Anchors: a way for FtsZ filaments to stay membrane bound

Abstract

Most bacteria use the tubulin homolog FtsZ to organize their cell division. FtsZ polymers initially assemble into mobile complexes that circle around a ring-like structure at the cell midpoint, followed by recruitment of other proteins that will constrict the cytoplasmic membrane and synthesize septal peptidoglycan to divide the cell. Despite the need for FtsZ polymers to associate with the membrane, FtsZ lacks intrinsic membrane binding ability. Consequently, FtsZ polymers have evolved to interact with the membrane through adaptor proteins that both bind FtsZ and the membrane. Here, we discuss recent progress in understanding the functions of these FtsZ membrane tethers. Some, such as FtsA and SepF, are widely conserved and assemble into varied oligomeric structures bound to the membrane through an amphipathic helix. Other less conserved proteins such as EzrA and ZipA have transmembrane domains, make extended structures, and seem to bind to FtsZ through two separate interactions. This review emphasizes that most FtsZs use multiple membrane tethers with overlapping functions, which not only attach FtsZ polymers to the membrane, but also organize them in specific higher-order structures that can optimize cell division activity. We discuss gaps in our knowledge of these concepts and how future studies can address them.

Graphical Abstract

Polymers of the bacterial tubulin homolog FtsZ are essential for division of most bacterial cells. Although these polymers need to associate with the cytoplasmic membrane for their function, they are unable to bind to the membrane themselves and instead have evolved to interact with a variety of tethering proteins that can bind to the cytoplasmic membrane. This Perspective summarizes our current understanding of FtsZ membrane tethers from diverse bacterial models.

We should not moor a ship with one anchor, or our life with one hope.

– Epictetus

1. Introduction

Bacterial cells divide and proliferate by using complexes of membrane-associated proteins that spatially and temporally organize synthesis of the cell wall, while constricting the cell membrane to split the mother cell into two daughter cells. Although the modes of cell division, and the proteins involved, are turning out to be as diverse as bacteria themselves, there are a few features shared by most species. For example, several proteins are conserved amongst most bacteria and comprise the core of the divisome, including the tubulin homolog FtsZ, the actin homolog FtsA, the FtsQBL complex, the FtsW glycosyltransferase, and the FtsI transpeptidase. FtsZ, the divisome’s master organizer, consists of mobile complexes of GTP-dependent dynamic protofilaments in a ring-like array called the Z ring that guide construction of the division septum and constriction of membrane(s) at the correct site, often at midcell, between segregated chromosomes. Although most divisome proteins are peripheral or integral membrane proteins, FtsZ itself has no intrinsic membrane binding ability and instead relies on a select group of FtsZ-interacting membrane-associated divisome proteins to tether it to the inner surface of the cytoplasmic membrane (CM). Acting like anchors, these tethers also help bundle dispersed FtsZ filaments into a condensed annular complex called the proto-ring (Fig. 1). Once the FtsZ filaments are moored snugly into the proto-ring, the maturing divisome is able to recruit other proteins to promote the septum formation and cytokinesis necessary for reproduction and continued life.

Fig. 1. Schematic overview of Z ring anchoring in several bacterial species.

Cytoplasmic monomers, dimers, and small multimers of FtsZ localize throughout a newborn, non-dividing cell (top row). Following their interaction with membrane anchors, FtsZ protofilaments further polymerize and condense at the nascent site of division (middle row), resulting in organization of the proto-ring (bottom row). In E. coli, FtsA and ZipA interact with each other (left/right arrow), and each contributes to tethering of FtsZ protofilaments to the cytoplasmic membrane (CM). The C-terminal MTH of FtsA inserts into the lipid bilayer and serves as the primary membrane anchor for FtsZ. The core polymerizing domain of FtsA interacts with the conserved FtsZ CCTP. The N-terminal TM domain of ZipA is essential for its function, while the bulk of the protein, the C-terminal cytoplasmic portion, binds directly to FtsZ. In addition to FtsA, B. subtilis possesses other accessory FtsZ membrane-anchors such as EzrA and SepF, each of which compete for interaction with the FtsZ CCTP. Unlike in E. coli or B. subtilis, FtsZ molecules localize at the cell pole containing the stalk (shown) and opposite of the flagellated cell pole (shown) in the asymmetrically dividing C. crescentus, prior to the formation of the Z ring at midcell. FzlC and FtsEX function as associate membrane anchors, while FtsA serves as the primary membrane tether. FzlC interacts with the FtsZ CCTP directly. FtsEX functions as a heterodimer with FtsE residing in the cytoplasm and FtsX traversing the CM through several TM domains.

Despite the high degree of conservation amongst FtsZ orthologs, its membrane anchoring proteins, and their mechanism of interaction with the Z ring, are highly diverse (Fig. 1). Therefore, FtsZ membrane anchors have evolved to adapt the relatively constant properties of FtsZ polymer dynamics to the requirements of individual species. Even within species, multiple proteins with distinct properties have evolved to tether FtsZ to the CM. Like the Greek philosopher Epictetus’ adage on sailing and human life, so too have bacteria evolved to not moor their hopes of reproduction and adaptability on one FtsZ anchor alone. In this perspective article, we focus mainly on the structural and functional properties of FtsZ membrane anchors from two divergent model bacterial species, and on the molecular basis of the interactions between these anchors and FtsZ.

2. Membrane tethers in the Gram-negative model Escherichia coli

In γ-proteobacteria such as E. coli, FtsA and ZipA tether FtsZ polymers to the CM, and are both essential for cell division (Table 1, Fig. 1). In the absence of FtsA or ZipA, Z rings still localize to midcell, indicating that either FtsA or ZipA alone is sufficient for tethering FtsZ polymers to the membrane. However, both proteins are required for cell division, as the absence of either results in the failure of the Z rings to assemble a functional divisome or a build a septum. When both FtsA and ZipA are inactivated, Z rings fail to form, demonstrating that FtsZ needs at least one of its tethers to form a Z ring structure at the membrane (Pichoff and Lutkenhaus, 2002). Although either is sufficient for FtsZ membrane anchoring, their localization to the division site requires a functional FtsZ. Consequently, the FtsZ-FtsA-ZipA triad can be considered a co-dependent protein complex, together forming the proto-ring, an essential precursor to the mature divisome (Ortiz et al., 2016).

