Engineered Interactions with Mesoporous Silica Facilitate Intracellular Delivery of Proteins and Gene Editing

Abstract

Intracellular delivery of functional proteins is a promising, but challenging, strategy for many therapeutic applications. Here, we report a new methodology that overcomes drawbacks of traditional mesoporous silica (MSi) particles for protein delivery. We hypothesize that engineering enhancement in interactions between proteins and delivery vehicles can facilitate efficient encapsulation and intracellular delivery. In this strategy, surface lysines in proteins were modified with a self-immolative linker containing a terminal boronic acid for stimulus-induced reversibility in functionalization. The boronic acid moiety serves to efficiently interact with amine-functionalized MSi through dative and electrostatic interactions. We show that proteins of different sizes and isoelectric points (pIs) can be quantitatively encapsulated into MSi, even at low protein concentrations. We also show that the proteins can be efficiently delivered into cells with retention of activity. Utility of this approach is further demonstrated with gene editing in cells, through the delivery of a CRISPR/Cas 9 complex.

Graphical Abstract

INTRODUCTION

Protein therapeutics have attracted lots of attention due to their potential to specifically supplement for a genetic deficiency, with low off-target effects.^1–3 Delivery of therapeutic proteins to the target sites has been the bottleneck in this practice.^1–11 As proteins are structurally fragile in general, minor perturbations in their environment can result in loss of activity. Large size and hydrophilicity also make proteins impermeable to the cell membrane. As a result, most therapeutic proteins are focused on the extracellular targets.^1–4 Note that ~70% of human proteins are intracellular and therefore there exists an enormous untapped potential to address biomedical challenges, if we were to gain the ability to deliver proteins inside cells.¹² An illustrative example involves the requirement for delivering the Cas9 protein in the CRISPR technology.^10–16

Indeed, various intracellular delivery strategies for proteins with diverse carriers, ranging from inorganic particles to polymeric materials, have been developed.^{12, 17–33} Among these, mesoporous silica (MSi) is promising because of its ready availability, simple preparation, ordered porous structure, and high specific surface area.^33–39 The ordered channels can be ideal host for the encapsulation of guest molecules. Further, silica is considered to be ‘Generally Recognized As Safe (GRAS)’ by the FDA.⁴⁰ Due to these desirable features, MSi materials have been widely used for small molecule drug delivery.^41,42 Recent developments in MSi synthesis has allowed for the generation of materials with various shapes and channel sizes, offering the opportunity to encapsulate much bigger molecules, including proteins.^35–37 However, intracellular protein delivery using MSi materials remains rare.

The dearth of reports on protein delivery using MSi systems could be attributed to low encapsulation efficiency. Strategies for encapsulating proteins inside MSi porous channels have been based on nonspecific physical absorption, which understandably results in weak binding and require high protein feed.^36–38 More importantly, the low affinity interaction results in unstable encapsulation that might cause proteins to prematurely diffuse out, especially from dilution effects in vivo. Furthermore, the large protein feed requirement makes it impractical for precious protein cargos. We hypothesize that engineered interactions between MSi and proteins could greatly increase the encapsulation efficiency for intracellular delivery. In this paper, we outline a simple method to enhance such an interaction and demonstrate the utility of such an approach in intracellular protein delivery (Scheme 1).

Scheme 1.

Protein encapsulation inside MSi-NH₂ and intracellular delivery with traceless release. The surface reversibly engineered protein can efficient interact with MSi-NH₂ through the amine-boronic acid dative bond and enhanced electrostatic interaction between them. The encapsulated protein can be delivered into the cytosol, where the pristine proteins will be released tracelessly under the intracellular stimuli to liberate the surface modified moieties.

RESULTS AND DISCUSSION

Dative bonding interactions between amines and boronic acids, along with potential electrostatic interactions, were targeted for encapsulating proteins within MSi particles. Proteins are polyampholytes, i.e. they have a mixture of both positively and negatively charged functionalities on their surfaces. For optimal electrostatic interactions, we hypothesized that modified protein surfaces present better opportunities. Accordingly, lysine moieties on the protein surface were modified with an aryl boronic acid moiety (Scheme 2).⁴³ In this process, we concurrently converted positively charged lysines to a negatively charged functional group, while also installing the boronic acid moiety for dative interactions onto the protein surface (Scheme 2a). Further, the linkage between boronic acid termini functional group and the proteins can offer reversibility through a self-immolative linker in this conjugation (Scheme 2b and 2c).^44–48 The reversible linker is based on a strategically placed disulfide moiety,⁴⁷ which should degrade and liberate pristine proteins under the reducing environment of the cytosol (Scheme 2b). We demonstrate the encapsulation and intracellular delivery of proteins of different sizes and pIs in three different cell lines. As a further demonstration of utility of this process, we deliver the ribonucleprotein (RNP) complex, composed of gRNA and Cas9 protein, into cells for efficient gene editing.

