Complete alanine scanning of the Escherichia coli RbsB ribose binding protein reveals residues important for chemoreceptor signaling and periplasmic abundance

The Escherichia coli RbsB ribose binding protein has been used as a scaffold for predicting new ligand binding functions through in silico modeling, yet with limited success and reproducibility. In order to possibly improve the success of predictive modeling on RbsB, we study here the influence of individual residues on RbsB-mediated signaling in a near complete library of alanine-substituted RbsB mutants. Among a total of 232 tested mutants, we found 10 which no longer activated GFPmut2 reporter expression in E. coli from a ribose-RbsB hybrid receptor signaling chain, and 13 with significantly lower GFPmut2 induction than wild-type. Quantitative mass spectrometry abundance measurements of 25 mutants and wild-type RbsB in periplasmic space showed four categories of effects. Some (such as D89A) seem correctly produced and translocated but fail to be induced with ribose. Others (such as N190A) show lower induction probably as a result of less efficient production, folding and translocation. The third (such as N41A or K29A) have defects in both induction and abundance. The fourth category consists of semi-constitutive mutants with increased periplasmic abundance but maintenance of ribose induction. Our data show how RbsB modeling should include ligand-binding as well as folding, translocation and receptor binding.

the concept, recent publications have been unable to reproduce the initial findings 16,17 , and it has been suggested that the designed mutants resulted in unstable proteins 18 . In addition, it is likely that the initial computations concentrated solely on calculating new PBP binding site specificities on the basis of existing crystal structures with or without bound ligand, but neglecting the other characteristics of a functional PBP. This may include processes such as: folding and translocation into the periplasm, dynamic ligand-binding (cleft opening and closing), and productive interactions of ligand-bound PBP to its membrane channel or receptor [19][20][21][22][23] . Possibly, computational design of binding specificities of dynamic proteins such as PBPs can be improved with better scoring functions that embed empirical data on each step of its functional cascade. As a start to better functionally characterize the importance of amino acid residues of RbsB in the complete ribose-induced signaling chain, we carried out an alanine scan of the near complete protein. Results from a limited alanine scan of RbsB have been published previously 24 . To proceed, a total of 235 RbsB mutants were produced by DNA synthesis in which the original residue was replaced by Ala. These mutants were individually expressed in an E. coli chassis with the Trz1 hybrid signaling pathway between the periplasmic part of the Trg chemoreceptor and the cytoplasmic part of EnvZ, enabling production of the GFPmut2 fluorescent protein 16,25 in response to ribose addition (Fig. 1). GFPmut2 expression was analyzed by flow cytometry in cells exposed or not to ribose. Additionally, RbsB levels in the periplasmic space of E. coli were quantified by nano liquid chromatography followed by direct peptide mass identification (LC-MS) in a selection of 25 mutants with the strongest defects plus the wild-type. Our results highlight a variety of critical amino acid residues for ribose-induced RbsB signaling also at seemingly inconspicuous positions, which help to understand previous irreproducible design results.

Results
Production of an Ala-substitution library of RbsB. A near-complete Ala-substitution library of the E. coli RbsB protein was produced from a combination of pooled mutant DNA synthesis and individual mutant recovery or construction. Despite multiple repeated attempts, three replacements could not be obtained (i.e., 135T, 136S and 145F). 37 positions (14%) of RbsB wild-type contain an Ala-residue and were not further changed in this work. All mutant rbsB genes were transferred to the generic vector pSTVP AA mcs for expression from the constitutive P AA -promoter 16 , and verified by sequencing again after final cloning. Noteworthy, several Alareplacements in RbsB led to appreciable slower growth of the E. coli expression strain, apparent as smaller colonies on plates and lower counts in flow cytometry analysis (Fig. S1).
