Spatial control of cell envelope biosynthesis in mycobacteria

Abstract

The mycobacterial cell envelope is a complex multilayered structure that provides the strength to the rod-shaped cell and creates the permeability barrier against antibiotics and host immune attack. In this review, we will discuss the spatial coordination of cell envelope biosynthesis and how plasma membrane compartmentalization plays a role in this process. The spatial organization of cell envelope biosynthetic enzymes as well as other membrane-associated proteins is crucial for cellular processes such as polar growth and midcell septum formation. We will highlight metabolic enzymes involved in the localized biosynthesis of envelope components such as peptidoglycan, arabinogalactan and outer/inner membrane lipids. The known and potential roles of cytoskeletal and coiled coil proteins in driving subcellular protein localization will also be summarized. Finally, we provide a comprehensive overview of known lateral heterogeneities in mycobacterial plasma membrane, with a particular focus on the intracellular membrane domain, recently revealed by biochemical fractionation and fluorescence microscopy. We consider how this dynamic and multifunctional membrane microdomain contributes to the subcellular localization of membrane proteins and spatially restricted cell envelope biosynthesis in mycobacteria.

cell envelope, membrane domain, metabolic compartmentalization, Mycobacterium, polar growth, septum

A BRIEF OVERVIEW OF CELL ENVELOPE STRUCTURE

The cell envelope is critical for mycobacteria to maintain cell shape, provide a permeability barrier and modulate the host immune response. The biosynthetic pathways and various aspects of its functional significance are discussed in other chapters in this issue (by Jordi Torrelles and Jessica Seeliger) and other recent reviews (Jackson 2014; Pavelka, Mahapatra and Crick 2014; Minnikin et al.2015; Jankute et al.2015; Källenius et al.2016; Alderwick et al.2015).

The mycobacterial cell envelope is composed of five main layers: an outer layer or capsule, the mycolic acid (MA)-rich outer membrane (OM) also known as mycomembrane, the arabinogalactan (AG) layer, followed by the core peptidoglycan (PG) sacculus and the plasma membrane (PM) (Fig. 1) (Minnikin 1982). The OM is rich in glycans and lipids, including trehalose dimycolate, phthiocerol dimycocerosate (PDIM), phenolic glycolipid (PGL), glycopeptidolipid (GPL), triacylglycerol, diacylglycerol, lipomannan and lipoarabinomannan (Bansal-Mutalik and Nikaido 2014; Jackson 2014; Jankute et al.2015). Some of these lipids are universally found in all Mycobacterium species, while some others such as GPLs, PGLs and PDIMs are only present in select species, conferring species-specific features (see reviews Daffé, Crick and Jackson 2014; Jackson 2014). The inner leaflet of the OM is proposed to be composed primarily of MA that can be up to C90 in length (Marrakchi, Lanéelle and Daffé 2014). These MAs are covalently linked to the AG layer beneath, which is composed primarily of 5- and 6- linked β-d-galactofuranose (Galf) residues and arabinose chains (Angala et al.2014; Jankute et al.2015). The AG is connected to the N-acetylmuramic acid residues of PG through the alpha-l-rhamnose-1,3-d-GlcNAc-1-phosphate linker (McNeil, Daffe and Brennan 1990). The entire crosslinked layers of MA, AG and PG are often referred to as the mycolyl-arabinogalactan-peptidoglycan (mAGP) complex.

Figure 1.

A model of mycobacterial cell envelope. The model shows each layer as follows: the plasma membrane (PM), peptidoglycan layer (PG), arabinogalactan (AG) layer and the outer membrane (OM). Major lipids present in the PM and the OM are as indicated in the model, but do not represent the true relative abundance. Some structural details of the lipids are omitted, and the thickness of each layer is not to scale. Some lipids such as GPLs and PDIMs are species-specific (see text for detail). GPL, glycopeptidolipid; LM, lipomannan; LAM, lipoarabinomannan; TMM, trehalose monomycolate; TDM, trehalose dimycolate; PI, phosphatidylinositol; CL, cardiolipin; PE, phosphatidylethanolamine; TAG, triacylglycerol; PDIM, phthiocerol dimycocerosate.

Underneath the PG layer is the periplasmic space followed by the PM, composed of major phospholipids such as phosphatidylethanolamine (PE), cardiolipin (CL), phosphatidylinositol (PI) and phosphatidylinositol mannosides (PIMs). Biosynthetic pathways of many cell envelope lipids and glycans start in the cytoplasm or the cytoplasmic side of the PM, followed by reactions taking place in the periplasmic space. For detailed metabolic reactions taking place in the PM, readers are directed to our previous review (Crellin, Luo and Morita 2013). In the current review, we will discuss the spatial coordination of the cell envelope biosynthesis and other biological processes in mycobacteria and how PM compartmentalization plays a role in these processes.

SPATIAL POSITIONING OF CELL ENVELOPE BIOSYNTHETIC MACHINERIES AND OTHER MEMBRANE-BOUND PROTEINS

Unlike other model eubacteria, mycobacteria and other actinobacteria elongate the cell envelope by zonal growth at the cell poles. To support this localized growth, cell envelope biosynthesis must take place in a spatially coordinated manner at the poles as well as the septum. The two poles are not equal. The polar growth is asymmetric in that the growth from the old pole is faster than that from the new pole (Aldridge et al.2012). Recent work suggests that the asymmetric growth is in fact a regulated process, as the divisome protein LamA negatively regulates elongation at newly formed poles (Fig. 2) (Rego, Audette and Rubin 2017). Cells that lack this protein grow more evenly from both new and old poles, and are interestingly more susceptible to killing by certain antibiotics (Rego, Audette and Rubin 2017). In this section, we will discuss how the cell envelope biosynthesis is coordinated spatially and will further highlight other cellular processes that are spatially segregated.

Figure 2.

