Tuning Targeted Liposome Avidity to Cells via Lipid Phase Separation

Abstract

The addition of both cell-targeting moieties and polyethylene glycol (PEG) to nanoparticle (NP) drug delivery systems is a standard approach to improve the biodistribution, specificity, and uptake of therapeutic cargo. The spatial presentation of these molecules affects avidity of the NP to target cells in part through an interplay between the local ligand concentration and the steric hindrance imposed by PEG molecules. Here, we show that lipid phase separation in nanoparticles can modulate liposome avidity by changing the proximity of PEG and targeting protein molecules on a nanoparticle surface. Using lipid-anchored nickel-nitrilotriacetic acid (Ni-NTA) as a model ligand, we demonstrate that the attachment of lipid anchored Ni-NTA and PEG molecules to distinct lipid domains in nanoparticles can enhance liposome binding to cancer cells by increasing ligand clustering and reducing steric hindrance. We then use this technique to enhance the binding of RGD-modified liposomes, which can bind to integrins overexpressed on many cancer cells. These results demonstrate the potential of lipid phase separation to modulate the spatial presentation of targeting and shielding molecules on lipid nanocarriers, offering a powerful tool to enhance the efficacy of NP drug delivery systems.

Graphical Abstract

1. INTRODUCTION

The physicochemical properties of nanoparticles (NPs) play an important role in their therapeutic efficacy when used in drug delivery systems. A critical step in NP administration is interaction with a target cell surface. NPs are often engineered with two classes of surface modifications that affect this interaction: targeting moieties to promote NP interactions with specific receptors and cell types,¹ and shielding polymers like polyethylene glycol (PEG) to increase in vivo circulation time and reduce the rate of biofouling in biological systems.² While many intrinsic properties of NPs affect ligand-receptor binding, such as stiffness,^3–5 ligand mobility,^6–8 and surface roughness,⁹ an important, often less explored feature is the spatial presentation of conjugated ligands on the NP surface. Ligand density is an important parameter that controls bond multivalency when targeting a NP to a cellular receptor.^10–13 This density affects the overall binding strength of the NP to the target cell, or NP avidity, and ultimately contributes to the physiological fate of the NP and overall efficacy of drug delivery. Avidity is affected by the binding affinity of the individual ligands, the valency of the nanoparticle, and the spatial arrangement. As a result, optimizing the spatial arrangement of both targeting ligands and shielding polymers is important for modulating the NP avidity to the target cell receptor.¹¹

Cells naturally modulate receptor avidity by dynamically rearranging membrane proteins and lipids in the cell membrane. Through the formation of lipid rafts, or highly ordered and cholesterol-rich regions of the cell membrane, cells can partition proteins into a lipid domain and exclude other proteins, which can control receptor density and function. This clustered arrangement of receptors is a common feature of immune cell signaling¹⁴ and facilitates not only receptor binding by significantly altering binding kinetics but also subsequent intracellular and intraorganelle signaling by lowering the activation threshold of receptors.^15,16 In addition, lipid rafts can exclude sterically hindering proteins and inhibitory signaling proteins that affect receptor binding and activation.¹⁷

Analogous to cellular membranes, the arrangement of ligands on a nanoparticle surface can also tune nanoparticle binding to target molecules. The clustered arrangement of ligands can enhance multivalent cooperativity^18–20 by facilitating the rapid recapture of unbound or dissociating ligands by cell receptors. As a result, the enhanced binding of NPs to cells can lead to more efficient activation of cellular receptors through subsequent cluster-dependent signaling.^21–24 One way to control ligand spatial presentation on nanoparticles is through lipid phase separation. By mixing saturated and unsaturated phospholipids with cholesterol, bilayer vesicles can undergo phase separation between the saturated liquid ordered (l_o) and unsaturated liquid disordered (l_d) phases, in a similar manner to cellular lipid rafts.²⁵ Harnessing lipid phase separation to control ligand presentation on NPs has previously been used to enhance delivery of therapeutic cargo to cells^26,27 and increase apoptotic signaling in target cells.²⁸ Because many ligand–receptor binding systems are dependent on the surface concentration of the receptor as well as the interligand spacing on the targeted surface,^29,30 we wondered if lipid phase separation can be more broadly used to tune lipid nanoparticle avidity to cells for other types of binding interactions. Understanding the capacity of phase separation to modify different ligand–receptor interactions would aid in the design of NPs for targeted drug delivery.

In this study, we demonstrate an approach for assembling lipid bilayer-based NP vesicles that display both targeting ligands and shielding PEG polymers with varying surface densities. By varying composition and lipid chain saturation, we assembled vesicles with varying sizes of membrane domains, which can spatially control localization of both ligands and PEG into distinct or colocalized domains. Using several phase-separated lipid compositions and ligands of varying binding strength, we identified key design principles for controlling and optimizing the avidity of vesicles to cells by controlling the location of both targeting and sterically inhibiting molecule on the NP surface. We then applied these design principles to tune the performance of a therapeutically relevant ligand in cancer research: integrin-binding tripeptide sequence Arg-Gly-Asp (RGD), which is known to be dependent on inter-ligand spacing.^29,31 Our modular and readily accessible approach presents a powerful tool for using lipid phase separation to tune NP avidity without changing overall concentration of ligand-bound lipid or engineering new molecules with different binding affinities. This technique can open the door to future technologies, including optimizing binding to target cells with lower overall concentrations of ligand per particle, as well as designing stimuli responsive NPs that can harness phase separation to dynamically change their avidity to a target cell based on environmental cues.

