A Structurally Diverse Heterocyclic Library by Decoration of Oxcarbazepine Scaffold

A library of new heterocyclic systems was synthesized starting from oxcarbazepine (OXC, Trileptal®, 10-oxo-10,11-dihydro-5H-dibenzo[b,f]azepine-5-carboxamide). The key for these transformations is the α-enolizable ketone present on the [d]-side of our starting material OXC, thus, an in depth investigation of the literature to find heteroannulation reactions for substrates carrying an α-enolizable ketone gave us a boost to discover an excellent derivatization strategy and [3+2], [4+2] and [4+1] approaches were successfully developed. Almost always a pre-functionalization was needed, but also the direct one-pot heterocycle construction was also explored.

Although the precise mechanism by which OXC exerts its antiseizure effect is unknown, it is well documented that OXC is completely adsorbed and extensively metabolized (approximately 70%) by reductase enzymes to its pharmacologically active 10-monohydroxy derivative form eslicarbazepine ((S)-licarbazepine, MHD, Figure 1) [5]. The structure-activity relationships to date have shown that the potency and selectivity of 1 are linked to the N-carboxamido group embodied in the twisted boat conformation of the dibenzoazepine core [6]. Accordingly, subsequent synthetic efforts focused on chemical manipulation of the 10-position, mimicking the active MDH with the aim of improving the pharmacological profile [7,8]. At present, an in-depth manipulation of 1 in order to find activities other than the anticonvulsant one has not been previously undertaken. Due to our interest in heterocyclic chemistry [9][10][11][12][13][14] and to the availability in large quantities of 1, we envisioned that 1 would be a versatile key intermediate for further decoration of the azepine subunit, to obtain a new library of compounds of high potential value in the drug research field. We report herein experimental procedures for the synthesis of compounds based on the template 1 (a large part of them with intact N-carboxamido functions). Most reactions were run only once and the reported yields are therefore unoptimized. A diverse set of 5-and 6-membered heterocyclic rings (e.g., oxazole, pyrrole, indole, thiazole, pyrazine, quinoxaline etc) were appended, thereby paving the way to [d]-fused heterocycles which have been largely neglected or are totally absent in the literature. This library of compounds will be subjected to high throughput screening for deciphering their biological activity and these results will be reported in due course.

Results and Discussion
The de novo synthesis of heterocyclic ring systems represents one of the most important topics and this goal was frequently achieved by cyclization strategies. Capitalizing on the presence of a carbonyl function on the [d]-side of OXC, the molecules described in this work could be structurally categorized in three classes depending on the type of pre-functionalization of 1 required prior to the heterocyclization step: pre-functionalization of position 10 (class a); pre-functionalization of position 11 (class b); no pre-functionalization required (class c) (Scheme 1). Scheme 1. Synthetic strategy en route to new fused heterocycles from oxcarbazepine (1).

Class a
The venerable (first reported over 120 years ago) Fischer reaction still stands out as one of the most efficient approach to indoles [15,16]. A number of literature methods for indolization of enolizable carbonyls with aryl hydrazines were examined in an effort to convert 1 to 3 (Scheme 2). Attempts to carry out this transformation (with phenylhydrazine and 1 as [C-C-N] and [C-C] units, respectively, in a [3+2] strategy) in a one-pot procedure under a variety of solvents and acid catalysts resulted in formation of dark mixtures in which only trace amounts of 3 could be detected.

