|Home | About | Journals | Submit | Contact Us | Français|
The heterodimer NF90–NF45 is an RNA-binding protein complex that modulates the expression of various cellular mRNAs on the post-transcriptional level. Furthermore, it acts as a host factor that supports the replication of several RNA viruses. The molecular mechanisms underlying these activities have yet to be elucidated. Recently, we showed that the RNA-binding capabilities and binding specificity of NF90 considerably improves when it forms a complex with NF45. Here, we demonstrate that NF90 has a substrate-selective RNA chaperone activity (RCA) involving RNA annealing and strand displacement activities. The mechanism of the NF90-catalyzed RNA annealing was elucidated to comprise a combination of ‘matchmaking’ and compensation of repulsive charges, which finally results in the population of dsRNA products. Heterodimer formation with NF45 enhances ‘matchmaking’ of complementary ssRNAs and substantially increases the efficiency of NF90’s RCA. During investigations of the relevance of the NF90–NF45 RCA, the complex was shown to stimulate the first step in the RNA replication process of hepatitis C virus (HCV) in vitro and to stabilize a regulatory element within the mRNA of vascular endothelial growth factor (VEGF) by protein-guided changes of the RNAs’ structures. Thus, our study reveals how the intrinsic properties of an RNA-binding protein determine its biological activities.
NF90 is a double-stranded (ds) RNA-binding protein that is crucially involved in RNA metabolism. NF90’s role as a post-transcriptional regulator of gene expression is best understood where it binds to structured 3′-untranslated regions (3′UTRs) of various mRNAs and modulates the RNA’s turnover and availability for translation (1–3). For example, under hypoxic stress, NF90 stabilizes the mRNA of vascular endothelial growth factor (VEGF) by associating with a defined structural element in the mRNA’s 3′-UTR (Supplementary Figure S1A) (4). NF90 also interacts with the mRNA-mimicking genomes of positive-strand RNA viruses. Thus, several members of the Flaviviridae family, e.g. hepatitis C virus (HCV), dengue virus (DV) and bovine viral diarrhea virus (BVDV), were found to exploit NF90 as a host factor that supports viral amplification (5–9). With HCV and BVDV, NF90 was shown to bind to defined RNA motifs close to the 3′-end of the viral RNA (6–8) (Supplementary Figure S1B) and to stimulate the viral RNA replication process (6–9). The molecular mechanisms underlying this pathogen-supporting activity of NF90 were hitherto unknown.
NF90 is 706 amino acids (aa) long and contains an N-terminal domain associated with zinc-fingers (DZF). Moreover, the protein contains a tandem pair of dsRNA-binding motifs (dsRBMs) and an RGG/RG motif in its otherwise structurally disordered C-terminus (aa 381–706). DsRBMs are known to associate with high affinity to dsRNA (10) while RGG/RG motifs preferentially interact with single-stranded (ss) RNA (11,12). Our own studies and those of others have shown that a dynamic and cooperative interplay of all three RBMs contributes to NF90’s overall RNA binding properties (13–15) and enables the protein to distinguish between different features of target RNAs (15).
NF90 is indicated to interact with several other protein partners (16); however, in most investigated cell types, it exists predominantly as a complex together with the 390 aa NF45 protein. The core of the NF90–NF45 complex is formed by the interaction between the N-terminal DZF of NF90 and the DZF of NF45, but the C-terminus of NF90 is also suggested to contribute to heterodimer formation (14,17–19). Recently, we reported that complex formation between NF90 and NF45 results in a mutual thermodynamic stabilization of the proteins, which leads to changes in the RNA-binding mode and significantly enhances the RNA-binding affinity of NF90 (19).
Several cellular proteins have been described to trigger the folding of functionally active conformations of RNA molecules and to regulate, via this function, diverse paths in cellular RNA metabolism (20). According to the current definition, proteins are referred to have an ‘RNA chaperone activity’, RCA, if they are capable of resolving mis-folded RNA structures in vitro and if they share activities in promoting RNA–RNA interactions by ‘RNA annealing’ or in separating RNA duplexes by ‘strand displacement’ (Figure (Figure1A)1A) (20,21). In the life cycle of RNA viruses, an essential step involves the folding of the viral RNAs into biologically relevant conformations (22). Thus, in several cases, cellular and viral RNA chaperones were suggested to assist viral infection processes by facilitating conformational transitions of RNA molecules (23–28). The exact mechanisms of how these proteins exert these activities have yet to be investigated.
Here, we report a hitherto unknown RCA of NF90, which manifests itself predominantly as an RNA annealing activity. Heterodimer formation between NF90 and NF45 strongly improves the efficiency of the RCA, and the underlying mechanism was characterized as a combination of ‘matchmaking’ and compensation of repulsive charges. Furthermore, we demonstrate that NF90–NF45, but neither a RCA-deficient complex variant nor NF90 alone, stimulates the first step of the HCV RNA replication process in vitro and that NF90–NF45 is able to elicit alterations in the structure of the viral RNA. Along the same line, we observed that NF90–NF45 increases the stability of a regulatory double-stranded region within the VEGF mRNA’s 3′UTR. Together, the findings of this study suggest that the RCA-driven changes in RNA structures represents a further essential functional determinant of the already diverse activities attributed to NF90 and NF90–NF45.
NF90, and variants of NF90 and NF45 were generated and purified as reported earlier (15,19). The proteins were produced in Escherichia coli BL21 (DE3) RP and isolated from inclusion bodies. Purification was performed under denaturing conditions applying a cation-exchange-chromatography column. Refolding of the individual proteins was achieved using a rapid-dilution/pulse-refolding approach in assay buffer [50 mM Na-phosphate, pH 7.2, 100 mM NaCl, 5% (v/v) glycerol, 1 mM TCEP] supplemented with arginine. Refolded protein solutions were finally applied to affinity and size-exclusion chromatography columns; each column was pre-equilibrated in assay buffer. Heterodimeric NF90–NF45 was produced by co-refolding equimolar amounts of NF90 and NF45 following the same procedure as with the individual proteins. The homogeneity of the prepared proteins and the heterodimeric complex was confirmed as described (15,19).
