Leveraging chromatin packing domains to target chemoevasion in vivo

Abstract

Cancer cells exhibit a remarkable resilience to cytotoxic stress, often adapting through transcriptional changes linked to alterations in chromatin structure. In several types of cancer, these adaptations involve epigenetic modifications and restructuring of topologically associating domains (TADs). However, the underlying principles by which chromatin architecture facilitates such adaptability across different cancers remain poorly understood. To investigate the role of chromatin in this process, we developed a physics-based mechanistic model that connects chromatin organization to cell fate decisions, specifically survival following chemotherapy. Our model builds on the observation that chromatin forms packing domains, which influence transcriptional efficiency through macromolecular crowding. The model accurately predicts chemoevasion in vitro, suggesting that changes in packing domains affect the likelihood of survival. Consistent results across diverse cancer types indicate that the model captures fundamental principles of chromatin-mediated adaptation, independent of the specific cancer or chemotherapy mechanisms involved. Based on these insights, we hypothesized that compounds capable of modulating packing domains, termed Transcriptional Plasticity Regulators (TPRs), could prevent cellular adaptation to chemotherapy. Using live-cell chromatin imaging, we conducted a compound screen that identified several TPRs which synergistically enhanced chemotherapy-induced cell death. The most effective TPR significantly improved therapeutic outcomes in a patient-derived xenograft (PDX) model of ovarian cancer. These findings underscore the central role of chromatin in cellular adaptation to cytotoxic stress and present a novel framework for enhancing cancer therapies, with broad potential across multiple cancer types.

Keywords: Biophysics; Cancer; Chemotherapy; Chromatin; Plasticity.

Fig. 1.

Linking chromatin to transcriptional plasticity and cell survival under cytotoxic stress using the Chromatin Dependent Adaptability (CDA) model. (A) Schematic illustrating the differential responses of cells with varying nuclear chromatin organization $D_n$ to cytotoxic stress. PWS microscopy images show representative low and high $D_n$ cells. The probability distribution function (PDF) of gene upregulation ($x=N_2/N_1$) is shown for both cell types, demonstrating increased mean and standard deviation of upregulated transcripts in high $D_n$ cells, leading to a higher survival probability. (B) Quantitative relationship between $D_n$, transcriptional malleability ($k$, purple), and transcriptional heterogeneity ($s$, green) for cell $b$ with fixed parameters $\mathrm{l}\mathrm{n}(E/\bar{E})=0$, ${\beta }_a=10$, and $t=7$ hours. (C) 3D plot showing the effect of transcriptional reactant (TR) concentration and local crowding (${\varphi }_{\text{in}}$) on the relative amount of mRNA produced. (D) Cell death probability (${\mathrm{\Theta }}_b$) as a function of $D_n$ for varying upregulation thresholds ($x_{\text{crit}}$) with fixed parameters $\mathrm{l}\mathrm{n}(E/\bar{E})=-2$, ${\beta }_a=10$, and $t=7$ hours. Inset shows individual curves for select $x_{\text{crit}}$ values. (E) Cell death probability $(\mathrm{\Theta })$ as a function of $D_n$ for different transcriptional amplification levels $\left({\beta }_a\right)$ with fixed parameters $\mathrm{l}\mathrm{n}(E/\bar{E})=-2$, $x_{\text{crit}}=5$, and $t=7$ hours. Inset displays individual curves for select ${\beta }_a$ values.

Fig. 2.

Chemotherapy induces alterations in chromatin across diverse cancer cell lines and treatment modalities. (A) Violin plots showing the distribution of $D_n$ in HCT116 cells over a 48-hour treatment with 15 μM oxaliplatin. The control population remains stable, while treated cells show significantly higher $D_n$ at 24 hours $\left(P<{10}^{-15}\right)$ and 48 hours $\left(P<{10}^{-32}\right.$). Sample sizes: $n=70-150$ cells per condition. (B) Scatter plot of changes in $D_n$ in individual HCT116 cell clusters after 48 hours of oxaliplatin treatment. Points represent average $D_n$ change per cluster vs. initial $D_n$ at 0 hours ($D_{n,0}$). Initial cluster size ranged from 2 to 5 cells, while final size ranged from 1 to 12 cells. Error bars represent standard error of the mean. (C) Violin plots showing $D_n$ distribution in surviving cells after 48-hour chemotherapy exposure across multiple cancer cell lines. Treatments include A2780, A2780.m248, MDA-MB-231 (vehicle, 5-fluorouracil, paclitaxel, oxaliplatin); HCT116 (vehicle, oxaliplatin); MES-SA, MES-SA/MX2 (vehicle, docetaxel, gemcitabine). Significance levels: *P<0.05, **P<0.01, ***P<0.001 (t-test against control, unpaired, unequal variance). (D) Representative PWS microscopy images of control and treated cells after 48-hour treatments: A2780 (0.5 μM 5-fluorouracil), A2780.m248 (5 nM paclitaxel), HCT116 (15 μM oxaliplatin), MDA-MB-231 (0.5 μM 5-fluorouracil), MES-SA (50 nM gemcitabine), and MES-SA/MX2 (5 nM docetaxel). Pseudocolor indicates that brighter red corresponds to higher $D_{\text{pixel}}$. Scale bars represent 15 μm.

