1. Biochemistry and Chemical Biology
  2. Microbiology and Infectious Disease
Download icon

Dynamic interactions between the RNA chaperone Hfq, small regulatory RNAs, and mRNAs in live bacterial cells

  1. Seongjin Park
  2. Karine Prévost
  3. Emily M Heideman
  4. Marie-Claude Carrier
  5. Muhammad S Azam
  6. Matthew A Reyer
  7. Wei Liu
  8. Eric Massé  Is a corresponding author
  9. Jingyi Fei  Is a corresponding author
  1. Department of Biochemistry and Molecular Biology, The University of Chicago, United States
  2. RNA Group, Department of Biochemistry, University of Sherbrooke, Canada
  3. Institute for Biophysical Dynamics, The University of Chicago, United States
Research Article
  • Cited 0
  • Views 1,262
  • Annotations
Cite this article as: eLife 2021;10:e64207 doi: 10.7554/eLife.64207
Voice your concerns about research culture and research communication: Have your say in our 7th annual survey.

Abstract

RNA-binding proteins play myriad roles in regulating RNAs and RNA-mediated functions. In bacteria, the RNA chaperone Hfq is an important post-transcriptional gene regulator. Using live-cell super-resolution imaging, we can distinguish Hfq binding to different sizes of cellular RNAs. We demonstrate that under normal growth conditions, Hfq exhibits widespread mRNA-binding activity, with the distal face of Hfq contributing mostly to the mRNA binding in vivo. In addition, sRNAs can either co-occupy Hfq with the mRNA as a ternary complex, or displace the mRNA from Hfq in a binding face-dependent manner, suggesting mechanisms through which sRNAs rapidly access Hfq to induce sRNA-mediated gene regulation. Finally, our data suggest that binding of Hfq to certain mRNAs through its distal face can recruit RNase E to promote turnover of these mRNAs in a sRNA-independent manner, and such regulatory function of Hfq can be decoyed by sRNA competitors that bind strongly at the distal face.

eLife digest

Messenger RNAs or mRNAs are molecules that the cell uses to transfer the information stored in the cell’s DNA so it can be used to make proteins. Bacteria can regulate their levels of mRNA molecules, and they can therefore control how many proteins are being made, by producing a different type of RNA called small regulatory RNAs or sRNAs. Each sRNA can bind to several specific mRNA targets, and lead to their degradation by an enzyme called RNase E. Certain bacterial RNA-binding proteins, such as Hfq, protect sRNAs from being degraded, and help them find their mRNA targets.

Hfq is abundant in bacteria. It is critical for bacterial growth under harsh conditions and it is involved in the process through which pathogenic bacteria infect cells. However, it is outnumbered by the many different RNA molecules in the cell, which compete for binding to the protein. It is not clear how Hfq prioritizes the different RNAs, or how binding to Hfq alters RNA regulation. Park, Prévost et al. imaged live bacterial cells to see how Hfq binds to RNA strands of different sizes.

The experiments revealed that, when bacteria are growing normally, Hfq is mainly bound to mRNA molecules, and it can recruit RNase E to speed up mRNA degradation without the need for sRNAs. Park, Prévost et al. also showed that sRNAs could bind to Hfq by either replacing the bound mRNA or co-binding alongside it. The sRNA molecules that strongly bind Hfq can compete against mRNA for binding, and thus slow down the degradation of certain mRNAs.

Hfq could be a potential drug target for treating bacterial infections. Understanding how it interacts with other molecules in bacteria could provide help in the development of new therapeutics. These findings suggest that a designed RNA that binds strongly to Hfq could disrupt its regulatory roles in bacteria, killing them. This could be a feasible drug design opportunity to counter the emergence of antibiotic-resistant bacteria.

Introduction

In all kingdoms of life, RNA-binding proteins (RBPs) orchestrate the post-transcriptional fates of RNAs by modulating their turnover, structure, and localization, and often as a companion of regulatory RNAs. As one of the most abundant RBPs in bacterial cells, Hfq is an important and prevalent post-transcriptional gene regulator (Brennan and Link, 2007; Vogel and Luisi, 2011; Updegrove et al., 2016). Acting as a chaperone of sRNA-mediated gene regulation, Hfq protects small RNAs (sRNAs) from degradation and promotes sRNA–mRNA duplex formation (Brennan and Link, 2007; Vogel and Luisi, 2011; Updegrove et al., 2016). Binding of sRNAs to target mRNAs further leads to changes in the translation activity and the stability of the mRNAs (Storz et al., 2011; Wagner and Romby, 2015). Moreover, other functions of Hfq in regulating translation and degradation of mRNAs independent of the sRNA-mediated regulatory pathway have also been reported (Urban and Vogel, 2008; Vytvytska et al., 2000; Pei et al., 2019; Hajnsdorf and Régnier, 2000; Mohanty et al., 2004). Loss of Hfq compromises the fitness of bacterial cells, especially under harsh conditions, and abolishes the virulence of bacterial pathogens (Sobrero and Valverde, 2012; Tsui et al., 1994).

Hfq binds broadly to cellular mRNAs, sRNAs, and ribosomal RNAs (rRNAs) (Sittka et al., 2008; Tree et al., 2014; Chao et al., 2012; Melamed et al., 2016; Melamed et al., 2020; Andrade et al., 2018; Dos Santos et al., 2020). Hfq interacts with RNAs through multiple interfaces of its homohexameric structure. The surface containing the N-terminal α-helices is referred to as the ‘proximal face’, whereas the opposite surface is referred to as the ‘distal face’, and the outer ring as the ‘rim’ (Figure 1a). The proximal face preferably binds U-rich sequences, while the distal face prefers A-rich sequences, with the exact composition of the A-rich motif varying from species to species (Horstmann et al., 2012; Link et al., 2009; Robinson et al., 2014; Salim et al., 2012; Someya et al., 2012). The rim can also interact with UA-rich RNAs through the patch of positively charged residues (Dimastrogiovanni et al., 2014; Murina et al., 2013; Peng et al., 2014). Finally, the unstructured C-terminal end of Hfq can also interact with certain RNAs to promote the exchange of RNAs (Robinson et al., 2014; Dimastrogiovanni et al., 2014; Santiago-Frangos et al., 2016). The most refined model describing the interactions between Hfq and sRNAs/mRNAs has sorted sRNAs into two classes (Schu et al., 2015). The proximal face of Hfq is generally important for the binding of the sRNAs through their poly-U tail of the Rho-independent terminator, independent of sRNA class. Class I sRNAs (such as RyhB and DsrA) then use the rim as the second binding site, whereas class II sRNAs (such as ChiX and MgrR) use the distal face as the second binding site (Schu et al., 2015). In addition, the preferred target mRNAs of the two classes of the sRNA are proposed to have the complementary binding sites on Hfq, i.e., class I sRNA-targeted mRNAs (or class I mRNAs) binding to the distal face, and class II sRNA-targeted mRNAs (or class II mRNAs) binding to the rim, to efficiently form sRNA–mRNA complexes (Schu et al., 2015). As a pleiotropic regulator, Hfq establishes additional interactions with other essential protein factors. Particularly, RNase E, the key ribonuclease in the bacterial inner membrane for processing and turnover of rRNAs and mRNAs, is known to interact with Hfq through its C-terminal scaffold region (Bruce et al., 2018; Ikeda et al., 2011; Morita et al., 2005; Worrall et al., 2008). The Hfq-RNase E interactions can promote the degradation of the sRNA-targeted mRNA (Morita et al., 2005; Afonyushkin et al., 2005; Pfeiffer et al., 2009; Prévost et al., 2011).

Figure 1 with 3 supplements see all
Diffusion and localization of Hfq during exponential growth.

(a) Schematic representation of Hfq with three RNA binding faces indicated. (b) A representative example of WT Hfq-mMaple3 in WT rne background in a single cell during exponential growth under no treatment (NT) condition. Nucleoid is stained with Hoechst in live cells. 2D reconstructed image of Hfq-mMaple3 is shown in the black background. Different diffusion trajectories from tracking algorithm are shown in different colors (‘Traj’). One-step displacement (osd) speed map (unit: µm/s) is shown as a scatter plot where different colors represent different speeds at each position, and the white curves represent the nucleoid regions detected by Hoechst staining. The scale bar represents 1 µm. (c) Enrichment of Hfq localization is calculated for cytoplasm, membrane, and nucleoid regions under NT condition. (d) Average osd speed of Hfq within the cytoplasm, membrane, and nucleoid regions under NT condition. Error bars in all plots represent the standard deviation (s.d.) from two experimental replicates, with each data set containing ~20,000 trajectories from ~80 cells.

Figure 1—source data 1

Single cell speed (b) and average enrichment and osd speed (c and d).

https://cdn.elifesciences.org/articles/64207/elife-64207-fig1-data1-v2.xlsx

While Hfq is an abundant RBP in bacterial cells (Ali Azam et al., 1999; Kajitani et al., 1994), it is still considered to be limiting, given the abundance of cellular RNAs. Particularly, in vitro studies on specific sRNAs demonstrate that Hfq binds RNAs tightly with a dissociation constant in the range of nanomolar, and the Hfq–RNA complexes are stable with a lifetime of >100 min (Fender et al., 2010; Salim and Feig, 2010Olejniczak, 2011). However, under stress conditions, induced sRNAs can regulate target mRNAs within minutes, raising a long-standing question of how sRNAs can rapidly access Hfq that might be tightly bound by pre-existing cellular RNAs. To address this question, a model of RNA exchange on Hfq, that is, an RNA can actively displace another RNA from Hfq, was proposed to account for the fast sRNA-mediated stress response (Vogel and Luisi, 2011; Updegrove et al., 2016; Wagner, 2013). While in vitro biophysical experiments can measure the affinity of RBP binding to many different RNAs under many different controllable conditions, it is difficult to replicate the concentrations, compartmentalization, crowding, competitive binding, and changes in cellular conditions that can affect the behavior and function of RBPs in live cells. Therefore, the mechanism(s) that can recycle Hfq from pre-bound RNAs in live cells remains to be elucidated.

To address this question in a cellular context, we measured the diffusivity of Hfq in live Escherichia coli cells, using single-molecule localization microscopy (SMLM) (Manley et al., 2008), with a rationale that the diffusivity is affected by the molecular weight of the molecules (Mika and Poolman, 2011), and therefore can report the interactions between Hfq with different cellular components. By measuring Hfq diffusivity under a variety of cellular conditions in combination with other biochemical assays, we demonstrate that Hfq dynamically changes its interactions with different RNAs. Specifically, the two classes of sRNA can gain access to mRNA pre-bound Hfq through different mechanisms. Finally, our data suggest that binding of Hfq to certain mRNAs through its distal face can recruit RNase E to promote turnover of these mRNAs in a sRNA-independent manner.

Results

Cellular Hfq freely diffuses in the absence of stress

Hfq was tagged by a photo-switchable fluorescent protein (FP), mMaple3 (Wang et al., 2014), at the C-terminus, and the fused hfq gene was integrated into the genomic locus to replace the wild-type (WT) hfq (denoted as ‘hfq-mMaple3’, Materials and methods). The strain harboring hfq-mMaple3 showed comparable growth curve as the WT strain, whereas Δhfq strain showed a growth defect (Figure 1—figure supplement 1). In addition, Hfq-mMaple3 showed activity comparable to WT Hfq protein, as revealed by northern blots of RyhB-mediated sodB mRNA degradation and MicA-mediated ompA mRNA degradation (Figure 1—figure supplement 2).

We performed single-particle tracking using SMLM in two dimensions (2D). Images were collected at a rate of 174 frames per second with 2.4 ms exposure time for each frame. Imaging conditions and parameters for applying tracking algorithm were optimized using fixed samples as the control (Figure 1—figure supplement 3). We first tracked Hfq-mMaple3 in live cells grown at exponential phase (referred to as ‘no treatment’, or ‘NT’ case). In the NT condition, Hfq-mMaple3 exhibited a relatively uniform distribution within the cell (Figure 1b), consistent with the distribution revealed by the earlier live-cell imaging with Hfq tagged by a different FP (Dendra2) (Persson et al., 2013). Quantification of Hfq-mMaple3 localization with DNA stained by Hoechst revealed a slightly higher cytoplasm enrichment than nucleoid or membrane localization in the NT condition (Figure 1c). We did not observe a helical organization along the longitudinal direction of the bacterial cell (Taghbalout et al., 2014), membrane localization (Diestra et al., 2009), or cell pole localization (Kannaiah et al., 2019), as reported in a few fixed-cell experiments. In addition, we calculated the one-step displacement (osd) speed of individual Hfq-mMaple3 protein at each time step and plotted osd speed as a function of the cellular coordinate in a diffusivity speed map (Figure 1b). The speed map and quantification of the average osd speed suggest that Hfq diffuses similarly within the nucleoid and cytoplasmic region, but slightly more slowly in the membrane (Figure 1d), which could be due to the association with RNase E in the inner membrane.

Binding of mRNAs to Hfq decreases its diffusivity primarily through the distal face of Hfq

We first tested the effect of mRNA on Hfq-mMaple3 diffusivity by treating the cells with rifampicin (Figure 2a). Rifampicin is an antibiotic that inhibits transcription and results in the loss of most cellular RNAs. We estimated the effective diffusion coefficient (D) by fitting the linear function to the mean squared displacement (MSD) as a function of time lag (Δt) (Figure 2b). Transcription inhibition increased the diffusivity of Hfq-mMaple3 (Figure 2b), suggesting that a fraction of Hfq-mMaple3 proteins is associated with cellular RNAs, consistent with a previous report (Persson et al., 2013). We also generated several control constructs: mMaple3 protein alone, mMaple3 fused to an engineered synthetic antibody sAB-70 (Mukherjee et al., 2018), and mMaple3-fused scFv-GCN4 used in the SunTag imaging approach (Tanenbaum et al., 2014), which are not reported to bind to any RNA. None of them showed any changes in diffusivity upon rifampicin treatment (Figure 2—figure supplement 1), confirming that the change in diffusivity is not due to the change in the cellular milieu. It should be noted that under the current rifampicin treatment (200 µg/mL concentration for 15 min), mRNAs, which have an average half-live of 1–4 min (Chen et al., 2015), are preferentially degraded, compared to tRNAs (Svenningsen et al., 2017) and rRNAs (Blundell and Wild, 1971). While many sRNAs show long half-lives when target-coupled degradation is reduced in the absence of mRNAs upon rifampicin treatment (Massé et al., 2003; Zhang et al., 2002), some sRNAs do have short half-lives (Vogel et al., 2003). Therefore, rifampicin treatment might also reduce the fraction of Hfq bound by sRNAs. However, our data suggest that binding of sRNA to RNA-free Hfq or to mRNA-associated Hfq does not change the diffusion coefficients of corresponding species (see sections below). Therefore, we interpreted that the change in the diffusion coefficient upon rifampicin treatment primarily reflected the binding of mRNAs to Hfq.

Figure 2 with 2 supplements see all
Binding of mRNAs to Hfq decreases its diffusivity primarily through the distal face of Hfq.

(a) A representative example of Hfq-mMaple3 with rifampicin treatment (Rif) in a single cell. 2D reconstructed image is shown in the black background (left), and different diffusion trajectories are shown in different colors (right). The scale bar represents 1 µm. (b) Mean squared displacement (MSD) is plotted against the time lag (Δt) for Hfq-mMaple3 under NT and Rif cases. The linear fitting lines are shown. (c) Ensemble diffusion coefficients are plotted for WT and six mutants of Hfq-mMaple3 under NT and rifampicin treatment conditions. (d) mRNA-associated fraction for WT and six mutants of Hfq under NT and Rif conditions. Error bars in all plots represent the s.d. from two or three experimental replicates, with each data set containing ~5,000 trajectories (for D value calculation) or ~20,000 trajectories (for mRNA-associated fraction calculation) from ~100 cells. All fitting results are reported in Supplementary file 1.

Figure 2—source data 1

Mean squared displacement (MSD) plot for NT and Rif (b).

https://cdn.elifesciences.org/articles/64207/elife-64207-fig2-data1-v2.xlsx

We next introduced point mutations on Hfq-mMaple3 that are shown to affect RNA binding: Q8A and F42A at the proximal face, R16A and R19D at the rim, and Y25D and K31A at the distal face (Schu et al., 2015; Zhang et al., 2013), and imaged these mutant Hfq-mMaple3 proteins under NT and rifampicin treated conditions. With rifampicin treatment, all Hfq mutants exhibited similar diffusivity. However, mutations on different faces changed Hfq diffusivity in the NT case to different levels, suggesting that mutations on different faces changed the ability of Hfq to bind cellular mRNAs. Specifically, both proximal face mutants (Q8A and F42A) exhibited similar diffusivity as the WT Hfq-mMaple3; both rim mutations (R16A and R19D) had a minor increase in diffusivity under NT condition; and both distal face mutations (Y25D and K31A) led to the largest increase in the diffusivity under NT condition, with the diffusion coefficients close to the rifampicin treated case (Figure 2c). Comparison of the WT and the mutant Hfq-mMaple3 proteins supports conclusions that Hfq binds mRNAs in the cell and that binding of mRNAs is primarily achieved through the interactions with the distal face of Hfq, whereas the rim also contributes to the mRNA binding in a minor way.

Majority of Hfq proteins are occupied by mRNAs in the cell during exponential growth

To analyze the subpopulations of Hfq-mMaple3 under different conditions, we plotted one-step squared displacement (osd2) in a histogram. Consistent with D values, distribution of osd2 overall shifted to larger values with rifampicin treatment compared to the NT case (Figure 2—figure supplement 2). We fit the cumulative probability density function (CDF) of osd2 with double populations (Figure 2—figure supplement 2; Bettridge et al., 2021). WT Hfq and all mutants with rifampicin treatment showed consistently 86–98% fast population with average osd2 of 0.048–0.053 µm2, and the remaining 2–14% slow population with average osd2 of 0.0022–0.0054 µm2 (Supplementary file 1). In the NT case, WT Hfq showed (57 ± 1)% slow population and (43 ± 1)% fast population. Consistent with the previous interpretation (Persson et al., 2013), we assigned the slow population as the mRNA-associated fraction, and the fast population as mRNA-free fraction. This result is consistent with the previous hypothesis that Hfq proteins are largely occupied in the cell (Vogel and Luisi, 2011; Updegrove et al., 2016; Wagner, 2013). Y25D and K31A mutants had the most compromised mRNA binding ability, with (23 ± 10)% and (26 ± 4)% of the population being mRNA-associated under NT condition, respectively (Figure 2d). It is worth noting that as sRNA binding to RNA-free Hfq or to mRNA-associated Hfq does not change the diffusion coefficients of corresponding species (see sections below), it is possible that a subpopulation of mRNA-free or mRNA-associated Hfq might be sRNA-associated Hfq, or a sRNA-mRNA-Hfq ternary complex, respectively. In addition, the remaining 2–14% slow population after rifampicin treatment may reflect Hfq interactions with rRNAs or possibly DNA (Malabirade et al., 2018; Orans et al., 2020).

sRNAs can access mRNA-associated Hfq in a face-dependent manner

We next examined the effect of sRNAs on the diffusivity of Hfq-mMaple3. We ectopically induced expression of different sRNAs, including RyhB, a class I sRNA, ChiX, a class II sRNA, and an sRNA that is less clearly defined between these two classes, SgrS (Schu et al., 2015) from the same vector. Whereas overexpression of RyhB or SgrS did not cause any noticeable changes in the Hfq-mMaple3 diffusivity or mRNA-associated fraction, overexpression of ChiX dramatically increased its diffusivity and lowered the mRNA-associated fraction (Figure 3a and b).