The 5′-fluorophor-labeled oligonucleotides used in this study (see Table Table1)1) were purchased from IBA GmbH (Goettingen, Germany). Concentrations were calculated by determining the absorbance at 260 nm using extinction coefficients provided by the supplier. The dsRNAs were prepared by mixing of the corresponding ssRNAs in equimolar ratio, heating to 95°C followed by slow cooling and final purification by non-denaturing PAGE before storage at –80°C.
10 nM of Cy3-labeled HCV-SL3 or RaSc (Table (Table1)1) were incubated with the indicated concentrations of protein in assay buffer for 5 min at 20°C. RNA annealing was induced by the addition of the complementary Cy5-labeled RNA strand (10 nM final) and readings were taken for 220 s. The reaction was monitored on a FluoroMax-4 (Jobin Yvon, France) by following a FRET signal, which occurred exclusively when we applied complementary RNA strands (see Supplementary Figure S2A and S2C). Excitation was done at 535 nm (Cy3) and emission was read at 680 nm (Cy5). Fluorescence intensities relative to intensities at t = 220 s were plotted versus time, and progress curves fitted according to a second-order reaction (Equation 1, (21)) using KaleidaGraph™ (Synergy Software) yielding the observed rate constants kobs.
F and F220 are the fluorescence intensity at t and 220 s. γ is the signal amplitude. c is the offset. kobs is the observed reaction rate constant.
kobs is the observed reaction rate constant at a protein or RNA concentration [S]0. KM is the protein concentration, where kobs is half-maximal and kcat is the catalytic rate constant at [S] 0 >> KM.
Catalytic efficiencies were calculated from kcat/KM.
Second-order rate constants kon were determined from the initial linear phase of the plot of kobs versus [S]0 according to Equation (3).
kobs, kon and koff represent the observed, the second-order rate constant and the off-rate, respectively. [S]0 is the protein or Cy5–ssRNA concentration.
Strand displacement reactions were performed according to Rajkowitsch and Schroeder (21). Briefly, RNA annealing reactions were allowed to proceed for 300 s prior to the addition of a 10-fold excess of unlabeled RNA strand together with indicated amounts of protein. The reaction was allowed to continue for another 300 s. kobs- and kon-values were obtained as described above for the RNA annealing reactions.
For all kinetic reactions, at least a duplicate of traces were averaged and applied to software-assisted calculations. The errors of resulting kinetic parameters that were obtained in the course of the fitting-routine were in the range of 15%.
RNA annealing reactions were performed as described above but at different temperatures. Apparent reaction rate constants kobs were determined from progress curves using Equation (1). Rate constants were normalized to the protein concentration to compensate for different specific activities of the proteins. Resulting second-order rate constants were further investigated by an Arrhenius analysis using the following Equations (4)-(7).
kobs and k are the observed and second-order rate constant, EA is the Arrhenius activation energy. R is the gas constant. T is the absolute temperature. C is constant. h is Planck's constant. kB is Boltzmann's constant. ΔG≠, ΔH≠ and ΔS≠ are the changes in free activation energy, in-activation enthalpy and in-activation entropy, respectively.
RNAs were transcribed with T7 RNA-polymerase from PCR products (standard procedures) that were generated with specific primers (Supplementary Table S1) from a pUC19-Con1 plasmid that contained the cDNA coding for a bicistronic replicon of HCV-1b (Con1 isolate) (30).
The assay was carried out in polymerase assay buffer [PAB; 20 mM HEPES/NaOH, pH 7.1, 50 mM NaCl, 20% (v/v) glycerol, 6.5 mM MgCl2, 0.5 mM MnCl2, 1 mM DTT, 0.5 mM r(ATP,UTP,GTP)] supplemented with 20 μCi [α-32P-]CTP, 30 nM template RNA and 100 nM NS5B (HCV-1b Con1) at 20 or 37°C in a final volume of 40 μl. The viral 3′-UTR plus the 5BSL3 region (5BSL3_3UTR), the X-tail only and poly-rC (1 μg) served as template RNA (Figure (Figure4A).4A). For reactions with poly-rC, [α-32P-]GTP was used instead of [α-32P-]CTP. When protein was investigated for an effect on the reaction, protein and template RNA were pre-incubated and allowed to equilibrate for 10 min at 20°C in PAB, before the reaction was started by the addition of NS5B. After 1 h, the reactions were stopped by phenol-chloroform extraction and radioactivity was measured applying scintillation counting.
30 nM of the viral RNA templates were incubated with the indicated concentrations of protein in PAB without MnCl2. Thermal denaturation was monitored in the range from 30 to 90°C at 0.4 K min−1 by recording the absorbance change at a constant wavelength of 268 nm on a Jasco-V550 equipped with temperature control unit ETC-505T. At this wavelength, maximal changes in the absorbance during unfolding were observed (not shown). Absorbance changes relative to the signal at 30°C were plotted against the temperature and transition curves were analyzed semi-quantitatively by calculating the melting points (TM) from the first derivative of each curve.
For VEGF HSR-SL RNA (IBA Goettingen, Table Table1),1), 500 nM of RNA were incubated with the indicated concentrations of protein in assay buffer. Thermal denaturation was conducted as described above but in the temperature range from 10 to 90°C at 0.4 K min−1. Absorbance changes were normalized and analyzed according to Riechmann et al. (31) using Equations (8) and (9) for analysis of the melting of VEGF RNA (unfolding of stem–loop).
A is the signal amplitude at temperature T, AN and AU are the signal amplitudes of the native and unfolded state extrapolated to T = 0 K, respectively. mN and mU are the slopes of the baselines of the native and unfolded state; TM is the transition midpoint, R is the gas constant, ΔCp is the difference in heat capacity between the native and unfolded state, and ΔG° and ΔH° are the changes in free energy and enthalpy in equilibrium, respectively.
