In many bacteria, mraY is found in the gene cluster with other mur genes and several genes involved in cell division (Boyle and Donachie, 1998; Mohammadi et al., 2007; Egan et al., 2020). In Anabaena, mraY (all4316) is surrounded by genes mostly encoding proteins of unknown function. The gene cluster includes 3 open-reading frames (ORFs) (Figure 1A). asl4317 and all4315 are found on the two sides of mraY, and all transcribed in the same direction (Figure 1A). The distance between the ORF of asl4317 and that of mraY is 78 bp, while the ORF of mraY and that of all4315 is separated by 404 bp.

According to the RNA seq data (Mitschke et al., 2011), each of the three genes possesses an independent promoter. To check if they could be cotranscribed, RT-PCR was performed, using oligonucleotide primers covering the intergenic regions. As shown in Figure 1B, co-transcript of asl4317 and mraY was identified by RT-PCR using primers P1/P2, but no signal was detected with primers P3/P4 (covering mraY and all4315) or P4/P5 (covering all three genes). This result suggests that mraY is cotranscribed with asl4317, but not with all4315. The RNA-seq data suggests that mraY may have a 10-times weaker promoter of its own compared with the shared one in front of asl4317 (Mitschke et al., 2011).

In order to make genetic analysis on the function of mraY, we initially tried to inactivate mraY via in-frame markerless deletion. However, no colonies were obtained on plate after conjugation, suggesting an essential function of mraY, consistent with previous reports of mraY in other bacteria (Boyle and Donachie, 1998; Bouhss et al., 1999; Pearcy et al., 2021). Using Cpf1-based gene editing technique (Niu et al., 2018), we created a conditional mutant of mraY (TRS-mraY) by replacing its weak and native promoter with a tunable synthetic riboswitch, R_TP, induced by theophylline (TP) (Nakahira et al., 2013; Figure 1A). When replaced by R_TP, the region corresponding to the binding site of the small RNA Yfr1, a negative regulator (Brenes-Álvarez et al., 2020), was deleted at the same time, making the expression of mraY only dependent on the translational control by R_TP. Translation of the mraY mRNA occurs normally in the presence of TP, but this process was blocked by the riboswitch when TP is removed. The obtained TRS-mraY mutant was first checked for gene segregation using specific primer targeting either the native promoter of mraY or R_TP. As shown in Figure 1C, for all three independent clones of strain TRS-mraY mutant, no WT copy could be detected using primers P6/P7, while a DNA fragment corresponding to the replaced promoter region could be amplified using primers P6/P8. These results indicate that mutants were fully segregated.

To test the growth capacity, TRS-mraY was first grown in the presence of 1 mM of TP in BG11 culture medium (with nitrate as nitrogen source), then equal culture volumes with comparable optical density were concentrated and spotted on BG11 or BG11₀ (deprived of combined nitrogen) plates with or without TP. After 4 and 7 days of incubation, the mutant spots were completely bleached compared with WT spots in the absence of TP, while cells maintained in the presence of 1 mM of TP were able to grow as indicated by the green color (Figure 2A). The mutant was also tested in liquid culture in flasks in both BG11 and BG11₀ (Figure 2B and Supplementary Figure 1). Similar result was observed that in the presence of 1 mM of TP, mutant cells were able to grow although slightly slower than the WT control, but cells without TP addition in the medium died completely (Figure 2B and Supplementary Figure 1). These results demonstrate that mraY is essential in Anabaena. The TRS-mraY strain could be maintained in the presence of 1 mM of TP as permissive conditions, and its phenotype could be analyzed upon transferring to non-permissive conditions by removal of TP from the growth medium.

To exclude any polar effect in TRS-mraY due to the replacement of the promoter region and further ascertain the essential function of mraY, complementation assays were carried out with two different plasmid constructs (Figure 3A). The first one, pP_mraY-mraY, carried mraY coding region placed behind the native strong promoter TSS 1 as depicted in Figure 1A; the second plasmid, pP_coaT-mraY, carried the coding region of mraY with a cobalt-inducible promoter of coaT (Peca et al., 2008). The strain carrying pP_mraY-mraY grows normally with or without TP in both BG11 and BG11₀. However, the strain with pP_coaT-mraY showed extensive cell lysis in the absence of any inducers (TP and Co²⁺), and this defect was corrected upon addition of TP that allows the translation of mraY from the chromosome or Co²⁺ that allows the expression of mraY from the replicative plasmid. In both complemented strains, heterocyst development could be seen, consistent with their growth under diazotrophic conditions in BG11₀ (Figure 3A).

