A Novel Approach to Highly Substituted -Carbolines via Reductive Ring Transformation of 2-Acyl-3-isoxazolylindoles

We have worked out a new approach to 1,3,4-trisubstituted -carbolines of pharmaceutical interest. As central building blocks we used 2-acylindoles, which are readily available from indole-2-Weinreb amides. Bromination at C-3, fol-

Previous research was mainly focused on the synthesis of -carbolines bearing various residues at C-1, eventually in combination with substituents on ring C, e.g. a methoxy group at C-7 in harmine, a well-documented inhibitor of monoamine oxidase A (MAO-A) and protein kinases like DYRK1A. [2] The most common approaches to -carbolines bearing residues at C-1 (reviewed in ref. [3] ) start from tryptamine or tryptophan (A), and pyridine ring A is built up using either an aldehyde (Pictet-Spengler reaction) [4] or a carboxylic acid (Bischler-Napieralski reaction). [5] Fully aromatic compounds are obtained in a final dehydrogenation step of the intermediate di-or tetrahydro-carbolines. Alternatively, 1-substituted -carbolines are available via Pd-catalyzed cross-coupling of 1-halogenated -carbolines (C), [6] reaction of 1,9-dimetalated -carboline with electrophiles, [7] and regioselective homolytic substitution reactions. [8] The precursors for these reactions are prepared from tryptamine as well. All of these approaches are primarily suitable for the construction of -carbolines bearing various residues at C-1, and, depending on the utilization of tryptamine precursors  The utilization of tryptamine building blocks in most of the common approaches towards -carbolines means that introduction of additional residues at C-3 and C-4 of the targetcarbolines is typically not possible. Tryptamines bearing additional residues in the side chain are available via electrophilic substitution of indoles at C-3 with appropriately substituted nitroalkenes in several steps. [3] For an application in the synthesis of the alkaloid (S)-brevicolline, see ref. [9] Tryptamine deriva-tives obtained by functionalization of gramine with α-(alkylideneamino)nitriles furnished 1,3-disubstituted -carbolines. [10] A convenient alternative approach was introduced by Larock [11] with the Pd-catalyzed cyclization of 3-alkynyl-2-formylindoles via the corresponding tert-butylimines (B, R 1 = H) to give 3substituted -carbolines. Rossi [12] published a variation starting from 2-acylindoles that provides 1,3-disubstituted -carbolines in a similar manner. Larock's methodology was extended to the synthesis of 3,4-disubstituted -carbolines by reacting internal alkynes with tert-butylimines of 3-iodoindole-2-carbaldehydes (D), but typically mixtures of regioisomers were obtained. [13] Jiao demonstrated that this iminoannulation can be accomplished even without the 3-iodo substituent on the indole-2carbaldimine, if oxygen was used as an oxidant. [14] However, all of these methods for construction of ring A starting from alkynes require protection of the indole nitrogen, and only few examples for readily removable protective groups (sulfonyl, MOM) have been presented by the authors. For some additional approaches to ring A-substituted -carbolines see ref. [15][16][17][18] In the past, we developed a number of -carbolines as inhibitors of protein kinases (CLK1, [19] DYRK1A, [2,19a] PIM1 [19a,20] ), and detected pronounced effects of the substitution patterns at both ring C, the indole nitrogen, and C-1 on the biological activities. However, in the group of fully aromatic -carbolines we worked on until now, substituents at C-3 and C-4 of ring A could not be investigated, since typically our building blocks were prepared from precursors that do not allow functionalization at C-3/C-4. In our attempt to establish broader structure-activity relationship (SAR) analysis of bioactive -carbolines, we saw a need for working out novel methodologies for the synthesis of -carbolines, which still can be modified freely at ring C (by selecting commercially available starting materials with manifold substitution patterns) and at C-1, but in addition the new protocol should enable us to introduce residues at C-3 and C-4 in a predictable manner and without formation of isomeric products.
