Onion-like glycodendrimersomes from sequence-defined Janus glycodendrimers and influence of architecture on reactivity to a lectin

Significance

The known role of the sugar d-mannose (Man) as a postal code in intracellular cargo routing has herein inspired the design of Man-presenting synthetic glycolipid-like mimics termed Janus glycodendrimers (GDs). Simple injection of a solution of Janus GDs prepared into a water-miscible solvent into buffer produces, via self-assembly, monodisperse multilamellar onion-like glycodendrimersomes (GDSs). Janus GD structural design impacts the resulting GDS architecture including surface display of Man. The latter is shown to tune reactivity to a lectin. Thus GDSs provide a model system to enable systematic studies of physiologically relevant glycan/lectin pairing.

Abstract

A library of eight amphiphilic Janus glycodendrimers (GDs) with d-mannose (Man) headgroups, a known routing signal for lectin-mediated transport processes, was constructed via an iterative modular methodology. Sequence-defined variations of the Janus GD modulate the surface density and sequence of Man after self-assembly into multilamellar glycodendrimersomes (GDSs). The spatial mode of Man presentation is decisive for formation of either unilamellar or onion-like GDS vesicles. Man presentation and Janus GD concentration determine GDS size and number of bilayers. Beyond vesicle architecture, Man topological display affects kinetics and plateau level of GDS aggregation by a tetravalent model lectin: the leguminous agglutinin Con A, which is structurally related to endogenous cargo transporters. The agglutination process was rapid, efficient, and readily reversible for onion-like GDSs, demonstrating their value as versatile tools to explore the nature of physiologically relevant glycan/lectin pairing.

Supramolecular chemistry has enormous potential to help resolve fundamental questions in the realm of cell biology. One of the key challenges is the design of programmable models for vesicles and cells and their surfaces as a means of establishing a chemical platform that mimics natural features in size and shape, and also allows customized implementation of bioactive epitopes, in structural and topological terms. Natural complexity can conveniently be reduced to simple systems, whose degree of diversity can then be rationally reconstituted in a stepwise process. Focusing on surface properties, the recently gained access to uniform populations of stable glycodendrimersomes (GDSs) by a simple injection method using a solution of amphiphilic Janus glycodendrimers (GDs) as building blocks for self-assembly in a water-soluble aprotic solvent (1–3), has afforded a promising opportunity to realize this concept. Interestingly, the resulting GDSs, which have tunable surface features, can cover the size range of naturally occurring vesicles such as endo- and exosomes.

When considering the choice of surface determinants, a natural class of molecules to study comes from the third alphabet of life, carbohydrates. In fact, glycan structures, unsurpassed in coding capacity relative to proteins and nucleic acids, serve as signals in many bioprocesses, from cell adhesion and growth regulation to diverse transport processes (4–9). Consequently, a wide array of synthetic products for the many roles of glycans must be prepared by diversity-oriented approaches (10–15). The self-assembly of Janus GDs into GDSs enables analysis of trans (bridging) contacts. Initial experiments with unilamellar GDSs have proven their reactivity with sugar receptors (lectins) (16). Thus, model studies of the impact of glycans on vesicle properties and on glycan–lectin recognition are now possible. The studies will be valuable in clarifying how glycan structure and topology of presentation on cell surfaces team up to achieve the physiological level of selectivity and specificity of lectin binding to distinct counter receptors (17).

In this article, generation of onion-like GDSs is demonstrated for some d-mannose (Man)-presenting Janus GDs. Of note, they are structurally related to physiological multilayered vesicles (18–20). The carbohydrate headgroup Man was selected because this sugar is relevant for cellular transport. A group of intracellular cargo transporters are mannose-specific lectins, i.e., ERGIC-53, ERGL, VIP36, and VIPL, and they share the β-sandwich fold and Ca²⁺ dependence (21–24). However, their nature as type-1 membrane proteins hampers studies on GDS aggregation (agglutination). Close similarities between the tissue and legume lectins, resulting in their classification as leguminous-like (L type) (25, 26), provide the motivation to run the assays with the tetrameric leguminous lectin Con A (ConA), a popular model often used in glycocluster research (7, 9, 10, 12). The distance between two lectin sites for ConA binding to the surface of the same Janus GDS is about 64 Å or 70 Å, and the cations are bound to ConA to hold amino acid side chains in place for ligand binding (25). These aggregation studies unravel the impact of sequence-defined surface display of Man on lectin-dependent bridging between vesicles (rate and extent) and on stability of the aggregates under conditions of impaired protein–carbohydrate interactions.

