Modulation of the M. tuberculosis cell envelope between replicating and non-replicating persistent bacteria

Abstract

The success of Mycobacterium tuberculosis as a human pathogen depends on the bacterium’s ability to persist in a quiescent form in oxygen and nutrient-poor host environments. In vitro studies have demonstrated that these restricting environments induce a shift from bacterial replication to non-replicating persistence (NRP). Entry into NRP involves changes in bacterial metabolism and remodeling of the cell envelope. Findings consistent with the phenotypes observed in vitro have been observed in patient and animal model samples. This review focuses on the cell envelope differences seen between replicating and NRP M. tuberculosis and summarizes the ways in which serine/threonine protein kinases (STPKs) may mediate this process.

Introduction

Mycobacterium tuberculosis is a highly specialized human pathogen and the causative agent of tuberculosis (TB). TB is still one of the leading causes of death due to a pathogenic disease despite the availability of antibiotics. A defining characteristic of M. tuberculosis is a complex cell envelope that is associated with pathogenesis and that provides a barrier against antibiotics and the host immune response. The mycobacterial cell envelope consists of an inner membrane and a mycomembrane separated by a peptidoglycan-arabinogalactan complex [1]. Based on studies in M. smegmatis, the mycobacterial inner membrane is composed of phospholipids and an abundance of diacyl phosphatidyl dimannosides (Ac₂PIM₂) [2]. The outer mycomembrane is made up of inner and outer leaflets. The periplasmic, inner leaflet is comprised of mycolyl arabinogalactan-peptidoglycan (mAGP), which consists of mycolic acids (MAs), 2-alkyl, 3-hydroxyl long chain (60–90C) fatty acids covalently bound to the arabinogalactan-peptidoglycan layer. The outer leaflet contains species-specific, non-covalently associated lipids, such as trehalose monomycolate (TMM), trehalose 6,6’-dimycolate (TDM), phthiocerol dimycocerosate (PDIM), sulfolipids (SLs), and di- and pentaacyl trehaloses (DATs, PATs) [3]. These surface-exposed lipids, or their biosynthetic precursors, are exported from the cytoplasm by mycobacterial membrane protein large (MmpL) proteins, which are therefore crucial contributors to mycobacterial physiology and pathogenesis [4].

The majority of infections result in latent TB infection (LTBI), where the bacilli enter a non-replicating persistent (NRP) state characterized by phenotypic drug tolerance. The NRP bacterial population can later resuscitate upon progression to active disease. As M. tuberculosis transitions between replication and NRP, the bacterium modulates both metabolism and its cell envelope lipid composition. These physiological changes are of great interest since the underlying cellular processes could be targets of novel therapeutics to treat LTBI. M. tuberculosis encodes 11 eukaryotic-like serine/threonine protein kinases (STPKs) that can mediate a post-translational response to the extracellular environment [5]. Phosphoproteomics of M. tuberculosis grown in different conditions or exposed to environmental stressors identified lipid metabolism and cell envelope/cell process proteins as major subsets of phosphorylated proteins [6]. In this review, we will focus on the differences between the cell envelope composition of replicating and NRP M. tuberculosis and the increasingly appreciated role of STPKs in cell envelope biogenesis and remodeling.

STPKs of M. tuberculosis

M. tuberculosis encodes 11 STPKs that mediate the bacterial responses to environmental conditions (comprehensively reviewed in [5]). Nine of the 11 STPKs are predicted transmembrane receptors with extracellular sensor domains and cytoplasmic kinase domains. They are therefore thought to transduce extracellular signals to intracellular substrate proteins and mediate cellular adaptation via alterations in protein function, protein-protein interactions, or protein stability. The two outliers, PknG and PknK, do not contain transmembrane domains. PknG can be secreted and contributes to the ability of M. tuberculosis to arrest phagosome and autophagosome maturation [7,8]. PknK is proposed to be anchored in the membrane based on subcellular fractionation data [9]. PknA, PknB, and PknG are the most studied via a combination of structural, biochemical and genetic approaches because of their requirement for M. tuberculosis viability (PknA/B) or virulence (PknG) [5].

The role of STPKs in M. tuberculosis biology is strongly supported by global phosphoproteomic, genetic and biochemical analyses. Notably, a seminal study by Prisic et al. identified 301 M. tuberculosis H37Rv phosphoproteins containing at least 516 phosphosites [6]. Phosphorylation occurred ~60% on threonines and 40% on serines. The authors compared phosphoproteins across mycobacteria cultured with different carbon sources, at different stages of growth, and exposed in vitro to host-associated stressors including NO, hydrogen peroxide, and hypoxia. While phosphoproteins belonged to all functional classes of proteins, most were annotated as involved in cell wall/cell processes and intermediary metabolism/respiration. Verma et al. subsequently analyzed two technical replicates of H37Ra and H37Rv harvested at logarithmic and stationary phases [10]. They identified 257 phosphorylated proteins out of a total of 2709 proteins. Of these, 94 proteins were differentially expressed in stationary vs. logarithmic phase. While their analyses focused on differences between H37Ra and H37Rv, our analysis of their data revealed 408 phosphorylated sites in 148 proteins that significantly differ in logarithmic vs. stationary phase of H37Rv, suggesting a potential role for phosphorylation in the transition to persistence. This study had limited numbers of replicates, but it confirmed many phosphorylation sites identified by Prisic et al., and identified 265 novel phosphosites. A recently published phosphoproteomic study identified 713 proteins as phosphorylated [11], demonstrating how extensive this post-translational modification is in M. tuberculosis.

