Glycosyl Sulfonates Beyond Triflates

Abstract

While glycosyl triflates are frequently invoked as intermediates in many chemical glycosylation reactions, the chemistry of other glycosyl sulfonates remains comparatively underexplored. Given the reactivity of sulfonates can span several orders of magnitude, this represents an untapped resource for the development of stereoselective glycosylation reactions. This personal account describes our laboratories efforts to take advantage of this reactivity to develop β-specific glycosylation reactions. Initial investigations led to the development of 2-deoxy-sugar tosylates as highly selective donors for β-glycoside synthesis, an approach which has been used to great success by our group and others for the construction of deoxy-sugar oligosaccharides and natural products. Subsequent studies demonstrate that “matching” the reactivity of the sulfonate to that of the sugar donor leads to highly selective S_N2-glycosylations with a range of substrates.

Graphical Abstract

1. Glycosyl Sulfonates – History

First described by Helferich and Gootz in 1929,^[1] glycosyl sulfonates readily react with a range of nucleophiles to form glycosidic linkages. The chemistry of these species was largely ignored for several decades, until seminal work by Conrad Schuerch in the 1970s and early 1980s on the chemistry of various sulfonates.^[2] While the chemistry yielded very promising results, including early examples of β-mannosylation and β-rhamnosylation,^[2c.d] the need to pre-generate these highly reactive species detracted from their wider adoption. Around this same time Koto and coworkers demonstrated that activating tetra-O-benzyl-glucopyranoside with a combination of AgOTf, 4-nritobenzenesulfonyl chloride (NsCl) and Et₃N in the presence of an alcohol led to the formation of β-linked products almost exclusively.^[3] Although no evidence for the intermediacy of a glycosyl nosylate was provided, additional experiments where the NsCl was replaced with methanesulfonyl chloride (MsCl) led to lower yields and mixtures of anomers, indicating that the nature of the sulfonate played a role in the selectivity of the reaction.

The use of glycosyl sulfonates gained much greater attention in the late 1990s when Crich and co-workers demonstrated that torsionally disarmed mannosyl triflates, generated in situ from the corresponding thioglycoside, underwent highly selective β-mannosylation reactions.^[4] Extensive mechanistic studies revealed that these glycosyl triflates reacted through an S_N2-like manifold.^[5] The use of the torsionally disarming benzylidene acetal is critical to selectivity in this reaction,^[6] and selectivity is less predictable with other donors, especially in cases where it is not possible to take advantage of neighboring group participation to control selectivity.^[7] For the most part, the use of other glycosyl sulfonates in the reaction had been mostly underexplored until recently, with the exception of elegant work form the Gin lab, who demonstrated that mannosyl benzenesulfonates underwent non-selective reactions with neutral alcohol acceptors.^[8] Over the past decade, however, our group and others have begun to re-examine the reactivity of other glycosyl sulfonates. These studies have shown that in the presence of activated alcohol nucleophilic acceptors, these glycosyl sulfonates can undergo S_N2-like reactions. These investigations continue to provide evidence that the S_N2-like nature of glycosylations using glycosyl sulfonate donors is more general than previously appreciated, especially if the electronics of the sulfonate leaving group are tailored to match the reactivity of the glycosyl donor in question.

2. 2-Deoxysugar Tosylates as Glycosyl Donors

Our initial foray into glycosyl sulfonate chemistry began in 2012, in an attempt to develop a selective approach for the synthesis of β-linked 2-deoxysugars. Our interest in these molecules was stimulated by the fact that they are common motifs in many bioactive natural products, however, the direct synthesis of these molecules was hampered by the high reactivity of the corresponding donors.^[9,10] This typically led to non-selective reactions, unless the donors were pre-modified with a temporary directing group at the C-2 position.^[11] In an effort to overcome this need for temporary prosthetic groups, we reasoned that a less reactive leaving group at the anomeric position of the activated sugar donor intermediate would possess more covalent character then a triflate, which could favor S_N2 pathways. After several failed attempts to generate the glycosyl tosylate from a variety of donors, we turned our attention to direct activation of a hemiacetal. Previously, the Shair lab had shown that metalating a 2-deoxysugar with KHMDS followed by treatment with allyl bromide, led to the formation of the allyl glycoside with extremely high levels of α-selectivity.^[12] By analogy, we reasoned that treating the metalated alkoxide with an electrophilic sulfonate source would lead to the in situ formation of a glycosyl sulfonate that could be trapped with another nucleophile to generate β-linked 2-deoxysugar products.

