Introduction

RNase E (RNE) is the main endoribonuclease in Escherichia coli, known for its role in RNA processing and mRNA degradation^1–3. It is an essential protein^4,5, and homologous proteins are found across many bacterial species^6–8. The essentiality stems from the N-terminal domain (NTD), or the catalytic domain⁹. The NTD is followed by a membrane targeting sequence (MTS) and the C-terminal domain (CTD), or macromolecular interaction domain², where RhlB (a DEAD-box RNA helicase), PNPase (a 3’→5’ exonuclease), and enolase (a glycolytic enzyme) bind to form the RNA degradosome complex⁶. The MTS forms an amphipathic α helix, responsible for the localization of RNE on the inner membrane^10,11. Interestingly, the membrane localization of RNE and the presence of the CTD are not essential in E. coli nor are they conserved across bacterial species, in contrast to the broad conservation of the NTD across bacteria as well as chloroplasts^7,12,13. This raises a question about the roles of membrane localization and the CTD in the in vivo function of RNE.

E. coli strains with cytoplasmic RNE (due to the removal of the MTS) are viable, although they grow more slowly than the wild-type (WT) cells^10,14. In vitro studies have shown that membrane binding of RNE does not necessarily increase its enzymatic activity^10,14. However, membrane localization is likely important for gene regulation in vivo because RNE becomes sequestered from the cytoplasmic pool of mRNAs, giving mRNAs time for translation. This idea is supported by our recent observation that the membrane-bound RNE limits the degradation of nascent mRNAs while cytoplasmic RNE (ΔMTS) can degrade nascent mRNAs during transcription¹⁵. We found that transcripts encoding membrane proteins can be an exception to this rule, in that they can experience co-transcriptional degradation assisted by the transertion effect¹⁵. These findings agree with results from a genome-wide study, indicating that the membrane localization of RNE allows for differential regulation of mRNA stability for genes encoding cytoplasmic proteins versus inner membrane proteins in E. coli¹⁶.

Previous studies have reported evidence that E. coli RNE can localize in the cytoplasm—for example, when cells were grown anaerobically¹⁷ or when membrane fluidity was reduced by changes in lipid composition¹⁸. These findings imply that RNE can dissociate from the membrane; however, the origin of its weak membrane binding remains unknown.

Across bacteria, several species within α-proteobacteria have cytoplasmic RNE¹⁹ while other species possess membrane-bound RNE. Among these, B. subtilis RNase Y (a functional homolog of RNE) associates with the membrane via a transmembrane (TM) motif²⁰, instead of an amphipathic motif used by E. coli and other γ-proteobacteria⁷. Given the diversity of membrane-binding motifs that have arisen through evolution, it should be possible to engineer E. coli RNE with a TM motif. Such a mutant would provide a useful model for investigating the impact of membrane-binding motifs on the localization, diffusion, and activity of RNE.

Lastly, the CTD of E. coli RNE is an intrinsically disordered region²¹ that, while nonessential for cell viability, enhances the enzymatic activity of the NTD^15,22–24. As the primary binding site for degradosome components⁶, the CTD is thought to facilitate mRNA degradation by recruiting these proteins near the catalytic NTD. However, our recent study showed that these associated proteins have minimal impact on the degradation rate of lacZ mRNA, whereas deletion of the CTD markedly stabilizes the transcript¹⁵, suggesting a possible intramolecular allosteric effect within RNE. In the present study, we further show that the CTD modulates the membrane binding affinity of RNE, possibly by affecting its conformation. These results reveal a previously underappreciated role of the CTD in regulating RNE function beyond degradosome assembly, with implications for how the spatial organization and structural dynamics of RNE fine-tune RNA degradation in bacteria.

In this study, we quantified the membrane binding percentage (MB%) of RNE in E. coli using single-molecule microscopy and showed that membrane association governs its diffusion and mRNA-degradation activity. Perturbing the native MTS, substituting it with LacY TM segments, and deleting the CTD collectively revealed how the MTS and CTD set RNE’s spatial organization and provided routes for tuning activity through subcellular control.

