Flexible Nanoparticles Reach Sterically Obscured Endothelial Targets Inaccessible to Rigid Nanoparticles

Abstract

Molecular targeting of nanoparticle drug carriers promises maximized therapeutic impact to sites of disease or injury with minimized systemic effects. Precise targeting demands addressing to subcellular features, among which caveolae, invaginations in the cell membrane implicated in transcytosis and inflammatory signaling, are appealing targets. Caveolar geometry was reported to impose a ~50nm size cutoff on nanocarrier access to plasmalemma vesicle associated protein (PLVAP), a marker found in caveolae in the lungs. We explored the use of deformable nanocarriers to overcome that size cutoff. Lysozyme-dextran nanogels (NGs) were synthesized with ~150nm or ~300nm mean diameter. AFM indicated deformability of NGs on complementary surfaces. Quartz crystal microbalance data indicated that NGs formed softer monolayers (~60kPa) than polystyrene particles (~8MPa). NGs deformed during flow through microfluidic channels, and modeling of NG extrusion through porous filters yielded sieving diameters of less than 25nm for NGs with 150nm and 300nm hydrodynamic diameters. In mouse lungs, where PLVAP is primarily localized to caveolae, NGs of 150nm and 300nm diameter targeted PLVAP while rigid polystyrene particles did not. Our data indicate a role for mechanical deformability in targeting large high-payload drug delivery vehicles to sterically obscured targets like PLVAP.

Intracellular delivery of drug carrier nanoparticles (NPs) and their cargo is an open challenge, where design of carriers and choice of target molecule represent active areas of research aimed at optimizing localization and impact of drugs.^[1-6] Goals of this research include delivery to specific endocytic pathways,^[3,7] to the cytosol,^[8] or across the endothelial barrier.^[9-11] Among factors in NP design for this aim, size of the particle stands out as a conundrum.^[1,2,6] Although larger particles are capable of carrying larger drug payloads, they may also have limited accessibility to target epitopes affected by steric blockade or concealed by features of the cell surface.^[1,2,6,12]

On the endothelial surface, the target epitope to which a nanocarrier is aimed might be obscured by localization in specific regions of the plasmalemma, including caveolae, plasma membrane invaginations characterized by an entry mouth of ~50nm diameter.^[9,10] Caveolae have transport and signaling functions in both normal and pathological conditions.^[13-17] Targeting drug delivery to caveolae has garnered attention as an attractive option for achieving uptake of therapeutic cargo in cells and, potentially, transcytosis of therapeutic cargo across the endothelial barrier.^[11,18]

Plasmalemma vesicle associated protein (PLVAP) is an endothelial surface determinant associated with the diaphragms of caveolae in the lungs.^[19-21] PLVAP is also a component of fenestral diaphragms in most fenestrated vessels and is critical to their barrier function.^[22-24] Given the roles of PLVAP in vascular integrity, its association with caveolae, and its abundance in the lungs, targeted drug delivery to this endothelial marker is an attractive goal. However, while antibodies to PLVAP have accumulated effectively in the pulmonary vasculature, PLVAP-targeted particles larger than 50nm have not accumulated in the lungs,^[25] suggesting a size constraint for targeting NPs to sterically constrained domains like caveolae.^[26,27]

Mechanical deformability may provide one way of overcoming size and geometry constraints on in vivo behavior of NPs.^[6,28,29] Elastic moduli of NPs can alter targeting affinity, specificity, and pharmacokinetics.^[30,31,32] Flexible particles can take on fluid properties, including capacity to translocate through small pores and to compress under shear to spread on target surfaces.^[31,33]

Nanogels (NGs) have garnered attention as polymeric NPs capable of shape and size change in response to mechanical force. The mechanical deformability of NGs may be controlled by properties of their constituent polymers, including crosslinking and polymer backbone properties.^[34,35] Lysozyme-dextran NGs have recently demonstrated potential as drug delivery vehicles targeted to endothelium.^[36] As synthesized, this type of NG has a low crosslinking density that fulfills the design requirements for low elastic modulus polymeric nanoparticles.

In this work, the mechanical deformability of lysozyme-dextran NGs was evaluated and compared to polystyrene spheres of similar size and surface chemistry. On confirming lysozyme-dextran NGs as a mechanically flexible counterpart to polystyrene spheres, both polystyrene and hydrogel NPs were targeted to PLVAP in mice. PLVAP-targeted NGs of two diameters, both greater than the previously inferred 50nm size cutoff for access to PLVAP, successfully targeted the lungs, whereas targeted polystyrene particles of similar size did not. We thus demonstrate the possibility of engineering mechanical flexibility of NPs to access sterically constrained targets like PLVAP.

