The list of strains, plasmids, and oligonucleotides used in this study is listed in Supplementary Tables 1–3, respectively. All strains were cultured at 37°C with shaking at 200 rpm. For S. aureus, the mid-exponential phase was defined as an OD₆₀₀ of 0.4–0.8. E. coli was cultured in Luria-Bertani (LB) broth or agar plus 100 μg/ml ampicillin. S. aureus strains were grown in tryptic soy broth (TSB) (Bacto) or agar (TSA), where required antibiotics were added at the following concentrations: 5 μg/ml erythromycin (Ery) 25 μg/ml lincomycin (Lin), 10 μg/ml chloramphenicol (Cm), 5 μg/ml tetracycline (Tet), 50 μg/ml kanamycin (Kan), and 100 μg/ml spectinomycin (Spec).

For growth curves, overnight S. aureus cultures were adjusted to OD₆₀₀ 0.05 in TSB and incubated at 37°C with shaking at 200 rpm to grow for 8 h. Samples were taken every hour, and OD₆₀₀ was measured. Direct cell counts were also performed by serial dilution in PBS and plating onto TSA. The number of colony-forming units (CFU) was directly counted after incubation. Growth curves were performed in triplicate. E. coli transformations and DNA manipulations were performed according to the previously described methodology (Sambrook and Russell, 2001).

Unless otherwise stated, all vectors were constructed in E. coli NEB5α (New England Biolabs, Ipswich, MA, United States) following previously described methods (Gibson et al., 2009; Lund et al., 2018) before passage through S. aureus RN4220 for DNA methylation (Novick and Morse, 1967). Finally, constructs were transduced into S. aureus SH1000 using phage Φ11. Transductions and transformations were confirmed by PCR. The genomic DNA of SH1000 was used as a template for S. aureus gene amplification. Genomic DNA was isolated by incubating S. aureus cells in 2.5 μg/ml lysostaphin prior to extraction using a Qiagen DNeasy Blood & Tissue Kit (Cat no. 69506) in accordance with the manufacturer's instructions. S. aureus transposon mutants were obtained from the NARSA library (Bae et al., 2008; Fey et al., 2013). Transposons were transduced from the library to the recipient strain and confirmed by PCR.

TEM was performed as mentioned in a study by Sutton et al. (2021). In brief, samples were fixed overnight in 2.5% (w/v) glutaraldehyde at 4°C. Samples were washed in PBS, and secondary fixation was performed with 2% (w/v) osmium tetroxide for 2 h. After washing, samples were dehydrated in incrementally increasing concentrations of ethanol and then incubated in propylene oxide. Samples were infiltrated overnight in a 50% (v/v) propylene oxide to 50% (v/v) Epon resin mixture overnight, which was then replaced with pure Epon resin for 4 h, which was then replaced with fresh resin for another 4 h. Polymerisation was then performed in fresh resin at 60°C for 48–72 h. Approximately 80 nm thin sections were taken and stained with 3% (w/v) aqueous uranyl acetate followed by Reynold's lead citrate. Sections were imaged using a FEI Tecnai T12 Spirit transmission electron microscope operating at 80 kV. Images were recorded using a Gatan Orius SC1000B bottom-mounted CCD camera. TEM images were analyzed using Fiji software (Schindelin et al., 2012). Cell wall thickness was measured as previously described by Sutton et al. (2021).

S. aureus strains were grown overnight in TSB (with appropriate antibiotics), which were used to inoculate fresh TSB to an OD₆₀₀ of 0.05. Cells were then grown to the mid-exponential phase (OD₆₀₀ of ~0.5) before being labeled. Samples were protected from light throughout the staining process until imaged. Then, 500 μM of 7-hydroxycoumarin-3-carboxylic acid-amino-D-alanine (HADA) or 1mM of azido-Dalanyl-D-alanine (ADA-DA) (Kuru et al., 2012b; Monteiro et al., 2015; Lund et al., 2018) was added to cells for 5 or 30 min and incubated at 37°C with shaking to label newly synthesized peptidoglycan. Cells were then washed twice in PBS at 4°C. The azide group of ADA-DA was labeled (post-fixation) with 5 μg ml⁻¹ Alexa Fluor 488 Alkyne using the Click-iT™ Cell Reaction Buffer Kit (Invitrogen) according to the manufacturer's instructions. To label and visualize the entire cell wall, cells were resuspended in PBS and incubated at 4°C for 5 min with 8 μg ml⁻¹ Alexa Fluor 555 NHS ester (Invitrogen, Waltham, MA, United States). Cells were then washed in PBS. After labeling, cells were incubated with 2% (w/v) paraformaldehyde (PFA) for 30 min at room temperature and then washed twice in water. After fixation, cells were resuspended in water containing 2 μg/ml DAPI (Sigma) for 5 min at room temperature on a rotary shaker to visualize DNA where appropriate. Samples were then washed twice in water before mounting.

