Beyond attachment: roles of DC-SIGN in dengue virus infection

Abstract

DC-SIGN, a C-type lectin expressed on the plasma membrane by human immature dendritic cells, is a receptor for numerous viruses including Ebola, SARS, and dengue. A controversial question has been whether DC-SIGN functions as a complete receptor for both binding and internalization of dengue virus (DENV) or whether it is solely a cell surface attachment factor, requiring either hand-off to another receptor or a co-receptor for internalization. To examine this question, we used four cell types: human immature dendritic cells and NIH3T3 cells expressing either wild type DC-SIGN or two internalization-deficient DC-SIGN mutants, in which either the three cytoplasmic internalization motifs are silenced by alanine substitutions or the cytoplasmic region is truncated. Using confocal and super-resolution imaging and high content single particle tracking, we investigated DENV binding, DC-SIGN surface transport, endocytosis, as well as cell infectivity. DC-SIGN was found colocalized with DENV inside cells suggesting hand-off at the plasma membrane to another receptor did not occur. Moreover, all three DC-SIGN molecules on NIH3T3 cells supported cell infection. These results imply the involvement of a co-receptor because cells expressing the internalization-deficient mutants could still be infected.

Graphical Abstract

Whether DC-SIGN functions as merely an attachment factor for dengue virus (DENV) or whether DC-SIGN plays further roles beyond attachment has been controversial. We use mammalian cell culture models, as well as primary dendritic cells, and high resolution, quantitative fluorescence microscopy to track the movements of DC-SIGN and DENV during viral entry. Our results support a model in which DC-SIGN captures DENV and participates, along with a co-receptor, in DENV internalization via clathrin-coated structures and subsequent trafficking to early endosomes.

Introduction

Dendritic cells (DCs) are professional and potent antigen-presenting cells in the human immune system. They sample pathogens from peripheral tissue, migrate to lymph nodes, and present antigens to activate both memory and naïve T cells to initiate immune responses ¹. However, DCs that are patrolling the peripheral tissues are also often the first targets of infection by viruses, such as Ebola, HIV and dengue ^2–4. The infected DCs transport viruses to lymph nodes and facilitate infection of other cells and systemic spread of the virus. It is estimated that about 400 million people are infected by dengue virus (DENV) each year ⁵, and some of these cases of dengue infection lead to fatal dengue hemorrhagic fever and dengue shock syndrome. More knowledge about the mechanism in which dengue infects DCs will provide new potential strategies for preventing infection at its earliest stages and aid the numerous ongoing efforts for vaccine development ⁶.

DCs display on their plasma membranes so-called “pattern recognition receptors” (PRRs) ⁷. These receptors bind to carbohydrates on pathogens. Dendritic cell-specific intercellular adhesion molecule-3-grabbing non-integrin (DC-SIGN) is a key PRR. This molecule, which is a single-pass transmembrane type II protein and contains a distal, extracellular lectin domain, is highly expressed on the surface of immature DCs and mediates the uptake of a variety of viral, bacterial, and yeast pathogens, by binding to their surface carbohydrates, for presentation to other immune cells ⁷. A number of pathogens use DC-SIGN binding to disrupt DC function and circumvent normal immune surveillance ^8,9. For example, SARS ⁸, Ebola ¹⁰, dengue ⁴ and other viruses ^11,12 use DC-SIGN as an initial cell attachment factor and/or entry receptor.

In the case of dengue infection, it has been controversial whether DC-SIGN is merely an attachment factor for DENV or plays further roles in viral entry ^4,13–17. The “attachment factor” mechanism was proposed after cells expressing internalization motif-deficient DC-SIGN were shown to still be infected by DENV. In this hypothesis, DC-SIGN binds DENV, but then the virus is handed over to another, as yet unidentified, co-receptor for DENV entry into the cells ¹⁵. However, it is also possible that the DENV/DC-SIGN complex does not dissociate during entry but, rather, acts in concert with a co-receptor.

To explore in greater detail the issue of DENV attachment and entry in DC-SIGN expressing fibroblast cells (MX-DC-SIGN) as well as primary human DCs in some cases, we employed quantitative imaging techniques including confocal imaging, single particle tracking with high-content, advanced analysis, and super-resolution imaging, combined with virus binding, internalization and infectivity assays. Previously, elementary single particle tracking was used to track early steps of DENV entry into cells that do not express DC-SIGN ^18,19. In our study, we found that DENV bound to cell surface DC-SIGN was laterally mobile in the plasma membrane, and that the virus/receptor complexes co-migrated to clathrin-coated structures (CCS) in both MX-DC-SIGN cells and primary human DCs. Contrary to the attachment factor only hypothesis, both DENV and DC-SIGN, in presumed complex, were found in early endosomes, in an internalization process involving microtubules (MTs). Alanine substitution of the three cytoplasmic internalization motifs or truncation of the DC-SIGN cytoplasmic tail reduced but did not completely abrogate both capture and endocytosis of virus, leading to reduced infection of cells. Overall, our results support a revised model in which DC-SIGN, in concert with a co-receptor as yet to be identified, functions as both an attachment factor and entry receptor for DENV.

Results

Cell surface DC-SIGN efficiently captures DENV

Confocal fluorescence microscopy was used to measure capture of dengue virions by NIH3T3 cells expressing human DC-SIGN, denoted as MX-DC-SIGN cells ^20,21. We have previously demonstrated that DC-SIGN expression renders these cells highly susceptible to productive DENV infection ²². To quantify DENV (serotype 1) attached to these cells during the steady-state of capture and endocytosis while constantly exposed to virus, cells were incubated with DENV for 30 minutes at 37 °C before fixation. After staining with dengue-specific and DC-SIGN-specific antibodies, 80 ± 12 % (N = 24 cells, ±SD) of cell-associated virions colocalized with DC-SIGN clusters (Fig. 1, upper row). A similar amount of DENV was captured by DC-SIGN expressed on the surface of human primary DCs (Fig. 1, bottom row) with a similar amount of colocalization (72 ± 8 %, N = 12 cells, ± SD). This degree of colocalization between DENV and DC-SIGN is well above the level expected from approximately random distributions of proteins with the same relative densities as DENV and DC-SIGN; for example, only 10 ± 3 % (N = 8 cells, ± SD) of DC-SIGN clusters colocalized with EGFP-clathrin in NIH3T3 cells expressing DC-SIGN (in the absence of DENV) (Fig. S1A). Control, wild type NIH3T3 cells (i.e., those not expressing DC-SIGN) did not effectively capture DENV (Fig. S1B). In over a dozen cell samples, only a few non-specifically binding virions were observed. These results complement our previous report that dengue serotype 2 virions colocalize with DC-SIGN at early time points after addition to cells ²² and are consistent with an earlier report that cell surface DC-SIGN efficiently captures different DENV serotypes ⁴.

Figure 1.

