|Home | About | Journals | Submit | Contact Us | Français|
A human immunodeficiency virus type 1 (HIV-1)-based vector expressing an antisense RNA directed against HIV-1 is currently in clinical trials. This vector has shown a remarkable ability to inhibit HIV-1 replication, in spite of the fact that therapeutic use of unmodified antisense RNAs has generally been disappointing. To further analyze the basis for this, we examined the effects of different plasmid-based HIV-1 long-terminal-repeat-driven constructs expressing antisense RNA to the same target region in HIV-1 but containing different export elements. Two of these vectors were designed to express antisense RNA containing either a Rev response element (RRE) or a Mason-Pfizer monkey virus (MPMV) constitutive transport element (CTE). In the third vector, no specific transport element was provided. Efficient inhibition of HIV-1 virus production was obtained with the RRE-driven antisense RNA. This construct also efficiently inhibited p24 production from a pNL4-3 provirus that used the MPMV CTE for RNA export. In contrast, little inhibition was observed with the constructs lacking an RRE. Furthermore, when the RRE-driven antisense RNA was redirected to the Tap/Nxf1 pathway, utilized by the MPMV CTE, through the expression of a RevM10-Tap fusion protein, the efficiency of antisense inhibition was greatly reduced. These results indicate that efficient inhibition requires trafficking of the antisense RNA through the Rev/RRE pathway. Mechanistic studies indicated that the Rev/RRE-mediated inhibition did not involve either nuclear retention or degradation of target mRNA, since target RNA was found to export and associate normally with polyribosomes. However, protein levels were significantly reduced. Taken together, our results suggest a new mechanism for antisense inhibition of HIV mediated by Rev/RRE.
Human immunodeficiency virus type 1 (HIV-1) is a complex retrovirus that encodes 15 viral proteins from its ~9-kb RNA genome (16). The virus uses a variety of strategies in order to optimize the utilization of the coding capacity of its genome (44). Alternative splicing gives rise to many different mRNAs that can be categorized into three distinct classes: unspliced, singly spliced, and multiply spliced RNA (58). The unspliced and singly spliced mRNAs retain introns. Normally, intron-containing RNAs are retained in the nucleus of the host cell (8, 67). However, HIV-1 has evolved a specific mechanism to overcome this problem, utilizing the Rev regulatory protein in conjunction with a structured cis-acting RNA element known as the Rev response element (RRE) (for a review, see reference 22). Rev is a small RNA-binding phosphoprotein encoded by a multiply spliced viral RNA (27). Rev localizes to the nucleus, where it binds the RRE and recruits the Crm1 nuclear export factor to help overcome nuclear retention (14, 15, 23, 28, 33, 47).
An alternative mechanism to overcome nuclear retention of mRNA with retained introns relies entirely on cellular proteins in conjunction with a structured RNA element first identified in Mason-Pfizer Monkey Virus (MPMV), known as the constitutive transport element (CTE) (4, 13). The CTE functions together with the cellular Tap/Nxf1 and Nxt1 proteins to overcome nuclear retention and enable export of mRNA with retained introns (20, 21). In addition to nuclear export, the proteins also enhance translation from CTE-containing RNAs (30). Experiments using leptomycin B, a Crm1-specific inhibitor, demonstrated that Tap/Nxf1-mediated nuclear export is Crm1 independent (53). Thus, the Rev/RRE and CTE represent two different trafficking pathways for export. However, despite this, the CTE can functionally replace Rev/RRE during HIV-1 replication (4, 79).
Current therapeutic strategies mainly use highly active antiretroviral therapy to control HIV-1 infection, typically using a triple-drug cocktail targeting reverse transcription and proteolytic processing (49, 56). In spite of the successes using highly active antiretroviral therapy, drug resistance remains a problem, requiring changes in drugs and treatment strategies. All currently approved drugs for HIV-1 treatment target proteins in order to prevent replication. Another avenue for controlling HIV-1 infection would be to target the viral RNA during replication. Although promising, no therapies targeting the HIV-1 viral RNA have been approved for wide clinical use.
Several approaches are currently being considered to target RNA directly, including the use of ribozymes, small interfering RNA (siRNA), and vector-delivered antisense RNA (11, 63). Ribozymes are catalytically active RNAs that recognize and cleave specific sequences within a target RNA. Several ribozyme-based strategies targeting HIV-1 sequences are in development, but none have achieved complete inhibition of HIV-1 replication (18, 32, 59, 72, 76). siRNA has been used successfully in the laboratory for a decade and is currently being tested as a viable therapeutic. Binding of siRNA to complementary mRNA leads to translation inhibition or degradation (10, 41, 48, 75). A drawback to the approaches outlined above is that they rely on short complementary sequences for inhibition. Resistance to siRNA has already been demonstrated, although the use of longer sequences has demonstrated greater tolerance for sequence changes (2, 34, 66). As a result of these problems, multiple combined RNA-targeting strategies have been suggested as essential (63).
