The Mycobacteria genus comprises over 100 species, with approximately 30 recognized as pathogens causing a spectrum of infectious diseases in mammals (1–4). These pathogens are broadly categorized into three groups: the M.tb complex (MTBC), the Mycobacterium leprae (M. leprae) complex (MLC), and the nontuberculous mycobacteria (NTMs) (1, 3, 5). Together, these groups significantly contribute to the global human health burden, with M.tb and M. leprae being the most impactful. According to the World Health Organization (WHO), M.tb causes tuberculosis (TB)in approximately 11 million people and causes over 1 million deaths annually making it the deadliest infectious disease worldwide (6). M. leprae, though less common, still causes over 200,000 new cases of leprosy each year, a disease that leads to severe disability due to the bacterium’s ability to infect peripheral nerve cells (7). In addition to M.tb and M. leprae, NTMs are emerging as significant pathogens (8, 9). Species such as Mycobacterium avium (M. avium), and Mycobacterium abscessus (M. abscessus) are environmental opportunists (10) that primarily infect individuals with underlying conditions like cystic fibrosis (11, 12), bronchiectasis (13), chronic obstructive pulmonary disease (COPD) (14), or immunodeficiencies (15, 16). However, infections can also occur in immunocompetent individuals (17). The ubiquitous presence of NTMs in the environment makes them difficult to control, and their infections are often underreported due to diagnostic challenges and the absence of systematic global surveillance (10).

The global persistence of mycobacterial diseases is exacerbated by the rise in drug-resistant strains. For M.tb, multi-drug resistance (MDR) remains a significant concern, causing approximately 3% of new TB cases in 2023 (6, 18, 19). Similarly, M. leprae eradication efforts are hindered by the emergence of drug-resistant strains, with around 10% of cases resistant to at least one drug in the standard treatment regimen (20, 21). NTMs present an even greater challenge due to species-specific resistance patterns. For instance, M. abscessus demonstrates resistance to nearly all available antimycobacterial therapies (22–25). The intrinsic resistance mechanisms of NTMs, coupled with their ability to form biofilms (26, 27) that impede drug efficacy, make these infections particularly difficult to treat.

Vaccine development against mycobacterial diseases has seen little success in recent decades. The Bacillus Calmette-Guérin (BCG) vaccine, first introduced in 1921, remains the only approved vaccine for M.tb (28). While BCG provides protection against severe extrapulmonary and meningeal TB in children, it has limited efficacy against adult pulmonary TB, the most prevalent form of the disease (29). This highlights the urgent need for a more effective vaccine. For M. leprae, the absence of a conductive in vitro culture system has hindered the development of attenuated or killed vaccines (30). Similarly, the genetic and pathologic diversity of NTMs poses a barrier to creating a universal vaccine for this group of pathogens. Though BCG offers some cross-protection against leprosy (30) and NTM (31, 32) infections, its efficacy is limited.

The development of effective vaccines and novel antimycobacterial therapies is crucial to achieving the goals of the WHO's End TB Strategy (33) and the Global Leprosy Strategy (34). Addressing the drug resistance crisis and improving diagnostic tools for NTMs are equally important. Advancements in biotechnology, including genomic tools, high-throughput drug screening, and innovative vaccine platforms, hold promise for tackling the persistent challenge of mycobacterial diseases.

Recent advances in antimycobacterial therapy have led to the approval of bedaquiline, delamanid, and pretomanid for TB treatment by the FDA (35, 36). Bedaquiline, which targets energy metabolism, has reignited interest in targeting bacterial metabolic pathways (37), particularly the central and essential secondary metabolic pathways in mycobacteria. This represents a paradigm shift, as most traditional antitubercular agents primarily target macromolecule synthesis. For instance, isoniazid and ethambutol inhibit the synthesis of the mycobacterial cell wall by targeting mycolic acid (38) and arabinogalactan/lipoarabinomannan (39) biosynthesis, respectively. Rifampicin and rifapentine disrupt transcription by inhibiting RNA polymerase (40), while fluoroquinolones such as moxifloxacin target DNA gyrase (41), an enzyme critical for DNA replication. Pyrazinamide acts by inhibiting the synthesis of coenzyme A, an essential molecule (42). Despite their success, these drugs face challenges such as resistance due to mutations and inactivity against non-replicating persisters (43–45). Consequently, the development of drugs targeting well-characterized and novel pathways remains a priority.

One promising avenue is the flavin biosynthetic pathway (FBP) and the deazaflavin biosynthetic pathway (DBP), both of which are highly conserved in mycobacteria. The FBP synthesizes flavins, including flavin mononucleotide (FMN), and flavin adenine dinucleotide (FAD), which are metabolites that serve as indispensable cofactors for a wide range of enzymatic reactions (46, 47) (Figure 1). DBP synthesizes the deazaflavins 7,8-didemethyl-8-hydroxy-5-deazariboflavin (F0) and F420, metabolites that are involved in a variety of enzymatic reactions (48) (Figure 1). FMN and FAD are derived directly from riboflavin, a vitamin that humans obtain through diet due to the absence of the FBP in eukaryotes. These flavins are critical in mycobacteria because of their unique redox versatility, transitioning between fully oxidized (quinone), one-electron reduced (semiquinone), and two-electron reduced (hydroquinone) states (49). This property renders FMN and FAD irreplaceable by other redox cofactors, such as NAD and NADP (49). The flavin biosynthesis pathway is also upstream of another essential pathway, the vitamin B12 biosynthetic pathway. The reduced form of FMN is used to produce 5,6-dimethylbenzimidazole (DMB) (50), which is the lower axial ligand of vitamin B12, which in turn is a cofactor for several enzymes in M.tb. F0 serves as the precursor for F420, an obligate two-electron carrying deazaflavin essential for oxidative homeostasis in mycobacteria (48). F420 is also necessary for activating the antibiotics delamanid and pretomanid (51), further underscoring its therapeutic importance. Unlike FMN and FAD, which serve as essential cofactors in human metabolism, F0 and F420 and their dependent enzymes are absent in humans (52). The essentiality of FMN, and FAD (46, 53), coupled with the role of F420 in surviving oxidative stress during infection (54, 55), positions the FBP and DBP as quintessential drug targets. Additionally, the absence of both pathways in mammals minimizes the risk of off-target effects, making it an attractive avenue for drug discovery.