To dissect the essential functions of mraY, we analyzed the morphology of cells and filaments of TRS-mraY under permissive and non-permissive conditions (Figure 3B). After 4 days of culture, the viability of TRS-mraY increased as the concentration of the inducer increased. With 1 or 4 mM of TP, the mutant grew almost as the WT, consistent with the growth curves as shown in Figure 2B. Without TP, or just with 0.1 mM of TP, extensive cell lysis, and filament fragmentation was found. In the presence of low concentrations of inducers, cell shape appeared to be different from that of the WT (Figure 3B).

We quantified cell shape changes as well as filament integrity in the WT and the TRS-mraY mutant in the presence of different concentrations of TP after 4 days of incubation. As shown in Figure 4, cell length based on analysis of 400 cells, displayed little changes with or without TP. However, cell width gradually increased as the concentrations of the inducers increased. In WT, the average width of the cells is about 2.84 μm. But with 0.1 mM of TP, the width of TRS-mraY increases to 4.92 μm. These changes translated into increased cell area in the conditional mutant incubated with low concentrations of TP as compared to the same strain with TP, or the WT (Figure 4C), which is also consistent with the microscopic images as shown in Figure 3. Consequently, under non-permissive conditions, cell shape of the mutant depleted of MraY became more rounded in shape compared to the ovoidal shape in WT.

Quantitative analysis demonstrates that filament integrity is strongly affected by the expression levels of MraY (Figure 4D and Supplementary Figure 2). While under non-permissive conditions, cells were completely lysed after 4 days of incubation, beginning with 0.1 mM of TP, some cells could survive after 4 days of culture. When either 0.2 or 0.4 mM of TP was added to induce the expression of mraY in the TRS-mraY mutant, some long filaments were found, but most of the filaments were still short with less than 50 cells. Compared to the WT in which most of the filaments had more than 100 cells per filament, TRS-mraY mutant had the majority of filaments with more than 100 cells per filament only when 1 mM of TP was added to the culture, indicating that this is the most suitable conditions to restore the phenotype of the TRS-mraY mutant. In the presence of 4 mM of TP, the TRS-mraY mutant had 35% of the filaments with more than 100 cells but also shorter filaments too, which may be resulted from overexpression of mraY. Thus, down regulation of mraY expression has a strong impact on cell shape, cell integrity and filament length.

Bacterial cell shape relies on PG synthesis under the control of the two protein complexes, the elongasome directing lateral PG synthesis along the side wall contributing to cell length, and the divisome for PG synthesis at the division site responsible for cell width (Szwedziak and Löwe, 2013). The increase in cell width of the TRS-mraY mutant under non-permissive conditions suggests enhanced PG synthesis activity at the division site. Therefore, we examined whether FtsZ-ring formation and PG biogenesis are altered in the TRS-mraY mutant under permissive and non-permissive conditions. For this, we constructed a ftsZ-cfp:TRS-mraY strain in which chromosomal copy of ftsZ was replaced by ftsZ-cfp fusion (Wang et al., 2021). The fluorescence of FtsZ-CFP (cyan fluorescent protein) was then observed. Under permissive condition (in the presence of 1 mM of TP), FtsZ-ring formation was similar to WT (Figure 5A). At 24 h after the removal of TP from the growth medium, extensive filament fragmentation and cell lysis occurred as already shown in Figure 3. Nevertheless, FtsZ-CFP localization could be found in short filaments, indicating that just before cell lysis, FtsZ was still targeted to midcell (Figure 5A). To visualize PG biosynthesis, we used HADA (7-hydroxycoumarin-amino-D-alanine), a fluorescent analog of the PG precursor D-Ala that can be incorporated into live cells through the transpeptidase activity (Hsu et al., 2017; Zhang et al., 2018). Under non-permissive conditions incubated for 24 h, we also observed the localization of HADA incorporation at midcell and cell-cell junctions (Figure 5B). In comparison to the WT control, most of the septa in the TRS-mraY strain under different conditions were wider, suggesting that cell division was not completed properly (Figure 5 and Supplementary Figure 3).

