Bacillus subtilis 168 encodes a pspA (BSU_06180) homolog under control of a σ^w-dependent promoter. Downstream of pspA, three structural genes, ydjGHI are positioned (Figure 1A). This operon is likely controlled via the promoter in front of pspA, however there is a Rho-independent terminator between pspA and ydjI. YdjI is a member of the SPFH-domain proteins. It shares the SPFH-domain with other bacterial flotillins, including FloA and FloT from B. subtilis (Figure 1B). Topology prediction analysis of YdjI suggests that the protein lacks a hydrophobic domain that could serve as a membrane anchor, consequently rendering YdjI soluble (Figure 1B). However, the two genes upstream of ydjI, ydjG, and ydjH encode for proteins that contain transmembrane helices and they are likely membrane integral proteins (Figure 1C). YdjI contains a low complexity region and a Zink-finger motif close to its C-terminus.

To study the subcellular localization of the B. subtilis YdjI protein, we constructed a translational YdjI-mNeonGreen fusion, replacing the native gene locus and, hence, expression of ydjI-mNeonGreen was still under native regulation. Furthermore, we used a PspA-GFP fusion to visualize PspA localization in vivo. PspA and YdjI were found to localize in discrete foci at the membrane of B. subtilis (Figures 2A,B). PspA expression was clearly enhanced under alkaline shock (Supplementary Figure 1), as expected since the promotor is known to be strongly induced by the ECF sigma factor σ^W (Huang et al., 1998; Wiegert et al., 2001). Western blotting or in-gel fluorescence revealed that PspA and YdjI fusion proteins were full length and only little degradation was observed (Supplementary Figure 1). While the cellular concentration of YdjI did not differ significantly under non-induced and stress induced conditions (stress was applied by addition of 15 mM NaOH), the concentration of PspA was greatly enhanced upon alkaline shock (Supplementary Figures 1A–C). These findings are in line with the presence of a Rho-independent terminator between pspA and ydjG (Figure 1A) leading to a substantial difference in the protein concentrations of PspA and YdjGHI after alkaline shock. The majority of YdjI was membrane associated, while the majority of PspA was found in the cytosol and not membrane associated (Supplementary Figures 1A–C). Although the PspA concentration was increased in the cell after alkaline shock, cell fractionation experiments show that the ratio of membrane associated and cytosolic PspA remained the same (Supplementary Figure 1C). Thus, PspA recruitment to the membrane was not increased after alkaline shock. The addition of 15 mM NaOH was sublethal for all individual deletion strains of pspA and the ydj genes tested here. In all cases addition of sodium hydroxide only marginally decreased growth rates compared to the unstressed control (Supplementary Figure 2). We conclude that alkaline shock with 15 mM NaOH does induce the pspA-ydjFGHI operon, but has no deleterious effect on the cells that might compromise localization studies.

In a next attempt we analyzed potential co-localization of PspA and YdjI. Both proteins formed membrane associated foci under stress conditions induced by 15 mM NaOH shock. We constructed a strain expressing PspA-GFP and YdjI-mCherry2 under control of their native promoters (strain ASB013). Microscopic examination of the cells revealed that PspA and YdjI readily form foci upon alkaline shock. The vast majority of YdjI foci co-localized with PspA foci (Figures 3A–C). Cells contained in average more PspA foci (Figure 3D and Table 1) and therefore not all PspA foci co-localized with a YdjI focus. However, a detailed analysis of co-localization using Pearson’s correlation revealed a high degree of co-localization for both proteins (Figure 3B).

