A bacterial two-hybrid analysis was carried out to shed light on the possible interaction between RibT and the two partner proteins of the heavy RF synthase complex, RibE and RibH. The hypothesis was that the acetyltransferase RibT interacts with these two proteins and transfers acetyl groups. Plasmid constructs containing the coding regions of each of the full-length enzymes RibT, RibE, and RibH were fused to the coding regions of the N- and C-terminus of the T18 fragment of adenylate cyclase. Similarly, plasmid constructs containing the same coding regions were also fused to the coding regions of the N- and C-terminus of the T25 fragment of adenylate cyclase. E. coli strains were transformed with two different plasmids, and all possible combinations between the Rib enzymes were tested. An interaction between the Rib enzymes would bring together the T18 and T25 fragments of adenylate cyclase-generating cAMP, which, in turn, would lead to expression of the lacZ reporter gene with its activity measured qualitatively on plates. Figure 1 shows one result from three different experiments (other two shown in Supplementary Figure S2).

As already known, we found interactions between RibH and RibE (Figure 1). RibE is known to exist as trimer, formed by interactions of C-terminal helices from each monomer, and occurs either solely or encapsulated in the cytoplasm (Bacher et al., 1980; Rotter et al., 2021). Furthermore, the C-terminus has been described to contain a localization tag required for encapsulation by RibH (Han and Woycechowsky, 2017). Interactions of RibH with RibE could be observed for three of eight possible cases (T18-RibH with RibE-T25, RibH-T18 with T25-RibE, as well as with RibE-T25).

Intriguingly, we also found interactions between RibE and RibT, as well as between RibH and RibT (Figure 1). Moreover, we found interaction between RibT itself, indicating possible dimer formation. Similarly, all fusions to RibH resulted in self-interaction. As expected, not all C- and N-terminal combinations of other fusion proteins revealed interactions: in case for self-interaction of RibT, we found interaction for the N-terminal fusion T25-RibT with the C-terminal fusion RibT-T18 only, but no interaction was detected for the N-terminal fusion T25-RibT with the other N-terminal fusion protein T18-RibT. This indicates either a spatial hindrance by the fused domains, shielding the monomers from self-interaction or by a distance that is too large between the split adenlylate cyclase, preventing generation of an active enzyme. Similarly, all other C-terminal fusion of T25-RibT completely abolished interaction with every protein shown, whereas the N-terminal fusion T25-RibT showed interactions for three cases (T18-RibT, T18-RibH, and RibT-T18), highlighting the importance of the C-terminus of RibT for the interactions found. It is of note that we did not find any interactions between RibT domain fusions and any Rib enzymes (RibAB or RibDG) other than RibE or RibH (results not shown), supporting that RibT acts specifically on RibH and RibE.

Knowing the interacting enzymes of RibT, we were interested in its localization and diffusion in B. subtilis using SPT under RF-producing conditions. For this purpose, we created a construct encoding for a C-terminal mVenus (mV) fluorescent protein fusion to the ribT gene, which was subsequently integrated into the original locus of the chromosome, such that its expression is driven from its native promotor(s). The presence of full-length RibT-mV has been verified by In-Gel fluorescence detection from cell lysates of the respective strain (Supplementary Figure S3). In order to study if interactions between RibT and RibH with RibE affect RibT-mV dynamics, we created mutant strains in which either the ribH or the ribE gene is substituted by an antibiotic resistance cassette (Koo et al., 2017). RF auxotrophy was addressed by the supplementation of RF to culture media for those two deletion strains. We first analyzed SPT datasets for all three strains by ensemble-averaged mean squared displacement (EAMSD). We found that overall average diffusion of RibT-mV (0.21 μm²/s) was increased by the deletion of either ribH (0.34 μm²/s) or ribE (0.4 μm²/s) (Figure 2A; Table 1), indicating that disruptions of interactions increase the number of freely diffusing RibT molecules. To determine potential diffusive subpopulations and their respective diffusion coefficients, we used a multiple Rayleigh Mixture Model-based approach for analysis of Jump distances (JD), derived from SPT trajectories. This way, we found that RibT-mV shows two diffusive states, a freely diffusive state characterized by diffusion in the range of 1.1–1.35 μm²/s and more than ten-fold lower diffusive states in the range of ~0.07-0.09 μm²/s (Figure 2B), indicating motion together with a large complex. Figure 2C shows that a two-population fit explains the observed data very well. In wt-like cells, the distribution of the two populations was almost equal, with 51% diffusing with 1.1 μm²/s and 49% diffusing with 0.08 μm²/s (Figure 2D; Table 1). However, deletion of either ribE or ribH leads to a shift in JD distributions toward increased diffusion (Figure 2B). The ratio of subpopulations relative to the wt-like condition changed such that, in case of deletion of ribE, the high-diffusing subpopulation increased to 83% (accounting for a relative increase of ~63%), or to 69% (accounting for a relative increase of ~35%) in the ribH deletion (Figure 2D). Accordingly, the low-diffusing subpopulation strongly decreased to 17% (diffusing with ~0.07 μm²/s) in ribE mutant cells, whereas deletion of ribH leads to a decrease in 31% in subpopulation size (diffusing with ~0.09 μm²/s), which accounts for relative decreases of ~65 and 37%, respectively (Figure 2D; Table 1). Note that the ribT gene has its own promoter, and fluorescence signal intensity of RibT-mV was not reduced in mutant cells.

