High-Throughput Mass Cytometry Staining for Immunophenotyping Clinical Samples

Abstract

As mass cytometry (MC) is implemented in clinical settings, the need for robust, validated protocols that reduce batch effects between samples becomes increasingly important. Here, we present a streamlined MC workflow for high-throughput staining that generates reproducible data for up to 80 samples in a single experiment by combining reference sample spike-in and palladium-based mass-tag cell barcoding. Although labor intensive, this workflow decreases experimental variables and thus reduces technical error and mitigates batch effects.

Conflict of interest statement

K.K. is currently an employee of bluebird bio. F.S.H. serves as a consultant to Genentech, Bristol-Myers Squibb, Merck, Novartis, Amgen, Sanofi, Bayer, Pfizer, EMD Serono, Verastem, Aduro, Celldex, and Incyte.

© 2020 The Authors.

Figures

Graphical abstract

Figure 1

Single Antibody Titration Data Generated by Mass Cytometry (A) Ex vivo samples stained with 89Y_CD45 and stimulated samples stained with 141Pr_CD45 are separated in downstream manual gating analysis by plotting the two channels on a biaxial plot. (B) CD69_144Nd single antibody titration data plotted as percent of total viable singlets.

Figure 2

Using Surface Antibody Master Mix to Analyze Mass Cytometry Titration Data CD11b titration data is used as an example for analyzing a marker with differential expression on cell types. This assists in placing the manually set gate on the viable cells. CD11b expression is expected to be positive on monocytes, with lower to no expression on B, T, and NKT cells. Increased background is observed as the antibody increases to 1:50 dilution in these cell types.

Figure 3

Comparing Storage Conditions of Surface Master Mix Metal-conjugated antibody surface master mixes were prepared and stored at -80°C or 4°C for 16 h and healthy donor PBMCs were stained. Clones and metals are the same as used in the adaptive and innate panels, as listed in Tables 3 and 4.

Figure 4

Plate Layout for Experimental Samples, Reference Samples, and CD45-Barcoded Reference Sample Spike-in (A) Experimental samples (yellow wells) and reference samples (blue) are plated. Wells C10 and F10 contain the ex vivo reference sample, and wells C11 and F11 contain the stimulated reference sample. Keeping the two stimulated conditions separate guides downstream analysis (Figure 8). (B) 1:1 mixed reference sample is “spike-in” to experimental sample (green wells). Each well now contains 2 million cells.

Figure 5

20-Plex Palladium Isotope Barcode Strategy A Barcode ID # (BC #) is assigned to each sample. Each BC # is positive for 3 of the six different Palladium isotopes (102, 104, 105, 106, 108, 110 Pd). In this schema, grayed out boxes indicate that this BC # is positive for that isotope. BC 1–18 are experimental samples with reference sample spike-in, and BC 19 and 20 are reserved for reference sample controls. Downstream gating of CD45 barcoding is shown on the right to indicate what samples are in which BC #.

Figure 6

Comparison of Injector Type on Signal Variability as Accessed by CD45_89Y Coefficient of Variable Identical samples were acquired in replicate 100,000 events each for two different acquisition modes, Wide Bore (WB) and High Throughput (HT) injectors. The WB injector significantly decreased the variability of the CD45_89Y signal. p-value=0.0001, Welch Two Sample T-test.

Figure 7

Tracking Increased Signal Drift Over Time during Long Sample Acquisitions (A) 89Y_CD45 positive reference cells contribute to an increase in background signal when resuspended over the course of an acquisition of at least 2 h. (B) Reference cells start as 141Pr_CD45 negative but contribute to increased background signal over time. (C) 89Y_CD45 negative cells are shown to be 141Pr_CD45 positive cells. These analyses were performed on “live, intact, single cells”, thus the increase in background signal in the 141Pr channel cannot be attributed to cell doublets.

Figure 8

Clean-up Gating with CD45 Barcoded Reference Sample Spike-in A representative single fcs file is cleaned-up to identify viable single cells with reference sample and experimental sample separated by manually gating on the CD45 channels, 89Y and 141Pr.

Figure 9

Tracking Cell Loss during Staining Protocol Cell count changes over the course from post-thawing and plating, to how many events acquired, and after clean-up gating are shown for three independent healthy donor PBMC samples. Three healthy donor PBMC samples were stained for MC using our full protocol. We observed ≈ 75% cell loss (mean values: pre-stain = 2 million, events acquired = 136k, events per sample after clean-up gating (“viable, single cells”) = 77k).

Figure 10

Independent Pd Barcode Ex Vivo and Stimulated Reference Samples Guides Gating Ex vivo and Stimulated reference sample are run as independent samples of the barcoding strategy. (A) Samples were analyzed using tSNE algorithm (viSNE maps created using cytobank.com) and show changes of major cell populations. Clusters indicate cell subsets and the major cell lineages are overlaid with colored dot plots. Each dot represents one cell. (B) Manual gating (Flowjo) of ex vivo and stimulated samples guides where to place positive threshold gate for dynamic markers that are expected to increase with stimulation.

Figure 11

Reference Sample Spike-In and Palladium Sample Barcoding Generates Reproducible Results Stimulated and ex vivo healthy donor PBMC samples were stained with either the adaptive or innate MC panels. Two sets of palladium sample barcodes were used for each panel. Each barcode set included one independent sample of stimulated and ex vivo and the remaining 18 barcodes were replicate samples of 1:1 mixed stimulated and ex vivo samples. (A) Manual gating was compared for cellular frequencies as percent of viable for 19 overlapping markers between the adaptive and innate panels for spike-in reference sample (n = 36 replicates spike-in reference samples, bar graph of mean with standard error of mean for error bars). (B) Percent CV graph of data in shown in (A), dotted line at 5%. (C) Comparison of 1:1 mixed stimulated and ex vivo samples as independent barcoded samples (n = 2/panel).

Figure 12

Stimulation Conditions Are Tested to Detect Markers in Mass Cytometry Antibody Panels Healthy donor PBMC samples were treated with various stimulation conditions at various timepoints: PMA + ionomycin, CD3/CD28 bead stimulation, or a combination of CD3/CD28 beads with PMA + ionomycin. Specific markers are highlighted here to represent the dynamic range of activating these markers. Top row: CD11c. Middle row: Lag-3. Bottom row: CD69.