|Home | About | Journals | Submit | Contact Us | Français|
Latently infected cell reservoirs represent the main barrier to HIV eradication. Combination antiretroviral therapy (cART) effectively blocks viral replication but cannot purge latent provirus. One approach to HIV eradication could include cART to block new infections plus an agent to activate latent provirus. NF-κB activation induces HIV expression, ending latency. Before activation, IκB proteins sequester NF-κB dimers in the cytoplasm. Three canonical IκBs, IκBα, IκBβ, and IκBε, exist, but the IκB proteins' role in HIV activation regulation is not fully understood. We studied the effects on HIV activation of targeting IκBs by single and pairwise small interfering RNA (siRNA) knockdown. After determining the relative abundance of the IκBs, the relative abundance of NF-κB subunits held by the IκBs, and the kinetics of IκB degradation and resynthesis following knockdown, we studied HIV activation by IκB knockdown, in comparison with those of known HIV activators, tumor necrosis factor alpha (TNF-α), tetradecanoyl phorbol acetate (TPA), and trichostatin A (TSA), in U1 monocytic and J-Lat 10.6 lymphocytic latently infected cells. We found that IκBα knockdown activated HIV in both U1 and J-Lat 10.6 cells, IκBβ knockdown did not activate HIV, and, surprisingly, IκBε knockdown produced the most HIV activation, comparable to TSA activation. Our data show that HIV reactivation can be triggered by targeting two different IκB proteins and that IκBε may be an effective target for HIV latency reactivation in T-cell and macrophage lineages. IκBε knockdown may offer attractive therapeutic advantages for HIV activation because it is not essential for mammalian growth and development and because new siRNA delivery strategies may target siRNAs to HIV latently infected cells.
Despite remarkable advances in understanding the biology of HIV, the pathogenesis of HIV disease, and the development of combination antiretroviral therapy (cART), which can lead to the restoration of immune function, vastly extend health and life, and dramatically decrease the risk of transmission, no currently available therapies can safely and effectively attack and eliminate the long-lived reservoirs of HIV latently infected cells. While other factors may play some role (1, 2), the main reason for the inability of cART to eradicate HIV infection is the existence of long-lived reservoirs of latently infected cells (3–10). One approach to attacking the reservoirs and potentially eradicating HIV involves activating HIV replication in latently infected cells while blocking new infection of cells with cART, an approach sometimes termed “shock (or “kick”) and kill” (reviewed in references 4 and 11–13).
To develop new ways to activate HIV, we must understand the viral and cellular factors that influence postintegration latency and HIV activation. Much interest has recently centered on developing HIV activation therapeutics that act epigenetically (reviewed in reference 14), such as via DNA methylation and histone acetylation. Some epigenetic agents are being studied in early clinical trials, but such strategies have the potential for significant off-target effects (OTE) (15–21), and some studies have suggested that these agents may have distinct activities in different latently infected model systems (22). Other studies have worked to identify new targets (23), such as genes that can be targeted to activate HIV expression when subjected to knockdown by RNA interference (RNAi) (24–26). Multiple agents acting through different mechanisms may be needed to effectively activate all latent HIVs, or certain latent HIVs may need multiple complementary agents for activation.
HIV gene expression depends initially on long terminal repeat (LTR) activation. Starting with the initial studies of HIV gene expression, investigators considered the LTR as an activatable T cell gene (27). Expression from the LTR initially requires important basal cellular transcription factors, plus inducible factors, critically members of the nuclear factor κB (NF-κB) family, and other cellular factors (27–36).
NF-κB is an inducible transcription factor that regulates a broad range of processes (37, 38). NF-κB transcription factors are dimers produced by the combination of five different monomers (RelA [p65], NF-κB1 [p50], NF-κB2 [p52], c-Rel, and RelB). All NF-κB family members share a 300-amino-acid N-terminal Rel homology domain (RHD) responsible for dimerization, DNA binding, nuclear translocation, and interaction with IκB proteins. Only p65, c-Rel, and RelB contain a transcription activation domain (TAD), which is responsible for gene expression activation. When p50 and p52 form homodimers, they act as transcription inhibitors; p50 and p52 activate gene expression as heterodimers when combined with a TAD-containing monomer. All NF-κB monomers can form homo- or heterodimers in vivo, except for RelB, which only can form heterodimers. There are 15 possible NF-κB dimers, but only 9 are potential transcriptional activators due to the absence in the other dimers (p50/p50, p52/p52, p50/p52, RelA/RelB, and c-Rel/RelB) of DNA binding or transcriptional activating activity (39). The different TAD-containing NF-κB subunits have distinct specificities and, presumably, functions within the cell (40, 41). The p50/p65 heterodimer is typically the most abundant activating dimer.