Results

Membrane binding percentage (MB%) of RNE

First, we investigated the subcellular localization of RNE in live cells. Previous fluorescence microscopy studies have shown that RNE is localized to the inner membrane in E. coli^10,11,16, but the percentage of membrane-bound molecules has not been quantitatively examined in live cells. To address this gap, we fused RNE with a photo-convertible fluorescent protein, mEos3.2²⁵ and imaged individual RNE molecules over time in two dimensions (Fig. 1A). The positions of fluorescent molecules were identified in each frame and linked into trajectories using the open-source software u-track²⁶ (Fig. 1A). In this section, we analyze localization, and diffusion dynamics are addressed in subsequent sections.

Figure 1:

Subcellular locations were calculated relative to the cell boundaries identified from bright-field images using another open-source image analysis package Oufti²⁷ (Fig 1B). To combine data from many cells, molecular positions along the short and long axes of a cell were normalized to the cell width and cell length, yielding xNorm and yNorm, respectively (Fig. 1C). Based on yNorm, molecules within the cylindrical part of the cell were selected, and their xNorm values were used to generate an xNorm histogram. Hereinafter, we focus on the xNorm histogram to compare the membrane enrichment across protein constructs.

The xNorm histogram of RNE shows two peaks corresponding to the inner membrane on each side of the cell (Fig. 1D, Fig. S1A), very similar to the xNorm histogram of LacY, obtained by imaging LacY-mEos3.2 using the same method. LacY is a membrane channel for lactose and is composed of 12 TM segments²⁸. It is expected to be inserted into the inner membrane during translation^29–31, such that all imaged LacY is expected to be localized in the inner membrane³².

We further quantified the percentage of molecules bound to the membrane (membrane-binding percentage or MB%) from the xNorm histogram. For this analysis, we first confirmed that proteins localized on the membrane and in the cytoplasm are detected with equal probability, despite differences in their mobilities (Fig. S1B-C). Next, we developed a mathematical model based on a 2D projection of molecules randomly distributed either on the surface of or within a cylinder. The model includes imaging effects that affect the shape of xNorm histograms: localization error, the limited focal depth of the quasi-TIRF illumination we used, and the location of the inner membrane relative to the cell boundary (Fig. S1D-G). Model fitting was performed using a Markov-Chain Monte Carlo algorithm (MCMC). For validation, we applied the model to xNorm histograms of LacY and LacZ, which serve as benchmarks for complete membrane binding and complete cytoplasmic localization, respectively. The MB% of LacY was 99% with a 95% confidence interval of [96%, 100%], and LacZ showed MB% of 3.4% [0.2%, 7.3%] (Fig. 1E-F). Both MB% agree with the expectations for membrane and cytoplasmic proteins. For RNE, we found an MB% of 93% [91%, 96%] (Fig. 1E). The xNorm histogram of the fastest 7% of the RNE population (based on the diffusion coefficient, as discussed below) exhibited a cytoplasmic localization pattern (without the two membrane-associated peaks), supporting the existence of a cytoplasmic RNE subpopulation (Fig. S2A).

We confirmed that the MTS (residue 568-582) is essential for the membrane binding of RNE, as deletion of the MTS sequence made the xNorm like that of LacZ (Fig. 2A-B, Fig. S2D). We note that the xNorm profile of RNE ΔMTS was slightly different from that of LacZ near the center line of the cell (x = 0), suggesting fewer RNE ΔMTS molecules were at the midline of the cell. Also, xNorm fitting yielded an MB% of 33%. These results may reflect nucleoid exclusion of RNE ΔMTS due to its large size³³. Consistently, a smaller cytoplasmic RNE variant generated by CTD truncation (RNE (1-529) or RNE ΔMTS ΔCTD) exhibited a xNorm profile closer to that of LacZ (Fig. 2B, Fig. S2E).