Overview of protein localization during cell growth. The divisome is assembled at the midcell driven by the formation of an FtsZ ring interacting with various proteins. DivIVA orchestrate the polar growth, interacting with the Acc complex. MabA and InhA localize to the pole as well for MA biosynthesis. ParA also interacts with DivIVA to ensure chromosomal DNA segregation. While there are more than 300 IMD-associated proteins identified by proteomic analysis, only some experimentally verified proteins are indicated. Exact morphological configuration of the IMD is unknown, and depicted as a membrane zone proximal to the polar DivIVA complex. See Fig. 3 for more IMD-associated proteins. IMD, intracellular membrane domain. Z, FtsZ; W, FtsW; BLQ, FtsB/FtsL/FtsQ.

PG synthesis occurs at the growth poles

Expansion of PG determines the rate of bacterial growth (Rojas, Huang and Theriot 2017). Unlike model organisms such as Escherichia coli and Bacillus subtilis, which elongate by inserting new PG along the sidewall (de Pedro et al.1997; Daniel and Errington 2003), mycobacteria grow from the cell poles (Hett and Rubin 2008; Aldridge et al.2012). In support, fluorescent protein fusions to a variety of enzymes involved in PG synthesis are preferentially enriched at the poles, including MurG (Fig. 2), the synthase for the membrane-linked precursor lipid II (Meniche et al.2014), as well as various transglycosylases and transpeptidases that respectively stitch together the glycan and peptide to create the mature PG (Hett, Chao and Rubin 2010; Plocinski et al.2011; Kieser et al.2015; Viswanathan et al.2017). These studies are supported by experiments that examined the localization of PG itself. Indeed, one of the first pieces of evidence indicating this growth model was the polar staining by a fluorescent conjugate of vancomycin, an antibiotic that binds uncrosslinked PG at the sites of synthesis (Thanky, Young and Robertson 2007). More recent bioorthogonal (click) metabolic labeling using structural analogs of PG biosynthetic precursors, such as alkyne-functionalized d-Ala, or d-Ala-d-Ala, further suggest that PG synthesis occurs at the poles of mycobacterial cells (Siegrist et al.2013; Meniche et al.2014; Boutte et al.2016, 2017; Rodriguez-Rivera et al.2017; Schubert et al.2017; Hayashi et al.2018).

Spatially coincident synthesis of AG layer and OM at the polar ends

Given the covalently linked mAGP complex, assembly of these structures is likely to be coordinated. This idea is supported by biochemical data showing that ligation of AG to PG occurs only after the latter has been crosslinked by transpeptidases (Hancock et al.2002). Several lines of evidence suggest that synthesis of mAGP occurs either at the same place or in close proximity within the cell. First, antibiotics that target different layers cause the same morphological defect, i.e. polar lysis (Bardou et al.1996; Makarov et al.2009; Kumar et al.2012; Botella et al.2017). Second, fluorescent protein fusions to enzymes involved in the cytoplasmic steps of AG and OM synthesis are preferentially enriched at the poles (Carel et al.2014; Meniche et al.2014; Hayashi et al.2016), and MmpL3, the transporter for the trehalose monomycolate precursor, also dynamically localizes to the growing poles in addition to its enrichment at the septa (Carel et al.2014). To our knowledge, localization of proteins involved in the extracellular steps of AG and OM synthesis has not yet been reported for mycobacteria. However, a fluorescent protein fusion to the AG-transferring protein LcpA colocalizes with metabolically labeled PG at the cell tips of the related species Corynebacterium glutamicum (Baumgart et al.2016). Furthermore, metabolic labeling experiments of the OM MA suggest that the assembly is enriched at the poles of mycobacteria (Backus et al.2011; Swarts et al.2012; Foley et al.2016). Taken together, these experiments insinuate that the mAGP synthesis is likely to be spatially coincident or in close proximity during elongation.

Cell envelope synthesis at the septum

Electron microscopy (EM) studies showed the formation of the mycobacterial septum near the midcell (Takade et al.1983; Vijay, Anand and Ajitkumar 2012). However, the precise molecular constituents of the electron-transparent and electron-dense layers remain unclear, and the dynamics of division must be inferred. Fluorescent labeling of both proteins and cell envelope precursors have begun to fill these gaps. It is important to realize that the biogenesis of the septum is a process to establish new poles for the daughter cells, and nearly all of the biomolecules described above for polar growth appear to localize to the midcell region as well. In mycobacteria, key PG biosynthetic enzymes, such as lipid II synthase MurG, a putative SEDS family PG glycosyltransferase FtsW (Meeske et al.2016) and transglycosylase/D,D-transpeptidases PBPA and PBPB (FtsI) are members of the divisome (Fig. 2) (Datta et al.2002; Rajagopalan et al.2005b; Datta et al.2006; Plocinski et al.2011) (see section ‘FtsZ, divisome, and division plane determination’ for more detail). These interactions are also essential in other bacteria (Pinho and Errington 2003; Aaron et al.2007; Mohammadi et al.2007), and recent studies point that PG synthesis is necessary for driving cytokinesis in E. coli and B. subtilis, possibly by providing constriction force during the onset of septum formation (Coltharp et al.2016; Monteiro et al.2018). Nevertheless, mycobacteria also have distinct proteins that regulate septal envelope synthesis. The membrane-bound CwsA and CrgA ensure proper PG assembly by influencing the localization of PBPs and DivIVA (Plocinski et al.2011, 2012), and because DivIVA interacts with enzymes responsible for MA biosynthesis (Meniche et al.2014) (see below), it is likely that these associations also drive septal MA production as illustrated by metabolic labeling (Swarts et al.2012; Foley et al.2016). Importantly, the coordinated activity of PG hydrolytic, modifying and synthetic enzymes localized to the septum ensures proper separation of the mother cell PG into two daughter cells (Chauhan et al.2006; Hett et al.2007, 2008; Hett, Chao and Rubin 2010; Vadrevu et al.2011; Senzani et al.2017).

Other membrane-bound proteins with spatially restricted localizations

In addition to cell envelope biosynthetic machineries localizing to restricted regions, there are additional proteins that show spatial organizations within the mycobacterial PM. Here we highlight some that are relevant to envelope biosynthesis.

