In this study, we demonstrated that tetraspanins in virus-producing cells prevent the formation of syncytia, which are multinucleated cells that result from fusion of virus producer cells with one or more target cells. We also documented that this fusion prevention depends on the presence of Gag and the recruitment of Env into TEMs. Further, intrinsically more fusogenic Envs require higher levels of tetraspanin expression for the repression of their fusion activity.
Accumulation of tetraspanins at HIV-1 budding sites is well documented, but their function as budding cofactors remains unclear (7
; Krementsov et al., unpublished data). Because of their well-established role in membrane fusion regulation, here we hypothesized that these proteins regulate HIV-1-induced fusion. In agreement with this hypothesis, we found that overexpression or ablation of individual members of the tetraspanin family in producer cells resulted in fusion repression or enhancement, respectively. This held true when either HeLa cells or more physiologically relevant Jurkat T cells served as virus producers (see Fig. ). We also document that this fusion regulation by tetraspanins is dose dependent, a finding that has bearing on the strain-specific sensitivity differences shown in Fig. (discussed below).
One member of the tetraspanin family, CD82, has previously been shown to inhibit cell-cell fusion mediated by the envelope glycoprotein of another retrovirus (human T-cell leukemia virus type 1) (37
). In that study, fusion mediated by HIV-1 Env (expressed independently from other HIV components) was also tested, but it turned out to be insensitive to the inhibitory action of that tetraspanin. We know now that CD82 can indeed repress HIV-1 Env-mediated membrane fusion (data not shown), but this repression, like the fusion inhibition by CD9 and CD63 (Fig. ), requires the presence of Gag. This dependence on Gag coexpression is likely due to the fact that Env colocalizes (and thus can either directly or indirectly associate with tetraspanins) only if Gag recruits it to TEMs (33
). Since immature Gag also represses the fusogenicity of Env (9
), how can we distinguish between fusion repression by uncleaved Gag and fusion regulation by tetraspanins? Strongly arguing that the two mechanisms are indeed different is the finding that tetraspanin overexpression also leads to repression of fusion mediated by virion-associated Env, i.e., in the context of processed Gag (as discussed below).
The findings reported here in Fig. and are in line with a recent report by Sato et al. (41
), as well as our own findings, which show that tetraspanins in virions inhibit virus-cell fusion, thus suggesting that these tetraspanins repress Env-mediated fusion in virions and in producer cells via a similar mechanism. However, we also note important differences. Sato et al. show that virus-cell fusion of ΔCT Env mutant NL4-3 can be inhibited by CD63 overexpression. In contrast, and as shown in Fig. , cell-cell fusion driven by ΔCT Env is insensitive to CD63. This discrepancy is most likely due to the fact that though ΔCT Env is “nonspecifically” incorporated into particles, once it is acquired by the virions, it is situated in close proximity to tetraspanins and thus can be repressed by them. In contrast, the majority of ΔCT Env at the cell surface is not in proximity to tetraspanins (as documented in Fig. ) and hence is not repressed. We conclude that the Gag-Env interaction is not required if tetraspanins and Env are already proximal (as in the case of ΔCT Env in virions). Why overexpression of CD9 not only fails to repress but even enhances ΔCT Env-driven fusion is unclear at this point, although it is evident that such enhancement does not require colocalization of CD9 and Env. It seems plausible that overexpression of CD9 diverts or sequesters some other cellular factor which would otherwise prevent fusion mediated by tail-less Env. The fact that only CD9 and not CD63 enhances ΔCT Env-driven fusion can serve as starting point for a genetic analysis aiming at the identification of such a hypothetical cellular factor.
Another apparent discrepancy between previously published data (41
) and our own data is displayed in Fig. . There, we show that cell-cell fusion mediated by Env of the R5 virus JRFL is inhibited by overexpression of the tetraspanin CD63. We also found that JRFL infectivity could be partially inhibited by expressing larger amounts of CD63 or CD9 (data not shown), suggesting that this virus is simply less sensitive to tetraspanin-mediated fusion inhibition than NL4-3. Interestingly however, and consistent with the results of Sato et al. (41
), both JRFL and NL4-3 were less sensitive to tetraspanin overexpression when the viruses were produced in 293T cells (in fact, JRFL was not inhibited at all at intermediate tetraspanin concentrations) (data not shown). Presumably, higher levels of virus production (relative to tetraspanins) in 293T cells together with apparently higher intrinsic fusion activity of JRFL Env (Fig. ) explain why the repression of virus-cell fusion was not observed previously.
Finally, we asked whether the fusion-preventive functions of tetraspanins in producer cells (described in this report) were independent of their entry-inhibiting activities in target cells (14
). The data presented in Fig. suggest that this is the case, as neither synergism nor mutual negative interference was observed in these overexpression experiments. These results are also in agreement with the fact that so far tetraspanins have be found to functionally interact with each other only in cis
, i.e., when situated within the same lipid bilayer, and not in trans
. Nevertheless, to formally exclude the latter possibility, we will still need to test whether simultaneous ablation of tetraspanins in producer and target cell shows they same, nonsynergistic outcome.
Did HIV-1 evolve to utilize tetraspanins in producer cells to prevent syncytium formation? To date, it remains unclear whether syncytia are beneficial for virus replication. Some reports answered that question positively by pointing to the facts that, in vitro, these multinucleated entities can shed virions at a high rate (see, e.g., reference 43
), that they can be formed at high levels in cocultures of infected and uninfected lymphocytes, and that they produce infectious virus for a short time prior to their death (39
). However, consistent with the idea that HIV-1 ultimately does not benefit from syncytium formation, syncytia are typically detected (if at all) only in late-stage AIDS, during which they are thought to contribute to the cytopathology of HIV-1. Also in agreement with the idea that syncytia are dispensable for virus spread is the finding that LFA-1-negative peripheral blood mononuclear cells do not form syncytia yet support HIV-1 replication (34
). Nevertheless, and unexpectedly at first sight, ablation of tetraspanins, which leads to increased syncytium formation as shown in this report, does not impede virus transmission in vitro (Krementsov et al., unpublished data). We reason that this is due to the fact that in our transmission assay, as in other currently used in vitro replication assays, producer and target cells are far less densely packed than in lymphoid tissue. Indeed, also compared to the two fusion assays used in this study, syncytia are rarely formed in the transmission assay. Consequently, even a considerable enhancement of syncytium formation, due to tetraspanin ablation, would not be expected to reduce the number of successfully infected target cells to a significant degree. Still, we hold that fusion of producer and target cells, in vitro as well as in vivo, is not likely to be beneficial for HIV-1, and this report now unequivocally shows that one of the functions of tetraspanins at viral exit sites is to prevent syncytium formation. Further, it characterizes some of the requirements for this regulation, thus providing the groundwork for understanding its mechanism.