2. MATERIALS AND METHODS

2.1. Materials.

Lipids including 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), 1,2-distearoyl-sn-glycero-3- phosphocholine (DSPC), cholesterol (Chol), 1,2-dioleoyl-sn-glycero-3-[(N-(5-amino-1-carboxypentyl)iminodiacetic acid)succinyl] (nickel salt) (DGS-NTA-Ni), 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (18:0 PEG2000-PE), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy-(polyethylene glycol)-2000] (ammonium salt) (18:1 PEG2000-PE), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(lissamine rhodamine B sulfonyl) (18:1 Rho), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(7-nitro-2-1,3-benzoxadiazol-4-yl) (ammonium salt) (18:1 NBD), 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[4-(p-(cysarginylglycylaspartate-maleimidomethyl)cyclohexane-carboxamide] (sodium salt) (DSPE-RGD), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-[4-(p-(cysarginylglycylaspartate-maleimidomethyl)cyclohexane-carboxamide] (sodium salt) (DOPE-RGD), and 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(cap biotinyl) (sodium salt) (18:1 Biotinyl Cap PE) were purchased from Avanti Polar Lipids. 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-(7-nitro-2-1,3-benzoxadiazol-4-yl) (16:0 NBD) was purchased from Invitrogen. Phosphate-buffered saline (PBS) tablets were obtained from Sigma-Aldrich. Cell media components, RPMI, DMEM, fetal bovine serum (FBS), and penicillin/streptomycin were purchased from Thermo Fisher (Gibco). Streptavidin magnetic beads, calcein AM, TrypLE, Trypsin–EDTA (0.25%), and NucBlue Hoechst 33342 were purchased from Thermo Fisher. RGD peptide (GRGDNP) was purchased from MedChemExpress. Triton-X-100 (LabChem) was purchased from Fisher Scientific.

2.2. Large Unilamellar Vesicle Formation.

Thin film hydration was used to prepare large unilamellar vesicles (LUVs). Lipid compositions for each vesicle formulation are provided in Tables S1–S6. Lipid formulations in chloroform solution were dried down under a nitrogen stream to create a thin film on the bottom of a glass tube. Lipid films were placed in vacuum for 2 h to remove excess chloroform. Films were rehydrated with 1× PBS (~290 mOsm) and incubated overnight at 60 °C. Vesicles were vortexed and extruded using an Avanti mini extruder through a 100 nm polycarbonate membrane for 7 passes. Vesicles containing saturated phospholipids were heated to 70 °C on a hot plate during the extrusion process. Vesicles were characterized with regard to size and zeta potential by dynamic light scattering (DLS) using a Malvern Zetasizer Nano. Phase separation was quantified via a Förster resonance energy transfer (FRET) assay (see Supporting Information, Experimental Section).

2.3. Cell Culture.

Jurkat and U937 cells were obtained from the American Type Culture Collection (ATCC) without further authentication. K562 cells obtained from ATCC were gifted by the Leonard Lab (Northwestern University). MDA-MB-231 cells were gifted by the Mrksich Lab (Northwestern University). Jurkat, U937 and K562 cells were cultured in RPMI 1640 supplemented with 10% FBS and 1% penicillin–streptomycin. MDA-MB-231 were cultured in DMEM supplemented with 10% FBS and 1% penicillin–streptomycin.

2.4. Vesicle-Cell Binding Studies.

Vesicle binding to cells was characterized with flow cytometry. For dose response studies, assays were performed in a 96-well round-bottom plate. For single concentration studies, assays were performed in Eppendorf tubes. Cells were suspended in flow buffer (PBS containing 1% FBS) and then seeded at a concentration of 100,000 cells per well/tube at 100 μL. Test vesicles were added to each well/tube at the specified concentrations and incubated with cells for 30 min at room temperature. Cells were washed 2 (tubes) to 3 (plates) times, where cells were spun down at room temperature in a centrifuge for 3 to 5 min until a pellet formed in each well/tube, and the supernatant was aspirated out and replaced with fresh flow buffer. Flow cytometry was performed on cells using a BD Fortessa LSRII using 550 nm excitation laser and 582/15 nm emission for rhodamine-labeled vesicles and 488 nm excitation laser 530/30 nm emission with 505 nm long pass filter for NBD-labeled vesicles. At least 5000 events (for plate-based) to 10,000 (for tube-based) flow cytometry measurement events were selected by gating cell population based on forward vs side scatter (FSC-A vs SSC-A). Doublets were ruled out by gating based on FSC-A vs FSC-H (Figure S1). Flow cytometry data including median fluorescence intensity (MFI) was analyzed using FlowJo software (Version 10.8.1). Fold change was calculated as the MFI of ligand-containing vesicle/MFI of the unconjugated control vesicle within a replicate (paired) and the means of at least 3-fold changes were calculated.³² For studies analyzing vesicle-binding over time, cells were incubated with an excess amount of 3:1 Ni-NTA vesicles (10 mM) for 30 min at room temperature, washed, resuspended in full media, and analyzed with flow cytometry at time points indicated.

