Subcellular Localization Defects Characterize Ribose-Binding Mutant Proteins with New Ligand Properties in Escherichia coli

ABSTRACT

Periplasmic binding proteins have been previously proclaimed as a general scaffold to design sensor proteins with new recognition specificities for nonnatural compounds. Such proteins can be integrated in bacterial bioreporter chassis with hybrid chemoreceptors to produce a concentration-dependent signal after ligand binding to the sensor cell. However, computationally designed new ligand-binding properties ignore the more general properties of periplasmic binding proteins, such as their periplasmic translocation, dynamic transition of open and closed forms, and interactions with membrane receptors. In order to better understand the roles of such general properties in periplasmic signaling behavior, we studied the subcellular localization of ribose-binding protein (RbsB) in Escherichia coli in comparison to a recently evolved set of mutants designed to bind 1,3-cyclohexanediol. As proxies for localization, we calibrated and deployed C-terminal end mCherry fluorescent protein fusions. Whereas RbsB-mCherry coherently localized to the periplasmic space and accumulated in (periplasmic) polar regions depending on chemoreceptor availability, mutant RbsB-mCherry expression resulted in high fluorescence cell-to-cell variability. This resulted in higher proportions of cells devoid of clear polar foci and of cells with multiple fluorescent foci elsewhere, suggesting poorer translocation, periplasmic autoaggregation, and mislocalization. Analysis of RbsB mutants and mutant libraries at different stages of directed evolution suggested overall improvement to more RbsB-wild-type-like characteristics, which was corroborated by structure predictions. Our results show that defects in periplasmic localization of mutant RbsB proteins partly explain their poor sensing performance. Future efforts should be directed to predicting or selecting secondary mutations outside computationally designed binding pockets, taking folding, translocation, and receptor interactions into account.

IMPORTANCE Biosensor engineering relies on transcription factors or signaling proteins to provide the actual sensory functions for the target chemicals. Since for many compounds there are no natural sensory proteins, there is a general interest in methods that could unlock routes to obtaining new ligand-binding properties. Bacterial periplasmic binding proteins (PBPs) form an interesting family of proteins to explore for this purpose, because there is a large natural variety suggesting evolutionary trajectories to bind new ligands. PBPs are conserved and amenable to accurate computational binding pocket predictions. However, studying ribose-binding protein in Escherichia coli, we discovered that designed variants have defects in their proper localization in the cell, which can impair appropriate sensor signaling. This indicates that functional sensing capacity of PBPs cannot be obtained solely through computational design of the ligand-binding pocket but must take other properties of the protein into account, which are currently very difficult to predict.

INTRODUCTION

Periplasmic binding proteins (PBPs) or solute-binding proteins are part of a large family of proteins with a typical and conserved bilobal structure (1–3). PBPs are involved in nutrient and mineral scavenging for bacterial cells, binding specific ligands (e.g., ribose, l-amino acids, or maltose), and presenting the bound molecules to specific transport channels and/or membrane receptors involved in chemotactic movement (3, 4). Their unique ligand-binding properties and sensitivity make PBPs an attractive class of proteins to use for biosensing (5). PBPs can be purified and either their ligand-binding capacity be exploited using in vitro measurements (e.g., isothermal titration calorimetry) (5–7) or the PBP ligand-binding activity can be integrated in an in vivo hybrid signaling chain, leading to fluorescent protein expression (8, 9). In that configuration, the presence of a ligand to the cell leads to stabilization of the PBP closed form, which binds to its membrane receptor and triggers the designed signaling chain. A synthetic signaling chain that has been perused for this purpose consisted of RbsB (the ribose-binding protein of Escherichia coli) and its hybrid receptor Trz1 (9, 10), formed by fusion between the C-terminal part of the E. coli EnvZ osmoregulation histidine kinase and the N-terminal part of the Trg chemotaxis receptor (8). Addition of ribose to cultures carrying the signaling chain starts a phosphorylation cascade that leads to reporter gene expression (9), which can be easily measured in a variety of ways (11). The native RbsB-Trz1 E. coli whole-cell biosensor has excellent selectivity and sensitivity for ribose, with a measured lowest detection limit of 50 nM in a 2-h assay (9).

With the aim to expand detection specificity of whole-cell biosensors, PBPs had been proposed as a scaffold to computationally design novel proteins with new binding capacities (10, 12). This conceptual design would permit utilization of a single constant hybrid chemoreceptor (such as Trz1) and reporter output while varying the PBP specificity (10, 12). Despite initial claims (10), however, designed PBPs with altered ligand-binding properties showed limited performance and poorly reproducible outcomes (6, 9, 13–15). So far, there has been only one report of new PBP variants for non-natural-compound recognition based on redesign of the RbsB ligand-binding pocket (16). In contrast, PBP ligand-binding reconstruction through grafting of existing binding pockets has been more successful (17, 18), and in vitro mutagenesis has enabled investigators to improve binding sensitivity of existing PBPs (17, 19, 20). One of the reasons for the relatively poor success in PBP engineering is the perceived poor stability and misfolding of the computationally designed proteins (15, 16). Since ligand prediction algorithms, mutagenesis, and screening methods have mostly focused on the ligand-binding properties of the PBP, other key steps in their signal transduction chain have been largely ignored. For a protein like RbsB in E. coli, this involves (i) proper translocation by SecB into the periplasm through the Sec pathway, (ii) proper folding in the periplasm, enabling ribose binding, and (iii) docking of the RbsB-bound ribose to the transport channels (RbsAC) and, partly, to the Trg chemoreceptor (3, 21–23). Ribose transport in E. coli involves the cytoplasmic membrane transporter RbsC and the cytosolic ATPase RbsA to provide energy. Ribose is presented to the channel in its bound form by RbsB and released into the cytoplasm by the ribokinase RsbK, with subsequent phosphorylation (23). The Trg membrane receptor controls chemotaxis of cells toward ribose upon binding by ribose-bound RbsB (24). This would indicate that for a successful performance of (designed) PBP variants in whole-cell biosensors, their capacity to engender proper interactions with the various signaling components must be taken into account.

Since this is relatively poorly understood, our primary goal in this study was to design an in vivo system to quantify the subcellular localization of RbsB in relation to its signaling chain in E. coli and test whether computationally designed RbsB-mutants with new ligand-binding properties show obvious localization defects that can explain their poorer signaling performance. We followed periplasmic localization of RbsB by using fluorescent protein fusions as proxies, imaged at individual cell level by quantitative epifluorescence microscopy. First, we examined which of different translocation signal peptides and configurations of mCherry gene fusions to rbsB would be best functional in E. coli. We imaged and quantified subcellular localization of wild-type RbsB-mCherry fluorescence in different E. coli backgrounds devoid of assumed components of its signaling chain (i.e., trg, rbsB, and rbsK), or with overexpressed hybrid Trz1 receptor. We then took advantage of a series of previously designed and selected RbsB variants for 1,3-cyclohexanediol binding and produced similar mCherry fusions to study subcellular localization. The initially computational designed RbsB variants (named DT002 and DT016) bind 1,3-cyclohexanediol and no longer recognize ribose, but they suffer from poor stability and or misfolding (16). Improved variants with up to 4.5-fold more inducibility by 1,3-cyclohexanediol had been selected from two (DT020, DT021, and DT021) or three (DT038) rounds of directed evolution (25), which we expected would show more wild-type RbsB behavior in terms of subcellular localization and expression. Finally, we examined the error-prone PCR mutant libraries after the second and third rounds using the same mCherry protein fusion technique, in order to understand whether mutants with more “wild-type”-like localization behavior could be detected. Our results show the main subcellular localization hot spots for RbsB-mCherry and demonstrate a variety of clear and more subtle defects of designed and selected mutants in signal chain localization within the context of similar mCherry-protein fusion constructs as proxies. We found evidence, supported by protein structure predictions, for improvement of PBP mutants to more wild-type behavior that may have contributed to better inducibility by the new ligand.

