Binding Kinetics of Liposome Conjugated E-selectin and P-selectin Glycoprotein Ligand-1 Measured with Atomic Force Microscopy

Abstract

Various ligand-functionalized liposomes have been developed for targeted therapies. Typically, the binding properties of the ligands and targeted proteins are measured with surface plasmon resonance (SPR), where the proteins are immobilized on a rigid surface. However, the difference of protein-ligand binding kinetics between liposome-conjugated protein and rigid surface-conjugated protein is not fully understood. In this work, the binding kinetics of P-selectin glycoprotein ligand-1 (PSGL-1) and E-selectin conjugated on liposome and on rigid surfaces are investigated with Atomic Force Microscopy (AFM). The results suggest that protein orientation and diffusion on liposomal membrane can alter the binding kinetics of the protein-ligand interaction. Specifically, the association and dissociation rate constant of AFM probe-conjugated E-selectin and glass-conjugated PSGL-1 are measured as 9.32×10⁴ M⁻¹s⁻¹ and 1.54 s⁻¹, respectively. While for the liposome-conjugated E-selectin and glass-conjugated PSGL-1, the kinetic constants are measured as 5.00×10⁷ M⁻¹s⁻¹ and 2.76 s⁻¹, respectively. Thus, there is an order’s magnitude increase of binding affinity (from k_d = 16.51 µM to k_d = 0.06 µM) when protein is attached to liposome compared to attached to a rigid surface. The results might provide better understanding and pave the way for the future design of the ligand-targeted liposomes.

Graphical Abstract

Introduction

Liposomes have been successfully used for cancer therapy, infection treatment, and vaccination [1]. In the last two decades, ligand-targeted liposomes, which have bioactive molecule-functionalized surfaces, have been extensively investigated for targeted drug delivery [2]. By taking advantages of the binding avidity between the coated biomolecules and surface proteins of targeted cells, the ligand-targeted liposomes can be enriched in targeted tissue, and deliver the therapeutic drugs with increased efficacy and less side effect [3].

To further improve its therapeutic efficacy in cancer treatment, dual-ligand liposomes have been tested for targeting two tumor cell receptors [4]. For example, a dual-antibody functionalized doxorubicin-containing liposome, which targeted CD19 and CD20, has been tested for treating B-cell lymphoma on mice model. Comparing to the single-ligand doxorubicin-containing liposomes, which were functionalized with either anti-CD19 or anti-CD20, this dual-antibody functionalized liposome increased the mice life about 50% [5]. In a different case, Zhang et al. reported their work on lung cancer treatment with dual-ligand liposome for gene delivery [6]. In their work, the transferrin and hyaluronic acid were used to functionalize the fusogenic liposome which targets the lung adenocarcinoma cells, a type of cells which usually has highly expressed transferrin receptors and hyaluronic receptor CD44. The in vivo study showed that this dual-ligand liposome generated 15% more transfection efficiency in lung adenocarcinoma-bearing mice. Additionally, the dual-ligand liposomes have been used to inhibit the metastasis. Before the cancer develops into metastasis stage, some tumor cells leave the original site and enter the blood circulation system to become circulating tumor cells (CTCs) [7]. These CTCs travel in blood vessels and some of them can adhere and go through the vessel wall to enter organs where they develop into secondary tumors. Mitchell et al. [8] reported an innovative CTCs killing strategy to inhibit this process. This strategy was based on E-selectin and tumor necrosis factor (TNF)-related apoptosis-inducing ligand (TRAIL) decorated liposome. In their work, the protein decorated liposome was intravenously administrated into the cancer-bearing mice, the liposome in blood stream then first bound to leukocytes via E-selectin-induced binding. In the next, by taking advantages of the big surface of leukocytes and their large population number in blood, the contact probability of liposome-leukocytes and CTCs was greatly increased. Once the CTCs were bound to the liposomes-leukocyte surface via E-selectin, TRAIL can further bind to death receptors on CTCs and subsequently trigger the apoptosis.

