A dynamic bactofilin cytoskeleton cooperates with an M23 endopeptidase to control bacterial morphogenesis

Bactofilins have emerged as a widespread family of cytoskeletal proteins with important roles in bacterial morphogenesis, but their precise mode of action is still incompletely understood. In this study, we identify the bactofilin cytoskeleton as a key regulator of cell growth in the stalked budding alphaproteobacterium Hyphomonas neptunium. We show that, in this species, bactofilin polymers localize dynamically to the stalk base and the bud neck, with their absence leading to unconstrained growth of the stalk and bud compartments, indicating a central role in the spatial regulation of cell wall biosynthesis. Database searches reveal that bactofilin genes are often clustered with genes for cell wall hydrolases of the M23 peptidase family, suggesting a functional connection between these two types of proteins. In support of this notion, we find that the H. neptunium M23 peptidase homolog LmdC interacts directly with bactofilin in vitro and is required for proper cell shape in vivo. Complementary studies in the spiral-shaped alphaproteobacterium Rhodospirillum rubrum again reveal a close association of its bactofilin and LmdC homologs, which co-localize at the inner curve of the cell, modulating the degree of cell curvature. Collectively, these findings demonstrate that bactofilins and M23 peptidases form a conserved functional module that promotes local changes in the mode of cell wall biosynthesis, thereby driving cell shape determination in morphologically complex bacteria.


Introduction
Bacteria come in a variety of different cell shapes, which can be further modified by the formation of cellular extensions such as branches or stalks (Kysela et al., 2016;Yang et al., 2016).Their morphology can change as a function of the cell cycle or in response to environmental cues to ensure optimal fitness in the given ecological niche or growth conditions (van Teeseling et al., 2017;Yang et al., 2016).In the vast majority of species, cell shape is determined by the peptidoglycan cell wall, a complex macromolecule composed of glycan chains that are crosslinked by short peptides (Typas et al., 2012;Vollmer et al., 2008).The synthesis of this mesh-like structure is achieved by an array of synthetic and lytic enzymes that are typically combined into multi-protein complexes and associated with regulatory factors and cytoskeletal elements to facilitate the coordination and spatiotemporal regulation of their activities (Egan et al., 2020;Rohs and Bernhardt, 2021).
Basic spherical, rod-like, and hyphal shapes are generated by the combined action of the cell division and cell elongation machinery (Margolin, 2009;Rohs and Bernhardt, 2021).In most bacteria, cell division is executed by the divisome, which is organized by the tubulin homolog FtsZ and mediates cell constriction and the synthesis of the new cell poles prior to cytokinesis (McQuillen and Xiao, 2020).Cell elongation, by contrast, can be achieved by various types of cell wall-biosynthetic complexes, including the so-called elongasome, which is organized by the actin homolog MreB and mediates the dispersed incorporation of new cell wall material along the lateral cell walls (Shi et al., 2018), or different cell pole-associated complexes that promote polar growth of the cell body (Brown et al., 2011).More complex cell shapes are generated with the help of accessory systems that either modulate the activity of the generic cell elongation machinery or have peptidoglycan biosynthetic activity on their own, thereby locally modifying the structure of the peptidoglycan layer (Taylor et al., 2019).
A particularly widespread family of cytoskeletal proteins implicated in cell shape modification are the bactofilins.They are characterized by a conserved central Bactofilin A/B domain (InterPro ID: IPR007607; Paysan- Lafosse et al., 2023) with a barrel-like β-helical fold that is typically flanked by short disordered terminal regions (Kühn et al., 2010;Shi et al., 2015;Vasa et al., 2015).Bactofilins polymerize spontaneously without the need for nucleotide cofactors (Koch et al., 2011;Kühn et al., 2010), driven by head-to-head and tail-to-tail interactions between the core domains of neighboring molecules (Deng et al., 2019).Lateral interactions between individual protofilaments can then give rise to higher-order assemblies, such as bundles or two-dimensional sheets (Kühn et al., 2010;Vasa et al., 2015;Zuckerman et al., 2015).Previous work has suggested that bactofilin polymers typically associate with the inner face of the cytoplasmic membrane and localize to regions of high membrane curvature (Caccamo et al., 2020;Hay et al., 1999;Kühn et al., 2010;Lin et al., 2017;Taylor et al., 2020).These structures have been co-opted as localization determinants and assembly platforms by several different morphogenetic systems.
For instance, bactofilin homologs were reported to contribute to rod-shape maintenance in Myxococcus xanthus (Koch et al., 2011), the establishment of helical cell shape in the human pathogen Helicobacter pylori (Sycuro et al., 2010;Taylor et al., 2020) as well as the modulation of cell helicity in the spiral-shaped bacterium Leptospira biflexa (Jackson et al., 2018).Apart from modulating general cell shape, they were found to have an important role in the formation of cellular extensions known as stalks, which are widespread among alphaproteobacterial species (Wagner and Brun, 2007).Stalks are elongated protrusions of the cell envelope that are filled with a thin thread of cytoplasm and grow through zonal incorporation of cell wall material at their base (Aaron et al., 2007;Randich and Brun, 2015).In Caulobacter crescentus and Asticcacaulis biprosthecum, bactofilin polymers were shown to localize to the stalk base to direct proper stalk formation.In C. crescentus, they recruit a cell wall synthase that contributes to stalk elongation, with their absence leading to a reduction in stalk length (Kühn et al., 2010).In A. biprosthecum, by contrast, bactofilin acts as a central topological regulator that is required to efficiently initiate stalk formation and limit peptidoglycan biosynthesis to the stalk base.Its absence leads to the development of pseudostalks, which are much shorter and wider than normal stalks and irregularly shaped, likely due to unrestrained peptidoglycan biosynthesis through the entire stalk envelope (Caccamo et al., 2020).
While stalks are often accessory structures with highly specialized functions (Klein et al., 2013;Persat et al., 2014;Wagner and Brun, 2007), stalked budding bacteria such as Hyphomonas neptunium and other members of the Hyphomonadaceae and Hyphomicrobiaceae use them as integral parts of the cell with key roles in cell growth and division (Moore, 1981).H. neptunium has a biphasic life cycle (Wali et al., 1980), in which a non-replicative, motile swimmer cell sheds its single polar flagellum and differentiates into a replicative, sessile stalked cell (Figure 1-figure supplement 1).Unlike most widely studied model species, it does not divide by binary fission but instead produces new offspring through the formation of buds at the tip of the stalk.As the terminal stalk segment gradually dilates, a flagellum is formed at the pole opposite the stalk.After DNA replication and translocation of one of the sister chromosomes through the stalk into the bud compartment (Jung et al., 2019), cytokinesis occurs at the bud neck, releasing a new swimmer cell.While the newborn cell first needs to differentiate into stalked cells to start replication, the stalked mother cell restores the stalk and then immediately re-enters the next budding cycle (Jung et al., 2019;Wali et al., 1980).The developmental program of H. neptunium involves several switches in the pattern of peptidoglycan biosynthesis (Cserti et al., 2017).After birth, swimmer cells increase in size by dispersed incorporation of new cell wall material throughout the cell body.Stalk formation is then achieved by zonal growth at the stalk base, followed by localized dispersed growth of the stalk-terminal bud compartment and, finally, zonal peptidoglycan synthesis at the site of cell division.The pattern of new cell wall synthesis is similar to the localization pattern of elongasome components (Cserti et al., 2017), suggesting the involvement of this machinery in all growth phases.However, the underlying regulatory mechanisms are still unknown.
In this study, we identify the bactofilin cytoskeleton as a central player in the regulation of cell growth of H. neptunium.We show that the lack of bactofilins causes severe morphological defects, resulting from unconstrained growth of the stalk and bud compartments.In line with this finding, bactofilin polymers localize dynamically to the stalk base and then to the incipient bud neck prior to the onset of bud formation, acting as a barrier that retains the cell wall biosynthetic machinery in the respective growth zones.Notably, in a broad range of species, bactofilin genes lie adjacent to genes encoding putative M23 peptidases.We show that the corresponding H. neptunium homolog, LmdC, is an essential bitopic membrane protein with peptidoglycan hydrolase activity that interacts directly with bactofilin through its N-terminal cytoplasmic tail.Its CRISPRi-mediated depletion also results in unconstrained cell growth, confirming its functional connection to bactofilin.The conservation of this interaction is further verified by studies of the spiral-shaped alphaproteobacterium Rhodospirillum rubrum.We show that the bactofilin and LmdC homologs of this species co-localize at the inner curve of the cell, forming filamentous assemblies that modulate the degree of cell curvature.Their interaction is abolished by amino acid exchanges in the cytoplasmic tail of LmdC, confirming a key role of this region in the recruitment of LmdC.Together, these results demonstrate a conserved functional interaction between bactofilins and M23 peptidases that is important for the control of cell growth in morphologically complex bacteria.