Morphological Characterization of Nanogels

Lysozyme-dextran NGs (Figure 1a), synthesized as previously described,^[36] were characterized via DLS. By varying the pH at which lysozyme-dextran conjugates formed particles, NGs formed with mean diameters of 149.5nm (PDI .073) (pH 11.34) or 289.3nm (PDI .128) (pH 10.70) (Figure 1b, Supplementary Figure 1a). After periodate oxidation of the dextran NG shells, antibodies, either IgG or mouse PLVAP-targeted, were added to the nanoparticles via aldehyde amine coupling. Antibody-coated NGs had hydrodynamic diameters of either 156.7nm (PDI .096) or 319.1nm (PDI .183), with addition of a layer of antibody accounting for the shift in size distribution (Figure 1b, Supplementary Figure 1a). Purchased aldehyde-latex particles had hydrodynamic diameters of 351.5nm (PDI .027) or 127.7nm (PDI .018) before antibody addition. After antibody addition, polystyrene nanoparticles shifted in diameter to 375.4nm (PDI .16) or 173.4nm (PDI .18), respectively (Supplementary Figure 1b). Similar antibody-conjugated NGs and polystyrene particles, traced with fluorescein tag, maintained stable size distributions during 2.5 hours incubation in platelet poor plasma (Supplementary Figure 1c, 1d).

Fig. 1.

Morphology and mechanical properties of lysozyme-dextran NGs. (a) Schematic of a lysozyme-dextran nanogel, with lysozyme core concealed by a loose dextran shell. (b) DLS number distributions indicate mean NG hydrodynamic diameters of 289.3nm (PDI .128) before antibody functionalization and 319.1nm (PDI .183) after antibody addition (c) AFM morphological characterization of NGs indicates spreading of NGs during surface adhesion. (d) Quantitative analysis of surface-grafted NG morphology showed spreading of 289.3nm diameter NGs to a mean height of 106.5nm. (e,f) QCM analysis of NGs or polystyrene particles (PS NPs). Dissipation factor (dotted lines, consistent with buildup of a viscoelastic film) increased concurrently with accumulation of nanoparticle mass (solid lines) on QCM surfaces. (g) QCM-D modeling of accumulated nanoparticles as viscoelastic films allowed determination of shear moduli of layers of adhered nanoparticles. NG layers had a shear modulus of 67.88 kPa. PS NP layers exhibited a significantly greater shear modulus of 8.47 MPa. (*) Indicates p<.01 difference between NGs and PS NPs in (g).

Oxidized NGs were added to (3-Aminopropyl)Triethoxysilane (APTES) coated glass surfaces to form layers of particles to be characterized by atomic force microscopy (AFM). Deposition of NGs was indicated by contact angle measurements (Supplementary Figure 2). Contact mode AFM determined the morphology of NG films on glass immersed in water (Figure 1c). Deposited NGs exhibited spreading indicative of flexibility of the particles. Individual NGs in a 20μm × 20μm AFM micrograph were evaluated in ImageJ, extracting values for mean particle height and particle footprint area. For particles determined by DLS to have a 289.3nm diameter, AFM determined a mean particle height of 106.5nm (Figure 1d) and a mean particle footprint diameter of 474.4nm. Notably, the average particle volume as evaluated by AFM corresponded to a spherical particle diameter of 325.7nm, matching well with the DLS-determined diameter of 289.3nm.

Mechanical Characterization of Nanogels

Further AFM studies with NGs immobilized on silicon wafers examined the mechanical deformability of NGs. After morphological AFM of NG layers as above (Supplementary Figure 3a), NGs were probed with a colloidal SiO₂ tip to obtain force-indentation curves. Hertzian modeling indicated NG Young’s moduli ranging from 48-71 kPa over a 2.5×2.5μm map of deposited NGs (Supplementary Figure 3b, 3c, 3d).

Quartz crystal microbalance (QCM-D) with APTES-functionalized quartz glass substrates was used to characterize deposition of 300nm NG layers. In separate experiments, aldehyde-presenting polystyrene NPs of similar diameter were exposed to APTES QCM-D cassettes under identical conditions. Change in resonant frequency of the quartz surface indicated steady deposition of NGs over 23 hours exposure to the APTES surface (Figure 1e, solid line). Analogous experiments with polystyrene particles indicated saturation of bead deposition at 15 hours exposure (Figure 1f, solid line).

Alongside QCM measurements of resonant frequency, measurement of dissipation factor indicated the viscoelastic properties of accumulating nanoparticle layers. The dissipation factor for NG films (Figure 1e, dotted line) exceeded that for polystyrene particles (Figure 1f, dotted line). Modeling of frequency and dissipation factor data was used to determine shear modulus of the NG and polystyrene particle layers as 67.88 kPa (n=6) and 8.47 MPa (n=4), respectively (Figure 1g).^[37] Additional modeling determined the viscosity of the NG films as .0027 kg m⁻¹s⁻¹ and the polystyrene films as .012 kg m⁻¹s⁻¹ (Supplementary Figure 4). Both the viscosity and shear modulus as modeled in QCM data indicate a significant difference in the relative flexibilities of lysozyme-dextran NGs and polystyrene particles. Notably, the modeled polystyrene shear modulus is less than that stated in the literature (~1GPa)^[38] and the modeled NG shear modulus is greater than that predicted in simulations (~1 kPa).^[39] These discrepancies may be attributable to shortcomings in applying a viscoelastic model and experimental technique designed for micron-sized films to thin NP layers that may be affected by water inclusion and spatial heterogeneities.

