|Home | About | Journals | Submit | Contact Us | Français|
Nuclear RNA processing events, such as 5′ cap formation, 3′ polyadenylation, and pre-mRNA splicing, mark mRNA for efficient translation. Splicing enhances translation via the deposition of the exon-junction complex and other multifunctional splicing factors, including SR proteins. All retroviruses synthesize their structural and enzymatic proteins from unspliced genomic RNAs (gRNAs) and must therefore exploit unconventional strategies to ensure their effective expression. Here, we report that specific SR proteins, particularly SRp40 and SRp55, promote human immunodeficiency virus type 1 (HIV-1) Gag translation from unspliced (intron-containing) viral RNA. This activity does not correlate with nucleocytoplasmic shuttling capacity and, in the case of SRp40, is dependent on the second RNA recognition motif and the arginine-serine (RS) domain. While SR proteins enhance Gag expression independent of RNA nuclear export pathway choice, altering the nucleotide sequence of the gag-pol coding region by codon optimization abolishes this effect. We therefore propose that SR proteins couple HIV-1 gRNA biogenesis to translational utilization.
From transcription to translation to cytoplasmic mRNA degradation, the sequential phases of gene expression are coupled physically and functionally. In humans, 94% of genes have more than one exon (79) and pre-mRNA splicing plays a central role in regulating transcription, nuclear stability, 3′ end formation, nuclear export, cytoplasmic trafficking, cytoplasmic stability, translation efficiency, and even posttranslational events, such as protein half-life (reviewed in references 56 and 75). The integration of pre-mRNA splicing with later steps in RNA metabolism can prevent the translation of deleterious RNAs in at least two ways. First, mRNAs containing functional introns are retained by the nucleus (14, 45), thereby preventing the translation of incompletely processed transcripts and the synthesis of deleterious protein products. Second, and less well understood, the coupling of splicing to the control of nuclear stability, export, and translation prevents the translation of unspliced parasitic and/or noncoding RNAs (reviewed in reference 6). Because viruses frequently depend upon the translation of unspliced viral RNAs, they need to have developed mechanisms to circumvent such barriers to gene expression.
All retroviruses use a combination of spliced and unspliced RNAs to express their proteins (reviewed in references 20 and 71). The nuclear export of a full-length, unspliced, intron-containing transcript is essential for the replication of all retroviruses, since this transcript is (i) the viral genome (genomic RNA [gRNA]) that is packaged into virions, (ii) the mRNA that is translated into Gag and Gag-polymerase (Pol), the polyproteins that yield the virion structural and enzymatic proteins, and (iii) a physical scaffold that helps promote virion assembly. Human immunodeficiency virus type 1 (HIV-1) has one of the most complicated gene expression strategies of all known retroviruses, where the Env glycoprotein and accessory proteins Vif, Vpr, and Vpu are translated from singly spliced mRNAs containing functional introns, and the Tat, Rev, and Nef proteins are expressed from mRNAs that have been fully spliced (reviewed in reference 61). The nuclear export of retroviral gRNAs has become a model example for how posttranscriptional RNA metabolism can overcome the aforementioned obstacles to expression imposed by a lack of splicing (20, 71).
For cellular transcripts, a heterodimer of NXF1 and NXT1/2 is necessary for the vast majority of mRNA nuclear export and is recruited to the RNA through multiple protein-protein interactions (reviewed in reference 42). Splicing enhances nuclear export by mediating the recruitment of multiple proteins, including the transcription export (TREX) complex (16, 77) and a subset of SR proteins including SF2/ASF, SRp20, and 9G8 (29, 30, 32, 44, 53). These adapters, possibly in a partially redundant manner, recruit NXF1-NXT and allow the nuclear export of fully processed mRNAs. Because retroviral gRNAs are unspliced and do not recruit mRNA export adapters deposited during pre-mRNA splicing, they must exploit alternative strategies to achieve efficient nuclear export (20, 71). Some retroviruses, such as Mason-Pfizer monkey virus (M-PMV), contain a cis-acting RNA structure called the constitutive transport element (CTE) that directly recruits NXF1 to the gRNA without the need for protein adapters. Other retroviruses, typified by HIV-1, use the Crm1-mediated nuclear export pathway. In particular, HIV-1 encodes the regulatory protein Rev, which binds both a viral cis-acting RNA sequence called the Rev response element (RRE) and Crm1, thus enabling the gRNA to connect to the Crm1 nuclear export pathway. In sum, retroviral gRNA nuclear export complexes allow splicing-independent gene expression by averting repressive mechanisms and specifically recruiting host nuclear export factors. Given the evident coupling between steps in RNA metabolism, it is not unexpected that such complexes also have the potential to regulate not only Gag translation but also posttranslational events, such as the packaging of gRNA into virions and the trafficking of Gag to the plasma membrane (10, 11, 17, 37, 38, 68, 71).
Several recent studies have begun to address the mechanism by which pre-mRNA splicing is coupled to translation (reviewed in references 34 and 56). Most notably, splicing increases translation efficiency and marks the mRNA for translation-dependent quality control processes, such as nonsense-mediated decay (NMD), via the recruitment of an exon junction complex (EJC) about 20 to 24 nucleotides (nt) upstream of the exon junction (22, 48, 57, 81). The recruitment of SR proteins during splicing is a second mechanism by which splicing marks mRNAs for efficient translation (8, 54, 64, 65, 67). As retroviral gRNAs are unspliced and will not have recruited EJCs, the means by which their translation is regulated is unknown.
