Decomposition of cell activities revealing the role of the cell cycle in driving biofunctional heterogeneity

The evolutions of both cell velocity and morphology stem from cell cycle progression

We first monitored whether NIH 3T3 fibroblasts possess significant phenotype transitions during the interphase of the cell cycle through a representative cell migratory trajectory and the corresponding cell shape (Fig. 1A). The results reveal that the trajectory is a straight line in the beginning but becomes curved later. Accordingly, the cell shape also transforms significantly. Thereafter, we scrutinized the motility and polarity of the cell at 3-min intervals over the whole cell cycle in terms of basic cell parameters, including instantaneous speed, area, aspect ratio, and circularity (Fig. 1B). The least-square criterion was also applied to classify the complete cell cycle into different periods based on the parameter similarities. The results suggest that the whole cell cycle could be divided into three distinct periods (see Materials and Methods).

The results show that the cell’s instantaneous speed fluctuates abruptly in the first and third periods but maintains consistently low in the second period. The cell area is initially small but continuously increases in the first period, steadies in the second period, and disperses vigorously in the third period. In terms of cell shape, the monitored cell displays a high aspect ratio and low circularity in the first period but contrasting results in the other periods, demonstrating that the cell shape transforms after the second period.

To verify that these results do not emerge from coincidence, we acquired 20 single-fibroblast movies, 20 h long each, ensuring inclusion of at least one cell division event. We then identified a movie containing a complete cell cycle with two consecutive cell divisions and set its timeframe as a benchmark to align the timeframes of the rest. If the cell division occurred at an earlier time in an aligned movie, then the division time of this movie would be superimposed with the first division of the standard one; otherwise, the division time would be overlaid with the second division of the standard one (Fig. 1C). Under these conditions, all cells possessed relatively similar timeframes with respect to the cell cycle. Hence, we could analyze cell phenotypes through the least-square criterion again using these four cell parameters over a reference timeframe by much larger sample sizes (Fig. 1D). We compared the individual trends of the mean values of those four parameters with their counterparts shown in Fig. 1B, left panel, the results suggest that those parameters in the cell cycle might exist the similar trends that can be distinguished into three periods with different performances.

The presence of three distinct phenotypes within the cell cycle seems to correlate with the progression of the cell cycle. Therefore, we further synchronized NIH 3T3 fibroblasts into different stages of the cell cycle (Fig. 1E and Supplementary Fig. 1) and randomly acquired 25 single-cell movies in each of these stages to see whether the phenotype transitions are coincident with the cell cycle progression (Fig. 1F). The results show that the instantaneous speeds are the highest in the G1 phase, reach a minimum in the S phase, and bounce back slightly in the G2 phase. The cell area displays a monotonically increasing trend until the late G1 phase and reaches a plateau after entering the S phase. The aspect ratio and circularity together suggest that cells have the lowest polarity in the S phase. In essence, cell phenotype transitions exhibit a strong correlation with the cell cycle phases, and these results lay the assumption that cell cycle stages could be the functional subpopulation of heterogeneity with phenotypic disparity.

Overview: CN-correlation assessment

The above assessments have built a loose correlation between phenotype transitions and the cell cycle progression. However, traditional cell migration analyses prevent detailed investigation of cellular dynamic behaviours. Those techniques obligate a long period of cell trajectories (~ 10 h)^28,29, which leads to a dilemma where the duration of the sampled cell trajectories must be long enough for statistically meaningful analysis, yet short enough to have the trajectories restrained within the same stage of the cell cycle phases. In addition, conventional assessments also fail to provide the underlying mechanisms explaining the causes of their results. These deficiencies led us to develop a novel biological model, termed the CN correlation analysis²⁶, to statistically decode the cell dynamic process into contributing subcellular migratory activities by weights. This novel analysis only requires short time (~ one hour) cell movies as the raw data; hence, we can obtain the cell migratory details within a given stage of the cell cycle.

This method is described in brief: We identified each known cell locomotion event of the NIH 3T3 fibroblasts in single-cell movies that were recorded at one-minute intervals. Consequently, the cell centroid displacement between each pair of adjacent video frames of the cell locomotion movie was analyzed and denoted as a CCD, and the relative nuclear centroid displacement along the CCD direction was analyzed and denoted as the NCD_//. Then, the pair of CCD and NCD_// determined a point in the CCD vs. NCD_// coordinate system (the CN plot), called the CN correlation datum (Fig. 2A). Through this procedure, we found that data generated by each specific locomotion event form a normal distribution pattern around a mean (peak) polar-angle in the CN plot. These cell locomotion are the nucleus-forward movement, which has the peak polar angle ~ 25°; the trailing-edge detachment, ~ 45°; the simultaneous protrusion and detachment, ~ 65°; the leading-edge protrusion, ~ 90°; the large-angle side-protrusion, ~ 115°; and the leading-edge retraction, ~ 145°. We denote these peak polar angles as signature polar angles for cell locomotion (Fig. 2B).

