Synthesis and in vitro cytotoxicity of acetylated 3-fluoro, 4-fluoro and 3,4-difluoro analogs of D-glucosamine and D-galactosamine

Background: Derivatives of D-glucosamine and D-galactosamine represent an important family of the cell surface glycan components and their fluorinated analogs found use as metabolic inhibitors of complex glycan biosynthesis, or as probes for the study of protein–carbohydrate interactions. This work is focused on the synthesis of acetylated 3-deoxy-3-fluoro, 4-deoxy-4-fluoro and 3,4-dideoxy-3,4-difluoro analogs of D-glucosamine and D-galactosamine via 1,6-anhydrohexopyranose chemistry. Moreover, the cytotoxicity of the target compounds towards selected cancer cells is determined. Results: Introduction of fluorine at C-3 was achieved by the reaction of 1,6-anhydro-2-azido-2-deoxy-4-O-benzyl-β-D-glucopyranose or its 4-fluoro analog with DAST. The retention of configuration in this reaction is discussed. Fluorine at C-4 was installed by the reaction of 1,6:2,3-dianhydro-β-D-talopyranose with DAST, or by fluoridolysis of 1,6:3,4-dianhydro-2-azido-β-D-galactopyranose with KHF2. The amino group was introduced and masked as an azide in the synthesis. The 1-O-deacetylated 3-fluoro and 4-fluoro analogs of acetylated D-galactosamine inhibited proliferation of the human prostate cancer cell line PC-3 more than cisplatin and 5-fluorouracil (IC50 28 ± 3 μM and 54 ± 5 μM, respectively). Conclusion: A complete series of acetylated 3-fluoro, 4-fluoro and 3,4-difluoro analogs of D-glucosamine and D-galactosamine is now accessible by 1,6-anhydrohexopyranose chemistry. Intermediate fluorinated 1,6-anhydro-2-azido-hexopyranoses have potential as synthons in oligosaccharide assembly.

The remarkable ability of acetylated fluoro analogs of GlcNAc and GalNAc to perturb the (glycosamino)glycan biosynthesis and their antiproliferative properties aroused our interest in developing a methodology for the preparation of a complete series of acetylated 3-fluoro, 4-fluoro-, and 3,4-difluoro analogs 1, 4-8 ( Figure 1) including previously unknown members of this class of hexosamine mimics: acetylated 3-fluoro-D-GalNAc 6, 3,4-difluoro-D-GlcNAc 7 and 3,4-difluoro-D-GalNAc 8, as well as 3-fluoro-D-GlcNAc 5, in which case the reported synthesis was troublesome and low-yielding [20]. To carry out the synthesis, flexible synthetic methods for the stereoselective introduction of fluorine at one or more designated positions of the hexosamine skeleton are necessary. The elaborated chemistry of 1,6-anhydrohexopyranose derivatives is suitable for this purpose [21][22][23]. Building on previous results from our [24] and other groups [25][26][27][28], we designed an approach based on stereoselective introduction of an azide as a masked amine group at C-2, and fluorine at C-3 and C-4 by nucleophilic displacement. Resulting 3-fluoro, 4-fluoro, and 3,4-difluoro analogs of 2-azido-1,6-anhydrohexopyranoses were then converted into the target fluoro analogs of D-glucosamine and D-galactosamine (Scheme 1). Dual protection of the anomeric and primary hydroxy groups in the form of the 1,6-anhydro bridge reduced the number of protecting groups, and the rigid bicyclic skeleton of 1,6-anhydrohexopyranoses enabled a high degree of regioand stereocontrol necessary for the introduction of heteroatomic substituents at C-2, C-3, and C-4. The synthesis of the analogs 6-8 has not yet been reported to our knowledge, while the synthesis of 1, 4 and 5 represents an alternative to the published procedures [2,20,29]. Herein we also report on the cytotoxicity of prepared fluoro analogs in the human ovarian cancer A2780 and prostate cancer PC-3 cell lines. Preliminary results for the synthesis of compounds 5 and 6 were communicated earlier in a letter [30].
