A Concise Synthesis of Glycolipids Based on Aspartic Acid Building Blocks

L-Aspartic acid building blocks bearing galactosyl moieties were used to synthesise glycolipid mimetics of variable hydrocarbon chain length. The glycolipids were readily prepared through amide bond formation using the TBTU/HOBt coupling methodology. It was observed that, under these conditions, activation of the α-carboxylic acid of the intermediates led to near complete racemisation of the chiral centre if the reaction was carried out in the presence of a base such as triethylamine. The enantiomerically pure glycolipids were obtained after careful consideration of the synthetic sequence and by performing the coupling reactions in the absence of base.


Introduction
Synthetic glycomimetics have been the subject of much research activity in the field of carbohydrate chemistry. The important role of carbohydrates in biological systems has prompted the development of different types of glycomimetics intended for diverse applications, such as therapeutic leads [1,2], novel materials [3], biosensors and diagnostic tools [4,5]. Glycolipid mimetics [6], in particular synthetic derivatives of biologically relevant ceramides (such as galactosyl ceramides, shown in Figure 1a), have attracted the attention of many carbohydrate chemists over recent years [7].

OPEN ACCESS
Amino acids that allow for side chain functionalization with glycosyl moieties, such as serine and aspartic acid, have been popular choices as the starting point for the preparation of glycolipid analogues [8,9]. The carboxylic acid present on the aspartic acid side chain offers the possibility for attachment of mono or oligosaccharides, while both the amino and carboxylic acid groups at the α-carbon allow for further functionalization. Due to the biological relevance of N-linked glycosides, this type of building blocks has been used predominantly in the synthesis of glycopeptides and glycopeptoids and hence, numerous examples of such compounds can be found in the literature [10][11][12][13]. In this study we report our investigations towards the synthesis of galactosylated building blocks based on: (i) orthogonally protected; (ii) enantiomerically pure and (iii) commercially available L-aspartic acid derivatives, as we intend to expand their application to the preparation of glycolipid mimetics. These non-natural glycolipids may be bioactive as neuroprotective agents [14] and/or may be used in materials or formulation science [15]. The nature of the building blocks should allow for a modular approach which could lead to the facile preparation of a small collection of glycolipids of different fatty acids chain lengths, such as 1-4, shown in Figure 1b. This feature of the glycolipid structure affects strongly its physicochemical characteristics, as well as its potential biological activity [16].  [17,18] and 6 [19] and of the commercially available L-aspartic acid derivatives 7 and 8, used for the modular synthesis of glycolipid mimetics 1-4. We have initially focused our attention on derivatives of decanoic acid (C-10), such as 1 and 3, and tetracosanoic acid (C-24), such as 2 and 4, as representative examples of medium and long fatty acid chain lengths. In glycolipids 1 and 2, the galactosyl moiety is connected to the aspartic acid by a flexible ethylene-type linker, while glycomimetics 3 and 4 resemble the native N-linked glycosides, as the acid conjugation occurs directly at the anomeric center. It is therefore expected that both sets of compounds would have different degrees of conformational freedom, which in turn may have an effect on their potential biological activities and physical properties.