Colocalization of DENV with cell surface DC-SIGN. Fixed-cell confocal immunofluorescence of DENV bound to cell-surface DC-SIGN clusters on MX-DC-SIGN cells and DCs. For MX DC-SIGN cells, the cells were incubated with DENV (MOI=10) for 30 minutes at 37 °C before fixation. For DCs, the cells were incubated with DENV (MOI=50) for 5 minutes at 37 °C before fixation. Upper row, DENV captured by MX-DC-SIGN cells. 80 ± 12 % (N = 24 cells, ± SD) of cell-associated virions colocalized with DC-SIGN clusters. Bottom row, DENV captured by DC-SIGN on human primary immature DCs. 72 ± 8 % (N = 12 cells, ± SD) of cell-associated virions colocalized with DC-SIGN clusters. DENV was labeled with 2H2-AlexaFluor488 mAb (left panels), and DC-SIGN was labeled with DCN46-AlexaFluor647 mAb (middle panels). Insets (right panels) are shown in the same colors as the whole cell image panels. Scale bar, 10 µm; inset scale bar, 1 µm.

Motions of DC-SIGN and DENV/DC-SIGN complexes on the cell surface

Measurements were carried out to determine if DC-SIGN and DENV moved together as a complex to sites of viral entry on the cell surface. Movements of AlexaFluor568-conjugated DENV and DCN46-AlexaFluor647-tagged DC-SIGN on the cell surface were tracked by live cell imaging using confocal microscopy. Both DENV and DC-SIGN movements were recorded simultaneously at ≈1 Hz for several minutes. Demonstrating true, prolonged complex existence, we observed not only frequent colocalization but also co-transport of DENV and DC-SIGN in over 20 cells with hundreds of DENV/DC-SIGN complex movements on cell surfaces. Statistically, ≈ 48 % of DENV/DC-SIGN complexes moved together on cell surfaces and disappeared together within 60 seconds, ≈ 33 % of complexes disappeared between 61–120 seconds, and ≈ 19 % of complexes disappeared after 120 seconds. S2 Video shows a representative example of co-transport of DC-SIGN clusters and DENV for about 20 seconds before the signal disappears, presumably due to endocytosis. Thus DC-SIGN and DENV are simultaneously and rapidly internalized after binding at the cell surface.

To further investigate the motions of DC-SIGN clusters on cell surfaces, ventral membranes of live cells were observed by using total internal reflection fluorescence (TIRF) microscopy. DC-SIGN was labeled with anti-DC-SIGN Fabs conjugated with AlexaFluor568, made from DCN46 mAb. Using u-track, a multi-particle tracking algorithm ²³, we focused on characterization of the dynamics of the entire population of DC-SIGN clusters to avoid potential for bias introduced by manual selection of fewer events/clusters. In addition to tracking the moving clusters, u-track also uses a trajectory classification scheme employed by Ewers et al. ²⁴ based on the earlier work of Ferrari et al. ²⁵. By using the moment scaling spectrum, trajectories are classified as Brownian diffusion, anomalous subdiffusion or anomalous superdiffusion. The latter two categories represent diffusive movement slower than or greater than Brownian diffusion, respectively. For example, subdiffusive behavior could reflect diffusion in a system of transient traps while superdiffusive behavior could reflect diffusion augmented by membrane flow or flow within a stationary membrane. In Table I, the fractions of each type of mobility as well as the lateral diffusion coefficient (D) for Brownian diffusers is given. The mean value of D for DC-SIGN clusters was about of 0.4 µm²/s, within the range typical of membrane proteins measured at the light microscope length scale. The mean value is somewhat larger than that measured by Manzo et al. on CHO cells but their measurements used quantum dots as the label ²⁶. Overall, this analysis shows that laterally mobile DC-SIGN clusters could transport DENV to coated pits by both Brownian and anomalous diffusion modes.

Table I.

DC-SIGN cluster diffusion measured by TIRF imaging and analyzed by U-track

Experiment number	Number of samples movies analyzed; one movie taken from one single cell	Fraction of Trajectories classified as anomalous subdiffusion (%)	Fraction of Trajectories classified as anomalous superdiffusion (%)	Fraction of Trajectories classified as Brownian diffusion (%)	D (for Brownian diffusion) (µm2/s)	SD	SEM
1	3	55	5	39	0.36	0.07
2	6	48	4	48	0.30	0.04
3	1	30	6	63	0.40	0.35
4	6	38	5	57	0.54	0.09
Total or Avg	16	44	5	51	0.40	0.10	0.05

DC-SIGN and DENV traffic together to clathrin-coated structures for endocytosis

We next examined if DENV/DC-SIGN complexes preferentially localized to clathrin-coated pits or closely associated clathrin-coated vesicles, collectively referred to as clathrin-coated structures (CCS), the main sites of DENV entry in a number of other cell types ^19,27–29. First EGFP-clathrin was introduced by transient transfection into MX-DC-SIGN cells one day before imaging. At the time of observation, cell surface DC-SIGN was labeled with DCN46-AlexaFluor647, and AlexaFluor568-conjugated DENV was used for the observation of DENV/DC-SIGN co-transport. Fig. 2A shows a sample image on fixed cells in which some DENVs are colocalized with just DC-SIGN, while other viruses are localized to regions that contain both DC-SIGN and clathrin. We observed that 73 ± 28 % (N=17 cells, ± SD,) of the CCSs that contained DENV also contained DC-SIGN. This observation, together with our result that 80 ± 12 % (N = 24 cells, ±SD) of cell-associated virions are colocalized with DC-SIGN, is consistent with the hypothesis that DENV bound to DC-SIGN on MX-DC-SIGN cells are preferentially targeted to CCSs. Furthermore, we also applied ground state depletion (GSD) super-resolution imaging to the three components to achieve better spatial resolution than confocal imaging. The result again confirmed the colocalization of all three components (Fig. 2B). Colocalization of DENV and DC-SIGN in the same CCS was 66 ± 20 % (N=22 cells, ± SD) in primary DCs (Fig. 2C). Confocal live cell imaging also showed that DENV/DC-SIGN complexes move to cell-surface CCSs (S3 Video) in MX-DC-SIGN cells. This representative video shows a DENV/DC-SIGN complex moving together to a clathrin-coated pit and then the three components are co-transported and finally disappear together. In all, our results demonstrate that DENV and DC-SIGN travel together as a complex to CCSs for endocytosis.

Figure 2.

DENV/DC-SIGN complexes translocate to CCS for endocytosis. A. Confocal imaging shows colocalization (white arrows) of DENV-AlexaFluor568 (red) and DC-SIGN (labeled with DCN46-AlexaFluor647; blue) in the same EGFP-clathrin-coated structures (green) on MX-DC-SIGN cell surfaces. Of all DENV colocalized with CCS, 73 ± 28 % (N=17 cells, ± SD) are coincident with DC-SIGN. Scale bar, 2 µm. B. Super-resolution GSD imaging shows that DENV-AlexaFluor488 (green), CCS (labeled with Clc mAb followed by goat anti-mouse IgG-AlexaFluor532; red), and DC-SIGN (labeled with DCN46-AlexaFluor647; blue) are colocalized on MX-DC-SIGN cells. Scale bar, 150 nm. C. Confocal imaging shows colocalization (white arrows) of DENV (labeled with 1M4 followed by goat anti-human IgG-AlexaFluor 488; green) and DC-SIGN (labeled with H-200 followed by goat anti-rabbit AlexaFluor647 Fab2; blue) in the same CCS (labeled with Clc mAb followed by goat anti-mouse IgG-AlexaFluor568; red) on primary DC surfaces. Of all DENV colocalized with CCS, 66 ± 20 % (N=22 cells, ± SD) are coincident with DC-SIGN. Scale bar, 5 µm. Inset panels (right) are shown in the same colors as the whole cell image panels (top row) with the merge of DENV with CCS in the left panel of the bottom row and the merge of all three colors shown in the middle panel of the bottom row; inset scale bar, 1 µm.

