A targeted gene signature stratifying mediastinal gray zone lymphoma into classical HL-like or PMBL-like subtypes

Mediastinal non-Hodgkin lymphomas with overlapping features between primary mediastinal B-
cell lymphoma (PMBL) and classical Hodgkin lymphoma (CHL) – defined as mediastinal gray
zone lymphoma (MGZL) – represent a unique, diagnosis-challenging entity. They typically exhibit
discordant morpho-phenotypic features between CHL and PMBL, a high rate of diagnostic
reclassification and poor therapeutic outcomes. There is an urgent need for practical tools exploiting
molecular traits of MGZL to facilitate their diagnostic stratification and guide optimal treatments.
By combining CIBERSORTx deconvolution with a nonnegative matrix factorization (NMF)-based
approach, we selected tumor- and microenvironment (TME)-related genes from bulk gene
expression data of CHL and PMBL. A panel of 2,913 genes with a striking discriminative capacity
between the two lymphoma subtypes underwent further feature selection toward a final signature of
168 genes retaining high discriminative power. NanoString Technology was then used to assess the
discriminative capacity of the signature on real-life cases, and its ability to molecularly place
MGZL within either CHL or PMBL subgroups. Here, we describe the development of a targeted
gene signature and a proof-of-concept of its usefulness for categorizing MGZL in support of current
morpho-phenotypic classification. If properly validated on larger cohorts, our approach may drive
the development of a molecular assay transferable into the routine clinical practice.