The subcellular localization pattern of YdjI and PspA was similar to the punctate membrane distribution of other bacterial SPFH-domain proteins, namely FloA and FloT (Donovan and Bramkamp, 2009; Lopez and Kolter, 2010). FloA and FloT were shown to localize into liquid ordered membrane regions (Bach and Bramkamp, 2013). We therefore wanted to address the preferred membrane region to which YdjI is recruited. To assess this, we performed in vivo co-localization studies of the YdjI-mNeonGreen fusion strain (strain ASB033) with Dil-C12, a well-known fluorescent probe for fluid membrane regions (Strahl et al., 2014). When B. subtilis cells were stained with Dil-C12, strictly punctuated fluorescent foci could be observed that overlapped significantly with the YdjI-mNeonGreen signal, with a Pearson’s correlation coefficient of 0.93 ± 0.022 (n = 23 cells) (Figures 4A,C). We also analyzed the co-localization pattern of FloT and YdjI by using a doubled tagged strain FloT-mNeonGreen; YdjI-mCherry2 (strain ASB154). In this strain both proteins were under control of their native promoter. YdjI did not co-localize with FloT in neither unstressed, or stressed conditions (Figures 4B,D). Fluorescence intensity correlation graphs for alkaline induced conditions in displaying a pixel by pixel correlation between Dil-C12 and YdjI-mNeonGreen (Figure 4C) or YdjI-mCherry2 and FloT-mNeonGreen (Figure 4D) revealed a clear anti-correlation between FloT and YdjI and a clear correlation between YdjI and Dil-C12.

Classical flotillins such as FloT can be enriched in so-called DRM extractions. We therefore aimed to compare the behavior of YdjI and FloT using DRM and DSM fractionation. Although, DRM/DSM fractionation is not suitable to make predictions about membrane domains in living cells, it allows the specific enrichment of proteins such as classical flotillins. Phase separation of DRM and DSM was achieved using a commercially available kit (see section “Materials and Methods”). Fractionation revealed that FloT is recovered in the DRM fraction as expected (Figure 4E). Although, part of the FloT protein was still found in the DSM fraction, a high proportion of the protein was recovered in the DRM fraction. On the contrary, YdjI was found exclusively in the DSM fraction (Figure 4E). This is an important difference, since currently, all SPFH-domain proteins have been associated with liquid ordered membrane regions and were thought to enrich in DRM fractions (Bramkamp and Lopez, 2015). In particular, the SPFH-domain itself was supposed to mediate contact and specificity to defined membrane regions. These data, together with the co-localization of YdjI with DilC-12 in vivo, confirm that YdjI behaves different compared to classical flotillins and likely resided in fluid membrane domains in the bacterial plasma membrane. Consequently, not all SPFH-domain proteins are associated with Lo domains.

While FloA and FloT bind autonomously to the membrane, the subcellular localization of YdjI might require other proteins from the PspA-YdjGHI operon for membrane association, as hypothesized by the protein domain prediction analysis of YdjG and YdjH (Figure 1). To test this, we used YdjI-mNeonGreen and PspA-GFP fusion strains that lacked the pspA and ydjI gene, respectively, and analyzed the in vivo localization of the tagged proteins. The results revealed that PspA was not recruited to the membrane when YdjI was absent, regardless if the cells were induced or not with alkali-shock (Figures 5A–D), showing a strong dependency and interaction between PspA and YdjI for membrane complex formation. Importantly, ectopic expression of YdjI fully complemented this phenotype (Supplementary Figure 3). Deletion of pspA led to an increase of YdjI expression, due to the removal of the Rho-independent terminator in this strain. In this case we observed YdjI focus formation and conclude that YdjI complex formation does not require PspA (Figure 5D).

Since YdjI is likely a soluble protein its membrane targeting should require an additional factor. We therefore reasoned that the two membrane proteins YdjG and YdjH that are encoded in the pspA-ydj operon are potential candidates for this membrane recruitment. Indeed, deletion of either ydjG or ydjH entirely abolished YdjI and PspA foci formation and membrane recruitment under normal growth and under alkali shock conditions (Figures 6A–D), hinting toward a direct interaction between the Ydj proteins. To test this hypothesis, we performed an extensive bacterial two hybrid interaction analysis. To this end, we fused the coding sequences of PspA, YdjI, YdjG, and YdjH to the split adenylate cyclase domains (Karimova et al., 1998). To minimize steric problems of the fusion proteins, we performed all possible combinations of N- and C-terminal fusions (Figure 6E). Using a blue-white screen on plate we observed a close interaction network between PspA and the Ydj proteins. PspA, YdjI and YdjH all showed self-interaction. We also detected a strong interaction between PspA and YdjI, corroborating the co-localization studies. Furthermore, we revealed an interaction of the two membrane components YdjG and YdjH with each other. Importantly, we also observed an interaction of YdjH with PspA and with YdjI, indicating that YdjH likely acts as the membrane anchor for the YdjI/PspA complex (Figure 6E).

