Assessing the transition state model for cellular differentiation in vivo: a case study of zebrafish neuromesodermal progenitors
Repository URI
Repository DOI
Change log
Authors
Abstract
During the process of cellular differentiation, stem or progenitor cells undergo a sequence of cell state changes before acquiring the characteristics of their destination fate. The transition state model proposes that a transient window of increased gene expression stochasticity and diversity precedes the coherent entry into the differentiated state. As the model has been assessed primarily via in vitro studies, I sought to explore whether a similar phenomenon can be observed within a cell population in vivo. In my thesis, I chose to study the neuromesodermal progenitors (NMps) in the zebrafish tailbud due to their relative biological simplicity. These progenitors co-express sox2 (neural marker) and tbxta (mesodermal marker) and differentiate into the posterior spinal cord or paraxial mesodermal lineages at the late somitogenesis stages, in the absence of other cellular processes such as proliferation, apoptosis and extensive mixing. Thus, I was able to relate cell states to cell fates as I analysed the gene expression heterogeneities in the NMp population in situ.
To assess whether the transition state model applies to zebrafish NMp differentiation in vivo, I established four experimental predictions. First, there should be an increase in the heterogeneities of sox2 and tbxta expression in the NMps as they enter the transition state. During mid to late somitogenesis, I found that the heterogeneities in sox2 and tbxta expression as well as the variability in cell number peaked at the 24-somite stage (24ss), which is when NMps contribute to the developing axis.
Second, as a transition state results from the flattening of the progenitor attractor during a bifurcation event, it should display several quantitative signatures of a critical transition. By analysing a high-dimensional single-cell RNA-seq dataset at 18ss, I observed an increase in the critical transition index and transcriptional noise in the NMp population relative to its derivatives.
Next, to follow the expression of neural and mesodermal markers within the tailbud at 18ss, 24ss and 28ss, I co-developed an image registration pipeline with my colleague in order to construct gene expression composite maps. Using these maps, I assigned individual NMps with an ‘NM index’ based on the cumulative expression of their neural and mesodermal markers. I used the NM index to assess my third prediction: that cell-cell variability in the NMp population should peak upon entry to the transition state. Consistent with this prediction, I found that the entropy of the NMp population is highest at 24ss.
Finally, during the transition state, a recent in vitro study proposed that ‘Rebellious’ cells may explore cell states that oppose their predicted cell fates. I noted an increase in the number of ‘Rebellious’ cells in my dataset, in the form of neural-biased cells that possess high Wnt/TCF activity residing in the mesoderm-fated domain at 24ss. To understand how these cells can arise under a stochastic regime, I explored a stochastic model of the genetic toggle switch and showed that by making the Wnt input time-dependent, I can recreate these rebellious cells in silico.
Taken together, my work demonstrates that several predictions of the transition state model hold true within an endogenous cell fate decision making event.