Insights into the Interaction between Immobilized Biocatalysts and Metal–Organic Frameworks: A Case Study of PCN-333

Abstract

The immobilization of enzymes in metal–organic frameworks (MOFs) with preserved biofunctionality paves a promising way to solve problems regarding the stability and reusability of enzymes. However, the rational design of MOF-based biocomposites remains a considerable challenge as very little is known about the state of the enzyme, the MOF support, and their host–guest interactions upon immobilization. In this study, we elucidate the detailed host–guest interaction for MOF immobilized enzymes in the biointerface. Two enzymes with different sizes, lipase and insulin, have been immobilized in a mesoporous PCN-333(Al) MOF. The dynamic changes of local structures of the MOF host and enzyme guests have been experimentally revealed for the existence of the confinement effect to enzymes and van der Waals interaction in the biointerface between the aluminum oxo-cluster of the PCN-333 and the -NH₂ species of enzymes. This kind of host–guest interaction renders the immobilization of enzymes in PCN-333 with high affinity and highly preserved enzymatic bioactivity.

Introduction

Metal–organic frameworks (MOFs) are a class of extended network materials constructed from a bottom-up building block approach by connecting metal-based clusters and organic ligands.¹⁻³ MOFs are well-known for their high degree of crystallinity, remarkable surface areas and pore volumes, and structural tunability.^4,5 These design features have motivated researchers from a broad range of scientific and engineering disciplines to explore the fundamental and applied aspects of MOF materials.⁶ Recently, the area of enzyme/MOF biocomposites has experienced considerable growth due to their potential applications in biotechnology.^7,8 Biomolecules can interact with MOFs via a wide variety of interactions including covalent attachment, coordination bonds, hydrogen bonding, π–π and hydrophobic interactions, and electrostatic attraction.⁹ The most widely employed strategy to synthesize enzyme/MOF biocomposites is to infiltrate the biomacromolecule within the MOF pore network.¹⁰ Due to their modular construction and chemical mutability,¹¹ MOFs allow for their pore environment to be optimized to house specific proteins to a level of precision that is not attainable for other porous materials,^12,13 such as silica matrixes,¹⁴ organic polymers,¹⁵ and DNA-based compartments.¹⁶ To date, these MOF-based biocomposites have been explored for applications in biocatalysis,⁷ biomedicine,¹⁷ controlled release and drug delivery,¹⁸ biopreservation,¹⁹ and cell and virus manipulation.⁸

However, the majority of studies, nowadays, have focused on the synthesis and performance characteristics of enzyme/MOF biocomposites;^20,21 little is known about the fundamental molecular interactions between the MOF and enzyme that govern the immobilization process. When an enzyme molecule is in contact with a MOF support, its structure, orientation, and conformational mobility can be modified, and this typically leads to a change in its native functionality such as a significant loss of catalytic activity.²² Thus, to maintain the activity of enzymes immobilized within MOF pores a detailed understanding of the host–guest interactions is necessary. A number of characterization techniques have been employed to examine how MOF pore environment influences the structure of an enzyme,²³ including, circular dichroism (CD),²⁴ electron paramagnetic resonance (EPR),^25,26 diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS),²¹ and Raman spectroscopy.²⁷ In recent years, methodological developments have led solid-state nuclear magnetic resonance (ssNMR) to be one of the most powerful tools for determining the local structure and dynamics of various materials, including proteins, catalysts, biomaterials, pharmaceuticals, metal–organic materials, and so on.²⁸⁻³⁰ For example, Fragai et al.³¹ showed that the three-dimensional structures of the enzymes were retained after immobilization in a peptide-templated silica gel. Indeed, ssNMR spectroscopy has been applied as a general approach to study MOFs and their interactions with small guest species³² (e.g., H₂,³³ xenon,³⁴ CO,³⁵ CO₂,^36,37 nitric oxide,³⁸ H₂O,³⁹ hydrocarbon,⁴⁰ aromatics,^41,42 etc.), providing complementary information for host–guest interaction.⁴³ For example, with the help of ssNMR, the coordination of methanol to the open metal sites in [Cu_3–xZn_x(btc)₂]_n (btc = 1,3,5-benzenetricarboxylate) has been proven and unveiled in detail at an atomic level.⁴⁴ In another study, the interactions between the Cr of the host MOF (Cr-MIL-101) and guest molecules (e.g., water, 2-aminopyridine, 3-aminopyridine, and diethylamine) have been revealed by a combination of density functional theory (DFT) calculations and ¹³C–¹H rotational-echo double-resonance (REDOR) NMR.⁴⁵ In addition, ssNMR is advantageous to probe the host–guest interaction in other porous materials such as zeolites,^46,47 amorphous silica–alumina,⁴⁸ and mesoporous organic materials.^49,50 However, ssNMR has not yet been employed to examine the atomic-level host–guest interactions in MOF-based biocatalysts.