Results
The bactofilin cytoskeleton is required for cell shape determination in H. neptunium The H. neptunium chromosome contains two open reading frames that encode bactofilin homologs, HNE_0444 and HNE_2629 (Badger et al., 2006).Reciprocal BLAST searches identified HNE_2629 as a potential ortholog of C. crescentus BacA (43% identity, 60% similarity), whereas HNE_0444 was only distantly related (<30% identity) to other so-far characterized bactofilin homologs.Based on these results, we propose to designate the two bactofilin homologs of H. neptunium BacA (HNE_2629) and BacD (HNE_0444), respectively.bacA forms a putative bicistronic operon with lmdC, an essential gene encoding an M23 peptidase homolog (Cserti et al., 2017;Figure 1A).The two genes overlap by 17 base pairs, suggesting that their expression is closely coupled.bacD, by contrast, is not part of an operon.Both BacA and BacD display the typical architecture of bactofilins, with a central polymerization domain flanked by short non-structured N-and C-terminal regions (Figure 1B).
To investigate the role of bactofilins in H. neptunium, we generated mutant strains in which bacA and bacD were deleted either individually or in combination and analyzed their phenotypes (Figure 1).The ΔbacA and ΔbacAD cells showed severe morphological defects, as reflected by irregularly shaped, elongated and/or oversized cells, buds directly fused with the mother cell body, branched stalks, and multiple wide protrusions that emerged from the cells in an apparently random fashion (Figure 1C-E), reminiscent of the pseudostalks reported for a bactofilin-deficient A. biprosthecum mutant (Caccamo et al., 2020).Surprisingly, these growth defects did not have any major defects on the doubling time (Figure 1F).The wild-type phenotype could be largely restored by expressing an ectopic copy of bacA under the control of a copper-inducible promoter, even though BacA accumulated to lowerthan-normal levels under this condition, confirming the absence of polar effects (Figure 1 The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1.Raw images for the immunoblot analysis in Figure 1G.ΔbacA background .These results demonstrate that BacA has a critical role in the regulation of cell growth in H. neptunium, whereas BacD might be an auxiliary factor of so-far unknown function, similar to BacB in C. crescentus (Kühn et al., 2010).
To gain more insight into the dynamics of cell growth in the ΔbacA mutant and identify the initial phenotypic defects induced upon BacA depletion, we imaged a conditional bacA mutant after its transfer from permissive to restrictive conditions on an agarose pad (Figure 2A and B).Following cells at the swimmer-to-stalked cell transition, we observed that stalk formation initially proceeded as in the wild-type strain.However, as BacA was gradually depleted, stalk elongation ceased and the stalk structure started to widen and eventually develop multiple bulges that kept on expanding in an apparently uncontrolled manner.Thus, BacA appears to be required to maintain the polar growth zone at the stalk base, with its absence leading to unconstrained growth of the stalk cell wall.To follow the fate of bactofilin-deficient cells over a prolonged period of time, we monitored ΔbacAD cells in a flow-cell system, which ensured optimal nutrient supply throughout the course of the experiment and led to a looser packing of cells, thereby facilitating their visual analysis (Figure 2-figure supplement 1 and Figure 2-video 1).In this setup, cells again started growth by the formation of irregularly shaped stalks that started to branch, with branches developing either into extensive hyphal-like structures or large, amorphous compartments that may represent morphologically aberrant buds.At irregular intervals, cells divided at the junctions between hyphal and bulged cellular segments, releasing smaller amorphous fragments and, occasionally, also cells with close-to-normal morphology.Wild-type cells, by contrast, showed the usual growth pattern when cultivated under these conditions (Figure 2-video 2).Together, these results underscore the importance of BacA in the spatiotemporal control of cell growth in H. neptunium.
To obtain more detailed information about the dynamics of cell wall biosynthesis in the absence of bactofilins, we performed pulse-labeling studies with the fluorescent D-amino acid HADA, which is rapidly incorporated into peptidoglycan when added to the culture medium, thus marking regions of ongoing cell wall biosynthesis (Kuru et al., 2015).Wild-type cells showed the typical succession of growth modes, with dispersed growth in swimmer cells, zonal growth at the stalk base in stalked cells, and localized dispersed growth in the nascent bud (Figure 2C).In the ΔbacAD background, by contrast, this switch in the growth modes was abolished.Cells with close-to-normal morphology that had just initiated stalk formation still showed a distinct fluorescent focus at the stalked pole, suggesting that the initial recruitment of the machinery responsible for stalk formation occurred in a bactofilin-independent manner.However, all other cell types, including amorphous cells with aberrant stalk-and bud-like extensions, only displayed diffuse fluorescence, which points to uncontrolled growth by dispersed incorporation of new peptidoglycan throughout the entire cell envelope.These findings support the notion that the bactofilin cytoskeleton is required to limit cell wall biosynthesis to the different growth zones of H. neptunium.

The H. neptunium bactofilin homologs assemble into stable filamentous structures
Despite their severe cell shape defects, the ΔbacA and ΔbacAD mutants showed close-to-normal growth rates.Moreover, muropeptide analysis revealed that the global composition of their peptidoglycan remained essentially unchanged (Figure 2-figure supplement 2 and Supplementary file 1).These findings suggested that BacA might affect cell wall composition at a local scale or act mainly by modulating the activity of the generic cell elongation machinery.To further investigate the role of bactofilins in H. neptunium, we first aimed to verify the ability of the proteins to assemble into polymeric scaffolds.A model of the structure of BacA generated with AlphaFold-Multimer (Evans et al., 2022) suggested that the protein adopted a β-helical fold and polymerized through head-tohead and tail-to-tail interactions between different subunits (Figure 3A), as revealed previously in experimental studies (Deng et al., 2019;Shi et al., 2015).Consistent with this prediction, purified BacA produced a mixture of long filaments, filament bundles and two-dimensional polymeric sheets in vitro, which could be readily visualized by transmission electron microscopy (Figure 3B).Moreover, upon heterologous co-expression in Escherichia coli, fluorescently tagged derivatives of BacA and BacD co-localized into extended filamentous structures that were associated with the cell envelope (Figure 3C and     https://elifesciences.org/articles/86577/figures#fig2video2structures when produced alone, whereas the BacD fusion condensed into tight foci in the absence of BacA (Figure 3-figure supplement 1B and C).These results confirm the polymeric nature of bactofilin assemblies in H. neptunium and suggest that BacA and BacD interact with each other in vivo.

Bactofilins localize dynamically to the boundaries of the H. neptunium growth zones
Next, we sought to investigate the localization dynamics of the bactofilin cytoskeleton in vivo.To this end, we generated strains producing fluorescently tagged BacA or BacD derivatives in place of the native proteins (Figure 4-figure supplement 1).Time-lapse microscopy analysis of cells producing a BacA-YFP fusion on agarose pads (Figure 4A) revealed that the protein was localized to the new cell pole in swimmer cells and remained associated with the stalk base during the initial phase of stalk formation.At some point, it appeared to attach to the stalk structure and then move away from the base as new cell wall material was inserted.Subsequently, stalk elongation ceased and the terminal stalk segment, delimited by the bactofilin assembly, started to swell and develop into a bud, gradually displacing BacA-YFP in the direction of the stalk base as its size increased.After cell division, both the mother cell and the newborn swimmer cells showed a fluorescent focus, indicating that the bactofilin assembly was split during cell division.To monitor the dynamics of the bactofilin cytoskeleton over multiple division cycles, we analyzed the same strain in a flow-cell system (Figure 4-figure supplement 2 and Figure 4-videos 1-3).Under these conditions, the small, newborn swimmer cells were washed away immediately after cytokinesis, preventing the formation of microcolonies around the mother cell.We observed that cell division occurred at a small distance from the BacA-YFP focus, leaving a short stalk-terminal segment in between the bactofilin assembly and the stalk tip.After cytokinesis, the stalk elongated again prior to the initiation of the next budding event.Notably, BacA-YFP