For our initial forays into this chemistry we examined N-tosyl imidazole as the sulfonate source (Scheme 1a).^[13,14] Our selection of this “promoter” was based on two simple reasons. First this compound is readily available, and the reagent could easily be prepared on large scale. Furthermore, we initially selected to sulfonyl imidazole reagent since we were concerned that the use of sulfonic anhydrides or sulfonyl chlorides would lead to either non-selective reactions through a Curtin-Hammett scenario, or the generation of glycosyl chlorides, which would be unreactive under our conditions.^[15] Initial investigations with thiophenol as a nucleophile quickly demonstrated that this was a serendipitous choice of reagent, and we were able to generate the desired product with moderate levels of β:α selectivity. Recognizing that this may be due to the sluggish nature of the activating agent, we next examined the use of tosyl 4-nitroimidazole in the reaction, and pleasingly, this latter reagent afforded the desired product in excellent yield as a single β-linked isomer.

Scheme 1.

Preliminary optimization of glycosyl tosylate chemistry.

Attempts to extend this chemistry to other less reactive nucleophiles failed to provide productive reactions. In order to attempt to drive the reaction with less nucleophilic thiols, we turned to pre-generating the thiolate ions using a variety of strong bases. These latter studies demonstrated that KHMDS was a superior metalating agent to other hexamethyldisilazide bases. Further extension of this chemistry to alcohol nucleophiles revealed that under the present conditions a strongly coordinating solvent, such as diglyme, was required in the reaction. Under these optimal conditions we were able to extend these β-selective glycosylation reactions to phenolic nucleophiles. Furthermore, this chemistry could be extended to 2,6-dideoxysugar donors, an important discovery since these species are what are typically encountered in natural products.

Encouraged by these preliminary studies, we next turned our attention to extending the scope of the reaction in include sugar acceptors.^[16] With the less reactive metalated sugar acceptors, we were able to obtain the desired products as β-isomers, albeit in moderate yield (Scheme 1b). A survey of other sulfonate leaving groups in the reaction failed to increase the yield of the reaction. While benzenesulfonyl 4-nitroimidazole led to a modest decrease in yield, the corresponding 4-bromobenzenesulfonate (brosylate) failed to produce any product, foreshadowing the importance of sulfonate selection on the outcome of the reaction.

Further attempts to optimize the reaction under a variety of conditions with the tosyl 4-nitoimidazole failed to improve the yield of the reaction. Having felt that we exhausted our efforts with the reagent, we turned our attention to using the much more reactive p-toluenesulfonic anhydride as a promoter in the reaction. The choice of a more reactive promoter proved to be prudent, and we were able to immediately improve the yield of the reaction to 86%. Having optimized the conditions for primary alcohol acceptors with a 2-deoxysugar, we turned our attention to more hindered acceptors. Pleasingly, these compounds all worked well in the reaction, with both 2-deoxyand 2,6-dideoxysugars, a fact that would open the door to using this chemistry in oligosaccharide synthesis.

Before moving forward with this chemistry, we wanted to establish that the reaction was indeed proceeding through the intermediacy of a glycosyl tosylate. Based off of precedent first from the Schuerch, and later the Crich and Gin labs,^[3,4,8] we used VT-NMR experiments to determine if the glycosyl tosylate was indeed the active species in the reaction. As expected, when we treated hemiacetal with KHMDS followed by Ts₂O at −78°C in THF_d8 with KHMDS we were able to identify a single species with a ¹H NMR resonance at δ6.11 ppm and a ¹³C resonance at δ102.3 ppm, indicative of the quantitative formation of a glycosyl sulfonate (Figure 1). These latter experiments further illustrated the importance of conducting the reaction at low temperature (which we had serendipitously initially adapted from the Shair protocol) as the tosylate readily decomposed above −60°C in a sealed tube under argon.

Figure 1.

HSQC spectrum confirming the existence of the glycosyl tosylate.

