Single-molecule modeling of mRNA degradation by miRNA: Lessons from data
© Sin et al.; licensee BioMed Central Ltd. 2015
Published: 1 June 2015
Recent experimental results on the effect of miRNA on the decay of its target mRNA have been analyzed against a previously hypothesized single molecule degradation pathway. According to that hypothesis, the silencing complex (miRISC) first interacts with its target mRNA and then recruits the protein complexes associated with NOT1 and PAN3 to trigger deadenylation (and subsequent degradation) of the target mRNA. Our analysis of the experimental decay patterns allowed us to refine the structure of the degradation pathways at the single molecule level. Surprisingly, we found that if the previously hypothesized network was correct, only about 7% of the target mRNA would be regulated by the miRNA mechanism, which is inconsistent with the available knowledge. Based on systematic data analysis, we propose the alternative hypothesis that NOT1 interacts with miRISC before binding to the target mRNA. Moreover, we show that when miRISC binds alone to the target mRNA, the mRNA is degraded more slowly, probably through a deadenylation-independent pathway. The new biochemical pathway proposed here both fits the data and paves the way for new experimental work to identify new interactions.
In living cells, the level of protein expression is thoroughly regulated. Many crucial processes for this regulation occur at the post-transcriptional level. In this context, control mechanisms acting on messenger RNAs (mRNAs) play a pivotal role. A number of biochemical pathways converging on cytosolic mRNAs serve to enhance or repress gene expression. These pathways are known to operate by enhancement of translation [1, 2], repression of translation [3, 4] or modulation of mRNA lifetimes [3–6]. The global picture emerging from the growing body of experimental evidence depicts a complex interaction network which affects the mRNAs available for translation. This network is composed of several biochemical pathways, often interwoven and cross-talking [7, 8], involving mRNA binding proteins as well as non coding RNAs [9–11]. While there are a number of mechanisms responsible for mRNA degradation in eukaryotic cells , the decay of messages mediated by micro-RNAs (miRNAs) plays a prominent role in the control of gene expression [3, 12, 13].
Despite extensive study, the topology and dynamics of miRNA-mediated mRNA degradation pathway are still unclear. One of the main challenges stems from the fact that intermediate states of the pathway are unknown or difficult to quantify; experimentally, it is only feasible to measure the decay patterns of the target mRNAs. Bridging the gap between observed decay patterns and degradation pathways is non-trivial , since the former refer to a population average and the latter refers to the single-molecule stochastic process of degradation. Here we apply a rigorous strategy to reconstruct the miRNA-mediated degradation pathway, starting from experimentally measured decay patterns. Surprisingly, we find the previously proposed pathway not consistent with the experimental data. We propose an alternative model which fits the decay pattern and allows us to gain some insight into the network topology.
The experimental data
The conclusion of this detailed experimental study is that NOT1 is a more relevant factor than PAN3 in destabilizing the mRNA . When only NOT1 is knocked down, the decay of F-Luc-Nerfin mRNA is significantly slower (yellow curve, Figure 2) than the control (red curve, Figure 2). In contrast, the effect of PAN3 knock down is less significant (blue curve, Figure 2). These findings apparently confirm that the degradation pathway through NOT1 in Figure 1 is the most prominent pathway for degradation of the target mRNA. Although this conclusion is relatively robust, the published analyses do not validate the hypothesized biochemical degradation pathway given in Figure 1. Indeed, in the negative control (green curve in Figure 2), the miRNA is knocked-down so that the formation of a specific silencing complex miRISC is suppressed, yet the target mRNA still decays. Additionally, when only the miRNA is expressed while PAN3 and NOT1 are knocked-down, the target mRNA decays (black curve in Figure 2), but is definitively more stable than in the negative control. Both of these cases suggest that the model hypothesized in Figure 1 should be expanded to include additional degradation pathways.
An important conceptual consideration is that Figure 1 depicts degradation from the single-molecule perspective whereas the curves in Figure 2 are averages as a function of time. Therefore, our strategy consists of starting with the network shown in Figure 1 and validating it against the experimental decay patterns. At the same time, we will propose alternative parsimonious extensions of the network when the validation fails. In particular, we will find that when the miRISC complex interacts with the mRNA alone, it seems to stabilize the mRNA and perhaps trigger a deadenylation independent degradation of the mRNA. Furthermore, we will show that the data supports the hypothesis that miRISC binds to NOT1 before recruiting the target mRNA and that there is a strong enhancement of mRNA recruitment when PAN3 is also present.
