Abstract

   Paramount to human health, symbiotic bacteria in the gastrointestinal
   tract rely on the breakdown of complex polysaccharides to thrive in
   this sugar-deprived environment. Gut Bacteroides are metabolic
   generalists and deploy dozens of polysaccharide utilization loci (PULs)
   to forage diverse dietary and host-derived glycans. The expression of
   the multi-protein PUL complexes is tightly regulated at the
   transcriptional level. However, how PULs are orchestrated at
   translational level in response to the fluctuating levels of their
   cognate substrates is unknown. Here, we identify the RNA-binding
   protein RbpB and a family of noncoding RNAs as key players in
   post-transcriptional PUL regulation. We demonstrate that RbpB interacts
   with numerous cellular transcripts, including a paralogous noncoding
   RNA family comprised of 14 members, the FopS (family of paralogous
   sRNAs). Through a series of in-vitro and in-vivo assays, we reveal that
   FopS sRNAs repress the translation of SusC-like glycan transporters
   when substrates are limited—an effect antagonized by RbpB. Ablation of
   RbpB in Bacteroides thetaiotaomicron compromises colonization in the
   mouse gut in a diet-dependent manner. Together, this study adds to our
   understanding of RNA-coordinated metabolic control as an important
   factor contributing to the in-vivo fitness of predominant microbiota
   species in dynamic nutrient landscapes.

   Subject terms: Microbiology, Biochemistry
     __________________________________________________________________

   Gut Bacteroides deploy several polysaccharide utilization loci (PULs)
   to forage diverse dietary and host-derived glycans. Here, the authors
   identify the RNA-binding protein RbpB and a family of noncoding RNAs as
   key players in post-transcriptional PUL regulation, further showing
   that ablation of RbpB in Bacteroides thetaiotaomicron compromises
   colonization in the mouse gut in a diet-dependent manner.

Introduction

   The obligate anaerobic, Gram-negative Bacteroidales represent a
   dominant order of the gut microbiota^[44]1,[45]2 and influence human
   health in various ways^[46]3. An important factor contributing to the
   success of these bacteria in colonizing the mammalian bowel is their
   immense metabolic capacities encoded on distinct polysaccharide
   utilization loci (PULs)^[47]4. PULs enable the catabolism of complex
   dietary polysaccharides and host mucus-derived glycans^[48]5. Each PUL
   encodes sets of membrane-spanning, glycolytic multi-protein complexes,
   which typically bind the respective substrate at the cell surface and
   cleave it into oligosaccharides. These oligosaccharides are imported
   into the periplasm through TonB-dependent transporters of the SusC
   family^[49]6, and further processed into simple sugars. As individual
   PULs are substrate-specific, their expression is tightly controlled and
   responds to the nutritional fluctuations that are commonly associated
   with the dynamic gut environment. Therefore, PULs often include
   dedicated regulatory factors that spur transcription of their specific
   PUL operons when the corresponding carbon sources are sensed. This
   includes SusR-like regulators^[50]7 and hybrid two-component
   systems^[51]8,[52]9 that combine both sugar-sensing and gene regulatory
   functions in a single polypeptide. Alternatively, transcription of PUL
   systems specific to the processing of host-derived glycans is typically
   governed by extracytoplasmic function sigma/anti-sigma factor
   pairs^[53]10. However, if and how Bacteroidales prevent ongoing
   translation of pre-existing PUL mRNAs once the inducing stimulus fades,
   is not currently known.

   Only recently did RNA-seq studies from us^[54]11,[55]12 and
   others^[56]13,[57]14 map the transcriptome of select Bacteroidales
   members at single-nucleotide resolution. This entailed the
   identification of hundreds of noncoding RNA candidates, including the
   small noncoding RNAs (sRNAs) DonS and GibS. DonS is divergently encoded
   to an N-glycan-specific PUL and was shown to repress the expression of
   this PUL in Bacteroides fragilis^[58]13, albeit through an unknown
   mechanism. The conserved sRNA GibS is induced in the presence of
   N-glycans, and binds and represses a glycoside hydrolase-encoding mRNA
   in Bacteroides thetaiotaomicron^[59]11. Given these examples and the
   general importance of RNA-based regulation, it is likely that
   Bacteroides spp. employ much more complex sRNA-based
   post-transcriptional regulatory networks to optimize fitness in
   response to dynamic nutrient levels. However, the identification of
   post-transcriptional regulatory circuits in these bacterial taxa that
   lack homologs of classical RNA chaperones has proven challenging.

   The present study reports the identification of an in-vivo phenotype,
   the RNA interactome, a metabolism-associated function, and the
   underlying molecular mechanism of the first global RNA-binding protein
   (RBP) in Bacteroides. Specifically, we found that a B. thetaiotaomicron
   mutant devoid of the RNA recognition motif 1 (RRM-1)-containing protein
   RbpB^[60]15,[61]16 failed to efficiently colonize the mammalian
   intestine in a diet-dependent manner. Cross-linking immunoprecipitation
   and sequencing (CLIP-seq) of B. thetaiotaomicron RbpB demonstrated that
   this protein is a global RNA binder. Analysis of the RbpB interactome
   led to the discovery of a complex RNA network, with a highly conserved
   multicopy sRNA family at its center: the ‘FopS’ cluster (for ‘family of
   paralogous sRNAs’). While the 14 FopS members of B. thetaiotaomicron
   exhibit partial sequence similarity to the previously characterized
   GibS sRNA^[62]11, we provide evidence of functional diversification
   between GibS and FopS. The FopS sRNAs function as post-transcriptional
   repressors of specific PUL operons by binding to the vicinity of the
   start codon of the first mRNA of target PUL polycistrons, typically
   corresponding to the cognate SusC homolog, impeding translation of the
   respective glycan transporter. As demonstrated for the high-mannose
   N-glycan-catabolizing PUL72, RbpB counteracts FopS activity, relieving
   translational repression. Together, this study reports a remarkably
   complex post-transcriptional control network that allows Bacteroides to
   switch off specific PULs in the absence of the cognate substrate or in
   the presence of an alternative, preferred carbohydrate source to
   optimize fitness.

Results

RbpB promotes mucus-foraging by B. thetaiotaomicron in the mammalian
intestine

   Bacteroides spp. encode an array of transcription factors that govern
   carbohydrate utilization in these gut
   bacteria^[63]9,[64]10,[65]17–[66]21. Conversely, there is currently
   little knowledge as to the extent to which post-transcriptional control
   impacts on Bacteroides metabolic competitiveness. Given that the
   deletion of the RbpB-encoding gene (BT_1887) in B. thetaiotaomicron led
   to the differential expression of several PUL genes^[67]16, we assessed
   the fitness of a B. thetaiotaomicron ∆rbpB mutant in the mammalian gut
   in response to distinct diets. As conventionally raised mice are
   resistant to B. thetaiotaomicron colonization^[68]22, we treated
   C57BL/6 mice with an antibiotic cocktail to promote B. thetaiotaomicron
   engraftment. We then inoculated mice fed a conventional diet with an
   equal ratio of the B. thetaiotaomicron wild-type strain and an isogenic
   ∆rbpB mutant, followed by determination of the abundance of each strain
   in cecal and colonic contents 6 days post-inoculation by plating on
   selective media (Fig. [69]1a).

Fig. 1. RbpB promotes mucus-foraging in the mammalian intestine.

   [70]Fig. 1
   [71]Open in a new tab

   a Schematic representation of the experimental outline. One week before
   antibiotic treatment, C57BL/6 J mice were switched to a low fiber diet
   or remained on the fiber-rich control diet until the end of the
   experiment. Antibiotic cocktails were administered by oral gavage daily
   for 5 days. Mice were inoculated with an equal mixture of 0.5 × 10^9
   CFU of the B. thetaiotaomicron wild-type and rbpB mutant strains. After
   6 days, cecal and colonic tissue was collected and the bacterial
   numbers determined by plating on selective agar. b The competitive
   index of B. thetaiotaomicron wild-type versus ΔrbpB or rbpB^+ in the
   cecal (left) or colonic (right) contents. The data are combined over
   two independent experiments, each comprised of four biological
   replicates (three biological replicates for WT/rbpB^+ competition).
   Black horizontal lines represent the geometric means. *p = 0.0245,
   degrees of freedom (df) = 14, 95% confidence interval (CI):
   1.169–14.56; ***p = 0.0007, df = 14, 95% CI: 5.268-15.67. c Same
   colonic data as in (b), separated by sex (female mice on the left, and
   male mice on the right). ****p < 0.0001, df = 6, 95% CI: 11.14–18.89;
   n.s., not statically significant based on two-tailed Student’s t-tests.
   Source data are provided as a Source Data file.

   Notably, the wild-type was outcompeted by the ∆rbpB mutant in the
   fiber-rich control diet (black dots in Fig. [72]1b). Remarkably,
   however, the ∆rbpB mutant displayed a significant fitness disadvantage
   compared to the wild-type strain when switched to a low fiber diet,
   which drives the bacteria to feed on host-derived glycans (red dots in
   Fig. [73]1b). Consistent with the notion that the colon has a thicker
   mucous layer than the cecum, the fitness defect was more pronounced in
   the colon. This phenotype was specific to RbpB as trans-complementation
   of the gene into the ∆rbpB background reverted the in-vivo fitness to
   that of the wild-type strain (triangles in Fig. [74]1b). Moreover, we
   observed a sex-specific difference, in which the fitness defect of the
   ∆rbpB mutant was more prominent in males than in females (Fig. [75]1c).
   The magnitude of this attenuation exceeded the relatively mild in-vivo
   phenotypes associated with individual PUL-encoded transcriptional
   regulators (e.g. refs. ^[76]23,[77]24) and rivaled the large impact of
   deleting the transcriptional master regulator of polysaccharide
   utilization, Cur^[78]25.

