SrmB Rescues Trapped Ribosome Assembly Intermediates

Abstract

RNA helicases play various roles in ribosome biogenesis depending on the ribosome assembly pathway and stress state of the cell. However, it is unclear how most RNA helicases interact with ribosome assembly intermediates or on other cell processes to regulate ribosome assembly. SrmB is a DEAD-box helicase that acts early in the ribosome assembly process, although very little is known about its mechanism of action. Here, we use a combined quantitative mass spectrometry/cryo-electron microscopy approach to detail the protein inventory, rRNA modification state, and structures of 40S ribosomal intermediates that form upon SrmB deletion. We show that the binding site of SrmB is unperturbed by SrmB deletion, but the peptidyl transferase center, the uL7/12 stalk, and 30S contact sites all show severe assembly defects. Taking into account existing data on SrmB function and the experiments presented here, we propose several mechanisms by which SrmB could guide assembling particles from kinetic traps to competent subunits during the 50S ribosome assembly process.

Graphical Abstract

Introduction

The ribosome is a large dynamic ribonucleoprotein (RNP) machine responsible for protein synthesis, a key process for all organisms. Disruption of protein synthesis leads to well-documented defects in cell growth and cell cycle control that are, in turn, translated to various disease states[1]. In Escherichia coli (E. coli), the 70S ribosome is composed of 54 ribosomal proteins (r-proteins) and three ribosomal RNAs (rRNAs) assembled into two subunits: the small (30S) and large (50S) subunits. The 50S large subunit is formed by association of the 23S and 5S rRNAs and 33 proteins (bL1 to bL36) (Fig. 1), and its functions include catalysis of the peptidyl transfer reaction and translocation along mRNA, preventing premature nascent chain hydrolysis, providing the binding site for tRNA and factors that assist in initiation, elongation and termination, and facilitation of protein folding after synthesis[2–5] While the functional states of the translating ribosome have been well characterized, comparatively little of the ribosome assembly process is understood.

Fig. 1.

Anatomy of the 50S subunit (PDB ID 4YBB) including (a) the uL1 stalk (orange), CP (red), and L7/12 stalk (yellow); (b) areas of contact between the 50S and 30S subunits (blue); (c) helices involved in the peptidyl tranferase center (red, individual amino acids involved colored dark red); (d) the SrmB binding site (rRNA in dark green, proteins uL4 and uL24 in green and light green, respectively); and (e) mutations that can ameliorate ΔsrmB (orange) which occur near the region of uL13 (red) and uL25 (dark red) binding.

RNA helicases are an important class of ribosome assembly factors, and the majority of RNA helicases found in E. coli are involved in ribosome biogenesis[6]. In particular, the DEAD-box family of proteins contains RNA helicases that are critically involved in nearly all aspects of cellular RNA metabolism[6]. The DEAD-box helicases are characterized by a highly conserved helicase core, containing at least 12 conserved amino acid motifs that participate in binding RNA and ATP substrates[6–11]. In vitro, DEAD-box proteins typically exhibit an RNA-dependent ATPase activity that is associated with RNA duplex dissociation, helix formation, protein displacement, or RNA secondary and tertiary structure rearrangements [6, 12, 13].

E. coli harbors five DEAD-box RNA helicases: DbpA, RhlB, SrmB, DeaD (CsdA), and RhlE [12, 14, 15]. Of these, SrmB has been implicated in the earliest stages of 50S ribosomal subunit biogenesis. Deletion of SrmB confers a cold-sensitive phenotype and leads to an accumulation of 40S particles corresponding to incompletely assembled 50S subunits[16]. SrmB forms a specific ribonucleoprotein complex in vivo and in vitro with the r-proteins uL4 and uL24 and h11–21 (nt 200–400) near the 5’ end of 23S RNA[17, 18] (Fig. 1d). Mutations in rRNA that suppress the phenotype of ΔsrmB were located far from the SrmB tethering site (Fig. 1e) but close to the 5S rRNA and 1024 G-ribowrench (which is a pseudoknot formed by helix 41 and an internal loop between helix 41 and helix 42), as well as uL13 and the L7/12 stalk. The 1024 G-ribowrench forms many contacts with uL13 and is a phylogenetically conserved pseudoknot[19]. One of the proposed functions of SrmB is to facilitate the folding of the 1024 G-ribowrench and h42 structures in 23S rRNA without a requirement for ATP hydrolysis[18].

The helicase activity of SrmB on noncognate substrates in vitro requires a short stretch of double stranded RNA with a long, single-stranded 5’ or 3’ overhang. It has been proposed that long RNAs may be required to bridge individual SrmB monomers for dimerization[20], although the oligomeric state of functional SrmB acting on ribosomal substrates is unknown. Recently, it has been found that interactions with RraA, a protein that binds to RNase E and regulates its endonucleolytic activity, can stimulate ATPase activity, although the roles of RraA and SrmB together in vivo are unclear[21]. Another potential role for SrmB could be as an rRNA chaperone, since SrmB is reported to have annealing activity[22]. More recent work by Iost and Jain [23] identified ΔsrmB suppressor mutations that map to the 5’-untranslated region of the uL13 and uS9 operon that cause the overexpression of these two proteins and alleviate the cold-sensitive phenotype of ΔsrmB. These finding suggest a previously unknown function of SrmB in regulating the expression levels of r-proteins and thus indirectly mediating ribosome assembly. However, it is difficult to integrate the results from previous work into a mechanistic picture including the precise nature of the ribosomal substrate for SrmB and the conformational changes that it might catalyze through direct interactions with ribosome assembly intermediates. Additionally, differences in strains, growth temperature, and growth medium previously reported in the literature make it hard to pinpoint the SrmB function(s) in assembly (Table S1).

