Metal–Organic Frameworks for Drug Delivery: A Design Perspective

Abstract

The use of metal–organic frameworks (MOFs) in biomedical applications has greatly expanded over the past decade due to the precision tunability, high surface areas, and high loading capacities of MOFs. Specifically, MOFs are being explored for a wide variety of drug delivery applications. Initially, MOFs were used for delivery of small-molecule pharmaceuticals; however, more recent work has focused on macromolecular cargos, such as proteins and nucleic acids. Here, we review the historical application of MOFs for drug delivery, with a specific focus on the available options for designing MOFs for specific drug delivery applications. These options include choices of MOF structure, synthetic method, and drug loading. Further considerations include tuning, modifications, biocompatibility, cellular targeting, and uptake. Altogether, this Review aims to guide MOF design for novel biomedical applications.

Graphical Abstract

INTRODUCTION

Metal–organic frameworks (MOFs), also known as porous coordination polymers, have been studied extensively for gas storage, catalysis, sensors, and membranes because of their high surface area/volume ratios and porosities.¹ More recently, MOFs have also been investigated for drug delivery, beginning with loading and controlled release of ibuprofen.^2,3 Subsequent applications of MOFs to drug delivery have still principally focused on the release of small molecule drugs, such as the antitumor agents doxorubicin and curcumin.^4–8 A technique for encapsulating and protecting macromolecules using MOFs, termed biomimetic mineralization, has been used to generate MOFs loaded with a variety of macromolecular therapeutics (e.g., gelonin, Cas9-loaded with sgRNA for CRISPR, plasmids, and siRNAs).^9–13 Using MOFs to deliver therapeutics of all classes, including small molecules, gasotransmitters, proteins, nucleic acids, viruses, and cells is a growing area of investigation.^10,14–18

General reviews describing MOFs and their development are widely available.^19,20 Here, we will focus on the application of MOFs for drug delivery. Specifically, our goal is to describe methods and techniques used to generate and characterize MOFs for drug delivery applications. Syntheses and modifications of MOFs that are useful for enhancing their utility for drug delivery are explored. A variety of drug loading techniques specific to MOFs will be discussed. Cell culture and in vivo evaluation of MOF-drug formulations will also be highlighted to demonstrate progress to date in translating MOFs to clinical practice.

STRUCTURES AND COMPOSITIONS OF MOFS

The structures of MOFs can be described on four levels (Figure 1). The first level is the chemical constituents used to construct the MOF, that is, a metal ion (node) and a coordinating ligand (linker). Multivalent metal ions are most commonly used; however, monovalent ions have also been used.²¹ Zirconium(IV), iron(III), and zinc(II) are the most prevalent ions used in MOFs intended for drug delivery applications (Table S1). The ligands used in MOF synthesis usually have multiple carboxyl or amine functional groups that extend from either an alkyl chain or a ring-based structure like benzene or imidazole. Coordination of the ligand with the ion results in a crystal-like lattice with regular repeating geometry. While most MOFs have rigid structures, others are known to demonstrate some structural flexibility.^22–24

Figure 1.

Levels of structure and composition for MOFs: level 1, node and linker; level 2, secondary building unit (SBU) and coordinatively unsaturated site (CUS); level 3, inner-framework structure; and level 4, morphology. Molecular models created with permission from ChemTube3D, images are adapted (http://www.ChemTube3D.com).

The second level of structure is referred to as the secondary building unit (SBU), which is the coordination site of multiple ligands with a metal ion into a relatively rigid geometry.²⁰ SBUs essentially serve as the template or unit cell for the growth of the MOF structure. The linking of multiple SBUs by bridging ligands, ligands linking two metal nodes, defines the internal framework, the third level of MOF structure. This third level encompasses the pores and cages (i.e., the void volume) of the MOF. The pore structure can generally be determined a priori, given a particular metal ion and ligand.²⁰

While the first three structural levels of MOFs are essentially predetermined by the coordinating metal and ligand,²⁵ the outer morphology (size, shape, orientation), the fourth structural level of a MOF, depends on how the internal framework grows. The synthesis procedures used and whether molecules (e.g., therapeutics) are being encapsulated during synthesis will affect the outer morphology of the MOF.^7,9 Also, MOFs contain coordinatively unsaturated metal sites (CUSs) that can act like Lewis acids and aid in loading molecules onto the surface and functionalizing the MOF.^6,25,26 The fine, multilevel control of the chemical and structural features of MOFs makes them highly desirable for use in drug delivery applications.

ADVANTAGES FOR DRUG DELIVERY

Though not originally developed with drug delivery in mind, MOFs have demonstrated their utility for drug delivery based on precise control over their size, structure, and pore dimensions; straightforward surface functionalization; high drug loading capacities; controlled release of therapeutics in biological environments; synergistic/dual drug loading/release; and protection/stabilization of biomolecular therapeutics. Synthetic methods can be altered to create nanosized MOFs or adjust the pore dimensions of the MOF to improve loading or control release.^27–30 Additionally, modifications can be made synthetically or post synthetically that further improve loading, targeting, and the stability of MOFs in biological environments.^{10,29,31–33}

With some of the most highly porous structures and largest surface areas reported for delivery vehicles,³⁴ MOFs can load substantially more drug (e.g., Figure 2; insulin–mesoporous silica³⁶ (26.1 wt %) vs MOF^11,35 (35–39.7 wt %)), resulting in high local concentrations of drug when delivered from MOFs.³⁷ Controlled release from MOFs has also been demonstrated.^38,39 Release from some frameworks is inherently triggered by stimuli (pH, ATP, UV light, etc.),^6,40–43 and other frameworks are easily modified with moieties that control release of therapeutics from the pores.^44–49 These stimuli triggered formulations have been reviewed.⁵⁰ MOF pores can also be designed to slow or accelerate the diffusion of the therapeutic cargo.^4,22 The rate of MOF degradation in biological environments and, hence, the rate of drug release can also be manipulated by choosing alternative chemical constituents.^4,22 Depending on the loading method and MOF composition, synergistic therapeutic effects can be achieved either by loading multiple therapeutics in a single MOF or by the release of the metal or ligand from the MOF along with the therapeutic. This synergism has been proposed for cancer, vaccines, tendon healing, wound healing, and osteopathic applications.^{14,40,51–56}

Figure 2.