Figure 3 with 2 supplements see all
sRNAs can displace mRNA from Hfq in a face-dependent way.

(a) Diffusion coefficients of WT Hfq-mMaple3 with control vector, RyhB, SgrS, WT ChiX, two ChiX mutants (with one or two AAN motif deleted) overexpressed, or Hfq Q8A-mMaple3 with WT ChiX overexpressed. All sRNAs were produced from an IPTG inducible promoter. (b) mRNA-associated fraction of Hfq, for the same cases as in (a). Error bars in all plots represent the s.d. from two experimental replicates, with each data set containing ~5,000 trajectories (for D value calculation) or ~20,000 trajectories (for mRNA-associated fraction calculation) from ~100 cells. (c) Sequences of WT ChiX and two ChiX mutants (with one or two AAN motif deleted). (d) ddPCR measurement of the abundance of RyhB, SgrS, WT ChiX, and ChiX ΔAANx1-2 mutants in the WT hfq-mMaple3 and WT ChiX in the hfq Q8A-mMaple3 background. The abundance of each sRNA is either normalized to the 16S rRNA (top) or to the control vector (bottom). Competition of (e) ChiX and (f) RyhB for mRNA-associated Hfq. 20 nM of a ptsG RNA fragment was pre-incubated with 100 nM Hfq before addition of increasing concentration of ChiX or RyhB sRNA. * marks the cases in which Hfq (100 nM) and ChiX or RyhB (100 nM) were simultaneously added to 20 nM ptsG fragment. Data is representative of three independent experiments. All fitting results are reported in Supplementary file 1.

As described above, the distal face is the primary binding site for mRNAs in the cell (Figure 2c and d). Since ChiX requires binding at both the proximal and distal faces, we expected the diffusivity of Hfq to increase after shifting from mRNA-associated Hfq to ChiX-associated Hfq. Due to the relatively small molecular weight of sRNAs (~50–300 nucleotides in length), sRNA-associated Hfq-mMaple3 has similar diffusivity as free Hfq-mMaple3. We then checked if ChiX could compete with mRNAs for binding to Hfq in vitro using electrophoretic mobility shift assay (EMSA). A radiolabeled fragment of ptsG mRNA was pre-incubated with purified Hfq protein and then chased with unlabeled ChiX. Consistent with the in vivo results, ChiX can effectively displace ptsG from Hfq (Figure 3a, b, and e).

Overexpression of RyhB or SgrS, in contrast, did not cause any significant changes in the Hfq-mMaple3 diffusivity or the corresponding mRNA-associated fraction (Figure 3a and b). We reasoned that there might be two possibilities. First, since class I sRNAs bind through the proximal face and the rim of Hfq, it can bind to the mRNA-free Hfq or co-occupy the mRNA-associated Hfq to generate sRNA-associated Hfq or sRNA–mRNA–Hfq ternary complex, respectively. Second, class I sRNAs cannot effectively compete against mRNAs for Hfq binding; therefore, most Hfq proteins remain associated with mRNAs. To distinguish these two possibilities, we examined the abundance of RyhB and SgrS compared to ChiX. Since the stabilities of RyhB and SgrS are highly dependent on Hfq (Massé et al., 2003), if the second hypothesis is correct, then we would expect a much lower cellular level of RyhB and SgrS compared to ChiX. We performed droplet digital PCR (ddPCR) in the same conditions as the tracking assays, and the result showed that RyhB or SgrS level was comparable to ChiX (Figure 3d). It should be noted that while ChiX level was almost fivefold of the RyhB level when normalized to the reads of 16S rRNA (Figure 3d, upper panel), ChiX level was about 50% of RyhB level when normalized to the reads from empty vector (representing the induction fold change) (Figure 3d, lower panel). We, therefore, reasoned that the difference between ChiX and RyhB when normalizing to the 16S rRNA was very likely due to the different efficiency during reverse transcription and PCR steps for these two targets. SgrS level was higher than ChiX with both normalizations. The observation supports that the stability of RyhB or SgrS is not compromised even though they do not displace mRNAs from Hfq, and therefore the first possibility that they either occupy the free Hfq or co-occupy on Hfq with the mRNA is more likely.

To further corroborate the observation that SgrS and RyhB can co-occupy Hfq with non-target mRNA, we performed the same EMSA competition assay using RyhB as an example (Figure 3f). When chasing with increasing concentration of RyhB, the band intensity of ptsG-Hfq complex decreased with an increased intensity of free ptsG and the appearance of an additional upper-shifted band that did not appear when chasing with ChiX (Figure 3f, red arrow). This result supports the possibility of the RyhB-ptsG-Hfq ternary complex formation. In the EMSA assay, we also observed direct displacement of ptsG fragment by RyhB, albeit less efficiently than by ChiX, which was not indicated by the in vivo imaging results. The exact cause of the discrepancy is unclear, but we speculate that it could result from the differences between the cellular conditions and in vitro setting. Nevertheless, the EMSA results still support that RyhB can have different mechanisms to gain access to mRNA-occupied Hfq, and that it is structurally possible to have RyhB co-occupy with a non-target mRNA on Hfq.

To summarize, our results collectively suggest that representative sRNAs for both class I and class II sRNAs can access mRNA-occupied Hfq in vivo. It is possible that the mechanisms can be generalized to other members of the two sRNA classes, that is, class I sRNAs can co-occupy the Hfq protein with mRNAs through different binding sites, whereas class II sRNAs can directly compete against the mRNAs at the distal face. Interestingly, fluorescence in situ hybridization (FISH) showed a much stronger signal for RyhB compared to ChiX (Figure 3—figure supplement 2), even though their levels were similar as revealed by ddPCR (Figure 3d). The weaker hybridization signal for ChiX is very likely a reflection of the larger protected region by Hfq on both distal and proximal faces, hindering FISH probe binding.

Class II sRNAs require interaction with the proximal face and a strong AAN motif to compete for Hfq binding

We next sought to understand the molecular features that made ChiX a strong competitor for Hfq binding. When overexpressed in the hfq Q8A-mMaple3 background (proximal face mutation), ChiX lost its capability to displace mRNAs from the mutant Hfq (Figure 3a and b), suggesting that additional binding affinity provided by the proximal face of Hfq is critical for displacing other RNAs from the distal face. E. coli Hfq prefers an (A-A-N)n sequence for distal face binding, where N can be any nucleotide, and each monomer binds to one A–A–N repeat (Robinson et al., 2014). ChiX contains four AAN motifs (Figure 3c). We tested the effect of AAN motifs on conferring the competitive binding to Hfq over mRNAs. We generated and overexpressed ChiX mutants with one or two AAN motif(s) deleted (Figure 3c) and found that the fraction of remaining mRNA-associated Hfq increased, when the number of AAN motifs decreased (Figure 3b and c). Notably, the levels of WT ChiX in the hfq Q8A-mMaple3 background, and the ChiX mutants in the WT hfq-mMaple3 background remained similar as the WT ChiX in WT hfq-mMaple3 background (Figure 3d), indicating that the observed difference was not due to a change in the cellular ChiX level.

Hfq is deficient in releasing mRNAs without interactions with RNase E

The C-terminal region of RNase E serves as a scaffold for the degradosome protein components (RNA helicase RhlB, enolase, and polynucleotide phosphorylase [PNPase]). Hfq has been demonstrated to interact with the C-terminal scaffold region of RNase E, although it is still under debate whether such interaction is direct or mediated by RNA (Bruce et al., 2018; Ikeda et al., 2011; Morita et al., 2005; Worrall et al., 2008). To study whether the interaction with RNase E affects the diffusivity of Hfq, we imaged Hfq-mMaple3 in two RNase E mutant strains. The rne131 mutant strain has RNase E truncated by the last 477 amino acid residues (Lopez et al., 1999); therefore, while it maintains its nuclease activity, this mutant cannot interact with Hfq. The rneΔ14 mutant has a smaller fraction of the C-terminal scaffold (residues 636–845) deleted, encompassing the Hfq, RhlB, and enolase binding regions and two RNA-binding domains (Leroy et al., 2002).

In both RNase E mutant backgrounds, the diffusivity of WT Hfq-mMaple3 became less sensitive to transcription inhibition by rifampicin compared to the WT rne background (Figure 4a). For WT Hfq-mMaple3, ~75% of Hfq-mMaple3 became mRNA-free upon rifampicin treatment in the RNase E mutant backgrounds compared to ~89% of mRNA-free Hfq in the WT rne background (Figure 4b). Hfq Y25D-mMaple3 (the distal face mutant), which is deficient in mRNA binding, showed minimal sensitivity to rifampicin treatment in the rne mutant backgrounds, the same as in the WT rne background (Figure 4a and b). These observations suggest that without the Hfq-RNase E interaction, more mRNAs remained bound to Hfq, and hint that Hfq-RNase E interaction may help recycle Hfq from the mRNA-associated form through degradation of mRNAs. To investigate whether the increased mRNA-associated form in the absence of Hfq-RNase E interaction is primarily contributed by the regulation of sRNAs, we imaged Hfq Q8A-mMaple3 in the rne mutant backgrounds. Q8A mutation in the proximal face of Hfq broadly disrupts the binding and stabilization of sRNAs (Schu et al., 2015). If sRNA-dependent regulation is the sole pathway in changing Hfq from the mRNA-associated form to the mRNA-free form, we would expect a great increase in the mRNA-associated fraction of the Q8A mutant in the rne mutant backgrounds compared to the WT Hfq with rifampicin treatment. However, we observed a minor difference in the mRNA-associated fraction between Q8A mutant (~33% considering both rne131 and rneΔ14 backgrounds) and the WT Hfq (~25%) (Figure 4b), suggesting that the recycling of Hfq from the mRNA-associated form to the mRNA-free form can be affected by Hfq-RNase E interaction in addition to the sRNA-dependent regulatory pathway.

Hfq-RNase E interaction contributes to the recycling of Hfq from the mRNA-associated form to the mRNA-free form.

(a) Diffusion coefficients are plotted for WT, Q8A, and Y25D Hfq-mMaple3 in the rne131 and rneΔ14 backgrounds under NT and Rif conditions. (b) mRNA-associated fraction of Hfq, for the same cases in (a). p-values are reported between WT rne and rne mutants under Rif condition. Error bars in all plots represent the s.d. from two experimental replicates, with each data set containing ~5,000 trajectories (for D value calculation) or ~20,000 trajectories (for mRNA-associated fraction calculation) from ~100 cells. All fitting results are reported in Supplementary file 1.

Figure 4—source data 1

Diffusion coefficients and mRNA-bound fractions of Hfq in the backgrounds of RNase E mutants.

https://cdn.elifesciences.org/articles/64207/elife-64207-fig4-data1-v2.xlsx

Hfq-RNase E interaction contributes to the degradation of Hfq-associated mRNAs

As our results above indicate that Hfq-RNase E interaction contributes to the recycling Hfq from the mRNA-associated form, likely through degradation, we hypothesized that Hfq-RNase E interaction might play a role in the turnover of specific Hfq-bound mRNAs. To test this hypothesis, we used northern blots to measure the half-lives of selected mRNAs that are known to interact with Hfq in four backgrounds: (1) WT hfq-mMaple3 + WT rne, (2) WT hfq-mMaple3 + rne131 mutant, (3) hfq Y25D-mMaple3 (distal face mutant) + WT rne, and (4) hfq Y25D-mMaple3 + rne131 mutant (Figure 5). To test whether such regulation can occur in the absence of the corresponding sRNAs, in addition to the genetic background of hfq-mMaple3 and rne, we also knocked out the corresponding sRNA regulators of the selected mRNAs (∆ryhBfnrS for sodB, ∆cyaRmicA for ompX and ∆ryhBspfΔrybB for sdhC; herein simplified as ‘∆sRNA’). The choice of knocked-out sRNAs covers all sRNAs identified in a global mapping of sRNA-target interactions for the selected mRNAs in log phase (Melamed et al., 2016). In the WT rne background, these three mRNAs showed 47% to 2.2-fold increase in the half-lives in the hfq Y25D-mMaple3 background compared to WT hfq-mMaple3 (Figure 5d). In the rne131 mutant background, while all mRNAs showed increased half-lives of 1.4 to 3.7-fold compared to the WT rne background, consistent with a compromised activity in the rne131 mutant (Lopez et al., 1999), the differences in the mRNA half-lives between WT hfq-mMaple3 and hfq Y25D-mMaple3 backgrounds were largely diminished (Figure 5e). This result indicates that in the absence of Hfq-RNase E interaction, association with Hfq or not does not change the mRNA turnover.

Figure 5 with 1 supplement see all
Hfq-RNase E interaction contributes to the regulation of mRNA degradation.

The abundance of (a) sodB, (b) ompX, and (c) sdhC mRNA in the presence of WT Hfq, Hfq Y25D, or Hfq Q8A in the WT rne or rne131 background. Corresponding sRNAs were knocked out for each of the mRNAs (ΔryhBΔfnrS for sodB, ΔcyaRΔmicA for ompX, and ΔryhBΔspfΔrybB for sdhC). Strains were grown in MOPS EZ-rich medium containing 0.2% glucose until OD600 = 0.5. Rifampicin was added as indicated by the arrow and total RNA was extracted at specific time points. ssrA, or 16S rRNA was used as loading controls. Relative abundance of mRNA quantified by densitometry as a function of time is presented in Figure 5—figure supplement 1. (d and e) Half-lives of the mRNAs determined from (a) to (c). Scatter plot represents the data points from individual replicates, and bar graph with error bars represents the mean and s.d. of four biological replicates. p-values from t-test are reported for each pairwise comparison.

To further exclude the contributions by potentially unknown sRNAs, we compared the lifetime of sodB, ompX, and sdhC in the hfq Q8A-mMaple3 background in addition to knocking out corresponding sRNAs (Figure 5). The half-lives of these mRNAs increased by 16–45% in the hfq Q8A background compared to WT hfq background, smaller than the increase observed in the hfq Y25D background (Figure 5d). The increase of mRNA half-life in the hfq Q8A background can either be due to contributions by unknown sRNA regulators or due to other possible regulatory pathways by Hfq through binding at the proximal face. One of such regulatory pathways may be Hfq-mediated polyadenylation, which involves binding of Hfq at the Rho-independent termination site and promotes mRNA degradation (Hajnsdorf and Régnier, 2000; Mohanty et al., 2004). Despite these two possibilities, the increase in the mRNA half-life due to Y25D mutation cannot be fully explained by sRNA-mediated regulation. These results collectively support that besides the sRNA-mediated pathway, Hfq can facilitate the turnover of certain mRNAs by binding to the mRNAs through the distal face and bridging them to RNase E for degradation.

RppH is not required in this Hfq-mediated mRNA turnover pathway

RNase E has two mechanisms for substrate recognition and subsequent endonuclease cleavage. RNase E can directly access and cleave RNA substrates with certain sequence preferences (Chao et al., 2017; Clarke et al., 2014). Alternatively, RNase E can recognize the 5’ monophosphate on RNA substrates for catalytic activation (Mackie, 1998; Jiang and Belasco, 2004; Bandyra et al., 2012). In this 5’-end dependent mechanism, RppH, a pyrophosphohydrolase, is needed to convert the 5’ triphosphate of the RNA substrate to 5’ monophosphate (Celesnik et al., 2007; Deana et al., 2008). To test whether this Hfq-mediated mRNA turnover is dependent on the 5’-end decapping, we used sdhC as an example, and compared its half-life in the backgrounds of ∆sRNA∆rppH and ∆sRNA∆rppH hfq Y25D.

In the ∆sRNA∆rppH background, sdhC’s half-life was ~2.6-fold of the half-life in the ∆sRNA background, suggesting that the action of RppH contributes to the endogenous turnover rate of sdhC in general (Figures 5d and 6a, b; Deana et al., 2008). However, in the ∆sRNA∆rppH background, additional Y25D mutation on Hfq caused ~90% increase in the half-life compared to that in the presence of WT Hfq (Figure 6a and b). The half-life increase caused by Hfq Y25D mutation in the ∆rppH background was comparable with the half-life increase in the WT rppH background (120% increase in half-life in the background of ∆sRNA hfq Y25D compared to ∆sRNA [Figure 5d] for sdhC). These results suggest that the decapping by RppH is not required for the Hfq-mediated regulation of mRNA turnover, at least for the case of Hfq regulation on sdhC mRNA.

Figure 6 with 1 supplement see all
Effect of RppH and ChiX on Hfq-mediated regulation on mRNA degradation.

(a) The abundance of sdhC mRNA in the presence of WT Hfq or Hfq Y25D in the ΔryhBΔspfΔrybBΔrppH background. Strains were grown in MOPS EZ-rich medium containing 0.2% glucose and rifampicin was added at OD600 = 0.5. (b) Half-life of sdhC mRNA determined from (a). (c) The abundance of sdhC mRNA in the presence of control vector, WT ChiX, and mutant ChiX with two AAN motif deleted. Strains were grown in MOPS EZ-rich medium containing 0.2% fructose and 1 mM IPTG was added at OD600 = 0.1 to induce ChiX for 1 hr before addition of rifampicin. (d) Half-life of sdhC mRNA determined from (c). Scatter plot represents the data points from individual replicates, and bar graph with error bars represents the mean and s.d. of three to four biological replicates. p-values from t-test are reported for each pairwise comparison.

sRNA that competes for Hfq binding can modulate Hfq’s ability to regulate mRNA turnover

As our model suggests that binding of Hfq to the mRNA through the distal face can regulate the mRNA turnover, we reasoned that sRNAs that can effectively compete for Hfq binding against mRNAs may decoy Hfq from this regulatory function. To test this, we again used sdhC as an example and measured its half-life in the presence of ChiX, which is a strong competitor for Hfq binding (Figure 3). In the presence of vector control, sdhC exhibited comparable half-life compared to the case without any plasmid (Figures 5d and 6c, d). The presence of WT ChiX increased the half-life by ~70%, whereas the mutant ChiX without two AAN motif did not cause a significant increase in the half-life of sdhC, consistent with its reduced binding ability to Hfq (Figures 3a, b, 6c and d). These results further support our model of Hfq-mediated regulation of mRNA turnover and demonstrate that the presence of strong Hfq binding sRNAs can modulate the strength of Hfq’s regulation.

Discussion

Using single-particle tracking, we resolved different diffusivity states of Hfq proteins in live cells, reporting on the interactions with different cellular RNAs. Specifically, free Hfq and sRNA-associated Hfq proteins (collectively termed as ‘mRNA-free Hfq’) have a high diffusivity, and association of mRNAs to form mRNA-associated Hfq or sRNA-mRNA-Hfq ternary complex (collectively termed as ‘mRNA-associated Hfq’) reduces the diffusivity of Hfq (Figure 7a). Our results are reminiscent of a previously proposed model of Hfq interacting with sRNAs and mRNAs in a face-dependent manner (Schu et al., 2015). During exponential growth, Hfq proteins are largely occupied by mRNAs. The distal face of Hfq is the primary binding site for cellular mRNAs, while the rim has a minor binding role. These observations suggest that the majority of the Hfq-bound mRNAs are class I mRNAs, and a minority are class II mRNAs, consistent with the previous findings that most of the sRNAs are class I sRNAs (Schu et al., 2015). Under the conditions when specific sRNAs are highly induced, both classes of sRNAs can easily access Hfq upon induction, albeit with different mechanisms (Figure 7b). Our data demonstrate that class II sRNAs, such as ChiX, can effectively displace class I mRNAs from the distal face, consistent with the proposed RNA exchange model. Interestingly, our data indicate that class I sRNAs do not necessarily need to displace mRNA from Hfq. Instead, they can directly co-occupy Hfq through the binding faces that are non-overlapping with the class I mRNA binding face. In both cases, the mRNA-associated Hfq proteins are in standby mode for sRNA binding if needed. The displacement of mRNA by the class II sRNA requires both the interactions at the proximal face of Hfq and higher AAN motif number to outcompete mRNAs for the binding at the distal face. In addition, we propose that the competitive binding by the class II sRNA is likely to occur stepwise, with binding at the proximal face happening first, followed by the displacement of mRNA from the distal face, which is supported by the observation that with the Hfq proximal face mutation, ChiX cannot displace mRNAs, even with a strong AAN motif.