Previously, we explored the effects of limiting an essential protein, bL17, by quantitative mass spectrometry (qMS) to determine r-protein occupancy and cryo-electron microscopy (cryo-EM) to examine structures of intermediates[24]. We have also developed methods to quantitatively measure the rRNA modification state [25]. Here, we us similar approaches to characterize the 40S intermediate present at 37°C in ΔsrmB in the E. coli K-12 BW25113 strain grown in minimal medium. Our decision to use 37°C is based on our de sire to facilitate comparison to the structures of assembly intermediates resolved previously [24], and the choice of a defined medium was driven by facilitation of isotope labeling (¹⁵N and ²H) for quantitative mass spectrometry. Our results suggest a possible mechanism through which SrmB aids ribosome assembly in vivo as an RNA chaperone to resolve misfolded and trapped RNA structures, and by stabilizing early r-protein and rRNA binding events in ribosome assembly.

Results

Multiple ribosome assembly intermediate populations are present in ΔsrmB

ΔsrmB has previously been shown to have a cold-growth defect in the E. coli K12 strain WJW45[16], indicated by the presence of a 40S peak during sucrose gradient ultracentrifugation. This defect is apparent in our E. coli ΔsrmB BW25113 strain grown at 37°C (Fig. S1), as well, although to a lesser extent; this defect at 37°C wa s also observed by Jagessar and Jain [26]. The composition of the 40S peak from ΔsrmB BW25113 was characterized using a qMS protein inventory experiment (Fig. S1, left workflow, and Fig. 2). The normalized protein levels for the control WT 50S particle represent fully assembled 50S particles with stoichiometric ribosomal protein occupancy (Fig. 2, black). On the other hand, three groups of r-proteins with differential occupancy were observed in the ΔsrmB 40S fraction (Fig. 2, pink). The first group, which contains, uL1, uL3, uL4, uL5, bL17, uL18, bL20, bL21, uL23, uL24, and uL29 were all found at levels that were stoichiometric compared to the control 50S particles. Many of these proteins (uL1, uL3, uL4, bL20, bL21, uL23, and uL24) are primary binding proteins that bind directly to the rRNA and nucleate ribosome assembly. The second group, with levels between 50 and 90%, include proteins that are important for the interactions of the 30S and 50S (bL2, uL14, and bL19, Fig. 1b.) The third group is composed of r-proteins that are present in less than 50% of the population. These proteins include many of the late binding proteins[27] (uL6, uL7/12, bL9, uL10, uL16, bL25, bL27, bL28, uL30, uL31, bL35, and bL36) as well as several early binding proteins such as uL13 and intermediary binding protein bL32. Many of these proteins surround the peptidyl transferase center (uL6, bL10, uL16, and uL31, Fig. 1c) and contain protein that are important for the proper docking of the central protuberance (bL25, Fig. 1a). uL22 was not identified in the 40S fraction. These data generally agree well with previous work by Charollais, et al.[16], although there are differences between the two datasets that could be due to differences in the strains, purification protocols, the temperature of the culture, or data collection and analysis procedures. For example, our work here utilizes relative quantification[28], whereas Charollais et al. use label-free quantification[16].

Fig. 2.

Relative quantification of 40S from ΔsrmB (pink) and WT 50S ribosome particles (black). The relative 50S uL20 protein levels were normalized to 1, and remain consistent throughout the experiment. (Pink) Relative levels of proteins in the 40S intermediate. The proteins fall into three groups: less than 0.4 (gray), between 0.4 and 0.92 (medium gray), and greater than 0.92 (dark gray). If no peptides were found, the protein is represented by an x; otherwise, data was taken in triplicate, data from only one peptide are shown as an open circle, and data from more than one peptide are shown as closed circles.

Our qMS data suggest that there are multiple assembly intermediates in the ΔsrmB 40S peak, given the large number of proteins that were found in substoichiometric amounts (Fig. 2 group 2 and 3). Furthermore, the data suggest that the binding site for SrmB, formed by uL24 and uL4, could be intact for all assembly intermediates in the 40S peak, as both are present at nearly stoichiometric levels. Proteins uL5 and uL18 are present at WT levels, but protein uL25 necessary for proper integration of 5S rRNA and docking of the central protuberance is present at substoichiometric levels. This could be explained by the presence of intermediates with a misdocked or misfolded central protuberance. Based on the protein composition data, the proposed SrmB binding site appears to be intact in the intermediates, but the 40S peak has a large amount of compositional heterogeneity that is presumably also reflected in structural heterogeneity.

Altered rRNA modification levels suggest effects of ΔsrmB near the peptidyl transfer site.

E. coli 23S rRNA contains 25 modified nucleosides spread over the functionally important sites of the ribosome, including the peptidyl-transferase center, exit tunnel, and the intersubunit bridges. These include 10 base and 3 ribose methylations, 9 pseudouridines, one hydroxycytidine, one dihydrouridine, and one methylated pseudouridine. Several of these modifications were recognized as important elements contributing to translation and the 50S assembly process[25, 29–31].

We previously determined the relative order of 23S modifications in MRE600 E. coli cells using a qMS analysis protocol[25]. By applying the same qMS workflow to the 40S samples from ΔsrmB (Fig. S1), we identified and quantified 22 out of 25 known 23S modifications relative to the 70S from WT BW25113 (Fig. 3). Based on these data there are three distinct groups of modifications, introduced consecutively during 50S assembly in WT cells (only two groups were resolved previously) (Fig. 3a and Fig. S2 for biological replicates). However, there are two specific modification events in the ΔsrmB dataset where the order of events is changed without globally perturbing the whole pathway (Fig. 3b). When SrmB is present, Cm(2498) and ho5C(2501) are introduced during intermediate stages of the 50S assembly process. In ΔsrmB, hydroxylation at 2501 by YdcPis largely delayed, and the ribose methylation at 2498 by RlmM is accelerated with respect to other modifications in the wild type.

