Computational modeling establishes mechanotransduction as a potent modulator of the mammalian circadian clock

Abstract

Mechanotransduction, which is the integration of mechanical signals from the external environment of a cell to changes in intracellular signaling, governs many cellular functions. Recent studies have shown that the mechanical state of the cell is also coupled to the cellular circadian clock. To investigate possible interactions between circadian rhythms and cellular mechanotransduction, we have developed a computational model that integrates the two pathways. We postulated that translocation of the transcriptional regulators MRTF (herein referring to both MRTF-A and MRTF-B), YAP and TAZ (also known as YAP1 and WWTR1, respectively; collectively denoted YAP/TAZ) into the nucleus leads to altered expression of circadian proteins. Simulations from our model predict that lower levels of cytoskeletal activity are associated with longer circadian oscillation periods and higher oscillation amplitudes, which is consistent with recent experimental observations. Furthermore, accumulation of YAP/TAZ and MRTF in the nucleus causes circadian oscillations to decay in our model. These effects hold both at the single-cell level and within a population-level framework. Finally, we investigated the effects of mutations in YAP or lamin A, the latter of which result in a class of diseases known as laminopathies. In silico, oscillations in circadian proteins are substantially weaker in populations of cells with mutations in YAP or lamin A, suggesting that defects in mechanotransduction can disrupt the circadian clock in certain disease states; however, reducing substrate stiffness in the model restores normal oscillatory behavior, suggesting a possible compensatory mechanism. Thus, our study identifies that mechanotransduction could be a potent modulatory cue for cellular clocks and that this crosstalk can be leveraged to rescue the circadian clock in disease states.

Keywords: Circadian clock; MRTF; Mechanotransduction; Systems biophysics; YAP/TAZ.

Fig. 1.

A coupled model of mechanotransduction and the circadian clock. (A) Schematic of pathways for YAP/TAZ and MRTF-mediated mechanotransduction and the mammalian circadian clock. Mechanical stimuli such as substrate stiffness and cell–substrate adhesion induce changes in cytoskeletal activity that lead to nuclear translocation of YAP/TAZ and MRTF. Cell–cell contacts also inhibit YAP/TAZ activation via a LATS-dependent pathway. Nuclear YAP/TAZ and MRTF are posited to induce changes in the expression of BMAL1, PER, CRY and REV-ERBα proteins, altering circadian oscillations. In parallel, increases in nuclear BMAL1 enhance transcription of genes encoding PER, CRY and REV-ERBα, while increases in nuclear REV-ERBα inhibit transcription of Bmal1 and increases in nuclear PER/CRY inhibit transcription of genes encoding PER, CRY and REV-ERBα. Dashed lines indicate delays associated with transcription, translation, post-translational modifications and nuclear translocation of BMAL1, PER/CRY and REV-ERBα. The schematic was created with Biorender.com. (B) Separation of timescales from mechanotransduction pathways to circadian oscillations. Events downstream of mechanical stimuli lead to changes in YAP/TAZ and MRTF nuclear concentrations within tens of minutes (lower plot), whereas changes in the concentration of circadian proteins extend over hours to days (upper plot). (C) Simulated dynamics of single-cell circadian oscillations. BMAL1, PER/CRY and REV-ERBα concentrations all oscillate with periods close to 1 day, as shown in the upper plot. Rates of expression of BMAL1 and PER/CRY are depicted in the lower plots. Conc., concentration; cyto, cytosolic; expr., expression; nucl., nuclear.

Fig. 2.

Bayesian parameter estimation for the mechanotransduction–circadian model. (A–E) Predicted oscillation period (top) and amplitude (bottom) in five experiments from Bayesian parameter estimation. Experimental data from Xiong et al. (2022) and model values are plotted for the effects of substrate stiffness (A), ROCK inhibitor (Y27632) treatment (B), cytochalasin D treatment (C), latrunculin B treatment (D) and jasplakinolide treatment (E). Experimental error bars denote s.d. of n=4 for each substrate stiffness, n=3 for all other treatment conditions. For the model, the darker blue region indicates interquartile range and the lighter blue region spans the 95% credible interval (region from the 2.5 percentile to 97.5 percentile), based on sampling from the posterior distributions for parameter values. To compute the estimates in this figure, 1000 samples were generated for each sequence of tests. In all cases, amplitude is normalized to that associated with the control condition (untreated cells on glass).

Fig. 3.

