Chemical synthesis and purification of AThDP and AThTP

We have previously shown that the condensation of ThDP and AMP by N,N′-dicyclohexylcarbodiimide leads to the synthesis of AThTP [5]. Here, this reaction has been further characterized with respect to the kinetics and composition of the reaction medium. In particular, we show that other side products, which have been unambiguously identified, are also formed during synthesis. Indeed, as shown in Fig. 2A, synthesis of AThTP (peak 5) is accompanied by the appearance of two other compounds in the reaction medium: AThDP (peak 3) and ThTP (peak 4). The small ThMP peak (peak 1) is essentially a contamination present in the commercially available ThDP used as the precursor (peak 2). However, AThTP synthesis proceeds through an optimum and after 3 h an accumulation of ThTP and AThDP is observed (Fig. 2C), although the amount of AThTP is much lower.

The presence of AThDP was further confirmed by the condensation of ThMP and AMP in the presence of N,N′-dicyclohexylcarbodiimide (Fig. 2B), which mostly leads to the formation of AThDP. However, the yield of AThDP synthesis according to this latter synthetic procedure is low: after 2 h, < 10% of the ThMP is converted to AThDP. Therefore, we routinely synthesized both compounds by condensing ThDP and AMP in the presence of N,N′-dicyclohexylcarbodiimide for 2 h. To purify AThDP and AThTP, large-scale synthesis was performed using an (AMP)/(ThDP) ratio of 1.5 rather than 1, because this resulted in higher yields of AThTP. After 2 h, thiamine derivatives were precipitated with diethyl ether and dissolved in water. ThTP, AThDP and AThTP were purified using several chromatographic steps. All thiamine phosphate derivatives, except ThTP [9], were retained on a AG 50W-X8 cation-exchange resin and eluted with ammonium acetate (0.2 m; pH 7.0). After lyophilization, the residue was dissolved in water and layered on a column filled with the anion-exchange resin AG-X1 equilibrated in water (Fig. 3). AThDP was eluted in 0.25 m ammonium acetate (pH 7.0) followed by 0.5 m ammonium acetate for the elution of AThTP. Both compounds were further purified on a Polaris C₁₈ HPLC column. The total yield was 5.3% for AThDP and 2.7% for AThTP (Table 2). The purity of the two preparations was checked by HPLC using UV and, after derivatization, fluorescence detection (Fig. 4).

Physicochemical characterization of chemically synthesized AThDP and AThTP using MS, NMR, fluorometry and molecular modeling

Both fractions were analyzed by positive ESI-MS (Fig. 5). As expected, the AThTP fraction contained a major cation with a m/z ratio of 754.1, as described previously [5]. In the AThDP fraction, the major peak had a m/z ratio of 674.1, as expected for AThDP (crude formula C₂₂H₃₀N₉O₁₀P₂S⁺, parent ion M⁺) with an average molecular mass of 674.5 Da (exact monoisotopic mass 674.1 Da). As for AThTP [5], ESI-MS/MS fragmentation of AThDP gave three main peaks: m/z 553.1 (a fragmentation product of AThDP obtained by loss of the pyrimidinium moiety, M⁺– 121 – pyrimidinium), m/z 348.1 (corresponding to AMP) and m/z 257.1. We were unable to assign the latter ion, which is obtained after fragmentation of both AThTP and AThDP and probably results from a molecular rearrangement.

NMR data for AThTP and AThDP, together with those for ThTP and ThDP, are listed in Table 1. They are clearly in accordance with the presence in AThDP and AThTP of a thiamine and an adenine moiety, as compared with thiamine and adenosine NMR data. The presence of three linked phosphates in AThTP is confirmed by three phosphorous signals in the ³¹P-NMR spectrum (two doublets and one triplet, as expected). Oddly, in AThDP, the two phosphates seemed to be equivalent, as only one signal was detected on the spectrum. However, the possibility that the molecule is adenosine thiamine monophosphate could be excluded on the basis of the molecular mass (Fig. 5). The linkage (C-15-triphosphate-C-5′) between the thiamine moiety and the adenine moiety of the molecule was proven by the presence of ¹³C–³¹P coupling constants for C-14 and C-15 and for C-5′ and C-4′.