In strains deleted for either ydjH or ydjG we observed a severe cell morphological defect under alkali shock conditions. While unstressed cells lacking YdjG or YdjH reveal a normal morphology, these cells showed curved cells and swollen or bulged cell poles upon NaOH addition (Figures 6A–D). These aberrant morphologies are reminiscent of certain mutations known from components involved in cell wall synthesis such as a gpsB mutation (Claessen et al., 2008) or mreBH mutations (Carballido-Lopez et al., 2006). Therefore, we used fluorescent staining of nascent peptidoglycan synthesis, in order to visualize localization of the cell wall synthesis machinery. Strains with clean deletions of either ydjG and ydjH were stained with HADA (Kuru et al., 2015) and cells were stressed with NaOH (15 mM). HADA localizations were often found at the bulging cell poles and in areas were the cell body seems bent and misshaped (Supplementary Figures 4A,B). In wild type B. subtilis cells the cell poles are usually inert and do not localize active cell wall synthetic complexes. Cells lacking YdjG show a more pronounced phenotype including occasional branching (Supplementary Figure 4A). We conclude that YdjG might couple cell envelope stress signaling and cell wall synthesis. In absence of these proteins, cell wall synthesis becomes deregulated under stress conditions. Interestingly, deletion of YdjI or PspA did not exhibit this strong morphological phenotype and, hence mis-localization of the PspA/YdjI complex might not be the (sole) reason for this cell shape defect in an YdjG/H mutant.

Anionic lipids and the cytoskeletal proteins MreB and RodZ define the spatio-temporal distribution and function of PspA in E. coli (Jovanovic et al., 2014). Based on these findings and the observed cell shape defect in ydjG/H mutants we wanted to analyze a potential involvement of the MreB cytoskeleton on PspA/YdjI complex formation. We constructed strains expressing YdjI-mNeonGreen and PspA-GFP in which we deleted the mreB gene (strains ASB155 and ASB123, respectively). In both strains we readily observed PspA and YdjI foci under alkali shock conditions, indicating that MreB is not essential for PspA or YdjI assembly (Figures 7A–D). However, a quantification of the PspA and YdjI foci per cell revealed that cells lacking MreB accumulate significant less foci compared to wild type cells (Table 1). While wild type cells had on average 1.6 PspA foci per cell, the mreB mutant strain had only 0.68 foci per cell. When cells were not stressed with sodium hydroxide we also observed a reduced number of PspA foci (0.21 foci in mreB mutant cells and 0.49 foci in wild type cells). Thus, absence of MreB leads to a 2.3-fold reduction of PspA foci formation under stressed and unstressed conditions. Similarly, the number of YdjI foci was reduced in absence of MreB in stressed and unstressed conditions (Table 1), lending support to the notion that PspA and YdjI foci formation is tightly connected. We next wanted to analyze whether PspA and YdjI foci are mobile in presence and absence of MreB. Time-lapse analysis revealed that in strains lacking mreB PspA and YdjI foci still retain their dynamic behavior (Supplementary Movies 1–8). Inspection of all movies revealed that many PspA and YdjI form foci that were mobile while others are rather static. A simple explanation might be that the static foci are acting on the membrane locally, maybe at damaged membrane areas. However, we did not observe a major difference in PspA or YdjI foci dynamics lacking MreB. We conclude that in B. subtilis, MreB is not essential for PspA and YdjI foci formation. However, deletion of MreB leads to altered foci numbers. We have shown that deletion of MreB led to a decrease in membrane fluidity (Figure 7E). A plausible explanation for the reduced foci number of PspA/YdjI might therefore a change in membrane fluidity.