Surface area,³⁴ pore volume,³⁴ and insulin loading^11,35 of MOFs in comparison to mesoporous silica.³⁶ The large surface areas and pore volumes allow the higher drug loading achieved by MOFs.

MOFs can encapsulate a wide variety of small molecule and macromolecular therapeutics. They are particularly useful for encapsulating poorly water-soluble cancer therapeutics (e.g., curcumin) and gasotransmitter molecules (e.g., nitric oxide (NO), carbon monoxide (CO), and hydrogen sulfide (H₂S)).^6,18,55–57 MOFs have also been shown to encapsulate macromolecular cargos (e.g., nucleic acids and proteins), stabilize their structures, and protect them from degradation in biological environments.^9,28 Even larger cargos, like vaccines and cells have been explored (see Case Studies below).^16,17,58

SYNTHESIS AND CHARACTERIZATION

The available chemical and physical properties space for MOFs is extensive, due to the number of possible ions and ligands (Table S2). Applying reticular chemistry (structure-guided synthesis) enables the design of MOF frameworks to suit the needs of a particular application and precludes exhaustive empirical investigations.²⁰ Subsequent modifications can be made to the ligand while maintaining the overall geometry of the MOF framework (aka isoreticular chemistry). These modifications can be used to enact incremental changes in pore size, void volume, and surface area.^28,59–64

Controlling the structure and chemistry of the framework also controls the storage and release of the therapeutic molecules.^{22,26–29,59} In numerous cases, this control was gained by altering the charge of the framework.^26,59,65 Multiple isoreticular MOF series have identified MOFs that are stable in biological fluids and have enhanced interactions with biomacromolecules.³³ Functionalizing the ligand with an azide allowed for click chemistry,⁶⁶ which was used in multiple studies to create stimuli responsive MOFs gated by DNA aptamers (oligonucleotides that bind a specific target through structural complementarity).^44–49

In addition to these molecular-scale design options, the macroscopic size, morphology, and internal framework of the MOF can be controlled. Macroscopic order is achieved through controlled syntheses that use reaction kinetics, reagent/solvent ratios, equilibrium, temperature, and modulating agents to achieve the desired size and shape.^{33,55,67–69} Modulating agents, which are chemically similar to the MOF ligand, can act as competing ligands in the synthesis or deprotonating reagents that alter the nucleation and growth of the MOF.^29,68 Furthermore, modulating agents can introduce defects into the MOF structure to achieve larger pore sizes or additional void space.^{7,29,65,70–73}

A variety of synthesis methods allows for flexibility in MOF fabrication (Figure 3). Synthesis methods vary based on the MOF, with some MOFs synthesized by multiple methods⁷⁴ (Table S2). Solvothermal and nonsolvothermal syntheses are commonly used for MOFs. Solvothermal synthesis is generally carried out above the boiling point or at high pressures to dissolve the reactants and promote synthesis. Nonsolvothermal synthesis is carried out below the boiling point of the solvent and is generally carried out in reaction conditions that favor nucleation. Nontraditional synthesis methods used for drug delivery applications include microwaving,⁷⁵ sonicating,^29,79 or mechanical grinding.^76–78 Many of these methods use organic solvents, which can remain in the MOFs after synthesis and which can be subsequently removed by a process termed “activation”.^74,80 To avoid solvent issues for biological applications, synthesis methods have been developed in aqueous conditions and using light alcohol solvents.^{22,33,41,67,69} Also, solvent-free, mechanical synthesis methods have been investigated.^76,78

Figure 3.

Methods of MOF synthesis. Solvothermal methods involve high temperatures and pressures; nonsolvothermal methods do not.⁷⁴ Microwaves can assist MOF syntheses.⁷⁵ Mechanochemical methods use ball milling as a means of synthesis.^76–78 Sonochemical methods use high-energy sonication to aid MOF formation.^29,79

After synthesis, MOFs are characterized by a variety of approaches. For example, a curcumin and zinc-based MOF, medi-MOF-1, was analyzed by powder X-ray diffraction (PXRD) to confirm its crystalline structure before and after exposure to different solvents.⁵² As measured by thermogravimetric analysis (TGA), medi-MOF-1 maintained structural stability up to 300 °C. Fourier transform IR (FTIR) was used to confirm bonding and, later, ibuprofen loading.⁵² Nitrogen adsorption was used to determine the porous structure characteristics of medi-MOF-1, such as Brunauer–Emmett–Teller (BET) surface area (3002 m²/g) and pore volume (0.902 cm³/g).⁵² Molecular modeling was used to reveal the coordination scheme, confirm the SBU, pore diameters (9.2–11 Å), free volume (1.51 cm³/g), and simulated X-ray diffraction pattern of medi-MOF-1, while light-based microscopy was used to determine its size (~80–100 μm).⁵² Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) are also commonly used to visualize the macroscopic morphology of MOF particles, depending on their size.^29,52

TUNING MOF CHARACTERISTICS FOR DRUG DELIVERY

Postsynthetic functionalization and modification of MOFs can be used to improve drug loading and release kinetics and control their behaviors in biological environments. MOFs can be modified to improve loading by adding moieties that provide favorable interactions for loading. For instance, cationic polymers, like polyethylenimine (PEI) and ethanolamine conjugated poly(glycidyl methacrylate) (PGME(EA)), have been added to the synthesis or attached postsynthetically to increase the loading of polyanionic nucleic acids.^15,81 Postsynthetic functionalization of the ligand to introduce a positive charge improved loading of the anionic anti-inflammatory drug diclofenac.⁵⁹ Similar ligand modifications have allowed for gasotransmitter storage.^55,56

