Iridium-Catalyzed Oxidant-Free Transfer Dehydrogenation of Carboxylic Acids

U nsaturated carbonyl moieties are often found in bioactive compounds that exhibit a broad range of functions, such as antioxidants, radical scavengers, or covalent inhibitors. 1 In addition, they also serve as valuable intermediates or precursors to access β or remote functionalized products. 2 Among various synthetic strategies, those that directly convert saturated carbonyl compounds to their unsaturated counterparts via dehydrogenation are highly attractive and have been widely utilized in the complex molecule synthesis. 3 However, compared to the diverse ways of desaturating aldehydes, 4 ketones, 5 amides, 6 and esters, 7 direct dehydrogenation of free carboxylic acids remains largely underdeveloped (Scheme 1A). The seminal work of Newhouse employed zinc enediolates as the key intermediate and achieved the Pd-catalyzed dehydrogenation of carboxylic acids with allyl acetate as the oxidant. 8 Later, the Yu group realized an elegant ligand-promoted Pdcatalyzed direct desaturation of carboxylic acids, which employed Ag 2 CO 3 , tert-butyl hydroperoxide, or 1,4-benzoquinone/O 2 as the oxidants. 9 In 2018, Huang and co-workers reported a Cp*Ir-catalyzed desaturation of γ,δ-unsaturated carbonyl compounds to 1,3-dienes using air as the oxidant, 10 and the γ,δ-olefin is critical for forming the key π-allyl intermediate.

On the other hand, the transition-metal-catalyzed dehydrogenation of aliphatic hydrocarbons, particularly promoted by pincer-ligated complexes, 11 represents an oxidant-free approach to transfer two hydrogens from the substrate to an acceptor alkene (Scheme 1B). While this type of catalysis typically does not tolerate polar functional groups, such as carboxylates, due to catalyst inhibition, 12,13 the unique nonoxidative feature of this H 2 -transfer approach motivated us to explore an alternative strategy for dehydrogenation of free carboxylic acids.

Based on our continued interest in carbonyl desaturation and its related transformations, particularly the recent development of an Ir-catalyzed byproduct-free β-alkenylation of ketones with alkynes, 14 here we describe our preliminary development of an Ir-catalyzed transfer-dehydrogenation of carboxylic acids (Scheme 1C). It was hypothesized that the carboxylate moiety first directs oxidative addition of Ir(I) into the β or γ C-H bond, and followed by β-H elimination, the resulting Ir(III) dihydride species then undergoes a hydridetransfer process with an acceptor alkene to deliver the desired unsaturated product and to regenerate the Ir(I) catalyst. It is anticipated that this transfer dehydrogenation approach would avoid strong bases to form enolates and avoid stoichiometric oxidants to turnover the catalyst.

To examine the feasibility of the proposed strategy, N-Me-3indolepropionic acid (1a) was employed as the model substrate. After systematic evaluation of various reaction parameters, the desired α,β-unsaturated carboxylic acid 2a was formed in 78% yield with >50:1 E/Z diastereoselectivity (Table 1, entry 1), with NBE as the hydrogen acceptor, 10 mol % [Ir(COD) 2 ]BArF as the catalyst, 24 mol % tri(3,5dimethylphenyl)phosphine L1 as the ligand, Cs 2 CO 3 as the base, and NaOAc as the additive. A series of control experiments was then performed to understand the role of each component. In the absence of the iridium catalyst, NBE or the ligand, the reaction gave almost no conversion (entries 2, 5 and 7), suggesting their critical roles in this reaction. Substitution of [Ir(COD) 2 ]BArF with [Ir(COD) 2 ]OTf or [Ir(COD) 2 ]NTf 2 led to a decreased yield (entries 3 and 4). In addition, a good yield was still obtained when reducing the NBE loading to 4 equiv (entry 6); but other alkenes or substituted NBEs gave almost no desired product (see Supporting Information). A much lower yield was observed when reducing the ligand loading to 12 mol %, suggesting that a 1:2 metal-to-ligand ratio is likely important for high catalyst efficiency. While the exact reason for the ligand effect remains to be uncovered, compared to the optimal ligand (L1), other monodentate phosphines without the 3,5-dimethyl groups (L2-L4) were significantly less efficient (see Supporting Information); in contrast, L5 with the additional 4-methoxy group gave comparable yield. Moreover, bidentate ligands (L6-L8), except dppf, exhibited almost no reactivity (entry 9 and Table S1), likely due to the sensitivity to the electronic and steric properties of the ligand. A survey of different bases suggested that Cs 2 CO 3 was much better than the others (entries 10 and 11). NaOAc is an effective additive, though it is not essential to the reactivity (entry 12). 15 PhF proved to be a better solvent, and slightly lower yields were observed with toluene and PhCF 3 (entries 13 and 14). Finally, lowering the reaction temperature led to a lower conversion (entry 15).

