The stringent response and Mycobacterium tuberculosis pathogenesis

Abstract

During infection, the host restrains Mycobacterium tuberculosis (Mtb) from proliferating by imposing an arsenal of stresses. Despite this onslaught of attacks, Mtb is able to persist for the lifetime of the host, indicating that this pathogen has substantial molecular mechanisms to resist host-inflicted damage. The stringent response is a conserved global stress response in bacteria that involves the production of the hyperphosphorylated guanine nucleotides ppGpp and pppGpp (collectively called (p)ppGpp). (p)ppGpp then regulates a number of cellular processes to adjust the physiology of the bacteria to promote survival in different environments. Survival in the presence of host-generated stresses is an essential quality of successful pathogens, and the stringent response is critical for the intracellular survival of a number of pathogenic bacteria. In addition, the stringent response has been linked to virulence gene expression, persistence, latency and drug tolerance. In Mtb, (p)ppGpp synthesis is required for survival in low nutrient conditions, long term culture and during chronic infection in animal models, all indicative of a strict requirement for (p)ppGpp during exposure to stresses associated with infection. In this review we discuss (p)ppGpp metabolism and how this functions as a critical regulator of Mtb virulence.

In this review we discuss (p)ppGpp metabolism and how this functions as a critical regulator of Mycobacterium tuberculosis pathogenesis.

(p)ppGpp SYNTHESIS IN BACTERIA

(p)ppGpp (also referred to as magic spot due to initial identification as a spot in chromotagraphy experiments (Cashel 1969; Cashel and Gallant 1969)) is the effector molecule of the stringent response. (p)ppGpp metabolism in bacteria is controlled by Rel/SpoT homolog proteins (RSHs), named for their sequence similarity to RelA and SpoT enzymes in Escherichia coli (Atkinson, Tenson and Hauryliuk 2011). RelA is a monofunctional (p)ppGpp synthetase while SpoT is a bifunctional enzyme with (p)ppGpp hydrolysis activity and weak (p)ppGpp synthetase activity. Despite being the basis for the nomenclature of RSH proteins, phylogenetic analyses indicate that the presence of two functionally divergent RSH homologs (RelA and SpoT) is unique to the γ- and β- proteobacteria lineages (Atkinson, Tenson and Hauryliuk 2011). Most other bacteria encode a single bifunctional RSH enzyme that contains both a (p)ppGpp synthetase and hydrolase domain as well as several regulatory domains. The common nomenclature for multi-domain RSH proteins is long RSHs. Orthologs of long RSHs are widely-distributed throughout bacteria and are believed to originate from a common ancestral long RSH (Atkinson, Tenson and Hauryliuk 2011). The E. coli homologs RelA and SpoT likely arose from a lineage-specific duplication.

Mtb encodes a single bifunctional long RSH named Rel_Mtb (Avarbock et al.1999) (Fig. 1) that is conserved in all Mycobacterium species and has been shown to complement an E. coli relA mutant for growth on minimal media (Ojha, Mukherjee and Chatterji 2000), confirming its ability to promote the stringent response. The rel_Mtb locus is syntenous with the rel locus in Bacillus subtilis, encoding apt (adenine phosphoribosyltransferase) upstream and cypH (cyclophilin) downstream (Avarbock et al.1999). Deletion of rel_Mtb(Δrel_Mtb) results in a (p)ppGpp null mutant ((p)ppGpp⁰) (Primm et al.2000), suggesting that Rel_Mtb is the only functional (p)ppGpp synthetase in Mtb. (p)ppGpp synthesis by Rel_Mtb is necessary for chronic Mtb infection in both mice and guinea pigs (Dahl et al.2003; Karakousis et al.2004; Klinkenberg et al.2010; Weiss and Stallings 2013), which is when the immune system is activated to impose growth restrictive stresses on the bacteria. In contrast, the Δrel_Mtb mutant is not attenuated for growth in THP-1 macrophages in cell culture (Primm et al.2000), indicating that the stringent response is essential for Mtb survival during stress conditions specific to the chronic phase of in vivo infection.

Figure 1.

Diagram of the full-length Rel_Mtb protein illustrating functional domains. The N-terminal domain (NTD) contains the catalytic domains and the C-terminal domain (CTD) contains the regulatory domains. The CTD contains two structural domains: the TGS domain (named for its initial identification in ThrRS, GTPase and SpoT) and the ACT domain (named for its initial identification in aspartate kinase, chorismite mutase and TyrA). The numbers designate the amino acid positions that form the borders of each domain.

Streptomyces, Frankia and Salinispora species encode a second long RSH homolog named RshA, which is believed to have arisen from an Actinobacteria-specific duplication but has subsequently been lost in Mycobacterium species (Sun, Hesketh and Bibb 2001; Atkinson, Tenson and Hauryliuk 2011). In addition to long RSHs, two other classes of RSH proteins have been discovered, the Small Alarmone Synthetases (SASs) and Small Alarmone Hydrolases (SAHs). In contrast to the long RSHs, SASs and SAHs are monofunctional, single-domain enzymes. Actinobacteria SASs belong to their own monophyletic clade (actRel) and are orthologous to the well-studied B. subtilis SAS proteins RelP (YjbM) and RelQ (YwaC) (Atkinson, Tenson and Hauryliuk 2011). Mtb encodes a single SAS (Rv1366), but the enzyme is believed to be non-functional (Bag et al.2014) and deletion of rv1366 in Mtb has no phenotypes in culture or during infection of mice (Weiss and Stallings 2013). In other bacteria, SAS activity is toxic in the absence of a functional (p)ppGpp hydrolase (Lemos et al.2007; Nanamiya et al.2008; Abranches et al.2009; Das et al.2009; Natori et al.2009), which is not observed in Mtb, further supporting that Rv1366 is not a functional (p)ppGpp synthetase (Weiss and Stallings 2013). The only other SAS that has been studied in mycobacteria is MS_RHII-RSD in Mycobacterium smegmatis (Msm), a non-pathogenic model organism for Mtb. Unlike Rv1366, MS_RHII-RSD has been shown to be a functional pppGpp synthetase and encodes an amino-terminal RNase HII domain (Murdeshwar and Chatterji 2012) that can hydrolyze R-loops (Krishnan et al.2016) that may form as a result of collisions between the replication and transcription machinery, indicating that MS_RHII-RSD has evolved functions distinct from Rv1366.