To study localization of RibT-mV, we created spot location heat-maps derived from all trajectories and projected those into normalized cells, resembling B. subtilis cells in the exponential growth phase (Figure 2E). We found that RibT-mV is distributed in the entire cytoplasm, which represents a tube-like structure (Figure 2E). When free and confined (molecules staying within a confined area for more than a predefined number of steps) motion was analyzed separately, we found that free diffusion of RibT-mV is found throughout the center of the cell, whereas confined diffusion is mostly found toward the periphery of the cytoplasm, including polar regions (Figure 2G). The area of confinement was determined according to the localization error of each dataset (Supplementary Figure S4; Table 3).

Deletion of ribH or of ribE leads to increased diffusion over central regions, as indicated by our speed-map analysis (Figure 2F). Comparing the relative amounts of static trajectories that are confined (at least three time intervals) revealed that deletion of ribE leads to a decrease in ~32%, from ~41% in wt-like cells, whereas the deletion of ribH had smaller impact on the percentage of static trajectories (~38%, Supplementary Table S1). For longer times of molecule confinement (six and nine time intervals), this trend continued, showing ~20% (~9% for nine time intervals) static trajectories of RibT-mV in wt cells, whereas deletion of ribE leads to a decrease to ~10% (~4% for nine time intervals), and deletion of ribH to a decrease in ~15% (~7% for nine time intervals) of static tracks. Because strong confinement leads to subdiffusion, we analyzed confined trajectories separately by pooling them for EAMSD analysis, including statistical significance tests for their mode of diffusion (Supplementary Figure S5). Using six time lags, we found that subdiffusion is only evident for the wt-like strain having no deletions, whereas deletion of either ribE or ribH leads to entirely Brownian motion (Supplementary Figure S5A). When nine time lags were considered for analysis, we found that deletion of ribE changed the mode of diffusion for RibT-mV from subdiffusive to Brownian, whereas deletion of ribH or wt-like conditions revealed subdiffusion, as well (Supplementary Figure S5B). Thus, the deletion of ribE and of ribH affects diffusion of RibT-mV. Overall, our results from SPT suggest that RibT-mV binds both partner proteins of the heavy RF synthase in vivo.

Because RibT-mV SPT dynamics are affected by deletion of the interacting gene products, we next examined how RibH-mV motion is affected by the absence or higher abundance of RibT. When we analyzed a non-fluorescent strain in which ribT was deleted, we could only find a very small effect on exponential growth using minimal media (Supplementary Figure S6). This finding indicates that the deletion of ribT does not affect RF yield such that this becomes severely growth limiting. Nevertheless, we measured a decrease in RF yield upon deletion of ribT (Supplementary Figure S6), indicating that the gene product optimizes RF biosynthesis in vivo.