Changes to circadian oscillations due to altered cytoskeletal activity. (A) Substrate-stiffness-dependent changes in actin, cytosolic stiffness (E_cyto), the YAP/TAZ nuclear-to-cytosolic (N/C) ratio and the MRTF N/C ratio. (B) Substrate-stiffness-dependent changes to oscillations in nuclear PER/CRY. (C) Cytochalasin D-induced changes to actin, cytosolic stiffness, and N/C ratios of YAP/TAZ and MRTF. Key as shown in A. (D) Cytochalasin D (CytD)-induced changes to oscillations in nuclear PER/CRY. (E) Jasplakinolide-induced changes to actin, cytosolic stiffness, and N/C ratios of YAP/TAZ and MRTF. F-actin increases dramatically upon jasplakinolide treatment, leading to substantial nuclear accumulation of YAP/TAZ and MRTF. Key as shown in A. (F) Jasplakinolide (Jas)-induced changes to oscillations in nuclear PER/CRY. Increased actin polymerization leads to a decrease in the oscillation period along with a decay in circadian oscillations (inset). (G) Bifurcation diagram showing the location of the Hopf bifurcation and the dependence of the circadian oscillation period on nuclear YAP/TAZ and MRTF. All points below the Hopf bifurcation curve correspond to sustained oscillations, whereas points above exhibit decaying oscillations. Individual markers plotted in the YAP/TAZ–MRTF phase plane correspond to the test cases shown in panels A–F. Conc., concentration.

Fig. 4.

Behavior of circadian oscillations in model cell populations with variability. (A) Kymograph depicting oscillations in nuclear REV-ERBα for 200 model cells on a soft (0.1 kPa) substrate. Cells exhibit consistent circadian oscillations, and variability is observed at the level of the population. (B) Kymograph depicting oscillations in nuclear REV-ERBα for 200 model cells on glass (10 GPa). Most cells show consistent, lower magnitude oscillations; in some cases, oscillations are weaker or nonexistent. (C,D) Distribution of oscillation period (C) and circadian power fraction (D) for cell populations (200 model cells each) on different substrate stiffnesses, as indicated. Violin plots show the distribution of data, with the central points and error bars marking the median and interquartile range, respectively. Compact letter display is used to denote statistical significance, where groups sharing a letter are statistically similar according to one-way ANOVA followed by Tukey's post hoc test with a significance threshold of P=0.05. Violin plots were generated using Violinplot in MATLAB (https://github.com/bastibe/Violinplot-Matlab).

Fig. 5.

Correlation between circadian power fraction and the N/C ratios of YAP/TAZ or MRTF. (A,B) Populations of model cells were subjected to the treatments tested by Abenza et al. (2023), and circadian power fraction was plotted against YAP/TAZ N/C ratio (A) and MRTF N/C ratio (B). Data points denote median values and error bars denote the range from the 40th to the 60th percentile for 200 model cells. In each case, Pearson's correlation coefficient (r) was assessed for the correlation between median power fraction and median N/C ratio on the log scale. Values of r and the P-value associated with the null hypothesis, r=0, are included on each graph. To generate distinguishable colors for each condition, we used the linspecer tool in MATLAB (https://www.mathworks.com/matlabcentral/fileexchange/42673-beautiful-and-distinguishable-line-colors-colormap). CytD, cytochalasin D; LatA, latrunculin A.

Fig. 6.

Effects of YAP or lamin A mutations on circadian oscillations. (A,B) Nuclear concentrations (conc.) of YAP/TAZ (A) and MRTF (B) compared across a wild-type (WT) population, YAP mutant population and LMNA mutant population, all on 30 kPa substrates. (C) Kymographs depicting the dynamics of nuclear REV-ERBα for a wild-type population, YAP mutant population and LMNA mutant population, all on 30 kPa substrates. Each plot shows 200 cells over 3 days. (D) Circadian power fraction compared across wild-type, YAP mutant and LMNA mutant cells on 30 kPa substrates. (E) Rescue of normal circadian oscillations (circadian power fraction) in YAP mutant and LMNA mutant cells via reduction of substrate stiffness, as indicated. Violin plots in A, B, D and E show the distribution of data, with the central points and error bars marking the median and interquartile range, respectively, for populations of 200 cells each. Compact letter display is used to denote statistical significance, where groups sharing a letter are statistically similar according to one-way ANOVA followed by Tukey's post hoc test with a significance threshold of P=0.05. Comparisons in D and E were conducted across all five groups. Violin plots were generated using Violinplot in MATLAB (https://github.com/bastibe/Violinplot-Matlab). (F) Summary of the causal flow from cell mechanotransduction to changes in the circadian clock. Upper colored boxes denote disruptions to mechanotransduction explored at different points in this article, either due to cytoskeleton-targeting drugs or mutations in YAP or LMNA. The lower box depicts the rescue of normal circadian oscillations in mutant cells via a reduction in substrate stiffness. Dashed arrows indicate possible feedback from the circadian clock to cell and tissue mechanics.