(p)ppGpp METABOLISM BY Rel_Mtb

Rel_Mtb is comprised of two catalytic domains, a (p)ppGpp hydrolysis domain (1-181 amino acids (aa)) and a (p)ppGpp synthetase domain (87-394 aa), that include an overlapping three helix bundle (87-181), as well as a regulatory C-terminal domain (CTD, 395-738 aa) (Fig. 1) (Avarbock et al.2005; Singal et al.2017). The (p)ppGpp synthetase domain transfers the 5΄-β,γ-pyrophosphate from ATP to the 3΄-OH of GDP or GTP to synthesize ppGpp and pppGpp, respectively (Avarbock et al.1999). Crystal structures of the N-terminus of Rel_Mtb (1-394 aa) and the N-terminus of a homolog from Streptococcus dysgalactiae subspecies equisimilis (Rel_Seq, 1-385 aa included in the crystal structure) have provided insight to the functional features of Rel catalytic domains (Hogg et al.2004; Singal et al.2017). The (p)ppGpp synthetase domain is comprised of five β-sheets surrounded by five α-helices (Singal et al.2017). The (p)ppGpp synthetase domain was determined to be structurally homologous to DNA polymerase β (pol β) and the D190 and D256 residues in pol β that are essential for coordinating the Mg²⁺ cofactor correspond to D265 and E325 in Rel_Mtb (Hogg et al.2004). Both D265 and E325 of Rel_Mtb are required for (p)ppGpp synthesis in vitro and Rel_Mtb rescue of an E. coli ΔrelAΔspoT ((p)ppGpp⁰) strain growth on minimal media (Hogg et al.2004; Bag et al.2014). Rel_Seq co-crystallized with GDP in a ligand-binding pocket partially formed by 176-185 aa within the synthetase domain. The conservation of the majority of the residues comprising the GDP binding pockets between Rel_Seq and Rel_Mtb suggests that Rel_Mtb and Rel_Seq recognize and bind to GDP in a similar fashion. The GDP pocket also includes Rel_Seq Y308 (Rel_Mtb Y309), which face-to-face stacks with GDP and is required for Rel_seq to rescue growth of an E. coli (p)ppGpp⁰ strain on minimal media (Hogg et al.2004; Bag et al.2014). The Rel_Mtb ATP binding pocket likely involves the positively charged residues R242, K244 and K252 coordinating the β and γ phosphates of ATP at a site proximal to the GDP binding pocket and the Mg²⁺ binding site (Hogg et al.2004; Singal et al.2017). The importance of the basic patch created by R242, K244 and K252 is supported by experiments showing that Rel_Mtb R242H is unable to rescue growth of an E. coli (p)ppGpp⁰ strain on minimal media (Bag et al.2014). In addition, Rel_Mtb G241 and H344, which were initially identified in E. coli RelA as being important for ATP binding (Gropp et al.2001), are required for pppGpp synthesis in vitro (Avarbock et al.2005). Additionally, H344 is required for Rel_Mtb to rescue growth of an E. coli (p)ppGpp⁰ strain on minimal media (Bag et al.2014) and mutation of Rel_Mtb H344 attenuates Mtb in mice (Weiss and Stallings 2013).

In addition to being able to synthesize (p)ppGpp, Rel_Mtb also hydrolyzes (p)ppGpp into pyrophosphate (PP_i) and GDP or GTP via an N-terminal catalytic HD superfamily domain (Avarbock et al.2005) (Fig. 1). HD superfamily members are phosphohydrolases with conserved histidine and aspartate residues that are involved in the coordination of divalent cations, which is essential for their activity (Aravind and Koonin 1998). The Rel_Mtb hydrolase domain is comprised of 11 α-helices, which includes a (p)ppGpp binding pocket spanning residues 41-53 located in the region connecting the second and third α-helices (α2 and α3). The HD motif comprised of the conserved histidine (H80) and aspartate (D81) residues is located in a β-turn that connects α4 and α5 (Singal et al.2017). Alanine substitution of H80 or D81 in Rel_Mtb abolishes hydrolase activity in vitro without affecting pppGpp synthesis (Avarbock et al.2005).

The Rel_Mtb catalytic domains have strict cation co-factor requirements where (p)ppGpp synthesis requires Mg²⁺ or Mn²⁺ and Rel_Mtb (p)ppGpp hydrolysis requires Mn²⁺. The Mg²⁺ and Mn²⁺ concentrations required for optimal (p)ppGpp synthesis is equal to the combined concentration of the substrates, ATP and GTP. Rel_Mtb synthetase activity is inhibited by concentrations of Mg²⁺ (or Mn²⁺) that exceed the GTP and ATP substrate concentrations due to an RXKD motif within the synthetase domain that is highly conserved among bifunctional Rel enzymes (Avarbock, Avarbock and Rubin 2000; Sajish et al., 2007, 2009). The model for how the RXKD motif sensitizes the Rel_Mtb synthetase catalytic activity to Mg²⁺ concentrations that exceed ATP and GTP substrate concentrations is that the negatively charged phosphates of GTP can bind to either the positive charges of Rel_Mtb RXKD motif (R and D) or positively charged Mg²⁺ ions. Mg²⁺ bound GTP is unable to bind the positive charges of RXKD in a way that is required for (p)ppGpp synthesis. Thus, Mg²⁺ ions compete with the Rel_Mtb RXKD motif for GTP binding. In contrast, E. coli RelA has an EXDD motif instead of the RXKD motif and is not sensitive to high concentrations of Mg²⁺. Instead, due to the presence of two negatively charged residues (E and D), the E. coli EXDD motif requires Mg²⁺ to bind GTP, allowing the E. coli RelA synthetase domain to function when Mg²⁺ concentrations are higher than the combined ATP and GTP concentration (Sajish et al., 2007, 2009).

REGULATION OF (p)ppGpp SYNTHESIS BY Rel_Mtb

Rel_Mtb is constitutively expressed at basal levels, likely via an almost consensus -10 promoter element upstream of the rel_Mtb gene that would be recognized by the housekeeping σ-factor σ^A (Jain et al.2005). However, in the absence of stress or stimuli, the Rel_Mtb synthetase activity is repressed by the presence of the CTD (Avarbock et al.1999; Jain et al.2006) and only a low level of (p)ppGpp is produced (Stallings et al.2009; Weiss and Stallings 2013). Although the level of (p)ppGpp produced in unstressed conditions is low, this low level production is still required for WT growth rates and a Δrel_Mtb mutant is attenuated for growth in unstressed, nutrient-rich media under aerobic conditions (Primm et al.2000; Weiss and Stallings 2013), suggesting that the low levels of (p)ppGpp produced in unstressed conditions still provide a growth advantage to Mtb. In addition, (p)ppGpp hydrolysis is essential for preventing toxic (p)ppGpp accumulation in Mtb in unstressed culture conditions, as well as during both acute and chronic stages of infection (Weiss and Stallings 2013). These observations demonstrate that Rel_Mtb constitutively produces (p)ppGpp and may act to maintain homeostasis in all growth conditions. This is also similar to observations in other bacteria where (p)ppGpp hydrolysis is essential for viability in unstressed conditions due to the presence of a constitutive level of (p)ppGpp synthesis (Lemos et al.2007; Nanamiya et al.2008; Das et al.2009).

