Dual-Selective Catalysis in Dephosphorylation Tuned by Hf6-Containing Metal-Organic Frameworks Mimicking Phosphatase.

Selective dephosphorylation is full of great challenges in the field of biomimetic catalysis. To mimic the active sites of protein phosphatase, Hf-OH-Hf motif-containing metal-organic frameworks (MOFs) were obtained and structurally characterized, which are assembled from [Hf48Ni6] cubic nanocages and exhibit good stability in various solvents and acid/base solutions. Catalytic investigations suggest as-synthesized Hf-Ni and Hf-Ni-NH 2 display accurate type-selectivity (selectively catalyzed P-O rather than S-O or C-O bonds) and position-selectivity (selectively catalyzed phosphomonoesters over phosphodiesters) for the hydrolysis of phosphoesters. Reaction kinetic studies further revealed the high activity of the catalytic sites in these catalysts, and the unique catalytic selectivity and high activity are comparable to phosphatase. Additionally, these MOF catalysts possess good recursivity and hypotoxicity. Control experiments (including Hf- and Zr-based isomorphous MOFs) and theoretical calculations indicate that both triplet nickel and Hf6 clusters play significant roles in the unique binding site and favorable binding energy. To our knowledge, this is the first example of selective dephosphorylation through MOF catalysts as mimic enzymes, which paves a potential way for the development of specific therapeutic MOFs.

Introduction

Phosphatase, a kind of important zinc-containing metalloprotein (Scheme S1a), not only serves in the hydrolysis of organophosphorus compounds such as pesticides and nerve agents[1,2] but also selectively regulates the phosphorylation of biomolecules in vivo,[3] in which abnormal hyperphosphorylation is closely related to a wide range of human diseases such as Alzheimer’s disease[4] and lung cancer.[5] However, some inherent drawbacks of enzymes, such as low stability, high cost, and difficulties in recycling, seriously impede advances in such fields. Therefore, it is important to develop a simulated phosphatase overcoming the deficiencies of natural phosphatase based on an understanding of its structure and mechanism.[6,7] To overcome the technique limitations and inherent complexity of natural enzymes,[8,9] biomimetic investigations on phosphatase as one of the effective strategies were extensively carried out, and up to now, some outstanding results associated with mimicking phosphatase have been reported. For example, Bera’s group synthesized three tetranuclear iron(III) and zinc(II) complexes with phosphatase-like activity using a model substrate of bis(p-nitrophenyl) phosphate (BNPP);[11] Hupp and Farha used UiO-66 to hydrolyze nerve agents (phosphate-based compounds).[10] However, the control and design of catalytic selectivity for the obtained biomimetic compounds remain unsolved.

High catalytic selectivity is indeed an important character of natural enzymes. In phosphatase, the Zn-OH-Zn motif serves as a catalytic active site (Scheme S1b), and the catalytic selectivity is considered to mainly originate from a magnesium ion in the third metal site.[12] Inspired by the structure–activity relationship of phosphatase, we propose to design Hf-OH-Hf motif-containing heterometallic metal–organic frameworks (MOFs) for mimicking phosphatase to explore the catalytic selectivity based on the following considerations: (1) MOFs are inherently suitable for biomimetic studies of metalloproteins since MOFs can provide fine-tunable structures for judiciously introducing catalytic sites with atomically composed organic linkers and metal ions/clusters.[13−17] (2) The Hf-OH-Hf motifs are similar to the Zn-OH-Zn active site in phosphatase, making it possible to serve as a biomimetic catalyst for the hydrolysis of phosphosubstrates.[10] (3) The Lewis acidic nature[18,19] and low toxicity of Hf(IV) centers help to promote the hydrolysis of phosphate esters and be applied in biology.[20] (4) Strong Hf–O bonds result in exceptional thermal and chemical stability of Hf-MOFs, which allows Hf-MOFs catalysts to be reused in reactions.[21−31] (5) Some transition metal complexes[32−37] were reported to significantly hydrolyze phosphate ester bonds in nerve agents, and the introduction of transition metal and Hf-OH-Hf motifs into MOFs maybe produces catalytic synergistic effects, which can provide a promising method to control the catalytic activity and/or enhance the catalytic selectivity.

