Structure of the Bacillus subtilis hibernating 100S ribosome reveals the basis for 70S dimerization

Abstract

Under stress conditions, such as nutrient deprivation, bacteria enter into a hibernation stage, which is characterized by the appearance of 100S ribosomal particles. In Escherichia coli, dimerization of 70S ribosomes into 100S requires the action of the ribosome modulation factor (RMF) and the hibernation‐promoting factor (HPF). Most other bacteria lack RMF and instead contain a long form HPF (LHPF), which is necessary and sufficient for 100S formation. While some structural information exists as to how RMF and HPF mediate formation of E. coli 100S (Ec100S), structural insight into 100S formation by LHPF has so far been lacking. Here we present a cryo‐EM structure of the Bacillus subtilis hibernating 100S (Bs100S), revealing that the C‐terminal domain (CTD) of the LHPF occupies a site on the 30S platform distinct from RMF. Moreover, unlike RMF, the Bs HPF‐CTD is directly involved in forming the dimer interface, thereby illustrating the divergent mechanisms by which 100S formation is mediated in the majority of bacteria that contain LHPF, compared to some γ‐proteobacteria, such as E. coli.

Introduction

The translational activity of the bacterial cell is able to respond rapidly to a variety of environmental cues. This is exemplified by the decrease in translational activity observed in bacteria entering into stationary growth phase due to stress conditions, such as nutrient deprivation. Under such circumstances, the decrease in translational activity is correlated with the appearance of 100S particles, which arise due to the dimerization of 70S ribosomes (Wada et al, 1990), reviewed by Yoshida and Wada (2014). In E. coli, 100S formation requires the presence of the ribosome modulation factor (RMF) and the hibernation‐promoting factor (HPF, previously referred to as YhbH; Yamagishi et al, 1993; Wada et al, 1995; Maki et al, 2000; Ueta et al, 2005, 2008). Stationary phase E. coli cells also express a homolog of HPF (Fig 1A), termed YfiA (also referred to as pY or RaiA), which binds and inactivates 70S ribosomes (Agafonov & Spirin, 2004; Vila‐Sanjurjo et al, 2004), and is antagonistic to RMF and HPF action by preventing 100S formation (Maki et al, 2000; Ueta et al, 2005). The hibernation state (Yoshida et al, 2002) appears to be important for bacterial survival since inactivation of the rmf gene leads to loss of viability in stationary phase cells (Yamagishi et al, 1993; Wada et al, 2000; Shcherbakova et al, 2015) as well as increased sensitivity to osmotic (Garay‐Arroyo et al, 2000), heat (Niven, 2004), and acid stress (El‐Sharoud & Niven, 2007).

Figure 1. Cryo‐EM reconstruction of the Bs70S‐30S subcomplex.

• A
  Schematic representation of the domain structure of Escherichia coli short form HPF (SHPF), RMF, and YfiA (C‐terminal extension in yellow) compared to Bacillus subtilis long form HPF (LHPF) harboring an N‐terminal (NTD, green) and C‐terminal domains (CTD, red).
• B, C
  Two views of the cryo‐EM map of the Bs70S‐30S subcomplex, with separated densities for the 30S‐A (yellow), 50S‐A (gray), 30S‐B (orange), and additional densities in green and red.

Phylogenetic analyses have revealed that the presence of RMF and HPF is restricted to a subset of γ‐proteobacteria, including E. coli, whereas the majority of other bacteria lack both RMF and YfiA, and instead contain a long form of HPF (LHPF; Fig 1A; Ueta et al, 2008, 2013; Yoshida & Wada, 2014). LHPFs comprise an N‐terminal domain (NTD) homologous to the short form HPF (SHPF) and a unique C‐terminal domain (CTD; Fig 1A), which was proposed to have weak homology with RMF (Ueta et al, 2010). LHPFs have been shown to be necessary and sufficient for 100S formation in a variety of different bacteria, including Staphylococcus aureus (Ueta et al, 2010, 2013; Basu & Yap, 2016), Lactobacillus paracasei, Thermus thermophilus (Ueta et al, 2010, 2013), Lactococcus lactis (Puri et al, 2014), and B. subtilis (Tagami et al, 2012; Akanuma et al, 2016). Unlike E. coli SHPF‐100S (Ec100S), low levels of LHPF‐containing 100S are also observed in exponentially growing cells (Ueta et al, 2010, 2013; Akanuma et al, 2016). Proteomics studies indicate that expression levels of BsLHPF increase under conditions of nutrient deprivation, but also in response to antibiotics, heat, salt, and ethanol stress (Drzewiecki et al, 1998; Reiss et al, 2012; Tagami et al, 2012). In Listeria monocytogenes, LHPF is necessary for tolerance of bacteria to aminoglycoside antibiotics during stationary phase (McKay & Portnoy, 2015) and for optimal fitness and pathogenesis (Kline et al, 2015).

Cryo‐EM and cryo‐electron tomography (cryo‐ET) structures of the Ec100S have revealed that the 70S monomers interact with each other via the back of the 30S subunits (Kato et al, 2010; Ortiz et al, 2010), consistent with earlier negative stain images (Wada, 1998; Yoshida et al, 2002). Unfortunately, the low resolution (18–38 Å) of these structures was insufficient to resolve the binding positions of the RMF and SHPF proteins within the Ec100S (Kato et al, 2010; Ortiz et al, 2010). However, structures of E. coli SHPF and RMF were subsequently determined on the T. thermophilus 70S ribosome by X‐ray crystallography (Polikanov et al, 2012), providing insight into how SHPF and RMF dimerize 70S ribosomes and inactivate translation in γ‐proteobacteria. To date, there is, however, little structural information available as to how LHPFs interact with 70S ribosomes to mediate 100S formation in the majority of bacteria other than E. coli and its close relatives.

Here we present a cryo‐EM structure of the B. subtilis 100S particle (Bs100S) revealing the binding site for the BsHPF (also referred to as YvyD). The BsHPF‐NTD binds in a position overlapping the mRNA, A‐ and P‐tRNAs, analogous to YfiA, SHPF, and the NTD of the LHPF from spinach chloroplasts (Vila‐Sanjurjo et al, 2004; Sharma et al, 2007, 2010; Polikanov et al, 2012; Graf et al, 2016; Bieri et al, 2017), indicating how LHPFs inhibit translation (Ueta et al, 2013; Basu & Yap, 2016). Unexpectedly, we observe that the BsHPF‐CTD forms a homodimer with the CTD of the BsHPF from the second 70S ribosome, thus providing a structural basis for LHPF‐mediated 100S formation. Our findings reveal that 100S formation mediated by RMF and HPF in γ‐proteobacteria, such as E. coli, is mechanistically unrelated to 100S formation mediated by LHPF in the majority of other bacteria.

