Targeting Multiple Binding Sites on Cholera Toxin B with Glycomimetic Polymers Promotes Formation of Protein–Polymer Aggregates

Abstract

The canonical binding site on the B subunit of cholera toxin (CTB) binds to GM1 gangliosides on host cells. However, the recently discovered noncanonical binding site on CTB with affinity for fucosylated molecules has raised the possibility that both sites can be involved in initiating intoxication. Previously, we showed that blocking CTB binding to human and murine small intestine epithelial cells can be increased by simultaneously targeting both binding sites with multivalent norbornene-based glycopolymers [ACS Infect. Dis. 2020, 6, 5, 1192–1203]. However, the mechanistic origin of the increased blocking efficacy was unclear. Herein, we observed that mixing CTB pentamers and glycopolymers that display fucose and galactose sugars results in the formation of large aggregates, which further inhibits binding of CTB to human granulocytes. Dynamic light scattering analysis, small-angle X-ray scattering analysis, transmission electron microscopy, and turbidimetric assays revealed that the facial directionality of CTB promotes interchain cross-linking, which in turn leads to self-assembly of protein–polymer networks. This cross-linking-induced self-assembly occurs only when the glycopolymer system contains both galactose and fucose. In an assay of the glycopolymer’s ability to block CTB binding to human granulocytes, we observed a direct correlation between IC₅₀ and self-assembly size. The aggregation mechanism of inhibition proposed herein has potential utility for the development of low-cost macromolecular clinical therapeutics for cholera that do not have exotic architectures and do not require complex synthetic sequences.

Graphical Abstract

INTRODUCTION

Cholera is still a life-threatening illness with an estimated annual incidence of 2.9 million cases and an estimated 95,000 deaths annually in endemic countries.¹ Underdeveloped countries bear the majority of the burden, owing to inadequate sanitation, poor hygiene and lack of vaccination in affected areas. The main sources of infection are drinking water and/or food contaminated with Vibrio cholerae. Although healthy individuals must receive a high dose (10⁸ organisms) to develop illness, infection can occur at lower doses among people with elevated gastric pH. Acute cholera infection lasts about 4–8 days, during which time Vibrio cholerae reside in the gastrointestinal tract, where they cause severe fluid loss and are excreted together with the characteristic rice-water stool; most of the fluid loss occurs in the jejunum.^2–5 After the bacteria colonize the small intestine, they secrete cholera toxin (CTX), a protein complex consisting of A (CTA) and B-subunits (CTB). CTB exists as a stable pentamer under physiological conditions.⁶ CTX binds to receptors on epithelial cells via the 56 kDa CTB pentamer. After being endocytosed, CTX undergoes retrograde transport to the endoplasmic reticulum, where it releases the noncovalently attached 28 kDa CTA. In the cytosol, CTA activates the G protein G_sα, leading to enhanced cyclic AMP production. The resulting increase in the intracellular cyclic AMP concentration triggers extreme ion secretion, which in turn creates an osmotic gradient, allowing water to transport across the epithelial cell membrane, eventually leading to massive fluid loss for the patient.^{5, 7}

The current treatment for all but the most severe cases is oral or intravenous rehydration. Antibiotics are generally not recommended as a first approach, because their use can lead to antimicrobial resistance among other microorganisms, but are used in more severe cases.^8–9 Killed, live-attenuated, and oral cholera vaccines are available. However, the vaccination efficacy is about 65–70% and even lower among children in endemic areas. Serotype–specificity of the vaccination, which does not result in cross-immunity toward multiple serogroups of V. cholerae, could also be a problem.^10–11 Neutralization of CTX by inhibitors that bind CTB has therefore been suggested and evaluated in many studies.^12–14 Monosialotetrahexosylganglioside (GM1) has been known as the canonical ligand for CTB due to its strong affinity for CTB (K_d = 43 nM).^15–16 Therefore, many studies have adopted GM1 as inhibitors for targeting receptor-binding process.^17–18

Although this strategy shows promise in vitro, GM1 gangliosides are sparsely expressed in human small intestinal epithelial cells.^19–20 Moreover, although GM1 was long thought to be the only receptor for CTB, a GM1-independent binding modality of CTX was recently discovered.²¹ In this pathway, CTB utilizes a noncanonical binding site to bind fucosylated glycans.²²

Previously, we demonstrated that glycopolymers displaying fucose block binding of CTB to human colonic cells.²³ Inhibitors with multivalent scaffolds can show greatly improved avidity toward toxins through additional binding mechanisms that include chelate effect, subsite binding, steric stabilization, statistical rebinding and receptor clustering.^{17–18, 24–25} This suggests the potential utility of a novel strategy involving inhibition of CTX intoxication by simultaneous targeting both the canonical and the noncanonical binding sites of CTB. It is highly unlikely that a small ligand could simultaneously block more than one type of binding site on a CTB, but simultaneous binding might be achievable by macromolecular architectures that are capable of accessing both binding sites.²⁶ Indeed, a functionalized dextran that presents both binding sugars has been reported recently although it is 10-fold less effective at blocking CTB binding to immobilized GM1 than dextran only displaying galactose.²⁷

