Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
Bringing a knowledge of enzymology into research in vivo and in situ is of great importance in understanding systems biology and metabolic regulation. The central metabolic significance of thiamin (vitamin B1) and its diphosphorylated derivative (thiamin diphosphate; ThDP), and the fundamental differences in the ThDP-dependent enzymes of metabolic networks in mammals versus plants, fungi and bacteria, or in health versus disease, suggest that these enzymes are promising targets for biotechnological and medical applications. Here, the in vivo action of known regulators of ThDP-dependent enzymes, such as synthetic structural analogs of the enzyme substrates and thiamin, is analyzed in light of the enzymological data accumulated during half a century of research. Mimicking the enzyme-specific catalytic intermediates, the phosphonate analogs of 2-oxo acids selectively inhibit particular ThDP-dependent enzymes. Because of their selectivity, use of these compounds in cellular and animal models of ThDP-dependent enzyme malfunctions improves the validity of the model and its predictive power when compared with the nonselective and enzymatically less characterized oxythiamin and pyrithiamin. In vitro studies of the interaction of thiamin analogs and their biological derivatives with potential in vivo targets are necessary to identify and attenuate the analog selectivity. For both the substrate and thiamin synthetic analogs, in vitro reactivities with potential targets are highly relevant in vivo. However, effective concentrations in vivo are often higher than in vitro studies would suggest. The significance of specific inihibition of the ThDP-dependent enzymes for the development of herbicides, antibiotics, anticancer and neuroprotective strategies is discussed.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
A number of reactions involving C–C bond formation/scission are catalyzed by enzymes employing the diphosphorylated derivative of thiamin, thiamin diphosphate, as a coenzyme (Table 1). Thiamin is synthesized by microorganisms, fungi and plants, but is an essential nutrient (vitamin B1) in animals [1, 2]. The isolation of thiamin, determination of its chemical structure and its in vitro synthesis were propelled by identification of its deficiency in humans as the cause of polyneuritis (beriberi). Thiamin is composed of pyrimidine (4-amino-2-methyl-5-hydroxymethylpyrimidine) and thiazole [4-methyl-5-(2-hydroxymethyl)-thiazole] rings linked by a methylene bridge (Fig. 1). All living organisms possess enzymatic systems to form a variety of phosphorylated and adenylated derivatives of thiamin (Fig. 1): thiamin monophosphate (ThMP), thiamin diphosphate (ThDP), thiamin triphosphate (ThTP), adenosine thiamin triphosphate and adenosine thiamin diphosphate . However, to date, the widely recognized biological roles have been ascribed to ThDP only. ThDP generally represents a major part of the total thiamin pool in vivo, and is well-known as a coenzyme of the enzymes of central metabolism. Recently, it was also shown to be a regulatory ligand binding to the riboswitch segment of premature mRNA, and affecting RNA transcription, translation or splicing in bacteria, plants and some fungi [4-7]. In view of the major contribution to the pool and recognized biological significance of ThDP, it is probably not surprising that of the multiple enzymes catalyzing the transformation of thiamin to its phosphorylated and adenylated derivatives (Fig. 1), only the two generating ThDP have been identified at the molecular level in animals. These are thiamin diphosphokinase (EC 184.108.40.206) and thiamin triphosphatase (EC 220.127.116.11), which catalyze the ATP-dependent diphosphorylation of thiamin and ThTP hydrolysis, respectively. Much less is known about the formation and biological roles of the other thiamin derivatives. ThMP is transported in blood plasma and is able to penetrate the blood–brain barrier . Remarkably, in erythrocytes, ThMP may be produced not only by ThDP hydrolysis, but also from thiamin using β-glycerol phosphate or creatine phosphate as the phosphate donors . ThMP is also the product of thiamin biosynthesis. The phosphorylation of ThMP by thiamin phosphate kinase (bacterial ThiN) in thiamin-synthesizing organisms adds to the ThDP formation from thiamin by thiamin diphosphokinase which is present in all species . Taken together, these data imply that ThMP likely plays a biological role. Furthermore, concerted changes in ThTP and its adenylated derivatives have been shown to be associated with the metabolic stress response in bacteria [10, 11]. In addition, it has long been known that the corelease of thiamin and acetylcholine facilitates synaptic transmission [[12, 13] and refs therein]. This non-coenzyme function of thiamin was suggested to involve as yet unidentified thiamin-binding proteins of synaptosomal plasmatic membrane. Supposedly, these proteins catalyze ThTP hydrolysis and/or phosphorylation of synaptic proteins with ThTP acting as a phosphate donor [13-17]. Finally, thiamin or ThMP-binding proteins of thiamin-synthesizing organisms with a transport/storage function have also been reported [[1, 18] and refs therein].
Table 1. Characterized ThDP-dependent enzymatic reactions of mammals vs other organisms
Commonly used abberviation
Metabolic pathway and/or catalyzed reaction
In photosynthetic organisms, a specific isoform of transketolase also exists and is involved in carbon fixation in the Calvin–Benson cycle.
Mammals and other kingdoms
Pyruvate dehydrogenase (component E1p of pyruvate dehydrogenase complex)
Pyruvate entry into the tricarboxylic acic cycle, oxidative decarboxylation of pyruvate
2-Oxoglutarate dehydrogenase (component E1o of 2-oxoglutarate dehydrogenase complex)
Tricarboxylic acic cycle, oxidative decarboxylation of 2-oxoglutarate
Carbohydrate catabolism, pentose fermentation, xylulose-5-phosphate cleavage into acetyl phosphate and glyceraldehyde-3-phosphate
Benzaldehyde lyase (benzoin aldolase)
Reversible conversion of (R)-benzoin (2-hydroxy-1,2-diphenylethanone) into two molecules of benzaldehyde
Thiamin deficiency, whether caused by dietary deficiency in animals or through a biosynthesis blockade in plants, bacteria and fungi, impairs central metabolism and is incompatible with life. Severe thiamin deficiency in humans is rarely observed nowadays. However, B1 hypovitaminosis does occur, often as a result of dietary deficiencies, the effects of certain diseases, excessive use of diuretics or alcohol abuse [19-21]. Even in highly developed societies, the risk of insufficient levels of thiamin remains significant in the elderly , patients after major surgery [23, 24], pregnant and breastfeeding women , smokers, diabetics and people eating in a high carbohydrate diet . Although the link between thiamin deficiency and serious neurological impairments, such as beriberi or Wernicke–Korsakoff syndrome has long been known, a number of recent studies have also revealed a significant impact of thiamin hypovitaminosis on chronic pathologies associated with systemic changes in metabolism. Lower concentrations of thiamin and its phosphate esters in whole blood have been associated with the occurrence of depressive symptoms . An appropriate dosage of thiamin inhibits the progression and reduces the symptoms of neurodegenerative diseases [28, 29]. By normalizing the metabolism of lipids and carbohydrates, thiamin may significantly delay vascular and metabolic complications caused by chronic diabetes . Clearly, thiamin has a major impact on the multitude of physiological functions of living organisms. Although the central role of the ThDP-catalyzed enzymatic reactions in homeostasis suggests that ThDP-dependent enzymes will be promising targets of metabolic regulation, the less-characterized systems operating with the non-coenzyme forms of thiamin must also be taken into account. In particular, this is important when considering the effects of thiamin on metabolism and/or developing tools to specifically affect metabolic reactions in which ThDP acts as coenzyme. The aim of this minireview is to summarize and critically assess the existing tools and the mechanistic interpretations of metabolic regulation achieved through targeting ThDP-dependent enzymes, including relevant issues of drug design and disease models.