The synthesis of mono-and difluoro analogs of 2-azido-2deoxy-1,6-anhydrohexopyranoses, which are key intermediates, is shown in Scheme 3. We first explored the reactions of azido alcohols 11, 12 and 15 (Scheme 3) with diethylaminosulfur trifluoride (DAST) to achieve the introduction of a nucleophilic fluorine atom. Reaction of 11 with DAST using a minor modification of the reported procedure [28] provided the D-gluco- 27 arising from solvent participation, and to a partial decomposition (indicated by TLC) at high temperatures necessary for the reaction to proceed. The low values of the vicinal coupling constant 3 J H3,H4 = 2.0 Hz and the large value of the geminal coupling constant 2 J C3,F = 29.6 Hz evidenced a trans-diaxial relationship between the C-3 and C-4 substituents in 26. The reaction of fluorohydrin 26 with DAST in benzene under heating afforded, as the main product, 3,4-difluoro-D-gluco analog 28 (46%), and the rearranged 2,6-anhydro compound 29 (12%) as a side product. Products 28 and 29 can be separated by careful chromatography and their structures were verified by single crystal X-ray analysis.
The reaction of the D-gluco-configured 3-hydroxy derivatives 11, and 26 with DAST, and the previously reported reactions of D-gluco-configured 3-hydroxy derivatives 33 [26], 34 [27], and 35 [25] (Scheme 4) with DAST are an important means of obtaining 3-deoxy-3-fluoro derivatives of D-gluco-configured aldohexopyranoses which are difficult to prepare otherwise. These reactions characteristically proceed with a clean retention of configuration which can be explained by an anchimeric assistance of the trans-diaxially positioned (with respect to C3-OH) polar groups at C-2 or C-4, or by an internal fluorine attack as in S N i substitution. A simple S N 2 displacement leading to configurational inversion is probably suppressed by the steric effects of the axially positioned groups at C-2 and C-4, and repulsive effects of their aligned dipoles [51]. Compounds 11, and 33-35, possess a trans-diaxially positioned benzyloxy group at C-4 capable of participation through an oxiranium intermediate species (Scheme 4A) [52][53][54]. The formation of the rearranged difluoride 29 from alcohol 26 (Scheme 3) suggests an anchimeric assistance of the vicinal C-2 azido group. Although rare, azide participation was postulated before [52]. The main product 28 and rearranged difluoride 29 can arise from the same intermediate species 39 (Scheme 4B, pathways a and b, respectively). Compound 19 can, in principle, be also formed from 11 through azide participation (not shown). In such a case the reaction is unexpectedly sensitive to minor steric alterations of the substrate because the C-4 epimer 15 (Scheme 3) did not react. An internal fluorine attack from the β-face of the tetrahydropyran ring through a concerted (Scheme 4C) or contact ion-pair (Scheme 4D) S N i mechanism cannot be ruled out [55,56] because the bulky Et 2 NSF 2 O substituent at C-3 might force the substrate to adopt boat B 3,O conformation [57] bringing the C-2, C-3 and C-4 substituents into a trans-equatorial arrangement unfavorable for anchimeric assistance.
With the D-gluco-and D-galacto-configured deoxofluoro derivatives of 2-azido-1,6-anhydrohexopyranoses in hand, we examined their conversion to the target acetylated hexosamine analogs. Initially, compound 19 was hydrogenated (Pd/C) and acetylated to afford acetamide 40 in modest yield (37%). Sulfuric acid-catalyzed [35] cleavage of the internal acetal with acetic anhydride gave a mixture containing 1,2-oxazoline 41 as the main product (Scheme 5). The structure of oxazoline 41 was confirmed by single crystal X-ray diffraction analysis which  also confirmed the retention of configuration during the preceding fluorine introduction.
