Total Synthesis of Natural Disaccharide Sambubiose

A practical and robust synthetic method to obtain the natural disaccharide sambubiose (2-O-β-D-xylopyranosyl-D-glucopyranose) is reported, exploring the key step in the synthesis, i.e., stereoselective O-glycosylation. Specifically, the best combinations of glycoside donors and acceptors were identified, stereospecific control of the reaction was achieved by screening several catalysts and protection/deprotection steps were evaluated and improved. The best result was obtained by coupling allyl 3,5,6-tri-O-benzyl-β-D-glucofuranoside with 2,3,4-tri-O-acetyl-D-xylopiranosyl-α-trichloro acetimidate in the presence of trimethylsilyl triflate as a catalyst giving the corresponding protected target compound as a correct single isomer. The latter was transformed accordingly into the desired final product by deprotection steps (deallylation, deacetylation, and debenzylation). Sambubiose was synthesized into a satisfactory and higher overall yield than previously reported and was also characterized.


Introduction
Chemopreventive phytochemicals (CPs) occur naturally in some plants and have been shown to inhibit all levels of carcinogenesis in both in vitro and in vivo models [1]. Among CPs, flavonoid 2-O-xylosylvitexin (Figure 1), which is constituted by carbohydrate (sambubiose, Figure 1) and aglycone (apigenin, Figure 1) components, was identified for the first time in Citrus sinensis leaves [2] and then in Beta vulgaris sub. cycla leaves and seeds [3] with the latter containing the most abundant and efficacious CPs.

Introduction
Chemopreventive phytochemicals (CPs) occur naturally in some plants and have been shown to inhibit all levels of carcinogenesis in both in vitro and in vivo models [1]. Among CPs, flavonoid 2-O-xylosylvitexin (Figure 1), which is constituted by carbohydrate (sambubiose, Figure 1) and aglycone (apigenin, Figure 1) components, was identified for the first time in Citrus sinensis leaves [2] and then in Beta vulgaris sub. cycla leaves and seeds [3] with the latter containing the most abundant and efficacious CPs.  2-O-xylosylvitexin shows strong antiproliferative activity against human colon cell cancer lines while also enhancing apoptosis, increasing cells in the G1 phase and reducing cells in the S phase as well as the proliferation of human fibroblasts [4].
Sambubiose, on the other hand, is formed by glucopyranose and xylopyranose units linked through a 1,2-trans bond in 2-position and was first isolated as a component of anthocyanins from elderberry (Sambucus nigra) [5], a plant with therapeutic properties [6,7]. In addition, sambubiose is also a constituent of the other major anthocyanins from black raspberries (Rubus occidentalis L.) [8], red currants, Davidson's plums, and small red beans, and is linked to minor anthocyanins from açai and Vaccinium padifolium blueberries [9].
To date, to the best of our knowledge, only one synthesis of sambubiose has been reported [10]; however, the method followed by Erbing and Lindberg [10] (Scheme 1) and subsequently applied by Dick et al. [11] has several drawbacks and improvements could be made in the reagents and conditions. Pharmaceuticals 2020, 13, x FOR PEER REVIEW 2 of 12 2-O-xylosylvitexin shows strong antiproliferative activity against human colon cell cancer lines while also enhancing apoptosis, increasing cells in the G1 phase and reducing cells in the S phase as well as the proliferation of human fibroblasts [4].
Sambubiose, on the other hand, is formed by glucopyranose and xylopyranose units linked through a 1,2-trans bond in 2-position and was first isolated as a component of anthocyanins from elderberry (Sambucus nigra) [5], a plant with therapeutic properties [6,7]. In addition, sambubiose is also a constituent of the other major anthocyanins from black raspberries (Rubus occidentalis L.) [8], red currants, Davidson's plums, and small red beans, and is linked to minor anthocyanins from açai and Vaccinium padifolium blueberries [9].
