Detection and Characterization of Rapidly Equilibrating Glycosylation Reaction Intermediates Using Exchange NMR

The stereoselective introduction of glycosidic bonds (glycosylation) is one of the main challenges in the chemical synthesis of carbohydrates. Glycosylation reaction mechanisms are difficult to control because, in many cases, the exact reactive species driving product formation cannot be detected and the product outcome cannot be explained by the primary reaction intermediate observed. In these cases, reactions are expected to take place via other low-abundance reaction intermediates that are in rapid equilibrium with the primary reaction intermediate via a Curtin–Hammett scenario. Despite this principle being well-known in organic synthesis, mechanistic studies investigating this model in glycosylation reactions are complicated by the challenge of detecting the extremely short-lived reactive species responsible for product formation. Herein, we report the utilization of the chemical equilibrium between low-abundance reaction intermediates and the stable, readily observed α-glycosyl triflate intermediate in order to infer the structure of the former species by employing exchange NMR. Using this technique, we enabled the detection of reaction intermediates such as β-glycosyl triflates and glycosyl dioxanium ions. This demonstrates the power of exchange NMR to unravel reaction mechanisms as we aim to build a catalog of kinetic parameters, allowing for the understanding and eventual prediction of glycosylation reactions.


Introduction:
The stereoselective introduction of glycosidic bonds (glycosylation) is one of the main challenges in the chemical synthesis of carbohydrates.In a chemical glycosylation reaction, an electrophile (the glycosyl donor) is activated by a chemical promotor and reacts with a nucleophile (the glycosyl acceptor).The nucleophile can add to the α-or β-face of reactive intermediates thereby leading to the formation of α-or β-diastereoisomers, respectively.Controlling the diastereoselectivity of glycosylation reactions can be achieved by the application of two main strategies.
First, a stereodirecting group, present on the donor molecule, can be employed to stabilize the glycosyl cation formed upon activation.2][3][4] Extension of this principle to acyl functionalities positioned on the C-3, C-4 or C-6 hydroxyl groups via NGP has also been suggested to direct the stereoselectivity of glycosylation reactions.][6][7][8][9][10][11][12][13][14][15][16][17][18][19] The second main strategy utilizes glycosyl donors that contain protecting groups that are less capable of neighboring group participation e.g.benzyl ethers.1][22] Most modern promotor systems give rise to the formation of glycosyl triflates and since these covalent adducts can exist in the α- (1) or β-form (6), reactions proceeding via these intermediates can in principle form the β-or α-product via an SN2-like reaction pathway, respectively. 22,23The nucleophilic displacement of α-glycosyl triflates 1 is likely to take place via an intermediate α-contact ion pair (CIP) 2 which maintains it stereochemical memory to form the β-product. 246][27] The conformation of monosaccharide derived SSIPs is dictated by the relative stereochemistry of its substituents 10 and is a crucial determinant of their stereoselectivity in glycosylation reactions. 25Finally, βglycosyl triflate CIP 5 and its corresponding covalent adduct 6 can form during glycosylation reactions form the SSIP or SN2-like displacement of the α-glycosyl triflate 1 by another triflate anion. 20-glycosyl triflates (on D-sugars) are not stabilized by the anomeric effect and hence less stable and more reactive.Since glycosylation reactions take place in a mechanistic continuum between these SN1-and SN2-like reaction pathways they are difficult to predict and control and very sensitive to parameters such as solvent 28,29 , type of monosaccharide, 17,30 strength of the nucleophile 31 and reaction temperature. 32,33o address this challenge, the characterization of reaction intermediates that drive product formation in glycosylation reactions is required.Glycosylation reaction intermediates can be characterized by employing Nuclear Magnetic Resonance (NMR) spectroscopy, and Infra-Red-Ion-Spectroscopy (IRIS), for example. 26,34,35These techniques have allowed for the characterization of intermediates such as glycosyl triflates 22,23,34 , dioxolanium ions 18,36,37 , dioxanium ions 17 and even oxocarbenium ions. 25,26,36,38However, IRIS is performed in the gas-phase and the structure of reaction intermediates under these conditions may not always be relevant to those formed in solution.While NMR can be used to detected reaction intermediates in solution under relevant reaction conditions, the observation of a reaction intermediate does not automatically mean it is a reactive intermediate and hence relevant to product formation.In many cases the exact reactive species driving product formation remain unknown as the product outcome cannot be explained by the primary reaction intermediate observed.In these cases, reactions are expected to take place via other low-abundance reaction intermediates that are in rapid

