Heterogeneous Nuclear Ribonucleoprotein A1 (hnRNP A1) and hnRNP A2 Inhibit Splicing to Human Papillomavirus 16 Splice Site SA409 through a UAG-Containing Sequence in the E7 Coding Region

Human papillomavirus type 16 (HPV16) belongs to the high-risk-group of HPVs and is causing a variety of anogenital cancers and head and neck cancer. The two HPV16 oncoproteins E6 and E7 prevent apoptosis and promote mitosis and are essential for completion of the HPV16 life cycle and for transformation of the infected cell and maintenance of malignancy. E6 and E7 are produced from two mRNAs that are generated in a mutually exclusive manner by alternative splicing. While E6 protein is made from the unspliced mRNA, E7 is made from the spliced version of the same pre-mRNA. Since sufficient quantities of both E6 and E7 are required for malignant transformation, this intricate arrangement of gene expression renders E6 and E7 expression vulnerable to external interference. Since antiviral drugs to HPV16 are not available, a detailed knowledge of the regulation of HPV16 E6 and E7 mRNA splicing may uncover novel targets for therapy.

this intricate arrangement of gene expression renders E6 and E7 expression vulnerable to external interference. Since antiviral drugs to HPV16 are not available, a detailed knowledge of the regulation of HPV16 E6 and E7 mRNA splicing may uncover novel targets for therapy. KEYWORDS E6, E7, hnRNP A1, papillomavirus, splicing H uman papillomaviruses (HPV) are small DNA viruses with a strict tropism for human epithelial cells. A subset of the HPVs is mainly sexually transmitted and has tropism for mucosal cells. The vast majority of these HPV infections are asymptomatic, and the infections are cleared by the immune system of the host. However, the high-risk HPVs can establish chronic persistent infections that last for many years or decades (1, 2). These infections may give rise to cervical lesions that progress to cervical cancer (3). More than 99% of all cervical cancers contain HPV DNA, and epidemiological studies have established that the most common high-risk HPV is HPV16 and that HPV16 is present in approximately 50% of all cervical cancers (4). HPV16 encodes at least eight distinct protein-expressing genes. Among those, the E6 and E7 genes are of particular interest. They not only carry out essential functions in the HPV16 replication cycle but also contribute to induction of cervical lesions as well as induction and maintenance of cervical cancer (5)(6)(7). Continued expression of HPV16 E6 and E7 is required for survival and proliferation of HPV16-driven cancer cells (8). This requirement is mainly attributed to the ability of HPV16 E7 to drive cell proliferation through its association with pRB and the liberation of E2F (9), while E6 targets p53 for degradation, thereby preventing apoptosis (10). However, the roles of E6 and E7 in carcinogenesis may be more complex than that, since both E6 and E7 interact with additional proteins.
Transcriptional regulation is important in the control of HPV16 gene expression (11)(12)(13). HPV16 makes extensive use of alternative mRNA splicing to produce mRNAs encoding each of the virally encoded proteins (14)(15)(16)(17)(18). The various splice sites on the HPV16 mRNAs are controlled by cis-acting RNA elements that interact with transacting cellular factors produced by the HPV16-infected cells. Splicing regulatory RNA elements in papillomaviruses were first discovered in bovine papillomaviruses (19)(20)(21). As a matter of fact, the HPV16 infection itself alters expression levels of cellular RNA binding proteins that control mRNA splicing, presumably to optimize intracellular conditions for the HPV16 gene expression program (22,23). Furthermore, HPV16 infection activates the DNA damage response machinery (DDR) (6), and DDR factors recruit RNA binding proteins that modulate HPV16 mRNA splicing and polyadenylation (24,25). The HPV16 E6 and E7 proteins are produced from two mRNAs that are generated in a mutually exclusive manner: the unspliced mRNAs encode the full-length E6 protein, and the spliced mRNAs encode E7 (26,27). The splice sites utilized to produce the E7 mRNAs are located within the E6 open reading frame (orf) and are named SD226 (5=-splice site) and SA409 and SA526 (3=-splice site) ( Fig. 1A to C) (14,16,17). The major and most abundant HPV16 mRNA produced in HPV16-infected cells as well as in HPV16-driven cancers is spliced from SD226 to SA409 (28)(29)(30)(31). This splicing event shortens the E6 orf to generate the E6*I-orf located upstream of the E7 orf on this mRNA (32). As a result of the shortened E6 orf and its poor ATG, ribosomes may reinitiate translation at the E7 ATG (26,33,34) or reach the downstream E7 orf by leaky scanning (35). Alternative ideas have been presented using less stringent experimental systems (36,37). In contrast, translation of the E7 orf does not seem to occur on mRNAs that contain an intact, full-length E6 orf upstream of E7 (26,36). The SD226-SA409 mRNA is specific for high-risk, cancer-associated HPV types and appears to be the major mRNA encoding the promitotic E7 protein. However, splicing to SA409 is not allowed to be too efficient, as that would prevent production of full-length E6 mRNAs. Since the utilization efficiency of SA409 determines the relative levels of the HPV16 E6 and E7 mRNAs, the control of splice site SA409 is of particular interest. We speculate that the intracellular concentrations of RNA binding proteins that control HPV16 E6 and E7 mRNA splicing is of paramount importance for a balanced expression of E6 and E7 mRNAs. Therefore,   (39), and hnRNP A1 and hnRNP A2 control HPV16 late gene expression (40)(41)(42)(43)(44). hnRNP A1 is upregulated in HPV16-infected cells, in particular in high-grade cervical lesions as well as cervical cancer (22). Furthermore, expression levels of hnRNP A1 are affected by HPV16 infection. Here, we show that hnRNP A1 and A2 interact with a C-less RNA element in the E7 coding region, thereby inhibiting HPV16 E7 mRNA splicing and HPV16 E7 production. Collectively, these results demonstrate that hnRNP A1 and A2 are major regulators of HPV16 mRNA splicing and HPV16 gene expression.

