Neutrophilic Extracellular Traps (NETs); A Subset of Smudge Cells Identifiable by Peripheral Smear Autoanalyzers in the Rising Era of Artificial Intelligence. (28th October 2020)

• Record Type: Journal Article
• Authors: Barouqa, Mohammad; Ikemura, Kenji; Billett, Henny; Kushnir, Margarita; Fedorov, Kateryna; Yin, David; Gil, Morayma Reyes
• Is Part Of: American journal of clinical pathology. Volume 154(2020)Supplement 1
• Page Start: S10
• Page End: S11
• Publication Date: 2020-10-28
• Subjects: Diagnosis, Laboratory -- Periodicals; Pathology -- Periodicals; 616.07
• DOI: 10.1093/ajcp/aqaa137.018
• Languages: English
• ISSNs: 0002-9173

Abstract

Abstract: Background: Digitized microscopy such as CellaVision® technology has revolutionized the laboratory. Smudge cells, also called basket cells, are usually seen in lymphoproliferative disorders representing remnants from degenerated lymphocytes (DLs). CellaVision® classifies DLs and web-like remnants as smudge cells. The morphology of the web-like remnants is compatible with Neutrophil Extracellular Traps (NETs) where extracellular decondensed DNA chromatin network is formed as one of several neutrophilic reactions to stress. Currently, we lack clinical tests that reliably identify and quantify NETs. Aims: To develop an in-vitro model for NETs formation in blood, create a library of their morphological changes at different maturation stages; correlate their presence to infections in absence of leukocytosis and develop an artificial intelligence platform (AI-Heme-1) for their detection. Methods: A library was built to develop AI-Heme-1 where NETs were induced with classic triggers (phorbol-myristate-acetate, lipopolysaccharide and ionomycin) in EDTA whole blood from normal subjects. Smears were prepared at 30 minutes intervals for 24 hours to identify NETs by Immunofluorescence and immunohistochemistry. WBC differentials were performed by CellaVision® to capture different stages of NETs. AI-Heme-1 was modified from Python online convolutional neural network. For the clinical correlation, smears with >20% smudge cells were classified morphologically as NETs vs. DLs compared to a control group, < 5% smudge cells. We used morphologic characteristics, immunohistochemistry, immunofluorescence and flow cytometry to differentiate NETs from DLs. Medical chart review performed by blinded investigators, included patient demographics, CBC and presence of microbial infection occurring < 1 week of sample collection. Statistical analyses included two sided t -test and chi square. Results: The classical triggers for Netosis showed consistent morphological changes following a canonical order: vacuolation, nuclear decondensation, degranulation and chromatin ejection. These cell remnants were positive for citrullinated histones, myeloperoxidase, leukocyte alkaline phosphatase and neutrophil elastase by immunofluorescence. On Wright Giemsa stain, web-like remnants resembling NETs stained for SytoxGreen. On flow cytometry, NETs were large with extracellular DNA and MPO. For the clinical study group of >20% smudge cells, 88 were morphologically designated as NETs, 8 as DL vs. 59 as control group. A random sampling from >20% smudge cells showed cases with NET subclassification stained strongly with myeloperoxidase, neutrophil elastase and SytoxGreen while DLs were negative. Comparing patients with > 20% smudge and NET sub-classification to < 5% smudge cells, the formers had higher incidence of bacterial and viral infections (p=0.009/0.005 and p=0.008/0.007). Conclusions: Our study was able to identify NETs on peripheral smears performed by a routine Hematology Autoanalyzer using a reliable set of morphologic characteristics, immunohistochemical stains and flow cytometry. It supports data that associate NETs with infections in the absence of leukocytosis. AI-Heme-1 was able to identify NETs on blood smears. This approach can provide a rapid, early and accurate tool to screen patients with infections.