Docking in Metal-Organic Frameworks

See allHide authors and affiliations

Science  14 Aug 2009:
Vol. 325, Issue 5942, pp. 855-859
DOI: 10.1126/science.1175441


The use of metal-organic frameworks (MOFs) so far has largely relied on nonspecific binding interactions to host small molecular guests. We used long organic struts (~2 nanometers) incorporating 34- and 36-membered macrocyclic polyethers as recognition modules in the construction of several crystalline primitive cubic frameworks that engage in specific binding in a way not observed in passive, open reticulated geometries. MOF-1001 is capable of docking paraquat dication (PQT2+) guests within the macrocycles in a stereoelectronically controlled fashion. This act of specific complexation yields quantitatively the corresponding MOF-1001 pseudorotaxanes, as confirmed by x-ray diffraction and by solid- and solution-state nuclear magnetic resonance spectroscopic studies performed on MOF-1001, its pseudorotaxanes, and their molecular strut precursors. A control experiment involving the attempted inclusion of PQT2+ inside a framework (MOF-177) devoid of polyether struts showed negligible uptake of PQT2+, indicating the importance of the macrocyclic polyether in PQT2+ docking.

The concept of architectural domains that operate independently, yet are interconnected, is common in biology but difficult to achieve in synthetic materials. We believe that this concept offers a useful strategy for achieving materials with higher complexity. We have focused our attention on the design and synthesis of porous crystals composed of several architectural domains, one of which is capable of docking molecules in a manner akin to the well-known docking of drug molecules.

Our design takes advantage of the emerging chemistry of metal-organic frameworks (MOFs) (13), which has been used effectively to assemble components with simple constitutions—specifically, organic struts and inorganic joints—into three-dimensionally ordered structures. The vast majority of porous MOFs prepared thus far can be regarded (Fig. 1) as having two important architectural domains: (i) the pore aperture, which is responsible for the shape- and size-selective binding of incoming molecules, and (ii) the internal surface of the pores, onto which gases or small molecules can be compacted and distributed with simple interaction sites covering the struts and joints; in some cases, the interaction is with open metal sites. These two domains are called the sorting domain (2) and the coverage domain (48), respectively.

Fig. 1

Classification of the different porous domains in metal-organic frameworks. In the sorting domain, guest molecules are selected according to their size at the orifices of the pores. The entry of H2 (orange) and concomitant exclusion of CO2 (black) and CH4 (purple) reflects the sieve-like action at the entrances to the pores. In the coverage domain, the guest molecules along the walls of the pores are disordered on account of their weak nonspecific interactions with the framework surrounding the pores. By contrast, the active domain has built-in recognition sites that help to maneuver and dock incoming guests in a highly selective and stereoelectronically controlled manner. These recognition sites (red) could be π-electron–rich and, as such, would seek out π-electron–deficient substrates (blue).

The synthesis of more complex MOFs, where more than two domains are present, has remained unexplored. Here, we show how molecular recognition components, much used in supramolecular chemistry (9, 10), can be integrated in a modular fashion into struts of MOFs, thereby creating recognition sites into which incoming guests will dock in a highly specific manner with stereoelectronic control. This third architectural domain—the active domain—combines shape, size, and electronic elements in the recognition of incoming guests, and brings order to otherwise highly disordered guests in conventional MOFs. Hence, this chemistry describes a class of MOFs with a level of complexity higher than that of known open reticulated geometries (1).

We used the primitive cubic topology of the archetypical MOF-5 (11), in which benzene struts are joined by Zn4O(CO2)6 cluster joints, as the target for our design. Initially, we demonstrated the feasibility of using the long 1/4DMBDA (1) to make MOF-1000 (12), which has the MOF-5 topology, albeit quadruply interpenetrated (Fig. 2A). This approach was extended to the more complex struts BPP34C10DA (2) and 1/5DNPPP36C10DA (3), which are known to act as electron-rich receptors for electron-deficient substrates (13), to make the corresponding MOF-1001A, MOF-1001, and MOF-1002 (Fig. 2, B to D). Each of the crown ether receptors in MOF-1001 is accessible, as evidenced by the docking of the paraquat dication (PQT2+) at every one of the receptor sites (see below). In contrast to known MOFs, where the frameworks are used mainly as passive platforms for the adsorption of gases and molecules, MOF-1001 not only has active components in precise recognition sites but also, by virtue of the openness of its structure, allows substrates to diffuse freely from solution, through the pores, and finally dock in these active domains.

