Determining Cell-Of-Origin Subtypes In Diffuse Large B-Cell Lymphoma Using Gene Expression Profiling On Formalin-Fixed Paraffin-Embedded Tissue – An L.L.M.P.P. Project

Abstract The diffuse large B-cell lymphoma (DLBCL) cell-of-origin (COO) distinction into germinal center B cell (GCB) and activated B cell (ABC) subtypes, as molecularly described by our group, has profound biological, prognostic, and potential therapeutic implications. New therapeutic agents with selective activity in ABC and GCB DLBCL are under development. An accurate diagnostic assay is urgently needed to qualify patients for clinical trials using targeted agents and as a predictive biomarker. Although the subtypes were originally defined using gene expression profiling on snap-frozen tissues (frozen-GEP), it has become common practice to use less precise but relatively inexpensive and broadly applicable immunohistochemical (IHC) methods in formalin-fixed paraffin-embedded tissue (FFPET). We sought to create a robust, highly accurate molecular assay for COO distinction using new GEP techniques applicable to FFPET. Studies were performed on centrally reviewed DLBCL FFPET biopsies using cases that had “gold standard” COO assigned by frozen-GEP using Affymetrix U133 plus 2.0 microarrays. The training cohort consisted of 51 cases comprising 20 GCB, 19 ABC and 12 Unclassifiable (U) cases. An independent validation cohort, consisting of 68 cases (28 GCB, 30 ABC, 10 U) drawn from the validation cohort of Lenz et al (NEJM 2008) had the typical proportions of COO subtypes seen in DLBCL populations. Nucleic acids were extracted from 10um FFPET scrolls. Digital gene expression was performed on 200ng of RNA using NanoString technology (Seattle, WA). All FFPET GEP studies were performed in parallel at two independent sites (BC Cancer Agency, Vancouver, and NCI, Frederick, MD) using different FFPET scrolls to determine inter-site concordance and assess the robustness and portability of the assay. To assign COO by IHC, tissue microarrays were made using 0.6mm duplicate cores from 60/68 validation cohort cases, and stained for CD10, BCL6, MUM1, FOXP1, GCET1 and LMO2. Two hematopathologists independently assessed the proportion of tumor cells stained, with consensus on discordant cases reached with a third hematopathologist. For the validation studies, those producing and analyzing the GEP and IHC data were blinded to the “gold standard” COO. All 119 FFPET biopsies yielded sufficient RNA. A pilot study using the training cohort identified 20 genes (15 genes of interest and 5 house keeping genes) whose expression, measured using NanoString, would allow accurate replication of the COO assignment model of Lenz et al (NEJM 2008). NanoString was then used to quantitate these 20 genes in the training cohort, allowing the COO model to be optimized. Despite the age of the FFPET blocks (6-32 years old), 95% (49/51) of the training samples gave gene expression data of sufficient quality. The model, including coefficients, thresholds and QC parameters was then “locked” and applied to the independent validation cohort. Ninety-nine percent (67/68) of the samples from the validation cohort (5-12 years old) provided gene expression of adequate quality. Three cases did not give interpretable IHC results. When considering the “gold standard” ABC and GCB cases, the COO assignments by the NanoString assay at the NCI site were 93% concordant, with 5% labeled U and 1 ABC misclassified as GCB (see table). This 2% rate of misclassification of ABC and GCB cases compares favorably with the 9%, 6% and 17% rates for the interpretable cases from the Hans, Tally and Choi algorithms, respectively. Furthermore, the 98% concordance of COO assignment (95% if “gold standard” U cases are also included) between the NCI and BC Cancer Agency sites indicates that, in contrast to the IHC algorithms, the assay is reproducible.TableNanoString GEP assay - NCIHans algorithmTally algorithmChoi algorithmGCBUABCGCBNon-GCBGCBABCGCBABCFrozen GEPGCB2800210183192U721552864ABC1325422026620 In summary, 119 well-characterized DLBCL cases from the LLMPP, previously subtyped by our published disease-defining algorithm using frozen-GEP, were used to develop a highly accurate and robust NanoString 20 gene assay, applicable to RNA from FFPET that is routinely obtained for diagnosis. This new assay shows excellent performance in archival FFPET, and the rapid turn-around time (<36 hours from FFPET block to result) will allow prospective implementation in future therapeutic trials and, ultimately, clinical practice. Disclosures: No relevant conflicts of interest to declare.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.