In HIV latency, p50 homodimers bind the κB sites in the HIV LTR and recruit histone deacetylase 1 (HDAC1), repressing expression (42). Upon stimulation, TAD-containing NF-κBs (e.g., p50/p65 heterodimers), replace p50 homodimers, activating expression (reviewed in references 5 and 43), which is also enhanced by basal and inducible transcription factors, such as Sp1 and nuclear factor of activated T cells (NFAT) (44, 45). Transcription activation leads to the production of the viral protein Tat, which binds the transactivating responsive element (TAR) in the nascent, paused HIV transcript. Tat recruits the pTEFb complex, which phosphorylates RNA polymerase II (Pol II), strongly increasing Pol II processivity and resulting in high levels of HIV gene expression.
HIV-activating approaches employing cytokines and chemokines, which act through the NF-κB axis, have been studied, including agents such as interleukin-2 (IL-2) alone and with OKT3 or gamma interferon (IFN-γ) (46–49). However, such studies showed unacceptable toxicities or ineffectiveness against latent reservoirs or both. Small molecules have also been used to activate HIV via NF-κB pathways. The best known of these is the powerful diacyl glycerol mimetic phorbol myristyl acetate (PMA; here, tetradecanoyl phorbol acetate [TPA]) (27) and its derivatives, such as prostratin (50, 51), but phorbol esters are oncogenic, and even the less toxic derivatives still show significant toxicity (50).
IκB proteins regulate NF-κB signaling. IκBs bind NF-κB dimers in the cytoplasm, preventing the NF-κB proteins from translocating to the nucleus to regulate gene expression. Five IκB proteins have been described in humans: the canonical proteins IκBα, IκBβ, and IκBε and the nonclassical proteins Bcl-3 and IκBζ. IκB members have 5 to 7 ankyrin (ANK) repeats that mediate interaction with NF-κB dimers. Some NF-κB dimers exhibit binding preferences for certain IκBs (52). Thus, the relative abundance of certain IκB proteins within the cell may affect the availability of specific NF-κB dimers for activation; activation and subsequent release of certain NF-κB dimers may differentially activate certain genes. Moreover, different cell types produce different complements of NF-κB dimers (53).
Two signaling pathways, the classical and alternative pathways, lead to nuclear translocation of NF-κB dimers. In the classical pathway, the ligand activates the NF-κB essential modulator protein (NEMO), mainly releasing RelA-, c-Rel-, and p50-containing dimers. The alternative pathway activates NF-κB-inducing kinase (NIK), only liberating p52/RelB heterodimers. Activation of the classical NF-κB signaling pathway induces phosphorylation of IκB proteins by the IκB kinase (IKK) complex, triggering its polyubiquitination and subsequent degradation by the 26S proteosome. NF-κB signaling is regulated by a negative-feedback loop where IκBα and IκBε (but not IκBβ) expression is induced by NF-κB, allowing precise modulation to counteract rapid environmental changes (54, 55). IκBβ and IκBε help damp oscillations of the IκBα–NF-κB feedback loop (56).
The IκBs have different activities and show differences in their requirement for normal cell function, growth, and development. IκBα is the best understood. NF-κB can activate IκBα expression, which serves a negative-feedback mechanism. Without IκBα, stimulation by TNF-α continues for longer times (57, 58). IκBα knockout (KO) mice have a severe phenotype (59), dying at 7 to 10 days of age with growth failure and severe dermatitis. IκBα-deficient fibroblasts still show a response to TNF-α and maintain NF-κB in the cytoplasm prior to stimulation, suggesting that the other IκBs can compensate for missing IκBα, but NF-κB nuclear localization is prolonged, suggesting that IκBα helps end the NF-κB response. IκBβ's function is less well established. However, IκBβ shows clear differences compared to IκBα in that the IκBβ promoter is not responsive to cell stimulation, and IκBβ binds p65 and c-Rel, but binds p50 less well (54, 60).
IκBε, discovered after IκBα (61, 62), has a physiologic function that is less well understood. IκBε is expressed predominantly in T cells of the thymus, spleen, and lymph nodes (63), so major sites and cell types of IκBε expression coincide with the locations of some of the main reservoirs of HIV latently infected cells. IκBε KO mice show a relatively normal phenotype, in contrast to the lethal phenotype observed with IκBα knockouts (59, 63). IκBε KO mice are identical to wild-type mice in appearance and histology and breed normally. The main differences between wild-type and IκBε KO mice are decreased CD44− CD25+ T cells, and increased production of IL-1α, IL-1β, IL-1 receptor α (IL-1Rα), and IL-6 mRNAs in macrophages. The minimal phenotype of IκBε KO mice suggests that targeting IκBε in an adult animal or human would be relatively safe.
In our work, we found that IκBα and particularly IκBε small interfering RNA (siRNA) activates HIV. Targeting IκBε represents a highly plausible and likely effective new approach to activating HIV expression that may find a role in attacking the reservoir of HIV latently infected cells.