In this work, we envision that comprehensive microscopy, ssNMR, and reaction study can provide data that will advance our fundamental understanding of the host–guest interactions in the enzyme–MOF biointerface allowing for these systems to be optimized for enhanced enzyme preservation and protection. Lipase and insulin were loaded in a mesoporous MOF, PCN-333(Al), to afford lipase@PCN-333 and insulin@PCN-333 biocatalysts. The routine characterizations (X-ray diffraction (XRD) and 77 K N₂ gas adsorption analysis) combined with scanning electron microscopy (SEM), confocal laser scanning microscopy (CLSM), and enzymatic tests have confirmed the successful immobilization of the targeted biocatalyst (lipase and insulin) in PCN-333. Furthermore, multinuclei 1D and 2D ssNMR and their multiple-quantum, double-resonance, and relaxation ssNMR measurements were employed to elucidate the local chemical environment and host–guest interactions between immobilized enzymes and the framework, PCN-333. The reaction results showed that the discovered host–guest interaction is able to confine the lipase inside PCN-333 while preserving its enzymatic bioactivity after immobilization.

Results and Discussion

PCN-333 has been reported to exhibit high porosity and exceptional chemical stability. These physical characteristics endow PCN-333 as an excellent platform for biomolecule immobilization. PCN-333 has been synthesized using a variety of metal precursors to realize metal nodes based on aluminum (Al), iron (Fe), and scandium (Sc).⁵¹ In this work, the Al version of PCN-333, PCN-333(Al), was selected for enzyme immobilization to facilitate the application of ssNMR techniques. The PCN-333 framework consists of [Al₃(μ₃-O)(OH)(H₂O)₂]^6– clusters (Al-oxo clusters) linked by tridentate TATB (4,4′,4″-s-triazine-2,4,6-triyl-tribenzoate) ligands, which form supertetrahedra (ST). These ST are connected via vertices to form a three-dimensional mtn network. This network topology affords two types of mesopores: a dodecahedral cage measuring 42 Å and a hexacaidecahedral cage measuring 55 Å (Figure 1b). PCN-333 was prepared and activated according to the literature procedure reported by Zhou et al.⁵¹ The permanent porosity of PCN-333 was assessed by performing 77 K N₂ gas adsorption isotherms. Application of the Brunauer–Emmett–Teller (BET) equation to the isotherm data afforded a surface area of 4198 m² g^–1 (Figure S1), which is consistent with the reported values.⁵¹ Recent publications have reported the successful immobilization of different enzymes within the pore network of PCN-333, including horseradish peroxidase (HRP),⁵¹ cytochrome c (Cyt c),⁵¹ microperoxidase-11 (MP-11),⁵¹ catalase (CAT),⁵² superoxide dismutase (SOD),⁵² and alcohol dehydrogenase (ADH).⁵³ In this study, we have selected two commercially available enzymes of different sizes for the immobilization experiments: lipase (3.0 × 4.0 × 5.0 nm³)⁵⁴ and insulin (1.3 × 1.3 × 3.4 nm³)⁵⁵ (Figure 1c). Lipase is a commercially important enzyme that catalyzes hydrolysis reactions,⁵⁶ and insulin is an important medical hormone responsible for regulating the concentration of glucose in blood plasma.⁵⁷ We note that both biomolecules, in their active conformation (in crystallographic dimensions), are smaller than that of the largest mesopore (ø = 5.5 nm) in PCN-333 (Figure 1b) and, thus, in principle, can infiltrate the MOF pores.

Figure 1.

(a) Simulated (gray) and experimental PXRD patterns of PCN-333 (black), lipase@PCN-333 (blue), and insulin@PCN-333 (red). (b) Pore size distributions of PCN-333 (black), lipase@PCN-333 (blue), and insulin@PCN-333 (red). Pore size distributions were calculated using NLDFT model based on the N₂ adsorption isotherms measured at 77 K (Figure S1). The salient pore features of PCN-333 are shown as insets. (c) Molecular structure of insulin (PDB 2A3G, insulin from bovine pancreas) and lipase (PDB 1OIL, lipase from Pseudomonas cepacia).

To immobilize the selected enzymes, PCN-333 crystals were soaked in an aqueous solution of lipase or insulin for 1 h at 4 °C (see Experimental Section for details). Control experiments confirmed the integrity of PCN-333 under these immobilization conditions (Figure S5). The resulting MOF-based biocatalysts were washed with water three times to remove the excess or loosely absorbed enzyme species on the MOF surface, affording biocatalysts denoted as lipase@PCN-333 and insulin@PCN-333. The loadings of lipase and insulin in lipase@PCN-333 and insulin@PCN-333 (P_loading¹⁰) were calculated to be 26.9 and 50.0 mg g^–1, respectively (Bradford assay,⁵⁸Figure S2). The values of loading efficiency (Y_P¹⁰) were then determined to be 53.8% and 100% for lipase and insulin, respectively. To determine if PCN-333 retained bulk crystallinity subsequent to the infiltration process, we performed powder X-ray diffraction (PXRD) measurements. Figure 1a shows that the PXRD patterns of both the lipase and insulin biocomposites are essentially identical to pristine PCN-333. The porosity of the biocomposites was examined by carrying out N₂ 77 K gas adsorption isotherms. As anticipated, the total N₂ uptakes for lipase@PCN-333 and insulin@PCN-333 decrease significantly as compared to that of PCN-333 (Figure S1). In addition, the normalized pore size distributions calculated via the nonlinear density function theory (NLDFT) show a disappearance of mesoporosity in lipase@PCN-333 and insulin@PCN-333 (Figure 1b). These results can be rationalized by the presence of large biomolecules (lipase or insulin) within the mesopores of PCN-333. Attempts to characterize structural features of the encapsulated biomolecules within the enzyme@PCN-333 samples using attenuated total reflection Fourier transform infrared spectroscopy (ATR-FTIR) were unsuccessful due to the interference of PCN-333 background peaks (Figure S3).