Effect of Ala-substitution on inducibility with ribose. The final 232 mutants were analyzed in triplicate assays by flow cytometry for GFPMut2 fluorescence as a result of induction by ribose from the hybrid Trz1-OmpR-ompCp-gfpmut2 signaling chain (Fig. 1), in comparison to E. coli expressing wild-type RbsB (strain 4175) as well as E. coli without rbsB (strain 4497) (Dataset 1). Inducibility with 1 mM ribose was measured on cultures grown and exposed in 96-well plates on different occasions; therefore, every plate contained a separate triplicate wild-type positive and a negative control. In general terms, we observed both partial and complete loss of inducibility by ribose, appearance of semiconstitutives (with higher values of GFPMut2 fluorescence than wild-type in cells not exposed to ribose), as well as similar inducibility as wild-type RbsB ( Fig. 2 and Dataset 1). Ten percent of Ala-substitutions led to a loss of ribose induction, which we defined from a fluorescence less than the average minus two standard deviations (SD), calculated from the data variation among all mutants and Ribose-binding protein (RbsB) captures its ligand ribose, leading to a conformational change. Ribose-RbsB binds the Trz1 hybrid transmembrane receptor, which causes a phosphorylation cascade leading to OmpR-P binding the ompC-promoter and consecutive gfpmut2 expression.
wild-type ( Fig. 2B and Dataset 1). Among those, half were not inducible at all, or even showed lower fluorescence in presence than in absence of ribose. Approximately, seven percent led to a statistically significant increased GFPMut2 fluorescence (i.e., higher than the average plus two SD) in absence of ribose compared to wild-type RbsB ( Fig. 2A and Dataset 1). Ribose induction profiles allowed to group the Ala-replacement mutants of RbsB loosely into three broad groups shown in Table 1, based on the criteria described above (e.g., no or weak induction, semi-constitutive).
RbsB mutant periplasmic space abundance. Since some of the differences in ribose induction compared to wild-type RbsB may be due to differences in RbsB-mutant stability, folding and/or translocation to the periplasmic space, we analyzed periplasmic protein abundance for a selection of 25 mutants with the strongest defects from the groups in Table 1, and of wild-type, grown under standard conditions as for ribose induction. Analysis of SDS-PAGE separated periplasmic protein fraction in the range of 28-40 kDa, trypsin-digested and quantified by nano LC-MS, revealed a total of 424 distinct proteins (Fig. S2). RbsB (mutant) peptides were identified and quantified from unique fragments (exclusive spectrum count), normalized by total spectrum count per sample and further by D-galactose-binding periplasmic protein (MglB) counts as internal standard (Table 2). Samples with a significantly lower abundance of RbsB mutant peptides than expected from variation of the MglB internal standard (F-test; p < 0.05) were considered carrying a mutation affecting folding, translocation and/or stability of the mutant RbsB in(to) the E. coli periplasm (Table 3). Assuming proportional decrease of GFPmut2 expression and the normalized RbsB periplasmic abundance (Table 3), we conclude that Ala-substitutions at D89, V106, I111, R141, G144, V186, G196, D215, P262, and D264 primarily have signaling defects, since their abundance is similar (or even slightly higher) as wild-type, but GFPmut2 fluorescence upon induction with ribose is lower (Table 3). In contrast, Ala-substitutions at I63, I86, N190 and I233 primarily seem to affect protein synthesis, folding, stability and/or translocation, since their fluorescence decreases (except for I63, a semi-constitutive) as much as their abundance in comparison to wild-type RbsB (Table 3). Ala-substitutions in L6, T10, K29, G31, L34, L88, N41 and L195 seem to affect both processes, since the loss in abundance is not sufficient to explain the decrease in normalized fluorescence through the signaling chain (again assuming proportionality between the two, Table 3). Of note that MglB abundance in mutants L6A, V106A and V186A itself was lower than expected from the variance across all samples (Table 3). Therefore, these RbsB mutants may fold to such a state that they block general protein translocation channels for other periplasmic binding proteins. Mutants P237A and I111A showed higher than