Evaluation and validation of automated multiplexed immunofluorescence for clinical lymphoma diagnoses Free

B-cell lymphomas encompass a wide range of malignancies, with clinical features ranging from indolent to aggressive. The diagnosis of these malignancies can routinely entail the evaluation of up to 20 immunohistochemistry (IHC) markers for complete characterization. An additional challenge is that analysis of these lymphomas requires inference of co-expression patterns from single-plex IHC on serial tissue sections, leading to significant subjectivity and inaccuracy. Increasingly, molecular and cytogenetic tests must also be performed placing additional demands on tissue material. At the same time, core needle biopsies, rather than full excisional biopsies, have become standard for lymphoma diagnosis, further limiting the material available for analysis (Seviar et al. 2020).

Here, we present the development, validation, and evaluation of a 22-marker sequential immunofluorescence (seqIF, Rivest et al. 2023) panel for the diagnosis and prognostication of B-cell lymphomas that can be performed on a single FFPE section on the Lunaphore COMET automated staining and imaging platform. Each marker was robustly validated through blinded hematopathologist review and compared to clinical IHC results from the same case. Validation was performed by converting each single marker expression from fluorescence to an RGB image with nuclei represented in a color similar to hematoxylin and the marker of interest in a color similar to DAB, which is routinely used in clinical practice. These sets of single-plex pseudo-colored images were then evaluated by a hematopathologist for diagnosis and classification, as well as comparison to the original IHC stains for that case. We further evaluated the effect of tissue section and tissue block age on performance of this panel and showed excellent results on tissue blocks more than10 years since collection. The COMET system can process four tissue slides in parallel in under 16 hours (overnight run) for this panel, which is compatible with expected clinical IHC turnaround times.

We next evaluated the utility of this seqIF panel in the classification and diagnosis of indolent and aggressive B-cell lymphoma types including small lymphocytic lymphoma, mantle cell lymphoma, follicular lymphoma, large B-cell lymphoma, and other subtypes. As controls, we included benign lymph nodes, as well as biopsies exhibiting progressive transformation of germinal centers, follicular hyperplasia, and interfollicular hyperplasia, all of which can mimic malignancy and present diagnostic challenges. We constructed a tissue microarray (TMA) from these cases (n=85) to recapitulate the limited diagnostic material received from core biopsies. Again, blinded hematopathology review of pseudo single-plex images was performed, leading to a 100% concordance (marker positive/negative) between the COMET generated data and diagnostic IHC at the single marker level. Furthermore, there was no discordance between classification of a tissue as malignant vs benign, or indolent vs aggressive between the seqIF images and the clinical pathology report.

Expression patterns of markers such as BCL2 and CMYC in malignant B-cells in large B-cell lymphoma have prognostic significance; however, other cell types within the tumor microenvironment can express these proteins, which makes the tumor specific quantification challenging. Our seqIF panel allows for accurate quantification of malignant B-cell co-expression of multiple prognostic markers, including BCL2, BCL6, CMYC, TP53, and KI67, as well as classification markers CD10, BCL6, and MUM1. With the implementation of cell segmentation, this process of co-expression analysis can be fully automated, producing precise quantifications from tens of thousands of cells.

Overall, we demonstrate the utility of a seqIF panel in the classification and prognostication of B-cell lymphomas. This panel has equal performance characteristics to single-plex clinical IHC in both qualitative marker assessment, tissue classification and B-cell lymphoma classification. Additionally, as the system performs the analysis on a single FFPE tissue section that can be reused for downstream applications, it preserves precious material for cytogenetic and molecular prognostic studies. Automated image analysis with the help of machine learning classifiers to assess for potential clinical diagnostic use and independent validation on a second set of diagnostic samples is ongoing.