Cells were mounted onto poly-L-Lysine coated slides (Sigma) using SlowFade^TM Gold antifade reagent (Thermo Fisher Scientific, Waltham, MA, United States) and then imaged using a Nikon Ti inverted microscope fitted with a Lumencor Spectra X light engine. Images were obtained with a 100x PlanApo (1.4 NA) oil objective 1.518 RI oil, and an Andora Zyla sCMOS camera was used for detection.

SIM was performed as previously described by Lund et al. (2018). In brief, coverslips (High-precision, 1.5H, 22 ± 22 mm, 170 ± 5 mm, Marienfeld) were sonicated in 1 M KOH for 15 min before being washed and then incubated in poly-L-Lysine solution for 30 min. Coverslips were then washed and dried before fixed cells (suspended in water) were dried onto coverslips and mounted with SlowFade^TM Gold antifade reagent (Thermo Fisher Scientific).

SIM was performed using a v4 DeltaVision OMX 3D-SIM system fitted with a Blaze module (Applied Precision, GE Healthcare, Issaquah, USA) with lasers used to illuminate samples. For each Z-slice (0.125 nm), images were taken in five phase shifts and three angles. To reconstruct images, the software Softworx (GE Healthcare, Issaquah, USA) was used with optimisation for a 1.516 immersion oil. The same software was used for deconvolution and image alignment.

All measurements from microscopy images were made using Fiji (Schindelin et al., 2012). Unless otherwise stated, micrographs presented are maximum intensity projections of Z-stacks.

Statistical analysis was performed using Prism version 9.31 (GraphPad, Boston, MA).

As DivIVA and GpsB are paralogues, they may be functionally related and therefore share a localization. A previous study has independently shown that DivIVA (Pinho and Errington, 2004) and GpsB (Eswara et al., 2018) localize to the S. aureus septum. To assess the co-localization of GpsB and DivIVA fusion, constructs were produced using pOB (GpsB-mCherry) and pKASBAR (DivIVA-GFP) and co-expressed in the same strain (SH1000 gpsB::gpsB-mCherry kanR ΔdivIVA geh::divIVA-gfp). Both GpsB-mCherry and DivIVA-GFP localize at the septum when compared to a HADA label, which allows visualization of new cell wall peptidoglycan synthesis by fluorescence microscopy (Kuru et al., 2012a) (Figures 1A, B). However, DivIVA-GFP has a ‘dotty' pattern of localization, which has previously been reported (Pinho and Errington, 2004), whereas GpsB-mCherry is seen to be forming smooth rings, suggesting that these proteins are not co-localizing. When viewing through the Z-stacks, it is clear that both GpsB-mCherry and DivIVA-GFP are forming independent patterns at the developing septum (Supplementary Video 1).

Previous reports have suggested that GpsB is an essential protein in S. aureus (Santiago et al., 2015; Eswara et al., 2018); however, in this study, we were able to construct a marked deletion of the gpsB gene in SH1000 using piMay, and thus, it is non-essential in this background. A transposon inactivation gpsB mutant can also be found within the NARSA transposon library (Fey et al., 2013). A markerless divIVA deletion mutant was also constructed using pMAD, as well as a double mutant (SH1000 gpsB::kan ΔdivIVA). No differences in growth were found between any of the mutants or when compared to the wild type (Supplementary Figure 1). Structured illumination microscopy (SIM) was used to analyse the cell volume of the mutants (Figure 1C and Supplementary Figure 2). No significant differences were observed in the volumes of the SH1000 ΔdivIVA, SH1000 gpsB::kan, or SH1000 ΔdivIVA gpsB::kan strains compared to the wild-type SH1000. Next, TEM was used to interrogate the cell wall structure of these mutants (Figure 1D and Supplementary Figure 1). divIVA was found to have a slightly thicker peripheral cell wall than wild-type cells (p = 0.0283), while gpsB had a slightly thinner cell wall (p = 0.0401). No difference in cell wall thickness was observed in divIVA gpsB compared to wild-type SH1000 (Figure 1D).