Natural mechanisms of metabolic regulation involve ThDP-dependent enzymes
The central metabolic significance of ThDP-dependent enzymes underlies certain mechanisms of metabolic regulation which exist in living systems. Analysis of published data indicates that metabolic reprogramming, such as occurring in cancer cells [31, 32] or chronic alcoholism , underlying circadian rhythms  or upon stress response [34-37], involves regulation of the thiamin metabolism and/or ThDP-dependent enzymes. Regulation is realized through different mechanisms, but eventually induces metabolic switches caused by the changed function of ThDP-dependent enzymes. In Corinebacteria and Mycobacteria interaction of the ThDP-dependent 2-oxoglutarate dehydrogenase (OGDH) with a protein inhibitor whose function is regulated by (de)phosphorylation, switches carbon flux between the glutamate synthesis and oxidation [38, 39]. Significant activation of the ThDP-synthesizing enzyme thiamin diphosphokinase, which is accompanied by increased transketolase activity, occurs in plants responding to abiotic stress . Under conditions of abiotic stress, the mRNAs of ThDP-dependent enzymes were shown also to increase in plants and yeast [35, 40]. In animals, too, metabolic reprogramming and adaptations are associated with increased availability of thiamin/ThDP which, in turn, affects the function of the ThDP-dependent enzymes. For instance, cancer cells increase expression of the thiamin transporter thereby elevating intracellular thiamin, and supposedly providing for stimulation of transketolase [31, 32, 41]. Chronic alcoholism activates the thiamin transport and diphosphorylation in rat brain, which may partially compensate for inactivation of pyruvate dehydrogenase complex (PDHC) and 2-oxoglutarate dehydrogenase complex (OGDHC) under these conditions . Different metabolic stresses, as well as i.p. injection of thiamin, were found to affect the ThDP-dependent OGDHC, with the level of the OGDHC activity in brain shown to be coupled to changes in the main excitatory neurotransmitter glutamate and heart performance .
In the thiamin-synthesizing organisms, metabolic regulation involves the ThDP riboswitch. It enables these living systems to respond to changes in concentration of ThDP by changing expression of proteins involved in the thiamin biosynthesis, transport and precursor salvage. These changes were, in turn, shown to affect core metabolism via the ThDP-dependent enzymes. For instance, in plants the ThDP riboswitch controls circadian rhythms in the thiamin levels which are coupled to changes in core metabolism through regulation of activities of the ThDP-dependent enzymes . In animals, thiamin transporter 1 was shown to be a direct transcriptional target for transcription regulator p53 , with the binding of p53 to DNA inhibited by ThDP . Recently, p53 was also shown to be involved in the upregulation of thiamin transporter 1, pyruvate dehydrogenase (PDH) and OGDH upon impairment of thiamin metabolism in mammalian SH-SY5Y cells . Thus, although the ThDP riboswitch is limited to thiamin-synthesizing organisms only, animals also possess mechanisms of the ThDP-dependent regulation of protein expression. All these mechanisms respond to perturbations in thiamin metabolism, with upregulation of the ThDP-dependent enzymes being an important component of the response. Thus it can be seen that ThDP is a universal systemic regulator at the transcriptional, translational and post-translational levels, acting through essential impact of ThDP-dependent enzymes on central metabolism. This strongly supports the idea that ThDP-dependent enzymes are promising targets for directed metabolic regulation to address medical, bioengineering and biotechnological goals. The existing approaches and examples of such regulation are considered below.
ThDP-dependent enzymes in the directed metabolic regulation
Non-mammalian ThDP-dependent enzymes as targets for herbicides, fungicides and antimicrobial compounds
As seen from Table 1, all kingdoms employ the ThDP-dependent enzymes in central catabolic pathways (Fig. 2A). By contrast, participation of ThDP-dependent enzymes in central anabolic pathways, such as the branched chain amino acid biosynthesis and nonmevalonate biosynthetic pathways or the photosynthesis-associated Calvin–Benson cycle of carbon fixation (Fig. 2B), has only been observed in non-mammalian species. Such distribution of ThDP-dependent enzymes between catabolic and anabolic pathways correlates with the ability of species to synthesize thiamin. Thus, the metabolism of mammals, which obtain thiamin exogenously, has not evolved to rely on ThDP-dependent biosynthesis. Conversely, essential ThDP-dependent biosynthetic processes are present in microorganisms, fungi and plants, which have the ability to synthesize thiamin. The latter group has also acquired the ThDP riboswitch which regulates thiamin biosynthesis according to metabolic demands [5-7]. This species-specific occurrence of both ThDP-dependent anabolic pathways and thiamin biosynthesis means that the thiamin-dependent enzymes of these pathways offer potential as targets for drug design.
An excellent example of this is the fact that plants, fungi and bacteria use the ThDP-dependent enzyme acetohydroxyacid synthase  in the biosynthesis of branched chain amino acids (Table 1, Fig. 2B). These amino acids are not synthesized by mammals, which lack acetohydroxyacid synthase (Table 1). The enzyme may therefore be a target for the development of herbicides, fungicides and antimicrobial compounds . Indeed, shortly after the sulfonylurea herbicides (e.g. metsulfuron methyl, Fig. 3) were discovered through an extensive screening program , acetohydroxyacid synthase was revealed as their target [48, 49]. Crystallization of enzyme–herbicide complexes has shown that the herbicides block the deep hydrophobic tunnel to the active site through multiple interactions with the amino acid residues (Fig. 4). Resistance to herbicides arises when these amino acids are mutated. The available structures of the enzyme–herbicide complexes provide a rational molecular basis to improve the drug design and avoid mutation-induced resistance [50, 51]. In an interesting development, some inhibitors of acetohydroxyacid synthase have also been shown to have antibacterial and particularly antitubercular effects .
Another herbicide target among the ThDP-dependent enzymes not present in mammals is 1-deoxy-d-xylulose 5-phosphate synthase . Involved in the nonmevalonate pathway in plants and microorganisms, this enzyme synthesizes deoxy-d-xylulose-5-phosphate, a key precursor for isoprenoids, thiamin and pyridoxal (vitamin B6) (Table 1, Fig. 2B). Occupying such a key position, the enzyme is inhibited by the herbicide, 5-ketoclomazone, which is also produced in vivo from another herbicide, clomazone [54, 55]. Suppression of the antibacterial activity of 5-ketoclomazone by 1-deoxyxylulose, whose in vivo phosphorylation may generate 1-deoxy-d-xylulose-5-phosphate, points to 1-deoxy-d-xylulose-5-phosphate synthase as the in vivo target of 5-ketoclomazone . It is notable that the inhibition kinetics exhibited by 5-ketoclomazone indicated that the binding site of the inhibitor is different to the binding sites of the two substrates, pyruvate and d-glyceraldehyde 3-phosphate . Because 5-ketoclomazone has a certain structural similarity to thiamin (Fig. 1) and its known antagonists (Fig. 3), the observed inhibition kinetics may be because of competition between 5-ketoclomazone and ThDP at the coenzyme binding site of 1-deoxy-d-xylulose-5-phosphate synthase.
