Mechanismen der Differenzierungsentscheidung einer eukaryontischen Zelle

The development and fate of a eukaryotic cell is controlled by complex networks of
interacting biomolecules. They regulate the differential expression of genes that drive the
process of cell differentiation or that are required to execute this process. There is a
tremendous body of information available on the regulatory mechanisms of gene expression in
eukaryotic cells and on the cell type-specific expression of transcription factors and other
regulators. However, the overall functional interplay of these molecules in determining the fate
of a differentiating cell or the identity of a stem cell which is able to differentiate into various
specialized cell types, is not really understood. Gene regulatory mechanisms in prokaryotes, the
lambda switch that controls the entry of the phage into the lytic cycle for example, are well-
established (Delbrück 1949; Kafri et al. 2013) and influence our thinking. The situation in
eukaryotes however is more complex with combinatorial control of gene expression. In
eukaryotes, many molecular factors can act together in binding to promoter regions, thereby
controlling the expression of a certain gene, often in a context-dependent manner. This variety
makes it very difficult to integrate the many pieces of knowledge on local molecular
mechanisms into a coherent picture or even into a mechanistic model of regulatory control that
would predict the differentiation behavior of the cell. Thus, while we can see expression
changes on a global scale and deduce connections between genes, we are currently only at the
beginning of discriminating and understanding cause and effect. This situation basically holds
for all eukaryotic cells, from Dictyostelium to mammalian cells. In well-studied yeasts for
example, interaction networks on gene expression and protein abundance levels were
described (e.g. Moignard et al. 2015), but even those detailed analyses yielded just static
pictures of potential hubs and interactions of components, while it is known that the dynamics of
regulation essentially determines the final outcome (Endres 2012; Rowland et al. 2012; Varusai
et al. 2015; Vilar et al. 2003).
Some regulatory networks are known to involve developmental switches that give rise to so
called commitment points, at which the decision on the cell fate is irreversibly made.
Commitment points in the cell cycle of yeasts are related, well-known paradigms for the control
of cell proliferation (Zachariae and Tyson 2016). Here, essential molecular players and their
fuctional interactions in generating irreversible switches are well understood. Co-regulated with
the cell cycle, the protein abundance is adjusted by modulating gene expression (Eser et al.
2011). In contrast however, commitment points that decide on the differentiation destiny of a
eukaryotic cell are currently not well understood.
There are two prevailing and in part contradicting views of how the differentiation of
eukaryotic cells is controlled, both dating back to the year of 1969 (Britten and Davidson 1969;
Kauffman 1969a; Kauffman 1969b) and both pursued and elaborated until today ((Thomas
1981; Abou-Jaoudé et al. 2016; Peter and Davidson 2011; Bornholdt and Kauffman 2019) and
references therein). The model by Britten and Davidson (Britten and Davidson 1969) assumes a- 2 -
hierarchical control based on uni-directional information flow (Peter and Davidson 2011)
assuming complex connections but without variability, acting like the rigid steering mechanics of
a technical device. The model by Kauffmann (Bornholdt and Kauffman 2019; Kauffman 1969a;
Kauffman 1969b) on the other hand explains the global dynamics of gene regulation by a
system of interconnected switches. While there are good arguments in favour for each of both
competing views, direct experimental proof or disproof is pending and obviously difficult
(Newman 2020).
Intrinsic transcriptional heterogeneity is widely observed in clonal populations of
mammalian cells in culture, but also occurs in intact tissues under physiological conditions
(Marco et al. 2014). Approaches to explain this heterogeneity while considering the global
dynamics which is expected for a complex gene regulatory network (and which the Britten-
Davidson model neglects), are based on the Kauffman model and have been metaphorically
illustrated by Waddington's epigenetic landscape (Huang et al. 2009; Waddington 1957). Here,
the global dynamics of the regulatory network is explained by a quasi-potential landscape, in the
following simply called Waddington landscape, that represents possible states of the dynamic
system while defining the probabilities for state transitions to occur ((Graf and Enver 2009;
Huang 2011; Huang et al. 2009; Macarthur et al. 2009; Moris et al. 2016; Wu et al. 2017; Zhou
and Huang 2011); Fig. 1A). Although there are theoretically sound formal frameworks that
principally allow to compute the quasi-potential landscape from a set of differential equations,
these approaches have currently still limited practical value simply because the molecular
interactions within considered regulatory networks are not sufficiently known and hence,
differential equations and their parameters are elusive. Experimental analysis, on the other
hand, would require the measurement of true time-series in individual cells but such approaches
are still in the fledgling stages, at least in mammalian cells.
The reconstruction of pseudo-time series from static snapshots taken of mammalian cell
populations depends on certain assumptions and unequivocal conclusions are hardly possible
for principle reasons (Weinreb et al. 2018). This limitation however, tends to be neglected or
even ignored for the sake of simplicity. Until recently, the Waddington landscape, and the
existence of attractors, accordingly remained a theoretical concept. It is supported by many
experimental observatios, while basic features, including the functional role of stochasticity, are
still a matter of pure speculation ((Moris et al. 2016) and references therein).
We have overcome this limitation by developing an experimental system that allows to
take true time series, i.e. to test the same cell all over again. True time series for individual cells
can be taken by repeated, non-destructive sampling retrieving just small parts of the stirred
cytoplasmic volume of the giant amoeba Physarum polycephalum. In these multinucleate cells
the cytoplasm is homogenous due to continuous mixing by the vigorous cytoplasmic streaming
(Guttes and Guttes 1961, 1964; Rusch et al. 1966; Sachsenmaier et al. 1972; Starostzik and
Marwan 1995a; Walter et al. 2013) (Pretschner et al. 2021). Based on our single cell data we
have developed an appropriate computational approach to identify attractors, to reconstruct the
Waddington landscape from gene expression time series, and to disentangle the complex
response revealing the differential regulation of the individual genes (Rätzel et al. 2020;
Werthmann and Marwan 2017; Pretschner et al. 2021). We have shown that cells, as predicted
by the model of the Waddington landscape, take individually different gene expression routes
(trajectories) to sporulation and that these routes converge to highly similar states of gene
expression. These findings however are only valid for the small set of 35 genes analysed in the
respective studies. Although our work resulted in a proof of principle, the molecular details of
developmental switching and especially of the commitment point have, due to the limited size of
the data set, not yet been identified. Neither in Physarum nor in mammalian cells it is clear
whether all cells of a population do cross the same commitment point or might use alternative
commitment points with different molecular signatures and mechanisms, that all might lead to
the same differentiated state. These fundamental questions are addressed in the proposed
project and the generation of the necessary data sets simply depends on funding.