Defining the Design Parameters for in Vivo Enzyme Delivery Through Protein Spherical Nucleic Acids.

The translation of proteins as effective intracellular drug candidates is limited by the challenge of cellular entry and their vulnerability to degradation. To advance their therapeutic potential, cell-impermeable proteins can be readily transformed into protein spherical nucleic acids (ProSNAs) by densely functionalizing their surfaces with DNA, yielding structures that are efficiently taken up by cells. Because small structural changes in the chemical makeup of a conjugated ligand can affect the bioactivity of the associated protein, structure-activity relationships of the linker bridging the DNA and the protein surface and the DNA sequence itself are investigated on the ProSNA system. In terms of attachment chemistry, DNA-based linkers promote a sevenfold increase in cellular uptake while maintaining enzymatic activity in vitro as opposed to hexaethylene glycol (HEG, Spacer18) linkers. Additionally, the employment of G-quadruplex-forming sequences increases cellular uptake in vitro up to fourfold. When translating to murine models, the ProSNA with a DNA-only shell exhibits increased blood circulation times and higher accumulation in major organs, including lung, kidney, and spleen, regardless of sequence. Importantly, ProSNAs with an all-oligonucleotide shell retain their enzymatic activity in tissue, whereas the native protein loses all function. Taken together, these results highlight the value of structural design in guiding ProSNA biological fate and activity and represent a significant step forward in the development of intracellular protein-based therapeutics.

Introduction

When materials are structured on the nanoscale, they exhibit unique properties that can be exploited for various applications. DNA is a notable illustration of this concept; when densely radially functionalized onto a nanoparticle core in the form of a spherical nucleic acid (SNA), DNA achieves high cellular uptake through the engagement of scavenger receptors and is more resistant to nuclease degradation than its component nucleic acids.[1−5] These properties have enabled the development of SNAs as diagnostic tools and in therapeutic applications, ranging from gene regulation to immunotherapy.[6−17] Because the three-dimensional architecture of the DNA shell is the primary driver for biological function, the SNA platform has been expanded to include biocompatible cores such as liposomes, polymers, and proteins.[18−21] Protein SNAs (ProSNAs) are especially attractive because they address an unmet need in biologics development: facilitating the cellular delivery of large, hydrophilic, and charged macromolecules while maintaining their biological function.

The notion that DNA shell design parameters dictate the biological fate of SNAs is well-established. One such example is that high oligonucleotide loadings correlate with enhanced cellular uptake.[22] In previous work with gold nanoparticle-based SNAs, hexaethylene glycol (HEG), also known as Spacer 18 (Sp18), were introduced at the interface between the nanoparticle core and its oligonucleotide shell to mitigate the electrostatic repulsion between the negatively charged gold surface and the DNA, enabling denser packing.[23] However, a pivot in DNA shell design is required when synthesizing an SNA architecture from a protein core. Proteins typically offer fewer attachment sites and thus may not be affected by electrostatic repulsion to the same degree. Moreover, with many proteins, chemical modification of their surfaces can either positively or negatively influence different aspects of their function. For example, in the development of protein-based therapies, biocompatible polymer modification, or PEGylation, has been utilized to extend the lifetime of susceptible proteins. However, such modifications generally lead to a loss in biological activity.[24−26]

In addition to the introduction of a linker region to the DNA shell of SNAs, another design rule identified with gold-based SNAs was their sequence-specific uptake behavior. Specifically that G-quadruplexes (GQs), which are guanine-rich sequences that can fold into unique secondary structures, contributed to higher cellular uptake and a difference in biodistribution.[27−29] This phenomenon is attributed to a favorable interaction between the scavenger receptor and a denser presentation of negative charge ascribed from the folding of the GQ.[30−32]

With insight into the potential benefits and detrimental effects of sequence design and polymer modifications, we sought to study the effects of incorporating (1) Spacer18 (a polymer mimic) in the DNA shell and (2) a GQ sequence on the cellular uptake, enzymatic activity, and biodistribution of ProSNAs, utilizing the previously established β-galactosidase (β-Gal) model system. We hypothesized that the nature and length of the linker would modulate the properties of the oligonucleotide shell of ProSNAs. Furthermore, the GQs will increase the uptake of the ProSNA akin to previous studies with the gold core. For this purpose, we used in vitro and in vivo models to investigate tunable parameters in the ProSNA DNA shell’s linker design space in the interest of developing a set of general DNA shell design rules for biologic drug development.