In our own work, we developed multivalent glycomimetic polymers that target both binding sites of CTB. We found that norbornene glycopolymers displaying both β-d-galactose and α-l-fucose (Gal₅₀Fuc₅₀) block CTB binding to both human and murine small intestine epithelial cells more effectively than glycopolymers displaying α-l-fucose (Fuc₁₀₀), glycopolymers displaying β-d-galactose (Gal₁₀₀), or a mixture of these two homoglycopolymers.^{23, 28} However, the mechanism for the greater effectiveness of Gal₅₀Fuc₅₀ relative to Gal₁₀₀ and Fuc₁₀₀ was unclear. Here, we propose that CTB and Gal₅₀Fuc₅₀ self-assemble into cross-linked CTB_n–glycopolymer networks, and we discuss how the mechanism for the self-assembly process might facilitate the development of potential therapeutics for cholera.

EXPERIMENTAL SECTION

Materials.

CTB was purchased from Sigma-Aldrich (C9903, United States). In this paper we have only used CTB as an approximation of the fully assembled CTX to avoid unnecessary toxicity risks.

Glycopolymers

Glycopolymers were prepared and characterized as previously described.²⁸

Small-Angle X-ray Scattering.

Small-angle X-ray scattering (SAXS) measurements were conducted at Brookhaven National Laboratory National Synchrotron Light Source II on the Life Science X-ray scattering beamline, 16-ID. Each sample (60 μL) and its matched buffer were exposed to X-rays for 1 s for 5 repeats at 25 °C. Buffer subtraction is performed based on the height of the water peak at ~2.0 Å⁻¹, and the transmitted beam intensity is measured by an intensity monitor built into the beam stop, a photodiode that measures the visible light converted from the direct X-ray beam by a YAG crystal.²⁹ Scattering image collection and data reduction were performed with the py4XS Python scripts. SAXS data were analyzed with the SasView 4.2.2 software package (http://www.sasview.org/). The scattering intensities were calculated as described in the literature³⁰

$$I\left(q\right)=\frac{1}{v}\int d\psi \int d\varphi \int p\left(\theta ,\varphi ,\psi \right)F^2\left(q,\alpha ,\psi \right)\text{sin}\left(\alpha \right)d\alpha$$

where

$$F^2\left(q,\alpha ,\psi \right)=2J_1\left(a\right)\text{sin}\left(b\right)/ab$$

and

$$a=q{r}^{\prime }\text{sin}\left(\alpha \right),b=q\frac{L}{2}\text{cos}\left(\alpha \right),$$

$$r^{\prime }=\frac{r_{\mathit{\text{minor}}}}{\sqrt{2}}\sqrt{(1+v^2)+(1+v^2\text{cos}\left(\psi \right)}$$

The angle θ, ϕ, ψ are defined as the cylinder axis relative to the beam angle, the rotation about the beam, and the rotation about the cylinder axis, which are fixed to the values 90°, 0°, 0°. The angle α is the angle between the axis of the cylinder and q. The elliptical cylinder model includes the length of the cylinder L, minor radius r_minor and major radius, r_major which is calculated as r_major = v · r_minor where v is the axis ratio.

Dynamic Light Scattering.

For dynamic light scattering (DLS) analysis, CTB was reconstituted in distilled deionized water to achieve a 1.0 % (w/v) solution. When CTB was reconstituted with water to a final concentration of 1 mg of CTB per mL, the solution contained 0.05 M Tris-buffer, pH 7.5, 0.2 M NaCl, 3 mM NaN₃ and 1 mM sodium EDTA. CTB in solution was considered as a pentamer based on the DLS and SAXS results. Glycopolymers were dissolved in distilled deionized water to achieve a 1.0 % (w/v) concentration and then diluted to 0.01 % (w/v) with the same buffer used for CTB reconstitution. Sample solutions (50–100 μL) were filtered through a 0.45 μm polyvinylidene difluoride filter and then transferred to disposable polystyrene cuvettes. CTB solution (50–100 μL) was then transferred to the cuvette containing glycopolymer solution and thoroughly agitated by pipetting the mixture. Mixtures of glycopolymers and CTB were allowed to settle for 2 min to assure a homogeneous dispersion. Samples were analyzed on a 90 Plus Particle Size Analyzer (Brookhaven Instruments, 35 mW red diode laser with a wavelength of 659.0 nm, scattering angle 90°, 25 °C). Each signal was recorded for 15 s, and the signals for 10 consecutive runs were averaged.

Transmission Electron Microscopy.

Transmission electron microscopy images were obtained at 80 kV with a transmission electron microscope (JEOL JEM-1400) equipped with a CCD camera (Gatan ORIUS SC1000B). Samples with glycopolymer alone were vapor stained using ruthenium tetroxide, 0.5% solution in water. Freshly prepared 1:1 (v/v) mixtures of a 0.01 % (w/v) glycopolymer (Glc₁₀₀, Gal₁₀₀, Fuc₁₀₀, or Gal₅₀Fuc₅₀) in distilled deionized water and 0.01 % (w/v) CTB in buffer were deposited onto formvar/carbon-coated copper grids (300 mesh) and vacuum dried prior to analysis. No staining was used for the samples containing CTB.