Fig. 2

Ball-and-stick drawings of single-crystal structures of MOF-1000, MOF-1001A, MOF-1001, MOF-1002, and their corresponding organic struts. Strut 1 was used to obtain MOF-1000 (A), which has a four-fold interpenetrating structure with different frameworks shown in four different colors. The crystal structure of MOF-1001A from strut 2 (B) is a triply interpenetrating cubic structure (shown in blue, gold, and gray), with polyethers represented by red balls and wires. (C and D) MOF-1001 from strut 2 (C) and MOF-1002 from strut 3 (D) share an identical cubic framework backbone, and crown ethers are placed precisely throughout the whole framework [Zn4O(CO2)6 polyhedra, blue; organic struts, gray; crown ethers, red]. Crown ethers in all the structures were modeled by Cerius2. All hydrogen atoms have been omitted for clarity.

Crystals of MOF-1000 (14) (Fig. 2A and Fig. 3A) were obtained by mixing a solution of strut 1 (15) with Zn(NO3)2·4H2O in N,N-diethylformamide under conditions previously used in the synthesis of MOF-5 (5, 11). Its crystal structure displays the same structural topology as does MOF-5. It is found to be four-fold interpenetrated because of the length and slender nature of the strut; the distance between the two carboxylate carbon atoms is 19.3 Å. The successful crystallization of MOF-1000 confirmed the practicality of creating MOFs with higher complexity by means of this synthetic protocol.

Fig. 3

Space-filling illustration of MOF-1000, MOF-1001A, MOF-1001, and MOF-1002. (A and B) The length of the struts (19.3 Å) allows the structures of MOF-1000 (A) and MOF-1001A (B) to interpenetrate. (C and D) In contrast, high volumes of open space were present in the noninterpenetrating MOF-1001 (C) and MOF-1002 (D) synthesized from struts with the same length. This feature ensures the full accessibility of the electron-donating receptors for the incoming substrates within the pores. The same color codes as in Fig. 2 were applied. All hydrogen atoms have been omitted for clarity.

Struts 2 and 3 respectively contain 34- and 36-membered polyether rings, which have been extensively used (13) as receptors for a wide range of electron-deficient substrates. These struts are ideally suited as molecular recognition modules for making MOFs. Strut 2 was prepared by means of a convergent synthetic approach (14) and was used under conditions similar to those used in the synthesis of MOF-1000 to yield MOF-1001A and MOF-1001 (14). The crystal structure of MOF-1001A is a triply interpenetrating framework (Fig. 2B and Fig. 3B), whereas that of MOF-1001 is the corresponding noninterpenetrating form (Fig. 2C and Fig. 3C); both have the MOF-5–type topology. The existence of MOF-1001A, despite its occasional appearance as a minor product, validates indirectly the high porosity of MOF-1001. The sheer openness of the structure, however, led us to further optimize the reaction conditions to successfully obtain MOF-1001 as a pure phase (14). MOF-1001 has Fm3¯m symmetry, with an exceptionally large unit cell parameter a = 52.93 Å.

We then extended the methodology to the synthesis of MOF-1002 (Fig. 2D and Fig. 3D) by using the 1,5-dioxynaphthalene–containing strut 3, which was produced by a divergent synthetic route (14). Single-crystal x-ray diffraction studies (14) indicate that MOF-1002 shares an identical cubic backbone with MOF-1001, affirming the generality of such a methodology for building a variety of crystalline structures with long struts capable of molecular recognition.