DENV/DC-SIGN complexes are transported along MT after endocytosis

Measurements were carried out to determine the mode of intracellular transport of endocytic vesicles containing DENV and DC-SIGN. Expecting a microtubule (MT) - dependent mode of transport ²⁷, we transfected MX-DC-SIGN cells with ensconsin MT-binding domain (EMBD)-3×GFP ³⁰. Live cells were treated sequentially with DC-SIGN antibody DCN46-AlexaFluor647 and DENV directly conjugated with AlexaFluor568 and examined by confocal imaging at 37 °C. Cells were imaged in three channels simultaneously for up to 5 minutes, and confocal z-stack scans were collected and used to reconstruct 3D cell images to ensure that the transport along MTs was intracellular. Fig. 3A shows a DENV/DC-SIGN complex being transported along MTs. (From 20 cells, we observed DENV/DC-SIGN being transported along MT in 17 cells.) Note that time and distance traveled do not equate to speed in these circumstances, as complexes periodically pause during transport. We frequently observed DENV/DC-SIGN complexes (presumably in the same vesicle) in the cytoplasm hopping onto MT bundles and then being transported along those MTs. S4 Video gives another example of a DENV/DC-SIGN complex being transported along a curved MT. We also observed some DENV alone (i.e., without DC-SIGN) undergoing vesicular transport along MTs, as well as separation of DENV/DC-SIGN complexes after some co-transport followed by continued MT-based transport of the two separate vesicles (Fig. 3B, indicated by the superimposed trajectories). These events, which may reflect a sorting process, occasionally occurred at MT intersections with each entity following a different MT after separation, suggesting that DC-SIGN dissociates from DENV in early endosomes and is quickly recycled back to the plasma membrane.

Figure 3.

DENV/DC-SIGN complexes are transported along MTs. A, transport of DENV/DC-SIGN complex along MTs. Top row, (green) EMTB-3×GFP labeled MTs; (red) DENV directly conjugated to AlexaFluor568; (blue) DC-SIGN labeled with DCN46-AlexaFluor647. Bottom row, DENV/DC-SIGN complex transports along MT at different time points (0–12s). White circles highlight the position of the DENV/DC-SIGN complex. Scale bar, 1 µm. B, separation of DENV/DC-SIGN complex along MTs. The left whole cell image shows intracellular MTs stained by the ENTB-3×GFP (green), DENV-AlexaFluor568 (red), and DC-SIGN labeled by DCN46-AlexaFluor647 (blue). The trajectories of one representative DENV&DC-SIGN separation event are shown in three colors: the yellow trajectory shows the DENV&DC-SIGN complex transported along MT (before separation, 5 frames as shown from 0s – 4.4 s), then the two separated, DC-SIGN transported towards the upper-left (light blue trajectory) and DENV went backward towards the upper-right (white trajectory). Scale bar, 1 µm.

Internalized DENV/DC-SIGN complexes are transported to early endosomes

Previous work has shown that DENV bound to unspecified receptors in BS-C-1 cells undergo clathrin-mediated endocytosis and subsequently appear in early and then late endosomes ¹⁹. To characterize the endocytic pathway utilized by the DENV/DC-SIGN complex, DENV, DC-SIGN, and early endosomes (Rab5 as the marker) were stained sequentially and imaged using confocal microscopy (see Methods). Fig. 4 shows a representative image of DENV in early endosomes that colocalized with DC-SIGN (arrows indicate three examples of three-color colocalization). Statistical analysis showed that 68 ± 17 % (± SD, N=28 cells) of DENV in early endosomes also had DC-SIGN in the same compartments, indicating that intact DENV/DC-SIGN complexes pass through early endosomes after internalization. As a negative control, an EGF receptor fused with mRFP (ErbB2-mRFP) was expressed in the MX-DC-SIGN cells (Fig. S1C). Those images showed that no early endosomes contained both internalized DENV as well as ErbB2-mRFP, supporting the specificity of DENV/DC-SIGN colocalization.

Figure 4.

DENV and DC-SIGN colocalize in early endosomes. Top row, arrows indicate representative sites of colocalized DENV (labeled with 1M4 mAb followed by goat anti-human IgG-AlexaFluor488; green, left panel), Rab5 (labeled with anti-Rab5 Ab followed by goat anti-rabbit F(ab’)₂-AlexaFluor568; red, middle panel), and DC-SIGN (labeled with DCN46-AlexaFluor647; blue, right panel). Bottom row: (left panel) DENV colocalization with Rab5 positive endosomes; (middle panel) DENV/DC-SIGN colocalization with Rab5 positive endosomes; (right panel) inset from middle panel magnified with top row showing left to right, DENV, Rab5, and DC-SIGN and bottom row showing, left to right, DENV colocalization with Rab5-positive endosomes and DENV/DC-SIGN colocalization with Rab5-positive endosomes. Of all early endosomes containing DENV, 68 ± 17 % (N=28 cells, ± SD) also contain DC-SIGN. Scale bar, 5 µm. Scale bar in the inset, 1 µm.

Disabling the DC-SIGN cytoplasmic domain significantly reduces, but does not completely abrogate, DENV capture, endocytosis, and cell infection

Our studies show that DC-SIGN not only captures DENV, but also co-migrates with DENV to, is endocytosed, and is then transported with DENV along MTs and trafficked through early endosomes. If DC-SIGN alone facilitates entry and intracellular processing of DENV, then one might postulate that internalization motifs in the DC-SIGN cytoplasmic tail are required for viral entry and subsequent cell infection.

To test this hypothesis, we established an NIH3T3 cell line stably expressing a mutant DC-SIGN, denoted as DC-SIGN-3A, which lacks the three trafficking motifs. A site-directed mutant of DC-SIGN was made in which the three internalization motifs (the dileucine (LL), the triacidic cluster (EEE) and the functional tyrosine (Y)), were silenced by changing those residues to alanines. We also established an NIH3T3 cell line stably expressing a cytoplasmic tail-truncated DC-SIGN, denoted as DC-SIGN-Δ35 ²¹. Note that removal of the LL motif essentially prevents internalization of monoclonal antibody bound to this mutant¹³ while removal of the cytoplasmic domain markedly reduces internalization of bound monoclonal antibody ²⁶.

