The structure of fadL mRNA and its interactions with RybB sRNA

Small bacterial RNAs (sRNAs) regulate translation by pairing with complementary sequences in their target mRNAs, in a process which is often dependent on the Hfq protein. Here, the secondary structure of a 95-nt long fragment of Salmonella fadL mRNA containing RybB sRNA binding site in the coding region was analyzed. The data indicated local rearrangements in this mRNA structure after the annealing of RybB. The filter retention data had shown that Hfq bound both RybB and the fadL mRNA fragment with tight affinities. Moreover, Hfq increased the rate of RybB annealing to fadL mRNA. These data indicate that Hfq directly participates in RybB interactions with the fadL mRNA.


INTRODUCTION
Bacterial trans-encoded small noncoding RNAs (sRNAs) participate in the cell's adaptation to changing environmental conditions and in the maintance of cellular homeostasis (Waters & Storz, 2009).They are also involved in the regulation of bacterial pathogenicity (Papenfort & Vogel, 2010;Tree et al., 2014).Small RNAs control gene expression by pairing with complementary sequences in the regulated mRNAs, which affects mRNA translation and decay (Wagner & Romby, 2015).In particular, small RNAs regulate the translation of the outer membrane proteins in Escherichia coli and Salmonella Typhimurium, which include the OmpC (Chen et al., 2004), OmpD (Pfeiffer et al., 2009), and OmpN porins (Bouvier et al., 2008).The sRNA-dependent regulation of the outer membrane composition affects such important processes as the uptake of nutrients (Figueroa-Bossi et al., 2009), response to membrane stress (Guo et al., 2014), biofilm formation (Jorgensen et al., 2013), and the resistance to antibiotics (Parker & Gottesman, 2016).
Translation regulation by many of sRNAs is dependent on the ring-shaped Hfq protein (Updegrove et al., 2016).Hfq contains three RNA binding sites on its surface, which allows it to use different binding modes in interactions with the sRNA and mRNA molecules (Zhang et al., 2013;Schu et al., 2015).The positively charged residues on the rim of the Hfq ring are necessary for efficient annealing of the sRNA and mRNA molecules (Panja et al., 2013;Zheng et al., 2016), while the negatively charged residues (Panja et al., 2015), and the C-terminal protein extensions (Santiago-Frangos et al., 2016) contribute to the specificity of the interactions.In the best studied interaction of E. coli DsrA sRNA with the rpoS mRNA, the role of Hfq is to rearrange the mRNA structure to facilitate the sRNA annealing (Soper & Woodson, 2008;Soper et al., 2011;Peng et al., 2014a;Peng et al., 2014b).A recent study of three sRNAs binding to Salmonella ompD mRNA had shown that the role of Hfq in rearranging sRNA and mRNA structures depends on the structural properties of the interacting RNAs (Wroblewska & Olejniczak, 2016b).
RybB is an Hfq-dependent sRNA that represses the synthesis of several outer membrane proteins in response to the envelope stress (Balbontin et al., 2010;Papenfort et al., 2010).The 5'-end of this 79-nt long sRNA functions as an autonomic regulatory domain, which is essential for sRNA binding to different mRNA targets.Among targets of RybB is the fadL mRNA, which encodes an outer membrane porin involved in the transport of long-chain fatty acids (Nunn & Simons, 1978;Black et al., 1987).RybB sRNA represses its translation by binding downstream of the AUG start codon, at positions +43 to +50 of the coding sequence (Papenfort et al., 2010).
To better elucidate the interactions between RybB sRNA and fadL mRNA, changes in the structure of a 95-nt long fadL mRNA fragment upon RybB binding were monitored.Moreover, the Hfq protein binding to both RNAs and its influence on the kinetics of RybB annealing to the coding sequence of fadL were studied.

MATERIALS AND METHODS
RNA preparation.DNA templates for in vitro transcription were obtained by Taq polymerase extension of chemically synthesized, overlapping oligonucleotides (oligo.pl, Warsaw, Poland).Salmonella RybB sRNA and fadL95 mRNA fragment were in vitro transcribed using T7 RNA polymerase and purified using denaturing PAGE as described (Milligan et al., 1987).RNA molecules were 5'-32 P-labeled using T4 polynucleotide kinase (Thermo Scientific) and purified on denaturing PAGE.After elution, the samples were ethanol precipitated, dried, and dissolved in water to a final 200 nM RNA concentration.
