Cytoplasmic Body Component TRIM5α Requires Lipid-enriched Microdomains for Efficient HIV-1 Restriction*

TRIM5α is a member of the tripartite motif (TRIM) family of proteins and affects both early and late phases of the retroviral life cycle. Although TRIM5α multimerizes to form cytoplasmic bodies, which are thought to play an important role in viral restriction, the identity of TRIM5α-containing cytoplasmic bodies remains elusive. To better understand TRIM5α cytoplasmic body constituents and the cellular proteins that could be involved in the TRIM5α-mediated antiviral activities, we sought TRIM5α-binding factors. We identified a lipid microdomain protein flotillin-1/Reggie-2 as an interacting partner of TRIM5α via co-immunoprecipitation. Immunohistochemistry studies confirmed the co-localization of rhesus monkey TRIM5α (TRIM5αrh) cytoplasmic bodies with flotillin-1/Reggie-2. Caveolin-1, another lipid microdomain-associated protein, also co-localized with TRIM5α cytoplasmic bodies. Intriguingly, disruption of cellular cholesterol by cyclodextrin perturbed TRIM5α cytoplasmic body formation. Furthermore, lipid starvation partially relieved the endogenous post-entry restriction of HIV-1 infection, which could be subsequently restored by lipid repletion. These observations indicate the involvement of cellular lipids in TRIM5α-mediated antiviral activities. Given that many viruses utilize cellular lipid microdomains for viral entry and assembly, it is plausible that lipid-enriched domains provide microenvironments where TRIM5α recognizes retroviral components.

The cytoplasmic body component TRIM5␣ is a key contributor of the host cell defense against a diverse family of retroviruses. TRIM5␣ includes the RING, B-box 2, coiled-coil, and B30.2(PRYSPRY) domains and blocks retroviral infection in a species-specific manner (1)(2)(3)(4). The sequences of the C-terminal B30.2(PRYSPRY) domain dictate the potency and specificity of the post-entry restriction (5)(6)(7)(8). Rhesus monkey TRIM5␣ (TRIM5␣rh) recognizes the mature core of incoming HIV-1 at a post-entry, pre-integration stage in the viral life cycle, before significant reverse transcription (3,4,9). Polyubiquitylation and rapid degradation of TRIM5␣ are dependent on the intact RING and B-box 2 domains (10). Although initial reports have indicated that TRIM5␣-mediated post-entry restriction occurs independently of the ubiquitin/proteosome system (11,12), more recent studies have suggested partial involvement of the proteosome system during the post-entry restriction (13,14). In addition to this well characterized antiviral activity, TRIM5␣rh has another antiviral activity to block the late phase of the HIV-1 life cycle (15)(16)(17). Although the late restriction activity of endogenous TRIM5␣rh remains controversial, transient expression of TRIM5␣rh potently blocks HIV-1 production in 293T cells by reducing the levels of HIV-1 Gag polyproteins in producer cells (16,18). In contrast to the post-entry restriction, this late restriction is dependent on the N-terminal RING, B-box 2, and coiled-coil domains rather than the B30.2(PRYSPRY) of TRIM5␣rh (16).
Both TRIM5␣rh and human TRIM5␣ (TRIM5␣hu) proteins typically display a diffuse cytoplasmic distribution marked by discrete cytoplasmic bodies and have the tendency to form larger cytoplasmic aggregates when overexpressed (4,12,19). Disruptions in the coiled-coil and adjacent linker 2 region, located between the coiled-coil and B30.2(PRYSPRY) domains of TRIM5␣rh, result in diffuse TRIM5␣rh distribution devoid of cytoplasmic bodies (11,17,20,21). Although cytidine deaminase enzyme APOBEC3G also forms cytoplasmic bodies, which co-localize with processing bodies (p-bodies) (22), TRIM5␣ cytoplasmic bodies are considered separate entities (23). Cytoplasmic bodies are also independent of actin, early endosomes, trans-Golgi, lysosomes, or aggresomes (23). Initial studies suggest that TRIM5␣ cytoplasmic body formation may not be required for HIV-1 post-entry restriction (5,12). However, the dynamic interaction between TRIM5␣rh and cytoplasmic HIV-1 preintegration complexes, including the rapid de novo formation of TRIM5␣rh cytoplasmic body-like structures around viral complexes, indicates a role for TRIM5␣rh cytoplasmic bodies in the TRIM5␣-mediated HIV-1 restriction (13,21). Moreover, the observation that higher order self-association through the TRIM5␣ B-box 2 domain increases TRIM5␣ avidity against the incoming retroviral core implies the importance of self-associated TRIM5␣rh protein in the post-entry restriction activity (24). However, the identity and constituents of TRIM5␣ cytoplasmic bodies remain elusive.
