The synthesis of 16-dehydropregnenolone acetate (DPA) from potato glycoalkaloids

The use of solanidine as starting material for the synthesis of steroid hormones was strongly stimulated by the possibility to isolate large amounts (ton scale) of potato glycoalkaloids from a waste stream of the potato starch production. A procedure is available to isolate these glycoalkaloids from the potato protein fraction and after hydrolysis solanidine is set free and can be made available as alternative for diosgenine as starting material for the production of dehydropregnenolon acetate (DPA). The conversion of solanidine to DPA was first tried by reinvestigation of several known methods like oxidation with Hg(OAc) 2 , the Cope reaction and the Polonovski reaction but none of these approaches were successful. The best option was to open the E,F-ring system using the Von Braun reaction. Besides the desired major E-ring opened compound also the minor F-ring opened compound was isolated. Alternatives for the hazardous Von Braun reagent BrCN were investigated but not found. Further degradation using the Hofmann reaction was successful in a ∆ 16 derivative, which led to the desired triene intermediate. Finally DPA could be obtained starting from solanidine in 9 reaction steps in 30% overall yield using the known conversion to spirosolane compounds followed by conversion to DPA. Extensive research


Introduction
Since the first structure elucidation of solanidine (1) about 70 years ago, research has been carried out to convert this aglycon to an intermediate for the synthesis of steroids.Renewed interest in this conversion was stimulated by the possibility to isolate large amounts (ton scale) of potato glycoalkaloids from a waste stream of the potato starch production.During the starch refining process the proteins together with the glycoalkaloids are separated from the starch.This protein fraction is then subjected to a number of refining steps in which the proteins are separated from the glycoalkaloids.The glycoalkaloids together with free amino acids, peptides and minerals end up in the so-called protamylasse fraction.The amount of glycoalkaloids present in this fraction varies from 200-2000 ppm dependent on the potatoes processed per campaign.Potato glycoalkaloids consist for more than 95% of α-chaconine and α-solanine, which both have the steroid-like solanidine (1) as aglycon, thus a large potential of starting material may become available for conversion to steroid hormones.
During the last 50 years diosgenine is used as the main starting material in the industrial synthesis of progestagens, androgens, estrogens, norsteroids, and a diuretic spironolactone.Uncertain external factors have often influenced the guaranteed supply of diosgenine and in the course of time many alternatives have been investigated.These alternatives should have in common that they can be implemented in existing production facilities for economic and pharmaceutical reasons.Solanidine (1) could be such an alternative on condition that it can be converted in an industrially attractive way to DPA (6), which is a key intermediate in the industrial syntheses of progesterone and cortisone derivatives.We here like to report on our results in this field.

Isolation and hydrolysis of potato glycoalkaloids
To obtain the starting material for our research, an improved simple and effective method has been developed for the isolation of the potato glycoalkaloids, α-chaconine and α-solanine, from the spray dried protamylasse fraction, which was obtained from AVEBE 1 .When this crude fraction was directly subjected to hydrolysis a complex product mixture was formed.Therefore a modified method of Friedman et al. 2 was used for the extraction of the potato glycoalkaloids from the protamylasse fraction.A crude mixture of α-chaconine and α-solanine was obtained by using aqueous ethanol for this extraction, and recrystallization from ethanol gave a clean mixture of both glycoalkaloids.In this way 221.1 g of spray-dried protamylasse yielded 10.8 g of glycoalkaloids, which is 85% yield based on a 5.7% glycoalkaloid content in the spray-dried protamylasse as determined by HPLC.Chemical hydrolysis was performed with acid and solanidine (1) was isolated from the acidic ethanol in an almost quantitative yield after precipitation with NH 4 OH.
Because both the isolation and hydrolysis of the glycoalkaloids can be performed on a large scale, attempts were made to combine both procedures.Since the protamylasse fraction has a pH of 5-6 it could be dissolved in water without addition of acid.After the aqueous solution was made basic with NH 4 OH and stored overnight at 4ºC, the slurry was centrifuged and the supernatant discarded.The pellet was again directly subjected to hydrolysis but also in this case a complex product mixture was formed, and extraction with ethanol proved to be necessary prior to hydrolysis.Thus the pellets were transferred to a Soxhlett apparatus and extracted with ethanol for 24 hours.After concentration of the extract, a crude glycoalkaloid fraction was obtained.This fraction was hydrolyzed in acidic ethanol for 2.5 hours, and the alkaloids were precipitated with NH 4 OH.The crude precipitate was filtered and recrystalized from ethanol to give pure solanidine (1).The hydrolysis of the glycoalkaloids is accompanied by the formation of solanidiene in 9% yield, but lowering the amount of acid in the hydrolysis step from 2M to 1M HCl in ethanol reduced the formation of solanidiene to a negligible amount.

