Hydrogen-bond-acceptor ligands enable distal C(sp3)-H arylation of free alcohols

Abstract

The functionalization of C-H bonds in organic molecules is one of the most direct approaches for chemical synthesis. Recent advances in catalysis have allowed native chemical groups such as carboxylic acids, ketones and amines to control and direct C(sp³)-H activation^1-4. However, alcohols, among the most common functionalities in organic chemistry⁵, have remained intractable because of their low affinity for late transition-metal catalysts^6,7. Here we describe ligands that enable alcohol-directed arylation of δ-C(sp³)-H bonds. We use charge balance and a secondary-coordination-sphere hydrogen-bonding interaction-evidenced by structure-activity relationship studies, computational modelling and crystallographic data-to stabilize L-type hydroxyl coordination to palladium, thereby facilitating the assembly of the key C-H cleavage transition state. In contrast to previous studies in C-H activation, in which secondary interactions were used to control selectivity in the context of established reactivity^8-13, this report demonstrates the feasibility of using secondary interactions to enable challenging, previously unknown reactivity by enhancing substrate-catalyst affinity.

Extended Data Fig. 1.. Aryl iodide scope with quaternized cyclobutane alcohol 3q.

Reactions were run according to General Procedure B. The reported yields are for isolated and purified products.

Extended Data Fig. 2.. Poorly performing aryl iodides and alcohol substrates.

Due to the extremely low yields or lack of detectable product formation in most of these reactions, we were unable to isolate analytically pure samples of products except where specifically noted A) Poorly performing aryl iodides. Reactions were run according to General Procedure B with modifications as indicated. Aryl iodide screening was performed using either alcohol 3q (for 1-chloro-2-iodobenzene and 1-bromo-2-iodobenzene) or alcohol 3a (for the remaining aryl iodides) as the substrate. B) Poorly performing alcohol substrates. Alcohol screening was performed using methyl-4-iodobenzoate as the coupling partner. Bolded bonds indicate relative stereochemistry. C) Products isolated in low yields from reactions run with linear alcohols. ^§12 mol% L27 used. ^&Reaction run at 0.1 M. ^#Reactions were run according to General Procedure A using L24 as the ligand. ^Based on crude NMR vs. CH₂Br₂ internal standard. *Reaction run using 0.1 mmol of methyl-4-iodobenzoate as the limiting reagent (0.1 M in DCE) with 10 equivalents of the alcohol substrate.

Extended Data Fig. 3.. Logic of double mutant cycle experiments.

Double mutant cycles are a classic tool used to quantify the energy of a non-covalent interaction of interest in a complex setting, including enzyme-substrate binding and fully synthetic systems such as molecular balances and zippers. The interaction under investigation is perturbed by removing each interacting partner (here the HBD and HBA) both individually (affording a pair of single mutants) and simultaneously (to give the double mutant). Each of these modifications will affect more than just the proposed H-bonding interaction, but the non-cooperative effects can be cancelled by adding and subtracting the energies of the four species using the equation in the figure above. The resulting energy will reflect differences in cooperative interactions between the two interacting partners across the four structures. Ideally, each mutation would fully knock out the interaction under investigation without creating any new interactions with the other partner, in which case the equation would directly report the (de)stabilization due to the interaction under investigation in the parent system. In practice, however, it can be difficult to design modifications which meet these criteria since there are many ways in which the mutated substituents could still interact cooperatively (e.g., other non-covalent interactions like dipole-dipole interactions or steric clash, or via through-bond effects, such as changes in the Lewis acidity of the metal center). Thus, double mutant cycles must be designed carefully to ensure that they provide information about the interaction of interest rather than other cooperative interactions. To address the potential for confounding factors, we have calculated four double mutant cycles each for TS-1, TS-δ−1, TS-cis-γ−1, and TS-trans-γ−1 (see Extended Data Fig. 4 and Figs. S78, S83, and S88 for details, discussion, and tabulated energies and Figs. 4E and S71–S77, S79–S82, and S84–S87 for the actual cycles) employing different mutations designed to have very different potential confounding factors. The qualitative agreement between these models provides strong support for significant stabilization by the proposed H-bonding interaction.

