Profiling the Immune Tumor Microenvironment of Newly Diagnosed Diffuse Large B Cell Lymphoma Identifies Three Major Immune Infiltration Patterns

Background: Diffuse large B-cell lymphoma (DLBCL) is a genetically heterogeneous disease of malignant B cells most often classified by tumor gene expression and/or mutations. DLBCL is also characterized by a tumor microenvironmental influence that has not been well described. Immuno-oncology-targeting agents such as immune checkpoint inhibitors have limited clinical activity in DLBCL, highlighting the need for a better understanding of the DLBCL tumor microenvironment for rational drug development, combinations, and disease stratification. Here, we systematically characterize the immune composition of more than 100 DLBCL tumors using 2 imaging technologies and provide insight into the complexity of DLBCL disease biology.

Methods: A total of 110 cases of newly diagnosed DLBCL were analyzed by multiplex immunohistochemistry (IHC; n=70) and multiplexed ion beam imaging (MIBI) (n=40), each with 10 common markers (CD20, CD3, CD8, Foxp3, CD56, CD163, CD11c, CD56, PD-1, PD-L1) and a few platform-specific markers. Both IHC and MIBI images were digitalized to generate marker-positive cell counts (including single, double, or triple positivity), cell density (cell count/mm²), and population fraction (% of total nucleated cells). Marker density was analyzed for the major components of tumor-infiltrating cell types and correlation between any pair of markers. In addition, RNAseq data and fluorescence in situ hybridization (FISH) data on MYC, BCL-2, and BCL-6 translocation were generated.

Results: In the IHC cohort, T cells (CD3⁺), dendritic cells (DCs; by CD11c⁺), and macrophages (CD163⁺)were the major immune components, with median population fractions of 22%, 16%, and 2.7%, respectively. Natural killer cells (CD56⁺CD20⁻) were a minor component at a median of 0.1%. A significant negative correlation was observed between tumor cells (CD20⁺) and CD4⁺ (Spearman ρ = −0.47; P = 1.3 × 10⁻⁰⁴), and CD8⁺ T cells (Spearman ρ = −0.42; P = 1.4 × 10⁻⁰³) cells, and an unexpectedly negative correlation between DCs (CD11c⁺) and macrophages (CD163⁺, r = −0.63; P = 9.8 × 10⁻⁰⁶) was found. Similar to follicular lymphoma, 2 PD-1⁺ T-cell populations were identified: PD-1^bright and PD-1^dim. The PD-1^bright cells co-stained with CXCR5, indicating T follicular helper cells (Tfh). The PD-1^dim were expressed on exhausted effector cells that co-stained with Tim3 or Lag3. The median population fraction of Tim3⁺ or Lag3⁺ among T cells was 25%, whereas that of PD-1^dim among T cells was 0.2%, indicating that PD-1 was not a useful marker for exhausted T cells. PD-L1 was predominantly found on DCs and macrophages, and the median population fraction of PD-L1⁺ among tumor cells was only 6.2%. By unsupervised hierarchical clustering on marker density, 3 major immune-infiltration patterns (P1, P2, and P3) were identified. The first 2 segments (P1 and P2) were 20% and 25% of the total cases, respectively. Both were characterized by high T-cell, macrophage, and DC infiltration. P1 was enriched for PD-1⁺ T cells, whereas P2 was void of any PD-1⁺ cells. The third segment (P3) that comprised 55% of the cases was predominantly tumor cells with low T-cell, macrophage, and DC infiltration. PD-L1⁺ cells were primarily found in segments P1 and P2 but rare in segment P3. Additional analysis on the associations between the 3 immune-infiltration patterns and prognostic DLBCL molecular features such as cell-of-origin, double-hit gene signature, and double-hit FISH will also be presented.

Conclusions: These data show the complexity of DLBCL disease biology and show classification of DLBCL at the immune-infiltration level as 3 distinct patterns. The overall low expression of PD-1⁺ T cells and the restricted pattern of PD-L1⁺ tumor cells provide a possible explanation for the lack of clinical activity of PD-1/PD-L1 blockade in DLBCL. We also observed other exhausted T cells expressing Lag3 and Tim3, suggesting alternative therapeutic opportunities in stratified populations. These data also highlight the opportunity to develop rational immuno-oncology-targeted agents based on the immune infiltration pattern of DLBCL and selection of patients who may respond more favorably to particular agents.