Abstract:
The stereoselective introduction of glycosidic bonds (glycosylation) is one of the main challenges in the chemical synthesis of carbohydrates.Glycosylation reaction mechanisms are difficult to control because in many cases the exact reactive species driving product formation cannot be detected and the product outcome cannot be explained by the primary reaction intermediate observed.In these cases, reactions are expected to take place via other low-abundance reaction intermediates that are in rapid equilibrium with the primary reaction intermediate via a Curtin-Hammett scenario.Despite this principle being well-known in organic synthesis, mechanistic studies investigating this model in glycosylation reactions are complicated by the challenge of detecting the extremely short-lived reactive species responsible for product formation.Herein, we report the utilization of the chemical equilibrium between low abundance reaction intermediates and the stable, readily observed α-glycosyl triflate intermediate in order to infer the structure of the former species by employing exchange NMR.Using this technique, we enabled the detection of reaction intermediates such as β-glycosyl triflates and glycosyl dioxanium ions.This demonstrates the power of exchange NMR to unravel reaction mechanisms as we aim to build a catalogue of kinetic parameters allowing for the understanding and the eventual prediction of glycosylation reactions.equilibrium with the primary reaction intermediate via a Curtin-Hammett scenario (Figure 1). 20,39Therefore, the stereochemical outcome of the reaction does not necessarily depend on the population of the intermediate leading to a given product but rather the overall barrier height when the barrier to intermediate interconversion is lower than the irreversible product-forming step. 20Despite this principle being well-known in organic synthesis, mechanistic studies investigating this model in glycosylation reactions are complicated by the challenge of detecting the extremely short-lived reactive species responsible for product formation.The low abundance and short lifetime of these intermediates, such as 2-6, complicate their characterization as they readily equilibrate back to form the more stable, readily observable but non-reactive α-glycosyl triflate intermediate 1.We have utilized the chemical equilibrium between low abundance reaction intermediates and the stable, readily observed α-glycosyl triflate intermediate in order to infer the structure of the former species. 17Using chemical exchange saturation transfer (CEST) NMR we demonstrated that high-energy or "invisible" mannosyl dioxanium ions, which are formed by intramolecular stabilization of a C-3 ester, are in chemical exchange with the highly-populated αglycosyl triflate intermediate. 17rein, we report the application of this principle to detect the presence of other virtually undetectable high energy reaction intermediates relevant to product formation, such as β-triflates and dioxanium ions formed by internal stabilization.We characterized the reactive intermediates for a systematic set of eight frequently used glycosyl donors 7-14 (Figure 1B).Not only the presence of reactive intermediates but also their exchange kinetics were measured, thereby providing valuable quantitative data to elucidate the formation mechanism of the reactive intermediates.We report an integrated exchange NMR workflow to measure the reactivity of α-glycosyl triflates by monitoring the dissociation of the triflate ion using 19 F exchange NMR spectroscopy (EXSY NMR).In addition, we established the mechanism of triflate dissociation with the same technique.A clear difference between mannose and glucose monosaccharides was observed in their triflate dissociation kinetics and mechanism.Mannosides were able to form dioxanium ions 7d and 8d via the participation of a C-3 acyl group whereas their glucoside counterparts were not and formed β-glycosyl triflates 11βOTf-14βOTf instead.We were able to indirectly detect the presence of these low-population intermediates via their chemical equilibrium with the observable α-glycosyl triflate using 13 C CEST, 1 H CEST, and 19 F CEST NMR.Finally we were also able to characterize selected examples of the dioxanium ion and β-glycosyl triflate using more classical NMR techniques to unequivocally establish their structure.These results demonstrate the power of chemical exchange NMR to detect fleeting reaction intermediates to build a catalogue of kinetic parameters which allows for the understanding and ultimately prediction of glycosylation reactions.We expect this technique to be applicable to various other types of glycosylation reaction intermediates such as additives and solvents commonly used in glycosylation reactions, both of which tend to be rich in NMR-active nuclei that are sensitive to changes in chemical environment.Finally, the application of the workflow laid out herein should be applicable to other types of reaction that are under Curtin-Hammett control.