As the chemistry of 1,3,4-trisubstituted -carbolines is rather underexplored, we worked out an unprecedented approach to this chemotype. For the central step in the construction of the pyridine ring (ring A) we selected a formal cyclocondensation of an 1,5-diketone (or an equivalent E thereof ) under incorporation of ammonia ( Figure 2). Comparable cyclocondensations have been published before by others [21] and by us (using an enol ether as carbonyl equivalent). [22a] As building block for rings B+C as well as C-1 of ring A and the substituent R 1 located there, we selected 2-acylindoles (G/H). These are readily available from indole-2-Weinreb amides, [22] which in turn can, with a broad variety of substituents at the benzene ring, be prepared in a few steps from aromatic aldehydes via the corresponding esters obtained by Hemetsberger-Knittel synthesis and other established methods of indole chemistry. [23] A novel approach to indole-2-Weinreb amides starting from cinnamic acid Weinreb amides has been reported by us recently. [24] The remaining part of ring A (the nitrogen atom, C-3, C-4, as well as the substituents located there) was to be introduced by means of one single building block. An appropriately substituted isoxazole (I/J) appeared most promising for this purpose. On the one hand it should be feasible to introduce isoxazoles at C-3 of the indole building block by means of Pd-catalyzed cross-coupling reactions, and on the other hand, reductive ring cleavage of isoxazoles is known to give Z-enamino ketones. [25] The primary enamino group obtained by reduction should undergo cyclocondensation with the acyl residue at C-2 of the indole directly (or after treatment with an ammonia source) to provide carbon atoms C-3 and C-4 of the envisaged -carboline. The other parts of the isoxazole building block would provide the substituents at C-3 and C-4 ( Figure 2). The inevitable acyl group at C-4 of the resulting -carbolines should offer the opportunity for a broad spectrum of consecutive modifications (reduction, oxidation, addition of nucleophiles) for the construction of novel, even more complex residues at this position. Figure 2. Attempted approach to highly substituted -carbolines via 2-acyl-3-(isoxazol-4-yl)indoles (F) and envisaged routes for Pd-catalyzed cross-coupling reactions of 2-acylindoles (G/H) and isoxazole (I/J) building blocks.
The first challenge in this project was to work out an effective protocol for the connection of the 2-acylindole and isoxazole building blocks. To the best of our knowledge the synthesis of 3-(isoxazol-4-yl)indoles (F) has been the subject of only very few investigations, and only indole-N-protected compounds were obtained in low yields by 1,3-dipolar cycloaddition reactions. [26] So we intended to work out an alternative access via palladium-catalyzed biaryl synthesis. Three options appeared promising for this purpose: route A) direct arylation of 4-unsubstituted isoxazoles (I) [27] with 3-halogenated indoles (G) under C-H activation, or routes B) and C) by Suzuki-Miyaura cross-coupling of building blocks H and J, being either a 3borylated 2-acylindole and a 4-haloisoxazole, or a 3-halogenated 2-acylindole and an isoxazole-4-boronic acid (or an ester thereof ) [28] (Figure 2).
