The Amyloid Region of Hfq Riboregulator Promotes DsrA:rpoS RNAs Annealing

Simple Summary RNA:amyloid protein interactions have been observed in the past few years. Nevertheless, the molecular basis and physiological implications of these interactions are still poorly understood. Here we focus on a bacterial amyloid protein, Hfq. This protein is a pleiotropic bacterial regulator that mediates many aspects of RNA metabolism. The protein notably mediates mRNA stability and translation efficiency by using stress-related small noncoding regulatory RNAs. This regulation contributes to bacterial adaptation to stresses. Our results show that the amyloid region of Hfq significantly influences the efficiency of annealing between DsrA small noncoding RNA to its target mRNA. This unexpected result opens perspectives for a novel physiological role of amyloids, including those associated with neurodegenerative diseases. Abstract Hfq is a bacterial RNA chaperone which promotes the pairing of small noncoding RNAs to target mRNAs, allowing post-transcriptional regulation. This RNA annealing activity has been attributed for years to the N-terminal region of the protein that forms a toroidal structure with a typical Sm-fold. Nevertheless, many Hfqs, including that of Escherichia coli, have a C-terminal region with unclear functions. Here we use a biophysical approach, Synchrotron Radiation Circular Dichroism (SRCD), to probe the interaction of the E. coli Hfq C-terminal amyloid region with RNA and its effect on RNA annealing. This C-terminal region of Hfq, which has been described to be dispensable for sRNA:mRNA annealing, has an unexpected and significant effect on this activity. The functional consequences of this novel property of the amyloid region of Hfq in relation to physiological stress are discussed.


Introduction
Bacteria are adapted to function in their normal physiological environment. Any change in environmental conditions such as temperature, pH, nutrients starvation, salts, and oxidation inflict stresses on bacteria [1,2]. Many regulatory systems help to respond to these changes and post-transcriptional regulation of mRNA translation and stability provides a rapid and efficient mechanism [3][4][5]. The Hfq protein, a chaperone for small noncoding RNAs (sRNAs), is considered a core component of a global post-transcriptional network in bacteria [6][7][8][9]. Hfq was discovered as a host factor required for replication of the bacteriophage Qβ in Escherichia coli [10,11]. Approximately 50% of sequenced bacterial genomes contain at least one gene coding for Hfq [12]. As a global regulator of E. coli metabolism, deletion of the hfq gene can cause pleiotropic effects, such as decreased growth rate, maladaptation to stress, altered cellular morphology, and increased cell length [13,14]. unfolded to base pair with the rpoS mRNA without Hfq at room temperature [59]. These results indicate that Hfq may play an important role in the unfolding of DsrA SL1.
Here, we used a biophysical approach, Synchrotron Radiation Circular Dichroism (SRCD), to study the effect of the Hfq-CTR on DsrA SL1 annealing to rpoS mRNA. Following RNA annealing using ElectroMobility Shift Assay (EMSA) is usually challenging for polymerizing proteins, as in the case of amyloids [64]. SRCD thus provides a new useful tool to observe RNA annealing in real time by amyloids (and in general by proteins that polymerize), but also to characterize annealing precisely. SRCD allows differentiation base pairing and base-stacking in a double-stranded helical transition [65]. Unexpected effects of the Hfq CTR amyloid-like region on the RNA annealing process are presented herein. This result reveals a perspective for a novel physiological role of amyloids, including those associated with neurodegenerative diseases.

Hfq CTR Peptide and Protein
Full-length Hfq and NTR were purified as described previously [66]. Hfq-CTR peptide was chemically synthetized (Proteogenix, Schiltigheim, France). This peptide corresponds to the amyloid CTR domain of Hfq (residues 64 to 102) and is referred to as Hfq-CTR throughout the manuscript. This peptide cannot be purified from E. coli bacteria as it is unstable when translated independently of the NTR region. The sequence of Hfq-CTR is SRPVSHHSNNAGGGTSSNYHHGSSAQNTSAQQDSEETE [43]. Before use, Hfq-CTR peptide was reconstituted in water at 20 mg/mL. We determined that the pH used in our condition (~5) was the most appropriate to form the complex with RNA. Indeed, the positive charge of the peptide at pH 5 allows its interaction with RNA, while increasing pH reduces or abolishes this interaction. We also chose to avoid the addition of salts (except those already present in RNA, protein, and peptide solutions) in order to allow a better investigation in deep-UV [45]. When the complex is analyzed in the presence of salts (or far-UV absorbing buffers), the spectral bandwidth accessible is limited, reducing the spectral information obtained [67]. We checked that presence of salts (NaCl 50 mM) does not change significantly the kinetics of annealing.

