Robust assessment of asymmetric division in colon cancer cells

TL;DR

This work tackles the challenge of quantifying partitioning fluctuations during asymmetric cell division in cancer cells by introducing a flow cytometry–based method coupled to a stochastic partitioning model. The authors show that the mean inherited component halves each generation, while the variance encodes the second moment of the partitioning distribution Pi(f), with the key relation mu_g = mu_0/2^g and sigma_g^2 = mu_0^2( E[f^2]^g - (1/2)^{2g}) + sigma_0^2 E[f^2]. They validate the approach against extensive live-cell microscopy and apply it to colon cell lines, revealing cell-type–specific cytoplasmic partition fluctuations linked to size bias in division. The method provides a scalable, accurate means to quantify partition noise across biological systems, offering insights into cancer heterogeneity, and can be extended to other organelles and adherent cell types to study fate diversification and plasticity.

Abstract

Asymmetric partition of fate determinants during cell division is a hallmark of cell differentiation. Recent work suggested that such a mechanism is hijacked by cancer cells to increase both their phenotypic heterogeneity and plasticity and in turn their fitness. To quantify fluctuations in the partitioning of cellular elements, imaging-based approaches are used, whose accuracy is limited by the difficulty of detecting cell divisions. Our work addresses this gap proposing a general method based on high-throughput flow cytometry measurements coupled with a theoretical framework. We applied our method to a panel of both normal and cancerous human colon cells, showing that different kinds of colon adenocarcinoma cells display very distinct extents of fluctuations in their cytoplasm partition, explained by an asymmetric division of their size. To test the accuracy of our population-level protocol, we directly measure the inherited fractions of cellular elements from extensive time-lapses of live-cell laser scanning microscopy, finding excellent agreement across the cell types. Ultimately, our flow cytometry-based method promises to be accurate and easily applicable to a wide range of biological systems where the quantification of partition fluctuations would help accounting for the observed phenotypic heterogeneity and plasticity

Figures (4)

• Figure 1: Modeling partitioning noise at cell division.a) Schematic representation of the growth and division process of a mother cell. The cell undergoes an initial phase of duplication, where its internal elements are multiplied, followed by a division phase, where these elements are partitioned between the two daughter cells. One daughter inherits a fraction $p$ of the mother's elements, while the other receives the remaining fraction $q = 1-p$. b) (blue) Idealized behavior of the components number of a cell element over time, considering three different noise terms: fluctuations in compound counts during growth, uncertainty in the timing of division, and noise in the compound’s partition fraction. (red) Same description, but considering only production and degradation processes. c) Microscopy images of a single cell division event for an HCT116 cell, whose cytoplasm is stained with Celltrace Yellow. d) Time evolution of the distribution of the components' number of a cellular element for a population of cells subjected only to partitioning noise. The distribution used for the simulation is a Gaussian distribution with mean $\mu = \frac{1}{2}$ and standard deviation $\sigma = 0.07$. e) Examples of partition distributions $\Pi$, with increasing coefficient of variation. f) Mean ($\mu_g$) and variance ($\sigma_g$) of the number of components as a function of the generations, g, for the proliferation of a population subject to different partition noise distributions. Different colors correspond to distributions with varying coefficients of variation, as represented in the top panels. The dashed line represents the theoretical behavior obtained from the model. Dots are colored according to the distributions shown in panel e). g) Behavior of the mean (left) and standard deviation (right) of the simulated dynamics compared to the expected theoretical behavior for a proliferating population assuming a sizer division strategy (see SI).
• Figure 2: Quantification of partitioning noise via population-level measurements.a) Time evolution of CellTrace-VioletTM fluorescence intensity distribution measured in a flow cytometry time course experiment for a population of HCT116 cells. Time progresses from the darkest shade of blue to the lightest, spanning from $[0, 84]$ hours (bottom line to top). b) Snapshots of the evolution of the distribution of CellTrace-Violet fluorescence intensity measured in a flow cytometry time course experiment for a population of HCT116 cells. Experimental data are represented by the light blue histogram, while the best-fit Gaussian Mixture Model is displayed as lines, with different colors representing different generations. c) Mean (left) and variance (right) of the intensity of the fluorescent markers as a function of generations, normalized to the initial population values. Each replica of the experiment is identified by a different color and a point marker's shape. The experiments are conducted on Caco2 cells, and the points correspond to the mean values for each generation. Dashed lines represent the best fit according to Equations \ref{['eq: sigmag']} and \ref{['eq: mug mu0']}, respectively. d) Division asymmetry of the $\Pi(f)$ obtained by fitting Equation \ref{['eq: sigmag']} to data for all experiments and cell lines. The division asymmetry is measured via the percentage of the coefficient of variation.
• Figure 3: Quantification of partitioning noise via single-cell measurements.a) Example of a recorded cell colony of HCT116 cells in brightfield (left) and on CTFR fluorescence (right). b) Cell cytoplasm fluorescence intensity as a function of time for a cell before and after division. Dark green circles correspond to the fluorescence intensity of the mother cell up to the division frame, and then to the sum of the daughters' fluorescence. Lighter green triangles and squares represent the fluorescence intensity of the daughter cells. Solid lines are the linear fit of the points. The intercepts of the linear fit are used to compute the fraction of tagged components inherited by the daughter cells. The time is counted from the start of the experiment. c) (bottom) Strip plot of the distribution of the inherited fraction of cytoplasm for the different cell lines. The points are randomly spread on the y-axis to avoid overlay. (top) Fit of the inherited fraction distribution with the sum of two Gaussians with a mean symmetric to $1/2$. d) Comparison of division asymmetry obtained with time-lapse fluorescent microscopy measures (striped bars) with the ones obtained from flow cytometry experiments (plain bars). The flow cytometry bars are obtained as the mean over the multiple conducted experiments.
• Figure 4: Cytoplasm partition fluctuations vs cells size.a) Behavior of the integral term $\Sigma(N)$ (Eq.\ref{['sigma_VERA']}) as a function of the number of dividing elements, $N$, at fixed $\sigma_{N_i} = 0.8$. Vertical lines mark typical values of cellular elements, like mitochondria. b) Theoretical value of the variance of $\Pi(f)$ in the binomial assumption for different levels of asymmetries and increasing values of $N$. c) Asymmetry of the partitioning distribution $\Pi(f)$ in the binomial limit as measured by the binomial bias, p. d) Sample cases of volume asymmetric division for different cell lines. Images show the overlay of consecutive times during the division dynamic. Time flows from the bottom left to the top right.