Despite the promising experiment results of ligand-target liposomes, their clinical usage is hindered by the immunogenicity in vivo and the limited understanding of their kinetics [9]. Typically, the binding affinity and binding kinetics of the protein-ligand pairs are measured with surface plasmon resonance (SPR) technique [10–13], in which the proteins are immobilized on a gold surface. However, when the SPR measured binding kinetic constants are directly used to estimate the behavior of liposome-conjugated proteins, the prediction might be aberrant from the actual results, as the protein immobilization may influence the binding kinetics [14, 15]. Therefore, to gain a better understanding of the binding kinetics of the ligand-target liposome to the cells, we study the binding kinetics between the liposome and the targeted-proteins expressed by the cells. In this study, E-selectin and PSGL-1 were selected as the protein-ligand interaction pair, and an Igor Pro-programmed AFM was utilized to measure the binding association rate constant, k_on, and dissociation rate constant, k_off, of the liposome-conjugated E-selectin and PSGL-1 on glass surface. The results obtained from this study should provide guidance for future ligand-targeted liposomes design and pave the way for the clinical applications.

Results

Unbinding force measured with AFM

The schematic of the AFM for unbind force measurement system is shown in Figure 1A. To investigate the binding kinetics of the liposome-conjugated E-selectin, three unbinding assays were designed. The unbinding force between E-selectin and PSGL-1, as well as E-selectin-liposome and PSGL-1 was measured for comparison, and the unbinding force between E-selectin-liposome and bovine serum albumin (BSA) was measured as a negative control. The schematic of experimental systems for the unbinding assays are shown in Figure 1B–D. For the first unbinding assay, E-selectin and PSGL-1 were immobilized by using a heterobifunctional polyethylene glycol (PEG) crosslinker, Acetal-PEG-NHS. For the second assay, the His-tagged E-selectin was attached to the liposome via chelation between Ni-NTA group and His-tag, the Ni-NTA group was introduced to liposome surface with Ni-NTA-DGS. And the E-selectin-liposome was attached to the neutravidin coated AFM probe by taking advantage of the biotin group on liposome surface. Both neutravidin and PSGL-1 were covalently immobilized to AFM probe and glass substrate with PEG crosslinker used above. For the third assay, the E-selectin-liposome and BSA were immobilized on AFM probe and glass substrate, respectively, with the same methods used in second assay. Based on the results of E-selectin-PSGL-1 unbinding assay, the maximum unbinding force of a single E-selectin-PSGL-1 complex was found to be 199.5 ± 10.5 pN. As the mean forces required to unbind His-tag and Ni-NTA was reported as 525 ± 41 pN [16], the His-tag-Ni-NTA bond should support the unbinding force measurement of E-selectin-liposome-PSGL-1. During the measurement, the AFM probe first moved down to the proteins on glass surface. Then, the probe may stay on glass surface for a short period, known as docking time, to give more time for protein binding. Next, the probe is retracted from the glass surface and returned to original position before the next measurement. This “approaching-retraction” process is recorded by the laser beam and photodiode to generate the force-displacement curve. By repeating this “approaching-retraction” cycles for hundreds of times, the unbinding force data can be collected for the following analysis.

Figure 1.

Schematic of the experimental system. (A) Diagram of AFM force measurement setup. (B) The AFM probe is functionalized with E-selectin protein for the E-selectin-PSGL-1 unbinding assay. (C) The AFM probe was functionalized with E-selectin-liposome for the E-selectin-liposome-PSGL-1 unbinding assay, the inset shows the structure of the constructed E-selectin-liposome. For assays of B and C, the PSGL-1 was immobilized on a glass substrate immerging in buffer solution. (D) As a control experiment, the constructed E-selectin-liposome was conjugated to AFM cantilever as in (C), and the BSA, instead of PSGL-1 was immobilized on the glass surface.

Three typical force-displacement curves of the retraction process are shown in Figure 2A. The upper curve (sample from “ESLP vs BSA” experiment) represents the no interaction force case which indicates there is no interaction between the AFM cantilever and the glass substrate. The middle curve (sample from “ES vs PSGL1” experiment) represents the unbinding process of E-selectin-PSGL-1 complex. The lower curve (sample from “ESLP vs PSGL-1” experiment) shows the unbinding process of E-selectin-liposome-PSGL-1 complex. The corresponding stages of the complex unbinding process are shown in Figure 2B.

Figure 2.