Optical Microscopy.

A 1.0 % (w/v) CTB solution in buffer (50 μL) was added to a 0.01 % (w/v) Gal₅₀Fuc₅₀ in buffer (50 μL) and allowed to settle for 2 min to assure a homogeneous dispersion. The solution of aggregates was cast onto glass microscopy slides and imaged on a Zeiss Axiovert microscope (0.55 NA, 20× objective).

Cellular Assessment of Glycopolymer Block

Blood donated anonymously through Sahlgrenska University hospital blood bank was used to investigate binding of CTB- and ovalbumin-biotin (biotinylated using the kit Lightning-Link Rapid Biotin, Innova) to granulocytes. The blood samples were processed to analyze CTB binding to cells in the presence or absence of glycopolymer. The majority of RBCs were lysed using ACK-buffer (Gibco, Thermo Scientific) according to the manufacturer’s instructions. The cells were stained with mAb against CD66a/c/e-PE (Biolegend) and CTB-biotin (that had been preincubated with or without glycopolymers) or OVA-biotin. After washing cells, streptavidin-BV421 and the live/dead marker Zombie Red (Biolegend) were added. Cells were then analyzed by flow cytometry on a Fortessa 20X (BD Biosciences) and the data analyzed using FlowJo software (Version 10, Tree Star).

Turbidimetric Assay.

Turbidimetric assays were performed with a Shimadzu UV–vis spectrophotometer (UV-2550). Solutions of Gal₅₀Fuc₅₀ and CTB were prepared by serial dilution of Gal₅₀Fuc₅₀ in water (1.0 w/v %) and CTB in buffer (1.0 w/v %). The glycopolymer solutions (40 μL) were added to the CTB solution (40 μL); the resulting solutions were briefly agitated, and their absorbance at 580 nm was monitored over time.

Acute Oral Toxicity Assay.

The animal experiment operated under the federal assurance #A3011–01. Five 6–8-week-old female CD-1 mice were treated by oral gavage with 100 μM Gal₅₀Fuc₅₀ in PBS (200 μL single oral gavage, fasted overnight prior to gavage). Control mice (five) received PBS (200 μL single oral gavage). All the mice were sacrificed after 24 h, and blood and jejunums were harvested for histopathology.

Cytotoxicity Assay.

The method outlined by Miret et al.³¹ was used to assay the cytotoxicities of the glycopolymers in human liver hepatocytes (HepG2 cells), which had originally been purchased from ATCC (HB-8065) and were cultured from HepG2 frozen stock grown in minimum Eagle medium (MEM, Corning) supplemented with 10% FBS and 0.1% Pen/Strep(Gibco). Briefly, HepG2 cells were suspended in 10 mL of supplemented MEM in a 100 mm sterile petri dish (BD Falcon). Once confluency was reached, the cells were detached from the dish with trypsin EDTA, and the solution was centrifuged at 2000 rpm for 5 min to pellet the cells. The cells were then resuspended in 2–5 mL MEM; a 200 μL aliquot of the cells was removed and stained with 20 μL of trypan blue; and the cells were counted with an automated cell counter (Cellometer Auto T4, Nexcelom Bioscience).

The remaining cells were then diluted to a volume of 10K cells/100 μL in supplemented MEM without Pen/Strep antibiotic and 100 μL aliquots of the diluted cells were then added to each of 60 wells of a sterile flat-bottom 96 well plate (Falcon), which was incubated overnight at 37 °C (humidified, 5% CO₂) to attach the cells to the wells. Glycopolymers were dissolved in sterilized distilled deionized water to yield a 20 mM stock solution. The stock solution (5 μL) was added to 995 μL of MEM in the first well of a 24 well plate to give a 100 μM solution, and MEM (500 μL) was added to the next nine wells. Serial dilution was performed, leaving the last well as a MEM-only well to serve as a no-compound control. The medium in each of the 60 cell-containing wells of the 96 well plate was carefully aspirated away, and 100 μL of each stock solution was added to the 96 well plate by means of a 200 μL multichannel pipettor. Technical triplicates were performed on each 96 well plate, allowing two compounds per plate with nine compound concentrations and a no-compound control per plate. Glycopolymers were incubated with the HepG2 cells for 72 h, at which point 10 μL of alamar blue reagent was added to each well. After 4 h incubation with the alamar blue, the UV absorbance of each well was measured with a microplate reader (BioTek) at 570 nm with a reference wavelength of 600 nm. Data were analyzed by determining the concentration of oxidized alamar blue in each well. Half-maximal inhibitory concentrations (IC₅₀) are report as micromolar drug concentrations.

