Roles of the leader-trailer helix and antitermination complex in biogenesis of the 30S ribosomal subunit

Abstract Ribosome biogenesis occurs co-transcriptionally and entails rRNA folding, ribosomal protein binding, rRNA processing, and rRNA modification. In most bacteria, the 16S, 23S and 5S rRNAs are co-transcribed, often with one or more tRNAs. Transcription involves a modified RNA polymerase, called the antitermination complex, which forms in response to cis-acting elements (boxB, boxA and boxC) in the nascent pre-rRNA. Sequences flanking the rRNAs are complementary and form long helices known as leader-trailer helices. Here, we employed an orthogonal translation system to interrogate the functional roles of these RNA elements in 30S subunit biogenesis in Escherichia coli. Mutations that disrupt the leader-trailer helix caused complete loss of translation activity, indicating that this helix is absolutely essential for active subunit formation in the cell. Mutations of boxA also reduced translation activity, but by only 2- to 3-fold, suggesting a smaller role for the antitermination complex. Similarly modest drops in activity were seen upon deletion of either or both of two leader helices, termed here hA and hB. Interestingly, subunits formed in the absence of these leader features exhibited defects in translational fidelity. These data suggest that the antitermination complex and precursor RNA elements help to ensure quality control during ribosome biogenesis.


INTRODUCTION
The ribosome is a two-subunit enzyme responsible for protein synthesis. In Escherichia coli, the large subunit consists of the 23S rRNA [2904 nucleotides (nt)], 5S rRNA (120 nt) and 34 ribosomal (r) proteins, while the small subunit consists of the 16S rRNA (1542 nt) and 21 r proteins. Ribosome assembly occurs co-transcriptionally and entails folding of the rRNA, binding of r proteins, modification of nucleotides and amino acid side chains, and processing of the rRNA by various RNases. Active ribosomal subunits can be reconstituted in vitro by mixing their constituent components (1,2), suggesting that these molecules themselves store all the necessary information for assembly. While subunits can be reconstituted in vitro, the process occurs slowly and requires non-physiological conditions. In the cell, non-ribosomal proteins known as assembly factors (AFs) facilitate the process (3)(4)(5)(6). There are multiple types of AFs, including ribonucleoprotein-binding proteins, heli-cases, chaperones, RNases, modification enzymes and GT-Pases. AFs are believed to facilitate ribosome assembly by preventing kinetic traps that would otherwise inhibit the process.
In most bacteria, the 16S, 23S and 5S rRNAs are cotranscribed along with one or more tRNAs (7). In Escherichia coli, there are seven of these operons, varying mainly with respect to the tRNA gene content (8). Flanking each rRNA are complementary sequences that form extended helices, known as leader-trailer helices (9). RNase III recognizes and cleaves the leader-trailer helices, releasing secondary precursors (pre-16S rRNA, pre-23S rRNA and pre-5S rRNA) from the primary transcript (10,11). Other RNases subsequently work to trim both ends of the precursor rRNAs, resulting in the mature length rRNAs (12). Processing and assembly are intimately intertwined, as reflected by the fact that precursor rRNAs are present in late-stage assembly intermediates (13). Moreover, immature particles containing precursor rRNAs accumulate when individual AFs are absent or when particular processing events are blocked (14,15).
A similar mechanism of antitermination is operational during transcription of the ribosomal RNA (rrn) operons. In this case, modification of RNAP involves NusA, NusB, NusE(S10), NusG, SuhB, and S4, and the resulting antitermination complex is termed rrnTAC (32)(33)(34)(35)(36). One set of cisacting elements--boxB, boxA, and boxC (5 to 3 )--lies at the beginning of the operon, just after promoter P2, and another lies upstream of the 23S rRNA gene (22,37). Of the RNA motifs, boxA is most important, being both necessary and sufficient for antitermination factor recruitment and rrnTAC formation (38). The presence of boxB in the context of rrn is puzzling, since there are no host-encoded versions of N, and the role of boxC remains unclear.
Cryo-EM structures of lambda N-TAC and E. coli rrnTAC at 3.7 and 4.0Å resolution, respectively, have shed new light on the mechanisms of antitermination (39,40). In both structures, the modifying proteins interact with one another, encircling the pore of the RNA exit channel. NusB heterodimerizes with NusE(S10) and interacts specifically with boxA of the nascent RNA. NusG adopts an extended conformation, with its N-terminal domain binding near the active site cleft of RNAP and its C-terminal domain binding NusE(S10). In rrnTAC, the SuhB dimer acts as a structural hub, interacting with NusA, NusE(S10), and NusG. While unrelated and structurally dissimilar to SuhB, N plays an analogous role in the case of N-TAC, interacting with both NusA and NusE(S10). In both complexes, NusA and NusG are stabilized in ways predicted to prevent RNAP pausing and inhibit backtracking, keeping RNAP in an elongationcompetent state. The bound antitermination factors probably prevent Rho-dependent and intrinsic termination by blocking Rho and NusA, respectively, from accessing the RNA exit channel.
In the early 1990s, Wagner and coworkers characterized numerous deletion and point mutations in the leader region of rrnB (41)(42)(43). Several mutations were found to reduce the 30S/50S ratio, 16S/23S ratio, growth rate, and activity of isolated subunits, implicating various portions of the leader in 30S subunit biogenesis. One caveat to these experiments is that they involved overexpression of rRNA to a high level (∼70% of the total rRNA), which on its own alters cell physiology. Their control strain grew with a doubling time of 40 minutes, nearly twice that of wild-type E. coli, and contained elevated levels of immature and/or misassembled ribosomes. Strains carrying particularly deleterious alleles of rrnB grew more poorly and gave rise to suppressors at high frequency. These issues, openly acknowledged by the authors, make interpretation of the reported data difficult. While the work of Wagner established the general importance of the leader RNA in 30S biogenesis, questions about the relative contributions of specific regions and features remain open.
In this work, we use an orthologous ribosome system to dissect the key functional elements of the leader and spacer regions. This system enables translation activity to be directly measured, over a >1000-fold range, without affecting cell growth (44,45). Various indicator strains, with specific codon changes in lacZ, allow the translational fidelity of ribosomes to be quantified in this system as well. Thus, we can obtain both quantitative (overall translation activity) and qualitative (translational accuracy) measurements of the product ribosomes, without worrying about complicating effects on cell physiology. We find that a leader-trailer helix of >17 basepairs (bp) is critical of formation of active 30S subunits in the cell. Other features, including boxA and leader helices hA and hB, are not strictly required. However, ribosomes formed in the absence of these elements are errorprone, suggesting that the antitermination complex and various intergenic RNA features contribute to the process and ensure quality control.
Note that throughout this article, we use the term subunit biogenesis to refer to the entire process of creating translationally-competent particles, including rRNA transcription.

