NAD⁺‐targeting by bacteria: an emerging weapon in pathogenesis

ABSTRACT

Nicotinamide adenine dinucleotide (NAD⁺) is a major cofactor in redox reactions in all life-forms. A stable level of NAD⁺ is vital to ensure cellular homeostasis. Some pathogens can modulate NAD⁺ metabolism to their advantage and even utilize or cleave NAD⁺ from the host using specialized effectors known as ADP-ribosyltransferase toxins and NADases, leading to energy store depletion, immune evasion or even cell death. This review explores recent advances in the field of bacterial NAD⁺‐targeting toxins, highlighting the relevance of NAD⁺ modulation as an emerging pathogenesis strategy. In addition, we discuss the role of specific NAD⁺‐targeting toxins in niche colonization and bacterial lifestyle as components of toxin/antitoxin systems and key players in interbacterial competition. Understanding the mechanisms of toxicity, regulation and secretion of these toxins will provide interesting leads in the search for new antimicrobial treatments in the fight against infectious diseases.

NAD⁺, NADase, ADP-ribosylation, bacterial virulence factor, secretion pathway, immune response evasion

INTRODUCTION

Discovered over a century ago as a coenzyme for yeast fermentation (Harden, Young and Martin 1906), nicotinamide adenine dinucleotide (NAD⁺) is one of the most abundant and essential molecules in organisms. After first being described by Sir Arthur Harden in 1906, two additional Nobel prizes went on to identify its role as a nucleoside sugar phosphate (Von Euler-Chelpin 1929) and in hydrogen transfer in fermentation (Warburg, Christian and Griese 1935). The function of NAD⁺ as a coenzyme in energy metabolism in eukaryotic cells was further elucidated by many scientists, including Krebs (Krebs and Veech 1969), detailing its importance in glycolysis, the tricarboxylic acid cycle and mitochondrial oxidative phosphorylation. To date, NAD⁺ has been implicated in hundreds of biochemical and biological processes, acting either as an electron carrier in oxidation/reduction reactions or being consumed as a substrate by NAD⁺‐consuming enzymes. Therefore, besides its role as a cofactor in electron transfer, NAD⁺ and its metabolites function as signaling molecules in regulating immune responses and inflammation, cell division, circadian rhythm, energy metabolism, neuronal function, DNA repair and stress resistance (Rajman, Chwalek and Sinclair 2018). It is well established that NAD⁺ levels decline with age for all species, impacting cellular metabolism and increasing susceptibility to disease (Zhu et al. 2015b). Indeed, maintenance of the intracellular NAD⁺/NADH ratio is vital for cell survival, and therefore, perturbation of NAD⁺ homeostasis can be associated with human disease. NAD⁺ levels change not only during physiological processes such as physical activity and diet but also in the context of metabolic disorders and several diseases, including cancer and neurodegenerative diseases (Rajman, Chwalek and Sinclair 2018).

Infectious diseases also negatively impact NAD⁺ levels in infected hosts through the action of bacterial NAD⁺‐utilizing toxins. Numerous pathogenic bacteria rely on toxins with ADP-ribosyltransferase activity to modify cellular functions. These toxins cleave NAD⁺ and transfer ADP-ribose, the hydrolysis product, onto a specific target protein (Simon, Aktories and Barbieri 2014; Catara et al. 2019), precisely remodeling a particular cellular function, such as protein synthesis, cytoskeleton dynamics or signal transduction. Some bacterial pathogens secrete toxins with a NAD⁺‐consuming activity, thus called NADases, that often lethally induce NAD⁺ depletion in targeted host cells.

In this review, we present first a brief overview of the central role of NAD⁺ in bacterial infections followed by a detailed discussion on the current knowledge of bacterial NAD⁺‐utilizing toxins, describing their pathogenic mechanisms, key structural features, regulation and secretion pathways. We also briefly discuss the potential of bacterial NAD⁺‐utilizing enzymes as promising targets for discovering and developing novel antimicrobial compounds.

CENTRAL ROLE OF NAD⁺ IN BACTERIAL SURVIVAL AND VIRULENCE

NAD⁺: friend and foe

The mammalian and bacterial NAD⁺ biosynthetic pathways are summarized in Figs 1 and 2 and the chemical structures of the main precursors in Box 1. In many bacterial pathogens, both de novo synthesis from L-aspartate and nicotinamide-dependent salvage pathway are needed for supplying nicotinic acid mononucleotide (NaMN), an essential precursor for NAD synthesis. Indeed, some of the enzymes in these pathways are highly conserved in bacterial genomes, namely those utilizing NaMN and participating in further downstream reactions generating NAD⁺ and its phosphorylated derivative (NADP⁺). These include NaMN adenylyltransferase (NadD), NAD synthetase (NadE) and NAD kinase (NadK) (Sorci et al. 2009a,b). However, some pathogens are NAD⁺ auxotrophs or present some restrictions in their ability to utilize biosynthetic NAD⁺ precursors. Thus, they have acquired the capacity to use the NAD⁺ from infected hosts, essential for their survival.

Figure 1.

NAD(P)⁺ biogenesis pathways and NAD⁺‐consuming enzymes in mammals. In mammals, the biosynthesis of NAD⁺ can occur through several pathways. The salvage pathway (green) is the principal generator of NAD⁺ in humans from nicotinamide (NAM). However, NAD⁺ can also be synthesized from the amino acid tryptophan by the de novo pathway (blue). In addition, nicotinic acid (NA), also known as niacin or nicotinate, can serve as a precursor of NAD⁺ through the shorter three-step Preiss–Handler pathway (purple). NAD⁺ can be reduced in NADH or transformed in NADP⁺ through the redox metabolism (pink). NAD⁺ can be consumed by eukaryotic and bacterial enzymes. Bacterial NADases are also capable of using NAD(P)⁺. NAD⁺: nicotinamide adenine dinucleotide, ACMS: α-amino-β-carboxymuconate-ε-semialdehyde, NaMN: nicotinic acid mononucleotide, NA: nicotinic acid, NaAD: nicotinic acid dinucleotide, NR: nicotinamide riboside, NMN: nicotinamide mononucleotide and NAM: nicotinamide; IDO: indoleamine 2,3-dioxygenase, TDO: tryptophan 2,3-dioxygenase, QPRT: quinolinate phosphoribosyltransferase, NAPRT: NA phosphoribosyltransferase, NMNAT: NMN adenylyltransferase, NADSY: NAD⁺ synthetase, NADK: NAD⁺ kinase, NRK: NR kinase, NAMPT: nicotinamide phosphoribosyltransferase, NMNAT: NMN adenylyltransferase and PRPP: phosphoribosyl pyrophosphate; SIRTs: sirtuins, PARPs: poly(ADP-ribose) polymerases, CD38: cluster of differentiation 38; and bARTs: bacterial ADP-ribosylating toxins.

Figure 2.

NAD(P)⁺ biogenesis pathways and redox metabolism in bacteria. In contrast to mammals, many bacteria can also synthesize NAD⁺de novo from the amino acid aspartate, and they use a salvage pathway where the substrate NAM is directly converted into NA by nicotinamidases, absent in mammals. NAD⁺ precursor NaMN is the first intermediate shared by the most common de novo (blue) and salvage pathways (green). Then, NaMN feeds the central biosynthetic pathway through route I or route II found in Bacillus anthracis or Francisella tularensis indicated by the fuchsia or the orange arrow, respectively (Sorci et al.2009a). There are two alternative routes to form NMN (dashed lines), via NadV (e.g. in F. tularensis) or NadR (e.g. Mycobacterium tuberculosis, Escherichia coli) (Sorci et al. 2009a; Rodionova et al. 2014). The redox metabolism of NAD(P)H is indicated in pink. NAD⁺: nicotinamide adenine dinucleotide, NA: nicotinic acid, NaMN: nicotinic acid mononucleotide, NaAD: nicotinic acid dinucleotide, NR: nicotinamide riboside, NMN: nicotinamide mononucleotide and NAM: nicotinamide; NadA: quinolinate synthase, NadB: L-aspartate oxidase, NadC: nicotinate-nucleotide pyrophosphorylase, NadD: NaMN adenylyltransferase, NadE: NAD synthetase, NadK: NAD kinase, NadR: NMN adenylyltransferases, NadV: NAM phosphoribosyltransferase, PncA,B: nicotinamidases.

Box 1. Chemical structures of the main NAD⁺ precursors.

Major NAD+ precursors	Chemical structure
Nicotinamide adenine dinucleotide (NAD+)
Nicotinamide adenine dinucleotide (NADH)
Nicotinamide (NAM)
Nicotinic acid (NA)
Nicotinamide mononucleotide (NMN)
Nicotinic acid mononucleotide (NaMN)
Nicotinic acid dinucleotide (NaAD)
Nicotinamide riboside (NR)

NAD⁺ auxotrophy has been shown for the Gram-negative bacteria Haemophilus influenzae and Shigella flexneri, which lack key genes for NAD⁺de novo synthesis pathway (Mantis and Sansonetti 1996; Kemmer et al. 2001). NAD⁺ precursor auxotrophy was also implicated in Mycobacterium tuberculosis pathogenesis, that is capable of using exogenous nicotinamide from the host through two NAD biosynthesis enzymes, NadD and NadE (Boshoff et al. 2008; Vilchèze et al. 2010). Thus, NadD and NadE deletion is deleterious for M. tuberculosis, and enzymatic knockdown reduces its survival during infection (Boshoff et al. 2008; Vilchèze et al. 2010), making them promising drug targets (Rodionova et al. 2014). Interestingly, in the case of S. flexneri, absence of NadA and NadB, both enzymes of the de novo NAD⁺ synthesis pathway, increases its virulence capacity suggesting pathoadaptive evolution of Shigella toward inactivation of these antivirulence loci to become nicotinic acid auxotroph (Prunier et al. 2007).