Signal transduction kinases show distinct localizations within mycobacterial PM. Mycobacteria have a repertoire of Ser/Thr protein kinases. Among them, PknA and PknB localize at the poles and septum, and regulate enzymes involved in synthesis of PG and other cell envelope components (Molle et al.2006; Thakur and Chakraborti 2006, 2008; Parikh et al.2009; Mir et al.2011; Gee et al.2012; Baer et al.2014; Kieser et al.2015). Other kinases (PknD, PknE, PknH, PknJ and PknL) also localize at the poles and septum, but their contribution to envelope synthesis remains less well-defined (Baer et al.2014). Mycobacteria also utilize two component regulatory systems albeit not as extensively as other bacteria. Among them, MtrAB is the only system known to be essential, and the sensor kinase MtrB is localized at the septum (Via et al.1996; Zahrt and Deretic 2000; Plocinska et al.2012). The response regulator MtrA activates genes involved in DNA replication, drug resistance, and cell wall biogenesis (Fol et al.2006; Li et al.2010; Rajagopalan et al.2010; Plocinska et al.2012; Purushotham et al.2015), and the disruption of the signaling pathway results in abnormal cell division, altered cell morphology and increased cell wall permeability (Möker et al.2004; Cangelosi et al.2006; Nguyen et al.2010; Plocinska et al.2012), implying its critical role in cell envelope biosynthesis. More detailed discussion on these kinases can be found in recent reviews (Parish 2014; Prisic and Husson 2014).

Another important mediator of bacterial signal transduction is cyclic diguanosine monophosphate (c-di-GMP), a ubiquitous second messenger, which can induce various physiological responses in the bacterial cell and community. In mycobacteria, c-di-GMP is critical for survival during carbon starvation (Gupta, Kumar and Chatterji 2010; Bharati et al.2012; Bharati, Swetha and Chatterji 2013), and downstream effectors include a number of proteins involved in regulating cell envelope metabolism (Li and He 2012; Deng et al.2014; Zhang et al.2017). DcpA, a bifunctional diguanylate cyclase and phosphodiesterase, orchestrates the c-di-GMP turnover, and localizes to the poles during early stationary phase (Sharma et al.2014). Experimentally altering the DcpA levels affects cell length and colony morphology of Mycobacterium smegmatis (Sharma et al.2014), implicating the importance of c-di-GMP signaling in controlling the cell division cycle and surface architecture.

Fatty acid synthesis is critical for numerous aspects of the cell envelope biogenesis. The initial stage of fatty acid synthesis is mediated by a multifunctional cytoplasmic enzyme, FAS-I, which produces two populations of fatty acyl-CoA carrying 16–18 or 24–26 carbons. The C_16–18 acyl-CoA is further elongated by the FAS-II system, leading to the production of the long meromycolate precursors. The β-ketoacyl-ACP reductase MabA and enoyl-ACP reductase InhA are part of the FAS-II system, and GFP fusions of these enzymes are located at the polar regions of growing cells (Fig. 2) (Carel et al.2014). In contrast, β-ketoacyl-ACP synthases KasA/B are more distributed throughout the cell body perhaps because the condensing enzymes must acquire C_16–18 acyl-CoA released from the cytoplasmic FAS-I system. C_24–26 acyl-CoA is modified to 2-carboxyacyl-CoA, an important precursor for MA biosynthesis, by an acyl-CoA carboxylase complex composed of AccA3, AccD5 and AccD4 (Gande et al.2007). The Acc enzyme complex is found at the polar ends of the cell, physically interacting with DivIVA (Fig. 2) (Meniche et al.2014; Xu et al.2014) (see below), further indicating that MA production takes place in the polar regions of mycobacterial cells.

CYTOSKELETAL PROTEINS AS POSITIONAL CUES FOR SUBCELLULAR PROTEIN LOCALIZATION

How membrane proteins localize within the mycobacterial cell remains an open question, but cytoskeletal proteins likely play important roles. For example, in dividing cells, tubulin-like FtsZ assembles in a ring-like structure near the midcell and serves as a platform to recruit other proteins required for cell division (Hong, Deng and Xie 2013). While many rod-shaped bacteria use the actin-like MreB to organize PG synthesis and to create and maintain their shape (Gitai, Dye and Shapiro 2004; Shiomi, Sakai and Niki 2008; Billings et al.2014), mycobacteria have no known ortholog of MreB. The lack of MreB demands a different mechanism for spatial positioning of cell envelope biosynthesis in mycobacteria. In this section, we describe known cytoskeletal proteins in mycobacteria as well as a group of coiled coil proteins that are likely involved in the early stage of determining the subcellular membrane protein localization.

FtsZ, divisome and division plane determination

FtsZ localizes near the center of mycobacterial cell where the future division plane is generated (Fig. 2) (Dziadek et al.2003; Datta et al.2006; Plocinski et al.2012; Eskandarian et al.2017; Rego, Audette and Rubin 2017). Depletion or overexpression of FtsZ in mycobacteria results in filamentous cells that fail to form septa (Dziadek et al.2003). In E. coli, GTP hydrolysis by FtsZ was pointed to generate force for the Z-ring to constrict membranes (Mukherjee and Lutkenhaus 1998; Lu, Reedy and Erickson 2000; Osawa, Anderson and Erickson 2008; Allard and Cytrynbaum 2009), but recent findings suggest that the GTPase activity is instead necessary for regulated assembly of FtsZ, and the septum closure and membrane constriction is driven by envelope synthesis (Coltharp et al.2016; Monteiro et al.2018). In mycobacteria, the GTPase activity of FtsZ is necessary for its function (Rajagopalan et al.2005a), but the Z-ring is more stable, with significantly slower GTPase activity than its E. coli counterpart (White et al.2000; Chen et al.2007). The physiological significance of these differences remains unknown.

