16.79) 1.25 (0.1–1.59) 5.51 (4.15–6.86) <0.00001 
CD3+CD8+ 18.92 (11.38–35.87) 5.24 (3.94–11.11) 1.27 (1.11–1.53) 4.83 (3.51–5.14) <0.00001 
CD3+CD4+CD8+ 9.67 (5.41–12.14) 4.55 (2.6–10.72) 1.2 (1.01–1.91) 2.47 (1.49–4.45) 0.00001533 
CD3+CD4+FOXP3+ 1.37 (0.2–2.53) 2.52 (1.13–4.95) 0.6 (0.17–0.76) 0.89 (0.7–1.3) 0.000085517 
CD3+CD8+FOXP3+ 0.61 (0.07–1.26) 0.14 (0–0.82) 0.1 (0.02–0.45) 0.1 (0–0.2) 0.0019902 
PD1+CD3+CD8+ 1.25 (0.17–5.98) 0.16 (0–0.86) 0.045 (0–0.13) 0.51 (0.2–0.84) 0.000062615 
PD-L1+CD3+CD8+ 0.04 (0–0.82) 2.17 (1.08–3.76) 0 (0–0) 0.64 (0.47–0.81) <0.00001 
PD-L1+CD68+ 2.35 (0.2–5.67) 4.19 (1.55–12.8) 0.8 (0.37–0.93) 1.08 (1.02–1.99) 0.00042488 
Figure 3.

Representative mIF images of the TME showing the “adaptive-2” pattern. PD-L1+ patient PDAC samples (n = 38) from the study cohort were assessed for PD-L1 (magenta), TCs (pancytokeratin, cyan), CD3+CD8+ T cells (green), PD-1 (red), CD68+ TAMs (orange), and FOXP3+CD4+ Tregs (white). Nuclear counterstain (DAPI, blue) also shown. Magnification, 200×.

Figure 3.

Representative mIF images of the TME showing the “adaptive-2” pattern. PD-L1+ patient PDAC samples (n = 38) from the study cohort were assessed for PD-L1 (magenta), TCs (pancytokeratin, cyan), CD3+CD8+ T cells (green), PD-1 (red), CD68+ TAMs (orange), and FOXP3+CD4+ Tregs (white). Nuclear counterstain (DAPI, blue) also shown. Magnification, 200×.

Close modal
Figure 4.

Representative mIF images of the TME showing the “combined” pattern. PD-L1+ patient PDAC samples (n = 38) from the study cohort were assessed for PD-L1 (magenta), TCs (pancytokeratin, cyan), CD3+CD8+ T cells (green), PD-1 (red), CD68+ TAMs (orange), and FOXP3+CD4+ Tregs (white). Nuclear counterstain (DAPI, blue) also shown. Magnification, 200×.

Figure 4.

Representative mIF images of the TME showing the “combined” pattern. PD-L1+ patient PDAC samples (n = 38) from the study cohort were assessed for PD-L1 (magenta), TCs (pancytokeratin, cyan), CD3+CD8+ T cells (green), PD-1 (red), CD68+ TAMs (orange), and FOXP3+CD4+ Tregs (white). Nuclear counterstain (DAPI, blue) also shown. Magnification, 200×.

Close modal

We further applied PCA, which resulted in distinct separation of both intraepithelial and stromal immune cell profiles across the different PD-L1 staining patterns (Fig. 1B; Supplementary Fig. S4B). Taken together, these results showed that PD-L1 expression on tumor and ICs associated with the immune contexture of the TME.

Immunoarchitectural features of MSI-H cases

MSI cases were characterized by an inflamed TME, with increased density of TILs, Tregs, and TAMs, which was more pronounced in the stromal compartment (Fig. 1; Supplementary Figs. S4 and S7A and S7B). The stromal infiltrates displayed a high density of all TIL subtypes. CD68+ macrophages were moderately increased in both intraepithelial and stromal compartments. There was some similarity with “adaptive-1” cases considering the abundance of T cells. However, CD68+ macrophages and FOXP3+ Tregs were more numerous in MSI cases compared with “adaptive-1” cases. All MSI cases were negative for PD-1 and PD-L1, although some degenerative PD-L1+ ICs were often seen within the glandular lumina (Supplementary Fig. S7A and S7B).