In addition to modifications that improve loading, modifications can be made to control release. Biomineralization of Ca₃(PO₄)₂ on the surface of a MOF limited the release of CpG oligonucleotides (immunostimulatory DNA sequences containing repeats of cytosine (C) followed by guanine (G)). When exposed to acidic environments, the calcium phosphate coating degraded and phosphate ions were liberated, triggering the release of CpG oligonucleotides from the surface of the MOF and inducing an immune response.⁶⁵ DNA-aptamer gated MOFs were created by using click chemistry to attach DNA to the surface of a MOF.^44–49 The DNA-aptamer blocked the release of drugs from the MOF until triggered by binding to its target. Similarly, addition of an amino group to a MOF ligand allowed attachment of a cytosine rich DNA, enabling the pH-triggered release of rhodamine.⁸² A reactive oxygen species (ROS) responsive polymer coating provided light responsive drug release and stabilization in biological environments.³⁷ MOF entrapment in a hydrophobic polymer scaffold permitted the prolonged release of NO.^55,56

Coatings, ranging from polymers and biomolecules to entire cell membranes,^{8,10,31,54,75} have been added to MOFs to improve targeting, colloidal stability, and biological half-life. Platelet membrane coated MOFs selectively delivered siRNAs to cancer cells in an in vivo mouse model.³¹ MDA-MB-231 extracellular vesicle (EV) membrane-coated MOFs showed homotypic uptake targeting to MDA-MB-231 cells and also were preferentially endocytosed by tumor cells in vivo.¹⁰ In both cases, membrane coatings also improved the colloidal stability and solubility of the MOF. Coating with CpG oligonucleotides stimulated an immune response in vivo and improved solubility.⁸³ Coating with albumin improved biorecognition, colloidal stability, and cell adsorption while reducing toxicity.⁸⁴ Chitosan and polyvinylpyrrolidone (PVP) coatings increased colloidal stability and biocompatibility.^32,79 Similarly, a capping of cetyltrimethylammonium bromide (CTAB), a surfactant, was used to maintain dispersity in suspension.⁸⁵ Methoxy poly(ethylene glycol)-block-poly(l-lactide) polymers were used to protect MOF-drug complexes for oral delivery of insulin from acidic degradation in the stomach (see Case Studies below).¹¹ MOF-polymer nanocomposites have been comprehensively reviewed elsewhere.⁸⁶

Mechanical and thermal modifications have also been applied in the manipulation of MOF structures. Ball milling and mechanical grinding have been used to decrease MOF size to take advantage of the enhanced permeability and retention (EPR) effect in cancer or to increase biocompatibility.^87,88 In one case, ball milling was used to create unsaturated Zn and N-sites on a MOF surface, which allowed for water binding and improved biocompatibility.⁸⁸ Mechanical grinding was used to control release kinetics by amorphizing the MOF and, potentially, blocking pore openings.^38,39 Similarly, thermal treatment, leading to partial pore collapse, was used to slow and extend drug release profiles.⁸⁹

DRUG LOADING, ENCAPSULATION, AND RELEASE

MOFs are attractive as drug delivery vehicles principally because of their exceptional drug loading capacity. Drug loading is governed by the physical properties of the MOF (i.e., pore size, surface area, void volume, and structural dynamics).² Variable sized pores and void volumes can be exploited for loading multiple therapeutics. Sequential loading in order of drug size (largest first) enables multiple drugs to be loaded into a single structure. For example, catalase (4.9 × 4.4 × 5.6 nm) was loaded into a MOF to fill 5.5 nm pores, followed by loading of superoxide dismutase (2.9 × 3.5 × 4.2 nm) to fill 4.2 nm pores. This structure with the combined antioxidant activity of the two enzymes effectively reduced ROS accumulation in stressed cells.⁷⁰

MOF pores may be rigid or more flexible, depending on the composition of the MOF and guest–host interactions.²² MOF pores exhibit breathing, swelling, ligand rotations, and subnetwork displacements.^23,90 MOF breathing, in which unit cell structures change upon binding to the loaded molecules, has been associated with loading and release of host molecules, like ibuprofen²⁷ and NO.²⁴ Swelling, where the unit cell expands while maintaining its shape, is also dependent on guest–host interactions and has been associated with release of ibuprofen.^90,91 Ligand rotation occurs around the metal coordination centers, which allows for expansion of the pore opening.⁹² Subnetwork displacements can occur when connecting forces are relatively weak and allow the relocation, drifting, and shifting of MOF components.²³ To make use of these dynamics, efforts have been made to understand which SBUs, ligands and metal ions allow dynamic flexibility in the structure of the MOF.²³ While ligand rotation and subnetwork displacements have not been explored in drug loading and release, they have in gas storage and may explain some release phenomena.⁹⁰

Chemical properties further govern the stabilization and loading of therapeutics on or within MOFs. Therapeutic molecules and MOFs can interact through van der Waals forces²⁸ or specific molecular features of the MOF, such as through Watson Crick base pairing with nucleic acid-based MOFs.⁸³ Aromatic structures found in many MOF ligands contribute to favorable π–π stacking interactions in loading molecules, such as 5-fluorouracil,⁹³ mitoxantrone,³³ rhodamine,⁷³ and doxorubicin.⁵ CUSs participate in surface coordination interactions with therapeutics, such as nucleic acids (phosphate-metal),^26,65,94,95 and small drug molecules, such as curcumin and doxorubicin.^5,6 CUSs are also critical in the loading and release of gasotransmitters; in some cases, CUSs can form chelating bridges with these gaseous therapeutics.⁹⁶ Electrostatic forces governed the loading of siRNAs and diclofenac.^26,59 Other small molecule drugs like oridonin leveraged hydrogen bonding to enhance loading.⁹⁷ As pH affects electrostatic interactions and hydrogen bonding, the pH during loading and the pH of the desired release environment should be considered when optimizing drug loading.