Results and Discussion

Impact of Linker Design on Cell Uptake and Activity

β-Gal ProSNAs were synthesized using protocols previously established by our group. The protein was tagged with Alexa Fluor 647 (AF647) to facilitate tracking in vitro and in vivo. To evaluate the impact of DNA linker identity and length on the cellular uptake and enzyme activity of a β-Gal ProSNA, DNA was conjugated to the protein through surface-accessible lysine residues, with either T4 or (sp18)2 (n = 1, 2, 3) near the conjugation site (Figure A). Each linker, T4 or (sp18)2, is approximately equivalent in length but varies significantly in chemical identity and flexibility, which may impact the radial orientation of the oligonucleotide shell and ultimately affect the biological properties of the resulting ProSNAs. UV–vis spectroscopy showed that the DNA loadings were identical (30 DNA/ β-Gal, 3.7 pmol/cm2), regardless of the composition of the shell (Figure S5). Moreover, covalent conjugation of the DNA to the protein core was confirmed via a denaturing SDS-PAGE gel, which shows mobility shifts commensurate with the increased mass upon DNA addition (Figure S6). Importantly, in contrast to classical gold-based SNAs, the absence of Spacer18 did not significantly affect loading density onto ProSNAs, suggesting that each surface-accessible lysine on β-Gal was readily available for conjugation, independent of linker type.

Figure 1

Impact of linker design on cellular uptake and enzymatic activity. (A) Cartoon representation of the surface of β-Gal at a lysine residue, which is covalently modified with a DNA strand. The “linker region”, at the interface of β-Gal and a T30 oligonucleotide strand, varies in either increasing T4 or 2 Spacer18 phosphoramidite increments. The representation was adapted from PDB ID 1PX3. (B) Cellular uptake of ProSNAs with increasing T4 (purple) or Spacer18 (pink) linkers, as determined by flow cytometry. Fluorescence was measured in HeLa cells (n = 3) 4 h after treatment with 5 nM enzyme in serum containing media. All ProSNAs have equal loading density of 30 DNA/β-Gal. (C) Relative kcat (n = 3) determined by an ONPG assay of ProSNAs with varying linker identity and the native protein. All ProSNAs have equal loading density of 30 DNA/β-Gal. The bars in the graphs represent the mean, and error bars show the standard error of the mean (SEM). An unpaired t test was performed with GraphPad Prism. n.s., not significant, ** P ≤ 0.01.

One of the defining features of the SNA platform is its robust and unaided cellular uptake, which has been demonstrated in more than 60 cell lines.[33] Specifically, akin to other SNAs, the ProSNA was found to be taken up by cells through scavenger receptor-A (SR-A) mediated endocytosis through a receptor blocking study.[21] To establish the role of the linker on the cellular uptake of ProSNAs, HeLa cells were incubated with the six β-Gal ProSNA variants described above, and their uptake was monitored using flow cytometry by tracking the fluorescence of the protein core (Figure B). Surprisingly, the ProSNA with no Spacer18 was taken up to the greatest extent. Furthermore, internalization dramatically decreased with increasing Spacer18 length, even though these modifications were made adjacent to the protein surface and thus not expected to be accessible to the cell surface. In contrast, increasing the length of the DNA linker imparted only minor changes in cellular uptake. We speculate that these results originate from differences between the linkers in terms of persistence length and ability of the shell to adopt a radial arrangement away from the protein surface, resulting in a decreased affinity for SR-A.[27] The linker’s deleterious uptake properties were further supported by switching the location of Spacer18 on the protein: cellular uptake recovers, confirming that placing Spacer18 near the protein surface is detrimental to internalization (Figures S11 and S12). These findings were further reproduced in another cell line (Figure S15). Based on these results, we find that while gold-based SNAs benefited from the addition of Spacer18 linkers, the functionalization of protein cores should be carried out using linkers that are composed entirely of DNA to maximize cellular uptake.

A critical aspect, unique to ProSNAs in the context of protein delivery, is the retention of protein function after conjugation. To assess the effect of conjugating a large number of DNA strands on the protein surface on function, the enzymatic activity of the β-Gal ProSNAs was measured utilizing an ortho-nitrophenyl-β-galactoside (ONPG) assay, and Michaelis–Menten constants were calculated. Both linkers caused a loss of enzymatic activity as their length increased; however, this effect is more pronounced for the Spacer18 linker (Figure C). We speculate that Spacer18 is more detrimental to enzymatic activity because of an increased affinity of the Spacer18 linker for the protein surface when compared to DNA. To test this hypothesis, we placed the Spacer18 linker away from the protein surface by appending it at the end of a T30 strand and found that activity was improved (Figure S18). Considering that the reported isoelectric point for β-Gal is approximately 5.5, the protein is overall negatively charged at physiological conditions. Similarly, the DNA linkers provide a high density of negative charges through their phosphate backbone. These two elements combined are thus expected to promote electrostatic repulsion and favor a radial orientation of the oligonucleotide shell. On the other hand, Spacer18 linkers are less densely charged than DNA and may have other interactions with the protein surface. Based on these results, we conclude that with this protein, short DNA-only linkers maximize the enzymatic activity of ProSNAs.

