Biomimetic Phosphate-Catalyzed Pictet–Spengler Reaction for the Synthesis of 1,1′-Disubstituted and Spiro-Tetrahydroisoquinoline Alkaloids

Tetrahydroisoquinoline (THIQ) alkaloids are an important group of compounds that exhibit a range of bioactivities. Here, a phosphate buffer-catalyzed Pictet–Spengler reaction (PSR) using unreactive ketone substrates is described. A variety of 1,1′-disubstituted and spiro-tetrahydroisoquinoline alkaloids were readily prepared in one-step and high yields, highlighting the general applicability of this approach. This study features the role of phosphate in the aqueous-based PSR and provides an atom-efficient, sustainable route to new THIQs.


■ INTRODUCTION
Tetrahydroisoquinolines (THIQs) are an important group of pharmacologically active compounds exhibiting for example, anticancer, antimalarial, analgesic, antithrombotic, and central nervous system-related activities. A subgroup of THIQs, the benzylisoquinoline alkaloids (BIAs) are a key family of natural products found in higher plants and mammals. 1 The BIAs in particular have attracted the attention of many researchers in recent years, giving rise to a range of patents and publications. 2−8 Amongst the THIQs, 1,1′-spiro compounds have proven to be synthetically challenging to synthesize, and they include the antiplasmodial compound 1, natural product ochotensines 2 and 3, and trabectedin 4 ( Figure 1). 9−11 Different strategies have been used to access THIQs, including natural product extraction from plant sources, affording the desired alkaloids in low yields. In recent years, efforts have also been focused on the expression of biocatalysts for use in the stereoselective synthesis of these alkaloids. For example, in plants norcoclaurine synthases (NCSs) catalyze the first committed step in the BIA biosynthetic pathway, between dopamine 5 and 4-hydroxyphenyl acetaldehyde (4-HPAA) and they have been used in the synthesis. 12−15 The Pictet−Spenglerase NCS has proven to be particularly tolerant towards a range of aldehydes and more recently has been reported to accept ketones. 16−20 While the use of NCS to generate 1,1′-substituted THIQs is a particularly powerful strategy in stereoselective synthesis, a complementary method to access achiral compounds, or an approach that does not necessitate the production of biocatalysts would also be highly valuable.
Many methods have been reported for synthesizing THIQs including the Pictet−Spengler or Bischler−Napieralski reactions, 21,22 but typically for these reactions, particularly when using ketones, high temperatures, strong acids, or superacids are required. 23,24 An example includes the reaction of 1-(3hydroxyphenyl)-2-aminoethanol with cyclohexanone to give a spirooxazolidine which when fused at 150°C produces 1,1′spiro-hexanotetrahydroisoquinoline. 25,26 Such reaction conditions are however incompatible with less stable phenethylamines such as catechols. To overcome some of these problems, Lewis acid catalysts such as o-benzenedisulfonamide, 27 titanium(IV) isopropoxide, 28 or calcium bis-1,1,1,3,3,3-hexafluoroisopropoxide 29 have been used (Scheme 1a,b), but there are inherent limitations associated with the preparation of such catalysts including high costs or the requirement of a drybox. 29 In addition, the key phenolic groups must be protected or side products can be formed. Phosphate salts are a useful alternative due to the lower costs and milder reaction conditions that can be used. 30 Indeed, we have previously reported the use of aqueous phosphate in a Pictet−Spengler reaction (PSR) for the synthesis of THIQs (Scheme 1c). 30 The reaction was extended to a range of aromatic, aliphatic, and heterocyclic aldehydes, but not with ketone substrates which are more sterically challenging and less reactive.