We used tumor necrosis factor alpha (TNF-α) (10 ng/ml; Invitrogen), trichostatin A (TSA; Sigma-Aldrich) (1.5 μM), phorbol myristyl acetate PMA (here, tetradecanoyl phorbol acetate [TPA]) (20 ng/ml; Sigma-Aldrich), and combinations of the agents as positive-control treatments to activate HIV-1 from promonocytic U1 (64) and cells of the T-lymphocyte Jurkat cell-derived line J-Lat 10.6 (65). Human IκBα and IκBε siRNAs and a nonspecific siRNA pool control (D-001206-14) were purchased from Dharmacon, and validated siRNA against IκBβ was acquired from Qiagen. We designed single siRNA sequences using computational algorithms supplied by the Dharmacon, Invitrogen, Qiagen, and Whitehead websites and used siRNAs generated by Integrated DNA Technologies (IDT) (Table 1). The target sequences of the selected siRNAs (catalog numbers in parentheses) are as follows: IκBα (D-004765-02), GGACGAGAAAGATCATTGA; IκBβ (S102654953), CCCGAGGGCGAGGATGAGAAA; and IκBε (D-004766-02), GGAAGGGAAGTTTCAGTAA.
U1 latently infected cells (64) (NIH AIDS Reagent Program), an HIV latently infected cell line often used in HIV latency studies (reviewed in references 66, 67, and 68), were derived from chronically infected U937 cells, a promonocytic cell line isolated from the pleural effusion of a 2-year-old with diffuse histiocytic lymphoma. U1 cells were seeded 48 h prior to transfection at a cell concentration of 0.2 × 106 cells/ml in RPMI (Atlanta Biologicals) with 1% l-glutamine (Invitrogen), 10% fetal bovine serum (FBS) (Invitrogen), 100 IU/ml penicillin, 100 μg/ml streptomycin (Sigma), and 0.4% β-mercaptoethanol (Sigma) in 5% CO2 at 37°C. On the day of transfection, specific siRNAs at the desired concentrations were mixed with 1.5 million U1 cells, resuspended in Nucleofector solution V (Lonza), and transfected with Amaxa program W1. The transfection efficiency, using the siGlo green transfection indicator (Thermo), for U937 was 48%. The concentration of transfected siRNAs ranged from 63 to 2,000 nM to generate a dose-response curve for each IκB siRNA. When combined, the different siRNAs targeting the three IκB genes were transfected at a concentration of 1,000 nM. TNF-α (10 ng/ml) was used as a positive control in siRNA transfections. Once transfected, U1 cells were transferred to a six-well plate that contained 1.4 ml of prewarmed 1× phosphate-buffered saline (PBS) per well. Plates were incubated for 3 h in 5% CO2 at 37°C. Cells were transferred to a 1.5-ml microcentrifuge tube and centrifuged at 1,400 rpm for 7 min. PBS was discarded, and cell pellets were resuspended in 2 ml RPMI–10% FBS and incubated in 6-well plates for 72 h at 37°C. Twenty-four hours posttransfection, 600 μl was used to extract total RNA using the RNeasy minikit (Qiagen), and 72 h posttransfection, cells were pelleted and supernatants were stored at −80°C for further use. The relative replication capacity of the virus was determined by measuring the amount of p24 antigen produced 72 h after transfection using an enzyme-linked immunosorbent assay (ELISA) (PerkinElmer). To determine cell viability, we used an MTS [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium] assay (Promega).
J-Lat 10.6 cells (65) (J-Lat full-length clone 10.6; NIH AIDS Reagent Program) were derived from the Jurkat T cell line by infection with an HIV in which nef sequences were replaced with green fluorescent protein (GFP) coding sequence and latently infected cells were cloned and determined to be latently infected by activation of GFP by HIV activators. Transfection efficiency, determined using siGlo, for Jurkat cells was 54%.
Total extracted RNAs (RNeasy minikit; Qiagen) from 0, 12, 24, 36, 48, 60, and 72 h posttransfection were quantified with a Nanodrop HD-1000 spectrophotometer (Thermo). Eight hundred nanograms of RNA per sample was used for cDNA preparation. The RNA was mixed with 50 μM random hexamers and 10 mM deoxynucleoside triphosphate (dNTP). The samples were incubated at 65°C for 10 min to denature the RNA and immediately chilled on ice. Eight microliters of reverse transcription-PCR (RT-PCR) mixture (1× buffer, 20 mM dithiothreitol [DTT], 200 U SuperScript II reverse transcriptase [Invitrogen], and 20 U SUPERase-In [Invitrogen]) was added to the reaction mixture, and the samples were further incubated for 10 min at 25°C for primer annealing and 50 min at 42°C and 15 min at 70°C for enzyme inactivation. Quantitative real-time RT-PCR (qRT-PCR) assays using the TaqMan system (Applied Biosystems) were used to determine the extent of knockdown of each siRNA relative to the siRNA control pool and the basal concentrations of the different IκB and NF-κB mRNAs (Table 2). We used glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and β-actin as endogenous controls. The final volume was 50 μl (1× TaqMan gene expression assay, 1× TaqMan gene expression master mix, 50 ng cDNA). The PCR conditions were two holds of 2 min at 50°C and 10 min at 95°C followed by 40 cycles of 15 s at 95°C and 1 min at 60°C. The detection was performed with a 7900HT Fast real-time PCR machine (Applied Biosystems).
