Establishing Rules for Self-Adhesion and Aggregation of N-Glycan Sugars Using Virus Glycan Shields

Abstract

The surfaces of cells and pathogens are covered with short polymers of sugars known as glycans. Complex N-glycans have a core of three mannose sugars with distal repeats of N-acetylglucosamine and galactose sugars terminating with sialic acid (SA). Long-range tough and short-range brittle self-adhesions were observed between SA and mannose residues, respectively, in ill-defined artificial monolayers. We investigated if and how these adhesions translate when the residues are presented in N-glycan architecture with SA at the surface and mannose at the core and with other glycan sugars. Two pseudotyped viruses with complex N-glycan shields were brought together in force spectroscopy (FS). At higher ramp rates, slime-like adhesions were observed between the shields, whereas Velcro-like adhesions were observed at lower rates. The higher approach rates compress the virus as a whole, and the self-adhesion between the surface SA is sampled. At the lower ramp rates, however, the complex glycan shield is penetrated and adhesion from the mannose core is accessed. The slime-like and Velcro-like adhesions were lost when SA and mannose were cleaved, respectively. While virus self-adhesion in forced contact was modulated by glycan penetrability, the self-aggregation of the freely diffusing virus was only determined by the surface sugar. Mannose-terminal viruses self-aggregated in solution, and SA-terminal ones required Ca²⁺ ions to self-aggregate. Viruses with galactose or N-acetylglucosamine surfaces did not self-aggregate, irrespective of whether or not a mannose core was present below the N-acetylglucosamine surface. Well-defined rules appear to govern the self-adhesion and -aggregation of N-glycosylated surfaces, regardless of whether the sugars are presented in an ill-defined monolayer, or N-glycan, or even polymer architecture.

Graphical Abstract

1. INTRODUCTION

The surface of eukaryotic cells and pathogens is covered with short, branched sequences of sugars known as glycans protruding from the proteins and lipids within the membrane.^1-5 More than 70% of these are N-glycans characteristically attached via the amine groups on proteins and having a core of two N-acetyl glucosamine (GlcNAc) followed by three mannose (man) sugars (Figure 1A). There are three types of N-glycans depending on the sequence of the sugar branches distal to the core: (i) high mannose type where all branches have mannose sugars, (ii) complex type where branches have repeat sequences of GlcNAc and galactose residues capped by sialic acid (SA), and (iii) hybrid type where one branch is high-mannose type and the other branch is complex type.^6-8 Fucose sugars may be attached to GlcNAc in the glycan core.⁹

Figure 1.

(A) Complex type of N-Glycans. (B) Experimental scheme for testing resilience of virus attachment to AFM probe. The probe with the attached virus is repeatedly pushed into opposite-charged APTES surface until 1 nN is reached and retracted. (C) A typical force curve observed when probe-virus is compressed against APTES surface has buckling features in approach and adhesion to surface in retraction. (D) Evolution of indentation thickness, a measure of virus compliance, as the probe-virus is repeatedly ramped into the APTES surface. (E) Range of peak adhesion or pull-out forces with oppositely-charged APTES surface endured by the virus during repeated ramping and retraction. (F) Range of adhesion energies resisting probe-virus pull-out from the APTES during the course of repeated ramping.

Glycans present sugars in conformations recognized by proteins or lectins. The protein binding of glycan sugars is a key event in the communication cascade between cells, cells and environment, and host and pathogens.^10-12 Debilitating malfunction of organ systems occurs in congenital disorders of glycosylation, a condition where glycan processing proteins are defective, indicating the criticality of glycans for biological function.^13-15 The branched architecture of N-glycans was also found critical for postnatal survival, protection against autoimmunity, and development of the nervous system.^16-18 The branching of N-glycans enhances biochemical signals, and it increases during tumor metastasis and apoptosis.¹⁹ The resulting multiplicity of galactose presentation, for instance, recruits clusters of galactose-binding lectins (galectins) close to the cell membrane, which cross-link neighboring membrane domains.^20-22 Reduced cell migration, growth arrest, and enhanced auto immunity are observed as a result. The sequence of sugars also plays a role. GlcNAc residues increase the recognition capacity of Mannose binding lectins.²³ Also each sugar in N-glycan sequence can participate in mutually exclusive biochemical pathways.²⁴ While the role of N-glycan architecture and presentation is increasing appreciated in the realm of protein-glycan binding and biochemical signaling, little is known about its impact on the sugar–sugar interactions which would constitute some of the first contacts between biological surfaces. The goal of the paper is to deconstruct how the sugars in complex N-glycan sequence and architecture determine the self-adhesion and self-aggregation of N-glycosylated surfaces.

The premise of the study is that well-defined rules govern the self-adhesion between N-glycosylated surfaces. The motivation for premise comes from initial self-adhesion studies between SA and mannose residues performed with single-sugar monolayers. Abeyratne-Perera et al.²⁵ reported “brittle” self-latching adhesions between monolayers of mannose (man). These self-adhesions can be described as Velcro-like; the release was short-range with a dominant peak at the separation interface between the layers. The Velcro-like releases also occurred when one of the mannose layers was present at the core of complex N-glycan shields under tiers of nonmannose sugars.²⁶ On the other hand, SA monolayers exhibited “tough” self-adhesion releases that can be distinguished as slime-like; the releases occurred over long distances with multiple unabating peaks (sawtooth pattern) and were mediated by salt. Accordingly, mannose-nanoparticles in solution coated viruses with complex N-glycan shields, whereas SA-nanoparticles aggregated the same viruses.²⁶ These observations with monolayers suggest that each sugar tier in the N-glycan complex may retain distinctly separate self-adhesion patterns, and that the self-adhesion character of the whole N-glycan complex can be predicted from the composition and sequence of the sugar tiers. Rules that relating biophysical adhesion to N-glycan composition will help us interpret the adhesion and aggregation behavior in cell–cell, cell–pathogen, and pathogen–pathogen complexes as N-glycans are widely present on these surfaces.^27-29 The problem is also tractable because N-glycan architecture is highly conserved and there are only five participating sugars to deconvolute. Decoding these rules is also important because the composition of N-glycans varies with age,^30,31 disease,^32-34 and stem-cell differentiation,³⁵ presumably changing not only the adhesion pattern of a surface but also the kind of glycosylated molecules that would accumulate around it.³⁶