Loading method is critical for maximizing loading and achieving desired release profiles. There are four common methods for loading MOFs with drugs: one-pot synthesis, biomimetic mineralization, postsynthetic encapsulation, and surface loading (Figure 4). One-pot syntheses involve the coprecipitation of the therapeutic molecule with the MOF during synthesis. This leads to relatively uniform distribution of drug molecules throughout the mesopores of the MOF.⁴ To avoid degradation of the therapeutic by the solvents used in most one-pot approaches, a mechanical one-pot synthesis method was developed, though its application has been limited to small molecule therapeutics.^76,77 One-pot synthesis is convenient for protecting the therapeutic and usually degradation of the MOF controls release, provided the pore sizes are sufficiently small to limit rapid diffusion of the drug from the MOF structure.^83,85

Figure 4.

Loading and encapsulation methods: (a) one-pot synthesis,⁸ (b) biomimetic mineralization,⁹ (c) postsynthetic pore encapsulation,²⁸ and (d) surface loading.⁶⁵ Reproduced with permission from refs 8 [2018 The Royal Society of Chemistry], ref 9 [2015 Springer Nature], and ref 28 [2018 Springer Nature].

Biomimetic mineralization is useful for loading of biomolecular therapeutics like proteins and nucleic acids. Similar to one-pot synthesis, biomimetic mineralization combines biomolecules and MOF base units in one reaction mixture. Distinct from one-pot synthesis, biomimetic mineralization relies on the biomolecule as a nucleation site for MOF crystallization.^9,98 Specifically, biomolecular moieties form favorable bonds/interactions with MOF building units thus facilitating nucleation. The biomolecule being encapsulated, thereby, determines the size, morphology, and crystallinity of the MOF, while simultaneously being encapsulated in a MOF shell. This encapsulation mechanism has been shown to protect biomolecules from harsh chemical environments, heat, and degrading enzymes.⁹ Because of the integration of the therapeutic into the MOF structure, its release relies on the degradation of the MOF, which can result in “slow” release and delayed activity of the encapsulated therapeutic.^15,16,99,100

Postsynthetic encapsulation involves the loading of therapeutic molecules inside the pores of the MOF after synthesis. This is typically achieved by mixing the MOF and therapeutic in a solvent followed by removal of the solvent via evaporation.^11,28,101 Excess therapeutic is then washed from the surface. Alternatively, sonication and mechanical grinding have also been used for postsynthetic encapsulation.^33,37,87 For gasotransmitter loading, gravimetric adsorption is used.^24,96,102 Postsynthetic encapsulation generally results in diffusive release that can be accelerated by degradation of the MOF or by changes in environmental factors like pH.^22,27,52,59 Alternatively, MOFs with smaller pore sizes have demonstrated only burst release associated with their degradation (i.e., no diffusion through pores; rapid release due to MOF degradation over a short time period).^22,52

Surface loading is generally governed by CUS interactions and electrostatic interactions but can depend on other interactions.^{5,83,85,94,95,103,104} Surface loading can also be achieved by linking the therapeutic to the surface of a polymer coating.¹⁰⁴ Surface loading often results in reduced drug loading compared to other methods and rapid release of the drug from the MOF.^4,5,8,81,85 For example, surface loading led to 4.9 wt % doxorubicin loading⁵ versus 14–20 wt % for one-pot loading⁴ in ZIF-8 (zeolitic-imidazole framework-8 composed of Zn²⁺ and 2-methlyimidazole). Interestingly, in this case, surface loading was not accompanied by rapid drug release, as doxorubicin complexes strongly and preferentially to the surface of ZIF-8. Surface loading is particularly useful for loading additional therapeutics onto a MOF already loaded with encapsulated therapeutic.⁹⁵ In general, regardless of the encapsulation approach, MOFs protect biomolecules from degradation and expand the potential routes that could be used for clinical administration. That said, proteins/enzymes may change conformation when adsorbed to the MOFs surface.¹⁰⁵

Drug loading is often confirmed using TGA.^{5,6,28,33,52,87} Protection of the drug from degradation can also be assessed by TGA.^6,106 Reduction in the available surface area and pore volume (e.g., as measured by nitrogen adsorption–desorption) has been used to confirm loading.^27,95 TEM can be used to confirm mesopores formed by incorporating therapeutics in the structure of the MOF or morphology/size changes.^4,65 Fluorescence measurements have also been used to confirm loading for calcein and biomolecules with fluorescent tags.^28,29,35,75 For surface loaded therapeutics, measuring the zeta potential and size of the MOF after loading serves as a way to determine the presence of the drug.^65,107 FTIR, NMR, and PXRD are also applied to confirm the fidelity of MOF synthesis and drug loading/encapsulation simultaneously (Figure 5a and 5b).^6,9,59,108 To confirm drug activity after loading, principally for macromolecular therapeutics, functional assays, such as enzyme-linked immunosorbent assays (ELISAs),^16,35 enzymatic assays,^9,70 and transfections,⁹⁹ have been performed after release of the therapeutic from the MOF (Figure 5c).

Figure 5.

