Anatomy of a viral entry platform differentially functionalized by integrins α3 and α6

During cell invasion, human papillomaviruses use large CD151 patches on the cell surface. Here, we studied whether these patches are defined architectures with features for virus binding and/or internalization. Super-resolution microscopy reveals that the patches are assemblies of closely associated nanoclusters of CD151, integrin α3 and integrin α6. Integrin α6 is required for virus attachment and integrin α3 for endocytosis. We propose that CD151 organizes viral entry platforms with different types of integrin clusters for different functionalities. Since numerous viruses use tetraspanin patches, we speculate that this building principle is a blueprint for cell-surface architectures utilized by viral particles.

. L1-7 vesicle staining intensity. Fluorescence intensity of L1-7 positive vesicles as counted in Figure 1C. Individual vesicle intensity values were averaged per cell. Cells without vesicles were excluded from the analysis. Values are given as means ± SD (n = 58 -60 cells collected from three biological replicates). Unpaired Student's t-test, comparing control to knockdown conditions (***, p < 0.001).

Figure S3. Direct plasmid transfection reveals secondary integrin knockdown effects on promoter activity.
The encapsidated plasmid used for Figure 1D was directly transfected into HaCaT cells. This circumvents the viral entry pathway and allows for analyzing secondary integrin knockdown effects on the viral promoter activity that are not related to the viral infection pathway. In brief, we transfected the siRNAs into HaCaT cells, and 1 day later the plasmid. After another day, we measured luciferase and lactate dehydrogenase activity. For normalization to cell number luciferase activity was normalized to dehydrogenase activity. Values are given as means ± SD (n = 12 technical replicates collected from 3 biological replicates). The knockdown conditions show that in the absence of integrin α3, luciferase activity is strongly diminished, whereas in the absence of integrin α6 its activity is strongly increased. Normalization of infection rates from Figure 1D to these secondary effects yield infection rates of 48% and 16% after integrin α3 and integrin α6 knockdown, respectively. The unpaired Student's t-test compares the control to the knockdown conditions (***, p < 0.001). Figure S4. Illustration of the gain in resolution by STED microscopy. Membrane sheets from cells overexpressing CD151-GFP were fixed and stained for integrin α3 and integrin α6. CD151-GFP fluorescence was enhanced by nanobodies. From the confocal channels, overviews and magnified views are shown, and from the magnified views also the STED micrographs. Magnified views in the left column show raw data, the right column illustrates smoothed images that were used for maxima detection. Circles indicate locations where maxima were detected. Images are shown using a linear greyscale lookup table employing an arbitrary scaling. Please note that STED microscopy allows for resolving larger spots into smaller entities (arrows point to a patch that is better resolved in the integrin channels).  Figure 3B, we show for the individual channels the distribution of maxima sizes as determined by Gaussian fitting. A Gaussian function was fitted to the intensity profile obtained from a 31-pixel long linescan, from which the central pixel was positioned to the maxima detected by the algorithm. Only maxima were included if the R 2 value of the fit was > 0.8, and if the determined center position of the Gaussian peak was within the central third of the linescan. As a result, the analysis includes 62.2% (22623 maxima), 39.1% (20209 maxima) and 42.7% (8300 maxima) of the initially detected integrin α3, integrin α6 and CD151 maxima, respectively. The histograms show lognormal distributions, in particular for the smaller integrin α3 maxima. Moreover, no sizes smaller than 50 nm are detected. This points towards the possibility that our STED microscope cannot resolve size differences within the fraction of smaller maxima and that the resolution limit does not reach the range of 50 nm. For verification, we determined the point spread function (PSF) of our STED system ( Figure S6). For determination of the point spread function we used PsVs for calibration. Papillomaviruses have a physical size of ≈ 60 nm 1 . After staining with primary and secondary antibodies, we estimate that the 'shell of fluorophores' increases the diameter by about 20 nm (as has been observed before for microtubuli 2 ) to roughly 80 nm. For sample preparation, one µl of PsVs, corresponding to ~ 2*10 7 viral genome equivalents (vge), was diluted in 1 ml of DMEM and the solution was pipetted onto a poly-Llysine coated glass-coverslip. PsVs were allowed to adsorb overnight at 4°C and stained as in Figure  6, just omitting the permeabilisation step and CD151 antibodies, and using instead of the K75 the 16L1-312F antibody. Secondary antibodies were labelled by the dyes Atto488 (used in Figure 3 for CD151-GFP labelling; linear green lookup table), AF594 (used in Figure 3 for integrin α3 labelling; linear red lookup table) and STAR RED (used in Figure 3 for integrin α6 labelling; linear magenta lookup table). PsVs were imaged employing the same STED laser powers and excitation laser powers as for the maxima analysis in Figure 3. All Images are displayed at arbitrary intensity scalings. Maxima were analysed as in Figure 3. Values are given as means ± SD (n = 10 images per dye, each obtained from one coverslip; in total, the analysis includes 582 -1461 maxima per dye). PSFs were calculated using the formula PSF = sqrt(measured size 2physical size 2 ), with a value of 80 nm taken as the real physical size. In the STED mode, the sizes of the PsVs are reduced from 236 -315 nm to 103 -127 nm. The calculated PSFs are 91 nm, 64 nm, and 97 nm for the CD151-GFP, integrin α3, and integrin α6 channels, respectively. Deconvolution with these PSFs yields physical sizes of the maxima of 189 nm (CD151), 84 nm (integrin α3), and 123 nm (integrin α6).  Figure  6A-C. After integrin α6 knockdown, the level of endogenous CD151 decreases by 33 % (Figure 6B). In the different conditions (overexpressed, low and normal endogenous level), doubling of the expression level roughly increases the maxima size by 10 -15 % and the density by 30 -50 %. Please note that absolute maxima size and density values cannot be compared throughout experiments as different staining and imaging conditions are applied. (E) CD151-GFP was transfected into HaCaT cells. After one day, membrane sheets were generated, fixed, permeabilized, immunostained for CD151 and imaged by epi-fluorescence microscopy. In case that overexpressed CD151-GFP and endogenous CD151 would segregate into different domains, the endogenous CD151 domains would not overlap with GFP. Images are displayed with linear green (CD151-GFP) and red (anti-CD151) lookup tables at arbitrary intensity scalings. The visual impression and the high Pearson correlation coefficient between the two channels (0.75 ± 0.09; value is given as mean ± SD; n = 36 membrane sheets collected from three biological replicates) suggest that both CD151-GFP and endogenous CD151 locate to the same domains. In fact, the Pearson correlation coefficient is higher than the value from an earlier experiment in which a double tagged protein was used (0.63; 3 ), which can be taken as a reference for perfect co-localization. (F) Plotting of CD151 immunostaining intensity versus GFP fluorescence intensity. In addition to the 36 membrane sheets analysed in E, supplementary membrane sheets, also without CD151-GFP, were included. The average immunostaining intensity of these membrane sheets marks the endogenous CD151. The graph suggests that upon overexpression the total CD151 expression level (endogenous + overexpressed) increases on average by 2.5 fold when compared to the endogenous level. In this experiment, compared to Figure 3, it was necessary to image in the GFP-channel brighter membrane sheets, in order to have comparable amounts of endogenous and overexpressed CD151 in the cell membrane.