The induction of (p)ppGpp synthesis by Rel homologs is best characterized during amino acid starvation where Rel is associated with ribosomes when an uncharged tRNA enters the A site of the ribosome and stimulates Rel-mediated (p)ppGpp production (Haseltine and Block 1973; Avarbock, Avarbock and Rubin 2000). Accordingly, Mtb accumulates (p)ppGpp in a Rel_Mtb-dependent manner during nutrient starvation in liquid culture (Primm et al.2000; Stallings et al.2009). In agreement with stringent response mutants in other bacteria, deletion of rel_Mtb reduces the viability of Mtb in nutrient-poor conditions (Primm et al.2000), supporting a role for the Mtb stringent response in responding to starvation. In addition to during starvation, (p)ppGpp accumulates in Mtb in response to hypoxia and oxidative stress (Primm et al.2000; Stallings et al.2009), and it is not known whether this is because these stresses cause amino acid starvation that triggers Rel_Mtb (p)ppGpp synthetase activity or if there are other ways to induce Rel_Mtb-mediated (p)ppGpp synthesis.

The induction of (p)ppGpp synthesis during amino acid starvation requires the Rel_Mtb CTD and can be recapitulated in vitro by adding the Rel activating complex (RAC) consisting of ribosomes, uncharged tRNAs, and mRNA (Avarbock, Avarbock and Rubin 2000; Avarbock et al.2005; Jain et al.2006). The Rel_Mtb C-terminal domain (CTD) contains two structural domains: the TGS domain (400-459 aa, named for its initial identification in ThrRS (threonyl tRNA synthetase), GTPase (Obg family of GTPases) and SpoT) and the ACT domain (657-733 aa, named for its initial identification in aspartate kinase, chorismite mutase and TyrA) (Wolf et al.1999; Chipman and Shaanan 2001) (Fig. 1). Although a crystal structure for the Rel_Mtb CTD has not yet been solved, multiple cryo-EM structures of E. coli RelA in complex with a ribosome with an uncharged tRNA (Phe) occupying the A site have provided a structure-based model for the function of TGS and ACT domains (Arenz et al.2016; Brown et al.2016; Loveland et al.2016). These structures suggest that the TGS domain acts as a sensor to distinguish charged versus uncharged tRNAs that enter the A-site of the ribosome. The sensor function of the TGS domain is accomplished by the ability of the TGS domain to interact with the 3’CCA of the A-site tRNA. If an aminoacylated tRNA occupies the A-site of the ribosome, the amino acid attached to the 3’CCA will sterically occlude the interaction of the 3’CCA with the TGS domain. However, when a deacylated tRNA occupies the A-site of the ribosome, the TGS domain is able to interact with the 3’CCA, resulting in structural changes within the RelA synthetase domain that activates (p)ppGpp synthesis. The ACT domain serves an anchor function, allowing RelA to bind the ribosome regardless of whether the A-site is occupied by a charged or uncharged tRNA. The ACT domain occupies a cavity between the A and P sites of the ribosome and is held in place through several interactions with the A-site finger (helix 38) of the 23S ribosomal RNA (rRNA), the L16 ribosomal protein and the P-site tRNA. The cryo-EM structures also revealed a domain between the TGS and ACT domains of RelA that resembles a zinc-finger domain. This previously uncharacterized domain provides a second binding interface between RelA and the ribosome, and is expected to be conserved in mycobacteria. Loveland et al. (2016) termed this new domain in RelA the Ribosomal-InterSubunit (RIS) domain because it bridges the 30S and 50S ribosomal subunits through interactions with the A-site finger and the hydrophobic patch on the 30S S19 ribosomal protein.

Induction of (p)ppGpp synthesis is accompanied by a repression of (p)ppGpp hydrolysis. The Rel_Seq crystal structure supports a model where Rel does not simultaneously catalyze (p)ppGpp synthesis and (p)ppGpp hydrolysis (Hogg et al.2004). In this study, two conformations of the Rel_Seq monomer (1-385) were observed, a hydrolase-OFF/synthetase-ON conformation and a hydrolase-ON/synthetase-OFF conformation. Transition between the two conformations is dependent on ligand binding to the respective active site and indicates that at any given time only one enzymatic function is active, thus preventing simultaneous synthesis and hydrolysis of (p)ppGpp (Hogg et al.2004). In support of this model, Avarbock et al. have shown that Rel_Mtb hydrolase activity is inhibited in the presence of synthetase substrates ATP and GTP (Avarbock, Avarbock and Rubin 2000).

In what appears to be a negative feedback loop, pppGpp has been shown to bind to the CTD in the region that separates the TGS and ACT of the Rel protein from Msm (Syal et al.2015). Binding of pppGpp to the Rel_Msm CTD represses pppGpp synthesis while increasing pppGpp hydrolysis. As such, Rel_Msm-mediated pppGpp synthesis is decreased at saturating pppGpp concentrations compared to synthesis at non-saturating pppGpp concentrations. Rel_Msm point mutants with lower affinity for pppGpp (F533A, R615A, F620A and R633A) lose this effect. pppGpp binding to Rel_Msm decreases protein compaction and causes Rel_Msm to become less structured, suggesting that conformational changes within Rel_Msm may lead to repression of the synthetase activity (Jain, Saleem-Batcha and Chatterji 2007).

Studies in vitro have suggested that Rel_Mtb could also be regulated through oligomerization (Avarbock et al.2005; Jain et al.2006; Singal et al.2017). The full-length Rel_Mtb has been shown to form trimers in gel filtration experiments, which is facilitated by the CTD, although this domain is not required for oligomerization (Avarbock et al.2005). The Rel_Mtb NTD (amino acids 1-394) alone formed dimers in small angle X-ray scattering analysis and X-ray crystallography experiments (Singal et al.2017). In both cases, oligomerization was shown to inhibit pppGpp synthesis (Avarbock et al.2005; Jain et al.2006). These in vitro observations have led to some speculation that this could be another way that Rel_Mtb activity is regulated, however, it is still unknown if oligomerization occurs in vivo.