   A 16S ribosomal RNA gene analysis of the fecal samples at the end of
   the experiment revealed significant differences in the resident
   microbiota composition between the diet groups in terms of taxa
   (Supplementary Fig. [79]1a), beta diversity (Supplementary Fig. [80]1b,
   c), and alpha diversity (Supplementary Fig. [81]1d, e). While we cannot
   rule out the influence of microbe-microbe interactions on B.
   thetaiotaomicron colonization, the competitive setup ensured the two
   experimentally introduced strains to experience the same nutritional
   environment and interactions with other members of the microbiota
   within individual mice. The difference in relative fitness, therefore,
   suggests a role of RbpB in promoting the mucus-foraging lifestyle of B.
   thetaiotaomicron, prompting us to investigate the cellular function of
   this protein in more detail.

RbpB acts as a global RNA-binding protein in B. thetaiotaomicron

   RbpB was reported to possess the ability to bind synthetic
   single-stranded RNA ‘pentaprobes’ in vitro^[82]16. However, whether
   Bacteroides RbpB acts as a global RNA binder in vivo—as opposed to a
   test tube—was previously not assessed. To explore the in-vivo functions
   of RbpB, we constructed a C-terminally FLAG-tagged version of RbpB and
   stably integrated it into the chromosome of B. thetaiotaomicron.
   In-vivo crosslinking and immunoprecipitation (CLIP) experiments
   (Supplementary Fig. [83]2a) revealed the characteristic
   crosslink-induced signal on an autoradiograph (Fig. [84]2a), which was
   sensitive to RNase I—but not to DNase I—treatment (Supplementary
   Fig. [85]2b), suggesting the protein interacts primarily with cellular
   RNA.

Fig. 2. RbpB acts as a global RNA-binding protein in B. thetaiotaomicron.

   [86]Fig. 2
   [87]Open in a new tab

   a Autoradiogram of radioactively labeled RNAs covalently bound to
   RbpB-3xFLAG after in-vivo UV crosslinking, immunoprecipitation, and
   transfer to a nitrocellulose membrane. Shown is one representative
   image of two independent replicate experiments. b Genomic positions of
   the identified RbpB CLIP peaks in the context of the B.
   thetaiotaomicron chromosome and plasmid. The positions of coding
   sequences (CDS) and noncoding RNA (ncRNA) genes were retrieved from
   NCBI and published literature^[88]11. The gibS gene and the 14
   FopS-encoding loci (which will become relevant below) are marked. The
   rbpB mRNA itself harbors a RbpB binding site, implying the potential
   for autoregulation. Inner circles refer to the minus strand and outer
   circles to the plus strand. c RNA class distribution of RbpB ligands.
   Numbers in brackets refer to statistically significant and manually
   confirmed peaks supported by two independent CLIP-seq experiments. Note
   that the sum of peaks mapping to the different genetic features exceeds
   the actual total peak number as some peaks mapped to overlapping
   annotations (such as 3′ UTRs and 3’-derived sRNAs). d Peak size
   distribution. Plotted are the frequencies for individual peak sizes of
   all unique peaks (yellow) and sRNA peaks (green). e Enriched sequence
   motif within RbpB sRNA ligands. Source data are provided as a Source
   Data file.

   To map the RbpB interactome at genomic scale, we subjected the
   co-purified RNA to high-throughput sequencing (CLIP-seq). Two
   independently performed replicate experiments showed clear read
   enrichments in crosslinked samples as compared to matched background
   controls (Supplementary Fig. [89]2c, d). We report a total of 213
   significantly enriched (p[adj] ≤ 0.05; log[2] FC > 2), manually
   confirmed peaks within 173 different mRNAs (12 peaks in 5’ UTRs, 175 in
   CDSs, 26 in 3’ UTRs), 64 peaks in 57 different sRNAs (two of which
   overlap with 3’ UTR peaks and three with 5’ UTR peaks, and one or two
   peaks, respectively, map to the 6S and 4.5S housekeeping RNAs), and 14
   peaks within intergenic regions (Fig. [90]2b, c; Supplementary
   Data [91]1). We integrated the RbpB binding site annotation in the
   open-access online resource ‘Theta-Base’
   ([92]https://bacteroides.helmholtz-hzi.de/jbrowse_new/)^[93]12 to
   facilitate interrogation by the public.

   Inspecting the peak size distribution revealed two overrepresented
   footprint lengths (Fig. [94]2d), reflecting the band pattern on the
   autoradiogram (Fig. [95]2a; Supplementary Fig. [96]2b). While a shorter
   peak size of ~30 nt was predominantly associated with mRNAs and
   intergenic regions, peaks within sRNAs exceeded 50 nt (Fig. [97]2d).
   Functional analysis of RbpB-bound mRNAs revealed an enrichment of peaks
   in transcripts encoding proteins involved in translational processes,
   such as the biosynthesis of aminoacyl-tRNAs and ribosomal proteins
   (Supplementary Fig. [98]2e), implying a translation-related role of
   this RBP.

   In Gram-negative species, global RBPs often stabilize their RNA ligands
   by shielding them from cellular nucleases^[99]26–[100]29. To test for a
   similar role of Bacteroides RbpB, we coupled rifampicin-mediated
   inhibition of de-novo transcription to the measurement of RNA decay
   kinetics^[101]30. This analysis suggested RbpB overexpression
   (Supplementary Fig. [102]3a) to reduce the cellular half-lives of some
   of the top-enriched mRNA ligands, whereas its deletion entailed more
   subtle effects (Supplementary Fig. [103]3b). Closer inspection of
   CLIP-seq peaks suggested the observed mRNA destabilizations upon RbpB
   overexpression to be independent of the relative position of the
   binding sites within those transcripts (Supplementary Fig. [104]3d).
   RbpB-associated sRNAs were generally very stable, regardless of the
   presence or absence of RbpB (Supplementary Fig. [105]3c). It therefore
   appears that the primary role of RbpB is not to protect its RNA ligands
   from degradation.

   With respect to RbpB-bound noncoding RNAs, drawing from available
   secondary structure information^[106]15, manual inspection revealed the
   protein to bind preferentially within single-stranded regions
   (Supplementary Fig. [107]2f). MEME analysis^[108]31 identified several
   primary sequence motifs enriched in RbpB peaks (Supplementary
   Fig. [109]4a), with the most significant comprising a 41 nt-long motif
   (Fig. [110]2e). In fact, many of the most strongly enriched sRNA
   ligands of RbpB contained this sequence around the RBP footprint
   (Supplementary Fig. [111]4b). Given this observation, we focused on
   this sequence motif and the sRNAs containing it.

A conserved multicopy sRNA family is associated with RbpB

   We next set out to determine the prevalence of the above sequence
   motif. A hidden Markov model-based iterative sequence homology search
   returned 14 paralogous sequences within the B. thetaiotaomicron
   VPI-5482 genome (Fig. [112]3a). They all fell within annotated sRNAs,
   which were significantly enriched in the RbpB CLIP-seq dataset. A BLAST
   search revealed this multicopy sRNA family to be highly conserved
   within Bacteroidaceae (Supplementary Fig. [113]5a), ranging from 12
   copies (in Bacteroides caccae and Bacteroides uniformis) to 15 copies
   per genome (in Bacteroides xylanisolvens) (Supplementary Fig. [114]5b).
   Given the presence of the sequence in a Caudovirales phage
   (Supplementary Fig. [115]5a)—predicted to target Bacteroides based on
   matching CRISPR spacers^[116]32—bacteriophages may have played a role
   in disseminating the corresponding sRNAs within and across
   Bacteroidaceae genomes. We term the cluster “family of paralogous
   sRNAs” (FopS).

Fig. 3. A fourteen-member family of paralogous sRNAs associates with RbpB.

   [117]Fig. 3
   [118]Open in a new tab

   a Sequence alignment of R1-containing sRNA genes from B.
   thetaiotaomicron VPI-5284. Red and blue letters indicate highly
   conserved and less-conserved ribonucleobases, respectively. Numbers
   denote the positions relative to the 5’ end of GibS (bent arrow).
   Canonical σ^ABfr promoters^[119]38 (green boxes), promoter motif
   2^[120]11 (‘PM2’; pale red boxes), and the R1 and R2 seeds (as
   determined in ref. ^[121]11) are indicated. R1’ refers to the
   3’-extended seed of FopS sRNAs (see Fig. [122]4e). Curated sRNA
   boundaries are marked by gray triangles (see Methods for details).
   Sequence identity of the 14 individual FopS paralogs with GibS ranges
   between 28.4-43.9%. b FopS consensus structure predicted using the
   Webserver for Aligning structural RNAs^[123]33. Secondary structure was
   visualized using the R2R software^[124]93. R1’ sequence is boxed. c In
   vitro-transcribed, radioactively labeled FopS-10 was incubated for 1 h
   with two defined concentrations of recombinant RbpB (‘+’: 10 µM; ‘++’:
   20 µM) and the resulting complexes resolved on denaturing gels. ‘WT’
   refers to the wild-type sRNA sequence and ‘Inv’ to a FopS-10 variant
   with an inverted RbpB-target sequence (see Supplementary Fig. [125]6c).
   d, e Expression profiling of GibS and FopS sRNAs. Relative expression
   of the indicated sRNAs over a range of carbon sources (relative to
   growth in minimal glucose-containing medium) and stress conditions
   (relative to an unstressed control) based on RNA-seq data from^[126]12
   (d) and northern blot-based validation (e). EEP, early exponential
   phase; MEP, mid-exponential phase; stat, stationary phase in TYG. The
   annealing site of the FopS-specific northern probe is indicated in
   panel a. The upper band corresponds to the primary transcript of 3’
   UTR-derived FopS-14. Positions of the marker bands are shown to the
   left. 5S rRNA served as loading control. Images representative of two
   (c) and three (e) independent replicates, respectively. Source data are
   provided as a Source Data file.

   The GibS sRNA was also pulled down together with RbpB (Fig. [127]2b).
   Previously, we had shown that this sRNA comprises two distinct seed
   regions, referred to as R1 and R2, which mediate basepairing with—and
   repression of—its two direct target mRNAs^[128]11. Interestingly, the
   FopS family members contain the R1 sequence of GibS but lack R2
   (Fig. [129]3a). We adopted a structural RNA alignment tool^[130]33 to
   infer the consensus secondary structure of the 14 FopS sRNAs of B.
   thetaiotaomicron. This structure comprised a single-stranded 5’ stretch
   followed by a Rho-independent terminator hairpin (Fig. [131]3b), and
   was deposited in the Rfam database^[132]34. The conserved R1 sequence
   is located within the predicted single-stranded region at the 5’ end of
   the FopS consensus structure, as would be expected if R1 acts as a seed
   sequence in these sRNAs.