As previously mentioned, the relationship between degradation pathways (such as the one in Figure 1) and decay patterns (such as those in Figure 2) is not trivial. If the decay pattern was exponential, the halftime of the mRNA population estimated from the decay pattern would be directly related to the rate of decay of single molecules. The analysis of the decays shown in Figure 2 shows that a model based on a single exponential results in a poor fit; more complex models are preferred even in light of evaluations based on the Akaike Information Criteria (AIC). Furthermore, if the decay curves could be well described by a model based on a single exponential, the traces would appear as straight lines when plotted in a linear-log scale (see Figure 7 in ).
While the network in Figure 3 results in a definitively better fit to the data and thus could be used to derive quantities such as the average lifetime and the age dependent degradation rate, it does not tell us if the network in Figure 1 is a suitable framework for the decay patterns observed in Figure 2. To address this question we employ a hierarchical strategy: (i) we start by fitting the negative control decay pattern (the green trace "Control (-)" in Figure 2) to the most parsimonious model (Figure 3) and fix the corresponding three rates; (ii) we then consider the next decay pattern with one additional decay factor active and enlarge the network of states to accommodate the additional decay factor. We continue until each curve has been evaluated and the corresponding network is built.
The details of the functions used to perform the fit can be found in Additional file 1: Section S1.1 provides the general aspects of the mathematical background required for the purpose of this paper, section S1.2 gives the explicit formulas used for the fit, section S1.4 provides parameters estimations.
Based on the hierarchical strategy above, we start with the "Control (-)" curve (green decay pattern, Figure 2). This curve describes the decay of the mRNA when none of the degradation factors (miRNA, NOT1 and PAN3) are present. In the framework of the hypothesized network in Figure 1, this corresponds to all downstream processes inactive, thus reducing the network to just the first state (green circle). This is obviously not sufficient to explain the observed decay, since a single state without decay would produce a horizontal line (i.e. no decay). This one-state scenario is also not consistent with biological reality. Indeed, even the most stable cellular macromolecule is eventually degraded. In fact, there are many biochemical pathways devoted to mRNA degradation .
The crisis of the original hypothesis
i.e., about 7% of the whole mRNA binds to miRISC complexes in the absence of NOT1 and or PAN3, based on the network shown in Figure 6. This discovery leads to two conclusions. First, the fraction of mRNA that can be manipulated after binding with miRISC is so small that an enlargement of the network by including a separate NOT1 and a separate PAN3 pathway downstream of miRISC binding becomes meaningless. Indeed, attempts to do so lead to very poor fitting of the remaining curves (see Additional file 1). Second, such a small fraction of miRNA-regulated mRNA (about 7%) would indicate that miRNA cannot be considered a strong mechanism of gene regulation, contrary to the experimental evidence that miRNA is a strong regulatory mechanism.
Therefore, we are forced to partially reject the hypothesis formulated in Figure 1 and revise it in search for other possible interactions between miRISC, PAN3 and NOT1. Note that the computed value of 7% is necessarily affected by some error due to the precision by which the data could be extracted from the plots originally published in Ref. . Nevertheless, this value is an indication that the model of degradation originally proposed in  would predict that only a very small fraction of mRNA is involved in miRNA mediated degradation. In the following we will present a parsimonious model of degradation that is able to predict more realistic figures of the relative amounts of mRNAs involved in the different degradation pathways.
Finally, the comparison between the decay pattern fitted in Figure 4 and 7 shows that binding of miRISC alone does stabilize the mRNA compared to when the miRNA is not expressed. This is a strong indication that miRISC "protects" the target mRNA from the action of alternative, competing degradation pathways.
A new hypothesis arises from the data
which emphasizes the strong role of NOT1 in the degradation of mRNA.
indicating that this model produces the strong regulatory effect of the miRNA on its target as expected.
The cooperative role of PAN3
Summary and discussion
In this paper, we show that the current hypothesis about the sequence of interactions between miRISC, its target mRNA and the factor NOT1 is not supported by the data. We have shown that the mRNA is also degraded when the miRNA is not expressed, indicating the existence of an alternative pathway, possibly competing with the miRNA pathway.