To analyze the localization pattern based on the images as shown in Figure 5, we quantified the percentage of cells that have FtsZ-CFP or HADA staining, either at midcell corresponding to active PG remodeling at the division site, or at cell-cell junctions reflecting PG precursor incorporation from the precedent cell cycle (Table 1 and Supplementary Table 4). In both WT and TRS-mraY mutant incubated in the presence of 1 mM of TP, 53.6 and 52.6% of the cells displayed FtsZ-CFP at midcell. However, 1 day after the removal of TP, only 30% of the cells showed FtsZ-CFP at midcell, indicating less cells initiating cell division without the induction of mraY, possibly due to the lethal effect of downregulation of mraY. In contrast, 44% of the cells retained FtsZ-CFP at cell-cell junctions, as compared to 23.7% of the cells of the same strain incubated with 1 mM of TP, a percentage similar to that of WT. Therefore, downregulation of mraY decreased initiation of cell division as indicated by the localization of FtsZ-CFP, but more FtsZ-CFP persisted at the cell-cell junctions than the controls (WT and mutant with TP), suggesting that the FtsZ targeted to the division sites at the precedent cycle was less likely to be disassembled at the cell-cell junctions in comparison to the control. Interestingly, even a concentration as low as 0.1 mM of the TP inducer could restore the capacity of the mutant in FtsZ-CFP localization at the midcell; however, persistent FtsZ-CFP localization at the cell-cell junction could be complemented only by the addition of 1 or 4 mM of TP; with decreasing concentrations of TP from 0.4 to 0.1 mM, more and more cell-cell junctions showed a FtsZ-CFP fluorescence. With 0.1 mM of TP, a condition allowing cells to grow, almost 52.9% of the cell-cell junctions have persistent FtsZ-CFP fluorescence, even higher than the mutant without TP, possibly because the mutant cell could no longer grow and lyse rapidly under such conditions (Table 1).

Although the TRS-mraY mutant failed to grow and rapidly lysed under non-permissive conditions (Figures 3–5), our data on FtsZ-CFP localization and HADA labeling in the TRS-mraY mutant in the presence of different concentrations of TP as inducer for mraY expression supported the results described above. With decreasing concentrations of the TP inducer, more and more cells with midcell HADA incorporation could be observed, indicating more persistent activity of PG synthesis at the division sites as mraY is downregulated. These observations provide an adequate explanation on the increase of cell width under similar conditions, as shown in Figures 3, 4.

In E. coli, elongasome and divisome share some common enzymes involved in PG synthesis such as PBP5, and components of the two complexes can even interact with each other (Mohammadi et al., 2007; Hugonnet et al., 2016). For example, MreB interacts with FtsZ in E. coli, and Lipid I and II synthesis enzymes including MraY are also found in both complexes (Hugonnet et al., 2016; Yoshii et al., 2019). Such interactions between the two complexes avoid competing PG synthesis activities and ensure cell growth and division in a spatiotemporal manner (Szwedziak and Löwe, 2013). Cyanobacterial divisome share some common components with those of other bacteria, but also has distinct features (Camargo et al., 2019; Xing et al., 2021). To understand the persistent PG synthesis activity following downregulation of MraY and its effect on the increase of cell width, we sought to determine whether MraY could interact directly with some components of the divisome or elongasome in Anabaena. For this purpose, MraY was expressed, respectively, in the two vectors of the bacterial two hybrid system (BACTH) and its interaction was tested with 31 proteins known to be involved in cell growth, division or elongation (Supplementary Table 1). MraY was found to interact with itself, as reported in other bacteria (White et al., 2010). As many cyanobacterial cell-division proteins (White et al., 2010; Camargo et al., 2019), MraY could interact with ZipN, a central scaffold of cyanobacterial divisome assembly. Another protein of the divisome, FtsQ, could also interact with MraY, at least when MraY was expressed from the T18 vector. In addition, we found that MraY also interacted with PBP1A (Figure 6), a component known as specific to the elongasome complex in E. coli. Interestingly, HetF, a newly identified component of the divisome in Anabaena necessary for heterocyst development, showed also interaction with MraY (Xing et al., 2021). HetF was required for septal PG synthesis under high-light conditions through interaction with FtsI. Thus, although we identified interaction between MraY and components of both elongasome and divisome (Figure 6), the interacting partners were different from those reported in other bacteria such as E. coli or Caulobacter crescentus (Mohammadi et al., 2007; White et al., 2010).