Figure S8. Rescue of infection by CD151 expression after CD151 knockdown.
CD151 was knocked down in HeLa cells by siRNA transfection targeting the non-coding 3'UTR region of the endogenous CD151 mRNA. After one day, cells were transfected with a control plasmid or plasmids for expression of non-tagged CD151 (CD151-WT) or GFP-tagged CD151 (CD151-GFP). One day later, cells were incubated with 100 vge per cell. Infection rate was measured 24 hours after PsVs addition in the luciferase based infection assay, normalizing luciferase activity to dehydrogenase activity. Values are given as means ± SD (n = 3 -4 biological replicates). Statistical analysis was performed employing the unpaired Student's t-test (*, p < 0.05; **, p < 0,01).  Figure 3A). The white box marks an area from which magnified views from the individual channels are shown. Images are displayed with the same lookup tables and scalings as in Figure 3. (B) Mean staining intensities of CD151-GFP, integrin α3 and integrin α6 quantified from membrane sheets generated from cells previously incubated for 5 h without (control) or with PsVs. Values are given as means ± SD (n = 60 membrane sheets collected from three biological replicates). Unpaired Student's t-tests compare control to PsV (**, p < 0.01).

Figure S10. PsV binding pattern imaged on intact cells or membrane sheets from cells overexpressing CD151-GFP.
(A) Basal membrane of a HaCaT cell fixed with methanol (from the experiment shown in Figure 4D -F). (B) Membrane sheet generated from a HaCaT cell overexpressing CD151-GFP (from Figure S11). Example images illustrate that independent from the fixation protocol and sample preparation in many cases integrin α3 (linear red lookup table) enriches at the cell periphery, occasionally in elongated structures (B). In particular viral particles (linear cyan lookup table), but also CD151 (linear green lookup table), enrich at the cell periphery. Images are displayed at arbitrary intensity scalings.   (A) Membrane sheets from control and CD151-GFP overexpressing HaCaT cells, stained with antibodies for GFP (green) and integrin α6 (red). Images are displayed at arbitrary intensity scalings (linear lookup tables). (B) From ROIs, the maxima per µm 2 of the two maxima types was determined, revealing that overexpression of CD151-GFP increases the number of integrin α6 maxima. Please note that the detected CD151-GFP maxima under non-overexpression conditions result from background signal. (C) Plotting from individual membrane sheets the density of integrin α6 against the CD151-GFP maxima suggests a trend towards more integrin α6 maxima with increasing CD151-GFP expression level. A regression line was fitted to the data (R 2 = 0.3296). (D) Average integrin α6 staining and (E) inter-maxima distances in control and overexpression condition. The distance between integrin maxima is reduced upon CD151-GFP overexpression due to the enhanced integrin maxima density. Values are shown as means ± SD (n = 59 membrane sheets collected from three biological replicates). The unpaired Student's t-test compares control to overexpression (***, p < 0.001; *, p < 0.05).     The blot was cut in two parts for staining of the upper part for integrin α6 and the lower part for actin ( Figure S19; both antibodies raised in rabbits). Figure S1A. The blot was cut in two parts for staining of the upper part for integrin α6 ( Figure S18) and the lower part for actin (both antibodies raised in rabbits).

Figure S19. Full Western blot for actin from
. Figure S20. Full Western blot for integrin 3 from Figure S1B.