Dendron-Functionalized Surface: Efficient Strategy for Enhancing the Capture of Microvesicles.

Microvesicles (MVs) are used by various types of cells in the human body for intercellular communication, making them biomarkers of great potential for the early and non-evasive diagnosis of a spectrum of diseases. An integrated analysis including morphological, quantitative, and compositional studies is most desirable for the clinical application of MV detection; however, such integration is limited by the currently available analysis techniques. In this context, exploiting the phosphatidylserine (PS) exposure of MVs, we synthesized a series of dendritic molecules with PS-binding sites at the periphery. PS-dendron binding was studied at the molecular level using NMR approaches, whereas PS-containing membrane-dendron interaction was investigated in an aqueous environment using plasmon waveguide resonance spectroscopy. As a proof of concept, polyethylene terephthalate surface was functionalized with the synthetic dendrons, forming devices that can capture MVs to facilitate their subsequent analyses.

Introduction

In the human body, microvesicles (MVs) like other kinds of extracellular vesicles play a pivotal role in both intercellular (adjacent or distant) communication and extracellular environment adaption (Henao Agudelo et al., 2017, Aupeix et al., 1997, Wang and Lu, 2017, Wyllie and Ramirez, 2017, Jeney, 2018, Paolicelli et al., 2019). The biogenesis of MVs is recognized as the budding of cytoplasmic membrane of eukaryotic cells upon activation (Tricarico, Clancy and D'Souza-Schorey, 2017). Widely adopted as carriers of biomolecules (proteins, mRNA, microRNA, lipids, etc.) from parent cells to even transkingdom recipients, they are omnipresent in all kinds of bodily fluids (Bruschi et al., 2019, Chavez et al., 2019, Clancy et al., 2015, Nievas et al., 2018, Owens and Mackman, 2011, Sampaio et al., 2017, Sun et al., 2018, Takahashi et al., 2017). In light of such knowledge, for novel diagnostic methods such as liquid biopsy, there are good reasons for MVs to become attractive subjects of study, namely, (1) MV release happens under cellular stress, preluding other symptoms of a same disease, therefore MVs of certain composition could act as an early biomarker (Jayachandran et al., 2008, La Marca and Fierabracci, 2017); (2) MV emission happens to a huge variety of cells from stem cells to parasites, making MV analysis useful for a spectrum of disease detection such as cancer, diabetes, viral infection, and cardio/cerebrovascular diseases (Aupeix et al., 1997, Owens and Mackman, 2011, Muralidharan-Chari et al., 2010, Andrews et al., 2016, Bergen et al., 2018, Holliday et al., 2019, Meckes and Raab-Traub, 2011); and (3) MV membrane composition is dependent on the cytoplasmic membrane composition of its parent cell, making them easier to be traced to the origin even when samples are collected at a distance to the malfunctioning tissue, allowing non-evasive sampling while improving the accuracy of the diagnosis (Jayachandran et al., 2008, Andrews et al., 2016).

For the application of MV detection in disease diagnosis, the quantity, size distribution, and composition (lipid composition, proteins, RNAs, etc.) of the MVs in a certain sample are desired. Unfortunately, none of the currently available approaches are able to achieve both criteria: flow cytometry failed to detect 99%–99.9% of the vesicle population due to its optical limitations, whereas atomic force microscopy and electron microscopy cannot provide any compositional information (Obeid et al., 2017, Rautou and Mackman, 2013). Western blot and qPCR, on the other hand, are able to give information on protein and RNA composition, respectively; however, both methods fail to quantify the MVs within a sample (Henao Agudelo et al., 2017, Bruschi et al., 2019, Sun et al., 2018, Clancy et al., 2015, Svedman et al., 2018).

A proper characterization method for MVs is therefore a combination of the above-mentioned methods, where MVs in a sample are prepared in such a manner that electron microscopy, qPCR, and western blot can all be performed easily. For this purpose, we herein propose a device that allows MVs be captured to a device surface, allowing the future analysis.