VIRxSYS has recently developed a lentivirus-based antisense RNA therapeutic to prevent HIV-1 replication (42) that has shown promise in a phase I clinical trial and is currently undergoing phase II testing (39). The vector delivering the antisense RNA is an HIV-1 vector, where the antisense RNA is expressed from an HIV-1 long terminal repeat (LTR) promoter. The RNA that is expressed includes a 937-nucleotide (nt) sequence from the env region of HIV-1 in an antisense orientation, as well as an RRE. A previous study demonstrated that HIV-1 was unable to easily develop resistance to inhibition by this RNA (42). However, after long-term passage of HIV-1 in antisense-expressing cells, virus particles were recovered. RNAs recovered from these particles were shown to contain multiple mutations in the env target region, making the virus noninfectious. Most of these mutations were A-to-G changes in the RNA, consistent with previous editing of the RNA through “adenosine deaminase that acts on RNA” (ADAR) activity (1). ADAR deaminates adenosines to inosine, which would be recognized by reverse transcriptase as guanosine, leading to subsequent A-to-G mutations in the genome. RNAs containing multiple inosines are typically retained in the nucleus and degraded (36, 57). Thus, the finding of multiple mutations in HIV-1 RNA, as a possible result of editing by ADAR, was an unexpected finding. However, previous work by Zhang and Carmichael (78), using a Xenopus oocyte model system, has demonstrated that hyperedited RNA containing an RRE was exported to the cytoplasm, efficiently overcoming nuclear retention, if Rev was provided. Together, these results suggest the possibility that Rev/RRE trafficking contributes to the observed efficiency of the antisense RNA approach.
In the present study, we specifically examine whether efficient inhibition of HIV-1 requires the antisense RNA to be trafficked through the Rev/RRE pathway. To do this, we constructed HIV-1-based plasmids that express antisense RNA containing either an RRE, CTE, or no transport element. The effect of expression of these RNAs on HIV-1 expression was then examined in experiments that include studies on the effects of antisense RNA on target RNA trafficking and expression.
To facilitate identification, all plasmids used in the present study were indexed as numbers in the form of pHRXXXX. Sequences from three different pNL4-3-derived proviral clones (pHR1272, pHR1498, and pHR2772) were used to make constructs expressing HIV-1 antisense RNA. The resulting constructs contain the 5′ and 3′ LTR, as well as the HIV-1 packaging signal. pNL4-3(nef−) (pHR1272) contains a 150-bp deletion at the start of nef that prevents Nef expression. pNL4-3(rev−)(RRE−)(CTE) (pHR1498) contains a rev with two point mutations that prevent Rev expression, as well as multiple third-base mutations in the RRE that prevents its function, but maintains the Env amino acid sequence (79). It also contains a copy of the MPMV CTE cloned into nef. pNL4-3(RRE−)(nef−) (pHR2772) contains the third-base mutations in the RRE, as well as the 150-bp deletion in nef that is also present in pHR1272. Antisense constructs were assembled by digesting pHR1272, pHR1498, or pHR2772 with NsiI (pNL4-3, nt 1249) and repair with the Klenow fragment, followed by digestion with NheI (pNL4-3, nt 7252). The resulting plasmid backbones lacked HIV-1 sequences from pNL4-3, nt 1249 to 7252, which contained gag/pol, vif, vpu, and vpr, as well as the first exon of tat and rev and regions of env 5′ to the RRE. A PCR-generated fragment corresponding to the same 937-bp region of env (pNL4-3, nt 6602 to 7538) that was expressed in original antisense vector (42) was then inserted into these backbones. The primers used to generate this fragment incorporated stop codons in the 5′ end of the fragment to prevent any protein expression from the insert region and an XbaI site in the 3′ end for cloning. The fragment was ligated to the backbone sequences in the antisense orientation, resulting in the following plasmids: RRE-driven antisense (pHR3476) derived from pHR1272, CTE-driven antisense (pHR3477) derived from pHR1498, and no-element antisense (pHR3478) derived from pHR2772 (see Fig. Fig.1A).1A). As a control, we also made a similar construct, RRE-driven sense (pHR3473), containing the RRE with the insert in the sense orientation.
A similar strategy was used to generate the antisense construct targeting green fluorescent protein (GFP), except that the primers amplified a 721-nt fragment containing the GFP open reading frame (ORF; Clontech pEGFP-N1, pHR1976). The primers incorporated either a SpeI or a NotI restriction site. The PCR product and backbone vectors were digested with SpeI and NotI, and the GFP fragment was used to replace the env sequence in pHR3476 and pHR3478 to generate the RRE-driven (pHR3755) and no-element (pHR3757) GFP antisense constructs.
All of the pNL4-3-derived proviruses that were used here contained either one or two point mutations in the Rev ORF (an AUG mutation and a nonsense mutation at Rev amino acid 12) that prevent Rev expression. pNL4-3(Rev−) (pHR1146) contains a functional RRE with the nonsense mutation in place of amino acid 12 (4). pNL4-3 (Rev−)(RRE−)(CTE) (pHR1371) contains multiple third-base mutations in the RRE that prevent its function but has the MPMV CTE cloned into nef. In some experiments an HXB2-derived provirus containing mutations in the myristylation site, frameshift region, and protease active site (pHR1320) was utilized. This was obtained from Casey Morrow (University of Alabama) (54). CMV-Rev (pHR30), CMV-Tat (pHR136), CMV-RevM10Tap (pHR2155), and CMV-Nxt1 (pHR2415), which express Rev, Tat, RevM10-Tap fusion protein, and Nxt1, respectively, have been described previously (21, 69). pCMV (pHR16) is the empty backbone vector for these expression constructs. The CMV-β-globin construct (pHR2643) expressing mouse β-globin RNA was obtained from Lynne Maquat (University of Rochester) (77). Further details about all of the proviral clones and plasmids used in the present study are available upon request.