Table 1.

Distribution and essentiality of FtsZ membrane anchors

Anchor	General Distribution	Essentiality (Typical)
FtsA	Gram-positive and Gram-negative bacteria	Essential
SepF	Gram-positive bacteria, Cyanobacteria, Euryarchaeota	Nonessential except in species lacking FtsA
ZipA	γ-Proteobacteria	Essential
EzrA	Bacillota (Firmicutes)	Nonessential
FzlC	α-Proteobacteria	Nonessential
FtsEX	Gram-negative and Gram-positive bacteria	Nonessential
GpsB (potential)	Bacillota (Firmicutes)	Nonessential
MapZ/LocZ	Streptococci, Lactococci, and Enterococci	Nonessential
SsgB	Streptomyces	Nonessential
CrgA (potential)	Actinomycetota	Nonessential

2.1. FtsA

FtsA is a bacterial homolog of actin and can assemble into actin-like filaments. It is conserved in many, but not all, Gram-negative and Gram-positive species. FtsA’s key ability to act as a membrane tether is dependent on an amphipathic helix at its extreme C-terminus that inserts into the lipid bilayer of the CM (Pichoff and Lutkenhaus, 2005) (Fig. 2). As with other E. coli proteins that have membrane-targeting helices (MTH), such as the spatial regulatory ATPase MinD (Szeto et al., 2002), FtsA likely evolved to bind reversibly to the membrane. This helix is conserved in FtsAs from a wide variety of species (Pichoff and Lutkenhaus, 2005), and is connected to the core polymerizing domain of FtsA through an intrinsically disordered linker (IDL) (Fig. 1, 2). E. coli FtsA has a short (~25 residue) IDL that can withstand some amino acid substitutions, but cannot be removed (Schoenemann et al., 2020). Most other FtsAs have similarly short IDLs, but several Gram-positive species including Streptococcus pneumoniae and Deinococcus radiodurans contain FtsA orthologs with considerably longer IDL domains (>100 additional linker residues in the case of D. radiodurans). These IDLs probably allow FtsA oligomers to assemble a short distance from the CM to increase their flexibility and dynamics. The longer FtsA IDLs in some species may have additional roles, including interactions with other proteins. Intriguingly, FtsZ proteins also have IDLs of variable length and sequence (Buske and Levin, 2013; Gardner et al., 2013; Shinn et al., 2022) (Fig. 1, 2) that function to connect the core N-terminal polymerizing domain of FtsZ to the conserved C-terminal peptide (CCTP, see below), and participate in interactions with other proteins (Barrows et al., 2020; Viola et al., 2022).

Fig. 2. Membrane topology and domain organization of FtsZ-membrane tethers.

Linear domains of E. coli FtsA and ZipA (Hale et al., 2000; Schoenemann et al., 2020), B. subtilis SepF (Duman et al., 2013) and B. subtilis EzrA (Haeusser et al., 2007; Cleverley et al., 2014) are depicted along with relevant residue numbers that demarcate the domains. Known crystal structures for ZipA, SepF and EzrA are shown to the right of each linear domain diagram along with their pdb accession numbers. The structure of Thermotoga maritima FtsA (1E4G), adapted from (van Den Ent and Löwe, 2000) with permission, is shown with its four subdomains highlighted in different colors, along with the interaction sites of ZipA and the FtsZ CCTP on the FtsA subunit (Szwedziak et al., 2012; Vega and Margolin, 2019). FZB denotes the FtsZ binding domain of each tether protein where it has been mapped; MTH denotes the membrane targeting helix; TM denotes a transmembrane domain; +/− and P/Q denotes the charged and IDL domains of ZipA. The EzrA N-terminal TM domain is followed by five spectrin repeat domains (R1–R5) and a C-terminal four-helix bundle (CH). The first of the five spectrin repeats within EzrA (R1) is an FtsZ binding domain, but the existence of at least one additional FtsZ binding site per EzrA monomer has been proposed. The linear domain organizations of FzlC and FtsEX were not depicted because of the lack of structural information or information about where these proteins bind to FtsZ.

The core polymerizing domain of FtsA contains an ATPase domain conserved in actin, Hsp70, and sugar kinases (Bork et al., 1992; van Den Ent and Löwe, 2000). Amino acid substitutions in the ATP binding pocket of FtsA often significantly impair FtsA’s function in cell division and comprise several well-studied thermosensitive alleles of ftsA including ftsA12 and ftsA27 (Herricks et al., 2014; Morrison et al., 2022). This core domain of FtsA is divided into four subdomains: 1A, 1C, 2A, and 2B (Fig. 2). Despite sharing structural homology with actin and another bacterial actin-homolog MreB in the 1A and 2A subdomains, FtsA harbors a unique 1C domain that replaces the conserved 1B subdomain in MreB and actin (Bork et al., 1992). An atomic structure of the FtsZ CCTP bound to FtsA showed that the 2B domain, which is distal from the ATP binding pocket and the FtsA-FtsA oligomeric subunit interface, forms a binding face for the FtsZ CCTP (Szwedziak et al., 2012). Some amino acid substitutions in the 2B domain of E. coli FtsA, such as R300E, perturb its interaction with FtsZ, preventing FtsA localization to Z rings and membrane tethering of FtsZ by FtsA (Pichoff and Lutkenhaus, 2007; Du et al., 2016).