Phosphatidylserine (PS) is the most abundant negatively charged phospholipid in human cellular membranes (Leventis and Grinstein, 2010). Despite being synthesized intracellularly, its distribution in healthy human cells is mostly limited to the plasma membrane, particularly the inner leaflet, by the function of flippases (Hankins et al., 2015, Segawa and Nagata, 2015). Its exposure to the extracellular leaflet is often considered as an early sign of programmed cell death, and its presence in MV outer membranes is also well documented (Owens and Mackman, 2011, Segawa and Nagata, 2015, Nagata et al., 2016, Iba and Ogura, 2018, van Engeland et al., 1998). MV capture devices using annexin V have been published since 1997 and have been improved over the years (Obeid et al., 2017, Gajos et al., 2017). However, this MV capture process depends on Ca2+ in the sample milieu by the concentration of millimoles (van Engeland et al., 1998). Such high concentration of Ca2+ would become problematic when MV capture is required before removal of cells within the sample: Ca2+ influx is a well-known stimulant for MV generation for a variety of cell types, causing the contamination of MV specimen captured onto the devices (Hugel et al., 2005, Taylor and Bebawy, 2019). An alternative approach for PS capture is the application of chemosensors, i.e., dipicolylamine-Zn2+ (DPA-Zn) complexes. First published as a sensor for phosphorylated peptides, DPA-Zn rapidly expanded into a family of chemosensors for the detection of negatively charged phospholipids (Ojida et al., 2002, Rice et al., 2016, Zwicker et al., 2019, Koulov et al., 2003). The synthetic chemosensors compared against annexin V have the obvious advantages of all chemosensors such as the ease of preparation in large quantity, relatively flexible transportation, and storage conditions. Most importantly, the binding of guest molecule to PS no longer requires the Ca2+-rich environment. For applications in MV capture, DPA-Zn molecules should be able to covalently conjugate with a supporting material. A strong interaction between DPA-Zn and PS is, of course, expected for the molecule to bind to MVs.

Herein, we wish to report a series of four dendron-functionalized surfaces with increasing dendricity and their use for the capture of MVs. To build these nanomaterials, and to evaluate the dendritic effects on the capture of MVs, we have designed four complexes with one, two, four, and eight peripheral DPA-Zn units (Scheme 1). For this purpose, our synthetic strategy is based on the dendron core connected to an alkyl spacer with a primary amine for surface functionalization. Multivalent binding of peripheral dendrons to PS containing MVs can be achieved to maximize the binding strength. Besides, it is suggested that spatially close DPA-Zn units have synergetic effect during PS binding (Ojida et al., 2002, Koulov et al., 2003). Phenoxyl repeating units will provide a structure wherein DPA-Zn moieties can be kept closer to each other spatially, hopefully achieving such synergy.

Scheme 1

DPA-Zn Molecules with Increased Dendricity

Results and Discussion

Dendron Synthesis

As illustrated in Scheme 2, 1DPAOH and 2DPAOH were synthesized by nucleophilic substitutions of dipicolylamine (DPA) with corresponding benzyl halide (chloride and bromide for 1DPAOH and 2DPAOH, respectively) alcohol. These reactions could proceed relatively easily in the presence of K2CO3. 4DPAOH and 8DPAOH, on the other hand, were synthesized using more strict reaction conditions: benzyl iodide was required for its higher reactivity during substitution reactions, whereas KOH was added to generate phenoxide before the addition of benzyl iodide. To facilitate higher yield in the synthesis of dendritic molecules, reactions were performed at low temperature (around −20°C), reducing the possible generation of side products. The convergent synthesis strategy facilitates the separation of starting material and the target compound, because the polarity of molecules increased significantly with the increase of the dendritic structure. This explained the high purity of complex 4. However, the convergent synthesis of 8DPAOH from the phenol dendron 4DPAOH led to modest yield, probably due to the great steric hindrance during the reaction and the high polarity of both compounds, which renders difficult the separation of compounds 4DPAOH and 8DPAOH by column chromatography. The steric hindrance can also be observed during the addition of the hexylamine spacers: for 1DPAOH and 2DPAOH, Boc-protected spacer can be easily attached with high yields and removed using trifluoroacetic acid also with high yields. In the case of 4DPAOH, however, Boc protection cannot be removed. Thus, Teoc protection had to be used instead because its deprotection only required fluoride ion, a significantly smaller reagent, which can easily find its way into the desired reaction site. For 8DPAOH, the attachment of a spacer is much more difficult because of the deeply buried phenol group within the dendritic structure, causing another significant drop in the final yield of C8 synthesis. The full characterization of all new compounds and the experimental procedures are given in Supplemental Information.

Scheme 2

Synthetic Routes of C1, C2, C4, and C8

For detailed synthesis protocols, please refer to “Transparent Methods in Supplemental Information, where corresponding NMR and mass spectra (Figures S4–S46) are also provided.

NMR Investigation of DPA-Zn Complex-PS Interaction

After acquiring the DPA-Zn complexes of increased dendricity, the zinc complex-PS interactions were first studied in solution. Although proposed in many literatures that PS interacts with DPA-Zn through electrostatic affinity of phosphate anion and Zn2+ of the complexes, and published X-ray crystallography structures of DPA-Zn and phosphate complexes always show bonding between Zn2+ and phosphate, surprisingly there is no experimental proof for this assumption yet (Zwicker et al., 2019, Plaunt et al., 2014, Selmeczi et al., 2007, Ngo et al., 2012, O’Neil and Smith, 2006). To address this issue as well as to preliminarily evaluate the binding ability of the complexes to PS, we adopted NMR investigation so that the peak shapes and chemical shifts can be used as evaluation criteria (Selmeczi et al., 2007).

