Highly Heterogeneous Cellular Composition and Neighborhood Architecture within the Follicular Lymphoma Tumor Microenvironment

The variable clinical course and response to treatment of Follicular Lymphoma (FL) is strongly influenced by its tumor microenvironment (TME). Indeed, cellular interactions dictate pro- and anti-tumor processes within the TME, although these mechanisms remain incompletely understood. Multiple bulk and single-cell studies have shed light on the TME with prognostic implications. For example, a macrophage gene signature as well as a ‘T cell depleted’ milieu have been associated with inferior clinical outcomes.

Imaging Mass Cytometry (IMC) permits the multiplexed assessment of over 40 markers concomitantly, superseding the limits of conventional immunohistochemistry, thus enabling detailed single cell spatial profiling of the TME. For comprehensive analysis of tumor immune architecture within FL, we first optimized a 31-plex panel to capture tumor, immune and stromal cell subsets, as well as functional information such as memory and effector status, proliferation, activation and exhaustion. Fixed formalin paraffin embedded (FFPE) biopsies from untreated FL patients enrolled on the GALLIUM trial (NCT01332968), assessing frontline chemo-immunotherapy regimens, were subjected to IMC and 381 high-dimensional images were acquired.

After quality control, we analysed >3.5 million cells across 340 high-dimensional images. Cell phenotyping identified ~78 cell populations including lymphoma cells, CD4+ and CD8+ T cell subsets (including T follicular helper (Tfh), memory, naïve, exhausted) and a range of non-lymphoid cells e.g. myeloid and stromal cells. As expected, CD20+ B cells were expanded in FL images compared with control lymphoid samples (~56% vs. ~15% respectively p<0.05). Consistent with previous literature, we found an expansion of CD4+ T cells in FL (~20%) compared with control tissue (~12% p<0.05). We have additionally identified sets of samples enriched for CD57+ T cells. The presence of these cells has previously been associated with inferior survival in FL. Furthermore, we have observed that parent lineage abundance (e.g. CD20+ cells, CD3+ cells) identified via our phenotyping of FFPE biopsies were consistent with CIBERSORTx deconvolution analysis from patient-matched RNA-seq samples, improving confidence in our phenotypic labels.

We next analysed cellular neighborhoods, defining ten distinct FL environments. Analysis was performed i) at the level of neighborhoodcellular composition, and ii) at the level of relative proportions of identified neighborhoods within images. In terms of neighborhood cellular composition, among visually similar tissue components, we found striking variability. For example, we assessed the cellular composition of the neighborhoods making up follicles (as defined by the presence of CD21+ meshworks). In some cases, these neighborhoods were very homogenous with the cellular infiltrate consisting almost exclusively of lymphoma CD20+ cells. In other cases, we found that such neighborhoods were highly heterogenous encompassing diverse immune cell composition.

By analysing the relative abundance of different neighborhoods, we found sets of variable architectures across images. For example, the ‘lymphoma exclusive’ neighborhood dominated many images. We found that images with high abundance of this neighborhood typically also contained a scattered ‘lymphoma+, CD4/8T+, CD16+’ neighbourhood additionally exhibiting immunosuppressive Tregs cells, as well as CD57+ T cells. In contrast, a separate notable pattern of neighbourhood abundances was observed in a set of images containing organised follicular structures with detectable CD21+ meshwork, admixed with CD57+/- Tfh-like cells and macrophages. Here, this follicular structure was outlined by a thin layer of CD4+ T cells, CD57+PD1+ T cells and lymphoma cells. These images are completed by an interfollicular region containing neighborhoods comprising mixed immune subsets.

Ongoing work seeks to examine associations between such variable spatial arrangements with this clinically annotated dataset. This will provide a detailed characterization of how the spatial context and crosstalk of individual cells amid neighbourhoods within the FL TME specifically influences disease behavior, response to treatment and clinical outcome, which will likely extend and compliment those findings already described at bulk and single cell level.