RESULTS

Fluorescent RbsB fusion protein expression and translocation.

In order to be able to quantify wild-type and mutant RbsB expression in single cells and distinguish differences in localization in the RbsB-mediated signaling cascade, we deployed mCherry fluorescence protein fusions. We first tested the functionality of C- or N-terminally fused mCherry and their effects on expression levels, driven from two different promoters in E. coli and in the presence or absence of two different translocation signal sequences (i.e., native wild-type RbsB translocation signal sequence [RbsBss] or the general TorA translocation signal sequence [TorAss]). RbsB/mCherry fusion constructs introduced on plasmids and expressed under the control of the arabinose-inducible P_BAD promoter grew normally in the absence of l-arabinose in E. coli DH5α. This held true for constructs that deployed either the RbsBss or the TorAss. Cells carrying the TorAss-RbsB/mCherry constructs induced with l-arabinose were significantly longer (8.46 ± 4.8 μm) than those expressing RbsBss-RbsB/mCherry (3.50 ± 0.7 μm; P = 0.000014, two-sided t test with 18 replicates, 10³ cells per replicate [Fig. 1A to D]). In case of the TorAss-RbsB C-terminally fused mCherry construct, induction with l-arabinose further yielded cells with swollen ends and visible cytoplasmic fluorescence, but without sign of periplasmic localization (Fig. 1A). In contrast, cells expressing the TorAss-mCherry-RbsB fusion protein yielded visible periplasmic fluorescence and, in some cases, showed visible fluorescent spots near the cell poles (Fig. 1B). Much clearer periplasmic fluorescence and polar fluorescent spots were obtained for the RbsBss-RbsB-mCherry fusion (Fig. 1C), whereas the RbsBss-mCherry-RbsB variant again led to very poorly visible periplasmic and polar fluorescence (Fig. 1D). In cultures without l-arabinose induction, no fluorescent signal from RbsBss-RbsB-mCherry was observed (Fig. 1E). Expression of RbsB-mCherry without the RbsBss translocation signal resulted in a homogenous distribution of fluorescence in the cytoplasm and no signs of periplasmic fluorescent spots (Fig. 1F). Other fusions expressed from the constitutive P_AA promoter with the native rbsB signal sequence (RbsBss [see Fig. S1 in the supplemental material]) as previously deployed (9) resulted in slow-growing E. coli colonies and frequently appearing mutants, irrespective of the position of the mCherry tag. This indicated that only in the presence of the proper signal sequence is RbsB-mCherry able to translocate to the periplasm and preferentially accumulate at the cell poles. Collectively, these results indicated that mCherry fused to the C-terminal end of RbsB and carrying the native RbsBss signal sequence was the optimal configuration for its expression and translocation into the periplasm.

FIG 1.

Effect of fluorescent protein fusion positioning and signal sequences on RbsB expression and translocation in E. coli DH5α. Panels show expression of RbsB fused with N-terminal TorA signal sequence and C-terminal mCherry (A), N-terminal TorAss and mCherry (B), RbsB fused with N-terminal RbsB signal sequence and C-terminal mCherry (C), N-terminal RbsBss and mCherry (D), RbsB fused with N-terminal RbsBss and C-terminal mCherry but without l-arabinose induction (E), and without RbsB signal sequence but with C-terminal mCherry in the presence of l-arabinose (F). Fusion constructs are schematically drawn on top of each micrograph (not to scale) and were all expressed under the control of the arabinose-inducible P_BAD promoter. Cells were incubated for 3 h and expression of fusion protein was induced with 0.5% l-arabinose (except in panel E). Hooked arrows indicate promoters and transcription direction. PhC, phase-contrast. mCHE, mCherry fluorescence. All fluorescence images (mCHE) are scaled to the same intensity (low-high).

Functionality of RbsB-mCherry.

Next we studied the functionality of the RbsB-mCherry fusion protein to respond to its native ligand ribose in E. coli by measuring induction of green fluorescent protein (GFP) formation from the Trz1-ompR-ompC′::gfp hybrid signaling chain (9). For this purpose, the different plasmids carrying mCherry C- or N-terminally fused to RbsB, with either the native RbsBss or the TorAss translocation signal sequence, were transformed into E. coli ΔrbsB cells, further containing a compatible plasmid (pSYK1) with the ribose-inducible Trz1-ompC′::gfp bioreporter system (9) (Fig. 2A). Cultures were amended with l-arabinose to induce expression of the RbsB-mCherry fusion protein and then incubated or not with 0.1 mM ribose for 2.5 h to stimulate GFP expression from the ompC promoter. For the RbsB-mCherry fusion protein translocated from the native RbsB signal sequence, but not for any of the three other configurations, this resulted in a 1.7-fold induction of GFP fluorescence compared to a noninduced control (P < 0.001, analysis of variance [ANOVA] followed by post hoc Tukey test [Fig. 2B]). Addition of ribose did not cause any change in RbsB-mCherry fluorescence (Fig. 2C). This 1.7-fold induction was lower than the 11-fold induction of wild-type RbsB in the same E. coli background (although expressed from the P_AA promoter [Fig. 2B]), but it showed that the periplasmic RbsB-mCherry protein retained part of its functionality. The absence of GFP induction by any of the other three mCherry fusion configurations (Fig. 2B) indicated that these proteins were nonfunctional or insufficiently present in the periplasm. This confirmed that C-terminally fused RbsB-mCherry is a partly functional protein and thus the best proxy to follow RbsB localization and signaling.

FIG 2.

Ribose-dependent induction of the Trz1-ompC′::gfp signaling chain in E. coli ΔrbsB by different plasmid-expressed RbsB/mCherry fusion constructs. (A) Principle of ribose-dependent RbsB-Trz1 induction of gfp from the ompC promoter. p, periplasmic space; c, cytoplasm. Trz1 is a hybrid receptor with Trg′ periplasmic and ′EnvZ cytoplasmic parts. (B) Fold induction of GFP fluorescence in the presence of 0.1 mM ribose for 2.5 h compared to the case with no ribose. Bars show the mean ratio from 12 biological replicates (black dots). Letters indicate significance groups in ANOVA followed by post hoc Tukey test (P_a,b < 0.001). As a comparison, induction by RbsB without mCherry tag but expressed from the constitutive P_AA promoter is shown. (C) Effect of ribose addition on mCherry signal intensities for the RbsBss-RbsB-mCherry fusion construct. Bars indicate the mean of population fluorescence means in flow cytometry across 12 replicates (black dots) in the absence (light gray) or presence (dark gray) of ribose. P value was derived from two-sided t test (n = 12). ns, not significant (P > 0.05).

Wild-type RbsB-mCherry protein localizes in the periplasm and to the cell poles.

