Promoter Architecture as a Design Principle for Buffering Transcriptional Noise and Diversifying Expression Patterns

Gene expression is inherently stochastic, and transcription initiation is a key source of variability across cells. While classical promoter models often assume linear state transitions, emerging evidence suggests more flexible promoter architectures. Here we introduce a generalized cyclic promoter model and compare it with the standard linear model using exact analytical solutions for initiation-time and nascent RNA distributions. Our results reveal that linear promoters produce only monotonic initiation-time statistics and a limited set of RNA expression patterns, whereas cyclic promoters generate non-monotonic initiation-time distributions and richer RNA profiles, including multimodal cases not achievable with linear architectures. We further show that cyclic promoters consistently buffer variability in initiation timing and RNA output, providing tighter control over transcriptional noise. Within the cyclic model, the number of exit pathways serves as a tunable parameter that shifts distributions from bimodal to unimodal and reduces noise, offering a potential mechanism for balancing robustness with flexibility in gene regulation. This framework highlights promoter topology as a critical determinant of transcriptional heterogeneity, bridges initiation dynamics with RNA-level variability, and generates testable predictions that can guide single-cell experiments probing promoter structure.