Besides NAD⁺ being essential for metabolic and energy homeostasis in pathogens, there is growing evidence that NAD⁺ biogenesis can be indispensable for their fitness and virulence. This is well illustrated by the transcriptional repressor NrtR from Pseudomonas aeruginosa. NrtR binds an intergenic sequence upstream of pncA encoding a nicotinamidase, whose subsequent inactivation leads to a sharp drop in the pathogen's competitive fitness in vitro and in vivo (Okon et al. 2017). More recently, the nadB gene encoding a putative L-aspartate oxidase in Coxiella burnetii required for de novo NAD biosynthesis was also shown to be essential for Coxiella-containing vacuole formation and subsequent replication (Bitew et al. 2018).

Bacterial mobilization and targeting of eukaryotic NAD⁺‐consuming enzymes during infection

Maintenance of host NAD⁺ levels is not only dependent on enzymes from the biosynthesis pathways as the indoleamine 2,3-dioxygenase (IDO) but also NAD⁺‐consuming enzymes including sirtuins (SIRTs), poly (ADP-ribose) polymerases (PARPs) and cluster of differentiation 38 (CD38). These enzymes are all interconnected and play vital roles inside the cell (Cantó, Menzies and Auwerx 2015) and are detailed in Box 2.

Box 2. Eukaryotic NAD⁺‐consuming enzymes

PARPs

PARPs carry out ADP-ribosylation, a reversible posttranslational modification that catalyzes the transfer of ADP-ribose from NAD⁺ to a substrate. In the nucleus, they have been shown to play an active role in DNA repair, transcription and maintenance of chromatin structure whereas in the cytosol they have been implicated in inflammation, cell death and endoplasmic reticulum stress (Gibson and Kraus 2012).

SIRTs

SIRTs belong to a family of deacetylases that can present additional enzymatic activities, including ADP-ribosylation. They generally respond to NAD⁺ availability, becoming particularly active upon energy deficiency or low carbohydrate levels but are also negatively regulated by components of the NAD biosynthesis pathways, such as NAM. They are involved in a variety of cellular processes, including mitochondrial biogenesis, stress adaptation, histone modifications and inflammation (Haigis and Sinclair 2010).

CD38

CD38 is a key cellular NADase involved in NAD hydrolysis in the cell. This transmembrane protein can sense and regulate extracellular levels of NAD⁺. Thanks to its cyclic ADP-ribose (cADP) synthase and hydrolase activities, CD38 cleaves NAD⁺ to generate cADP or ADP-ribose, important second messengers that modulate Ca²⁺ release from intracellular storage sites and both innate and adaptive immune responses. CD38 has also been implicated in other signaling cascades, such as cell cycle and insulin (Aksoy et al. 2006; Malavasi et al. 2008).

PARPs, SIRTs and CD38 all play an important role in the control of bacterial infections. This is nicely illustrated by the case of Salmonella enterica serovar Typhimurium, that was shown to upregulate nuclear PARP1 in epithelial cells (Qi et al. 2017). In macrophages, PARP1-dependent necrotic cell death during Salmonella infection has been described (Ro et al. 2018). More recently, Matalonga et al. have highlighted a new role played by the nuclear receptor LXR in modulating host NAD⁺ levels in a CD38-dependent manner (Matalonga et al. 2017). LXR activates the transcription of the gene encoding the CD38 enzyme, preventing Salmonella Typhimurium from inducing actin cytoskeleton rearrangements necessary for invasion.

SIRT2 is also upregulated during SalmonellaTyphimurium infection of dendritic cells, promoting inflammation and nitric oxide production, and eliciting a robust antibacterial response (Gogoi et al. 2018). The subcellular location of SIRT2, crucial for its function, is significantly impacted in a number of bacterial infections (North et al. 2003; Michan and Sinclair 2007; Bhaskar et al. 2020). Listeria monocytogenes infection results in dephosphorylation of SIRT2, inducing its translocation to the nucleus and its association with chromatin. Subsequent histone modifications and targeted gene repression mediated by SIRT2 favors Listeria virulence (Pereira et al. 2018). The role of SIRT2 has also been recently studied during M. tuberculosis infection where SIRT2 is upregulated and translocated to the nucleus leading to important modification in host gene expression (Bhaskar et al. 2020). SIRT2 inhibitors have been shown to be effective in limiting M. tuberculosis infection by reducing bacterial burden and increasing immune responses in a mouse model (Bhaskar et al. 2020).

Given the importance of these eukaryotic NAD⁺‐consuming enzymes for cellular homeostasis, it is not surprising that some pathogens have evolved to hijack them to control NAD⁺ levels, causing metabolic dysfunction and promoting virulence. This strategy is used by different classes of intracellular pathogens including bacteria, virus and parasites to modulate the functions of immune cells, impeding microbial clearance in both animals (Zerez et al. 1990; Murray, Nghiem and Srinivasan 1995; Ba et al. 2010; Moreira et al. 2015) and plants (Pétriacq et al. 2013).

Salmonella, for example, targets SIRT1 to gain specific control of the host acetylation machinery. Salmonella Typhimurium directs SIRT1 and associated proteins to lysosomes for degradation, ultimately enabling the bacteria to evade autophagy (Ganesan et al. 2017). Other pathogens display increased virulence through the alteration of SIRT1 expression. Cheng et al. have demonstrated M. tuberculosis infection downregulates SIRT1 expression and reduces NAD⁺ levels, including in patients with active tuberculosis (Cheng et al. 2017). Although extensive depletion of intracellular NAD⁺ is due to a specific toxin secreted by M. tuberculosis belonging to the NADase family (see the section 'SPN and TNT toxicity in the eukaryotic host: two textbook examples'), activation of SIRT1 activity using specific molecules inhibited intracellular growth of the bacterium. SIRT1 activation was accompanied by an increase of host cell autophagy and phagosome-lysosome fusion in vitro as well as a reduction of inflammation in vivo.

Multiple pathogens have been shown to directly target PARP-1, including Chlamydia trachomatis that injects an effector protein called CPAF into host cells, which cleaves PARP-1 to reduce inflammation (Paschen et al. 2008). Helicobacter pylori also secretes a specific protease that directly activates PARP-1, to generate the catabolite poly(ADP-ribose), regulate the host immune response and potentially enhance cell death (Nossa et al. 2009).

Crosstalk between NAD⁺ and the immune response

Mounting evidence strongly suggests that NAD⁺ coordinates host innate and adaptive immune responses via several mechanisms implicating: (i) the tryptophan-catabolizing enzyme IDO, (ii) the NAD⁺‐biosynthesis enzyme NAMPT and (iii) NAD⁺‐utilizing enzymes PARPs and SIRTs.

IDO is known to induce immunosuppression and subsequent impaired proliferation and inhibition of T cell functions, relevant in the context of infectious diseases (for complete review, see Schmidt and Schultze 2014). For example, IFN-γ activity enhanced by TNF-α induces IDO, inhibits Staphylococcus aureus replication in human brain endothelial cells and leads to bacteriostasis (Schroten et al. 2001). Tryptophan catabolism by IDO was also shown to be involved in limitation of tissue damage caused by Clostridium difficile. The knockout of IDO increases the destruction of the murine mucosa by this pathogen and the production of IFN-γ by neutrophils (El-Zaatari et al. 2014). It is therefore not surprising that some pathogens exploit the immunomodulatory activities of IDO as it has been observed for uropathogenic Escherichia coli (UPEC). After infection of uroepithelial cells, UPEC induce IDO overexpression resulting in attenuation of the immune response to promote bacterial survival and proliferation (Loughman and Hunstad 2012).

Likewise, other NAD⁺‐biosynthesis enzymes coordinate host immune responses against bacterial infection. Besides its role in the NAD⁺ salvage pathway, NAMPT can be a relevant actor in inflammatory diseases. NAMPT can trigger cytokine secretion from inflammatory cells, including TNF-α, IL-1β and IL-6 (Brentano et al. 2007) and delay neutrophil apoptosis (Jia et al. 2004). Interestingly, therapeutic NAMPT targeting has an anti-inflammatory action by reducing neutrophil respiratory oxidative burst, without impacting host defense against bacterial pathogens such as S. aureus (Roberts et al. 2013). Modulation of NAMPT may prove useful in the context of bacterial sepsis.

In the case of NAD⁺‐utilizing enzymes there is a clear reciprocal dialogue essential for NAD⁺ dependent signaling and metabolism. This is particularly true for PARP1 and SIRT1, whose functions are intimately related. PARP1 inhibits SIRT1 by limiting NAD⁺ levels and preventing SIRT1 transcription. Furthermore, while SIRT1 deacetylates RelA/p65 resulting in inhibition of NF-κB, PARP1 induces inflammation through the transcriptional activation of NF-κB. Interestingly, both SIRTs and PARPs also contribute to regulation of cell death by modulation of p53 activity. The cross-talk existing between NAD⁺ and immune response has been extensively reviewed elsewhere (Cantó et al. 2015; Mesquita et al. 2016; Singhal and Cheng 2019).

We would like to highlight several recent studies that have strengthened the link between NAD⁺ homeostasis and the immune response in infected cells. The sterile alpha and Toll/interleukin-1 receptor motif-containing 1 (SARM1) is a well-characterized Toll-like receptor (TLR) adaptor, involved in innate immune responses and also plays a critical role in axonal degeneration in neurodegenerative diseases. In the context of axonal degeneration, which is characterized by severe NAD⁺ depletion, Essuman and colleagues highlighted the NADase activity of SARM1, more precisely its TIR domain (Essuman et al. 2017). Horsefield et al. and Wan et al. went on to confirm these results, revealing that the TIR domain-containing plant resistance (R) proteins and SARM function as NAD⁺‐cleaving enzymes (Horsefield et al. 2019; Wan et al. 2019). Their enzymatic activity is triggered by specific pathogen recognition leading to activation of cell death-dependent plant defence responses via depletion of NAD⁺. These studies highlight a new family of NAD⁺‐consuming enzymes relying on a TIR domain, highly conserved in animal and plant cell innate immune signaling pathways, including all TLRs, corresponding adaptors and IL1 receptor.