Beyond its metabolic significance, flavin biosynthesis produces specific intermediates presented by the non-classical presenting molecule MHC Class I-Related Protein (MR1) (56, 57). These metabolites are recognized by MR1-restricted T cells (MR1T cells), a unique subset of donor-unrestricted T cells (DURTs). Unlike conventional T cells that recognize peptide antigens, MR1T cells respond to FBP and DBP intermediates through semi-invariant T-cell receptors (TCRs) (57). These MR1T cells can be further subclassified into Mucosal Associated Invariant T (MAIT) cells, which are defined by their expression of the semi-invariant alpha-chain TRAV1-2, and non-TRAV1–2 expressing MR1T cells, which have a more diverse TCR repertoire (58, 59). MAIT cells have been shown to reside in mucosal surfaces (60, 61), including the respiratory tract, the primary infection site of M.tb and several NTMs. Emerging evidence suggests that MAIT cells play a crucial role in controlling mycobacterial infections (62–66) and represent exciting targets for novel therapeutics and vaccines against mycobacteria (67). However, advancing this field requires prioritizing the identification and characterization of the antigenic metabolic produced by mycobacteria.

Historically, the FBP garnered significant attention during the mid-20th century due to its essentiality in bacterial survival (68–74). However, the discovery of alternative drug targets and antimicrobials led to a decline in interest. Moreover, limitations in biotechnological tools at the time hindered deeper exploration of this pathway. With recent technological advances, such as genome editing, high-throughput drug screening, structural biology, and metabolomics, the potential of the FBP as a drug target can now be fully realized. These tools will enable precise investigations into the metabolic capabilities of mycobacteria and the therapeutic exploitation of the FBP.

In this review, we aim to comprehensively describe flavin and deazaflavin biosynthesis and metabolism in mycobacteria, emphasizing its potential as a target for drug discovery and vaccine development. We will highlight existing research gaps that need to be addressed to harness this pathway for therapeutic innovation. Additionally, we will discuss how modern biotechnological tools can accelerate the exploration and exploitation of this vital metabolic pathway.

The FBP and DBP together consist of nine enzymes (Figures 1, 2). Three of these enzymes (RibA2, RibG and an uncharacterized phosphatase) are shared between the two routes (88–90). These shared enzymes are required for the conversion of guanosine triphosphate (GTP) to 5-Amino-6-(ribityl-amino) uracil (5-A-RU) (88). Subsequently, 5-A-RU is converted into riboflavin by the two enzymes, RibH and RibC and eventually FMN and FAD by RibF (47). Alternatively, for the synthesis of F0 and F420, four enzymes are required, namely FbiA, FbiB, FbiC and FbiD to which 5-A-RU serves as the starting material (91–93). Here, we go over the core enzymes of this pathway and the similarity and differences to well characterized organisms such as Bacillus subtilis (B. subtilis) and Escherichia coli (E. coli). Biochemical characterization of these pathways is beyond the scope of this review, but has recently been reviewed elsewhere (94, 95).

The FBP and DBP are highly conserved across Mycobacteria, as illustrated in Figure 2. In M.tb and Mycolicibacterium smegmatis (M.smeg), three key genes of this pathway—ribC, ribA2, and ribH—are organized in an operon known as the rib operon (46). The ribG gene is located separately from the operon with two intervening open reading frames (ORFs) and appears to have no transcriptional relationship with the rib operon (46). Additionally, ribF is expressed as a standalone gene, situated distally from the other pathway genes. In some mycobacterial species, two ORFs that are not involved in riboflavin biosynthesis are found embedded within the rib operon. The high degree of synteny in the arrangement of these operonic genes across different mycobacterial species suggests that they may also be expressed as an operon in other members of this genus. This operonic arrangement of riboflavin biosynthesis genes are a common feature in eubacteria. For example, in B. subtilis, all the genes of the riboflavin biosynthetic pathway (RBP) are clustered in a single operon regulated by a FMN riboswitch (101). The FMN riboswitch serves as a negative feedback regulatory mechanism to prevent the excess production of riboflavin due to the high energy cost of this process and to ensure redox homeostasis (102). A similar mechanism exists in E. coli, despite the genes being scattered across its genome (102). Bioinformatic studies have suggested that mycobacteria lacks a FMN riboswitch regulating the RBP (103), which may indicate that there are other mechanism(s) in place to mitigate the problem of excess riboflavin. One such mechanism could be flavin sequestering proteins that is highly conserved in mycobacteria, which will be discussed later in this review (104, 105). Another reported mechanism is the redox homeostatic system (RHOCS) made up of protein kinase G (PknG), ribosomal protein L13 and RenU, a Nudix hydrolase. The disruption of the RHOCS prevents the degradation of FAD and NAD(P)H by RenU, leading to their accumulation (106). However, the RHOCS is not specific for alleviating redox stress due to flavins as this system primarily senses high levels of NADH. It is also probable that the basal rate of riboflavin biosynthesis provides the right quantity of cofactors (FMN and FAD) needed to drive the high flavin dependence of mycobacteria (107).

A common feature of the RBP in bacteria is the presence of redundant systems to ensure sufficient riboflavin supply. These mechanisms include duplicate pathway genes, which may afford protection from inhibitory molecules targeting riboflavin biosynthesis (108), as well as the ability of some pathogenic bacteria to also encode a riboflavin uptake mechanism (109). The presence of redundant supplies of riboflavin has been shown to be important in the colonization of the host by certain pathogens with the dependence on exogenous or endogenous source of riboflavin varying based on environmental conditions (108). Bioinformatic annotation of the M.tb genome indicate the presence of two ribA genes, rv1940 and rv1415, as well as a second putative deaminase, rv2671 (110). However, functional studies confirmed that only rv1415 encodes a functional enzyme (46), and that rv2671 encodes a dihydrofolate reductase (DHFR) rather than a deaminase (111), thus indicating the presence of a sole gene for all the steps of this pathway. Conversely, in the non-pathogenic M.smeg, a redundant gene encoding a lumazine synthase (RibH) was observed (46). In terms of riboflavin uptake mechanisms, of the nine different families of riboflavin importers, none has been bioinformatically observed in mycobacteria (112).