Hfq protein purification.The Salmonella C-terminally His 6 -tagged Hfq protein was overexpressed and purified as previously described for E. coli Hfq (Malecka et al., 2015).In short, after overexpression from pET15b plasmid, the cells were lysed by sonication, and the protein was purified by Ni 2+ affinity chromatography (HisTrap crude column, GE Healthcare), followed by treatment with DNase I and RNase A to remove any bound nu-cleic acids.Next, the size exclusion purification was performed on a HiLoad 16/60 Superdex 200 size exclusion column (GE Healthcare) equilibrated with storage buffer (50 mM HEPES 7.5, 250 mM NH 4 Cl, 1 mM EDTA, and 10% glycerol).The Hfq concentration was determined using absorbance at 280 nm (Olejniczak, 2011).
In vitro structure probing and footprinting.Just before setting up the reactions, the 32 P-fadL95 RNA was denatured at 90°C for 1 min and then incubated at room temperature for 10 min.The structure probing experiments were performed essentially as previously described (Wroblewska & Olejniczak, 2016b).The reactions with RNase T1 were carried out in 12 mM Tris-HCl, pH 7.2, 48 mM NaCl, and 1.2 mM MgCl 2 at RT for 10 min.The RNase T2 reactions were performed in 10 mM Tris pH 7, 100 mM KCl, and 10 mM MgCl 2 , in the presence of 1 μg of yeast tRNA, and incubated at RT for 15 min.Structure probing reactions with nuclease S1 were performed in 40 mM sodium acetate pH 4.5, 300 mM NaCl, and 2 mM ZnSO 4 , and incubated at RT for 10 min.Reactions with RNase T1, T2 and S1 were quenched by addition of 10 μL of stop buffer (8 M urea and 20 mM EDTA).To obtain alkaline hydrolysis ladder, 32 P-fadL95 was incubated in formamide at 100°C for 1 h, followed by cooling on ice.RNase T1 ladder was obtained by incubation of 32 P-fadL95 in 50 mM sodium citrate (pH 4.3) and 7 M urea at 55°C for 10 min.To monitor changes in RNase T1 accessibility upon RybB binding, 32 P-fadL95 was mixed with increasing concentration of RybB in T1 reaction buffer.After 20 min of incubation, 1 μL of RNase T1 was added to each tube.
Reactions were incubated at RT for 10 min, and then the reactions were stopped and processed as described above.Samples were separated on 10% polyacrylamide gels.The gels were frozen at -20°C, exposed to phosphor screens overnight and analyzed using a phosphorimager.
The binding reactions were initiated by mixing 15 μL of 5'-32 P-labeled RNA at 0.02 nM concentration with 15 μL of Hfq dilution followed by incubation for 1 h at RT. 25 μL aliquots of each reaction were withdrawn, filtered and washed with 100 μL of binding buffer.Filters were dried, exposed to phosphor screens, and the data were quantified using a phosphorimager and MultiGauge software.
The binding data were fit to the Michaelis-Menten binding isotherm.
To monitor the equlibrium binding of 32 P-labeled fadL95 to RybB sRNA, an electrophoretic mobility shift assay was used.Prior to use, RNAs were treated as described above.The equilibrium binding reactions were prepared by mixing 15 μL of 32 P-labeled fadL95 at 1 nM final concentration with 15 μL of RybB solution followed by incubation for 1 h at RT. 5 µl of each sample was loaded on a 6% native polyacrylamide gel in 1× TBE buffer.The gels were run at 5 W at 4°C in a coldroom, dried, and analyzed using a phosphorimager.The fraction bound data were fit to the Michaelis-Menten binding isotherm.
Annealing assays.The kinetics of annealing of RybB sRNA to fadL95 mRNA fragment in the presence or absence of Hfq protein was monitored by gel mobility shift  (Wroblewska & Olejniczak, 2016b) with the following modifications.The annealing reactions were prepared in 80 µL total volume by mixing 1 nM 32 P-fadL95 with 50 nM RybB sRNA in the presence or absence of 3 nM Hfq (hexamer concentration).The reactions were incubated at 25°C in a thermomixer.5 µL aliquots were withdrawn at indicated time points and separated on 6% native polyacrylamide gels in 1× TBE buffer at 4°C in a coldroom.Control reactions were prepared in the same way, except that to the reactions with 32 P-fadL95 and Hfq, the unlabeled fadL95 RNA at 2 nM concentration was added to prevent the precipitation of the sample in the wells.The gels were run continuously during experiments.After quantifying the data using a phosphorimager, the fraction bound values of RybB-fadL95 or RybB-fadL95-Hfq complexes were plotted versus time (0.5-60 minutes).Observed association rates (k obs ) were determined by fitting the data to the single exponential equation.