Intratumoral heterogeneity

Our results underline a high degree of intertumoral heterogeneity. However, to better understand the degree of intratumoral heterogeneity, we further analyzed the density and distribution of the immune cell infiltrates in four regions of interest (ROI) by creating virtual annotations corresponding to a diameter of 1 mm (area of 0.79 mm2) in randomly selected areas of each tumor (Supplementary Fig. S8A). All cases displayed significant heterogeneity regarding the distribution pattern and density of the immune cell infiltrates. “Adaptive-1” and “constitutive” cases presented mostly quantitative differences among ROIs with significant variation in the counts of the immune cell infiltrates (Supplementary Fig. S8B). However, the characteristics of the immune response were preserved, with clear predominance of TILs over Tregs and TAMs in all ROIs of “adaptive-1” tumors and predominance of TAMs and Tregs over TILs in the sparse immune cell infiltrates in all ROIs of the “constitutive” cases. On the other hand, the “adaptive-2,” and to a lesser degree “combined,” cases displayed not only quantitative differences in the density of the immune cell infiltrates, but also significant variation regarding the characteristics of the immune response. Thus, the relative proportion of both TILs and TAMs showed great variation among the different ROIs of the same tumor. In some ROIs, TAMs were the predominant cell population, whereas in ROIs with a higher density of stromal CD3+ T cells, a significant proportion of the CD3+CD4+ T cells coexpressed FOXP3, corresponding to Tregs (Supplementary Fig. S7). In the “adaptive-2” pattern, the density of the immune infiltrates was higher in stromal areas surrounding PD-L1+ TCs. The five MSI cases showed, similarly to “adaptive-1” cases, mostly quantitative differences among the ROIs; however, their microenvironment displayed higher numbers of almost all ICs compared with the “adaptive-1” cases (Supplementary Fig. S8D). When we compared the median values of the immune cell infiltrates across all four ROIs to those of the WTSs, no statistically significant differences were observed in all cases (Supplementary Fig. S8C and S8D).

Proximity analysis

Proximity histograms revealed that all immune cell populations were significantly closer to PD-L1 than to PD-L1+ TCs (P < 0.0001; Supplementary Table S5). However, PD-L1+CD3+CD8+ T cells were closer to PD-L1+ TCs (median distance = 90.11 μm) compared with their PD-L1CD3+CD8+ T-cell counterparts (median distance = 336.34 μm; P = 0.001; Supplementary Table S4). Representative histogram plots depicting the distance of PD-L1+ and PD-L1 cytotoxic T-cell populations to PD-L1+ and PD-L1 TCs are presented in Fig. 5.

Figure 5.

Representative proximity histograms of the PD-L1+ PDAC samples of the study cohort (n = 38). The data were binned at set distances (concentric circles of 50 μm, radius of 25 μm), and histogram plots for each immune cell population were created (HALO image analysis platform; Indica Labs). The y-axis indicates the number of the PD-L1CD8+ or PD-L1+CD8+ T cells, and the x-axis indicates the distance of the above ICs from the PD-L1 and PD-L1+ TCs in increments of 10 μm. Each bar represents the number of CD8+ T cells found in a certain distance from the PD-L1 and PD-L1+ TCs.

Figure 5.

Representative proximity histograms of the PD-L1+ PDAC samples of the study cohort (n = 38). The data were binned at set distances (concentric circles of 50 μm, radius of 25 μm), and histogram plots for each immune cell population were created (HALO image analysis platform; Indica Labs). The y-axis indicates the number of the PD-L1CD8+ or PD-L1+CD8+ T cells, and the x-axis indicates the distance of the above ICs from the PD-L1 and PD-L1+ TCs in increments of 10 μm. Each bar represents the number of CD8+ T cells found in a certain distance from the PD-L1 and PD-L1+ TCs.

Close modal

Prognostic significance of PD-1/PD-L1 patterns

Kaplan–Meier analysis of patients' outcome was performed. Significant differences were observed among the OS of the different PD-1/PD-L1 staining patterns in both study and validation cohorts (P = 0.015 and P < 0.0001, respectively; Fig. 6A), with the “constitutive” pattern exhibiting the worse outcome and “adaptive-1” pattern showing improved outcome in both cohorts. Compared with “adaptive-2” cases, “adaptive-1” cases exhibited significantly longer OS in the validation cohort (P = 0.009; Fig. 6B), although this difference did not achieve statistical significance in the fewer cases of the study cohort (P = 0.382). PD-L1+ tumors had longer OS than PD-L1 PDACs, whereas MSI cases had better OS than both PD-L1 categories (Fig. 6C). This, however, should be carefully interpreted due to the very small number of MSI cases. To determine if the immunologic features of the PD-1/PD-L1 patterns could independently predict OS, we performed multivariate Cox regression analysis, including PD-1/PD-L1 staining patterns as distinct variables and primary clinical parameters, including postoperative adjuvant chemotherapy (Supplementary Table S6). In addition to clinical variables such as T- and N-stages, PD-1/PD-L1 patterns were identified as independent predictive variables.

Figure 6.

Kaplan–Meier curves. A, Kaplan–Meier curves comparing the OS of PDAC cases with “adaptive-1” (light blue; study, n = 13; validation, n = 21), “adaptive-2” (magenta; study, n = 18; validation, n = 34), “constitutive” (green; study, n = 4; validation, n = 6), “combined” (orange; study, n = 3; validation, n = 6), and negative (yellow green; study, n = 80; validation, n = 159) PD-L1 staining patterns for the study and validation cohorts. B, Kaplan–Meier curves comparing the OS of PDAC cases with “adaptive-1” (light blue; study, n = 13; validation, n = 21) and “adaptive-2” (magenta; study, n = 18; validation, n = 34) PD-L1 staining patterns for the study and validation cohorts. C, Kaplan–Meier curves comparing the OS of all PD-L1–positive (light blue, n = 105) and all PD-L1–negative (magenta, n = 239) PDAC cases without (left) and with (right) MSI cases (green, n = 5). Statistical comparisons were performed using the log-rank test.

