Supplementary Materials and Methods Stabilization of Hfq-mediated translational repression by the co-repressor Crc in Pseudomonas aeruginosa

Abstract In Pseudomonas aeruginosa the RNA chaperone Hfq and the catabolite repression control protein (Crc) govern translation of numerous transcripts during carbon catabolite repression. Here, Crc was shown to enhance Hfq-mediated translational repression of several mRNAs. We have developed a single-molecule fluorescence assay to quantitatively assess the cooperation of Hfq and Crc to form a repressive complex on a RNA, encompassing the translation initiation region and the proximal coding sequence of the P. aeruginosa amiE gene. The presence of Crc did not change the amiE RNA-Hfq interaction lifetimes, whereas it changed the equilibrium towards more stable repressive complexes. This observation is in accord with Cryo-EM analyses, which showed an increased compactness of the repressive Hfq/Crc/RNA assemblies. These biophysical studies revealed how Crc protein kinetically stabilizes Hfq/RNA complexes, and how the two proteins together fold a large segment of the mRNA into a more compact translationally repressive structure. In fact, the presence of Crc resulted in stronger translational repression in vitro and in a significantly reduced half-life of the target amiE mRNA in vivo. Although Hfq is well-known to act with small regulatory RNAs, this study shows how Hfq can collaborate with another protein to down-regulate translation of mRNAs that become targets for the degradative machinery.

In Bacteria, the uptake and utilization of carbon compounds is controlled in a hierarchical manner by a mechanism known as carbon catabolite repression (CCR). Whereas in Enterobacteriaceae and Firmicutes CCR is predominantly controlled at the level of transcription (19), in Pseudomonas spp. CCR operates at the post-transcriptional level (20). In Pae Hfq was shown to act as a translational repressor of uptake and degradative functions required for utilization of carbon and nitrogen sources other than the preferred one (2). Several studies revealed that Hfq forms a repressive complex together with the catabolite repression control protein Crc on target mRNAs, e.g. amiE mRNA encoding an aliphatic amidase (21,22). As Crc was shown to promote Hfq-mediated translational repression of several mRNAs (2), it can be regarded as a translational corepressor. Moreover, a recent ChIP-seq study and a combined RNA-seq/proteomics approach identified 100- (1) and 244 mRNAs (23), respectively, to be regulated by Hfq and Crc. Cryo-EM studies showed that a segment encom-2 Nucleic Acids Research, 2021 passing the ribosome binding site of amiE mRNA is sandwiched between Hfq and Crc, thus rationalizing the auxiliary function of Crc in Hfq-mediated translational repression by obstructing initiating ribosomes (24).
The response to different carbon sources is mediated through different levels of the regulatory RNA CrcZ (25,26). As CrcZ displays several Hfq binding motifs, it sequesters Hfq when CCR is alleviated (2,22), which in turn directs metabolism to use less preferred carbon and nitrogen sources. CrcZ expression is under control of the alternative sigma factor RpoN and the two component signalling system CbrA/B (25,27). As CrcZ acts as a decoy for Hfq, it interferes not only with direct translational repression by Hfq/Crc (2) but also with Hfq-mediated riboregulation by sRNAs (13), with impact on biofilm formation (28) and antibiotic susceptibility (8,9).
In isolation, the Crc protein has no intrinsic affinity for either RNA (29) or Hfq (22). However, it can cooperate with Hfq to assemble a complex on RNAs that contain (AAN) n or (ARN) n repeats, in which A is an adenine, R is a purine (A/G) and N is any nucleotide, which align to engage the six tripartite binding pockets on the distal face of Hfq (22,(30)(31)(32). Although the presence of Crc did not significantly enhance the affinity of Hfq for an amiE derived RNA octadecamer, encompassing five consecutive AAN triplets followed by one ARN triplet (amiE 6ARN ), the Hfq/Crc/amiE 6ARN assembly displayed an increased stability when compared with the Hfq/amiE 6ARN complex alone (22). Structural studies on Hfq/Crc/amiE 6ARN complex supported and explained these findings. The assembly was formed by two stacked Hfq hexamers with each bound to one amiE 6ARN motif and two interconnecting Crc protomers, predominantly contacting the RNA molecules bound to the Hfq hexamers (24).
Here, we have studied Hfq/Crc complexes formed on a longer amiE mRNA sub-sequence (amiE102 RNA; Figure  1A), encompassing the translation initiation region and the proximal coding sequence. Single-molecule co-localization experiments revealed that the presence of Crc did not change the amiE102 RNA-Hfq interaction lifetimes. However, Crc increased the probability of forming stable complexes, which is supported by Cryo-EM studies, showing that the presence of Crc increases the compactness of the repressive complexes. In vitro translation assays corroborate these findings as the presence of Crc resulted in stronger translational repression, which can be reconciled with the significantly reduced stability of amiE mRNA in Pae strain PAO1 when compared with PAO1Δcrc.