Results and Discussion
We started by investigating the stability and reactivity of glycosyl triflates derived from 7-14 to assess their likelihood of acting as reactive intermediates in glycosylation reactions.Previous experiments have investigated the decomposition temperature of glycosyl triflates as a measure of their stability/reactivity.While indicative of their thermal stability, such metrics do not necessarily speak to their relevance in product formation during a glycosylation event. 40In addition, 1 H EXSY NMR has been used to monitor the interconversion of α-and β-glycosyl triflates 23 and mesylates 41 providing kinetics of their interconversion.However, a major limitation of 1 H EXSY is that it requires both interconverting species to be visible in 1D NMR.This means that for the vast majority of glycosyl donors, intermediate exchange kinetics cannot be recorded due to the low populated state of the highly reactive βtriflate intermediates.Another means of measuring α-glycosyl triflate stability and reactivity is by observing the kinetics of triflate dissociation in the absence of an acceptor using 19 F EXSY NMR.Since both the α-glycosyl triflate and unbound triflate anion, which results from the activation step (vide infra), are always observed and exist as strong signals, we reasoned that if we could observe the interconversion between bound-and free-triflate then that would indicate the presence of an unobserved intermediate.Hence, we started by investigating the stability/reactivity of eight glycosyl triflates derived from 7-14 using 19 F EXSY NMR, which revealed all eight species underwent triflate exchange.The preparation of the thioglycoside precursors used to generate glycosyl triflates 7αOTf-14αOTf is described in the supporting info (SI page S12-S31).The glycosyl triflates were generated by activating the corresponding glycosyl sulfoxide donor with triflic anhydride (Tf2O) in the presence of the non-nucleophilic base 2,4,6-tri-tertbutyl-pyrimidine (TTBP) in CD2Cl2 at -80 o C inside an NMR tube (please see page S7-S10). 42Clean formation of glycosyl triflates (7αOTf-14αOTf) was observed in all cases (Figure S4-11).Dissociation of the anomeric triflate to unbound triflate could be monitored using selective 19 F EXSY NMR (SI page S2-S4, S7).Magnetization of the selectively-excited CF3 resonance of the α-glycosyl triflate is transferred to the unbound triflate ion upon triflate dissociation during the ZZ-exchange mix time (Figure 2A).By plotting the mix time vs the extent of magnetization transfer, the rate of triflate dissociation was measured, which we propose is an indicator of αglycosyl triflate stability (Figure 2B).In order to be compare measurements across different sugars and samples while also taking into account multiple mechanisms, the reported rates in Figure 2 are normalized by dividing the measured rate by the initial -triflate concentration (see SI page S2-4).By subsequently repeating this process at different temperatures, the normalized rate of triflate dissociation as a function of temperature was established (Figure 2C, Figure S49-S85).A few important considerations are needed to ensure reliable kinetic data to emerge from the 19 F EXSY experiments.First, the exchanging system needs to be in the slow-exchange regime which his defined by the rate of exchange being much lower than the frequency difference between the interconverting species (k1 + k-1 << A-B).Secondly, to reduce the complexity of modelling the kinetics, we chose to measure initial rates for the exchange processes, which puts a rough limit to the maximum normalized rate of ~100 s -1 (10% conversion at 1 ms mix time).Furthermore, the slowest measurable rate is related to the T1 of the observed nucleus.Hence, the T1 of the nuclei, and more importantly, the difference in T1 between the resonances of interest, should be taken into consideration when setting the maximum mix time.For 19 F, the T1s of the -triflate 10αOTf and soluble triflate salt were measured at -80 °C where exchange was frozen and found to be roughly identical (0.31 s and 0.42 s, respectively, Figure S48).This allowed us to make the approximation that T1 losses for each were roughly equivalent and hence could be ignored thereby allowing us to extend our mix range past T1.Therefore, we were able to measure normalized rates accurately down to ~0.1 s -1 .Third, the population of the glycosyl triflate and free triflate should not significantly change from the beginning to the end of experiment.Therefore, once thermal decomposition begins to take hold, these experiments become unreliable and thus limits the upper limit of the temperature window.