With this idea in mind, here we designed and synthesized two heterometallic Hf6-containing networks {Hf6(μ3-O)4(μ3-OH)4Ni3(L)12Cl6} (L = isonicotinic acid (Hf–Ni); 3-aminoisonicotinic acid (Hf–Ni–NH)) by rationally regulating the organic linker. Both of them possess robust three-dimensional frameworks with ftw topology (Figure S1) and are assembled by [Hf48Ni6] cubic nanocages. Hf–Ni and Hf–Ni–NH feature superior stability in various solvents and acid/base solutions. With a biomimetic strategy, their catalytic activity and selectivity mimicking phosphatase in dephosphorylation were systematically investigated. The results reveal that, even when compared to their enzyme counterparts, these heterometallic MOFs possess excellent catalytic activity and accurate catalytic dual-selectivity for p-NPP among diverse substrates, which contain various types of chemical bonds (p-NPA, p-NPS) and different substitutional positions (BNPP). Control experiments and theoretical calculations indicate that both triplet nickel and Hf-MOF played significant roles in the unique binding site and favorable binding energies, which are in good agreement with the observed reactivity and selectivity. For the first time, this work realizes the biomimetic catalysis with catalytic dual-selectivity by a MOF platform, which will facilitate the understanding of natural biocatalytic systems and provide a new route for the design of specific biomimetic materials.

Results and Discussion

These MOFs were harvested from the reaction of isonicotinic/3-aminoisonicotinic acid, HfCl4, and Ni(NO3)2 in a DMF solvent system with acetic acid at 100 °C. Crystallographic analyses revealed they are isomorphous,[38] which also can be verified by powder X-ray diffraction (PXRD, Figure S2). Hf–Ni was selected as a representative example to describe the structure. Single-crystal X-ray diffraction studies indicate that compound Hf–Ni crystallizes in cubic space group Pm3m with unit cell a = b = c = 15.2592(6) Å. Each Hf(IV) ion is coordinated by four μ3-O2-/OH– and four carboxylic O atoms from the ligand. Six Hf(IV) ions are connected by eight O atoms, forming an octahedral [Hf6O4(OH)4(COO)12] cluster (Figure a), in which each triangular face is capped by one μ3-O or μ3-OH group. The Hf6 cluster with O symmetry is further coordinated by 12 carboxylates, which empirically can increase the robustness of MOFs.[39,40] The Ni2+ ion is coordinated with two Cl– and four ligand anions, forming a porphyrin-like ligand-unit (Figure b). Eight Hf6 clusters are knitted together by six ligand-unit anions to fabricate one cubic [Hf48Ni6] nanocage with an outer edge length as large as ∼2.2 nm (Figure c). Every Hf6 clusters act as one vertex of the cube, and each face is capped by one square ligand-unit. Interestingly, each Hf6 cluster bridges 12 ligand-unit anions, and each such ligand-unit anion is surrounded by four Hf6 clusters, resulting in the formation of the (4,12)-connected ftw-topology frameworks (Figure d). Upon removal of guest molecules, the corresponding solvent-accessible free volume for Hf–Ni is estimated to be 42.6% with PLATON software.[41] A reversible type-I isotherm for Hf–Ni was clearly exhibited from the N2 gas adsorption at 77 K (Figure S4), confirming the permanent porosity of compound Hf–Ni. The apparent Brunauer–Emmett–Teller surface area is estimated to be 1040 m2 g–1, and the maximum N2 sorption capability (P/P0 = 1.0) is 432.8 m3 g–1. The pore size distribution of Hf–Ni according to the desorption data further demonstrated that most of the resultant micropores are around 0.74 nm.

Figure 1

Structures of Hf–Ni. (a) The 12-connected [Hf6O4(OH)4(COO)12] cluster and the simplified model. (b) Ball–stick and simplified models of the porphyrin-like ligand-unit with isonicotinic acid ligands, Ni2+ cation, and Cl– anions. (c) Ball–stick model of the cubic cage built from eight [Hf6O4(OH)4(COO)12] building blocks and six ligand-unit anions. (d) Three-dimensional frameworks with ftw topology assembled by cubic nanocages. Violet cubic blocks indicate the cavity inside the cages, and hydrogens are omitted from the structure for clarity.

For compound n class="Chemical">Hf–class="Chemical">pan> class="Chemical">Ni, high connectivity of the Hf6 cluster was found to be one of the most stable building units for MOF construction, and the high charge density (Z/r) of Hf(IV) may endow its high solvent stability and thermostability.[42] To confirm the idea, as-synthesized crystals of Hf–Ni were immersed in various organic solvents and aqueous solutions of different pH values (from pH = 1 to 13) for 6 h at room temperature, respectively. Experimental PXRD results are very consistent with the simulated one, indicating excellent solvent stability of Hf–Ni (Figures S5 and S6). Thermogravimetric analysis (TGA) exhibits a weight loss of 25.5% from room temperature to 249 °C, in accordance with the release of guest molecules. Then, the framework begins to collapse at approximately 365 °C, indicative of good thermal stability of compound Hf–Ni (Figure S7). Variable temperature PXRD of Hf–Ni demonstrates that the crystal phase remains unchanged until 200 °C (Figure S8), further confirming the high thermal stability of Hf–Ni.