DsrA and rpoS Sequences
For SRCD analysis, we focused on the core of the RNA-RNA interactions and used DsrA and rpoS fragments described in Hwang et al. [68]. Two fragments of DsrA (DsrA core WT and mutant) and two fragments of rpoS (rpoS rbs and rpoS reg ) were used ( Figure 1).
Before use, the oligonucleotides (Eurogentec) were heated at 90 • C for 3 min and then slowly cooled down at 20 • C to allow proper folding. Duplexes of oligonucleotides (i.e., DsrA core :rpoS reg , rpoS rbs :rpoS reg . . . ) were formed using the same protocol after stoichiometric addition of the two oligonucleotides in the same tube. Base pairing of DsrA with the leader sequence of rpoS enhances its translation by exposing the ribosome binding site (rbs) and start codon (AUG), which are sequestered in the inhibitory stem structure. This process requires the RNA chaperone Hfq. Until now, Hfq-CTR was described as dispensable for this activity.
Before use, the oligonucleotides (Eurogentec) were heated at 90 °C for 3 min and then slowly cooled down at 20 °C to allow proper folding. Duplexes of oligonucleotides (i.e., DsrAcore:rpoSreg, rpoSrbs:rpoSreg…) were formed using the same protocol after stoichiometric addition of the two oligonucleotides in the same tube.

Synchrotron Radiation Circular Dichroism (SRCD)
SRCD measurements were carried out on the DISCO beamline at SOLEIL Synchrotron as described previously [45] (proposal 20201013). Samples (~4 µL) were loaded into a CaF2 circular cell of 10 µm pathlength. Spectral acquisitions of 1 nm steps at 1.2 s integration time were recorded in triplicates between 320 and 180 nm. (+)-camphor-10-sulfonic acid (CSA) was used to calibrate amplitudes and wavelength positions of the experiment. Data analyses (averaging, baseline subtraction, smoothing, scaling, and spectral summations) were carried out with CDtoolX [69]. Spectra are presented in units of mdeg versus nm maintaining the same molar ratios for all presented samples. Due to the origin of absorption, spectra of mixed samples (polynucleotides + peptide) could not be standardized to ∆ε.
Two types of experiments have been performed: the effect of Hfq or CTR were tested either on individual RNA, or on duplexes such as DsrAcore:rpoSreg, rpoSrbs:rpoSreg. For all SRCD experiments, RNA concentrations were fixed at 20 mM (while Hfq or CTR (monomeric) concentration were 0.2 or 0.4 mM, respectively (full-length Hfq concentration was lower due to its lower solubility). RNAs were in excess relative to proteins. The effect of Hfq or CTR on RNAs was analyzed at 15 °C. For kinetic measurements, an apparent catalytic Base pairing of DsrA with the leader sequence of rpoS enhances its translation by exposing the ribosome binding site (rbs) and start codon (AUG), which are sequestered in the inhibitory stem structure. This process requires the RNA chaperone Hfq. Until now, Hfq-CTR was described as dispensable for this activity.