The conversion of solanidine to DPA
The conversion of solanidine (1) to DPA (6) was first tried by the recently published method using Hg(OAc) 2 as oxidation reagent. 3Electrochemical 4,5 oxidations of solanidanes have shown that the ∆ 22(N) -iminium salt is exclusively formed in acetone, while the ∆ 16(N) -iminium salt is the sole product in CH 2 Cl 2 , in both cases pyridine is added as base.Chemical 6,7 oxidation of solanidine with Hg(OAc) 2 in acetone or CH 2 Cl 2 showed the same preference in formation of the ∆ 16(N) -iminium and ∆ 22(N) -iminium salts.In 1997 Gaši et al. 3 reported the conversion of solanidine (1) to DPA (6) via iminium ion 2 and the enamine intermediates 3 and 4 (Scheme 1).Enamine 4 is oxidized with NaIO 4 , NaI, and NaHCO 3 in a mixture of water and t-BuOH to ketolactam 5, and elimination of the lactam moiety in 5 gives DPA (6) in an overall yield of 28%.Before reinvestigation of the degradation of solanidine (1), it was converted to its acetate prior to oxidation with Hg(OAc) 2 .Because the Hg(OAc) 2 contained some acetic acid 7 , enamine 4 was directly formed in 90% yield, which made the separate isomerization step superfluous.However, oxidation of enamine 4 according to the procedure of Gaši et al. 3 was unsuccessful in our hands and the starting material was recovered almost quantitatively.Many other oxidation reagents were tried (ozone, 8 19 ), but in all cases no oxidation product could be obtained.To rule out a possible involvement of the ∆ 5,6 double bond during the oxidation, 20 1 was transformed into solanidan-4-en-3-one ( 7) by treatment of 1 with Al(i-PrO) 3 in toluene in the presence of cyclohexanone (Scheme 2).Oxidation of enone (7) with Hg(OAc) 2 again showed the exclusive formation of the corresponding ∆ 20,22 -enamine, which was subjected again to a range of oxidation reactions, 3,8,10- 16,18,19,21 but all failed in our hands.
According to Gaši and co-workers 3 the modest yield of the oxidation of enamine 4 was due to its instability.However, Mopac 22 PM3 calculations 23,24 showed that enamine 4 (∆H f = -76.57kcal) is more stable than enamine 3 (∆H f = -68.86kcal).The energy difference is large enough to make isomerization during the oxidation reaction unlikely and the difficulties in the oxidation reaction can not only be imputed to the instability of enamine 4.These calculations also show that enamine 4 is not really an enamine but more an isolated double bond and a separate amino group.The bond order of the C-N bond is 1.03, which indicates that there is nearly a single bond between C22 and N. The ∆ 20,22 double bond forces the five membered E-ring to be completely flat, and as a consequence the D-ring is bent in such a way that C18 and the six membered ring shield its top and bottom side, respectively (Figure 1).The methyl group (C21) lies in the plane of the ∆ 20,22 double bond making it even more difficult to approach.In our opinion, steric hindrance is the main reason for the low reactivity of this ∆ 20,22 double bond.The Cope and Polonovski reactions In the second attempt to convert solanidine (1) to DPA (6) the Cope and Polonovski reactions were studied and to do so it was necessary to synthesize a solanidine N-oxide first.To avoid epoxidation problems solanidine (1) was converted to solanidi-4-en-3-one (7), which could be oxidized with MMPP to the corresponding solanidi-4-en-3-one N-oxide (8) in good yield (Scheme 2).It turned out that the Cope reaction did not give the desired results because a 5membered planar transition state, necessary for the Cope reaction, is not possible in this molecule.In such cases deoxygenation, being a competitive process, takes over and indeed, solanidi-4-en-3-one (7) was recovered in all attempts.
The Polonovski reaction can be carried out with Ac 2 O or (CF 3 CO) 2 O under rather extreme conditions.With Ac 2 O no reaction was observed for solanidine N-oxide (8) or 3acetoxysolanidine N-oxide (9).Treatment of 9 with (CF 3 CO) 2 O yielded compound 10 as the result of elimination, isomerization of iminium ion 2 and trifluoroacylation (Scheme 3).This is a new result, but no further attempts have been undertaken to convert compound 10 in DPA (6).