RESULTS AND DISCUSSION

We prepared glycopolymers by means of ring-opening metathesis polymerization (ROMP) of monosaccharide-bearing norbornene derivatives, because this method allows for the synthesis of glycopolymers with controlled molecular weights and narrow dispersities.³² In addition, glycopolymers prepared by means of ROMP have shown various advantages for biological applications; especially notable is their ability to induce protein clustering.^{25, 33–35}

β-d-Galactose and α-l-fucose were chosen as the two major CTB-binding ligands on the basis of their affinity toward the two binding sites on CTB: β-d-galactose binds to the canonical binding site, α-l-fucose binds to the noncanonical site, and β-d-glucose does not bind to either site. We hypothesized that a random copolymer system comprising both β-d-galactose and α-l-fucose (Gal₅₀Fuc₅₀), would simultaneously occupy both CTB-binding sites and thus show the highest inhibitory efficacy against CTX-mediated intoxication. Ruthenium-catalyzed ROMP of the peracetylated glycomonomers followed by deacetylation afforded multivalent glycopolymers (Figure 1A, Glc₁₀₀, Gal₁₀₀, Fuc₁₀₀, and Gal₅₀Fuc₅₀). As a control, we also prepared a random copolymer consisting of β-d-glucose and α-l-fucose (Glc₅₀Fuc₅₀). Detailed procedures for the synthesis of the peracetylated glycomonomers and the peracetylated and deacetylated glycopolymers have previously been reported.²⁸ All of the glycopolymers had 50–100 copies of one or two monosaccharides randomly distributed along the polymer backbone. The distribution of monosaccharide side chains is expected to yield a statistical copolymer. The molecular weights of the peracetylated glycopolymers are listed in Table 1, and the molecular weights of the corresponding deacetylated glycopolymers were calculated by subtracting the molecular weights of the appropriate number of acetyl groups (based on the DP of the glycopolymer) from the molecular weights in Table 1. Complete deacetylation was confirmed by ¹H nuclear magnetic resonance spectroscopy.

Figure 1.

(A) Glycopolymer structures and (B) representative cartoon for Gal₅₀Fuc₅₀.

Table 1.

Molecular Weights of Peracetylated Glycopolymers.

Polymer	Mtheor	Mna	Mwa	Ða	DPa
pGlc100b	51,300	35,600	44,000	1.28	69
pGal100b	51,300	55,400	61,300	1.11	108
pFuc100b	45,500	59,500	66,300	1.12	131
pGlc50Fuc50c	48,400	51,700	65,800	1.27	107
pGal50Fuc50b	48,400	54,800	62,800	1.15	113

^aDetermined by means of gel permeation chromatography with static light scattering and refractive index detectors.

^bPolymer preparations were previously reported.²⁸

^cPrepared for this study. The GPC trace is presented in Figure S1.

We estimated the maximum glycopolymer size attainable based on a fully stretched linear chain. Specifically, we used a C=C–C unit length of 0.116 nm and a C–C–C unit length of 0.126 nm, with torsion angles of 120° and 109.5°, respectively. Accordingly, the maximal linear length of a norbornyl polymer repeat unit comprising one C=C–C unit and four C–C–C units would be 0.62 nm. Therefore, the maximum length of an uncoiled glycopolymer with a DP of 100 would be 62 nm. However, it is highly unlikely that such a polymer would remain uncoiled, owing to the presence of many amide bonds and pyranoses, which would form hydrogen bonds to form irregularly coiled structures that are shorter than 62 nm.

We obtained SAXS data for the glycopolymers to gather information about their conformations in distilled deionized water. Specifically, we investigated the solution behavior of 1.0 w/v % Glc₁₀₀, Gal₁₀₀, Fuc₁₀₀, Gal₅₀Fuc₅₀, and mixtures of each glycopolymer with CTB. The spectrum of water was used as the background signal, which was subtracted from the sample data. Data points with large uncertainties due to a low signal in the high-q range were omitted from the analysis. Detailed analysis of SAXS data was performed over a q range of 0.005 to 0.6 Å⁻¹, which corresponds to structures with length scales of approximately 0.628 to 125.6 nm. Figure 2 shows SAXS data for polymers and mixtures with fits to an elliptical cylinder form factor model and the results are summarized in Table 2.

Figure 2.

SAXS plots of 1.0 % (w/v) glycopolymers and glycopolymer–CTB mixtures and fits to the elliptical cylinder model. The spectra are vertically offset by 30% with respect to Gal₅₀Fuc₅₀ for clarity, and at q = 0.5 the I value is identical for all samples and equal to 0.3.

Table 2.

SAXS data fitted to an elliptical cylinder form factor model.^a

Sample	rminor (Å)	rmajor (Å)	Length (Å)
Glc100	8.7 ± 0.1	48.0 ± 0.5	185.8 ± 1.0
Glc+CTBb	8.2 ± 0.1	46.1 ± 1.2	184.8 ± 2.1
Gal100	9.0 ± 0.1	44.8 ± 0.7	198.5 ± 1.7
Gal+CTBb	8.3 ± 0.1	44.6 ± 0.5	225.2 ± 1.5
Fuc100	9.5 ± 0.1	32.6 ± 0.6	145.1 ± 1.1
Fuc+CTBb	9.1 ± 0.1	35.7 ± 0.4	175.3 ± 1.0
Gal50Fuc50	7.7 ± 0.1	44.6 ± 0.9	189.8 ± 1.7
GalFuc+CTBb	8.2 ± 0.1	47.8 ± 0.5	232.0 ± 1.6

^aErrors on fit parameters were estimated with SASView software and are based on the goodness-of-fit.