Metronidazole (Fig. 3) is an antibiotic used to fight certain bacterial and protozoan infections. This antibiotic occupies the thiazole-binding site of the pathogen thiaminase (EC 18.104.22.168 for thiaminase I), which is absent in mammalian species. Thiaminase substitutes the thiazole ring in thiamin for metronidazole, producing the thiamin analog (1-[(4-amino-2-methyl-5-pyrimidinyl)methyl]-3-(2-hydroxyethyl)-2-methyl-4-nitr-oimidazole) which efficiently inhibits thiamin diphosphokinase [56, 57]. The inhibition of thiamin diphosphokinase is supposed to cause thiamin deficiency which impairs ThDP-dependent enzymes. Furthermore, the N-alkylation of metronidazole by thiaminase increases drug electrophilicity, thereby stimulating the reduction of the product to chemically reactive species. This mechanism may also contribute to the antipathogen action of metronidazole . Because the thiamin analog is only synthesized by pathogens, this provides for the species-specific differences needed for the antibiotic application to treat infections. Nevertheless, the prolonged use of metronidazole is known to have side effects [56, 57]. It is therefore likely that accumulation of the metronidazole-generated thiamin analog may affect the thiamin-synthesizing gut microflora and the host organism itself. In particular, a side effect of DNA damage was shown in human lymphocytes. The effect was ascribed to the metronidazol one-electron reduction to a nitro radical anion, whose subsequent reoxidation by oxygen generated reactive oxygen species (ROS) . However, no significant cytotoxic or genotoxic effects on cultured human cells have been shown in other studies . The discrepancy suggests that the reactive metabolites of intracellular reduction of metronidazol do not always accumulate to a level that would induce cell damage. Most probably, the cytotoxic and genotoxic effects in vivo require the enzymatic alkylation of metronidazole, which both augments electrophilicity of the heterocyclic weak base and induces thiamin deficiency .
Structural analogs of thiamin as tools of directed metabolic regulation
In view of the central role of ThDP in the homeostasis of living systems (Fig. 2), it has long been realized that compounds interfering with the thiamin synthesis, transport, phosphorylation or ThDP coenzyme action might be efficient drugs affecting viability [60-62]. The majority of such compounds (Fig. 3) are structurally similar to thiamin or its fragments, and compete for the thiamin/ThDP-binding sites of living systems. The variety of the structural analogs of thiamin presented in Fig. 3 requires compound-specific synthetic schemes. In general, iminium salts (such as pyrithiamin or amprolium in Fig. 3) are easily formed during alkylation of the correspondingly substituted pyridines by appropriate 5-chloromethylpyrimidines [63, 64]. A similar alkylation of the substituted thiazole was also used for synthesis of a number of the thiamin analogs with modified pyrimidine ring, such as N3′-pyridil thiamin [65, 66]. Although this approach is the most universal, other routes were exploited as well. For instance, oxythiamin was first synthesized by the condensation of 4-hydroxy-5-thioformamidomethyl-2-methylpyrimidine with 3-bromo-5-acetoxypentan-2-one . Later it was shown that oxythiamin can also be obtained from thiamin after boiling in diluted HCl . Tetrahydrothiamin was produced with a good yield by the reduction of thiamin with NaBH4 in an aqueous medium . Thiamin thiazolone was also prepared by modification of thiamin . For the synthesis of 5-ketochlomazone and 3-dezazathiamin, specific heterocyclization reactions were used in the final steps [71, 72]. Metronidazole is the product of the nucleophilic opening of ethylene oxide by 2-methyl-5-nitroimidazole in acidic medium .
Studies on the antimetabolic (anticoenzyme) action of the thiamin structural analogs, so-called thiamin antagonists, have been performed in vitro (isolated enzymes), in situ (cellular organelles, cell cultures, tissues) and in vivo (living organisms) (Tables 2 and 3). In vitro, the thiamin structural analogs were routinely used to decipher the mechanism of catalysis by ThDP-dependent enzymes (reviewed in ). However, to date, little work has been done that couples enzymology and systems biology to the extent necessary for implementing the target-specific thiamin analogs into in vivo studies. This is caused by several problems. First, enzymological studies revealed varied sensitivities of different ThDP-dependent enzymes  or of different catalytic activities of a particular ThDP-dependent enzyme  to certain thiamin modifications (e.g. diphosphorylated tetrahydrothiamin and 3-dezazathiamin in Table 2). However, Table 2 exposes insufficiency of the available information on the affinities of the diphosphorylated thiamin analogs to ThDP-dependent enzymes. Even the enzymes of central metabolism were not systematically characterized in this regard to reveal the enzyme-specific analogs for in vivo applications. Moreover, as Table 3 shows, in vivo studies have mostly been carried out with analogs that do not exhibit significant selectivity to target proteins. Second, the analogs of thiamin, which are most commonly used in situ and in vivo (pyrithiamin, oxythiamin and amprolium, Table 3), are poorly characterized regarding not only their potential interactions with the multitude of thiamin/ThDP-binding systems, but also their intracellular transformation. For example, pyrithiamin is a strong inhibitor of thiamin diphosphokinase. As such, it is widely used to create animal models of thiamin deficiency under the erroneous belief that it is not diphosphorylated by the kinase. However, the strong inhibition of the thiamin diphosphorylation by pyrithiamin (Table 3) does not mean that pyrithiamin itself is not phosphorylated. Indeed, that pyrithiamin acts as a substrate of the thiamin diphosphokinase has long been known from in vitro studies . Recently, mass spectrometry was used to confirm the enzyme-catalyzed formation of pyrithiamin diphosphate, whereas the X-ray structure (Fig. 5A) shows that pyrithiamin diphosphate is formed at the active site after binding pyrithiamin and ATP to thiamin diphosphokinase . Recent in vivo findings have also demonstrated that the amount of pyrithiamin diphosphate formed by thiamin diphosphokinase is physiologically significant, as pyrithiamin diphosphate competes with ThDP for the ThDP riboswitch in the pyrithiamin-exposed bacteria (reviewed in [53, 56]). Some discrepancies that may be attributed to pyrithiamin diphosphorylation, also exist in animal models of pyrithiamin-induced thiamin deficiency. For example, among mammalian ThDP-dependent enzymes, OGDHC has the strongest affinity to ThDP. Nevertheless, after pyrithiamin treatment, OGDHC is the slowest to recover activity when thiamin levels are restored. If the pyrithiamin action was solely because of reduced ThDP levels (as occurs in alimentary thiamin deficiency), OGDHC should be the last to be inactivated by losing its tightly bound ThDP and the first one to be restored after thiamin repletion. It would appear that the persistent inactivation of OGDHC upon in vivo pyrithiamin administration is because of a strong interaction of the ThDP-binding sites of OGDHC with pyrithiamin diphosphate formed in vivo rather than just an increase in OGDHC with the ThDP-binding sites being empty [134 and refs therein]. However, as Table 3 shows, there are no experimental data on the interaction of pyrithiamin diphosphate with mammalian ThDP-dependent enzymes. The gaps should therefore be filled in to correctly interpret the results of in vivo studies involving pyrithiamin. Unlike dietary thiamin deficiency, the thiamin antagonists that can be diphosphorylated in vivo may cause effects beyond just depleting the ThDP pool. Furthermore, although certain quantitative features in the thiamin uptake inhibition by oxythiamin, pyrithiamin and amprolium have been noticed [8, 78, 79], the overall data published (Table 3) do not allow us to experimentally resolve their action on thiamin uptake and thiamin-dependent intracellular processes in vivo, especially at higher doses/treatment times. Earlier studies suggested that oxythiamin does not penetrate the blood–brain barrier [79, 80]. This was probably caused by a short treatment time in these experiments, as later studies have disproved this assumption . Nevertheless, concerning the selectivity of the in vivo effects, amprolium has an important advantage over pyrithiamin and oxythiamin. Notably, amprolium (Fig. 3) does not have the hydroxyethyl group undergoing diphosphorylation in thiamin (Fig. 1). Therefore, no intracellular diphosphorylation of amprolium can occur, in contrast to oxythiamin and pyrithiamin. Because of this feature, amprolium could be considered as the analog affecting the thiamin-dependent processes rather than the ThDP-dependent ones. Indeed, the amprolium-induced blockade of parasitic thiamin transport is used in veterinary applications to fight coccidiosis . In addition, amprolium also inhibits the thiamin diphosphorylation (Table 3). Obviously, these two effects of amprolium deplete the intracellular ThDP pool. However, the lack of a diphosphorylation site guarantees that binding of amprolium to the ThDP-dependent enzymes is not efficient. For this reason, the in vivo action of amprolium would be a better model of dietary thiamin deprivation than the widely used model of thiamin deficiency induced by pyrithiamin.