For adherent MDA-MB-231 cells, a slightly altered protocol was used. Cells were plated in either 6 well plates (400,000 cells/well) or 24 well plates (100,000 cells/well) and allowed to adhere overnight. Then, 500 μM vesicles were added to cells in full media for 4 h in a 37 °C incubator. After incubation, cells were detached with PBS supplemented with either 10 mM EDTA-PBS or TrypLE. Cells were then analyzed for vesicle binding using flow cytometry.

2.5. Vesicle-Bead Binding Assay.

Vesicle binding to streptavidin beads was also characterized via flow cytometry. 2 μL of Pierce streptavidin magnetic beads was added to PBS, and test vesicles were added to each tube at 500 μM lipid and incubated with beads in a 100 μL reaction for 30 min at room temperature. Beads were drawn down to the bottom of the solution with a magnet and the supernatant was replaced with fresh PBS, repeated three times. 10,000 flow cytometry measurement events were selected by gating bead population based on forward vs side scatter (FSC-A vs SSC-A), selecting for larger beads to rule out dust and debris. Doublets were eliminated by gating based on FSC-A vs FSC-H (Figure S2). Flow cytometry data including MFI was analyzed using FlowJo software.

2.6. Confocal Microscopy.

Cells were stained with 500 ng/mL calcein AM for 30 min, washed, and resuspended in full media. Vesicles were added to the cells at 500 μM final lipid concentration to 100,000 cells in 100 μL of 50% complete media, 50% PBS and incubated for 2 h at 37 °C for binding and internalization. After 2 h, cells were washed 2× and resuspended in full media. NucBlue was added to the cells and imaged using confocal microscopy (Nikon Eclipse Ti).

2.7. Cell Internalization Studies.

Vesicle-cell internalization studies were performed similarly as previously described but with slight alterations. Vesicles labeled with NBD were used specifically for these studies. Jurkat cells were seeded at 200,000 cells/tube in 200 μL, and 500 μM NBD-labeled vesicles were added to each tube. Vesicle binding was performed at 4 °C and 37 °C for 1 h to disable or enable internalization, and the solution was then washed 2× with PBS. After washing, 100,000 cells (100 μL) were incubated with 100 μL of trypsin–EDTA (0.25%) for 10 min. Cells in trypsin were then washed 2× with flow buffer, and all tubes were resuspended in 300 μL of flow buffer. Cells were then analyzed with flow cytometry. For trypan blue quenching, after untreated cells were analyzed via flow cytometry, 20 μL of trypan blue was added to the cells, vortexed, and then analyzed again via flow cytometry.

2.8. Statistical Analysis.

All significance tests, including 2-way ANOVA, multiple comparisons, and non-linear fits, were performed using Prism (GraphPad). For fold change data, all statistical analyses were conducted on the log-normal distribution [log(fold change)] in order to assume normal distribution.³²

3. RESULTS

3.1. Phase Separation Increases the Avidity of a Model Weak-Binding Ligand.

To probe how ligand density affects liposome-cell binding, we prepared 100 nm vesicles with homogeneously distributed and phase-separated lipids. Several different vesicle formulations were prepared with 0:1, 1:1, 2:1, 3:1, and 1:0 ratios of 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC):1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), a saturated and unsaturated phospholipid, respectively, and with a constant amount of cholesterol (30 mol %) and DSPE-PEG2000 (1 mol %) (Table S1). These formulations have been previously tested and characterized by our group using FRET analysis (see the Supporting Information, Experimental Section) to confirm production of phase-separated vesicles. Vesicles with 0:1 and 1:0 DSPC:DOPC that contain only DOPC and cholesterol or DSPC and cholesterol have a uniform distribution of the component lipids, while vesicles with ternary mixtures of DOPC, DSPC, and cholesterol show more phase separation as the ratio of DSPC:DOPC increases.²⁸

To study whether phase separation can enhance vesicle avidity through ligand clustering, we chose to investigate Ni²⁺-nitrilotriacetic acid (Ni-NTA) as a model weak-binding ligand. Ni-NTA is a positively charged molecule commonly used in protein purification due to its interaction with polyhistidine tags and has also been used to conjugate His-tag proteins to vesicles.^33–35 Most proteins will adsorb to Ni-NTA with low affinity, however, due to histidine interactions. Additionally, high Ni-NTA concentration on vesicles has been shown to increase nonspecific binding to cells,^36,37 which we suspect is due to the low affinity interactions between Ni-NTA and membrane proteins expressed on cell surfaces.^38,39 When added to the vesicle formulation, DGS-NTA(Ni), an unsaturated phospholipid covalently bound to Ni-NTA, should localize to the smaller, unsaturated DOPC domain, resulting in a higher local density of DGS-NTA(Ni) on the vesicle surface. We hypothesized that the higher density of DGS-NTA(Ni) would increase vesicle avidity (Figure 1a) to cells via these low affinity interactions.

Figure 1.