Keeping in mind these considerations, the rates of triflate dissociation from the parent α-glycosyl triflates 7αOTf-14αOTf at various temperatures were obtained (Figure 2C).Interestingly, clear differences in triflate dissociation were observed that related to the relative stereochemistry of the monosaccharide (mannose vs glucose) and the protecting group pattern.Mannosyl triflate 7αOTf and 8αOTf carrying a single benzoate ester at C-3 showed the fastest triflate dissociation.Interestingly, the benzylidene protected mannosyl analogue 8αOTf carrying a C-3 benzoate ester was second fastest in the mannose series.Benzylidene protected monosaccharides are typically classified as disarmed due to the torsional 43,44 and electronic effects 45 induced by the fused benzylidene ring system. 40Indeed triflate dissociation for 8αOTf was slower than the corresponding benzylated analogue 7αOTf but still faster than the fully benzylated compound 9αOTf.This suggests a role of the C-3 ester in driving triflate release, overpowering the disarming effect induced by the benzylidene.This is confirmed by the fact that the benzylidene protected mannosyl triflate containing benzyl ethers at C-2 and C-3 was slowest in the mannose series in triflate dissociation.Interestingly, the opposite trend was observed in the glucose series.The difference in rate for the C-3 benzoyl 11αOTf and 12αOTf vs C-3 benzyl 13αOTf and 14αOTf analogues was much smaller with the former being slower, particularly at higher temperature.As expected, the benzylidene protected analogues 12αOTf and 14αOTf showed slower triflate dissociation compared to their benzylated counterparts 11αOTf and 13αOTf.Overall these results indicate that the presence of a benzylidene acetal compared to the benzylated analogue slows down triflate dissociation for both the mannose and glucose series.However, the introduction of a C-3 benzoyl group speeds up triflate dissociation in the mannose series and slows down triflate release in the glucose series.
The striking differences in triflate dissociation rates are likely the result of a different mechanism of triflate dissociation, but these cannot be established from the rates alone.Hence, we set out to investigate the mechanism of triflate dissociation of 7αOTf-14αOTf to explain their dissociation rate differences.We foresee three main equilibria that would lead to the dissociation of the triflate anion (Figure 3A).First, dissociation of the triflate anion could lead to a SSIP.Secondly, NGP of the C-3 acyl substituent present in molecules 7αOTf-8αOTf could take place to form a dioxanium ion with or without an intermediary SSIP.Third, a triflate anion could displace the αglycosyl triflate in an SN2-like manner to form the corresponding βtriflate.To dissect the mechanism(s) responsible for triflate dissociation we investigated the rate of this process as a function of the triflate anion concentration.The rate of triflate dissociation from the α-glycosyl triflate via SSIP and/or dioxanium ion formation should be insensitive to the triflate concentration.The rate of triflate dissociation from the α-glycosyl triflate via SN2-like displacement to form the corresponding β-glycosyl triflate should be first order with respect to the triflate concentration.In case both processes operate simultaneously, the sum of the rates would constitute the overall rate of triflate dissociation (Figure 3A).The temperature for the study was chosen so that if a first-order dependence was found, it would not move the kinetics outside the window for EXSY and at the same time being suitably fast enough to study if no dependence was discovered.Next, using the aforementioned 19 F EXSY scheme, we measured the triflate dissociation as a function of triflate concentration at a single temperature by the addition of tetrabutylammonium triflate (TBAT).To this end, the triflate concentration was increased in steps by taking the sample out of the probe and adding 1 M TBAT solution in dry CD2Cl2 at -80 °C to approximately double the triflate concentration in each step.The exact triflate concentration was determined by comparison to the trimethyl(4-(trifluoromethyl)phenyl)silane internal standard.The results are plotted in Figure 3B-C and show a clear dichotomy in the mechanism of triflate dissociation, consistent with the observed temperature dependent rate differences listed in Figure 2C.C-3 Benzoyl-containing mannosides 7αOTf-8αOTf showed a 0 th -order rate dependence with respect to the triflate concentration.These results indicate that internal stabilization of the C-3 benzoyl, which forms the dioxanium ion, drives triflate dissociation.In contrast, all glucosides (11αOTf-14αOTf) and both mannosides lacking the C-3 benzoyl group (9αOTf-10αOTf) show a 1 st order triflate concentration dependence for the rate of triflate dissociation (Figure 3B-C).These data suggest that these molecules likely undergo a SN2-like triflate displacement to form the β-glycosyl triflate.Furthermore, this demonstrates that internal stabilization by the C-3 acyl group is reserved to the mannose donors while the glucose counterparts likely form βglycosyl triflates.The above 19 F EXSY studies established the rates of triflate dissociation as well as provided evidence for its mechanism but does not confirm the presence of the intermediates proposed in such mechanisms.Since EXSY can only be applied if both intermediates are readily visible in 1D NMR, it cannot be used to study the expected dioxanium ions and/or β-glycosyl triflate intermediates due to their low population at equilibrium.Hence, to solve this challenge, a different technique is needed.As mentioned above, chemical exchange saturation transfer (CEST) NMR is well suited to detect very low population "invisible" reaction intermediates that are in a chemical equilibrium with a visible reaction intermediate.This technique, originally developed as a contrast approach in MRI, 46 has been applied by Gschwind et al. to detect iminium ions and by us to detect mannosyl dioxanium ions, for example. 