Table 2. Ki values (μm) for the inhibition of ThDP-dependent enzymes by selected diphosphorylated thiamin analogs in vitro
Thiamin analog (diphosphorylated)
In the 2–hydroxy-3-oxoadipate synthase reaction catalyzed by OGDH.
Table 3. Characteristics of the actions of oxythiamin, pyrithiamin and amprolium in vitro (extracted or purified enzymes), in situ (cell fractions, cell cultures, tissues) and in vivo (yeast, plants, animals). For the in vivo experiments in yeast, the concentration in the culture medium is indicated in μm. The order of the in situ and in vivo data in each subsection corresponds to increasing total dose, estimated as the product of the maximal concentration and time of treatment
23% inhibition of PDHC, 45–56% decrease in acetyl-CoA level
Finally, although specific molecular events and biochemical reactions underlying thiamin involvement in its non-coenzyme function in living systems are not well characterized, these aspects cannot be ignored when considering the consequences of thiamin deficiency or designing models of thiamin-related pathologies. For example, recent data have shown that adenosine thiamin triphosphate can inhibit poly-ADP-ribose polymerase 1 with Ki = 10 μm. This links the non-coenzyme thiamin derivatives to the well-known regulator of cell death and hence pharmacological target poly-ADP-ribose polymerase 1 . However, current use of thiamin structural analogs as pharmacological agents or in studies on disease mechanisms does not take into consideration the possibility that, in vivo, such analogs may be not only diphosphorylated, but also transformed to other derivatives (Fig. 1). The interaction with the proteins beyond the ThDP-dependent enzymes is not taken into account either.
Similar to the inhibition of thiamin diphosphorylation by thiamin diphosphokinase in the presence of the enzyme alternative substrate pyrithiamin, observation of the inhibition of thiamin transport by thiamin analogs does not necessarily imply that the analogs themselves are not transported. Indeed, all of the thiamin analogs presented in Table 3 are transported and may therefore affect currently unidentified enzymes that produce intracellular derivatives of thiamin (Fig. 1). A true blocker of the thiamin transporter is unknown, although this type of compound would be the best one to model thiamin deprivation. If the compound does not enter a cell, it will not compete with either thiamin or its natural derivatives (Fig. 1) for any of the binding sites. Furthermore, in vivo, such a compound would prevent the tissue redistribution of thiamin. This process is known to interfere with thiamin deprivation in the brain , complicating experimental perturbation of thiamin status in the nervous system. An alternative approach to thiamin deprivation in mammalian cells, which excludes the cellular uptake of thiamin analogs, was developed using extracellular thiamin degradation by exogenous thiaminase, which is absent in mammals .
Representing a summary of published research on thiamin antagonists, Table 3 exposes a certain preference for the use of oxythiamin as an anticancer agent, pyrithiamin as an inducer of thiamin deficiency and amprolium as the thiamin uptake inhibitor. However, our comparative analysis of the data accumulated over time (Table 3) indicates that this selectivity has been based on nonvalidated presumptions regarding the mechanisms of action of the thiamin antagonists in vivo. Because of this, most of the in situ/in vivo studies with thiamin antagonists suffer from a focus on certain effects of interest, which may be irrelevant to the true mechanisms of action. Incorrect mechanistic interpretations decrease the predictive capacity of models employing the thiamin antagonist and hence therapeutic treatments based on their results. By contrast, the development and incorporation of enzymological knowledge into studies in vivo should greatly increase the successful outcomes of such studies. In particular, structural work revealed different binding of thiamin in thiamin-metabolizing (Fig. 5, see also ) and ThDP-dependent enzymes (Figs 6, 7, see also ). A specific sequence motif G(D/E)(G/A)X22–26NN, binding the diphosphate group of ThDP through the magnesium ion [85, 86], has been revealed in ThDP-dependent enzymes only. The motif interaction with the diphosphate residue of ThDP is important for not only coenzyme binding, but also the enzyme catalytic action [85, 87-89]. This may explain conservation of the motif throughout the ThDP-dependent enzymes and its absence from thiamin-metabolizing enzymes. Indeed, it is not found in thiamin triphosphatase which dephosphorylates thiamin triphosphate to ThDP , nor in thiamin diphosphokinase  which phosphorylates thiamin to ThDP. Furthermore, the energetically unfavorable V-conformation of ThDP (Fig. 6, see also ), its analogs (Fig. 7A) or adducts (Fig. 7B) is found in all ThDP-dependent enzymes. The relative position of the two rings of thiamin in this conformation supports the intramolecular catalysis of deprotonation of the catalytic center of ThDP, C2 atom of thiazole ring, by the 4′-NH2 group of the pyrimidine ring. By contrast, thiamin diphosphokinase binds thiamin (Fig. 5B) as well as its diphosphorylated analog pyrithiamin diphosphate (Fig. 5A) in the alternative F-conformation, where the 4′-NH2 group does not interact with the C2 atom . Docking thiamin triphosphate into the active site of thiamin triphosphatase does not reveal the thiamin V-conformation either . These structural data provide the basis for a rational design of the thiamin analogs with different affinities to the ThDP-dependent and thiamin-metabolizing enzymes. Moreover, in the active sites of ThDP-dependent enzymes, the enzyme-specific conformational changes occur during catalysis, which also affect the ThDP conformation [87, 88]. If similar conformational changes are observed upon binding of a thiamin analog, the complex between the analog and this particular ThDP-dependent enzyme is strengthened. For instance, binding of thiamin thiazolone diphosphate to pyruvate dehydrogenase (Fig. 7A) is tighter than that of ThDP because of significant conformational transition of the enzyme–inhibitor complex . The PDH-specific conformational change upon binding of thiamin thiazolone diphosphate may also explain the 100-fold higher affinity of the inhibitor to PDH compared with pyruvate decarboxylase (Table 2). Thus, the structural characterization of the thiamin-binding modes and the catalytic mechanisms in different enzymes provides a prerequisite for creating the enzyme-specific analogs of thiamin. The accompanying minireviews by Hailes et al.  and Andrews and McLeish  show the latest contributions to this line of research regarding ThDP-dependent enzymes, while the minireview by Bettendorff and Wins  contributes to our understanding of the thiamin-binding mode in thiamin-metabolizing enzymes. However, beyond thiamin diphosphokinase and thiamin triphosphatase, the thiamin binding and transformation in thiamin-metabolizing enzymes remain unresolved because the enzymes are not identified at the molecular level. As pointed out in the Introduction, most of the enzymatic transformations of thiamin to its natural derivatives shown in Fig. 1 have been characterized as biochemical activities of cellular fractions only, whereas the gene products possessing these activities are still largely unknown.
