Reduced Immune-Cell Infiltration in MHC Class I Negative DLBCL

Abstract

Background Diffuse large B-cell lymphoma (DLBCL) is an aggressive malignancy and the most common type of non-Hodgkin B-cell lymphoma. It is a genetically, biologically and clinically heterogeneous disease and current research focuses on identifying stratification biomarkers to identify high-risk patients and guide selection of therapy. The genetic landscape of DLBCL is well characterized, but how these mutations affect the lymphoma microenvironment is not fully understood. We aimed to categorize DLBCL cases based on tumor microenvironment (TME) features and to identify how tumor cell features including loss of MHC expression shape the TME. Methods Diagnostic biopsies from 58 high-risk de novo DLBCL NOS from two Nordic phase II clinical trials (PMID: 32380536, PMID: 23247661) were imaged by Hyperion imaging mass cytometry. Single-cell segmentation was followed by clustering of 14 lineage markers and manual annotation to identify eight cell types. Using CytoMAP (PMID: 32320656), the cells were divided into 20-µm-radius neighborhoods and clustered to define six types of tissue regions (Figure 1). Unsupervised hierarchical clustering of the region prevalence in each image was used to classify different types of tumor microenvironments. To identify tumor cell features, tumor-specific expression of markers related to immune evasion (CD54, PD-L1) and prognostic markers (FOXP1, Ki67) were quantified by manual gating of positive cells. Cases were classified as MHC class I loss when less than 50% of the tumor cells expressed HLA A or HLA B, and Ki67-positive when more than 10% of the tumor cells were gated as positive. The LymphGen algorithm was used to classify tumor genotypes. Results Clustering of the region prevalence in each image revealed six types of tumor microenvironments; immune-cell depleted (47%), tumor/immune cell mixed (29%) and four TMEs with low tumor-cell fractions that were dominated by CD4 T cells (7%), CD8 T cells (7%), M1/M2 macrophages (5%) or collagen-positive stromal cells (5%). In cases classified as immune-cell depleted, more than 70% of the cells were tumor cells and less than 15% were infiltrating immune cells. The four groups with low tumor-cell fractions were merged for statistical analyses. TME subclasses were not correlated with tumor genotypes, but DLBCL cases classified as immune-cell depleted were overrepresented among the Ki67-positive cases. Loss of MHC class I was seen in 34% of the cases and was associated with the immune-cell depleted TME subclass. Furthermore, MHC class I loss correlated with lower infiltration of CD4 and CD8 T cells. Tumor-cell specific expression of MHC class II, PD-L1, CD54, or FOXP1 did not significantly correlate with the TME subclasses. Conclusion In conclusion, our study contributes to improved understanding of the DLBCL microenvironment through mapping of the diversity of immune-cell composition. We found that loss of MHC class I expression was associated with an immune cell-depleted TME. The prominent loss of MHC has potential impact for the selection of patients for CAR T-cell therapy vs. bispecific antibodies/T-cell engagers. Together, refined classification tools of DLBCL based on genetic aberrations and microenvironment architecture may improve strategies of personalized medicine.