Results and Discussion

An accelerated modular synthetic strategy was used to generate a library of eight Man-presenting Janus GDs. The library was extended from the mono- and bivalent and mixed-type headgroups of the amphiphilic Janus GDs, previously shown to result in ConA-reactive unilamellar GDSs (27, 28), to five further sequence-defined Janus GDs. Details on their synthesis are shown graphically in SI Appendix, Figs. S1–S3. Methoxytriethoxy fragments were used as spacers or space-filling chains to minimize chemical surface heterogeneity, except for the lectin-reactive group. Indeed, for various types of lectins, this surface design has been found to be inert (16, 27–29). Also, the presence of methoxytriethoxy has previously been shown to be conducive to the formation of onion-like dendrimersomes (3). This result prompted the use of Janus GDs as building blocks for obtaining onion-like GDSs. As shown in Fig. 1 and SI Appendix, Fig. S5, the surface density of the sugar and its relative positioning on the scaffold result in different surface topologies after the Janus GDs self-assembled into bilayers. This change on an otherwise constant chemical background was significant for yielding new products. The self-assembly process of several Janus GDs of this panel resulted in nonunilamellar vesicles. From 5a-Man onward and 3-Man, onion-like GDSs, resembling the structural organization of multilamellar vesicles, were obtained (Fig. 2 and SI Appendix, Fig. S6), and 4-Man formed according to structural analysis results to be published elsewhere, bicontinuous glycodendrimercubosomes (SI Appendix, Fig. S6).

Fig. 1.

Summary of Man-presenting amphiphilic Janus GDs. Their diameter (D_DLS, in nm) and polydispersity (in the parentheses) were measured by dynamic light scattering (DLS) with 0.1 mM of Man in 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (Hepes) buffer (10 mM, pH = 7.4).

Fig. 2.

Selected cryo-TEM images of onion-like GDSs self-assembled from 0.1 mM Janus GD 5a-Man 3EO (1,2,3,4,5,6,7)-3EOMan(8)-3EO(9) and their 3D intensity-plotting images with different numbers of bilayers and diameters.

The diameter of the onion-like GDSs is linearly related to the number of bilayers (Fig. 3 and SI Appendix, Fig. S7) and directly connected to the Janus GD concentration (SI Appendix, Table S1 and Fig. S8). Overall, the new findings are that the nature and concentration of a Janus GD’s headgroup determine the size and architecture of the GDSs. The various arrangements of the bilayers and the GDSs resulting from using a fixed concentration of 0.1 M Man in Janus GDs are presented in Fig. 3. At the same concentration, Janus GDs with d-lactose (Lac) established GDSs that were invariably unilamellar (16, 29). Intriguingly, when the monosaccharide d-galactose (Gal) was incorporated as part of the amphiphilic Janus GD 3-Gal (SI Appendix, Fig. S4), the occurrence of several bilayers was noted (SI Appendix, Fig. S12). This observation demonstrates a previously unknown influence of seemingly subtle changes in headgroup structure of Janus GDs for the properties of GDSs. Of conspicuous relevance due to their occurrence on membranes, the chemical nature of the sugar appears to be relevant for the mode of structural organization of Janus GDs into GDSs, with the possibility for a similar activity of glycans in glycolipids in cellular vesicles. As outlined in the Introduction, a toolbox of reagents (here Janus GDs with different headgroups) and a simple test system, starting with morphological assessment, can help to dissect physiologically important parameters toward reaching an understanding of their interrelationship. The current lack of understanding of the details of the structure–activity relationship that underlies self-assembly of Man- (and Gal-) presenting Janus GDs into the obtained multilamellar vesicles notwithstanding, the produced populations of bi- to multibilayered GDSs enables the study of their ligand properties, i.e., their bioactivity with the leguminous lectin, to answer questions of how sugar presentation affects bridging of GDSs/glycodendrimercubosome by ConA.

Fig. 3.

