Circadian rhythm disruption is linked to various diseases, including cancer. The circadian clock regulates cancer hallmarks, suggesting that treatment timing could be personalized. Colorectal cancer (CRC) is a significant health concern, and chemotherapy, while effective, often causes severe side effects. Chronotherapy aims to mitigate these side effects by aligning drug administration with a patient's circadian rhythm. Previous studies have shown improved treatment outcomes by considering drug administration time. This study builds upon a previous model, refining it to fit data from three CRC cell lines (HCT116, SW480, SW620) and core-clock knockouts. The improved model integrates external Zeitgebers to potentially fine-tune toxicity and facilitate time-dependent treatment in clinical practice by aligning endogenous rhythms with clinic routines.

The introduction extensively reviews the connection between circadian rhythms and cancer, highlighting the potential of chronotherapy. It cites studies demonstrating the impact of circadian rhythms on drug metabolism and the success of chronotherapy in improving cancer treatment outcomes, particularly with irinotecan. The authors mention their previous work establishing a mathematical model linking circadian gene expression with irinotecan toxicity. The literature review underscores the need for a more comprehensive model that accounts for various cell types and external factors influencing circadian rhythms.

The study utilized RNA-seq data from three colorectal cancer cell lines (HCT116 wild-type and three core-clock knockouts: PER2^KO, NR1D1^KO, ARNTL^KO) and microarray data from SW480 and SW620. A mathematical model was developed, integrating a transcription-translation network for circadian gene expression with a pharmacokinetics and pharmacodynamics (PK-PD) model for irinotecan. The model was refined from a previous version by Hesse et al. (2021), incorporating new network connections and biologically motivated temporal effects of treatment on gene expression, such as an increase in UGT1A1 and a transient increase in apoptosis rate. The model was fitted to the mRNA expression data using the CMA-ES evolutionary algorithm, with a cost function based on the squared error between data and model simulation. LASSO regularization was employed to minimize parameter divergence between knockout and wild-type cell lines, identifying parameters that significantly differed between the conditions. For the PK-PD model, the parameters of the equations for the number of living and dead cells, and the cell death modulation were fitted to cytotoxicity data. The model was further extended to include the influence of external Zeitgebers, such as light pulses, on the predicted toxicity profile. Light exposure was implemented as a transient increase in the maximal expression rate of PER, and the effects of varying pulse duration, strength, and timing were investigated. The effects of NR1D1 pulses were also explored.

The refined model successfully fitted data from all six cell lines, showing improved accuracy compared to the previous model (R² increased from 0.23 to 0.29). Paralog compensation was observed for PER2 and NR1D1, highlighting the importance of modeling the sum of paralogous gene expression for personalized networks. LASSO regularization effectively reduced parameter divergence between knockout and wild-type cell lines, indicating that the knockouts primarily affected a small subset of parameters. The model predicted differences in toxicity profiles between HCT116 and SW480 cell lines, attributable to variations in irinotecan metabolism genes (UGT1A1, CES2, ABC transporters). The inclusion of external Zeitgebers demonstrated that light pulses could shift the circadian phase and consequently the timing of maximal toxicity, with the magnitude of the shift depending on pulse timing, strength, and duration. Pulses in NR1D1 also induced phase shifts, though the optimal timing for phase shifts differed from that of PER pulses. The model predicted differences in toxicity profile shifts between HCT116 WT and KO cell lines in response to light pulses.

The study demonstrates the feasibility of personalizing cancer treatment timing using a mathematical model that integrates circadian gene expression and drug metabolism. The findings highlight the importance of considering paralog compensation and using regularization techniques for fitting such models to diverse cell lines. The model's ability to predict toxicity profiles and their modification by Zeitgebers suggests its potential for optimizing treatment schedules. The observed differences in toxicity profile shifts between HCT116 WT and KO cell lines suggest that patient-specific circadian profiles significantly impact the effects of chronotherapy. The study's success in incorporating external Zeitgebers demonstrates the potential of interventions such as light therapy to align a patient's circadian rhythm with an optimal treatment schedule.

This study presents a significantly improved mathematical model for predicting irinotecan toxicity in CRC cells, incorporating circadian rhythms and the effects of external Zeitgebers. The model's ability to accurately fit diverse cell lines and predict the impact of Zeitgebers on toxicity profiles underscores its potential for personalizing cancer treatment timing. Future research could expand the model to include metabolic dynamics and explore its application in clinical settings to optimize treatment strategies and minimize side effects.

The study uses in vitro cell lines, which may not fully reflect the complexity of in vivo human responses. The model's accuracy relies on the quality of the input data, and the generalizability of the findings may be limited by the specific cell lines studied. While the model incorporates external Zeitgebers, the implementation of light as a transient increase in PER expression might be simplified and could benefit from a more refined representation of light perception and its downstream effects on the circadian clock.