We have shown before that deletion of mreB or flotillins has an effect on membrane fluidity. Membrane fluidity can be experimentally addressed by using the generalized polarization (GP) values after a Laurdan staining. Laurdan intercalates into the membrane bilayer and its fluorescent spectra are sensitive to polarity and dipole dynamics. This allows a correlation of measured GP values and lipid packing in the membrane. Using Laurdan GP analysis we showed that deletion of mreB reduced the overall GP, rendering the membrane slightly more fluid. In contrast deletion of YdjI significantly increased the overall GP value. Thus, similar to the situation observed for the deletion of the flotillins FloA and FloT, deletion of the SPFH domain protein YdjI led to a rigidification of the bacterial membrane (Figure 7E).

All B. subtilis strains used in this study are derived from strain 168 and are listed in Table 2. Construction of new strains was based on natural competence of B. subtilis (Harwood and Cutting, 1990). Gene integration or deletion was validated by colony PCR whereas the expression and localization of the fluorescent fusions were additionally validated by microscopy. Cells were grown in lysogeny broth (5 g/L yeast extract; 5 g/L NaCl; 10 g/L tryptone) at 37°C and 200 rpm. Induction of the pspA promoter was triggered by addition of 15 mM NaOH (Wiegert et al., 2001). Cell cultures were supplemented with spectinomycin (50 μg ml^–1), chloramphenicol (5 μg ml^–1), kanamycin (5 μg ml^–1), erythromycin (1 μg ml^–1), neomycin (10 μg ml^–1), magnesium chloride (MgCl₂, 25 mM), and/or sucrose (0.3 M) when necessary.

Plasmids were constructed using standard cloning techniques and Golden Gate Cloning Assembly (Engler et al., 2009). For DNA amplification via PCR, Phusion DNA polymerase was used. Restriction enzymes were purchased from New England Biolabs (NEB, Ipswich, MA, United States) and used accordingly with their respective protocols. E. coli colonies were checked by colony PCR, using EconoTaq polymerase (Lucigen) and all plasmids were verified by sequencing. All strains, primers and plasmids are listed in Tables 2–4, respectively. Plasmid cloning was carried out in E. coli DH5α or in DH5α Turbo (New England Biolabs). The strains FloT-PAmCherry and YdjI-mNeonGreen were constructed as follows: PAmCherry fragment was amplified from the plasmid pPAmCherry1-N1 using the primers pamCherry-F and pamCherry-R. The fragment of mNeonGreen was amplified from pNCS-mNeonGreen plasmid using the primers mNeonGreen-F and mNeonGreen-R. The BsaI site in pUC18 plasmid was mutated using the primers pUC18mut-F and pUC18mut-R resulting in the plasmid pUC18mut. YdjI and FloT including their endogenous promoter and ribosomal binding site were amplified from genomic B. subtilis DNA using the primers pairs YdjI-GG-F/YdjI-GG-R and FloT-GG-F/FloT-GG-R and cloned into pUC18mut. FloT-PAmCherry construct was done by Golden Gate assembly of the fragments yuaGIN, PAmCherry, cat, yuaGDown and pUC18BsaI, respectively resulting in the plasmid pUC18mut-yuaGIN-PAmCherry-cat-yuaGDown. For the ectopic expression of ydjI, we constructed plasmid pJPR1-ydjI. The coding sequence of ydjI including its own ribosomal binding site was amplified from genomic B. subtilis DNA using the primer pair XbaI_RSB_ydjI_UP and EagI_Stop_ydjI_DOWN and subsequently cloned into pJPR1. The final plasmid was transformed into B. subtilis and integration into the amyE locus was confirmed.

All oligonucleotides resulting in the respective fragments can be found in Table 4. All plasmids were transformed into B. subtilis was based on natural competence of B. subtilis (Harwood and Cutting, 1990) and screened for double cross over. Gene integration or deletion was validated by colony PCR whereas the expression and localization of the fluorescent fusions was additionally validated by microscopy.