Phase separation enhances Ni-NTA vesicle avidity to cells. (a) Schematic of the vesicle compositions used in these experiments. By varying the ratio of DSPC:DOPC, the local surface density of DGS-NTA(Ni) can be increased, which increases avidity to cells. (b,c) Flow cytometry histograms of vesicle avidity to (b) Jurkat and (c) U937 cells at 500 μM. Black line plot represents unstained cells, gray plot represents unconjugated vesicles, and colored plots represent NTA-conjugated vesicles. (d,e) Fold change of median fluorescence intensity of (d) Jurkat and (e) U937 cells incubated with Ni-NTA vesicles over no-NTA control vesicles. Error bars represent one-sided SEM for three biological replicates conducted on different days (n = 3). Significance testing represents results of two-way ANOVA over concentration and compositions with Dunnett’s multiple comparison tests with respect to DOPC. *p < 0.05, **p < 0.01, ****p < 0.0001.

We incubated fluorescent phase-separated vesicles with and without 1 mol % DGS-NTA(Ni) with Jurkat and U937 cells, which are human T-cell and monocyte lines, respectively. Because 1 mol % Ni-NTA did not increase nonspecific binding to Jurkat and U937 cells when uniformly present on DOPC vesicles, we chose this condition to study the effect of increasing the local density of DGS-NTA(Ni) through phase separation on cell binding (Figure S3). Vesicle binding was analyzed via flow cytometry, and median fluorescence intensity (MFI) values for cells incubated with DGS-NTA(Ni) vesicles were normalized to background MFI values from corresponding cells incubated with non-functionalized vesicles to study the effects of DGS-NTA(Ni) clustering while ruling out any differences in avidity due to lipid composition.⁴⁰ We observed that 2:1 and 3:1 DSPC:DOPC phase-separated vesicles, the compositions we expected to have the highest surface density of Ni-NTA ligands, enhanced DGS-NTA(Ni)-mediated cell binding. However, inclusion of DGS-NTA(Ni) did not increase vesicle binding when added to 1:1 DSPC:DOPC and uniform DOPC vesicles (Figure 1b,c), indicating that a certain threshold of ligand density is required to observe enhanced avidity. We also tested vesicle binding to U937 cells and observed similar trends, indicating that enhancing DGS-NTA(Ni)-mediated cell binding via phase separation is not a phenomenon limited to one cell type (Figure 1d,e). Avidity effects were prominent at concentrations of 500 μM lipid vesicles and higher, and therefore 500 μM was chosen for future studies. We expected that by increasing the ratio of DSPC:DOPC, we would increase the local concentration of Ni-NTA on the vesicle surface in a similar manner to when we increased the global concentration of Ni-NTA in vesicles (Figure S3). Previous work has shown that the Ni-NTA-polyhistidine interaction is a weak bond, which is not stable over time, especially in biological fluids.^33,34 We therefore also looked at 3:1 Ni-NTA vesicle binding to Jurkat over time in full media and saw that vesicles did indeed unbind from cells after 2 h (Figure S4). Together, this data demonstrates that lipid phase separation provides a route to enhance vesicle-to-cell avidity for a low affinity, charge-based interaction by increasing the clustering of DGS-NTA(Ni) lipids, and a critical threshold of phase separation is required to observe enhanced binding.

3.2. Phase Separation Can be Used to Modulate the Steric Effects of PEG on a Model Weak-Binding Ligand.

Next, we investigated how modulating the location of PEG on Ni-NTA-functionalized vesicles affects binding to cells. Besides improving nanoparticle stability and reducing biofouling, PEG molecules can undesirably shield ligands through steric effects.⁴¹ Cleavable PEG systems have been developed to take advantage of this effect by better exposing ligands after encountering a disease-specific stimuli.^42,43 Motivated by this shielding capacity, we hypothesized that segregation of PEG away from the ligand on the vesicle surface could improve ligand accessibility. To study the effect of PEG segregation on binding, we created Ni-NTA vesicles and incorporated varying ratios of unsaturated and saturated PEG lipids (DOPE-PEG and DSPE-PEG, respectively) at an overall constant PEG concentration of 1 mol % (Figure 2a). We added varying ratios of PEG lipids into both phase-separated Ni-NTA vesicles (3:1 DSPC:DOPC) and uniform vesicles (DOPC) and conducted binding studies (see Table S2 for lipid composition). In uniform vesicles, vesicle binding to cells was similar for all DOPE-PEG:DSPE-PEG ratios, consistent with our hypothesis that PEG lipids are evenly distributed across the vesicle surface (Figure 2b,c). As anticipated, in phase-separated vesicles, the presence of DOPE-PEG in the same phase as the Ni-NTA groups decreased vesicle avidity and replacing DOPE-PEG with DSPE-PEG to sequester the PEG molecules in a separate domain from the Ni-NTA groups enhanced vesicle binding (Figure 2b,c). Because the addition of PEG-modified lipids can dissociate lipid domains,⁴⁴ we conducted FRET studies to confirm that lipid domains existed at all PEG ratios (Figure S5, Table S5). Taken together, our results demonstrate that by simply changing the location of PEG on the surface of lipid vesicles, we can tune phase-separated vesicle avidity to target cells, independent of the ligand’s binding affinity. Modifying the location of PEG or other shielding polymers away from targeting ligands could be used in lieu of ligand engineering techniques to tune overall nanoparticle avidity.

Figure 2.

