Purpose:

Neoadjuvant therapy (neoTx) has dramatically improved the prognosis of patients with locally advanced and borderline resectable pancreatic ductal adenocarcinoma, yet its mechanisms of action on tumor cells and the tumor microenvironment are still unknown. Here, we aimed to characterize the multiple facets of neoTx-induced alterations in the pancreatic cancer microenvironment.

Experimental Design:

We performed the currently most comprehensive histopathologic analysis of desmoplasia, angiogenesis, neural invasion, and immune cell infiltration at the tumor–host interface of pancreatic cancer after neoTx (n = 37) versus after primary resection (n = 37) through quantitative IHC and double immunofluorescence using automated and software-based quantification algorithms.

Results:

We demonstrate that, independently of the applied pretreatment, neoadjuvant regimes are able to reverse the immunosuppressive behavior of malignant cells on pancreatic cancer microenvironment. Here, neoTx-driven selective depletion of regulatory T cells and myeloid-derived suppressor cells was associated with enrichment of antitumor immune cells in the peritumoral niche, decreased stromal activation, and less neural invasion. Importantly, the degree of this antitumor immune remodeling correlates to the degree of histopathologic response to neoTx. Survival analysis revealed that the tumor proliferation rate together with the activation of the stroma and the intratumoral infiltration with CD4+ T cells and natural killer cells constitute as independent prognostic factors for neoadjuvantly treated pancreatic cancer.

Conclusions:

NeoTx is not only cytotoxic but has pleiotropic, beneficial effects on all cellular and noncellular components of pancreatic cancer. Combinational approaches including immunotherapy may unleash long-term and more effective antitumor responses and improve prognosis of pancreatic cancer.

This article is featured in Highlights of This Issue, p. 1

Translational Relevance

Neoadjuvant therapy (neoTx) has dramatically improved the prognosis of patients with locally advanced and borderline resectable pancreatic ductal adenocarcinoma. Despite its broad use in clinical practice, its underlying mechanisms of action are not yet clarified. In this study, we carried out the most comprehensive histopathologic analysis of the pancreatic cancer microenvironment to date and uncovered a compelling immunostimulatory effect of neoTx on pancreatic cancer. Here, the selective depletion of regulatory T cells and myeloid-derived suppressor cells mediated by neoTx was associated with enrichment of antitumor immune cells in the peritumoral niche, decreased stromal activation, and less neural invasion. Hence, the beneficial effect of neoTx may not rely, as traditionally believed, on the direct cytotoxicity of the chemotherapeutical agents but on the restoration of the immune cell–mediated antitumor response. Future combinational therapies with immunotherapy on a neoadjuvant setting could therefore substantially improve the prognosis of pancreatic cancer.

Over the past decade, neoadjuvant therapy (neoTx) has dramatically improved prognosis of patients with locally advanced and borderline resectable pancreatic cancer (1, 2); thus, understanding its mechanisms of action is of high and immediate clinical relevance. Although traditionally considered to be immunosuppressive, growing evidence suggests that neoTx exerts its beneficial effects by restoration of the local antitumoral immune response (3–5). Furthermore, neoadjuvant approaches have been shown to enhance systemic immunity and eradicate residual metastasis more effectively than adjuvantly applied regimes on preclinical settings (6, 7). To design novel combinational therapies, a comprehensive understanding of the immunostimulatory properties of neoTx on pancreatic cancer tumor microenvironment is needed (6).

Throughout carcinogenesis, pancreatic cancer cells promote peritumoral infiltration of immunosuppressive cells such as regulatory T cells (Treg), myeloid-derived suppressor cells (MDSC), and alternatively activated M2 macrophages. They all suppress the activity of effector cells and support tumor progression (8, 9). This characteristic immunosuppressive milieu in pancreatic cancer is partly a result of the dense desmoplastic reaction and the sequestration of CD8+ T cells by activated pancreatic stellate cells (PSC; ref. 10). Endothelial cells also play a central role on intratumoral immune cell trafficking mediating selective apoptosis of CD8+ T cells but not of Tregs during the process of extravasation (11, 12). Whether the stromal components are affected by neoTx and how this influences the local immune response are still unexplored.

For this reason, we performed the largest systematic, quantitative analysis of desmoplasia, angiogenesis, neural invasion, tumor proliferation, and immune infiltration at the tumor–host interface after neoTx in comparison with patients with primary resected pancreatic cancer.

Patients and histologic specimens

Pancreatic tissue samples for IHC and immunofluorescence were collected from patients with conventional pancreatic ductal adenocarcinoma, who underwent surgical resection with curative intention between 2008 and 2015 in our institution. In this comprehensive study, 37 patients with different neoadjuvant regimes were matched according to age, sex, and tumor stage with 37 up-front resected cases. The indication for neoTx was based on the presence of radiographically borderline resectable or locally advanced disease. The clinical characteristics for both groups and the applied neoadjuvant treatments are summarized in Supplementary Table S1. All patients were informed, and written consent was obtained for tissue collection. This study was approved by the Ethics Committee of the Technical University of Munich (Munich, Germany; Nr. 549/16s).

IHC and double immunofluorescence labeling

We used IHC and immunofluorescence double staining to systematically analyze the histopathologic features of neoadjuvantly treated and primary resected pancreatic cancer samples. Two different regions of each resection specimen were analyzed for each histologic feature. Consecutive 3-μm paraffin-embedded pancreatic cancer sections were analyzed for stromal activation using alpha smooth muscle actin (αSMA), as well as for angiogenesis and neurogenesis with the cluster of differentiation/CD31 as a specific marker for endothelial cells and the glial marker S100, respectively. The proliferation rate of tumor cells was analyzed using the nuclear marker Ki67. Furthermore, for the analysis of the immunologic response after neoTx, we used inflammatory cell surface markers including CD45 as pan-leucocyte marker, CD8 for the CTLs, CD20 as a marker for B cells, CD4 to label Th lymphocytes, FOXP3 as a marker for Tregs, CD68 as a pan-macrophage marker, and HLA-DR and CD206 to distinguish between M1 and M2 macrophage polarization, respectively. We also characterized MDSCs with CD11b and CD33, and dendritic cells (DC) using the most commonly applied specific markers CD8 and CD103. Natural killer (NK) cells were immunolabeled against CD56, and residual intrapancreatic neuroendocrine cells were excluded on a morphologic basis (Supplementary Table S2). The sections were incubated with primary antibodies at the indicated dilutions overnight in a humid chamber at 4°C (Supplementary Table S2). Respective isotype control antibodies were used as negative control.