Scheme 2.

a) protein modification with different linkers via lysine conjugation, b) self-immolation chemistry of linker on protein, c) general structure of the linkers used in this study.

With boronic acids on the surface of the proteins (Figure S1), we next prepared different types of porous MSi particles containing amine functionalities with different sizes and morphologies. First, MCM-41 MSi particles with a non-uniform size of ~900 nm μm (Figure 1a–c) was prepared.³⁸ These MSi exhibited well defined axial ordered channels with a diameter of ~6 nm (Figure S2). The surface charge of the particles changed from −30 mV to +12.3 mV, upon functionalization (Figure 1d), confirming the successful incorporation of amines on the MSi-NH₂ surface.

Figure 1.

Characterization of MSi and complexation with proteins. a) and b) TEM image of MSi-NH₂ with different magnifications; c) DLS measurements and d) zeta potentials of MSi, MSi-NH₂ and FITC labeled MSi-NH₂; e) SDS-PAGE gel for RNaseA complexation (Lane 1. MSi + RNaseA, Lane 2. MSi + RNaseA-BA, Lane 3. MSi-NH₂ + RNaseA, Lane 4. MSi-NH₂ + RNaseA-BA, Lane 5. RNaseA, Lane 6. MSi-NH₂-FITC+ RNaseA-BA); f) Rhodamine b labeled BSA for complexation (1. MSi + BSA, 2. MSi + BSA-BA, 3. MSi-NH₂ + BSA, 4. MSi-NH₂ + BSA-BA). g) TEM image of DMSi-NH₂, h) DLS measurement and i) zeta potential of DMSi-NH₂.

The possibility of encapsulating the boronic acid functionalized proteins into MSi-NH₂ was tested. Gel electrophoresis shows that the boronic acid modified RNaseA (from linker 2) can efficiently interact with MSi-NH₂, as evidenced by the much-diminished RNaseA band (Figure 1e). None of the control samples, containing RNaseA with MSi or boronic acid modified RNaseA (RNaseA-BA) with MSi or RNaseA with amine functionalized MSi, had any effects on the RNaseA band. These results show that the boronic acid – amine interaction forms the basis for efficient loading of proteins within MSi particles. The complexes showed the same size as the host MSi-NH₂ particles with a reduced zeta potential (Figure S3). Different linkers (1-3) result in similar complexation (Figure S3c and S3d). Further, the encapsulated RNase A can be efficiently released from the complex under reduction condition from the reversible linker 2 (Figure S4).

The importance of this interaction was further explored and confirmed using rhodamine b labeled bovine serum albumin (BSA), which allowed for direct visualization of the complexation through sequestration from solution to the sedimented particles (Figure 1f). The encapsulation was found to be efficient with BSA-BA@MSi-NH₂ even at μg concentrations, as all the BSA-BA in the solution can be encapsulated into the MSi-NH₂ particles, while other combinations resulted in incomplete encapsulation. The loading efficiency was further quantified by fluorescence of the unencapsulated protein in the supernatant based on a calibration curve (Figure S5). The encapsulation efficiency was nearly quantitative (>97%) with a loading capacity of ~20%.

To evaluate the delivery ability of this complex, we first checked the cellular uptake efficiency using FITC-labeled MSi-NH₂. We confirmed that the FITC-modification didn’t affect MSi-NH₂ interaction with proteins as demonstrated by DLS, zeta-potential and SDS-PAGE (Figure 1). The MSi-NH₂ particles were efficiently taken up by cells as the strong green fluorescence spread in the cytosol, around the nucleus, throughout HeLa cells in just 4 hours (Figure S6). To evaluate the possibility of protein delivery, HeLa cells were incubated with BSA-BA@MSi-NH₂ complex for 4 h. A diffused red fluorescence of the rhodamine-labeled BSA was similarly observed throughout the cells (Figure 2a and S7). Further, z-stacked orthogonal projection from CLSM experiments demonstrated the internalization of the protein inside the cells (Figure S8). By comparison, negligible fluorescence from BSA without the boronic acid modification was observed (BSA@MSi-NH₂). With a pI of 5.8, BSA is considered to be negatively charged at pH 7. Thus, the lack of uptake by the cells of BSA without the boronic acid functionality indicates that the electrostatic interaction between BSA and amine functionalized MSi alone is not sufficient for intracellular delivery. Similarly, both BSA and BSA-BA cannot efficiently access the cellular interior either by themselves or in the presence of unmodified MSi (Figure 2c, 2d and S7). Even though the boronic acid modification on proteins were reported to enhance the cellular uptake,⁴⁵ the delivery efficiency is not high. Together, these results show that the enhanced interaction between protein and MSi vehicles can efficiently facilitate the delivery of protein cargos inside cells. Variations in linkers (1-3) did not alter complexation (Figure S3); the complexes also showed the similar cell uptake efficiency (Figure S9). These results show the efficiency of our engineered interactions in forming protein complexes with different triggerable release features.