In order to understand the dependency of the RbsB-mCherry localization on known factors in its signaling chain, we compared fluorescence expression and localization in a number of E. coli backgrounds (Fig. 3A). Notably, we tested the effect of coexpression of the Trz1 hybrid (Trg′-′EnvZ) chemoreceptor from the pSYK1 plasmid and the effects of deletion of the native chromosomal rbsB gene in the MG1655 background, of the ribose-transport-involved rbsK gene (26, 27), and of the native trg chemoreceptor gene (28–30). RsbB-mCherry in all host backgrounds mostly localized to the cell pole regions, with new poles shortly before and after cell division being devoid of but then quickly regaining mCherry fluorescence (Fig. 3A, arrowheads). A faint fluorescence trace covering the cell envelope could be detected in most cells (Fig. 3B), indicating periplasmic presence of RbsB-mCherry, which was dependent on l-arabinose induction (Fig. 3B). Coexpression of Trz1 both in the E. coli DH5α background and in E. coli BW25133 Δtrg increased polar fluorescence levels (P = 0.0352 for mean top 5% pixel fluorescence levels [n = 6 biological replicates] [Fig. 3B and C]); inactivation of rbsB and rbsK also resulted in significantly higher polar fluorescence levels (Fig. 3C). In contrast, deletion of trg itself did not decrease polar fluorescence of RbsB-mCherry (Fig. 3C). Most cells in all hosts displayed one polar RbsB-mCherry fluorescent spot, and 70 to 100% also displayed fluorescence at the opposite pole (Fig. 3D). Very few cells showed additional fluorescent foci outside polar regions (Fig. 3E). In summary, these data indicated that mutations and alterations in the signaling pathway components did change the fluorescence intensity but not so much the positioning of RbsB-mCherry in individual cells. Differences in observed RbsB-mCherry spot intensities may be due to changed equilibria of potential available binding sites of chemoreceptors (e.g., Trg and Trz1), competitor protein (e.g., nonlabeled RbsB), and ribose transport (i.e., RbsK effects, even though all the experiments were conducted in the absence of added ribose).

FIG 3.

RbsB-mCherry localization in different E. coli backgrounds. (A) Representative cell micrographs in mCherry fluorescence and phase-contrast. Fluorescence images are autocontrasted and cropped. (B) Detail of periplasmic space and polar region fluorescence of RbsB-mCherry in E. coli Δtrg background. Shown are a z-projection of the mean top 20 percentiles per pixel across all cells in a biological replicate (n = 103) standardized to the same cell length and width and a heat map representation of fluorescence intensity in the cells in presence or absence of arabinose (ARA) or coexpressed Trz1. Blue and orange colors represent low and high intensities, respectively. Scale is not comparable between heat maps. (C) Polar (cyan and blue) and mid-cell (green) fluorescence from RbsB-mCherry normalized for image background (bg nrmz) of individual cells in one technical replicate in different E. coli hosts, manually segmented (each dot a single cell). P values compare the extracted mean top 5% pixel values (representative for foci) from automated segmentation across 5 or 6 biological replicates (each with 10 images) of that host to MG1655 (unpaired one-sided t test). (D) Manually extracted proportions of cells in panel C with one or both fluorescent polar regions. (E) Proportions of cells with only polar foci and those with one or two additional foci outside the cell pole regions.

Altered localization of mutant RbsB-mCherry protein.

In previous work, we had isolated a number of RbsB mutants with gain of functional binding to 1,3-cyclohexanediol and loss of ribose as a ligand (16, 25). Although these proteins had de novo ligand-binding properties, biochemical data showed them to be impaired in stability and translocation (16). Continued selection by directed evolution approaches yielded variants with improved induction ratios but still not close to those for wild-type RbsB and ribose (25). In order to better understand if and where in the signaling pathway those RbsB mutants might be impaired, we constructed similar C-terminal end fusions as described above and compared mCherry fluorescence localization to that of wild-type RbsB-mCherry. We concentrated on six RbsB mutants, DT002 and DT016, from a protein design survey (16), and DT020, DT021, DT022, and DT038, obtained from further directed evolution screenings (25). Notably mutant DT038 had shown a 4.5-times improvement in inducibility with 1,3-cyclohexanediol compared to DT016 and DT002, with DT020, DT021, and DT022 showing 2-fold-higher induction (25).

l-Arabinose-induced individual cells of the various E. coli hosts expressing DT-mCherry protein variants visually still showed fluorescent spots in cell pole regions, but to a lower extent and with a variety of other effects, such as higher cell fluorescence background, deformed or lysed cells, multiple fluorescent foci at other places in the cell, and generally higher cell-cell variability (Fig. S3). Manually segmented cells, foci, and polar regions of the same host backgrounds indicated mCherry fluorescence generally lower than in the wild type (e.g., DT002-mCherry), generally higher (e.g., DT021- and DT038-mCherry), or lower and higher depending on the host (e.g., DT016- and DT020-mCherry) (Fig. S4). Across all DT-mCherry expressed variants, there was a tendency for mCherry fluorescence values to be higher in the Trz1 background than in the absence of trg: not only for cell pole regions but also in the cell background (Fig. 4A and B). To quantify this across a larger number of cells and biological replicates, we proxied cell pole fluorescence intensities in individual culture replicates by the mean of the top 5% fluorescent pixels per cell across all segmented cells in the images and cell background as the median pixel fluorescence (n = 6 to 10 technical replicates; 2 to 6 biological replicates [Fig. 4C]). As an example, the mean top 5% fluorescence of E. coli DH5α or E. coli BW25113 Δtrg expressing RbsB-mCherry increased 10-fold upon induction with l-arabinose (P = 0.00018, two-sided t test; n = 3 biological replicates, each 10³ cells [Fig. 4C]). Deleting the RbsBss signal sequence resulted in a strong decrease of mean top 5% values, as expected from the images themselves (P = 0.0248; n = 3 [Fig. 4C]).

FIG 4.

Aberrant localization behavior of DT mutant-mCherry fusions in E. coli hosts. (A and B) Manually segmented cells of one technical replicate showing background normalized polar fluorescence (cyan and blue) or mid-cell fluorescence (green) in E. coli Δtrg or Δtrg with coexpressed pSYK1 plasmid (Trz1 expression). Note the higher cell-cell variability and higher cell background fluorescence in many of the DT mutants in comparison to wild-type RbsB (WT). (C) Mean top 5% fluorescence pixel extraction as proxy for manual polar focus determination. Bars show replicate means (dots are individual replicates) for the indicated E. coli host and expressed wild-type RbsB-mCherry (n = approximately 1,000 cells per replicate). (D) Comparison of mean top 5% and cell background (i.e., median fluorescence) in E. coli MG1655 ΔrbsB or ΔrbsK of expressed wild-type RbsB- to DT016- and DT022-mCherry. Bars represent replicate means (dots are individual replicate values), each from 200 to 1,000 cells. P values are from unpaired one-sided t test comparisons to wild-type RbsB-mCherry in that host (red, top 5% comparisons; blue, cell background). (E) Proportions of cells with zero, one, or both polar fluorescent foci for the various expressed DT mutant- or RbsB-mCherry in different E. coli hosts, obtained from the manually segmented images. (F) Same as panel E, but for the proportions of cells with additional foci.