Nearly all cells expressed both mutants (Fig. 5A). Cells were selected by fluorescence-activated cell sorting (FACS) to have expression levels comparable to that of MX-DC-SIGN cells. The expression level of the DC-SIGN-3A mutant was about 70% of that of wild-type expressing cells (Table II and Fig. S5A). The expression level of the DC-SIGN-Δ35 mutant was about 140% of that of wild-type expressing cells (Table II and Fig. S5B). Representative confocal images were qualitatively consistent with these expression levels (Fig. S5C).

Figure 5.

Expression of the DC-SIGN mutants and their capability to capture DENV. A. FACS assays show that after sorting the percentages of cells expressing DC-SIGN-3A (upper panel) or DC-SIGN-Δ35 (lower panel) are comparable to that of the wt DC-SIGN expressing cells. B. 3D quantitative imaging shows that DC-SIGN-3A cells capture about 10% of DENV as compared to wt DC-SIGN cells (upper panel), while DC-SIGN-Δ35 cells capture about 50% of DENV as compared to wt DC-SIGN cells (lower panel).

Table II.

Expression of DC-SIGN and its mutants, and DENV internalization and infectivity mediated by DC-SIGN and its mutants

DC-SIGN	Expression relative to wt by flow (%)	DENV binding relative to MX- DC-SIGN cells after 5 min incubation at MOI=50 (%)	DENV binding relative to MX-DC-SIGN normalized for differences in expression of DC-SIGN mutants (%)	% DENV Internalized after 15 mina	DENV infectivity mediated by DC-SIGN or its mutants (% relative to MX-DC-SIGN for cells infected at MOI=3)
WT DC-SIGN	100	100	100	68	100
DC-SIGN-3A	69	12	17	24	27
DC-SIGN-Δ35	139	50	36	53	46

^aCells were incubated with DENV at MOI 50 for 5min, washed to remove unbound DENV, and then incubated without virus for another 10 min.

Cells expressing the wt or mutant versions of DC-SIGN were exposed to DENV for 5 minutes at a given multiplicity of infection (MOI), washed, fixed, and labeled such that surface-attached and internalized DENV could be distinguished by immunofluorescence (Fig. S5D). Virus could not be detected by immunofluorescence after mock (i.e., no virus) infection of MX-DC-SIGN cells (supplemental Figure S5 E). After a five minute incubation with DENV, most of the cell-associated viruses remained on the surfaces of cells expressing the wt, DC-SIGN-3A and DC-SIGN-Δ35 receptors although much less virus bound to 3A cells even after normalization for different expression levels (Fig. 5B, upper panels; Table II). By contrast, the DC-SIGN-Δ35 mutant was able to bind about one-half the number of viruses compared to wt DC-SIGN (Figure 5B, lower panels; Table II). Manzo et al. similarly found that DC-SIGN-Δ35 bound much less of an artificial HIV mimic ²⁶.

After an additional 10-min incubation with the already-attached virus (i.e., unbound virus was removed by washing at the 5 min time point), cells expressing DC-SIGN-3A internalized only ≈ 24 % of bound DENV, whereas cells expressing wt DC-SIGN internalized ≈ 68 % of bound DENV (Fig. 6A, top panel; Table II), indicating that wt DC-SIGN internalizes DENV more than twice as efficiently as DC-SIGN-3A. A similar comparison for the DC-SIGN-Δ35 mutant indicates that it has internalized nearly 80% as rapidly as wt DC-SIGN (Fig. 6A, bottom panel; Table II). Note the percentages are calculated based on the ratios of the numbers of internalized vs. initially bound viruses. Thus, these ratios reflect only the internalization efficiency; i.e., reduced internalization is not a consequence of reduced binding.

Figure 6.

DENV internalization and infectivity in wt DC-SIGN compared to mutant cells. A. % virus internalized for DC-SIGN-3A cells as compared to wt DC-SIGN cells (upper panel); the same comparison for DC-SIGN-Δ35 cells (lower panel). B. DENV infection at 24h of wt DC-SIGN, DC-SIGN-3A (upper panel) and DC-SIGN-Δ35 (lower panel) expressing cells at MOIs of 1, 3 and 10.

We compared the ability of cells expressing the wt and mutant DC-SIGN receptors to be productively infected by DENV. As described above for the entry assays, the cells were pulsed with DENV for 5 minutes, washed to remove unbound virus, and incubated for 24 hours before measuring cell infection by flow cytometry. In these experiments we used a low MOI (1–10) and not the higher MOI of 50 required for imaging studies. Compared to cells expressing the wt DC-SIGN receptor, DC-SIGN-3A cells were considerably less susceptible to infection at all MOIs tested (Fig. 6B, top panel; Table II) yet infection still occurred in significant amounts. Similarly, infection occurred in DC-SIGN-Δ35 expressing cells but at reduced levels (Fig. 6B, bottom panel; Table II).

Discussion

The mechanisms of DENV infection have been a subject of considerable debate among cell biologists and virologists. The crucial role of DC-SIGN in rendering cells susceptible to DENV infection is universally acknowledged, as is the importance of DC-SIGN’s capacity to bind the virus on cell surfaces. Indeed, in this work, quantitative confocal fluorescence microscopy confirmed that cell surface DC-SIGN is the major receptor for DENV on MX-DC-SIGN cells as well as on human primary DCs, which are one of the first targets of DENV infection. Moreover, DC-SIGN is found inside the cell in apparent compartmental colocalization with DENV. In addition, the fact that DC-SIGN mutants lacking either internalization motifs in their cytoplasmic domains or almost the entire cytoplasmic domain still support DENV infection, albeit in lower amounts, indicates that a co-receptor of some sort must be involved.

By applying quantitative fluorescence imaging techniques, we observed that DC-SIGN captures DENV and remains associated during lateral transport in the plasma membrane to CCS and subsequent internalization by endocytosis. Presumably, the lateral transport of the DENV-DC-SIGN complex is enabled by the considerable diffusional mobility of DC-SIGN (Table I) in the plane of the membrane. Once inside the cell, the complex is found in Rab5-positive early endosomes. In fact, Rab5-containing early endosomes are required for DENV infection ³¹. Intracellular transport of vesicles to and from endosomes occurs on MTs ^27,32. As expected, vesicles containing DENV/DC-SIGN complexes also moved on MTs during viral entry. An important remaining question that we continue to investigate is the point at which DC-SIGN releases its cargo. We have described observations of dissociation before and during MT trafficking, and we have measured reduced DENV/DC-SIGN colocalization in early endosomes compared to that on the cell surface. These results strongly suggests that dissociation occurs during endosome maturation, likely triggered by the dropping pH as previously described for Man₃₀-BSA binding to isolated DC-SIGN CRDs ³³. It has been suggested that, in the case of the phleboviruses which use DC-SIGN as an entry receptor, that after dissociation in the early endosomes, DC-SIGN recycles back to cell surfaces ¹¹.

