Non-essentiality of canonical cell division genes in the planctomycete Planctopirus limnophila

Most bacteria divide by binary fission using an FtsZ-based mechanism that relies on a multi-protein complex, the divisome. In the majority of non-spherical bacteria another multi-protein complex, the elongasome, is also required for the maintenance of cell shape. Components of these multi-protein assemblies are conserved and essential in most bacteria. Here, we provide evidence that at least three proteins of these two complexes are not essential in the FtsZ-less ovoid planctomycete bacterium Planctopirus limnophila which divides by budding. We attempted to construct P. limnophila knock-out mutants of the genes coding for the divisome proteins FtsI, FtsK, FtsW and the elongasome protein MreB. Surprisingly, ftsI, ftsW and mreB could be deleted without affecting the growth rate. On the other hand, the conserved ftsK appeared to be essential in this bacterium. In conclusion, the canonical bacterial cell division machinery is not essential in P. limnophila and this bacterium divides via budding using an unknown mechanism.


Results
We focused here on the P. limnophila homologs of the canonical cell division genes ftsI, ftsW, ftsK and mreB, detected by us and others 17,48,49 , supporting the identification of the homologs. P. limnophila FtsI (Uniprot ID: D5SXQ2) is homologous to the E. coli protein with 34% identity and similar length: 561 vs 572 amino acids (aa). P. limnophila FtsW (ID: D5STY0) shares 34% identity with the protein from E. coli with similar length, 406 vs 414 aa. P. limnophila contains only one FtsK homolog (ID: D5SYV8) with 46% identity to E. coli protein and size similar to most FtsK homologs (750-900 aa). The MreB homolog (D5SQI2) shares 47% identity to E. coli with similar size, 348 vs 347 aa, and the corresponding gene seems to be in the same transcriptional unit as mreC and mreD, as they are separated by 74 and 3 nucleotides, respectively.
The chromosome and PG are two important cell components inherited during cell division. FtsK is an ATP-dependent DNA pump, most likely responsible for the transfer of the genetic material from mother to daughter cell. It is highly conserved and is an essential component of the divisome in model bacteria 50 . Despite the punctuated pattern of presence of most of the canonical cell division genes in the planctomycetal genomes, ftsK is conserved across the phylum 17 , and semi-quantitative RT-PCR assays showed that it is expressed at the same level throughout the cell cycle, as compared to the control gyrA gene, constitutively expressed in most bacteria (Fig. 1). We were unable to generate insertion or deletion mutants of this gene in P. limnophila, suggesting that FtsK plays an essential role in this bacterium (identical results were obtained for G. obscuriglobus ftsK, supporting the proposed essentiality of the gene; ERM unpublished).
Bacterial cell division is tightly linked to PG synthesis. In model organisms, PG incorporation into the growing PG layer is performed by the divisome at the septum and by the elongasome elsewhere. In the periplasm, the PG precursor is incorporated into the cell wall by various penicillin-binding proteins (PBP). FtsI (PBP3) is one of the PG transpeptidases involved in septal PG synthesis 51,52 . FtsI and FtsW have been shown to interact and to be essential in model bacteria 53 . The genes coding for FtsI and FtsW are conserved in roughly half of the planctomycetal proteomes 17 . In order to decipher if these genes are expressed in P. limnophila, we first performed semi-quantitative PCR, demonstrating expression of ftsI and ftsW in the wild type strain at the same level in exponential and stationary phase (Fig. 1). As suggested by their patchy conservation in Planctomycetes 17 , and in contrast to the situation observed in other bacteria, we were able to generate single deletion mutants for both ftsW and ftsI. Growth curves of the deleted strains did not differ from the wild type ( Fig. 2A). Similarly, budding division could not be differentiated from the wild type strain when examined by bright field microscopy (Fig. 2C).
By measuring various cell morphology indexes, the most significant changes were detected for the ftsW deletion mutant. The length of the cells of the ftsW mutant decreases, while their width increase, resulting in a change of circularity (all with pairwise Wilcox test P value < 10 −15 ; Fig. 2B). These changes resulting in slight modifications of the area of the cells (Wilcox test P value > 10 −3 ). Nevertheless, as stated above, no differences could be detected in the growth curves or the division modes and thus, the ftsW gene is not essential for cell division.
The actin homolog MreB is the main component of the elongasome. It is expected to play a scaffolding role during lateral growth to guide cell-wall elongation in rod-shaped bacteria 30 . Similar to ftsI and ftsW, mreB showed  a patchy distribution across the phylum 17 and we could generate a deletion mutant. Growth curves, morphology and budding phenotype did not differ from the wild type (Fig. 2).
