Lineage topology, replication kinetics and cell cycle synchronization reveal regulated growth dynamics in human bone marrow stromal cell colonies

TL;DR

Fundamental mechanisms governing BMSC heterogeneity and growth dynamics that may inform strategies to control their regenerative potential are revealed.

Abstract

Bone marrow stromal cells (BMSC) -- which include skeletal stem cells -- are a promising tool in regenerative medicine. However, their heterogeneous and unpredictable in vivo behaviour remains a critical barrier preventing the development of standardized therapeutic approaches for skeletal tissue regeneration. Several studies have attempted to identify in vitro features that could correlate with the in vivo differentiation properties, yet the mechanisms ruling BMSC heterogeneity remain poorly understood. Here, using time-lapse imaging, we lineage-trace 32 single-cell-derived BMSC colonies through seven generations. We observe significant inter-colony and intra-colony heterogeneity in lineage topology (determined by the number of senescent or apoptotic cells) and in replicative kinetics (measured from proliferating cells only). Interestingly, topology and kinetics result strongly correlated, suggesting the existence of regulatory factors linking the non-dividing/apoptotic subpopulations with proliferating cells. Furthermore, BMSCs display highly synchronized cell cycles during early generations, indicating stage-specific regulatory mechanisms through which cells influence each other. By employing a non-interacting population growth model, we demonstrate that the observed synchronisation cannot be explained by an uncorrelated branching process; instead, cell-to-cell correlation of division times must exist. Our findings reveal fundamental mechanisms governing BMSC heterogeneity and growth dynamics that may inform strategies to control their regenerative potential.

Figures (12)

• Figure 1: Time-lapse of a BMSC colony.(A-K) Images follow the evolution in time of colony 20241112_05 from its single originating cell. Panels (A-J) are different scale crops of images acquired at magnification $20\times$. (A) The originating cell $90$ minutes before its mitosis. (B) During the Mitotic Rounding Phase (MRP), cells assume a bright and almost spherical shape. We use the MRP as a marker for mitosis. The MRP of the originating cell marks the temporal beginning of the colony development defining the time origin ($t=0$) for the colony. (C) Half an hour after the beginning of MRP, the two daughter cells still have a spherical shape. (D) At $t=90$ minutes, the two daughters are distinguishable by their well-separated nuclei. (E-H) Within the first $48$ hours, the colony expands to the fourth generation. (I-J) During the third and the fourth days the colony expands to the fifth and sixth generation. (K) Overview at magnification $4\times$ of the entire dish at day five ($8\times 8$ photo tessellation). The white contour encloses the colony described in panels (A-J). Yellow contours indicate isolated colonies that we still follow on day five; red contours indicate infiltrated colonies (colonies that got in contact with each other), which we stop recording. Purple denotes colonies derived from cells that we did not select at the beginning of the experiment because of their close proximity. Finally, black curves correspond to cells that were selected but never formed colonies. Bottom panels. Samples of G$_0$ and active cells at the same magnification. G$_0$ cells have a different --flat-- morphology, and are larger than the spindle-shaped active cells.
• Figure 2: Topology --- Lineage trees.a) Abstract representation of the lineages of six BMSC colonies in the dataset. Each node (black dots) represents a mitosis and each link between two nodes represents a cell; the central node is the mitosis of the originating cell (the originating cell is not represented in the tree). All links have the same length, while their color represents the division time of the corresponding cell: light-green for short division times, dark-blue for long times (the mean division time across the whole dataset is about 20 hours). The thin elliptical lines separate different generations. All colonies have been followed up to the seventh generation, which means that the last divisions we observe are those between generation $k = 6$ and $k = 7$; hence, we do not measure the division times of the $k = 7$ cells, which are therefore colored grey. In some colonies all cells divide up to the seventh generation, giving rise to $2^7=128$ cells (20230113_07, top left), but in most colonies some inactive cells (G$_0$ or dead) emerge. We put a short black cap at the end of the link to represent a G$_0$ cell, or a red dot for a dead cell. Inactive cells do not have a division time and we color them in black. b) To show the significant inter-colony heterogeneity of the topology, we report for each lineage in the dataset the number of inactive cells and the number of missing cells at the seventh generation. Both quantities have fluctuations that are almost as large as the mean.
• Figure 3: Kinetics --- Statistics of the division times.a) We report the probability density of the cells' division times, $\tau$, for the same six colonies as in Fig. \ref{['fig:lineage-trees-topo']}a. The vertical green line in each histogram indicates the $84$ hours threshold that we use to define G$_0$ cells. In all colonies there is a threshold below which we find no division times. For this reason, a reasonably good fit of the data (full lines) is given by the gap-exponential form (see text), which is characterised by a minimum division time $\tau_\mathrm{min}$ and by a rate $\gamma$ of decay of the exponential part of the function (red line). b) We report the relative single-colony fluctuations of the rate $\gamma$ and of the minimum time $\tau_\mathrm{min}$ defined as $(x - \bar{x})/\bar{x}$, where $x$ represents the observable (either $\tau_\mathrm{min}$ or $\gamma$) and the average $\bar{x}$ is computed over all the colonies. As in the topology, a significant heterogeneity is detected also in the kinetics, in particular in the fluctuations of the rate $\gamma$, while the minimum division time, $\tau_\mathrm{min}$, fluctuates much less. The orange and purple shading represent the standard deviation of the single-colony fluctuations of $\gamma$ and $\tau_{min}$ respectively over all the colonies.
• Figure 4: Variability with experiment, donor, age, sex, and passage. We report here the mean division time for each colony in the dataset, ordered according to experiment, donor, age, sex, and the passage of the cell culture. Each vertical bar corresponds to a different colony; colour represents passage: P0-green, P1-orange, P2-pink, P3-cyan. Fluctuations within the same experiment are of the same order as the variability across the whole dataset. The only (very weak) correlation in the data is between division time and passage (Spearman correlation $\rho=0.30$, $P$-value$= 0.046$). We notice that the only colony that seems an outlier in this plot is 20240627_06 (largest division time of experiment 20240627); interestingly, this is also the colony most separated from the bulk of points in the topology vs kinetics plot (see Fig. \ref{['fig:correlation-topo-kin']}).
• Figure 5: The correlation between topology and kinetics.a) We plot the number of inactive cells of each colony, $N_\mathrm{inactive}$ as a function of the mean division time of that colony, $\overline{\tau}$ (orange points). The Spearman correlation between $N_\mathrm{inactive}$ and $\bar{\tau}$ is both strong and significant: correlation coefficient $\rho=0.825$ and $P$-value $< 10^{-6}$. b) A second characterisation of topology is given by the number of missing cells at generation $k=7$, $N_\mathrm{missing}$, which also happens to be strongly correlated to the mean division time, $\overline{\tau}$ (Spearman $\rho=0.761$, P-value $=1.0 \times 10^{-6}$). The dashed line represents the maximum number of missing cells, namely $2^7=128$. c-d) The rate $\gamma$ extracted from the fit of the division times histograms is a second characterisation of kinetics. We plot $N_\mathrm{inactive}$ and $N_\mathrm{missing}$ vs. $1/\gamma$ and find in both cases a strong and significant correlation ($\rho=0.815$, P-value $< 10^{-6}$ and $\rho=0.784$, P-value $< 10^{-6}$, respectively). Shaded areas are just a guide for the eye, corresponding to an amplitude of 1$\sigma$ (darker orange) and 2$\sigma$ (lighter orange) around a sigmoidal fit of the data, where $\sigma$ is the standard deviation of the residuals.