MicroRNA-155 Reinforces HIV Latency*

Background: Curing HIV will require suppression of viral replication and clearance of the transcriptionally silent latent reservoir. Results: We have discovered a novel cellular pathway involving a latency-promoting host miRNA and HIV-activating TRIM protein. Conclusion: miR-155 inhibits the HIV-activating effects of TRIM32 and thus may promote a return to latency in reservoir cells transiently expressing HIV. Significance: Preventing reversion to latency could provide a novel approach for eliminating the latent reservoir. The presence of a small number of infected but transcriptionally dormant cells currently thwarts a cure for the more than 35 million individuals infected with HIV. Reactivation of these latently infected cells may result in three fates: 1) cell death due to a viral cytopathic effect, 2) cell death due to immune clearance, or 3) a retreat into latency. Uncovering the dynamics of HIV gene expression and silencing in the latent reservoir will be crucial for developing an HIV-1 cure. Here we identify and characterize an intracellular circuit involving TRIM32, an HIV activator, and miR-155, a microRNA that may promote a return to latency in these transiently activated reservoir cells. Notably, we demonstrate that TRIM32, an E3 ubiquitin ligase, promotes reactivation from latency by directly modifying IκBα, leading to a novel mechanism of NF-κB induction not involving IκB kinase activation.

With the advent of combination ART, 2 hopes were high that these new and potent agents might cure HIV-infected subjects (1). However, despite consistently reducing viral loads to undetectable levels, a cure was not achieved (2) because of the persistence of latent but integrated HIV proviruses in a drug-insensitive reservoir formed at least in part by memory CD4 T cells. Considerable effort is now focused on devising ways to attack this latent reservoir. One approach termed "shock and kill" involves the identification of a mixture of inducing agents that would be used to activate the latent virus under the cover of ART. These agents need to be strong enough to induce the virus throughout the reservoir but not so strong that a toxic cytokine storm is induced. Ideally, these agents will induce the death of the reservoir cells by a viral cytopathic effect or immune clearance.
Although the latent reservoir appears quite stable over time, it is not static. Intermittently, latently infected cells spontaneously activate and produce small amounts of virus. These viruses contribute to the low viral load observed in subjects receiving ART. Whether the transiently activated cells return to a latent state remains an intriguing but unanswered question.
The elimination of HIV latency is in its simplest terms a transcriptional problem. HIV transcription is controlled by the long terminal repeat (LTR), which contains multiple binding sites for cellular transcription factors, such as NF-B, nuclear factor of activated T cells (NFAT), Sp1, and AP1 (3). NF-B-inducing agents such as TNF␣ and prostratin are strong activators of HIV-1 transcription (4), but in keeping with the pleiotropic effects of NF-B, these agents alter the expression of many cellular genes.
The study of HIV latency in vivo is complicated by the rarity of latently infected cells and a current inability to purify such cells. Many studies have relied heavily on cell-line models of chronic HIV infection. One such model is the J-Lat model of post-integration latency developed by infecting Jurkat cells using recombinant envelope-deleted virus containing a GFP reporter (5). Infected cells were then screened for latently infected cells that were transcriptionally inactive in the absence of cellular stimulation. Transcription of the HIV genome could be induced by a number of biological and chemical activating agents including NF-B activators like PMA, prostratin, phytohemagglutinin (PHA), and TNF␣. However, more physiological activation via the TCR complex with anti-CD3 antibodies was ineffective, likely because of down-regulation of TCR complex components. We recently developed an additional clonal cell line, J-Lat 5A8, that retains surface CD3 expression and responds to TCR stimulation (6). Of note, when responsiveness of a variety of latency cell models including those formed in primary CD4 T cells was compared with resting CD4 T cells from aviremic patients, J-Lat 5A8 cells along with a primary CD4 T cell model developed by Dr. Sharon Lewin's laboratory displayed the strongest similarity to the reactivation profile exhibited by patient cells (7). These data also suggest that the cell signaling pathways within J-Lat 5A8 cells are similar to ex vivo cells from patients. In the current study, we have employed J-Lat 5A8 cells and primary CD4 T cells to study select aspects of the regulation of HIV latency.
MicroRNAs (miRNAs) are evolutionarily conserved RNAs that regulate a wide variety of biological processes including the immune response to viruses. These miRNAs correspond to 20 -25-nucleotide-long non-coding RNAs that modulate gene expression through base pairing of the miRNA seed sequence to its target mRNA (usually located within the 3Ј-UTR). This interaction leads to either translational repression or mRNA cleavage, thereby reducing the final amount of target protein produced. Host miRNAs have inhibit HIV through cellular regulation of PCAF (8), cyclin T1 (9), and other HIV-1 factors involved in trafficking and/or importing pre-integration complexes into the nucleus (10). Cellular miRNAs also regulate HIV-1 by directly targeting the 3Ј-UTR of HIV-1 mRNA (11,12). Although miRNAs clearly modulate HIV infection and replication, whether miRNAs regulate viral latency is still unclear. In this study, we identify multiple miRNAs that inhibit HIV-1 reactivation and uncover a novel miRNA-target interaction that reinforces latency in infected cells.
Tripartite motif-containing (TRIM) proteins are E3 ubiquitin ligases containing a RING finger domain, one or two B-box domains, and a coiled-coil region. TRIM32, a member of the TRIM-NHL family (named after the NCL-1, HT2A, and LIN-41 proteins), contains a C-terminal domain believed to mediate protein binding. Specifically, the NHL domain of TRIM32 binds to Ago1, which activates certain miRNAs required for neural differentiation (13). In addition, TRIM32 regulates the induction of type I IFNs and the cellular antiviral response by activating STING via Lys-63-linked ubiquitination (14). Interestingly, TRIM32 expression also activates NF-B (15). A more recent study demonstrates that certain TRIM proteins (including TRIM32) that induce NF-B also promote HIV-1 LTR expression (16). These studies highlight the importance of TRIM32 in NF-B-mediated transcriptional activation of HIV-1. However, it is unknown whether TRIM32 plays a role in NF-B signaling in a manner that antagonizes HIV latency. In this study, we explore the role of TRIM32 as an antagonist of HIV latency and counter-regulation of TRIM32 by miR-155.

