An efficient asymmetric synthesis of Enigmols (1-deoxy-5-hydroxysphingoid bases), an important class of bioactive lipid modulators

An enantioselective synthesis of the four 2 S -diastereomers of enigmol (2-amino-octadecane-3,5-diols) from L -alanine is reported. The procedure described is robust, highly stereoselective at each step, and suitable for preparation of gram quantities of material.


Introduction
Sphingolipids (ceramides, sphingomyelins, cerebrosides, gangliosides, etc.) are a structurally diverse class of complex membrane lipids, composed of long-chain 2-amino-1,3-diols (sphingoid bases), amide-linked fatty acids, and numerous phosphoric acid and carbohydrate polar head groups, connected to the primary hydroxyl.Studies performed over the past two decades have shown that the low molecular weight sphingolipid metabolites (ceramide 3, sphingosine 4, sphingosine 1-phosphate (S1P, 5)) are involved in many critical signaling pathways that regulate, inter alia, cell growth, cell differentiation and apoptosis. 1Furthermore, sphingolipids function as modulators for growth factor receptors and as second messengers for a growing list of agonists. 1,2phingolipid modulators are now emerging as potential therapeutics, and the first candidates targeting S1P signaling have just begun to come on to the market.For example, the Novartisdeveloped drug candidate Gilenia (previously Fingolimod) 6 just received FDA approval.This compound is a potent S1P antagonist and has shown to be beneficial to patients having relapsing multiple sclerosis in clinical studies.Similarly, safingol 7, an unnatural sphinganine diastereomer and potent sphingosine kinase inhibitor, has been shown to provide some benefit in breast cancer patients when used topically. 5

Figure 1
During the early stages of our own sphingolipid modulator program, we were intrigued by the apparent bifurcation in the biological functions of the sphingolipid biosynthesis products shown in Scheme 1.

Scheme 1
Specifically, the sphingoid bases 1 and 4, as well as their amine metabolites (N-methylation and fatty acid N-acylation adducts 2 and 3) appeared to promote anti-mitotic activity, while sphingosine-1-phosphate (S1P, 5) appeared to be pro-mitotic. 1Thus, from a structural perspective, phosphorylation of the C-1 hydroxyl group seemed to be the key differentiating element that resulted in two vastly different biological outcomes.Perhaps, we hypothesized, this observation could serve as the basis for designing sphingoid base analogs that could only elicit anti-mitotic behavior.The attractiveness of this proposal was enhanced by the knowledge that 1deoxy sphingoid bases could not be substrates for decomposition by sphingosine-1-phosphate lyase, and that 1-deoxysphinganine 9, an inhibitor of sphingosine kinase, lowers levels of endogenous S1P, thereby exerting an additional, albeit indirect, anti-mitotic effect.
While the conceptual excision of the C-1 hydroxyl group would likely achieve our goals from a mechanism of action perspective, we worried that removal of the hydroxyl group would convert an already "greasy" compound into an even more hydrophobic entity that would have poor drug-like properties.To circumvent this problem, we focused instead on compounds containing the 2-amino-3,5-diol head-group (later termed 'enigmols'), which would eliminate the pro-mitotic and anti-apoptotic properties of naturally occurring sphingoid bases without altering the overall hydrophobicity and (presumably) compartmentalization properties of the new analogs.Furthermore, the addition of a new chiral center at C-5 would allow us, in principle, to examine the biological properties of eight different stereoisomers (four pairs of diastereomers), thereby providing us with a straightforward approach for tuning the biological specificity of these compounds.