Quantitative analysis of the peritumoral immune cell distribution and response to neoTx

In a preliminary analysis of the samples, an independent investigator from the department of pathology (A. Muckenhuber) selected those areas with tumor cells on hematoxylin and eosin–stained pancreatic cancer sections. We then transferred this label on to the rest of immunofluorescence slides. Digitalized images of eight juxtatumoral hotspots were then taken with a digital microscope (Keyence BioRevo BZ-9000) at high resolution and 20× magnification. Cellprofiler Software was used to determine the immune cell infiltration. For this purpose, we elaborated a pipeline that enabled the automated identification and quantification of double immunolabeled cells by their color, shape, size, and staining intensity (Supplementary Fig. S1). To measure the immune cell infiltration, we quantified not only the density [number of cells per field of view (FOV)] of the cells but also the ratio (proportion of the analyzed cell subtype to the total leucocyte population), as preliminary tests revealed that the latter may be more representative of the biological characteristics of the respective infiltrating cells.

The response to neoTx was measured in analogy with the present scoring systems that are all based on the proportion of remaining tumor cells on histologic sections (13, 14). To have a more precise and quantitative measurement, we divided the proportion of cytokeratin-19–immunopositive cancer cells to the total tissue area after scanning of the entire tumor bed that was previously marked by the pathologists (A. Muckenhuber and B. Konukiewitz). We termed this proportion the “tumor index” as an indicator of the remaining tumor cells after neoTx.

Quantitative analysis of desmoplastic reaction, angiogenesis, neurogenesis, neural invasion, and tumor proliferation rate

Slides were digitally and entirely scanned with a digital brightfield microscope (Keyence BioRevo BZ-9000) at high resolution and 10× magnification. The software-based histopathologic analysis of the desmoplastic reaction, angiogenesis, and tumor proliferation rate was carried out with ImageJ (v1.51e, NIH, Bethesda, MD). Briefly, to analyze the stromal activation, we used the threshold function to select the area occupied by the immunostained pixels, after exclusion of color artifacts and major arteries. Once the optimal detection threshold was set, these values were kept constant throughout the analysis. We then created a mask of the pixels stained with αSMA and anillin representing the quiescent and activated component of the stroma and calculated the proportion of αSMA and anillin-stained areas to the total area of the sample (Supplementary Fig. S2A–S2D). We also determined the already established activated stromal index (ASI; ref. 15; Supplementary Table S3). For the analysis of angiogenesis, we calculated the proportion of endothelial cells stained with CD31 (using again an automated detection upon a preset threshold, and the luminal area of the vessels was selected manually on each sample) to the total area of the tissue. We performed this analysis in the peritumoral niche defined as the circular area with a 150-μm radius around the tumor cells, as well as on the whole tissue area (Supplementary Fig. S2A and S2B; Supplementary Table S3). To determine the proliferation rate of cancer cells and to have a tumor area independent and thus more accurate estimate of the cancer cell proliferation rate, we selected the area of the tumor manually and automatically quantified the proportion of Ki67+ proliferating nuclei to both the absolute number of DAPI+ tumor cell nuclei and the tumor area using the threshold tool of ImageJ (Supplementary Fig. S3B–S3D; Supplementary Table S3). Finally, the neural size, neural density, and neural invasion in our pancreatic cancer human specimens were calculated using the Keyence Analyzer Software, which enables the detection of nerves stained with S100. Furthermore, we designed an algorithm for the automated identification and quantification of perineural invasion (PNI), defined as the presence of cancer cells along the perineural sheet without destroying its bundle integrity, and endoneural invasion (ENI), where cancer cells invade the endoneural sheet (Supplementary Fig. S3A; Supplementary Table S3).

Statistical analysis

Statistical analysis was performed using the IBM SPSS Statistics v25 and the GraphPad Prism 5 software was used to create all graphics. The unpaired t test was applied for two-group analysis, followed by multiple test correction (bootstrapping method). Results are expressed as mean ± SD. Overall patient survival was calculated from the date of therapy start to the date of last follow-up (censored) or date of patient's death (event). Median values were taken as cut-off limits in the comparison of two groups. Univariate survival analyses were calculated using the Kaplan–Meier method for estimation of event rates and the Breslow–Wilcoxon test for survival comparisons between patient groups. The multivariate analysis was performed applying a Cox regression model. The estimations of HRs were presented with 95% confidence intervals. Comparison of demographic data was made using the χ2 test. To examine the correlation between immune cell subtypes and other members of the stroma, we used the Pearson coefficient. Two-sided P values were always computed, and an effect was considered statistically significant at a P value of ≤0.05.

NeoTx induces an immunologic shift toward a proinflammatory and antitumor immune milieu in pancreatic cancer

Pancreatic cancer is frequently associated with pronounced inflammatory infiltrates in the tumor tissue (16). To uncover the differences in the immune cell infiltration patterns between neoadjuvantly treated and primary resected tumors, we performed a histopathologic in-depth analysis of the immune response in the juxtatumoral niche.

First, we analyzed the intratumoral infiltration of the utmost studied cell populations in pancreatic cancer, that is, CD8+ cytotoxic T cells, CD4+ T cells, and FOXP3+ Tregs. The density of infiltrating antitumorigenic CD8+ T cells was similar in neoadjuvantly treated and primary resected specimens (56.20 ± 41.59 vs. 64.43 ± 54.48 cells/FOV, P = 0.425); however, the total number of CD45+ leucocytes was notably decreased in neoTx patients (267.7 ± 140.1 vs. 188.7 ± 142.7 cells/FOV, P = 0.027). As such, the proportion of CD8+ T cells to the entire leucocyte population almost doubled after neoTx (0.2265 ± 0.1139 vs. 0.3713 ± 0.1019, P = 0.001; Fig. 1A and B). Although the density of CD4+ T cells remained constant (26.38 ± 13.95 vs. 23.26 ± 14.20 cells/FOV, P = 0.443), a subpopulation of highly immunosuppressive FOXP3+CD4+ Tregs tended to be reduced after neoTx in both its density (15.04 ± 8.857 vs. 10.21 ± 10.41 cells/FOV, P = 0.057) and in its ratio to the whole CD4+ T-cell population (0.5792 ± 0.2272 vs. 0.4200 ± 0.2011, P = 0.003; Fig. 2A and B). Furthermore, the intratumoral infiltration of CD8+ T cells was inversely correlated with the infiltrating FOXP3+ Tregs (r = −0.3502, P = 0.0022, Fig. 2C) and positively correlated with DCs (P = 0.0041) and M1 macrophages (P = 0.0018, Fig. 1C).

Figure 1.

