Gold Nanoparticles Decorated with Mannose-6-phosphate Analogues

Herein, the preparation of neoglycoconjugates bearing mannose-6-phosphate analogues is described by: (a) synthesis of a cyclic sulfate precursor to access the carbohydrate head-group by nucleophilic displacement with an appropriate nucleophile; (b) introduction of spacers on the mannose-6-phosphate analogues via Huisgen’s cycloaddition, the Julia reaction, or the thiol-ene reaction under ultrasound activation. With the resulting compounds in hand, gold nanoparticles could be functionalized with various carbohydrate derivatives (glycoconjugates) and then tested for angiogenic activity. It was observed that the length and flexibility of the spacer separating the sugar analogue from the nanoparticle have little influence on the biological response. One particular nanoparticle system substantially inhibits blood vessel growth in contrast to activation by the corresponding monomeric glycoconjugate, thereby demonstrating the importance of multivalency in angiogenic activity.


Introduction
The ensuing article was initially motivated by the biomedical importance of mannose 6-phosphate (M6P) [1]. While two different mannose 6-phosphate receptors (M6PR) recognize the M6P residues and mediate the endocytosis of extracellular M6P-containing ligands, only the larger of these OPEN ACCESS (CI-M6PR, 275 kDa) has been reported to also bind retinoic acid and IGF-II [2]. The biological importance of this receptor is found in numerous processes and it has been reported that the angiogenic action of proliferin was mediated by this receptor [3]. We have recently described the synthesis of a series of mannose-6-phosphate (M6P) analogues, showing for the first time that these monosaccharides play a role in angiogenesis [4][5][6][7]. The replacement of the phosphate head-group by analogues, mostly bioisosteres, was intended to provide a better understanding of the chemical factors involved in the modulation of angiogenic activities. It is known, however, that a monovalent carbohydrate ligand possesses only a weak binding affinity toward its associated receptor protein [8][9][10]. To impart biological relevance to such interactions Nature often utilizes multivalency [11,12]. Therefore, interest in designing multivalent carbohydrate systems has been growing [13]. In particular, glyconanoparticles (GNPs), that offer useful tools for investigating carbohydrate-mediated interactions, have been developed [14]. The purpose of the present study was: (a) to synthesize new glycoconjugates bearing M6P-like groups and (b) to insert these compounds onto the surfaces of gold particles via a spacer for angiogenic testing. Our objective, therefore, is to investigate the effect of clustered sugar derivatives on angiogenesis and to determine whether or not the spacer has an influence on the biological response. The choice of the M6P analogues has been guided by previous results conducted in our laboratory including the synthesis of carboxylate and azido analogues (with 123% and 125% relative angiogenic activity, respectively, compared to phosphate buffer saline (PBS) as control in an egg membrane assay) [4]. Additional considerations include varying the length, hydrophilic or hydrophobic nature, and flexibility of the spacer between the sugar headgroups and the nanoparticle core. In this manner we could modify the presentation of the carbohydrates and, consequently, affect their accessibility during the molecular recognition events. Many of the mannose derivatives with their different spacers were assembled using the "click chemistry" strategy introduced by Huisgen and improved by Sharpless and co-workers in 2001 [15,16]. Within a short time-frame, the click chemistry reaction has proven to be of remarkable utility and broad scope, not only in organic synthesis but in chemical biology and drug discovery [17,18]. Although 1,3-dipo lar cycloaddition reaction is central to click chemistry, the resulting creation of a triazole moiety may have an adverse influence on a biological response. For this reason two other reactions were used for chain elongation or for conjugation of two synthons: the Julia reaction and the thiol-ene reaction that was run under unprecedented ultrasound activation.

Results and Discussion
The preparation of the neoglycoconjugates we describe herein took place in three major steps: (a) the synthesis of a cyclic sulfate precursor to access the ligand head-group by nucleophilic displacement with the appropriate nucleophile; (b) the introduction of the spacers on the M6P via either Huisgen's cycloaddition, the Julia reaction, or the thiol-ene reaction under ultrasound activation; (c) the coupling between the spacer and the sugar moiety. By this means gold nanoparticles as functionalized by various carbohydrates could be compared for their effect on angiogenic processes. Although preliminary biological data are presented at the end of the paper, the emphasis here will be on the synthetic challenges involved in obtaining the necessary neoglycoconjugates.