Strains and Plasmids
The strains and plasmids used in these studies are indicated throughout the text and newly constructed plasmids are described in Supplementary Materials and Methods.

Protein purification
The P. aeruginosa Hfq/Hfq S65C -Flag heterooligomer (Hfq*) used in the single-molecule assays was pro-duced in E. coli strain JW4130F´ hfq(pKEHfq Pae ; pBADHfq PaeS65C -Flag). The cells were grown at 37 • C in LB medium. At an OD 600 of 0.7, the synthesis of Hfq* was induced by addition of IPTG and arabinose to a final concentration of 1 mM and 0.02%, respectively. The cells were harvested after 2 h, the cell pellet was resuspended in lysis buffer (50 mM Tris-HCl pH 8.0, 1.5 M NaCl, 250 mM MgCl 2 , 1 mM PMSF, 1 mM ␤-mercaptoethanol, 1 mM EDTA, 10 g/ml DNAse I, 10 g/ml RNaseA), and the cells were lysed by sonication. First, Hfq* was highly enriched by heating and ammonium sufate precipitation followed by three consecutive chromatography steps as described in detail by Beich-Frandsen et al. (33). To further determine the ratio between wild type (wt) Hfq and Hfq S65C -Flag in the Hfq* mixture, Hfq* was captured with anti-Flag M2 magnetic beads (Sigma Aldrich) as specified by the manufacturer. Then, given the difference in size between Hfq wt and Hfq S65C -Flag, a western-blot analysis was performed with anti-Hfq antibodies to determine the ratio of Hfq wt and Hfq S65C -Flag. Both, the column purified Hfq* and Hfq*, captured with the anti-Flag M2 magnetic beads, showed a 1:1 ratio of Hfq wt and Hfq S65C -Flag (Supplementary Figure S1A). For the Cryo-EM studies, Pae Hfq was purified from E. coli strain JW4130F´ hfq harboring plasmid pKEHfq Pae as described before (33).

Single-molecule experiments
The RNA molecules used for the single-molecule experiments were prepared by in vitro transcription using T7 RNA polymerase, and then purified on 8 M urea 8% polyacrylamide gels. The DNA templates for transcription were obtained by extension of chemically synthesized overlapping oligodeoxyribonucleotides (Thermo Fisher Scientific) with Q5 DNA polymerase (NEB). The template for the amiE102 sub-sequence, which comprises nt -40 to +60 of amiE mRNA relative to the A (+1) of the AUG start codon, plus two additional Gs at the 5 end ( Figure 1A) was generated by hybridization and extension of the amiE F (5 -TAATACGACTCACTATAGGTTTTTTCGTCCCG AAAAAATAACAACAAGAGGTGATATCCATGC GTCACGGCGATATTTCCAGCAGC-3 ) and amiE R (5 CCTGTGTCCTGTGTGTCCTGTCCAAAGTGTGT CGTCCTGTAGTTGACCACCGCCACTCCGACGG TGTCGTTGCTGCTGGAAATATCGCCG-3 ) oligonucleotides, while the amiE 5 AANmut RNA (Figure 2A   Single-molecule data were collected using a custom-built prism-based total internal reflection fluorescence (TIRF) microscope with an EMCCD camera (Andor). A 640 nm laser was used for Cy5 excitation. A selected field of view was recorded upon flow of Cy5-RNA for 3 minutes. The Cy5 emission signals were collected through a 60 × water immersion objective using custom software developed in the Ha Lab (smCamera).
The amiE102 RNA (250 nM) or variants thereof were added to the same concentration of the SA5-Cy5 DNA nucleotide and annealed by denaturation at 75 • C for 5 min in 1× TNK buffer (10 mM Tris-HCl pH 7.5, 50 mM NaCl, 50 mM KCl), refolded at 37 • C for 15 min, and equilibrated at 20 • C. DDS-coated quartz slides were pre-treated with biotinylated BSA (0.2 mg/ml), Tween-20 (0.2%) and Neutravidin (0.1 mg/ml) as previously described (35). Biotinylated anti-FLAG antibodies (1mg/ml) were 12.500× diluted in imaging buffer (10 mM Tris-HCl pH 7.5, 50 mM NaCl, 50 mM KCl, 4 mM Trolox, 0.01% octaethylene glycol monododecyl-ether (Nikkol), 0.8% glucose, 2 U RNasin Plus) and applied to the slide for 5 min. Unbound antibody was removed by washing with imaging buffer. The immobilized antibody was incubated with 20 nM Hfq* for 5 min, then washed with imaging buffer supplemented with 165 U/ml glucose oxidase. We used unlabeled protein because we noticed that labeling of Pae Hfq* changes the kinetics of RNA binding. The RNA/DNA complexes were diluted to 1 nM in imaging buffer containing glucose oxidase and added to the slide channel in the absence or presence of 100 or 500 nM Crc. As judged from native gel mobility shift assays, the Cy5 DNA oligomer was not displaced from the RNA by Hfq (not shown). The time trajectories of single molecules were extracted from the CCD images using custom IDL code developed in the Ha lab. The Cy5 molecules were selected based on spots appearing in the acceptor channel during the whole duration of the movie.

