The principle and basic setup of a prism-based TIRFM system (Figure 1A) has been described in detail elsewhere [18, 50]. For the measurements described here, fluorescence was excited by the evanescent wave generated by internally reflecting the 488 nm line of an argon ion laser (Coherent Innova 90-3). Unless otherwise specified, the power of the incident beam was 300 μW. The beam was passed through a fused silica prism optically coupled to the microscope slides with glycerol, so that the incidence angle at the interface of the microscope slide and inner volume was greater than the critical angle. The incident beam was always s-polarized, the evanescent wave depth was ≈85 nm, and the 1/e2-radii of the elliptically shaped Gaussian illumination area were approximately 65 µm × 20 µm. Evanescently excited fluorescence was collected through a 60×, 1.4 NA oil immersion objective on an inverted microscope (Zeiss Axiovert 35) coupled below the sandwich coverslip, passed through an appropriate dichroic mirror and barrier filter, and imaged by an EMCCD camera (Andor iXon DU-871E) driven by Andor iQ software. The pixel size of this camera was 16 µm which, when a 60× 1.4 NA objective is employed, means that the PSF is undersampled. However, because we are measuring the emitted intensity of single molecules or domains that are widely separated as opposed to creating detailed images of those objects, this was not considered to be a limitation. Unless otherwise specified, the EMCCD was cooled to −70°C, the gain was set at 260, and images were collected at 512 × 512 pixels per frame with a 300 milliseconds exposure time. Depending on the photostability of the fluorophores, 100–500 frames per video were recorded to observe single-step photobleaching and calculate the brightness of single fluorophores. For imaging DC-SIGN microdomains on cell surfaces, a single frame was collected.
Molecular counting data analysis
To determine the numbers of DC-SIGN molecules in microdomains, the brightnesses of microdomains and single fluorescent molecules were compared. Ideally, this approach requires a 1:1 labeling ratio of single fluorophores to single DC-SIGN molecules, as well as identical optical parameters for both single-molecule and microdomain imaging. The extent to which deviations from these ideal conditions might have skewed the reported results is discussed above.
Prism-based TIRFM employing an incident argon ion laser beam creates a surface-associated evanescent wave with an intensity profile in the sample plane that has an elliptically Gaussian shape  as shown in Figure 1B. The evanescent excitation intensity is higher at the middle than at positions away from the center; thus, fluorophores towards the middle of the illuminated area will be brighter than those in peripheral regions. To account for this effect, the local background luminescence arising from substrate impurities and close to single fluorescent molecules or microdomains was employed as an indicator of the local excitation light intensity. This procedure can in general be applied to any illumination scheme, although it is particularly important for prism-based TIRFM.
For single fluorescent molecules, first, a time sequence of 100–500 images was acquired. ImageJ software was then used to roughly localize the single molecules and generate plots of the time-dependent, total intensities within selected regions of interest surrounding the molecules (Figure 1C). These plots were examined to determine the time of the single-step bleach (for GFP) or the last single-step bleach (for mAb). A frame preceding the single or last bleach and having an integrated spot intensity approximately equal to the time-average immediately before this bleach was then selected for further analysis, as described below.
Single fluorescent molecules or single DC-SIGN microdomains were localized in individual frames by a particle finding algorithm in the ‘localizer’ plugin of Igor Pro software (kindly provided by Dr. Peter Dedecker at the University of Leuven, Belgium; the plugin can be downloaded at: www.igorexchange.com/project/Localizer). The criteria by which spots were identified are described elsewhere [16, 51]. The counts per pixel surrounding the kth localized centroid for single molecules, Zsm(i,j,k), or the mth localized centroid for single microdomains, Zdomain(I,j,m), where the integers i and j denote the pixel position, were then fit to two-dimensional Gaussian functions; i.e.