Synchrotron Radiation Circular Dichroism (SRCD)
SRCD measurements were carried out on the DISCO beamline at SOLEIL Synchrotron as described previously [45] (proposal 20201013). Samples (~4 µL) were loaded into a CaF 2 circular cell of 10 µm pathlength. Spectral acquisitions of 1 nm steps at 1.2 s integration time were recorded in triplicates between 320 and 180 nm. (+)-camphor-10-sulfonic acid (CSA) was used to calibrate amplitudes and wavelength positions of the experiment. Data analyses (averaging, baseline subtraction, smoothing, scaling, and spectral summations) were carried out with CDtoolX [69]. Spectra are presented in units of mdeg versus nm maintaining the same molar ratios for all presented samples. Due to the origin of absorption, spectra of mixed samples (polynucleotides + peptide) could not be standardized to ∆ε.
Two types of experiments have been performed: the effect of Hfq or CTR were tested either on individual RNA, or on duplexes such as DsrA core :rpoS reg , rpoS rbs :rpoS reg . For all SRCD experiments, RNA concentrations were fixed at 20 mM (while Hfq or CTR (monomeric) concentration were 0.2 or 0.4 mM, respectively (full-length Hfq concentration was lower due to its lower solubility). RNAs were in excess relative to proteins. The effect of Hfq or CTR on RNAs was analyzed at 15 • C. For kinetic measurements, an apparent catalytic kinetic constant k cat app was determined during initial rate conditions. This constant was expressed in mdeg·M −1 ·min −1 and depends on protein concentration (CD units cannot be normalized to protein concentration as they are not expressed in the same unit, mdeg vs. M).
For melting curves, triplet SRCD spectra were acquired every 3 • C between 15 • C and 81 • C. Averaged SRCD values of the maximum of the peak around 180 nm are presented as a function of temperature to measure the melting point (T m ). Note that a shift of the peak may occur in some cases and that the maximum is not always precisely at 180 nm [70] (Supplementary Materials Figure S1). A Boltzmann sigmoid equation, which assumes a two-state model, was used for fitting of melting curves: y = Bottom + (top − bottom)/(1 + e((T m − x)/slope)).

ElectroMobility Shift Assay (EMSA)
The Electromobility Shift Assays were performed with labelled RNAs (Cy3-DsrA core and Cy5-rpoS reg ) or unlabeled RNA (DsrA core , DsrA core mut ). After incubation with CTR for one hour, samples were loaded on a polyacrylamide gel (10% or 4-20%) and subjected to electrophoresis in Tris-Acetate-EDTA (TAE) buffer. The gels were scanned using a fluorescence imager.

