To understand the relative contribution of cell contact to viral dissemination, we set out to carefully measure how much virus is passed from T cell to T cell by comparing direct-contact routes to cell-free routes. Previous immunofluorescence microscopy studies have documented the relocalization of viral and cellular receptors to the site of infected-cell-to-uninfected-cell contact and revealed accumulated Gag in putative target T cells (28
). To quantify viral uptake, we established an ultrasensitive flow cytometry assay that employed a novel infectious molecular clone of HIV, HIV Gag-iGFP (27
). This infectious viral construct carries an internal insertion of GFP into the core structural protein, Gag, stoichiometrically loading each virus particle with an estimated 5,000 GFP molecules (6
). This enables direct measurements of viral uptake with exceptional sensitivity.
Highly efficient transfer of fluorescent HIV-1 into target cells mediated by cell-to-cell contact.
To measure the efficiency of cell-to-cell viral transfer, we modified a previously described assay utilizing distinct donor and target CD4+
T cells (28
). The donor cells were Jurkat CD4+
T cells expressing HIV Gag-iGFP, and the target cells were purified primary human CD4+
T cells labeled with the red fluorescent dye CMTMR. The red fluorescent dye was used to distinguish HIV-naïve target cells from the HIV Gag-iGFP-expressing donor cells. To synchronize our infections and to facilitate studies of viral mutants, we transduced Jurkat T cells with HIV Gag-iGFP by Amaxa nucleofection. Transducing the donor cells also ensured that the virus that we studied was produced in the donor cells and was not derived from carryover of high-dose, infectious inocula.
HIV-expressing donor cells were carefully washed to remove cell-free virus and were incubated with CMTMR-labeled target cells. Within 3 h, 22% of the CD4+ target T cells acquired high levels of GFP fluorescence (Fig. ). Analysis of the forward- and side-scatter plots of the red target cells that had acquired green fluorescence revealed that they were not simple aggregates with infected donor cells. When control GFP-expressing Jurkat donor cells were mixed with the same target cells, no transfer of GFP was observed (Fig. ). These control Jurkat cells express GFP fluorescence at an intensity comparable to that of the brighter HIV Gag-iGFP-expressing cells and were subjected to an Amaxa transduction protocol without proviral DNA.
FIG. 1. Massive VS-mediated transfer of HIV to target cells measured by flow-cytometry-based detection of HIV Gag-iGFP. (A) Target cells, which were CMTMR-labeled CD4+ primary T cells (R1 gate), were mixed with donor cells, which were HIV Gag-iGFP-expressing (more ...)
In contrast to the findings for GFP staining after exposure to infected cells, only faint GFP staining was observed when cells were exposed to cell-free viral stocks of HIV Gag-iGFP containing 100 ng/ml of p24 (Fig. ). Incubation of the target cells with supernatants collected over a 24-h period from HIV Gag-iGFP-nucleofected Jurkat cells (5 ng/ml p24) did not produce a detectable increase in cellular fluorescence (data not shown). These results strongly suggest that during coculture, the levels of cell-free viral transfer are very small relative to those of cell-to-cell transfer.
To examine whether the virus acquired by cells through cell-to-cell or cell-free exposure were bound to the cell surface, we digested the cells with trypsin to strip away cell surface viral proteins after the 3-h exposure to cell-free virus or HIV-expressing cells. The VS-mediated transfer was resistant to trypsin treatment (Fig. ), whereas the small amount of fluorescence acquired by cell-free virus was largely trypsin sensitive (Fig. ). To ensure that the trypsin treatment was effective, we conducted immunostaining of target cells treated with trypsin and found a total loss of cell surface CD4 (data not shown). We conclude that VS-mediated viral transfer results in rapid sequestration of viral particles into a trypsin-resistant compartment. Both the magnitude of fluorescence and the resistance to trypsin distinguish the VS-mediated transfer from cell-free viral transfer (Fig. ).
Because Amaxa nucleofection could lead to extraordinarily high levels of viral gene expression, we compared the fluorescence of Amaxa-nucleofected Jurkat cells to that of the Jurkat cells subjected to spinoculation by infectious HIV Gag-iGFP. The levels of HIV Gag-iGFP fluorescence in the two conditions were not dramatically different (Fig. ), yet the Amaxa nucleofection allowed us to avoid using large input of cell-free virus to initiate the infections. We have found that MT4 cells are highly permissive for replication of HIV Gag-iGFP (27
). With infected MT4 cells as donor cells, they were able to engage in efficient VS-mediated viral transfer when coincubated briefly with primary CD4+
T cells (Fig. ).