While the monolayer studies provide support for a premise that well-defined rules could exist between N-glycan composition and adhesion pattern, the sugar presentation in these studies was not biologically relevant.^25,26 The monolayers were generated by inserting linker molecules into the sugars resulting in random linkages and orientations and without the branched architecture and accompanying sugars characteristic of N-glycans. Therefore, the goal of the present study is to determine how the self-adhesion between individual sugars translates when they are presented in N-glycan architecture and how the self-adhesion behavior of the whole complex can be constructed from the those of the individual sugars. The model N-glycosylated surface used is the dense glycan shield of human immunodeficiency virus 1 (HIV-1). The interaction between two virus glycan shields is interrogated with atomic force spectroscopy for the first time. The HIV virus glycan shield is a good model of a glycosylated surface because it is contributed predominantly by sugars attached to one type of membrane protein in the envelope membrane, and there will be no interfering adhesions from sugar–protein and ligand–protein interactions. In this study, the virus is pseudotyped (i.e., engineered to display a desired envelope protein) with the envelope protein of vesicular stomatitis virus (VSV-G). The VSV-G protein is a trimer and each spike has two glycosylation sites, both of which have been reported as complex N-glycans (Supporting Information 1).^37-42 The complex N-glycans are predominantly of the diantennary type,³⁸ and the extent of terminal sialic acid might change in some host cells.⁴³ A dense gel-like glycan shield can be sampled separately in indentation studies of these pesudotyped viruses.²⁶ The complex N-glycan shields of two viruses are brought into forced contact with force spectroscopy (FS) or into free contact by diffusion in solution. The changes in adhesion forces and aggregation pattern are tracked as layers of sugars are shaved off the glycan shield with site-specific glycosidases. The rules for self-adhesion between multitiered and branched N-glycan sugars are deconstructed along with insight from previous sugar monolayer studies.

We note that though early studies have reported the two glycans on each VSV-G spike to be of the complex type, it is possible there are dispersed hybrid and oligomannose glycans in the shield. Similarly while the surface adsorption of the virus on positive-charged surfaces is in agreement with a SA-rich glycan surface, it is possible that some of the complex glycans are not terminated with SA and instead display galactose. These possible scenarios are also considered below when experimental results are primarily interpreted as the interaction between two complex glycan shields.

2. MATERIALS AND METHODS

2.1. Materials.

293T cells and 293 cells were purchased from ATCC (Manassas, VA). The pHEF-VSVG expression vector (courtesy of Dr. Lung-Ji Chang) and pNL4-3.Luc.R⁻E⁻ (Courtesy of Dr. Nathaniel Landau) were obtained from the NIH AIDS Research and Reference Reagent Program. Glycosidases and lectins were purchased from Millipore Sigma Inc., St. Louis, MO. The glycosidase β-N-acetylglucosaminidase (NAG) from Canavalia ensiformis was used to expose the mannose core. The enzyme β-mannosidase from Helix pomatia was used to remove the mannose core. Neuraminidase (NA) from Clostridium perf ringens was used to expose galactose residues, and α-mannosidase from Canavalia ensiformis was used to cleave between the two tiers of mannose in the core. β-galactosidase (β-gal) from Aspergilus oryzae was used to remove galactose residues. Lectins from Maackia amurensis (MAA) and Concanavalin A (Con A) were used for detecting terminal SA and mannose, respectively. Lectin and glycosidase solutions were prepared in PBS and supplemented with ions to achieve final concentrations recommended by the supplier. Solutions with ConA were prepared to have final Ca concentrations of 100 μM.

2.2. Production and Characterization of Virus.

Vesicular stomatitis virus G protein (VSV-G)-pseudotyped pNL4-3.Luc.R-E-virus was prepared as previously described.⁴⁴ In brief, HEK293T cells were cotransfected with pNL4-3.Luc.R-E-⁻ HIV-1 genomic vector and pHEF VSV-G expression vector. The former does not express the HIV envelope protein Gp120, whereas the latter encodes the VSV-G envelope protein. Media with virus was collected at 72 h post transfection, concentrated, and stored at −70 °C. In a previous study, the pseudotyped HIV-1 R⁻E⁻/VSV-G virus was found morphologically similar to wild type HIV-1 virus in that it contains a coneshaped capsid within a membrane envelope, but the interactions of the glycan shield was consistent with the presence of complex N-glycans from VSV-G envelope proteins.²⁶

2.3. Virus Attachment to AFM Probe.

The virus was attached to a gold-coated probe using dithiobis succinimidyl propionate (DSP) linkers (Thermo Scientific). Dithiobis succinimidyl propionate (DSP) (4 mg) was dissolved in 1 mL of dimethyl sulfoxide (DMSO) and applied on gold-coated AFM probes for 30 min at room temperature. The disulfide bond is known to be cleaved upon binding to a gold surface and the molecules self-assemble into a monolayer via thiol groups at one terminus⁴⁵ to display free NHS ester groups at the other terminus. The probes were rinsed with water. HIV-1 R⁻E⁻/VSV-G virus in phosphate-buffered saline (PBS) was immediately added to the probe and incubated for 1 h at room temperature to allow the free NHS terminus to react with amines on the virus surface.⁴⁶ The probe was rinsed with PBS to remove all cross-linker products including NHS leaving groups and unconjugated virus and used immediately for experiments. The grafting schematic is presented in Supporting Information 2. By assembling the linkers on the surface and then overlaying the virus, the latter’s surface probed in FS was not exposed to cross-linking reactants and organic solvents.