We then applied scanning electron microscopy (SEM) to ascertain the crystal size and morphology of PCN-333 and the enzyme@PCN-333 biocomposites. Close inspection of the micrographs shows that PCN-333, lipase@PCN-333, and insulin@PCN-333 samples were dispersed octahedral crystals (∼300 nm in diameter, Figure S4). The morphological homogeneity and crystal size triggered us to visualize the spatial distribution of lipase within PCN-333 using confocal laser microscopy (CLSM). In order to visualize the lipase within the biocomposite crystals using CLSM, a fluorescent tag needs to be attached to the enzyme. Fluorescein-tagged lipase (FITC-lipase) was prepared and used to synthesize FITC-lipase@PCN-333. The synthesis and washing protocol for FITC-lipase@PCN-333 were analogous to those of lipase@PCN-333. CLSM images (Figure S6) show that the FITC-lipase molecules were present throughout FITC-lipase@PCN-333 crystals; however, due to the resolution of this technique, we are unable to differentiate between signal from surface-absorbed FITC-lipase and encapsulated FITC-lipase. Supplementary experiments with HRP@PCN-333 confirmed the successful infiltration/encapsulation of the biomolecules within the mesopores of PCN-333 (Figures S13–S17).

To assess the enzymatic performance of the biocomposite, we measured the catalytic hydrolysis of p-nitrophenyl butyrate (PNPB) using free FITC-lipase and FITC-lipase@PCN-333 biocatalyst (Figure 2a). A noteworthy result is that FITC-lipase@PCN-333 showed comparable activity to that of free FITC-lipase (Figure 2b). Approximately 91.6% of the bioactivity (calculated based on the slope of the reaction kinetics) of FITC-lipase was retained upon immobilization in FITC-lipase@PCN-333.

Figure 2.

(a) Biocatalytic hydrolysis of p-nitrophenyl butyrate (PNPB). (b) Catalytic activity of FITC-lipase and FITC-lipase@PCN-333 biocomposite. The amount of FITC-lipase available was the same for each experiment (25 μg based on FITC-lipase). (c) Cycling runs for the hydrolysis of 4-nitrophenyl butyrate in the presence of FITC-lipase@PCN-333 after a reaction time of 2 min. Experiments were performed in triplicate with standard error bars shown for each run.

Given that the catalytic activity of the immobilized lipase is preserved for FITC-lipase@PCN-333 (Figure 2b), we sought to examine recyclability, an important performance factor for practical applications. The relative bioactivity of FITC-lipase@PCN-333 biocomposites was calculated by comparing the relative activity with that of the first run. We successfully cycled FITC-lipase@PCN-333 six times with a 22.4% reduction in enzymatic activity in the sixth cycle (Figure 2c). This observed reduction is actually due to enzyme leaching from the FITC-lipase@PCN-333 samples during the washing procedure between cycles (Figure S7 and Table S1–S2). Interpreting these results, it is concluded that the lipase biomolecules retained their bioactivity and are confined in PCN-333 framework in relatively high affinity.

To elucidate the interactions between the enzyme and PCN-333, we studied PCN-333, lipase@PCN-333, and insulin@PCN-333 using ssNMR techniques. A unique characteristic of MOFs is that their structural flexibility enables their molecular architectures to exhibit local motions under external stimuli.^59,60 Upon pore-filling with large biomolecules, we anticipate a subtle difference in the local flexibility of the PCN-333 framework. Solid-state NMR has been proven to be a powerful tool to investigate the molecular dynamics in a variety of materials by applying relaxation or diffusion spectroscopy.⁶¹⁻⁶³ In this work, in order to understand the structural flexibility of the framework, the ¹H spin–lattice relaxation (¹H T₁) was measured by applying the inversion–recovery pulse sequence. For ¹H T₁ analysis, as the magnetization of ¹H nuclei can rapidly diffuse throughout the entire spin network of the solid, a relatively uniform T₁ value for all protons in the material is expected. In addition, due to the strong homonuclear dipolar–dipolar interactions and the high natural abundance of the ¹H nuclei, the ¹H T₁ value of the whole set of ¹H peaks can be utilized to represent the molecular dynamics of the entire network. Therefore, the peaks across the whole spectrum in the ¹H spectra obtained via the application of the inversion–recovery pulse sequence were selected to acquire the relaxation curves, and the ¹H T₁ values are obtained accordingly (Figure S8 and Table 1).

Table 1. Molecular Dynamics Measurements for PCN-333, Lipase@PCN-333, and Insulin@PCN-333 via ¹H and ²⁷Al Relaxation Spectroscopy.

nuclei	materials	T1 (ms)
1H	PCN-333	936
	insulin@PCN-333	2323
	lipase@PCN-333	2342
27Al	PCN-333	14.92
	insulin@PCN-333	11.23
	lipase@PCN-333	11.05