We place special emphasis on the C-2 proton of the thiazolium ring which is required for the catalytic activity of ThDP [10]. This proton is particularly labile and is completely exchanged with deuterium within a few minutes [11]. The experimental shift was 9.61, 9.55 and 9.55 p.p.m. respectively for ThMP, ThDP and ThTP (pH 7.4), values higher than expected for aromatic protons (in general 7.5–8.5 p.p.m.). In AThDP and AThTP, we saw a decrease in the shift (9.14/9.18 p.p.m.) compared with ThMP, ThDP or ThTP, indicating a modification of the electronic environment of the C-2 proton, probably as a consequence of a through-space interaction with the adenine moiety. This would suggest a U-shaped folding of AThDP and AThTP.

Molecular modeling was applied on a model of the free (without influence of the environment) molecules without any counterion. The phosphate groups are neutralized by hydrogens and the whole system bears a positive charge because of the thiazolium fragment. Calculations showed that a U-folded conformation is energetically accessible for both di- and triphosphorylated derivatives. A possible structure for each derivative is shown in Fig. 6. In this conformation, the C-2 proton is embedded in the closed environment formed by the aromatic adenine and aminopyrimidine rings. Such a folded structure for adenylated thiamine derivatives is not in favor of a cofactor role that requires a highly reactive C-2 proton [10].

Like free thiamine, AThDP and AThTP can be readily oxidized to highly fluorescent thiochrome (Thc) derivatives. AThcDP and AThcTP gave practically identical fluorescence spectra with an optimum of 353 nm for excitation and 439 nm for emission (Fig. 7). However, when we compared the fluorescence properties of AThcDP and AThcTP with those of nonadenylated thiochromes (Thc, ThcMP, ThcDP and ThcTP, which have roughly the same fluorescence) [12,13], we found some interesting differences. First, the optimum emission wavelength was slightly lower for AThcDP and AThcTP than for thiochrome (439 versus 443 nm; Fig. 7C). More importantly, we observed that AThcDP and AThcTP solutions gave peaks with areas approximately twice as large as thiochrome solutions of the same molarity. These differences were confirmed by comparing the fluorescence of thiochromes obtained before and after the enzymatic hydrolysis of AThTP and AThDP. We have previously shown that complete hydrolysis of AThTP by bacterial membranes yields ThMP as the sole thiamine-containing product [5]. When we incubated synthetic AThDP with a membrane fraction obtained by centrifuging sonicated E. coli, we also observed hydrolysis of this compound with ThMP as product. In these experiments we found that after derivatization, the fluorescence ratios AThcDP/ThcMP and AThcTP/ThcMP were, respectively, 2.1 ± 0.1 and 2.4 ± 0.3, in agreement with a higher fluorescence for adenine thiochrome derivatives than other thiochrome derivatives. The higher fluorescence of adenylated thiochrome derivatives may be caused by either a higher quantum yield for the latter compounds or higher self-quenching in nonadenylated thiochromes. The first possibility seems unlikely because an interaction between adenosine and thiamine moieties, as suggested above, would probably lead to decreased, rather than increased fluorescence. A more probable explanation would be self-quenching in thiochrome, ThcMP, ThcDP and ThcTP, caused by stacking of the molecules; this is possible because of the planar structure of the conjugated thiochrome part. In adenosine thiamine derivatives, because of the U-shaped structure, such stacking would be unlikely to occur.

Is AThDP a natural compound?