Results

Cryo‐EM structure of Bs100S

Bs100S ribosomal particles were isolated from lysates of late exponential phase cells using sucrose density gradient centrifugation (Fig EV1A, see Materials and Methods). Negative stain electron microscopy images of the isolated Bs100S revealed the characteristic dimer arrangement of 70S monomers interacting via their 30S subunits (Fig EV1B), as observed previously for B. subtilis (Tagami et al, 2012), Lactococcus lactis (Puri et al, 2014), but distinct from Ec100S (Wada, 1998; Yoshida et al, 2002; Kato et al, 2010). The presence of the BsHPF (YvyD) in the Bs100S was further confirmed using mass spectrometry. The LHPF‐containing 100S particles were then subjected to single particle cryo‐EM analysis (see Materials and Methods). Processing of the Bs100S was performed by aligning the 70S ribosomes within each 100S to a vacant 70S reference. The box size was maintained large enough so that the majority of the small 30S subunit of the second 70S ribosome in the dimer would also be represented during the reconstruction. The initial reconstructions revealed significant flexibility in the 100S, which was indicated by a stable aligned ribosome (70S‐A) with a blurred density for the second 70S ribosome (70S‐B). By implementing in silico sorting procedures, we were able to obtain a subpopulation of 100S particles with better‐defined density for the 70S‐B ribosome (Fig EV2). Subsequent refinement yielded a cryo‐EM reconstruction of the Bs70S‐30S subcomplex (Fig 1B and C) with an average resolution of 3.8 Å (Fig EV3A–D and Table EV1). Local resolution calculations indicate that the resolution for the 70S‐A monomer ranges in the core between 3.5 and 5.0 Å, whereas, as expected, the resolution for 70S‐B is worse, ranging between 5.0 and 10 Å (Fig EV3B and C). The cryo‐EM map was fitted with the molecular model of the B. subtilis 70S ribosome (Sohmen et al, 2015), revealing that the 70S‐A monomer adopts a classic non‐rotated state, as observed previously (Sohmen et al, 2015). Moreover, the swivel of head observed when E. coli SHPF and RMF bind to T. thermophilus 70S ribosomes (Polikanov et al, 2012) is not observed in the Bs100S, indicating that dimerization of B. subtilis 70S ribosomes, unlike E. coli, does not require head movement. After fitting of the 70S models, two unassigned densities remained, one located within the intersubunit space of the 70S‐A ribosome and a second located on the back of the 30S platform at the interface of the 70S‐A and 70S‐B ribosomes (Fig 1B and C).

Figure EV1. Isolation of Bacillus subtilis 100S and sequence alignments of BsHPF with EcHPF‐NTD and CaCTD.

1. Sucrose density gradient profile of B. subtilis extract from late log phase cells, with 30S, 50S, 70S, 100S, and polysome peaks indicated.
2. Negative stain electron microscopy images of purified Bs100S from (A), with selected 70S dimers circled in yellow.
3. PROMALS3D (Pei et al, 2008) sequence alignment of BsHPF‐NTD with Escherichia coli HPF (PDB 4V8H)(Polikanov et al, 2012) that was used to generate the homology model for BsHPF‐NTD.
4. PROMALS3D (Pei et al, 2008) sequence alignment of BsHPF‐CTD with Clostridium acetobutylicum HPF‐CTD (CaCTD; PDB ID 3KA5) that was used to generate the homology model for BsHPF‐CTD.

Data information: In (C) and (D), fully conserved residues are indicated with “9” and are bold in the Consensus_aa, whereas similar residues are indicated with a “+”. Consensus_ss indicates β‐sheet (e) and helical (h) regions.

Figure EV2. In silico sorting and refinement scheme for the Bs70S‐30S subcomplex and complete Bs100S.

• A–C
  253,905 particles were sorted into 10 classes. Class 8 had the most defined density for the 70S‐B and was taken for further refinement using (B) a box size that includes the 70S‐A ribosome and the 30S part of the 70S‐B, and (C) a larger box size that encompasses both the 70S‐A and 70S‐B ribosomes.

Figure EV3. Resolution of 70S‐A in the Bs70S‐30S subcomplex.

• A
  Overview of the Bs70S‐30S subcomplex with 30S‐A (yellow), 50S‐A (gray), and 30S‐B (orange), as well as BsHPF‐NTD (green) and BsHPF‐CTD (red).
• B, C
  Overview (B) and transverse section (C) of the Bs70S‐30S subcomplex colored according to the local resolution, as calculated using ResMap (Kucukelbir et al, 2014).
• D
  Fourier‐shell correlation curve of the refined cryo‐EM map, indicating the average resolution of 70S‐A in the Bs70S‐30S subcomplex is 3.8 Å.
• E
  Fit of models to maps. FSC curves calculated between the refined model and the final map (black), with the self‐ and cross‐validated correlations in orange and red, respectively. Information beyond 4 Å was not used during refinement and preserved for validation.
• F–I
  Map density for the (F, G) BsHPF‐NTD and (H, I) BsHPF‐CTD, which are (F, H) colored according to the local resolution, as calculated using ResMap (see Materials and Methods), or (G, I) shown as a gray mesh with molecular models (G) for BsHPF‐NTD (green) or (I) BsHPF‐CTD for 70S‐A (red) and 70S‐B (blue), using the same respective view as in (F, H).

Binding site of the BsHPF‐NTD on the small 30S subunit

The additional map density within the intersubunit space located between the head and body of the 30S subunit was assigned to the N‐terminal domain of BsHPF (BsHPF‐NTD; Fig 2A). This was based on the high sequence similarity of the BsHPF‐NTD with E. coli YfiA and HPF (Fig EV1C), both of which were shown to bind to this region of the ribosome (Vila‐Sanjurjo et al, 2004; Polikanov et al, 2012). The local resolution of the BsHPF‐NTD ranged between 3.5 and 5.0 Å (Fig EV3F–G), enabling an unambiguous fit of the homology model to the density (Fig 2B). Aligning the E. coli SHPF‐70S structure (Polikanov et al, 2012) to the 70S‐A ribosome in the Bs100S based on the 16S rRNA revealed the expected similarity in their binding positions (Fig 2C). As noted previously for E. coli YfiA and HPF (Vila‐Sanjurjo et al, 2004; Polikanov et al, 2012) and for the NTD of the LHPF from Spinach chloroplast (Sharma et al, 2007, 2010; Graf et al, 2016; Bieri et al, 2017), the binding position of BsHPF‐NTD overlaps with the mRNA and anticodon‐stem loop regions of tRNAs bound in the ribosomal A‐ and P‐sites (Fig 2D), thus explaining the observed inhibitory effect by LHPFs when added to in vitro translation assays (Ueta et al, 2013; Basu & Yap, 2016). The BsHPF‐NTD is connected by a 34 aa linker to the CTD (Fig 1A). Map density for the linker region was not observed in the cryo‐EM map of the Bs100S, indicating that it is highly flexible. An exception is the 5–6 aa stretch of the linker region that directly follows the terminal α‐helix of the BsHPF‐NTD (Fig 2B). Map density for this N‐terminal part of the linker passes, analogous to mRNA, through the opening created by the β‐hairpin of ribosomal protein S7 and helix h23 of the 16S rRNA, and extends in the general direction of the platform cavity at the back of the 30S subunit (Fig EV4).