Dynamic interactions between Hfq and cellular RNAs.

(a) Hfq facilitates the degradation of certain Hfq-bound mRNAs through the recruitment of RNase E. (b) Class I or Class II sRNAs can get access of mRNA-associated Hfq through co-occupying different binding sites of Hfq simultaneously, or displacing mRNA from the distal face of Hfq respectively.

We observed from live-cell tracking experiments that recycling of Hfq from the mRNA-associated form to the mRNA-free form upon rifampicin treatment is compromised in the RNase E mutant backgrounds, where regions including the Hfq binding site are deleted. This observation suggests that RNase E can facilitate the recycling of Hfq from the mRNA-associated form through mRNA degradation. Additional half-life measurements on a few selected mRNAs under various genetic backgrounds further demonstrate that Hfq can facilitate the turnover of certain mRNAs through binding with its distal face and recruiting RNase E, and this regulation is most likely to be independent of their corresponding sRNA regulators. Interestingly, we observed that sRNA competitors, such as ChiX, which can outcompete mRNAs for binding at the distal face, can decoy Hfq from regulating mRNA turnover, the same effect as the distal face Y25D mutation. Similar observation was reported previously that ChiX can titrate Hfq from translationally repressing transposase mRNA (Ellis et al., 2015). Considering our mutational work on ChiX-Hfq interaction, it should be possible to engineer synthetic sRNAs to tune Hfq-RNA interactions and Hfq regulatory functions in vivo.

Mechanisms of sRNA-independent Hfq-mediated regulation on mRNA turnover have been reported. First, binding Hfq, or Hfq in complex with other proteins such as Crc, at the ribosome binding site of the mRNA can repress translation (Urban and Vogel, 2008; Vytvytska et al., 2000; Pei et al., 2019), therefore indirectly increasing the mRNA degradation due to de-protection of the mRNA by translating ribosomes against RNase E. Since this translation-dependent regulation of Hfq does not require Hfq-RNase E interaction, if this mechanism applied to the mRNAs we tested, we would expect the difference in half-life between the cases of WT hfq and hfq Y25D to be similar in the WT rne and rne131 backgrounds. The observation that the difference in half-life due to Hfq binding is eliminated in the rne131 background suggests that the turnover of these mRNAs is not primarily through regulation at the translation level. Second, binding of Hfq may recruit polyA polymerase (PAP) and PNPase to stimulate polyadenylation at the 3’ end, and therefore promote degradation (Hajnsdorf and Régnier, 2000; Mohanty et al., 2004), an action that may also involve interactions with the C-terminal scaffold region of RNase E. However, this mechanism also cannot fully explain our results, as Hfq-stimulated polyadenylation prefers Hfq binding at the 3’ termini of mRNAs containing Rho-independent transcription terminator (Mohanty et al., 2004), whereas in our selected mRNAs, they do not all utilize Rho-independent termination, and the binding sites of Hfq on the tested mRNAs are within the 5’ UTR and CDS region containing A-rich motif, based on the CLIP-seq analysis of Hfq (Tree et al., 2014). Therefore, it is more likely that the regulation for the selected mRNAs is through recruitment of RNase E rather than through Hfq-stimulated polyadenylation mechanism. Nevertheless, we expect Hfq can potentially regulate mRNA turnover through a combination of these mechanisms in a gene-specific manner, which explains our observation that depending on the specific mRNA, Hfq can facilitate the turnover rate through distal face binding to different extent.

The endonuclease activity of RNase E has been shown to either be dependent on 5’-monophosphate of the RNA substrate, or independent, with the former mechanism requiring RppH to convert 5’-triphosphate cap to 5’-monophosphate. Using sdhC mRNA as an example, our results demonstrate that the 5’-monophosphate dependent pathway contributes insignificantly to this specific Hfq-mediated regulation of mRNA turnover. While our results reveal that Hfq binding contributes to mRNA turnover through recruitment of RNase E, more mechanistic details remain to be further elucidated. First, it is under debate whether Hfq-RNase E interaction is direct or mediated by RNAs (Bruce et al., 2018; Ikeda et al., 2011; Morita et al., 2005; Worrall et al., 2008). While our data show that Hfq can promote mRNA degradation without their matching regulatory sRNAs, it remains to be investigated whether other cellular RNAs may participate in bridging the interaction. Second, while our data demonstrate that deletion of the scaffold region of RNase E abolishes the Hfq-mediated regulation on mRNA turnover, the same scaffold region also includes the binding sites of other degradomes components; thus we cannot exclude the possibility that these protein components also participate in the regulation. Further experiments are needed to answer these mechanistic questions.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Chemical compound, druga-a'-dipyridyl crystallineFisher ScientificCatalog #: D95-5
Chemical compound, drugD-fructoseBioshopCatalog #: FRC180
Chemical compound, drugIPTG (for in vitro study)BioshopCatalog #: IPT001
Chemical compound, drugIPTG (for microscopy imaging)Goldbio367-93-1
Chemical compound, drugL(+)-arabinoseBioshopCatalog #: ARB222
Chemical compound, drugRifampicin (for in vitro study)BioshopCatalog #: RIF222
Chemical compound, drugRifampicin (for microscopy imaging)Sigma-AldrichR3501-1G
Commercial assay, kit2× QX200 ddPCR EvaGreen SupermixBio-RadRRID:SCR_019707Catalog #: 1864034
OtherEZ-rich define mediaTeknovaCatalog #: M2105
Bacterial growth media
OtherHiTRAP Heparin column HP (resin)GE Healthcare Life SciencesCatalog #: 17-0406-01
Protein purification column
Peptide, recombinant proteinTranscriptor reverse transcriptaseRocheCatalog #: 3531317001
Software, algorithmGraphPad PrismGraphPad PrismRRID:SCR_002798Version 8.2.1
Software, algorithmImage studio liteLi-CORRRID:SCR_013715Version 5.2
Software, algorithmMATLABMathWorksVersion R2019b
Software, algorithmNIS-ElementNikonVersion 4.51
Software, algorithmQuantaSoftBio-RadCatalog #: 1864011

Bacterial strains

Request a detailed protocol

Transfer of the Linker-mMaple3-Kan sequence at the 3’ end of chromosomal hfq gene was achieved by following the PCR-based method (Datsenko and Wanner, 2000) with a few modifications. First, a PCR (PCR1) was performed using plasmid pZEA93M as template to amplify the mMaple3 sequence (oligos EM4314-4293). Then, to add sequence homology of hfq gene, a second PCR (PCR2) was performed using PCR1 product as template (oligos EM4313-4293). The final PCR product (PCR3), containing a flippase recognition target (FRT)‐flanked kanamycin resistance cassette, was generated from pKD4 plasmid as template with PCR2 product and oligo EM1690 as primers carrying extensions homologous to the hfq gene. PCR3 was then purified and transformed to WT (EM1055), hfq Y25D (KK2562), hfq Q8A (KK2560), hfq K31A (AZZ41), or hfq R16A (KK2561) strains containing the pKD46 plasmid using electroporation, to obtain strains with hfq-Linker-mMaple3-Kan, hfq Y25D-Linker-mMaple3-Kan, hfq Q8A-Linker-mMaple3-Kan, hfq K31A-Linker-mMaple3-Kan, and hfq R16A-Linker-mMaple3-Kan, respectively. Hfq mutations F42A and R19D were obtained by performing PCRs on fusion strains KP1867 (Hfq-linker-mMaple3) with oligos EM4704-1690 (Hfq F42A) or EM4705-1690 (Hfq R19D). Fragments were then transformed to WT (EM1055) containing the pKD46 plasmid, following induction of the λ Red. P1 transduction was used to transfer the linked FP and the antibiotic resistance gene into a WT (EM1055), rne131 (EM1377), or rne∆14 (EM1376) strains. fnrS, micA, spf, chiX, and rppH knockouts were obtained through transformation of PCR products into EM1237 after induction of λ red and selecting for kanamycin or chloramphenicol resistance. P1 transduction was used to transfer the knockout mutations and the antibiotic resistance gene into appropriate strains. Selection was achieved with kanamycin, chloramphenicol, or tetracycline. When necessary, FRT-flanked antibiotic resistance cassettes were eliminated after transformation with pCP20, as described (Datsenko and Wanner, 2000). P1 transduction was also used to transfer hfq-Linker-mMaple3-Kan, hfq Y25D-Linker-mMaple3-Kan and hfq Q8A-Linker-mMaple3-Kan in the ΔsRNA or rne131-ΔsRNA strains. All constructs were verified by sequencing and are listed in Supplementary file 2. Oligonucleotides used for generating constructs are listed in Supplementary file 3.

Plasmids

E. coli MG1655 sRNA genes sgrS, ryhB, and chiX were inserted into the pET15b vector (kind gift from Perozo lab) to create plasmids pET15b-RyhB, pET15b-SgrS, and pET15b-ChiX using Gibson Assembly (in house) using oligos listed in Supplementary file 2. The chiX ΔAAN mutants were made using site directed mutagenesis. Primers EH159, EH160, and EH161 homologous to chiX while excluding the AAN domain were used to amplify the plasmid. The products were phosphorylated (NEB M0201S) and ligated (NEB M0202S) before transformation.

Cloning of pBAD-micA was performed by PCR amplification of micA (oligos EM2651-2652) on WT strain (EM1055). The PCR product was digested with SphI and cloned into a pNM12 vector digested with MscI and SphI. Two DNA fragments encoding the Fab fragment of the sAB-70 synthetic antibody were PCR amplified from pRH2.2-70-4D5-EA plasmid with OSA996/OSA997 and OSA998/OSA999 primer pairs. The third fragment (plasmid backbone) was amplified with OSA992/OSA993 primer pair from pZESA93M plasmid, and these three DNA fragments were joined to construct pZEFabM with RAIR assembly (Watson and García-Nafría, 2019). Construction of pZEGCN4MSD plasmid encoding the scFv-GCN4 binding protein was performed by amplifying the scFv-GCN4 reading frame (with oligos OSA1010/OSA1011) from pHR-scFv-GCN4-sfGFP-GB1-NLS-dWPRE plasmid. The plasmid backbone was PCR amplified with OSA1008/OSA1009 DNA oligos using pZESA93M as a template. The fragments were joined into pZEGCN4MSD with RAIR assembly. All constructs were verified by sequencing and are listed in Supplementary file 2. Oligos used for generating the constructs are listed in Supplementary file 3.

Growth conditions for imaging experiments

Request a detailed protocol

Overnight cultures of E. coli strains were diluted by 1:100 in MOPS EZ-rich defined medium (Teknova). 0.2% glucose was used as the carbon source for imaging Hfq-mMaple3 WT and mutants under NT and rifampicin treated conditions. 0.2% fructose was used as the carbon source with 100 µg/mL ampicillin for cases with sRNA overexpression, mMaple3 control, mMaple3 fused sAB-70, and mMaple3 fused scFv-GCN4. Cultures were grown at 37°C aerobically. Plasmid-encoded sRNAs were induced by 1 mM IPTG when the OD600 of the cell culture was ~0.1. Induced cells were grown for ~45 min before imaging. Plasmid-encoded mMaple3 protein (with 100–400 μM IPTG), mMaple3 fused sAB-70 (with 1 mM IPTG), and mMaple3 fused scFv-GCN4 (with 1 mM IPTG) were expressed and imaged in the same way. For the rifampicin treatment, rifampicin was added to a final concentration of 200 μg/mL when the OD600 of the cell culture was ~0.2, and the cells were incubated for 15 min before imaging.

Growth curve measurement

Request a detailed protocol

The bacterial strains were grown overnight in LB or MOPS EZ-rich medium containing 0.2% glucose. Cultures were diluted to 6 × 106 cells/mL in their respective medium and samples were prepared in triplicate by mixing 50 µL of cells and 50 µL of fresh medium to obtain 3 × 106 cells/mL. Assay was performed in Microtest plate, 96-well, flat base, polystyrene, sterile (Sarstedt) and growth was monitored using Epoch 2 Microplate Spectrophotometer reader (BioTek) with the following settings: OD = 600 nm, Temperature = 37°C, Reading = every 10 min for 22 hr, Continuous shaking.

RNA extraction and northern blot analysis

Request a detailed protocol

Total RNA was extracted following the hot-phenol protocol as described (Aiba et al., 1981). To test the function of the mMaple3-tagged Hfq and compare that with the WT Hfq, cells were grown in LB to the OD600 of 0.5 and either RyhB was induced by adding of 2.2′-dipyridyl in a WT hfq or in an hfq-mMaple3 background, or MicA was induced by addition of 0.1% arabinose (ara) in a ΔmicA WT hfq or in a ΔmicA hfq-mMaple3 background (pBAD-micA).

Determination of RNA half-life was performed in MOPS EZ-rich defined medium (Teknova) with 0.2% glucose by addition of 500 μg/mL rifampicin to the culture at the OD600 of 0.5 before total RNA extraction. Northern blots were performed as described previously (Desnoyers and Massé, 2012) with some modifications. Following total RNA extraction, 5–10 μg of total RNA was loaded on polyacrylamide gel (5% acrylamide 29:1, 8 M urea) and 20 μg was loaded on agarose gel (1%, 1× MOPS). Radiolabeled DNA and RNA probes used in this study are described in Supplementary file 3. The radiolabeled RNA probes used for northern blot analysis were transcribed with T7 RNA polymerase from a PCR product to generate the antisense transcript of the gene of interest (Desnoyers et al., 2009). Membranes were then exposed to phosphor storage screens and analyzed using a Typhoon Trio (GE Healthcare) instrument. Quantification was performed using the Image studio lite software (LI-COR).

The decay rate of mRNA degradation was calculated as previously described (Moffitt et al., 2016). Briefly, the intensity of northern blot at each time point upon adding rifampicin was normalized to the intensity at time zero and was fit by a piecewise function in the log space:

lnI(t)={lnI(0), t αlnI(0)k(tα), t>α

where I(t) is the normalized intensity at time t, I(0) is the normalized intensity at time zero, k is the rate of exponential decay, and α is the duration of the initial delay before the exponential decay begins. The reported half-lives (τ) are calculated by τ=log (2)/k .

Droplet digital PCR

Request a detailed protocol

Droplet digital PCR (ddPCR) was performed on total RNA extracted following the hot-phenol protocol (Aiba et al., 1981) from cells grown in MOPS EZ-rich defined medium containing 0.2% fructose (Teknova) with 50 µg/mL ampicillin. 1 mM IPTG was added at OD600 = 0.1 for 1 hr before total RNA extraction. Samples were treated with 8 U Turbo DNase (Ambion) for 1 hr. RNA integrity was assessed with an Agilent 2100 Bioanalyzer (Agilent Technologies). Reverse transcription was performed on 1.5 µg total RNA with Transcriptor reverse transcriptase, random hexamers, dNTPs (Roche Diagnostics), and 10 U of RNase OUT (Invitrogen) following the manufacturer’s protocol in a total volume of 10 µL.

ddPCR reactions were composed of 10 µL of 2× QX200 ddPCR EvaGreen Supermix (Bio-Rad), 10 ng (3 µL) cDNA, 100 nM final (2 µL) primer pair solutions, and 5 µL molecular grade sterile water (Wisent) for a 20 µL total reaction. Primers are listed in Supplementary file 3. Each reaction mix (20 µL) was converted to droplets with the QX200 droplet generator (Bio-Rad). Droplet-partitioned samples were then transferred to a 96-well plate, sealed, and cycled in a C1000 deep well Thermocycler (Bio-Rad) under the following cycling protocol: 95°C for 5 min (DNA polymerase activation), followed by 50 cycles of 95°C for 30 s (denaturation), 59°C for 1 min (annealing), and 72°C for 30 s (extension) followed by post-cycling steps of 4°C for 5 min and 90°C for 5 min (Signal stabilization) and an infinite 12°C hold. The cycled plate was then transferred and read using the QX200 reader (Bio-Rad) either the same or the following day post-cycling. The concentration reported is copies/μL of the final 1× ddPCR reaction (using QuantaSoft software from Bio-Rad) (Taylor et al., 2015).

Hfq purification

Request a detailed protocol

Hfq was purified following the previously described procedure (Prévost et al., 2007) with modifications. Briefly, strain EM1392 containing pET21b-hfq was grown at 37°C in LB medium supplemented with 50 μg/mL ampicillin and 30 μg/mL chloramphenicol until it reached an OD600 = 0.6. Hfq expression was induced by addition of 5 mM IPTG (Bioshop) for 3 hr. Cells were pelleted by centrifugation (15 min, 3825 g) and resuspended in 4 mL Buffer C (50 mM Tris-HCl pH 7.5, 1 mM EDTA, 50 mM NH4Cl, 5% glycerol) (Zhang et al., 2002) supplemented with 30 U Turbo DNase (Ambion). Cells were lysed by sonication for 4 min (amplitude 25%, cycles of 5 s sonication, 5 s on ice) and samples were cleared by centrifugation (45 min, 12,000 g). The supernatant was incubated at 80°C for 10 min, centrifuged again (20 min, 12,000 g), and cleared by filtration.

The protein extract was loaded onto a 1 mL HiTRAP Heparin column HP (GE Healthcare Life Sciences, 17-0406-01) equilibrated with Buffer A (50 mM Tris-HCl pH 8.0, 50 mM NaCl, 50 mM KCl, 1 mM EDTA, 5% glycerol). After washes, the protein was eluted with a linear NaCl gradient (0.05–1 M NaCl) in Buffer A. Fraction samples were loaded on SDS-PAGE and stained with Coomassie-Blue. Hfq-containing fractions were dialyzed against a dialysis buffer (50 mM Tris-Cl pH 7.5, 1 mM EDTA pH 8.0, 5% Glycerol, 0.25 M NH4Cl). Glycerol concentration was brought up to 10% and protein content was quantified by BCA assay (Thermo Scientific).

EMSA

Request a detailed protocol

DNA templates containing a T7 promoter were synthesized by PCR amplification on genomic DNA using oligonucleotides EM88-EM1978 (T7-ryhB), T7-ChiX(F)-T7-ChiX(R) (T7-chiX), or T7-ptsG(F)-T7-ptsG(R) (T7-ptsG). Briefly, templates were incubated for 4 hr at 37°C in RNA Transcription Buffer (80 mM HEPES-KOH pH 7.5, 24 mM MgCl2, 40 mM DTT, 2 mM spermidine) in the presence of 5 mM NTP, 40 U porcine RNase Inhibitor (in house), 1 μg pyrophosphatase (Roche), and 10 μg purified T7 RNA polymerase (in house). Samples were treated with 2 U Turbo DNase (Ambion) and purified on polyacrylamide gel (6% acrylamide:bisacrylamide 19:1, 8 M urea). When necessary, transcripts were dephosphorylated using 10 U Calf Intestinal Phosphatase (NEB) and were 5’ end-radiolabeled with [γ-32P]-ATP using 10 U T4 polynucleotide kinase (NEB). Radiolabeled transcripts were purified on polyacrylamide gel (6% acrylamide:bisacrylamide 19:1, 8 M urea).