Histopathologic impact of neoTx on antitumorigenic immune cell subsets in human pancreatic cancer. A, Representative immunofluorescence images (40×) of the intratumoral infiltration by cytotoxic T cells, DCs, M1 macrophages, and NK cells (arrows) on PR versus neoTx patients with pancreatic cancer. CD45+ leucocytes (white markers) were analyzed together with CD8+ T cells. B, The pairwise comparison of antitumorigenic immune cell infiltration, including CD8+ cytotoxic T cells, CD8+CD103+ DCs, HLA-DR+CD68+ M1 macrophages, and CD56+ NK cells between PR and neoadjuvantly treated patients, showed an immunologic reactivation of the tumor microenvironment and an unexpected decrease of NK cells after neoTx. Each data point represents a single patient (median score of all hotspots (n = 8) and lines represent the SD). C, Positive correlation analysis between the cytotoxic T-cell ratio and the density of DCs and M1 macrophages, whereas it showed to be negatively correlated to NK cells. FOV, field of view; NS, not significant; PR, primary resected.

Figure 1.

Histopathologic impact of neoTx on antitumorigenic immune cell subsets in human pancreatic cancer. A, Representative immunofluorescence images (40×) of the intratumoral infiltration by cytotoxic T cells, DCs, M1 macrophages, and NK cells (arrows) on PR versus neoTx patients with pancreatic cancer. CD45+ leucocytes (white markers) were analyzed together with CD8+ T cells. B, The pairwise comparison of antitumorigenic immune cell infiltration, including CD8+ cytotoxic T cells, CD8+CD103+ DCs, HLA-DR+CD68+ M1 macrophages, and CD56+ NK cells between PR and neoadjuvantly treated patients, showed an immunologic reactivation of the tumor microenvironment and an unexpected decrease of NK cells after neoTx. Each data point represents a single patient (median score of all hotspots (n = 8) and lines represent the SD). C, Positive correlation analysis between the cytotoxic T-cell ratio and the density of DCs and M1 macrophages, whereas it showed to be negatively correlated to NK cells. FOV, field of view; NS, not significant; PR, primary resected.

Close modal
Figure 2.

Histopathologic impact of neoTx on protumorigenic immune cell subsets in human pancreatic cancer. A, Representative photomicrographs (40×) of immunolabeled Tregs, B cells, M2 macrophages, and MDSCs of PR versus neoTx patients with pancreatic cancer. White arrows show double-immunolabeled cells; white arrowheads indicate CD45+ cells. B, Scatterplots comparing the protumorigenic immune cell infiltration, including FOXP3+CD4+ Tregs, CD20+ CD45+ B lymphocytes, CD206+CD68+ M2 macrophages, and CD33+CD11+ MDSCs between PR and neoadjuvantly treated patients. These demonstrated a significant decrease of Treg and MDSCs densities after neoTx, and these cell subsets arose as the most potent immunosuppressive cells in the pancreatic cancer microenvironment. Each data point represents a single patient [median score of all hotspots (n = 8) and lines represent the SD]. C, Inverse correlation analysis between the cytotoxic T-cell ratio and the density of MDSCs and Tregs ratio. D, Relative distribution of tumor infiltrating immune cells in PR versus neoTx patients. FOV, field of view; NS, not significant; PR, primary resected.

Figure 2.

Histopathologic impact of neoTx on protumorigenic immune cell subsets in human pancreatic cancer. A, Representative photomicrographs (40×) of immunolabeled Tregs, B cells, M2 macrophages, and MDSCs of PR versus neoTx patients with pancreatic cancer. White arrows show double-immunolabeled cells; white arrowheads indicate CD45+ cells. B, Scatterplots comparing the protumorigenic immune cell infiltration, including FOXP3+CD4+ Tregs, CD20+ CD45+ B lymphocytes, CD206+CD68+ M2 macrophages, and CD33+CD11+ MDSCs between PR and neoadjuvantly treated patients. These demonstrated a significant decrease of Treg and MDSCs densities after neoTx, and these cell subsets arose as the most potent immunosuppressive cells in the pancreatic cancer microenvironment. Each data point represents a single patient [median score of all hotspots (n = 8) and lines represent the SD]. C, Inverse correlation analysis between the cytotoxic T-cell ratio and the density of MDSCs and Tregs ratio. D, Relative distribution of tumor infiltrating immune cells in PR versus neoTx patients. FOV, field of view; NS, not significant; PR, primary resected.

Close modal

We then analyzed a subset of innate lymphoid cells, traditionally considered to be able to recognize and destroy tumor cells, that is, CD56+ NK cells. Unexpectedly, neoTx induced a notable depletion of these cells (14.67 ± 12.24 vs. 2.596 ± 2.654 cells/FOV, P = 0.002), and infiltrating CD8+ T cells showed a negative correlation to the amount of NK cells (r = −0.4756, P = 0.0001; Fig. 1B). Another lymphocytic cell subpopulation that has been reported to play a protumorigenic role in pancreatic cancer is the CD20+ B cells. Although not significant, we observed a decreased density of these cells (66.89 ± 63.17 vs. 44.23 ± 58.57 cells/FOV, P = 0.079) as well as a reduction in their ratio to the entire leucocytic population after neoTx (0.1754 ± 0.1215 vs. 0.1196 ± 0.1234, P = 0.067; Fig. 2B).

We then extended our histopathologic analysis to myeloid cells, including DCs, macrophages, and MDSCs. Our results showed that DCs, considered the most potent antigen-presenting cells in tumor immunity, presented a great expansion after neoTx, nearly doubling their density in the neoadjuvantly treated patients (10.07 ± 5.879 vs. 21.72 ± 18.01 cells/FOV, P = 0.002; Fig. 1B). In contrast, the intratumoral infiltration of MDSCs, a population of immature myeloid cells with the ability to suppress T-cell activation (17), was significantly reduced after neoTx (17.49 ± 15.05 vs. 5.692 ± 5.460 cells/FOV, P = 0.002; Fig. 2B). As the regression of “alternatively activated” tumor-infiltrating M2 macrophages back to M1 antitumorigenic phenotype was reported in the literature (18), we analyzed the possible role of neoTx in M2-specific inactivation. Neither the density of M1- (13.61 ± 9.968 vs. 17.26 ± 10.71 cells/FOV, P = 0.155) and M2-polarized macrophages (12.77 ± 12.42 vs. 8.398 ± 6.659 cells/FOV, P = 0.065), nor the proportion of the protumorigenic phenotype to the whole macrophage population (0.5215 ± 0.1783 vs. 0.4530 ± 0.1637, P = 0.065) was altered after neoTx (Figs. 1B and 2B).

This immunologic shift was mostly independent of the applied neoadjuvant regimen, and only the tumor infiltration with Tregs seemed to be more susceptible to FOLFIRINOX compared with gemcitabine-based and neoadjuvant radiotherapy treatment (Treg density: 15.04 ± 8.857 vs. 11.56 ± 13.09, P = 0.0424; Treg ratio: 0.091 ± 0.072 vs. 0.3629 ± 0.2173, P = 0.0035; Supplementary Fig. S4A).