   Electrophoretic mobility shift assay (EMSA) with recombinant B.
   thetaiotaomicron RbpB (Supplementary Fig. [133]6a) and—as a
   representative sRNA family member—in-vitro-transcribed FopS-10
   supported the formation of stable ribonucleoprotein (RNP) particles
   (Fig. [134]3c; Supplementary Fig. [135]6b). Inverting the sequence of
   the 55 nt-long RbpB binding site (as deduced from CLIP-seq) predicted
   to largely maintain the secondary structure of FopS-10 (Supplementary
   Fig. [136]6c) impeded the formation of RNP complexes (‘Inv’ in
   Fig. [137]3c; Supplementary Fig. [138]6b). EMSAs also validated that
   RbpB binds to GibS (Supplementary Fig. [139]6d). In the case of both
   tested sRNAs, the affinity of the protein to its ligand was in the low
   micromolar range, and thus an order of magnitude lower than what is
   typically observed in interactions of pseudomonadotal Hfq and ProQ with
   their cognate sRNA partners^[140]35–[141]37. Taken altogether, these
   data support the CLIP-seq results and confirm that RbpB binds to FopS
   and GibS sRNAs in a sequence-specific and concentration-dependent
   manner.

FopS expression responds to bile salts

   Oftentimes, expression profiling of a given sRNA provides a glimpse
   into its function. While GibS transcription is driven from a
   non-canonical promoter associated with stationary phase-induced
   genes^[142]11, inspection of the regions upstream of the fopS genes
   revealed the presence of the canonical σ^ABfr promoter^[143]38 (red or
   green boxes, respectively, in Fig. [144]3a). In accordance with an
   independent transcriptional activation, published RNA-seq data^[145]12
   showed an anticorrelated expression pattern of GibS and the FopS sRNAs
   (Fig. [146]3d). As confirmed by northern blotting (Fig. [147]3e), the
   steady-state level of GibS increased during growth in rich TYG medium
   up to stationary phase, but was highest when bacteria fed on
   N-Acetylglucosamine (GlcNAc)-containing carbohydrates (as previously
   reported^[148]11), particularly when mucin was the sole carbon source.
   In contrast, GibS levels dropped when bacteria were exposed to bile
   salts. Under those conditions, B. thetaiotaomicron induced expression
   of the FopS sRNAs, as reflected by two, relatively distinct bands that
   matched the predicted length of the FopS sRNAs (two clusters of ~65 and
   ~70 nt) on a northern blot (Fig. [149]3e) using a probe against the
   universal FopS 5’ region (Fig. [150]3a). An additional, high molecular
   band of ~1,400 nt was most likely derived from BT_4331, i.e. the
   parental transcript of the only 3’ end-derived member of this sRNA
   family, FopS-14 (Supplementary Fig. [151]6e). Concentration range
   experiments coupled to northern blotting revealed FopS expression to
   peak between 0.03 and 0.1 mg/mL of bile salts (Supplementary
   Fig. [152]6f, g), falling within the range of their in-vivo
   concentrations in the human large intestine^[153]39,[154]40.
   Responsiveness to this in-vivo-relevant stimulus corroborates the
   relevance of the identified sRNA-RBP cluster for B. thetaiotaomicron
   within its host niche.

Established GibS targets are refractory to FopS-mediated regulation

   We next sought to functionally characterize the FopS sRNA family.
   Previously we found that GibS represses BT_0771 and BT_3893, which code
   for a glucan-branching enzyme that belongs to the glycoside hydrolase
   13 family ([155]http://www.cazy.org/) and a hypothetical protein,
   respectively^[156]11. Mechanistically, GibS binds the translation
   initiation regions of BT_0771 and BT_3893 mRNAs, the latter of which by
   engaging both the R1 and R2 seeds^[157]11. To assess if these GibS
   targets are also regulated by the FopS cluster, we compared their
   steady-state transcript levels between wild-type and three independent
   fopS deletion mutants (∆fopS-09, ∆fopS-10, ∆fopS-13; i.e., sRNAs, whose
   5’ end matches the FopS consensus, yet reflect downstream sequence
   diversity at positions 35-60). In stationary phase TYG cultures,
   endogenous GibS and FopS sRNAs were relatively highly expressed
   (Fig. [158]3e). However, both established GibS target mRNAs were
   derepressed only in the absence of GibS and not in the absence of
   individual FopS sRNAs (Fig. [159]4a). Consistent with the notion that
   GibS is the primary regulator of these mRNAs, in-vitro-transcribed 5’
   ends of BT_3893 and BT_0771 mRNAs annealed with radiolabeled GibS, but
   not (BT_3893) or less efficiently (BT_0771) with FopS-10 (Supplementary
   Fig. [160]7a). Therefore, although functional redundancy amongst the
   paralogous sRNAs might partially compensate for effects derived from
   single fopS deletion, it appears from EMSAs that FopS sRNAs are
   relatively ineffective in regulating GibS targets, despite possessing
   one of the two seed regions of GibS.

Fig. 4. FopS and RbpB constitute a post-transcriptional layer of PUL control.

   [161]Fig. 4
   [162]Open in a new tab

   a qRT-PCR analysis of direct GibS targets^[163]11 in stationary phase
   in TYG. Bars denote the mean from six replicate measurements.
   Significance was assessed using two-way ANOVA (Sidak’s multiple
   comparisons test; shown are significant changes [p < 0.05] relative to
   wild-type [dashed line]). Target locus representations at the top, with
   experimentally mapped GibS binding sites^[164]11 in red. b Pearson’s
   correlation coefficients (r) for individual sRNA-mRNA pairs based on
   RNA-seq data^[165]12. As an example, the (anti-)correlation of BT_3983
   with FopS and other predicted targets is boxed. c qRT-PCR measurement
   of susCD^BT_3983-84, analogous to panel (a), except that strains were
   exposed for 2 h to 0.5 mg/mL of bile salts. Locus representation on
   top, with FopS binding site (determined in panels e, f) in green. d
   Left: growth kinetics of a multi-fopS mutant (‘∆∆+gRNA3’) and its
   parental wild-type on either glucose or Man[9]GlcNAc[2] (mean of two
   biological replicates). Right: expression of selected sRNAs and mRNAs
   at 48 h of growth (relative to wild-type in glucose). For underlying
   expression data, see Supplementary Fig. [166]8d-f. e RNA:RNA
   interaction predictions. Start codons are red and gray shadings denote
   coding regions. R1, R2: seed regions of GibS^[167]11; R1’: 3’-extended
   seed of the FopS’s. Coordinates relative to 5’ ends. f In-line probing
   of 0.2 pmol ^32P-labeled susC^BT_3983 in absence or presence of either
   20 nM or 200 nM GibS, FopS-10, or both sRNAs, and 20 µM of recombinant
   RbpB. ‘Control’ and ‘–‘ refer to the radiolabeled RNA substrate in
   water or inline buffer, respectively. Partially RNase T1- (‘T1’) or
   alkali-digested (‘OH’) substrates were included as ladders. g In-vitro
   translation of 5’ susC^BT_3983-gfp-3xFLAG or gfp-3xFLAG (control). 1 µM
   (lanes 1–4) or 0.5 µM (lanes 5–10) of mRNA was in vitro-translated with
   reconstituted E. coli 70S ribosomes, with or without 20 µM purified
   RbpB and 25 µM in-vitro-transcribed FopS-10. Translation products were
   detected via western blotting and Ponceau staining served as loading
   control. Images representative of two (f) or three (g) independent
   replicates. Source data are provided as a Source Data file.

FopS sRNAs repress PUL72 with phenotypic consequences during growth on
high-mannose N-glycans

   To identify the bona fide function of the FopS's, we applied the
   IntaRNA algorithm^[168]41 using the consensus FopS 5’ sequence
   (position 1 to 21) as query (Supplementary Data [169]2). The proposed
   target candidates were strongly enriched in mRNAs encoding outer
   membrane proteins, particularly SusC-like transporters of PUL systems.
   Interrogation of available RNA-seq data^[170]12 revealed a strong
   anticorrelation between the expression of individual FopS sRNAs and
   that of their in-silico-predicted, common targets (Fig. [171]4b),
   implying negative regulation to prevail amongst putative FopS-mediated
   activities.

   For further characterization of FopS-mediated target control, we
   focused on PUL72, a functionally and structurally characterized PUL
   specific for mammalian high-mannose-type
   N-glycans^[172]20,[173]42,[174]43. IntaRNA predictions were
   corroborated by the steady-state transcript levels of BT_3983, encoding
   the SusC homolog of this PUL. In line with a repressive effect of the
   FopS sRNAs on this SusC-like transporter, its mRNA level was elevated
   in individual ∆fopS strains during growth under bile stress, and
   returned to basal level when the respective fopS was complemented in
   trans (Fig. [175]4c). The genes of the PUL72-encoded surface
   glycan-binding SusD homolog (BT_3984) and, to lesser extent, of
   BT_3985-88 (glycoside hydrolase genes) that are co-transcribed with the
   susC homolog, showed a similar FopS-dependent expression profile
   (Fig. [176]4c; Supplementary Fig. [177]7b), suggesting that FopS
   binding to the 5’ region of susC^BT_3983 also represses the downstream
   genes in the corresponding operons. The expression of other PUL72
   genes, encoded from the opposing DNA strand, was hardly altered by FopS
   depletion and deletion of gibS did not affect the steady-state levels
   of any of these transcripts (Supplementary Fig. [178]7b).