Illustration of unilamellar GDSs (1-Man, 2-Man), onion-like multilayer GDSs (3-Man, 5a-Man, 5b-Man, 6a-Man, and 6b-Man), and glycodendrimercubosomes (4-Man) self-assembled from Man-presenting Janus GDs (0.1 mM of Man) in Hepes. The diameter (D_DLS, in nm) and polydispersity (in the parentheses) were measured by DLS. The numbers of bilayers of onion-like GDSs were calculated from diameter D_DLS and interbilayer distance d as detailed in SI Appendix, Fig. S7 from the linear correlation between the number of bilayers and the diameter of GDS determined by cryo-TEM.

GDSs formed by injection of THF solutions of Janus GDs in Hepes (Methods) are stable in time and upon dilution with additional Hepes (SI Appendix, Figs. S9–S11), indicating their nonequilibrium states. Testing was carried out in Hepes buffer (pH = 7.4) containing 1 mM Ca²⁺ and 1 mM Mn²⁺ to keep the cation-binding sites saturated and ConA active (see Fig. 5 B and C). The lectin-dependent cross-linking of the GDS/glycodendrimercubosome preparations was routinely monitored spectrophotometrically at 450 nm, following an increase of turbidity due to aggregation with time (up to 400 s). No turbidity increase was measured in the absence of ConA, excluding self-aggregation of GDSs. A carbohydrate-independent GDS/glycodendrimercubosome reactivity to the lectin was excluded by a lack of change in absorbance after mixing GDSs/glycodendrimercubosome bearing the noncognate sugar Lac with ConA (Fig. 4A). This control aspect, i.e., to check for carbohydrate-independent aggregation, was further backed by addition of a β-sandwich protein with reactivity toward Lac, i.e., human galectin-3 (Gal-3) to Man-presenting GDSs. A measurable turbidity in the solution was not induced over a period of 1,000 s, although this lectin is an agglutinin (SI Appendix, Fig. S14). Thus, the response was strictly dependent on the presence of the cognate pair of Man-bearing GDSs with ConA, and the presence of inhibitors Man (100 mM)/EDTA (20 mM) completely abolished the activity (please see below). Structural properties other than the binding of Man are therefore not relevant for ConA-dependent bridging between GDSs. The formation of such aggregates was determined in terms of the plateau level and the initial rate.

Fig. 5.

Rate of change in turbidity, k, of GDSs and glycodendrimercubosomes with ConA calculated from the curves in of GDSs (0.1 mM of Man, 900 μL) with ConA (0.5 mg·mL^–1, 100 μL) in Hepes (1.0 mM CaCl₂ and 1.0 mM MnCl₂) as in Fig. 4A at t_1/2, where t_1/2 is the time at which the observed absorbance is equal to half of the plateau absorbance (A). Each rate of GDS was the average value from triplicate measurements. Representative conformation (B) of tetrameric structure of ConA with Ca²⁺ (green), Mn²⁺ (purple), and (C) its binding site loaded with methyl α-d-mannopyranoside (Protein Data Bank ID code 5CNA).

Fig. 4.

Agglutination assays between different Man-presenting GDSs and glycodendrimercubosomes with ConA. (A) Man-presenting GDSs with identical concentration (0.1 mM of Man in 900 µL of Hepes, 1.0 mM CaCl₂ and 1.0 mM MnCl₂), (B) Man-presenting GDSs with identical sizes by using the method detailed in SI Appendix, Fig. S8 (145 ± 15 nm, in 900 µL of Hepes, 1.0 mM CaCl₂ and 1.0 mM MnCl₂) were incubated with ConA (0.5 mg·mL^–1 in 100 µL of Hepes, 1.0 mM CaCl₂, and 1.0 mM MnCl₂). The molar attenuation coefficient ε = A/(cl), adapted from Beer–Lambert law, where A = plateau OD value, c = molar concentration of Man, and l = semimicro cuvette path length (0.23 cm). Control experiments were carried out by incubating GDSs generated from 3-Lac 3EO(1,2,3)-3EOLac(4), 0.1 mM of Lac in 900 µL of Hepes (1.0 mM CaCl₂ and 1.0 mM MnCl₂) with ConA (0.5 mg·mL^–1 in 100 µL of Hepes, 1.0 mM CaCl₂ and 1.0 mM MnCl₂). Chemical structure of 3-Lac was described in SI Appendix, Fig. S4.