Having established that this chemistry was able to activate olivose-derived hemiacetals for β-specific disaccharide synthesis (such as 5, Scheme 2), we wanted to both extend the scope of these reactions and demonstrate that it could be used in total synthesis. Before doing so, however, we had to address two issues with the Ts₂O promoter system. Specifically, the reagent itself is a hydrolytically unstable solid, which could lead to issues with batch-to-batch reproducibility. Furthermore, in anticipation of the fact that the tosylate would not provide a one-size-fits-all solution to the problem of controlling selectivity in glycosylation reaction,^[17] we felt that the lack of commercial availability of other sulfonic anhydrides would present a hurdle to others who may want to adopt the chemistry.

Scheme 2.

Glycosyl tosylate chemistry in the synthesis of FD-594. TTBP=2,4,6 tri-tert-butylpyrimidine.

In our search for a more general reagent, we decided to examine the much more common sulfonyl chlorides in the reaction. These reagents were appealing to us both from the standpoint of cost and stability, however, we were aware of the possibility that the chloride ion byproduct of hemiacetal activation could react with the glycosyl tosylate intermediate to generate a fairly unreactive glycosyl chloride. Nevertheless, we moved forward with these studies, and found that TsCl was not only as effective a promoter as the Ts₂O, but the reaction with the former reagent could be run on gram scale.^[18] The ability to reliably scale up the reaction allowed us to begin to explore the scope of the chemistry in more complex oligosaccharide synthesis. As proof of principle, we used the chemistry to synthesize the trisaccharide of FD-594 and the tetrasaccharide fragment of kigamicin E (Scheme 2).

During the course of these latter studies, we made an observation that the stereochemical outcome of reactions using 2,3,6-trideoxy sugars was entirely dependent on the protecting group on the C4 hydroxyl group. Specifically, amicetose donors bearing a 2-naphthylmethyl protecting group at the C-4 position underwent glycosylation reactions to afford predominantly β-linked products, however, when the corresponding C4 acetate protected donor was used the reaction was moderately α-selective. Reasoning this result may be due to the instability of the sulfonate intermediate, we next examined the use of the more electron-rich 2,4,6-triisopropylbenzenesulfonyl chloride as a promoter in the reaction. Much to our surprise, this latter reagent led to an α-specific reaction when amicetose donor 11 was used as a donor in the reaction (Scheme 3). This was our first concrete evidence that the electronics of the sulfonate played a crucial role in the reaction. Nevertheless, the selectivity opened up the possibility of applying this chemistry to the construction of much more complex molecules than we had previously anticipated.

Scheme 3.

α-Selective glycosylations with trideoxy sugars. TrisylCl=2,4,6-triisopropylbenzenesulfonyl chloride.

3. Applications to Deoxy Sugar Oligosaccharide Synthesis

Having established conditions for β-selective olivose construction and α-selective trideoxy-sugar synthesis, we sought targets that would allow us to determine the utility of this chemistry in a more complex setting. To this end we were drawn to the hexasaccharide from landomycin A^[19] and the pentasaccharide from saquayamycin Z.^[20] These targets appealed to us for different reasons. In the case of the saquayamycin pentasaccharide, this highly sensitive structure had not been previously synthesized. In contrast, there had been four previously reported syntheses of the landomycin hexasaccharide, as well as one report of the total synthesis of the natural product. Notably, however, all of these approaches afforded the target product in less than 1% overall yield.^[21,22] Thus, this target served as something of a proving ground that would allow us to determine if our chemistry did indeed provide an advantage over existing methods.

We envisioned that the saquayamycin pentasaccharide could be obtained through a [2 + 2 + 1] approach where the non-reducing rhodinose 14 would be appended to a tetrasaccharide at the end of the synthesis (Scheme 4). This tetrasaccharide could in turn arise from a [2 + 2] reaction using intermediates derived from 15.^[23]

Scheme 4.

Retrosynthetic analysis of the saquayamycin z pentasaccharide. PMP = 4-methoxyphenol.