Methods to confirm drug loading and function. (a) Methods used to characterize Ni-IRMOF-74-II loading with ssDNA.²⁸ (b) Methods used to characterize isMOF (UiO-66 surface loaded with CpGs and coated with Ca₃(PO₄)₂.⁶⁵ (c) Functional assays to confirm macromolecular therapeutic function after loading and release from MOFs. Reproduced with permission from ref 28 [2018 Springer Nature].

ssDNA loading into Ni-IRMOF-74-x was confirmed by shifts in the TGA curves relative to the unloaded Ni-IRMOF-74 MOF (Figure 5a).²⁸ Nitrogen adsorption analysis showed that pore volume had decreased from 0.78–0.45 cm³/g because of the ssDNA loading. The appearance of an FTIR peak at 1050 cm⁻¹ confirmed ssDNA incorporation, a conclusion reinforced by quenching of the fluorescence probe on the ssDNA. Lastly, PXRD analysis showed additional peaks appearing after ssDNA loading.²⁸ Dynamic light scattering and zeta potential were used to measure loading of CpGs (negatively charged) onto isMOF (Figure 5b).⁶⁵ Morphology changes and size changes, determined by TEM, further confirmed surface loading of CpGs. The appearance of peaks related to DNA in the NMR spectrum further confirmed CpG loading.⁶⁵

BIOCOMPATIBILITY, CELLULAR TARGETING, AND UPTAKE

To be useful as clinical drug delivery vehicles, MOFs have to be biocompatible with limited cytotoxicity and immunogenicity. The cytotoxicity of MOFs is complicated by the fact that any of the components can be toxic alone but not when structured in the MOF and vice versa.^10,89,109 Moreover, cytotoxicity has also been shown to be cell type dependent.¹¹⁰ The metal ions and ligand alone need to be biocompatible as each can leach into the biological fluid/tissue over time.^33,35,111 In fact, toxicity of the same MOF depended on the metal ion (Fe vs Cr).¹¹⁰ The most widely used MOF in biological applications, ZIF-8 (Table S1), was evaluated in laboratory buffers for structural changes and leakiness.¹⁰⁰ Structural changes were common with most buffers, but leakiness was only an issue with cell medium, serum, and 0.1 M bicarbonate solutions. In cytotoxicity studies, ZIF-8 was well tolerated up to 100 μg/mL in cell culture,¹⁰⁹ with similar biocompatibility seen for the related MOF, ZIF-90.⁴¹ MIL(Fe) MOFs were tolerated up to 2 mg/mL in cell culture.¹⁰⁹ The difference in toxicity between ZIF and MIL(Fe) MOFs (Table S3) was hypothesized to be due to Zn²⁺ competition with Fe²⁺ and Ca²⁺ for ion channels and DNA damage caused by excess Zn²⁺.¹⁰⁹ Zirconium based MOFs showed a broad range of biocompatibilities (50 μg/mL−1.6 mg/mL).^37,89,109 A gadolinium-based MOF was biocompatible to 300 μg/mL in a mouse model; minimal Gd³⁺ leaching was reported in mimicked biological fluid (450 parts per billion),⁸⁷ mitigating concerns about Gd³⁺ toxicity (associated with cell iron transport), which can occur at ppm concentrations.^112,113

In general, immune stimulation from drugs is undesired as it leads to increased clearance rates and inflammation.^73,114 Mixed results have been reported in regard to the immunogenicity of MOFs. Some studies find uncoated MOFs initiate no significant immune response compared to untreated cells^73,115 while other studies showed that coated MOF systems (by EVs,¹⁰ by chitosan,¹¹⁵ or by heparin^103,116) were less immunostimulatory than uncoated MOFs. Note, these findings depended on the immune response inspected (e.g., macrophage internalization and expression,^{10,73,103,116} Th1 immune response,^{104,107,115,116} or complement system response^115,116). For some therapeutic applications (e.g., subunit vaccines), immunogenicity of the delivery vehicle is an advantage.^40,65,83,104 In subunit vaccine applications, CpG coatings are commonly included for immune activation.^40,65,83,104 Continued study of immunogenicity will be necessary in the development of clinical MOF formulations.

For clinical application, MOFs need to maintain their structure, dispersion, and chemical functionality in biological fluids, which has been achieved by some MOFs^{29,30,33,35,117} but not others.^79,88 Ligand functional groups and surface charge (as measured by zeta potential) can determine colloidal stability of the MOFs in biological fluids.^26,33 Formation of a protein corona improved the colloidal stability of MOFs^26,33,38,84 and their cellular uptake.⁸⁴ Membrane coatings also improved colloidal stability.^10,31,75 PVP, chitosan, polyethylene glycol (PEG), hydrogels, and 1,2-dioleoylsn-glycero-3-phosphocholine (DOPC) have all been used as coatings to prevent aggregation.^{11,32,37,47,79,110} Even nucleic acids loaded onto the surface of a MOF have been shown to increase colloidal stability.^51,83

Limiting systemic toxicity is an important consideration for drug development. One approach to targeting MOF-based drugs to only the diseased cells is through surface functionalization with a targeting moiety or coating. Functionalization of a MOF surface can be achieved readily by reacting primary amines found on MOF ligands.¹¹⁸ MOFs have been functionalized with folic acid,^119,120 HER2/Neu antibody,¹²¹ RGD-peptide,¹²² hyaluronic acid,^8,123 and AS1411 aptamer^48,49,124 or coated with cellular membrane materials^10,31 for targeting cancer cells. A MOF was functionalized with MK6240, a tau positron emission tomography tracer, to target neurofibrillary tangles for the diagnosis and treatment of Alzheimer’s disease.¹²⁵ Novel passive targeting strategies have also been used. Temporal, in vivo aggregation of MIL-100 allowed for passive targeting of lungs.¹²⁶ Successful targeting methods will be important for reducing systemic toxicity and translation of MOFs to clinical applications.