The LTR-driven GFP construct was created by digesting pEGFP-N1 (Clontech, pHR1976) with HindIII and NotI. A fragment containing the EGFP ORF was then cloned into pBluescript that had also been digested with HindIII and NotI, resulting in pBS-GFP (pHR3753). To generate pLTR-GFP (pHR3754), a fragment that contained the HIV-1 5′ LTR was then inserted into pBS-GFP. This was done by first digesting pNL4-3 (pHR1145) with PstI, followed by repair with T4 DNA polymerase and further digestion with HindIII to obtain the LTR fragment. pBS-GFP was then digested with ClaI, repaired with T4 DNA polymerase, and then further digested with HindIII. This facilitated the cloning of the LTR fragment directly between the blunt ended ClaI site and the HindIII site of pBS-GFP.
293T/17 cells were maintained in Iscove minimal essential medium supplemented with 10% bovine calf serum and 0.1% gentamicin. Transient transfections were performed by using the calcium phosphate method (19). For analysis of RNA stability, cells were treated with 5 μg of actinomycin D (Sigma, St. Louis, MO)/ml for 0, 3, or 6 h (40), followed by the harvest of total RNA using TriReagent (Molecular Research Center, Cincinnati, OH) according to the manufacturer's instructions.
Supernatants from transfected cells were collected 48 h after transfection and diluted 100- to 1,000-fold before application to the enzyme-linked immunosorbent assay (ELISA) plate. The ELISA was performed using a p24 monoclonal antibody (catalog no. 1513) and pooled human anti-HIV-1 immunoglobulin G (catalog no. 3957) that were obtained from the AIDS Research and Reference Reagent Program, following a protocol developed by Bruce Chesebro (National Institute of Allergy and Infectious Diseases, Rocky Mountain Laboratories) (73).
Methods for cytoplasmic RNA extraction, poly(A) mRNA selection, and Northern analysis have been described previously (23, 24). Total RNA was extracted by using TriReagent according to the manufacturer's instructions. Polyribosome analysis was performed as described in detail by Bor et al. (3). Specific radiolabeled DNA probes to detect GagPol and antisense RNA were generated by PCR incorporating [α-32P]dCTP (Perkin-Elmer Life Sciences, Boston, MA). The probe template was derived from the pNL4-3 gag region and recognizes an ~500-bp sequence that is present in the 5′ portion of both the GagPol viral RNA and the antisense RNA. The probe used to detect mouse β-globin mRNA was generated by PCR incorporating [α-32P]dCTP, using a template corresponding to exon 3 of the β-globin mRNA. Visualization and quantitation was performed by using a Molecular Dynamics PhosphorImager and ImageQuant software.
Proteins were separated by using sodium dodecyl sulfate (SDS)-8% polyacrylamide gel electrophoresis (acrylamide-bisacrylamide, 37.5:1), and Western blot analysis was performed essentially as previously described (23). Proteins were transferred to Immobilon-FL membrane by electrotransfer and blocked using 5% milk in phosphate-buffered saline. Mouse monoclonal antibodies were used to detect Pr160 (catalog no. 1513; AIDS Research and Reference Reagent Program) and Nef (provided by Bernhard Meier) (43). A commercial polyclonal antibody to human β-tubulin (Abcam, Cambridge, MA) was used detect cellular β-tubulin as a loading control. After incubation with secondary antibody (IRDye800 anti-mouse or Alexa Fluor 680 anti-rabbit antibodies; Rockland Immunochemicals, Gilbertsville, PA), blots were visualized by using the Odyssey Infrared Imaging System (Li-Cor Biosciences, Lincoln, NE) and analyzed by using the Odyssey software package.
To test whether the Rev/RRE pathway is important for efficient antisense inhibition of HIV-1, we constructed three nearly identical plasmids that express the same antisense RNA targeting the env region of HIV-1. The plasmids differ only in the export element they contain (RRE, CTE, or no specific export element; see Materials and Methods). These plasmids were all based on sequences from the pNL4-3 provirus and contain the 5′ and 3′ LTR, a portion of gag sequence, and a fragment from the env gene in the antisense orientation which targets a 937-bp region of pNL4-3. This is the same region that was targeted by the previously described lentivirus vector (42). The no-element plasmid is identical to the CTE plasmid, except that no CTE was inserted. As a control, we also made a similar RRE-containing construct that had the target sequence in the sense orientation. A schematic diagram of these plasmids is shown in Fig. Fig.1A1A.
We first compared the level of RNA expression from the three different antisense plasmids in the absence of an HIV-1 target. To do this, 293T cells were transfected in triplicate with the antisense constructs containing no element, the CTE, or the RRE together with CMV promoter-driven plasmids expressing the HIV-1 Tat and Rev proteins. Tat protein is essential for high expression from the HIV-1 LTR promoter, and Rev was provided to ensure good expression from the plasmid containing the RRE. A CMV-β-globin plasmid was also cotransfected to serve as a normalization control. After 48 h, total RNA was harvested, and poly(A)+ RNA was analyzed on Northern blots (Fig. (Fig.1B).1B). The experiment was performed in triplicate. Quantitation of the blots, with normalization to the β-globin signal, demonstrated that the no-element construct and the RRE-containing antisense RNA were expressed at similar levels, while the levels of the CTE-containing RNA were two- to threefold higher (Fig. (Fig.1C1C).