In the forward direction, coupling of olivose donor 16 with rhodinose acceptor 17 proceeded smoothly to afford the required disaccharide in excellent selectivity. Notably, despite the basic nature of the reaction, we found that the non-nucleophilic proton scavenger 2,4,6-tri-tert-butylpyrimidine was needed for the reaction to proceed with good levels of selectivity. The underlying basis for the need for this reagent is currently unclear. Moving forward, disaccharide 15 could be converted into donor 18 or acceptor 19 through treatment with DDQ and β-pinene or CAN, respectively.

With both donor and acceptor in hand we subjected them to our previously reported optimized conditions for α-selective glycosylation with trideoxy sugars. Surprisingly, selectivity in the reaction was fairly modest, affording the desired product as a 5:1 α:β mixture of isomers (Scheme 5). Further investigations confirmed the need for the C-4 acetate protecting group on rhodinose to obtain selectivity, in line with our previous observations with amicetose. Indeed, once the two tetrasaccharide isomers were separated and the desired anomer deprotected, the final glycosylation with acetate protected rhodinose 14 proceeded to provide the product as a single α-isomer. Finally, global deprotection through a two-step sequence afforded the target compound in 2.5% overall yield from commercial material.

Scheme 5.

Synthesis of 13 using sulfonate-mediated dehydrative glycosylation.

Concurrent with these studies, we also studied the application of our sulfonate-mediated glycosylation chemistry to the synthesis of the landomycin A hexasaccharide (Scheme 6). We envisioned that this target could be obtained through a [3 + 3] approach where the two coupling partners were derived from the same trisaccharide precursor.^[24] In the forward direction the synthesis commenced with the coupling of olivose donor 16 with acceptor 6 to afford disaccharide 24 in 81% yield as a 16:1 (β : α) mixture of anomers. Removal of the Nap protecting group using DDQ and β-pinene as an acid scavenger afforded acceptor 25.^[25] This compound could be easily coupled to rhodinose donor 14 to afford the desired product as a single α-isomer in excellent yield. Interestingly, this in this latter reaction TsCl proved to be a superior promoter to TrisylCl for the α-selective glycosylation, in to contrast to our earlier studies.

Scheme 6.

Sulfonyl chloride mediated synthesis of the landomycin A hexasaccharide.

Trisaccharide 26 could readily be converted to donor 27 and acceptor 28 under standard conditions, setting the stage for the final [3 + 3] coupling. The actual coupling event did produce the desired hexasaccharide target as judged by crude ¹H- and ¹³C-NMR, however, all attempts to isolate the desired product failed due to its instability. Reasoning that the arming benzyl protecting groups may be the culprit for the low stability of the molecule, we decided to immediately subject the crude product to hydrogenolysis with Raney Ni, followed by global acetate protection. Pleasingly, this sequence worked, and we were able to obtain the desired hexasaccharide product in 54% yield over three steps as a single isomer. Finally, removal of the acetate protecting groups afforded the desired target in 8.9% overall yield from commercial starting material. In terms of overall yield, this approach was ten times more efficient than previously reported approaches, demonstrating the power of the methodology.

A few months after completing the synthesis of this hexasaccharide, the Rhee lab reported the synthesis of several 11-deoxylandomycin analogs.^[26] Central to this synthesis was the Pd-mediated hydroxyalkoxylation/RCM approach to trideoxy sugar synthesis. However, we were pleased to see that this group had adopted our Ts₂O-mediated glycosylation to synthesize the olivose-olivose core disaccharide of the molecule and attach it to the angulcycline aglycone with excellent selectivity. Shortly after, Gao and co-workers reported a further application of our chemistry in the total synthesis of FD-594.^[27] Here, the trisaccharide fragment of the molecule was assembled on gram scale using our TsCl chemistry. The authors then used the Ts₂O-mediated chemistry to attach this trisaccharide to the aglycone, after more conventional approaches failed to deliver the natural product. More recently, the Gao group used this chemistry with a permethylated quinovose (6-deoxy glucose) donor in their synthesis of calixanthomycin A.^[28]

4. Beyond Deoxy-Sugars

While the tosylate chemistry provided a reliable method for the construction of β-linked deoxy sugars, it was unclear that it could be extended to other classes of glycosyl donors. Notably, Mong and co-workers had reported that 2-deoxy glucose was approximately 500 times more reactive than glucose itself.^[29] This presented a potential hurdle, since the selectivity we observed in the glycosylation reactions involved maintaining a balance between the reactivity of the donor and the leaving group.

