Diels-Alder in Aqueous Molecular Hosts: Unusual Regioselectivity and Efficient Catalysis

See allHide authors and affiliations

Science  14 Apr 2006:
Vol. 312, Issue 5771, pp. 251-254
DOI: 10.1126/science.1124985

This article has a correction. Please see:


Self-assembled, hollow molecular structures are appealing as synthetic hosts for mediating chemical reactions. However, product binding has inhibited catalytic turnover in such systems, and selectivity has rarely approached the levels observed in more structurally elaborate natural enzymes. We found that an aqueous organopalladium cage induces highly unusual regioselectivity in the Diels-Alder coupling of anthracene and phthalimide guests, promoting reaction at a terminal rather than central anthracene ring. Moreover, a similar bowl-shaped host attains efficient catalytic turnover in coupling the same substrates (although with the conventional regiochemistry), most likely because the product geometry inhibits the aromatic stacking interactions that attract the planar reagents to the host.

Effective synthetic homogeneous catalysts have generally been structurally simple small molecules, which act by binding to substrates at or near the reaction site. In contrast, enzymes are much larger and more complex and derive much of their selectivity by bonding substrates through multiple interactions in elaborate pockets, thereby forcing the substrates into orientations that favor specific reaction paths (1, 2). In the past decade, chemists have made substantial progress in building molecular hosts that emulate these enzymatic pockets (3, 4). Self-assembly of carefully constructed organic and/or metallic building blocks in solution produces hollow host structures that can bind small molecule guests (5, 6). Among the many potential advantages of this strategy is the creation of hydrophobic reaction environments in aqueous solution, widening the scope of accessible reactivity in ecologically friendly media. However, these synthetic hosts have rarely conferred the orientational precision necessary to guide reactions along otherwise unfavored pathways. Moreover, catalytic turnover has been inhibited because the hosts bind products as effectively as reactants, if not more so. In earlier reports by Rebek (7, 8), Sanders (9), and our group (10), the Diels-Alder and related cycloadditions are significantly accelerated in synthetic pockets, but the product inhibition prevents the reactions from showing turnover and the stereochemical courses are not well controlled by the pockets. For catalytic reactions by self-assembled hosts, there have appeared only a few examples, including the Diels-Alder (11), epoxidation (12), and the aza-Cope rearrangement (13). Controlling reaction pathways by encapsulation has been discussed in the excited-state chemistry of aromatic guests (14) and also realized by a regioselective cycloaddition (8).

We investigated the host-mediated Diels-Alder coupling of anthracenes and phthalimides. The Diels-Alder reaction of anthracenes in the absence of hosts is extremely well studied and generally yields an adduct bridging the center ring (9,10-position) of the anthracene framework (1517) as a consequence of the high localization of π-electron density at that site (18, 19). We find that an appropriately designed cage structure can alter this well-established selectivity to favor adduct formation at a terminal ring (1,4-position). This unusual regioselectivity likely stems from topochemical control induced by the proximity of the 1,4-position of the anthracene to the dienophile in the cage. The 1,4-selective Diels-Alder of anthracenes has been previously reported only for a benzyne addition (20) and for the addition with 9,10-diarylanthracenes (21). We further find that the same reaction, through conventional regioselectivity, can be catalyzed with efficient turnover by a related, bowl-shaped host. As in enzymatic reactions (2, 22), the product geometry, bent at the 9,10-position, precludes the aromatic stacking interactions that underlie the host's affinity for the reagents.

The coordination hosts we used here are octahedral cage 1 and square-pyramidal bowl 2 (Fig. 1) (2325). Both of them assemble from cis end-capped Pd(II) ions and triazine-cored tridentate ligands in a surprisingly efficient manner (100°C, <5 min, quantitative yields). In aqueous solution, these structures provide an efficient hydrophobic pocket capable of binding a variety of neutral organic compounds. Cage 1 features a three-dimensionally enclosed cavity, which binds substrates in precisely fixed positions. Geometry-fixed encapsulation (26), isomer separation (27), pairwise selective inclusion of two guests (28), and stereocontrolled cyclic siloxane formation (29) have been reported. Bowl 2 has an open cavity that facilitates rapid binding and dissociation of substrates (25).

