Aminomethylation/hydrogenolysis as an alternative to direct methylation of metalated isoquinolines – a novel total synthesis of the alkaloid 7-hydroxy-6-methoxy-1-methylisoquinoline

Introduction

The isoquinoline ring system has considerable relevance in drug development, since it is found in both complex [1] and “simple” alkaloids (variously substituted derivatives of the native isoquinoline) isolated from different plants [2], and antiviral and antimicrobial activities have been found for numerous isoquinoline alkaloids [3]. Further, isoquinoline is regarded as a “privileged scaffold” in drug design, and a large number of drug candidates containing this partial structure are in clinical development [4]. Consequently, synthetic approaches enabling a free variation of substituents on this heteroaromatic ring system are required. Numerous methods have been published over the decades for the construction of highly substituted isoquinolines [4]. Alternatively, subsequent functionalization of isoquinolines is feasible, e.g., via Pd-catalyzed C–H functionalization [5] or regioselective direct ring metalation (at C1) [6-8]. General aspects of the direct methylation of electron-deficient N-heterocycles have been reviewed [9].

In a project aimed at the synthesis of tri- and tetracylic alkaloids containing the isoquinoline scaffold, we were interested in isoquinoline building blocks which bear a methyl group at C1, since this group should, due to its intrinsic C–H acidity, open opportunities for further functionalization. Since most isoquinoline alkaloids bear hydroxy and/or methoxy substituents at the positions 6 and 7, we regarded 7-hydroxy-6-methoxy-1-methylisoquinoline (1) as well as analogues containing a protected 7-hydroxy group as highly attractive building blocks for our project. A literature search revealed that compound 1 is in fact a natural product. It was isolated from the trunk bark of the Taiwanese tree Hernandia nymphaeifolia (Hernandiaceae) in 1996 [10]. No data on the biological activities of this alkaloid have been reported so far. Further, this literature search showed that two total syntheses of isoquinoline 1 had been published even before its identification as a natural product (Figure 1). In 1963, Franck and Blaschke [11] obtained 1 by dehydrogenation of its 1,2,3,4-tetrahydro analogue (which itself had to be prepared in several steps) with MnO₂ in poor yield (24% crude product; yield of the final crystallization step for purification not given). In 1965, Bruderer and Brossi [12] reported on an approach whose central step was a Pictet–Gams cyclization of N-acetylated β,3-dimethoxy-4-benzyloxyphenethylamine, which gave the 7-O-benzyl derivative of alkaloid 1 in 27% yield. The required precursor is available from commercially available precursor 2 via a nitrostyrene intermediate in three steps in 45% overall yield [13]. In conclusion, no efficient approach to this “simple” alkaloid 1 has yet been published.

This prompted us to work out a novel synthetic access to alkaloid 1. Our new strategy was fueled by our previous findings in the course of new approaches to benzylisoquinoline, oxoaporphine [7], and oxoisoaporphine alkaloids [8], where we could demonstrate the power of regioselective direct ring metalations of isoquinolines at C1 with sterically hindered amide bases like TMPMgCl∙LiCl (Knochel–Hauser base) [6]. For our present purpose, an appropriate 1-metalated isoquinoline species was to be converted into the corresponding 1-methyl product. Since the 7-hydroxy group of target alkaloid 1 is not compatible with the metalation reagent, the corresponding benzyl ether 3 was selected as central building block.

Results and Discussion

In our previous work we prepared isoquinoline 3 in a three-step procedure starting from commercially available O-benzylisovanillin (2) in a modified Pomeranz–Fritsch reaction published by Reimann and Renz [14]. In the present investigations, we even optimized this procedure, ending up with a protocol that does not afford purification of any of the intermediates, and gives isoquinoline 3 in a straightforward operation in 65% isolated yield. For this purpose, the starting aldehyde 2 was subjected to a reductive amination with aminoacetaldehyde dimethyl acetal and NaBH₄, followed by N-tosylation and hydrochloric acid-mediated cyclization under concomitant N-detosylation and aromatization. Direct ring metalation of 3 with TMPMgCl∙LiCl was performed as described by us previously [7]. First attempts for a direct methylation at C1 with iodomethane failed completely. Only upon addition of catalytic amounts of CuCN∙2LiCl [15] significant methylation took place. The crude reaction mixture contained almost equimolar amounts of the desired 1-methylisoquinoline 4 and starting material 3. Due to the very similar polarities of 3 and 4, chromatographic separation was very tedious, and only 34% of methyl compound 4 was isolated, accompanied by about 30% of starting material 3 and mixed fractions. Debenzylation of 4 by catalytic hydrogenation in methanol solution under palladium catalysis gave the alkaloid 1 in almost quantitative yield (Scheme 1).