Following confirmation of previously reported data on 1-Man, 2-Man, and 3-Man at a constant concentration of Man (27), experiments under identical conditions revealed a strong activity of the glycodendrimercubosome from 4-Man and the onion-like GDSs consisting of 5a/b-Man and 6a/b-Man, as graded by the molar attenuation coefficient (Fig. 4A). When experiments were run at a constant size of 145 ± 15 nm and data normalized, a similar grading was obtained, with 4-Man holding the prominent position with a coefficient of 40 (Fig. 4B). Among the onion-like GDSs, relative positioning (sequence) of Man on the scaffold (5a/b-Man) and length of the linker (6a/b-Man) indicate a clear impact of both sequence and spacer length on agglutination (Fig. 4) and its rate (Fig. 5). This is in in agreement with results on Lac-presenting GDSs (29). In unilamellar GDSs (1-Man, 2-Man) the concentration of Man on the outside and inside part of assembly can be statistically estimated to be 50%. However, in multilamellar GDSs the Man concentration on the outside part of the assembly must be very low. Despite this, multilamellar GDSs with identical diameter are much more efficient than the unilamellar assemblies (Fig. 4B) although the onions are underestimated. Therefore, the difference in efficiency is due to Man density, sequence, and spacer length when compared at identical diameter (Fig. 4B). In contrast, without its implementation into an intimate structural context that likely has some spatial flexibility, the Man headgroup of 4-Man is best suited to adapt to binding to a lectin, especially when strengthening an initial contact by a second pairing. In the case of the tetravalent ConA, two sites on each side have a fixed distance (Fig. 5B); when fully loaded with ligand, this yields a contact pattern of two Man moieties bound by two of the four ConA sites per GDS in an aggregate. Such a spatial adaptation process probably takes some time to occur so that it is reasonable that the rate of change in turbidity was lower for the 4-Man than onion-like GDSs from 5a-, 5b-, 6a-, and 6b-Man (Fig. 5A). Conversely, if the lectin also introduces spatial adaptability to the system, as natural or engineered human tandem-repeat-type galectins with a linker peptide do, then plateau levels are increased relative to linkerless variants (28) and the difference in the attenuation coefficient is reduced (29).

In addition to the concentration- and size-normalized extent of aggregation and its rate, another parameter of GDS–lectin interaction is the stability of the aggregates in an environment with blocking compounds. As mentioned above, no aggregation occurred when adding a mixture of the cognate sugar for saturating lectin sites (Man) and the chelating agent (EDTA) for disrupting the carbohydrate-binding site by cation removal (Fig. 5C). Interestingly, a markedly reduced but not lost activity was traced for GDSs from 1-Man/2-Man and 3-Man in the presence of Man but not the Man/EDTA mixture (Fig. 6 and SI Appendix, Fig. S13). Evidently, a high local density of surface Man can compete with the free sugar for lectin binding. Fittingly, addition of Man only partially led to aggregate dissociation of 1-Man/2-Man (Fig. 6A and SI Appendix, Fig. S13A). Under these conditions, the locked conformation of active ConA (26) was apparently sufficiently stable, and EDTA addition only led to a minor, if any, enhancement (Fig. 6B and SI Appendix, Fig. S13B). Slow formation with low plateau level was coupled with resistance to aggregate dissociation by inhibitors. An inverted relationship was found in the other aggregation processes of GDSs from 4-Man onward (Fig. 6 and SI Appendix, Fig. S13). Although secondary interactions cannot be precluded (30), a kinetically favored mechanism involving a decrease in the macroscopic off-rate resembling mucin precipitation by the soybean agglutinin (31) can be operative. Lowering cognate sugar density will then increase the off-rate, a boon for transient recognition phenomena, when present in context with physiological inhibitors, e.g., glycan branches on a cell surface. Alternatively, smaller GDSs (1-Man and 2-Man) may be more resistant to the addition of Man/EDTA due to their larger number of inter-GDSs contacts via ConA linkers. By comparison, larger GDSs may be involved in fewer surface-to-surface interactions and therefore would be expected to agglutinate and deaggregate faster. Clarification of this mechanism will be reported elsewhere. Physiologically, dense clusters can be attained in microdomains, with association of glycolipids to glycoproteins with their branched N- and O-glycans. Thus, the panel of sequence-defined Janus GDs was instrumental to trace density as a molecular switch between comparatively slow aggregate formation resistant to dissociation and rapid generation with high susceptibility to inhibitor presence.

