Spatially Resolved Single-Cell Transcriptomics of T Cells in Follicular Lymphoma Reveals Distinct Immune Ecosystems

Background: Recent studies have shown that the extent of T-cell infiltration strongly correlates with a low frequency of early relapse in patients with follicular lymphoma (FL). This suggests that a deeper understanding of T-cell heterogeneity and biology could lead to the development of novel FL biomarkers. Our previous single-cell RNA sequencing (scRNA-seq) analysis (Abe et al. ASH 2023) revealed significant heterogeneity in T cells, particularly in follicular T cell components in FL. This study aimed to elucidate the spatial characteristics of T cells in FL, focusing on the follicular T cell subsets relevant to FL biology.

Methods: We performed a spatially resolved single-cell transcriptomic analysis of six formalin-fixed paraffin-embedded samples of newly diagnosed FL using the Xenium In Situ system. We used a custom gene panel comprising 289 genes curated from the differentially expressed genes detected in the scRNA-seq analysis of various immune and stromal cells in FL. After DAPI-based cell segmentation, single-cell transcriptome data were integrated and subjected to unsupervised clustering analysis to reproduce the results of the scRNA-seq analysis. The spatially smoothed chemokine gene density was calculated using the kernel-smoothing method. Multiplex digital spatial profiling (MDSP) was performed on 242 FL samples from three cohorts to validate the Xenium analysis findings.

Results: We analyzed the transcriptome data from 516,657 cells detected using Xenium. Unsupervised clustering analysis detected T cell components, in addition to malignant (B_malig) and non-malignant B cells, monocytes/dendritic cells, plasma cells, and stromal cells, including endothelial cells. Sub-clustering analysis detected T cell subsets annotated as lymphoma follicular regulatory (_LT_FR) and CD4 (_LT_FC4) or CD8 (_LT_FC8) lymphoma follicular cytotoxic T cells in our previous scRNA-seq analysis, as well as conventional T cell subsets, including follicular helper (T_FH), regulatory (T_reg), and cytotoxic (T_ct) T cells. The transcriptional features of these subsets were consistent with their scRNA-seq profiles. Remarkably, the _LT_FR and _LT_FC cells were preferentially observed within and at the edges of neoplastic follicles (NFs), respectively. The minimum distance from _LT_FR cells to NFs or B_malig cells was shorter than that from T_reg cells.Similarly, compared to T_ct cells, _LT_FC cells were localized closer to NFs or B_malig cells. Additionally, compared to _LT_FC cells, _LT_FR cells tended to be closer to NFs or B_malig cells. The spatially smoothed densities of CXCL13 and CCL19 were relatively high in the NFs and T-cell zones, respectively, as reported previously. We also confirmed that the CXCL13 and CCL19 densities were the highest at the coordinate points of T_FH and naïve T cells, respectively. Among the non-T_FH T cell populations, the highest CXCL13 and lowest CCL19 densities were detected at the coordinates of _LT_FR cells, whereas the joint density of CXCL13 and CCL19, which estimated their overlap, was the highest at _LT_FC cell positions, supporting the distribution patterns of these cell subsets. The spatial and distance relationships depicted using the MDSP echoed those suggested by the Xenium analysis. Correlation analysis of cell localization identified remarkably high correlations between _LT_FR and T_FH cells and between _LT_FC4 and _LT_FC8 cells. Cellular neighborhood analysis revealed that _LT_FR and T_FH cells formed distinct cellular neighborhoods, suggesting that _LT_FR cells specifically suppressed T_FH cells activity. Additionally, the _LT_FC4 and _LT_FC8 cells formed neighborhoods, consistent with their biased distribution at the edge regions of the NFs. Literature-based upstream regulator and Xenium spatial gene density analyses suggested that interleukin-21 (IL-21) was the most relevant inducer of _LT_FR, _LT_FC4, and _LT_FC8 cells. Consistently, the phenotypes of these cells were reproduced in cell culture in the presence of IL-21, suggesting that the IL-21-predominant FL microenvironment, enriched with activated T_FH cells, induced these non-T_FH cell subsets and elicited anti-tumorigenic immunity.

Conclusions: Our spatially resolved single-cell analysis of T cells revealed the presence and characteristic distribution patterns of distinct follicular T cell subsets in FL. This approach highlights the self-regulatory immune ecosystems that may underlie FL biology and clinical behavior.