Our conclusion that DC-SIGN is more than simply an attachment factor but also enters the cell with DENV is in apparent contradiction with hypotheses involving handoff of DENV to another receptor in the plasma membrane for entry, e.g. ¹⁵. The original proposition of the “attachment factor only” hypothesis was based on the fact that cells expressing internalization motif-deficient DC-SIGN can still be infected by DENV ¹³. However, these investigators did not examine the details of DENV interactions with cell surface DC-SIGN and the subsequent transport and entry kinetics of DENV, all of which indicate that wt DC-SIGN has a role, remaining to be defined in detail, in DENV infection downstream of attachment. Indeed, it has been suggested ³³ that the cytoplasmic domain motifs of DC-SIGN not only function as an internalization signal but also as endosomal/lysosomal targeting signals, consistent with our observation of DENV/DC-SIGN complexes in endosomal compartments.

Moreover, in these earlier studies of the role of DC-SIGN in DENV infection, host cells were exposed to virus at an MOI of 10 for 2 hours before washing and then virus proliferation was measured after 40 hours of subsequent incubation ¹³. Such a long exposure at a high MOI allows ample time for uptake of infectious virions via a less-efficient surrogate pathway such as adventitious endocytosis. Furthermore, 40 hours of post-exposure incubation would most likely obfuscate significant differences in early cell infection events, such as the reduced (but not eliminated) virus binding and endocytosis efficiencies that we observed in DC-SIGN-3A and DC-SIGN-Δ35 cells.

In fact, quantitative 3D counting by imaging showed that DC-SIGN-3A captured only about 1/10 as much DENV as wt DC-SIGN after a 5-minute exposure while DC-SIGN-Δ35 captured about ½ as much viruses (Table II). Since we sorted cells so that expression levels were roughly comparable to wt DC-SIGN (Fig. S5A, S5B and Table II) and qualitatively confirmed the sort by confocal microscopy (Fig. S5C and Table II), the observed differences in binding were not, in the main, due to reduced surface expression of the mutant proteins.

Why is DENV binding to the mutants less efficient and in the case of DC-SIGN-3A much less efficient? We hypothesize that DENV binding to DC-SIGN is stabilized by the same conformational changes (à la induced-fit) that initiate downstream signaling through the cytoplasmic domain; without these key internalization motifs or the entire cytoplasmic domain, such conformational changes would be incomplete, yielding a less stable DENV/DC-SIGN complex. It also remains possible that alterations in DC-SIGN cluster micro- or nano-architecture ³⁴ resulting from removal of the internalization motifs or the entire cytoplasmic domain could decrease binding efficiency. From the point of view of membrane biology and immunology, these issues remain as important topics for future study.

In addition, we observed at ≤15 minutes post-incubation with virus less than half of the endocytosis rate by DC-SIGN-3A cells compared to cells expressing wt DC-SIGN (Fig. 6A, top panel; Table II). These cells were also infected less efficiently (Fig. 6B, top panel; Table II). Similarly, DENV bound to DC-SIGN-Δ35 was internalized less efficiently than DENV bound to wt DC-SIGN (Fig. 6A, bottom panel; Table II) and reduced, but not completely abrogated, infectivity was found for cells expressing DC-SIGN-Δ35 (Fig. 6B, bottom panel; Table II). Thus, we conclude that the presence of the intact, unmodified DC-SIGN cytoplasmic tail increases both the efficiency of capture and entry of DENV into cells and subsequent infection when compared to cells expressing a mutated or truncated DC-SIGN cytoplasmic domain. Our results are consistent with a previous report finding that mutating the cytoplasmic internalization motifs of DC-SIGN results in markedly reduced receptor activity and phagocytosis/endocytosis of mannosylated ligands in human myeloid cells ³⁵.

It should be noted that while internalization is a prerequisite for infection, the latter will require additional steps, which may be rate limiting. wt DC-SIGN is able to bind much more DENV than the 3A mutant. And at a high MOI of 50, wt bound to DENV is clearly internalized more effectively than 3A mutant as would be expected since the latter lacks the internalization motifs. However, when considering infectivity, virion binding and internalization, while obligatory are not necessarily rate limiting so the differences in infectivity may not be as pronounced as expected from differences in the binding and internalization. Thus, even though fewer virus particles per cell bind to DC-SIGN-3A, virus bound to the mutant DC-SIGN does enter the cell and leads to infection, albeit to a lesser degree compared to wt DC-SIGN expressing cells.

Finally, while our results strongly suggest that DC-SIGN functions as both an attachment and entry receptor for DENV, what can we conclude about the involvement of a co-receptor in DENV endocytosis via DC-SIGN? As mentioned, it would appear that DC-SIGN does not hand the virus over to a separate entry receptor in the plasma membrane, because DC-SIGN is trafficked intracellularly with virus. However, although the binding, endocytosis efficiency and infection rates of DC-SIGN-3A and DC-SIGN-Δ35 expressing cells are significantly impaired, DENV still can enter the cells and lead to productive infection. This indicates that even without DC-SIGN’s internalization motifs, viral internalization can still occur via a presumed co-receptor that remains to be identified. (Confocal imaging of anti DC-SIGN mAb or Fab revealed that essentially insignificant amounts of the mutant forms of DC-SIGN were internalized in 5 min at 37 C (data not shown). These amounts were far less that the 10% of the 3A mutant bound to DENV that was internalized after 5 minutes of incubation with DENV so our hypothesis that a co-receptor must be involved stands.) At this point, the question of whether the putative co-receptor is pre-associated with DC-SIGN or is recruited after DENV binding remains open.

Our results expand the role of DC-SIGN and its putative co-receptor in the physiological setting where rapid and highly efficient capture and endocytosis of pathogen cargo is a “raison d’être” of the DCs.

Materials and Methods

Reagents

Antibodies for DC-SIGN Labeling

For DC-SIGN immunostaining we used either DCN46 mAb (BD Biosciences) or 120507 mAb (R&D Systems). These mAbs were either directly conjugated with AlexaFluor488, AlexaFluor568, or AlexaFluor647 using the respective AlexaFluor Antibody Labeling Kits (Cat # A-20181, A-20184, and A-20186, Life Technologies), or digested into Fab fragments using the Pierce Fab Micro Preparation Kit (Pierce) and then conjugated with AlexaFluor dye. Except where noted below, all anti-DC-SIGN antibodies and antibody fragments were used at roughly 8–25 µg/mL, which was empirically determined to ensure sufficient cell-surface saturation of DC-SIGN. These were the minimum concentrations sufficient to achieve maximum fluorescence within a region of interest, as described previously ³⁶. All anti-DC-SIGN antibodies and antibody fragments were tested with NIH3T3 cells not expressing DC-SIGN to insure specificity (e.g., Fig. S1B). In some measurements (e.g., Fig. S1B), cells were treated first with unlabeled DCN46 mAb and then with goat anti-mouse F(ab’)₂-AlexaFluor568 (Life Technologies).

DENV and DENV Labeling

DENV (serotype 1, West Pac 74) was grown in Vero cells and purified as described previously ³⁷. The MOI we use in the text refers to the amount DENV added to a known number of Vero cells to obtain a set number of plaque forming units (PFU). Thus, the MOI is a relative measure for our studies and is not easily related to the actual number of virus particles added to the cells, because DENV is produced in cell culture as a mixture of mature, partially mature and immature forms ³⁸.