Single-molecule data analysis
The dwell times were collected using custom Matlab code. To estimate the lifetimes and associated errors, un-binned distributions were fit using maximum likelihood methods (36) to equations containing two or three exponential terms (Equations 1 and 2, respectively). In both equations, t m is the minimum resolvable time interval in the experiment, t x represents the duration of the experiment, 1 , 2 , 3 represent characteristic lifetimes and a, a 1 , and a 2 are the fitted amplitudes; a 3 = 1a 1a 2 . The amplitudes of each term of an exponential distribution represent the proportion of detected events that occur with a given rate constant (1/ ).
The amplitude values are between 0 and 1.
Errors in fitted parameters were determined by bootstrapping 1000 random samples of initial data, fitting the resultant values with a normal distribution, and determining the standard deviation, . The error for a 3 was obtained by error propagation (Equation 3).
Error bars in the histograms visualizing the maximum likelihood fits represent the standard deviation (Equation 4), where N is the number of events and P is the event probability.
To assess the significance of the observed changes in the absence versus the presence of Crc, a two-tailed, unpaired t-test was performed to calculate the P value between given conditions using GraphPad Prism 6.

Cryo-EM sample preparation, imaging and data analysis
For the Cryo-EM studies, amiE105 RNA (nt -45 to +60 of amiE mRNA with respect to the A (+1) of the AUG start codon was prepared by in vitro transcription using T7 RNA polymerase, and then purified on a 8M urea 8% polyacrylamide gel. The DNA template was amplified by PCR using the primer pair R185 (5 -TCTAGACGTAATACGACTCACTATAGGGCCTT-TTTTCGTCCCGAAAAAATAACAAC-3 ) and Z172 (5 -GTAGTTGACCACCGCCACTC-3 ), wherein the forward primer R185 contained the T7 promoter sequence. When compared with amiE102, the yield obtained with the amiE105 template was ∼2-fold higher. The RNA amiE105 is identical to amiE102 except for the 3 extension and that the artificial two G-nucleotides at the 5 -end of amiE102 RNA are replaced by 5 -GGGCC-3 .., which is identical to the authentic sub-sequence of amiE. For Hfq and Hfq/Crc assembly on amiE105, the RNA was annealed at 50 • C for 3 minutes. Hfq and Hfq/Crc were mixed at 2:1 and 2:11 ratios prior to addition of amiE105 (800 nM final concentration). After incubation on ice for 1 h, the mixture was diluted 7-fold before loading onto grids. Graphene oxide (GO) grids were prepared from regular Quantifoil r1.2/1.3 grids. A 2 mg/ml graphene oxide dispersion (Sigma Aldrich) was diluted 10-fold and spun at 300 × g for 30 s to remove larger aggregates. The dispersion was then diluted 10-fold before applying 1 l to glow-discharged grids (0.29 mbar, 15 mA, 2 min, Pelco Easiglow glow discharger). After drying, 3 l of the sample was applied to the GO grids and after 30 s of incubation, excess sample was blotted away and frozen in liquid ethane (Vitrobot markIV (FEI)). The grids were screened on a 200 kV Talos Arctica (FEI) (Cryo-EM facility, Department of Biochemistry, University of Cambridge), and datasets collected on a 300 kV Titan Krios (FEI) with a K3 (Gatan) direct electron detector (BioCem facility, Department of Biochemistry, University of Cambridge). For the 2 Hfq/1 Crc/1 amiE105 dataset, a set of 1007921 particles was extracted by Warp (37), and further processed with CryoSparc v2.15 (38) via multiple rounds of 2D classification. Due to significant preferential orientation bias of the particles in vitreous ice, as illustrated by the large discrepancy in particle numbers contributing to the 2D class averages in Figure 3A, no 3D reconstructions were generated. For the 2 Hfq/11 Crc/1 amiE105 dataset, 2757685 particles were extracted by Warp (37), and further processed with CryoSparc v2.15 (38) via multiple rounds of 2D classification and 3D heterogeneous refinement. Prior to heterogeneous refinement, a 3D ab initio reconstruction was generated in CryoSparc v2.15 from a subset of the particles. The maps from heterogeneous refinement were lowpass filtered to 15Å for depiction, and a selection of 2D class averages was used to measure compaction of the Hfq/amiE105 intermediate upon Crc binding ( Figure 3B).