In addition, we demonstrated that the density of myeloid cells subclasses was also correlated with the CD8+ T-cell infiltrate. We detected a positive correlation of CD8+ T cells with DCs (r = 0.3301, P = 0.0041) and M1 macrophages (r = 0.3570, P = 0.0018), whereas it was inversely correlated to MDSCs (r = −0.2663, P = 0.0218; Figs. 1C and 2C).

Histologic response to neoTx for pancreatic cancer is associated with higher numbers of CD8+ T cells and lower number of MDSCs in the tumor tissue

To correlate the observed immunogenic remodeling in the pancreatic cancer tissues after neoTx to the degree of histologic response to neoTx, we introduced the “tumor index,” which is a quantitative parameter that expresses the proportion of present tumor cells after neoTx to the total tumor tissue area on each section. As expected, we observed a markedly reduced tumor index in the tumor tissues after neoTx when compared with tissues obtained upon primary resection, independently of the neoadjuvant regimen used (Fig. 3A). Interestingly, though, the tumor index and thus the histologic response to neoTx did not translate into better overall survival when the patients were stratified into low versus high tumor index groups based on the median tumor index of the cohort (Fig. 3B). Furthermore, the main histopathologic features showed no correlation to degree of histologic response (Fig. 3C; Supplementary Fig. S5A and S5B). This observation suggests that the extent of response to neoTx did not markedly influence the overall survival after resection; it was the presence of response to neoTx per se that was associated with improved survival.

Figure 3.

Response to neoTx for pancreatic cancer is associated with higher numbers of CD8+ T cells and lower number of MDSCs in the tumor tissue. A, We introduced the “tumor index” as a quantitative parameter that expresses the proportion of tumor cells after neoTx to the total tumor tissue area on each section. B, Overall survival based on the CD8/MDSC proportion and on the tumor index and thus the histologic response to neoTx with survival when the patients were stratified into low versus high tumor index groups according to the median value of the cohort. C, The degree of histologic response to neoTx was analyzed in comparison with the proportion of infiltrating CD8+ T cells and MDSC. D, The proportion of CD8+ T cells to MDSC (CD8/MDSC) in neoTx versus PR patients. PR, primary resected.

Figure 3.

Response to neoTx for pancreatic cancer is associated with higher numbers of CD8+ T cells and lower number of MDSCs in the tumor tissue. A, We introduced the “tumor index” as a quantitative parameter that expresses the proportion of tumor cells after neoTx to the total tumor tissue area on each section. B, Overall survival based on the CD8/MDSC proportion and on the tumor index and thus the histologic response to neoTx with survival when the patients were stratified into low versus high tumor index groups according to the median value of the cohort. C, The degree of histologic response to neoTx was analyzed in comparison with the proportion of infiltrating CD8+ T cells and MDSC. D, The proportion of CD8+ T cells to MDSC (CD8/MDSC) in neoTx versus PR patients. PR, primary resected.

Close modal

We then correlated the degree of histologic response to therapy to the proportion of infiltrating CD8+ T cells and MDSCs, as the interrelation between these two cell populations is assumed to affect the rate of tumor spread (19). Indeed, we detected a greater density of CD8+ T cells in the tumor tissues with lower tumor index, thus lower amounts of remaining cancer cells (Fig. 3C), which reflects the antitumor role of CD8+ T cells in the pancreatic cancer microenvironment. Accordingly, the proportion of CD8+ T cells to MDSCs (CD8/MDSC) was remarkably higher in patients treated with neoTx than in primary resected patients (Fig. 3D). However, a higher CD8/MDSC proportion did not result in a better overall survival (Fig. 3B).

The pancreatic cancer microenvironment after neoTx is characterized by reduced stromal activation and diminished neural invasion

One of the aims of our study was to determine whether neoTx is able to alter the characteristic strong activation of the stroma and collagen deposition in pancreatic cancer. Although both the αSMA content (0.07514 ± 0.03273 vs. 0.02094 ± 0.01692; P = 0.734) and the amount of “quiescent” stroma marked by collagen content remained constant (0.1593 ± 0.1741 vs. 0.2763 ± 0.3248; P = 0.394) after neoTx, the ASI (15) turned out to be significantly reduced in neoadjuvantly treated patients (0.8581 ± 0.7822 vs. 0.1151 ± 0.1136; P = 0.003; Fig. 4A and B). The pronounced stromal activation surrounding pancreatic cancer cells has been proposed to impede lymphocyte infiltration (12). This hypothesis was supported by our correlation analyses, as the activation of PSCs was inversely correlated with the amount of CD8+ T cells (r = −0.3279, P = 0.0043) in the tumor vicinity (Supplementary Fig. S6A). Both the endothelial area (0.002855 ± 0.001931 vs. 0.002326 ± 0.001461 μm2; P = 0.162) and microvessel lumen area remained constant after neoTx (0.006258 ± 0.005861 vs. 0.009823 ± 0.01289 μm2; P = 0.265; Fig. 4C–E). However, we could demonstrate a strong correlation between the proportion of cytotoxic T cells and Tregs with the vascularity of the tumor (r = 0.2471, P = 0.0167 and r = −0.3009, P = 0.0092, respectively), suggesting a possible role of angiogenesis on immune cell trafficking, which is independent of pancreatic cancer (Supplementary Fig. S7A and S7C).

Figure 4.

Histopathologic impact of neoTx on the desmoplastic reaction and angiogenesis of pancreatic cancer. A, Representative photomicrographs of the IHC staining of activated/quiescent stroma with αSMA and anillin, respectively. Note the increased amount of activated PSCs stained with αSMA on the PR patients (top row). Scale bars indicate 2,000 μm, 100 μm, and 50 μm from the left to the right, respectively. B, Pairwise comparisons of collagen area, αSMA area, and ASI on both neoadjuvantly treated and PR patients revealed a significant stromal suppression after neoTx. C, Representative images of CD31-stained areas reflecting the vascularization of the stroma (left, scale bars indicate 50 μm) and the respective analysis on the peritumoral area (right, scale bars indicate 100 μm) on patients treated with neoTx versus upfront resected cases. The discontinuous line determines an area of 150 μm around tumor cells defining the peritumoral niche. D, The correlation analysis between angiogenesis and neural invasion showed no relation between the two histopathologic features. Pearson correlation coefficient (r) and statistical significance are presented for both correlations. The unit for the y-axes is μm2. E, Scatterplots showing no difference in endothelial area and microvessel density in the total and peritumoral areas, respectively, after neoTx. The unit for the y-axes is μm2. NI, neural invasion; NS, not significant; PR, primary resected.

Figure 4.