Figure 2:

Is there a critical residue(s) within the MTS required for membrane binding? The MTS forms an amphipathic α-helix, in which hydrophobic residues are expected to align on one side of the helix (Fig. 2C)¹⁰. Previous studies suggested that replacing one of the hydrophobic residues with a hydrophilic amino acid can disrupt membrane binding of RNE^10,11. We revisited the point mutations discussed in Khemici et al¹⁰ and analyzed MB% by mEos3.2 imaging. We found that the F574A F575A double mutation (noted as F574AA) did not significantly affect MB%, consistent with the fact that the substituted amino acids remain hydrophobic (Fig. 2D-E, Fig. S2F). In contrast, F582E and F575E mutations reduced MB% to 47% and 50%, respectively (Fig. 2E), and their xNorm histograms resembled that of RNE ΔMTS, suggesting predominantly cytoplasmic localization (Fig. 2D, Fig. S2D, S2G-H). These findings indicate that the phenylalanine residues at positions 575 and 582 are critical for membrane association of RNE.

Effect of membrane binding on the diffusion of RNE

The membrane localization of RNE likely limits its diffusion and interaction with mRNA targets. Additionally, interaction with ribosome-bound mRNAs can further slow the diffusion of RNE. Here, we examined how these factors contribute to the diffusion dynamics of RNE.

To measure the diffusion of RNE, we analyzed the trajectories of individual RNE-mEos3.2 imaged at a 21.7-ms acquisition interval (Fig. 1A). We calculated the diffusion coefficient D by fitting the mean-squared displacement (MSD) of each trajectory to the equation MSD = 4Dτ + b, where τ is lag time and b accounts for both dynamic and static localization errors³⁴ (Fig. 3A). We obtained D_RNE = 0.0184 ± 0.0002 µm²/s (mean ± SEM). This value was well above the lower detection limit of our microscope, determined using stationary, surface-immobilized mEos3.2, D = 0.0020 ± 0.0001 μm²/s (Fig. S3A, Supplementary Discussion). Notably, D_RNE was comparable to that of ribosome-bound mRNAs, estimated to be D ∼0.015 µm²/s based on the diffusion of ribosomal protein L1 (Fig. S3B).

Figure 3:

To assess how membrane association affects diffusion, we compared D of WT RNE and the ΔMTS mutant. These two proteins have similar molecular masses, as the MTS comprises only 15 of the 1061 residues in a monomer of RNE (Fig. 2A). Therefore, any difference in D can be attributed to their subcellular localizations (membrane vs cytoplasm) rather than mass. We found that D_ΔMTS is ∼5.5 times that of D_RNE (Fig. 3B).

We examined another pair of RNE mutants that differ in localization (membrane vs cytoplasm) but are similar in size: membrane-bound RNE (1-592) and cytoplasmic RNE (1-529), both lacking the CTD. These truncated variants diffused faster than their full-length counterparts due to reduced mass, but their D values still differed by a factor of ∼5.3 due to localization (Fig. 3B). Together, these results suggest that the membrane binding reduces RNE mobility by a factor of 5.

Diffusion of RNE in the absence of mRNA substrates

When some of RNE molecules interact with mRNA, their diffusion can slow due to the added mass of mRNA and ribosomes, possibly yielding slower RNE subpopulations. Such mobility-based subpopulations have been observed for RNA polymerases and ribosomes in E. coli and used to estimate the fraction of molecules interacting with RNA^35–37.

To test the effect of mRNA substrates on RNE diffusion, we treated cells with rifampicin (rif), which blocks transcription initiation, thus depleting cellular mRNAs³⁸. In rif-treated cells, D_RNE increased to 0.0270 ± 0.0003 µm²/s, which is 1.47 ± 0.010 times that in untreated cells (Fig. 3C). We note that D_LacY also increased to 1.25 ± 0.01 times relative to untreated cells (Fig. 3D). Although this increase is relatively small, the increase in D_LacY was unexpected because LacY is not an RNA-binding protein. The increase in D_LacY likely results from the depletion of mRNAs near the membrane (e.g., the mRNAs undergoing transertion), which could otherwise hinder the diffusion of membrane proteins^39,40. The similar fold change in D_RNE and D_LacY upon rif treatment suggests that the change in RNE diffusion may largely be attributed to physical changes in the intracellular environment (such as reduced viscosity or macromolecular crowding^41,42), rather than a loss of RNA-RNE interactions.

