Transcriptional heterogeneity predicts and enables clonal selection in ageing haematopoiesis

Technical Summary: Transcriptional Heterogeneity Predicts and Enables Clonal Selection in Ageing Haematopoiesis

Problem Statement
A central puzzle in stem cell and ageing biology is why specific cellular clones dominate ageing tissues while phenotypically similar neighbors do not. Current understanding suggests that the average transcriptional state of a cell poorly predicts this fate. While single-cell technologies have revealed increased transcriptional and epigenetic heterogeneity in ageing haematopoietic stem cells (HSCs), and lineage-tracing studies have mapped clonal expansions, these fields have largely progressed separately. A critical gap remains in linking pre-existing molecular states to long-term clonal output. It is unclear whether clonal expansion in ageing results from stochastic drift, de novo acquisition of advantageous states, or the selective amplification of pre-existing heterogeneity within the HSC pool. Furthermore, the relative contributions of intrinsic HSC aging programs versus extrinsic niche signals in shaping clonal evolution are not fully disentangled.

Methodology
The authors employed a multi-modal approach combining heritable lineage tracing with single-cell RNA sequencing (scRNA-seq) to prospectively link HSC molecular states to clonal fate across different ageing contexts.

• Experimental Models: The study utilized CellTag barcoding (a heritable lentiviral barcode system) to track individual HSC clones. Experiments were conducted using both young (2 months) and old (20–22 months) C57BL/6J mice.
• Transplantation Designs: The authors performed homochronic (young-to-young, old-to-old) and heterochronic (young-to-old, old-to-young) transplantation assays. This design allowed for the separation of intrinsic donor effects from extrinsic host microenvironment effects.
• Longitudinal Tracking: scRNA-seq was performed at two time points: Day 0 (in vitro, pre-transplantation) and Day 60 (in vivo, post-engraftment). This enabled the correlation of baseline transcriptional states with functional outcomes.
• Computational Framework (scCloneVar): To quantify transcriptional heterogeneity at clonal resolution, the authors developed a new computational framework, scCloneVar. This tool estimates mean-adjusted gene expression variance to identify Differentially Variable Genes (DVG), distinguishing variability-driven signals from conventional differential expression (mean shifts).
• Validation and Expansion: Findings were validated in unmanipulated (non-transplanted) mouse bone marrow and extended to human datasets (including a lifespan atlas spanning ages 18 to 77). Machine learning models (Random Forest) were trained to predict clonal self-renewal based on early transcriptional states.
• Clinical Correlation: Gene signatures derived from the study were tested for prognostic value in The Cancer Genome Atlas (TCGA) Acute Myeloid Leukemia (LAML) cohort and enriched against GWAS loci for clonal haematopoiesis of indeterminate potential (CHIP) and myelodysplastic syndromes (MDS).

Key Contributions and Results

1. Ageing Increases Transcriptional Heterogeneity: The study confirms that ageing increases transcriptional entropy and variability in HSCs. Using scCloneVar, the authors identified a set of Differentially Variable Genes (DVG) that exhibit significantly higher variance in old HSCs compared to young ones. This increased heterogeneity is observed both within and between clones and is not merely stochastic noise but represents structured biological variation enriched for pathways regulating stem cell self-renewal, quiescence, and leukemogenesis (e.g., E2F, c-MYB, Rho GTPase signaling).
2. Clonal Adaptation to Age-Matched Niches: In heterochronic transplantation experiments, sister clones (sharing a common origin) exhibited divergent fates depending on the host age. Clones from old donors showed significantly higher self-renewal and fitness in old hosts compared to young hosts, and vice versa for young donors. This indicates that ageing niches impose selective pressures that favor clones pre-adapted to the specific age-matched microenvironment, rather than uniformly degrading HSC function.
3. Predictive Power of Pre-existing Heterogeneity: A major finding is that transcriptional variability measured prior to transplantation (Day 0) predicts clonal self-renewal and fitness at later time points (Day 60). Models trained on DVGs (variability) performed comparably or superiorly to models trained on differentially expressed genes (mean expression). This suggests that intra-clonal heterogeneity encodes latent information about a clone's future adaptability and fitness.
4. Temporal Establishment of Variability: Analysis of human lifespan datasets revealed that the signature of increased transcriptional variability begins to emerge in middle age (35–53 years), well before the typical onset of age-associated haematopoietic disorders like CHIP or leukemia. This variability continues to accumulate into late life.
5. Clinical Relevance: DVG-associated programs are conserved between mouse and human HSCs. Signatures derived from these variability genes are prognostic in AML patients (higher expression associated with worse survival) and are significantly enriched in GWAS loci associated with CHIP and MDS risk. Pathway analysis implicated MAPK, EGFR, TNF, and NFκB signaling as potential mechanisms by which variability confers selective advantage in aged environments.

Significance and Claims
The paper claims to provide a unifying framework for understanding stem cell ageing and age-associated clonal disorders. The authors posit that ageing does not merely degrade stem cell function globally but acts as a selective process that amplifies pre-existing transcriptional heterogeneity.

• Predictive Model: The fate of an HSC clone is partially predictable not just from its mean transcriptional state, but crucially from its intra-clonal expression heterogeneity. This establishes variability as a distinct, functionally meaningful axis of HSC state.
• Mechanism of Selection: The findings support a model where ageing expands the distribution of stem cell states, rendering clonal fate predictable from early cellular states. Selection operates on this pre-existing distribution, favoring clones whose transcriptional programs (characterized by specific variability patterns) are compatible with the aged niche.
• Clinical Implications: The identification of DVGs linked to malignancy risk suggests that monitoring transcriptional variability could serve as an early indicator of clonal selection and risk stratification for myeloid malignancies, potentially decades before mutational drivers reach detectable frequencies.
• Therapeutic Outlook: The authors suggest that interventions targeting age-selected stem cell programs (rather than attempting global rejuvenation) may be a more effective strategy to preserve tissue function and limit age-associated clonal disorders.

The study concludes that intrinsic heterogeneity provides the substrate for both ageing-associated selection and malignant evolution, bridging the gap between molecular noise and functional clonal dominance.