FtsA’s MTH is essential for its function, as even single residue changes that reduce the amphipathic properties of the helix can prevent cell division. For example, changing W408, which is near the border of the helix (Fig. 2), to a glutamate (W408E) prevents FtsA from associating with the CM and blocks cell division, despite retaining the ability to interact with FtsZ through subdomain 2B (Pichoff and Lutkenhaus, 2005; Shiomi and Margolin, 2008). Interestingly, overproduction of FtsAW408E (or FtsA with the entire MTH deleted) triggers the assembly of large, stable filaments of FtsA in the cytoplasm that are hundreds of nanometers in length (Gayda et al., 1992). These cytoplasmic “bars” cause cells to grow into striking crescent-shapes, presumably by locally inhibiting cell wall and/or CM biosynthesis. These bars also have been used as an in vivo assay for the ability of FtsA mutant proteins to self-interact and polymerize: residue changes at the oligomer subunit interface of MTH-defective FtsA derivatives, which presumably cannot oligomerize efficiently, also fail to form the cytoplasmic bars (Pichoff et al., 2012; Herricks et al., 2014). However, these FtsA bars also can bind to FtsZ (Morrison et al., 2022), suggesting that their polymerization may be dependent on FtsZ. In any case, it is unlikely that wild-type FtsA ever forms such structures, as it is likely always associated closely with the CM. However, the MTH of FtsA serves only as a generic membrane anchor for FtsZ, independently of any specific sequence identity. Replacing FtsA’s MTH with another membrane attachment still allows FtsZ tethering in vitro and in vivo (Shiomi and Margolin, 2008; Radler et al., 2022).

The activity of FtsA in the divisome relies heavily on its oligomeric state on the CM. When purified E. coli FtsA is added to a lipid monolayer, it forms striking 12-membered mini-rings on the membrane (Krupka et al., 2017). In contrast, mutant FtsA proteins that hyperactivate the divisome (called FtsA*, see also next section) rarely form mini-rings, and instead assemble into short, curved oligomers or straight, double-stranded oligomers (Krupka et al., 2017; Schoenemann et al., 2018). As visualized by electron tomography, the MTH of FtsA in these mini-rings assemble into a second mini-ring inside the lumen of the main mini-ring, all connected through their IDLs, which can be visualized as spokes of a wheel (Krupka et al., 2017). Each of these FtsA structures can tether protofilaments of purified FtsZ to the membrane. However, the higher-order structure of tethered FtsZ is distinct amongst the different FtsA structures. Whereas FtsZ protofilaments are aligned, but kept apart, when tethered to mini-rings of wild-type FtsA, non-ring FtsA structures (including the curved oligomers and double stranded filaments) pack the FtsZ protofilaments close together in bundles, probably because of the higher packing density on the membrane of the FtsA tethers in this form (Krupka et al., 2017; Radler et al., 2022). The cytoplasmic domain of FtsN, an essential divisome protein that interacts with FtsA’s unique 1C domain (Rico et al., 2004; Busiek et al., 2012), can convert the mini-rings to the double stranded oligomers (Nierhaus et al., 2022). Although these different FtsA oligomeric states have yet to be visualized directly in E. coli cells, the data suggest that increasing the lateral interactions between FtsZ protofilaments stimulates cell division. This idea is supported by recent studies indicating that condensation of FtsZ filaments is required for activation of division septum synthesis in Gram-positive Bacillus subtilis (Squyres et al., 2021).

2.2. ZipA

Unlike ftsA, clear orthologs of zipA are mainly found only in the γ-proteobacteria (Raychaudhuri, 1999), although distantly related ZipA homologs with similar functional and structural characteristics, including interaction with FtsZ and partial complementation of E. coli ZipA function, have been described in the β-proteobacterial species Burkholderia cenocepacia (Trespidi et al., 2020) and Neisseria gonorrheae (Du and Arvidson, 2003) (Fig. 3). Originally isolated as an FtsZ binding protein, ZipA exhibits unusual topology for an E. coli integral membrane protein, with its N-terminal transmembrane (TM) domain in the CM and the remainder of the protein in the cytoplasm (Hale and de Boer, 1997). If ZipA’s TM is replaced with an analogous domain from another protein with similar membrane topology, the hybrid protein can localize to Z rings but no longer functions in cell division, indicating that the specific sequence of ZipA’s TM is essential for its function (Hale et al., 2000).

Fig. 3.

ZipA orthologs from γ-proteobacteria and β-proteobacteria. (A) The globular FZB domains of ZipA from the β-proteobacteria Neisseria meningitidis and Burkholderia cenocepacia were aligned with those from the γ-proteobacteria Pseudomonas aeruginosa and E. coli using Clustal Omega. Residue 1 of E. coli ZipA in the alignment corresponds to residue 192 of the native protein, near the N terminus of the FZB domain. Residues were coloured according to their chemical similarity. Orange rectangles denote E. coli ZipA residues that directly contact the E. coli FtsZ CCTP in the atomic structure (Mosyak et al., 2000). * = complete identity; : = strong identity; . = some identity. (B) Similarity of ZipA structures and contacts with the FtsZ CCTP. Shown are 180° rotated atomic structures of FtsZ CCTP bound to ZipA (1F47), along with structures of E. coli and N. meningitidis ZipA predicted by AlphaFold (Jumper et al., 2021), suggesting strong structural similarity despite divergent primary sequences.

