Surface Presentation of Hyaluronic Acid Modulates Nanoparticle-Cell Association

HA presentation on liposomal surfaces alters NP-cell association

To generate liposomes with HA coupled onto the particle surface, we installed an alkyne handle onto the reducing end of 5 kDa HA via reductive amination. We confirmed that the functionalization occurred with high efficiency using ¹H NMR and FT-IR (see Methods). We fabricated liposomes with azide-terminated phospholipids to install the HA-alkyne via a click reaction involving the copper-catalyzed azide-alkyne cycloaddition.⁴⁹ We elected to use 5 mol% of an azide-modified lipid in the liposomes to have our total surface modification in line with clinically approved formulations (e.g. Doxil).⁵ Though we were unable to quantify the number of HA molecules conjugated on the liposome, we assume that the binding sites are fully saturated because the conjugation was performed with excess HA and with extended reaction time.

In parallel, we generated HA LbL NPs where the HA is the outermost layer.³³ We successfully layered poly-L-arginine (PLR) onto an anionic liposomal core (Figure S1), followed by 5 kDa HA as the terminal layer in our NP constructs. We used 5 kDa HA as an anionic layer to directly compare NPs that present HA polysaccharides of the same molecular weight. The coupled and layered HA constructs yielded stable NPs with relatively similar bulk properties; Z-average diameters ranging from 122 ± 2.8 to 144 ± 1.1 nm, polydispersity indices ranging from 0.130 ± 0.081 to 0.153 ± 0.034, and negative zeta potentials of −18 ± 0.3 mV to −56 ± 1.7 mV, respectively (Figure 2A). The less negative zeta potential of the layered NP is likely due to the underlying cationic charge from PLR. Since 5 kDa HA only has ~10-12 carboxylate functional groups available to convert the surface charge, high weight equivalents of 5 kDa HA were required to achieve stable NPs (vide infra).

We next performed flow cytometry to evaluate the extent of NP-cell association. The covalently coupled HA NPs associated significantly worse than the layered NPs. The layered NPs had superior associated with two immortalized high-grade serous ovarian cancer cell lines with high relative CD44 expression levels:⁵⁰ COV362 and OVCAR8 (Figure 2B). These two cell lines were chosen since tumors derived from them are known for their metastatic progression, complicating drug delivery.⁵¹ We hypothesized that HA covalently appended to NPs will have limited accessibility, and therefore will be less capable of engaging CD44 via multivalent binding. Indeed, the cell association of the surface-coupled HA NPs was only modestly increased (10- and 8-fold at 24 h, respectively) relative to unmodified liposomes.

In contrast, the 5 kDa layered HA NPs had notable, statistically significant increases in association with both cell lines compared to the unmodified liposome or the coupled HA (99- and 22-fold at 4 h, respectively). This effect was magnified with longer incubation times (105- and 74-fold respectively at 24 h). We attribute this difference to the more accessible HA in the layered configuration. Specifically, typical LbL assemblies have polyion loops and tails that extend away from the colloidal surface.⁵² Thus, the layered LbL assemblies would have a more thorough surface coating and more accessible HA. This results in displaying HA in a configuration that is more likely to bind multiple CD44 receptors with high avidity. Indeed, the crystal structure of HA engaging with CD44 reveals that binding occurs through multiple hydrogen bonding events in a buried binding pocket, giving credence to the requirement of available loops of HA from the NP surface.⁵³ These data highlight the advantages of electrostatic LbL adsorption as a robust multivalent surface functionalization strategy.