In the last two decades, TIR domains have also been described in various bacterial proteins displaying immune-suppressive effects in eukaryotic cells during infection. Thus, Essuman et al. postulated that the TIR domain's ancestral role is to regulate host metabolic pathways by modulating NAD⁺ levels instead of counteracting immune cell signaling by direct interaction with TLRs or adaptors (further discussed in the section 'Bacterial TIR domain-containing toxins: an atypical NADase family').

BACTERIAL NAD⁺‐UTILIZING TOXINS AS KEY VIRULENCE FACTORS

Host NAD⁺ modulation by bacterial toxins is a widespread activity among bacterial species. Categorizing bacterial NAD⁺‐utilizing toxins is challenging because they cause distinct toxic effects through different molecular processes and enzymatic activities. This is further complicated by the fact that a single toxin can have multiple effects, as illustrated for the Streptococcus pyogenes SPN toxin, which exhibits three different enzymatic activities to degrade NAD⁺ (Ghosh et al. 2010). Another example is the NarE toxin from Neisseria meningitidis displaying either a ADP-ribosyltransferase activity in host cells or a NAD⁺‐glycohydrolase activity in the absence of a substrate (Carlier et al. 2011). However, we decided to split the different bacterial NAD⁺‐utilizing toxins into two main groups according to their function in the target host: covalent modification of a target protein mediated by ADP-ribosylating toxins (bARTs) or depletion of the NAD⁺ levels by bacterial NADases. We have further divided the NADase family into several groups according to their distinct structural patterns and functions inside the host, with focus on the secretion pathways and the regulatory mechanisms involved.

Bacterial ADP-ribosylating toxins (bARTs)

A number of bacterial toxins are capable of damaging their host by modulating the activity of host proteins through covalent modifications. ADP-ribosylation by the Diphtheria toxin was the first one to be described (Collier 1975). As mentioned above, this posttranslational modification entails transferring of the ADP-ribose moiety of NAD⁺ to a target molecule, releasing the nicotinamide (NAM) moiety. ADP-ribosylation is widely performed by bacterial and eukaryotic enzymes.

bARTs play an important role in bacterial pathogenicity, including in P. aeruginosa, E. coli, Bordetella pertussis, Salmonella sp., Clostridium botulinum and Corynebacterium diphtheriae (Simon, Aktories and Barbieri 2014). Indeed bARTs are the causative agents of many diseases affecting insects, plants and humans (Visschedyk et al. 2010; Feng et al. 2016; Catara et al. 2019). A recent and exhaustive review nicely summarized the eukaryotic mechanisms modulated by ADP-ribosylating virulence factors, such as modulation of the cytoskeleton to promote host cell death in both eukaryotic and prokaryotic organisms (Catara et al. 2019). Thus, we will present herein a selection of bARTs associated with bacterial virulence and exhibiting impact on human health as examples for this family of toxins (see Table 1).

Table 1.

Selection of bARTs discussed in this review. For a complete table of bARTs, see Catara et al. (2019).

Most bARTs are characterized by two domains (A:B), in which the A domain corresponds to the catalytic ADP-ribosylating turn-turn (ARTT) domain and the B domain is involved in cell receptor binding. As such, one can distinguish two main families of bARTs based on their target proteins and structural features: the diphtheria-like toxins (DT) and the cholera-like toxins (CT). In the DT-like group, the A and B domains are present in a single chain, whereas in the CT-like group, the catalytic domain is generally noncovalently bound to five B domains. However, some exceptions have been reported such as for the Mycoplasma pneumoniae community-acquired respiratory distress syndrome toxin (Becker et al. 2015). The CT-like toxins can be further subdivided in C2-like, with the two polypeptides independently expressed constituting a binary toxin, and the C3-like exotoxin subgroup consisting of a single A domain subunit (Simon, Aktories and Barbieri 2014).

The NAD⁺ binding pocket is characterized by a conserved arginine residue for CT-like toxins and a histidine for DT-like toxins, essential for the active-site integrity. Another important conserved region consists of a serine-threonine-serine (STS) or a tyrosine-x-tyrosine (YX₁₀Y) motif of CT-like or DT-like toxins, respectively, known to stabilize the NAD⁺ binding site. Also, a key feature of the bARTs family is the presence of a conserved glutamic acid in the active site containing the ARTT motif: the histidine-tyrosine-glutamic acid (HYE) and the arginine-serine-glutamic acid (RSE) triads from DT-like toxins and CT-like toxins family, respectively, are essential for transferase activity (Holbourn, Shone and Acharya 2006; Simon, Aktories and Barbieri 2014; Catara et al. 2019).

DT-like toxins are essentially involved in the modulation of the protein synthesis machinery by modifying the eukaryotic elongation factor-2, as ExoA, ChT and DTX from P. aeruginosa, V. cholerae and C. diphtheriae, respectively (Iglewski 1977; Wilson and Collier 1992; Jørgensen et al. 2008). CT-like toxins are implicated in intracellular trafficking, actin cytoskeleton modulation and G protein modification. The best-known CT-like toxin is the CTX toxin itself from V. cholerae, which upon entry into the cytoplasm of intestinal cells modifies Gαs through ADP-ribosylation (Kopic and Geibel 2010). Gαs are heterodimeric G proteins (or guanine nucleotide-binding proteins) subunit that activate the cyclic AMP-dependent pathway by stimulating the production of cAMP from ATP (Bomsel and Mostov 1992). Gα modification leads to increased cAMP levels, efflux of sodium, potassium and water into the intestinal lumen resulting in severe diarrhea (Kopic and Geibel 2010). A similar mechanism has been reported for the heat-labile toxin (LT) from E. coli ETEC (Chang et al. 1987), whereas the pertussis toxin (PT) from B. pertussis (Katada and Ui 1982) has been shown to target an inhibitory G protein (Gαi) that inhibits adenylate cyclase activity (Table 1).

More recently, Patry et al. have highlighted the involvement of CT-like toxins in bacterial gut competition thus identifying a new way for bARTs to impact human health (Patry et al. 2019). The B domain of the cholerae toxin is known to display high affinity for ganglioside cell receptors (MacKenzie et al. 1997) and some gut pathogens like Campylobacter jejuni can synthesize mimics of the GM1 ganglioside receptor (Parker et al. 2005). The CT and LT toxins can bind C. jejuni GM1-mimicking lipooligosaccharides leading to C. jejuni growth arrest. This mechanism was not restricted to C. jejuni as oral administration of the B subunit of CT and LT toxins resulted in modification of the gut microbiota in chickens (Patry et al. 2019). However, it is important to note that the GM1 binding and the impact on bacterial competition is independent of the NAD-utilizing activity of these toxins. In contrast, a new family of ADP-ribosyltransferase toxins has been described displaying a role in interbacterial competition in a NAD-dependent manner. Ting et al. have highlighted the implication of the toxin Tre1 (type VI secretion ADP-ribosyl transferase effector 1) from Serratia proteamaculans in the targeting of the cell division protein FtsZ leading to cytoskeletal modifications and finally to death of the competing bacterial cell (Ting et al. 2018).

Nonetheless, little is known on the consequences for the host of the NAD⁺ pool decrease caused by ADP-ribosylating activity of bARTs, contrary to NADases. Additional work on a broader group of bacterial pathogens is now essential to better understand how bART modulation of host NAD⁺ levels contributes to pathogenesis.

NADases

Some bacterial toxins that display NAD⁺‐glycohydrolase activity can lead to the killing of host cells by NAD⁺ depletion and are often called NADases. This family of toxins hydrolyze NAD⁺ at much higher rates and without requiring ADP-ribose transfer. This process usually is irreversible, and inhibition of the NADase activity involves binding to an immunity factor. The list of bacterial NADase toxins has been growing the last few years and includes some very well-characterized toxins as the S. pyogenes SPN toxin (Ghosh et al. 2010; Smith et al. 2011), M. tuberculosis TNT toxin (Sun et al. 2015), P. aeruginosa Tse6 toxin (Whitney et al. 2015) and PFL_6209 from Pseudomonas protegens (Tang et al. 2018). More recently, two new families have been described due to some structural and enzymatic differences, namely the Arginine (R) Glutamate (E) Serine (S) or RES domain, and the TIR domain-containing toxins (Essuman et al. 2018; Skjerning et al. 2019). All NADases described in this review are listed in Table 2.

Table 2.

Bacterial NADases and their implication in human pathogenesis and interbacterial competition.

SPN and TNT toxicity in the eukaryotic host: two textbook examples

Some NADases have been extensively studied in the context of eukaryotic host pathogenicity. One such example is the NADase SPN from S. pyogenes or group A Streptococcus (GAS), also known as NADase group A (Nga), which is secreted with streptolysin O (SLO) that facilitates translocation of SPN into target cells by forming membrane pores (Madden, Ruiz and Caparon 2001) (Fig. 3).

Figure 3.