Thiamin antagonists to target ThDP-dependent enzymes in anticancer therapies
The significance of thiamin for cancer cell proliferation has long been known. Indeed, the intensity of the nucleic acid biosynthesis required for the proliferation greatly depends on the availability of ribose-5-phosphate and NADPH, which is controlled by the ThDP-dependent transketolase functioning in the pentose phosphate pathway [41, 62, 91, 92]. Hence, different strategies using either thiamin antagonists (Table 3) or thiamin deprivation  have been suggested to deplete thiamin from cancer cells. Certainly, such strategies should take into account that the well-characterized complications in cancer patients, such as heart failure, are known to result from a thiamin deficiency in healthy cells , which may be promoted by efficient pumping of thiamin into cancer cells . Nevertheless, some treatments employing oxythiamin have given promising results (Table 3). Although Table 3 shows that, in animals (in vivo), the effective doses of oxythiamin (starting from 300 μmol·kg−1 body weight) were never as low as observed in some experiments with animal cells (in situ) (starting from 0.25 μm), applied histotoxicity analysis of liver, heart and kidney from mice after oxythiamin treatment showed no signs of toxicity [92, 94]. Besides, recent data indicate that the combination of oxythiamin with other drugs may be useful in therapy for drug-resistant cancers [91-95]. For example, a combination of oxythiamin with dehydroepiandrosterone (glucose-6-phosphate dehydrogenase inhibitor) was effective in arresting metatrexate-resistant cancer cell proliferation (human colon adrenocarcinoma M6-HT29). Combination therapy using oxythiamin and imatinib (tyrosine kinase inhibitor used in the treatment of chronic myeliod leukemia) led to the reduction of in vitro growth in imatinib-resistant tumors, enhancing the efficacy of imatinib in primary chronic myeloid leukemia cells isolated from patients.
It is worth noting that most of the work on anticancer strategies employing the thiamin antagonists was based on the current assumption that metabolic reprogramming in cancer requires the upregulation of transketolase and downregulation of pyruvate dehydrogenase [41, 62, 92]. However, although the data are still accumulating, they do not support such a simplified view, at least for all types of cancer. Table 3 shows that the responses of different cancer cells to oxythiamin differ greatly, both in situ and in vivo. For instance, oxythiamin decreased the viability of human colon adenocarcinoma cells at 10−3–10−2m (IC50 of 5.4 mm), whereas inhibition of proliferation of Mia pancreatic carcinoma cells was achieved at concentrations three orders of magnitude lower, i.e. 10−7–10−5m (Table 3). These findings point to the need for more detailed studies on the involvement of different ThDP-dependent enzymes in the reprogramming of different cancer cells. In particular, it has long been known that tumor pyruvate dehydrogenase may not be totally inactivated, but rather abnormally active in catalyzing one of its side reactions, the nonoxidative production of acetoin . This stimulation of nonphysiological enzymatic reactions may be caused by toxic intermediates of the diseased state participating in a vicious cycle of physiological impairment . Furthermore, recent structural and mutagenesis studies questioned the catalytic competence of the tumor-specific isoform of transketolase, transketolase-like protein 1 [98, 99]. Catalytic incompetence of transketolase-like protein 1 may explain the results of an independent in vivo study which showed that overexpression of transketolase-like protein 1 in cancer cells was not strongly correlated with their proliferation rate, glucose transporter-1 expression and response to oxythiamin . In another study, the thiamin antagonist N3′–pyridyl thiamin (Fig. 3) completely suppressed the activity of transketolase in HTC-116 tumor cells in vivo and in vitro, but this had no effect on the tumor cell growth . Remarkably, the same study showed a poor effect of N3′-pyridyl thiamin on the OGDHC activity. It is worth noting in this regard that the gene coding for the recently characterized isoform of OGDH, OGDHL [85, 101], was one of the five genes for which the cancer-specific methylation in breast tissues was established . OGDHL gene downregulation by hypermethylation has also been seen in a number of different cancer types. A recent study  investigated the molecular mechanisms of OGDHL involvement in the tumorigenicity of cervical cancer cells. It was shown that restoration of OGDHL expression in these cells lacking endogenous OGDHL, inhibited cell proliferation, invasion and soft agar colony formation. This was associated with increased ROS production leading to apoptosis. The apoptosis occurred through the downregulation of protein kinase B signaling mediated by caspase 3 and decreased NF-κB phosphorylation. OGDHL silencing had opposite effects. Thus, decreased OGDHL expression in cervical cancer was shown to contribute to tumorigenesis via activation of the protein kinase B signaling pathway. As a result, OGDHL was shown to be an important antiproliferative gene in cervical cancer. Another ThDP-dependent enzyme with high structural similarity to OGDH, dehydrogenase E1 and transketolase domain-containing 1 (DHTKD1 gene), is highly expressed in brain tumors, yet only a hypothetical function of this enzyme has been suggested . Unresolved biological roles of the natural thiamin derivatives beyond ThDP should not be ignored with regards to cancer cell reprogramming. In particular, thiamin triphosphatase was decreased in melanoma, being expressed inversely to the melanoma tumor antigen p97 . Clearly, the significance of ThDP-dependent enzymes in cancer goes beyond the accepted views and requires more experimental investigation. The multitude of enzymes dependent on thiamin and its derivatives, and the selectivity of thiamin antagonists to potential in vivo targets are important issues to be considered in order to advance research into the role of thiamin and ThDP-dependent enzymes in cancer cell reprogramming.
Phosphonate analogs of 2-oxo acids as tools to specifically target ThDP-dependent enzymes in vivo
Synthetic analogs of both coenzymes and substrates may be used to regulate enzyme functions. In this section, we consider the enzymological background and examples of the in vivo application of phosphonate analogs of 2-oxo acids to regulate ThDP-dependent enzymes.
The phosphonate analogs of pyruvate were first described by Vvedenskij in 1888 , followed by Brooks in 1912 , but the structures were not unambiguously proved in these studies. To the best of our knowledge, synthesis of the 2-oxo phosphonate esters of carboxylic acids (IIa, Fig. 8A) according to Arbuzov's reaction was first published by Kabachnik & Rossijskaja in 1945 . The synthesis of fully esterified phosphonate analogs of dicarboxylic acids (IIe, Fig. 8A) followed in 1967 . In the late 1970s, Kluger & Pike  and Khomutov et al.  introduced the free phosphonic acid (I, Fig. 8A) and its monoesters (IIIa,b, Fig. 8A) using de-esterification of the methyl phosphonates by NaI or Pd-catalyzed hydrogenation of the benzyl ester. De-esterification to obtain compounds I was also performed through the formation of intermediate trimethylsilyl esters [111, 112]. In later procedures, acid  or base  hydrolysis was employed to obtain the free phosphonic acids. Acethylphosphinates (structures IV and V in Fig. 8B) were synthesized according to a different scheme using orthoesters of carboxylic acids and hypophosphorous acid or methyldichlorophosphine (Fig. 8B) . The synthesis of a number of new acylphosphonic acids similar to I and their monomethyl esters similar to III (Fig. 8A) followed . According to this study, the length of the C(O)-PO3H2 bond in compounds I–III (Fig. 8A) is 1.86 Å. This is significantly longer than the σ-sp2–sp2 bond in the CO–CO moiety (~ 1.47 Å), and the average energy of the C–P bond scission is known to be lower than that of the C–C bond. However, the chemical stability of the C–P bond to hydrolysis and oxidation is nevertheless rather high. An essential factor determining the stability is the nature of substituents at the phosphorous atom. The stability of the C–P bond in free phosphonic acids under acidic and basic conditions is generally higher compared to that of the corresponding esters .