When we compared EAMSD plots for RibH-mV from wt-like cells (DC of ~0.34 μm²/s), we found little change in comparison to cells having ribT deleted (DC of ~0.38 μm²/s) (Figure 3A). However, induced overexpression of ribT from an ectopic locus resulted in an overall increased diffusion (~0.43 μm²/s), and thus higher mobility of RibH-mV. In order to analyze datasets at the level of diffusive subpopulations, we further compared them by JD analysis. In case for the ribT deletion, we found three existing diffusive subpopulations for RibH-mV resembling those of the wt-like strain (Figure 3B). Here, the triple fit yielded better results than the double fit (Figure 3C). As suggested before (Rotter et al., 2021), the medium-mobile subpopulation best agrees with comprising capsids of RibH-mV, and was increased by ~11% in the absence of RibT, relative to the wt cells (Figure 3D). Accordingly, the fast-mobile subpopulation thought to consist of RibH-mV pentamers (Figure 3E), diffusing with ~0.87 μm²/s −1.05 μm²/s, is relatively decreased by ~13% for the ribT deletion strain. Furthermore, the lowest diffusing subpopulation is also decreased (8% subpopulation size) for the ribT deletion strain in comparison to the wt-like strain (12% subpopulation size) (Figure 3D), which would account for a relative decrease in the confined subpopulation of 33%. Strikingly, we found no such low diffusive subpopulation for RibH-mV when ribT was expressed ectopically (Figure 3D). However, we found a large increase in the subpopulation thought to represent pentamers (62% subpopulation size), diffusing with 0.87 μm²/s, relative to wt-like cells, and a respective decrease in the fast-mobile subpopulation (38% subpopulation size), diffusing with 0.35 μm²/s. The change of the putative pentamer fraction is, therefore, quite striking in comparison to wt cells with a relative increase of 169.5%, whereas the change of the capsid subpopulation accounts for a relative decrease of 41.5%. These results indicate that RibT changes dynamics of the heavy RF synthase quite remarkably. Changes in molecule speeds can be seen in Figure 3H. RibH-mV is accumulated at the cell poles and in the cell middle (representing the future cell poles), and thus excluded from nucleoids (Rotter et al., 2021). This localization pattern is less pronounced in the absence of RibT and lost upon overproduction of RibT (Figure 3G, distribution quantified in Figure 3F). This effect was more evident when we further differentiated trajectories according to their behavior being classified as confined or freely diffusive (Supplementary Figure S7); confinement at polar regions is less pronounced for shorter events (three steps) and more so for six to nine steps.

Our findings can be best explained by an increase of assembled capsids and a decrease in pentamers by the deletion of ribT. Furthermore, the subpopulation considered as capsids showed decreased polar confinement for shorter times (three steps) compared to wt-like conditions, while the deletion of ribT had a noticeable positive effect on capsid abundance as judged from a relative increase of ~11% in comparison to wt-like cells. Thus, our findings indicate a role for RibT in capsid assembly and, more precisely, in increasing disassembly toward pentamers.

Residue K29 has been shown to be a target for acetylation (Carabetta et al., 2019), but the responsible acetyl transferase has not been identified. Therefore, we aimed at analyzing if K29 of RibH is a substrate for acetylation carried out by RibT. We employed a fluorescence-based detection assay in order to show the activity toward RibH. The assay is based on the detection of a fluorescence adduct that is formed upon reaction with free CoA in solution. Thus, the amount of free (fluorescent) CoA is proportional to the number of acetylated amine groups, and, thereby, the activity of any acetyl transferase utilizing AcCoA as a donor molecule can be measured indirectly by fluorescence readout at the appropriate wavelength (RibH^K29-NH3+ + RibT + Ac-S-CoA < -> RibH^K29-NHAc + CoA-SH + H⁺+RibT, CoA-SH + dye -> fluorescent product).

We expressed RibH and mutant RibH^K29E, which cannot be acetylated, as N-terminal Strep-tagged versions, as well as RibT as a C-terminal His-tagged protein (RibT-His₆), using E. coli as heterologous system. Purification was done by two-step purification using Strep-tactin and size exclusion chromatography (SEC), which is shown for Strep-RibH as an example in Supplementary Figure S8. Similarly, we also purified RibT-His₆ by Ni-NTA affinity chromatography and SEC (Supplementary Figure S9). In order to probe the importance of C112 for catalysis, we also created and tested activity of RibT^C112A as a His₆-tagged version (RibT^C112A-His₆) toward RibH. As a control to prove the specificity of RibT toward K29 of RibH as a substrate, we made use of a mutant from RibH in which K29 is substituted by E, which cannot be acetylated due to the missing ε-amine group. Additionally, we applied bovine serum albumin (BSA) in similar concentrations as negative acceptor control.