A variety of DENV labeling procedures were employed. For some measurements, DENV was labeled by direct conjugation with AlexaFluor488 or AlexaFluor568 by using Microscale Protein Labeling Kits (Life Technologies). For other measurements, the anti-DENV mAbs 2H2, 1M4, and 1F4 were used. The mAb 2H2 was commercially obtained (EMD Millipore), whereas the 1M4 and 1F4 mAbs were isolated and prepared as described ³⁹. In these cases, DENV was labeled with these mAbs at concentrations ranging from 8–20 µg/mL. For direct immunofluorescence observations, 2H2 was conjugated with AlexaFluor488 and 1F4 was conjugated with AlexaFluor568. For indirect immunofluorescence observations, cell treatment with 1M4 mAb was followed by treatment with goat anti-human IgG-AlexaFluor488 (Life Technologies).

Other Labeling

In some measurements, clathrin was labeled by using a clathrin light-chain (Clc) specific mouse mAb (Santa Cruz Biotechnology) followed by goat anti-mouse IgG-AlexaFluor532 (Life Technologies). In other measurements, MX-DC-SIGN cells were transiently transfected with pEGFP-clathrin, a kind gift from Deborah Brown at Stony Brook University (NY) ⁴⁰. For visualization of MTs, cells were transiently transfected with ensconsin MT-binding domain (EMTB)-3×GFP; the construct was a gift from William Bement (Addgene plasmid # 26741 and ⁴¹). ErbB2-mRFP construct was a kind gift from Dr. Ichi N Maruyama at Genome Institute of Singapore (Singapore). For visualization of Rab5, rabbit anti-Rab5 Ab (Ab18211) was purchased from Abcam (Cambridge, MA). Cells were treated with this Ab followed by goat anti-rabbit F(ab’)2-AlexaFluor568 or AlexaFluor647 conjugates (Life Technologies).

Other Reagents

Cells were fixed using 4 % paraformaldehyde (PFA). Cell permeabilization buffer (Perm buffer) was 2 % BSA, 0.1 % saponin, 0.02 % NaN₃ in sterile DPBS.

Construction of DC-SIGN-3A

Construction of the internalization motifs-silenced mutant, DC-SIGN-3A, was based on the DC-SIGN LL/AA in pTRIP vector, a kind gift from Dr Ali Amara at Unité d’Immunologie Virale ¹³. DC-SIGN LL/AA EEE/AAA was made by using the Q5 Site-directed mutagenesis kit (NEB) with the following two primers: 5’ GGCACAGCTGAGAGGCCTTGGATTCCGACAG 3’, and 5’ GCCGCCGCGGCGCCCAGCTGCTG 3’. The resulting DC-SIGN LL/AA EEE/AAA was then used as the template to make the three motifs-silenced construct, DC-SIGN LL/AA EEE/AAA Y/A (abbreviated as DC-SIGN-3A), via an overlap PCR strategy. The first of two PCR runs employed two sets of primers and two templates as follows: S1 5’ ACCTCCATAGAAGACACCGACTC 3’, and S2 5’ CACCCTGCTAAGCTCTTGGCTCCTCGAGTCTGTCGG 3’, in which primer S2 contains the site mutation of Y/A (in italics) using pTRIP DC-SIGN LL/AA EEE/AAA as the template; L1 5’ CCGACAGACTCGAGGAGCCAAGAGCTTAGCAGGGTG 3’, in which L1 contains the Y/A mutation (in italics), and L2 5’ ATCGTCGACAAAAGGGGGTGAAGTTCTG 3’, using pMX DC-SIGN as the template. The resulting two PCR products were then purified and mixed as templates for the overlap PCR, using S1 and L2 as primers. The final PCR product was then digested with BamHI and SalI, and inserted into the pMX DC-SIGN vector digested with the same two restriction enzymes. The final construct was confirmed by DNA sequencing.

Cells and transfections

Culture conditions of NIH3T3 cells are described elsewhere ²². NIH3T3 cells stably expressing DC-SIGN (denoted as MX-DC-SIGN cells) were obtained through the NIH AIDS Research and Reference Reagent Program, Division of AIDS, NIAID (NIH: NIH3T3/MX-DC-SIGN, Cat # 9947, from Drs. Li Wu and Vineet N. KewalRamani ²⁰). The pMX-DC-SIGN-Δ35 plasmid was a kind gift from Dan Littman, New York University and Howard Hughes Medical Institute, New York. NIH3T3 cells were transduced with the two plasmids, pMX-DC-SIGN-Δ35 and pMX-DC-SIGN-3A, using the Phoenix-ECO retroviral packaging system (ATCC CRL-3214), respectively. The cells were then subjected to FACS after staining with 120507-AlexaFluor488 mAb; cells expressing DC-SIGN-3A exhibited a similar expression level compared to that of the MX-DC-SIGN cells; cells expressing relatively high amounts of DC-SIGN-Δ35 were selected for subsequent proliferation. For co-expression of DC-SIGN with either EGFP-clathrin, EMTB-3×GFP, or ErbB2-mRFP, the plasmids were introduced into cells by using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol.

Human monocyte-derived immature DCs were prepared as described elsewhere ^21,22,42. Isolated monocytes were seeded on 35 mm glass-bottom dishes (MatTek) and cultured in RPMI-1640 media containing 10 % fetal bovine serum (FBS), 500 U/ml human IL-4 and 800 U/ml human GM-CSF (Peprotech, Rocky Hill, NJ) for one week in a 37 °C incubator with 5 % CO₂.

Labeling for fluorescence microscopy

For colocalization between DENV and cell surface DC-SIGN clusters (Fig. 1), DENV was thawed from −80 °C and incubated with MX-DC-SIGN cells at MOI=10 for 30 min at 37 °C in a CO₂ incubator. The cells were then washed thoroughly with DPBS, and fixed with 4 % PFA for 20 min. After that, the cells were washed several times with DPBS, and blocked with 1 % NMS for at least 30 min. Finally, the cells were sequentially stained with 2H2-AlexaFluor488 mAb for DENV and DCN46-AlexaFluor647 for DC-SIGN, with sufficient washing between each staining. For colocalization between DENV and cell surface DC-SIGN clusters on primary DCs, DENV was incubated with DCs at MOI=50 for 5 min at 37 °C in a CO₂ incubator, then cells were fixed and stained in the same way as for the MX-DC-SIGN cells. Here we used a lower MOI but longer incubation time for MX-DC-SIGN cells, and a higher MOI but shorter incubation time for DCs; both approaches were designed to get sufficient DENV binding on cell surfaces. Moreover, the high MOI does not indicate very large amounts of DENV bound to cells, because a large portion of the DENV remained in the media and was washed away after the incubation. As negative controls, native NIH3T3 cells were stained with the same antibodies described here and showed no non-specific staining (Fig. S1B).

