Excited-State Palladium-Catalyzed α-Selective C1-Ketonylation

Abstract

C-Glycosides are important carbohydrate mimetics found in natural products, bioactive compounds, and marketed drugs. However, stereoselective preparation of this class of glycomimetics remains a significant challenge in organic synthesis. Herein, we report an excited-state palladium-catalyzed α-selective C-ketonylation strategy using readily available 1-bromosugars to access a range of C-glycosides. The reaction features excellent α-selectivity and mild conditions that tolerate a wide range of functional groups and complex molecular architectures. The resulting α-ketonylsugars can serve as versatile precursors for their β-isomers and other C-glycosides. Preliminary experimental and computational studies of the mechanism suggest a radical pathway involving the formation of palladoradical and glycosyl radical that undergoes polarity-mismatched coupling with silyl enol ether, affording the desired α-ketonylsugars. Insight into the reactivity and mechanism will inspire new reaction development and provide straightforward access to both α- and β-C-glycosides, greatly expanding the chemical and patent spaces of glycomimetics.

Graphical Abstract

Introduction

Glycoconjugates are important constituents of living organisms where they regulate indispensable functions in a wide range of biological processes, including cell-cell recognition, cell-matrix interactions, and detoxification processes.¹ Over the past several decades, intensive efforts have been made to develop methods of constructing and utilizing C─O and C─C glycosidic bonds.^2-8 C─C glycosidic bonds are not subject to hydrolysis and are inert toward the hydrolytic enzymes in vivo, enabling C-glycosides to be used as artificial surrogates for potential therapeutic agents by mimicking their native O-glycoside structure.^9-12 For example, the C-glycoside analog of KRN7000 is metabolically stable and exhibits excellent antitumor activity (Figure 1A).¹³ Other bioactive C-glycosides include Pro-Xylane,¹⁴ (−)-neodysiherbaine A,¹⁵ (+)-varitriol,¹⁶ and (+)-ambruticin S.¹⁷ Although chemical methods to construct the anomeric C─C bond have been reported, the development of a synthetic strategy that has a broad substrate scope and enables stereoselective access to both α- and β-anomers remains a challenge in the field.^{2, 18-21}

Figure 1.

Development and exploration of ketonylation of carbohydrates.

C1-ketonylation of carbohydrates (i.e., replacing the C1 leaving group with a ketone) is a fundamental strategy with which to synthesize C-glycosides.² The resulting C1-ketonylsugars could serve as versatile precursors for the preparation of a range of C-glycosides. Catalytic C1-ketonylation reactions that have a high level of α-selectivity are highly desired because the corresponding α-ketonylsugars can be epimerized to thermodynamically stable β-ketonylsugars,^22-23 granting selective access to both α- and β-glycosides from the same starting materials. Most of the existing strategies, including Knoevenagel condensation,^24-28 Horner-Wadsworth-Emmons olefination/Michael addition cascade reactions,²⁹ Claisen rearrangements,³⁰ and nucleophilic substitutions^31-34 are non-catalytic transformations and afforded the thermodynamic β-ketonylsugars. Catalytic C1-ketonylation is rare, and the current state-of-the-art approach involves Au-catalyzed nucleophilic substitution of an in situ generated oxocarbenium ion by silyl enol ethers through an ionic, 2-electron reaction pathway, producing the desired C1-ketonylsugars with 2:1 to 10:1 α/β-selectivity (Figure 1B).³⁵ This elegant strategy took place at room temperature and finished within 30 min, but substrates bearing C2-O-acyl protected moiety (disarmed sugars) failed.^36-39 Presumably, this is due to the formation of the oxocarbenium ion intermediate, which is disfavored by the electron withdrawing property of the acyl protecting group and complicated by the neighbouring group participation of the C2-OAc/OBz moiety.^40-41 Given that radical reactions often proceed under mild reaction conditions and the corresponding glycosyl radicals stereoelectronically favor the formation of the α-epimer,^{8, 42-45} we posited that an open-shell, radical C1-ketonylation of carbohydrate derivatives could provide a complementary strategy to achieve C-glycosylation with high levels of α-selectivity and functional group compatibility.

