Human immunodeficiency virus type 1 splicing at the major splice donor site is controlled by highly conserved RNA sequence and structural elements

Transcription

1 Journal of eneral Virology (2015), 96, DOI /jgv Short ommunication Human immunodeficiency virus type 1 splicing at the major splice donor site is controlled by highly conserved RN sequence and structural elements Nancy Mueller, Bep Klaver, Ben Berkhout and tze T. Das orrespondence tze T. Das a.t.das@amc.uva.nl Laboratory of Experimental Virology, Department of Medical Microbiology, enter for Infection and Immunity msterdam (INIM), cademic Medical enter, niversity of msterdam, msterdam, The Netherlands Received 11 June 2015 ccepted 16 September 2015 Human immunodeficiency virus type 1 (HIV-1) splicing has to be strictly controlled to ensure the balanced production of the unspliced and all differently spliced viral RNs. Splicing at the major 59 splice site (59ss) that is used for the synthesis of all spliced RNs is modulated by the local RN structure and binding of regulatory SR proteins. Here, we demonstrate that the suboptimal sequence complementarity between this 59ss and 1 small nuclear RN (snrn) also contributes to prevent excessive splicing. nalysis of a large set of HIV-1 sequences revealed that all three regulatory features of the 59ss region (RN structure, SR protein binding and sequence complementarity with 1 snrn) are highly conserved amongst virus isolates, which supports their importance. ombined mutations that destabilize the local RN structure, remove binding sites for inhibitory SR proteins and optimize the 1 snrn complementarity resulted in almost complete splicing and accordingly reduced virus replication. Human immunodeficiency virus type 1 (HIV-1) produces a single primary RN transcript. The full-length transcript functions as RN genome that is packaged into virions and as mrn for translation of the ag and Pol proteins. HIV- 1 RN contains several splice donor (SD) [59 splice site (59ss)] and splice acceptor [39 splice site (39ss)] sites. Differential usage of these sites results in the production of a variety of spliced RNs that encode the other viral proteins. Splicing has to be strictly regulated for the balanced production of all viral RNs and proteins. Formation of the spliceosome complex is initiated by binding of 1 small nuclear ribonucleoprotein (snrnp) to the 59ss, which is mediated by the complementarity between the 11 single-stranded nucleotides at the 59 end of the 1 small nuclear RN (snrn) component and the 59ss sequence (Freund et al., 2003). The major 59ss located in the 59 untranslated leader of HIV-1 transcripts is used for the production of all spliced mrns. s both spliced and unspliced transcripts are required for a productive replication cycle, splicing at this 59ss has to be controlled to prevent complete processing. Previous studies indicated that both RN structure and sequence elements are involved in this control. The 59ss region can fold a stem loop structure, the splice donor (SD) hairpin, that partially Two supplementary tables are available with the online Supplementary Material. occludes the 1 annealing site and reduces splicing frequency (bbink & Berkhout, 2008; Mueller et al., 2014). Splicing is also influenced by binding of several SR proteins to splice regulatory elements (SREs) in the SD region (sang et al., 2012; Mueller et al., 2015). The SD hairpin structure may also influence binding of these SR proteins. In addition, the extent of sequence complementarity between the 59ss and 1 may influence snrnp binding and splicing efficiency. Previous studies demonstrated that viral RN structures and sequence motifs with an important role in viral replication have been highly conserved during HIV-1 evolution. Therefore, the major 59ss region of different HIV-1 isolates was analysed for the capacity to form an SD hairpin stem loop structure, presence of putative SR protein-binding sites and sequence complementarity with 1 snrn. In total, 1184 sequences of HIV-1 group M isolates described in the Los lamos HIV sequence database (Web alignment 2012; were analysed and compared with the subtype B LI strain that is used as prototype in our studies (Table S1, available in the online Supplementary Material). The ability to form an SD stem loop structure was analysed by mfold RN structure analysis (Zuker, 2003). The majority of the isolates (780 sequences) have an SD hairpin sequence (nt ) identical to that of the LI strain and can fold the same structure (Fig. 1a, structure ). Most isolates The uthors Printed in reat Britain 3389