Similar to RbsB-mCherry, all DT-mCherry mutants showed higher mean top 5% cell fluorescence pixel values, but they also showed higher cell median fluorescence in the E. coli Δtrg plus pSYK1 background than in E. coli Δtrg (P = 1.97 × 10⁻⁴ and P = 0.0001, paired t test [Fig. S5A and B]). Mutants DT002 and DT016 showed less, whereas DT021, DT022 and DT038 showed higher top 5% fluorescent levels than wild-type RbsB-mCherry in a Δtrg background (Fig. S5A), whereas all except DT021 showed higher top 5% and background fluorescence when Trz1 was expressed from pSYK1 (Fig. S5B). Top 5% and cell median fluorescence levels of DT016 and DT022 were higher than for wild-type RbsB in the MG1655 background (Fig. 4D). Median fluorescence levels (cell backgrounds) were increased for DT016 and DT022 in the ΔrbsB background, whereas DT016 increased and DT022 decreased both top 5% and median fluorescence in E. coli ΔrbsK compared to RbsB (Fig. 4D). This suggested that DT mutants behave suboptimally compared to wild-type RbsB in a number of features that become visible as higher and more variable (polar) foci fluorescence and in a higher cell background. The rather dramatic reduction from highly variable cell backgrounds for DT016 in the ΔrbsK background, but much lower for DT022 (and similar to wild-type RbsB [Fig. S5C]), may be an indication for its improved wild-type-like behavior.

The aberrant behavior of DT mutant-mCherry fusions compared to that of the wild-type RbsB-mCherry fusion was even more apparent from the localization and numbers of fluorescent foci. Whereas RbsB-mCherry was characterized by fluorescence almost exclusively at both cell poles (Fig. 3D and E), the DT-mCherry fusions scored consistently lower with large proportions (10 to 60%) of cells without any foci and only 10 to 75% of cells having two foci (Fig. 4E). This difference was less for E. coli strains that are more proficient for gene cloning stability, such as DH5α, than for MG1655 derivatives (Fig. 4E). First-generation designed mutants (e.g., DT002 and DT016) consistently performed poorer (i.e., more cells without polar foci or with only one) than further-evolved variants (Fig. 4E and Fig. S6A and B). In addition, all mutants in all backgrounds showed significantly more fluorescent foci at other positions than the cell pole, in comparison to hosts expressing RbsB-mCherry (P = 1.01 × 10⁻⁴, nonpaired t test [Fig. 4F]).

Collectively, these results demonstrated that DT mutant-mCherry fusions tend to be more poorly localized to the cell poles and more often lead to cells with high fluorescent background and no visible foci and to more frequent appearance of cells with foci at other positions than at the poles. Within the context of mCherry-fusion proteins as proxies for RbsB/DT protein behavior, these are indications for defects of the DT mutant proteins in their periplasmic translocation and binding to the appropriate receptors. This may at least partly explain their poorer-than-wild-type behavior in inducing the Trz1-ompCp signaling chain.

Improvement of wild-type-like localization behavior in evolved mutant RbsB libraries.

In order to detect if directed evolution approaches for better ligand induction would also improve overall signal chain behavior of the DT mutants, we screened the three mutant libraries generated by error-prone PCR that led to the selection of second-generation (DT020, DT021, and DT022) and third-generation (DT038) mutants (25). For this, we isolated the rbsB gene variant pool and fused it C-terminally to mCherry as for the individual mutants. The library named epDT016 was started using the dt016 gene as the template (and used to recover DT020 to DT022), whereas epDT021 and epDT022 were started from dt021 and dt022, respectively (from which DT038 had been recovered [25]). Libraries were transformed into E. coli DH5α cells because of their higher efficiency than, e.g., E. coli MG1655. Although DT016-mCherry still produced fluorescent spots in E. coli after induction with l-arabinose (Fig. S3), samples from the epDT016 library barely showed cells with fluorescence (Fig. 5A). The estimated proportion of cells with properties expected from a cell expressing RbsB-mCherry (i.e., cell width range, low cell background, polar foci, and no side foci) was below 5%, whereas for DT016-mCherry this amounted to 20% (Fig. 5B). It is of note that the automated procedure is conservative and discards many poorly segmented cells, which can be seen from the fact that even images with RbsB-mCherry-expressing cells yielded, on average, only 50% “true” wild-type cells of all segmented cell objects (Fig. 5B). Spiking known amounts of RbsB-mCherry cells to the epDT016 library increased the proportion of detected wild-type cells, suggesting that the actual proportion of mutants with wild-type behavior is indeed very low (Fig. 5B). These results with epDT016 thus suggested that most error-prone introduced mutations are actually deleterious, which is not unexpected. In contrast, mutants DT021- and DT022-mCherry showed a higher proportion of wild-type-like cells, and although the corresponding evolved libraries, epDT021 and epDT022, reduced this again, the proportion of cells with wild-type-like subcellular fluorescence localization increased compared to the case with epDT016 (Fig. 5B). The final mutant recovered from those two libraries (i.e., DT038) indeed showed more wild-type-like behavior than DT022 in the proportion of cells with polar foci (n = 200 to 1,000 automatically segmented cells per replicate [Fig. 5C]). We concluded that the behavior of DT038 had improved compared to that of first- and second-generation mutants, which may have helped to gain better inducibility with the new ligand 1,3-cyclohexanediol.

FIG 5.

Wild-type-like improvement in successive DT mutant error-prone libraries. (A) Representative images (autocontrasted and cropped) of RbsB-mCherry-expressing E. coli DH5α cells and library samples of the first-generation (epDT016) and second-generation (epDT021 and epDT022) error-prone libraries. Note the (still) large fraction of individual cells in error-prone libraries that have completely lost DT-mCherry fluorescence. (B) Estimation of the proportion of cells behaving like E. coli DH5α expressing RbsB-mCherry, based on similar cell width, low cell background, and high polar fluorescence and absence of side foci on automatically segmented cells from images. Bars are replicate means with black dots representing single replicate values (combined from up to 10 technical replicates, typically 200 to 1,000 cells). “+WT” refers to proportions of spiked E. coli DH5α expressing RbsB-mCherry to the respective error-prone library. (C) Proportions of cells within cell width range and with only detectable polar foci for E. coli DH5α expressing RbsB-mCherry (WT), DT022-mCherry, or DT038-mCherry. Bars are replicate means with dots representing single replicate values. P values were derived from unpaired one-side t test on the shown grouped individual biological replicates.

DISCUSSION

We developed and tested an in vivo system to follow RbsB and mutant RbsB expression and their translocation into and localization within the periplasmic space in E. coli, with the objective to better understand the poor inducibility of RbsB mutants designed and selected for non-natural compound biosensing. As proxy for protein localization, we deployed C-terminally fused mCherry, and we quantified subcellular mCherry fluorescence signals in cytoplasm, periplasm, and aggregated forms (“foci”). Although RbsB-mCherry was not completely functional in its capacity to be induced by ribose, its subcellular localization in the periplasmic space and at the cell poles was consistent with the current assumptions on its behavior (3, 21). Therefore, within the context of subcellular localization, this was the best proxy to compare wild-type to mutant behavior in terms of expression, translocation, and signal chain localization. Whereas wild-type RbsB-mCherry in the absence of ligand was mainly present in the periplasmic space and its cell poles, the mutants underwent a variety of defects that appear as increased cytoplasmic fluorescence levels, more cells with homogenously high (and low) fluorescence, and increased numbers of non-polar-localized fluorescent spots. Our results with mutants and mutant libraries taken at different “evolutionary” steps further suggested that their overall behavior within the RbsB-Trz1 signaling chain has become more wild-type-like and consequently has contributed to their better inducibility with 1,3-cyclohexanediol (25).