While developing the excited-state Pd-catalyzed C2-ketonylation of 1-bromosugars,⁴⁶ we observed the formation of C1 α-ketonylated adduct as a side-product. We questioned whether we could establish a general and highly α-selective C1-ketonylation through excited-state palladium catalysis.⁴⁷ This catalytic approach allows access to both close- and open-shell species under mild conditions and has recently been utilized in a number of radical transformations of aryl or alkyl halides.^48-63 Herein, we report our success in exploiting this excited-state catalytic platform and readily accessible 1-bromosugars to generate the palladoradical intermediate ([Pd^I]Br) and glycosyl radicals, which can undergo polarity-mismatched coupling with nucleophilic silyl enol ethers, selectively forming the desired α-coupling product (Figure 1C). Preliminary mechanistic studies suggest that the subsequent reaction proceeds through bromine atom transfer followed by H-Br elimination, furnishing the silyl enol ether that is hydrolyzed to deliver the desired C1-ketonylated carbohydrates. In addition to these mechanistic insights, the reaction (i) tolerates disarmed sugars,^64-65,66-67 (ii) exhibits broad substrate scope and wide functional group compatibility, (iii) features excellent α-selectivity, (iv) is amenable to late-stage functionalization of complex molecules, and (v) allows rapid access to a range of α- and β-C-glycosides via various post-functionalization reactions.

Results and Discussion

At the outset of our investigation, we selected the challenging combination of a readily available acetylated α-glucosyl bromide (1a) and acetophenone trimethylsilyl enol ether (2a) as model substrates for initial optimization studies, expecting that a broad reaction scope, including disarmed sugars would be observed (Table 1). When 1a and 2a were treated with 5.00 mol% Pd(PPh₃)₄, 6.00 mol% Xantphos [(9,9-dimethyl-9H-xanthene-4,5-diyl)bis(diphenylphosphane)], 1.50 equiv KOAc, and 10.0 equiv H₂O in dioxane (0.10 M) at room temperature (rt) under irradiation from 24 W blue LEDs for 24 h (entry 1), the desired product (3a) was obtained in 95% yield with >20:1 α-selectivity. The Pd(PPh₃)₄ catalyst was critical for the desired reactivity: no reaction occurred in its absence, and only 14% of the desired product was formed when it was replaced with Pd(OAc)₂ (entries 2 & 3). Removing Xantphos or replacing it with BINAP decreased the reaction yield (entries 4 & 5). While other common photocatalysts, such as Ru(bpy)₃(PF₆)₂ and Eosin Y free acid failed, Ir(ppy)₃ gave only 24% yield (entries 6-8). Using MeCN as the solvent instead of dioxane significantly lowered the yield (entry 9). Water, which is presumed to hydrolyze the silyl enol ether intermediate to form the ketone product, was crucial for a high yield of the product (entry 10). Finally, oxygen-free conditions and light were essential for high reaction efficiency (entries 11 and 12).

Table 1.

Selected optimization experiments^a

Entry	Deviation from standard conditions	Yield (%)	α/β
1	None	95	>20:1
2	Without Pd(PPh3)4	N.R.	-
3	Pd(OAc)2 instead of Pd(PPh3)4	14	-
4	Without Xantphos	69	>20:1
5	BINAP instead of Xantphos	70	18:1
6	Ir(ppy)3 as photocatalyst	24	-
7	Ru(bpy)3(PF6)2 as photocatalyst	N.R.	-
8	Eosin Y free acid as photocatalyst	N.R.	-
9	MeCN as solvent	7	-
10	Without H2O	45	-
11	Air	N.R.	-
12	Keep in dark	N.R.	-

^aSee Supplementary Information (SI) for experimental details. Reaction yields and α/β ratios were determined by ¹H-NMR using CH₂Br₂ as an internal standard. Ac, acetyl; BINAP, 2,2′-bis(diphenylphosphino)-1,1′-binaphthyl; LED, light-emitting diode; N.R., no reaction.

With the optimized reaction conditions in hand, we examined the generality of the reaction and found that a range of silyl enol ethers reacted with α-glucosyl bromide, forming the desired products in moderate to excellent yields (Table 2A). Different silyl enol ethers with electron-neutral (2a), electron-withdrawing (2b-2e), or electron-donating (2f) substituents on the aryl ring delivered the corresponding α-ketonyl glucosides (3a-3f) in 55-87% yields and with up to 20:1 α-selectivity. Aryl silyl enol ethers with an extended conjugation (2g) or multiple substituents (2h-2i) were compatible. Notably, medicinally relevant heteroaryl derivatives, including pyridyl (2j), furanyl (2k), and thienyl substituted silyl enol ethers (2l), were viable substrates, furnishing the desired products (3j-3l) in good to excellent yields and with high levels of α-selectivity.⁶⁸ The reaction is amenable to gram scale synthesis with similar efficiency (3l, Fig. S1). The absolute stereochemistry of the product 3a was confirmed by a single-crystal analysis,⁶⁹ as shown in Table 3, and 2D-NMR spectroscopy. The stereochemistry of the other products was tentatively assigned by analogy.

Table 2.

Scope of excited-state palladium-catalyzed α-selective ketonylation of carbohydrates^a

^aSee SI for experimental details. Isolated yield and α/β ratio are indicated below each entry

Table 3.