Vesicle avidity can be tuned by controlling the colocalization of PEG and Ni-NTA lipid. (a) Schematic depicting some of the vesicle compositions examined in this experiment. In uniform vesicles, changing the ratio of DSPE-PEG to DOPE-PEG does not affect the spatial distribution of PEG and does not affect vesicle avidity. In phase-separated vesicles, controlling this ratio modulates the degree of colocalization which can increase or decrease binding. (b) Flow cytometry histograms of uniform and phase-separated vesicle binding to Jurkat cells with varying ratios of DSPE-PEG:DOPE-PEG. Dotted line represents approximated median of uniform vesicle binding to cells. (c) Quantitative and statistical analysis of the flow cytometry data between phase-separated and uniform vesicles as well as within each type of vesicle with varying ratios of PEG lipids. Errors bars represent SEM for n = 3. Significance tests represent the results of two-way ANOVA using Tukey’s multiple comparison tests. ****p < 0.0001, *p < 0.05.

3.3. Phase Separation Has a Weaker Effect on Avidity in a Model Strong-Binding Ligand.

After studying the effect of lipid phase separation and PEG location on the avidity of a weak-binding, vesicle-bound ligand, we set out to investigate how this behavior compared with a strong-binding ligand. We used biotin–avidin interactions for these studies by measuring binding of biotinylated vesicles to streptavidin-coated magnetic beads. Uniform and phase-separated vesicles with 0.1 mol % unsaturated, biotinylated lipid (18:1 Biotinyl Cap PE) with varying ratios of DSPE- to DOPE-bound PEG were incubated with avidin-coated magnetic beads, and binding was measured via flow cytometry (see Table S3 for lipid compositions). Vesicle binding to avidin beads was enhanced when the vesicles were phase-separated compared to uniform vesicles, similar to the effect observed in most Ni-NTA vesicles (Figure S6). These results indicate that clustering through phase separation may enhance many different ligand–receptor systems across a wide range of binding affinities. Contrary to what we found with Ni-NTA, however, the ratio of DOPE-PEG to DSPE-PEG did not have a significant effect on the binding of biotin vesicles to avidin beads (Figure S6). One explanation for this is that the strength of the biotin–avidin interaction may be strong enough to overpower the steric effects of PEG. However, high molecular weight PEG has also been shown to have steric hindrance effects on biotin–avidin systems in some cases, including on biotinylated vesicles,^45,46 suggesting PEG location relative to targeting ligands may still modulate stronger binding protein interactions. Another possibility is that beads present molecular targets statically, while cell membranes present molecular targets dynamically in a fluid medium, which could explain the differences we saw between the binding studies.

3.4. Phase Separation Can Enhance Binding of Vesicles with a Therapeutically Relevant Ligand, RGD.

To demonstrate translational utility of this technology, we sought to confirm that phase separation can enhance cell-binding of vesicles functionalized with a therapeutically relevant ligand. The RGD peptide motif interacts with α₅β₁ and α_vβ₃ integrins, which are overexpressed on many cancers, and therefore RGD-functionalized liposomes have been studied extensively for drug delivery.⁴⁷ While the cyclic form of RGD is typically preferred due to its higher stability and affinity to integrin, we used the linear form for these studies as it is more readily available in the lipid-conjugated form. Furthermore, a recent study demonstrated giant unilamellar vesicles functionalized with linear RGD were able to bind Jurkat cells.⁴⁸ We therefore wanted to test whether phase-separated vesicles functionalized with RGD would enhance RGD-mediated binding (Figure 3a).

Figure 3.

Phase separation can enhance RGD-mediated cell binding to Jurkat and K562 cells in the l_d phase. (a) Schematic depicting the vesicle compositions explored in this experiment. In uniform DOPC vesicles, the lipid anchor on the PEG and the RGD does not affect their local surface density. In phase-separated vesicles, RGD can be localized to either the l_o phase (yellow) or the l_d phase (purple), depending on the lipid anchor. (b,c) Flow cytometry histograms of uniform and phase-separated DOPE- and DSPE-RGD vesicle binding to (b) Jurkat and (c) K562 cells as well as the no RGD controls. (d,e) Quantitative analysis of fold increase in MFI of (d) Jurkat and (e) K562 cells incubated with RGD vesicles over no RGD control vesicles. Error bars represent SEM for three (DSPE-RGD vesicles) to six (DOPE-RGD vesicles) biological replicates conducted on different days. Significance testing represents the results of two-way ANOVA with Tukey’s multiple comparison tests. *p < 0.05, **p < 0.01. (f) Confocal microscopy images of Jurkat and K562 cells treated with uniform and phase-separated DOPE- and DSPE-RGD vesicles. Cells are stained with the nuclear dye NucBlue (blue) and the viability dye calcein AM (green). Vesicles are doped with 0.1 mol % 18:1 Liss Rhod PE (red). Scale bar = 10 μm.

We created uniform and phase-separated vesicles with 10 mol % DOPE-RGD, confirmed phase separation using FRET studies (Figure S7), and tested binding to Jurkat and K562 cells, cell lines that both highly express the RGD-binding integrin α₅β₁.⁴⁹ In uniform vesicles, 10 mol % RGD vesicles exhibited minimal binding on both Jurkat and K562 cells compared to unfunctionalized control vesicles (Figure 3b–e). When RGD was concentrated into lipid domains, binding was greatly enhanced to Jurkat (~15 fold) and K562 (~5 fold) (Figure 3b–e). To verify that vesicle binding was due to specific interactions between vesicle-bound RGD and cellular receptors, we added a soluble RGD peptide that competitively binds to integrins. The addition of the soluble RGD peptides significantly reduced phase-separated vesicle binding, supporting the specificity of the vesicle-to-cell binding interaction (Figure S8). Interestingly, we did not see vesicle binding to Jurkat cells when RGD was uniformly distributed on the vesicle surface as previously reported, even at 10 mol % RGD.⁴⁸ The discrepancy with previous data could be due to differences in vesicle size as the previous study used larger vesicles (2 μm diameter), which have more RGD ligands and a larger surface for multivalent binding. In summary, phase separation enhances RGD-mediated vesicle binding and supports the applicability of phase separation as a tool to enhance interactions of nanoparticle-bound, therapeutically relevant ligands with target cells.

