Rh(I)–Bisphosphine-Catalyzed Asymmetric, Intermolecular Hydroheteroarylation of α-Substituted Acrylate Derivatives

Abstract

Asymmetric hydroheteroarylation of alkenes represents a convenient entry to elaborated heterocyclic motifs. While chiral acids are known to mediate asymmetric addition of electron-rich heteroarenes to Michael acceptors, very few methods exploit transition metals to catalyze alkylation of heterocycles with olefins via a C–H activation, migratory insertion sequence. Herein, we describe the development of an asymmetric, intermolecular hydroheteroarylation reaction of α-substituted acrylates with benzoxazoles. The reaction provides 2-substitued benzoxazoles in moderate to excellent yields and good to excellent enantioselectivities. Notably, a series of mechanistic studies appears to contradict a pathway involving enantioselective protonation of a Rh(I)–enolate, despite the fact that such a mechanism is invoked almost unanimously in the related addition of aryl boronic acids to methacrylate derivatives. Evidence suggests instead that migratory insertion or beta-hydride elimination is enantiodetermining and that isomerization of a Rh(I)–enolate to a Rh(I)–heterobenzyl species insulates the resultant α-stereocenter from epimerization. A bulky ligand, CTH-(R)-Xylyl-P-Phos, is crucial for reactivity and enantioselectivity, as it likely discourages undesired ligation of benzoxazole substrates or intermediates to on- or off-cycle rhodium complexes and attenuates coordination-promoted product epimerization.

Introduction

Catalytic, enantioselective addition of a C–H bond of a heterocycle across an alkene represents a conceptually simple and atom economical method for the preparation of elaborated heterocyclic scaffolds. This concept has been implemented in a formal sense in the asymmetric Friedel–Crafts alkylation of electron-rich heteroarenes, such as indoles, with Michael acceptors.¹ Yet methods exploiting transition metals to mediate asymmetric hydroheteroarylation (HH) of alkenes via a C–H activation, insertion sequence remain quite elusive.^2,3 This deficiency is somewhat surprising given the diverse methods for asymmetric hydroarylation of olefins with activated arenes⁴ or with arenes containing directing groups for C–H functionalization.⁵ In the early 2000s, Bergman and Ellman pioneered the achiral, intramolecular HH of unactivated alkenes with a Rh(I)–phosphine catalyst.^3a This discovery was expanded in a great body of work to the intermolecular HH reaction of alkenes⁶ and to several discrete asymmetric, intramolecular HH reactions.⁷ In 2012, Shibata provided an early example of an asymmetric intermolecular HH reaction mediated by a transition metal (TM):⁸ an Ir(I)–SDP-catalyst promotes the branched-selective alkylation of N-benzoylindole and styrene in 42% ee (Figure 1, eq 1). Notably, alkylation occurs at the indole 2-position, whereas functionalization typically proceeds at the 3-position under Friedel-Craft conditions.¹ Though only modestly selective, Shibata’s example foreshadows that TM-catalyzed HH may eventually serve as a selective and general complement to established methods using chiral acids. Indeed, Hartwig and Sevov described in short succession the asymmetric HH of norbornene with diverse heterocycles using a chiral Ir(I) catalyst (Figure 1, eq 2).⁹ Most recently, Hou and co-workers reported the enantioselective alkylation of 2- substituted pyridines with unactivated, terminal alkenes using a chiral, half-sandwich scandium complex. (Figure 1, eq 3).¹⁰

Figure 1.

TM-catalyzed, asymmetric, intermolecular hydroheteroarylation reactions previously reported in the literature.

While the work of Hartwig and Hou provides a powerful proof of concept, room for complementary asymmetric HH methods remains. Specifically, we sought to expand the scope of the olefin coupling partner. Hartwig’s HH reaction is demonstrated only with the strained cyclic alkene, norbornene,⁹ and Hou’s pyridine alkylation appears limited to relatively unfunctionalized, electron neutral alkenes.¹⁰ Herein, we describe a Rh(I)-catalyzed asymmetric alkylation of benzoxazoles with acrylate derivatives (Figure 2, eq 4). To our knowledge, this work represents the first example of an enantioselective, transition-metal-mediated, intermolecular HH of acyclic, electron-deficient alkenes. Moreover, the described reaction makes products of potential medicinal value; isosteres for purine bases and certain amino acids, 2-substituted benzoxazoles are known to exhibit tremendous biological activity.¹¹

Figure 2.