NF90 and/or NF90–NF45 were earlier reported to be subunits of various nucleic acid-rearranging protein complexes (6,7,32,33). This led to the hypothesis that NF90, besides acting as a classical RNA-binding protein, might also function as an RNA chaperone. To test this, we took advantage of the assay developed by Rajkowitsch and Schroeder (21) using short, complementary Cy3/Cy5-fluorophor labeled ssRNAs that enable the monitoring of RNA annealing and strand displacement processes by time-dependent fluorescence resonance-energy transfer (FRET; Figure Figure1A).1A). On the one hand, we applied the same RNA molecules as these authors (referred here to as ‘RaSc’). On the other hand, we tested RNAs, which corresponded to a fully double-stranded version of the SL3 stem–loop motif that was identified by us to be part of the NF90–NF45 binding region in the HCV RNA (6), referred to here as ‘dsHCV-SL3’; Table Table11 and Supplementary Figure S1B). Control ‘RNA-only’ reactions validated the assay. By mixing the complementary ssRNAs, a signal change was measured, and the apparent rate constants (kobs) of RNA-RNA interactions were determined. When we performed this assay with the RaSc RNAs, our measurements correlated well with the data reported by Rajkowitsch and Schroeder (Figure (Figure1B1B and Supplementary Figure S2).
Interestingly, in succeeding experiments, which were carried out with the Cy-labeled RNAs and in the presence of homogenous preparations of NF90 (isoform NF90b) or the NF90b–NF45 heterodimer, the RNA annealing reactions were found to be significantly accelerated (Figure (Figure1C).1C). This was not observed in control reactions, which applied either the individual NF45, an unrelated RNA-binding protein, hnRNP H1 (28), or NF90b or NF90b–NF45 and non-complementary RNAs (Supplementary Figure S2). The apparent rate constant kobs approximated a hyperbolic ‘saturation’ behavior, which, in the case of the RNA-only reactions was limited by the concentrations of the RNAs (Figure (Figure1B).1B). In the presence of NF90b or NF90b–NF45, it depended on the protein concentration (Figure (Figure1C).1C). Since the reaction applied two ssRNAs, we explained this by the formation of a reaction intermediate involving both RNA strands (29,34,35). Accordingly, the linear part of the plot of kobs versus the RNA or protein concentration (Figure (Figure1B1B–C) could be investigated to yield the second-order rate constant kon as a measure of the initial association step of the two RNA strands. At higher concentrations, the decay of the intermediate became rate limiting as kcat; catalytic efficiencies were calculated from kcat/KM (further explained in Materials and Methods).
The analyses revealed that with both the RaSc and the dsHCV-SL3 RNAs, the heterodimer NF90b–NF45 showed a 10-fold increase in kcat and a 50-fold increase in kon as compared to the NF90b monomer (Table (Table2).2). Thus, the efficiency (kcat/KM) of the RNA annealing activity (RAA) of NF90b–NF45 versus the monomeric NF90b was enhanced by 13-fold with dsHCV-SL3, and 70-fold with RaSc (Table (Table2).2). The interesting fact that NF90–NF45 displayed different RAAs with the two types of RNA substrates was even more apparent when we correlated the ‘protein-catalyzed’ with the ‘RNA-only’ reactions. Thus, NF90b–NF45 improved the RNA annealing efficiency of RaSc by 62-fold, while it was 2.5-fold with dsHCV-SL3 (Table (Table2).2). In contrast, the NF90b monomer apparently did not improve RNA annealing efficiency of RaSc and impaired the RNA annealing of dsHCV-SL3 by 5-fold in comparison to the ‘RNA only’ reaction. With the applied RNA substrates, and under the chosen experimental conditions, neither NF90b nor NF90b–NF45 showed a strand-displacement activity.
Previous chemical unfolding studies of NF90b and NF90b–NF45 had demonstrated that the formation of the complex results in a thermodynamic stabilization of structural elements of the NF90b and NF45 components (19). To test whether the improved RCA of NF90b–NF45 correlated with this increased stability of the heterodimer, we assayed the RAA of NF90b–NF45 with dsHCV-SL3 in the presence of 0 to 4 M urea. Within this concentration range of the denaturant, the stabilization effect of heterodimer formation was earlier shown to be significant, and the individual as well as the NF45-complexed NF90b were known to bind RNA under these conditions (19). Control ‘RNA-only’ experiments revealed a negligible influence of the denaturant on the RNA-RNA interactions (Figure (Figure2A).2A). However, the acceleration of the RNA-annealing reaction by NF90b and NF90b–NF45 was only detectable up to 2–3 M of denaturant; at higher concentrations, the kobs values approximated the numbers that were measured with the ‘RNA-only’ reactions (Figure (Figure2A).2A). To compare directly the different RAA of NF90b and NF90b–NF45, kobs was related to the protein concentration and plotted as ln (kobs/[protein]) against the respective concentrations of urea (Figure (Figure2B).2B). This describes a linear free energy method (LEM) (15,19), which is established to investigate correlations between structure and function of different enzymes (19,36,37). Thus, a linear correlation of the LEM indicates the integrity of structural elements of the protein that are involved in the enzymatic reaction. We investigated the data obtained with NF90b and NF90b–NF45 applying an F-test that compared a line versus an exponential model with a false-rejection of P < 0.01. As shown in Figure Figure2B,2B, this revealed a linear correlation for the RAA of the NF90b–NF45 heterodimer in the range of 0–2 M urea, while this was not the case for NF90b. Hence, we considered these results as an indication that functional elements of NF90 and/or NF45, which contribute to the RAA, are stabilized within the heterodimer.
The above findings directed us to evaluate next the contributions of the individual RNA-binding motifs (RBMs) of NF90b in the RNA annealing reaction. For this, we examined two complex variants, NF90bS651E–NF45 and NF90bF432,559A–NF45 that had been characterized earlier, and in which amino acid substitutions in NF90b impaired the RGG motif or both dsRBMs, respectively (19). The RGG motif was modified by pseudophosphorylation of S651 (exchange to glutamic acid) while the dsRBMs were inactivated by the exchange of functionally important phenylalanines to alanines (10,19,38–40).