Nanogel Response to Shear in Flow

Suspensions of NGs or polystyrene NPs were subjected to shear in a 4mm × 0.2mm ektacytometry flow channel. In ektacytometry experiments, diffraction patterns generated by flowing particles or cells indicate the degree of elongation of the particles or cells along the direction of flow.^[40] In high-shear flow (~1200 s⁻¹), NGs generated an asymmetrical diffraction pattern indicating elongation of the particles. As shear was continuously lowered, the NG diffraction progressed towards a circular pattern, indicative of more symmetrical particle geometry for NGs at rest (Supplementary Movie 1). Under the same conditions, the diffraction pattern generated by polystyrene particles was an unchanging circle (Supplementary Movie 2), indicating no particle deformation. Plotting the relative elongation of the diffraction pattern as a function of shear stress further indicated NG morphology is mechanically responsive to flow (Supplementary Figure 5a), but PS bead morphology is not (Supplementary Figure 5b).

To demonstrate that NG deformation under flow can permit access to nanostructured environments, NGs and polystyrene NPs were extruded through polymer membranes. When pressed against filters with a rated pore diameter of 100nm, NGs with hydrodynamic diameters of ~150nm and ~300nm passed through at rates exceeding 80%. Similar quantities of NG passage were observed for different flow rates, corresponding to shear ranging from sub-physiological (~15 s⁻¹) to venous (~400 s⁻¹) to arterial (~1000 s⁻¹) (Figure 2a,b, red/orange bars).^[41] Polystyrene NPs, however, were excluded from passage through the membranes. NPs of 250nm diameter did not traverse the membranes and even NPs of 100nm diameter only traversed 100nm cutoff membranes at ~10% for venous shear and at ~25% for arterial shear (Figure 2a,b, green/yellow bars). Similar experiments explored passage of nanoparticles through membranes oriented parallel to flow direction, where ~150nm diameter NGs were compared to 100nm diameter polystyrene NPs (Figure 2c). At a shear rate of 50 s⁻¹, NGs eluted through tubing with 50nm pores at an 85% rate, while polystyrene NPs were excluded (Figure 2d).

Fig. 2.

NG deformation during extrusion through pores. (a,b) At different shear rates, dextran NGs of ~150nm and ~300nm diameter (red and orange bars, respectively) passed through membranes with 100nm pore size in proportions greater than 80%. ~100nm and ~250nm diameter PS NPs only exceeded a maximum of 20% passage through 100nm pores. (c,d) 150nm NG or 100nm PS NP suspensions were flowed parallel to 50nm cutoff porous membranes at a shear rate of 50 s⁻¹. NGs passed through the membranes at a rate greater than 80% and PS NPs were completely excluded by the membranes. (e) The fraction of NGs or PS NPs passing through membranes with different rated pore size (flow rate 0.5 mL min⁻¹) was modeled as a function of pore size, allowing extrapolation of an effective nanoparticle sieving diameter. (f) NGs of 150nm and 300nm diameters as measured by DLS exhibited effective diameters of ~3 nm. After addition of antibody to the nanogels, a sieving diameter of ~20nm was determined. PS NPs of 100nm DLS diameter had an effective diameter of ~50nm.

Filters presenting a range of pore sizes (1, 5, 10, 20, 100, 220, 450, and 800nm) were employed against NG and PS particles suspensions flowing at a rate of 0.5 mL min⁻¹. The eluted fraction vs. pore size relationship for each type of particle was modeled according to an equation derived by Kirtane et al. to express filtration efficiency in terms of pore size and an “effective particle diameter,” indicative of how well filtrate particles pass through pores.^[42] Best fit determination of the sieving parameter for NGs revealed values of 3.3nm for NGs with a 149.5nm DLS diameter and 2.8nm for NGs with a 289.3nm DLS diameter. The effective diameter of 100nm polystyrene NPs, 52.1nm, better matched that determined by DLS. Addition of IgG to 149.5nm NGs reduced the efficiency of particle passage through pores, yielding an effective diameter of 23.4nm (Figure 2e, 2f). The higher effective diameter may be attributable to the IgG layer added to the NGs. The IgG layer itself would be more rigid than the dextran shell of the bare NGs.^[43] Additionally, each IgG presents multiple amines to the multivalent aldehyde-dextran NG shell and therefore has a theoretical possibility of increasing dextran crosslinking density, and thus rigidity.^[39]