STPKs likely contribute to transition into, survival during, and resuscitation from NRP. In particular, the roles for the essential kinases PknA and PknB and the PknG kinase in NRP have been investigated. PknA and PknB are essential for M. tuberculosis growth in most conditions [12,13]. Several studies demonstrate that modulation of PknA and B alters phenotypes associated with NRP: There was significant overlap between the transcriptional profile of PknA/PknB inhibitor-treated bacteria and M. tuberculosis during hypoxia, an environmental signal associated with dormancy [14]. PknB levels decrease during dormancy and are restored upon resuscitation [15]. Ortega and colleagues showed that PknB is key for the transition of M. tuberculosis out of in vitro dormancy conditions. PknB inhibition resulted in a hypoxic M. tuberculosis culture that failed to grow upon reaeration, and overexpression of PknB resulted in growth defects upon reaeration [15]. Changes in PknB levels also impacted replicating, aerated cultures [15]. A study of individual and dual PknA-PknB depletion strains revealed a consistent loss in acid-fastness and increased antibiotic susceptibility [11].

As mentioned above, PknG is secreted and impacts M. tuberculosis pathogenesis [7,8]. PknG also functions in the bacterial cytoplasm. PknG phosphorylation of GarA in response to nutrient availability regulates the TCA cycle [16]. PknG is also a component of the redox homeostatic system (RHOCS) that contributes to oxidative stress resistance [17]. Relevant to NRP, an M. tuberculosis pknG mutant had decreased survival in the Wayne model of hypoxia, and the authors attributed this phenotype to the inability of this strain to remodel central carbon metabolism [18]. M. tuberculosis pknG mutants had reduced virulence in a guinea pig model of infection and specifically showed defects in establishing granulomas, a hallmark of LTBI [18].

Finally, though less well-studied, PknH may also contribute to M. tuberculosis transition into NRP. Mutants lacking PknH replicate to a higher load than wild-type M. tuberculosis during chronic infections in mice [19]. Proteomics analysis of this mutant demonstrated dysregulation of the DosR regulon that is associated with NRP [20]. Using in vitro kinase assays and gel shift assays, Chao et al. demonstrated that phosphorylation of DosR by PknH enhanced the DNA binding activity of this transcriptional regulator. Combined, these results suggest that PknH is important for adaptation to the host environment and establishment of NRP phenotypes via cross talk with the DosRST two-component regulatory system.

Cell envelope and metabolic differences in replicating versus NRP M. tuberculosis

Studies utilizing clinical samples and animal models show that M. tuberculosis physiology differs between the replicating and NRP states. M. tuberculosis is termed acid-fast because the cell envelope causes the bacteria to be resistant to destaining by acid alcohol solutions. While the components of the M. tuberculosis cell envelope that contribute to acid-fastness are still being elucidated, it is thought to be due primarily to MAs and glycolipid components of the cell envelope [21]. Acid-fastness is commonly used to detect M. tuberculosis in clinical specimens and samples from experimental infections. However, in some samples where acid-fast bacilli are not visible, M. tuberculosis can be detected by other means. For instance, late stage murine pulmonary TB samples (>30 weeks post infection) were dramatically less acid-fast than the early infection counterparts despite maintaining a high bacterial load [22]. Similarly, these investigators found that human lung samples from patients with acute, active infection were acid-fast, while samples from LTBI patients were negative for acid-fastness. In a separate study, 39% of sputum samples collected from TB patients in San Francisco were negative for acid-fast staining [23]. These patients may represent a less infectious group, but they were still responsible for an estimated 17% of transmission. In addition to a loss of acid-fastness, intracellular lipophilic inclusions (ILIs) were observed within mycobacteria obtained from patient samples but not in aerated in vitro cultured cells [24]. Roughly 45±20% of M. tuberculosis cells recovered from patient sputum samples contain ILIs, and the percent of ILIs containing bacteria in a sputum sample correlated with time to positivity in a diagnostic liquid culture [25].