Firstly, the purification of benzyl 3,5,6-tri-O-benzyl-α-D-glucofuranoside (4a) to eliminate highboiling benzyl alcohol from the reaction crude mixture was extremely difficult both by distillation and/or by chromatography. Erbing and Lindberg [10], therefore, had to employ acetylation and deacetylation as additional synthetic steps via benzyl 2-O-acetyl-3,5,6-tri-O-benzyl-α-Dglucofuranoside (3a) to purify 4a (Scheme 1, "iii" and "iv" steps). Secondly, the use of the intrinsically unstable donor 2,3,4-tri-O-acetyl-α-D-xylopyranosyl bromide (5a), which must be freshly prepared and recrystallized because of its low stability, even when refrigerated, was also problematic. Moreover, Erbing and Lindberg [10] used a stoichiometric amount of mercuric cyanide as a glycosylation promoter. This reactant should not be used because of its high toxicity and dangerousness. Furthermore, Hg(CN)2 does not allow a stereospecific control of glycosylation: 6 [12] must be separated from its anomer since the output of the reaction is an α-and β-mixture. Finally, and importantly, the authors did not fully characterize the natural disaccharide sambubiose [10]. Hence, the aim of this investigation was to develop an improved efficient strategy to synthesize our target molecule overcoming the abovementioned drawbacks. The key synthetic step (i.e., the Oglycosylation between the appropriate protected glucose and activated/protected xylose) must be thoroughly explored. In particular, the leaving group of the donor should be screened, and the correct catalyst/promoter should be fine-tuned.
Herein, as a part of our ongoing investigations of disaccharide chemistry [13][14][15][16], we describe an efficient synthetic method that allowed us to obtain sambubiose in a satisfactory and higher overall Firstly, the purification of benzyl 3,5,6-tri-O-benzyl-α-D-glucofuranoside (4a) to eliminate high-boiling benzyl alcohol from the reaction crude mixture was extremely difficult both by distillation and/or by chromatography.
Erbing and Lindberg [10], therefore, had to employ acetylation and deacetylation as additional synthetic steps via benzyl 2-O-acetyl-3,5,6-tri-O-benzyl-α-D-glucofuranoside (3a) to purify 4a (Scheme 1, "iii" and "iv" steps). Secondly, the use of the intrinsically unstable donor 2,3,4-tri-O-acetyl-α-D-xylopyranosyl bromide (5a), which must be freshly prepared and recrystallized because of its low stability, even when refrigerated, was also problematic. Moreover, Erbing and Lindberg [10] used a stoichiometric amount of mercuric cyanide as a glycosylation promoter. This reactant should not be used because of its high toxicity and dangerousness. Furthermore, Hg(CN) 2 does not allow a stereospecific control of glycosylation: 6 [12] must be separated from its anomer since the output of the reaction is an αand β-mixture. Finally, and importantly, the authors did not fully characterize the natural disaccharide sambubiose [10]. Hence, the aim of this investigation was to develop an improved efficient strategy to synthesize our target molecule overcoming the abovementioned drawbacks. The key synthetic step (i.e., the O-glycosylation between the appropriate protected glucose and activated/protected xylose) must be thoroughly explored. In particular, the leaving group of the donor should be screened, and the correct catalyst/promoter should be fine-tuned. Herein, as a part of our ongoing investigations of disaccharide chemistry [13][14][15][16], we describe an efficient synthetic method that allowed us to obtain sambubiose in a satisfactory and higher overall yield than previously reported [10] as well as to achieve its full characterization. The proposed methodology ensured an effective stereospecific glycosylation control and exploited strategies of protection and deprotection of functional groups which also guide the regioselective reaction progress.

Results and Discussion
The study started with the optimization of the key step, consisting of the O-glycosylation reaction between the appropriately protected glucose (acceptor) and protected/activated xylose (donor) (Scheme 1, step "v"). In particular, our attention was first focused on the synthesis of several glucose acceptors and xylose donors to overcome the abovementioned problems [10]. With regard to the glucose acceptor, the synthesis of compounds 9-12 ( Figure 2) was designed and explored aiming to avoid the acetylation and deacetylation steps necessary for the purification of 4a (Scheme 1).
Pharmaceuticals 2020, 13, x FOR PEER REVIEW 3 of 12 yield than previously reported [10] as well as to achieve its full characterization. The proposed methodology ensured an effective stereospecific glycosylation control and exploited strategies of protection and deprotection of functional groups which also guide the regioselective reaction progress.