2.4. Nanoindentation of HIV-1 Pseudotyped Virus Attached to Virus Probe.

The HIV-1 R-E-/VSV-G virus surface has negatively charged SA sugars which adsorb on positively-charged APTES-coating on mica surface. The APTES coating was generated by placing 100 μL of 5 mM (3-aminopropyl) triethoxysilane (APTES) on freshly cleaved mica surface and incubated for 30 min, followed by washing with water and then PBS. HIV-1 R⁻E⁻/VSV-G (6 μL in PBS) was applied to the APTES surface for about 8 min. In a previous imaging study, we established that these adsorption parameters produced complete coverage of the APTES surface with a monolayer of virus.²⁶ Then the surface was washed and covered with about 100 μL of PBS before mounting onto the sample stage of a Multimode/PicoForce system (NS-V controller, Bruker Nanosurfaces Inc., Santa Barbara, CA). A fluid cell retained the buffer during indentation experiments. NPG-10 cantilevers (Bruker Inc., Billerica, MA) were functionalized with HIV-1 R⁻E⁻/VSV-G as described in Figure 2 and mounted on the AFM cantilever holder. The spring constant (k) of the cantilever was determined by standard thermal tuning methods. The contact force for initiating cantilever retraction (i.e., maximum indentation force) was set at ~1 nN and the ramp size at ~500 nm. Force spectroscopy data were analyzed using in-house MATLAB code to quantify descriptive features of each curve. Indentation thickness is calculated as the difference between the separations where indentation force becomes higher than zero and where the maximum indentation force is reached (see Figure 3B). Adhesion length is calculated as the separation span over which the adhesion releases. Adhesion energy is the work done to release an adhering AFM probe and is calculated as the area under the negative adhesion force during retraction.

Figure 2.

SA–SA and man–man adhesions appear at high and low rates, respectively, of virus–virus interaction. (A) Experimental setup where two viruses are brought into contact up to a force of 1.2 nN and retracted. (B–D) Forces representative of Velcro-like, mixed, and slime-like adhesions were predominantly observed at low (50 nm/sec), intermediate, and high (>300 nm/sec) rates. Velcro-like adhesions are characterized by a single dominant peak close to the contact point between the two surfaces and were previously observed between man–man monolayers. Slime-like adhesions are distinguished by a sawtooth-like profile of release with several peaks which continue undiminished away from the contact point between the two surfaces and were previously observed in SA–SA monolayers. The mixed type has features of both. (E) The frequency of occurrence of the three adhesion types depends on the indentation rate with slime-like adhesions occurring toward the higher rates and Velcro-like adhesions occurring at lower rates. (F) The distribution of adhesion lengths measured during retraction is dependent of the indentation rate, with the longer lengths occurring toward the higher rates. (G) The distribution of adhesion energies measured during retraction is dependent on the indentation rate with higher energies occurring toward the higher rates. (H) The distribution of indentation lengths was not dependent on indentation rate, indicating similar virus populations were being sampled at all rates (n > 65 for each ramp rate, 3 samples and 2 runs).

Figure 3.

(A) Retraction forces between the two virus glycan shields observed at different indentation rates and on the same location; the adhesion patterns during retraction changes from man–man Velcro-like at low rates to SA–SA slime-like at high rates. (B) Schematic of the bare probe indentation of a surface adsorbed virus. (C–F) Virus force response to bare-tip indentation at different rates; as the indentation rate decreases regions with negative slope appear during approach (black arrow in blue curve) and negative adhesions occur during retraction (black arrow in red curve). (G) Frequency plot of incidence of negative slopes in the approach curves, where 0 incidence is more common at higher rates (blue arrow) and 3–4 incidence is more frequent at the lower rates (orange arrow).

2.5. Cleaving Sugars on Bound and Free Virus.

For force spectroscopy, cleavage was performed on HIV-1 R⁻E⁻/VSV-G attached to AFM probes or those adsorbed on APTES mica surfaces. The virus surface was rinsed with 100 μL of PBS. Sixty microliters of 1U/μL glycosidase and 2.5 mg/mL of BSA in PBS was overlaid on the virus surface. A time of 45 min to 1 h incubation was found sufficient to cleave sugar tiers without leaving significant interaction traces from uncleaved sugars. The enzyme mixture and cleaved sugars were washed from the virus surface using PBS. Because sugar cleavage was performed on surface-bound virus, cleaved products and enzymes could be removed by simple washing of the surface. Moreover the structural stability of the virus would also be retained to some extent. For cleaving virus in solution, stock enzyme solutions were prepared as per manufacturer recommendations and mixed with the virus to obtain a final concentrations of ~0.15 U/ml. The completion of cleavage was monitored from virus aggregation with DLS and by increasing the enzyme concentration until there was no more shift of the DLS curves. Virus aggregate sizes observed in DLS were stabilized for 1 h for the concentrations used in this study.

2.6. Dynamic Light Scattering (DLS).

DLS measurements were performed in 40 μL disposable cuvettes with a Malvern Zetasizer ZS (Malvern Instruments, Inc., Westborough, MA) as described previously.²⁶ Hydrodynamic diameters D_H were extracted from the correlation curves using an in-house Matlab software based on the Sequential Extraction of Late Exponentials (SELE) algorithm.⁴⁷ All virus incubation with lectins and glycosidases was performed at room temperatures, except for that with SA lectin NA which was performed at 37 °C. DLS studies of the virus with glycosidases and lectins were repeated for multiple concentrations of the proteins and included a control without the virus to optimize the protein concentration, time of incubation, and to rule out aggregation originating from other sources (e.g., protein–protein aggregation) (see Supporting Information 5).

