Lineage tracing on transcriptional landscapes links state to fate during differentiation

By Caleb Weinreb, Alejo E. Rodriguez-Fraticelli, F. Camargo, Allon M. Klein. 2018

638 citations recorded by the source corpus.

Mapping cell fate during hematopoiesis Biologists have long attempted to understand how stem and progenitor cells in regenerating and embryonic tissues differentiate into mature cell types. Through the use of recent technical advances to sequence the genes expressed in thousands of individual cells, differentiation mechanisms are being revealed. Weinreb et al. extended these methods to track clones of cells (cell families) across time. Their approach reveals differences in cellular gene expression as cells progress through hematopoiesis, which is the process of blood production. Using machine learning, they tested how well gene expression measurements account for the choices that cells make. This work reveals that a considerable gap still exists in understanding differentiation mechanisms, and future methods are needed to fully understand—and ultimately control—cell differentiation. Science, this issue p. eaaw3381 Single-cell barcoding allows the tracking of gene expression states over time as blood cells differentiate during hematopoiesis. INTRODUCTION During tissue turnover, stem and progenitor cells differentiate to produce mature cell types. To understand and ultimately control differentiation, it is important to establish how initial differences between cells influence their ultimate choice of cell fate. This challenge is exemplified in hematopoiesis, the ongoing process of blood regeneration in bone marrow, in which multipotent progenitors give rise to red cells of the blood, as well as myeloid and lymphoid immune cell types. In hematopoiesis, progenitor cell states have been canonically defined by their expression of several antigens. However, as in several other tissues, recent transcriptome analysis by single-cell RNA sequencing (scSeq) showed that the canonically defined intermediate cell types are not uniform, but rather contain cells in a variety of gene expression states. scSeq also showed that the states of hematopoietic progenitors form a continuum, differing from classic depictions of a discrete stepwise hierarchy. RATIONALE In this study, we set out to establish how variation in transcriptional state biases future cell fate and whether scSeq is sufficient to completely distinguish cells with distinct fate biases. Directly linking whole-transcriptome descriptions of cells to their future fate is challenging because cells are destroyed during scSeq measurement. We therefore developed a tool we call LARRY (lineage and RNA recovery) that clonally tags cells with DNA barcodes that can be read using scSeq. Using LARRY, we aimed to reconstruct the genome-wide transcriptional trajectories of cells as they differentiate. RESULTS We linked transcriptional progenitor states with their clonal fates by barcoding heterogeneous cells, allowing cell division, and then sampling cells for scSeq immediately or at later time points after differentiation in culture or in transplanted mice. We profiled >300,000 cells in total, comprising 10,968 clones that gave information on lineage relationships at single time points and 2632 clones spanning multiple time points in culture or in mice. We confirmed that clonal trajectories over time approximated the trajectories of single cells and were thus able to identify states of primed fate potential on the continuous transcriptional landscape. From this analysis, we identified genes correlating with fate, established a lineage hierarchy for hematopoiesis in culture and after transplantation, and revealed two routes of monocyte differentiation that give rise to distinct gene expression programs in mature cells. The data made it possible to test state-of the-art algorithms of scSeq analysis, and we found that fate choice occurs earlier than predicted algorithmically but that computationally predicted pseudotime orderings faithfully describe clonal dynamics. We investigated whether there are stable cellular properties that have a cell-autonomous influence on fate choice yet are not detected by scSeq. By analyzing clones split between wells or transplanted into separate mice, we found that the variance in cell fate choice attributable to cell-autonomous fate bias was greater than what could be explained by initial transcriptional state. Less formally, sister cells tended to be far more similar in their fate choice than pairs of cells with similar transcriptomes. These results suggest that current scSeq measurements cannot fully separate progenitor cells with distinct fate bias. The missing signature of future fate choice might be detectable in the RNA that is not sampled during scSeq. Alternatively, other stable cellular properties such as chromatin state could encode the missing information. CONCLUSION By integrating transcriptome and lineage measurements, we established a map of clonal fate on a continuous transcriptional landscape. The map revealed transcriptional correlates of fate among putatively multipotent cells, convergent differentiation trajectories, and fate boundaries that could be not be predicted using current trajectory inference methods. However, the map is far from complete because scSeq cannot separate cells with distinct fate bias. Our results argue for looking beyond scSeq to define cellular maps of stem and progenitor cells and offer an approach for linking cell state and fate in other tissues. Lineage and transcriptome measurements allow fate mapping on continuous cell state landscapes. A tool we named LARRY labels cell clones with an scSeq-compatible barcode. By barcoding cells, letting them divide, and then sampling them immediately or after differentiation, it is possible to link the initial states of cells with their differentiation outcomes and produce a map of cell fate bias on a continuous transcriptional landscape. A challenge in biology is to associate molecular differences among progenitor cells with their capacity to generate mature cell types. Here, we used expressed DNA barcodes to clonally trace transcriptomes over time and applied this to study fate determination in hematopoiesis. We identified states of primed fate potential and located them on a continuous transcriptional landscape. We identified two routes of monocyte differentiation that leave an imprint on mature cells. Analysis of sister cells also revealed cells to have intrinsic fate biases not detectable by single-cell RNA sequencing. Finally, we benchmarked computational methods of dynamic inference from single-cell snapshots, showing that fate choice occurs earlier than is detected by state-of the-art algorithms and that cells progress steadily through pseudotime with precise and consistent dynamics.

Subfields: Biophysics, Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Biophysics

In Molecular Biology

Citations in the corpus

Understanding How Dimension Reduction Tools Work: An Empirical Approach to Deciphering t-SNE, UMAP, TriMAP, and PaCMAP for Data Visualization (2020, methodology): “This is the most common case it is used in single-cell transcriptomic data (see, e.g. Wagner et al., 2018; Weinreb et al., 2020).”
A Unifying Perspective on Neighbor Embeddings along the Attraction-Repulsion Spectrum (2020, background): “…ForceAtlas2 (Jacomy et al., 2014) has recently gained popularity in the single-cell transcriptomic community to visualize datasets capturing cells at di ﬀ erent stages of development (Weinreb et al., 2018, 2020; Wagner et al., 2018a; Tusi et al., 2018; Kanton et al., 2019; Sharma et al., 2020).”
Probabilistic harmonization and annotation of single‐cell transcriptomics data with deep generative models (2019, positive): “In the simplest scenario, we have access to one dataset where states have been annotated either ab initio , or using additional experimental measurements (e.g., protein expression (Zheng et al , 2017; Stoeck-ius et al , 2017) or lineage tracing (Weinreb et al , 2020)) and another, unannotated…”
CellRank for directed single-cell fate mapping (2020, neutral): “To overcome the deficiencies of splicing data, it may be possible to extend the CellRank model to epigenetic information such as chromatin accessibility, leveraging the directional information coded by the typical delay between epigenetic and transcriptional changes 61,62 .”
Mapping Transcriptomic Vector Fields of Single Cells (2022, positive): “We also arrive at similar conclusions when analyzing data from a recently published study (Weinreb et al., 2020) with sequential clone cell tracing (Figures S5G–S5I).”