Multiple deficiencies in HIV-1 replication have been described to occur in mouse cells, and their analyses have been instrumental in identifying and characterizing cellular factors that promote or inhibit viral replication. These replication blocks encompass most steps of the life cycle, including entry, reverse transcription, integration, transcription, pre-mRNA splicing, Gag translation, virion assembly, and infectivity (4, 7, 24, 26, 49, 50, 52, 72, 80, 82, 84). Because gRNA metabolism has features that distinguish it from that of most cellular mRNAs, we have postulated that species-specific differences in cellular RNA binding proteins may affect the noted posttranscriptional deficiencies in HIV-1 replication in murine cells. Indeed, we have previously shown that mouse NIH 3T3 (3T3) cells are inherently defective for Rev-dependent Gag translation and Gag trafficking to the plasma membrane and that both defects can be substantially rescued by changing the cis-acting signal for gRNA nuclear export element from the RRE to four copies of the M-PMV CTE (4xCTE) (68, 72). The mechanism by which 4xCTE enhances these steps is unknown, and since the murine replication blocks can be complemented by fusion of mouse and human cells (7, 18, 51, 76), we sought to use a genetic overexpression screen of human genes to identify cellular proteins that modulate the posttranscriptional regulation of HIV-1 gene expression. Here, we show that SR proteins can complement the RRE-dependent Gag translation defect in murine cells. Since SR proteins couple splicing and translation (reviewed in references 47 and 85), we hypothesize that they mark HIV-1 gRNA for translation and therefore compensate for the lack of splicing.
NIH 3T3 and HeLa cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 1% penicillin-streptomycin, and 1% l-glutamine. pGag-Pol-RRE, 1xCTE, and 4xCTE were generated by first inserting the human cytomegalovirus (CMV) immediate-early (IE) promoter-HIV-1 5′ untranslated region (UTR) fusion from pBC12 (21) into pcDNA3.1 containing the RRE (72). HIV-1NL4-3 Gag-Pol was then inserted as a SacI-EcoRI fragment where the EcoRI site was artificially inserted 3′ to the pol reading frame (NL4-3 nt 455 to 4610 inserted in total). 1xCTE or 4xCTE was then swapped in place of the RRE as EcoRI-StuI fragments. The pGag-Pol(PR D25A) constructs were made by replacing the SacI-AgeI fragment from pNL4-3(PR D25A) (28) into the same sites in the wild-type versions. The pGag-Pol(MA G2A), pGag-Pol(MA L8A), and pGag-Pol(MA L21S) plasmids have previously been described (68). The pGag-Pol-RRE-LTR plasmid was cloned using pGL4 (Promega) as the vector backbone. The CMV promoter and HIV-1 5′ UTR-Gag-Pol are identical to the pGag-Pol constructs described above, the 3′ long terminal repeat (LTR) encompasses nt 8887 to 9709 of the HIV-1NL4-3 isolate, and the RRE was cloned into the unique XhoI site in this vector. pGag-RRE, 1xCTE, and 4xCTE were generated by replacing the SphI-EcoRI fragment from pGag-Pol-RRE with a SphI-EcoRI fragment, with the EcoRI site artificially inserted at the 3′ terminus of the gag reading frame. The RRE was then swapped with 1xCTE and 4xCTE as described above. pluciferase-FLAG was generated by inserting the luciferase coding region from pGL4 (Promega) into pCMV-Tag4A (Stratagene) to add the FLAG tag. The luciferase-FLAG sequence was then inserted into pcDNA3.1 as an NheI-EcoRI fragment. pCO-Gag-Pol was generated by excising codon-optimized Gag-Pol from pHDMHgpm2 as a XbaI-PvuI fragment, filled in using T4 DNA polymerase, and then ligated into a pcDNA3.1 backbone previously digested with NheI-EcoRI and filled in using T4 DNA polymerase. pcRev and pluciferase were previously described (72).
pT7-SF2/ASF, pT7-SRp20, pT7-9G8, pT7-SC35, pT7-SRp40, pT7-SRp55, pT7-SRp40ΔRS, pT7-FF40, pT7-F40F, pT7-F4040, and pT7-40F40 were previously described (12, 83). All further pT7 constructs were prepared by replacing SC35 in pT7-SC35 with the insert described. pT7-luciferase was generated by inserting the Renilla luciferase coding region into the pT7 vector. pT7-40FF and pT7-4040F were generated using overlapping PCR. pT7-SRp55ΔRS was made by PCR amplifying SRp55 to insert a stop codon after amino acid P183. pT7-mSRp55 was generated by amplifying mSRp55 from the IMAGE:2648111 construct. pT7-mSRp40 was prepared by amplifying mSRp40 from cDNA derived from RNA extracted from 3T3 cells. pNXF1 and pNXT1 were previously described (39).