In general, increasing mobility in a solid will result in reduced ¹H spin–lattice relaxation times.⁶⁴ As shown in Table 1, after enzyme immobilization, lipase@PCN-333 and insulin@PCN-333 show a much longer spin–lattice relaxation time compared to PCN-333 (T₁ = 2.342, 2.323, and 0.936 s for insulin@PCN-333, lipase@PCN-333, and PCN-333, respectively, Table 1 and Figure S8a). These results indicate that the mobility of protons in lipase@PCN-333 and insulin@PCN-333 decreases as compared to those in PCN-333. Since these ¹H T₁ values reflect the flexibility of the entire network, a longer T₁ value indicates a reduction in local framework flexibility upon enzyme immobilization. This effect is expected and can be rationalized by the confinement of the large enzyme biomolecules in lipase@PCN-333 and insulin@PCN-333.⁶⁵ A similar phenomenon has been observed in the case of benzene loaded UiO-66(Zr), where a decrease in the internal dynamics of π-flips of the terephthalate ligands was observed.⁴¹

T₁ relaxation times of ²⁷Al were also measured with the purpose of comparing the local mobility of the Al sites in different PCN-333 samples. Interestingly, as shown in Table 1 and Figure S8b, after enzyme immobilization, a decrease in the ²⁷Al T₁ relaxation time is observed for both lipase@PCN-333 and insulin@PCN-333. The decrease, instead of increase, in the ²⁷Al T₁ relaxation times after enzyme immobilization clearly demonstrates that the local flexibility of Al centers in the PCN-333 framework has a different dynamic behavior compared with the entire framework (Table 1).⁶⁴ This observation points to the possible existence of interaction between the Al-oxo cluster and enzymes, which relax the existing coordination bonds of Al-oxo cluster in PCN-333 and contributed to a lower ²⁷Al T₁ value.

In order to further illustrate the structural change of PCN-333 after enzyme immobilization in detail and understand the nature of the host–guest interactions at the atomic scale, one-dimensional and two-dimensional ssNMR spectroscopies were conducted to investigate the local chemical structure of the PCN-333 biocomposites. The pristine PCN-333 (Figure 3) was first analyzed by ssNMR to establish a reference point. By application of ¹H and ¹³C magic-angle spinning (MAS) NMR spectroscopy, all the proton and carbon species in PCN-333 (Figure 3) were successfully assigned. The broad ¹H signal at 7–9 ppm is partially attributed to the coordinated water on the trimeric Al-oxo cluster (Al···OH₂, denoted as H4 in Figure 3a), which shows a higher ppm value in comparison with that of the free H₂O (δ_¹H = ca. 5 ppm). This assignment is supported by recent ssNMR studies.^66,67 In one study of MIL-110, Haouas et al. reported a higher ¹H chemical shift value for the coordinated water (δ_¹H = 5.8 ppm) than that of the free water (δ_¹H = 4.4 ppm) in the framework.⁶⁶ In another case, the coordinated water (Al–H₂O) in mineral alunite signaled at a lower chemical field shift in the ¹H NMR spectrum (δ_¹H ≈ 7.4 ppm) in comparison with the free H₂O (δ = 4.8 ppm).⁶⁷ In addition, these broad ¹H peaks (δ_¹H = 7–9 ppm) are associated with the aromatic C–H species (H1 and H2) of the TATB ligand (Figure 3a,b).⁶⁸ The ¹H signal at δ_1H = 2.4 ppm is assigned to the proton of the coordinated hydroxyl (H3, Figure 3a) species in the Al-oxo cluster (Al₃(μ₃-O)(OH)(H₂O)₂).⁶⁹ A broad ¹H signal at around 5 ppm is assigned as the free H₂O molecule.^38,69 These assignments are consistent with an ssNMR study of MIL-100(Al).⁶⁹ However, no proton signal was identified in the region of ∼2 ppm for MIL-100(Al), as an extra-framework btc ligand instead of a hydroxyl group is present in MIL-100(Al) to balance the charge of the Al-oxo cluster. In another study, in the absence of extra-framework btc ligand in MIL-100(Al), hydroxyl protons (−OH) peaks appear at δ_¹H = 1.5 and 0.3 ppm.³⁸Figure 3c (gray line) shows the ¹³C CP MAS NMR spectra of the PCN-333 samples. The aromatic C2 and C5 of the TATB ligand can be observed at δ_¹³C = 138 ppm, while C3 and C4 sit at δ_¹³C = 129 ppm. Deconvolution of the broad peak at 170–174 ppm confirms the existence of the C1 (δ_¹³C = 173 ppm) and C6 species (δ_¹³C = 171 ppm) (Figure 3d and Table S3). Other small peaks originating from the residue solvent were also observed (N,N′-dimethylformamide (DMF), Figure 3b,c).⁷⁰

Figure 3.

(a) Molecular representation of PCN-333(Al). (b) ¹H one-pulse MAS NMR spectra. (c) ¹³C cross-polarization MAS NMR spectra of PCN-333 (black), lipase@PCN-333 (blue), and insulin@PCN-333 (red). Asterisks represent sidebands; filled circles indicate the residue solvent signals (DMF). (d) Deconvolution of ¹³C peaks in the region of 150–200 ppm.

