pDC, Siglec-H- pre-cDC, CD115+ CDP, CD127+ CLP, Ly6D+ lymphoid progenitors, Ly6D+ Siglec-H+ lymphoid progenitors, lo-lo and lo-hi were sorted one cell per well across nine 96-well plates, 96 cells per population, with each population spread across two plates. Using shared HVGs for analysis was sufficient to remove batch effects between plates. Libraries for scRNA-seq were prepared following the plate-based mcSCRBseq protocol as described. Single-read 50 bp sequencing was performed on the HiSeq1500 platform with a target sequencing depth of 50.000 reads per cell. Barcode/UMI-filtering, mapping and counting of the raw data was performed using the zUMIs pipeline (version 2.5.6). Within zUMIs, barcode sequences were quality filtered, allowing up to 2 bases below Phred quality score of 20. Remaining reads were mapped to the mouse genome (build mm10) using STAR (version 2.6.0a). Gene identities were obtained from Ensembl annotations (GRCm38.75). Velocity-tagged zUMIs output from the scRNA-seq data was processed using the Python velocyto v0.17 pipeline with specified barcodes. The resulting loom file was used for downstream RNA velocity analysis using the scvelo package v0.2.2. Cells with abnormally high or low gene counts were excluded from the analysis. Genes present in less than 10 cells were excluded as well. Cells showing high level expression of mast cell genes (Prss34, Prg2, Mcpt8) were excluded from the analysis. Cells with a total transcript number below 900 were excluded. Final analysis was performed on 675 cells.

pDC, Siglec-H- pre-cDC, CD115+ CDP, CD127+ CLP, Ly6D+ lymphoid progenitors, Ly6D+ Siglec-H+ lymphoid progenitors, lo-lo and lo-hi were sorted one cell per well across nine 96-well plates, 96 cells per population, with each population spread across two plates. Using shared HVGs for analysis was sufficient to remove batch effects between plates. Libraries for scRNA-seq were prepared following the plate-based mcSCRBseq protocol as described. Single-read 50 bp sequencing was performed on the HiSeq1500 platform with a target sequencing depth of 50.000 reads per cell. Barcode/UMI-filtering, mapping and counting of the raw data was performed using the zUMIs pipeline (version 2.5.6). Within zUMIs, barcode sequences were quality filtered, allowing up to 2 bases below Phred quality score of 20. Remaining reads were mapped to the mouse genome (build mm10) using STAR (version 2.6.0a). Gene identities were obtained from Ensembl annotations (GRCm38.75). Velocity-tagged zUMIs output from the scRNA-seq data was processed using the Python velocyto v0.17 pipeline with specified barcodes. The resulting loom file was used for downstream RNA velocity analysis using the scvelo package v0.2.2. Cells with abnormally high or low gene counts were excluded from the analysis. Genes present in less than 10 cells were excluded as well. Cells showing high level expression of mast cell genes (Prss34, Prg2, Mcpt8) were excluded from the analysis. Cells with a total transcript number below 900 were excluded. Final analysis was performed on 675 cells.