A combination of transcriptional and microRNA regulation improves the stability of the relative concentrations of target genes

Abstract

It is well known that, under suitable conditions, microRNAs are able to fine tune the relative concentration of their targets to any desired value. We show that this function is particularly effective when one of the targets is a Transcription Factor (TF) which regulates the other targets. This combination defines a new class of feed-forward loops (FFLs) in which the microRNA plays the role of master regulator. Using both deterministic and stochastic equations, we show that these FFLs are indeed able not only to fine-tune the TF/target ratio to any desired value as a function of the miRNA concentration but also, thanks to the peculiar topology of the circuit, to ensure the stability of this ratio against stochastic fluctuations. These two effects are due to the interplay between the direct transcriptional regulation and the indirect TF/Target interaction due to competition of TF and target for miRNA binding (the so called "sponge effect"). We then perform a genome wide search of these FFLs in the human regulatory network and show that they are characterized by a very peculiar enrichment pattern. In particular, they are strongly enriched in all the situations in which the TF and its target have to be precisely kept at the same concentration notwithstanding the environmental noise. As an example we discuss the FFL involving E2F1 as Transcription Factor, RB1 as target and miR-17 family as master regulator. These FFLs ensure a tight control of the E2F/RB ratio which in turns ensures the stability of the transition from the G0/G1 to the S phase in quiescent cells.

Figure 1. A. Schematic description of the circuits discussed in the paper.

NM1: direct regulation; NM2: open motif in which the microRNA regulates only the transcription factor; NM3: open motif in which the microRNA regulates only the target; NM4: Open motif in which the microRNA regulates both the TF and the target but the TF-target link is missing; NM5, open motif in which two different microRNAs regulate separately the TF and the target. In the box we show the activactory micFFL whose deterministic and stochastic behavior we studied in the paper. B. Schematic view of the general miRNA controlled Feed Forward Loops (combining both activactory and repressive TF-target interactions) mined in the bioinformatic analysis discussed in the paper. C. Schematic description of the chemical reactions which must be taken into account to describe the miRNA-mediated feedforward loop with a miRNA-target titrative interaction.

Figure 2. A. Randomization of miRNA-target links.

Distribution of the number of FFLs for 1000 simulations obtained with JASPAR TFs list and confirmed by at least 4 miRNA databases (Z = 49,4). B. Randomization of miRNA-target links. Distribution of the number of FFLs for 1000 simulations obtained with ENCODE TFs list and confirmed by at least 4 miRNA databases (Z = 23,3). C. Randomization of TF-target links. Distribution of the number of FFLs for 1000 simulations obtained with JASPAR TFs list and confirmed by at least 4 miRNA databases (Z = −20,8). D. Randomization of TF-target links. Distribution of the number of FFLs for 1000 simulations obtained with ENCODE TFs list and confirmed by at least 4 miRNA databases (Z = −18,1).

Figure 3. The ratio of the target and TF concentrations as a function of for the micFFL and the NM2 and NM3 null models for three values and of the Hill exponent.

Figure 4. Heat map of the correlation for the micFFL and NM3,NM4 and NM5 Null Models.

In each plot the values of is mapped as a function of the miRNA concentration and of the interaction strength . While for NM3 and NM5 the fluctuation of TF and T are almost uncorrelated, both NM4 and the micFFL show a well defined region of large correlation. This correlation occurs for rather low miRNA concentrations and for almost any value of the miRNA-mRNA interaction strength.

Figure 5. Comparison of switch-on (A) and switch-off (B) response times between micFFL and direct regulation (NM1).

Figure 6. The network of micFFLs involving E2F1 as transcription factor and RB1 as target.