RESULTS
Inhibition of HPV16 3=-splice site SA409 by hnRNP A1 results in production of unspliced E6 mRNAs, whereas inhibition of SA409 by hnRNP A2 results in splicing to alternative 3=-splice site SA742. The hnRNP A1 and A2 proteins were shown previously to be involved in the regulation of HPV16 gene expression at the level of RNA processing (38)(39)(40)(41)(42)(43)(44). To further investigate the effects of hnRNP A1 and hnRNP A2 on HPV16 mRNA splicing, we wished to determine the effect of hnRNP A1 on the splicing of HPV16 mRNAs produced by the HPV16 subgenomic reporter plasmid pC97ELsL ( Fig. 1A and B). This plasmid was transfected into HeLa cells, and RNA was extracted and subjected to reverse transcription-PCR (RT-PCR). The various alternatively spliced HPV16 mRNAs produced by pC97ELsL were monitored by RT-PCR using various primer pairs (Fig. 1C to E). The RT-PCR gels in Fig. 1C to E show alternatively spliced HPV16 mRNAs produced in HeLa cells transfected with pC97ELsL. In addition to identifying the mRNAs produced by pC97ELsL, the results also revealed that mRNAs utilizing HPV16 SD226 are efficiently spliced to SA3358 (Fig. 1D), whereas mRNAs utilizing SD880 are more efficiently spliced to SA2709 (Fig. 1E). Absence of detectable plasmid DNA contamination of the RNA samples was confirmed by RT-PCR experiments performed in the absence of reverse transcriptase for all HPV16 primer pairs (RT-minus) ( Fig. 1C to E). All RT-PCR products were sequenced to determine their identity. These results identify the HPV16 mRNAs detected by the various RT-PCR primers.
To monitor the effect of hnRNP A1 or hnRNP A2 on the various alternatively spliced HPV16 mRNAs, serially diluted expression plasmids for hnRNP A1 or hnRNP A2 were cotransfected with HPV16 reporter plasmid pC97ELsL into HeLa cells. First, the mRNAs spliced from HPV16 5=-splice site SD880 to either SA2709 or SA3358 were monitored by RT-PCR. As can be seen from Fig. 2A, B, D, and E, mRNAs spliced from SD880 to either SA2709 or SA3358 were largely unaffected by hnRNP A1 or A2 overexpression. In contrast, mRNAs spliced from SD226 and analyzed with primers 97S and E2AS were strongly reduced by hnRNP A1 but only to a minor extent by hnRNP A2 (Fig. 2C). Absence of detectable plasmid DNA contamination of the RNA samples was confirmed by RT-PCR experiments performed in the absence of reverse transcriptase for all HPV16 primer pairs (RT-minus) ( Fig. 2A to D). Since many of the mRNAs detected by primers  Our results indicated that hnRNP A1 and A2 affected splice sites in the E6 and E7 coding regions, which is in line with previously published results on HPV16 (38) and HPV18 (39). Next, we analyzed the effect of hnRNP A1 and A2 on the HPV16 splice sites SD226, SA409, SA526, and/or SA742 in the E6 and E7 coding region using primer pair 97S and 880A that specifically covers this area (Fig. 1C depicts the structures of the E6 and E7 mRNAs and the location of these primer pairs). Overexpression of hnRNP A1 inhibited splicing to splice sites SA409 and SA526 and resulted in a significant increase in unspliced mRNAs encoding the full-length E6 open reading frame ( Fig. 3A and B). Levels of mRNAs spliced from SD226 to SA742 (E6^E7 mRNAs) also increased at the expense of mRNAs spliced to SA409 and SA526 ( Fig. 3A and B). Thus, high levels of hnRNP A1 favored unspliced E6 mRNAs and mRNAs spliced SD226 to SA742 over mRNAs spliced to SA409 and SA526. Overexpression of hnRNP A2 also inhibited production of mRNAs spliced to SA409 and SA526, but, in contrast to hnRNP A1 overexpression, hnRNP A2 overexpression did not promote production of unspliced E6 mRNAs but strongly enhanced alternative splicing from SD226 to SA742 ( Fig. 3A and B). This occurred at the expense of the unspliced E6 mRNAs as well as mRNAs spliced to SA409 and SA526 encoding E6*I and E6*II ( Fig. 3A and B). Similar results were obtained after transfection of 293T cells (Fig. 2F). Unspliced mRNAs are known to produce primarily E6 protein, while mRNAs spliced to SA409 produce primarily E7 protein, although they are also predicted to produce the E6*I protein, and mRNAs spliced to SA742 produce the E6^E7 protein. As predicted, overexpression of either hnRNP A1 or hnRNP A2 reduced E7 protein production (Fig. 3C). There was a 4-to 5-fold reduction in E7 protein levels (Fig. 3D). The E7 antibody specifically detected the HPV16 E7 protein in pC97ELsL-transfected cells and not in untransfected cells (Fig. 3E). As an additional control for the E7 antibody specificity, we transfected HeLa cells with plasmid pE6E7F, from which Flag-tagged E7 is produced and stained with E7 antibody. High levels of E7 protein were observed in transfected cells, whereas E7 protein was undetectable in untransfected cells (data not shown), further supporting the specificity of the antibody. Although hnRNP A1 is predicted to enhance E6 production, we were unable to detect the E6 protein (data not shown). We concluded that hnRNP A1 and A2 both inhibited splicing to SA409 and SA526 and thereby inhibited production of E7 protein, but the alternative splicing outcome of this inhibition differed between the two proteins hnRNP A1 and hnRNP A2.