With regard to the donor glycoside, only compounds that are more stable than 5a (Scheme 1), such as fluoride [18] and trichloroacetimidate [19] derivatives, two of the most appealing and promising glycoside donors for O-glycosylation, were taken into account. Hence,  With regard to trichloroacetimidate 14a [22], this donor type showed some advantages such as a relatively high stability, easy purification, and, most importantly, a potential stereochemical control of the glycosylation reaction. In addition, 14a was synthesized by treating 17 with trichloroacetonitrile (CCl3CN) and 1,8-diazabicyclo [5.4.0]undec-7-ene (DBU) obtaining only αanomer [23]. As previously reported, DBU promoted thermodynamic control of the reaction, which led to a more stable α-anomer 14a (Scheme 3). One-pot attempts to obtain 14a from 16a were also explored; however, such attempts proved not to be advantageous due to the fact of their low yields (data not shown).
Finally, the key synthetic step, consisting in the O-glycosylation reaction, was explored. The initial substitution of mercuric salts [10] with silver oxide or silver carbonate as an activator was ineffective any advantage because the yields in these conditions fell drastically (data not shown). The use of safer, greener catalysts such as boron trifluoride diethyl etherate (BF3 . OEt2) or trimethylsilyl trifluoromethanesulfonate (TMSOTf, particularly suitable for the glycoside donor activation bearing the trichloroacetoimidate group as the leaving group), warrant further investigation as alternatives to mercury compounds. Dichloromethane was used as a non-polar solvent; therefore, the stereoselection of glycosylation was led by the 2-O-acyl vicinal group on the donor (anchimeric assistance). All the prepared glycoside donors and acceptors were reacted in different conditions in the presence of powdered molecular sieves (MS) 4Å (Scheme 4) at −20 °C, and the main results are reported in Table 1.  In the case of fluoride 13 [20], the desired hemiacetal intermediate 2,3,4-tri-O-acetyl-D-xylopyranoside (17, Scheme 3) was prepared starting from D-(+)-xylose (15) through an initial acetylation (Ac 2 O, Py, 0 • C, 6 h, 99%) followed by monohydrolysis of the corresponding D-xylopyranose 1,2,3,4-tetracetate (16a) [21] with ammonium acetate (DMF dry, rt, 22 h, 70%). Glycoside donor 13 was then obtained by fluorination of 17 with diethylaminosulfur trifluoride (DAST) (CH 2 Cl 2 , −30 • C to rt, 1 h, 98%) as a mixture of anomers (α/β 1:4), which could be easily isolated by a flash chromatography.
With regard to trichloroacetimidate 14a [22], this donor type showed some advantages such as a relatively high stability, easy purification, and, most importantly, a potential stereochemical control of the glycosylation reaction. In addition, 14a was synthesized by treating 17 with trichloroacetonitrile (CCl 3 CN) and 1,8-diazabicyclo [5.4.0]undec-7-ene (DBU) obtaining only α-anomer [23]. As previously reported, DBU promoted thermodynamic control of the reaction, which led to a more stable α-anomer 14a (Scheme 3). One-pot attempts to obtain 14a from 16a were also explored; however, such attempts proved not to be advantageous due to the fact of their low yields (data not shown).
Finally, the key synthetic step, consisting in the O-glycosylation reaction, was explored. The initial substitution of mercuric salts [10] with silver oxide or silver carbonate as an activator was ineffective any advantage because the yields in these conditions fell drastically (data not shown). The use of safer, greener catalysts such as boron trifluoride diethyl etherate (BF 3 . OEt 2 ) or trimethylsilyl trifluoromethanesulfonate (TMSOTf, particularly suitable for the glycoside donor activation bearing the trichloroacetoimidate group as the leaving group), warrant further investigation as alternatives to mercury compounds. Dichloromethane was used as a non-polar solvent; therefore, the stereoselection of glycosylation was led by the 2-O-acyl vicinal group on the donor (anchimeric assistance). All the prepared glycoside donors and acceptors were reacted in different conditions in the presence of powdered molecular sieves (MS) 4Å (Scheme 4) at −20 • C, and the main results are reported in Table 1.