31P NMR spectroscopy was used to examine the interaction between complexes and PS in solution, because the chemical environment of phosphorous atom in 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine (POPS) will be the most affected should the complexes bind to the phosphate of POPS. Considering the similarity of the binding events (DPA-Zn to POPS), the changes in chemical shift can also roughly represent the strength of such interactions (it is assumed that the observed chemical shift is the mole fraction-weighted average of the shifts of the free and DPA-Zn-bound POPS) (Fielding, 2000). As shown in Figure 1A, when 1 equiv. of different complexes is added into 1 equiv. of POPS in solution, the changes of the 31P NMR chemical shifts varies significantly. C1, C2, and C4 make the signal shift to high field, whereas C8 makes the signal go to low field. From C1 to C4, with the increase of molecular dendricity, the change in chemical shift also increases, indicating a stronger binding ability with more binding sites. This can be easily interpreted as the result of the increase in DPA-Zn concentration in solution. For instance, even though the molarities of the complexes are the same, C4 brings 4 times the amount of DPA-Zn into the solution compared with C1, pushing the reaction equilibrium further to the formation of (DPA-Zn)-POPS complex. However, the difference between different DPA-Zn molecules is not only limited to the increase of DPA-Zn moiety numbers. With the increase in DPA-Zn numbers, one obvious difference is the dendritic scaffolds to which the DPA-Zns are attached. To investigate the effect of the molecular scaffolds, another experiment was performed where the molar ratio of DPA-Zn units to POPS was fixed at 1:1. The 31P NMR spectra of the corresponding mixtures are summarized in Figure 1B. Again a similar trend in chemical shift can be observed: the changes in chemical shift increase with the increase in the number of DPA-Zn units within complexes, in the order of C1, C2, and C4. The differences in chemical shifts between C1 and C2 complexes bearing a single phenyl ring as a molecular scaffold are less significant, whereas C4, possessing a dendritic structure, shows more significant changes in the 31P NMR chemical shift of POPS. Regardless of the molar ratios, C8 behaves quite differently than the other molecules. It is most likely due to the so-called negative dendritic effect, in which the dendritic structure acts as a shield of steric hindrance, reducing the binding ability of DPA-Zn. In all cases, the changes are in good agreement with the assumption that DPA-Zn complexes bind to the phosphate groups on POPS, whereas C4 has the strongest ability to bind to POPS.

Figure 1

31P Spectra of Complex-POPS Mixture

(A) The mixture of complexes and POPS in 1:1 stoichiometry.

(B) The mixture of complexes and POPS with stoichiometry of 1 DPA-Zn unit to 1POPS.

Inset tables: mixture compositions.

Although the 31P NMR can provide preliminary information about DPA-Zn complex-POPS interactions, such as the interaction site on POPS and the interaction strength, the involvement of other atoms during complex-POPS binding is yet to be determined. For this purpose, 1H NMR is another useful technique. With the involvement of heteroatoms, nearby protons will experience change in the chemical environment, causing the shift of 1H NMR signals. However, the 1H NMR of synthesized complexes has too many overlapping regions with POPS, causing the difficulty of spectra interpretation. Heteronuclear single quantum correlation (HSQC), on the other hand, is able to distinguish different proton signals of similar chemical shifts according to the carbon backbone. It was therefore used to investigate the interaction from the perspectives of both POPS and complexes.

To investigate DPA-Zn-binding site on POPS, 1 equiv. of different complexes was added into POPS solution and the HSQC spectra of corresponding mixtures were recorded. Figure 2 shows the part of POPS HSQC NMR spectrum interpretation related to the head group. Looking at the glycerol moiety of the molecule (protons at positions a, c, and d), there is an obvious trend that the change in chemical shifts as well as the broadenings also increase with dendricity (in the cases of C1, C2, and C4) for the proton signals. For C8-POPS mixture, there are also shifts in glycerol signals; however, these shifts are less prominent compared with C4, again highlighting the negative dendritic effect. Unsurprisingly, this is in good agreement with 31P NMR spectra. Both 31P and HSQC indicate that the strongest (DPA-Zn)-POPS interaction is achieved with C4. Unfortunately, signals of the serine moiety (signals b and e) cannot be observed clearly after the complexation; therefore, involvement of the serine moiety is still unclear. Considering the change in 31P signal and the change in glycerol signals, the involvement of the phosphate group during complexation can now be confirmed.

Figure 2

HSQC Spectrum of POPS Head Group

Pure POPS, gray; C1+POPS, blue; C2+POPS, red; C4+POPS, green; and C4+POPS, purple. Inset table: the chemical shifts of the glycerol proton signals.