Figure 2
This potential utility of our design strategy was reinforced by the discovery of Fumonisin B1 8, a naturally occurring mycotoxin produced by the fungus Fusarium moniliforme, that shares structural similarities to the sphingoid bases. 6Soon after its discovery, Fumonisin B1 was shown to be an inhibitor of ceramide synthase, a key enzyme in the sphingolipid biosynthetic pathway (see Scheme 1). 7Since addition of exogenous ceramide synthase inhibitors, like Fumonisin B1, elevates the endogenous levels of sphingoid bases (primarily sphinganine) to toxic levels, this provided yet another mechanistic rationale for studying these compounds. 8

Figure 3
In the early 1990's, we performed the first total synthesis of the four 2S-enigmol diastereomers 10-13 as simplified structural analogs of Fumonisin B1, starting from L-alanine. 9hile this route served its purpose, a more efficient synthesis was required to prepare larger quantities of each compound.Herein, we describe a scalable synthesis of 2S,3S,5S-enigmol 10 and its C-3 and C-5 diastereomers 11-13.The synthesis is centered around a highly diastereoselective boron-mediated aldol reaction, which is a modification of a methodology developed in our lab, between (S)-3-(dibenzylamino)butan-2-one 16 (derived from L-alanine) and tetradecanal 17. 10 Interest in this series of compounds is evident by recent publications describing the synthesis of 1-deoxy-5-hydroxy sphingoid bases. 11

Results and Discussion
The synthesis of the enigmols drew from the chiral pool, starting from L-alanine.Conversion into (S)-3-(N,N-dibenzyl)-alanine 14 was accomplished following literature procedures. 12Further manipulation of 14 to obtain the methyl ketone (S)-3-(dibenzylamino)butan-2-one 16 was achieved in two steps via the intermediate Weinreb amide in nearly quantitative yield, a significant improvement over direct conversion of the carboxylic acid.The aldol reaction of ketone 16 with tetradecanal 17 in the presence of (-)-diisopinocampheyl chloroborane ((-)-DIP-Cl) provided the corresponding (2S,5S)-keto derivative 18 exclusively.The extremely high diastereoselectivity (the 5R product is not observed) of this aldol reaction is driven by the bulky N,N-dibenzyl substituent at C-2, a somewhat counterintuitive finding when using a chiral boron reagent.However, the same stereochemistry (5S) was observed independent of boron-ligand chirality.Reduction of ketone 18 with sodium borohydride proceeded smoothly to give (2S,3S,5S)-aminodiol 19 without the observed formation of other products.To gain access to the inverse reduction product, Luche conditions were used, producing a 4:1 mixture of the anti 20 to syn 19 1,3-diol, which could be easily separated by flash chromatography.The enigmols 10 and 12 were then obtained from 19 and 20 following standard hydrogenolysis conditions with Pearlman's catalyst under 1 atmosphere of hydrogen.
In the early stages of development, the aldol chemistry between methyl ketone 16 and tetradecanal 17 10b proved to be quite versatile for small-scale synthesis.However, when we attempted to utilize this approach on a gram scale, we observed partial-to total-loss of optical activity.The degree to which optical activity was reduced was substrate-and batch-dependent.After careful investigation it became apparent that the methyl ketone 16 and the 3-ketoderivative 18 are prone to racemization at C-2 under chromatographic conditions (on silica or neutral alumina).This conclusion was reinforced by the observation that initial fractions of methyl ketone 16 on silica gel produced a clear oil that had a greater specific rotation than later fractions, which had no optical rotation and became solid after drying.The epimerization of compound 18 occurred more slowly on silica than 16, and this process could be further retarded by neutralizing the silica with 1% triethylamine.To circumvent this problem, we developed a synthetic route to the methyl ketone 16 through the Weinreb amide 15, which obviated the need for purification and therefore racemization, before using 16 in the aldol reaction.Also, reduction of 18 was performed immediately after simple work-up, also avoiding racemization.The methyl ketone 16 can be stored at -20 ˚C without racemization for several months.Chiral chromatography was used to verify that enantiopure methyl ketone 16 yielded only one enantiomer of N,N-dibenzyl-enigmol 19 after two consecutive steps without isolation of the intermediate compound 18 (>99.8%ee, HPLC).
Access to the remaining two 5R-diastereomers was accomplished by performing a Mitsunobu inversion at that center.First, the free base 10 was regioselectively converted to oxazolidinone 21 with carbonyldiimidazole, tying up the C-2 and C-3 functionality.Next, the hydroxyl group at C-5 was inverted using standard conditions.Finally, microwave irradiation with lithium hydroxide at 120 ˚C for 1.5 hours gave the corresponding (2S,3S,5R)-amino-diol 11. 13 Similar reaction conditions were used to convert 12 into 13.