Model of SPN and SLO secretion inducing cell death of macrophages infected by S. pyogenes (GAS). SPN entry into host cell can occur through two distinct mechanisms. (1A) S. pyogenes or GAS may be phagocytosed or (1B) GAS can secrete SPN into the extracellular milieu. (2) SPN and SLO are translocated from the bacterium through a cytolysin-mediated translocation (CMT) pathway (Madden, Ruiz and Caparon 2001). SLO is a cholesterol-dependent cytolysin secreted by the Sec pathway forming a pore into the phago(lyso)some membrane, enabling SPN to gain access to the macrophage cytosol. It has been hypothesized that SPN and SLO navigate the bacterial membrane through a T3SS-like protective channel (Magassa et al. 2010). However, whether GAS-containing phagosomes fuse with lysosomes remains controversial (Bastiat-Sempe et al. 2014; Chandrasekaran and Caparon 2015). (3) SLO damages the phagosome membrane preventing its acidification (Hsieh et al. 2018) and SPN release triggers depletion of energy stores, thereby inhibiting cellular repair of SLO-mediated membrane damage. The pore-forming activity of SLO also induces activation of PARP-1, which leads to a decrease of NAD⁺, ATP and PARylation (PAR) (Chandrasekaran and Caparon 2015). (4) Altogether, these events trigger the release of the high mobility group box-1 (HMGB1) from the nucleus to the macrophage cytosol. Ultimately, HMGB1, also known as 'alarmin', is released into the extracellular milieu to interact with TLR2, 4 and RAGE receptors on phagocytic cells. This results in the induction of inflammation through the translocation of NF-κB into the nucleus. (5) Furthermore, an extracellular and functional NADase activity of SPN has been shown to inactivate the P2×7 receptor (P2×7R) disturbing the signaling cascade that normally results in IL-1β interleukin release. Thus, P2×7R inactivation prevents caspase-1 activation of the NLRP3 inflammasome blocking the production of mature IL-1β resulting in inhibition of this host defense mechanism (Westerlund et al. 2019). Finally, the combination of all these events enables GAS survival and replication in macrophages. Note that the nonsecreted immunity protein IFS protects GAS from self-poisoning (Smith et al. 2011). Each black line represents a step of the process. Dotted line shows a trafficking or a translocation event. IM: bacterial inner membrane, PDG: peptidoglycan, PM: plasma membrane, NAM: nicotinamide, cADPR: cyclic ADP-ribose.

To date, SPN is the only bacterial NADase reported to have glycohydrolase, ADP-ribosyltransferase and ADP-ribosyl cyclase activities producing either nicotinamide, ADP-ribose and cyclic ADP-ribose (cADPR), respectively. Such a property has only been described for eukaryotic proteins such as SARM-1 (Essuman et al. 2018). However, in vitro studies using recombinant SPN suggest that it functions exclusively as a NAD⁺‐glycohydrolase (Ghosh et al. 2010). Indeed, SPN displays a functional ARTT motif containing the catalytically conserved glutamate residue (Ghosh et al. 2010). Ghosh et al. suggested that NAD⁺‐glycohydrolases may share a common ancestry with ADP-ribosyltransferases. However, SPN does not display any other conserved motifs normally found in the active sites of bARTs, such as the STS or the YX₁₀Y motif. SPN is an atypical NADase containing some bART motifs that could account for ribosyltransferase activity but also a particular catalytic domain and additional structural elements essential for glycohydrolase activity and immunity binding (see the section 'Immunity-like proteins').

The activity of SPN results in a substantial reduction of cellular NAD⁺ and ATP levels, in essence depleting the host cell energy supply. This not only leads to growth arrest and cell death (Michos et al. 2006) but has also been shown to help bypass the host immune system. Several mechanisms have been described in the past few years, including: (i) enhancement of GAS extracellular survival (Bricker et al. 2002), (ii) promotion of macrophages apoptosis (Timmer et al. 2009; Chandrasekaran and Caparon 2015), (iii) modification of several host NAD⁺‐dependent enzymes including PARP-1 (Chandrasekaran and Caparon 2015), (iv) prevention of phagolysosome acidification (Bastiat-Sempe et al. 2014; Hsieh et al. 2018) and autophagy killing (Sharma et al. 2016; Hsieh et al. 2018) and (v) inhibition of the bactericidal actions of neutrophils and leukocytes (Zhu et al. 2015a, 2017). The main NAD⁺‐dependent effects of SPN are summarized in Fig. 3. The SPN toxin has also been shown to contribute to extracellular modulation of the inflammasome, independently of SLO (Hancz et al. 2017). Recently, Westerlund et al. showed that the expression of a functional SPN allows the bacterium to avoid the activation of the P2×7 receptor leading to a decrease of secretion of mature IL-1β cytokine (Westerlund et al. 2019). Indeed, P2×7R mediates NLRP3 inflammasome activation enabling pro-IL-1β to be processed into an active form (Di Virgilio et al. 2017). Curiously, it has been demonstrated that P2×7 activation promotes the elimination of other intracellular pathogens such as Chlamydiasp. (Coutinho-Silva et al. 2003).

All these toxic phenotypes driven by NAD⁺‐consuming activity of SPN contribute to an increase in GAS virulence, resulting in chronic infections due to treatment failure. However, it remains unknown to what extent each enzymatic activity exhibited by SPN influences these phenotypes or whether they act synergistically. Interestingly, Toh et al. have recently shown that SPN inhibits the caveolin1-mediated internalization of GAS into epithelial cells in a NADase-independent manner but dependent on SLO. Furthermore, subtypes of SPN with inactive enzymatic activity found in clinical isolates still display cytotoxicity suggesting this toxin may have NADase-independent functions (Toh et al. 2019).

Another NADase has been identified in the Streptococcus genus. Whitney et al. recently characterized the toxin exported by the Esx secretion system with LXG domain B (TelB) as an interbacterial NADase secreted by the human commensal and opportunistic pathogen Streptococcus intermedius isolated from patients with abscesses (Whitney et al. 2017). TelB shares no homology to the SPN nor Tne1 nor Tne2 families. Instead it is part of a previously described group of NADases, known as DUF4237 (Tang et al. 2018), that includes the well-studied tuberculosis necrotizing toxin (TNT) from M. tuberculosis (Sun et al. 2015).

Unlike some bacterial pathogens, M. tuberculosis does not rely on classical toxins for pathogenesis. Indeed, resistance to macrophage killing is a key feature of M. tuberculosis pathogenesis (Russell, Barry and Flynn 2010). Pajuelo et al. highlighted that NAD⁺ hydrolysis by TNT is a primary toxicity mechanism in infected macrophages as the catalytically inactive mutant shows complete loss of toxicity (Pajuelo et al. 2018). Once released into the macrophage cytosol, TNT induces NAD⁺ depletion leading to macrophage death (Pajuelo et al. 2018). The details on the secretion and function of TNT are summarized in Fig. 4.

Figure 4.

Model of CpnT secretion and TNT-induced macrophage cell death following infection by M. tuberculosis (Mtb). (1) Mycobacterium tuberculosis (Mtb) enters macrophages by phagocytosis and then (2), once established in the phagosome, Mtb secretes Esat-6 and CFP10 in an ESX-1-dependent manner to perforate the phagosomal membrane allowing Mtb to exit the phagosome and TNT to translocate to the cytosol and preventing phagosome maturation and membrane repair. (3) Secretion of CpnT/TNT is not fully understood. CpnT translocation requires the membrane-spanning complex EsxEF and ESX systems to reach the bacterial outer membrane and expose TNT at the surface (Izquierdo Lafuente et al. 2021; Tak, Dokland and Niederweis 2021). However, how TNT is released from CpnT into the macrophage cytosol remains to be discovered. Note that when CpnT stays in the Mtb cytoplasm, the protein is either degraded or bound to its immunity protein IFT to protect Mtb from self-damage (not shown in the figure). (4) Once in the macrophage cytosol, TNT cleaves NAD⁺ to form NAM and ADPr. Thus, the NAD⁺ level drops, leading to macrophage death via three major pathways (Pajuelo et al. 2018). (5) By sensing either high level of NAM and NADr or low level of NAD⁺, the receptor-interacting serine-threonine kinase 3 (RIPK3) phosphorylates the mixed-lineage kinase domain-like protein (MLKL), which induces necrosis by extensive formation of pores on the mitochondrial and plasma membranes. (6) Bcl-xL is also recruited by RIPK3 and both accumulate on mitochondria preventing caspase-8 activation. Instead, they induce the mitochondrial permeability transition (MPT) pore resulting in loss of mitochondrial potential (Δψ) and mitochondrial depolarization. (7) It is known that the NAD⁺/NADH ratio is much lower in mitochondria compared with the cytosol and its maintenance is crucial for cell survival through the control of the overall cellular redox state. Thus, depletion of the cytosolic NAD⁺ pool disturbs mitochondrial energy metabolism impairing ATP synthesis and the electron transport chain (ECT). (8) Cytosolic Mtb has also been shown to take up NAM, which is then integrated in the NAD biosynthesis pathway through three nicotinate phosphoribosyltransferase enzymes (Boshoff et al. 2008). Altogether, these events enable Mtb survival and replication in macrophages. IM: inner membrane; AGPG: arabinogalactan and peptidoglycan; OM: outer membrane; NAM: nicotinamide; ADPr: ADP-ribose; ATPS: ATP synthase.

Members of the TNT family are widely distributed in pathogenic bacteria and in fungi. In contrast to SPN and other bARTs, TNT is an effective NAD(P)⁺ hydrolase due to distinct structural properties (Tak et al. 2019). Structural alignment of TNT with SPN and DT toxins highlighted strong architectural similarities in the NAD⁺‐binding core but TNT lacks the ARTT motif implying a distinct catalytic mechanism (Sun et al. 2015). A potential NAD⁺‐binding pocket was identified, containing four residues, including a highly conserved glutamine (Tak et al. 2019). Thus, TNT family members represent a distinct NADase family especially due to the presence of a four glycine-formed cradle important for TNT stability and three conserved residues (Phe, Arg, Arg) essential for hydrolysis. Although TNT share a similar catalytic activity with NADases like SPN, TNT is much smaller (150 versus 450 residues) as the protein displays the smallest NAD⁺‐binding site described yet maybe to facilitate its secretion, as hypothesized by Tak et al. (2019) (see the section 'Secretion mechanisms associated with Gram-positive bacteria' for further discussion on the secretion mechanisms).