Experimental Procedures
qRT-PCR-Total J-Lat cell RNA was purified using TRIzol reagent (Invitrogen), and the RNA concentration was determined with an ND-1000 spectrophotometer (NanoDrop). Taq-Man low density array (TLDA) human miRNA assays (versions 2.1 and 3.0; Applied Biosystems) were used to calculate changes in miRNA expression between latent cells and reactivated cells collected from four independent experiments. Briefly, cDNA was generated using the TaqMan reverse transcription kit and Megaplex primer pools A (v2.1) and B (v3.0). TLDA cards A and B were loaded with reverse transcription products and PCR was performed on a 7900HT Fast Real-Time PCR system (Applied Biosystems). Fold changes in the expression of each miRNA were calculated (⌬⌬Ct method), and differences in expression between latent and reactivated cells were analyzed using moderated t statistics. Linear contrasts were used to make all pairwise comparisons between groups. Follow-up analyses of specific miRNAs were performed using TaqMan microRNA assays. RNU6 was used as an endogenous control. TaqMan gene expression assays (Applied Biosystems) were used to quantify the expression of mRNA transcripts. The following primers and probes were used in gene expression assays: DGCR8 (Hs00256062_m1), Dicer (Hs00229023_m1), and TRIM32 (Hs00705875_s1). GAPDH or ␤-actin was used as an endogenous control for ⌬⌬Ct calculations.
Lentiviral Infection-Lentiviral particles were produced as described (17). For J-Lat infections, 100,000 cells were incubated with 4 g/ml Polybrene (Sigma), RPMI, and viral suspension for ϳ2 h at 37°C. After 24 h, the cells were washed and cultured in RPMI.
Infection of CD4 T Cells with Dual Reporter Virus-Primary CD4 T cells and peripheral blood mononuclear cells were purified from healthy donor blood (Blood Centers of the Pacific, San Francisco, CA, and Stanford Blood Center). CD4 T cells were isolated by negative selection using the RosetteSep human CD4

miR-155 and TRIM32 Exert Opposing Effects on HIV-1 Latency
T cell enrichment cocktail (StemCell Technologies). Peripheral blood mononuclear cells were purified by Histopaque-1077 density gradient. Purified CD4 T cells from peripheral blood were cultured in RPMI 1640 medium supplemented with 10% FBS, L-glutamine (2 mM), penicillin (50 units/ml), and streptomycin (50 mg/ml). Purified CD4 T cells isolated from peripheral blood were stimulated with ␣CD3/␣CD28 activating beads (Life Technologies) at a concentration of 1 bead/cell in the presence of 30 units/ml IL-2 (PeproTech) for 3 days. All cells were spinoculated with HIV Duo-Fluo I at a concentration of 100 ng of p24 per 1 ϫ 10 6 cells for 2 h at 1,200 ϫ g at 37°C. After spinoculation, all cells were returned to culture in the presence of 30 units/ml IL-2. Infected CD4 T cells were sorted based on GFP and mCherry fluorescence 4 days after infection using a FACSAria II (BD Biosciences), and RNA was extracted immediately using the Ambion PARIS kit (Life Technologies).
Transfections-J-Lat 5A8 cells were infected with lentivirus expressing shRNA targeting Dicer and sorted by the mCherrypositive population. These sorted shDicer-infected J-Lat cells were cultured for ϳ1 week and were nucleofected (Lonza) using Solution R and the program O-028 with 2 M miRIDIAN microRNA mimics (Dharmacon), unless otherwise noted. After 24 h, cells were stimulated with ␣-CD3 and ␣-CD28 or with TNF␣ overnight or for 24 h, and analyzed by flow cytometry. For immunoprecipitation experiments, 4 million 293T cells were transfected with 3.5 g of TRIM32, IB␣, pCMV constructs, and 0.7 g of His-ubiquitin using FuGENE HD for 24 h. For Dual-Luciferase assays, the indicated luciferase plasmid (150 ng) was transfected into HeLa cells in a 24-well plate with or without miR-155 mimic (10 pmol) using Lipofectamine 2000 (Invitrogen). The Dual-Luciferase reporter assay system (Promega) was used to measure the levels of both firefly and Renilla luciferase 24 h after transfection.
Plasmid for Luciferase Assay-Luciferase reporter construct was generated using the psiCHECK2 vector (Promega). The TRIM32 3Ј-UTR was ligated into the XhoI/NotI sites of psiCHECK2. Sequencing was performed to confirm that construct contained the expected DNA sequence.
In Vitro Kinase Assay-5A8 cells were unstimulated or treated (30 min) with PMA (20 nM) and ionomycin (2 M). A total of 15 million cells were lysed in whole cell lysis buffer, and the assay was performed as described previously (4).
EMSA-Nuclear extracts were prepared (4 g/sample) and were incubated with poly(dI-dC) and salmon sperm DNA for 15 min, followed by the addition of [␥-32 ]ATP-labeled consensus B or Oct-1 oligonucleotides (Promega). For cold-probe competition, 100ϫ cold B probes were added 10 min prior to labeled probes. Samples were separated on non-denaturing gels and analyzed by autoradiography.
Cloning, Expression, and Purification of TRIM32 and TRIM25-TRIM cDNA was cloned between the KasI and XhoI restriction sites of pPROEX-HTa. The TRIM32 I22E mutant was cloned by site-directed mutagenesis. A saturated culture of ArcticExpress (DE3) cells (Agilent) harboring pPROEX-HTa-TRIM32 was diluted to an A 600 of ϳ0. Reactions were stopped by the addition of SDS sample loading buffer, resolved on a 10% SDS-acrylamide gel, and exposed on an autoradiography screen.

