For decades after its introduction, the mechanisms of action of the front-line antituberculosis therapeutic agent isoniazid (INH) remained unclear. Recent developments have shown that peroxidative activation of isoniazid by the mycobacterial enzyme KatG generates reactive species that form adducts with NAD+ and NADP+ that are potent inhibitors of lipid and nucleic acid biosynthetic enzymes. A direct role for some isoniazid-derived reactive species, such as nitric oxide, in inhibiting mycobacterial metabolic enzymes has also been shown. The concerted effects of these activities – inhibition of cell wall lipid synthesis, depletion of nucleic acid pools and metabolic depression – drive the exquisite potency and selectivity of this agent. To understand INH action and resistance fully, a synthesis of knowledge is required from multiple separate lines of research – including molecular genetic approaches, in vitro biochemical studies and free radical chemistry – which is the intent of this review.
Tuberculosis (TB), caused by infection with Mycobacterium tuberculosis, kills over 2 million people per year, with between 1 billion and 2 billion people latently infected worldwide (World Health Organization, 2002). Not only has the unfortunate synergy between TB and HIV increased the already high human life toll, but also the emergence of multidrug resistant strains, which are both difficult and very costly to treat, poses an additional public health hazard and further roadblock in effective control of TB. Although there has been a renewal of activity in the development of antimycobacterial agents, such as nitroimidazo (Stover et al., 2000) and diarylquinoline derivatives (Andries et al., 2005), there remains as much need for new drug discovery as there is for fundamental research from which to delineate mechanisms and targets of already proven and effective agents. Many of the drugs still used to treat TB were discovered many years ago, before the advent of newer powerful molecular techniques. Thus, the mechanisms of action of many older TB drugs were poorly defined, yet many are still part of first- or second-line therapies. So, in addition to high-throughput library screening, and rational identification of new mycobacterial targets, another promising approach would be to elucidate the mechanisms of action of these older, tried drugs. These activities could then be rationally screened for and optimized in development programmes. No drug is more suited to such an approach than isoniazid (isonicotinic acid hydrazide, INH, 1 in Fig. 1), a potent, highly selective agent that is still a centrepiece of therapy (Youatt, 1969; Deretic et al., 1996; Zhang et al., 1996) some 50 years after its discovery yet with mechanisms of action that have remained contentious. We use the plural, mechanisms, deliberately, as a range of potent mechanisms have recently been uncovered that may act additively or synergistically to explain the exceptional and highly selective potency of INH against M. tuberculosis. Here we review these mechanisms.
INH is a prodrug activated by KatG
INH enters the mycobacterial cell by passive diffusion (Bardou et al., 1998). INH itself is not toxic to the bacterial cell, but acts as a prodrug and is activated by the mycobacterial enzyme KatG (Zhang et al., 1992), a multifunctional catalase-peroxidase that has other activities including peroxynitritase (Wengenack et al., 1999) and NADH oxidase (Singh et al., 2004). While the protective activities of KatG against host phagocyte NADPH oxidase-derived peroxides appear important in the absence of INH (Ng et al., 2004), strong selection for INH-resistant mutants occurs during treatment, with a major site for INH resistance mutations being the katG gene (Zhang et al., 1992; Heym et al., 1995; Slayden and Barry, 2000). These mutations include selective point mutations that result in a partially active protein that retains some of the bacterial survival-supporting activities of KatG while reducing INH toxicity (e.g. S315T), although a range of other mutations are also known (Musser et al., 1996; Rouse et al., 1996; Marttila et al., 1998).