TESOTf as a catalyst for acetolysis gave better results than sulfuric acid in terms of product purity. Hydrogenolysis of 42 on palladium in ethanol/HCl followed by acetylation of the amino group furnished the target acetylated 3-fluoro-D-GlcNAc 5 as a chromatographically separable mixture of anomers ( Table 1). Addition of HCl was found necessary to effect a clean hydrogenolytic removal of the O-benzyl group in 42. The D-gluco configuration of 42 and 5 was reflected in the large values of 3 J H3,H2/4 (7.6-10.2 Hz), and the chair inversion accompanying the cleavage of the 1,6-anhydro bridge ( 1 C 4 → 4 C 1 ) in a decrease of the geminal fluorine-carbon coupling value 2 J C4,F (31.5 Hz in 40 → 18.4 Hz in 5 (α-anomer)) [59]. 1 H and 13 C NMR spectra of 5 reported in [20] are comparable with our data for the α-anomer (5α). A discrepancy in the values of specific optical rotation (reported +10°, +81°o btained by us) can be explained assuming that the value for the β-anomer (5β) was reported in [20]. Our synthesis of 5 is more practical (55% from 10 in 3 steps) than the earlier preparations which were reported troublesome owing to low yields (28% and 10%, respectively) of the key fluorination step [18,20].
For the remaining fluoro derivatives of 1,6-anhydro-2-azidohexopyranoses, TESOTf-catalyzed acetolysis and subsequent hydrogenation (Pd/C in a mixture of ethanol and acetic anhydride) was applied to obtain the target acetylated fluoro analogs (Table 1). Compound 21 was acetolyzed to 43 which was obtained as a mixture of anomers and characterized by NMR and finally hydrogenated to furnish acetylated 3-fluoro-GalNAc 6 isolated as a separable mixture of anomers. The D-galacto configuration of 6 is manifested by the lower 3 J H3,H4 coupling value (3.4 Hz, α-anomer) in comparison with that of its C-4 epimer 5 (8.6 Hz, α-anomer). Acetolysis of the internal acetal in 26 proceeded more slowly and beside the desired product 44 (78% yield), 3-O-acetyl derivative 45, in which the 1,6anhydro bridge remained intact, was also isolated. Hydrogenation of 44 furnished the known acetylated 4-fluoro-D-GlcNAc 1 [2]. The large values of the vicinal coupling constants 3 J H3,H2/4 and 3 J H4,H5 (8.9-11.1 Hz) confirmed the D-gluco configuration of 1 and the chair inversion when going from 26 to 1, which was also manifested by a decrease in the geminal coupling constant 2 J C3,F (29.6 → 18.9 Hz, α-anomer). Acetolysis and subsequent hydrogenation of 31 provided the known peracetylated 4-deoxy-4-fluoro-D-galactosamine 4 [29]. The chair inversion of the tetrahydropyran ring associated with the cleavage of the 1,6-anhydro bridge is indicated by the increase in the coupling value 3 J H2,H3 (1.4 → 11.6 Hz) and decrease in the coupling value 2 J C5,F (27.4 → 18.4 Hz) when going from 31 to 4 (α-anomer). Acetolysis of 28 afforded diacetate 47 containing chromatographically inseparable impurities (NMR). They were removed in the next hydrogenation step to yield the peracety- To study the influence of 1-O-deacetylation on the cytotoxicity, the monofluorinated analogs 1, and 4-6 were subjected to anomeric deacetylation (Scheme 6). Compound 5 provided 1-Odeacetylated product 49 by treatment with BnNH 2 in THF. Since acetylated 4-fluoro-D-GlcNAc 1 under these conditions did not react cleanly, we used piperidine-promoted [60] deacetylation to prepare 2 in 74% yield. Similarly, acetylated 4-fluoro-D-GalNAc 4 gave 50 in 60% yield. The attempted anomeric deacetylation of 3-fluoro-D-GalNAc 6 by treatment with piperidine followed by chromatography gave a fraction containing an inseparable side-product in addition to the expected deacetylated product 51. The side-product showed no fluorine resonance in 19 F NMR and its molecular formula C 17 H 28 N 2 O 7 assigned by LC-HRMS corresponded to a formal displacement of fluorine by piperidine, leading probably to compound 53. When pure 51 (prepared by another method, see below) was reacted with excess piperidine, high resolution ESIMS analysis of the reaction detected transient formation of an adduct ion corresponding to a supposed intermediate enal 52 (Scheme 6), while the adduct ion corresponding to 53 was the final product (see Supporting Information File 1). Presumably, piperidine as a relatively strong base effected dehydrofluorination of 51 to enal 52 which then added piperidine to give 53 as a byproduct (Scheme 6). To avoid the action of basic amines, a silica gel mediated anomeric deacetylation, recommended for 2-aminosugars [61], was tried. The reaction proceeded extremely slowly with our substrate 6 and the product 51 was obtained in only 40% yield after chromatography and recrystallization.