Figure 4 continued on next page
was only occasionally detected at the stalked pole during the stalk restoration phase, suggesting that the bactofilin cytoskeleton no longer plays a major role in stalk growth once the stalk structure has been established.To verify the behavior observed in the time-lapse studies, we performed a population-wide analysis of the BacA-YFP localization pattern, based on snap-shot images of exponentially growing cells.A demographic analysis confirmed that the protein localizes to the new cell pole in swimmer cells, remains at the stalk base in cells with short stalks, and then moves to a position close to the distal end of the stalk before the onset of budding, remaining associated with the bud neck up to the point of cell division (Figure 4B).A very similar behavior was observed for a BacD variant fused to the yellow fluorescent protein Venus (Nagai et al., 2002;Figure 4C), consistent with the notion that the two bactofilin paralogs interact.In support of this hypothesis, studies of a strain in which both proteins were fluorescently labeled showed that BacA and BacD indeed co-localized to a large extent during all stages of the developmental cycle (Figure 4D and Figure 4-figure supplement 3).However, at the onset of bud formation, the signals of the two proteins were not always fully superimposable, with BacD assemblies occasionally detectable both at the bud neck and in the mother cell (Figure 4figure supplement 4).In the absence of BacA, BacD-Venus still formed distinct foci, albeit at apparently random positions within the cell (Figure 4-figure supplement 5).Notably, a BacA-YFP variant carrying a previously reported mutation that disrupts the polymerization interface (F130R) (Deng et al., 2019;Vasa et al., 2015;Zuckerman et al., 2015) was evenly distributed in the cell and unable to functionally replace the wild-type protein, indicating that the formation of polymeric assemblies is essential for proper BacA localization and function (Figure 4-figure supplement 6).Given that the bactofilin assemblies are localized to the boundaries of the stalk and the bud and required to prevent unconstrained growth of these cellular structures, we hypothesized that the bactofilin cytoskeleton might serve to limit the cell wall biosynthetic machinery to the different growth zones of H. neptunium.
To further investigate the role of bactofilin in the spatial regulation of cell wall biosynthesis, we generated derivatives of the wild-type and ΔbacAD strains that produced a fluorescently (YFP-) labeled variant of the membrane-bound elongasome protein RodZ, a core structural component of the complex connecting the cytoplasmic MreB cytoskeleton to the periplasmic enzymatic machinery (Bendezú et al., 2009;van den Ent et al., 2010;Alyahya et al., 2009).In the wildtype background, YFP-RodZ formed dynamic foci in the mother cell, at the (incipient) stalk base and the nascent bud (Figure 5 and Figure 5-videos 1 and 2), thus localizing to the different growth zones detected by HADA labeling (Figure 2C).These results are in line with previous findings identifying the elongasome as a key determinant of cell growth in H. neptunium (Cserti et al., 2017).Importantly, however, YFP-RodZ foci were never observed in the stalk, suggesting that elongasome complexes are unable to cross the stalk base and the bud neck, even though their free components appear to be able to diffuse through the stalk to facilitate the de novo assembly of elongasomes in the incipient bud compartment.In the ΔbacAD mutant, the localization behavior of YFP-RodZ resembled that in the wild-type background as long as the cells were the early stages of the cell cycle, with dynamic foci detectable throughout the cell envelope and at the (future) stalk base (Figure 5 and Figure 5-video 3).However, at later stages, the elongasome was no longer excluded from the nascent stalk structure and, in many cases, even appeared to be enriched this region, leading to its remodeling into an increasingly amorphous cellular extension (Figure 5 and Figure 5-videos 3-5).The bactofilin cytoskeleton itself or the high degree of positive inner cell curvature induced in its presence at the stalk base thus appear to act as a barrier that excludes elongasome complexes from the stalk and thus facilitates proper morphogenesis and bud development.The assembly state of BacA changes at different stages of H. neptunium development Since bactofilins form highly stable polymers in vitro (Kühn et al., 2010;Zuckerman et al., 2015), the dynamic localization observed for BacA and BacD was unexpected.To further characterize the dynamics of the bactofilin cytoskeleton in H. neptunium, we followed the movement of individual BacA-YFP molecules in swimmer, stalked, and budding cells (Figure 6-video 1).First, we used single-molecule tracking data to generate high-resolution images of the bactofilin assemblies in each of the three cell types.The results confirmed the cell cycle-dependent localization patterns observed by widefield fluorescence microscopy (Figure 6A and Figure 6-figure supplement 1).In addition, they revealed clusters in medial regions of the stalk, which may represent transient polymers formed during the reshuffling of bactofilin molecules between the stalk base and the (incipient) bud neck.When calculating the average mean squared displacement of the tracked BacA-YFP molecules, we found that their mobility decreased gradually from the swimmer over the stalked cell to the budding cell stage (Figure 6B).For each cell type, the distribution of step sizes in the single-particle tracks suggests the existence of two distinct diffusion regimes, with a static (D=0.02 ± 0.0004 µm 2 s -1 ) and a mobile (D=0.35 ± 0.004 µm 2 s -1 ) population, likely representing the polymerized and freely diffusible states, respectively (Figure 6C and  with the notion that it is impaired in polymerization but still able to assemble to some extent after undergoing a structural change that generates an alternative polymerization interface (Deng et al., 2019).Collectively, these results show that, in H. neptunium, the assembly state of the bactofilin cytoskeleton changes as cells progress through their developmental cycle.Moreover, despite the inherent stability of bactofilin polymers, all cell types display a sizeable fraction of mobile BacA-YFP molecules, which could potentially reflect the dynamic reorganization of bactofilin assemblies during cell growth.

Bactofilin genes are often clustered with genes encoding M23 peptidases
Having identified a critical role of BacA in H. neptunium morphogenesis, we set out to gain further insight into its mechanism of action.Notably, in several species in which bactofilins play a role in cell shape determination, the corresponding bactofilin genes are located adjacent to genes encoding putative M23 peptidases homologous to LmdC (Figure 7A).To determine whether this gene arrangement was more widely conserved, we performed a comprehensive bioinformatic analysis in which we searched all bacterial genome sequences available for putative bactofilin genes that were located immediately upstream or downstream of open reading frames encoding proteins with a predicted   M23 peptidase domain.This analysis identified a total of 226 species from a wide variety of bacterial phyla (Figure 7B and Figure 7-source data 1), suggesting a conserved functional association between bactofilins and M23 peptidases.

The M23 peptidase LmdC of H. neptunium has peptidoglycan hydrolase activity in vitro
Like many of its homologs, H. neptunium LmdC is a predicted bitopic membrane protein with a short N-terminal cytoplasmic region, a transmembrane helix, large periplasmic region composed of a predicted coiled-coil domain, and a C-terminal M23 peptidase domain (Figure 8A).Members of the M23 peptidase family usually have Zn + -dependent hydrolase activity and cleave bonds within the peptide side chains of the peptidoglycan meshwork (Firczuk et al., 2005;Grabowska et al., 2015).However, there are also various representatives that have lost their enzymatic activity because of mutations in residues required for metal cofactor binding and have instead adopted regulatory roles in cell wall biosynthesis (Figueroa-Cuilan et al., 2021;Goley et al., 2010;Gurnani Serrano et al., 2021;Möll et al., 2010;Poggio et al., 2010;Uehara et al., 2010).An amino acid alignment showed that the M23 peptidase domain of LmdC features all residues reported to be critical for Zn 2+  binding, suggesting that it could act as a genuine peptidoglycan hydrolase (Figure 8B).In a structural model of LmdC generated with AlphaFold2 (Jumper et al., 2021), the periplasmic coiled-coil and M23 peptidase domains form an elongated, rigid unit with two flanking non-structured regions that is flexibly linked to the transmembrane helix (Figure 8C).Notably, LmdC homologs were shown to form a distinct clade among the M23 peptidases that is broadly conserved among species but particularly enriched in alpha-and deltaproteobacteria (Figueroa-Cuilan et al., 2021).
To clarify whether LmdC was indeed catalytically active, we purified a C-terminal fragment of the protein including the M23 peptidase domain, and assayed its activity in vitro.Peptidoglycan from H. neptunium shows only a low degree of cross-linkage and hardly any pentapeptides (Cserti et al., 2017), After retrieval of the taxonomy IDs of the corresponding species from the National Center for Biotechnology Information (NCBI) website, a phylogenetic tree of the species was created using the iTOL server (Letunic and Bork, 2021).Abbreviations: DT (Deinococcus-Thermus group), FCB (Fibrobacteres-Chlorobi-Bacteroidetes group), β (β-proteobacteria), δ (δ-proteobacteria).The full list of species used and details on the genes identified are provided in Figure 7-source data 1.
The online version of this article includes the following source data for figure 7: Source data 1.List of bactofilin-M23 peptidase gene clusters identified in bacterial genomes.
likely due to a high level of autolytic activity.We, therefore, used normally crosslinked, pentapeptideenriched peptidoglycan from E. coli strain D456 (lacking PBPs 4, 5, and 6) (Edwards and Donachie, 1993) as a substrate to enable a comprehensive analysis of the cleavage specificity of LmdC.Upon treatment with the protein, the relative proportion of dimeric muropeptide species (Tetra-Tetra and Tetra-Penta) strongly decreased, regardless of the length of the crosslinked peptides and without the  formation of major additional monomer peaks (Figure 8D).The activity of LmdC was particularly high at a pH value of 5, but the physiological significance of this effect remains to be determined.Collectively, these results demonstrate that LmdC is a DD-endopeptidase cleaving the bond between the meso-diaminopimelic acid residue of one peptide and D-alanine at position 4 of the other peptide, thereby reducing the degree of cross-linkage within the peptidoglycan layer (Figure 8E).