Fluorescent protein fusion expression, even with wild-type RbsB, was actually quite sensitive for cell morphology and viability in the various E. coli backgrounds. Among the different tested genetic configurations, only the C-terminally fused mCherry, the native RbsB signal sequence, and inducible expression from P_BAD resulted in viable cells with proper visibly periplasmic localized fluorescence and fluorescent foci in the cell pole regions. The other configurations led to abnormal cells, possibly because of mistranslocation and blockade of the Sec or Tat translocation channels. Periplasmic fluorescence and focus visibility at both cell poles were l-arabinose dependent and dependent on a translocation signal sequence, from which we conclude that RbsB-mCherry forms not cytoplasmic but periplasmic polar foci. The absence of cytoplasmic RbsB-mCherry foci in constructs without translocation signal further indicates that there is no a priori autoaggregation of mCherry fluorescent protein that causes foci to appear. The localization at the (periplasmic) cell poles is most likely the result of RbsB-mCherry binding to the (Trg) chemoreceptors, which preferentially concentrate at the cell pole (28–30). However, RbsB-mCherry also produced polar foci in an E. coli Δtrg background; this mutant does not synthesize the Trg chemoreceptor and therefore was expected to show less or no polar fluorescence. This would thus indicate that polar accumulation of RbsB-mCherry not only is the result of a direct interaction between RbsB and Trg but is also due either to other chemoreceptors or to the ribose transporter channel or is a result of intrinsic membrane curvature, as previously reported for other fluorescent proteins (29, 31–36). On the other hand, there must be a definitive interaction between the periplasmic exposed Trg part and RbsB-mCherry, since expressing the hybrid chemoreceptor Trz1 (a hybrid protein composed of the periplasmic Trg and the cytoplasmic EnvZ part [8]) in a Δtrg background increased RbsB-mCherry fluorescence in the polar regions. The increased polar fluorescence of RbsB-mCherry in the Δtrg background with coexpressed Trz1 would then also dictate that the majority of Trz1 is located at the cell pole (despite having EnvZ cytoplasmic tails). Expression in both E. coli ΔrbsB and ΔrbsK backgrounds led to increased RbsB-mCherry fluorescence at the poles, which we attribute to less competition for chemoreceptor binding sites from native RbsB and, possibly, to decreased competitive binding to the ribose transporter, which is interrupted in the ΔrbsK mutant (26, 27). The positive effect of a ΔrbsK inactivation on ribose inducibility has been seen before (9), and this may thus be due to increased contact of RbsB to the Trz1 chemoreceptor. Although the majority of RbsB-mCherry located to the cell poles, a distinct diffuse halo was visible in the periplasmic space, suggesting circulating RbsB-mCherry complexes and/or ephemeral binding to transporter channels. We did not manage to capture any dynamic changes in RbsB-mCherry localization upon instant availability of the ligand ribose, which would require different experimental setups, but this might be interesting to study further.

The main objective of the work presented here was to understand if the previously observed poor inducibility of RbsB mutants designed and selected to become activated by nonnatural ligands (16) could be partly explained by impaired translocation and altered periplasmic localization properties. We concentrated on six mutants that were obtained in an effort to change ribose as a ligand for RbsB to 1,3-cyclohexanediol (16). Two of these (DT002 and DT016) had been obtained by computational redesign of the RbsB binding pocket (16). Although both proteins showed genuine gain of function to bind 1,3-cyclohexanediol and lost capacity to bind ribose, their induction was very small (16). Subsequently, both proteins were used as scaffolds for directed evolution through error-prone PCR and fluorescence-assisted cell sorting, from which we obtained three second-generation mutants (DT020, DT021, and DT022) and one third-generation mutant (DT038) that showed up to 4.5-fold-increased inducibility with 1,3-cyclohexanediol (25). Purified DT002 and DT016 had shown reduced thermostability compared to RbsB, the more severe for DT002, which suggested an inherent propensity for un- or misfolding (16). Although the other proteins have not been purified, we compared structural properties from AlphaFold predictions (37, 38) (Fig. S7). All predicted structures varied very little (average per-residue local distance difference test [pLDDT] scores close to 90), but we noticed that the per-residue values clearly varied at the mutated residues in the DT variants compared to RbsB (Fig. S7A). Interestingly, consistently 1 out of the 5 best models also predicted an “open” (i.e., non-ligand-bound) configuration for DT variants, in contrast to (predicted) RbsB itself (Fig. S7B). All DT002 models differed more strongly from the wild type than those of the three others (DT016, DT022, and DT038 [Fig. S7C]), which would be in line with previous thermostability measurements of purified DT002 and DT016 (16). Therefore, although AlphaFold does not explicitly inform about protein stability (39), these predictions thus still suggested more structural instability (or flexibility between open and closed configurations) for DT variants than for the wild type, which may influence inducibility and receptor binding.

Both thermostability of purified protein and structure predictions provided some evidence for increasing wild-type-like behavior from DT002 to DT016, but less so in further-selected mutants DT022 and DT038. However, localization results suggested that DT038-mCherry fluorescence localization behaves more like RbsB-mCherry (defined from the proportion of cells with only polar foci, low cytoplasmic fluorescence background, and cell width) than second- and first-generation mutants. Localization defects were clearly observed in cells expressing the first-generation DT mutant-mCherry fusion proteins. Indeed, the DT002-mCherry protein was less abundant in the periplasmic space and polar regions than wild-type RbsB-mCherry, which is in agreement with previous periplasmic space mass spectrum analysis (16). The larger proportion of cells with high cytoplasmic fluorescence may be the result of partially blocked Sec translocation channels by misfolded DT-mCherry proteins. Finally, a larger fraction of cells with non-polar-side fluorescent foci may indicate properties of the DT-mCherry proteins to autoaggregate in periplasmic space or to improperly bind to other cytoplasmic membrane channels. Both cases would lead to loss of signal in a bioreporter chassis such as the one deploying Trz1-ompCp::gfpmut that we use to test ribose or 1,3-cyclohexanediol induction (9). On the other hand, all DT mutant-mCherry proteins accumulated more strongly in polar regions when Trz1 was coexpressed, suggesting some affinity for the hybrid receptor. However, also the much more apparent cell-to-cell variability in polar and cell background fluorescence values in the case of DT mutant-mCherry expression, as opposed to that of wild-type RbsB-mCherry, indicates suboptimal performance between the mutant periplasmic binding proteins and the different components of the signaling pathway (e.g., chemoreceptors).