Our HH reaction of benzoxazoles and α-substituted acrylates and precedent inspiring its development.

We found inspiration for the described HH reaction in chemistry developed by Chang et al.^3j This group reported the HH of acrylates and acrylate derivatives with benzheterocycles or pyridine oxides (Figure 2, eq 5). Chang et al. invoke catalysis by a Rh(I)–acetate species—acetate counterion mediates C–H activation, while liberated acetic acid protonates an eventual C–Rh bond (Figure 2, eq 6). We envisioned that use of a substituted acrylate in a system related to Chang’s would enable the asymmetric preparation of branched products (Figure 2, eq 7). Notably, the Rh(I)–dppe system used by Chang et al. lends itself to enantioselective modification: in contrast to relatively scarce chiral cyclopentadienyl ligands ubiquitous in Rh(III) catalysis,^5d,5e,5h chiral bisphosphine ligands abound.¹²

Despite the overt similarity between the known and proposed reactions, several complications could accompany the envisioned asymmetric method. The mechanism proposed by Chang invokes protonation of Rh–enolate II (Figure 2).^3j While protonation of C-bound II could provide enantioenriched products, protonation or ligand exchange of O-bound III at oxygen would give racemic product. Additionally, β-H elimination and dissociation of resultant conjugated alkene would furnish undesired Heck product.^3j Indeed, success of Hartwig’s and Hou’s chemistry may be understood in light of these anticipated difficulties; the privileged nature of norbornene in eq 2 (Figure 1) likely derives in part from the fact that presumed intermediate I cannot undergo β-H elimination. Hou’s pyridine alkylation (Figure 1, eq 3) is also presumably more insulated from β-H elimination than a Rh(I)-system, since the enhanced thermodynamic stabilization of metal–hydrogen bonds over metal–carbon bonds is smaller for early TMs than for late ones.¹³

While we were aware that the described pitfalls could plague our desired reaction with low stereo- or product-selectivity, work by Reetz, Genet, and others offered hope that these obstacles would not be insurmountable.¹⁴ These groups report that a Rh(I)–chiral bisphosphine system mediates the asymmetric hydroarylation of α-substituted acrylates with boronic acid derivatives (Figure 2, eq 8). Importantly, this reaction is presumed to intercept analogous Rh–enolate intermediate IV.^14b−14d Similar opportunities for stereochemical scrambling or Heck reactivity exist for IV as for our presumed Rh–enolate II. Yet these pathways must not be competitive in the described systems, since saturated products are obtained in good to excellent enantioselectivities.¹⁴ These groups invoke asymmetric protonation of Rh–enolate IV or O-bound Rh-isomer to explain high product enantioselectivities,^14,15 but aside from Genet et al.,^14e none provide rigorous mechanistic evidence in favor of this claim (vide infra).

Results and Discussion

Encouraged that our asymmetric HH could succeed, we decided to begin by investigating mechanistic aspects of the parent, achiral reaction (Figure 2, eq 5). The first question we sought to address was the role of the CsOAc. If, as Chang and co-workers postulated, CsOAc serves to generate a Rh(I)–acetate catalyst in situ, then perhaps the same reactivity could be accomplished with a premade Rh(I)–acetate catalyst. Chatani and co-workers have indeed observed that [Rh(cod)OAc]₂ can be used in place of a KOAc–[Rh(cod)Cl]₂ system in the directed hydroarylation of acrylates with 8-aminoquinoline-derived benzamides.^16,17 We prepared [Rh(cod)OAc]₂ by treating [Rh(cod)Cl]₂ with KOAc in refluxing acetone according to a known procedure.¹⁸ Recrystallization from EtOAc provided X-ray quality crystals of the air-stable, orange solid. These were characterized by X-ray crystallography to provide what we believe is the first reported crystal structure of the complex (see Supporting Information).¹⁹ As predicted, [Rh(cod)OAc]₂ performs with equal efficiency as Chang’s in situ generated catalyst in the HH of several benzheterocycles 1 with tert-butyl acrylate (Chart 1). CsOAc thus appears to serve primarily as an acetate source in Chang’s chemistry.

