6S in E. coli is physiologically produced by the ssrS gene (Hsu et al., 1985; Chae et al., 2011), controlled by two promoters, the proximal σ⁷⁰-dependent promoter and the distal σ⁷⁰/σ^S-dependent promoter (Lee et al., 2013). Consequently, two transcripts can be produced, a long precursor of 404 nt and a short precursor molecule of 194 nt. The 5′ end of both precursor transcripts is processed by RNases to produce the 183 nt mature 6S form, hereafter referred to as 6S (Kim and Lee, 2004; Fadouloglou et al., 2015). We cloned the 6S into the pet16b vector and transformed BL21 (DE3) cells, both routinely used for protein overexpression using a T7 polymerase-based system. Three different designs of the vector constructs were examined. The first construct, termed naked6S, contains the mature 6S sequence cloned between two restriction sites. In this design, the transcription initiation is controlled by the T7 promoter and the termination by elements in the vector under the assumption that the RNA is processed by internal RNases to produce the canonical mature 6S form (Figure 2A). The second construct, termed prom6S, contains the mature 6S sequence along with flanking 5′ and 3′ sequences containing both the proximal ssrS promoter and most of the precursor 6S sequence on the 5′ end and the rho terminator on the 3′ end. In this manner, the transcription can be controlled by the T7 promoter found upstream of the gene in the vector as well as the proximal ssrS promoter while the features that the endogenous E. coli RNases recognize to process this chimeric RNA transcript to the mature 6S are retained (Figure 2B). For the last construct, termed synth6S, we followed a drastically different approach independent of E. coli-specific processes that can be easily applied to other RNAs. In this construct the mature 6S sequence is flanked by the self-cleaving hammerhead (Prody et al., 1986) and HDV (Kuo et al., 1988) ribozyme sequences at the 5′ and 3′ends respectively. The transcription is controlled by the T7 promoter and the “maturation” of the transcript results from processing by the two ribozymes (Figure 2C).

The naked6S construct did not produce appreciable 6S expression (not shown), possibly reflecting a limited ability of the construct to properly process the transcript to the canonical 6S sequence. On the other hand, both prom6S and synth6S constructs produced high levels of 6S expression higher even than most other RNAs in E. coli (Figure 2D; Lanes prom6S and synth6S, respectively), exceeding by far endogenous 6S expression (Figure 2D; Lane M). The synth6S construct showed one more prominent band of a larger size, probably a result of incomplete processing of the self-cleaving ribozymes. For the subsequent experiments, we decided to proceed with prom6S, although the results of the synth6S are very promising and applicable to different types of RNAs.

Bacterial cultures containing the prom6S construct were induced by different IPTG concentrations (0, 0.05, 0.1 and 0.5 mM) at varying durations (1, 2, 4 h and overnight) (Supplementary Figure S1) at 37°C to assess 6S expression. Unexpectedly, the highest 6S expression yield was observed in the overnight cultures with no IPTG addition. The relative intensity of larger RNA bands (e.g., rRNA) was also reduced under these conditions compared to endogenous 6S expression (Figure 2D). Overnight incubation is likely beneficial to 6S expression because more cells enter the static phase increasing its stability. Following these observations, we decided to carry out the 6S RNA expression without adding IPTG at 37°C overnight where the optimum level of 6S RNA was achieved. The 6S RNA purification protocol consisted of two chromatographic steps, i.e., anion exchange (Q-sepharose) and size exclusion chromatography (SEC, Sephacryl S-200) (Nelissen et al., 2012; Edelmann et al., 2014) (Supplementary Figure S2) after heating and refolding of the originally produced RNA material. By this protocol the yield of pure 6S was up to ∼10 mg for a 1 L overnight Terrific Broth culture, orders of magnitude higher than a typical reaction of in vitro transcription. In addition, the quality of the product was sufficiently good to allow for the acquisition of excellent quality SAXS data for the structural analysis of 6S. We also examined a longer protocol including hydroxyapatite and arginine-sepharose affinity chromatography. However, the extra steps resulted in a substantially prolonged protocol (and lower yields) without significantly improving the purity of the final RNA.