Fig. 3.

Quantitative MS reveals the relative order of individual 23S rRNA modification events. (a) Sucrose density gradient profile of WT. Fractions are denoted by vertical lines. Relative rRNA modification levels were calculated across the sucrose gradient and normalized to the levels of unmodified 23S in each fraction. Modifications are classified as early (red), intermediate (green), or late (cyan). (b) Sucrose density gradient profile of ΔsrmB. Cm(2498) and ho5C(2501), which differ in order between WT and ΔsrmB strains, are labeled for each plot shown. (c) Relative position of C2498 and C2501 within the crystal structure of the bacterial ribosome. r-proteins are shown in grey. Helix 89 of the 23S rRNA, which contains C2498 (red spheres) and C2501(cyan spheres), is shown in yellow. Helices 41–44 of the 23S rRNA, which contain the sites of mutations that stabilize the growth defect at cold temperatures in ΔsrmB strain (A1039 and A1027) are in green. C27 is the site of another stabilizing mutation in the 5S rRNA (grey). Structure used was PDB ID 4YBB.

The decoupling between modifications at two closely spaced residues in the peptidyl transferase center (at the base of h89, Fig 3c) of the 50S is surprising. We speculate that SrmB is required for proper timing of Cm(2498) and ho5C(2501) modifications and possibly plays an active role in displacing RlmM, which in turn allows 2501 to be hydroxylated. Furthermore, close examination of our raw LC-MS data suggest presence of both 2501-modified and 2501-unmodified RNase U2 fragments (Fig. S3). In the mature 70S, this residue is known to be partially modified depending on the strain and growth conditions [32, 33]. The levels of ho5C(2501) were estimated to be 85% in the wild type cells and 60–70% in ΔsrmB ribosomes. This reduction might be a direct consequence of the delay in the 2501 modification step when SrmB is unavailable. In summary, our findings using ΔsrmB suggest specific alterations in the order of Cm(2498) and ho5C(2501) modifications and possibly in the assembly of the peptidyl transferase center.

Cryo-EM of ΔsrmB 40S reveals three assembly intermediates

Cryo-EM analysis of the particles in the 40S peak were performed using the same general approach that was used for bL17 limitation strain as previously described [24]. Three distinct classes were identified by 3D classification, at resolutions of 4.7–5.8Å (Fig S4.), from the combined 40S fractions of the sucrose gradient ultracentrifugation experiments (Fig. 4). The three classes are arranged in order of increasing maturity and represent 36% (A class, Fig. 4a), 39% (B class, Fig. 4b), and 25% (C class, Fig. 4c) of the classified particles, respectively. While these resolutions are insufficient to confidently build models into the density, we did not use additional processing and refinement strategies in order to increase resolution. The cryo-EM density maps presented in Fig 4. have been filtered to 10Å in order to facilitate data analysis, as we are interested in exploring broad structural features at the level of protein occupancies rather than small conformational changes or subtle differences in individual amino acids or rRNA bases. The low-pass filtered maps are accordingly used for subsequent analyses. The unfiltered maps with the corresponding FSC curves are presented in Fig. S4 and were used to confirm findings from the filtered dataset (see Table S2 for statistics). To broadly compare the maps, an occupancy matrix was calculated based on quantitative comparison of the experimental map to a reference map (PDB ID 4YBB), which has been segmented into a standard set of rRNA helices and r-protein densities, as previously described[24]. By clustering the protein and rRNA helix occupancy, we observed five major structural blocks that assemble cooperatively (Fig. 4d).

Fig. 4.

Analysis of the three 40S intermediate cryoEM structure. All cryoEM classes have been filtered to 10Å and are shown with a soft mask. (a) The least mature class: A class. (b) The intermediate class, B class. (c) The most mature class observed in the data, the C class. Blocks of structural elements in (a-c) are colored according to the occupancy analysis in d. (d) Occupancy analysis of the r-proteins and rRNA helices as referenced to PDB ID 4YBB. The largest Block 1 (Fig. S5) is not shown here, and has 100% occupancy across all structures. (e) Structural elements of the rRNA helices defined by the occupancy analysis mapped on to the 2D 23S rRNA domain structure. Colors correspond to the blocks in (d). Block 1 is shown in red. rRNA domain contacts that span across domains are shown in solid lines.

For the most part, the protein abundances observed in the qMS data (Fig. 2, pink) match well to the protein occupancies observed in the cryo-EM dataset (Fig. 4d). Differences between the qMS data and the cryo-EM occupancy analysis can be caused by several reasons. First, r-proteins that are present in the qMS data may not appear in the cryo-EM density because they are bound to flexible structural elements that are not resolved by cryo-EM. Conversely, proteins present in the cryo-EM maps that have low abundance by qMS can be explained by non-native structures, such as misfolded rRNAs, misdocked r-protein binding, or unidentified cellular factors bound., In fact, the reference-based nature of the cryo-EM occupancy analysis is blind to non-native features. Furthermore, qMS is a bulk measurement of the average of protein occupancy of many different intermediates, while the cryo-EM occupancies are calculated for the individual solved intermediate structures. Therefore, proteins that are substoichiometric by the qMS analysis can be found in one or more of the observed cryo-EM intermediates. Finally, some proteins, like bL31, uL1, and uL7/12, are not included in the 4YBB crystal structure or, like bL9, are in different conformations due to the crystallization process. Regardless, the cryo-EM occupancy analysis provides a quantitative means by which to interpret structural similarities and differences of the ribosome assembly states For example, by mapping the blocks revealed by the occupancy analysis to the 2D- structure of the 23S (Fig. 4e), we observe that the folding blocks map to multiple rRNA domains, that are connected by tertiary contacts. We have previously observed this “block” behavior in our work under the r-protein bL17 limitation[24].