Chart 1. HH Using Chang’s Established Conditions (Red)^3j or [Rh(cod)OAc]₂ (Blue)a,b.

^a To ensure uniformity for comparison, all reactions were performed by the first author.

^b Yields were determined with respect to 4,4′-di-tert-butylbiphenyl (DTBB) by ¹H NMR of the reaction mixture.

With [Rh(cod)OAc]₂ in hand, we screened the asymmetric HH of ethyl methacrylate (3a) and 4-methylbenzoxazole (1c) (Table 1), since this heterocycle proved most reactive in the achiral reaction with tert-butyl acrylate (Chart 1, vide supra). Ligands resembling dppe were chosen at the outset. In PhMe at 120 °C, 1c and 3a react in the presence of a Rh(I)–prophos (L1) catalyst to deliver α-substituted product 4ca in quantitative yields and 29% ee (Table 1, entry 1). Ees remain modest with Chiraphos (L2) and Me–Duphos (L3) (entries 2 and 3). Significant improvement in ee is achieved with Binap (L4), but yields of 4ca suffer. Since bite angle is known to have a pronounced effect on reaction selectivity and efficiency,²⁰ we examined Binap derivatives, Synphos (L5) and Segphos (L6), whose bite angles we hoped would compare more favorably to dppe.^21,22 Gratifyingly, a Rh(I)–Segphos system delivers product 4ca in acceptable 56% yield, and good selectivity (85% ee, entry 6). A twofold increase in acrylate concentration further increases reactivity, providing comparable yields in 24 h to what is obtained in 60 h with lower acrylate concentrations (entries 6–9). Concurrently, a solvent and temperature screen (entries 9–17) revealed acetonitrile (CH₃CN) to be optimal for selectivity (95% ee, entry 11). Combining results, execution of the HH reaction in CH₃CN with 8 equiv of acrylate 3a and 5 mol % rhodium dimer provides satisfactory yields of 4ca in excellent enantioselectivity (entry 18).

Table 1. Initial Reaction Optimization.

entry	ligand	solvent	equiv 3a	T (°C)	time (h)	4caa (%)	eeb (%)
1	L1	PhMe	4	120	60	100	29
2	L2	PhMe	4	120	60	95	–47
3	L3	PhMe	4	120	60	39	57
4	L4	PhMe	4	120	60	9	–78
5	L5	PhMe	4	120	60	20	84
6	L6	PhMe	4	120	60	56	85
7	L6	PhMe	4	120	24	19	89
8	L6	PhMe	6	120	24	29	85
9	L6	PhMe	8	120	24	58	77
10	L6	PhMe	4	100	24	17	88
11	L6	CH3CN	4	100	24	15	95
12	L6	TFE	4	100	24	<5	16
13	L6	DCE	4	100	24	<5	95
14	L6	DME	4	100	24	6	91
15	L6	DMF	4	100	24	22	88
16	L6	PhCF3	4	100	24	10	95
17	L6	o-DCB	4	160	24	7	17
18c	L6	CH3CN	8	100	24	58	95

^aDetermined with respect to DTBB by LC analysis of the reaction mixture on a chiral stationary phase.

^bDetermined at the same time as % yield by LC analysis of the crude reaction mixture on a chiral stationary phase.

^cReaction conducted with 5 mol % [Rh(cod)OAc]₂ and 10 mol % L6.

Although we were pleased with this result, we anticipated that reaction efficiency would need to be further improved in order to extend the substrate scope to less reactive heterocycles. For instance, when benzoxazole 1a is reacted under the conditions shown in entry 2 of Table 1 (which provide nearly quantitative yields of 4ca), no discernible product 4aa is obtained (eq 9). Before refining our conditions, we sought to understand what made 4-methylbenzoxazole (1c) so much more reactive than its unsubstituted or 6-substituted counterparts (Chart 1, 1a–1b, and 1d). Yields displayed in Chart 1 fail to adequately capture this striking reactivity difference—while reaction of 1c is complete in 3 h, reaction of 1a, 1b, and 1d stall at about 50% after 60 h. To gain insight into this disparate reactivity, we performed two competition experiments—one between 1b-D and 1c-H (Figure 3 and eq 10),²³ and one between 1b-H and 1c-H (eq 11).