Knockdown of Enzymes Involved in miRNA Biogenesis
Increases Reactivation of Latent HIV-Because various miRNAs block HIV replication, we hypothesized that select miRNAs promote HIV latency in infected CD4 T cells. To test this possibility, miRNA-deficient J-Lat 5A8 cells were prepared by knocking down the expression of two enzymes required for miRNA biogenesis, DGCR8 and Dicer. Knockdown was achieved by introduction of an shRNA lentiviral vector (pSicoR-MS1) containing an mCherry reporter into the J-Lat

miR-155 and TRIM32 Exert Opposing Effects on HIV-1 Latency
5A8 cells. The mCherry-positive cells, which were successfully infected with the lentivirus, were isolated by fluorescent cell sorting followed by qRT-PCR to assess target gene expression. Knockdown of both DGCR8 and Dicer exceeded 90% in the mCherry-positive 5A8 cells as compared with levels obtained in cells receiving a negative control (5A8 shScramble) (Fig. 1A). Additionally, DGCR8 and Dicer levels were much lower in the knock-out cells than in cells receiving no shRNA, and further each knockdown appeared specific as expression of the counterpart microRNA biogenic enzyme was not affected, nor was the expression of GAPDH altered (Fig. 1B). To assess the effects of diminished miRNAs on HIV reactivation, the infected cells were cultured alone or stimulated with ␣-CD3/CD28 or TNF␣. Both DGCR8 and Dicer knockdown cells exhibited significantly greater increases in HIV-1 reactivation with both agonists as demonstrated by increased GFP expression as compared with the shRNA scramble control knockdown (Fig. 1C). GFP expression in the absence of stimulation was not markedly affected by knockdown of either enzyme. These findings suggest that one or more miRNAs naturally function as inhibitors of latent HIV reactivation.
To identify the specific miRNAs that contribute to this inhibition, we used a TLDA to profile 754 unique human miRNAs (supplemental Table 1). We identified miRNAs that were differentially expressed between reactivated and latent cells (Table 1), validating top candidates in individual TaqMan assays (data not shown).

Introducing miR-155 into Dicer-deficient Cells Decreases the
Level of HIV Reactivation-To determine whether candidate miRNAs inhibited HIV-1 reactivation, each candidate miRNA TABLE 1 TLDA identifies miRNAs involved in HIV-1 latency J-Lat 5A8 cells were stimulated with ␣CD3/CD28 and sorted for either GFP-positive (reactivated) or GFP-negative cells (latent) cells. TLDA was performed, and fold changes were calculated using the ⌬⌬Ct method. Differences in expression between latent and activated cells were analyzed using moderated t statistics. Samples in bold indicate the top candidates that were individually validated and carried into future experiments. RNU6 (row in italics) was used as a housekeeping control. Data represent four independent experiments.

FIGURE 1. Stable knockdown of DGCR8 and Dicer promotes HIV-1 reactivation in the J-Lat 5A8 T cell model of HIV latency. A, knockdown of DGCR8 and
Dicer mRNA in J-Lat 5A8 cells infected with lentiviruses expressing these specific shRNAs or a scrambled shRNA control was measured by qRT-PCR. Values are normalized to scrambled shRNA control; error bars represent ϩS.E., **, p Ͻ 0.01 (two-tailed t test), n ϭ 3. B, knockdown as described in A was also measured at the protein level by immunoblotting with specific antibodies. Immunoblotting with anti-GAPDH antibodies was used to assess comparability of the different lysates. C, increased HIV-GFP reactivation following stable knockdown of DGCR8 or Dicer assessed by flow cytometry 24 h after stimulation. Error bars represent ϩS.E., ***, p Ͻ 0.001 (two-way ANOVA), n ϭ 3.
was reintroduced into Dicer-deficient cells. Although knocking down Dicer nearly doubled the level of reactivation in stimulated cells (Fig. 1C), introducing the top four candidate miRNAs from both GFP-positive and GFP-negative cells countered this effect, restoring a lower level of reactivation as compared with cells receiving a negative control miRNA mimic ( Fig. 2A).
Introducing miR-155 produced the greatest decline in HIV-1 reactivation (78.3% rescue) ( Fig. 2A), occurring in a dose-related manner (Fig. 2B). Two of the miRNAs tested, miR-1290 and miR-505* (asterisk indicates passenger strand), did not significantly affect reactivation levels, indicating that this effect is specific and not associated with the introduction of every

