DNA-Functionalized Metal–Organic Framework Nanoparticles for Intracellular Delivery of Proteins

Abstract

Due to their large size, charged surfaces, and environmental sensitivity, proteins do not naturally cross cell-membranes in intact form and, therefore, are difficult to deliver for both diagnostic and therapeutic purposes. Based upon the observation that clustered oligonucleotides can naturally engage scavenger receptors that facilitate cellular transfection, nucleic acid–metal organic framework nanoparticle (MOF NP) conjugates have been designed and synthesized from NU-1000 and PCN-222/MOF-545, respectively, and phosphate-terminated oligonucleotides. They have been characterized structurally and with respect to their ability to enter mammalian cells. The MOFs act as protein hosts, and their densely functionalized, oligonucleotide-rich surfaces make them colloidally stable and ensure facile cellular entry. With insulin as a model protein, high loading and a 10-fold enhancement of cellular uptake (as compared to that of the native protein) were achieved. Importantly, this approach can be generalized to facilitate the delivery of a variety of proteins as biological probes or potential therapeutics.

Proteins play key roles in living systems, and the ability to deliver active proteins to cells is attractive for both diagnostic and therapeutic purposes.¹ Potential uses involve the evaluation of metabolic pathways,² regulation of cellular processes,³ and treatment of disease involving protein deficiencies.^4–6 During the past decade, a series of techniques have been developed to facilitate protein internalization by live cells, including the use of complementary transfection agents, nanocarriers,^7–9 and protein surface modifications.^10–13 Although each strategy has its own merit, none are perfect solutions; they can cause cytotoxicity, reduce protein activity, and suffer from low delivery payloads.¹⁴ For example, we have made the observation that one can take almost any protein and functionalize its surface with DNA to create entities that will naturally engage the cell-surface receptors involved in spherical nucleic acid (SNA) uptake.^13,15–17 While this method is extremely useful in certain situations, it requires direct modification of the protein and large amounts of nucleic acid, on a per-protein basis, to effect transfection. Ideally, one would like to deliver intact, functional proteins without the need to chemically modify them, and to do so in a nucleic-acid efficient manner.

Metal organic frameworks (MOFs) have emerged as a class of promising materials for the immobilization and storage of functional proteins.¹⁸ Their mesoporous structures allow for exceptionally high protein loadings, and their framework architectures can significantly improve the thermal and chemical stabilities of the encapsulated proteins.^19–24

However, although MOF NPs have been recognized as potentially important intracellular delivery vehicles for proteins,^25–27 their poor colloidal stability and positively charged surfaces,^28,29 inhibit their cellular uptake and have led to unfavorable bioavailabilities.^30–33 Therefore, the development of general approaches for reducing MOF NP aggregation, minimizing positive charge (which can cause cytotoxicity), and facilitating cellular uptake is desirable.^34,35

Herein, we report a new method for the intracellular delivery of proteins that relies on nucleic acid–MOF NP conjugates (Scheme 1A).^34–37 In this protocol, two water stable zirconium mesoporous MOFs, NU-1000 (Zr₆(μ₃O)₄(μ₃OH)₄(OH)₄(H₂O)₄(TBAPy)₂, H₄-TBAPy = tetraethyl 4,4′,4″,4″′-(pyrene-1,3,6,8-tetrayl)tetrabenzoic acid) and P C N - 2 2 2 / M O F - 5 45 ( Z r ₆ ( μ ₃ - O ) ₄ ( μ ₃ OH)₄(OH)₄(H₂O)₄(TCPP-H₂)₂, H₄-TCPP-H₂ = tetrakis(4carboxyphenyl)porphyrin),^38–40 were synthesized in nanoparticle form and used to encapsulate insulin, a model protein for the studies described herein (Scheme 1B).^41,42 Next, via modification of literature procedures, these insulin@MOF NPs were surface functionalized with terminal phosphate-modified DNA to yield insulin@DNA-MOF NPs (Scheme 1C).³⁶ The 3D oligonucleotide shell creates a steric and electrostatic barrier to stabilize MOF NPs in high dielectric media and renders them functional with respect to cellular entry.³⁴ In principle, this strategy can be generalized to MOFs with different pore sizes and topologies, thereby creating an arsenal of nucleic acid–MOF-based delivery vehicles for transporting functional enzymes across cellular membranes with high payloads.

Scheme 1.