Scheme 2. [3+2] Fischer indolization like strategy.
Considerably better results were obtained starting from the known 10-phenylhydrazone 2 [8] which, on exposure to refluxing acetic acid for 1 hour, underwent the desired Fischer indolization and 3 was isolated in 79% yield. However, much to our surprise, exposure of pyrid-2-yl hydrazone 7 to the same reaction conditions did not afford the anticipated product 8 but rather yielded an intractable mixture of unidentified products. This reaction was attempted several times, producing, in all cases, similar results.
We planned to apply the Fischer strategy using the 10-hydrazone 4 [8] as [C-C-N] unit and cyclohexanone and OXC (1) itself as [C-C] units. In this context we found the hydrazone formation/Fischer indolization (MeOH, HCl, 68 °C, 30 min.) was the most effective route and therefore 5 and 6 were prepared without significant complications in 69% and 58% yield, respectively. Compound 6 is formally the product of a Piloty-Robinson reaction on the transient azine [17,18].
The use of oxime 9 [8] as [C-C-N] unit in a [3+2] approach facilitated considerably the preparation of oxazole-and pyrrole-fused analogues (Scheme 3). In the event, cyclocondensation of 9 with benzoyl chloride in the presence of pyridine (PhCl, 132 °C, 30 min.) [19] gave oxazole 10 in fair (40%) yield. Furthermore, 9 is ideally suited to serve as a starting material to prepare pyrrolo [2,3-d]azepines 11 by using the modified methodology of Trofimov [20]. Accordingly, 11a was obtained by reaction of 11 with methyl propiolate catalysed by TEA in MeCN-DMF (4:1) at r.t. for 1 hour. The transient O-vinyl oxime derivative underwent a tandem [3,3]-sigmatropic rearrangement-cyclocondensation to 11a (57% isolated yield). This reaction is usually reported [21] to require more drastic conditions (DMSO, >90 °C) than those reported here. It was found that this approach could be successfully extended to other electron-deficient alkynes (e.g., 1-phenyl-2-propyn-1-one and its 4-nitrophenyl derivative) leading to 11b and 11c in 71 and 78% yield, respectively. It is worth mentioning that reaction of more electrophilic DMAD with 9 under the same conditions failed to provide the expected pyrrole derivative 11d, a consequence presumably of the overwhelming steric effect of the additional CO 2 Me in the six-membered cyclic transition state.   The [4+1] approach is rarely used in the annulation of a carbocyclic scaffold, but it represents the basic mode of accessing heterocycles embodying a N-N-X subunit (Scheme 4). To this end, hetero-cyclization of semicarbazone 12 (as [C-C-N-N] unit) [8] to the corresponding 1,2,3-thiadiazole 9 (83%) was readily accomplished with SOCl 2 ([S] unit) [22]. By analogy to the above achieved Hurd-Mori reaction, the treatment of 4 with SeO 2 [23] in refluxing AcOH also proceeded to furnish primarily the selenadiazole 10 in 48% yield. A 2007 report by Huazhou et al. [24] describes the high-yielding generation of 2H-1,2,3-diazaphosphole (starting from semicarbazones) through the agency of a POCl 3 -SOCl 2 mixture and subsequent treatment with EtOH. However, exposure of 4 to these conditions failed to provide the corresponding diazaphosphole 15 and no further time was invested to understand this failure.

Class b
In the second approach (pre-functionalization of the position 11) α-bromo ketone 15 served as pivotal intermediate for further elaboration. It is well known that α-halocarbonyls have been successfully exploited in the building of numerous 5-and 6-membered heterocycles [25]. The bromo ketone 16 was prepared by a literature procedure [26] and its reaction with thiourea (Hantzsch reaction) in refluxing EtOH proceeded smoothly and consistently to provide the thiazole derivative 17 in a pleasing 43% yield. Spurred by this finding, we set out to explore the reactions of 15 with other 1,3 N,S-bis(nucleophiles) (i.e., benzothioamide b and ethyl 2-amino-2-thioxoacetate c) ( Figure 2).