After enzyme immobilization, a drastic change in the ¹H spectra can be observed for enzyme@PCN-333 as compared to that of PCN-333 (Figure 3b). In lipase@PCN-333 and insulin@PCN-333, a sharp ¹H signal emerges at 5 ppm (Figure 3b). The sharp ¹H signal (δ_¹H = 5 ppm) can also be identified in ¹H MAS NMR spectra of free enzymes (Figure S9). Compared to the broad water proton signal in PCN-333 (δ_¹H ≈ 5 ppm, Figure 3b), the sharp ¹H peak centered at 5 ppm for lipase@PCN-333 and insulin@PCN-333 can be assigned to the -NH₂ species (Figure 3b) in enzyme biomolecules⁶⁹ and the free H₂O in the biocomposites. In the ¹³C cross-polarization (CP) MAS NMR spectra (Figure 3c), no distinguishable changes are observed for the aromatic carbon species (C2, C3, C4, and C5) of the TATB ligand before and after enzyme immobilization. ¹³C NMR signals originating from carbonyl species (C=O) in enzyme molecules can be found at 173 ppm,⁷¹ which overlaps with the C1 ¹³C CP NMR signal from PCN-333 (Figure 3d). In addition, only tiny ¹³C resonance peaks originating from the aliphatic carbon of immobilized protein molecules are also detected in Figure 3c (δ_¹³C = 20–40 ppm) for lipase@PCN-333 and insulin@PCN-333.⁷¹ The coexistence of ¹H and ¹³C signals of the PCN-333 framework and the guest biomolecule in enzyme@PCN-333 samples confirms the successful immobilization of lipase and insulin. This observation is in agreement with the characterization data in Figure 1, Figure S1, and Figure S6. The absence of new carbon species in the ¹³C CP MAS NMR spectra for PCN-333 biocomposites samples implies that there are no covalent host–guest interactions between the MOF and the enzyme biomolecules. However, after enzyme immobilization, the signal related to the carboxylate carbon (C6) shifts to a higher field (δ_¹³C = 169 ppm) for both PCN-333 biocomposite samples (see Figure 3d and Table S3 for details). These chemical shifts of the carboxylate C6 species are reported to be an indicative probe for studying the variation in the coordination environment of metal nodes in a metal–organic complex.⁷² For example, Ke et al. studied the ssNMR shift of carboxylates carbon of benzoate in [Cu₂(C₆H₅CO₂)₄]_n to determine the nature of the interaction between the Cu-paddle wheel and urea.⁷³ The authors claimed that upon coordination of urea to the open Cu sites, the chemical signal of the carboxylate carbons in the benzoate ligand shifted to high-fields.⁷³ In the present study, a high-field shift in ¹³C signal of carboxylate carbons (C6) is also observed in enzyme@PCN-333 samples (from 171 ppm in PCN-333 to 169 ppm in lipase@PCN-333 and insulin@PCN-333, Figure 3d and Table S3). The high-field shift of C6 is consistent with the literature, indicating a change of the local chemical environment of the Al-oxo clusters in PCN-333 and possible host–guest interaction between the Al-oxo clusters of host PCN-333 and the guest biomolecules. This result is also in good agreement with the information obtained from the ²⁷Al T₁ ssNMR relaxation measurements (Table 1 and Figure S8).

To reveal the existence of interaction between the PCN-333 and the immobilized biomolecules, ²⁷Al MAS NMR technique^74,75 was applied to characterize the local chemical environment of Al nuclei in PCN-333 samples. In the ²⁷Al MAS NMR spectrum, only one broad peak is observed in the region of 8–15 ppm for PCN-333 (Figure 4a and Figure S10). The ²⁷Al peak position indicates the presence of hexa-coordinated Al³⁺ (Al^VI) cations^51,76 in PCN-333. After enzyme immobilization, the width of the ²⁷Al peak decreased (Figure S10, as measured by the values of the full width at half-maximum (fwhm), Table S4). This difference in the shape of ²⁷Al spectra further confirms the change in the chemical environment of the Al centers in PCN-333 after enzyme immobilization. These results are consistent with the information obtained via ²⁷Al NMR relaxation spectra (Table 1, Figure S8b). Since ²⁷Al is a quadrupolar nucleus, the strong second-order quadrupolar interaction can lead to broad peaks in the 1D ssNMR spectra and hinder the resolution of subspecies.⁷⁷ To obtain detailed information on the Al-oxo cluster, two-dimensional multiple-quantum ²⁷Al MAS NMR spectroscopy (2D ²⁷Al MQ MAS NMR) was applied to resolve the different Al species (Figure S11a–c). For PCN-333, two contour centers for Al^VI species with different isotropic chemical shifts are observed (Figure S11a and Table S5), denoted as Al^a (δ_iso²⁷Al = 2.7 ppm) and Al^b (δ_iso²⁷Al = 10.0 ppm). Based on the parameters obtained from 2D ²⁷Al MQ MAS NMR spectra and eqs 1–4 (see Experimental Section for details), the anisotropic 1D ²⁷Al MAS NMR spectra were deconvoluted (Figure 4a). As the peak area of the Al^a is approximately double that of the Al^b species (Al^b/Al^a = 0.48), the Al^a and Al^b species are assigned to Al···OH₂ and AlOH groups, respectively, in the Al-oxo cluster (Al₃(μ₃-O)(OH)(H₂O)₂). The identification of the two isotropic resonances in the spectra is consistent with the observation reported by Loiseau et al.,⁷⁸ who also observed two Al^VI species originated from AlOH and Al···OH₂ groups in MIL-110(Al). Importantly, this deconvolution result also matches well with the crystallographic data.⁵¹ After biomolecule loading, no new Al species is observed (Figure S11a–c). Both Al^a and Al^b species remain at similar isotropic chemical shifts with slightly increased quadrupole coupling constant (C_QCC) (Table S5). The nonshifted Al^a and Al^b species in peak position confirm the absence of strong coordination interaction at the biointerface. The change in C_QCC values of the Al sites points to only a slight distortion in the symmetry of the Al coordination environment (Table S5), which is not strong enough to cause a severe change in the electric field gradients on the Al sites. The small change in quadrupolar coupling constant further confirms the appearance of a new weak interaction between the Al-oxo cluster of the PCN-333 and the biocatalysts. In order to visualize and reveal the nature of the weak interaction between the Al-oxo cluster of the PCN-333 and the biocatalysts in-depth, ¹H/²⁷Al transfer of population in double-resonance (TRAPDOR) NMR technique was applied to study the correlation between the Al atoms in PCN-333 and their neighboring protons.⁷⁹Figure 4b shows the amplified ¹H/²⁷Al TRAPDOR NMR spectra of PCN-333 samples, in which both Al···OH₂ (δ_¹H = 7–8 ppm) and AlOH (δ_¹H = 2.4 ppm) can be observed. This observation further confirmed the proton assignment in the ¹H spectra (Figure 3b). Interestingly, a strong correlation between the -NH₂ (narrow signal δ_¹H = 5 ppm) and Al nuclei was observed for lipase@PCN-333 and insulin@PCN-33 (Figure 4b). Since the physisorbed water molecule is extremely dynamic compared to other molecules, the majority of the ¹H–²⁷Al dipolar couplings between the water molecules and Al-oxo cluster are averaged away. The absence of a sharp peak at δ_¹H = 5 ppm in PCN-333 rules out the possibility that the ¹H–²⁷Al correlation observed at 5 ppm for enzyme@PCN-333 samples was derived from the signal between the exchangeable water molecule and Al sites of PCN-333 (Figure 4b and Figure S12). Thus, the identified correlation can be assigned to the ¹H–²⁷Al dipolar couplings between the -NH₂ group of the enzymes (δ_¹H = 5 ppm) and Al nuclei of PCN-333. This result clearly indicates the presence of proximity and host–guest interaction between the -NH₂ group of the immobilized biocatalyst and the Al-oxo cluster of the PCN-333. Since no shift of peak position is observed for the -NH₂ group in the ¹H/²⁷Al TRAPDOR spectra, there is no strong coordination, hydrogen bonds, or covalent interactions between the Al centers and the -NH₂ functionalities. Thus, with weak van der Waals interaction between the Al sites and -NH₂ sites, the biocatalysts were confined inside the cavities of the PCN-333 and immobilized on the PCN-333 crystal surface in relatively high affinity. In addition, since the signal intensity of the -NH₂ is much stronger than those of the AlOH and Al···OH₂ groups, we conclude that the interaction between the enzyme and PCN-333 is primarily governed by the interactions between the Al-oxo clusters of the host PCN-333 and the -NH₂ moieties of the guest biocatalysts.