3T3 or HeLa cells were transfected using FuGENE 6 (Roche) according to the manufacturer's directions, using a ratio of 3 μl FuGENE/1 μg DNA. A standard transfection mixture contained 1 μg Gag expression vector, 0.25 μg pcRev (for RRE-containing constructs) or pGFP (for 1xCTE- or 4xCTE-containing constructs), and 0.5 μg pT7-luciferase or pT7-SR protein added to a 35-mm well seeded at ~50% confluence. When the amount of pGag-Pol plasmid was titrated down, pluciferase was used as filler DNA. In the experiment whose results are shown in Fig. Fig.6C,6C, the standard transfection mixture was altered so that 0.0625 μg of pGag-Pol(PR D25A)-RRE was added with pluciferase as filler DNA to give 1 μg, and the amounts of pcRev and pT7 were as described above.
Approximately 40 h after transfection, the medium was removed, filtered through a 0.45 μM filter, centrifuged through a 20% sucrose cushion for 2 h at ~21,000 × g, and then resuspended in 0.5% Triton X-100 in phosphate-buffered saline (PBS). The cells were lysed in radioimmunoprecipitation assay (RIPA) buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.1% SDS, 1% Triton X-100, 1% sodium deoxycholate), and the proteins were resolved by SDS-PAGE. After transferring to a nitrocellulose membrane, Gag was detected using an antibody that recognized p24Gag (25) and a secondary antibody conjugated to IRdye800 (Li-Cor Biosciences) for quantitative immunoblotting using an Odyssey infrared scanner (Li-Cor Biosciences). Hsp90 was detected using rabbit anti-Hsp90 antiserum (Santa Cruz Biotechnology), and luciferase-FLAG was detected using the mouse anti-FLAG antibody (Stratagene). The amount of p24Gag in the medium was quantified using a p24Gag enzyme-linked immunosorbent assay (ELISA) kit (Perkin-Elmer).
3T3 and HeLa cells were analyzed using [35S]methionine-cysteine metabolic labeling as previously described (68). RNA was isolated from 3T3 cells and analyzed by Northern blotting as previously described (72), except that the probe used herein was generated to specifically recognize the RRE (nt 7708 to 8058 of the HIV-1HXB3 provirus) or 5′ UTR-Gag (nt 759 to 1457 of the HIV-1NL4-3 provirus).
There are multiple posttranscriptional HIV-1 replication deficiencies in murine cells, including gRNA oversplicing, suboptimal Gag translation, and ineffective Gag trafficking to the plasma membrane (7, 15, 43, 50, 52, 62, 68, 72, 84). To analyze the regulation of Gag translation and virion production in mouse cells, we developed an experimental system to examine surrogate gRNAs containing the full HIV-1 5′ UTR and Gag-Pol coding region and either the RRE, one copy of the CTE (1xCTE), or four copies of the CTE (4xCTE) as the cis-acting RNA export element (Fig. (Fig.1A).1A). Efficient transcription was ensured by using the hCMV-IE promoter with the HIV-1 5′ UTR fused at the transcriptional start site (21). These constructs contain a single splice donor (in the 5′ UTR) and a single splice acceptor (SA; the acceptor for the vif transcript in pol) and, importantly, yield sufficient pools of unspliced RNA in mouse cells for their fate and function to be explored.
As demonstrated previously, cotransfection of pGag-Pol-RRE and pcRev into 3T3 cells (plus a luciferase control vector as filler) resulted in a low level of Gag expression, as detected by immunoblot analysis using a CA-specific antibody (Fig. (Fig.1B,1B, lane 1) (68, 72). Gag is a 55-kDa polyprotein precursor that is cleaved into p17Gag (matrix [MA]), p24Gag (capsid [CA]), p7Gag (nucleocapsid [NC]), and p6Gag by the viral protease (PR) after Gag binds the plasma membrane and viral particles start to assemble (reviewed in reference 1). As Gag trafficking to the plasma membrane is inefficient in 3T3 cells, little or no intracellular Gag processing was detected. Similarly, a very low level of p24Gag was detected by ELISA in the culture supernatant, indicative of minimal virion-like-particle (VLP) production. Because simply switching the nuclear export element from the RRE to 4xCTE increased Gag expression 5- to 10-fold and VLP production ~50-fold in mouse cells (68, 72) (Fig. (Fig.1B,1B, compare lanes 1 and 8), we hypothesized that one or more murine regulatory RNA binding proteins may not interact with RRE-Rev-dependent gRNA. Complementation of this deficiency with human cDNAs could serve as an approach for identifying factors important for the posttranscriptional regulation of HIV-1 gRNA metabolism and/or virus particle assembly. We therefore performed such a gain-of-function screen focusing on candidates previously known to bind mRNAs in the nucleus and influence their cytoplasmic regulation (55). We tested over 50 RNA binding proteins, most of which have been described to modulate pre-mRNA splicing or translation.
While the full results of this screen will be reported elsewhere, the proteins with the greatest ability to increase intracellular Gag levels belonged to the SR protein family. As can be seen in Fig. Fig.1B,1B, lanes 2 to 7, all six of the human SR proteins that we tested increased intracellular Gag abundance and supernatant VLP levels. More specifically, SF2/ASF, SRp20, 9G8, and SC35 each increased intracellular Gag levels and VLP production ~5-fold. SRp40 and SRp55 were far more efficacious and increased intracellular Gag ~30- and ~90-fold, respectively, and VLP production up to 200-fold in the case of SRp55. Differences in activity between the SR proteins did not correlate with their expression, since SRp40 and SRp55 both accumulated to lower levels than SF2/ASF and SC35 (see supplemental Fig. S1 at http://www.kcl.ac.uk/malim). Titration studies in which the levels of transfected pT7-SRp40 or pT7-SRp55 vector were lowered from 500 ng to less than 20 ng showed that Gag expression still increased ~20-fold, indicating that the effects of these SR proteins on HIV-1 gene expression are potent (see supplemental Fig. S2). Human and mouse SRp40 and SRp55 share 98% and 96% sequence identity, respectively, and the human and mouse cDNAs exhibited similar enhancing effects on Gag expression and VLP production (see supplemental Fig. S3), though the mouse orthologs tended to be moderately less active (not shown).