In accord with its peroxidase activity, KatG activates INH by peroxidation to produce intracellular, INH-derived damaging species, and activation here refers to this formation of reactive INH-derived species. A range of oxidants support KatG oxidation of INH, including superoxide (Ghiladi et al., 2005), hydrogen peroxide (Zhao et al., 2006) and simple alkyl hydroperoxides (Wengenack and Rusnak, 2001). Even in the absence of added oxidants, the in vitro auto-oxidation of INH (Winder and Denneny, 1959) and NADH, when used (Zhao et al., 2006), can provide sufficient oxidants to allow INH activation by KatG. It is likely that the oxidant in vivo is a low flux of hydrogen peroxide that might form within the bacteria as a by-product of aerobic metabolism (Zhao et al., 2006). However, intracellular-formed superoxide might also have a direct role (Ghiladi et al., 2005). As both these superoxide- and low-flux hydrogen peroxide-oxidizing systems show the expected decrease in the ability of mutant S315T KatG to activate INH in vitro (in contrast to alkyl hydroperoxides which do not) (Wengenack and Rusnak, 2001), resolution of exactly which of these species is the oxidant in vivo is not yet possible. It would also appear that other model-oxidizing systems are capable of activating INH, such as horseradish peroxidase (Zinner et al., 1977; Sinha, 1983) or even inorganic manganese ions (Nguyen et al., 2002a; Broussy et al., 2003). However, it is uncertain whether these model oxidants accurately simulate the exact species formed by M. tuberculosis KatG, as even highly related KatG enzymes can differ in the products they produce (Kang et al., 2006). Additionally, a role for manganese in vivo is unlikely as although it can reach high concentrations in some bacteria (Archibald, 1986; Tseng et al., 2001) it does not in mycobacteria (Wagner et al., 2005). The crystal structures of KatG (Bertrand et al., 2004; Pierattelli et al., 2004) and recently the S315T variant (Zhao et al., 2006) provide better insight into the mechanisms of INH activation. The narrowing of the haem access channel from 6 in the wild type to 4.7 in the S315T variant suggests that decreased INH access to the oxidizing site of KatG might be key in resistance, in accord with previous spectroscopic investigations (Sherman et al., 1999; Lukat-Rodgers et al., 2000; Yu et al., 2003).
Reactive species formed by KatG activation of INH
A range of elegant studies have characterized the stable products of INH oxidation by both KatG and model oxidants (Johnsson and Schultz, 1994; Johnsson et al., 1995; Magliozzo and Marcinkeviciene, 1996; Bodiguel et al., 2001), but such studies can only be used to infer the nature of the key reactive INH-derived intermediates that bring about INH activity. In the absence of NAD+ or NADP+, no significantly antimycobacterial INH-derived stable products are formed during in vitro activation. Thus, INH-derived reactive intermediates are central to INH action. KatG oxidatively activates INH via the production of a range of carbon-, oxygen- and nitrogen-centred free radical species, with the formation of acyl, acylperoxo and pyridyl radical adducts of phenylbutylnitrone (PBN) proposed from results of spin trapping experiments (Wengenack and Rusnak, 2001). However, the discrimination of different adducts of PBN by hyperfine coupling constant alone is difficult, and so these assignments remain somewhat tentative (Buettner, 1987). The spin trap 5,5-dimethyl-1-pyrolline-N-oxide (DMPO) often allows better assignments than PBN, as the range of hyperfine coupling constants of its spin adducts are wider. Accordingly, INH-derived species assigned as carbon-centred and alkoxyl adducts of DMPO have been observed upon KatG oxidation of INH (Timmins et al., 2004a), with an additional peroxyl radical species observed in the presence of molecular oxygen, resulting from reaction of one of these radicals with O2. The nitric oxide radical (NO·) has also been trapped from KatG oxidation of INH, with 15N isotopic labelling of INH hydrazide resulting in 15NO· (Timmins et al., 2004a,b). The results from other oxidizing systems suggest that the hydrazyl radical (2 in Fig. 1) (Sipe et al., 2004) can be readily formed by INH oxidation and is likely an initial product with KatG also, although confirmation with authentic KatG is required. Despite these advances, there remains a need to characterize the nature of these radical intermediates, and in particular, to demonstrate definitively the production of the isonicotinoyl radical (3 in Fig. 1) (Nguyen et al., 2002b), as this is the key intermediate that adds to NAD+ and NADP+ to produce a range of powerful inhibitors. There are several important antimycobacterial effects of these INH-derived radicals that will be discussed in detail.
Direct actions of INH-derived radicals
Mycobacterium tuberculosis is a natural ‘oxyR knockout’ (Deretic et al., 1995). This results in a high sensitivity of M. tuberculosis, and other mycobacteria with defective oxyR (Dhandayuthapani et al., 1996) to oxidative stress and to INH (Deretic et al., 1996), as well as to nitric oxide (Chan et al., 1992; Yu et al., 1999; Master et al., 2002). Furthermore, some KatG catalytic activities can be inhibited by the presence of INH (Marcinkeviciene et al., 1995) and, instead, KatG acts as an efficient producer of the range of INH-derived radicals previously discussed. This led to the hypothesis that these INH-derived oxygen- and carbon-centred free radicals are directly important in mycobacterial cell killing (Sinha, 1983; Shoeb et al., 1985a,b,c) because they damage a range of cellular components such as lipids, protein and nucleic acids. However, the focus upon mechanisms by which INH inhibits cell wall lipid synthesis, coupled with the findings that inhibitory INH adducts of NAD+/NADP+ were formed from the isonicotinoyl radical (see next heading), led the field away from this area. Paradoxically, Mycobacterium smegmatis mutants deficient in synthesis of mycothiol, the major low-molecular-weight antioxidant in M. tuberculosis, became highly resistant to INH (Rawat et al., 2002), while their sensitivity to other oxidants showed the expected increase, apparently contradicting a direct role for INH-derived radicals.