Because the rif-induced change in D_RNE is largely physical, we next examined why eliminating RNA-RNE interactions does not further increase RNE mobility, using the ribosome as a benchmark. The diffusion of ribosomal protein L1 became 7 times as fast as that in rif-treated cells (Fig. 3E), consistent with previous reports^35,36. This big change can be explained by the fact that ribosomes form polysomes, where the effective mass of L1 protein in untreated cells would be 2 or more times that in rif-treated cells (where it remains as a free subunit). In the case of RNE, it forms the RNA degradosome complex, whose mass can be from 450 kDa⁶ to 2.3 MDa depending on the occupancy of the RNA degradosome component proteins (RhlB, PNPase, enolase)^2,43–46 (Supplementary Discussion). Even at its largest size, the RNE complex is smaller than a 70S ribosome (∼2.5 MDa⁴⁷). This means that if RNE interacts with an mRNA associated with n ribosomes, the total mass of the RNE complex would increase by a factor of n or more. Thus, a substantial increase in D_RNE would be expected upon mRNA depletion, assuming that a significant fraction of RNE is engaged with mRNA-ribosome assemblies.

To explain the marginal increase in D_RNE upon rif treatment, we considered two possibilities; (1) only a small percentage of RNE molecules interacts with mRNAs at a given time and/or (2) RNE interacts with mRNAs only briefly, unlike ribosomes which spend an order of 10-100 s in a polysome state during translation elongation⁴⁸. To distinguish between these possibilities, we attempted to increase the cellular pool of polysomes, either by treating cells with a translation elongation inhibitor chloramphenicol³⁷ or by overexpressing lacZ mRNA from a high-copy plasmid (Fig. S3D-E). In both cases, a larger fraction of RNE would engage with mRNAs, potentially increasing the fraction of RNE in the slow-diffusing state. However, D_RNE remained unchanged compared to untreated cells (Fig. S3F). This result rules out the possibility that only a small percentage of RNE interacts with mRNAs and instead weighs in favor of the scenario that RNE-mRNA interactions are brief. Specifically, if RNE interacts with mRNAs for ∼20 ms or less, the slow-diffusing state would last shorter than the frame interval and remain undetected in our experiment.

Diffusion and localization of the MTS and TM segments

Unlike RNE in E. coli, RNase Y, a functional homolog of RNE in B. subtilis, is localized to the membrane via a TM domain²⁰. We wondered if there are differences between a peripheral motif (like the MTS of E. coli’s RNE) and a TM motif in terms of membrane localization and mobility. To address this question, we created RNE mutants in which the MTS was replaced with a TM domain. For the TM domain, we used TM segments of LacY, a native E. coli protein, rather than using the TM motif from B. subtilis RNase Y, which might interact with the E. coli membrane in a non-native manner.

Before creating the RNE mutants, we characterized the MB% and the diffusion of individual short membrane-binding motifs. Native LacY contains 12 TM segments, arranged into two groups of six²⁸. We successfully expressed constructs containing the first two TM segments (LacY (1-74) or LacY2), the first six TM segments (LacY (1-193) or LacY6), and the full-length LacY (LacY12), each fused to mEos3.2 and expressed from a chromosomal IPTG-inducible promoter (Fig. 4A). We then imaged their membrane localization and diffusion. Both the MTS segment and LacY-derived TM segments showed a strong membrane enrichment (Fig. 4B). In terms of diffusion, LacY2 and LacY6 diffused faster than the MTS segment (Fig. 4C), contrary to expectations based on size (or mass)-dependent diffusion (Fig. 4D).

Figure 4:

According to the Stokes-Einstein relation for diffusion in simple fluids⁴⁹ and the Saffman-Delbruck diffusion model for membrane proteins⁵⁰, D decreases as particle size increases, albeit with different scaling behaviors. Specifically, the Saffman-Delbruck model predicts that D for membrane proteins decreases logarithmically with increasing radius of the membrane-embedded region, assuming a constant membrane environment⁵⁰. Thus, if size (or mass) were the primary determinant of diffusion, LacY2 and LacY6 would diffuse more slowly than the smaller MTS. The observed discrepancy instead implies that D may be governed by how each motif interacts with the membrane. For example, the way that TM domains are anchored to the membrane may facilitate faster lateral diffusion with surrounding lipids.