In order to assess the ability of the antisense RNA to inhibit HIV-1 production, we cotransfected 293T cells with a pNL4-3 provirus containing a nonfunctional Rev gene, a plasmid expressing functional Rev (pCMV-Rev), and the indicated amounts of antisense plasmids. This allowed us to determine a dose-response range for antisense inhibition. In this experiment, we used a Rev− provirus and supplied Rev in trans to allow direct comparison with the Rev− provirus containing the CTE that was used as a target in later experiments. Virus production was measured by analyzing the levels of p24 in the medium supernatants of the transfected cells at 48 h posttransfection. The experiment was performed in triplicate.
As shown in Fig. Fig.2A2A (where levels of p24 are shown relative to control cultures in the absence of antisense RNA), the RRE-driven antisense construct efficiently inhibited HIV-1 production even at a very low target/antisense molar ratio (5:1). At a ratio of 5:1 the levels of p24 were only ca. 20% of that in control cultures, at a ratio of 1:1 they were only ca. 3%, and at a ratio of 1:5 they were only ca. 0.3%. In contrast, the CTE-driven antisense did not significantly inhibit particle production from the provirus, even at a 1:5 molar ratio, despite the fact it expressed more antisense RNA per microgram of plasmid (see Fig. Fig.1C).1C). The antisense plasmid with no export element also failed to inhibit particle production. These data indicate that the presence of an RRE in the antisense RNA is essential for efficient antisense inhibition and that the CTE fails to substitute for the RRE.
Since the target virus in Fig. Fig.2A2A utilized the RRE and was Rev dependent, the efficient inhibition observed with the RRE antisense plasmid compared to the other antisense plasmids could simply reflect cotargeting of the target and antisense RNA to the same cellular compartment, making antisense-target RNA association more efficient. To further address this, we analyzed the effects of the antisense constructs on particle production from a provirus that was Rev-independent and utilized a different export pathway. This provirus carries mutations in rev that make it nonfunctional and also contains a nonfunctional RRE with synonymous mutations that do not alter the overlapping Env protein sequence. To overcome the block to RNA export that would otherwise be present, the provirus has the MPMV CTE cloned into the nef region. This allows full-length and singly spliced viral mRNA to be exported via the MPMV CTE pathway. The provirus has previously been shown to be replication competent (4).
We cotransfected the CTE-containing provirus with different amounts of the RRE-driven antisense plasmid (with or without pCMV-Rev) or with different amounts of CTE-driven antisense plasmid. As before, this experiment was performed in triplicate. As can be seen in Fig. Fig.2B,2B, the CTE-driven antisense was still rather inefficient at inhibiting p24 production, although significant inhibition (ca. 80%) could be seen at higher concentrations (Fig. (Fig.2B).2B). This suggests that cotargeting is a factor for the CTE-driven antisense RNA at these higher concentrations, since the antisense RNA would now be expected to traffic along the same pathway as the provirus, in contrast to the experiment depicted in Fig. Fig.2A.2A. However, surprisingly, the RRE-driven antisense was still much more efficient in inhibiting the CTE-driven provirus than was the CTE-antisense RNA, when Rev was supplied (square symbols). This cannot be attributed to cotargeting, since the antisense and target RNA utilized different export pathways. However, antisense inhibition by the RRE-driven antisense was less efficient on the CTE-virus than the effect on the RRE-virus (compare Fig. 2A and B, square symbols), suggesting that cotargeting may also play a role in this case. No significant inhibition of virus production was observed with the RRE-driven antisense when Rev was not supplied (triangle symbols). These results demonstrate that both Rev and the RRE are important for efficient inhibition by the RRE-containing antisense even when the RNA produced from the target provirus does not utilize the Rev pathway.
Previously, we showed that an RRE-containing RNA can be redirected from the Rev pathway to the Tap/Nxf1pathway used by the MPMV CTE when a RevM10-Tap fusion protein in conjunction with the Tap cofactor Nxt1 is utilized in place of Rev (21). The RevM10 protein is a Rev mutant that contains a mutation in the nuclear export signal but maintains the ability to bind to the RRE (46). This allows the RevM10-Tap protein to export the RRE-RNA through the Tap/Nxf1 portion of the protein using the CTE pathway. RevM10-Tap-mediated RNA export is not sensitive to leptomycin B inhibition (21), confirming that it does not use the Crm1 export pathway that is essential in the case of Rev/RRE (15, 74).
To further test the hypothesis that the Rev/RRE pathway is essential for efficient antisense inhibition, we cotransfected the pNL4-3 proviral clone lacking a functional rev gene, plasmids expressing RRE- or CTE-driven antisense, a plasmid expressing the RevM10-Tap fusion protein, and a plasmid expressing the Nxt1 protein. As can be seen in Fig. Fig.2C,2C, this significantly reduced the inhibition that was obtained with RRE antisense, especially at lower antisense/target ratios (compare to Fig. Fig.2A,2A, square symbols). Furthermore, in this experiment the inhibition curves obtained with the CTE- and RRE-driven antisense constructs were virtually overlapping, with inhibition occurring only at the highest ratio of antisense. This result is not unexpected, since both of the antisense RNAs, as well as the target RNA, now use the Tap/Nxf1 pathway. Taken together, these results clearly demonstrate that trafficking of the antisense RNA along the Rev/RRE pathway is essential for efficient antisense inhibition, independent of a cotargeting component that is also a contributor.
