Norovirus GII.4 Virus-like Particles Recognize Galactosylceramides in Domains of Planar Supported Lipid Bilayers

A sticky situation: Domain-dependent recognition of the glycosphingolipid galactosylceramide by norovirus-like particles (see picture; red/yellow) is shown using supported lipid bilayers (purple) as model membranes. Optimal ligand presentation is found to promote strong binding to GalCer. This presentation can be found at the edges of the glycosphingolipid-enriched domains (green) and binding is repressed in the absence of these domains.

The original TLC chromatogram binding assay using radiolabelled Sendai virus was published 1984 [7] and was recently used for studies of norovirus. [2b] In short, GSLs (1-2 µg of pure GSLs and 10-20 µg for mixtures) were applied to aluminum backed silica gel 60 HPTLC plates (Merck, Darmstadt, Germany) and chromatographed with chloroform:methanol:water (60:35:8, by volume). Each GSL sample was applied at least twice on each TLC-plate. The plate was cut into sections, which were either chemically stained by anisaldehyde:sulfuric acid:acetic acid (1:2:97, by volume) or plasticized by dipping them in polyisobutylmethacrylate (P28, 0.3 % in hexan:diethylether (1:1, by volume)) and dried thoroughly. The plasticized plates were blocked by immersion in 3 % BSA, 0.05 % Tween-20 in PBS at room temperature (RT) for 1 h. Thereafter, VLPs (1 µg/mL in CBA dilution buffer (0.5 % BSA, 0.05 % Tween in PBS) were added to the surface and incubated for 2 h at RT. The plates were washed in CBA washing buffer (0.05 % Tween-20 in PBS), followed by addition of rabbit antiserum against norovirus (diluted 1: 3000 in CBA dilution buffer) to the plates, incubation for 1.5 h at RT and washed. Secondary antibodies (alkaline phosphatase conjugated goat anti-rabbit IgG, Sigma (St Louis, MO, USA)) were added (1:1000 in CBA dilution buffer). After washing, the staining was developed using Sigma Fast BCIP/NBT.

Fluorescence microscopy
Fluorescence microscopy images were acquired using a Nikon Eclipse Ti-E inverted microscope and a 60X magnification (NA = 1.49) oil immersion objective (Nikon Corporation, Tokyo, Japan). The microscope was equipped with a mercury lamp (Intensilight C-HGFIE; Nikon Corporation), a FITC and TRITC filter cube (Nikon Corporation) and an Andor iXon+ EMCCD camera (Andor Technology, Belfast, Northern Ireland). Fluorescence recovery after photobleaching experiments were performed by bleaching circular areas with a diode pumped solid state laser at 475 nm (BWB-475-20E; B&W Tek Inc., Newark DE, USA).
Experiments were performed in glass-bottom microwells (96-well plate, MatTek Corporation, Ashland MA, USA) cleaned for at least 2 hours in 2% Hellmanex cleaning solution. The supported lipid bilayer was formed at 37 ⁰C by addition of a pre warmed vesicle suspension (0.1 mg/ml, TRIS buffer, 30 min incubation) followed by extensive rinsing with warm buffer, taking care not to dry the surface at any time. Temperature treatments were performed in a laboratory stove and the samples were always imaged at 22 ⁰C. The diffusion coefficient of fluorescent lipids was determined by fluorescence recovery after photobleaching experiments (FRAP). FRAP measures the lateral mobility of fluorescent lipids by bleaching the fluorescent tags in a well-defined circular area of the bilayer and monitoring lateral diffusion and healing of the bleached spot due to mixing of bleached and unbleached lipids.
The diffusion coefficient was determined using the Hankel transform method previously developed in our laboratory. [8] The diffusion coefficient and immobile fraction values are either the medians, when n=3, or the mean±standard deviation, when n=4. (n= number of individual experiments) VLP binding experiments were performed by incubating the substrate 30 min with VLPs (66 µg/ml), followed by rinsing, addition of the rhodamine vesicles containing H type 1 (total lipid concentration: 30 µg/ml) and incubation for another 30 min. The sample was imaged after extensive rinsing.