Following, the CN correlation analysis was applied to an undesignated cell movement event to reveal the percentages of individual cell locomotion contributing to the event. We recorded twenty-five randomly selected single NIH 3T3 fibroblast for one hour at one-minute intervals as individual cell movies to decipher the 1500 CN correlation data, which were plotted together to form the CN profile²⁶. Then, the automatic univariate normal mixtures (UNM) algorithm was applied to the profile to statistically classify the histogram of the polar angles into appropriate normal distributions based on the signature polar angles to reveal the weight of each cell locomotion in the migratory event²⁷ (Fig. 2C). Hence, the CN profile is a statistical descriptor that can analyze the unique cell migration pattern. This information can be utilized to further connect the underlying signaling pathways involved in each cell locomotion.

Cell motility is carried out by effective migration, in which a cell and its coupled nucleus must move in the same direction. In the CN correlation analysis, every datum with a low polar angle (i.e., < 75°) possesses a nuclear displacement that effectively contributes to migration, while that with a high polar angle having an irrelevant NCD_//. Hence, only the CN correlation data located within the polar angles ranging from 0° to 75° are considered to contribute to the effective migration and the values of all NCD_// with a low polar angle can be summed together to robustly estimate the cell motility of the probed cells, called the cell migration potential index (CMPI) (Fig. 2D). Our previous study²⁶ has proved that a greater CMPI value indicates a greater long-term motility for a cell.

The CN correlation analysis possesses excellent applicability and reliability for cell migration assessment²⁶. When we incorporated cell polarity into the CN correlation analysis, it can evaluate the persistence of mesenchymal cell migration. The consistency of the short-term CCD orientations concerning the long-term cell displacement indicates that the probed cell possesses high polarity (Fig. 2E). Hence, the CN correlation analysis can decompose the cellular phenotypes into functional subcellular events and accordingly assess cell migration from both the motility and polarity perspectives.

Cell migration patterns are distinct and highly homogeneous in each stage of the cell cycle phase

To clarify the role of the cell cycle progression in driving cell phenotype transformation, we first assessed the cell dynamics that results in appearing cell morphology in every interphase stage of the cell cycle phases. We acquired cell images of the same cell under 3-min intervals and compared them through every 10° of cell boundaries to characterize the changes in cell morphology. The results reveal the dynamic status of the cells in each stage of the cell cycle phases, as shown in the representing cells (Fig. 3A). Cells in the early G1 phase are highly motile and move along a consistent direction; however, this consistency gradually diminishes during the late G1 phase. In the S phase, cells are relatively still with the peripheral ruffling in random directions. Interestingly, the cells in the G2 phase cell motion resurges by notable side-protrusions (i.e., these motions do not align along a fixed axis).

Consequently, we asked how these short-term cell dynamics transform into long-term cell motility. The CN correlation analysis^26,27 was applied to the movies of single NIH 3T3 fibroblasts, synchronized within each stage of the cell cycle phases, to systematically characterize their migration patterns (Fig. 3B & Table 1). This analysis not only specifies unique motion patterns by the cell cycle stages but also illustrates the evolution of phenotypic transition of cells throughout the cell cycle (Fig. 3C).

In the early G1 phase, cells exhibit a standard, highly polarized fibroblast behaviour, where detachment events take ~ 30% of the time and protrusion events take ~ 60% (Fig. 3D). In the late G1 phase, the cells deviate from the initial directional migratory mode and their polarity fades, judging by the increase of side protrusion events (from none to 5%) and the reduction of CCD and NCD_//. In the S phase, the detachment events drop significantly to 16%, while the protrusion events are dominant at 70%. Meanwhile, the sizes of CCD and NCD_// are both at a minimum, indicating the cells have lost their polarities and intend to remain stationary. Notably, upon reaching the G2 phase, the cells only perform pure protrusion 16% of the time. In contrast, 52% of the time the cells are predominantly occupied by simultaneous detachments and protrusions (also see Table 1). In addition, the cells also exhibit large-angle side protrusion 20% of the time. This irregular dynamic pattern is illustrated by the G2-cells shown in Fig. 3A, where the cell displays neither a polarized nor isotropic, but rather, an irregular shape.