The cytoplasmic portion of ZipA consists of a charged domain, rich in basic amino acid residues (residues 26–84), followed by a proline-glutamine-rich IDL (Ohashi et al., 2002) that connects the charged domain to a ~140 residue globular domain (residues 185–328) (Fig. 2). This globular FtsZ Binding (FZB) domain binds directly to FtsZ (Hale and de Boer, 1997; Raychaudhuri, 1999; Mosyak et al., 2000; Moy et al., 2000). Originally thought to bind only the FtsZ CCTP, recent in vivo crosslinking has shown that another face of the FZB domain binds to the core polymerizing domain of FtsZ (Cameron et al., 2021). Single amino acid changes in both the domain of ZipA that binds the FtsZ CCTP (e.g. F269S) and the ZipA domain that binds the FtsZ core (e.g. V299D) reduce ZipA activity in cell division. Remarkably, the F269S change abolishes the ability of ZipA to localize to Z rings and to function in cell division, indicating that optimal ZipA-FtsZ CCTP interaction is crucial for membrane tethering. Residue changes within the second binding face of the FZB domain such as V299D and M308K inhibit ZipA’s cell division activity, although these mutant ZipA proteins retain their ability to localize to the Z ring (Cameron et al., 2021). This suggests that FtsZ membrane tethering is initially mainly achieved by the ZipA-CCTP interaction, but the second interaction between ZipA and the FtsZ core domain may regulate a downstream activity of FtsZ, such as FtsZ protofilament bundling, recruitment of other divisome proteins, or FtsZ recycling. This last point is important, because FtsZ leaves the CM at the septum constriction site earlier than other divisome proteins (Soderstrom et al., 2016). Whether it is the increased membrane curvature itself, or a specific change in oligomeric state, FtsZ release from the membrane and its membrane tethers is likely a regulated process.

2.3. Interactions amongst FtsA, ZipA and FtsZ

As discussed above, ZipA and FtsA tether FtsZ filaments to the CM mainly through interactions with the FtsZ CCTP located (Wang et al., 1997; Din et al., 1998; Liu et al., 1999; Ma and Margolin, 1999). As a result, ZipA and FtsA may compete for binding, although there is sufficient cellular FtsZ (~5000 molecules/cell) to permit FtsA (~1000 molecules/cell) and ZipA (~2000 molecules/cell) to each bind independently. The competition for CCTP sites can be observed when either FtsA or ZipA are artificially overproduced, which results in cell division inhibition that can be reversed by concomitant overproduction of the FtsZ target. Along the same lines, excess FtsA inhibits cell division more severely in a strain with a compromised ZipA, such as the thermosensitive zipA1 allele (at the permissive temperature) compared with wild-type cells. Excess ZipA similarly inhibits cell division more in ftsA thermosensitive strains at the permissive temperature compared with wild-type strains (Herricks et al., 2014). Despite their ability to bind to the same small conserved sequence in the FtsZ CCTP, the CCTP-binding regions of ZipA and FtsA have no obvious sequence or structural homology (van Den Ent and Löwe, 2000; Mosyak et al., 2000).

In addition to separately binding the same site on FtsZ under normal conditions, ZipA and FtsA also interact with each other through α-helix 7 in subdomain 2B of FtsA (Vega and Margolin, 2019) (Fig. 2). This interaction was shown by in vivo crosslinking using a short-range crosslinker, suggesting that the interaction is not merely because FtsA and ZipA are nearby, interacting with adjacent CCTPs within an FtsZ, but interact directly. As ZipA may help in converting FtsA from an inactive to an active conformation (see below), it is not surprising that they bind to each other, although it is likely that both need to be bound to the Z ring (and thus be in close proximity) for this binding to occur efficiently.

The role of ZipA as a membrane tether for FtsZ is clearly secondary to that of FtsA. An initial hint for this was that the genes for ftsA and ftsZ map to the main division-cell wall (dcw) cluster that is conserved amongst many bacteria, and that includes genes for the two septum synthesis enzymes FtsI and FtsW, in addition to their key protein interactors: FtsQ and FtsL. In contrast, the zipA gene is located far from the dcw locus in the E. coli chromosome. More importantly, gain-of-function mutations in ftsA, collectively called ftsA*, can bypass the normal requirement for ZipA in E. coli cell division (Geissler et al., 2003; Geissler and Margolin, 2005; Pichoff et al., 2012). Although ZipA in wild-type cells is required for further assembly of the divisome after the proto-ring forms (Hale and de Boer, 1999), this requirement is likely due to ZipA’s ability to change FtsA’s oligomeric state into an FtsA*-like active form. This conversion of FtsA conceivably requires the direct interaction between ZipA and helix 7 of FtsA as mentioned above. Interestingly, a gain-of-function mutant of FtsZ (FtsZ*, L169R) also bypasses the need for ZipA in divisome assembly (Haeusser et al., 2015). FtsZ* forms double-stranded filaments in vitro and is generally in a more bundled state than wild-type FtsZ in vivo. This bundling probably mimics the action of FtsA* on wild-type FtsZ, and the hypermorphic properties of a hyperbundled FtsZ is entirely consistent with the need for a more condensed Z ring to drive septum formation, as discussed earlier.

2.4. Why two membrane anchors?

Although the E. coli proto-ring has been known for over 20 years, it is still a mystery why it evolved to require two membrane tethers for FtsZ. Yet other bacteria, as we discuss below, also seem to have multiple FtsZ membrane tethers, so E. coli is not unique in this regard. Although ZipA can be easily bypassed by alleles that circumvent the checkpoint, including hyperfission “star” alleles in other divisome proteins (Geissler et al., 2003; Tsang and Bernhardt, 2015), it does have some unique nonessential functions not provided by FtsA. First, ZipA seems to protect FtsZ from Clp proteases, which FtsA or even FtsA* cannot do (Pazos et al., 2013). Second, ZipA interacts with, and regulates, the activity of PBP1A and PBP1B, nonessential proteins that participate in the initial stages of E. coli septum synthesis prior to the action of FtsW and FtsI (Pazos et al., 2018). Because it has a much longer flexible IDL than FtsA (Lee et al., 2019), ZipA is also potentially more useful in fine tuning the dynamics of treadmilling FtsZ protofilaments than FtsA, although this function is dispensable in lab-grown E. coli. Although conversion of FtsA to the active form seems to promote FtsZ protofilament bundling, ZipA does not seem to stimulate FtsZ bundling. The ability of ZipA to support active FtsZ protofilament treadmilling is still being debated (Krupka et al., 2018; García-Soriano et al., 2020).