(A) Schematic Illustration of Insulin Encapsulation in the Mesoporous Channels of MOF NPs Followed by DNA Surface Functionalization; (B) Crystal Structures of Two Mesoporous Zr MOFs: NU-1000 and PCN-222/MOF-545 and Their Respective Organic Linkers; (C) DNA Functionalization of Insulin Encapsulated MOF NPs Using 3′ Terminal Phosphate Modified Nucleic Acids

MOF NP syntheses^43,44 and insulin encapsulations⁴¹ were realized via literature protocols. Specifically, NU-1000 MOF NPs [180(20) × 70(10) nm] were synthesized via a solvothermal reaction of zirconium chloride (ZrCl₄) with H₄TBAPy ligands, modulated by acetic acid in N,N-dimethylformamide (DMF) at 90 °C (Figure 1A). Similarly, PCN-222 NPs [210(30) × 50(10) nm] were synthesized via a solvothermal reaction between zirconyl chloride octahydrate (ZrOCl₂·8H₂O) and H₄-TCPP-H₂ ligands, modulated by dichloroacetic acid in DMF at 130 °C (Figure 1B). Next, the thermally activated crystals of NU-1000 were treated with a bis-tris-propane buffer (BTP, pH = 7) solution of insulin (0.4 mg/mL). The MOF NP insulin encapsulation efficiencies were determined by measuring the S [for insulin] and Zr (for MOFs) contents by inductively coupled plasma-optical emission spectroscopy. The maximum insulin loadings of 34 and 63 wt % were determined for NU-1000 and PCN-222 NPs, respectively, which are consistent with our previous report.⁴¹ The excess insulin in the supernatant was removed by sequential washing steps with DI water.

Figure 1.

Scanning electron microscopy (left) and transmission electron microscopy (right) images of as-synthesized NU-1000 NPs (A) and PCN-222 NPs (B). (C–D) Colloidal stability of NU-1000 and PCN-222 NPs in cell medium, as determined by DLS without (left) and with DNA surface modification (right). Scale bars = 100 nm.

The insulin@MOF NPs were functionalized with nucleic acids by coordinating the terminal phosphate-modified oligonucleotides to the surface Zr SBUs.^36,45 The sequence used here, 5′ (dGGT)₁₀-phosphate 3′, was chosen because it is known with SNAs that a G-rich shell, relative to poly dT shells, facilitates higher cellular uptake.⁴⁶ In a typical NP functionalization experiment, excess oligonucleotides were added to a colloidal dispersion of MOF NPs and incubated for 4 h (Supporting Information). Particle DNA coverage was quantitively determined by measuring the P to Zr ratio by ICP-OES (8 ± 1 nmol/mg for NU-1000 NPs and 10 ± 1 nmol/mg for PCN-222 NPs). Powder X-ray diffraction (PXRD) and scanning electron microscopy (SEM) confirmed that the crystallinity and morphologies of the MOF NPs were maintained, post-DNA functionalization (Figures S2 and S3). Importantly, dynamic light scattering (DLS) verified that DNA surface functionalization significantly increases MOF NP colloidal stability in cellular media (90% DMEM buffer +10% fetal bovine serum) for at least 24 h; for comparison, unfunctionalized NU-1000 NPs aggregated in less than 1 h, hampering further in vitro use (Figure 1C,D, EM image of aggregated NPs: Figure S5).

In addition to colloidal stability, the intra- and extracellular stability of protein delivery vehicles in serum and serum free but biologically relevant matrices is important. Indeed, the ability to control degradation could be useful in the development of temporally controlled drug delivery applications. Under physiological conditions, intracellular fluid exhibits significantly higher inorganic phosphate concentration (5–10 mM) as compared to that of serum (∼1 mM).^47,48

Therefore, the degradation profiles of insulin@DNA-NU-1000 NPs and insulin@DNA-PCN-222 NPs were evaluated by exposing them to solutions designed to emulate both extracellular and intracellular conditions (Supporting Information). To simulate serum, MOF NPs were incubated with 90% DMEM buffer +10% blood serum (pH = 7.0) at 37 °C with gentle shaking (400 rpm), where less than 5% of degradation occurred within 12 h for both vehicles, and less than 20% within 96 h, suggesting DNA-MOF NPs exhibit excellent stability and may be compatible with blood (Figure 2C, dashed). In contrast, when the same MOF NPs were incubated in an intracellular medium simulant (1 × phosphate buffered saline, pH = 7.0) at 37 °C with gentle shaking, the particles degrade at much faster rates (Figure 2C, solid) due to the high phosphate content, which competitively binds to Zr clusters. Interestingly, DNA-PCN-222 NPs exhibit a faster degradation rate (half-life = 1 h) when compared to that of DNA-NU-1000 NPs (half-life = 40 h). Such degradation kinetics could be useful for in vivo purposes by providing a means to control the temporal release of proteins from particles, once inside cells.

Figure 2.

