The major splice donor in the 5’LTR is required for FV gagexpression
FVs possess a poly(A) signal and site in each LTR. Consequently, both repression of the poly(A) signal in the 5′LTR and promotion of polyadenylation in the 3′LTR are required to express full-length 5′LTR-derived transcripts. Surprisingly, previous experiments showed that inactivation of the MSD by site-directed mutagenesis resulted in complete loss of gag expression ([29, 32] Löchelt and Bodem, unpublished observation). To confirm these results, we transfected baby hamster kidney (BHK-21) cells with either the proviral pHSRV2 plasmid or the 5′LTR MSD mutant clone pHSRV2-SDm1 [32]. The latter carries a single nucleotide exchange in the MSD (Figure 1A, SDm1). This leaves 6 nucleotides complementary to the cellular U1snRNA but disrupts the continuous binding site of 5 nucleotides in the MSD (Figure 1A). Two days after transfection, cells were harvested, and gag expression was analysed by Western blotting (Figure 1B). The SDm1 mutation was inserted into the proviral pHSRV13 backbone [33] for cloning reasons, and all other proviral constructs of this study were based on pHSRV13, too. As Gag protein levels were undetectable in cells transfected with the pHRSV13-SDm1 construct (Figure 1B), we used Northern blotting to analyse expression of gag-encoding genomic RNAs (size 11 kb). In cells transfected with the pHSRV13-SDm1 plasmid, neither gag-encoding genomic RNA nor pol or env RNA was retrieved (Figure 1C, lane 5), indicating that the mutation in SDm1 might activate cleavage and polyadenylation similar to inactivation of the HIV-1 MSD [16, 18]. Signals below p68 are due to an unspecific reactivity of the serum.
To further investigate the MSD mutant phenotype, we introduced a different single nucleotide mutation (SDm2) into the MSD at the 5′LTR (Figure 1A). SDm2 also encodes 6 nucleotides complementary to U1snRNA (Figure 1A). No LTR-derived transcripts were observed in cells transfected with pHSRV13-SDm2 (Figure 1C, lane 3). To correlate this effect to polyadenylation, we mutated the poly(A) signal in the 5′LTR (wild-type, AAUAAA; p(A)m, UAAUAA) in the wild-type and the SD2 mutant (SDm2+p(A)m). This inactivation of the poly(A) signal should restore expression only if polyadenylation was activated by the MSD mutation (Figure 1A). Transfection of cells with plasmids containing p(A)m resulted in increased expression of LTR-derived transcripts (Figure 1C, lane 2), indicating that some of the transcripts were already polyadenylated at the poly(A) site in the wild type 5′LTR. The inactivation of the poly(A) signal in the 5′LTR in the pHRSV13-SDm2+p(A)m plasmid restored expression of LTR transcripts to wild-type levels, indicating that the pHRSV13-SDm2 mutation might have activated polyadenylation at the 5′LTR (Figure 1C, lane 4). The tas/bet expression was similar in all isolated RNAs, as both genes are expressed from the internal promoter. To analyse influences of a strong MSD, an additional MSD mutant (SDm4), encoding 11 nucleotides complementary to U1snRNA, was generated as well (Figure 1A). In cells transfected with pHSRV13-SDm2, Gag, the Pol precursor and integrase were undetectable by Western blotting, but expression of these proteins was restored by the additional inactivation of the poly(A) signal (Figure 1D, lanes 4 and 5). Cells transfected with pHSRV13-SDm4 expressed pol, but Gag was undetectable (Figure 1D, lane 6). The additional inactivation of the poly(A) signal did not restore Gag expression, which could be assigned to enhanced splicing (data not shown). These results show that the MSD is required for expression of LTR-derived transcripts. In addition, the results with the poly(A) signal mutants support the hypothesis that the MSD is essential for suppression of polyadenylation or RNA cleavage at the 5′LTR.