9

Figure 3.

¹H and ²H NMR of competition experiment between 1c-H and 1b-D in PhMe implicates reversible C–H activation.

From the former, the following significant observations are made: (a) crossover substrates 1b-H and 1c-D are observed by ¹H and ²H NMR (Figure 3); (b) ²H is incorporated into the alkyl backbone of both products 2b and 2c (eq 10); and (c) ²H is incorporated predominantly at the β-position of both products (eq 10). From this data, we propose a mechanistic cycle similar to that offered by Chang et al. (Figure 4).^3j,24 A Rh–acetate catalyst mediates reversible C–H activation of heteroarene 1 (observation a) to provide Rh–heteroaryl complex V. Migratory insertion (MI) across the terminal acrylate (R = H) furnishes Rh–enolate VI, which isomerizes via a β-H elimination, hydrorhodation sequence to heterobenzyl-Rh VIII (observation c). Protonation appears to occur predominantly from VIII (or the N-bound isomer, vide infra). Protonation likely proceeds via an outer-sphere mechanism (observation b), but an inner-sphere mechanism after D–H exchange cannot be ruled out.

Figure 4.

Proposed mechanistic cycle for the HH of terminal (R = H) or α-substituted (R ≠ H) acrylates.

10

Competition between 1b-H and 1c-H provides further mechanistic insights (eq 11). When reactive 1c and sluggish 1b (Chart 1) are subjected to the standard conditions, products 2b and 2c form in roughly equal rates (eq 11). We rationalize the identical rates of formation of 2b and 2c in one of two ways, both of which invoke the different ligating abilities of 1b and 1c. Given that C–H activation is reversible, one explanation assumes that there exists one or more irreversible steps before the turnover-limiting step (TLS) of sluggish substrate 1b.²⁵ In the context of the mechanism shown in Figure 4, we assume that MI is irreversible and therefore product determining and that protonation of 1b-derived intermediates VI or VIII is turnover limiting. Sluggish protonation of 1b-derived VI or VIII is understood by invoking coordination of the heterocycle to rhodium in 1b-derived intermediate VI. Ligation blocks a free coordination site necessary for either protonation of VI or isomerization to VIII via β-H elimination. While unhindered azoles such as 1b, 1a, and 1d can presumably bind in the fashion described, A[1,3]-strain would disfavor analogous coordination of 1c-derived IX, accelerating the reactivity of 1c relative to its unsubstituted counterparts. Indeed, ¹⁵N NMR studies suggest that bulky substitution adjacent to the coordinating nitrogen of various oxazoles impedes their coordination to Rh(II)-complexes.²⁶ To sum up, then, so long as the C–H activation, MI sequence proceeds at roughly equal rates for both substrates, products 2b and 2c will form in a one-to-one ratio, since all catalyst will eventually funnel to 1b-derived VI.

11

In perhaps a more simple explanation, strongly coordinating 1b (and 1a and 1d) but not weakly coordinating 1c acts as a competitive ligand toward important intermediates on or off the catalytic cycle, slowing catalysis of both 1b and 1c.

Although it would be difficult to discriminate between these two explanations—one invoking an intramolecular coordination event and one invoking an intermolecular coordination event—both suggest similar avenues for reaction optimization. Specifically, if deleterious coordination of the heteroarene were responsible for low reactivity of 1a–1b and 1d, then perhaps it could be discouraged by increasing the bulk of the bisphosphine ligand. We were optimistic that increasing ligand bulk might offer additional advantages. A congested coordination environment could also encourage a difficult MI event for steric reasons, since MI necessarily reduces the metal coordination number by one.²⁷

To this end, we sought to further optimize the reaction of ethyl methacrylate (3a) and 1c by screening bulky Segphos derivatives (Table 2). While DTBM-Segphos (L8) is fairly unreactive (entry 3), DM-Segphos (L7) improves yields by about 20% relative to Segphos (Table 2, entries 1 vs 2). With the arene held constant, exploration of the phosphine backbone revealed CTH-(R)-Xylyl-P-Phos (L11) to be a superior ligand.²⁸ It provides quantitative yield of product 4ca in excellent enantioselectivity after 24 h (entry 6). A control reaction confirms that the acetate counterion is crucial for reactivity—no product is obtained under optimal conditions when [Rh(cod)Cl]₂ is used.²⁹

Table 2. Reaction Optimization with Second Generation, Bulky Bisphosphine Ligands.