Figure 6.

Kaplan–Meier curves. A, Kaplan–Meier curves comparing the OS of PDAC cases with “adaptive-1” (light blue; study, n = 13; validation, n = 21), “adaptive-2” (magenta; study, n = 18; validation, n = 34), “constitutive” (green; study, n = 4; validation, n = 6), “combined” (orange; study, n = 3; validation, n = 6), and negative (yellow green; study, n = 80; validation, n = 159) PD-L1 staining patterns for the study and validation cohorts. B, Kaplan–Meier curves comparing the OS of PDAC cases with “adaptive-1” (light blue; study, n = 13; validation, n = 21) and “adaptive-2” (magenta; study, n = 18; validation, n = 34) PD-L1 staining patterns for the study and validation cohorts. C, Kaplan–Meier curves comparing the OS of all PD-L1–positive (light blue, n = 105) and all PD-L1–negative (magenta, n = 239) PDAC cases without (left) and with (right) MSI cases (green, n = 5). Statistical comparisons were performed using the log-rank test.

Close modal

In the present study, we showed that in patients with resected, MSS pancreatic cancer, patient outcomes and immunoarchitectural features of the TME are associated with the patterns of PD-L1 expression by tumor and ICs. This has potential to stratify patients in predictive and prognostic subgroups.

The “adaptive-1” pattern displayed an inflamed TME with high TIL counts, enriched for cytotoxic CD3+CD8+ T lymphocytes in both the intraepithelial and stromal compartments and the best survival in our cohort. We and others (21, 22) have shown that up to 40% of resected human PDACs are rich in cytotoxic CD8+ T cells and that these T cells exhibit features of recent or prolonged T-cell receptor signaling (23). This “hot” TME may be attributed to both T-cell-mediated recognition of tumor antigens and the recent genomic identification of a putative immunogenic subset of PDAC, although the gut microbiota has also been reported to play an important role in shaping immune responses (24–26). CD8+ T-cell abundance is predictive of better outcome in PDAC patients (19, 27) and is known to favor clinical responses following immune-checkpoint inhibitors in many solid tumors (28, 29). Nevertheless, a significant subset of the CD3+CD8+ T cells with the “adaptive-1” pattern displayed PD-1+CD3+CD8+ phenotypes, known to dampen excessive T-cell activation (30, 31). A possible explanation for this could be that although tumor antigen-specific T cells can accumulate in the TME of PDAC, a low precursor frequency of neoepitope-specific T cells may result in preservation of neoantigens and induction of T-cell exhaustion. This was shown by Burrack and colleagues (32) in a mouse PDAC model. PD-1+ cells generally represent the target population for anti-PD-1 therapies (33). However, it has also been shown that features of exhaustion are predictive of relatively more functionality compared to non-exhausted, likely non-responding, T cells. For example, in human tumors, TILs expressing markers of exhaustion, such as PD-1, are more likely to express cytokines such as IFNγ (34). These data suggest that it is specifically the “exhausted” CD8+ T cells that are exerting residual control over tumor growth (35). According to recent reports (36), there exist distinct subpopulations of CD8+ T cells exhibiting a range of PD-1 expression, with high PD-1 expression (PD-1highCD8+) potentially predictive for anti–PD-1 therapeutic response. However, in our study, PD-1+ cells constituted merely a small percentage of the CD3+CD8+ T-cell population in all cases, (i.e., up to 16% in the intratumoral and up to 3% in the stromal compartment). This finding is consistent with reported results by Liudahl and colleagues (33) and may, at least partly, explain the lack of response to checkpoint inhibitor therapies observed in PDAC patients. PD-1high T cells were even less, constituting only between 2% and 7% of the PD-1+CD8+ population, in all patterns, including “adaptive-1”, whereas the remaining PD-1+CD8+ T cells belonged to the PD-1low T-cell population. However, PD-1high T cells were frequently (50%) observed in double-positive CD4+CD8+ T cells, which was significantly expanded in “adaptive-1” cases. It has been shown that double-positive T cells have an effector memory phenotype and express CD38, 4–1BB, and HLA-DR, suggesting antigen-driven expansion (37). Their frequent PD-1 positivity suggests that double-positive CD4+CD8+ T cells might represent tumor-specific T cells that could be reactivated by checkpoint inhibitors.

The “adaptive-1” pattern was characterized by low numbers of PD-L1+CD3+CD8+ T cells and absence of PD-L1+ TCs. This is in line with the results of Danilova and colleagues (38), who, by using clinical survival and RNA expression data from The Cancer Genome Atlas, showed that the combination of PD-L1 and high CD8 expression identified a subtype with favorable survival in PDAC. Together, all the above suggests that, immune-checkpoint blockade alone might not be appropriate for these patients, whereas treatments including agents to promote immunogenicity and priming of T cells might be more effective (39, 40).