The first block of assembly (Fig. 4e and Fig. S5) corresponds to proteins and rRNA helices that have >90% density observed in all three cryo-EM maps, which includes the SrmB binding site (Fig. 5, row 1). This block also supports a fully-formed peptide exit tunnel and contains most of the rRNA helices from Domains I and III. Block 2 (Fig. 4d,e and Fig. S5, yellow) represents proteins and rRNA helices that are present in the B and C classes, but are not fully occupied in the A class. bL20, bL21, uL15 and many of the rRNA helices in Domain II belong to this class, thus indicating that some of the earliest r-proteins and helices necessary for stabilizing the 5S rRNA are not stable in the A class. In block 3 (Fig. 4d,e and Fig. S5, green) are proteins and helices that are almost fully occupied in the A and C classes, but not fully occupied in the B class, which include rRNA elements of Domain V, bL28, and bL35. Closer inspection of the A class reveals that much of this density is due to misdocked rRNA or r-protein structures that do not conform to the density that would be expected for the native elements (according to PDB ID 4YBB), confirming that the A class is less mature than the B class.

Fig. 5.

Specific areas pertinent to SrmB binding in reference to the theoretical cryoEM density of 4YBB. From left to right: the A class, B Class, C class, and the theoretical 4YBB EM representations, and the 4YBB model (right column) are represented in gray. The original, non-filtered cryoEM densities are used for visualization. Color scheme for elements in each row is at the end of the row and in the first column. (Row 1) The SrmB binding site. L4 is colored in olive green, L24 is colored in lime green, and nt200–400 are colored in sea green. (Row 2) Sites of known SrmB deletion suppressor mutations. h42 is colored in light blue, h41 is in cyan and the region of the 5S is colored in light steel blue. (Row 3) Areas of known disruption in ΔsrmB. uL13 colored in red, bL25 in dark red, h42 is dark orange, and the dark red arrow points to density blocking uL13 binding. (Row 4) Areas of non-native density that could be extensions of local rRNA helices. h41 is in purple, bL20 is in magenta, bL21 is in dark magenta, and h97 is in pink.

Block 3 also contains occupancy at the position of uL13, which is present at a low stoichiometry in the qMS data. Upon closer examination of this region in the maps, we observe that there is a “bridge” of non-native rRNA density spanning from the base of the uL7/12 stalk to the uL13 binding site in the B class and the C class (Fig. S6) and non-protein density in the A class (Fig. 5, row 2). This bridge partially occludes the uL13 binding site, thus accounting for the seemingly anomalous density observed in the cryo-EM occupancy analysis. Block 4 (Fig. 4d,e and Fig. S5, blue) represents proteins and rRNA helices that do not have any occupancies in the cryo-EM maps. These proteins and rRNA helices involve late formation of the peptidyl transfer center (h89, h91) and important sites of interactions between the 30S and 50S subunits, such as (h68, h69, h71) uL16 and h42. Others are important for the formation of the uL7/12 stalk (uL10, uL11, h42–44),. Block 5 (Fig. 4d,e and Fig. S5, purple) represents the central protuberance proteins (L5,L18, L25) 5S rRNA, and helixes 38, 83–85 which directly contact 5S RNA, which is only stable enough for density to be observed in the C class. Although we only observe density for the central protuberance in the C class, qMS data suggests that the central protuberance is at least partially formed, but flexible and misdocked, as many of the r-proteins necessary for stable central protuberance docking are observed in the qMS data but the density does not average into a structure in the cryo-EM data.

This analysis, combined with qMS data, reveals that the 40S peak is conformationally and compositionally heterogenous. The three structurally distinct can be naturally organized into a putative linear assembly pathway of A->B->C->50S, proceeding from the least to the most mature intermediate, in the absence of SrmB. The SrmB-B class and SrmB-C class are broadly similar to the major L17-C classes and L17-E classes from the bL17-limitation strain data, implying that these structures are likely common on-pathway intermediates for further maturation, whereas the presence of the SrmB-A class is unique to SrmB deletion. Although the peptidyl exit site seems to be well-formed for all three classes, there are severe defects in the peptidyl transferase center, the central protuberance, and intersubunit bridge contacts. Interestingly, many of structural perturbations center along the sites of misdocked or missing density for rRNAs that suppress SrmB deletion (Fig. 5, row 3). Furthermore, there is also a “bridge” of non-native density between h97 and h41 that could be due to non-native RNA tertiary interactions or could be an extension of local rRNA helices (Fig. 5, row 4).

Discussion

The SrmB binding site is formed in the absence of SrmB

The SrmB binding site, which is composed of uL4, uL24, and nt 200–400 (h11–21) [17] is intact in all three cryo-EM structures (Fig 5. row 1) and uL4 and uL24 are at stoichiometric levels in the protein qMS data, which indicates that the SrmB binding site has no clear deficiencies during the assembly process at 37°C. Thus, the earl iest nucleation stages of ribosome assembly are able to proceed without the intervention of SrmB. Our observations are fully consistent with previous studies by the Dreyfus group [14, 16–18, 20] and suggest that SrmB acts distally to its binding site by acting as a chaperone for proper rRNA docking (e.g. h42) and r-protein binding events (e.g. uL13 and uL25, Fig 5. row 2, Fig. 6b) or by providing rigidity to the nascent assembling ribosome so that later steps of assembly can occur on a solid base (Fig. 6c).