Histopathologic impact of neoTx on the desmoplastic reaction and angiogenesis of pancreatic cancer. A, Representative photomicrographs of the IHC staining of activated/quiescent stroma with αSMA and anillin, respectively. Note the increased amount of activated PSCs stained with αSMA on the PR patients (top row). Scale bars indicate 2,000 μm, 100 μm, and 50 μm from the left to the right, respectively. B, Pairwise comparisons of collagen area, αSMA area, and ASI on both neoadjuvantly treated and PR patients revealed a significant stromal suppression after neoTx. C, Representative images of CD31-stained areas reflecting the vascularization of the stroma (left, scale bars indicate 50 μm) and the respective analysis on the peritumoral area (right, scale bars indicate 100 μm) on patients treated with neoTx versus upfront resected cases. The discontinuous line determines an area of 150 μm around tumor cells defining the peritumoral niche. D, The correlation analysis between angiogenesis and neural invasion showed no relation between the two histopathologic features. Pearson correlation coefficient (r) and statistical significance are presented for both correlations. The unit for the y-axes is μm2. E, Scatterplots showing no difference in endothelial area and microvessel density in the total and peritumoral areas, respectively, after neoTx. The unit for the y-axes is μm2. NI, neural invasion; NS, not significant; PR, primary resected.

Close modal

Pancreatic cancer is characterized by prominent intrapancreatic neuropathic changes including increased neural density and hypertrophy, neural invasion, and neuritis (20). Neoadjuvantly treated patients presented a 3-fold decrease in neural invasion severity (30.51 ± 21.75 vs. 9.422 ± 12.70; P = 0.001) whereas neural density (3.3 ± 2.3 vs. 3.9 ± 2.5 nerves per mm2; P = 0.230) and neural size (15,400 ± 36,610 μm2 vs. 9,772 ± 10,550 μm2; P = 0.831) remained constant when compared with up-front resected patients (Fig. 5A and B; Supplementary Fig. S8A). Both ENI and PNI underwent a significant decrease after neoTx (12.39 ± 11.48 vs. 4.064 ± 5.011, P < 0.003; 18.37 ± 15.25 vs. 5.214 ± 9.584, P = 0.001, respectively, Fig. 5B). In addition, we demonstrated a strong correlation between the size of the nerves and NI severity and the activation of the stroma (ASI; r = 0.3636, P = 0.014, respectively; Fig. 5C; Supplementary Fig. S8B). However, the amount of activated PSCs did not correlate with the density of nerves (P = 0.1009; Supplementary Fig. S8B). In line with the novel theories that account the beneficial effect of neoTx in pancreatic cancer to a redistribution of the stromal components rather than to a direct cytotoxic effect, our results showed no changes on the tumor proliferation rate when proportioned to tumor area (0.0002033 ± 0.0001594 vs. 0.0002000 ± 0.0002874, P = 0.961) or to the total cell count (0.07383 ± 0.06005 vs. 0.06491 ± 0.06338, P = 0.633) between primarily resected and neoadjuvantly treated patients with pancreatic cancer (Fig. 5D and E).

Figure 5.

Histopathologic impact of neoTx on neural invasion and tumor proliferation rate of pancreatic cancer. A, Representative photomicrographs of intrapancreatic nerves immunolabeled with the glial marker S100. Note that endoneural (arrows) and perineural (asterisk) invasions are encountered more commonly in PR cases. Scale bars indicate 50 μm. B, Scatterplots comparing the severity of neural invasion (NI), ENI, and PNI in neoadjuvantly treated patients versus PR cases demonstrate diminished infiltration of nerves by tumor cells after neoTx. C, Positive correlation between the activation of the stroma and neural invasion. Pearson correlation coefficient (r) and statistical significance are presented for both correlations. D, Representative images of Ki67+ proliferating tumor cells (arrows) within the tumor area (green). E, Pairwise comparison of the tumor proliferation rate using the tumor area (top graphic, the unit for the y-axis is number of cells per μm2) and DAPI+ tumor cells (bottom graphic) in PR resected versus neoadjuvantly treated patients with pancreatic canecr showed no change in the mitotic rate after neoTx. NS, not significant; PR, primary resected.

Figure 5.

Histopathologic impact of neoTx on neural invasion and tumor proliferation rate of pancreatic cancer. A, Representative photomicrographs of intrapancreatic nerves immunolabeled with the glial marker S100. Note that endoneural (arrows) and perineural (asterisk) invasions are encountered more commonly in PR cases. Scale bars indicate 50 μm. B, Scatterplots comparing the severity of neural invasion (NI), ENI, and PNI in neoadjuvantly treated patients versus PR cases demonstrate diminished infiltration of nerves by tumor cells after neoTx. C, Positive correlation between the activation of the stroma and neural invasion. Pearson correlation coefficient (r) and statistical significance are presented for both correlations. D, Representative images of Ki67+ proliferating tumor cells (arrows) within the tumor area (green). E, Pairwise comparison of the tumor proliferation rate using the tumor area (top graphic, the unit for the y-axis is number of cells per μm2) and DAPI+ tumor cells (bottom graphic) in PR resected versus neoadjuvantly treated patients with pancreatic canecr showed no change in the mitotic rate after neoTx. NS, not significant; PR, primary resected.

Close modal

Stromal activation and intratumoral infiltration with CD4+ T cells and NK cells constitute independent prognostic factors for pancreatic cancer

NeoTx did not show a beneficial effect on the clinical outcome of patients with pancreatic cancer with a median survival of 24 months from the day of therapy initiation against 14 months by primary resected patients (P = 0.1181; Supplementary Fig. S9A). We next sought to identify possible predicting factors for survival based on clinical and histopathologic features of pancreatic cancer. Median values of the cohorts were used as cutoffs except for neural invasion and CD4+ T cells, where preliminary analysis demonstrated that the use of percentiles 75 and 25, respectively, was able to predict more accurately tumor progression. Our univariate analysis showed that histopathologic features such as ASI (P = 0.0157), neural invasion (P = 0.0392), and the intratumoral infiltration with CD4+ T cells (P = 0.0475), and as per tendency also the NK cells (P = 0.0514) were able to predict survival (Fig. 6A–F) in a similar way to traditionally recognized prognostic factors like surgical margins (P = 0.0232), lymph node status (P = 0.0793), and grade of differentiation (P = 0.0316; Supplementary Fig. S9B-D). Unexpectedly, the tumor proliferation rate (proportion of Ki67+ tumor cells) rose as the most significant prognostic marker for pancreatic cancer (P = 0.0013, Fig. 6F). Independently of the treatment group, patients with high numbers of mitotic cells presented a much lower median survival (14 months) than patients with low tumor proliferation rate (40 months, Fig. 6F). Survival analysis using the Kaplan–Meier method supported this hypothesis showing a significant survival benefit for all the abovementioned histopathologic features. To assess the interdependence of these histopathologic features, we performed a Cox regression analysis. Multivariate analysis revealed that high versus low tumor proliferation rate (P = 0.017), ASI (P = 0.027), high versus low infiltration of CD4+ T cells (P = 0.052), and NK cells (P = 0.038) were independent prognostic factors in pancreatic cancer, independent of the pretreatment status (Supplementary Table S4).