So this new metalation/methylation protocol offered a new approach to alkaloid 1 with an overall yield (22% from commercially available precursor 2) much higher than in the above mentioned previously published methods [11,12]. However, the unsatisfactory separation of 1-methylisoquinoline intermediate 4 from its unmethylated precursor 3 rendered this protocol unattractive. This is a problem that is common to numerous methylation protocols, including methylation of metalated arenes [16], methylation of arenes bearing directing groups with methylmagnesium bromide [17], and direct methylation using radical reactions [9,18,19].

Hence, we searched for a more viable method for the introduction of the methyl group. A prerequisite was that the organometallic intermediate should be trapped by an electrophile which could later be transformed into a methyl group, but the trapping product should be easily separable from the starting material 3. For this purpose we selected Eschenmoser’s salt (N,N-dimethylmethyleniminium iodide) as an electrophile. Trapping of the metalated species was expected to give the N,N-dimethylaminomethyl derivative 5 in reasonable yield. Later on, hydrogenolytic cleavage of the generated benzylamine-type group at C1 should give the desired 1-methyl moiety. Related reductive cleavage reactions have been published earlier by Möhrle [20] for phenolic Mannich bases in the course of the total synthesis of the naphthalene-derived natural product plumbagin.

In fact, trapping 1-metalated isoquinoline 3 with Eschenmoser’s salt gave the aminomethyl derivative 5 in 37% yield. Chromatographic separation from starting material 3 (recovered yield: 32%) was unproblematic.

Surprising results were obtained in our hydrogenolysis experiments with 5, which were aimed at simultaneous O-debenzylation at the 7-position and conversion of the N,N-dimethylaminomethyl group at C1 into a methyl group. Hydrogenation in presence of palladium as catalyst at 1 bar in the presence or absence of small amounts of sulfuric acid gave the phenolic product 6 in high yield with unchanged N,N-dimethylaminomethyl group. The same result was obtained at high pressure (40 bar) and upon addition of formic acid for accelerating hydrogenolysis [21]. Obviously, and in contrast to earlier reports on related naphthol Mannich bases [20], the benzylamine moiety of 5 is resistant to hydrogenolysis, whereas the benzyl ether is readily removed. This order of reactivity is known from previous work [22]. Since it is further known that quaternary benzylammonium compounds undergo hydrogenolysis easier than the corresponding tertiary benzylamines [22], we converted amine 5 into methoiodide 7 by treatment with iodomethane. This salt was obtained in pure form in 78% yield by simply collecting the precipitate, and there was no indication of an undesired methylation of the ring nitrogen of the isoquinoline. High-pressure hydrogenolysis of 7 gave only minor amounts (about 7%) of the desired alkaloid 1, but 1-methyl compound 4 with intact O-benzyl residue was isolated in 54% yield. The surprising stability of the O-benzyl residue in this experiment might be due to the iodide counterion, which is known to be a poison for palladium catalysts. Cleavage of the highly reactive benzylammonium residue still takes place, but O-debenzylation is predominantly suppressed by this catalyst poison. Finally, poisoning of the catalyst was prevented by simply passing a solution of the methoiodide 7 through a chloride-loaded ion exchanger prior to catalytic hydrogenation. In this manner alkaloid 1 was obtained in 94% yield (18% overall yield from commercially available precursor 2, Scheme 2).

Conclusion

In conclusion, we have worked out two novel protocols for the introduction of a methyl group at C1 of isoquinolines. These were applied to the total synthesis of the alkaloid 7-hydroxy-6-methoxy-1-methylisoquinoline (1), but should also be of value for the synthesis of other 1-methylisoquinolines. We could demonstrate that the aminomethylation of metalated arenes with Eschenmoser’s salt followed by hydrogenolytic cleavage is a highly attractive alternative to direct ring methylations. Further, new insights into selectivity of O- and N-debenzylation reactions should be useful for future natural product and drug syntheses.

Experimental

For experimental procedures and copies of ¹H and ¹³C NMR spectra of all compounds see Supporting Information File 1.