Despite the prevalence of peripheral membrane proteins⁵¹, how they interact with the membrane and how this differs from TM proteins remain poorly understood. To further explore this, we conducted all-atom molecular dynamics (MD) simulations of the MTS and the LacY variants interacting with the E. coli membrane using the NAMD software⁵². In the simulations, protein motion was calculated for 1 µs. Although the absolute D values were higher than experimental values (possibly due to the absence of mEos3.2 in the model), the overall trends were preserved; among the LacY series, larger constructs diffused more slowly (Fig. 4E-F, Fig. S4A). Most importantly, the MTS again diffused more slowly than LacY2 in silico (Fig. 4E, Fig. S4A). By calculating membrane-protein interaction energies, we found that the MTS-membrane interactions were more stable than those of LacY2 (Fig. S4B). These results suggest that the slower diffusion of the MTS is due to stronger interactions with lipid head groups compared to membrane-embedded TM segments.

RNE mutants carrying a TM motif

Since LacY2 and LacY6 showed strong membrane enrichment similar to LacY12 (Fig. 4B), we replaced the MTS in RNE with LacY2, LacY6, and LacY12 in the presence or absence of the CTD (Fig. 5A-B and Fig. S5). All chimeric RNE mutants were expressed from the native chromosomal locus as the only copy of rne, with mEos3.2 fused at the C terminus for imaging. The resulting strains exhibited no noticeable differences in growth rate compared to the WT strain (Table S3), suggesting that the RNE mutants were functionally active.

Figure 5:

xNorm histograms of the ΔCTD mutants indicated membrane localization similar to LacY (Fig. 5C). However, mutants containing the CTD showed noticeable cytoplasmic subpopulations when LacY2 and LacY6 were used in place of the MTS (Fig. 5D, Fig. S2J-K, Fig. S6B). Mathematical model fitting of the xNorm histograms estimated the MB% of RNE-LacY2-CTD and RNE-LacY6-CTD to be 69% [66%, 73%] and 86% [84%, 90%], respectively (Fig. 5F). We note that imperfect membrane localization was observed only in mutants containing the CTD; the same protein without the CTD showed MB% of 100% (Fig. 5E-F, Fig. S6A-B). These findings suggest that the CTD may contribute to unstable membrane binding of RNE. Supporting this idea, previously characterized RNE MTS point mutants with MB% <50% (Fig. 2D-E) also exhibited increased MB% upon the CTD removal (Fig. S6A-B). Such a difference in MB% between the CTD-containing and the CTD-lacking mutants was not observed in the chimera based on LacY12 (Fig. 5E-F), possibly due to the stable membrane insertion by LacY12.

Next, we examined whether the D of chimeric RNE mutants varied depending on the type of membrane-binding motifs. For example, the MTS diffused more slowly than LacY2 and LacY6, despite being smaller in size (Fig. 4C-D). Based on this, we expected the chimeric RNE with LacY2 or LacY6 to diffuse faster than RNE with MTS. Indeed, in the absence of the CTD, we found that the D of LacY2-based RNE was 1.33 ± 0.01 times as fast as the MTS-based RNE (Fig. 5G). However, LacY6-based RNE did not diffuse faster than the MTS-based version (Fig. 5G). This result may be due to the high TM load (24 helices) created by four LacY6 anchors in the RNE tetramer. Although all constructs are tetrameric, the 24-helix load (LacY6), compared with 8 (LacY2) and 4 (MTS), likely enlarges the membrane-embedded footprint and increases drag, thereby changing the mobility advantages assessed as standalone membrane anchors.