Differential effect of hnRNP A1 and hnRNP A2 on alternative splicing of the HPV16 L1 mRNAs. We have previously shown that hnRNP A1 and A2 bind to splicing silencers on the HPV16 L1 mRNAs (40)(41)(42)(43). Therefore, we investigated if overexpression of hnRNP A1 or A2 affected alternative splicing of HPV16 L1 mRNAs produced from pC97ELsL. Although the majority of HPV16 L1-and L2-encoding mRNAs are produced from the HPV16 late promoter, HPV16 L1-and L2-encoding mRNAs initiated at the early promoter also have been identified in HPV16-infected cells (45). Analyses of the low levels of L1 mRNAs produced by pC97ELsL were performed with primers 97S and L1A or 773S and L1A (data not shown). The results revealed that overexpression of hnRNP A1 inhibited inclusion of the central exon located between SA3358 and SD3632 on the HPV16 L1 mRNAs and favored production of L1 mRNAs spliced from either SD226 or SD880 directly to SA5639 (data not shown). In contrast, overexpression of hnRNP A2 favored inclusion of the central exon between SA3358 and SD3632 on all L1 mRNAs detected with primers 97S and L1A or 773S and L1A (data not shown). Therefore, these results support the idea that hnRNP A1 and A2 differ in their effects on HPV16 mRNA splicing.
hnRNP A1 acts on splicing inhibitory RNA sequences upstream of HPV16 nucleotide position 604 in the HPV16 E7 coding region. Having established that high levels of hnRNP A1 and A2 inhibited splicing between HPV16 splice sites SD226 h n R N P A 1 h n R N P A 2 p U C U n tr a n s fe c e te d and SA409, we wished to the map the target site for hnRNP A1 and hnRNP A2 on the HPV16 E6 and E7 mRNAs. To this end, we generated a number of subgenomic HPV16 expression plasmids from pC97ELsL ( Fig. 4A and B). Plasmids pX1200, pX1060, pX960, pX856F, and pX556F were transfected into HeLa cells in the presence of empty pUC plasmid (Ϫ) or plasmids encoding hnRNP A1 or hnRNP A2, RNA was extracted, and RT-PCR was performed with primers 97S and 438A (Fig. 4C). The numerical digits in the plasmid names represent the endpoint of each deletion and refer to nucleotide position in the HPV16R sequence. Analysis of RT-PCR products from the deletion mutants cotransfected with empty pUC plasmid revealed that mRNAs produced from pX556F were efficiently spliced between SD226 and SA409 (Fig. 4C), whereas HPV16 mRNAs produced from the longer plasmids pX1200, pX1060, pX960, and pX856F were less efficiently spliced (Fig. 4C). Quantitation of the bands and plotting percent unspliced versus the endpoint in the HPV16 genome of the various plasmids revealed that only around 10% of the mRNAs produced by the smallest plasmid, pX556F, were unspliced when cotransfected with empty pUC plasmid (Fig. 4D). Percent unspliced mRNA increased when downstream sequences were included, and it was around 30% with plasmid pX856F and peaked at around 55% for plasmids pX960, pX1060, and pX1200 ( Fig. 4C and D), indicating that multiple splicing inhibitory RNA sequences were present downstream of HPV16 nucleotide position 556. RT-PCR bands were undetectable in control PCR experiments performed on the same RNA samples in the absence of reverse transcriptase (data not shown). These splicing inhibitory RNA sequences may be the target sequences for hnRNP A1 and A2.
To determine how hnRNP A1 and hnRNP A2 affected splicing of mRNAs produced from the various HPV16 deletion mutants pX1200, pX1060, pX960, pX856F, and pX556F ( Fig. 4B), RNA was extracted from HeLa cells transfected with these plasmids in the presence of hnRNP A1 or hnRNP A2 and analyzed by RT-PCR with primers 97S and 438A (Fig. 4B). The results revealed that overexpression of hnRNP A1 and A2 did not inhibit splicing from plasmid pX556F ( Fig. 4C and D), demonstrating that the target sequences for hnRNP A1 and A2 were absent from pX556F. In contrast, overexpression of hnRNP A1 and A2 efficiently inhibited splicing of mRNAs produced from plasmids pX1200, pX1060, pX960, and pX856F ( Fig. 4C and D). These results suggested that overexpression of hnRNP A1 and A2 (hnRNP A1 in particular) acted through splicing inhibitory sequences located downstream of HPV16 nucleotide position 556.