In vivo studies have shown MOF delivery systems to be well tolerated with and without modifications. In many cases, in vivo biocompatibility was achieved despite observed cytotoxicity in cell culture,¹⁰ perhaps because of the formation of a protein corona in vivo.^33,84,111 ZIF-8 delivery systems without modifications showed no significant toxicity or pathology and achieved good therapeutic outcomes.^10,16 Additionally, Gd-pDBI, ZrDTBA, UiO-66, MIL-100 (Fe), HUSKT-1, and porphyrin based MOFs have all been tested in vivo without evidence of toxicity.^{6,37,73,87,119,126} Similarly, coated MOFs were also biocompatible when delivered subcutaneously, intravenously, or orally and also demonstrated desired therapeutic effects.^{10,11,16,37,40,55,56,83,120,122,124,125,127}

Because some therapeutics, like siRNAs, need to be internalized by cells to achieve a therapeutic effect, cellular uptake of MOF-drug complexes has been investigated.^{10,11,41,54,65,108} To date, few generalizations can be made about endocytosis of MOF-drug complexes. MOF endocytosis has been shown to be energy dependent.^54,128 Furthermore, MOF size, coatings, and surface charge affect the mechanism of endocytosis.^65,89,128 MOF size has a significant effect on uptake, with one study finding 90 nm to be optimal.²⁹ Surface modifications, such as addition of bioactive agents like CpG oligonucleotides or targeting moieties like triphenyl phosphonium ions, can improve intracellular accumulation.^65,108 NU-1000 (a Zr⁴⁺ and 1,3,6,8-tetrakis(p-benzoate) pyrene-based MOF) has been shown to primarily use caveolae mediated endocytosis in HeLa cells.⁸⁹ However, it is likely that the route of endocytosis will vary by cell type and influence the efficacy of the delivered cargo.¹²⁹

Endosomal escape is critical for function of delivered therapeutics.³⁰ As with uptake, endosomal escape of MOFs depends on the MOF chemistry, structure, and modifications. Endosomal escape has been attributed to the “proton sponge effect”,^10,12,31,32 where endosomal acidification results in swelling and rupture, and metal-ion-mediated disruption of phosphate groups in endosomal membranes.^51,95 Other MOF systems require the assistance of other molecules to aid their endosomal escape (e.g., NU-1000 based delivery of siRNA required NH₄CL and KALA peptide).^30,108

CASE STUDIES

This section showcases the unique capabilities MOFs offer for drug delivery. Specifically, the precision structural tuning of MOFs, the biomolecular and cellular protection MOFs offer, and their application to gasotransmitter release and catalytic nanomedicine will be discussed. Information on other MOFs and their delivery applications have been tabulated (Table S1).

Fine Control of MOF Pore Dimensions for Delivery of Specific Therapeutic Cargos.

The pore structures of MOFs can be finely tuned to accommodate specific therapeutic cargos as described by Peng and colleagues.²⁸ This group generated an isoreticular series of Ni-MOF-74 by extending its salicylic acid ligand with phenylene units, which created the tailored porosity and pore size needed to encapsulate and control the release of ssDNA (Figure 6). Specifically, Ni-IRMOF-74-II (pore size 2.2 nm) was found to precisely control ssDNA incorporation (governed by van der Waals interactions) and release (triggered by the presence of complementary DNA). Ni-IRMOF-74-II protected ssDNA from nucleases in 10% fetal bovine serum and achieved a loading of 6.9 wt % (as compared to commercial lipid reagents LipoGene 2000 (0.02%) and Neofect (0.1%)). This formulation matched or bettered commercial reagents for transfection of macrophages, breast cancer cells, and CD4+ T cells and B cells, while causing significantly less cell death. This study comprehensively showed the promise MOFs offer transfecting conventionally difficult to transfect cell lines and showcases the precision tuning of MOFs for precise control of loading and release of large biomolecules.²⁸

Figure 6.

MOF ligand expansion used to increase pore size and tailor the loading of ssDNA.²⁸ Reproduced with permission from ref 28 [2018 Springer Nature].

MOFs for Oral Delivery of Protein Therapeutics.

Oral administration of therapeutics remains the most straightforward approach for maximizing patient compliance. Barriers to oral protein delivery include the gastrointestinal environment (acidic and proteolytic enzymes) and low permeability of protein drugs across biological membranes in the intestines.¹³⁰ Two studies have investigated MOFs as a solution for oral delivery of protein therapeutics, specifically insulin. Chen et al. developed a Zr-based MOF (NU-1000) insulin delivery system (Figure 7a), which is acid resistant, has pores that favorably interact with insulin, and disassembles in the presence of phosphate ions (as in blood). NU-1000 was able to load 39.7 wt % insulin via postsynthetic pore encapsulation. Insulin loaded NU-1000 was stable in a simulated gastrointestinal environment and readily released insulin at physiological conditions (pH 7.0 in PBS).³⁵A polymer microsphere system was developed that successively encapsulated insulin and SDS in MIL-100¹¹ (Figure 7b). The methoxy poly(ethylene glycol)-block-poly(l-lactide) polymer coating protected the MOF from degrading in the gastrointestinal environment, and the SDS increased the permeability through the intestinal membrane. In monolayer, Caco-2 cell culture, the MIL-100 NP increased endocytosis of insulin and demonstrated good permeability of the monolayer. In a BALB/c mouse, type I diabetes model, this oral delivery system, at 50 IU/kg, reached a maximum plasma insulin level (~50 mIU/mL) at 4 h and remained elevated for 8 h. This system reduced glucose levels more slowly and for longer than subcutaneous injections. Furthermore, insulin accumulation in the liver suggested that insulin released from the MOF circulated through the portal veins to the liver and subsequently the cardiac tissue, closely mimicking endogenous insulin circulation patterns.¹¹

Figure 7.

MOF formulations for oral insulin delivery: (a) NU-1000 system³⁵ and (b) MIL-100(Fe) formulation.¹¹

MOFs for Long-Term, Ambient Storage of Viruses and Cells.