Notably, phase-separated RGD vesicles also exhibit higher background adhesion to both Jurkat and K562 cells compared to their uniform counterparts (Figure 3b,c). One explanation could be due to the PEG coverage. Phase separation of PEG results in certain regions of the cell membrane that are not blocked by PEG. Another reason could be the effects of lipid fluidity based on the composition.⁴⁰ Background adhesion with control vesicles was also much more prominent in K562 cells compared to Jurkat cells, which suggests that different cell types are more or less sensitive to these effects. Background adhesion can affect the specificity of a targeted therapeutic nanoparticle and hence would be an important consideration in the design of a phase-separated drug delivery vesicle system.

3.5. Lipid Saturation of RGD Conjugation Affects Cell Binding.

We also investigated whether clustering RGD lipids in the l_o phase could enhance RGD-mediated cell binding. In our previous experiments, the ligand was conjugated to an unsaturated lipid, and therefore the ligand should have localized into the l_d phase during phase separation. Therefore, we wanted to compare how concentrating the ligand into the l_o phase affected vesicle avidity compared to the l_d phase since the fluidity of the local membrane affects the mobility of the ligand and therefore has the potential to affect vesicle binding (Figure 3a).^6–9 To investigate this, we created uniform and phase-separated vesicles with DSPE-RGD, which should localize into the l_o phases of vesicles (Figure 3a and Table S4 for full composition). We observed that uniform vesicles containing DSPE-RGD and DOPE-RGD differentially bound to cells. In particular, uniform vesicles containing DSPE-RGD bound significantly stronger to Jurkat cells compared to uniform DOPE-RGD vesicles (Figure 3b,c), while in K562 cells, this effect was less pronounced (p = 0.0503, Figure 3d,e). The differences in integrin expression between these cells could play a role in the differences in binding, as different types of integrins might each have a unique optimal receptor spacing.⁴⁹ Additionally, phase-separated vesicles containing DSPE-RGD did not bind stronger than uniform DSPE-RGD vesicles in either cell type. FRET analysis to confirm lipid domains demonstrated that uniform DSPE-RGD vesicles do not exhibit domain formation while phase-separated DSPE-RGD vesicles do (Figure S7). However, phase-separated DSPE-RGD vesicles also exhibited a lower melting temperature compared to phase-separated DOPE-RGD vesicles, which could indicate a morphological or chemical stability difference between domains that might affect binding. Furthermore, it is possible that DSPE-RGD segregates into small clusters that are undetectable by FRET even in compositions that should lead to uniform DSPE-RGD distribution, which is supported by our previous studies.²⁸ Headgroups also affect domain localization and could also affect how well RGD lipids phase separate, as seen similarly with fluorophores.⁵⁰ Regardless, our results revealed that using phase-separation to increase avidity of DSPE-RGD is not effective in contrast to DOPE-RGD vesicles, suggesting that the differences between the saturation of the ligand-bound lipid and local lipid environment may have a strong effect on vesicle-to-cell binding.

Finally, we studied RGD vesicle interactions with cells via confocal microscopy to visualize vesicle-to-cell binding. Jurkat and K562 cells were incubated with the same RGD vesicle compositions mentioned above for 2 h (red) and stained with a nuclear dye (NucBlue, blue). Images on the confocal microscope confirmed similar trends to those observed with flow cytometry, where phase separation increased binding of DOPE-RGD vesicles but not DSPE-RGD vesicles in both cell types (Figure 3f and Figure S9). Moreover, our images showed improved interactions of DOPE-RGD phase-separated vesicles compared to their uniform controls, suggesting that phase separation of ligands on nanoparticles to improve vesicle avidity to a target cell may also have the potential to enhance delivery of therapeutic cargo. Some vesicles are located near the nuclei, which could indicate that vesicle uptake and internalization could be occurring.