This study showed an increased presence of CD8+ Tregs (CD3+CD8+FOXP3+ T cells) in the TME of tumors with an “adaptive-1” pattern, whereas CD4+Tregs were more abundant in the TME of “adaptive-2” cases. Interestingly, CD8+ Tregs have been shown to control memory responses more efficiently than CD4+ Tregs, whereas CD4+ Tregs were more efficient than CD8+ Tregs at controlling naïve immune responses (41, 42). These findings underscore the distinct TME profiles between the two adaptive patterns, suggesting strategic differences in the quality of the immunologic response. In mutant mice that harbor a point mutation in the MHC class Ib molecule, Qa-1, which impairs CD8+ Treg suppressive function, the resulting disruption of CD8+ Treg activity led to delayed B16 melanoma tumor growth and associated with expansion of follicular T helper cells and germinal center B cells, high titers of antitumor autoantibodies, and an increase in the ratio of CD8+ T effectors to CD8+ Tregs, indicating a robust antitumor immune response (43). Thus, inactivation of CD8+ Tregs may represent a further strategy to enhance antitumor immunity in patients with an “adaptive-1” pattern.

Similarities regarding the presence of an inflamed TME were observed between “adaptive-1” and MSI cases, which were more pronounced in the stromal compartment. However, although TILs were the predominant leukocyte population in both “adaptive-1” and MSI cases, the latter additionally displayed an increase in almost all immune cell populations, including TAMs and Tregs, creating an exceptionally “hot” and cellular TME. MSI cases were not found to express PD-1 or PD-L1 in viable tumor or ICs, although some degenerative PD-L1+ ICs were often observed within the lumina of the neoplastic glands. This might be a further indication of an immune reaction which eliminates dysfunctional ICs.

The “adaptive-2” pattern combined the presence of moderately increased counts of PD-L1–expressing immune and TCs. Data from preclinical models suggest that PD-L1 expression on TCs and CD8+ T cells can modulate T-cell function to a similar degree, equally inducing an immunosuppressive TME (44). This pattern was additionally characterized by increased numbers of CD68+ TAMs, including subpopulations of PD-L1+CD68+ TAMs. Myeloid cells, including TAMs, are known to regulate the expression of PD-L1 on TCs, which then suppress antitumor immune responses mediated by CD8+ T cells (45). Alternatively, previous studies have shown that PD-L1+ T cells can induce an M2 macrophage polarization, which has further inhibitory effects on adaptive antitumor immunity (46), indicating cross-regulation across different immune cell subsets. In support of this, we showed that the TME of these tumors was particularly enriched with stromal CD68+CD206+ macrophages (putative M2 macrophages). The phenotype and maintenance of tumor-infiltrating myeloid cells, including TAMs, are determined by cellular interactions and signals within the TME, and CSF1/CSF-1R is an important factor for the survival of resident macrophages (47). Blocking CSF-1R in vivo produces a loss of intratumoral macrophages in PDAC, leading to improved antitumor responses with immune-checkpoint inhibitors (48, 49). This suggests that especially in PDACs with an “adaptive-2” pattern, effective T-cell responses may be additionally restricted by immunosuppressive TAMs, a finding also observed in several preclinical settings (44, 49, 50). Together, these findings indicate that PD-L1+ ICs can promote a tolerogenic, tumor-promoting TME in various ways and render patients with an “adaptive-2” pattern potential candidates for treatment strategies which combine PD-L1 checkpoint blockade with CSF-1R inhibitors or anti-CD40 antibodies, which would help repolarizing macrophages toward an M1 phenotype.

We additionally examined the regional variability of the immune cell infiltrates by creating four virtual annotations in randomly selected areas of each tumor and found significant intratumoral heterogeneity in all cases. This was even more striking in “adaptive-2” and “combined” tumors, which revealed considerable spatial variability regarding both the counts and the relative abundance of the immune cell infiltrates among the annotated tumor areas. Our findings are in agreement with Liudahl and colleagues (33), who, by profiling the density and spatial distribution of myeloid and lymphoid cells in 135 PDACs at single-cell resolution, provided an immune atlas of PDAC heterogeneity. Although in our study the median values of the immune cell infiltrates of all four annotated regions closely resembled those of the WTSs, these findings indicate regional differences in the recruitment of immune cell populations, which by creating different selection pressures may guide tumor progression and impede therapeutic effects.