Mutations in the MSD of the 5’LTR lead to premature cleavage
In order to analyse repression of polyadenylation at the 5′LTR in a quantitative way and to exclude influences of the 3′LTR, we constructed reporter plasmids encompassing the complete pHSRV13 5´LTR encoding either the wild-type MSD or the SDm1 or SDm2 mutants in the pGL3 vector backbone. Thus, the U3 promoter drives firefly luciferase expression (Figure 2A). The resulting construct possesses two poly(A) sites, one in the 5´LTR and a second SV40-derived poly(A) signal 3′ of the luciferase gene. If cleavage at the LTR poly(A) site is suppressed, firefly luciferase should be expressed (Figure 2A and Additional file 1: Figure S1). On the other hand, if the LTR poly(A) signal is active, the RNA should be cleaved at the LTR poly(A) site, and luciferase expression should be impeded (Additional file 1: Figure S1).
These constructs were used to analyse the SDm1 and SDm2 LTR variants by ribonuclease protection assays (RPAs) (Figure 2B). For the RPA, three antisense RNA probes complementary to nucleotides +1 to +250 – encoding the wild-type, the SDm1, or the SDm2 MSD – were produced. A specific probe for each construct was necessary to avoid cleavage of the RNA probe at the mutated MSD due to non-pairing. All transfections included a Tas expression plasmid (pCMVTas) as expression of the viral transactivator Tas is required to activate the LTR promoter. Transcripts cleaved/polyadenylated at the LTR poly(A) site should result in a protected 193-nucleotide fragment (Figure 2A), whereas suppression of this site should result in a 250-nucleotide fragment. The RPAs showed that suppression of the poly(A) site at the 5′LTR is incomplete and that suppression of polyadenylation acts at the first step of polyadenylation, i.e. RNA cleavage is inhibited. The majority of all transcripts were cleaved at the LTR poly(A) site (Figure 2B). Reporters carrying SDm1 (pGL3SDm1) or SDm2 (pGL3SDm2) showed strong increases in RNAs cleaved at the LTR poly(A) site compared to the wild-type (Figure 2B). This experiment indicates that 1) about 40% of all transcripts are prematurely cleaved in the wild-type context, and 2) the SDm1 and SDm2 mutations result in a further increase in transcripts cleaved at the LTR polyadenylation site, confirming that the MSD indeed suppresses RNA cleavage.
To analyse the impact of the essential G/U-rich DSE in the U5 region on incomplete suppression of polyadenylation we cloned either the U3- or the U3R-promoter regions in the pGL3 backbone. In this set of experiments, a CMV-promoter-driven Renilla luciferase expression plasmid was co-transfected to allow normalization of transfection efficiencies. Two days after transfection, cellular lysates were prepared, and both firefly and Renilla luciferase activities were measured (Figure 2C). The deletion of the U5 region (Figure 2C, second bar (U3R)), which includes the deletion of the DSE required for polyadenylation, resulted in an approximately 2.5-fold increase in the luciferase activity, whereas a plasmid encoding only the U3 region exhibited an approximately 2-fold increase (Figure 2C, third bar (U3)). These findings, along with the increase of the genomic transcript with the SDm2+p(A)m double mutant (Figure 1C), support the view that suppression of the FV polyadenylation at the 5′LTR is incomplete and that the U5 region indeed contains a DSE.
To show that the short transcripts are not only cleaved but also polyadenylated at the 5´LTR, an oligo d(T) primed RT-PCR was performed with RNA of cells transfected with the reporter plasmids (Additional file 1). FV cDNAs were amplified with oligo d(T) and the +1 primer. The PCR products were blotted and hybridised to an antisense RNA probe complementary to nucleotides +250 to +1 to verify the FV origin of the PCR products. This analysis revealed strong amplicons of transcripts polyadenylated at the 5´LTR from cells transfected with the wild-type, the SDm1 or SDm2 mutants (Figure 2D, lanes 1, 3, and 4) showing an almost complete polyadenylation at the LTR for both MSD mutants (Figure 2D, lanes 3 and 4). The inactivation of the poly(A) signal resulted in the loss of RNA species polyadenylated at the 5´LTR (Figure 2D, lanes 2, 5 and 7). This shows that the short transcripts are indeed both cleaved and polyadenylated. Furthermore, it supports our hypothesis that the suppression of polyadenylation in the wild-type LTR is incomplete and is regulated via the MSD, possibly by U1snRNP interaction.