Fig. 6.

Potential pathways for SrmB to act in ribosome assembly. (a) Sequence of the C-terminal tail of SrmB. The blue box indicates predicted α-helix content, otherwise, the rest of the C-terminus is predicted to be uncoiled. Red stars indicate positively charged amino acids that could participate in RNA binding. The predicted minimum length of an α-helix, followed by random coils, has a minimum end-to-end length of 112Å, which is long enough to span the front of the ribosome from the uL1 stalk to the uL7/12 stalk. (b) SrmB could regulate the expression of uL13, which causes the ribosome assembly pathway to stall under cold-stress conditions (c) SrmB could stabilize the uL4:uL24:nt 200–400 interactions, thus providing a stable base that could prevent misfolding in further rRNA folding events. In this case, it is only the binding of SrmB that stabilizes the rRNA, and the C-terminal tail is not needed for further structural rearrangements. (d) The C-terminal tail of SrmB could act as a folding chaperone to ensure proper central protuberance docking and rRNA folding around the uL13 binding site after SrmB docks into its binding site. (e) SrmB could also act as a dock for other proteins, either by means of its C-terminal tail or by some other protein: protein interactions. In this case, SrmB would guide the rRNA into the proper conformations for docking and release of other proteins, such as methyltransferases (such as RlmM) or hydroxylases.

Structural defects near ΔsrmB suppressor sites, at peptidyl transfer center and the uL7/12 stalk are present in ΔsrmB

Density is missing for uL13 and helix 42 in all three of our cryo-EM maps, and bL25 is only present in the C class (Fig. 5, row 2). These structural elements are located near the proposed site of action for SrmB (Fig. 5, row 3). uL13 directly interacts with the 1024 G-ribowrench and bL25 interacts with h42. Given that uL13 and bL25 are at levels three times lower than their corresponding levels in WT 50S ribosomes in the qMS data, it likely that these proteins are not present in the ribosome assembly intermediates, and we observe non-native densities that would not allow for proper protein docking. These non-native densities form a “bridge” that could be an extension or alternative conformation of h41 or h97 (Fig. 5 row 4). h41, which participates in 1024 G-ribowrench formation, is slightly misdocked, although density is clearly observed for bL20 and bL21, both of which contact the outside of this helix. h97, while maintaining at least 50% occupancy in all three classes, is not well-resolved at the area where the “bridging” density occurs. Furthermore, the r-proteins and rRNAs that support the uL7/12 stalk are missing in all three structures (L11, L10, h42- h44, Fig 4d–e), indicating that SrmB plays a key role in its formation.

The changes in the order of Cm(2498) and ho5C(2501) modifications suggest alternative conformations of the rRNA in the peptidyl transferase center region, which alter the availability of rRNA for modification. These modifications occur at the base of h89, which is not resolved in any of the cryo-EM data, although the C class exhibits extra density at the base of h89. These data suggest that there are severe defects at the peptidyl transferase center. Defects in peptidyl transferase center formation were previously observed in 50S assembly intermediates isolated with other strains, including bL17-lim [24, 34, 35] and are thought to ensure that these intermediate particles, if mistakenly integrated into a 70S ribosome, are unable to translate proteins.

SrmB alleviates kinetic traps for ribosome assembly

Our data qMS and cryo-EM data reveal that, in the absence of SrmB, there are multiple sites of defects in ribosome assembly: the peptidyl transferase center (Fig. 1b), the central protuberance (Fig. 1a), and areas of contact between the 30S and 50S subunits (Fig. 1c), and L7/L12 stalk. Although SrmB deletion is less severe for ribosome assembly at 37°C, compared to cold temperatures (<30°C), ribosome assembly pro ceeds with the accumulation of intermediates (classes A-C). Overall, our data suggest that it is likely that one of the functions of SrmB is to act against rRNA falling into a misfolded kinetic trap (Figure 6).

Given the data presented here, along with the current literature, it is likely that SrmB has multiple roles depending on the stress conditions of the cell. First, given the recent work by Iost and Jain [23], SrmB acts in a mechanism to regulate uL13 production, and the 40S particles thus represent a “uL13-limited” assembly pathway(Fig. 6b). However, our whole cell proteomics analysis (Fig. S7) revealed that there is no significant depletion (~10% reduction with respect to other r-proteins in ΔsrmB) of uL13 at either 37°C or 18°C, and so it i s unlikely that a “uL13-limited” assembly pathway is represented in our data. Second, given that in vitro, SrmB can facilitate structural rearrangements without ATP hydrolysis[18, 22], we hypothesize that the chaperone function of SrmB are through its C-terminal tail (Fig. 6a) which facilitates correct docking ofthe central protuberance and the peptidyl transferase center. C-terminal tail may also assist, uL13, bL25, and h42, and they would be able to form their proper contacts and dock the rest of the rRNA and r-proteins into a fully-functional 50S (Fig. 6c). However, given that the C-terminal tail may not be necessary for ribosome assembly at 30°C[17] (at least in the context of SrmB overexpressed from a plasmid), it is possible that SrmB has a different mechanism of action under cold-stress conditions. Alternatively, SrmB could act to stabilize the nascent interactions of uL24, uL4, and nt200–400. The rigidity provided by these initial interactions could prevent non-productive rRNA structures from forming, thus ensuring the success of ribosome assembly, even at colder temperatures, where rRNA could become trapped in unfavorable intermediary states (Fig. 6d). Third, SrmB could act as a nexus for other proteins, such as RlmM and the newly discovered YdcP responsible for ho5C2501 [36] to bind in the correct order, thus ensuring that ribosome rRNA modifications occur at the proper times (Fig. 6e). The evidence from our qMS, rRNA modification MS, and cryo-EM data from ΔsrmB E. coli grown at physiological temperatures lead to the conclusion that SrmB anchors to the uL24, uL4, and nt 200–400 region in order to perform chaperone activities through physical interactions with ribosome assembly intermediates. However, it seems likely that SrmB plays multiple roles in ribosome assembly depending on the stress conditions of the cell.