Molecular weight of the terminal HA layer affects ovarian cancer cell association of LbL NPs

Our determination that HA presented on the LbL assembly enhanced NP-cell association prompted an investigation of the effects of terminal layer HA molecular weight. To this end, we assembled a panel of layered NPs using HA of increasing molecular weight (ranging from 5 kDa to 100 kDa). Cationic layered PLR-adsorbed liposomes were coated with these varying molecular weights of anionic HA. To determine the HA quantities needed for stable charge conversion, routinely defined as approximately |30 mV|,⁵⁴ we titrated increasing amounts of HA onto the PLR-layered liposome (Figure 3). All HA polymers above 10 kDa required 0.5 weight equivalents (wt. eq.) to NP to convert the underlying cationic nanoparticle charge to at least −25 mV with diameters < 210 nm (Figure 3A). However, the lowest molecular weight, 5 kDa HA, required higher wt. eq. for similar stability and charge conversion; only by layering with 2 wt. eq. of 5 kDa HA did the charge reach a value of −21.7 mV (Figure 3A). We hypothesize that, while the |30 mV| zeta potential for these NPs was not reached, the highly hydrated HA chains provide additional steric stability for these nanoparticles, even at a lower net charge. The 100 kDa HA NPs tend to have larger hydrodynamic diameters (Figure 3A), most likely due to HA hydration and subsequent swelling, as measured using dynamic light scattering (DLS).

We next measured the NP-cell association of the increasing HA molecular weight panel of LbL NPs. NPs layered with HA molecular weights > 5 kDa HA had significantly higher association with both cell lines than the LbL NPs layered with 5 kDa HA (Figure 3B). This trend was apparent at both time points (4 and 24 h incubation periods). The linear HA used in this study is highly anionic at physiological ionic strengths, prompting discussions about its exact surface morphology.^55–57 The 5 kDa HA contour length matches its persistence length suggesting this molecular weight HA may have a stiff rod-like morphology.⁵⁷ Therefore, when low molecular weight HA is displayed on NP surfaces, the HA oligomers may function as consecutively adsorbed segments, also known as trains, which lay “flatly” on the NP surface. This presentation mode would leave fewer HA repeat units free to interact with cell surface receptors, thereby attenuating the benefits of the multivalent nanoparticle surface. Furthermore, previous studies have shown that HA polymers with fewer than 18 repeat units exhibit monovalent and reversible CD44 binding. In contrast, HA composed of more than 18 repeat units can bind to multiple copies of CD44 and therefore interact with higher avidity.⁴¹ The 18 repeat units needed for multivalent binding correspond to a molecular weight of roughly 7 kDa, which matches the increased cellular association we observed for HA polymers of 10 kDa and above. This apparent molecular weight threshold may also explain the plateau¹⁸ we observed in cellular uptake of NPs layered with higher molecular weights of HA. Once there are HA loops off the NP surface, the layered architecture likely does not change considerably with longer HA polysaccharides. Thus, as long as the NPs are generated using HA of sufficient length (≥ 10 kDa), their cellular interactions are similar.

Bottlebrush glycopolymer HA constructs effectively engage cell surfaces

Our results concerning increased HA binding with greater molecular weight prompted us to test whether an LbL assembly displaying a bottlebrush arrangement of HA would afford even more effective binding. The bottlebrush architecture offers a pendant display of linear HA molecules, which we hypothesized would increase the availability of binding partners relative to NPs displaying linear HA. To enable direct comparisons, we used the previously synthesized HA-alkyne as the side chain by appending it to a poly(norboronene) polymer backbone bearing azide handles. We leveraged the ring-opening metathesis polymerization to generate poly(norboronene) polymers of controllable size and low dispersity.⁵⁸ We then performed a post-polymerization functionalization involving displacement of the α-chloro amide groups on the polymer with sodium azide to afford the parent polymer (see Supplemental Information). This azide-functionalized polymer was then exposed to the 5 kDa HA-alkyne under standard azide-alkyne cycloaddition click chemistry conditions^{45, 59} to furnish the target bottlebrush HA polymer. Using this polymer in an LbL NP assembly, we generated monodisperse particles of equivalent size and zeta potential to those previously tested (Figure S2).