To understand the difference in DPA-Zn-POPS interaction among the synthesized molecules, study of the complexes before and after POPS binding is also performed. As only synthesized complexes contain aromatic groups, 1H NMR can be used directly for investigating the effect of POPS binding to the complexes without the interference of POPS signals. To compare the effect of PS binding on individual DPA-Zn moieties, 1, 0.5, 0.25, and 0.125 equiv. of C1, C2, C4, and C8 were added to POPS solution, respectively, to acquire the corresponding 1H NMR spectra. For C1, C2, and C4, the pyridine proton signals (Figure 3, peaks a, b, c, and d in each spectrum) become broader and the spectrum details are also lost after the binding to POPS. In the three complexes, the change in chemical shifts are most prominent for the α protons, which locate closest to the Zn2+ ions. This result confirms that the synthesized molecules interact with POPS through the complexed Zn2+ ions. Most interestingly, the aromatic proton signals of benzyl ether backbone in each molecule (Figure 3, peaks e, f, and g) also change significantly after binding to POPS. Considering their electron neutrality, the benzyl ether backbones are unlikely to interact directly with POPS. The chemical shift changes could come from either the change of the electron environment of the backbone after DPA-Zn of each molecule binds to POPS or the conformational change of the entire molecule after POPS binding. Although the mechanism is not clear, the indirect involvement of the benzyl ether backbones in POPS binding is confirmed. The differences of the 31P NMR spectra in Figure 1B for C1, C2, and C4 could be explained by the involvement of benzyl ether backbones in POPS binding, whereas C4 with the dendritic structure has the strongest positive influence on the change of POPS 31P chemical shift. Unlike the positive dendritic effect shown with C4, C8 with even higher dendricity showed little spectral change after the addition of POPS, again indicating a negative dendritic effect during POPS binding.

Figure 3

1H Spectra of Complex Aromatic Groups

(A–D) C1 (A), C2 (B), C4 (C), and C8 (D). Inset table: the chemical shifts of proton signals.

Plasmon Waveguide Resonance Study of DPA-Zn Complex-Model Membrane Interaction

Although solution NMR provides information on the molecular interactions between POPS and zinc complexes, PS in biological systems is not solubilized and can only be found on lipid membranes. The goal of the complex design is therefore the improved interaction strength of synthesized molecules to PS-containing biological membranes. To evaluate the interaction strength between POPS-containing membranes and synthesized DPA-Zn complexes, a homemade plasmon waveguide resonance (PWR) spectrometer was adopted. The detailed setup of PWR and lipid bilayer constitution used herein is described elsewhere (Calmet et al., 2016, Reimhult et al., 2003). A PWR spectrometer has a setup quite similar to that of an Surface plasmon resonance (SPR) spectrometer with one important difference: the plasmon-generating metallic thin film (~50 nm silver) is coated with another dielectric layer (~460 nm silica) as waveguide (Salamon et al., 1997a, Salamon et al., 1997b). The metallic layer itself is only able to generate plasmon resonance upon the excitation of linearly polarized light that has oscillation direction perpendicular to the metallic surface (p-polarized light), whereas the waveguide layer can generate waveguide resonance with light of both perpendicular and parallel (s-polarized light) polarizations (Salamon et al., 1997a, Salamon and Tollin, 2004). Once adhered to the silica surface, the optical properties of lipid membrane can be probed at both perpendicular and parallel directions using p- and s-polarized light, respectively. When DPA-Zn complex is interacting with such lipid membrane, the changes in s- and p-polarization signals are able to yield information on the change of membrane mass as well as anisotropy (Salamon and Tollin, 2004, Alves and Lecomte, 2019). Using a titration experiment, the peak positions of s- and p-polarized light can be fitted to acquire the Kd values of such interactions (Harte et al., 2014). After the formation of POPC(1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine):POPS = 5:1 (w/w) bilayer (PSPC membrane) on the prism surface, a titration experiment is performed for each complex. Briefly, each complex was dissolved in HEPES buffer and then titrated into the Teflon sample chamber. As equilibrium was reached in each addition, the complex concentration, the spectral shifts of s-polarized light (parallel to the membrane surface), and the spectral shifts of p-polarized light (perpendicular to the membrane surface) were recorded. The titration stopped when addition of complex solution did not induce further change in spectral positions. The complex concentration was then plotted against the resonance angle, and the plot was fitted using GraphPad Prism 5 (Figure 4). The acquired Kd values were recorded in Table 1.

Figure 4

Fittings of Complex-PSPC Membrane Titrations

(A–F) (A and B) Fitting of C1-PSPC membrane titration in s- and p-polarizations, respectively; (C and D) fitting of C2-PSPC membrane titration in s- and p-polarizations, respectively; (E and F) fitting of C4-PSPC membrane titration in s- and p-polarizations, respectively.