expected abundance but lower normalized fluorescence and, therefore, are also impaired in signaling. Finally, mutants I132A and N105A showed a higher level of periplasmic abundance than expected from wild-type RbsB, but only N105A expressed constitutively higher GFPmut2 fluorescence in absence of ribose (Table 1). Positions of Ala substitution effects. The Ala-substitution mutants with observed effects cover any part of RbsB except a few helices (Fig. 3A,B). Not surprisingly, several mutants are located within the ribose binding pocket, notably D89A, N190A, D215A and R141A, and to a lesser extent L88A (Fig. 3C). At the cleft entry are two mutants T10A and N41A. On the back side of the binding pocket is the flexible hinge region (Fig. 3B) where six of the mutants can be found. These are D104A, V106A, I233A, P237A, P262A and D264A (Fig. 3D). Furthermore, two mutations causing a semi-constitutive phenotype (S103A and N105A) are also positioned in the hinge region. Several more mutants locate in the N-domain of RbsB (Fig. 3B); notably in a peripheral corner in a helix on the "top" (K29A, G31A) and within beta-sheets (L6A, L34A, I86A, Fig. 3D-F). Furthermore, prominent arms in the C-domain display two groups of consecutive mutants (G144A, which is close to the R141 in the binding site; and L195A, G196A and V186A in the arm supporting the N190 binding site residue, Fig. 3D-F). Interestingly, two of the positions for which no substitutions could be obtained (T135, S136) locate on the edges of the cleft of RbsB in the structure (Fig. 3E), which may therefore be also critical for its folding and functioning.
Several of the critical positions we found here are not absolutely conserved among members of the cd06323:PBP1_ribose_binding-domain containing family (Fig. S3). Interestingly, the third missing Ala-replacement mutant (F145) is at a conserved position, as is its neighbor G144 (G144A caused semi-constitutive expression). Vice-versa, many of the conserved amino acids in this taxonomically extremely widespread domain family can be substituted in RbsB by Ala without noticeable effects (Fig. S4, hash-signs).

Mutant nature Position
Amino acid  Table 1. Ala-replacement mutants of RbsB causing significant change in ribose induction. a Average cell fluorescence in flow cytometry after 2 h induction with 1 mM ribose, averaged from biological triplicates ± SD. b Fold-induction is the ratio of the average cell fluorescence with ribose (induced) and without ribose (uninduced). c Fold-induction at or below 1 in presence of 1 mM ribose. d Positions in bold type-face indicate mutants whose expression in the periplasmic space was verified by quantitative peptide mass spectrometry.
e Induced values lower than the average induction across all mutants including wild-type, minus 2× the SD across all mutant induction data. f Uninduced values minus their SD higher than the average uninduced value across all mutants including wild-type, plus 2× the SD across all mutant uninduced data.

Discussion
One of the major hurdles in bacterial bioreporter development is the choice of the chemical target receptor protein. Whereas most current bioreporters have exploited naturally existing transcription regulators or sensory proteins with their cognate ligand specificities 26 , for detection of new (often manmade) molecules it will be necessary to find or engineer new ligand-binding properties 27 . One of the proposed possibilities for engineering new ligand-binding properties has been the periplasmic-binding-protein scaffold, for which crystal structures in absence and presence of ligands are known 14,21,22,25 . Predictions and experimental data from a decade ago suggested that completely new ligand specificities could be obtained on scaffolds such as the ribose-binding protein (RbsB) of E. coli, changing, for example, from ribose to trinitrotoluene 15 . Key results from this work could not be reproduced and have been seriously challenged 16 . It thus appears that our knowledge on the various steps of RbsB-ligand signaling, such as binding of the ligand and binding of the ligand-protein complex to the receptor, as well as RbsB folding, periplasmic translocation and stability itself, must be as yet incomplete. Hence, attempts for prediction of new ligand-binding specificities solely on the basis of modeling of the PBP binding pockets may lead to false results.