To provide further support for the results that indicated that both hnRNP A1 and A2 inhibited HPV16 mRNA splicing from HPV16 plasmid pX856F (the smallest HPV16 plasmid that responded to hnRNP A1 and A2), serial dilutions of hnRNP A1 or A2 plasmid were cotransfected with pX856F, RNA was extracted, and RT-PCR was performed. The results revealed that hnRNP A1 and A2 both inhibited splicing between SD226 and SA409 in a dose-dependent manner ( Fig. 5A and B). RT-PCR bands were undetectable in control PCR experiments performed on the same RNA samples in the absence of reverse transcriptase (data not shown). As expected, hnRNP A1 promoted production of the unspliced HPV16 E6 mRNA ( Fig. 5A and D), whereas hnRNP A2 promoted production of HPV16 mRNAs spliced from SD226 to SA742 ( Fig. 5A and C). In conclusion, both hnRNP A1 and A2 acted on HPV16 sequences located upstream of HPV16 nucleotide position 856 to inhibit splicing between SD226 and SA409. However, the outcome of this splicing inhibition differed between hnRNP A1 and A2. displayed as fold difference of HPV16 RT-PCR band intensity obtained with RNA extracted from cells transfected with pC97ELsL in the presence of phnRNP A1 or A2 over band intensities in cells transfected with pC97ELsL in the presence of empty pUC plasmid (Ϫ). Band intensity ratios are plotted against micrograms of transfected phnRNP A1 or phnRNP A2 plasmid DNA. (C) Western blotting on cell extracts from HeLa cells transfected with pC97ELsL (0.5 g) and 2-fold serially diluted phnRNP A1 or phnRNP A2 plasmid (1.5 g, 0.75 g, 0.38 g, and 0.19 g). Cells were harvested 20 h posttransfection. Blots were stained with antibody specific for HPV16 E7 and actin as described in Materials and Methods. (D) Western blotting on cell extracts from HeLa cells transfected with pC97ELsL (0.5 g) and 1.5 g of phnRNP A1 or phnRNP A2 plasmid. The E7 and actin bands were quantified, and fold reduction of E7 protein by hnRNP A1 or hnRNP A2 overexpression is indicated below gel pictures. (E) Western blotting on cell extracts from HeLa cells transfected with pC97ELsL (0.5 g) and 1.5 g of phnRNP A1 or phnRNP A2 plasmid. Cell extract from untransfected cells is included. MW, molecular weight marker.
To more precisely map the target sequence of hnRNP A1 and A2, additional deletions were introduced in the 3= end of plasmid pX856F, generating plasmids pX644, pX616, pX604, pX581, and pX579. The numerical digits in the plasmid names represent the endpoint of each deletion and refer to nucleotide position in the HPV16R sequence. As unspliced p X 1 0 6 0 p X 9 6 0 p X 1 2 0 0 p X 8 5 6 F p X 5 5 6 F p X 1 0 6 0 p X 9 6 0 p X 1 2 0 0 p X 8 5 6 F p X 5 5 6 F p X 1 0 6 0 p X 9 6 0 p X 1 2 0 0 p X 8 5 6 F p X 5 5 (62). Transcription of the HPV16 sequences in the pC97ELsL plasmid is driven by the human cytomegalovirus immediate-early promoter (CMV). HPV16 splice sites are indicated. Numbers refer to the HPV16 reference strain HPV16R (45). Early and late polyadenylation signals pAE and pAL are indicated. IRES, internal ribosome entry site; sluc, secreted luciferase. (B) Schematic representation of HPV16 subgenomic reporter plasmids pX1200, pX1060, pX960, pX856F, and pX556F. Transcription of the HPV16 sequences is driven by the human cytomegalovirus immediate-early promoter (CMV). RT-PCR primers 97S and X525A are indicated. HPV16 sequences in the deletion mutants pX1200, pX1060, pX960, pX856F, and pX556F end at HPV16 nucleotide positions 1200, 1060, 960, 856, and 556. HPV16 splice sites are indicated. Nucleotide positions refer to the nucleotide position of the HPV16 reference genome HPV16R. (C) RT-PCR on RNA extracted from HeLa cells transfected with plasmids pX1200, pX1060, pX960, pX856F, or pX556F in the absence (Ϫ) or presence of phnRNP A1 or phnRNP A2. The bands representing unspliced HPV16 E6 mRNAs and HPV16 mRNAs spliced from SD226 to SA409 are indicated to the right. cDNAs were amplified by RT-PCR with primers 97S and X525A. RT-minus, RT-PCR performed in the absence of reverse transcriptase. (D) Quantitations were performed on triplicates. Percent unspliced HPV16 mRNA is calculated for each HPV16 plasmid transfected with either pUC plasmid or hnRNP A1 or hnRNP A2 expression plasmid and plotted against the transfected HPV16 expression plasmid. hnRNP A1 and A2 Control HPV16 E6 and E7 mRNA Splicing Journal of Virology seen from the RT-PCR results obtained in cotransfections with empty pUC plasmid, splicing of mRNAs produced from pX556F, pX579, and pX581 was highly efficient, whereas splicing from plasmids pX604, pX614, and pX644 was less efficient ( Fig. 6A and C). These results indicated that splicing inhibitory sequences extended to HPV16 nucleotide position 604 ( Fig. 6A and C). RT-PCR bands were undetectable in control PCR experiments performed on the same RNA samples in the absence of reverse transcrip-  tase (data not shown). Furthermore, similar results were obtained after transfection of the deletion mutants pX556F, pX579, pX604, pX614, and pX644 into HPV-negative cervical cancer cell line C33A (data not shown). Next, the same plasmids were transfected in the presence of phnRNP A1 plasmid. hnRNP A2 overexpression was not analyzed here, since HPV16 3=-splice site SA742 was absent from the reporter plasmids.
