Plasmid p287MS2 carries the rrnB operon downstream from the lambda P_L promoter (Youngman et al., 2004). Single mutations were made in the ASD region of p287MS2, generating the plasmids of Table 1 (pMDxx; where “xx” represents a unique number). To test for dominant lethal/negative phenotypes, each pMDxx plasmid was transformed into DH10 (pcI857), and transformants were evaluated for growth at 30°C and 43°C, as described (Samaha et al., 1995; Cochella et al., 2007). To test the ability of the mutant ribosomes to support cell growth, the Δ7 strain SQZ10 was employed (Qin et al., 2007; Quan et al., 2015). Each pMDxx plasmid was transformed into SQZ10, selecting for ampicillin resistance (100 μg/ml). The resulting transformants were grown in liquid media, and cells were spread onto plates containing ampicillin and sucrose (5%), to select against the resident plasmid pHKrrnC-sacB. Successful plasmid replacement was evident by a high frequency of sucrose resistant (and kanamycin sensitive) colonies, and subsequently confirmed by plasmid purification and DNA sequencing (Qin et al., 2007). Unsuccessful plasmid replacement was indicated by a low frequency of sucrose resistant colonies; i.e., more than four orders of magnitude lower than the control (p278MS2) case. Most of these colonies retained kanamycin resistance, and any rare isolates sensitive to kanamycin were found to contain the wild-type 16S rRNA gene, presumably due to homologous recombination.

All F. johnsoniae strains (Table 2) were grown on rich CYE medium at 30°C. F. johnsoniae plasmids (Table 2) were transformed into E. coli strain E726 and then moved into F. johnsoniae via tri-parental mating as described previously (McBride and Kempf, 1996).

Mutations to the F. johnsoniae chromosome were made using precise allelic replacement (Zhu et al., 2017). Alleles were cloned into the Bam HI and Sph I restriction sites of the suicide vector pYT313 (Zhu et al., 2017). This vector has two selectable markers, bla (expressed in E. coli) and ermF (expressed in F. johnsoniae), as well as the counter-selectable sacB gene (expressed in F. johnsoniae). Alleles were generated by separately amplifying ∼1 kb regions from the F. johnsoniae chromosome both up- and downstream of the target site. These two fragments were then inserted into pYT313 using Gibson Assembly (Gibson et al., 2009). Resulting plasmids were moved into F. johnsoniae, and erythromycin (Em, 100 μg/ml) resistant transconjugants were selected. Colonies were then screened for plasmid integration at the appropriate chromosomal locations using PCR. Confirmed recombinants were then grown overnight in the absence of Em, to allow for loss of the plasmid via a second recombination event, and then cells were plated on 5% sucrose for counterselection. Sucrose resistant/Em sensitive colonies were then screened via colony PCR for replacement of the wild-type allele for the mutant allele. rrnF was deleted from F. johnsoniae by removal of the chromosomal region 5,118,368 to 5,124,329. rrnA was deleted by removal of the chromosomal region 24,082 to 30,164. rrnB was deleted by removal of the chromosomal region 49,9700 to 50,6103.

The F. johnsoniae rrnA operon (chromosome positions: 29,556–23,688) was cloned into the Bam HI and Sph I restriction sites of expression vector pSCH710 (Baez et al., 2019), downstream of the inducible ompA promoter, to generate pZM06. The marker mutation C1451U (phenotypically silent) was introduced into the plasmid-encoded 16S gene, using site-directed mutagenesis, to yield pZM14. Other rrn alleles were similarly cloned into pSCH710 using Gibson Assembly. Plasmids were moved into F. johnsoniae strains via tri-parental mating and selecting for Em (100 μg/ml) resistance.

