Repair of a Rev-Minus Human Immunodeficiency Virus Type 1 Mutant by Activation of a Cryptic Splice Site

ABSTRACT We isolated a revertant virus after prolonged culturing of a replication-impaired human immunodeficiency virus type 1 (HIV-1) mutant of which the Rev open reading frame was inactivated by mutation of the AUG translation initiation codon. Sequencing of the tat-rev region of this revertant virus identified a second-site mutation in tat that restored virus replication in the mutant background. This mutation activated a cryptic 5′ splice site (ss) that, when used in conjunction with the regular HIV 3′ ss #5, fuses the tat and revreading frames to encode a novel T-Rev fusion protein that rescues Rev function. We also demonstrate an alternative route to indirectly activate this cryptic 5′ ss by mutational inactivation of an adjacent exon splicing silencer element.

Human immunodeficiency virus type 1 (HIV-1) transcripts contain multiple splicing signals that are utilized by the cellular splicing machinery to produce a large set of subgenomic mRNAs. More than 30 mRNA species are produced by differential usage of the 5Ј and 3Ј splice sites (ss) that are present in the HIV pre-mRNA (reviewed in reference 20). This splicing process is highly regulated by mechanisms that control the efficiency by which these splicing signals are recognized by components of the splicing machinery (19). For instance, suboptimal branch point sequences are used for lariat formation (12), and splicing enhancer and silencer elements regulate the activity of splice signals (1,2,22). The viral Rev protein stimulates the nuclear export of unspliced and partially spliced HIV-1 transcripts through its binding to a structured RNA motif termed the Rev-responsive element (RRE) located in the env gene (20). At low Rev levels, primary HIV transcripts will be multiply spliced, producing transcripts that encode the early viral proteins Tat, Nef, and Rev itself. Upon accumulation of Rev protein in the nucleus, unspliced and singly spliced transcripts appear in the cytoplasm. The latter mRNAs encode the Vif, Vpr, Vpu, and Env proteins. The full-length, unspliced HIV RNA serves two functions. As mRNA, it produces Gag and Gag-Pol polyproteins, and as viral genome it is packaged in virus particles. A balanced and properly timed expression of viral transcripts and proteins is required to ensure efficient production of new, infectious HIV virions. The Rev protein is critical for this regulation and is therefore an essential viral function.
This study describes how a replication-impaired virus mutant with an inactivated Rev start codon can be repaired by activation of a cryptic 5Ј ss in the tat gene. This leads to the synthesis of a functional Tat-Rev fusion protein. A large set of mutants was previously generated in a study on the structure and function of the HIV-1 Tat protein (26). The mutant proteins were tested for transcriptional activity in transfections with reporter constructs and in virus replication assays. Some of the mutations that target the tyrosine (Y) residue at codon 47 also affect the overlapping Rev translation initiation codon. For instance, the same Tat amino acid substitution is present in Y47H1 and Y47H2, but only the latter mutation destroys the overlapping Rev start codon (26). The Y47H1 virus replicates suboptimally due to a partially active Tat protein. The Y47H2 virus has the same Tat problem, but this mutant is replication impaired because expression of the essential Rev protein is abolished. We tried to select for virus revertants through prolonged culturing of this replication-impaired Y47H2 virus (25). Despite multiple attempts, we generated only one candidate Y47H2 revertant. Sequencing of the Tat-Rev region identified a second-site mutation (Q17K) in the tat gene, upstream of the original Y47H2 mutation that was maintained (Fig. 1A). The revertant tat gene, with the point mutation that alters Tat codon 17 from CAG (glutamine; Q) to AAG (lysine; K), was subsequently cloned in a eukaryotic expression vector. Transient transfection of this plasmid with a long terminal repeat-chloramphenicol acetyltransferase reporter construct in SupT1 T cells revealed that the Q17K second-site mutation does not improve Tat function (data not shown and reference 25).