Covariation analysis
The genome assemblies and annotations for each of the 508 GTDB (46)(47)(48)(49) members of the Enterobacteriaceae family listed in GTDB as NCBI type material (Supplementary Table S1) were downloaded from NCBI. For each annotated 16S rRNA, first the trailer sequence was identified by locating the conserved sequence AAGUCGUAA-CAAGGUA from the vicinity of the 3 end of the mature 16S rRNA within the range from 100 nt upstream to 200 nt downstream of the annotated 16S rRNA 3 end and then retaining the sequence from position 51 to position 107 downstream of this match. Then, the 5 leader sequence was identified by searching for boxC (UCUGUGUGGG; with up to two mismatches) in the range from 257 nt to 90 nt upstream of the annotated 5 end of the 16S rRNA and keeping the 3 most match if multiple matches to boxC are found. The leader is then defined as the sequence from 23 nt upstream of the boxC motif to the 18th nucleotide of the annotated 16S rRNA. Sequences for which the total length of leader and trailer together was below 350 nt were discarded resulting in 1441 total rRNA leader/trailer pairs (Supplementary Table S1). The leaders and trailers were concatenated with NNNNNN representing the position of the mature 16S rRNA. Clustering of all 1441 sequences was performed via RNAclust with default parameters. The resulting tree was visualized (Supplementary Figure S1) using iTOL (50). After manual inspection of the tree, it was divided into the 15 groups shown in Figure 1 (group assignments in Supplementary Table S1). Structural alignments and conservation analysis of each group was performed using mlocarna (51)(52)(53) with default parameters.