Figure 4.

(a) Deconvolution of ²⁷Al MAS NMR spectra of PCN-333 samples. (b) ¹H/²⁷Al TRAPDOR difference spectra of PCN-333 samples (10 times amplified). (c) Schematic illustration of enzyme–PCN-333 interaction in the biointerface.

All experimental data are consistent with the hypothesis that the confined biocatalysts in PCN-333 are predominantly governed by the interactions between the Al sites of PCN-333 and the -NH₂ functionalities of biocatalysts. The strong ¹H–²⁷Al correlation undoubtedly suggests that the host–guest interactions are with -NH₂ functionalities to Al-oxo cluster (δ_¹H = 5.0 ppm in ¹H/²⁷Al TRAPDOR spectra). We speculate that surface-exposed lysine or arginine residues of lipase and insulin play a key role in the enzyme–MOF interaction (Figure 4c). Meanwhile, the overall dynamic change (T₁ of ¹H NMR) of enzyme@PCN-333 implies the existence of the confinement effects for immobilized enzymes in MOFs.

Conclusion

In conclusion, enzyme biomolecules (lipase and insulin) have been successfully immobilized in PCN-333 MOF and maintained high biocatalytic activity. The change of both the relaxation times and chemical shifts in ssNMR studies indicates that the immobilization of proteins affects both the flexibility and local structure of PCN-333. After enzyme immobilization, the overall framework flexibility was decreased (increase in ¹H T₁ values) due to the confinement effects to enzymes from pores, while the ²⁷Al spin–lattice relaxation time for enzyme@PCN-333 showed a different trend due to the interaction between the Al-oxo cluster in PCN-333 and the biocatalysts. Analysis of the enzyme@PCN-333 revealed that a distinct signal related to the -NH₂ group of the biocatalyst in the ¹H ssNMR spectra is concomitant with a subtle chemical shift of carboxylate ¹³C spectral feature that is associated with a change of the coordination environment in the Al sites of PCN-333. These data suggest the dominant interactions between the biomolecules and the Al oxo-cluster of PCN-333. Via both ¹H/²⁷Al TRAPDOR NMR and multiple quantum ²⁷Al MAS NMR, the immobilization mechanism was further revealed, where the spectroscopic data support the existence of weak interactions between the Al atoms in MOF and the -NH₂ sites on the biocatalysts via van der Waals forces. After combining all the experimental evidence, we, for the first time, have experimentally visualized the immobilization of enzymes in PCN-333 and deduced the possible host–guest interaction mode between the MOF’s Al oxo-cluster and the biocatalyst’s -NH₂ sites as well as the pore confinement effect. This work shed light on the complex relationship between enzyme and MOF support in the biointerface. In addition, this fundamental knowledge will help to guide the design of other supported enzymatic systems with optimum catalytic activity and stability.