3. Anchoring the Z ring in other Gram-negative bacteria

In the α-proteobacterium Caulobacter crescentus, which divides asymmetrically mainly by membrane constriction instead of a septal wall, ftsA is essential (Sackett et al., 1998). Additionally, FzlC and FtsEX proteins also function as membrane anchors (Meier et al., 2016; Meier et al., 2017) (Fig. 1, Table 1). FzlC is conserved throughout most α-proteobacteria and binds directly to the CCTP of FtsZ, although it is not yet clear how it binds to the CM. FtsE and FtsX are ABC transporter homologs that function as a unit and are conserved throughout a diverse group of Gram-negative and Gram-positive species. In E. coli, FtsE interacts with FtsZ, while FtsX interacts with FtsA (Corbin et al., 2007; Du et al., 2019). FtsE and FtsX are nonessential, but deletion of ftsE in C. crescentus results in a severe cell chaining phenotype indicative of a late cell division defect (Meier et al., 2017).

In striking contrast to E. coli, C. crescentus FtsA is expressed late in the cell division time course relative to FtsZ (Sackett et al., 1998). This strongly suggests that FtsA does not anchor FtsZ to the membrane at the initial stages of Z ring formation, or perhaps at any stage. Notably, FtsZs from the α-proteobacteria, including C. crescentus and Sinorhizobium meliloti, have much longer IDLs than those of other species (Margolin et al., 1991; Sundararajan et al., 2015), and these may be important for flexible membrane tethering of FtsZ protofilaments at the early stages of division in the absence of FtsA. Alternatively, these longer IDLs may play a role in the special constriction mode of cytokinesis in these species, such as providing more inward force for the CM to invaginate (Poindexter and Hagenzieker, 1981). It is tempting to speculate that FtsA is specifically recruited later in division for this purpose.

In another well-characterized α-proteobacterium, Agrobacterium tumefaciens, FtsZ forms a Z ring prior to the arrival of FtsA. This is reminiscent of the C. crescentus situation, suggesting that another membrane anchor is used for proto-ring assembly (Zupan et al., 2013). An A. tumefaciens homolog of FzlC (Howell et al., 2019) likely has a similarly crucial role in early Z ring assembly. Notably, A. tumefaciens and close relatives such as S. meliloti harbor additional FtsZ homologs that lack the CCTP, suggesting that these FtsZs may not be membrane-associated, but instead might regulate assembly of the primary FtsZ (Margolin and Long, 1994; Howell et al., 2019).

4. Membrane tethers in the Gram-positive model Bacillus subtilis

Although missing in some Gram-positive families, FtsA is conserved in the Gram positive Bacillota family (formerly Firmicutes). Like E. coli, Bacillota species primarily rely on FtsA as a membrane anchor for FtsZ and for cytokinesis, but often contain additional proteins that have overlapping functions as well as additional roles. Two such proteins are EzrA and SepF (YlmF), each discovered and first characterized in B. subtilis (Levin et al., 1999; Ishikawa et al., 2006; Gamba et al., 2009) (Fig. 1, 2). In Actinobacterial species including Mycobacterium tuberculosis (Mtb), Mycobacterium smegmatis, and Corynebacterium glutamicum, where both EzrA and FtsA are missing, SepF assumes the essential role of membrane anchor for FtsZ and divisome coordination (Gola et al., 2015; Sogues et al., 2020; Dey and Zhou, 2022).

4.1. EzrA

Though Bacillota such as B. subtilis lack a ZipA homolog, they have a conserved unrelated protein, EzrA, that has the same unusual orientation in the CM as ZipA, with a short N-terminal transmembrane (TM) domain linked to a longer cytoplasmic C-terminal domain (Fig. 2). Originally discovered as an inhibitor of aberrant FtsZ polymer assembly whose loss permits formation of extra Z rings at polar and quarter positions in B. subtilis (Levin et al., 1999), EzrA paradoxically also localizes to the early divisome with FtsA, SepF, and ZapA through direct interactions with FtsZ (Ishikawa et al., 2006; Gamba et al., 2009). This suggests that EzrA could play a role in tethering FtsZ to the membrane through EzrA’s TM domain, while aiding in the stability of divisome scaffolding and/or promoting FtsZ dynamics through EzrA’s inhibition of FtsZ assembly. The ability of EzrA to serve as a membrane anchor for FtsZ is most apparent in cells lacking both SepF and FtsA. Though unable to divide and non-viable, these cells still have FtsZ localized to the membrane at midcell, predominantly in uncondensed spirals (Ishikawa et al., 2006).

Consistent with such membrane-bound uncondensed FtsZ spirals, both in vivo and in vitro analysis of EzrA has supported its proposed inhibition of FtsZ filament condensation (bundling) (Levin et al., 2001; Haeusser et al., 2004). However, in the absence of EzrA, photobleaching recovery studies showed only modest effects on Z ring dynamics at midcell (Anderson et al., 2004) and single molecule imaging revealed no change in FtsZ treadmilling (Squyres et al., 2021). Although EzrA and the other early divisome proteins that affect ring condensation seem to have no effect on FtsZ filament treadmilling, EzrA uniquely decreases FtsZ subunit lifetime within filaments to shorten them (Squyres et al., 2021), suggesting that it plays an important balancing role along with FtsA, SepF, and ZapA during initial FtsZ filament condensation in the proto-ring. Additional supporting evidence for EzrA’s role in promoting midcell proto-ring formation and stabilizing divisome function is suggested by synthetic lethality of an ezrA knockout with loss of other positively-acting divisome proteins (ZapA, SepF, or GpsB) (Gueiros-Filho and Losick, 2002; Hamoen et al., 2006; Claessen et al., 2008) and EzrA’s role in proper recruitment of the bi-functional glycosyltranserase/transpeptidase PBP1to the septum (Claessen et al., 2008).