Fig. 1.

Self-assembled coordination cages (1 and 2), which are prepared by simple mixing of an exo-tridentate organic ligand and an end-capped Pd(II) ion in a 4:6 ratio in water.

When 9-hydroxymethylanthrancene (3a) and N-cyclohexylphthalimide (4a) (6.0 mM) were suspended in an aqueous solution of cage 1 (5.0 mM) at room temperature, the inclusion complex Embedded Image (this set notation denotes that 1 includes 3a and 4a) formed selectively within 5 min (Fig. 2A). A 1H nuclear magnetic resonance (NMR) analysis confirmed the encapsulation, with the resonances of 3a and 4a shifted far upfield because of interaction with the cage (fig. S1a) (30). No signals indicating Embedded Image or Embedded Image (n ≤ 2) were observed in the NMR spectrum. On heating the solution at 80°C for 5 hours, the signals derived from 3a and 4a disappeared and were replaced by resonances consistent with a Diels-Alder adduct, distributed between 6.8 and –2.1 parts per million (ppm) (fig. S1b). Sixteen signals in the 9.7- to 8.4-ppm range were observed for cage 1, indicating the desymmetrization of the cage from Td to C3 symmetry (26). This symmetry agreed with the restricted motion of a noncentrosymmetric product along the C3 axis, which is perpendicular to one of the triazine ligands (30). After insoluble solids were removed by filtration, the product was extracted into CDCl3 and fully assigned as the syn isomer of 1,4-Diels-Alder adduct 5 (fig. S4). No other regio- or stereoisomers (1,9-adduct or anti-1,4-adduct) were detected. The yield of 5 was estimated to be >98% (based on 1) from the 1H NMR spectra (30). In contrast, in the absence of 1, the reaction gave only the conventional 9,10-Diels-Alder adduct in 44% yield based on 3a.

Fig. 2.

(A) Pair-selective encapsulation of two types of reactants, 9-hydroxymethylanthrancene (3a) andN-cyclohexylphthalimide (4a), within cage 1 and the subsequent Diels-Alder reaction leading to syn isomer of 1,4-adduct 5 within the cavity of 1. (B) Syn-1,4-regioselective Diels-Alder products within cage 1. The structures of theadductsandtheyields are shown.

The unusual structure of the 1,4-Diels-Alder adduct was unambiguously determined by x-ray crystallographic analysis of Embedded Image (Fig. 3A). A single crystal suitable for x-ray analysis was obtained by the slow evaporation of water from an aqueous solution of Embedded Image over 5 days (30). The crystal structure displays the syn stereochemistry of 1,4-adduct 5, which is tightly accommodated in the cavity of 1 via π-π stacking interactions (3.3 Å) between the naphthalene ring of 5 and a triazine ligand of 1.

Fig. 3.

(A) Crystal structure of Embedded Image 5 and (B) optimized structure of Embedded Image by a force-field calculation.

Because the Diels-Alder reaction has an early transition state (15), the unusual regio- and stereoselectivities can be explained by the fixed orientation of the guests before the reaction. The geometries of 3a and 4a in the Embedded Image complex were modeled by force-field calculation (31). Randomly oriented 3a and 4a guests in several initial structures converged in all cases to a parallel orientation with the Embedded Image bond of 4a in close contact with the 1,4-position of 3a (Fig. 3B). The center-to-center distance between the two reaction centers is only circa (ca.) 3.8 Å, which is comparable to the sum of van der Waals radii. Because of the steric restrictions induced by the cage, the Embedded Image bond of 4a hardly interacts with the 9,10-position of 3a (ca. 4.7 Å). It is also interesting that the cavity of 1 directs exo-selective addition of 4a to the 1,3-diene moiety of 3a, yielding only exoselective syn adduct 5.