Post-functionalization of α-ketonylsugars^a

^aSee SI for experimental details. Isolated yield and α/β ratio are indicated below each entry.

We evaluated the scope of α-bromosugars under the standard reaction conditions. A broad array of α-bromosugars derived from D-galactose, D-xylose, D-glucose, D-mannose, L-fucose, and L-rhamnose (3m-3q, 3s, 3t) reacted with silyl enol ether 2a, affording the desired products in 64-87% yields and with up to >20:1 α-selectivity (Table 2B). Protecting groups such as tert-butyldiphenylsilyl, acetal, and ketal were well tolerated (3o-3q). A D-mannofuranose derivative reacted smoothly to generate the desired product 3r in 95% yield and with complete α-selectivity. An analog of D-glucuronic acid also proved to be compatible with the standard conditions (3u). Acetyl-protected disaccharides, such as cellobiose, maltose, and melibiose, underwent C1-ketonylation, furnishing the corresponding products 3v-3x in 74-87% yields and with 10:1 to >20:1 stereoselectivity.

Late-stage modification of complex molecules is often a key to identify medicinal agents.⁷⁰ To demonstrate the applicability of the excited-state Pd-catalyzed C1-ketonylation to late-stage syntheses, natural product- and drug-conjugated sugars were subjected to the standard reaction conditions (Table 2C). For example, α-bromosugar derivatives of L-menthol, Febuxostat, Adapalene, Probenecid, Ibuprofen, Indomethacin, oleanolic acid, and Zaltoprofen reacted, affording the desired products 5a-5h in 58-90% yields and with up to >20:1 α-selectivity.

The α-ketonylsugars produced in this reaction are versatile synthetic intermediates that can be used to prepare other valuable C-glycosides. For instance, the α-isomers could be epimerized to their β-isomers under mild basic conditions with up to 93% yield and >20:1 β-selectivity (Table 3A). The epimerization proceeds through the base-mediated ring-opening followed by a Michael addition reaction, aligning with our computational results (Fig. S13). Moreover, the ketonylsugars could also undergo hydrogenation, olefination, Baeyer-Villiger oxidation, Beckmann rearrangement, and reduction to afford the corresponding alkylated C-glucosides in moderate to good yields (Table 3B-3E and Fig. S1).

To shed more light on the reaction mechanism and the origin of the stereoselectivity, we conducted a series of experimental and computational studies. UV-Vis measurements showed that a mixture of Pd(PPh₃)₄ and Xantphos had a strong absorption at 347 nm (Figure 2A). This observation suggested that the ligand exchange forms [Pd⁰(PPh₃)₂Xantphos], which might be the active catalyst.⁵⁵ Irradiating the reaction mixture with blue light for 8 min shifted the λ_max from 347 nm to 360 nm, which implicates the formation of [Pd^II]-species.⁶³ Stern-Volmer quenching studies demonstrated that α-glucosyl bromide (1a) quenches the excited palladium species more efficiently than phenyl silyl enol ether (2a, Figure 2B). The addition of a radical scavenger such as butylated hydroxytoluene (BHT) or 2,2,6,6-tetramethylpiperidine 1-oxyl (TEMPO) inhibited the reaction (Figure 2C). The TEMPO-glucosyl adduct (12) was also detected by HRMS (Fig. S5). In addition, the radical clock experiment using cyclopropyl silyl enol ether 13 afforded the ring-opening product 14 in 22% yield (Figure 2D). These results suggest that the reaction likely proceeds through a radical pathway. Moreover, radical chain propagation is unlikely because the quantum yield of the reaction is 0.02 (Fig. S10). We did not observe the formation of TMS-protected silyl enol ether products, presumably, the labile TMS group was deprotected under the reaction. Indeed, performing the reaction using TBS-protected silyl enol ether 2a’ in the absence of water additive furnished silyl enol ether product 3a’ in 61% yield accompanied by 22% ketone product 3a, suggesting that 3a’ is an intermediate en route to the ketone product (Figure 2E). Intermolecular kinetic isotope effect studies using 1:1 of 2a’:2a’-d₂ furnished the product 3a’:3a’-d₁ in a 1.1:1 ratio, showing that cleavage of the C-H bond is not the rate-determining step (Figure 2F).

Figure 2.

Mechanistic studies of C1-ketonylation of carbohydrates. ^aThe dip at around 350 nm is the instrumental artifact as the instrument switches the light sources. ^bAn additional 22% of ketone product 3a was obtained as well. ^cDFT calculations were performed at the M06/SDD-6-311+G(d,p)/SMD//B3LYP-D3/SDD-6-31G(d) level of theory using a simplified model of the glucosyl radical(1), where OMe groups were used in place of the OAc groups at the C3, 4, and 6 positions of the pyranose ring.