The crystal structure of EzrA’s cytoplasmic domain reveals a linear extension of five antiparallel triple-helical bundles that bend into a semicircle with a diameter of 120 Å (Cleverley et al., 2014), reminiscent of eukaryotic actin-associated spectrins (Fig. 2). Results of multiple studies now argue that EzrA interacts with FtsZ through at least two separate domains, one in the N-terminus of EzrA, and one in its C-terminus (Haeusser et al., 2007; Cleverley et al., 2014; Land et al., 2014). This two-pronged interaction with FtsZ is reminiscent of its topological cousin, ZipA. Notably, interactions of these two distinct regions of EzrA with FtsZ seem to correspond to the dual activities of EzrA in both inhibiting FtsZ assembly and in promoting/scaffolding Z ring condensation. While finer mapping remains to be done, the N-terminal cytoplasmic portion of EzrA seems responsible for inhibitory interactions with FtsZ (Cleverley et al., 2014; Land et al., 2014). Conversely, the C-terminal region of EzrA contains a conserved “QNR patch” essential for EzrA localization to the divisome (Haeusser et al., 2007). This finding also suggests that, as with ZipA, EzrA’s TM domain does not impart any specificity to EzrA’s localization. Supporting this, substitution of residues within EzrA’s TM domain with other TM sequences, including that from ZipA, does not alter EzrA recruitment to FtsZ at midcell or function in B. subtilis cell division (Land et al., 2014). Thus in contrast to ZipA, the specific sequence of EzrA’s TM is not required for function.

Models using the EzrA crystal structure suggest that the arc of EzrA’s semicircle shape places the “QNR patch” close to the membrane (Cleverley et al., 2014). This indicates that EzrA may localize to the Z ring through interactions with membrane components or other membrane-associated factors, rather than a direct interaction with FtsZ. Its semi-circular arrangement would also give a EzrA a seat-belt-type shape that could constrain single FtsZ filaments from associating laterally during initial proto-ring condensation, and potentially promote filament condensation as cytokinesis proceeds. Interestingly, such a model is similar to models for SepF and FtsA activity in the divisome (see above and below), except that FtsZ polymer filaments reside on the outside of SepF or FtsA arcs in these models in contrast to FtsZ polymers being confined within EzrA arcs. Yet, intriguingly, such a simple difference in where the FtsZ polymers fit in with the arcs would be consistent with the opposing effects that EzrA and SepF/FtsA appear to have on FtsZ bundling.

On the other hand, truncations of EzrA that would presumably destroy the full arc (strap) geometry, but that retain the inhibitory N-terminal regions, still function in preventing aberrant FtsZ assembly and related phenotypes (Land et al., 2014). Further work is necessary to characterize the details of which specific residues of EzrA interact with FtsZ. Likewise, while one study has shown that EzrA only interacts with FtsZ’s promiscuous CTTP (Singh et al., 2008a), additional investigation appears warranted given the recent discovery of a second interaction between ZipA and the FtsZ core domain (Cameron et al., 2021).

4.2. SepF

SepF proteins are the most widely conserved FtsZ membrane anchors in Gram-positive bacteria, and remain conserved in the Euryarchaea that use FtsZ for cytokinesis (Pende et al., 2021). Additionally, although they are Gram negative, the Cyanobacteria cluster from a common ancestor with Gram-positive species and their Archaeal offshoot, and also encode sepF homologs (Miyagishima et al., 2005). Although SepF has no structural or sequence homology with FtsA and is not an actin homolog, SepF is analogous to FtsA in that it binds to the FtsZ CCTP (Cendrowicz et al., 2012) and has a terminal amphipathic helix that tethers it to the membrane (Duman et al., 2013) (Fig. 2). As with FtsA and ZipA of E. coli, FtsA, SepF, and EzrA of B. subtilis all compete for interaction with the FtsZ CCTP (Gao et al., 2017) (Fig. 1). In contrast to FtsA, SepF’s membrane-targeting helix is at the N-terminus of the protein, separated from the FtsZ binding domain at the C terminus by an IDL. The conservation of sepF closely downstream of ftsA and ftsZ within the cell division gene operon of many Gram-positive bacteria (White and Eswara, 2021) also underscores SepF’s importance to cytokinesis. Although depleting B. subtilis FtsA causes a severe cell division defect, overexpression of SepF can rescue this defect (Ishikawa et al., 2006), indicating the overlapping analogous functions of FtsA and SepF. Moreover, a B. subtilis strain depleted for FtsA and lacking EzrA can still divide, albeit poorly, consistent with the idea that SepF can act as a redundant FtsZ membrane anchor (Duman et al., 2013).

There are additional parallels between SepF and FtsA. B. subtilis SepF self-assembles into polymeric rings ~50 nm in diameter that are larger than the ~20 nm wide E. coli FtsA mini-rings (Krupka et al., 2017) and the SepF rings can stack to make tubes (Gundogdu et al., 2011). The C-terminal ~60% of SepF is sufficient for polymerization and interaction with FtsZ (Duman et al., 2013). Both SepF and FtsA can deform lipid vesicles to which they are bound (Duman et al., 2013; Conti et al., 2018). Finally, the MTHs of both SepF and FtsA can be replaced with a heterologous MTH from the MinD protein and retain full function (Pichoff and Lutkenhaus, 2005; Duman et al., 2013).

SepF, like FtsA, seems to have an important role both in stabilization of FtsZ during proto-ring condensation, perhaps through FtsZ filament bundling activity (Singh et al., 2008b) and during the latter stages of division during membrane constriction and septal synthesis (Hamoen et al., 2006). A recent study using chimeras constructed from a variety of SepF homologs that form rings ranging from ~19–44 nm in diameter suggested that curvature of membrane-bound SepF at the leading edge of septal wall formation controls septal thickness through divisome confinement (Wenzel et al., 2021). This is reminiscent of a model for E. coli FtsA, based on planar lipid layers in vitro, where FtsZ protofilaments bind on top of FtsA mini-rings (Krupka et al., 2017). In these conditions FtsA mini-rings are forced to lie flat on the planar lipid monolayer. But, in a dividing cell, FtsA polymers may act like those in the proposed SepF model, and follow – or constrain – the positive curvature of the invaginating septum. This could help to guide FtsZ polymers as they arrange themselves in a circumferential direction during condensation into the proto-ring and/or during the latter stages of division (Nierhaus et al., 2022).