The scope of the Ir-catalyzed α,β-dehydrogenation was subsequently explored (Table 2A). It should be mentioned that some products were converted to the corresponding methyl esters for convenience of purification. For the indolederived substrates, first we found the substituent on the nitrogen does not significantly influence the reactivity (2a-2c). In addition, the substrates with substituents at the C2, C4, C5, C6, and C7 positions, including methyl (2i and 2j), phenyl (2d), methoxy (2e and 2f), and chloro (2g and 2h) groups, were all compatible in this transformation. Besides the indole framework, substrates bearing aryl rings (2k-2t) are also suitable for this transformation. 3-Arylpropanoic acids containing electron-donating or -withdrawing groups on the arene can afford the desired α,β-unsaturated carboxylic acids (2l-2s) in moderate yields. In general, these reactions were very clean, showing high mass balances. Note that compound 2m is the precursor for the synthesis of vanilloid receptor-1 antagonist RPV-1. 16 Moreover, a styrenyl group (2p) was tolerated under the transfer dehydrogenation conditions, indicating that the hydride transfer process is much faster with NBE than with other alkenes. In addition to simple phenyl rings, products 2t and 2u bearing bicyclic moieties could also be synthesized. Gratifyingly, α-substituted carbox-

Scheme 1. Transition-Metal-Catalyzed Dehydrogenation Reactions Table 1. Selected Reaction Optimization with Substrate 1a a ylic acids can also be used as the substrate (1v); although the yield is low, only the E isomer of the product was observed. The application of 2,3-dihydro-1H-indene-2-carboxylic acid as the starting material resulted in forming indene 2w. Finally, substrates featuring other heteroarenes, such as pyrrole, furan, and thiophene (2x-2z), were competent for this transformation. Note that attempts to desaturate esters and amides were not fruitful under the current reaction conditions (see Supporting Information). 17 When one-carbon elongated carboxylic acids were used as the substrates, this method can be extended to realize the β,γtransfer dehydrogenation (Table 2B). 9c The α,α-disubstituted substrates worked notably better; however, the one without any α-substituent also delivered the desired product (4n), albeit in a lower conversion. This suggests that the Thorpe-Ingold effect likely benefits the C-H activation step in this reaction. The substrates bearing an electron-rich arene gave higher conversion than electron-deficient ones, although the exact reason is unclear. In addition, carboxylic acids bearing diverse functional groups, such as tertiary amine (4c), ethoxy (4d), benzyl ether (4e), phenyl ether (4h) and naphthalene (4i and 4j), could give the desired β,γ-unsaturated carboxylic acids in excellent E/Z diastereoselectivity. Moreover, besides α-dimethyl groups, diethyl-or unsymmetrically α-disubstituted Unless mentioned otherwise, all reactions were carried out with 0.1 mmol of 1 or 3 for 24 h under the standard conditions. All yields are isolated yields. b The subsequent esterification was carried out with MeI (2 equiv), Cs 2 CO 3 (2 equiv), acetone, 60 °C, 4 h. c The recovered amount of the unreacted ester of 1 is given within parentheses. d In PhF (0.33 M). e In 1,4-dioxane (0.33 M). f At 150 °C.

substrates (4l and 4m) also worked. Like in the former α,βdehydrogenation, the β,γ-dehydrogenation reactions typically do not form side products, and the modest yields were mainly due to incomplete conversions. 18 Furthermore, apart from the aryl-substituted carboxylic acids, the cyclohexene-derived substrate 3o is also competent to afford the corresponding diene product 4o in 74% yield.

To explore the potential utility of this method, first a scaleup reaction was carried out, and the desired product was obtained in good yield on a 1 mmol scale (Scheme 2A). In addition, simple transformations of the dehydrogenated products were demonstrated (Scheme 2B). For example, the α,β-unsaturated carboxylic acid (2a) can be converted to the corresponding secondary amide 2aa and tertiary amide 2ab uneventfully. In addition, Michael addition of a thiophenol followed by esterification furnished β-functionalized ester 2ac. Furthermore, treatment of a β,γ-unsaturated carboxylic acid (4a) with N-bromosuccinimide (NBS) led to bromolactonization in 88% yield with excellent diastereoselectivity.

