Bacterial strains and plasmids are listed in Table 1. Xac was cultivated at 30°C in various media: nutrient yeast glycerol broth (NYGB, peptone 5 g/L; yeast extract 3 g/L; glycerol 20 g/L) or on plates containing agar (15 g/L, NYGA) supplemented with D-glucose (1% w/v) or L-arabinose (0.05% w/v) when required; nutrient yeast citrate broth (NYCB, peptone 5 g/L; yeast extract 3 g/L; glycerol 2.46 g/L; trisodium citrate 0.44 g/L); Xam1 medium (per liter: 2.46 g glycerol, 0.247 g MgSO₄·7H₂O, 1.0 g (NH₄)₂SO₄, 4.5 g KH₂PO₄, 10.5 g K₂HPO₄, 0.5 g Na₃C₆H₅O₇·2H₂O, 0.3 g casaminoacids, 1 g BSA, pH 5.4 with HCl); or Xamg1 medium which is identical to Xam1 except for glycerol which is at 20 g/L. LB 0% is Lysogeny Broth without NaCl (tryptone 10 g/L; yeast extract 5 g/L), whereas LB 0.5% contains NaCl 5 g/L.

For the cloning steps, we used E. coli DH10B (Invitrogen) cultivated in LB 0.5%-agar/LB 0.5% at 37°C (Sambrook et al., 1989). The antibiotics kanamycin, gentamicin, and ampicillin were used at the concentration of 20 μg/mL.

The minC gene was amplified by PCR using Xac genomic DNA as a template and the primers minC_pLAL1F/minC_pLAL1R (all oligonucleotides are listed Table S1). The resultant DNA fragment was digested with EcoRI/HindIII and ligated into the Xac expression vector pLAL1 (GenBank KP696472; Lacerda et al., 2017) digested with the same enzymes, which resulted in pGCD1C.

Two DNA fragments flanking the genomic sequence of minC, minCup (870 bp), and minCdown (929 bp) were obtained by PCR using the primer pairs minCupF/minCupR and minCdownF/minCdownR, respectively. The PCR products minCup and minCdown were digested with the restriction enzymes BamHI/XbaI and XbaI/HindIII, respectively, and ligated to pNPTS138 (kindly donated by Lucy Shapiro, Stanford University, USA) digested with BamHI/HindIII, which generated pNPTSΔminC.

GFP-MinC was expressed using the integrative vector pGCD21 (this work), a derivative of pHF5Ca (Ucci et al., 2014), which enables the expression of GFP (gfp-mut1, GenBank ADF80258.1) fusion proteins in Xac under the control of the arabinose promoter. To construct pGCD21, gfp was removed from pPM7g (Martins et al., 2010) using the restriction enzyme BamHI/XbaI; the isolated gfp cassette was ligated to the backbone of pHF5Ca digested with the same enzymes, giving rise to pGCD19 (GenBank KJ619486). Next, the arabinose repressor and promoter (araC-para) were amplified by PCR using pEB304 (Gully et al., 2003) as a template and the primer pair pARAF/pARAR. The PCR product was digested with EcoRI/BglII and ligated to the backbone of pGCD19/EcoRI/BglII, generating pGCD21 (GenBank KU678206). In order to clone Xac minC in pGCD21, the gene was isolated by PCR using genomic DNA and the primers minCF/201402minCR. The PCR product was digested with NotI/XbaI prior to ligation into pGCD21/NotI/XbaI, which resulted in pGCD2C.

All the constructs were checked by DNA sequencing.