In the presence of the CTD, the D of LacY2 and LacY6-based RNE became 3.33 ± 0.004 and 1.14 ± 0.01 times that of the MTS-based RNE counterpart (i.e. the WT RNE), respectively (Fig. 5H). This is likely influenced by the presence of a cytoplasmic population (∼31% for LacY2 and ∼14% for LacY6; Fig. 5F, Fig S2J-K), which diffuses more rapidly than membrane-bound molecules (possibly by a factor of five, based on Fig. 3B). Taken together, our data suggest that the CTD weakens the membrane association of RNE and small TM motifs can facilitate the diffusion of RNE.

Functional consequence of subcellular localization and diffusion of RNE

To check the functional consequence of cytoplasmic localization of RNE, we measured lacZ mRNA degradation in various RNE mutants presented in this study. Recently, we developed an assay to quantify both co-transcriptional and post-transcriptional degradation rates of lacZ mRNA by inducing its transcription for only 75 s, thereby capturing the degradation of nascent mRNA¹⁵. In WT cells, we found that the co-transcriptional degradation rate (k_d1) is about 10 times slower than the post-transcriptional degradation rate (k_d2)¹⁵. In cells expressing RNE ΔMTS, however, k_d1 increases by a factor of ∼3, suggesting that cytoplasmic RNE can freely diffuse and degrade nascent mRNAs¹⁵. This result led us to hypothesize that RNE variants exhibiting a cytoplasmic subpopulation (Fig. 2E and 5F) may exhibit a larger k_d1 compared to more membrane-bound variants.

We repeated this assay in a strain expressing the WT RNE fused to mEos3.2 (Fig. S7A). Note that this strain is different from the one we used for imaging because a monocistronic lacZ gene is needed for the transient induction assay¹⁵. The relative abundances of 5’ lacZ mRNA (Z5) remained constant prior to the rise in 3’ lacZ mRNA (Z3) levels (between ∼100 and 210 s), confirming negligible co-transcriptional degradation in WT cells¹⁵ (Fig. 6A). However, in cells expressing RNE-LacY2-CTD (MB% = 69%), Z5 levels exhibited a downward trend during the same time window (Fig. 6B). The estimated k_d1 was close to what was observed in RNE ΔMTS¹⁵ (gray line, p = 0.28; Fig. 6C). A similarly high k_d1 was also observed in RNE MTS point mutations, F582E and F757E, which exhibited a ΔMTS-like xNorm profile (Fig. 2D), supporting the idea that the cytoplasmic subpopulation of RNE enables co-transcriptional mRNA degradation in E. coli (Fig. 6C). Notably, RNE-LacY6-CTD, which also exhibited a cytoplasmic subpopulation (∼14% from MB% of 86%), did not exhibit a significant increase in k_d1 (Fig. 6C), suggesting a critical amount of cytoplasmic population may be needed to facilitate co-transcriptional mRNA degradation

Figure 6:

We next examined whether the post-transcriptional mRNA degradation rate (k_d2) is limited by the slow diffusion of membrane-bound RNE. In the presence of the CTD, k_d2 did not significantly vary across membrane-targeting variants (Fig. 6D). For example, even though the LacY12 motif slows RNE diffusion, its k_{d 2} was similar to that of WT RNE (Fig. 6D). A critical control for this comparison is RNE abundance: because RNE autoregulates its expression by degrading its own transcript⁵³, slower diffusion could elevate RNE levels and mask the negative effect of the reduced mobility. We tested this explicitly and found that over-expression of WT RNE does not significantly change k_d1 or k_d2 (Fig. S7B). Thus, when the CTD is present, neither copy number nor diffusion of membrane-bound RNE limits k_d2.