   Intending to inactivate multiple FopS family members simultaneously, we
   generated a ∆∆fopS-09/10 double-knockout mutant. However, attempts to
   delete additional fopS genes in this strain background failed after
   repeated attempts. We therefore opted for a CRISPR interference
   (CRISPRi)-based approach^[179]44 to transcriptionally silence the
   remaining family members. Evaluating the knockdown efficiencies of
   three different guide RNAs (gRNAs) targeting the universal sequence at
   the FopS 5’ end (Supplementary Fig. [180]8a), we identified one strain
   (expressing gRNA #3; termed “∆∆+gRNA3”) in which FopS aggregate levels
   reduced to ~30% upon inducing the CRISPRi system (Supplementary
   Fig. [181]8b, c). When grown in minimal medium containing the PUL72
   substrate high-mannose N-glycan as the sole carbon source, FopS
   expression was ~3-fold reduced as compared to growth on glucose
   (Fig. [182]4d, right; Supplementary Fig. [183]8e). Importantly, the
   ∆∆+gRNA3 strain exhibited enhanced growth on this carbon source
   relative to a wild-type culture (Fig. [184]4d, left). This is likely a
   result of elevated PUL72 activity in the mutant background, as the
   transcript levels of the PUL72-associated SusCD homologs increased two-
   to three-fold in the ∆∆+gRNA3 compared to the wild-type strain
   (Fig. [185]4d, right; Supplementary Fig. [186]8f). Collectively, these
   data therefore suggest that the FopS sRNAs—but not GibS—repress PUL72
   and this regulation entails phenotypic consequences when B.
   thetaiotaomicron consumes high-mannose N-glycans.

A panel of PUL polycistrons are targeted by FopS at their 5’ end

   In-silico prediction of RNA-RNA interaction suggested that the FopS 5’
   end basepaired with the translation initiation region of susC^BT_3983
   (Fig. [187]4e). This involves a 3’-extended version of the R1 region
   within FopS (that we term R1’)—a sequence that includes several
   mismatches in GibS (Fig. [188]3a), which might explain the divergent
   outcomes of deleting these sRNAs on levels of the susC^BT_3983 mRNA
   (Fig. [189]4c). Indeed, inline probing of the radiolabeled 5’ region of
   a susC^BT_3983 mRNA fragment (between the 5’ end and the 69^th codon of
   the coding sequence) upon incubation with increasing concentrations of
   in-vitro-transcribed GibS or FopS-10 showed the predicted targeting
   site is shielded by FopS-10, but not by GibS (Fig. [190]4f; compare
   lanes 10 and 11, or 6 and 7, respectively, with lane 4). EMSA likewise
   confirmed binding of FopS-10 to the 5’ region of susC^BT_3983 mRNA,
   whereas the addition of up to 1 µM of GibS did not result in an upshift
   of the radioactively labeled mRNA fragment (Supplementary
   Fig. [191]7c). Point-mutating eight nucleotides within the predicted
   FopS-10 targeting region of susC^BT_3983 mRNA (‘Mut’) fully abrogated
   sRNA binding in vitro and introduction of the corresponding
   compensatory mutations into FopS-10 restored binding to the mutant
   susC^BT_3983 variant (Supplementary Fig. [192]7d), thus supporting the
   in-silico-predicted interaction site (Fig. [193]4e).

   Finally, EMSAs confirmed the predicted interaction of FopS-10 with
   eight additional target candidates in vitro, including the 5’ regions
   of mRNAs of five additional SusC homologs from various PULs (PUL12,
   -51, -77, -80, -88) and the mRNAs of three glycoside hydrolases, two of
   which are also encoded within PULs (Supplementary Fig. [194]7e). These
   latter PULs (PUL20 and -44) exhibit an atypical genomic architecture,
   wherein the respective FopS target glycoside hydrolase (rather than the
   cognate SusC homolog) is encoded by the first operon
   gene^[195]4,[196]12. Taken together, these data suggest the FopS sRNA
   family to repress multiple, mostly host glycan-catabolic^[197]10 PUL
   systems by direct basepairing to the translation initiation region at
   the 5’ end of the respective polycistronic mRNA.

FopS-10 suppresses translation initiation of susC^BT_3983 and is antagonized
by RbpB

   Current paradigm holds that RNA-binding proteins can repress target
   mRNA translation by facilitating sRNA binding, in which the specificity
   of the binding is determined by the basepairing between the sRNA and
   the target. To investigate whether RbpB assumes similar roles, we
   performed three-component EMSAs. Compared to the affinities of GibS and
   FopS-10 to their respective targets in the absence of RbpB (black
   curves in Supplementary Fig. [198]9a), supplementation of a fixed
   concentration (1 µM) of the protein did not affect sRNA-mRNA duplex
   formation efficiency in vitro (yellow curves in Supplementary
   Fig. [199]9a). Gradually increasing the RbpB concentration to up to
   20 µM suggested the formation of a ternary complex consisting of GibS,
   its target mRNA (BT_3893), and RbpB, whereas the protein was not
   engaged in a stable ternary complex with FopS-10 and most of its mRNA
   targets (except for BT_2970) under the same conditions (Supplementary
   Fig. [200]9b, c, d). Instead, high concentrations of the protein
   ( > 5 µM) titrated FopS-10 away from its targets in vitro
   (Supplementary Fig. [201]9c, d). Likewise, overexpression of the
   protein in vivo resulted in increased, and rbpB deletion in reduced
   steady-state levels of the PUL72 susCD pair (Fig. [202]4c), suggesting
   that RbpB counteracts FopS-mediated target repression.

   EMSAs also revealed that the susC^BT_3983 and BT_3893 mRNAs themselves
   are ligands of RbpB—even in the respective absence of FopS or GibS
   (Supplementary Fig. [203]9b, c). Note that none of these genes was
   expressed during mid-exponential growth in rich medium^[204]11,
   providing a plausible explanation why the corresponding mRNAs were not
   recovered in the CLIP-seq experiment (Fig. [205]2). This
   sRNA-independent interaction was further supported by inline probing,
   revealing substantial structural rearrangements in the 5’ UTR of
   susC^BT_3983 in the presence of RbpB (compare lanes 4 and 5 in
   Fig. [206]4f). Inline probing of the interaction between RbpB and this
   5’ fragment in the additional presence of FopS-10 uncovered one further
   structural change (a band present only in lanes 12 and 13 in
   Fig. [207]4f). This rearrangement occurred at position 72 relative to
   the 5’ end, which coincides with position –6 with respect to the start
   codon of the susC^BT_3983 mRNA (see Supplementary Fig. [208]10a for a
   model of the structural rearrangements) and thus, falls within the
   critical window for efficient translation initiation in the
   Bacteroidota^[209]45.

   In-vitro translation of a susC^BT_3983-gfp-3xFLAG mRNA template was
   unaffected by the addition of recombinant RbpB alone (Fig. [210]4g,
   lanes 4 vs. 3). In contrast, addition of FopS-10 to the reaction mix
   abrogated target protein synthesis (Fig. [211]4g, lanes 9 vs. 8). This
   effect was sequence-specific as evidenced by the point-mutated FopS-10
   variant (see Supplementary Fig. [212]7d), which failed to repress the
   translation of wild-type susC^BT_3983, yet effectively inhibited that
   of a template harboring the respective compensatory mutations
   (Supplementary Fig. [213]10b). In the additional presence of
   recombinant RbpB, FopS-10-mediated translational repression was
   relieved (Fig. [214]4g, lane 10). Translation of a control template,
   consisting of an Escherichia coli ribosome binding site fused to the 5’
   region of the GFP open reading frame and a triple-FLAG tag, was largely
   unchanged in analogously prepared reactions (Fig. [215]4g, lanes 1, 2,
   5–7). Based on these data, we propose that FopS-10 sRNA blocks
   translation of the susC^BT_3983 mRNA by binding adjacent to the start
   codon and providing steric hindrance to the initiating ribosome. This
   repression is counteracted by RbpB, possibly by titrating FopS-10 away
   from its target mRNA and/or by opening up the translational enhancer
   region of susC^BT_3983 that is otherwise occluded by FopS-10
   (Supplementary Fig. [216]10a). Altogether, our mechanistic data suggest
   that, in contrast to well-characterized RNA-binding proteins, RbpB
   assumes a role in excluding the interaction of sRNAs with a range of
   target transcripts, serving as a post-transcriptional hub in
   Bacteroides metabolic control.

Discussion

   The composition of the gut microbial community and, therefore, the
   pivotal functions these microbes provide to human health hinges on
   their ability to persist in the face of nutritional fluctuation.
   Understanding how gut commensals adapt their metabolism to the daily
   variations in feeding rhythm and types of diet has thus become an
   important branch of microbiota research^[217]46,[218]47. As one of the
   predominant bacterial genera in the healthy human gut microbiota,
   Bacteroides possess dozens of multi-protein complexes encoded on
   dedicated genomic loci—the PULs—to bind, clip, and import specific
   polysaccharides. Historically, PUL function was first studied in B.
   thetaiotaomicron^[219]48,[220]49, serving as a paradigm for
   polysaccharide breakdown in the human microbiome, with practical
   applications^[221]50. Complementing previous studies that focused on
   transcriptional control
   mechanisms^[222]7–[223]10,[224]21,[225]51,[226]52, the present work
   revealed a remarkably complex RNA-based regulatory circuit governing
   PUL regulation in B. thetaiotaomicron. At the heart of this network are
   the RRM-1 protein RbpB and a conserved family of paralogous sRNAs that
   optimize fitness by orchestrating mutually exclusive expression of
   opposing catabolic processes (Fig. [227]5). Given the conservation of
   RRM-1 proteins^[228]16 and R1 sequence-containing sRNAs (Supplementary
   Fig. [229]5a) across the Bacteroidaceae, this regulatory circuit is
   likely prevalent in a substantial fraction of mammalian intestinal
   microbiota members.

Fig. 5. Working model for the functional diversification between
RbpB-associated sRNAs—GibS and the FopS sRNA cluster—in B. thetaiotaomicron
metabolism control.