Figure 6.

Survival analysis of patients with pancreatic cancer based on the prognostically relevant histopathologic features. A and B, High stromal activation (ASI) correlated with decreased survival in both the whole cohort and when the cohort was stratified by neoadjuvant therapy versus upfront resection. C, High severity of neural invasion (NI) was associated with decreased overall survival. D, High intratumoral infiltration with CD4+ T cells and DCs showed improved clinical outcomes. E, In contrast, the infiltration with NK cells was correlated with poorer prognosis. F, High tumor proliferation rate was associated with worsened survival and showed the strongest potential as prognostic marker. High and low ASI and intratumoral infiltration values were divided on the basis of the median value of the variables, except for neural invasion where the 75th percentile was used and the CD4+ cell density, where the 25th percentile showed to predict survival more accurately than the mean. Significance was calculated using the log-rank test. *P < 0.05, **P < 0.01.

Figure 6.

Survival analysis of patients with pancreatic cancer based on the prognostically relevant histopathologic features. A and B, High stromal activation (ASI) correlated with decreased survival in both the whole cohort and when the cohort was stratified by neoadjuvant therapy versus upfront resection. C, High severity of neural invasion (NI) was associated with decreased overall survival. D, High intratumoral infiltration with CD4+ T cells and DCs showed improved clinical outcomes. E, In contrast, the infiltration with NK cells was correlated with poorer prognosis. F, High tumor proliferation rate was associated with worsened survival and showed the strongest potential as prognostic marker. High and low ASI and intratumoral infiltration values were divided on the basis of the median value of the variables, except for neural invasion where the 75th percentile was used and the CD4+ cell density, where the 25th percentile showed to predict survival more accurately than the mean. Significance was calculated using the log-rank test. *P < 0.05, **P < 0.01.

Close modal

Differential spatial distribution of the components of the tumor microenvironment according to the anatomic localization of pancreatic cancer

Numerous studies have demonstrated a survival benefit of tumors located in the head of the pancreas, compared with those in the body and the tail (21). Our in-depth analysis of the tumor microenvironment revealed substantial spatial differences in the immune milieu of pancreatic cancer (Supplementary Fig. S10). Although pancreatic head cancers presented significantly higher numbers of intratumoral CD45+ leucocytes (255.9 ± 144.9 vs. 175.8 ± 138.2, P = 0.029), especially cytotoxic CD8+ T cells (72.83 ± 52.65 vs. 36.18 ± 25.89, P = 0.027) as well as antitumor immune cells such as DCs (17.68 ± 14.42 vs. 9.064 ± 5.574, P = 0.008) and M1 macrophages (17.56 ± 11.23 vs. 11.42 ± 6.983 P = 0.012; Supplementary Fig. S10A), tumors of the body and the tail were characterized by a pronounced immunosuppressive environment where higher counts of Tregs (0.4648 ± 0.2236 vs. 0.5618 ± 0.2307, P = 0.054) were accompanied by increased stromal activation (αSMA area: 0.04188 ± 0.03070 vs. 0.06248 ± 0.04890, P = 0.020; Supplementary Fig. S10B).

This study aimed at shedding light on the underlying mechanisms that are responsible for the survival benefit after neoTx in patients with borderline resectable and locally advanced pancreatic cancer. Through a detailed histopathologic analysis, we demonstrated the ability of neoadjuvant approaches to boost the antitumor immune response in the microenvironment of pancreatic cancer, which arose in parallel with the suppression of stromal activation and neural invasion (Supplementary Fig. S11; Supplementary Table S5). These findings strongly suggest that the presence of an antitumoral immune response induced by neoTx could restrain tumor growth and reduce cancer aggressiveness, highlighting the urgent need of new therapeutic strategies, especially neoadjuvant immunotherapy (22, 32–35).

The efficacy of immunotherapy in pancreatic cancer is widely hampered by its low immunogenic potential compared with melanoma (23) and its largely immunosuppressive microenvironment dominated by Treg and MDSC infiltration (16, 23). Our results, however, demonstrate that neoTx can reverse these premises and reactivate the local immune response. Here, we hypothesize that neoadjuvant settings may unleash a more potent and prolonged antitumor immune response compared with adjuvant therapies due to the presence of the tumor burden serving as antigen source for DC-mediated priming and expansion of T cells (6). On advanced melanoma and pancreatic cancer, preclinical studies on different immunotherapy regimes (anti-PD1, anti-PDL1, anti-CD137, and anti-CD25) have shown that the survival benefit is undoubtedly associated with neoTx timing (6, 8). Thus, we consider that only combinational therapies that exploit the immunologic benefits of neoTx and complement this with systemic immunotherapy will be able to achieve a significant therapeutic success in pancreatic cancer.

The first study that analyzed the immunologic impact of neoTx on pancreatic cancer was carried out by Homma and colleagues and reported significantly higher numbers of CD8+ and CD4+ T cells and a reduction of the density of FOXP3+ Tregs (24, 25) after therapy. Later reports on pancreatic cancer, however, underscore the selectivity of neoTx regimes in the depletion of FOXP3+CD4+ Tregs, which secondarily leads to increased activity of effector CD8+ T cells (25, 26). However, Tregs are not the only cell type involved in tumor-induced immunosuppression (17); MDSCs have also been identified as major modulators of tumor tolerance (27). Recent evidence demonstrated that gemcitabine and 5-fluorouracil are able to induce a transient selective depletion of MDSCs, resulting in the enhancement of CD8+ T-cell antitumor activity (17). Moreover, unlike in breast and prostate cancer, the number of mitotic tumor cells kept constant after neoTx in pancreatic cancer (28, 29), which suggests that the direct cytotoxicity of chemotherapeutical agents on pancreatic tumor cells accounts merely a minor role and seems to be subordinated to the immune cell–mediated antitumor response.

A recent study on pancreatic cancer reported that neoTx is able to redirect the polarization of macrophages into the “classically activated” M1 phenotype, providing evidence that macrophages might develop an antitumorigenic role if properly educated by therapeutics (18). Our results, however, suggest a minor role of tumor-associated macrophages on the immunologic reactivation mediated by neoTx on pancreatic cancer, as the density of both M1- and M2-polarized macrophages remained unaltered after therapy (15).