We also show that when only miRNA is expressed (with NOT1 and PAN3 knocked down), the target mRNA is stabilized, probably because it is protected from the action of an alternative miRNA-independent pathway. We postulate that the binding between miRISC and mRNA is irreversible and leads to deadenylation independent decay of the target message in agreement with recent experimental studies. However, this assumption is not obligatory. Indeed, one could have hypothesized that binding to miRISC is reversible, and that the presence of miRNA alone just slows down the action of the alternative pathway. With the present data it is not possible to distinguish between these two alternatives.
Finally, our analysis indicates that the miRISC complex and NOT1 interact with each other before interacting with the mRNA. We assume that this discovery is not limited to the special miRNA-mRNA pair studied in  and is therefore a new general mechanism of mRNA control. Our analyses confirm the conclusions in  that PAN3 without NOT1 does not lead to an identifiable destabilization of the mRNA. Nevertheless, we see a strong cooperative effect between PAN3 and NOT1, where PAN3 is able to strongly enhance the binding of the miRISC+NOT1+PAN3 complex to the target mRNA compared to the miRISC+NOT1 complex alone.
Experimentally, one should be able to detect the presence of miRISC+NOT1 complexes in the absence of target mRNA in order to verify our findings. Moreover, steady state relative amounts of mRNA in the different biochemical states can provide further validation data for our networks and additional information to unveil further details of miRNA-mediated mRNA degradation.
Publication of this article has been funded by the European Union FP7 project Marie Curie ITN "NICHE".
This article has been published as part of BMC Systems Biology Volume 9 Supplement 3, 2015: Proceedings of the Italian Society of Bioinformatics (BITS): Annual Meeting 2014: Systems Biology. The full contents of the supplement are available online at http://www.biomedcentral.com/bmcsystbiol/supplements/9/S3.
- Orom UA, Nielsen FC, Lund AH: MicroRNA-10a Binds the 5'UTR of Ribosomal Protein mRNAs and Enhances Their Translation. Mol Cell. 2008, 30 (4): 460-471. 10.1016/j.molcel.2008.05.001.View ArticlePubMedGoogle Scholar
- Vasudevan S, Tong Y, Steitz JA: Switching from repression to activation: MicroRNAs can up-regulate translation. Science. 2007, 318 (5858): 1931-1934. 10.1126/science.1149460.View ArticlePubMedGoogle Scholar
- Huntzinger E, Izaurralde E: Gene silencing by microRNAs: Contributions of translational repression and mRNA decay. Nature Reviews Genetics. 2011, 12 (2): 99-110. 10.1038/nrg2936.View ArticlePubMedGoogle Scholar
- Izaurralde E: Elucidating the temporal order of silencing. EMBO Rep. 2012, 13 (8): 662-663. 10.1038/embor.2012.91.PubMed CentralView ArticlePubMedGoogle Scholar
- Repetto E, Briata P, Kuziner N, Harfe BD, McManus MT, Gherzi R, Rosenfeld MG, Trabucchi M: Let-7b/c enhance the stability of a tissue-specific mRNA during mammalian organogenesis as part of a feedback loop involving KSRP. PLoS Genetics. 2012, 8 (7): e1002823-10.1371/journal.pgen.1002823.PubMed CentralView ArticlePubMedGoogle Scholar
- Cairrao F, Halees AS, Khabar KSA, Morello D, Vanzo N: AU-rich elements regulate Drosophila gene expression. Mol Cell Biol. 2009, 29 (10): 2636-2643. 10.1128/MCB.01506-08.PubMed CentralView ArticlePubMedGoogle Scholar
- Ma F, Liu X, Li D, Wang P, Li N, Lu L, Cao X: MicroRNA-466l upregulates IL-10 expression in TLR-triggered macrophages by antagonizing RNA-binding protein tristetraprolin-mediated IL-10 mRNA degradation. Journal of Immunology. 2010, 184 (11): 6053-6059. 10.4049/jimmunol.0902308.View ArticleGoogle Scholar