3.6. Studying Internalization of Phase Separated RGD Vesicles.

Finally, we wanted to further investigate what appeared to be internalization of our vesicles shown in Figure 3f. To study internalization, we used three different methods: temperature variations, trypan blue quenching, and trypsin digestion (Figure 4a).⁵¹ While temperature affects the rate of endocytosis, trypsin and Trypan blue addition remove signal from uninternalized vesicles. Specifically, trypsin cleaves surface proteins to remove vesicles that are still bound to integrins but not yet internalized and Trypan blue quenches the fluorescence of vesicles that have not internalized and can also quench vesicles fused to the cell membrane.⁵² For these studies, we changed the vesicle dye from rhodamine, which we had been using for vesicle binding studies, to NBD, to enable trypan quenching of the vesicle-bound fluorophore.⁵¹ We first looked at RGD vesicle binding to Jurkat cells at 4 °C, where endocytosis pathways are inhibited, and 37 °C, where endocytosis pathways are highly active. Here, we saw that at 37 °C than at 4 °C, suggesting that internalization is most likely occurring (Figure 4b). Again, we observed that DOPE-RGD phase separation enhances vesicle uptake at both temperatures, but we also observed DSPE-RGD phase separation enhances binding at 37 °C, contrary to Figure 3. Furthermore, we did not see binding of uniform RGD vesicles to cells and we saw reduced levels of nonspecific adhesion of control vesicles when labeled with NBD compared to Figure 3 experiments with vesicles labeled with rhodamine (see Discussion). After binding vesicles to cells, we either quenched the vesicles with Trypan blue or cleaved proteins on the surface with trypsin to determine the amount of vesicles internalized (Figure 4c,d and Figure S10 and S11). As expected, we only saw significant internalization for phase-separated vesicles that exhibited the most vesicle binding. At both 4 and 37 °C, phase-separated DOPE-RGD vesicles were internalized at similar percentages. However, only at 37 °C did we witness significant internalization of DSPE-RGD vesicles. This data indicates that phase-separated vesicles not only bind Jurkat cells to a greater extent than uniform vesicles, but they also are internalized to a greater extent.

Figure 4.

Phase-separated vesicles are partially internalized by Jurkat cells. (a) Schematic depicting the methods for determining extent of internalization. Jurkat cells are treated with 500 μM vesicles with NBD for 1 h at 4 °C or 37 °C. Externally bound vesicles are either quenched with Trypan blue or cleaved via trypsin. (b) Fold change of Jurkat cells bound to RGD vesicles and control vesicles at 37 °C normalized to the vesicle signal on Jurkat cells at 4 °C. Error bars represent SEM for n = 3 different experiments performed on different days. Significance testing represents the results of two-way ANOVA with Tukey’s multiple comparison tests between 4 and 37 °C. (c, d) Vesicle fluorescence from internalized vesicles was calculated for two temperatures: at (c) 4 °C and (d) 37 °C by taking the difference between data in panel (b) (total vesicle fluorescence) and subtracting the fluorescent signal for the same condition after treatment with Trypan blue and trypsin to remove surface-bound vesicles. Percentages listed above specific bars are the percent internalized fluorescence relative to the vesicle signal in panel (b) for the given temperature. Error bars represent SEM for n = 3 different experiments performed on different days. Significance testing represents the results of two-way ANOVA with Tukey’s multiple comparison tests between untreated cells vs treated cells. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Finally, we also wanted to see whether phase separation enhances RGD vesicle binding to adherent cells. We used MDA-MB-231 cells, which have been previously shown to bind to RGD nanoparticles.⁵³ Curiously, we did not see any significant binding of RGD vesicles to MDA-MB-231 cells and instead found that control vesicles exhibited more background adhesion than RGD-containing vesicles (Figure S12). RGD nanovesicles do not bind as strongly as larger vesicles previously demonstrated, however, indicating that the size of the nanoparticles could also be a confounding variable.⁴⁸ This data suggests that adherent cells may be less responsive to phase separation and ligand presentation on nanoparticles and merits further investigation across a broader range of cell types.

4. DISCUSSION

Cells spatially control the presentation of receptors on their surface to better control binding and signaling processes, and we theorize that mimicking these properties could be beneficial to therapeutic nanoparticles. While overexpression of receptors is one route a cell might use to increase signaling, this overexpression can also lead to disease through constitutive activation.⁵⁴ To overcome the need to use gene expression changes—a slow process—to induce changes in signaling kinetics, cells utilize their membranes to rapidly reorganize and cluster receptors together in response to certain cell states or environmental conditions so that they are better poised to initiate intracellular and intraorganelle signaling and reduce the effects of noise. This phenomenon is observed in T cells, which needs to balance receptor affinity and expression in order to ultimately optimize their overall functional avidity.^55,56 To do that, T cells utilize receptor clustering to dynamically change their avidity,⁵⁷ which has also been harnessed in CAR-T cell engineering.^58,59 Clustering of low affinity ligands or receptors to change the overall avidity of ligand–receptor interactions is therefore a potent, biologically relevant strategy to not only prevent off-target signaling but also dynamically enhance signaling in response to environmental changes.

Here, we showed how these same cellular principles of modulating cellular avidity through clustering of lower affinity receptors can be applied to nanoparticles. Our group recently demonstrated that clustering ligands on vesicles using lipid phase separation is an effective route to improve intracellular signaling in a target cell.²⁸ This current work demonstrates that phase separation can enhance binding interactions between a wider range of biologically relevant, low affinity interactions and furthermore be harnessed to enhance binding avidity of vesicles by changing the location of sterically shielding molecules. Because many therapeutic ligands are expensive to produce, phase separation in nanoparticles is a readily achievable route to enhance local ligand concentration and require less total ligand per particle. We envision that phase separation can be harnessed as a tool to design vesicle technologies with multimodal functionalities.⁶⁰