Although the phosphonate and carboxyl groups cannot be formally considered as isosteric groups, becausae they have different number of substituents and electrons at the P and C atoms, respectively, they show an overall structural similarity to an extent that allows the 2-oxo phosphonates to mimic their corresponding 2-oxo acids in the active sites of enzymes (Fig. 7B). The low-energy barriers of the rotation around the C–P bond, shown by semi-empirical quantum mechanical (MNDO/H) calculations for benzoylphosphonic acid, its esters and anions , may also contribute to the accommodation of the phosphonates in the enzyme active sites.
It is worth noting that the phosphonic (I–III, Fig. 8A) and phosphinic (IV, V, Fig. 8B) acids have similar structures and properties. However, the phosphinates are more easily oxidized, generating the phosphonates. In addition, several features of the phosphinates, compared with phosphonates might be responsible for the greater inhibitory power of the phosphinates, observed in enzymatic studies (Table 4). First, the charge of the phosphinate group (-1) is identical to that of the carboxylic group, whereas the phosphonates may have a higher charge (-2) at physiological pH. Second, steric hindrance for the carbonyl group addition, which is created by the tetrahedral phosphorous atom compared with the carbon atom in the carboxyl group, should be less pronounced in the phosphinates, with their hydrogen atom substituting for the hydroxyl group in the phosphonates. Third, compared with the phosphinates with one hydroxyl group, the partial positive charge at the phosphorus atom in the phosphonates with two hydroxyl groups is less because of the mesomeric (+M) electron-donating action of the two oxygen atoms. This, in turn, decreases the partial positive charge and reactivity of the neighboring carbonyl group, especially in anionic forms of the phosphonates at physiological pH. Nevertheless, the general chemical reactivity of the phosphinates and phosphonates is rather similar, and similar to that of carboxylates. The carbonyl group of the 2-oxo phosphonates is known to react with typical nucleophiles (e.g. borohydride, amines, hydroxylamine) with formation of the corresponding derivatives (alcohols, imines, oximes). Accordingly, the 2-oxo group of the phosphonates may undergo transamination or oxidoreduction, inherent in the 2-oxo acids. However, the affinity of the phosphonates to the corresponding enzymes is rather low compared with the carboxylic acids and hence enzymatic reactions of this type are not efficient [113, 114, 117]. In vitro, a minor transamination of succinyl phosphonate (SP; 100 μm) by the aspartate and branched chain amino acid transaminases (0.6 and 0.016% of the corresponding physiological reaction rates) was observed . Acetyl phosphonate and its monomethyl ester were also poor substrates of lactate dehydrogenase, producing corresponding 2-hydroxy phosphonates with, at pH 6.0, Km values of 15 and 10 mm, respectively .
Table 4. Comparison of the in vitro (extracted or purified enzymes), in situ (mitochondria, cell cultures, tissues) and in vivo (bacteria, plants, animals) action of the phosphonate analogs of 2-oxoglutarate and pyruvate
As with the 2-oxo acids, the phosphonates undergo addition to the C-2 atom of ThDP (Fig. 7B). Introduced by Kluger & Pike in 1977 , this type of enzymatic reaction has enabled researchers to employ the phosphonate analogs of 2-oxo acids to test the details of catalytic mechanisms, an application similar to that for the structural analogs of ThDP considered above. Recently, the varied structures of phosphonic and phosphinic analogs of pyruvate and 2-oxoglutarate (Table 4) have been synthesized to extend the enzymological studies [114, 118-120]. This has enabled assessment of the catalytic details, the kinetics of the enzyme conformational changes during catalysis and the role of the active site amino acid residues [87, 114, 118, 119, 121-128]. A slow (compared with catalysis) conformational transition has been shown upon phosphonate binding to different ThDP-dependent enzymes [87, 114, 117, 122-124]. The transition contributes to the complex kinetics of the phosphonate (SP) inhibition of the enzymes (E) according to Reaction 1:
The second step of Reaction 1 is the kinetically slow step, and the phosphonates often require a preincubation with their cognate enzymes to exhibit maximal inhibition. The conformational transition is associated with multiple rearrangements in the active site , enabling the tight binding of the catalytic intermediate. Although the equilibrium is shifted to the formation of tightly bound intermediate E*SP, in most cases Reaction 1 is fully reversible [109, 114, 121, 124]. The inability of the ThDP-dependent dehydrogenases and decarboxylases to cleave the C–P bond in the phosphonates has proved advantageous for crystallographic studies of substrate binding to the ThDP-dependent enzymes at the pre-decarboxylation step [87, 129, 130]. Eventually the C–P bond splitting also was revealed in such studies. This occurred in the complex between benzoyl phosphonate and benzoylformate decarboxylase, and was accompanied by irreversible inactivation of the enzyme . The inactivation was caused by phosphorylation of the active site with formation of the usual product P and modified enzyme E-PO32− according to Reaction 2:
In good accord with previous data  which showed the dependence of the C–P bond stability on the type of substitutions at the phosphorous atom, this finding not only adds a novel aspect to the chemistry of the ThDP-bound intermediates, but also points to a mechanism which in some active sites might provide for the irreversible inhibition by the phosphonate analogs of the 2-oxo acids (Reaction 2).
Overall, significant data on the interaction between the phosphonate analogs of 2-oxo acids and ThDP-dependent enzymes have been accumulated to permit rational inhibitor design.
Introduction of the phosphonate analogs of pyruvate as transition state analogs of PDH [109, 121] and of 2-oxoglutarate as potent inhibitors of OGDH  opened a way to study metabolic significance of particular ThDP-dependent enzymes through their selective targeting in vivo. Because the enzymatic transition state analogs strongly bind only to the enzymes whose transition state they mimic, they are much more selective to their targets (Table 4) compared with thiamin antagonists considered above (Tables 2 and 3). For instance, the structural analog of pyruvate, acetyl phosphinate, has a ≥ 1000-fold higher affinity for PDH compared with OGDH. Similarly, the structural analog of 2-oxoglutarate, SP, is a potent inhibitor of OGDH, but does not inhibit PDH (Table 4). Even for enzymes performing the same catalytic step of the 2-oxo acid decarboxylation, the sensitivity to the same phosphonate analog of substrate differs, depending on the fate of the decarboxylated intermediate. Thus, Table 4 shows that acetyl phosphonate is a much more powerful inhibitor of the oxidative transformation of pyruvate, catalyzed by PDH/PDHC, compared with both oxidative and nonoxidative transformations performed by pyruvate oxidase and pyruvate decarboxylase, respectively . Similarly, SP is a much more effective inhibitor of the oxidative decarboxylation of 2-oxoglutarate by OGDH/OGDHC than it is of nonoxidative 2-oxoglutarate decarboxylation by MenD (Table 4). Also structural variations in the phosphonate group may have different impacts on the phosphonate binding to the 2-oxo acid dehydrogenases and decarboxylases. Table 4 shows that, ethyl ester of SP has a 10-fold higher affinity for the 2-oxoglutarate decarboxylase MenD than does SP . However, for OGDHC, SP binds ~ 10-fold more tightly than its ethyl ester . The methyl ester of SP was also found to be ~ 10-fold less efficient than SP in the overall physiological reaction catalyzed by OGDHC . It therefore appears that the esterification-induced changes in the binding of the phosphonates within active sites are defined by the combined contributions of steric factors and any additional interactions formed with the added methyl or ethyl groups of the esters. Overall, it seems that spatial constraints in the active site of OGDH determine the better binding of SP compared with the phosphonate methyl and ethyl esters. Similarly, substitution of a methyl group for the hydrogen in the phosphinate analog of pyruvate, increases its affinity to PDHC (Table 4). Again, this suggests that accommodation of the phosphinate group in the active site of PDH may be limited by spatial considerations. By contrast, when these constraints are not so strong, as seems to be the case of MenD, additional interactions between the active site and the ethyl moiety of the SP ester may increase the stability of the enzyme–inhibitor complex (Table 4).