Growing evidence is pointing to activated PSCs and endothelial cells as key regulators of the immune cell infiltration into the tumor microenvironment (10). The paucity of functional vasculature close to pancreatic cancer tumor cells is well established (11) and according to our results, it is not altered by neoadjuvant approaches. Although a strong correlation was shown between angiogenesis and the tumor infiltration with CD8+ T cells and Tregs, which supports the hypothesis of a pancreatic cancer–specific CD8+ trafficking defect within pancreatic cancer microvessels (11), this does not seem to be the principal mechanism of the immunologic reactivation after neoTx. Pancreatic neuropathy is one of the hallmarks of pancreatic cancer and is an important way of cancer spread and tumor recurrence (30, 31). The presence of numerous hypertrophic nerves, however, seems to be part of even the early stages of pancreatic cancer neuropathy, as, based on our results, they are irreversible by the time of chemotherapeutical administration. In accordance with the existing data, we showed a prolonged overall survival of patients with low severity of neural invasion, which again highlights the importance of the generation of neural invasion–targeting therapies (31).

In conclusion, we performed the currently most comprehensive histopathologic analysis that examined the role of neoTx on borderline resectable and locally advanced pancreatic cancer and demonstrated the ability of neoTx to enrich antitumor immune cells in the tumor microenvironment. This study was not able to detect a dependence of this effect from the type of applied pretreatment regimen, and as a limitation, it included only patients who per se responded to therapy, albeit at a varying extent. Still, this effect was particularly characterized by selective depletion of Tregs and MDSCs, which was associated with increased T-cell antitumor activity. The possibility of combining traditional neoTx with immunotherapy may unleash long-term and more effective antitumoral immune responses, beneficially alter the multiple immune–stroma-activating cross-talks, and eventually improve the prognosis of borderline resectable and locally advanced pancreatic cancer.

W. Weichert reports receiving commercial research grants from Roche, MSD, Bristol-Myers Squibb, and Bruker, and speakers bureau honoraria from Roche, MSD, Bristol-Myers Squibb, AstraZeneca, Pfizer, Merck, Lilly, Boehringer, Novartis, and Takeda. H. Friess is an advisory board member/unpaid consultant for McKinsey. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C. Mota Reyes, G.O. Ceyhan, I.E. Demir

Development of methodology: C. Mota Reyes, S. Teller, G.O. Ceyhan, I.E. Demir

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Mota Reyes, A. Muckenhuber, B. Konukiewitz, W. Weichert, I.E. Demir

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Mota Reyes, A. Muckenhuber, B. Konukiewitz, O. Safak, H. Friess, G.O. Ceyhan, I.E. Demir

Writing, review, and/or revision of the manuscript: C. Mota Reyes, S. Teller, O. Safak, W. Weichert, H. Friess, G.O. Ceyhan, I.E. Demir

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Mota Reyes, O. Safak, H. Friess, I.E. Demir

Study supervision: S. Teller, H. Friess, G.O. Ceyhan, I.E. Demir

I.E. Demir and G.O. Ceyhan were supported in this project by a grant of the German Cancer Aid (Deutsche Krebshilfe, no. 70112897). The results presented in this work are part of C. Mota Reyes's doctoral thesis.