We next performed an experiment to determine the localization of the antisense RNA and whether expression of antisense RNA resulted in nuclear retention and/or significant degradation of the proviral target RNA. To do this, we cotransfected the pNL4-3 proviral clone lacking a functional rev gene with pCMV-Rev and the RRE or CTE-driven antisense constructs at a 5:1 molar ratio. At this molar ratio, the RRE-driven antisense inhibits ca. 80% of particle production (see Fig. Fig.2A).2A). As a control, we also performed a similar experiment using an RRE-containing plasmid with the target region in the “sense” orientation, since this RNA does not inhibit particle production. In all of these experiments, a CMV-β-globin plasmid was cotransfected to provide a normalization control. To control for the quality of subcellular fractionation, we also transfected the provirus with or without pCMV-Rev in the absence of sense/antisense constructs. Since the GagPol RNA produced from the provirus requires Rev for export, in the absence of Rev, GagPol RNA would not be expected to be found in the cytoplasmic fractions.
At 65 h posttransfection, total and cytoplasmic RNA was extracted from the transfected cells, oligo(dT) selected, and analyzed by Northern blotting for GagPol, sense/antisense vector, and β-globin RNA. For this analysis, we compared total and cytoplasmic RNA levels in the control cells transfected with the sense RNA vector to the levels of RNA in cells transfected with the antisense vectors. A probe that detected sequences in the GagPol region in HIV-1 that is shared between the provirus and the sense/antisense plasmids was used for the GagPol and sense/antisense plasmids, and a separate probe was also included to detect β-globin RNA (see Materials and Methods). The results from this experiment are shown in Fig. Fig.33.
Analysis of the data in Fig. Fig.33 shows that the levels of total and cytoplasmic GagPol RNA in the presence of antisense were within 2.5-fold of the GagPol levels in the presence of sense RNA (see GagPol RNA). The total and cytoplasmic levels of the CTE- and RRE-antisense RNAs were also similar (see sense or antisense RNA). The presence of significant amounts of this RNA in the cytoplasm indicates that both the RRE- and CTE-antisense RNA are efficiently exported to the cytoplasm. Figure Figure3C3C shows proper fractionation during preparation of cell extracts since, in the absence of Rev, GagPol RNA did not localize to the cytoplasm. Thus, this experiment demonstrates that significant nuclear retention or degradation of target HIV-1 RNA in the presence of antisense RNA does not occur and that considerable amounts of antisense RNA can be found in the cytoplasm.
To more directly determine whether expression of RRE-driven antisense RNA had any effects on the stability of GagPol RNA and whether antisense RNA was degraded in the presence of target RNA, we cotransfected 293T cells with the pNL4-3 proviral clone lacking a functional rev gene, pCMV-Rev, and RRE-driven antisense plasmid at a 5:1 molar ratio of provirus and antisense vector (which causes an 80% reduction in p24 levels). To be able to compare the fate of sense and antisense RNAs in the presence or absence of target, we also performed a separate experiment in which the same amount of sense or antisense plasmid was cotransfected with only pCMV-Rev and pCMV-Tat. In all cases, a CMV-β-globin plasmid was cotransfected for normalization. After 48 h, the cells were harvested or treated with 5 μg of actinomycin D/ml for 3 or 6 h, to inhibit RNA polymerase II transcription, before harvesting. Total RNA was extracted, and oligo(dT)-selected RNA was prepared and analyzed by Northern blotting (Fig. 4A and C). The GagPol, sense, and antisense RNA levels present in each sample were then quantitated by using ImageQuant software after normalization to β-globin RNA levels (Fig. 4B, D, and E).
The results show that in the absence of target RNA, neither the antisense RNA nor the sense RNA was significantly degraded relative to β-globin RNA (Fig. 4A and B). There also was no significant change in the stability of the antisense RNA in the presence of target RNA, and the GagPol RNA was also stable (see Fig. 4C, D, and E). Taken together, these results demonstrate that antisense, in combination with target expression, does not lead to an apparent degradation of either antisense or target RNA. These results argue against RNA degradation as a contributing factor to the observed antisense effects.
The results from the RNA analysis indicated that the presence of the RRE-driven antisense RNA, which led to a very significant reduction of p24 in the tissue culture supernatant, did not reduce the GagPol mRNA levels in the cytoplasm of transfected cells. Since p24 is expressed from the GagPol RNA, this suggests that RRE-driven antisense RNA either results in the expression of low levels of protein from the GagPol mRNA or interference with later steps in virus particle assembly or release. To distinguish between these two models for antisense inhibition, we used a modified HIV-1 proviral clone (pHR1320), derived from the HIV-1 HXB2 isolate, as a target. This clone contains a glycine-to-alanine myristylation site mutation, a deletion at the GagPol frameshift sites, and an inactivating mutation in the protease active site. The result of these three mutations is that the clone produces a GagPol precursor protein (Pr160) that cannot target to the membrane and is trapped within the cell (54). Since the protein is not processed further, its levels can be easily measured and represent a more direct readout of the GagPol mRNA translation efficiency than the processed p24 protein. The target nucleotide sequence in pNL4-3 and HXB2 are >98% identical, and the antisense constructs inhibit particle production from a wild-type HXB2 proviral clone to the same extent as pNL4-3 targeted by antisense RNA (data not shown).