Steady-state expression of the wild-type, mutant, and revertant Tat protein was verified by Western blot analysis of lysates from COS cells that were transiently transfected with a Tat/ Rev expression vector (Fig. 1B). Expression of wild-type Tat from this subgenomic expression construct produces two bands ( Fig. 2A, lane 2). The upper band corresponds to the full-length, 86-amino acid Tat protein that is expressed from the spliced transcript (Tat 86). The lower band represents a truncated form of Tat (Tat 72) that is translated from unspliced mRNA. This latter Tat form predominates, due to the Rev-mediated nuclear export of the unspliced mRNA (the RRE is also encoded by this construct We next sought to determine whether the Q17K mutation is able to restore replication of a virus containing the inactivating Y47H2 mutation. To test this, we cloned the revertant tat gene, containing the Y47H2 and Q17K mutations, back into the pLAI molecular clone of HIV-1 and measured virus replication after transfection of SupT1 cells (Fig. 2B). Several appropriate control constructs were transfected in parallel, and CA-p24 production in the culture supernatant was monitored for several days. Wild-type LAI virus spreads rapidly, and this culture was stopped at 5 days posttransfection because of massive, virus-induced cell death. In contrast, we did not measure any virus replication for the Rev-minus Y47H2 virus, which encodes a partially active Tat protein. Delayed virus replication was measured for the Y47H1 virus, which encodes the same Tat mutant but with normal Rev expression. The Y47H2 Q17K double mutant replicated to almost the same extent as the Y47H1 mutant, which is consistent with the finding that Rev expression is restored.
It is not immediately obvious how the Q17K mutation in Tat can rescue Rev expression, as the mutation does not simply create a new translational start codon for the Rev open reading frame. We therefore analyzed the wild-type and revertant sequences for differences in splicing potential. The sequence around codon 17 of both the wild-type and revertant tat gene was entered into a computer program that predicts 5Ј and 3Ј ss (http://www.fruitfly.org). Whereas the wild-type sequence was scored unlikely to be recognized as a splicing signal (P Ͻ 0.10; . .AAGUCAGC. .; with the putative 5Ј ss core nucleotides underlined), the revertant sequence scored as a high probability 5Ј ss (P ϭ 0.91; . . .AAGUAAGC, altered nucleotide in bold). Several downstream 3Ј ss can be used as acceptors to combine with this putative 5Ј ss to generate novel mRNA species (Fig.  3A, top). In fact, usage of the putative 5Ј ss in combination with the downstream 3Ј ss #5 that is normally used to express the Env-Nef genes will fuse the Tat and Rev open reading frames to encode a new Tat-Rev fusion protein (Fig. 3A, bottom). In this way, translation of the functional Rev domains would be under control of the Tat AUG start codon and thus bypass the inactivated Rev start codon.
To test this hypothesis, we transfected SupT1 T cells with molecular clones of wild-type, mutant, and revertant genotype and isolated viral RNA at 2 days posttransfection. Sequences encompassing the tat gene were amplified in a reverse transcription (RT)-PCR (primer positions are indicated in Fig.  3A). A prominent, slowly migrating RT-PCR product was observed for all virus variants (Fig. 3B). The size of this DNA fragment corresponds to the expected size of the regular tat exon 1-2 splice product (using 5Ј ss #4 and 3Ј ss #7), and this was verified by sequence analysis. For the revertant virus, several additional, faster-migrating RT-PCR products were observed (new splices; lane 7). These DNA fragments were excised from gel, amplified, and cloned. Sequence analysis of several clones revealed that the novel 5Ј ss, generated by the Q17K change in the tat gene, was utilized in all cases. The heterogeneity in size of the cDNA fragments stems from the combination of this new 5Ј ss with alternative 3Ј ss 4a, 4b, or 5. Of the 4 clones that were sequenced, the new 5Ј ss was combined twice with 3Ј ss #4b and once with #4a and #5. The splices with 3Ј ss #4a/4b yield mRNAs that encode an internally deleted Tat protein that is likely to be inactive due to the absence of the cysteine-rich domain. The splicing event with 3Ј ss #5 is special in that an mRNA is produced that encodes a Tat-Rev fusion protein that we have coined T-Rev (Fig. 3A). The transcript encodes the first 15 amino acids of Tat, an Arg codon that is generated at the splice, and the nearly complete Rev protein (residues 4 to 116). The T-Rev protein lacks the functional Tat domains but is predicted to have Rev function. T-Rev contains all essential Rev domains: the arginine-rich domain that is required for nuclear localization and binding to the RRE (residues 35 to 50), sequences that mediate Rev multimerization, and the leucine-rich domain (residues 75 to 83) that functions as an activation domain-nuclear export sequence.