The Von Braun reaction
The third option was to open the E,F-ring system using the Von Braun reaction, which was performed on 3-acetoxysolanidine (11) according to the method of Beisler and Sato. 25,26Next to a 68% yield of 12, the F-ring opened side-product 13 was isolated in 10% yield, which must be the result of bromide attack at C26 (Scheme 4).Although not described in the literature, it is most likely that this product is always formed in the Von Braun reaction but it has never been mentioned before.
Because the use of BrCN in industry requires special safety measures, less toxic alternatives such as acetyl chloride, 27,28 ethyl chloroformate, [29][30][31][32][33][34][35] trichloroethyl chloroformate, 36 benzoylchloride, 37 and benzyl chloride 38 have been investigated.Although these reagents give good results with common tertiary amines, no reaction was observed with 3-acetoxysolanidine (11).Other attempts with Ac 2 O and (CF 3 CO) 2 O, 39 trichlorotriazine, chlorodimethoyxytriazine 40- 42 and TMSCl, NaI, Ac 2 O 43 were also unsuccessful.The fact that ringopening can only be achieved with BrCN can be explained by the unique electronic and steric properties of this reagent.Besides, a striking difference between an N-atom bearing a nitrile group, and an acylated N-atom is the formal positive charge of the N-atom as shown by MOPAC PM3 calculations.The N-atom of the intermediate N-nitrilium ion possesses a formal charge of +0.75 while the N-atom of the corresponding N-acylium ion has a charge of only +0.38.This makes the neighboring C-atoms in the N-nitrilium ion much more susceptible to nucleophilic attack than in the N-acylium ions.
Despite all attempts, BrCN remains the only reagent until now, capable to open the indolizidine ring system of 3-acetoxysolanidine (11).Although the objectives against large-scale industrial application of BrCN remain, the reaction itself gives a good yield of 12, which has possibilities for further transformation to DPA (6).

The Hofmann degradation
One of the possibilities for further transformation of the ringopened product 12 is the Hofmann degradation.To avoid disturbing side reactions under the basic reaction conditions, the bromide was removed, replaced by an acetate or eliminated.Treatment of 12 with Bu 3 SnH and AIBN in benzene at reflux temperature, gave 14 in 94% yield.Treatment of 12 with KOAc in DMF at 90ºC gave the acetate 15 in 83% yield 26 and introduction of the ∆ 16,17 double bond was achieved by treatment of 12 with s-collidine to afford 16 in 98% yield. 7Direct methylation of these nitriles was unsuccessful as expected, so the nitrile was removed by reduction with Red-Al to give the corresponding secondary amines.Methylation with MeI and Na 2 CO 3 in water proceeded smoothly and gave the ammonium salts in high yields, which were subjected to treatment with base without further purification.This resulted in formation of the demethylated products 17 and 18 in 76% and 32% yield, respectively.Treatment of the unsaturated ammonium salt 20 with KOH, 44 t-BuOK, 45,46 NaOH, 47,48 NaOMe, 49 or Et 3 N, 50 gave products which immediately decomposed during the isolation process.Only when 20 was treated with LDA the desired product 21 together with the N-monomethylated product 19 could be isolated in 32% and 26% yield, respectively (Scheme 5).The formation of demethylated products can be explained by difficulties in the proton abstraction, which is necessary for ringopening.In these cases the competitive nucleophilic substitution resulting in demethylation is strongly favored over the Hofmann degradation.LDA is a small and strong enough base to abstract the proton from C20, but only in the case of 20 compound 21 was formed.The presence of the ∆ 16,17 double bond makes H20 to an allylic proton, which is more prone to abstraction.The further degradation of 21 to DPA (6) requires several protection and deprotection steps as shown by Maitra and Breslow, 51 which are not very attractive for industrial application.Besides the use of several hazardous reagents is another reason to look for alternatives.A better industrially applicable alternative may be found in the conversion of solanidine (1) to spirosolanes.In 1971 Schramm and Riedl 26 already mentioned that the degradation of solanidine (1) to DPA (6) can be accomplished via the tomatidine series but up to now the complete procedure has never been published.Therefore 3-acetoxysolanidine (11) was first submitted to the Von Braun reaction, which gave 12 in 68% yield 25 (Scheme 6).Substitution of the bromide at C16 with KOAc yielded acetate 15 in 83%. 26Subsequent reduction of 15 with Red-Al then gave 22 in 93% yield 52 .4][55][56][57][58] Acetylation of tomatidenol with Ac 2 O in pyridine gave 25 in 98% yield.Treatment of 25 with HOAc at reflux temperature gave compound 26 in 85% yield.Subsequent oxidation with CrO 3 in HOAc and elimination of the resulting C16-ester gave DPA ( 6) in 76% yield.The overall yield starting from 3acetoxysolanidine (11) to DPA (6) over 9 steps was 30%.