Finally, route C was successful. The 4-borylated isoxazoles required for this protocol can be built up de novo by regioselective 1,3-dipolar cycloaddition of nitrile oxides with alkynylboronates, [32] by Blum's oxyboration of ynone oximes, [33] or are obtained from 4-bromoisoxazoles via bromine-lithium exchange/trapping with trialkylborates [28] or by Pd-catalyzed Miyaura borylation. [34] Initial experiments were performed with commercially available isoxazole pinacolboronate 4a and 2-acetyl-3-bromoindole (1a). [22a] Only poor yields (<15 %) were obtained in a first series of cross-coupling experiments using Pd(0) and Pd(II) catalysts in combination with different phosphine ligands (PPh 3 , dppf, SPhos) and bases (DIPEA, K 2 CO 3 ). These frustrating experiments gave rise to the question whether the free NH group of the indole prevented cross-coupling. So we introduced the SEM protective group, which we had identified as a very useful indole protective group in previous work. [35] But even SEM derivative 1b gave only slightly improved yields (up to 20 %) under the cross-coupling conditions examined with 1a before. Finally, further modifications of the catalyst system led to the Pd(PPh 3 ) 4 /Cs 2 CO 3 system, which gave a 61 % yield of the desired biaryl 5. With this intermediate in hands the ring transformation reaction was examined. Reductive opening of the isoxazole ring was first attempted under conditions (catalytic hydrogenation with Pd/C catalyst in ethanolic KOH) which had given best results in our previously published approach to canthin-4-ones involving a related ring transformation. [36] But only a complex mixture of products was obtained. Fortunately, hydrogenation in ethanol without added base resulted in clean conversion (79 % yield) to the expected Z-enamino ketone 6. The Z-configuration of the enamino ketone, which is stabilized by an intramolecular hydrogen bond, obviously prevents direct cyclization to the desired -carboline 7 due to steric reasons. This prompted us to perform the hydrogenation in presence of ammonium acetate, hoping that nucleophilic addition/elimination of ammonia to the vinylogous ketone moiety should in situ give an E-enamino ketone that can undergo cyclocondensation with the acetyl group at C-2 of the indole to give the -carboline 7. Cyclization did not occur under these conditions, and once again the Z-enamino ketone 6 was isolated. Ring closure to the -carboline 7 was achieved by heating the crude enamino ketone with ammonium acetate in ethanol/acetic acid at 60°C (87 % overall yield from 5). Finally, we found that hydrogenation in presence of cesium carbonate directly leads to the desired -carboline 7. Deprotection of the SEM group was achieved in 55 % yield (not optimized) under standard conditions (TBAF, THF) to give the 1,3,4trisubstituted -carboline 8 (Scheme 2).
Having proven the feasibility of the new approach we intended to reduce the number of steps of synthesis. To our delight the optimized Suzuki-Miyaura cross-coupling conditions (Pd(PPh 3 ) 4 /Cs 2 CO 3 catalyst system) also worked with N-unprotected 2-acyl-3-bromoindoles. Comparable results were obtained with sodium carbonate here, but in contrast to cesium carbonate, this base did not dissolve completely in the reaction mixture. With alkyl, phenyl and heteroaryl (2-thienyl) ketones (1a, 9a-g) the cross-coupling with 3,5-dimethylisoxazole-4pinacolboronate (4a) worked in acceptable yields (50-69 %) to give biaryls 10, 11a-g. In contrast to the above mentioned N-SEM derivative 5, the N-unsubstituted intermediate 10 did not undergo reductive ring cleavage of the isoxazole moiety under neutral conditions and with added ammonium acetate, however in presence of Cs 2 CO 3 once again smooth reductive ring cleavage occurred, and fortunately, the formed Z-enamino ketones underwent direct ring closure to the target -carbolines 8, 12a-g under the reaction conditions in yields ranging Finally, we investigated other isoxazole-4-boronates bearing different aliphatic, aromatic and heteroaromatic residues at C-3 and C-5. The boronates 4b/4c were obtained from the corresponding 3,5-disubstituted isoxazoles by bromination at C-4, [37] followed by bromine-lithium exchange with n-butyllithium, and trapping with B-isopropoxy-pinacolborane. [38] Subjecting these boronates to cross-coupling with bromoindoles 1a/9d, followed by reductive ring transformation (yields 68-88 %) gave the highly substituted -carbolines 12h-k. To our surprise, the pyridyl ketones arising from 5-pyridylisoxazole building block 4c were further reduced to give the corresponding pyridyl carbinols 12j and 12k (Scheme 3).