In order to show qualitatively that acetylation occurs using RibH and RibT, we performed endpoint assays with varying concentrations (25, 50, and 100 nM) of RibT-His₆ or mutant RibT^C112A-His₆ and a fixed concentration of acceptor protein Strep-RibH (10 μM; Figure 4A). We started the endpoint assay by addition of the GNAT to reaction mixtures and incubated reaction mixtures for 30 min at room temperature before these were quenched, and the fluorescence was read out at a microplate reader. We found that fluorescence increased with increasing GNAT concentrations, indicating GNAT activity of RibT-His₆ toward Strep-RibH (Figure 4A). We also found activity toward Strep-RibH for the mutant RibT^C112A-His₆, however, much lower than for the wt enzyme; A two-fold decrease in GNAT activity is very likely the result of an increased dissociation constant for the substrate AcCoA (as also shown by Srivastava et al., 2018).

Next, we compared kinetics of acetylation using a kinetic endpoint assay for RibT-His₆ and RibT^C112A-His₆, applying 10 μM of Strep-RibH as an acceptor. However, when using mutant RibT^C112A-His₆ with the same concentration (100 nM) as the non-mutated enzyme, we did not observe any detectable activity (results not shown). Therefore, we increased the concentration of the mutant GNAT by two-fold (200 nM) to ensure that the fluorescence signals derived from the reactions with free CoA do not fall below the detection range. We found that mutant RibT^C112A-His₆ has activity toward Strep-RibH, as well, but to a much lower extent than the wt enzyme (Figure 4B).

From the kinetic end point assay, we can already conclude that RibH is, indeed, the acceptor molecule for acetylation by the GNAT RibT. Nevertheless, to prove its specificity for residue K29 in RibH, we made use of the mutant RibH^K29E, which cannot be acetylated at this particular residue. Thus, we performed the kinetic end point assay using mutant Strep-RibH^K29E, Strep-RibH, and BSA for two different concentration magnitudes (~10-123 μM) of acceptor probes using a fixed GNAT concentration of 100 nM. For the kinetic end point assay in which we used the lower acceptor concentration range, we found increasing fluorescence over time in case for Strep-RibH in linear dependency for the first four time increments (Figure 5A), underlining GNAT activity toward Strep-RibH. The later time points reveal a plateau, indicating that acetylation of Strep-RibH is completed after 40 min at this concentration. Contrarily, for mutant Strep-RibH^K29E, which lacks the acceptor amino acid for acetylation, we observe a decrease in fluorescence over all time points, instead of an expected fluorescence distributed around zero (Figure 5A). Possibly, relatively high fluorescence values of the measured blank values lead to false-negative fluorescence after their subtraction.

At higher concentrations of acceptor substrates, and keeping GNAT concentrations constant at 100 nM, relative fluorescence of Strep-RibH^K29E did not increase over time, but, as expected for this mutant, was distributed around values of zero. In contrast, Strep-RibH showed a clearly increased relative fluorescence (~5- × -10³ RFU) for the first five time increments (Figure 5B). For the negative control BSA, we observed negative values for fluorescence (Figure 5B), because measured blank values were higher than analyzed samples. Similar to RibH^K29E, fluorescence for BSA remained almost constant over time, strongly indicating no activity of RibT-His₆ toward BSA. Thus, it is evident that Strep-RibH showed increased fluorescence in comparison to its mutant Strep-RibH^K29E, or to BSA, indicating that RibH is, indeed, acetylated by RibT at K29.

Because we found that changes in RibT levels have an effect on diffusive subpopulations of RibH-mV, we anticipated that the interaction between RibH and RibT, resulting in an acetylation event of RibH at residue K29, might change RibH dynamics. Thus, we visualized two mutants of RibH not capable of acetylation at the suspected residue (K29E and K29R); we hypothesized that mutation from K to R leads to mimicry of K that cannot be acetylated. We speculated that this change should result in a phenotype observed by SPT that resembles the phenotype of a ribT deletion strain, as presented before (Figure 3). On the other hand, mutant RibH^K29E-mV introduces negative charges into the three-fold symmetry axis of the 60-meric capsids (Supplementary Figure S1), and might form less stable capsids due to the repulsive clash of the three negative charges. Overall DCs of mutants RibH^K29R-mV (~0.34 μm²/s) and RibH^K29E-mV (~0.34 μm²/s) by EAMSD analysis were found to be similar to RibH-mV (~0.34 μm²/s) (Table 1). However, analysis of JDs by multiple Rayleigh Mixture models resulted in three different subpopulations (Figures 6B,C), and showed that all three strains differed in their relative RibH proportions (Figure 6A). RibH^K29R-mV revealed a relative increase of ~8% in the medium-mobile subpopulation considered as capsids (70% subpopulation size) in comparison to RibH-mV (65% subpopulation size), accompanied by a relative decrease of 24% in subpopulation considered as pentamers (19% subpopulation size) in comparison to the non-mutated strain (25% subpopulation size) (Figure 6C). Although these changes are small, they are significant because they are based on a very large data set. Comparison of trajectory localizations and visualization of diffusions using normalized cell representations for the acetylation mutants revealed that NO seen for the K29R mutant is comparable to that of the wt-like enzyme, and that diffusion is also distributed in a similar manner over the cytoplasm, being higher over central, nucleoid regions and lower toward the cytoplasmic periphery (Supplementary Figures S10A,B). Contrarily, for the K29E, mutant hotspots of trajectory localizations are visible in heat-maps and coincide with spaces of low diffusion, indicating that the K29E mutation leads to an altered localization (Supplementary Figures S10A,B).