Here, to better understand the identity of TRIM5␣ cytoplasmic bodies, we sought cellular binding partners of TRIM5␣rh. We found that lipid microdomain protein flotillin-1/Reggie-2 co-immunoprecipitated with TRIM5␣rh. Immunostaining revealed the co-localization of the TRIM5␣rh cytoplasmic bodies with flotillin-1/Reggie-2, as well as caveolin-1, another lipid microdomain-associated protein. Flotillin-1/Reggie-2 knock-down did not have remarkable effects on TRIM5␣rh-mediated cellular restriction; however, depletion of cellular lipids and cholesterol disrupted TRIM5␣ cytoplasmic body formation and impaired the TRIM5␣rh-mediated post-entry restriction. Subsequent lipid repletion completely restored the TRIM5␣rhmediated post-entry restriction of HIV-1. Similar effects were observed in the late restriction upon lipid depletion. Thus, our results demonstrate that target cell lipids play a critical role in the cytoplasmic body formation and antiviral activities of TRIM5␣.
Immunoprecipitation-293T cells were co-transfected with 4.0 g of pRhT5␣ and 0.5 g of pNL4-3 in 15-cm culture plates. 40 h post-transfection, cells were treated with 40 M MG115 for 6 h (Sigma). Producer cells were washed with ice-cold PBS, and cell lysates were harvested in 1.0 ml of RIPA lysis buffer. Cell debris was removed by centrifugation, and the supernatant was kept for analysis. Immunoprecipitation of HA-tagged TRIM5␣ proteins was performed using agarose beads conjugated with mouse anti-HA antibody (Sigma) and were incubated at 4°C overnight with rotation. The beads were washed at least 20 times with RIPA buffer and pelleted. Immunoprecipitated proteins were eluted off the HA-agarose beads through a 30-min room temperature incubation with 100 g of HA peptide (Sigma), according to the manufacturer's instructions. Immunoprecipitation of flotillin-1/Reggie-2 was performed using monoclonal antibody against human flotillin-1/Reggie-2 protein (BD Transduction Laboratories). To remove any nonspecific binding of proteins, the antibody against flotillin-1/Reggie-2 was added to the supernatant at a 1:250 concentration after a 30 min rotation with protein G beads (Sigma) and rotated overnight at 4°C. After more than 20 washes with RIPA buffer, the protein-antibody complex was pulled down using protein G beads. Laemmli sample buffer with ␤-mercaptoethanol was directly added to the beads, and protein samples were heat-denatured for immunoblot analysis.

RESULTS
Lipid Microdomain-associated Protein Flotillin-1/Reggie-2 Coimmunoprecipitates with TRIM5␣rh-To gain an insight into the identity of TRIM5␣ cytoplasmic bodies where TRIM5␣rh appears to interact with incoming HIV-1 capsid (13, 23), we examined cellular factors that may associate with TRIM5␣. 293T cells were transfected with a plasmid expressing a C-terminally HA-tagged TRIM5␣ with or without pNL4-3. Cell lysates were immunoprecipitated with anti-HA antibody as described under "Experimental Procedures." Eluted proteins were then separated in a 4 -15% Tris-HCl gradient gel and silver-stained for protein visualization. A representative silver stain is shown (Fig. 1A). Prominent bands and the corresponding areas in the control lane were excised and analyzed using tandem MS. Bands B, F, and H appeared more prominently following longer development times (data not shown). Peptide assignments from the tandem MS spectra were accepted if they could be established at greater than 99.9% probability as specified by the PeptideProphet algorithm. Keratin was excluded as it is known as a common contaminant in proteomic analysis. The identified proteins are listed in Fig. 1B. Of the several proteins identified, flotillin-1/Reggie-2, ␣-tubulin, and heat shock protein 70 (HSP70) were further analyzed as their peptides were identified at least nine times (Fig. 1B). HIV-1 p24 was not detected upon TRIM5␣rh immunoprecipitation (data not shown). To verify the HA immunoprecipitation results, we used an anti-flotillin-1/Reggie-2 monoclonal antibody to immunoprecipitate endogenous flotillin-1/Reggie-2 in TRIM5␣rhand/or NL4-3-expressing 293T cells. Although endogenous flotillin-1/Reggie-2 was below detectable levels via immunoblot analysis, flotillin-1/ Reggie-2 immunoprecipitation enriched flotillin-1/Reggie-2 signals (Fig. 1C). Furthermore, we were able to detect co-immunoprecipitated TRIM5␣rh-HA upon flotillin-1/Reggie-2 pulldown (Fig. 1C). We could not confirm the bindings of TRIM5␣rh with ␣-tubulin or HSP70 by using antibodies against respective proteins (data not shown).
