Tissue preparation

PBMCs were collected prior to surgery. Resected tumor, normal adjacent (Nadj) pancreas, and spleen were collected immediately following surgery and placed into ice cold complete media (DMEM containing 10% FBS + 1% Penicillin/Streptomycin). Tissues were subdivided for flash frozen (RNA/DNA), formalin (immunohistochemistry), optimum cutting temperature compound (O.C.T., immunofluorescence), or were further processed to generate single cell suspensions of mononuclear cells. In the latter case, tumors were weighed, minced with a razor blade, incubated for 30 min in collagenase (1 mg/ml), mechanically filtered (70 μM), and washed 2X in complete media. Peripheral blood was collected in heparin-containing tubes, mixed 1:1 with PBS/EDTA, layered over histopaque, and centrifuged (2,000 rpm for 20 minutes). PBMCs were isolated at the interface of the gradient, washed 1X, counted, and cryopreserved in 10% DMSO for subsequent analysis.

Immunohistochemistry

Tissues were fixed in 10% formalin, paraffin-embedded, and 5 μM-thick sections stained with hematoxylin and eosin (H&E) or prepared for immunohistochemistry by optimized standard antigen-retrieval methods. Each antibody was titrated and compared with isotype-matched controls. Primary antibodies used include: CD8 (Abcam ab4055, 1:200), Foxp3 (Abcam 236A/E7, 1:100), CD20 (eBioscience L26, 1:100), iNOS (Abcam ab15326, neat), CD163 (Novocastra 10D6, 1:200), DC-Lamp (Dendritics 1010E1.01, 1:100), pan-HLA (eBioscience A4, 1:250), HLA-DR (Dako M0746, 1:400), Galectin-3 (eBioscience M3/38, 1:100) and Galectin-9 (Abcam ab69630, 1:250). A 20X objective using a Nanozoomer Digital Pathology slide scanner (Hamamatsu; Bridgewater, New Jersey) was used to scan stained slides. In Fig. 4C and Supplementary Fig. S1F, a project-specific configuration based on a threshold of pixel values (Visiopharm software, Hoersholm, Denmark) detected immunoreactivity. In Figs. 1D and 1E, CD8⁺ and Foxp3⁺ T-cell numbers were counted in three to five areas and mean T-cell number normalized to area (mm²) was plotted. In Figs. 4A and 4B, staining intensity in cell types was based on morphology. For each specimen, three to five regions were graded on a scale (0, no staining detected; 1, light; 2, moderate; 3, high; 4 very high), and averaged for each patient sample. All staining and analyses were reviewed by a pathologist R.H.P.

Open in a separate window

Figure 1

Effector and regulatory T-cell abundance and localization in PDA. A, Immune cell frequency in normal individual blood (Nl blood), PDA patient blood (PDA blood) and resectable tumor specimens (PDA) was assessed by flow cytometric analysis of mononuclear cells. B, CD8⁺, CD4⁺Foxp3^- and CD4⁺Foxp3⁺ Treg cell number in Nadj and resected tumors normalized to gram of tissue. C, Immunohistochemistry (IHC) for CD8 and Foxp3 in representative Nadj pancreas, 2 tumors that differ in the CD8 T-cell infiltrate (PDA-1 and PD1-2) and TLS in PDA. Ac, acinar; TEC, tumor epithelial cell nests; St, stroma; TLS, tertiary lymphoid structures; and black line, demarcating TEC vs. stroma vs. TLS. Scale bar represents 25 μm. D, CD8⁺ T-cell number per mm² in Nadj, tumor epithelial cell nests (TEC), stroma (St) and TLS. Each circle represents the mean number of cells per mm² in the specified location for each resected specimen. E, Foxp3⁺ T-cell number per mm² in Nadj, tumor epithelial cell nests (TEC), stroma (St) and TLS. Each circle is the mean number of cells per mm² in the specified location for each resected specimen. F, Immunofluorescent staining of TLS in PDA for CD31 and PNAd for HEVs. Arrowhead, CD31⁺ vessel in tumor stroma; arrow, CD31⁺PNAd⁺ HEVs in TLS; and dashed white line demarcates TLS vs. stroma. Scale bar represents 25 μm. G, CD8:Treg ratio was determined in TLS and in tumor stroma that were either TLS⁺ or TLS^- by IHC. H, Mean fluorescence intensity (MFI) of Foxp3 in Treg isolated from Nadj pancreas and tumors was determined by flow cytometry. I, Correlation between CD8 T-cell and Treg numbers in PDA was determined by flow cytometry. Green circles represent tumors with TLS.