Calculations of the volumes of open space within the MOF structures confirmed the highly open nature of these crystals (86.9% space unoccupied by MOF-1001 framework atoms, as assessed by a model using the program Cerius2, version 4.2). The inherent flexibility of the macrocyclic polyether substructure was evident from the single-crystal x-ray analysis of MOF-1001. The bismethylenedioxy units of the tetraethylene glycol loops in the substructure are found to be highly disordered. Nonetheless, the positions of all the atoms in the inorganic joints and the rigid backbone of the links are unambiguous, as judged by comparison of the resulting bond distances and angles with the model structure (16). On the basis of the overall geometry and stoichiometry of the MOF framework, we can conclude that the crown ether receptors—capable of the complexation behavior required (17) for molecular recognition—are integrated precisely and periodically inside a robust framework. Thus, the extended framework provides the basis for their strategic placement so that they are exposed to the maximum accessibility to guests in three-dimensional space.

To date, a number of reports (1820) have appeared on the synthesis and structure of hybrid organic-inorganic compounds with macrocycles and mechanically interlocking components. The MOFs presented here combine the precise positioning of the active domains with docking as an expression of molecular recognition. This property was revealed by examining the molecular recognition behavior of the macrocyclic polyethers 2 and 3 as docking sites. When MOF-1001 crystals were introduced into a saturated solution of PQT·2PF6 in acetone, the crystals immediately turned red, and the color intensified over 60 min (21) (Fig. 4, A to E, and movie S1)—a typical behavior for this binding event that indicates charge-transfer interactions (22) between PQT2+ and crown ether rings. This observation points to the formation of MOF-1001 pseudorotaxanes (23) by threading of PQT2+ through the middle of the crown ether. The reversibility of such a process was evidenced by the reappearance of the original light yellow color upon rinsing with acetone, where 60% of PQT2+ could be removed after rinsing MOF-1001 pseudorotaxanes (2.8 mg) four times with 1 ml every 30 min (14). The complexed MOF-1001 maintained the original high crystallinity of the parent framework, as confirmed by coincident powder x-ray diffraction patterns.

Fig. 4

X-ray diffraction and solid-state NMR spectroscopic studies on MOF-1001, MOF-1001 pseudorotaxanes, and their molecular analogs. (A to C) MOF-1001 [(A) and (B)] maintained its crystallinity after docking of PQT2+, a single crystal–to–single crystal transformation revealed by the x-ray diffraction pattern (C). Dimensions of the cubic crystals varied from 0.05 to 0.45 mm. (D to F) This quantitative threading to form MOF-1001 pseudorotaxanes [(D) and (E)] was confirmed by the 1:1 stoichiometry of PQT2+ and strut 2 (F). The crystal structure of MOF-1001 and the simulated MOF-1001 pseudorotaxanes structure are illustrated in ball-and-stick models. (G) This docking phenomenon resulted in the upfield shifts of the 15N CP/MAS signals (11) relative to free PQT2+. (H and I) The molecular pseudorotaxane analog [PQT⊂2]·2PF6 (H) was found to have the same upfield shift trend (I) (Δδ = 4.9 ppm). Color code: Zn4O(CO2)6 polyhedra, gold; organic struts, gray; crown ethers, red; PQT2+, blue. All hydrogen atoms and counterions have been omitted for clarity.

Further evidence of complexation was obtained by examining the 1H nuclear magnetic resonance (NMR) spectrum of the MOF-1001 pseudorotaxanes after dissolution in DCl (14). Integration of the peaks appearing at 7.96 ppm (d, 4H, Ar-Ha in 2, 3J = 8.5 Hz; Fig. 4F) and 4.60 ppm (s, 6H, N-CH3 in PQT2+) revealed the expected 1:1 ratio of strut 2 and PQT2+, indicating that the docking phenomenon of PQT2+ does indeed take place at every crown ether ring throughout the whole MOF framework (fig. S10). Solid-state 15N NMR spectroscopy—a technique that is highly sensitive to the environment of the nitrogen (15N) in PQT2+—provided further evidence for docking in MOF-1001. Isotope-labeled PQT2+ (14) with 25% abundance of 15N was used to make the MOF-1001 pseudorotaxanes, and the resulting solid was examined by 15N cross-polarization magic angle spinning (CP/MAS) spectroscopy (24). The spectrum of the uncomplexed PQT2+ has a 15N signal centered on 207.2 ppm, whereas the spectrum of PQT2+ bound within the crown ether rings in MOF-1001 shows an upfield shift to 204.6 ppm for the 15N resonance resulting from docking into the macrocyclic polyether units of the struts (Fig. 4G).