To rule out the possibility that the C-terminal mCherry tag affects the TRIM5␣rh subcellular localization, FRhK4T5␣HA cells, stably expressing C-terminally HA-tagged TRIM5␣rh, were also generated. FRhK4T5␣HA cells were immunostained with anti-HA and anti-flotillin-1 primary antibodies and visualized using secondary antibodies conjugated with Texas Red and FITC, respectively. Both TRIM5␣rh and flotillin-1/Reggie-2 formed cytoplasmic bodies, and the co-localization coef-ficients between the signals were nearly identical to those observed in the FRhK4T5␣Ch cells (Fig. 2C, co-localization coefficient ϭ 0.598).
Next, to assess the influence of flotillin-1/Reggie-2 knockdown on the TRIM5␣-mediated late restriction, the shRNAcarrying FRhK4 cells were transfected with an HIV-1 infectious molecular clone, pNL4-3, and HIV-1 production was monitored. No remarkable difference in TRIM5␣rh-mediated restriction of HIV-1 production was observed upon flotillin-1/ Reggie-2 knockdown (Fig. 3C). We also tested the effects of flotillin-1/Reggie-2 knockdown on the late restriction in TRIM5␣rh-overexpressing 293T cells. 293T cells were co-transfected with the flotillin-1/Reggie-2-specific shRNAexpressing lentiviral vector plasmids along with the HIV-1 A, 293T cells were transfected with an HA-tagged TRIM5␣rh-expressing plasmid with or without proviral HIV-1 plasmid pNL4-3. TRIM5␣rh was immunoprecipitated using anti-HA-agarose beads. Proteins were eluted off the beads as described under "Experimental Procedures" and were subjected to SDS-PAGE. Silver-stained bands that were excised and analyzed by MS/MS analysis are designated by the letters A-H. B, prominent peptide identifications from MS/MS analysis. Identifications were accepted if they could be established at greater than 99.9% probability, as specified by the Peptide Prophet algorithm, and contained at least nine identified peptides. Letters correspond to the band identified in the silver-stained gel in A. No HIV-1 Gag was detected in the samples. C, endogenous flotillin-1/Reggie-2 was immunoprecipitated from 293T cells expressing TRIM5␣rh (T5␣rh) with or without NL4-3. Anti-HA antibody was used to detect HA-tagged TRIM5␣rh, although anti-flotillin-1 antibody was used to detect flotillin-1/Reggie-2. Arrows indicate appropriate protein size. NOVEMBER 5, 2010 • VOLUME 285 • NUMBER 45

Lipid Starvation Increases HIV-1 Vector Infectivity in FRhK4 Cells-
The disruption of TRIM5␣ cytoplasmic bodies through cholesterol depletion suggested that cellular lipids may play an important role in the TRIM5␣-mediated restrictions. To address the possible role in the restriction, we examined the susceptibility of target cells upon serum starvation and depletion conditions.
Because VSV-Gpseudotyped viruses are largely unaffected by lipid depletion (40 -42), we challenged target cells with VSV-G-pseudotyped HIV-1 vectors. After incubation in 5% LPDSsupplemented media for 48 h, FRhK4 cells were treated with mock (LPDS) or 10 mM 2OHp␤CD (LPDS ϩ CD) for 45 min, followed by GFP-expressing HIV-1 vector infection (Fig. 5A). Intriguingly, LPDS and LPDS ϩ CD-treated FRhK4 cells showed increased permissivity to HIV-1 by 10-fold when compared with control (Fig. 5A). When LPDS ϩ CD-treated cells were lipid-replenished using 10% FBS media (repletion), HIV-1 infectivity was reduced to that of control FRhK4 cells (Fig.  5A). Target cell cytotoxicity did not appear to be an issue as LPDS or 2OHp␤CD treatment of FRhK4 cells did not affect the infectivity of a permissive SIV MAC -based vector (Fig.  5A). Similar results were obtained when LPDS ϩ CD-treated cells were lipid-repleted with excess LDL or water-soluble cholesterol (data not shown).