Pulling traces of the unbinding assays. (A) AFM pulling trace sample of unbinding assays. The upper trace (sampled from ESLP vs BSA test) involves no interaction. The middle trace shows the rupture force of the E-selectin-PSGL-1 complex. The lower trace shows the rupture force of the E-selectin-liposome-PSGL-1 complex. F_u is the rupture force, and k is the system spring constant calculated from the slope of force-displacement curve. (B) Demonstration of protein-ligand complex stretching and rupturing during the AFM cantilever retraction. When the cantilever moves up from the lowest point, the E-selectin-PSGL-1 complex will first return to its original length, then get stretched until rupture. (C) Adhesion frequency of different interaction pairs. Error bars are the standard derivation of the frequency measured under different retraction speed. Here, E-selectin is noted as “ES”, E-selectin liposome as “ESLP”, and PSGL-1 as “PSGL1”.

In the last two traces, the unbinding force, F_u, can be derived from the sudden jump of the force curve. The system spring constant, k, can be derived from the slope of the force peak. With the system spring constant, the retraction speed (also known as loading rate) which correlates to the unbinding force, can be calculated [17]. Comparing the “ESLP vs. PSGL1” trace to the “ES vs. PSGL1” trace, the system spring constant was reduced, which was caused by the liposome deformation during the AFM cantilever retraction.

The average binding probabilities of each interaction pairs under different retraction speeds are shown in Figure 2C. The control group has a low binding probability, and the two experimental groups have much higher binding probabilities which indicate the interaction in these groups are specific. In a typical AFM unbinding experiment, not only the force-displacement curve of unbinding force (as middle and lower curve in Figure 2A) can be observed, but also the force-displacement curve of no interaction and non-specific interaction, such as electrostatic interaction, can be observed. As when the AFM cantilever shortly touches the glass substrate, the protein and ligand on each surface may not be able to contact and bind during the short time. Although the no-interaction and non-specific interaction exist in all unbinding experiments, the ligand-protein interaction pairs usually have higher binding probability as they can form specific binding while the protein pair in control group cannot [18]. Therefore, the following kinetics analysis only concerns the data collected from specific interaction groups, E-selectin-PSGL-1, and E-selectin-liposome-PSGL-1.

K_off estimation

To further analyze the kinetics of the interactions involved in the assays, the unbinding forces and their corresponding loading rates were used to determine the dynamic force spectrum (DFS) (see Figure 3A). The unbind force-loading rate pairs measured in each assay were evenly separated in groups based on their loading rates. For each group, the average loading rate was taken and plotted with the most likely unbinding force determined by the histogram plot (see Figure S1 and S2), and two examples are shown in Figure 3B. Thus, a series of loading rate-unbinding force pairs were extracted for each assay, and the DFS which describes the linear correlation between the logarithm of loading rate and unbinding force was determined with regression analysis. Over a loading rate of 200 to 20,000 pN/s, the unbinding forces of E-selectin-PSGL-1 and E-selectin-liposome-PSGL-1 range from 64 to 210 pN and 30 to 104 pN, respectively. The DFS indicates that the unbinding force of E-selectin-liposome and PSGL-1 is about half of the E-selectin and PSGL-1.

Figure 3.

AFM measurement of E-selectin-PSGL-1 and E-selectin-liposome-PSGL-1 interactions. (A) The dynamic force spectrum of the interaction pairs, which describes the correlation of the unbinding force and loading rate. The error bars show the half-width of the bin size in each unbinding force histogram. (B) Histogram examples of the unbinding forces at an average loading rate. Upper panel shows the histogram used to determine the most likely unbinding force at 1156 pN/s for E-selectin-PSGL-1 interaction. Bottom panel shows the histogram which determines the most likely unbinding force at 1683 pN/s for E-selectin-liposome-PSGL-1 interaction.

By fitting the DFS to the Bell-Evans model, the dissociation rate constant in the absence of an external force, $k_{off}^0$, and activation barrier width, γ*_, were calculated for both interaction pairs. Bell-Evans model describes how the dissociation rate, k_off, increases with the loading rate, r_f [17]. The relation is described by the following equation:

$$f=\frac{k_{B}T}{{\gamma }^{\ast }}\ln \left(\frac{{\gamma }^{\ast }}{k_{off}^0{k}_{B}T}\right)+\frac{k_{B}T}{{\gamma }^{\ast }}\ln \left(r_f\right)$$

in which T is the absolute temperature, k_B is the Boltzmann constant, f is the external pulling force under. The $k_{off}^0$ for E-selectin-PSGL-1 and E-selectin-liposome-PSGL-1 was determined as 1.54 s⁻¹ and 2.76 s⁻¹, and the γ* was 0.11 nm and 0.26 nm, respectively. The results suggested that the E-seletin-liposome-PSGL-1 complex dissociates faster.