Interestingly, we found binding differences when the ligand was localized into the liquid ordered or liquid disordered phases and when the saturation of the lipid bound to RGD was changed (Figures 3 and 4). When RGD was in the liquid disordered phase, vesicle binding was enhanced relative to when RGD was in the liquid ordered phase. This enhanced binding could be due to fluidity effects, as it is known that greater membrane fluidity and corresponding ligand mobility associated with the liquid disordered phase enhances binding.^7,61 Previous work has shown that vesicles with the capacity to phase separate can enhance liposomal delivery and vesicles may not need to be phase-separated prior to interactions with cells to enhance vesicle-to-cell binding.²⁶ Because cellular binding events can cluster receptors on the surface of a target vesicle, the binding event can lower the energy barrier of lipid phase separation and induce phase separation. This binding-induced phase separation may explain why we still found phase separation enhancements in cellular internalization of DSPE-RGD vesicles, where RGD is segregated to the ordered domain, though this effect was significantly weaker than vesicle internalization observed with DOPE-RGD vesicles, where RGD is segregated to the disordered domain (Figure 4). Taken together, our results support the inclusion of therapeutic or targeting ligands into lipid disordered phases to better bind and internalize into target cells.

A curious phenomenon we observed was the difference in RGD binding and vesicle background adhesion based on the lipid-conjugated dye we used to track vesicle-to-cell binding. While both rhodamine and NBD-labeled vesicles showed phase-separated vesicles enhanced ligand-mediated binding, rhodamine-labeled vesicles resulted in greater cell binding by the uniform RGD vesicles as well as more background adhesion with ligand-free vesicles. In contrast, NBD-labeled vesicles had negligible background adhesion to cells and, when RGD was uniformly presented on vesicles, demonstrated no significant cell binding. While the experiments used slightly different conditions, 25 °C/30 min for rhodamine and 4 °C and 37 °C/1 h for NBD, we would only expect minor differences but similar binding trends. Our differential results with the two lipid dyes suggest that rhodamine could be the culprit enhancing background adhesion with non-modified vesicles and with vesicles that uniformly present RGD. It is known that lipid-anchored fluorescent probes can influence the chemical and mechanical properties of vesicles.⁶² It is also important to note that NP binding is a multistep, dynamic process that is influenced by both specific and nonspecific interactions.⁶³ Changes in the lipid dye, vesicle base composition, and RGD saturation could all be changing how these nonspecific interactions synergize with the specific interactions of RGD. Further investigation into these phenomenon as well as careful selection and characterization of lipid dyes used for vesicle tracking is important going forward.

4.1. Limitations of Study.

There are a few limitations of our study that should be addressed in future studies. First, we needed to use different lipid compositions to study the effects of ligand clustering via phase separation. Future studies should look at using a single base composition for phase-separated vesicles, and varying the ligand-lipid conjugate so the ligand is either uniformly distributed or locally concentrated into domains. Using this technique, we would reduce composition-dependent background adhesion as a possible confounding variable in vesicle-to-cell binding. Another limitation is that we have not explored the effect of phase separation on ligands conjugated with a PEG spacer. Traditionally, most proteins or antibodies conjugated onto a vesicle will contain a PEG spacer located between the ligand and the membrane.⁶⁴ This spacer helps with conjugation efficiency and can be used to tune the distance of the protein from the vesicle surface. While there are examples in immune signaling where minimizing the ligand distance is important,^65,66 most therapeutic nanoparticles will include a PEG spacer for ligand conjugation. Previous work has demonstrated that phase separation can still occur in proteins conjugated through a PEGylated lipid and future studies should look at the effect of PEG length on phase-separated vesicle binding.⁶⁷

Finally, we did not observe enhancement of cell binding with phase-separated vesicles when we looked at adherent cells. In fact, we saw that background adhesion was the larger driving factor for vesicle uptake (Figure S12). As of now, we have only witnessed phase-separated vesicles enhancing binding to suspension cells, which could be useful to target immune cells in circulation. To better investigate binding to adherent cells, future studies should use stronger, more specific ligands than those used here.

5. CONCLUSIONS

In this work, we demonstrated that lipid phase separation can be used to spatially control ligands on lipid vesicles, which can modulate the overall avidity of the vesicle to a target cell. Phase separation can increase NP binding to cells by both enhancing ligand binding through clustering and adjusting the steric hindrance of PEG molecules. Finally, we demonstrated that the type of lipid anchor used to localize a ligand to a vesicle can impact the capacity of phase separation to affect vesicle-to-cell binding and should also be considered when creating lipid-based nanoparticles. Similar to how cells rearrange their membranes and ligands to activate receptor signaling, we expect nanoparticles that mimic this feature should better bind and activate target cells, which should be useful for applications like drug delivery.

Beyond mimicking the cell membrane, phase separation could also be used to engineer stimuli-responsive vesicle avidity in order to enhance drug delivery. Phase separation can be induced through multiple stimuli including tension,⁶⁸ pH,⁶⁹ temperature,⁷⁰ small molecules,⁷¹ peptides,⁷² proteins,^73,74 and association with soluble oligomerizing molecules.^75,76 Phase-separated vesicles could therefore be used in conjunction with these stimuli to dynamically change the presentation of nanoparticle-bound ligands. For example, with a temperature responsive vesicle,⁷⁰ temperature changes could be used to change ligand presentation, cell binding, and release of encapsulated cargo. Furthermore, PEG phase separation could be used to dynamically modulate steric shielding events in response to environmental conditions of the nanoparticle. By using lipid phase separation, new classes of stimuli-responsive nanoparticles can be developed for enhancing drug delivery by dynamically controlling the presentation of both targeting and shielding molecules.