Cell-to-Cell Transmission of HIV-1 Is Required to Trigger Pyroptotic Death of Lymphoid-Tissue-Derived CD4 T Cells

Most studies examining innate immune recognition of HIV-1 have utilized cell-free particles and characterized responses occurring in dendritic cells or macrophages (Gao et al., 2013, Hayashi et al., 2010, Jakobsen et al., 2013, Lahaye et al., 2013, Manel et al., 2010, Sun et al., 2013, Yan et al., 2010). More recently, attention has focused on resting CD4 T cells in lymphoid tissues, which are mostly non-permissive for productive HIV infection. We previously have shown that the massive death of lymphoid CD4 T cells that are abortively infected with HIV-1 requires close interaction between uninfected target and HIV-producing cells (Doitsh et al., 2010). These findings were consistent with in vitro (Garg et al., 2007, Holm and Gabuzda, 2005) and in vivo (Finkel et al., 1995) studies showing that dying non-productively infected cells in human lymph nodes often cluster near productively infected cells (Finkel et al., 1995). In contrast, we found that cell-free virions accumulating in the supernatants of HIV-infected HLACs, even at high concentrations, were much less efficient at inducing killing of resting target cells by abortive infection. One potential explanation for these differences was that transfer of cell-free particles may not generate sufficient incomplete reverse DNA transcripts to induce a cytopathic response in target CD4 T cells. Cell-to-cell spread increases infection kinetics by two to three orders of magnitude by directing virus assembly and obviating the rate-limiting step of extracellular diffusion required for cell-free virus to find and engage a susceptible target cell (Jolly, 2011, Martin and Sattentau, 2009, Sato et al., 1992, Sourisseau et al., 2007).

To test this hypothesis, we used spinoculation to emulate efficient cell-to-cell spread of virus (Geng et al., 2014). Spinoculation accelerates the binding of cell-free virions to target cells, facilitates synchronized delivery of a large number of particles into the cells (O’Doherty et al., 2000, Saphire et al., 2002), and enhances accumulation of cytoplasmic reverse DNA transcripts (Figure 1A) (O’Doherty et al., 2000, Pace et al., 2012). As expected, spinoculation of HLACs with free HIV-1 promoted high levels of HIV-1 fusion into target lymphoid CD4 T cells (Figure S1). Spinoculation also caused extensive and selective depletion of target CD4 T cells (Figure 1B). The relative proportion of CD8 T cells was unaltered. CD3⁺/CD8⁻ T cells were similarly depleted, indicating that cell loss was not an artifact of downregulated surface expression of CD4 following direct infection (not shown). Consistent with our previous reports (Doitsh et al., 2010,Doitsh et al., 2014, Monroe et al., 2014), loss of CD4 T cells was prevented by addition of efavirenz, an NNRTI that allosterically inhibits HIV-1 reverse transcriptase, and by AMD3100, an entry inhibitor that blocks gp120 engagement of the CXCR4 coreceptor. However, unexpectedly and not in keeping with our previous reports, addition of raltegravir, an integrase inhibitor also blocked CD4 T cell death (Figures 1B and S5). Because cell death involves viral life cycle events occurring prior to viral integration, raltegravir should act too late to affect the abortive infection process that triggers the pyroptotic response.

To further investigate this surprising result, carboxyfluorescein succinimidyl ester (CFSE)-labeled target CD4 T cells were co-cultured with productively infected HLACs, and raltegravir was added at the time of mixing of productively infected and target CD4 T cells (Figure 1C). Under these conditions, raltegravir had no effect on target CD4 T cell death, while efavirenz and AMD3100 blocked the response (Figure 1D).

Based on these contrasting effects of raltegravir, we hypothesized that in the co-culture experiments involving mixing of HIV-infected HLACs with target cells, raltegravir had no effect on the ensuing death of target CD4 T cells that became abortively infected, because the culture already contained productively infected cells. Conversely, in the spinoculation experiments, raltegravir blocked cell death, because it prevented the establishment of a productively infected subset of cells needed for cell-to-cell spread of the virus to target CD4 T cells. To test this hypothesis, we spinoculated HLACs with either single-round or multiple-round viruses containing a GFP reporter (NLENG1I) (Levy et al., 2004). These viruses permit the dynamics of HIV-1 infection and T cell depletion to be simultaneously monitored in the spinoculated cultures. Four days after spinoculation, we observed a similar number of GFP-positive, productively infected cells with both the single-round and multiple-round viruses, indicating that viral spread was not required to establish an initial population of productively infected cells. However, we observed a massive loss of CD4 T cells only in cultures spinoculated with the multiple-round virus. Notably, spinoculation with an integrase-deficient GFP HIV-1 (NLENG1I-D116N) (Gelderblom et al., 2008) resulted in no productive infection and no CD4 T cell death (Figure 2A). These results suggested that viral spread from productively infected cells, but not spinoculation of cell-free virions, is promoting the death of non-permissive lymphoid CD4 T cells. In agreement with this conclusion, addition of the AMD3100 entry inhibitor 4 hr after spinoculation efficiently blocked the ensuing death response while not affecting the number of GFP-positive productively infected cells (Figure 2B). Moreover, treatment with the viral protease inhibitor saquinavir, which acts during the budding stage of HIV-1 replication, did not inhibit productive infection but prevented CD4 T cell death by newly released HIV-1 virions (Figure S2). These findings further indicated that CD4 T cell death occurs after establishment of productive infection, but not during infection with cell-free viruses.