As expected, overexpression of hnRNP A1 strongly inhibited splicing from the plasmids containing splicing inhibitory sequences (pX604, pX614, and pX644) and shifted splicing from a majority of spliced mRNAs (226^409) to a majority of unspliced E6 mRNAs ( Fig. 6A and B). In contrast, hnRNP A1 had no or very little effect on splicing of mRNAs produced from plasmids pX556F, pX579, and pX581 that did not contain splicing inhibitory sequences (Fig. 6A and B). These results suggested that hnRNPA1 acted on splicing inhibitory sequences located immediately upstream of HPV16 nucleotide position 604. The hnRNP A1 and A2 proteins target a UAG-encoding HPV16 sequence. The sequence between HPV16 nucleotide positions 581 and 604 that inhibited splicing and that mediated the splicing-inhibitory effect of hnRNP A1 and hnRNP A2 encodes a UAG motif (Fig. 6C). Since hnRNP A1 has been shown to bind UAG-containing sequences, we introduced nucleotide substitutions in this sequence, resulting in plasmid pM604 (Fig.  6C). As can be seen, these point mutations inactivated the splicing silencer and enhanced splicing (Fig. 6D). Furthermore, the mutant plasmid pM604 responded less well to hnRNP A1 and hnRNP A2 overexpression than the wild-type plasmid pX604 (Fig.  6D), demonstrating that the UAG motif was required for splicing silencing and suggesting that hnRNP A1 and hnRNP A2 recognized this sequence.
The hnRNP A1 and A2 proteins interact specifically with splicing inhibitory RNA sequences in the HPV16 E7 coding region. To determine if hnRNP A1 and A2 interact with the splicing inhibitory sequences immediately upstream of HPV16 nucleotide position 604, we performed RNA-mediated protein pulldowns using nuclear extract from HeLa cells and a set of nested, biotinylated HPV16 RNA oligonucleotides attached to Streptavidin-coated beads (Fig. 7A). The pulled-down factors were subjected to Western blotting with antibodies to hnRNP A1 or hnRNP A2. Only RNA oligonucleotides extending to HPV16 nucleotide position 604 pulled down hnRNAP A1, while oligonucleotides extending no further than to position 573 or 588 did not (Fig. 7C). Interestingly, hnRNP A2 interacted with the same RNA sequences (Fig. 7D). To more precisely map the target sequence for hnRNP A1 and A2, two shorter RNA oligonucleotides, named 604AS1 and 604BS1 (Fig. 7B), were used in the pulldown assay. Since hnRNP A1 and A2 were pulled down by 604AS1, the results presented in Fig. 7C and D collectively mapped the target site for both hnRNPA1 and A2 to sequences between HPV16 nucleotide positions 594 and 604. We concluded that hnRNP A1 and A2 interact with an 11-nucleotide, C-less sequence downstream of SA409 to inhibit utilization of this splice site. To determine if hnRNP A1 interacted directly with this sequence, glutathione S-transferase (GST) or GST-hnRNP A1 was used instead of nuclear extract in the pulldown assay ( Fig. 7E and F). As can be seen, GST-hnRNP A1, but not GST, interacted with RNA oligonucleotide nucleotide 604AS1 (Fig. 7G). This interaction was GST-hnRNP A1 dose dependent (Fig. 7H). If GST-hnRNP A1 was preincubated with nonbiotinylated 604 RNA oligonucleotide, pulldown of GST-hnRNP A1 with the  hnRNP A1 and A2 Control HPV16 E6 and E7 mRNA Splicing Journal of Virology biotinylated 604AS1oligo was reduced (Fig. 7I), while preincubation with RNA oligonucleotide 588, in which the primary binding site of hnRNP A1 was absent, did not affect pulldown of GST-hnRNP A1 with biotinylated 604AS1oligo (Fig. 7I). Thus, hnRNP A1 interacted specifically and directly with the splicing inhibitory RNA sequence immediately upstream of HPV16 nucleotide position 604. The hnRNP A1 and A2 proteins interact with an UAG-encoding sequence. Since we had shown in Fig. 6 that point mutations that changed the UAGAU sequence in HPV16 to GCCAC inactivated the splicing silencer in transient-transfection experiments, we introduced the same mutations in biotinylated RNA oligonucleotide 604AS1 (Fig.  8A), resulting in RNA oligonucleotide 604AM1 (Fig. 8A) (available on request). Two additional mutant oligonucleotides with substitutions immediately upstream of the UAGAU sequence, named 604AM2 and 604AM3 (Fig. 8A), were also utilized. As can be seen from the pulldown of proteins from HeLa nuclear extracts followed by Western blotting with antibodies to hnRNP A1 and hnRNP A2, mutations in the UAGAU sequence abolished hnRNP A1 and hnRNP A2 binding, whereas the upstream mutations did not ( Fig. 8B and C). Similar results were obtained with both hnRNP A1 and hnRNP A2 ( Fig. 8B and C), with the exception that mutations in 604AM2 immediately upstream of the UAGAU sequence reduced pulldown of hnRNP A1 but not hnRNP A2, suggesting that the hnRNP A1 binding site extends to the 5= end of the UAGAU sequence. We concluded that nucleotide substitutions that abolished the splicinginhibitory effect of the HPV16 splicing silencer in transfected cells also abolished pulldown in vitro of hnRNP A1 and hnRNP A2 by HPV16 RNA oligonucleotides.
In addition, we mutated the UAG-containing sequence in the context of the subgenomic HPV16 reporter plasmid pC97ELsL, resulting in pC97ELsLM1 (Fig. 8D). Analysis of RNA from HeLa cells transfected with pC97ELsL or pC97ELsLM1 revealed that pC97ELsLM1 produced more mRNA spliced from SD226 to SA409 than pC97ELsL ( Fig. 8E and F). RT-PCR bands were not detected in the absence of reverse transcriptase (Fig. 8G). These results indicated that the UAG-containing splicing silencer is a major determinant of splicing to HPV16 SA409.