Time course experiments were performed after transfection. U1 cells were washed twice with 1× PBS and resuspended with lysis buffer. After 30 min on ice, cells were pelleted for 30 min at 13,000 rpm. Supernatants were stored at −80°C until further use. All samples were resuspended in 2× Laemmli buffer, heated at 95°C for 10 min, and normalized to cell count. Samples were fractionated on a 4 to 12% sodium dodecyl sulfate-polyacrylamide gel (NuPAGE; Invitrogen) for 2 h (Novex minicell; Invitrogen), using 1× MOPS (morpholinepropanesulfonic acid)-SDS running buffer, transferred onto a polyvinylidene difluoride (PVDF) membrane (Invitrogen) for 1 h at 90 V (Mini-Protean 3 cell; Bio-Rad), and blocked (1× PBS, 0.1% Tween 20, 5% skim milk) overnight at 4°C. Specific primary antibody antiserum (IκBα, sc-1643; IκBβ, sc-945; and IκBε, sc-7275) (Santa Cruz Biologicals), diluted 1/200, was used to detect the IκB proteins as well as basal protein concentration with a horseradish chemiluminescent system (ECL-Plus; PerkinElmer). The NF-κB subunit basal concentration was determined using primary antibodies to p65 (436700; Invitrogen), p50 (ab72138; Abcam), and p52 (05-361; Millipore). As a loading control, β-actin was used and detected with antibody sc-1615 (Santa Cruz). Protein quantification was performed using ImageJ software (http://rsbweb.nih.gov/ij/).
To determine viral induction using siRNA or the positive-control activating agents, p24 antigen was measured in U1 and J-Lat 10.6 latently infected cell supernatants using an enzyme-linked immunosorbent assay (ELISA) (Perkin-Elmer). Seventy-two hours posttransfection, cells were pelleted at 1,400 rpm for 7 min. Three hundred microliters of supernatant was stored at −80°C until p24 was determined, as described by the manufacturer.
For coimmunoprecipitation (co-IP), a total of 10 × 106 U1 cells were collected and washed twice in chilled 1× PBS at 1,400 rpm for 7 min. Cells were resuspended in 1 ml of lysis buffer (1% Triton X-100, 3 mM EDTA, 2 mM DTT, 1 mg/ml bovine serum albumin [BSA], 20 mM Tris-HCl [pH 7.0], 50 mM NaCl, and protease inhibitor tablet Complete Ultra Mini [Roche]) and incubated on ice for 1 h with mixing every 15 min. Samples were centrifuged at 14,000 rpm for 30 min at 4°C, and supernatants were transferred to a new microcentrifuge tube. Cell lysates were precleared with True Blot anti-rabbit immunoprecipitation (IP) beads (Thermo Scientific) for 15 min at 4°C. We incubated 400 μl of postclear supernatant (4 × 106 cells/reaction) overnight at 4°C with the specific IκB antibodies or isotype controls (5 μg/reaction). After incubation with the primary antisera, the samples were incubated for 4 h at 4°C with True Blot anti-rabbit IP beads and washed twice with lysis buffer and two more times with wash buffer (1/10 lysis buffer). Supernatants were discarded, and beads were frozen and stored at −80°C for further use. All samples were resuspended in 2× Laemmli buffer, heated at 95°C for 10 min, and normalized to cell count. Samples were processed as described above for immunoblot analysis. Specific primary antibodies to IκBα (sc-1643), IκBβ (sc-945), and IκBε (sc-7275) (Santa Cruz), were used for the immunoprecipitation. For the NF-κB immunoblots, the following mouse monoclonal antibodies were used: p65, 436700 (Invitrogen); p50, ab72138 (Abcam); and p52, 05-361 (Millipore).
To determine the different IκB protein half-lives (t1/2), we calculated the first-order decay using a linear regression, t1/2 = log2/k, with k, the slope of the linear regression, as the decay constant. To determine the correlation between p24 fold increase and mRNA knockdown for the different siRNAs targeting IκBα and IκBε, Spearman's ρ, was generated using the statistical package R (http://www.R-project.org). STATA11 (StataCorp) was used to generate Gompertz 4 component nonlinear logistic regression plots.