2 N. Mueller and others (a) 5v'ss E11 TH2 M TH2 p 4 2/B77/E6/E /205/BF3 p 2 B42a B52/.D B D E B27/E6 131 B45 B13 E5 B2/159 B2 E5 21 B2 B65/73/B1.D 23 E4/B1/B B75 1D11 47/90 BF1/BF17 B20/B66 BF1/BF17 F B20 BF17 D3 3 H I J K L 1D1/ 1D2 D16 D13 DF1 D16 DF1 D9/D10/ 1D M H1/1DHK H1/H2/H3 4 N 1 O P Q R S T V (b) 5'ss sequence Structure Isolates SR protein-binding site SRSF2 (1) SRSF2 (2) SRSF2 (3) SRSF5 SRSF6 (1) SRSF6 (2) HBS F P K (c) SRSF5 72.6% SRSF2 (2) 98.2% SRSF6 (1) 99.8% SRSF6 (2) 95.6% 278 SRSF2 (1) SRSF2 (3) % 95.4% (d) 3' 1 snrn 5' Fig. 1. Evolutionary conservation of the structure and sequence characteristics of the major 59ss. (a) The capacity of the major 59ss RN region of 1184 HIV-1 group M isolates described in the Los lamos sequence database (Web alignment 2012) to fold an SD stem loop structure was analysed with mfold RN structure analysis software. The SD hairpin structure 3390 Journal of eneral Virology 96

3 Regulation of HIV-1 splicing for the HIV-1 LI strain is shown (structure ). The identical sequence and structure is observed for 780 isolates. The sequence and structure of different hairpin prototypes is shown (structures B V) with nucleotides differing from the HIV-1 LI sequence boxed and nucleotide insertions encircled. The frequency of every structure variant is indicated. Observed nucleotide differences in the prototypic sequences are indicated with the abbreviations referring to the sequence names shown in Table S1. The 59ss nucleotides to which the 1 snrn binds, are encircled in grey (cleavage site indicated by arrowhead). In two isolates with structure (with an R variation at position 286) an additional base pair (not shown) can be formed in the lower stem region as a result of a nucleotide co-variation at positions 281 and 301 (59 isolate: ) or a single nucleotide difference at position 281 (56 isolate; ). (b, c) The major 59ss region (nt in LI) was analysed with ESEfinder for the presence of putative SR protein-binding sites and with the HBond Score software to calculate the hydrogen bond score (HBS) of the 11 nt 59ss sequence (highlighted in grey; exon and intron nucleotides indicated in upper and lower case, respectively). (b) The 10 most frequently observed sequences are shown. Nucleotides that differ from the most frequently observed sequence (corresponding to the HIV-1 LI strain; upper row) are indicated in bold and underlined. The number of isolates with the corresponding sequence, the predicted structure ( V, as shown in a), the presence (+) or absence (2) of putative SR protein-binding sites and the HBS value are indicated. The complete analysis of all sequences is shown in Table S2. (c) The position of the putative SR protein-binding sites in the HIV-1 LI SD region and the percentage of virus isolates carrying each site are shown (59ss nucleotides to which the 1 snrn binds are boxed in grey). (d) Sequence variation in the 59ss motif observed in a collection of human 59ss, as described by Roca et al. (2013). The height of each nucleotide reflects its conservation at the corresponding position. with a single or a few nucleotide differences can fold the same or a very similar hairpin structure (Fig. 1a, structures B V). The most frequently observed mutation is an R change at position 286, which results in a instead of an base pair (Fig. 1a, structure ). Several isolates have nucleotide differences that at first glance are not compatible with the SD hairpin structure, but mfold analysis revealed that they can fold a slightly different hairpin structure (Fig. 1a, structures J V)thatpresentsthe59ss in a very similarwayintheloop.inmostsubtypedisolates(and recombinants thereof) a R mutation at position 282 is combined with a R at position 300, which results in a base pair co-variation (Fig. 1a, structures L: to base pair). Thus, despite significant sequence variation in the 59ss region, nearly all group M isolates can fold an SD hairpin structure that partially occludes the 1 snrn annealing site in the base-paired stem. sang et al. (2012) previously used ESEfinder, an algorithm based on SELEX (systematic evolution of ligands by exponential enrichment)-identified RN binding sites for splicing regulatory proteins (artegni et al., 2003; Smith et al., 2006), to identify potential binding sites for the SR proteins SRSF2 (S35), SRSF5 (SRp40) and SRSF6 (SRp55) in the major 59ss region. Binding of the SRSF2 protein was confirmed experimentally and shown to reduce splice site usage, but the other interactions remain putative. ESEfinder analysis of the HIV-1 group M sequences indicated that the previously described SR protein-binding sites are present in 776 of 1184 isolates (Fig. 1b, c, Table S2). The SRSF6 (1) site is observed in nearly all isolates (99.8 %), whereas the overlapping SRSF6 (2) site is detected in 95.6 % of the isolates. ll but one of the 1184 isolates have at least one of these SRSF6 sites. Similarly, the overlapping SRSF2 (1) and SRSF2 (2) sites are detected in 92.1 and 98.2 % of the isolates, respectively, with 98.9 % of the isolates having at least one of these SRSF2 sites. The SRSF2 (3) and SRSF5 sites are observed in 95.4 and 72.6 % of the isolates, respectively. Thus, the presence of these putative SR protein-binding sites is highly conserved amongst group M isolates, with the possible exception of the SRSF5 site. The 11 nt 59ss sequence to which 1 snrn binds varies only slightly amongst different HIV-1 isolates (Fig. 1b, Table S2). The most frequently observed sequence ( ) deviates only at position 22 from the consensus sequence for eukaryotic 59ss elements (Fig. 1d). Whereas most HIV-1 isolates have a at position 22 that cannot pair with 1 snrn (Fig. 1c),.50 % of eukaryotic 59ss elements have an that can interact with 1 (Fig. 1d). The hydrogen bond score (HBS) (Freund et al., 2003) reflects the base-pairing potential between the 1 snrn and the three exonic (nt 23 to 21) and eight intronic nucleotides (nt +1 to +8) at the 59ss. The HBS can vary from 1.8 for a 59ss that contains only the essential nucleotides to 23.8 for a 59ss with optimal complementary to 1. To compare the complementarity between 1 and the major 59ss of the different HIV-1 isolates, we calculated the HBS for each variant with the HBond Score Web-Interface ( The most frequently observed 59ss sequence has an HBS value of 17.5 (Fig. 1b, Table S2), which reflects a high, but incomplete sequence complementarity with 1. Due to the small sequence variation in the 59ss sequence, the HBS score varied slightly for other HIV-1 isolates, and a value between 17.3 and 17.7 was calculated for 98.9 % of the M group isolates (1171 of 1184). These results demonstrate that the sequence complementarity between 1 snrn and the 59ss is also highly conserved during HIV-1 evolution. Our in silico analysis demonstrated that the SRSF2-binding sites are highly conserved amongst HIV-1 isolates, which indicates a role for these SREs in splicing control, although in a previous study the deletion of the SRSF2 (1) site did not

4 N. Mueller and others (a) LTR 3 R 5 5 ss gag gag luciferase 3 ss luciferase Spliced RN nspliced RN DN (b) 5 ss WT WT 4 1 snrn a 3b 3a 3b SDa SDaa SDa SDaa (c) SR protein-binding site SRSF2 (1) SRSF2 (2) SRSF2 (3) SRSF5 SRSF6 (1) SRSF6 (2) Δ kcal mol 1 HBS nspliced (%) Spliced (%) WT a b SDa SDaa (d) 3 Luciferase (RL) 2 1 No Tat Plus Tat 0 WT 4 3a 3b SDa SDaa Fig. 2. Mutation of the SD hairpin region affects splicing. (a) In the pltr-gag-flag-luc construct the HIV 59 LTR and leader RN sequences including the major 59ss (with +1 indicating the transcription start site) are coupled to the gene encoding a fusion protein consisting of the N-terminal 25 aa of ag, Flag-tag peptide and firefly luciferase. The luciferase transcript contains a cryptic 39ss. The unspliced RN encodes an active luciferase protein, whereas the spliced RN does not. The sequences of the WT and mutant 59ss regions are shown. The 59ss nucleotides to which the 1 snrn anneals are boxed in grey (cleavage site indicated by arrowhead). Nucleotide substitutions are boxed in black. (b) The structure of the WT and mutant SD hairpins as predicted with the mfold RN structure analysis software are shown. The 59ss nucleotides are encircled in grey. (c) The WT and mutant 59ss regions were analysed with ESEfinder. The presence (+) or absence (2) of the putative SR protein-binding sites is shown. In addition, the RN structure stability (D; structures shown in b), HBS 3392 Journal of eneral Virology 96