Performing analogous experiments as in Figure Figure11 with the dsHCV SL-3 RNA, we determined a second-order rate constant kon with NF90bF432,559A–NF45, which was only slightly reduced in comparison to that of NF90bwt–NF45, (Figure (Figure3A3A and Table Table2).2). While kcat was slightly increased (Table (Table2),2), the efficiency of the RNA-annealing reaction of NF90bF432,559A–NF45 was nevertheless comparable to that of the wild-type complex (Table (Table2).2). Conversely, the pseudophosphorylated complex NF90bS651E–NF45 showed an evident decrease in kon and kcat by 5- and 2-fold, respectively, i.e., the efficiency of the RAA of this variant was about a factor of 8 lower than that of the wild-type complex (Table (Table2).2). As these data indicated an important role of the C-terminal region of NF90 in the RAA, we further investigated the RCA of other, previously characterized NF90 variants, namely NF90bΔC63 and NF90bΔC46 (15). These contain major C-terminal truncations, which included (ΔC63) or did not include (ΔC46) the RGG motif (Figure (Figure3A).3A). Annealing experiments demonstrated that NF90bΔC63 did not accelerate the RNA annealing, while NF90bΔC46 showed a very weak catalysis (Table (Table2).2). Subsequent tests with the respective NF90bΔC63–NF45 and NF90bΔC46–NF45 heterodimers revealed that the presence of NF45, while increasing the efficiency of the variant's RAA, it did not restore the low efficiency of the NF90 variant's RCA to wild-type levels (Table (Table2).2). Hence, these findings supported and extended earlier observations demonstrating that the individual contributions of NF45 to RNA binding and RNA annealing is rather low (19) and that it requires full-length NF90 for most efficient RNA annealing. Furthermore, the data underpinned that NF90’s RNA-binding motifs contribute differently to the RCA.
Several of NF90’s native dsRNA targets are not fully double-stranded but contain internal or terminal mismatches. Accordingly, the next experiment investigated the RAA of NF90b/NF90b–NF45 with RNA substrates that show considerably different structural properties. To generate a fork-like (referred to as HCV-SL3-fork) or bulged structure (referred to as HCV-SL3-bulge; see Figure Figure3B),3B), we introduced 4 nt-substitutions at the termini and in the center of the dsHCV-SL3 RNA, respectively. Initially, we characterized the thermodynamic properties of these RNAs by performing thermal denaturation studies to determine the melting enthalpy (ΔHTM). As would be anticipated from the lower proportion of base pairings, both the fork and bulge RNAs showed significantly lower ΔHTM values than the dsHCV-SL3. Still, both modified dsRNAs showed thermodynamic stabilities, which were sufficient to perform annealing measurements under the applied assay conditions (see Supplementary Results and Table S2).
In ‘non-catalyzed’ RNA-only annealing reactions that applied the fork- or bulge substrates, the dependency of kobs on the RNA concentration turned out not to have a hyperbolic profile (data not shown). For that reason, we investigated only kon. This revealed that the non-catalyzed RNA-annealing reactions of HCV-SL3-bulge and HCV-SL3-fork proceeded with second-order rate constants that were ca. 30-fold reduced in comparison to the fully complementary dsHCV-SL3 (Table (Table33).
With HCV-SL3-fork, both NF90bwt and NF90bwt–NF45 accelerated the RNA-RNA interaction in a concentration-dependent manner but the corresponding kon-values were about 10- and 5-fold reduced with respect to the reactions that applied dsHCV-SL3 (Table (Table3).3). However, analogous to what we observed earlier with dsHCV-SL3, the heterodimer showed an increase in kon by about 100-fold as compared to the individual NF90. Interestingly, none of the proteins accelerated the RNA annealing of the HCV-SL3-bulge RNAs (for each reaction kobs = 0.01 s−1).
As these findings suggested a selectivity of NF90b in the RNA annealing reaction, we were prompted to examine a strand displacement (SD) activity with the modified RNA substrates. Considering that neither NF90bwt nor NF90bwt–NF45 performed a strand displacement on fully complementary substrates, we performed the assay in such a way that the mismatched strand was displaced by a fully complementary strand (see Figure Figure1A).1A). Thus, in the absence of protein, HCV-SL3-fork was not susceptible to a strand displacement, while HCV-SL3-bulge readily underwent displacement with a first-order rate constant of kobs = 0.02 s−1.
Strikingly, in the presence of NF90bwt or NF90bwt–NF45, we observed catalysis of the strand displacement reaction in a concentration-dependent manner on both RNA substrates (Table (Table33 and Supplementary Figure S3). With HCV-SL3-bulge, the kon-values that were measured with NF90bwt and NF90bwt–NF45 were very similar to those of the corresponding annealing reactions with dsHCV-SL3 (Table (Table3).3). In contrast, the rate constants and kon values that were measured for strand displacement with HCV-SL3-fork were much lower than those measured for the annealing of dsHCV-SL3: e.g. with NF90b–NF45 it was 250-fold lower (Table (Table3).3). As in the annealing reaction, the heterodimer showed strand displacement activities that were 10-fold (HCV-SL3-fork) or 35-fold (HCV-SL3-bulge) increased in comparison to those of the NF90 monomer. In conclusion, these experiments revealed the importance of structural features of the RNA substrates for the protein-catalyzed RNA annealing. Moreover, NF90b and NF90b–NF45 were demonstrated to anneal, but also to displace, RNA strands and, therefore, to meet both the criteria of RNA chaperones (see Introduction).
We next sought to define the underlying mechanism of the NF90-catalyzed reaction. For this, we performed Arrhenius-analyses (‘transition state-analyses’), which investigated the temperature-dependence of the RNA-annealing reactions in the absence and presence of protein. Apparent rate constants kobs were normalized to the RNA or protein concentrations according to Doetsch et al. and the calculation of parameters was done by using Equations (4)-(7) (41). Thus, ‘RNA only’ reactions with the diverse HCV-SL3 and RaSc RNA variants (ds, fork and bulge) revealed that these shared the same properties in their individual transition states under the applied conditions. The activation energies EA were between 4 and 7 kcal mol−1 and changes in entropy ΔS# ranged from –20 to –8 cal mol−1 K−1 (Supplementary Figure S4 and Table S3). Hence, we concluded that these data reflected a common situation for interactions of short, unstructured ssRNAs (41–43).