Granulomas are characteristic of LTBI and are generally characterized as nutrient- and oxygen-restricted [26]. In vitro models that incorporate these physiological conditions were developed to study the mechanisms underlying NRP. In the Wayne model, M. tuberculosis is grown in screw-cap tubes in Dubos Tween-albumin medium with an air volume to media volume ratio, or Head Space Ratio (HSR), of 0.5. The bacteria slowly use up the oxygen via aerobic respiration and enter NRP stage I when the oxygen saturation in the HSR reaches 1%. Bacteria continue through this microaerophilic stage and move into full anaerobic conditions defined by less than 0.06% oxygen saturation in the HSR, referred to as NRP stage II. This is traditionally considered NRP [27]. The Hamilton model is a long-term adaptation model of nutrient starvation that uses a chemostat to maintain bacterial cultures at optimal oxygen concentration and pH while they gradually transition into nutrient-depleted NRP conditions that can be maintained for upwards of 60 days [28]. A more rapid carbon starvation model grows M. tuberculosis in nutrient-rich medium, then washes and resuspends the culture in PBS in sealed containers with limited head space. The bacteria rapidly experience nutrient and oxygen depletion and show phenotypic resistance to antibiotic treatment [29,30]. This model is frequently utilized in vitro to determine whether novel therapeutics are effective against NRP bacteria [31–33]. Finally, we and others also use biofilms as a model for NRP [34–36]. Bacteria are cultured in tightly sealed polystyrene bottles in Sauton’s medium lacking Tween-80. After two to three weeks, the lids are loosened and cultures incubated for an additional two weeks, at which point a pellicle forms at the air-liquid interface and reticulates. This model gradually transitions bacteria into NRP as nutrients are depleted and aggregated bacteria in the pellicle experience nutrient and oxygen restriction. The NRP population exhibits phenotypic resistance to antibiotics [34]. In vitro biofilms are relevant to TB disease because extracellular M. tuberculosis are observed in necrotic granulomas of humans and model organisms, and the bacterium develops a corded/multicellular complex functionally similar to biofilms [37]. Additionally, data from stationary phase bacteria is informative because these bacteria encounter similar environments as NRP bacteria such as nutrient depletion, and many genes that are upregulated in NRP are upregulated during stationary phase [38].

The NRP-associated phenotypes identified in vivo were also observed using the above in vitro models: In a multiple-stress NRP model of acidic pH, low oxygen concentration, high carbon dioxide concentration, and nutrient deprivation, M. tuberculosis gradually lost acid-fastness [39]. Consistent with loss of acid-fastness, levels of MAs were reduced in the cell envelope of bacteria exposed to a modified Wayne model of hypoxia compared to replicating bacteria [40]. Mycobacteria also accumulated ILIs in vitro using hypoxia and multiple stress models of NRP [25,39]. ILIs from in vitro models contained triacylglycerols (TAGs) of different fatty acid moieties depending on the growth medium [24]. In infected macrophages, M. tuberculosis metabolized host TAGs to synthesize new bacterial ILIs [41]. Bacteria accumulated lipid inclusion bodies over time, regardless of oxygen supply, albeit more extensively in hypoxic conditions.

M. tuberculosis mutants lacking TAGs and wax esters (WEs) have defects in NRP. TAGs and WEs are considered storage molecules for M. tuberculosis during NRP and reactivation. Utilization of TAGs yields more energy than other storage compounds like glycogen [42]. There are 15 genes in M. tuberculosis that encode diacylglyerol acyltransferases capable of synthesizing TAGs or WEs when expressed in Escherichia coli. Several, including the primary TAG synthase Tgs1, are induced in M. tuberculosis upon transition into NRP via hypoxia, NO treatment, or nutrient deprivation [43]. Production of TAGs during hypoxia restricted the growth rate of M. tuberculosis by shifting carbon usage away from the tricarboxylic acid (TCA) cycle [44]. WEs are required for the transition of M. tuberculosis into an NRP state using an in vitro multiple stress model. Mutants lacking the Fcr1 and Fcr2 fatty acyl-CoA reductases that generate the fatty alcohol precursors for WEs continued to replicate and failed to establish phenotypic antibiotic resistance [45]. The authors of this study proposed that the accumulation of WEs also attenuates growth by limiting nutrient uptake by the pathogen in stress conditions. In mycobacterial biofilms, we previously showed that mycolate wax esters (MWEs) and long-chain TAGs (LC-TAGs) accumulate in the cell envelope and are isolated with surface exposed lipids. MWEs and LC-TAGs are exported by MmpL11, and mmpL11 mutants have impaired biofilm formation and reduced survival and/or resuscitation from the PBS starvation model of NRP [35]. These data support the model that LC-TAGs and MWEs play important roles in M. tuberculosis persistence, but there is little data to show how these external storage lipids are utilized during resuscitation. We are currently testing the model that upon resuscitation, LC-TAGs and WEs are hydrolyzed and reimported by M. tuberculosis for use as a carbon source and/or incorporated into newly synthesized cell envelope lipids.