The 1,4-regioselective Diels-Alder reaction also proceeded with varied substrates. Carboxyl-, cyano-, and vinyl-substituted anthracenes coupled with phthalimide 4a to give the corresponding 1,4-adducts in 92, 88, and 80% yields, respectively (Fig. 2B) (30). Unsubstituted anthracene also afforded only the 1,4-adduct in 55% yield. The moderate yield for this substrate is due not to reduced regio- or stereoselectivity but to the less efficient inclusion process before the reaction. The steric bulkiness of the N-substituent on the dienophile is crucial to the 1,4-selectivity. When sterically less demanding N-propylphthalimide (4b) was used, only the 9,10-adduct was formed.

We turned next to investigating Diels-Alder mediation by bowl-shaped host 2 and, strikingly, observed efficient catalytic turnover. Only 10 mole percent (mol %) of 2 sufficed to catalyze the Diels-Alder reaction of 3a and N-phenylphthalimide (4c) (Fig. 4). When 3a (10.0 μmol) and N-phenylphthalimide (4c, 10.0 μmol) were suspended in an aqueous solution of 2 (1.0 μmol in 1.0 ml) at room temperature for 5 hours (Fig. 5, A and B), the Diels-Alder adduct formed quantitatively (>99% based on 3a), as evaluated by NMR analysis of the product (32). NMR analysis also indicated that the reaction took place at the normal 9,10-position of anthracene to give 6 (Fig. 5C). In the absence of bowl 2, the reaction hardly proceeded (only 3% yield) under the same conditions. Surprisingly, even in the presence of 1 mol % of 2, adduct 3a was obtained in >99% yield after 1 day as estimated by the NMR spectrum in CDCl3. Moreover, the metal component, (en)Pd(NO3)2 (where en is ethylenediamine) alone, did not catalyze the reaction (30). Therefore, the data support promotion of the reaction by the hydrophobic pocket of 2.

Fig. 4.

Catalytic Diels-Alder reaction of 9-hydroxymethylanthracene (3a) and N-phenylphthalimide (4c) in the aqueous solution of bowl 2, leading to 9,10-adduct 6.

Fig. 5.

The 1H NMR spectra (500 MHz, room temperature) of the catalytic Diels-Alder reaction of 9-hydroxymethylanthrancene (3a) and N-phenylphthalimide (4c) in an aqueous solution of bowl 2. (A) Before and (B) after the reaction at room temperature for 5 hours (red circles, 3a; blue circles, 4c; and green squares, 6). (C) Diels-Alder product 6 after extraction with CDCl3. (D) Schematic representation of the catalytic Diels-Alder reaction of anthracenes and phthalimide in the presence of bowl 2. Autoinclusion of substrates into 2 (step a) and autoexclusion of the product from 2 (step c) underlie the efficient catalytic Diels-Alder reaction.

Bowl 2 also efficiently catalyzed Diels-Alder coupling of a variety of anthracene and phthalimide derivatives (30). When 3a and N-propyl- or N-benzylphthalimide were suspended in an aqueous solution of 2, the corresponding Diels-Alder products were obtained in almost quantitative yields after 5 hours at room temperature. In addition, 9-methyl and 9-viynylanthracene reacted with 4c in the presence of a catalytic amount of 2 (10 mol %).

Product inhibition has been a serious problem in previous examples of cavity-promoted Diels-Alder reactions with synthetic hosts (711). Because of the entropic disadvantage arising from the need to bind two reactant molecules, the encapsulated product has generally been a thermodynamic sink. Therefore, the reactions require near-stoichiometric quantities of host. It is noteworthy that, in contrast to previous examples, the present Diels-Alder reaction involves an exclusion step in the catalytic cycle. Before the reaction, anthracene can stack onto the triazine ligand of 2, gaining considerable stabilization via aromatic-aromatic or charge-transfer interactions (Fig. 5D, step a). The reactant-like transition state is similarly stabilized. However, once the reaction is complete, the product framework is bent at the 9,10-position, cutting off the host-guest aromatic stacking interaction (Fig. 5D, step b). Accordingly, the encapsulated product is considerably destabilized and smoothly replaced by incoming reagents (Fig. 5D, step c → a). In this sense, the affinity of the host for reactive substrates and the disaffinity for product is markedly similar to enzymatic behavior.

Supporting Online Material

Materials and Methods

Figs. S1 to S18

Tables S1 and S2

References and Notes

View Abstract

Navigate This Article