REGULATION OF TRANSCRIPTION BY (p)ppGpp

Once (p)ppGpp is synthesized, it has been shown to bind multiple targets in bacteria to modify their function and promote changes in physiology that allow for survival during stress. The most well-characterized outcome of the stringent response in bacteria is a downregulation of rRNA and ribosomal proteins accompanied by an upregulation of amino acid biosynthesis genes to restore amino acid levels during starvation (Eymann et al.2002; Traxler et al.2008). Hallmarks of this canonical stringent response appear to be conserved in Mtb. In a microarray experiment comparing expression profiles of a Δrel_Mtb strain to WT Mtb starved in Tris-buffered saline with Tween-80 (TBS-T), expression of 54 of the 58 ribosomal proteins was downregulated during starvation in WT Mtb but not in the Δrel_Mtb mutant (Dahl et al.2003). In addition, quantitative reverse transcription-PCR (qRT-PCR) experiments showed that the Δrel_Mtb strain was unable to downregulate rRNA levels during starvation (Stallings et al.2009). These data support the conservation of the canonical stringent response involving a Rel-dependent repression of genes encoding the translation machinery. However, WT Mtb starved in TBS-T does not induce expression of amino acid biosynthesis genes (Betts et al.2002; Dahl et al.2003). One potential explanation for Mtb failing to induce expression of amino acid biosynthesis genes is that the studies with Mtb used a general carbon starvation condition by incubating Mtb in TBS-T, whereas studies in E. coli and B. subtilis have used amino acid analogs to directly interfere with tRNA charging. Therefore, it is possible that Mtb would upregulate amino acid biosynthesis genes in response to specifically interfering with tRNA charging, but this has yet to be tested. Alternatively, Zhang and Rubin (2013) attribute this observation to the relatively unique amino acid prototrophy of Mtb where the bacteria constitutively synthesizes its own amino acids and these pathways may have evolved to become independent of stress signaling.

The mechanisms by which (p)ppGpp alters gene expression during starvation have been most well-characterized in the gram-negative γ-proteobacteria E. coli and the gram-positive firmicute B. subtilis. Numerous differences in the mechanisms used by (p)ppGpp in these two organisms has established two divergent paradigms of the stringent response: the E. coli paradigm, which is utilized by most proteobacteria, and the B. subtilis paradigm, which occurs in most other bacteria species (Boutte and Crosson 2013; Hauryliuk et al.2015; Liu et al.2015). One of the primary differences between the two stringent response paradigms is the direct binding target of (p)ppGpp. In E. coli, (p)ppGpp directly binds to RNA polymerase (RNAP) and changes gene expression by destabilizing RNAP-promoter open complexes (RP_o) that form during transcription initiation (Barker, Gaal and Gourse 2001; Ross et al.2013; Zuo, Wang and Steitz 2013). Two distinct ppGpp binding sites have been identified on the E. coli RNAP. The first ppGpp binding site (Site 1) is a pocket in between the β’ and ω subunits of RNAP (Ross et al.2013; Zuo, Wang and Steitz 2013). ppGpp binding to Site 1 is thought to allosterically prevent the RNAP from making stable contacts with the promoter DNA, thus destabilizing RP_o (Ross et al.2013; Zuo, Wang and Steitz 2013). A second ppGpp binding site (Site 2) is introduced by the binding of the DksA protein to RNAP (Ross et al.2016). DksA binds to the β’ subunit of RNAP and inserts its coiled-coil domain into the RNAP secondary channel, creating (p)ppGpp binding Site 2, located at the interface between the RNAP β’ subunit and DksA. Recent crystal structures revealed that ppGpp binding stabilizes the DksA-RNAP interaction by relieving the mechanical stress in RNAP caused by DksA binding (Molodtsov et al.2018). In addition, ppGpp binding induces a conformational change in DksA, allowing it to insert its coiled-coil domain further into the RNAP secondary channel to potentially interfere with RNAP catalytic activity (Molodtsov et al.2018). Binding of DksA is necessary for the full effects of ppGpp on RP_o stability in vitro (Paul et al.2004; Paul, Berkmen and Gourse 2005; Lennon et al.2012; Molodtsov et al.2018). (p)ppGpp and DksA are thought to regulate transcription initiation by destabilizing the transition states between the RNAP-promoter closed complex (RP_c) and RP_o complex, thus decreasing the activation energy of this transition (Paul, Berkmen and Gourse 2005; Rutherford et al.2009). At promoters that form intrinsically unstable RP_o, such as E. coli rRNA promoters, this effect of ppGpp and DksA negatively regulates transcription by facilitating RP_o collapse back to RP_c (Paul et al.2004). In contrast, at promoters where transcription is not rate-limited by RP_o stability, such as promoters of the E. coli amino acid biosynthesis genes, this effect of ppGpp and DksA positively regulates transcription by allowing the isomerization of RP_c to RP_o to occur more readily (Paul, Berkmen and Gourse 2005). Thus, (p)ppGpp is able to exert differential effects on transcription initiation based on the kinetics of the RNAP complex at a given promoter.

In B. subtilis, (p)ppGpp exerts its effects on transcription indirectly through changes in intracellular guanosine nucleobase metabolism (Kriel et al.2012). (p)ppGpp directly binds and inhibits enzymes involved in both de novo and salvage pathways of GTP synthesis. Specifically, (p)ppGpp potently inhibits the activity of guanylate kinase (GMK), which phosphorylates GMP to GDP (Kriel et al.2012; Liu et al.2015), and HprT, which converts hypoxanthine to inosine monophosphate (IMP) in the purine salvage pathway (Kriel et al.2012), collectively inhibiting GTP biosynthesis. At high (millimolar) concentrations, (p)ppGpp also inhibits the enzymatic activity of IMP dehydrogenase (IMPDH or GuaB), which converts IMP to xanthosine monophosphate (XMP) and controls a branch point between ATP and GTP synthesis in the purine salvage pathway, in both B. subtilis (Kriel et al.2012) and E. coli (Pao and Dyess 1981). Inhibition of the GTP biosynthesis pathways upon nutrient starvation alters the metabolite landscape, most notably causing depletion of GTP and concordant accumulation of ATP (Kriel et al.2012; Liu et al.2015). In addition, pppGpp synthesis depletes free GTP by using it as a substrate for pppGpp synthesis, further decreasing intracellular GTP levels and increasing the intracellular ATP:GTP ratio. ATP and GTP are the most common initiating nucleotide triphosphates (iNTP) used in B. subtilis and since binding of the iNTP to the transcription initiation complex stabilizes RP_o (Kuzmine, Gottlieb and Martin 2003), changes in the ATP:GTP ratio impacts gene expression. Consequently, B. subtilis genes that initiate with ATP, such as amino acid biosynthesis genes, are up-regulated while genes that initiate with GTP, such as rRNA genes, are down-regulated (Krásný et al.2008; Tojo et al.2010; Kriel et al.2012). Depletion of GTP also activates a starvation response through the transcription factor CodY (Handke, Shivers and Sonenshein 2008; Kriel et al.2012). In nutrient replete conditions, GTP binds CodY and blocks transcription of stress adaptive genes such as those involved in sporulation and branched chain amino acid biosynthesis (Ratnayake-Lecamwasam et al.2001; Belitsky and Sonenshein 2008). When GTP levels decrease during the stringent response, CodY-mediated repression is relieved and transcription of these genes is activated. Together, the ATP:GTP ratio and CodY contribute to the gene expression profile that results from the B. subtilis stringent response. However, the role of guanosine metabolites in the stringent response is likely more complex because GTP depletion alone is not sufficient for starvation survival (Liu et al.2015).