Synthesis of the Glycolipids
Both the galactosyl amines 5 [17,18] and 6 [19] used in the syntheses described herein are readily prepared from D-galactose pentacetate following procedures described in literature. Our initial approach to the glycolipid mimetics 1-4 involved a convergent synthesis (Scheme 1), whereby the N-Boc--benzyl ester protected L-aspartic acid 7 was coupled to tetradecylamine using standard TBTU/HOBt activation conditions in the presence of triethylamine. Subsequent removal of the N-Boc protecting group with TFA afforded the amine 9a, which was acylated with decanoic acid using the above mentioned TBTU/HOBt methodology. Hydrogenolysis of the side chain benzyl ester was carried out at 50 °C to enhance solubility and it afforded carboxylic acid 10a, which was then coupled to the primary amine of galactosyl derivative 6, to yield the acetyl protected glycolipid 11a. The 1 H-NMR spectrum of 11a showed distinct duplication of every expected signal in a 1:1 ratio. To rule out possible conformational exchange equilibrium, variable temperature 1 H-NMR spectra of compound 11a were recorded in d 6 -DMSO. No coalescence of the signals was observed at temperatures as high as 80 °C, which confirmed that glycolipid 11a was, in fact, a mixture of diastereoisomers. The unexpected racemisation of the chiral α-carbon of the L-aspartic acid derivative 7 takes place in the first step of the synthesis. Although the use of TBTU and HOBt as coupling reagents is a very standard procedure in peptide synthesis [20], the activation of the α-carboxylic acid under these conditions is likely to increase the acidity of the α-proton in 7 and it may be abstracted in the presence of a base such as triethylamine. This is further supported by the disappearance of the optical activity of compound 9a [[α] 22 D = 0 (c 1.55, CHCl 3 )], while if the same coupling reaction is carried out in the absence of triethylamine, a specific optical rotation value is obtained for the L-enantiomer, compound 9b [[α] 22 D = +2.5 (c 1.55, CHCl 3 )]. The effects on reaction yields and racemisation of the products, caused by different bases and activating reagents commonly used in peptide couplings, have been extensively reviewed in the literature [21]. Most of the published procedures reporting amide bond formation of N-Boc aspartic acid 7 involve the use of carbodiimide-type coupling reagents [22], formation of activated esters, such as pentafluorophenyl derivatives [23], or mixed anhydrides [24]. However, no compromise of the optical purity of the resulting aspartate derivatives when using uronium-type reagents (such as TBTU or HBTU) has been explicitly reported so far, to the best of our knowledge [25,26]. The mixture of D and L diastereoisomers of glycolipid 11a could not be separated by flash column chromatography or by recrystallization.
The same synthetic sequence as described above was carried out on the L-enantiomer 9b. Although this route allowed access to sufficient amounts of diastereomerically pure 11b, we decided to investigate a different synthetic sequence that may result in an overall higher yield for the enantiomerically pure glycolipids, as outlined in Scheme 2. In the first step of the reviewed scheme, the free amino galactosyl derivative 6 was coupled to the N-Boc aspartic acid benzyl ester 8, which bears the free carboxylic acid at the side chain, to give the orthogonally protected compound 12. The benzyl ester on 12 was removed by hydrogenolysis and the resulting carboxylic acid at the α-carbon was then carefully reacted again with the TBTU/HOBt system, followed by the addition of tetradecylamine. To avoid racemisation of the chiral carbon in this crucial step, this reaction was carried out in the absence of base. Under these conditions, enantiomerically pure 13 was successfully obtained, albeit in a moderate yield (51% over two steps). This building block was then reacted with TFA to cleave the N-Boc group and the corresponding amine was acylated with pre-activated decanoic acid (stirred with TBTU/HOBt prior to addition) to lead to the protected glycolipid 11b. Acylation of the amine derived from 13 by treatment with TBTU/HOBt and tetracosanoic acid instead gave the longer C-24 compound 14.The hydrolysis of the acetyl protecting groups on the galactosyl moiety of both derivatives 11a and 14 was initially attempted following standard procedures, such as the Zemplén deprotection or reaction with hydrazine [27]. However, these conditions proved to be rather harsh, resulting in amide bond hydrolysis and degradation of the glycolipids. Enzyme catalysed acetolyisis was also considered, using both immobilized enzymes (such as CALB, Candida antarctica lipase [28], immobilized as Novazym 435) and soluble lipases (such as CRL, Candida rugosa lipase) [29]. These and many other lipases have been reported to chemoselectively achieve total or partial deacetylation of protected glycosides. However, the success of enzyme-catalysed reactions is often highly dependent on substrate structure, and we found that, perhaps due to the steric bulk imposed by the hydrocarbon chains, not even partial deacetylation of any of the glycolipids could be achieved. The deprotection of 11b and 14 was most successfully carried out with mild base catalysis in a heterogenous mixture of triethylamine and dichloromethane/methanol/water at 40 °C, to give the corresponding glycolipids 1 and 2. Similar considerations regarding the preservation of the chirality of the L-aspartic acid asymmetric α-carbon were observed in the reviewed syntheses of the anomeric N-linked glycolipid analogues 3 and 4 (Scheme 3). In this case, the synthesis starts again with the direct coupling of galactosyl amine 5 with the N-Boc aspartic acid benzyl ester 8 to give building block 16 [30]. In order to avoid side reactions due to the increased acidity of the resulting anomeric amide, the reaction was carried out with TBTU/HOBt but in the absence of an added base. The next step involved the removal of the N-Boc group of 15 with TFA, and the corresponding amine was acylated with either pre-activated decanoic or tetracosanoic acid, to give the corresponding intermediates 16 and 17, respectively. It was expected that the presence of the long hydrocarbon chains would introduce steric hindrance and minimize the risk of intramolecular cyclization to yield aspartimide-type by-products when attempting the coupling of the α-carboxylic acid, as this is a well known side reaction in glycopeptides and glycoprotein synthesis [31,32]. Indeed, after 16 and 17 underwent hydrogenolysis, the corresponding carboxylic acids were subjected to reaction with tetradecylamine mediated by TBTU/HOBt to give the enantiomerically pure glycolipids 18 and 19, and no significant formation of cyclic products could be observed. Since Zemplén deacetylation may have involved too harsh conditions for the final deprotection of the glycolipids 11b and 14 described earlier, we used again the mildly basic hydrolysis method described above to access derivatives 3 and 4. It must be noted that the solubility of the C-24 tetracosanoic acid derivatives, 2 and 4 is very poor (both in water and in most common solvents), when compared to that of the C-10 decanoic glycolipids 1 and 3. This is likely to hamper potential applications of the longer chain analogues.