RESULTS
RybB sRNA binds fadL mRNA at the sequence which is localized + 43 to +50 nt from the translation start site (Papenfort et al., 2010).To analyze RybB interactions with fadL, a 95-nt long fragment of this mRNA was used (fadL95) (Fig. 1D).It included a 35-nt long sequence of 5'-untranslated region (5'-UTR) of this mRNA and a 60-nt long region of its coding sequence, including the RybB binding site.This fragment contained a 16-nt long AU-rich sequence in the 5'-untranslated region, which could be described as (ANN) 5 repeated sequence.Similar sequences were shown as functional Hfq binding sites in other mRNA molecules (Soper & Woodson, 2008;Schu et al., 2015;Wroblewska & Olejniczak, 2016b).It also included a shorter 8-nt long AU-rich sequence in the coding region (Fig. 1D).As a control, a longer, 216-nt fragment including the whole 100-nt 5'-UTR of fadL mRNA was also tested; however, its multiple complexes with Hfq could not be well separated on the gel (data not shown).Hence, the fadL95 mRNA fragment was used in this study, which contained both the region of RybB binding and a lengthy AU-rich sequence in the 5'-UTR.
To determine the secondary structure of fadL95 mRNA the probing with RNase T2 and nuclease S1 was applied to detect single-stranded regions, while the comparison of cleavages generated by RNase T1 under the denaturing and native conditions was used to define double-stranded regions.The structural information provided by enzymatic probing was used to predict the secondary structure of fadL95 using the RNAstructure software (Reuter & Mathews, 2010).The most thermodynamically stable of the two structures predicted by the software based on the experimental constraints is presented in Fig. 1D.
The secondary structure of the region of fadL mRNA surrounding the translation start site contained three stem-loop structures, named SL1 to SL3 (Fig. 1A, D).The double-stranded character of SL1 and SL3 was supported by decreased RNase T1-induced cleavage under native conditions of the SL1 G26, G27, G32, G33, and G40 residues, and the SL3 G66 and G78 residues.The fact that guanosine residues at positions 83 and 84 were quite accessible to RNase T1 under native conditions, as well as the A86 -U88 residues to RNases S1 and T2, suggested that they form an elongated single-stranded region.Residue A45, susceptible to RNase T2, is located within the 3' part of the internal asymmetric loop of SL1.Additionally, cleavages induced by RNase T2 and nuclease S1 at positions U55-A57 defined the apical loop of SL2, while U68 and C73 the apical loop of SL3 (Fig. 1A, D).
Both AU-rich sequences in fadL95 mRNA are located in conformationally dynamic regions, while the RybB binding site partially overlaps with a stem-loop structure (Fig. 1A, D).The long AU-rich sequence in the 5'-UTR is composed of unpaired nucleotides and it could be accessible for binding by the Hfq protein (Fig. 1A, D).The short AU-rich sequence in the coding region is also single-stranded and forms a loop of SL2.The GAGG Shine-Dalgarno sequence is partly involved in the formation of the SL1, which could limit its availability to the ribosome, while the AUG start codon is located in the apical loop of SL1.The predicted binding site of RybB sRNA is located within the stem of SL3 structure and in the following single-stranded region (Fig. 1A, D).
Annealing of RybB sRNA to fadL95 induced changes in the structure probing pattern in the area of its binding site, which is consistent with the unfolding of SL3 (Fig. 1B, C).The G81 residue in the SL3, as well as G83 and G84 in the following single-stranded region, which are part of the sequence complementary to the 5'-terminus of RybB sRNA, were increasingly protected from RNase T1 induced cleavage with the increase of RybB concentration (Fig. 1B, C).As a control, the cleavage intensities of a distantly located G44 residue were also compared, and have shown only a small change upon RybB annealing.On the other hand, the G72 residue, which is located in the apical loop of SL3, has displayed an increased susceptibility to RNase T1-induced degradation, when RybB concentration was increased (Fig 1B , C, D).These data are consistent with the unfolding of the SL3 stem-loop upon RybB binding and improved accessibility of residues located within the SL3 loop for RNases.Overall, these results confirm the predicted RybB sRNA binding site in the coding sequence of fadL mRNA (Papenfort et al., 2010), and suggest that the annealing of RybB induces local rearrangements in the structure of fadL.