Intriguingly, Mtb does not appear to strictly follow either stringent response paradigm, and the mechanisms underpinning the Mtb stringent response remain mostly unknown. In terms of the E. coli paradigm of the stringent response, the sites known to be targeted by ppGpp on the E. coli RNAP are largely not conserved in the Mtb RNAP. Site 1 where ppGpp binds the E. coli RNAP involves numerous residues at the interface of the β’ and ω subunits, including β’ K615, β’ I619, β’ D622, β’ Y626, β’ R632, β’ R417, ω A2, ω R3, ω V4 and ω T5 (Ross et al.2013; Zuo, Wang and Steitz 2013). The Mtb RNAP shares identity or similarity with only two of these residues, E. coli β’ D622 (Mtb β’ D714) and E. coli β’ R362 (Mtb β’ K437). Site 2 on the E. coli RNAP that is targeted by ppGpp is DksA-dependent (Ross et al.2016) and Mtb does not encode DksA. In addition, the Site 2 residues located in the E. coli β’ subunit are not conserved in the Mtb RNAP.

Nonetheless, there is evidence that pppGpp directly binds the Mtb RNAP, raising the possibility that although the features of the (p)ppGpp-RNAP interaction in Mtb may be different than described in E. coli, this interaction could still occur. Addition of pppGpp to in vitro transcription reactions inhibited RP_o formation by Mtb RNAP at the Mtb rRNA rrnAP3 promoter (China et al.2012) and transcript formation by Mtb RNAP at the Mtb rrnAP3 and gyrB promoters (Tare, China and Nagaraja 2012; Tare, Mallick and Nagaraja 2013). Addition of Rv3788, a protein that binds near the secondary channel of the Mtb RNAP, blocks the effects of pppGpp on RP_o formation at the Mtb rrnAP3 promoter (China et al.2012), suggesting that the effects of pppGpp require its interaction with the RNAP. The effect of pppGpp on transcript formation by Mtb RNAP at the Mtb rrnAP3 promoter was also dependent on a thymine positioned three nucleotides downstream of the -10 element and L232 in σ^A, both of which affected RP_o stability, thus supporting a link between pppGpp binding to the RNAP and RP_o stability (Tare, Mallick and Nagaraja 2013). However, in all of these cases, effects of pppGpp were only observed at concentrations ≥100 μM, and often only at ≥500 μM, whereas the IC₅₀ of ppGpp on the E. coli RNAP at the rrnBP1 promoter is 12 μM (Ross et al.2013). Therefore, differences exist in how (p)ppGpp impacts RNAP activity in E. coli and Mtb and future studies will be necessary to determine whether the high concentrations of pppGpp used in vitro are physiologically relevant in vivo in Mtb.

Mycobacteria also share some similarities with the B. subtilis stringent response paradigm. For example, when M. smegmatis was engineered to accumulate higher levels of (p)ppGpp through mutation of the Rel_Msm hydrolase domain, ATP levels were increased and GTP levels were decreased (Weiss and Stallings 2013), indicating that (p)ppGpp synthesis affected purine metabolism. Accordingly, rRNA production is downregulated in mycobacteria during the stringent response and the two promoters (AP1 and AP3) upstream of the Mtb rRNA locus both initiate with GTP, which mirrors the pattern for transcriptional start sites at rRNA operons that has been reported in B. subtilis (Krásný et al.2008). Nonetheless, it has yet to be directly tested whether the identity of the iNTP for Mtb promoters is sufficient to determine the direction of gene expression during the stringent response. Unlike B. subtilis, pppGpp is not able to inhibit Mtb GMK (Rv1389) in vitro, even though GMK homologs from other Actinobacteria such as Streptomyces coelicolor were inhibited by pppGpp (Liu et al.2015). This is possibly explained by the lack of conservation in multiple residues in the Rv1389 active site that are required for the interaction of GMK with pppGpp in B. subtilis and S. coelicolor (Fig. 2). These polymorphisms are exclusive to Mtb, as even the (p)ppGpp-insensitive proteobacteria GMK homologs are conserved at these residues and their (p)ppGpp insensitivity is due to polymorphisms elsewhere in the protein (Liu et al.2015). Another difference between the B. subtilis paradigm and Mtb is that mycobacteria do not encode a CodY homolog. The effect of (p)ppGpp on the Mtb homologs of other B. subtilis (p)ppGpp targets in the purine biosynthesis pathway (HprT and IMP dehydrogenase) has not yet been tested and, therefore, could be (p)ppGpp targets in Mtb and explain the decreased GTP levels during (p)ppGpp synthesis. Alternatively, (p)ppGpp synthesis could be sufficient to deplete GTP by using it as a substrate. Together these data indicate that the classic model of (p)ppGpp impacting RP_o to alter gene expression during the stringent response could be conserved in Mtb, but the mechanistic features of this response are unique from model organisms.

Figure 2.

Sequence alignment of GMK homologs from Mycobacterium tuberculosis (Mtb), Streptomyces coelicolor and Bacillus subtilis. Yellow boxes indicate residues that interact with pppGpp and are conserved in all three species. Red boxes indicate residues that interact with pppGpp and are polymorphic in Mtb.

(p)ppGpp, POLYPHOSPHATE AND PERSISTENCE

In addition to RNAP and enzymes involved in GTP metabolism, polyphosphatase, a key regulator of polyphosphate (polyP) metabolism, has been investigated as a target of (p)ppGpp in Mtb. PolyP chains are linear polymers of phosphates that can reach hundreds to thousands of phosphates in length (Rao, Roberts and Torriani 1985). Mtb accumulates polyP chains when exposed to nutritional, nitrosative, oxidative, acidic and antibiotic stresses as well as during its transition to stationary phase (Thayil et al.2011; Singh et al.2013; Chuang et al.2015), overlapping with conditions that induce (p)ppGpp accumulation (Primm et al.2000; Stallings et al.2009). PolyP accumulation results in an overhaul of Mtb metabolism that arrests growth and facilitates Mtb persistence. In Mtb, polyP is synthesized primarily by the polyphosphate kinase PPK1 (Rv2984). Mtb encodes another polyphosphate kinase PPK2 (Rv1026) that can synthesize polyP, but PPK2 more efficiently carries out the reverse reaction as a polyphosphate dependent nucleoside diphosphate kinase that utilizes polyP to produce ATP and GTP (Sureka et al.2009; Shum et al.2011; Chuang, Belchis and Karakousis 2013). In addition, Δppk1 mutants produce only negligible amounts of polyP (Singh et al.2013), supporting that PPK1 serves as the primary polyP synthase in Mtb. Mtb encodes two polyphosphatases, PPX1 (Rv0496) and PPX2 (Rv1026), that hydrolyze polyP by cleaving phosphates from the termini of polyP chains. PPX1 and PPX2 differ in their substrate preference where PPX1 preferentially catabolizes short polyP chains and PPX2 specializes in breaking down long polyP chains (Choi et al.2012; Chuang et al.2015). (p)ppGpp directly inhibits the hydrolytic activity of Mtb PPX1 and PPX2 during in vitro exopolyphosphatase assays (Choi et al.2012; Chuang et al.2015). Inhibition of PPX1 or PPX2 would lead to the accumulation of polyP in Mtb and this has been proposed to contribute to the role of the stringent response in persistence (Thayil et al.2011; Chuang et al.2015). In support of the physiological relevance of this mechanism, deleting rel_Mtbcauses mid- and late- log phase cultures of Mtb to accumulate less polyP (Singh et al.2013).