3. RESULTS AND DISCUSSION

3.1. Confirming Virus Attachment to AFM Probe.

NHS-terminated linkers were assembled on gold-coated AFM probes via chemisorption of thiol group at the other end. The NHS group traps overlaid virus by reacting with amines on membrane proteins. Since the virus is ~80 nm in diameter, that is, larger than the spacing between linkers in assembled monolayers, each particle is expected to bind to the gold surface of an AFM probe with several short linkers. To verify if the virus remains structurally intact through the attachment procedure and can survive the repeated push and pulls in force spectroscopy experiments, the following test was performed.

The probe-attached virus was repeatedly pushed into APTES-coated mica until 1 nN of indentation force was reached and withdrawn (Figure 1B). When the probe approaches the APTES surface (i.e., separation between the probe and surface decreases), the indentation force rises as the virus compresses against the surface (blue curve in Figure 1C). The force rise is not steady, however, but flattens for several nanometers before rising again. A leveling midway in the indentation force typically signifies buckling-like deformations in shell and micelle materials, indicating that the membrane and capsid structure in the probe-bound virus are physically intact. The resistance profile of the probe-bound virus being pushed into a surface is similar to that observed when surface-adsorbed virus was indented by a bare probe,²⁶ confirming that virus was attached to the AFM probe without significant alterations in structure.

Indentation thickness is the separation or depth the virus is compressed before ~1 nN of resistance is registered in the approach curve. The indentation thickness would drop to zero if virus becomes dislodged from the probe or completely squished. Over the course of repeated pounding into the APTES surface (~90 ramps shown in Figure 1D), the indentation thickness of the virus remained around 40 nm. The overall consistency in indentation thickness confirms that the probe-bound virus was able to sustain repeated ramping without detaching, rupturing, or collapse. The indentation thickness showed a modest downward trend with ramp number (Figure 1E), suggesting that in this experiment the continuous pounding could be irreversibly compressing or stiffening the virus.

When the virus-bound probe was retracted after compressing into the APTES surface, a negative adhesion force was observed (Figure 1C, red curve), indicating that the positively charged APTES surface was resisting the virus pull out. The distribution of adhesion forces and energies endured by the virus during the course of the repeated ramping is summarized in Figure 1E,F. Since the indentation thickness did not shift to zero (Figure 1D), the attachment between the probe and virus can survive the 0.9 nN of pull-out force and 40 KJ of adhesion energy encountered in this experiment without detaching. In cases where the virus was pushed into the APTES surface with a greater indentation force of 2 nN, greater adhesion forces were registered during retraction, without dislodging or disrupting the virus on the probe (data not shown). From the range of adhesion forces survived in these experiments, the attachment between virus and probe can be considered stronger than the adhesion forces to be sampled between glycans in subsequent experiments.

3.2. Rate-Dependent Shift from Slime- to Velcro-like Self-Adhesion between Complex N-Glycan Shields.

We tested the adhesion between the complex N-glycan shields on two viruses, one bound to the probe and the other adsorbed on APTES (Figure 2A). The viruses were brought into contact until an indentation force of ~1.2 nN was registered, so that only the interactions between glycan shields were sampled.²⁶ As the two virus glycan shields retracted after contact, the nature of adhesion release changed with the indentation rate. At the lower rates (toward 50 nm/sec), the release occurred in one predominant peak around the point of pullout between the two viruses (Figure 2B). Such short-range “brittle” or Velcro-like adhesion releases were observed earlier between two monolayers of mannose residues.²⁵ At the higher indentation rates (toward 500 nm/sec), however, the adhesion released in multiple peaks which increased in magnitude even further from the contact point between the two surfaces (Figure 2D). Such long-range sawtooth-like releases were observed earlier between two SA monolayers,and can be characterized as “slime-like” (releasing in spurts over long-range). In between the higher and lower rates, the adhesion release was usually a mix of the two types (Figure 2C).

In >500 curves (multiple viruses, adsorption surfaces, and stocks) analyzed per indentation rate, the slime-like adhesions occurred predominantly at higher ramp rates (above 300 nm/sec) whereas “Velcro-like” adhesions were frequent at the lower ramp rates (Figure 2E). At intermediate rates, adhesion curves with both features (“mixed”) were likely to occur. There were no instances where the two viruses retracted without adhesion.

Quantifiable features of the force curves like adhesion length and adhesion energy also exhibited rate-dependent trends (Figure 2F,G). The adhesion lengths clustered around 100 nm at the higher indentation rates (consistent with long-range slime-like release) and around 20 nm at the lower rates (consistent with short-release Velcro-like release at these rates) (Figure 2F). The adhesion energy distributed around larger values at the higher indentation rates (consistent with long-range release) and shifted to lower values at the lower rates (consistent with short-range Velcro-like adhesions) (Figure 2G). The indentation thickness, however, did not change between the different rates (Figure 2H), indicating that a similar population of viruses was sampled at all the rates

3.3. Rate-Dependent Shift in Adhesion Pattern Relates to Rate-Dependent Penetrability of Complex N-Glycan Shield.

VSV-G glycoproteins have two glycosylation sites in each trimer, and early sequencing studies identified both of these sites to have complex N-glycans, irrespective of the cell in which the proteins are produced.⁴³ Therefore, at first approximation, the VSV-G glycan shield is expected to display SA residues at the surface and mannose residues within the core, and therefore slime-like adhesions observed previously between SA–SA monolayers was expected to occur between the two glycan shields at all rates. Instead Velcro-like adhesions observed previously between the man–man monolayer occurred at low rates of indentation, and the SA–SA-like adhesions occurred at only the higher rates. It is possible that high-mannose and hybrid N-glycans displaying surface mannose residues were also present in the glycan shield, and patches with high surface mannose content were being being sampled more at the lower indentation rates. To rule out the possibility that this was the reason for the rate-dependent shift in adhesion patterns, the same spot on the glycan shield of two viruses was approached and retracted at multiple rates. The Velcro-like release at lower rates shifted to slime-like sawtooth release at higher rates even when at the same location (Figure 3A), confirming that the rate-dependence shift was not due to spatial variations in glycan types. Lectin binding studies in solution (Section 3.6) also verified that there were no patches with high surface mannose content on the virus.