Overnight cultures of F. johnsoniae UW101 and ZAM11 were used to seed fresh CYE medium both separately (wild-type and mutant only) or mixed (∼1:1, eight replicates). Inoculated cultures were grown up and back-diluted 200-fold to seed another culture, a process repeated daily for 36 days. Aliquots were taken from saturated cultures for use as template for PCR to quantify the fraction of prfB mutants. Because the allele of the prfB mutant is effectively shortened by the removal of a single base, amplification of the prfB gene around the frameshift site resulted in two different size PCR products for the mixed cultures. PCR was done with Phusion DNA polymerase (NEB), since this enzyme leaves clean blunt-ended PCR products. Primers prZM53 (TAT​TGT​GGA​GCG​CCT​TGG​TGC​GTT) and prZM55 (ATT​TCG​ATT​AGC​TTG​GCA​TCA​ACG​TC) were used to amplify the prfB alleles, producing a 64 bp product for the wild-type allele and 63 bp for the mutant allele. Radiolabeled prZM53 was included in the reaction. Briefly, prZM53 was 5′ end-labeled using γ-[³²P]-ATP and T4 polynucleotide kinase (NEB) and purified from free γ-[³²P]-ATP by Sephadex G-25 (Amersham Biosciences) chromatography. PCR products were resolved by denaturing 8% PAGE. Gel imaging and quantification were performed with a Typhoon FLA 9000 phosphorimager (GE Healthcare) and associated software (ImageQuant 5.2).

A list of 997 organisms from the orders Cytophagales, Bacteroidales, Chitinophagales, Flavobacteriales, and Sphingobacteriales marked as GTDB species representatives and as NCBI type material were downloaded from GTDB (Parks et al., 2021) (Table S2). 740 of these genome assemblies were successfully downloaded and ARFA (Bekaert et al., 2006) was run on these assemblies with default parameters. A prfB gene was identified in 726 of these assemblies. Out of the 524 of the 726, in which ARFA detects a frameshift, we manually inspected all five for which the E-value for the detection of the gene fragment upstream of the frameshift (ORF0) was above 0.05, and identified one [Weeksella virosa, which had the highest of all E-values (0.32) for ORF0] that was miscalled by ARFA as frameshifted. We visualized the phylogenetic tree of the 726 organisms with a detected prfB gene using iTOL (Letunic and Bork, 2021).

For each strain, cells from overnight cultures were diluted 100-fold into CYE medium. If included, erythromycin was added to a final concentration of 100 μg/ml and IPTG was added to a final concentration of 1 mM. Cultures were shaken at 250 rpm at 30°C, and aliquots were regularly taken throughout growth to measure the optical density (OD) at 600 nm. Doubling times were determined by fitting the data of the logarithmic phase of growth.

Ribosomal particles were fractionated using methods described previously (Qin and Fredrick, 2013). Briefly, cells were grown to mid-log phase (OD₆₀₀ 0.4–0.7), poured over crushed ice, and harvested via centrifugation. The cell pellet was resuspended in lysis buffer [10 mM Tris-HCl (pH 8.0), 10 mM MgCl₂, 1 mg/ml lysozyme] and flash frozen three times in liquid nitrogen to lyse the cells. Deoxycholate was added (0.3% final), cell debris was pelleted, and clarified lysate (0.4 ml) was loaded onto an 11 ml 10–40% (wt/vol) sucrose gradient and subjected to ultracentrifugation for 3.5 h at 35,000 rpm in an SW41 rotor (Beckman Coulter). Gradients were pumped using a syringe-pump system (Brandel) with in-line UV absorbance detector (UA-6, ISCO; 254 nm), and 1 ml fractions were collected.

Ribosomes were precipitated from sucrose fractions with ethanol, pelleted, and dissolved in 200 µL extraction buffer [0.3 M sodium acetate (pH 6.5), 0.5% SDS, 5 mM EDTA]. RNA was extracted twice with water-saturated phenol and twice with CHCl₃/isoamyl alcohol (24:1). Extracted RNA was then precipitated with ethanol, pelleted, and dissolved in water.