- Jing Q, Huang S, Guth S, Zarubin T, Motoyama A, Chen J, et al: Involvement of MicroRNA in AU-Rich Element-Mediated mRNA Instability. Cell. 2005, 120 (5): 623-634. 10.1016/j.cell.2004.12.038.View ArticlePubMedGoogle Scholar
- Houseley J, Tollervey D: The Many Pathways of RNA Degradation. Cell. 2009, 136 (4): 763-776. 10.1016/j.cell.2009.01.019.View ArticlePubMedGoogle Scholar
- Tay Y, Rinn J, Pandolfi PP: The multilayered complexity of ceRNA crosstalk and competition. Nature. 2014, 505 (7483): 344-352. 10.1038/nature12986.PubMed CentralView ArticlePubMedGoogle Scholar
- Kartha RV, Subramanian S: Competing endogenous RNAs (ceRNAs): New entrants to the intricacies of gene regulation. Frontiers in Genetics. 2014, 5: 8-PubMed CentralView ArticlePubMedGoogle Scholar
- Belasco JG: All things must pass: Contrasts and commonalities in eukaryotic and bacterial mRNA decay. Nature Reviews Molecular Cell Biology. 2010, 11 (7): 467-478. 10.1038/nrm2917.PubMed CentralView ArticlePubMedGoogle Scholar
- Lu J, Clark AG: Impact of microRNA regulation on variation in human gene expression. Genome Res. 2012, 22 (7): 1243-1254. 10.1101/gr.132514.111.PubMed CentralView ArticlePubMedGoogle Scholar
- Deneke C, Lipowsky R, Valleriani A: Complex Degradation Processes Lead to Non-Exponential Decay Patterns and Age-Dependent Decay Rates of Messenger RNA. PLoS One. 2013, 8 (2): 1-View ArticleGoogle Scholar
- Braun JE, Huntzinger E, Fauser M, Izaurralde E: GW182 proteins directly recruit cytoplasmic deadenylase complexes to miRNA targets. Mol Cell. 2011, 44 (1): 120-133. 10.1016/j.molcel.2011.09.007.View ArticlePubMedGoogle Scholar
- Wu Q, Smith-Miles K, Zhou T, Tian T: Stochastic modelling of biochemical systems of multi-step reactions using a simplified two-variable model. BMC Syst Biol. 2013, 7 (Suppl 4): S14-10.1186/1752-0509-7-S4-S14.PubMed CentralView ArticlePubMedGoogle Scholar
- Valleriani A, Liepelt S, Lipowsky R: Dwell time distributions for kinesin's mechanical steps. EPL. 2008, 82 (2): 28011-10.1209/0295-5075/82/28011.View ArticleGoogle Scholar
- Keller P, Valleriani A: Single-molecule stochastic times in a reversible bimolecular reaction. J Chem Phys. 2012, 137 (8): 084106-10.1063/1.4747337.View ArticlePubMedGoogle Scholar
- Valleriani A, Li X, Kolomeisky AB: Unveiling the hidden structure of complex stochastic biochemical networks. J Chem Phys. 2014, 140 (6): 064101-10.1063/1.4863997.PubMed CentralView ArticlePubMedGoogle Scholar
- Li X, Kolomeisky AB, Valleriani A: Stochastic kinetics on networks: when slow is fast. J Phys Chem B. 2014, 118 (35): 10419-10425. 10.1021/jp506668a.PubMed CentralView ArticlePubMedGoogle Scholar
- Li X, Kolomeisky AB, Valleriani A: Pathway structure determination in complex stochastic networks with non-exponential dwell times. J Chem Phys. 2014, 140 (18): 184102-10.1063/1.4874113.PubMed CentralView ArticlePubMedGoogle Scholar
- Helfer S, Schott J, Stoecklin G, Forstemann K: AU-rich element-mediated mRNA decay can occur independently of the miRNA machinery in mouse embryonic fibroblasts and Drosophila S2-cells. PLoS One. 2012, 7 (1): e28907-10.1371/journal.pone.0028907.PubMed CentralView ArticlePubMedGoogle Scholar
- Zekri L, Kuzuoglu D, Izaurralde E: GW182 proteins cause PABP dissociation from silenced miRNA targets in the absence of deadenylation. EMBO Journal. 2013, 32 (7): 1052-1065. 10.1038/emboj.2013.44.PubMed CentralView ArticlePubMedGoogle Scholar
- Nishihara T, Zekri L, Braun JE, Izaurralde E: miRISC recruits decapping factors to miRNA targets to enhance their degradation. Nucleic acids research. 2013, 41 (18): 8692-8705. 10.1093/nar/gkt619.PubMed CentralView ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.