Fig. 6.

Agglutination assays between the Man-presenting GDSs from 2-Man, 3-Man, 5a-Man, and 6b-Man with ConA. Man-presenting GDSs (0.1 mM of Man in 900 μL of Hepes, 1.0 mM CaCl₂ and 1.0 mM MnCl₂) were incubated with ConA (0.5 mg·mL^–1 in 100 μL of Hepes, 1.0 mM CaCl₂ and 1.0 mM MnCl₂) with 100 mM of Man (blue line) or Lac (red line). A high concentration (100 mM) of Man solution in Hepes (100 μL) was added at t = 50 s into Man-presenting GDSs (0.1 mM of Man in 900 μL of Hepes, 1.0 mM CaCl₂ and 1.0 mM MnCl₂) with ConA (black line) (A). Man-presenting GDSs (0.1 mM of Man in 900 μL of Hepes, without Ca²⁺ and Mn²⁺) were incubated with ConA (0.5 mg·mL^–1 in 100 μL of Hepes, without Ca²⁺ and Mn²⁺) (blue line) or with Man (100 mM) and EDTA (20 mM) in Hepes (100 μL, without Ca²⁺ and Mn²⁺) as inhibitors (red line). Man (100 mM) with EDTA (20 mM) solution in Hepes (100 μL, without Ca²⁺ and Mn²⁺) was added at t = 50 s into Man-presenting GDSs with ConA (black line) (B).

Conclusions

In the quest to simulate the natural multivalent surface display of cellular glycans, the self-assembly of specific Janus GDs into onion-like GDSs is intriguing considering the presence of naturally occurring multilamellar vesicles. Importantly, the type of sugar and the chemical context of its presentation govern the generation of the onion-like structures. Results in this article establish a basis for systematically studying the influence of the physiological sugar alphabet on functionality. On GDS surfaces, sugar determinants are presented to facilitate rapid lectin-dependent aggregation, especially when the headgroup is spatially adaptable to lectin sites in a fixed constellation. These experiments with a leguminous lectin ConA demonstrate the bioactivity of the GDS-surface-presented ligands. They therefore inspire further experiments with cellular cargo transporters, bivalent galectin-4 being a prominent candidate (32). The sensitivity of the test system, i.e., the aggregation assay, has already been documented to be high. It can even detect a difference in the capacity for aggregate formation upon slight changes of surface density (by sequence-defined display or by using Janus GD mixtures for GDS assembly) or of protein properties (by testing natural single-site variants due to polymorphism at the gene level) (28, 29, 33). With regard to cargo transport by GDSs, the proven resistance of proteins such as the β-sandwich galectins to exposure to an aprotic solvent (34) makes the consideration of corresponding GDS loading possible. The versatility of chemical programming of the bilayer composition will enable the discovery of novel structure–activity relationships of glycans as structural organizers and as docking sites for tissue lectins on GDSs.

Methods

GDSs and glycodendrimercubosomes were generated by injection of 100 μL of the stock solution of amphiphilic Janus GDs in freshly distilled THF into 2.0 mL Hepes buffer, followed by about 5 s vortexing.

Three-dimensional intensity plotting images of multilayer GDSs were generated from their cryogenic transmission electron microscopy (cryo-TEM) images by ImageJ (v1.50a) with interactive 3D plotting plugin in invert fire LUT mode.

Agglutination assays of GDSs or glycodendrimercubosomes with ConA were monitored in semimicro disposable cuvettes at 23 °C at single wavelength λ = 450 nm by using a Shimadzu UV-vis spectrophotometer UV-1601 with Shimadzu/UV Probe software with kinetic mode. One hundred microliters of Hepes solution of ConA was injected into 900 μL of Hepes solution of GDSs or glycodendrimercubosomes. The cuvette was shaken for about 2 s before starting recording the absorbance change in time, with the same GDSs or glycodendrimercubosomes solution in reference cuvette. Hepes solutions of ConA were kept at 0 °C ice bath before the agglutination assays.