FIG. 2. HIV Gag-iGFP-infected cells can engage in VS-mediated transfer. (A) Levels of HIV Gag-iGFP expressed by Amaxa-nucleofected Jurkat cells are comparable to those of HIV Gag-iGFP-infected Jurkat cells at 48 h after nucleofection/infection. Infection of Jurkat (more ...)
To quantify the overall magnitude of viral transfer into the population of recipient T cells, we calculated the relative fluorescence transfer (RFT) index as the product of the average increase in fluorescence intensity of the target cells and the percentage of GFP+ cells (Table ). The relative fluorescence intensity of the GFP-positive fraction was normalized to the background fluorescence of the GFP-negative population. VS-mediated viral transfer resulted in a large fraction (20%) of target cells that acquired a 10-fold-greater average fluorescence than nonexposed cells (Table ). Exposure of target cells to high concentrations of cell-free virus (100 ng/ml of p24 antigen) resulted in a very low percentage of cells (1.0%) with a weak increase in fluorescence over background. Decreasing the input of cell-free virus resulted in small decreases in the RFT index (Table ). The RFT index from cell-to-cell conditions was 178, which was 93-fold higher than that obtained with 100 ng/ml of cell-free virus (RFT of 1.91). The large quantities of cell-free virus tested were well beyond the levels of p24 obtained from infected Jurkat cells in vitro, but they were required to observe small, dose-dependent shifts in fluorescence. Jurkat cell viral supernatants harvested more than 24 h posttransfection, which contained 5 ng of p24 antigen, did not result in a detectable shift in fluorescence when exposed to target cells (data not shown).
Quantitative comparison of cell-to-cell viral transfer to cell-free viral transfer
Under VS-forming conditions, we found that the amount of viral p24 released by the donor cells during the 3-h coculture period was 0.5 ng, a value 200-fold lower than the concentration of cell-free virus tested. Assuming that dilution of these lower levels of cell-free virus is likely to result, at best, in a linear decrease in cell-free uptake, we can estimate the difference between the efficiency of cell-free transfer and that of cell-to-cell transfer to be 200-fold greater than the calculated difference in the RFT index, or 18,600-fold (93-fold × 200-fold). We conclude that the efficiency of viral transfer from a cell-cell mode is at least 92- to 18,600-fold more efficient than that of a cell-free mode of uptake.
We next examined the time course of VS-mediated viral transfer. The rate of transfer was rapid and efficient, with the onset of transfer occurring after 1 h and increasing steadily over a 4-h time course. When mixed at a fixed ratio of donor Jurkat cells to acceptor CD4+ T cells of 1:1, a 4-h incubation resulted in more than 20% of the target cells exhibiting strong green fluorescence (Fig. ). In addition, we altered the ratio of infected donor cells to target cells to test its effect on the efficiency of viral transfer. At donor-to-target ratios of 1:3, 1:1, and 3:1, the efficiency of transfer increased as the donor-to-target ratio was increased (Fig. ). We thus observed a rapid, cell-ratio-dependent increase in virus-associated fluorescence in target cells.
FIG. 3. VS-mediated viral transfer is rapid and dependent on donor target ratio. (A) A 1:1 ratio of HIV Gag-iGFP-expressing donor cells and CD4+ target cells was mixed and incubated for 0, 1, 2, 3, or 4 h. The percentage of GFP+ cells is plotted (more ...) Dependence of VS-mediated HIV-1 transfer upon Env and CD4 interactions.
Previous studies have illustrated that the viral Env protein is relocalized to the site of cell-cell contact during VS formation (28
). We therefore examined mutant HIV clones to determine the dependence of viral transfer upon Env. An Env-deficient HIV-1 clone, HIV Gag-iGFP ΔEnv, was nucleofected to generate donor cells that abundantly express all HIV-1 genes except for Env (Fig. ). Jurkat cells were nucleofected with the Env-deleted construct with an efficiency comparable to that of native HIV Gag-iGFP (Fig. ). However, when HIV Gag-iGFP ΔEnv-expressing donor cells were mixed with target cells, only a background level of HIV Gag-iGFP transfer to target cells was observed (Fig. ). The massive transfer of virus to target cells therefore is completely dependent on expression of Env in the donor cell.