Bacterial TIR domain-containing toxins: an atypical NADase family

Whereas degrading NAD⁺ in the infected host seems to be collateral damage exerted by bARTs, depleting the NAD⁺ pool level is the prime goal of NADases to disturb the host. Recently, bacterial TIR-containing proteins (Tcps) have emerged as a new family of NADases.

The TIR domain, present in all TLRs and associated adaptors as well as IL-1 receptor, is a critical component of the innate immune system (Akira, Uematsu and Takeuchi 2006). After pathogen detection by TLRs recognizing pathogen-associated molecular patterns (PAMPs), TIR domain interactions will initiate and trigger propagation of TLR signaling cascade and activation of an appropriate immune response. The TIR domain is composed of 125–200 residues with a conserved secondary structure of five stranded parallel β-sheets from βA to βE and five α-helices, from αA to αE. These are connected by eight surface-exposed loops named with the letters corresponding to the connecting elements, such as AA, AB, BB, until EE (Xu et al. 2000). The surface-exposed loops are essential for homotypic TIR domain interactions in TLR/IL-1R signaling. Analysis of TIR domains from TLRs and bacterial Tcps shows they share a conserved core structure with variable surface loops (Xu et al. 2000; Jang and Park 2014). In the case of bacterial Tcps, although they share a structurally conserved TIR domain (either C-terminal or N-terminal positioned), the rest of the protein (about half) is highly variable in terms of structure. Amino acid sequence alignment of multiple bacterial Tcps showed high conservation of the WxxxE motif in bacterial Tcps (Essuman et al. 2018; Coronas-Serna et al. 2020) as well as eukaryotic TLR proteins and TIR adaptor protein SARM. The conserved glutamate residue of this motif is implicated in the NADase activity of SARM1 and the bacterial TcpC from uropathogenic E. coli, TirS from Staphylococcus aureus, BtpA and BtpB from Brucella abortus. Recently, Bratkowski et al. have highlighted how the structure of SARM1 influences its NAD⁺ hydrolase activity through TIR domain dynamics (Bratkowski et al. 2020). SARM1 is produced under an active or an autoinhibited form and the latter contains an ARM domain that locks the TIR domain. ARM domain mutation or increased NMN concentration induces a conformational change leading to dimerization of the TIR domains. This results in a switch of TIR domains to an active conformation, triggering the NAD⁺ hydrolysis.

In the past years, many groups have been revealing the role of a family of bacterial effectors containing a TIR domain that hijack early signaling events to evade the eukaryotic immune system. These bacterial Tcps have been shown to mediate interactions directly with TLRs or specific adaptors and consequently perturb these signaling cascades and down-modulate inflammatory response during infection. However, Tcps have also been implicated in additional functions. TcpC from uropathogenic E. coli, first identified to block TLR4 and aggravate pyelonephritis (Cirl et al. 2008), was also shown to target NACHT leucin-rich repeat PYD protein 3 (NLRP3) to control the inflammasome and caspase-1 (Waldhuber et al. 2016). Two Tcps from Brucella abortus, named BtpA and BtpB, modulate microtubules dynamics and activate the unfolded protein response, in addition to their role in the down-modulation of dendritic cell maturation (Salcedo et al. 2008; Smith et al. 2013; Felix et al. 2014). In the context of nosocomial pathogens, the Tcp PumA was shown to be a key virulence factor in P. aeruginosa PA7, able to target specific TLR adaptors and also ubiquitin-associated protein 1 (UBAP1), a component of the endosomal-sorting complex for transport I (ESCRT-I), resulting in inhibition of cytokine receptor trafficking (Imbert et al. 2017).

As mentioned previously, while pursuing their work on axon degeneration, Essuman et al. discovered the TIR-containing adaptor protein SARM1 cleaves NAD⁺ to generate axon destruction upon injury (Essuman et al. 2018). They showed the TIR domain of several Tcps from both bacteria and archaea exhibits an NAD(P)⁺‐glycohydrolase activity in vitro by producing nicotinamide, ADP-ribose and even cADPR in some cases. Moreover, they showed that transfection of the full-length TIR domain from the Tcp TirS from the nosocomial pathogen S. aureus induces NAD⁺ loss in mammalian cells.

Consistent with these findings, recent work showed that BtpA and BtpB from B. abortus deplete NAD⁺ in infected mammalian epithelial cells and macrophages (Coronas-Serna et al. 2020). It is possible that many of the phenotypes observed in host cells exerted by Tcps of other bacterial pathogens may be dependent on the NADase activity conferred by their TIR domains but this hypothesis remains to be tested. In that sense, Essuman et al. postulated that the major function of Tcps is to modulate metabolic and bioenergetic pathways through alteration of NAD⁺ level (Essuman et al. 2018).

Further work is now necessary to better understand the extent to which the modulation of host NAD⁺ homeostasis by TIR domains during bacterial infection directly affects TLR signaling and the immune response. We could hypothesize that, besides their function in hijacking the first steps of the immune response, TIR domain-containing effectors might prevent NF-κB translocation through its glycohydrolase activity, which converts NAD⁺ to NAM. Indeed, Crowley et al. have highlighted NAM's capacity to decrease intracellular levels of NF-κB and the endoplasmic reticulum chaperone GRP78/BiP (Crowley et al. 2000). In the case of BtpA, its impact on NAM could affect the pool of GRP78 and lead to initiation of the unfolded protein response previously described (Lee 2005). Furthermore, NAD⁺ also plays a role in controlling the actin cytoskeleton (Venter et al. 2014). However, it is unclear how BtpA depletion of NAD⁺ would help to stabilize microtubules, a proposed role for this effector based on transfection experiments (Felix et al. 2014), given that microtubule depolymerization is blocked by increasing intracellular NAD⁺ levels (Harkcom et al. 2014).

The presence of TIR domains in archaea and nonpathogenic bacteria as plant symbionts and human gut commensals (Zhang et al. 2011) raises the question that they may also participate in bacterial population dynamics. Furthermore, Tcps may have nonvirulence-related roles such as regulation of metabolic pathways as supposed for SPN, because they produce ADPR and cADPR, two second messengers controlling calcium homeostasis. It is still unclear how the enzymatic activity of Tcps in bacteria is regulated or turned-off as for other NADases. A specific immunity mechanism could be conferring protection, as Essuman et al. showed that purified TirS is enzymatically active (Essuman et al. 2018). Although TIR domain-containing NADases TirS, TcpC, BtpA and BtpB from S. aureus, E. coli and B. abortus, respectively, share no structural similarities to the other previously described NADase families, they present structural similarities with nucleotide enzymes such as cytidine monophosphate (CMP) hydrolase and nucleoside 2-deoxyribosyltransferase (Essuman et al. 2018). However, even if the number of Tcps confirmed to have NADase activity in the context of bacterial pathogenesis is still small, it is possible this is a conserved function for this family of bacterial effectors.

The Tne families

Some NADases have not been described for their toxicity to the eukaryotic host but rather in a context of bacterial competition. Based on their divergent structures and secretion pathways, two NADase effector families have been recognized named type VI secretion NADase effector families 1 and 2 (Tne1 and Tne2). The Tne1 family is present in the Gram-negative phyla Proteobacteria and Bacteroidetes comprising its archetypal member Tse6, a T6SS-dependent effector. The Tne2 family includes NADases predicted to be exported by both Gram-negative and Gram-positive bacteria through T6SS and T7SS, respectively (Tang et al. 2018).

Tse6 was the first NAD⁺‐utilizing effector described as a NADase (Whitney et al. 2015) and then Tang et al. identified PFL_6209 as a Tne2 family member (Tang et al. 2018). Both Tse6 and Tne2-NADases catalyze the hydrolytic removal of the nicotinamide moiety of NAD⁺ and NADP⁺. This hydrolysis has been described both in vitro and in the bacteria and results in growth arrest rather than death (Whitney et al. 2015; Tang et al. 2018).

Structural analysis indicated that the closest structural homologs of the cytotoxic region of Tse6 (Tse6_CT) are the catalytic domains of bARTs (Whitney et al. 2015). However, residues of the NAD⁺ binding pocket of Tse6 differ from those of bARTs including the presence of a nonacidic residue (Gln) important for interaction with its immunity protein suggesting its relevance in the enzymatic activity (Whitney et al. 2015). Tse6 orthologous proteins share a conserved acidic residue (Asp) positioned similarly to the conserved glutamic acid found in bARTs, and that is essential for Tse6 NADase activity (Whitney et al. 2015). According to the authors, Tne1 and Tne2 are two distinct families of NADases because despite structural similarities, they exhibit some differences especially in their NAD⁺‐binding pocket (Whitney et al. 2015; Tang et al. 2018). Compared to the Tne1 family, the NAD⁺‐binding pocket of Tne2 comprises four conserved residues (Phe, Arg, Lys and Gln) while the catalytic residue has not been characterized yet (Tang et al. 2018). Also, the immunity proteins, Tsi6 and Tni2, interact with their cognate effector in a specific manner that further highlights differences between the two toxins. In summary, structural analyses on Tne-NADases suggest they have the fold of bARTs enzymes but with unique binding and catalytic motifs essentials for their NADase activity. At this stage, we cannot exclude the possibility that Tne-NADases may also exert toxicity in eukaryotic and even archaeal cells. Indeed, other Pseudomonas effectors have already been shown to be translocated into both bacterial and eukaryotic cells via the T6SS (Jiang et al. 2016).