miR-155 and TRIM32 Exert Opposing Effects on HIV-1 Latency
miRNA. To further evaluate miR-155 levels when cells were stimulated, individual TaqMan assays were performed. miR-155 was expressed at 50-fold higher levels in GFP-positive cells (ϳ50-fold more) as compared with unstimulated cells. miR-155 levels were also increased in GFP-negative cells following stimulation, albeit to a lesser extent (Fig. 2C). Together, these findings raise the possibility that specific miRNAs abundantly expressed in reactivated cells serve to counter viral activation and perhaps promote a return to latency in cells undergoing transient virus production.
To determine the levels of miR-155 in a model of HIV-1 latency formed with primary CD4 T cells, we employed a dualfluorescence virus, HIV-DuoFluo, to identify latent and productively infected primary CD4 T cells occurring after HIV infection. This model has been recently described (20). In this model, HIV-DuoFluo (R7GEmC) expresses GFP under control of the HIV promoter and expresses mCherry under control of a constitutive EF1␣ promoter. Following integration of this virus into primary CD4 T cells, mCherry is constitutively expressed and marks all successfully infected cells. Active viral expression occurs in a subset of these cells indicated by the expression of GFP. Latently infected cells are identified by the expression of mCherry in the absence of GFP expression. CD4 T cells from two blood donors (Donor 9600 and Donor 9601) were infected with R7GEmC and sorted for uninfected cells, latently infected cells (mCherry-positive, GFP-negative), and actively infected cells (mCherry-positive, GFP-positive). Both donors exhibited high levels of miR-155 in cells that were TCR-stimulated but uninfected. This finding is consistent with prior studies describing up-regulation of miR-155 in activated lymphoid and myeloid cells (21,22). Levels of miR-155 were much higher in virus-producing cells (mCherry-positive, GFP-positive) as compared with latently infected cells (mCherry-positive, GFPnegative) (Fig. 2D). These findings suggest that miR-155 may principally act within an activated cellular environment. Interestingly, miR-29b levels did not change in any of the tested cell populations. This microRNA has been implicated in the regulation of cyclin T1, a participant in the PTEF-b complex that functions as a key cofactor for HIV Tat (9).
MicroRNA-155 Targets TRIM32-Next, we sought to identify host genes whose action was suppressed by miR-155 using the TargetScan prediction algorithms. These analyses identified TRIM32 as a high-value hit because it had a favorable aggregate P CT and contextϩ score. In addition, the 3Ј-UTR of TRIM32 contains an 8-mer site that precisely matches the seed region (positions 1-8) of miR-155 (Fig. 3A). Furthermore, TRIM32 represented an intriguing potential target because of a prior report of its direct binding to the activation domain of HIV Tat (21). To study whether miR-155 directly binds to the 3Ј-UTR of TRIM32, the full 3Ј-UTR of TRIM32 including the predicted target site was cloned downstream of a Renilla luciferase reporter construct. In this assay, a decrease in luciferase activity in the presence of miR-155 demonstrates binding of the miRNA and subsequent inhibition of luciferase expression. (Fig. 3A). Indeed, miR-155 inhibited luciferase expression of the reporter containing the full 3Ј-UTR of TRIM32 (Fig. 3B). These results confirm that TRIM32 is a target of miR-155.
Studies were next performed to assess miR-155 effects on endogenous TRIM32 levels in J-Lat 5A8 cells. The introduction of miR-155 decreased levels of TRIM32 mRNA in these cells by ϳ70% as compared with a negative control miRNA (Fig. 3C). Recent studies in the Rudensky laboratory (22) have analyzed endogenous targets of miR-155 within the entire murine transcriptome using AGO differential HITS-CLIP (dCLIP) and mRNA changes in CD4 T cells isolated from wild type or miR-155 knock out mice. Binding maps were made accessible online for all 3Ј-UTRs on CLIP Base. We searched for TRIM32 in CLIP Base and found that the putative miR-155 binding site in the TRIM32 3Ј-UTR is in fact bound by the miR-155-AGO complex in wild type CD4 T cells, but not in CD4 T cells lacking miR-155 due to gene knock-out (Fig. 3D, image obtained from CLIP Base). Together, these findings provide strong evidence supporting TRIM32 as a bona fide cellular target of miR-155.
TRIM32 Regulates HIV-1 Latency-Next, we assessed the effects of shRNA knockdown of TRIM32 in latently infected J-Lat 5A8 cells. After infection with a TRIM32 shRNA, TRIM32 mRNA and protein levels were both reduced by ϳ40% (Fig. 4, A  and B). Conversely, TRIM32 mRNA and protein levels were not reduced in cells infected with virus encoding a scrambled control shRNA (Fig. 4, A and B). In agreement with one or more miRNAs posttranscriptionally regulating TRIM32 expression, TRIM32 protein, but not mRNA, was increased when Dicer was knocked down (Fig. 4, A and B). TCR activation of J-Lat 5A8 cells with anti-CD3 and anti-CD28 antibodies decreased GFP expression when TRIM32 expression was knocked down but increased expression following Dicer knockdown (Fig. 4C). These findings suggest that TRIM32 promotes reactivation of latent HIV-1 and that its activity is regulated by miRNAs, notably miR-155.
Next, we examined the effects of lentivirus-mediated overexpression of TRIM32 (pCDH-TRIM32) in J-Lat 5A8 cells. By both qRT-PCR and immunoblotting TRIM32 levels were significantly higher in pCDH-TRIM32 cells as compared with cells infected with the empty lentiviral vector (pCDH) (Fig. 4, D and E). TRIM32 alone promoted HIV-1 reactivation in the absence of stimulation. Treatment with anti-CD3/CD28 and TNF␣ resulted in increased levels of latent virus reactivation in cells expressing TRIM32 (Fig. 4F). Together, these findings confirm TRIM32 as an HIV activator and effective host antagonist of HIV latency.