FIG. 4. Efficient VS-mediated viral transfer is dependent upon Env and Gag. (A) Western blot of cell lysates of viral mutants nucleofected into Jurkat T cells and probed with anti-Env antibodies (top) or anti-HIV antisera (bottom). (B) HIV Gag-iGFP fluorescence (more ...)
Prior studies suggest that interactions between Env and Gag enhance the cell-to-cell spread of HIV-1 (13
). We therefore tested viral clones carrying deletion mutations in either the cytoplasmic tail of Env (2
) or the globular head of the MA domain of Gag (54
), both of which attenuate viral replication but are dispensable for viral production. These viral mutants, called HIV Gag-iGFP ΔCT and HIV Gag-iGFP ΔMA, were efficiently nucleofected into donor cells (Fig. , D, and E). The levels of supernatant p24 released from cells nucleofected with the HIV Gag-iGFP ΔCT construct were close to wild-type levels (about 2 ng/ml), while HIV Gag-iGFP ΔMA produced roughly threefold less virus than wild-type HIV Gag-iGFP (data not shown). Surprisingly, deletion of the cytoplasmic domain of Env resulted in only a small reduction in viral transfer (Fig. ), while the MA domain mutation abolished viral transfer completely (Fig. ). The ΔMA mutation illustrates that the viral transfer is not driven simply by an Env-mediated adhesion or fusion event but that it may require proper targeting of Gag to the synapse. VS-mediated transfer therefore is likely to involve coordinated interactions between Gag and Env, yet the presence of the full cytoplasmic tail is not essential for the process.
It has been shown that human intestinal epithelial cells can transcytose HIV-1 in a CD4-independent manner following contact with HIV-infected T cells (4
). We therefore tested whether VS-mediated viral transfer occurred specifically only with CD4+
lymphocytes. Purified primary CD8+
T cells were CMTMR labeled and tested for their ability to capture HIV Gag-iGFP when cocultured with HIV Gag-iGFP-expressing donor cells. CD8+
T cells did not engage in viral transfer with the donor cells (Fig. ). The level of transfer of HIV Gag-iGFP into the CD8 cells was slightly higher than that of the background observed for GFP-expressing cell controls (Fig. ). These data showed that VS-mediated transfer was specific to CD4+
To further examine the role of CD4 in VS-mediated viral transfer, we incubated the donor cells with the HIV-blocking, anti-CD4 antibody Leu 3a. The pretreatment of target cells with Leu 3a at 0.25 μg/ml blocked more than 80% of the VS-mediated transfer (Table ). Similarly, treatment of the cells with monoclonal antibody against the CD4 binding site on gp120, IgG1b12, inhibited 50% of VS-mediated transfer (Table ). Incubation with soluble CD4 blocked more than 60% of the viral transfer (Table ). Thus, three inhibitors of the interaction between Env and CD4 all had inhibitory effects on VS-mediated viral transfer. We conclude that VS-mediated transfer requires the engagement of target cell CD4 by Env on the surface of HIV-expressing donor cells.
Inhibition of VS-mediated transfer
HIV-1 transfer using an R5-tropic Env does not require CCR5 on target cells.
To examine the efficiency of R5-tropic Env in mediating cell-to-cell transfer, we nucleofected Jurkat cells with an HIV Gag-iGFP clone carrying the Env from the R5 molecular clone, JRFL. When the HIV Gag-iGFP(JRFL)-expressing Jurkat cells were mixed with primary CD4+ T cells, a large fraction (31.8%) became highly GFP positive (Fig. , panel ii). This fraction of cells was higher than the fraction of CD4+ T cells that expressed high levels of CCR5 (12.6%) (Fig. , panel iv). The transfer also could be blocked efficiently with the anti-CD4 antibody Leu 3a. This suggests that CCR5 expression is not required for VS-mediated transfer of an R5-tropic virus into primary CD4+ T cells. Viral transfer mediated by the R5 Env also was observed in CCR5-negative MT4 cells (Fig. ) and was not greatly enhanced by the stable expression of CCR5 in these cells (Fig. ). While R5-tropic Env was very efficient at directing cell-to-cell transfer of virus, the presence of the CCR5 coreceptor was not required for VS-mediated viral transfer.
FIG. 5. Viral transfer mediated through R5-tropic Env from molecular clone JRFL occurs efficiently in CCR5+ or CCR5− CD4+ T cells. (A) VS-mediated transfer from HIV Gag-iGFP(JRFL)-expressing Jurkat cells into primary CD4+ T cells (more ...) HIV-1 transfer to CD4+ T cells requires cytoskeletal rearrangements but is not blocked by inhibitors of viral membrane fusion.