Bacterial adenylate cyclase-based two-hybrid (BACTH) assay was performed according the protocol described by Karimova et al. (1998). Strain E. coli BTH101 (Euromedex) was co-transformed with plasmids containing genes of interest fused either to T18 or T25 fragments of adenylate cyclase. The adenylate cyclase is divided into two parts, each of them either N- or C-terminal present on the vectors pUT18/pKT25N or pKT25/pUT18C, respectively. Since the interaction of the proteins can be influenced by the position of the adenylate cyclase domains, all plasmid combinations were used (Table 3). After transformation, 10 μl of the cell suspensions were spotted on agar plates containing ampicillin (100 μg ml^–1), kanamycin (50 μg ml^–1), IPTG (0.5 mM), and X-Gal (40 μg ml^–1) and incubated at 37°C for approximately 40 h. Blue colonies indicate potential protein-protein interactions. Empty vectors or vectors containing in frame fusions of the leucine zipper of GCN4 to the T18/T25 fragments were used as negative and positive controls, respectively. The plates were stored at 4°C protected from light for at least 4 days to increase contrast and intensity of the colony color.

Detergent resistant membrane and detergent soluble membrane fractions were separated as described previously by Salzer and Prohaska (2001) with modifications. To this end, the kit CelLytic^TM MEM Protein Extraction Kit from Sigma-Aldrich was used. Overnight cultures of strains YdjI-mCherry2 and FloT-PAmCherry were diluted (1:100) into 50 ml of LB medium and incubated at 37°C (200 rpm) until an OD_{600 nm} between 0.5 and 0.8 was reached. The cells were induced with 15 mM NaOH (final concentration) and incubated for an additional hour under the same conditions. The cells were washed in cold saline phosphate buffer containing 1 mM EDTA and concentrated to a final volume of 1.75 ml. 10 μl of the provided protease inhibitor cocktail was added and the cells were cracked with 7 cycles of sonication (30% amplitude, 0.5 s/pulse ON, 0.5 s/pulse OFF, 30 s total), waiting on ice between the runs. The cell debris was pelleted with a bench top centrifuge (18,000 × g, 10 min, 4°C) and 1 ml of the supernatant was transferred into ultracentrifugation tubes. Membranes were pelleted at 235,000 × g for 40 min at 4°C. The membrane pellet was resuspended in 600 μl lysis and separation buffer supplemented with protease inhibitor cocktail, and incubated for 10 min on ice. The suspension was centrifuged for an additional 5 min (4°C, 18,000 × g). For the separation of the DSM and the DRM fraction the supernatant was heated at 30°C for 3–5 min. Both layers were separated at 3000 × g for 3 min (>20°C). Samples were taken from the upper hydrophilic phase (DSM) and from the lower hydrophilic phase (DRM) and used in Western-blot downstream analysis.