The roles of gpsB and divIVA in other species and the results of bacterial two-hybrid screens previously described in the literature (Steele et al., 2011; Bottomley et al., 2017) allowed us to determine a potential role for DivIVA and GpsB in various aspects of cell growth and division. As single and double mutants of gpsB and divIVA did not show any clear phenotype, further mutations, from the NARSA transposon library (Fey et al., 2013) (unless otherwise stated), were added in combination with divIVA gpsB to provide information about DivIVA and GpsB.

S. aureus encodes three LytR-CpsA-Psr (LCP) proteins within its genome: lcpA, lcpB, and lcpC (Over et al., 2011; Chan et al., 2013), all of which have a putative interaction with GpsB in a bacterial two-hybrid system (Kent, 2013). The LCP family of proteins catalyzes the transfer of WTA intermediates to the cell wall (Kawai et al., 2011; Chan et al., 2013). Severe phenotypic defects of lcpA mutants prevented the establishment of multiple mutant strains (Over et al., 2011; Chan et al., 2013), and no significant difference could be found in cell volume when comparing SH1000 lcpB::ery and SH1000 lcpB::ery ΔdivIVA gpsB::kan (Supplementary Figure 3A). An lcpC mutation, which has previously been shown to have the smallest impact on cell growth and morphology (Over et al., 2011; Chan et al., 2013), was also tested. SH1000 lcpC::ery was found to be significantly smaller than wild-type SH1000, while SH1000 lcpC::ery ΔdivIVA gpsB::kan was significantly larger than SH1000 (Figure 2A). This increase in cell size was observed in all stages of the cell cycle (Supplementary Figure 4A). Strains SH1000 ΔdivIVA lcpC::ery and SH1000 gpsB::kan lcpC::ery were produced to deconvolve these results. Both divIVA lcpC and gpsB lcpC showed no significant differences in cell volume to SH1000 (Figures 2C, D). lcpC was smaller than the wild type, while divIVA lcpC and gpsB lcpC were the same size as SH1000. However, divIVA gpsB lcpC was bigger than the wild type. This suggests that a divIVA or gpsB mutation causes an increase in the lcpC mutant size (as lcpC alone is smaller than SH1000), and divIVA gpsB together have a cumulative effect upon the size increase of cells lacking lcpC. divIVA gpsB lcpC was complemented using a pKASBAR plasmid, containing native gpsB, which integrates into the geh locus of S. aureus. A control of an empty pKASBAR was also used. As expected, both divIVA gpsB lcpC and divIVA gpsB lcpC geh::pKASBAR were significantly larger than wild-type SH1000, whereas divIVA gpsB lcpC geh::gpsB was the same size (Supplementary Figure 5A).

TarO, the first enzyme in the biosynthetic pathway of wall teichoic acids (Soldo et al., 2002; Atilano et al., 2010), was shown to interact with both DivIVA and GpsB (Kent, 2013). SH1000 tarO::ery (Salamaga et al., 2021) and divIVA gpsB tarO both show a significantly increased cell volume compared to SH1000, and divIVA and gpsB mutations did not alter this phenotype (Supplementary Figure 3B).

Due to the apparent interaction between DivIVA and GpsB and proteins involved in WTA synthesis and display, a potential role for LTA, which has been shown to be involved in cell division (Gründling and Schneewind, 2007), was also investigated. As the synthesis of LTA is essential via the action of LtaS (Gründling and Schneewind, 2007), ypfP, which has an 87% reduction in LTA content (Fedtke et al., 2007), was investigated. SH1000 ypfP::ery has a significantly greater volume than SH1000, while divIVA gpsB ypfP was significantly smaller than wild type (Figure 2B), suggesting that LTA may be important in DivIVA or GpsB function. The decrease in cell volume was observed in all stages of the cell cycle (Supplementary Figure 4C). divIVA ypfP and gpsB ypfP were constructed to further analyse this phenotype. Both divIVA ypfP and gpsB ypfP show an increase in cell volume compared to SH1000 (Figures 2C, D), the same phenotype as ypfP. Therefore, the loss of divIVA, gpsB, and ypfP are all required for the reduced volume phenotype of the triple mutant. Complementation of divIVA gpsB ypfP with pKASBAR expressing native gpsB restored the increased cell volume phenotype (Supplementary Figure 5).