SAXS data for free 6S and 6S:pRNA were obtained at medium salinity KCl buffers (200 mM), shown in Figure 3A. To prepare the 6S:pRNA complex a chemically synthesized 20 nt RNA (Steuten and Wagner, 2012) was mixed in molar excess (1:3 and 1:7) to the 6S RNA. The samples were also run through an inline size exclusion chromatography system (SEC-SAXS) at the beamline to better assess concentration effects in the SAXS patterns and remove the excess pRNA from the complex (Figure 3B). As expected for strongly negatively charged molecules, there was a decrease of the scattering intensity at higher concentrations due to strong repulsive interactions between the molecules (Figure 3A; inset). The calculated molecular masses were ∼63 kDa and ∼74 kDa, very close to the expected 59 kDa and 66 kDa of the free 6S and the complex, respectively. Conversely, the R _g were 60 Å and 49 Å for the free 6S and the complex, respectively, indicating a strong compaction of the molecule upon pRNA binding. A shift of the SEC peak to larger elution volumes for the 6S:pRNA complex compared to free 6S indicates a decrease in the hydrodynamic radius of the complex while the homogeneity of the samples is evidenced by the stability of their R _g throughout their respective peaks (Figure 3B). The distance distribution functions (Figure 3C) further illustrate the compaction occurring upon binding of pRNA as a much smaller D _max is observed for the complex (240 Å vs. 200 Å). Interestingly, the dimensionless Kratky plot (Durand et al., 2010) of the free 6S corresponds to a typical elongated, very anisometric, “rod-like” particle whereas the dimensionless Kratky plot of the complex corresponds to a less anisometric “disk-like” particle with decreased rigidity, indicating the pRNA synthesis may increase the flexibility of 6S (Figure 3D), possibly as a result of the engorgement of the central domain bulge.

A better understanding of the 6S RNA can be achieved by the analysis of its three-dimensional structure. SAXS, despite being a low resolution method, can discern differences between structures, especially in terms of the overall shape and size of the molecule. Fortunately, RNA has fewer types of building blocks compared to proteins and reasonable three-dimensional models can be built on the basis of secondary structure assumptions. For this purpose, we build 3D models of several secondary structure arrangements of free 6S and the 6S:pRNA complex with the help of RNAComposer (Popenda et al., 2012). The secondary structure arrangements we examined are shown in Figure 4A and declared as folds I to V. The first free 6S arrangement (fold I) contains no stem loops in the central domain bulge (Wassarman, 2007; Chen et al., 2017). The second arrangement (fold II) contains a commonly reported (Barrick et al., 2005; Steuten and Wagner, 2012) 3′ small stem loop in the central domain bulge at ∼140 nt (colored in yellow) and the third one (fold III) a much larger 3′ stem loop, involving partial unfolding of the closing stem, commonly associated with pRNA synthesis (Chen et al., 2017). For the 6S:pRNA complex, two alternative arrangements were considered differing only in the absence (fold IV) or presence of a 5′ stem loop (fold V) at ∼60 nt (Panchapakesan and Unrau, 2012; Steuten and Wagner, 2012). The differences between the arrangements may not appear very significant at secondary structure level but they can be very dramatic at tertiary structure level because the small hairpin loops can introduce tensions that completely change the shape and dimensions of the molecule, as shown by the 3D RNAComposer-derived models (Figure 4B).