Targeting of Nanoparticles to PLVAP in Cell Culture

A murine brain endothelioma cell line (bEnd3) was employed for in vitro testing of NGs functionalized with antibody to mouse PLVAP. Cell culture experiments verified ability of particles coated with PLVAP antibody to engage with PLVAP on the cell surface, rather than ability of particles to access PLVAP in specific domains like caveolae.^[44]

NGs synthesized with rhodamine-dextran and conjugated to PLVAP antibody produced fluorescent signal associated with bEnd3 cells after 30 minutes incubation at 37°C (Figure 3a). Confocal imaging of bEnd3 cells showed targeted NG fluorescence primarily located within cell boundaries inferred by location relative to actin staining (Supplementary Movie 3,4). Epifluorescence and confocal imaging indicated lesser adhesion of IgG NGs as compared to PLVAP-targeted particles (Figure 3b, Supplementary Movie 5, 6). Using NGs coated with biotinylated PLVAP antibody, membrane-impermeable avidin staining indicated that PLVAP-targeted NG internalization in bEnd3 cells (Supplementary Figure 6a, Supplementary Movie 7) was inhibited when lipid rafts were disrupted with filipin pretreatment (Supplementary Figure 6b, Supplementary Movie 8), implying that active internalization of NGs in bEnd3 cells is in part mediated by mechanisms consistent with caveolar endocytosis.

Fig. 3.

NG and PS NP targeting to PLVAP in vitro and in vivo. (a,b) NGs (red) with PLVAP antibody adhere to and are taken up in bEnd3 cells. IgG NGs do not adhere to bEnd3 cells. Membrane impermeable Fc label is in green and DAPI nuclear stain in blue. (c) ¹²⁵I-labeled PLVAP-targeted NGs (red) adhered to and co-precipitated with bEnd3 cells in suspension in greater quantities than IgG NGs (pink). (d) Both PS NPs and NGs with PLVAP antibody adhere to bEnd3 cells expressing PLVAP. (e) 150nm anti-PLVAP NGs (red bars) specifically target mouse lungs. 150nm anti-PLVAP PS NPs do not target the lungs (green bars). (f) Similar results are obtained with larger nanoparticles, where 300nm anti-PLVAP NGs specifically target mouse lungs, but analogous 350nm PS NPs do not. (g,h) Specificity for the lungs (determined by the anti-PLVAP:IgG adhesion ratio) is observed for NGs (orange, red), but not PS NPs (yellow, green). Lung immunospecificity indices for two sizes of NG, two sizes of PS NP, and PLVAP antibody indicate that NGs partially recover the size-dependent targeting observed for PLVAP antibody. (i,j) Proposed mechanism for different lung PLVAP targeting dynamics for PS NPs and NGs. The geometry of caveolae limits PLVAP access for large rigid bodies (i). As depicted in (j), deformation of NGs, including under mild shear force, has been shown to reshape the flexible particle, and NGs were shown to successfully target PLVAP at the caveolar entrance in mouse lungs.

To quantify the engagement of targeted particles with PLVAP, ¹²⁵I-labeled NGs were incubated in a suspension of bEnd3 cells on ice. A trace amount of NGs and an excess of cells were used to best assure that a significant proportion of successfully targeted particles would bind to the cells. After 1h, 150nm PLVAP-targeted NGs precipitated with the bEnd3 cells in significantly greater proportions than IgG NGs (n=5) (Figure 3c). 300nm NGs also specifically targeted to PLVAP on bEnd3 cells at 4°C (n=5) (Supplementary Figure 7a, 8).

Noting that PLVAP on bEnd3 cells is not exclusively localized to membrane domains obscured in caveolae, polystyrene NPs were also targeted to PLVAP on bEnd3 cells on ice (n=5).^[44] Under these conditions, 150nm polystyrene NPs bound to PLVAP similarly to 150nm NGs (Figure 3d). Likewise, both 150nm and 350nm polystyrene NPs were capable of targeting to PLVAP on bEnd3 cells (Supplementary Figure 7b, 8). On the surface of transformed endothelial cells, where PLVAP is found outside sterically constrained domains, both NGs and rigid counterpart anti-PLVAP particles can adhere to PLVAP.

Targeting Nanogels to Caveolar PLVAP in the Lungs

Polystyrene NPs and NGs coated with PLVAP antibody were tested in mice to evaluate targeting to the lungs, where PLVAP is specifically associated with introits of caveolae.^[22] At 30 minutes after IV bolus injection, PLVAP-targeted ~150nm NGs (n=7) targeted the lungs (Figure 3e), at ~40% levels observed for free PLVAP antibody (n=3) (Supplementary Figure 9) and with ~10-fold advantage over IgG NGs of equivalent size (n=5). Analogous polystyrene particles were not able to target PLVAP in the lungs (n=3) (Figure 3e, Supplementary Figure 10a). ~300nm PLVAP-targeted NGs (n=5) also demonstrated ligand-dependent uptake in the lungs, with ~3-fold advantage over ~300nm IgG NGs (n=5). No lung targeting was observed for analogous 350nm PLVAP-targeted polystyrene particles (n=3) (Figure 3f, Supplementary Figure 10b).