To determine if the ability of the SR proteins to enhance Gag expression and VLP production was absolutely dependent on RRE-mediated RNA nuclear export, all six SR protein constructs were individually cotransfected into 3T3 cells with pGag-Pol-4xCTE (Fig. (Fig.1C).1C). SC35, SRp40, and SRp55 increased intracellular Gag levels while SF2/ASF, SRp20, and 9G8 had little effect. When the input pGag-Pol-4xCTE plasmid was reduced (1 μg to 0.125 μg DNA), SRp40 and SRp55 induced a greater enhancement of intracellular Gag expression (data not shown). Therefore, overexpression of SRp40 and SRp55 substantially increased intracellular Gag abundance in 3T3 cells regardless of which RNA nuclear export element was present.
9G8 has recently been shown to modulate HIV-1 3′ end formation when the poly(A) signal is derived from the HIV-1 3′ LTR but not from the bovine growth hormone (BGH) transcript (78). The HIV-1 Gag-Pol constructs tested so far all contain the BGH poly(A) signal, which is a strong poly(A) signal (60) known to be resistant to differential regulation of 3′ end formation (77). To determine if SRp40 and SRp55 enhance RRE-dependent Gag expression and VLP production in 3T3 cells when the poly(A) signal is derived from the HIV-1 3′ LTR, pGag-Pol-RRE-LTR (Fig. (Fig.1A)1A) was cotransfected with pcRev and either pT7-luciferase, pT7-SRp40, or pT7-SRp55 (see supplemental Fig. S4). Both SRp40 and SRp55 enhanced intracellular Gag abundance and VLP production, demonstrating that Gag expression is rescued irrespective of whether the poly(A) signal is from the viral 3′ LTR or a cellular transcript.
The SRp40/55-mediated induction in Gag expression could be due to increased Gag translation or extended intracellular half-life. To address this directly, we carried out [35S]methionine-cysteine pulse experiments where 3T3 cells transfected with pGag-Pol-RRE, pcRev, and an SR protein expression construct (or control) were metabolically labeled for 10 min and Gag synthesis was monitored by immunoprecipitation. As shown in Fig. Fig.2A,2A, samples in which SRp40 or SRp55 were overexpressed displayed, respectively, an ~20- or ~100-fold increase in the amount of Gag translated compared to the luciferase control (compare lanes 1, 4, and 5). These increases in translation rates correlated well with the increases in steady-state intracellular Gag levels shown in Fig. Fig.1B1B and demonstrated that changes in Gag translation rates largely accounted for the changes in steady-state abundance.
Because SR proteins affect multiple steps of mRNA biogenesis and metabolism, we next addressed the consequences of ectopic SR protein expression for gRNA expression from the pGag-Pol-RRE vector by Northern blot analysis of cytoplasmic RNA isolated from transfected 3T3 cells. An RRE-specific probe was used so that both the spliced and the unspliced RNA would be recognized (Fig. (Fig.2B),2B), and the total amount of HIV-1 RNA (the unspliced and spliced bands added together and normalized to pGag-Pol-RRE-luciferase at 100%) as well as the percentage of unspliced RNA (the amount of the unspliced band divided by the total HIV-1 RNA in that lane) was quantified (Fig. (Fig.2C).2C). As expected, even in the presence of Rev, most of the gRNA was spliced in the control sample (lane 1). When certain SR proteins were overexpressed, the total amount of HIV-1 RNA changed substantially, which has previously been noted for gRNA derived from intact proviruses (35, 36). Only SRp55 substantially increased the abundance of unspliced RNA (the template for Gag), and then only by ~5- to 10-fold, which is much less than the ~90-fold increase in steady-state Gag levels (Fig. (Fig.1B)1B) or the ~100-fold increase in Gag translation rate (Fig. (Fig.2A).2A). SRp40 increased the unspliced RNA abundance only about 2-fold, which is also much less than the ~30-fold change it induces in steady-state Gag levels and the ~20-fold change it causes in translation rate. In addition to moderately altering the abundance of the unspliced HIV-1 RNA, SRp40 and SRp55 also increased the percentage of unspliced RNA (Fig. (Fig.2C).2C). Similar results were seen for Northern blots analyzing RNA isolated from cells 24 and 48 h after transfection (data not shown). We therefore conclude that SRp40 and SRp55 each promote Gag translation substantially but that they also increase the abundance of the unspliced transcript.