In vitro translation
In vitro translation was performed with the PURExpress in vitro protein synthesis kit (New England BioLabs) in the presence of 20 U RiboLock RNase Inhibitor (Thermo Fisher Scientific). 5 pmol of in vitro transcribed amiE-flag mRNA (2) was used in a 12.5 l reaction and 5 pmol of either in vitro transcribed amiE-flag and oprD-flag mRNA were used in a 25 l reaction. For in vitro transcription of oprD-flag mRNA, first, a PCR fragment was generated with the primer pair Y133 (5 -TCTAGACGTAATACGA CTCACTATAGGCTAGCCGTCACTGCGGCAC-3 ) and A131 (5 -TTTTTTCTATTATCAtcacttgtcgtcatcgtcttt gtagtcCAGGATCGACAGCGG -3 ). The forward primer contained a T7 promoter sequence and the reverse primer contained the Flag-tag encoding sequence (lower-case letters). The amplified PCR fragment was use as template in an in vitro transcription reaction using the AmpliScribe T7-Flash Transcription Kit (Epicentre Biotechnologies). Increasing amounts of purified Hfq protein as indicated in Figure 4 were added alone or in the presence of a 2-fold molar excess of Crc protein. The protein amount was doubled in the reactions were both mRNAs were present. The proteins and mRNAs were added simultaneously. After 1 h of incubation at 37 • C, 5 l of the reaction was mixed with 5 l protein loading buffer and AmiE-Flag and OprD-Flag synthesis, respectively, was assessed by Western-blotting. The proteins were separated on a 12% SDS-polyacrylamide gel and blotted onto a nitrocellulose membrane. The blot was blocked with 5% dry milk in TBS buffer, and then probed with rabbit anti-Flag-tag polyclonal antibodies (Antibodies Online GmbH) and anti-rabbit IgG HRP-linked antibody (Cell Signaling Technology). The antibody-antigen complexes were visualized with the Clarity Max TM Western ECL Substrate (BioRad). The quantification of the signals was performed with the Volume Tools of ImageLab software 5.2.1 (BioRad).

Determination of amiE stability
The stability of amiE mRNA was determined in strains PAO1 (39) and PAO1 crc (25) grown in BSM medium supplemented with 40 mM acetamide and 40 mM succinate (+ CCR) to an OD 600 of 1.0 followed by addition of rifampicin (100 g/ml, final concentration). The samples were withdrawn at several times thereafter ( Figure 5 and Supplementary Figure S4), and total RNA was extracted by using the hot phenol method (40). 2 g of total RNA was used for cDNA synthesis of amiE and 16S rRNA. Finally, 2 l of cDNA were used as a template for PCR amplification of amiE and 16S rRNA (internal control) as previously described (2). The quantification of the PCR signals was performed with the Volume Tools of ImageLab software 5.2.1 (BioRad). The values for the amiE signals were divided by the corresponding values of the 16S rRNA signals and time 0 was set as 100%. The degradation of amiE mRNA in PAO1 followed a first order kinetics and the half-life for was calculated by t 1/2 = ln 2/k decay , whereas the k decay values were extracted from the exponential trendlines (line of best fit) to the graphs determined in Excel. The amiE mRNA was not degraded with a first order kinetics in strain PAO1 crc. Therefore, the half-life was estimated by determining the slope (k* decay ) from a linear trendline and the half-life was calculated by t 1/2 = 50/k* decay .