^bCTB was introduced as a solution in buffer.

The glycopolymer sizes determined by means of SAXS were in the 20 nm range, which is consistent with the coiling of 62 nm glycopolymers into elliptical cylinders. SAXS data for CTB were also analyzed and fitted to stacked disks model, which is the model closest to the actual structure of CTB; the result was a disk with a thickness of 3.0 nm and a diameter of 6.6 nm (Table S1, Figure S2). These values are consistent with the X-ray crystal structure of CTB (5ELB).^{22, 36} The lengths of Glc₁₀₀ and Gal₁₀₀ were proportional to their molecular weights, but Fuc₁₀₀ was the shortest glycopolymer despite having the highest molecular weight. This discrepancy in the size of Fuc₁₀₀ may have been due to the fact that fucose lacks a C-6 hydroxy group, making the glycopolymer less hydrophilic than the others and thus more prone to chain collapse and a consequent decrease in size.

Next, we treated the glycopolymers, 0.01 % (w/v), with excess CTB, 1.0 % (w/v), in Tris-buffer at pH 7.4, and we then used SAXS to evaluate the effect of CTB binding on the size of the resulting complexes. No change in the size of Glc₁₀₀ was observed upon treatment with CTB. In contrast, Gal₁₀₀, Fuc₁₀₀, and Gal₅₀Fuc₅₀ increased slightly in size, which is consistent with their ability to block CTB binding to cell surfaces. However, subtle changes in SAXS profile were detected after addition of CTB. Although the structures of the glycopolymers (Glc₁₀₀, Gal₁₀₀, Fuc₁₀₀, and Gal₅₀Fuc₅₀) were unaffected by the presence of CTB, the sizes were affected to some extent.

Interestingly, we observed the formation of a precipitate when CTB was mixed with the random copolymer Gal₅₀Fuc₅₀ but not when it was mixed with a homopolymer (Glc₁₀₀, Gal₁₀₀, or Fuc₁₀₀). We believe that our SAXS experiments detected species that remained in solution after the precipitate had formed; that is, the insoluble material was outside the SAXS detection window. Unfortunately, the insolubility of the precipitate prevented us from performing structural analyses that require the analyte to be in solution. Therefore, we undertook light scattering studies to analyze the CTB-glycopolymer complex structures.

We used dynamic light scattering (DLS) to explore the effect of CTB binding on the mean hydrodynamic diameters (D_h) of the glycopolymer particles (Figure 3). The mean hydrodynamic diameters of the glycopolymer particles in Tris-buffer solution ranged from 299 to 437 nm suggesting that the particles consist of 16–18 entangled chains considering the length of glycopolymers from SAXS. When the homopolymers were mixed with CTB, no obvious change in particle size was observed (Figure 3A–C), which is consistent with the SAXS data fitted to elliptical cylinders. In contrast, mixing CTB with Gal₅₀Fuc₅₀ increased the particle size by an order of magnitude indicating large aggregate formation (Figure 3E). To determine whether both galactose and fucose were required for aggregate formation, we prepared Glc₅₀Fuc₅₀, which contains β-d-glucose rather than β-d-galactose, and we mixed it with CTB.

Figure 3.

Histograms of measured hydrodynamic diameters (D_H) of glycopolymers (colored bars) and their mixtures with CTB (dotted colored bars). The D_H of CTB was measured to be 5.5 nm (Figure S3). CTB solution (1.0 w/v% in buffer) was added to glycopolymer solutions (0.01 w/v%) and the DLS signal acquired for 2.5 minutes. For the experiment shown in panel F, CTB solution was added to a premix of equimolar amounts of Gal₁₀₀ and Fuc₁₀₀.

Under these conditions, aggregate formation was not observed, which is consistent with β-d-glucose having no affinity for CTB. In addition, the size change upon addition of CTB was negligible (Figure 3D), a result that implies that both galactose and fucose were required for aggregate formation. Furthermore, to determine whether aggregate formation required that the galactose and fucose be within a single chain, we prepared an equimolar blend of Gal₁₀₀ and Fuc₁₀₀ and mixed the blend with CTB. Surprisingly, this experiment resulted in the formation of aggregates that were half the size of those formed by Gal₅₀Fuc₅₀, suggesting that β-d-galactose and α-l-fucose independently interacted with CTB to form a lower degree of cross-linking. (Figure 3E,F).

We next assessed the capacity of the glycopolymers to block the binding of CTB to natural ligands. Previously, we reported that Fuc₁₀₀ and Gal₅₀Fuc₅₀ block CTB binding to a fucosylated glycan, triLe^x, but Gal₁₀₀ does not and that Gal₁₀₀ blocks CTB binding to GM1 but Gal₅₀Fuc₅₀ does not.²⁸ The incomplete blocking of GM1 binding to CTB by Gal₁₀₀ may have been due to the fact that the Gal of CTB-Gal₁₀₀ does not have multiple points of interaction like GM1 has with CTB.³⁷ Blocking of CTB binding to human primary intestinal epithelium can be accomplished with both galactosylated and fucosylated polymers.²⁸

As accessibility to primary human intestinal samples is limited we opted for using primary human granulocytes from blood and assessed by flow cytometry the capacity of the different glycopolymers to block binding of CTB to these cells.