1,3Dipolar Cycloaddition
Huisgen's 1,3-dipolar cycloaddition is the primary example of a "click reaction". It is the reaction between a 1,3-dipole (an azide) and a dipolarophile (an acetylene) to form a five-membered heterocycle. The classical reaction proceeds by a concerted mechanism under thermal conditions to afford a mixture of 1,4-and 1,5-disubstituted [1,2,3]-triazole regioisomers [19], but when the reaction is catalyzed by Cu(I), only the 1,4-substituted-triazole is obtained [20]. We selected this reaction as one means for securing our mannose-6-phosphate analogues. Thus to prepare the nanoparticles, the carbohydrate moiety had to bear either an azide or alkyne function. The linker chain, in turn, would provide the complementary group. The cyclic sulfate strategy, utilized in our laboratory to prepare M6P analogues, demanded that the carbohydrate possess the azide group because an alkyne function would become oxidized during the preparation of the sulfate. Thus, peracetylated mannose has been coupled in very good yield to 2-bromoethanol, under classical conditions [21], in the presence of boron trifluoride etherate (Scheme 1). The azide group was then introduced with sodium azide, and the acetate protecting groups were removed under Zemplen conditions [22] to give the 2-azidoethyl-α-D-mannopyranoside 3. After selectively introducing isopropylidene protection at the 2 and 3 positions of the mannose, the cyclic sulfate 5 was prepared according to a modified published procedure [23,24]. Compound 4 was converted via thionyl chloride into the cyclic sulfite which was then oxidized by ruthenium oxide (prepared in situ) into the corresponding cyclic sulfate 5.
A "spacer" refers to a chain that can be used to join our sugar derivatives to the gold particles. One of the spacers, possessing an alkyne unit for reaction with a sugar-azide, was designed on the basis of its flexibility and aqueous solubility (Scheme 2). Thus, the reaction of 5-bromopentene with a slight excess of 50% sodium hydroxide and hexaethylene glycol provided the monoether 6 [25]. Photochemical addition of thioacetic acid to the double bond gave the thioacetate in good yield [26]. The next step was to introduce the alkyne function on the spacer in the presence of NaH, but the acetate protecting groups, being sensitive to hydrides, were first replaced by 4-methoxytrityl. Thus, compound 7 was deacetylated by concentrated hydrochloric acid in ethanol to avoid the formation of disulfide under basic conditions. The thiol was then protected by reaction with 4-methoxytrityl chloride. Finally, the free hydroxyl of 8 was reacted with 3-bromopropyne in the presence of sodium hydride in anhydrous THF to introduce the alkyne function required for the click reaction. Scheme 2. Preparation of the alkyne 9 for the click reaction.
The literature describes a variety of ways in which the Huisgen cycloaddition can be performed to join two entities. Sources of copper (I) catalyst can be produced in situ by reduction of copper (II) salts [20] or obtained through disproportionation of Cu (0) and Cu (II) salts [27]. Cu (I) can also be introduced as copper (I) salts such as CuI or obtained from oxidation of Cu (0) salt [28][29][30][31]. In search for the optimal reaction conditions, we initially tested the most commonly employed system, namely CuSO 4 ·5H 2 O and sodium ascorbate as source of copper (I) in tert-BuOH/H 2 O [32]. Interestingly, no reaction was observed after 24 h. In addition, several parameters were altered without success: increasing the concentration of reactants, changing the ratio copper/sodium ascorbate, using a co-solvent (acetonitrile), or substituting tBuOH with pyridine. The click reaction was also attempted using cuprous iodide in pyridine as catalyst [33]. Despite many modifications to the original protocol the desired product was never obtained in good yield. Thus, another copper catalyst system consisting of formation of Cu(I) by oxidation of copper metal was investigated. The oxidative cycloaddition of Cu(0) with ammonium chloride [34] in a mixture of tert-BuOH/H 2 O was also unsuccessful. It should be noted that heating to 40-60 °C, and increasing the reagents'concentration, failed to improve the performance, as they often do in many examples of click chemistry reactions, but led only to degradation. Ultrasound in place of classical activation was carried out again without success. Only a system using copper powder, rarely encountered in the literature, gave positive results, giving compound 10 in 60% yield (Scheme 3). Starting from compound 10 two mannose-6-phosphate analogues were prepared with only slight modification to the previously reported protocols [4]. First, the azide function was easily introduced on the cyclic sulfate 10 by reaction with sodium azide to afford compound 11. Although isopropylidene and trityl are usually deprotected under acid conditions, our assays did not allow simultaneous cleavage of the two functions. The final ligand 12 was therefore obtained in two separate steps. The trityl group was first cleaved by ceric ammonium nitrate (a redox reaction) [35,36] prior to deprotection of the isopropylidene and the sulfate groups via acidic ion exchange resins. To afford the carboxylic acid analogue of M6P 14, sodium cyanide was first reacted with the cyclic sulfate 10, and the nitrile function was then hydrolyzed with sodium hydroxide in a 30% solution of hydrogen peroxide to give the corresponding carboxylic acid. The ligand 14 was obtained using the same deprotection conditions as described for the azide analogue 12 (Scheme 3).