The free parameters were A(k,m) (the amplitude of the emitted fluorescence), i0(k,m) and j0(k,m) (a more accurate center), μ(k,m) (the spatial width), and B(k,m) (the local emitted background luminescence). In the following discussion, we use the parameter s to denote the length of one pixel with s equal to one; hence, s2 also equals one. Parameters i, j, i0(k,m), j0(k,m), and μ(k,m) have the units of s. Zsm(i,j,k), Zdomain(i,j,m), A(k,m), B(k,m), and C have the units of (detected counts)(s)−2(τ)−1, where τ was the exposure time (300 milliseconds). More explicitly, the A(k,m) have the units of (detected fluorescence counts)(s)−2(τ)−1 and the B(k,m) have the units of (detected background luminescence counts)(s)−2(τ)−1. C, representing the sum of the background luminescence which is independent of the excitation intensity and the dark count of a single pixel on the EMCCD camera, was measured far from the evanescently illuminated area to have an average of 165 (detected counts)(s)−2(τ)−1 and fixed at this number during curve-fitting. This analysis gave three lists (for the different single-molecule types) containing the best-fit values of A(k), B(k), i0(k), j0(k) and μ(k), where the index k denotes the kth single molecule of a given type. The analysis also gave six lists (for the different microdomain types) containing the best-fit values of A(m), B(m), i0(m), j0(m) and μ(m), where the index m denotes the mth microdomain of a given type.
For a given single fluorescent molecule or DC-SIGN microdomain, the local evanescent excitation intensity was accounted for in the following manner. The intensity-dependent background luminescence's can be expressed as
where I(k) and I(m) are the local excitation intensities for the kth single molecule or the mth microdomain, respectively, in units of (excitation photons)(s)−2(τ)−1, and β is a proportionality constant independent of k and m with the units of (detected background luminescence counts)(excitation photons)−1. Single molecule and microdomain powers (spatially integrated fluorescence emission intensities) were calculated for each single molecule or domain, from the best-fit values of A(k,m) and μ(k,m), as
in units of (detected fluorescence counts)/τ. These parameters are also given by
where Q is a proportionality constant independent of k and m, with the units of (s2)(detected fluorescence counts)(excitation photons)−1 and N(m) is the number of DC-SIGN molecules in the mth microdomain. Corrected powers, which account for the local excitation intensity, were calculated for each single molecule or microdomain as
Thus, referring to eqn (2) and (4), one finds that
The corrected powers do not depend on the local excitation intensity and have the units of (s2) (detected fluorescence counts)(detected background luminescence counts)−1.
In the idealized case in which other sources of noise are not present, for GFP, ηsm,c(k) should not depend on the particular kth single molecule (eqn (6)). Thus, an average, ηGFP,c, was calculated as
where ηsm,c(k) denotes the measured corrected power for the kth single GFP molecule and TGFP denotes the number of single GFP molecule images analyzed. Then, for each microdomain in which the fluorescence was reported via GFP, the number of DC-SIGN molecules in this microdomain was computed as (see eqn (6))
For the two different mAbs, the situation is a bit more complex. Computing experimentally obtained, average corrected powers for the two types of mAbs gives
where ηsm,c(k) denotes the measured corrected power for the kth single mAb of a given type and TmAb denotes the number of single mAb molecule images analyzed. Because the experimental values of ηsm,c(k) were measured for the last single-step bleach, ηmAB,c = [ηmAB,c]expt is the measured, average corrected power for a single AlexaFluor488 conjugated to a mAb. In general, for an average labeling ratio of γ ≈ 1 of AlexaFluor488 probes per mAb, and assuming a Poisson distribution for the number of fluorophores per mAb, the average corrected power is, theoretically,
Thus, because γ ≈ 1 and because only frames immediately prior to the last single-step bleach for the mAbs were used, the fact that some mAb have 0, 1, 2 or more conjugated fluorophores can be accounted for. Worth noting is that a similar procedure can be used when γ ≠ 1, by multiplying [ηmAb,c]expt by γ. For each microdomain in which the fluorescence was reported via a mAb, the number of DC-SIGN molecules in this microdomain was computed as (see eqn (6))
The spot widths (for single molecules) or microdomain widths are denoted by δsm(k) and δdomain(m), respectively, and were calculated in nm as δsm(k) = μ(k)σ or δdomain(m) = μ(m)σ where σ = (16 µm)/(60) = 270 nm was the pixel size (16 µm is the pixel dimension of the camera and the objective was 60×). Apparent microdomain areas were determined as Adomain(m) = πδdomain2(m). As noted in Figure 1D, large ill-defined microdomains were excluded from analysis as it was impossible to ascertain whether such domains were a collection of smaller microdomains.