We employed time-lapse deconvolution microscopy to visualize cell-association differences in our family of NPs. We labeled COV362 cells with a CellTracker dye that is known to pass through cell membranes. We then monitored uptake of our NPs, imbued with fluorescently tagged lipid cores, up to 30 min post-addition (Figure 4A). Differences in NP interactions with COV362 cells (Figure 4A) were readily apparent. The covalently coupled HA NPs have minimal cell-surface association, whereas the linear HA-layered and bottlebrush HA-layered NPs show improved engagement. The extent of internalization increased over the 30 min time course for both layered HA NP constructs, but little internalization of the covalently coupled HA NPs was observed. In representative cell images taken after one hour (Figure 4B), the bottlebrush layered NPs colocalized with endo-lysosomes. In contrast, little, if any, colocalization of the linear layered particles was observed. These notable results led us to further explore the bottlebrush polymer-modified NPs design space.

Bottlebrush HA-glycopolymer outperforms linear HA on LbL NPs, providing a modular platform for chemical modification

Given the effectiveness of the HA bottlebrush-coated NPs in previous studies, we investigated how the polymer composition influenced NP-cell association. We synthesized bottlebrush copolymers substituted with a lower mole fraction of HA by decreasing the molar equivalents of HA-alkyne included in the cycloaddition relative to propargyl-PEG₃-alcohol (PEG) (Figure 5A). The click reaction proceeded in high yield for both alkyne and PEG, enabling the generation of three bottlebrush copolymers: 3:1, 1:1, and 1:3 HA:PEG (Figure 5A). Decreasing the density of HA in the copolymer led to a concomitant decrease in anionic charge and water solubility. Consistent with the reduced charge density of bottlebrush copolymers by including higher PEG functionalization, the weight equivalents required for stable, charge-converted particles, as indicated by both nanoparticle sizes < 200 nm and zeta potentials < −11 mV, doubled from 0.25 wt. eq. for the 100% HA bottlebrush and 3:1 HA:PEG to 0.5 wt. eq. for the 1:1 and 1:3 HA:PEG bottlebrush (Figure 5B). Similar shifts in zeta potential arise when plotting data in terms of equivalent carboxylate units (Figure S3). Again, we note that the rule of thumb of |30 mV| zeta potential for stable colloids is not reached, probably due to the steric stabilization of the large, surface absorbed bottlebrush polymer.

We next evaluated the stability and physiochemical properties of these NPs. NPs layered with 1:1 and 1:3 HA:PEG bottlebrush polymers were unstable after purification. Tangential flow filtration, which was used to purify all other NPs, resulted in aggregation, most likely from the stripping of polymer at high shear rates. Dialysis in a buffered solution also failed to afford stable NPs from these bottlebrush polymers (data not shown). This instability likely arises from their reduced charge density, as seen by the lower net negative zeta potential of the corresponding NPs. With less anionic NPs, electrostatic repulsion is decreased, causing attractive forces to drive aggregation.⁶⁰ In contrast, layering and purifying the 100% HA bottlebrush and 3:1 HA:PEG bottlebrush polymers resulted in relatively similar size, zeta potential, and stability as linear layered NPs (Figure S2).

Intriguingly, both the layered bottlebrush NPs had a statistically significant higher degree of cell association than the analogous layered linear NPs—both 5 kDa and 40 kDa HA—after 24 hours of incubation (Figure 5C). Furthermore, layered bottlebrush NPs containing no PEG or 3:1 HA:PEG ratio in the HA bottlebrush afforded statistically similar extents of cell association. We postulate that the pendant display of 5 kDa HA from the bottlebrush architecture may allow these small HA chains to act as tails projecting off the NP surface. Both steric and rotational limitations of the poly(norbornene) backbone may hinder the electrostatic coupling of some HA side chains onto the nanoparticle surface and promote their availability for cell surface receptor binding. Thus, NPs presenting ligands on a bottlebrush scaffold can offer increased accessibility and, therefore, more efficient cell targeting.

We note that for the batch of NPs presented in Figure 5, the PLR layer was adsorbed at a lower weight ratio than the previous NPs tested. Previous figures used a 0.3 weight equivalent of PLR, yielding NPs with a highly charged surface of 64.5 ± 1.63 mV. In subsequent batches, a determined, optimal 0.086 weight equivalent of PLR was used, resulting in a less positive surface charge of 45.9 ± 2.27 mV (Figure S1). The reduction in underlying positive charge may enable more HA loops off the NP surface—even for low molecular weight HA—leading to an enhanced relative cellular association. This model is corroborated by the observed NP cellular association plateau with increasing HA molecular weight, suggesting there is a maximum number of available loops off the NP surface above a given molecular weight. These findings further emphasize that linear 5 kDa HA is near the limit of molecular weight for forming stable particles.

