Clinical prognosticators and targets in the immune microenvironment of intrahepatic cholangiocarcinoma

Abstract

Background: Intrahepatic cholangiocarcinoma (ICC) is a disease with poor prognosis and limited therapeutic options. We investigated the tumor immune microenvironment (TIME) to identify predictors of disease outcome and to explore targets for therapeutic modulation.

Methods: Liver tissue samples were collected during 2008-2019 from patients (n = 139) diagnosed with ICC who underwent curative intent surgery without neoadjuvant chemotherapy. Samples from the discovery cohort (n = 86) were immunohistochemically analyzed on tissue microarrays (TMAs) for the expression of CD68, CD3, CD4, CD8, Foxp3, PD-L1, STAT1, and p-STAT1 in tumor core and stroma areas. Results were digitally analyzed using QuPath software and correlated with clinicopathological characteristics. For validation of TIME-related biomarkers, we performed multiplex imaging mass cytometry (IMC) in a validation cohort (n = 53).

Results: CD68+ cells were the predominant immune cell type in the TIME of ICC. CD4+^high T cell density correlated with better overall survival (OS). Prediction modeling together with validation cohort confirmed relevance of CD4+ cells, PD-L1 expression by immune cells in the stroma and N-stage on overall disease outcome. In turn, IMC analyses revealed that silent CD3+CD4+ clusters inversely impacted survival. Among annotated immune cell clusters, PD-L1 was most relevantly expressed by CD4+FoxP3+ cells. A subset of tumors with high density of immune cells ("hot" cluster) correlated with PD-L1 expression and could identify a group of candidates for immune checkpoint inhibition (ICI). Ultimately, higher levels of STAT1 expression were associated with higher lymphocyte infiltration and PD-L1 expression.

Conclusions: These results highlight the importance of CD4+ T cells in immune response against ICC. Secondly, a subset of tumors with "hot" TIME represents potential candidates for ICI, while stimulation of STAT1 pathway could be a potential target to turn "cold" into "hot" TIME in ICC.

Keywords: Immune cell prognosticators; immunomodulation; intrahepatic cholangiocarcinoma; tumor immune microenvironment.

Figure 1.

Tumor immune cell infiltration (TIME) in ICC. (a). Flow diagram characterizes the patient selection of surgically treated patients from 2008-2019 at Charité Universitätsmedizin Berlin with histopathological confirmed diagnosis of intrahepatic cholangiocellular carcinoma (ICC). In a two-step trial design patients were randomly allocated into discovery and validation cohort. (b). Representative delimitation of whole TMA section (WTS) into tumor core (TE) and stromal area (ST) following the histological evaluation of HE stainings by two independent pathologists. (c). Kruskal-Wallis multiple comparison between groups showing higher densities of lymphocytes in stromal area in comparison to tumor core. ***, p < 0.001. (d). Correlation matrix among immune cell densities, PD-L1 expression, and STAT1 (Spearman’s rho). Most immune cell populations correlated with each other (heat map represents correlation coefficient r). (e). CD68+ TE densities with standard errors of the mean (SEM) for each UICC stage. (f). Lymphocyte densities for each UICC stage. Two way ANOVA analysis with Tukey’s multiple comparison among stages with significant differences on patients with distant metastases. **, p < 0.01.

Figure 2.

The impact of T-cell infiltration on patient survival. (a). Representative micrographs of “high” and “low” immune cell infiltration of CD4+ cells in WTS, TE, and ST (scale bars represent 100 µm). (b). Kaplan-Meier analysis for overall survival: CD4+high vs. CD4+low WTS. (c). Kaplan-Meier analysis for disease-free survival: CD4+high vs. CD4+low WTS.

Figure 3.

The role of PD-L1 on tumor and tumor-associated immune cells. (a). PD-L1 expression on tumor cells (membranous staining) and PD-L1 staining on tumor associated immune cells (scale bars represent 100 µm). (b). Overall survival Kaplan-Meier curves for PD-L1 tumor expression and PD-L1 expression on immune cells (negative vs. positive). (c). Kaplan-Meier curves representing overall survival for patients with PD-L1 positive expression on tumor cells and adjuvant chemotherapy response.

Figure 4.

Immune cell clusters in ICC. (a). Immune cell densities (cells/mm²) for each cluster. Color intensity represents input predictor importance. (b). PD-L1 expression on tumor cells and immune cells for each TME cluster (Mann-Whitney U test,**, p < 0.01, and ***, p < 0.001, respectively). (c). Immune “cold” subset survival depending on PD-L1 expression on tumor associated-inflammatory cells.

Figure 5.

Spatial IMC analysis of TIME in the validation cohort. (a). Kruskal-Wallis multiple comparison between groups showing predominant CD45+CD68+ macrophages and major lymphocyte populations in WTS, TE and ST. (b). Spatial segmentation of WTS sections for identification of tumoral (TE) and stromal (ST) areas (red=te, blue=st). (c). Unbiased T cell clustering performed by Phenograph using specified panel (on the right) after sorting of 38,673 CD45+CD3+CD19– cells across validation cohort identified 14 major immune cell sub-classes. (d). Left: Representative IMC micrographs (P124, P122) representing the major immune cell subtypes (CD45+CD68+, CD3+CD4+, CD3+CD8+, CD3+CD4+FoxP3+) in a immunologically “cold” versus “hot” TIME. Right: pie charts with fraction of clustered “hot” versus “cold” patients. (e). Kaplan-Meier analysis for overall survival in the validation cohort: CD4+high vs. CD4+low WTS. (f). Neighborhood analyses of pooled clusters displayed longer distances of CD4+ clusters to CD4+FoxP3+ and CD68+ clusters in CD4+^high tumors. *, p < 0.05. (g). Correlation analyses with cumulative densities (cell counts per mm2) of silent CD4+ clusters plotted against the post-operative survival in days in WTS, ST and TE. (h). Expression analyses of PD-L1 in pooled clusters identified CD4+FoxP3+ cells with significantly higher expression compared to other T cell or macrophage clusters.**, p < 0.01; 2 way ANOVA analysis with Bonferroni correction.

Figure 6.

STAT1 axis relevant for immune cell modulation. (a). Representative micrographs after STAT1 and p-STAT1 staining. (b). STAT1 staining with distinct intensities defined with by QuPath software (yellow= low, orange= moderate, red=high). (c). H-score STAT1 intensity groups (low, moderate, high) in correlation with lymphocytes densities. “Moderate” STAT1 levels showed more infiltration of lymphocytes compared to “low” and “high” groups (2 way ANOVA with Tukey´s multiple comparison). (d). H-score STAT1 intensity vs. PD-L1 expression on tumor cells ranks (columns with standard error of mean). Kruskal-Wallis test with Dunn’s multiple comparisons showed significant higher expression of PD-L1 in tumors with “moderate” and “high” STAT1 scores in comparison with “low” score (p < 0.05). (e). H-score for PD-L1 expression on tumor cells (negative: 110.8 vs. positive: 193.7). Mann-Whitney U test, p < 0.001. (f). OS for STAT1high with PD-L1 positive expression on tumor cell population vs. STAT1high with PD-L1 negative expression on tumor cell population.