The lack of redundancy of riboflavin supply and dependence on a sole ORF for each step of riboflavin biosynthesis in pathogenic mycobacteria suggests that targeting this pathway for therapeutics should be relatively straightforward. Additionally, distinct features of the mycobacterial RBP could be leveraged to develop targeted therapeutics that selectively inhibit mycobacteria while sparing the host microbiome. One major concern with antimicrobial therapies is the unintended disruption of commensal bacteria, which can have significant health consequences. However, structural and functional differences in the mycobacterial RBP compared to other bacteria present an opportunity to design highly specific inhibitors. In some riboflavin-competent organisms, the function of hydrolyzing the GTP ring and synthesizing 3,4-dihydroxy-2-butanone-4-phosphate (DHBP) is carried out by two separate enzymes, RibA and RibB (108, 113), whereas in mycobacteria, this process is consolidated into a single multifunctional enzyme (114). These distinctions could serve as a basis for designing inhibitors that exploit the structural and mechanistic uniqueness of mycobacterial riboflavin biosynthesis while avoiding off-target effects on beneficial microbiota which is a strategy currently employed by bedaquiline, an antimicrobial that specifically targets the mycobacterial ATP synthase. By targeting these species-specific variations in flavin biosynthesis, it may be possible to develop antimycobacterial agents that effectively combat infections without the collateral damage associated with broad-spectrum antibiotics. This approach underscores the importance of detailed biochemical and structural characterization of the mycobacterial RBP for the development of therapeutics.

Given the essentiality of flavin biosynthesis and the flavin intense lifestyle of mycobacteria, it is intriguing that pathogenic mycobacterial species do not encode redundant mechanisms to ensure the supply of these vital molecules (46, 107, 115). This phenomenon can likely be attributed to the absence of evolutionary pressure to develop redundancy. Pathogenic mycobacteria colonize traditionally sterile anatomical sites, such as the lungs or peripheral nerves, where they encounter minimal microbial competition. This sterility reduces the likelihood of exposure to inhibitory molecules or metabolic competition from other bacteria, thereby diminishing the selective pressure to evolve backup systems for flavin production. By contrast, environmental mycobacteria and other bacteria that coexist in competitive ecosystems are often subjected to such pressures, which may drive the evolution of metabolic redundancy or alternative pathways to ensure survival (10). For example, photolumazines, which have only been observed in M.smeg, can serve as inhibitors of riboflavin synthase which may provide an advantage in the midst of other environmental microorganisms (116). This distinction emphasizes the unique metabolic adaptations of pathogenic mycobacteria, shaped by their specialized niches within the host, and further highlights flavin biosynthesis as a vulnerable and attractive target for therapeutic development.

Interestingly, the initial genome annotation of M.tb in 1998 did not include genes required for deazaflavin biosynthesis (110). In 2001, fbiA and fbiB were identified as essential for F420 biosynthesis in Mycobacterium bovis (M. bovis) through PA-824 (now pretomanid)-induced selection of transposon mutants (117). The same group later identified fbiC using a similar approach (92), and the discovery of fbiD followed, linked to mutations in its ORF that conferred resistance to pretomanid and delamanid (93). Like the rib operon, two of the fbi genes, fbiA and fbiB, are juxtaposed and have been shown to be co-transcribed (118), while fbiC and fbiD are located at separate loci (Figure 2). Intriguingly, the genetic architecture of the deazaflavin pathway in mycobacteria differs from the archaeal pathway. Five genes instead of four are required for F420 synthesis in archaebacteria with the role of fbiC requiring two separate genes cofG and cofH (89). Also, the length of the polyglutamate tail of F420 has been shown to be shorter in the archaeal organisms in comparison to mycobacteria (90, 119). The distinct mechanisms through which different organisms are able to regulate the length of the glutamate tail is yet to be elucidated. The length of the polyglutamate tail has been shown to impact the kinetics, the turnover rate, and the interaction of F420 with oxidoreductases (120). Shorter tail length has also been linked to resistance (121), however, whether mycobacteria can vary the tail length depending on environmental conditions has yet to be fully understood.

Initial studies characterizing deazaflavin biosynthesis had shown that L-Lactyl-2-diphospho-5`-guanosine (LPPG), a metabolite made from the guanylylation of 2-Phospho-L-lactate via FbiD, was required for the synthesis of F420 (122, 123). However, a study by Bashiri et al. showed that PEP rather than LPPG was required for F420 biosynthesis in mycobacteria (99). This clarification of the substrate of FbiD was due to the lack of genetic and biochemical evidence supporting the synthesis of 2-Phospho-L-lactate in mycobacteria, while PEP was a well characterized product of the glycolytic pathway. This discovery challenged the long-standing schema of this pathway. Recently, the diverse nature of this pathway has become more obvious, with different organisms using molecules other than PEP for F0 synthesis (89). For example, the Gram-negative bacterium Paraburkholderia rhizoxinica utilizes 3-phopsho-D-glycerate (3PG) instead of LPPG or PEP (124). The use of PEP rather that LPPG in mycobacteria was further supported by the presence of a FMN binding domain in FbiC which catalyzes the reduction of dehydro-F420-0 (99), the product of FbiA if PEP is used. Also, the synthesis of F0 was thought to require the condensation of 5-A-RU and 4-hydroxyphenylpyruvate (4-HPP). However, the biochemical characterization of FbiC enabled better elucidation of F0 synthesis and demonstrated the need for tyrosine instead of 4-HPP.

Beyond the production of flavins and deazaflavins, 5-A-RU may participate in the synthesis of additional metabolites. Notably, 5-A-RU condenses non-enzymatically with glyoxal and methylglyoxal, secondary products of metabolism, to form 5-(2-oxoethylideneamino)-6-D-ribitylaminouracil (5-OE-RU) and 5-(2-oxopropylideneamino)-6-D-ribitylaminouracil (5-OP-RU) (125). These unstable metabolites are potent ligands for MR1. Additionally, other studies have demonstrated the formation of lumazine compounds through the non-enzymatic condensation of 5-A-RU with transamination products (126, 127). The enzymatic modification of 5-A-RU to provide starting material for a tangential metabolic pathway has also been observed in other organisms (128). The promiscuity of 5-A-RU and the non-enzymatic production of its derivatives suggest that this pathway may contribute to the synthesis of other metabolic products. Although most known products of 5-A-RU outside flavins and deazaflavins have only been characterized as MR1 antigens, it is necessary to determine whether these metabolites have other physiological roles in mycobacteria. In addition, the distinct metabolome of mycobacteria (129, 130) compared to other organisms may provide an avenue for the formation of novel secondary metabolites derived from 5-A-RU.