In conclusion, we demonstrate specific localization patterns mirroring the expected periplasmic targets of RbsB (studied by its proxy of a C-terminal mCherry fusion protein) in E. coli, and apparent defects in focus positioning and intensity and proportions of focus numbers per cell across clonal populations, which are evidence for poorly functioning mutant RbsB proteins with new ligand-binding properties. Quantification of mutant fusion protein expression and subcellular localization indicated overall improvement to become more wild-type RbsB-like over the course of multiple rounds of directed evolution, as a result of secondary mutations outside the direct binding pocket (25). Even though it is currently too challenging to isolate single cells from microscopy studies with improved subcellular localization behavior, our screening method permitted us to obtain a global idea of improvement within generated mutant libraries. The RbsB family of periplasmic binding proteins has been proclaimed as an ideal platform for protein ligand design by computational tools (10), but subsequent studies, including those by us, have tempered this enthusiasm (9, 14, 15). Partly this was due to poor and irreproducible designs (9), but partly, as we show here, it is due to ignoring of the complex interactions of PBPs with periplasmic and integral membrane signal chain components that are currently not captured in purely computational ligand pocket designs.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

All strains and plasmid constructions used in this study are listed in Table 1. Five E. coli strains were used as hosts for expression of RbsB- or mutant RbsB-fluorescent protein fusion constructs. E. coli BW25133 strains with insertional inactivation of the rbsB, trg, and rbsK genes were obtained from the Keio Collection (40). Wild-type E. coli MG1655 was obtained from the E. coli Genetic Stock Center (Yale University; CGSC 8237). Expression of the hybrid Trz1 chemoreceptor was provided by transformation of the plasmid pSYK1 (9) into the respective host background. E. coli cultures were grown in LB medium supplemented with 30 μg chloramphenicol mL⁻¹ and 100 μg ampicillin mL⁻¹ for strains containing plasmids pSTV and pSYK1, respectively. To maintain the P_BAD promoter in the off state (41), 0.4% (wt/vol) glucose was added to the LB culture medium (LB-Glc). To activate expression from P_BAD, overnight cultures were diluted 100 times in fresh LB medium without glucose but supplemented with 0.5% (wt/vol) l-arabinose. Coated microscopy slides for live-cell observations contained M9-basal medium without carbon source, jellified with 1% agarose.

TABLE 1.

Strains used in this study

Strain	E. coli host	Plasmids	Relevant characteristics	Reference or source
3044	DH5α λPir		Host for plasmid propagation	48
3671	DH5α	pSTVPAA_mcs	pSTVPAA to clone rbsB and its derivatives	9
4076	BW25113 ΔrbsB		Genomic deletion of rbsB gene	40
4172	BW25113 ΔrbsB	pSYK1	Host strain containing the Ptac-trzI, PompC-gfpmut2 bioreporter system	9
4175	BW25113 ΔrbsB	pSTVPAA_rbsB, pSYK1	RbsB expression with RbsBss for periplasmic translocation	9
4498	MG1655		Wild type, motile	E. coli Genetic Stock Center, Yale (CGSC 8237)
4501	BW25113 Δtrg		Genomic deletion of trg gene	40
4505	BW25113 ΔrbsK		Genomic deletion of rbsK gene	40
6686	BW25113 ΔrbsB	pSTVPAA_RbsBss-rbsB-mCherry, pSYK1	RbsB-mCherry expression with RbsBss for periplasmic translocation	This work
6952	DH5α λPir	pSTVPBAD_RbsBss-rbsB-mCherry	Inducible rbsB-mCherry expression and translocation (RbsBss)	This work
6953	DH5α λPir	pSTVPBAD_torAss-rbsB-mCherry	Inducible rbsB-mCherry expression and translocation (TorAss)	This work
6955	DH5α λPir	pSTVPBAD_RbsBss-mCherry-rbsB	Inducible mCherry-rbsB expression and translocation (RbsBss)	This work
6957	DH5α λPir	pSTVPBAD_torAss-mCherry-rbsB	Inducible mCherry-rbsB expression and translocation (TorAss)	This work
6958	DH5α λPir	pSTVPBAD_RbsBss-DT002-mCherry	As for 6952, but wild-type RbsB was replaced by DT002 mutant	This work
6959	DH5α λPir	pSTVPBAD_RbsBss-DT016-mCherry	As for 6952, but wild-type RbsB was replaced by DT016 mutant	This work
6960	BW25113 Δtrg	pSTVPBAD_RbsBss-rbsB-mCherry	Same as 6952 in host 4501	This work
6962	BW25113 Δtrg	pSTVPBAD_RbsBss-rbsB-mCherry, pSYK1	Same as 6972 in host 4501	This work
6963	BW25113 Δtrg	pSTVPBAD_RbsBss-DT002-mCherry	Same as 6958 in host 4501	This work
6965	BW25113 Δtrg	pSTVPBAD_RbsBss-DT002-mCherry, pSYK1	Same as 6973 in host 4501	This work
6967	BW25113 Δtrg	pSTVPBAD_RbsBss-DT016-mCherry	Same as 6959 in host 4501	This work
6969	BW25113 Δtrg	pSTVPBAD_RbsBss-DT016-mCherry, pSYK1	Same as 6974 in host 4501	This work
6972	DH5α λPir	pSTVPBAD_RbsBss-rbsB-mCherry, pSYK1	Inducible rbsB-mCherry expression and translocation (RbsBss), Ptac-trzI, PompC-gfpmut2 bioreporter system	This work
6973	DH5α λPir	pSTVPBAD_RbsBss-DT002-mCherry, pSYK1	As for 6972, but wild-type RbsB was replaced by DT002 mutant	This work
6974	DH5α λPir	pSTVPBAD_RbsBss-DT016-mCherry, pSYK1	As for 6972, but wild-type RbsB was replaced by DT016 mutant	This work
6975	DH5α λPir	pSTVPBAD_RbsBss-DT020-mCherry	As for 6952, but wild-type RbsB was replaced by DT020 mutant	This work
6976	DH5α λPir	pSTVPBAD_RbsBss-DT021-mCherry	As for 6952, but wild-type RbsB was replaced by DT021 mutant	This work
6977	DH5α λPir	pSTVPBAD_RbsBss-DT022-mCherry	As for 6952, but wild-type RbsB was replaced by DT022 mutant	This work
6991	DH5α λPir	pSTVPBAD_RbsBss-DT020-mCherry, pSYK1	As for 6972, but wild-type RbsB was replaced by DT20 mutant	This work
6992	DH5α λPir	pSTVPBAD_RbsBss-DT021-mCherry, pSYK1	As for 6972, but wild-type RbsB was replaced by DT021 mutant	This work
6993	DH5α λPir	pSTVPBAD_RbsBss-DT022-mCherry, pSYK1	As for 6972, but wild-type RbsB was replaced by DT022 mutant	This work
6996	BW25113 Δtrg	pSTVPBAD_RbsBss-DT020-mCherry	Same as 6975 in host 4501	This work
6997	BW25113 Δtrg	pSTVPBAD_RbsBss-DT021-mCherry	Same as 6976 in host 4501	This work
6998	BW25113 Δtrg	pSTVPBAD_RbsBss-DT022-mCherry	Same as 6977 in host 4501	This work
6999	BW25113 Δtrg	pSTVPBAD_RbsBss-DT020-mCherry, pSYK1	Same as 6991 in host 4501	This work
7000	BW25113 Δtrg	pSTVPBAD_RbsBss-DT021-mCherry, pSYK1	Same as 6992 in host 4501	This work
7001	BW25113 Δtrg	pSTVPBAD_RbsBss-DT022-mCherry, pSYK1	Same as 6993 in host 4501	This work
7007	BW25113 ΔrbsB	pSTV PBAD_RbsBss-RbsB-mCherry, pSYK1	Inducible RbsBss-RbsB-mCherry expression for periplasmic translocation in host 4172	This work
7035	DH5α λPir	pSTVPBAD_RbsB-mCherry	Inducible rbsB-mCherry expression and without RbsBss for translocation	This work
7146	BW25113 ΔrbsB	pSTVPBAD_RbsBss-mCherry-RbsB, pSYK1	As 7007, but for RbsBss-mCherry-RbsB	This work
7147	BW25113 ΔrbsB	pSTVPBAD_torAss-rbsB-mCherry, pSYK1	As 7007, but for torAss-rbsB-mCherry	This work
7148	BW25113 ΔrbsB	pSTVPBAD_torAss-mCherry-RbsB, pSYK1	As 7007, but for torAss-mCherry-RbsB	This work
7189	BW25113 ΔrbsK	pSTVPBAD_RbsBss-rbsB-mCherry	Same as 6952 in host 4505	This work
7190	BW25113 ΔrbsK	pSTVPBAD_RbsBss-DT016-mCherry	Same as 6959 in host 4505	This work
7191	BW25113 ΔrbsK	pSTVPBAD_RbsBss-DT022-mCherry	Same as 6977 in host 4505	This work
7192	MG1655	pSTVPBAD_RbsBss-rbsB-mCherry	Same as 6952 in host 4498	This work
7193	MG1655	pSTVPBAD_RbsBss-DT016-mCherry	Same as 6959 in host 4498	This work
7194	MG1655	pSTVPBAD_RbsBss-DT022-mCherry	Same as 6977 in host 4498	This work
7195	BW25113 ΔrbsB	pSTVPBAD_RbsBss-rbsB-mCherry	Same as 6952 in host 4076	This work
7196	BW25113 ΔrbsB	pSTVPBAD_RbsBss-DT016-mCherry	Same as 6959 in host 4076	This work
7197	BW25113 ΔrbsB	pSTVPBAD_RbsBss-DT022-mCherry	Same as 6977 in host 4076	This work