Crc increases the population of stable Hfq/RNA complexes
Hfq/Crc assemblies are most likely forming on nascent mRNA (1), and thus on a longer sub-sequence than used in our previous studies (24). Therefore, we used a 100 nt amiE fragment to measure the protein binding dynamics by single molecule fluorescence co-localization. This RNA includes the authentic amiE mRNA sequence spanning nucleotides -40 to +60 with respect to the A (+1) of the AUG start codon, two extra 5 -terminal G nucleotides, and a 3 extension for annealing to a Cy5 labeled DNA oligonucleotide. (hereafter amiE102); Figure 1A and B). The amiE102 RNA comprises the proximal amiE 6ARN motif as well as further AA(R)N motifs for the distal site of Hfq that are located downstream of the AUG (2; Figure 1A).
To assess the impact of Crc on the Hfq/amiE102 complex a single-molecule fluorescence assay was established to monitor RNA binding to Hfq with 0.1 s resolution ( Figure 1B). For these assays, we used mixed hexamers of Hfq wt and Hfq S65C -Flag (Hfq*) that contained both protomers in a 1:1 ratio (Supplementary Figure S1A). Neither the C-terminal Flag-tag nor the S65C substitution affected the interaction with Crc and amiE102 RNA, i.e. Hfq*/Crc/amiE102 complex formation (Supplementary Figure S1B) or the ability to repress a translational amiE::lacZ reporter gene (Supplementary Figure S1C). Hfq* was immobilized on a microscope slide via biotinylated anti-FLAG antibodies. The amiE102 mRNA frag-Nucleic Acids Research, 2021 7 ment was transcribed with a 3 extension that was hybridized to a Cy5-labeled DNA oligomer. Therefore, RNA binding to immobilized Hfq* could be observed by localization of the Cy5-labeled amiE102 RNA/DNA complex near the slide surface ( Figure 1B).
The binding assays of amiE102 to immobilized Hfq* revealed multiple Cy5 signals appearing in the field of view ( Figure 1C). They represent specific Hfq*/amiE102 interactions, as these events were not observed in control experiments lacking Hfq* or the anti-FLAG antibodies (Supplementary Figure S2A). Maximum likelihood fits of the dwell times for these interactions revealed that amiE102 forms two types of complexes with Hfq; unstable complexes with a lifetime of ∼2 s, and stable complexes with ∼60 s lifetime ( Figure 1E and Supplementary Figure S2C). The unstable complexes were slightly more common than long-lived complexes during the first 3 min after RNA injection (a short = 0.56 ± 0.05; Supplementary Table S1). The Cy5 photobleaching lifetime (173 s; Supplementary Figure S2B) was ∼3 times L (long) (∼60 s), indicating that photobleaching might lead us to slightly underestimate the lifetimes of the most stable complexes, but does not account for the shortlived complexes that we observed.
Next, we tested whether Crc alters the binding of Hfq* to amiE102. The presence of 100 or 500 nM Crc did not change the lifetimes of the stable and unstable Hfq*/amiE102 interactions, suggesting that Crc does not affect the mode of Hfq/RNA binding ( Figure 1D and E). However, Crc increased the population of stable complexes (a long = 0.7 ± 0.04; Figure 1F). The higher concentration of Crc (500 nM) did neither affect the lifetimes nor the fractions of unstable and stable complexes ( Figure 1E and F), suggesting that the complexes were saturated with Crc. Thus, in the presence of Crc, a higher fraction of the Hfq*/amiE102 complexes are long-lived, indicating that Crc locks Hfq onto amiE102 RNA, preventing the complexes from dissociating.

Upstream and downstream AA(R)N motifs in amiE contribute to stable interactions with Hfq
Previous work revealed that the amiE 6ARN motif present in the translation initiation region of amiE ( Figure 1A) is required for translational repression (2,25) and that it interacts with the distal face of Hfq (2,22). In addition, recent structural studies suggested that Hfq binding to the amiE 6ARN motif presents the RNA to Crc, thereby permitting the assembly of higher order Hfq/Crc/RNA complexes (24). To further assess the role of the amiE 6ARN motif for Hfq*/Crc/amiE102 assembly, three of the AAN triplets in the amiE 6ARN motif were deleted (hereafter, amiE 5 AANmut; Figure 2A). This deletion is not expected to impose significant changes in the predicted RNA secondary structure. The dwell time analyses revealed that the vast majority of the Hfq/amiE 5 AANmut binding events were transient and likely non-specific (∼0.2 s, Figure 2C, E, F). We also observed some short and long-lived complexes with lifetimes similar to those of amiE102 RNA ( Figure 2E and Supplementary Figure S3). However, these short and long-lived complexes were much less prevalent for amiE 5 AANmut RNA than for amiE102 RNA ( Figure 2F, Supplementary Table S1).  . At an OD 600 of 1.0, rifampicin was added to a final concentration of 100 g/ml and samples were withdrawn for total RNA extraction at the times indicated. The remaining amounts of amiE mRNA in strains PAO1 (blue graph) and PAO1 crc (red graph) were normalized to that of 16S rRNA at different times after addition of rifampicin (see also Supplementary Figure S4). The average and standard-deviation from two biological replicates are presented.