Extended Data Fig. 4.. Summary of computational double mutant cycles examined for benzylic C–H activation.

A) The HBA can be knocked out by replacing the acyl sulfonamide with a carboxylate or an amidate. The amidate preserves the identity of the chelating atom, but may significantly alter the electronics at Pd, whereas the carboxylate is more similar to the acyl sulfonamide electronically but changes the atom bound to Pd. B) The HBD can be knocked out by replacing the DG with a methyl ether or an alkoxide. The methyl ether preserves the charge and nature of the DG, but significantly alters the steric properties of the directing group, whereas the alkoxide significantly alters the electronic character of the DG, but presents a similar steric profile to the alcohol. C) Tabulated data for the four cycles generated by the pairwise combinations of the ligand and substrate mutations described above (the tabulated data is taken from the cycles shown in Figs. 4E and S71–S73). While there are modest quantitative differences in the interaction energy measured for each cycle, they are all in qualitative agreement on the stabilization afforded by the interaction between the alcohol and the sulfonamide. Since the four cycles were chosen to have different potential confounding factors, the qualitative agreement strongly suggests that the observed stabilization is due at least in large part to the proposed H-bonding interaction.

Fig. 1:. Free-alcohol-directed C(sp³)–H activation: value, challenges, and strategy for enabling reactivity.

A) Native functional group directed C–H activation enhances synthetic efficiency. Alcohol-directed C(sp³)–H activation is a significant unmet challenge. B) The weak coordination of alcohols to Pd and their lack of rigidity disfavor formation of an agostic complex and CMD. With commonly used L,X chelating ligands, alcohol coordination is further destabilized by charge separation. Pd(II)-alkoxides readily undergo undesired side reactions such as oxidation and fragmentation. C) (This work) Bis-anionic ligands containing an H-bond acceptor and a CMD-active base enable free-alcohol-directed δ-C(sp³)–H arylations.

Fig. 2:. Model system and ligand optimization.

A) Trace reactivity is observed with Pd(OAc)₂, but is suppressed by added base. B) L,X ligands inhibit ROH-directed reactivity. C) Simple X,X ligands provide similar reactivity to Pd(OAc)₂. D) X,X-HBA ligands allow for high-yielding alcohol-directed C(sp³)–H arylation. Yields in Fig. 2 A–D are determined by NMR relative to CH₂Br₂ as an internal standard. E) Reaction scope of benzylic C(sp³)–H arylation with isolated yields. ^Reaction run at 100 °C for 3 days; *Reaction run at 85 °C for 3 days. ^#NMR yield relative to a CH₂Br₂ internal standard. For information on reaction setup, see General Procedure A in the Supporting Information.

Fig. 3:. Ligand optimization and scope for alcohol-directed δ-arylations of cyclobutanes.

A) Model reaction for δ-arylations of cyclobutane alcohols and the identification of pyridone-triflamide ligand L27. B) Scope studies for the L27-enabled arylation of cycloalkane alcohols with isolated yields. For detailed information on reaction setup, see Supporting Information General Procedure B. ^§Reaction performed on 1 mmol scale. *1.2 equiv. of Ar–I used. ^24-hour reaction time. ^#L24 used instead of L27. Bolded bonds indicate relative stereochemistry.

Fig. 4:. Evidence for H-bonding.

A) Crystal structure of a Pd(II)-L24 complex bound to a transition state analog showing the proposed H-bonding interaction. B) Isoelectronic “HBA-knockout” variants of L12 and L14 are ineffective. C) Methyl ether “HBD-knockout” substrates are unreactive. (a) Pd(OAc)₂ (10 mol%), L24 (12 mol%), p-Tol-I (3 equiv.), AgOAc (2 equiv.), DCE (0.1 M), 90 °C, 48 h; (b) Pd(OAc)₂ (10 mol%), L27 (10 mol%), p-Tol-I (3 equiv.), AgOAc (2 equiv.), DCE (0.05 M), 100 °C, 48 h. D) H-bonding stabilizes the lowest energy transition structure for C(sp³)–H cleavage with substrate 1k (TS-1). E) A computational double mutant cycle provides further evidence that the H-bonding interaction contributes significant stabilization to TS-1. See the Supporting Information for computational details.