In the absence of the CTD, k_d1 was high in RNE variant F582E (MB% = 67%; Fig. S6A), reaching levels close to its cytoplasmic counterpart, the ΔMTS ΔCTD mutant (gray line, p = 0.7; Fig. 6E). k_d2 values for ΔCTD variants were, in general, lower than those of their CTD-containing counterparts (Fig. 6F), indicating the importance of the CTD for the catalytic activity. These findings are consistent with previous reports on RNE constructs based on the MTS^15,22–24. Among ΔCTD variants with different membrane motifs, point mutants F575E and F582E (MB% = 91% and 67%, respectively; Fig. S6A) exhibited higher k_d2 than the MTS-based variant, in agreement with the fact that completely cytoplasmic ΔMTS ΔCTD exhibited higher k_d2 than the MTS-containing ΔCTD (Fig. 6F). However, the LacY2-based RNE variant, which diffuses faster than the MTS version (Fig. 5G), did not show a corresponding increase in k_d2 (Fig. 6F). Plus, LacY6 and LacY12 versions showed even lower k_d2 than MTS-based RNE ΔCTD. Overall, these results indicate that the reduced k_d2 caused by ΔCTD cannot be rescued by faster diffusion of membrane-bound RNE. In fact, it may be further impaired by large and slow membrane motifs (such as LacY12). The presence of the CTD appears to buffer the effects of large membrane-binding motifs on RNE’s catalytic activity, helping to maintain efficient post-transcriptional mRNA degradation (k_d2).

Discussion

Our study establishes the membrane enrichment and slow diffusion of RNE in E. coli. This supports the notion that sequestration of RNE on the membrane confers the spatial and temporal separation between synthesis and decay of mRNAs¹⁵. Furthermore, the processing of rRNA^54–58 and tRNA^59,60 and small RNA-based gene regulations^61,62 mediated by RNE likely take place on the membrane.

For WT RNE, our analysis showed MB% of 93%, close to 91% previously reported using immunogold labeling and freeze-fracture electron microscopy⁶³. Whether the MB% of RNE changes under different growth conditions remains to be tested. As a case study, we examined the MB% of WT RNE when cells were grown in M9 minimal medium with succinate as the sole carbon source without supplements, a condition different from that used in our primary experiments (M9 glycerol with supplements). The MTS segment exhibited a reduction in MB% from 100% to 83% (Fig. S6C), suggesting that MTS-mediated membrane binding can be sensitive to growth conditions. However, the MB% of WT RNE was only marginally affected by the same media change (Fig. S6C). We speculate that tetramer formation stabilizes RNE membrane localization, even when individual MTS motifs exhibit a weak membrane affinity. A similar phenomenon has been observed with MinD; although its amphipathic motif alone is cytoplasmic, MinD becomes membrane-associated upon dimerization, indicating that oligomerization enhances the membrane binding of this peripheral membrane protein⁶⁴. By analogy, the tetrameric structure of RNE may reinforce membrane association, buffering against changes in cellular metabolism or lipid environment that would otherwise weaken MTS-mediated binding.

While individual TM motifs exhibited strong membrane association regardless of growth conditions (Fig. S6D), we found LacY2-based RNE can lose membrane binding affinity, unlike MTS-based RNE. The loss of MB% in LacY2-based RNE was observed only in the presence of the CTD (Fig. S6D), suggesting that the CTD negatively affects membrane binding of RNE, possibly by altering protein conformation. In fact, all ΔCTD RNE mutants we tested exhibited higher MB% than their CTD-containing counterparts (Fig. S6A-B). For example, WT RNE (containing CTD) showed an MB% of 93%, whereas its ΔCTD version, RNE (1-592), showed 100%. Similarly, the chimeric RNE-LacY6-CTD showed an MB% of 86%, while its ΔCTD version showed 100% (Fig. 5E-F, Fig. S6A-B). A similar trend was observed for MTS point mutants (Fig. S6A-B), further supporting that the CTD decreases membrane association across RNE variants. We speculate that this effect may be related to the CTD’s role in promoting phase-separated ribonucleoprotein condensates, as observed in Caulobacter crescentus¹⁹. In E. coli, we also observed a modest increase in the clustering tendency of RNE compared to ΔCTD (Fig. S8).

The D_RNE we measured in E. coli (0.018 μm²/s) is comparable to those measured in other bacterial species. For example, in C. crescentus, the D of its cytoplasmic RNE was shown to be about 0.03 μm²/s⁶⁵. The diffusion of RNase Y in B. subtilis was found in either a slow (0.031 μm²/s) or fast (0.3 μm²/s) population⁶⁶, with the slow population corresponding to RNase Y bound to mRNA and/or the putative RNA degradosome and the fast population representing freely diffusing RNase Y. In the case of E. coli RNE, a two-population fit of the D histogram based on MB% identified the slow and fast subpopulations whose D values differed by a factor of 4 (Fig. S3C). This agrees with our finding that cytoplasmic RNE ΔMTS diffuses ∼5 times as fast as WT RNE on average (Fig. 3B).