Nucleic Acids Research, 2021 9
The amiE102 RNA also comprises Hfq binding motifs situated downstream of the start codon ( Figure 1A). We therefore asked whether these motifs also contribute to the interactions with Hfq. To address this, a variant of amiE102 was generated that lacked the (ARN) 3 motif and the distal hairpin comprising one putative ARN motif in the loop (hereafter, amiE 3 ARNmut; Figure  2B). The amiE 3 ARNmut RNA also formed both shortand long-lived complexes ( Figure 2D and E). However, like amiE 5 AANmut, amiE 3 ARNmut formed noticeably fewer stable complexes with Hfq than amiE102, albeit marginally more than amiE 5 AANmut (a L , 0.14 ± 0.05 versus 0.05 ± 0.03; Figure 2F). Thus, Hfq is more likely to stably interact with the upstream amiE 6ARN motif than with the downstream (ARN) 3 motif. Nevertheless, most binding events between amiE 3 ARNmut and Hfq were transient and non-specific (∼0.4 s; Figure 2F), indicating that both A-rich motifs are recognized by Hfq and that they both contribute to its stable association with amiE102.
When Crc was injected together with amiE 5 AANmut RNA, no change in binding was noted ( Figure 2E and F; Supplementary Table S1). Crc only slightly decreased the fraction of non-specific amiE 3 ARNmut RNA complexes in favor of stable Hfq complexes (a L , 0.23 ± 0.04 versus 0.14 ± 0.05; Supplementary Figure S3A). These results indicated that Crc can only stabilize Hfq/RNA complexes when Hfq suitably interacts with the AA(R)N motifs located up-and downstream of the amiE start codon, indicating that Crc repression depends on sequence-specific recognition of the mRNA by Hfq.

The presence of Crc results in compact Hfq/Crc/amiE assemblies
Next, Cryo-EM studies were performed to visualize the impact of Crc on Hfq/amiE105 complexes. 2D classification of Hfq/amiE105 assemblies indicated the presence of up to three Hfq hexamers that interacted with amiE105 RNA. The extended assemblies appeared as 'loose beads on a string' and revealed different relative orientations of Hfq in the folding intermediate ( Figure 3A). The preferred particle orientations on the grid did not allow for interpretable 3D maps to be reconstructed, but the extendedness of the subassemblies could be measured directly from the 2D class averages with 158Å and 131Å for the top and bottom classes in Figure 3A, respectively.
When Crc was added to Hfq and amiE105, its presence resulted in more compact Hfq/Crc/amiE105 assemblies, consisting of complexes with apparent compositional stoichiometries of 3 Hfq: 4 Crc: 1 amiE105 ( Figure 3B; Dendooven et al., unpublished). This increased compactness of the Hfq/Crc/amiE105 assemblies was apparent from a decreased extendedness of the 2D class averages with 106 and 104Å for the top and bottom class averages in Figure 3B, respectively. The 2D class average in the top panel of Figure 3B displays the greatest intramolecular length in the Hfq/Crc/amiE105 assembly, which is significantly lower than the sizes measured from the 2D class averages of the Hfq/amiE105 assembly ( Figure 3A). Lastly, the compactness is also rationalized by preliminary 3D reconstructions ( Figure 3C), where Crc molecules form proteinogenic bridges between the Hfq hexamers in the complex, as was recently observed for the Hfq/Crc/amiE 6ARN assembly (24; Figure 3D).

Increased translational repression in the presence of Crc
Previous in vivo studies with the Pae strains PAO1 wt and PAO1 crc revealed that translational repression of different Hfq/Crc regulated reporter gene fusions was reduced ∼ 50% in the absence of Crc, qualifying Crc as a translational co-repressor (2,9). To assess more directly the contribution of Crc in Hfq-mediated translational repression, full length amiE transcript abutted to a Flag-encoding sequence was translated in vitro in the presence of increasing molar ratios of Hfq and Hfq/Crc, respectively. Although, a 4-fold molar excess of Hfq-hexamer over amiE-flag mRNA did result in significant translational repression, translation was not completely abolished ( Figure 4A, lane 5). In contrast, a 3and 4-fold molar excess of Hfq resulted in complete translational shut off in the presence of Crc ( Figure 4A, lanes 9 and 10). To verify that Hfq/Crc repression is restricted to Crc targets such as amiE, an in vitro translation competition assay programmed with amiE-flag mRNA and oprDflag mRNA was performed. The oprD translation was recently shown to be down-regulated by two Hfq-dependent small RNAs (10). Hfq alone or the presence of Hfq/Crc did not impact translation of oprD (9). Consistent with these observations, translation of oprD-flag mRNA was hardly affected in the presence of increasing concentrations of Hfq and Crc, whereas translation of amiE-flag mRNA ceased in the presence of Hfq/Crc ( Figure 4B). These results suggested that co-binding of Hfq and Crc and the presumed assembly of higher order Hfq/Crc/RNA complexes, similar to those shown in Figure 3B and C, ensures efficient translational repression of the target mRNA.

