Model-guided quantitative analysis of microRNA- mediated regulation on competing endogenous RNAs using a synthetic gene circuit Ye Yuana,1, Bing Liua,1, Peng Xiea, Michael Q. Zhanga,b, Yanda Lia, Zhen Xiea,2, and Xiaowo Wanga,2 aMinistry of Education Key Laboratory of Bioinformatics and Bioinformatics Division, Center for Synthetic and Systems Biology, Tsinghua National Laboratory for Information Science and Technology/Department of Automation, Tsinghua University, Beijing 100084, China; and bDepartment of Molecular and Cell Biology, Center for Systems Biology, University of Texas, Dallas, TX 75080-3021 Edited by Wing Hung Wong, Stanford University, Stanford, CA, and approved January 27, 2015 (received for review July 23, 2014) Competing endogenous RNAs (ceRNAs) cross-regulate each other be affected by the miRNA-mediated stoichiometric mechanism at the posttranscriptional level by titrating shared microRNAs through a ceRNA effect or vice versa. (miRNAs). Here, we established a computational model to quan- Currently, the ability to systematically elucidate features of the titatively describe a minimum ceRNA network and experimentally ceRNA effect is impeded by the complexity of natural miRNA– validated our model predictions in cultured human cells by using ceRNA regulatory networks. Synthetic biology provides a com- synthetic gene circuits. We demonstrated that the range and plementary approach to investigate miRNA-mediated regu- strength of ceRNA regulation are largely determined by the relative lations in a controlled and largely isolated biological setting by abundance and the binding strength of miRNA and ceRNAs. We using engineered genetic circuits (11). Synthetic circuits have found that a nonreciprocal competing effect between partially and been constructed to help elucidate the underlying design principles perfectly complementary targets is mainly due to different miRNA of network motifs that combine transcriptional and miRNA-medi- loss rates in these two types of regulations. Furthermore, we ated regulations (12–14). It has also been demonstrated that miRNAs showed that miRNA-like off targets with high expression levels can generate a threshold in target gene expression by using a and strong binding sites significantly diminish the RNA interfer- bidirectional reporter assay in mammalian cells (15). However, ence efficiency, but the effect caused by high expression levels construction and implementation of a complex synthetic gene could be compensated by introducing more small interference circuit in mammalian cells to experimentally investigate the RNAs (siRNAs). Thus, our results provided a quantitative under- aforementioned ceRNA effect remain a great challenge. standing of ceRNA cross-regulation via shared miRNA and implied In this study, we aimed to obtain a comprehensive and quan- an siRNA design strategy to reduce the siRNA off-target effect in titative understanding of miRNA regulation principles on com- mammalian cells. peting RNAs. First, we formulated a coarse-grained model for a minimum miRNA–ceRNA system composed of one miRNA microRNA regulation | competing endogenous RNA | quantitative species and two competing RNA targets. Then, we engineered and biology | RNA interference efficiency | synthetic gene circuits implemented a corresponding genetic circuit in cultured human embryonic kidney 293 (HEK293) cells to quantify the ceRNA effect under various conditions by using a multifluorescent flow icroRNAs (miRNAs) are a class of ∼22-nt short noncoding RNAs that are loaded onto RNA-induced silencing com- M Significance plexes (RISC) and subsequently bind to their target RNAs. In mammalian cells, the perfect pairing of miRNA to target RNAs causes RNA cleavage through the RNA interference We established a minimum competing endogenous RNA (ceRNA) (RNAi) pathway, whereas partial pairing results in translational model to quantitatively analyze the behavior of the ceRNA regulation and implemented multifluorescent synthetic gene repression and RNA destabilization (1, 2). miRNA-mediated circuits in cultured human cells to validate our predictions. Our regulation can be triggered by only 6-nt complementarity of the results suggested that the ceRNA effect is affected by the miRNA 5′-end “seed region” to the target RNA, which confers abundance of microRNA (miRNA) and ceRNAs, the number and each miRNA species the capacity to interact with multiple RNA affinity of binding sites, and the mRNA degradation pathway species, including gene-coding mRNAs (3, 4), long noncoding determined by the degree of miRNA–mRNA complementarity. RNAs (5), and circular RNAs (6). Similarly, each RNA species Furthermore, we found that a nonreciprocal competing effect can interact with multiple miRNA species