Because there was no indication that variations in surface residues were causing the rate-dependent shift from SA- to mannose-like adhesions, it is possible that these rate-dependent shifts may be due to a rate-dependent access to the deeper mannose core or rate-dependent penetrability of the glycan shield. To test this premise, virus adsorbed on APTES surfaces was indented at different ramp rates with a bare AFM probe up to a maximum force of ~1 nN (Figure 3B). At the higher ramp rates (toward 500 nm/sec), the resistance to indentation increased steadily in approach and decreased steadily with retraction (Figure 3C,D). Though there was hysteresis (i.e., the approach and retraction curves do not overlap due to permanent deformation occurring in the indented virus), an indentation force response that is proportional to the indentation strain indicates that the virus material is being compressed as a whole at these rates, without significant penetration of the material. However, at lower indentation rates (toward ramp rates of 50 nm/sec) the force rise in approach was interspersed by plateaus and negative slopes (black arrows in Figure 3E,F). Sharp negative slopes typically indicate forced penetration of the material layers of the virus layers. The first penetration slope occurring at ~400 pN is likely to be in the soft glycan shield layer of the virus. Correspondingly, the retraction curves showed sharp negative forces (red arrow in Figure 3F), which occur if the bare tip is stuck at a penetrated layer and a buildup in retraction force is required to release it. Figure 3G is a frequency plot of the incidents of negative slopes in virus approach force curves taken at different ramp rates. The higher rate approach curves were more likely to have 0–1 incidence of negative slopes (blue arrow), whereas the lower rates were more likely to have 3–4 incidents of negative slope (orange arrow). Therefore, a possible explanation for the rate-dependent shift in glycan–glycan adhesion patterns is that the deeper mannose residues are being sampled at the lower rates where the glycan shield can be penetrated, whereas the surface SA residues are being sampled at the higher ramp rates where the glycan shield deforms as a whole. However, this explanation supposes that the surface SA residues are responsible for the slime-like adhesion and deeper mannose residues are responsible for the Velcro-like adhesion. We verify this supposition in the following experiment.

3.4. Slime-like and Velcro-like Self-Adhesions Are Localized at SA and Mannose Sugars, Respectively.

β-N-acetylglucosaminidase (NAG) exposes the mannose core by cleaving the β (1–2) glycosidic bonds between mannose and the GlcNAc residues above it (Figure 4A). When two viruses with exposed mannose core were brought in contact and retracted, only Velcro-like adhesions were seen at low and high rates (Figure 4B-D). The disappearance of the sawtooth negative forces confirms that the cleaved SA residues are the source of the slime-like adhesion and not mannose residues. β-mannosidase enzyme was used to remove the mannose core by cleaving the β (1–4) glycosidic bonds between the mannose sugars and the GlcNAc residues below the core (Figure 4E). When two viruses lacking mannose core but having exposed GlcNac residues were brought into contact and the retracted, there was no self-adhesion at all rates (Figure 4F-H). The complete disappearance of Velcro-like adhesions when the mannose core is removed validates the latter to be the source of these short-range adhesions between N-glycan shields (Figure 4D).

Figure 4.

Nature of adhesive pull outs between two viruses when their mannose cores are exposed (right) and removed (left). (A) Complex N-glycan shield is cleaved with enzyme β-N-acetylglucosaminidase to expose the mannose core. (B) Representative indentation and retraction force curves between two viruses with exposed mannose core show Velcro-like adhesions at lower ramp rates (~50 nm/sec). (C) Representative FS force curves between two viruses with exposed mannose core exhibit Velcro-like adhesions also at high ramp rates (~300 nm/sec). (D) Distribution of adhesion types between two mannose cores as a function of ramp rate showing man-man Velcro-like interactions occurring at all rates. (E) Complex N-glycan shield is cleaved with glycosidase β-mannosidase to remove mannose core and expose underlying GlcNAc residues. (F) Representative indentation and retraction force curves between two viruses with no mannose core and with exposed GlcNAc residues and showing no adhesion at low ramp rates (~50 nm/sec). (G) Representative force curve between two GlcNAc exposed viruses showing no adhesion also at higher ramp rates (~300 nm/sec) rates. (H) No self-adhesions were evident at all ramp rates between viruses with no mannose core and exposed GlcNAc residues.

The rate dependence of quantifiable features of the force response in both cases is shown in Supporting Information 3. The adhesion length and adhesion energy hovered around smaller values for all rates when the mannose core is exposed (consistent with man–man forces being short-range) and became negligible when the mannose core is removed (consistent with mannose core being the source of the Velcro-like interactions). The indentation thickness was distributed similarly at all rates, confirming that comparable virus populations were being sampled. The indentation thicknesses for mannosidase-treated viruses (i.e., no mannose core) were however larger than the uncleaved (buried mannose core) and NAG-treated (exposed mannose core) virus, indicating a possible decrease in the structural stiffness of the virus following removal of the mannose core.