TRIM32 Promotes Reactivation of Latent HIV-1 by Stimulating NF-B
Signaling-TRIM32 has been reported to interact with the activation domain of HIV-1, HIV-2, and equine infectious anemia virus Tat proteins (21). To assess whether the activating effects of TRIM32 on latent HIV required the presence of Tat, TRIM32 was expressed in J-Lat A72 cells that lack Tat (23). TRIM32 effectively induced latent provirus expression alone or in combination with anti-CD3/CD28 antibodies or TNF␣ (Fig. 5A). These results indicate that the TRIM32 is able to activate the HIV LTR in a Tat-independent manner. Recent studies have shown that TRIM32 induces NF-B in 293 cells (14,16). We examined whether TRIM32 similarly activates NF-B using Jurkat B-dsRed cells, which contain a basal promoter and five B enhancer elements controlling expres- MAY 29, 2015 • VOLUME 290 • NUMBER 22

JOURNAL OF BIOLOGICAL CHEMISTRY 13741
sion of a dsRed reporter (6). When TRIM32 was expressed in these cells, the level of dsRed expression increased as compared with the empty vector control in both the absence and the presence of stimulation (Fig. 5B). These findings confirm the NF-B-inducing activity of TRIM32.
Next, nuclear and cytoplasmic extracts from TRIM32-expressing cells were immunoblotted. In the absence of concomitant PMA and ionomycin stimulation, TRIM32 expression was associated with an increase in nuclear RelA expression and a decrease in cytoplasmic IB␣ levels (Fig. 5C). Based on the pattern of nuclear Sp1 and cytoplasmic IB␣, nuclear and cytoplasmic fractions were of high quality (Fig. 5C). To confirm TRIM32-induced nuclear translocation of NF-B, EMSA were performed with nuclear extracts and 32 P-labeled B and Oct1 DNA probes. Unstimulated cells expressing TRIM32 displayed increased binding activity in the form of a more slowly migrating complex (p50/RelA heterodimer) that was fully competed with unlabeled wild type B probe but not by a mutant B probe. (Fig. 5D). In contrast, Oct1 DNA binding did not change in the presence or absence of TRIM32. Together, these results indicate that TRIM32 expression induces NF-B activation in CD4 T cells.
To further investigate the mechanism through which TRIM32 induces NF-B, an IB kinase (IKK) assay was performed (Fig. 5E). Unexpectedly, the same level of IKK activity was found in cells expressing TRIM32 versus the negative con- . miR-155 directly targets TRIM32. A, prediction of the binding site of miR-155. The TRIM32 3Ј-UTR site is predicted to bind miR-155 perfectly at the seed region. This region of the 3Ј-UTR of TRIM32 is well conserved among many mammals. Hsa, Homo sapiens; Ptr, Pan troglodytes; Mml, Macaca mulatta; Mmu, Mus musculus; Rno, Rattus norvegicus. B, luciferase binding assay shows that miR-155 binds to TRIM32 3Ј-UTR. The full TRIM32 3Ј-UTR was cloned into psi-Check2, transfected into HeLa cells (with or without miR-155), and analyzed for levels of luciferase expression 24 h after transfection. Expression of Renilla luciferase was normalized to that of firefly luciferase for each sample. Error bars represent S.E., ***, p Ͻ 0.001 (one-way ANOVA), n ϭ 3. ns, not significant. C, miR-155 decreases endogenous TRIM32 levels. Negative control miRNA and miR-155 mimics were nucleofected into 5A8 shDicer cells. Cells were collected 24 h after nucleofection, and analyzed by qRT-PCR. ***, p Ͻ 0.001 (two-tailed t test), n ϭ 3. D, binding map demonstrates that miR-155 binds to TRIM32 3Ј-UTR in wild type miR-155 mice, but not to miR-155 knock-out mice. The y-axis represents read counts, and the x-axis represents nucleotide position relative to the beginning of the 3Ј-UTR. Blue indicates the number of reads from miR-155 (WT) replicates, and orange indicates reads from miR-155-deficient (knock-out) mice. The figure was adapted from a database assembled by the Rudensky laboratory (CLIP Base).