   [230]Fig. 5
   [231]Open in a new tab

   a GibS transcription is activated during growth on GlcNAc-containing
   glycans (especially mucin) through an elusive transcriptional factor
   recognizing a previously identified sequence motif in the gibS promoter
   (‘PM2’^[232]11). GibS associates with RbpB (two distinct CLIP peaks)
   and represses the glycoside hydrolase 13 family protein BT_0771 as well
   as the uncharacterized protein BT_3893 through basepairing with its two
   distinct seeds (R1 and R2) to the translation initiation region of the
   corresponding mRNAs. In contrast, transcription of the 14 FopS sRNAs is
   driven from canonical σ^ABfr-dependent promoters and possibly further
   enhanced by an unknown transcriptional regulator during exposure to
   bile salts and potentially other membrane-assaulting stresses. The FopS
   sRNAs also associate with RbpB (each one CLIP peak in the 5’ portion of
   the sRNAs). Through their extended seed sequence R1’, the FopS sRNAs
   bind and repress translation of the mRNA for the outer membrane porin
   SusC^BT_3983 of mammalian high-mannose N-glycan-catabolic PUL72. Given
   the cross-talk of the SusCD system of PUL72 with polysaccharide capsule
   biosynthesis^[233]20, the FopS sRNAs may indirectly also influence the
   surface glycan structure of B. thetaiotaomicron. While the present
   study’s focus was on PUL72, our data suggest FopS regulation to extend
   to numerous additional PULs (experimental support for the in-vitro
   interaction of FopS with seven additional PUL polycistrons [namely
   PUL12, -20, -44, -51, -77, -80, -88] and various further PUL genes
   amongst the top-scoring in-silico-predicted FopS target candidates
   [Supplementary Data [234]2]). b The proposed role of RbpB in
   post-transcriptional PUL72 regulation. In vitro, FopS-10 inhibited
   SusC^BT-3983 translation and addition of recombinant RbpB was
   sufficient to overcome repression. In vivo, expression of susC^BT_3983
   was upregulated in an rbpB overexpression strain and decreased in the
   absence of the protein. This implies that RbpB counteracts
   FopS-mediated target repression, either by titrating the sRNA away from
   the target mRNA or by evoking structural changes within the
   translation-enhancing region of susC^BT_3983.

   Paralogous sRNAs are ubiquitously present in the bacterial kingdom, but
   have mostly been studied in Pseudomonadota (formerly Proteobacteria)
   and Bacillota (Firmicutes)^[235]53. In these phyla, multicopy sRNAs are
   frequently involved in regulating lifestyle transitions. E. coli and
   Salmonella enterica, for example, encode the paralogous sRNA pair
   OmrA/B to reinforce the transition between motility and biofilm
   formation^[236]54–[237]57. A pair of sibling sRNAs governs the switch
   from cataplerotic to anaplerotic metabolism in
   Neisseria^[238]58–[239]61. The Qrr sRNA family integrates
   quorum-sensing signals to coordinate biofilm formation and virulence in
   Vibrio spp^[240]62,[241]63. The five Cia-dependent sRNAs of
   Streptococcus pneumonia repress competence^[242]64,[243]65 and no less
   than seven sRNA paralogs (LhrC1-7) regulate virulence programs in the
   food-borne pathogen Listeria monocytogenes^[244]66,[245]67.

   What are the functional benefits that prompt bacteria to maintain
   multicopy sRNAs in their genome despite the evolutionary pressure to
   minimize energy expenditure? Since they share a substantial degree of
   sequence and structure, sRNA paralogs typically regulate an overlapping
   set of target genes, yet may also have exclusive targets. Non-redundant
   regulatory functions may additionally arise from differential sRNA
   activation, e.g. the expression of paralogous sRNAs may be governed by
   different σ factors. Within the FopS sRNA cluster (excluding GibS), we
   did not observe strong indication of sub-diversification among
   individual family members, except for a few SNPs in the R1’ seed region
   of four FopS sRNAs (FopS-04, FopS-08, FopS-11, FopS-14; Fig. [246]3a).
   However, coexpression of the FopS sRNAs might have implications for
   dosage-dependent effects. For example, multiplexing FopS transcription
   across multiple loci could minimize the response time upon sensing a
   certain stress or nutrient cue, and rapidly rearrange cell-surface
   components needed to cope with hostile conditions and/or outpace
   metabolic competitors. Of note, the FopS cluster is not the only
   multicopy sRNA family in Bacteroides that is associated with RbpB. For
   example, the second enriched sequence motif within its ligands
   (Supplementary Fig. [247]4a) is contained in another paralogous sRNA
   family, consisting of six members (BTnc020, BTnc054, BTnc075, BTnc172,
   BTnc213, BTnc266) that are each encoded adjacent to
   transposases/invertases. CRISPR-based combinatorial perturbation
   screens that build upon existing technology^[248]44 might in the future
   be leveraged to systematically dissect additive effects and functional
   redundancies within the FopS cluster and other paralogous sRNA families
   in Bacteroides and beyond.

   The Bacteroidota diverged from the common line of eubacterial descent
   before other groups^[249]68. Consequently, these bacteria are
   fundamentally different from the other major Gram-negative phylum, the
   Pseudomonadota, with its long-standing model species E. coli and S.
   enterica. For example, Bacteroides spp. evolved unique
   transcription^[250]69 and translation^[251]70 initiation signals.
   Importantly, although they encode hundreds of noncoding
   RNAs^[252]11–[253]13,[254]15, these bacteria lack the classical
   chaperones of pseudomonadotal sRNAs^[255]71. Despite recent attempts to
   infer Bacteroidota RNA chaperones from in-silico prediction and
   in-vitro experimentation^[256]15,[257]16, up to now the only global RBP
   that had been functionally characterized in this phylum was the
   transcription termination factor Rho^[258]72.

   Applying CLIP-seq, we here demonstrated that B. thetaiotaomicron RbpB
   directly associates with >170 mRNAs and ~60 sRNAs. Among the latter are
   the functionally characterized GibS^[259]11, MasB^[260]12, and
   BatR^[261]44, which all act via basepairing to complementary stretches
   near the start codon of their respective target mRNAs. In fact, similar
   to the situation with GibS (Supplementary Fig. [262]9b), we observed
   ternary complexes composed of RbpB, MasB or BatR, and their respective
   target mRNAs in vitro (Supplementary Fig. [263]10c). It therefore
   appears plausible that RbpB may oligomerize in vivo, offering multiple
   binding sites for RNA. Besides, we observed sRNA-independent binding of
   RbpB to the 5’ UTRs of selected mRNAs in vitro and according to our
   CLIP-seq data, RbpB preferentially binds to AU-rich sequences that are
   frequently found around the Bacteroides ribosome binding site^[264]70.
   A pathway enrichment analysis further indicated translation-associated
   processes to be overrepresented amongst the functional annotations of
   RbpB mRNA ligands. In the one tested example, the addition of
   recombinant RbpB did not alter the translational output of susC^BT_3983
   in the absence of a repressing sRNA. Structural analysis of RbpB,
   including the determination of its RNA-binding residues, might provide
   valuable insights needed to better understand the molecular
   mechanism(s) of this global RNA binder in the future.

   In summary, the present work uncovered an RNA network controlling PUL
   translation in Bacteroides—genetic loci that are important for
   efficient colonization of the mammalian gut and were previously known
   to be regulated transcriptionally. More generally, our study serves as
   a model for RNA-mediated metabolism control by gut commensals and
   proposes RRM-1 domain-containing proteins as excellent candidates to
   identify additional post-transcriptional hubs in the microbiota. This
   knowledge could spur ongoing endeavors to exploit Bacteroides RNA
   elements as tools for microbiome engineering^[265]73.

Methods

Mouse experiments

   Mouse experiments were conducted in accordance with the policies of the
   Institutional Animal Care and Use Committee at Vanderbilt University
   Medical Center (protocol # M2300019-00). C57BL/6 J wild-type (cat#
   000664), were obtained from Jackson Laboratory. Mice were housed in
   sterile cages under specific pathogen-free conditions on a 12 h light
   cycle, with ad libitum access to food and sterile water at Vanderbilt
   University Medical Center. Seven to nine-week-old female or male mice
   were randomly assigned into treatment groups before the experiment. One
   week before antibiotic treatment, mice were switched to a low fiber
   diet (TD.86489) or remained on a control diet 5010 (LabDiet 0001326)
   until the end of the experiment. Antibiotic cocktails (ampicillin
   [Sigma-Aldrich], metronidazole [Sigma-Aldrich], vancomycin [Chem Impex
   International], and neomycin [Sigma-Aldrich]; 5 mg of each per mouse)
   were administered by oral gavage daily for 5 days. After antibiotic
   treatment, mice were inoculated with an equal mixture of 0.5 × 10^9 CFU
   of the B. thetaiotaomicron wild-type strain and 0.5 × 10^9 CFU of the
   respective rbpB mutant for 6 days. After euthanasia, cecal and colonic
   tissue was collected in sterile PBS, and the abundance of B.
   thetaiotaomicron strains was quantified by plating serial-diluted
   intestinal contents on selective agar.

16S rDNA sequencing

   Total DNA was isolated from colonic contents using the Qiagen PowerSoil
   Pro Kit and the libraries were prepared by amplifying the V3-V4
   hypervariable region of 16S ribosomal RNA-encoding sequences. The
   library was purified and sequenced on an Illumina MiSeq platform,
   generating 250 bp single-end reads. The downstream analysis was
   performed with QIIME2 version 2024.5^[266]74. Taxonomic profiling was
   performed using the SILVA taxonomy database version 132^[267]75. The
   sequencing reads generated during the current study are available at
   the European Bioinformatics Institute repository under accession No.
   PRJEB80083 (secondary accession ERP164134).