Materials and Methods

Bacterial Strains and plasmids

WT BW25113 E. coli and the ΔSrmB BW25113 E. coli strain from the Keio Knockout Collection, were obtained from the E. coli Genetic Stock Center [37].

Cell growth

ΔsrmB cells were grown at 37°C in M9 glucose minimal medium supplemented with trace metals in the presence of either 1 g/L of ¹⁴N ammonium sulfate (EM, RNA modification and whole cell proteomics analysis) or using a mixture of 0.5 g/L ¹⁴N ammonium sulfate and 0.5 g/L ¹⁵N ammonium sulfate (QMS protein inventory). This was the sole nitrogen sources in order to facilitate isotope labeling and relative quantitation of ribosomal proteins and RNA modifications’ ΔsrmB BW25113 cultures were grown to OD₆₀₀ 0.5–0.6, were quenched on an equal volume of ice and were harvested by centrifugation at 5000 rpm for 15 min. Cell pellets were either immediately lysed or stored at −80 °C prior to lysis.

Sucrose Gradient Purification of Ribosomal Particles

Frozen cell pellets for ΔsrmB cultures were then thawed and resuspended in 20 mM Tris–HCl, pH 7.5, 100 mM NH₄Cl, 10 mM MgCl₂, 0.5 mM ethylenediaminetetraacetic acid (EDTA), 0.5 mM dithiothreitol [DTT], 1 mM phenylmethylsulfonylfluoride (PMSF), and 20 U/ml DNase I (Sigma), and EDTA-free protease inhibitor cocktail (Roche Applied Science). Cells were lysed in a bead beater (BioSpec Products, Inc., Bartlesville, OK) using 0.1-mm zirconia/silica beads (3×40 second pulses with 2 minutes on ice in between). Insoluble debris was removed by two centrifugation steps: a low-speed spin at 6000 rpm for 10 min and then a high-speed spin centrifugation step at 16,000 rpm (31,000g) for 40 min.

The clarified supernatant was loaded onto a 13–51% (w/v) or 10–40% (w/v) non-dissociating linear sucrose gradient (50 mM Tris–HCl, pH 7.8, 10 mM MgCl₂, and 100 mM NH₄Cl) and centrifuged in a Beckman SW32 rotor at 26000 rpm for 18 or 16 hours at 4°C. For protein inventory QMSanalysis, approximately 84 fractions were collected from each sucrose gradient using a Brandel gradient fractionator. Based on the UV 254 nm trace, gradient fractions corresponding to the 40S ribosome peak were pooled together. For EM analysis, the combined fraction was diluted to 3X volume in gradient buffer (50 mM Tris–HCl, pH 7.8, 10 mM MgCl₂, and 100 mM NH₄Cl) and buffer exchange 3 times using a 3 kDa cutoff concentrator (Amicon) to remove most of the sucrose The concentrated to 50 μl sample was used to estimate RNA concentration using OD260.

ESI-TOF Mass Spectrometry

Each experiment was performed in triplicate. The data show a representative replicate (Fig. 2). For protein inventory experiments, 50 pmol of both ¹⁵N WT-70S and ¹⁴N WT-70S were added to 50 pmol of 40S subunit purified via sucrose gradient centrifugation. 70S from WT cells were purified using a sucrose cushion. Samples were incubated on ice for 12 h in 13% trichloroacetic acid (TCA). The protein precipitate was pelleted by centrifugation at 13,000g for 30 min at 4 °C. The supernatant was removed, and th e pellets were rinsed first with 10% TCA and then with ice- cold acetone, dried in a Speed-Vac concentrator, and then resuspended in 40 μL of 100 mM ammonium bicarbonate (pH 8.5) in 5% acetonitrile (ACN). A 4-μL aliquot of 50 mM DTT was added, and the samples were incubated at 65 °C for 10 min. Cysteine residues were modified by the addition of 4 μL of 100 mM iodoacetamide followed by incubation at 30 °C for 30 min in the dark. Proteolytic digestion of the proteins was carried out by the addition of 4 μL of 0.1 μg/mL sequencing grade porcine trypsin (Promega, Co., Madison, WI) with incubation overnight at 37 °C. Undigested proteins were precip itated by adding 1/3 volume of 20% ACNin 2% trifluoroacetic acid and removed by centrifugation. The supernatant was loaded to a PepClean C18 spin column (Thermo Fisher Scientific Inc., Rockford, IL) to remove salts and concentrate the samples. The eluant was dried in a Speed-Vac concentrator and the peptides were re-dissolved in 10 μL of 5% ACN in 0.1% formic acid. An 8 μL of aliquot was used for the electrospray ionization time of flight (ESI-TOF) analysis.

The peptide samples were analyzed on an Agilent 1100 Series HPLC instrument coupled to an Agilent ESI-TOF instrument with capillary flow electrospray (Agilent Technologies Inc., Santa Clara, CA). The digested ribosomal proteins were injected using an autosampler onto an Agilent Zorbax SB C18 150mm × 0.5mm HPLC column. The mobile phases used were buffer A (H2O, 0.1% formic acid) and buffer B (acetonitrile, 0.1% formic acid). Peptides were separated at a flow rate of 7 μL/min using the linear gradient(step 1: 5–15% buffer B over 10 min; step 2: 15–50% buffer B over 70 min, and step 3: 50–95% buffer B over 4 min). Data were collected using positive polarity over the m/z range of 300–1300.