Julia Reaction
Among the olefination reactions to form a regio-and stereoselective alkene, the Julia olefination is one of the well-known methods, along with the Wittig reaction [37,38], the Wittig-Horner reaction [39][40][41], the Horner-Wadsworth-Emmons [42,43], the Peterson reaction [44][45][46] and the Johnson reaction [47]. The classical Julia olefination, also known as the Julia-Lythgoe olefination, was developed fourty years ago and is based on a reductive elimination process of β-acyloxysulfones [48]. Since its discovery, significant improvements have been made to the methodology of this reaction, and it has become a crucial step in the synthesis of many natural products. A new variant of the classical Julia reaction, the Julia-Kocienski olefination, also called modified or one-pot Julia olefination, has recently emerged as a powerful tool for olefin synthesis [49][50][51]. The process involves the replacement of the aryl sulfone moiety, traditionally used in the classical reaction, with different heteroaryl sulfones, thus allowing a direct olefination process.
In our Julia olefination, a carbohydrate block was derivatized with an allyl bromide function (to be later joined with a sulfone-bearing linker). The initial steps in the sugar portion of the molecule followed the same strategy as described for compound 5 (Scheme 4). The methyl α-D-mannopyranoside was previously protected with two O-isopropylidene groups on the 2,3 and 4,6 positions using 2,2-dimethoxypropane and para-toluenesulfonic acid. After the selective opening of isopropylidene at the 4,6 positions with a mixture of AcOH/H 2 O, the cyclic sulfite was obtained by reaction with thionyl chloride and triethylamine, and subsequent oxidation afforded the cyclic sulfate 15 in good yield. In contrast to the chemistry in Scheme 1 and Scheme 3, the azide and carboxylic acid analogues of M6P were prepared prior to the coupling reaction. Therefore, sodium azide was reacted with the cyclic sulfate 15 to give compound 16. A solution of acetic acid in water led to the cleavage of the isopropylidene and the sulfate. The replacement of the anomeric methyl group by an acetyl group led to compound 17 in 83% yield. The allyl bromide unit required to perform the coupling reaction was then introduced by glycosylation with cis-1-bromo-but-2-en-4-ol. The same strategy was applied to form the carboxylic acid analogue of M6P.  To create the Julia spacer (Scheme 5), geraniol was reacted with phosphorus tribromide to give geranyl bromide 23 in good yield followed by a reaction with sodium phenylsulfinate to provide the desired sulfone 24. Next, functionalization of the other side of the linker was carried out in one step via oxidation of a terminal methyl by selenium dioxide. The strategy described by Sharpless using tert-butyl hydroperoxide as an oxidant [52] was utilized: 24 was reacted with a catalytic amount of selenium oxide in the presence of tert-butyl hydroperoxide which enables the recycling of selenium dioxide. A 50/50 mixture of alcohol and aldehyde was obtained and, after purification, the aldehyde was reduced with sodium borohydride to give compound 26 in 63% yield. The alcohol 26 was then brominated with tetrabromomethane and triphenylphosphine after which the thiol group was introduced by reaction with potassium thioacetate. A "click-type" reaction was then performed between 18 or 22 and 27 in the presence of lithium bis(trimethylsilyl)amide (LiHMDS) in THF from which compounds 28 and 29 were obtained in 15% and 17% yield, respectively. Deprotection of the acetates and removal of the sulfone group under basic conditions gave the desired final compounds 30 and 31 in almost quantitative yield (Scheme 6). The linker in this case is polyunsaturated.