When considering the clearance of each type of NG from the blood pool, where larger NGs cleared more rapidly, the distinction between targeting with 150nm NGs and 300nm NGs becomes less profound; 150nm PLVAP-targeted NGs had a lung:blood ratio of 6.14 and 300nm PLVAP-targeted NGs had a ratio of 3.69 (Supplementary Figure 11). The PLVAP:IgG specificity ratio for the clearance-normalized parameter was 19.17 for 150nm NGs and 4.03 for 300nm NGs or 9.73 for 150nm NGs and 2.72 for 300nm NGs for lung uptake without consideration of blood clearance (Figure 3g, 3h). No significant PLVAP-conferred lung specificity was observed for polystyrene particles of either tested size, even considering clearance in the calculation (Figure 3g, 3h). As measured by specificity index, 150nm NGs targeted the lungs at ~45% levels observed for free PLVAP antibody (Figure 3h).

Finally, towards demonstrating that the failure to target PLVAP was not unique to aldehyde-latex NPs or previously reported data with poly(4-vinylphenol) NPs,^[25] we conjugated PLVAP antibody or IgG to PLGA NPs.^[45] For ~400nm PLGA NPs (Supplementary Figure 12a), nearly identical lung uptake was observed for PLVAP-targeted and IgG PLGA NPs (Supplementary Figure 12b, 12c). By clearance-normalized lung specificity index, PLGA NPs exhibit a value of 1.08, compared to 4.03 for NGs and 1.19 for PS NPs of similar size.

Discussion

The failure of large (>100nm) rigid NPs to target PLVAP in the lungs, previously reported and confirmed here, comports with the theory of caveolae as size-restrictive gatekeepers for certain endocytosis and transcytosis pathways (Figure 3i).^{[11,18,25,27]} Our targeting data with mechanically flexible NGs adds an exception to this theory, providing evidence that larger objects targeted to PLVAP can access the mouths of caveolae if they can deform to present an aspect smaller than the caveolar neck. Indeed, AFM morphological analysis shows that NGs can spread during adhesion to a height under half their hydrodynamic diameter. Further data indicate that NGs have a morphology that is responsive to shear in flowing buffer. Under physiological shear, NGs elongated and accessed porous environments in a manner that would be expected for particles of <25nm diameter. Taken together, these data suggest NGs in the vasculature may be capable of accessing sterically concealed cell surface features by deforming and, in effect, exploring the “nooks and crannies” of the vascular surface (Figure 3j).

Ultimately, the goal of targeting to PLVAP with large NPs is more robust and specific drug delivery with the potential to impact inflammatory pathways affected by caveolar signaling.^{[9,10,14-17,44]} The current study focuses on basic modulation of nanomaterials properties with an eye towards accessing a cell surface feature previously thought inaccessible. Our demonstration of large particles accessing PLVAP, specifically, obviates studies that introduce large NGs or other similarly flexible NPs^[32] containing drugs, including mediators of inflammation, to this targeting route.

Towards this end, previous work has demonstrated that lysozyme-dextran NGs in particular, and NGs in general, may be versatile drug carriers.^[34,36] Osmotic loading of dextran NGs with dexamethasone indicated that small molecule drugs can be transiently retained in the porous hydrophilic volume of the NG’s dextran shell.^[36] Indeed, we have explored loading of NGs with a variety of small molecule drugs, finding that the catalase mimetic EUK-134, the corticosteroid dexamethasone, the NADPH-oxidase inhibitor MJ33, and the tyrosine kinase inhibitor imatinib mesylate can be osmotically loaded at a capacity of approximately 5% NG mass with approximately 50% loading efficiency (Supplementary Figure 13). Delivery of protein therapeutics with NGs is also available, given the facile nature of protein conjugation to NGs.^[35,37] This study demonstrates that different macromolecules (antibodies) can be conjugated to NGs, but we have also evaluated loading of NGs, in the absence of conjugation, with superoxide dismutase at a capacity matching that observed with small molecules (Supplementary Figure 13). Anti-inflammatory drug delivery to pulmonary endothelium, including in TLR4-driven inflammatory conditions that may be associated with caveolar signaling, has shown promise in previous animal studies.^[44,46] A larger quantity of anti-inflammatory cargo, still delivered specifically, may be enabled by the mechanical deformability strategy explored here for large NPs.