Experimental Section

Syntheses

All chemicals and solvents were purchased from commercial sources and used as received without further purification.

Ligand Synthesis

4,4′,4″-s-triazine-2,4,6-triyl-tribenzoate (TATB) was synthesized according to the literature.⁵¹

PCN-333 Synthesis

PCN-333 was synthesized according to the literature.⁵¹ The as synthesized MOF materials were then washed sequentially with DMF (3 × 2 mL) and ethanol (3 × 2 mL) over 2 days. Thereafter, the PCN-333 was collected by centrifugation and dried in vacuum at room temperature. The dried PCN-333 was stored in a sealed vial at 4 °C before further usage and characterization.

Enzyme@PCN-333 Synthesis

Enzyme@PCN-333 samples were synthesized by mixing lipase (1 mL of 5 mg mL^–1 lipase aqueous solution) or insulin (1 mL of 5 mg mL^–1 insulin stock solution) with the as-synthesized PCN-333 (100 mg) at 4 °C for 1 h. The insulin stock solution was prepared by dissolving 5 mg of insulin in 1 mL of diluted HCl solution (pH 3.0, 20 mM). Thereafter, the solid was recovered by centrifugation and washed with H₂O (3 × 1 mL) to remove the excess biomolecule. All supernatants were collected. The concentration of biomolecule in the supernatant was determined by the Bradford assay. The amount of biomolecule immobilized in PCN-333 was calculated by the difference between the concentration used in the synthesis and that in the collected supernatant.

Fluorescein-Tagged Lipase (FITC-Lipase)

A solution of fluorescein isothiocyanate (FITC, 0.005 M, 0.0003 mmol) in aqueous carbonate/bicarbonate buffer solution (60 μL, 0.1 M, pH 9.2) was added to a solution of lipase (10 mg, Sigma-Aldrich 62288 lipase from Pseudomonas cepacia, 30 units mg^–1) in aqueous carbonate/bicarbonate buffer solution (1.94 mL, 0.1 M, pH 9.2) at 0 °C. The solution was covered with aluminum foil and stirred at 0 °C for 4 h. The reaction solution was then filtered through Sephadex G-25M, PD-10 column (GE Healthcare, 17085101), which had been equilibrated with water, eluting with aqueous carbonate/bicarbonate buffer solution (0.1 M, pH 9.2). The fractions containing FITC-lipase were concentrated through a Pall corporation Microsep Advance 10K membrane by centrifugation (4000 rpm; multiple batches were required ∼20 min). The concentrated FITC-lipase was washed twice with deionized water and combined to give a final volume of 2 mL, which was stored at 4 °C, protected from light, until required. The stock solution was assumed to contain 5 mg mL^–1 of lipase.

FITC-Lipase@PCN-333 Synthesis

A solution of aqueous FITC-lipase (100 μL, 5 mg mL^–1) was added to PCN-333 powder (10 mg) at 0 °C. The suspension was stirred at 0 °C for 1 h. The suspension was transferred to an Eppendorf tube, with deionized water and centrifuged (6000 rpm, 2 min). The supernatant was removed, the pellet was resuspended in deionized water and centrifuged as above (×3) to remove the excess FITC-lipase. The pellet was suspended in deionized water (1 mL) to afford a concentration of 0.5 mg_FITC-lipase mL^–1 with an assumption of 100% FITC-lipase immobilization in PCN-333. The complete immobilization of FITC-lipase was supported by the absence of fluorescence signal (from the FITC tag in the FITC-lipase) in the supernatant from the immobilization mixture. The suspension was stored at 4 °C, protected from light, until required.

Characterization

Gas Sorption

N₂ sorption isotherms were measured on an autosorb iQ (Quantachrome INSTRUMENTS Inc.). Approximately 50 mg of the powdered solid was loaded into a glass analysis tube and outgassed under dynamic vacuum at 60 °C for 12 h. N₂ adsorption and desorption isotherms were measured at 77 K, and data were analyzed using Brunauer–Emmett–Teller (BET) analysis models⁸⁰ to determine the surface area. The pore size distribution calculations were carried out using the ASiQwin software DFT package (Quantachrome INSTRUMENTS Inc.).

Powder X-ray Diffraction (PXRD)

PXRD measurements were performed on a PANalytical XPert Pro powder diffractometer fitted with a solid-state PIXcel detector (45 kV, 40 mA, 1° divergence and antiscatter slits, and 0.3 mm receiver and detector slits) using Cu Kα (λ = 1.5406 Å) radiation.

Catalytic Assay of FITC-lipase and FITC-lipase@PCN-333

The catalytic activity of the free FITC-lipase or FITC-lipase@PCN-333 was measured according to a modified protocol from Sigma-Aldrich. UV–vis measurements were recorded on a Shimadzu UV-3600i Plus UV–vis–NIR spectrophotometer between the wavelengths of 600 and 300 nm. The enzymatic activity of free FITC-lipase or FITC-lipase@PCN-333 was calculated from the change of absorbance at 400 nm. In a specific experiment with free FITC-lipase, an aqueous solution of FITC-lipase (5 μL, 5 mg_FITC-lipase mL^–1) and a solution of 4-nitrophenyl butyrate in acetonitrile (10 μL, 50 mM, 0.5 mmol) were added sequentially into an aqueous Tris-HCl buffer solution (985 μL, 50 mM, pH 7.4) at room temperature. The reaction was left to stand at room temperature for the allocated reaction time before being transferred to a 1 mL cuvette for UV–vis measurement. The absorbance at 400 nm was recorded and used to measure the enzymatic activity of the materials.