We then estimated the cell motility in each stage of the cell cycle phases using CMPI values. The results suggest that the motility of the cells in the G2 phase is low, comparable to in the S phase (Fig. 3E). However, the instantaneous speeds (3-min CCD) of the cells in the G2 phase are relatively high and similar to those in the late G1 phase. Hence, the high instantaneous speeds are not translated to the long-term motility for the G2-phase cells, a typical phenomenon for poorly polarized phenotypes. This conclusion is further supported by the persistency evaluation (Fig. 3F), which shows that the CCD directions alter more frequently for the cells in the G2 phase than in the G1 phase. It is noteworthy that the CN correlation analysis is highly consistent²⁷. The convergence analysis here verified the homogeneity of cell migration patterns within each stage of the cell cycle phases (Fig. 3D & Supplementary Fig. 2). Since the cell morphology and dynamics are both governed by the stages of the cell cycle, these results strongly suggest that the heterogeneity of cell phenotype in the same cell type originates from the distribution of the cells in different stages of the cycle phases. In addition, the tight connection between the cell dynamics and the appealing morphology supports that the outcomes of cell dynamic measurements can present the phenotype heterogeneity.

The actin-cytoskeletal remodeling leads to the cell-cycle-dependent migration pattern

After every phase transition, the cells transformed into another morphology with new dynamics. As such, we continued evaluating how these dynamic processes occur. Cell dynamics are controlled by cytoskeletal remodeling. Migration-related subcellular activities, including protrusion, detachment, contraction, adhesion, etc., all stem from spatiotemporally assembling and disassembling of actin filaments^30,31. Hence, we examined the actin-cytoskeleton and cell adhesions of the canonical NIH 3T3 fibroblasts in different stages of the cell cycle phases through fluorescence microscopy (Fig. 4A). The representing micrographs show that the early G1-phase cells have an amorphous actin-network and sparsely distributed adhesion plaques, which hallmarks the motile pattern^32,33. In contrast, in the S and the G2 phase, cells form a defined cytoskeleton with highly organized stress fibers and dense-patched cell adhesions, signifying a stable, immotile pattern³⁴.

To study the cell-cycle-phase-dependent migration patterns further, we scrutinized the stress fibers (SFs) and focal adhesions (FAs) in detail. The number, size (area), and density of the SFs and FAs were measured from the corresponding cell images (Fig. 4B), providing insightful cytoskeletal evolution throughout the cell cycle (Fig. 4C). In the early stage of the G1 phase, small adhesion plaques are scattered over the whole cell bodies. These adhesion plaques are not matured and could be disassembled easily; thus, the cells are highly motile³⁵. In the S phase, the overall numbers and sizes of both FAs and SFs increase in a cell, but their densities drop over the enlarged cell body, indicating that the adhesion plaques are merged into matured patches and the extended SFs become stable, which lead to immotile cells. Even though the G2-phase cells continuously possess thick SFs and matured FAs from the S phase, the density of FAs is significantly reduced. This change inspires the speculation for the occurrence of focal adhesion turnover.

To clarify whether focal adhesion turnover had occurred, we analyzed individual cells to normalize their cell shape and size dependence and statistically characterized the FAs within the cells (Fig. 4D). We utilized the polar coordinate system, set the origin at each cell’s centroid, the major axis of the cell in the direction of the leading edge as polar angle 0˚ for each cell, and the radii as the normalized distances from the cell centroid to the radiated cell edges. Following this format, FA distributions in different cells were overlaid on a rose plot with a colormap representing the density (Fig. 4E). The rose plots demonstrate that cells in the S phase possess the highest density of adhesions. Therefore, the comparison of local adhesion densities between cells in the S and the G2 phase supports the idea that the FAs diminish in certain regions of the G2 cells and focal adhesion turnover occurs.

The occurrence of focal adhesion turnover provides a rational justification for the 52% of prevalent “unconventional” cell motion and 20% of side protrusion, deciphered by the CN correlation analysis. As a result, parts of the cell periphery are freed up so that pseudopods can form and give rise to an irregular morphology and limited local migration. This type of migration shifts the cells locally in a random direction and is highly dynamic with the cell instantaneously speed comparable to the cells in the G1 phase.

To study how the cytoskeletal features impact cell polarity over the cell cycle, we also analyzed the distributions of FAs and SFs through their polar histograms in different stages of the cell cycle phases (Fig. 4F). After the polar histograms were determined through the rose plots with 10˚ resolution, the data showed that both distributions are anisotropic in all cell cycle phases. The degree of isotropy can be determined by the aspect ratio (AR) of the histograms, where a greater value indicates that the probed cytoskeletal features are distributed with more bias, leading to a strong polarity. The results reveal the G1-phase cells have the highest AR value, followed by the G2-phase cells, then the S-phase cells with the lowest AR value. Thus, the polar histogram analysis restates that cells lose polarity when approaching the S phase, and regain some polarity in the G2 phase through focal adhesion turnover. Since cytoskeleton remodeling is the driving machinery of cell morphological dynamics, our results reiterate the postulation that a good portion of cell heterogeneity originates from the subpopulations of cells distribute in different stages of the cell cycle phases.