Increased cytotoxicity as a result of 1-O-deacylation was noted for a variety of acylated (nonfluorinated) D-mannosamine and D-glucosamine derivatives [62,63]. Acylated hexosamine derivatives were subsequently studied as possible templates for the development of anticancer therapeutics [64,65]. While the ability of hexosamine derivatives and analogs to inhibit cell growth creates an avenue for their use in the development of anticancer drugs, it also limits their utility as agents to modify the cellular glycome [62]. The cytotoxic activity of peracetylated monofluoro analogs 1, and 4-6, their 1-O-deacetylated derivatives 2, and 49-51, difluoro analogs 7 and 8, and oxazoline 41 was therefore tested for 24 h on the human prostate cancer PC-3 cell line, and human ovarian cancer A2780 cell line using the MTT assay, and the obtained IC 50 values were compared with those obtained for cisplatin and 5-fluorouracil.
All of the tested compounds induced only moderate to weak inhibition of A2780 cell proliferation (IC 50 values ranging from 78 μM to 327 μM, Table 2) in comparison to cisplatin (IC 50 12.9 μM). Anomeric deacetylation resulted in a higher activity in the case of 3-fluoro-D-GlcNAc (49, IC 50 84 μM, Table 2, entry 6), and especially 4-fluoro-D-GlcNAc (2, IC 50 78 μM, ca. 4-fold higher activity than that of the peracetylated 1,  Table 2, entry 13). The effect of anomeric deacetylation was particularly pronounced in 3-fluoro-D-galactosamine 51 owing to inactivity of the corresponding peracetate 6. Finally, difluoro analogs 7 and 8, and oxazoline 41 exhibited weak (compound 7) to none (compounds 8 and 41) activity in the PC-cell line ( Table 2, entries 9-11). Taken together, our results seem to corroborate in part the observations by Yarema at al. that the liberation of the anomeric hydroxy group in acylated hexosamines resulted in an enhancement of their cytotoxicity [62,63]. The effect of anomeric deacetylation was, however, found much more cell-line and substratespecific for our fluoro analogs than it was for acylated natural hexosamines [62].

Conclusion
We have developed a synthetic route for the preparation of a series of 3-and 4-deoxofluorinated analogs of D-glucosamine and D-galactosamine. The choice of O-benzylated-1,6:2,3-dianhydro-β-D-mannopyranose 10 and 2,3-isopropylidene-Dmannosan 16 as starting material permits regio-and stereoselective introduction of both fluorine and nitrogen (as an azide) by nucleophilic substitution before acetolysis of the 1,6-anhydro bridge. The characteristic feature of the synthesis is the introduction of fluorine at C-3 by the reaction of D-gluco-configured 3-hydroxy derivatives with DAST with retention of configuration. The 1-O-deacetylated 3-fluoro and 4-fluoro analogs 51 and 50 of acetylated D-galactosamine were shown to be more cytotoxic in the PC-3 cell line than cisplatin and 5-fluorouracil.
Most of the other fluoro analogs displayed moderate to low cytotoxicity. Fluoro analogs 6-8 are new compounds and their influence on the cell-surface glycan biosynthesis is currently being studied. We anticipate that 1,6-anhydro-2-azido-fluorohydrins 20, 21, 26 and 31 will find use as building blocks for the synthesis of fluorinated oligosaccharides and other glycoconjugates because they can be immediately employed as glycosyl acceptors or readily converted into glycosyl donors. Research in this way is now in progress in our laboratory.