Figure 2. Interaction of the BsHPF‐NTD with the ribosome.

• A
  Interface view of cryo‐EM map of the 30S‐A (yellow) from the Bs70S‐30S subcomplex with separated BsHPF‐NTD density (green).
• B
  Map density (gray mesh) with model of BsHPF‐NTD (green).
• C, D
  Comparison of BsHPF‐NTD (green) with (C) Escherichia coli SHPF (EcHPF, blue; Polikanov et al, 2012), and (D) mRNA (yellow surface), A‐ (cyan), P‐ (gray), and E‐tRNAs (orange; Jenner et al, 2011).

Figure EV4. BsHPF linker region approaches the 30S platform cavity.

• A
  Overview of the 30S cavity region showing BsHPF‐NTD (green) and BsHPF‐CTD (red) and 30S (yellow), except S2 (cyan), S7 (blue), and S18 (purple).
• B
  Zoom of (A), showing map density (gray mesh) for the N‐terminal part of the linker region of BsHPF (green) as well as for the 3′ end of the 16S rRNA.
• C
  Coomassie (upper panel) and Western blot of cell extracts (CE) and ribosome pelleted fractions (R) of the wild‐type Bs168 (wt) strain or the ∆BsHPF strains expressing either wild‐type BsHPF, BsHPF‐(GS)₅, or BsHPF‐(GS)₁₀.
• D, E
  Sucrose gradient profiles of cell extracts from the (D) Bs168 ∆BsHPF amyE::BsHPF‐(GS)₅ strain and (E) Bs168 ∆BsHPF amyE::BsHPF‐(GS)₁₀ strain, in the absence (I−) or presence (I+) of IPTG.

BsHPF‐CTD is present as a dimer on the small 30S subunit

Given the general direction of the linker, we assigned the additional density located on the back of the 30S platform to the BsHPF‐CTD (Fig 3A and B). It was possible to generate a homology model for the BsHPF‐CTD based on the deposited crystal structure of the LHPF‐CTD from a closely related Firmicute, Clostridium acetobutylicum (PDB ID 3KA5; Fig EV1D). Curiously, the C. acetobutylicum LHPF‐CTD is present as a dimer in the crystal, and it was possible to make an unambiguous rigid body fit of the homology model of the BsHPF‐CTD dimer into the unassigned map density of the cryo‐EM map (Fig 3C). We note that while the structurally conserved L. monocytogenes HPF‐CTD (PDB ID 3K2T) appears as a monomer in the asymmetric unit, the homodimer forms across the crystallographic twofold symmetry. This suggests that the LHPF‐CTDs are not only dimeric on the ribosome, but are likely to be dimeric in solution. To investigate this further, we performed size‐exclusion chromatography (SEC) on the recombinantly expressed and purified wild‐type BsHPF and BsHPF variants (see Materials and Methods). Analysis of the full‐length BsHPF and the BsHPF‐CTD revealed that they have apparent molecular masses of 56 and 14 kDa, respectively, rather than the expected 23 kDa and 8 kDa (Fig 3D–G), indeed suggesting that both proteins are dimeric in solution as well as on the ribosome. The apparent migration behavior of BsHPF on SEC reflects the elongated shape of the dimer as also seen in our cryo‐EM structure of the Bs100S. Based on the structures of the dimeric C. acetobutylicum and L. monocytogenes LHPF‐CTD, we rationalized that the highly conserved Phe160 in the BsHPF‐CTD is critical for dimerization (Fig 3H). Phe160 is present within the very hydrophobic dimer interface where it forms stacking interactions with Phe160 of the second monomer (Fig 3H). We predicted that a mutation of Phe160 to Glu (F160E) would disrupt the dimer interface via introduction of a negative charge into the hydrophobic environment. To test this, we also subjected the full‐length BsHPF‐F160E protein to SEC (Fig 3D and E), revealing that the protein eluted with an apparent molecular mass of 40 kDa, smaller than the 56 kDa observed for the wild‐type BsHPF (Fig 3G). Although 40 kDa is larger than the expected size of 22.8 kDa, we believe this is due to retardation of the NTD and subsequent linker. Indeed, a BsHPF variant lacking the CTD (BsHPF‐NTD) eluted with an apparent molecular mass of 28 kDa (rather than the expected 13.1 kDa; Fig 3G). This observation is in good agreement with structural information on the NTDs of other hibernation factors showing a non‐globular shape (Polikanov et al, 2012). Our conclusions based on SEC were also confirmed using static light scattering (SLS), revealing the full‐length BsHPF had an absolute molecular mass of 42.8 ± 0.9 kDa, corresponding with a dimer (46 kDa), whereas the mass of the BsHPF‐F160E variant (28 ± 2.1 kDa) was more consistent with a monomer (22.8 kDa; Fig 3G). Taken together, our biochemical data clearly show that BsHPF forms a homodimer in solution that is mediated via its CTD.

Figure 3. Binding site of dimeric LHPF‐CTD on the Bs70S‐30S subcomplex.

• A
  Cryo‐EM map of the 30S‐A (yellow) from the Bs70S‐30S subcomplex with separated LHPF‐CTD density (red).
• B, C
  Density (gray mesh) with fitted model of dimeric LHPF‐CTD with monomers from 70S‐A and 70S‐B colored red and blue, respectively.
• D
  Gel‐filtration profiles of full‐length BsHPF (blue), BsHPF‐F160E (yellow), BsHPF‐NTD (green), and BsHPF‐CTD (red). Arrows indicate the molecular mass in kDa of the size standard.
• E
  Standard curve with estimated molecular masses for full‐length BsHPF (blue), BsHPF‐F160E (yellow), BsHPF‐NTD (green), and BsHPF‐CTD (red). Arrows indicate the molecular mass in kDa of the size standard.
• F
  Coomassie‐stained SDS–PAGE of the peak fractions containing BsHPF or its variants.
• G
  Table summarizing the actual and apparent molecular mass of proteins in (D‐F). Size‐exclusion chromatography (SEC) and static light scattering (SLS) determined the apparent and absolute MWs, respectively. “Stoichiometry” indicates whether BsHPF and its variants exist as mono‐ or homodimer.
• H
  Homology model of the BsHPF‐CTD homodimer illustrating the position of Phe160 (F160) at the dimer interface.