For colocalization between EGFP-clathrin and DC-SIGN (Fig. S1A), plasmids were introduced into NIH3T3 cells as described above. One day after transfection, the cells were fixed with 4 % PFA as described above, and cell surface DC-SIGN was stained with DCN46 mAb followed by goat anti-mouse F(ab’)₂-AlexaFluor568. For three-color colocalization of DENV, DC-SIGN, and clathrin on MX-DC-SIGN cells (Fig. 2A), EGFP-clathrin was introduced into cells by using Lipofectamine 2000 reagent one day before imaging. Purified DENV was directly conjugated to AlexaFluor568 as described above and incubated with MX-DC-SIGN cells for 30 min. Then the cells were washed and fixed with 4 % PFA, blocked, and DC-SIGN was stained with DCN46-AlexaFluor647. For three-color colocalization of DENV, DC-SIGN, and clathrin on DCs (Fig. 2C), DENV was incubated with DCs at MOI=50 for 5 min at 37 °C in a CO₂ incubator. Then the dishes were washed, media was replaced, and the dishes were incubated for another 5 min before fixation, permeabilization, and blocking with 1 % goat serum. The cells were then stained sequentially with sufficient washes in between each staining: DENV was stained with primary 1M4 mAb (against E protein, from de Silva lab) followed by secondary goat anti-human IgG AlexaFluor488 conjugate; clathrin was labeled using a clathrin light-chain (Clc) specific mouse mAb (Santa Cruz Biotechnology) followed by goat anti-mouse IgG-AlexaFluor568 (Life Technologies); DC-SIGN was stained with H-200 Ab (Santa Cruz) followed by goat anti-rabbit AlexaFluor647 Fab2 conjugate (Life Technologies).

For GSD super-resolution imaging and three-component colocalization of DENV, DC-SIGN, and clathrin (Fig. 2B), MX-DC-SIGN cells were plated on #1.5 18 mm circular cover glasses (Sigma) one day before imaging. DENV was directly conjugated with AlexaFluor488, endogenous clathrin was labeled with the Clc mAb as described above followed by secondary goat anti-mouse AlexaFluor532 IgG, and DC-SIGN was stained with DCN46-AlexaFluor647. DENV-AlexaFluor488 were incubated with MX-DC-SIGN cells for 30 min, then washed with DPBS and fixed with 4 % PFA. The cells were then permeabilized using Perm buffer, and placed in blocking buffer (1 % NMS in Perm buffer) for 1–3 h at room temperature or overnight at 4 °C. After that, the cells were sequentially stained with Clc mAb followed by goat anti-mouse AlexaFluor532 IgG, then DCN46-AlexaFluor647, with sufficient washing (> 5 times) and blocking between each staining step. The cover slips with stained cells were then mounted on depression slides with freshly prepared 10 mM β-mercaptoethylamine (MEA) in PBS (adjusted to pH 7.4) for GSD imaging.

For confocal three-component imaging and colocalization of DENV, DC-SIGN, and early endosomes or DENV, ErbB2, and early endosomes (Fig. 4 and Fig. S1C), MX-DC-SIGN cells were plated on 35 mm (#1.5) MatTek dishes 1–2 days before experiments. For ErbB2 imaging, the ErbB2 plasmid was introduced into cells by using Lipofectamine 2000 reagent one day before imaging. Cells were then incubated with DENV at MOI=50 for 5 min, washed with fresh media, and incubated a further 10 min at 37 °C. After that, the cells were washed with DPBS and fixed with 4 % PFA, permeabilized with Perm buffer, and blocked with 1 % goat serum (in Perm buffer) overnight at 4 °C. At the time of observation, cell samples were sequentially stained as follows (with sufficient washing and blocking between each step): (1) 1M4 mAb followed by goat anti-human IgG AlexaFluor488 conjugate; (2) anti-Rab5 Ab followed by goat anti-rabbit F(ab’)₂-AlexaFluor568 conjugate (or AlexaFluor647 conjugate when imaging ErbB2-mRFP); (3) DCN46-AlexaFluor647 (or nothing when imaging ErbB2-mRFP).

For all confocal live cell imaging, including surface co-transport of DENV with DC-SIGN (S2 Video) or DENV/DC-SIGN/CCS together (S3 Video) and intracellular transport of DENV & DC-SIGN along MTs (Fig. 3), EGFP-clathrin and EMTB-3×GFP plasmids, respectively, were introduced into MX-DC-SIGN cells by using Lipofectamine 2000 reagent one day before imaging, and DENV directly conjugated with AlexaFluor dyes was used. DC-SIGN was incubated with DCN46 mAb-AlexaFluor647 or DCN46 Fab-AlexaFluor568 for 5–10 min, washed, and maintained in phenol red-free RPMI-1640 (Invitrogen) during imaging.

For TIRF microscopic measurements of DC-SIGN surface diffusivity (Table I), before imaging, media was supplemented with 15 µM chlorpromazine for 30 minutes, followed by DCN46 Fab-AlexaFluor568 for 10 minutes, and finally replaced with RPMI-1640 containing 15 µM chlorpromazine. Imaging was carried out at 37 °C using an environmental chamber integrated with the microscope stage.

Confocal imaging and data analysis

Confocal imaging was carried out on a Fluoview FV1200 laser scanning microscope (Olympus) equipped with an environmental heating chamber for live cell imaging. An oil-immersion objective (60X, NA 1.35) was used, and a multi-coating dichroic mirror (DM450/488/559/635) was used to separate excitation laser lines. The excitation/emission filter setups for each channel are as follows: AlexaFluor488 (or GFP/EGFP) channel, excitation wavelength 488 nm, emission wavelength 500–540 nm; AlexaFluor568 (or mRFP) channel, excitation 559 nm, emission 590–630 nm; AlexaFluor647 channel, excitation wavelength 635nm, emission 655–755 nm. To minimize crosstalk between channels, images were taken sequentially by channel. For all fixed samples, cells after sufficient washing were immersed in DPBS; for live cell imaging, an environmental heating chamber was used and cells were in phenol red-free RPMI-1640. For counting of total virus uptake by each cell, 3D z-stack imaging was performed with an optimized z dimension (0.52 µm/section).

Colocalization analysis was carried out by using the ImageJ plugin JACoP ⁴³. For two-color colocalization, first thresholds were set for both channels after loading the two images. As both DENV particles and DC-SIGN clusters are distinctive objects on images, we chose object-based analysis to quantitatively calculate colocalization percentages. Object-based methods compare the intensity centroids of two objects in two channels, therefore eliminating the effect of size differences between the two objects that would otherwise affect colocalization percentage calculation in other methods like pixel-based analysis. Our criterion for colocalization was that one object centroid falls into the area of the other object after image segmentation. Object-based analysis gives both the fraction of DENV particles colocalized with DC-SIGN and also the fraction of DC-SIGN clusters with DENV bound (i.e. colocalized with DENV). As the number of DENV on cells is usually much smaller than the total number of DC-SIGN clusters, we calculated colocalization percentages by dividing the number of DENV colocalized with DC-SIGN by the total number of DENV on cells. For three-color (three-component) colocalization, we first analyzed the two-color (green and red) colocalization using object-based analysis then used the images that show only colocalized particles to compare with the third color image (blue in Fig. 2 and Fig. 4). Still, as DC-SIGN, CCS and endogenous early endosomes (Rab5-marked) are much more abundant than DENV particles, we calculated the colocalization percentages by dividing the number of DENV colocalized with both DC-SIGN & CCS (or both DC-SIGN & Rab5-marked endosomes) by the total number of DENV colocalized with CCS (or total number of DENV colocalized with Rab5-marked endosomes).