Another point in the FBP that may lead to the synthesis of other novel metabolites is from DMRL. DMRL, the direct precursor for riboflavin, belongs to the broad class of nitrogen-containing heterocycles known as pteridines or lumazines. The ribityl lumazine motif serves as a core structure for several naturally occurring compounds in various bacterial species, often carrying diverse additional moieties that contribute to their biological functions. One notable class, photolumazines, has been shown to be produced by M.smeg (85, 86), and is yet to be observed in any pathogenic species. Photolumazines may play a role in protecting M.smeg in light-exposed environments as these compounds function in association with lumazine binding proteins as optical transponders (131). Therefore, the presence of photolumazines and their functional role in pathogenic mycobacteria such as M.tb and M. leprae is likely minimal or absent, given their evolution to survive in the absence of light. However, NTMs, often sourced from environmental niches (10), may produce photolumazines. Since both DMRL and 5-A-RU have been implicated in the formation of lumazines in other organisms, it is critical to investigate pathogenic mycobacteria for the presence of such substrates and to explore their potential physiological and immunological relevance.

FBP also provides a precursor essential for cobalamin biosynthesis. Specifically, reduced flavin mononucleotide (FMN-H₂) undergoes reduction by BluB, a nitroreductase-like enzyme, yielding DMB and erythrose-4-phosphate (E4P) (50). Subsequently, DMB is incorporated into cobalamin through additional enzymatic reactions (50). While cobalamin is critical for mycobacterial metabolism, members of the MTBC typically obtain it exogenously from their host due to an incomplete de novo synthesis pathway (132, 133). In contrast, NTMs possess the complete machinery to synthesize cobalamin independently (132), relying on DMB derived from FMN. The broader implications of disrupting the synthesis of cobalamin lie beyond the scope of this review but have been previously addressed elsewhere (134).

With the wealth of information now available about these pathways and the expanding repertoire of therapeutic strategies beyond traditional small-molecule inhibitors, there is significant potential to develop novel antimycobacterial agents. Additionally, the conservation of flavin and deazaflavin biosynthesis among mycobacteria suggests that inhibitors designed to target this pathway in one species will likely be effective against others. For example, bedaquilline targets ATP synthase, a highly conserved function in mycobacteria, and has demonstrated potent antibacterial activity against several mycobacterial species (135–137). This broad applicability enhances the feasibility of developing a single therapeutic agent capable of targeting multiple pathogenic mycobacteria, making flavin and deazaflavin biosynthesis an attractive target for the next generation of antimycobacterial drugs.

There are currently over 200 families of enzymes that depend on FMN or FAD as cofactors (153) and 8 families that depend on deazaflavins (90). These proteins are involved in a plethora of functions in nature and in the mycobacteria species. We have highlighted some of these functions in mycobacteria in Figure 3.

In this section, we provide an overview of the function of key flavin dependent enzymes, flavoproteome, and deazaflavin dependent enzymes in the context of features essential to the pathogenicity of mycobacteria.

The flavin biosynthetic pathway presents several druggable targets, as five known enzymes within this pathway (RibA2, RibG, RibH, RibC, and RibF) are essential for bacterial viability. Among these, only RibF, a bifunctional riboflavin kinase/FAD synthase, has a homolog in the human host. Since humans are incapable of synthesizing riboflavin, they rely on dietary intake. This riboflavin is subsequently converted to FMN and FAD by separate enzymes, a riboflavin kinase and FAD synthase respectively, whereas both of these functions are carried out by the RibF enzyme in prokaryotes (213). The conservation of the riboflavin kinase domain of RibF between prokaryotes and eukaryotes in both sequence and structure makes it a less druggable target, however the FAD synthase domain displays significant structural divergence which may serve as a drug target (214).

Efforts to design inhibitors targeting this pathway have largely focused on lumazine synthase (RibH) and riboflavin synthase (RibC), primarily through the development of substrate analogs that mimic their natural ligands (215–223). Following the structural and biochemical characterization of these enzymes in B. subtilis and E. coli respectively, an early study successfully identified an inhibitory molecule, which, although a weak inhibitor, served as potent scaffold for further optimization (216). Subsequent studies were able to identify molecules, derivatives of ribitylaminolumazines (218) and purinetrione (217), that exhibited improved binding to the lumazine and riboflavin synthase of E. coli and B. subtilis. These compounds provided a useful scaffold that guided the design of inhibitors against M.tb, with three alkyl phosphate (Figure 4A) derivatives of purinetrione demonstrating nanomolar inhibition of M.tb RibH (224). The resolution of the crystal structure of RibH of M.tb catalyzed the structure-based design and the virtual screening campaigns of libraries for enzyme inhibitors (225–227). The binding mode of the three alkyl phosphate derivatives to M.tb RibH was then determined in order to optimize the design of better inhibitors (228). Subsequently, two derivatives of ribityllumazinediones (Figure 4B) (215) were shown to inhibit the enzymatic activity and bind the RibH of M.tb with high affinity, however their bactericidal activity was not characterized. Beyond the ribityllumazine and purinetrione scaffold, additional novel inhibitors of lumazine and riboflavin synthase have been developed. Derivatives of a sulfur nucleoside analogue of the RibH substrate 5-A-RU (Figure 4C), inhibited both RibH and RibC of M.tb (220). In a follow-up study by the same group, derivatives of O, S, and C nucleoside analogues of 5-A-RU (Figure 4C) also demonstrated inhibitory activities against RibH and RibC (221). Further, a series of three 3-alkyl phosphate derivatives of pyrazolopyrimidine analogues (Figure 4D) were shown to be potent inhibitors of M.tb RibH (223). Another series of oxalamic acid derivatives of 5-A-RU (Figure 4E) was shown to bind RibH and RibC of M.tb with moderate inhibition (Ki) of these enzymes (222). Using a high throughput screen to identify inhibitors of lumazine synthase in Schizosaccharomyces pombe, an analog of 5-A-RU with a substituted ribityl side chain was shown to inhibit M.tb RibH (229, 230) (Figure 4F).