Next, we examined the effect on heterocyst development with different concentrations of inducer in the conditional mutant strain TRS-mraY. In the WT used as a control, mature heterocysts with distinct morphology were formed about 24 h after the deprivation of combined nitrogen (Figure 7). At 12 h after the heterocyst induction, proheterocysts, enlarged cells that can be weakly stained with alcian blue, an agent specific for heterocyst-specific polysaccharide layer, could be revealed. Under non-permissive conditions in TRS-mraY mutant, because of the extensive fragmentation, we could only examine proheterocyst formation with alcian blue staining at 12 h. At that time point, no alcian-blue stained cells with cell enlargement could be found. At 24 h, most cells lysed, and no single heterocysts which are usually more resistant to cell lysis, could be identified. However, even with a low level of TP (0.1 mM), heterocyst development could be found despite more and more cell fragmentation occurred with time. At 48 h, free and resistant heterocysts detached from the filament due to fragmentation could be observed. Heterocyst formation was found with the addition of different concentrations of TP tested, and the distribution of heterocyst pattern was similar to WT (Figure 8). However, a high level of aberrant hetercoysts were observed in the mutant, especially at 0.1 mM of TP in which such aberrant heterocysts account up to 16.5% of all heterocysts formed after 48 h of nitrogen starvation (Table 2). With the addition of 1 or 4 mM of TP, the frequency of such aberrant heterocysts reduced to only about 3.5%. When closely examined, some aberrant heterocysts appeared to be at various stages of cell division (Figure 7B). Indeed, some of them had a large septum just formed, while others were more or less constricted, with a long-neck connecting the two cells. These doublets rather resemble those already observed in a conditional mutant of polA encoding DNA polymerase I (PolI) under non-permissive conditions where problem in DNA segregation prevented septum closure (Xing et al., 2020).

All strains used in this study are listed in Supplementary Table 2. Anabaena strains were cultivated in BG11 (Stanier et al., 1971) or BG11₀ (BG11 without nitrate) in a shaker (30°C at 180 rpm) with illumination at 30 μmol photons m^–2 s^–1. In order to maintain TRS-mraY, the TRS-mraY mutant was cultivated and stored in BG11 or BG11₀ with 1 mM of theophylline (TP). 100 μg ml^–1 neomycin, or of 5 μg ml^–1 spectinomycin and 2.5 μg ml^–1 streptomycin, were added to the cultures when needed. Absorbance at 750 nm was measured at the indicated times upon inoculation in BG11 or BG11₀ medium to measure the growth curve of different strains.

RNA was isolated from filaments of Anabaena strains grown in BG11 medium for 5 days. RNA (20 ng) was used for reverse transcription with the Quantitec Reverse Transcription kit (Qiagen). To check if asl4317, all4316 (mraY), and all4315 could be cotranscribed, the obtained cDNA was used for RT-PCR, using oligonucleotide primers covering the intergenic regions. The primer pairs used were: Pall4316F280m/Pall4316R220, Pall4315F452m/Pall4315R48, Pall4315F1702m/Pall4315R48, respectively (Supplementary Table 3). The number of cycles at which the PCR reaction was in the exponential range was empirically determined. Samples were taken and analyzed by electrophoresis in agarose gels.