The lack of a division phenotype for the ΔmreB mutant additionally presented the opportunity to test the effect of the A22 drug in an mreB-negative background (Fig. 3). This compound is classically used as an MreB-inhibitor despite the fact that the possible existence of off-target effects has been raised 54 . Before fission, the P. limnophila wild type cells increase in cell volume. At the apex of size, the daughter cell bud is formed, and the mother cell decreases in total cell volume. When exposed to A22, wild type and mreB deletion mutants, still increased in size (Supp. Fig. S1A) but no longer initiated cell division (Supp. Fig. S1B). These results confirm that, at least in P. limnophila, the drug has effects on targets other than MreB, corroborating the proposed unspecific effect of A22.

Discussion
We report the first evidence of non-essentiality for the divisiome and elongasome components FtsI, FtsW and MreB. This raises the possibility that the divisome and elongasome complexes, essential in the vast majority of bacteria, are absent in some members of the Planctomycetes phylum.
Current knowledge about bacterial cell division is predominantly based on model organisms. However, recent studies on various non-model organisms have begun to reveal that some of the statements derived from model organisms do not apply to bacteria universally. A wide variety of different cell division mechanisms is bound to broaden our perspective about bacterial cell division 34,55 .
New evidence has questioned the central role of FtsZ as the predominant force generator during cell division 4,56,57 . An emerging model proposes that assembled FtsZ is a platform to recruit PG synthesis enzymes, and that it is PG synthesis per se that directly contributes the forces for septum formation 58 . As all planctomycetes are devoid of FtsZ, this protein cannot be the force generator during cytokinesis in this phylum. However, P. limnophila divides by budding, not by binary fission. On the other hand, the planctomycetal class Phycisphaerae and order "Candidatus Brocadiales" (i.e. anammox bacteria) divide by a mechanism similar to binary fission without FtsZ. This raises the question of whether there is a different main protein driving cell division in these organisms.
In P. limnophila, ftsI and mreB deletion do not produce phenotypes. However, we observed that ftsW deletion leads to cells that are more spherical compared to the ovoid wild type, without any detectable modification of growth pattern or division mode. FtsW is essential in most organisms containing a single copy of this gene. Depletion of FtsW in rod-shaped bacteria leads to a block in cell division and formation of elongated cells [59][60][61] . In the polyploid spherical cyanobacterium Synechocystis sp. strain PCC 6803, a ΔftsW heteroploid mutant (still retaining wt copies despite deletion of some ftsW genes) grows slower and generates giant cells 62 . No such phenotype is observed for the mutants generated in P. limnophila which is ovoid and divides by budding. www.nature.com/scientificreports www.nature.com/scientificreports/ MreB is essential in model bacteria as well as in bacteria with a divergent division mode, like the longitudinally dividing bacteria of the genus Thiosymbion 55 . However, some exceptions exist, such as the ovoid Streptococcus pneumoniae and the rod-shaped Corynebacterium glutamicum, both devoid of MreB 63 . Many rod-shaped alphaproteobacteria and actinobacteria also lack MreB 64 . On the other hand, some coccoid cyanobacteria contain the mreB gene.
Planctomycetes belongs to the PVC superphylum that also comprises the Chlamydiae, Verrucomicrobia and Lentisphaerae. The latter two contain FtsZ and divide by binary fission. As for Planctomycetes, the chlamydial genomes do not encode for FtsZ and it has recently been reported that a few members of the Chlamydiae also exhibit asymmetric, budding-like cell division 35 . In these species, MreB localizes at the cell division site 65 . In most species lacking FtsZ, it is proposed that MreB might substitute for its function 66 . This possibility is revoked here as cells lacking MreB do no display any phenotype.
The non-essentiality of divisome and elongasome genes raises the question of the role of PG itself in Planctomycetes and its essentiality to cell integrity and division, despite the importance of these functions in other bacteria. Non-natural forms of PG-deprived cells can be induced in osmoprotective conditions by inhibition of PG synthesis in diverse bacteria 67 . These so-called L-forms divide and can be maintained indefinitely. It has been shown that ftsZ becomes non-essential in L-form bacteria, at least in B. subtilis 68 . It has been proposed that, in these cell wall-deficient, propagation occurs by an extrusion-resolution mechanism. In addition, the involvement of "some kind of cytoskeletal system" has been suggested and some MreB homologs were directly pinpointed 68 . Interestingly, L-form bacteria appear to be resistant to A22 when grown in the presence of a beta-lactam antibiotic, suggesting that MreB is not required for the growth of spherical L-form-like cells 69 . This is related to our report of A22 susceptibility of the ΔmreB mutant which confirms previous suspicion of side-effects of this widely used drug 70 .