EMSA were performed as previously described (Morita et al., 2012). To determine binding affinity of Hfq to RyhB, ChiX, and ptsG, radiolabeled RNA was heated for 1 min at 90°C and put on ice for 1 min. RNA was diluted to 20 nM in modified Binding Buffer 2 (10 nM Tris-HCl pH 8.0, 1 mM DTT, 1 mM MgCl2, 20 mM KCl, 10 mM Na2HPO4-NaH2PO4 pH 8.0, 12.5 μg/mL yeast tRNA) and mixed with specific concentrations of Hfq (0–200 nM). Samples were incubated for 15 min at 37°C and reactions were stopped by addition of 1 μL of non-denaturing loading buffer (1× TBE, 50% glycerol, 0.1% bromophenol blue, 0.1% xylene cyanol). For competition assays, 20 nM of radiolabeled ptsG was first incubated for 15 min at 37°C with 100 nM Hfq (as described above). Specific concentrations of RyhB or ChiX (0–100 nM) were added to the samples and incubation was carried out for 15 min at 37°C. Reactions were stopped by addition of 1 μL non-denaturing loading buffer. Samples were loaded on native polyacrylamide gels (5% acrylamide:bisacrylamide 29:1) in cold TBE 1X and migrated at 50 V, at 4°C. Gels were dried and exposed to phosphor storage screens and analyzed using a Typhoon Trio (GE Healthcare) instrument. When applicable, quantification was performed using the Image studio lite software (LI-COR) and data was fitted using nonlinear regression (GraphPad Prism).

Fluorescence in situ hybridization

Request a detailed protocol

Sample preparation for fixed cells was performed mostly according to the protocol previously reported (Park et al., 2018a; Fei et al., 2015). Briefly, ~10 mL of cell culture was collected and fixed with 4% formaldehyde in 1× PBS for 30 min at room temperature (RT). The fixed cells were then permeabilized with 70% ethanol for 1 hr at RT. Permeabilized cells can be stored in 70% ethanol at 4°C until the sample preparation. FISH probes were designed and dye-labeled as in the previous reports (Park et al., 2018a; Fei et al., 2015). Hybridization was performed in 20 μL of hybridization buffer (10% dextran sulfate [Sigma D8906] and 10% formamide in 2× SSC) containing specific sets of FISH probes at 30°C in the dark overnight. The concentration of FISH probes was 50 nM. After hybridization, cells were washed three times with 10% FISH wash buffer (10% formamide in 2× SSC) at 30°C.

Live-cell single-particle tracking and fixed cell SMLM imaging

Request a detailed protocol

Imaging was performed on a custom built microscopy setup as previously described (Park et al., 2018b). Briefly, an inverted optical microscope (Nikon Ti-E with 100× NA 1.49 CFI HP TIRF oil immersion objective) was fiber-coupled with a 647 nm laser (Cobolt 06–01), a 561 nm laser (Coherent Obis LS) and a 405 nm laser (Crystalaser). A common dichroic mirror (Chroma zt405/488/561/647/752r-UF3) was used for all lasers, but different emission filters were used for different fluorophores (Chroma ET700/75M for Alexa Fluor 647 and Chroma ET595/50M for mMaple3). For imaging Hoechst dye, a LED lamp (X-Cite 120LED) was coupled with a filter cube (Chroma 49000). The emission signal was captured by an EMCCD camera (Andor iXon Ultra 888) with slits (Cairn OptoSplit III), enabling fast frame rates by cropping the imaging region. During imaging acquisition, the Z-drift was prevented in real time by a built-in focus lock system (Nikon Perfect Focus).

For live-cell single particle tracking, 1 mL of cell culture was centrifuged at 1500 g for 5 min and 970 µL of the supernatant was removed. The remaining volume was mixed well and ~1.5 µL was covered by a thin piece of 1% agarose gel on an ethanol-cleaned-and-flamed coverslip sealed to a custom 3D printed chamber. The agarose gel contained the same concentration of any drug or inducer used in each condition. Exceptions include rifampicin, which was at 100 μg/mL in the gel, due to high imaging background caused by high concentration of rifampicin, and IPTG for mMaple3 alone control culture, which was eliminated in the gel, due to the high abundance mMaple3 already induced by IPTG in the culture. The power density of the 561 nm laser for single-particle tracking was ~2750 W/cm2, and the power density of the 405 nm laser was ~7 W/cm2 (except for mMaple3 alone control where ~4.5 W/cm2 was used due to high abundance of mMaple3). 1.5× Tube lens was used for the microscope body, and 2 × 2 binning mode was used for the camera. In this way, the effective pixel size became larger (173 nm instead of original 130 nm), receiving 77% more photons per pixel. Ten frames with 561 nm excitation were taken after each frame of 405 nm photo-conversion. About 13,000 frames were collected per movie at a rate of 174 frames per second. For fixed-cell control experiment for tracking parameter optimization, imaging was performed using the exact same imaging parameters as in live-cell measurements for a fair comparison. In cases imaging DNA together, Hoechst dye (Thermo 62249) was added to the ~30 µL of cell culture before imaging at ~20 µM final concentration and imaged by the LED lamp (12%) with 500 ms exposure time. Imaging acquisition was conducted by NIS-Element (Nikon) software, at RT.

Image reconstruction

Request a detailed protocol

The SMLM images are reconstructed as previously descried (Fei et al., 2015), by a custom code written in IDL (Interactive Data Language). Briefly, all the pixels with an intensity value above the threshold were identified in each frame. The threshold was set at three times of the standard deviation of the individual frame pixel intensity. Among those pixels, the ones having larger values than surrounding pixels in each 5 × 5 pixel region were identified as possible peak candidates, and 2D Gaussian function was fit to a 7 × 7 pixel region surrounding these candidates. Candidates with failed fitting were discarded, and precise peak positions were defined for the remaining ones. The horizontal drift, which often occurred during the imaging acquisition, was corrected by fast Fourier transformation analysis.

Tracking analysis

Request a detailed protocol

We used a MATLAB coded tracking algorithm to generate diffusion trajectories, which was modified by Sadoon and Yong (Sadoon and Wang, 2018) based on the previously developed code (Crocker and Grier, 1996). Per each time step of ~5.76 ms, 400 nm was empirically chosen to be the maximum one-step displacement to reduce artificial diffusion trajectories connected between different molecules, using a fixed cell sample as a control (Figure 1—figure supplement 3B). Trajectories longer than five time steps were used to calculate effective diffusion coefficient (D). MSD as a function of time lag (Δt) was fit with a linear function (MSD = D×Δt). D values are reported in related figures. For analysis using one-step displacement (osd), trajectories longer than three time steps were used.

Enrichment calculation

Request a detailed protocol

Enrichment at a certain region (nucleoid, membrane, or cytoplasm) of a cell is defined as follows:

(#oflocalizationsintheregion)/(total#oflocalizationsinthecell)(areaoftheregion)/(totalareaofthecell)

Here the area of a cell refers to the two-dimensional area of the cell from the differential interference contrast (DIC) image. The area of the nucleoid region was defined from the Hoechst image (nucleoid staining) and calculated by our custom MATLAB code (Reyer et al., 2018). Membrane region was determined as the boundary region from the DIC image, and the cytoplasm region was defined as the total cell region minus the nucleoid and the membrane regions.

Population analysis

Request a detailed protocol

The analysis of mRNA-associated and mRNA-free population of Hfq was performed by double population fitting of the cumulative probability density function (CDF) of one-step squared displacement (osd2) according to a previous report (Bettridge et al., 2021):

CDF(osd2)=1i=1nPieosd2/4Dit

where n is the number of diffusion states, Di is the diffusion coefficient of ith state, Pi is the fraction of ith state population, and i=1nPi=1. We found that a two-state model (n=2) fit better than one-state model (n=1), whereas a three-state model (n=3) did not further improve the fitting (Figure 2—figure supplement 2a). Therefore, we used a two-state model for fitting all Hfq tracking data, and the fast-diffusing state (D1, P1) was assigned as the mRNA-free fraction and the slow-diffusing state (D2, P2) assigned as the mRNA-associated. CDFs of rifampicin treatment cases in the WT rne background were fit first, the D1 range of rifampicin treatment cases under WT rne, that is, <D1, rif>±2×std(D1,rif), was then used to constrain the D1 values in the fitting for all other cases (Figure 2—figure supplement 2c and Supplementary file 1). All CDF fittings were conducted in OriginPro with Levenberg-Marquardt iteration algorithm. Osd speed was calculated as osd/Δt0 (Δt0 is the time interval between two consecutive frames, i.e., 5.76 ms). For comparison with D values from the linear fitting of MSD, one-step diffusion coefficient Di≡4D’i values were reported in Supplementary file 1 and Figure 2—figure supplement 2d.

Data availability

All the numeric data for each plot/graph and fitting results are provided in Supplementary file 1 or as source data. The MATLAB scripts for analysis are provided as source code.

References

    1. Vytvytska O
    2. Moll I
    3. Kaberdin VR
    4. von Gabain A
    5. Bläsi U
    (2000)
    Hfq (HF1) stimulates ompA mRNA decay by interfering with ribosome binding
    Genes & Development 14:1109–1118.

Decision letter

  1. Gisela Storz
    Senior and Reviewing Editor; National Institute of Child Health and Human Development, United States
  2. Jie Xiao
    Reviewer; Johns Hopkins University, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The Hfq RNA chaperone protein modulates the stabilities and activities of base pairing small RNAs as well as mRNAs in many bacteria. Park et al. use single molecule tracking in combination with genetic manipulations and supporting biochemical experiments to study Hfq and its interactions with small RNAs, mRNAs and the RNase E ribonuclease in vivo, gaining further insights into the mechanism of Hfq-mediated regulation.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Dynamic interactions between the RNA chaperone Hfq, small regulatory RNAs and mRNAs in live bacterial cells" for consideration by eLife. Your article has been reviewed by a Senior Editor, a Reviewing Editor, and three reviewer. The following individual involved in review of your submission has agreed to reveal their identity: Jie Xiao (Reviewer #2).

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife at this point.

This study examining the diffusion of Hfq in vivo is important given the discrepant reports regarding Hfq localization and interactions with RNase E. However, the discrepant reports also are the reason it is so critical that the conclusions drawn in the current paper be strongly supported. All three of the reviewers raised significant but valid concerns. If you think you can address their comments (which will require additional experiments and analyses and is expected to take longer than two months), you could resubmit a substantially revised version of this study for consideration as a new manuscript at eLife.

Reviewer #1:

This manuscript provides data on the diffusion of Hfq in vivo under a number of conditions, using a tagged Hfq and super-resolution imaging. The authors compare the movement of Hfq with and without treatment with rifampicin, as well as in a number of Hfq mutants, and find differences in diffusion that they use to conclude that the protein is generally associated with mRNAs, with the distal face contributing most of the binding. in vitro and previous functional tests of sRNA degradation and regulation are in agreement with many of the findings here- that mRNAs mostly bind to the distal face of Hfq, that there is likely in vivo competition between some (class II) sRNAs and mRNA binding, and that RNase E contributes to degradation of the mRNAs. Work published by Elf and coworkers (Persson et al., 2013) and cited here showed faster diffusion after rifampicin treatment, as found here. In the previous work the authors suggested three states, the slowest moving due to Hfq interacting with mRNA during transcription, discussed briefly here; understanding to what extent the data presented here may be affected by co-translational regulation would be very useful. Other conclusions reached in the current work (role of Hfq in degradation of mRNAs, independent of sRNAs and suggestion of the presence of monomeric forms of Hfq binding mRNA) are more novel but will require further analysis to be fully convincing.

1) The control for the movement of tagged Hfq is the free mMaple3. Given that Hfq is a hexamer, and presumably thus carries 6 tags as well as the six copies of Hfq, the MW of these two proteins will be very different. It would be useful to have a more comparable control (another hexamer that doesn't interact with RNA?). This becomes particularly important in interpreting the effect of rifampicin, central to much of the paper; is it clear that the effect on Hfq movement is not reflecting a general change in the cellular milieu?

2) Subsection “Binding of mRNAs to Hfq decreases its diffusivity primarily through the distal face of Hf”: Is the interpretation in Figure 2B that with rifampicin, there is no mRNA-bound Hfq?

3) Subsection “Most Hfq proteins are occupied by mRNAs in the cell during exponential growth”: I did not understand the argument being made here about Hfq existing as a monomer. This is supposed to be only in the absence of RNA binding? Absence of mRNA binding? There is reason to believe, from previous rifampicin chase experiments, that sRNAs will still be bound to Hfq after 15' of rifampicin treatment. Do these still bind to this subset of monomers? Where did the number for a 2-4 fold change in MW come from? Why is that 50-100 kDa (of monomer MW? Hfq with single tag? Or from somewhere else?)? Is the same conclusion made for the distal site mutations of Hfq, in which (Figure 2D), apparently only one third are associated with mRNA. In this case, again, sRNAs should be continuing to bind to Hfq (in the absence of rifampicin). If this suggestion of monomers for Hfq existing in the cell is of importance/significance, it needs to be better explained.

4) Figure 3, Interactions with and role of RNase E: There are multiple issues here, and it is difficult to sort them out as presented.

a) The rne131 and rne14 mutant D values seem to be similar to those shown in Figure 2 for rne+ cells, although the direct comparison was not provided. Does that suggest that RNase E interactions are not affecting diffusion (is not part of a complex)? It is a large enzyme, supposed to associate with the membrane. This should be commented on.

b) In Figure 3b, the mRNA associated fraction of 40-50% after rifampicin treatment is noted; is there data or just an assumption that this value is extremely low in the rne+ cell? I could not find that data anywhere.

c) Why isn't there a comparison for the mRNA associated fraction (and D values) in the combination of Y25D and the rne mutants? This might help to determine if the effects of the rne mutants reflect more RNA in general or the extent of mRNA bound to Hfq.

d) Figure 3C: Please relabel; left panel grey bars is not a ratio but just the half-life of cells WT for hfq and rne, correct? This was not immediately obvious. The interpretation of the (very modestly) increased half-life in the Y25D cells effects on the half-lives in Figure 3C is interpreted as sRNA-independent since the well-studied sRNA regulators have been deleted. What is the result for Q8A (unable to bind most known sRNAs) vs. Q8A rne? Is the half-life similar to fully WT, as expected if this were truly sRNA independent? Is it totally clear that there are no other (3' UTR) sRNA regulators? Something further is necessary to interpret this result as the authors would like to.

e) In the model the authors are suggesting, should all of sodB or most of sodB, for instance, be bound to Hfq for RNAse-E dependent turnover?

5) Does displacement of mRNAs from Hfq (as with ChiX in Figure 4) lead to a change in the half-life of the mRNAs as the data in Figure 3 might predict?

6) Figure 5: Is it surprising that the levels of ChiX are the same in the Q8A mutant and in the cases where AANs are deleted? Does this reflect the very high level overexpression? Are the levels of the RNAs measured under conditions that at all parallel the imaging that is providing the diffusion data, in terms of growth and induction conditions and time? I found it difficult to determine this.

7) Discussion: This statement (that the authors directly observed the transition of major interaction partners in response to cellular changes) is misleading. The authors only looked at exponential growth (what is the evidence that there are fewer sRNAs expressed then?), and do not really have any evidence that sRNAs are not also present on Hfq – they just cannot detect them by their method. The one changing condition tested is rather drastic – with rifampicin. Are there differences seen under more physiological conditions?

8) Is it possible to extrapolate the data to determine how many Hfq hexamers are detectable, in particular whether all classes of Hfq (bound to mRNA or not) are being seen? Does the slow-diffusing, presumably mRNA or mRNA/ribosome bound Hfqs represent all Hfqs?

9) Discussion: This statement (that the observations here will provide possible tools for use in eukaryotes) has very little relationship to the work presented and is not supported by anything in this paper. I really don't think it belongs here.

Reviewer #2:

In this work, Park et al., characterize the interactions between RNA chaperone Hfq with both mRNAs and sRNAs using single molecule tracking under genetic and drug manipulations, which are supported by biochemical assays. The major findings are that Hfq can regulate the stability of some mRNAs through its interaction with RNase E, that class I sRNAs are able to co-occupy mRNA-associated Hfq molecules, and that class II sRNAs displace mRNAs from Hfq, which requires an AAN motif. The provides strong in vivo evidence of several previously proposed mechanisms of Hfq and sRNA-mediated post-transcriptional regulation. However, some analyses could be redone to better support the author's conclusions. In addition, some critical information is missing. I recommend the work for publication on the condition these concerns are addressed.

- The authors need to provide sufficient statistics for the reviewer to determine validity and significance. For example, how many trajectories were collected from how many cells for each condition? How many cells were imaged using FISH? How many trajectories and/or single step displacements were used for the fitting? These numbers need to be provided in the figure legends or supplementary table.

- The authors plot the osd distribution and use this to calculate the mRNA-associated fraction of Hfq, but use the overall MSD vs time to calculate the D of Hfq. The distribution of osd appeared to be lognormal sine the authors used Gaussian of log(osd) to do the fitting. If it is indeed lognormal, it means that most likely there are more than one Gaussian distributed population. The authors should use osd distribution (not log) or CDF to calculate Ds and subpopulations, and compare with that obtained from MSD, which is an average measurement. Such analysis should be conducted for all conditions.

- The authors should exert caution to use the power law distribution of diffusion coefficient and molecular weight in bacterial cells to interpret the association of Hfq-bound mRNA with ribosomes. The D for Hfq under NT condition is too small (~ 0.5 um2/s) compared to what would be expected form the molecular weight of Hfq-mRNA complex (~ 550 kD), therefore the authors interprets that a fraction of the associated mRNAs are translated by ribosomes. Additionally, because rif-treated Hfq showed higher diffusion coefficient compared to what should be expected from its MW, the authors concluded that Hfq could dissociate into monomers when not bound to RNAs. It is known that proteins with different surface charges and mRNA molecules diffuse differently in bacterial cells and also respond to different stress responses. They do not necessarily scale with the expected MW especially at high MW. See Kumar et al., 2010, Lampo et al., 2017, and Schavemaker et al., 2017. To substantiate the authors' claims, translation inhibitor and Hfq hexamer mutants should be used. Similarly, subsection “sRNAs can displace Hfq from mRNAs in a face-dependent manner” "sRNA-bound Hfq-mMaple3 has similar diffusivity as free Hfq-mMaple3" based on similar MW is not substantiated.

Reviewer #3:

This manuscript provides single-particle tracking data that describes the movement of the RNA-binding protein Hfq in living E. coli cells. The authors investigate how the diffusion of Hfq is affected by the presence of RNA, the ability of Hfq to interact with RNA, and mutations in the major endoribonuclease RNase E. The authors also investigate how overexpression of different small RNAs affects Hfq diffusion.

Although the tracking data is of high quality, this manuscript suffers from several problems. Most worrying is that the authors interpret their data in an often careless manner. The manuscript contains many instances of over-interpretations, and some interpretations of data that are entirely incorrect.

This reviewer is particularly concerned about the emphasis on the role for Hfq in mRNA destabilization via recruitment of Hfq. It seems this is the only new message in this paper and thus the most critical and important one. As detailed below, most aspects of this model hinge upon assumptions that are contradicted by available biochemical data from the Luisi lab, or have not been addressed by clear-cut unambiguous experiments in this manuscript.

