Eq calibration beads

Fresh blood was processed and stained using the Maxpar Direct Immunoprofiling Assay (MDIPA) (Fluidigm or The Longwood Medical Area CyTOF core, Boston, MA). Following processing and staining, samples were stored at −80° C. The frozen stained blood samples were placed at room temperature (RT) for 30–50 min until fully thawed and then incubated with 1× Thaw-Lyse buffer (Smart Tube Inc.) at RT for 10 min. The lysis steps were repeated up to 3 times until the pellet turned white evidencing complete lysis of red blood cells. The cells were then incubated with Maxpar Fix and Perm Buffer (Fluidigm) containing 125 nM Cell-ID Intercalator-Ir solution (Fluidigm) at 4° C overnight or up to 48 h before sample acquisition. Groups of samples (4–8/day) were assessed by Helios mass cytometry (Fluidigm) in 4 independent experiments using a flow rate of 45 μI/min in the presence of EQ Calibration beads (Fluidigm) for normalization. An average of 400,000 ± 50,000 cells (mean ± SEM) from each sample were acquired and analyzed by Helios. Gating was performed on the Cytobank platform (Cytobank, Inc. Santa Clara, CA) and FlowJo 10.5.3 (BD Biosciences).

CyTOF assay was performed according to the manufacturer’s instruction. Briefly, 3×10⁶ cells were stained with 5μM Cell-ID™ Cisplatin (Fluidigm, San Francisco, CA) for 5min and quenched with MaxPal Cell Staining Buffer (Fluidigm) using 5× the volume of the cell suspension. After centrifugation, cell suspensions (50μl) were incubated with 5μl of human Fc-receptor Blocking solution (Biolegend, San Diego, CA) for 10min and 50μl of pre-mixed antibody cocktail for 30 min. After washing, cells were incubated with 1ml of cell intercalation solution (125 nM MaxPal Intercalator-Ir into 1ml MaxPal Fix and Pem Buffer, Fluidigm) overnight at 4°C. Cells were centrifuged with MaxPal Water and pelleted. The pelleted cells were suspended with EQ Calibration Beads (Fluidigm) and cell events were acquired by a CyTOF II instrument (Fluidigm). The data were analyzed using online software Cytobank, viSNE map and SPADE.

Raw data were normalised with the EQ Calibration beads and individual sample de-convolution (using the barcodes) was performed with the commercially available software (Fluidigm). Initial data quality and gating for intact single and alive cells were determined using traditional cytometry visualisation with software available from Cytobank^{70 (link)} (Supplementary Fig. 1a). In-gate events were imported in R/Bioconductor (v. 3.5), arcsinh transformed and filtered based on the median marker signal to remove cells with extremely low or high median values (±2 × IQR). Samples with a low number of events were discarded and cosine normalisation was applied. Downsampling was performed before tSNE and clustering analysis. A summary of samples and the number of events for all the experiments are reported in Supplementary Table 4.
In the ‘PDTX characterisation’ and ‘PDX target engagement’ experiments, where more than one batch had to be combined, reference samples were included in each batch (STG139, AB551 and STG335, respectively) and the absence of significant bias was verified by tSNE analysis and signal distribution evaluation (Supplementary Figs. 2c, d, 7b, respectively).