Table 1

Complex-Membrane Interaction Dissociation Constants (Kds) Acquired Using PWR

Complex		C1	C2	C4
Kd (M)	S	1.25×10−4±3.163×10−5	2.092×10−5±9.509×10−6	1.771×10−7±9.617×10−8
				8.622×10−7±1.496×10−7
				9.346×10−6±2.565×10−6
				1.217×10−6±3.904×10−7
	P	1.498×10−4±5.054×10−5	7.232×10−6±2.178×10−6	1.42×10−7±2.811×10−8
				2.282×10−7±2.811×10−8
				3.489×10−6±1.849×10−7
				6.056×10−6±1.208×10−6

s, Kd values acquired using s-polarized light; p, Kd values acquired using p-polarized light.

In the PWR titration experiment, binding events can be monitored by following the changes in both p- and s-spectra for each incremental addition of the complex. As explained by Salamon and Tollin, the resonance angle change in either polarization linearly correlates with the mass density (mass per unit area) changes on the adsorbed lipid membrane (Salamon and Tollin, 2004). Two different scenarios can generally occur: (1) complex binding to the membrane leads to a hyperbolic increase in the resonance shift upon complex addition and (2) complex binding results in a shift decrease. The first scenario is the most common for molecules that bind and accumulate in the membrane surface without much change in the membrane organization. Therefore, increasing accumulation of the molecule at the membrane level leads to a hyperbolic response with saturation being reached. The data can be fitted to obtain a binding affinity. The second scenario can also be observed and reflects an impact of the interacting molecule on the lipid membrane organization. Indeed, such decrease in the resonance shifts can only be explained by mass removal from the system as a result of a “detergent”-type effect of the molecule on the membrane. A binding affinity can also be obtained in this case. It is to be noted that this is an apparent binding affinity that reflects both the binding of the molecule investigated and the accompanying lipid reorganization. The complexes show very different binding behaviors between each other. The resonance angles of both p- and s-polarized lights experienced continuous decreases upon the addition of C1, and the decreases stabilized at about 11 mdeg and 15 mdeg, respectively, for p- and s-polarizations. The final equilibrium was only reached when the concentration of C1 in the chamber reached submillimolar level, and the interaction Kds determined by the fitting of the titration curves were at 10−4 M for both polarizations. When C2 was titrated to a membrane of same constitution until equilibrium, increasing spectral shifts of same scale happened for both polarizations. The Kd values were determined to be around 10−5 M, indicating a stronger complex-membrane interaction. In the case of C4, the titration curves indicated a four-stage reaction with the increase of complex concentration. The first three stages showed consecutive increases, with two Kd values as low as 10−7 M and the third at 10−6 M. The last stage showed decreasing resonance angles of both polarities with Kd values also at the 10−6 M level. The p- and s-polarized resonances stabilized at 60 and 37 mdeg, respectively, a shift more significant than either C1 or C2.

Fabrication and Characterization of the MV Capture Devices

The MV capture devices were then prepared by functionalizing polyethylene terephthalate (PET) sheets using the synthesized DPA-Zn complexes. The purchased PET sheets were chemically treated using similar procedures reported previously with minor adaptions as shown in Figure 5 (Chollet et al., 2007, Chollet et al., 2009). Briefly, the PET sheets were hydrolyzed and then oxidized to produce abundant carboxyl groups onto the surfaces. The carboxyl groups were activated using EDC (N-(3-Dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride)/NHS (N-hydroxysuccinimide) thereafter to assist the amide bond formation so that the complexes with amine-ending spacers can be attached to the PET surfaces covalently. For each step of the surface functionalization, the carboxyl group surface density was monitored with toluidine blue O (TBO) test, whereas the atomic percentage of the material surface was followed by X-ray photoelectron spectroscopy (XPS).

Figure 5

Surface Functionalization of PET Sheets Using Synthesized Complexes

As shown in Figure 6, the oxidized surface has the highest carboxyl group density. After the activation step, there is a drop of 43.06 pmol/mm2 in COOH density, resulting from the substitution of hydrogen in COOH by NHS. A further drop in the COOH density can be observed after the grafting of the complexes onto the surface. This drop can result from two events: first, there is a reaction of amine and COOH in the presence of complexes, further eliminating the carboxyl groups on PET surface; second, when grafted on to PET surface, positively charged complexes repel the TBO+ ions, preventing them from adsorbing onto PET.

Figure 6

Carboxyl Group Density on Treated PET Surfaces

Oxy., oxidized PET; Act., NHS-activated PET; C1, C1 functionalized PET; C2, C2 functionalized PET; C4, C4 functionalized PET; C8, C8functionalized PET. Unit, pmol/mm2. See also Figure S1 and Table S1.