Conclusion
In conclusion, we have worked out a new and general protocol for the synthesis of 1,3,4-trisubstituted -carbolines (with the option for including additional substituents in the carbocyclic ring C, if substituted indole precursors are utilized). Readily available 2-acylindoles (aliphatic, aromatic, heteroaromatic acyl residues are equally suitable) are easily brominated at C-3. Suzuki-Miyaura cross-coupling with variously substituted and readily available isoxazole-4-pinacol boronates gives biaryls, which can be directly converted into the desired -carbolines upon Pd-catalyzed hydrogenation in presence of Cs 2 CO 3 in the sense of a reductive isoxazole-pyridine ring transformation. This new method favorably compares with previously developed methods for the synthesis of highly substituted -carbolines, since it utilizes readily available building blocks, proceeds in a small number of steps, and gives complex compounds with predictable substitution patterns. In contrast to previously published protocols for the synthesis of 3,4-disubstituted -carbolines via reaction of internal alkynes and tert-butylimines of (3iodo)indole-2-carbaldehydes [13,14] via masked 1,5-dicarbonyl compounds [21b] our method is advantageous since it needs neither protection of the keto group at C-2 of the indole nor of the indole nitrogen. This new protocol should be of high interest for the total synthesis of complex -carboline alkaloids and synthetic -carbolines as drug candidates, hence opening new opportunities for systematic investigations of structure-activity relationships also including broad variation of substitution patterns at ring A.

Experimental Section
Solvents used were of HPLC grade or p.a. grade and/or purified according to standard procedures. All chemicals used were of analytical grade. Tetrahydrofuran was dried with sodium and distilled before usage. Chloroform was dried with molecular sieves 3 Å. Melting points were determined by open tube capillary method on a Büchi melting point B-450 apparatus and are uncorrected. IR measurements were carried out with a Perkin-Elmer FTIR Paragon 1000 spectrometer or with a Jasco FT/IR-4100 as KBr pellets or as films. NMR spectra were recorded on Avance III HD 400 MHz Bruker Bio-Spin and Avance III HD 500 MHz Bruker BioSpin spectrometers. Spectra were recorded in deuterated solvents and chemical shifts are reported in parts per million (ppm). J values are given in Hertz. Multiplicities are abbreviated as follows: s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet. Signal assignments were carried out based on 1 H, 13 C, DEPT, HMQC, HMBC and COSY spectra. NMR spectra were analyzed with the NMR software MestReNova, Version 5.1.1-3092. Mass spectra were performed by electron impact (EI) at 70 eV on a Thermo Finnigan MAT 95 or Jeol GCmate II spectrometer or by electrospray ionization (ESI) using a Thermo Finnigan LTQ FT Ultra Fourier Transform Ion Cyclotron resonance mass spectrometer. All reactions were monitored by thin-layer chromatography (TLC) using pre-coated plastic sheets POLYGRAM® SIL G/UV254 from Macherey-Nagel. Chromatographic purification of products was performed by flash column chromatography (FCC) on Merck silica gel 60 (0.015-0.040 mm). HPLC purities were determined using a HP Agilent 1100 HPLC with a diode array detector and an InfinityLab Poroshell column (120 EC-CN, 4.6 × 150 mm, 4.6 Micron). The column flow was 1.0 mL/min and the temperature 50°C. Either acetonitrile/water, 70:30 (method a), 50:50 (method b) or 50:49.9 and 0.1 part 1.0 M formic acid (method c) was used as eluent.
The 2-acyl-3-isoxazolylindole 10,11a-k (0.15-0.30 mmol), palladium on carbon (10 %) (100 mg/mmol starting material) and an excess of cesium carbonate (1.5-3.0 equivalents) were disperged in anhydrous ethanol (2 mL). The reaction mixture was hydrogenated at 35 bar and 40°C for 19 h. After cooling to room temperature, the catalyst was filtered off through a pad of celite, washed with ethyl acetate and the filtrate was evaporated under reduced pressure. Purification was accomplished by FCC.