miR-155 and TRIM32 Exert Opposing Effects on HIV-1 Latency
trol. However, brief stimulation with PMA and ionomycin increased IKK activity. Furthermore, there was no evidence of phosphorylated IB␣ in the lysate of cells expressing TRIM32 (Fig. 5E). These findings suggest that TRIM32 acts downstream of the IKKs in the NF-B pathway. Because our prior experiments indicated that TRIM32 expression promotes degradation of IB␣, we hypothesized that TRIM32 directly binds to and ubiquitinates IB␣. To test this possibility, TRIM32 and IB␣ interaction was tested in cells expressing HA-tagged IB␣ and FLAG-tagged full-length TRIM32 or FLAG-tagged mutant TRIM32 containing a deletion in the RING domain (TRIM32⌬RING), thereby removing its E3 ubiquitin ligase activity. When anti-TRIM32 antibodies were used in the immunoprecipitations, both IB␣ and the IB␣ super-repressor containing S32A/S36A mutations corresponding to the key IKK phosphoacceptor sites (24) were coimmunoprecipitated. However, when anti-IB␣ antibodies were used, lower amounts of the TRIM32⌬RING protein were coimmunoprecipitated as compared with wild type TRIM32 (Fig. 5F). The ability of TRIM32 to ubiquitinate IB␣ was next tested. In the presence of the ␤-lactone proteasome inhibitor, full-length TRIM32 promoted ubiquitination of IB␣, whereas TRIM32-⌬RING did not. In addition, TRIM32 also promoted ubiquitination of the IB␣ super-repressor, but at slightly lower levels as compared with wild type IB␣ (Fig. 5G). To confirm these findings, we assayed the ability of recombinant TRIM32 (wild FIGURE 4. TRIM32 regulates reactivation of latent HIV-1 in J-Lat 5A8 cells. A, lentivirus expressing shRNAs targeting TRIM32, Dicer, and a control scramble shRNA were introduced into J-Lat 5A8 cells. TRIM32 levels were measured by qRT-PCR. Error bars indicate S.E., ***, p Ͻ 0.001 (one-way ANOVA), n ϭ 3. B, knockdown of TRIM32 as described in A assessed at the protein level by immunoblotting with ␣TRIM32 antibodies. GAPHD1 was used a loading control for the various lysates. C, knockdown of TRIM32 reduces levels of reactivation following stimulation of J-Lat 5A8 cells. Cells were infected with lentivirus expressing the indicated shRNAs, stimulated with ␣CD3/CD28 for 24 h, and analyzed by flow cytometry. Error bars indicate S.E., **, p Ͻ 0.01 (one-way ANOVA), n ϭ 3. No Stim, unstimulated cells. D, cells were infected with lentivirus expressing TRIM32 (pCDH-TRIM32) or empty vector (pCDH), and levels of TRIM32 were measured by qRT-PCR. Error bars indicate S.E., ***, p Ͻ 0.001 (one-way ANOVA), n ϭ 3. E, following expression of TRIM32 protein as described in D, lysates were immunoblotted with ␣TRIM32 antibodies to assess levels of expression. Hsp90 immunoblotting was performed to control for protein loading. F, TRIM32 expression promotes HIV-1 reactivation alone and in combination with ␣CD3/CD28 or TNF␣ stimulation. 5A8 pCDH-TRIM32 or pCDH cells were either not stimulated or stimulated with ␣CD3/CD28 or TNF␣ as indicated. Error bars depict S.E., ***, p Ͻ 0.001 (two-way ANOVA), n ϭ 3.
type versus catalytic-dead mutant) to ubiquitinate radiolabeled IB␣ in vitro. Wild type TRIM32 mediated ubiquitination of IB␣, whereas the I22E mutant, lacking ubiquitin transfer activity, did not (Fig. 5H). Furthermore, we tested the ability of another NF-B-activating TRIM protein, TRIM25, to directly ubiquitinate IB␣. We found that recombinant TRIM25 ubiq-uitinated IB␣ in vitro at levels similar to that of TRIM32 (Fig.  5I). Together, these results suggest that both TRIM32 and TRIM25 bypass many of the usual upstream NF-B signaling steps by directly binding to and ubiquitinating IB␣ and that this mechanism may be conserved in other NF-B-activating TRIM E3 ligases.