To test if the Hfq protein could be involved in the interactions between RybB and fadL, binding affinities of Hfq to RybB and fadL95 RNA molecules were measured using a high-throughput filter retention assay (Fig. 2).The data indicates that each of these two RNAs bound Hfq very tightly, with sub-nanomolar affinities.RybB bound Hfq with an equilibrium dissociation constant (K d ) value of 0.073±0.016nM, while fadL95 bound Hfq with a K d value of 0.039±0.0086nM.The K d value of RybB binding to Hfq was the same, within the error range, as the value previously reported for this interaction in a buffer containing 2 mM Mg 2+ ions (Wroblewska & Olejniczak, 2016b).The K d value of Hfq binding to fadL95 was similar as previously reported for Hfq binding to RNA-IN mRNA (Ross et al., 2013), and ompD mRNA (Wroblewska & Olejniczak, 2016b), and somewhat tighter than that observed for rpoS mRNA (Peng et al., 2014b), and glmS mRNA (Salim et al., 2012).These data suggest specific interactions of Hfq with both RNAs.
To test if Hfq affects the kinetics of RybB annealing to fadL95 mRNA fragment, a native gel mobility shift assay was used (Fig. 3).At first, the affinity of RybB to fadL95 was measured, which provided the K d value of 90±13 nM (Fig. 3A, B, C).This value is similar to that previously reported for DsrA sRNA binding to rpoS mRNA (Soper & Woodson, 2008;Soper et al., 2011).Next, the rates of RybB annealing to fadL95 (k obs ) were measured at 1 nM concentration of 32 P-fadL95 and 50 nM unlabeled RybB in the presence or absence of 3 nM Hfq (Fig. 3D, E, F).The unbound 32 P-fadL95, its binary complexes with Hfq or RybB, and the ternary fadL95-RybB-Hfq complex migrated in the gel with distinct rates, which allowed quantifying the individual complexes.In the presence of 3 nM Hfq, mostly the ternary 32 P-fadL95-RybB-Hfq complex was formed, with a rate that was more than 300-times faster than the rate of the binary 32 P-fadL95-RybB complex formation in the absence of Hfq (Fig. 3D, E, F).The k obs value of the ternary 32 P-fadL95-RybB-Hfq complex formation was 3.1±0.082,while that of the binary 32 P-fadL95-RybB complex formation in the absence of Hfq was 0.0078±0.0035.This range of Hfq influence on the rate of sRNA annealing is similar to that previously observed for other sRNAs (Peng et al., 2014b;Wroblewska & Olejniczak, 2016b).

DISCUSSION
The binding site of RybB sRNA in the coding sequence of fadL mRNA was predicted based on sequence analysis and confirmed using compensatory mutations in a reporter assay in vivo (Papenfort et al., 2010).Regulation of the fadL mRNA translation by RybB is among several examples of mRNAs which are regulated by sR-NAs binding in the mRNA coding sequence, outside of the footprint of the initiating ribosome (Wroblewska & Olejniczak, 2016a).Here, the direct binding of RybB to the complementary sequence in the coding region of fadL was further supported by the changes induced in the structure probing pattern of the 95-nt long fragment of fadL mRNA upon RybB annealing (Fig. 1).The decreased intensities of RNase T1 cleavages occurred in the area of predicted base-pairing of the RybB 5'-end in the fadL coding sequence, confirming its binding site (Fig 1).Moreover, the increased T1-induced cleavage, which occurred at a guanosine residue located closely upstream of the RybB binding site, was consistent with the local structural changes upon RybB binding (Fig. 1B,  C, D).
The RNase T1 probing data presented here suggested that the annealing of RybB sRNA induced local structural rearrangements in the fadL mRNA coding sequence (Fig. 1B, C, D).However, as only one RNA-structure specific enzyme was used to monitor the conformational change, it is likely that the application of other structure probes, such as Pb 2+ ions, could lead to a more detailed picture of this structural transition (Ciesiolka et al., 1989).In further studies, it would be also important to test the role of the structural context of a longer fadL mRNA fragment on the conformational changes in the area of RybB annealing.Local structural rearrangements induced by sRNA annealing to the mRNA coding sequence have been previously observed for other sRNAs, including MicF binding to the lpxR mRNA (Corcoran et al., 2012), and MicC binding to the ompD mRNA (Wroblewska & Olejniczak, 2016b).Much larger conformational changes were observed for DsrA annealing to the 5'-UTR of rpoS mRNA, where the binding of sRNA opens up an inhibitory structure and enables the access of the ribosome to mRNA (Lease & Woodson, 2004;Soper & Woodson, 2008).