Open in a separate window

Figure 4

Antigen-presentation molecule and inhibitory ligand expression in human pancreatic tumors. A, Representative IHC of MHC class I (HLA-A), MHC class II (HLA-DR), Galectin-3 and Galectin-9 in Nadj pancreas, PDA and TLS. is, islet; Ac, acinar; d, normal duct. Scale bar represents 25 μm. B, Quantification of IHC staining in acinar cells (Ac), islets (Is) and ducts (D) from Nadj pancreas, PNET tumor epithelial cells (TEC, Pn), PDA TEC and PDA TLS. C, Evaluation of the correlation between CD8 T-cell staining and ligand expression in PDA.

Immunofluorescence

Tissues were embedded in O.C.T. compound (Tissue-Tek), frozen, and stored at -80°C. 7 μM sections were cut and fixed in cold (-20°C) acetone for 15 min. Sections were rehydrated with PBS +2% BSA, blocked with 2% goat serum and incubated for 1 h at room temperature with primary antibody diluted in PBS + 2.5% BSA. Primary antibodies used include peripheral lymph node addressin (PNAd)-FITC (eBioscience MECA79, 1:100), rabbit polyclonal CD8 (Abcam ab4055, 1:25), pan-cytokeratin-FITC (Sigma-Aldrich F3418, 1:200), CD15 (Thermo Scientific MMA/LeuM1, 1:25), αSMA (DAKO 1A4, 1:100) and CD31 (eBioscience WM-59, 1:100). Slides were washed 3X in PBS+2% BSA and incubated with species-specific secondary Alexa-conjugated antibodies (Invitrogen, 1:1000) for 30 min, washed, and mounted with Prolong Gold anti-fade plus DAPI (ThermoFisher Scientific).

Multiplex staining and image analysis

4 μM formalin-fixed paraffin-embedded tissue sections were baked for 1 h at 60°C, dewaxed (Dewax Solution, Leica), and stained on a Leica BOND Rx stainer. Antigen retrieval and antibody stripping steps were performed at 100°C using Epitope Retrieval Solution 2 and Bond Wash Solution (Leica). All other steps were performed at room temperature. Endogenous peroxidase was blocked with 3% H₂O₂ for 8 min followed by protein blocking with TCT buffer (0.05M Tris, 0.15M NaCl, 0.25% Casein, 0.1% Tween 20, pH 7.6) for 30 m. A high stringency wash of high-salt TBST solution (0.05M Tris, 0.3M NaCl, and 0.1% Tween-20, pH 7.2-7.6) was performed after the secondary and tertiary applications. The first primary antibody (position 1) was applied for 60 min followed by the secondary antibody application for 10 min and the application of the tertiary TSA-amplification reagent (PerkinElmer OPAL fluor) for 10 min. The primary and secondary antibodies were stripped with retrieval solution for 20 min before repeating the process with the second primary antibody (position 2) starting with a new application of 3% H₂O₂. The process was repeated for all 6 positions. Antibody position, clone and concentration are provided in Supplementary Table 1. Slides were stained with Spectral DAPI (Perkin Elmer) for 5 min, rinsed for 5 min, and coverslipped with Prolong Gold Antifade. OPAL Polymer HRP Mouse plus Rabbit (PerkinElmer) was used for all secondary applications. Slides were cured for 24 h at room temperature, then representative images from each slide were acquired on PerkinElmer Vectra 3.0 Automated Imaging System. Images were spectrally unmixed using PerkinElmer inForm software and exported as multi-image TIFF files and analyzed using HALO (Indica Labs). Two sections per tumor were sampled at 100%. Regions of interest were defined based on the tumor cell stain and cell morphology.