Orthogonal ribosome system
The translation activity of orthologous ribosomes was measured as described previously (44,54,55). Plasmid pDQ207 contains the 16S rRNA gene, rrsB, with the alternative ASD sequence 5 -GGGAT-3 , under transcriptional control of the P BAD promoter. This plasmid lacks the 23S and 5S genes; however, expression of 16S rRNA in this context has no apparent effect on growth rate or sucrose gradient sedimentation profile. The marker mutation T1451A was engineered into pDQ207 to generate pBW022. This mutation was used to track plasmid-encoded rRNA by primer extension. Mutations and deletions were constructed using QuikChange (Stratagene) or Phusion site-directed mutagenesis (New England Biolabs) (Supplementary Table S2, Supplementary Figure S2). Plasmid pBW024 ( T) was generated by Gibson Assembly (NEB), where the vector and a sequence encoding the hammerhead ribozyme of Schistosoma mansoni (56)(57)(58) were PCR amplified with overlapping ends. Stem III of the hammerhead was engineered such that cleavage occurs 2 nt downstream of the mature 16S rRNA. To construct the HH@ series of plasmids, DNA encoding the hammerhead ribozyme was inserted at position 1612 via restriction sites engineered in plasmid pBW039, a derivative of pBW022. Then Phusion site-directed mutagenesis was used to remove a given trailer segment, fusing the hammerhead to the position denoted. With this set of constructs, stem III of the hammerhead is constant, so 7 nt (5 -GGGCAUC-3 ) is left, appended to the pre-rRNA, after cleavage. These plasmid-borne 16S rRNA alleles were expressed in indicator strains KLF2674 (55), KLF2672, KLF2723, KLF3361 (45) and BRW299 (see below), enabling overall translation activity and translational fidelity to be measured, as described previously (45,55,59).

Beta-galactosidase activity measurements
Cells from an overnight culture were diluted 300-fold into 3 ml of fresh Luria broth (LB) containing ampicillin (Amp; 100 g/ml), kanamycin (Kan; 50 g/ml), and L-arabinose (5 mM) and grown for 4 h at 37 • C. The cells were washed once in 1 ml Z buffer (100 mM sodium phosphate at pH 7.0, 10 mM KCl, 10 mM MgSO 4 ), and ␤-galactosidase activity was measured as described (54). Specific activity was defined by the equation: 1 unit = 1000·(A 574 )/(OD 600 ·v·t), where A 574 is absorbance at 574 nm (characteristic of the product of CPRG cleavage), OD 600 is optical density of the cell suspension used, v is the volume of the cell suspension used (in milliliters), and t is time of incubation (in minutes) at room temperature. Strain KLF2674(pBW022) was used as the WT control and KLF2674(pBAD18) was used as the vector-only control.

Quantification of plasmid-encoded rRNA
Cells from an overnight culture were diluted 300-fold into fresh LB (25 ml) containing Amp (100 g/ml), Kan (50 g/ml) and L-arabinose (5 mM) and grown for 4 h at 37 • C. Cells were pelleted at 4 • C, resuspended in Z buffer, transferred to a 1.5 ml Eppendorf tube, and then re-pelleted. TRIzol (Invitrogen) reagent (1 ml) was added, and the pellet was resuspended by pipetting and vortex mixing (5 min) at room temperature. Chloroform was added, mixed, then the aqueous phase was transferred to a new tube. RNA was precipitated with isopropanol then pelleted. The RNA pellet was washed with 70% ethanol and dissolved in water. Relative amounts of P-16S rRNA were determined using poisoned primer extension, essentially as previously described (44,60). In a 20 l reaction, primer #1456 (5 -[Cy5]-AAAGTGGTAAGCGCCCT-3 ) (0.42 pmol) was incubated with RNA (5 g) at 50 • C for 10 min in AMV buffer (NEB), AMV reverse transcriptase (NEB, 3 units), ddATP, dCTP, dGTP, and dTTP (3.5 nmol each) were added, and the reaction was incubated at 42 • C for 1 h. Products were resolved by denaturing 8% PAGE, gels were imaged using a Typhoon 5 (Cytiva), and data were quantified using Image-Quant (Cytiva).