Here we focused on obtaining a better understanding of the influence of the individual RbsB residues on ligand-induced signaling as well as RbsB periplasmic space abundance. A near complete series of Ala-substitution mutants of non-Ala residues of RbsB was produced and characterized for loss of ribose-induced signaling in E. coli (Table 1). A total of 23 mutants was found, which resulted in stronger decrease than expected by variation of GFPmut2 expression from the Trz1-OmpR hybrid signaling pathway upon ribose induction. Several other mutations were found that resulted in a significant increase of fluorescence signal irrespective of ribose addition. Mutants displaying significant defects in GFPmut2 induction were further examined for RbsB-mutant protein abundance in the periplasmic space (Table 2). From the combined results we can infer whether an Ala-substitution acts on the signaling pathway itself (hence: ligand and/or receptor binding), on protein abundance (assumed to be the result of defects in either translation, folding and/or translocation, or stability), or both. Our assumption here was that a decrease in periplasmic abundance might lead to a proportionally diminished fluorescence reporter output from the signaling chain, although this relation may be more complex for individual RbsB mutants. First we want to highlight the presumed ribose binding pocket, which has already been subject of a previous investigation by Vercillo et al. who constructed 13 alanine-substitution mutants in residues on the basis of their proximity to bound ribose in the crystal structure 24 . In the ribose-binding pocket of RbsB we found five mutants leading to a defect in the behavior of the bioreporter (Fig. 3C). Mutants D89A, D215A, R141A, and N190A showed a strong loss of inducibility, but the periplasmic abundance of the N190A mutant was proportionally as low as the loss in ribose-induced GFPmut2 fluorescence. This suggests that N190 is not so much involved in ribose binding but a key residue influencing protein folding, stability and/or translocation into the periplasm. Vercillo et al. 24 demonstrated that Ala-substutions at positions 89, 215, and 190, cause loss of ribose binding by the purified protein. Purified N190A mutant protein displayed a significant decrease in stability 24 , which is in agreement with our findings. R141A was not part of their mutant selection. In contrast, we did not find statistically significant effects at 1 mM ribose induction of Ala-substitutions on F16, F164, F214 and Q235 as well as N64, which were previously reported as losing ribose binding capacity 24 .
Mutants N13A and F15A, which are located close to the entrance of the cleft, displayed a semi-constitutive behavior, similar to I132A and N105A (Table 1 and Dataset 1). Due to their location, N13A and F15A might change the ratio between the open and closed conformation of the protein, and, in this way, lead to constant recognition by the Trz1 chemoreceptor and thus higher background fluorescence. According to Vercillo et al. the purified N13A and F15A mutant proteins (in an L265C RbsB background) are stable and bind ribose, albeit (slightly) diminished compared to wild-type 24 . This may explain their semi-constitutive behavior in our reporter background. Mutation of the nearby residue I132, located at the entrance of the pocket, produced an even stronger semi-constitutive phenotype as N13A or F15A, which is in agreement with previous results that purified I132A has similar stability and ribose-binding as wild-type RbsB 24 . Position I132 had also been shown by Björkman et al. to play a crucial role in the intra-protein connection between the two domains of RbsB (Fig. 3B)    in order to facilitate the closed conformation 22 . The connection partner for the formation of the van der Waals contact is position N41 22 . An alanine substitution at this place led to a strong growth defect (Fig. S1), a significant decrease of the periplasmic abundance of the protein and a complete loss of induction (Tables 2 and 3). We therefore conclude that this position is both important for protein stability and for ribose-or Trz1 binding (Fig. 3F). Notably, also two residues for which no substitution could be obtained, locate in the entry cleft (T135, S136, Fig. 3E,F). On the opposite side of the ligand entrance is the hinge region (Fig. 3B). Here we detected two groups of mutation effects, which may therefore influence RbsB's ability for adopting conformational changes. The first group, consisting of D104A, V106A, I233A, P237A, P262A and D264A, led to a significant loss of induction by ribose, which became more dramatic the closer the residue's location to the hinge (Tables 1 and 3, Fig. 3D,E). The other group, with S103A and N105A, resulted in opposite, more constitutive behavior, but with maintenance of induction by ribose ( Table 1). Four of the mutants in this region, D264A, P262A, P237A and N105A, even displayed higher periplasmic abundance than wild-type RbsB (Tables 2 and 3), suggesting a decreased cytoplasmic pre-folding and thus a potentially higher secretion rate 28 . Although we did not specifically examine S103A's abundance in the periplasm, we assume it to be similar to N105A because of its high fluorescence value ( Table 1). The higher background of S103A and N105A mutations, but maintenance of ribose induction potential may be due to more limited hinge movement but adoption of a configuration favorable for contact to the receptor. This would be in agreement with the measured wild-type stability and similar wild-type binding constant of ribose for purified S103A protein 24 . In contrast, the effect of the mutants D264A, P262A and P237A may be the result of an unfavorable change to the hinge region, which prevents adoption of the closed form upon binding of ribose or impedes subsequent contacts to the Trz1 receptor (none of these mutations have been previously analyzed).