Catalytic activities of these MOFs were evaluated by monitoring the hydrolysis of p-NPP. The formation of nitrophenol (p-NP) from cleavage of the P–O bond in p-NPP was directly monitored using UV/vis spectroscopy with an absorbance at 400 nm (Figure a). The catalytic conditions were first optimized using four buffer systems (HEPES buffer, Tris-HCl buffer, carbonate buffer, and water). (4-(2-Hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) buffer (pH 7.4, 50 mM) exhibited the best performance for this reaction at room temperature (Figure S11) and was applied as a default buffer in subsequent experiments. As illustrated in Figure b, the reaction rate of Hf–Ni–NH and Hf–Ni was 263 and 194 μM h–1, respectively, showing that Hf–Ni–NH displays a higher catalytic activity than Hf–Ni. This can be attributed to the amino moieties acting as a base or a proton-transfer agent to enhance the reaction rate, which is consistent with reported results.[43] Importantly, the reaction rates catalyzed by heterometallic MOFs were both higher than that of UiO-66(Hf) (31 μM h–1). Transition Ni2+ ions may also have important roles during catalysis except for the amino moieties and hafnium-bridging hydroxy ligands (similar to the catalytic center of phosphatase). In order to verify this assumption, two isomorphic compounds Hf–Cu and Hf–Cu–NH were obtained and structurally characterized using copper atom displaced nickel atoms. PXRD measurements clearly indicated the isomorphic structure of Hf–Cu, Hf–Cu–NH, Hf–Ni, and Hf–Ni–NH (Figure S2). Interestingly, the reaction rates of Hf–Cu and Hf–Cu–NH were 55.3 and 70.6 μM h–1, which were both smaller than that of Hf–Ni or Hf–Ni–NH, suggesting that Ni-based nodes are more suitable than Cu for hydrolysis of phosphate ester. In addition, Zr-based isomorphic MOFs (Figure S3) were also synthesized and applied in this reaction, while all of them exhibited a lower reacted efficiency than their corresponding Hf-base MOFs (Figure b), revealing the Hf6 center as a catalytic site has a higher activity in the dephosphorylation reaction. It is worth noting that biomimetic catalysts of Hf-based MOFs applied in this reaction have never been reported up to now. Furthermore, the raw material of as-synthesized MOFs, including HIn-NH2, HIn, Ni(NO3)2, and Cu(NO3)2, had no effect on the hydrolysis of p-NPP; only HfCl4 exhibited slight activity, but it was far lower than that of Hf–Ni and Hf–Ni–NH (Figure S12). These results indicate that Hf–O–Hf motifs in frameworks are the main catalytic sites, and the 3D framework can facilitate enhancement of the catalytic activity because the porous framework can enrich the substrates and keep catalytic sites highly dispersed during the reaction. In addition, Ni nodes and amino groups also have a positive synergistic effect on this reaction based on the control experiments of Hf–Cu, Hf–Cu–NH, Hf–Ni, and Hf–Ni–NH.

Figure 2

Catalytic activity. (a) Increase of the 4-nitrophenolate band at 400 nm after addition of Hf–Ni to a p-NPP solution. Inset shows the standard curve for p-NPP. (b) Hydrolysis profiles of p-NPP in the presence of Hf–Ni, Hf–Ni–NH, Hf–Cu, Hf–Cu–NH, UiO-66(Hf), and Zr-based isomorphic MOFs. (c) Catalysis profiles of p-NPP. Hf–Ni was removed by centrifugation at 25 min and put back at 60 min. (d) Cycling runs of hydrolysis of p-NPP in the presence of Hf–Ni. The concentration of Hf–Ni and p-NPP was 0.2 mM and 1 mM, respectively.