When comparing subcellular localizations for trajectories classified as confined or freely diffusive (Figure 6D), the RibH^K29R-mV mutant showed very similar distributions for both classes of trajectories compared to the wt-like enzyme, retaining a pattern of nucleoid occlusion (NO). Contrarily, the medium-mobile population decreased for RibH^K29E-mV, while the high-mobile fraction increased (Figure 6C). Importantly, NO was lost for the mutant protein, best seen for six and nine steps of confined motion (Figure 6D). Thus, mutation of K29 leads to small but noticeable changes in single particle dynamics of RibH, possibly reflecting altered dissociation rates of pentamers from the assembled capsids. For the small low-mobile fraction observed (Figures 3D, 6C), we have proposed that these represent ribosome-associated assembly structures for capsids (Rotter et al., 2021). This fraction is not affected in its size by mutating K29 (Figure 6C).

All bacterial strains used in this study are listed (Table 2). Bacterial strains were routinely grown using liquid lysogeny broth (LB) media (10 g/l tryptone, 5 g/l yeast extract, 10 g/l sodium chloride, pH 7.5) or solid LB-agar plates (liquid LB, 15 g/l agar). Molecular cloning of plasmids and their propagation was achieved using E. coli strain XL-1 Blue (Stratagene) as a host strain. Strains in this study were routinely cultivated in liquid LB media at 37°C under constant shaking (200 rpm), with addition of the respective antibiotics (Supplementary Table S3), ensuring selective pressure. Plasmid Mini Kit (Qiagen) was used for preparation of plasmids according to the manufacturer's instructions. Preparation and heat-shock transformation of competent E. coli strains were achieved by standard protocols (Sambrook and Russell, 2006a).

B. subtilis strains created in course of this study were derived from the prototrophic laboratory strain PY79, the auxotrophic wt strain 168 or its respective deletion mutants ordered from the Bacillus Genetic stock center (BGSC) (Koo et al., 2017). Strains were routinely grown in liquid LB media at 30°C (200 rpm) for overnight culturing, with addition of the respective antibiotics. For SPT microscopy, overnight cultures were diluted 75-fold in glass tubes with 1 ml of freshly prepared S7₅₀ minimal growth media, containing glucose as a carbon source. In case of inducible expression by xylose using the xylose promoter (P_Xyl), glucose was quantitatively substituted for the same amount of fracture in the growth medium, and expression of P_Xyl was induced with 0.1% (w/v) of xylose. Cultures were grown under constant shaking at 30°C till reaching an exponential growth phase (OD₆₀₀:0.4-0.6) until samples of live cells were taken for SPT microscopy experiments. S7₅₀ minimal media was freshly prepared according to published procedures (Jaacks et al., 1989).