Accelerated decay of amiE mRNA in the presence of Crc
Translational repression of bacterial mRNAs very often results in their rapid degradation (41,42). Therefore, we next tested the stability of the amiE transcript in the presence and absence of Crc in strains PAO1 and PAO1 crc. Under conditions of CCR, the half-life of amiE mRNA was 1.7 ± 0.1 min in strain PAO1, which is in the presence of Crc. In contrast, the half-life increased >6-fold to 10.4 ± 0.9 min in strain PAO1 crc ( Figure 5 and Supplementary Figure S4). Together with the results shown above, we conclude that Crc compacts and stabilizes Hfq/Crc/RNA assemblies, strengthening translational repression, which leads to rapid degradation of target mRNAs during CCR when their encoded functions are dispensable.

DISCUSSION
In Pae, the concerted action of Hfq and Crc steers cell metabolism by governing the hierarchical utilization of carbon sources. Previous studies on the Hfq/Crc repressome strongly indicated that Crc works only in concert with Hfq, i.e. Hfq acts as the principle translational repressor and Crc has an ancillary role as a co-repressor (1,22,23). RNA-seq based comparative transcriptome analyses with the Pae mutant strains PAO1hfq-and PAO1Δcrc revealed many overlapping mRNA targets, regulation of which is governed by both, Hfq and Crc during CCR (22). In addition, a ChIP-seq approach showed that Hfq and Crc co-associate with nascent transcripts, whereas Crc was not observed on RNAs in the absence of Hfq (1). These in vivo findings are in accord with previous in vitro results, showing that Crc is devoid of RNA binding activity (29), and that the association of Crc with RNA depends on Hfq (21,22). Moreover, we have performed an RNA-seq based comparative transcriptome analysis with the Pae strains PA01 hfq and PA01 hfq crc to verify that Crc does not exert an independent regulatory function. As shown in Supplementary  Table S2, using standard thresholds of significance for the analysis only eight transcripts were differentially abundant in the presence of Crc in strain PA01 hfq when compared with strain PA01 hfq crc under conditions of CCR. As none of these transcripts was identified by ChIP-seq using anti-Crc antibodies (1), it is rather unlikely that these genes are directly regulated by Crc. However, at this stage we cannot exclude that Crc interacts with another regulatory protein(s) that is involved in regulation of these transcripts.
Although these studies underlined the cooperative action of Hfq and Crc, the question arises as to why two proteins are needed to efficiently repress translation of a target mRNA. To gain deeper insight into the mechanism of regulation, we developed a single-molecule assay that visualizes individual binding events on time scales of milliseconds. Because single-molecule studies enable characterization of transition states, they are a valuable tool for understanding molecular processes. In the context of Hfq, this technique was recently used to demonstrate unfolding of structured mRNAs during base-pairing with small RNAs (43).
Here, the single-molecule assay revealed that the majority of Hfq/amiE102 complexes are intrinsically perturbed, which leads to a high dissociation rate ( ∼2 s; Figure 6B). Our assays strongly corroborate the previous notion (24) that Crc engages transient, pre-organized Hfq/amiE102 complexes and shifts the equilibrium towards assemblies with increased stability ( ∼ 60 s) so that the effector complex blocks translation more efficiently ( Figure 6C). Thus, the cooperation of Hfq with Crc effectively stabilizes the repressive complex, excluding the 30S ribosome more effectively than can be achieved by Hfq alone ( Figure 6C).
When compared with the Hfq/amiE105 complex (Figure 3A), the increased stability of the Hfq/Crc/amiE102 assemblies determined by the single molecule experiments can be readily reconciled with the 2D and 3D Cryo-EM averages of the complex formed with both abundant and sparse concentrations of Crc ( Figure 3B and C). In the 3D reconstructions, 3 Hfq hexamers and 4 Crc protomers can be discerned that build a compact assembly ( Figure 3C). In fact, the observed compaction is anticipated to increase buried surface areas of the complex and is a hallmark of increased thermodynamic stability. Moreover, the arrangement of the Hfq hexamers on amiE105 suggests that Hfq binding is not restricted to the amiE 6ARN motif upstream of the AUG but involves as well cognate sequences of the immediate coding region. This observation is supported by the single molecule experiments with amiE 3 ARNmut RNA that implicated the distal Hfq-binding motif(s) in formation of stable Hfq/amiE102 complexes (Figure 2).