In the “adaptive-2” and “combined” patterns, the expression of PD-L1 on the TCs was associated with presence of numerous immunosuppressive PD-L1+CD3+CD8+ T cells. Although all immune cell populations were found to be significantly closer to PD-L1 compared with PD-L1+ TCs, the PD-L1+ subset of the CD3+CD8+ T cells was significantly closer to PD-L1+ TCs than their PD-L1CD3+CD8+ T-cell counterparts. The above findings suggest that the upregulation of PD-L1 in a subset of PDACs is temporally and spatially correlated to the presence of PD-L1 on cytotoxic T cells, analogous to similar observations in other solid tumors, including melanoma and lung cancer (7, 11, 51). This further implies that adaptive immune resistance may be partly responsible for the PD-L1 upregulation in a number of PDACs. However, the overall counts of TILs, including CD3+CD8+ and CD3+CD4+ T-cell subpopulations and double-positive CD4+CD8+ phenotypes, decreased with increasing presence of PD-L1 in the TCs. This is best illustrated in the “constitutive” pattern, which was characterized by extensive expression of PD-L1 on the TCs, accompanied by an immune-deserted TME, rendering immune-checkpoint inhibition ineffective in these tumors. Moreover, in the “adaptive-1” pattern and in MSI cases, an inflamed TME was present, however, this was not accompanied by PD-L1 upregulation in TCs. These findings are consistent with the results of Winograd and colleagues (52) in murine PDACs and suggest that PD-L1 expression on PDAC cells is, unlike other solid tumors (7, 11–13), not necessarily activated through cytokine-regulated adaptive immune responses, following increasing immune pressure in the TME. In fact, oncogenic alterations are known to modulate PD-L1 expression (15–17). Similar to previous reports in other tumor types (53, 54), TP53 and PIK3CA mutations were found to be more frequent in PD-L1+ than in PD-L1 tumors in our cohort; however, these values did not achieve statistical significance. They, nevertheless, showed that the PIK3CA/PTEN/AKT pathway may be altered in as many as 11% of PD-L1+ PDACs, indicating a potential additional benefit with PI3K/AKT/mTOR inhibitors for these patients (55). Notwithstanding, in the few PDACs with a “combined” pattern, the extensive PD-L1 expression on the TCs was accompanied by a moderately inflamed TME and an improved OS compared with the “constitutive” pattern cases. This suggests that these patients might be potential candidates for treatments that include PD-1/PD-L1 antagonists.

Overall, our findings underscore the genetic and biological variability that exists in PDAC and indicate that a subset of patients with MSS pancreatic cancer might derive clinical benefit from treatments, including checkpoint inhibitors. Because immunotherapies continue to be refined, these findings may help in paving the path toward more individualized and precision treatments for pancreatic cancer patients.

No disclosures were reported.

The funders had no involvement in the study design; in the collection, analysis, and interpretation of the data; in the writing of the report; and in the decision to submit the paper for publication.

E. Karamitopoulou: Conceptualization, resources, supervision, funding acquisition, methodology, writing–original draft. A. Andreou: Data curation, software, formal analysis, validation, writing–review and editing. A. Pahud de Mortanges: Data curation. M. Tinguely: Resources, data curation, writing–review and editing. B. Gloor: Resources, funding acquisition, writing–review and editing. A. Perren: Conceptualization, resources, writing–review and editing.