Identical experimental conditions were applied to TE671 cells, in which infection of N-MLV, but not NB-MLV, is strongly blocked by human TRIM5␣ at the post-entry stage. LPDS-and LPDS ϩ CD-treated cells showed a 60-fold increase in N-MLV titers, although lipid repletion partially reduced N-MLV infectivity to that of control TE671 cells (Fig. 5B). In contrast, NB-MLV infectivity was not strongly affected by the lipid starvation/depletion and the following lipid repletion. These data suggest that target cell lipid status plays a role in the cellular restriction against incoming HIV-1 and N-MLV.
Lipid Modulation Affects TRIM5␣rh-mediated Post-entry Restriction of HIV-1-Target cell lipid starvation and cholesterol depletion rendered FRhK4 and TE671 cells permissive to HIV-1-or N-MLV-based vectors. To address the involvement of TRIM5␣ in this observation, we generated stable TRIM5␣ knockdown cell lines and examined the effects of lipid modula- tion on their permissivity to viral infection. FRhK4 cells were transduced with a lentiviral vector carrying the shRNA sequence H4, which targets the B30.2(PRYSPRY) domain of TRIM5␣rh. From multiple puromycin-resistant clones, we selected one control LKO.1 clone (LKO.1) and three LKO.1-H4 clones (H4.2, H4.4, and H4.6), which showed similar growth properties to parental FRhK4 cells, for further characterization. The H4 clones showed substantially reduced levels of TRIM5␣rh-specific mRNA upon RT-PCR analysis (Fig. 6A). Band intensities of the H4 clones were similar to that of the 1/32-diluted control; therefore, we estimate H4 clones expressed ϳ3% of the normal level of TRIM5␣ mRNA (Fig. 6, A  and B). TRIM5␣ knockdown clones showed up to a 50-fold increase in permissivity to HIV-1 than the parental FRhK4 cells or the control LKO.1 clone, although TRIM5␣ knockdown showed little effect on the unrestricted SIV MAC vector infectivity in FRhK4 cells (Fig. 6C).
Using the TRIM5␣ knockdown clone H4.2, we assessed the effects of lipid depletion on HIV-1 restriction status. Lipid depletion experiments were performed as described earlier. Control LKO.1 clone behaved similarly to parental FRhK4 cells   under lipid starvation conditions (Figs. 5A and 6D), where LPDS and LPDS ϩ CD treatment relieved the viral restriction by 10-fold, and lipid repletion restored viral infectivity to that of controls (Fig. 6D). In contrast, the H4.2 clone did not show little differences in permissivity against HIV-1 upon lipid depletion (Fig. 6D). Two other TRIM5␣ knockdown clones also showed similar results (data not shown). These observations indicate that modulation of cellular lipids affect HIV-1 infectivity only in the presence of TRIM5␣rh.
Lipid Modulation Does Not Affect TRIM5␣rh Expression-To rule out the possibility that lipid modulation affects TRIM5␣ expression in the target cells, we examined the levels of HA-tagged TRIM5␣rh expression upon lipid modulation using FRhK4T5␣HA cells. Under conditions where TRIM5␣mediated restriction activities were disrupted, TRIM5␣rh pro-tein expression remained largely unchanged (Fig. 7A). When mRNA levels of endogenous TRIM5␣ were examined using FRhK4 cells under these same conditions, lipid depletion showed no notable effects on TRIM5␣ mRNA levels (Fig. 7B). These data suggest that lipid depletion conditions do not affect TRIM5␣ expression. We therefore concluded that cellular lipids support the TRIM5␣rh-mediated antiviral activity against HIV-1 infection.