To examine total GagPol Pr160 production, we cotransfected 293T cells with the mutant HXB2 proviral construct, pCMV-Rev, and the RRE-driven sense or antisense plasmids, using a range of provirus/antisense molar ratios between 50:1 and 1:1. After 48 h, the cells were harvested and lysed, and the extracts were run on SDS-polyacrylamide gels, followed by electrotransfer and Western blotting with antibodies to the p24 CA portion of Pr160 and β-tubulin (to provide a loading control) (Fig. (Fig.5A).5A). The blot was then scanned and quantitated by using an Odyssey infrared imaging system, and GagPol Pr160 levels were normalized to β-tubulin. The results are shown in Fig. 5A and B.
The data show that expression of RRE-driven antisense RNA efficiently reduced Pr160 levels, even at an extremely low 25:1 provirus/antisense molar ratio. In contrast, the RRE-driven sense construct did not reduce Pr160 levels at all, except at the highest concentration used. The reduction of intracellular Pr160 levels by the RRE-driven antisense was similar to the observed inhibition of particle production measured by p24 in the supernatant in previous experiments. Thus, these results indicate a direct effect of antisense on protein levels rather than an effect on virus particle assembly or release.
To exclude the possibility that the reduction of Pr160 levels was the result of nondirect effects of the antisense vectors on the cell, such as interferon induction, we examined the effects of the antisense RNA on the HIV-1 Nef protein. The HIV-1 Nef accessory protein is encoded by a multiply spliced RNA in which the antisense target sequence has been removed by splicing. Thus, the RNA encoding Nef would not be expected to be directly affected by antisense RNA expression, unless the antisense RNA targeted the Nef mRNA precursor in the nucleus prior to splicing.
To determine whether expression of antisense RNA had any effects on Nef protein levels, we cotransfected 293T cells with the pNL4-3 proviral clone lacking a functional rev gene, pCMV-Rev, and increasing amounts of either the RRE or CTE-driven antisense construct. After 48 h the cells were harvested and lysed, and the extracts were separated on SDS-polyacrylamide gels, followed by electrotransfer and Western blotting with antibodies to Nef and β-tubulin (Fig. (Fig.5C).5C). The blot was then scanned and quantitated by using the Odyssey infrared imaging system, and Nef levels were normalized to β-tubulin levels. As can be seen in Fig. Fig.5D,5D, Nef protein levels remained constant in the presence of either the CTE- or RRE-driven antisense constructs (Fig. (Fig.5D).5D). This result demonstrates that the antisense effect is specific for mRNA that contains the target sequence and speaks against nonspecific effects on protein production or degradation. It also suggests that the antisense RNA does not exert its function on the target RNA prior to splicing.
Although our results clearly indicate that the target RNA reaches the cytoplasm, the lack of stable protein expression could indicate that in the presence of antisense RNA, the RNA traffics to a compartment away from the translation machinery, such as P-bodies (55). This would lead to a failure of the RNA to appear in polyribosome complexes. To determine whether the HIV-1 GagPol mRNA localized to polyribosomes in the presence of antisense RNA, we cotransfected 293T cells with the pNL4-3 provirus lacking a functional rev gene, pCMV-Rev, and the RRE-driven sense or antisense plasmids. The provirus and sense or antisense plasmids were transfected at the same 5:1 molar ratio of provirus and vector used in most of our other experiments. After 48 h the cells were harvested for cytoplasmic RNA, and extracts were applied to a sucrose gradient, which was subjected to ultracentrifugation as previously described (3). During fractionation, the gradient was continuously analyzed using a BioComp Gradient Master equipped with a UV monitor, which recorded absorbance at 254 nm. In vitro-transcribed Gag RNA (IVT Gag) was added to each fraction to control for recovery during subsequent analysis. RNA was then isolated from the collected fractions and analyzed by Northern blotting for GagPol, vector, β-globin, and IVT Gag RNA. Northern blots were analyzed and quantitated by using ImageQuant software. The blots are shown in Fig. 6A and B, and the quantitation is shown in Fig. Fig.6E.6E. To confirm that antisense inhibition was effective in this experiment, we also assayed cell supernatants from the cells for particle production. The p24 levels are indicated above the UV traces (Fig. 6A and B). As indicated, p24 levels were 53 ng/ml in the presence of antisense RNA versus 385 ng/ml in the control experiment. Thus, the 84% inhibition observed was consistent with earlier experiments.
As can be seen in Fig. 6A, B, and E, the GagPol mRNA showed a similar localization to fractions containing polyribosomes in the presence of either sense or antisense RNA. The sense and antisense RNAs were also localized to the polyribosome region of these gradients. Although these results suggested that, in the presence of antisense RNA, the target RNA still associated with polyribosome complexes, it was possible that this reflected association with a similar-size complex that cosedimented with polyribosomes. It has previously been demonstrated that treatment with 15 mM EDTA disrupts polyribosomes without disrupting most non-rRNA-protein complexes (6, 31). For this reason, we also treated cell lysates with 15 mM EDTA prior to fractionation (Fig. 6C and D). The loss of polyribosomal complexes in the presence of EDTA can be observed in the UV trace (Fig. 6C and D). In the presence of EDTA, GagPol, vector, and β-globin RNA all localized only to fractions that sedimented more slowly than the normal position of polyribosomes and, again, the GagPol target RNA showed a similar localization both in the presence and the absence of antisense RNA (Fig. (Fig.6F).6F). The lower amounts of GagPol RNA recovered in these experiments likely reflect degradation due to the RNA not being protected by association with the ribosomal machinery. Taken together, these results suggest that the target RNA is indeed localized to polyribosomes in the presence of antisense RNA. However, based on the results described above, it is clear that this association does not result in normal amounts of stable Gag protein.