17,47The main advantage of CEST NMR is that no prior information about the chemical shift of the low-populated intermediate resonance is required since the experiment scans a given frequency domain by incrementing the saturation offset frequency while monitoring the transfer of saturation to the main observable species via chemical exchange (Figure 4A).Furthermore, by varying saturation times the interconversion kinetics for these intermediates can be quantified. 48The limitations of CEST are similar to EXSY: the chemical exchange is slow on the NMR timescale (k1 + k-1 << A-B), and the detection window is defined by a combination of relative population exchanging species and longitudinal relaxation rate (R1). 46,49Since the dioxanium ions can only form from the C-3 benzoyl containing compounds 7-8 we investigated this set using 13 C CEST NMR.To this end, 13 C labelled substrates 7 13C and 8 13C were prepared to enable the sensitive detection of the carbonyl quaternary carbon (SI page S12-31).By incrementing the saturation offset frequency while monitoring the transfer of saturation to a reporter peak ( 13 C=O) on the main observable species (α-glycosyl triflate) via chemical exchange we investigated the detection of mannosyl and glucosyl dioxanium ion formation (Figure 4, Top, SI Figure S13-17).Apart from the use of a 13 C-labelled starting material, the CEST NMR experiments were performed under identical conditions as the EXSY experiments.As expected, the mannosyl triflate 7 13C αOTf showed saturation transfer at the chemical shift expected for the dioxanium ion (δC = 177 ppm, Figure 4A).In contrast, the mannosyl benzylidene derivative 8 13C αOTf did not (Figure 4B), even though the formation of a dioxanium ion was expected based on an observed 1 st -order triflate dependence.Due to the presence of the benzylidene, the molecule is less flexible and requires the formation of a tricyclic dioxanium ion which is likely much slower and less stable than the non-benzylidene derivative 7d.This may explain why CEST NMR was unable to pick up the dioxanium ion as the CEST NMR technique is limited by the population of the minor exchanging species and the rates of dioxanium ion formation and consumption (ka→α and kd→α respectively).If the population of the intermediate is well below 1% or the exchange has moved into the intermediate or fast regime (kα→d + kd→α > d), then the minor exchanging species cannot be detected.Due to the intrinsically high reactivity of a tricyclic dioxanium ion, it is likely that dioxanium ion consumption (kd→α) is exceedingly fast, hence, limiting detection of the ion by 13 C CEST NMR.The corresponding glucose series (11 13C and 12 13C ) did not show dioxanium ion formation via CEST (Figure S14 and S16), fully consistent with the 1 st -order triflate dependence (Figure 3C).In light of the conflicting CEST results with the triflate order in the case of mannosyl triflate x, we reasoned that we could be approaching the limitation of the CEST NMR viewing window outlined above.Therefore, we wanted to boost the observable population of the dioxanium ion by preparing the corresponding p-anisoyl derivatives 15 13C and 16 13C of the mannose series as the electon-donating methoxy group should increase the stability of the dioxanium ion and slow down the α-glycosyl triflate formation.Additionally, we performed the activation of 15 13C and 16 13C in the absence of TTBP to minimize the free triflate anion concentration in an effort to further boost the population of dioxanium ion. 17Under these conditions, saturation transfer of the dioxanium ion was observed in both cases (Figure 4C and D).For benzylidene derivative 16 13C , it was now possible to detect the dioxanium ion with 13 C CEST NMR, which aligned with its 0 th -order triflate dependence.Furthermore, mannosyl donor 15 13C showed a much higher population (~45%) of dioxanium ion compared to benzoyl derivative 7. Due to the very high population of dioxanium upon activating donor 15 13C , we were able to use the dioxanium ion signal as the reporter peak of the CEST NMR experiment in order to detect if there were additional intermediates chemical exchange with it.Identical to the benzoyl-derivative, the α-glycosyl triflate was the only detectable species in chemical exchange with the dioxanium ion.Fortunately, due to the high population of ion 15 13C d, we could fully characterize it using 1 H- 13 C HMBC and 1 H-1 H COSY experiments (Figure 4, Bottom).We had previously reported the characterization of a similar mannosyl dioxanium ion but these results now demonstrate its formation with relevant protecting groups which are routinely used in oligosaccharide synthesis.Hence, although not directly observed, we expect the characteristic saturation transfer at 175 ppm for benzylidene derivative 16 13C to correspond to the tricyclic mannosyl dioxanium ion 16 13C d.To further solidify this hypothesis we measured the triflate dissociation of 16 13C αOTf with 19 F EXSY and dioxanium ion formation using 1 H and 13 C CEST NMR.The measured rate from these techniques were 0.28± 0.02 s -1 , 0.28± 0.01 s -1 and 0.26± 0.004 s -1 , respectively, indicating that the processes for forming the dioxanium and dissociating the triflate group are clearly coupled (Figure 4D).From these results it is clear that mannoside 15 13C and 16 13C showed formation of the C-3 dioxanium ion, which is consistent with the α-selective glycosylation behaviour of 7 and 8 and the observation that their triflate dissociation rates are irrespective of the triflate anion concentration.
