Extended Data Fig. 3: SPLINTR identifies both LSC and non-LSC like transcriptional signatures in disease causing clones.

a, Schematic of an in vivo experiment to identify the transcriptional features of dominant murine MLL-AF9 + Kras^G12D leukaemic clones that drive disease. b, Correlation plot between the frequencies of SPLINTR barcoded bone marrow (left) and spleen (right) clones at the Baseline and Disease timepoints (n = 4 mice). Inset values indicate adjusted R² and p-values using the F-test for linear regression on frequencies of barcodes detected at both timepoints. Clone barcode frequencies were quantified by barcode-seq. Dashed line indicates a threshold frequency (2%) above which a clone is considered dominant. (n = 4 mice) c, UMAP projection of scRNA-seq data from SPLINTR barcoded MLL-AF9 + Kras^G12D cells taken at the baseline timepoint prior to transplantation. Top row, left to right: Louvain clusters, cells comprising dominant clones, cells coloured according to module enrichment scores for genes upregulated in leukaemic stem cells (LSC Signature - UP). Bottom row, left to right: Cells coloured according to module enrichment scores for genes upregulated in normal long-term haematopoetic stem cells (LT-HSCs), common myeloid progenitors (CMPs) and granulocyte/macrophage progenitors (preGM/GMPs). q10 = 10^th quantile, q90 = 90^th quantile. Module enrichment scores for genes upregulated in LSCs (d), normal LT-HSC (e) or GMPs (f) or CMPs (g) for each dominant clone vs. all other cells (cells per group left to right: n = 24, 34, 29, 147, 115, 32, 13115). Asterisks indicate the result of a two-sided T-test using Holm multiple testing correction for each clone vs. all other cells. Number of cells shown above each category. Adjusted p-values (left to right) for (d): 6e⁻⁰⁷, 3.2e⁻⁰⁷, 0.33, 0.25, 3e⁻³⁴, 0.92. (e): 1, 1, 1, 0.00012, 2.2e⁻¹⁴, 1. (f): 2.6e⁻⁰⁷, 9.1e⁻⁰⁷,1, 1, 1.2e⁻⁴⁵, 1. (g): 1.5e⁻¹⁰, 9.1e⁻¹¹, 0.76, 6.9e⁻⁰⁶, 3.7e⁻⁵⁵, 0.76. **** = p-adj < 0.001. Boxplots for (d) to (g) span the upper quartile (upper limit), median (centre) and lower quartile (lower limit). Whiskers extend a maximum of 1.5x IQR. h, Integration of scRNA-seq data sets from MLL-AF9 primary leukaemia cells from (c) and healthy progenitor HSPC populations from Dahlin et al. 2018. Top-left panel: UMAP projection showing Louvain clusters. Top-middle panel: Cells labelled according to their sample of origin. Primary MLL-AF9 baseline timepoint cells (MLL-AF9), Dahlin et al. lineage negative, c-Kit positive cells (LK), and Dahlin et al. lineage negative, Sca-1 positive and c-Kit positive cells (LSK). Top-right: UMAP projection with dominant primary MLL-AF9 clones highlighted. Bottom row: UMAP projections showing gene set module enrichment scores for genes upregulated in Long Term Haematopoietic Stem Cells (LT-HSC) (left), Common Myeloid Progenitors (CMPs) (middle) and Granulocyte/Macrophage progenitors (preGM/GMPs) (right). q10 = 10^th quantile, q90 = 90^th quantile. i, Hypergeometric test for Louvain clusters enriched for dominant MLL-AF9 clones. Negative log₁₀ transformed P values per cluster are shown. Red clusters are significantly enriched for dominant clones. Dashed line indicates a p = 0.05 significance threshold. j, The percentage of the 3 different cell types comprising each of the Louvain clusters in the integrated dataset. Clusters enriched for dominant MLL-AF9 clones are shown in red.