The peak areas after fittings of the high-resolution XPS spectra were summarized in Table 2. The XPS results well correspond to each treatment of PET surface: After hydrolysis and oxidation steps, the ester bonds were hydrolyzed and then oxidized into carboxyl groups; therefore the highest value of COO signal among all the samples is observed. In the activation step, NHS molecules substituted the hydrogen of carboxyl groups. In the process, the number of surface carboxyl groups remains the same, whereas the NHS brings nitrogen onto the surface, leading to significant increases in the N1s and N-C=O signals without change in the –COO signal intensity. After the functionalization with complexes, multiple changes in the XPS spectra can be observed. First, the emergence of Zn indicates the successful attachment of complexes onto material surface. Second, the significant increase in (C1s N-C=O)/(C1s COO) and (N1s 399.9)/(O1s O=C) can be interpreted as the successful replacement of NHS by complexes, further confirming the covalent attachment of the complexes to PET surface. Third, as shown in Table 3, the molecular percentages (MPs) of the complexes (which is calculated by dividing the atomic percentage of an atom by the number of the same atom within a single molecule) calculated according to Zn2p3 and N1s all indicate an increasing difficulty for larger molecules attaching to the PET surface. The MP calculated with N1s is always higher than that calculated with Zn2p3, which could come from two facts: one is that there is NHS residue left on the surface because of incomplete substitution by the complexes, and the other is that during the intensive washing process, water behaved as a complexing agent and washed away the Zn2+ ion from the complex, whereas the ligands being covalently attached and insoluble in water, remained on the PET surface. This effect can become more prominent with the increase of dendricity, as shown in MPZn/MPN in Table 3. Zn2+ cations are forced closer to each other in space with the increase of dendricity, experiencing more repelling force from each other, making them easier to come off.

Table 2

XPS-Determined Atomic Percentages of PET Surfaces

Name	Oxy.	Act.	C1	C2	C4	C8
Si2p	0.33	0.38	0.53	0.39	0.49	0.63
S2p	0.33	0.33	0.27	0.23	0.23	0.29
Cl2p	0.15	0.11	0	0	0	0
C-C	32.20	26.62	34.18	28.02	31.97	28.25
C-CO	12.36	14.64	15.46	18.06	15.08	23.28
C-O	12.39	14.44	13.73	13.55	14.15	13.69
N-C=O	0.24	1.55	2.29	2.19	1.38	2.85
COO	12.09	11.65	7.33	9.40	8.38	5.32
291.6	1.77	1.34	0.80	0.96	1.31	1.02
N1s 398.0	0	0.27	0.13	0.09	0.23	0.01
N1s 399.9	0	1.77	3.05	2.85	3.42	5.30
N1s 401.9	0	1.05	0.28	0.29	0.28	0.40
O1s 530.7	0	0	1.32	0.33	1.22	0.75
O1s O=C	13.41	14.02	11.63	11.81	11.32	10.37
O1s O-C	14.72	11.84	8.49	11.32	10.06	7.56
Zn2p3	0	0	0.51	0.50	0.50	0.27

Oxy., oxidized PET; Act., NHS-activated PET; C1, C1 functionalized PET; C2, C2 functionalized PET; C4, C4 functionalized PET; C8, C8 functionalized PET.

Table 3

XPS-Determined Molecular Percentages

Surfaces	C1	C2	C4	C8
MPZn	0.51	0.25	0.13	0.034
MPN	0.92	0.47	0.32	0.216
MPZn/MPN	0.55	0.53	0.40	0.16

MP = AP/n, where MP is molecular percentage, AP is atomic percentage, and n is the theoretical number of the corresponding atom per molecule.

C1, C1 functionalized PET; C2, C2 functionalized PET; C4, C4 functionalized PET; C8, C8 functionalized PET.

Unfortunately, the exact molecular densities of all the complexes grafted to PET surface are yet to be determined. To the best of our knowledge, there is no technique available for such measurement. We tried to come up with an approximate value using the combination of the TBO measurement as a qualitative approach and the XPS as a semi-quantitative approach. TBO test indicated the carboxyl group densities. For the activation step, it was able to quantify the NHS groups fixed onto PET surface through the decrease of carboxyl group density; however, it failed to quantify the substitution of NHS by the complexes in the next substitution step. The further decreases in carboxyl group densities were suggestive for the substitution reaction, but they no longer serve as quantitative measurement for the reaction. The MP of a grafted complex calculated from the XPS results is quantitative; however, as it measures the total amount of atoms with a penetration about 10 nm into the bulk material, such percentage is still unable to transfer directly into molecular density at the surface of the material. Despite the above-mentioned disadvantages, TBO was able to suggest that all the synthesized complexes exhibit similar degree of carboxyl group “blocking” ability, whereas the XPS MPs of all complexes grafted onto PET surface suggest that different complexes bind to PET surface with the same order of magnitude.

For impurities appearing in the XPS spectra, Si most likely comes from the glass containers in which we processed all the PET sheets; S is from the manufacturing of the material as its concentration is relatively stable throughout the grafting procedures; last, Cl in oxidized and activated surface comes from the hydrochloric acid used in the surface washing of PET.