Lipidomic analyses of NRP cultures have defined additional cell envelope modifications that accompany the transition to NRP (Figure 1). In hypoxic cultures, levels of DATs/PATs and SLs were unaltered while PDIM levels declined [46]. M. tuberculosis cultured in the Hampshire model chemostat under nutrient depletion gradually produced and accumulated free mycolates, which corresponded with a decrease in both TMM and TDM [38]. These results suggested that TMM and TDM are hydrolyzed in NRP bacteria to release free MAs. In addition, levels of lipoarabinomannan (LAM) vs lipomannan (LM) increased over time, with an increase in the arabinose:mannose ratio in LAM composition [38]. Subsequent metabolomic studies corroborated some of these results. Metabolomic analysis of hypoxic M. tuberculosis cultures demonstrated that extracellular TMM and TDM were catabolized to trehalose and free mycolates. While the fate of the MAs constituent was not investigated, trehalose was imported by the trehalose specific ABC transporter LpqY-SugABC [47] and redirected to generate a pool of pentose phosphates that drive peptidoglycan (PG) synthesis upon reaeration [48]. Metabolomic analysis of biofilm-associated M. tuberculosis also showed that trehalose was utilized as a carbon source for NRP organisms [49]. Combined, altered MAs and LAM/LM composition likely impact virulence and the ability of M. tuberculosis to establish chronic infections. The hydrolysis of TDM and TMM, known pro-inflammatory molecules, reduced immunoreactivity of the organism in mouse macrophages [48], and an increase in the LAM:LM ratio was correlated with increased virulence via inhibition of phagosomal maturation [50].

Figure 1. M. tuberculosis cell envelope composition in replicating vs. NRP cultures.

Replicating M. tuberculosis produced peptidoglycan, arabinogalactan, and PDIM, and readily converted exported TMM into TDM. In NRP M. tuberculosis, levels of DATs/PATs and SLs were unaltered while PDIM levels declined. TMM and TDM levels also decreased as they were converted into free mycolic acids. NRP M. tuberculosis accumulated TAGs and WEs. The ratio of LAM to LM was increased in NRP. MmpL3, MmpL7 and MmpL11 are phosphorylated. MmpL3 and MmpL7 export TMM and PDIM, respectively, which are present in replicating and NRP M. tuberculosis. MmpL11 accumulates in late stationary conditions and contributes to NRP through transport of TAGs and WEs. Colored circles next to proteins indicate known STPKs. *, Differentially phosphorylated in stationary phase, (Verma et al., 2017). **, Differentially phosphorylated in hypoxia, (Prisic et al., 2010).

STPK regulation of cell envelope components

There is strong evidence that STPKs play a role in regulating M. tuberculosis cell envelope biology. Prisic et al. found a large number of phosphoproteins belonging to the cell wall/cell processes category. Subsequent studies confirmed some of their results and delved more deeply into the impact of phosphorylation for specific targets. Phosphorylation can impact M. tuberculosis protein function and/or the nature of protein-protein or protein-ligand interactions by either steric hindrance or inducing a conformational change in the protein. The impact of phosphorylation has been demonstrated for several M. tuberculosis proteins in cell envelope biogenesis. In the next section, we will summarize how STPK modification of specific cell envelope components alters their function.

Peptidoglycan (PG) and Arabinogalactan peptidoglycan (AGP)

PG provides structural integrity to the mycobacterial envelope and anchors mAGP. Post-translational modification by STPKs, in particular PknB, contributes to PG regulation. The extracellular domain of PknB consists of 4 PASTA domains, a structural element predicted to bind PG monomers [51]. PknB binds various muropeptides, having the highest affinity for muropeptides containing diaminopimelic acid (DAP) at the 3rd position and D-isoglutamine at the 2nd position [52]. Attachment to N-acetylmuramic acid (MurNAc) was essential for this interaction. PknB also interacts with mDAP-Lipid II via the PASTA 3/4 linker region [53]. Mutation of the ligand-binding residues S556, K557, N559, and Q560 delocalized PknB from the poles and septa, phenocopying a M. tuberculosis mutant lacking the entire PknB extracellular domain [52,53]. The tetra-mutant also resulted in hyper-auto and substrate phosphorylation [53], which led the authors to suggest that PG precursors act as negative regulators of PknB activity.