^a,bSee footnotes for Table 1.

^cWith 2 mol % [Rh(cod)OAc]₂, 4 mol % L8, 4 equiv 3a in PhMe at 120 °C for 60 h: these conditions give 4ca in 56% yield and 85% ee when L6 is used as a ligand.

With these second-generation conditions in hand, we sought to examine the substrate scope of our HH reaction (Chart 2).³⁰ Variation of the ester group provides products 4ca–4cc in excellent yields and selectivities. Methacrylonitrile (3d) participates in moderate yield and good enantioselectivity. The HH reaction is also tolerant of diverse acrylate backbones, although α-substitution appears crucial—racemic product 4ce is obtained in low yield from the reaction of 1c and ethyl crotonate (3e). Acrylates with benzyl, n-butyl, and sterically bulky isobutyl substituents at the α-position react in good yield to give products 4cf–4ch in very high enantioselectivities despite the opportunity for β-H elimination into the alkyl backbone. Dimethyl itaconate (3i) provides good yields of functionalized product 4ci albeit in modest enantioselectivity. Acrylate 3j containing a protected alcohol reacts without difficulty to give silyl ether 4cj in excellent enantioselectivity.

Chart 2. Scope of the Rh(I)–P-Phos-Catalyzed HH of Benzoxazoles and Methacrylate Derivativesa,b.

^a Isolated yields after column chromatography on silica gel.

^b Ees of isolated products determined by LC analysis on chiral stationary phase.

^c Reaction run for 24 h.

^d Yield determined with respect to 4,4′-di-tert-butylbiphenyl by LC analysis of the crude reaction mixture on a chiral stationary phase.

^e Reaction run for 80 h.

^f Yield determined with respect to 4,4′-di-tert-butylbiphenyl by ¹H NMR of the crude reaction mixture.

^g Ee determined by LC analysis of the crude reaction mixture on a chiral stationary phase.

Notably, it was found that addition of 25 mol % CsOAc is necessary to promote reactivity for these more hindered acrylates—indeed, no product is obtained from the reaction of benzyl-substituted 3f in its absence (Chart 2).³¹ While the beneficial effect of CsOAc is not fully understood, acetate rather than cesium ion appears to be responsible for the yield improvement, since no product is obtained from the reaction of 3f and 1c when CsI is used in the place of CsOAc.

Finally, and much to our gratification, variation of the benzoxazole backbone is possible with bulky P-Phos ligand L11. Unsubstituted benzoxazole 1a reacts smoothly; chloro- and fluoro-products 4ea–4fa are assembled in high ees albeit in diminished yields. Isomeric methoxy products 4ga–4ha are obtained in moderate yield and moderate to high enantioselectivities. While addition of 25 mol % CsOAc also appears to accelerate reactions with these benzoxazole substrates, its effect is less pronounced (4aa, 50% vs 67%). The HH reaction is not without limitations. Acrylates substituted with aryl or secondary alkyl groups do not participate effectively, nor do α,β-disubstituted acrylates or acrylates containing β-leaving groups (Figure 5).

Figure 5.

Acrylates that do not provide product in the HH reaction with benzoxazoles.

