Determination of in vivo target search kinetics of regulatory noncoding RNA

See allHide authors and affiliations

Science  20 Mar 2015:
Vol. 347, Issue 6228, pp. 1371-1374
DOI: 10.1126/science.1258849

RNA kinetics may define regulatory hierarchy

The double-helical structure of DNA suggests immediately how nucleic acid polymers can recognize and bind to homologous sequences. Target recognition by RNA is vital in many biological processes. Fei et al. used super-resolution microscopy of tagged RNAs and computer modeling to understand how RNA-RNA base-pairing reactions occur in vivo. They studied a small RNA (sRNA) that targets a messenger RNA (mRNA) for degradation in bacteria. They observed a slow rate of association as the sRNA searched for its mRNA target, but thereafter a fast rate of dissociation. This explains the need for high concentrations of sRNA to cause mRNA degradation. The sRNA found different target mRNAs at different rates, allowing the generation of a regulatory hierarchy.

Science, this issue p. 1371


Base-pairing interactions between nucleic acids mediate target recognition in many biological processes. We developed a super-resolution imaging and modeling platform that enabled the in vivo determination of base pairing–mediated target recognition kinetics. We examined a stress-induced bacterial small RNA, SgrS, which induces the degradation of target messenger RNAs (mRNAs). SgrS binds to a primary target mRNA in a reversible and dynamic fashion, and formation of SgrS-mRNA complexes is rate-limiting, dictating the overall regulation efficiency in vivo. Examination of a secondary target indicated that differences in the target search kinetics contribute to setting the regulation priority among different target mRNAs. This super-resolution imaging and analysis approach provides a conceptual framework that can be generalized to other small RNA systems and other target search processes.

Base-pairing interactions between nucleic acids constitute a large category of target recognition processes such as noncoding RNA-based gene regulation [e.g., microRNAs (1) and long noncoding RNAs (2) in eukaryotes and small RNAs (sRNAs) in bacteria (3)], bacterial adaptive immunity [e.g., the clustered regularly interspaced short palindromic repeat (CRISPR) system (4)], and homologous recombination (5). Although target search kinetics by transcription factors has been studied in vivo (6), the rate constants for target identification via base-pairing interactions in vivo are not known for any system. Here, we developed a super-resolution imaging and analysis platform to assess the kinetics of base-pairing interaction-mediated target recognition for a bacterial sRNA, SgrS. SgrS is produced upon sugar-phosphate stress, and its function is dependent on an RNA chaperone protein Hfq. SgrS regulates several target mRNAs posttranscriptionally through base-pairing interactions that affect mRNA translation and stability (7). We combined single-molecule fluorescence in situ hybridization (smFISH) (8) with single-molecule localization-based super-resolution microscopy (9) to count RNAs and obtain information on subcellular localization. High spatial resolution is required for accurate quantification of the high-copy-number RNAs and sRNA-mRNA complexes. Here, simultaneous measurements of sRNA, mRNA, and sRNA-mRNA complexes together with mathematical modeling allow determination of key parameters describing sRNA target search and downstream codegradation of sRNA-mRNA complexes.

We first studied the kinetic properties of SgrS regulation of ptsG mRNA, encoding a primary glucose transporter. SgrS binds within the 5′ untranslated region (UTR) of ptsG mRNA, blocks its translation, and induces its degradation (10). We induced stress and SgrS production in Escherichia coli strains derived from wild-type MG1655 (table S1) using a nonmetabolizable sugar analog, α-methyl glucoside (αMG) (10, 11). Fractions of cell culture were taken at different time points after induction and fixed (12). Oligonucleotide probes (table S2) labeled with photoswitchable dyes, Alexa 647 and Alexa 568, were used to detect SgrS (9 probes) and ptsG mRNA (28 probes), respectively, using smFISH (8). We then imaged the cells using two-color three-dimensional (3D) super-resolution microscopy (9, 12) (Fig. 1A; compare to diffraction limited images in Fig. 1B).

Fig. 1 Super-resolution imaging and analysis.

(A) 3D Super-resolution images of SgrS and ptsG mRNA labeled by smFISH projected in 2D planes. (B) Diffraction-limited fluorescent images of SgrS and ptsG mRNA. Cell boundaries imaged by differential interference contrast in (A) and (B) are depicted by white solid lines. (C) Examples of clustering analysis with comparison of raw data (left) and clustered data (right). (D) Comparison of average RNA copy number per cell measured by super-resolution imaging and qPCR. (E) Kinetic scheme of SgrS-induced ptsG mRNA degradation. Kinetic steps are described in the main text. [p], [S], and [Sp] are the concentrations of ptsG mRNA, SgrS, and their complex, respectively, in the mass-action equations.