Synthesis of PG occurs via four steps: elongation of the muropeptide chain, modification to form Lipid II, translocation to the periplasm, and maturation and incorporation into the mAGP. Phosphoregulation by STPKs is implicated at each stage of synthesis, though muropeptide elongation appears heavily influenced by phosphoregulation. Phosphorylation by PknB of GlmU, which catalyzes the formation of uridine diphosphate-N-acetylglucosamine (UDP-GlcNAc), reduced its acetyltransferase activity in vitro [54]. MurA, which catalyzes that primary elongation reaction, is indirectly activated by PknB via interaction with a phosphorylated, cytosolic CwlM [13,55]. In addition, MurC and MurD are substrates of PknG and PknA, respectively, but the functional consequences of those modifications are unknown [56,57]. Phosphorylation of Wag31 by PknB or PknA upregulates the production of Lipid II in vitro and increases staining for nascent PG at the poles in vivo [58,59]. The flippase MviN, which translocates Lipid II to the periplasm, is also a PknB substrate and its phosphorylation results in the accumulation of nascent PG due to increased protein-protein interactions with FhaA, a conserved signal transducer [60]. Several of the MviN phosphopeptides appear to be differentially phosphorylated in logarithmic vs. stationary phase M. tuberculosis [10]. PknB-mediated phosphorylation of the cytoplasmic tail of PonA1, a PG assembly protein, reduces cellular elongation, potentially through negative regulation of its transglycosylation domain [61]. Phosphoproteomic data from PknA or PknB genetic depletion strains support the above in vitro data, demonstrating differential phosphorylation of these peptidoglycan synthesis enzymes across wildtype and depletion strains [11]. Based on the above results and the knowledge that PknB levels decrease in dormancy conditions, reduced phosphorylation of the above targets would decrease PG synthesis and levels of PG intermediates as the bacterium transitions into NRP. On the other hand, phosphorylation of Wag31, CwlM and MviN by PknB upon resuscitation would jumpstart PG synthesis.

AG synthesis may also be modulated by STPKs. RmlA, involved in rhamnose synthesis, and the rhamnosyl transferase Wbbl2 are phosphorylated by PknG in vitro [62]. Enzymatic assays further showed that phosphorylation of RmlA by PknG in vitro reduces its activity. There was no in vivo data on the phosphostatus of these enzymes in published datasets. Combined, these data suggest that phosphorylation of PG and AG biosynthetic enzymes may mediate growth slowing and remodeling in the transition into and out of NRP.

Mycolic acids (MAs)

The mycomembrane consists largely of MAs covalently attached as mAGP and as the trehalose glycoconjugates TMM and TDM [3]. The biosynthetic pathway of MAs is well understood and consists of de novo synthesis of fatty acids by FAS I, elongation of those by the FAS II complex, followed by a series of modifications before export by MmpL3 as TMM [63,64]. MA biosynthesis is controlled transcriptionally by MabR [65], and post-translationally by STPKs (Table 1, Figure 2). The FAS II elongation cycle is completed by the actions of six enzymes: MabA (FabG1), HadAB, HadBC, InhA, KasA, and KasB. Each of these enzymes are substrates of at least 4 STPKs and are catalytically impaired by phosphorylation in vitro [66–70]. In addition, four enzymes that precede (FabH, FadD) or follow (PcaA, FadD32) the FAS II cycle are also regulated by STPK phosphorylation in vitro [66,71–73]. Phosphorylation of FabH, MabA, HadB, KasA, InhA, PcaA, and FadD32 also occurs in replicating M. tuberculosis cells [11,14]. Differential phosphorylation status between wildtype and PknA or PknB depleted strains demonstrated that FabH and FadD32 are likely PknB substrates, HadB is a PknA substrate, and MabA contains distinct targets of both PknA and PknB [11]. The mycolyl transferase FbpA was also found to be phosphorylated in replicating M. tuberculosis cells [14]. Combined, these data suggest a biosynthetic pathway that is heavily targeted by phosphorylation, and in vitro biochemical data demonstrate a reduction in activity for phosphorylated enzymes (Table 1). However, there is little data regarding differential phosphorylation between replication and NRP or functional implications of phosphorylation in NRP.

Table 1:

Phosphoregulation of mycolic acid synthesis

Substrate Protein	STPKs	Phospho-acceptors	In vitro catalytic impact of phosphorylation	In vivo impact of phosphorylation	REF
FabD (Rv2243)	A,B,D,E,F,H	-	-	-	[66]
FabH (Rv0533)	A,F,D,H,E	T45*, S38*	T45D: 42% reduced condensation activity, 90% reduced decarboxylation activity	-	[11,70]
MabA (Rv1483)	B,A,D,E,L	T191 (Primary site), T21, T114** T188*, T2*	MabAT191D: 90% reduced activity, reduced NADPH affinity	Constitutive overexpression of MabAT191D impaired growth and inhibited MA synthesis	[11,14,70]
KasA (Rv2245)	A,B,D,E,F,H	Unspecified Thr T52*, S70*, S227*, S41*	Phospho-KasA: 3–4 fold reduction in Vmax, Km unchanged	KasA phosphorylation remained constant during growth and a tri-phosphorylated KasA appeared in stationary phase	[11,14,66]
KasB (Rv2246)	A,B,D,E,F,H	T334,T336	Phospho-KasB: increased Vmax and Km for malonyl-AcpM	ΔkasB and KasBT334D,T336D strains have 3–4C shorter MAs that lacked trans-cyclopropanation; ΔkasB and KasBT334D,T336D strains are attenuated in mouse model,	[66,80]
HadAB/HadBC (Rv0635–0637)	A,B,D,E,F,H	HadA:T79,T118, T138 HadB:S7,T19**,T88, T119 HadC:S150,S159	Phospho-HadAB: 70% reduced activity Phospho-HadBC: 90% reduced activity	Authors reported that HadBC was more phosphorylated in stationary phase	[11,14,69]
InhA (Rv1484)	A,B,H,F/A,B,E,L	T79,T253,T254,T266/ T266*,S73*, S143*	Phospho-InhA: 5-fold reduced activity, reduced affinity for NADH InhAT266E: 70% reduced activity, reduced affinity for NADH	Msm: Overexpression of InhAT266E impaired growth rate, expression of InhAT266E did not complement ΔinhA mutant MA synthesis defect. InhAT266E is lethal mutation in M. bovis BCG and M. tuberculosis	[14,68,81]
PcaA (Rv0470c)	F,D,E,H	T168,T183*, T185*	Phospho-PcaA: 50–60% reduced activity	M. bovis BCG ΔpcaA has a small colony phenotype, an altered MA profile and reduced replication in macrophages. Expression of WT and PcaAT168A alleles complement, but PcaAT168D does not.	[11,72]
FadD32 (Rv3801c)	B,A,D,F	T281**, T552***	Phospho-FadD32: 40% reduced activity		[11,14,73]
CmaA2 (Rv0503c)	A	T7*, T8*, T198*			[11]
MmaA3 (Rv0643c)	A	S11*, T118*, S180*, T206*			[11,14]

Underlining indicates if a phosphosite is targeted by a specific protein kinase

Asterisks, indicate identification of a phosphosite from phosphoproteomics of M. tuberculosis cells. Number of astrisks indicates how many studies identified the phospho-peptide.

Figure 2. Mycolic acid biosynthesis is controlled by STPK phosphorylation.

Phosphoproteomic data from PknA and PknB studies indicates that several mycolic acid synthesis proteins are substrates of PknA and/or PknB. In vitro and biochemical data suggests that all enzymes represented above are down regulated by phosphorylation and may be substrates of multiple protein kinases. Closed circles indicate if PknA or PknB phosphorylate each enzyme based on phosphoproteomic data of whole cell lysates from WT and PknA or PknB depletion mutants (Carette et al., 2018; Zeng et al., 2020), and open circles indicate which kinases phosphorylate each enzyme in vitro. * indicate that an enzyme has not been identified as a phosphoprotein in cells.

PDIM and PGL

As noted earlier, PDIM levels decreased under hypoxic conditions [46], suggesting its biosynthesis and/or transport is reduced in NRP. Phosphorylation may regulate the partially overlapping PDIM and phenolic glycolipid (PGL) biosynthesis pathways: A number of PDIM and PGL synthesis and transport proteins are targets of STPKs [6,14]. Notably, the mycocerosic acid synthesis enzyme Mas was phosphorylated at the C-terminal S2111 residue only under hypoxic conditions [6]. This study also identified PGL synthesis enzymes Pks1 and FadD22 as each having single, ambiguous phosphosites under logarithmic and peroxide stress conditions, respectively. Phosphorylation of the PDIM biosynthetic enzymes FadD28 (T556), PpsE (Y1429, S882), PpsC (Y195), and ABC transporter protein DrrA (T319, S321, T323) was reported [14]. Phosphorylation at DrrA at T319 and S321 were also observed by Zeng et al. [11]. Interestingly, none of these proteins were differentially phosphorylated across wildtype and PknA or PknB inhibited/depleted conditions, suggesting that they may be targets of other STPKs [11,14]. Based upon phosphorylation of MmpL7 by PknD (discussed below), we speculate that PknD is a likely suspect. Further investigation is needed to determine whether phosphorylation regulates the PDIM and/or PGL synthesis pathways and how this impacts lipid composition and virulence.

Cell envelope exporters and flippases

Phosphorylation may also modulate cell envelope composition by regulating lipid transporters such as MmpL proteins. The majority of MmpL family members are predicted to be >100-kDa and consist of 11–12 transmembrane (TM) domains and two periplasmic loop domains. MmpL3 and 11 have large C-terminal (CT) cytoplasmic domain, whereas MmpL7 possesses a large Domain 2 (D2) located between TM domains TM7 and TM8 (Figure 3). Based on TMHMM prediction algorithms, MmpL7 D2 was initially predicted to localize in the periplasm. To gain better insight, we modeled MmpL7 on our recently solved crystal structure of M. smegmatis MmpL3 using Phyre2 [74,75]. Superimposition of the two structures gives a root-mean-square-deviation (r.m.s.d.) of 0.886 Å. This structural modelling approach gave a high-confidence prediction that MmpL7 D2 is cytoplasmic. These intracellular domains of the MmpL proteins are likely regions of interaction with cytoplasmic proteins, and they are also the domains of MmpL3_TB, MmpL7_TB, and MmpL11_TB that were phosphorylated (Table 2) [6,76].