The Amyloid CTR Region of Hfq Triggers RNA Annealing
The effect of Hfq-CTR on RNA annealing was analyzed at 15 • C. This low temperature is physiologically relevant because DsrA is expressed at low temperature allowing a cold-shock response [71]. In addition, measurements at 15 • C avoid spontaneous annealing between RNA molecules occurring at 37 • C with RNA fragments (even if this spontaneous annealing at 37 • C is slow). Kinetics of annealing at 15 • C were thus longer than at 37 • C and were recorded for few hours to ensure accurate measurements. As seen in Figure 2, we clearly observe an effect of Hfq-CTR on sRNA:mRNA annealing, shown by the different increase of amplitudes observed at 180 nm and 260 nm with and without the Hfq-CTR. Indeed, the spectral region between 320 and 170 nm contains several electronic transitions of interest. Around 260 nm, the positive CD signal shows base-pairing and base-stacking [72,73]. On the other hand, the positive band at 185 nm and the negative one between 200 and 210 nm are indicative for the formation of double-stranded right-handed RNA molecules (A-and B-forms). The increase of amplitudes of these peaks signifies the formation of base pairs and right-handed double stranded RNA. In addition, denaturation of the RNA induces a shift of the spectral maximum from~180 to~192 nm [70]. Spectroscopically the batho-chromic shifting (red-shift) of a maximum corresponds to a change (reduction) of absorption energy of the chromophore (RNA). This shift therefore indicates a clear structural change or the existence of two states of the chromophore and can also be used to confirm the structural change of RNAs (Supplementary Materials Figure S1). Therefore, the dynamics of RNA annealing can be qualitatively and quantitatively followed. This effect of Hfq-CTR on RNA annealing was also confirmed using EMSA (Supplementary Materials Figure S2). Nevertheless, due to the high molecular weight of the complex with the amyloid form of the CTR, that stays on the top of the gel, it is difficult to discriminate solely with an EMSA the complex (i.e., triplex) CTR:Cy3-DsrA core :Cy5-rpoS reg from CTR:Cy3-DsrA core and CTR:Cy5-rpoS reg duplexes. They all migrate at the same position.
In addition, precise kinetics measurements are not possible using EMSA as they are with SRCD analysis.
SRCD kinetics of rpoS reg annealing have been measured using SRCD for WT DsrA core and for a mutated form called DsrA core mut , harboring a mutated tetraloop (GGAA vs. AUUUC pentaloop) (Figure 1). DsrA core forms a stem-loop structure with a typical A-form helix and with a dynamic AUUUC pentaloop [59]. Many other pentaloops can form stable structures with unusual interactions, while AUUUC pentatloop is unstructured. To test whether the unstructured dynamic AUUUC pentaloop is important for the annealing, we replaced it with a very stable GGAA tetraloop. At 180 nm, an apparent Hfq-CTR catalytic kinetic constant k cat app of 0.083 mdeg·M −1 ·min −1 for DsrA core was determined compared with 0.0173 mdeg·M −1 ·min −1 for DsrA core mut (Figure 3). In comparison, for full-length Hfq the kinetics constant k cat app measured was for 0.039 mdeg·M −1 ·min −1 for DsrA core (Supplementary Materials Figure S3); nevertheless, Hfq was two-times less concentrated than Hfq-CTR, thus CTR seems to have an activity similar to that of the full-length protein to promote DsrA core :rpoS reg RNA annealing. Note that when measured at 260 nm, the k cat app measured was 0.0257 mdeg·M −1 ·min −1 for DsrA core and 0.0069 mdeg·M −1 ·min −1 for DsrA core mut , thus with a similar decrease by a factor~5 compared with WT DsrA core . This suggests that the formation of base pairs/base stacking and helical structure during annealing occurs at the same rate regardless of the sequence of the loop. We also tried to analyze Hfq-NTR in SRCD experiments but failed to have this form of the protein at Biology 2021, 10, 900 6 of 14 a concentration sufficient to measure annealing kinetics with SRCD. This shows a limit of the technique: for SRCD, short pathlengths (two to several tenths of microns) are used permitting very small loading volumes (3-4 µL). Therefore, high concentrations are chosen following Beer-Lambert law. This has the advantage of reducing overall precious sample consumption as well as extending the spectral-band down to 170 nm. With Hfq-NTR we could not reach a concentration of~0.2 mM and increasing its concentration clearly promoted precipitation of the protein which is incompatible with SRCD measurements because it results in diffusion of the UV light. We tried to measure the annealing activity of Hfq-NTR using lower concentrations but, unfortunately, were not able to detect a significant signal change for SRCD, because the signal to noise ratios were dropping considerably below 10 in contrast to above 200 for high concentration and short pathlengths. We thus presume NTR activity is significantly lower than that of the full-length protein, even if this region is able to promote annealing in vivo [38]. This result is in agreement with the significant activity of the CTR. SRCD kinetics of rpoSreg annealing have been measured using SRCD for WT Ds and for a mutated form called DsrAcore mut , harboring a mutated tetraloop (GGAA vs UUC pentaloop) (Figure 1). DsrAcore forms a stem-loop structure with a typical Ahelix and with a dynamic AUUUC pentaloop [59]. Many other pentaloops can form s structures with unusual interactions, while AUUUC pentatloop is unstructured. T whether the unstructured dynamic AUUUC pentaloop is important for the annealin replaced it with a very stable GGAA tetraloop. At 180 nm, an apparent Hfq-CTR cat kinetic constant kcat app of 0.083 mdeg·M −1 ·min −1 for DsrAcore was determined compared 0.0173 mdeg·M −1 ·min −1 for DsrAcore mut (Figure 3). In comparison, for full-length Hf kinetics constant kcat app measured was for 0.039 mdeg·M −1 ·min −1 for DsrAcore (Supple tary Materials Figure S3); nevertheless, Hfq was two-times less concentrated than CTR, thus CTR seems to have an activity similar to that of the full-length protein to mote DsrAcore:rpoSreg RNA annealing. Note that when measured at 260 nm, the kcat app m ured was 0.0257 mdeg·M −1 ·min −1 for DsrAcore and 0.0069 mdeg·M −1 ·min −1 for DsrAc Wavelength (nm) Wavelength (nm) CD (mdeg) CD (mdeg) Figure 2. Observation of DsrA core and rpoS reg annealing by Hfq-CTR using SRCD (from blue t = 0 min to red t = 300 min). Inset: control without the protein; no significant RNA annealing occurs without the protein.
The effect of Hfq and Hfq-CTR on the stability of the complex formed was evaluated using meting curves. As seen in Figure 4 and Table 1, the DsrA core :rpoS reg complex is more stable in the presence of both Hfq and CTR. This confirms that the full protein stabilizes the sRNA:mRNA complex, but this result was unexpected for the CTR.
Note that DsrA core mut is significantly more stable than DsrA core in the absence or in the presence of CTR (Table 1). This effect is probably due to the presence of the tetraloop vs. the pentaloop. As shown in Figure 3, the effect of the replacement of the pentaloop by a tetraloop affects the rate of annealing that is decreased by a factor 5. We next analyzed the complex between CTR and DsrA core or DsrA core mut in the absence of rpoS.
centrations but, unfortunately, were not able to detect a significant signal change for SRCD, because the signal to noise ratios were dropping considerably below 10 in contrast to above 200 for high concentration and short pathlengths. We thus presume NTR activity is significantly lower than that of the full-length protein, even if this region is able to promote annealing in vivo [38]. This result is in agreement with the significant activity of the CTR. The effect of Hfq and Hfq-CTR on the stability of the complex formed was evaluated using meting curves. As seen in Figure 4 and Table 1, the DsrAcore:rpoSreg complex is more stable in the presence of both Hfq and CTR. This confirms that the full protein stabilizes the sRNA:mRNA complex, but this result was unexpected for the CTR.  . Melting curves at ~180 nm of DsrAcore and rpoSreg complexes annealed without protein (red), with full-length Hfq (green), with Hfq-CTR (blue), and heat pre-hybrided DsrAcore:rpoSreg without protein (black), respectively. For a better comparison of various sigmoids, they were normalized between 0% and 100%. The corresponding Tm were 36.6 ± 0.8 °C, 37.0 ± 0.9 °C, 33.6 ± 2.2 °C, and 39.9 ± 0.4 °C.  . Melting curves at~180 nm of DsrA core and rpoS reg complexes annealed without protein (red), with full-length Hfq (green), with Hfq-CTR (blue), and heat pre-hybrided DsrA core :rpoS reg without protein (black), respectively. For a better comparison of various sigmoids, they were normalized between 0% and 100%. The corresponding T m were 36.6 ± 0.8 • C, 37.0 ± 0.9 • C, 33.6 ± 2.2 • C, and 39.9 ± 0.4 • C.