Figure 2.

Intracellular delivery of different proteins. a-d) Confocal images of BSA delivery based on different strategies (red color represents BSA, blue color represents nucleus): a) BSA-BA&MSi-NH₂, b) BSA&MSi-NH₂, c) BSA&MSi, d) BSA-BA&MSi; e) and f) cell uptake mechanism study; g) endosome escape experiment; h) information of the proteins (β-gal is tetramer); Confocal images of i) GFP; and j) RNaseA delivery.

Next, we evaluated the mechanism of cellular entry using inhibitors of known uptake pathways. Amiloride (AMI, micropinocytosis), chlorpromazine (CPZ, clathrin-mediated endo-cytosis), filipin and nystatin (FL & NYS, caveolae-mediated endocytosis), methyl-β-cyclodextrin (Me-β-CD, lipid raft mediated endocytosis), and dynasore (DYN, dynamin-mediated endocytosis) were employed in this effort. As indicated by flow cytometry (Figure 2e and 2f), all the inhibitors can result in decreased cell uptake (AMI: 50%; FL: 60%; CPZ:60%; NYS: 40%). However, dynasore significantly reduced cell uptake (down to ~10%) suggesting dynamin-mediated endocytosis as the dominant pathway.

Note however that the targeted intracellular location for these protein cargos is the cytosol. Therefore, we investigated whether these assemblies escape the endosome by labeling the cells with green lysotracker dyes (Figure 2g). Proteins do efficiently escape the endosome after 4 h incubation, as the red color from the rhodamine labeling is seen throughout the cell with very little co-localization with the endosome/lysosome (green color). The efficient endosome escape may be due to the ‘proton sponge effect’ of MSi-NH₂ particles.^50,51

Although the proteins are delivered inside the cells, these studies do not yet show whether the cargo is intracellularly released from the MSi carrier. We evaluated the temporal evolution of rhodamine fluorescence from the BSA cargo and that of fluorescein from the carrier. Indeed, both of these were colocalized at earlier times, suggesting that the carrier is involved in ferrying the protein into the cells, but are together (Figure S10). After 24 h however, the dominance of red color shows that the protein is released from the MSi particles inside the cells.

To test the versatility of this methodology, we further tested this approach using two other proteins with variations in both size and pI. Rhodamine B-labeled RNaseA (Mw=13.7 kDa, pI=8.93) and GFP (Mw=26.9 kDa, pI=5.67) were modified with boronic acid for complexation. Both RNase-BA@MSi-NH₂ and GFP-BA@MSi-NH₂ are efficiently delivered into HeLa cells in just 4 hours (Figure 2h and i). All control samples that lack the engineered protein-particle interaction exhibited very low uptake (Figure S11 and S12).

Traditional MCM-41 MSi particles, with relatively small channels, are not suitable for large protein cargos. With advances in MSi syntheses, these channel sizes can be tuned.^35–38 We chose amine-functionalized dendritic MSi (DMSi-NH₂) with the channel size of ~20 nm³⁹ to test the possibility of utilizing this methodology for larger proteins. 490 nm DMSi-NH₂ particles, with a channel size of 15–30 nm and surface charge of +13.6 mV, were synthesized (Figure 1g–i). The characterizations of the complex were detailed in Figure S13. With these particles too, the boronic acid modified BSA and GFP can be efficiently delivered into cells, compared to the controls (Figure S14). These results demonstrate the generality of our protein delivery strategy in that it is not limited to a specific morphology or size of the MSi particles and the cell uptake mechanism is similar to that in the MCM particles (Figure S15 and S16). Also, the big channel size of this dendritic MSi will be suitable for big sized proteins, which will broaden the application. We further prepared a small sized dendritic MSi (SDMSi, d = ~80 nm)⁵² for protein encapsulation and intracellular delivery (Figure S17–S19). The SDMSi particles showed very similar results as MSi and DMSi, which can efficiently encapsulate the proteins and further deliver them inside cells with our strategy. Utilizing particles of different sizes for encapsulation and delivery shows the potential utility of our strategy in diverse biological applications.

To test if larger proteins can be delivered with retention of function inside the cells, we used β-gal (Mw ~464 kDa) for study. MSi-NH₂ particles were used to encapsulate and deliver modified β-gal, the intracellular activity of which was tested using the X-gal assay. The delivery efficiency of our strategy is highlighted by the fact that various cell lines uniformly demonstrated high enzymatic activity (Figure 3a), compared to other controls that lacked the required modifications for the enhanced interactions (β-gal with MSi, β-gal with DMSi-NH₂ and β-gal-BA with MSi, Figure 3d–3f and Figure S20–S21). As a further testament to the strategy’s generality, the high cell uptake efficiency and enzymatic activity of β-gal was consistently present in all three cell lines (HeLa, MDA-MB-231, MCF-7) tested (Figure 3 and S21).