Flow cytometry

Mononuclear cell preparations were stained with the following fluorophore-conjugated monoclonal antibodies obtained from eBioscience: Lag3 (3DS223H), CD244 (eBioDM244), CD25 (BC96), CCR7 (3D12), CD45RO (UCHL1), Foxp3 (236A/E7), CD137 (4B4-1), CD62L (DREG-56), CD11b (1CRF44), HLA-DR (L243), Eomes (WD1928), and Tim-3 (F38-2E2); from BD Biosciences: T-bet (4B10), CD4 (RPA-T4), and PD-1 (M1H4); and from Biolegend: CD15 (SSEA-1), CD45RA (HI100), and CD160 (B755). For detection of transcription factors Foxp3, Eomes, and T-bet, cells were stained first for cell surface antigens and subsequently fixed and permeabilized using eBioscience fixation buffer. The number of positively stained cells was normalized to gram of tissue for enumeration.

TCRβ CDR3 amplification and sequencing

The DNeasy Blood and Tissue Kit (Qiagen) was used to extract genomic DNA from PBMC (n=7) and flash-frozen tumor specimens (n=9). Samples were treated with Proteinase K and RNAse and genomic DNA purified using the spin column method. High-throughput sequencing of the TCR V-beta CDR3 region (TRB) was performed at Fred Hutchinson Cancer Research Center using the hsTCRB kit (Adaptive Biotechnologies) and an Illumina HiSeq system (Illumina), a method that has been previously validated (31). Sequencing reads were analyzed using the ImmunoSEQ software (Adaptive Biotechnologies). TCR repertoire diversity was assessed using the clonality metric. Clonality score was derived from Shannon’s entropy, which measures the uncertainty associated with a random variable. In the case of CDR3 TCRβ sequences, Shannon’s entropy calculates the uncertainty in predicting a sequence from a large dataset of total CDR3 TCRβ sequences within a sample. To normalize between samples that contain different numbers of total CDR3 TCRβ sequencing reads, entropy was divided by log₂ of the number of unique productive sequences. This method has been previously described (32) and used to obtain clonality, where clonality = 1 – (normalized entropy). Clonality values range from 0 (most diverse, every T-cell in a sample contains a unique TCR) to 1 (monoclonal T-cell population). The sequencing product was organized and nonproductive TCR sequences (premature stop or frame-shift) were excluded from the final analyses. TCR sequence data is publicly available (https://doi.org/10.21417/B7305D).

TLS quantification

R.H.P. evaluated 3 Nadj pancreas samples, 7 PNET, and 20 PDA tumors histologically for TLS using H&E and a large fraction (n=16) were further validated by CD20, DC-Lamp, and PNAd staining. Specimens underwent serial sectioning to cover >60 μm of tumor depth within each sample.

Gene expression analysis

Gene expression data were generated from RNA prepared from flash frozen tumor or Nadj pancreas and directly measured via counts of corresponding mRNA in each sample using nCounter PanCancer Immune Profiling Panel at the Fred Hutchinson Cancer Research Institute. The detection and counting of genes was done via reporter gene-specific nucleic acid probes appended with fluorophore barcodes and biotinylated capture probes that attach to microscopic beads (33). Gene expression was normalized to internal housekeeping genes and was analyzed using the nSolver software and the Advanced Analysis package (Nanostring). The Cancer Genome Atlas (TCGA) of pancreatic adenocarcinoma (https://gdc.cancer.gov/) was used to determine correlation between T-cell and myeloid genes.