For the analyses of the protein-catalyzed reactions, we focused on the dsHCV-SL3 and RaSc substrates and tested the unmodified NF90b and NF90b–NF45 as well as the modified NF90bS651E–NF45 complex. Interestingly, in the presence of each of the protein preparations, the activation energy EA dropped to 0 kcal mol−1. Moreover, and independently of the applied RNA substrates, the reactions were accompanied by a large decrease in the change of entropy, in which NF90bwt showed the largest, NF90bwt–NF45 showed the lowest, and NF90bS651E–NF45 showed an intermediate magnitude in |ΔΔS#| (Figure (Figure3C3C and Table Table4;4; for data with RaSc see Supplementary Figure S4C). In close congruence with the earlier shown annealing and strand displacement data, this yielded a stabilization of the transition state (ΔG#) of 1 kcal mol−1 by NF90bwt–NF45 and only 0.2 kcal mol−1 by NF90bS651E–NF45. In the presence of NF90bwt, the transition state was actually destabilized by 0.6 kcal mol−1 as compared to the RNA-only situation (Table (Table4).4). Taken together, the data indicate that the RNA chaperone activity of NF90b depends on the presence of a functional C-terminus of the protein and on the formation of a heterodimer with NF45.
To gauge the biological relevance of the RCA of NF90, we addressed the protein's proposed role in the life cycle of HCV. As explained, NF90–NF45 was earlier shown to interact with the SL3 motif within the HCV genome's 3′UTR. SL3, together with SL2 and SL1 form the so-called ‘X-tail’ region of the HCV RNA at the immediate 3′-terminus, which is highly conserved among different HCV subtypes and essential for replication (44,45). The HCV 3′UTR additionally includes a variable polyU/UC stretch and a highly variable region at the 5′-end, which may form a labile stem-loop structure, termed here SLstop ((6); Figure Figure4A).4A). The 3′UTR was further shown to form complex, tertiary structures with RNA elements in the 3′-end of the viral protein-coding region. In particular, the stem-loop motif 5BSL3.2, which also interacts with the neighbored stem-loop 5BSL3.1 (46,47), forms a kissing-loop interaction with SL2 that is important for effective viral replication ((48,49); summarized by Sagan et al. (50); see Figures Figures4A4A and 5A).
NF90 was suggested to positively affect the first replication step of HCV, which, with all positive-strand RNA viruses involves the synthesis of a complementary negative-strand RNA intermediate by the viral RNA-dependent RNA polymerase (RdRp, (51)). Thus, to examine NF90 and NF90–NF45 for a direct effect on this step, we applied a well-established RdRp assay, which reconstitutes negative-strand RNA synthesis with the purified HCV RdRp NS5B and a viral RNA template in vitro (52–54). Recently, we used a similar approach to reveal the functional importance of an RCA activity of a cellular host factor involved in the replication of the HCV-related West Nile Flavivirus (28,55). The assay applied here was performed with an RNA template, which encodes the 3′UTR as well as the aforementioned 5BSL motifs of the HCV RNA (referred here to as ‘5BSL3_3UTR’; Figure Figure4A),4A), i.e. the template contained all cis-acting replicative elements (CRE) known to be essential for the de novo initiation of HCV negative-strand RNA synthesis (50). As control-templates served a segment of the HCV RNA that encodes exclusively the 3′X-tail region (‘X-tail’). As a positive control for the activity of NS5B, we applied poly(rC) RNA, which was known to represent a reasonable template of flaviviral polymerases (28,55).
Control ‘template-only’ RdRp assays confirmed that, with the 5BSL3_3UTR RNA and with the poly(rC) RNA nucleotide, incorporation/RNA synthesis was clearly detectable (Supplementary Figure S5). In contrast, the assays that were performed with the X-tail template or with the radiolabeled nucleotide alone yielded only background values indicating that a terminal transferase activity of the NS5B RdRp was undetectable under the chosen conditions (Supplementary Figure S5). Next, we pre-incubated the template RNAs with NF90bwt, with NF90bwt–NF45 or with the pseudophosphorylated NF90bS651E–NF45 and then started the RdRp reaction by adding the NS5B polymerase. Interestingly, with the 5BSL3_3UTR RNA and in the presence of NF90bwt–NF45, the RdRp-catalyzed negative-strand-RNA synthesis was significantly stimulated, with a maximal increase of 2.5-fold. At lower concentrations of complex, the stimulatory effect correlated with the amounts of complemented NF90bwt–NF45, while it was not detectable at very high concentrations of the heterodimer (Figure (Figure4B).4B). In contrast, neither NF90bwt nor NF90bS651E–NF45 displayed a stimulatory effect on negative-strand-RNA synthesis with the 5BSL3_3UTR RNA. Moreover, in assays that were performed with the X-tail or the poly(rC) templates, none of the proteins caused changes in product yields, which excluded a direct effect of NF90bwt–NF45 on the HCV NS5B RdRp (Figure (Figure4C4C–D).
These data demonstrated an intrinsic specificity of NF90bwt–NF45 as the complex stimulated the RdRp catalyzed synthesis of negative-strand RNA synthesis exclusively with the 5BSL3_3UTR template. Considering that NF90bwt as well as NF90bS651E–NF45 each showed a lower RCA than NF90bwt–NF45, the stimulatory effect on negative-strand-RNA synthesis correlated with the level of the protein's RNA chaperone activities.
The fact that the 3′-terminus of the HCV genome folds into complex secondary and tertiary RNA structures that involve the 3′UTR as well as the 3′-part of the NS5B coding region (Figure (Figure4A)4A) suggests that the initiation process of viral RNA synthesis by the RdRp is decisively determined by the RNA conformation (47,50,56). Along this line, we hypothesized that NF90bwt–NF45 should be more capable than NF90bwt and NF90bS651E–NF45 of changing the conformation of the viral RNA in such a way that this leads to an increase in replication-competent RNA conformers.