Since the total HIV-1 RNA levels as well as the ratio of spliced to unspliced RNA changed in response to different human SR proteins, it was difficult to differentiate between effects on splicing and consequences for transcription, 3′ end processing, or mRNA stability (31, 47, 85). To clarify this, we next analyzed whether the 3′ splice acceptor in our construct was necessary for the enhancement of intracellular Gag abundance. We compared the effects of SR protein expression in 3T3 cells on a protease (PR)-deficient (where the aspartic acid residue at position 25 was replaced with alanine) Gag-Pol-RRE vector and a Gag-RRE vector where the 3′ SA had been removed (Fig. (Fig.1A).1A). Consistent with the idea that alterations in Gag translation are independent of any effects on gRNA splicing, cotransfection of pT7-SRp40 and pT7-SRp55 with either pGag-Pol(PR D25A)-RRE plus pcRev or pGag-RRE plus pcRev led to similar increases in intracellular Gag levels (Fig. (Fig.3A).3A). Therefore, functional 3′ splice signals in Gag-Pol mRNA are not necessary for the SR protein-mediated induction of Gag expression.
We then compared the amount of Gag RNA expressed by Northern blotting and the amount of Gag protein translated by pulse-labeling in matched transfections 24 h after 3T3 cells were cotransfected with pGag-RRE plus pcRev and either pT7-luciferase, pT7-SF2/ASF, pT7-SRp40, or pT7-SRp55 (Fig. (Fig.3B).3B). As a control, pGag-4xCTE was also transfected with pT7-luciferase. SRp55 had a dramatic effect on the amount of Gag translated (~20-fold), while the increase in RNA abundance was less substantial (less than 5-fold). Also, SF2/ASF and SRp40 consistently have similar effects on RNA levels (Fig. (Fig.2B2B and and3B),3B), while SRp40 always causes a much greater increase in the amount of Gag translated (Fig. (Fig.2A2A and and3B)3B) and in its steady-state levels (Fig. (Fig.1B1B and and3A).3A). SR proteins had similar, though slightly more pronounced, effects on Gag RNA abundance 48 h after transfection, and no difference was detected whether a whole cell or the cytoplasmic RNA fraction was tested (see supplemental Fig. S5; also data not shown). Therefore, while some SR proteins, and in particular SRp55, moderately increase the abundance of Gag RNA, a subset of them, most notably SRp40 and SRp55, substantially increase the translational efficiency of the RNA.
To establish the substrate specificity for the SR-mediated enhancement of Gag expression in 3T3 cells, we coexpressed our panel of SR proteins with codon-optimized HIV-1 Gag-Pol or luciferase expression vectors that have the same promoter and polyadenylation elements as our surrogate gRNAs (Fig. (Fig.1A).1A). The codon-optimized Gag-Pol gene has 1,170 changes in the RNA sequence but maintains the identical protein-coding sequence. No substantial changes in luciferase (Fig. (Fig.4A)4A) or codon-optimized-Gag (Fig. (Fig.4B)4B) expression were observed upon addition of the human SR proteins, demonstrating that the SR protein-mediated phenotypes discussed above are specific for wild-type Gag and Gag-Pol coding sequences. To ensure that excessive intracellular Gag levels were not masking an effect of SRp40 or SRp55 on codon-optimized Gag-Pol, we titrated input plasmid DNA levels from 1 μg to 0.0625 μg (see supplemental Fig. S6). At all levels of transfected pCO-Gag-Pol DNA, only minor changes in intracellular Gag levels were observed in the presence of pT7-SRp40 or pT7-SRp55. In sum, the potent effect of SR proteins on Gag expression specifically reflects the viral identity of the transcript.
We have previously reported that the HIV-1 Gag translation and virion assembly deficiencies are independent in 3T3 cells (68, 72). The assembly defect in mouse cells is predominately at the step of Gag trafficking to the plasma membrane (7, 15, 43, 52, 62, 68, 72). Gag is myristylated at the amino-terminal glycine, and this myristate can either be sequestered in the globular head of MA or be exposed to interact with the plasma membrane (1). It is unclear how this myristyl switch is regulated, but we and others have hypothesized that the myristate is constitutively sequestered in 3T3 cells (27, 68). As overexpression of SRp40 or SRp55 substantially enhanced Rev-dependent Gag expression well beyond the level of 4xCTE-dependent Gag expression (which produces Gag that is assembled efficiently) (Fig. (Fig.1B),1B), we wished to determine the effects of SR proteins on particle assembly. We therefore titrated the amount of transfected pGag-Pol-RRE against a constant level of pT7-SRp55 and compared the intracellular Gag processing and VLP production to those seen with pGag-Pol-4xCTE (Fig. (Fig.5A).5A). We quantified all of the Gag bands and determined both the assembly efficiency (the total amount of VLPs produced divided by the total amount of intracellular Gag, normalized to the level for 4xCTE at 100%) and the percentage of Gag processing (the total amount of processed Gag bands divided by the total amount of Gag) (Fig. (Fig.5B).5B). At all Gag expression levels tested, the assembly efficiency of RRE-dependent Gag-Pol in the presence of human SRp55 was lower than that of 4xCTE-dependent Gag-Pol. When the intracellular levels of Gag were similar for RRE- and 4xCTE-dependent Gag-Pol (Fig. (Fig.5A,5A, compare lanes 5 and 7), the assembly efficiency of RRE-dependent Gag was negligible.
