Page 25 - 2019_10 resto del Mondo_web

Extended myeloid-based model of lineage commitment
tions representing pure differentiation bifurcation points; or (ii) lineage commitments occur only in HSC, whereas each progenitor population is a mixture of committed cells sharing the same cell surface markers at the time of isolation. Since single-cell differentiation analysis from CMP and LMPP showed that a minor part of these pop- ulations can produce multilineage mature cells,14,15 there might be further subpopulations that represent actual bifurcation points of erythroid-myeloid or lymphoid- myeloid differentiation. However, whether or not these hypothetical subpopulations can be defined using addi- tional cell surface markers remains unclear. Before addressing these two possibilities, we need to stop and consider the potential limitations of recent studies using single-cell analyses. A single-cell transcriptomic analysis is a ‘snapshot’ observation. Therefore, if a set of genes shows dynamic fluctuations in expression with coherent patterns in cells of a specific subpopulation, these cells might be considered heterogeneous. However, these subpopulations can be homogenous when time-depen- dent fluctuations are considered, like those observed in neural progenitors.38 In addition, despite the importance of the microenvironment for HSPC biology,39 an ex vivo single-cell transcriptomic analysis is devoid of anatomi- cal information. Furthermore, single-cell in vivo or in vitro differentiation analyses can only examine the differenti- ation potential under stress and/or artificial conditions (i.e., cell sorting, culture and transplantation into irradiat- ed mice), which can skew the original differentiation tra- jectory of progenitor cells,40 possibly by altering activities of critical TF whose expression is thought to be main- tained to some extent for multilineage priming,41 a state in which multiple, conflicting lineage-affiliated genes can be induced or co-expressed. In other words, there is a chance that progenitor cells with unilineage output potential in perturbed conditions still possess multilin- eage output potential in an unperturbed condition. Recent studies using single-cell analyses may, therefore, lack information about the dynamics (time and three- dimensional information) of lineage commitment,42 espe- cially regarding unperturbed hematopoiesis. Potential effects of circadian rhythm in HSC differentiation might also have to be considered.43 The analysis of entropy in gene expression within single cells44 and the three- dimensional detection of transcriptomics45 might be helpful. Remarkably, recent in vivo barcoding analyses give new support to the existence of a hierarchical devel- opment model in hematopoiesis.46,47
We must therefore reconsider the actual point at which lineage commitment occurs. An alternative approach to define such a point involves using the regulatory mecha- nisms of the differentiation of HSPC. To this end, the precise understanding of gene regulatory networks gov- erned by TF may provide a dynamic view of lineage commitment.
This leads us to the second point that should be con- sidered: how are the differentiation trajectories shaped and restricted along the path of differentiation? Several models of lineage commitment have been proposed, showing that TF are critical to shaping and resolving the patterns of lineage-affiliated gene expression.48-50 One model features a network of two TF, each promoting dif- ferentiation into a specific lineage. If the expression of these two TF is inhibited in a mutual manner and thus they induce their own expression, they can define two
cell types with distinct expression patterns of the two TF and thus their downstream target genes (Figure 1).48 Machine-learning methods using single-cell transcrip- tomic data support the notion that gradual, stochastic changes in a few TF have a strong influence on the line- age commitment of progenitor cells.50 Such a gene regu- latory network may therefore dictate lineage commit- ment.
However, it has been unclear how one or the other of these TF are initially upregulated or downregulated upon lineage commitment. Stochastic fluctuation in these TF may be involved,48 but the output of hematopoiesis should be dynamically tuned in response to diverse stres- sors, as HSPC produce huge numbers of mature cells daily in a fine balance, as noted above. This property of the hematopoietic system may not therefore be fully explained merely by the stochastic fluctuation of TF. The differentiation trajectory of HSPC must be tightly con- trolled by responding to environmental changes in order to maintain homeostasis. This means that environmental factors, including pathogen-associated molecular pat- terns (PAMP) and damage-associated molecular patterns (DAMP),51 may affect the cell-intrinsic TF of gene regula- tory networks that control the differentiation trajectory. It is therefore important to understand how cell-intrinsic systems of TF are connected to extrinsic signals.
The gene regulatory network for erythroid lineage commitment
Erythroid cells are derived from progenitor cells that possess the ability to differentiate into erythroid or myeloid cells.1,52 CMP have long been considered to rep- resent a bifurcation point of erythroid-myeloid differen- tiation.32 However, single-cell analyses have challenged this notion. A single-cell RNA sequencing analysis of c- kit+Sca1-lineage- bone marrow cells revealed at least seven different subpopulations with lineage priming at the transcriptomic level.16 Importantly, no subpopula- tions with multilineage priming were observed in that
Figure 1. Stochastic model of lineage commitment. Transcription factors (TF) A and TF B play important roles in determining cell differentiation. If these two TF activate themselves and work in a mutually exclusive manner, slight sto- chastic fluctuations that alter the ratio of TF A to B can affect cell fate.
haematologica | 2019; 104(10)
1921