Figure 3.

X-gal assay. Delivery of β-gal-BA@MSi-NH₂ toward different cell lines: a) HeLa, b) MDA-MB-231, c) MCF-7. Control experiments for delivery of β-gal toward HeLa cells: d) β-gal+MSi, e) β-gal-BA+MSi, f) β-gal+MSi-NH₂. g and h) photographs for x-gal assay: g) HeLa cells, (1. β-gal+MSi, 2. β-gal-BA+MSi, 3. β-gal+MSi-NH₂, 4. β-gal-BA+MSi-NH₂). h) MDA-MB-231 and MCF-7 cell lines, (1&5. β-gal+MSi, 2&6. β-gal-BA+MSi, 3&7. β-gal+MSi-NH₂, 4&8. β-gal-BA+MSi-NH₂).

Finally, we wanted to demonstrate the use of our strategy to deliver a protein cargo with significant therapeutic function. CRISPR/Cas9 has great potential in addressing genetic malfunctions due to its gene editing capabilities. A challenge associated with this system is that the positively charged protein (Cas9) must be delivered in conjunction with a negatively charged single guide RNA (sgRNA). Note that in our system, boronic acid modification of Cas9 will reduce the positive charge of the protein, rendering the overall CRISPR complex, Cas9/sgRNA, to be more negatively charged. To test this approach, we initially used the linker 2 for modifying Cas 9. However, this modification caused Cas 9 to lose its DNA editing activity (Figure S22). We surmised that this could be due to the hydrophobicity of the modified Cas9. To address this, we synthesized and used a short oligoethylene glycol-based linker 3, which is more hydrophilic but still redox-sensitive (Scheme 2). The resultant Cas9-BA/sgRNA retained its original DNA cleavage activity (Figure 4a) when modified with linker 3. Next, we complexed Cas9-BA/sgRNA with MSi-NH₂ and tested its gene editing ability. We designed the sgRNA to target a gene responsible for the expression of a destabilized version of GFP (deGFP) in HEK293 cells. In the event of functional delivery of the CRISPR complex, we should observe loss of fluorescence of GFP as a result of successful deGFP gene editing. After incubation of the cells with Cas9-BA/sgRNA@MSi-NH₂ for 3 h, with subsequent gene-editing time of 3 days, the fluorescence of GFP was analyzed and compared with untreated cells (Figure 4 and S14). The gene editing efficiency was found to be dose dependent, with a maximum of ~45% at ~50 pmol of Cas9-BA/sgRNA feed (Figure 4b) and ~30% at 24 pmol feed (Figure 4c). In comparison, other controls such as bare Cas9/sgRNA complex, Cas9-BA/sgRNA, Cas9/sgRNA@MSi or MSi-NH₂ did not exhibit any significant gene editing. The editing efficiency was comparable to a commercially available vehicle, Lipofectamine (Figure S23). Further, the T7E1 mutation detection assay was used to confirm gene editing at the target DNA site. Treated cells showed ~30% genome editing efficiency toward the intended DNA site (Figure 4d). Together, these results suggest that our delivery strategy based on the efficient interaction between the cargo and the carrier can transport cell impermeable proteins into cells to perform enzymatic functions, including gene editing.

Figure 4.

CRISPR/Cas 9 delivery for gene editing. a) Agarose gel for activity test toward CRISPR complex after linker modification, b) Flow cytometry analysis of cell populations achieved after GFP knockdown in deGFP expressing HEK 293 cells, c) quantified gene editing efficiency with a dose dependent profile; d) T7E1 assay to detect on-target DNA mutations after treatment with CRISPR-BA@MSi-NH₂.

CONCLUSION

In summary, we have developed an efficient method for intracellular protein delivery by engineering an enhanced interaction motif between protein cargos and mesoporous silica-based delivery vehicles. Specifically, we have shown that with the surface modification of the proteins with boronic acid groups and the MSi delivery scaffolds with amine groups, the interaction between the carrier and proteins can be significantly enhanced through dative bond formation and increased electro-static interactions. We demonstrated the utility and versatility of this approach by successfully delivering proteins of different sizes and pIs with applications, ranging from increasing a specific enzymatic activity inside cells to effective gene editing. We expect that our strategy can benefit many intracellular protein-based therapeutic technologies. We chose general mesoporous silica for this approach, because of its high surface area and tunable pore size. The strategy described here can be expanded beyond MSi scaffolds and applied to other inorganic/organic materials in future for efficient delivery of therapeutic proteins and other cargos.