At this point in our studies, we wanted to better understand the origin of enantioselectivity of our HH reaction. Asymmetric protonation of a Rh–enolate (e.g., IV or O-bound isomer, Figure 2, eq 8) is classically invoked as the enantio-determining step of the Rh(I)–bisphosphine-mediated addition of boronic acids to α-substituted acrylates, although mechanistic evidence is sparse.¹⁴ We chose to test plausibility of this enantio-determining step with a labeling study using deuterated 1c (1c-D) (Figure 6, eq 12). Were our HH mechanism to proceed via protonation of a Rh-enolate (e.g., II or III, Figure 2; or VI, Figure 4), then we should see D-incorporation at the α-position of product 4ca, since 1c is the terminal proton source. Contrary to this expectation, reaction of 1c-D with 3a to 42% conversion under standard conditions provides product 4ca, in which D is incorporated exclusively at the β-position (eq 12). 1c is recovered with 33% H incorporation, consistent with a reversible C–H activation event. The proton source responsible for formation of 1c-H in eq 12 is presumably solvent: indeed, when the experiment is repeated in CD₃CN, virtually no H–D exchange in 1c-D is observed (eq 13). All ²H from 1c-D is accounted for in product 4ca, since CH₃CN cannot serve as a competitive proton source (eq 13). β-deuterium incorporation in 4ca does not likely arise from in situ generation and subsequent preferential reaction of β-deutero 3a, since the reciprocal reaction of 1c-H and 3b-d₈ gives 4ba with ¹H-incorporation at the β-position exclusively (eq 14).

Figure 6.

Labeling experiments rule out a mechanism involving enantioselective protonation of a rhodium enolate.

These labeling studies provide considerable insight into the reaction mechanism. First, they give grounds for dismissal of several possible elementary steps. For instance, protonation of a Rh–enolate cannot be enantiodetermining, as protonation takes place predominantly at the β- rather than the α-position.

The labeling study also seems to contradict a mechanism involving migratory insertion of a Rh(III)–heteroarene (in a 3,2 sense) or a Rh(III)–hydride (in a 2,3 sense) across acrylate 3 followed by reductive elimination to form a C–H or C–C bond respectively—this mechanism, too, would deliver products deuterated at the α- not the β-position.³² To account for the results of our labeling experiment, then, we propose a mechanism analogous to that proffered by Chang and co-workers for the hydroheteroarylation of terminal acrylates (Figure 4, R ≠ H).^3j Reversible C–H activation liberates a molecule of acetic acid and gives a Rh–heteroaryl complex V, which undergoes MI across the acrylate. At this point, a β-H elimination, hydrorhodation sequence isomerizes resultant Rh–enolate VI to alkyl–Rh VIII, which is protonated by acetic acid, regenerating RhOAc complex.

We believe that the proposed isomerization event is crucial for the high enantioselectivities obtained in our reaction. In our preferred mechanism, enantiodetermining MI delivers C-bound Rh–enolate X in a stereodefined fashion (Figure 7). One might imagine that C-bound X could equilibrate with O-bound isomer XI(1–2). Protonation or ligand exchange of XI on O would deliver racemic product, and ees would suffer to the extent that this path is operative. Isomerization of Rh–enolate X to isomer XII, then, insulates the α-stereocenter from epimerization, as long as isomerization is stereospecific. Stereospecificity is guaranteed if the β-H elimination, hydrorhodation steps take place from the same face of alkene XIII, or said another way, if Rh–H intermediate XIII stays bound to the alkene in a sigma fashion. Indeed, β-H-elimination, hydrometalation sequences mediated by late transition metals have been shown to preserve with high fidelity the stereochemistry set by MI events.^4m

Figure 7.

Rationale for isomerization of a rhodium enolate intermediate.

This mechanism may also help explain why α-substituted acrylates are privileged substrates for our HH reaction and perhaps even for the Rh(I)–bisphosphine-mediated asymmetric hydroarylation reported by Darses and others.¹⁴ When an α-substituted acrylate is used, C-bound Rh–enolate X is tetrasubstituted (Figure 7), and O-bound isomer XI experiences significant allylic strain, either between the ester OR group and the heterobenzylic carbon (red, XI-1) or between rhodium and the α-R substituent (blue, XI-2). Sterics may thus discourage formation of XI and promote isomerization to less hindered trisubstituted alkyl rhodium XII. Trisubstituted XII is further stabilized as the heterobenzyl complex. Protonation or ligand exchange may be facilitated by isomerization to Rh–enamido complex XIV.³³