Cloning of mCherry fusion protein constructs.

To study RbsB and mutant RbsB localization in the cell, we translationally fused rbsB and mCherry open reading frames in different configurations and under the control of different promoters (Fig. S1). To favor translocation, the reading frames were N-terminally fused to sequences coding for either the cognate RbsB signal peptide (RbsBss) or the TorA signal peptide (TorAss).

For constructs with the constitutive P_AA promoter (42), we recovered the wild-type rbsB gene (without the stop codon but including or not the signal peptide coding sequence) or the respective dt gene variant (RbsB mutants with designed 1,3-cyclohexanediol recognition [16, 25]) by PCR amplification from their respective plasmid vector, pSTV28PAA_ (9). The gene coding for mCherry was amplified from pARS1003 (43). Primers used to amplify rbsB/dt and mCherry add a unique SpeI restriction site and included a coding sequence for an 8-amino-acid linker (GGSGSGSR) between the two proteins in both C- and N-terminal fusion positions (Table 2), which we previously showed functioned well for chemoreceptor localization studies (44). Amplicons were then cloned into vector pSTV28PAAmcs (9), digested with NdeI and SalI using a ClonExpress kit according to the manufacturer’s instructions (Vazyme, China), and heat shock transformed (180 ng) into E. coli DH5α cells.

TABLE 2.

List of primers used to produce the different fusion constructs

Fusion construct	Target	Primer	DNA sequence (5′–3′)
pSTVPAA_ RbsBss-RbsB-mCherry	RbsBss-rbsB (without stop codon)	190402 F	ACTTTAAGAAGGAGATATACATATGAACATGAAAAAACTGG
		190401 R	AGAGCCAGAGCCACCACTAGTGTGGTGGTGGTGGTGGTGCTCGAGCTGCTT
	mCherry	190404 F	ACTAGTGGTGGCTCTGGCTCTGGCTCGAGAATGGTGAGCAAGGGCGAGGAGGA
		190403 R	CAAGCTTGCATGCCTGCAGGGGAATTGGGGATCGGAAGCT
pSTVPBAD_ RbsBss-RbsB-mCherry	araC/PBAD	200102 F	CTGCAGGTCGACGGAGCTCGTGTTTGACAGCTTATCATCGATGCATAATGTGCCTG
		200103 R	GCTAGCCCAAAAAAACGGGTATGG
	RbsBss-rbsB (without stop codon)	200104 F	ACCCGTTTTTTTGGGCTAGCGAAGGAGATATACATATGAA
		190401 R	AGAGCCAGAGCCACCACTAGTGTGGTGGTGGTGGTGGTGCTCGAGCTGCTT
	mCherry	190404 F	ACTAGTGGTGGCTCTGGCTCTGGCTCGAGAATGGTGAGCAAGGGCGAGGAGGA
		200101 R	TATGACCATGATTACGAATTGGAATTGGGGATCGGAAGCT
pSTVPBAD_ TorAss-RbsB-mCherry	araC PBAD + torAss	200102 F	CTGCAGGTCGACGGAGCTCGTGTTTGACAGCTTATCATCGATGCATAATGTGCCTG
		200106 R	GGCCGCTTGCGCCGCAGTCGCACG
	rbsB (without stop codon)	200105 F	CGACTGCGGCGCAAGCGGCCATGGCAAAAGACACCATCGCGCTG
		190401 R	AGAGCCAGAGCCACCACTAGTGTGGTGGTGGTGGTGGTGCTCGAGCTGCTT
	mCherry	190404 F	ACTAGTGGTGGCTCTGGCTCTGGCTCGAGAATGGTGAGCAAGGGCGAGGAGGA
		200101 R	TATGACCATGATTACGAATTGGAATTGGGGATCGGAAGCT
pSTVPBAD_ RbsBss-mCherry-RbsB	PBAD region + RbsBss	200102 F	CTGCAGGTCGACGGAGCTCGTGTTTGACAGCTTATCATCGATGCATAATGTGCCTG
		200121 R	CGCATTCGCACTGACGGTGGCGCTTAGCGCAACAGCGGAAACCAGGGTAGCCAGT
			TTTTTCATGTTCATATGTATATCTCCTTCGCTAGCCCAAAAAAACGGGTATGG
	mCherry (without stop codon)	200109 F	CCACCGTCAGTGCGAATGCGATGGTGAGCAAGGGCGAGGAGGATAAC
		200110 R	AGAGCCAGAGCCACCACTAGTCTTGTACAGCTCGTCCATGCCGCCGGTGGA
	rbsB	200111 F	ACTAGTGGTGGCTCTGGCTCTGGCTCGAGAATGGCAAAAGACACCATCGCGCT
		200107 R	TATGACCATGATTACGAATTTTGCATGCCTGCAGGTCGACTCAGTGGT
pSTVPBAD_ TorAss-mCherry-RbsB	araC/PBAD + torAss	200102 F	CTGCAGGTCGACGGAGCTCGTGTTTGACAGCTTATCATCGATGCATAATGTGCCTG
		200105 R	CGACTGCGGCGCAAGCGGCCATGGCAAAAGACACCATCGCGCTG
	mCherry (without stop codon)	200112 F	CGACTGCGGCGCAAGCGGCCATGGTGAGCAAGGGCGAGGAGGATAAC
		200110 R	AGAGCCAGAGCCACCACTAGTCTTGTACAGCTCGTCCATGCCGCCGGTGGA
	rbsB	200111 F	ACTAGTGGTGGCTCTGGCTCTGGCTCGAGAATGGCAAAAGACACCATCGCGCT
		200107 R	TATGACCATGATTACGAATTTTGCATGCCTGCAGGTCGACTCAGTGGT

The resulting plasmids (Table 1) were used to produce derivative constructs in which expression was under the control of the inducible P_BAD promoter (41). The rbsB/mCherry or dt/mCherry gene variants were recovered from their pSTV28PAAmcs vectors by double digestion with EcoRI (which removes the P_AA promoter region). The P_BAD promoter region with or without torA signal peptide sequence was amplified by PCR from pCRO4 (44) and fused by ClonExpress cloning as described above. Plasmids were purified from their E. coli hosts and the different fusion constructs were analyzed by sequencing to verify their integrity as designed. Plasmids carrying different fusion constructs were then transformed into the various E. coli backgrounds or cotransformed with pSYK1 to express Trz1.