Here, we describe the discovery that aqueous phosphate reaction conditions can also be used to synthesize THIQs using ketones. The reaction conditions were optimized and applied to variously decorated phenethylamines and a wide range of cyclic and acyclic ketones in high yield (Scheme 1d). Some of the THIQs synthesized have also been reported to possess interesting bioactivities, including antiparasitic properties. 10 ■ RESULTS AND DISCUSSION Aqueous phosphate media have previously been used to produce norcoclaurine from dopamine 5 and 4-HPAA via a PSR, and the reaction was extended to a range of aldehyde substrates. 30 Despite this facile approach of using catechols to generate THIQs, an extension to ketone substrates has received little attention, other than the work by Tono et al. who noted that acetone might couple with dopamine in phosphate buffer (0.1 M, pH 7.2), although the product was not characterized. 31 In addition, the extraction of dopamine from the Chinese yam with acetone was believed to form a THIQ and the characterization of plant acetone extracts from Aristolochia arcuata highlighted that a THIQ was formed from dopamine and acetone. 32,33 The main challenge with the PSR reaction using ketones is the relatively low reactivity, which when combined with the sensitivity of catechols can lead to the formation of many side products. However, inspired by the interesting NCS mediated PSR reaction with ketones it was decided to investigate whether an aqueous biomimetic phosphate-based approach could also be developed. 19 First, to avoid side reactions because of the oxidation of catechols, such as dopamine 5, the antioxidant sodium ascorbate was added to the reactions. 20 Cyclohexanone 6 was then used in a reaction with potassium phosphate (KPi, 1 M) with acetonitrile as the cosolvent at pH 6 and 70°C and surprisingly, the PSR product 7 was formed in 11% yield (Table 1, entry 1). Because different phosphate anions predominate at different pHs, at pH 6 the major ion is − , alternative pHs were explored. Unsurprisingly, no product 7 was observed at pH 4, due to dopamine protonation, and pH 12 due to the oxidation of dopamine which is prevalent at high basic pHs. However, at pH 9 the yield increased to 16% (Table 1, entry 4; see Supporting Information Figure S1 for further data on the effect of pH). A pH of 9 was then used as a suitable starting point for further reaction optimization. Density functional theory (DFT) and MP2 methods have been used in a theoretical study exploring the phosphate-mediated PSR with 3-hydroxypehenethylamine and formaldehyde, which suggested that the lowest energy pathway required both HPO 4 2− and H 2 PO 4 − for optimal catalysis in the phosphate biomimetic reaction. 34 At pH 9, HPO 4 2− is the predominant anion present (with some H 2 PO 4 − ) and can facilitate both the deprotonation of the meta-phenolic OH and abstraction of 8a-H in the isoquinolone intermediate.
Typically, a cosolvent was required when developing the phosphate PSR reaction with aldehydes where aldehyde solubility was poor so other cosolvents were investigated, also with a view to enhancing yields. The polar aprotic solvents dimethyl sulfoxide (DMSO), dimethylformamide (DMF), and 1,4-dioxane together with the polar protic solvents methanol, ethanol, and isopropanol were investigated (Table 1 and Supporting Information Figure S2). While DMSO had little effect on the yield, it decreased when using DMF, 1,4-dioxane, and isopropanol. However, the reaction yields were higher with ethanol (Table 1, entry 6) and in particular methanol which was superior to other cosolvents (Table 1, entry 7), giving a 43% yield of 7. Further studies established that lower proportions of methanol (10−40% v/v, data not shown) generated similar yields of 7, and in the absence of a cosolvent, 36% of 7 was formed. This presumably reflects the fact that in the case of 6 there were no solubility issues in water.
To improve the synthetic utility, the reaction temperature was increased to 80 and 90°C with little effect. Upon

The Journal of Organic Chemistry
Article increasing the number of equivalents of 6 from 1.3 to 2, 5, and 10 equiv, in the latter case, 7 was formed in quantitative yield ( Table 1, entry 11), so this was used in further experiments. The effect of using lower KPi concentrations was also investigated using 5 and 6 (Supporting Information, Table  S1). Concentrations of 0.1−1 M KPi gave 7 in 97% yield, and below 0.1 M the yield of 7 decreased. To ensure sufficient KPi was present, a 0.3−0.5 M KPi concentration was therefore selected. The effect of the dopamine 5 concentration in the reaction was also explored (Supporting Information, Table  S2). The yield remained constant with concentrations of up to 0.3 M, decreasing to 48% with 1 M of 5 when solubility issues were encountered. With optimized reaction conditions in hand, we were curious to understand the requirements for phosphate in the reaction. Alternative buffers or bases were used at pH 9 including KOH, KHCO 3 −K 2 CO 3 (0.5 M), Na 3 BO 3 (0.5 M), and Na 2 SO 3 (saturated) together with water alone (Supporting Information, Table S3). When water, water adjusted to pH 9 using KOH, or Na 3 BO 3 solution was used, 7 was formed in trace amounts in reactions with 10 mM of 5 (Supporting Information, Table S3). These increased to 8−15% yields in reactions using 50 mM of 5 (Supporting Information, Table  S4). Interestingly, with KHCO 3 −K 2 CO 3 and Na 2 SO 3 at pH 9 yields of 24 and 43% with 10 mM 5 were noted, increasing to 63 and 89% with 50 mM 5, respectively. At both dopamine concentrations the highest yields were still observed with KPi. The role of phosphate in previous experiments for aldehyde PSRs has been probed using DFT and MP2 methods in a theoretical study. 