The protocol described in the article Khusainov et al. (2016b) was used for 70 S purification (Khusainov et al., 2016a). S. aureus cells (RN6390 strain) were grown at 37°C (180 rpm) in brain-heart infusion broth (BHI) and harvested in the early logarithmic phase (1 OD⁶⁰⁰ / ml). Then cells were washed in buffer A (20 mM Hepes-KOH pH 7.5, 100 mM NH₄Cl, 21 mM Mg(OAc)₂, 1 mM DTT), pelleted at 4,750 g and kept frozen at −80^oC. Typically, 5 g of cells were obtained from 2 L of culture. Lysis of the cells (5 g) was performed in buffer A in the presence of 1 mM EDTA, lysostaphin (Sigma-Aldrich), DNase I (Roche), and protease inhibitor cocktail (Roche) at 37^°C for 45 min followed by centrifugation at 30,000 g for 90 min.

Ribosomes were precipitated through two stages by adding to the supernatant PEG 20,000 (Hampton Research) with final concentrations of 2.8 and 4.2 % w/v. Solutions were centrifuged at 20,000 g for 5 and 10 min, respectively. The pellet was then resuspended in 35 ml of buffer A and layered to 25 ml cushion of buffer B (10 mM Hepes-KOH pH 7.5, 500 mM KCl, 25 mM Mg(OAc)₂, 1.1 M Sucrose, 0.5 mM EDTA, 1 mM DTT), followed by centrifugation at 158,420 g for 15 h using a Beckman Type 45 Ti rotor.

The ribosomal pellet from sucrose cushion was resuspended in buffer E (10 mM Hepes-KOH pH 7.5, 100 mM KCl, 10.5 mM Mg(OAc)₂, 0.5 mM EDTA, 1 mM DTT), and loaded on 7–30 % w/v sucrose gradients at a concentration up to 7 mg/ ml and centrifuged at 38,694 g for 15.5 h using a Beckman SW28 rotor. Magnesium was adjusted to 25 mM in the pooled fractions. Ribosomes were precipitated by adding PEG 20,000 to a final concentration 4.5 % w/v and centrifuged at 20,000 g for 12 min. The pellet was gently resuspended in buffer G (10 mM Hepes-KOH pH 7.5, 30 mM NH₄Cl, 10 mM Mg(OAc)₂, 1 mM DTT), flash-frozen in liquid nitrogen and stored at −80^oC in small aliquots.

The small ribosomal subunit was isolated from the intact 70 S ribosome under various conditions described below. First, the 70 S ribosome was dialysed in dissociation buffer (1 mM Mg(OAc)₂, 200 mM NaCl₂ or 200 mM KCl, 10 mM Hepes-KOH pH7.5, 1 mM DTT) for 4 h at 4^°C. Then, the subunits were separated with 0–30 % w/v sucrose gradients followed by centrifugation at 35,606 g for 17 h at 4^oC using Beckman SW28 rotor. Fractionation of the sucrose gradient is shown on Supplementary Figure S1A,B. Fractions from the sucrose gradient in 200 mM KCl were pooled and loaded again on 5–20 % w/v sucrose gradients (35,606 g for 17 h, Beckman SW28 rotor, Supplementary Figure S1C). The search for optimal salt conditions for 30 S purification was carried out in 100 and 200 mM KCl or NH₄Cl (1 mM Mg(OAc)₂, 10 mM Hepes-KOH pH7.5, 1 mM DTT) and 400 mM NaCl (6 mM or 10 mM Mg(OAc)₂, 10 mM Hepes-KOH pH7.5, 1 mM DTT). Dialysis for 4 h at 4^oC was followed by subunits separation step with 0–30 % w/v sucrose gradient and centrifuged at 44,556 g for 15 h using Beckman SW41 rotor. Selected fractions were pooled, concentrated with Amicon 0.5 ml MWCO 100K, and analyzed on a 15 % SDS-PAGE (Supplementary Figures S2A,B). Re-screening for optimal conditions was performed in 30 and 50 mM NH₄Cl (1 mM Mg(OAc)₂, 10 mM Hepes-KOH pH7.5, 1 mM DTT). The 70S ribosome was divided into subunits with 0–30 % w/v sucrose gradient followed by centrifugation at 46,932 g for 14 h 14 min using Beckman SW41 rotor. The obtained sucrose gradient profiles and respective SDS-PAGE profiles of pooled 30 S peak samples are shown in Figures 1A,B.