Although most of these compounds identified as inhibitors of RibH have provided important insights into the binding mode of this enzyme, there has been little success in translation to pharmacological agents, likely due to their inactivation by phosphatases (224) or inability to permeate the mycobacterial cell wall (231). However, recent progress in targeting this pathway has led to the identification of novel inhibitors of RibH in M.tb. A recent study using molecular docking of a ~600,000 compound library identified three novel RibH inhibitors with potent antimycobacterial activity and synergy with first-line anti-TB drugs (Figure 4G) (227). Interestingly, these compounds possessed unique chemical scaffolds distinct from traditional substrate mimics, marking a significant shift in the design strategy for RibH inhibitors and potentially for inhibitors of other enzymes in this pathway.

There has also been some success in identifying molecules targeting RibC solely. Kaiser et al. (232) designed an enzymatic photometric read out high-throughput screen for inhibitors of specific riboflavin pathway enzymes. Applying this screening format against a commercial library consisting of 100,000 compounds, they identified 127 hits, with two of them (Figure 4H) showing significant inhibition of M.tb RibC and antimicrobial activity against replicating and non-replicating M.tb (233). However, on-target activity of the compounds had not been confirmed. Serer et al. also developed a unique high-throughput screening platform to search for RibC inhibitors against Brucella sp (234). They employed a mobility shift assay coupled with a microfluidic system, where inhibition was observed as a reduction in the riboflavin fluorescent signal. Screening a 44,000-compound library led to the identification of 10 inhibitors of which three demonstrated moderate growth inhibition even in the presence of riboflavin (234). However, activity of these molecules against M.tb RibC is yet to be determined.

An alternative strategy to inhibit flavin biosynthesis involves targeting the regulation of flavin biosynthesis. Roseoflavin (RoF), a natural riboflavin analog synthesized by Streptomyces exhibits antimycobacterial activity (235). RoF can bind to the FMN riboswitch, leading to the downregulation of RBP genes and consequently reducing riboflavin synthesis (236). Another report documented five riboflavin derivatives, modified at the ribityl moiety that displayed moderate antimycobacterial activity (237). Through in silico docking, and by demonstrating binding affinity to the FMN riboswitch aptamer of Lactobacillus plantarum, FMN riboswitch binding was proposed as one of their mechanisms (237). However, as the presence of an FMN riboswitch in mycobacteria has not been confirmed, the alternative mechanism of RoF and these riboflavin derivatives may explain their antimycobacterial properties. In the case of RoF, it mimics riboflavin and gets converted to inactive cofactors Roseoflavin-5`-monophosphate (RoFMN) and Roseoflavin adenine dinucleotide (RoFAD) which get incorporated into flavoenzymes, either inhibiting or reducing their activity (238). A similar mechanism was also purported for the riboflavin derivatives (237), however, the clinical use of these molecules is impeded by their potential to also disrupt the host flavoproteome. Other inhibitors of riboflavin biosynthesis have also been reported. Ribocil (239), a synthetic small molecule, and 5FDQD (240), a flavin analog, were both shown to be bactericidal to E. coli and Clostridium difficile respectively via repression of RBP genes by binding to the FMN riboswitch. Similar to RoF, the use of these molecules in mycobacteria is impeded by the lack of an FMN riboswitch. However, an approach of disrupting the regulation of flavin biosynthesis in mycobacteria may be employed for drug targeting.

From the studies highlighted above, it is clear that RibH and RibC are the most widely targeted enzymes in the flavin pathway. This preferential targeting likely stems from the better stability and malleability of their substrates and a better understood mechanism of action than the RibA2 and RibG enzymes (231). Although, several inhibitors for RibH and RibC have been identified, most of these compounds were identified through cell-free assays using recombinant enzymes, hence their efficacy as antimycobacterial agents remains uncertain. In vitro and in vivo validation is necessary to assess their translational potential; however, such studies have yet to be reported. Also, studies demonstrating the on-target inhibitory effect on the flavin biosynthetic pathway of these molecules will be essential to demonstrating their mechanism of action. With the advent of tools to better understand the structure and function of these enzymes, the discovery of inhibitors of this pathway should be more approachable. Also, allosteric inhibition of these enzymes may serve as an alternative to substrate mimicry which may facilitate targeting other components of this pathway beyond RibH and RibC.

Four major considerations must be addressed collectively when targeting the flavin biosynthetic pathway pharmacologically. First, the relative vulnerability of mycobacteria to inhibition at different steps of the pathway should be evaluated to identify the most effective enzymatic target. For instance, the presence of two functional lumazine synthase homologs in M.smeg may complicate pharmacological targeting of this enzyme. It is therefore crucial to assess whether similar functional redundancy exists in other pathogenic mycobacterial species. Second, the ideal target should be the component of the pathway most resistant to mutational escape, thereby minimizing the likelihood of developing antimicrobial resistance. Third, given the pathway’s involvement in the biosynthesis of MR1 ligands, the immunological impact of inhibiting different enzymes on MR1-restricted T cell responses must be considered when selecting a therapeutic target. Finally, the impact of targeting the shared pathway (RibA2, RibG) on the administration of deazaflavin dependent prodrugs must be considered. It is well established that disruption of deazaflavin biosynthesis confers resistance to pretomanid and delamanid (93). Therefore, co-administering pretomanid and delamanid with inhibitors of RibA2 or RibG would be impractical. Collectively, the unique mechanism of action associated with flavin pathway inhibitors offers a promising avenue to circumvent existing drug resistance mechanisms and contribute to sustained therapeutic efficacy.

MR1 is the most conserved antigen presenting molecule in mammals (262, 263). For instance, the critical antigen binding domains of MR1, the α1 and α2 domains are 90% conserved between mice and humans (263, 264). This high degree of conservation suggests that MR1 plays a fundamental and evolutionarily important function in the immune system that has therefore been maintained by evolutionary pressure (262). MR1T cells were first described in 1993, as a common type of CD4^-CD8^-cell expressing an invariant TCR α-chain (TRAV1-2/TRAJ33) (265). These TRAV1–2 expressing cells were later termed MAIT cells given their invariant α-chain, enrichment in mucosal surfaces and later demonstrated to be restricted by MR1 (266). We now recognize that MAIT cells are only a subset of the broader class of MR1T cells. In adults, MAIT cells are the most common MR1T cell and are defined by their expression of an invariant TCR α-chain (TRAV1-2/TRAJ33/20/12 in humans), paired with a limited array of Vβ segments (267, 268), and often express high levels of CD161 (269, 270) and CD26 (271). MR1T cells have been shown to play an important role in microbial defense, providing protection against many different riboflavin producing pathogens, including Klebsiella pneumoniae (272), E. coli (260), Francisella tularensis (273), Streptococcus pneumoniae (274), and mycobacterial pathogens (64, 65, 275).