Considering potential biomedical applications, the cytotoxicity of these Hf-M-MOFs was evaluated. An MTT assay was performed using a 3T3 cell line. The EC90 values for them were 0.4 mg/mL (Hf–Ni), 0.08 mg/mL (Hf–Cu), 0.25 mg/mL (Hf–Ni–NH), and 0.16 mg/mL (Hf–Cu–NH), respectively (Figure S13), indicating that all of these catalysts have low cytotoxicities. Especially, Hf–Ni exhibits relatively less toxic than Hf–Cu, Hf–Ni–NH, and Hf–Cu–NH, suggesting that copper metal and an organic linker with amino groups may increase the cytotoxicity, which is consistent with a previous report.[44] It is worth noting that Hf–Ni–NH has a higher catalytic activity, but the cytotoxicity is much higher than Hf–Ni. Hence, Hf–Ni is more suitable for medical applications, and Hf–Ni–NH can be applied in in vitro experiments. We also observed neither significant density nor morphology differences of cells upon incubation with Hf-M-MOFs for 4 h. Furthermore, results of RBC hemolysis assays also indicated that these MOFs induced no hemolysis in red blood cells (RBC) even at a concentration of 1 mg/mL after 1 h incubation at room temperature (Figure S14). These results together revealed good biocompatibility and system security of the designed Hf-M-MOFs.

The nature of the catan class="Chemical">lysis reactionclass="Chemical">pan> has also been explored. After removal of the MOF catalyst filtered using a 220 nm syringe filter, no reaction was observed after filtration at 30 min (Figure c). Subsequently, the reaction restarted after putting the MOF catalyst into the reaction system at 60 min. Furthermore, very little Hf or Ni elements in the filtered solution were detected by inductively coupled plasma (ICP) analysis (Table S5), which excluded the possibility of Hf and Ni ions leaching, confirming the heterogeneous catalytic nature of this reaction. In view of the significance of reutilization for a heterogeneous catalyst,[45−47]Hf–Ni catalysts were reused four times with same conditions and had slight decreas in catalytic performance (Figure d), which was probably due to a small loss of the solid catalysts during the reuse process. The solid UV/vis absorption curve and PXRD pattern after the reaction were in accordance with the original ones (Figures S15 and S16), indicating these catalysts still maintain integrity after catalytic cycles. Additionally, the morphology of Hf–Ni after the reaction based on scanning-electron microscope (SEM) measurement did not changed much except for the appearance of slight cracks caused by stirring (Figure S17).

Steady-state kinetics were empn class="Chemical">loyed to further inclass="Chemical">pan>vestigate the phosphatases-like catalytic activity and kinetic parameters of Hf–Ni. Within the range of 0.05–6 mM of p-NPP,a typical Michaelis–Menten curve was observed (Figure ), and a Lineweaver–Burk plot could be obtained with a nearly linear relationship (inset of Figure ), from which the kinetic parameters Km, Kcat, and Vmax can be achieved. Km shows the substrate affinity of the enzyme, and smaller Km value indicates a higher enzyme affinity to substrate. Kcat represents the maximum number of substrate molecules turned over per catalyst molecule per unit time under optimal conditions. It can be regarded as the optimum turnover rate, giving a direct measure of the catalytic activity. As shown in Table S6, the values of Vmax and Km for Hf–Ni are 1.67 × 10–3 mM min–1 and 0.23 mM, respectively. For the hydrolysis of p-NPP, the derived Kcat of Hf–Ni catalyst was 8.34 × 10–3 min–1, which is 3.79 × 103 times higher than that of cerium oxide nanoparticles (CeNPs) (2.20 × 10–6 min–1),[48] indicating the high catalytic activity of Hf–Ni. Moreover, the derived Km value of 0.23 mM for Hf–Ni is lower than that of natural protein tyrosine phosphatases (PTP) enzyme (0.38 mM),[49,50] which demonstrated a better affinity of the substrate to Hf–Ni.

Figure 3

Steady-state kinetic assays of Hf–Ni; inset: double-reciprocal plots of activity of Hf–Ni. Error bars show the standard error derived from three repeated measurements.

Steady-state kinetic assays of n class="Chemical">Hf–class="Chemical">pan> class="Chemical">Ni; inset: double-reciprocal plots of activity of Hf–Ni. Error bars show the standard error derived from three repeated measurements.