Bacterial cultivation and genetics

   Liquid cultures of B. thetaiotaomicron strains were prepared in complex
   tryptone-yeast extract-glucose (TYG) medium (20 g L^−1 tryptone,
   10 g L^−1 yeast extract, 0.5% glucose, 5 mg L^−1 hemin, 1 g L^−1
   cysteine, 0.0008% CaCl[2], 19.2 mg L^−1 MgSO[4]·7H[2]O, 40 mg L^−1
   KH[2]PO[4], 40 mg L^−1 K[2]HPO[4], 80 mg L^−1 NaCl, 0.2% NaHCO[3]) or
   minimal medium (1 g L^−1 L-cysteine, 5 mg L^−1 hemin, 20 mg L^−1
   L-methionine, 4.17 mg L^−1 FeSO[4], 0.2% NaHCO[3], 0.9 g L^−1
   KH[2]PO[4], 0.02 g L^−1 MgCl[2]·6H[2]O, 0.026 g L^−1 CaCl[2]·2H[2]O,
   0.001 g L^−1 CoCl[2]·6H[2]O, 0.01 g L^−1 MnCl[2]·4H[2]O, 0.5 g L^−1
   NH[4]Cl, 0.25 g L^−1 Na[2]SO[4]) supplemented with 0.5% of the
   indicated carbon sources. For propagation on plates, brain heart
   infusion-supplemented (BHIS) agar (52 g L^−1 BHI agar powder, 1 g L^−1
   cysteine, 5 mg L^−1 hemin, 0.2% NaHCO[3]) was used. Cultures were
   incubated at 37 °C in an anaerobic chamber (Coy Laboratory Products) in
   presence of an anoxic gas mix of 85% N[2], 10% CO[2], 5% H[2]. E. coli
   strains were cultured aerobically in Luria-Bertani (LB) broth (10 g
   L^−1 tryptone, 5 gL^−1 yeast extract, 10 g L^−1 NaCl) at 37 °C with
   shaking or statically on LB agar plates. All bacterial strains,
   plasmids, and oligonucleotides used in this study are listed in
   Supplementary Data [268]3.

   Defined B. thetaiotaomicron deletion mutants (∆rbpB, ∆fops-08,
   ∆fopS-09, ∆fopS-10, ∆∆fopS-09/10) were generated using the pSIE1
   plasmid system^[269]76. In brief, flanking regions ( ~750 bp) of the
   intended deletion site were assembled into the linearized pSIE1 plasmid
   using NEBuilder® HiFi DNA Assembly kit (NEB, E2621). Plasmids were
   transformed into electro-competent E. coli S17-1 λpir and further
   conjugated into B. thetaiotaomicron. Conjugants were selected on
   BHIS^gent/erm plates. Subsequently, single colonies were inoculated in
   TYG medium without antibiotics and cultured overnight. The next day,
   cultures were serially diluted (10^−1 to 10^−3) and plated on BHIS agar
   plates supplemented with 100 ng mL^−1 anhydrotetracycline (aTC). Colony
   PCR and Sanger sequencing were performed to confirm the intended
   deletions. Complementation strains under control of the respective
   native promoter (fopS-08^+, fopS-09^+, fopS-10^+) were constructed via
   Gibson assembly using the pWW3452^[270]77 plasmid (AWP-015). For
   overexpression of rbpB (strain rbpB^++), constructs were generated
   analogous to the complementation strains, except that the native
   promoter was replaced by a strong phage promoter. For CLIP-seq, rbpB
   was expressed in frame with a downstream 3xFLAG and 6xHis tag, stably
   integrated into the wild-type chromosome. Cloning gRNAs for the
   CRISPRi-based fopS knockdown was performed as described in ref.
   ^[271]44. Briefly, oligonucleotides containing the 20-nt spacer
   sequence and flanking regions on both ends 5’
   (5′ATCTTTGCAGTAATTTCTACTGTTGTAGAT) and 3’
   (5′CCGGCTTATCGGTCAGTTTCACCTGATTTA) were inserted into the
   SmaI-linearized vector (AWP-029) encoding an inactive Cas12a nuclease
   (dPb2Cas12a) under control of an isopropyl-β-d-1-thiogalactopyranoside
   (IPTG)-inducible promoter. Oligonucleotides and the linearized vector
   were joined in a 10 µL reaction with an oligonucleotide:vector ratio of
   20:1. Assembled constructs were transformed into electrocompetent E.
   coli S17-1 λpir and subsequently conjugated into the B.
   thetaiotaomicron double-deletion strain ∆∆fopS-09/10. Constructs used
   for trans-complementation, CLIP-seq, and CRISPRi are variants of the
   pNBU2 integrative vector system^[272]78, which allows their insertion
   into the Bacteroides genome at one of the two tRNA^ser sites.

UV cross-linking and immunoprecipitation (CLIP)

   CLIP experiments were performed as described in ref. ^[273]79 with
   minor modifications. In brief, bacterial cultures were grown to
   mid-exponential phase (OD[600] = 2.0) in TYG. For each biological
   replicate experiment, 200 OD equivalents were harvested, of which 50 mL
   were irradiated with UV light (254 nm) at 800 mJ/cm^2 on a
   22 cm x 22 cm plastic tray (crosslinked), while the remaining 50 mL
   were not (non-crosslinked control). Cells were pelleted (4000 x g,
   40 min, 4 °C) and snap-frozen in liquid nitrogen. Each pellet was
   resuspended in 800 µL of NP-T buffer (50 mM NaH[2]PO[4], 300 mM NaCl,
   0.05% Tween 20, adjusted to pH 8.0). Cells were lysed mechanically in a
   standard Retsch apparatus (30 1/s, 10 min), using 1 mL of 0.1 mm glass
   beads for grinding. To clear the lysates, samples were centrifuged
   twice at 16,000 x g for each 15 min at 4 °C.

   Cleared lysates were mixed with equal volumes of NP-T buffer containing
   8 M urea and incubated for 5 min at 65 °C, with shaking at 900 rpm.
   Each lysate was then diluted (1:10) in pre-cooled NP-T buffer.
   Subsequently, 30 µL of anti-FLAG M2 magnetic beads (Sigma-Aldrich) were
   washed and equilibrated in 800 µL of NP-T buffer, added to the samples,
   and incubated for 1 h at 4 °C, rotating. Beads were collected by
   centrifugation at 1500 x g for 1 min at 4 °C and washed twice with each
   2 mL of high-salt buffer (50 mM NaH[2]PO[4], 1 M NaCl, 0.05% Tween 20,
   adjusted to pH 8.0) and 2 mL of NP-T buffer. Each sample was
   resuspended in NP-T buffer, containing 1 mM MgCl[2] and 25 U Benzonase
   and incubated for 10 min at 37 °C, 900 rpm, followed by a 2 min
   incubation on ice. After washing the beads once with 1 mL of high-salt
   buffer and twice with 1 mL of CIP-buffer (100 mM NaCl, 50 mM Tris-HCl
   [pH 7.4], 10 mM MgCl[2]), 100 µL of CIP-mix containing 10 U of calf
   intestinal alkaline phosphatase (New England Biolabs) in CIP-buffer
   were added and beads incubated for 30 min at 37 °C, 800 rpm.
   Subsequently, one wash with 500 µL of high-salt buffer and two washes
   with 500 µL of 1x PNK buffer (Reaction buffer A, ThermoScientific) were
   performed. For labeling, a PNK mix was prepared containing 98 µL of 1x
   PNK buffer, 10 U of T4 polynucleotide kinase (ThermoScientific), and 10
   µCi γ^32P-ATP. Beads were resuspended in the PNK mix and incubated for
   30 min at 37 °C. Finally, 10 µL of non-radioactive ATP (1 mM) were
   added and samples incubated for 5 min at 37 °C, before washing the
   beads two times with 1 mL of NP-T buffer. For elution, beads were
   resuspended in 15 µL of 1x protein loading buffer with 50 mM DTT and
   incubated for 5 min at 95 °C. Magnetic beads were collected on a
   magnetic separator and the supernatant was transferred to a fresh tube.
   The elution step was repeated, the two supernatant fractions pooled and
   loaded (total volume of 30 µL) on a 15% sodium dodecyl sulfate
   (SDS)–polyacrylamide (PAA) gel.

   Protein-RNA complexes were transferred on a Portran 0.45 µm NC membrane
   (Amersham). The protein ladder was labeled with a radioactive marker
   pen and membranes were exposed to a phosphor screen overnight. The
   autoradiogram was visualized, printed, and aligned with the membrane to
   properly excise the RNA-protein complexes from the membrane. Each
   membrane piece was cut into smaller pieces and transferred to a
   low-binding tube. To each tube, 200 µL of PK mix were added, containing
   2x PK buffer (100 mM Tris-HCl [pH 7.9], 10 mM EDTA, 1% SDS), 1 mg/mL
   Proteinase K (Fermentas) and 10 U SUPERaseIN (Life Technologies), and
   incubated for 1 h at 37 °C, 800 rpm. Additionally, 100 µL of PK buffer
   containing 9 M urea were added and incubated for another hour at 37 °C,
   800 rpm. The solution was cleared from membrane pieces and one volume
   of P:C:I (ROTI phenol:chloroform:isoamyl alcohol) was added. Samples
   were incubated in phase-lock tubes for 5 min at 30 °C, 1000 rpm,
   centrifuged for 12 min at 16,000 x g, 4 °C, and the aqueous phase
   transferred to a fresh tube. RNA was precipitated with three volumes of
   30:1 mix (ethanol:3 M NaOAc, pH 6.5) and 1 µL of GlycoBlue
   (ThermoScientific) overnight at -20 °C. Subsequently, RNA was pelleted,
   washed with 80% (vol vol^−1) ethanol, resuspended in 10 µL of H[2]O and
   dissolved at 65 °C for 5 min, shaking (800 rpm). Samples were stored at
   -20 °C until sequencing.

CLIP-seq protocol and data analysis

   The preparation of cDNA libraries and high-throughput sequencing was
   performed at vertis Biotechnologie AG, Freising, Germany. First,
   oligonucleotide adapters were ligated to the 3’ and 5’ ends, followed
   by a first-strand cDNA synthesis using M-MLV reverse transcriptase and
   the 3’ adapter as primer. The resulting cDNAs were PCR-amplified using
   a high-fidelity DNA polymerase with a varying number (18-24) of PCR
   cycles per sample. Amplified cDNA was purified using the Agencourt
   AMPure XP kit (Beckman Coulter Genomics). For sequencing, the samples
   were pooled in equimolar amounts and paired-end-sequenced on an Ilumina
   HiSeq system with 2x150 bp read lengths.