MR1T cells have many features that make them well suited for defense against a wide variety of pathogens. They are enriched in mucosal surfaces, including the lung (276) and gastrointestinal tract (277), the first contact point for most infections. They do not require priming and are therefore able to respond much more rapidly than conventional T cells. They respond with production of critical inflammatory cytokines including IFNγ and TNF, but also with a robust cytotoxic response, more so than conventional T cells and this may be particularly important for intracellular pathogens such as M.tb (276). Insights from key studies support the immunological importance of MR1 in M.tb defense, including both animal models and human studies. In murine models, MR1 knockout mice exhibit worse control of BCG infection, showing the role of MR1 in mycobacterial defense (275). In non-human primates (NHPs), blocking CD8a to deplete all CD8⁺ cells, including non-conventional T cells like MR1T cells, had a profound impact on bacterial control and dissemination, whereas blocking CD8b to only deplete conventional CD8-positive cells had only a modest effect on lymph node bacterial control (278). Using a similar model, this same group demonstrated that the protection given to NHPs from IV BCG was lost with CD8a depletion, but not CD8b depletion (279). These studies highlight the important role non-conventional CD8-positive T cells, such as MR1T cells, play in control of M.tb, but also in vaccine-mediated protection from M.tb. Human studies further reinforce MR1’s role in TB defense, with MR1T cells being higher in a Ugandan cohort of TB resisters (280). Additionally, MR1 polymorphisms have been associated with severe and disseminated TB, underscoring the clinical relevance of MR1 in TB pathogenesis (257).

Since the initial discovery that intermediates of the RBP serve as ligands for MR1T cells, many other ligands have been discovered, with some functioning as agonists (261, 264) and others as antagonists (87, 258). Among the most potent agonists are 5-OP-RU and 5-OE-RU (Figure 5) which bound to majority of the known MR1T cells TCRs (125). This potency was leveraged to develop a MR1-tetramer loaded with 5-OP-RU that allowed researchers to study MR1T cells that do not express canonical markers attributed to MR1T cells (267). The use of MR1-tetramers led to the discovery of TRAV1-2 (–) MR1T cells including MR1T cells that recognized the riboflavin auxotroph, Streptococcus pyogenes, in an MR1-dependent manner. These findings started to disprove the hypothesis that the TCR of these T cells were highly invariant. These findings have also spurred investigations to determine the full repertoire of TCRs utilized by MR1T-cells as well as how this impacts the recognition of ligands presented by MR1.

Following the development and expansion of MR1T cells over time has contributed to the observed heterogeneity in their TCR repertoire. Although the majority of MR1T cells in adults are identified as belonging to the subclass of MAIT cells and express the TRAV1–2 chain, this has been shown to differ from what is observed at birth. Using the MR1/5-OP-RU tetramer, the majority of the tetramer(+) cells at birth were identified to be TRAV1-2 (–) (65). Recent work performing TCR sequencing of MR1/5-OP-RU tetramer(+) MR1T cells at birth has further shown that this TCR diversity extends beyond just alternate TRAV usage, with tremendous diversity in TCR usage similar to what is observed in conventional T cells (281). These more diverse cord blood derived MR1T cells were less able to recognize riboflavin producing microbes, suggesting that they may recognize a more diverse array of ligands. Further demonstrating this heterogeneity of MR1T TCRs and the antigens they recognize, work done by Gold et al. showed that different clones of MR1T cells were capable of recognizing riboflavin-producing organisms (S. typhimirium, C. albicans and M.smeg) in an MR1 dependent manner. However, only some of these clones could recognize the ligand 6,7-dimethyl ribityl lumazine (RL-6,7-diMe) (282). This finding indicates that, even in the context of the ligands derived from the RBP, there is a diversity of ligands.

Similar to the discovery of the diversity of TCR of MR1T cells, understanding of the molecules thought to be presented by MR1 has changed drastically over the years with the discovery of different classes of molecules including folic acid derivatives (57, 125), pyridoxal derivatives (261), nucleobase derivatives (283–285). The diversity of ligands presented by MR1 was further explored by Harriff et al. who used mass spectrometry to examine the E.coli and M.smeg ligand repertoire presented by MR1 (258). Some ligands were only observed in M.smeg and these activated some, but not all, of the MR1T cells clones used in the study. This finding suggests that MR1T cells are capable of discriminating between pathogen-specific ligand repertoires through TCR dependent recognition. Collectively, these studies support the idea that MR1T cells can mount tailored responses to distinct microbial exposures, rather than acting as a uniform invariant population responding to a single conserved antigen. Ultimately, these findings legitimizes the need to identify MR1 ligands, in this case, ligands specific to the mycobacteria genus and species.

Since the discovery of 6-formylpterin (5-FP) and 5-OP-RU, the first MR1 antagonist and agonist respectively (57, 125), the list of MR1 ligands has expanded to include both naturally occurring and synthetic molecules. With the recent uncovering of the heterogeneity of TCR utilized by MR1T cells and the binding pocket of MR1 shown to be potentially receptive to other molecules beyond those currently identified, the MR1 ligandome is likely to keep expanding. Here, we will discuss currently identified MR1 ligands in mycobacteria, the current state of research probing the mycobacterial MR1 ligandome and their potential role as immunomodulators.

