ceRNA crosstalk stabilizes protein expression and affects the correlation pattern of interacting proteins

Abstract

Gene expression is a noisy process and several mechanisms, both transcriptional and post-transcriptional, can stabilize protein levels in cells. Much work has focused on the role of miRNAs, showing in particular that miRNA-mediated regulation can buffer expression noise for lowly expressed genes. Here, using in silico simulations and mathematical modeling, we demonstrate that miRNAs can exert a much broader influence on protein levels by orchestrating competition-induced crosstalk between mRNAs. Most notably, we find that miRNA-mediated cross-talk (i) can stabilize protein levels across the full range of gene expression rates, and (ii) modifies the correlation pattern of co-regulated interacting proteins, changing the sign of correlations from negative to positive. The latter feature may constitute a potentially robust signature of the existence of RNA crosstalk induced by endogenous competition for miRNAs in standard cellular conditions.

Figure 1. Model schematics and main parameters.

A single miRNA species negatively controls the expression of the ceRNA species ceRNA_T (‘target’, level m_T) and ceRNA_C (‘competitor’, level m_C), to which it can bind with rates and , respectively. The functional products of the ceRNAs, proteins p_T and p_C, can eventually interact to form a complex C_P. The key control parameter is the target’s transcription rate b_T. See Fig. 7 for a detailed scheme that includes all processes.

Figure 2. Dependence of mean protein expression levels and relative fluctuations (CV) on the transcription rate b_T of the target.

Panels (B) and (C) describe the case of a simple miRNA-regulated target, shown in panel (A). In (B), increases in the direction of the arrow (specifically, for orange, purple, blue, black curves respectively). Panels (E) and (F) describe the case of a target regulated through ceRNA competition, depicted in (D), for and . Note that no PPI is considered in this case.

Figure 3. ceRNA competition can stabilize highly expressed proteins.

(A) Coefficient of variation of p_T as a function of the mean protein level for a post-transcriptionally unregulated protein (black line, and ), a miRNA-regulated protein (red line, and ) and a ceRNA-regulated protein (blue line, and ). (B) Capacity of the target’s expression channel as a function of the miRNA-competitor interaction strength. Color code same as in panel A. (C) Derepression size Δ_C of the competitor as a function of the miRNA-competitor interaction strength in the case of ceRNA regulation (same parameters as panel B).

Figure 4. Noise reduction in miRNA-mediated circuits due to the effective miRNA-recycling at the target node.

Fast and slow (high and low κ_T, see Methods) miRNA recycling scenarios at the target node are shown, respectively, by dashed and solid lines for a miRNA-regulated protein (red, and ) and a ceRNA-regulated protein (blue, and ). The black curve describes the case of an unregulated target ( and ).

Figure 5. Correlation patterns for interacting proteins measured by the Pearson coefficient ρ as a function of the mean target level.

(A) Case of non-interacting proteins translated from competing transcripts ( and ). (B) Case of interacting proteins translated from post-transcriptionally unregulated transcripts (). (C) Case of interacting proteins translated from competing transcripts ( and ). Lines (blue and orange) correspond to analytical results obtained by the Linear Noise Approximation (see Methods), markers (black and purple) to results from stochastic simulations by the Gillespie algorithm.

Figure 6. Capacity of the protein complex synthesis channel, I_max(C_p, b_T), as a function of the miRNA-ceRNA binding strengths of the target () and the competitor ().

To allow for comparisons, simulations were performed at fixed output variation range (with ) and variable b_T.

Figure 7. Schematic representation of the model.

Two proteins (p_T and p_C) translated from 2 distinct ceRNAs (m_T and m_T) that are regulated by the same miRNA (μ). Proteins p_T and p_C associate and dissociate to a protein complex C_p with a rate k⁺ and k⁻ respectively, miRNA binds (unbinds) to the ceRNAs m_T and m_C with the rates and respectively forming ceRNA:miRNA complexes c_T and c_C. Species m_T, m_C, μ, p_T, p_C are synthesized (degraded) with the rates b_T, b_C, β, α_T, α_C (d_T, d_C, δ, δ_T, δ_C) correspondingly. Protein complex C_p undergoes spontaneous degradation with a rate δ_p. Finally, c_T and c_C decay catalytically (ceRNA cleavage and miRNA recycling) with the rates κ_T and κ_C respectively. Figure-specific values of the kinetic parameters are reported in Table 1.