As a result, structural modifications of the phosphonate and phosphinate analogs of 2-oxo acids are known which may be used for further attenuating their selectivity. Depending on substitutions in the phosphonate molecule  and the active site architecture, C–P bond cleavage may occur, resulting in an irreversible modification of the active site (Reaction 2). A case in point is the reaction of benzoylformate decarboxylase with methylbenzoyl phoshonate and benzoyl phosphonate . Reaction of the former provided a stable pre-decarboxylation intermediate. Conversely, reaction with benzoyl phosphonate resulted in the phosphorylation of an active site serine residue. Thus, not only the structural variations in the phosphonate molecules, but also specific mechanistic details of the catalysis by ThDP-dependent enzymes determine the phosphonate binding and transformations in the active sites. All these features may contribute to the species-specific action of the phosphonate analogs of 2-oxo acids. Therefore, it is very important to understand how ThDP-dependent enzymes specifically bind and transform their substrates. More detailed discussion of this is provided in the accompanying minireviews by Hailes et al.  and Andrews & McLeish .
Initially, the idea was developed that the phosphonate analogs of pyruvate may be used as efficient herbicides or antibacterial and antifungal agens . However, in contrast to the high selectivity of the phosphonates for individual ThDP-dependent enzymes, the necessary species-selective action has not been seen [117, 118, 132, 133].The phosphonates were strong inhibitors in not only plants, fungi and bacteria, but also in mammalian systems (Table 4). As alluded to previously, accumulating a body of enzymological knowledge might provide solutions to this problem. That said, species-specific action is not required for phosphonate inhibitor applications in the rapidly developing fields of metabolic regulation and systems biology. These research areas may benefit greatly from the ability of the phosphonate analogs of 2-oxo acids to target particular ThDP-dependent enzymes in vivo in a highly specific manner. For instance, the phosphonate analogs of 2-oxoglutarate were suggested as a tool to model impaired function of OGDHC or to regulate the enzyme in vivo [132, 134]. Application of these analogs to plant systems revealed them to be efficient in vivo regulators of plant OGDHC. The studies showed that the inhibition of OGDHC by the phosphonates strongly affected ADP-stimulated respiration and induced widespread metabolic effects in both primary and secondary metabolism [135-138]. Extensive effort was undertaken in these studies to experimentally test the high selectivity of the phosphonate analogs of 2-oxoglutarate to OGDHC, which is predicted theoretically. There was no significant inhibition by the phosphonate analogs of 2-oxoglutarate of a number of enzymes working with structural analogs of 2-oxoglutarate, the 2-oxoglutarate transporters and the regulatory PII protein interacting with 2-oxoglutarate [114, 135, 136, 139]. This high in vivo selectivity was confirmed by studies of related fluxes using isotopes and metabolomics [135, 136, 139]. Comparative metabolomics has provided further evidence of the specific action of the phosphonate analogs of 2-oxoglutarate upon OGDHC in vivo. The characteristic changes in the metabolic profiles after the phosphonate application were inherent only in the systems possessing the 2-oxoglutarate dehydrogenase, being absent in cyanobacteria which possess 2-oxoglutarate decarboxylase instead . Finally, in plants, the OGDHC function is not essential for development and survival, and hence the OGDH gene could be manipulated more easily than in animals. This allows one to reveal that the plant mutants with a low level of OGDH recapitulated all the main features of the metabolomic and flux changes known from OGDH inhibition by the phosphonates . In animals, only a mild reduction in OGDHC was possible by genetic manipulation, and a limited number of metabolites were tested . Nevertheless, metabolic changes in the brain of the mouse mutant of OGDHC also revealed certain similarities with metabolic changes in neurons exposed to SP (discussed in Ref. ).
Known transformations of the phosphonate analogs of 2-oxo acids in vivo should be mentioned. First, physiologically relevant de-esterification of the ethyl esters of SP by esterases present in vivo was implied by the fact that tri- and diethylated SP, inactive in vitro, mimicked the SP effects in vivo [114, 142]. At the same time, comparison of the effects caused by SP with its carboxyethyl ester in plant systems [135, 136, 139], or SP with its phosphonoethyl ester in neurons [143, 144], or phosphonoethyl with carboxyethyl SP in neurons  has revealed certain quantitative features. They suggest that the monoethylated derivatives may act without de-esterification, but differ in penetration. In addition, the reactivity of phosphonoethyl and carboxyethyl SP to esterases may be different. As noted above, minor transamination and oxidoreduction of the phosphonates by some transaminases and oxidoreductases cannot be excluded. However, the bulk of the evidence obtained supports very high selectivity of the phosphonate analogs of 2-oxoglutarate to OGDHC in vivo. Fewer studies have been carried out in this regard on the phosphonate analogs of pyruvate, but the available information suggests them to be efficient and specific inhibitors of PDHC in vivo, as well (Table 4).
In general, effective phosphonate concentrations are significantly higher in situ than in vitro (Table 4). For instance, OGDHC extracted from brain mitochondria was half-inhibited by 5 μm SP at substrate saturation (i.e. IC50SP = 5 μm in the presence of 2-oxoglutarate at 2 mm ~ 10 × Km) , but the half-saturation of the SP effects in neurons required a concentration at least one order of magnitude higher (IC50SP ~ 50 μm) . Similarly, methyl acetyl phosphinate was much more efficient on the purified PDHC than in mitochondria . As discussed above, the efficiency of the enzymatic transformation of the phosphonates is low compared with the naturally occurring carboxylic acids, Hence, the increase in effective concentrations in situ/in vivo compared with in vitro is unlikely because of the phosphonate metabolism. Rather, it could be explained by in situ/in vivo competition of the phosphonates with the natural substrates for both the transporters and enzyme active sites. In some systems, however, strong problems with intracellular delivery of the phosphonates were detected. For example, Table 4 shows that acetyl phosphinate, which is a very efficient inhibitor of bacterial and plant PDHC and PDH in vitro, had no effect on plant metabolism up to 1 mm. However, 1-aminoethyl phosphinate, which generates acetyl phosphinate in the transamination reaction, did interfere with both bacterial and plant metabolism at 1–100 μm. If transamination was blocked, the metabolic effects were not observed, which points to acetyl phosphinate as the metabolically active compound . Because bacteria and probably plants as well possess natural systems for the biosynthesis of phosphonopyruvate (including 3-phosphonopyruvate decarboxylase, Table 1), tighter control of the phosphonate-dependent systems may exist in plants and bacteria than in mammals. This would explain the inefficiency of the external acetyl phosphinate to affect bacterial and plant metabolism, in contrast to its efficient action after intracellular formation from the transported aminophosphonate precursor. It has also been shown  that bacteria generate methylacetylphosphonate from dehydrophos, a broad-spectrum peptide antibiotic containing the vinyl-phosphonate moiety. In this case, dehydrophos is transported by nonspecific oligopeptide permeases and hydrolyzed by intracellular peptidases. The resulting phosphonate analog of dehydroalanine undergoes spontaneous rearrangement and hydrolysis, generating the phosphonate analog of pyruvate. The latter is most probably the active species responsible for the antibiotic action of dehydrophos .