Low frequency of T cells productively responding to antigens in tumor environment

The phenotype of T cells was next assessed to address potential tumor-reactivity and also the extent that inhibitory receptors are elevated on tumor-infiltrating T cells. Although most circulating T cells were naive (T_N, CD45R0^-CCR7⁺), most tumoral CD8⁺ T cells were antigen-experienced (CD45RO⁺CCR7^- or CD45RO⁺CCR7⁺), independent of the presence or absence of TLS (Fig. 2B). Regarding markers of TCR signaling and activation (41, 42), expression of both 41BB and PD-1 was significantly elevated on tumor-infiltrating T cells compared to circulating T cells (Fig. 2C). A small subset of tumors contained a high percentage of CD8⁺ T cells that co-expressed 41BB and PD-1 (Fig. 2C). Lag-3 was also significantly elevated on tumor-infiltrating T cells compared to circulating T cells (Fig. 2C), and T cells that expressed PD-1 often also expressed Lag-3 (Fig. 2D and E). The inhibitory receptors Tim-3 and CD244 were not significantly different between tumroral and circulating CD8⁺ T cells (Fig. 2C). CD244 was significantly elevated on circulating CD8⁺ T cells from PDA patients compared to normal individuals (Fig. 2C and Supplementary Fig. S2). TLS presence did not correlate with either 41BB or PD-1 expression by tumoral T cells (Fig. 2F and G).

Open in a separate window

Figure 2

Activation marker and inhibitory receptor expression on tumor-infiltrating T cells. A, Naïve (CD45RO^-CCR7⁺), effector memory (CD45RO⁺CCR7^-) and central memory (CD45RO⁺CCR7⁺) CD8⁺ T-cell frequency in normal blood (Nl), circulation of PDA patients (PDA blood) and in tumors (PDA). Percentages were determined by flow cytometric analysis of CD8⁺ T cells. Each symbol represents an independent sample. B, Representative histogram of CD45RO on CD8⁺ T cells in blood, spleen and TLS^- or TLS⁺ tumors. C, Percentage of CD8⁺ T-cell that express the indicated markers in normal blood (Nl blood), PDA patient blood (PDA blood), PDA patient spleens (PDA spleen) and tumors (PDA) was determined by flow cytometry. Each symbol represents an independent sample. D, Flow plots of CD8⁺ T cells from patient in which the majority CD8⁺ T cells expressed PD1 in tumors. E, CD8⁺PD1⁺ T cells in PDA co-express 41BB, Lag3 and to a lesser extent Tim3. F, Histogram overlay of 41BB and PD1 by CD8⁺ T cells isolated from TLS^- or TLS⁺ tumor. G, Percentage of CD8⁺ T cells that express 41BB or PD1 in TLS^- or TLS⁺ tumors. H, Flow plots demonstrating Eomes and T-bet expression by CD8⁺PD1⁺ T cells from 5 different tumors. I, Percentage of CD8⁺PD1⁺ T cells in PDA that are T-bet^hiEomes^lo or T-bet^loEomes^hi subset. Each dot represents an independent sample.

Functionally distinct subsets PD-1⁺ exhausted T cells can be identified by differential expression of the transcription factors Eomes and T-bet (43, 44). In the majority (5/7) of samples analyzed, CD8⁺PD-1⁺ T cells were predominantly T-bet^hiEomes^lo (Fig. 2H and 2I), the subset that can proliferate in response to PD-1 blockade. Thus, the phenotype of most CD8⁺PD-1⁺ T cells is not consistent with having acquired a terminally exhausted state refractory to PD-1 blockade.

TCRβ CDR3 sequencing shows limited T cell clonal expansion

To address the extent that T cells had undergone clonal expansion, an indicator of antigen recognition, we performed deep sequencing of the complementary determining region 3 (CDR3) of the TCRβ chain in T cells from PDA patients. Because each T-cell expresses a unique rearranged TCRβ variable (V), diversity (D), and joining (J) region that comprise CDR3, deep sequencing of TCRβ CDR3 can molecularly barcode T-cell clonotypes in vivo to determine if T-cell infiltrates are comprised of many unique low frequency clones (polyclonal) and/or contain a few high frequency T-cell clones (oligoclonal) (31). The clonality metric measures T-cell diversity and is derived from the Shannon entropy of the specimen normalized to the log₂ number of unique clones. Clonality ranges from 0 to 1, with 0 representing a purely polyclonal population and 1 representing a purely monoclonal population (31). One caveat of comparing clonality in tumor vs. blood is that areas of inflammation such as tumors attract and are preferentially enriched in CD45RO⁺ effector T cells (Fig. 2A) rather than naïve T cells, which will bias for less T-cell diversity (45) and therefore the appearance of enhanced clonality. However, we found that T-cell clonality was not significantly different between T cells in the highly polyclonal circulating population vs. the tumor (Fig. 3A, mean clonality PDA: 0.112 ± 0.03; mean clonality blood: 0.0721 ± 0.03) despite an increased number of total CDR3 templates in circulation (54,584 ± 25,458) compared to tumor (20,596 ± 8,628) and an 8-fold increase in unique CDR3 templates (clonotypes) in circulation compared to tumor (42,674 ± 21,059 vs. 5,687 ± 1125, Figs. 3B, 3C and Supplementary Table 3). T-cell clonality in PDA was 30% lower than T-cell clonality in colorectal cancer as determined by comparison with the public dataset (46).