Quantitative mass spectrometry data was processed and analyzed as previously described[38–41]

Whole cell Proteomics

Whole cell relative abundances of the r-proteins were obtained by culturing ΔsrmB and WT cells in M9 at 37°C or at 18°C. ¹⁴N- and ¹⁵N-ammonium sulfate were used to label ΔsrmB and WT cells respectively. ΔsrmB and WT cells were mixed at ~ 1:1 ratio based on the OD measurements, lysed overnight in 20% TCA at 4°C, an d the precipitated proteins digested with trypsin as described above. LC-MS/MS data were collected on a Sciex 5600+ Triple TOF coupled to the Eksigent expert nanoLC [24]. ¹⁴N/¹⁵N ratios for each ribosomal peptide were normalized to the medium over ~2500 E. coli peptides identified and quantified in each sample (Fig. S7).

Analysis of 23S RNA modifications.

MS-based quantitative measurements of 23S rRNA modifications were done as described in the prior work [25]. Wild-type and ΔsrmB rRNAs were metabolically labeled with ¹⁴N- or ¹⁵N-ammonium sulfate in the presence of 5,6-D-uracil (Cambridge Isotope Laboratories) enabling identification of pseudouridines in both the sample and the reference. Abundances of individual modifications were assessed using nucleolytic fragments detected using RNase T1, A, or U2 cleavages, relative to their abundances in the 23S standard purified from the mature WT ribosomes:

$$\text{Relative}\text{modification}\text{level}\text{=}\frac{\text{Peak}\text{intensity}\text{of}\mathrm{\Delta }\mathit{\text{srmB}}}{\text{Peak}\text{intensity}\text{of}\mathrm{\Delta }\mathit{\text{srmB}}+\text{Peak}\text{intensity}\text{of}\text{WT}}$$

where the ΔsrmB and WT are differentially labeled with ¹⁴N and ¹⁵N isotopes.

Two additional U2/T1 digestion fragments were added to the previously reported list of 23S specific oligonucleotides[25]: 2604-ΨΨCG-2607, used to monitor Ψ(2604) and Ψ(2605), and chemically identical 2456-CΨG-2458 and 2579-CΨG-2581, reporting on the presence of both Ψ(2457) and Ψ(2580).

Electron Microscopy Sample Preparation and Data Acquisition

Gold grids were a generous gift from Dr. Bridget Carragher and were prepared as in [42]. Sample was diluted to ~225 nM in gradient buffer and applied to a plasma cleaned (6s, Gatan Solarus) gold grid in humidified CP3 chamber (FEI). Sample was blotted automatically for 2.5s and plunged into liquid ethane, then stored in liquid nitrogen until imaged.

Data were collected on an Thermo Fisher Scientific (formerly FEI) Titan Krios electron microscope operating at 300 kV equipped with a Gatan K2 Summit detector using the Leginon software [43] with an estimated underfocus ranging from 1.0 μm to 3.5 μm (distributed in an approximately Gaussian manner). The total dose was 45 e-/ Ų, fractionated over 50 raw frames collected over a 10 second exposure time (200 ms per frame), with each frame receiving a dose of ~7 e-/ Ų. 1958 movies were recorded at a calibrated magnification for the position of the detector of 38,167 (nominal magnification of 22,500), corresponding to a pixel size 1.31 Å. To overcome problems of preferred orientation on the grid, particles were imaged at different tilt angles (0, 10, 20, 30, 40, 50 degrees) to obtain different views [44].

Image Processing

All pre-processing was performed within the Appion pipeline[45] and individual programs used within the pipeline are cited below. Frames were aligned using MotionCor2[46], and then used for processing. All micrographs were manually masked using the masking tool to remove regions corresponding to the gold grid bars and large aggregates. The contrast transfer function (CTF) for all micrographs was estimated using CTFFind3 and CTFTilt[47]. 284,420 particles were selected using low-pass filtered 50S ribosomal subunit templates from the micrographs using the FindEM package [48]. A phase-flipped, contrast-inverted, 2x-binned stack was created from these picks with a box size of 128 and pixel size 2.62 Å and was used until the final 3D refinements. Reference-free 2D alignment of this stack was accomplished using ML2D [49] followed by RELION [50]. The 2D classes were visually inspected, and any classes that were clearly 30S, 70S, or other known cellular structures were removed. The remaining 199,381 particles were sorted using projection-matching into one of seven 3D maps (30S, 70S, and five different 50S assembly intermediates obtained during pre-processing). The 129,271 particles that were matched to the 50S assembly intermediates were then subjected to 3D classification in RELION using 5 classes [50] and a 50S map filtered to 60 Å as an initial model. After removing classes that either did not produce an interpretable map or clearly belonged to a 30S or 70S particle (and was obscured in the prior, coarse-grained classification), 60,488 particles remained and resulted in three broadly different maps comprising three classes described herein (Fig. 4). The angles and class occupancies were refined within Frealign[51]. Final maps were sharpened using cisTEM [52] by flattening the amplitude spectrum between 10Å and the resolution of the individual maps.

