Molecular Recognition in the Selective Oxygenation of Saturated C-H Bonds by a Dimanganese Catalyst

See allHide authors and affiliations

Science  30 Jun 2006:
Vol. 312, Issue 5782, pp. 1941-1943
DOI: 10.1126/science.1127899


Although enzymes often incorporate molecular recognition elements to orient substrates selectively, such strategies are rarely achieved by synthetic catalysts. We combined molecular recognition through hydrogen bonding with C-H activation to obtain high-turnover catalytic regioselective functionalization of sp3 C-H bonds remote from the –COOH recognition group. The catalyst contains a Mn(μ-O)2Mn reactive center and a ligand based on Kemp's triacid that directs a –COOH group to anchor the carboxylic acid group of the substrate and thus modify the usual selectivity for oxidation. Control experiments supported the role of hydrogen bonding in orienting the substrate to achieve high selectivity.

Selective functionalization of C-H bonds at sites in molecules remote from more reactive substituents is a major challenge in synthetic chemistry (1). In biological systems, monooxygenases and fatty acid desaturases have long been known to combine molecular recognition with reactive catalytic metal centers and thereby oxidize hydrocarbons with very high regio- and stereoselectivity (29). Multiple amino acid residues interact with the substrate via noncovalent interactions, such as hydrogen bonding and π-stacking, to position a specific reaction site in a precisely favorable orientation relative to a (typically iron-based) catalytic center. The ubiquitous hydroxylating cytochrome P450–based enzymes, for example, having an iron-porphyrin in the active site, constitute the most widely distributed class of regioselective monooxygenases (2, 3). Noncovalent recognition is important to ensure adequate lability for catalytic turnover. This strategy is hard to implement in synthetic systems that, unlike proteins, must rely on much smaller, simpler scaffolding structures to influence substrate orientation. In one case, Breslow and co-workers grafted porphyrin-based manganese catalysts to cyclodextrin groups, which anchored substrates in a favorable orientation through hydrophobic interactions (1013). Regioselective oxidation has also been achieved with other catalysts, but in this case substrates were aligned through covalent bonds, which prevented turnover (1416).

Here we report a nonporphyrin di-μ-oxo dimanganese compound that catalyzes the highly regioselective oxygenation of saturated C-H bonds in ibuprofen and (4-methylcyclohexyl) acetic acid (cis + trans) with >100 turnovers. The selectivity arises from noncovalent molecular recognition via reversible H-bonding between a carboxylic acid group of the catalyst and a carboxylic acid group of the substrate.

The complex [H2O(L)Mn(μ-O)2Mn(L)OH2](NO3)3 (1a, L is 2,2′:6′,2″-terpyridine) has earlier been reported as a very active catalyst for oxidation chemistry with Oxone (peroxomonosulfate) as primary oxidant (17, 18). Good evidence was obtained in this system to exclude the intermediacy of freely diffusing radicals, which would otherwise degrade the selectivity (19). Our present study uses the same di-μ-oxo dimanganese core, but with a different ligand, 2 (Fig. 1). Our ligand design addressed several key criteria: the need to stabilize high-valent manganese, an oxidation-resistant framework, and a –COOH group properly directed for molecular recognition. In ligand 2, the terpy group was chosen to stabilize high-valent manganese (20, 21). The phenylene linker provides a spacer between the docking element and the remote site of functionalization. The Kemp's triacid fragment provides a U-turn motif, a –COOH group suitably oriented for the molecular recognition function. The structure of the ligand was confirmed by x-ray crystallography (Fig. 1) (22). A suitable catalyst precursor, [(2)MnCl2], was prepared by refluxing a solution of ligand 2 in acetonitrile and a saturated aqueous solution of excess MnCl2. Subsequent oxidation of [(2)MnCl2] with 0.80 eq of Oxone in CH3CN-H2O mixture (1:1) gave [H2O(2)Mn(μ-O)2Mn(2)OH2](NO3)4 (1b) (22, 23).

Fig. 1.

ORTEP diagram of ligand 2.

Molecular modeling allowed us to predict which C-H bond in the substrate would be expected to come closest to the active site. The geometry of the proposed H-bonded catalyst-substrate complex was obtained by importing the crystal structure parameters of the ligand 2 (Fig. 1) and the di-μ-oxo dimanganese core into the model (17, 18, 23), docking ibuprofen, and then minimizing the energy (Fig. 2) (MM2, CAChe 5). We chose ibuprofen [2-(4-isobutyl-phenyl)-propionic acid] (3) as an appropriately rigid substrate with two sites of attack (indicated by bold arrows in Fig. 3). Oxidation occurs at the remote benzylic carbon to give 4 and at the vicinal position to give 5 (Fig. 3). If the catalysis operates via the catalyst-substrate complex predicted by the model (Fig. 2), then 4 should be the major product, with any initial –CH(OH) intermediate being rapidly oxidized further to the ketone (Fig. 3). With another rigid substrate, (4-methylcyclohexyl) acetic acid (6c: cis isomer; 6t: trans isomer; 6: 6c + 6t) docked via hydrogen bonding to 1b, the model (fig. S17) suggests that the remote tertiary C-H bond at C6 (indicated by the solid arrows in Fig. 3) should undergo oxidation preferentially to give the corresponding alcohol 7 (22).

Fig. 2.

Molecular model of catalyst 1b (Chem 3D) docked with H-bonded ibuprofen (substrate).

Fig. 3.

(A) Oxidation products of ibuprofen with 1b as catalyst. Bold arrows indicate alternate sites of attack. (B) Oxidation products of 6 (6c + 6t) with 1b as catalyst. Solid arrows indicate possible sites of attack according to molecular modeling.