1) The authors base many of their interpretations on the assumption that Hfq interacts directly with RNase E. From reading the manuscript, it appears as if a direct Hfq-RNase E interaction is a well-established fact. This is not correct. In fact, the interaction between Hfq and RNase E is highly controversial. According to work by the Aiba lab, Hfq and RNase E interact directly without a requirement for RNA (e.g. Ikeda et al., 2010). In contrast, the Luisi lab has shown that removing RNA from purified Hfq abolishes the Hfq-RNase E interaction (Bruce et al., 2008), and further showed by carefully conducted biochemical experiments that re-constituted Hfq-RNase E complexes only form when Hfq is pre-bound to an sRNA. According to Luisi's data, it is the sRNA (not Hfq) that directly interacts with RNase E (Worrall et al., 2018). Unfortunately, the authors have based a large part of their experiments and conclusions on this controversial assumption. Therefore, the authors need to explicitly state that it is unclear whether Hfq can bind RNase E directly, or whether this interaction is mediated through RNA. The data in this manuscript need to be re-interpreted in light of the uncertainty of the Hfq-RNase E interaction.

2) On the same topic, subsection “Hfq is deficient in releasing mRNAs without interactions with RNase E”: "Hfq has been demonstrated to interact with the C-terminal scaffold region of RNase". As mentioned above, it is unclear whether this interaction is direct or occurs via RNA. In fact, one of the papers (Bruce et al.,) cited after this sentence claims that the interaction is dependent on RNA, which is the complete opposite of the authors' statement.

3) The authors claim that rifampicin "inhibits transcription and results in the loss of most cellular mRNAs". A more correct wording would be that rifampicin treatment inhibits transcription and leads to a global loss of RNA. For instance, many Hfq-binding sRNAs have half-lives shorter than 15 minutes (e.g. Vogel et al., 2003). While the measured increased diffusion of Hfq after rifampicin treatment most likely indicates a higher fraction of Hfq free from RNA, it does not inform on the classes of RNA that contribute to this binding. Either the authors need to explicitly show that mRNAs are preferentially lost over other RNA classes after 15 minutes of rifampicin treatment, or the term "mRNA-associated fraction" used throughout the manuscript should be changed to "RNA-associated fraction".

4) The authors claim that mutations in the distal phase of Hfq lead to "a large increase in the diffusivity under NT condition". This is a clear overstatement; according to Figure 2C, the increase compared to WT is less than two-fold.

5) Subsection “Most Hfq proteins are occupied by mRNAs in the cell during exponential growth”: "Considering the average length of bacterial mRNAs to be 1 kb (~330 kDa), and Mw of bacterial ribosome (~2.5 MDa), this reduction in D supports the interpretation that a significant fraction of WT Hfq proteins are associated with mRNAs in the NT case, and that a fraction of the associated mRNAs are translated by the ribosomes." Do the authors imply that the majority of Hfq is associated with mRNA alone? An alternative explanation would be that most Hfq hexamers are simultaneously bound to mRNA and sRNA. The data does not inform on the nature of the complexes Hfq is involved in, other than that they contain RNA (according to the measured changes in diffusivity during rifampicin treatment). This should be clearly stated in the text.

6) Subsection “Hfq is deficient in releasing mRNAs without interactions with RNase E”: "The rneΔ14 mutant has a smaller fraction of the C-terminal scaffold (residues 636-845) deleted, encompassing the Hfq binding region." The deleted part of RNase E also encompasses two RNA-binding domains, as well as binding sites for proteins RhlB and enolase. The differences in activity between WT and rneΔ14 could thus stem from impairment of many different interactions and do not specifically report on the loss of a potential interaction with Hfq.

7) Subsection “Hfq is deficient in releasing mRNAs without interactions with RNase E”: "In both RNase E mutant backgrounds, the diffusivity of Hfq-mMaple3 became less sensitive to transcription inhibition by rifampicin compared to the WT rne case (Figure 3A and Figure 2C)." The authors should show the data for strains with WT and mutant RNase E in the same graph for easier comparison. In addition, a statistical analysis should be provided to test whether the claimed differences are significant.

8) Subsection “Hfq is deficient in releasing mRNAs without interactions with RNase E”: "40-50% of Hfq-mMaple3 remained mRNA associated upon rifampicin treatment in the RNase E mutant backgrounds (Figure 3B)." I could not find the corresponding numbers for the strain with WT RNase E. Please provide these numbers in the figure and/ or in this part of the text for clarity.

9) Subsection “Hfq is deficient in releasing mRNAs without interactions with RNase E”: "These observations suggest that without the Hfq-RNase E interaction, more mRNAs remain bound to Hfq, indicating that Hfq may help deliver the associated mRNA to RNase E for degradation." This is a very far-reaching interpretation of the data presented in Figure 3. There is no evidence for Hfq-mediated delivery of RNase E in Figure 3. I strongly advise the authors to use a more careful interpretation of the data.

10) In Figure 3C, the authors show half-life measurements of several mRNAs that are regulated by sRNAs. For these experiments, strains with corresponding sRNA gene deletions were used. I assume that the rationale for this was to avoid putative differences due to impaired sRNA regulation. This is not a stringent strategy, as it is unknown whether other sRNAs target these mRNAs. Moreover, from the bar charts in Figure 3C, it appears that mRNA half-lives increase when Hfq carries a distal face mutation. However, in the Northern blots used to create the bar charts (corresponding supplementary figure), there seem to be very small (if any) differences in half-lives between WT and mutant Hfq. For transparency, and easier interpretation for the reader, the authors should (instead of bar charts) plot the log10-transformed relative band intensities versus time, and show all data points (not only error bars). They should also include the corresponding Northern blots in Figure 3 along with the quantifications.

11) The authors propose that Hfq, through an interaction with RNase E, promotes mRNA degradation in an sRNA-independent fashion. They also suggest that "mRNA-occupied Hfq proteins are in standby mode for sRNA binding if needed". If this were correct, one would expect that a high cellular concentration of a Class II sRNA, that is a strong competitor for Hfq's interaction with mRNAs, would result not only in displacement of Hfq from many mRNAs, but also thereby increase their stability. ChiX is a strong competitor for distal phase binding that should cause such an effect. The authors interpret their data to imply that ChiX overexpression results in displacement of Hfq from many mRNAs. However, they do not provide evidence that this displacement results in general mRNA stabilization, which should be the outcome according to their model. In fact, induction of ChiX to a high intracellular concentration from a plasmid resulted in downregulation of one specific mRNA (ybfM), rather than a stabilizing effect on many mRNAs (Rasmussen et al., 2009). It is very surprising that the authors did not cite this paper.

12) Several reports have shown that the 5' moiety of RNA substrates largely influences on both, RNase E cleavage efficiency and specificity of cleavage site selection (e.g. Mackie, 1998, Jiang and Belasco, 2004, Bandyra et al., 2012); RNAs carrying a 5' monophosphate are substantially better substrates than those with a 5' tri-phosphate. Regarding the Hfq-dependent mRNA degradation proposed by the authors, do these mRNAs need to be decapped for RNase E to degrade them? Or does the degradation go through the substantially less efficient internal cleavage route? The authors should discuss both their data and their models with respect to what is known about the cleavage activity/ specificity of RNase E. They should also cite the most seminal papers on this subject.

13) Regarding the EMSA shown in Figure 4. Subsection “sRNAs can displace Hfq from mRNAs in a face-dependent man”: "Results show that RyhB cannot displace the radiolabeled ptsG from Hfq, but rather generates an additional upper-shifted band compared to the band of ptsG-Hfq complex, supporting that RyhB and ptsG can co-occupy Hfq". This is a very creative interpretation of an inconclusive result. What can be deduced from the gel picture is that the band representing the Hfq-ptsG complex becomes weaker at the two highest RyhB concentrations. The reason for this could be either that RyhB displaces ptsG, or that a ternary complex is formed. The design of the experiment makes it impossible to judge whether the former is happening; since the majority of ptsG mRNA is not in complex with Hfq (even in the absence of RyhB), it is impossible to judge whether high concentrations of RyhB results in increased free ptsG mRNA. Regarding the latter possibility (which is put forward by the authors), the "upper-shifted band" is barely visible and do not by any means reach the intensity of the ptsG-Hfq band, which would be the case if the major effect of RyhB addition would be the formation of a ternary complex.

14) Subsection “sRNAs can displace Hfq from mRNAs in a face-dependent man”: "In addition, droplet digital PCR (ddPCR) performed in the same conditions as the diffusivity assays showed that RyhB level was comparable to ChiX (Figure 4D)." This is not correct. According to Figure 4d ChiX is almost ten times more abundant than RyhB. The same incorrect statement is repeated later in this subsection.

15) Subsection “sRNAs can displace Hfq from mRNAs in a face-dependent man”: "EMSA and ddPCR results suggest that both in vitro and in vivo, RyhB can effectively access mRNA-occupied Hfq through co-occupying Hfq from the proximal face." This is an unsubstantiated and probably incorrect interpretation of the results. The EMSA does not provide evidence for a ternary Hfq-ptsG-RyhB complex. In what way does the ddPCR result inform on binding of RyhB to mRNA-bound Hfq?

16) Conceptually, there is no apparent reason why Hfq would not interact with RNA undergoing transcription. A previous Hfq tracking study in E. coli indeed reported a three-state model in which the slowest state was interpreted as Hfq bound to RNAs during transcription (Persson et al., 2013). The authors should comment on this finding with regard to their own results. Do the data presented in the current manuscript fit with the previous model? If not, why not?

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "Dynamic interactions between the RNA chaperone Hfq, small regulatory RNAs and mRNAs in live bacterial cells" for consideration by eLife. Your article has been reviewed by Gisela Storz as a Reviewing and Senior Editor and three reviewers. The following individual involved in review of your submission has agreed to reveal their identity: Jie Xiao (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

As the editors have judged that your manuscript is of interest, but as described below that additional experiments are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Summary:

In this resubmission, Park et al., use single molecule tracking in combination with genetic manipulations and supporting biochemical experiments to probe Hfq-mediated mRNA regulation in live E. coli cells. They added some critical control experiments, including investigating whether RppH (i.e. 5'-end decapping) is required for RNase E-Hfq mediated mRNA degradation, and whether ChiX overexpression can modulate mRNA turnover independently of Hfq-RNase E interaction. The connection between their single-molecule tracking results with actual underlying biology is strong. The authors also revised some of their analyses. Additionally, the interpretations of the data are much more careful, and alternative interpretations are discussed. The uncertainty regarding a direct interaction between Hfq and RNase E is clearly stated. The authors also acknowledge that the slow state of diffusion may represent not only Hfq-mRNA complexes but also Hfq-mRNA-sRNA or other complexes. Overall, these changes substantially improved the strength of the arguments made, and the study will contribute significantly to the understanding of Hfq-mediated regulation of mRNA.

Essential revisions:

However, two critical points still need to be addressed:

1) One of the major findings put forward by the authors, which is highlighted both in the abstract and in the schematic Figure 8, is the suggestion that binding of Hfq to mRNAs can recruit RNase E for (sRNA-independent) mRNA degradation. The authors provide live cell Hfq tracking data in strains with mutations in the Hfq distal and proximal phases, combined with full-length or mutant RNase E (Figure 4). The tracking data show that the differences in Hfq diffusion, which are observed between WT and mutant RNase E upon rif treatment, are abolished when Hfq has the distal phase mutation. These data are sound and convincing but do not directly address whether the observed effects are indeed due to mRNA degradation.

To directly monitor effects on mRNA degradation, the authors provide Northern blot data monitoring the half-life of three selected mRNAs (Figure 5). According to the authors, and many previous studies, the distal face mutation primarily impairs Hfq-mRNA interactions, while the proximal face mutant primarily impairs Hfq-sRNA interactions, at least when considering Class I sRNAs. In Figure 5, the presented data show that, in the strain expressing WT RNase E, both the distal and the proximal face mutants lead to increased mRNA half-lives. It is well established that sRNA-dependent (Class I sRNAs) regulation requires Hfq to contact both the sRNA and the mRNA. If Hfq-dependent degradation of mRNAs where to be mediated strictly through sRNAs, the increase half-life should be the same for the distal and proximal mutants (binding to both RNAs are required). However, according to the authors, the data presented in Figure 5 shows a greater increase in mRNA half-lives with the distal face mutant compared to the proximal face mutant, and is interpreted as representing Hfq-dependent and sRNA-independent degradation by RNase E. From this follows that the contribution of a sRNA-independent effect on mRNA half-lives can be deduced from the difference between the values obtained in the distal mutant and in the proximal mutant, while the difference between WT Hfq and the distal face mutant is the sum of the sRNA-dependent and sRNA-independent effects on mRNA degradation. In other words, if the authors' hypothesis is correct, there should be a significant difference in mRNA half-life between the distal and proximal face mutants, and this difference should be abolished if Hfq cannot interact with RNase E.

Unfortunately, the data provided in Figure 5 do not provide unequivocal evidence for the following reasons: (i) too few replicates: some of the mean values were calculated from only two data points, therefore the standard deviations are not meaningful (ii) there is no information on how many replicates where used for calculating each specific mean value, making it impossible to judge the how reliable specific mean values and standard deviations are, (iii) there is no statistical analysis provided to ensure that the differences are significant (which on the other hand is not possible with only two data points). For these reasons, it is impossible to judge whether the differences between Y25D and Q8A in the WT and rne131 mutant are significant. To test whether the hypothesis is correct, the authors need to provide at least three (four would be advisable) replicates for each mean value, and use statistical tests to assess whether the proposed differences are indeed significant. To increase the possibility for the reader to judge the data, each bar showing mean values in Figure 5d should to be overlaid with the value of each data point used for calculating the mean.

2) While the analysis is improved, there appeared to be a misunderstanding on how to use CDF to fit displacement to extract the diffusion coefficients and population percentages. The authors used the CDF fitting of the apparent speed of one step displacement (osd) to extract the average speed and population percentage of Hfq. The authors cited the Yang, 2019 paper as the source. Do note that in Yang, 2019, the authors were measuring the directional moving speed but not random diffusion. For non-processively diffusing molecules, the diffusion coefficient D should be used. While the speed CDF from osd can also be used to extract different populations based on the difference in speed, it is different from classifying molecules based on their apparent diffusion coefficients (note that only osd2, but not osd, is proportional to D). Two molecules diffusing at different speeds (with random orientations) will have different Ds, but the quantitative difference between the two Ds is not the same as that between the two speeds (again, osd2 , but not osd, is proportional to D). Therefore, the classification of fast and slow diffusing molecules based on diffusion coefficient could be different from that based on speed. Furthermore, the mean squared displacement (MSD) measures how far the molecule diffuses away over different time lags. MSD is also an averaged measurement of molecules of different Ds, and not particularly accurate when there are at least two populations of different Ds, which is the case the authors are trying to establish. See https://link.springer.com/protocol/10.1007/978-1-59745-513-8_14 for a review, and Bettridge, et al., https://doi.org/10.1111/mmi.14572, in the supplemental notes for a practical guide. The fractionation of subpopulations based on diffusion coefficient would be important for the authors to make the argument of whether mRNA binding would lower the D or vice. versa. If the difference in speed is fairly large, I suspect that the major conclusions should still hold if the authors switch to the analysis of CDF of osd2, but the correct analysis should be provided.

https://doi.org/10.7554/eLife.64207.sa1

Author response

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

We would like to thank all three reviewers for their careful consideration and constructive suggestions, particularly under the difficult situation of COVID-19 outbreak. We have now significantly revised the manuscript based on the reviewers’ suggestions, including new data and analysis. As some of the major concerns are shared by more than one reviewer, we would like to summarize the corresponding changes first before providing point-to-point responses on specific critiques.

Essential revisions:

(1) Data analysis for determining mRNA-associated and mRNA-free fraction of Hfq

Several critiques raised by reviewers 1 and 3 regarding the mRNA-associated and mRNAfree fraction of Hfq under various conditions are due to our fault of not presenting the analysis in a clear enough fashion. In our previous analysis, we assumed under rifampicin (Rif) treated cases, Hfq proteins were all free of mRNAs, and fit the one-step displacement (osd) histogram for Rif treated cases with single Gaussian. Then with the mean osd value for the Rif treated case, we fit the no treatment (NT) case with two Gaussians to estimate the mean osd value for the mRNA-bound Hfq, and the associated fraction mRNA-free and mRNA-associated Hfq. We thank Dr. Jie Xiao (reviewer 2) for her suggestion on alternative analysis. We have accordingly applied the fitting of cumulative probability density function (CDF) of the one-step directional (osd) speed to recalculate the mRNA-associated and mRNA-free fraction of Hfq, as described in (Yang et al., https://doi.org/10.1101/850073). differences in the analysis are summarized as below:

(1) Data are all described well with two-population models, better than single-population fitting. Comparison is presented in Figure 2—figure supplement 2.

(2) When applying the new analysis method, we made one additional change in building trajectories from the raw imaging: changing the maximum one-step distance from the original cutoff of 250 nm to 400nm. We made this change, because we realized our initial distance cutoff of 250 nm was too stringent, which caused significant loss of fast-diffusing population, particularly for Rif treatment cases. We therefore increased the cutoff to 400 nm. As presented in Figure 1—figure supplement 3, for NT and Rif cases, particularly Rif case, changing the cutoff distance from 250 nm to 400 nm included significantly more fast-diffusing population (the right four plots in panel b). As a control, we compared the two distance cutoffs for fixed cells, and the difference was minimal (the leftmost two plots).

(3) The analysis and fitting are presented in Figure 2—figure supplement 2 and Supplement file 1. Briefly, CDFs of all Rif cases (for WT Hfq and all Hfq mutants) were fitted with double lognormal populations, and we obtained the relative populations (probability values) of the fast-diffusing state and the slow-diffusing states, which we assigned as mRNA-free and mRNA-bound Hfq populations, respectively. Rif cases for all mutants gave consistent fitting results. To better estimate the percentage of the two populations for the NT cases, we constrained the fitting parameters for the fast population during fitting. Importantly, the population average of the diffusion coefficient from the new analysis is consistent with the ensemble diffusion coefficient from linear fitting of MSD vs Δt curve in all cases (panel d in Figure 2—figure supplement 2).

All conclusions remain the same with the new analysis, with mostly minor changes in the actual numbers (diffusion coefficient and population percentage). One difference worth noticing is the mRNA-associated fraction (slow population) in Rif case, which on average is around 40%. mRNA-associated fraction in the Rif case is likely to be overestimated, or mRNA-free fraction in the Rif case is likely to be underestimated, because even with 400 nm distance cutoff, we may still likely miss some fast-diffusing population (shown in the figure above). However, we decided not to further increase the distance cutoff in the analysis, in order to prevent creating artificial trajectories. Nevertheless, this does not affect our conclusions, since all the trends of Hfq diffusion changes remained the same with the new fitting method.

(4) Related to the population analysis, reviewers 1 and 3 also brought up the analysis in a previous study on Hfq, in which an HMM-based method (named vbSTP) was used to estimate different diffusion states. Particularly, in addition to mRNA-free Hfq, and mRNAbound Hfq, vbSTP also revealed a third state representing Hfq associated with nascent mRNA during transcription. Indeed, when we applied vbSPT to our tracking data, we actually got very consistent results (Author response image 1).

Author response image 1

In this figure, vbSPT plots for WT Hfq, NT case (left) and Rif case (right) are presented. Upper plots are from vbSPT fitting (software package is from the Elf’s lab) to our datasets, and the lower plots are from the corresponding paper (Persson et al., 2013, Figure 4). Please note that the software output as well as the manuscript by Persson et al., use the factor of 4 for MSD fitting, i.e., MSD=4*D*Δt. Thus, to compare with the diffusion coefficients in the above plots, our diffusion coefficients presented in the revised manuscript should be divided by 4. For example, in “WT Hfq, Rif case, we report 9±0.8 µm2/s and 0.9±0.4 for the faster and slower diffusion coefficients from CDF fitting (Figure 2—source data 1), which translate to 2.3 µm2/s and 0.23 µm2/s with the consideration of the factor of 4. These numbers are very comparable to our vbSPT outcomes, 2.1 µm2/s and 0.304 µm2/s Author response 1, top right). In the same way, we compare all these numbers in Author response image 2.