G-Quadruplex Shell Enhances Cell Uptake in Vitro

We next assessed sequence-dependent cellular uptake and enzyme activity trends of ProSNAs under physiological conditions. Templated by their DNA sequence, particular strands will fold into secondary structures that can bind preferentially to receptors on the cell membrane. In previous work, our group demonstrated that SNAs constructed using guanine-rich sequences resulted in significantly higher uptake compared to thymine-, adenine-, or cytosine-rich control SNAs. G-rich sequences, which can fold into GQs, are a natural ligand for SR-A due to the proposed favorable electrostatic interaction between the GQ and the lysine-rich collagenase region of the SR-A.[30−32] Because GQs are expected to interact preferentially with SR-A, we hypothesized that ProSNAs with a DNA shell that can form GQs would be able to enter cells in more significant amounts than SNAs composed of other nucleotides (Figure A). GQs exist in various topologies based on their sequences; therefore, we investigated the impact of GQ folding on cellular uptake of the ProSNA in vitro. To facilitate attachment to β-gal, all GQ sequences were synthesized with a T4 DNA-only linker, which was found to maximize both cellular uptake and enzymatic activity (vide supra).

Figure 2

Impact of G-quadruplex sequences on cellular uptake and enzymatic activity. (A) Cartoon representation of the surface of β-Gal at a lysine residue which is covalently modified with a DNA strand. The inset represents either a T-rich and G-rich sequence structure. In the G-quadruplex topology, the guanine bases are selectively stabilized by a potassium cation. The representation was adapted from PDB ID 1PX3 and 2N3M. (B) Cellular uptake of ProSNAs with different G-quadruplex topologies, as determined by flow cytometry. 2KF7 (antiparallel basket), 1KF1 (parallel propeller), 148D (antiparallel chair), and 2JPZ (mixed chair) are named after their PDB ID. Fluorescence was measured in HeLa cells (n = 3) 4 h after treatment with 5 nM enzyme in serum containing media. (C) Cellular uptake of ProSNAs at different loading densities with either a T4T30 or T4(GGT)10 sequence, as determined by flow cytometry. Fluorescence was measured in HeLa cells (n = 3) 2 h after treatment with 10 nM enzyme in serum containing media. (D) Relative kcat (n = 3) determined by an ONPG assay of ProSNAs (both 20 DNA/β-Gal) and the native protein. The bars in the graphs represent the mean, and error bars show SEM. An unpaired t test was performed with GraphPad Prism. **P ≤ 0.01, ****P ≤ 0.0001.

To begin, we characterized all GQ strands by circular dichroism to confirm the chosen sequences fold into the appropriate GQ structure after the inclusion of a T4 linker and DBCO-dT conjugation group (Figure S9). Then we synthesized ProSNAs using the previously reported method but under different buffer conditions. The G-quartet structure in a GQ’s tetrad is selectively stabilized by a K+ ion, which coordinates to each of the guanine bases. To linearize the strand, thus enabling more efficient loading, we replaced K+ with Li+ as the counterion in the reaction buffer. Lithium is the least stabilizing among type Ia and IIa cations; therefore, the GQ does not form under these conditions. After conjugation and ProSNA purification, the buffer was exchanged back to PBS (a potassium phosphate buffer) to facilitate GQ formation (Figure S10).

To determine which GQ topology results in the most significant enhancement in cellular internalization, HeLa cells were treated with each of the ProSNA GQ variants in serum containing media. According to flow cytometry, all GQs demonstrated an approximately twofold increase in uptake compared to the T-rich ProSNA (Figure B). More notably, the T4(GGT)10 ProSNA, which folds into a parallel GQ, increased uptake by fourfold. This enhancement in uptake with the T4(GGT)10 ProSNA was demonstrated over a range of loading densities with the highest cellular uptake seen at a loading of 20 DNA/β-Gal (Figure C).