For structural analysis of the 30 S subunit, we first purified 70 S particles as described above, with tiny modification. At the last step, the 70 S ribosomes were dissolved in buffer G’ (10 mM Hepes-KOH pH 7.5, 30 mM NH₄Cl, 1 mM Mg(OAc)₂, 1 mM DTT) omitting the freezing step ribosomes were directly loaded on 0–30 % w/v sucrose gradients equilibrated in buffer G’ and centrifuged at 61,739 g for 14.5 h using a Beckman SW28 rotor. The concentration of Mg(OAc)₂ was adjusted to 10 mM in the selected pooled fractions. In addition, the sample was supplemented with 2.5 mM of spermidine and concentrated using 100K Amicon ultra centrifugal filters (Merck Millipore). Aliquots were flash-frozen in liquid nitrogen and stored at −80°C.

S. aureus spa mRNA, encoding protein A, has UUG start codon and a strong Shine and Dalgarno (SD) sequence AGGGG. The sequence of the full-length spa mRNA used in the study is shown in Supplementary Figure S2C. The RNA was transcribed from a plasmid as previously described in (Benito et al., 2000; Huntzinger et al., 2005). In brief, the plasmid was linearized by BamHI during 2 h at 37°C, and the mRNA was in vitro transcribed during 3 h at 37°C with T7 RNA polymerase. The mRNA was separated on a 6 % polyacrylamide midi-sized gel and eluted in a solution containing 16 % phenol pH 4.5–5 with 50 mM ammonium acetate and 1 mM EDTA.

The 30 S ribosomal subunits were incubated with spa mRNA during 15 min at 37°C in the buffer containing 10 mM Hepes-KOH pH 7.5, 30 mM NH₄Cl, 10 mM Mg(OAc)₂ in the presence or absence of 2.5 mM spermidine.

4 µL of samples (containing the 30 S bound to mRNA) at 90 nM was applied on Quantifoil R2/2300-mesh holey carbon grids covered with carbon in a temperature- and humidity-controlled Vitrobot Mark IV (T = 4°C, humidity 100%, blotting time 2 s, blotting force 5, waiting time 30 s). The data acquisitions were performed on a Talos Arсtica instrument (FEI Company) at 200 kV using the EPU software on the Falcon three direct detector device (FEI Company). Data were collected at a nominal under focus of −0.5 to −2.7 μm at a magnification of 120,000 X yielding a pixel size of 1.24 Å.

Drift and gain correction, and dose weighting were performed using MotionCor2 (Zheng et al., 2017). A dose weighted average image of the whole stack was used to determine the contrast transfer function with the software Gctf (Zhang, 2016). The following process has been achieved using RELION 3.0 (Zivanov et al., 2018). Particles were picked using a Laplacian of Gaussian function (min diameter 180 Å, max diameter 290 Å). For the 30S-mRNA complex without polyamine, after 2D classification, 256,000 particles were extracted with a box size of 248 pixels and binned four folds for 3D classification into five classes (final = 5.3 Å resolution). For the 30−S-mRNA complex with spermidine, after 2D classification, 529,602 particles were extracted with a box size of 270 pixels and binned three-fold for 3D classification into six classes. Three classes depicting high-resolution features have been selected for refinement. The obtained structure has been refined up to 3.6 Å resolution. Individual focused refinement of the head and the body of the small ribosomal subunit led to the resolution 3.6 Å and 3.4 Å respectively (Supplementary Figures S3A,B). Local resolution estimation was performed in RELION 3.0 using RELION implementation and visualized in Chimera using Surface color option (Supplementary Figure S3C).

As the initial model, we used 30−S extracted from the S. aureus vacant 70S ribosome (PBD 5LI0 [https://doi.org/10.2210/pdb5LI0/pdb]). The initial coarse fitting was performed using the NAMDinator web service (Kidmose et al., 2019), which implements the algorithms of molecular dynamics flexible fitting (MDFF) (Trabuco et al., 2009). The default parameters that were used for flexible fitting (start temperature = 298 K; G-force scaling factor = 0.3; minimization steps = 2000; simulation steps = 20,000). Then the real-space refinement was performed in PHENIX (Afonine et al., 2018) (starting temperature = 800 K; cool rate = 100 K). Ribosomal RNA was corrected in ERRASER web service (Chou et al., 2013), which uses enumerative real-space refinement assisted by electron density under Rosetta protocol. Obtained model was corrected manually in Coot (Supplementary Figure S5); (Emsley and Cowtan, 2004). The model validation was done in MolProbity web service (Chen et al., 2010). Figures featuring cryo-EM densities as well as atomic models were visualized with UCSF Chimera (Pettersen et al., 2004) and ChimeraX (Goddard et al., 2018). The coordinates of the head and the body parts were rigid body fitted into the 30S-mRNA with spermidine density map to build the full 30 S model followed by manual curation in coot.