   For analysis, raw reads were filtered and trimmed with BBDuk
   (paired-end mode, phred score cutoff: 20, minimal read length: 12 nt).
   To remove putative PCR duplicates, reads were deduplicated with
   FastUniq 1.1^[274]80. Mapping was done with the READemption 1.0.5
   pipeline using segemehl 0.3.4^[275]81. READemption align was run in
   paired-end mode with an accuracy of 80%. Peak calling was performed
   with PEAKachu pipeline version 0.2.0^[276]82. First, normalization
   factors for peak calling were calculated with an in-house R script as
   described in ref. ^[277]28. In brief, the core positions of all
   libraries were isolated using an exploratory analysis of read counts
   summarized per position. Positions with low read counts were filtered.
   Normalization factors were then calculated based on the background
   positions, which are represented by high read counts in both
   crosslinked and non-crosslinked libraries. PEAKachu adaptive was run in
   paired-end and paired-replicates mode together with the calculated
   normalization factors. Maximum fragment size was set to 50 and mad
   multiplier to 0. Only peaks with a log[2]-fold-change (log[2] FC) ≥ 2
   and an adjusted p-value (p[adj], Benjamini-Hochberg corrected) ≤0.05
   were considered significant. The annotation of those peaks and the read
   coverages from the sequenced CLIP-seq libraries can be retrieved
   through our public database, Theta-Base^[278]12, by first clicking on
   the JBrowse icon and then selecting tracks from the “RbpB_CLIP” tab to
   the left. Users can jump to individual CLIP peaks by typing the
   respective peak ID into the search field at the top of the page.

   As per default^[279]79, we first considered only uniquely mapped reads
   in our CLIP-seq quantification and discarded multi-mapped reads.
   However, given the observed sequence similarity amongst the identified
   RbpB targets (e.g. the FopS sRNAs), we re-ran the analysis, this time
   also considering multi-mapped reads. While not affecting the overall
   results, this adaptation of the peak calling pipeline led to a further
   increase in the number of significantly enriched FopS sRNAs (from 10
   FopS sRNAs when multi-mapped reads were discarded to all 14 FopS
   members when also these reads were included).

Pathway enrichment analysis of CLIP-seq peaks

   The genes containing at least one significant peak were checked for
   enrichment of gene sets belonging to a custom annotation (containing GO
   terms, KEGG pathways and modules, and functional information parsed
   from the literature^[280]12). Enrichment was computed with the enricher
   function of the clusterProfiler R package version 3.14.3^[281]83.

RbpB purification

   Expression and purification of RbpB was performed at the recombinant
   protein expression facility of the Rudolf-Virchow-Center, Würzburg,
   Germany. In brief, constructs containing the coding sequence of RbpB
   together with a His-Sumo3 tag were cloned in an E. coli expression
   vector (pETM11) and transformed into E. coli Bl21 (DE3). For
   large-scale purification, cultures were grown to an OD[600] of 0.6 in
   8 L of LB medium, induced with 0.5 mM IPTG, and incubated overnight at
   18 °C. Cells were resuspended in 5–10 mL of lysis buffer (150 mM NaCl,
   50 mM NaH[2]PO[4], pH 7.0, 10% sucrose, 10 mM imidazole, 1 mM TCEP,
   1 mM MgCl[2], protease inhibitor, DNase) per gram cell pellet and lysed
   by sonication. Lysates were cleared by centrifugation (30,000 x g,
   30 min, 4 °C) and incubated with 4 mL of Ni-NTA (equilibrated to lysis
   buffer) for 1 h at 6-8 °C followed by immobilized metal affinity
   chromatography (IMAC) purification. Elution was performed seven times
   with each 4 mL of IMAC elution buffer (150 mM NaCl, 20 mM NaH[2]PO[4],
   pH 7.0, 10% sucrose, 250 mM imidazole, 1 mM TCEP, 1 mM MgCl[2]).

   IMAC eluates were pooled and the tag was cleaved off using SenP2
   Sumo-protease. Pooled eluates were dialyzed overnight against 2 L of
   dialysis buffer (150 mM NaCl, 20 mM NaH[2]PO[4], pH 7.0, 10% sucrose,
   1 mM TCEP, 1 mM MgCl[2]) at 6-8 °C. Finally, the dialysate was
   concentrated (10 mL total volume), cleared by centrifugation (16,000 x
   g, 20 min, 4 °C), and loaded on a HiLoad Superdex 75 16/600 pg for size
   exclusion chromatography. Eluate fractions were collected (2 mL each),
   pooled from two runs, and analyzed by SDS-PAGE.

In-vitro transcription and radiolabeling of RNA

   For in-vitro transcription, DNA templates were amplified from genomic
   DNA using primer pairs carrying a T7 promoter (Supplementary
   Data [282]3). The ensuing in-vitro transcription reaction was performed
   using the MEGAscript T7 kit (ThermoFisher Scientific). Excess DNA was
   removed by DNase I digestion (1 U, 37 °C for 15 min) and the RNA
   product purified from a 6% (vol vol^−1) PAA-7M urea gel using a
   LowRange RNA ladder (ThermoFisher Scientific) for precise sizing. The
   in-vitro-transcribed RNA was eluted in RNA elution buffer (0.1 M NaOAc,
   0.1% SDS, 10 mM EDTA) overnight on a thermoblock at 8 °C and 1400 rpm,
   subsequently precipitated using ethanol:NaOAc (30:1) mix, washed with
   75% ethanol, and resuspended in 20 µL of water (65 °C for 5 min).

   For radioactive labeling, 50 pmol of the in-vitro-transcribed RNA were
   dephosphorylated using 25 U of calf intestine alkaline phosphatase
   (NEB) in a 50 µL reaction volume. After 1 h incubation at 37 °C, RNA
   was extracted with a phenol:chloroform:isoamyl alcohol mix (P:C:I,
   25:24:1) and precipitated as described above. Finally, 20 pmol of the
   dephosporylated RNA were 5’ end-labeled (20 µCi of ^32P-γATP) in a
   20 µL reaction for 1 h at 37 °C using 1 U polynucleotide kinase (NEB).
   Labeled RNA was purified on a G-50 column (GE Healthcare) and extracted
   from a PAA gel as described above.

Electrophoretic mobility shift assay (EMSA)

   Protein-RNA and RNA-RNA EMSAs were performed in a final reaction volume
   of each 10 µL, containing 1x RNA structure buffer (SB; Ambion), 1 µg
   yeast RNA (~4 µM final concentration), and 5’ end-labeled RNA (4 nM
   final concentration). Reactions were either incubated with increasing
   concentrations of in-vitro-transcribed target mRNA segments (0; 8; 16;
   32; 64; 128; 256; 512; 1,024 nM final concentration) or increasing
   concentrations of purified RbpB (0; 0.15; 0.3; 0.6; 1.25; 2.5; 5; 10;
   20; 30; 40; 80 µM final concentration). Reactions were incubated for
   1 h at 37 °C and stopped by adding 3 µL of 5x native loading dye (0.2%
   bromophenol blue, 0.5x TBE, 50% glycerol) and loaded on a native 6%
   (vol vol^−1) PAA gel in 0.5x TBE buffer at 4 °C and run at 300 V for
   3 h. The gel was dried at 80 °C for 1 h on a Gel Dryer 583 (Bio-Rad),
   exposed overnight, and visualized on a phosphorimager (FLA-3000 Series,
   Fuji).

   Three-component EMSA was carried out in a 15 µL reaction volume,
   containing 1x RNA structure buffer (SB; Ambion), 1 µg yeast RNA (~ 4 µM
   final concentration), and 5’ end-labeled RNA (4 nM final
   concentration). Depending on the experimental setup, reactions were
   either incubated with in-vitro-transcribed target mRNA segments or
   sRNAs (500 nM final concentration). Increasing concentrations of
   purified RbpB were added to the reactions (0; 0.15; 0.3; 0.6; 1.25;
   2.5; 5; 10; 20 µM final concentrations).

Rifampicin treatment to halt de-novo transcription

   To assess RNA stability, single colonies of B. thetaiotaomicron strains
   AWS-001 (WT), AWS-323 (ΔrbpB), and AWS-218 (rbpB^++) were inoculated in
   liquid TYG medium and grown for ~16 h, subsequently sub-cultured (1:100
   dilution), and grown to mid-exponential phase (OD[600] = 2.0). To halt
   de-novo transcription, rifampicin was added (500 µg/mL final
   concentration). Samples were taken immediately before the addition of
   rifampicin (0 min) and at indicated time points after the treatment (3;
   6; 9; 12; 45; 60; 90 min). Total RNA was extracted and decay assessed
   by northern blot analysis or qRT-PCR (see below).

RNA extraction and removal of genomic DNA

   Generally, total RNA was isolated by hot-phenol extraction from culture
   aliquots. To this end, 4 OD equivalents of culture were harvested,
   mixed with 20% vol. stop mix (95% vol vol^−1 ethanol, 5% vol vol^−1
   water saturated phenol, pH >7.0), and snap-frozen in liquid nitrogen.
   Lysis of the bacterial cells was mediated by the addition of 600 µL of
   lysozyme (0.5 mg mL^−1) and 60 µL of 10% SDS, followed by an incubation
   for 2 min at 64 °C, before 66 µL of 3 M NaOAc (pH 5.2) were added. For
   extraction, phenol was added (750 µL; Roti-Aqua phenol) and samples
   incubated for 6 min at 64 °C, followed by the addition of 750 µL of
   chloroform. After centrifugation, RNA was precipitated from the aqueous
   phase overnight at -20 °C with twice the volume of 30:1 (ethanol:3 M
   NaOAc, pH 6.5). Subsequently, RNA was pelleted, washed with 75% (vol
   vol^−1) ethanol, and resuspended in 50 µL of H[2]O. To remove
   contaminating genomic DNA, 40 µg of RNA were treated with 5 U of DNase
   I (Fermentas) and 0.5 µL of Superase-In RNase Inhibitor (Ambion) for
   45-60 min at 37 °C in 50 µL reaction volumes. Finally, RNA was purified
   through phenol-chloroform extraction (ROTI phenol:chloroform:isoamyl
   alcohol) and resuspended in 30 µL of H[2]O.

