Dihydropyranone Formation by Ipso C–H Activation in a Glucal 3-Carbamate-Derived Rhodium Acyl Nitrenoid

Abstract

By using (N-tosyloxy)-3-O-carbamoyl-D-glucal 10, which removes the need for a hypervalent iodine(III) oxidant, we provide evidence for rhodium nitrenoid-mediated ipso C–H activation as the origin of a C3-oxidized dihydropyranone product 3. This system may be especially susceptible to such a pathway due to the ease of forming a cation upon hydride transfer to the rhodium-complexed acyl nitrene.

In the C–H insertion chemistry of metal carbenoids (usually rhodium- but also copper- and ruthenium-stabilized), anomalous products can form when hydride transfer to the carbenoid results in a particularly well stabilized cation.¹ For example, Clark and co-workers found acetal byproducts in dirhodium(II)-catalyzed reactions of certain α-diazoketones (eq 1).^1c,i Mechanistic investigations led to the conclusion that hydride transfer, probably to the rhodium center but possibly to the carbenoid carbon (shown below), provided an oxocarbenium ion en route to the acetal. In the area of metalla nitrene chemistry, meanwhile, there have been hints of related reactivity, but examples remain limited.² Herein we report our investigations on a system that is apparently well set up for intramolecular hydride transfer to a rhodium acyl nitrenoid moiety.

(1)

During studies on the dirhodium(II)-catalyzed oxidative cyclization of D-glucal 3-carbamates to oxazolidinone-protected 2-aminoglycosides (e.g., 1→2), we discovered concurrent formation of C3-oxidized dihydropyranones such as 3.³ Our mechanistic hypothesis^3b was that 2 and 3 both stemmed from a common acyl nitrenoid intermediate (5), which could undergo either alkene insertion or C3–H oxidation (Scheme 1). However, the use of an iodine(III) oxidant⁴ in conjunction with the known Lewis acidic properties of Rh(II) carboxylates⁵ left open the possibility that 2, 3, or both might be forming via initial reaction of the hypervalent iodine species at the enol ether π bond. In this Note we provide evidence, including examples where 2 and 3 arise from a pre-N-oxidized substrate in the absence of iodine(III), that both the amino sugar (2) and C3-oxidation (3) products come from rhodium-complexed nitrene 5, and we discuss the formation of 3 by analogy to Clark’s hypothesis (eq 1).

Scheme 1.

Amidoglycosylation and C3 carbonyl formation from an acyl nitrenoid intermediate

^aCombined yield of both anomers after silica gel chromatography.

^bDetermined by ¹H NMR analysis of the crude reaction mixture. Ratios are the average of five experiments.

Our motivation to investigate the origin of 3 under our reaction conditions was reinforced by Kirschning’s thorough studies^4a–c on dihydropyranone formation from glycals by hydroxytosyloxyiodobenzene (Koser’s reagent⁶). This process, which transforms a variety of 3O-protected glycals, including tri-O-acetyl-D-glucal 6, entails initial addition of the iodine(III) oxidant at the glycal alkene. While Koser’s reagent is considerably more reactive than the iodosobenzene used in our reactions, a Kirschning-type route to the 3 formed under our conditions, perhaps with Lewis acid activation of PhIO by dirhodium(II)acetate, was a viable mechanistic option (Scheme 2).⁷ Furthermore, it was possible that 2 could also form through initial olefin activation by the iodine(III) oxidant, followed by cyclization of the carbamate nitrogen and glycosylation.

Scheme 2.

Possible route to dihydropyranone 3 via reaction of the glycal alkene with the iodine(III) oxidant

Initially, we engineered a series of control experiments to examine these options. Neither 2 nor 3 formed from glucal 3-carbamate 1 in the absence of rhodium.⁸ We also subjected non-carbamate glycals 6 and 7a–b⁹ (Figure 1) to our conditions, but observed no dihydropyranone formation, even though substrate 7a contained a 3O-silyl ether, making it particularly prone to Kirschning’s oxidation.^4a Furthermore, there was no sign of oxidative cyclization or C3–H oxidation when we tested glucal- or allal-derived N-tosyl or N-phenyl carbamates 7c–e.⁹

Figure 1.