Our previous studies have indicated that human granulocytes and intestinal epithelial cells have similar CTB binding profiles.³⁸ Figure 4 shows CTB binding plotted against increasing glycopolymer concentration where 100% binding shows the level of CTB binding to granulocytes without polymer addition. We found that in human granulocytes, glycopolymers containing either galactose or fucose significantly blocked CTB binding, whereas Glc₁₀₀ did not. Gal₅₀Fuc₅₀ showed the highest blocking potency among the glycopolymers. The capacity of the different glycopolymers to block CTB binding to granulocytes thereby mimics our previous results obtained with human primary intestinal epithelial cells.²⁸ An equimolar blend of Gal₁₀₀ and Fuc₁₀₀ was almost as potent as Gal₅₀Fuc₅₀ at blocking CTB binding to human granulocytes, a result that is consistent with the different complexes formed as observed by DLS. When we calculated IC₅₀ values for the glycopolymers (Figure 4), we found that even though all the glycopolymers had the same number of ligands per chain, systems containing both galactose and fucose (that is, Gal₅₀Fuc₅₀ and the equimolar blend of Gal₁₀₀ and Fuc₁₀₀) were significantly more potent than any of the homopolymers (Gal₁₀₀, Fuc₁₀₀) by two orders of magnitude.

Figure 4.

Effect of glycopolymer concentration on blocking of CTB binding to human granulocytes obtained from anonymous blood donors. Cells were stained with markers for granulocytes and CTB (0.5 μg/mL) that had been preincubated in the absence of glycopolymers or in their presence at various concentrations. The fluorescence signal from labeled CTB on the cells was then measured by flow cytometry (Fortessa X20, BD), and data were analyzed with the FlowJo software package (Tree Star). Only live single-cell granulocytes were included in the analysis. The graph shows data from three human donors, and the signal (geometric mean fluorescence index, gMFI) of unblocked CTB binding was set as 100% for each donor. Error bars indicate SDs. n.a.; Not available, no significant blocking was observed for Glc₁₀₀. Data for Gal₁₀₀ could not be calculated, because 100% blocking could not be achieved with this glycopolymer. An equimolar mixture of Gal₁₀₀ and Fuc₁₀₀ was used for Gal₁₀₀+Fuc₁₀₀. Statistical analysis was performed using 2-way ANOVA.

We hypothesize that the better inhibitory efficacy of Gal₅₀Fuc₅₀ relative to that of homopolymers arises from complex formation that is mediated by facial directionality. Furthermore, we observed a correlation between IC₅₀ and CTB–glycopolymer complex size: lower IC₅₀ values were observed for the larger complexes. This suggests that the aggregates that formed sequester CTB units from solution, which in turn might halt initiation of intoxication. Pieters and colleagues have found that multivalent dextran-based CTB inhibitors that contain both fucose and a meta-nitrophenyl α-galactoside show lower efficacy than the latter alone.^{27, 39} Moreover, the IC₅₀ values of these dextran-based inhibitors are far higher than those of our norbornene-derived glycopolymers (Figure 4).²⁷ Gibson et al^40–41 investigated the effect of glycan density and observed that lowering galactose density increases the inhibition efficacy. However, we do not observe a similar trend in our system. Mixing two homopolymers with similar glycan density is almost as effective as a copolymer displaying each ligand at half the glycan density of the homopolymers. We assume that this difference is due to the sugar display on a flexible, cylinder backbone that supports extended structures and thus the cross-linked aggregates to form The greater hydrophobicity of the norbornene-based glycopolymers may also play an important role in the binding of CTB to glycopolymers.⁴²

To explain the trend we observed, we propose the model system shown in Figure 5. In this system, the ligands of the glycopolymer point radially outward to maximize surface tension (Figure 1B). This orientation provides no directionality when the glycopolymer binds to CTB. In contrast, the two types of ligand-binding site are located on opposite faces of CTB: the canonical binding sites are on the bottom face of the CTB, and the noncanonical sites are located on the anterior sides of the top face. (Figure 5A).^21–22 As a consequence, the orientation of CTB relative to its bound glycopolymer depends on the binding site occupied. Three scenarios are possible. In one scenario, there is no interaction between the glycopolymers and CTB (Figure 5B), as is the case for Glc₁₀₀, which does not have affinity for either of the binding sites. No noticeable change in DLS signal was observed after addition of CTB in DLS (Figure 3A). In another scenario, interaction occurs but does not involve interchain cross-linking (Figure 5C, D). We contend that both Gal₁₀₀ and Fuc₁₀₀ bind to CTB in this mode.

Figure 5.

(A) Facial directionality of CTB and (B–F) proposed modes for binding of glycopolymers to CTB: (B) Glc₁₀₀, no interaction with CTB, (C) Gal₁₀₀, interchain cross-linking is quenched by CTB, (D) Fuc₁₀₀, interchain cross-linking is quenched by CTB, (E) Gal₁₀₀+Fuc₁₀₀, interchain cross-linking occurs, and (F) Gal₅₀Fuc₅₀, interchain cross-linking occurs.

