The Double-Hit Gene Expression Signature Defines a Clinically and Biologically Distinct Subgroup within GCB-DLBCL

Recognizing biological heterogeneity in diffuse large B-cell lymphoma (DLBCL), significant effort has been made to define distinct molecular subgroups of prognostic importance which harbor potentially targetable biology. Reflecting this, in the recent revision of the WHO classification, DLBCL was divided into cell-of-origin molecular subtypes and a new entity was defined - high-grade B-cell lymphoma with MYC and BCL2 and/or BCL6 rearrangements (HGBL-DH/TH). ~8% of tumors with DLBCL morphology are HGBL-DH/TH and all HGBL-DH/TH with BCL2 translocations (HGBL-DH/TH-BCL2) are of the GCB molecular subtype. To explore specific biology operating in HGBL-DH/TH-BCL2, we analyzed RNAseq data from 157 de novo GCB DLBCL tumors (25 being HGBL-DH/TH-BCL2) with the aim of defining gene expression signatures that distinguish such cases from other GCB-DLBCLs.

We identified 104 genes that were most significantly differentially expressed between HGBL-DH/TH-BCL2 and other GCB-DLBCLs, defined as having a 95% confidence interval of the Importance Score that did not cross 0. A model constructed from the expression of these genes clustered 42 tumors into one group ("double-hit signature" positive - DHITsig pos), and 115 tumors into the DHITsig neg group. 22 tumors were HGBL-DH/TH-BCL2 within the DHITsig pos group compared with only 3 tumors in the DHITsig neg group.

We next assessed the clinical impact of the DHITsig within a uniformly R-CHOP treated cohort of de novo GCB-DLBCL drawn from a population-based registry, which included the discovery cases. The DHITsig pos group had significantly inferior outcomes for time to progression (TTP) and overall survival (OS) (P < 0.001 and P = 0.01, respectively) similar to ABC-DLBCL (Figs A, B). Notably, the non-HGBL-DH/TH-BCL2 cases sharing the DHITsig showed the same poor prognosis as the HGBL-DH/TH-BCL2 cases. A multivariate Cox model of TTP revealed that DHITsig remained prognostic, independent of IPI and MYC/BCL2 dual protein expression (HR = 3.1 [1.5 - 6.4], P = 0.002). We then applied this gene expression model to GCB-DLBCL in an independent dataset (n = 262 GCB-DLBCLs; Reddy et al,Cell 2017). Validating the prognostic significance, the DHITsig pos group had significantly inferior OS compared with other GCB-DLBCLs (P < 0.001) similar to ABC-DLBCL (Fig C).

We then sought to determine whether differentially expressed genes, according to DHITsig, could inform on the biology of the DHITsig pos group. Gene set enrichment analysis (GSEA) strongly suggested a germinal centre dark-zone (DZ) cell-of-origin for the DHITsig pos tumors with significant enrichment of DZ and light-zone (LZ) gene signatures (Victora et al, Blood 2012) in DHITsig pos and neg tumors, respectively (FDR = 0.002 and < 0.001). Furthermore, the DHITsig pos group had up-regulation of pathways related to mitochondrial metabolism and RNA synthesis (both FDR < 0.001).

We separately identified mutations associated with DHITsig pos cases within GCB-DLBCL. In addition to the expected enrichment of MYC and BCL2 mutations, chromatin modifiers EZH2 and CREBBP, as well as RFX7 and DDX3X (mutated in Burkitt lymphoma), were more frequently mutated in DHITsig pos tumors. In contrast, mutations of TNFAIP3, MYD88 and IRF4, more typical of ABC-DLBCLs, were more prevalent in DHITsig neg tumors.

To enable application to FFPE biopsies, the DLBCL90 NanoString assay was developed by translating the DHIT gene expression signature into a 30-gene module that was then added to the Lymph2Cx assay. The DLBCL90 assay was applied to 171 DLBCL tumors (including 156 from the discovery cohort), yielding 26% DHITsig pos, 64% DHITsig neg, and 10% unclassified, with a frank misclassification rate of 3% against the RNAseq comparator. The prognostic significance of the groups was maintained (Fig D). Importantly, the DHITsig neg group had a disease specific survival of 91% at 5 years. To validate the association between the DHITsig and HGBL-DH/TH-BCL2 tumors, the DLBCL90 assay was applied to 113 transformed follicular lymphoma tumors. Within the DHITsig pos group, 19/34 tumors were HGBL-DH/TH-BCL2 compared with 0/58 in the DHITsig neg group.

In conclusion, we have identified a clinically and biologically distinct subgroup of GCB-DLBCL tumors that are defined by the HGBL-DH/TH-BCL2 gene signature. The translation to an assay applicable to FFPE allows exploration of its utility to guide patient management within the context of clinical trials.

Figure.

View largeDownload slide

Figure.

View largeDownload slide

Close modal