Depletion of Cellular Cholesterol Abrogates the TRIM5␣rhmediated Late Restriction in 293T Cells-We next examined whether depletion of cellular cholesterol affects the TRIM5␣rh-mediated late restriction activity in FRhK4 and 293T cells. When the influence of cholesterol depletion was examined in FRhK4 cells, 2OHp␤CD treatment increased HIV-1 production up to 3-fold (Fig. 8A). Moreover, when HIV-1 NL4-3 virions were generated in 293T cells transiently expressing TRIM5␣rh, 2OHp␤CD treatment increased HIV-1 titers by up to 20-fold (Fig. 8B). In the absence of TRIM5␣rh, 2OHp␤CD-mediated cholesterol depletion did not affect HIV-1 production in 293T cells (Fig. 8B). When SIV MAC , which is resistant to TRIM5␣rh-mediated late restriction (16), was produced in the presence or absence of TRIM5␣rh, 2OHp␤CD showed no notable effects (Fig. 8B). Similar to the effects of 2OHp␤CD treatment on the protein expression of FRhK4T5␣HA cells (Fig. 7A), treatment of 293T cells transiently expressing TRIM5␣rh with 10 mM 2OHp␤CD did not show any remarkable differences in protein expression (Fig.  8C). Similar to our observations during the post-entry restriction, these observations indicate that efficient TRIM5␣rh-mediated late restriction requires cellular cholesterol.  Protein content in each well was normalized via Bradford colorimetric assay prior to SDS-PAGE. Anti-HA antibodies were used to detect TRIM5␣rhHA, and ␤-actin was used as a control. B, total RNA was harvested from FRhK4 WT cells using TRIzol solution. Superscript III kit was used for first strand cDNA synthesis, and PCR was performed using a primer set for TRIM5␣. ␣-Tubulin was used as a control.

DISCUSSION
HIV-1 restriction factor TRIM5␣ resides in host cell cytoplasmic bodies, where TRIM5␣ is thought to interact with viral components during viral infection (13,23). TRIM5␣ also forms cytoplasmic bodies during the late restriction (16,17); however, the details of TRIM5␣ cytoplasmic body constituents, as well as the ensuing mechanisms of TRIM5␣-mediated restriction, remain poorly defined. Here, we demonstrated that TRIM5␣rh binds to cellular lipid-associated protein flotillin-1/Reggie-2 and co-localizes as cytoplasmic bodies with flotillin-1/Reggie-2 and caveolin-1. Target cell lipid starvation and depletion altered TRIM5␣rh cytoplasmic body formation and impaired the endogenous post-entry restriction activity against HIV-1 in FRhK4 cells. Lipid modulation also impaired the TRIM5␣rhmediated late restriction activities against HIV-1. Our data therefore demonstrated the importance of cellular lipids in TRIM5␣ cytoplasmic body formation and efficient TRIM5␣mediated antiviral activities.
Although flotillin-1/Reggie-2 unlikely plays a role in TRIM5␣ restrictions, our results strongly suggest that TRIM5␣rh cytoplasmic bodies are lipid-enriched. Lipid-enriched cellular microdomains are known to be multifunctional. They are involved in the transport of intracellular lipids (46), the converging of proteosomal and autophagic degradation pathways (50 -52), and the removal and storage of excess proteins from cellular compartments (53). For example, aberrantor overexpression of caveolin-1, -2, or -3, which are involved in many functions ranging from endocytosis to signal transduction, leads to the sequestering of these proteins to these lipid bodies (34,54,55). This raises the concern over the TRIM5␣rh localization in lipid bodies as an artifact due to stably expressed TRIM5␣. In this regard, the recent article by Kim et al. (56) is notable. The authors used a biochemical analysis and revealed the presence of endogenous human TRIM5␣ in lipid-enriched microdomains from human fibroblast cell lysates (56). These data rule out the possible artifacts due to sequestering of an overexpressed TRIM5␣ protein in lipid microdomains. In the course of this study, Hwang et al. (57) reported that TRIM5␣rh was present in detergent-insoluble fractions when expressed in 293T cells, which further support our current findings.