Figure 3. Membrane topology of MmpL3, MmpL11, and MmpL7.

Table 2.

Phosphorylated MmpL3, MmpL11, and MmpL7 residues identified in H37Rv

MmpL3 Peptide (#, identified residue)	AA	Differentially phosphorylated in stationary phaseb
AAGDPRPPHDPT#HPLAESPRPAR b, e	T789
SSPASSPELTPALEATAAPAAPS#GASTTR b	S823
MQIGSST#EPPTTR b	T836	yes
MQIGSSTEPPT#TR a, e	T840
SVQSPAS#TPPPTPTPPSAPSAGQTR e	S855
SVQSPAST#PPPTPTPPSAPSAGQTR b	T856	yes
SVQSPASTPPPT#PTPPSAPSAGQTR a	T860
SVQSPASTPPPTPT#PPSAPSAGQTR b	T862	yes
SVQSPASTPPPTPTPPS#APSAGQTR e	S865
SVQSPASTPPPTPTPPSAPS#AGQTR e	S868
SVQSPASTPPPTPTPPSAPSAGQT#R a, e	T872
SVQSPASTPPPTPTPPS#APSAGQTR b	S865	yes
STDAAGDPAEPT#AALPIIR a, b, e	T893	yes
STDAAGDPAEPT#AALPIIRS#DGDDSEAATEQLNAR b, f	T893, S901
STDAAGDPAEPT#AALPIIRSDGDDSEAAT#EQLNAR b	T893, T910
SDGDDS#EAATQLNAR b	S906	yes
SDGDDSEAAT#EQLNAR a, b, e, f	T910	yes
MmpL7 Peptide	Start AA
KVESAAWPAGVPWTDASLSSAAGRLADQLGQQAGc	467
ISAISTQTMSALSSAPRMVAQMR c	611
HVRESMFSSDGTATR c	693
MmpL11 Peptide	AA
VAGAAQVDVGGPT#ALIKDFDDR d	T514
LAPDAICVTDPLAFT#GCGCDGKALDQVQLAYR a, d	T765
RLAVALDALQT#TTWECGGVQTHR d	T821
LAVALDALQT#T#TWECGGVQTHR d	T821, T823
RCLS#VAVAMLEEAR d	S956

^a.Prisic et al, 2010. Only unambiguous residues shown.

^b.Verma et al., 2017. Only unambiguous residues shown.

^c.Perez et al, 2006. This study did not identify exact residue phosphorylated.

^d.Melly et al., 2020.

We have a longstanding interest in MmpL11. Prisic et al. demonstrated that MmpL11 was phosphorylated at residue T765 when grown planktonically in medium containing acetate as the sole carbon source [6]. We recently showed that MmpL11 function is modulated by phosphorylation at T765, and found that the ability of MmpL11_TB alleles to complement biofilm and lipid phenotypes of M. smegmatis and M. tuberculosis depended on the phosphorylation state of T765 [77]. Furthermore, when expressed under control of the native MmpL11_TB promoter, phospho-mimetic MmpL11_TB(T765E) protein levels fail to accumulate over time, suggesting that non-functional MmpL11_TB is unstable.

Since phosphorylation of T765 impacted MmpL11_TB function in biofilms, we examined the in vivo phosphorylation status of MmpL11_TB purified from M. smegmatis mmpL11 mutant biofilms. T765 was not phosphorylated in biofilm cultures, consistent with the observation that phosphorylation abrogates MmpL11 activity while de-phosphorylation enables MmpL11-mediated lipid transport and biofilm formation. We also identified six additional MmpL11 phosphopeptides that were not previously known (Table 2). Our data showed that not all phosphorylation sites will be important for MmpL proteins in the conditions tested. For instance, MmpL11_TB T771 was phosphorylated in our analysis (Table 2). We tested the relative importance of MmpL11_TBT771 in biofilm formation and found that the phospho-status of T771 did not impact biofilm formation or cell envelope lipid profiles. These results emphasize that the role of phosphorylation must be determined experimentally.

While we provided evidence that the phospho-status of T765 regulates MmpL11_TB-mediated lipid transport and protein levels, we have not yet determined the underlying mechanism. We predict that phosphorylation may impact protein stability, protein function, and the proteins with which MmpLs interact. Given that the CT of MmpL11 is dispensable for lipid transport [77], we hypothesize that phosphorylation at this site decreases MmpL11 protein accumulation through the activation of protein degradation mechanisms. We also note that the CT domain of MmpL11 (and MmpL3) is intrinsically unstructured [74], and phosphorylation of such domains may stabilize secondary or tertiary structure and thereby influence function or interacting partners [78]. Current investigations focus on determining the kinase or kinases responsible for MmpL11_TB phosphorylation and deciphering the different metabolic cues that trigger phosphorylation or dephosphorylation of MmpL11. PknA and PknB are likely suspects since they have a demonstrated role in mediating the response to non-replicating persistence and influence cell envelope biology [15]. Unfortunately, MmpL11 was not present in either the Carette et al. or the Zeng et al. datasets.