Since RNA is expected to exhibit some structural plasticity, we decided to account for it and expand the conformational space available to the SAXS analysis by running short MD simulations on the RNAComposer-derived models. We used snapshots of the simulation as models, under the constraint of retaining the original secondary structure. These models were used as a pool of structures (each secondary structure arrangement was represented by the same number of structures) from which the EOM program (Bernadó et al., 2007) can select a subset that is consistent with the SAXS data as an ensemble. Interestingly, very good agreement of the ensembles to the experimental SAXS data (Figures 5A, B) was achieved (χ = 3.4 and 1.6 for the free 6S and 6S:pRNA complex, respectively), significantly improved over the best-fitting single models (Figure 6). Representative ensembles are shown in Supplementary Figures S3, S4. Comparison of the R _g histograms of the selected ensembles vs. the original pool of structures (Figure 5C) illustrates that only the most elongated models (i.e., fold I) are compatible with the SAXS data in the case of free 6S, in agreement with our previous work (Fadouloglou et al., 2015) and the 3.8 Å cryo-EM model by Chen et al. (Chen et al., 2017) which clearly shows that 6S adopts fold I-like conformation when in complex with Eσ⁷⁰ (without pRNA). Strikingly, no such bias is observed for the 6S:pRNA complex (Figure 5D). To better understand the results, we can look at the composition of the selected ensembles (Table 1; Supplementary Figures S3, S4) with respect to the examined structures (folds I to V) of Figure 4. Free RNA is predominantly found in a conformation with no stem loops in the central domain bulge, superficially resembling a “fully double helical” structure with some kinks (fold I). The hybridization of even a few nucleotides, while looking inconspicuous at the secondary structure level, requires the formation of a small double helix which, in turn, causes a significant compaction of the molecule (fold II) incompatible with our experimental data. This effect is even more pronounced in the presence of a larger loop (fold III). On the contrary, the best fit to the SAXS data of the complex is achieved when one considers a more equimolar ratio of the two conformations presented in Figure 4B (fold IV and fold V). Both 6S:pRNA arrangements have a very similar overall shape, despite their differences at the secondary structure level, with extended “double-helical” closing and internal stems and a “swollen” central domain due to the presence of the pRNA. MM-GBSA free energy analysis of the 3D conformations corroborate the above observations (Table 2). While the secondary structure mFold analysis (Table 2) shows no strong preference for the free 6S conformation (with fold II having the lowest mFold ΔG), the 3D structure free energy of Fold I is significantly lower than II and III, suggesting that it is the most stable free 6S form, in full agreement with the EOM observation. In the case of the 6S:pRNA complex, the free energy difference of the 3D structures of the two conformations is smaller (while the mFold difference is larger than in the case of free 6S), with fold IV being preferable to fold V, also similar to the EOM analysis. Binding free energy analysis of pRNA to 6S (Table 2) shows that pRNA binding is favorable for both fold IV and V.

The free, flexible, fold I-like 6S can accommodate the structural changes required for the change from the solution form to the Eσ⁷⁰–bound open promoter-like form (Barrick et al., 2005; Chen et al., 2017) while this is hindered by the “swollen” central domain of the 6S:pRNA complex and the reduced structural plasticity, imposed by the presence of the pRNA. In Figure 7, we superimpose on the 6S:Eσ⁷⁰ experimental model (Chen et al., 2017) a fold IV model of the 6S:pRNA complex in order to compare the two predominant states of 6S (Table 1, fold I and fold IV) for their fit into the Eσ⁷⁰. The presence of the 3′ stem loop at ∼140 nt (Figure 7, colored in yellow) very likely makes the 6S shape incompatible with binding to Eσ⁷⁰ and additionally, withdraws a significant number of nucleotides from interactions with the protein by engaging them to nucleotide pairing interactions. In the 6S:Eσ⁷⁰ experimental model (Chen et al., 2017) the nucleotides U134-G143 are single stranded and in direct interaction with the protein. These same nucleotides are the ones that form the 3′stem loop. Thus, the 6S:pRNA complex can hardly fit into the Eσ⁷⁰ cavity (Figure 7) when 6S adopts fold IV (and impossible when adopts fold V) and even then it would first require the disengagement of σ⁷⁰ from RNA polymerase. This renders the re-integration of the complex back to the Eσ⁷⁰ holoenzyme very unlikely, rescuing it after pRNA transcription and allowing it to resume transcription of genes.

Three different constructs were designed for overexpression in E. coli and cloning in the pet16b vector (Figures 2A–C) between the XbaI and BamHI restriction sites. The first construct, naked6S, only contains the mature 6S sequence between the two restriction sites. The second construct, prom6S, contains the mature 6S sequence as well as 5′ and 3′ flanking sequences, including the proximal promoter of the 6S gene. The third construct, synth6S, was synthesized (integrated DNA Technologies) and the mature 6S sequence is incorporated between two self-cleaving ribozymes; 5′ hammerhead (Prody et al., 1986) and 3′ HDV (Kuo et al., 1988) (shown in italics below). All sequences were verified by sequencing. Restriction sites are shown in bold.