Finally, loss of PG synthesis can in some cases lead to non-dividing cells that may enlarge considerably. This has been observed in various species, including E. coli and Vibrio cholera. Recently it has been shown that giant bacteria could be formed by eliminating essential functions needed for, or treatment with antibiotics targeting PG synthesis in Acinetobacter baylyi, a diderm gammaproteobacteria. However, these giant cells did not proliferate 71 . The relevance of these artificially induced cells and division modes and how they compare to Planctomycetes division is unclear.
The non-essentiality of the divisome genes for planctomycetal division might appear to be consistent with the previously suggested unusual status of phylum Planctomycetes and related PVC superphylum bacteria within the bacterial domain. It seems probable that PVC bacteria are derived from diderm bacteria and that their unknown division modes may have evolved from FtsZ-based binary fission. The deciphering of these alternative division modes and their evolution is important for our sampling of the biodiversity and our understanding of evolutionary cell biology.
Altogether, our results demonstrate that Planctomycetes use divergent division modes, urging for the characterization of the molecular mechanisms of division in P. limnophila and in other species of this phylum.
Plasmid description and genetic modification. Plasmids used for gene deletion in a double event of homologous recombination were derived from the pEX18Tc vector 72 . To construct knockout plasmids for ftsI, ftsW, ftsK and mreB genes, 1000-1200 bp upstream and downstream fragments of the target gene were amplified by PCR from genomic DNA using the primer pairs listed in Supplementary Table 1. The upstream and downstream fragments were digested with the appropriate enzymes and then cloned into pEX18Tc by three-way ligation. Finally, the kanamycin resistance gene amplified from the pUTminiTn5km plasmid 73 was subsequently cloned as a BamHI fragment between the two flanking regions. These plasmids enable the deletion of each of the complete genes.
Genetic transformation of P. limnophila was performed by electroporation as described before 74 . The cells were then plated onto modified PYGV plates supplemented with kanamycin 50 μg ml −1 and were incubated at 28 °C until colony formation (7-9 days). Colonies were transferred to fresh selection plates and genotyped by Southern Blotting and sequencing.
Mutants sequencing. Deletion mutants were verified by paired-end sequencing on an Illumina MiSeq machine upon library preparation with the Nextera XT DNA Library Prep Kit (Illumina, San Diego, USA). Pre-assembly processing of the reads was done employing Trimmomatic v0. 35 75  www.nature.com/scientificreports www.nature.com/scientificreports/ performed with a DNA-free kit (Ambion). The samples were purified using RNAeasy columns (Quiagen) and RNA quality was confirmed by non-denaturing agarose gel electrophoresis. The absence of contaminating DNA was then confirmed by PCR amplification.
Reverse transcription of the RNA samples was performed using the High-Capacity cDNA Archive Kit (Applied Biosystems), with random hexamers as primers to generate cDNAs. The resulting cDNA samples were amplified by semi-quantitative RT-PCR using 1 mM of each primer (Supplementary Table 2). Genomic DNA was used as a positive control of the PCR. Samples were visualized in 10% acrylamide gels.
Microscopy. Bacteria from 2 ml of exponentially growing culture (OD 600 ~0.4) were harvested (12000 g, 3 min) and resuspended in 100 μl of fresh medium. A sample of 2 μl was spotted on a glass-bottom dish (MatTek) and covered with a 1% agarose in M3 medium cushion, as described by 82 . Bright field images were acquired using a 100×/1.46 objective through an 1.6X amplification lens and an EMCCD Andor iXon camera mounted on a Zeiss microscope, resulting in a pixel size of 0.1 × 0.1 µm. image analysis. An automatic analysis workflow was design using FIJI until a satisfactory segmentation of the cells was achieved. The image analysis workflow runs as follows under FIJI software: Image Acquisition → Subtract Background → Gaussian Blur → Invert → Enhance Contrast → Unsharp Mask → Watershed Thresholding → Convert to mask → Binary Watershed → Analyze Particles (with a size upper and lower limit of 21 • 10 −3 to 3.5 • 10 −3 µm. Any particle out of this limit was not considered as a cell. Afterwards, the area of the cells was measured and fitted to an ellipse. In order to extract cell's width and length, we identified them as the ellipse's minor and major axis, calculating subsequently the circularity. The values of each single cell were exported for further statistical analysis with R. A22 inhibition assay. Exponentially growing cultures (OD 600 ~0.4) were loaded on a CellASIC ONIX plate for bacteria (EMD Millipore). The plate was then flushed with fresh medium without A22 for at least 12 hours to get the cells accustomed to their new environment. The plate was then flushed with fresh medium constituted with A22 (20 μg ml −1 ). Images were acquired every five minutes using a Leica DMi8 microscope with a 1000x phase contrast objective using an 8 micron Z-stack.
Statistical methods. All statistical analyses were done with R with >250 observations for each features.
Normality was tested with the Bartlett test. The identity of the distributions was evaluated with the kruskal test, and the groups compared pairwise with the Wilcox test for non-parametric statistical tests.