3.5. Velcro-like Adhesion between Mannose- and SA-Terminated Complex Glycan Shields.

It was observed previously that there were no adhesions between monolayers of SA and mannose.²⁶ To understand how the monolayer findings with single sugars translated to multisugar glycan shields with terminal SA and mannose residues, we performed the following study. The virus on the AFM probe was cleaved with NAG to expose the mannose core (i.e., terminal mannose) and brought in contact with uncleaved virus (terminal SA) (Figure 5A). Velcro-like adhesions were observed at all rates between the SA-terminated and man-terminated complex glycan shield (Figure 5B). A sample force curve is shown in Figure 5C. The adhesion energy, for instance, distributed toward Velcro-like adhesion values at all rates (Figure 5D). Velcro-like adhesions were also predominantly observed when the same position on terminal mannose- and SA-glycan shields was contacted at different rates (Figure 5E). Even though the two glycan shields have terminal mannose and SA residues which are nonadhering in monolayers, penetrability and access to the mannose core appear to determine their net adhesion in glycan architecture. Frequency plots of the adhesion force and indentation thickness quantified at multiple rates are provided in Supporting Information 4. The observation is similar to a previous report where pure mannose monolayers were able to penetrate complex N-glycan shields to elicit Velcro-like or brittle adhesions with the latter’s mannose core.²⁶ Break throughs in the indentation force, indicating penetration, were only seen when the virus glycan shield was probed with mannose monolayers, but not with the SA monolayers. We had also showed there were no cross interactions between SA–man and gal–man monolayers. Taken together, there appears to be a general design theme that mannose-terminal surfaces can penetrate SA-terminated glycan shields in forced contact to latch on to the latter’s mannose core.

Figure 5.

Cross adhesion between complex glycan shields that are SA- and man-terminated. (A) Virus attached to probe was cleaved by NAG to expose mannose core. (B) Representative interaction curve showing the man-man Velcro-like interaction that occurs at all rates. (C) Frequency plot showing adhesions classified as man-man dominate at all rates. (D) Distribution of adhesion energy quantified from force spectroscopy curves showing similar distribution at all rates (n > 20 for each ramp rates, 3 samples, and 2 runs). (E) Change in adhesion profile at the same point when indentation rates are changed; man-man adhesion releases were dominant at low rates whereas few mixed release occurred at higher rates in this case.

3.6. Virus Self-Aggregation in Solution Is Determined by Surface Residues.

We investigated how the self-adhesion patterns in force spectroscopy translated to self-aggregation in solution of viruses with different sugars of the N-glycan shield terminally exposed. Glycosidases were added to virus solution, and virus aggregation was tracked after 1 h incubation by changes in the DLS correlation curves. Lectins were added to verify the identity of the terminally exposed sugars. Virus aggregation and binding with lectins would increase the size of the diffusing species in solution (i.e., the hydrodynamic diameter D_H) which is detected in DLS by a rightward shift of the virus correlation curve.

Sialic Acid Exposure (Figure 6A).

Figure 6.

(A) Verifying SA exposure on virus glycan shield by tracking shifts in the virus DLS correlation curve. The virus solution aggregated with SA lectin MAA (correlation curve shifts to the right) and in the presence of Ca ion (virus correlation curve shifts to right) but not with mannose–lectin ConA (no rightward shift of correlation curve). (B–F) Self-aggregation propensity of each tier of complex N-glycan sugars in solution at 25 °C. As each tier of N-glycan sugar is exposed by glycosidases (shown on left), the DLS curve moves to the right if self-aggregation of viruses with the exposed sugar tiers occurs and increases the size of the diffusing particles. Aggregation by SA–lectin MAA and Ca²⁺ were also indicated by rightward shifts of the virus DLS curve, and it indicates the presence of terminal SA residues, whereas rightward shifts produced by ConA addition indicate the presence of terminal mannose residues.

SA exposure on the virus was tested by the interaction with SA lectin MAA, mannose lectin ConA, and by aggregation with Ca²⁺ ions. The hydrodynamic diameter of the virus is about ~200 nm in PBS. When MAA is added to the virus solution, the DLS correlation curve shifts to the right with the appearance of diffusing species in the range of 3 μm and indicating aggregation of the virus population. When Con A (with 100 μM Ca ions) was added to virus solution, no rightward shift of the DLS curve indicating aggregation was observed. The lectin binding results confirm that the HIV-1 R-E-/VSV-G virus has terminal SA residues and no significant patches of terminal mannose that would be aggregated by mannose lectins. While surface-bound viruses exhibited slime-like self-adhesion in force spectroscopy, there was no corresponding self-aggregation between the freely diffusing virus in solution. It is possible that the virus self-aggregation in solution is prevented by the charge repulsion between surface SA residues preventing close approach. Ca²⁺ ions, which can screen charge repulsion, are known to effectively bind the glycerol side chain of SA.⁵² Large-scale self-aggregation of the virus occurred with Ca addition, consistent with both the glycan shield being covered by SA residues and the presence of slime-like adhesions between SA residues. The DLS correlation curve shifted rightward as aggregates of 2 μm D_H appeared in solution.

Galactose Exposure (Figure 6B).

When Neuraminidase was added to virus solution to cleave SA and expose galactose residues, there was no rightward shift in the DLS curves even after 1 h incubation, indicating absence of gal–gal self-aggregation. The complete removal of SA residues was confirmed by the absence of aggregation with SA–lectin MAA and Ca²⁺ ions (no rightward shift). There was no aggregation with mannose–lectin ConA either. The absence of aggregation between Gal-terminated viruses is consistent with the force spectroscopy observation of no adhesion between two galactose monolayers. It also confirms that only the surface SA residues were contributing to the slime-like adhesions; even if there were terminal galactose residues on the uncleaved virus, gal–gal interactions would not contribute to the self-adhesions between the virus.

Mannose Exposure (Figure 6C,D).