5 Regulation of HIV-1 splicing value and splicing efficiency (estimated level of unspliced and spliced RN) of the WT and mutant SD sites are shown. (d) 293T cells were transfected with one of the pltr-gag-flag-luc constructs with or without a Tat-expressing plasmid. The intracellular luciferase production was measured after 48 h. Data represent mean SD relative light units (RL) of three experiments. Statistical analysis demonstrated that the mutant values differed significantly (P,0.05) from the WT value (both with and without Tat). affect splicing efficiency when tested in the context of a destabilized SD hairpin structure (Mueller et al., 2015). However, two putative SRSF2 sites [SRSF2 (1) and SRSF2 (2) ] are present upstream of the 59ss. We therefore generated novel SD mutants in which both SRSF2 sites were inactivated by introducing 3 or 4 nt substitutions on the left side of the SD hairpin stem (3b and 4 mutants; Fig. 2). In addition, we generated the 3a mutant with a 3 nt substitution that removed only the SRSF2 (1) site. These mutants fold the same destabilized RN structure (D525.2 kcal mol 21 ). Splicing analysis in the HIV-1 virus context is very complicated as 59ss mutations will affect expression of all viral proteins, including the viral Tat and Rev proteins that influence RN production (transcription) and processing (nuclear export). We therefore used a previously described LTR-driven leader RN luciferase construct (Mueller et al., 2014) to analyse the effect of SD mutations on 59ss usage (Fig. 2a). We previously showed that SD hairpin mutations that decrease or increase splicing in this simplified reporter delay virus replication in the complete HIV-1 genome context. Furthermore, these mutations only affected the local RN structure and did not affect other gene expression processes, such as RN production, polyadenylation and translation. The reporter constructs with a WT or mutated SD element were transfected into 293T cells as described previously (Mueller et al., 2014). The cells were co-transfected with a Tat-expressing plasmid to activate transcription from the LTR promoter. The effect of the mutations on splicing was analysed by measuring the luciferase activity in transfected cells. Whereas the unspliced RN encodes an active luciferase protein, the spliced RN does not (Fig. 2a). Incomplete splicing of the WT transcript results in detectable luciferase expression (Fig. 2d). Mutants 3a, 3b and 4 showed a decrease in luciferase activity, which reflects increased splicing. Similar effects were observed when cells were transfected without Tat plasmid, when a lower level of transcription was established by basal LTR promoter activity. Previously, we demonstrated that,85 % of the transcripts containing a WT SD hairpin are spliced, which results in,15 % unspliced RN (Mueller et al., 2014). The observed luciferase levels (Fig. 2d) were used to estimate the unspliced and spliced RN level of the mutant constructs (unspliced RN mutant 5 RL mutant /RL WT 615 %; spliced RN mutant 5100 %2unspliced RN mutant %). Splicing of the 3a mutant was slightly increased to,90 % (,10 % unspliced RN), whereas the 3b and 4 mutants showed a further increase to,94 95 % splicing (,5 6 % unspliced RN; Fig. 2c). These latter mutants, lacking both repressive SRSF2-binding sites upstream of the 59ss, thus show the highest level of splicing. s the thermodynamic stability of all mutant RN structures is predicted to be the same, it seems likely that removal of both splicing-inhibitory SRSF2 sites was responsible for the increased splicing frequency of the 3b and 4 mutants. These results confirm that 