Hfq-CTR Stabilizes DsrA Secondary Structure
One possibility is that annealing efficiency depends on the DsrA stem-loop which must be melted to associate with rpoS reg (Figure 1). We thus tested the effect of Hfq-CTR on DsrA core stability. Unexpectedly, we observed that Hfq-CTR does not melt DsrA core SL1, but in contrast it stabilized it significantly ( Figure 5). This can be observed by the increase of peaks at 180 and 270 nm ( Figure 6) and by T m measurements (Table 1). This is in contrast with DsrA core mut where CTR does not have any effect ( Figure 6). As the pentaloop of DsrA-SL1 is similar to single strand RNA (e.g., like AU 6 A in the linker between SL1 and SL2), we suspect this loop is important to form the complex. This is confirmed using EMSA where we show that Hfq-CTR binds DsrA core mut but less than DsrA core (Supplementary Materials Figure S4). The SL1 pentaloop is likely an important determinant for rpoS annealing to DsrA by Hfq-CTR, as it is the case of full Hfq [59]. with DsrAcore mut where CTR does not have any effect ( Figure 6). As the pentaloop of DsrA-SL1 is similar to single strand RNA (e.g., like AU6A in the linker between SL1 and SL2), we suspect this loop is important to form the complex. This is confirmed using EMSA where we show that Hfq-CTR binds DsrAcore mut but less than DsrAcore (Supplementary Materials Figure S4). The SL1 pentaloop is likely an important determinant for rpoS annealing to DsrA by Hfq-CTR, as it is the case of full Hfq [59].  Effect of Hfq-CTR on DsrAcore mut spectral evolution. In this case no significant stabilization is observed. Note the amplitude of 18 mdeg measured at 180 nm. As DsrAcore ( Figure 5) and DsrAcore mut were used at the same concentration and pathlength, this confirms that DsrAcore mut has a stronger (3-fold) structuration than DsrAcore. This was expected due to the presence of the structured tetraloop.