Both DivIVA and GpsB interact with PBP4 (Kent, 2013). SH1000 pbp4::ery had no significant difference in cell volume compared to parental SH1000, whereas divIVA gpsB pbp4 has a significantly greater volume than SH1000 (Figure 2B). This increase in cell volume was only seen in cells with no or an incomplete septum (Supplementary Figure 4B). The results were deconvolved by constructing and analyzing divIVA pbp4 and gpsB pbp4. divIVA pbp4 shows a significantly greater increase in cell volume compared to SH1000 (Figure 2C). However, gpsB pbp4 shows no significant difference in cell volume compared to the wild type (Figure 2D), demonstrating that only a loss of divIVA is required for the phenotype. This result was complemented using pKASBAR expressing native divIVA (Supplementary Figure 5B).

S. aureus has previously been shown to elongate during the cell cycle (Monteiro et al., 2015), resulting in a long axis and a short axis. Calculating the ratio between these two axes allows the extent of elongation to be calculated as previously reported (Reichmann et al., 2019) and acts as a measure of circularity. In this study, we calculated the ratio by dividing the short axis (axis perpendicular to the long axis) by the long axis (axis perpendicular to the septum). Using the short/long-axis ratio, a value of 1 indicates that the cell is perfectly circular, whereas the smaller the ratio, the more elongated the cell is.

The short/long-axis ratio was calculated for divIVA and gpsB mutants (Figure 3A). divIVA had no significant difference in short/long-axis ratio compared to SH1000, while both gpsB and divIVA gpsB had a significantly greater ratio, meaning that the cells are more circular, or less elongated, than wild-type cells. Both gpsB and SH1000 divIVA gpsB have significantly higher short-/long-axis ratios with incomplete septa compared to SH1000 and divIVA, but there is no difference with no septa (Figure 3B). Complementation with gpsB being expressed from the geh locus using pKASBAR (Bottomley et al., 2014) restored the elongation phenotype, with no difference in short/long-axis ratio between the wild-type and complemented strains (Figures 3C, D).

It has previously been reported that RodA and PBP3 form a cognate pair that is important for peripheral PG synthesis and the elongation of S. aureus (Reichmann et al., 2019). All single, double, and triple mutant permutations were created for rodA, gpsB, and pbp3 and compared (Figure 3E). pbp3 and rodA pbp3 cells were significantly more circular than SH1000. gpsB was found to be significantly more circular than SH1000, rodA, pbp3, and rodA pbp3. gpsB showed no differences in circularity to rodA pbp3 gpsB (Figure 3E and Supplementary Table 4).

The localization of PG synthesis was determined by measuring the fluorescence ratio (FR) at the septum and the periphery of cells that have been sequentially labeled with ADA-DA and HADA (Figure 4A) for 5 min each to follow septal development (Tinajero-Trejo et al., 2022). The higher the FR, the greater the PG synthesis at the septum compared to the periphery. gpsB had a significantly higher FR than wild-type SH1000 (Figure 4B).

As PG synthesis is altered in gpsB, we determined PBP localization at the septum compared to the cell periphery using PBP fluorescent reporter fusions. PBP2 is the major PG transpeptidase in S. aureus responsible for the bulk of PG synthesis (Pinho et al., 2001a,b). Imaging SH1000 and gpsB expressing GFP-PBP2 (Figure 4C) showed a significant increase in FR for gpsB (Figure 4D), indicating that there is more GFP-PBP2 in the septum of dividing cells in the absence of GpsB.

PBP3 plays a role in the elongation of S. aureus due to off-septal PG synthesis (Reichmann et al., 2019). A GFP-PBP3 construct was made in pLOW under the control of the Ppcn promoter, which was then transduced into SH1000 pbp3::spec and SH1000 pbp3::spec gpsB::kan; therefore, the only expressed copy of PBP3 was GFP-PBP3. To test whether the fusion was functional, the circularity of SH1000 (Supplementary Figure 7A), pbp3 (Supplementary Figure 7B), pbp3 gfp-pbp3 (Supplementary Figure 7C), and pbp3 gpsB gfp-pbp3 (Supplementary Figure 7D) was compared. pbp3 is significantly more circular than SH1000. pbp3 gfp-pbp3 was found to be significantly less circular than pbp3 and was found to have no difference compared to SH1000 but could not complement the increased circularity of gpsB (Supplementary Figure 7E). pbp3 gpsB gfp-pbp3 (Figure 4E) had a significantly higher FR for GFP-PBP3 than pbp3 gfp-pbp3 (Figure 4F), demonstrating increased PBP3 at the septum in the absence of GpsB.