How changes in polyP levels affect Mtb has been investigated using Mtb mutants that produce lower levels of polyP (Δppk1) or over-accumulate polyP (ppx1:Tn, ppx2 knockdown, and ppk2:Tn) (Table 1). The Δppk1 and ppx1:Tn strains grow slower in liquid culture and the Δppk1, ppx1:Tn, ppx2 knockdown and ppk2:Tn mutants all reach a lower maximum optical density at stationary phase (Thayil et al.2011; Chuang, Belchis and Karakousis 2013; Singh et al.2013; Chuang et al.2015). Δppk1, ppx1:Tn, and ppk2:Tn mutant strains are also attenuated in vivo (Thayil et al.2011; Chuang, Belchis and Karakousis 2013; Singh et al.2013). The Δppk1 and ppx1:Tn strains both showed decreased lung CFU starting at 14 days post-infection in the guinea pig infection model (Thayil et al.2011; Singh et al.2013). The ppk2:Tn strain showed decreased lung CFU at 14 days but grew back to levels comparable to WT Mtb in a mouse infection model (Chuang, Belchis and Karakousis 2013). Unlike the Δrel_Mtb mutant, the Δppk1, ppx1:Tn, and ppk2:Tn strains were impaired for survival in macrophages (Thayil et al.2011; Chuang, Belchis and Karakousis 2013; Singh et al.2013), suggesting that the virulence mechanisms mediated by polyP and the stringent response are not completely overlapping. Unexpectedly, the ppx2 knockdown strain grew better in macrophages (Chuang et al.2015), suggesting that the role of polyP metabolism in virulence requires further study to reconcile these seemingly contradictory findings. Nonetheless, these data demonstrated that proper regulation of polyP levels is important for Mtb virulence and the ability of (p)ppGpp to modulate polyP levels by direct inhibition of polyphosphatases could be one mechanism by which the stringent response impacts pathogenesis.

Table 1.

Summary of the phenotypes of polyP accumulating and polyP deficient Mtb strains. References are provided within the main text.

Phenotype	Δppk1	ppx1:Tn	ppx2 knockdown	ppk2:Tn
polyP levels	Decreased	Increased	Increased	Increased
Growth in liquid culture	Decreased growth rate in late log and stationary phase	Decreased growth rate	Decreased cell density during stationary phase	Decreased cell density during stationary phase
Growth in macrophages	Decreased survival in macrophages	Decreased survival in macrophages	Increased survival in macrophages	Decreased survival in macrophages
Virulence	Decreased virulence in guinea pigs	Decreased virulence in guinea pigs	Unknown	Decreased virulence in mice
Biofilm formation	Unknown	Decreased biofilm formation	Decreased biofilm formation	Decreased biofilm formation
Antibiotic and Stress tolerance	Increased sensitivity to isoniazid, levofloxacin, rifampicin and nitrosative stress	Decreased sensitivity to isoniazid. Increased sensitivity to clofazimine and hypoxia	Decreased sensitivity to isoniazid, heat, cell surface and acid stress	Decreased sensitivity to isoniazid. Increased sensitivity to naphthoquinone plumbagin and meropenem
Metabolism	Unknown	Decreased levels of peptidoglycan and phospholipid metabolism intermediates. Increased levels of TCA cycle and arginine metabolism intermediates	Decreased levels of glycolysis, fatty acid metabolism, phospholipid metabolism and pentose phosphate pathway intermediates	Decreased levels of peptidoglycan and phospholipid metabolism intermediates. Increased levels of arginine metabolism intermediates

Metabolomics studies revealed that the ppx1:Tn, ppk2:Tn and ppx2 knockdown mutants that accumulate polyP contain significantly less glycerol-3-phosphate (G3P), a major scaffold for phospholipids, accompanied by differences in the expression of enzymes involved in G3P metabolism. The enzymes responsible for G3P synthesis from dihydroxyacetone phosphate and glycerol, Rv0564c and GlpK, respectively, were significantly downregulated in the ppx1:Tn mutant and the enzyme responsible for the reverse reaction, Rv1692, which hydrolyses G3P to glycerol, was significantly upregulated in the ppk2:Tn strain. In addition, an enzyme involved in the recycling of glycerophospholipids for G3P synthesis, Rv2182c, was significantly downregulated in both the ppx1:Tn and ppk2:Tn mutants (Chuang et al., 2015, 2016). These data suggest that polyP accumulation signals for changes in gene expression that are linked to a reduction in de novo G3P synthesis, a reduction in recycling pathways for G3P synthesis, and increased G3P turnover. The ppx1:Tn, ppk2:Tn, and ppx2 knockdown mutants also exhibit decreased levels of metabolites involved in peptidoglycan synthesis and both ppx1:Tn and ppk2:Tn mutants express lower transcript levels for the peptidoglycan L,D-transpeptidases ldtA and ldtB (Chuang et al., 2015, 2016). Several metabolites involved in NADH metabolism, the TCA cycle and arginine metabolism also accumulate in the ppx1:Tn and ppk2:Tn mutants (Chuang et al.2016), while fatty acid biosynthesis and nucleotide biogenesis metabolites are reduced in the ppx2 knockdown strain (Chuang et al.2015). The accumulation of metabolites involved in the TCA cycle also occurs during exposure to hypoxia, a condition that induces the formation of drug and stress tolerant persisters (Wayne and Hayes 1996; Watanabe et al.2011; Eoh and Rhee 2013). Overall, the decrease in central metabolism and anabolic pathways during polyP accumulation is indicative of a shift towards metabolic quiescence.

PolyP accumulation in Mtb also affects antibiotic tolerance. The Δppk1 mutant that does not produce polyP showed increased susceptibility to rifampicin, levofloxacin and isoniazid (Singh et al.2013). In contrast, the polyP accumulating ppx1:Tn, ppk2:Tn and ppx2 knockdown mutants displayed decreased sensitivity to isoniazid (Thayil et al.2011; Chuang, Belchis and Karakousis 2013; Chuang et al.2015). However, the ppx1:Tn strain displayed increased sensitivity to clofazimine, which is thought to kill Mtb through membrane destabilization and altered redox cycling (Cholo et al.2017), while the ppk2:Tn strain displayed increased sensitivity to meropenem, which inhibits peptidoglycan L,D-transpeptidases (Chuang et al.2016). Therefore, although polyP production is important for antibiotic tolerance, over-accumulation of polyP results in changes in physiology that compromise antibiotic tolerance, indicating the polyP levels must be tightly regulated. Together, the changes that result from polyP accumulation coordinate a program that favors persistence, stress resistance and antibiotic tolerance, which would be induced when (p)ppGpp inhibits PPX1 and PPX2 and polyP accumulates.

