Transcriptional heterogeneity predicts and enables clonal selection in ageing haematopoiesis

A general puzzle in stem-cell and ageing biology is why a few cellular clones come to dominate an ageing tissue while otherwise similar neighbours do not, a fate that the average transcriptional state of a cell predicts poorly. Here we ask whether the variability between sister cells of a clone, rather than their transcriptional state, is the property that predicts ageing-associated clonal selection, using the haematopoietic stem cell (HSC) as a tractable test case. We combine heritable lineage tracing with single-cell RNA sequencing across heterochronic and homochronic transplantation models to link early transcriptional states of individual HSC clones to their long-term functional output in vivo. To quantify transcriptional heterogeneity at clonal resolution, we developed a computational framework (scCloneVar) that estimates mean-adjusted gene expression variance and identifies differentially variable genes (DVG). We found that ageing increases transcriptional heterogeneity at both the cellular and clonal levels, reflected by elevated variability in gene expression programs that regulate stem cell activity. We observe polyclonal expansion of HSC independently of the age of the host or the donor mice; however, individual clones in heterochronic transplantations show reduced self-renewal and fitness compared to sister clones in homochronic transplantations, indicating better adaptation of HSC clones in age-matched microenvironments. Strikingly, transcriptional features measured prior to transplantation predict clonal self-renewal at later time points, with transcriptional variability, captured by DVG, providing predictive power beyond that captured by mean expression differences. DVG-associated programs are conserved across mouse and human HSC, are established by middle age, and are enriched in pathways relevant to clonal haematopoiesis and myeloid malignancy risk. Together, our findings support a model in which ageing expands transcriptional heterogeneity that tracks with subsequent clonal selection, rendering clonal fate partially predictable from early cellular states.