Unraveling the Role and Spatial Architecture of Cancer-Associated Fibroblasts in Classical Hodgkin Lymphoma

Introduction: Cancer-associated fibroblasts (CAFs) constitute an important stromal component of the tumor microenvironment (TME). They are a heterogeneous population of cells, which can modulate the immune system and have both pro-tumorigenic and anti-tumorigenic effects in a context-dependent manner. In solid tumors, the impact of CAFs in shaping the TME is well recognized, although challenges, including lack of specific CAF markers, still exist. In classical Hodgkin lymphoma (cHL), the role of CAFs has remained largely undefined. Here, we aimed to characterize distinct CAF subsets as well as their interactions with other TME cells and associate the findings with clinical characteristics and outcome of patients with previously untreated cHL.

Methods: We used a tissue microarray (TMA) containing formalin-fixed paraffin-embedded samples from 115 diagnostic cHL tumors. The TMA was stained with a 7-plex immunohistochemistry panel to characterize CAFs [platelet-derived growth factor (PDGFR) -alpha and -beta, fibroblast-activating protein (FAP), secreted protein acidic and rich in cysteine (SPARC)], macrophages (CD68), Hodgkin-Reed Sternberg (HRS) cells (CD30), and leukocytes (CD45). Image processing and quality control were performed by combining Ilastik and CellProfiler softwares, and nuclei segmented by a pretrained deep learning segmentation model. Single cell features from the images were extracted using HistoCAT software. We utilized the Phenograph clustering algorithm for cell phenotyping and a permutation test by HistoCAT for neighborhood analysis.

Results: We identified a total of 952,099 single cells, which were split into 25 distinct phenotype subsets by the Phenograph clustering spanning CAFs, macrophages, leukocytes, and HRS cells. CAFs were classified into seven distinct subsets based on the status of the different CAF markers. Median proportion of all CAFs was 28% (range 2-80%), the proportion being higher in the nodular sclerosis subtype compared to the others. In general, higher proportions of CAFs associated with favorable freedom from treatment failure (FFTF) independently of the subtype, age, and stage (P<0.01). On the contrary, a subcluster of CD45+ immune cells with strong FAP-positivity, which were characterized as macrophages, was enriched in other than nodular sclerosis subtype (P<0.001) and associated with worse FFTF independently of age, stage, and subtype (P=0.01). The neighborhood analysis allowed identification of colocalization and statistically significant interaction or avoidance between pairs of cell phenotypes. For instance, higher proportions of FAP+PDGFRalpha+ CAFs interacting with other cells predicted better FFTF (P<0.01), whereas interactions of FAP+ macrophages with other cells were associated with worse FFTF. Despite the positive impact of CAF proportions on cHL outcome, patients with CD30+ HRS cells surrounded by PDGFRbeta+ CAFs had worse FFTF (P<0.001).

Conclusions: We have characterized distinct CAF subsets in cHL and demonstrate their favorable clinical impact on cHL outcome. We also identified a novel subset of FAP-positive macrophages, the proportions and interactions of which translated to poor outcome. Our data highlight that not only the cell proportions, but the cell-cell interactions and spatial context play a crucial role. Further validation of the findings is ongoing.