Kinetic studies revean class="Chemical">led that compounclass="Chemical">pan>d Hf–Ni exhibited excellent phosphatase-like catalytic activity, while UiO-66(Hf) showed inactivity with the same conditions. The derived Km value of Hf–Ni is lower than that of natural protein tyrosine phosphatases (PTP) enzyme, which demonstrates that Hf–Ni has a better affinity with the substrate. Considering the fast catalytic rate, the diffusion of these substrates limited the complete utilization of active centers, which shows that the actual Kcat values of catalytic centers in Hf–Ni could be even higher.[51] And the excellent catalytic performance of Hf–Ni can be attributed to the high density of active centers in the framework, which provides an active site per 3053 Da. In contrast, 37 000 Da is necessary for each active site in PTP.[52]

P–O, S–O, and C–O bonds are three types of important chemical bonds in vivo. Therefore, 4-nitrophenylphosphate (p-NPP), bis(4-nitrophenyl) phosphoester (BNPP), 4-nitrophenyl acetate (p-NPA), and 4-nitrophenyl sulfate (p-NPS) were selected as models to explore the catalytic selectivity of Hf–Ni (Figure a). The formation of reaction products for different substrates was also detected by UV–vis spectra. Among all the tested reactions, Hf–Ni has an outstanding rate of 194 μM h–1 for substrate p-NPP with a monosubstituted PO43– group (Figure b). While for substrate BNPP containing a P–O bond with different substitutional positions, the reaction rate is about 10 times lower than that of p-NPP, which could be attributed to the steric hindrance of BNPP.[53] The molecular size of BNPP is 8.60 Å × 4.28 Å × 8.60 Å, the height size of which is larger than p-NPP molecules (8.60 Å × 4.28 Å × 2.53 Å) but slightly smaller than the hexagonal channels of Hf–Ni connected by 8.37 Å × 7.23 Å × 4.17 Å windows (Figure S22). The dimensional difference makes p-NPP easier to diffuse through the channel of compound Hf–Ni, resulting in the higher catalytic efficiency of p-NPP than BNPP. Moreover, Hf–Ni indicates almost no catalytic activity for p-NPA with the type of C–O bond, which may originate from the different geometry of the substrate. And for p-NPS containing a S–O bond, which has analogously tetrahedral geometries with p-NPP,[54]Hf–Ni still has poor activity. This could be because the hydrolysis process of p-NPP was via a fairly compact transition state with solvent destabilization, while the hydrolysis process of p-NPS was via a far more expansive transition state with a much smaller solvent effect.[55] Furthermore, the other two model substrates (p-nitrophenylbutyrate and Tris(4-nitrophenyl)phosphate) have been employed with the catalyst Hf–Ni, and the results showed a less significant catalytic effect (Figure S19 and Table ). These results reveal that Hf–Ni possesses accurate type-selectivity and position-selectivity, which can not only selectively catalyze the hydrolysis of P–O bond but can also catalyze the hydrolysis of the P–O bond with a specific substitutional position. This phenomenon also suggested the potential application of Hf–Ni in the field of medicine because precisely regulating dephosphorylation of biomolecules is an important method to treat many diseases.

Figure 4

Catalytic selectivity. (a) Model substrates used in this research. (b) The catalysis profiles of p-NPP, BNPP, p-NPA, p-NPS by Hf–Ni and Hf–Ni–NH. (c) The catalytic performance of Hf–Cu and Hf–Cu–NH for different substrates. (d) The catalytic performance of HfCl4 and Ni(NO3)2 for different substrates.

Table 1

Comparison of the Hydrolysis Rate of Various Substrates with Hf–Ni Catalyst

Furthermore, several control experiments were performed to explore the factors for catalytic selectivity in this reaction. Ligands HIn/HIn-NH2 and Ni(NO3)2/Cu(NO3)2 were first tested and have almost no catalytic activity and selectivity (Figure S18), proving that individual ligands or transition metal salts are ineffective for this reaction. HfCl4 possesses a poor catalytic activity and selectivity (Figure c), indicating Hf(IV) nodes are the main catalytic sites and exert a positive effect on selectivity. Besides, isomorphic compounds Hf–Cu and Hf–Cu–NH also display good selectivity but lower than Hf–Ni and Hf–Ni–NH (Figure d), indicative of the important roles of coordinate nickel ions and a three-dimensional (3D) framework for improving catalytic activity and selectivity. Furthermore, nickel ions in the framework have positive effects on polarizing the phosphoryl group and altering the structure of the transition states, leading to enhancement of the catalytic activity, which was also verified by different catalytic performances of Zr–Ni, Zr–Cu, Zr–Ni–NH and Zr–Cu–NH. Moreover, compound Hf–Ni–NH has more obvious selectivity than Hf–Ni, proving the positive effect of the amino functional group, which can be explained as the synergistic effects between the amine group on the linker and the Hf6 node acting as a proximal base and Lewis acidic center, respectively. It was also reported that, through the formation of the H-bond between the amine and phosphate ester, proximal aromatic amines could facilitate the delivery of phosphate ester substrates to nearby catalytic active sites.[43] In addition, Hf–Ni is a 3D network with ftw-topology containing mesoporous hexagonal channels connected by 8.37 Å × 7.23 Å × 4.17 Å windows, which allow for faster diffusion of substrate/product. All results mentioned above mean that Hf(IV) nodes and coordinate nickel ions, as well as amino functional groups in the 3D porous framework, have a positive synergistic effect on this reaction. This is the first time MOFs have been used for selective biomimetic dephosphorylation.