Our immunoprecipitation and tandem MS analysis also identified ␣-tubulin as a potential TRIM5␣rh binding partner (Fig. 1, A and B). However, no prominent co-localization between TRIM5␣rh and ␣-tubulin was observed (Fig. 2F). Because ␣-tubulin is found to be tightly associated with flotillin-1/Reggie-2 in membrane raft domains from rat brain homogenates (58), ␣-tubulin may have been pulled down along with flotillin-1/Reggie-2. The identification of HSP70 (Fig. 1B), one of the best characterized molecular chaperone proteins (59), as a potential binding partner was intriguing, because HSP70 is specifically incorporated into primate lentiviral virions (60). Although we were unable to see prominent co-localization of HSP70 with TRIM5␣rh cytoplasmic bodies in FRhK4 cells (Fig.  2G), a previous study identified partial HSP70 co-localization with TRIM5␣rh at distinct cytoplasmic bodies in HeLa cells (61). Because HSP70 can be translocated to lipid bodies upon stimulation (62), the discrepancy between our data and the reported HSP70 co-localization data may be partially explained by the differences in cell types and/or the dynamicity of subcellular lipid compartments.
Lipids and lipid-associated proteins are reported to be involved in the initial stages of the HIV-1 life cycle. For instance, both primary HIV-1 receptor CD4 (40,63) and HIV-1 co-receptor CCR5 (64) reside on host cell membrane raft domains. Del Real et al. (65) reported that initial HIV-1 entry into host cells is crippled when the primary CD4 receptor is translocated to non-raft domains on the plasma membrane. To avoid the possible influence of lipid depletion on HIV-1 and MLV entry, we here used VSV-G pseudotyped viruses, which are largely unaffected by lipid depletion (40 -42). Lipid starvation resulted in reduced post-entry restriction against HIV-1 (Fig. 5A) and N-MLV (Fig. 5B), which could be restored by subsequent lipid repletion (Fig. 5, A and B). Lipid starvation/ depletion conditions did not affect HIV-1 permissivity in TRIM5␣rh-knockdown FRhK4 cells, which indicate the increase in permissivity observed upon lipid modulation is dependent on TRIM5␣ expression. Because lipid modulation affected TRIM5␣ cytoplasmic body formation without affecting TRIM5␣ expression (Figs. 4, 7A, and 8C), it is plausible that lipid depletion may disrupt TRIM5␣ compartmentalization with viral components in lipid microdomains, thereby reducing the availability of TRIM5␣ during the restrictions. Given that the proteosomal degradation pathway is involved in the initial TRIM5␣-mediated post-entry restriction step (13,14), these events may occur in lipid microdomains, leaving the possibility that lipid depletion disrupts the ensuing proteosomal degradation pathways in the TRIM5␣-mediated restriction. Further studies of the target-cell lipid biology would likely elucidate the details of cellular lipid involvement in the TRIM5␣-mediated restrictions.
The association of TRIM5␣ with subcellular lipid microdomains may not be surprising, as several cellular restriction activities against lentiviruses occur at lipid-enriched microdomains (66 -69). Involvement of cellular lipids in the post-entry restriction of HIV has been reported as Lv2 activity, where a yet-to-be-identified restriction factor determines the cellular tropism of two related HIV type 2 (HIV-2) isolates (66 -68). Lv2 restriction is determined by the viral capsid and envelope protein sequences, and Lv2 lentiviral restriction activity is determined by the route of viral entry; the restriction is only prominent after the restriction-sensitive viral core reached a subcellular compartment where Lv2 activity is accessible (67,68). Although Lv2 restriction appears to be independent of human TRIM5␣-mediated post-entry restriction (66), these observations indicate the involvement of cellular lipids in the post-entry restriction of lentiviruses.
Cellular lipids are also required for HIV-1 Vpu to counteract CD317/BST-2/tetherin (69). Previous reports suggest that these lipid microdomains can act as scaffolds or platforms for protein-protein interactions and protein compartmentalization (70,71). Similar to our observations with VSV-G pseudotyped vectors, several reports suggest that endocytic routes of viral entry into target cells are dependent on cellular cholesterol but are independent of flotillin-1/Reggie-2, caveolin-1, or clathrin (72,73). Thus, lipid microdomains may provide subcellular microenvironments where antiviral factors, such as TRIM5␣, Lv2, and CD317/BST-2/tetherin, can recognize and/or sense viral components. Elucidating the roles of cellular lipids and lipid-associated proteins in TRIM5␣-mediated antiviral activities may result in a better understanding of the underlying mechanisms of TRIM5␣ restrictions, ultimately leading to novel antiviral strategies.