Firstly, the reactivity of fluoride glycoside 13b was explored with the methoxy derivative 9b or 9a as a glycoside acceptor, in the presence of BF 3 . OEt 2 at −20 • C to room temperature for 4 h.
The reaction gave the corresponding βor α-disaccharide in 80% or 50% yields, respectively, and a complete trans stereocontrol (entries 1 and 2). A complete recovery of the acceptor (9b or 9a) was pointed out in both reactions. The lower yield for glycosylation obtained using the acetate 9a could be attributed to the steric hindrance of the methoxy group present in the same part of the plane of the sugar of the reactive 2-OH group. Similarly, the allyloxy acceptor 12b gave the corresponding coupling product in 30% yield, and its anomer 12a did not react at all and was completely recovered (entries 6 and 3, respectively). The absence of reactivity of 12a could be explained by the greater steric hindrance of its 1-allyloxy substituent compared to that of 1-methoxy present on 9a (entry 3 versus 2). On the other hand, the α-anomer fluoride donor 13a did not react with either highly reactive acceptor 9b or 12b in the same conditions (entries 4 and Pharmaceuticals 2020, 13, 198 5 of 13 5, respectively). The last donor to be investigated was trichloroacetimidate 14a, which was reacted with 9b (entry 7) in the presence of BF 3 . OEt 2 , giving the corresponding β-linked disaccharide in 44% yield or with 12b (entry 8) using TMSOTf as a promoter and obtaining a better yield (60%), complete stereocontrol, and high reproducibility. Surprisingly, the use of the latter conditions (donor 14a and promoter TMSOTf) and very low reactive donor 12a allowed coupling with the formation of the corresponding disaccharide; however, with a lower yield (45%) and longer reaction time than anomer 12b (entry 9). Hence, with regard to O-glycosylation, the best overall result in terms of yield, stereocontrol, and reproducibility was obtained by reacting donor 14a and acceptor 12b promoted by TMSOTf.
of the glycosylation reaction. In addition, 14a was synthesized by treating 17 with trichloroacetonitrile (CCl3CN) and 1,8-diazabicyclo [5.4.0]undec-7-ene (DBU) obtaining only αanomer [23]. As previously reported, DBU promoted thermodynamic control of the reaction, which led to a more stable α-anomer 14a (Scheme 3). One-pot attempts to obtain 14a from 16a were also explored; however, such attempts proved not to be advantageous due to the fact of their low yields (data not shown). Finally, the key synthetic step, consisting in the O-glycosylation reaction, was explored. The initial substitution of mercuric salts [10] with silver oxide or silver carbonate as an activator was ineffective any advantage because the yields in these conditions fell drastically (data not shown). The use of safer, greener catalysts such as boron trifluoride diethyl etherate (BF3 . OEt2) or trimethylsilyl trifluoromethanesulfonate (TMSOTf, particularly suitable for the glycoside donor activation bearing the trichloroacetoimidate group as the leaving group), warrant further investigation as alternatives to mercury compounds. Dichloromethane was used as a non-polar solvent; therefore, the stereoselection of glycosylation was led by the 2-O-acyl vicinal group on the donor (anchimeric assistance). All the prepared glycoside donors and acceptors were reacted in different conditions in the presence of powdered molecular sieves (MS) 4Å (Scheme 4) at −20 °C, and the main results are reported in Table 1.  Firstly, the reactivity of fluoride glycoside 13b was explored with the methoxy derivative 9b or 9a as a glycoside acceptor, in the presence of BF3 . OEt2 at −20 °C to room temperature for 4 h. The reaction gave the corresponding β-or α-disaccharide in 80% or 50% yields, respectively, and a complete trans stereocontrol (entries 1 and 2). A complete recovery of the acceptor (9b or 9a) was pointed out in both reactions. The lower yield for glycosylation obtained using the acetate 9a could be attributed to the steric hindrance of the methoxy group present in the same part of the plane of the sugar of the reactive 2-OH group. Similarly, the allyloxy acceptor 12b gave the