Quantitative real-time PCR (qRT-PCR)

   qRT-PCR was performed as described in^[283]11. Briefly, a reaction mix
   was prepared for each well of a 96-well plate, containing 10 ng of
   DNase I-treated RNA, 5 µL of master mix (No ROX SYBR MasterMix blue
   dTTP kit, Takyon), 0.1 µL of each forward and reverse primer (10 µM
   each), and 0.08 µL of reverse transcriptase (One-Step Kit converter,
   Takyon). Analysis was on a CFX96 instrument (Biorad).

Identification of FopS paralogs and homologs

   The 14 paralogous sequences that were identified by sequence homology
   were manually examined for the presence of annotated transcription
   start and termination sites on Theta-Base^[284]12 and the existence of
   putative sORFs within each locus was excluded^[285]84. Originally, the
   annotated 5’ ends of FopS-01 (BTnc025) and FopS-04 (BTnc032), as well
   as the 3’ end of FopS-13 (BTnc188) were supported by only few RNA-seq
   reads^[286]11. We therefore manually curated the transcript boundaries
   to the previously identified secondary transcription start
   sites^[287]11 in case of FopS-01 and FopS-04, and a 3’ end downstream
   of a predicted terminator hairpin in case of FopS-13 (curated sRNA
   boundaries marked by gray triangles in Fig. [288]3a). Based on these
   accurately defined transcript boundaries, we classified the individual
   FopS family members into intergenic sRNAs (n = 9), antisense RNAs
   (n = 4), and 3’ UTR-derived sRNAs (n = 1).

   BLAST analysis of FopS homologs from different Bacteroides species was
   performed using the BLASTn suite and FopS-08 as a representative of the
   B. thetaiotaomicron family. Run parameters were adjusted to filter for
   a maximum of 250 target sequences, with a word size of 28 and a
   match/mismatch score of 1/-2. The resulting table was grouped by
   species and the Supplementary Fig. [289]5a was generated using the heat
   map function of GraphPad Prism version 10.0.0 for Windows (GraphPad
   Software, Boston, Massachusetts USA, [290]www.graphpad.com). The same
   procedure was followed to retrieve the RRM protein-encoding homologs.

   The percent identity of the 14 individual FopS paralogs of B.
   thetaiotaomicron with GibS was calculated using Clustal Omega multiple
   sequence alignment^[291]85 and visualized by passing the output to
   MView v1.67^[292]86.

In-silico prediction of the FopS sRNA consensus structure

   Curation of the B. thetaiotaomicron FopS consensus structure was done
   as described in detail in^[293]15 and originally reported in^[294]87.
   Briefly, alignments of the 14 FopS sRNA sequences were generated with
   the WAR webserver^[295]33 and the maximum consistency alignment and
   structure were downloaded in stockholm format. We then used the RALEE
   emacs “RNA editor mode”^[296]88 to manually curate the alignment and
   optimized it with R-scape^[297]89.

Northern blot analysis

   For northern blot analysis, 5 µg total RNA were denatured for 5 min at
   95 °C, incubated on ice for 5 min, and separated on a 6% (vol vol^−1)
   PAA-7 M urea gel at 300 V for ~2 h. Electroblotting of the RNA onto a
   Hybond-N+ membrane (GE Healthcare Life Sciences) was performed at 50 V,
   4 °C for 1 h. Membranes were UV-crosslinked (0.12 J/cm^2),
   pre-hybridized in 15 mL of Hybri-Quick buffer (Carl Roth AG) at 42 °C
   for 1 h, and incubated with ^32P-labeled gene-specific oligonucleotides
   at 42 °C. Blots were washed with decreasing concentrations (5×, 1×,
   0.5×) of SSC buffer (20× SSC: 3 M NaCl, 0.3 M sodium citrate, pH 7.0),
   exposed as required, and visualized on a phosphorimager (FLA‐3000
   Series, Fuji).

sRNA target prediction

   To predict FopS targets, the IntaRNA program^[298]41,[299]90 was
   employed. The query input consisted of a 35 nt sequence comprising the
   complete R1’ region at a consensus of 70%
   (GTGTTTTTCATAGTATTAGATTTAAGGTTAACAAA) and the program was executed
   using its default settings against the B. thetaiotaomicron VPI-5482
   genome.

B. thetaiotaomicron growth kinetics in minimal medium containing defined
carbon sources

   Growth assays on different carbon sources were performed in minimal
   medium (1 g L^−1 L-cysteine, 5 mg L^−1 hemin, 20 mg L^−1 L-methionine,
   4.17 mg L^−1 FeSO[4], 0.2% NaHCO[3], 0.9 g L^−1 KH[2]PO[4], 0.02 g L^−1
   MgCl[2]·6H[2]O, 0.026 g L^−1 CaCl[2]·2H[2]O, 0.001 g L^−1
   CoCl[2]·6H[2]O, 0.01 g L^−1 MnCl[2]·4H[2]O, 0.5 g L^−1 NH[4]Cl,
   0.25 g L^−1 Na[2]SO[4]) supplemented with 0.5% of the indicated carbon
   sources. A single colony of the indicated strains was used to inoculate
   5 mL of minimal medium-glucose and pre-cultured for 24 h. Subsequently,
   1 mL of the bacterial culture was pelleted (2000 x g for 3 min) and
   resuspended in the same volume of minimal medium without any carbon
   source. This served as the inoculum (1:100) for fresh minimal medium
   containing either high-mannose (Man[9]GlcNAc[2]; NatGlycan LLC) or
   glucose. For measurement, diluted cultures were transferred to
   individual wells of a 96-well plate and optical density was recorded
   every 20 min over a time-course of 48 h.

Inline probing

   Inline probing was performed as described in ref. ^[300]11. Briefly,
   0.2 pmol 5’ end-^32P-labeled RNA were incubated for 40 h at room
   temperature in 2x inline probing buffer (100 mM KCl, 20 mM MgCl[2],
   50 mM Tris-HCl, pH 8.3). Reactions were stopped by the addition of
   10 µL of 2x gel-loading solution (10 M urea, 1.5 mM EDTA, pH 8.0). To
   prepare the RNase I ladder, 0.4 pmol 5’ end-^32P-labeled RNA were
   denatured in 1x sequencing buffer (Ambion) at 95 °C for 1 min. RNase TI
   (0.1 U) was added and incubated at 37 °C for 5 min. The alkaline
   hydrolysis ladder was prepared by incubating 0.4 pmol 5’
   end-^32P-labeled RNA in 9 µL of 1x alkaline hydrolysis buffer (Ambion)
   for 5 min at 95 °C. To stop the reaction, 12 µL of loading buffer II
   were added to both ladders and stored on ice. Samples were resolved on
   a 10% (vol vol^−1) PAA-7 M urea sequencing gel at 45 W for 2-3 h. Gels
   were dried and visualized as described above.

In-vitro translation assay and western blotting

   In-vitro translation assays were performed using a reconstituted E.
   coli protein synthesis system (PURExpress, New England Biolabs).
   Reactions were performed following manufacturer’s instruction with a
   few modifications based on our prior expertise^[301]91. In brief, 0.5
   or 1 µM in-vitro-transcribed mRNA (5’susC^BT_3983-gfp-3xFLAG,
   gfp-3xFLAG) were incubated in presence or absence of 25 or 50 µM
   in-vitro-transcribed FopS-10 sRNA for 1 min at 95 °C and chilled on ice
   for 5 min. RbpB (20 µM) was added and the reaction pre-incubated for
   10 min at 37 °C before the PURExpress components were added. After
   ~4 h, reactions were stopped by adding 5 µL of 5x protein loading
   buffer and the whole sample volume (30 µL) was loaded on a 12% SDS-PAA
   gel. Proteins were transferred onto a Portran 0.2 µm NC membrane
   (Amersham) for 1.5 h, 350 mA at 4 °C under semi-dry conditions. To
   assess equal sample loading, membranes were stained with Ponceau S
   (Sigma-Aldrich), visualized, and then destained with 0.1 M NaOH prior
   to blocking in TBS-T with 10% powdered milk (1 h, room temperature).
   Monoclonal anti-FLAG (Sigma-Aldrich) antibody was added (1:1000 in
   TBS-T with 10% powdered milk) and incubated overnight at 4 °C, with
   shaking. Membranes were washed three times in TBS-T for each 10 min and
   incubated with anti-mouse IgG, HRP (ThermoFisherScientific) diluted
   1:10,000 in TBS-T with 10% powdered milk for 1 h at room temperature.
   After rinsing, ECL detection substrate (Amersham) was added and HRP
   activity detected using a CCD imager (Amersham ImageQuant 800 systems).

   The translational fusion of target sequences to gfp allowed for a
   fluorescence-based kinetic readout, alternatively to the above
   end-point determination via immunoblotting. To this end, 0.5 µM of
   in-vitro-transcribed FopS target mRNA variants
   (5’susC^BT_3983-gfp-3xFLAG, wild-type or point-mutated sequences) were
   incubated in presence or absence of 25 µM in-vitro-transcribed FopS-10
   sRNA (wild-type or mutated) for 1 min at 95 °C and chilled on ice for
   5 min. Components of the PURExpress Kit were added and reactions
   adjusted to a final volume of 12 µL. Each reaction was prepared in two
   technical replicates and thus split into 5 µL aliquots and distributed
   into individual wells of a 96-well plate (V-bottom) for fluorescence
   measurements. Kinetics were recorded for 16 h at 37 °C, with
   fluorescence measurement in 3-min intervals.

Reporting summary

   Further information on research design is available in the [302]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [303]Supplementary Information^ (31.9MB, pdf)
   [304]41467_2024_55383_MOESM2_ESM.docx^ (16.6KB, docx)

   Description of Additional Supplementary Files
   [305]Supplementary Dataset 1^ (70.4KB, xlsx)
   [306]Supplementary Dataset 2^ (32.4KB, xlsx)
   [307]Supplementary Dataset 3^ (37.3KB, xlsx)
   [308]Reporting Summary^ (103.3KB, pdf)
   [309]Peer Review file^ (943.4KB, pdf)

Source data

   [310]Source Data^ (40.3MB, xlsx)

Acknowledgements