Open in a separate window

Figure 3

Deep-sequencing of TCRβ CDR3 and expression of T-cell activation and inhibitory genes in PDA. A, Clonality of T cells in blood (B) vs. PDA. B, Number of productive TCRβ CDR3 templates in blood (B) vs PDA. C, Number of productive unique TCRβ CDR3 templates in blood (B) vs. PDA. D, Frequency of the top 15 T-cell clones (○) and clonality scores ( ) from 9 different tumors. Presence of TLS is shown. Dashed line represents a clonality score of 0.10; blue arrows, tumors with a clonality of >0.10. E-F, Gene expression in Nadj pancreas, tumors with clonality >0.10 (n=5) or <0.10 (n=4) was determined by Nanostring.

Since assessing the whole population might obscure limited expansion of a few T-cell clonotypes, we quantified the frequency of the top 15 T-cell clones across 9 different PDAs. Only 1 out of 9 tumors had a T-cell clone that represented > 5% of the T-cell population (Fig. 3D and Supplementary Table 3). The presence of TLS also did not correlate with increased T-cell clonality (Fig. 3D). To investigate if clonality correlated to T-cell activation gene signature, we segregated tumors based on clonality score of a clonality of 0.10, Fig. 3D, Supplementary Table 3). Tumors with >0.10 clonality showed a trend for increased CD4, CD8A, and CD8B T-cell transcripts and significant enrichment of FOXP3 transcripts (Fig. 3E). HLA-A and PD-L1 transcripts were not significantly different compared to Nadj (Fig. 3E). Tumors with clonality >0.10 contained significantly more transcripts for a variety of TCR signaling and effector genes that are expressed on recently activated but not resting T cells (Fig. 3F). Genes involved in T-cell inhibition, including LAG3, TIGIT, HAVCR3, and CTLA-4, were also significantly increased in tumors with a clonality score of >0.10 (Fig. 3G). These data are consistent with flow cytometric data that showed a subset of tumors contained some T cells that expressed markers indicative of recent antigen encounter (Fig. 2) and suggest that clonality score of >0.10 may provide a metric for identification of tumors that may contain a suppressed yet potentially tumor-reactive T-cell response.

Expression of antigen presentation molecules and inhibitory ligands

We next sought to determine the expression of antigen presentation molecules and inhibitory ligands that bind receptors elevated on tumor-infiltrating T cells in tumors. PDA tumor cells expressed moderate to high amounts of HLA that were significantly higher than normal ducts in Nadj pancreas and also significantly higher than PNET tumor cells (Fig. 4A and 4B). HLA-DR, which binds Lag-3 (47), was low (Figs. 4A and 4B). Galectin-3, another putative ligand for Lag-3 (48), and Galectin-9, which binds Tim-3 (49), were also highly expressed by both Nadj ducts and tumor cells in PDA, but not PNET (Fig. 4A and B). HLA staining intensity in tumor cells showed the greatest inter- and intratumoral heterogeneity (Fig. 4B) and did not correlate with CD8 T-cell abundance (Fig. 4C). Although we observed a trend for Galectin-3 and Galectin-9 expression to negatively correlate with CD8 T cells, these comparisons were not significant (Fig. 4C).