LmdC is critical for morphogenesis in H. neptunium and physically interacts with BacA
To clarify whether LmdC was also required for proper growth in H. neptunium, we aimed to generate a mutant strain lacking LmdC activity.However, all attempts to delete the lmdC gene or the entire putative lmdC-bacA operon or to generate non-functional truncated lmdC alleles were unsuccessful, in line with a previous report suggesting that lmdC is essential for viability (Cserti et al., 2017).It was also not possible to generate a conditional lmdC mutant producing the gene under the control of a copper-inducible promoter, suggesting that its expression level needs to be precisely regulated.
To address this issue, we developed a CRISPR interference (CRISPRi) system (Larson et al., 2013) that enabled the inducible down-regulation of target genes in H. neptunium.It is based on a newly constructed integrative plasmid (pdCas9Entry) that carries the gene for a nuclease-deficient variant of Streptococcus pyogenes Cas9 (dCas9) under the control of a copper-inducible promoter as well as a cassette for the construction and constitutive expression of a gene-specific small guide RNA (sgRNA) (Figure 9A and Figure 9-figure supplement 1A).A derivative of this plasmid encoding an sgRNA directed against the 5' region of the lmdC gene was then inserted into the H. neptunium chromosome (Figure 9B).Induction with copper led to the formation of an sgRNA-dCas9 complex that targeted the non-coding strand of lmdC, generating a road-block that prevented the movement of RNA polymerase and, thus, gene expression (Figure 9C and D).Importantly, the block of lmdC expression led to cell shape defects very similar to those observed for the ΔbacA mutant, as reflected by a large proportion of distorted and amorphous cells, which still showed close-to-normal growth rates .Induced cells of a control strain carrying the empty pdCas9Entry plasmid did not show any obvious phenotype, indicating that the presence of dCas9 alone did not have any adverse effects on cell growth or viability (Figure 9D and E and Figure 9-figure supplement 1C).These results suggest that LmdC and BacA are part of the same morphogenetic pathway.
To further investigate whether the peptidoglycan hydrolase LmdC was functionally linked to bactofilins in H. neptunium, we aimed to conduct in vivo co-localization studies.However, all attempts to generate functional fluorescent protein fusions to full-length or C-terminally truncated LmdC failed, because the fusion proteins were either unstable or did not produce any fluorescence signal in vivo.As a complementary approach, we analyzed the interaction between LmdC and BacA in vitro by biolayer interferometry.To this end, a synthetic peptide comprising the predicted cytoplasmic region of LmdC (amino acids 1-38) was immobilized on a biosensor and probed with increasing concentrations of purified BacA protein.The results showed that BacA interacts with the LmdC peptide with an apparent equilibrium dissociation constant (K D ) of ~15 µM.By contrast, an unrelated peptide used as a negative control was not bound with appreciable affinity (Figure 9G and H).This result confirms a direct interaction between BacA and LmdC and identifies the N-terminal cytoplasmic region of LmdC as the interaction determinant for BacA association.

R. rubrum BacA recruits LmdC to the inner curve of the cell to modulate cell curvature
Our bioinformatic analysis suggests that bactofilins and M23 peptidases may be functionally associated in a large number of species.To further explore this possibility, we turned our efforts to the spiral-shaped bacterium Rhodospirillum rubrum, which contains a putative lmdC-bacA operon similar to that in H. neptunium (Munk et al., 2011;Figure 10A).The R. rubrum BacA homolog Rru_A1867 (BacA Rs ) is 37% identical (57% similar) to BacA of H. neptunium and also consists of a central Bactofilin A/B domain flanked by terminal non-structured regions (Figure 10B).The LmdC homolog Rru_A1868 (LmdC Rs ) is 39% identical (55% similar) to H. neptunium LmdC and has a similar predicted molecular structure (Varadi et al., 2022).The deletion of bacA Rs led to a noticeable increase in cell curvature (Figure 10C and Complementation studies showed that the expression of a plasmid-borne copy of bacA under the control of a weak constitutive promoter attenuated the curvature defect, albeit only partially, likely due to inadequate expression levels (Figure 10-figure supplement 2).The generation of an analogous plasmid to complement the deletion of lmdC Rs failed, because it was not possible to introduce the construct into Escherichia coli for plasmid propagation, suggesting that the expression of lmdC Rs was toxic to the cells.To further test this hypothesis, we constructed a plasmid enabling the expression of the gene under a tightly controlled, inducible promoter in E. coli.Upon induction, a large part of the cell population lysed, whereas control cells carrying an empty plasmid did not show any phenotypic defects (Figure 10-figure supplement 3).These results indicate that, similar to its H. neptunium homolog, LmdC Rs is an active enzyme with peptidoglycan hydrolase activity.
To further test the hypothesis that LmdC was functionally linked to bactofilin in R. rubrum, we set out to perform in vivo co-localization studies.However, all attempts to generate fluorescently labeled variants of full-length LmdC failed, because the corresponding R. rubrum expression plasmids appeared to be toxic to E. coli.Importantly, our in vitro studies had shown that the interaction of LmdC with the bactofilin cytoskeleton was mediated by its N-terminal cytoplasmic region (Figure 9G and H).To avoid the issues caused by the full-length protein, we, therefore, generated a reporter fusion encoding only the predicted N-terminal cytoplasmic region and the transmembrane helix (amino acids 1-80) of LmdC Rs fused to the green fluorescent protein mNeonGreen (Shaner et al., 2013) (LmdC N Rs -mNG) (Figure 10-figure supplement 4A-C).A low-copy plasmid carrying the corresponding allele under the control of the native lmdC Rs promoter was then introduced into an R. rubrum ΔlmdC Rs mutant producing a fully functional fusion of BacA Rs to the red fluorescent protein mCherry (BacA Rs -CHY) (Shaner et al., 2004) in place of the wild-type protein (Figure 10-figure supplement 2).We found that BacA Rs -CHY and LmdC N Rs -mNG were consistently localized to the same subcellular positions, forming patchy or filamentous structures that were preferentially, but not exclusively, positioned at the inner curve of the cell (Figure 10E).A very similar behavior was observed in the presence of wild-type LmdC Rs (Figure 10-figure supplement 4D).To further validate the enrichment of the two fusion proteins at the inner curve, we quantified their fluorescence signals along the inner and outer curves of multiple individual cells, concatenated the resulting fluorescence profiles, and ordered them according to their total length.In both cases, cells consistently displayed higher signal intensities at the inner curve (Figure 10F and G), supporting the notion that BacA Rs -LmdC Rs assemblies promote cell wall biosynthesis in the inner curve to reduce overall cell curvature.Notably, a BacA Rs -mNG fusion showed a reduced enrichment in the absence of LmdC Rs , whereas LmdC N Rs -mNG completely lost its asymmetric distribution in the ΔbacA Rs background (Figure 10H-J).These results identify BacA Rs polymers as recruitment platforms for LmdC Rs , although an interaction between the two proteins appears to be required for normal BacA Rs localization.
imaged by DIC and fluorescence microscopy.Bar: 2 µm.(F) Quantification of the proportion of phenotypically abnormal stalked and budding cells in the cultures of strain SP236 (lmdC::lmdC-HA pJH13) analyzed in panel E (n=100 cells per condition).(G) Bio-layer interferometric analysis of the interaction between cytoplasmic domain of LmdC and BacA.Sensors derivatized with a biotinylated synthetic peptide comprising the N-terminal cytoplasmic region of LmdC (amino acids 1-38) were probed with the indicated concentrations of BacA.After the association of BacA, the sensors were transferred to protein-free buffer to induce BacA dissociation.The interaction kinetics were followed by monitoring the wavelength shifts resulting from changes in the optical thickness of the sensor surface during association or dissociation of the analyte.The extent of non-specific binding of BacA to the sensor surface was negligible (control).(H) Affinity of the BacA-LmdC interaction.The maximal wavelength shifts measured in the experiment shown in panel G were plotted against the corresponding BacA concentration.The data were fitted to a one-site binding model, yielding an apparent K D value of ~15 µM.
The online version of this article includes the following source data and figure supplement(s) for figure 9: Source data 1.Raw images for the immunoblot analysis in Figure 9D.

LmdC recruitment depends on the conserved cytoplasmic β-hairpin motif
The N-terminal cytoplasmic region of LmdC Rs contains a β-hairpin motif (Figure 11A) that is conserved among LmdC homologs.To determine whether this structural feature was involved in BacA binding,     we exchanged two small clusters of amino acids that protruded from its surface and might thus serve as interaction determinants, using LmdC N Rs -mNG as a reporter.Cluster I (R25, H26, L27) was located within the first β-strand, whereas cluster II (R30, S31) was close to the β-turn (Figure 11B).The modifications in cluster I completely abolished the enrichment of LmdC at the inner curve of the cell, indicating that the β-hairpin motif is a critical part of the BacA binding site (Figure 11C and D).Notably, the loss of interaction between the two proteins also led to a reduced inner-curve enrichment of BacA Rs -CHY, as previously observed in the ΔlmdC Rs background (Figure 10I and J).This observation confirms that LmdC Rs binding is required for BacA to adopt its proper spatial organization.Similar defects were caused by the modification of cluster II (Figure 11C and E).However, in this case, BacA Rs -CHY was also no longer enriched at the inner curve, suggesting that the mutant protein may still show a residual, yet aberrant interaction with BacA Rs that has a negative effect on BacA Rs localization.Collectively, these findings confirm a functional association between the bactofilin cytoskeleton and LmdC Rs in R. rubrum.Moreover, they support the notion that bactofilins and LmdC-like M23 peptidases form a conserved morphogenetic module that is involved in the establishment of complex bacterial cell shapes.