Binding of U1snRNP is required for suppression of polyadenylation
To show that U1snRNP binding to the MSD regulates poly(A) suppression, we performed experiments with a mutated U1snRNA that was complementary to 7 nucleotides of SDm2 (Figure 3A). Expression of this U1snRNA mutant should restore suppression of polyadenylation only if snRNP binding is a determinant for suppression of polyadenylation. Cells were co-transfected with a plasmid encoding the wild-type U1snRNA or the mutant U1snRNA (U1snRNAm2) and with the luciferase reporter constructs (Figure 3A). A CMV-promoter-driven Renilla luciferase expression plasmid was co-transfected to allow normalization of transfection efficiencies. Both firefly and Renilla luciferase activities were measured (Figure 3A). The reporter carrying the pGL3SDm2 mutant showed strongly reduced luciferase activity compared to the wild-type LTR construct, similar to the reduction observed in the RPA (Figure 3A). As described before [34], we observed that over-expression of the wild-type U1snRNA lowered the luciferase expression of the wild-type pGL3LTR significantly (p = 0.006) (Figure 3A, compare bars 1 and 3), indicating that U1snRNA over-expression exerts some side effects. However, the ratio of luciferase activity seen with the pGL3LTR wild-type and the SDm2 reporter with and without co-transfection of the wild-type U1snRNA remained unchanged (Figure 3A, compare reduction from bar 1 to 2 (p = 0.006) and from bar 3 to 5 (p<0.00001)). Co-transfection of the U1snRNAm2 construct strongly increased expression of the SDm2 construct (Figure 3A, compare bars 5 and 6), showing that U1snRNA binding can reverse the impact of the SDm2 mutation. This result supports the hypothesis that U1snRNA binding is required for suppression of transcript cleavage and subsequent polyadenylation.
To analyse whether expression of 5´LTR-derived transcripts could be restored by U1snRNAm2 expression in the context of proviral MSD mutant constructs, BHK-21 cells were co-transfected with the proviral clones pHSRV13 or pHSRV13-SDm2 and the U1snRNA or U1snRNAm2 expression constructs. We co-transfected a Tas-encoding plasmid to compensate for splicing defects, which might effect Tas expression. The foamy viral transcripts were visualized by Northern blotting using a tas-specific probe (Figure 3B). Co-expression of U1snRNA or the mutated U1snRNA did not influence the ratio of 5´LTR-derived transcripts of pHSRV13 (Figure 3B, lanes 3 and 5). In contrast, co-transfection with the U1snRNAm2 construct enhanced the LTR-promoter-derived gag expression of pHSRV13-SDm2, as seen in the luciferase model. To further verify these data, quantities of Gag expression were analysed by Western blotting with a Gag-specific monoclonal antibody (Figure 3C). The pHSRV13-SDm2 mutant did not express a significant amount of Gag. The Gag expression levels of pHRSV13 and its SDm2 mutant were not affected by over-expression of the wild-type U1snRNA, but expression of U1snRNAm2 restored Gag expression of pHSRV13-SDm2 to wild-type levels (Figure 3C).
The experiments with the proviral plasmids gave rise to similar results on the RNA and protein levels and show that U1snRNA is required for the expression of LTR-derived transcripts. Furthermore, the results correlate well with the quantitative data obtained with the luciferase-reporter-based model system. The higher sensitivity of the reporter system allowed us to detect effects of the mutated U1snRNA on the wild-type MSD that could not be visualized by Western or Northern blotting.
Suppression of the poly(A) site is independent of splicing
In order to confirm that suppression of the poly(A) site is independent of splicing but dependent on U1snRNP binding, a pGL3LTR reporter plasmid encoding an inactive splice donor mutant (SDm5) was constructed. This mutant encodes an ideal U1 binding site with the exception of the G/G dinucleotide. This dinucleotide was mutated to G/C, which has been shown to inhibit splicing (Figure 1A) [35]. BHK-21 cells transfected with pGL3SDm5 showed a slight decrease in luciferase activity of 23% compared to the wild-type (p = 0.01) (Figure 4A), likely due to the mismatch in U1snRNA-MSD binding (for luciferase data on SDm4 see S1). Nevertheless, the splicing-incompetent SDm5 suppressed 5´LTR polyadenylation compared to SDm2, showing that splicing is not required for suppression of polyadenylation.