How does Crc engage as a translational co-repressor? Recent Cryo-EM studies showed that there are limited contacts between Crc and the rim of Hfq (24). Crc mainly contacted exposed bases at the 'N' position of the ARN motif and the RNA backbone in the Hfq/Crc/amiE 6ARN Nucleic Acids Research, 2021 11 complex (24). Consequently, we suggested that complex assembly occurs in a checklist-like manner through numerous small contacting surfaces, and when the RNA target is presented by Hfq in a specific, well-defined configuration. It can be assumed that the location of Hfq binding motifs as well as the length and structure of an RNA molecule determine the meandering of it around one or more Hfq hexamers. Thus, the presentation mode of the RNA substrate by Hfq might impact Crc assembly. Our studies with the amiE 5 AANmut and amiE 3 ARNmut RNAs support this notion. Hfq bound to these RNAs dynamically, forming mostly transient interactions ( ∼ 0.2-0.4 s; Figure 6A). The lack of the three upstream AAN or the three downstream ARN motifs most likely prevented proper presentation of the RNA by Hfq, therefore the addition of Crc did not result in increased stability of the complex. Hence, this observation underlines the notion that Crc can only engage preformed Hfq/RNA complexes.
Autogenously acting translational repressors generally employ molecular mimicry and operate either by competing with initiating ribosomes or by an entrapment mechanism, whereby the formation of a productive translation initiation complex is impeded (44). A well-studied example of the first type is the E. coli threonyl-tRNA-synthetase (ThrRS), which binds as a dimer to the operator on its own mRNA so that the N-terminal domain sterically prevents ribosome binding (45). The second type of translational repressors are exemplified by E.coli ribosomal proteins, such as S4 and S15. The S4 autoregulatory feedback mechanism involves allosteric changes in the mRNA structure, affecting the distribution of translationally active and inactive mRNA (46). Similarly, S15 binding to a pseudoknot structure on its own mRNA prevents formation of an active 30S complex by abrogating productive interactions of the anti-codon of initiator tRNA and the start codon on mRNA (47). Regardless of the different mode of action, the affinity of S4 (48) and of ThrRS (49) for their own mRNA is similar as that for the 30S subunit for a typical ribosome binding site (48,50). Thus, low concentrations of the repressor may allow efficient feedback regulation.
In contrast, effective translational repression by Hfq/Crc of undesirable mRNAs, such as amiE during CCR, or translational regulation by RsmA (51) would benefit from a higher affinity of the repressor than that of the 30S subunit for target mRNAs. Indeed, the affinity of Hfq/Crc and RsmA for target mRNAs was ∼5-fold (22) and ∼2fold higher (3), respectively, than that of 30S subunits for a typical ribosome binding site (48,50). The formation of higher order complexes as observed for Hfq/Crc (Figure 3) or dimers as inferred for Pae RsmA (52) and shown for P. fluorescens RsmE (18) appear to contribute to the stability of the complexes, and thus to an increased competition with 30S subunits. Moreover, these stable repressive complexes will most likely result in a translationally inactive mRNA that is kinetically trapped and subject to rapid degradation as shown for amiE mRNA (Figures 4, 5 and 6C).
Although Hfq is well-known to act with sRNAs, this study shows how Hfq can collaborate with another protein to down-regulate mRNA expression. Thus, to effect robust translational control Hfq appears to need a partner--either a sRNA or another protein--to lock it onto the mRNA.
Finally, assembly of Hfq/Crc on nascent mRNAs may provide a means for efficient competition with small RNA substrates of Hfq. In the complexes shown in Figure 3B and C, it is most likely that sRNA binding sites on the proximal side and the rim of Hfq (53) are shielded by the winding mRNA. We hypothesize that several AA(R)N motifs in the 5 UTR and/or the immediate 5 coding region of a mRNA ( Figure 1A) result first in assembly of several Hfq hexamers followed by suitable presentation of the mRNA to and assembly of Crc protomers. As previously suggested (22), the sequestration of sRNA binding sites on Hfq would allow the formation of stable repressive Hfq/Crc/mRNA complexes that are inaccessible to invading sRNAs.