Dimerization of 70S ribosomes via the BsHPF‐CTD

While the limited resolution of the BsHPF‐CTD (Fig EV3H and I) does not allow a detailed analysis of the contacts with the ribosomal components to be made, the fitted model nevertheless enables a general description of the interaction mode (Fig 4A). The BsHPF‐CTD appears to interact exclusively with ribosomal proteins S2 and S18 and does not establish contact with the 16S rRNA. Importantly, each BsHPF‐CTD monomer contacts S2 from the 70S to which the corresponding BsHPF‐NTD is bound, whereas the interaction with the N‐terminal extension of S18 is from the second 70S ribosome (Fig 4A). The 100S dimer is also stabilized by direct interactions between the 70S‐A and 70S‐B monomers (Fig 4A and B). In addition to the contacts established between the N‐terminal helix of S2 and the N‐terminal extension of S18, the N‐terminal β‐hairpin and proximal region of the α2‐helix of S2 establish a large interaction surface with the stem‐loop of helix h26 of the 16S rRNA of the second 70S (Fig 4B). Thus, the dimerization of the HPF‐CTDs stabilizes and facilitates direct interaction between the 70S‐A and 70S‐B monomers in the Bs100S. Our findings highlight the importance of the BsHPF‐CTD for 70S dimerization, and therefore 100S formation, which is in complete agreement with biochemical studies demonstrating that truncation of the CTD from LHPF leads to loss of 100S formation (Puri et al, 2014; Basu & Yap, 2016). Moreover, it was reported that the CTD of the LHPF from Lactococcus lactis can dimerize E. coli 70S ribosomes, but only when acting in concert with the SHPF from E. coli (Puri et al, 2014). This observation supports to some extent the previous assertion that the HPF‐CTD functions analogously to RMF; an assertion that was partly based on proposed sequence homology between HPF‐CTD and RMF (Ueta et al, 2010). However, comparison of the structures of BsHPF‐CTD with that of RMF on the ribosome (Polikanov et al, 2012) reveals that there is no structural similarity in terms of the protein fold and, despite both binding at the platform region at the back of the 30S subunit, there is no overlap in their binding sites on the ribosome (Fig 4C). The binding position of RMF was suggested to inhibit translation by sterically preventing formation of the Shine‐Dalgarno‐helix (SD‐helix) between the 5′ end of the mRNA and the 3′ end of the 16S rRNA (Polikanov et al, 2012). In contrast, the HPF‐CTD does not overlap with the SD‐helix (Fig 4D), although we cannot exclude the possibility that the flexible linker of BsHPF traverses the RMF binding site since it was not visualized in the cryo‐EM map.

Figure 4. Dimerization interface of the Bs70S‐30S subcomplex.

• A, B
  Distinct views of the dimer interface between 30S‐A (yellow) with BsHPF‐CTD‐A (red) and 30S‐B (gray, darker yellow with dashed line in zoomed panel) with BsHPF‐CTD‐B (blue). Ribosomal proteins S2 (cyan), S18 (purple), and 16S rRNA are shown only, and the surface outline of the 30S subunit is included schematically for reference.
• C, D
  Binding site of BsHPF‐NTD (green) and dimeric BsHPF‐CTD (red, blue) relative to (C) RMF (orange; Polikanov et al, 2012) and (D) SD–anti‐SD helix (yellow‐purple surface; Sohmen et al, 2015). The dashed line indicates the linker and is shown only to illustrate that the 34 amino acids are more than sufficient to connect the NTD and CTD; however, no density for the linker was observed, suggesting it does not adopt a defined conformation on the ribosome.

Importance of the linker–CTD for 100S formation

To assess the importance of the linker and CTD of BsHPF for 100S formation in vivo, we generated a B. subtilis 168 strain where the yvyD gene was inactivated (∆BsHPF), as confirmed by Western blotting using antibodies specific to BsHPF (Fig 5A). We then re‐introduced the wild‐type yvyD gene, as well as yvyD variants, into the amyE locus and monitored the IPTG‐induced expression of the BsHPFs (Fig 5A). To investigate the importance of the linker between the NTD and CTD of BsHPF, we generated ∆BsHPF strains expressing BsHPF deletion variants lacking 10 aa (BsHPF‐L∆10AA, lacking residues 110–119) or 20 aa (BsHPF‐L∆20AA, lacking residues 105–124) from the central region of the linker (Fig 5A). In addition, we generated a BsHPF variant bearing the F160E mutation in the CTD (BsHPF‐F160E), which interferes with homodimerization (Fig 3G). Western blotting of cell extracts from stationary phase bacteria indicated that all BsHPF variants inserted into the amyE locus were expressed in the presence of IPTG at similar levels to wild‐type BsHPF observed in the parental Bs168 strain (Fig 5A). Pelleting experiments indicated that full‐length BsHPF co‐migrated with the ribosome fraction as expected, as did the BsHPF‐L∆10AA variant (Fig 5B). In contrast, the BsHPF‐L∆20AA and BsHPF‐F160E variants had significantly reduced association with the ribosomal pellets (Fig 5B), suggesting that the deletion of 20 aa within the linker or preventing homodimerization via the CTD disrupts the interaction of BsHPF with the ribosome. This is consistent with previous studies using S. aureus LHPF where C‐terminal truncations of 42 aa (∆CTD) and 90 aa (∆linker–CTD) led to progressive loss in ribosome binding (Basu & Yap, 2016).

Figure 5. Monitoring 100S formation in vivo for BsHPF variants.

• A
  Western blot using antibodies raised against BsHPF to assess the levels of BsHPF in cell extracts of wild‐type Bs168 (wt), ∆BsHPF, and ∆BsHPF strains expressing either wild‐type BsHPF or BsHPF‐L∆20AA, BsHPF‐L∆10AA, BsHPF‐F160E variants.
• B
  Coomassie (above) and Western blot of cell extracts (CE) and ribosome pelleted fractions (R) of the wild‐type Bs168 (wt) strain or the ∆BsHPF strains expressing either wild‐type BsHPF, BsHPF‐L∆10AA, BsHPF‐L∆20AA, and BsHPF‐F160E.
• C
  Sucrose gradient profiles of cell extracts from the wild‐type Bs168 (wt) strain in exponential phase (blue) or heat stressed (red), compared with the extract from the Bs168 ∆BsHPF strain (dashed line).
• D–G
  Sucrose gradient profiles of cell extracts from the (D) Bs168 ∆BsHPF amyE::BsHPF strain, (E) Bs168 ∆BsHPF amyE::BsHPF‐L∆20AA strain, (F) Bs168 ∆BsHPF amyE::BsHPF‐L∆10AA strain, and (G) Bs168 ∆BsHPF amyE::BsHPF‐F160E strain in the absence (I−) or presence (I+) of IPTG. The dashed line of the Bs168 ∆BsHPF strain from (C) is shown for reference.
• H
  Growth curves illustrating the recovery from stationary phase of the wild‐type Bs168 (wt), ∆BsHPF, and ∆BsHPF strains expressing either wild‐type BsHPF or BsHPF‐L∆20AA, BsHPF‐L∆10AA, BsHPF‐F160E variants.