In addition to (p)ppGpp promoting polyP accumulation, polyP also promotes Rel_Mtb expression and the production of (p)ppGpp through a signaling cascade involving the two component system MprAB and the alternative σ-factor σ^E (Sureka et al.2007) (Fig. 3). PolyP serves as a phosphodonor to the histidine kinase MprB, which then phosphorylates the response regulator MprA. Phosphorylated MprA activates expression of σ^E, which binds to a promoter sequence upstream of the rel_Mtb gene to activate transcription of rel_Mtb (He et al.2006; Sureka et al.2007). Accordingly, increasing intracellular polyP levels via mutation of ppx1 also increased transcript levels of rel_Mtb (Thayil et al.2011). However, the Δppk1 mutation does not affect the expression of rel_Mtb, suggesting that polyP is not necessary for Rel_Mtb expression but can induce higher expression (Singh et al.2013). This signaling cascade is amplified by several layers of positive feedback loops that rest on three molecular fulcrums: σ^E, MprA and (p)ppGpp. MprAB activates its own expression in addition to sigE (He and Zahrt 2005; He et al.2006) and σ^E activates expression of ppk1, the mprAB operon, and itself (Manganelli et al.2001; Sanyal et al.2013), thereby positively regulating this signaling cascade. (p)ppGpp also positively regulates its own production by inhibiting polyphosphatases to promote polyP accumulation. Thus, crosstalk between the polyP metabolism and stringent response activation pathways results in synchronized promotion of polyP and (p)ppGpp production.

Figure 3.

Signaling network involving Rel_Mtb, (p)ppGpp and polyP in Mtb. The network shows the proteins and molecules comprising the signaling cascades that positively and negatively regulate (p)ppGpp and polyP accumulation. Polyphosphate kinase 1 (PPK1), Polyphosphate kinase 2 (PPK2), Polyphosphatase 1 (PPX1) and Polyphosphatase 2 (PPX2) are shown. Green arrows indicate positive regulation. Red lines indicate negative regulation.

PolyP accumulation and the stringent response are both linked to Mtb persistence. The ability of Mtb to persist in the host is attributed to the formation of persister cells that exhibit decreased replication, altered metabolism, increased antibiotic tolerance and increased stress resistance (Gomez and McKinney 2004; Keren et al.2011). Persister cell formation occurs through bacterial bistability, a strategy used by bacterial populations to maintain phenotypically heterogeneous subpopulations that protect the population from environmental insults (Dubnau and Losick 2006; Smits, Kuipers and Veening 2006). Cells transition between phenotypic states through stochastic changes in gene expression that are reinforced by feedback loops. In E. coli, a subpopulation of persister cells is maintained through phenotypic heterogeneity in relA expression (Balaban et al.2004). Similarly, the mprAB-sigE-rel feedback loops underpin the maintenance of ‘bistable’ rel expression in mycobacteria (Sureka et al.2008; Ghosh et al.2011). When Sureka et al. sorted a population of M. smegmatis cells based on GFP expression from the rel_Msm promoter, they observed a bi-modal distribution of GFP expression with ‘high’ and ‘low’ subpopulations (Sureka et al.2008). As the cultures of M. smegmatis approach stationary phase, the ‘low’ rel_Msm-expressing subpopulation declines and the ‘high’ population increases. When the mprAB-sigE-rel feedback loop is disrupted by mutations to the σ^E binding site in the rel_Msm promoter, the ‘high’ subpopulation does not form (Sureka et al.2008), suggesting that this loop is necessary for maintaining high rel_Msmexpression. Although the existence of bistable rel_Mtb expression has not yet been examined in Mtb, the mprAB-sigE-rel feedback loop presents a model that connects the stringent response, polyP and mycobacterial persistence.

THE STRINGENT RESPONSE AND Mtb VIRULENCE GENE EXPRESSION

(p)ppGpp accumulates in Mtb during exposure to multiple stresses, including starvation, hypoxia and oxidative stress (Primm et al.2000; Stallings et al.2009), which are all conditions Mtb is exposed to during infection (Stallings and Glickman 2010). As a result, stress survival strategies and virulence mechanisms have converged such that activating the stringent response leads to expression of multiple virulence factors (Dahl et al.2003; Dalebroux et al.2010). A microarray study examining the effect of nutrient starvation on gene expression in WT Mtb versus the Δrel_Mtb mutant identified a Rel_Mtb regulon as the set of 159 genes whose expression changed significantly during starvation in WT bacteria but not in the Δrel_Mtb mutant (Dahl et al.2003). Genes involved in lipid metabolism comprised a significant proportion of the Rel_Mtb regulon indicating that Rel_Mtb may coordinate changes in the production of Mtb lipids and lipoproteins that can play important protective and immunomodulatory roles during infection (Dahl et al.2003; Cambier et al.2014). With a few exceptions, the trend of gene expression suggests that Rel_Mtb may coordinate a shift from lipid anabolism to catabolism during starvation, which could represent the bacteria liberating nutrients from lipids but may also affect lipid modulators of virulence. Indeed, the Rel_Mtb regulon also included several polyketide synthases (PKS) and polyketide associated protein (Pap) genes that are required to synthesize complex lipids (Dahl et al.2003; Gokhale et al.2007). Notably, during starvation, Rel_Mtb upregulates pks2 and papA1, which are responsible for synthesizing sulfolipid-1 (SL-1) (Dahl et al.2003; Bhatt et al.2007). SL-1 is the most abundant sulfolipid in the Mtb cell wall and purified SL-1 has been reported to have immunomodulatory effects (Pabst et al.1988). PKS-synthesized surface glycolipids are also critical for biofilm formation in vitro (Pang et al.2012; Cambier et al.2014). Accordingly, loss of Rel_Mtb-mediated (p)ppGpp synthesis resulted in delayed formation of biofilms and pellicles in static liquid cultures (Weiss and Stallings 2013). The relevance of biofilm formation for Mtb pathogenesis is currently uncertain, but Mtb during chronic infection shares many characteristics with Mtb within biofilms, including decreased replication rates, and increased stress tolerance (Ojha et al.2008; Richards and Ojha 2014). In addition, multiple groups have also reported the observation of extracellular communities of Mtb that resemble biofilms during infection (Lenaerts et al.2007; Wong and Jacobs 2016).

