Imaging Mass Cytometry Reveals a Highly Heterogenous Cell Makeup of the Follicular Lymphoma Tumour Microenvironment

Abstract

The highly variable clinical course and response to treatment of follicular lymphoma (FL) is strongly influenced by the tumor microenvironment (TME). While both pro- and anti-tumor effects depend on cell-cell interactions within the TME, these mechanisms remain incompletely understood. Bulk (Dave et al. NEJM 2004) and single-cell RNA-seq studies have identified signatures associated with prognosis, e.g., ‘T cell depleted’ microenvironments that are linked with inferior clinical outcomes (Han et al. Blood Cancer Discov. 2022). While such research has shed light on the TME, single-cell studies often lack spatial context and conventional immunohistochemistry is limited by a low number of parameters. Imaging mass cytometry (IMC) allows multiplexed analysis of up to 40 markers in tissue simultaneously, offering a unique opportunity to study the spatial features of single cells within the TME, informing both disease biology and predictors of clinical outcome. To our knowledge, IMC has more commonly been applied to solid malignancies. FL presents unique challenges in this remit, such as its morphologic variability and phenotyping cells within highly structured, dense lymph-node tissue composed of inherent tight-knit microenvironments. To analyse FL tissue by IMC, we first optimized a 31-plex panel to identify key immune and non-immune lineages including tumor cells, T and NK cells, macrophages, fibroblasts, dendritic, endothelial and reticular cells. Additionally, we wished to identify functional states such as memory, effector or regulatory phenotypes in addition to other parameters including activation, exhaustion and proliferation. Formalin-fixed paraffin embedded (FFPE) biopsies from 100 untreated FL patients enrolled on the Gallium trial (NCT01332968) of frontline chemo-immunotherapy, were subjected to IMC. In total, 381 highly multiplexed images were acquired; 13 were excluded due to suboptimal image quality. Following image pre-processing and deep-learning based cell segmentation, we analysed >3.5 million cells across 368 high-dimensional images (mean=9659 cells, SD=4600). To effectively phenotype large numbers of cells, we constructed a workflow incorporating ‘ASTIR’ (Geuenich et al. Cell Syst. 2021) - a publicly available marker-based algorithm - and cell clustering to identify population subsets. We identified 95 cell populations including Lymphoma subsets (variable CD10, CD11c, CD80 expression), CD4+ and CD8+ T cell subsets (memory, naïve, exhausted etc) and a range of non-lymphoid cells including myeloid and stromal cells. Of particular interest, we found a CD4+ CD3+ PD1+ ICOS+ population which are likely T follicular helper cells (Tfh). On average, Tfh-like cells constitute 1.9% of cells per image (SD=3.1%), although we a found considerable range in their presence. Given previous associations with Tfh cells and FL prognosis we are investigating this variability in further detail. In addition, we have identified ~30 samples enriched for CD57+ T cells, a phenotype suggestive of functional senescence. Amongst substantial heterogeneity of identified cell phenotypes (Fig A), we have identified variable sample sets characterised by enrichment or depletion of i) Lymphoma vs. T-cells (Fig B) and ii) Macrophage vs. T-cell subsets, consistent with previous single cell studies. Here, we report the use of IMC to examine the TME of untreated FL, observing marked heterogeneity of cell phenotypes across FL. Future work seeks to examine prognostic associations between the frequency of cell subsets and their spatial interactions, with this clinically annotated dataset derived from the front-line GALLIUM study. This work will provide a detailed characterization of how the spatial context of FL TME influences disease behaviour, response to treatment and survival. Given the high-throughput and high-plex nature of this study, ultimately, we aim to distil this multitude of information into condensed biomarker(s), to provide potential insights into enhancing diagnostic and/or prognostic capabilities. Therefore, we aim to derive meaningful clinical correlates associated to outcome that can be ultimately translated into practice.