Hfq CTR Does Not Affect rpoS Secondary Structure
Another possibility to explain the effect of Hfq-CTR would be that it unwinds the rpoS stem-loop. This possibility was tested using a prehybrided rpoSreg:rpoSrbs. As shown in Figure 7, Hfq-CTR does not have any significant effect on the rpoSreg:rpoSrbs complex and thus does not influence rpoS structure. This was also confirmed by Tm measurement; Hfq-CTR does not affect rpoSreg:rpoSrbs stability (Table 1).

Figure 7.
Observation of the effect of Hfq-CTR on rpoSreg:rpoSrbs stability using SRCD (from blue t = 0 min to yellow t = 300 min). No significant stabilization is observed.

Wavelength (nm) CD (mdeg)
Wavelength (nm) Figure 6. Effect of Hfq-CTR on DsrA core mut spectral evolution. In this case no significant stabilization is observed. Note the amplitude of 18 mdeg measured at 180 nm. As DsrA core ( Figure 5) and DsrA core mut were used at the same concentration and pathlength, this confirms that DsrA core mut has a stronger (3-fold) structuration than DsrA core . This was expected due to the presence of the structured tetraloop.

Hfq CTR Does Not Affect rpoS Secondary Structure
Another possibility to explain the effect of Hfq-CTR would be that it unwinds the rpoS stem-loop. This possibility was tested using a prehybrided rpoS reg :rpoS rbs . As shown in Figure 7, Hfq-CTR does not have any significant effect on the rpoS reg :rpoS rbs complex and thus does not influence rpoS structure. This was also confirmed by T m measurement; Hfq-CTR does not affect rpoS reg :rpoS rbs stability (Table 1).
Biology 2021, 10, x 10 of 15 Figure 6. Effect of Hfq-CTR on DsrAcore mut spectral evolution. In this case no significant stabilization is observed. Note the amplitude of 18 mdeg measured at 180 nm. As DsrAcore ( Figure 5) and DsrAcore mut were used at the same concentration and pathlength, this confirms that DsrAcore mut has a stronger (3-fold) structuration than DsrAcore. This was expected due to the presence of the structured tetraloop.

Hfq CTR Does Not Affect rpoS Secondary Structure
Another possibility to explain the effect of Hfq-CTR would be that it unwinds the rpoS stem-loop. This possibility was tested using a prehybrided rpoSreg:rpoSrbs. As shown in Figure 7, Hfq-CTR does not have any significant effect on the rpoSreg:rpoSrbs complex and thus does not influence rpoS structure. This was also confirmed by Tm measurement; Hfq-CTR does not affect rpoSreg:rpoSrbs stability (Table 1).