As each Tne-NADase has a specific Tni immunity protein (Tang et al. 2018), Tne-NADases are certainly promoting environment colonization by limiting the growth of surrounding bacteria that do not express the same Tne-Tni effector–immunity pair. Moreover, the exchange of bacteriostatic toxins within a population could promote the formation of persister bacterial cells known as a source of many recurrent infections, notably affecting patients in intensive care. Such diseases represent a considerable challenge for treatment because most antibiotics are ineffective against nongrowing persister cells (Kim et al. 2018).

NAD⁺‐phosphorylase toxins

Most NADases are associated with an immunity protein to which they bind, to prevent NAD⁺ access and protect bacteria from self-poisoning. This has been shown for SPN, TNT, Tne-NADases and also for the bART Tre1. This strategy resembles the toxin–antitoxin (TA) systems (Harms et al. 2018). Four major types of TA modules (type I–IV) have been reported, depending on the nature of the antitoxin and TA interactions. TA modules consist of a toxin and a cognate antitoxin. In most cases, they are encoded in an operon that responds to environmental stimuli to induce dormancy, persistence and biofilm formation under hostile conditions (Wen, Behiels and Devreese 2014). These systems are also directly involved in bacterial pathogenesis of, for example, E. coli or M. tuberculosis (Ramage, Connolly and Cox 2009; Norton and Mulvey 2012).

Two recent studies have described for the first-time toxins belonging to TA modules that are NAD⁺‐phosphorylase enzymes. Skjerning et al. demonstrated that the RES domain of RES-xenobiotic response element (RES-Xre) TA module degrades NAD⁺, arresting bacterial cell growth (Skjerning et al. 2019). RES domain homologs are present in both Gram-positive and Gram-negative bacteria. Despite structural similarities with ADP-ribosyltransferase toxins, the RES domain has some specificities (see the section 'Immunity-like proteins') that suggest it may function as a NAD⁺‐degrading toxin rather than an ADP-rybosyltransferase. An additional RES-Xre TA system was described in M. tuberculosis, called mycobacterial cidal toxin/antitoxin (MbcTA) (Freire et al. 2019). MbcT catalyzes the degradation of NAD⁺ through a NAD⁺‐phosphorylase activity. The MbcA antitoxin is required for M. tuberculosis survival in human macrophages and to promote virulence in the mouse model of infection. Recently, it has been demonstrated that the mbcTA operon was expressed under multi-stress conditions reflecting those likely found inside macrophages (Ariyachaokun et al. 2020).

The closest structural relatives to the RES domain-containing toxins such as MbcT are the bART family although without an ARTT motif (Freire et al. 2019; Skjerning et al. 2019). Residues Arg27, Tyr28 and Tyr58 have been predicted to form the NAD⁺‐binding site of MbcT. Also, the arginine residue Arg47 within the RES motif has been revealed to be essential for the NAD⁺‐phosphorylase activity of MbcT suggesting a similar role of the RES domain in all known RES-Xre TA systems. These observations highlight that MbcT-like toxins belong to a particular NADase family but further investigations are needed to fully decrypt the molecular mechanisms involved. Interestingly, structural patterns and activity of RES domain-containing toxins suggest a unique mechanism among TA systems and a range of functions much broader than first expected. This NAD⁺‐phosphorylase activity does not seem to be restricted to RES domain-containing toxins. Shidore et al. have indeed revealed that the avirulence factor AvrRxo1 from Xanthomonassp. phosphorylates NAD⁺in planta and functions as a type III secretion effector and a TA system (Shidore et al. 2017). Recognition depends on the presence of a pair of matching genes, an avirulent (avr) gene in the pathogen and a resistance (R) gene of the plant. Thus, in several plants expressing the resistance protein Rxo1, AvrRxo1 triggers an hypersensitive resistance response (HR) and it has also been shown to promote virulence of Xanthomonas oryzae, a major rice pathogen (Zhao et al. 2004; Han et al. 2015).

NAD⁺ consumption and regulation

Substrate specificity and feedback

Bacterial glycohydrolases can cleave both NAD⁺ and NADP⁺, unlike ADP-ribosyltransferases that cleave only NAD⁺. However, some specificity has been described. Whereas SPN is significantly less efficient hydrolyzing NADP⁺ compared to NAD^+, nuclear magnetic resonance studies showed NADP⁺ is the preferred substrate for TNT (Tak et al. 2019). Furthermore, their rate of hydrolysis is very distinct, with NADases rapidly cleaving NAD⁺, in contrast to bARTs that have a slow turnover (Ghosh et al. 2010; Whitney et al. 2015). However, NAD⁺ consumption rates can show some variability within the same family (Ghosh et al. 2010; Essuman et al. 2018).

It is important to note that often, regulatory feedback mechanisms are present to control enzymatic activities. For example, in the case of bARTs, its activity can be inhibited by ADP-ribose, one of the key reaction products. This is also the case for the NADase SPN, inhibited by ADP-ribose (Ghosh et al. 2010). In contrast, the production of ADP-ribose or NAM by TNT does not influence its activity (Tak et al. 2019).

NAD⁺ pathway metabolites have been shown to play a role in eukaryotic NADase regulation. NAM, the reaction product of NAD hydrolysis, inhibit the enzymatic activity of SARM1 (Bratkowski et al. 2020) but also of NAD⁺‐utilizing enzymes such as PARPS and sirtuins (Avalos, Bever and Wolberger 2005; Gibson and Kraus 2012; Essuman et al. 2017). Nicotinamide mononucleotide (NMN) enhances the NAD hydrolase activity of SARM1, illustrating the regulatory influence of NAD⁺ pathway metabolites on the function of SARM1, particularly important in axon degeneration (Bratkowski et al. 2020).

ADP-ribosylation, a reversible modification

Many factors affect the turnover of ADP-ribosylation. In particular, it has been shown that ADP-ribosylhydrolase (ARH) enzymes can catalyze ADP-ribose removal (Cohen and Chang 2018). Recently, Tri1 was shown to protect bacteria from self-intoxication by the bART Tre1 (Ting et al. 2018). Tri1 is able to reverse the ART-catalyzed modification via its ADP-ribosylhydrolase (ARH) activity. This is the first immunity-like protein reported among the bART family as well as the first reported ADP-ribosyltransferase toxin with a role in interbacterial competition. Bioinformatic analyses suggest this kind of ART/ARH effector/immunity pair is widespread and is likely to constitute an important family of effectors in bacterial competition warfare.

Although immunity proteins or antitoxins allow for the shut-down of NAD⁺‐degrading activities of most known NADases, a different mechanism applies to control of bART ADP-ribosyltransferase activity. In some cases, a feedback loop is initiated for negative regulation of the enzymatic activity, as for the ExoS from Pseudomonas and the mammalian protein CD38, where ADP-ribose alone or coupled with nicotinamide inhibits catalysis (Riese et al. 2002; Ghosh et al. 2010).

Immunity-like proteins

The presence of cognate immunity-like proteins to each characterized NADase differentiates this family from other NAD⁺‐utilizing toxins. Indeed, most bART producing bacteria do not express any cognate immunity-like protein (except for Tre1 and AvrRxo1, as described above). The core binding sites of bARTs and NADases display major structural differences, allowing bARTs to accommodate target proteins and NADases to bind an immunity protein. However, no immunity-like proteins belonging to the TIR domain-containing NADases have yet been identified.

Although SPN presents an atypical ARTT motif on its active site as other bARTs, the toxin does not accommodate protein substrates. This is also the case for the NADases TNT and MbcT. Indeed, the main difference between bARTs and NADases is that bARTs directly transfer an ADP-ribose from NAD⁺ onto a target protein. Structural analysis from several bARTs have shown that the active site has to be exposed to allow access to NAD⁺ by the toxin. On the contrary, NADases expressing a NAD⁺‐glycohydrolase or NAD⁺‐phosphorylase activity such as SPN or MbcT, respectively, exhibit a helical linker region near the NAD⁺‐binding site that impedes binding of target proteins (Smith et al. 2011; Skjerning et al. 2019). Amino acid residues 77 to 100 within the RES domain form a helical extension placed in a similar location to the SPN linker. This strongly strengthens the hypothesis that RES domain-containing toxins may be NADases rather than bARTs (Skjerning et al. 2019). Consistently, the active site of Tse6 shows limited accessibility to potential acceptor proteins suggesting a similar mechanism is involved in preventing the binding of target molecules (Whitney et al. 2015). The presence of this type of linker may instead contribute to the binding of the corresponding immunity or antitoxin protein typical of the NADase family as demonstrated for the immunity factor of SPN (IFS), and supposed for the antitoxin of MbcT (MbcA) (Smith et al. 2011; Freire et al. 2019; Skjerning et al. 2019).

As mentioned above, immunity factors block access of NAD⁺ by binding to the NAD⁺‐binding pocket of the toxin. In this way, the toxin's active site and the corresponding immunity factor form a stable complex. This is nicely illustrated for IFT, the immunity factor of TNT, that functions as a competitive inhibitor of the NAD⁺ substrate for which the structure of the TNT–IFT complex has been solved (Sun et al. 2015). As expected, the amino acid residues involved in NAD⁺ binding and hydrolysis are also essential for the toxin–immunity interaction. The level of specificity of each immunity and toxin pair is very high. For example, Tsi6 (Tni1) and Tni2 are highly divergent at the amino acid sequence and structural levels and only neutralize their specific cognate NADase toxins (Tang et al. 2018). This ensures the environmental cohabitation of bacterial species harboring a compatible immunity or antitoxin and efficient exclusion of other bacteria.