Discussion
Achieving a cure for HIV will require both the complete suppression of active viral replication in vivo and the clearance of the transcriptionally silent proviral reservoir. Current antiviral drugs effectively target the active virus but have no effect on the latent reservoir. One potential approach for attacking the latent reservoir is to identify combinations of agents that activate latent proviral transcription without inducing full T cell activation. These agents would promote viral replication under the cover of ART without inducing a toxic cytokine storm (25). Attacking the latent reservoir is further complicated by the fact that only a fraction of the latent proviruses appear to respond to these inducing agents (26). The variegated nature of the response suggests that repeating cycles of induction will likely be required. The half-life of productively infected cells is usually quite short, often measuring less than 24 h (27). However, recent studies suggest that induction of virus production in the latent reservoir does not result in the death of the virus-producing memory T cells (28). The production of viral proteins, however, does render this population of cells transiently visible to the immune system and thus potentially vulnerable to elimination by cytotoxic T cells or antibodies mediating antibody-dependent cellular cytotoxicity (28,29). Latently infected cells do not appear to remain continuously latent. Rather, these cells are intermittently stimulated by cytokines or other uncharacterized signals to produce a burst of virus sometimes culminating in a visible "blip" in the viral load (30). Furthermore, low levels of virus persist in subjects on ART, below the normal detection level of standard assays (31). These viruses likely reflect intermittent production of virus by cells within the latent reservoir. However, after producing virus, these cells appear able to retreat back to a quiescent state with reestablishment of viral latency.
We have explored the potential involvement of miRNAs as key regulators of HIV latency. This notion was reinforced when knockdown of two essential enzymes required for miRNA biogenesis, DGCR8 and Dicer, was shown to lead to the activation of latent HIV proviruses. These findings are also consistent with prior studies describing an overall increase in HIV replication following knockdown of Drosha and Dicer in multiple cell lines and peripheral blood mononuclear cells (8,12). Our findings suggest that miRNAs predominantly promote inhibi-tion HIV-1 replication and thus function as pro-latency factors. However, we certainly do not exclude the possibility that some individual miRNAs promote viral reactivation by blocking the action of inhibitory transcription factors.
We now describe a new gene target for miR-155 and demonstrate that this target, TRIM32, is sufficient to promote transcriptional activation of latent HIV-1. TRIM32 acts, at least in part, by a Tat-independent mechanism stimulating nuclear translocation of NF-B. However, our studies reveal that TRIM32 activates NF-B in a novel manner involving direct ubiquitination of IB␣. Specifically, TRIM32 induction of NF-B proceeds independently of IKK activation within signalosomes (Fig. 6). Our studies of the potential role of microRNAs in the regulation of HIV latency have revealed an interesting network of regulation highlighting how an HIV transcriptional activator antagonizing latency through the induction of NF-B is counter-regulated by an NF-B-inducible microRNA that suppresses its expression.
We focused on miR-155 in this study because, as compared with other miRNAs, it exerted the most potent blocking effects on HIV-1 reactivation in J-Lat 5A8 cells. However, our studies suggested that multiple miRNAs may play a role in blocking the reactivation of latent HIV. These other miRNAs may also be responsible for the increased levels of reactivation we demonstrate in shDGCR8 and shDicer cells following stimulation. It would be interesting to follow up on these other miRNAs as they could lead to additional miRNA host targets that regulate HIV latency. As noted previously, among all of the continuous T cell lines, J-Lat 5A8 cells most closely mirror the pattern of responsiveness found in latently infected cells from patients suppressed with ART (7). Reflecting its induction by NF-B, miR-155 levels are higher in activated than in resting T cells (32)(33)(34)(35). Prior studies have implicated miR-155 in the regulation of cellular proteins involved in trafficking and/or nuclear import of HIV pre-integration complexes including ADAM10, TNFO3, Nup153, and lens epithelium-derived growth factor (LEDGF)/p75 (10). Our findings now extend the effects of miR-155 to a post-integration level where it may promote reestablishment of latency in reservoir cells transiently producing virus.
miR-155 is known to play a key role in the regulation of viruses other than HIV. For example, EBV strongly induces the FIGURE 5. Expression of TRIM32 promotes HIV-1 reactivation by inducing NF-B activation. A, J-Lat A72 cells, which lack Tat, were reactivated by pCDH-TRIM32, but not pCDH (empty vector) in the absence of other stimuli or in the presence of either ␣CD3/CD28 or TNF␣. GFP expression was detected by flow cytometry. Error bars indicate S.E., ***, p Ͻ 0.001 (two-way ANOVA), n ϭ 3. No Stim, unstimulated cells. B, TRIM32 promotes NF-B activation in Jurkat B-dsRed cells. Cells were infected with pCDH-TRIM32 or pCDH lentivirus and stimulated with either ␣CD3/CD28 or TNF␣, and then analyzed by flow cytometry. Error bars depict S.E., ***, p Ͻ 0.001 (two-way ANOVA), n ϭ 3. C, TRIM32 expression increases RelA levels in the nucleus and decreases IB␣ levels in the cytoplasm. Nuclear and cytoplasmic extracts were isolated for unstimulated 5A8 cells, and 5A8 cells treated with PMA and ionomycin (Iono) for 30 min. D, TRIM32 specifically promotes the formation of nucleoprotein complexes on B sites but not mutant (Mut) B sites. 5A8 cells were infected with pCDH or pCDH-TRIM32 lentivirus, and nuclear extracts were isolated. An EMSA was performed using 32 P-labeled B enhancer probes or negative control 32 P-labeled Oct1 probes. In addition, nuclear lysates were pre-incubated with a 100-fold excess of either unlabeled wild type or mutated B probes. E, IKK activity is not increased in unstimulated TRIM32-expressing 5A8 cells as compared with pCDH controls, nor is phospho-IB␣ detected. Positive controls show that IKK kinase activity is increased following 30-min treatment with PMA, and ionomycin and phospho-IB␣ also increase. F, ␣FLAG immunoprecipitations (IP) of FLAGtagged TRIM32 or TRIM32⌬RING and HA-tagged IB␣ or IB␣ super-repressor S32A/S36A (SS/AA) in 293T cells indicate that these proteins interact with each other. In contrast, when IB␣ or its corresponding super-repressor (S32A/S36A) is first immunoprecipitated with ␣HA antibodies, TRIM32 is coimmunoprecipitated, but the TRIM32⌬RING binding is greatly reduced. G, in the presence of a proteasome inhibitor, immunoprecipitation of IB␣ followed by blotting for His-tagged ubiquitin shows that wild type TRIM32 is capable of ubiquitinating IB␣ (wild type and S32A/S36A) in 293T cells. The TRIM32⌬RING mutant does not produce similar ubiquitination, but this could reflect its diminished recovery in IB␣ coimmunoprecipitations. H, IB␣ is ubiquitinated by wild type TRIM32 in vitro, but not by an I22E mutant deficient in E2 binding. Recombinant TRIM32 (wild type or I22E) was purified and incubated with 35 S-labeled IB␣ for 0 or 1 h. Ubiquitination reactions were visualized by autoradiography. I, both wild type TRIM32 and TRIM25 ubiquitinate IB␣. The experiment was performed as described for H. MAY 29, 2015 • VOLUME 290 • NUMBER 22 expression of miR-155, leading to increased growth and transformation of B cells (36,37). In addition, Kaposi sarcoma-associated herpesvirus and Marek disease virus type 1 encode viral mimics of miR-155, which are thought to be involved in oncogenic transformation of infected cells (38). Kaposi sarcomaassociated herpesvirus miR-K11 and Marek disease virus type 1 miR-M4 both contain the full identical seed sequence of miR-155, and therefore likely regulate many of the same cellular targets (39,40). Furthermore, miR-155 regulates a variety of targets involved in mammalian immunity (41). For example, during thymic differentiation, increases in miR-155 stimulate T regulatory cell fitness and proliferative potential by targeting the suppressor of cytokine signaling 1 (Socs1) (42). Also, miR-155 is required for the development of inflammatory Th17 cells in experimental autoimmune encephalomyelitis (43,44). However, the precise role of miR-155 in immune processes is not well understood because miR-155, like many miRNAs, interacts with a large number targets, some of which exert opposing actions.