59ss splicing is inhibited by binding of SRSF2 proteins in the SD region. Removal of both binding sites did not result in complete splicing, which indicates that other limiting factors are present. The suboptimal sequence complementarity between the 59ss and 1 snrn may be one of the factors that cause incomplete splicing. To investigate this possibility, we constructed two SD mutants with increased 1-binding capacity. In the SDa mutant, 22 was replaced by to increase the HBS to 21.1 (Fig. 2). This mutation slightly destabilized the SD hairpin structure (D527.9 kcal mol 21 ; Fig. 2c). In the SDaa mutant, in addition the 23 was replaced by, because an is frequently observed at this position in eukaryotic 59ss (Fig. 1d). This double mutation resulted in an HBS of 19.4 and further destabilized the hairpin structure (D525.5 kcal mol 21 ). ESEfinder analysis indicated that the SDa and SDaa mutations inactivated both putative SRSF2-binding sites upstream of the cleavage site (Fig. 2c). ells were transfected with WT or SD-mutated luciferase constructs and the Tat plasmid, and luciferase production was measured after 48 h. Both mutants showed around fivefold decrease in luciferase activity, which reflects a significant increase in splicing compared with WT (Fig. 2d). Similar effects were observed when the analysis was performed at basal transcription levels without Tat. Based on these luciferase levels, we estimate that splicing of the SDa and SDaa mutants is increased to,97 %, yielding only,3 % unspliced RN (Fig. 2c). This high splicing level can be explained by the combinatorial effects: the improved 59ss 1 snrn complementarity, removal of both SRSF2 (1) and SRSF2 (2) sites, and a destabilized RN structure that results in a better 59ss exposure. Splicing of the SDa mutant with a slightly destabilized SD hairpin (D527.9; HBS 21.1; 97 % splicing) was higher than that of 3b and 4 that lack the same SRSF2 sites and have a more dramatically destabilized RN structure (D525.2; HBS 17.5; % splicing). This difference in splicing frequency suggests an important role of the 59ss 1 snrn complementarity (reflected by the HBS) in splicing efficiency. Splicing of the SDaa mutant (D525.5; HBS 19.4; 97 % splicing) was also higher than that of mutants 3b and 4, whilst the RN

6 N. Mueller and others (a) teto TR m 3 R 5 5' LTR (b) -p24 (ng ml 1 ) 100 5'ss gag pol vpr vpu tat rev rtt structures varied only slightly in stability, which confirms the important role of 59ss 1 snrn complementarity. Taken together, these results indicate that HIV-1 splicing at the major 59ss is restricted by local RN structure, binding of SRSF2 and a suboptimal 59ss sequence. By removing these restrictions, we generated a very efficient splice site that mediated nearly complete RN splicing. We previously demonstrated that SD hairpin mutations that either increase or decrease the splicing frequency reduce the HIV-1 replication capacity (Mueller et al., 2014). Such effects were evident in sensitive virus competition experiments, where WT virus consistently outcompeted the mutant upon prolonged culturing. When replication of the WT and mutant HIV-1 variants was tested in parallel cultures, the negative effect of the mutations on viral fitness was sometimes more difficult to observe. This is possibly due to our in vitro virus culture system, which is based on vif Day 4 Day 5 WT 4 3b SDa env SDaa teto TR m 3 R 5 3' LTR Fig. 3. Mutations that alter the splicing frequency reduce virus replication. (a) Proviral genome of the Dox-controlled HIV-rtT variant. In HIV-rtT, the Tat-TR transcriptional activation mechanism was inactivated by mutations in Tat and TR (TR m ) and functionally replaced by the Dox-inducible transcription mechanism by introducing teto-binding sites in the 3 promoter region and by replacing the nef gene with the rtt gene. s a consequence, transcription and replication of this virus are dependent on Dox administration. (b) SupT1 T-cells were infected with equal amounts of WT and mutant virus (-p24 input: 3 ng ml 21 ) and cultured with 25 ng Dox ml 21 to activate viral replication. Virus replication was monitored by measuring the -p24 level in the culture