To determine the relative amount of mutant 16S rRNA in each fraction, poison primer extension was used as described (Abdi and Fredrick, 2005). Primer prZM66 (GTT​ACC​AGT​TTT​ACC​CTA​GGC​A) was designed to anneal to 16S rRNA at a position 3′ of the marker mutation C1451U such that extension of the primer in the presence of dideoxyadenosine triphosphate (ddATP) results in distinct extension products that reflect the fraction of templates containing the mutation. Briefly, prZM66 was 5′ end-labeled using γ-[³²P]-ATP and T4 polynucleotide kinase (NEB) and purified from free γ-[³²P]-ATP by Sephadex G-25 (Amersham Biosciences) chromatography. In a 10-μL reaction containing 50 mM HEPES (pH 7.6) and 100 mM KCl, labeled primer was annealed to ∼1.5 pmol 16S rRNA by heating the reaction to 95°C for 1 min and then allowing it to cool slowly. After a brief centrifugation to recover condensation, 10 μL of 2X extension mix [260 mM Tris-HCl (pH 8.5), 20 mM MgCl₂, 20 mM DTT, 6 U AMV reverse transcriptase (Life Sciences Advance Technologies Inc.), 340 μM of ddATP, and 340 μM of each other deoxynucleotide triphosphate) was added and the reaction was incubated for 10 min at 42°C. Finally, the primer extension products were passed through a Sephadex G-25 column, dissolved in loading solution (95% formamide, 20 mM EDTA, 0.05% xylene cyanol FF, and 0.05% bromophenol blue), and resolved by denaturing 8% PAGE. Gels were then dried and imaged as described above.

The ASD region of the E. coli ribosome has been targeted in several previous studies (Hui and de Boer, 1987; Jacob et al., 1987; Lee et al., 1996; Rackham and Chin, 2005; Hui et al., 1988; Yassin et al., 2005). Most of these studies aimed to generate functionally orthogonal ribosomes and hence entailed the simultaneous substitution of multiple nucleotides (e.g., 1,535–1,540). While certain single mutations have been analyzed (Jacob et al., 1987; Yassin et al., 2005), to our knowledge no one has performed a comprehensive analysis of single substitutions across this critical region. We did so here, targeting nine positions (1,534–1,542) of the 16S rRNA gene in plasmid p287MS2 (Youngman et al., 2004). This plasmid contains the ribosomal RNA operon rrnB downstream from the P_L promoter, allowing temperature-dependent transcription in cells containing a labile form of lambda repressor (cI857). Each plasmid was moved into E. coli strain DH10 (pcI857) (Samaha et al., 1995; Durfee et al., 2008), and cell growth was assessed at 43°C, conditions of P_L de-repression (Table 1, Supplementary Figure S1). Production of 16S rRNA substituted at position 1,535, 1,536, 1,537, 1,538, or 1,539 conferred dominant negative effects. Certain variants (C1536G, U1537G, C1538G, C1539A, and C1539U) were especially deleterious. Presumably, these strong effects stem from altered specificity of the mutant ribosomes during initiation, consistent with widespread proteomic changes seen in analogous studies (Jacob et al., 1987). Next, each plasmid was tested for its ability to support the growth of SQZ10, an E. coli strain lacking all seven chromosomal rrn operons (Δ7) (Qin et al., 2007; Quan et al., 2015). Most alleles substituted at positions 1,540–1,542 were able to complement the Δ7 strain, whereas alleles with any mutation further upstream could not (Table 1). Collectively, these data indicate the functional importance of nucleotides 1,534–1,539 in E. coli and suggest that CCUCC represents the core ASD.