To address this, we decided to investigate the 5BSL3_3UTR RNA for global structural changes by analyzing its thermal melting behavior in the absence and presence of protein (57). Thus, the protein-free 5BSL3_3UTR RNA turned out to exhibit a flattened thermal denaturation profile (Figure (Figure5A5A and B), and the transition midpoint was determined at 68°C from the first derivative of the transition curve (Figure (Figure5C).5C). Interestingly, in the presence of NF90bwt–NF45, the thermal transition of the RNA exhibited at least two pronounced transition parts with midpoints at 58 and 68°C, respectively (Figure (Figure5B5B and C). A third transition was indicated at T > 80°C, but protein precipitation was observed at T > 85°C, which hampered further interpretation. With NF90bwt, similarly changed transition parts were detected, but, in comparison to NF90bwt–NF45, the transition that occurred at TM = 68°C was less pronounced, less cooperative, and the transition at T > 80°C was not detected (Figure (Figure5B5B and C). With NF90bS651E–NF45, both transitions at TM = 58 and 68°C were detectable, and the amplitude of the latter transition ranked in between that of NF90bwt–NF45 and that of NF90bwt. The transition at T > 80°C was also undetectable with NF90bS651E–NF45 (Figure (Figure5B5B and C).
Since SHAPE (selective 2-hydroxyl acylation analyzed by primer extension, (58)) experiments failed to unambiguously detect NF90–NF45 mediated conformational transitions in the 5BSL3_3UTR RNA (Supplementary Figure S6), we assumed that the protein complex targets higher order rather than secondary RNA structures. To clarify this further, we analysed the thermal melting of individual segments of the RNA (Figure (Figure5A)5A) in the absence of protein. Thus, RNA transcripts that contained only SL1, SL1 + SL2 or the X-tail (X) did not display a cooperative transition while a segment, termed here as UX, which contained the X-tail and the polyU/UC region, revealed a pronounced profile with midpoints at 58 and 68°C, respectively (Figure (Figure5D5D and E). A larger 3UTR segment, which additionally contained SLstop (see Figure Figure5A),5A), displayed a similar cooperative thermal transition with a less prominent midpoint at 58°C and a pronounced midpoint at 68°C. Thus, it turned out that the melting behavior of the RNA segments UX and 3UTR resembled that of the 5BSL3_3UTR RNA in the presence of NF90wt–NF45 (see Figure Figure5D5D and E). Since 5BSL3_3UTR RNA only did not follow this melting profile, this suggested that NF90–NF45 mainly affected the fold of the 3′-portion of the viral RNA (see Discussion).
In conclusion, it can be stated that an evident change in thermal melting of the 5BSL3_3UTR RNA occurred in the presence of each of the purified proteins. The least alterations were monitored with the NF90b monomer, while intermediate changes occurred with NF90bS651E–NF45, and most pronounced changes were detected with NF90bwt–NF45. As this order is congruent with the increasing efficiency of the protein's RCA, these findings strongly suggest a correlation between the protein's RCA and the observed structural changes in the viral RNA.
As already outlined, NF90 was demonstrated to determine VEGF protein synthesis under hypoxia by binding to a regulatory element, designated as ‘hypoxia stability region’, HSR, within the mRNA’s 3′UTR (Supplementary Figure S1A) (4). Interestingly, the HSR represents a riboswitch, i.e. its secondary structure may adopt at least two different conformations that display a lower or higher degree of base-pairing, the latter of which supports the translation of the mRNA (Figure (Figure6A).6A). NF90 was indicated to be an active part of this switch by stabilizing double-stranded conformations of the HSR (32). A key role in these transitions is played by a 63 nt element of the HSR, HSR-SL, which is capable of forming an extensive stem-loop structure (4; Figure Figure6A;6A; Supplementary Figure S1A). Thus, for a further understanding of the relevance of NF90–NF45’s RCA, the last set of experiments of this study aimed at determining its impact on the HSR-SL structure. For this, we followed the same approach as with the HCV-derived RNAs, namely by measuring transition curves in thermal melting experiments to define the thermodynamic parameters of HSR-SL in the absence and presence of NF90wt–NF45.
Since changes of the ionic strength affect base-paring and, accordingly, dsRNA formation (57,59,60), we measured first the thermal transitions of the HSR-SL at increasing ionic strengths in assay buffer only (Figure (Figure6B).6B). As expected, the melting temperature TM increased with increasing concentrations of NaCl from 44°C at 0 mM to 58°C at 400 mM (Figure (Figure6B).6B). The corresponding ΔHTM values, which were obtained by applying Equations (8) and (9) (see Materials and Methods), interestingly followed a non-linear but hyperbolic dependency on the NaCl concentration that reached a maximum at above 200 mM (Figure (Figure6C).6C). ΔHTM is proportional to the number of melting base pairs with a general approximation of ΔH = 4.6–6.1 kcal mol−1 per bp (61,62). In the absence of NaCl, the ΔHTM value of 71.5 kcal mol−1 accordingly corresponded to a melting of 11–15 bp, while at 400 mM NaCl, the ΔHTM value of 100 kcal mol−1 corresponded to a melting of 16–21 bp (Figure (Figure6C).6C). Thus, in agreement with earlier findings of Ray et al. (63), these data indicated that at concentrations below 200 mM NaCl, the predominant conformation of the HSR-SL lacks the formation of 5 bp (Figure (Figure6A).6A). Interestingly, when we next measured the thermal transition of the HSR SL in the presence of increasing concentrations of NF90wt–NF45 at 100 mM NaCl, the TM shifted from 50 to 58°C (Figure (Figure6D6D and E). As this value corresponded exactly to that of the thermodynamically more stable conformation of the HSR-SL in the absence of protein, we took these data as an indication that NF90wt–NF45 actively populates and stabilizes a HSR conformation that contains a higher degree of base pairing (Figure (Figure6A6A).