We have preliminarily evaluated the role of lipid rafts in internalization of NGs in bEnd3 cells and the data indicate that targeted NGs may be taken up in endothelial cells via caveolae. As to whether or not flexible nanoparticles can exploit the potential of caveolae as a conduit to the parenchyma, further work is needed to more specifically identify the disposition of NGs and their cargo after engagement with PLVAP. In vivo, it is not known whether the NGs are deposited in endothelial cytosol, the lung parenchyma, or parts unknown. Noting previous work on the disposition of PLVAP in lungs,^{[21,22,26,27]} tracing of in vivo cellular and subcellular localization of PLVAP-targeted NGs in future work may provide additional insight into the transport behavior of this unique endothelial surface marker. Unique transport may also be explored with the limited alternative options for markers associated with caveolae. Aminopeptidase P, for instance, has been demonstrated as a caveolar target with a role in transcytosis in the lungs.^[18]

Beyond delivery to PLVAP and caveolae, our results with deformable NGs indicate new dimensions to the role of mechanical properties in design of nanomaterials for drug delivery. Several advantages of low modulus NCs as targeted delivery tools have been identified. Soft NPs have prolonged circulation times compared to hard counterparts.^[30,47] Highly flexible polymeric nano- and microgels have been able to integrate in flexible in vivo environments like solidifying clots.^[31,48] And softer NPs may have advantages in affinity targeting, due to ability to avoid non-specific cellular uptake (e.g. by macrophages) and to spread on target surfaces.^[30,31,49] However, our data represents the first example to our knowledge of NP mechanical deformability specifically enabling targeting to topologies that are completely inaccessible to rigid counterpart particles; we have pointed out the possibility of squeezing 300nm of cargo into a 50nm hole.

Experimental Section

Nanoparticle Synthesis

Lysozyme-dextran NGs were synthesized as described.^[36] Briefly, 1:1 mol:mol preparations of chicken egg white lysozyme (Sigma) and 70 kDa dextran from Leuconostoc ssp. (Sigma), ~70 kDa Rhodamine B isothiocyanate-dextran from Leuconostoc ssp. (Sigma), or ~70 kDa Fluorescein isothiocyanate-dextran from Leuconostoc ssp. (Sigma) were prepared in water and adjusted to pH 7.1 by .1 N sodium hydroxide. The solutions were lyophilized and the dry powder was reacted at 60°C over a saturated KBr solution in a desiccator for 24h. The reacted powder was dissolved in water at 5 mg mL⁻¹ and the solution pH was adjusted to either 10.70 or 11.34 with .1 N sodium hydroxide, prior to gelation at 80°C for 30 min. After synthesis, DLS (Malvern Instruments) was performed for .02 mg mL⁻¹ solutions of NGs in water. NGs were conjugated to antibodies following periodate oxidation of the dextran component, as described.^[36] NGs were concentrated and resuspended to initial volume in a 1:1:2 (vol:vol:vol) solution of .01 M NaIO₄, 25% NaCl, and water, and reacted for 72h at room temperature prior to removal of NaIO₄ and NaCl via two sequential centrifugations against a 10 kDa cutoff concentrator (EMD-Millipore), followed by resuspension to initial volume in water. Oxidized NGs (presenting aldehyde groups) were reacted with antibody (presenting primary amines) overnight at 4°C in a 1:15 (mass:mass) antibody:NG solution. Excess antibody was removed by twice centrifuging NGs at 16000xg for 15min and removing natant. DLS was performed as above for antibody-NG complexes. Nanosight (Malvern) tracing of antibody-NGs following incubation in platelet poor plasma at 0.5 mg mL⁻¹ and at 37°C indicated stability of NG size. Quantity and efficiency of antibody addition to NGs was confirmed by tracing ¹²⁵I label on added antibodies. For experiments tracing antibody-NG internalization in cells via avidin stain, antibodies were tagged with Biotin N-hydroxysuccinimide ester prior to conjugation to NGs.

Purchased aldehyde-presenting latex NPs (Life Technologies) were incubated with antibody at 1:15 (mass:mass) antibody:NP for 1h at 4°C. Excess antibody was removed by twice centrifuging NPs at 10000xg for 12min and removing natant. Poly(D,L-lactide)co-glycolide (PLGA) double emulsion polymeric NPs were prepared and conjugated to antibody as described in the literature.^[44] NP size was determined by DLS before and after antibody coupling. For Nanosight tracing of antibody-NP stability in plasma, fluorescein-tagged carboxylate presenting polystyrene NPs were conjugated to Ab over 3 hours at 4°C and at 1:15 (mass:mass) antibody:NP following 3 minutes sulfo-NHS exposure and 15 minutes EDCI exposure to carboxylate-NPs.

To produce antibody for PLVAP-targeted particles, hybridoma secreting rat anti-mouse PLVAP IgG2a, clone MECA32, was obtained from the Developmental Studies Hybridoma Bank (developed under the auspices of the NICHD and maintained at the University of Iowa, Department of Biology). Cells were grown in PFHM-2 and antibody to murine PLVAP was purified using Protein G-Sepharose 4 Fast flow. Control particles were prepared with rat IgG2a isotype control (Invitrogen).