Discussion
Although bactofilins are widely conserved among bacteria, their cellular functions are still incompletely understood.Here, we show that BacA homologs critically contribute to cell shape determination in two morphologically distinct members of the alphaproteobacteria, the stalked budding bacterium H. neptunium and the spiral-shaped bacterium R. rubrum.In both cases, they functionally interact with LmdC-type DD-endopeptidases to promote local changes in the pattern of peptidoglycan biosynthesis.Bactofilins thus complement the activities of the MreB and FtsZ cytoskeletons in the regulation of cell growth and act as versatile cell shape modifiers that facilitate the establishment of complex bacterial morphologies.
In H. neptunium, the bactofilin cytoskeleton localizes dynamically to the stalk base and the bud neck, promoting stalk growth and subsequent bud formation (Figure 4).While flanked by zones of active growth, the stalk itself is usually devoid of cell wall biosynthetic machinery (Cserti et al., 2017; Figure 5).In bactofilin-deficient cells, however, the stalk and bud growth zones are no longer confined and expand into the stalk envelope, leading to stalk widening and the formation of irregular bulges due to uncontrolled peptidoglycan incorporation and bud expansion (Figure 2 and Figure 5).Bulges or larger amorphous segments at some point separate from the mother cell and then continue to grow as independent cells, indicating that chromosome replication and segregation as well as cell division still occur under these conditions.Together, these observations suggests that bactofilin polymers establish barriers that normally prevent the entry of elongasome complexes from the mother cell or nascent bud compartments into the stalk.
The precise mechanism underlying the function of bactofilin in cell morphogenesis remains to be clarified.It is conceivable that bactofilin polymers act by tethering peptidoglycan biosynthetic proteins or establishing physical barriers that hinder the mobility of elongasome complexes (Figure 12A).However, we observed a close functional association of bactofilins with M23 peptidases that appears to be widely conserved among species, suggesting that their activity may be intimately tied to cell wall hydrolysis.The transition zones between the stalk and the adjacent mother cell and bud compartments are characterized by a high degree of positive inner cell curvature, which is in stark contrast to the negative curvature in the remaining parts of the cell.The hydrolytic activity of H. neptunium LmdC may be critical to remodel the cell wall in these zones and thus facilitate the extrusion of the stalk.Importantly, MreB is thought to move along regions of negative inner cell curvature (Hussain et al., 2018;Wong et al., 2019).The positively curved transition zones generated by the bactofilin-LmdC assemblies could, therefore, represent topological barriers that are difficult to cross by elongasome complexes, thereby helping to restrict their movement to the mother cell and bud compartments.Moreover, the narrow diameter stalks and their exceedingly high degree of negative inner cell curvature may reduce the stability of elongasome complexes that may form spontaneously in the stalk region.These effects may also be important during the stalk elongation phase following cell division, when bactofilin assemblies are no longer present at the stalk base.
Notably, a role in stalk formation has also been reported for the bactofilin homologs of C. crescentus and A. biprosthecum.In C. crescentus, the major bactofilin BacA was shown to act as a localization determinant for the bifunctional penicillin-binding protein PbpC, a cell wall synthase that contributes to stalk elongation (Kühn et al., 2010) and the proper sorting of a stalk-specific protein (Hughes et al., 2013).However, the absence of BacA only leads to a moderate reduction in stalk length and leaves overall cell shape unaffected, indicating that this protein has only a minor role in C. crescentus stalk formation.In A. biprosthecum, by contrast, the deletion of bacA completely abolishes stalk formation in rich medium and leads to the formation of wide cellular protrusions, named pseudostalks, under phosphate-limiting conditions (Caccamo et al., 2020).Similar to the amorphous extensions observed for the H. neptunium ΔbacAD mutant, these structures grow through disperse incorporation of new peptidoglycan and develop into viable offspring.Their formation was also attributed to the entry of cell wall biosynthetic proteins into nascent stalks (Caccamo et al., 2020), suggesting that the H. neptunium and A. biprosthecum BacA homologs share the same barrier function during stalk formation.However, in H. neptunium, another layer of regulation has been added in which bactofilins establish a second barrier close to the stalk tip that enables the formation of stalk-terminal buds.Bud expansion requires the relocation of elongasome complexes from the mother cell to the nascent bud at the onset of bud formation, as likely reflected by the fact that components of the elongasome (MreB, RodZ) and new cell wall biosynthesis can be detected within the stalk during a short time window at the end of the stalk elongation phase (Cserti et al., 2017).This process may be facilitated by the (partial) disassembly of the bactofilin complex at the stalk base and the diffusion of its components to the bud compartment, but the mechanistic details remain to be investigated.Another open question concerns the factors that control the dynamic localization of bactofilin to the stalked pole and the future bud neck in the terminal segment of the stalk.In A. biprosthecum, the recruitment of BacA assemblies to the sites of stalk biosynthesis is dependent on the cell wall hydrolase SpmX (Caccamo et al., 2020).However, an H. neptunium mutant lacking this protein shows normal morphology (Leicht et al., 2020), suggesting the existence of a different localization determinant.
Intriguingly, bactofilins not only mediate stalk formation and stalk-terminal budding but also the formation of curved cell shapes.We show that both BacA and LmdC are required to establish the normal degree of cell curvature in R. rubrum, with their absence leading to hyper-curved cells (Figure 10C and D).This result suggests that the two proteins form an accessory module that counteracts the activity of a so-far unknown system responsible for generating spiral cell shape in R. rubrum (Figure 12B).Cell curvature was shown to promote cell motility (Martínez et al., 2016).It is tempting to speculate that the expression of bacA and lmdC could be controlled in response to external cues to ensure optimal cell curvature and, thus, fitness in different environments.Our results show that R. rubrum cells contain multiple BacA-LmdC complexes that are distributed along the entire cell  (top) and ΔbacAD cells (bottom).In wild-type cells, BacA polymers form a complex with the DD-endopeptidase LmdC at the future stalked pole.The hydrolytic activity of LmdC, possibly stimulated by its interaction with BacA, helps to curve the cell wall at the incipient stalk base and thus determine stalk morphology.The zone of high positive inner cell curvature and, potentially, the physical barrier constituted by the bactofilin polymer prevent the movement of elongasome complexes from the mother cell body into the stalk, thereby limiting peptidoglycan biosynthesis to the stalked cell pole.At a later stage, the bactofilin-LmdC complex localizes close to the tip of the stalk and again generates a ring-shaped zone of positive cell curvature, thereby promoting the remodeling of the stalk tip into a spherical bud.At the onset of the budding process, elongasome complexes accumulate in the nascent bud by a so-far unknown mechanism.The positively curved bud neck and the bactofilin polymers prevent the movement of elongasome complexes from the bud into the stalk, thereby limiting cell growth to the bud compartment.In the ΔbacAD mutant, cells fail to concentrate LmdC at the future stalk base and bud neck.As a consequence, peptidoglycan biosynthesis is no longer limited to the different growth zones, leading to pseudo-stalk formation and unregulated bud expansion.(B) Bactofilin-mediated modulation of cell curvature in R. rubrum.BacA Rs (yellow) recruits LmdC Rs (red) to the inner curve of the cell.The hydrolytic activity of LmdC ultimately stimulates peptidoglycan biosynthesis at this position, leading to a reduction in cell helicity.
envelope but localize preferentially to the inner curve (Figure 10E and G).In this system, it is immediately obvious that BacA acts as a localization determinant for LmdC, since the preferential localization of LmdC to the inner curve is abolished in its absence (Figure 10I and J).To reduce cell curvature, the hydrolytic activity of LmdC must ultimately stimulate the insertion of new peptidoglycan at the inner curve of the cell, which increases the rate of cell elongation in this region and thus straightens the cell body.A similar mechanism may also be at work in the spirochete L. biflexa, where the loss of the bactofilin homolog LbbD, whose gene lies adjacent to a gene for putative M23 peptidase (LEPBI_I1430) (Figure 7A), was found to yield severely hyper-curved and more tightly twisted cells (Jackson et al., 2018).A different variation of this theme is found in H. pylori.There, the bactofilin homolog CcmA also forms multiple assemblies along the cell envelope, which interact with a membrane-spanning protein complex including an M23 peptidase (Csd1) homologous to LmdC (Sichel et al., 2022;Taylor et al., 2020).However, these assemblies are enriched at the outer curve of the cell, where they stimulate peptidoglycan synthesis to locally increase the rate of cell elongation over that at the inner curve, leading to twisting of the cell body.
Notably, in all systems characterized so far in molecular detail, the function of bactofilin-M23 peptidase complexes involves the stimulation of cell wall biosynthesis in a confined region of the cell envelope that entails a local change in cell envelope curvature.In H. neptunium, these complexes are only localized to narrow bands at the stalk base and bud neck, generating sharp bends at the transition zones between the stalk and the adjacent mother cell and bud compartments.In spiral-shaped bacteria, by contrast, they are scattered along the entire length of the cell, thereby establishing an elongated zone of increased longitudinal growth that changes overall cell curvature.The same mechanistic principle may also explain the morphogenetic role of bactofilins in so-far uninvestigated systems.
Collectively, our results underscore the role of bactofilins as versatile modulators of bacterial cell shape.In association with M23 peptidases, they form cell wall biosynthetic complexes that introduce local changes in cell envelope curvature.In doing so, they complement the activities of the tubulin and actin cytoskeletons in cell shape determination and thus expand the morphogenetic potential of bacteria, enabling the generation of complex cell shapes that go beyond the generic rod-like or spherical morphologies.It will be interesting to investigate the molecular function of bactofilins in a range of morphologically diverse species to obtain a comprehensive picture of the functionalities provided by these cytoskeletal proteins and the conservation of their mode of action.

