Synthesis of (+)-(R)-Tiruchanduramine

The absolute stereochemistry of the marine alkaloid (+)-(R)-tiruchanduramine was established via a convergent total synthesis in six steps and 15.5% overall yield from Fmoc-D-Dab(Boc)-OH.


Introduction
We have previously reported [1] a racemic synthesis of the alkaloid tiruchanduramine 1 (Figure 1) isolated by Ravinder et al. [2] from the ascidian Synoicum macroglossum and shown to be a α-glucosidase inhibitor. Glucosidases have a number of roles in the body, including digestion, glycoprotein processing and degradation, and have differing locations and substrate specificities [3]. They are targets for treatments to viral infections, diabetes, lysosomal storage disorders, and cancer [4]. Glucosidases involved in digestion have attracted particular interest as they release glucose and inhibiting them slows rises in blood sugar following food consumption and medications such as Acarbose [5] and Miglitol [6] have been developed.
We wished to determine the absolute configuration of 1 by synthesis from a chiral pool precursor and were initially attracted by the synthesis reported by Ravinder et al. [2]. This work reported the coupling of β-carboline-3-carboxylic acid 2 [7], easily prepared in four steps from tryptophan, with the amine 3, prepared in five steps from but-3-en-1-ol. Amide 4 was converted into the hydrochloride salt of 1 via a four-step protocol involving the introduction of the guanidine under Mitsunobu conditions (Scheme 1). Our racemic synthesis [1] differed from this as it was a convergent approach in which the guanidine heterocycle 5 is prepared separately then coupled to the β-carboline-3-carboxylic acid 2 (Scheme 1). and shown to be a α-glucosidase inhibitor. Glucosidases have a number of roles in the body, including digestion, glycoprotein processing and degradation, and have differing locations and substrate specificities [3]. They are targets for treatments to viral infections, diabetes, lysosomal storage disorders, and cancer [4]. Glucosidases involved in digestion have attracted particular interest as they release glucose and inhibiting them slows rises in blood sugar following food consumption and medications such as Acarbose [5] and Miglitol [6] have been developed. We wished to determine the absolute configuration of 1 by synthesis from a chiral pool precursor and were initially attracted by the synthesis reported by Ravinder et al. [2]. This work reported the coupling of β-carboline-3-carboxylic acid 2 [7], easily prepared in four steps from tryptophan, with the amine 3, prepared in five steps from but-3-en-1-ol.