All plasmids were transformed in Xac by electrotransformation (Ferreira et al., 1995). First, Xac was transformed with the replicative vector pGDC1C, which provides an additional copy of minC. Afterwards, Xac/pGCD1C was transformed with pNPTSΔminC, which mediates removal of chromosomal minC by allele exchange. Mutant strains carrying one of the first crossover events (integration of pNPTSΔminC into minCup or minCdown) were selected on NYGA/kanamycin. To obtain the second crossover, deletion of minC, individual colonies were cultivated for 16 h in NYGB supplemented with 0.01% arabinose and gentamycin only, which is the marker of pGCD1C. The final selection of Xac/pGCD1C ΔminC was carried out in NYGA supplemented with 3% sucrose. Deletion of minC was verified by diagnostic PCR and Southern Blot experiments using minCD as a probe. To cure pGCD1C, Xac/pGCD1C ΔminC was cultivated for ~20 generations without gentamycin (Xac ΔminC).

Xac ΔminC was complemented with GFP-MinC by transformation with pGCD2C. The vector was stably integrated into the amy locus (Xac ΔminC amy::gfp-minC). For a GFP only control strain, Xac ΔminC was transformed with pGCD21. The vector was stably integrated into the amy locus (Xac ΔminC amy::gfp). To test whether GFP-MinC was expressed as full-length protein and not proteolytically cleaved inside the cell, cell extracts of cells expressing GFP only or GFP-MinC were analyzed by SDS-PAGE/Western Blot using antibodies against GFP (Figure S1).

Xac cells were grown to exponential phase in either NYGB or Xam1 medium and directly imaged or labeled with a fluorescent dye prior to imaging.

DAPI (4′,6-diamidino-2-phenylindole) was used to image DNA. Exponential phase cells were harvested, resuspended in 10 μM DAPI in Phosphate Buffered Saline (PBS) buffer, pH 7, for 30 min, washed twice in PBS and imaged.

HADA (Hydroxycoumarin-carboxylic acid-Amino-D-Alanine) labeling, to detect sites of peptidoglycan synthesis, was done as previously described by Kuru et al. (2015). HADA was added to 125 μM in exponential phase Xac cells, labeling time was 24 min in Xam1 or 10 min in NYGB (8% of doubling time under these growth conditions). Cells were either imaged directly when simultaneous detection of GFP was required, or fixed with ice-cold ethanol for 10 min, and washed before imaging.

Cells were imaged using a Nikon Ti-E microscope (Nikon Instruments, Tokyo, Japan) equipped with a Hamamatsu Orca Flash 4.0 camera. Image analysis was performed using the software ImageJ (http://rsb.info.nih.gov/ij/).

Xac amy::gfp (wild type control) and Xac ΔminC amy::gfp were analyzed to observe MinC deletion effects. Xac ΔminC amy::gfp-minC was analyzed to see if GFP-MinC could complement the ΔminC phenotype. GFP producing strains were used as the cytosolic GFP allowed for an easier determination of whether a cell had divided or not. First, images were manually inspected for short filamentation, minicells and branching. Two biological replicates were analyzed with at least 200 cells each. Subsequently, cell lengths of rod-shaped cells (not minicells or branched cells) were measured using the public domain program Coli-Inspector, which runs under ImageJ in combination with the plugin ObjectJ, written by Norbert Vischer (http://simon.bio.uva.nl/objectj/). After the automatic selection process, a manual check was done to make sure only single cells were selected. Cells with a length equal to or longer than 3.15 μm were defined as short filaments. Two biological replicates of each strain were analyzed with at least 500 cells per replicate.

The nucleoid length in cells of Xac parB::parB-gfp and Xac ΔminC parB::parB-gfp stained with DAPI was manually measured using ObjectJ.

Xac amy::gfp-zapA and Xac ΔminC amy::gfp-zapA were were analyzed to observe co-localization of HADA (peptidoglycan incorporation) with GFP-ZapA (division sites), and possible MinC deletion effects. After the pictures were taken, the fluorescence channels CFP (HADA) and FITC (GFP-ZapA) were merged. Linescans were made of the cells displayed using the plot profile function of ImageJ. The maximum fluorescence intensity of each cell was defined as 100%, the minimum at 0%, and plots were made relative to those levels (Figure S5). Linescans confirmed that colocalization could be scored manually from image overlays, which was done for a total of more than 100 cells per strain.