Five regions within the CT of the TMM exporter MmpL3 were identified as unambiguous phosphosites [6]. Three of the five regions identified were phosphorylated in all conditions tested, including in logarithmic and stationary cultures and under reactive oxygen and nitrogen stress, hypoxia, and in different carbon sources. The remaining two regions lacked phosphorylation in logarithmic stage cultures or when exposed to reactive oxygen. Abundance of several phosphosites was reduced in stationary phase cultures (Table 2) [10]. Whether the phospho-status of these residues is important for the transition between replication and NRP remains to be seen.

There is robust data showing that MmpL3 is phosphorylated by PknA and PknB. Carette and colleagues used a multipronged approach combining small-molecule inhibitors of PknA and PknB with multiple -omic analyses to identify phosphorylation targets and downstream effects of PknA and PknB [14]. This study demonstrated that the MmpL3 T893 is phosphorylated by PknA or PknB in M. tuberculosis. This result was in agreement with another recent study using conditional expression of PknA and PknB in M. tuberculosis that found decreased phosphorylation of MmpL3 T893 and T910 upon knockdown of PknA [11]. Furthermore, they showed that PknA depletion increased TMM levels. The authors attributed this result in part to decreased expression of the fpbB gene that encodes the mycolyltransferase responsible for generating TDM from TMM via phosphorylation and modulation of the two-component regulator MtrA. The authors acknowledged that it is also possible that MmpL3 function itself is impacted, but this was not explored experimentally.

MmpL3 was also identified as a target of PknB_TB in M. smegmatis via overexpression of PknB [79]. Phosphoproteomic analysis showed that PknB_TB phosphorylated a number of enzymes involved in MA biosynthesis as well as M. smegmatis MmpL3, which was phosphorylated at T984. Consistent with this result and our model that phosphorylation of MmpL proteins modulates their function, lipid analysis of the M. smegmatis PknB_TB overexpression strain demonstrated that MA biosynthesis was impacted, with the accumulation of TMM and reduction of TDM levels. Work by Le et al. may support the use of this non-pathogenic model system to identify residues whose phosphorylation modulates MmpL3 function.

MmpL7_TB was phosphorylated in replicating wild-type M. tuberculosis compared to the pknD mutant, suggesting that PknD phosphorylates MmpL7 [76]. Three of four phosphopeptides were located in the D2 domain and were not identified by Prisic et al. One peptide of MmpL7 was observed by Verma et al, but it was not phosphorylated [10]. Unfortunately, Perez et al. did not test if MmpL7_TB is phosphorylated in any other growth condition, if phosphorylation altered the function of MmpL7_TB, or if MmpL7_TB was phosphorylated by PknD or other kinases. Therefore, the impact of phosphorylation remains unknown.

To summarize, several studies obtained a snapshot of the global mycobacterial phosphoproteome and showed MmpL proteins are phosphorylated. That being said, data on the phosphorylation status of MmpL proteins are limited, likely due to difficulties solubilizing and extracting these large membrane proteins. These studies provide a starting point, but do not provide mechanistic insights on the conditions in which the MmpL proteins are differentially phosphorylated, or the effect that phosphorylation has on the substrate protein.

Conclusions and future directions

There are major alterations in the M. tuberculosis cell envelope composition as the bacterium transitions into NRP. Phosphorylation by STPKs plays a role in this remodeling by modulating some lipid biosynthetic enzymes and likely through altering exporter function. As described above, phosphorylation can impact M. tuberculosis protein function and/or the nature of protein-protein or protein-ligand interactions. Phosphorylation of lipid biosynthetic and metabolic enzymes may therefore be an important mechanism for sensitive control of essential pathways to ensure pathogen survival in the various environments encountered in vivo. As described above, the impact of phosphorylation was demonstrated for several M. tuberculosis proteins in cell envelope biogenesis. Less is known about the impact of phosphorylation on lipid exporters. Based on our recent work, we propose that phosphorylation of these proteins will impact their function and/or levels in the bacterium thereby facilitating cell envelope biogenesis and remodeling. We would also note that a global perspective of the phosphoproteome in replicating and NRP M. tuberculosis and its role in cell envelope remodeling is lacking. We suspect that global interrogation of the phosphoproteome in replicating M. tuberculosis and in bacteria cultured under the in vitro models of NRP discussed herein will reveal additional points where STPKs modulate cell envelope composition.