NAD⁺ antiviral defence systems

Interestingly, NAD⁺ degradation has been proven to be beneficial for bacteria in the context of antiphage protection (Ka et al. 2020). The bacterial antiviral Thoeris defense system relies on two genes, called thsA and thsB. Interestingly, TshA contains a sirtuin-like domain, while ThsB contains a Toll/interleukin-1 receptor (TIR) domain. In several recent studies, sirtuins and TIR-containing proteins exhibited NAD⁺‐cleavage activities (discussed in greater detail in the section ‘Bacterial TIR domain-containing toxins: an atypical NADase family') suggesting that the Thoeris system could be related to NAD⁺ modulation. Ka et al. showed that only the TshA was able to cleave NAD⁺, and that NAD⁺‐binding and/or cleavage was essential for the antiphage function of the system. However, it remains unclear how NAD⁺‐cleavage triggers a defence mechanism as the expression of TshA with or without TshB in E. coli was not toxic suggesting this system does not result in direct intracellular NAD depletion (Ka et al. 2020).

Very recently, Morehouse et al. have highlighted the contribution of the cGAS-STING pathway in bacteria to protect them from viral attacks (Morehouse et al. 2020). Stimulator of interferon genes (STING)-containing proteins are cytosolic sensors used by mammalian cells to induce immunity pathways during microbial infections. In contrast to the TIR domain, phylogenetic reconstructions indicate that STING domain has a bacterial origin. Interestingly, most STING proteins are found fused to a TIR domain in bacteria and these STING-TIR proteins constitute a bacterial immune defence mechanism against viral infections (Morehouse et al. 2020). Indeed, following an unknown viral cue, the cGAS proteins produce the second cGAMP messenger (cyclic-GMP-AMP) that binds to the STING-TIR proteins resulting in assembly of long filaments. This configuration activates the TIR domain-enzymatic activity leading to the degradation of NAD⁺. The latter may represent a suicide strategy for bacteria to prevent virus spreading inside the population (Morehouse et al. 2020).

SECRETION PATHWAYS FOR DELIVERY OF BACTERIAL NAD⁺‐UTILIZING TOXINS

The translocation of effector proteins into target cells is a crucial strategy used by bacterial pathogens to promote virulence. Delivery of bacterial toxins requires a secretion mechanism with a dedicated translocator. These complex machinery called secretion systems can be classified as type I–IX secretion systems (T1-9SS) (Rapisarda et al. 2018). Mycobacteria, which have a particular cell envelope more closely resembling Gram-negative bacteria, encode a T7SS, which has not been observed in the Gram-negative phylum. The secretion of effectors across the bacterial membranes involves either a one-step (T1SS, T3SS, T4SS and T6SS) or two-step (T2SS, T5SS) secretion mechanisms. Substrates of either the T3SS, T4SS or T6SS gain access directly to the cytoplasm of the target host.

Secretion mechanisms associated with Gram-negative bacteria

The T6SS, one of the most recently identified, is remarkably well characterized. It is well established that effectors can become associated with secreted structural components of the T6SS apparatus. Small effectors with a single domain interact with the hemolysin coregulated protein (Hcp) whereas large effectors containing multiple domains interact with the valine-glycine repeat protein G (VgrG) family (Cianfanelli, Monlezun and Coulthurst 2016). Hcp and VgrG form, respectively, the inner tube and the spike of the T6SS structure. Most effectors interacting with VgrGs contain a conserved N-terminal adaptor domain recognized by VrgG such as the Pro-Ala-Ala-Arg (PAAR) domain present in Tne-NADases and also in the ADP-ribosyltransferase Tre1 (Tang et al. 2018; Ting et al. 2018).

For some bARTs, two additional secretion systems have been reported, namely the T2SS and the T3SS. The ability to secrete the CT toxin to the extracellular environment is essential for V. cholerae virulence (Sandkvist, Bagdasarian and Howard 2000; Fieldhouse et al. 2010). Secretion of CT occurs via a T2SS, that first transports the toxin into the periplasm through the Sec pathway where the N-terminal signal peptide is removed. This is followed by translocation across the outer membrane by a dedicated Extracellular Protein Secretion (Eps) machinery. Noticeably, the T2SS is called Eps in V. cholerae but general secretion pathway (Gsp) in enterotoxigenic E. coli, and Xcp in P. aeruginosa (Costa et al. 2015).

Effectors that are translocated through the T3SS are targeted to the secretion machine by a set of secretion signals, in many cases encoded within the first 100 residues of the protein as described for ExoS and ExoT (Barbieri 2000). These secretion signals are characterized by low frequency of charged and hydrophobic residues and an enrichment in serine and threonine. They ensure specificity of the binding site for specific chaperon proteins preventing nonspecific interactions from taking place. In addition, these signals also maintain the substrate in a stable conformation suitable for secretion (Galán et al. 2014). T3SS secretion delivery has been described for some bARTs including ExoS and ExoT from P. aeruginosa (Yahr et al. 1997), AexT from Aeromonas salmonicida (Braun et al. 2002) and SpvB from Salmonellasp. (Cheng and Wiedmann 2019).

Unlike the T3SS and T6SS, the T4SS is capable of transferring both DNA and protein molecules into target cells. The process of substrate secretion through the T4SS apparatus is still under investigation, with important advances being made on the structural features of this system (Grohmann et al. 2018). Translocation of both BtpA and BtpB from B. abortus fused to the CyA reporter system suggest a VirB T4SS-dependent secretion mechanism (Salcedo et al. 2013). To date, this is the only example of a secretion system implicated in the injection of TIR-containing NADases. In addition, the T4SS has been reported to secrete the PT toxin (Weiss, Johnson and Burns 1993).

OMVs are recognized as an alternative secretory system in both Gram-negative and Gram-positive bacteria (Brown et al. 2015; Toyofuku, Nomura and Eberl 2019). OMVs consist of spherical particles ranging in size from 20 to 500 nm in diameter, including an outer leaflet composed of lipopolysaccharides and an inner bilayer of phospholipids, which is derived essentially from the outer membrane of Gram-negative bacteria.

Although a NADase activity has not yet been demonstrated for the Tcp TirE from E. faecium, an OMV-dependent delivery has been proposed for this effector (Wagner et al. 2018), suggesting this may be an important mechanism to take into account for delivery of bacterial NADases.

Secretion mechanisms associated with Gram-positive bacteria

Cytolysin-mediated translocation (CMT) is a secretion apparatus in Gram-positive organisms functionally resembling the T3SS of Gram-negative bacteria (Madden, Ruiz and Caparon 2001). CMT has been observed in the pathogen S. pyogenes, which injects at least one virulence factor, the NADase SPN, into eukaryotic cells. The secretion of SPN is also facilitated by another toxin, SLO, which is a cholesterol-dependent cytolysin that binds cholesterol on the surface of eukaryotic cells, inserts into their membranes and leads to pore formation (Madden, Ruiz and Caparon 2001; Tweten 2005). As illustrated in Fig. 3, following pore creation by SLO, SPN is translocated across the bacterial inner membrane by the Sec pathway. Curiously, SPN secretion does not seem to result from simple diffusion across the pore established by SLO. Instead, it has been hypothesized that a protected channel enables direct transfer between the bacterium and the eukaryotic plasma membrane, reminiscent of the T3SS (Madden, Ruiz and Caparon 2001). It is important to note that some studies suggest that SPN translocation can occur independently of cholesterol and SLO. However, SLO is required for SPN-mediated cytotoxicity, suggesting a synergistic action (Magassa, Chandrasekaran and Caparon 2010).

The T7SS is specific to Gram-positive bacteria and Mycobacterium (Costa et al. 2015; Ates, Houben and Bitter 2016). The first T7SS identified was designated as ESX-1, encoded by esx operons, with four additional T7SSs since then described, ESX-1 to 5, all with distinct roles in the bacteria (Stanley et al. 2003).

The Tne2 family and the NAD⁺‐glycohydrolase TelB from S. intermedius, both sharing an N-terminal LXG domain rely on the T7SS for delivery (Whitney et al. 2017; Tang et al. 2018). However, their secretion partners (such as chaperones), folding and molecular interactions remain to be discovered.

Secretion of TNT by M. tuberculosis

The T7SS is capable of translocating effectors across both M. tuberculosis membranes. The M. tuberculosis genome harbors five T7SS clusters (ESX-1 to ESX-7) with distinct physiological functions. ESX-1 is crucial for bacterial virulence and ESX-5 participates in nutrient uptake, host immunomodulation and immune evasion (Izquierdo Lafuente et al. 2021). After inhalation of M. tuberculosis into the lung, and uptake by alveolar macrophages, bacteria evade killing by blocking the fusion of its phagosome with lysosomes (Sun et al. 2010; Wong et al. 2011). This is achieved in part by exporting specific proteins (ESAT-6 and CFP10 also known as EsxA and B, respectively) via the ESX-1 secretion system (Xu et al. 2007) leading also to pore formation on the phagosomal membrane required for M. tuberculosis escape into the macrophage cytosol (Simeone et al. 2012) (Fig. 4). TNT constitutes the C-terminal domain of the outer membrane protein CpnT (channel protein with necrosis-inducing toxin), whereas the N-terminal domain is required for secretion as it contains the secretion motif YxxxE found in other ESX substrates. Indeed, two very recent studies revealed that the secretion of TNT involves the T7SS and other key factors (Izquierdo Lafuente et al. 2021; Tak, Dokland and Niederweis 2021). The cpnT gene is found in an operon upstream the immunity-encoding ift gene and downstream of the esxF and esxE genes. EsxE and EsxF belong to the WXG-containing Esx proteins family including EsxA and B. Using electron microscopy, Tak et al. revealed the pentameric structure of the EsxEF complex forming a central pore and able to bind mycobacterial membranes. Furthermore, authors showed that mutation of the WXG abrogate secretion of CpnT and that EsxE and ExsF allow the translocation of CpnT to the mycobacterial surface (Tak, Dokland and Niederweis 2021). They suggested that EsxEF forms a transmembrane channel perhaps in association with an ESX system. Lafuente et al. also observed the surface localization of CpnT during macrophage infection. In addition to having shown that EsxEF was important for CpnT translocation, they demonstrated that CpnT is an ESX-5 substrate in bacterial cultures but its secretion is dependent on EXS-1, ESX-4 and EXS-5 during macrophage infection (Izquierdo Lafuente et al. 2021). However, there are still unanswered questions regarding the interplay between the different ESX systems involved in CpnT transport and the ExsEF complex, and how TNT is released from CpnT inside the macrophage cytosol. By combining both studies, it is possible that EsxE, EsxF and CpnT use an ESX system to reach the periplasm where the EsxEF complex forms a pore in the outer membrane enabling translocation and thus anchoring of CpnT in the outer membrane. However, further work is required to decipher the molecular mechanism involved and if additional proteins are implicated in the secretion of TNT (Fig. 4).