These results indicate that a point mutation in the 5Ј part of the tat gene creates a novel splicing signal that rescues Rev expression in the context of the Y47H2 mutant. The wild-type HIV-1 Tat sequence is not utilized as 5Ј ss. This may in part be due to the presence of an exon splicing silencer (ESS) element immediately upstream from this site (Fig. 4A) (1). To test this possibility, we inactivated the ESS element by four base substitutions as indicated in Fig. 4A. Two effects were observed upon inactivation of the ESS enhancer (Fig. 4B). First, there is a selective increase in splicing to the Tat ss, as evidenced by the increase in Tat1 RNA product. Second, a novel splice product was observed by RT-PCR. Cloning and sequencing of the unique cDNA product revealed that the GU dinucleotide at tat codon 16 had served as 5Ј ss in combination with 3Ј ss #5, thereby creating the T-Rev open reading frame.
The analysis of spontaneous revertant viruses that appear in long-term cultures of replication-impaired HIV-1 mutants is a powerful genetic approach to dissect complex regulatory mechanisms and to study nucleic acid and protein motifs that control virus replication (7). We constructed an HIV-1 mutant with a single amino acid substitution in the Tat protein (Y47H2) that partially inactivates Tat function. However, this same mutation inactivates the start codon of the overlapping rev gene (AUG to ACG), thereby causing a severe replication defect. The Tat-Rev coding region of a revertant virus was sequenced. Surprisingly, no repair of the Rev start codon was apparent, but instead a second-site mutation was observed in the upstream Tat sequences, changing codon position 17 in Tat (Q17K). No improved Tat function was measured in transient assays, but we could nevertheless demonstrate that this mutation in the tat gene was responsible for restored virus replication. Furthermore, the reversion mutation was able to restore Rev expression and Rev function as measured in sensitive reporter assays. Interestingly, we detected novel spliced transcripts by RT-PCR, and sequence analysis revealed that a cryptic 5Ј ss was activated in the revertant. The reversion mutation improves the sequence of the cryptic 5Ј ss (AA/GUC AGC to AA/GUAAGC; the consensus 5Ј ss is AG/ GUAAGC). This new 5Ј ss is linked to common 3Ј ss further downstream in the HIV-1 genome. Combination with the vpuenv 3Ј ss (3Ј ss #5) produces a new open reading frame in which the N terminus of Tat (amino acids 1 to 15) is fused to most of Rev (amino acids 4 to 116). Thus, Rev expression is restored by a novel splicing event that places the Rev open reading frame in register with the Tat translational start codon, generating a novel fusion protein that we termed T-Rev. Another Tat-Rev fusion protein has been described previously and was termed Tnv (21) or Tev (4). This Tev-Tnv fusion protein is much different from T-Rev because it fuses the first coding exon of Tat and the second coding exon of Rev, joined together by sequences derived from the env gene. The Tnv-Tev protein exhibits both Tat and Rev activity and can be detected in cells that are latently infected with HIV-1 as well as in T cells that have been transfected with HIV-1 molecular clones.
The T-Rev fusion protein could not be detected directly by Western blotting because the new 5Ј ss is used very inefficiently (approximately 2 to 6% of the regular Tat splice). For instance, we used a monoclonal antibody directed against the Tat Nterminal 15 amino acids, which readily detects the Tat protein itself but not the T-Rev protein (data not shown). Thus, a low level of T-Rev splicing is apparently sufficient to rescue virus replication. It seems likely that more frequent use of the new splice pattern will have a negative impact on virus replication because it will compete with the synthesis of Tat-encoding mRNAs. These results indicate that very little Rev protein is needed, although it is possible that T-Rev translation is more efficient due to the stronger translation start signal provided by the tat AUG than by the usual rev AUG.