Homologs of known divisome proteins, which modulate the dynamic activities of FtsZ at different levels, have been identified in mycobacteria. Among them, SepF is an FtsZ-interacting protein that is crucial to form and stabilize the Z-ring (Fig. 2) (Gupta et al.2015), consistent with its role in other Gram-positive bacteria as a critical membrane-bound protein that recruits FtsZ (Duman et al.2013). FipA, another FtsZ-interacting protein, forms a ternary FtsZ-FipA-FtsQ complex (Fig. 2) (Sureka et al.2010). It is needed only under oxidative stress to stabilize the Z-ring and FtsZ-FtsQ interaction, and the FipA-FtsZ interaction depends on phosphorylation of FipA by PknA. Nonetheless, the FtsZ-FtsQ interaction can also be mediated by CrgA (Fig. 2) (Plocinski et al.2011), possibly explaining why FipA is dispensable for the FtsZ-FtsQ association under normal growth. In other bacteria, FtsQ is a component of FtsBLQ subcomplex important for septal ring activity (Liu et al.2015; Tsang and Bernhardt 2015). FtsB and FtsL were recently identified in mycobacteria, and their septal localization is dependent on FtsQ (Wu et al.2018), indicating that mycobacterial FtsBLQ might play a similar role in septal ring formation. One more potential partner of FtsZ that is needed for proper septation is WhmD (Gomez and Bishai 2000; Raghunand and Bishai 2006). WhmD-GFP showed cytoplasmic localization, which led to the initial conclusion that it is not associated with the divisome, but recent in vitro studies suggested that it interacts and stabilizes FtsZ (Fig. 2) (Bhattacharya, Kumar and Panda 2017), warranting further investigations to determine its precise role. In addition to these FtsZ-interacting positive regulators, there are negative regulators that interfere with cell division (Dziedzic et al.2010; England, Crew and Slayden 2011). Taken together, the mycobacterial divisome is being revealed as a complex protein machinery with intricate regulatory systems to precisely control cell envelope biosynthesis and remodeling.

The Min and nucleoid occlusion systems ensure precise septum formation at midcell in E. coli and B. subtilis by using either polar or DNA-bound proteins that inhibit FtsZ polymerization (Dajkovic et al.2008; Monahan et al.2014). Mycobacteria, however, lack the Min system, and the presence of a nucleoid occlusion system is still being investigated. One study showed that the Z-ring forms randomly, independently of DNA positioning, which resulted in cells septating at subpolar regions and/or containing no DNA (Singh et al.2013). Other groups have found that division occurs at or near midcell, resulting in daughter cells with variable but comparable lengths (Takade et al.1983; Rajagopalan et al.2005b; Joyce et al.2012; Vijay, Anand and Ajitkumar 2012; Santi et al.2013; Gola et al.2015; Eskandarian et al.2017; Rego, Audette and Rubin 2017; Senzani et al.2017). In particular, a recent study in M. smegmatis visualized the in vivo dynamics of the chromosomal terminus of replication in addition to the replisome and the chromosomal origin of replication (oriC) during cell cycle (Logsdon et al.2017). The authors provided rigorous data sets, demonstrating that the positioning of the chromosomal terminus as well as the replisome prior to division is off-centered, being slightly closer to the new pole. This off-centered positioning of two replicated chromosomes predicts the site of septum formation, implying the potential presence of a nucleoid occlusion system. In addition, another recent report provided evidence that the Z-ring positioning is not random (Eskandarian et al.2017). Using correlated optical and atomic-force microscopy, the authors revealed the undulating surface of mycobacterial cells and further demonstrated that the Z-ring forms on the waveform troughs of their surface. Remarkably, these waveform troughs are present as morphological features in the mother/grandmother cells long before Z-ring formation. The authors proposed that these morphological landmarks serve as licensed locations for the Z-ring formation. Using a mutant defective in DNA partitioning (lacking the ParAB system, see section ‘ParA and segregation of DNA’), they further showed that the presence of the nucleoid can negatively influence the Z-ring formation, supporting the idea that a nucleoid occlusion system does exist in mycobacteria.

ParA and segregation of DNA

ParA and MinD comprise a class of cytoskeletal proteins that are uniquely found in bacteria (Gitai 2007; Cho 2015). While MinD is absent in actinobacteria, the ParAB system is widely distributed. Chromosomal DNA segregation is an active, vital process that needs to be executed reliably each time cells divide. Recent studies revealed that M. smegmatis mutants lacking either parA or parB, while being viable, show chromosomal segregation defects resulting in more frequent appearance of anucleate cells (Jakimowicz et al.2007; Ginda et al.2013). OriC-proximal parS sequences is the signature recognized by mycobacterial ParB (Jakimowicz et al.2007), and phosphorylation of ParB regulates its DNA binding and polar localization within the cell (Baronian et al.2015). ParA also facilitates the interaction of ParB and parS in vitro (Jakimowicz et al.2007). In vivo, ParA physically interacts with DivIVA (see below), transiently localizes to the new pole and drives the separation of the duplicated ParB-parS complex to the opposite poles, indicating its critical role in spatial segregation of chromosomal DNA (Donovan et al.2010; Ginda et al.2013, 2017).

Donovan and colleagues also revealed ParA-like proteins involved in division site determination, Rv1708 and Rv3213c in M. tuberculosis. While overproduced GFP fusions of M. tuberculosis ParA and Rv1708 showed similar polar enrichment, GFP-tagged Rv3213c showed more homogeneous membrane fluorescence (Maloney, Madiraju and Rajagopalan 2009). These authors further demonstrated that overproduction of these M. tuberculosis proteins in M. smegmatis made the cells filamentous with multiple nucleoids. These observations suggest that these ParA homologs may also play roles in spatially controlling cell-cycle progression.