Occupancy Matrix Analysis

The relative occupancy of ribosomal proteins and rRNA helices n ΔsrmB classes was calculated using a combination of 3^rd party programs and in-house scripts. The experimental maps were prepared for quantitative comparison to a reference map generated from PDB entry 4YBB. Briefly, the unsharpened, unmasked maps from Frealign were filtered to 10Å, and the standard deviation (σ) of voxel values for each map was calculated using relion_image_handler, as an estimate of the noise in the maps. The average σ for the three final maps was 2.25±0.22. The real-valued voxels for the maps were binarized to 0 or 1 using the relion_image_handler command, with the threshold for binarization set to 3*σ for each individual map without any soft edges. The reference map from the E. coli 50S subunit crystal structure (PDB ID 4YBB) was segmented into 140** elements comprised of individual ribosomal proteins and of rRNA helices according to 23S secondary structure [53]. Theoretical density of 4YBB was then calculated for each element at 10 Å using the pdb2mrc command from EMAN[54]. The density for each voxel was then binarized as 0 or 1 using a threshold value of 0.016, which is the threshold that gave approximately correct molecular weight values for individual r-proteins and rRNA helices. Finally, the relative volumes in the binarized experimental and reference maps were calculated for each of the 140 reference elements, resulting in a fractional occupancy between 0 and 1 for each element. Observed occupancy values were then clustered across rows (classes) and columns (rRNA/protein elements) using unsupervised hierarchical clustering, with a squared Euclidean distance metric and Ward’s linkage method, implemented in Mathematica[55].

Supplementary Material

1

Fig. S1. Combined efforts of qMS and cryo-EM to determine how SrmB acts in 50S ribosome assembly. The protein levels report on the total amount of each protein relative to an internal standard compared to an intact ribosome as an external standard (spike). Similarly, the qMS levels report the relative amount and identity of rRNA modification levels relative to 23S standard. CryoEM from selected fractions is then used to explore the structures of the 40S ribosome assembly intermediates.

2

Fig. S2. Several biological replicate measurements were obtained using fractions spanning the entire pre-50S : 50S : 70S regions of a gradient (Ex. 1 and 2 in WT and Ex.1 in ΔsrmB) or across the pre-50S : 50S region (Ex. 3 and 4 in WT and Ex. 2 and 3 in ΔsrmB), where the largest changes in modification occupancy were found. The resulting data were merged for hierarchical clustering analysis performed using Euclidian distance metric and average linkages. Based on the dendograms, 23S modification were divided into three groups of early (red), intermediate (green), and late (cyan) events during ribosome assembly in wild type and ΔsrmB cells.

3

Fig S3. C(2501) is partially modified in 23S of WT and ΔsrmB. Cm(2498) was monitored via RNase A fragment 2497-A(Cm)C-2499. RNase U2 fragments 2498-(Cm)CU (ho5C)G-2502 and 2498-(Cm)CU(C)G-2502 were used to profile ho5C (2501) and C(2501), known to be partially modified. While (Cm)CU (ho5C)G-2502 is dominant, some quantities of 2498-(Cm)CU(C)G-2502 are resolvable in both WT and ΔsrmB, enabling us to roughly estimate fraction of 2501 modified in 70S ribosomes. (a) Representative spectra and their least-squares fits reporting on the presence of Cm(2498), ho5C(2501), and C(2501) in sample 80 from ΔsrmB. (b) 14N/15N ratios were calculated for each of the four samples from 50S-70S region, with ΔsrmB particles being 14N labeled and 23S-WT being 15N labeled. After normalizing to the amount of unmodified 23S present in the sample and the reference (e.g., 14N/15N = 1.38, sample 80), we found that Cm(2498) is stoichiometrically present (14N/15N = 1.45 vs. 1.38), however unmodified 2501 is 2.5 more abundant in ΔsrmB, and ho5C(2501) is slightly substoichiometric. Furthermore, we made two assumptions: first is that 2498 is 100% modified in both ΔsrmB and WT; second that ho5C(2501) and C(2501) fractions add up to 1. Using these assumptions, fraction C(2501) hydroxylated were calculated for each of the four samples, and are shown together with average and sd. The analysis suggests, that 50S and 70S particles in ΔsrmB have larger quantities of unmodified C(2501) than WT 70S.

4

Fig. S4. Unfiltered CryoEM maps. (a-c) Shown from left to right: front view, back view rotated 180°, Euler angle plot, and plots of global half-ma p 3DFSCs, three 3DFSC isosurfaces at a cutoff of 0.5 in three axial orientations describe the isotropy (directional resolution) of the refined maps. (a) SrmB-A class. (b) B class. (c) C class. (d) FSC curves with the colors of the curves corresponding to (A-C). The resolutions of the classes are 5.7Å (Class A), 4.9Å (Class B), and 4.7Å (Class C). (e) Representative micrograph. (f) cryo-EM data analysis workflow.

5

Fig. S5. Assembly blocks from Fig. 4 shown on PDB ID 4YBB. (a) The full complement of assembly blocks are shown on the top, followed by Block 5 (purple), Block 4 (blue), Block 3 (green), Block 2 (yellow), and Block 1 (red, not shown in Fig. 4). (b) Occupancy matrix for Group 1.

6

Fig. S6. (a) B class with non-native “bridge” highlighted in purple. (b) C class with non-native “bridge” highlighted in purple.

7

Fig. S7. Relative levels of the SSU and LSU r-protein in ΔsrmB vs WT measured using whole cell proteomics. Red arrows indicate levels of S9 (a,b) and uL13 (c,d) that were found to be significantly depleted in (Iost and Jain 2019), but they are not found to be significantly depleted here. (a) Ratio of ΔsrmB/WT SSU proteins at 37°C. (b) Ratio of ΔsrmB/WT SSU proteins at 18°C. (c) Ratio of ΔsrmB/WT LSU proteins at 37°C. (d) Ratio of ΔsrmB/WT LSU proteins at 18°C.

Highlights.

• SrmB deletion causes compositionally and conformationally diverse intermediates

• Non-native density observed in the uL13 binding site and suppressor mutation sites

• Defects in central protuberance, uL7/12 stalk, peptidyltransferase center

• SrmB deletion causes reordering of rRNA modifications

• SrmB might reroute stalled complexes with structural defects