Inflammatory monocytes and granulocytes correlate with tumoral CD8⁺ T cells

Various suppressive immature myeloid subsets are elevated in genetically engineered mouse models of PDA (9, 11, 12, 50). Inflammatory monocytes have been reported to be elevated in human PDA and correlate with a poor prognosis (50). We found monocytes (CD45⁺CD11b⁺CD14⁺HLA-DR^low) to be significantly elevated in circulation but not the tumor of PDA patients (Fig. 5A and B). CD11b⁺CD15⁺CD33⁺ granulocyte frequency was quite variable in patients (Figs. 5C-5E). Tumors with a high CD8 T-cell infiltrate also contained elevated numbers of granulocytes and inflammatory monocytes (Fig. 5F); conversely, tumors with a low number of granulocytes had few CD8⁺ T cells (Fig. 5G). The neutrophil chemokines CXCL1 and IL8, and the chemokine receptor CXCR2, were enriched 3.3-fold, 2.7-fold and 7.5-fold, respectively, in tumors that had a clonality score of >0.10 compared to tumors with a clonality score of <0.10 (Fig. 5H). Gene expression analyses from the TCGA dataset of CD33 and CXCR2 myeloid markers also showed a positive correlation with CD8⁺ T cells (Fig. 5I). The correlation between CD8⁺ T cells and a neutrophil signature in human PDA was unexpected, because studies in mouse models suggest that granulocytes may play a role in excluding T-cell infiltration from PDA (8, 9, 11, 13), although the inability to map specific myeloid cell subsets from one species to the other may also explain these differences.

Open in a separate window

Figure 5

Spectrum of inflammatory monocyte and granulocyte accumulation in PDA and correlation with CD8⁺ T cells. A, CD14⁺HLA-DR^low monocyte frequency in circulation (Bl) of normal (Nl) individuals and in circulation (Bl), Nadj pancreas and tumors (Tu) from PDA patients. B, Number of CD14⁺HLA-DR^low in Nadj pancreas and PDA normalized to gram of tissue. C, Representative FACs plots of CD33⁺CD15⁺ granulocytes (gated on CD45⁺ cells) from a patient. D, CD11b⁺CD33⁺CD15⁺ granulocyte frequency in circulation (Bl) of normal (Nl) individuals and in circulation (Bl), Nadj pancreas and tumors (Tu) from patient samples. E, CD11b⁺CD33⁺CD15⁺ granulocyte number in Nadj pancreas and PDA normalized to gram of tissue. F, CD8⁺ T-cell number and myeloid subsets were determined by flow cytometric analysis. G, Multicolor immunofluorescence of CD8⁺ T cells, CD15⁺ granulocytes and cytokeratin (CK⁺) tumor cells in two PDAs that differ in the T-cell stromal infiltrate. Representative PNET sample is shown. Scale bar represents 25 μm. H, Expression of granulocyte genes in tumors that had a clonality score of >0.10 or <0.10. I, Analysis of TCGA dataset show a correlation between CD8 T-cell transcripts and genes involved in granulocyte trafficking in PDA.

We thank Ashley Dotson, Markus Carlson, Kathleen DelGiorno, Libing Feng, Yen Ho and Joseph Ryan for technical assistance processing human tissue; Megan Larmore and Brian Johnson for histology and immunohistochemistry services at the University of Washington, Seattle, WA;Kimberly S. Smythe for technical assistance with multiplex development and HALO image analysis software; and Deborah Banker and Shelley Thorsen for manuscript review. This work was supported by an Irvington Institute Fellowship Program of the Cancer Research Institute (to I.M.S), Swim Across America (to I.M.S.), the Jack and Sylvia Paul Estate Fund to Support Collaborative Immunotherapy Research (to I.M.S.), the Giles W. and Elise G. Mead Foundation (to S.R.H.), NIH National Cancer Institute (CA018029 and CA033084 (to P.D.G.), CA161112 (to S.R.H.) and a P30CA015704 Supplement (S.R.H. and P.D.G.)), Pancreatic Cancer Action Network (16-65-GREE to P.D.G.) and a research agreement with Juno Therapeutics (to P.D.G.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.