Control compounds that gave neither dihydropyranone formation nor oxidative cyclization under the amidoglycosylation conditions [PhIO (1.8 equiv), Rh₂(OAc)₄ (0.1 equiv), 4-penten-1-ol (5 equiv), 4Å MS, CH₂Cl₂].

In an alkene activation mechanism, the Rh₂(OAc)₄ catalyst would act as a Lewis acid to increase the electrophilicity of the iodine(III) center.^4d,e,10 Were this the case, other Lewis acids, despite lacking the nitrene-stabilizing capacity of dirhodium(II), should also promote formation of 2 and 3. To the contrary, a number of Lewis acids [Zn(OTf)₂, Sm(OTf)₃, La(OTf)₃] failed to catalyze the reaction of 1.

Since the control experiments indicated that 2 and 3 do not arise through initial C=C activation by the iodine(III) species, we next sought to probe whether 2 and 3 form when a rhodium-complexed nitrene 5 is generated in the absence of any iodine(III) oxidant. The means to conduct such a test came from Lebel and co-workers who have used N-tosyloxycarbamates as a pre-oxidized nitrogen atom source for dirhodium(II)-catalyzed C–H and C=C insertions.¹¹ Both the Fleming and Lebel groups have demonstrated copper-catalyzed variants,¹² and Donohoe has applied a similar strategy, using N-acyloxycarbamates in osmium-catalyzed tethered aminohydroxylations.¹³

We prepared what turned out to be a sensitive N-tosyloxy glucal 3-carbamate 10 from silylene-protected alcohol 8 ^9a (Scheme 3). The intermediate N-hydroxycarbamate 9 was a stable solid. However, attempted tosylation under Lebel’s and Fleming’s conditions (TsCl, Et₃N, Et₂O) provided a complex mixture, from which we isolated primary carbamate 1 as the principal component. Evidently, as the desired N-tosyloxycarbamate 10 formed, it underwent base-induced, metal-free nitrene formation and subsequent hydrogen atom abstraction.^11c,14 Using an alternative procedure (TsCl, pyr, CH₂Cl₂), however, provided clean conversion to the desired 10. This material could be purified by chromatography on silica gel, but attempts to store it neat, even at −20 °C, resulted in decomposition to a black tar, probably indicating formation of O-p-toluenesulfonylhydroxylamine¹⁵ upon decarboxylative loss of the C3 substituent. Nevertheless, 10 was tractable when freshly prepared, purified by chromatography, and used promptly in CH₂Cl₂ solution.

Scheme 3.

Preparation of N-tosyloxy glucal 3-carbamate 10 and 3-azidoformate 11

Upon treatment with dirhodium(II)acetate¹⁶ and K₂CO₃ in CH₂Cl₂, N-tosyloxy glucal 3-carbamate 10 gave a complex reaction mixture, from which we identified dihydropyranone 3 and primary carbamate 1 by comparison with authentic samples (Table 1, entry 1). This shows that the C3 carbonyl of 3 can form from an acyl nitrenoid precursor under conditions where the Kirschning-type mechanism is impossible. The production of 1 under these conditions may occur via the nitrene by hydrogen atom abstraction. Lebel has observed analogous products in reactions of other N-arylsulfonyloxycarbamates. ^11c In the absence of dirhodium(II), decomposition of the starting material 10 occurred, but neither 1 nor 3 was detected by ¹H NMR analysis of the reaction mixture.¹⁷

Table 1.

