Heat Shock Factor 1 Reprograms the DLBCL Microenvironment to Evade Immune Surveillance and Support Tumor Growth

The primary components of the stromal lymphoma microenvironment (LME) are the cancer-associated fibroblasts (CAF) and the extracellular matrix (ECM) they produce. CAFs are derived from healthy fibroblasts that have been hijacked and transcriptionally reprogrammed by cancer cells into a novel biological entity to promote tumorigenesis. While the key effectors of these programs are largely unknown, recent data indicates that in a variety of solid tumors and T-acute lymphoblastic leukemias an important transcriptional programmer of the microenvironment is the Heat Shock Factor 1 (HSF1). Hence, we sought to determine the contribution of HSF1 within the stromal LME to the acquisition of lymphomagenic features in diffuse large B-cell lymphoma (DLBCL).

We first analyzed the activity of HSF1 in 80 DLBCL patients by RNA-sequencing. Patients were segregated into having "high" and "low" HSF1 activity based on the expression of the canonical stress response target genes. We found distinct patterns of stromal-associated genes between DLBCL having "low" vs. "high" HSF1 activity. Furthermore, ChIP-sequencing data of lymphoma cells and CAFs showed the emergence of distinct transcriptional programs that are differentially orchestrated by HSF1. To delve deeper into the mechanism of these effects, we co-cultured lymphoma cells with fibroblasts having either HSF1 WT or HSF1 null in 3D collagen-containing lymphoma organoids. We found that only organoids harboring HSF1 WT fibroblasts were supportive of DLBCL cell proliferation. Moreover, DLBCL cells responsive and non-responsive to HSF1 downregulation in a cell-autonomous manner, failed to proliferate when cultured in HSF1-deficient LME organoids.

To further elucidate this mechanism, we implanted a murine HSF1-positive DLBCL cell line (A20) into HSF1 WT (n=13) and HSF1 null (n=11) syngeneic mice and monitored lymphoma development for up to 14 days. We observed a continuous growth of HSF1-positive murine DLBCL in HSF1 WT mice; however, DLBCL in HSF1 null mice underwent spontaneous collapse after day 7, leading to almost complete tumor eradication. We further investigated the lymphoma microarchitecture, ECM biomechanics and composition in relation to the cellular components in serial tumor sections at several time points. At day 7, when no significant difference in tumor sizes exists between HSF1 WTs and nulls, we found significant differences in tissue stiffness (p=0.017), collagen microarchitectureby fiber alignment (p<0.001) and elastic fiber area (p<0.001). ECM proteomics by HPLC-MS/MS of the ECM-enriched fractions of these murine DLBCLs at days 5, 7 and 13 followed by network analysis demonstrated highly dynamic changes in ECM composition. Specifically, comparing the "matrisomes" of HSF1 WT and null tumors at day 7, we identified differentially regulated ECM modules that facilitate collagen functionalization and abundance of small leucine-rich proteoglycans; suggesting a failure of the HSF1 null stroma in efficiently remodeling the ECM. We identified dynamic changes in specific matrisome components with known roles in tumor proliferation and immune activity. These biochemical and biophysical changeswere in turn capable of modulating the cellular composition of DLBCL tumors. By day 7, the LME in HSF1 null was less supportive of lymphoma growth as measured by lower cellularity (p=4.5e-18) and proliferation of lymphoma cells by Ki67 (p=7.6e-6); and increased apoptosis by TUNEL (p=5.6e-5). In regards to CD31 positivity based on endothelial cells and vascularization, we observed no significant differences between HSF1 WT and HSF1 null tumors, however, there were significant changes in the repertoire of immune infiltrating cells. Specifically, HSF1 null LME harbored higher fractions of effector cd3+ T-cells (p=3.0e-6) and mac2+macrophages (p=9.4e-11); and a decrease in the suppressive cells cd3+foxp3+Tregs (p=6.6e-19) and cd11b+ gr1+myeloid-derived suppressor cells (p=0.0467).

Taken together, our data in DLBCL specimens and cell lines, lymphoma-fibroblasts organoids, and a murine DLBCL model, all demonstrate that HSF1 activity in DLBCL drives an ECM program that confers microarchitectural and biophysical properties to the LME that are supportive of lymphoma cell proliferation and infiltration by immune suppressive populations.