All oligonucleotide primers are listed in Supplementary Table 3. All employed plasmids (Supplementary Table 4) generated in this study were verified by Sanger sequencing. To construct TRS-mraY strain, plasmid pTRS-mraYR10m-sp was first generated in a similar way as previously described using pCpf1-sp plasmid (Niu et al., 2018; Xing et al., 2021). The repair template, in which the weak and native promoter of mraY was replaced by a tunable synthetic riboswitch inducible by TP, was generated by fusing the upstream and downstream region of the target sequence using primers listed in Supplementary Table 3 (Pall4316F970m/Pall4316R40m, Priboswitch2/Pall4316F1c/Pall4316R1140). The spacer sequence was designed according to the described rules (Niu et al., 2018) and prepared by annealing two complementary primers (cr_all4316R10mF/cr_all4316R10mR). To generate the plasmid that knocks out a specific sequence, the respective repair template and the spacer sequence were sequentially cloned into pCpf1-sp at the sites of BglII-BamHI and AarI-AarI. For plasmid pP_mraY-mraY, upstream (native promoter of asl4317 and mraY) and downstream (mraY) sequences were amplified from Anabaena gDNA using primers Pasl4317F243m/Pasl4317R1m and Pall4316F1f/Pall4316R1107b, respectively, and ligated into linearized pCT (using primers PpCT-R2979/PV_20) by Gibson assembly. For plasmid pP_coaT-mraY, upstream (Co²⁺ or Zn²⁺ inducible promoter from Synechocystis sp. PCC 6803) sequence was amplified from Synechocystis sp. PCC 6803 gDNA using primers Pslr0797F1193ma/Pslr0797R1mb, and downstream (mraY) sequence was amplified from Anabaena gDNA using primers Pall4316F1g/Pall4316R1107b, respectively, and these two fragments were ligated into linearized pCT (using primers PpCT-R2979/PV_20) by Gibson assembly.

Plasmid pTRS-mraYR10m-sp was transferred by conjugation to WT or the ftsZ-cfp strain with selection for Sm^r Sp^r to get TRS-mraY and ftsZ-cfp:TRS-mraY strain, respectively. Plasmids pP_mraY-mraY and pP_coaT-mraY were transferred by conjugation to TRS-mraY mutant strain with selection for Nm^r to get TRS-mraY:P_mraY-mraY and TRS-mraY:P_coaT-mraY strain, respectively.

The Bacterial Adenylate Cyclase Two-Hybrid System Kit (BATCH) based on the reconstitution of adenylate cyclase was used for testing protein-protein interaction (Battesti and Bouveret, 2012). 32 genes in this study were amplified with primers listed in Supplementary Table 3, and individual PCR products were assembled into linearized pUT18C and pKT25 vectors. All the resulting plasmids were verified by PCR and Sanger sequencing. The plasmids were co-transformed into strain BTH101, and the transformants were plated on solid LB medium containing 50 μg L^–1 ampicillin, 25 μg L^–1 kanamycin, 0.5 mM L^–1 IPTG, and 40 μg L^–1 5-bromo-4-chloro-3-indolyl-b-D-galactopyranoside (X-gal) to estimate the strength of interactions.

A Sdptop EX30 microscope was used to take bright-field images, and an Sdptop EX40 epifluorescence microscope was used to take fluorescence images. A filter [exciter (EX) 379–401, dichroic beamsplitter (DM) 420LP, emitter (EM) 435–485] was used to image HADA fluorescence (exposure time of 200 ms). A filter (EX426-446, DM455LP, EM460-500) was used to image CFP fluorescence (exposure time of 1 s). Fluorescence images were taken with an oil immersion lens objective (100/1.28). All images were processed using ImageJ without deconvolution. Cell length, cell width, cell area and filament length (cells per filament) were analyzed using the ImageJ software of microscopy images. Statistical tests and plotting of data were performed with the GraphPad Prim software or Origin.

For HADA labeling, Anabaena WT was grown in BG11 liquid medium, and TRS-mraY mutant was grown in BG11 liquid medium with 1 mM of TP. WT and TRS-mraY were washed three times in BG11 and incubated with 150 μM of HADA in BG11 without TP or with different concentrations of TP to observe PG synthesis for 24 h at standard growth conditions by microscopy.