We recently showed that the known antimycobacterial free radical NO· is formed from INH by KatG-mediated oxidation, that an NO· scavenger can partially protect against INH and that known targets of NO·, key mycobacterial respiratory enzymes, are affected in INH-treated mycobacteria (Timmins et al., 2004a,b). As addition of NO·-releasing moieties to other antibiotics increases their antimycobacterial effects (Ciccone et al., 2003), we proposed that this INH-derived NO· might synergize with the other actions of INH-derived species to contribute to overall INH action. We further hypothesize that some of the seemingly contradictory effects of mycothiol can be explained through NO·. The diffusional path length of NO· is such that, after formation, most would simply exit the mycobacterial cell and be oxidatively detoxified in the growth medium (Lancaster, 1997; Thomas et al., 2001; Lancaster and Gaston, 2004) to produce nitrite/nitrate that the bacilli are well adapted to deal with. However, the presence of large amounts of intracellular mycothiol causes a large amount of NO· to become ‘trapped’ as intracellular S-nitrosomycothiol that can then transnitrosylate and affect a range of intracellular molecules (Hess et al., 2001; 2005) (Fig. 2). Lowered mycothiol levels could thus allow efflux of INH-derived NO·, and help explain the paradoxical INH resistance of mycothiol mutants. The presence of an S-nitrosomycothiol reductase (Vogt et al., 2003) confirms that S-nitrosomycothiol is a noxious species that mycobacteria have evolved specific detoxification mechanisms to remove. Thus, although the major effects of carbon- and oxygen-centred INH derived radicals are mediated through reactions with NAD+ or NADP+, INH-derived NO· appears able to act directly to complement enzyme inhibitions caused by these species. However, further delineation of the pathway of NO· formation, its mycobacterial targets and of the extent to which NO· contributes to the overall INH effect are needed for a complete assessment of its role.
Formation and actions of NAD+ and NADP+ adducts
It had long been known that INH disrupts synthesis of both mycolic acids (Winder et al., 1970; Takayama et al., 1972) and nucleic acids (Gangadharam et al., 1963) although it took some time to determine the mechanisms. The finding that oxidation of INH in the presence of NADH and InhA led to covalent INH-NADH adducts that are powerful inhibitors of InhA (Rozwarski et al., 1998) was a major advance. InhA is an enoyl acyl carrier protein reductase (Banerjee et al., 1994; Dessen et al., 1995) involved in the synthesis of mycolic acids, unique and important mycobacterial cell wall lipids, and so its inhibition is in accord with the unique sensitivity of mycobacteria to INH. It was later shown that these adducts could be formed during oxidation of INH in the presence of NADH in free solution, through Minisci addition of the isonicotinoyl radical to NAD+ (Wilming and Johnsson, 1999; Zhao et al., 2006) that occurs at rates ∼106 M−1 s−1 (Minisci et al., 1980) (Fig. 3). The role of NAD+ in formation of INH-NAD adducts is also supported by the metabolic effects of NADH/NAD+ ratios on INH sensitivity in mycobacteria (Miesel et al., 1998), in which lowered NAD+ levels conferred INH resistance, and also by mutations in the NADH oxidase gene ndh in some INH resistant isolates (Lee et al., 2001). INH-NAD adducts capable of InhA inhibition are produced both in simple Mn/INH/NADH mixtures (Bodiguel et al., 2001; Nguyen et al., 2002a,b) and in KatG-catalysed reactions. The addition of the isonicotinoyl radical to NAD+ generates a stereochemical centre in the INH-NAD adduct, and it is the S isomer that binds to InhA (Rozwarski et al., 1998) as a tight-binding (Ki = 0.75 nM) inhibitor (Rawat et al., 2003). Although subsequent cyclization of these two initial INH-NAD adducts generates a range of diastereoisomers (Nguyen et al., 2002a; Broussy et al., 2003; 2005), these do not appear to have the biological activity of the acyclic S isomer in inhibiting InhA. It would be predicted that addition of activated INH to NADP+ could also occur and in vitro these INH-NADP adducts strongly inhibit MabA, an NADPH-dependent β-ketoacyl-ACP reductase that is also central in mycolic acid biosynthesis (Ducasse-Cabanot et al., 2004). Thus, INH adducts of both NAD+ and NADP+ could inhibit different steps in cell wall lipid synthesis, although an in vivo role for MabA inhibition awaits demonstration. Furthermore, it has recently been shown that the acyclic 4R isomer of the INH-NADP adduct binds