NF90 is an important post-transcriptional regulator of gene expression that exists predominantly in a complex with NF45. We showed earlier that the RNA-binding capacity of the heterodimer is considerably increased in comparison to the monomeric NF90. This was explained by conformational changes in NF90 upon binding of NF45 that increase the co-operative interplay between NF90’s RBMs. Phosphorylation of Ser-651 within the C-terminal RGG/RG region was suggested to inhibit these conformational changes and demonstrated to be a clear-cut negative regulator of RNA-binding by NF90 and NF90–NF45 (15,19). During heterodimer formation, we furthermore observed a mutual stabilization of the proteins, which, however, could not explain the improved RNA-binding affinities (19).
Here, we obtained unequivocal evidence that NF90 has an inherent RNA chaperone activity, which functions particularly to anneal RNA. Interestingly, the RNA-annealing activity of NF90 also significantly improves by interaction with NF45, i.e. the efficiency of ssRNA hybridization by NF90–NF45 was found to be 50-fold increased in comparison to the monomeric NF90. On the one hand, we explained this by the improved RNA-binding affinity of the complex (19). On the other hand, we observed that complex formation stabilizes structural elements of NF90 that contribute to the RAA and acquired evidence that heterodimerization directly improves catalytic steps of the reaction (Figures (Figures22 and 3, Table Table4).4). In sum, these data provided solid evidence that the formation of the NF90–NF45 complex aids NF90’s RCA.
Interestingly, both the NF90–NF45 heterodimer, as well as the monomeric NF90, display different RAA efficiencies on two different RNA pairs (Figure (Figure1),1), which apparently arise from different on-rates (kon) of substrate binding. This suggested a putative specificity in the reaction, and this idea was further strengthened when we performed the annealing assays with partly mismatching ssRNA substrates. These experiments demonstrated that structural features of the formed dsRNAs considerably affect the efficiency of the protein-catalyzed RNA-annealing reactions. NF90 and NF90–NF45 were further revealed to perform strand-displacement reactions only on mismatched dsRNAs to yield dsRNAs with a higher degree of base-pairing (Figure (Figure3).3). NF90 and NF90–NF45 are accordingly suggested to function as substrate-selective RNA chaperones, which require a minimum region of consecutively formed base pairs in order to catalyze RNA annealing in the most efficient way.
The mechanisms of non-catalyzed RNA-annealing reactions of short, unstructured ssRNAs were earlier characterized to involve charge-repulsion (ΔH# > 0) and a decrease in microstate dynamics (ΔS# < 0), (41–43,64). We observed similar characteristic for RNA substrates used in this study. In contrast, in the presence of the different NF90 variants, a complete compensation of charge-repulsion was achieved (ΔH# ~ 0 kcal mol−1), and the dynamics of the system decreased even further (ΔS#NF90 << ΔS#RNA). A similar situation was described for the RNA-annealing protein gBP21, which stabilizes ssRNA by compensating the negative charges of the RNA’s backbone (ΔH# ~ 0 kcal mol−1) and induces local conformational changes within the ssRNA. The activity of gBP21 was interpreted as a ‘matchmaker mechanism’, i.e. the protein was proposed to render the ssRNA annealing-competent upon binding (64). Considering the earlier analyses of NF90’s RNA binding properties (15), our data point to a highly rigid protein–RNA complex as the RNA-annealing reaction progresses and suggest that these complexes form a closed conformation-like state during the reaction (shown as a model in Figure Figure7).7). This would support a matchmaker mechanism by providing an annealing-competent microenvironment and result in an increase in kcat. Interestingly, this was observed only for the catalysis by variants of NF90–NF45, supporting the conclusion that the NF90–NF45 interaction specifically fosters the formation of a closed conformation, as suggested earlier (19), and in this way directly facilitates the provision of an annealing-competent system (Figure (Figure7).7). The importance of the C-terminal region of NF90 during these processes is highlighted by the observation that the pseudo-phosphorylated NF90bS651E–NF45 displays a lower tendency to form such a rigid protein-RNA complex and that it has a low RCA. In accord with this, NF90bΔC63 does not show an RAA. Conversely, the dsRBM-impaired NF90bF432,559A–NF45 has a wild-type-like activity indicating that NF90’s dsRBMs do not contribute to the actual catalysis of RNA annealing. However, considering the capability of dsRBMs to bind specifically A-helical RNA, it is conceivable that these motifs play a role after the rate-limiting step, for example, by stabilizing the dsRNA product (Figure (Figure77).
The significance of the RCA of NF90 became apparent when we investigated the role of NF90 and NF90–NF45 during the initial step of the RNA replication process of HCV, the synthesis of viral negative-strand replicative intermediate, which can be reasonably reconstituted in vitro. The key enzyme catalyzing HCV RNA replication is the NS5B RdRp (52). Although the binding process of NS5B to HCV RNA and the template-dependent production of dsRNA product were studied already in detail (50,65–68), the molecular determinants that contribute to the de novo initiation of HCV RNA synthesis are not yet completely characterized. For example, the NS5B ‘promoter region’ for the initiation of negative-strand RNA synthesis has not yet been defined. Our thermal melting assays with the 5BSL3_3UTR RNA template revealed that NF90bwt, the pseudo-phosphorylated NF90bS651E–NF45, as well as NF90bwt–NF45, each significantly alter the RNA’s conformation. As explained, the extent of these changes firmly correlated with the RCA of NF90. It was most pronounced with NF90bwt–NF45, which, exclusively, stimulates NS5B-mediated RNA synthesis (Figure (Figure4).4). Thus, it is tempting to conclude that a main function of the RCA of NF90–NF45 involves the population of a replication-permissive HCV RNA conformer. Along this line, NF90–NF45 may cooperate with other proteins such as EWSR1, which was proposed as having a similar function (69).