AFM
Atomic force microscopy (AFM) images were acquired in intermittent contact mode with a Nanowizard I BioAFM (JPK Instruments, Berlin, Germany) using MikroMasch NSC 18/Cr-Au cantilevers (MikroMasch, Tallin, Estonia). Experiments were performed in a temperaturestabilized Biocell scanning chamber (JPK Instruments, Berlin, Germany) on glass coverslips. The image in Figure 2c was acquired on a Si Wafer (University wafers, south Boston, USA) mounted in a homemade scanning chamber. The substrates were cleaned for at least 2 hours in a 2 wt % sodium dodecyl sulfate (SDS) solution followed by a 30 min UV-ozone treatment.
Bilayers were formed at 37 ⁰C with a procedure similar to the one described for fluorescence microscopy.
Height mode images were further processed using the SPM image processing software (JPK Instruments,Germany), the Scanning Probe Image Processor, SPIP (Image Metrology A/S, Denmark) and ImageJ (Image processing and analysis in Java, National Institute of Health; USA). The average domain coverage was estimated from images from two independent experiments. The total area imaged was 10 4 µm 2 .

QCM-D experiments
Quartz crystal microbalance with dissipation monitoring (QCM-D) was carried out to monitor bilayer formation and VLP attachment in real time. This technique relies on monitoring the change of the fundamental resonance frequency and its overtones in thickness-shear modes upon mass adsorption on the surface of an oscillating quartz crystal. [9] A decrease in frequency corresponds to an increase in mass and thickness on the surface where the mass estimated from QCM-D includes the surrounding liquid associated with the oscillation. Additionally, QCM-D measures the damping of the oscillation which provides structural information as it can be related to changes in viscoelastic properties of the adlayer: the softer the film, the higher the ∆D values. For rigid films, this mass is directly proportional to the frequency shift (∆f). [10] For viscoelastic systems, the relationship is no longer linear and the adsorbed mass is obtained by applying the Voigt model to ∆f and ∆D. [11] The QCM-D experiments were conducted with a Q-sense E4 instrument (Q-sense AB, Västra Frölunda, Sweden) equipped with 4 temperature-stabilized chambers. The SiO 2 -coated QCM-D crystals were purchased from Q-Sense AB and cleaned overnight in 2 % (w/w) SDS and 45 min in a UV-ozone cleaner. All experiments were performed in TRIS buffer. After baseline stabilization the vesicle suspension was introduced in the measurement chamber at a flow rate of 15 μl/min. After completion of the bilayer formation process, the chamber was rinsed with buffer. GalCer bilayers were always formed at 37 ⁰C, H type 1 bilayers at 22 ⁰C. When applicable, the following temperature treatment was performed: 1 hour at 50 ⁰C followed by cooling to 22 ⁰C at a rate of 0.5 ⁰C/min. The VLPs (8 μg/ml) were introduced in the measurement chamber under constant flow 50 (μl/min) and rinsed after 10 or 15 min. For GalCer bilayers, VLP injection was preceded by a 10 min blocking step with BSA (0.1 mg/ml). In the manuscript, we report the normalized frequency shifts (∆f 5 /5) and energy dissipation (∆D 5 ) of the 5 th overtone. These values are either the medians, when n=3, or the mean±standard deviation, when n=6 (n= number of individual experiments).
Fluorescence microscopy FRAP experiment were further performed in order to complemented our results with quantitative data on lipid mobility. In bilayers containing 10% GalCer, i.e. under conditions identical to those used for the AFM and QCM-D analyses, the diffusion coefficient of NBD-PC was 1.5 times lower than for NBD-PC in POPC (1.38±0.3 μm 2 s -1 vs 1.99 μm 2 s -1 ) with an around two times higher immobile fraction (6.6±3.3% versus 3.5%). For comparison, gel-phase-free bilayers containing 5% H type1, had D= 1.82±0.3 μm 2 s -1 and IF=1.16±1.46 μm 2 s -1 . (See fig. S1.) Fig. S1: FRAP experiment on bilayers containing 1% NBD-PC: Hankel transform of the fluorescence intensity of a bleached spot for a POPC bilayer (black), a bilayer containing 10% GalCer (red) and a bilayer containing 5% H type 1 (blue). All experiments were performed at room temperature ( ~21 o C).   ∆D vs. ∆F plots Fig. S6: A comparison of ∆D vs ∆f for both GSL systems provides further insight on the binding behavior. For H type 1, VLP binding was essentially linear, while binding to GalCer deviated from the linear behavior for ∆f < -2 Hz (Figure 3b). The successively lower ratio between ∆D to ∆f is most likely attributed to a local crowding of VLPs since the amount of water mechanically coupled to the oscillation decreases when the surface is increasingly saturated with virus particles. [12] It cannot be excluded, however, that VLP deformation or penetration into the bilayer contributes to the observed deviation.