This work was supported in part by the Foundation for Clinical-Experimental Tumor-Research (to E. Karamitopoulou) and Insular Foundation (to B. Gloor). The authors also wish to thank Animesh Acharjee for expert statistical support and Miguel Hinojosa for expert technical assistance.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Beatty
GL
,
Eghbali
S
,
Kim
R
. 
Deploying immunotherapy in pancreatic cancer: defining mechanisms of response and resistance
.
Am Soc Clin Oncol Educ Book
2017
;
37
:
267
78
.
2.
Brahmer
JR
,
Tykodi
SS
,
Chow
LQ
,
Hwu
WJ
,
Topalian
SL
,
Hwu
P
, et al
Safety and activity of anti-PD-L1 antibody in patients with advanced cancer
.
N Engl J Med
2012
;
366
:
2455
65
.
3.
Aglietta
M
,
Barone
C
,
Sawyer
MB
,
Moore
MJ
,
Miller
WH
 Jr
,
Bagalà
C
, et al
A phase I dose escalation trial of tremelimumab (CP-675,206) in combination with gemcitabine in chemotherapy-naive patients with metastatic pancreatic cancer
.
Ann Oncol
2014
;
25
:
1750
5
.
4.
Balachandran
VP
,
Beatty
GL
,
Dougan
SK
. 
Broadening the impact of immunotherapy to pancreatic cancer: challenges and opportunities
.
Gastroenterology
2019
;
156
:
2056
72
.
5.
Sahin
IH
,
Askan
G
,
Hu
ZI
,
O'Reilly
EM
. 
Immunotherapy in pancreatic ductal adenocarcinoma: an emerging entity?
Ann Oncol
2017
;
28
:
2950
61
.
6.
Lutz
ER
,
Wu
AA
,
Bigelow
E
,
Sharma
R
,
Mo
G
,
Soares
K
, et al
Immunotherapy converts nonimmunogenic pancreatic tumors into immunogenic foci of immune regulation
.
Cancer Immunol Res
2014
;
2
:
616
31
.
7.
Taube
JM
,
Galon
J
,
Sholl
LM
,
Rodig
SJ
,
Cottrell
TR
,
Giraldo
NA
, et al
Implications of the tumor immune microenvironment for staging and therapeutics
.
Mod Pathol
2018
;
31
:
214
34
.
8.
Gajewski
TF
,
Schreiber
H
,
Fu
YX
. 
Innate and adaptive immune cells in the tumor microenvironment
.
Nat Immunol
2013
;
14
:
1014
22
.
9.
Vesely
MD
,
Kershaw
MH
,
Schreiber
RD
,
Smyth
MJ
. 
Natural innate and adaptive immunity to cancer
.
Annu Rev Immunol
2011
;
29
:
235
71
.
10.
Taube
JM
,
Anders
RA
,
Young
GD
,
Xu
H
,
Sharma
R
,
McMiller
TL
, et al
Colocalization of inflammatory response with B7-h1 expression in human melanocytic lesions supports an adaptive resistance mechanism of immune escape
.
Sci Transl Med
2012
;
4
:
127
37
.
11.
Velcheti
V
,
Schalper
KA
,
Carvajal
DE
,
Anagnostou
VK
,
Syrigos
KN
,
Sznol
M
, et al
Programmed death ligand-1 expression in non-small cell lung cancer
.
Lab Invest
2014
;
94
:
107
16
.
12.
Cimino-Mathews
A
,
Thompson
E
,
Taube
JM
,
Ye
X
,
Lu
Y
,
Meeker
A
, et al
PD-L1 (B7-H1) expression and the immune tumor microenvironment in primary and metastatic breast carcinomas
.
Hum Pathol
2016
;
47
:
52
63
.
13.
Yanik
EL
,
Kaunitz
GJ
,
Cottrell
TR
,
Succaria
F
,
McMiller
TL
,
Ascierto
ML
, et al
Association of HIV status with local immune response to anal squamous cell carcinoma: implications for immunotherapy
.
JAMA Oncol
2017
;
3
:
974
8
.
14.
Topalian
SL
,
Taube
JM
,
Anders
RA
,
Pardoll
DM
. 
Mechanism-driven biomarkers to guide immune checkpoint blockade in cancer therapy
.
Nat Rev Cancer
2016
;
16
:
275
87
.
15.
Xu
C
,
Fillmore
CM
,
Koyama
S
,
Wu
H
,
Zhao
Y
,
Chen
Z
, et al
Loss of Lkb1 and Pten leads to lung squamous cell carcinoma with elevated PD-L1 expression
.
Cancer Cell
2014
;
25
:
590
604
.
16.
Parsa
AT
,
Waldron
JS
,
Panner
A
,
Crane
CA
,
Parney
IF
,
Barry
JJ
, et al
Loss of tumor suppressor PTEN function increases B7-H1 expression and immunoresistance in glioma
.
Nat Med
2007
;
13
:
84
8
.
17.
Van Allen
EM
,
Golay
HG
,
Liu
Y
,
Koyama
S
,
Wong
K
,
Taylor-Weiner
A
, et al
Long-term benefit of PD-L1 blockade in lung cancer associated with JAK3 activation
.
Cancer Immunol Res
2015
;
3
:
855
63
.
18.
Pardoll
DM
. 
The blockade of immune checkpoints in cancer immunotherapy
.
Nat Rev Cancer
2012
;
12
:
252
64
.
19.
Postow
MA
,
Sidlow
R
,
Hellmann
MD
. 
Immune-related adverse events associated with immune checkpoint Blockade
.
N Engl J Med
2018
;
378
:
158
68
.