Evaluation of drug loading in and release from NGs was performed as previously described.^[36] Oxidized NGs at 5 mg mL⁻¹ concentration were incubated in a reservoir of small molecule drugs at .5 mg mL⁻¹ concentration for two days at 37°C. For loading of NGs with the antioxidant enzyme superoxide dismutase (SOD), non-oxidized NGs were incubated with the drug in order to exclude conjugation of SOD to the NGs. Following small molecule drug loading, NG-drug solutions were centrifuged against molecular weight cutoff 10 kDa concentrators to remove small molecule drugs not associated with NGs. Following SOD loading, NG-drug solutions were pelleted at 16000xg to remove free SOD. Drug-loaded NGs were resuspended to initial solution volume and subsequently filtered or centrifuged at various times after resuspension to assess the time course of drug release into buffer. Solutions of free small molecule drugs released from NGs were reserved for HPLC quantification. Briefly, eluted drugs were passed through a C18 column with a 60:40 acetonitrile:water and.1% trifluoroacetic acid mobile phase. Optical absorption was monitored at 382nm for EUK-134, at 240nm for dexamethasone, at 270nm for MJ33, and at 230nm for Imatinib Mesylate. For eluted SOD, concentrations were evaluated through inhibition of cytochrome C oxidation in the presence of xanthine and xanthine oxidase. The time course of drug release was modeled according to an extended Langmuir model^[36] as (with a, b, and c as fitting parameters):

$$\%DrugReleased=\frac{ab{x}^{(1-c)}}{1+b{x}^{(1-c)}}$$

Atomic Force Microscopy

For morphological analysis, silicon wafers or glass slides were treated with piranha solution for 30 min at 80°C, treated in a UVO-cleaner for 30min, and exposed to 1 mL separately contained (3-Aminopropyl)Triethoxysilane (APTES) in a sealed vessel at 70°C for 6h to facilitate APTES vapor deposition. APTES-coated wafers were submerged in solutions of periodate-activated NGs (.5 mg mL⁻¹) and NaBH₃CN (3 mg mL⁻¹) overnight, followed by threefold rinsing with water. Water droplet contact angle measurements verified modification of Silicon with APTES and with NGs. NG monolayers were characterized with atomic force microscopy (Asylum), employing contact mode imaging with a triangular Au/Pt probe with a 700 pN nm⁻¹ spring constant. NGs were maintained in water at all times and aqueous AFM protocols were employed. For analysis of NG height, images were processed in Image SXM 198 and ImageJ. Borders of individual particles were identified manually and maximum particle heights were determined relative to a mean background height determined for a 10nm square field of view adjacent to each particle. For presentation in Fig. 1c, but not for analysis, the substrate was flattened by interpolative smoothing as implemented in Asylum software.

For Young’s modulus mapping, the above procedure for NG immobilization was adapted without piranha solution treatment of wafers, with overnight exposure of wafers to APTES vapor, and with exposure to 10 mg mL⁻¹ NGs without NaBH₃CN. 600nm SiO₂ colloid tips with 219 pN nm⁻¹ spring constant were used to obtain force-indentation maps across 2.5×2.5μm fields at 2500 point resolution. Young’s moduli were obtained via Hertzian modeling as implemented in Asylum software and moduli reflecting NGs were selected by thresholding according to height retrace in the same field of view.

Quartz Crystal Microbalance

Au/SiO₂ quartz crystal microbalance (QCM) sensors (Biolin Scientific) were coated with APTES, as described for Silicon wafers above, after UVO cleaning. After heating the sensor to 37°C, baseline measurements were obtained for the dry sensor and the sensor under exposure to DI water flowing at 7 μL min⁻¹. (QCM-D, Biolin Scientific). Solutions of NGs or PS NPs in water (.5 mg mL⁻¹), were introduced to the sensors after acquisition of solvent baseline measurements and measurements of resonant frequency and dissipation factor were recorded over ~24h, followed by 2h rinsing with water. Modeling of raw data to obtain deposited masses, shear moduli, and film viscosities was completed with QTools (Biolin Scientific) viscoelastic modeling tool and via application of the Sauerbrey Equation to third, fifth, and seventh harmonic resonant frequencies.^[37]