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.
Zhang
J
,
Xu
X
,
Shi
M
,
Chen
Y
,
Yu
D
,
Zhao
C
, et al
CD13(hi) Neutrophil-like myeloid-derived suppressor cells exert immune suppression through Arginase 1 expression in pancreatic ductal adenocarcinoma
.
Oncoimmunology
2017
;
6
:
e1258504
.
2.
Strobel
O
,
Neoptolemos
J
,
Jager
D
,
Buchler
MW
. 
Optimizing the outcomes of pancreatic cancer surgery
.
Nat Rev Clin Oncol
2019
;
16
:
11
26
.
3.
McCoy
MJ
,
Hemmings
C
,
Miller
TJ
,
Austin
SJ
,
Bulsara
MK
,
Zeps
N
, et al
Low stromal Foxp3+ regulatory T-cell density is associated with complete response to neoadjuvant chemoradiotherapy in rectal cancer
.
Br J Cancer
2015
;
113
:
1677
86
.
4.
Gupta
A
,
Probst
HC
,
Vuong
V
,
Landshammer
A
,
Muth
S
,
Yagita
H
, et al
Radiotherapy promotes tumor-specific effector CD8+ T cells via dendritic cell activation
.
J Immunol
2012
;
189
:
558
66
.
5.
Nejati
R
,
Goldstein
JB
,
Halperin
DM
,
Wang
H
,
Hejazi
N
,
Rashid
A
, et al
Prognostic significance of tumor-infiltrating lymphocytes in patients with pancreatic ductal adenocarcinoma treated with neoadjuvant chemotherapy
.
Pancreas
2017
;
46
:
1180
7
.
6.
Liu
J
,
Blake
SJ
,
Yong
MC
,
Harjunpaa
H
,
Ngiow
SF
,
Takeda
K
, et al
Improved efficacy of neoadjuvant compared to adjuvant immunotherapy to eradicate metastatic disease
.
Cancer Discov
2016
;
6
:
1382
99
.
7.
Gurlevik
E
,
Fleischmann-Mundt
B
,
Brooks
J
,
Demir
IE
,
Steiger
K
,
Ribback
S
, et al
Administration of gemcitabine after pancreatic tumor resection in mice induces an antitumor immune response mediated by natural killer cells
.
Gastroenterology
2016
;
151
:
338
50
.
8.
Dunn
GP
,
Old
LJ
,
Schreiber
RD
. 
The three Es of cancer immunoediting
.
Annu Rev Immunol
2004
;
22
:
329
60
.
9.
Protti
MP
,
De Monte
L
. 
Immune infiltrates as predictive markers of survival in pancreatic cancer patients
.
Front Physiol
2013
;
4
:
6
.
10.
Watt
J
,
Kocher
HM
. 
The desmoplastic stroma of pancreatic cancer is a barrier to immune cell infiltration
.
Oncoimmunology
2013
;
2
:
e26788
.
11.
Ene-Obong
A
,
Clear
AJ
,
Watt
J
,
Wang
J
,
Fatah
R
,
Riches
JC
, et al
Activated pancreatic stellate cells sequester CD8+ T cells to reduce their infiltration of the juxtatumoral compartment of pancreatic ductal adenocarcinoma
.
Gastroenterology
2013
;
145
:
1121
32
.
12.
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
.
13.
Lee
SM
,
Katz
MH
,
Liu
L
,
Sundar
M
,
Wang
H
,
Varadhachary
GR
, et al
Validation of a proposed tumor regression grading scheme for pancreatic ductal adenocarcinoma after neoadjuvant therapy as a prognostic indicator for survival
.
Am J Surg Pathol
2016
;
40
:
1653
60
.
14.
Evans
DB
,
Rich
TA
,
Byrd
DR
,
Cleary
KR
,
Connelly
JH
,
Levin
B
, et al
Preoperative chemoradiation and pancreaticoduodenectomy for adenocarcinoma of the pancreas
.
Arch Surg
1992
;
127
:
1335
9
.
15.
Erkan
M
,
Michalski
CW
,
Rieder
S
,
Reiser-Erkan
C
,
Abiatari
I
,
Kolb
A
, et al
The activated stroma index is a novel and independent prognostic marker in pancreatic ductal adenocarcinoma
.
Clin Gastroenterol Hepatol
2008
;
6
:
1155
61
.
16.
Songchuan Guo
MC
,
Miller
G
,
Leichman
L
,
Wu
J
. 
Immunotherapy in pancreatic cancer: unleash its potential through novel combinations
.
World J Clin Oncol
2017
;
8
:
230
40
.
17.
Vincent
J
,
Mignot
G
,
Chalmin
F
,
Ladoire
S
,
Bruchard
M
,
Chevriaux
A
, et al
5-Fluorouracil selectively kills tumor-associated myeloid-derived suppressor cells resulting in enhanced T cell-dependent antitumor immunity
.
Cancer Res
2010
;
70
:
3052
61
.
18.
Di Caro
G
,
Cortese
N
,
Castino
GF
,
Grizzi
F
,
Gavazzi
F
,
Ridolfi
C
, et al
Dual prognostic significance of tumour-associated macrophages in human pancreatic adenocarcinoma treated or untreated with chemotherapy
.
Gut
2016
;
65
:
1710
20
.
19.
Kalluri
R
. 
The biology and function of fibroblasts in cancer
.
Nat Rev Cancer
2016
;
16
:
582
98
.
20.
Ceyhan
GO
,
Bergmann
F
,
Kadihasanoglu
M
,
Altintas
B
,
Demir
IE
,
Hinz
U
, et al
Pancreatic neuropathy and neuropathic pain—a comprehensive pathomorphological study of 546 cases
.
Gastroenterology
2009
;
136
:
177
86
.
21.
Dreyer
SB
,
Jamieson
NB
,
Upstill-Goddard
R
,
Bailey
PJ
,
McKay
CJ
, et al
,
Australian Pancreatic Cancer Genome Initiative
, 
Defining the molecular pathology of pancreatic body and tail adenocarcinoma
.
Br J Surg
2018
;
105
:
e183
e91
.
22.
Parra
ER
,
Villalobos
P
,
Behrens
C
,
Jiang
M
,
Pataer
A
,
Swisher
SG
, et al
Effect of neoadjuvant chemotherapy on the immune microenvironment in non-small cell lung carcinomas as determined by multiplex immunofluorescence and image analysis approaches
.
J Immunother Cancer
2018
;
6
:
48
.
23.
Rosenberg
A
,
Mahalingam
D
. 
Immunotherapy in pancreatic adenocarcinoma-overcoming barriers to response
.
J Gastrointest Oncol
2018
;
9
:
143
59
.
24.
Homma
Y
,
Taniguchi
K
,
Murakami
T
,
Nakagawa
K
,
Nakazawa
M
,
Matsuyama
R
, et al
Immunological impact of neoadjuvant chemoradiotherapy in patients with borderline resectable pancreatic ductal adenocarcinoma
.
Ann Surg Oncol
2014
;
21
:
670
6
.
25.
Shibuya
KC
,
Goel
VK
,
Xiong
W
,
Sham
JG
,
Pollack
SM
,
Leahy
AM
, et al
Pancreatic ductal adenocarcinoma contains an effector and regulatory immune cell infiltrate that is altered by multimodal neoadjuvant treatment
.
PLoS One
2014
;
9
:
e96565
.
26.
Tsuchikawa
T
,
Hirano
S
,
Tanaka
E
,
Matsumoto
J
,
Kato
K
,
Nakamura
T
, et al
Novel aspects of preoperative chemoradiation therapy improving anti-tumor immunity in pancreatic cancer
.
Cancer Sci
2013
;
104
:
531
5
.
27.
Liu
Q
,
Liao
Q
,
Zhao
Y
. 
Chemotherapy and tumor microenvironment of pancreatic cancer
.
Cancer Cell Int
2017
;
17
:
68
.
28.
Tretiakova
MS
,
Wei
W
,
Boyer
HD
,
Newcomb
LF
,
Hawley
S
,
Auman
H
, et al
Prognostic value of Ki67 in localized prostate carcinoma: a multi-institutional study of >1000 prostatectomies
.
Prostate Cancer Prostatic Dis
2016
;
19
:
264
70
.
29.
Ghanim
B
,
Klikovits
T
,
Hoda
MA
,
Lang
G
,
Szirtes
I
,
Setinek
U
, et al
Ki67 index is an independent prognostic factor in epithelioid but not in non-epithelioid malignant pleural mesothelioma: a multicenter study
.
Br J Cancer
2015
;
112
:
783
92
.
30.
Lagarde
SM
,
Phillips
AW
,
Navidi
M
,
Disep
B
,
Immanuel
A
,
Griffin
SM
. 
The presence of lymphovascular and perineural infiltration after neoadjuvant therapy and oesophagectomy identifies patients at high risk for recurrence
.
Br J Cancer
2015
;
113
:
1427
33
.
31.
Chatterjee
D
,
Katz
MH
,
Rashid
A
,
Wang
H
,
Iuga
AC
,
Varadhachary
GR
, et al
Perineural and intraneural invasion in posttherapy pancreaticoduodenectomy specimens predicts poor prognosis in patients with pancreatic ductal adenocarcinoma
.
Am J Surg Pathol
2012
;
36
:
409
17
.
32.
Lin
HJ
,
Lin
J
. 
Seed-in-soil: pancreatic cancer influenced by tumor microenvironment
.
Cancers
2017
;
9
:
pii: E93
.
33.
Chang
JH
,
Jiang
Y
,
Pillarisetty
VG
. 
Role of immune cells in pancreatic cancer from bench to clinical application: an updated review
.
Medicine
2016
;
95
:
e5541
.
34.
Jang
JE
,
Hajdu
CH
,
Liot
C
,
Miller
G
,
Dustin
ML
,
Bar-Sagi
D
. 
Crosstalk between regulatory T cells and tumor-associated dendritic cells negates anti-tumor immunity in pancreatic cancer
.
Cell Rep
2017
;
20
:
558
71
.
35.
Pylayeva-Gupta
Y
,
Das
S
,
Handler
JS
,
Hajdu
CH
,
Coffre
M
,
Koralov
SB
, et al
IL35-Producing B cells promote the development of pancreatic neoplasia
.
Cancer Discov
2016
;
6
:
247
55
.

Supplementary data