All plasmid constructs in this study with respect to integration at the original gene loci were derived from pSG1164-linker-mVenus (Rotter et al., 2021), which was created from the integrative single-crossover plasmid pSG1164 (Lewis and Marston, 1999). Integration at ectopic amy Elocus was achieved using pSG1193-linker-mVenus (Rotter et al., 2021), which was created from pSG1193 (Feucht and Lewis, 2001). To generate plasmids used to transform B. subtilis, we amplified the respective DNA sequences by PCR with dNTP (NEB) and Phusion High Fidelity DNA Polymerase (NEB) using genomic DNA of B. subtilis PY79 as a template, and using the corresponding oligonucleotides (Sigma-Aldrich) listed in Supplementary Table S5. PCR products were purified by gel extraction using QIAquick Gel Extraction Kit (Quiagen), successively digested with restriction enzymes AvrII and ApaI (NEB) at the specified optimal temperature, followed by heat inactivation of enzymes at 65°C for 20 min, and then purified using QIAquick spin columns (Qiagen). The plasmids pSG1164-linker-mVenus or pSG1193-linker-mVenus were digested accordingly and, furthermore, dephosphorylated by treatment with Calf intestine phosphatase (CIP; NEB), similarly purified as described before, subsequently ligated using T4 Ligase (NEB) with the respective linear DNA insert fragment in a 1:3 molar ratio, and subsequently transformed into E. coli XL-1 Blue, as described in Bacterial strains and growth conditions to yield the plasmids listed in Supplementary Table S4. Correct insert sizes were confirmed by analytical restriction digestion of the respective plasmid using XbaI (NEB), and correct insert sequences were verified by Sanger sequencing (Eurofins).

Deletion strains that were used in the course of this study have been created by transformation of B. subtilis PY79 wt with chromosomal donor DNA of the respective 168 deletion strain (Table 2). Those 168 strains were obtained from the BGSC (Koo et al., 2017), and their chromosomal DNA was prepared using the phenol-chloroform extraction method (see preparation of chromosomal DNA). Transformation was achieved using an aliquot of an overnight culture from the respective PY79 strain. Overnight cultures were grown at 30°C with shaking (250 rpm) in liquid LB media. Using an appropriate aliquot, we inoculated 10 ml of freshly prepared liquid modified competence media (MCM) in a 200-ml shaking flask to have an initial optical density of 0.08–0.1 measured at 600 nm (OD₆₀₀). MCM was prepared freshly according to published procedures (Zafra et al., 2012). The culture was grown in MCM at 37°C under constant shaking (200 rpm) till it reached the stationary growth phase indicated by an OD₆₀₀ of 1.4–1.6. For transformation of plasmid or chromosomal DNA, we used an aliquot of 1 ml and added a total of 1 μg of either plasmid or chromosomal DNA; as a control, we used 1 ml of the same culture without any addition of DNA. Each culture was further incubated for two more hours at 37°C with constant shaking, followed by streaking out different amounts of culture aliquots onto fleshly prepared, solid LB-agar plates containing the appropriate antibiotics (Supplementary Table S3). To allow for growth of RF-deficient strains, we had to add minimal amounts of RF (1 μg/ml, Sigma-Aldrich) to all growth media used for their cultivation. Following incubation for 1–2 days, the resulting transformed colonies were selected for Kanamycin resistance and the PY79 phenotype in case of transformations with chromosomal DNA from 168, stored, and further used for transformation of integrative pSG1164-linker-mVenus or pSG1193-linker-mVenus-derived plasmids (Supplementary Table S4).

Chromosomal DNA from B. subtilis 168 strains used as templates for PCR or for transformation purposes was prepared using the phenol-chloroform extraction method (Green and Sambrook, 2017a). Following extraction, high-molecular DNA was subjected to precipitation by 1-propanol for concentration (Green and Sambrook, 2017b), subsequently washed with 70% ethanol (w/v), air dried, and suspended in appropriate amounts of a TE buffer (100-mM TRIS-HCl, 10-mM EDTA, pH 8.0).

Integration of fusion enzyme coding sequences into the amyE locus of B. subtilis was probed using amylase assay. For this purpose, we streaked single colonies from a transformation onto single wells that have been prepared with solid LB agar media, containing 1% (w/v) of starch and the respective antibiotics, if needed. These plates were incubated overnight at 30°C to allow for colony growth. As a positive control, we used untransformed PY79 wt cells. The next day, wells were covered with Lugol's iodine and incubated at 30°C for 5-10 min to allow for visual screening of non-starch hydrolyzing colonies.

For analysis of interactions between RibH, RibT, and RibE, either one of the plasmids, e.g., pRibHT25 with the fused T25 domain (Supplementary Table S4), was used to transform the E. coli reporter strain BTH101 (Table 2), followed by secondary transformation of any of the T18 domain containing plasmids, e.g., pRibHT18 (Supplementary Table S4). A positive interaction will create a functioning adenylate cyclase that synthesizes cAMP and subsequently stimulates expression of lacZ in E. coli. Growing colonies were qualitatively analyzed for the development of blue color dependent on the reporter gene product β-galactosidase (LacZ) using an overlay assay with X-Gal staining. The control protein Zip (encoded on pKT25-zip or pUT18C-zip) was used as a negative control in all cases.