DFT calculations showed that α-radical addition to the silyl enol ether via transition state TS1α to form radical intermediate IIIα, is 1.4 kcal/mol more favorable than the formation of the β anomer IIIβ via TS1β (Fig. S11). This data corroborates the experimental results and literature reports.^{8, 42-44} The stereoselectivity originates from the more favorable chair-like conformation of TS1α, in which the silyl enol ether approaches from the axial position of the glycosyl radical. On the other hand, the chair conformer of the β-radical addition transition state TS1β′, where the silyl enol ether approaches from the less favorable equatorial position (ΔG^‡ = 13.4 kcal/mol, Fig. S12), is less stable. The most stable conformer of the β-radical addition transition state has a twist-boat geometry (TS1β), and is 1.4 kcal/mol higher in energy than TS1α. Although the polarity-mismatched addition of nucleophilic radicals to electron-rich alkenes is rare, DFT calculations showed that the addition of nucleophilic anomeric radicals to silyl enol ethers is energetically feasible (ΔG^≠ = 10.8 kcal/mol) and the formation of stabilized α-OTMS alkyl radical IIIα drives the stereo-determining, irreversible C-C bond-forming event (Fig. S11).⁷¹ Subsequent silyl enol ether formation from radical intermediate IIIα could proceed through at least four distinct reaction pathways: (P1) single electron transfer (SET) from IIIα to [Pd^I]Br to form carbocation followed by deprotonation; (P2) recombination of radical IIIα with [Pd^I]Br to form Pd^II intermediate IV followed by β-hydride elimination; (P3) a palladoradical β-H-atom abstraction;^72-81 and (P4) bromine atom transfer from [Pd^I]Br to IIIα followed by HBr elimination (Figure 2G). DFT calculations showed that P1 is highly disfavored with ΔG = 66.6 kcal/mol with respect to IIIα (Fig. S14). For the P2, the recombination of radical IIIα with [Pd^I]Br to form Pd^II intermediate IV is endergonic by 4.2 kcal/mol (Fig. S11). Here, the [Pd^II]-C bond is weakened by steric repulsions between the tertiary carbon center and the large-bite-angle of the Xantphos ligand (BDE = 11.6 kcal/mol, Table S2). From IV, subsequent β-hydride elimination (TS4) requires a relatively higher barrier (ΔG^‡ = 16.8 kcal/mol with respect to IIIα) because one of the P-arms of the Xantphos ligand needs to be dissociated prior to the β-hydride elimination.⁸² Consequently, intermediate IV is prone to undergo a backward reaction to form radical IIIα and [Pd^I]Br, especially under the photoexcitation conditions. DFT calculations also showed that the palladoradical β-H-atom abstraction (P3) has a higher energy barrier with ΔG^‡ = 22.3 kcal/mol with respect to IIIα. These computational results implied that the rate-determining step of both the pathways P2⁸³ and P3⁸⁴ involve the C-H bond cleavage. However, the intermolecular KIE studies gave a non-first order KIE of 1.1 (Figure 2E),⁸⁵ suggesting that mechanistic pathways P2 and P3 are unlikely. On the other hand, the non-first order KIE is more in line with the bromine atom transfer followed by a rapid H-Br elimination (P4).^86-88 Although we failed to locate the transition state of the bromine atom transfer, DFT calculations showed that the formation of the α-OTMS alkyl bromide from [Pd^I]Br and radical IIIα is endergonic by only 4.6 kcal/mol, which is energetically feasible.

While a precise reaction mechanism awaits further study, a plausible catalytic cycle is shown in Figure 2. A photoexcited [Pd⁰]* species abstracts the bromine atom from 1-bromosugar 1 to afford a [Pd^I]Br complex and 1-glycosyl radical intermediate IIa, which is in equilibrium with IIb under visible-light irradiation conditions. Guided by a stereoelectronic effect,^{8, 42-45} II adds to silyl enol ether 2 with a high level of α-selectivity, furnishing intermediate IIIα. Subsequent bromine atom transfer followed by H-Br elimination liberates [Pd⁰] and enol ether 3’’, which is hydrolyzed to deliver the desired α-selective C1-ketonylated product.

Conclusion

In summary, we reported a visible-light-induced excited-state palladium-catalyzed α-selective radical C-glycosylation of carbohydrates using readily available 1-bromosugars and silyl enol ethers. Preliminary experimental and computational studies of the reaction mechanism suggested a non-chain radical mechanism. The reaction features excellent stereoselectivity, broad substrate scope, and mild conditions that tolerate various functional groups and complex molecular structures. The resulting α-ketonylated products could serve as precursors for a range of valuable C-glycosides. This general and α-selective strategy, which allows facile access to both α- and β-glycosides, should find applications in drug discovery and chemical biology research.