To study the role of the Min system in Xac we deleted minC in a strain that expresses gfp. This resulted in a strain that exhibited the classical Δmin phenotype: next to rod-shaped cells of normal length, minicells and longer cells, both the result of polar divisions, could be observed (Figure 1Bb,c). However, in addition to these phenotypes, branched cells were also observed in the minC mutants (Figure 1Ba). Similar results were obtained in a strain that did not carry a copy of the gfp gene (Figure 1D). To show that the phenotype was due to the deletion of minC, and not caused by any polar effects, we complemented the strain with gfp-minC at the ectopic amy locus. This complementation restored the wild type phenotype (Figures 1A,C, Table 2).

Branches and minicells were not observed in the wild type strain nor in the strain complemented with gfp-minC. Branches and minicells could not be detected automatically, thus the frequency of their occurrence in the minC deletion strain was scored by visual inspection, which revealed that around 20% of the cells were either minicells or branched cells (Table 2, see Section Branching in the minC Deletion Strain is Dependent on Growth Medium for more details on branching). Subsequently, the length of the rod-shaped cells was determined by automatic length analysis using Coli-inspector (methods). This showed that the length distribution of the minC deletion strain was a lot broader than that of the wild type and gfp-minC complemented strain (Figure S2). This is most likely because the longer cells in the minC deletion population are the result of the outgrowth from an asymmetric division. It is impossible to precisely determine which cells are derived from an asymmetric division. Therefore, we defined “short filaments,” which are likely to represent such cells, as cells with a length of over 3.15 μm (this is 1.5 times the median length of cells in the wild type strain). These short filaments comprised around 16% of the cells observed in the minC deletion strain, whereas the wild type strain, and the strain complemented with gfp-minC had less than 3% of these cells. As the gfp-minC complemented strain is so similar to the wild type, while the minC knockout is strikingly different, and as GFP-MinC does not show signs of degradation even when overexpressed (Figure S1), we conclude that GFP-MinC is fully functional.

The strain Xac ΔminC amy::gfp-minC, in which gfp-minC is the only copy of minC, was studied using fluorescence microscopy. As expected, GFP-MinC was localized in a gradient throughout the cell with the maximum at one of the poles (Figure 2, Figure S4). To see whether MinC oscillates from pole to pole, as previously described for E. coli and recently S. elongatus (MacCready et al., 2016), we imaged the cells with 10 s intervals. These short time lapses clearly show that GFP-MinC localization is dynamic, and that the protein oscillates from pole to pole (Figure 3, Movie S1). Following several oscillations in various cells over time revealed that GFP-MinC oscillates with a periodicity of roughly 65 s, which is comparable to the oscillations described in E. coli (Raskin and De Boer, 1999a).

The branching phenotype of E. coli min mutants has been described in two studies from the 1990's (Akerlund et al., 1993; Gullbrand et al., 1999), but the phenotype, to paraphrase a paper from the Young lab, has received much less attention than other questions related to bacterial morphology (Potluri et al., 2012). Branching of min mutants was shown to be dependent on the composition of the growth medium. Some branching was observed on rich (LB) medium, whereas growth on minimal salt (M9) media supplemented with casaminoacids and either succinate or acetate, but not glucose, resulted in branching in 5–20% of the cells depending on the E. coli strain studied (Akerlund et al., 1993; Gullbrand et al., 1999). We tested the medium dependency of the branching phenotype of the Xac minC deletion strain and observed similar results: hardly any branches on the rich NYG/CB media and roughly 12% of branches on the minimal salt Xam1 medium. An intermediate phenotype was observed on LB where cells showed bulges and abnormal extensions at the poles. These cells were scored as branching cells (Table 3; Figure 4). We would like to remark that it is formally possible that minicells do not only originate from polar divisions but also from budding of branches. However, as the frequency of minicelling under non-branching conditions is almost the same as on branching conditions we deem this unlikely.