Source data are available online for this figure.

We next employed sucrose density gradient centrifugation to monitor the formation of 100S ribosomes using the different Bs168 strains (Fig 5C–G). As controls, the wild‐type Bs168 strain was harvested during exponential phase, where a large 70S peak and lots of polysomes were observed, but little or no 100S were evident (Fig 5C). In contrast, a short heat treatment of the wild‐type cells led to a complete loss of polysomes and the appearance of a prominent 100S peak (Fig 5C), as observed previously for B. subtilis (Akanuma et al, 2016). Formation of 100S was never observed in the ∆BsHPF strain (Fig 5C) regardless of the stress conditions tested, in agreement with the strict dependence on BsHPF for 70S dimerization (Akanuma et al, 2016). However, when the yvyD gene was reintroduced into the amyE locus of the ∆BsHPF strain, 100S formation (and loss of polysomes) was observed, but only when BsHPF expression was induced by the presence of IPTG (Fig 5D). No significant increase in the 100S peak, nor reduction in polysomes, was observed when expression of the BsHPF‐L∆20AA variant was induced (Fig 5E), consistent with the lack of ribosome binding (Fig 5B). Surprisingly, similar results were obtained for BsHPF‐L∆10AA (Fig 5F), suggesting that although the BsHPF‐L∆10AA can still bind to the ribosome (Fig 5B), it is impaired in 100S formation. BsHPF variants where the 10 aa or 20 aa were substituted (rather than deleted) by glycine‐serine (GS) repeats, creating BsHPF‐(GS)₅ or BsHPF‐(GS)₁₀, respectively, also led to both a reduction in ribosome binding and 100S formation (Fig EV4C–E), suggesting that the sequence and not just the length of the linker is critical for BsHPF activity. Lastly, we also monitored 100S formation in the Bs168 strain expressing the BsHPF‐F160E variant. As expected, no increase in the 100S peak or decrease in the polysome peaks was observed upon BsHPF‐F160E induction (Fig 5G), indicating that BsHPF‐CTD homodimerization is necessary for 100S formation.

Further support for the loss of activity of the BsHPF‐L∆20AA and BsHPF‐F160E variants comes from growth assays. Compared to the wild‐type Bs168 strain, the ∆BsHPF strain exhibits a lag phase when stationary phase cells are diluted into fresh media (Fig 5H), as reported previously (Akanuma et al, 2016). The lag phenotype can be restored by expression of wild‐type BsHPF, but not by BsHPF‐L∆20AA and BsHPF‐F160E variants (Fig 5H). Curiously, the BsHPF‐L∆10AA variant also rescued the growth phenotype (Fig 5H), suggesting that ribosome binding rather than 100S formation may be important for efficient stationary phase recovery.

Distinct arrangement of 70S monomers in the Bs100S

In order to obtain a reconstruction of the complete Bs100S particle to compare with previous Ec100S reconstructions, we also reprocessed the cryo‐EM data using a larger box size that completely encompassed both 70S monomers (Fig EV2). Despite the inherent flexibility between the 70S monomers, we were able to obtain a reconstruction of the Bs100S (Fig 6A) with an average resolution of 6.2 Å (Fig EV5A–C). The relative orientation of the 70S‐A and 70S‐B monomers within the Bs100S was related by a 180° rotational symmetry with the axis of rotation centered on the dimeric BsHPF‐CTD (Fig EV5D). As expected, we observed additional density for the HPF‐NTD within the intersubunit space and for the HPF‐CTD at the back of the 30S subunit (Fig EV5E and F). Comparison of the Bs100S with the previous cryo‐EM and cryo‐ET reconstructions of the Ec100S (Fig 6B) revealed a dramatically different monomer arrangement (Fig 6C). While Ec100S dimerization involves a “back‐to‐back” interaction of the 30S subunits of each 70S monomers, Bs100S dimerization involves a more “side‐to‐side” (platform‐to‐platform) interaction of the 30S subunits. In the Ec100S, dimerization is proposed to be stabilized by contacts between S2 on one 70S with the cavity formed by S3/S4/S5 on the other (Kato et al, 2010), which may be facilitated by a swivel movement of the head of the 30S subunit that was observed upon RMF binding (Polikanov et al, 2012). In contrast, the head position of the Bs100S is identical to that observed in the classic post‐translocational state ribosome (Sohmen et al, 2015) and, unlike RMF, the BsHPF‐CTD directly comprises part of the dimerization interface. The spatial orientation of the 70S monomers in the Bs100S (Fig 6A) could be considered intermediate between that observed in the Ec100S (Fig 6B; Kato et al, 2010) and the orientation observed in the cryo‐ET reconstructions of E. coli polysomes (Fig 6D; Brandt et al, 2009).

Figure 6. Spatial organization of Bs100S, Ec100S, and polysomes.

• A–D
  Comparison of the 70S‐A and 70S‐B monomer arrangement in (A) Bs100S, compared with (B, C) Ec100S (Ortiz et al, 2010) and (D) Escherichia coli polysomes (Brandt et al, 2009). The 30S‐A (yellow), 30S‐B (orange), 50S (gray), BsHPF‐NTD (green), and BsHPF‐CTD (red) are colored for reference, and schematics of the Ec100S are presented in (C) and (D) for ease of comparison.

Figure EV5. Resolution of the complete dimeric Bs100S.

• A
  Fourier‐shell correlation curve of the refined cryo‐EM map, indicating the average resolution of 70S‐A, 70S‐B, and the complete Bs100S is 5.3, 6.9 and 6.2 Å, respectively.
• B, C
  Cryo‐EM map of the dimeric Bs100S colored according to local resolution showing (B) overview and (C) transverse section of the complete 100S disome.
• D
  The 70S‐A and 70S‐B monomers in the Bs100S are related by rotational symmetry of ˜180°.
• E, F
  Cryo‐EM map of the (E) dimeric Bs100S with 30S‐A (yellow), 30S‐B (orange) and 50S (gray), and (F) transverse section of (E) highlighting the densities for the BsHPF‐NTD (green) and BsHPF‐CTD (red).