Small aggregates are formed if a single ligand is present on the polymer chain. Several CTB can bind along the linear span of a Gal100 chain to form these small aggregates. However, the converse cannot occur. Multiple Gal100 chains are prevented from binding to a single CTB due to the close proximity of the Gal binding sites located on a single face of CTB. Thus, upon Gal ligand binding to one or two canonical sites the remaining Gal sites are blocked by the steric bulk of the polymer chain.²⁵ Statistical and chelate effects ensure that CTB remains bound to the same polymer chain even if individual binding interactions are weak (Figure 5C). Losing the ability to attract another free Gal₁₀₀ chain minimizes the formation of large aggregates. Due to the facial orientation of the fucose-binding sites, the non-canonical binding sites remain accessible to additional polymer chains. The third possible scenario involves interchain cross-linking (Figure 5E, F). In this mode, unoccupied binding sites on a heteroglycopolymer–CTB complex can be occupied by the ligands of another heteroglycopolymer chain to form interchain cross-links. For example, Gal₅₀Fuc₅₀ can form a complex with CTB in which unoccupied canonical and noncanonical ligands are randomly distributed, facing outward along the glycopolymer chain (Figure 5F). Unlike the homopolymer–CTB complexes, Gal₅₀Fuc₅₀–CTB complexes can thus utilize their unoccupied binding sites to bind additional glycopolymer chains. This means that the binding is no longer limited by the facial directionality of CTB. As a result, CTB and Gal₅₀Fuc₅₀ can stack to form a continuous network of alternating protein and glycopolymer (Figure 5F).

Transmission electron microscopy (TEM) showed that in buffer solution, CTB exists in the form of spherical particles with a diameter of 6 nm (Figure 6A), which is consistent with the CTB crystal structure and our DLS data. For better contrast, glycopolymers were stained using ruthenium tetroxide. Regardless of sugar composition, TEM images of the glycopolymers showed amorphous spheres which suggest the glycopolymer chains are entangled and collapsed due to the absence of solvent molecules. (Figure 6B, C). We observed mixed amorphous sphere and lamellar morphologies of glycopolymers when stained (Figure S4). The lamellar structures are more visible at lower magnification. We also obtained TEM images of glycopolymers displaying bicontinuous morphology without staining (Figure S5). The bicontinuous domains may be due to the two volume fractions that the glycopolymers are composed of: that is, 64.3% hydrophilic sugars and 35.7% hydrophobic norbornene. Also, the atactic glycopolymers may contribute to the uneven size and order of the domains. The volume fractions may have been altered during the oxidative staining process and this could affect the morphology observed. After carefully examining TEM images of both unstained and stained glycopolymers, we concluded that the lamellar structures developed from bicontinuous structures during the oxidative staining process. The amorphous spheres may result from dense aggregation of glycopolymers upon drying which then leads to high contrast staining. Therefore, we analyzed CTB-glycopolymer complexes without staining to better distinguish CTB aggregation from entangled glycopolymer chains. TEM images of the Gal₁₀₀–CTB_n complexes suggest that they locally segregated into small aggregates as each Gal₁₀₀ chain became saturated by CTB (Figure 7A, D). This binding mode would result in only a slight increase in particle size upon mixing of the glycopolymer with CTB, which is consistent with the DLS results (Figure 3B). Similarly, a Fuc₁₀₀ chain can be saturated with CTB at noncanonical binding sites to form a Fuc₁₀₀–CTB_n complex, leaving only the canonical, galactose-binding sites open (Figure 5D). The same segregation behavior was observed for Fuc₁₀₀ (Figure S6). In the absence of interchain cross-linking, the complexes that form consist of a single homoglycopolymer with multiple CTBs.

Figure 6.

TEM images of (A) CTB without staining, (B) Gal₁₀₀ with RuO₄ staining, and (C) Gal₅₀Fuc₅₀ with RuO₄ staining.

Figure 7.

TEM images without staining of (A, D) Gal₁₀₀–CTB complex and (B, C, E) Gal₅₀Fuc₅₀–CTB complex and (F) confocal laser scanning microscopy image of Gal₅₀Fuc₅₀–CTB complex.

In contrast, TEM images of Gal₅₀Fuc₅₀–CTB_n complexes show that they have hyperbranched structures (Figure 7B,C,E) similar to snowflake-like structures formed by polymer-capped gold nanoparticles.⁴³ These structures are as large as 100 μm and can be visualized by means of confocal laser scanning microscopy (Figure 7F). Likewise, a blend of Gal₁₀₀ and Fuc₁₀₀ homopolymers and CTB form a glycopolymer–protein network with a similar snowflake-like structure as expected based on the observed aggregates in DLS (Figure 3F).

Although both Gal₅₀Fuc₅₀ and the blend of Gal₁₀₀ and Fuc₁₀₀ generate large complexes with CTB, the complexes generated from the heteroglycopolymer are much larger. Because the formation of both Gal₁₀₀–CTB and Fuc₁₀₀–CTB complexes produce unoccupied binding faces accessible to their other sugar ligand, the complexes most likely consist of alternating layers of glycopolymer chains (Figure 5E). The random presentation of galactose and fucose within a single glycopolymer chain results in more interchain cross-linking and greater immobilization of the toxin by forming complexes that precipitate out of solution, which in turn, reduces toxin available to initiate intoxication.