Since our ultimate goal was to assess the role of individual IκB and NF-κB proteins in HIV activation and, in particular, to determine whether targeting IκB proteins would activate HIV, to better understand how the different components of the NF-κB pathway control HIV expression in U1 cells, we quantified the relative abundance of mRNA and proteins of the different IκBs (Fig. 1A) and NF-κBs (Fig. 1B). Protein and RNA values for the IκBs were normalized to what we determined to be the least abundant IκB, IκBε; protein values for the NF-κB subunits were normalized to what we determined to be the least abundant subunit at the RNA level, RelB. We found that the most abundant IκB species was IκBβ, with RNA present at a 21.4-fold-greater amount than that of IκBε and protein present at a 7.6-fold-greater amount than that of IκBε. IκBα RNA was present at an 8.1-fold-greater amount than IκBε, and protein was present at a 6.5-fold-greater amount than IκBε. The least abundant NF-κB subunit was RelB, followed by RelC, with RNA present at a 1.01-fold-higher level than RelB and protein at a 0.91-fold-lower level than RelB. p52 had an RNA level 146.1-fold higher than RelB and a protein level 92-fold higher than RelB. p50 had an RNA level that was 220.4-fold higher than RelB and a protein level that was 438-fold higher than RelB. p65 had an RNA level that was 883-fold higher than RelB and a protein level that was 540-fold higher than RelB. The very low relative levels of protein and RNA in these cells suggest that RelB and c-Rel play a limited part in NF-κB-mediated HIV activation.
After making estimates of the relative abundance of the different IκBs and NF-κBs, we next determined how the individual NF-κB subunits interact with the different IκBs using a qualitative coimmunoprecipitation (co-IP) assay (Fig. 1C). We found that IκBα strongly interacted with p65, p52, and p50, while IκBβ and IκBε interacted strongly with p65 and interacted less strongly with p52 and p50. If different NF-κB subunits have significantly different effects on one or another biological process, then since IκBs hold different populations of NF-κB subunits, releasing the NF-κB subunits from particular IκBs—say by specifically targeting a particular IκB for degradation—might be likely to have specific biological effects.
One of the key determinants of how activation through the NF-κB pathway initiates HIV replication must involve the kinetics of IκB destruction and replenishment following exposure of the host cell to an activating stimulus. To determine how the different IκB proteins respond when the NF-κB pathway is activated, we treated cells with TNF-α and assayed at serial times for IκB protein abundance using immunoblots (Fig. 1D). The three canonical IκB proteins showed substantially different decay and replenishment kinetics, as expected. The results showed that IκBα showed a steep decline, becoming undetectable by 5 min following TNF-α treatment. IκBβ and IκBε showed a more moderate reduction. IκBα recovered faster than IκBβ and IκBε. By 90 min after TNF-α treatment, IκBα had increased to more than double its initial amount. Both IκBβ and IκBε showed similar kinetics after TNF-α treatment, both showing a much smaller reduction in abundance than IκBα, reaching a nadir at 15 min and then increasing slowly in abundance, but not returning to the initial amounts by 120 min. These results indicate that IκBα has a much faster feedback mechanism than IκBβ or IκBε. These differences in kinetics might result from physiologically different roles of the three IκBs. IκBβ and IκBε recover more slowly than IκBα, so the effects resulting from their destruction would last longer. The co-IP results show that the pool of NF-κB dimers binding to the three IκBs is different, reinforcing the model that each IκB has a specific role in holding distinct populations of NF-κB subunits inactive in the cytoplasm or releasing distinct subunits following activation. Physiologically, IκBα may therefore play a more important role in short-term changes, while IκBβ and IκBε may influence more long-term processes in these cells.