If not otherwise indicated, B. subtilis strains were aerobically grown at 37°C in a 2x Spizizen's minimal medium (Spizizen, 1958), supplemented with 0.02% casamino acids, 2% (w/v) yeast extract, and 10% (w/v) glucose or in liquid LB. B. subtilis 168 (Saito et al., 1979) is a wild-type strain with regard to riboflavin biosynthesis and uptake and was used as a control. Riboflavin levels in the supernatant of aerobic B. subtilis cultures in the time course of growth was monitored as follows: an aliquot from the culture containing cells and a medium (500 μl) was combined with 465-μl 4 N NaOH and vigorously mixed for 1 min. The mixture was neutralized by adding potassium phosphate (1 M, pH 8.0) and centrifuged for 5 min at 13,000 x g at room temperature. Riboflavin in the supernatant was determined by using a standard procedure employing HPLC (Mack et al., 1998). Using the above-described protocol, the total amount of riboflavin (intracellular riboflavin and riboflavin present in the fermentation broth) was measured. For the statistical analysis of the data, Student's T-test was applied, comparing two unknown means based on independent samples.

To detect fusion enzymes by in-Gel fluorescence, we created cell lysates of the respective strains and performed SDS-PAGE of those samples. For this purpose, B. subtilis strains were grown till reaching the exponential growth phase as already described (Bacterial strains and growth conditions). Cells have been repeatedly pelleted and washed in minimal media and subsequently stored as a frozen pellet. Next day, the pellets were suspended in a buffer (100-mM NaCl, 50°mM EDTA, pH 7.5) and DNAse; RNASe and lysozyme were added to yield final concentrations of 1 mg/ml. This suspension was incubated for 30 min at 37°C. An SDS-sample buffer was added to yield a ten-fold concentrated lysate. In order not to irreversibly quench mVenus-derived fluorescence for further imaging, the samples were not cooked as usually done. Those samples were finally loaded on 12% SDS-PAGE gel for electrophoretic separation for 1 h at 4°C using 50 mA and 120 V. The mVenus-derived fluorescence from fusion proteins was detected with a Typhoon TRIO Imager System (GE Healthcare) using excitation at 488 nm and optimal integration time.

SDS-PAGE was performed according to standard protocols (Sambrook and Russell, 2006b) using 15% polyacrylamide gels.

Strains of PY79 used for SPT were grown as described before. For imaging of single particles in live cells, 3 μl of cell culture suspension were spotted on clean coverslips (25 mm, Menzel) and covered using freshly prepared 1% (w/v) ultra-pure agarose pads. Those agarose pads were made by melting the appropriate amount of agarose in a fresh S7₅₀ minimal medium and by further sandwiching ~80-μl melted agarose solution between two small 12-mm coverslips (Marienfeld). To avoid drying, pads were freshly prepared and kept humid until further use. The coverslips used for microscopy were intensively cleaned by ultra-sonication in Hellmanex II solution (2% v/v) for 30 min and intensively rinsed in distilled water, followed by a second round of ultra-sonication in double-distilled water until they were dried using high-pressure airflow.

SPT experiments in this study were conducted using an Olympus IX-71. This microscope was equipped with a high numerical aperture (NA) Total internal reflection (TIRF) objective (UAPON × 100, Oil, NA = 1.49), which was coupled to a light-emitting diode laser LuxX 515-100 (515 nm/100 mW) to achieve excitation of fluorophores, which were detected using a back-illuminated Electron Multiplying Charge Coupled Device (EMCCD) iXon Ultra camera (Andor Solis). The laser beam was focused onto the back focal plane and operated during image acquisition with up to 2 mW (60 W/cm² at the image plane). Using Andor Solis, 4.21 stream acquisitions of 1,500 frames having 13.76-ms interval times (12 ms integration time) were acquired from live cell samples. Those streams were equally cropped according to photobleaching curves to yield 1,000 frames, subsequently adjusted for pixel sizes of 100 nm and time increments using Fiji (Schindelin et al., 2012). Tracking of single particles was done using u-track 2.2 (Jaqaman et al., 2008). Trajectories were considered only for further analysis if they had a length of at least five steps. Data analyses were done using SMTracker 2.0 (Oviedo-Bocanegra et al., 2021).