prrn strain construction and analysis
A recombination-deficient 7 prrn strain was generated by moving (recA-srl)306 srlR::Tn10 into SQZ10 (61) via P1 transduction, resulting in strain BRW246. Mutations to the leader and trailer region of rrsB were introduced into plasmid p278MS2 (62) to generate variants listed in Table 1. These plasmids were transformed into SQZ10 (or BRW246), and transformants were plated on sucrose (5%) to select against the resident plasmid pHKrrnC-sacB (54). Strains were verified by plasmid purification and sequencing. For growth rate measurements, overnight cultures were diluted 200-fold into fresh LB Amp, and growth was monitored at 37 • C by measuring OD 600 as a function of time. LocARNA was used to align 1441 sequences and generate a phylogenetic tree. Sequences from each of 15 clades were then subjected to co-variation analysis, resulting in the models shown. Color-coding indicates the number of different basepairs observed at a given position: red, 1; yellow, 2; green, 3; cyan, 4; blue, 5; magenta, 6. Tint level reflects the number of mismatches, with white indicating three or more mismatches. Larger high-resolution images of these models are provided in Figures S4-S18. A two-tailed t test was used to evaluate differences from the WT: *P < 0.05; **P < 0.005; ***P < 0.0005.
Variants of strain BRW246 (WT, boxA, 1575-1598, hA and hB) were grown to mid-log phase and subjected to sucrose gradient sedimentation analysis as described (63). Fractions (0.5 ml) were collected across the gradient, and RNA from the pre-30S, 30S and 70S regions of the gradient was extracted and analyzed by denaturing PAGE as previously described (14,64). Gels were stained with SYBR-Gold Nucleic acid stain (Invitrogen) and scanned using a Typhoon 5 (Cytiva). Data were quantified using Image-Quant (Cytiva). Young and Steitz (1978) proposed the first secondary structure model of the pre-16S rRNA (9), which includes a long (44 bp) leader-trailer helix interrupted by two adjacent ∼10 bp leader helices (Supplementary Figure S3A). Ten years later, Schlessinger and co-workers (12) proposed a similar model, with a continuous leader-trailer helix of 30 bp and two leader helices of 4 and 19 bp ( Supplementary Figure S3B). To evaluate these models, we applied covariation analysis to 1441 sequences of the family Enterobacteriaceae. Unlike the 16S rRNA itself, which is highly conserved, the flanking regions are quite variable. The RNAclust software based on LocARNA (51-53), which uses both primary sequence alignments and RNA folding algorithms, was used to build a phylogenetic tree of these sequences (Supplementary Figure S1), which we then split into 15 different clades. Sequences of each clade were then fed into LocARNA to generate a secondary structure model, with basepairs color coded to indicate the degree of covariation ( Figure 1, Supplementary Figures S4-S18). One common feature of all models is a long (25-40 bp) leader-trailer helix (hLT), well supported by covariation in clades Escherichia, Xenorhabdus, Proteus, Yersinia, Serratia, Budvicia and Pectobacterium (Figure 1). In the middle of the 5 strand of hLT lies a highly-conserved sequence, boxC. Despite boxC being nearly invariant across the family (Supplementary Table S3), covariation in this portion of hLT is still observed, due to interchangeable Watson-Crick and wobble basepairs. Another conserved helix, named here h0, forms between nt −11 to −3 and nt 7-17 of the pre-30S particle (Figure 2A). During final maturation of the subunit, h0 melts and nt 9-13 repairs with nt 21-25 (65). Xenorhabdus, Providencia and Dickeya show substantial covariation within h0 (Figure 1, Supplementary Figures S4-S18). A short helix of the leader, termed here hA, is seen in Escherichia, Cronobacter, Proteus, Yersinia, Serratia, and Budvicia structures but is absent from Photorhabdus, Plesiomonas, Providencia, Pectobacterium and Dickeya structures. Another helix of the leader, hB, ranges in size from 16-22 bp and is seen in the majority of clades analyzed. Overall, the Escherichia structure closely resembles that of Mixta, Xenorhabdus, Cronobacter, Proteus and Frischella. Other clades show distinguishing features, such as an additional helix between h0 and hB (Yersinia, Serratia, Pectobacterium, and Dickeya). The Escherichia clade structure agrees with the model of Schlessinger. However, our analysis lends no support for a conservation of base pairing between nucleotides −99 to −96 and 1554-1557 of hLT or between nucleotides −1 to −2 and 7-8 of h0 across the Escherichia clade (Supplementary Figure S3).

Effects of various mutations on 30S subunit activity
To functionally characterize the leader-trailer structure in E. coli, we used an orthogonal ribosome system described previously (44,54,55). In this system, plasmid-encoded pre-16S rRNA with an alternative anti-Shine-Dalgarno forms 30S subunits that specifically translate chromosomallyencoded lacZ mRNA. Deletion of the full leader [ L; (−209 to −1)] resulted in complete loss of activity (Figure 2A,C). Removal of the full trailer ( T), achieved by replacing the nt 1543-1589 with an efficient self-cleaving ribozyme from Schistosoma mansoni (56)(57)(58) to generate the near-precise 3 end of the 16S rRNA ( Figure 2B), also resulted in complete loss of translation activity (Figure 2A-C). Cells carrying a construct with both mutations ( L T), which should generate the mature 16S rRNA, also exhibited no detectable activity (Figure 2A-C). (−59 to −12)] reduced activity by ∼40% ( Figure 2C). These data suggest that hA and hB play relatively minor roles in 30S subunit biogenesis. The fact that these elements are functionally dispensable is consistent with their variable presence in the Enterobacteriaceae (Figure 1, Supplementary Figures S4-S18).