K_on estimation

To further investigate the difference in association rate constant, k_on, between the liposome-conjugated E-selectin and rigid surface-conjugated E-selectin, a method developed by Hinterdorfer et al. was used to estimate the k_on of both interaction pairs [19]. This method assumes the protein-ligand binding follows a pseudo first-order kinetics. Therefore, the association rate can be described as k_on = (τ C_eff)⁻¹, in which τ is the interaction time, and C_eff is the effective concentration of the binding pairs.

The interaction time, τ, can be estimated from the plot of binding probabilities against the docking time (see Figure 4A). The binding probability usually increases with the docking time, t, as the longer contact time between the AFM cantilever and glass slide will allow more protein-protein bonds to form, this relation can be described by P = A(1−exp (−(t−t₀)/τ)), where P is the binding probability, A is the maximum binding probability and t₀ is the shortest docking time, both A and t₀ are determined by experiments [20]. As shown in Figure 4A, the binding between E-selectin-liposome and PSGL-1 reached to the saturation plateau earlier than the E-selectin and PSGL-1 on the rigid surfaces, which suggested a higher association rate. By fitting the binding probability-docking time plot to the equation, the τ were calculated as 0.15 s and 0.10 s for E-selectin-PSGL-1 and E-selectin-liposome-PSGL-1 assay.

Figure 4.

k_on measurement for E-selectin-PSGL-1 and liposome-E-selectin-PSGL-1. A. Binding probability of both interaction pairs are plotted as a function of contact time, thus the binding time constant can be estimated for on-rate estimation. B. Histograms of the unbinding forces measured at longest contact time. The multiple-peak Gaussian distribution was fitted to each histogram to estimate N_b for on-rate estimation.

For the k_on estimation, the other value, effective concentration, C_eff, is defined as the molar concentration of effective binding pair in a spherical effective volume, V_eff. Therefore, the C_eff = N_b/(N_A∙V_eff), in which the N_b is the number of binding pair in V_eff, and N_A is the Avogadro constant. The V_eff is calculated from the effective radius, r_eff, which represented the total length of protein/liposome protein and cross-linker in this study. In this study, the r_eff were 30 and 180 nm for E-selectin-PSGL-1 and liposome-E-selectin-PSGL-1 assay (see Figure S3)[19, 21–23], respectively. The N_b can be determined by applying multiple-peak Gaussian analysis to the unbinding force distribution at the longest docking time (see Figure 4B). The multiple Gaussian analysis shows that there are 5 peaks for both E-selectin-PSGL-1 interactions and 3 peaks for liposome-E-selectin-PSGL-1 interactions. Therefore, the N_b was determined as 5 and 3 for the two interactions, respectively. With these parameters (see Table 1), the estimated k_on for E-selectin-PSGL-1 interaction was 7.77×10⁴ M⁻¹s⁻¹, for E-selectin-liposome-PSGL-1 interaction was 3.00×10⁷ M⁻¹s^-1.

Table 1.

Parameters for association rate estimation

	reff/(nm)	nb	𝜏/(s)	kon/(M−1s−1)
ES vs PSGL-1	30	5	0.146	9.32×104
ESLP vs PSGL-1	180	3	0.098	5.00×107