Single-round and integrase-deficient HIV-1 clones are not competent for cell-to-cell dissemination following spinoculation with HLACs. To confirm that the mode of viral transfer influenced the death response of target CD4 T cells, we modified the infection system by overlaying HLACs on a monolayer of 293T cells that had been transfected with these single-round proviral clones (Figure 2C). Interestingly, when these single-round viruses were transferred to HLACs by direct interaction with virus-producing 293T cells, a massive killing of target lymphoid CD4 T cells was observed (Figure 2D). These results demonstrate that recapitulating the cell-to-cell mode of viral transfer is sufficient to restore the killing capacity of these single-round clones.

To further explore whether cell-cell contact was needed to induce death of CD4 T cells, we repeated the co-culture assay using productively infected and target CFSE-labeled HLACs. However, in this experiment, the cells were co-cultured under conditions of increasing surface area, thereby reducing the likelihood of cell-cell interactions. Using flow cytometry, we analyzed the levels of viable target CD4 T cells in the plates every 24 hr during 4 days of co-culture. The death of target CD4 T cells decreased as the surface area of the culture increased (Figure 3A), even in samples where the volume of culture medium remained constant (Figure 3B). These data suggest that the physical distance between HIV-producing and target cells directly affects the kinetics of CD4 T cell depletion and argue further against a role for free virions released into the medium in the death response.

Cell-to-cell spread of HIV-1 predominantly takes place across specialized contact-induced structures known as virological synapses (Agosto et al., 2015,Jolly et al., 2004, Jolly et al., 2007, Jolly and Sattentau, 2004). These synapses facilitate efficient transmission of virus toward the uninfected and engaged target cell. The synapse gains stability through actin-mediated recruitment of adhesion molecules, such as the integrin leukocyte function-association antigen 1 (LFA-1) and its cognate ligand ICAM-1 to the junction point of cellular interaction (Jolly et al., 2007). To examine whether virological synapse formation between HIV-infected and target cells is required to promote CD4 T cell death, productively infected and target CFSE-labeled HLACs were co-cultured in the presence of blocking antibodies against ICAM-1 or CD11a, the α-subunit of the LFA-1 heterodimer. Addition of either the anti-ICAM-1 (Figure 3C) or anti-CD11a (Figure 3D) antibodies, but not isotype-matched control antibodies, effectively blocked depletion of target CD4 T cells in the mixed cultures as efficiently as the antiviral drug efavirenz. These findings suggest that the death response involves adhesion molecules that are required for virological synapse formation, indicative of a requirement for close cell-cell contact in mediating pyroptotic cell death. Because cell-free virions also express LFA-1 and ICAM-1 (Bounou et al., 2002,Fortin et al., 1997), we cannot completely rule out an additional effect on HIV virions. However, when combined with all of the data (Figure S2), we conclude that cell-to-cell transmission is critical for the induction of pyroptosis.

Western blotting analysis of HLAC revealed high expression levels of ICAM-1 in B cells, but not in CD4 or CD8 T lymphocytes. However, activated CD4 T cells, which correspond to those that become productively infected with HIV-1, express high levels of this adhesion molecule (Figure 3E). In contrast to ICAM-1, CD11a expression levels were high in both resting and activated CD4 T cells (Figure 3F). Thus, synapse formation between activated CD4 T cells expressing ICAM-1 and target CD4 T cells (either activated or resting) expressing LFA-1 may occur regularly in lymphoid tissues, independently of viral infection.

Most CD4 T cells in lymphoid tissues infected with HIV die by caspase-1-mediated pyroptosis triggered by abortive viral infection (Doitsh et al., 2014). To test whether caspase-1 is induced by cell-free HIV-1 particles or by cell-to-cell spread of HIV-1, we spinoculated HLACs with single-round or multiple-round clones of a DsRedExpress reporter virus (NLRX-IRES) (Gelderblom et al., 2008) and analyzed intracellular caspase-1 activity using cell-permeable fluorogenic caspase-1-specific substrates (CaspaLux1) (Komoriya et al., 2000, Packard and Komoriya, 2008). Consistent with our previous reports (Doitsh et al., 2014,Monroe et al., 2014), spinoculation with multiple-round HIV-1 particles triggered high levels of intracellular caspase-1 activity in target CD4 T cells. In contrast, spinoculation with single-round or integrase-deficient HIV-1 particles produced only background levels of caspase-1 activity (Figure 4A). Inhibition of cell-to-cell spread using the viral protease inhibitor saquinavir, or by treatment with AMD3100 4 hr after spinoculation, also markedly inhibited caspase-1 activation induced by multiple-round HIV-1 particles.

Consistent with pyroptosis as the pathway of programmed cell death, spinoculation with multiple-round HIV particles resulted in the release of the intracellular enzyme lactate dehydrogenase (LDH) (Fink and Cookson, 2005) (Figure 4B). Further, the release of LDH was completely blocked when AMD3100 was added 4 hr after spinoculation or when single-round or integrase-deficient HIV-1 particles were used for initial infection. Together, these findings indicate that infection with cell-free HIV-1 particles does not lead to caspase-1 activation, despite apparent abortive infection of lymphoid CD4 T cells. Rather, capsase-1 activation and the induction of pyroptosis require the generation of productively infected cells and successful cell-to-cell spread of HIV-1 to quiescent bystander lymphoid CD4 T cells.