Author response image 2

(Since we only fit two states for the CDF fitting method, there are only D1 and D3 values for NT case in our fitting). Overall, the manuscript by Persson et al., generally gave slightly faster diffusion coefficients, but these numbers are comparable for each case. Please note that these two studies used two different fluorescent protein taggings (mMaple3 for ours, and Dendra2 for Persson et al.,), possibly contributing to the differences of vbSPT outputs.While we got roughly consistent results using vbSPT, we had concerns on using this method to estimate the fraction of different population due to two reasons:

First, we found vbSPT may overfit the data sometimes, partially because the algorithm does not consider the localization uncertainty in imaging. For example, when we analyzed more than 10000 trajectories (well above the recommended minimal number of trajectories, 3000), from the NT case, vbSPT generates 4 states. The diffusion coefficient of the slowest state corresponds to a one-step displacement of 20nm, actually below our localization uncertainty (Author response image 3) (0.075 μm2/s of D value corresponds to (0.075x0.00576)/0.00576=3.6µm/s of diffusion speed, so between two imaging frames, this corresponds to 3.6 x 0.00576 = 0.02 µm = 20 nm of displacement).

Author response image 3

Similarly, Hfq-mMaple in fixed cells was fit by vbSPT into four states. Again, the one-step displacement in the slowest state is beyond our imaging accuracy.

Author response image 4

Second, vbSPT does not allow users to constrain fitting parameters. In both their and our data, vbSPT tends to fit a smaller diffusion coefficient for the mRNA-free population in the NT case compared to Rif case (D3 value in NT vs D2 value in Rif), which would lead to an overestimation of the mRNA-free population in the NT cases. Based on these reasons, we prefer to use the analysis method suggested by reviewer 2.

(2) Reviewer 1 and 3 raised several concerns on the data and interpretations that lead to the conclusion that Hfq can regulate mRNA turnover in a sRNA-independent way. We thank both reviewers for their suggestion of several new experiments. We are summarizing the new experiments below:

(1) We added new imaging experiments on Y25D and Q8A Hfq mutants in the rne mutant backgrounds. The data are presented in the new Figure 4. The description of the results is as below:

“Hfq Y25D-mMaple3, as it is deficient in mRNA binding, showed minimal sensitivity to rifampicin treatment in the rne mutant backgrounds, the same as in the WT rne background (Figure 4A and B). These observations suggest that without the Hfq-RNase E interaction, more mRNAs remained bound to Hfq, and hint that Hfq-RNase E interaction may help recycle Hfq from the mRNA-associated form through degradation of mRNAs. To investigate whether the reduction of mRNA releasing from Hfq in the rne mutant background is primarily contributed by sRNAs, we imaged Hfq Q8A-mMaple3 in the rne mutant backgrounds. Q8A mutation of Hfq broadly disrupts the binding and stabilization of sRNAs27. If sRNA is the dominating factor in driving the mRNA releasing from Hfq, we would expect a significant increase in the mRNA-associated fraction of the Q8A mutant in the rne mutant backgrounds compared to the WT Hfq with rifampicin treatment. However, we did not observe any significant difference in the mRNAassociated fraction between Q8A mutant and the WT Hfq (Figure 4B), suggesting that Hfq-RNase E interaction can also drive the release of mRNA from Hfq independently of sRNAs.”

2) We also added new mRNA half-life measurements in the background of Hfq Q8A in addition to knockout of the relevant sRNAs. The data are presented in new Figure 5. The description of the results in the revised manuscript is as blow:

“To further exclude the contributions by potentially unknown sRNAs, we compared the lifetime of sodB, ompX and sdhC in the hfq-mMaple3 Q8A background in addition to knocking out corresponding sRNAs (Figure 5). The half-lives of these mRNAs increased by 19%-46% in the hfq Q8A background compared to WT hfq background, smaller than the increase observed in the hfq Y25D background (Figure 5D). The increase of mRNA half-life in the hfq Q8A background can either be due to contributions by unknown sRNA regulators, or due to other possible regulatory pathways by Hfq through binding at the proximal face. One of such regulatory pathways may be Hfq-mediated polyadenylation, which involves binding of Hfq at the Rho-independent termination site and promotes mRNA degradation9,10. Despite these two possibilities, the increase in the mRNA half-life due to Y25D mutation cannot be fully explained by sRNA-mediated regulation. These results collectively support that besides the sRNA mediated pathway, Hfq can facilitate the turnover of certain mRNAs by binding to the mRNAs through the distal face and bridging them to RNase E for degradation.”

Based on the new results, we think our original interpretation still holds, that Hfq can regulate mRNA turnover in a small-RNA independent manner in addition to sRNA-mediated pathway. However, we make it more specific that this regulation is mediated through binding of mRNA at the distal face of Hfq.

Reviewer #1:

This manuscript provides data on the diffusion of Hfq in vivo under a number of conditions, using a tagged Hfq and super-resolution imaging. The authors compare the movement of Hfq with and without treatment with rifampicin, as well as in a number of Hfq mutants, and find differences in diffusion that they use to conclude that the protein is generally associated with mRNAs, with the distal face contributing most of the binding. in vitro and previous functional tests of sRNA degradation and regulation are in agreement with many of the findings here- that mRNAs mostly bind to the distal face of Hfq, that there is likely in vivo competition between some (class II) sRNAs and mRNA binding, and that RNase E contributes to degradation of the mRNAs. Work published by Elf and coworkers (Persson et al., 2013) and cited here showed faster diffusion after rifampicin treatment, as found here. In the previous work the authors suggested three states, the slowest moving due to Hfq interacting with mRNA during transcription, discussed briefly here; understanding to what extent the data presented here may be affected by co-translational regulation would be very useful. Other conclusions reached in the current work (role of Hfq in degradation of mRNAs, independent of sRNAs and suggestion of the presence of monomeric forms of Hfq binding mRNA) are more novel but will require further analysis to be fully convincing.

We thank Reviewer 1 for her/his overall positive comments and useful feedback. As this reviewer mentioned above, the finding that Hfq might contribute to the regulation of mRNA turnover in the absence of the corresponding sRNA regulators is novel. In addition, our results provide strong in vivo support for some of the previous in vitro biochemical and functional tests and conclusions on Hfq-mRNA and Hfq-sRNA interactions. Finally, we also found that while class II sRNAs can directly compete with mRNA for binding to Hfq, class I sRNAs do not necessarily need to displace the mRNA, but can co-occupy with mRNA on Hfq through different binding sites. These results provide more concrete mechanisms to the previously suggested RNA exchange model.

1) The control for the movement of tagged Hfq is the free mMaple3. Given that Hfq is a hexamer, and presumably thus carries 6 tags as well as the six copies of Hfq, the MW of these two proteins will be very different. It would be useful to have a more comparable control (another hexamer that doesn't interact with RNA?). This becomes particularly important in interpreting the effect of rifampicin, central to much of the paper; is it clear that the effect on Hfq movement is not reflecting a general change in the cellular milieu?

We thank the reviewer for the good suggestion. To my knowledge, I have not seen similar control being used in any studies using single-particle tracking. We reasoned that to make it a solid control, it is better to find a protein that does not interact with DNA (since rifampicin also causes changes in DNA morphology) or RNA, nor naturally exists in E. coli. However, after discussing with colleagues with expertise in protein purification, we found it difficult to express a recombinant protein that meets these criteria, and is either hexamer or with a size of 200 kDa. Nevertheless, we generated two more controls with a total size of 80 kDa each, one being the scFv protein used in the SunTag technique, the other one being a synthetic antibody sAB-70. Both of them behave the same as the mMaple3 only control, i.e. do not show a difference in the diffusivity upon rifampicin treatment. The new data are included in Figure 2—figure supplement 1. We hope reviewer 1 can find these new controls satisfying. In fact, it is unlikely that the observed difference in Hfq-mMaple3 is due to change in the cellular milieu, as the difference we observed upon rifampicin treatment varies between different Hfq mutants. For the Y25D mutant, it is also most completely insensitive to rifampicin treatment, arguing against the possibility that the difference we observed in WT Hfq and other Hfq mutants are due to a general change in the cellular milieu.

2) Subsection “Binding of mRNAs to Hfq decreases its diffusivity primarily through the distal face of Hf”: Is the interpretation in Figure 2B that with rifampicin, there is no mRNA-bound Hfq?

We apologize for not being clear in the previous manuscript. In the previous analysis, we assumed that with high concentration of rifampicin treatment for 15 minutes, the cells will be depleted with mRNAs and therefore mRNA-bound Hfq is close to 0%. This assumption helped us on the fitting of the mRNA-bound fractions in the no treatment case. We have now reanalyzed all the data as suggested by reviewer 2, and the mRNA-bound fractions are reported now for all conditions in the new Figure 2 and Figure 4. The details on the new data analysis method are described in the new Figure 1—figure supplement 3 and Figure 2—figure supplement 2, and summarized in the Essential revision part 1 above.

3) Subsection “Most Hfq proteins are occupied by mRNAs in the cell during exponential growth”: I did not understand the argument being made here about Hfq existing as a monomer. This is supposed to be only in the absence of RNA binding? Absence of mRNA binding? There is reason to believe, from previous rifampicin chase experiments, that sRNAs will still be bound to Hfq after 15' of rifampicin treatment. Do these still bind to this subset of monomers? Where did the number for a 2-4 fold change in MW come from? Why is that 50-100 kDa (of monomer MW? Hfq with single tag? Or from somewhere else?)? Is the same conclusion made for the distal site mutations of Hfq, in which (Figure 2D), apparently only one third are associated with mRNA. In this case, again, sRNAs should be continuing to bind to Hfq (in the absence of rifampicin). If this suggestion of monomers for Hfq existing in the cell is of importance/significance, it needs to be better explained.

We apologize for the confusion. Our previous interpretation was based on the power-law relationship between molecular weight (Mw) and the diffusion coefficient (D) (D = aMwx, where x = -0.33) that was experimentally observed in bacterial cells (reference 42). Comparing the diffusion coefficients of Hfq-mMaple3 and free mMaple, and the molecular weight of free mMaple3, we estimated that the molecular weight of Hfq-mMaple3 is (50-100kDa) based on the power-law relationship, which is smaller than Hfq-mMaple3 hexamer. Therefore, we concluded that part of the Hfq-mMaple3 may exist in monomer form. However, as reviewer 2 suggested, inference of Mw just based on D may not be robust, since diffusion coefficient is also dependent on other factors, such as surface charge of the protein. In the revised manuscript, we have removed the part of the Discussion, as it does not contribute to the major conclusions of our manuscript.

4) Figure 3, Interactions with and role of RNase E: There are multiple issues here, and it is difficult to sort them out as presented.

a) The rne131 and rne14 mutant D values seem to be similar to those shown in Figure 2 for rne+ cells, although the direct comparison was not provided. Does that suggest that RNase E interactions are not affecting diffusion (is not part of a complex)? It is a large enzyme, supposed to associate with the membrane. This should be commented on.

We thank the reviewer for bringing up this point. We improved the analysis on the one-step diffusion speed as a function of cell coordinates (Figure 1B-D), in which we further separated the cell into nucleoid, cytoplasm and membrane region. Indeed, Hfq diffusion speed in the membrane region is slightly lower compared to nucleoid and cytoplasm, which is likely due to the interactions with RNase E localized in the inner membrane. However, since the population of Hfq diffusing into the inner membrane region is not significant compared to the total population of Hfq, the minor slowing down of diffusion speed is overweighed by Hfq population in nucleoid and cytoplasm. We cannot robustly resolve the populations of Hfq interacting with RNase E in the presence and absence of mRNAs. For visualization of the data, we provided a direct comparison of D values under WT and mutant RNase E in the new Figure 4.

b) In Figure 3B, the mRNA associated fraction of 40-50% after rifampicin treatment is noted; is there data or just an assumption that this value is extremely low in the rne+ cell? I could not find that data anywhere.

This is the same issue as the point #2 brought up by the reviewer -- we assumed that the mRNA associated fraction is close to zero after rifampicin treatment in the WT rne background. With this assumption, fitting of the rifampicin treated case in the rne131 background generated 4050% of mRNA-associated fraction. We have now reanalyzed all the data based on reviewer 2’s suggestion. The mRNA-bound fractions are reported now for all conditions in the new Figure 2 and Figure 4. The details on the new data analysis method are described in the new Figure 1—figure supplement 3 and Figure 2—figure supplement 2, and summarized in the Essential revisions part 1 above.

c) Why isn't there a comparison for the mRNA associated fraction (and D values) in the combination of Y25D and the rne mutants? This might help to determine if the effects of the rne mutants reflect more RNA in general or the extent of mRNA bound to Hfq.

We thank the reviewer for the good suggestion. We have now added data in the background of Y25D/rne131 background (new Figure 4). Without rifampicin treatment, Hfq Y25D-mMaple3 in the rne mutant background demonstrated low mRNA-associated fraction, the same as the no treatment case of the Y25D in theWT rne background. With rifampicin treatment, Y25D in the rne background also showed the same efficiency in mRNA releasing as in the WT rne background, suggesting that this Y25D mutation does not contribute to the regulation of mRNA turnover by RNase E, presumably due to its deficiency in mRNA binding.

d) Figure 3C: Please relabel; left panel grey bars is not a ratio but just the half-life of cells WT for hfq and rne, correct? This was not immediately obvious. The interpretation of the (very modestly) increased half-life in the Y25D cells effects on the half-lives in Figure 3C is interpreted as sRNA-independent since the well-studied sRNA regulators have been deleted. What is the result for Q8A (unable to bind most known sRNAs) vs. Q8A rne? Is the half-life similar to fully WT, as expected if this were truly sRNA independent? Is it totally clear that there are no other (3' UTR) sRNA regulators? Something further is necessary to interpret this result as the authors would like to.

We have added the new half-life measurement using Q8A mutant in addition to small RNA knockout as the reviewer suggested. The new data are presented in new Figure 5, and the description is as below:

“To further exclude the contributions by potentially unknown sRNAs, we compared the lifetime of sodB, ompX and sdhC in the hfq-mMaple3 Q8A background in addition to knocking out corresponding sRNAs (Figure 5). The half-lives of these mRNAs increased by 19%-46% in the hfq Q8A background compared to WT hfq background, smaller than the increase observed in the hfq Y25D background (Figure 5D). The increase of mRNA half-life in the hfq Q8A background can either be due to contributions by unknown sRNA regulators, or due to other possible regulatory pathways by Hfq through binding at the proximal face. One of such regulatory pathways may be Hfq-mediated polyadenylation, which involves binding of Hfq at the Rho-independent termination site and promotes mRNA degradation9,10. Despite these two possibilities, the increase in the mRNA half-life due to Y25D mutation cannot be fully explained by sRNA-mediated regulation. These results collectively support that besides the sRNAmediated pathway, Hfq can facilitate the turnover of certain mRNAs by binding to the mRNAs through the distal face and bridging them to RNase E for degradation.”

e) In the model the authors are suggesting, should all of sodB or most of sodB, for instance, be bound to Hfq for RNAse-E dependent turnover?

Our measurement cannot determine the percentage of individual mRNA species that are associated with Hfq, i.e., we do not know the percentage of sodB mRNAs that bind to Hfq. In fact, since many mRNAs compete for Hfq, it is actually unlikely that sodB mRNAs are close to being completely bound by Hfq. In this sense, our model only suggests that Hfq-RNase E-dependent degradation “contributes” to the endogenous turnover of some mRNAs, but is not the sole pathway for these mRNAs. Other RNase-E mediated degradation pathways clearly also contribute. We have emphasized this point in the second before the last paragraph of the Discussion.

5) Does displacement of mRNAs from Hfq (as with ChiX in Figure 4) lead to a change in the half-life of the mRNAs as the data in Figure 3 might predict?

We thank the reviewer for the good suggestion. We added new lifetime measurements in the presence of WT ChiX and ChiX with two AAN motif deleted, and found that the presence of WT ChiX can indeed increase the sdhC lifetime, similar to Y25D Hfq mutant, whereas the mutant ChiX that has reduced Hfq binding ability is not able to increase the lifetime as significantly as the WT ChiX. The new data is presented as new Figure 6.

“As our model suggests that binding of Hfq to the mRNA through the distal face can regulate the mRNA turnover, we reasoned that sRNAs that can effectively compete for Hfq binding against mRNAs may decoy Hfq from this regulatory function. To test this, we again used sdhC as an example, and measured its half-life in the presence of ChiX, which is a strong competitor for Hfq binding (Figure 3). In the presence of vector control, sdhC exhibited comparable half-life compared to the case without any plasmid (Figure 5D, Figure 6C and D). The presence of WT ChiX increased the half-life by ~70%, whereas the mutant ChiX without two AAN motif deleted only increased the half-life by ~42%, consistent with its reduced binding ability to Hfq (Figures 3a, 3b, 6c and 6d). These results further support our model of Hfq-mediated regulation of mRNA turnover, and demonstrate that the presence of strong Hfq binding sRNAs can modulate the strength of Hfq’s regulation.”

6) Figure 5: Is it surprising that the levels of ChiX are the same in the Q8A mutant and in the cases where AANs are deleted? Does this reflect the very high level overexpression? Are the levels of the RNAs measured under conditions that at all parallel the imaging that is providing the diffusion data, in terms of growth and induction conditions and time? I found it difficult to determine this.

The cell growth media and inducer concentration are the same for ddPCR and live-cell imaging as described in the Materials and methods. Minor difference is that cells were imaged after 45 min induction, whereas the total RNA was extracted after 1 hour induction. The reason for this is that imaging the live-cell samples took 10-15 minutes. Therefore, the actual induction times were comparable between these two experiments. Both ddPCR and FISH experiments confirmed that the relative levels of WT and mutant ChiX were very comparable (Figure 3D and Figure 3—figure supplement 2), perhaps due to plasmid-expressing levels being higher than endogenous abundance.

7) Discussion: This statement (that the authors directly observed the transition of major interaction partners in response to cellular changes) is misleading. The authors only looked at exponential growth (what is the evidence that there are fewer sRNAs expressed then?), and do not really have any evidence that sRNAs are not also present on Hfq – they just cannot detect them by their method. The one changing condition tested is rather drastic – with rifampicin. Are there differences seen under more physiological conditions?

“Cellular changes” also include the cases where we expressed specific sRNAs, and we observed that the sRNAs can compete for Hfq binding against mRNAs. We have removed that sentence from Discussion.

8) Is it possible to extrapolate the data to determine how many Hfq hexamers are detectable, in particular whether all classes of Hfq (bound to mRNA or not) are being seen? Does the slow-diffusing, presumably mRNA or mRNA/ribosome bound Hfqs represent all Hfqs?

Unfortunately, we are not able to determine the absolute copy number of Hfq, as the photoconversion of the fluorescent protein is not 100% during the time window of our imaging experiments. Since the changes we introduced, such as rifampicin treatment and sRNA expression, more specifically affect the association of Hfq with the mRNAs, we interpret that the shift from the slow population to fast population is due to removal of the mRNA from Hfq and the slow-diffusing population represent the mRNA-bound Hfq. It is possible that some fraction of the slow-diffusing population also includes other Hfq-involved complexes, such as DNA associated Hfq. But one needs to introduce mutation specific affecting Hfq-DNA interaction to extrapolate that population, which is beyond the scope of the current manuscript.