The enhancement in cellular uptake with ProSNAs composed of a GQ shell stems from the DNA’s unique secondary structure. However, this special folding imparts a small inhibitory effect on the enzyme’s activity. An ONPG assay was used to calculate the kcat of the enzyme for ProSNAs with either T4T30 or T4(GGT)10 DNA shells. After loading both ProSNAs equally (20 DNA/β-Gal), we calculated the kcat of the enzyme and found that there is an ∼30% loss in activity with a GQ shell (Figure D). This small loss in activity may stem from steric hindrance by the dense folding of the DNA around the enzyme. Nevertheless, it is subject to the intentions of the application whether this marginal loss in activity depreciates the gain in cellular uptake for a G-quadruplex DNA shell.

ProSNAs Demonstrate Enhanced Pharmacokinetics

We assessed whether these shell design considerations would translate from in vitro cell studies to a more relevant in vivo mouse model. Previous work using gold core SNAs highlighted the benefits of the oligonucleotide shell in enabling the biodistribution of SNAs to all major organs, which have been applied for the treatment of many diseases, including glioblastoma and many other forms of cancer.[12,16] ProSNAs are a special case because there is an approximate order of magnitude difference in loading density (3.7 pmol/cm2) when comparing to the classical gold SNAs (30–60 pmol/cm2). Moreover, naked proteins are susceptible to degradation and typically show low cellular uptake. Therefore, we evaluated if this loading density would be sufficient to enable applications of ProSNAs in vivo and if these newly devised design rules would be able to amplify these effects further.

Based on cellular uptake and enzymatic activity findings in vitro, we hypothesized that Spacer18 would impede the internalization of ProSNAs within the tissue. For this purpose, we started by comparing constructs bearing either a T30 or (sp18)6T30 DNA shell, such that these two constructs would vary only in the presence of a Spacer18 linker. We assessed both the blood circulation time and temporal biodistribution following a single tail vein intravenous injection of either β-Gal ProSNAs or native protein in CD-1 mice, utilizing the Alexa Fluor 647 fluorescent-tag on the protein to track its distribution.

First, we measured the distribution half-life of these ProSNAs in plasma to determine the clearance rate of these protein constructs. Notably, both ProSNAs exhibit an increase in circulation time (60 min), which is twice that of the native protein (30 min), regardless of the linker identity (Figure A). These similarities in the pharmacokinetic profiles of the SNAs highlight the influence of the DNA shell, regardless of linker, in transforming the biological properties of an SNA’s nanoparticle core.

Figure 3

Pharmacokinetic profile of the ProSNA and native protein. (A) The clearance of two ProSNA variants with and without Spacer18 linker and native protein were studied by measuring the fluorescence in plasma. Raw data points as well as the geometric mean at each time point are depicted. (B, C) After a 1 h treatment with 4 mg of enzyme/kg mouse via a tail-vein injection; mice (n = 3) were sacrificed, perfused, and their organs dissected and fixed. Using IVIS, a region of interest was drawn around each organ, and the fluorescence counts were quantified. The bars in the graphs represent the mean, and error bars show SEM.

To evaluate the distribution of β-Gal in tissue, we measured the fluorescence of dissected and perfused organs utilizing an in vivo imaging system (IVIS). Quantification of radiant efficiency revealed that a majority of both ProSNA and native protein were sequestered by the liver, akin to previous observations with other nanoparticle systems.[34] However, the ProSNAs, in particular the T30 variant, exhibited distinct differences in biodistribution compared to the native protein (Figure B). Consistent with our hypothesis, we observed that the native protein is found primarily in the liver with limited distribution to other organs. In contrast, the addition of the oligonucleotide shell enabled a broader distribution of both ProSNAs. Importantly, an enhanced tissue accumulation was observed with the T30 ProSNA, mirroring our previous in vitro cellular uptake results (vide supra). This finding is of significance because it highlights the importance of using an all-DNA shell when designing ProSNAs because it dictates their in vivo behavior.

In addition to linker identity being a determinant of uptake for the ProSNA, DNA sequence was also established to enhance cellular uptake in vitro. Consistent with the previous experiment, we utilized IVIS to study the distribution of the ProSNA with two different sequence identities: T34 or T4(GGT)10. Both ProSNAs were loaded equally at 20 strands per protein, which we demonstrated resulted in the highest internalization for the GQ ProSNA in HeLa cells. CD-1 mice were treated with each construct via the tail-vein, and after a 1 h treatment, mice were sacrificed and perfused, and organs were dissected for IVIS analysis, which was used to track the AF647 fluorophore covalently conjugated to the protein. Based on fluorescence counts, both ProSNAs accumulated in the dissected organs equally (Figure C). From this experiment, it is apparent that the nature of the linker is a more important determinant of in vivo ProSNA behavior.