Discussion
Using SRCD analysis, we show herein that, unexpectedly, the amyloid region of Hfq-CTR is able to promote DrsA:rpoS annealing with an efficiency similar to that of the full-length protein at 15 • C. This work thus provides a new useful tool to analyze label-free RNA annealing by an amyloid protein, an analysis that is not trivial using classical EMSA (Supplementary Materials Figure S2). This spectroscopic analysis provides an alternative to fluorescence measurements without RNA labelling, that in addition gives information on base-pairing, base-stacking, and type of helix formed [72,73]. In turn, SRCD could also allow observation of the formation of the amyloid structures [45].
We show that the effect of Hfq-CTR is not due to the melting of the SL1 loop of DsrA ( Figure 5) nor to the melting of the rpoS stem (Figure 6), which contrasts with the mechanism observed for the full Hfq including the Sm core [62]. Indeed, ribonuclease footprinting shows full Hfq binds the AUUUC DsrA pentaloop and that it melts DsrA-SL1 [62].
As for the recognition of DsrA by Hfq-CTR, DsrA-SL1 contains a capping AUUUC pentaloop (Figure 8, left). All these five nucleotides are conserved in Escherichia coli, Salmonella typhimurium, and Klebsiella pneumoniae, and are all complementary to the rpoS mRNA. Previous mutational analysis of DsrA-SL1 showed that DsrA sRNA lost its activity to activate rpoS translation with most of the mutations in this pentaloop [54]. NMR studies show that A11 and U12 are stacked in a helical environment, while U13, U14, and C15 are highly flexible [59]. All five nucleotides of DsrA-SL1 participate in the formation of an Aform helix in the structure of the DsrA:rpoS complex. These studies suggest that the highly dynamic character of DsrA-SL1 plays an important role in base-pairing with rpoS mRNA and in regulating its translation [59]. To investigate the role of DsrA-SL1 in regulating rpoS translation, we thus designed a mutated form of DsrA core called DsrA core mut in which AUUUC pentaloop is replaced by GGAA tetraloop in DsrA-SL1. The GGAA tetraloop belongs to the GNRA tetraloop family (N = any nucleotide, R = A or G), which is extremely widespread, comprising one-third of the tetraloops in ribosomal RNA [74]. The GNRA tetraloop is particularly stable in comparison with other RNA loops [75]. The GGAA tetraloop forms an asymmetric loop structure, where only the first G is stacked on the 5 side of the stem and the last three nucleotides stack on the 3 side of the stem [76]. It also contains several hydrogen bonds. The first G and fourth A form a sheared base pair with two hydrogen bonds. Additional hydrogen bonding occurs between the 2 -OH of the first G and the base R7 of the third A [77]. The extensive base stacking and the intramolecular interactions provide GGAA tetraloops with a high thermodynamic stability. We also confirm this effect with our T m measurement (Table 1).