Enzymes possess better sen class="Chemical">lectivity thanclass="Chemical">pan> conventional chemical catalysts presumably because the geometric constraints of enzymes can accelerate their reaction and promote the selectivity of the unique product.[56] Hereby, to understand critical factors for the reactivity and selectivity of the Hf–Ni catalyst, we carried out density functional theory (DFT) calculations at the SMD-M06/6-31++G(d,p)/Lanl2dz//6-31G(d,p)/Lanl2dz level of theory without any constraint. As shown in (Figure a-1), the computational model of complex Hf–Ni-p-NPP displays three key interactions between the substrate p-NPP and MOF catalyst: (1) the nitryl group of the substrate coordinates to metal Ni of the catalyst, (2) a π–π stacking of aromatic ring between the nitrophenol and isonicotinic acid ligand, (3) the phosphate group of the substrate interacts with Hf6 involving a P=O4···Hf1 interaction and an O···H hydrogen bond. Spin multiplicities of complex Hf–Ni-p-NPP were investigated by DFT calculations, and the calculated triplet complex is 26.0 kJ/mol more stable than that in the singlet state (Figure a-2). For the triplet complex, most of the spin density located on the Ni metal and pyridine ligands dispersed partial spin density (Figure b). Through the inspection of the frontier molecular orbital, the dz2 orbital of Ni has a favorable orbital overlap with the π orbital of the nitryl group (Figure c), and the distance of O1–Ni is 2.38 Å. The second-order perturbative estimate shows 16.6 kJ/mol donor–acceptor interaction from lone pair electrons of O1 to lone vacant orbital of Ni (Table S9). Because of the fully occupied dz2 orbital of Ni rests in the singlet state, it is difficult to form axial interactions with the nitryl group, which can be verified by the longer distance of O1–Ni (3.95 Å) (Figure S21). Furthermore, a favorable π–π stacking of the aromatic ring was predicted in the triplet complex between the nitrophenol and isonicotinic acid ligand with nearly parallel orientation, and the distance between two aromatic centers is 3.43 Å, while there is no obvious π–π stacking in the singlet complex. Accordingly, the stronger interaction of O1–Ni and π–π stacking jointly contribute to the superiority of the triplet state of the Hf–Ni catalyst. The case when Ni (triplet) was replaced with Cu (doublet) was also calculated for comparison. And it is more stable for p-NPP to bind to Hf–Ni than to Hf–Cu (ΔΔE = 14.4 kJ/mol), which indicates the significant binding contribution of metal Ni for the catalytic reactivity (Table S10).

Figure 5

Computational analysis. (a-1) The computational model of complex Hf–Ni-p-NPP depicted in two parts: the square planar ligand-unit and the Hf6 cluster unit. (a-2, a-3) There are two possible spin multiplicities of Ni (singlet and triplet) and one of Cu (doublet), and three possible modes of p-NPP binding to Hf6 cluster: a, one O of p-NPP interacts with one Hf without breaking any initial O–Hf bond; b, one O of p-NPP interacts with one Hf after breaking one initial O–Hf bond; b + c, p-NPP bonds with two Hf of a Hf6 cluster after the dissociation of one HCOO–. The relative energy (ΔE) and relative binding energy (ΔΔE) are given. (b) Spin density of Hf–Ni-p-NPP (triplet). (c) Frontier molecular orbital of Hf–Ni-p-NPP (triplet). (d-1) Sketch of complexes of Hf–Ni(-NH) and different model substrates. (d-2) Correlation diagram of the length of the O4–Hf bond and the relative binding energy (ΔΔE) after removal of Hf–Ni-p-NPA (hollow point) for its poor correlation.