On its single chromosome, F. johnsoniae contains six virtually identical rrn operons that each encode 16S rRNA, tRNA^Ile (anticodon GAU), tRNA^Ala (anticodon UGC), 23S rRNA, and 5S rRNA. Starting with ZAM11, a strain which constitutively produces RF2 (see below), we began to progressively delete rrn operons (named rrnA-F, based on chromosome position; Figure 1). Loss of one (ΔrrnF) or two (ΔrrnFΔrrnA) operons had little if any effect on growth (Figure 1C, Supplementary Table S1). However, loss of three operons (ΔrrnFΔrrnAΔrrnB) slowed growth considerably, increasing the doubling time from 70 to 90 min. These findings are reminiscent of E. coli studies which showed that a minimum of four chromosomal operons are needed to sustain rapid growth (Quan et al., 2015).

An rrn operon with a marker mutation (C1451U; a base substitution in the tetraloop of h44, predicted to be phenotypically silent) was cloned downstream of an engineered IPTG-inducible promoter in the shuttle vector pSCH710. The resulting plasmid (pZM14) was moved into the Δ3 strain of F. johnsoniae, and growth in the absence and presence of IPTG was measured (Figure 2, ZAM26; Supplementary Table S1). The doubling time decreased from 90 to 80 min in the presence of inducer (1 mM), indicating clear albeit partial complementation by the plasmid-borne rrn operon. Expression of tRNA^Ile and tRNA^Ala only (Figure 2, ZAM28) had no effect, indicating the importance of rRNA in the complementation. The marker mutation C1451U allowed us to track the plasmid-encoded 30S subunits in various ribosome fractions, using primer extension with dideoxy-ATP (Figure 3, ZAM26). These subunits accounted for ∼25% of the total and were distributed evenly across all fractions of the sucrose gradient. This distribution pattern shows that the plasmid-encoded subunits are as active as chromosomally-encoded subunits in these cells.

Using this complementation system, we began to evaluate mutations to the ASD core. Derivatives of pZM14 harboring various single mutations were made and introduced into the Δ3 strain. Growth of the resulting strains (ZAM41-43) in the absence and presence of IPTG was measured, and the data were indistinguishable from that of the ZAM26 control (Figure 2, Supplementary Table S1). In other words, the mutant 30S subunits carrying C1535G, C1536A, or U1537A rescued the growth defect of the Δ3 strain as well as the control subunits and hence have similar activity. Notably, these same mutations confer dominant negative phenotypes in E. coli (Table 1).

As a separate control, we introduced the A-site mutation A1492U into plasmid pZM14 and moved the resulting plasmid into the Δ3 strain (Figure 2, ZAM50; Table S1). Mutation A1492U targets the 30S A site and eliminates translation activity in E. coli (Abdi and Fredrick, 2005). Expression of 16S (A1492U) rRNA in the presence of IPTG strongly inhibited growth, increasing the doubling time to 104 min. This dominant negative phenotype is in line with analogous experiments done in E. coli (Powers and Noller, 1990; Cochella et al., 2007) and indicates that, in this F. johnsoniae system, plasmid-encoded 16S rRNA is expressed at levels high enough to confer such phenotypes.

Next, we heavily mutagenized the ASD and tested the ability of the corresponding mutant ribosomes to restore growth of the Δ3 strain (Figure 2, Supplementary Table S1). Production of subunits with quadruple substitutions within the ASD (CCUCC to GAAGC or AUUGG; mutations underscored) reduced doubling times from 90 to ∼82 min, rescues nearly as robust as that provided by control (CCUCC) subunits. Subunits in which the core ASD is replaced with AAAAA also stimulated growth, albeit to smaller degree. Thus, ribosomes lacking the ASD sequence can translate endogenous mRNA in F. johnsoniae.

The activity of various mutant ribosomes in the cell was evaluated by quantifying the proportion of plasmid-encoded 16S rRNA in ribosomal particles fractionated by sucrose gradient sedimentation (Figure 3). Subunits carrying four or five substitutions in the ASD exhibited similar profiles, with overrepresentation in the subunit region (fractions 3–4) and underrepresentation in the polysome region (fractions 8–10). These data indicate that wild-type ribosomes outcompete the mutant ribosomes for mRNA loading. This could stem from a defect in initiation or assembly, as in either case 30S particles would accumulate and fewer ribosomes would enter the actively-translating pool. Notably, A₂₆₀ traces of the lysates were similar across the board (Supplementary Figure S2), indicating little or no effects on the overall proportions of subunits, monosomes, and polysomes in these strains.