34 The proposed reaction scheme for 3hydroxyphenethylamine and formaldehyde involved both HPO 4 2− and H 2 PO 4 − having a role in the formation of a complex with the iminium intermediate, and after cyclization, H 2 PO 4 − assisting in the deprotonation at 8a-H of isoquinolone, while also being complexed to the amine proton. The lower yields observed with KHCO 3 −K 2 CO 3 of 24% (10 mM 5) and 63% (50 mM 5) may reflect that the trigonal planar geometry of carbonate is unlikely to be able to perform a similar role of deprotonation and amine proton complexation as readily as H 2 PO 4 − (or potentially HPO 4 2− ) in the rearomatization step, compared to tetrahedral phosphate. The relatively high yields with Na 2 SO 3 were interesting and may be due to its trigonal pyramidal structure, enabling the rearomatization step in a similar fashion to KPi, combined with the reducing environment it provides. Overall, KPi gave rise to the highest yields and so was used in further experiments. It was also noted that in the reaction between 5 and 6 for the KPi concentrations used, no ortho-regioisomer was detected presumably because of steric reasons. 35 With suitable reaction conditions established, the general applicability of the reaction was explored for the synthesis of a range of 1,1′-disubstituted and spiro-THIQ alkaloids (Scheme 2). Reactions were monitored by 1 H NMR spectroscopy because of the reported challenges with product isolation. 29 In addition, to confirm the NMR data, for several reactions, the yields were confirmed by high-performance liquid chromatography (HPLC) analysis. The phosphate mediated PSR could tolerate a range of methyl ketones, giving 8−12 in 54−95% Scheme 2. Application of the Phosphate-Catalyzed PSR to Synthesize 1,1′-Disubstituted and Spiro-THIQ Alkaloids a a Typical reaction conditions: dopamine 5 (15−25 mM), ketone (20−150 mM), and sodium ascorbate (1.0 equiv relative to dopamine or the corresponding phenethylamine) were reacted together on a 1 mL scale in 0.3 M KPi, pH 9 and methanol (50% v/v of methanol and ketone combined) at 70°C. Small scale reactions were performed in duplicate or triplicate. Yields were determined by 1 H NMR spectroscopy with an internal standard (maleic acid), and also by HPLC analysis (see Supporting Information, Table S5 for HPLC retention times) for several examples to confirm the data. b 10 equiv of the corresponding ketones were used. c 50 equiv of the corresponding ketones were used. All products were isolated (yields are in brackets) either from these small scale reactions or larger reactions and purified, using the acid−base extraction method or preparative HPLC, for characterization purposes.

The Journal of Organic Chemistry
Article yield when using aliphatic ketones. The reaction yields were greater for 8 and 9, reflecting the influence of the increasing alkyl chain length and associated steric effects in the intermediates and subsequent cyclizations to give products 10−12. For the aromatic ketones acetophenone or phenyl acetone giving 13 and 14, respectively, again poor reactivities were observed, reflecting unfavorable steric interactions in the cyclization step to generate the THIQs. The introduction of a trifluoromethyl group, however, was readily achieved giving 15 in 93% yield.
The introduction of a third cyclic or heterocyclic ring could more widely be introduced into the THIQ scaffold, following the model reaction to give 7. Using 3-or 4-methyl cyclohexanone readily gave 16 and 17 in 70 and 80% yields, respectively. Similarly, with 4-isopropyl cyclohexanone, 18 was generated in 75% yield and using tetrahydro-4H-pyran-4-one 19 was formed in 83% yield. When using cyclobutanone and cyclopentanone THIAs 20 and 21 were formed in 97 and 95% yields, comparable yields to when using 6, highlighting that other ring sizes can be readily used in the phosphate-catalyzed Pictet−Spengler condensation. Furthermore, the use of 1-or 2indanone afforded the tetracyclic ochotensine 9 derivatives 22 and 23 in 70 and 21% yields, respectively.
In addition to using alternative ketone substrates, selected substituted phenethylamines were investigated (Scheme 2). 2-(3-Hydroxyphenyl)ethylamine 20 was used with 6 to synthesize 24, a compound previously synthesized in 2 steps in 10% yield, and reported to possess antimalarial properties. 10 A yield of 74% was achieved and the product was isolated by preparative HPLC in 53% yield. The fluorinated analogue 4-fluoro-3hydroxy phenethylamine was also used to give 25 in 55% yield. A regioisomer of dopamine, 5-(2-aminoethyl)benzene-1,3diol, 36 which will provide more unfavorable steric interactions was investigated and gave 26 in a reasonable 57% yield.