5. Anchoring the Z ring in other Gram-positive bacteria

Other model Bacillota species, including Staphylococcus aureus and S. pneumoniae, also use FtsA, SepF, and EzrA as membrane anchors for FtsZ. The essentiality of ftsA in both S. aureus (Santiago et al., 2015) and S. pneumoniae (Lara et al., 2005) highlights its importance as the primary anchor for FtsZ in these species. Additionally, complete depletion of FtsA in S. pneumoniae leads to Z ring loss (Mura et al., 2016), suggesting SepF and EzrA cannot normally suffice as anchors on their own. Notably, FtsA in S. pneumoniae remains localized with FtsZ throughout constriction, where it also seems to play a role in regulation of the slight elongation the ovoid species undergoes through septal cell wall growth associated with the divisome (Mura et al., 2016).

While SepF is not essential in either species (Fadda et al., 2003; Santiago et al., 2015), a S. pneumoniae sepF null allele phenocopies partial depletion of FtsA, with elongated and enlarged cells that still form Z rings that fail to progress. Overexpression of FtsA suppresses these defects in septum formation (Mura et al., 2016). This suggests the primary role of SepF in S. pneumoniae is redundancy with FtsA in the latter stages of cytokinesis. However, any other investigations into potential redundancies between FtsA and SepF (and EzrA) remain to be explored in S. pneumoniae, and no research to date has explored SepF in S. aureus in any detail beyond its association in complex with FtsZ, FtsA, and EzrA (Bottomley et al., 2017).

Though initially reported to be essential in S. aureus (Steele et al., 2011), as it is in S. pneumoniae (Thanassi et al., 2002; van Opijnen et al., 2009; Liu et al., 2017), later reports characterized ezrA as non-essential in S. aureus (Jorge et al., 2011; Santiago et al., 2015; Stamsås et al., 2018). However, loss of EzrA in either species results in a similar cell enlargement phenotype with significantly reduced Z ring formation, where the rare rings that do form are misplaced (Steele et al., 2011; Liu et al., 2017; Perez et al., 2021). Though consistent with a role for EzrA in helping anchor FtsZ, this phenotype, along with the extra Z rings in S. pneumoniae cells overproducing EzrA (Perez et al., 2021), stands in stark contrast to the extra Z rings originally observed in B. subtilis cells lacking EzrA (Levin et al., 1999). This opposite effect likely arises from the absence of potential polar division sites in S. pneumoniae where FtsZ assembly would need to be inhibited, compared to the rod-shaped B. subtilis that also sporulates through alternatively asymmetric, polar septation. Beyond anchoring FtsZ at the membrane and promoting Z ring formation, EzrA additionally serves to recruit additional divisome proteins as in B. subtilis.

As in B. subtilis, EzrA interacts with GpsB in both S. aureus and S. pneumoniae to regulate septal cell wall synthesis (Fleurie et al., 2014b; Eswara et al., 2018; Cleverley et al., 2019). However, unlike in B. subtilis, S. aureus GpsB directly interacts with FtsZ early during proto-ring formation, and alters FtsZ assembly dynamics (Eswara et al., 2018). Intriguingly, putative membrane-targeting sequences have been identified in some GpsB family members, with the GpsB of Listeria monocytogenes demonstrated to directly bind membranes (Rismondo et al., 2016). To our knowledge, it is unknown if GpsB of L. monocytogenes interacts with FtsZ or if GpsB of S. aureus interacts with the membrane, but GpsB has the potential to be another membrane anchor in these and other species (Table 1).

Another membrane anchor for FtsZ called MapZ (also called LocZ), is conserved in the streptococci, lactococci, and enterococci (Holeckova et al., 2014; Fleurie et al., 2014a) (Table 1). MapZ contains a single TM domain with a cytoplasmic N-terminal topology and extracellular C-terminal topology required for its localization to sites of nascent division, prior to localization of FtsZ and FtsA. Regulated by phosphorylation and peptidoglycan interactions at the division site, MapZ rings recruit FtsZ, FtsA, SepF, and EzrA to guide septum placement. Duplicate MapZ rings subsequently migrate away from the Z ring of a newborn cell, following future division sites as they grow outward during ovoid cell elongation. Although nonessential, MapZ loss results in asymmetric cytokinesis from mislocalized Z rings, suggesting that while it may serve as an anchor, it does not have the predominance of FtsA in that role. The expendability of MapZ for S. pneumoniae cell division also may be due to its redundancy with EzrA, which can compensate for septal localization in the absence of MapZ (Perez et al., 2019).

As discussed above, the Actinomycetota generally lack FtsA and EzrA, but use SepF as the primary membrane anchor for FtsZ. The role of Actinomycetota SepF homologs in cytokinesis has been studied in C. glutamicum (Sogues et al., 2020), the Mycobacteria (Gola et al., 2015), and the Streptomyces (Zhang et al., 2023). Notably, Streptomyces coelicolor and Streptomyces venezuelae contain two SepF paralogs (SflA and SflB, also known as SepF2 and SepF3, respectively) in addition to the canonical SepF encoded by the dcw gene cluster (Schlimpert et al., 2017; Zhang et al., 2023). This is consistent with redundancy in FtsZ anchoring in the absence of an FtsA homolog, though SepF2 of S. venezuelae lacks conserved N-terminal residues important for the amphipathic MTH structure of B. subtilis SepF, suggesting this paralog may not bind membranes (Schlimpert et al., 2017). The loss of SflA and SflB in S. coelicolor leads to increased hyphal branching due to mislocalization of FtsZ and DivIVA (Zhang et al., 2023), consistent with their putative role later in sporulation division. Despite some study of these paralogs, no detailed investigation into the canonical SepF of Streptomyces species has yet been undertaken.