Effect Position
Mutants L6A, L34A, I86A, and I63A locate within the two β-sheets in the interior of the protein N-domain (Fig. 3D,E) and showed a significantly reduced periplasmic abundance, suggesting a strong influence on RbsB folding, stability or translocation. The loss of ribose-induced signaling for L6A and L34A was proportionally stronger than can be explained from their decreased periplasmic abundance. Therefore, this position may have an additional influence on the signaling process. Interestingly, strains expressing L34A also grew poorer (as N41A), which may point to these proteins clogging E. coli's translocation systems 28 . Surprisingly, mutant I63A showed a semi-constitutive response despite a lower periplasmic abundance. Noteworthy, the strain expressing mutant L6A showed a complete absence of MglB (galactose-binding protein) in its periplasm, in comparison to the other strains (Table 3). This mutant may thus cause extreme folding defects that inhibit translocation of other PBPs. The location of many Ala-replacement mutants with lower RbsB periplasmic space abundance in the N-domain is in agreement with the location of most spontaneous RbsB suppressor mutants that overcame a defect in SecB-dependent translocation 19 . It suggests that the N-domain (as that part of the protein that passes first through the translocation channel) is particularly sensitive to misfolding.
Positions I111, G144 and F145 are situated at two alpha-helices of the C-terminal part (Fig. 3E). Ala-substitutions in two of these residues caused a significant loss of induction, which could not (I111A) or only partially (G144A) be explained by reduced periplasmic abundance. Mutant F145A could not be obtained, despite multiple trials. This is a conserved residue among a wide range of periplasmic-binding proteins of the cd06323:PBP1_ribose_binding protein domain family, suggesting it plays a pivotal role (Fig. S3). Similarly, substitutions in L195, G196, and V186 in other helices and sheets in the C-domain caused drastic effects on induction and/or on protein abundance (L195A, Table 3). This suggests some global structural change of the protein resulting in a perturbation of the hinge region and thus influencing the relative positions of the residues on the cleft side, similar as has previously been suggested on an analysis of the RbsB mutant I111R 23 .
One would expect to have produced substitutions causing a defect in ribose-bound-RbsB binding to the Trz1 receptor, leading to decreased signaling and, hence, decreased Gfpmut2 fluorescence. Amidst the list and location of mutants causing effects that cannot be more simply explained by mechanistic details of ribose binding or hinge-movement, or residues of structural importance (e.g., Gly-or Pro-residues), we find two conspicuous mutations: K29A and E255A. Both of these locate to the left outside part of the protein's N-domain (Fig. 4), but whereas substitution in K29A abolishes induction, that in E255A causes a semi-constitutive phenotype (Dataset 1 and Table 1). Both residues are charged, locate on the outside (Fig. 4) and change their relative orientation between open and closed configuration (Fig. 3B). On the other hand, these residues have not been detected in previous spontaneous or constructed RbsB mutants with inferred receptor or permease-binding defects [19][20][21][22][23] . The RbsB region contacting Trz1 has not been characterized, but, according to previous work analyzing RbsB mutants impaired in ribose chemotaxis, it would cover residues I111 and L107 on the "back" of the protein (Fig. 4) and A44, A70, N73 and G134 on the "front" (Fig. 4) 23 . Part of these residues (A44, A70, N73, I111, L107) overlap with the proposed region on RbsB binding the ribose permease 20,22,23,29 (Fig. 4). None of the Ala-substitutions to those residues (Fig. 4), except I111A, caused measurable loss of Trz1 signaling in our experiments, but D52A and R166A showed elevated background expression (Dataset 1), which may be due indirectly to altered ribose transport 20 .