To confirm these results, Northern blotting analysis using a probe encompassing the R region of the pGL3SDm5- and SDm5+p(A)m-derived transcripts was performed. RNAs were extracted using an miRNA isolation procedure (Figure 4B). The mutation SDm2 led to an increase in polyadenylation at the 5´LTR poly(A) site and a reduction of the read-through transcript (Figure 4B), which is in line with the results of the RPA. Consistent with the results of the luciferase assay, the Northern blot analysis revealed that SDm5 suppresses 5´LTR polyadenylation similar to the wild-type (compare lanes 1 and 3), indicating that splicing is not a prerequisite for poly(A) suppression. Nevertheless, transcript cleavage at the 5´LTR was not fully suppressed by SDm5, which contains 10 nucleotides complementary to the U1snRNA. A control transfection with inactivation of the 5´LTR poly(A) signal led to the expected polyadenylation at the vector’s SV40 polyadenylation signal (Figures 4B, lane 4). In addition, we confirmed by RT-PCR that SDm5+p(A)m supports polyadenylation at the SV40 polyadenylation site (Figure 2D, lane 7). In summary, we provide evidence that splicing is not a prerequisite for suppression of polyadenylation at the FV 5’LTR.
Regulation of polyadenylation is promoter-independent
Transcription, splicing, and poly(A) addition are coupled processes [1]. Since the HIV-1 U3 promoter and the CMV i.E. promoter recruit specific RNA-polymerase complexes II (Pol II) which display differences in both processivity and splicing [36], an analysis of the regulation of the FV polyadenylation concerning the promoter-dependency was desirable. The U3 promoter was excised from the pGL3LTR, -SDm2, and the respective poly(A) signal mutant constructs and replaced with the CMV-promoter fragment of pcHSRV2 [37] (Figure 5A). In these plasmids, the transcriptional start site of the constitutive CMV promoter is identical to the PFV transcriptional start site. Cellular luciferase activities after transfection with the U3 plasmid were more than 2-fold higher compared to cells transfected with the CMV plasmids, showing either a higher processivity of the recruited Pol II-complexes or a higher initiation rate at the FV LTR promoter. But the regulation of the polyadenylation was unaffected. The reduction of luciferase activity of the SDm2 transfected cells was in the same range as those transfected with the LTR promoter, and the additional poly(A) signal mutants displayed comparable increases in luciferase activities. This increase might be due to an inactive polyadenylation signal and to suppressed splicing by the SDm2 mutant. These results imply that the suppression of the 5’LTR polyadenylation of the 5’LTR is independent of the promoter.
Regulation of polyadenylation at the 3′LTR
In HIV-1, the MSD is located downstream of the 5′LTR. Therefore, polyadenylation at the 3′LTR, which lacks a downstream MSD, is not inhibited. In contrast, FVs have to prevent suppression of polyadenylation by the MSD at the 3′LTR because the R regions of both FV LTRs harbour an MSD. In order to determine the requirements for polyadenylation at the 3′end, we analysed whether the splice donor is essential for the regulation of polyadenylation. Either the wild-type LTR or the SDm2 mutants were inserted between Renilla and firefly luciferase genes in the pRL vector (Figure 6A). In addition, to find out whether a stronger MSD would suppress polyadenylation, we created a LTR mutant with 11 nucleotides of the MSD complementary to the U1snRNA (SDm4) by site-directed mutagenesis (Figure 1A) and inserted it into the 3′LTR reporter construct. The resulting constructs encode two poly(A) sites: 1) the FV LTR polyadenylation site (transcript size 2174 nts) and 2) the vector-derived SV40 late poly(A) site (transcript size 4125 nts) (Figure 6A). BHK-21 cells were transfected with the reporter constructs, and RNAs were analysed by Northern blotting using a probe encompassing Renilla luciferase (Figure 6B). These experiments were performed in the absence of Tas; however, further experiments showed that addition of Tas did not change the polyadenylation pattern nor did Tas activate the U3 promoter in these constructs, possibly indicating that Tas is unable to bind to 3′LTR sequences. The Northern blots showed that RNA was polyadenylated at the LTR and that the polyadenylation was independent of a functional splice donor (Figure 6B, lanes 2 and 3). Neither the weak splice donor SDm2 nor the strong SDm4 had any influence on polyadenylation site selection. These results were in striking contrast to all experiments with the LTR at the 5′position. To identify signals that support polyadenylation and render the splice donor non-relevant, we analysed the effects of U3 region deletions. The U3 region of pHSRV13 encompasses 777 nucleotides. Five additional reporter plasmids encompassing the RU5 region alone or RU5 and the U3 regions from −350, -200, -100, or −13 to +1 were constructed (Figure 6A). RNAs from BHK-21 cells transfected with these plasmids were analysed by Northern blotting (Figure 6B). All constructs showed a preferential polyadenylation at the LTR poly(A) site, indicating that the region from −13 to +1 and other upstream sequences relieve suppression of LTR polyadenylation. In addition, transcripts of the construct encoding only RU5 were polyadenylated at the LTR (Figure 6B, lane 7), supporting the hypothesis that U3 sequences or even upstream exons activate polyadenylation.