Disruption of the min system in E. coli (Jaffé et al., 1988; Mulder et al., 1990) results in aberrant nucleoids, and in the case of branching cells large masses of DNA are often located at branching points (Akerlund et al., 1993). To study nucleoid distribution and organization we made use of strains carrying a functional GFP fusion to ParB (Ucci et al., 2014). In a wildtype background, GFP-ParB localizes to the origin-proximal parS site on the chromosome. Upon parS duplication during chromosome replication and segregation, ParB bound to the second parS copy moves from one pole to the other during the cell cycle (Ucci et al., 2014; Figure 5A). We introduced the parB-gfp allele into the ΔminC background to construct the strain Xac ΔminC parB::parB-gfp, and studied nucleoid distribution with DAPI and nucleoid organization using ParB-GFP. In branching cells, we observed longer nucleoid length and the occasional accumulation of DNA at branch points (Figures 5B, 6), although the accumulation at branch points was not a typical feature of branching cells (9 out of 36 branching cells on Xam1 had DNA at the branch point). The ParB-GFP pattern appeared less clearly defined on Xam1 medium (Figures 5A,B) with more faint fluorescent spots visible in the cells compared to the one or two spots per cell seen with NYGB medium (Figures 5C,D).

In non-branching ΔminC cells the organization of the nucleoid looked more like wild type—the DNA was clearly more condensed—although in some cells it appeared that divisions were initiated over non-segregated nucleoids (Figure 5D). Combined, these results suggest that chromosome segregation/organization is affected in the absence of MinC.

In E. coli, branch formation is the result of defects in positioning the cell division machinery, notably FtsZ (Potluri et al., 2012). In min mutants the FtsZ ring can form at the poles, but there is also an increased zone for FtsZ polymerization at midcell, which also occurs in cells with defective PBP5 (Potluri et al., 2012). In E. coli some peptidoglycan (PG) is synthesized in an FtsZ-dependent manner at a site that will form the new pole, and strikingly this PG is not subject to turnover of material, which is why it is called inert PG (iPG) (de Pedro et al., 1997). iPG is always found at the poles and at the tips of branches, which originate from places on the side wall that contained iPG. We studied the relation between cell division site placement and peptidoglycan synthesis by using GFP-ZapA (Martins et al., 2010) as a proxy for division site localization and the fluorescent D-amino acid analog HADA (Kuru et al., 2012) to label sites of active, ongoing PG synthesis. Cells were grown for 24 min (8% of the doubling time) in the presence of HADA and analyzed. In wildtype cells, most PG synthesis occurred at division sites at midcell, which could also be identified by fluorescent bands of GFP-ZapA (Figures 7A,C). This pattern is similar to HADA labeling in other rod-shaped bacteria such as E. coli and B. subtilis (Kuru et al., 2012). Sometimes, spots of HADA were seen at some poles, which could be due to the fact that the HADA was incorporated in the septum just before cell division. In non-dividing cells, GFP-ZapA localized throughout the cytosol as described (Martins et al., 2010). Consistent with strong PG incorporation at division septa, 97% of the cells that had clear labeling with both HADA and GFP-ZapA showed an overlapping signal (Table 4). In the minC deletion mutant, this colocalization of GFP-ZapA and HADA incorporation at division sites was lost (Figures 7B,D, Table 4). The minC deletion strain when grown in conditions that promote branching, shows less complete GFP-ZapA rings, which also do not appear to be fully perpendicular to the cell axis, and strong GFP-ZapA signals are not overlapping with bands of HADA that can be observed at sites that show constrictions that would be indicative of ongoing divisions. When these cells are grown under non-branching conditions, this defect in colocalization of GFP-ZapA and HADA was not rescued (Table 4). Recent work has pointed at a role for metabolism in the control of cell division, with pyruvate levels in B. subtilis stimulating efficient Z ring formation under nutrient rich conditions through pyruvate dehydrogenase E1α (Monahan et al., 2014; Sperber and Herman, 2017). We tested whether the addition of pyruvate to the medium in which branching occurs can rescue Z ring positioning and branching, but this was not the case (Figure S3).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.