To highlight the utility of this methodology, the synthesis of 7 and 20 as selected compounds were then performed on a 1 g scale. The reaction yields by 1 H NMR spectroscopy were 97%, and at this larger scale it was possible to isolate the materials using an acid−base extraction method to give 7 in 40% yield and 20 in 51% yield.
The use of a ketone possessing ester functionality for subsequent ring cyclization was also investigated to establish a reaction cascade to form an additional ring, as this strategy with an aldehyde had been successfully used to generate the antibacterial trolline. 20 Using the ketone KPi reaction conditions established, 5 was reacted with commercially available methyl levulinate 27 to form a linear intermediate 28 with a C-1 quaternary carbon. Spontaneous cyclization by nucleophilic attack of the amine at the ester carbonyl, following the favored 5-exo-trig mode, afforded lactam 29 in 50% yield (by 1 H NMR spectroscopy) (Scheme 3).

■ CONCLUSIONS
In summary, a biomimetic approach for the synthesis of 1,1′disubstituted and spiro-THIQs has been established using a PSR with unactivated ketones. The reaction is readily performed in methanol/phosphate buffer at pH 9 and 70°C ; these are significantly milder reaction conditions than those previously reported. The reaction successfully used methyl ketones, cyclic ketones, and aromatic ketones with phenethylamines bearing a meta-hydroxyl group. Overall, this study provides a low cost and sustainable way of making C-1,1′ disubstituted THIQ alkaloids using phosphate buffer, to afford a range of products in high yields using a facile procedure.
■ EXPERIMENTAL SECTION General Information. All chemicals were obtained from commercial suppliers and used as received. Small-scale reactions were heated using a BIOER Mixing block MB-102, and the scale-up reactions were heated using a Heidolph MR Hei-Tec type heating mantle. Thin layer chromatography was carried out using Merck TLC silica gel 60 F 254 plates, and the products were visualized using combinations of UV light (254 nm), potassium permanganate, and phosphomolybdic acid staining solutions. Column chromatography was carried out using silica gel (particle size 40−60 μm). Infrared (IR) spectra were recorded using a PerkinElmer Spectrum 100 FT-IR spectrometer or a Bruker ALPHA PLATINUM-ATR, operating in the ATR mode. 1 H NMR spectra were recorded on Bruker AVANCE 400, 500, 600, and 700 MHz spectrometers at 25°C, using the residual protic solvent stated as the internal standard. Chemical shifts are quoted in ppm to the nearest 0.01 ppm using the following abbreviations: s (singlet), d (doublet), t (triplet), q (quartet), qn (quintet), sext (sextet), dd (doublet of doublets), dt (doublet of triplets), and m (multiplet) defined as all multipeak signals, where the overlap or complex coupling of signals make definitive descriptions of peaks difficult. The coupling constants are defined as J and measured in Hz. 13 C{ 1 H} NMR spectra were recorded at 100, 125, 150, or 175 MHz on Bruker AVANCE 400, 500, 600, and 700 MHz spectrometers at 25°C using the stated solvent as the standard. Chemical shifts are reported to the nearest 0.1 ppm. Melting points were measured with a Gallenkamp apparatus, and were uncorrected.
Mass spectra were obtained using a Waters Acquity UPLC SQD (using a linear gradient of 5−95% of acetonitrile over 5 min, with a C8 column, and flow rate of 0.6 mL/min) and the Waters LCT Premier XE ESI Q-TOF mass spectrometer in the Department of Chemistry, UCL.
Determination of Yields Using 1 H NMR Spectroscopy. The reactions were set up in duplicate (1 mL scale) in Eppendorf tubes which were shaken at 70°C for 20 h. The reaction mixture was transferred to a round-bottomed flask and evaporated to remove methanol. Water was then removed using a freeze-drier to obtain a solid, which was dissolved in a solution of maleic acid (CD 3 OD/D 2 O = 1:3, 1 mL) and characterized by 1 H NMR spectroscopy. The concentration of the maleic acid was half of the starting material dopamine. The peak integration of maleic acid proton (δ = 5.89 ppm) was calibrated as 1.00, and the integration (×100%) of one of the product aromatic protons was used to determine the NMR yield.
Preparative HPLC Methods. Preparative HPLC was carried out on a Dionex 580 HPLC system, including a P580 Pump, ASI-100 Automated Sample Injector, and PDA-100 Photodiode Array Detector. A C 18 reverse phase column, Agilent ZORBAX 300SB-C18 (250 mm × 9.4 mm, 5 μm) was used for purifications. The detection wavelength was 280 nm and the flow rate was 2 mL/min. Chromeleon Client Program software was used to monitor the HPLC traces.