DivIVA and asymmetric cell envelope elongation

DivIVA (also known as Wag31 or Ag84 in mycobacteria) is a widely conserved coiled coil protein in bacteria. In actinobacteria, it was first shown to be essential for apical growth and enriched at the cell poles in Brevibacterium lactofermentum and Streptomyces coelicolor (Flärdh 2003; Ramos et al.2003). Accumulating evidence suggests that DivIVA is key for orchestrating polar growth in mycobacteria (Fig. 2). First, DivIVA localizes asymmetrically at the cell tips and is enriched at the older, faster growing pole (Nguyen et al.2007; Kang et al.2008; Jani et al.2010; Meniche et al.2014; Botella et al.2017). Second, the phosphorylation of the protein positively regulates its localization and correlates with optimal growth and more intense polar staining by fluorescent vancomycin (Kang et al.2008; Jani et al.2010). Finally, silencing or depletion of DivIVA results in bulged and spherical cells, suggesting that it is important for maintaining the rod morphology (Kang et al.2008; Meniche et al.2014). Given that the protein physically interacts with enzymes required for MA precursor synthesis (Carel et al.2014; Meniche et al.2014; Xu et al.2014), it appears that at least one function of DivIVA is to organize the envelope biogenesis to promote polar growth and rod shape. Recruitment of DivIVA, in turn, has been hypothesized to be dependent on a curvature-sensing mechanism and on the availability of envelope precursors (Meniche et al.2014). In the future, one strategy for untangling the complex relationships between cell shape, cytoskeletal-like proteins and envelope synthesis may be to ‘start from scratch’ and examine these interactions during de novo pole formation (Billings et al.2014; Meniche et al.2014; Ranjit, Jorgenson and Young 2017).

FilP filament and polarisome

FilP is a coiled coil protein found in S. coelicolor. It is distantly homologous to DivIVA, and has been shown to form a filament both in vitro and in vivo. FilP interacts with DivIVA, the main component of apical protein complex known as polarisome in Streptomyces (Fuchino et al.2013). The filP deletion mutant shows morphological defects and reduced tolerance to mechanical stress (Bagchi et al.2008), suggesting its roles in structural integrity of the cellular architecture. A FilP homolog is found in many species of Mycobacterium, although it is notably missing in M. smegmatis. The M. bovis homolog, Mb1709, has been shown to form a filament in vitro (Bagchi et al.2008). While M. tuberculosis ortholog Rv1682 is predicted to be non-essential based on genome-wide transposon mutagenesis studies (Griffin et al.2011), it is intriguing if the protein binds to DivIVA and plays roles in cell morphology and mechanical strength as observed in S. coelicolor.

SepIVA, a DivIVA-like protein at the divisome

A novel coiled-coil protein termed SepIVA (MSMEG_2416), reported recently in M. smegmatis (Wu et al.2018), is a distant homolog of DivIVA. The gene is predicted to be essential in M. tuberculosis (Griffin et al.2011), and found widely in mycobacteria, including the degenerate genome of M. leprae. It was identified as a protein associated with the divisome (Fig. 2), and its depletion results in cells elongating without dividing, leading to long branched filamentous cells. SepIVA becomes enriched at the septum transiently during cell division, and septal enrichment of FtsZ precedes that of SepIVA. When cells are not dividing, SepIVA appears to associate with the intracellular membrane domain (IMD), a spatially distinct domain of the PM (see below), suggesting a dynamic property to change its subcellular localization during cell cycle.

PspA and membrane stress response

PspA/RsmP are another important group of coiled coil proteins. RsmP, a corynebacteria-specific protein, which forms a filament in vitro and in vivo, is upregulated upon DivIVA depletion, and is required for polar growth of C. glutamicum (Fiuza et al.2010). PspA, in contrast, is widely conserved in bacteria and mediates stress response to membrane damage (Joly et al.2010; Flores-Kim and Darwin 2016; Manganelli and Gennaro 2017). Mycobacterial pspA is part of a regulon controlled by mprAB-sigE and clgR (Manganelli et al.2001; He et al.2006; Estorninho et al.2010). The MprAB two-component system and Sigma factor E (σE) sense cell envelope stress, positively regulate each other, and activate the transcription factor ClgR that induces the expression of the clgR-pspA-rv2743c-rv2742c operon (Manganelli et al.2001; He et al.2006; Barik et al.2010; Tiwari et al.2010; Bretl et al.2014). The observation that PspA is part of this regulon suggests a link between envelope stress signaling and envelope maintenance. Indeed, PspA interacts with mycobacterial membranes (Rumschlag et al.1990; Mawuenyega et al.2005; Datta et al.2015), and the membrane localization is mediated through a complex formation with two membrane proteins, Rv2743c and Rv2742c (White et al.2011; Datta et al.2015). This protein complex is proposed to have an envelope-stabilizing function similar to that observed in Gram-negative bacteria (Provvedi et al.2009; Datta et al.2015). Nonetheless, another study reported that pspA deletion mutant shows unaltered resistance to envelope stress, and suggested that PspA is involved in lipid droplet homeostasis (Armstrong et al.2016), possibly implying a divergent role of PspA in mycobacteria.

MEMBRANE HETEROGENEITY AND ITS LINK TO DIFFERENTIAL PROTEIN LOCALIZATION

There is accumulating evidence that bacterial PM is not homogeneous (Barák and Muchová 2013; Farnoud et al.2015; Matsumoto et al.2015; Huang and London 2016; López and Koch 2017). However, the contribution of membrane heterogeneity to the spatial coordination of the cellular processes in mycobacteria remains largely unknown. Here, we will discuss the current understanding.

Microscopic evidence supporting the presence of PM heterogeneity

M. tuberculosis was one of the first microbes examined by EM. Numerous early studies suggested that mycobacteria, like many other bacteria, have intracellular membrane-bound structures that were variably called intracytoplasmic membranous configurations, mitochondrial equivalents or mesosomes. The earliest studies reporting mycobacterial mesosomes were published in 1957 (Shinohara, Fukushi and Suzuki 1957; Zapf 1957), and investigation on these structures continued for the next few decades in mycobacteria as well as in other bacteria. However, the physiological roles of these structures were never proven and cryo-EM technologies eventually revealed these structures as artifacts of chemical fixation (Nanninga et al.1984; Ryter 1988). Since then, there has been no convincing EM evidence for the presence of lateral heterogeneity of the mycobacterial PM.