corresponding coupling product in 30% yield, and its Firstly, the reactivity of fluoride glycoside 13b was explored with the methoxy derivative 9b or 9a as a glycoside acceptor, in the presence of BF3 . OEt2 at −20 °C to room temperature for 4 h. The reaction gave the corresponding β-or α-disaccharide in 80% or 50% yields, respectively, and a complete trans stereocontrol (entries 1 and 2). A complete recovery of the acceptor (9b or 9a) was pointed out in both reactions. The lower yield for glycosylation obtained using the acetate 9a could be attributed to the steric hindrance of the methoxy group present in the same part of the plane of the sugar of the reactive 2-OH group. Similarly, the allyloxy acceptor 12b gave the corresponding coupling product in 30% yield, and its Firstly, the reactivity of fluoride glycoside 13b was explored with the methoxy derivative 9b or 9a as a glycoside acceptor, in the presence of BF3 . OEt2 at −20 °C to room temperature for 4 h. The reaction gave the corresponding β-or α-disaccharide in 80% or 50% yields, respectively, and a complete trans stereocontrol (entries 1 and 2). A complete recovery of the acceptor (9b or 9a) was pointed out in both reactions. The lower yield for glycosylation obtained using the acetate 9a could be attributed to the steric hindrance of the methoxy group present in the same part of the plane of the sugar of the reactive 2-OH group. Similarly, the allyloxy acceptor 12b gave the corresponding coupling product in 30% yield, and its anomer 12a did not react at all and was completely recovered (entries 6 and 3, respectively). The Firstly, the reactivity of fluoride glycoside 13b was explored with the methoxy derivative 9b or 9a as a glycoside acceptor, in the presence of BF3 . OEt2 at −20 °C to room temperature for 4 h. The reaction gave the corresponding β-or α-disaccharide in 80% or 50% yields, respectively, and a complete trans stereocontrol (entries 1 and 2). A complete recovery of the acceptor (9b or 9a) was pointed out in both reactions. The lower yield for glycosylation obtained using the acetate 9a could be attributed to the steric hindrance of the methoxy group present in the same part of the plane of the sugar of the reactive 2-OH group. Similarly, the allyloxy acceptor 12b gave the corresponding coupling product in 30% yield, and its anomer 12a did not react at all and was completely recovered (entries 6 and 3, respectively). The Firstly, the reactivity of fluoride glycoside 13b was explored with the methoxy derivative 9b or 9a as a glycoside acceptor, in the presence of BF3 . OEt2 at −20 °C to room temperature for 4 h. The reaction gave the corresponding β-or α-disaccharide in 80% or 50% yields, respectively, and a complete trans stereocontrol (entries 1 and 2). A complete recovery of the acceptor (9b or 9a) was pointed out in both reactions. The lower yield for glycosylation obtained using the acetate 9a could be attributed to the steric hindrance of the methoxy group present in the same part of the plane of the sugar of the reactive 2-OH group. Similarly, the allyloxy acceptor 12b gave the corresponding coupling product in 30% yield, and its anomer 12a did not react at all and was completely recovered (entries 6 and 3, respectively). The absence of reactivity of 12a could be explained by the greater steric hindrance of its 1-allyloxy 9b BF 3 . OEt 2 6 18b -

13a
Pharmaceuticals 2020, 13, x FOR PEER REVIEW 5 of 13 Firstly, the reactivity of fluoride glycoside 13b was explored with the methoxy derivative 9b or 9a as a glycoside acceptor, in the presence of BF3 . OEt2 at −20 °C to room temperature for 4 h. The reaction gave the corresponding β-or α-disaccharide in 80% or 50% yields, respectively, and a complete trans stereocontrol (entries 1 and 2). A complete recovery of the acceptor (9b or 9a) was pointed out in both reactions. The lower yield for glycosylation obtained using the acetate 9a could be attributed to the steric hindrance of the methoxy group present in the same part of the plane of the sugar of the reactive 2-OH group. Similarly, the allyloxy acceptor 12b gave the corresponding coupling product in 30% yield, and