AThTP has only very recently been shown to occur naturally in bacteria where it accumulates during carbon starvation [5]. Concerning AThDP, to date, there is no reference to the compound in the scientific literature. However, it was mentioned in at least two patents in 1969 and 1970 [7,8], but no further data have become available since then. It was claimed [7] that some bacteria (Corynebacterium ammoniagenes or C. glutamicum) are able to synthesize AThDP in the presence of suitable precursors (thiamine, ThMP, adenine, adenosine) added to the medium. Under these conditions, the inventors reported that the bacteria accumulated large amounts of AThDP (0.5–1 mg·mL⁻¹) in the culture medium. It was not clear whether AThDP was synthesized inside the bacteria and then excreted or whether it was synthesized in the periplasmic space and then diffused into the fermentation liquor. We repeated these experiments with C. glutamicum and E. coli, but we did not observe any accumulation of AThDP inside or outside the bacteria. Because of the poor description of the methods used in the patent and the lack of any description of the compound synthesized, it is difficult to draw a conclusion concerning the reasons for our failure to reproduce these results. We were also unable to find any mention of AThDP in subsequent patents and any reference to this compound in the peer-reviewed literature.

However, in E. coli, we observed a transient appearance of AThDP, when the bacteria grown overnight were diluted in Luria–Bertani medium (Fig. 8A). The amounts observed were quite variable, ranging from a few pmol·mg⁻¹ of protein to ∼ 50 pmol·mg⁻¹ of protein, representing a maximum of 2–3% of total thiamine. For comparison, much larger amounts of AThTP could be observed in E. coli under conditions of carbon starvation, i.e. when the bacteria were transferred to a minimal medium without a carbon source. Under these conditions, AThTP slowly accumulates and, after a few hours, reaches a maximum corresponding to ∼ 10–15% of total thiamine [5]. Concerning AThDP, to date, we have no evidence that its appearance might be linked to some kind of cellular stress.

We then looked for the presence of adenylated thiamine compounds in eukaryotes. We have previously shown that AThTP may be detected in small amounts in yeast, the roots of plants and in several organs in the rat [5]. In the mouse, significant amounts of AThDP were found in the liver (Fig. 8B), although it was below the limits of detection in the brain, heart, kidney and skeletal muscle (Table 3). We also found very small amounts (near the detection limit) of AThDP in quail liver (Fig. 8C), but not in other quail tissue (brain, heart, skeletal muscle). In contrast to mouse tissues, AThTP was never observed in any quail tissues. ThTP, however, was found in relatively high amounts in quail brain (4.6% of total thiamine) and skeletal muscle (1.9% of total thiamine) [1], in small amounts in quail heart (0.15% of total thiamine, this study, not shown), and was hardly detectable in quail liver (≤ 0.1% of total thiamine) (Fig. 8C).

In cultured mammalian cells, we found significant amounts of AThTP in 3T3 mouse fibroblasts (Fig. 8D and Table 4), but these cells contained no detectable amounts of AThDP or ThTP. In contrast to 3T3 fibroblasts, Neuro2a neuroblastoma cells contained significant amounts of ThTP but no AThTP. AThDP was not found in any of these cell lines, although it seemed that the commercially available Dulbecco’s modified Eagle’s medium contained a small amount (< 0.01% of thiamine) of this compound.

Determination of thiamine compounds by HPLC

Thiamine compounds, including AThTP and AThDP, were determined by HPLC using a PRP-1 column, as described previously, after transformation to fluorescent thiochrome derivatives [5,32]. Prior to analysis, an 80-μL aliquot was oxidized with 50 μL of 4.3 mm potassium ferricyanide in 15% NaOH. AThTP and AThDP were also quantified using UV detection (254 nm, 535 HPLC detector; Bio-Tek Instruments, Winooski, VT, USA) after separation on a 5-μm Polaris C₁₈ column (150 × 4.6 mm; Varian Benelux, Middelburg, the Netherlands). The mobile phase was composed of 50 mm ammonium acetate adjusted to pH 7.0 and 5% methanol. The flow rate was 1 mL·min⁻¹. All solutions were prepared using milli-Q water (Millipore S.A./N.V., Brussels, Belgium) and all the solvents used for HPLC were of HPLC grade (Biosolve, Valkenswaard, the Netherlands).