through various miRNA between partial and perfect complementary targets is mainly response elements (MREs) (7). due to different miRNA loss rates in these two types of The complex interaction network of miRNAs and their target repressions, which sheds light on utilizing such a competing RNAs has been shown to allow indirect cross-regulation between model for rational design of effective siRNA. different competing endogenous RNAs (ceRNAs) by seques- tering shared miRNAs, which is essential for regulating many Author contributions: Y.Y., B.L., P.X., M.Q.Z., Y.L., Z.X., and X.W. designed research; Y.Y. biological functions (7). The strength of ceRNA regulation is largely and B.L. performed research; Y.Y., Z.X., and X.W. analyzed data; and Y.Y., B.L., Z.X., and determined by the relative abundance and binding strength of X.W. wrote the paper. ceRNAs and miRNAs and whether the miRNA-bound ceRNA The authors declare no conflict of interest. decays through a stoichiometric mechanism or a catalytic mecha- This article is a PNAS Direct Submission. nism (8–10). The threshold-like behavior of the ceRNA regulation Freely available online through the PNAS open access option. has been experimentally observed by measuring the abundance Data deposition: The RNA sequencing data reported in this paper have been deposited in of two ceRNAs, phosphatase and tensin homolog (PTEN) and the NCBI Sequence Read Archive, www.ncbi.nlm.nih.gov/sra/ (accession no. SRP052983). vesicle-associated membrane protein (VAMP)-associated protein A 1Y.Y. and B.L. contributed equally to this work. (VAPA) across various cell lines (8). Nevertheless, many quan- 2To whom correspondence may be addressed. Email: [email protected] or titative predictions deduced from miRNA–ceRNA computational [email protected]. models have not been experimentally validated. Another intriguing This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. question is whether the miRNA-mediated catalytic mechanism can 1073/pnas.1413896112/-/DCSupplemental. 3158–3163 | PNAS | March 10, 2015 | vol. 112 | no. 10 www.pnas.org/cgi/doi/10.1073/pnas.1413896112 Downloaded by guest on October 1, 2021 cytometry. Based on computational and experimental results, we A protein1g B demonstrated that the relative abundance of miRNAs and com- 1 P R 1 1 – k P peting RNAs, the miRNA target binding free energy, and the g k iRFP TagBFP number of MREs have strong impacts on miRNA-mediated R1 CAG ceRNA1 1 hEF1α g ceRNA regulation in suitable molecular environments. Further- 1 miR-21 R1 k α more, we proposed a special ceRNA system consisting of one 1+ (1-α)g target with partially paired MREs (miRNA-type target) and the 1 1 rtTA k CAG other one with perfectly paired MREs (RNAi-type target). We 1- C1 k rtTA found a nonreciprocal competition between partially and perfectly S S miRNA complementary targets that share the same miRNA species. This Dox competing effect was largely due to the low miRNA loss rate in S g k C RNAi-type regulation and the high miRNA loss rate in miRNA- 2+ 2 mKate EYFP type regulation. In addition, we demonstrated that the RNAi ef- k (1-α)2 g 2 hEF1α pTRE k 2- 2 R2 g (ceRNA1) (ceRNA2) ficiency of siRNA can be significantly reduced by miRNA-like off R2 2 targets with high expression levels and strong binding sites. In- ceRNA2 α 2 k creasing siRNA concentrations diminished the competing effect R g P 2 g caused by highly expressed off targets, but severely repressed off P targets with strong binding sites. This finding provided us an protein2 2 siRNA design strategy to reduce the siRNA off-target effect. Fig. 1. Minimum miRNA–ceRNA model and experimental design. (A) The Results minimum miRNA–ceRNA model. The parameters are described in detail in SI Materials and Methods.(B) Schematic representation of a minimal miRNA– – Minimum miRNA ceRNA Model and Experimental Design. To analyze ceRNA synthetic gene circuit. Arrows represent up-regulation; lines with bars the quantitative behavior of the ceRNA effect, we established a designate down-regulation. For simplicity, all adjacent miRNA response ele- minimum miRNA–ceRNA model in which two mRNA species ments (MREs) in the 3′-UTR region are shown as a green box. are regulated by one miRNA species (Fig. 1A). This model was inspired by previous quantitative studies on the repression of small RNA in bacteria (16), miRNA regulation on a single target 3′-UTR of a near-infrared fluorescent protein (iRFP) gene, us- in mammalian cells (15), and in silico simulations on ceRNA ing a similar method to that described previously (13, 18). Four effects (8). Following Mukherji et al. (15), we assumed that only tandem repeats of miR-21 target sites (MREs) were fused to the the free mRNAs that are not bound by miRNAs can translate 3′-UTRs of both