In prokaryotic cell biology, the use of fluorescent probes had been limited because of cell size and resolution constraints of light microscopy. Christensen and collaborators were the first to describe the lateral heterogeneity of mycobacterial membrane using fluorescent lipid probes (Christensen et al.1999). Probes with different hydrophobic and amphiphilic properties revealed different labeling patterns of mycobacterial cell envelope and membranes (Christensen et al.1999), suggesting differences either in lateral composition of the cellular membranes or in the physical state of the lipids.

FM4-64 is a hydrophobic red fluorescent dye that has been used to visualize PM domains in E. coli (Fishov and Woldringh 1999). It has been used to visualize mycobacterial cell membranes as well, but unlike E. coli, the dye shows homogeneous annular fluorescence in mycobacteria with more intense accumulation in the septal region, without indication of lateral membrane heterogeneity (Maloney, Madiraju and Rajagopalan 2009; Plocinski et al.2011; Singh et al.2013; Fay and Glickman 2014; Kieser et al.2015; Sharma et al.2016; Wu, Gengenbacher and Dick 2016). This result may suggest that mycobacteria do not form PM domains. However, FM4-64 has been used a membrane-impermeant dye to examine the membrane dynamics during Bacillus sporulation (Pogliano et al.1999; Sharp and Pogliano 1999), and there has been no verification of the exact subcellular locations FM4-64 is staining in mycobacteria. Therefore, it is possible that the dye is trapped in the OM without reaching the PM.

Nonyl acridine orange (NAO) is a fluorescent dye thought to be specific to CL. A recent study suggested that NAO actually binds not only to CL but also other lipids such as phosphatidylglycerol (Oliver et al.2014), but a specific biophysical mechanism of polar phosphatidylglycerol accumulation in E. coli remains unclear (Matsumoto et al.2015). NAO has been used to visualize membrane domains in both M. smegmatis and M. tuberculosis, and the staining was found at the poles and septa (Maloney et al.2011). This finding is consistent with the idea proposed in model bacteria that NAO visualizes CLs that are enriched at areas of negative curvature such as septa and poles (Mileykovskaya and Dowhan 2000; Kawai et al.2004). Because CL is not a major component of the OM (Bansal-Mutalik and Nikaido 2014), these observations imply that the PM of mycobacteria is laterally heterogeneous.

While sterols are rarely found in bacteria, cholesterol glycolipids exist as lipid rafts in the OM of Borrelia burgdorferi (LaRocca et al.2010). In addition, Bacillus and Staphylococcus are known to form lipid rafts, which are enriched in squalenes, a class of molecules structurally similar to cholesterols (López and Kolter 2010; García-Fernández et al.2017). So far, there is no evidence suggesting the presence of such a liquid-ordered ‘rigid’ membrane microdomain in mycobacteria. A recent study indicated that various fluorescent cholesterol analogs and their catabolites stain the PM of M. smegmatis and M. tuberculosis homogeneously (Faletrov et al.2017), suggesting that there are no membrane domains to which such cholesterol analogs can preferentially accumulate. Both M. smegmatis and M. tuberculosis catabolize cholesterols, and these cholesterol analogs can be oxidized from 3β-hydroxysterols to 3-ketosteroids. This reaction is thought to be mediated by 3β-hydroxysteroid dehydrogenase or cholesterol oxidase. These enzymes are cytoplasmic FAD/NAD-dependent oxidoreductases (Yang et al.2007), providing convincing evidence that the fluorescent cholesterol analogs do reach the cytoplasmic side of the PM. Indeed, during the time course of BODIPY-cholesterol labeling in M. smegmatis, homogeneous annular fluorescent patterns transition to intracellular lipid body labeling in 48 h, further supporting the transient PM labeling of the cholesterol dye (Faletrov et al.2017).

PM heterogeneity revealed by centrifugation

Studying the mycobacterial cell using subcellular fractionation techniques dates back to 1960s (Yamaguchi 1960). The main motivation was to characterize the functional differences between mesosomes and conventional PM. Later, the focus of the pursuit shifted to the purification of the OM from the underlying cell wall and the PM. Despite continuous efforts from many laboratories, the complete purification of the OM remains challenging. The proteomic analysis of the OM preparation reported in the most recent study still showed significant enrichment of PM proteins (Chiaradia et al.2017), suggesting the tight association of the PM and cell wall. As detailed below, we have used density gradient fractionation not to separate the OM from the PM, but to separate a membrane domain from the conventional PM.

To reveal whether enzymes involved in cell wall biosynthesis are compartmentalized in the PM, we applied sucrose density gradient fractionation to crude cell lysates of growing mycobacteria. Initially, two distinct fractions enriched in PM phospholipids, termed PMf and PM-CW, were isolated (Morita et al.2005). PMf denotes a fraction of the plasma membrane that is free of cell wall components. This nomenclature was based on the chemical analysis of fractions using thin layer chromatography and mass spectrometry, showing that the PMf fraction is devoid of cell wall components such as GPLs and AG. In contrast, the PM-CW is a fraction containing both PM and cell wall components. More recently, proteomics and lipidomics demonstrated that the PMf is compositionally distinct from the PM-CW, and fluorescence microscopy further showed that the fluorescent protein-tagged PMf-associated proteins (GlfT2 and Ppm1, see below) form discrete regions within the cell with particular enrichment at the growing cell poles (Hayashi et al.2016). In contrast, a fluorescent protein-tagged PM-CW-associated protein (PimE, see below) showed annular fluorescence, suggesting that the PM-CW is the conventional PM. Based on these observations, we proposed that the PMf is a membrane domain in mycobacteria (Hayashi et al.2016), and renamed the PMf as the IMD (Hayashi et al.2018).