9) Discussion: This statement (that the observations here will provide possible tools for use in eukaryotes) has very little relationship to the work presented and is not supported by anything in this paper. I really don't think it belongs here.

We have deleted this part.

Reviewer #2:

In this work, Park et al., characterize the interactions between RNA chaperone Hfq with both mRNAs and sRNAs using single molecule tracking under genetic and drug manipulations, which are supported by biochemical assays. The major findings are that Hfq can regulate the stability of some mRNAs through its interaction with RNase E, that class I sRNAs are able to co-occupy mRNA-associated Hfq molecules, and that class II sRNAs displace mRNAs from Hfq, which requires an AAN motif. The provides strong in vivo evidence of several previously proposed mechanisms of Hfq and sRNA-mediated post-transcriptional regulation. However, some analyses could be redone to better support the author's conclusions. In addition, some critical information is missing. I recommend the work for publication on the condition these concerns are addressed.

We thank reviewer 2 for her positive comments and useful feedback, particularly on the data analysis. We have changed our analysis based on her suggestion, described in Part 1 of Essential revisions above, and also addressed all her specific points in the manuscript.

- The authors need to provide sufficient statistics for the reviewer to determine validity and significance. For example, how many trajectories were collected from how many cells for each condition? How many cells were imaged using FISH? How many trajectories and/or single step displacements were used for the fitting? These numbers need to be provided in the figure legends or supplementary table.

We thank the reviewer for pointing out the issues. We have now added all the information to the source data and figure captions.

- The authors plot the osd distribution and use this to calculate the mRNA-associated fraction of Hfq, but use the overall MSD vs time to calculate the D of Hfq. The distribution of osd appeared to be lognormal sine the authors used Gaussian of log(osd) to do the fitting. If it is indeed lognormal, it means that most likely there are more than one Gaussian distributed population. The authors should use osd distribution (not log) or CDF to calculate Ds and subpopulations, and compare with that obtained from MSD, which is an average measurement. Such analysis should be conducted for all conditions.

We thank the reviewer for the good suggestion. We have reanalyzed all the data using the suggested method. We summarize the results from the new analysis in Essential revisions point 1, to make it easier to read for all three reviewers.

- The authors should exert caution to use the power law distribution of diffusion coefficient and molecular weight in bacterial cells to interpret the association of Hfq-bound mRNA with ribosomes. The D for Hfq under NT condition is too small (~ 0.5 um2/s) compared to what would be expected form the molecular weight of Hfq-mRNA complex (~ 550 kD), therefore the authors interprets that a fraction of the associated mRNAs are translated by ribosomes. Additionally, because rif-treated Hfq showed higher diffusion coefficient compared to what should be expected from its MW, the authors concluded that Hfq could dissociate into monomers when not bound to RNAs. It is known that proteins with different surface charges and mRNA molecules diffuse differently in bacterial cells and also respond to different stress responses. They do not necessarily scale with the expected MW especially at high MW. See Kumar et al., 2010, Lampo et al., 2017, and Schavemaker et al., 2017. To substantiate the authors' claims, translation inhibitor and Hfq hexamer mutants should be used. Similarly, subsection “sRNAs can displace Hfq from mRNAs in a face-dependent manner” "sRNA-bound Hfq-mMaple3 has similar diffusivity as free Hfq-mMaple3" based on similar MW is not substantiated.

We thank the reviewer for pointing this out. We agree that the interpretation on Mw based on the diffusion coefficient with the power-law relationship is not robust. Since the issue is brought up by all three reviewers, we removed this discussion from the manuscript, which does not affect any major conclusions of the manuscript.

Reviewer #3:

This manuscript provides single-particle tracking data that describes the movement of the RNA-binding protein Hfq in living E. coli cells. The authors investigate how the diffusion of Hfq is affected by the presence of RNA, the ability of Hfq to interact with RNA, and mutations in the major endoribonuclease RNase E. The authors also investigate how overexpression of different small RNAs affects Hfq diffusion.

Although the tracking data is of high quality, this manuscript suffers from several problems. Most worrying is that the authors interpret their data in an often careless manner. The manuscript contains many instances of over-interpretations, and some interpretations of data that are entirely incorrect.

This reviewer is particularly concerned about the emphasis on the role for Hfq in mRNA destabilization via recruitment of Hfq. It seems this is the only new message in this paper and thus the most critical and important one. As detailed below, most aspects of this model hinge upon assumptions that are contradicted by available biochemical data from the Luisi lab, or have not been addressed by clear-cut unambiguous experiments in this manuscript.

We thank this reviewer for his/her critiques and constructive suggestions. We have addressed the comments and concerns from the reviewer as shown below, and significantly revised the manuscript. However, we would still like to emphasize the novelty and significance of our study here. In addition to the new finding that Hfq might contribute to the regulation of mRNA turnover in the absence of the corresponding sRNA regulators, as the reviewer mentioned, our results provide strong in vivo support for some of the previous in vitro biochemical and functional tests and conclusions on Hfq-mRNA and Hfq-sRNA interactions. Finally, we also found that while class II sRNAs can directly compete with mRNA for binding to Hfq, class I sRNAs do not necessarily need to displace the mRNA, but can co-occupy with mRNA on Hfq through different binding sites. These results provide more concrete mechanisms to the previously suggested RNA exchange model.

1) The authors base many of their interpretations on the assumption that Hfq interacts directly with RNase E. From reading the manuscript, it appears as if a direct Hfq-RNase E interaction is a well-established fact. This is not correct. In fact, the interaction between Hfq and RNase E is highly controversial. According to work by the Aiba lab, Hfq and RNase E interact directly without a requirement for RNA (e.g. Ikeda et al., 2010). In contrast, the Luisi lab has shown that removing RNA from purified Hfq abolishes the Hfq-RNase E interaction (Bruce et al., 2008), and further showed by carefully conducted biochemical experiments that re-constituted Hfq-RNase E complexes only form when Hfq is pre-bound to an sRNA. According to Luisi's data, it is the sRNA (not Hfq) that directly interacts with RNase E (Worrall et al., 2018). Unfortunately, the authors have based a large part of their experiments and conclusions on this controversial assumption. Therefore, the authors need to explicitly state that it is unclear whether Hfq can bind RNase E directly, or whether this interaction is mediated through RNA. The data in this manuscript need to be re-interpreted in light of the uncertainty of the Hfq-RNase E interaction.

2) On the same topic, subsection “Hfq is deficient in releasing mRNAs without interactions with RNase E”: "Hfq has been demonstrated to interact with the C-terminal scaffold region of RNase". As mentioned above, it is unclear whether this interaction is direct or occurs via RNA. In fact, one of the papers (Bruce et al.,) cited after this sentence claims that the interaction is dependent on RNA, which is the complete opposite of the authors' statement.

We thank the reviewer for the two points above. It is possible that although the regulation on the selected mRNA does not require the corresponding sRNA regulators, some other RNAs are needed to bridge the Hfq-RNase E interaction. Our data only suggest the Hfq-RNase E interaction is needed, but cannot specify whether RNA is involved in the interaction, which we leave as an open question in the Discussion. We have now clarified the Hfq-RNase E interaction in the Results and Discussion as below:

“The C-terminal region of RNase E serves as a scaffold for the degradosome protein components (RNA helicase RhlB, enolase, and PNPase). Hfq has been demonstrated to interact with the C-terminal scaffold region of RNase E, although it is still under debate whether such interaction is direct or mediated by RNA28–31

“While our results reveal that Hfq binding contributes to mRNA turnover through recruitment of RNase E, more mechanistic details remain to be further elucidated. First, it is under debate whether Hfq-RNase E interaction is direct or mediated by RNAs28–31. While our data show that Hfq can promote mRNA degradation without their matching regulatory sRNAs, it remains to be investigated whether other cellular RNAs may participate in bridging the interaction. Second, while our data demonstrate that deletion of the scaffold region of RNase E abolishes the Hfqmediated regulation on mRNA turnover, the same scaffold region also includes the binding sites of other degradomes components, thus we cannot exclude the possibility that these protein components also participate in the regulation. Further experiments are needed to answer these mechanistic questions.”

3) The authors claim that rifampicin "inhibits transcription and results in the loss of most cellular mRNAs". A more correct wording would be that rifampicin treatment inhibits transcription and leads to a global loss of RNA. For instance, many Hfq-binding sRNAs have half-lives shorter than 15 minutes (e.g. Vogel et al., 2003). While the measured increased diffusion of Hfq after rifampicin treatment most likely indicates a higher fraction of Hfq free from RNA, it does not inform on the classes of RNA that contribute to this binding. Either the authors need to explicitly show that mRNAs are preferentially lost over other RNA classes after 15 minutes of rifampicin treatment, or the term "mRNA-associated fraction" used throughout the manuscript should be changed to "RNA-associated fraction".

We thank the reviewer for pointing this out, and indeed for this exact reason, we prefer to term it “mRNA-associated” and “mRNA-free” Hfq, as the diffusion coefficient change is more robustly depending on the presence of mRNA. From the sRNA expression data sets, we can see that the change in diffusion coefficient cannot distinguish sRNA-mRNA-associated Hfq from mRNAassociated Hfq, or sRNA-Hfq from mRNA-free Hfq. However, we think it is fair to name so as sRNA-mRNA-associated form is a subset of the mRNA-associated form, and sRNA-Hfq is a subset of mRNA-free form. Please note that mRNA-free form is not equal to free Hfq. To make this clear in the manuscript, we revised the text as below:

“It should be noted that under the current rifampicin treatment (200 µg/mL concentration for 15 minutes), mRNAs, which have an average half-live of 1-4 min50, are preferentially degraded, compared to tRNAs51 and rRNAs52. While many sRNAs show long half-lives when targetcoupled degradation is reduced in the absence of mRNAs upon rifampicin treatment53,54, some sRNAs do have short half-lives55. Therefore, rifampicin treatment might also reduce the fraction of Hfq bound by sRNAs. However, our data suggest that binding sRNA to RNA-free Hfq or to mRNA-associated Hfq did not change the diffusion coefficients of corresponding species (see sections below). Therefore, we interpreted that the change in the diffusion coefficient upon rifampicin treatment primarily reflected the binding of mRNAs to Hfq.”

“However, based on the discussion above, it is possible that a subpopulation of mRNA-free or mRNA-associated Hfq might be sRNA-associated Hfq, or sRNA-mRNA-Hfq tertiary complex, respectively.”

4) The authors claim that mutations in the distal phase of Hfq lead to "a large increase in the diffusivity under NT condition". This is a clear overstatement; according to Figure 2C, the increase compared to WT is less than two-fold.

We have revised the sentence as below. “Large” was referring to the comparison with the mutations in distal face or rim.

“..and both distal face mutations (Y25D and K31A) led to the largest increase in the diffusivity under NT condition, with the diffusion coefficients close to the rifampicin treated case …”

5) Subsection “Most Hfq proteins are occupied by mRNAs in the cell during exponential growth”: "Considering the average length of bacterial mRNAs to be 1 kb (~330 kDa), and Mw of bacterial ribosome (~2.5 MDa), this reduction in D supports the interpretation that a significant fraction of WT Hfq proteins are associated with mRNAs in the NT case, and that a fraction of the associated mRNAs are translated by the ribosomes." Do the authors imply that the majority of Hfq is associated with mRNA alone? An alternative explanation would be that most Hfq hexamers are simultaneously bound to mRNA and sRNA. The data does not inform on the nature of the complexes Hfq is involved in, other than that they contain RNA (according to the measured changes in diffusivity during rifampicin treatment). This should be clearly stated in the text.

We thank the reviewer for pointing this out. All three reviewers raised the same issue on the interpretation and discussion on the molecular weight. Our previous interpretation was based on the power-law relationship between molecular weight (Mw) and the diffusion coefficient (D) (D = aMwx, where x = -0.33) that was experimentally observed in bacterial cells (reference 42). However, as reviewer 2 suggested, inference of Mw just based on D may not be robust, since diffusion coefficient is also dependent on other factors, such as surface charge of the protein. In the revised manuscript, we have removed the part of the discussion, as it does not contribute to our major conclusions of the manuscript.

It is entirely possible that the mRNA-free Hfq contains both free Hfq and sRNA-bound

Hfq, however, the change in the diffusion coefficient caused by sRNA binding compared to free Hfq is not distinguishable. We therefore use “mRNA-free” Hfq to refer to the fraction of Hfq that does not associate with mRNA. Please refer to the response to point (3) above.

6) Subsection “Hfq is deficient in releasing mRNAs without interactions with RNase E”: "The rneΔ14 mutant has a smaller fraction of the C-terminal scaffold (residues 636-845) deleted, encompassing the Hfq binding region." The deleted part of RNase E also encompasses two RNA-binding domains, as well as binding sites for proteins RhlB and enolase. The differences in activity between WT and rneΔ14 could thus stem from impairment of many different interactions and do not specifically report on the loss of a potential interaction with Hfq.

We have revised the text for the description of the rneΔ14 mutant, and added discussion regarding the Hfq-RNase E interaction.

“The rneΔ14 mutant has a smaller fraction of the C-terminal scaffold (residues 636-845) deleted, encompassing the Hfq, RhlB and enolase binding regions and two RNA-binding domains59.”

“While our results reveal that Hfq binding contributes to mRNA turnover through recruitment of RNase E, more mechanistic details remain to be further elucidated. First, it is under debate whether Hfq-RNase E interaction is direct or mediated by RNAs28–31. While our data show that Hfq can promote mRNA degradation without their matching regulatory sRNAs, it remains to be investigated whether other cellular RNAs may participate in bridging the interaction. Second, while our data demonstrate that deletion of the scaffold region of RNase E abolishes the Hfq mediated regulation on mRNA turnover, the same scaffold region also includes the binding sites of other degradomes components, thus we cannot exclude the possibility that these protein components also participate in the regulation. Further experiments are needed to answer these mechanistic questions.”

7) Subsection “Hfq is deficient in releasing mRNAs without interactions with RNase E”: "In both RNase E mutant backgrounds, the diffusivity of Hfq-mMaple3 became less sensitive to transcription inhibition by rifampicin compared to the WT rne case (Figure 3a and 2c)." The authors should show the data for strains with WT and mutant RNase E in the same graph for easier comparison. In addition, a statistical analysis should be provided to test whether the claimed differences are significant.

We have reanalyzed all the data according to reviewer 2’s suggestion and replotted all the figures for better comparison (new Figure 4 for this specific case). P-values are provided for the comparison.

8) Subsection “Hfq is deficient in releasing mRNAs without interactions with RNase E”: "40-50% of Hfq-mMaple3 remained mRNA associated upon rifampicin treatment in the RNase E mutant backgrounds (Figure 3B)." I could not find the corresponding numbers for the strain with WT RNase E. Please provide these numbers in the figure and/ or in this part of the text for clarity.

We apologize for not being clear in the original manuscript -- this same issue was also brought up by the reviewer 1. In the previous analysis, we assumed that the mRNA-associated fraction is close to zero after rifampicin treatment in the WT rne background. With this assumption, fitting of the rifampicin treated case in the rne131 background generated 40-50% of mRNA-associated fraction. We have now reanalyzed all the data based on reviewer 2’s suggestion. The mRNAbound fractions are reported now for all conditions in the new Figure 2 and Figure 4. The details on the new data analysis method are described in the new Figure 1—figure supplement 3 and Figure 2—figure supplement 2, and summarized in the Essential revisions part 1 above.

9) Subsection “Hfq is deficient in releasing mRNAs without interactions with RNase E”: "These observations suggest that without the Hfq-RNase E interaction, more mRNAs remain bound to Hfq, indicating that Hfq may help deliver the associated mRNA to RNase E for degradation." This is a very far-reaching interpretation of the data presented in Figure 3. There is no evidence for Hfq-mediated delivery of RNase E in Figure 3. I strongly advise the authors to use a more careful interpretation of the data.

We have revised the text as below:

“These observations suggest that without the Hfq-RNase E interaction, more mRNAs remained bound to Hfq, and hint that Hfq-RNase E interaction may help recycle Hfq from the mRNA-associated form through degradation of mRNAs.”

10) In Figure 3C, the authors show half-life measurements of several mRNAs that are regulated by sRNAs. For these experiments, strains with corresponding sRNA gene deletions were used. I assume that the rationale for this was to avoid putative differences due to impaired sRNA regulation. This is not a stringent strategy, as it is unknown whether other sRNAs target these mRNAs. Moreover, from the bar charts in Figure 3C, it appears that mRNA half-lives increase when Hfq carries a distal face mutation. However, in the Northern blots used to create the bar charts (corresponding supplementary figure), there seem to be very small (if any) differences in half-lives between WT and mutant Hfq. For transparency, and easier interpretation for the reader, the authors should (instead of bar charts) plot the log10-transformed relative band intensities versus time, and show all data points (not only error bars). They should also include the corresponding Northern blots in Figure 3 along with the quantifications.

The differences in the mRNA half-life are 50% to 2-fold for the tested three mRNAs, which are indeed not a big difference, but correctly reflect the Northern blot image. We do think there is a very noticeable difference between Y25D and WT Hfq even from the Northern blot. We also include a more stringent strategy as also suggested by reviewer 1, using Hfq Q8A mutant. The detailed description can be found in point 2 of Essential revisions above.

We have moved the Northern blot to the main figure (now in the new Figure 5). In the corresponding figure supplement, we plot individual data points from each replicate, and the mean ± standard deviation calculated from all replicates in log scale. We also show the fitting curves to extract the half-lives. The equation we used for half-life fitting is described in the method as previously published.

11) The authors propose that Hfq, through an interaction with RNase E, promotes mRNA degradation in an sRNA-independent fashion. They also suggest that "mRNA-occupied Hfq proteins are in standby mode for sRNA binding if needed". If this were correct, one would expect that a high cellular concentration of a Class II sRNA, that is a strong competitor for Hfq's interaction with mRNAs, would result not only in displacement of Hfq from many mRNAs, but also thereby increase their stability. ChiX is a strong competitor for distal phase binding that should cause such an effect. The authors interpret their data to imply that ChiX overexpression results in displacement of Hfq from many mRNAs. However, they do not provide evidence that this displacement results in general mRNA stabilization, which should be the outcome according to their model. In fact, induction of ChiX to a high intracellular concentration from a plasmid resulted in downregulation of one specific mRNA (ybfM), rather than a stabilizing effect on many mRNAs (Rasmussen et al., 2009). It is very surprising that the authors did not cite this paper.

We thank the reviewer for the good suggestion. We added new lifetime measurement in the presence of WT ChiX and ChiX with two AAN motif deleted, and found that the presence of WT ChiX can indeed increase the sdhC lifetime, similar to Y25D Hfq mutant, whereas the mutant ChiX that has reduced Hfq binding ability was not able to increase the lifetime as significantly as the WT ChiX. The new data is presented as the new Figure 6D. It is worth mentioning that sdhC is not a known target of ChiX, therefore the increase in the stability of sdhC is an indirect effect due to competitive binding of ChiX to Hfq. ybfM is a specific target for ChiX, the destabilization of ybfM is a direct result of ChiX-mediated regulation. As in our specific case, since we are not studying ChiX targets, we do not find a good place for citing this particular paper even with the new data.