When the virus was cleaved with NAG to expose the mannose core, the hydrodynamic size increased from ~200 to ~500 nm (Figure 6C). Removal of sugars is expected to decrease virus size, but the observed increase suggests that the viruses with exposed mannose core aggregate. Neither the SA–lectin MAA nor Ca ions aggregated NAG-cleaved virus, verifying there was no uncleaved terminal SA. Addition of ConA produced large aggregation, verifying that mannose residues are exposed. These results confirm that the Velcro-like self-adhesion observed between mannose cores in FS also manifest in solution. To check if the branched presentation of mannose in the N-glycan core is critical for Velcro-like self-adhesion, we removed the higher mannose residues with α-mannosidase (Figure 6D). Again, there was progressive aggregation, indicating that it is mannose residues that foster the self-aggregation and not necessarily the branched core architecture. SA-lectin and Ca ions did not aggregate the α-mannosidase cleaved virus verifying complete removal of SA sugars, and mannose–lectin Con A aggregated the virus consistent with mannose exposure.

GlcNac Exposure (Figure 6E).

When β-mannosidase was added to the virus solution to remove the mannose core, no size increase corresponding to the aggregation of viruses was observed. Cleavage of the distal residues was confirmed by the absence of aggregation with MAA, ConA, and Ca²⁺⁺ ions. The lack of aggregation is consistent with force-spectroscopy findings in Figure 4E-H where there were no self-adhesions between GlcNAc residues on surface bound virus.

Figure 7A summarizes the glycan linkage at which the different glycosidases cleave and the terminal sugar is exposed as a result. The above results suggest that, unlike in force spectroscopy where penetrability of the glycan shield determined self-adhesion, only the terminal sugar determines self-aggregation of virus in solution. To further check this observation, we compared the self-aggregation of viruses with terminal GlcNAc residues that were either above the mannose core (produced by galactosidase cleavage) or below the mannose core (produced by β-mannosidase cleavage) (Figure 7B). Should glycan shield penetrability determine self-aggregation in solution, then the former with a mannose core beneath would exhibit different aggregation patterns than the latter which does not have a mannose core. The absence of self-aggregation in both cases, however, confirms that in solution the nonadhesive character of the terminal GlcNAc residues supersedes any penetrated adhesion from the mannose core. Figure 7C is a summary of the changes in D_H of the virus solutions as each tier of N-glycan sugars is exposed. Self-aggregation occurs (i.e., large jump in D_H) only when mannose residues are exposed or when SA residues are in the presence of Ca²⁺ ions.

Figure 7.

Self-aggregation pattern of N-glycan sugars in solution depends on the terminal sugar and not on penetrability of the glycan shield. (A) Summary of the different cleavages performed and the terminal sugars exposed in solution as a result. (B) DLS correlation curves comparing virus solutions with GlcNAc-terminal viruses having mannose core (g, generated by galactosidase cleavage; left schematic in blue box) and GlcNAc-terminal viruses having no mannose core (m, generated by mannosidase cleavage; right schematic in black box). Both GlcNAc-terminal viruses do not show self-aggregation (i.e., no rightward shift of the DLS curves) indicating that penetrability and access to mannose core does not govern virus self-aggregation in solution. (C) Hydrodynamic diameters of the diffusing species in solutions of viruses with different sugar terminals showing that only mannose and SA-terminated virus exhibit self-aggregation but with the latter requiring calcium ions.

4. CONCLUSION

The goal of this study was to investigate how the placement of sugars in N-glycans affects the adhesion properties of biological surfaces. All N-glycans have a core of mannose residues, and the complex-type N-glycans have surface sialic acid residues. An HIV virus pseudotyped with VSV-G proteins presenting complex N-glycans was used as a biologically relevant platform for studying sugar interaction. The morphological and mechanical characterization of this HIV-1 R- E-/VSV-G virus was described earlier.²⁶ The low virus indentation forces (<2 nN) observed in this study generally support the conclusions of the earlier one that pseudotyping with VSV-G proteins weakens the structural stiffness of the native HIV-1 virus. The complex N-glycan shields of viruses were bought into contact by force spectroscopy (FS) and by free diffusion in solution. Superficial sugars were cleaved using glycosidases to expose underlying sugars. Changes in self-adhesion in FS and self-aggregation in solution were tracked to understand the rules of sugar–sugar adhesion in N-glycan presentation.

Slime-like and Velcro-like Adhesions among N-Glycan Sugars.

The interactions between two N-glycan shields followed previous findings with single sugar monolayers that there were two types of self-adhesion among the constituent sugars: (i) brittle or Velcro-like self-adhesions which release in single peaks close to the interface and originate from mannose residues; and (ii) tough or slime-like adhesions which release in multiple spurts over distances beyond the interface and originate from SA residues.²⁶ Removal of the respective sugars from the N-glycan shield led to loss of the respective adhesion patterns. It appears that sugars retain their self-adhesion identity even when present in complex branched structure with other sugars, or that the N-glycan architecture is specifically designed to foster and not interfere with the self-adhesion propensity of its constituent sugars. The N-glycan architecture, however, can restrict access to the deeper residues based on the rate and force of the interaction.

Glycan Shield Penetrability Determines Adhesion Patterns in Force Spectroscopy.

When both SA and mannose are presented in N-glycan architecture, the rate-dependent penetrability of the glycan shield appears to determine the type of self-adhesion manifested. At higher rates of indentation, the penetrability of the complex glycan shield decreases, a bare AFM tip compresses the glycan shield as a whole, and the slime-like self-adhesions from the surface SA residues dominate. At low rates, the penetrability of the glycan shield increases, a bare tip indents the shield with sharp rise and falls in force indicating penetration, and Velcro-like self-adhesions between buried mannose cores dominate.