Confocal imaging and colocalization analysis
For DENV and DC-SIGN microdomain colocalization analysis, NIH3T3 cells expressing DC-SIGN plated on 35 mm MatTek dishes were first incubated with endocytosis inhibitors (10 mm NaN3, 2 mm NaF, and 5 mm 2-deoxy-d-glucose) for 2 min, then incubated with DENVs at 15.7 MOI for 10 min, thoroughly washed several times with Dulbecco's phosphate-buffered saline (DPBS) and fixed with 2% paraformaldehyde (PFA) for 20 min. After fixation, the cell dishes were separated into two groups: non-permeabilized and permeabilized. Non-permeabilized cells were used to image only cell-surface DENV and DC-SIGN microdomains for surface colocalization analysis. For this group, the cells were washed three times with DPBS, and submerged in 1% normal mouse serum (NMS) in DPBS for 30 min for blocking. Permeabilized cells were used to image both surface and internalized DENVs and DC-SIGN. For this group, the cells were washed three times with DPBS, submerged in Perm Buffer (2% BSA, 0.1% saponin, 0.02% NaN3 in sterile DPBS) and washed twice with Perm Buffer. After permeabilization the cells were incubated with blocking buffer (1% NMS in Perm Buffer) for 30 min. After blocking, antibodies for staining DENVs or DC-SIGN were diluted either in 1% NMS in DPBS for non-permeabilized cells, or in 1% NMS in Perm Buffer for permeabilized cells. The cells were stained with anti-DENV 2H2-AlexaFluor488 at saturation concentration for 1 h at 37°C, washed thoroughly several times with DPBS, incubated with primary anti-DC-SIGN H-200 IgG at 6 µg/mL for 20 min, washed thoroughly several times with DPBS, treated with anti-rabbit (Fab')2 AlexaFluor647 for 20 min, and finally washed thoroughly several times with DPBS.
Confocal imaging of DENV and DC-SIGN on NIH3T3 cells was carried out on a Fluoview FV1200 laser scanning microscope (Olympus) with an oil-immersion 60x NA 1.35 objective. An excitation wavelength of 488 nm at 2% power was used to image DENVs stained by 2H2-AlexaFluor488 mAb and an excitation wavelength of 635 nm at 2% power was used to image DC-SIGN stained with secondary anti-rabbit AlexaFluor 647 (Fab')2. The emission detection wavelength ranges were 500–550 nm and 650–710 nm, respectively. Image frames contained 1024 × 1024 pixels, with a pixel size of 69 nm.
Colocalization analysis was carried out with a plugin named JACop on ImageJ . After preprocessing and setting the thresholds in both green and red channels, Manders' coefficients were calculated and read out as indications for colocalization percentages. Manders' overlap coefficients are based on Pearson's correlation coefficients , and range from 0 (no colocalization) to 1 (100% colocalization). Since, in our images the numbers of DENVs were much less than the numbers of DC-SIGN microdomains on cell surfaces, we used M1, the ratio of green DENV signals that overlap with red DC-SIGN signals to the total green signals. Calculation of the number of DENV particles was done by using the Mosaic Particle Tracker 2D/3D plugin in Image J .
Infectivity assay by FACS
Two different concentrations of DENVs were used (15.7 and 1.57 MOI). NIH3T3 cells stably expressing DC-SIGN were plated on 35 mm cell culture dishes, and incubated with the two concentrations of DENVs for 24 h, 48 h, and 72 h, respectively. At the time of observation, cells were washed several times with DPBS, trypsinized, centrifuged at 450 × g for 5 min, resuspended in DPBS, centrifuged at 450 × g for 5 min, and resuspended in 2% PFA for 20 min. After fixation, the cells were washed with DPBS, and permeabilized with Perm buffer. Staining of DENVs using 2H2-AlexaFluor488 and staining of DC-SIGN using H-200/anti-rabbit AlexaFluor647 was performed as described above in Confocal imaging and colocalization analysis section. The cells were finally transferred into a 96-well plate and infectivity assays were carried out on a Guava easyCyte 8HT flow cytometer (Guava). The data analysis was performed using Guava Express software (Guava).