Overnight cultures containing the appropriate antibiotic and/or supplements were diluted 1:100, incubated at 37°C and induced with 15 mM NaoH when an OD_{600 nm} between 0.5 and 0.8 was reached. Controls were monitored so that they did not exceed mid-log phase. After 1 h induction, cell densities were normalized, pelleted in a bench centrifuge (max. speed for 10 min, 4°C), and washed with PBS supplemented with 1 mM EDTA. Cell pellets were stored at −80°C for further use. For separation into cytosol and membrane cell pellets, the samples were resuspended in 500 μl of the same buffer (PBS + 1 mM EDTA), supplemented with protease inhibitor cocktail (Sigma Aldrich). The suspensions were homogenized (7 cycles of sonication at 30% amplitude, 0.5 s/pulse ON, 0.5 s/pulse OFF, 30 s total). The cell debris were pelleted in a bench centrifuge (18,000 × g, 10 min, 4°C) and 500 μl of the supernatant was transferred into ultracentrifuge tubes. Pellets were washed once and resuspended in same volume as before. Samples of cytosol and membrane were incubated for 20 min at room temperature with 4× SDS loading dye. Samples were subjected to SDS-PAGE through a 10 or 12 % gel and blotted onto a PVDF membrane (Bio-Rad) at 200 mA for 2 h and blocked overnight with 5% milk solution in buffer (50 mM Tris–HCl pH 7.5/150 mM NaCl) supplemented with or without Triton X 100 under slight agitation at 4°C. The blots were incubated with anti-α-mCherry (1:1000) or, anti-mNeonGreen (1:1000) antibodies and incubated at room temperature for at least 1 h with mild shaking. The blots were then washed five times with buffer and incubated with the corresponding secondary antibody (anti-rabbit IgG for α-mCherry and anti-α-GFP; anti-mouse IgG for anti-mNeonGreen) conjugated with the corresponding alkaline phosphatase (1:2000) or horse radish peroxidase conjugate (1:5000, Invitrogen). Immunoblots were again washed with buffer and developed with NBT/BCIP or Pierce^TM ECL Western Blotting Substrate (Thermo Scientific) using the Chemi Hi Sensitivity program with signal accumulation of the ChemiDoc^TM MP. For in-gel fluorescence of GFP fusion proteins, SDS-PAGE gels were rinsed in distilled water and subsequently subjected to Blue Epi illumination in the ChemiDoc^TM MP. Emission was detected by a 530/28 filter. For in-gel fluorescence of ΔmreB pspA-GFP cultures were sonicated and directly applied to acrylamide gel-electrophoresis. Fluorescence of GFP was visualized in gels using a Bio-Rad ChemiDoc imager.

Overnight cultures of the strains were inoculated 1:100 into LB medium with supplements if necessary and grown to OD_{600 nm} 0.5–0.8. To induce PspA promoter, alkaline shock condition (15 mM NaOH, final concentration) was added. Cells were washed and resuspended in medium supplemented, when appropriate, with Dil-C12 solution (2.5 μg ml^–1, final concentration) for staining of fluid membrane regions as described before (Strahl et al., 2014), or with HADA (Kuru et al., 2015), as a peptidoglycan synthesis machinery probe. Exponentially growing cells were immobilized on 1% agarose-pads (w/v) in the appropriate medium mounted onto microscope slides using Gene Frames (1.0 × 1.0 cm chamber, 0.25 mm thickness, from Thermo Scientific) as described before (de Jong et al., 2011). Samples were covered with high precision microscopy cover glasses (170 ± 5 μm, Paul Marienfeld GmbH & Co. KG, Germany) and directly taken for microscopic observations.

Standard fluorescence microscopy was carried out using an Axio Zeiss Imager M1 fluorescence microscope (EC Plan-Neofluar 100×/1.30 Oil Ph3 objective) equipped with an AxioCam HRm camera. For membrane fluidity experiments, a Delta Vision Elite microscope (Applied Precision, GE Healthcare) equipped with an Insight SSI Illumination, an X4 Laser module, a CoolSnap HQ CCD camera and a temperature-controlled chamber set up at 37°C was used. Laurdan images were taken with an Olympus UplanSApo 100×/1.4 oil objective. Data processing was performed with SoftWoRx Suite 2.0 and FIJI Software’s.

Laurdan (6-Dodecanoyl-N, N-dymethyl2-naphthylamine, Sigma-Aldrich) was used to detect the liquid ordering in the membrane, as described (Bach and Bramkamp, 2013), with modifications. Cells were grown in LB until late exponential phase. Laurdan, dissolved in dimethylformamide (DMF), was added at 10 μM final concentration and cells were incubated for 10 min in the dark at 37°C, 200 rpm. Cells were then washed twice in PBS buffer supplemented with 0.2% (w/v) glucose and 1% (w/v) DMF, and resuspended in fresh pre-warmed medium. Laurdan was excited at 360 ± 20 nm, and fluorescence emission was captured at 460 ± 25 nm (exposure time: 500 ms) and at 535 ± 25 nm (exposure time: 1 s) (Strahl et al., 2014). The image analysis including the generation of GP maps was carried out using FIJI Software (Schindelin et al., 2012) in combination with the macro tool CalculateGP designed by Norbert Vischer.¹ The overall GP values were measured for at least 100 individual cells from two biological replicates, after background subtraction.