For experiments with FITC-lipase@PCN-333, a suspension of FITC-lipase@PCN-333 in deionized water (50 μL, 0.5 mg_FITC-lipase mL^–1) was added to a solution of 4-nitrophenyl butyrate in acetonitrile (10 μL, 50 mM, 0.5 mmol) in aqueous Tris-HCl buffer solution (940 μL, 50 mM, pH 7.4) at room temperature. The reaction mixture was left to stand at room temperature before being centrifuged (13 500 RPM, 30 s) before the allocated time. The supernatant was transferred into a 1 mL cuvette, and the absorbance at 400 nm was measured.

Recycling Catalytic Activity of FITC-lipase@PCN-333

Recycling experiments of FITC-lipase@PCN-333 were completed as per the experimental protocol of the catalytic activity of FITC-lipase@PCN-333 mentioned above. Absorbance at 400 nm was collected after a reaction time of 2 min. Between each run, the FITC-lipase@PCN-333 pellet was resuspended in deionized water, centrifuged (13 500 RPM, 30 s) and the supernatant removed (× 2). The next cycle was started upon the addition of a mixture of the 4-nitrophenyl butyrate and Tris-HCl buffer solutions cited above.

Solid-State Magic Angle Spinning NMR Experiments

Prior to the ¹H, ¹³C, and ²⁷Al MAS NMR experiments, all samples were fully hydrated by exposure to the atmosphere for a week. The single pulse MAS NMR experiments for ¹H, ¹³C, and ²⁷Al were performed on a Bruker Avance III 500WB spectrometer at a resonance frequency of 400.1 MHz using a 4 mm MAS rotor. For ¹H single-pulse MAS NMR, spinning rate of 12 kHz, 64 scans were recorded for the spectra after single-pulse π/2 excitation with repetition times of 20 s. ¹³C and ²⁷Al MAS NMR investigations were conducted at a resonance frequency of 125.7 and 130.3 MHz, respectively. Noteworthy, a spin rate of 10 kHz has been applied for ¹³C MAS NMR experiments with the purpose of avoiding the overlap of the spinning sidebands from the peaks, while the 12 kHz spinning rate was applied for ²⁷Al MAS NMR measurements. Five thousand scans were recorded to obtain ¹³C CP MAS NMR spectra with a contact time of 4 ms and a repetition time of 4 s. Single-pulse ²⁷Al MAS NMR spectra were recorded after a single pulse π/6 excitation with a repetition time of 0.5 s. In order to obtain readable ²⁷Al MAS NMR spectra, 200 scans have been recorded. The two-dimensional (2D) ²⁷Al MQ spectra were acquired using the three-pulse z-filter MQ MAS pulse sequence.⁸¹ The spectra were sheared with TopSpin2.1 xfshear program. The repetition time for the 2D ²⁷Al MQ experiments was 1 s, and 5000 scans were recorded. Parameters of isotropic chemical shift (δ_iso), quadrupolar constant (C_QCC), and asymmetric parameter (η) have been obtained by applying eqs 1–4 for the 2D MQ MAS spectra⁷⁷ with the purpose to obtain information for the decomposition of ²⁷Al MAS NMR spectra:

1

2

3

4

The values of I (nuclear spin of the quadrupolar nucleus) and V_o (Larmor frequency of the quadrupolar nucleus) were 5/2 and 104.3 MHz, respectively. SOQE represents the second order quadrupolar effect parameter, and δ_F1 and δ_F2 are the centers of gravity of the MQ MAS signal in the F₁ and F₂ dimension.

Spin–lattice relaxation times (T₁) were determined via ¹H and ²⁷Al detection applying inversion recovery and saturation recovery, with a list of t₁ delay of 10, 5, 4, 3, 2, 1, 0.5, 0.25, 0.1, and 0.01 s for ¹H T₁ value measurements and 120, 100, 80, 40, 20, 10, 5, 2, 1, 0.5, 0.2, 0.1, 0.05, and 0.02 ms for ²⁷Al T₁ value measurements. In the ¹H T₁ experiment, a 180° pulse was applied first to get a fully inverted M_z. A fully recovered equilibrium magnetization (M_z > 99%) can be achieved after a long recycle delay (D₁ = 50 s). Another 90° pulse was then applied to flip the magnetization. For the ²⁷Al T₁ measurement, a 90° pulse was first applied to saturate the system; before the second 90° pulse, a 2 s recycling delay was applied for the magnetization recovery. The corresponding spectra were recorded as a 2D data where the t₁ delay was varied (t₁ delay list). The ¹H T₁ and ²⁷Al T₁ spectral deconvolutions were applied to obtain readable information using the TopSpin and DMFit software.⁸² By application of eq 5, the corresponding T₁ values were obtained. In eq 5, M_t represents the original magnetization length, t is the applied t₁ delay, and M represents the magnetization recovery after a specific t₁ delay.⁸²

5

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.1c00226.

• Full analytic details, including SEM images, CLSM images, gas sorption isotherms, and ssNMR data (PDF)

Author Contributions

All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.

Dedication

Dedicated to Professor Jianzhong Chen on the occasion of his 70th birthday.