Conclusions
Using the synthetic methodology described above, we have successfully prepared all four 2Senigmol diastereomers.Although others have attempted to synthesize the enigmol amino-diol headgroup via a boron-mediated aldol reaction, 11d this is the first example of a high yielding, completely diastereoselective application.In addition, we have successfully prepared (2S,3S,5S)and (2S,3R,5S)-enigmols in gram quantities for biological evaluation, proving the scalability and reproducibility of our procedure.

Experimental Section
General.All reactions were run under inert atmosphere unless otherwise stated.All chemicals used were purchased from Sigma-Aldrich Chemical Co. and used without further purification, unless otherwise stated.All solvents termed "dry" were dried via a Glass Contour solvent Page 269  ARKAT USA, Inc.
purification system (SG Water, USA LLC).All melting points were obtained on a 200W MelTemp apparatus and are uncorrected. 1H NMR spectra were obtained on a 400 MHz Varian Spectrometer, referenced to residual solvent or to TMS. 13  The mixture was stirred for 20 minutes at -20 ˚C, followed by addition of granulated sodium borohydride (850 mg, 22.5 mmol) in 9 portions of ~100 mg each with continued stirring until no starting material remained.The resulting mixture was quenched by the dropwise addition of glacial acetic acid (~2 mL), and the resulting mixture was concentrated in vacuo, diluted with Et2O, and neutralized with saturated sodium bicarbonate.The solids were then filtered off, and the organic layer was isolated and washed with water, brine, dried over magnesium sulfate, and concentrated under vacuum to give the crude product as an oil (12.4 g).
After this material was subjected to sublimation at 0.2 mm Hg at 50 o C for 24 hours to remove pinene byproducts derived from DIP-Cl, the remaining oil (7.5 g) was purified by flash chromatography (SiO2, hexanes: ethyl acetate, both containing 2% triethylamine) to yield 20 as a waxy colorless solid (3.60 g, 67%).Compound 19 was also isolated as a waxy colorless solid (761 mg, 11%).Rf = 0.27 (4:1 hexanes: ethyl acetate); IR (νmax, cm carbonyldiimidazole (121 mg, 401 µmol) in dry THF (2 mL), dropwise, at 0 o C.After addition was complete, the reaction was allowed to warm to ambient temperature over 3.5 hours.The reaction mixture was quenched by addition of MeOH and the solvents were evaporated under reduced pressure.The crude material was purified by flash chromatography (SiO2, hexanes: ethyl acetate) to give a white solid (156 mg, 86%).Rf = 0.48 (9:1:  23).This compound was synthesized by the method reported in literature. 24To a solution of oxazolidinone 21 (610 mg, 1.86 mmol) in dry THF (10 mL) was added triphenyphosphine (1.46 g, 5.55 mmol) and benzoic acid (1.12 g, 9.16 mmol) at 0 o C. The resultant mixture was allowed to stir at 0 o C for 15 minutes, to which diisopropyl azodicarboxylate (1.90 mL, 9.16 mmol) was added.After 1 hour at 0 o C, the mixture was then diluted with Et2O, washed with the aqueous saturated sodium bicarbonate, brine, dried over magnesium sulfate, and concentrated in vacuo to give the crude product (4.9 g).The resulting material was taken into hexanes and filtered, then concentrated and purified by flash chromatography (SiO2, hexanes: ethyl acetate) to yield 23 as an off-white powder (736 mg, 92%).Rf = 0.40