Media and growth conditions
H. neptunium LE670 (ATCC 15444) (Leifson, 1964) and its derivatives were grown in Artificial Sea Salt (ASS) medium at 28 °C under aerobic conditions, shaking at 210 rpm in baffled flasks.ASS medium was composed of 0.5% Bacto Peptone (Thermo Fisher Scientific, USA), 0.1% yeast extract, 1 mM MgSO 4 , 0.5 mM CaCl 2 and 1.5% Instant Ocean Sea Salt (Spectrum Brands, USA), dissolved in deionized water.For solid media, 1.5% agar was added prior to autoclaving.When appropriate, antibiotics were used at the following concentrations (µg/mL in liquid/solid medium): rifampicin (1/2), kanamycin (100/200).The expression of genes under the control of the copper-inducible P Cu promoter or the zinc-inducible P Zn promoter (Jung et al., 2015) was induced by the addition of CuSO 4 or ZnSO 4 to the concentrations indicated in the text.To assess the growth of H. neptunium, cells were grown to exponential phase, diluted in fresh medium to an optical density at 580 nm (OD 580 ) of 0.02, and transferred into 24-well polystyrene microtiter plates (Becton Dickinson Labware, USA).Growth was then followed at 32 °C under double-orbital shaking in an EPOCH 2 microplate reader (BioTek, USA) by measuring the OD 580 at 20 min intervals.

Determination of growth rates
Exponentially growing cultures were diluted to an OD 580 of ~0.01, transferred to a 96-well plate, and incubated at 30 °C with double-orbital shaking at 355 cpm in an Epoch 2 microplate reader (BioTek, Germany).Growth was monitored by measuring the OD 600 at 10 min intervals until the stationary phase was reached.OD 600 measurements obtained in the exponential phase were used to calculate the growth rate µ (increase in OD 600 per hour).The doubling time t d was then calculated using the equation t d = ln(2)/µ.

Plasmid and strain construction
The bacterial strains, plasmids, and oligonucleotides used in this study are listed in Supplementary file 3. E. coli TOP10 (Thermo Fisher Scientific, USA) was used as a host for cloning purposes.All plasmids were verified by DNA sequencing.H. neptunium and R. rubrum were transformed by conjugation using the DAP-auxotrophic strain E. coli WM3064 as a donor (Jung et al., 2015).The integration of non-replicating plasmids at the chromosomal P Cu or P Zn locus of H. neptunium was achieved by single homologous recombination (Jung et al., 2015).Gene replacement in H. neptunium and R. rubrum was achieved by double-homologous recombination using the counter-selectable sacB marker (Cserti et al., 2017).Proper chromosomal integration or gene replacement was verified by colony PCR.

Live-cell imaging
Cells of H. neptunium and R. rubrum were grown to exponential phase in the appropriate medium and, if suitable, induced with CuSO 4 prior to imaging.For depletion experiments, cells were grown in the presence of inducer, washed three times with inducer-free medium, and then further cultivated in the absence of inducer for the indicated period of time prior to analysis.To acquire still images, cells were transferred onto 1% agarose pads (in water).For time-lapse analysis, cells were immobilized on pads made of 1% agarose in ASS medium, and the cover slides were sealed with VLAP (1:1:1 mixture of vaseline, lanolin, and paraffin) to prevent dehydration.Imaging was performed using a Zeiss Axio Imager.M1 microscope equipped with a Zeiss alpha Plan-Apochromat 100 x/1.40Oil DIC objective and a pco.edge 3.1 sCMOS camera (PCO, Germany) or a Zeiss Axio Imager.Z1 microscope equipped with a Zeiss alpha Plan-Apochromat 100 x/1.46Oil DIC M27 or a Plan-Apochromat 100 x/1.40Oil Ph3 M27 objective and a pco.edge 4.2 sCMOS camera (PCO, Germany).An X-Cite 120PC metal halide light source (EXFO, Canada) and ET-DAPI, ET-YFP, or ET-TexasRed filter cubes (Chroma, USA) were used for fluorescence detection.Microfluidic experiments were performed using a CellASIC ONIX EV262 Microfluidic System, equipped with an F84 manifold and B04A microfluidic plates (Merck Millipore, Germany), which were flushed with PBS buffer for 30 min before usage.Exponentially growing cells were flushed into the flow cells, cultivated under continuous medium flow (0.4 ml/h), and imaged at regular intervals using the Axio Image.Z1 microscope described above.To visualize sites of ongoing peptidoglycan synthesis, cells were grown to the exponential phase, incubated for 1 min with 1 mM HADA, and washed four times in ASS medium prior to imaging.
To analyze the distribution of fluorescently tagged BacA and LmdC variants along the outline of R. rubrum cells, the fluorescence intensity along the outer and inner curve was determined for 200 cells per strain using Fiji (Schindelin et al., 2012).After concatenation of the two fluorescence profiles obtained for each cell, the combined profiles were plotted as demographs, using R 4.2.3 (R Development Core Team, 2012) and the Cell-Profiles package (https://github.com/ta-cameron/Cell-Profiles/wiki/Vignette; Cameron et al., 2014).

Single-particle tracking and diffusion analysis
Cells were cultivated overnight in ASS medium, transferred into fresh medium, and grown to exponential phase prior to imaging by slimfield microscopy (Plank et al., 2009).In this approach, the back aperture of the objective is underfilled by illumination with a collimated laser beam of reduced width, generating an area of ~10 μm in diameter with a light intensity high enough to enable the visualization of single fluorescent protein molecules at very high acquisition rates.The single-molecule level was reached by bleaching most molecules in the cell for 100-1,000 frames, followed by tracking of the remaining and newly synthesized molecules for ~3,000 frames.Images were taken at 30 ms intervals using an Olympus IX-71 microscope equipped with a UAPON 100 x/ NA 1.49 TIRF objective, a backilluminated electron-multiplying charge-coupled device (EMCCD) iXon Ultra camera (Andor Solis, USA) in stream acquisition mode, and a LuxX 457-100 (457 nm, 100 mW) light-emitting diode laser (Omicron-Laserage Laserprodukte GmbH, Germany) as an excitation light source.The laser beam was focused onto the back focal plane and operated during image acquisition with up to 2 mW (60 W/cm 2 at the image plane).Andor Solis 4.21 software was used for camera control and stream acquisition.Prior to analysis, frames recorded before reaching the single-molecule level were removed from the streams, using photobleaching curves as a reference.Subsequently, the streams were cropped to an equal length of 2000 frames and the proper pixel size (100 nm) and time increment were set in the imaging metadata using Fiji (Schindelin et al., 2012).Single particles were tracked with u-track 2.2 (Jaqaman et al., 2008).Trajectories were only considered for further statistical analysis if they had a length of at least five steps.Data analysis was performed using SMTracker 2.0 (Oviedo-Bocanegra et al., 2021).An estimate of the diffusion coefficient and insight into the kind of diffusive motion exhibited were obtained from mean-squared-displacement (MSD)-versus-time-lag curves.In addition, the frame-to-frame displacements of all molecules in x and the y direction were fitted to a twopopulation Gaussian mixture model to determine the proportions of mobile and static molecules in each condition (Oviedo-Bocanegra et al., 2021).

Transmission electron microscopy
To image H. neptunium cells, 10 μl of early exponential cell cultures were applied onto glow-discharged electron microscopy grids (Formvar/Carbon Film on 300 Mesh Copper; Plano GmbH, Germany) and incubated for 1 min at room temperature.The grid was manually blotted with Whatman filter paper to remove excess liquid.Subsequently, the cells were negatively stained for 5 sec with 5 μl of 1% uranyl acetate.After three washes with H 2 O, the grids were dried and imaged in a 100 kV JEM-1400 Plus transmission electron microscope (JEOL, USA).To image BacA polymers, purified BacA-His6 was dialyzed against low-salt buffer (50 mM HEPES pH 7.2, 10 mM NaCl, 5 mM MgCl2, 0.1 mM EDTA, 1 mM ß-mercaptoethanol) for 16 hr.The protein was spotted onto carbon-coated grids and allowed to settle for 2 min.The grids were blotted dry, stained with 1:2 diluted supernatant of saturated 2% uranyl acetate (in H 2 O) for 1 min, dried and imaged using a Zeiss CEM902 electron microscope, operated at 80 kV and equipped with a 2048 × 2048 pixel CCD camera.Image processing was carried out using Photoshop CS2 and Illustrator CS5 (Adobe Systems, USA).