Discussion

In this work, AFM was used to investigate the binding kinetics of PSGL-1 and E-selectin conjugated on liposome membrane and on rigid surface. The interaction kinetics of rigid surface-attached E-selectin and PSGL-1 was first measured to build a baseline. The k_off for E-selectin-PSGL-1 is similar to AFM measured dissociation rate constant for P-selectin-PSGL-1 (k_{off_P-selectin} = 1.21 s⁻¹) [24], which is reasonable due to the structural similarity of E-selectin and P-selectin [10, 11]. Then the kinetic constants of liposome-conjugated E-selectin and PSGL-1 on glass surface were measured. Comparing the kinetic constants measured from the two scenarios, both k_off and k_on increased when the E-selectin was conjugated to liposome and the results are summarized in Table 2. The k_off measured with liposome-conjugated E-selectin is about twice of the dissociation rate constant measured with rigid surface-conjugated E-selectin, while the k_on was increases for ~536 times with liposome-conjugated E-selectin. The significantly increased association rate constant could be a result of protein orientation. Huang et al. reported their work on quantifying the molecular orientation effects on 2D protein-ligand binding kinetics with micropipette adhesion experiments, and the work concluded that the uniformly orientated molecules generate an increased association rate but wouldn’t influence the dissociation rate [25]. In their study, the monoclonal antibody, which only binds to the specific site of receptor proteins, was used to conjugate the receptors on cell membrane in a uniformly oriented manner, while the Cr³⁺ ion was used to couple the receptor to the membrane in a randomly oriented manner. As shown in Figure 5, in this study, when the E-selectin was conjugated to AFM cantilever via interaction of crosslinker and random amine residue groups of the proteins, the E-selectin molecules should be randomly immobilized to the surface, as multiple amine residues present on E-selectins. The liposome-conjugated E-selectin, on the other hand, was attached to liposome surface via chelation of Ni-NTA headgroup and His-tag on E-selectin C-terminal. As each E-selectin molecule only has one His-tag on its C-terminal, this chelation-based conjugation only happened at the C-terminal of the E-selectin so that the E-selectins are more uniformly oriented on liposome surface.

Table 2.

Summary of the interaction kinetics

	koff/ (s−1)	kon/(M−1s−1)	kd/(µM)
ES & PSGL-1	1.54	9.32×104	16.51
ESLP & PSGL-1	2.76	5.00×107	0.06

Figure 5.

Diagram of the E-selectin orientation in each scenario. Due to the different immobilization methods, E-selectin conjugated on AFM cantilever for E-selectin-PSGL-1 unbinding assay has a random orientation. E-selectin chelated to liposome for E-selectin-liposomePSGL-1 unbinding assay is more oriented.

For the k_off measurement, it is possible that the slightly increased dissociation rate was caused by the E-selectin diffusion on liposome membrane. As the E-selectin was attached to Ni-NTA-DGS, it could passively move around when the Ni-NTA-DGS diffuse on the outer leaf of liposome membrane. Consequently, when the diffusible E-selectin binds to the anchored PSGL-1, the diffusion movement may apply an extra pulling force to the E-selectin-PSGL-1 complex, thus increases the dissociation rate constant as demonstrated by the Bell-Evans model.

In addition, the randomly immobilized PSGL-1 was used in both measurement scenarios. Thus, the immobilized PSGL-1 was not uniformly oriented and cannot move on membrane surface as compared to the PSGL-1 expressed on cell surface. As discussed above, the PSGL-1 randomly conjugated on substrate can lead to reduced association rates in both scenarios, and increased dissociation rate when it binds to diffusible E-selectins. Moreover, the PSGL-1 expressed on cell surface has constant lateral diffusion, which allows the proteins to associate with lipid rafts [26–29] and promotes the cell-cell adhesion[30][31]. The thermal fluctuation of the cell membrane can also alter the binding kinetics between membrane proteins[32][33]. To further investigate the influence of these factors, the PSGL-1 immobilized on rigid surface can be replaced with PSGL1 expressed on cell membrane in the future study.

Overall, with the increased k_on and k_off, our experimental results show that the liposome-conjugated E-selectin has a decreased dissociation constant, k_d, of 0.06 µM, which is 298 times smaller than the rigid surface-conjugated E-selectin (k_d = 16.51 µM). The significantly reduced k_d indicates the higher binding affinity between liposome-conjugated E-selectin and PSGL-1. In the future, we will test the binding affinity of free E-selectin and liposome-conjugated E-selectin to the cells to further validate the increased binding affinity.

Conclusion

With the AFM force measurement experiment, it is observed that the liposome-conjugated ligands can have a significantly aberrant binding kinetics and binding affinity comparing to the ligands immobilized on rigid surface. Thus, for the ligand-targeted liposome design, it is important to use the binding parameters measured in proper conditions to achieve desired delivery efficiency.