AcOH
ROH Δ Surprisingly, it was not possible to prepare the corresponding thiazoles 18 and 19. Reaction conditions (refluxing EtOH) similar to those for 17 gave no reaction, while higher temperatures (up to 120 °C in a heavy wall-pressure tube) or prolonged reaction times (up to 6 days) gave inconsistent or messy results. It seemed likely that this failure was due to product/reagent decomposition which ensued at the high temperatures and extended reaction times necessary to heterocyclization, so more robust reagents might be desirable. Accordingly, we chose to examine a structurally diverse set of 1,3 N,N-(i.e., 2-amino thiazole e, 2-aminopyridine f and 2-aminopyrimidine g), 1,3 N,O-(urea d) and 1,3 C,N-bis(nucleophiles): (2-methylpyridine h) ( Figure 2). These substrates were expected to be even less reactive, but these difficulties have been circumvented by switching to butan-1-ol as a solvent with ideal boiling point, solubilization characteristics, and polarity.
We found that heating 16 with 2-aminothiazole e in refluxing n-BuOH for 6 h did indeed effect cyclization, providing primarily the thiazole 21 as the only identifiable product in 56% yield. The optimized reaction conditions were applied to the other representative N,N-, N,O-and C,N-bis(nucleophiles). Generally, the yields are satisfactory (32%-53% range), the exception being 2-methylpyridine h, which did not perform well in the Tschitschibabin indolizine synthesis (viz. condensation of α-haloketones with 2-alkylpyridines) [27]. The fact that h did not give any noticeable quantity of the expected 25 is rather difficult to interpret, and we cannot explain this discrepancy at the moment. Interestingly, routes a and b provided access to an array of diversely substituted derivatives with the N-carboxamido group still attached to azepine core.
Another approach to compounds belonging to class b entailed the installation of an α-dicarbonyl function at the [d]-edge of the azepine subunit in 1 (to 25) followed by the heterocyclization step (Scheme 5). The two-fold benzylic oxidation of carbamazepine CBZ in the presence of a metal salt and N-hydroxyphthalimide has been shown by Alsters et al [28] to give the required 25 (in only 8% at best!) and, as far as we are aware, this constitutes the only synthesis of this compound. Owing to the availability in large quantities of 1, efforts were taken to oxidize it using the Alsters conditions. However, it appeared that this methodology is of limited applicability even to our case. Different methods for oxidation of enolizable carbonyl to α-diketones were tested, but they uniformly failed under a range of conditions: the starting material either proved inert or underwent a messy reaction. On exposure to freshly sublimed SeO 2 in refluxing dioxane (7 h) the desired 25 was not observed. Instead, α-diketone 26 was produced as the single reaction product, which crystallized from the crude reaction mixture in 90% yield. OXC reacted also with benzeneseleninic anhydride in chlorobenzene at 120 °C for 6 h to provide the same diketone (87%). Unfortunately, despite considerable efforts, the oxidation of 1 without compromising the pendant N-carboxamido group proved to be fruitless and it was therefore decided to pursue this approach using 26 as a starting material. Actually, reaction of derivatives devoid of N-carboxamido function with chlorosulphonyl isocyanate (CSI) would help address this problem [29]. 1,2-Dicarbonyl compounds are versatile intermediates for further elaboration leading, i.e., to five-to seven-membered ring systems and 26 proved to be no exception. Specifically, treatment of a solution of 26 in AcOH with o-phenylenediamine (118 °C, 1 h) effected conversion to the quinoxaline derivative 27 [30].
Following this procedure, either diaminomaleonitrile or semicarbazide hydrochloride participated as 1,4 N,N-bis(nucleophiles) leading to the isolation of pyrazine 28 and triazinone 29 in 65 and 19% yield, respectively. Further structural diversity could be achieved via MultiComponent Reactions (MCRs). In this context, the reaction of 1,2-dicarbonyls with an aldehyde and NH 3 , originally developed by Debus and Radziszewski [31], is a remarkably effective method for the preparation of 1H-2,4,5-trisubstituted imidazoles. Thus, the 1H-imidazole 30 could be accessed, albeit in a modest 28% yield, from reaction of 26 with benzaldehyde and NH 4 OAc (as a source of NH 3 ) in AcOH at 80 °C.
The disappointing yields obtained by reacting 26 under acidic conditions represent the manifestation of the vexing problem associated with ring contraction-dehydration process leading to 31 as an end-point product.
Nevertheless, the participation of 1 in other cyclocondensation reactions proved to be a tricky undertaking. The application of Gewald conditions [32] to the reaction of 1 with ethyl cyanoacetate and sulphur in the presence of diethylamine (MeOH, 65 °C, 20 h) provided only extensive decomposition products and no isolable quantities of the expected 2-aminotiophene derivative 33. Furthermore, attempts at replacing sulphur by selenium under the same conditions to achieve a seleno-Gewald multicomponent reaction [33] also gave a complex mixture of products, none of which was the desired 2-aminoselenophene 34.
In a similar vein, when 1 was subjected to the conditions under which other cyclic ketones were successfully cyclized (N-methyl-N-propargylamine, PhMe, 110 °C, sealed tube) to annulated [b]pyrroles through the intermediacy of a N-vinyl-N-propargylamine [34], no traces of 35 were detected in the reaction mixture over a 8-day time period. Finally, the reluctance of 1 to participate in these reactions was also witnessed by the failure met in the Pfitzinger protocol (isatin, KOH, H 2 O, 100 °C) [35], so this strategy was not pursued further.

Experimental
All reactions were performed using standard glassware and IKA num heating plates. Reactions utilizing air-sensitive reagents were performed in dried glassware under a nitrogen atmosphere. Solvents were used as received without further purifications. All reagents, if not otherwise specified, were used as received and, if necessary, stored under inert gas. Oxcarbazepine was supplied by Trifarma SpA, Ceriano Laghetto, Italy. For thin-layer chromatography (TLC) analysis Macherey-Nagel Polygram ® sil G/UV 254 pre-coated plates were used. Column chromatography was performed on silica gel 60A (70-200 µm) (Carlo Erba Reagents, Arese, Italy). Melting point (mp) determinations were performed by using a Gallenkamp melting point apparatus (Weiss-Gallenkamp, Loughborough, UK). 1 H-NMR (400 MHz) and 13 C-NMR (100 MHz) were measured on a AV400 spectrometer (Bruker Corporation, Billerica, MA, USA). Chemical shifts (δ) are expressed in parts per million (ppm) and coupling constants are given in hertz (Hz). Splitting patterns are indicated as follows: s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, br = broad, dd = doublet-doublet, td = triplet-doublet. Chemical ionization mass spectra (+ve mode) (CI + -MS) were performed on a Finnigan-MAT TSQ70 instrument with isobutane as the reactant gas. Elemental analyses were performed on a Perkin Elmer Series II CHNS/O Analyzer 2400. (2). The compound was prepared following the procedure present in the literature and the spectroscopic data were in accordance with those reported [8]. Yield: 88%.

Conclusions
To summarize, a structurally diverse heterocyclic library sharing a common dibenzo[b,f]azepine scaffold, decorated with [d]-fused heterocycles has been reported, thus providing a foundation for the further investigation of their biological activities.