miR-155 and TRIM32 Exert Opposing Effects on HIV-1 Latency
Although cellular miRNAs may be capable of directly targeting HIV (12), our focus was on uncovering host gene products that participate in the regulation of HIV-1 latency. In general, miRNAs are regarded as fine-tuners of gene expression (45). Having identified miR-155 as the most HIV-repressive microRNA that was induced in our analysis, we used Target-Scan (46) for in silico prediction of its cellular targets. Through this process, we identified TRIM32 as a potential miR-155 target and experimentally confirmed the ability of miR-155 to interact with the 3Ј-UTR of TRIM32 to suppress its expression. Of note, the Rudensky laboratory (22) also performed experiments with 12 miR-155 wild type mice and 12 miR-155 knockout mice, identifying miR-155 target genes in CD4 T cells. When we searched for TRIM32 on their CLIP database, we found that the evolutionarily conserved TRIM32 binding sequence did bind to an AGO-miR-155 complex in vivo. Taken together, these findings confirm TRIM32 as a bona fide miR-155 host gene target.
As we predicted, when levels of TRIM32 were reduced with shRNA, lower reactivation of latent HIV was observed. Conversely, when TRIM32 was overexpressed in cells harboring latent provirus, increased viral activation was detected. Notably, TRIM32 expression induced reactivation, even in the absence of other stimuli. Interestingly, knockdown of DGCR8 and Dicer in J-Lat 5A8 cells does not result in a spontaneous increase in the level of reactivation in the absence of stimulation. Considering that the levels of both TRIM32 and miR-155 are low in unstimulated cells, we would not expect a further decrease in miR-155 levels to promote reactivation in the absence of a stimulatory signal.
These findings may be relevant in the context of aviremic patients on highly active antiretroviral therapy who commonly experience intermittent biological and statistical fluctuations in levels of viremia, termed blips (30). It is possible that the natural down-regulation of virus expression reflects changes in the levels of miR-155 expression, which in turn could result in decreased TRIM32 expression. These events could contribute to resilencing of the virus and maintenance of the latent reservoir. Similarly, increases in miR-155 levels during shock and kill could impair latent virus reactivation, thereby avoiding a viral cytopathic effect or allowing the cell to remain undetected by the immune system.
A prior study indicated that TRIM32 interacts with Tat but did not identify any functional consequences of this interaction. Although we certainly do not exclude a role for TRIM32-Tat interaction in HIV biology, our findings indicate that TRIM32 can function independently of Tat. Specifically, we have shown

miR-155 and TRIM32 Exert Opposing Effects on HIV-1 Latency
in biochemical and functional experiments that TRIM32 expression promotes nuclear NF-B translocation and engagement of the cognate B enhancer sites present in the HIV-1 LTR. Consistent with our findings, a number of TRIM proteins including TRIM32 are capable of inducing the expression of NF-B, AP-1, and activating the HIV-1 LTR in HEK293 cells (16). In that study, knockdown of TAK1 (upstream of IKK) decreased the level of NF-B induction following stimulation with eight different TRIMs including TRIM32. Surprisingly, we observed that TRIM32 expression did not activate the IKK complex. Instead, we found that TRIM32 is able to activate NF-B by direct ubiquitination of either wild type IB␣ or IB␣ super-repressor lacking IKK phospho-acceptor site promoting release and nuclear translocation of the NF-B heterodimer. Thus, IB␣ serves as a novel target of the E3 ligase TRIM32.
Supporting these findings, we have also shown that TRIM32 can directly ubiquitinate IB␣ in vitro, whereas the introduction of a single point mutation (I22E) in TRIM32, which abrogates the activity of ubiquitin transfer without destabilizing the overall fold of the RING domain, prevents IB␣ ubiquitination. Furthermore, to investigate whether this mechanism occurs in other TRIMs, we performed an in vitro ubiquitination assay and found that TRIM25, an NF-B-activating E3 ligase, is also capable of directly ubiquitinating IB␣. This suggests that this mechanism may be more widespread among other TRIMs that activate NF-B. Although over 600 human RING-based E3 ligases control many cellular processes, we have only a rudimentary understanding of their functions, substrates, and mechanisms of action (47). In the canonical NF-B pathway, ubiquitination of IB is normally carried out by the SCF-␤TrCP E3 ligase (SKP1-CUL1-F-box ligase containing the F-box protein ␤TrCP) (48 -50). We propose a novel form of NF-B activation, independent of IKK activation, in which TRIM32, TRIM25, and perhaps other TRIMs directly bind to and ubiquitinate IB␣.
It is important to note that TRIM32 targets a number of substrates involved in diverse pathways (13)(14)(15)18), and modulation of these pathways may also promote NF-B signaling. For example, TRIM32 targets STING for Lys-63-linked ubiquitination, thereby promoting its activation and subsequent induction of NF-B and production of IFN-␤ in response to Sendai virus and Herpes simplex virus 1 infection (14). This suggests that TRIM32 (via STING) plays an important role in the innate immune response against these viruses. However, in the context of HIV-1, activation of the NF-B pathway by TRIM32 leads to the production of virus. It seems possible that the suppressive interplay of the inducible miR-155 with the similarly inducible TRIM32 was designed as a means to control the overall stimulatory effects of TRIM32. In the case of HIV, inhibition of TRIM32 by miR-155 may have the untoward effect of promoting a return of viral latency, thereby providing a mechanism for HIV persistence in infected reservoir cells transiently producing virus.
It is clear that a better understanding of the mechanisms that govern transcriptional activation of HIV is critical to develop new treatments that will force the reactivation of latent provirus and subsequent eradication of HIV-1 from patients. Identification of novel host regulatory pathways that govern HIV-1 transcription enhances our understanding of the molecular basis of HIV latency and may offer avenues for pharmacological intervention.