Determining cell-of-origin subtypes of diffuse large B-cell lymphoma using gene expression in formalin-fixed paraffin-embedded tissue

Abstract

The assignment of diffuse large B-cell lymphoma into cell-of-origin (COO) groups is becoming increasingly important with the emergence of novel therapies that have selective biological activity in germinal center B cell-like or activated B cell-like groups. The Lymphoma/Leukemia Molecular Profiling Project's Lymph2Cx assay is a parsimonious digital gene expression (NanoString)-based test for COO assignment in formalin-fixed paraffin-embedded tissue (FFPET). The 20-gene assay was trained using 51 FFPET biopsies; the locked assay was then validated using an independent cohort of 68 FFPET biopsies. Comparisons were made with COO assignment using the original COO model on matched frozen tissue. In the validation cohort, the assay was accurate, with only 1 case with definitive COO being incorrectly assigned, and robust, with >95% concordance of COO assignment between 2 independent laboratories. These qualities, along with the rapid turnaround time, make Lymph2Cx attractive for implementation in clinical trials and, ultimately, patient management.

Figure 1

Performance of the Lymph2Cx assay in the independent validation cohort. (A) The Lymph2Cx model is shown in the form of a gene expression heatmap (upper) with 67 patient samples from the independent validation cohort arrayed left to right in ascending order of the assay score. The 20 genes that contribute to the model are shown at the left, with the top 8 genes being overexpressed in ABC, the middle 5 genes being housekeeping genes, and the lower 7 genes being overexpressed in GCB. (Lower) The cell-of-origin assignments are shown for the assay, the gold standard method using previously published algorithms on gene expression from FT and 3 immunohistochemistry-based algorithms. The Lymph2Cx results shown are from the Molecular Characterization Laboratory (FNLCR, Frederick, MD), with 1 of the 68 cases in the independent validation cohort having failed. Results from the Centre for Lymphoid Cancer, BC Cancer Agency, Vancouver, BC Canada, are shown in supplemental Figure 2. (B) Comparison of the Lymph2Cx scores in the validation cohort from the 2 independent laboratories: the Molecular Characterization Laboratory (MoCha) (Frederick National Laboratory for Cancer Research) and the CLC, BC Cancer Agency. The dotted lines represent the thresholds between GCB, unclassified, and ABC. The 66 of 68 cases where both laboratories generated results are shown. The 3 cases that gave discordant COO assignments are shown in red. The concordance is 98%, when considering the ABC and GCB cases by the gold standard method, and 95% if the unclassified cases are included. The R² is 0.996, and the slope of the line of best fit is 1.015. Comparisons in the training and total cohorts are shown in supplemental Figure 3.

Figure 2

Patient outcomes according to COO in the independent validation cohort. (A) Progression-free survival in the COO groups as determined by the Lymph2Cx assay. (B) Overall survival in the COO groups as determined by the Lymph2Cx assay. (C) Progression-free survival in the COO groups determined by the gold standard method applying the previously described model to gene expression on FT. (D) Overall survival in the COO groups determined by the gold standard method. The P values are from log-rank tests comparing the ABC and GCB groups. The log-rank tests are 1 sided in the direction of greater hazard for ABC. RR, relative risk (with the 95% confidence interval in brackets) associated with the ABC group compared with the GCB group. The groupings in A and B are from the results at the Molecular Characterization Laboratory (Frederick National Laboratory for Cancer Research). Results from the Centre for Lymphoid Cancer, BC Cancer Agency, Vancouver, BC Canada, are shown in supplemental Figure 4.