Chemical synthesis and purification of AThDP and AThTP

AThTP was synthesized by modification of a previously published method [9] for the synthesis of ThTP and nucleoside triphosphates. All products and solvents were from Sigma-Aldrich NV/SA (Bornem, Belgium). Preliminary tests were made using either 0.7 mmol ThDP (acid form) and 0.7 mmol 5′-AMP (acid form) or 0.7 mmol ThMP (acid form) and 0.7 mmol 5′-AMP. The compounds were dissolved in 700 μL tributylamine and 750 μL H₂O and mixed until a translucent, slightly viscous, solution was obtained. We diluted 5 μL of this mixture in 500 μL dimethylsulfoxide mixed with 450 μL pyridine and finally added 0.15 g N,N′-dicyclohexylcarbodiimide (dissolved in 50 μL pyridine) to start the synthesis. The reaction was allowed to proceed at room temperature and aliquots were taken at different time intervals, diluted 2000 times in water and analyzed by HPLC. Three main compounds (ThTP, AThDP and AThTP) appeared in the mixture.

For purification of the compounds, the synthesis was made on a larger scale: 2.25 mmol ThDP (acid form), 3.5 mmol 5′-AMP (acid form), 3.5 mL (14.5 mmol) tributylamine and 3 mL H₂O were mixed and dissolved in 500 mL dimethylsulfoxide and 445 mL pyridine. Finally, 45 g N,N′-dicyclohexylcarbodiimide (dissolved in 15 mL pyridine) was added and the mixture was incubated for 2 h at room temperature. Addition of 3 L diethyl ether to the mixture led to the precipitation of synthesized compounds. The suspension was centrifuged (1000 g, 10 min) and the precipitate was dissolved in 40 mL H₂O. This solution was applied on a column (8 × 2.5 cm) filled with AG 50W-X8 cation-exchange resin (H⁺ form; Bio-Rad Laboratories, Nazareth Eke, Belgium) equilibrated in water (pH 4.5 with HCl). The column was washed with 100 mL H₂O and 8-mL fractions were collected (flow rate 2 mL·min⁻¹). During this step, ThTP was eluted [9]. All other thiamine derivatives were eluted with 480 mL (60 × 8 mL fractions) ammonium acetate (0.2 m, pH 7.0). Fractions 20–60 were pooled (320 mL) and lyophylized. The powder was dissolved in 25 mL H₂O and layered on a column (8 × 2.5 mL) filled with AG-X1 resin (Cl⁻ form; Bio-Rad). The resin was washed with 120 mL H₂O during which the yellow form was eluted. Residual ThDP and some AThDP were also eluted at this stage. AThDP was eluted with 250 mL ammonium acetate (0.25 m, pH 7.0). The fractions containing AThDP were pooled and lyophilized. AThTP was eluted with 500 mL ammonium acetate (0.5 m, pH 7.0) and lyophilized. The residue was dissolved in 3 mL H₂O and filtered on a Millex-GP filter unit (0.22 μm, dia. 25 mm; Millipore). Aliquots of 100 μL of the pool were then purified on a Polaris C₁₈ HPLC column. The mobile phase consisted of 50 mm ammonium acetate and 5% methanol in water and the flow rate was 1 mL·min⁻¹. AThTP was eluted with a retention time of 7 min, and AThDP was eluted after 14 min. The peaks were collected, lyophilized and used for MS analysis and NMR.

Identification of AThTP and AThDP by ESI tandem MS

Experiments were performed on a Micromass Q-TOF Ultima Global apparatus (Waters Corp., Zellik, Belgium) operated in nano-ESI positive ion mode. The synthesized compounds were injected at a concentration of 150 μm in 50% water/50% acetonitrile. The source parameters were: capillary voltage, 1.8 kV; cone voltage, 100 V; RF lens 1, 90 V; source temperature, 80 °C; collision energy, 6 eV. The fragmentation pattern of the m/z 674.1 was obtained with 30 V acceleration voltage.

