Purpose: The critical role of the immune system in cancer development and progression has become increasingly evident in the last years with the advent of immunomodulators as cancer therapeutics in clinic. Deep phenotyping will allow the development of individually tailored drugs so that patients will benefit from higher efficacy and diminished adverse effects. Recently, a variety of cytometry technologies have arisen that allow the measurement of previously unprecedented number of biomarkers. Image cytometry systems provide additional data like cell morphology, subcellular biomarker localisation and increased sample stability. However, the throughput of image cytometers still lags behind that of flow cytometers and enrichment of rare cell populations by cell sorting is precluded in most cases, making analysis of rare cell events like circulating tumour cells or rare immune cell populations extremely cumbersome, if not impossible. Here we present and evaluate a method called “virtual sorting” that largely alleviates this problem and allows rare event analysis even with technologies where cells are mounted on slides, rendering conventional cell sorting impossible.
Methods: Peripheral blood mononuclear cells (PBMCs) were isolated from blood of healthy volunteers and stimulated in vitro either with phorbol 12-myristate 13-acetate (PMA)/ionomycin or with CD3/IL-12 to mimic different types of immune cell activation. Unstimulated PBMCs were applied to microfluidic slides together with low numbers of the stimulated cells to achieve spiking of “normal” blood with rare activated cells, followed by paraformaldehyde fixation. The cells were then stained with a multicolour set of 5 fluorophore-conjugated biomarkers and the entire slides were scanned using chipcytometry. After calculation of fluorescence intensities, cells of interest (activated cells) were gated on a 2D plot and all positions on the slides that didn’t contain cells of interest were excluded from further analysis. Deep phenotyping was iteratively performed with another 25 biomarkers, scanning only the remaining positions. Coefficient of variation (CV) was then calculated on the replicates.
Results: Spiked cells that constituted 0.1% of the entire cell population were successfully detected during the first scan. Application of the virtual sorting algorithm diminished scan time for the remaining 20 biomarkers of the oncoimmunology panel by 80%. Among others, the activated cells were analysed for a variety of checkpoint inhibitors and activation biomarkers (PD-1, PD-L1, LAG-3, OX-40, OX40L, CTLA-4). At the same time, the marker panel allowed the identification and additional phenotyping of relevant, but relatively rare cell populations like dendritic cells and regulatory T cells. The CVs for the smallest cell populations were below 20%.
Conclusion: We demonstrate the feasibility, applicability and high precision of the so-called virtual sorting for detection and characterisation of rare cell populations in samples mounted to slides and deeply phenotyped by image cytometry with 25 biomarkers including cancer-related checkpoint inhibitors. This method will prove to be useful for analysis of patient samples for very rare events, thus supporting diagnosis, patient stratification and drug development for cancer therapy.
Anja Mirenska– Lab Manager, ZELLKRAFTWERK