We described activation of the cryptic 5Ј ss by a single nucleotide substitution. The cryptic splicing signal is too weak in the wild-type HIV-1 context. This is due in part to the presence of an adjacent ESS element, because the cryptic 5Ј ss Total RNA was isolated, DNase I treated, and used as template in RT-PCR with the HIV-specific 3Ј Tat1 primer (5Ј TTTGAATTCTAATCGAATGGATCTGTCTC 3Ј) and Tat-AUG (5Ј ATGGAGCCAGTAGATCCTAG 3Ј) in a standard 35-cycle PCR. Products were resolved on a sequencing gel (a standard dideoxy sequencing reaction was run in parallel as a marker). Several new, low-abundance splice signals, specific for the Y47H2 Q17K revertant, are indicated. These products were purified from gel, cloned, and sequenced using standard techniques.
can also be triggered by inactivation of this ESS element. Interestingly, a recent study of pseudoexons indicated that inactivation of cryptic ss by ESS elements may be a more general mechanism (24).
Activation of a cryptic 5Ј ss has also been reported upon inactivation of the major 5Ј ss in the untranslated leader (10,19). It thus appears that the HIV-1 genome is riddled with spare ss that may help the virus to overcome deleterious mutations. It has been proposed that the cryptic splice signals may have an active regulatory function. These elements may act as a cis-acting repressor sequence to establish Rev dependence of the viral RNA (10). These combined results underscore both the complexity and plasticity of the HIV-1 genome. Coordinated viral gene expression results from the interplay of multiple regulatory signals in the retroviral genome, but complex gene expression strategies are also likely to be important for the more simple animal retroviruses (3,11,15). Simple retroviruses have certainly evolved strategies for maintaining appropriate levels of splicing. In avian cells, it has been shown that the two 3Ј ss of Rous sarcoma virus (env and src) are both suboptimal because of nonconsensus polypyrimidine tracts or branch point sequences (15,27). In addition, other cis elements upstream and downstream of the 3Ј ss, as well as within the gag gene, are necessary to control the levels of splicing (3,FIG. 4. (A) Schematic of a subgenomic cytomegalovirus-driven cassette expressing the untranslated HIV-1 leader RNA, containing the major 5Ј ss, and coupled to the Tat-Rev coding region. The various 5Ј and 3Ј splice signals in this construct are indicated, as well as a well-characterized ESS element (black rectangle in 5Ј half of the tat gene). Mutant ESS4 contains specific mutations in this element that abolish the silencing function. Horizontal arrows indicate the 5Ј to 3Ј direction of the primers used for the analysis of spliced products in the Tat-Rev coding regions. (B) RT-PCR analysis of HIV-1-specific transcripts from transfected cells. The RT-PCRs were performed essentially as described elsewhere (9). The sequence around the novel splice junction and the new T-Rev open reading frame is indicated. 6,13,14,18,23). In mammalian cell types, a cryptic 5Ј ss within the env gene is activated in the formation of a double-spliced mRNA containing sequences from both the env and src genes (5,17). Although the intricate gene expression circuit of HIV-1 is vulnerable to mutational inactivation, the plasticity of the HIV-1 genome allows subsequent repair strategies by acquisition of spontaneous reverse transcriptase errors and subsequent selection of beneficial second-site mutations. This plasticity is underscored by the forced evolution experiments presented in this and other in vitro studies (8,16), but the plasticity is also apparent from inspection of naturally evolved HIV-1 isolates that belong to the different subtypes (9). In fact, inspection of the Los Alamos HIV-1 sequence database (http: //hiv-web.lanl.gov/) indicates that activation of the 5Ј ss may also occur infrequently in vivo. Among 776 viral sequences, we identified four isolates with the Q17K mutation in the tat gene (UG273, subtype A; 98SE-MP1213, subtype AG; and MVP5180 and BCF06, both group O viruses).