Characterization of AThDP and AThTP by ¹H-NMR, ¹³C-NMR and ³¹P-NMR

One-dimensional ¹H-NMR, ¹³C-NMR and ³¹P-NMR spectra were recorded at 25 °C (pH 7.4) on a Bruker Avance 500 spectrometer (Bruker Belgium S.A./N.V., Brussels, Belgium) operating at a proton NMR frequency of 500.13 MHz, using a 5-mm probe and a simple pulse-acquire sequence (30° pulses for ¹H and ³¹P and 90° pulse for ¹³C). Several 2D spectra were also recorded using standard Bruker parameters. NMR data for ThDP, AThDP, ThTP and AThTP are presented in Table 1.

Determination of fluorescence characteristics of thiochrome derivatives of AThDP and AThTP

Fluorescence excitation (emission set at 439 nm) and emission (excitation set at 353 nm) spectra were taken using a SFM 25 fluorescence spectrophotometer (Kontron Instruments, Milan, Italy). One milliliter of a 10 μm thiamine, AThDP or AThTP solution was mixed with 500 μL of 4.3 mm potassium ferricyanide in 15% NaOH and the spectra were taken immediately. AThDP and AThTP could be hydrolyzed by a crude membrane preparation from E. coli as described earlier [5].

Molecular modeling

All calculations were performed at the quantum chemistry level using the B3LYP functional [33] with the polarized double ζ basis set 6-31G(d) [34] and the gaussian 03 suite of programs (Gaussian Inc, Wallingford, CT, USA, 2004). All the degrees of freedom of the geometry have been fully optimized. Several extended and folded conformations have been generated and located as true energy minima.

Bacterial cultures

E. coli (strain BL 21) were grown in Luria–Bertani medium as previously described [5]. C. glutamicum were from the Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (DSM-No. 1412, Braunschweig, Germany) and grown in MMTG medium (glucose 5 g·L⁻¹; tryptone 10 g·L⁻¹; yeast extract 10 g·L⁻¹, NaCl 5 g·L⁻¹ and methionine 0.2 g·L⁻¹) at 30 °C (250 r.p.m.). The bacteria were then incubated in a medium containing glucose (100 g·L⁻¹), urea (6 g·L⁻¹), K₂HPO₄ (10 g·L⁻¹), MgSO₄·7H₂O (10 g·L⁻¹), CaCl₂·2H₂O (0.1 g·L⁻¹), yeast extract (10 g·L⁻¹) and biotin (30 μg·L⁻¹) as described in Sankyo Company, Ltd [8]. After 48 h at 30 °C, ThMP (2 g·L⁻¹) and adenosine (2 g·L⁻¹) were added and thiamine derivatives were determined by HPLC in the fermentation liquor after 24 h after protein precipitation by 12% trichloroacetic acid. Similar experiments were performed with E. coli in Luria–Bertani medium with either thiamine, ThMP, adenine or adenosine as substrates for AThDP synthesis.

Eukaryotic cell culture

Different established cell lines (mouse fibroblast 3T3, human malignant glioma cell line LN-18 and mouse neuroblastoma cell line Neuro2a) were grown at 37 °C in a humidified atmosphere of 95% air, 5% CO₂, in 10-cm Petri dishes in 10 mL of Dulbecco’s modified Eagle’s medium (Neuro2a, 3T3) or RPMI (LN-18) supplemented with fetal bovine serum (10%) and penicillin (100 U·mL⁻¹). Cells were subcultured to a fresh culture dish every 2–3 days. At the third day of culture, the medium was discarded and the cells were detached with trypsin (3T3 and LN-18). After centrifugation at 4000 g for 4 min, the pellets were resuspended in 200 μL of trichloroacetic acid 12%. After centrifugation (5000 g, 3 min), the trichloroacetic acid was extracted with diethyl ether. The pellets obtained were dissolved in 0.8 m NaOH for the protein assay.

Presence of AThDP and AThTP in mouse and quail tissues

Mice (Mus musculus, C57BL6/129SvJ mixed genetic background) and quails (Coturnix japonica japonica) were decapitated and tissue extracts were prepared as previously described [1]. All animal experiments were made in accordance with the directives of the committee for animal care and use of the University of Liège, in accordance with the European Communities Council Directive of 24 November 1986 (86/609/EEC).