“As our model suggests that binding of Hfq to the mRNA through the distal face can regulate the mRNA turnover, we reasoned that sRNAs that can effectively compete for Hfq binding against mRNAs may decoy Hfq from this regulatory function. To test this, we again used sdhC as an example, and measured its half-life in the presence of ChiX, which is a strong competitor for Hfq binding (Figure 3). In the presence of vector control, sdhC exhibited comparable half-life compared to the case without any plasmid (Figure 5D, Figure 6C and D). The presence of WT ChiX increased the half-life by ~70%, whereas the mutant ChiX without two AAN motif deleted only increased the half-life by ~42%, consistent with its reduced binding ability to Hfq (Figure 3A, 3B, Figure 6C and 6D). These results further support our model of Hfq-mediated regulation of mRNA turnover, and demonstrate that the presence of strong Hfq binding sRNAs can modulate the strength of Hfq’s regulation.”

“…we observed that sRNA competitors, such as ChiX, which can outcompete mRNAs for binding at the distal face, can decoy Hfq from regulating mRNA turnover, the same effect as the Y25D mutation. Similar observation was reported previously that ChiX can titrate Hfq from translational repressing transposase mRNA67.”

12) Several reports have shown that the 5' moiety of RNA substrates largely influences on both, RNase E cleavage efficiency and specificity of cleavage site selection (e.g. Mackie, 1998, Jiang and Belasco, 2004, Bandyra et al., 2012); RNAs carrying a 5' monophosphate are substantially better substrates than those with a 5' tri-phosphate. Regarding the Hfq-dependent mRNA degradation proposed by the authors, do these mRNAs need to be decapped for RNase E to degrade them? Or does the degradation go through the substantially less efficient internal cleavage route? The authors should discuss both their data and their models with respect to what is known about the cleavage activity/ specificity of RNase E. They should also cite the most seminal papers on this subject.

We thank the reviewer for the good suggestion. We have performed additional experiments to measure the lifetime of shdC, as an example. The revised manuscript is quoted as below.

“To test whether this Hfq-mediated mRNA turnover is dependent on the 5’-end decapping, we used sdhC as an example, and compared its half-life in the backgrounds of ∆sRNA∆rppH and ∆sRNA∆rppH hfqY25D.[…] These results suggest that the decapping by RppH is not required for the Hfq-mediated regulation of mRNA turnover, at least for the case of Hfq regulation on sdhC mRNA.”

13) Regarding the EMSA shown in Figure 4. Subsection “sRNAs can displace Hfq from mRNAs in a face-dependent man”: "Results show that RyhB cannot displace the radiolabeled ptsG from Hfq, but rather generates an additional upper-shifted band compared to the band of ptsG-Hfq complex, supporting that RyhB and ptsG can co-occupy Hfq". This is a very creative interpretation of an inconclusive result. What can be deduced from the gel picture is that the band representing the Hfq-ptsG complex becomes weaker at the two highest RyhB concentrations. The reason for this could be either that RyhB displaces ptsG, or that a ternary complex is formed. The design of the experiment makes it impossible to judge whether the former is happening; since the majority of ptsG mRNA is not in complex with Hfq (even in the absence of RyhB), it is impossible to judge whether high concentrations of RyhB results in increased free ptsG mRNA. Regarding the latter possibility (which is put forward by the authors), the "upper-shifted band" is barely visible and do not by any means reach the intensity of the ptsG-Hfq band, which would be the case if the major effect of RyhB addition would be the formation of a ternary complex.

We thank the reviewer for pointing this out. The purpose of the EMSA assay is to show the feasibility of forming ptsG-RyhB-Hfq tertiary complex, as ptsG and RyhB are not matching sRNA-mRNA pair. The purpose of the EMSA is not to quantitatively determine the fraction or efficiency of tertiary complex formation, as it is difficult to fully recapitulate the in vivo condition in an in vitro experiment. But we agree with the reviewer that the quality of the previous gel is not good. We therefore repeated the experiments starting with a higher ptsG-Hfq complex to chase with RyhB or ChiX. The new results are presented in the new Figure 3E and 3F. The gel consistently showed the formation of the ptsG-RyhB-Hfq complex when chased with RyhB, but not ChiX. However, in the EMSA assay, we also observed direct displacement of ptsG fragment by RyhB, which was not indicated by the in vivo imaging results, as our in vivo imaging did not show a shift of Hfq to a fast-diffusing population when overexpressing RyhB. We do not know the exact cause of the discrepancy, but reason that it can largely result from the difference between the cellular conditions and in vitro setting. Nevertheless, the EMSA results still support that class I and class II sRNAs can have different mechanisms to gain access to mRNA-occupied Hfq, and that it is structurally possible to have a class I sRNA co-occupy with a non-target mRNA on Hfq.

14) Subsection “sRNAs can displace Hfq from mRNAs in a face-dependent man”: "In addition, droplet digital PCR (ddPCR) performed in the same conditions as the diffusivity assays showed that RyhB level was comparable to ChiX (Figure 4d)." This is not correct. According to Figure 4d ChiX is almost ten times more abundant than RyhB. The same incorrect statement is repeated later in this subsection.

We thank the reviewer for pointing this out. We added a different normalization method, against the empty vector (results shown in the new Figure 3D). While ChiX level was ~5-fold of the RyhB level when normalized the reads of 16S rRNA , ChiX level was about 50% of RyhB level when normalized to the reads from empty vector (representing the induction fold change). We therefore reason that the difference between ChiX and RyhB when normalizing to the 16S rRNA was very likely due to the different efficiency during RT and PCR steps for these two targets. We have explicitly explained this in the revised manuscript.

15) Subsection “sRNAs can displace Hfq from mRNAs in a face-dependent man”: "EMSA and ddPCR results suggest that both in vitro and in vivo, RyhB can effectively access mRNA-occupied Hfq through co-occupying Hfq from the proximal face." This is an unsubstantiated and probably incorrect interpretation of the results. The EMSA does not provide evidence for a ternary Hfq-ptsG-RyhB complex. In what way does the ddPCR result inform on binding of RyhB to mRNA-bound Hfq?

We apologize for not explaining this well in the previous version. We hope that through the response to point (13), we can convince the reviewer that the EMSA provides in vitro evidence that it is feasible to form Hfq-ptsG-RyhB complex. Our logic is as follows: (1) our live-cell tracking results suggest that RyhB expression does not change Hfq diffisity to a fast-diffusing population, in contrast to the case of ChiX, therefore RyhB does not displace mRNA from Hfq effectively. (2) Since RyhB does not displace mRNA from Hfq, it would either be unstable in the cell if it does not bind Hfq at all, therefore showing low cellular abundance, or it can bind to Hfq with mRNA bound at the same time. Using ddPCR and FISH, we showed the RyhB (or SgrS, which shows the same behavior in tracking experiments) level is not significantly lower than ChiX (SgrS level is even higher than ChiX using both normalization methods). We therefore think that RyhB can occupy Hfq together with mRNA. (3) We use EMSA to show the feasibility of forming the tertiary complex, even though the actual fraction of the tertiary complex could be different between in vivo and in vitro experiments. We have significantly revised the manuscript to improve the reasoning here.

“Overexpression of RyhB or SgrS, in contrast, did not cause any significant changes in the HfqmMaple3 diffusivity or the corresponding mRNA-associated fraction (Figure 3A and B). […] Nevertheless, the EMSA results still support that class I and class II sRNAs can have different mechanisms to gain access to mRNA-occupied Hfq, and that it is structurally possible to have a class I sRNA co-occupy with a non-target mRNA on Hfq.”

16) Conceptually, there is no apparent reason why Hfq would not interact with RNA undergoing transcription. A previous Hfq tracking study in E. coli indeed reported a three-state model in which the slowest state was interpreted as Hfq bound to RNAs during transcription (Persson et al., 2013). The authors should comment on this finding with regard to their own results. Do the data presented in the current manuscript fit with the previous model? If not, why not?

We have indeed tried to use the same method as in the Nature Methods paper to analyze our data, and got consistent results with their published results on the WT Hfq. Specifically, we also observed a third slowest diffusion state of Hfq, which corresponds to the interpreted Hfq fraction that binds to mRNAs co-transcriptionally. However, due to technical reasons, we decided not to use this analysis method, but used the method as suggested by reviewer 2. For the detailed reason please refer to the Essential revisions part 1. In our method, however, we do not separate Hfq binding to mature mRNAs and mRNA co-transcriptionally. But this does not affect any of our current conclusions.

[Editors’ note: what follows is the authors’ response to the second round of review.]

Essential revisions:

However, two critical points still need to be addressed:

1) One of the major findings put forward by the authors, which is highlighted both in the abstract and in the schematic Figure 8, is the suggestion that binding of Hfq to mRNAs can recruit RNase E for (sRNA-independent) mRNA degradation. The authors provide live cell Hfq tracking data in strains with mutations in the Hfq distal and proximal phases, combined with full-length or mutant RNase E (Figure 4). The tracking data show that the differences in Hfq diffusion, which are observed between WT and mutant RNase E upon rif treatment, are abolished when Hfq has the distal phase mutation. These data are sound and convincing but do not directly address whether the observed effects are indeed due to mRNA degradation.

To directly monitor effects on mRNA degradation, the authors provide Northern blot data monitoring the half-life of three selected mRNAs (Figure 5). According to the authors, and many previous studies, the distal face mutation primarily impairs Hfq-mRNA interactions, while the proximal face mutant primarily impairs Hfq-sRNA interactions, at least when considering Class I sRNAs. In Figure 5, the presented data show that, in the strain expressing WT RNase E, both the distal and the proximal face mutants lead to increased mRNA half-lives. It is well established that sRNA-dependent (Class I sRNAs) regulation requires Hfq to contact both the sRNA and the mRNA. If Hfq-dependent degradation of mRNAs where to be mediated strictly through sRNAs, the increase half-life should be the same for the distal and proximal mutants (binding to both RNAs are required). However, according to the authors, the data presented in Figure 5 shows a greater increase in mRNA half-lives with the distal face mutant compared to the proximal face mutant, and is interpreted as representing Hfq-dependent and sRNA-independent degradation by RNase E. From this follows that the contribution of a sRNA-independent effect on mRNA half-lives can be deduced from the difference between the values obtained in the distal mutant and in the proximal mutant, while the difference between WT Hfq and the distal face mutant is the sum of the sRNA-dependent and sRNA-independent effects on mRNA degradation. In other words, if the authors' hypothesis is correct, there should be a significant difference in mRNA half-life between the distal and proximal face mutants, and this difference should be abolished if Hfq cannot interact with RNase E.

Unfortunately, the data provided in Figure 5 do not provide unequivocal evidence for the following reasons: (i) too few replicates: some of the mean values were calculated from only two data points, therefore the standard deviations are not meaningful (ii) there is no information on how many replicates where used for calculating each specific mean value, making it impossible to judge the how reliable specific mean values and standard deviations are, (iii) there is no statistical analysis provided to ensure that the differences are significant (which on the other hand is not possible with only two data points). For these reasons, it is impossible to judge whether the differences between Y25D and Q8A in the WT and rne131 mutant are significant. To test whether the hypothesis is correct, the authors need to provide at least three (four would be advisable) replicates for each mean value, and use statistical tests to assess whether the proposed differences are indeed significant. To increase the possibility for the reader to judge the data, each bar showing mean values in Figure 5D should to be overlaid with the value of each data point used for calculating the mean.

We agree with the reviewers on the interpretation of the data. We have now included 4 replicates for each data set in Figure 5. In addition, we also added 1 or 2 additional replicates for Figure 6 (3 or 4 replicates in total). We have provided p-values for each pairwise comparison. Each bar graph was overlaid with each data point for readers’ judgement. The addition of new replicates changes the mean values slightly, but does not change the conclusions.

2) While the analysis is improved, there appeared to be a misunderstanding on how to use CDF to fit displacement to extract the diffusion coefficients and population percentages. The authors used the CDF fitting of the apparent speed of one step displacement (osd) to extract the average speed and population percentage of Hfq. The authors cited the Yang, 2019 paper as the source. Do note that in Yang, 2019, the authors were measuring the directional moving speed but not random diffusion. For non-processively diffusing molecules, the diffusion coefficient D should be used. While the speed CDF from osd can also be used to extract different populations based on the difference in speed, it is different from classifying molecules based on their apparent diffusion coefficients (note that only osd2, but not osd, is proportional to D). Two molecules diffusing at different speeds (with random orientations) will have different Ds, but the quantitative difference between the two Ds is not the same as that between the two speeds (again, osd2 , but not osd, is proportional to D). Therefore, the classification of fast and slow diffusing molecules based on diffusion coefficient could be different from that based on speed. Furthermore, the mean squared displacement (MSD) measures how far the molecule diffuses away over different time lags. MSD is also an averaged measurement of molecules of different Ds, and not particularly accurate when there are at least two populations of different Ds, which is the case the authors are trying to establish. See https://link.springer.com/protocol/10.1007/978-1-59745-513-8_14 for a review, and Bettridge, et al. https://doi.org/10.1111/mmi.14572, in the supplemental notes for a practical guide. The fractionation of subpopulations based on diffusion coefficient would be important for the authors to make the argument of whether mRNA binding would lower the D or vice. versa. If the difference in speed is fairly large, I suspect that the major conclusions should still hold if the authors switch to the analysis of CDF of osd2, but the correct analysis should be provided.

We thank the reviewer for pointing out the correct way of analyzing the data. As the reviewer wrote, our raw data (one-step displacement) are from random diffusion. Thus, the correct fitting functions for this case are not the error functions (which are for directional motions) as in the previous submission, but the exponential functions as expressed in https://doi.org/10.1111/mmi.14572. All the mRNA-associated fraction plots were updated based on the new fitting, and new fitting values are listed in the Supplementary file 1. Details of the fitting method are presented in the updated Materials and methods section. The new analysis does not change the conclusions of the manuscript.

https://doi.org/10.7554/eLife.64207.sa2

Article and author information

Author details

  1. Seongjin Park

    Department of Biochemistry and Molecular Biology, The University of Chicago, Chicago, United States
    Contribution
    Data curation, Formal analysis, Validation, Visualization, Methodology, Writing - review and editing
    Contributed equally with
    Karine Prévost
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4298-1770
  2. Karine Prévost

    RNA Group, Department of Biochemistry, University of Sherbrooke, Sherbrooke, Canada
    Contribution
    Data curation, Formal analysis, Validation, Visualization, Methodology, Writing - review and editing
    Contributed equally with
    Seongjin Park
    Competing interests
    No competing interests declared
  3. Emily M Heideman

    Department of Biochemistry and Molecular Biology, The University of Chicago, Chicago, United States
    Contribution
    Data curation, Formal analysis, Validation, Visualization, Writing - review and editing
    Contributed equally with
    Marie-Claude Carrier
    Competing interests
    No competing interests declared
  4. Marie-Claude Carrier

    RNA Group, Department of Biochemistry, University of Sherbrooke, Sherbrooke, Canada
    Contribution
    Data curation, Formal analysis, Validation, Visualization, Methodology, Writing - review and editing
    Contributed equally with
    Emily M Heideman
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3878-4141
  5. Muhammad S Azam

    Department of Biochemistry and Molecular Biology, The University of Chicago, Chicago, United States
    Contribution
    Data curation, Validation, Writing - review and editing
    Competing interests
    No competing interests declared
  6. Matthew A Reyer

    Institute for Biophysical Dynamics, The University of Chicago, Chicago, United States
    Contribution
    Software, Formal analysis, Validation, Visualization
    Competing interests
    No competing interests declared
  7. Wei Liu

    Department of Biochemistry and Molecular Biology, The University of Chicago, Chicago, United States
    Contribution
    Formal analysis
    Competing interests
    No competing interests declared
  8. Eric Massé

    RNA Group, Department of Biochemistry, University of Sherbrooke, Sherbrooke, Canada
    Contribution
    Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Project administration, Writing - review and editing
    For correspondence
    eric.masse@usherbrooke.ca
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5253-1415
  9. Jingyi Fei

    1. Department of Biochemistry and Molecular Biology, The University of Chicago, Chicago, United States
    2. Institute for Biophysical Dynamics, The University of Chicago, Chicago, United States
    Contribution
    Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    jingyifei@uchicago.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9775-3820

Funding

National Institutes of Health (1DP2GM128185-01)

  • Jingyi Fei

Searle Scholars Program

  • Jingyi Fei

National Institutes of Health (R01 GM092830-06A1)

  • Eric Massé

Canadian Institutes of Health Research (MOP69005)

  • Eric Massé

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank CK Vanderpool and X Ma for sharing the plasmid containing mMaple3 template, E Perozo for sharing the plasmid containing Lac repressor and operator system, AA Kossiakoff for sharing the plasmid containing the synthetic antibody, Y Wang for sharing MATLAB codes for extracting tracking trajectories, and the MRSEC Shared User Facilities at the University of Chicago (NSF DMR-1420709) for providing 3D printed imaging chambers. We thank the Service de Purification de Protéines de l’Université de Sherbrooke (SPP) for Hfq purification and the Plateforme RNomique de l’Université de Sherbrooke for ddPCR. J Fei acknowledges the support by the Searle Scholars Program and NIH Director’s New Innovator Award (1DP2GM128185-01). Work in E Massé Lab has been supported by an operating grant MOP69005 from the Canadian Institutes of Health Research (CIHR) and NIH Team Grant R01 GM092830-06A1.

Senior and Reviewing Editor

  1. Gisela Storz, National Institute of Child Health and Human Development, United States

Reviewer

  1. Jie Xiao, Johns Hopkins University, United States

Publication history

  1. Received: October 21, 2020
  2. Accepted: February 19, 2021
  3. Accepted Manuscript published: February 22, 2021 (version 1)
  4. Version of Record published: March 23, 2021 (version 2)

Copyright

© 2021, Park et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,262
    Page views
  • 269
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Biochemistry and Chemical Biology
    2. Cell Biology
    Wan Hua Li et al.
    Research Article

    SARM1 regulates axonal degeneration through its NAD-metabolizing activity and is a drug target for neurodegenerative disorders. We designed and synthesized fluorescent conjugates of styryl derivative with pyridine to serve as substrates of SARM1, which exhibited large red-shifts after conversion. With the conjugates, SARM1 activation was visualized in live cells following elevation of endogenous NMN or treatment with a cell-permeant NMN-analog. In neurons, imaging documented mouse SARM1 activation preceded vincristine-induced axonal degeneration by hours. Library screening identified a derivative of nisoldipine as a covalent inhibitor of SARM1 that reacted with the cysteines, especially Cys311 in its ARM domain and blocked its NMN-activation, protecting axons from degeneration. The Cryo-EM structure showed that SARM1 was locked into an inactive conformation by the inhibitor, uncovering a potential neuroprotective mechanism of dihydropyridines.

    1. Biochemistry and Chemical Biology
    Renata O Pereira et al.
    Research Article

    Adrenergic stimulation of brown adipocytes alters mitochondrial dynamics, including the mitochondrial fusion protein optic atrophy 1 (OPA1). However, direct mechanisms linking OPA1 to brown adipose tissue (BAT) physiology are incompletely understood. We utilized a mouse model of selective OPA1 deletion in BAT (OPA1 BAT KO) to investigate the role of OPA1 in thermogenesis. OPA1 is required for cold-induced activation of thermogenic genes in BAT. Unexpectedly, OPA1 deficiency induced fibroblast growth factor 21 (FGF21) as a BATokine in an activating transcription factor 4 (ATF4)-dependent manner. BAT-derived FGF21 mediates an adaptive response, by inducing browning of white adipose tissue, increasing resting metabolic rates, and improving thermoregulation. However, mechanisms independent of FGF21, but dependent on ATF4 induction, promote resistance to diet-induced obesity in OPA1 BAT KO mice. These findings uncover a homeostatic mechanism of BAT-mediated metabolic protection governed in part by an ATF4-FGF21 axis, that is activated independently of BAT thermogenic function.