Special Anatomical Subtypes of Diffuse Large B-Cell Lymphoma Have Distinct Tumor Microenvironments

Background The heterogeneity of diffuse large B-cell lymphoma (DLBCL) and its diverse clinical course encouraged the World Health Organization (WHO) to recognize specific subtypes based on anatomical location. These anatomical subtypes are accompanied by specific mutational profiles and gene-expression patterns. For example, the poor-prognostic primary DLBCL of immune-privileged sites (IP-DLBCL) originating in the central nervous system or testis have frequent NF-κB-activating mutations and a predominantly activated B-cell (ABC) phenotype. In contrast, we recently showed that good-prognostic bone DLBCL (B-DLBCL) have frequent mutations in immune-modulating and epigenetic regulating genes and a germinal center B-cell (GCB) phenotype (PMID: 34478526). We hypothesize that these remarkable anatomical site-specific enrichments for tumors with a distinct molecular signature are associated with different local anti-tumor immune responses. Here, IP-DLBCL and B-DLBCL could be considered opposing ends of a spectrum and we compared the composition of the tumor microenvironment (TME) of these entities to each other and to DLBCL with nodal localization only (nodal-DLBCL).

Method This retrospective study used formalin fixed and paraffin embedded biopsy specimens that were selected from clinically, histologically and molecularly well-annotated cases of IP-DLBCL (n=29), B-DLBCL (n=21) and nodal-DLBCL (n=18). All cases were diagnosed according to the revised 2016 WHO classification. High grade B-cell lymphomas (double/triple hit) were excluded. Imaging mass cytometry (Hyperion) was used to define the TME at single-cell resolution and with intact spatial relationships. The panel comprised 41 antibody-markers for deep phenotyping of four major cellular compartments: tumor cells, tumor-infiltrating lymphocytes, myeloid- and stromal cells. Image output was processed to create single-cell data, as described before (PMID: 34196108). Cell phenotyping was performed by supervised clustering after dimension reduction using opt-SNE algorithm (OMIQ software). Cluster abundances were depicted as median (±IQR) per thousand of total cells per patient(‰). Kruskal-Wallis testing was used to determine statistical differences between abundances.

Results In total 1,7M single cells (range 7-49k) were obtained and 90 unique cellular phenotypes were defined. Differences in composition of the four cellular compartments were compared between IP-DLBCL and B-DLBCL (Figure 1). The lymphocyte compartment of IP-DLBCL was significantly enriched with granzyme B expressing lymphocytes, mainly cytotoxic T cells (IP 9,4‰ ± 8,8 versus B-DLBCL 2,7‰ ± 3,8; p<0,01), including activation related markers like CCR4 and Ki-67. In contrast, B-DLBCL was characterized by a substantial influx of T cells, more specifically T-helper cells (B-DLBCL 49,9‰ ± 32,5 versus IP 35,4‰ ± 24,6; p<0,01), including T-regulatory cells (B-DLBCL 2,7‰ ± 4,4 versus IP 0,9‰ ± 1,5; p<0,01) and cells expressing ICOS. T-helper 1 cells, expressing TBET, were less abundant (B-DLBCL 0,3‰ ± 0,5 versus IP 1,0‰ ± 0,6; p<0,01). The myeloid compartment was generally increased in B-DLBCL and comprised of HLADR and/or CD11c expressing dendritic cells (B-DLBCL 84,3‰ ± 39,8 vs IP 32,5‰ ± 31,9; p<0,01). In the stromal compartment, a significantly higher expression of activation markers related to the "stromal1" gene signature (PMID: 34597589) in B-DLBCL was identified (B-DLBCL 15,4‰ ± 12,4 vs IP 10,98‰ ± 9,0; p=0,04). Immune checkpoint markers like PD1, LAG3, TIM3, TIGIT and VISTA were generally low in expression within the entire cohort. In nodal-DLBCL both IP- as well as B-DLBCL-like phenotypes were identified.

Conclusion This study shows significant differences in the tumor microenvironment of DLBCL subtypes based on anatomical site. In IP-DLBCL, substantial presence of cytotoxic T cells-with activated phenotypes-was identified. In contrast, in bone-DLBCL the TME was generally non-cytotoxic and exhibited immune regulatory elements. Altogether, knowledge on these distinct tumor microenvironments may help understanding heterogeneity between these subtypes and advocates correlation with efficacy of emerging treatments like bispecific antibodies (for the cytotoxic TME) and CAR T-cell therapy (for the non-cytotoxic/immune regulatory TME).

Griffioen:Miltenyi Biotec B.V & Co. KG: Research Funding. Nijland:Genmab: Consultancy; Takeda: Research Funding; Roche: Research Funding. Mutseaers:Glaxo Smith Kline: Consultancy; Astra Zeneca: Research Funding; BMS: Consultancy. Chamuleau:Genmab: Research Funding; Abbvie: Honoraria; Novartis: Honoraria; Roche: Honoraria; BMS/Celgene: Honoraria, Research Funding; Gilead: Research Funding. Kersten:BMS/Celgene: Honoraria, Research Funding; Kite, a Gilead Company: Honoraria, Research Funding; Adicet Bio: Honoraria; Takeda: Honoraria; Roche: Honoraria, Research Funding; Novartis: Honoraria; Miltenyi Biotech: Honoraria. Diepstra:Takeda: Membership on an entity's Board of Directors or advisory committees, Research Funding. Bovée:Tracon pharmaceuticals: Research Funding.