LmdC enzymatic activity assay
To test the enzymatic activity of LmdC, peptidoglycan from E. coli D456 (Edwards and Donachie, 1993) was mixed with 10 µM LmdC in buffer A (20 mM Hepes/NaOH pH 7.5, 50 mM NaCl, 1 mM ZnCl 2 ) or buffer B (20 mM sodium acetate pH 5.0, 50 mM NaCl, 1 mM ZnCl 2 ) in a final volume of 50 µL and incubated at 37 °C for 16 hr in a thermal shaker set at 900 rpm.A mixture of peptidoglycan in buffer B without LmdC was used as a control.The reactions were stopped by heating at 100 °C for 10 min.The peptidoglycan was then further digested overnight with cellosyl, and the reactions were stopped by heating at 100 °C for 10 min.After centrifugation of the samples at 14,000 rpm for 10 min, the supernatants were recovered, reduced with sodium borohydride, acidified to pH 4.0 -pH 4.5 with dilute 20% phosphoric acid and subjected to HPLC analysis as previously described (Glauner, 1988).

Bio-layer interferometry
Bio-layer interferometry analyses were conducted using a BLItz system equipped with High Precision Streptavidin (SAX) Biosensors (Sartorius, Germany).As a ligand, a custom-synthetized N-terminally biotinylated peptide comprising residues Met1 to Gln38 of LmdC (GenScript, USA) was immobilized on the biosensors.After the establishment of a stable baseline, association reactions were monitored at various analyte concentrations.At the end of each binding step, the sensor was transferred into an analyte-free buffer to follow the dissociation kinetics.The extent of non-specific binding was assessed by monitoring the interaction of analyte with unmodified sensors.All analyses were performed in BLItz binding buffer (25 mM HEPES/KOH pH 7.6, 100 mM KCl, 10 mM MgSO 4 , 1 mM DTT, 10 mM BSA, 0.01% Tween).

Bioinformatic analysis
Protein similarity searches were performed with BLAST (Altschul et al., 1990), using the BLAST server of the National Institutes of Health (https://blast.ncbi.nlm.nih.gov/Blast.cgi).Transmembrane helices were predicted with DeepTMHMM (Hallgren et al., 2022), coiled-coil regions with PCOILS (Lupas, 1996).The positions of conserved functional domains were determined using the PFAM server (Mistry et al., 2021).AlphaFold2 (Jumper et al., 2021) and AlphaFold-Multimer (Evans et al., 2022), as implemented in the AlphaFold.ipynb notebook on Google Colab, were used to predict the tertiary or quaternary structure of proteins, respectively.SuperPlots (Lord et al., 2020) were used to visualize cell length distributions and to evaluate the statistical significance of differences between multiple distributions, employing the PlotsOfData web app (Postma and Goedhart, 2019).

Statistics
All experiments were performed at least twice independently with similar results.No data were excluded from the analyses.Statistical analyses were performed using Excel 2019 (Microsoft), unless indicated otherwise.To quantify imaging data, multiple images were analyzed per condition.The analyses included all cells in the images or, in the case of high cell densities, all cells in a square portion of the images.The selection of the images and fields of cells analyzed was performed randomly.

Material availability
All biological materials generated in this study are available from the corresponding author upon request.

Figure 1 .
Figure 1.The bactofilin homolog BacA is required for proper cell morphology in H. neptunium.(A) Schematic representation of the two bactofilin genes present in the H. neptunium genome.bacA lies adjacent to the M23 peptidase gene lmdC.Arrows indicate the direction of transcription.(B) Domain organization of BacA and BacD from H. neptunium.The bactofilin polymerization domain (colored boxes) is flanked by non-structured N-and C-terminal regions.(C) Morphology of H. neptunium bactofilin mutants.Shown are representative cells of strains LE670 (wild type), EC28 (ΔbacA), EC23 (ΔbacD), and EC33 (ΔbacAD), imaged by differential interference contrast (DIC) microscopy.Bar: 4 µm.(D) Transmission electron micrographs of representative ΔbacA cells at the early stalked-cell stage.Bar: 1 µm.(E) Quantification of the proportion of phenotypically abnormal stalked and budding cells in cultures of the strains analyzed in panel C (n=100 cells per strain).(F) Doubling times of the indicated H. neptunium strains.Bars represent the mean (± SD) of three independent experiments (dots).Statistical significance was determined using an unpaired two-sided Welch's t-test (*p<0.05;n.s., not significant).(G) Immunoblot analysis of the strains shown in panel C, performed using anti-BacA antibodies.

Figure supplement 2
Figure supplement 2-source data 1.Raw images for the immunoblot analysis in Figure 1-figure supplement 2A.

Figure 2 .
Figure 2. Lack of BacA leads to uncontrolled growth of the stalk and bud compartments.(A) Immunoblot showing the levels of BacA in strain EC41 (ΔbacA P Cu -bacA) over the course of BacA depletion.Cells were grown in coppercontaining medium, washed, and transferred to inducer-free medium.At the indicated time points, samples were withdrawn and subjected to immunoblot analysis.Strains LE670 (wild-type) and EC28 (ΔbacA) were included as controls.The position of BacA is indicated.(B) Morphological defects induced by BacA depletion.Cells of strains LE670 (wild-type) and EC41 (ΔbacA P Cu -bacA) were grown in medium containing 0.5 mM CuSO 4 , washed, and incubated for 6 hr in inducer-free medium before they were transferred onto ASS-agarose pads lacking inducer and imaged at the indicated time points.Shown are representative images.Bar: 2 µm.(C) Changes in the growth pattern of H. neptunium in the absence of bactofilins.Cells of strains LE670 (wild-type), and EC33 (ΔbacAD) were stained with the fluorescent D-amino acid HADA prior to analysis by fluorescence microscopy.Shown are representative images of cells at different developmental stages.Bars: 2 µm.The online version of this article includes the following video, source data, and figure supplement(s) for figure 2: Source data 1.Raw images for the immunoblot analysis in Figure 2A.

Figure supplement 1 .
Figure supplement 1.Growth of ΔbacAD cells in a microfluidic flow cell.

Figure 3 .Figure 4 .
Figure 3. BacA and BacD assemble into filamentous structures.(A) Model of a BacA trimer generated with AlphaFold-Multimer (Evans et al., 2022).Only residues P34-D138 are shown for each subunit.(B) Visualization of BacA polymers.Purified BacA-His 6 was stained with uranyl acetate and imaged by transmission electron microscopy (TEM).Arrowheads point to BacA filaments.Asterisks indicate filament bundles and sheets.Bars: 200 nm.(C) Copolymerization of BacA and BacD after heterologous co-expression in E. coli.Cells of E. coli BL21(DE3) transformed with plasmid pEC121 (P T7 -bacA-eyfp P T7 -bacD-ecfp) were grown in LB medium containing 5% glucose and induced with 0.5 mM isopropyl-β-D-1-thiogalactopyranoside (IPTG) prior to imaging.Shown are representative cells.Bar: 3 µm.The online version of this article includes the following figure supplement(s) for figure 3:Figure supplement 1.Localization patterns of BacA-YFP and BacD-CFP after heterologous overproduction in E. coli.

Figure supplement 1
Figure supplement 1-source data 1.Raw images for the immunoblot analyses in Figure 4-figure supplement 1A and B.

Figure supplement 4 .
Figure supplement 4. Partial independence of BacA and BacD localization at the onset of the budding process.

Figure supplement 5 .
Figure supplement 5. Random localization of BacD complexes in the absence of BacA.

Figure 5 .
Figure 5.The bactofilin cytoskeleton is required to confine the elongasome to the mother cell and bud compartments.Localization of YFP-RodZ in the H. neptunium wild-type and ΔbacAD backgrounds.Shown are representative cells of strain EC93 (rodZ::eyfp-rodZ) and SP221 (ΔbacAD rodZ::eyfp-rodZ) imaged by DIC and fluorescence microscopy and arranged according to their developmental state.Arrowheads indicate cellular extensions that show YFP-RodZ foci.Bar: 2 µm.The online version of this article includes the following video(s) for figure 5:

Figure 6 .
Figure 6.The single-particle dynamics of BacA change over the course of the cell cycle.(A) Representative heat maps showing the sum of single-particle positions observed in live-cell single-particle localization microscopy studies of a swimmer, stalked, and budding cell producing BacA-YFP (EC61) or a mutant cell producing the polymerization-deficient BacA F130R -YFP variant (MO78).Insets show the corresponding bright-field images.Bars: 1 µm.(B) Mean-squared displacement (MSD) analysis of the mobility of BacA-YFP in wild-type swimmer (n=100), stalked (n=105), and budding (n=101) cells (EC61).Cells (n=110) producing the BacA F130R -YFP were analyzed as a control (MO78).Error bars indicate the standard deviation.(C) Bubble plots showing the proportions of mobile and immobile particles (size of the bubbles) and the corresponding average diffusion constants (y-axis) in the cells analyzed in panel B. The online version of this article includes the following video and figure supplement(s) for figure 6:Figure supplement 1. Single-particle tracking analysis of BacA-YFP.

Figure 7 .
Figure7.Putative operons of including adjacent BacA and M23 peptidase genes are widely conserved among bacteria.(A) Arrangement of bacA and lmdC genes in species whose BacA homolog has a proven role in cell morphogenesis.(B) Co-conservation of BacA and M23 peptidase genes across different bacterial phyla.Bioinformatic analysis was used to identify genomes that contained a bactofilin gene located next to an M23 peptidase gene.After retrieval of the taxonomy IDs of the corresponding species from the National Center for Biotechnology Information (NCBI) website, a phylogenetic tree of the species was created using the iTOL server(Letunic and Bork, 2021).Abbreviations: DT (Deinococcus-Thermus group), FCB (Fibrobacteres-Chlorobi-Bacteroidetes group), β (β-proteobacteria), δ (δ-proteobacteria).The full list of species used and details on the genes identified are provided in Figure7-source data 1.

