Trimodal single-cell profiling reveals principles of cell fate acquisition in human erythropoiesis. Free

Hematopoiesis is a highly dynamic and hierarchical process that gives rise to all blood cell types from multipotent hematopoietic stem cells. This system must simultaneously maintain self-renewal, allow for multilineage output, and respond rapidly to environmental changes. Over the last decade, single-cell transcriptomic studies have provided a high-resolution view of differentiation trajectories, showing that lineage commitment occurs gradually and not as discrete steps. However, by focusing solely on RNA expression, these studies have blurred the molecular definition of classical cell populations, traditionally identified by functional assays and surface markers. Moreover, scRNA-seq cannot capture epigenetic priming events that often precede transcriptional change. Lineage tracing studies have shown that cell fate decisions are made earlier than transcriptomic data alone can predict, suggesting that integrating chromatin accessibility is essential for a comprehensive understanding of differentiation. To overcome these limitations, we applied a trimodal single-cell assay, TEA-seq, which simultaneously profiles gene expression, chromatin accessibility, and surface protein levels in the same individual cells. We generated a high-resolution map of human hematopoiesis using TEA-seq on hematopoietic stem and progenitor cells isolated from umbilical cord blood. Cells were collected at four time points across a fourteen-day erythroid differentiation culture system. Each sample was stained with 158 barcoded antibodies targeting surface proteins, followed by joint measurement of the transcriptome and epigenome. Data from all three modalities were integrated using MultiVI, a probabilistic model that learns a shared latent space across cells, accounting for batch effects.

The resulting map reconstructs the erythroid trajectory and links progressive fate acquisition with specific chromatin and protein signatures. Early in differentiation, transcriptional changes are primarily driven by fluctuations in transcription factor abundance, with minimal changes in chromatin accessibility. In intermediate progenitors, chromatin remodeling becomes the dominant regulatory mechanism. At later stages, transcriptional dynamics once again take precedence, acting on a largely pre-established chromatin landscape. These observations suggest a model in which chromatin accessibility serves as a scaffold for future transcriptional activation, decoupling stemness loss from immediate lineage commitment.

To further investigate how these molecular changes are orchestrated, we reconstructed enhancer-based gene regulatory networks using SCENIC+, which integrates TF expression, chromatin accessibility, motif enrichment, and enhancer-gene associations. We identified dynamic rewiring of gene regulatory networks (GRNs) along the differentiation trajectory. In early hematopoietic stem and progenitor cells, GRN transitions were driven by transcriptional changes. In intermediate populations, chromatin remodeling was the major driver. In late-stage progenitors, transcription factor abundance once again played a critical role. These distinct regulatory phases indicate that cell identity is shaped by the dynamic rewiring of the GRN, involving a complex interplay between transcription factor abundance and chromatin accessibility.

Notably, stemness-related GRNs remain active well into the erythroid trajectory, consistent with retained plasticity in late erythroid progenitors. These findings demonstrate that loss of stemness and acquisition of lineage identity are not synchronous events but are governed by temporally and mechanistically distinct programs. Chromatin remodeling at cell-specific enhancers occurs gradually and establishes a permissive state, while changes in transcription factor availability trigger abrupt transcriptional shifts. This sequential architecture likely serves to buffer against stochastic activation and ensures robust lineage commitment.

Together, our work establishes the first dynamic, enhancer-based gene regulatory networks built from simultaneous single-cell measurements of chromatin accessibility, gene expression, and protein abundance in human hematopoiesis. This trimodal approach unifies classical immunophenotyping with transcriptomic and epigenetic profiles. Our interactive web portal enables further exploration of the GRN and provides a resource for understanding cell fate decisions in human hematopoiesis.