Recent work by the Gassensmith group has explored the enhanced stability and controlled delivery of a MOF-encapsulated tobacco mosaic virus (TMV). Interestingly, their controlled biomimetic mineralization process created a rod-shaped nanocoating over TMV^98,131 (Figure 8a). This coating was put under the stress of protein destabilizing agents (methanol, ethyl acetate, 6 M guanidinium chloride) and heat (100 °C) and was found to maintain the structure of TMV as determined by ELISA. The vaccine@ MOF formulation was further demonstrated to elicit an antibody response in Balb/C mice comparable to the naked TMV, with no apparent toxicity.¹⁶ This study demonstrates the possibility of MOF-based vaccine formulations that are stable without refrigeration, alleviating the significant cold-chain requirements of current vaccine formulations¹³² (including mRNA-based COVID-19 vaccine formulations).

Figure 8.

Biomimetic mineralization for the encapsulation and delivery of (a) TMV¹⁶ and (b) cytoprotective exoskeletons¹⁷ for living cells. (c) SupraCell construction for encapsulating living cells.⁵⁸ Reproduced with permission from refs 17 [2016 John Wiley and Sons] and 58 [2019 John Wiley and Sons].

An emerging application of MOFs involves their use in creating cytoprotective, diffusion controlling exoskeletons. A ZIF-8 coating on Saccharomyces cerevisiae (baker’s yeast) was shown by Liang and colleagues to control molecular trafficking to the cell and prevent division, inducing an artificial hibernation state¹⁷ (Figure 8b). Upon exfoliation of the ZIF-8 exoskeleton, the yeast regained full function. Furthermore, studies have encapsulated mammalian cells in MOF-based exoskeletons named SupraCells (Figure 8c). These coatings have been generated using ZIF-8, MIL-100, and UiO-66 MOFs and tannic acid. The exoskeleton coatings cause a quiescent cell state preventing replication or adherence to surfaces and conferring resistance to extreme environmental conditions. Exoskeleton coated cells were shown to resist osmotic stress, ROS, pH, and UV exposure and return to normal activities after exfoliation.⁵⁸ One clear application for these coated cells would be delivery of probiotic bacterial supplements for gut health.

MOF-Based Gasotransmitter Delivery.

NO, CO, and H₂S are gaseous signaling molecules that are endogenously generated, freely traverse cellular environments, and play important roles in normal physiological processes.¹³³ Because their balance in physiological processes is so delicate, their delivery as a therapeutic has been difficult.¹⁸ MOFs have been identified as excellent candidates for controlled gasotransmitter delivery,²² and applications have since been pursued and reviewed.^{18,55,56,96,102}

In one case, a copper MOF-based NO delivery system was developed to support tendon regeneration (Figure 9).⁵⁶ NO was encapsulated in the MOF (HKUST-1), which was further embedded in a hydrophobic polycaprolactone/gelatin scaffold to control water triggered NO release. This system sustained controlled release of NO over 15 days at 1.67 nM h⁻¹ while simultaneously releasing Cu²⁺ from the degrading MOF. The combined NO and Cu²⁺ release synergistically supported angiogenesis and collagen formation. This system supported tendon healing in vivo over a period of 70 days. A similar system was also applied to diabetic wound healing.⁵⁵ These examples notwithstanding, gasotransmitter delivery represents a small portion of the MOF-based therapeutic literature and will require considerably more study for translation to clinical application.¹⁸

Figure 9.

MOF-based nitric oxide (NO) delivery.⁵⁶ (a) HKUST-1(HK) after postsynthetic modification with 4-methylamino pyridine and loading of NO becomes NMHK. (b) HK is embedded in a polycaprolactone (PCL)/gelatin (gel) scaffold to generate NMPGA. (c) The scaffold material was applied to an in vivo tendon defect model and a cultured endothelial cell model. Reproduced with permission from ref 56 [2021 Springer Nature].

Catalytic MOF-Based Nanomedicine.

Catalytic nanomedicine is a unique and growing application of MOFs, as they are able to act as catalysts themselves or provide a scaffold for other catalytic species. Prior reviews have covered MOF-based catalytic nanomedicine applications in tumor therapies, bacterial disinfection, tissue regeneration, and biosensors.^134–136 Cancer-based applications represent the bulk of the catalytic nanomedicine literature. MOF-based catalytic nanomedicine therapies take advantage of the tumor microenvironment (acidosis, high-glucose conditions, hydrogen peroxide, lactate, and glutathione overproduction) to initiate a series of cytotoxic chemical reactions.^{54,119,127,136–141}

One zirconium and iron(III) meso-tetra(4-carboxyphenyl) porphine chloride (TCPP(Fe)) MOF was used as a scaffold for gold nanoparticles (AuNP) while also being loaded with the chemotherapeutic drug camptothecin (CPT). The loaded MOFs were then coated with PEG (top of Figure 10).⁵⁴ These nanoMOFs (nMOFs), when delivered intravenously in mice, exploited the EPR effect as a means of passively targeting a xenograft tumor. Once present in the tumor environment, the nMOF scaffold degraded quickly in the presence of phosphate ions, thereby releasing the TCPP(Fe) ligand, AuNPs, and CPT and initiating a catalytic cascade. The released AuNPs functioned as a glucose oxidase mimicking catalyst, breaking down glucose into gluconic acid and hydrogen peroxide. The TCPP(Fe) ligand reacted with the hydrogen peroxide in a Fenton type reaction generating cytotoxic hydroxyl radicals (Figure 10, bottom). nMOFs significantly inhibited tumor growth (85.6% inhibition of growth compared to an untreated control) over the course of 20 days.⁵⁴ MOF-based, catalytic nanomedicine is an emerging application of MOFs with great promise to reduce the cost of drug development, reduce systemic toxicity, and overcome drug resistance.¹³⁶

Figure 10.

Catalytic MOF-based cancer therapy: synthesis (top) and mechanism of action (bottom).⁵⁴ Reproduced with permission from ref 54 [2020 John Wiley and Sons].