As the Gag polyprotein is not processed until after it binds the plasma membrane (1), the amount of intracellular Gag processing is an accurate indicator for membrane trafficking. Even at high intracellular Gag levels, the percentage of Gag processing seen for pGag-Pol-RRE/SRp55 samples is lower than that seen for 4xCTE-dependent Gag, and at similar levels of Gag, 4xCTE-dependent Gag is processed much more efficiently (Fig. (Fig.5B,5B, lanes 5 and 7). Therefore, while the translation defect in 3T3 cells is overcome by SRp55 overexpression, the Gag that is expressed is relatively nonfunctional and fails to form VLPs efficiently. Similar results were obtained with SRp40 (data not shown).
To characterize further the effect of SR proteins on virion assembly, we analyzed the effect of SR protein overexpression with either the wild-type Gag protein or Gag proteins with previously characterized amino acid mutations in MA. When pT7-luciferase, pT7-SRp40, or pT7-SRp55 was cotransfected into 3T3 cells with pGag-Pol-RRE, pGag-Pol(MA G2A)-RRE, pGag-Pol(MA L8A)-RRE, or pGag-Pol(MA L21S)-RRE plus pcRev, a clear role for the MA sequence emerged. While SRp40 and SRp55 enhance the intracellular Gag abundance of the G2A mutant protein, in contrast to the wild-type protein, very little VLP production was observed (Fig. (Fig.5C,5C, compare lanes 1 to 3 and 4 to 6), consistent with the observation that myristylation of MA is necessary for virion production (1). It should be noted that even though high steady-state intracellular Gag concentrations are present with the G2A mutant protein when SRp40 and SRp55 are overexpressed, very little intracellular Gag processing is detected, which further demonstrates that the efficiency of Gag processing is not dependent on the intracellular Gag abundance.
We have previously shown that the L8A and L21S mutations in MA have opposite effects on virion assembly in 3T3 cells, likely through their effects on the myristyl switch in Gag (68). The L8A mutation sequesters the myristate in the globular head of MA, preventing membrane binding and virion production (58, 63), while the L21S mutant Gag protein assembles much more efficiently than the wild-type protein (58). While SRp40 and SRp55 substantially enhance intracellular Gag expression for L8A and L21S Gag, these mutations still have opposite effects on VLP production. SRp40 and SRp55 increase the VLP production for L8A Gag to levels that are still slightly lower than those seen for wild-type Gag (Fig. (Fig.5C,5C, compare lanes 1 to 3 and 7 to 9), supporting our hypothesis that, in 3T3 cells, wild-type Gag is mostly in a myristate-sequestered state. However, the enhancement in VLP production that the SR proteins mediate is additive with the L21S-associated phenotype (compare lanes 1 to 3 and 9 to 12). Therefore, even though the substantial increase in Gag expression due to SR protein overexpression increases VLP production, this does not overcome the intrinsic inefficiency of Rev-dependent Gag assembly in 3T3 cells, and VLP production can be further increased by a MA mutation previously described to increase virion assembly in mouse cells.
In some human cells, HIV-1 Gag-Pol mRNA containing a single copy of the CTE (1xCTE) as the export element displays a Gag translation defect that is reminiscent of the translation deficiency of RRE-dependent Gag-Pol mRNA in 3T3 cells (19, 72). This 1xCTE-associated defect can be overcome in human cells by overexpression of NXF1/NXT or 9G8 (39, 73), raising the possibility that interactions between SR proteins and NXF1 underlie this effect (73). To address this further, we analyzed our panel of SR proteins for their ability to enhance 1xCTE-dependent Gag expression in transfected HeLa cells. Remarkably, similar rank orders of Gag induction levels were observed in cotransfection experiments with pGag-Pol(PR D25A)-1xCTE (Fig. (Fig.6A,6A, top panels) and those with pGag-1xCTE (Fig. (Fig.6A,6A, bottom panels), as had been noted earlier for pGag-Pol(PR D25A)-RRE in 3T3 cells (Fig. (Fig.3A);3A); cotransfection with pNXF1/pNXT (but no SR protein expression vector) gave a strong induction of Gag expression, similar to what was found for SRp55, and served as a positive control.
We also carried out a pulse-labeling study with [35S]methionine-cysteine and confirmed that increases in Gag expression levels correlated with increased translation rates (Fig. (Fig.6B).6B). Thus, in two different experimental scenarios, using two distinct RNA export pathways, ectopic expression of SR proteins potently stimulates the translation of HIV-1 Gag-Pol mRNA. Interestingly, SC35, SRp40, and SRp55 overexpression could even enhance RRE-dependent Gag expression in HeLa cells, though to a lesser extent than in 3T3 cells (Fig. (Fig.6C).6C). Finally, as with 3T3 cells, overexpression of the human SR proteins had minimal impact on the expression of either luciferase (see supplemental Fig. S7A) or codon-optimized Gag-Pol (see supplemental Fig. S7B) in HeLa cells, again demonstrating specificity for natural HIV-1 mRNA substrates.
To map the domains in SR proteins that are necessary for stimulating intracellular Gag expression, we analyzed a series of chimeras between SRp40 and SF2/ASF. As shown in Fig. Fig.7A,7A, both proteins have similar domain organizations of two RNA recognition motifs (RRMs) and carboxy-terminal domains consisting predominately of arginine and serine repeats (the RS domain) that may also bind RNA and/or interact with protein cofactors (reviewed in references 47 and 85). Despite similarities in domain structure, overexpression of SRp40 is substantially more potent at inducing expression from pGag-Pol-RRE in 3T3 cells, even though SF2/ASF is expressed at higher levels (Fig. (Fig.7B,7B, lanes 2 and 3). Each domain was individually swapped between the two proteins.