Our results clearly demonstrate that the Rev/RRE pathway is required for efficient antisense inhibition of a target region in HIV-1 env. To determine whether this effect could be extended to non-HIV-1 target regions, we constructed a GFP expression vector that uses the HIV-1 5′ LTR as a promoter (see Materials and Methods). This construct does not contain splice sites and is expected to generate only one RNA transcript. In addition, it is not known to contain any additional elements that control RNA splicing or stability.
A portion of the GFP coding region was also used to make antisense constructs which contained either an RRE or had no additional export element. These constructs were essentially the same as the ones used to target the HIV env region in our previous experiments, except that a 721-bp GFP antisense sequence replaced the 927-bp env antisense sequence.
293T cells were transfected with LTR-GFP, pCMV-Rev, and antisense vectors which contained or lacked an RRE. A 5:1 and a 1:1 molar ratio of expression plasmid to antisense vector were used. After 48 h, cells were harvested and analyzed for GFP and β-tubulin expression (Fig. (Fig.7).7). Quantitation of the GFP after normalization to β-tubulin demonstrated that the RRE-driven antisense inhibited GFP expression from the LTR-GFP construct more efficiently than the antisense vector that contained no element. The inhibition observed was somewhat less potent than the inhibition seen with the HIV target provirus that trafficked through the RRE pathway (compare to Fig. Fig.2A)2A) but was very similar to the inhibition seen with the HIV target provirus that trafficked through the CTE pathway (compare to Fig. Fig.2B).2B). Thus, we conclude that trafficking of the antisense RNA through the Rev/RRE pathway promotes efficient inhibition, even on a non-HIV target sequence.
In this study, we present experiments demonstrating efficient inhibition of HIV-1 by unmodified antisense RNA. Our data clearly show that the expression of an antisense RNA targeting a region in env is able to significantly reduce particle production, even at very low ratios of antisense to target HIV-1 RNA. Although this was an unexpected finding, based on many previous studies showing the inefficiency of similar approaches to target HIV and other viruses, as well as cellular genes (26, 38, 60), it confirms the recent studies of Lu et al. targeting HIV-1 using a similar RRE-containing antisense RNA (42). Although our study was performed in 293T cells, the studies of Lu et al. (42) were performed in CD4+ lymphocytes. It is thus clear that efficient inhibition of HIV by antisense also occurs in the primary cells that are the natural host for HIV.
Our data specifically show that efficient targeting requires trafficking of the antisense RNA along the Rev/RRE pathway. In support of this conclusion, we demonstrated that the inhibition of HIV-1 was virtually abolished when the antisense RNA lacked an RRE or when Rev was not provided. In addition, the efficiency of antisense inhibition was significantly reduced when the RRE-containing antisense RNA was redirected to the Tap/Nxf1 pathway that normally is used to export CTE-containing RNA and many cellular mRNAs. We conclude that the efficient inhibition achieved using HIV-1-derived lentiviral vectors can be explained by the specific RNA trafficking pathway utilized by HIV-1.
Cotargeting of the antisense and HIV-1 target RNA to the same cellular compartment in the nucleus has recently been proposed as a possible mechanism for the efficiency of the vector that is currently in clinical trials (61). It was proposed that this would trigger extensive adenosine deamination of the HIV-1-antisense duplex, resulting in nuclear retention of the resulting dsRNA complexes (36, 78). However, our results show that, although cotargeting may contribute to antisense efficiency, it is clearly not essential, since an HIV-1 proviral clone that was altered to use an MPMV CTE, rather than an RRE, was also efficiently inhibited by the RRE-driven antisense RNA in the presence of Rev. In this case, the antisense RNA used the Rev/RRE pathway, whereas the HIV-1 target RNA was exported through the Tap/Nxf1 pathway. Previous work from many laboratories have shown that these export pathways are separate and use different cellular factors (15, 21, 53).
If cotargeting were the dominant reason for the observed antisense inhibition, the efficiency would also have been expected to remain the same when RevM10-Tap was used to replace Rev in the export of the HIV-1 target and antisense RNA. Instead, the inhibition was significantly reduced, demonstrating that trafficking through the Rev pathway is a major determinant of efficient antisense inhibition. However, nuclear cotargeting may still contribute to antisense inhibition, since the CTE-driven antisense more efficiently targeted the RRE-virus that was forced to use the Tap/Nxf1 pathway. In addition, trafficking of both RRE and CTE-driven antisense RNAs and targets on the Tap/Nxf1 pathway also led to nearly identical inhibition profiles.
Our results clearly demonstrate that nuclear retention of HIV-1 target RNA did not contribute to the antisense inhibition that was achieved with the RRE-driven antisense RNA in the presence of Rev. Even when p24 levels were efficiently reduced, the GagPol RNA was still exported to the cytoplasm and present in polyribosomal complexes. In addition, our experiments with a proviral HIV-1 clone that did not give rise to virus particles demonstrated that the inhibition did not occur at the level of particle assembly or release. Thus, the antisense effects are manifested at the cytoplasmic level, after the association of the mRNA with the translation machinery, but before particle assembly.