We performed a turbidimetric assay to monitor the formation of Gal₅₀Fuc₅₀–CTB_n complexes (Figure 8). Three different CTB/Gal₅₀Fuc₅₀ ratios (100:1, 50:50 and 1:100) were tested to determine the rate of complex formation. We monitored absorbance at 580 nm as an indication of increased light scattering by particle formation. The absorbance increased immediately when Gal₅₀Fuc₅₀ was introduced to a buffer solution of CTB, and an emulsion began to form. The rate of absorbance changed depended on the CTB/Gal₅₀Fuc₅₀ ratio. Next, we titrated CTB with Gal₅₀Fuc₅₀. A decrease in absorbance was observed after each addition of Gal₅₀Fuc₅₀ once the titration reached the cloud point indicating precipitation of large aggregates (Figure S7). The turbidimetric assay suggests that complex formation is fast (≤17 min) and generates large insoluble protein-polymer complexes.

Figure 8.

Turbidimetric assay curves, showing the temporal dependence of the absorbance at 580 nm at three CTB/Gal₅₀Fuc₅₀ ratios. (0.1 w/v% CTB:0.01 w/v% Gal₅₀Fuc₅₀, 0.05 w/v% CTB:0.05 w/v% Gal₅₀Fuc₅₀, and 0.01 w/v% CTB:0.1 w/v% Gal₅₀Fuc₅₀)

With the goal of determining the amount of CTB captured in the CTB–Gal₅₀Fuc₅₀ complexes and the rates of CTB capture and release, we attempted to carry out some complexation experiments using CTB labeled with a commercially available fluorescent dye. However, we found that the labeled CTB did not form aggregates with Gal₅₀Fuc₅₀. The fluorophore may have been large enough to interfere with CTB–glycopolymer binding because of the relative sizes of the glycopolymers and the ligands. This observation raises doubts about the validity of using fluorescent-dye-labeled proteins in certain assays. Therefore, it is worth noting that the detection of fluorescence might not be the best way to evaluate CTB binding when such a high degree of interchain cross-linking is involved.

An acute oral toxicity test in mice showed that a single dose of Gal₅₀Fuc₅₀ administered by oral gavage had no adverse effects. Jejunum histology and blood chemistry performed 24 h after administration of Gal₅₀Fuc₅₀ showed no significant abnormalities, including no epithelial cell rupture or inflammation of the jejunal tissues. The cytotoxicity of Gal₅₀Fuc₅₀ was greater than 100 μM. This result, combined with the nanomolar IC₅₀, suggests that Gal₅₀Fuc₅₀ has potential utility as a cholera therapeutic and that the strategy of targeting the two different binding sites on a toxin, as described herein for CTX, may be useful for treating diseases in which carbohydrate recognition is involved in the intoxication process.

CONCLUSION

The large-scale preparation of GM1 and other bioactive oligosaccharides is challenging due to the complexity of their synthesis.^44–45 To overcome this limitation, investigators have developed GM1 mimics.^{17, 46–51} These precisely tailored inhibitors with exotic architectures have matching valency to the binding sites on CTB for better positioning and are known to have particularly high potency.^52–56 Nevertheless, the cumbersome synthesis of such scaffolds is often expensive, thus limiting their cost-effectiveness as therapeutics in developing countries.

Herein, we have described a strategy involving concurrent targeting of the two different binding sites on CTB with the goal of inhibiting intoxication induced by binding of CTB to two different native ligands. This strategy was implemented with glycomimetic polymers that promoted the formation of cross-linked CTB_n–glycopolymer networks that blocked CTX internalization process. The size of the networks could be increased if the appropriate ligands were present and could be tuned further by modifying the pendant ligands or by varying the glycopolymer composition of a mixture. A random Gal₅₀Fuc₅₀ copolymer showed the highest degree of cross-linking and consequently formed mega-aggregates in the presence of CTB. Such “protein capturing agents” can be expected to be effective against other lectins that, like CTB, have independent binding sites on opposite faces. The glycopolymers described herein could be expected to capture and immobilize CTB by precipitating it in the form of complexes that would be too large to be endocytosed by epithelial cells.^57–59 Glycopolymers with low toxicity could be orally ingested and would form aggregates with CTB in the human intestine, and the complexes eliminated by passage through the digestive tract. Simple multivalent glycopolymers that maintain the required structural characteristics without exotic architectures, and that do not require complex synthetic procedures can be expected to be relatively inexpensive. Such a cholera mitigation therapy would be accessible to developing countries in which cholera outbreaks are both recurrent and severe.

ABBREVIATIONS

CTA

cholera toxin A subunit

CTB

cholera toxin B subunit

CTX

cholera toxin

DLS

dynamic light scattering

DP

degree of polymerization

GPC

gel permeation chromatography

ROMP

ring-opening metathesis polymerization

SAXS

small-angle X-ray scattering

TEM

transmission electron microscop