Catalytic runs were carried out with 1 eq of substrate, 0.001 eq of 1b, and 5 eq of Tetrabutylammonium Oxone in acetonitrile. At room temperature, the reaction was quenched after 2 hours by addition of excess NaHSO3; products and unreacted substrate were extracted into ether. To confirm the role of docking, we performed control experiments using [H2O(L′)Mn(μ-O)2 Mn(L′)OH2](ClO4)3 (1c, L′ is 4′-phenyl-2,2′:6′, 2″-terpyridine) (23), a catalyst lacking the key –COOH group; under identical conditions, selectivity was lost. We also confirmed that no Oxone-induced reactivity occurs in the absence of the manganese catalyst (22).

When ibuprofen was treated with catalyst 1c, which lacks the key –COOH group, the ratio of remote to proximal oxidation products (4:5) was roughly 3:1. With catalyst 1b, however, the recognition functionality raised the selectivity for 4 more than 10-fold (Table 1). In time-dependent studies, the ratios of the products remained constant, confirming that the regioselectivity is not due to degradation of the alternative product 5 (22, 24, 25).

Table 1.

Product distribution from ibuprofen oxidation [by 1H–nuclear magnetic resonance (1H-NMR) spectroscopy].

TemperatureCatalystConversion4 (Favored by recognition)5 (Disfavored by recognition)Total turnoversView inline
20°C 1b 50% 97.5% 2.5% 50
1c 53% 77% 23% 53
0°C 1b 53% 98.5% 1.5% 53
1c 54% 78% 22% 54
-20°C 1b 53% 98.5% 1.5% 53
1c 54% 77% 23% 54
20°C 1b View inline 56% 75% 25% 56
1c View inline 58% 77% 23% 58
0°C 1b View inline 58% 98.5% 1.5% 580
20°C 1b View inline ,View inline 71% 96.5% 3.5% 710
  • View inline* Total turnovers = mol products per mol catalyst; subtrate:catalyst:oxidant = 100:1:500.

  • View inline Solutions contained excess acetic acid (400% with respect to substrate).

  • View inline With CD3CN as solvent instead of CH3CN.

  • View inline§ Subtrate:catalyst:oxidant = 100:0.1:500.

  • We ascribe this high regioselectivity to the modeled docking H-bonding between the –COOH groups of ibuprofen and catalyst 1b. To test this hypothesis further, we added acetic acid to the reaction medium, expecting that it would largely displace the ibuprofen from the recognition site. If so, the catalytic selectivity would then resemble that observed in the absence of recognition. Indeed, the 4:5 ratio found in this case fell precisely to the level observed with the control catalyst 1c (Table 1).

    Oxidation of the alkyl carboxylic acid substrate 6, using recognition catalyst 1b, led not only to regioselective oxygenation at the remote tertiary C-H bond, but also to diastereoselection of a single isomer of 7: the trans isomer 7t (Table 2). Our control experiment with catalyst 1c yielded several other oxidation products in addition to 7t (22). This diastereoselectivity can be rationalized on the basis of a model (Fig. 4) in which the stereochemical outcome is determined in the “rebound” step. In this model, the carbon radical putatively formed by H-atom abstraction from the substrate adopts a chairlike conformation (26), which leads to the observed product isomer 7t on transfer of an OH group from the Mn-OH intermediate to the carbon radical center. Transfer of OH to the opposite side of the ring, to yield the other product isomer, is not possible via any plausible intermediate conformations that we could identify from modeling work.

    Fig. 4.

    Molecular model of the intermediate, resulting from H-atom abstraction from C6 of 6 in a distorted chair conformation, docked to 1b (Chem 3D).

    Table 2.

    Product distribution from 6 (6c + 6t) (by 1H-NMR spectroscopy).

    TemperatureCatalyst (0.1%)Conversions7t (favored)7c (disfavored)Other productsTotal turnoversView inline
    20°C 1b 13% >99% <1% <1% 130
    20°C 1c ∼19% ∼30% ∼30% ∼40% 190
    20°C 1b View inline 18% >99% <1% <1% 180
  • View inline* Total turnovers = mol products per mol catalyst; subtrate:catalyst:oxidant = 100:0.1:500.

  • View inline With CD3CN as solvent instead of CH3CN.

  • In a control experiment that rules out undesirable autoxidative mechanisms involving atmospheric molecular oxygen, the same results were obtained under a nitrogen atmosphere as were obtained in air. Any rebound must, therefore, be fast because the proposed carbon radical is not trapped by ambient molecular oxygen. Oxidation of cis-stilbene yielded cis-stilbene epoxide, not the trans epoxide, which is usually the major product from a freely diffusing radical mechanism. This result suggests that a freely diffusing radical mechanism is unlikely (19, 27). Further mechanistic studies are in progress.

    Initially, we obtained low catalytic turnovers (∼50) for ibuprofen oxidation using 1% catalyst (relative to substrate), probably because catalyst degradation halted the reaction. To remedy this problem, we lowered the catalyst concentration, thereby reducing the probability of bimolecular catalyst self-oxidation. With a 0.1% catalyst to substrate ratio, a much higher total turnover number of 580 was attained (catalyst 1b, 0°C) with no loss of regioselectivity (98.5%) (Table 1). For oxidation of 6, we used 0.1% catalyst in all experiments. Further improvement of the catalytic turnover has been attained by replacing the CH3CN solvent with more oxidation-resistant CD3CN, thereby reducing the competitive solvent oxidation. Use of CD3CN raised total turnover numbers even further without adversely affecting the regioselectivity (and diastereoselectivity in the case of 6) (Tables 1 and 2).

    The general strategy may have wide application in chemical catalysis.

    Supporting Online Material

    Materials and Methods

    SOM Text

    Figs. S1 to S17


    References and Notes

    View Abstract

    Stay Connected to Science

    Navigate This Article