We also showed that the reduction in p24 was not due to a general effect on protein synthesis, since Nef protein levels were unaffected by antisense expression. This specificity of inhibition for the HIV-1 target was somewhat surprising, since in mammalian cells, long double-stranded RNAs often activate protein kinase R and 2′5′-oligoadenylate synthetase, which leads to the interferon response and a general downregulation of translation and RNA degradation (29, 64). This suggests that trafficking of the antisense RNA through the Rev/RRE pathway somehow allows the interferon response to be bypassed. An analogous bypass mechanism has been suggested for inhibition by some siRNAs (12, 64).
Although in most cases, the appearance of an RNA in polyribosome complexes leads to production of protein, a reduction in the rate of initiation concomitant with a decrease in the rate of elongation can give rise to significantly reduced levels of protein (“ribosome stalling”) that could explain the reduced GagPol protein levels seen (68). Another possibility is that protein is produced but rapidly degraded. Both of these mechanisms have been proposed to function in miRNA-mediated translation inhibition (51). Alternatively, the high-molecular-weight complexes could represent complexes that are EDTA sensitive and appear to be polyribosomes but are actually recently described pseudo-polyribosomal complexes (70). Further experiments will be needed to distinguish between these possibilities.
Independent of the detailed mechanism utilized, our results clearly point to a novel, previously unknown, mechanism for antisense inhibition. One model would be that Rev and the RRE allow the target/antisense RNA complex to be exported to the cytoplasm, where the antisense RNA functions to inhibit protein production. Intriguingly, there have been several reports that some of the more complex retroviruses produce natural antisense transcripts (5, 7, 37, 45, 50). The best characterized of these RNAs in HIV-1 appears to initiate from multiple transcription start sites 5′ of the 3′ LTR and extend into the pol region, where a novel polyadenylation site has recently been described (37). Although additional studies are needed to validate the presence of antisense RNA in HIV-1-infected cells, the evidence for the existence of an antisense transcript in the human T-cell leukemia retrovirus is much more compelling (7, 62, 71). In light of our finding that antisense transcripts can be potent inhibitors of gene expression in the HIV-1 system, further studies on the role of these natural transcripts in the regulation of HIV-1 and other retroviruses seem warranted.
Although we do not know whether a similar mechanism of antisense inhibition normally operates in the host cell to function in gene regulation, recent evidence indicates that expression of natural antisense RNA to normal gene transcripts may be a common occurrence (9, 17, 25). Although most mRNAs probably do not traffic down the Crm1 pathway used by Rev and the RRE, this pathway has been reported to be utilized by some mRNAs (35, 65). Thus, it is possible that mechanisms similar to the one we have described are utilized to regulate expression of cellular mRNAs.
In the case of cellular, as well as viral mRNA, it has been shown that regions of double-stranded RNA, resulting from the presence of inverted sequences in the RNA or association with antisense RNA, are subject to deamination by ADAR, leading to multiple inosines in the RNA (1). Such RNAs are normally retained in the nucleus through interaction with nuclear matrix proteins and eventually degraded (36, 57, 78). However, a previous study using the Xenopus oocyte export model showed that edited RNA was exported to the cytoplasm if the RNA contained an RRE and Rev was provided in trans (78). In addition, a previous study by Lu et al. using a lentivirus antisense vector reported that HIV-1 RNA recovered from cells expressing antisense RNA showed multiple mutations in the antisense target region of the HIV-1 genome that were consistent with ADAR activity (42). These results suggest that at least some of the genomic RNA, which was complexed with antisense RNA and edited as a result of formation of dsRNA, was eventually exported to the cytoplasm and packaged. Therefore, it seems possible that editing activity plays a role in antisense inhibition. It also follows that naturally occurring antisense could help provide retroviruses with an additional pathway for sequence diversification.
However, to date, reverse transcription-PCR sequencing of cytoplasmic RNA in cells transfected with proviral clones and plasmids expressing antisense RNA has failed to detect any mutations indicating ADAR editing (data not shown). Thus, we do not believe that editing is directly connected to the reduced protein levels we observe. Also, the target region is downstream of the GagPol ORF, and mutations in the target region, the 3′ untranslated region, will thus not affect the GagPol protein per se. However, editing of even a small amount of the RNA could potentially lead to the production of miRNA from the antisense RNA (52). Inhibition by a miRNA-mediated mechanism would be consistent with the efficient inhibition we observe, and miRNA often exerts its effect at the translation level (55).
Irrespective of the mechanism utilized for antisense inhibition, our results are of importance for future development in the gene therapy field. The data suggest that it will be advantageous to ensure that any long antisense RNA designed to combat HIV-1 contains the RRE to allow trafficking along the Rev/RRE pathway, since this will likely significantly increase the efficiency of antisense inhibition. Our data also show that an RRE-driven antisense RNA, in combination with Rev, is able to efficiently inhibit a target that utilizes the CTE pathway. This raises the possibility that Rev/RRE trafficking of antisense RNA could also be exploited to make antisense RNA inhibition more efficient for non-HIV-1 applications.
We thank Lynne Maquat, Casey Morrow, and the AIDS Research and Reference Reagent Repository for providing valuable reagents. Joy Morgenegg provided expert tissue culture assistance, Yeou-Cherng Bor provided technical advice, and Ann Beyer provided a critical reading of the manuscript.
This study was supported by National Institutes of Health grants CA097095 and AI054335 to M.-L.H. and grants AI054213 and AI068591 to D.R. Salary support for M.-L.H. and D.R. was provided by the Charles H. Ross, Jr., and Myles H. Thaler Endowments at the University of Virginia.
Published ahead of print on 29 October 2008.