MR1 ligands are currently classified as antagonist and agonists based on their ability to induce MR1 expression and interact with the TCR of MR1T cells. For both agonists and antagonists, these molecules can be loaded on MR1 in the endoplasmic reticulum (ER) of antigen presenting cells (286, 287), where unloaded MR1 molecules reside, and induce the egress of MR1 to the cell surface where it presents the loaded molecule to the cognate TCR (288). At the point of interaction between the loaded MR1 and the TCR comes the distinction between agonists and antagonists. Agonists induce a response via the TCR activation of MR1T cells leading to cytotoxic activity and the release of proinflammatory cytokines (289). Conversely, antagonists may interact with the TCR but do not induce an immune response (290). This antagonistic property has been shown to prevent TCR dependent activation of MR1T cells via agonists as antagonist loaded MR1 are unable to present agonistic ligands. This competitive inhibition of MR1 agonists presentation by antagonists has been shown experimentally and may play a role in immunomodulation (290). Conversely, the antagonist induced egress of MR1 to the cell surface has also been shown to facilitate the loading of agonist from the extracellular milieu via a yet-to-be-defined exchange mechanism (291). However, it is still unclear how much this exchange mechanism contributes to MR1 antigen presentation during infection. Recently, another mechanism of MR1 antagonism was discovered where antagonists retain MR1 in the ER and prevent its egress to the cell membrane (292). As opposed to antagonists that induce surface expression, antagonists that prevent egress prevent antigen loading both in the ER and the exchange at the plasma membrane. Currently, only synthetic molecules have been shown to utilize the egress-inhibition mechanism. However, the evidence of such mechanism indicates a likelihood of a mycobacteria derived ligand possessing such property.

The FBP and DBP in mycobacteria have been shown to produce both agonistic and antagonistic ligands of MR1 (Figure 5). DMRL, is a well-established MR1 agonist, albeit to a lesser extent in comparison to 5-OP-RU (290, 293). F0 and riboflavin have also been shown to be loaded on MR1, although activation studies showed that they had antagonistic properties (258). Beyond these intermediates, photolumazines have also been observed in the environmental M.smeg, as previously discussed. Harriff et al. identified Photolumazine I and III using recombinant MR1 in insect cells (258) (Figure 5). Krawic and Ladd et al. also identified four isomers of Photolumazine IV in M.smeg using a similar system (294) (Figure 5). Although none of these ligands have been identified in pathogenic mycobacteria, their discovery provides some context on the expanding family of MR1 ligands in mycobacteria. When these ligands were discovered, they were initially thought to be derivatives of DMRL. However, two recent studies indicate that they are likely derived from 5-A-RU instead. As earlier stated, 5-A-RU was shown to form lumazines, including Photolumazine III, through condensation reactions with transamination products in E. coli (126). Also, we recently showed the centrality of 5-A-RU to produce MR1 ligands in mycobacteria. Mutants of M.smeg and M.tb that lacked RibA2 or RibG were unable to activate MR1T cell clones (295). Interestingly, the loss of DMRL production had little to no impact on MR1T cell activation. Since 5-A-RU itself is not loaded onto the MR1 ligand groove (57), it is hypothesized that 5-A-RU primarily serves as a primer for the synthesis of MR1 ligands. This hypothesis is further corroborated by the ubiquitous nature of 5-A-RU in all riboflavin competent organisms (231). However, 5-A-RU serving as an antigen does not account for the unique MR1 ligandome of different organisms. The findings from these studies, as well as those in the previous section, highlight the need to discover these mycobacterial ligands and understand how they impact MR1 immunology during infection and how the metabolic status of mycobacteria impacts their abundance.

The pathogenic success of mycobacterial species, particularly M.tb, partially relies on their capability to modulate the host immune response during infection. The production of MR1 agonists (such as DMRL) and antagonists (such as F0) by mycobacteria highlights their potential role as immunomodulatory agents. Furthermore, during infection of a host, mycobacteria can assume several phenotypically distinct states due to the plasticity of its metabolic landscape (296, 297). This plasticity enables mycobacteria to establish itself as a highly heterogenous population dependent on stressors encountered in host (298–302). These populations differ in cell wall composition, replication dynamics and kinetics, and the status of CCM and accessory pathways. This metabolic plasticity exhibited by mycobacteria enables dynamic alterations in metabolic antigenicity, thus influencing host-pathogen interactions. It is possible that mycobacteria utilize a similar mechanism to modulate surveillance by MR1T cells as its reliance on flavins may change in some of these states. Proteomic analysis of M.tb during hypoxia showed a transient increase in the abundance of all the proteins in the riboflavin and deazaflavin pathway except for fbiB (249). In the same study, they also observed an increased abundance of FMN and decreased abundance of 5-A-RU and DMRL as oxygen levels depleted (249). It is unclear how these changes may impact MR1 antigenicity as the flux of these intermediates toward downstream products cannot be predicted from this data. Hence, further investigation is necessary to determine how these changes in the proteome and metabolome during hypoxia and other stress-induced dormancy impact the MR1 ligandome of mycobacteria.

It is also clear that the immunological pressure faced by mycobacteria influences its genetic evolution (303, 304). Given the importance of MR1 during mycobacterial infection, it will be important to determine evolutionary changes that might have impacted flavin and deazaflavin biosynthesis in mycobacteria. It was observed that a clinical strain of Salmonella enterica had evolved to overexpress ribB, the gene that encodes the equivalent of the DHBP synthase activity of ribA2 in mycobacteria, evaded recognition by MAIT cells (305). Although, it was uncertain how the overexpression of this gene impacted MR1 ligand production, the investigators observed higher levels of extracellular riboflavin and FMN in the ribB overexpressor. It is possible that the overproduction of DHBP may drive the pathway toward the synthesis of DMRL and riboflavin and prevent the buildup of 5-A-RU, the required intermediate for MR1 antigens. Similarly, a pathogenic strain of F. tularensis was shown to harbor mutations in the ribA2 gene which negatively impacted MAIT cell activation, a phenotype shown to enhance virulence (306). Another study investigating the impact of overexpressing the rib pathway genes in mycobacteria showed increased MAIT activation and decreased virulence in the mouse model when ribA2 was overexpressed (307). This finding was shown to be MAIT cell dependent as infection of CAST/EiJ mouse model, which has an increased frequency of MAIT cells, led to better protection. Earlier studies in the 1930s attempted, without success, to link riboflavin production with virulence (68). Given recent findings, their original hypothesis may indeed have merit and warrants renewed investigation.