The spirosolane routes
An alternative procedure for the conversion of tomatiedol (24) in DPA consists of nitrosation of tomatidenol followed by decomposition of the nitroso compound, oxidation and elimination 59 .

Shortcuts to tomatidenol (24) starting from 3-acetoxysolanidine (11)
To compete with existing industrial processes this route should be shortened and expensive reagents should be avoided.Shortcuts were attempted to convert compounds from the first part of the route (11→24) to compounds from the second part of the route (24→6).
A first improvement was found in the replacement of Red-Al by activated Zn in HOAc 60 in the reduction of 15, which was already mentioned in a patent (Scheme 7).Best results were obtained by using freshly prepared activated Zn 61 and the acetate 27 was obtained in a yield of 90%.LiAlH 4 62 also reduces nitrile 15 to 22 but the yield (64%) is lower than in the cases of Red-Al or Zn/HOAc. 52Chlorination of 27 with NCS leads to chloride 28 63 in 95% yield and subsequent treatment of 28 with NaOMe in MeOH gives tomatidenol (24) in 95% yield [53][54][55][56][57][58]64,65 . The clorination/dehydrochlorination process is an economically unfavorable two-step process and direct introduction of the ∆ 22,N double bond would be more efficient.This has been tried i) by elimination of HCN from 15 (Scheme 8), ii) by oxidation of amide 30 (Scheme 9), iii) by elimination of chlorine from 28, and iv) by oxidation of amine 27.
Direct elimination of the nitrile group from 15 has been tried by treatment with NaOMe, which led to a single product that was identified as 29.A mechanistic explanation for the formation of 28 is depicted in Scheme 8. 66