The Performances of the MV Capture Devices

The crucial parameters for vesicles being adsorbed on material surface in solution are the chemical composition of the substrate, temperature, and the osmotic pressure difference between the inside and the outside of the vesicles (Granqvist et al., 2014, Isono et al., 2007, Reimhult et al., 2003). In our case, to compare the complex-functionalized surfaces for their MV capture abilities, the most important parameter is the surface chemical environment. As the surface densities of complexes fixed onto PET are of the same magnitude for all four complexes, the MV capture abilities of functionalized surfaces can therefore be compared using the above-mentioned PET sheets. Cryo-scanning electron microscopy (SEM) and fluorescence (FL) micrograph of each surface after incubation with mesenchymal stem cell-generated MVs were used to visualize the comparison (for the FL imaging of MVs, CellMask Deep Red was used to stain the MVs beforehand).

Figures 7A and 7B show how the MVs responded to the C1-grafted surface. In the FL micrograph, only a small amount of donut-shaped red fluorescent spots can be observed. These structures were of sizes around 1 μm, much larger than the diameter of MVs. Considering the size distribution of the MVs acquired by nanoparticle tracking analysis (Figure S2), such phenomenon can only result from the aggregation of MVs. On the cryo-SEM image, the PET surface was blank and of a smooth morphology. No vesicle structure can be observed. MV capture using C2 functionalized PET is shown in Figures 7C and 7D. The whole material surface was covered with membrane structures, although inhomogeneity was observed throughout the surface. In the FL micrograph, although large areas of the surface were covered with fluorescent membranes, there were also bright dots of MV size in the less-fluorescent regions. Very bright membrane aggregations with size over 1 μm and of irregular shapes can also be observed. Observation using cryo-SEM also reveals that the majority of the PET surface was covered by coalesced membrane structures, whereas at the boundaries of such aggregations, holes in the membrane and a small amount of vesicles attached to the PET surface can be observed as well. As shown in Figures 7E and 7F, both FL micrograph and cryo-SEM show that the MVs uniformly cover the whole surface functionalized with C4. MVs were found individually attached to the material surface. There are indeed some aggregations of MVs; however, such aggregation does not cause the fusion of MVs, leaving intact vesicles or vesicle aggregations captured onto the PET surface. FL micrograph and cryo-SEM in Figures 7G and 7H both show that particles uniformly cover the whole surface functionalized with C8. On the zoomed cryo-SEM image (Figure 7H), the particles were found to be of the same size as MVs; however, these structures resemble rough membrane flakes rather than smooth vesicles.

Figure 7

Micrographs of Functionalized Surfaces after 15 min of MV Incubation

(A) Fluorescence micrograph of C1 functionalized surface incubated with CellMask Deep Red-stained microvesicles; scale bar, 8 μm.

(B) Cryo-SEM image of C1 functionalized surface incubated with microvesicles; scale bar, 20 μm.

(C) Fluorescence micrograph of C2 functionalized surface incubated with CellMask Deep Red-stained microvesicles; scale bar, 8 μm.

(D) Cryo-SEM image of C2 functionalized surface incubated with microvesicles; scale bar, 2 μm.

(E) Fluorescence micrograph of C4 functionalized surface incubated with CellMask Deep Red-stained microvesicles; scale bar, 10 μm.

(F) Cryo-SEM image of C4 functionalized surface incubated with microvesicles; scale bar, 3 μm.

(G) Fluorescence micrograph of C8 functionalized surface incubated with CellMask Deep Red-stained microvesicles; scale bar, 10 μm.

(H) Cryo-SEM image of C8 functionalized surface incubated with microvesicles; scale bar, 1 μm

See also Figure S2 and S3.