Finally, to confirm that hnRNP A1 and A2 interact with HPV16 early mRNAs in living cells, we immunoprecipitated RNA protein complexes from HeLa cells transfected with pC97ELsL in the presence of hnRNP A1 or A2 using monoclonal antibodies to either hnRNP A1 or hnRNP A2. RNA was extracted from the immunoprecipitations and subjected to RT-PCR with HPV16-specific PCR primers. As can be seen, monoclonal antibodies to hnRNP A1 or hnRNP A2 immunoprecipitated HPV16 mRNAs, whereas IgG did not (Fig. 8H). Thus, HPV16 early mRNAs interact with hnRNP A1 and hnRNP A2 in human cells.
The C terminus of hnRNP A1 is required for efficient inhibition of HPV16 3=-splice site SA409 and for production of HPV16 unspliced E6 mRNAs. Both hnRNP A1 and A2 inhibit splicing between SD226 and SA409 and interact with the same splicing-inhibitory RNA element located between HPV16 nucleotide positions 594 and 604. Despite these similarities, overexpression of hnRNP A1 induced unspliced E6encoding mRNAs, whereas overexpression of hnRNP A2 promoted E6^E7-encoding mRNAs that were spliced between SD226 and SA742. hnRNP A1 and A2 are two closely related hnRNP proteins and display structural similarity with two conserved N-terminal (G) Western blotting of GST-hnRNP A1 protein pulled down by Streptavidin-coated magnetic beads without RNA oligonucleotide or carrying biotinylated RNA oligonucleotide 604AS1. Filter was stained with antibody to hnRNP A1. (H) Serially diluted GST-hnRNP A1 was incubated with Streptavidin-coated magnetic beads bound to biotinylated RNA oligonucleotide 604S1 followed by magnetic pulldown of beads and Western blotting of hnRNP A1. (I) GST-hnRNP A1 was preincubated with nonbiotinylated RNA oligonucleotide 604 or 588 followed by incubation with Streptavidin-coated magnetic beads bound to biotinylated RNA oligonucleotide 604AS1. Pulled-down proteins were subjected to Western blotting with anti-hnRNP A1 antibody. MW, molecular weight markers; beads, pulldowns using beads only; NE, nuclear extract. Ig G a n ti -h n R N P A 1 in p u t Ig G a n ti -h n R N P A 2 in p u t RNA recognition motifs (RRMs), followed by a less conserved C-terminal glycine-rich domain that also harbors an RGG box and nuclear import and export signals. The N-terminal halves of the two proteins are relatively conserved, whereas the C-terminal part is more variable. To delineate various roles of hnRNP A1 and A2 protein domains in the control of HPV16 mRNA splicing, we generated hybrids between hnRNP A1 and A2 in which the two RRMs were derived from hnRNP A1 or A2 and the C terminus from either hnRNP A1 or A2, generating plasmids pA1NA2C and pA2NA1C (Fig. 9A). Western blotting confirmed that the HA-tagged hybrid proteins were expressed in HeLa cells (Fig. 9B). Next, the hybrids between hnRNP A1 and A2 were cotransfected with HPV16 reporter plasmid pC97ELsL, and RNA was extracted and analyzed by RT-PCR. Interestingly, the A1A2 hybrid, in which the two RRMs were derived from hnRNP A1 and the C terminus from hnRNP A2, inhibited splicing from SD226 to SA409 and promoted production of unspliced E6 mRNA, albeit with lower efficiency than wild-type hnRNP A1 (Fig. 9C). RT-PCR bands were undetectable in control PCR experiments performed on the same RNA samples in the absence of reverse transcriptase (data not shown). Thus, replacing the hnRNP A1 C terminus with the hnRNP A2 C terminus reduced efficiency of splicing inhibition, demonstrating an important role for the hnRNP A1 C terminus in splicing inhibition and production of HPV16 unspliced E6 mRNAs. On the other hand, the A2A1 hybrid, which contained the two RRMs from hnRNP A2 and the C terminus from hnRNP A1, inhibited splicing from SD226 to SA409 and promoted production of unspliced E6 mRNA in a manner that was similar to that of the wild-type hnRNP A1 (Fig.  9C), further corroborating the importance of the C terminus of hnRNP A1 in the generation of HPV16 unspliced E6 mRNAs. However, the A2A1 hybrid also induced splicing to SA742, although not as efficiently as wild-type hnRNP A2 (Fig. 9C). Thus, efficient splicing to SA742 required both N terminus and C terminus from hnRNP A2.
Since hnRNP A1 and A2 may act on multiple sites on HPV16 mRNAs, we next cotransfected the A1A2 and A2A1 hybrids with HPV16 reporter plasmid pX604 that contains only one target site for both hnRNP A1 and hnRNP A2. As expected, hnRNP A1 inhibited splicing between SD226 and SA409 more efficiently than hnRNP A2 (Fig. 9D and E), whereas no effect on pX579 was observed ( Fig. 9D and E), as expected, since pX579 lacks the target site for hnRNP A1 and A2. Interestingly, the A2A1 hybrid inhibited splicing to the same extent as hnRNP A1 (Fig. 9D and E), whereas the ability of the A1A2 hybrid to inhibit splicing was lessened compared to that of wild-type hnRNP A1 (Fig. 9D and E). Combined, our results suggested that the conserved N terminus of hnRNP A1 and A2 interacted with the same HPV16 splicing inhibitory RNA element, and that the C terminus of hnRNP A1 was a more efficient inhibitor of splicing between HPV16 SD226 and SA409 than the C terminus of hnRNP A2. The results obtained with the hnRNP A1 and hnRNP A2 hybrids also strongly suggested that the activation of HPV16 SA742 by hnRNP A2 most likely was mediated by interactions of hnRNP A2 with sequences present elsewhere on the HPV16 pre-mRNAs than the splicing inhibitory between nucleotides 594 and 604. These other sequences were present in pC97ELsL but not in pX604. Taken together, hnRNP A1 and hnRNP A2 both inhibited splicing between the major HPV16 splice sites SD226 and SA409, but hnRNP A1 primarily promoted the production of unspliced E6-encoding mRNAs, while hnRNP A2 primarily promoted splicing to SA742, perhaps by interacting with other HPV16 sequences.