Proteomic and lipidomic analyses suggest that the IMD is an organizing center for the biosynthesis of specific metabolites (Hayashi et al.2016). A notable feature of the IMD is that it is enriched in particular steps of PM lipid biosynthesis. For example, AcPIM6 is synthesized from PI by sequential additions of mannose residues. As shown in Fig. 3A, PimB’ and PimE are two mannosyltransferases involved in transferring the second and fifth mannoses, while PatA is an acyltransferase involved in transferring a fatty acid to one of the mannose residues upon the synthesis of PIM2. GDP-mannose-dependent PimB’ is a cytoplasmically oriented enzyme, while PimE is dependent on polyprenol-phosphate-mannose, suggesting that its active site is on the periplasmic side. The AcPIM6 biosynthesis can be examined in vitro using a radioactive sugar nucleotide, GDP-[³H]mannose. When each density gradient fraction was incubated with GDP-[³H]mannose, the IMD fractions were found enriched in the initial biosynthetic steps up to AcPIM2, whereas the PM-CW fractions were enriched in the later steps to form AcPIM6 (Fig. 3A) (Morita et al.2005). Furthermore, the IMD localization of endogenous PimB’ and the PM-CW localization of an epitope-tagged PimE were confirmed by western blotting (Morita et al.2006; Sena et al.2010; Hayashi et al.2016). A proteomic analysis additionally revealed that the IMD is enriched in PimB’ as well as PatA.

Figure 3.

IMD-associated metabolic reactions. (A) Proposed PIM biosynthesis pathway. The AcPIM2 biosynthesis takes place in the IMD, whereas AcPIM6 is formed in the PM-CW. There is no experimental evidence for the exact PIM intermediate that flips across the membrane. (B) PA and PE biosynthesis. PA is synthesized in the IMD, while PE biosynthesis spans both the IMD and PM-CW. (C) Proposed galactan biosynthesis pathway. The galactan intermediate is synthesized in the IMD and translocated to the periplasm for further modification. Neither the flippase for the galactan precursor nor the exact structure of the flippase substrate is known. (D) PG precursor biosynthesis. Lipid II is formed in the IMD prior to continuing to the periplasmic PG biosynthesis. The putative flippase MurJ and penicillin-binding proteins (PBPA/PBPB) are found in the PM-CW proteome, but additional verification is needed. (E) PPM biosynthesis. Ppm1 acts on PP forming the mannose donor PPM in the IMD. OM, outer membrane; AG, arabinogalactan; PG, peptidoglycan; IMD, intracellular membrane domain; PM-CW, plasma membrane with cell wall; PI, phosphatidylinositol; PIMs, phosphatidylinositol mannosides; PA, phosphatidic acid; PS, phosphatidylserine; PE, phosphatidylethanolamine; LM, lipomannan; LAM, lipoarabinomannan.

IMD: a multifunctional membrane domain

There are many additional biosynthetic reactions that appear to be enriched in the IMD. In 2014, a seminal paper by Meniche et al. reported the concept of the subpolar space, which is spatially separate from the polar DivIVA-anchored Acc enzymes and enriched for cell envelope biosynthetic enzymes such as MurG, GlfT2 and Pks13 (Meniche et al.2014). As detailed below, MurG and GlfT2 are also found in the IMD proteome, suggesting that the physical nature of the subpolar space is at least partially fulfilled by the IMD. We also highlight other IMD-associated metabolism, for which more than one piece of experimental evidence is available to support the IMD localization.

Phosphatidic acid (PA) is a key biosynthetic intermediate of phospholipid biosynthesis. It is produced by sequential transfer of fatty acids from acyl-CoA to glycerol 3-phosphate. The two acyltransferases likely involved in the PA biosynthesis are glycerol-3-phosphate O-acyltransferase (GPAT; MSMEG_4703/Rv1551) and 1-acylglycerol-3-phosphate O-acyltransferase (AGPAT; MSMEG_4248/Rv2182c) (Larrouy-Maumus et al.2013; Law and Daniel 2017). These two enzymes were enriched in the IMD proteome (Hayashi et al.2016). Furthermore, lipidomic analysis showed that PA was enriched in the IMD (Hayashi et al.2016), consistent with the enrichment of the PA biosynthetic enzymes (Fig. 3B). These data together support the notion that PA biosynthesis takes place in the IMD.

Similar to AcPIM6 biosynthesis, the final two steps of PE biosynthesis have been examined by a cell-free system using [³H]serine and [³H]phosphatidylserine (PS). PS synthase reaction was enriched in the PM-CW, while the PS decarboxylase (Psd) reaction was found in the IMD (Morita et al.2005) (Fig. 3B). Proteomic analysis supported the radiolabeling experiment: Psd was found enriched in the IMD (Hayashi et al.2016).

Galactan, a component of AG, is synthesized as a polyprenol-linked precursor. The linker between galactan and PG is composed of l-Rhap-d-GlcNAc-PO₄ where phosphate is attached to the glycan backbone of PG (Bhamidi et al.2008). This linker is synthesized as a part of the polyprenol-linked galactan precursor, and sequentially mediated by d-GlcNAc-1-phosphate transferase (Rv1302), rhamnosyl transferase WbbL (Rv3265c) and galactosyltransferase GlfT1 (Rv3782) (Mikusova et al.1996; Weston et al.1997; Mills et al.2004). The polymerization of galactan chain beyond the addition of the first two galactose residues by GlfT1 is mediated by a processive galactosyltransferase GlfT2. The IMD proteome is enriched in WbbL, GlfT1 and GlfT2 (Fig. 3C). An M. smegmatis strain expressing HA-mCherry-GlfT2 at the endogenous locus further demonstrated the enrichment of the fusion protein in the density gradient fractions corresponding to the IMD (Hayashi et al.2016). Fluorescent microscopy also revealed a patchy distribution of GlfT2 at the sidewall plus polar enrichment (Fig. 2) (Meniche et al.2014; Hayashi et al.2016).