Figure 8 .
Figure 8. LmdC is a peptidoglycan hydrolase with DD-endopeptidase activity.(A) Predicted domain architecture of H. neptunium LmdC.The predicted positions of the transmembrane helix (TM), the three coiled-coil regions (CC), and the M23 peptidase domain are indicated.(B) Alignment of the amino acid sequences of multiple M23 peptidases showing the conservation of the catalytic residues in LmdC.Residues required to coordinate the catalytic Zn 2+ ion of LytM from S. aureus (Firczuk et al., 2005) are indicated by arrowheads.The proteins shown are EnvC from E. coli (UniProt: P37690), NlpD from E. coli (P0ADA3), LytM from S. aureus (O33599), MepM from E. coli (P0AFS9) and LmdC from H. neptunium (Q0BYX6).(C) Predicted molecular structure of H. neptunium LmdC, generated with Alphafold2 (Jumper et al., 2021).The different domains of the protein and the position of proline 66, which terminates the N-terminal fragment of LmdC used for the in vivo analyses in this study (LmdC 1-66 ), are indicated.(D) Representative HPLC traces showing the muropeptide profile of peptidoglycan treated with LmdC.Cell walls were incubated with the isolated M23 peptidase domain of LmdC at pH 5.0 or pH 7.5.Subsequently, muropeptides were released with cellosyl, reduced, and separated by HPLC.A sample lacking LmdC (No protein) was analyzed as a reference.The nature of the major products is indicated above the peaks.Tri, Tetra, and Penta stand for N-acetylglucosamine-N-acetylmuramitol tripeptide, tetrapeptide, and pentapeptide, respectively.(E) Structure of a Tetra-Tetra muropeptide.Abbreviations: G (N-acetylglucosamine), M (N-acetylmuramic acid).The cleavage site of LmdC is indicated by a red arrowhead.

Figure 9 .
Figure9.LmdC interacts with BacA to mediate H. neptunium morphogenesis.(A) Schematic representation of the integrative plasmid pdCas9Entry, enabling CRISPR interference (CRISPRi) in H. neptunium.The plasmid harbors a chromosomal fragment containing the copper-inducible P Cu promoter, followed by a gene encoding a nuclease-deficient variant of Cas9 (dCas9).It also contains an sgRNA expression cassette composed of the constitutive P HNE0038 promoter, a BbsI cloning site for the insertion of target-specific DNA fragments, a sequence encoding an optimized small guide RNA (sgRNA) handle(Chen et al., 2013), and a transcriptional terminator (ter).Other features of the plasmid are a replication initiator gene (rep), a kanamycin resistance gene (npt1), an origin of transfer (oriT), and a strong transcriptional terminator (rrnBT1T2) that is placed upstream of the P Cu -containing fragment.The DNA sequence of the sgRNA expression cassette is given in Figure9-figure supplement 1. (B) Schematic showing the integration of a pdCas9Entry derivative into the H. neptunium chromosome by single homologous recombination at the P Cu locus.(C) Mechanism of the CRISPRimediated repression of lmdC expression.In the presence of inducer, dCas9 interacts with a constitutively produced sgRNA that targets bases 51-72 of the non-coding strand of lmdC (sglmdC).The formation of the interference complex creates a roadblock that blocks the transcription of lmdC by RNA polymerase.(D) Immunoblot analysis verifying the CRISPRi-mediated depletion of LmdC.Strain SU34 (lmdC::lmdC-HA), which produces a fully functional variant of LmdC carrying the HA affinity tag inserted in between the transmembrane helix and the coiled-coil domain, was transformed with pdCas9Entry (empty vector) or plasmid pJH13 (sglmdC).The resulting strains (SP249 and SP236) were grown for 24 hr in the absence (-) or presence (+) of 0.3 mM CuSO 4 and subjected to immunoblot analysis with an anti-HA antibody.The wild-type strain (LE670), the parental strain SU34 as well as the uninduced pre-cultures (pre) of strains SP249 and SP236 used for the experiment were analyzed as controls.A non-specific band served as a loading control (LC).(E) Morphology of H. neptunium cells depleted of LmdC.Shown are representative cells of strains SP236 (lmdC::lmdC-HA pJH13) (sglmdC) and SP249 (lmdC::lmdC-HA pCas9Entry) (empty vector) grown for 24 hr in the absence (control) or presence (+dCas9) of 0.3 mM CuSO 4 .Cells were Figure 9 continued on next page

Figure supplement 1 .
Figure supplement 1. Characterization of the phenotype induced by the CRISPRi-mediated depletion of LmdC in H. neptunium.
Figure 9 continued

Figure 10 .
Figure 10.BacA and LmdC contribute to cell morphogenesis in R. rubrum.(A) Schematic representation of the bacA-lmdC operon in R. rubrum.(B) Domain organization of LmdC and BacA from R. rubrum.LmdC Rs consists of a transmembrane helix (TM) followed by a coiled-coil-rich region and a C-terminal M23 peptidase domain (M23).The bactofilin polymerization domain (green box) is flanked by non-structured N-and C-terminal regions.Numbers indicate the first and last amino acid of a domain.(C) Phenotypes of R. rubrum wild-type (S1), ΔbacA Rs (SP70), ΔlmdC Rs (SP68), and ΔbacA Rs ΔlmdC Rs (SP116) cells, imaged using phase-contrast microscopy.Bar: 5 µm.(D) Superplots showing the distribution of cell sinuosities in populations of the indicated R. rubrum strains.Small dots represent the data points, large dots represent the median values of the three independent experiments shown in the graphs (dark blue, light blue, gray).The mean of the three median values is indicated by a horizontal red line.Whiskers represent the standard deviation.***p<0.005;ns, not significant; unpaired two-sided Welch's t-test.(E) Co-localization of BacA Rs -CHY and LmdC N Rs -mNG in the ΔlmdC Rs background (SP119).The Pearson's Correlation Coefficient (PCC) of the two fluorescence signals in a random subpopulation of cells is 0.916.Scale bar: 5 µm.(F) Schematic showing the approach used to quantify the distribution of proteins along the outer and inner curve of a cell in panels G and J. (G) Demographs showing the enrichment of BacA-CYH and LmdC N -mNG at the inner curve the populations of strain SP119 analyzed in panel E (n=200 cells per strain).The white dashed line represents the point of transition from the outer to the inner curve.(H) Localization of BacA Rs -mNG in the ΔlmdC Rs background (SP98).Bar: 5 µm.(I) Localization of LmdC Rs -mNG in the ΔbacA Rs background (SP118).Bar: 5 µm.(J) Demographs showing the outer-versus-inner curve distribution of BacA Rs -mNG in the ΔlmdC Rs background (SP98; n=200 cells) and LmdC N Rs -mNG in the ΔbacA Rs background (SP118; n=200 cells).The white dashed line represents the point of transition from the outer to the inner curve.The online version of this article includes the following source data and figure supplement(s) for figure 10:Figure supplement 1. Analysis of the role of BacA and LmdC in R. rubrum morphogenesis.

Figure supplement 3 .
Figure supplement 3. Lysis of E. coli upon heterologous expression of full-length R. rubrum lmdC.

Figure supplement 4 .
Figure supplement 4. Analysis of R. rubrum strains producing fluorescent protein fusions.

Figure supplement 4
Figure supplement 4-source data 1.Raw images for the immunoblot analyses in Figure 10-figure supplement 4B and C.

Figure 11 .
Figure 11.The β-hairpin in the cytoplasmic region of LmdC is critical for BacA binding.(A) Predicted structure of LmdC N , comprising the first 80 amino acids of LmdC.The transmembrane domain is colored light blue.The cytoplasmic domain includes a conserved β-hairpin motif.(B) Clusters of amino acids mutated for the analysis of BacA-LmdC interaction.Mutated amino acids are shown as red sticks.(C) Localization of BacA Rs -CHY and the indicated LmdC N Rs -mNG variants in the ΔlmdC Rs background.The Pearson's Correlation Coefficient (PCC) of the two fluorescence signals in a random subpopulation of cells is 0.914 for the wild-type fusion protein (SP119), 0.887 for the R25A-H26A-L27A variant (SP237) and 0.827 for the R30A-S31A variant (SP238).Scale bar: 5 µm.(D,E) Demographs showing the distribution of BacA-mCherry and (D) R25A-H26A-L27A variant (SP237) or R31A-S31A variant (SP238) of LmdC N -mNG along the outer and inner curve of the cell (n=200 cells per strain).The white line marks the point of transition from the outer and the inner curve.The analysis was performed as described in Figure 10F.
Model of the roles of BacA and LmdC in H. neptunium and R. rubrum cell morphogenesis.(A) Cell morphogenesis in H. neptunium wild-type