General Methods
All chemicals purchased were reagent grade and used without further purification unless stated otherwise. Dichloromethane was freshly distilled over CaH 2 prior use. Anhydrous dimethylformamide (DMF) was purchased from Sigma Aldrich. Molecular sieves (MS) used for glycosylation and coupling reactions were 8-12 mesh and were flame dried prior to use. Reactions were monitored with thin layer chromatography (TLC) on Merck Silica Gel F 254 plates, using mixtures of hexane/ethyl acetate unless otherwise stated. Detection was effected either by visualisation in UV light and/or charring in a mixture of 5% sulphuric acid-EtOH or phosphomolybdic acid-EtOH. NMR spectra were obtained on a Bruker Avance 300 spectrometer. Proton and carbon signals were assigned with the aid of 2D-NMR experiments and DEPT experiments for novel compounds. The 2D-NMR experiments included COSY and HCCOSW, which is an HSQC type of experiment. Better resolution of the signals was observed when using the HCCOSW experiments than with conventional HSQC experiments. Chemicals shifts for 1 H-NMR are reported in ppm relative to residual solvent proton. Flash chromatography was performed with Merck Silica Gel 60, using adjusted mixtures of hexane/ethyl acetate unless otherwise stated. Optical rotations were obtained using an AA-100 polarimeter.
[α] 25 values are given in 10 −1 cm 2 ·g −1 . The melting points were obtained using a Stuart Scientific SMP1 melting point apparatus and are uncorrected. High resolution mass spectrometry (HRMS) were performed on an Agilent-LC 1200 Series coupled to a 6210 Agilent Time-Of-Flight (TOF) mass spectrometer equipped with an electrospray source both positive and negative (ESI+/−) or in a MALDI-QTOF Premier MS SYSTEM, using an α-cyano-4-hydroxy cinnamic acid matrix. Infrared spectra were obtained as a film on NaCl plates in the region 4000-400 cm −1 on a Nicolet Impact 400D spectrophotometer. (12). HOBt (0.09 g, 0.68 mmol), followed by NEt 3 (13). To a solution of 12 (0.120 g, 0.17 mmol) in ethyl acetate (6 mL), Pd/C 10% w/w (0.012 g, 10% w/w) was added. The resulting slurry was stirred under H 2 gas for 4 h. The mixture was then filtered through a Celite cake and the filtrate was concentrated under vacuum to afford the corresponding carboxylic acid as an off-white solid, which was used without further purification (0.094 g, 90%). 1 , 1 H, H-1), 4.40-4.39 (m, 1 H, H-α)

Conclusions
In summary, we present a short and convenient route to access glycolipid mimetics from suitably protected and commercially available L-aspartic acid building blocks and easily synthesized galactosyl amines. A small collection of compounds of diverse structural characteristics has been prepared. The design of suitably assembled building blocks and careful consideration of the synthetic sequence, to avoid undesired side reactions, will allow for the next generation of glycolipid mimetics bearing different mono or oligosaccharides, as well as fatty acid derivatives of different chain lengths and saturation patterns.