Our study largely expands previous more limited mutation analyses of RbsB 20,23,24 , resulting in a more complete view on the importance of RbsB residues for protein stability and transport, as well as for proper signaling in the Trz1 transduction pathway. In addition to the more obvious residues implicated in ribose binding, we were able to highlight the importance of various other inconspicuous residues effecting the folding, structure or transport of the protein and also the signal transduction towards the chemoreceptor. Several of the positions we identified as crucial for proper RbsB functioning had been in fact targeted by the aforementioned RbsB protein design study, such as D89(S), R141(S), N190(F) or D215(S) 15 . Although it is possible that the negative effect of an Ala-substitution is absent in another substitution, this suggests that great caution has to be taken when targeting residues that are critical for ligand binding (D89, R141, D215) or protein stability and transport (N190). On the other hand, it suggests that mutant ligand-binding RbsB with poor structural stability may be improved by further mutating residues implicated in folding or translocation, the importance of which was shown here. We expect that our results may help future computational design of RbsB and PBPs more in general, by improving the scoring functions used in the force field calculations.

Methods
Strains and growth conditions. All E. coli strains used for this work are listed in Supplementary Table 1. For cloning purposes, E. coli strains were cultured at 37 °C on Luria Bertani (LB) medium 30 , supplemented with appropriate antibiotics to select for plasmid maintenance. In case of ampicillin (Amp), a concentration of 100 µg ml −1 was used; for chloramphenicol (Cm), we used 30 µg ml −1 . Culturing conditions for protein overexpression and for reporter assays are specified below.
Library construction. The mutant library was produced by DNA synthesis (DNA2.0, CA, USA) on the basis of the sequence provided in Fig. S4. The gene sequence further encoded a C-terminal hexahistidine (His 6 ) tag framed by MfeI sites, restriction sites for NdeI (N-terminal), SalI (C-terminal), and for NcoI at the end of the signal sequence (Fig. S4). In first instance, the pooled DNA synthesis product was digested with NdeI and SalI and ligated into vector pSTV28P AA mcs, cut with the same enzymes (Fig. S5). This reporter construct library was cotransformed with pSYK1 into E. coli BW25113ΔrbsB 16 . The transformants were plated and 800 single colonies were picked, verified by sequencing and organized in a 96-well format. This screening covered 74% of the mutants in the library. Further mutants were then recovered individually from the original synthesized templates by PCR, which were cloned into pSTV28P AA mcs, cotransformed, verified and stocked as before. A total of 17 mutants which could not be recovered from the synthesized mutant library were produced individually using the DpnI method 31 . Plasmid pAR3 (with wild-type rbsB in pSTV28P AA mcs; ref. 16) was amplified in the PCR using overlapping but reverse complementary primers with point mutations to create Ala-replacements at the desired codons in RbsB. The PCR product was then digested by DpnI at 37 °C for 1 h to remove exclusively the (methylated) template plasmid DNA. After inactivation of the enzyme at 80 °C for 20 min the PCR product was transformed in E. coli DH5a and transformants were selected on LB with Cm. Ten transformants were verified by sequencing of the re-amplified rbsB-insert for the presence of the intended mutation in rbsB. If correct, the insert was recovered by restriction enzyme digestion and ligated to pSTV28P AA mcs. Those constructs were again verified by sequencing and if correct, the mutant plasmids were cotransformed into E. coli BW25113ΔrbsB together with pSYK1 as before. , residues identified here that affect protein abundance and signaling; magenta (V106, P262, D264), residues affecting signaling; gold (I233), residue affecting protein abundance only; orange (E255, S103 -just below P262), Ala-substitutions causing more constitutive signaling; green (A44, A70, N73, I111, L107), residues affecting chemotactic receptor and ribose transport channel binding 23 ; tan (N12, K45, D67), residues affecting ribose transport 23 ; iceblue (D52, G134, R166), residues affecting contact to ribose transport channel 20 . Image panels produced using VMD visual molecular dynamics editor 32 .