Similar studies carried out on strut 2 were used as a molecular analog for comparison with MOF-1001 complexation experiments. Here, addition of PQT·2PF6 to an acetone solution of strut 2 led to the formation of a pseudorotaxane, [PQT⊂2]·2PF6. The binding affinity (Ka = 829 M–1) (fig. S2) between PQT2+ and strut 2 in solution was obtained from spectrophotometric titrations. Single-crystal x-ray diffraction of the [PQT⊂2]·2PF6 (Fig. 4H) clearly shows the insertion of the π-electron–deficient bipyridinium dication through the middle of the macrocyclic polyether. π-π stacking and [C–H···O] interactions are reflected in the interplanar separation of 3.6 Å between the bipyridinium unit of PQT2+ and the hydroquinone rings. The same upfield shift trend in the 15N NMR spectra observed for MOF-1001 pseudorotaxanes was also evident in the 15N NMR spectra of [PQT⊂2]·2PF6 in the solid state (14) (Fig. 4I) as well as in solution (fig. S5). Control experiments were carried out by attempting to introduce PQT·2PF6 into porous MOF-177 crystals (25), the pore dimensions (d = 11.8 Å) of which were expected to allow the free movement of PQT2+ within the pores. We found that fewer than 0.06 PQT2+ molecules per strut of MOF-177 were incorporated in the pores (fig. S11). These results clearly show that specific stereoelectronic host-guest interactions, rather than simple diffusion and adsorption, are responsible for the all but quantitative formation of the MOF-1001 pseudorotaxanes.

Supporting Online Material

Materials and Methods

Figs. S1 to S24

Tables S1 to S11

Schemes S1 to S13


Movie S1

References and Notes

  1. We assign the MOF-1000 numbering scheme to MOFs that are constructed from struts capable of stereoelectronically controlled binding.
  2. See supporting material on Science Online.
  3. Strut 1 was synthesized (14) by the attachment of 4-(carboxyphenyl)ethynyl groups to the 2- and 5-positions of a central 1,4-dimethoxybenzene ring using the Pd-catalyzed alkyne-aromatic Sonogashira coupling (26).
  4. It has not escaped our attention that the phenylene rings that incorporate the struts in 2 and 3 also support planes of chirality. As a consequence, there is yet another fundamental source of disorder associated with the polyether loops in MOF-1001A, MOF-1001, and MOF-1002. Moreover, there is also the prospect of being able to construct chiral MOFs where the elements of chirality are planar (27) in origin.
  5. Crystals of MOF-1001 were first immersed in acetone to exchange the N,N-dimethylformamide (DMF) guests and unreacted BPP34C10DA. This process was repeated nine times by decanting and refreshing with acetone (5 ml) every 30 min to ensure full exchange of DMF.
  6. The experiments were done by introducing acetone-exchanged MOF-1001 into acetone solutions of PQT·2PF6 with different amounts of PQT2+. After sitting for 6 hours, solvent was then removed by evaporation and the residue was further dried under vacuum (10−2 torr) overnight at room temperature. A 15N CP/MAS NMR spectrum was acquired on the solid sample and the loading of PQT2+ was determined by solution-state 1H NMR spectroscopy after digestion of the solid. See (14).
  7. This work was supported by the U.S. Department of Defense (Defense Threat Reduction Agency grant HDTRA1-08-10023) and Northwestern University. We thank S. Kabehie for assistance with emission spectroscopy. Crystallographic data for [PQT⊂2]·2PF6, MOF-1000, MOF-1001A, MOF-1001, and MOF-1002 have been deposited into the Cambridge Crystallographic Data Centre under deposition numbers CCDC 728413 to 728420.
View Abstract

Stay Connected to Science

Navigate This Article