Final evidence for our proposed mechanism is provided by epimerization studies (Figure 8). We wanted to know why the reaction of 1c appeared significantly more selective than the reaction of other benzoxazole substrates, particularly 1h. We speculated that epimerization over the long reaction time might be partially responsible, but we struggled to rationalize why 4ha would epimerize more quickly than other products: the most simple racemization pathway that can be imagined is deprotonation–reprotonation of the α-stereocenter by an acetate–acetic acid couple. Yet electronics of the benzoxazole backbone should not affect acidity of the remote stereocenter. Nevertheless, we resubjected low (4ha), intermediate (4ga), and high (4ca) ee products to the reaction of 1c and an appropriate acrylate (Figure 8, eqs 15–17). When low ee product 4ha is resubjected to the reaction of 1c and 3a under standard conditions, it is indeed found to epimerize to 50% ee (eq 15). In contrast, the ee of product 4ca drops to only 93% ee when it is resubjected to the reaction of 1c and benzyl methacrylate 3c under identical conditions (eq 17).³⁴ Yet epimerization does not appear to be solely responsible for the low ees of 4ha, since intermediate ee product 4ga also shows significant stereochemical scrambling under the reaction conditions (eq 16).

Figure 8.

Epimerization experiments of 4ha, 4ga, and 4ca.

That rates of epimerization of product 4 depend crucially on the benzoxazole backbone challenges an epimerization mechanism via traditional base-assisted deprotonation of the α-stereocenter. Tenuousness of this racemization pathway is reinforced by the fact that product 4ha epimerizes at the same rate in the presence or absence of added base (eq 15)³⁵ and that CsOAc alone fails to epimerize product 4ha even after prolonged heating (data not shown).

In light of insights gained from labeling studies in eqs 12–14, we wondered whether epimerization takes place by the microscopic reverse of the mechanism proposed in Figure 7: coordination of the benzoxazole nitrogen to rhodium acidifies the heterobenzylic H of product 4, which is abstracted by acetate (Figure 9, step 1).³⁶ Resultant Rh–enamido complex XVI, which is in equilibrium with C-bound XVII (step 2), isomerizes back into the acrylate backbone via a series of β-H-elimination, hydrorhodation events (steps 3–5) to eventually give O-bound Rh–enolate XX. Enolate XX is shown as, but need not exist as, the rhodacycle. Protonation or ligand exchange of XX at oxygen epimerizes the α-stereocenter of product 4 (step 6).³⁷ While intermediate XVII is shown with a specific stereochemistry at the carbon bearing rhodium, this is only intended to illustrate that no stereochemical scrambling of the α-stereocenter occurs prior to formation of O-bound XX if alkene XVIII remains coordinated to rhodium (i.e., the stereochemistry of the starting material is relayed to the stereochemistry of C-bound XIX).

Figure 9.

Proposed epimerization mechanism.

We tested credence of this mechanism by treating product 4ha (75% ee) with [Rh(cod)OAc]₂ and CTH-(R)-xylyl-P-Phos in CD₃CN (Figure 10, eq 18), since we knew CD₃CN to be a competent proton source (Figure 6, eqs 12–13). If epimerization were occurring via a typical deprotonation–reprotonation sequence at the α-carbon, then we should see ²H incorporation at the α-position of product 4ha. On the other hand, if the epimerization mechanism depicted in Figure 9 were operative, we would see ²H incorporation at both β- and α-positions of product. In accord with our hypothesis, 4ha is isolated from the reaction in eq 18 in 20% ee with significant deuterium incorporation at the α-position and predominant deuterium incorporation at the β-position.

Figure 10.

Epimerization–labeling experiment.

While this data cannot unequivocally debunk a mechanism by which deuteration at the α- and β-positions occurs by independent deprotonation–reprotonation events at vicinal carbons, the level of D incorporation at the α-position of product 4ca strongly suggests that the two incorporation events are coupled by a common intermediate. Specifically, 21% ²H at the α-position of 4ca does not nearly account for a 55% loss in ee of 4ca (eq 18).³⁸ Thus, 4ca must epimerize by at least one other mechanism besides protonation. We propose that Rh–enolate intermediate XX has two opportunities to scramble α-stereochemistry. It can, as already discussed, protonate or undergo ligand exchange on oxygen to give enantiomeric product (Figure 9, step 6). Yet protonation is not necessary for epimerization to occur. To the extent that the α-stereochemistry of C-bound XIX is lost in O-bound XX, then isomerization back to the C-bound isomer should be able to deliver diastereomeric complex XXI in which α-stereochemistry is inverted (step 7). A reverse sequence of elimination, addition events relays XXI to enantiomeric product (step 8).