Modeling the impaired function of neuronal and brain OGDHC by enzyme inhibition in situ and in vivo: implications for neurodegeneration
The fact that OGDHC is impaired in the brain of patients with neurodegenerative diseases [147, 148] promoted the in situ [114, 137, 141-145] and in vivo [36, 149, 150] application of the phosphonate analogs of 2-oxoglutarate to understand the role of the OGDHC impairment in the neuronal and brain function. These studies revealed a dual effect of OGDHC inhibitors on different physiological parameters in vivo and in situ. On the short-term (within 1 h) application of moderate concentrations (100–200 μm added to neuronal medium), SP and some of its esters caused neuroprotective effects under conditions of glutamate excitotoxicity in situ. The effects included: (a) a decrease in the glutamate-induced mitochondrial ROS ; (b) alleviation of calcium deregulation [143, 144]; (c) protection from irreversible mitochondrial depolarization [143, 144]; and (d) decreased neuronal death . Positive effects were observed at moderate concentrations of the OGDHC inhibitors, which did not decrease total neuronal ATP or NAD(P)H in situ, and neither resulted in obvious energetic impairment in vivo [141, 142, 149]. The influence of SP on glutamate-induced ROS production was biphasic and did not correlate with the effect of SP on the mitochondrial potential (Fig. 9). It was noted that, after an initial decrease, elevation of ROS was observed, in spite of the fact that increasing SP monotonously increased the depolarization of neuronal mitochondria . By contrast, the triethylated SP derivative depolarized mitochondria less efficiently than SP and did not induce the secondary increase in ROS production. With increasing triethylated SP, both mitochondrial polarization and ROS production decreased monotonously . First, these data provide evidence on the contribution to glutamate neurotoxicity of ROS produced by OGDHC [97, 151, 152]. Second, they indicate that strong inhibition of OGDHC by SP under conditions of glutamate excitotoxicity activates the ROS sources other than the 2-oxoglutarate-dependent reaction of OGDHC, which is inhibited by SP, and mitochondrial potential/NADH-dependent reactions of the respiratory chain, which are inhibited by decreases in ΔΨ/NADH (Fig. 9).
Although at high SP doses the neuronal ATP and NAD(P)H levels in situ did decrease compared with the lower doses, the primary effect of the OGDHC inhibition in neuronal culture was not the energetic failure. Instead, impaired oxidation of 2-oxoglutarate in the tricarboxylic acid cycle resulted in systemic disturbance in the neuronal amino acid pool. Changes in the brain glutamate levels were also detected in SP-treated animals under conditions when no energetic impairment was obvious . It was noted that, only when the SP-exposed animals were additionally challenged by alcohol exposure, their locomotor activity showed impairments, which could be attributed to an energy deficit. In studies on neuronal cultures, the SP-induced decrease in ATP or NAD(P)H was also stimulated by an additional metabolic stress, such as alcohol exposure [36, 141]. The physiological consequences of the SP inhibition of OGDHC in situ and in vivo are thus in line with the existing notion that the energy level in living organisms is highly buffered by multiple systems of the short-term regulation dependent on the universal energy indicator, the adenine nucleotide phosphorylation. By contrast, the levels of other metabolites are less ‘buffered’. Therefore, OGDHC inhibition is first of all evident not as an energy deficit, but as an increase in 2-oxoglutarate. This perturbation of metabolic flux in the tricarboxylic acid cycle induces significant changes in the amino acid pool because of the transaminase and other reactions involving 2-oxoglutarate and amino acids . Remarkably, many of these reactions may substitute for the OGDHC-catalyzed NADH production, which explains that there was no significant energy impairment observed. Some of the reactions may also generate another OGDHC product, succinyl-CoA. However, all these compensatory pathways which shunt the block on OGDHC, switch the metabolism to the oxidation of amino acids. The accompanying decrease in neuronal protein is in good accordance with the lack of amino acids, which must impair protein synthesis [137, 141]. The impaired biosynthesis is supported by concomitant observation of an ATP increase. This paradoxical finding in SP-treated neurons appears to be caused by a neuronal inability to perform the required level of protein synthesis when the pool of amino acids is strongly perturbed. Remarkably, the OGDHC inhibition in pyrithiamin-treated animals was also associated with perturbations in the amino acid pool  and a paradoxical increase in ATP in brain . Thus, substantial and/or long-term inhibition of OGDHC impairs the metabolism of amino acids, many of them neurotransmitters or their precursors, and proteins. Clearly, these two consequences of OGDHC inhibition may cause perturbed cognitive function. Indeed, reduced cognition correlates with OGDHC impairment in patients with neurodegenerative diseases . New knowledge on the molecular mechanisms underlying this correlation, which was obtained by modeling the impaired function of OGDHC using the synthetic phosphonate analogs of 2-oxoglutarate, should benefit future research on therapeutic strategies.
Interestingly, low doses of SP induced preconditioning effects in animals exposed to hypoxia, which was associated with an increase in their brain OGDHC activity. The increase in OGDHC was ascribed to a compensatory response resulting from OGDHC inhibition by SP [36, 149, 150]. In this regard, a temporary increase in neuronal ATP which was observed under low SP doses in situ, may contribute to the preconditioning in vivo. In independent experiments, ischemic preconditioning of neurons was also associated with an increase in neuronal ATP , similar to that after exposure to low SP doses. It is therefore possible that OGDHC upregulation in response to the short-term inhibition of the enzyme, either by hypoxia or SP, may occur in situ and in vivo as a natural response to metabolic stress caused by impaired OGDHC. This response may contribute to preconditioning effects. These and other data [36, 149, 150] suggest that the action of the phosphonate analogs of 2-oxoglutarate on OGDHC may provide the mechanistically defined model of chemically induced hypoxia. Currently, such models employ cyanide to inhibit respiration and iodoacetate to inhibit glycolysis . The highly selective and reversible action of SP on mitochondrial OGDHC provides significant advantages over the unspecific action of cyanide and iodoacetate, which may irreversibly modify many more targets beyond the enzymes of respiratory chain and glycolysis, respectively.
Correct mechanistic interpretations of the action of synthetic analogs of the enzyme substrates and coenzymes in complex biological systems require comparative in vitro study of relative specificities of the compounds to their potential in vivo targets. Mimicking the catalytic transition states, phosphonate analogs of 2-oxo acids are highly selective inhibitors of the corresponding 2-oxo acid dehydrogenases in vivo. Regarding the thiamin structural analogs, more detailed characteristics of their targets and metabolic transformations in vivo are required to introduce the target-selective compounds. The molecular identification and enzymological characterization of the thiamin-metabolizing systems are strongly needed. Synthetic regulators of the enzymes dependent on thiamin and its biological derivatives have significant potential as pharmacological tools for metabolic regulation, including the design of not only drugs, but also therapeutic approaches. The combination of enzymological and in vivo studies provides a key to successfully solve biotechnological and medical goals.
VIB and NVL acknowledge the RFBR support of their work on the synthesis of the phosphonate analogs of 2-oxo acids and their action in vivo and in situ (current grant N 12-04-01541a to VIB and grant N 11-03-00265a to NVL). The authors thank Prof. S. Strumilo (University of Bialystok, Poland) for interesting discussions regarding certain aspects of the thiamin antagonist action; Dr T. Wagner (MPI Marburg, Germany) for his contribution of Fig. 6 and Dr G. Raddatz (DKFZ, Heidelberg, Germany) for valuable advices on preparation of Figs 5 and 7. Copy-editing of the manuscript by Ms L. Millstine (Propep SARL, Paris, France) and Dr M. McLeish (Indiana University-Purdu University, Indianapolis, USA) is greatly acknowledged.