Effects of progressive truncations of the leader and trailer strands
To determine the minimal leader-trailer helix needed to form functional 30S subunits, we progressively deleted portions of the leader in the presence or absence of boxBA ( Figure 2D, top and bottom graphs). Both series showed a similar trend, with incremental decreases in activity as deletions extend from −128 to −114, and complete loss of activity when deletions extend to −109 or beyond. We also progressively deleted the trailer, by fusing the hammerhead ribozyme at various downstream positions (Figure 2D, top graph). Truncations of the trailer strand to position 1589 or 1584 had no effect; truncations to 1579 or 1574 reduced activity by ∼75%; and truncations to 1569 or further resulted in complete loss of activity. Leader and trailer deletions coincide remarkably well and show that basepairs formed by nt −113 to −89 of the leader and nt 1548-1570 of the trailer are most critical (Figure 2A, D). The base of hLT is least important, as nucleotides downstream of 1584 and upstream of −124 can be removed without loss of 30S activity.

Effects of mutations in boxC
The sequence of boxC is highly conserved (Supplementary Table S3) yet its precise role remains elusive. We targeted boxC and the complementary portion of the trailer ( Figure  3A). Disruption of base pairing by substitution of 4 or 6 nucleotides of the leader or trailer resulted in substantial loss of activity (15-to 100-fold) in all cases ( Figure 3B). Surprisingly, mutations in the trailer strand had slightly more deleterious effects than those in the leader strand (i.e. boxC) ( Figure 3B). In all cases, when base pairing in hLT was restored, so were levels of translation activity in the cell (Figure 3B). These data show that the structure of hLT is much more important than the sequence of boxC.

Effects of increasing the distance between boxA and boxC
Based on their cryo-EM structure of rrnTAC, Wahl and coworkers hypothesized that positioning of boxC close to the exit channel of RNAP is important for 30S subunit assembly (40). According to their model, boxC remains stationary during transcription, anchored by boxA binding to NusB, and the nascent 16S rRNA loops out as it is synthesized. When the trailer emerges from the RNAP exit channel, boxC is poised to pair with the trailer sequence, enabling formation of hLT. In the Enterobacteriaceae, boxC is consistently positioned downstream of boxA, with 18 or 19 nucleotides between these elements in 98% of the cases (Supplementary Table S4). To directly test the importance of this spacing, we inserted 6 or 12 nt downstream of boxA (Figure 2A), which would be predicted to move boxC out of the rrnTAC channel and into the solvent. The 6-nt insertion (Ins 6) resulted in a 50% reduction in translation activity, while the 12-nt insertion (Ins 12) resulted in a 25% reduction ( Figure 2C). These effects are somewhat smaller than that of boxA, suggesting that rrnTAC function may only partially depend on boxA-boxC spacing.

Effects in the absence of RNase III
The RNase III recognition site overlaps with boxC in hLT. The rnc gene is dispensable in E. coli (67,68), ruling out an essential role for RNase III in ribosome biogenesis. However, loss of translation activity coincides with removal of the RNase III site (Figure 2A, D), raising the question of whether (or to what degree) loss of RNase III activity is responsible for these phenotypes. To address this, we regenerated the complete set of orthogonal ribosome strains in the rnc background and re-measured translation activity. Generally lower ␤-galactosidase levels were seen in the rnc strain set (Supplementary Figure S19). While the basis of this effect remains unclear, some change in our readout came as no surprise. RNase III is known to cleave multiple RNA targets besides pre-rRNA in the cell, including the mRNA encoding PNPase; consequently, widespread changes to the transcriptome are evident in rnc cells (69). Importantly, comparison of the normalized data showed that the effects of leader/trailer mutations are remarkably similar whether RNase III is present or absent (Supplementary Figure S19). In both strain sets, effects of progressive truncations of either the leader or trailer strand predicted the same critical portion of hLT. Interestingly, rnc appeared to suppress the effects of certain leader or trailer mutations, reminiscent of data reported previously for pre-23S rRNA (70). Collectively, these observations suggest that the leader-trailer helix itself plays a critical role for 30S biogenesis, independent of RNase III.