Single-particle TIRF imaging and data analysis

The TIRF microscope was selected because the relative flatness of the ventral membrane (as compared to the dorsal) is better suited for two-dimensional tracking and motion analysis and because the minimal illumination depth of TIRF provides excellent signal over cytoplasmic background for membrane observations. Moreover, there is an exponential decay of evanescent wave intensity in the direction normal to the coverslip so that major axial motions would result in a change in spot brightness and therefore would be less likely to be linked in the tracking. To further minimize DC-SIGN departure from the membrane, an endocytosis inhibitor (15 µM chlorpromazine) was included in all experiments ^28,44–46. To minimize potential antibody cross-linking artifacts, DC-SIGN was labeled with Fab fragments. In these experiments, a large majority of the observed fluorescence could be quenched using potassium iodide ⁴⁷, demonstrating that we were only observing DC-SIGN on the plasma membrane (i.e., we were not exciting DC-SIGN that might have been in endocytic vesicles).

TIRF images were collected by a Hamamatsu Image-EM CCD with 16 µm pixels mounted on an automated Olympus IX83 inverted microscope with a UAPON 100 × Oil immersion TIRF objective with Correction Collar (NA 1.49; Olympus). Cells were maintained at 37 °C and 5 % (v/v) CO₂ in a Tokai Hit environmental chamber, and an Olympus Zero Drift Compensation (ZDC2) laser autofocus system was employed during video recordings. Videos were recorded under constant laser illumination at a frequency of 25 Hz (i.e., 40 ms exposures).

Particle tracking and motion analysis were done using u-track 2.1.1 (Harvard) ^23,48 with MATLAB R2013a (MathWorks). In brief, u-track uses a trajectory classification scheme employed by Ewers et al. ²⁴ based on earlier work of Ferrari et al. ²⁵. Moments (µ_m), where m is the order of the moment and ranges from 0 to 6, are calculated for a generalized displacement equation as a function of lag time (tau). A plot of log [µ_m(tau)] vs log tau yields α_m as the slope and log[4 D_m] as the y intercept where D_m is the generalized diffusion coefficient. The plot of α_m vs m is the moment scaling spectrum (MSS). For strongly self-similar processes, the curves are linear where S_MSS is the slope. In this analysis, S_MSS =0.5 is random, Brownian diffusion; S_MSS < 0.5 is anomalous subdiffusion and S_MSS > 0.5 is anomalous superdiffusion. Because of the linearity of the MSS for strongly self-similar processes, this procedure affords less error in fitting and a clearer distinction between motion types is possible ²⁴. The MSS slopes that are classified as sub-, Brownian, or super diffusive depend on the length of the particular trajectory and are automatically assigned by the u-track software.

Trajectories and motion analysis results consisting of motion type (e.g., Brownian) and parameters (e.g., diffusion coefficient) for each track were exported from MATLAB. Microsoft Excel was used to merge data sets from ≥3 cells and remove anomalous (i.e., subdiffusive and superdiffusive) data before statistical analysis. Plots which are summarized in Table I were generated and Kruskal-Wallis nonparametric one-way ANOVA with Dunn’s multiple-comparison post-test was performed using GraphPad Prism version 5.00 for Windows, GraphPad Software, San Diego California USA, www.graphpad.com.

Super-resolution GSD imaging

GSD imaging of the three-component colocalization (DENV, DC-SIGN, and clathrin-coated structures) was carried out using a Leica SR GSD 3D microscope (Leica). The excitation laser lines for the three channels were 488 nm, 532 nm, and 633 nm, respectively. A custom filter cube (GSD Quad, Leica) was used to separate excitation and emission wavelengths, and the signal was collected by an Andor iXon Ultra 897 EMCCD camera. Image collection and analysis was carried out using Leica software on the computer supplied with the GSD microscope.

Imaging DENV entry kinetics

For imaging of DENV entry kinetics and subsequent analysis (Fig. 6; Fig. S5D), MX-DC-SIGN cells, NIH3T3 cells, and either DC-SIGN-3A or DC-SIGN-Δ35 cells were plated on 35 mm (#1.5) MatTek dishes one day before experiments. The cells were incubated with DENV at MOI=50 for 5 min, then washed, then either directly fixed with 4 % PFA, or further incubated for 10 min and then fixed with 4 % PFA. For both the 5-min and 15-min sample sets, cells were first blocked and stained with 2H2-AlexaFluor488 mAb (green, for cell surface-bound DENV staining), then permeabilized with Perm buffer and stained with 1F4-AlexaFluor568 (red, for whole cell DENV staining). In overlay images, surface-bound DENV appeared yellow as both 2H2-AlexaFluor488 and 1F4-AlexaFluor568 stained these viruses, while internalized DENV appeared red as 1F4-AlexaFluor568 was added after permeabilizing the cells.

Vero cells employed in our study produce a mixture of mature, partially mature and immature viruses. Richter et al. ⁴⁹ had shown previously that DC-SIGN can mediate the uptake of immature DENV into DC. The mAb 2H2 recognizes PrM on immature and partially mature DENV, whereas the 1F4 mAb recognizes the E glycoprotein on all virus types. Few surface-bound viruses were pure red (Fig. S5D), indicating that the large majority of viruses bound on cell surfaces are immature or partially mature. In other words, the Vero cells appear to produce very little purely mature DENV (serotype 1) and, apparently, partially mature virus can still be infective (see below). It is possible that mature DENV could use other entry pathways.

For DENV entry kinetic analysis, we counted viruses on the surface compared to those both on the surface and inside. Virus counting was performed using the 3D objects counter in ImageJ. 3D z-stack confocal images of whole cells were loaded and split into separate channels, and the numbers of particles in each channel was counted. The counts from AlexaFluor488 (green channel) represented the surface-bound DENV while the counts from AlexaFluor568 (red channel) represented the total number of both surface-bound and internalized DENV. The number of internalized DENV was computed by subtraction. The percentage internalized was computed from the ratio of the number of DENV internalized to the total number of surface-bound and internalized DENV.

Infectivity assay

For comparison of DENV infectivities (Fig. 6 B), MX-DC-SIGN, DC-SIGN-Δ35 or DC-SIGN-3A, and wt NIH3T3 (as a negative control) cells were plated on 35 mm cell culture petri dishes at 3 × 10⁴ cells/dish one day before infection. Cells were exposed to DENV at MOI = 1, 3 or 10 for 5 min, then washed and replaced with fresh low-FBS (2 %) media. (The reference MOI was measured in Vero cells ⁵⁰. At 24 h, cells were fixed, permeabilized, stained with 2H2-AlexaFluor488 mAb, and assayed with a Guava easyCyte 8HT flow cytometer as described before ⁵¹. Data were analyzed using Guava Express software (Guava).