Benzoylated Uronic Acid Building Blocks and Synthesis of N-Uronate Conjugates of Lamotrigine

A chemoenzymatic approach towards benzoylated uronic acid building blocks has been investigated starting with benzoylated hexapyranosides using regioselective C-6 enzymatic hydrolysis as the key step. Two of the building blocks were reacted with the antiepileptic drug lamotrigine. Glucuronidation of lamotrigine using methyl (2,3,4-tri-O-benzoyl-α-d-glycopyranosyl bromide)uronate proceeded to give the N2-conjugate. However, lamotrigine-N2-glucuronide was most efficiently synthesised from methyl (2,3,4-tri-O-acetyl-α-d-glucopyranosyl bromide)uronate. Employing nitromethane as solvent with CdCO3 as a base lamotrigine-N2 glucuronide was prepared in a high yield (41%). Also methyl (2,3-di-O-benzoyl-4-deoxy-4-fluoro-α-d-glucosyl bromide)uronate underwent N-glucuronidation, but the product was unstable, eliminating hydrogen fluoride to give the corresponding enoate conjugate.

Lamotrigine (LTG) is a so-called second-generation antiepileptic drug that is being used in a variety of epileptic syndromes as well as in bipolar disorder. LTG is metabolized primarily by conjugation with glucuronic acid, mainly by uridine 5'-diphospho-glucuronosyltransferase 1A4 (UGT-1A4), and presumably also by UGT-2B7 [19,20]. Approximately 70% of an oral dose of LTG is found in the urine as the corresponding N2-glucuronide. A N5-glucuronide has been postulated [21], but has, to our knowledge, never been demonstrated. LTG is subject to marked pharmacokinetic drug-drug interactions with enzyme-inhibiting and -inducing drugs. Moreover, its serum concentrations decline considerably during pregnancy [22]. To investigate the underlying mechanisms for these observations, it is desirable not only to quantify the parent compound, but also the main metabolite. On this background we have undertaken the synthesis uronic acid building blocks and have tested their use in preparation of LTG-uronic acid conjugates.

Uronic Acid Building Blocks
To obtain benzoylated uronic acid derivatives the route shown in Scheme 1 was employed. The starting materials 1a-c were prepared by standard benzoylation [23], whereas 1d was made by fluorination of methyl 2,3,6-tri-O-benzoyl -D-galactopyranoside using DAST or Deoxo-Fluor TM [24]. Enzymatic hydrolysis with Candida rugosa lipase in dioxane/water afforded a regioselective debenzoylation at position 6 giving compounds 2a-d. Scheme 1. Synthetic route to the uronic acid building blocks.
In CRL catalysed hydrolysis of the 4-deoxy-4-fluoro derivative 1d, inhibition phenomena where encountered and full conversion was not reached by a standard single run experiment or by the addition of extra fresh enzyme during the reaction. However, by performing an extractive work-up and restarting the hydrolysis, 90% conversion and 76% isolated yield of 2d was obtained in gram scale synthesis. Ethanolysis of 1d to 2d in hexane was also tested (Scheme 2). A 29% conversion (23% isolated yield) was obtained after 48 h. reaction time. Due to the low rate of reaction this strategy was not investigated further. No benzoyl migration was observed for these compounds under storage as solids or in chloroform solution.
Oxidations of carbohydrates can be performed regioselectively using laccases and TEMPO. In our hands, the model compound phenyl α-D-glucoside was efficiently converted to the corresponding acid. However, no reaction could be observed when 2a was subjected to the same reaction conditions, possibly due to insolubility of the substrate and phase transfer limitations. We therefore turned our attention to alternative oxidation systems. Employing RuCl 3 and potassium periodate as oxidant [25] in a water/acetone solvent mixture, only trace amount of product was observed after 48 h. Changing the solvent to a mixture of acetonitrile, water and carbon tetrachloride proved successful and full conversion was obtained in 4 h reaction time. A Jones oxidation [26], with CrO 3 -H 2 SO 4 was also slow when performed in acetone or in acetonitrile/water/CCl 4 mixture. However, when performed in acetonitrile/CCl 4 , full conversion was obtained in 6-7 h. By this method the uronic acids 3a-d were obtained in 74-79% isolated yield.
Methylation of the carboxylic acid function was performed using trimethylsilyl diazomethane. In pure dichloromethane the reaction did not proceed to completion. However, by applying methanol as a co-solvent full conversion was obtained in 45 min at 0 °C. This gave the uronic acid methyl esters 4a-d in 73-82% isolated yield after crystallisation.
Functional group interconversion of the methyl acetal function to the corresponding bromide was performed by two methods. In a small 50 mg scale, 4a-c could be reacted with HBr/acetic acid to give the 1-bromo compounds 5a-c in 44-59% yield after crystallisation. The anomeric configuration was independent of the configuration of the starting material. Thus, both α-anomer 4b and β anomer 4c upon bromination gave 5b, in accordance with the high preference for axial positioning of the bromo substituents at C-1 [27,28]. Reactions at a 0.5 g scale proceeded more slowly and more impurities were observed, which complicated purification by crystallisation. Silica gel column chromatography resulted in decomposition of 5a-d and a low yield. Therefore, zinc bromide in combination with trimethylsilane bromide [29], was tested in brominations of 4a and 4d. The reaction proceeded to give 5a and 5d as sole products, and allowed for the isolation of 5a and 5d in 75 and 77% yield by crystallisation from ethanol.
As an alternative building block for preparation of the N-2-glucuronide of lamotrigine, the acetylated glucuronic acid derivative 8 was made from D-glucuronolactone (6) by methylation, acetylation and bromination (Scheme 3). Using the method as described by Caldwell et al. [30] more than 30 g of 8 was synthesised starting from 50 g of 6.