Overexpression of hnRNP A1 or hnRNP A2 inhibits splicing between SD226 and SA409 on mRNAs produced from episomal HPV16 DNA. To investigate if G (Life Technologies) for 1.5 h at 4°C, followed by washing four times in RIP buffer supplemented with 300 mM NaCl. The immunoprecipitated RNA then was extracted as described previously and subjected to RT-PCR by HPV16-specific primers 773s and E42AS, which detect HPV16 mRNA spliced between HPV16 splice sites SD880 and SA3358. Input represents RT-PCR with the same primers on RNA extracted from 5% of the input in the immunoprecipitation incubation. RT-minus, RT-PCR in the absence of reverse transcriptase. M, molecular weight marker; *, free primers.   (pA1NA2C and pA2NA1C). RT-PCR primers 97S and X765A were used (for the location of these RT-PCR primers, see (Continued on next page) hnRNP A1 and A2 Control HPV16 E6 and E7 mRNA Splicing Journal of Virology overexpression of hnRNP A1 and A2 could inhibit splicing from SD226 to SA409 on HPV16 mRNAs produced from episomal genomic HPV16 DNA, we used plasmid pHPV16AN, which encodes the full-length HPV16 genome (Fig. 10A). The genomic plasmid pHPV16AN contains two lox sites that flank the HPV16 genome. Their presence allows the release and formation of the episomal form of the HPV16 genome by cotransfection with a cre enzyme-expressing plasmid (Fig. 10A). Plasmid pHPV16AN was cotransfected with a cre-expressing plasmid in the absence or presence of hnRNP A1 or A2 expression plasmid. Cre enzyme-mediated release and formation of circular episomal HPV16 DNA was confirmed by DNA extraction and DNA PCR using primers 16S and 16A flanking the loxP sites (Fig. 10B). As can be seen, the episomal form of the HPV16 genome was efficiently produced in the presence of the cre enzyme (Fig. 10B). Overexpression of hnRNP A1 or A2 inhibited HPV16 mRNA splicing and promoted production of unspliced E6 mRNAs over HPV16 mRNAs spliced between SD226 and SA409 (Fig. 10C). Quantitation of the RT-PCR bands is shown in Fig. 10D. We concluded that overexpression of hnRNP A1 or A2 inhibited splicing between HPV16 SD226 and SA409 on mRNAs produced from the episomal form of the HPV16 genome. Knockdown of hnRNP A1 enhances splicing between HPV16 SD226 and SA409 on mRNAs produced from episomal HPV16 genome. Since overexpression of hnRNP A1 or hnRNP A2 inhibited splicing between SD226 and SA409, knockdown of hnRNP A1 or hnRNP A2 should have the opposite effect. Transfection of full-length HPV16 genome pHPV16ANϩcre with scrambled short interfering RNAs (siRNAs) or siRNAs directed to either hnRNP A1 or hnRNP A2, revealed that knockdown of hnRNP A1 resulted in enhanced splicing between SD226 and SA409 ( Fig. 11A and B), whereas knockdown of hnRNP A2 had no measurable effect on splicing between SD226 and SA409 ( Fig. 11A and B). Western blotting revealed that the siRNAs specifically reduced levels of hnRNP A1 or hnRNP A2 protein in the siRNA-transfected cells (Fig. 11C). Taken together, these results confirmed that hnRNP A1 controlled HPV16 E6 and E7 mRNA splicing and supported the idea that high levels of hnRNP A1 inhibit splicing between SD226 and SA409, thereby promoting production of full-length HPV16 E6 mRNA.
Knockdown of hnRNP A1 in HPV16-positive cervical cancer cell line SiHa alters HPV16 E6/E7 mRNA splicing and increases HPV16 E7 protein levels. To investigate if knockdown of hnRNP A1 or A2 affected splicing of HPV16 mRNAs in HPV16-driven cancer cells, HPV16-positive cervical cancer cell line SiHa was transfected with siRNAs to hnRNP A1 or hnRNP A2. As can be seen, knockdown of hnRNP A1 enhanced levels of the HPV16 E7 mRNAs spliced between SD226 and SA409 ( Fig. 12B and C), whereas knockdown of hnRNP A2 caused a reduction of unspliced E6 mRNAs without substantially affecting HPV16 mRNAs spliced between SD226 and SA409 ( Fig. 12B and C). As predicted, knockdown of hnRNP A1 also enhanced production of the HPV16 E7 protein, whereas knockdown of hnRNP A2 had no monitorable effect on E7 (Fig. 12D, E, and F). We concluded that hnRNP A1 contributed to the control of HPV16 E7 protein production in HPV16-positive cancer cells by regulating HPV16 mRNA splicing.

DISCUSSION
hnRNP A1 is a modular RNA binding protein consisting of two RNA recognition motifs (RRM1 and RRM2) and a glycine-rich C-terminal region rich in Arg-Gly-Gly (RGG) motifs that regulates alternative splicing and contributes to mRNA nuclear export, translation, and other RNA processing events (46). hnRNP A1 functions as an inhibitor     (47)(48)(49). In contrast, in HPV16, hnRNP A1 promotes retention of an intron that also encodes a portion of the E6 protein. Thus, splicing inhibition and intron retention by hnRNP A1 in HPV16 generates the only HPV16 mRNA with ability to produce full-length E6 protein.