Lipid II is the polyprenol-linked precursor of PG biosynthesis. MurG is the key enzyme that transfers GlcNAc to lipid I resulting in lipid II, and is enriched in the IMD proteome (Fig. 3D) (Hayashi et al.2016). In addition, MurG shows IMD-like fluorescence pattern (Meniche et al.2014), supporting its IMD association. Furthermore, as described above, substantial evidence suggests that PG biosynthesis is concentrated at the growing cell poles. Therefore, polar enrichment of the IMD-associated MurG may be important for the spatially controlled production of lipid II. It would be important to determine the subcellular localization of other proteins involved PG biosynthesis.

Polyprenol phosphate mannose (PPM) is a polyprenol-linked mannose donor utilized by mannosyltransferases that act on the periplasmic side of the PM. It is synthesized by a heterodimer of Ppm1 and Ppm2. Ppm2 is a multitransmembrane protein that anchors the catalytic subunit Ppm1 to the membrane (Baulard et al.2003). Ppm1 was found enriched in the IMD proteome (Fig. 3E) (Hayashi et al.2016). Furthermore, Ppm1-mNeonGreen-cMyc expressed from the endogenous locus was enriched in the IMD fraction in a density gradient fractionation. Fluorescence microscopy also revealed the colocalization of Ppm1 and GlfT2 (Hayashi et al.2016), supporting that Ppm1 is an IMD-associated protein (Fig. 2).

Remarkable features of the IMD

A remarkable feature of the IMD is that its associated proteins identified by proteomic analysis are dominated by proteins with no predicted transmembrane domains (Hayashi et al.2016). With a few exceptions, most transmembrane proteins are excluded from the IMD. This is in contrast to many transmembrane proteins identified in the PM-CW, suggesting that the lack of these proteins is not due to the technical limitation of mass spectroscopic fingerprinting. For some IMD proteins, peripheral membrane association is supported by experimental evidence. For example, crystal structure of GlfT2 demonstrated hydrophobic and positively charged patches on the surface of the protein, which are proposed to interact with membrane phospholipids (Wheatley et al.2012). Two enzymes involved in the PIM biosynthesis, a mannosyltransferase PimB’ and an acyltransferase PatA, were also suggested to have positively charged and hydrophobic residues for peripheral association with anionic PM lipids (Guerin et al.2009; Albesa-Jové et al.2016; Sancho-Vaello et al.2017).

Another notable feature is that the IMD is dynamic and spatially relocated under stress conditions (Hayashi et al.2018). Because the IMD is enriched near the poles when the cell envelope actively elongates, we were prompted to examine the IMD localization when cells are no longer growing. Stationary phase, starvation in phosphate-buffered saline supplemented with 0.05% Tween-80 and antibiotic treatment were tested as stress models. In all cases, the IMD delocalized from the pole and redistributed to the sidewall. The IMD could still be biochemically purified by density gradient fractionation, indicating that the IMD is maintained as a membrane domain even under these stress conditions. Furthermore, polar IMD enrichment can be restored when the starved cells were replenished with nutrient-rich medium. These data collectively demonstrated the dynamism of the IMD as a membrane domain. However, the molecular mechanisms of how the IMD responds to the environmental conditions and relocates within the cell remain completely unknown.

FUTURE VISIONS

Mechanistic understanding of the (a)symmetry of mycobacterial elongation is an important challenge for the future. The controversy over this topic may be in part because of challenges in marking division events and subsequent outgrowth (Aldridge et al.2012; Joyce et al.2012; Santi et al.2013; Singh et al.2013; Wakamoto et al.2013; Kieser and Rubin 2014; Siegrist et al.2015; Botella et al.2017). One complicating factor may have been the use of fluorescent DivIVA fusions in several studies to mark the cell division cycle (Santi et al.2013; Carel et al.2014). Expression of these constructs from a strong promoter leads to more even distribution of the protein (Meniche et al.2014) and of metabolically labeled PG (Botella et al.2017) at either pole. As discussed above, the discovery of the divisome protein LamA is beginning to unveil the molecular mechanism of growth asymmetry (Rego, Audette and Rubin 2017). Together, these data raise the possibility that the degree of mycobacterial growth asymmetry may depend on environmental conditions, which in turn could influence the expression levels of LamA, DivIVA or other, yet-identified deterministic, factors.

Fluorescent fusions have been a powerful tool, but have the potential to alter the function and/or localization of the tagged proteins in unintended ways (Landgraf et al.2012; Swulius and Jensen 2012; Meniche et al.2014; Botella et al.2017). Historically, the function of the fluorescent protein fusions was often not experimentally verified. Moving forward, it will be important to perform functional assays on mutant strains, and, when possible, opt for monomeric fluorescent tags that are less prone to clustering artifacts (Landgraf et al.2012). In many cases, particularly for envelope biogenesis, small molecule probes will offer a complementary means for imaging enzyme function (Foss, Eun and Weibel 2011; Kocaoglu and Carlson 2013; Siegrist et al.2015).

How lateral heterogeneity of the PM contributes to the positioning of divisome and formation of the apical protein complex (or mycobacterial polarisome) remains to be determined. Does the enrichment of CL at the septum drive DivIVA localization? Is the polar localization of so many proteins solely dependent on DivIVA? Is the subpolar IMD enrichment dependent on DivIVA? How do so many proteins localize to the IMD? How does the IMD contribute to mycobacterial cell physiology? In Bacillus, the presence of detergent-resistant lipid raft as well as membrane region of increased fluidity (RIF) have been proposed (López and Kolter 2010; Müller et al.2016). Lipid rafts are liquid ordered structures that have low fluidity, while RIF is a conceptually opposite highly fluid region within the same PM. Whether the IMD is similar to lipid rafts or RIFs is an important biophysical question. Interestingly, daptomycin, which disrupt the RIF in Bacillus, causes the formation of ectopic poles in mycobacteria (Meniche et al.2014). The profound effect of daptomycin on mycobacteria might imply the presence of similar RIF-like region in the PM of mycobacteria.

Conflicts of interest. None declared.