Visualization of individual cell division history in complex tissues using iCOUNT

Summary The division potential of individual stem cells and the molecular consequences of successive rounds of proliferation remain largely unknown. Here, we developed an inducible cell division counter (iCOUNT) that reports cell division events in human and mouse tissues in vitro and in vivo. Analyzing cell division histories of neural stem/progenitor cells (NSPCs) in the developing and adult brain, we show that iCOUNT can provide novel insights into stem cell behavior. Further, we use single-cell RNA sequencing (scRNA-seq) of iCOUNT-labeled NSPCs and their progenies from the developing mouse cortex and forebrain-regionalized human organoids to identify functionally relevant molecular pathways that are commonly regulated between mouse and human cells, depending on individual cell division histories. Thus, we developed a tool to characterize the molecular consequences of repeated cell divisions of stem cells that allows an analysis of the cellular principles underlying tissue formation, homeostasis, and repair.

The length of the amplification phase is not known precisely, and indeed may vary across different RGPs; but the MADM clone data suggests n s ∼ 5.
In the neurogenic phase (ca. E12-E16), RGPs follow a pattern of invariant asymmetric fate, with each cell division giving rise to one RGP and either a neuron (with probability q) or an intermediate progenitor, P (with probability 1 − q), i.e. for an RGP at generation g in the neurogenic phase, Once again, the precise range of the asymmetric phase is uncertain, but an estimate of n p ∼ 5 is consistent with the clone data.
Finally, while the fate behaviour of intermediate progenitors is also difficult to resolve precisely, with their output limited to at most three neurons, we can capture their dynamics by assuming that they choose between asymmetric division (with probability p 1) and terminal division (with probability 1−p), giving rise to neuronal progenies, Therefore, in the neurogenic phase, for each round of asymmetric division, an RGP generates an average output of From MADM tracings from the E12 time-point, analysis of sister clones suggests that q 0.3 and, from the measured average neurogenic output of a single RGP division n 1.87, it follows that p 0.2.
Finally, after n p rounds of asymmetric division, a fraction of RGPs then progress into a gliogenic phase (at around E16), where they give rise to astrocytes and oligodendrocytes. Since we are interested here in the neuronal outputs of RGPs, it is not necessary to consider further the fate behaviour within this phase.
Cell cycle number distribution: Based on this dynamics, as a starting point, we first considered the expected distribution of cell cycle number of neuronal outputs based on the induction of a RGP that is at some generation g within the neurogenic phase. Defining Q n as the probability of finding a neuron that has experienced precisely n rounds of division following induction of its RGP ancestor, we have that Here, the first term under the sum represents the one-half probability that, following the RGP division, the lineage follows a RGP fate rather than a N/P fate. The second term represents the probability d r that, once a N/P fate is adopted, there are r rounds of division before terminal differentiation. Note that here, since the progenitor differentiates through terminal division, there is a factor of 2 that enters the second component of the sum in the definition of d n .
In practice, if the RGP is labelled early in the neurogenic phase, the majority of neurons will derive from a generation n n p . In this case, we can make the approximation n p → ∞, whereupon In this limit, the average number of cell divisions experienced before neuron pro-duction is given by Then, for a RGP labelled at generation g during the amplification phase, a total of 2 g−ns+1 RGPs will enter into the neurogenic phase. So, if all labelled RGPs were synchronized, the distribution of cell cycle number of neurons gener-ated from such RGPs would be simply Q n+g−ns . However, in practice, RGPs are not synchronized in their progression through the amplification phase, but instead belong to a distribution of generation number which, empirically, fits well with the following dependence (Gao et al. 2014), where a(t) scales linearly with time, k(t) denotes the probability that, at the time of induction, the RGP is already in the neurogenic phase, and C(t) = ns g=1 2 g−a(t) e −2 g−a(t) denotes the total normalization. At the E12 induction time, measurements based on the MADM system suggest a figure of k(t) 0.72, i.e. some 72% of RGPs have already entered into the neurogenic phase.
Based on this definition, the distribution of cell cycle count number for the neuronal progenies of induced RGPs is given where the normalization is given by A −1 = ns g=0 2 g F g . Note that, here, the factor of 2 g accounts for the amplification of RGP number before entry into neurogene-sis. we obtain a prediction that fits remarkably well with the observed distributions ( Figure 4C).

Predictions
Based on the MADM tracings, for the E11.5 induction, it is expected that some 50% of RGPs are expected to be in their amplification phase (with a(t) 0), while the remaining RGPs have already progressed to the neurogenic phase. With tissue fixed at E19.5, the vast majority of these RGPs should have completed their neurogenic program. Then, using the same estimates for the probabilities p and q, the predicted cell cycle number distribution is shown in Figure 4G alongside the estimated values from the iCOUNT system. Here, there is generally good agreement between the model and the data, with the principal peak positioned around 3-4 rounds of division. However, the model predicts an excess of weight at higher numbers of cell divisions, consistent with the marking of cells deeper within the amplification phase.
This exposes a challenge for experiment since cells that have undergone multiple rounds of division would express very low levels of mCherry expression, where small fluctuations of intensity would alter dramatically the predicted cell cycle number. Nevertheless, the general quantitative agreement between the model pre-dictions (derived from the clonal data) and the experimental results obtained by the iCOUNT system emphasizes the potential of the latter to serve as a useful quantitative assay.