Discussion

The appearance of hibernating 100S ribosomes is a near universal response of bacteria to adapt to a variety of stress conditions, in particular nutrient limitation (Ueta et al, 2013; Yoshida & Wada, 2014). Under these circumstances, bacteria employ second messenger signaling molecules, such as (p)ppGpp and cyclic AMP (cAMP), to reprogram the cellular activity network, down‐regulating genes associated with translation and up‐regulating stress response and amino acid biogenesis pathways (Hauryliuk et al, 2015; Steinchen & Bange, 2016). In E. coli, transcription of rmf, the gene encoding RMF, is up‐regulated by (p)ppGpp when amino acids become limiting (Izutsu et al, 2001) and by cAMP upon carbon starvation (Shimada et al, 2013). Transcription of yvyD, the gene encoding BsLHPF, is under the control of the sigma factors σ^H and σ^B (Drzewiecki et al, 1998; Tam le et al, 2006; Akanuma et al, 2016), and up‐regulated by the presence of the alarmone (p)ppGpp (Eymann et al, 2001; Tagami et al, 2012; Shimada et al, 2013; Fig 7A). Similarly, in the cyanobacterium Synechococcus elongatus, LHPF is also up‐regulated by (p)ppGpp to enable dark adaptation (Hood et al, 2016).

Figure 7. Model for BsHPF‐induced 100S formation.

• A
  Stress conditions, such as nutrient deprivation, lead to elevated levels of (p)ppGpp, which up‐regulates expression of the LHPF (NTD, green; CTD, red). The LHPF‐CTD can interact to form homodimers in solution and therefore may also be present as dimers in the cell.
• B
  The long linker of the dimeric LHPF enables the LHPF‐NTD to interact with two independent 70S ribosomes and by bringing them in to close proximity stabilizes the 70S dimers, forming 100S.
• C–E
  Following removal of the stress conditions, BsHPF levels decline leading to (C) dissociation of 100S into 70S ribosomes and (D) eventually ribosome splitting into 30S and 50S subunits, or (E) alternatively directly in 30S and 50S subunits.

The up‐regulation of LHPF leads to increased 100S formation, indicating that LHPF competes effectively with translation factors, as evidenced by LHPF inhibition of in vitro translation systems (Ueta et al, 2013; Basu & Yap, 2016). Since we observed that BsHPF is dimeric in solution, we favor a model whereby dimeric BsHPF interacts independently with two 70S ribosomes (Fig 7B). In this model, we propose that BsHPF utilizes the free NTDs and long linker to initially bring 70S ribosomes into close proximity, and then further stabilizes the 70S dimer using the BsHPF‐CTD ribosome interface (Fig 7B). However, we cannot exclude that at a fraction of BsHPF resides as a monomer in vivo, and these BsHPF monomers bind separately to the 70S ribosome, such that 100S formation could then occur concomitantly with BsHPF‐CTD homodimerization. Moreover, it remains unclear how hibernating 100S ribosomes exactly provide protection against stress. What is clear however is that in the absence of 100S, 70S ribosomes are slowly degraded leading to early cell death, suggesting that hibernating 100S are less susceptible to degradation by RNases (Fukuchi et al, 1995; Wada, 1998; Niven, 2004; Shcherbakova et al, 2015; Akanuma et al, 2016). Because 100S formation does not significantly alter the large rRNA surface exposed to RNases, we believe LHPF binding and 100S formation may interfere with a specific ribosome degradation pathway, rather than preventing non‐specific RNase action on ribosomes. The identification of BsHPF variant, such as the BsHPF‐L∆10AA, which binds to the ribosome but does not promote 100S formation, may allow the contribution of these activities to ribosome protection to be dissected further.

In E. coli, disassembly of 100S is rapid and occurs within 1 min upon transfer to fresh medium, suggesting that an active mechanism exists to remove EcHPF and RMF from the ribosome (Wada, 1998; Aiso et al, 2005). In contrast, Bs100S are more stable than Ec100S (Ueta et al, 2013) and upon transfer to fresh media significant dissociation of Bs100S was only observed after 120 min, where LHPF protein levels were also significantly decreased (Akanuma et al, 2016). Nevertheless, recycling factors, such as IF3, RRF, and EF‐G, which have been reported to remove LHPF (PSRP‐1) from Spinach chloroplast ribosomes (Sharma et al, 2010), might also be involved in BsHPF release and 100S dissociation (Fig 7C–E).

In conclusion, the high conservation of the LHPF proteins suggests that most, if not all, LHPF proteins are present as dimers in the cell, with the implication that the majority of bacteria are likely to utilize an identical mechanism to induce 100S formation as we have described here for B. subtilis (Fig 7).

Materials and Methods

Cloning of BsHPF and BsHPF variants for protein purification

The yvyD gene encoding BsHPF was amplified from B. subtilis PY79 genomic DNA by polymerase chain reaction using Phusion High‐Fidelity DNA Polymerase (NEB) according to the manufacturer's manual. The forward primer encoded a hexa‐histidine tag in‐frame with the DNA sequence of yvyD. The fragment was cloned via NcoI/XhoI restriction sites into a modified pGAT2‐vector incorporating a GST‐tag N‐terminal of His₆‐BsHPF. BsHPF‐CTD (amino acids 130–189 of BsHPF), BsHPF‐NTD (amino acids 1–104 of BsHPF), and BsHPF(F160E) containing an N‐terminal hexa‐histidine tag were amplified by PCR as described above and cloned via NcoI/XhoI restriction sites into pET24d(+) vector (Novagen). Mutations within BsHPF were introduced by overlapping PCR.

Protein production and purification for SEC and SLS

Escherichia coli BL21(DE3) cells (NEB) carrying the expression plasmid were grown in lysogeny broth (LB) medium supplemented with ampicillin (100 μg/ml) or kanamycin (50 μg/ml) and D(+)‐lactose‐monohydrate (12.5 g/l) for 16 h at 30°C under rigorous shaking (180 rpm). The cells were harvested (3,500 × g, 20 min, 4°C), resuspended in lysis buffer (20 mM HEPES‐KOH, pH 8.0, 20 mM KCl, 20 mM MgCl₂, 500 mM NaCl, 40 mM imidazole) and lysed using a M‐110L Microfluidizer (Microfluidics). After centrifugation (47,850 × g, 20 min, 4°C), the clear supernatant was loaded on a HisTrap HP 1 ml column (GE Healthcare) equilibrated with 15 column volumes (CV) of lysis buffer. After washing with 15 CV of lysis buffer, the protein was eluted with 5 CV of elution buffer (lysis buffer containing 500 mM imidazole). The GST‐tag was removed from BsHPF variants by incubation with 100 U of bovine thrombin (Merck Millipore) for 2 h at 20°C. After dilution with 12 volume parts of lysis buffer without imidazole, BsHPF variants were resubjected to Ni‐NTA affinity chromatography as described above and the elution fraction containing BsHPF were collected. BsHPF and BsHPF variants were then concentrated using an Amicon Ultracel‐10K or 3K, respectively (Merck Millipore), and applied to size‐exclusion chromatography (HiLoad 26/600 Superdex 75 pg, GE Healthcare) equilibrated in SEC buffer (20 mM HEPES‐KOH, pH 8.0, 20 mM KCl, 20 mM MgCl₂, 500 mM NH₄Cl). Protein‐containing fractions were pooled and concentrated to ~500 μM as determined by a spectrophotometer (NanoDrop Lite, Thermo Scientific).