Scientific RepoRTS | 7: 8245 | DOI:10.1038/s41598-017-08035-5 RBP-based bioreporter assays using the Trz1-OmpR hybrid signaling chain. In order to measure the capacity of RbsB or its alanine-mutants to induce the Trz1-hybrid-OmpR ompCp-gfpmut2 signaling chain in the presence of ribose, we used E. coli BW25113ΔrbsB cotransformed with pSTV-based plasmids (pAR3-K1A -pAR3-L272A for the alanine mutants, or pAR3, for wild-type rbsB) and plasmid pSYK1 (to provide the hybrid signaling chain) 16 . Upon induction, these strains produce GFPmut2, the fluorescence intensity of which was measured using flow cytometry. The bioreporter assay was optimized for minimal background GFPmut2 expression and medium fluorescence, according to ref. 16. Strains were inoculated in 96 well plates (F96 Cert.Maxisorp, Nunc, Denmark) containing 200 µL of minimal medium with Amp and Cm 16 and 20 mM fumarate as sole carbon and energy source, with cells from organized stocks stored at −80 °C. Cultures were incubated overnight at 37 °C with rotary shaking at 700 rpm. The next morning, a volume of 1 µL of each culture was transferred into 200 µL of fresh minimal medium with fumarate in a new 96 well plate 16 , and incubated for 2 h at 37 °C with rotary shaking at 700 rpm. An aliquot of 90 µL of each culture was then removed and mixed with 10 µL of ribose solution (10 mM, D-(-)-ribose, Aldrich, USA) in the wells of a new sterile 96-well plate to reach a final ribose concentration of 1 mM. The remaining cell volume (110 µl) in the wells of the previous (uninduced assay) plate were filled with 100 µL of PBS to dilute the cell concentration and measured immediately. The ribose containing assay (induced) was incubated for another 2 h at 37 °C with rotary shaking and then sampled. Assay plate wells were auto-sampled, and values of individual cell forward scatter (FSC) and GFPmut2 fluorescence (FITC-channel) were recorded by a Becton Dickinson Fortessa flow cytometer (LRS FortessaTM, Becton Dickinson, USA). The flow rate was set to 3 µl s −1 and the cell density was between 100-1000 cells µL −1 . Sensitivities for the FSC and the FITC channels were set at 350 V and 676 V, respectively. Recorded data were gated to remove background particles. The mean fluorescence values of the gated uninduced or induced populations were calculated. All experiments were carried out in independent triplicate incubations on the 96-well plate. In all plates, triplicate incubations of E. coli strain 4175 (carrying wild-type rbsB on pAR3 plus pSYK1) and E. coli strain 4497 (with empty vector pSTV28P AA mcs plus pSYK1) served as simultaneous positive and negative controls for ribose induction, respectively. Reported fluorescence values are not normalized for plate-to-plate variations.
Periplasmic space RbsB abundance. The abundance of 25 alanine mutants in the E. coli periplasmic space was analyzed using direct peptide mass identification, as described previously 16 . After extraction and separation of the periplasmic protein fraction by SDS-PAGE, proteins in a size region of 28-36 kD were excised from the gel. Subsequently they were digested with trypsin and peptides were separated on an Ultimate 3000 Nano LC System (Dionex), followed by detection in a Thermo Scientific LTQ-Orbitrap XL mass spectrometer (Thermo Fisher Scientific, Waltham, MA). Mass spectra were analyzed by Scaffold Viewer (http://www.proteomesoftware. com/), using protein and peptide identification thresholds of 99.9% and 99.99%, respectively. The minimum number of peptides for identification was 1.
Crystal structure mutant positioning. Positions of selected Ala-substitutions were visualized on the