Single-cell genotyping

Single-cell genotyping was performed following the procedure previously described² Briefly, allele-specific TaqMan probes (ThermoFisher) were designed for selected mutations representing the previously calculated clusters (VAF-based clonal evolution)³³, probes labeled with VIC detected WT allele, and probes labeled with FAM detected mutant allele in every case. For JAK2 V617F, IDH2 R140Q, KRAS G12R, NRAS Q61P, commercially available probes were used (ThermoFisher). In Supplementary Tables ²–⁴ and Supplementary Data ³ a summary of designed probes, sequences, and quality control assessments are outlined. Lin-CD34 + cells were single-cell sorted in 96 well plates, lysed and DNA was pre-amplified with the multiplexed-specific TaqMan probes and the TaqMan Preamplification Master Mix from ThermoFisher. Pre-amplified material was used for high-throughput qPCR reactions carried on a BioMark HD using 192.24 dynamic array plates (Fluidigm). Data was collected and images were inspected manually.

For each signal, manual inspection of amplification curves and amplification threshold setting was done. Reactions under the threshold were considered to be negative. Reactions with negative values for both probes, were considered failed and cells that presented at least one failed reaction were discarded. Wells with no reaction were considered empty. A summary of number of cells processed and the total number of cells used for each patient and experiment is shown in Supplementary Table ⁴.

For each sample, an additional calibration plate was sorted in parallel for the estimation of sorting errors as shown in Supplementary Fig. ¹³. Copy Number TaqMan probes were used for the estimation of doublets rates (Supplementary Table ²). False-discovery rates were determined using the K562 cell line, and estimating proportion of false-positive single-cells per probe (Supplementary Table ³). Plate processing was carried out simultaneously for each sample. An overview of the entire experimental design and procedure is shown in Supplementary Fig. ¹⁴.

CALR mutations, IDH2 R140Q, and ARMCX5 E546E could not be detected in the Biomark HD system. In the case of CALR mutation type B, IDH2 R140 and ARMCX5 E546E, standard qPCR reactions were performed (Applied Biosystems) and for CALR mutation type A, a ddPCR was used (Bio-Rad). Reaction data from passing cell reactions were coded as follow, 0: WT, 1: heterozygous, 2: LOH (homozygous for the mutant). Data matrices were the used as input for the estimation of the clonal composition and phylogenetic analysis.