Finally, we probed how NP surface architectures might impart increased binding avidity by carrying out competition studies. We incubated OVCAR8 cells with exogenous, high molecular weight HA (100 kDa) for thirty minutes prior to nanoparticle dosing to bind to, and block, cell surface CD44 receptors. We subsequently added each of our nanoparticle formulations to monitor their ability to competitively bind. After two hours of incubation, we observed a decreased cell association across all HA nanoparticle formulations, but no effect in the unfunctionalized, untargeted liposomes (Figure 5D). In accord with the live-cell imaging and flow cytometry data, the bottlebrush HA-layered NPs required the highest concentrations of exogenous HA to outcompete and block nanoparticle-cell associations. These results suggest that multivalent interactions of the dense HA side chains increase the avidity of nanoparticle-cell binding, promoting higher overall association. The effective IC₅₀ values of exogenous HA against bottlebrush HA nanoparticles was higher than that for either the linear HA or 3:1 HA:PEG bottlebrush nanoparticles (Table 1). The presence of neutral PEG side chains for the 3:1 HA:PEG bottlebrush nanoparticles resulted in a reduced avidity compared to linear HA, but higher overall cell association (Figure 5C). Thus, the optimized NPs were highly effective at both associating and avidly binding CD44-positive ovarian cancer cells.

The modularity afforded by this NP design approach presents many opportunities. Specifically, we used PEG within the bottlebrush, but alkynes bearing different functional groups could readily be incorporated. We envision conjugates presenting orthogonal receptor-ligand pairings including neutral moieties, like many saccharides, which could increase the specificity of cell surface targeting. In addition, the layered bottlebrush polymers could be modified to display small molecule fluorophores, drugs, or other therapeutics for imaging or delivery to a tumor.

Materials

All chemicals were purchased from Millipore Sigma (Burlington, MA) unless otherwise specified. All sodium hyaluronate molecular weights used, 5 kDa, 10 kDa, 20 kDa, 40 kDa, and 100 kDa, were purchased from Lifecore Biomedical (Chaska, MN). Propargyl-PEG3-alcohol was purchased from BroadPharm (San Diego, CA). Anhydrous solvents were obtained from a solvent purification system by Pure Process Technology (Nashua, NH). Regenerated cellulose dialysis tubing for polymer purification was purchased from Thermo Fisher Scientific (Waltham, MA). Lipids and plant-based cholesterol were purchased from Avanti Polar Lipids, Inc (Birmingham, AL), and the chloroform used to resuspend them was purchased from Macron Fine Chemicals (Randor, PA). Cholesterol was resuspended in a 50 mg/mL solution of chloroform. 1,2-distearoyl-sn-glycero-3-phosphocholine (18:0 DSPC), 1,2-distearoyl-sn-glycero-3-phospho-(1’-rac-glycerol) (18:0 DPSG), 1,2-distearoyl-sn-glycero-3-phosphoethanolamine (18:0 DSPE), and 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-(6-azidohexanoyl) (ammonium salt) (18: azido PE) were resuspended in a 50 mg/mL chloroform solution, 25 mg/mL chloroform solution, 25 mg/mL solution of 65:35 vol:vol chloroform:methanol solution, or a 5 mg/mL of 65:35 chloroform:methanol solution, respectively. Sulfo-Cyanine5 succinimidyl ester dye was purchased from Lumiprobe (Hunt Valley, MD). Membranes used for liposomal extrusion were purchased from Whatman (Maidstown, England). Tangential flow filtration and dialysis membranes for nanoparticle (NP) purification were purchased from Spectrum Labs (Rancho Dominguez, CA), and tubing used for particle purification was purchased from Saint-Gobain (Worchster, MA). 38.5 kDa Poly-L-arginine was purchased from Alamanda Polymers (Hunstville, AL).