supernatant at 4 and 5 days after infection. Data represent mean SEM of four experiments. J2 the efficiently replicating HIV-1 LI strain and the SupT1 T-cell line that may only allow the detection of relatively large differences in replication capacity. We therefore tested the effect of the SD mutations on virus replication in the context of a doxycycline (Dox)-dependent HIV-1 variant. In this HIV-rtT variant, the viral Tat-TR transcription activation mechanism was inactivated and functionally replaced by the components of the Dox-inducible Tet-On gene expression system (Das et al., 2004a) (Fig. 3a). HIV-rtT replication can be modulated by varying the Dox concentration, lacking replication in the absence of Dox and very efficient replication at 1000 ng Dox ml 21 (Das et al., 2004b). To detect small replication differences, we used a suboptimal Dox level. SupT1 cells were therefore infected with the WT and SD-mutated HIV-rtT variants andculturedat25ngdoxml 21. The J2 mutation that stabilizes the SD hairpin and reduces splicing (Mueller et al., 2014) was included as control. When compared with the HIV-rtT with a WT SD region, infection with the J2 variant resulted in a,40 % reduced -p24 level in the culture supernatant at 4 and 5 days after infection (Fig. 3b), which is in agreement with the previously observed replication defect of the corresponding HIV-1 mutant (Mueller et al., 2014). Replication of the 4, 3b, SDa and SDaa mutated HIV-rtTs resulted in a % lower -p24 level compared with the WT HIV-rtT, which shows that these SD mutations (that increase the splicing frequency) decrease virus replication. Taken together, the present and previous studies demonstrate that splicing at the major 59ss is regulated by the stability of the local RN structure, the presence of SREs to which SR proteins can bind, and the base-pairing potential between the 59ss and the 1 snrn. The analysis of a large set of HIV-1 sequences revealed that all three regulatory features of the 59ss region are highly conserved in different group M viruses, which supports an important function during virus replication. Modulation of the splicing frequency reduces HIV-1 replication and, as previously discussed (Mueller et al., 2015), the HIV-1 splicing process is an interesting target for antiviral therapy. References bbink, T. E. & Berkhout, B. (2008). RN structure modulates splicing efficiency at the human immunodeficiency virus type 1 major splice donor. J Virol 82, sang,., Erkelenz, S. & Schaal, H. (2012). The HIV-1 major splice donor D1 is activated by splicing enhancer elements within the leader region and the p17-inhibitory sequence. Virology 432, artegni, L., Wang, J., Zhu, Z., Zhang, M. Q. & Krainer,. R. (2003). ESEfinder: a web resource to identify exonic splicing enhancers. Nucleic cids Res 31, Das,. T., Verhoef, K. & Berkhout, B. (2004a). conditionally replicating virus as a novel approach toward an HIV vaccine. Methods Enzymol 388, Das,. T., Zhou, X., Vink, M., Klaver, B., Verhoef, K., Marzio,. & Berkhout, B. (2004b). Viral evolution as a tool to improve the 3394 Journal of eneral Virology 96

7 Regulation of HIV-1 splicing tetracycline-regulated gene expression system. J Biol hem 279, Freund, M., sang,., Kammler, S., Konermann,., Krummheuer, J., Hipp, M., Meyer, I., ierling, W., Theiss, S. & other authors (2003). novel approach to describe a 1 snrn binding site. Nucleic cids Res 31, Mueller, N., van Bel, N., Berkhout, B. & Das,. T. (2014). HIV-1 splicing at the major splice donor site is restricted by RN structure. Virology , Mueller, N., Berkhout, B. & Das,. T. (2015). HIV-1 splicing is controlled by local RN structure and binding of splicing regulatory proteins at the major 59 splice site. J en Virol 96, Roca, X., Krainer,. R. & Eperon, I.. (2013). Pick one, but be quick: 59 splice sites and the problems of too many choices. enes Dev 27, Smith, P. J., Zhang,., Wang, J., hew, S. L., Zhang, M. Q. & Krainer,. R. (2006). n increased specificity score matrix for the prediction of SF2/SF-specific exonic splicing enhancers. Hum Mol enet 15, Zuker, M. (2003). Mfold web server for nucleic acid folding and hybridization prediction. Nucleic cids Res 31,