NAD⁺ METABOLISM-BASED THERAPEUTICS

Enzymes involved in NAD⁺ metabolism have been widely described as potential targets for drug development against many human pathologies including cancer, neurodegenerative diseases, aging, metabolic and autoimmune disorders (Khan et al. 2007; Pankiewicz et al. 2008; Cantó, Menzies and Auwerx 2015; Verdin 2015; Rajman, Chwalek and Sinclair 2018; Singhal and Cheng 2019). Mechanistic pathways in mammals associated with beneficial effects of targeting NAD⁺ metabolism have been well characterized. For example, while degenerated axons have shown a decrease of NAD⁺ levels, cancer cells have been shown to have increased NAD⁺ levels as DNA damage can stimulate NAD⁺ biosynthesis to trigger repair process through boosting of the salvage pathway (Maldi et al. 2013). For example, FK866 is a potent small-molecule inhibitor of human nicotinamide phosphoribosyltransferase (NAPRT) that reduces NAD⁺ levels and can lead to apoptosis in tumor cells with little toxic effect on normal cells (Hasmann and Schemainda 2003).

In contrast, targeting the NAD⁺ network to counteract infectious diseases has been much less explored. Research for new antimicrobial agents is absolutely vital as antibiotics have been increasingly showing their limits due to accumulated resistance rates around the world and the collateral damage on our commensal microbiota. Therefore, alternative solutions are needed to fight against bacterial infections.

Two main approaches have been explored to modulate the host NAD⁺ balance and thus health: (i) activation or inhibition of NAD⁺ biosynthesis through either control of NAD⁺ biosynthetic enzymes or supplementation of NAD⁺ precursors and (ii) inhibition of NAD⁺ degradation.

The targeting of key enzymes involved in the biogenesis of NAD⁺ and NAD cofactors (NADP, NADPH, NADH) in bacterial cells appears to be a powerful approach to counteract bacterial virulence. Enzymes driving the downstream conversion of NaMN to NAD cofactors are present in almost all bacterial genomes and targeting all three enzymes of this pathway NadD, NadE and NadK was proposed more than a decade ago as a promising weapon to fight against a wide range of bacteria (Sorci et al. 2009b). They may even constitute good targets for intracellular pathogens as, for example, the putative L-aspartate oxidase NadB, involved in de novo NAD biosynthetic pathway, is required for intracellular replication of C. burnetii (Bitew et al. 2018). The fact that these bacterial enzymes are substantially divergent from a structural point of view from their human counterparts further strengthens their potential for use as antimicrobial targets.

In silico and in vitro screens enabled the identification of specific inhibitors of NadD from E. coli and Bacillus anthracis that show no effect on the human enzymes (Sorci et al. 2009b). It has also been demonstrated that induced degradation of the M. tuberculosis NadD and NadE enzymes resulted in disruption of both carbon and energy metabolism, resulting in pronounced bactericidal effects (Rodionova et al. 2014). Recently, Osterman et al. have screened a collection of more than a thousand compounds able to interfere with M. tuberculosis NadD, identifying a few with significant inhibitory capacity (Osterman et al. 2019). Preventing NAD cofactors biogenesis by inhibiting NadD may provide a mechanism for elimination of dormant and resistant M. tuberculosis strains.

In reality, the most frequent approach thus far to treat human disorders and diseases related to NAD, has been the pharmacological use of NAD⁺ precursors also known as NAD⁺ boosters. Rajman et al. have reviewed the potential use of NAD⁺‐boosting molecules, such as Nicotinamide ribose (NR), NAM and Nicotinic acid (NA), for the cure of diverse diseases (Rajman, Chwalek and Sinclair 2018). In summary, NR has been shown to be more efficiency in boosting NAD⁺ than NA and NAM, maybe due to a better solubility and increased uptake. However, NA is the one that has shown the greatest diversity of beneficial effects on health such as diminution of risk of myocardial infarct or stroke. In the case of Alzheimer's disease, which is characterized by impaired DNA repair mechanisms, use of DNA repair-deficient mice with low NAD+/NADH ratio in brain tissue revealed that treatment with nicotinamide riboside significantly improved cognitive functions (Hou et al. 2018). However, use of NAD⁺‐boosting molecules to cure infectious diseases is yet to be explored.

It is still unclear how tissues take up NAD⁺ and its precursors, an important factor to take into account for drug design. In the case of NAM, it rapidly diffuses across the plasma and mitochondrial membranes (Hara et al. 2007). Upon administration, NAM is rapidly converted to NAD⁺ through the salvage pathway stimulating SIRT1 activity, beneficial for cellular and organ function. In the context of infectious diseases, knock out of SIRT1 activity in macrophages has been associated with enhanced M. tuberculosis virulence (Singhal and Cheng 2019).

An alternative way to enhance NAD⁺ levels is to prevent its degradation by inhibiting endogenous NAD⁺‐consuming enzymes such as eukaryotic NADases. For example, XAV939 has been identified as SARM1 inhibitor in vitro thus its capacity to treat neurological disorders is currently under development (Rajman, Chwalek and Sinclair 2018). This kind of strategy has been recently successfully applied to bacterial infections, as inhibitors of SIRT2 were shown to effectively limit M. tuberculosis infection in vivo (Bhaskar et al. 2020).

In a similar way, to fight infectious diseases, researchers are currently investigating novel antibacterial approaches that rely on targeting specific bacterial toxins by targeting their enzymatic activity (Ivarsson, Leroux and Castagner 2012; Czaplewski et al. 2016).

As described above, bARTs share a structurally conserved catalytic domain. Thus, targeting of bART activity has been focused on using chemical compounds to prevent their interaction with NAD⁺ as to block their ribosyltransferase activity. Several recent screens have been carried out to identify bART inhibitors, including those of CTX, ChT, ExoA and ExoS (Turgeon et al. 2011; Cherubin et al. 2016; Saleeb et al. 2018). More recently, the screening of >1500 chemical compounds resulted in the identification of two molecules that inhibit PT toxin-catalyzed ADP-ribosylation of Gα proteins (Ashok et al. 2020).

However, a key problem of some of these studies is that they relied on established chemical compound libraries specifically designed for targeting a variety of mechanisms in mammals. They therefore most certainly lack the specificity that would enable to counteract bARTs without impacting mammalian equivalent enzymes, which is essential for successful therapeutics.

Although the majority of NADases share a common glycohydrolase activity, they display specific immunity proteins that could be a promising research axis to discover new bactericidal compounds. Indeed, structural divergences between immunity proteins exposed in this review represent an attractive advantage to ensure a specificity of antibacterial actions and protect, for example, the host microbiota.

Structural analysis of toxin–immunity complexes has revealed features that could be exploited for the development of inhibitors to prevent the interaction of immunity proteins with the NADase, releasing its toxic activity against itself. Some authors have already raised this possibility for inhibiting IFS by preventing the refolding of peculiar surface-exposed structures essential to bind SPN (Smith et al. 2011). Since this study and in contrast to the bART family, not much work has been reported for finding solutions to impede toxic effects generated by bacterial NADases.

Targeting TA systems by disrupting TA complexes or inactivating the antitoxin in bacterial pathogens to develop new antibacterial strategies is currently under investigation. This may constitute a powerful approach as TA systems are widely expressed in bacteria and not in humans (Williams and Hergenrother 2012). In this context, the MbcTA NAD⁺‐phosphorylase toxin system has been proposed as a promising antibacterial target to directly attack intracellular M. tuberculosis (Freire et al. 2019; Ariyachaokun et al. 2020).

CONCLUSION

This is an exciting time to study NAD⁺ metabolism and its relevance in infectious diseases. The increasing recognition of the vital role of NAD⁺ and its related cofactors in pathologic processes has led to new efforts at discovering and developing drugs to counteract enzymes involved in NAD⁺ metabolism, including eukaryotic and bacterial NAD⁺‐utilizing toxins. We summarized in this review how NAD⁺ influences a wide variety of biological processes from DNA repair to metabolic functions in humans, which makes the therapeutic approach using NAD⁺‐boosting molecules more complicated to apply in fighting infectious diseases. We think that targeting bacterial NAD⁺‐utilizing toxins could become a powerful weapon to combat pathogens but also curb antibiotic resistance that submerges the world today. We expect that we will see in the coming years a growing interest and development of highly potent and selective inhibitors against these toxins greatly helped by an increasing identification of essential structural features to be targeted.

It is compelling to see the list of bacterial NADase toxins and effectors growing, suggesting further exciting discoveries are still to come regarding the ability of bacteria to modify the metabolic balance of the host but also of the host's microbiota through the modulation of NAD⁺ levels during infection.

ACKNOWLEDGMENTS

All the figures were created with BioRender.com.

FUNDING

We would like to thank the Fondation pour la Recherche Médicale (FRM) for funding the team and this work (grant DEQ20180339215). MR was supported by a fellowship from the FRM (FDT202012010464) and SPS by Inserm.

Conflict of interest

None declared.