Despite these challenges, while we were working on extending our chemistry to fully substituted sugars reports began to emerge in the literature that the process should be feasible. Notably in 2016 Taylor and co-workers reported that glycosyl mesylates, generated in situ from the corresponding hemiacetal, reacted with donors under borinic acid catalysis to afford β-linked products with good to excellent levels of selectivity.^[30] The borinic acid catalyst was required to activate the acceptor for glycosylation, and in its absence, the reaction provided products with moderate levels of selectivity. Through a combination of ¹H-NMR and NMR exchange spectroscopy, the authors showed that the catalyst activated the acceptor for direct displacement of α-linked mesylates. In its absence a Curtin-Hammett scenario arose, where the α- and β-linked mesylates were in equilibrium, with the latter reacting with the acceptor faster than the former.

The need to activate the acceptor to obtain β-linked products was further highlighted in 2018 by the Walzcak group’s studies on the activation of anomeric stannanes for glycosylation using Koser’s reagent^.[31] In order to understand the basis of the β-selectivity in these reactions, they turned to NMR studies that found the reaction proceeded through the formation of an intermediate which displayed characteristic resonances of a glycosyl tosylate. It was hypothesized that this intermediate reacted with the tin ether of the acceptor, a species with enhanced reactivity.

In our own work, it quickly became apparent that the tosylate was not the optimal promoter for activation of 2,3,4,6-tetra-O-benzylglycopyranoside. Reasoning that changing the electronics of the leaving group could improve the reaction, we conducted a screen of sulfonyl chlorides and found that both NsCl and 3,5-bis(trifluoromethyl)benzene sulfonyl chloride were competent promoters for β-selective reactions between donor 31 and primary acceptor 3 (Scheme 7A).^[32] The importance of the sulfonate became clear when the less reactive acceptor 33 was used in the reaction. In these cases, NsCl led to the production of 34 with excellent β-selectivity, while the 3,5-bis(trifluoromethyl)benzene sulfonyl chloride provided the same product with only moderate levels of selectivity (Scehem 7B). To better understand this, we turned to VT-NMR to study the stability of the active intermediates. Through these studies, we observed that the glycosyl nosylate was stable up to 0°C, although it did begin to decompose at room temperature (Figure 2). In contrast, the 3,5-bis(trifluoromethyl)benzene sulfonate rapidly decomposed at 0°C (Figure 3). Based off of these studies, we hypothesized that the nosylate was undergoing an S_N2-like glycosylation while the 3,5-bis(trifluoromethyl)benzene sulfonate was undergoing both S_N2 and competitive S_N1-like glycosylations.

Scheme 7.

The effect of different sulfonyl chloride promoters on reactions with primary (A) and hindered (B) acceptors.

Figure 2.

VT-NMR study of stability of glycosyl nosylate intermediate.

Figure 3.

VT NMR study of stability of glycosyl-3,5-bis(trifluoromethyl) benzene sulfonate.

In order to verify the S_N2-like nature of the reaction we decided to carry out primary ¹³C KIE studies using the Jacobsen modified Singleton method.^[33] To minimize background signals, we conducted these studies on the reaction between per-methyl analogue of donor 31 and primary acceptor 3. These studies turned out to be less straightforward than we initially anticipated as the per-methyl analogue of 31 was an excellent substrate for the reaction, which went to completion in 5 minutes at −30°C. This was problematic, since the Singleton method requires that the reaction be run at low conversion. To get around this, we demonstrated that the selectivity in the reaction was independent of the stoichiometry of the coupling partners, and were able to run the reaction at “low” conversion using an excess of the donor. Through this we were able to measure KIE values of 1.035 at C1, a value that is consistent with other chemical and enzymatic S_N2 glycosylations.^[5c,34,35] Computational modelling of the transition state using the B3LYP–D3(BJ)/6–31G*/PCM method revealed that the reaction may be proceeding through a pre-organized transition state where the sodium cation both counterbalances the charge on the alkoxide and acts as a Lewis acid to activate the sulfonate for nucleophilic displacement (Figure 4).