Micrographs indicate that the PET surfaces functionalized with C1 have very limited ability to capture MVs. This result correlates well with the PWR and NMR experiment results indicating that C1 has less ability to bind to PS. In comparison, C2, C4, and C8 functionalized surfaces were able to capture MVs due to the higher PS-binding abilities, although MVs behave differently when in contact with the three surfaces. When captured by C2 functionalized surface, vesicles are prone to rupture and then fuse with each other into membranes. Because of the destructed morphology of the MVs, their contents were suspected to be lost during the capture process, making C2 surfaces less favorable for the future applications for the examination of the vesicle contents. A similar conclusion can be drawn for C8 functionalized surface as it also caused the rupture of MVs. So far, C4 functionalized surface seems to be the best candidate for MV capture because vesicle morphology and their contents are well preserved. A possible explanation for such differences can be attributed to the dendritic backbones of the three molecules: the morphology of the membrane on C2 functionalized surface is highly reminiscent of the so-called supported lipid bilayers, where the vesicles (synthetic or naturally existing in biological systems) adhere to a hydrophilic substrate (such as silicon, glass, or silica) and form a bilayer via vesicle fusion (Granqvist et al., 2014, Isono et al., 2007). Such fusion is considered to be a process related to the chemical properties and most importantly the hydrophilicity of the surface in contact (van Weerd et al., 2015). C2 being the simplest of the three, its DPA-Zn moiety has the largest exposure to the aqueous environment with the highest density on PET surface yet with limited molecular flexibility. After C2 functionalization of PET, C2-PS interaction invites the vesicle adhesion to the PET surface, whereas the hydrophilic nature of C2 and the stiffness of PET substrate induce the vesicle to rupture, lie flat onto the surface, and fuse with each other into a larger membrane. Compared with C2, C4 is a relatively flexible molecule: as shown in Figure 3C, when binding to PS, the aromatic signals of the benzyl ether backbone also experience significant chemical shift, indicating the conformational change of C4 when binding to PS. Besides, C4 is a less hydrophilic molecule than C2, thus vesicle rupture is less likely to happen on the C4 functionalized surface. High-generation cationic dendrimers such as poly(amidoamine) (PAMAM) are known to cause membrane destruction. Mecke et al. found that PAMAM of generation 3 (G3) does not induce 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) membrane destruction, whereas G5 and G7 PAMAMs are able to destroy lipid bilayer by removing lipid from membrane (Mecke et al., 2004, Mecke et al., 2005). Through computer simulation, Wang et al. discovered that PAMAMs of lower generations have more flexible structures (Wang et al., 2012). When in contact with DMPC membranes, the smaller PAMAMs are able to adjust their conformation and cause little disruption of DMPC bilayer. In contrast, the higher-generation dendrimers' conformations are limited by their interbranch steric hindrance; thus at the dendrimer-lipid bilayer interface, DMPC membranes are forced to adopt the curvature of individual dendrimer molecules, causing possible membrane destruction. Similarly among our DPA-Zn complexes, C8 is the dendron of highest generation and thus has the highest conformation rigidity. When in contact with C8 functionalized surface, MVs are prone to destruction.

In conclusion, we have synthesized a series of multivalent DPA-Zn complexes with increasing dendricity. In 31P NMR studies of complex-POPS interactions, dendritic C4 showed the strongest POPS binding ability. This result is in good agreement with the fittings of PWR titrations where the complex-membrane (POPS containing) interaction Kd values were determined. C1 being the complex with only one DPA-Zn unit has Kd values of 10−4 M; C2 of similar molecular scaffold but with two DPA-Zn units showed Kd values of 10−5 M. C4 on the other hand, with four DPA-Zn units attached to a dendritic scaffold, showed a four-stage binding process in the titration experiment, with Kd values as low as 10−7 M. Such improvement in binding strength is beyond the simple multivalent binding effect. To understand the exponential increase of complex-membrane interaction with the increase of DPA-Zn units, we performed HSQC and 1H NMR experiments to investigate the interaction at the molecular level. The HSQC experiments indicated that the complexes bind to the phosphate group of POPS, whereas the 1H NMR spectra reveal the DPA-Zn involvement in the POPS-binding process as hypothesized in many literatures. The 1H chemical shifts of the benzyl ether scaffold where DPA-Zn units are attached also showed significant changes after POPS binding, indicating the scaffolds are also involved in the binding processes. In the case of C2, the two DPA-Zn units showed synergetic binding to the POPS molecules, and C4 with a dendritic structure further enhanced such effect by conformational change upon PS binding. Unfortunately, the enhancement effect cannot be applied to molecules of higher dendricity of the same repeating units. C8 of eight DPA-Zn units has been proved to be difficult to synthesize, while its PS-binding performance is no better than any smaller DPA-Zn complex according to 31P NMR. All the synthesized DPA-Zn complexes were able to be attached to the model PET surface covalently with an amino hexane spacer to form a device for MV capture, while the binding molecular densities on PET surfaces were semi-quantitatively determined with TBO and XPS experiments. The MPs were found to be at the same order of magnitude for all complexes, although a decreasing trend can be found with the increase in dendricity. The MV capture performance for each device was then evaluated with FL microscopy and cryo-SEM. Dendron C4 functionalized surface appears to be the best candidate for MV capture, as it not only successfully captured MVs onto the surface but also has maintained the morphology of the MVs. Such surface could be useful for future diagnostic applications because the device can capture biochemical information of both MV membrane and the inner content for analysis purpose. This interfacial engineering technique should be substrate independent, a high added value being that the same reactive platform can be applied across a spectrum of substrate materials used in medical analytical and testing laboratories.

Limitations of the Study

The dendron enhancement of MV capture onto material surface is discussed in this article as a proof of concept. Besides molecule dendricity, other parameters such as material substrate, spacer length, and dendron scaffolds other than the polybenzyl ethers we proposed herein may also affect the performance of the capture devices.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.