Results and Discussion
We thus attempted to repeat the work reported by Ravinder [2] by preparing the known [8] chiral amine (S)-3 from commercially available (S)-malic acid 6 using literature procedures (esterification [8] (MeOH, H + ), reduction [8] (H3B·SMe2, NaBH4, THF), protection [8] (2-methoxypropene, H + ), reduction [8] (LiAlH4, Et2O), azide formation [9] (TPP, DIAD, DPPA), and reduction [10] (TPP, H2O), see supplementary materials for experimental details). Coupling of the amine (S)-3 with carboxylic acid 2 was attempted using DCC and DMAP as reported by Ravinder [2]. Whilst successful, the reaction was hampered by difficulties associated with the removal of the DCCU byproduct from the desired product 7. To circumvent this, we repeated the reaction by activating the acid 2 with CDI followed by the addition of the amine (S)-3. After 24 h, work up and chromatography gave 7 in a 70% yield and the structure was confirmed by NMR and MS analysis. Following this, the β-carboline NH was protected using Boc-anhydride and trimethylamine, to secure 8 in a 56% yield, again, its structure was confirmed by NMR and MS analysis. Finally, deprotection of the acetonide using PPTSA in methanol gave 9 in 71% yield. NMR data had been reported [2] for this compound and comparison of this gave a close agreement despite several signals being missing in the original literature report. MS data was as expected for the desired compound. With diol 9 in hand, we attempted the formation of the cyclic guanidine 10 using the conditions suggested in the original racemic synthesis. [2] Unfortunately, in our hands we were unable to effect any reaction between 9 and N,N′,N′′-tri-Boc-guanidine 11 using DEAD and triphenylphosphine. We varied the concentration of the reaction and the number of equivalents of each reagent and used the alternate reagent DIAD as well as differing solvents. In all cases only the unreacted diol 9 was recovered after chromatography (Scheme 2).
To circumvent this, we repeated the reaction by activating the acid 2 with CDI followed by the addition of the amine (S)-3. After 24 h, work up and chromatography gave 7 in a 70% yield and the structure was confirmed by NMR and MS analysis. Following this, the β-carboline NH was protected using Boc-anhydride and trimethylamine, to secure 8 in a 56% yield, again, its structure was confirmed by NMR and MS analysis. Finally, deprotection of the acetonide using PPTSA in methanol gave 9 in 71% yield. NMR data had been reported [2] for this compound and comparison of this gave a close agreement despite several signals being missing in the original literature report. MS data was as expected for the desired compound. With diol 9 in hand, we attempted the formation of the cyclic guanidine 10 using the conditions suggested in the original racemic synthesis. [2] Unfortunately, in our hands we were unable to effect any reaction between 9 and N,N ,N -tri-Boc-guanidine 11 using DEAD and triphenylphosphine. We varied the concentration of the reaction and the number of equivalents of each reagent and used the alternate reagent DIAD as well as differing solvents. In all cases only the unreacted diol 9 was recovered after chromatography (Scheme 2). Frustrated by this synthetic approach, we returned to our original convergent strategy and sought a synthetic route to enantiopure guanidine 5. Our synthesis began with the commercially available orthogonally protected diaminocarboxylic acid Fmoc-D-Dab(Boc)-OH (12). Reduction of the mixed anhydride of 12 with sodium borohydride gave alcohol 13 in an 86% yield [11]. Deprotection of 13 using piperidine in dichloromethane gave the amine 14 which was guanidinylated [1] with 15 to give the alcohol 16 in 94% yield. Cyclization of this with a mixture of I2/dppe/PPh3/imidazole [1] in dichloromethane to give the cyclic guanidine 17 which was used in the next steps without purification as it was found to undergo decomposition on silica gel. Crude 17 was deprotected using aqueous HCl (3M) and after freeze drying gave the desired guanidine (R)-5. This compound gave identical data to the previously reported racemic compound [1] and the synthetic methods adopted have preserved the original stereogenic center found in compound 12. The preparation of (R)-1 was achieved by treatment of the carboxylic acid 2 with an excess of 1,1′-carbonyldiimidazole (CDI) in DMF for 30 min followed addition of the free base of (S)-5.
[1] After 24 h the mixture was freeze dried and purified by repeated column chromatography to give (R)-1.HCl in 18% yield over three steps. Synthetic 1.HCl gave identical NMR data to that previously reported [1] for racemic 1.HCl, however it gave an [α]D 24 of +27 (MeOH, c = 1.0,) which is in close agreement to the value for natural 1 which was reported [2] as [α]D +31 (c 0.5, MeOH) with an unspecified counterion (Scheme 3). Frustrated by this synthetic approach, we returned to our original convergent strategy and sought a synthetic route to enantiopure guanidine 5. Our synthesis began with the commercially available orthogonally protected diaminocarboxylic acid Fmoc-D-Dab(Boc)-OH (12). Reduction of the mixed anhydride of 12 with sodium borohydride gave alcohol 13 in an 86% yield [11]. Deprotection of 13 using piperidine in dichloromethane gave the amine 14 which was guanidinylated [1] with 15 to give the alcohol 16 in 94% yield. Cyclization of this with a mixture of I 2 /dppe/PPh 3 /imidazole [1] in dichloromethane to give the cyclic guanidine 17 which was used in the next steps without purification as it was found to undergo decomposition on silica gel. Crude 17 was deprotected using aqueous HCl (3M) and after freeze drying gave the desired guanidine (R)-5. This compound gave identical data to the previously reported racemic compound [1] and the synthetic methods adopted have preserved the original stereogenic center found in compound 12. The preparation of (R)-1 was achieved by treatment of the carboxylic acid 2 with an excess of 1,1 -carbonyldiimidazole (CDI) in DMF for 30 min followed addition of the free base of (S)-5.
[1] After 24 h the mixture was freeze dried and purified by repeated column chromatography to give (R)-1·HCl in 18% yield over three steps. Synthetic 1·HCl gave identical NMR data to that previously reported [1] for racemic 1·HCl, however it gave an [α] D 24 of +27 (MeOH, c = 1.0,) which is in close agreement to the value for natural 1 which was reported [2] as [α] D +31 (c 0.5, MeOH) with an unspecified counterion (Scheme 3). Ravinder 2 reported that tiruchanduramine 1 was a potent inhibitor of yeast α-glucosidase (IC 50 78.2 µg/mL, 0.24 mM) as compared with Acarbose (IC 50 100 µg/mL, 0.16 mM) as the standard. We have previously reported that racemic tiruchanduramine 1 is a weak inhibitor of yeast α-glucosidase and now with a full synthesis of the natural product (+)-(R)tiruchanduramine 1, we confirm that the compound is not a potent inhibitor of either yeast α-(IC 50 0.7 mM) or almond β-glucosidase (IC 50 1.0 mM). Our assays show the same order of magnitude inhibition to the reported figures which correlates with Ravinder's report, but not with the conclusion that the compound is a potent inhibitor for which much lower µM IC 50 values would be expected. If there is a greater inhibition seen by Ravinder then we propose that, as with many natural products, the original natural product may have had a minor contaminant which produced the greater apparent inhibition. The comparison with acarbose by Ravinder may not have been appropriate since Acarbose does not potently inhibit yeast α-glucosidase [12]. Ravinder does not report which specific α-glucosidase they investigated but cite the method of Kim et al. who used a yeast α-glucosidase [13]. It is noteworthy that the synthetic analogues of tiruchanduramine 1 we prepared [14] gave considerably better IC 50 values against yeast α-glucosidase, for example compounds 18a (n = 4) and 18b (n = 5) IC 50 values of 12 µM and 7.5 µM respectively. (Figure 2) We can thus propose that although the natural product is not a very good inhibitor of yeast α-glucosidase, it represents a useful starting point for the design of much more potent inhibitors.
compound gave identical data to the previously reported racemic compound [1] and the synthetic methods adopted have preserved the original stereogenic center found in compound 12. The preparation of (R)-1 was achieved by treatment of the carboxylic acid 2 with an excess of 1,1′-carbonyldiimidazole (CDI) in DMF for 30 min followed addition of the free base of (S)-5.
[1] After 24 h the mixture was freeze dried and purified by repeated column chromatography to give (R)-1.HCl in 18% yield over three steps. Synthetic 1.HCl gave identical NMR data to that previously reported [1] for racemic 1.HCl, however it gave an [α]D 24 of +27 (MeOH, c = 1.0,) which is in close agreement to the value for natural 1 which was reported [2] as [α]D +31 (c 0.5, MeOH) with an unspecified counterion (Scheme 3). Ravinder 2 reported that tiruchanduramine 1 was a potent inhibitor of yeast α-glucosidase (IC50 78.2 μg/mL, 0.24 mM) as compared with Acarbose (IC50 100 μg/mL, 0.16 mM) as the standard. We have previously reported that racemic tiruchanduramine 1 is a weak inhibitor of yeast α-glucosidase and now with a full synthesis of the natural product (+)-(R)-tiruchanduramine 1, we confirm that the compound is not a potent inhibitor of either yeast α-(IC50 0.7 mM) or almond β-glucosidase (IC50 1.0 mM). Our assays show the same order of magnitude inhibition to the reported figures which correlates with Ravinder's report, but not with the conclusion that the compound is a potent inhibitor for which much lower µ M IC50 values would be expected. If there is a greater inhibition seen by Ravinder then we propose that, as with many natural products, the original natural product may have had a minor contaminant which produced the greater apparent inhibition. The comparison with acarbose by Ravinder may not have been appropriate since Acarbose does not potently inhibit yeast α-glucosidase [12]. Ravinder does not report which specific α-glucosidase they investigated but cite the method of Kim et al. who used a yeast α-glucosidase [13]. It is noteworthy that the synthetic analogues of tiruchanduramine 1 we prepared [14] gave considerably better IC50 values against yeast α-glucosidase, for example compounds 18a (n = 4) and 18b (n = 5) IC50 values of 12 µ M and 7.5 µ M respectively. (Figure 2) We can thus propose that although the natural product is not a very good inhibitor of yeast α-glucosidase, it represents a useful starting point for the design of much more potent inhibitors.