Results using 10 and 11 as metal nitrenoid or nitrene precursors

Entry	SM	Conditions	%1a	%2 (α:β)a	%3a
1	10	No added alcohol Rh2(OAc)4	10b	N/A	11b (3:1 = 1.1:1b)
2	10	(5 equiv) Rh2(OAc)4	10b	35b (3.0:1b)	13b (2:3 = 2.8:1b)
3	10	(5 equiv) (CuOTf)2•PhH	12	30 (2.4:1)	6
4	11	hνc	-d-	44 (1.5:1)	None
5	11	hνc, Rh2(OAc)4	-d-	20 (2.4:1)	None
6	11	100 °C	None	2 of βe	None
7	11	100 °C, Rh2(OAc)4	5	8 (3.0)	2

^aYields and anomeric ratios determined by ¹H NMR analysis of the crude reaction mixture with mesitylene as an internal standard.

^bThe reported value is the average of three experiments.

^c254 nm, Vycor filter.

^dNot detected, but regions in ¹H NMR spectra of the crude reaction mixture that would have contained resonances for 1 were obscured by other signals.

^eThe α anomer was also detected by ¹H NMR analysis, but the α:β ratio could not be reliably determined due to overlapping signals.

Dirhodium(II)acetate-catalyzed reaction of 10 in the presence of 4-penten-1-ol as a nucleophile provided α and β amidoglycosylation products 2, dihydropyranone 3, and primary carbamate 1 (Table 1, entry 2). The ratio of amidoglycosylation to dihydropyranone formation (2:3) and the α:β ratio of 2 were similar to the results using the primary carbamate 1 with iodosobenzene oxidant (cf. Scheme 1).¹⁸ This is further evidence that both the iodine(III)-mediated process with 1 and the reaction starting from pre-N-oxidized substrate 10 proceed through the same intermediate, the rhodium acyl nitrenoid 5, which partitions between C=C insertion and C3–H activation. Control experiments including 10 and 4-penten-1-ol, but omitting either the Rh₂(OAc)₄ or the inorganic base, consumed the N-tosyloxycarbamate but did not lead to detectable formation of products 1, 2, or 3.¹⁷ Separately, we verified that a sample of dihydropyranone 3, prepared by PDC oxidation of allylic alcohol 8, was stable to the combination of Rh₂(OAc)₄ and K₂CO₃ in CD₂Cl₂, either in the absence or presence of 4-penten-1-ol.

We obtained similar results (Table 4, entry 3), including formation of dihydropyranone 3, with the copper catalyst used in Fleming’s study,^12a suggesting that the analogy between carbenoid and nitrenoid chemistry extends both to rhodium and copper catalysis. Indeed, copper carbenoid reactions corresponding to the process in eq 1 are known.^1k,l,m

Finally, we prepared glucal 3-azidoformate 11 (Scheme 3) via the acyl imidazole under aqueous conditions,¹⁹ and investigated it as a nitrene source. Photolysis without or with Rh₂(OAc)₄ (Table 1, entries 5 and 6) gave the amidoglycosylation product 2, but no trace of 3.²⁰ The thermal reactions led only to very limited amounts of products 1–3.²¹ The catalyst-free thermal reaction (entry 6) gave none of the dihydropyranone 3, while a detectable level of this material did form when Rh₂(OAc)₄ was included. From our experiments with 11, we conclude that the metal-free nitrene undergoes the amidoglycosylation process to 2,^14a but that it does not provide C3 oxidized 3. Also, the rhodium-complexed nitrene is not constituted photochemically, but does arise to a small extent thermally, as suggested by formation of 3 in the reaction of entry 7.

The significance of the findings outlined in this Note lies in clarifying the origin of the ipso-oxidized 3 rather than from the synthetic utility of the process. In fact, the use of either N-tosyloxy glucal 3-carbamate 10 or azidoformate 11 is considerably less effective for the desired alkene insertion process than using primary carbamate 1 with iodosobenzene and Rh₂(OAc)₄ as depicted in Scheme 1.