In the case of A1492U, mutant particles accumulated in the 70S region (fractions 6–7) and were present at reduced levels in both the subunit and polysome regions (Figure 3). As this mutation targets an A-site residue critical for decoding, these mutant ribosomes are presumably trapped as 70S initiation complexes, unable to transition to elongation. Their presence in polysomes can be explained by one or more wild-type ribosomes downstream on the mRNA, and their relative underrepresentation in the subunit region might be explained by an inability of “stuck” 70S complexes to dissociate.

In many bacteria, the gene encoding RF2, prfB, is autoregulated via a +1 programmed frameshifting mechanism (Craigen and Caskey, 1986; Weiss et al., 1987; Baranov et al., 2002). In E. coli, the frameshift site corresponds to AGGGGGUAUCUU UGA C, where the slippery sequence (underscored) overlaps the in-frame stop codon (bold italics), just downstream from a SD-like sequence (bold). When RF2 levels are low in the cell, ribosomes containing peptidyl-tRNA^Leu:CUU in the P site and codon UGA in the A site pause. Because the SD-like sequence is closely juxtaposed to the P codon, pairing to the ASD causes tension on the mRNA that promotes slippage of peptidyl-tRNA^Leu from CUU (zero frame) to UUU (+1 frame) (Devaraj and Fredrick, 2010). Continued translation in the +1 frame allows production of full-length RF2. As RF2 levels rise, the rate of termination at the in-frame UGA increases, down-regulating further production of the factor.

One might expect Bacteroidia to lack this autoregulatory mechanism because it depends on mRNA-rRNA pairing. However, we noticed that the prfB gene of F. johnsoniae contains a +1 programmed frameshift site, virtually identical to that of E. coli but with a “perfect” SD-like sequence: AGGAGGUAUCUU UGA C (annotated as above). To evaluate the prevalence of prfB programmed frameshifting across the class, we used the tool ARFA (Bekaert et al., 2006). Over 700 representative genomes were analyzed, and in 72% of the cases (523/726), prfB contains the frameshift. This value is in line with frequencies calculated previously (70–87%), using organisms from multiple phyla (Baranov et al., 2002; Bekaert et al., 2006). Thus, the autoregulatory mechanism seems no less common in the Bacteroidia. Supplementary Figure S3 shows occurrences of programmed frameshifting projected onto the GTDB phylogenetic tree (Parks et al., 2021). The frameshift is present in all Sphingobacteriales analyzed and absent from all Weeksellaceae analyzed. However, the other clades show considerable variability, implying evolutionary loss and/or gain of the autoregulatory mechanism.

Before embarking on genetic analysis of the ASD (described above), we decided to remove the prfB frameshift, because we were mainly interested in roles of the ASD beyond this autoregulatory mechanism. We replaced the frameshift site (codons 18–22) with the sequence CGT AGA TAT CTT GAC. The resulting strain, ZAM11, which contains an undisrupted prfB open reading frame encoding the wild-type RF2 protein, exhibited no obvious growth defect. To further characterize the strain, we co-cultured ZAM11 and its parent strain UW101 in eight replicate experiments, growing the cells in CYE medium and passaging them every day for 36 days (Figure 4). Samples were removed at each passage, and PCR was used to quantify the abundance of ZAM11 versus UW101. The ZAM11/UW101 ratio increased or decreased slowly as a function of time, depending on the particular replicate and time window. In six of the eight experiments, ZAM11 predominated by day 36. Hence, this mutation, which effectively removes prfB autoregulation, confers no obvious loss of fitness, at least under these laboratory conditions.