In addition to SepF, the morphologically complex Actinomycetes such as Streptomyces also contain a unique membrane anchor, SsgB (Xu et al., 2009) (Table 1), that localizes with its targeting partner SsgA at future division sites to directly recruit FtsZ and promote its assembly into a proto-ring (Willemse et al., 2011). Recruitment and membrane anchoring of FtsZ by SsgB is further stabilized by interaction in complex with other membrane components including SepG (Zhang et al., 2016), prokaryotic dynamin homologs DynA and DynB (Schlimpert et al., 2017), as well as SepF and its paralogs (Cantlay et al., 2021).

CrgA is another membrane-associated protein conserved in the Actinomycetota that interacts directly with FtsZ, and therefore could be considered a membrane anchor (Table 1). However, while the loss of CrgA in M. smegmatis leads to cytokinesis defects, they manifest in latter stages after Z ring formation (Plocinski et al., 2011). Additionally, CrgA in S. coelicolor has been proposed to be a part of the large complex with SsgB-anchored FtsZ described above, and its overexpression leads to reductions in Z ring formation (Del Sol et al., 2006).

6. Questions and challenges

Despite impressive progress in understanding how FtsZ polymers are tethered to the CM during bacterial cell division, many questions and challenges remain. Why, for example, can another cytoskeletal filament-forming protein in bacteria, MreB, bind to the CM directly as polymers using its own MTH, while FtsZ polymers require adaptor proteins for their attachment to the CM? The most likely answer is that FtsZ polymers move by treadmilling, whereas MreB polymers move but do not treadmill (van den Ent et al., 2014). Treadmilling requires that subunits dissociate from one end of a polarized polymer and re-associate at the other end, which necessitates frequent release from the membrane. However, MinD protein has its own MTH and is capable of frequent release from the membrane in response to MinE, followed by MinD rebinding the membrane (Hu and Lutkenhaus, 2001; Huang et al., 2003). Moreover, FtsZ fused to an MTH allowed full GTP-dependent filament dynamics in vitro, without any adapter protein (Ramirez-Diaz et al., 2018). Indeed, early pioneering studies in the reconstitution of proto-rings in liposomes fused E. coli FtsZ directly to a MTH, with the goal of attaching FtsZ to the membrane without the complications of adapter proteins (Osawa et al., 2008). These FtsZ-MTH fusions formed rings inside liposomes and even constricted them, but, crucially, full liposome fission required including FtsA* as a membrane tethering adapter for untagged FtsZ (Osawa and Erickson, 2013). This indicated that having a separate membrane tether was important for FtsZ polymers to fully function in Z ring organization and membrane constriction. Advances in synthesis of divisome proteins in vitro have recently allowed FtsZ and FtsA (or FtsA*) to be added to liposome lumens, where they form contractile Z rings tethered by their natural FtsA anchors (Godino et al., 2020; Kohyama et al., 2022). These studies support the idea that a membrane tether optimizes FtsZ polymer activity and allows for greater flexibility in interaction with other proteins, including those such as the spatial regulator MinC that also bind FtsZ’s CCTP (Ortiz et al., 2016).

A related question posed earlier is why multiple membrane tethers seem to be needed in most species. The answer is probably similar to why FtsZ needs an adaptor protein in the first place—to provide maximum flexibility and responsiveness. Using a tether protein, and optimally more than one, allows FtsZ polymers to bind to the membrane with different affinities, different geometrical constraints, and different local concentrations, depending on the state of the adaptor. This is manifested by the distinct higher order structures FtsZ displays in response to different oligomeric states of FtsA on the membrane, as well as the dual binding interactions between FtsZ and ZipA that probably have different affinities. For example, ΔzipA cells expressing FtsA* divide quite efficiently, but not as efficiently as wild-type cells (Geissler et al., 2003), and significantly less efficiently than zipA⁺ cells expressing FtsA* (or other hyperfission alleles such as ftsL*), which are ~20% shorter than wild-type cells (Geissler et al., 2007; Liu et al., 2015; Tsang and Bernhardt, 2015). These results suggest that a functional ZipA is still needed to optimize cell division.

Other proteins not defined strictly as FtsZ membrane tethers, including the E. coli FtsE and Mtb FtsW proteins, may help to modulate FtsZ-CM interactions during the course of cell division. As FtsE interacts directly with both FtsZ and the integral membrane protein FtsX, it is possible that FtsE acts as a molecular bridge between FtsZ and the membrane after the proto-ring stage (Du et al., 2019). Similarly, FtsZ has been reported to interact with FtsW directly in Mtb (Datta et al., 2002), implicating that interaction in membrane anchoring for FtsZ. In actinobacteria such as Mtb, where FtsA is lacking and SepF is essential, an FtsZ-FtsW interaction may be important for FtsW activation as well as an alternative to SepF for FtsZ membrane anchoring, under conditions yet to be defined. Nonetheless, this also may introduce a complication, as FtsZ and FtsW move at different speeds, at least in E. coli (Yang et al., 2021). Perhaps this speed differential between the treadmilling FtsZ and the slower moving septum synthesis machine is not an issue in slow-growing Mtb. Furthermore, we still do not know how intracellular levels of FtsZ membrane anchors and other associated proteins are regulated throughout the bacterial life cycle under different conditions such as nutrient limitation, exposure to environmental stress, and during infection of host cells. Clearly we have just scratched the surface of our understanding of the diversity of bacterial cell division mechanisms and the interplay between FtsZ membrane anchors and their divisomes.

DATA AVAILABILITY STATEMENT

This article does not contain new data.