Effects of leader and trailer mutations on translation fidelity
A growing body of evidence suggests that defects in 30S assembly can result in error-prone ribosomes in the trans-Nucleic Acids Research, 2023, Vol. 51, No. 10 5249 lationally active pool (71)(72)(73)(74)(75)(76). To investigate the effects of leader/trailer mutations on the fidelity of the subunits produced, we moved a subset of our constructs into indicator strains KLF2672, KLF2723 and KLF3361 (45). These strains are isogenic to KLF2674 (used above) but carry specific mutations in lacZ, enabling errors in start codon selection, decoding, and frame maintenance to be quantified. Leader mutations boxBA, boxA, U-141G, hA, hB and Ins 6 each caused significant defects in initiation and elongation fidelity (Figure 4). These mutations increased AUC initiation by 2-to 4-fold, UGA readthrough by 2-to 4-fold, and + 1 frameshifting by 2-to 5-fold. Mutation Ins 12 also conferred defects in start codon selection and frame maintenance, increasing error rates by ∼2fold ( Figure 4A, C). The trailer mutation HH@1574 also caused fidelity defects, increasing error rates in all three indicator strains by 3-to 5-fold (Figure 4). Mutations that alter the primary sequence of hLT (T4L4, T6AL6A, and T6BL6B) increased frameshift errors modestly but had no significant impact on AUC initiation or UGA readthrough ( Figure 4A, B). These data suggest that intergenic RNA elements and the rrnTAC contribute not only to the efficiency of subunit biogenesis but also to the fidelity of the subunits produced.

Levels of rRNA in the trailer mutant strains
To further explore the consequences of these mutations, we used primer extension in the presence of ddATP to quantify the levels of plasmid-encoded rRNA in various KLF2674 transformants. Strains which produced control (WT) orthogonal ribosomes exhibited ∼15% plasmidencoded rRNA (Figure 5), in line with previous work (44). Strains carrying mutations HH@1579 or HH@1574, which produce 75% fewer active ribosomes ( Figure 2D), showed 30% reduced levels of rRNA ( Figure 5C). Strains which had no translation activity, HH@1544 and HH@1569 ( Figure  2D), had 70% reduced levels of rRNA ( Figure 5C). All these mutations lie well downstream of the promoter and are unlikely to influence transcription initiation. We envisage that these mutations confer assembly defects, and the misassembled particles are targeted for degradation, explaining the lower steady-state levels of 16S rRNA observed.

Effects of various mutations on E.coli strain 7 prrn
To further evaluate the role of the rrnTAC on ribosomal subunit biogenesis, we moved the boxB, boxA and boxBA mutations into the 7 prrn strain, which lacks chromosomal rRNA operons and is supported by a single frameshifting (C) of ribosomes made from various constructs (as indicated) were determined. Data represent the quotient of two means ± standard error from three or more independent biological replicates. A two-tailed t test was used to evaluate differences from the WT: *P < 0.05; **P < 0.005; ***P < 0.0005. n.s., not significant. plasmid-borne rrn operon (61,77). Deletion of boxB and/or boxA had no significant effect on the strain's doubling time (Table 1). We also moved hA and hB into 7 prrn and again, no obvious change in doubling time was seen. These data provide further evidence that these elements are dispensable for the assembly of 30S subunits.
Next, we tried to move various deletions of the trailer region into 7 prrn. Only the smallest deletion tested, 1575-1598 could be introduced (Table 1), and the resulting strain grew slowly, forming very small colonies on solid media. Attempts to measure doubling were hampered by suppressors, which arose at high frequency. We suspected that these suppressors might stem from recombination events between the introduced plasmid and ptRNA67 (77), which contains a portion of rrnB including the pre-16S trailer. So, we moved (recA-srl)306 srlR::Tn10 into SQZ10, making a recombination-deficient (Rec − ) 7 prrn strain, and then replaced the resident plasmid (pHKrrnC-sacB) with each of several rrnB-containing plasmids. In this Rec − background, pBW136 ( 1575-1598) supported slow growth, and there was no indication of suppressor mutations. This enabled us to measure its doubling time, which was 23 min longer than the control strain (Table 1). Rec − 7 prrn strains harboring boxA, hB or hA grew nearly as fast as the control, with slightly higher doubling times measured in the former two cases (Table 1).
To further evaluate these mutations, we subjected the Rec − 7 prrn strains to sucrose gradient sedimentation analysis (Supplementary Figure S20). In all cases, including the WT control, subunit peaks were considerably larger than polysome peaks, in contrast to typical profiles from wild-type E. coli cells (78). This suggests that ribosome biogenesis or its regulation is compromised in 7 prrn, which may help explain its reduced growth rate (1.7 doublings/hour) compared to wild-type E. coli (2.4 doublings/h) (78). None of the mutations conferred a significant change in the proportion of subunits in the context of 7 prrn, although the mean values for 1575-1598 and hB are suggestive of small increases ( Supplementary Figure S20B). RNA was extracted from the sucrose gradient fractions and analyzed by PAGE to evaluate pre-16S pro-cessing (Supplementary Figure S21). An RNA shorter than 16S rRNA is observed in several of the mutant strains, being most prominent in 7 prrn ( 1575-1598). This product is reminiscent of 16S*, a truncated form of 16S rRNA, seen in ΔrsgA and ΔrimM strains, which are defective in 30S assembly (14,64,79). This 16S* band may stem from an error in pre-rRNA processing or reflect the initial degradation of dysfunctional subunits. Either way, the presence of 16S* in these strains provides further evidence that 30S biogenesis is negatively impacted by these mutations.