Conclusions
We have succeeded in the first enantioselective synthesis of (+)-(R)-tiruchanduramine 1 as its hydrochloride and formate salts in 15.5% overall yield (6 steps) from a commercially available starting material 12 and have established its absolute stereochemistry as R. We also attempted to repeat reactions reported in the original synthesis which proved capricious in our hands.

Experimental
Column chromatography was carried out on silica gel (60Å , 40-63 μm) and TLCs were conducted on precoated Kieselgel 60 F254 (Art. 5554) (Merck Life Science UK Limited, Watford, UK) with the eluent specified in each case. All non-aqueous reactions were conducted in oven-dried apparatus under a static atmosphere of argon. Diethyl ether, THF, and dichloromethane were dried by a Pure Solv MD-3 solvent purification system (One Industrial Way, Amesbury, MA, USA). Dry methanol and DMF and all starting materials were purchased from Aldrich (Merck Life Science UK Limited, Gillingham, UK ). Chemical shifts are reported in δ values relative to residual chloroform (7.26/77.16 ppm),

Conclusions
We have succeeded in the first enantioselective synthesis of (+)-(R)-tiruchanduramine 1 as its hydrochloride and formate salts in 15.5% overall yield (6 steps) from a commercially available starting material 12 and have established its absolute stereochemistry as R. We also attempted to repeat reactions reported in the original synthesis which proved capricious in our hands.