Fig. 3.

Model predictions of population-level chromatin dynamics, cell survival, and chemotherapy efficacy over time closely match experimental results. (A) Probability density functions (PDFs) of $D_n$ in a population of HCT116 cell clusters at 0 (purple), 24 (blue), and 48 (teal) hours after treatment with 15 μM oxaliplatin, showing a progressive shift toward higher $D_n$ values over time. (B) Comparison of experimental data (blue) and model predictions (purple) showing the increase in mean $D_n$ of oxaliplatin-treated HCT116 clusters over 48 hours. Error bars represent standard error of the mean for experimental data and propagated error for model predictions. (C) Cell death probability (Θ) as a function of $D_n$, with experimental data (blue points) derived from tracking HCT116 cell clusters over 48 hours of oxaliplatin treatment. The CDA model fit (purple line) to experimental data was optimized using three free parameters, resulting in a mean squared error (MSE) of 0.012. (D) Effective inhibition rate (EIR) per day, representing cumulative cell death from oxaliplatin treatment over 48 hours, as a function of mean cluster $D_n$. Both experimental data (blue points) and model predictions (purple line) are shown, with error bars derived from error propagation.

Fig. 4.

Transcriptional Plasticity Regulators (TPRs) modulate chromatin structure to enhance chemotherapeutic efficacy in vitro. (A) Schematic depicting differential survival in cancer cell populations with distinct $D_n$ distributions under chemotherapy. Magenta and green curves represent control and TPR-treated populations, respectively. Dashed grey line indicates survival probability ($1-\mathrm{\Theta }$). (B) TPR candidate drug screen in live A2780 cells, showing $D_n$ distributions for control and potential TPR treatments. Dashed red line denotes $D_n$ threshold below which $\mathrm{\Theta }>0.999$. **P<0.01, ***P<0.001 (t-test against control, unpaired, unequal variance). (C) Representative PWS microscopy images of control and celecoxib-treated A2780 cells. Pseudocolor indicates $D_n$, with brighter red corresponding to higher values. Scale bars: 15 μm. (D) Correlation between TPR-induced chromatin change (TPR index) and increased cell death (inhibition index) upon combined chemotherapy and TPR treatment, showing that a reduction in $D_n$ correlates with enhanced efficacy. The relationship follows $y=2.0844{\mathrm{e}}^{2.0107x}$, with $R^2=0.984$. (E) Comparison of cell death in A2780 cells treated with paclitaxel alone (${\mathrm{\Theta }}_{\text{Pac}}$) versus paclitaxel combined with celecoxib normalized by celecoxib alone (${\mathrm{\Theta }}_{\text{Combo-Cele}}$). Blue dots represent experimental data with error bars as standard error of the mean. Solid purple line indicates model prediction using exact equation (Eq. 11) with the parameters $\mathrm{l}\mathrm{n}(E/\bar{E})\approx -3$, ${\beta }_a\approx 6$, and mean squared error (MSE) = 0.031. Dashed purple line is the prediction using approximate equation (Eq. 12).

Fig. 5.

In vivo validation of TPR-enhanced chemotherapy efficacy using patient-derived xenograft (PDX) models of ovarian cancer. (A) Schematic illustrating the treatment regimen for ovarian PDX studies and predicted outcomes, highlighting expected tumor growth patterns under different treatment conditions over 30 days. (B) Growth curves of ovarian carcinoma PDX tumors under various treatments. Co-treatment with celecoxib (25 mg/kg) and paclitaxel (1.7 mg/kg) resulted in minimal growth over 30 days compared to monotherapy with paclitaxel (1.7 mg/kg), celecoxib (25 mg/kg), or vehicle (DMSO). Animals were treated orally with celecoxib and intraperitoneally with paclitaxel daily for one week. Points represent mean tumor volume normalized by the volume at day 0 ± standard error of the mean, while lines indicate linear regression fits. (C) Effective Inhibition Rate (EIR) for paclitaxel alone and the paclitaxel + celecoxib combination over time, calculated using vehicle as a reference for paclitaxel and celecoxib as a reference for the combination. EIR was calculated using the equation $\mathrm{E}\mathrm{I}\mathrm{R}=\left[\mathrm{l}\mathrm{n}\left(V_{t,\text{control}}/V_{0,\text{control}}\right)-\mathrm{l}\mathrm{n}\left(V_{t,\text{treated}}/V_{0,\text{treated}}\right)\right]/t$. Points show mean EIR ± standard error of the mean, with lines representing linear regression fits. (D) Normalized tumor growth rate over time for paclitaxel alone and the paclitaxel + celecoxib combination. Points indicate experimental data (mean ± standard error of the mean). Solid lines show best fits using the adaptive inhibition model (Eq. 13).