Analysis of oligomerization states of BsHPF variants by SEC and SLS

The apparent molecular weight was analyzed by size‐exclusion chromatography using a HiLoad 26/600 Superdex 75 pg column (GE Healthcare) equilibrated in SEC buffer. A standard curve for molecular mass determination was obtained using BSA (66.5 kDa), ovalbumin (chicken, 44.3 kDa), myoglobin (horse, 17 kDa), and vitamin B12 (1.35 kDa). The absolute molecular weight was determined by static light scattering (SLS) with a DelsaMax CORE (Beckmann Coulter) according to the manufacturer's instructions.

Cloning of BsHPF and BsHPF variants for in vivo studies

Full‐length yvyD was amplified from genomic DNA by PCR as described above with the forward primer encoding the strong ribosomal binding site of the gsiB gene (AGGAGGAATTCAAA) and cloned via SalI/SphI restriction sites into the pDR111 plasmid (Ben‐Yehuda et al, 2003). The BsHPF‐LΔ10AA, BsHPF‐LΔ20AA, and BsHPF‐F160E mutation were introduced by overlap extension PCR and cloned via SalI/SphI restriction sites as described above. The resulting plasmids were linearized by digestion with ScaI and transformed into naturally competent B. subtilis cells. Proper integration into the amyE locus was checked by growing selected transformants on LB‐Agar containing 1% starch overnight and staining the plates with a solution of 0.5% (w/v) iodine, 1% (w/v) potassium iodine. Strains and oligonucleotides used in this study are presented in Tables EV2 and EV3.

Western blotting of BsHPF variants

Strains expressing HPF variants in trans were grown in LB medium supplemented with 1 mM IPTG with rigorous shaking to until the mid‐exponential phase (OD₆₀₀ of ~0.8), harvested by centrifugation at 11,000 × g, 4°C for 5 min, washed once in TE buffer (10 mM Tris–HCl, 1 mM EDTA, pH 8.0), and disrupted by sonication three times for 30 s on ice in TE buffer supplemented with 1 mM PMSF. The soluble protein was cleared from cell debris by centrifugation at 11,000 × g, 4°C for 5 min. 10 μg of the protein extract (as determined by the Bradford assay) was analyzed by SDS–PAGE and Western Blotting onto a nitrocellulose membrane. As controls, equally treated samples from a stationary phase overnight culture of B. subtilis wild‐type or Δhpf cells were loaded. The BsHPF protein was detected using a polyclonal antibody raised against BsHPF (Pineda Antibody Service) and a polyclonal Goat anti‐Rabbit IgG alkaline Phosphatase conjugated antibody (Antikörper Online). Western blotting using an antibody against the malate dehydrogenase (MDH) was used as a loading control. The ECF reagent (GE Healthcare) was used as a substrate according to the manufacturer's manual, and chemifluorescent signals were detected using a cooled CCD camera in a ChemiBIS 4.2 Bioimaging system (DNR).

Binding assay for BsHPF variants with pelleted ribosomes

Bacillus subtilis cells were grown in 200‐ml LB medium supplemented with 1 mM IPTG with rigorous shaking (200 rpm) to the mid‐exponential phase (OD₆₀₀ ~0.8) and harvested by centrifugation at 15,300 × g, 10 min, 4°C. Ribosomes were pelleted as described in (Schmalisch et al, 2002). Briefly, cells were washed once in buffer A (20 mM Tris–HCl, 100 mM NH₄Cl, 10 mM MgCl₂, 10 mM 2‐mercaptoethanol, pH 7.5), resuspended in 3 ml of the same buffer with 1 mM PMSF and disrupted in a French Pressure Cell three times at 1,000 psi. The lysate was cleared from cell debris by centrifugation for 30 min at 29,953 × g, 4°C (SW55‐Ti, Beckman Coulter), layered on top of a 8 ml 1.1 M sucrose cushion in buffer A, and centrifuged for 16 h at 119,307 × g, 4°C (SW40‐Ti, Beckman Coulter). The cell pellet was washed three times in buffer A and resuspended in buffer B (20 mM Tris–HCl, 100 mM NH₄Cl, 6 mM MgCl₂, 2 mM DTT). The suspension was centrifuged at 10,000 × g, 10 min, 4°C, and the supernatant containing the ribosomes was collected. 10 μg of the total soluble protein (“CE”, as determined by the Bradford assay) and an equal volume of the ribosome suspension (“R”) was subjected to 15% SDS–PAGE and subsequent stained with Coomassie using standard procedures or Western blotting as described above.

Growth recovery from stationary phase

Precultures of B. subtilis cells were grown in 5‐ml LB medium supplemented with 0.5 mM IPTG at 37°C for 18 h, to ensure the cells reached the stationary growth phase. The cultures were then diluted to an OD₆₀₀ of 0.05 into 20‐ml fresh LB medium and grown at 37°C with rigorous shaking. The cell growth was monitored by determining the optical density at 600 nm (OD₆₀₀) at regular intervals.

Sucrose density gradient centrifugation analysis

Analysis of 100S formation was performed as described previously for B. subtilis (Akanuma et al, 2016). Briefly, 50‐ml LB medium was inoculated at a 1:100 dilution with an overnight culture. Expression was induced using 1 mM IPTG at an OD₆₀₀ of 0.4. Cells were harvested at the stationary phase by centrifugation at 4,000 × g for 10 min at 4°C (Hettich Rotanta 46R) and the cell pellet re‐suspended in buffer C (50 mM HEPES‐KOH, pH7.4, 100 mM KOAc, 25 mM Mg(OAc)₂, 6 mM β‐mercaptoethanol). Cells were lysed using the sonifier three times, with each cycle consisting of 30 s at 30% power followed by centrifugation at 16,000 × g for 15 min at 4°C to remove cellular debris. A total OD₂₆₀ of 10 of the cleared lysate was loaded onto sucrose density gradients (10–60% sucrose in buffer C) by centrifugation at 154,693 × g (SW‐40 Ti, Beckman Coulter) for 3 h at 4°C and then analyzed using a Gradient Station (Biocomp) with an Econo UV Monitor (Bio‐Rad) and a FC203B Fraction Collector (Gilson).