To gain some insight into the dehydrogenation mechanism, several control experiments were next conducted. First, to elucidate the role of NBE, less volatile benzo-fused NBE 5 was used as the hydrogen acceptor (Scheme 3A). As expected, under the standard conditions, the corresponding hydrogenated product 6 was detected in a comparable yield to that of the dehydrogenated product (2a), strongly supporting the idea that NBE serves as the hydrogen acceptor in this reaction. Next, the kinetic orders of the soluble components of the reaction were measured (Scheme 3B). The reaction exhibits a first-order dependence on the concentration of the Ir catalyst and substrate 1a, and a first-order dependence on NBE at low NBE concentration (<0.4 M), which dropped to zero when the concentration of NBE was higher than 0.4 M. These kinetic data suggest that the iridium catalyst, 1a and NBE should be involved in the turnover limiting step (TLS). Moreover, deuterium labeling experiments were conducted. The reaction of the α-deuterated substrate 1o-α-D gave product 2o-α-D with 4% D at the β position and 62% D at the α position (eq 1), while the β-deuterated substrate 1o-β-D gave product 2oβ-D with 31% D at the β position and 8% D at the α position (eq 2). These experiments suggest that C-H activation and βhydrogen elimination are likely reversible processes; otherwise, deuterium loss and scrambling should have not taken place. Subsequently, parallel kinetic isotope effect (KIE) was investigated, which showed a small KIE value (1.6) at the α position and a slightly higher value (2.2) at the β position (eqs 3 and 4). Altogether, these experiments collectively imply that cleavage of the αand β-C-H bonds may occur before the TLS. 19 Finally, the complete reaction profile was calculated by density functional theory (DFT) study, and a tentative mechanism is proposed (Figure 1). The initial ligand exchange of [Ir(COD) 2 ] + with the two phosphine ligands and the substrate is downhill by 17.3 kcal/mol, leading to a stable intermediate (Int1). 20 It is also possible to have NBE coordination to the Ir center to give Int1-nbe under high NBE concentration, which could explain the observed saturation kinetics of NBE. The subsequent sp 3 C-H oxidative addition forms the five-membered iridacycle Int3, followed by Scheme 3. Experimental Mechanistic Studies Journal of the American Chemical Society pubs.acs.org/JACS Communication

the β-H elimination to realize the desaturation process. Both steps are reversible with relatively low barriers, which is consistent with the deuterium labeling results. The reversibility of the dehydrogenation process also suggests the product can serve as a competing hydrogen acceptor (see the Supporting Information). The following NBE-hydrogenation process starts with ligand exchange between the α,β-unsaturated carboxylate ( 8) and the substrate (7), which is thermodynamically favored by 1.4 kcal/mol. The reason is probably because the unsaturated carboxylic acid is notably more acidic than the saturated counterpart (pK a in water: 4.44 for 2k versus 4.66 for 1k 21 ), thus the unsaturated carboxylate is a better leaving group. This also corroborates the observation that adding an unsaturated carboxylic acid to this reaction system does not inhibit the α,β-dehydrogenation process (see the Supporting Information). The subsequent NBE coordination and migratory insertion of the Ir-H intermediate from the exo face of NBE gives Int8. The final sp 3 C-H reductive elimination step requires an overall activation barrier of 28.7 kcal/mol, which is considered as the TLS of the whole transformation. Hydrogenation with the unsaturated carboxylate as the ligand was also considered, albeit with a ca. 3 kcal/ mol higher barrier in the final reductive elimination step (see Supporting Information). The fact that the ligand exchange takes place prior to the hydrogenation step is also supported by the first-order kinetics of the substrate. The overall transformation is thermodynamically favored by 10.3 kcal/mol. In summary, we have realized the first transfer dehydrogenation of free carboxylic acids that also provides a new strategy for carbonyl desaturation. This protocol is redox neutral and strong base free, thus, complementary to all the previous desaturation approaches. Such a nonoxidative reaction condition allows tolerance of diverse functional groups. The mechanistic insights obtained here suggest the sp 3 C-H reductive elimination as a surprising TLS, which could be valuable for designing a more efficient catalytic system in the future. While the current substrate scope is limited, efforts to expand this strategy to more general carboxylic acid and other carbonyl substrates are ongoing.