In the background of SF2/ASF, changing the RS domain or RRM2 to that of SRp40 partially increased Gag expression (compare lane 2 to lanes 4 and 5), while changing RRM1 had no effect (lane 6). Similarly, in the context of SRp40, changing RRM1 to that of SF2/ASF had no effect (compare lane 3 to lane 7), while changing the RS domain to that of SF2/ASF substantially decreased its ability to enhance intracellular Gag abundance (compare lane 3 to lane 9). The chimera containing just RRM2 from SF2/ASF was poorly expressed, and therefore, its effect was not interpretable (lane 8). When the RS domain of SRp40 was deleted, a modest amount of activity was lost, though an analogous deletion in SRp55 resulted in a dramatic loss of function (lanes 11 and 12). Similar results were observed when pGag-RRE was tested as the substrate. In sum, the RS domains of SRp40 and SRp55 play central roles in the induction of HIV-1 translation, and the RS domain, together with RRM2, at least for SRp40 and SF2/ASF, confers specificity.
The nuclear history of a cellular mRNA is critical for its stability and translational competency in the cytoplasm. The mRNA maturation steps of 5′ capping, pre-mRNA splicing, and 3′ polyadenylation are all essential for efficient translation; for instance, the 5′ cap and 3′ poly(A) tail act as scaffolds for proteins that mediate translation initiation (reviewed in reference 69). Many viral RNAs are not modified in the same manner as cellular mRNAs, and this presents a conundrum for achieving efficient translation of viral proteins. For viral RNAs that do not have a 5′ cap and 3′ poly(A) tail, one well-characterized solution to this problem is the exploitation of an internal ribosomal entry site (IRES) (69). How other viral mRNAs, particularly those that are synthesized in the nucleus but are not processed by splicing, undergo efficient translation in the absence of proteins recruited during splicing is mostly unknown.
As approximately 2,400 Gag molecules are required for each virion (13), efficient Gag translation is a prerequisite for viral replication. In the absence of EJC deposition, it is predicted that the HIV-1 gRNA must employ alternative mechanisms to ensure efficient translation of Gag and Gag-Pol. In murine cells, Gag is very inefficiently translated, and we hypothesize that one or more cellular proteins that mark the gRNA for productive translation do not do so effectively. To identify such factors, we used this nonpermissive system to screen human RNA binding proteins for their ability to increase Rev-dependent Gag expression. Overexpression of SR proteins profoundly increased intracellular Gag abundance, reproducibly in excess of 50-fold in the case of SRp55 (Fig. (Fig.1).1). The SR protein family is well established as being necessary for constitutive and alternative splicing of all pre-mRNAs and also for playing roles in the regulation of many other steps of mRNA metabolism, including transcription, 3′ end formation, nuclear export, translation, and NMD (47, 85). We therefore conclude that SR proteins act to mark the HIV-1 gRNA for productive translation.
Overexpression of RNA binding proteins in 3T3 cells has the potential to rescue HIV-1 Gag expression in two ways. First, species-specific sequence differences between the human and mouse orthologs may render the mouse ortholog incapable of serving as a cofactor for HIV-1 replication: adding the human version would rescue the defect. Second, many RNA binding proteins are maintained at limiting, local concentrations such that overexpression of RNA binding proteins has been recognized to enhance their normal function: for instance, HuR and hnRNP D/AUF1 overexpression can modulate the stability of AU-rich-element (ARE)-containing RNAs, and Staufen overexpression enhances cytoplasmic mRNA trafficking in neurons (23, 46, 59, 74). To date, we have not detected substantial differences in activity between human and mouse orthologs of SRp40 and SRp55 in promoting Gag translation in 3T3 cells (see supplemental Fig. S3), though minor variations are apparent on occasion. Also, overexpression of human SR proteins enhanced Gag expression from 1xCTE- and RRE-containing transcripts in HeLa cells (Fig. (Fig.6),6), leading us to propose that the availability of SR proteins for promoting gRNA translation is broadly limiting but that this deficiency is most apparent for RRE-mediated export in 3T3 cells and 1xCTE-mediated export in human cells. The molecular basis for these dramatic phenotypes is unknown but has allowed us to demonstrate that HIV-1 translation intersects with SR protein function.
Moreover, it is clear that there is specificity for SRp40/55-mediated enhancement of HIV-1 translation since these proteins did not enhance the expression of codon-optimized Gag-Pol or luciferase (Fig. (Fig.4;4; see also supplemental Fig. S6 and S7). We do not know which cis-acting elements in Gag-Pol regulate this specificity, though multiple sequences in Gag-Pol have previously been implicated in negatively regulating Gag expression via decreased RNA stability or translation efficiency, often in the absence of a functional retroviral export element (reviewed in reference 61). How, or if, these negative elements regulate Rev-dependent Gag expression in murine cells is unclear, but SR protein overexpression may be able to counteract the restriction they impart. This would be consistent with the observation that SRp55 overexpression increases both Gag RNA abundance and translation efficiency (Fig. (Fig.22 and and33).