Mature 6SRNA sequence:

5′-AUU​UCU​CUG​AGA​UGU​UCG​CAA​GCG​GGC​CAG​UCC​CCU​GAG​CCG​AUA​UUU​CAU​ACC​ACA​AGA​AUG​UGG​CGC​UCC​GCG​GUU​GGU​GAG​CAU​GCU​CGG​UCC​GUC​CGA​GAA​GCC​UUA​AAA​CUG​CGA​CGA​CAC​AUU​CAC​CUU​GAA​CCA​AGG​GUU​CAA​GGG​UUA​CAG​CCU​GCG​GCG​GCA​UCU​CGG​AGA​UUC-3′

Naked6S primers:

P1: 5′-GTG​GGCTCT​AGAATT​TCT​CTG​AGA​TGT​TCG​CAA​GC-3′

P2: 5′-GTT​GATGGA​TCCGAA​TCT​CCG​AGA​TGC​CGC​C-3′

Prom6S primers:

P1: 5′-GTG​GGCTCT​AGA​CTAC​GCG​GCA​AGT​ATG​GAA​C-3′

P2: 5′-GTT​GATGGA​TCCGAG​AGG​AAT​ACA​GCG​ACC​GT-3′

Synthetic Gene with flanking HammerHead and HDV ribozymes:

5′-TCT​AGA GGG​AGA​GAG​AAA​TCT​GAT​GAG​TCC​GTG​AGG​ACG​AAA​CGG​TAC​CCG​GTA​CCG​TCATT​TCT​CTG​AGA​TGT​TCG​CAA​GCG​GGC​CAG​TCC​CCT​GAG​CCG​ATA​TTT​CAT​ACC​ACA​AGA​ATG​TGG​CGC​TCC​GCG​GTT​GGT​GAG​CAT​GCT​CGG​TCC​GTC​CGA​GAA​GCC​TTA​AAA​CTG​CGA​CGA​CAC​ATT​CAC​CTT​GAA​CCA​AGG​GTT​CAA​GGG​TTA​CAG​CCT​GCG​GCG​GCA​TCT​CGG​AGA​TTCGGG​TCG​GCA​TGG​CAT​CTC​CAC​CTC​CTC​GCG​GTC​CGA​CCT​GGG​CTA​CTT​CGG​TAG​GCT​AAG​GGA​GAA​G GGA​TCC-3′

The recombinant plasmid was transformed into E. coli BL21 (DE3) competent cells for RNA expression. A single colony was cultivated in 20 mL LB medium containing 50 μg/mL ampicillin (LB amp⁵⁰) at 37°C, 250 rpm and used to inoculate 2 L of Terrific Broth Medium (TB) containing 50 μg/mL ampicillin (TB amp⁵⁰). It was subsequently incubated overnight at 37°C, 250 rpm. The cells were pelleted and washed with 20 mM Tris-HCl pH 7, 200 mM NaCl. The RNA was isolated by acid-guanidinium thiocyanate-phenol-chloroform extraction in an analogous manner to that previously described by Chomczynski (Chomczynski and Sacchi, 2006). The RNA was analyzed by electrophoresis on an analytical denaturing 7.2% urea polyacrylamide gel (Urea-PAGE) in 1X TBE containing 8M urea (Figure 2D).

RNA pellets from 2 L of culture were diluted to 300 mL of a buffer containing 50 mM K₂HPO₄/KH₂PO₄ pH 7.0 and 0.1 mM EDTA, heated to 95°C for 5 min and subsequently placed on ice for 20 min to refold and centrifuged to precipitate protein remains. RNAs were purified using an Akta purifier system (Amersham), Q Sepharose anion exchange column and a Sephacryl S-200 column (GE Healthcare). The Q Sepharose column was pre-equilibrated with buffer A (20 mM Tris HCl pH 8.0, 400 mM NaCl, 100 mM KCl, 0.1 mM EDTA). After loading the sample onto the column 400 mL of buffer A was used to wash the column and subsequently a 800 mL gradient of buffer B (20 mM Tris HCl pH 8.0, 620 mM NaCl, 100 mM KCl, 0.1 mM EDTA) up to 100% (620 mM NaCl) was performed while collecting 8.5 mL fractions. The system was cleaned with 200 mL of 100% buffer B to wash off remaining uncleaved RNA and DNA remains. Absorbance was monitored continuously at 260 nm. The fractions were analyzed by 7.2% Urea-PAGE in 1X TBE containing 8 M urea. Fractions with the 6S RNA were merged, heated to 95°C for 5 min and subsequently placed on ice for 20 min to refold. After centrifugation, the sample was loaded onto a Sephacryl S-200 column (GE Healthcare), and 6S RNA eluted with a buffer containing 20 mM Tris HCl pH 8.0, 200 mM KCl and 5 mM MgCl₂. The purified 6S RNA was concentrated by ultrafiltration with 10 kDa cut-off filter and analyzed by 7.2% Urea-PAGE in 1X TBE containing 8 M urea. RNA quantification was performed as described before (Chomczynski and Sacchi, 2006).