Expression of RbsB- and DT-mCherry fusions for epifluorescence microscopy.

For single-cell microscopy analysis, we started with a single E. coli colony grown on a selective plate to maintain the respective plasmid(s). The colony was inoculated in 5 mL LB-Glc medium (supplemented with the appropriate antibiotics) and grown overnight at 37°C with shaking at 180 rpm. The next morning, the culture was diluted 100 times in LB-Glc medium with antibiotics. Expression from P_BAD was induced by addition of 0.5% (wt/vol) l-arabinose to the culture, which was incubated at 37°C for 2 to 3 h and with shaking at 180 rpm until reaching exponential phase (culture turbidity, i.e., optical density [OD], at 600 nm of 0.5). Samples of 4 μL were pipetted on M9-agarose-coated standard microscopy slides (Menzel-Gläser) and covered with a coverslip. Images were acquired with a Zeiss Axioplan II epifluorescence microscope with a 100× Plan Apochromat oil objective (Carl Zeiss, Jena, Germany) and a Sola SE light engine (Lumencor, USA). A SPOT Xplorer slow-can charge-coupled-device camera (1.4 megapixels, monochrome, without infrared; Diagnostic Instruments) fixed on the microscope was used to capture images. Up to 10 images at different positions were acquired using Visiview software (Visitron Systems GmbH), with exposure times set to 10 ms for phase-contrast (PhC) and 500 ms for mCherry. All constructs were imaged in at least biological triplicates, either on different days with triplicates grown independently or as triplicates examined on the same day.

Flow cytometry.

To test the functionality of different configurations of RbsB/mCherry variants to be induced by ribose, we quantified appearance of Gfpmut2 fluorescence by flow cytometry, as previously described (45). In short, E. coli ΔrbsB pSYK1 strains expressing the rbsB/mCherry plasmid constructs were grown for 16 h in M9 medium with 20 mM fumarate, diluted to 10 replicates, regrown for 2 h in the same medium, then mixed with 0.1 mM ribose or not, incubated for 2 h, and sampled for analysis on an Acea NovoCyte 3000 flow cytometer (Bucher Biotec, Switzerland). Gfpmut2 fluorescence was detected in the FL1-H channel, whereas corresponding mCherry fluorescence was detected in the Texas Red instrument channel. Data are reported as the mean of replicate population mean fluorescence above cell threshold.

Experiments with spiked mutant libraries.

Mutant libraries had previously been generated by error-prone PCR in the presence of MnCl₂ (0.025 to 0.06 mM) using the RbsB variant genes dt016, dt021, and dt022 as the template, targeting a frequency of 1 to 3 sense mutations per 1,000 bp (25). Mutant gene fragments were extracted as a DNA pool and fused at the 3′ end to the mCherry coding sequence, and at the 5′ end to the P_BAD-RbsBss promoter and signal peptide sequence, as explained above. ClonExpress reactions (∼180 ng DNA) were heat shock transformed into E. coli DH5α and selected by antibiotic resistance, after which the libraries were stored as a mixture with 15% (vol/vol) glycerol at −80°C. For microscopy analysis, small aliquots were thawed and precultured as described above for single pure cultures and then supplemented with 0.5% (wt/vol) l-arabinose to induce the synthesis of the mCherry fusion protein. Library mixtures were imaged directly or spiked with induced pure culture of E. coli DH5α expressing wild-type P_BAD-RbsBss-RbsB-mCherry in cell-cell proportions of 50%, 10%, 1%, and 0.1%.

Image analysis.

One mCherry image of each E. coli host background and construct (20 to 100 cells per series) was segmented manually using ImageJ (46) and the corresponding phase-contrast image as the template. Image background, cell length, cell pole regions, cell background, and additional foci per cell were outlined as individual objects, of which the mean pixel fluorescence was measured (Fig. S2). Polar regions were arbitrarily divided to maximum and minimum values and counted as “foci” if their mean fluorescence was higher than the mean cell background. Sixteen-bit images were opened in Adobe Photoshop (Vs 2020), “autocontrasted,” cropped to a region of 800 by 600 pixels (312 pixels cm⁻¹) and saved as 8-bit TIFF files for display.

In order to analyze a larger number of images and culture replicates, we deployed automated cell segmentation on phase-contrast images to create binary masks, which were then transposed to the fluorescence image; as implemented in SuperSegger (47). SuperSegger stores and quantifies relevant single cell features, which were extracted using MATLAB (v R2016a; MathWorks). Notably, this consisted of cell length, width, orientation, median and mean top 5% fluorescence, and focus information (e.g., number of foci per cell and focus position, score, and intensities). Note that fluorescent foci in SuperSegger are determined by Gaussian fitting, which is a slightly different quantification than the manual outlining of cell poles and causes more noise from spurious foci. However, focus superpositioning across hundreds of cells is similar to manual outlines (see, e.g., Fig. S2). Cell and focus lists were combined to further extract median fluorescence and means of the top 5 percentile fluorescence values (MATLAB v R2016a), which are representative for the cytoplasmic mCherry expression and foci fluorescence intensity, respectively (Fig. S2). Foci with a normalized intensity threshold of <3 were considered spurious foci on the basis of uninduced cells and were removed (Fig. S2). Focus positions were counted as polar if their position on the long cell axis was within 10 pixels of the cell tip; otherwise, they were considered side foci. Cells were counted as having two polar foci if the paired focus distances were at least 0.75 times the measured cell long axis. In order to count cells with wild-type-RbsB properties, we first filtered objects to the cell width range (10 to 20 pixels) observed for E. coli DH5α cells expressing P_BAD-RbsBss-RbsB-mCherry (covering ∼70% of segmented cells [Fig. S2]). Next, these cells were filtered to having distinguishable foci and cell fluorescence above image background; in other words: their mean top 5% fluorescence should be higher than the median cell fluorescence plus twice the observed standard deviation, and the mean of all automatically detected foci should be higher than the median cell fluorescence plus observed standard deviation. Finally, in order to qualify as a wild-type RbsB-mCherry-expressing cell, the cell should not display any detectable side foci and display only polar foci (criteria explained above).

Structure prediction.

We took advantage of the recent developments in machine-learned protein structure predictions using the AlphaFold server (37, 38) to compare DT mutant proteins and the RbsB wild type. Amino acid sequences (without signal sequence) were uploaded to the GitHub server (https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/beta/AlphaFold2_advanced.ipynb) and analyzed using default settings, resulting in the prediction of the best five model structures. Per-residue local distance difference test values (pLDDT) were extracted from produced .pdb files, averaged from five replicates and subtracted from those of wild-type RbsB (predicted using the same pipeline) to show positional differences. The mean absolute difference was then taken as a measure for the overall dissimilarity to RbsB.

Statistical analysis.

Statistical analysis was done using GraphPad Prism software (v 8.4.3). Fold induction differences of cells expressing fusion constructs in the presence of ribose were tested using one-way ANOVA followed by post hoc Tukey test. Differences in mCherry fluorescence were tested by comparing the extracted mean top 5% fluorescence and median fluorescence values from individual biological replicates in cultures with and without l-arabinose induction in paired and nonpaired t tests. Numbers of replicates were variable and are indicated in figures or their legends.