The Hfq protein has been shown to increase the rates of association of several sRNAs to the 5'-untranslated regions of their target mRNAs (Wagner & Romby, 2015).However, the molecular mechanism used by Hfq to contribute to the sRNA-mRNA interaction has been explained only for the annealing of DsrA sRNA to the rpoS mRNA (Lease & Woodson, 2004;Peng et al., 2014a;Peng et al., 2014b).Hfq rearranges the rpoS mRNA structure to enable the pairing of DsrA to a complementary sequence in the mRNA (Soper & Woodson, 2008;Soper et al., 2011).The data presented here had shown that Hfq increased the annealing rate of RybB sRNA to the coding sequence of fadL mRNA, and bound tightly to both of these RNAs (Fig. 2, 3).The role of Hfq in accelerating sRNA annealing to the mRNA coding sequence has been reported before for the lpxR mRNA regulation by MicF (Corcoran et al., 2012), and the ompD mRNA regulation by RybB, SdsR and MicC sRNAs (Wroblewska & Olejniczak, 2016b).It was reported that Hfq strongly increased the annealing of both, RybB and MicC, to their complementary sites located in different structural contexts in the ompD mRNA coding sequence (Wroblewska & Olejniczak, 2016b).However, its role for RybB annealing was to overcome the energetic barriers of both RybB and ompD mRNA structures, while for MicC annealing it served only to overcome the barrier of MicC sRNA structure.The role of Hfq in sRNA annealing to the coding region depended on the (ANN) 8 sequence in the 5'-UTR of ompD mRNA (Wroblewska & Olejniczak, 2016b).It is possible that the single-stranded AU-rich, (ANN) 5 sequence in the 5'-untranslated region of fadL mRNA could be accessible for Hfq binding (Fig. 1D).However, not all Hfq binding sites are functional in the sRNA annealing.To explain the molecular mechanism used by Hfq to promote the RybB sRNA annealing to the fadL mRNA, further studies are needed which will involve mutants of the fadL mRNA, and Hfq variants with mutations in its RNA binding sites.

Figure 1 .
Figure 1.The annealing of RybB sRNA induces changes in the secondary structure of the fadL95 mRNA fragment.(A) In vitro structure probing of 32 P-fadL mRNA fragment was performed using RNase T1 and T2, and the S1 nuclease (indicated above the lanes).(B) The RNase T1 structure probing of 32 P-fadL mRNA fragment in the presence of increasing concentration of RybB.The guanosine residue more susceptible to RNase T1-induced cleavages is indicated by a red triangle, while the residues less susceptible to this cleavage are indicated by blue reverse triangles.RNase T1 was used under native (T1 N) and denaturing (T1 D) conditions, C is an untreated control, while OH indicates the formamide ladder.Numbers on the left side of the gel correspond to the G residue positions.(C) Changes in nucleotide susceptibility to cleavage upon RybB binding were plotted versus RybB concentration.(D) Secondary structure of the fadL95 mRNA based on the structure probing results presented in (A).Residues indicated in red and green were constrained as single-or double-stranded, respectively, to generate the structure model using the RNAstructure program.The region complementary to RybB sRNA is marked by lines, AUG start codon is indicated by red font.The position of increased cleavage with RNase T1 upon RybB binding is indicated by a red triangle, and the positions of decreased cleavage with blue reverse triangles.

Figure 2 .
Figure 2. The binding of the RybB sRNA and fadL95 mRNA to the Hfq protein.The bound fraction data provided by filter retention assay (raw data shown under the plot) were plotted versus the concentration of Hfq.The fits of data to the Michaelis-Menten binding isotherm provided the K d value of 0.092 nM for 32 P-RybB-Hfq, and 0.026 nM for 32 P-fadL95-Hfq.Average K d values from three independent experiments are provided in the text.

Figure 3 .
Figure 3. Hfq increases the rate of RybB annealing to the fadL95 mRNA fragment.(A) The secondary structure of RybB sRNA showing the complementary sequence of fadL mRNA.(B) The analysis of RybB sRNA binding to 32 P-fadL95 using the native mobility shift assay.(C) The plot of 32 P-fadL95 binding data from (B) versus the concentration of unlabeled RybB sRNA.The data were fit to the Michaelis-Menten equation and provided the K d value of 89 nM.An average K d value is provided in the text.(D) The kinetics of annealing of 1 nM 32 P-fadL95 mRNA fragment to RybB (50 nM concentration) in the absence (E) or presence of 3 nM Hfq hexamer.Free fadL95 is denoted as F, fadL95-Hfq complex as F-H, fadL95-RybB complex as F-R, fadL95-RybB-Hfq ternary complex as F-R-H.The control reaction, in which 32 P-fadL95 alone was bound to Hfq, additionally contained 2 nM unlabeled fadL95; (F) The data from D and E were plotted versus time.The fitting of data to the exponential equation provided the k obs values of 0.0055 min -1 and 4.0 min -1 , in the absence and presence of 3 nM Hfq, respectively.The average k obs values are provided in the text.