Limiting the impact of protein leakage in single-cell proteomics

Abstract

Limiting artifacts during sample preparation can significantly increase data quality in single-cell proteomics experiments. Towards this goal, we characterize the impact of protein leakage by analyzing thousands of primary single cells from mouse trachea. The cells were prepared either fresh immediately after dissociation or first cryopreserved and prepared at a later date. We directly identify permeabilized cells by imaging a cell permeable dye and use the data to define a signature for protein leakage. This signature is similar across diverse cell types and reflects increased leakage propensities for cytosolic and nuclear proteins compared to membrane and mitochondrial proteins. A classifier based on the signature allowed for the accurate identification of permeabilized cells across cell types and species. The classifier is integrated into QuantQC ( scp.slavovlab.net/QuantQC ) to support its application to diverse samples and workflows.

Fig. 1. Quantifying protein leakage artifacts.

a Images of cells in the CellenONE nozzle taken with brightfield and with the green fluorescent channel. The two cells shown are not clearly distinguishable from the brightfield but one cell is permeable and thus positive for the Sytox green fluorescent dye. b The UMAP dimensionality reduction shows co-clustering of cells from fresh and frozen batches that were recorded as permeable via Sytox green. c Percentages of permeable cells for the two sample handling conditions are shown in the dot plot. d Differences in protein fold changes between permeable and intact cells for all proteins. e Differences in protein fold changes between permeable and intact cells are categorized by subcellular compartment.

Fig. 2. Examining and predicting cell permeability in human cell lines.

a Log2 average protein fold changes between permeable and intact Club cells and Fibroblasts are plotted against each other and show Pearson correlation of 0.66. The heatmap summarizes correlations between fold changes for all cell types is shown. b ROC curve for a classifier trained on permeability status of single cells using protein abundance profiles of the top 75 most significantly leaking proteins. Results for the model trained and tested on the same cell type are in red and trained and tested on different cell types are in black. c PCA projection of single cells from Leduc et al., 2024. Cells are colored by cell type or their permeability score from the classifier. The cells towards the center of the two dimensional space of the first two principal components were enriched for high permeability scores. d When comparing the fold changes between cells with probability over 0.2 and under 0.2 to be permeable versus intact fold changes from the primary mouse tracheal cells, the fold changes strongly agree, Pearson correlation 0.50.