Transcriptomics in Multiple Myeloma Demonstrates an Association between Survival and Expression of T Cell Co-Signaling Ligands in Bone Marrow Derived Myeloma Plasma Cells

Introduction

The introduction of novel drugs has resulted in better outcome in Multiple Myeloma (MM). Disease progression, however, remains a major problem in the majority of patients. Recently, checkpoint inhibitor therapies have been applied, but T cell-mediated control is still poorly understood in MM. Parameters that control local anti-tumour immunity include, among others: presence of immune effector and suppressor cells; immune and metabolic checkpoints; and antigen presentation. These parameters are defined by unique immune markers, and recent technical advances have boosted typing of tumours according to these markers. We set out to evaluate whether overall survival in a large cohort of MM patients is associated with different categories of T cell evasion using transcriptomic analysis.

Methods

In the current study, we have analyzed gene expression data from 8 large MM trials, with a total number of 1654 patients. The HOVON65/GMMG-HD4, UAMS-TT2, MRC-IX non-intensive, and APEX trials represented a discovery set (1045 patients), whereas the HOVON87/NMSG18, MRC-IX intensive, UAMS-TT3, and UAMS-TT6 trials represented a validation set (609 patients). We defined gene sets for 7 categories of T cell evasion, including: presence of immune suppressor cells, immune checkpoints (and their ligands), metabolic checkpoints, stromal cells and their products, compromised antigen presentation, tumor cell death, and occurrence of oncogenic pathways. Gene expression analysis from pre-treatment bone marrow samples, purified for plasma cells followed 2 stages: first, we obtained ridge regression Cox models for overall survival (OS) for sets of genes that represent T cell evasion in the discovery set which were applied to the validation set; second, we tested each individual gene per gene set versus OS in Cox regression analysis. Bonferroni-Holm multiple testing adjustments were applied. For multivariate analyses age and ISS were included, which were available for all sets, except the TT6 trial.

Results

Gene models were trained for each gene set in the discovery set. Six models (all except antigen presentation) reached a significant association when applied to the validation set and were used for downstream analyses. The adjusted p values in the multivariate models are: immune checkpoints (and ligands) p= 6.5x10^-6, cell death p = 1.9x10^-5, oncogenic signaling p = 3.3x10^-5, stromal cells and their products p = 3.8x10^-5, metabolic checkpoints p = 0. 011, and immune suppressor cells p = 0.020. Analysing individual genes resulted in 34 genes that showed a significant association with OS in the discovery cohort and a remaining 7 genes in the validation cohort. These genes predominantly belonged to the family of immune checkpoint ligands. These genes were VISTA (HR 0.71; CI 0.62-0.82), CD40 (HR 0.75; CI 0.63-0.88), HVEM (HR 0.81; CI 0.71-0.92) and PDL1 (HR 0.78; CI 0.69-0.92). Notably, in a multivariate model, VISTA came out as an independent prognostic marker with a hazard ratio (HR): 0.72 [CI: 0.61-0.83].

Conclusion

Immune transcriptomic analysis in a large cohort of patients reveals that OS in MM patients is related to expression of T cell co-signaling ligands, such as VISTA, CD40, HVEM, and PDL1. The expression of the checkpoint molecule VISTA remained independently associated with OS. In addition, another 5 categories of T cell evasion were shown to be significantly related to OS in both a discovery and validation set of patients. Collectively, these results stress the important role of T cell-mediated control of tumor cells in MM patients, and urge for a more detailed analysis in subgroups and individual patients in order to select the most effective (combination) therapies.