While our results do not provide detailed information about the nitrene-mediated C3–H oxidation process, possibilities include C–H insertion followed by fragmentation of the resulting four-membered-ring carbamate 12, as suggested in the work of Du Bois.^2a,b Alternatively, particularly in 5 where the C3–H bond is vinylogously anomeric and favorably aligned with the enol ether π system, the oxidation may involve hydride transfer (5→13), leading to the resonance-stabilized cationic portion of zwitterion 13. Given the analogous processes previously identified for dirhodium(II) carbenoids,¹ hydride transfer to the acyl carbonyl, to the rhodium center, or directly to nitrogen (as shown in Scheme 4) are all viable alternatives. Consistent with hydride migration from C3 to the metallanitrene, we have recently reported that dihydropyranone formation is suppressed by inductive and conformational factors when the 4O and 6O protecting groups are electron-withdrawing.^3b

Scheme 4.

Proposed formation of 3 by hydride transfer or C–H insertion from the rhodium acyl nitrene

In conclusion, this study presents evidence that rhodium carbamoyl nitrenoids can undergo ipso carbonyl formation. Certain glycal 3-carbamate substrates are particularly prone to this pathway because of the electron-rich character of the C3–H bond and the stability of the resulting vinylogous oxocarbenium ion. A better understanding of mechanism will aid efforts to improve the chemoselectivity of nitrenoid C=C and C–H insertion processes.

Experimental Section

4,6-O-Di-tert-butylsilylene-3-O-(N-hydroxy)carbamoyl-d-glucal (9)

To a solution of alcohol 8 ^9a (0.501 g, 1.75 mmol) in CH₂Cl₂ (20 mL) was added 1,1’-carbonyl diimidazole (0.425 g, 2.62 mmol). After 2 h at room temperature, the mixture was diluted with CH₂Cl₂ (80 mL) and washed with satd aq NH₄Cl (3 × 80 mL). The organic layer was dried (MgSO₄), filtered, and concentrated, providing the carbonyl imidazole intermediate as a light yellow foam. Without further purification, this material (assumed 1.75 mmol) was dissolved in pyridine (4.0 mL), and hydroxylamine hydrochloride (0.365 g, 5.25 mmol) was added. After stirring 60 h at room temperature, the reaction mixture was diluted with CH₂Cl₂ (80 mL) and washed with water (2 × 80 mL) and brine (80 mL). The organic layer was dried (MgSO₄), filtered, and concentrated. The crude product was chromatographed (30% EtOAc/hexanes, 100 mL SiO₂), yielding N-hydroxycarbamate 9 as a white solid (0.470 g, 78%). mp 110.0 °C; R_f = 0.21 (30% EtOAc/hexanes); IR (thin film) 3315, 1731, 1649 cm⁻¹; ¹H NMR (300 MHz, CDCl₃) δ 7.44 (s, 1H), 6.92 (very br s, 1H), 6.33 (dd, J = 6.0, 1.5 Hz, 1H), 5.34 (dt, J = 7.5, 1.8 Hz, 1H), 4.80 (dd, J = 6.1, 2.1 Hz, 1H), 4.25-4.10 (m, 2H), 4.04-3.85 (m, 2H), 1.05 (s, 9H), 0.99 (s, 9H); ¹³C NMR²² (75 MHz, CDCl₃) δ 160.0 (s), 145.2 (o), 100.3 (o), 74.4 (o), 73.6 (o), 72.8 (o), 65.6 (t), 27.4 (o), 26.8 (o), 22.7 (s), 19.8 (s); HRMS (FAB) m/z calcd for C₁₅H₂₈NO₆Si (M+H)⁺ 346.1686, found 346.1688.