Currently, we can only speculate about how NF90bwt–NF45-guided structural alterations in the viral RNA stimulate HCV RNA synthesis. While we cannot exactly define the RNA elements that are modulated by NF90–NF45, the thermal melting data suggest a scenario in which the X-tail region of the 5BSL3_3UTR RNA forms a conformation that is unraveled by NF90–NF45. Considering the fact that the SHAPE reactivity of the RNA remained unchanged in the presence of the proteins (Supplementary Figure S6), the activity of NF90–NF45 was suggested to affect the RNA’s tertiary structure, while the secondary structure elements are maintained. This idea is supported by the observation that those RNA transcripts, which contained the X-tail and the poly U stretch (UX and 3UTR), exhibited a similar melting profile in the absence of protein as the 5BSL3_3UTR RNA in the presence of the NF90–NF45 complex (Figure (Figure5).5). In the absence of protein, the melting profile of 5BSL3_3UTR was not indicative of major transitions. Earlier protein–RNA interaction studies indeed revealed NF90–NF45 to associate with SL3 within the X-tail region, but only in the presence of polyU/UC (6).
Strikingly, the stimulatory effect of NF90bwt–NF45 on NS5B-mediated RNA synthesis was only detectable with RNA templates that contained all cis-replicative elements (CRE) within the 3′-end of the viral ORF and the 3′UTR, which were characterized to be essential for effective viral RNA replication (46–49). Besides excluding a direct effect of NF90–NF45 on NS5B, these data together with the melting analyses fuel the speculation that the heterodimer changes the earlier-mentioned pseudoknot interactions of the stem-loop motif 5BSL3.2 with the X-tail region of the HCV RNA, which, in turn, may increase the efficiency of negative-strand RNA synthesis by the viral polymerase. This idea is supported by our data as well as by findings of other laboratories, which showed that the polymerase acts efficiently on RNA molecules that include the CRE, while the X-tail alone is a poor template (70–72). The complex tertiary structure formed by the CRE within the 3′-end of the viral ORF and the 3′UTR may be accordingly suggested as an important element of the NS5B promotor. The situation of the HCV 3′-end and the activity of NF90–NF45 could resemble what has been described with plant-infecting Tombusviruses. The genomic 3′-ends of these viruses encode a so-called ‘replication silencer element’ (RSE), a pseudoknot structure, which masks the promoter of the Tombusvirus RdRp (73). The Tombusvirus-encoded protein p33 was indicated to act as an RNA chaperone that rearranges the RSE to enable the polymerase to initiate negative-strand RNA synthesis efficiently (27).
Our findings further substantiate the idea that NF90 and NF90–NF45 have distinct functions in the cell. As NF45 significantly modulates the RNA-binding specificity (19), as well as the RNA-annealing activity of NF90 (this report), it may be reasonable to assume that most of the previously described functions of NF90 are exerted by the NF90–NF45 heterodimer. This notion fits to the earlier proposal of Guan et al. that NF45 has an important, hitherto disregarded, regulatory function by adjusting the cellular level and by steering the activity of the NF90–NF45 complex (17).
Combined with our earlier findings, the data presented here provide a general understanding of the activity of NF90–NF45. The modular architecture of NF90 facilitates a cooperative interplay of its RNA-binding domains, which enables specific RNA binding and a discrimination of structural RNA features (19). In turn, the mode of RNA-binding of NF90 determines its RCA by which the protein may change the conformation of the bound RNA. In this way, the heterodimer may affect inter- as well as intramolecular RNA-RNA interactions and accelerate cellular processes, which are otherwise trapped by non-functional or incorrect RNA structures. As several RNA targets of NF90–NF45 are likely to contain highly structured regions, association of the complex might also change the RNA’s dynamics or even the distribution between different conformers of the bound RNA.
An interesting example, which was also addressed here, is the VEGF mRNA, where NF90–NF45 was suggested to bind to the HSR riboswitch element (4,63). Ray et al. obtained experimental evidence that, under hypoxia, this region transitions from a translation-silencing into a translation-permissive conformation (63), and studies of Yao et al. further indicated that NF90, being part of the HILDA protein complex, mediates the flipping of the riboswitch (32). In direct support of this, our data suggest that NF90wt–NF45 changes the structure of the HSR element in vitro such that, according to its substrate-selective RCA, the resulting conformer contains an increased number of base pairs. An increase of the double-stranded fraction of the HSR riboswitch was earlier shown to support the translation-permissive state of the VEGF mRNA (63).
A similar concept to what we propose here for NF90–NF45 was recently suggested to occur during early steps of ribosome assembly, where the ordered association of protein components to the maturing ribosome appears to be orchestrated by sequential protein-mediated conformations of the 16S ribosomal RNA (74). Protein-guided RNA dynamics were also shown to regulate processes in prokaryotes. The E. coli protein Hfq, one of the best-investigated proteins with RNA chaperone activity, was shown to be an efficient RNA ‘annealer’ (75) and demonstrated to crucially affect the efficiency of translation initiation and stability of target mRNAs, both by hybridizing small regulatory RNAs to their target regions, as well as by stabilizing these complexes (76).
Very recently, NF90 was shown to promote the biogenesis of circular RNAs by binding to complementary intronic RNA pairs juxtaposing the exons that form the mature circular RNA (77). We suppose that the RCA of NF90–NF45 is critically involved in this process.
In conclusion, our data strongly support the opinion that the RCA of NF90 represents a major determinant of the protein's diverse biological activities. As shown here, the level of the RCA is regulated by post-translational processes such as phosphorylation, and by the association of NF45.
We thank Gary Sawers for critically reading of the manuscript and Christine Hamann for excellent technical assistance.
Present address: Tobias Schmidt, Medical Research Council Toxicology Unit, Hodgkin Building, Lancaster Road, Leicester LE1 9HN, UK.
Supplementary Data are available at NAR Online.
Deutsche Forschungsgemeinschaft [GRK 1026 to T.S., R.G.; BE1885/7, BE1885/12-1 to S.F., S.E.B.]; Martin Luther University Halle-Wittenberg (to R.G., S.F., S.E.B.). Funding for open access charge: Deutsche Forschungsgemeinschaft, Martin Luther University Halle-Wittenberg.
Conflict of interest statement. None declared.