M. tuberculosis dihydrofolate reductase (DHFR), with a Ki of under 1 nM (Argyrou et al., 2006). As DHFR is central in nucleic acid biosynthesis to make nucleotide pools, and not to mycolic acid synthesis, INH-NAD/NADP adducts might be involved in inhibition of a range of cellular processes, and elucidation of further targets, especially for the more recently characterized adducts of NADP+ would seem worthwhile. Although it appeared that an INH adduct of KasA, another mycolic acid synthetic enzyme, might occur through an INH-induced 80 kDa covalent complex comprising KasA, AcpM and an INH-derived isonicotinic acyl fragment (Mdluli et al., 1998), it seems likely that effects on KasA are mediated via INH-NAD adduct inhibition of InhA (Kremer et al., 2003) as a result of its known close regulation with InhA (Slayden et al., 2000). The relative importance of InhA and KasA as targets has recently been further clarified by transferring clinically observed mutations, such as inhA(S94A) into wild-type mycobacteria (Vilcheze et al., 2006). Introduction of inhA(S94A) increased INH resistance, while kasA mutants G269S and F413L did not.
Thus, a range of powerful INH-NAD/NADP inhibitors have been characterized, and in general, the requisite genetic modifications of target mycobacteria have resulted in the expected alteration in INH sensitivity (Table 1). However, these in vitro systems used a high concentration of NAD+/NADP+ and a very low concentration of other molecules capable of reacting with the isonicotinic acyl radical, thereby ‘forcing’ the isonicotinoyl radical to react with NAD+ or NADP+. However, the isonicotinoyl radical will be highly reactive with a wide range of reactants as for other acyl radicals (Boger and Mathvink, 1992; Brown et al., 1995) such predicted competitors would include proteins, mycothiol and unsaturated lipids (Davies, 2005). So, when formed inside the mycobacteria, the isonicotinoyl radical will instead react predominantly with molecules other than NAD+ or NADP+, which are present at only ∼0.4 mM and 0.2 mM, respectively, in M. tuberculosis (Gopinathan et al., 1963). Although the accumulated evidence is convincing, final proof of the role of these INH-NAD and INH-NADP species will require their isolation from INH-treated mycobacteria at concentrations consistent with a significant inhibitory effect from their known Kis.
Table 1. Summary of the evidence supporting different adducts of NAD+ and NADP+ as inhibitors of differing targets.
Target of inhibition
Confirmatory experimental evidence
For clarity, references are as in text.
In vitro binding and inhibition,inhA mutations confer INH resistance, genetic insertion of S94A inhA confers INH resistance
In vitro binding and inhibition, overexpression of DHFR confers INH resistance
In vitro binding and inhibition
Although delineation of the mechanisms of action of isoniazid required many years, it can be seen that convincing progress has been made in the last decade through the combined approaches of bacterial genetics, biochemical work and detailed free radical chemistry. The inhibition of both cell wall lipid, and nucleic acid synthesis by INH-NAD and INH-NAPD adducts, together with respiratory inhibition by INH-derived NO·, provides a potent antituberculosis cocktail that is so effective because it attacks multiple targets. Looking ahead, it would appear that INH is already an optimal substrate for KatG for conversion to species such as the isonicotinoyl radical, and so logical avenues for drug development involving INH could include:
(i) developing agents that produce the isonicotinoyl radical, independently of KatG, for treatment of KatG mutant drug-resistant TB, although it may be difficult to achieve sufficient mycobacterial specificity;
(ii) screening for molecules that increase mycobacterial levels of NAD+ or NADP+ for use in co-administration with INH, to maximize the yield of INH-NAD(P) adducts from the limited amount of isonicotinoyl radical generated in variants such as S315T;
(iii) using the structure of INH-NAD(P) adducts to design more drug-like molecules to inhibit specifically mycobacterial enzymes such as InhA or DHFR; and
(iv) developing inhibitors of arylamine N-acetyltransferase, a mycobacterial enzyme that can inactivate INH and that may be involved in resistance (Sholto-Douglas-Vernon et al., 2005).
Thus, the fundamental understanding of mechanisms, while important in itself, will also have allowed the development of new treatment strategies.
This work was supported by NIH Grants AI42999 (V.D.) and AI63486 (G.S.T.).