4,6-O-Di-tert-butylsilylene-3-O-(N-tosyloxy)carbamoyl-D-glucal (10)

N-hydroxycarbamate 9 (105 mg, 0.304 mmol) was dissolved in CH₂Cl₂ (1.5 mL) and pyridine (74 µL, 0.92 mmol) was added, followed by p-toluenesulfonyl chloride (89 mg, 0.47 mmol). The solution was stirred at 25 °C during 4.5 h, diluted with CH₂Cl₂ (25 mL), and washed with water (2 × 20 mL) and brine (1 × 20 mL). The organic layer was dried (MgSO₄), filtered, and concentrated. The crude material was chromatographed immediately (20→25→30% EtOAc/Hexanes, 50 mL SiO₂). Because 10 was prone to decomposition when kept neat for extended periods of time, the yield was determined by weight after concentration on the rotovap, followed by 5 min on the vacuum line. The resulting product 10, a colorless foam (−120 mg, 80%), consequently contained traces of solvent, and the reported yield is therefore approximate. After weighing, 10 was dissolved in CH₂Cl₂ (1.0 mL) and used immediately. R_f = 0.52 (30% EtOAc/hexanes); IR (thin film) 3281, 1775, 1736, 1647, 1597 cm⁻¹; ¹H NMR (300 MHz, CD₂Cl₂) δ 8.20 (s, 1H), 7.86 (apparent d, J = 8.4 Hz, 2H), 7.37 (apparent d, J = 8.0 Hz, 2H), 6.31 (dd, J = 6.0, 1.4 Hz, 1H), 5.21 (dt, J = 7.4, 1.8 Hz, 1H), 4.53 (dd, J = 6.1, 2.0 Hz, 1H), 4.16 (dd, J = 9.6, 4.4 Hz, 1H), 4.03 (dd, J = 10.2, 7.4 Hz, 1H), 3.95 (t, J = 9.9 Hz, 1H), 3.87 (td, J = 10.1, 4.5 Hz, 1H), 2.45 (s, 3H), 1.05 (s, 9H), 0.96 (s, 9H); ¹³C NMR²² (75 MHz, CD₂Cl₂) δ 155.9 (s), 147.0 (s), 146.1 (o), 131.0 (s), 130.4 (o), 130.0 (o), 100.0 (o), 75.7 (o), 73.9 (o), 73.4 (o), 66.2 (t), 27.7 (o), 27.2 (o), 23.1 (s), 22.1 (o), 20.2 (s); HRMS (FAB) m/z calcd for C₂₂H₃₄NO₈SiS (M+H)⁺ 500.1774, found 500.1760.

Reaction of 10 including 4-penten-1-ol

Potassium carbonate (96 mg, 0.69 mmol) and Rh₂(OAc)₄ (11 mg, 0.025 mmol) were combined in a 10 mL round-bottom flask and 4-penten-1-ol (125 µL, 1.23 mmol) was added, followed immediately by a solution of freshly prepared N-tosyloxycarbamate 10 (~114 mg, 0.228 mmol) in CH₂Cl₂ (1.0 mL). The carbamate-containing pear-shaped flask was rinsed with CH₂Cl₂ (2 × 1.0 mL) with the rinsings being added to the reaction mixture. The well stirred mixture turned from a blue-green-grey to a purple color within 20–30 min. Stirring was continued 16 h and the purple mixture was filtered through a tightly packed pad of Celite, rising with EtOAc (80 mL). The filtrate was concentrated and the crude material analyzed by ¹H NMR (separately in CDCl₃ and acetone-d₆), comparing to authentic samples of 1, 2, and 3. The 2:3 and 2-α:2-β ratios were best measured in acetone-d₆ from the resonances for H3 of 2-α (δ 4.56), H1 of 2-β (δ 4.86), and H2 of 3 (δ 5.33). The yields were determined by ¹H NMR analysis of the crude in CDCl₃ using the H1 signals for 2-α and 2-β (δ 4.81 and δ 4.69, respectively), the H2 signal for 3 (δ 5.41), and the H3 signal for 1 (δ 5.27) versus mesitylene added as an internal standard. The reported ratios and yields are the averages from three separate runs. ²³