Treatment with the powerful inducer TNF-α strongly affected all three canonical IκB proteins, as shown in Fig. 1D. To assess the role that the different IκB proteins and their bound NF-κB subunits play in HIV activation, we knocked down the different IκB mRNAs with siRNAs specific for each IκB. In a preliminary experiment, we used a pool of four siRNAs against IκBα and IκBε purchased from Dharmacon and a single siRNA targeting IκBβ purchased from Qiagen. The IκBβ siRNA knockdown reached a level of 78%, while IκBα and IκBε knockdowns reached levels of 49% and 35% (not shown). To increase the knockdown effect of IκBα and IκBε, we assayed for knockdown produced by the individual siRNAs included in the pool and generated new ones using computational algorithms available on the Invitrogen, Qiagen, and Whitehead websites. For IκBα, we tested three more siRNAs and tested eight more for IκBε, screening a total of seven siRNAs for IκBα and 12 siRNAs for IκBε (Table 1 and Fig. 2A and andB).B). We determined mRNA knockdown in U1 cells at 24 h posttransfection using qRT-PCR. The siRNAs that achieved the best knockdown were no. 7 for IκBα and no. 9 for IκBε. We used those siRNAs for subsequent experiments. Both siIκBα and siIκBε siRNAs targeted the 3′ (UTR) region of the specific mRNAs, while the siRNA against IκBβ was directed to the coding region (Fig. 3A). None of the siRNAs had cytotoxic effects: there was no decrease in cell viability following and IκB siRNA transfection over 72 h (Fig. 3B). We determined the knockdown kinetics of the IκB mRNAs using qRT-PCR (Fig. 3C). mRNA knockdown peaked at 12 h posttransfection and slowly recovered, returning to about 70% of baseline mRNA abundance by 72 h. We calculated the half-lives of the different IκB proteins (Fig. 3D) based upon abundances determined by immunoblotting, which also enabled us to establish on- and off-target effects for the different IκB siRNAs (Fig. 3E). The protein kinetics, determined by immunoblotting, showed patterns that differ from the kinetics observed for IκB RNA (Fig. 3E). IκBα siRNA produced a rapid decay in IκBα protein levels, with levels reaching less than 20% of their initial abundance by 12 h. IκBβ siRNA decreased IκBβ protein to levels that were also about 20% of baseline, but for IκBβ, the nadir was achieved after 48 h. IκBε siRNA also achieved a nadir at about 48 h, but the IκBε siRNA did not reduce IκBε protein levels to the extent IκBα, nor did IκBβ siRNA. None of the siRNAs had significant off-target effects at the protein level: that is, each IκB siRNA affected only the RNA and protein levels of the targeted IκB, with only minimal effects on the other IκBs, making it unlikely that the activation observed would be due to the knockdown of a nontargeted IκB. The data sets made it possible to estimate the half-lives of the different IκBs in the U1 cells (Fig. 3D). We found that IκBα had the shortest half-life, 5 h, while we found that IκBβ had a half-life of 17 h and IκBε had a half-life of 33 h. Thus, while the knockdown of IκBε was not as quantitatively large as knockdown of IκBα, IκBε knockdown persisted much longer. These results demonstrate that our siRNAs can specifically knock down individual IκB proteins and that IκB protein decay and replenishment kinetics following siRNA knockdown are comparable to those observed following treatment with TNF-α (Fig. 1C). This suggests that the IκB siRNA knockdown produced changes in IκB protein levels that, while not as quantitatively extensive as those seen with TNF-α, nevertheless shared the basic kinetic characteristics seen with TNF-α treatment. The longer half-lives and slower replenishment kinetics of the IκBβ and IκBε proteins reinforced the hypothesis that there might be significant differences in the biological effects of knocking down the different IκB proteins, suggested by the finding that the different IκBs hold different populations of NF-κB subunits. In addition to the release of different populations of NF-κB subunits, the differences in the kinetics of knocking down the individual IκB proteins further suggest that targeting individual IκB proteins may have different biological effects. The specificity observed in knocking down the individual IκB proteins indicates that the specific IκB siRNAs could be used to dissect out the different roles that the IκB proteins and the NF-κB subunits bound to those IκBs play in biological processes, including HIV activation.
Having established that our siRNAs could knock down individual IκBs at the RNA and protein levels and having characterized the performance characteristics of those siRNAs, we could then use those reagents to study the involvement of individual IκBs and the NF-κB subunits held by those IκBs in the maintenance of HIV latency and HIV reactivation. We performed a series of transfections to determine the activation dose response with specific IκB siRNAs (Fig. 4A). For each concentration, we determined the extent of mRNA knockdown as well as the increase in cell supernatant HIV p24 antigen, determining the fold increase seen with the transfection, normalized to the siRNA control. There appeared to be a clear dose response associating HIV activation with the extent of the IκB knockdown for both IκBα and IκBε siRNAs (Fig. 4B). Although IκBβ siRNA achieved the greatest knockdown, reaching almost 75% of baseline, IκBβ siRNA produced no viral reactivation. However, IκBα and IκBε siRNAs both reactivated HIV from latency. With IκBα siRNA, knockdown reached levels of ~60%, which produced substantial activation, with p24 levels increasing by >5-fold. Surprisingly, the IκBε siRNA had a larger effect. With the IκBε siRNA, we were able to achieve knockdown levels of about 75%, which produced activation of more than a 10-fold increase in HIV p24 antigen in the cell supernatant. At equivalent achieved levels of knockdown (up to 60%), IκBε siRNA appeared to be a more potent activator, which was unexpected given the relatively minor role that IκBε had been thought to play in the regulation of NF-κB-controlled gene expression and the relatively smaller amounts of IκBε present in the cell. A threshold effect was also apparent for IκBε knockdown, with a large increase in HIV activation when IκBε knockdown exceeded about the 50% level.