Discussion
Using SRCD analysis, we show herein that, unexpectedly, the amyloid region of Hfq-CTR is able to promote DrsA:rpoS annealing with an efficiency similar to that of the fulllength protein at 15 °C. This work thus provides a new useful tool to analyze label-free RNA annealing by an amyloid protein, an analysis that is not trivial using classical EMSA (Supplementary Materials Figure S2). This spectroscopic analysis provides an alternative to fluorescence measurements without RNA labelling, that in addition gives information on base-pairing, base-stacking, and type of helix formed [72,73]. In turn, SRCD could also allow observation of the formation of the amyloid structures [45].
We show that the effect of Hfq-CTR is not due to the melting of the SL1 loop of DsrA ( Figure 5) nor to the melting of the rpoS stem ( Figure 6), which contrasts with the mechanism observed for the full Hfq including the Sm core [62]. Indeed, ribonuclease footprinting shows full Hfq binds the AUUUC DsrA pentaloop and that it melts DsrA-SL1 [62].
As for the recognition of DsrA by Hfq-CTR, DsrA-SL1 contains a capping AUUUC pentaloop (Figure 8, left). All these five nucleotides are conserved in Escherichia coli, Salmonella typhimurium, and Klebsiella pneumoniae, and are all complementary to the rpoS mRNA. Previous mutational analysis of DsrA-SL1 showed that DsrA sRNA lost its activity to activate rpoS translation with most of the mutations in this pentaloop [54]. NMR studies show that A11 and U12 are stacked in a helical environment, while U13, U14, and C15 are highly flexible [59]. All five nucleotides of DsrA-SL1 participate in the formation of an A-form helix in the structure of the DsrA:rpoS complex. These studies suggest that the highly dynamic character of DsrA-SL1 plays an important role in base-pairing with rpoS mRNA and in regulating its translation [59]. To investigate the role of DsrA-SL1 in regulating rpoS translation, we thus designed a mutated form of DsrAcore called DsrAcore mut in which AUUUC pentaloop is replaced by GGAA tetraloop in DsrA-SL1. The GGAA tetraloop belongs to the GNRA tetraloop family (N = any nucleotide, R = A or G), which is extremely widespread, comprising one-third of the tetraloops in ribosomal RNA [74]. The GNRA tetraloop is particularly stable in comparison with other RNA loops [75]. The GGAA tetraloop forms an asymmetric loop structure, where only the first G is stacked on the 5′ side of the stem and the last three nucleotides stack on the 3′ side of the stem [76]. It also contains several hydrogen bonds. The first G and fourth A form a sheared base pair with two hydrogen bonds. Additional hydrogen bonding occurs between the 2′-OH of the first G and the base R7 of the third A [77]. The extensive base stacking and the intramolecular interactions provide GGAA tetraloops with a high thermodynamic stability. We also confirm this effect with our Tm measurement (Table 1). Here we show that Hfq-CTR binds to WT DsrA SL1, and also to the mutated form of the SL1 with a tetraloop, but less solidly than WT (Supplementary Materials Figures S4,  3, and 6). We also show that the presence of the tetraloop in DsrA SL1 does not allow Here we show that Hfq-CTR binds to WT DsrA SL1, and also to the mutated form of the SL1 with a tetraloop, but less solidly than WT (Supplementary Materials Figure S4, Figures 3 and 6). We also show that the presence of the tetraloop in DsrA SL1 does not allow efficient annealing to rpoS (probably due to less base pairing with rpoS reg ) and more important stabilization of DsrA (Figures 2 and 6). Thus, we conclude that the effect of the CTR is directly linked to the efficiency of sRNA:mRNA annealing and not to complex stabilization nor SL1 melting. As Hfq-CTR binds the tetraloop mutant less than WT pentaloop, and cannot stabilize the mutant structure, it may indicate that CTR binds the AUUUC pentaloop and then promotes sRNA:mRNA annealing.
Finally, EMSA shows that Hfq-CTR can bind free DsrA core , free rpoS reg , and DsrA core :rpoS reg and suggests that Hfq-CTR prefers to bind to the DsrA core :rpoS reg complex (Supplementary Materials Figure S2). However, it has been shown that Hfq cannot bind the DsrA:rpoS complex very well after annealing [68]. This result should be investigated further as it could indicate that the NTR/Sm domain of Hfq may transfer the DsrA:rpoS complex to CTR after annealing [40]. Unfortunately, it was not possible to analyze Hfq-NTR activity using SRCD due to the poor solubility of this form of the protein.

Conclusions
The main and unexpected result of this study is that an amyloid-like region promotes RNA annealing. This result opens a perspective for a new physiological role of amyloids, including those associated with neurodegenerative diseases. In addition, we show here that a biophysical method, SRCD, can be used to follow RNA annealing by an amyloid protein, which is not trivial using EMSA. The role of amyloids on deoxyribonucleic (DNA) and ribonucleic (RNA) acid conformational changes has been observed previously [78]. In particular, amyloids may bind non-coding RNA [79]. They also may sequester RNA [80]. As stress-induced RNA activates amyloidogenesis in vivo [81], the effect of stress induced sRNA such as DsrA on amyloidogenesis may also occur, such as with DNA [45], and should be investigated further.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/biology10090900/s1, Figure S1: Observation of the shift of the SRCD peak maximum (λ max ) from~180 to 190 nm for DsrA core and DsrA core mut , Figure S2: EMSA analysis to confirm DsrA core rpoS reg annealing reaction by Hfq-CTR, Figure S3: Observation of DsrA core and rpoS reg annealing by Hfq full-length using SRCD, Figure S4: Analysis of the complex between DsrA core mut and CTR. Informed Consent Statement: Not applicable.

Data Availability Statement:
The SRCD data that support the findings of this study are available on request from the corresponding authors.

Abbreviations
CTR/NTR C/N-terminal region IDP Intrinsically Disordered Proteins rbs ribosome binding site sRNA small noncoding RNAs SL stem-loop SRCD Synchrotron Radiation Circular Dichroism.