Recombinant protein production was achieved using E. coli strain BL21 (DE3) Star (Table 2) as a host for transformation of pASK-IBA7-, pET-28a-, pET-Duet-1- or pCDF-Duet-1-derived plasmids (Supplementary Table S4). Cloning of plasmids and cryoconservation of the respective strains were achieved using E. coli strain XL-1 Blue as a host (Table 2). Correct insert sizes of all plasmids were verified by analytical restriction digestion, and correct sequences were verified using sanger sequencing (Eurofins). For a detailed description of all clonings and purification steps, please see Supplementary Data.

Purified, monodisperse proteins (Strep-RibH and mutant Strep RibH^K29E) were probed as targets for enzymatically catalyzed acetylation by purified RibT-His₆ using a fluorescence-based assay specific for Acetyl-CoA (AcCoA)-dependent acetyl transferases (ENZO Life sciences) in combination with a fluorescence microplate reader (TECAN Infinite 200). The fluorescence readout at 520 nm relies on excitation at 380 nm of a fluorescent adduct that is formed upon reaction with free CoA when detection solution was added. The reaction mixtures (containing AcCoA) were set up in black 96-well plates in duplicate according to the manufacturer's instructions using different concentrations (~10 μM, ~100 μM) of probed acceptor substrates (Strep-RibH or Strep-RibH^K29E) and varying concentrations (25, 50, 100, 200 nM) of acetyl transferase RibT-His₆ (or mutant RibT^C112A-His₆) for endpoint assays. Blank samples have been set up using only the respective acceptor substrates in a buffer. In case of kinetic endpoint assay, fixed concentrations of ~10 μM or ~100 μM acceptor substrate and 100 nM of acetyl transferase have been used. Chloramphenicol acetyl transferase (CAT) supplied by the manufacturer was used as positive control for acetylation of chloramphenicol in order to qualify the assay. As a negative control, we used bovine serum albumin (BSA) and RibT-His₆ in kinetic endpoint assays. Following incubation at room temperature with gentle orbital shaking (100 rpm) for different time intervals (a total of 60 min with 10 min or 20 min time intervals), reactions were quenched by addition of ice-cold 1-propanol. After all reactions were quenched, detection solution was equally added, followed by development for 10 min at room temperature, avoiding incident light and shaking. Finally, the samples were excited using 380 nm, and fluorescence was detected at 520 nm. Data were recorded using 20-μs integration time and optimal gain.

For imaging of purified Strep-RibH capsids and the respective mutants by transmission electron microscopy (TEM), we applied carbon-coated grids (400 mesh), which were lyophilized by glow discharging (PELCO easiGlow, Ted Pella, USA). Staining was done using 2% (w/v) uranyl acetate, which was spotted onto the grids together with 1 μl of 100-μg/ml protein solution in a buffer (50-mM sodium-phosphate, 150-mM NaCl, pH 7.2), followed by a short washing step using ultrapure H₂O. Those samples were analyzed with a JEOL JEM-2100 transmission electron microscope using an acceleration voltage of 120 kV. For image acquisition, an F214 FastScan CCD camera (TVIPS, Gauting) was used. Image analysis was done using ImageJ (Schindelin et al., 2012). For the determination of capsid radii, well contrasted capsids have been selected to draw circular regions of interest, largely avoiding black contrast outlines. Radii have been calculated from the measured area using Excel 2016 (Microsoft), and values have been statistically analyzed using BoxPlotR (Spitzer et al., 2014).

To account for differences in localization precision between compared SPT datasets, we calculated estimates for the average localization errors using the intersection point of the ordinate axis and the linear fit for five lag times in individual TAMSD plots for the respective SPT datasets. The average localization errors determined that way are listed for each SPT dataset used in this study (Table 3).

Based on the radii of confinement (Table 3), we have developed trajectory heat maps, displaying trajectories of an SPT dataset that have been classified either as confined or freely diffusive in a normalized cell representation (1 μm × 3 μm), resembling a typical cell of B. subtilis in the mid-exponential growth phase. A trajectory is considered as confined if it has at least one dwell event for a given number of steps, which we fixed to be either three, six, or nine. This confinement can be total (a confined track for N steps), or absent (freely diffusive for N steps), leading to their classification.