its anomer 12a did not react at all and was completely recovered (entries 6 and 3, respectively). The absence of reactivity of 12a could be explained by the greater steric hindrance of its 1-allyloxy substituent compared to that of 1-methoxy present on 9a (entry 3 versus 2). On the other hand, the α- Firstly, the reactivity of fluoride glycoside 13b was explored with the methoxy derivative 9b or 9a as a glycoside acceptor, in the presence of BF3 . OEt2 at −20 °C to room temperature for 4 h. The reaction gave the corresponding β-or α-disaccharide in 80% or 50% yields, respectively, and a complete trans stereocontrol (entries 1 and 2). A complete recovery of the acceptor (9b or 9a) was pointed out in both reactions. The lower yield for glycosylation obtained using the acetate 9a could be attributed to the steric hindrance of the methoxy group present in the same part of the plane of the sugar of the reactive 2-OH group. Similarly, the allyloxy acceptor 12b gave the corresponding coupling product in 30% yield, and its anomer 12a did not react at all and was completely recovered (entries 6 and 3, respectively). The absence of reactivity of 12a could be explained by the greater steric hindrance of its 1-allyloxy substituent compared to that of 1-methoxy present on 9a (entry 3 versus 2). On the other hand, the α- O-Methoxy group deprotection in the disaccharides 18a and 18b were first investigated using strong acids such as hydrochloride acid (HCl, rt or −15 • C, 10 h) or BF 3 . OEt 2 (0 • C, 10 h). Unfortunately, demethylation was not obtained in any of these cases, and degradation of the starting material or no reaction was observed. Also treating disaccharides with trityl tetrafluoroborate [(C 6 H 5 ) 3 C . BF 4 ] [24] (rt or −15 • C, 4-12 h), usually employed in the deprotection of benzyl and methyl ethers on anomeric carbons, was found to be detrimental (data not shown). Hence, 18b was also first debenzylated by hydrogenolysis (10% Pd/C, EtOAc/EtOH 1:2, 32%) leading to the derivative β-methyl 2-O-(2,3,4-tri-O-acetyl-β-D-xylopyranosyl)-D-glucofuranoside (20b) which, in turn, was deacetylated (NaOMe, MeOH, 7 h, 70%) to give β-methyl 2-O-β-D-xylopyranosyl-D-glucofuranoside (21b). Attempts to obtain sambubiose by deprotection of 20b and 21b by cleaving the methoxy group also failed (data not shown).

Chemicals, Materials, and Methods
All reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA), Alfa Aesar (Haverhill, MA, USA), or TCI (Tokyo, Japan) in the highest quality commercially available. The solvents were RP grade. The melting points were determined using the Büchi (Flawil, Switzerland) B-540 apparatus. The purity of the products was evaluated unequivocally through MS, 1 H NMR, 13 C NMR, IR, and [α] 20 D. The MS (ESI) spectra were recorded with a Waters (Milford, MA, USA) Micromass ZQ spectrometer in a positive mode using a nebulizing nitrogen gas at 400 L/min and a temperature of 250 °C, cone flow 40 mL/min, capillary 3.5 K volts, and cone voltage 60 V; the revelation was performed either in ESI (+) and/or ESI (−), from 100 to 800 units of mass, spectrophotometrically measured using a diode array spectrophotometer. The 1 H NMR and 13 C NMR spectra were recorded Synthesis of 1,2,3,4-tetra-O-acetyl-α-D-xylopyranose (16a). To a solution of 15 (1 g, 6.66 mmol) in pyridine (5 g, 4.5 mL, 62 mmol), Ac 2 O was added (5.5 g, 4.5 mL, 52.74 mmol). The mixture was stirred at 0 • C for 6 h, extracted with Et 2 O and washed with H 2 O and CuSO 4 saturated solution. The combined organic layers were dried with Na 2 SO 4 , filtered and concentrated. Purification of the residue by column chromatography (cyclohexane/EtOAc 9:1) gave 16a as a yellow oil which was dried under N 2 atmosphere because of its thermolability. Yield: 75% (1.59 g, 5 mmol  13 C NMR, and IR data are in agreement with those reported in the literature [21]. Synthesis of 2,3,4-tri-O-acetyl-D-xylopyranose (17). To a solution of 16a (1 g, 3.14 mmol) in dry DMF (9 mL), AcONH 4 (0.485 g, 6.29 mmol) was added and the mixture was stirred at room temperature for 22 h, extracted with EtOAc, and washed with H 2 O and saturated aq. ppm. 13