(A) Representative confocal fluorescence micrographs of 10 μm insulin@DNA-NU-1000 particles verified the colocalization of insulin (AF647 channel) and DNA (TAMRA channel). (B) Z-stack image of a single 10 μm insulin@DNA-NU-1000 crystal. (C) Degradation profiles of DNA-NU-1000 NPs and DNA-PCN-222 NPs incubated in extracellular medium (dashed lines) and in simulated intracellular medium (solid lines) at 37 °C with 400 rpm shaking.

To directly visualize nucleic acid-modified, insulin encapsulated MOF NPs, we employed confocal laser scanning microscopy to image them. Due to the resolution limits of confocal microscopy, larger particles (2.8 μm × 10 μm for NU1000), AlexaFluor 647 dye (AF647)-labeled insulin, and TAMRA-labeled DNA were used. With such particles, the colocalization of AF647 and TAMRA signals can be clearly observed, verifying the encapsulation of insulin and DNA surface functionalization of the MOF (Figure 2A). To obtain detailed information regarding relative distribution of insulin and DNA, Z-stack images of a single MOF particle were taken, where TAMRA signal (DNA) was observed to preferentially occupy the periphery while AF647 (insulin) was present throughout the particle (Figure 2B and Figure S7). Brighter AF647 signals were observed at both ends of the particle as compared to the center section of the MOF, consistent with the previous observation that proteins diffuse into NU-1000 through its 1D channels.⁴² Due to the large diameter of the MOF pores (3.2 nm for NU-1000 and 3.7 nm for PCN-222),⁴⁰ single stranded DNA was also expected to penetrate through the MOF pores and functionalize the internal surface, leading to fluorescence signal inside the particles. As verified by N₂ adsorption isotherms, reduced N₂ uptake capacity was observed postinsulin encapsulation for both MOFs, and further loss of porosity was observed post-DNA functionalization (Figures S9 and S10). Furthermore, an enzyme-linked immunosorbent assay (ELISA) was employed to determine whether insulin would leach from the MOF NP pores and/or lose catalytic activity during the DNA functionalization process. In both cases, no appreciable leaching and/or insulin activity loss was observed for insulin@DNA-NU-1000 and insulin@DNA-PCN-222 constructs (Figure S8).

As previously stated, a key characteristic of SNA-NP conjugates is their ability to effectively enter cells. Therefore, we tested whether insulin@DNA-MOF NPs exhibited enhanced cellular uptake. Specifically, NU-1000 and PCN-222 NPs were encapsulated with AF647-labeled insulin and functionalized with TAMRA-labeled DNA and incubated with human ovarian adenocarcinoma cells, SKOV-3, for 0.5, 2, 6, and 24 h (Supporting Information). As a control group, a mixture of free TAMRA-labeled DNA and AF-647-labeled insulin was incubated with cells at the same concentration. Confocal laser scanning microscopy confirms the enrichment of insulin in cellular vesicles, as evidenced by strong colocalization of AF647 and TAMRA signals in cellular vesicles (Figures 3A–C). The Z-stack images confirm that the insulin@DNA-MOF NPs are internalized by the cells, as opposed to attached to their membranes. Consistent with this conclusion, flow cytometry showed a 10-fold increase in fluorescence in cells treated with insulin@DNA-MOF NPs as compared to those treated with the free insulin + DNA control group (Figure 3D). The insulin@DNA-MOF NPs exhibits similar levels of enhancement in cellular uptake, as compared to that of conventional SNA-NP conjugates.⁴⁹ Finally, MTT assays show that the particles result in no apparent cytotoxicity or antiproliferative effects (Figures 3E).

Figure 3.

(A–C) Flow cytometry plots and confocal fluorescence micrographs of SK-OV cells after treatment with free insulin + DNA (A), insulin@DNA-NU-1000 (B), and insulin@DNA-PCN-222 (C). (D) Cellular uptake of insulin delivered in different constructs as determined by flow cytometry. Fluorescence at 647 nm was measured in SK-OV cells after treatment with insulin at various incubation time (0.5 and 2 h). (E) MTT assay verifies no appreciable cytotoxicity induced by insulin@DNA-PCN-222 and insulin@DNA-NU-1000 NPs. Scale bar = 10 μm.

In conclusion, we have developed a facile strategy for using nucleic-acid modified MOF NPs to deliver proteins across cell membranes at high payloads and negligible cytotoxicity. This work is important since it highlights how clustered surface oligonucleotides on these modular materials can be used to make them colloidally stable in physiological environments and useful for intracellular biological applications. Future design iterations will allow for encapsulating various proteins by tuning the MOF pore sizes,^42,50,51 and potentially codelivery of protein and nucleic acid targets that are important for many purposes, including in vivo imaging,² gene regulation,³⁵ therapeutics,⁵ and the study of fundamental cellular processes.⁴