In the wild-type strain (table S1), we observed production of SgrS and corresponding reduction of ptsG mRNA over a few minutes (Fig. 1A), consistent with SgrS-mediated degradation of ptsG mRNA (10). In a strain producing an SgrS that does not base pair with ptsG mRNA due to mutations in the seed region (13, 14) and in an Hfq deletion (Δhfq) strain (table S1), ptsG mRNA reduction was not observed (figs. S1 and S2). To quantify the copy number of RNAs in each cell, we employed a density-based clustering algorithm to map single-molecule localization signal to individual clusters corresponding to individual RNAs (12, 15, 16) (Fig. 1C and movies S1 and S2). The absolute copy number quantification was validated by quantitative polymerase chain reaction (qPCR) (12) (Fig. 1D).

We next built a kinetic model containing the following kinetic steps: transcription of SgrS (with rate constant αS) and ptsGp), endogenous degradation of ptsG mRNA (with rate constant βp), degradation of SgrS in the absence of codegradation with ptsG mRNA (βS,p), binding of SgrS to ptsG mRNA (with rate constant kon), dissociation of SgrS from ptsG mRNA (koff), and ribonuclease E (RNase E)–mediated codegradation of SgrS-ptsG mRNA complex (kcat) (Fig. 1E). We independently measured βp and the total SgrS degradation rate, including endogenous and mRNA-coupled degradation [table S4, fig. S3, and supplementary materials section 1.9 (SM 1.9)]. Because ptsG mRNA levels remained constant in the absence of SgrS-mediated degradation, as observed in the base-pairing mutant strain (fig. S1), we determined αp as the product of βp and ptsG mRNA concentration before SgrS induction (table S4 and SM 1.10)

To determine kon and koff, it is necessary to count the SgrS-ptsG mRNA complexes. Colocalization of ptsG mRNA and SgrS at the 40-nm resolution was rarely observed in the wild-type strain (up to ~5%, similar to ~3% colocalization by chance, estimated using the base-pairing mutant as a negative control) (Fig. 2). This is possibly because SgrS regulates several other target mRNAs (7) and/or the SgrS-ptsG mRNA complex may be unstable due to rapid codegradation or disassembly. In an RNase E mutant strain, in which codegradation is blocked (17, 18) (table S1), ptsG mRNA levels stayed the same as SgrS levels increased (fig. S4) (17, 18), and a fraction of ptsG mRNA colocalized with SgrS, increasing over time to reach ~15% (Fig. 2 and fig. S5). A positive control using ptsG mRNA simultaneously labeled with two colors (Fig. 2 and SM 1.8) showed a high degree of colocalization (~70%), similar to the reported detection efficiency of colocalization by super-resolution imaging (19).

Fig. 2 Colocalization analysis of SgrS-ptsG complex.

(A) Example of colocalization under various conditions. (B) Quantification of colocalized fraction of ptsG mRNA in cases (ii), (iii) (at 10 min after SgrS induction), and (iv) (at 10 min after SgrS induction). Error bars are standard deviations from three to eight images. (C) Time-course changes in the fraction of colocalized ptsG mRNA with SgrS. Error bars are standard errors from 200 to 600 cells from two independent measurements.

We then applied these measured parameters (αp and βp), used total SgrS degradation rate as a constraint for βS,p, and determined the remaining parameters (αS, βS,p kon, koff, and kcat) by fitting equations (Fig. 1E) to the six time-course changes of SgrS, ptsG mRNA, and SgrS-ptsG mRNA complex in both the wild-type and the RNase E mutant strains (Fig. 3A, table S4, and SM 1.10). We further validated the model by changing experimental conditions to vary only the transcription rates of SgrS (with lower αMG concentration) and/or ptsG mRNA (in the absence of glucose in the growth media), and the model could account for the data with the same set of kon, koff, and kcat values (table S4, figs. S6 to S8, and SM 2.2).

Fig. 3 Estimation of kinetic parameters.

(A) Modeling of time-course changes of SgrS, ptsG mRNA, and SgrS-ptsG complexes. Average copy numbers per cell are plotted as a function of time. Rate constants and weighted R2 for modeling are reported in tables S4 and S5. (B) Extraction of KD for SgrS-mRNA complex formation. The ratio of mRNA in complex with SgrS to free mRNA is plotted against average SgrS copy number and the slope of the linear fitting reports 1/KD. Error bars in (A) and (B) report standard errors from 200 to 600 cells from two independent measurements.

We can now quantitatively examine the effect of Hfq, which functions by stabilizing the sRNA or promoting its annealing with the target mRNA (20). In the Δhfq strain, the degradation rate of SgrS increased by a factor of 20, whereas the SgrS-ptsG mRNA association rate decreased slightly (table S4, figs. S1 and S8, and SM 2.2). Therefore, for the SgrS-ptsG mRNA pair, the primary effect of Hfq in regulation kinetics is in SgrS stabilization.