20.
Vasaturo
A
,
Galon
J
. 
Multiplexed immunohistochemistry for immune cell phenotyping, quantification and spatial distribution in situ
.
Methods Enzymol
2020
;
635
:
51
66
.
21.
Wartenberg
M
,
Cibin
S
,
Zlobec
I
,
Vassella
E
,
Eppenberger-Castori
S
,
Terracciano
L
, et al
Integrated genomic and immunophenotypic classification of pancreatic cancer reveals three distinct subtypes with prognostic/predictive significance
.
Clin Cancer Res
2018
;
24
:
4444
54
.
22.
Mahajan
UM
,
Langhoff
E
,
Goni
E
,
Costello
E
,
Greenhalf
W
,
Halloran
C
, et al
Immune cell and stromal signature associated with progression-free survival of patients with resected pancreatic ductal adenocarcinoma
.
Gastroenterology
2018
;
155
:
1625
39
.
23.
Carstens
JL
,
Correa de Sampaio
P
,
Yang
D
,
Barua
S
,
Wang
H
,
Rao
A
, et al
Spatial computation of intratumoral T cells correlates with survival of patients with pancreatic cancer
.
Nat Commun
2017
;
8
:
15095
.
24.
Stromnes
IM
,
Hulbert
A
,
Pierce
RH
,
Greenberg
PD
,
Hingorani
SR
. 
T-cell localization, activation, and clonal expansion in human pancreatic ductal adenocarcinoma
.
Cancer Immunol Res
2017
;
5
:
978
1
.
25.
Bailey
P
,
Chang
DK
,
Nones
K
,
Johns
AL
,
Patch
AM
,
Gingras
MC
, et al
Genomic analyses identify molecular subtypes of pancreatic cancer
.
Nature
2016
;
531
:
47
52
.
26.
Riquelme
E
,
Zhang
Y
,
Zhang
L
,
Montiel
M
,
Zoltan
M
,
Dong
W
, et al
Tumor microbiome diversity and composition influence pancreatic cancer outcomes
.
Cell
2019
;
178
:
795
806
.
27.
Puleo
F
,
Nicolle
R
,
Blum
Y
,
Cros
J
,
Marisa
L
,
Demetter
P
, et al
Stratification of pancreatic ductal adenocarcinomas based on tumor and microenvironment features
.
Gastroenterology
2018
;
155
:
1999
2013
.
28.
Balachandran
VP
,
Łuksza
M
,
Zhao
JN
,
Makarov
V
,
Moral
JA
,
Remark
R
, et al
Identification of unique neoantigen qualities in long-term survivors of pancreatic cancer
.
Nature
2017
;
551
:
512
6
.
29.
Cristescu
R
,
Mogg
R
,
Ayers
M
,
Murphy
E
,
Yearley
J
,
Sher
X
, et al
Pan-tumor genomic biomarkers for PD-1 checkpoint blockade-based immunotherapy
.
Science
2018
;
362
:
eaar3593
.
Erratum in: Science 2019;363:eaax1384
.
30.
Tumeh
PC
,
Harview
CL
,
Yearley
JH
, et al
PD-1 blockade induces responses by inhibiting adaptive immune resistance
.
Nature
2014
;
515
:
568
71
.
31.
Ambler
R
,
Edmunds
GL
,
Tan
SL
,
Cirillo
S
,
Pernes
JI
,
Ruan
X
, et al
PD-1 suppresses the maintenance of cell couples between cytotoxic T cells and target tumor cells within the tumor
.
Sci Signal
2020
;
13
:
eaau4518
.
32.
Burrack
AL
,
Spartz
EJ
,
Raynor
JF
,
Wang
I
,
Olson
M
,
Stromnes
IM
. 
Combination PD-1 and PD-L1 blockade promotes durable neoantigen-specific T cell-mediated immunity in pancreatic ductal adenocarcinoma
.
Cell Rep
2019
;
28
:
2140
55
.
33.
Liudahl
SM
,
Betts
CB
,
Sivagnanam
S
,
Morales-Oyarvide
V
,
da Silva
A
,
Yuan
C
, et al
Leukocyte heterogeneity in pancreatic ductal adenocarcinoma: phenotypic and spatial features associated with clinical outcome
.
Cancer Discov
2021
;
11
:
2014
31
.
34.
Gros
A
,
Robbins
PF
,
Yao
X
,
Li
YF
,
Turcotte
S
,
Tran
E
, et al
PD-1 identifies the patient-specific CD8+ tumor-reactive repertoire infiltrating human tumors
.
J Clin Invest
2014
;
124
:
2246
59
.
35.
Collier
JL
,
Weiss
SA
,
Pauken
KE
,
Sen
DR
,
Sharpe
AH
. 
Not-so-opposite ends of the spectrum: CD8+ T cell dysfunction across chronic infection, cancer and autoimmunity
.
Nat Immunol
2021
;
22
:
809
19
.
36.
Kim
HD
,
Song
GW
,
Park
S
,
Jung
MK
,
Kim
MH
,
Kang
HJ
, et al
Association between expression level of PD1 by tumor-infiltrating CD8+ T cells and features of hepatocellular carcinoma
.
Gastroenterology
2018
;
155
:
1936
50
.
37.
Menard
LC
,
Fischer
P
,
Kakrecha
B
,
Linsley
PS
,
Wambre
E
,
Liu
MC
, et al
Renal cell carcinoma (RCC) tumors display large expansion of double positive (DP) CD4+CD8+ T cells with expression of exhaustion markers
.
Front Immunol
2018
;
9
:
2728
.
38.
Danilova
L
,
Ho