To show that knocking down IκBα or IκBε specifically produced HIV reactivation, and to explore whether some nonspecific effect of the siRNAs studied in the experiments of Fig. 3 and and44 might have led to HIV activation, we determined the HIV reactivation capabilities of the different IκBα and IκBε siRNAs we initially tested for IκB mRNA knockdown ability (Fig. 2). These two sets of siRNAs targeted diverse regions of the mRNAs and exhibited a spectrum of both IκB knockdown activity and, when tested for activation, HIV activation ability (Fig. 5). We observed a clear correlation by Spearman's ρ (IκBα, ρ = 0.96; and IκBε, ρ = 0.93) between the ability of the siRNAs to knock down IκB mRNA and activate HIV (Fig. 5B and andD),D), although the two IκBε siRNAs that were most effective at knocking down IκBε mRNA had a somewhat disproportional effect on HIV activation, reinforcing the findings of the dose-response study (Fig. 4B), in which we observed a threshold effect for increased HIV activation. The findings that many different siRNAs targeting many different regions of the mRNA were effective in both knocking down mRNA and activating HIV and that the HIV-activating activity correlated with the IκBε knockdown activity argue that the HIV activation was caused by the IκB mRNA knockdown and not primarily by some other nonspecific or off-target effects of the siRNAs.
After establishing that both IκBα and IκBε siRNAs effectively activate HIV in latently infected cells, we wanted to determine whether the IκBα and IκBε siRNAs would have additive effects when transfected together and to compare the effects of the IκBα and IκBε siRNAs with those of conventional HIV activators. We also wanted to confirm that the HIV-activating ability of the IκB siRNAs was not limited to a single HIV latently infected cell line or to only monocytic cells. We therefore transfected the IκBα and IκBε siRNAs alone and in pairwise combinations into both the U1 monocytic and Jurkat-derived J-Lat 10.6 lymphocytic cell lines and also treated the cells with positive-control activators, TNF-α, TSA, and TPA, which activate HIV through both the NF-κB pathway and by other mechanisms (Fig. 6). TNF-α produced the greatest activation (325- and 80-fold increases in U1 and J-Lat 10.6 cells, respectively) followed by TPA (100- and 30-fold increases) and TSA (10- and 2-fold increases). The experiments showed that IκBα and IκBε activate HIV in both the monocytic U1 cells, as seen in the previous figures and in the lymphocytic J-Lat 10.6 cells. While the HIV-activating abilities of the IκBα and IκBε siRNAs were substantially less than those of the well-known highly effective (and toxic) HIV activators TNF-α, and TPA, the IκBα and IκBε siRNAs activated HIV about as well as the histone deacetylase inhibitor TSA, a representative of a class of compounds that has attracted wide interest as potential agents to attack and purge the reservoir of HIV latently infected cells. The results confirmed that IκBβ siRNA had no effect on viral reactivation in either cell line when transfected into the cells alone and did not add to the activating effect of either IκBα or IκBε siRNAs when transfected into the cells in combination with those siRNAs. Interestingly, IκBα and IκBε siRNAs showed additive effects when both IκB siRNAs were transfected into both cell lines.
NF-κB was recognized very early on as a key activator of HIV expression, so when interest began to build in finding ways to activate HIV as a component of a strategy to deplete the reservoir of HIV latently infected cells, much research activity initially focused on HIV activation via NF-κB pathways. However, recognition that the available NF-κB-directed activators were toxic and relatively nonspecific, together with an increasing appreciation for the importance of other levels of HIV expression control, such as epigenetic modifications, led to a shift away from studies of the NF-κB pathway as a likely effective approach for HIV activation and reservoir depletion. Our finding that IκBε siRNA can activate HIV suggests not only that IκBε might be a reasonably specific and effective target for HIV activation, but also more broadly that specificity and selectivity for HIV activation may yet be achieved by selectively targeting specific elements of NF-κB pathways. While our IκBε siRNAs activated HIV reasonably well, particularly compared to some of the activation strategies that are currently the subject of considerable research, like HDAC inhibitors (Fig. 6), it would certainly be better if IκBε siRNAs activated HIV even more effectively. Since we observed transfection efficiencies of only ~50%, employing more effective delivery methods should substantially increase HIV activation by IκBε siRNAs. However, effective HIV activation even at relatively low levels of transfection efficiency also suggests that IκBε knockdown might, in fact, prove to be a reasonable clinical strategy, since in vivo clinical siRNA delivery can be challenging.
Our finding that IκBε siRNA can activate HIV was unexpected. IκBε was thought to function mostly to modulate and dampen potentially extreme autoregulatory feedback loops created by IκBα's strong response to NF-κB (56). The fact that IκBε comprises a small minority of the IκBs within a cell initially suggested that targeting IκBε alone would not have much of any effect at all, an idea reinforced by the observation that IκBε KO mice have a minimal phenotype (63), in contrast to the lethal phenotype observed with IκBα KO mice (59).