Preparation of Bacillus subtilis S12 extract

Bacillus subtilis S12 extract was prepared as described (Sohmen et al, 2015). Briefly, an “INFORCE HT minifors” bench top fermenter was used to grow B. subtilis strain 168 cells to an OD₆₀₀ 4.5 in 2× YPTG medium (16 g/l peptone, 10 g/l yeast extract, 5 g/l NaCl, 22 mM NaH₂PO₄, 40 mM Na₂HPO₄, 19.8 g/l glucose (sterile filtered)), with extra glucose feeding at 37°C while maintaining a pH 7.0 and oxygen level (60%). After collecting cells at 5,000 × g at 4°C for 15 min, they were washed 3× in cold Buffer A (10 mM Tris–acetate (pH 8.2), 14 mM magnesium acetate, 60 mM potassium glutamate, 1 mM dithiothreitol, and 6 mM β‐mercaptoethanol). Cells were then snap‐frozen in liquid nitrogen and stored at −80°C. 15 g of cells was thawed on ice, resuspended in 10 ml of cold buffer B (buffer A missing β‐mercaptoethanol), and lysed 3× at 15,000 psi in an “microfluidics model 110I lab homogenizer”. The lysate was cleared at 12,000 × g and 4°C for 10 min and incubated in a water bath for 30 min at 37°C. The cell extract was aliquoted, snap‐frozen, and stored at −80°C. Extracts were analyzed on sucrose density gradients (10–50% sucrose in buffer C), by centrifugation at 89,454 × g (SW‐28, Beckman Coulter) for 4 h at 4°C. For 100S purification, 100S fractions were collected using a Gradient Station (Biocomp) with an Econo UV Monitor (Bio‐Rad) and a FC203B Fraction Collector (Gilson). Purified 100S ribosomes were concentrated by centrifugation at 92,159 × g for 2.5 h at 4°C (TLA110 rotor, Beckman Coulter).

Negative stain electron microscopy

Ribosomal particles were diluted in buffer C to a final concentration of 0.2 OD₂₆₀/ml. A 3.5 μl sample was applied onto a carbon‐coated grid. After 30 s, the grids were washed with distilled water and then stained with 2% aqueous uranyl acetate for 15 s. The remaining liquid was removed by touching the grid with filter paper. Micrographs were taken using a Morgagni transmission electron microscope (FEI).

Cryo‐electron microscopy and single particle reconstruction

A total of 4 OD₂₆₀/ml Bs100S sample were applied to 2 nm pre‐coated Quantifoil R3/3 holey carbon‐supported grids and vitrified using Vitrobot Mark IV (FEI Company). Data collection was performed using EM‐TOOLS (TVIPS GmbH) on a Titan Krios transmission electron microscope equipped with a Falcon II direct electron detector (FEI Company) at 300 kV at a pixel size of 1.084 Å and a defocus range of 0.7–2.2 μm. Ten frames (dose per frame of 2.5 e⁻/Å) were aligned using Motion Correction Software (Li et al, 2013). Power‐spectra, defocus values, and astigmatism were then determined using CTFFIND4 software (Rohou & Grigorieff, 2015). Micrographs showing Thon rings beyond 3.5 Å were manually inspected for a good areas and power‐spectra quality. Automatic particle picking was then performed using SIGNATURE (Chen & Grigorieff, 2007), and single particles were windowed out in small box able to contain a 70S ribosome together with the majority of the small 30S subunit of the neighboring 70S ribosome. The particles were then further processed using FREALIGN (Grigorieff, 2007). The 253,905 particles were first subjected to an extensive 3D classification (Fig EV2A and B), and the selected 24,546 Bs100S particles of class 8 were then subjected to refinement using 30S‐70S mask resulting in a final reconstruction of 3.8 Å (0.143 FSC) average resolution (Figs EV2C and EV3). Local resolution was finally calculated using ResMap (Kucukelbir et al, 2014). For the processing of the complete Bs100S, the coordinates of the selected 24,546 particles were carefully re‐inspected in order to remove particles that were within close proximity of another particle, so as not to include particles twice in the final reconstruction; 5,511 particles were identified and removed from class 8, and the rest of particles were windowed out using a larger box size that encompassed two 70S ribosomes (Fig EV2D). The remaining 19,335 particles were then realigned and refined, resulting in a final reconstruction with an average resolution of 6.2 Å (0.143 FSC; Fig EV5A–C).

Molecular modeling, refinement, and validation

The molecular model for the ribosomal proteins and rRNA of the 70S ribosome of the Bs100S was based on the molecular model from the recent cryo‐EM reconstruction of the B. subtilis 70S ribosome (PDB ID 3JW9; Sohmen et al, 2015). The molecular model was fitted as a rigid body into the cryo‐EM density maps using UCSF Chimera (Pettersen et al, 2004). For BsHPF‐NTD domain, a homology model was generated using HHPred (Soding et al, 2005) based on the HPF protein template from E. coli (PDB ID 4V8H; Polikanov et al, 2012; Fig EV1C). Molecular models were fitted and adjusted by using COOT (Emsley & Cowtan, 2004) and refined in Phenix using phenix.real_space_refine (Adams et al, 2010). Model over‐fitting was evaluated through its refinement against one cryo‐EM half map as described previously (Brown et al, 2015). FSC curves were calculated between the resulting model and the half map used for refinement, as well as between the resulting model and the other half map for cross‐validation (Fig EV3E). The final refinement statistics were determined using MolProbity (Chen et al, 2010) and are provided in Table EV1. For BsHPF‐CTD domain, a homology model was generated using HHPred based on the template from C. acetobutylicum (PDB ID 3KA5; Fig EV1D). The molecular model was rigid body fitted using UCSF Chimera (Pettersen et al, 2004).

Figure preparation

Figures showing map densities and atomic models were generated using UCSF Chimera (Pettersen et al, 2004).

Accession numbers

The cryo‐EM map of the Bs70S‐30S subcomplex and the complete Bs100S have been deposited in the EMDB with the accession codes EMD‐3656 and EMD‐3664, respectively. Atomic coordinates have been deposited in the Protein Data Bank with accession code PDB ID 5NJT.

Author contributions

DNW, GB, and KT designed and supervised the study. MA prepared the Bs100S sample for cryo‐EM and performed all sucrose gradient analyses. OB collected the cryo‐EM data. BB, MA, and SA processed the cryo‐EM data. BB, MA, RB, and DNW interpreted the cryo‐EM data. WS cloned and purified the BsHPF protein variants and performed the SEC and SLS. HS generated all BsHPF expression strains and performed Western blotting, growth curves, and ribosome pelleting assays. DNW, GB, and KT wrote the manuscript with comments from all authors.

Conflict of interest

The authors declare that they have no conflict of interest.

Note added in proof

The recent cryo‐EM structure of the Staphylococcus aureus 100S determined by Khusainov et al (2017) reveals that the mechanism of 70S dimerization mediated by the S. aureus long‐form HPF appears to be similar to that observed here for Bacillus subtilis.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

Table EV3

Review Process File

Source Data for Figure 5

The EMBO Journal (2017) 36: 2061–2072

See also: I Khusainov et al (July 2017) and RL Gonzalez Jr (July 2017)