DISCUSSION
In this work, we find that a leader-trailer helix of 17 or more basepairs is necessary for generating active 30S subunits in E. coli. The fact that absolutely no functional subunits are detected in the absence of hLT came as a surprise, because mature 16S rRNA supports efficient assembly of subunits in vitro. How can this apparent discrepancy be explained? A growing body of evidence suggests that the free energy landscape of RNA folding in the cell differs from that in the test tube (80,81). Molecular crowding, osmolytes, liquidliquid phase condensates, and various proteins and enzymes distinguish the cytoplasmic environment from that of standard reconstitution reactions, and all these parameters impact RNA folding and dynamics. Global chemical probing studies have revealed that RNA molecules tend to be considerably less structured in vivo than in vitro (82,83). Depletion of intracellular ATP leads to increased RNA structure, suggesting that enzymes such as helicases continually promote unfolding events in the cell (83). Other recent work has shown that binding of Pumilio (PUM1/PUM2) protein to its single-stranded RNA targets is predictably inhibited by RNA structure in vitro but not in vivo, further evidence that mechanisms and/or conditions in the cell promote RNA unfolding (84). We envisage that hLT limits the conformational dynamics of pre-16S rRNA molecule by connecting its 5 and 3 ends. This may effectively counter the cellular drivers of unfolding and open favorable routes of 30S subunit formation.
Hammerhead cleavage leaves a 5 hydroxyl on the ribozyme and a 2 ,3 -cyclic phosphate on the upstream RNA product. This raises the question of whether this phosphate (2 ,3 -or 3 -linked) might interfere with subsequent 16S rRNA processing in our system. Five enzymes have been implicated in 3 end maturation: an endonuclease, YbeY, and four 3 -to-5 exonucleases--RNase II, RNase R, RNase PH, and PNPase. These exonucleases act in a redundant fashion to ensure that the 3 tail of 16S rRNA is fully trimmed (85). The activity of RNase PH on pre-tRNA substrates is strongly inhibited by a 3 phosphate group (86), and the homologous PNPase may be similarly susceptible to such inhibition. By contrast, RNase II and RNase R are largely unaffected by the presence of a 3 -or 2 -3 -phosphate (87,88). Of the four exonucleases, RNase II and RNase R contribute most to 16S rRNA maturation, based on analysis of various triple mutant strains (85). In fact, either RNase II or RNase R is sufficient for 16S rRNA maturation (in cells containing YbeY). Given this insight, we consider it highly unlikely that hammerhead cleavage itself contributes to any substantial way to the observed effects on subunit ac-tivity. Consistent with this view, the relative production of active subunits from constructs HH@1589, HH@1584, or HH@1579 is just as high in the absence of RNase III as in its presence (Supplementary Figure S19).
In the cell, there exist pathways to target and degrade defective or damaged ribosomes. These mechanisms, while poorly understood, act on long-lived assembly intermediates and off-pathway (misassembled) particles (89)(90)(91). In the absence of hLT, slow or defective 30S biogenesis may lead to rapid degradation of 16S rRNA in assembly intermediates, further hampering production of active subunits. Indeed, we see reduced levels of plasmid-encoded rRNA when hLT cannot form, in line with active rRNA turnover. Notably, rRNA from constructs HH@1544 and HH@1569 is present at ∼30% the control level and yet no translation activity is observed, indicating some major problem in subunit assembly. Whether detectable levels of active subunits could form without the leader-trailer helix in cells lacking one or more RNases implicated in quality control remains an open question, one worth pursuing in future studies.
A growing body of evidence indicates that defects in 30S biogenesis can give rise to error-prone ribosomes. Mutation or loss of particular ribosomal proteins (rpsE-G28E, ΔrpsO), assembly factors (ΔrimM, ΔksgA, ΔrbfA, ΔlepA), or endonucleases (Δrng, ΔybeY) leads to increased rates of stop codon read-through, frameshifting, and/or spurious initiation (71)(72)(73)(74)(75)(76). Compelling evidence suggests that quality control mechanisms are compromised in these strains, enabling premature or defective 30S particles to enter the translationally active pool. For example, Culver and coworkers found that, in strains defective in 30S assembly, precursor 17S rRNA is present in 70S ribosomes, and the level of 17S coincides with error rate (74,75). Moreover, Sharma and Anand showed that 30S subunits purified from ribosomes of ΔksgA or ΔrbfA cells have low affinity for IF3, indicating that both KsgA and RbfA are needed to enable and/or ensure IF3's functionality on the subunit (76). Here, we show that deletion of either leader helix ( hA or hB) or shortening of hLT by 15 bp (HH@1574) leads to reduced initiation and elongation fidelity. These data suggest that precursor RNA structures and/or dynamics also contribute to quality control. In line with this view, mutations Δrng and ΔybeY alter leader and trailer strand processing, respectively, and both mutations lead to reduced translational fidelity (71,74). How precursor RNA elements help ensure quality control remains an open question. They may for example facilitate AF action, rRNA folding, and/or rRNA processing. In general, mechanisms of quality control in bacterial ribosome biogenesis remain poorly defined. Future analyses of the mutations described here may shed light on these mechanisms.