Why would targeting IκBε have such a disproportionate effect on HIV activation? Our data suggest a few possibilities. First, the relatively low abundance of IκBε RNA and protein may mean that given amounts of siRNA can lead to a greater fractional knockdown of IκBε, a hypothesis supported by the observation that there appears to be a threshold effect in IκBε knockdown, with greater increases in HIV activation observed with knockdown levels beyond 50%. Second, the kinetics of IκBε destruction and replenishment may make IκBε knockdown more effective at activating HIV. Turning on the “lytic switch” for HIV replication requires essentially two hits: an increase in initial basal transcription (“hit 1”), which must be sustained long enough for some Tat to be translated in the cytoplasm and then transported back to the nucleus to bind TAR on the nascent HIV transcript (“hit 2”), which results in the massive increase in expression that initiates lytic replication. The prolonged IκBε kinetics that we observed may favor a relatively prolonged activation of basal HIV expression produced by the NF-κB subunits released by IκBε, an activation prolonged sufficiently to switch HIV onto the lytic replication pathway. This kinetic argument may also help explain the synergy observed when cells were transfected with IκBα plus IκBε siRNA. Since we observed that IκBα abundance decreased early, while IκBε abundance decreased later after siRNA transfection, the combination of both siRNAs should produce a more prolonged stimulus of the HIV LTR. Third, some of the effect may be due to the specific NF-κB subunits released from IκBε. We found that IκBε holds proportionately more p65 than p52 or p50, compared to IκBα. The inhibitory NF-κB dimer bound to the quiescent HIV LTR is a p50 homodimer (69). The principal activating NF-κB dimer for the LTR is a p50/p65 heterodimer. If knocking down IκBε predominantly releases p65 for transit to the nucleus, then the p65 released from IκBε may convert the inhibitory p50 homodimers into activating p50/p65 heterodimers. While these explanations are plausible, a definitive determination of the explanation will have to await further experiments.
We found that both IκBε and IκBβ hold disproportionately more p65 and p50, and our siRNAs were at least as effective at knocking down IκBβ, when normalized to the baseline amount of IκB, as they were at knocking down IκBε: so why then didn't the IκBβ siRNA also activate HIV? One possible explanation lies in the relative abundance of the different IκBs. IκBβ was more than 20-fold more abundant than IκBε. If there is some absolute threshold effect—some threshold below which the IκB has to be knocked down to activate HIV—then it may simply not be possible for the siRNA to knock down the IκBβ below that level. In addition, IκBβ traffics between the nucleus and cytoplasm differently from other IκBs, or at least differently from IκBα (69, 70; reviewed in reference 71). IκBα has a nuclear export signal (NES), while IκBβ does not, and unlike IκBα, IκBβ can mask the nuclear localization signal (NLS) of bound NF-κB (69), so IκBβ does not shuttle between nucleus and cytoplasm as IκBα does (70), and IκBβ, once degraded, is first resynthesized as a hypophosphorylated form (60, 72). If one of the mechanisms for HIV activation by IκBα and IκBε siRNAs, but not IκBβ siRNA, is the conversion of inhibitory p50 homodimers by relatively larger amounts of p65 released by the IκBs, the lack of shuttling by IκBβ could also help account for the lack of activation observed with IκBβ siRNA.
While HIV activation via the NF-κB pathway initiated by IκBε siRNA is attractive, it is unlikely to be the one and only HIV activation approach needed to effectively attack all of the HIV latently infected cells within the many disparate reservoirs that exist within an HIV-infected person. Indeed, other activators that work via the NF-κB system and activators that employ other mechanisms, such as activators that act epigenetically, activate HIV in some model systems, but not in others (22, 73). Some activators can even exhibit antagonistic effects for HIV activation in some systems and additive or synergistic effects in others (22). Even NF-κB-mediated activation sometimes requires help. In some resting CD4 memory T cells, NF-κB is not sufficient for activation; activation requires both NF-κB and induction of P-TEFb (74–76).
While several different modes of attack may be required to effectively activate HIV replication for most or all HIV latently infected cells within an HIV-infected patient, our findings suggest that activation through the NF-κB pathway, achieved by specific targeting of IκBε, may make a helpful contribution. The specificity of IκBε siRNA for HIV activation and the potentially high therapeutic index of IκBε-active agents may also be enhanced by the ability to specifically target siRNAs to particular cell types. Microparticles and micelles complexed with single-chain variable fragments (scFvs) (77, 78) and aptamers (79) and other approaches are attracting increased attention as ways of directing specific siRNAs to the desired cell types (reviewed in references 80 and 81). Specific activation of HIV latently infected cells may therefore be provided by both the relative specificity for HIV activation of IκBε siRNA plus the potential to target IκBε siRNA specifically to cells likely to harbor a latent HIV provirus.
We thank Tae-Wook Chun and Michael Bukrinsky for helpful discussion. IDT provided in-kind support with siRNAs and other oligonucleotides. The cell lines U1 and J-Lat 10.6 were obtained through the NIH AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH, through donations by Tom Folks and Eric Verdin, respectively.
This work was supported in part by the District of Columbia Developmental Center for AIDS Research (P30AI087714).
Published ahead of print 30 January 2013