HIV-1 VS formation is thought to involve active cytoskeletal rearrangements that stabilize adhesive interactions and promote the concentration of viral and cellular receptors at the VS (28
). We therefore examined the extent to which viral transfer was sensitive to the actin inhibitor cytochalasin D. Cytochalasin D treatment of the donor cells resulted in a 50 to 60% decrease in viral transfer, suggesting that active cytoskeletal rearrangements are necessary for the process (Table ). Another inhibitor of actin polymerization, latrunculin B, reduced VS-mediated viral transfer by 87% (Table ). These results confirm that cytoskeletal rearrangements are required for VS-mediated transfer. Interestingly, the same cytoskeletal inhibitors had little effect on cell-free infection (Table ), showing that the requirements for cell-to-cell transfer are distinct from those of cell-free infection.
To examine the dependence of viral transfer upon viral membrane fusion, we tested the ability of a peptide inhibitor of Env-mediated membrane fusion, T20, to block VS-mediated viral transfer. T20 had no measurable effect on VS-mediated transfer, suggesting that the uptake of virus into target cells does not require triggering of viral membrane fusion (Table ). Similarly, treatment of cells with the broadly neutralizing antibody 2F5 did not block VS-mediated transfer (Table ). Control experiments using these inhibitors at the same or lower concentrations showed that they were effective at blocking cell-free infection (Table ). Since T20 and 2F5 both target fusion intermediates in gp41, we conclude that the highly efficient VS-mediated transfer is not dependent on activation of viral membrane fusion, or that it may occur in a manner that blocks the accessibility of these inhibitors. Because these inhibitors typically are effective at blocking the formation of syncytia, the results also suggest that syncytia do not explain the massive transfer of HIV-1 into target cells. In a previous study, treatment of cells with the CXCR4 antagonist AMD3100 did not block conjugate formation, but it did inhibit CD4 polarization during VS formation (28
). We found that high concentrations of AMD3100 had little effect on viral transfer (Table ). Based on our tests with AMD3100, engagement of CXCR4 by gp120 is not required for highly efficient VS-mediated viral transfer.
Resistance of VS-mediated viral transfer to patient-derived neutralizing antisera.
Viral transmission in the dendritic cell to T-cell-infectious synapses has been shown to mediate efficient infection of T cells by a mechanism that is resistant to neutralizing antibodies (18
). We therefore examined whether the VS-mediated viral transfer in our system is sensitive to antibody-mediated neutralization. We tested two well-studied neutralizing antisera against a control nonneutralizing serum (61
) and found that viral transfer was not affected at a 1:50 dilution, a concentration that effectively blocked infection by homologous HIV-1(NL4-3) under cell-free conditions (Table ). Interestingly, the neutralizing serum did partially block the viral mutant HIV Gag-iGFP ΔCT, reducing the percentage of viral transfer by 40% relative to the level of transfer of the nonimmune control serum (Table ). In recent studies, it has been found that deletions or mutations in the cytoplasmic tail of Env can expose neutralizing epitopes in its ectodomain that are otherwise masked (31
). These results suggest that the patient sera can recognize Env with a truncated cytoplasmic tail on the surface of cells and can block its ability to mediate viral transfer, but they do not recognize native HIV Env on the surface of cells. The cytoplasmic tail therefore may play a role in limiting the exposure of neutralizing epitopes on the cell surface.
Resistance of VS-mediated viral transfer to patient neutralizing antibodies
Dynamic VS imaging reveals enhanced T-cell adhesion, polarity, and chemokinesis.
The rapid and high level of HIV Gag-iGFP fluorescence transferred into target cells measured by flow cytometry suggested that we should be able to visualize the virus as it is being transferred into target cells and to capture dynamic aspects of the living VS. Red-fluorescent-dye-labeled target cells were mixed with HIV Gag-iGFP-expressing donor cells and were imaged with two-color, time-lapse, confocal microscopy. To perform long-term imaging of live samples, we use a sealed, gas-permeable microchamber (Ibidi Biosciences) maintained at 37°C. A fibronectin-coated microchamber provided the T cells with a two-dimensional substrate for attachment and migration. Simultaneous DIC and confocal green and red fluorescence images were acquired. Confocal images were acquired every 10 s for 3 h.