The pRNA in our experiments is the 20 nt synthetic sequence 5′-AUC​GGC​UCA​GGG​GAC​UGG​CC-3′ (Steuten and Wagner, 2012). For the complex formation, purified 6S RNA solution (in 20 mM Tris HCl pH 8.0, 200 mM KCl and 5 mM MgCl₂ buffer) was mixed with molar excess of pRNA in ratios 1:3 and 1:7 to a final volume of 75 μL. The mixture was heated at 94°C for 4 min and subsequently placed at 37°C for 1.5 h.

Preliminary SAXS data of 6S were collected at the EMBL Hamburg P12 undulator beamline of the Petra III storage ring in DESY (Hamburg, Germany) using a Pilatus 2M (DECTRIS) photon counting pixel detector (Blanchet et al., 2015). The measurements were performed at 10°C using the automated sample changer. The sample-to-detector distance was 3.1 m, covering a range of momentum transfer 0.02 < s < 4.8 nm⁻¹ (s = 4π sinθ/λ, where 2θ is the scattering angle, and λ = 1.24 Å is the X-ray wavelength). Primary data reduction, radial averaging, averaging and subtraction were performed on-site with the beamline software (SASFLOW, v. 3.0, Hamburg, Germany). The majority of SAXS data of 6S as well as the 6S:pRNA complex were collected at the SWING Beamline of Synchrotron SOLEIL (Gif-sur-Yvette, France) with an Aviex charge-coupled device detector (David and Pérez, 2009). The measurements were performed at 15°C for several different concentrations of 6S (up to 8 mg/mL) using the automatic sample changer. Both free 6S and 6S:pRNA complex samples were also run through an Agilent HPLC system with a gel filtration column to assess the behavior of the samples at lower effective concentrations and for the 6S:pRNA complex to remove unbound pRNA. The sample-to-detector distance was 3.1 m, covering a range of momentum transfer 0.007 < q < 0.614 Å⁻¹ (q = 4π sinθ/λ, where 2θ is the scattering angle, and λ = 1.033 Å is the X-ray wavelength). Using the Foxtrot software, the data were averaged radially and converted to absolute units, analyzed for radiation damage, averaged and subtracted. Subsequent analysis was performed with the ATSAS program suite (Petoukhov et al., 2012). PRIMUS (Konarev et al., 2003) was used to merge data from different concentrations, and for the calculation of the radius of gyration R _g and the forward scattering intensity I(0) (proportional to the number of electrons of the particle) from the slope of Guinier plot (lnI(q) versus q ²) (Guinier, 1939). GNOM (Svergun, 1993) was used to calculate the pair distance distribution function p(r) and to estimate the maximum particle dimension (D _max). The molecular mass (MM) of the solute was estimated from the SAXS data from the I(0). Twenty dummy beads models of the SAXS data were created for both free 6S and 6S:pRNA complex with DAMMIF (Franke and Svergun, 2009) and averaged with DAMAVER (Volkov and Svergun, 2003).

A few alternative secondary structure arrangements (illustrations produced with the VARNA program) (Darty et al., 2009) of free 6S and 6S:pRNA complex were produced and the corresponding free energies (ΔG) were calculated with mFold (Zuker, 2003). These structures were used as input for the online 3D structure prediction program RNAComposer (Popenda et al., 2012). To better assess the conformational space explored by the molecules, short (100 ns) all-atom Molecular Dynamics simulations were performed with the AMBER14 program suite (Case et al., 2005) with the parameters of the ff14SB force field (Maier et al., 2015). All simulations were performed in 2 fs steps at 300K and constant pressure of 1 atm in TIP3P water boxes with periodic boundary conditions after an initial minimization and stepwise heating in constant volume. Snapshots of the simulations were taken at regular intervals and the SAXS patterns were calculated with CRYSOL (Svergun et al., 1995). The Ensemble Optimization Method (EOM) program (Bernadó et al., 2007) was used to select the subset of models with the best fit to the experimental SAXS data. R _g values of the models were also calculated and used to create histograms to compare, in a general manner, the population of the pool of original models (i.e., all the models) vs. the subset of the selected ensembles. MM-GBSA free energies (G) and binding free energies of pRNA to the 6S:pRNA complex (ΔG) were calculated with the MMPBSA module (Miller et al., 2012) of AMBER14 from 20 snapshots taken from the last 20ns of each simulation.