Uronic Acid Conjugates
Quaternary N-glucuronides of 1-phenylimidazole [31], cotinine [30], and other nicotine derivatives [32], have previously been synthesised using methyl (2,3,4-tri-O-acetyl-α-D-glucopyranosyl bromide) uronate (8) in melt form without a base. Using compound 8 as starting material in reaction with lamotrigine (9), only a low 8% of the N-glucuronide 11 was obtained by this method. Instead, reactions with CdCO 3 , previously reported to be an efficient base in N-glycosylations, were tested on both 5a and 8 [33][34][35]. In both cases a rapid conversion towards the product was evident when the coupling was performed in refluxing nitromethane (Scheme 4). Starting with 8, hydrolysis of the acetyl functions was performed without isolation of the intermediate using lithium hydroxide followed by pH adjustment with AcOH. The lyophilized product was first purified by cation exchange to a purity of 95%. A final purification by preparative HPLC using a SB-C18 column resulted in a purity of >99% (HPLC-UV) in 41% yield. This compares very favourably with all previous reported synthesis of quaternary N-glucuronides. The reason for the conversion improvement upon using CdCO 3 is not firmly established. However, it has been reasoned that the in situ formation of a Cd 2+ halide is responsible for a heterogeneous catalysis involving these speices [33]. Starting with the benzoylated 5a an 80% conversion to 10 was obtained in 3 h. Conversion could not be improved by extending the reaction time. Following extractive work-up, preparative TLC and crystallisation, 51% of a material indicated by NMR and MS to be a mixture of the α-and β-anomer of 10. Methanolysis of 10 gave the corresponding N-glucuronide 11, but we were not able to purify this material as efficiently as when 8 was used as starting material. Possibly, the challenges encountered with 5a as a building block in this reaction are related to a high anomerisation rate, and the presence of the open form of the hexose as observed for other benzoylated uronic acid derivatives [36].
In analysis of drug metabolites such as glucuronide conjugates, structurally related standards would be highly desirable. The 4-deoxy-4-fluoroglucuronide might in this respect be useful. An alternative starting material is 5d, which was reacted with lamotrigine (9) as shown in Scheme 5. The reaction gave, according to MS and NMR the N-glucuronate 12, however, the product proved unstable under the reaction conditions, and the defluorinated enoate 13 was also formed. The ratio of 12/13 depended on degree of conversion, and a higher amount of 12 was observed when lowering the amount of base. However, these conditions gave low overall conversion and did not allow for efficient isolation of the synthesised products. By employing refluxing conditions and short reaction time, 12 and 13 were isolated semipure in 16% and 13% yield respectively. Elimination reactions have also previously been reported for various galacto derivatives [37][38][39][40], and given the challenges also with non fluorinated analouges, this strategy was discontinued. Rhy et al. [12] has prepared disaccharides containing a 4-fluorinated uronic acid, by performing 6-C oxidation after glycoside coupling. Provided mild conditions could be identified, this might be an alternative strategy to such fluorinated glucuronides.

NMR Characterisation of Lamotrigin Glucuronide
The LTG-glucuronide 11 was subjected to high field NMR analysis. The chemical shift values are summarized in Table 1. Table 1. Proton and carbon NMR chemical shift assignments for Lamotrigine-N2-glucuronide.

Chemicals and Equipment
Candida rugosa lipase was from Sigma-Aldrich (type VII, 700 units/mg solid). Compounds 1a-c [23], 1d [24], 2a-c [41] were prepared as described previously. Silica gel column chromatography was performed using silica gel 60A from Fluka, pore size 40-63 μm. Accurate mass determination (ESI) was performed on an Agilent G1969 TOF MS instrument equipped with a dual electrospray ion source. Melting points were measured by a Mettler FP 5 melting point apparatus and are uncorrected. Optical rotations were measured using sodium D line at 589 nm on a Perkin-Elmer 243 B polarimeter.