The Δrel_Mtb mutant also exhibited lower expression of the mce1, mce3 and mce4 (mammalian cell entry) operons that have been associated with macrophage entry and are critical for Mtb infection (Zhang and Xie 2011) as well as lower expression of the major secreted antigen ESAT-6 (Dahl et al.2003; Ganguly et al.2007; Wang et al.2009; Sreejit et al.2014). Additionally, multiple genes belonging to the PE-PPE (Pro-Glu-rich) family of Mtb proteins were up- or downregulated in the Δrel_Mtb mutant (Dahl et al.2003). PE-PPE proteins are either secreted or are cell surface proteins that are thought to contribute to Mtb virulence (Singh et al.2016). Collectively, this data suggests that the Mtb stringent response has crosstalk with numerous virulence pathways. The connection between the stringent response and virulence suggests that Mtb may have coopted sensing changes in the host environment to activate virulence mechanisms and changes in metabolism to promote adaptation and persistence.

TARGETING Rel_Mtb AS AN ANTI-VIRULENCE STRATEGY TO TREAT TUBERCULOSIS

There has been significant interest in identifying compounds that target (p)ppGpp synthesis as an anti-virulence and anti-persistence strategy in many pathogenic bacteria, including Mtb (Zhang, Yew and Barer 2012). Relacin, a 2΄-deoxyguanosine-based analogue of ppGpp with the original pyrophosphate moieties at positions 5΄ and 3΄ replaced by glycyl-glycine dipeptides, was designed based on the Rel_Seq crystal structure and shown to inhibit E. coli RelA in vitro and (p)ppGpp synthesis in B. subtilis (Wexselblatt et al., 2010, 2012). However, Relacin is unable to penetrate the mycobacterial cell envelope to access its cytoplasmic target Rel_Mtb. More recently, Syal et al. designed and synthesized analogs of Relacin that had activity in mycobacteria by functionalization of the amine group at C-2 position of the guanine base in Relacin (Syal et al.2017). These next generation Rel inhibitors were able to inhibit long-term survival and biofilm formation of mycobacteria. However, all compounds that have been designed to target the stringent response so far are analogs of ppGpp, which has limits in terms of improving potency and cell permeability. Therefore, the identification of new chemical matter will likely be important to translate (p)ppGpp synthesis inhibitors to clinical use.

SUMMARY AND OPEN QUESTIONS

A model has emerged where Rel_Mtb-synthesized (p)ppGpp alters gene expression profiles, purine metabolism and polyP metabolism to enhance Mtb stress responses, virulence and persistence (Fig. 4 and Table 2). In addition to the RNAP, GTP synthesis enzymes, and polyphosphatases, other molecular targets of (p)ppGpp have been identified in non-mycobacterial species, but have yet to be explored in mycobacteria (Table 3) (Kanjee, Ogata and Houry 2012; Hauryliuk et al.2015; Liu, Bittner and Wang 2015). Many of these additional (p)ppGpp targets share a common theme of being GTPases or GTP-binding proteins with binding pockets for other guanine nucleotides that likely provide a natural binding site for (p)ppGpp. In addition to the GTPases or GTP-binding proteins, (p)ppGpp has been shown to target E. coli lipid and amino acid metabolism enzymes (Merlie and Pizer 1973; Polakis, Guchhait and Lane 1973; Stein and Bloch 1976; Heath, Jackowski and Rock 1994; Kanjee et al.2011; Kanjee, Ogata and Houry 2012). Whether (p)ppGpp binds and modulates similar enzymes in Mtb remains an open question for further study, but it is possible that (p)ppGpp effects on Mtb physiology are much more complex than our current understanding. In particular, our current understanding of how the Mtb stringent response promotes pathogenesis is based on transcriptional responses that are lost in the Δrel_mtb mutant. Importantly, there exists a clear precedent for (p)ppGpp interacting with and affecting the activity of proteins post-translationally, so a purely transcriptome-level approach may miss key mechanistic details of the stringent response. Nonetheless, it is clear that by some mechanism (p)ppGpp metabolism is important for survival during stress, antibiotic treatment and virulence in animal models, highlighting the stringent response as a key mediator of Mtb pathogenesis.

Figure 4.

Summary of the effects of (p)ppGpp in Mtb. The RelMtb synthetase domain transfers the 5΄-β,γ-pyrophosphate from ATP to the 3΄-OH of GDP or GTP to synthesize (p)ppGpp, which can then be hydrolyzed by the RelMtb hydrolase domain. (p)ppGpp accumulates in response to a number of stresses, including starvation and oxidative stress, leading to alterations in purine metabolism, gene expression and polyP metabolism. It is also possible that there are other unidentified targets of (p)ppGpp in Mtb. These effects of (p)ppGpp accumulation contribute to enhanced virulence, persistence and stress responses.

Table 2.

Molecular targets of (p)ppGpp that have been investigated in Mycobacterium tuberculosis (Mtb).

Molecular targets	Cellular process	Effects in Mtb	References
RNAP	Transcription	Inhibited at high [(p)ppGpp] in vitro	China et al. (2012), Tare, China and Nagaraja (2012) and Tare, Mallick and Nagaraja (2013)
GMK, Guanlyate Kinase	Purine metabolism	Insensitive to (p)ppGpp	Liu et al. (2015)
PPX1 and PPX2, Polyphosphatases	Polyphosphate metabolism	Inhibited by (p)ppGpp in vitro	Choi et al. (2012) and Chuang et al. (2015)

Table 3.

Molecular targets of (p)ppGpp that have been identified in other bacteria but not yet explored in mycobacteria.

Cellular process	Molecular targets	References
DNA replication	DnaG	Wang, Sanders and Grossman (2007) and Maciag et al. (2010)
Translation	IF2, EF-G, EF-Tu, EF-Ts, Obg, RsgA	Legault, Jeantet and Gros (1972), Rojas et al. (1984) and Persky et al. (2009)
Nucleotide metabolism	AprT, HprT, GprT, Upp, GuaB, PurA, GdpP	Gallant, Irr and Cashel (1971), Hochstadt-Ozer and Cashel (1972), Fast and Sköld (1977), Pao and Dyess (1981) and Rao et al. (2010)
Amino acid metabolism	HisG, LdcI, LdcC, SpeC	Kanjee et al. (2011)
Lipid metabolism	PlsB, PgsA, AccA, AccD, FabA, FabZ,	Merlie and Pizer (1973), Polakis, Guchhait and Lane (1973) and Stein and Bloch (1976)
Central metabolism	GdhA, Pcc	Pao and Dyess (1981) and Maurizi and Rasulova (2002)
Glycogen biosynthesis	GlgC	Dietzler and Leckie (1977)

FUNDING

CLS is supported by a Burroughs Wellcome Fund Investigator in the Pathogenesis of Infectious Disease Award and National Institutes of Health (NIH) grants GM107544 and AI111696. JP is supported by the National Institute of General Medicine of the National Institutes of Health (NIGMS) Cell and Molecular Biology Training Grant GM007067 and the Stephen I. Morse Graduate Fellowship.

Conflict of interest. None declared.