To analyze the effect of ionic conditions on stability of the 30 S subunit, we performed sucrose gradient sedimentation assays of the 70 S ribosome equilibrated in H₁₀K₂₀₀M₁ (10 mM Hepes-KOH pH 7.5 (at 25^oC), 200 mM KCl, 1 mM Mg(OAc)₂) or H₁₀Na₂₀₀M₁ (10 mM Hepes-KOH pH 7.5 (at 25^oC), 200 mM NaCl, 1 mM Mg(OAc)₂) buffers. The ribosomes sedimented in H₁₀K₂₀₀M₁ buffer dissociated into 50 S and 30 S subunits (Supplementary Figure S1A). Conversely, in presence of 200 mM NaCl, the dissociation of the 70 S ribosomes into the two subunits was not observed (Supplementary Figure S1A). However, the shift of the peak towards light fractions of the gradient and the peak asymmetry suggested the loss of structural integrity of these subunits (Supplementary Figure S1B). To further monitor the integrity of the dissociated ribosomal subunits, the 50 S and 30 S peaks were loaded onto sucrose gradients equilibrated in a buffer containing 200 mM KCl. A bifurcation of the 30 S peak was detected with the appearance of particles with a lower molecular weight than 30 S (Supplementary Figure S1C, right panel) suggesting the inability of the small subunit to withstand 200 mM KCl for a prolonged period of time.

Therefore, we repeated experiments with amended ionic conditions using 100 and 200 mM of NH₄Cl or KCl, respectively at constant 1 mM magnesium acetate. Other conditions included 400 mM NaCl with either six or 10 mM magnesium acetate. For each experiment, the content of the ribosomal proteins within the 30 S peak was analyzed using SDS-PAGE analysis. Despite similarities in the sucrose gradient profiles (Supplementary Figure S1A), higher salt concentrations, especially with 200 mM KCl or 400 mM NaCl, led to the loss of several ribosomal proteins (Supplementary Figure S1B). Taken together, all further experiments were conducted using NH₄Cl, which has the least dissociating effect on the binding of several ribosomal proteins.

After performing a fine analysis of Mg/NH₄Cl balance effect on ribosome dissociation, we selected 1 mM Mg²⁺ in combination with 30 mM NH₄Cl to isolate 30 S from 70S for further cryo-EM analysis (Figures 1A,B). These conditions provided an efficient dissociation of the 70 S ribosome into subunits and concomitantly had their mildest effect on the loss of ribosomal proteins.

Using optimized salt concentrations for 30 S purification, we solved the cryo-EM structure of the 30 S in complex with S. aureus spa mRNA at a resolution of 3.6 Å. The complex was formed in H₁₀NH₃₀M₁₀ (10mM Hepes-KOH pH7.5 (at 25^oC), 30mM NH₄Cl, 10 mM Mg(OAc)₂ and 1mM DTT). Initially, two different datasets were collected in presence (dataset 2) and in absence (dataset 1) of spermidine.

In silico sorting the complex lacking spermidine was divided into five classes (Supplementary Figure S3). In classes 2, 3, and 4, the upper part of h44 near the decoding center (nucleotides 1,414–1,431 and 1,490–1,508) was tilted by about 32°, whereas the lower part of h44 (1,430–1,489) was only partially visible (Figures 2B–D). In addition, in classes 1, 2, and 3, only partial densities of uS2 and bS6 proteins were detected (Supplementary Figure S4).

In the presence of spermidine, h44 acquired more stable conformation (Figure 2A). It is also worth noting that all small subunit ribosomal proteins except bS21 were present in the structures. Indeed, the density of bS21 protein was not identified due to the positioning of spa mRNA in the region where this protein was previously determined (Korostelev et al., 2007; Khusainov et al., 2017). In addition, in both datasets, a high concentration of Mg²⁺ was used for complex formation, which allows us to conclude that polyamines and magnesium ions at these concentrations have different effects on obtained structures.

Open head conformation was observed in all 3D classes obtained from polyamine-deficient 30 S particles; however, mRNA was barely discernible only in class 4 (Figure 3A). When spermidine was added, the mRNA channel was closed by rotation of the 30 S head, allowing spa mRNA to fully accommodate in the channel (Figures 3B,C) similarly to the 70S·mRNA·fMet-tRNA^fMet complex (Golubev et al., 2020). The well-defined density found in the Shine-Dalgarno-anti-Shine-Dalgarno (SD:aSD) region allowed us to fairly accurately build the 30S-mRNA model in this region.