WJ
,
Zhu
Q
,
Vithayathil
T
,
De Jesus-Acosta
A
,
Azad
NS
, et al
Programmed cell death ligand-1 (PD-L1) and CD8 expression profiling identify an immunologic subtype of pancreatic ductal adenocarcinomas with favorable survival
.
Cancer Immunol Res
2019
;
7
:
886
95
.
39.
Vonderheide
RH
. 
The immune revolution: a case for priming, not checkpoint
.
Cancer Cell
2018
;
33
:
563
9
.
40.
Morrison
AH
,
Diamond
MS
,
Hay
CA
,
Byrne
KT
,
Vonderheide
RH
. 
Sufficiency of CD40 activation and immune checkpoint blockade for T cell priming and tumor immunity
.
Proc Natl Acad Sci U S A
2020
;
117
:
8022
31
.
41.
Long
X
,
Cheng
Q
,
Liang
H
,
Zhao
J
,
Wang
J
,
Wang
W
, et al
Memory CD4+ T cells are suppressed by CD8+regulatory T cells in vitro and in vivo
.
Am J Transl Res
2017
;
9
:
63
78
.
42.
Flippe
L
,
Bézie
S
,
Anegon
I
,
Guillonneau
C
. 
Future prospects for CD8+regulatory T cells in immune tolerance
.
Immunol Rev
2019
;
292
:
209
24
.
43.
Arias
DAA
,
Kim
HJ
,
Zhou
P
,
Holderried
TA
,
Wang
X
,
Dranoff
G
, et al
Disruption of CD8+ Treg activity results in expansion of T follicular helper cells and enhanced antitumor immunity
.
Cancer Immunol Res
2014
;
2
:
207
16
.
44.
Lau
J
,
Cheung
J
,
Navarro
A
,
Lianoglou
S
,
Haley
B
,
Totpal
K
, et al
Tumour and host cell PD-L1 is required to mediate suppression of antitumour immunity in mice
.
Nat Commun
2017
;
8
:
14572
.
45.
Zhang
Y
,
Velez-Delgado
A
,
Mathew
E
,
Li
D
,
Mendez
FM
,
Flannagan
K
, et al
Myeloid cells are required for PD-1/PD-L1 checkpoint activation and the establishment of an immunosuppressive environment in pancreatic cancer
.
Gut
2017
;
66
:
124
36
.
46.
Diskin
B
,
Adam
S
,
Cassini
MF
,
Sanchez
G
,
Liria
M
,
Aykut
B
, et al
PD-L1 engagement on T cells promotes self-tolerance and suppression of neighboring macrophages and effector T cells in cancer
.
Nat Immunol
2020
;
21
:
442
54
.
47.
Candido
JB
,
Morton
JP
,
Bailey
P
,
Campbell
AD
,
Karim
SA
,
Jamieson
T
, et al
CSF1R+macrophages sustain pancreatic tumor growth through T cell suppression and maintenance of key gene programs that define the squamous subtype
.
Cell Rep
2018
;
23
:
1448
60
.
48.
Peranzoni
E
,
Lemoine
J
,
Vimeux
L
,
Feuillet
V
,
Barrin
S
,
Kantari-Mimoun
C
, et al
Macrophages impede CD8 T cells from reaching tumor cells and limit the efficacy of anti–PD-1 treatment
.
Proc Natl Acad Sci U S A
2018
;
115
:
E4041
50
.
49.
Kim
W
,
Chu
TH
,
Nienhüser
H
,
Jiang
Z
,
Del Portillo
A
,
Remotti
HE
, et al
PD-1 signaling promotes tumor-infiltrating myeloid-derived suppressor cells and gastric tumorigenesis in mice
.
Gastroenterology
2021
;
160
:
781
96
.
50.
Nywening
TM
,
Belt
BA
,
Cullinan
DR
,
Panni
RZ
,
Han
BJ
,
Sanford
DE
, et al
Targeting both tumour-associated CXCR2(+) neutrophils and CCR2(+) macrophages disrupts myeloid recruitment and improves chemotherapeutic responses in pancreatic ductal adenocarcinoma
.
Gut
2018
;
67
:
1112
23
.
51.
Dougan
SK
,
Dougan
M
. 
Regulation of innate and adaptive antitumor immunity by IAP antagonists
.
Immunotherapy
2018
;
10
:
787
96
.
52.
Winograd
R
,
Byrne
KT
,
Evans
RA
,
Odorizzi
PM
,
Meyer
AR
,
Bajor
DL
, et al
Induction of T-cell immunity overcomes complete resistance to PD-1 and CTLA-4 blockade and improves survival in pancreatic carcinoma
.
Cancer Immunol Res
2015
;
3
:
399
411
.
53.
Sun
H
,
Liu
SY
,
Zhou
JY
,
Xu
JT
,
Zhang
HK
,
Yan
HH
, et al
Specific TP53 subtype as biomarker for immune checkpoint inhibitors in lung adenocarcinoma
.
EBioMedicine
2020
;
60
:
102990
.
54.
Biton
J
,
Mansuet-Lupo
A
,
Pécuchet
N
,
Alifano
M
,
Ouakrim
H
,
Arrondeau
J
, et al
TP53, STK11, and EGFR mutations predict tumor immune profile and the response to anti-PD-1 in lung adenocarcinoma
.
Clin Cancer Res
2018
;
24
:
5710
23
.
55.
Barrett
MT
,
Lenkiewicz
E
,
Malasi
S
,
Basu
A
,
Yearley
JH
,
Annamalai
L
, et al
The association of genomic lesions and PD-1/PD-L1 expression in resected triple-negative breast cancers
.
Breast Cancer Res
2018
;
20
:
71
.