Current research on MR1 ligands in mycobacterial species has predominantly focused on M.smeg and M.tb. However, the emerging evidence of a broader MR1 ligandome than previously recognized, with species-specific variations, highlights the need to expand ligand characterization efforts to NTMs and M. leprae, which exhibit distinct metabolic profiles compared to M.tb (8). Growth rate differences between rapid- and slow-growing mycobacteria likely impact the dependence on and the flux of the intermediates through the flavin/deazaflavin biosynthesis. This metabolic divergence suggests pathogen-specific ligand signatures which may influence the role of MR1T cells during infection. With recent advancements in mass spectrometry and bioinformatic tools, the identification of novel natural compounds has become significantly more feasible. It is now an opportune moment to leverage these powerful technologies to systematically identify and characterize MR1 ligands. Discovering both flavin-related and non-flavin-related ligands represents an essential step toward elucidating their roles in MR1-mediated immune responses during mycobacterial infections. Furthermore, pinpointing these ligands will facilitate optimal targeting of MR1-restricted immune cell populations, particularly those residing at the site of infection, enhancing both immunotherapeutic and vaccine development strategies.

MR1T cells have many features that make them attractive vaccine targets. In addition to being enriched in mucosal surfaces (276), the main sites of most initial infections, they are activated more rapidly than conventional T cells, and they have robust cytotoxic capabilities (Figure 6) (308). Furthermore, MR1T cells are also active against a broad array of pathogens for which no effective vaccine currently exists. Importantly, unlike the highly polymorphic Major Histocompatibility Complex (MHC), the antigen-presenting molecules that restrict these unconventional T-cells are non-polymorphic (262, 266, 308). This has major implications for vaccine design. Genetic variations in Human Leukocyte Antigen (HLA) genes influence immune responses to vaccination, posing a challenge to developing universal vaccine strategies. In contrast, MR1’s non-polymorphic nature, and the consequent donor-unrestricted characteristics of MR1T cells, suggest that genetic variability is less likely to affect vaccine efficacy.

In mouse models, 5-OP-RU was explored as a vaccine candidate to activate MR1T cells for protection against M.tb and while there was robust expansion of MR1T cells in the lung, this was not associated with protection (309). This was also attempted in NHP models, and similarly, 5-OP-RU did not provide protection against M.tb, and in fact led MR1T cells to upregulate PD-1 and lose the ability to produce important cytokines such as IFN-γ (310). Although these initial attempts have been unsuccessful, we are only just beginning to understand the complex biology and MR1T cell activation, and utilizing a very broad activating antigen such as 5-OP-RU may not be an appropriate strategy. Work from Riffelmacher et al. (274) has shown that vaccination with the live attenuated Salmonella vaccine strain, BRD509, in mouse models led to an expansion of lung-resident antigen adapted MR1T cells with enhanced effector programs. This expansion was associated with a protection against subsequent Streptococcus pneumoniae infection, providing proof of concept for an MR1T-based vaccine strategy capable of conferring cross-species protection in mice (274). Additionally, several studies have shown the capacity of MR1T cells to assume an innate memory-like phenotype (273, 276, 277, 311–316). In particular, recent work by Kain et al. showed that following BCG vaccination, there was an expansion of MR1T meta-clonotypes with increased expression of pro-inflammatory and cytotoxic genes, indicative of a recall-like response to prior vaccination (316). Systematically boosted MR1T cell were shown to induce protection against Francisella tularensis a month after vaccination with the MR1 agonist, 5-OP-RU (273). Using a similar vaccination strategy, the expansion of lung resident MR1T cells a month prior to infection with Legionella longbeachae led to a significant reduction in bacterial burden (315). A recent study showed that Listeria monocytogenes, a riboflavin auxotroph, engineered to produce riboflavin conferred protection against F. tularensis when used as a vaccine (312). These findings provide evidence of sustainable MR1T cell expansion enabling the development of vaccination strategies targeting MR1T cells. Since most of the studies assessing the vaccine-targeting potential of MR1T cells have evaluated immune response within only a few weeks to a month, further research is required to determine the long-term persistence of the innate, memory-like phenotype of MR1T cells. Understanding the duration and stability of this response will be critical for defining the most effective vaccination strategy that can optimally engage and sustain MR1T cell-mediated immunity.

With the growing recognition of the critical role of MR1T cells during M.tb infection, several studies have examined the impact of BCG vaccination on MR1T cell-mediated (316–319). BCG vaccination was shown to modulate the MR1T cell landscape at infancy suggesting its capacity to induce an MR1T cell-mediated response (316). Similarly, BCG revaccination in adults was shown to expand the MR1T cell population, a finding that was recapitulated in non-human primates (317, 319). However, although BCG clearly induces an MR1T cell response, the longevity and the extent to which this response contributes to protection against M.tb remains unclear. Furthermore, confirming the similarity between the MR1 ligandome and TCR repertoires elicited by BCG and those triggered by M.tb infection is essential to establish effective cross-induction. Finally, identifying the optimal route of BCG administration to maximize the activation of both conventional and unconventional T cells, including MR1T cells, remains an important consideration for vaccine optimization (320).

Although 5-OP-RU broadly activated the majority of MR1T cells, it has been shown that MR1T cells have antigen selectivity (258, 282), and thus the choice of antigen that leads to expansion could play an important role in subsequent protection from infection in a manner akin to conventional T cells. Furthermore, MR1T cells are also cytokine responsive, and activation of MR1T cells in the setting of different cytokine exposure can influence the functional ability of MR1T cells. For example, Wang et al. have recently shown that MR1T cell clones can be induced to switch from IL-17-producing clones to IFNγ-producing clones (311). Thus, it is entirely possible that it is not just the antigen that activates MR1T cells that is important in inducing protection from these cells, but also the cytokine environment that shapes their eventual function and ability to protect against subsequent infection.

Given these facts, identifying MR1 ligands unique to mycobacteria will proffer several advantages in the design of vaccines targeting MR1T cells, Firstly, it will be able to optimize the delivery of the antigens to the right population of immune cells and anatomical region. Secondly, using tetramers loaded with these ligands, it will be feasible to track how different vaccines strategies impact the expansion and function of cognate MR1T cells. Most importantly, the best strategy to ensure the development of memory MR1T cells can be determined. Furthermore, metabolite recognition offers advantages over peptide antigen recognition. Peptide-based immune responses are influenced by post-translational modifications, protein-protein interaction and protein secondary structures, whereas metabolite recognition is comparatively simpler. Additionally, synthesizing and manipulating small molecules is more feasible than working with peptides, making them attractive targets for vaccine development (321, 322).