Materials and Methods

Liposome preparation

To prepare the liposome, the detergent remove method was used. Cholesterol (Sigma Aldrich, USA), L-alpha phosphatidylcholine (PC) (Sigma Aldrich, USA), biotinyl phosphoethanolamine (biotin-PE) (Avanti Polar Lipids, USA), and Ni-NTA-DGS (Avanti Polar Lipids, USA) were first dissolved in 60 ˚C N-Octylglucoside (Sigma Aldrich, USA) solution to make lipid stock solutions. The liposome for the experiment was generated by mixing the lipids with a ratio of PC: cholesterol: Ni-NTA-DGS: biotin-PE = 16:2:2:1, to achieve a total lipid concentration of 800 mM. After mixing, 540 ul Tris-buffered saline (0.01 M, pH 7.4, 150 mM NaCl) was added to the mix and followed by eliminating N-Octylglucoside with the Pierson detergent removal column (Thermo Fisher, USA). Then, the detergent-free lipid mix was vortex and extruded through 0.2 μm syringe filter at 60 ˚C for 15 times to construct the liposome with unified size.

To functionalize the liposomes, the His-tagged human recombinant E-selectin (10335-H03H, Sino Biological, USA) was used. 45 μl of 0.58 μM E-selectin solution was added to the liposome solution. To collect the E-selectin conjugated liposome, the solution was centrifuged at 100,000xg for 3 hours at 4 °C. After the centrifuge, the supernatant was removed, and the pellet was resuspended in 100 μl Tris-buffered saline. The liposome samples were stored at 4 °C and used within 24 hours.

Cantilever preparation/glass slide preparation

The AFM cantilevers (MLCT-BIO-DC, Bruker, USA) and coverslip were functionalized with the method developed by Gruber et al. [10]. In brief, the AFM cantilevers were silanized with (3-aminopropyl)-triethoxysilane (Sigma Aldrich, USA) to obtain surface amine group. Then, the surface of the AFM cantilevers and pre-functionalized NH₂- glass slides (Nanocs, USA) were decorated with heterobifunctional polyethylene glycol (PEG) crosslinker, Acetal-PEG-NHS (Institute of BioPhysics, Johannes Kepler University).

For both unbinding assays, 100 μl of 0.3 μM recombinant human PSGL-1 (13863-H08H, Sino Biological, USA) solution was used to immobilize PSGL-1 to the coverslip. The protein-conjugated coverslips were stored at 4 °C in Tris-buffered saline and used within 8 hours.

For the E-selectin-PSGL-1 unbinding assay, 50 μl of 0.35 nM E-selectin solution was used to conjugate E-selectin to each AFM cantilever. For the liposome-E-selectin-PSGL-1 unbinding assay, 50 μl of 1.6 μM Neutravidin (Thermo Fisher, USA) solution was first incubated with AFM cantilever for 2 hours to allow the Neutravidin to be coated to the cantilever surface, then the free proteins were rinse off. The Neutravidin-conjugated AFM cantilevers were stored in Tris-buffered saline at 4 °C and used within 8 hours. Before the experiment, 50 μl of liposome prepared above was used to conjugate liposome-E-selectin to each AFM cantilever, the free liposome was rinse off after 10 minutes incubation.

Unbinding force measurement

A custom-designed AFM was used to measure the unbinding force of E-selectin-PSGL-1 and liposome-E-selectin-PSGL-1. To estimate the k_off and γ*, the unbinding forces were measured under different loading rates by controlling the AFM cantilever retraction speed, while the docking time of cantilever remained at 0.2 second. For E-selectin-PSGL-1 unbinding assay, the retraction speeds range from 0.94 to 7.52 μm/s. For liposome-E-selectin-PSGL-1 case, the speeds range from 0.94 to 7.52 μm/s. To measure the kon, the cantilever retraction speed was set at 3.76 μm/s, the unbinding measurements were performed at various docking time ranging from 0 to 0.6 second. The E-selectin-PSGL-1 unbinding assay was conducted in Tris-buffered saline containing 5 mM CaCl₂ at room temperature. The liposome-E-selectin-PSGL-1 unbinding assay was conducted in Tris-buffered saline containing 5 mM CaCl₂ and 5 mM NiCl₂.

Highlights.

• Unbinding forces of PSGL-1/E-selectin and PSGL-1/E-selectin liposome were measured with AFM

• Binding kinetics of the two binding pairs were derived from measured unbinding forces

• Compared to E-selectin, E-selectin liposome has much higher association rate constant to PSGL-1

• E-selectin on liposome has much higher binding affinity to PSGL-1 than E-selectin