The C-3 acylated mannosides above are a privileged class of glycosides as all of the other cases (9αOTf-14αOT) displayed a 1 st -order triflate anion concentration dependence for triflate dissociation.These observations are consistent with an equilibrium between an α-and β-glycosyl triflate.β-glycosyl triflates are exceedingly difficult to detect and have only been observed in case of benzylidene-protected methylated glucose and allose donors using 13 C labelled monosaccharides. 23Only in case of the allose, a large enough population of β-glycosyl triflate was formed that allowed for the investigation of its exchange kinetics using 1 H EXSY. 23 Furthermore, an equatorial triflate not stabilized by the anomeric effect formed via a conformational ring flip was observed by Van der Marel and co-workers. 50To the best of our knowledge, these two reports forms the summative collection of observed glycosyl triflates not stabilized by the anomeric effect.To enable the detection of unstable and very low populated β-glycosyl triflates we again applied CEST NMR but focused on the 1 H and 19 F nuclei.For 1 H CEST NMR, the H-1 signal belonging to the α-glycosyl triflate was used as a reporter peak and the saturation offset frequency was scanned while monitoring saturation transfer to this peak (Figure 5, Top).Upon detection of a possible β-glycosyl triflate, saturation transfer from an upfield peak position (~5.5-5.7 ppm) would be expected to be observed.None of the mannoside donors displayed a chemical equilibrium with an upfield resonance with the exception of benzylidene derivative 10, which we tentatively assigned as β-glycosyl triflate 10βOTf (Figure 5A-H).These results are consistent with the triflate orders obtained for three of these compounds 7 and 8 (0 th order, no β-glycosyl triflate detected) and 10 (1 st order, β-glycosyl triflate detected).Only for perbenzyl mannosyl donor 9, we did not detect a β-glycosyl triflate even though its triflate dissociation rate is 1 st order in triflate concentration.This could be due to a very low population of βglycosyl triflate at equilibrium or the system has moved out of the slow-exchange regime in which case it would be difficult or impossible to detect with 1 H CEST NMR.In contrast, all of the glucose series donors (11-14) showed clear saturation transfer via an upfield peak that we presume is a β-glycosyl triflate.These results are fully consistent with their 1 st order triflate dependence (Figure 3B-C).We validated these results by repeating β-glycosyl triflate detection using the 19 F channel.For 19 F CEST NMR, the CF3 signal belonging to the -glycosyl triflate ion was used as a reporter peak and the saturation offset frequency was scanned while monitoring saturation transfer to this peak (Figure S19-S26).As a consequence, all 19 F CEST NMR spectra identify chemical exchange with the "free" triflate anion pool which can arise from either the α-or β-glycosyl triflate.The 19 F CEST NMR results are fully consistent with those obtained from the 1 H CEST NMR experiments.For the mannose series, chemical exchange of the α-glycosyl triflate with the unbound triflate anion could be clearly observed via saturation transfer monitored from the α-glycosyl triflate at -75.9 ppm (Figure S22).Only for the benzylidene derivative 10 an additional saturation transfer from a presumed β-glycosyl triflate was observed at -75.0 ppm, similar to the 1 H CEST NMR.Moreover, in case of the glucose series, saturation transfer from β-glycosyl triflates could be detected in all cases using 19 F CEST NMR as was the case for the corresponding 1 H experiments.