Only ProSNAs Exhibit Enzymatic Activity in Tissue

The ultimate test to determine the potential of the SNA architecture as a delivery method for biologics was to assess whether the delivered enzyme retains its biological activity in tissue. Therefore, we performed a colorimetric activity assay specific to β-Gal on cryo-sectioned organs, allowing us to gain information on the suborgan compartmentalization of the SNA platform and its overall structural integrity. For this purpose, mice were intravenously administered ProSNAs or native protein. After 1 h, mice were sacrificed and perfused, and organs were cryosectioned and subsequently stained with 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) to visualize the enzyme’s activity via the appearance of a blue-colored insoluble product.[35] Strikingly, we observed blue coloring in both T34 ProSNA and T4(GGT)10 ProSNA treated mice (n = 3), which confirmed the successful delivery of active enzymes to tissues, while all organs treated with the native proteins (n = 3) exhibited no coloring (Figure ). This is the first time that the ProSNA strategy has been tested in a living animal model, opening the door to various applications for relevant disease models.

Figure 4

In vivo catalytic activity and tissue distribution of native β-Gal and ProSNA. Representative light micrographs of histology slides after incubation with the β-Gal substrate, X-Gal. The blue color apparent in tissue dissected from mice (n = 3) 1 h after intravenous injection of 6.5 mg enzyme/kg mouse results from the hydrolysis of X-Gal and the formation of an insoluble blue product. Scale = 250 μm.

In this context, we looked at the distribution of the blue stain within each organ, which revealed details about tissue localization of the ProSNA. A first striking observation was the presence of blue in the ProSNA-treated liver indicative of enzyme activity, whereas it was completely absent in the case of the native protein. This was unexpected since IVIS fluorescence signals in the liver from all treatment groups were nearly identical, suggesting that the oligonucleotide shell on the ProSNA was able to protect its structural integrity. The liver and lung displayed diffuse staining throughout the organ, whereas in the kidney and spleen, it was localized to tissue substructures. Specifically, the ProSNA is trapped within the glomerulus in the kidney, which is a structural unit that acts as a filter with an average pore size of 3 nm.[36] We hypothesize that the ProSNAs are retained within this glomerular space and are thus prevented from entering the kidney tissue due to their inherent size. However, when the filtration size thresholds of the organs are larger, as is the case with the spleen, the ProSNA readily enters. Indeed, the spleen exhibits two distinct regions based on Nuclear Fast Red staining: red and white pulp. The white pulp is predominantly a sheath of lymphocytes (T and B) surrounding the artery, while the red pulp envelops the white pulp and consists mostly of macrophages, plasma, and a few lymphocytes.[37] Based on these images, the ProSNA is preferentially sequestered in the red pulp of the spleen. In particular, there is a difference in spleen distribution between the T-rich and GQ ProSNAs. The T-rich ProSNA distributes evenly throughout the red pulp. However, for the GQ ProSNA, the enzyme predominantly accumulates in a marginal zone at the junction between the red and white pulp. Regardless, we hypothesize that both ProSNA variants are taken up by macrophages (either marginal zone or red pulp), which take up the SNA due to the abundance of scavenger receptors on their cell membranes.[38]

Conclusions

We established the importance of DNA shell design in tuning the activity, cellular uptake, and overall biodistribution of ProSNAs. We find that the chemical composition of the linker is a crucial determinant of ProSNA properties: specifically, the replacement of the HEG linker with DNA leads to a recovery in both cellular uptake and enzymatic activity at no cost to loading density. Furthermore, we can harness the programmable topology of DNA structures to modulate cell interactions, as demonstrated with the GQ ProSNAs acting as preferential ligands for the scavenger receptor-A. In terms of biodistribution, the removal of the HEG linker resulted in a greater distribution in organs other than the liver. Conversely, both types of sequences facilitated protein delivery to a similar extent. Because the most critical parameter in protein delivery consists in maintaining enzymatic activity, we demonstrate for the first time that the ProSNA strategy can be used to deliver functional enzymes to tissues, thus highlighting the relevance of this approach for biologics development. Considering the facile synthesis and conjugation of DNA onto proteins as well as its biocompatibility, these findings emphasize the relevance of applying the SNA architecture to therapeutically relevant proteins. We foresee that ProSNAs can be tailored to a breadth of applications from immunotherapy to enzyme replacement therapy in macrophage-relevant disease models.