Experimental
Column chromatography was carried out on silica gel (60Å, 40-63 µm) and TLCs were conducted on precoated Kieselgel 60 F254 (Art. 5554) (Merck Life Science UK Limited, Watford, UK) with the eluent specified in each case. All non-aqueous reactions were conducted in oven-dried apparatus under a static atmosphere of argon. Diethyl ether, THF, and dichloromethane were dried by a Pure Solv MD-3 solvent purification system (One Industrial Way, Amesbury, MA, USA). Dry methanol and DMF and all starting materials were purchased from Aldrich (Merck Life Science UK Limited, Gillingham, UK). Chemical shifts are reported in δ values relative to residual chloroform (7.26/77.16 ppm), methanol (3.31/49.0 ppm), and DMSO (2.50/39.52 ppm) as internal standards. Proton and carbon NMR spectra were recorded in CDCl 3 on a Bruker AC400 spectrometer (Coventry, UK) unless otherwise stated and spectra are presented in the Supplementary Materials for each compound. Mass spectra data were obtained at the EPSRC Mass Spectrometry Service Centre at the University of Wales, Swansea. Infrared spectra were recorded as thin films (oils) on a Bruker Tensor 27 series instrument (Coventry, UK). Optical rotations were performed on a Bellingham and Stanley Ltd. ADP400 TM polarimeter (Bellingham+Stanley Ltd, Kent, UK). Melting points were performed on a Stuart SMP10 apparatus (Cambridgeshire, UK) and are uncorrected.

Tert-Butyl (R)-(4-hydroxy-3-(bis-tert-butyl-guanidine)butyl)carbamate 16
Alcohol 13 (1.74 g, 4.08 mmol, 1.0 equiv.) was dissolved in 20% piperidine in chloroform (v/v, 40 mL) and stirred. After 2 h TLC (100% EA) indicated the reaction was complete and the solvent was evaporated under high vacuum. The resulting white solid was triturated with hexane (2 × 50 mL) to remove Fmoc by-products. After drying, the resulting crude amine 14 and compound 15 (1.90 g, 6.12 mmol, 1.5 equiv.) were dissolved in dry methanol (20 mL) and NEt 3 (1.7 mL, 1.24 g, 12.2 mmol, 3.0 equiv.) was added. After stirring for 48 h, the solvent was removed under reduced pressure and the crude product purified by column chromatography (30-40% EA in PE) to give 16 (1.72 g, 3  Guanidine 16 (740 mg, 1.66 mmol, 1.0 equiv.) was dissolved in dichloromethane (15 mL) and cooled (−40 • C) whereupon dppe (990 mg, 2.49 mmol, 1.5 equiv.) was added in one portion and the mixture stirred until complete dissolution was achieved. Imidazole (395 mg, 5.80 mmol, 3.5 equiv.) and iodine (631 mg, 2.50 mmol, 1.5 equiv.) were then added sequentially and the resultant mixture stirred for 3 h with cooling (−40 to −20 • C). Chloroform (50 mL) was then added, followed by ammonium chloride solution (aq., sat., 100 mL). After warming to rt, the organic layer was separated, and the aqueous layer extracted with chloroform (2 × 20 mL). The combined extracts were washed with water (50 mL) and brine (50 mL) then dried (MgSO 4 ), filtered and concentrated under reduced pressure. The resulting residue was dissolved in a minimum volume of dichloromethane (4-5 mL) and then diethyl ether was added to precipitate any DPPE by-products. The supernatant liquid was removed by filtration and after evaporation this process of precipitation was repeated twice. This gave 17 (555 mg, contaminated with dppe byproducts) as a sticky white gum which was used in the next step without further purification.  Hydrochloric acid solution (5.0 mL, aq, 3M) was added to a cooled (0 • C) flask containing guanidine 17 (488.0 mg) and the mixture stirred to rt over 48 h. The mixture was passed through a short Celite© pad in a pipette to remove dppe derived impurities and the pad washed with water (ca. 2-3 mL). The filtrate was freeze dried to give crude (R)-5 (259 mg) as a hygroscopic off-white solid which was used in next step without any further purification.