To obtain additional proof that the saturation transfer originates from exchange with a the β-glycosyl triflate, we set out to further characterize this low population species.From the glucose set, benzylidene derivative 14 displayed the clearest formation of this species and we thus investigated this molecule further.The only reported direct spectroscopic evidence of a β-glycosyl triflate was obtained for a very similar compound, 4,6-benzylidene-2,3-di-Omethyl glucosyl-β-triflate by Asensio and co-workers. 23We prepared the corresponding benzyl protected compound with a C-1 13 C isotope label (14 13C-1 ) in order to measure the 1 JC1-H1 and 3 JH1-H2 coupling constants, which are indicative of the stereochemistry at C-1. Upon activation, this derivative formed a small (1%, SI, Figure S46) population of β-glucosyl triflate which could be characterized using a 1 H- 13 C HSQC NMR experiment where the 13 C decoupler was not applied during the acquisition period (Figure 6).This is experiment serves as a simple method for measuring the 1 JCH as well as allowing for long F2 acquisition times to obtain highresolution peaks in the 1 H dimension suitable for measuring 1 H-1 H couplings.The 1 JC1-H1 coupling constant for the minor species was determined to be 175 Hz compared to 183 Hz for the corresponding α-derivative (Figure 6).This 8 Hz decrease in coupling constant is indicative of an axial H-1 found in β-configured molecules. 51Most importantly, the 3 JH1-H2 coupling constant was measured to be 7.5 Hz, which is consistent with the axial-axial coupling expected for a β-glycosyl triflate intermediate.C decoupling in order to measure 1 JC1-H1 and 3 JH1-H2.
The systematic set of exchange NMR experiments described above allows for the development of a rationale for the reaction pathways leading to the stereoselective formation of glycosylation product(s) for each of the eight glycosyl donors 7-14 studied.For all eight examples, the main reaction intermediate formed after activation is the α-glycosyl triflate.A baseline of SN2-like displacement pathway leading to the β-product can therefore be expected in all cases.The SSIPs involved in the reactions could not detected due to their instability but for both the mannose 10,25 and glucose 25 series are expected to react via their α-selective 4 H3 half-chair conformers.Additional pathways proceeding via C-3 participation and β-glycosyl triflate intermediate can also lead to α-selective product formation.The rate differences of reaction intermediate interconversion and the ensuing product forming step therefore dictate the stereospecific outcome as outlined by Lemieux and coworkers. 20The stereoselectivity for all eight glycosyl donors 7-14 as a function of acceptor reactivity was investigated separately. 18tarting with the mannose series, C-3 benzoyl mannosyl donor 7 provides α-mannosides upon glycosylation. 17,18The corresponding α-glycosyl triflate 7αOTf is the main observable reaction intermediate but does not give direct access to the αproduct via an SN2-like pathway.Hence, glycosylation likely takes place via a rapidly equilibrating dioxanium ion 7d and/or β-glycosyl triflate 7βOTf which would afford the α-product.Dioxanium ion 7d was observed via CEST NMR while the β-glycosyl triflate was not nor was it expected to be based on its 0 th -order triflate dependence.Therefore, we confidently propose that the reactive intermediate in this case is the dioxanium ion and that the system is under Curtin-Hammett control (Figure 7).
The corresponding C-3 benzoyl-4,6-benzylidene mannoside 8 shows a similar profile.The α-mannoside is formed upon glycosylation which cannot be explained via the observable αglycosyl triflate intermediate. 52CEST NMR was unable to confirm the presence of dioxanium ion 8d or β-triflate 8βOTf.However, the rate of triflate dissociation from 8αOTf was independent of the triflate concentration pointing towards triflate dissociation via C-3 participation.By substituting the C-3 benzoyl for anisoyl in order to further stabilize the cation, the resulting dioxanium ion was detected for the first time.The existence of this intermediate has long been debated, 5,6 and frequently omitted from consideration. 5,14,52However, the experiments support its existence as the chemical shift, triflate order and glycosylation stereoselectivity all support the existence of dioxanium ion 8d and its role as a product-forming intermediate (Figure 7).Perbenzyl mannoside 9 is a much less selective glycosylation donor.The 1 st order of the triflate concentration on triflate dissociation support the formation of a β-glycosyl triflate although this species could not be detected via the exchange NMR experiments.We hypothesize that a dynamic equilibrium of α-, β-glycosyl triflates and possibly the SSIP is present leading to various product forming pathways and hence mixed stereoselectivity.