Figure 4.

Minimum energy of transition state calculated at the B3LYP–D3(BJ)/6–31G*/PCM level. The yellow atom is the sulphur of the sulfonate, and the purple sphere is the sodium ion. The ion bridges both the nucleophile and the electrophile, pre-organizing the transition state.

Having established that the reaction is proceeding through an S_N2 manifold, we choose to examine the reaction with other classes of glycosyl donors. From these studies it quickly became apparent that a single sulfonate is not a one-size fits all solution to controlling selectivity in the reaction. Rather, the electronics of the sulfonate leaving group need to be matched to the reactivity of the donor. Fortuitously, we found that there was a linear relationship between the reactivity of the donor (as assessed by Wong’s RRV parameters)^[36] and the sigma constant of the para substituent on the aryl sulfonate leaving group (Figure 5). This has opened up the intriguing possibility of predicting the proper sulfonate to use for a β-selective reaction by measuring the donor’s RRV. Such information will be useful in further optimizing reactions with substrates that have proven to be challenging, such as the protected saccharosamine 35 which undergoes α-selective reactions under our standard conditions for glycosylation (Scheme 8).^[37]

Figure 5.

Correlation between the relative reactivity of a glycosyl donor and the Hammett sigma constant of the ideal promoter for β-selective glycosylations. RRV values are relative to mannose pentaacetate as measured by Wong et. al.

Scheme 8.

Exceptions for the rule that need further study. α-Selective glycosylation with saccharosamine.

5. Summary and Outlook

Compared to triflates, other classes of glycosyl sulfonates have been relatively underexplored as donors in carbohydrate synthesis. Recent work from our lab, and others, is quickly beginning to reveal the utility of these compounds. Through proper selection of sulfonate, it is possible to obtain highly stereo- and regioselective transformations without recourse to using directing groups. This has in turn permitted the construction of glycosidic linkages that were traditionally considered to be difficult to synthesize, such as β-linked deoxy sugars, with relative ease. This is best exemplified by the fact that the glycosyl tosylate chemistry has permitted the construction of the landomycin hexasaccharide on ten times higher yield than previously reported syntheses. Indeed, owing to its mild nature and predictable selectivity, other groups are beginning to adopt the tosylate chemistry for natural products synthesis.

Through these studies we have also found that there is no single sulfonate to provide selectivity with all classes of glycosyl donors. Rather, the electronics of the sulfonate need to be matched to the reactivity of the donor (and at times the acceptor!) in order to promote selective reactions through S_N2 manifolds. This is exemplified by elegant work from the Yu group and the Jiménez-Osés, Boutureira, and Bernardes groups who have shown that in the absence of proper sulfonate selection, selectivity in glycosylation reactions can be highly substrate dependent.^[38] Recent studies from our lab are beginning to demonstrate that the process of selecting a particular sulfonate should be predictable.

In a larger sense, these studies fit nicely in with elegant work from a number of other labs, who have begun to show that glycosylation can reliably be thought of as an S_N2 process. This is a tremendous change from when we started these studies over 12 years ago. At that time, much of the community assumed that chemical glycosylation proceeded through oxocarbenium ion intermediates, and that S_N2 glycosylation reactions were the exception not the norm. Increasing evidence has shown this not to be the case. While there is still a considerable amount of work that needs to be done before oligosaccharide synthesis can be performed as routinely as peptide and nucleic acid synthesis, an increased understanding of the physical organic chemistry of glycosylation reactions is providing a light at the end of the tunnel. There is still much mechanistic work to be done. However, as more and more mechanistic insights about the glycosylation reaction are revealed, we are coming closer to the day when stereoselective glycosylation reactions will be predictable processes for all classes of sugars.

Biography

Clay S. Bennett received his B.A. in Chemistry from Connecticut College in 1999. He then obtained his Ph.D. under the direction of Amos B. Smith, III at the University of Pennsylvania in 2005. Following postdoctoral studies with Chi-Huey Wong at the Scripps Research Institute, he started his independent career at Tufts University in 2008, where he is currently Professor of Chemistry. His current research focuses on the development of new glycosylation methodologies, their use in automated synthesis, and their application to the synthesis of antimicrobial oligosaccharides.