Tumour-Intrinsic Features Shape T-Cell Differentiation through Myeloma Disease Evolution

Background. Virtually every MM patient transitions through the asymptomatic precursor conditions MGUS and SMM prior to developing MM. This provides a window for intervention to prevent progression to frank malignancy, and also exciting opportunities to investigate mechanisms underlying such progression. In contrast to solid tumours, relatively little is known about how the immune system is shaped by, and in turn shapes, MM through progression from precursor conditions to frank malignancy.

Aim. To define the contribution of tumour features to T cell differentiation during disease evolution from precursor disease to MM.

Methods. We generated the “pre-myeloma atlas”, the largest single-cell RNA sequencing (scRNA-seq) dataset of precursor disease and MM to date. It encompassed >1,000,000 cells from 295 newly-generated and published scRNA-seq samples from MM (n=54 patients), precursor conditions (n=78), and non-cancer controls (n=102). We enhanced the atlas through including single-cell T cell receptor (TCR) sequencing (42% samples), bone marrow and peripheral blood sampling, and modelling clinical data including age and tumour burden. Malignant clones (tumour cells) were identified by clonal immunoglobulin expression and scored with pan-cancer transcriptional pathways. Viral-reactive TCRs were acquired from VDJdb and neoantigen-reactive TCRs predicted with NetMHCpan4.1 and TEINet.

Results. Within the pre-myeloma atlas, we identified CD4+ and CD8+ T cell subsets across the spectrum of differentiation from naïve (SELL+CCR7+) to terminal memory (GZMB+) phenotypes. Notably, we identified exhausted T cells in only a single MM patient (1 of 54), suggesting T cell exhaustion is not a pervasive feature of MM. Comparing controls with precursor and symptomatic patients, we found T cell composition was comparable through disease evolution. However, the abundance of T cell subsets was altered: in MM, earlier differentiated subsets such as naïve CD8+ were depleted relative to controls (P<0.001) whereas terminal memory subsets were enriched (P<0.003). Principal component analysis of T cell composition captured this shift from early to terminal subsets, which we termed T cell skewing. T cell skewing rose with disease severity and was enhanced in SMM and MM independent of patient age (SMM P=0.004; MM P<0.001). Longitudinal sampling of precursor patients who subsequently progressed revealed T cell skewing occurs longitudinally at progression. Together, this suggests exaggerated T cell memory differentiation occurs in MM.

TCR repertoire clonality (indicative of T cell proliferation in response to antigen) was higher in MM relative to controls (P=0.06) and corrected with T cell skewing (R=0.71, P<0.001). Additionally, the frequency of TCR clones possessing similar antigen-recognition domains (indicative of shared antigen-specificity) rose between SMM and MM (P<0.002) and correlated with T cell skewing (R=0.4, P=0.02). Importantly, in silico analyses of T cell specificity suggested (1) TCR clones annotated as viral-reactive were not associated with T cell skewing, and (2) TCR clones predicted to bind autologous tumour neoantigens occupied terminal memory states. Together, these results suggest antigen, possibly tumour in origin, drive T cell differentiation in MM.

Finally, to interrogate mechanisms of tumour-immune cross-talk, we compared T cell and tumour features. We observed a positive correlation between T cell skewing with paraprotein (R=0.43, P=0.04), that serves as a proxy for tumour burden. Additionally, effector T cells (IFNG+TNF+) were associated with tumour marrow infiltration (aspirate % CD138+; R=055, P<0.001). When analysing the transcriptional activity of tumour cells, we found effector T cells also correlated with cell death-associated pathways (R=0.66, P<0.001). We noted sharing of TCRs between these two T cell subsets, suggesting a differentiation process between these subsets possibly regulated by tumour-intrinsic processes and tumour burden.

Conclusion. Our results show that, unlike solid cancers, MM is not associated with T-cell exhaustion and instead defined by a pattern of T-cell differentiation resembling antigen-driven terminal memory differentiation. T cell associations with tumour burden and antigen could suggest anti-tumour immunity drives a novel form of cancer-associated T-cell differentiation in MM.