Scheme 8
The amide 30 could be obtained in 95% yield by treatment of nitrile 15 with Zn in anhydrous HOAc. 26,67The introduction of the double bond in 30 was attempted with CAN 68 , MnO 2 , 10,11 and HCl(aq.) 69but unfortunately without success.These results indicate that an imine can only be formed by chlorination and subsequent dehydrochlorination of 28 with NaOMe, but under these circumstances the acetates are saponified also, and a further reaction to tomatidenol (24) can not be avoided.To achieve formation of the ∆ 22(N) bond and to prevent further reaction to tomatidenol, it will be necessary to maintain the acetate group at C16. Therefore 28 was treated with several bases (NaOAc/EtOH 70 , K 2 CO 3 /DMF, 71 LiBr/Li 2 CO 3 /DMF, 71 NaOMe/toluene, t-BuOK, 72 Et 3 N, 73 Et 2 Nli, 74 KHMDS, 74,75 NaH/DMSO 76 ) but all attempts were unsuccessful.It proved to be impossible to eliminate HCl from 28, when an acetate group at C16 is present 65 .On the other hand, elimination of HCl proved to be relatively easy if a free hydroxyl group is present at C16 as was demonstrated by the conversion of 23 to tomatidenol (24) in 73% yield with DBU. 77An intramolecualr elimination as depicted in Scheme 10 53,54,56,58,78 can explain these results.Direct oxidation of 27 with CrO 3 /pyridine, 79 KMnO 4 , 12 CAN, 68 or KMnO 4 /Al 2 O 3 68 was in all cases unsuccessful and only the starting material was recovered.Adam and Huong 80 described several successful spirosolane formations (tomatidine (24), solasodine, soladulcidine, and solasodenone) through oxidation of the secondary amine with MnO 2 81 followed by stereospecific cyclization. 54,82However, several attempts with differently activated MnO 2 80,83 failed to give any of the desired imine 31 and only with MnO 2 freshly prepared according to the method of Attenburrow et al. 81 some imine 31 could be detected in the reaction mixture but the major product was still the starting material 27.
Although the diacetylated imine 31 could not be obtained from nitrile 15, amide 30, chloride 28, or amine 27, it can become available from tomatidenol (24) by treatment with ZnCl 2 and Ac 2 O in HOAc. 84The availability of 31 gives the possibility to investigate two alternative degradation routes toward DPA (6).However, the limited amount of solanidine (1), available from the spray-dried protamylasse fraction, and the fact that tomatidenol (24) is not commercially available gave severe problems for further research.On the other hand, solasodine, which differs from tomatidenol only in the configuration at C25, is commercially available and can be considered as an acceptable model compound to investigate two alternative degradation routes toward DPA (6).Any successful results could then be repeated with tomatidenol (24)  itself.Treatment of solasodine with ZnCl 2 in Ac 2 O and HOAc in the same manner as previously described for tomatidenol (24) gave imine 33 in 96% yield 85 (Scheme 11).
In the first route, isomerization of the endocyclic ∆ 22(N) double bond to the exocyclic ∆ 20,22 position was achieved after methylation of 33 followed by treatment with aqueous NaHCO 3 in acetone 85 to 34 in 80% overall yield (Scheme 11).Oxidation of 34 was attempted with CrO 3 /HOAc, 86 KMnO 4 /Al 2 O 3 , 13 ozone, 87 and O 2 /CuCl 88 but all attempts were unsuccessful.Hydrolysis of 34 with HCl or HBr in HOAc was tried but failed as well.The reasons for these failures are probably the same as for the oxidation of 3-acetoxysolanidi-5,20-ene (4), namely the steric hindrance around the ∆ 20,22 double bond.Mopac calculations on 37 indicate that it becomes difficult for other reagents to approach the ∆ 20,22 double bond due to C18, the F-ring itself, and the N-methyl.Similar negative results were found by Sato and Ikekawa in their attempt to oxidize a compound similar to 34 with an acetyl instead of a methyl group attached to the nitrogen. 89 89,90Deprotection of the C16 acetate with methanolic K 2 CO 3 followed by addition of HOAc gave a 82% yield of 37 instead of the expected 26. 89A renewed attempt to eliminate MeOH from 37 with HOAc at reflux temperature also failed.
Treatment of 36 with K 2 CO 3 in a mixture of water and EtOH overnight gave hemiacetal 38 in 78% yield.The formation of 38 in about the same yield was also achieved upon treatment of 36 with NaOH in dioxane.In this case the elimination of water from 38 with HOAc at reflux temperature 89 also failed.This result is in agreement with that of Cambie et al., 91 who was not able to dehydrate compound 39.However a similar dehydration could be achieved by Sato and Ikekawa. 89It thus turned out that properties of imine 33 as an intermediate in the synthesis of DPA (6) were not redeemed.Despite all our efforts, shortening of the route from 3acetoxysolanidine (11) via tomatidenol (24) to DPA (6) as depicted in Scheme 6, could not be accomplished.

Experimental Section
General Procedures.Dry reactions were performed under a steady stream of dry nitrogen or argon with glassware dried at 140ºC.All 1 H and 13 C NMR spectra were measured with a Bruker AC-E 200 spectrometer.400 MHz 1 H and 100 MHz 13 C NMR were measured with a DPX 400 spectrometer.Chemical shifts are reported in parts per million (δ) relative to tetramethylsilane (δ 0.0).MS and HRMS data were obtained with a Finnigan Mat 95 spectrometer.FT-IR spectra were measured with a BIO-RAD FTS-7 infra-red spectrometer.Optical rotations were measured with a Perkin-Elmer 241 polarimeter with the concentrations denoted in units of g/100 ml.Analytical data were obtained using a Carlo Erba Elemental Analyzer 7206.Melting points are uncorrected.Solvents were freshly distilled by common practice.Product solutions were dried over Na 2 SO 4 prior to evaporation of the solvent under reduced pressure by using a rotary evaporator.For flash chromatography, Merck Kieselgel silica 60 (230-400 Mesh) or Baker Alumina was used.Reactions were monitored with TLC using Merck silica gel 60F254 plastic sheets.Compounds were visualized on TLC by UV detection and by spraying with acid and subsequent heating.
To a suspension of freshly activated Zn (210.7 mg) in HOAc (1.3 ml) and H 2 O (3.0 ml) was added 15 (107.6 mg, 0.21 mmol) and the reaction mixture was heated at reflux temperature.After 2 hours, the reaction mixture was filtered through Hyflow and evaporated in vacuo.The residue was dissolved in CHCl 3 , washed with aqueous NaOH (0.05 M) and brine, dried, and evaporated in vacuo to obtain a white amorphous solid.Purification by column chromatography with CHCl 3 /MeOH 9/1 followed by crystallization from EtOAc gave 27 (92.0mg, 90%) as a white solid.M.p. 170-173ºC (172-173ºC 100 ); 1
The NMR data are identical to those obtained with method A.