Ektacytometry

NGs or polystyrene spheres at 10 mg mL−1 were suspended in 5.5% polyvinylpyrrolidone (FW 40kDa, Sigma) solution. 500 μL of nanoparticle suspension was added to ektacytometer cassettes with 6mmx0.2mm channels (RheoMeditech, Korea). Nanoparticle suspensions were extruded through microfluidic channels under air pressure, with diffraction patterns being generated at regular intervals as the applied pressure varied. Each diffraction pattern generated by the NG or polystyrene particle suspensions was analyzed as a best-fit ellipse determined in ImageJ. Asymmetry in the ellipse was quantified by calculation of the elongation index, (L-W)/(L+W), where L is the length of the horizontal axis of the ellipse and W is the length of the vertical axis. Elongation index vs. applied stress curves were fitted via the Streekstra-Bronkhorst model for elongation of cells in flow:

$$EI=\left\{(E{I}_{max}-E{I}_{min})\ast \frac{(\frac{s}{s{s}_{half}}{)}^m}{1+{\left(\frac{s}{s{s}_{half}}\right)}^m}\right\}+E{I}_{min},$$

where S is the applied stress, EI is the elongation index, EI_min is the best fit elongation index at zero stress, EI_max is the best fit elongation index at infinite stress, SS_half is the applied stress at which the elongation index reaches half its maximum, and m is a fitting parameter accounting for variations in slope of the curve.^[40]

Nanoparticle Filtration

Fluorescent nanoparticle suspensions were prepared at .5 mg mL⁻¹ in water and pressed through PTFE or PVDF syringe filters (EMD-Millipore) at volumetric flow rates determined by syringe pump. The number of pores per filter was estimated by dividing the filter area by the stated cross-sectional pore area and volumetric flow rate in each pore was taken as the total volumetric flow rate in the syringe, as set by syringe pump controls, divided by the estimated number of pores per filter. The standard equation for laminar flow through a cylinder, $\gamma =\frac{4\mathit{Q}}{\pi r^3}$, was used to relate shear rate for passage through pores, γ, to volumetric flow rate per pore, Q, by way of pore radius, r. For smaller pore sizes (1, 5, and 10nm), pore sizes were interpolated based on theoretical globular protein size, given molecular weight cutoff values for the filters (10, 50, and 100 kDa). 0.5 mL of eluent was collected from each filter and fluorescence in the eluent was compared to a fluorescence standard curve extrapolated from the .5 mg mL⁻¹ nanoparticle stock to determine relative quantity of nanoparticles eluted through the filters. Effective particle diameters were obtained according to Kirtane’s model for sieving coefficients,^[42] expressing the relationship between fraction of particles filtered f, effective particle diameter d_particle, and pore diameter d_pore.

$$\begin{array}{c}\hfill a=\frac{d_{particle}}{d_{pore}}\hfill \\ \hfill Ifa\ge 1.0,f_{filtered}=0.00\hfill \\ \hfill Ifa<1.0,f_{filtered}=(1-a{)}^2\times \left(\frac{1+3.867a-1.907a^2-0.834a^3}{1+1.867a-0.741a^2}\right)\hfill \end{array}$$

Cell Culture

bEnd3 murine endothelial cells were obtained from ATCC (Manassas) and were grown in DMEM with 10% fetal bovine serum and antibiotic/antimycotic. bEnd3 cells were grown on 12mm glass coverslips treated with 1% gelatin. For internalization studies, after treatment with NGs, cells were washed, fixed, and treated with Alexa Fluor 488 goat anti-rat IgG or with Alexa Fluor 488 avidin. For actin co-localization experiments, cell samples were fixed with ice-cold 2% paraformaldehyde for 15min and permeabilized with 0.2% Triton X-100 for 15min prior to staining with Phalloidin. For lipid raft inhibition, cells were treated with filipin (3 μg mL⁻¹) 30min prior to introduction of NGs. For immunoreactivity experiments, cells were fixed and trypsinized before suspension in PBS on ice. Samples of 10⁷ cells were incubated with 2.5 μg of NGs or PS NPs.

After staining, samples were mounted using ProLong Gold Antifade Reagent with DAPI (Molecular Probes, ThermoFisher). For initial internalization and targeting studies, fluorescence images were acquired using a Nikon Eclipse TE2000-U fluorescence microscope equipped with a Plan Apo x40/1.0 oil objective (Nikon). Microscope controlling and image processing were carried out using Image-Pro Plus 4.5.1.27 (Media Cybernetics). Localization studies were performed on a confocal laser scanning microscope Leica TCS-SP8 (Leica).

In Vivo Nanoparticle Targeting

Animal experiments were performed according to protocol approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Pennsylvania. ¹²⁵I-IgG was conjugated to nanoparticles at 1:10 mol:mol with either MECA32 anti-PLVAP or unlabeled IgG in order to prepare ¹²⁵I-labeled nanoparticles. After noting the total ¹²⁵I activity in prepared particles, 200 μg NP doses were injected IV in normal C-57BL/65 mice (The Jackson Laboratory) via jugular vein. After 30min, mice were sacrificed by terminal exsanguination and organs were harvested, rinsed with saline, blotted dry, and weighed. Tissue radioactivity in organs and ~300 μl samples of blood was determined in a Wallac 2470 Wizard gamma counter. The percentage of injected ¹²⁵I dose in each organ was noted to prepare biodistribution charts.

Statistical Analysis

Where noted, significance indicates p<.01 as determined either by two-tailed paired t-test. All presented error bars indicate standard error of the mean.