The RNA-region at the 3’LTR is preceded by the whole genomic pre-mRNA including pre-selected splice sites etc., whereas the RNA at the 5’poly(A) signal only encompasses the R-region. To further investigate whether 3′polyadenylation is influenced by upstream sequences, a simplified reporter was constructed by inserting nucleotides −13 to +1 into the pGL3-CMV-RU5 clones described above (Figure 5). In this construct, nucleotide −13 is positioned directly at the start site of the CMV promoter. Thus, the transcript is free of upstream coding regions, but encodes minimal sequences of U3. Transfection experiments showed that the additional 13 nucleotides of U3 caused a significant reduction in luciferase activity to 34% of the wild-type (p=0.005) (Figure 6C), which is comparable to the reduction seen with the SDm2 mutant (Figure 6C). This indicated that polyadenylation at the 3’LTR might have been activated by the 13 nucleotides of U3. In summary, these experiments show that U3 upstream sequences are able to activate polyadenylation at the 3’LTR.
Differences in the RNA structure at the 5′ and 3′LTRs presumably regulate splice donor recognition
To determine the differences in both polyadenylation and splice donor dependence at the 5′ and 3′LTRs, we analysed the RNA secondary structure of two RNA fragments, one representing the 3′LTR RNA (nucleotides −13 to +198) and one representing the 5′LTR RNA (nucleotides +1 to +198) by RNA SHAPE (Figure 7 and Additional file 1: Figure S2). The 3′ends of the two RNAs including the poly(A) signal and poly(A) site show identical secondary structural folds. However, we observed major differences at the 5′end of the two RNAs. Compared with the 5′LTR sequence (Figure 7A), the first stem loop of the 3′LTR is extended and the second stem loop is shortened (Figure 7B). The MSD is located between stem loops one and two in the 3′LTR, and only two nucleotides complementary to the U1snRNA are unpaired. The MSD of the 5′LTR is part of its extended second stem loop and forms a bulge. This leaves four U1snRNA-binding nucleotides unpaired. This structure is strikingly similar to the U1A-stem structure conserved in all mammals [38]. Additionally, we predicted the RNA secondary structures of the SDm1 and SDm2 mutants.
The single 5’LTR mutation of SDm1 completely disrupts the local RNA fold and modifies the RNA structure to a 3’LTR-like fold (Additional file 1: Figure S3). The 5’LTR SDm2 mutation repositions the bulge of the MSD by one nucleotide further upstream. This leads to substantial changes in the stem loop containing the MSD: a) the stem of the mutated RNA upstream of the bulge consists of six instead of five base paired nucleotides; b) the mutated base is no longer complementary to the U1snRNA; c) only two instead of three unpaired bases present in the bulge are complementary to the U1snRNA. In addition, structure prediction of a Renilla luciferase-R-U5 construct (Additional file 1: Figure S3C) representing the one used in Figure 6B (lane 7) shows disruption of 5’LTR MSD.
Taken together, our data provide evidence that adoption of deviating RNA structures in the 5’LTR MSD leads to premature polyadenylation.