Thiol-ene Reaction
One reaction that is emerging as an attractive click-type process is the century-old addition of thiols to alkenes [53], which is currently called thiol-ene coupling. In fact, the thiol-ene reaction is simply the hydrothiolation of a C=C bond, and proceeds by a radical mechanism, induced photochemical or thermally, to give an anti-Markovnikov-type thioether [54,55]. The reaction discovered in 1905 by Posner [56] has been widely used in the mid-nineteenth century, especially in polymer chemistry. However, the thiol-ene reaction has recently attracted researchers in other areas of synthesis due to recognition of its ''click-type'' characteristics: highly efficient and orthogonal to a wide range of functional groups, as well as compatible with water and oxygen. Thus, the thiol-ene reaction enables the establishment of a rapid ligation between two entities assisted by the stability of the thioether linkage in a wide range of chemical environments. To perform the reaction, the thiol function was placed on the sugar moiety while the spacer carried the vinylic group. As before, the M6P derivatives were prepared using the cyclic sulfate strategy prior to performing the click-style reaction (Scheme 7). The synthesis began by replacing the anomeric acetate with bromine on compounds 17 and 21 described previously. This was accomplished with a solution of hydrobromic acid/acetic acid in quantitative yields. The thiol function was then introduced in two steps, first via thiourea in acetone then removal of the nitrogens with sodium metabisulfide. Only thiosugars (34 and 35) having the β configuration were obtained.  Having chosen to synthesize a fully flexible spacer (Scheme 8), the triethylene glycol was coupled to allyl bromide in the presence of 50% aqueous sodium hydroxide.  Then, to facilitate the reaction, and to avoid formation of byproducts, the free hydroxyl of the linker was brominated and thioacetylated only at the end of the synthesis. Actually, coupling can be accomplished between a protected thiol group and an alkene or between a thiol and an alkene. However, the final thioether was obtained with better results when the anomeric thiol was not protected (Scheme 9). The initiation step can be triggered in several ways, by simple heating or by ultraviolet irradiation. Another method of initiation has been developed in the laboratory which is to perform the coupling under ultrasound (Table 1). When THF was replaced by dioxane, yields increased by 10%.

Gold Nanoparticles
Research in developing new synthesis protocols to generate gold nanoparticles (AuNPs) with desired properties has received immense attention due to their considerable applications in biomedical field [57]. One of the primary prerequisites for using AuNPs in biomedical application is that they are non-toxic and biocompatible to both in vitro and in vivo environments. Secondly, AuNPs should be coated with a protective layer to prevent aggregation. Thirdly, AuNPs need to be labeled with biologically relevant biomolecules to impart specificity for their potential application. The two most interesting and common methods to prepare AuNPs are the Brust method [58] utilizing NaBH 4 (which can't be used in our case because NaBH 4 would reduce the azide function of our derivatives) and the citrate method. This latter method includes only three starting materials, namely, auric acid, sodium citrate (the reducer), and water. Following a report by Turkevich et al. in 1951 [59], this synthetic scheme has been widely studied and often used for the preparation of AuNPs-based materials [60][61][62][63]. We have developed a protocol by adjusting the gold-to-citrate ratio to obtain 10 nm AuNPs (Table 2). Details are given in the Experimental section.