NMR Spectroscopy
1 H-and 13 C-NMR spectra were recorded with Bruker Avance 400 or 600 spectrometers. 19 F-NMR was performed on a Bruker Avance 500 operating at 470 MHz. For 1 H-and 13 C-NMR chemical shifts are in ppm rel. to TMS, while for 19 F-NMR the shift values are relative to hexafluorobenzene. Coupling constants are in Hertz.
Compound 11: 1 H-and 13 C-NMR spectra were recorded and assigned by using IP-COSY, NOESY, HSQC and HMBC experiments. 1 H-NMR, 13 C-NMR, NOESY and HSQC spectra were recorded on a Bruker Avance 600 FT-NMR Spectrometer, equipped with a TCI CryoProbe. IP-COSY and HMBC spectra were recorded on a Bruker DPX 400 FT-NMR Spectrometer equipped with a PADUL probe. The NMR solvent (for LTG-glucuronide) used was H 2 O added 10% D 2 O to provide a lock signal. Acetone was used as internal standard ( 1 H shift 2.218 ppm, 13 C shift 30.89 ppm). For the 1 H-NMR, IP-COSY and NOESY experiments, excitation sculpting was used to suppress the water signal. For all experiments, the number of transients was varied to obtain the required signal-to-noise ratio. IP-COSY data were acquired as 2,048 × 128 complex points with spectral widths of 6410 Hz for both frequency domains and 1.5 s relaxation delay. The NOESY data were acquired as 2048 × 256 complex points with spectral widths of 6002 Hz for both frequency domains, a mixing time of 500 ms and 1.0 s relaxation delay. The HSQC data were acquired as 2,048 × 256 complex points with spectral widths of 9,615 Hz and 24,148 Hz in F 2 and F 1 respectively. The experiment had a relaxation delay of 1.0 s and was optimized for an assumed direct coupling of 145 Hz. The HMBC data were acquired as 1,024 × 256 data points with spectral widths of 4,006 Hz and 20,124 Hz in F 2 and F 1 respectively. This experiment has a relaxation delay of 0.9 s and was optimized for an assumed long-range coupling of 8 Hz.

Oxidation to Uronic Acid Derivatives 3a-d
Under an argon atmosphere the benzoylated methyl pyranoside (0.4 mmol) was dissolved in MeCN/CCl 4 (3 mL, 1/1 by vol.) and cooled to −10 °C. Then, a solution of CrO 3 -H 2 SO 4 (200 μL, 0.48 mmol) was added dropwise over 5 min. The reaction mixture was stirred at the room temperature until full conversion as analysed by TLC (6-7 h). The mixture was then filtered, diluted with chloroform (15 mL), washed with saturated NaHCO 3 (10 mL), NaCl (10 mL) and water (5 × 10 mL). Drying over MgSO 4 and concentration under reduced pressure gave a syrup, which was crystallised from CHCl 3 /pentane to give a colourless solid.

Lamotrigine N-2-glucuronide 11
Procedure A: A mixture of lamotrigine (50.0 mg, 0.197 mmol) and bromo-2,3,4-tri-O-acetyl-α-Dgluco pyranuronic acid methyl ester (50.0 mg, 0.125 mmol) was heated at 90 °C under argon. After 20 h a new portion of bromo-2,3,4-tri-O-acetyl-α-D-glucopyranuronic acid methyl ester (50.0 mg, 0.125 mmol) was added and the reaction continued for another 20 h. The melt was then cooled to room temperature, dissolved in methanol (15 mL) and filtered. The methyl ester and acetyl groups were hydrolyzed by addition of lithium hydroxide (0.1 M) until the pH reached 9.0 and stirring was continued at room temperature for 30 min. The alkaline solution was extracted with diethyl ether (3 × 70 mL) and the ether extract containing unrelated bromo-sugar and degradation products were discarded. The aqueous solution was diluted with glacial acetic acid until pH reached 6.0 and the mixture was then lyophilized at high vacuum. The solid material was dissolved in water, added to a Dowex 50W2-400 strong cation-exchange resin, and products eluted with 2 M aquas ammonia. Fractions with UV-response was pooled and lyophilized, before a final separation using preparative HPLC to give LTG-N2-glucuronide (6.2 mg, 0.014 mmol, 8% yield).

Conclusions
Benzoylated uronic acid building blocks have been prepared by a chemo-enzymatic approach using Candida rugosa for selective hydrolysis at C-6. The uronic acid derivatives might have different applications, one being the synthesis of glucuronides. Methyl (2,3,4-tri-O-benzoyl-α-D-glycopyranosyl bromide) uronate could be reacted with lamotrigine to give the N2-conjugate. The N2-glucuronide of lamotrigine was however most easily synthesised from the acetylated precursor using CdCO 3 as base in nitromethane. This gave 41% yield and represents a major improvement to previously published synthetic procedure for producing quaternary N-glucuronides. Complete 1 H-and 13 C-NMR-assignments of the synthesized product have proven the existence of lamotrigine N2-glucuronide. Reaction of lamotrigine with the 4-deoxy-4-fluoro derivative 5d was hampered by fluorine elimination as a side reaction, thus indicating the limitations of this building block in general.