All SR proteins can modulate splicing, but it is unknown how different family members may regulate translation (47, 85). Our results showing that SRp40 and SRp55 are potent inducers of HIV-1 translation may offer a system in which to address this question. While all SR proteins tested are able to enhance HIV-1 Gag expression at least moderately, SRp40 and SRp55 have the most activity. They, along with SRp75, form a specific subfamily of SR proteins (3), though it is not known how the varied functions described for SR proteins are conserved within subfamilies. SR proteins can bind RNA through their RRM and RS domains and also interact with a number of proteins, such as U1-70K, U2AF35, mTOR, and NXF1, though there is specificity in the interactions with RNA targets and protein cofactors (47, 85).
One initial motivation for screening the SR proteins was their interaction with NXF1. We had speculated that 4xCTE-dependent Gag translation may be more efficient than RRE-dependent Gag translation because 4xCTE recruits SR proteins via NXF1. In support of this hypothesis, NXF1-binding SR proteins, such as SRp20 and 9G8, have been reported to enhance 1xCTE-dependent Gag translation (73). However, our data show that the connection between retroviral nuclear export elements and efficient translation is likely more complicated. In particular, SRp40 and SRp55 enhance Gag production regardless of the export element used (Fig. 1B and C and and6A).6A). Moreover, there is no correlation between an SR protein's capacity to increase Gag expression and its ability to shuttle: SRp40 does not shuttle and yet enhances 1xCTE- and RRE-dependent Gag expression more efficiently than SF2/ASF, SRp20, or 9G8, all of which shuttle and can bind NXF1 (12, 29, 44, 66).
To identify domain determinants of SR protein activity, we analyzed a panel of chimeras between SF2/ASF and SRp40 (Fig. (Fig.7).7). Both proteins have two RRMs and a RS domain. The identity of the first RRM did not affect the protein's activity. In terms of increasing SF2/ASF activity, swapping either RRM2 or the RS domain with that of SRp40 had a similar partial effect; only when both regions from SRp40 were present did the chimera act like SRp40. Also, deletion of the RS domains of SRp40 or SRp55 decreases their abilities to enhance Gag expression. Further studies will be required to determine precisely which functional elements of RRM2 and the RS domain, which could mediate RNA binding specificity or cofactor recruitment, underlie the differential activities of SR proteins.
Very little is known about the translational regulation of HIV-1 Gag, and even the mode of initiation is controversial, with some groups proposing a cap-dependent model and others implicating the involvement of an IRES (reviewed in reference 9). There are at least three different mechanisms by which SR proteins can regulate translation initiation. First, SR proteins increase the efficiency of the pioneering round of translation, thereby increasing overall protein production and the efficiency of NMD if a premature stop codon is present in the appropriate context (67, 83). The pioneering round of translation is biochemically distinct from the steady-state translation and is dependent on CBP80-20 and CTIF instead of eIF4E (33, 41). A role for SR proteins in enhancing the pioneering round of Gag translation would be consistent with our finding that nonshuttling SR proteins are functional since the pioneering round of translation has been reported to occur immediately after nuclear export, possibly when the mRNA is still associated with the nucleus (34). Accordingly, SR proteins that do not shuttle (as measured in heterokaryon assays) may actually accompany an mRNA through the nuclear pore and enhance translation without ever fully dissociating from the nucleus.
A second mechanism by which SR proteins can regulate translation initiation is operative during steady-state translation. SF2/ASF binds the mTOR kinase to modulate the phosphorylation of 4E-BP1 and allow eIF4E-dependent translation (54). While SF2/ASF is a much less marked enhancer of Gag translation than SRp40 or SRp55, we note the RRM2 from all three proteins contains the SWQDLKD motif that is necessary to bind mTOR. Third, SRp20 has been reported to enhance IRES-dependent translation initiation (5). SRp20 is also not very active in our experiments; however, it is unknown if other SR proteins can also modulate this type of translation. Our finding that SR proteins induce HIV-1 Gag translation will be a useful system for addressing fundamental questions such as whether HIV-1 Gag translation is cap dependent or independent, whether the pioneering or steady-state round of gRNA translation is inefficient in the nonpermissive systems we have described, and whether the mTOR pathway regulates Gag translation.
gRNA regulation diverges from that of most cellular mRNAs, which may present novel opportunities for antiretroviral drug development. Small molecule inhibitors have been developed for SR proteins, and these have been proposed to be antiretroviral compounds, though mostly as splicing inhibitors (2, 40, 70). Identifying the mechanism by which SRp40/55 regulate HIV-1 Gag translation may lead to the development of specific small molecule inhibitors that inhibit HIV-1 protein production.
We thank Adrian Krainer for the T7-tagged SR protein constructs, Eric Freed for NL4-3(PR D25A), Marie-Louis Hammarskjold for NXF1 and NXT1 cDNA constructs, and John Gray for NL4-3 codon-optimized Gag-Pol. We also thank Edmund Newman for cDNA generated from RNA extracted from 3T3 cells.
This work was supported by the United Kingdom Medical Research Council. C.M.S. is a Research Councils United Kingdom Academic Fellow, N.M.S. is a Long-Term Fellow (ALTF 176-2007) of the European Molecular Biology Organization, and M.H.M. is an Elizabeth Glaser Scientist.
Published ahead of print on 28 April 2010.