Two CDKIs—p27^Kip1 and p21^Cip1 guide the RhoA and Rac1 signaling in migration evolutions

The alteration of cell phenotypes across cell cycle stages should be inexorable from a molecular-level perspective. Under such a condition, there must be certain molecules that connect the cell cycle to cell dynamics beyond Rho GTPases, the primary molecular switches for cytoskeletal remodeling¹⁰. These molecules must be associated with regulators of the cell cycle, the cyclin-dependent kinases (CDKs)^6,7. To our best knowledge, the only known molecules directly interact with both Rho GTPases and the CDKs are two CDKIs: p27^Kip1 and p21^Cip1. After cells entering the S phase, p27^Kip1 is exported from the nucleus to the cytoplasmic region to bind and inhibit RhoA, while p21^Cip1 binds and downregulates the ROCK^13,14.

To observe whether their spatiotemporal appearances agree with the activity profiles of RhoA and Rho kinase (ROCK) throughout the interphase of the cell cycle, we monitored the distributions of p27^Kip1 and p21^Cip1 in cells of different cell cycle phases. p27^Kip1 and p21^Cip1 were immuno-stained in the synchronized NIH 3T3 fibroblasts of different cell cycle stages to quantitatively determine their cytoplasmic concentration profiles (Fig. 5A). The results shows that the cytoplasmic concentration of p27^Kip1 elevates along the cell cycle progression, while cytoplasmic p21^Cip1 exhibits a mild concave evolution with the lowest concentration appearing in the S phase (Fig. 5B). Consequently, we conducted quantitative fluorescence microscopy to determine the activity profiles of RhoA and Rac1, the regulators of actomyosin contraction and membrane protrusion, respectively³⁶ (Fig. 5C). The results show that the RhoA activity monotonically drops throughout the interphase of the cell cycle; whereas the Rac1 activity slightly varies from the early G1 phase to the S phase, but elevates significantly to a maximum in the G2 phase (Fig. 5D). The trends of these profiles are supported through the RhoA and Rac1 pull-down/Western blot assessments (Supplementary Fig. 4). We also conducted Western blotting against phosphorylated myosin phosphatase target subunit 1 (MYPT1, Thr696), which indicates the ROCK activity, to determine the activity profile of ROCK in the interphase of the cell cycle (Fig. 5E). The results show that the ROCK activity significantly reduces after entering the G2 phase, echoing the p21^Cip1 increasing in the cytoplasmic region and the diminishing of RhoA activity.

The comparison of these profiles shows that a spatiotemporal agreement for the intermediate roles of p27^Kip1 and p21^Cip1 in the crossroad between the cell cycle and the cell movement. The increase of cytoplasmic p27^Kip1 and p21^Cip1 in cells from the S phase to the G2 phase is correlated with the decrease of RhoA and ROCK activities, respectively. Since the inhibition of the RhoA pathway leads to the disassembly of focal adhesion³⁷, the weakening of RhoA pathway activity in the G2-phase cells explains the concurrence of focal adhesion turnover as we observed. In addition, the depleting of RhoA pathway activity only occurs after the cells entering the S phase where the p27^Kip1 and p21^Cip1 are released from the CDK-cyclin complex and become cytoplasmic^{13,14,15,16,38}.

The brief cell-cycle-phase-dependent cell morphology model

Immediately after cytokinesis, a newly divided cell enters the early G1 phase and spreads on the extracellular matrix (ECM), the cell explores its microenvironment via lamellipodia, which are formed upon the Rac1 activity. In this period, the cell is also polarized along with the ECM that offers dense anchorage sites. Meanwhile, in the presence of mitogen, RhoA is activated, further promoting focal adhesions maturation in the polarized sites^36,39,40. Eventually, the formation of dense, stable focal adhesions causes the hindrance of cell movement and the cell reaches the late G1 phase. During this time, the migration retention leads to prolonged substrate contacts around the cell periphery, inducing lateral protrusions through the activation of proximal Rac1 activity. When the cell periphery is fully occupied by stable adhesions, the cell becomes stationary and enters the S phase. Starting around the same time, the presence of cytoplasmic CDKIs gradually diminishes the RhoA pathway activity, eventually causing focal adhesion turnover in the G2 phase. Hence, some parts of the cell periphery are freed up again and available for new cell-substrate contact developments, allowing the cell to form pseudopod-like features through re-surged Rac1 activity. This cytoskeletal remodelling alters cell morphology to be a form suitable for mitosis. Together, the evolution of cell phenotype is highly connected to the cell cycle progression that renders the fact that cell cycle stages can be considered as subpopulations of cell heterogeneity with distinct functional significance (Fig. 5F).