Computationan class="Chemical">l anclass="Chemical">pan>alysis. (a-1) The computational model of complex Hf–Ni-p-NPP depicted in two parts: the square planar ligand-unit and the Hf6 cluster unit. (a-2, a-3) There are two possible spin multiplicities of Ni (singlet and triplet) and one of Cu (doublet), and three possible modes of p-NPP binding to Hf6 cluster: a, one O of p-NPP interacts with one Hf without breaking any initial O–Hf bond; b, one O of p-NPP interacts with one Hf after breaking one initial O–Hf bond; b + c, p-NPP bonds with two Hf of a Hf6 cluster after the dissociation of one HCOO–. The relative energy (ΔE) and relative binding energy (ΔΔE) are given. (b) Spin density of Hf–Ni-p-NPP (triplet). (c) Frontier molecular orbital of Hf–Ni-p-NPP (triplet). (d-1) Sketch of complexes of Hf–Ni(-NH) and different model substrates. (d-2) Correlation diagram of the length of the O4–Hf bond and the relative binding energy (ΔΔE) after removal of Hf–Ni-p-NPA (hollow point) for its poor correlation.

For interactions between the phosphate group of p-NPP and Hf6, there are three possible binding modes depending on whether the carboxylate of Hf6 is dissociated from the MOF complex (mode a, b, b+c in Figure a-3). Failure to locate the complex in mode a indicates that one vacant binding site at least is required to form the phosphate–Hf interaction. Moreover, during structural optimization, mode a would easily transform into the most favorable mode b that involves the interaction of P=O4···Hf1 and a hydrogen bond to stabilize the adduct of p-NPP and Hf6. The binding energy of the complex in mode b is 19.5 kJ/mol more stable than that in mode b+c, suggesting that the carboxylate group will not fully be dissociated from Hf–Ni catalyst during the binding with p-NPP. After establishing the mode of Hf–Ni-p-NPP, we would be able to probe the selectivity of the Hf–Ni catalyst for different model substrates in Figure d-1 based on the calculated binding energies. The sequence of stability for complexes in terms of different substrates is p-NPP ≫ BNPP > p-NPA > p-NPS, which supports the experimental observation in Figure d. As for the amino-substituted Hf–Ni catalyst, Hf–Ni–NH, it has stronger binding ability to p-NPP than Hf–Ni (ΔΔE = −5.8 kJ/mol). According to the structure of different complexes, the O4–Hf interaction seems to operate the mentioned binding energies for varying substrates and catalysts. A good correlation (R2 > 0.9) of relative binding energies and O4–Hf distances reveals that the selectivity of the Hf–Ni catalyst is highly dependent on the binding affinity of the X-O4 and Hf (Figure d-2). Through the natural population analysis (NPA) charge and the second-order perturbative interaction for different model substrates (Tables S7 and S8), the O4 atom in p-NPP exerts the strongest interaction between O4 and Hf, which stabilizes the complex Hf–Ni-p-NPP. The introduction of -NH2 enabled the forming of the intramolecular hydrogen bond N–H···O3, which thus improved the binding affinity of Hf–Ni–NH.

Further calculations focus on the mechanism of the hydrolysis of p-NPP by Hf–Ni (Figure ). A triangular bipyramid transition state was formed as a concerted O–P–O mode with the free energy barrier, ΔG1, is 76.7 kJ/mol, and the total thermodynamic free energy variation, ΔGrxn, is −10.5 kJ/mol. More inspections to the structure of the transition state indicate that all of the O1–Ni, O4–Hf, π–π stacking, and hydrogen bonds play important roles to stabilize the transition state, supported by the shorten distance of related atoms or functional groups, which is in agreement with our previous discussion in Figure .

Figure 6

Mechanism of the hydrolysis process. Sketch of essential steps of the reaction for the hydrolysis of p-NPP on Hf–Ni, and the structure of the transition state (TS) depicting the process of nucleophilic attack for H2O toward P.

Mechanism of the hydron class="Chemical">lysis process. Sketch of essenclass="Chemical">pan>tial steps of the reaction for the hydrolysis of p-NPP on Hf–Ni, and the structure of the transition state (TS) depicting the process of nucleophilic attack for H2O toward P.

Experimental Section

Synthesis of Compound Hf–Ni

HfCl4 (64 mg), HIn (62 mg), and Ni(NO3)2·6H2O (29 mg) were dissolved in DMF (6 mL) and acetic acid (1.5 mL) mixture solvent while stirring. Then the starting materials were sealed in a 20 mL glass vial. Subsequently, the resulting solution was kept at 100 °C for 3 days and then cooled down to room temperature at the rate of 1.5 °C per hour to acquire pale-blue cube-shaped crystals. Then, the resulting crystals were directly washed with DMF three times and air-dried naturally.

Notes: The experimental operation of dissolving HfClshould be conducted in a fume hood.