Diffusion is affected by the size of the particle in a given medium⁵⁰, and it has been used to identify different size forms of a protein due to biochemical interactions or complex formation^36,67. Related, RNA substrates interacting with RNE can also increase the effective mass of RNE and lower its D value. However, when cellular mRNAs were depleted by rifampicin treatment, RNE diffusion increased less than that of other RNA-binding proteins related to transcription and translation. For example, the large and small ribosomal subunits^35,36,68, tRNA⁶⁹, and RNA polymerase³⁷ showed a large (about 10-20 fold) increase in D upon rifampicin treatment. Interestingly, Hfq, an RNA chaperon involved in small RNA regulation together with RNE, showed a moderate increase (∼2 fold) in D when RNA was depleted by rifampicin⁷⁰. We note that our result is consistent with previous studies that examined the effect of rifampicin on RNE diffusion in E. coli^11,71 as well as RNE in C. crescentus⁶⁵ and RNase Y in B. subtilis^66,72. One possible explanation is that RNA-bound RNE (and RNase Y) is short-lived compared to our frame interval (∼20 ms), unlike other RNA-binding proteins related to transcription and translation, interacting with RNA for ∼1 min for elongation⁴⁸.

Lastly, the slow diffusion of the MTS in comparison to LacY2 and LacY6 suggests that MTS is less favorable for rapid lateral motion in the membrane. Our MD simulations demonstrated that this reduced diffusivity may originate from a stronger interaction energy between the protein and the membrane. Whether this property is specific to the MTS or generalizable to other peripheral versus integral membrane motifs remains to be tested. We speculate that this can be a general phenomenon, as peripheral motifs interact with lipids orthogonally while integral membrane motifs align parallel to the bilayer and may diffuse more freely by coupling with the motion of surrounding lipids. The diffusion behavior of membrane-bound proteins reflects underlying protein-lipid interactions and membrane dynamics⁷³. Accordingly, future work may analyze how protein-lipid interaction strength and membrane dynamics contribute to differences in lateral diffusion between peripheral and integral membrane proteins.

Altogether, our work highlights strategies to modulate the MB% and diffusion of RNE to possibly affect mRNA degradation rates. For example, membrane-bound RNE lacking the CTD can stabilize mRNAs and increase protein expression. This idea has been used in a commercial E. coli BL21 strain (Invitrogen’s One Shot BL21 Star) to increase recombinant protein yield²². Additionally, RNE variants with environmentally responsive MB%, such as RNE-LacY2-CTD, could be harnessed to tune mRNA half-lives and protein expression levels under different growth conditions.

Acknowledgements

We thank Drs. Mark Arbing, Agamemnon Carpousis, Johan Elf, and Christine Jacobs-Wagner for strains. We thank Maggie Liu, Kavya Vaidya, and Zach Wang for their contributions in the early phase of this work and the members of Kim lab for critical reading of the manuscript. This work was supported by the NSF Center for Physics of Living Cells (1430124), NSF Science and Technology Center for Quantitative Cell Biology (2243257), NIH (R35GM143203; R24GM145965), and Searle Scholars Program.

Additional information

Author contributions

L.T. and S.K. designed the research; L.T. and Y.W. performed imaging and analysis; L.T. and Se.K. measured mRNA lifetimes; LT, Se.K., and S.K. performed genetics; L.T., Y.W., and J.W. developed data analysis methods; S. performed all-atom MD simulations under supervision of E.T.; B.R. performed epi-fluorescence imaging; L.T. and S.K. wrote the manuscript with input from all authors; S.K. supervised the study.

Funding

NIH (R35GM143203)

NIH (R24GM145965)

NSF (2243257)

NSF (1430124)

Additional files

Supplemental Information