We wondered how the epimerization mechanism depicted in Figure 9 could account for the very different fates of low ee product 4ha and high ee product 4ca when they are resubjected to our Rh–bisphosphine system. Interestingly, when highly enantioenriched product 4ca (95% ee) is treated with rhodium and ligand under identical conditions to those described for 4ha, it also deuterates considerably at the β-position (Figure 10, eq 19). In contrast to 4ha, however, product 4ca epimerizes quite slowly (to 91%) even at high dimer loading, and it shows no discernible ²H incorporation at the α-position. We provide two possible explanations to account for the data in eqs 18–19, but alternatives are possible. As illustrated in Figure 9, deprotonation of 4 gives Rh–enamido complex XVI (step 1). It is possible that A[1,3]-strain between the axial methyl of 4ca and rhodium shortens the lifetime of XVI such that a rapid backward reaction—protonation of XVI—outcompetes isomerization into the acrylate backbone (step 2).

An alternative explanation invokes differential stability of 4ha and 4ca Rh–enolate complexes XX (Figure 9). Whereas coordination of the heterocyclic nitrogen to rhodium could stabilize a 4ha-derived Rh–enolate XX, A[1,3]-strain would prevent analogous stabilization of 4ca-derived XX. In either case, relative coordinating abilities of 4ca and other benzoxazoles appear to crucially influence product epimerization rates. If this is true, then our bulky P-Phos ligand may serve an additional service: it may discourage ligation-promoted racemization.

Summary

In summary, mechanistic insights gained from a known reaction of heterocycles and tert-butyl acrylate^3j have enabled development of an asymmetric, hydroheteroarylation reaction of benzoxazoles and α-substituted methacrylate derivatives. The reaction is mediated by a Rh(I)–acetate precatalyst and bulky bisphosphine ligand, CTH-(R)-xylyl-P-Phos, and it delivers diverse elaborated benzoxazole products in moderate to excellent yields and good to excellent enantioselectivities. Mechanistically, the reaction is thought to proceed via a C–H activation, MI, and protonation sequence in which acetate serves as a proton shuttle. Labeling studies implicate MI as a possible enantiodetermining step, after which stereospecific isomerization to a Rh–heterobenzyl complex insulates the newly formed stereocenter from epimerization. Products that are good ligands for rhodium can epimerize by a reverse sequence: coordination and subsequent C–H activation at the heterobenzylic position provide a Rh–enamido complex. A series of β-H elimination, hydrorhodation events relays the enamido complex to O-bound Rh–enolate, in which α-stereochemistry is lost. Our proposed mechanism differs importantly from those implicated in studies describing the related Rh(I)–bisphosphine-mediated hydroarylation of α-substituted acrylates with boronic acids.¹⁴ These studies invoke protonation of a rhodium enolate as the enantio-determining step of the reaction. Since little mechanistic evidence is provided in these studies, it is conceivable that an isomerization pathway such as ours is operative in these systems. Finally, a bulky bisphosphine ligand is found to be crucial for reactivity and selectivity in our HH reaction, as it likely discourages deleterious coordination of benzoxazole substrates to on- or off-cycle intermediates, accelerates a difficult MI step, and discourages coordination-initiated epimerization. In short, careful mechanistic analysis has enabled the development of an efficient and highly selective catalytic, asymmetric HH of readily accessible reagents to produce chiral compounds of high biological importance.

Supporting Information Available

Experimental procedures; characterization data, ¹H NMR, ¹³C NMR, and HPLC spectra for new compounds; crystallographic data for [Rh(cod)OAc]₂. This material is available free of charge via the Internet at http://pubs.acs.org.

The authors declare no competing financial interest.

Funding Statement

National Institutes of Health, United States