It is also not clear what the source of the rate dependence in glycan penetration is, but the rate dependence disappeared when SA and Gal residues were together removed from at least one of the N-glycan shields. For instance, mannose-terminated surfaces (be it NAG-cleaved virus or mannose monolayers) were able to penetrate SA-terminated complex glycans on uncleaved virus to elicit Velcro-like adhesions with the latter’s mannose core at all the rates sampled.

Adhesion rules between N-glycan complexes were found different for free (solution) and forced (force spectroscopy) contact. This conclusion suggests that aggregation patterns between virus and other glycosylated entities would change between solutions that are in stagnant (where entities will be freely diffusing), laminar (where entities will be in low-rate forced contact approach), and turbulent (where entities will be in high-rate forced contact approach) flow conditions. It is well documented that cells aggregate differently in different shear conditions, but the role of glycan adhesion in facilitating these changes has not been explored.^48-51

Self-Aggregation for Freely Diffusing Virus in Solution Determined by Terminal N-Glycan Sugar.

When freely diffusing in solution, only the virus with mannose and SA-terminal sugars self-aggregated, which is also consistent with only these two sugars showing self-adhesion in FS. Mannose-terminated glycans, however, did not require Ca²⁺ ions to self-aggregate whereas SA-terminated required Ca²⁺ ions. Mannose surfaces were observed to self-adhere both in the presence and absence of ions²⁵ in monolayer and glycan presentation.²⁶ SA surfaces, however, required Na⁺ ions to self-adhere in FS, and Ca²⁺ ions additionally to self-aggregate in solution.²⁶ The other two N-glycan sugars, galactose and GlcNAc, did not promote self-aggregation in solution when displayed at the surface. This solution-level observation is also consistent with the FS observations for these sugars: Galactose monolayers did not self-adhere in FS,²⁵ and neither did N-glycan shields with only GlcNAc residues (Figure 7E-H). It appears that when the virus is freely diffusing in solution, the terminal sugar alone determines the self-aggregation pattern. This conclusion is also supported by the observation that GlcNAc terminated viruses did not self-aggregate in solution irrespective of whether a mannose core was present beneath the GlcNAc layer or not. For instance if the virus aggregation was driven by the entire sequence of glycan sugars, the sequence with a mannose core would show different characteristics than the sequence without a mannose core. However, this was not the case. Therefore, unlike in FS, penetrable access to the mannose core did not determine self-aggregation patterns in freely diffusing virus. The present study did not explore how the ramp-rate dependence of penetration observed in FS translated to shear-rate dependence of aggregation in solution.

Rules of Interaction Applicable beyond N-Glycan Presentation of Sugars.

It appears that the conformation of mannose does not appear critical for the Velcro-like self-adhesion to manifest. For instance, viruses with terminal mannose self-aggregated in solution irrespective of whether the first or second tier of the mannose core was exposed. Artificial monolayers of disordered mannose residues also exhibited self-adhesion.²⁵ We can also look at another system where mannose is abundant but in a different conformation. The endosperm of seeds (dates, coffee beans, ivory nuts, and so forth) is rich in polymers of mannose called mannans which serve as dense stores of energy for the seeds.⁵³ X-ray crystallography reveals organized hydrogen-bonding across stacked mannose faces as responsible for the tight packing of these polymers.^54-56 Similar short-range hydrogen bonding interactions could be responsible for the Velcro-like adhesions between mannose residues in glycans. The tight packing among mannose polymers is loosened in mucilaginous materials (guar gum, locust bean gum, cassia gum) by interspersing the polymer with galactose.^53,57 The greater the galactose/mannose ratio is, the higher the swelling or loose-packing of the galactomannan polymers is.⁵⁸ The relation between galactomannan swelling and galactose content agrees with our findings that galactose does not adhere to other galactose or to mannose, and therefore interferes with the otherwise tight-packing between mannose polymers. The aggregation pattern of galactose and mannose in galactomannans is consistent with their adhesion patterns in N-glycans, supporting the conclusion that the rules of self-adhesion between sugars observed in this study may be applicable beyond the N-glycan system.

Rules for Sugar–Sugar Interaction Have Biomedical and Bioengineering Implications.

It appears that the precise presentation of the sugars is not as critical a factor for sugar–sugar adhesions as it is for protein–sugar binding. The rules for sugar self-adhesion have been consistent through monolayer–monolayer, monolayer–glycan shield, and glycan shield–glycan shield studies, even thought the presentation patterns changed between the systems. In the case of the protein–sugar binding, however, the presentation of the sugar is known to be critical because it determines how the sugars fit within binding pockets on lectins.^59-62 This would imply that the rules for sugar–sugar interactions can be translated to other engineered systems having different presentations of sugars. Sugar-coated nanomedicines have been used for dispersing gold nanoparticles and inorganic magnetic nanoparticles in biosensing and theranostic applications,^63,64 for targeting labeling probes and gene cargo to specific cells based on their sugar receptor profile,^65-68 and for adsorbing and stabilizing bioactive drugs on ceramic cores in aquasome delivery vehicles.^69-71 Sugar–sugar adhesion rules could help understand how these sugar-coated vehicles would distribute nonspecifically on the glycosylated surfaces encountered in the host. The composition of N-glycans changes with age and disease.⁷² SA and mannose residues, which influence how neurons interconnect, are diminished in neural disorders such as multiple sclerosis.^73-75 Distal SA decreases during aging but distal galactose increases.^76-78 Mannose expression increases in breast cancer⁷⁹ and SA expression is altered in cancer.⁸⁰ Environmental stresses also change the N-glycan composi-tion.^81-83 Our findings suggest that the self-adhesion patterns between cells and between cells and pathogens will change drastically with changes in the glycosylation sequence. Establishing guiding rules for how changes in sugar composition alter the interfacial adhesion environment will provide a new perspective for interpreting the biological phenomena occurring in these conditions.