4,6-benzylidene mannoside 10 is known to provide β-products upon glycosylation. 39This can be explained by an SN2-like displacement of the corresponding α-contact ion pair. 24Hence, in this case, the main observed reaction intermediate is assigned as the reactive intermediate.Interestingly, CEST NMR did show the presence of an equilibrium with the β-triflate intermediate, which is consistent with its 1 st order triflate dependence.Interestingly, this equilibrium is not relevant for product formation with most nucleophiles and therefore represents a situation where the triflate is the less reactive species (Figure 7).
The glucose series showed very different stereoselectivity in glycosylations compared to the corresponding mannose series as noted earlier by Crich and co-workers. 30In sharp contrast to the mannose series, no evidence for dioxanium ion formation by a C-3 acyl neighboring group in 11 and 12 was found (Figure 8).The origin of this striking difference was investigated further in a separate study.Both molecules form β-glycosyl triflates whereas the mannosides do not, but these lead to a lower α-selectivity than the dioxanium ions in the case of their mannose counterparts. 18Also the origin of this clear divergence was investigated further in a separate study.A role for the SSIP, which is expected to adopt a 4 H3 halfchair conformer and drives the formation of α-product as reported by Codée and co-workers, cannot be excluded. 25The simultaneous operation of multiple product forming pathways likely leads to a lower stereoselectivity in these cases.The two glucosides 13 and 14 lacking the C-3 benzoyl group show a similar, moderate stereoselectivity compared to their C-3 benzoyl counterparts (Figure 8). 18This underscores the lack of stereodirecting capability of the C-3 substituent in glucose. 18The benzylidene derivative 14 has been demonstrated to be α-selective in contrast to its mannose derivative. 30Since 14 also clearly showed to be in equilibrium with the β-glycosyl triflate, we assign this as a major reactive intermediate driving α-product formation (Figure 8).

Conclusions
We were able to directly and indirectly detect the presence of lowpopulation reaction intermediates via their chemical equilibrium with the readily observable α-glycosyl triflate for a collection of eight of gluco-and mannosides bearing either C-3 acyl or benzyl substituent.The stability of α-glycosyl triflates and their mechanism for dissociation could be readily measured using 19 F EXSY NMR.This allowed for their role in glycosylation reactions to be ascertained.Furthermore, equilibria with low-population intermediates such as dioxanium ions and β-glycosyl triflates could be detected using 13 C CEST, 1 H CEST, and 19 F CEST NMR.Finally, selected examples of the dioxanium ion and β-glycosyl triflate were characterized using classical NMR techniques to unequivocally establish their identity.The results provide the first mechanistic proof for the existence of mannosyl dioxanium ions and β-glucosyl triflates utilizing relevant protecting groups and under actual reaction conditions.These observations allow for the rationalization of their observed αselectivity and demonstrate the power of chemical exchange NMR to detect transient intermediates.Ultimately, we aim to build a catalogue of kinetic parameters allowing for the understanding and the eventual prediction of glycosylation reactions.We expect this technique to be applicable to various other types of glycosylation reaction intermediates such as additives (e.g.DMF, PPh3O) and coordinating solvents (e.g.Et2O, ACN).Finally, the application of the workflow laid out herein should be applicable to other types of reactions that are under Curtin-Hammett control.

Figure 1 :
Figure 1: A) Glycosylation reaction intermediates and their characteristic resonances monitored by exchange NMR B) Glycosyl donors used in this study.

Figure 2 :
Figure 2: A) Working principle behind 19 F EXSY NMR to monitor triflate dissociation.The α-glycosyl triflate is selectively excited and then observed over time transforming into free triflate.B) Determination of the initial rate of triflate dissociation.C) Summary of normalized triflate dissociation rates (s -1 ) across the studied glycoside series as a function of temperature.

Figure 3 :
Figure 3: A) SN1 and SN2-like triflate dissociation mechanisms.B) Reaction order determination for triflate dissociation in the Mannose series and C) Glucose series.

Figure 4 :
Figure 4: Top) Mechanism of 13 C CEST NMR to detect neighboring group participation.A-D) CEST profiles for 13 C-labeled C-3 acylprotected mannosyl and glucosyl donors.Bottom) Characterization of mannosyl dioxanium ion using HMBC and COSY.