X-ray Crystallographic Analyses

Data were collected on an Oxford SuperNova (TM) CCD diffractometer with a SuperNova X-ray source (Mo–Kα). The structures were solved and refined using the SHELXS-97 program. Detail analysis of data collection and refinement can be viewed in Supporting Information.

Characterization Methods

All compounds were characterized using general experimental techniques, such as powder X-ray diffraction, thermogravimetry analyses, IR, XPS, and ICP. Details are provided in Supporting Information.

General Catalytic Procedures

The reaction was conducted in HEPES (50 mM, pH 7.4). The concentrations of class="Chemical">pan> class="Chemical">Hf–Ni and p-NPP were 0.2 mM and 1 mM, respectively. Catalytic activity was monitored by measuring the absorbance at 400 nm for p-nitrophenolate using a LAMBDA 950 UV/vis spectrophotometer.

Simulated Phosphatase Activity Assays

p-Nitrophenyl phosphate was used as a substrate to describe the phosphatase mimetic activity of Hf–Ni. All parallel experiments were conducted at least three times with triplicates. Hf–Ni (500 μM) was dispersed in HEPES buffer in advance. A 300 μL catalyst solution was directly added to buffer in a 3 mL cuvette and followed by the addition of various concentrations of p-NPP (0.05 mM to 6 mM). After an appropriate shaking, the reactive substance was examined with a UV/vis spectrophotometer as quickly as possible. Then, we obtained the data which were collected using the kinetic mode and the plots of absorbance to time.

Cell Culture and Cytotoxicity Assays

The cytotoxic activity of Hf–Ni was analyzed by the MTT assay. The cells were seeded in 96-well plates at a density of 5 × 103 cells per well and incubated overnight. The treatments were added to the cells at different concentrations (from 3.2 mg mL–1 to 0.025 mg mL–1) and kept 4 h at 37 °C with a 5% CO2 atmosphere. The cytotoxicity was determined after 4 h by adding the MTT. The plates were read at 562 nm.

Hemolysis Assays

Hemolysis assays were performed as follows: 50 μL of Hf-M-MOFs (suspended in PBS), PBS (negative control), or Triton X-100 (positive control) are added to 450 μL of RBC solution to form a suspension, which was incubated at room temperature for 1 h. Then, the suspension was centrifuged at 10000g over 5 min, and the supernatant was read in a 96-well plate at 540 nm. The % hemolysis was calculated as H (%) = (OD540 nm sample – OD540 nm PBS)/(OD540 nm Triton X-100 1% – OD540 nm PBS) × 100. Positive and negative controls induced 100% and 0% of lysis, respectively.

Computational Methods

The M06 functional with an uclass="Chemical">pan> class="Chemical">ltrafine integration grid was employed as implemented in the Gaussian09 code for all of the DFT calculations. Geometry optimizations and harmonic frequencies were obtained with the 6-31G** basis set for main-group elements and the Lanl2dz basis set and pseudopotentials for Hf, Ni, and Cu. To improve the accuracy, energies were further refined via the addition of diffuse functions in the main-group basis set (6-31++G**) and the correction of the SMD (H2O) model. A simplified model of the Hf6 cluster was adopted in this work, including one porphyrin-liked ligand-unit with isonicotinic acid ligands, metal Ni2+ (or Cu2+), and Cl–, while other linkers to the Hf6 cluster were replaced with formate capping ligands.

More detailed experimental procedures and the characterization data are available as Supporting Information.

Conclusion

In summary, two porous frameworks assembn class="Chemical">led by [class="Chemical">pan> class="Chemical">Hf48Ni6] cubic nanocages were synthesized and structurally characterized by regulating an organic linker, which feature high stability in various solvents and acid/base solutions. With a biomimetic strategy, these MOFs were used to mimic protein phosphatase for dephosphorylation, and results revealed that both Hf–Ni and Hf–Ni–NH have excellent type-selectivity and position-selectivity for hydrolysis of p-NPP. Moreover, the sustainable catalysts are relatively active in mild catalytic reactions even when compared to their enzyme counterparts and possess good reusability as well as hypotoxicity. The reactivity and selectivity of these catalysts were further verified by control experiments and theoretical DFT calculations through calculated binding energies and structural details of reaction intermediates and transition states. For the first time, this study realizes the accurate catalytic dual-selectivity in the MOFs system and provides valuable clues for the design of biomimetic materials with controllable selectivity, which also paves a possible way for the development of specific therapeutic MOFs. Especially, establishment of biomimetic platforms with clarified principles and tunable properties (e.g., selectivity) to probe detailed information that cannot be facilely investigated in their natural enzyme counterparts can greatly promote the study and applications of this field.