This in vivo determination of base pairing–mediated target search kinetics revealed two important characteristics of SgrS-mediated ptsG mRNA degradation. First, the target search process is characterized by slow association [kon = (2.0 ± 0.2) × 105 M−1 s−1] and fast dissociation (koff = 0.20 ± 0.04 s−1), resulting in a dissociation constant (KD = koff/kon) of 1.0 ± 0.2 μM (Fig. 3B and SM 1.11). To get a comparable apparent association rate, ka,app (kon×[S]), and koff, about one thousand SgrS molecules per cell need to be produced. The large KD explains the need for excessive production of sRNA molecules to enable rapid regulation when cells experience high levels of stress. Despite the crowded cellular environment and large excess of non–target RNA molecules, the kon is within the wide range of Hfq-mediated sRNA and target mRNA association rates reported by in vitro measurements. In contrast, koff is 1 to 2 orders of magnitude larger than in vitro estimates (2123). The large KD for target search in vivo is likely due to the limited availability of key players in the cell. For example, Hfq was suggested to be limited in the cell due to the dynamic competition for Hfq among different sRNAs (24, 25). Second, kcat and koff are comparable such that both codegradation and dissociation can occur with similar probabilities upon target binding. Disallowing dissociation by setting koff to zero cannot explain our experimental data (fig. S9 and SM 2.1). The observed fast kcat (0.4 ± 0.1 s−1) may be due to the formation of a ribonucleoprotein complex comprised of SgrS, Hfq, and RNase E, as suggested by biochemical studies (18); if so, once SgrS-Hfq-RNase E binds the ptsG mRNA, RNase E would be readily available for codegradation.

The kinetic model suggests that the overall rate of ptsG mRNA degradation is limited by its association with SgrS: At early time points after SgrS induction, when the copy number of SgrS is on the order of tens per cell, ka,app is about two orders of magnitude smaller than the codegradation rate kcat. The nonhomogenous spatial distribution of the sRNA and its target mRNA may also contribute to the slow target search. We observed membrane localization of ptsG mRNA, whereas SgrS is primarily localized in the cytoplasm (fig. S10). Further modeling incorporating the spatial information and stochastic gene expression may improve the kinetic analysis.

Regulation prioritization among multiple targets by one sRNA was suggested by computational modeling (26, 27) and experimental observation (28). However, how the kinetic prioritization is achieved remains to be elucidated. We propose that the combination of kon, koff, and kcat is characteristic of a specific sRNA-mRNA pair and determines the regulatory outcome. kcat may reflect the regulatory mode (codegradation versus translation repression) and target search kinetics (kon and koff) could contribute to the regulatory specificity and priority among many targets. To investigate this possibility, we examined manXYZ mRNA, which encodes a general sugar transporter for mannose and glucose and is also negatively regulated by SgrS. Compared with ptsG, manXYZ mRNA showed slower degradation kinetics (28) (figs. S11 to S13). The prioritization of ptsG over manXYZ by SgrS is consistent with the observation that SgrS regulation of ptsG (but not manXYZ) is absolutely essential for continued cell growth under most stress conditions (29). Using the RNase E mutant strain, we found that formation of SgrS-manXYZ mRNA complexes is slower than SgrS-ptsG mRNA complex formation (fig. S13C). The KD for SgrS binding to manXYZ mRNA, 2.3 ± 0.2 μM, was also higher than 1.0 ± 0.2 μM for SgrS binding to ptsG mRNA (Fig. 3B and SM 1.11). This result indicates that the slower regulation kinetics observed for manXYZ may, at least partially, originate from the differences in target search kinetics.

Overall, our results indicate that the formation of sRNA-mRNA complexes is reversible and highly dynamic in the cell, providing additional layers for regulating individual targets. Our kinetic model highlights the importance of target search kinetics on regulation prioritization. This super-resolution imaging and analysis platform provides a conceptual framework that can be generalized to other sRNA systems and potentially to other target search processes.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S18

Tables S1 to S5

Movies S1 to S3

References (3054)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank L. A. Sepulveda, L. H. So, M. Bates, X. Zhuang, Y. Liu, J. E. Stone, K. Schulten, H. Aiba, Z. Luthey-Schulten, T. E. Kuhlman, Y. Wang, and T. Lu for discussion, reagents, and software (12). This work was supported by grants from the National Science Foundation (PHY 082265 and PHY 1147498), the National Institutes of Health (GM 112659, GM065367, GM082837, and GM092830), the Welch Foundation (Q-1759), and the Jane Coffin Childs Memorial Fund for Medical Research.
View Abstract

Navigate This Article