CONCLUSIONS AND FUTURE PERSPECTIVES

MOFs hold great promise for the delivery of biomacromolecular and cellular therapeutics. It is straightforward to envision using MOFs as a component of a delivery system that specifically serves the purpose of loading and protecting biomolecules/cells. Then, postloading modifications, like polymeric nanoparticle encapsulation or membrane coating, would be used to formulate the final therapeutic. These types of composite vehicle approaches are beginning to show promise in vivo.^10,11 Finally, continuing studies of endocytosis and trafficking of MOF-based therapeutics will be required, especially for nucleic acid therapeutics, to ensure proper intracellular processing of the delivered cargo.

Going forward, it is critical that studies emphasize in vivo evaluation of MOF drug delivery vehicles, given the differences found in vitro versus in vivo. Likewise, much work exists using mock drug molecules. Future work should focus on actual drug compounds to accelerate progress toward clinical applications. Further proof-of-concept studies still need to be performed for vaccines and cell-based therapies, as delivery of these species remains understudied. Also, the potential synergism of MOFs with drugs needs to be further explored.^14,53,55,56 The development of MOFs for clinical applications will also require new disease targets to be explored, specifically beyond cancer, and potentially new targeting methods. The immunogenicity of MOFs likewise needs to be studied more comprehensively. The application of MOFs to drug delivery has resulted in research targeting development of therapeutics relying on catalytic nanomedicine⁵⁴ and gasotransmitters.^55,56 These studies should continue as MOFs are uniquely suited for these applications.

ABBREVIATIONS

Biological Test System

293T

human embryonic kidney

3T3

Swiss albino mouse embryo fibroblast

4T1

mouse mammary carcinoma

A2780

human ovarian carcinoma

A2780/CDDP

human ovarian carcinoma cisplatin resistant

A549

human adenocarcinoma alveolar basal epithelial cells

APC

antigen-presenting cell

B16-F10

murine skin melanoma

BT-474

human mammary ductal carcinoma

BXPC-3

human pancreatic carcinoma

CACO-2

human colorectal epithelial adenocarcinoma

CAD

murine catecholaminergic neuronal tumor

CHO

Chinese hamster ovary

COS7

monkey kidney fibroblast-like cell

DC2.4

murine dendritic cell

H460

large cell lung carcinoma

HACAT

human keratinocyte

HASMC

human aortic smooth muscle cells

HDF

human dermal fibroblasts

HEK-293

human embryonic kidney

HEKN

human epidermal keratinocytes neonatal

HELA

human cervical cancer adenocarcinoma

HEPG2

human hepatocellular carcinoma

HL-60

human promyelocytic leukemia

HL7702

human liver cell

HMSC

human mesenchymal stem cell

HT-29

human colorectal adenocarcinoma

HUVEC

human umbilical vein endothelial cell

J774.A1

murine monocyte macrophage

L02

human liver cell

L929

murine fibroblasts

MC3T3

murine osteoblast precursor

MCF-10A

human preneoplastic mammary epithelial cells

MCF-7

human breast adenocarcinoma

MCF-7/T

human breast adenocarcinoma taxol resistant

MDA-MB-231

human epithelial breast cancer

MDA-MB-468

human pleural effusion metastatic breast cancer

MGC-803

human gastric mucinous adenocarcinoma

MH-S

murine alveolar macrophages

NCI-H292

human lung carcinoma

NIH-3T3

murine embryonic fibroblasts

NOD/SCID MICE

nonobese diabetic/severe combined immunodeficiency mice

PBL

human peripheral blood lymphocytes

PBMC

human peripheral blood mononuclear cells

PC-12

rat pheochromocytoma

PC-3

human prostate adenocarcinoma

RAW264.7

murine macrophage

RBC

red blood cell

SH-SY5Y

human neuroblast

SK-BR-3

human breast adenocarcinoma

SKOV3

human ovarian adenocarcinoma

SMMC-7721

human hepatocellular carcinoma

SW480

human colorectal adenocarcinoma

T1D RAT

type 1 diabetes rat

THP-1

human monocyte (acute monocytic leukemia)

U937

human histiocytic lymphoma

U-87

human glioblastoma

Cargo

ALPHA-CHC

A-cyano-4-hydroxycinnamic acid

BETA-GAL

beta-galactose

BSA

bovine serum albumin

CPG

5′-C-phosphate-G-3′

CYT C

cytochrome C

GFP

green fluorescent protein

GMP

gemcitabine monophosphate

HRP

horse radish peroxidase

HSA

human serum albumin

MP-11

microperoxidase-11

OVA

ovalbumin

PQQ-GDH

glucose dehydrogenase

RAPTA-C

Ru(η⁶-p-cymene) Cl₂(PTA)

RNASE A-NBC

ribonuclease A ROS responsive modification

SBHA

suberohydroxamic acid

SOD

superoxide dismutase

VEGF

vascular endothelial growth factor

MOF

HUSKT

Hong Kong University of Science and Technology

IRMOF

isoreticular MOF

MIL

Matériaux de l’Institut Lavoisier

MIP

Materials of the Institute of Porous Materials of Paris

NCP

nanoscale coordination polymer

NU

Northwestern University

PCN

porous coordination network

UIO

Universitetet i Oslo

UMCM

University of Michigan Crystalline Material

ZIF

zeolitic imidazolate framework

ZJU

Zhejiang University

Viability Tests

CCK8

cell counting kit-8

MTT

3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide

MTS

3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium

SRB

sulforhodamine B

WST

water-soluble tetrazolium salts

LDH

lactate dehydrogenase

H&E

hematoxylin and eosin stain

TUNEL

transferase-mediated dUTP nick end-labeling

XTT

2,3-bis(2-methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5-carboxanilide