Previous studies have shown that boxA is the key cisacting RNA element, both necessary and sufficient for rrnTAC formation (22,31,92). Here, we show that deletion or mutation of boxA reduces 30S activity by about 3fold, indicating that transcription antitermination moderately enhances the production of 30S subunits, in line with earlier work (66). Additionally, we find that these boxA mutations lead to defects in translational fidelity. This implies that rrnTAC not only increases transcriptional processivity (66) but also contributes in some way to the quality control of ribosome biogenesis. How rrnTAC acts in this regard remains unclear. It has been proposed that rrnTAC has RNA chaperone activity and can facilitate the folding of nascent rRNA (40). Mutations Ins 6 and Ins 12 phenocopy mutations boxA and U-141G, which lends general support to the idea that tethering of boxA to NusB may indeed promote hLT formation (40). It is also possible that the role of rrnTAC is less direct. For example, transcription elongation speed and/or cadence may be optimal for ribosome assembly in the presence of rrnTAC, and suboptimal in its absence. Further studies will be needed to understand the functional link between the rrnTAC and the fidelity of the product ribosomes.
Like that of 16S rRNA, the precursor of 23S rRNA is bookended by a long leader-trailer helix. In E. coli, this 27 bp helix is processed by RNase III, RNase T, RNase AM and other RNases during 50S maturation (93), leaving the 8 bp helix H1 behind. Helix H1 lies on the solvent side of the subunit, well away from active centers of the ribosome. In some bacteria, which naturally lack H98, H1 is fully removed during 50S maturation (94). These observations suggest that the basepairs of H1 function solely in 50S biogenesis, as part of the long leader-trailer helix. Using the erythromycin-resistance mutation A1067U to distinguish plasmid-and chromosomally-encoded rRNA, Liiv and Remme performed a mutational analysis of the leader and trailer region of pre-23S rRNA in E. coli (70). They found that base substitutions or deletions that disrupt the leader-trailer helix have serious consequences on 50S biogenesis, akin to our current results for pre-16S rRNA. One caveat to the Liiv and Remme study is that their constructs, which carry deletions within the leader or trailer region, express pre-23S rRNA molecules that still contain 5 and 3 flanking RNA. These flanking sequences, which vary by construct, can potentially interact, complicating interpretation of the data. It may be worthwhile to re-visit mutagenesis of pre-23S rRNA, using a promoter and the hammerhead ribozyme to generate defined 5 and 3 ends, and compare the minimal pairing free energy requirements for the leader-trailer helices of pre-16S versus pre-23S particles.
Finally, we also present secondary-structure models for 15 distinct leader-trailer structures found within the Enterobacteriaceae. In all cases, a long (>30 bp) leader-trailer helix is observed, well-supported by covariation in multiple clades. Aside from this common feature, considerable structural diversity is evident (Figure 1), even though ribosomes of these organisms are highly similar. In fact, diverse leader-trailer structures are seen even within the same organism. For example, Providencia heimbachae has seven rrn operons which encode nearly identical 16S rRNA (>99.6% identity in all pairwise comparisons). Two of these operons encode the Providencia I leader-trailer structure, characterized by three leader helices of 10-15 bp; whereas, the other operons encode the Providencia II leader-trailer structure, characterized by five leader helices of 3, 6, 8, 10 and >30 bp (Figure 1). Several other organisms (Providencia rustigianii, Serratia symbiotica, Shigella boydii, Shimwellia blattae, Tatumella citrea, Xenorhabdus cabanillasii, Yersinia aleksiciae, Yokenella regensburgei) similarly have two different leader-trailer structures represented among their rrn operons (Supplementary Table S1). In future studies, it will be worthwhile to address whether such distinct structures play unique or common roles in the cell.

DATA AVAILABILITY
The custom scripts used during data analysis are available at doi://10.5281/zenodo.7589875.