Biological Assays
AuNPs functionalized with M6P analogues have been subjected to angiogenic assays using an experimental model, the avian chorioallantoic membrane assay (CAM) [64][65][66]. Paper discs were saturated with a phosphate buffer saline dispersion of coated AuNPs (60 mg/mL) in PBS or a control (phosphate buffer saline) and then deposited on chorioallantoic membranes of 7-day-old chicken embryos for 4 days in ovo at 38 °C. Sutent ® (sunitinib, a non-proteic inhibitor) and endothelial cell growth supplement (ECGS) were used at 60 mg/mL as negative and positive stimuli, respectively. Quantification of the angiogenic response was carried out by measuring the area of neo-vascularization on each particular membrane ( Figure 1) using Image J software. The experiments have been repeated at least four times, and the results were reproducible (see experimental part for details). These experiments demonstrate that all our prepared AuNPs are CAM-inhibitors. Study of the three azide-AuNPs synthesized according to the coupling methods (NP1: thiol-ene 58%, NP2: Julia 59%, NP3: Huisgen 65%) revealed that the length and flexibility of spacers have little influence on the observed biological response. Interestingly, the azide sugar-monomer is a good angiogenic activator (125%), whereas the functionalized -N 3 nanoparticles, representing a multi-valent collection of sugars, show a strong inhibitory effect (58%-65%). Similar results were obtained for the three carboxylic acid-AuNPs (NP4: Huisgen 86%, NP5: Julia 87%, NP6: thiol-ene 88%). Comparison of the activating effect of the carboxylic acid analogue (123% observed in previous work) [4] and the inhibitory effect of the carboxylic acid-AuNPs (86% to 88% compared to the control) indicates that multi-valency can do more than qualitatively affect the magnitude of blood vessel formation; it can convert a significant catalyzed process into an inhibition.

Preparation of Citrate-Reduced Gold Nanoparticles
Hydrogen tetrachloroaurate trihydrate (60 mg, 0.15 mmol) was dissolved in water (250 mL) to give a pale-yellow solution. A second solution with sodium citrate (150 mg, 0.58 mmol) dissolved in water (10 mL) was prepared. Both solutions were heated to 60 °C for 10 min then the sodium citrate solution was rapidly added to the gold solution. The temperature was then increased to 120 °C with continuous stirring for 2 h 30 min. A deep-red solution was formed. The solution was warmed to RT and each thiol-derivatized carbohydrate (50 mg) dissolved in methanol (1 mL) was added to freshly prepared citrate-reduced gold nanoparticles. Self-assembly was facilitated by leaving the solution under stirring for 48 h. To remove any unbound carbohydrate, the solution was diluted with brine to precipitate nano-objects and left to rest over night. Then the supernatant was removed and nanoparticles were resuspended in water. They were then centrifuged for 30 min at 14,000 rpm. The centrifugation step was repeated three times to ensure complete removal of any unbound carbohydrate.

CAM Assays
Paper discs were saturated with a phosphate buffer saline dispersion of coated AuNPs (60 mg/mL) in PBS or a control (phosphate buffer saline) and then deposited on chorioallantoic membranes of 7-day-old chicken embryos for 4 days in ovo at 38 °C. Sutent ® (sunitinib, a non-proteic inhibitor) and ECGS (endothelial cell growth supplement) were used at 60 mg/mL as negative and positive stimuli, respectively. Quantification of the angiogenic response was carried out by measuring the area of neo-vascularization on each particular membrane ( Figure 2). The vascularization was evaluated using Image J software, and are given in pixels compared to phosphate buffer saline (PBS, control).

Conclusions
We have reported the synthesis of a series of gold glyconanoparticles bearing diverse M6P neoglycoconjugates. These M6P analogues were synthesized either by Huisgen cycloaddition, by the Julia olefination, or by thiol-ene coupling. The thiol-ene reaction strategy under ultrasound activation proved to be the most efficient in terms of yields and ease of implementation. The angiogenic activities of the AuNPs have been tested by the CAM assay, and all possess angiogenic activities via the M6P receptor with no apparent toxicity. The results of this study have allowed us: (a) to demonstrate that the activity is not dependent of the structure of the linker between the nanoparticles and the carbohydrate and (b) to identify the inhibitory multivalent effect of M6P derivatives on gold surfaces compared with the corresponding monomeric activators. Although our biological results are obviously in a preliminary stage, the work described herein is valuable in that it provides synthetic access to some potentially useful multi-functional and biologically active systems.