A large majority of infected donor cells was stably associated with one or more target cells (Fig. ; also see movies S1 and S2 in the supplemental material). T cells engaged in synapses assumed a polarized morphology and interacted with target cells through uropod-like structures (Fig. ). A polarized morphology is characteristic of motile T cells (53
) and is distinct from the morphology of T cells engaged in immunological synapses (15
). In the donor cells, higher concentrations of Gag-iGFP were observed at the site of cell-cell contact (Fig. , left). When threshold settings were set to measure low-intensity green signals in target cells, we readily observed accumulations of green fluorescence on the target cells engaged in synapses (Fig. , bottom panels). We also observed spots of green fluorescence on target cells that were not engaged with donor cells, which likely represented cells that already had participated in a VS. These spots also localized asymmetrically in the cell in a location that appeared to be the cell uropod (Fig. ).
FIG. 6. Live imaging of VS formation reveals polarized HIV-infected and uninfected cells engaging in long-lived contacts. (A) An HIV Gag-iGFP-expressing donor cell bound to a target cell through uropod-like structures. (Left) At the VS, increased concentrations (more ...)
Neighboring Jurkat cells not expressing HIV-1 were observed to interact with target cells; however, these interactions usually were relatively short lived. To illustrate the interactions of representative donor cells over the duration of the movie, we converted the temporal sequence of images into a three-dimensional kymograph. In the kymograph, a green donor cell outline (or control cell) is plotted in an x-y plane with the outlines of all target cells with which it interacts during a 40-min imaging period. Time is plotted on a third z plane. HIV-1-infected cells typically stayed engaged with target cells for most of the imaging period (Fig. ), whereas uninfected control cells engaged in multiple short-term interactions with target T cells during the same time frame (Fig. ).
To determine whether the small green spots observed microscopically in the target cells were representative of the same cells that had acquired massive viral fluorescence as measured by flow cytometry, we acquired images of the GFP-positive, CMTMR-positive target cells after they were flow sorted. Confocal images revealed the presence of similar intense green spots localizing to the perimeter of the cells in dots (Fig. ; also see movie S3 in the supplemental material). Most cells appeared to contain more than one spot. Importantly, in flow-sorted target cells, we did not find multicell aggregates or syncytia. By comparing the images of the fixed, flow-sorted target cells to the live images of cells undergoing VS, we confirmed that the target cells that had acquired bright dots of green fluorescence resembled those measured by flow cytometry.
Measurement of Env-dependent adhesive interactions between infected and uninfected T cells.
To quantify the number of cell-to-cell conjugates formed by HIV-infected versus noninfected donor cells, we analyzed the live images on a frame-by-frame basis. In every frame, each infected cell was classified by the number of target cells that it had engaged. An equal number of uninfected, non-GFP-expressing Jurkat cells were analyzed as controls. The frequency of HIV-expressing cells that had formed stable cell conjugates was roughly twofold higher than that of nonexpressing cells (Fig. ). Infected cells therefore were twice as likely to be engaged with one or more target cells. In addition, a large fraction of HIV-expressing donor cells interacted with two or three cells at a time (Fig. ). In contrast, control cells that were not expressing HIV were much less likely to interact with more than one target cell (Fig. ). The average duration of interaction with the target cells also was fourfold greater for the infected cells (Fig. ), although the number of cell contacts initiated was lower than that of uninfected controls (Fig. ). Infected cells engaged fewer target cells than uninfected cells over a period of time, but the interactions with HIV-expressing cells were much more durable.
FIG. 7. Durable VS adhesion is driven by Env. (A to E) HIV Gag-iGFP-expressing donor cells were mixed with purified target CD4+ T cells and imaged for 40 min. (A) The conjugate fraction of control or HIV Gag-iGFP-expressing cells engaged with one, two, (more ...)
We conducted the same experiment with an Env-deficient viral clone, HIV Gag-iGFP ΔEnv, to determine the role of Env in mediating these adhesive contacts. The HIV Gag-iGFP ΔEnv-expressing cells bound to target cells with a lower level of efficiency than uninfected controls (Fig. ). This suggests that the expression of all HIV-1 genes in the absence of Env does not enhance cell-to-cell adhesion (Fig. ). The duration of the typical interaction of the HIV ΔEnv-expressing cells was less than that of neighboring control cells (Fig. ), and the average number of interactions per donor cell was not affected by expression of HIV ΔEnv (Fig. ). These results demonstrate that cell surface Env functions as an adhesion molecule to upregulate the duration of contact between infected and uninfected cells.