The mechanism by which hnRNP A1 causes intron retention in HPV16 is unknown but may be similar to the exon-skipping mechanism, in which it has been proposed that initial binding of hnRNP A1 to a high-affinity binding site promotes cooperative binding of hnRNP A1 to the same mRNA, thereby extending hnRNP A1 RNA binding to splicing enhancers nearby and blocking their function (46). Alternatively, hnRNP A1 and splicing-enhancing factors are antagonists that compete for the same or overlapping binding sites. It remains to be determined how hnRNP A1 promotes HPV16 E6 intron retention.   The retained intron in the HPV16 E6 coding region is located between HPV16 splice sites SD226 and SA409. When 5=-splice site SD226 is not used, the HPV16 mRNAs will contain at least one unutilized 5=-splice site (SD226) and at least three unutilized 3=-splice sites (SA409, SA526, and SA742). Such splice sites may cause nuclear retention of these mRNAs, as the mRNAs are deemed immature and not ready for further processing and export to the cytoplasm. In that regard, it is interesting that hnRNP A1 has a shuttling function and potentially aids in the nuclear export of the HPV16 E6 mRNAs with retained introns. The PY-NSL sequence (also called M9) in the glycine-rich region of hnRNP A1 interacts with transportin 1 and 2 and appears to act as a bidirectional nuclear-cytoplasmic shuttling signal, probably with the help of transportin, as no export-specific partner of hnRNP A1 has been identified.
hnRNP A1 binding sites have been determined previously by SELEX (5=-UAGGGA/ U-3=) (50) as well as by transcriptome-wide studies (5=-UAG-3=) (51,52). Both RRM1 and RRM2 can bind RNA simultaneously and with similar sequence specificity and affinity. The preferred binding site for RRM1 was shown to be 5=-U/CAGG-3=, and that for RRM2 was 5=-U/CAGN-3= (53). Binding of hnRNP A1 to these sequences was particularly sensitive to C-substitutions, particularly AG to CG (54,55). In line with these previous observations, we mapped the hnRNP A1 binding site in HPV16 to the 5=-UAUGUUAG AUU-3= sequence in the HPV16 E7 coding region, downstream of 3=-splice site SA409. This is an 11-nucleotide, C-less sequence with a UAG motif in the middle, just like many of the previously mapped hnRNP A1-binding sites and splicing regulatory RNA elements (Fig. 7).
The HPV16 3=-splice site SA409 in the E6 coding region is a hallmark of the high-risk, cancer-associated HPV types and is a very efficiently utilized splice site. It generates the HPV16 E7 mRNA by shortening the upstream E6 orf by splicing to create an mRNA that is efficiently translated into E7 protein (26,36). One can speculate that this splice site is of significance, primarily since it generates the E7 mRNAs of the high-risk HPV types, although it has been suggested that this mRNA produces a shorter protein, named E6*I, that has a distinct function (32). Interestingly, the regulatory RNA element identified here (5=-UAUGUUAGAUU-3=) is located in the E7 coding region, even though it controls splice site SA409, which is located in the E6 coding region. According to previous reports, the E7 sequence is particularly well conserved, even compared with E6 (56). Therefore, we performed a sequence alignment of the 11-nucleotide RNA element and hnRNP A1 binding site that we had identified. Even though the sequence is not well conserved among all high-risk types (Fig. 13A), the UAGAU sequence that appears to be of key importance for hnRNP A1 binding is strictly conserved in all high-risk HPV types in the HPV alpha-9 species (Fig. 13B). Interestingly, this sequence is only moderately conserved in the HPV alpha-7 species that includes HPV18, despite the fact that it has been shown that hnRNP A1 inhibits the splice sites in the HPV18 E6 region. The binding site for hnRNP A1 in HPV18 (39) was mapped to sequences upstream and downstream of the hnRNP A1 site identified here in HPV16. Since hnRNP A1 now has been shown to inhibit splicing in the E6 region of both HPV16 and HPV18, it is reasonable to speculate that hnRNP A1 plays a significant role in the control of HPV16 and HPV18 E6 and E7 expression. hnRNP A1 has a splicing inhibitory role and promotes production of unspliced E6-encoding mRNAs while inhibiting E7 mRNA and protein production. However, it is the spliced mRNA that produces E7 protein that is the most common of the E6 and E7 mRNAs (28,29,31). It has also been suggested that splicing factor SRSF2/SC35 stabilizes all alternatively spliced HPV16 E6 mRNAs and the E7 mRNA (57). Even if hnRNP A1 is required for upholding the levels of unspliced E6 mRNAs, another cellular factor must operate in high-risk HPV16-infected cells to promote efficient splicing to 3=-splice site SA409 in the E6 region. Alternatively, hnRNP A1 or other cellular factors contribute to control of HPV16 E6/E7 mRNA splicing by interacting with the branch point or with SF1, which recognizes the branch point located upstream of the 3=-splice sites (33). One can speculate that knockdown or overexpression of hnRNP A1 and/or A2 indirectly affects HPV16 mRNA splicing by altering expression of other cellular RNA binding proteins, such as SR proteins, that contribute to control of HPV16 mRNAs splicing. Further experiments are required to understand how expression of high-risk E6 and E7 expression is controlled by alternative splicing.
The HPV16 E7 protein is the major driver of cell proliferation in HPV16-infected cells, including infected keratinocytes and HPV16-driven cancer cells (9). As unscheduled cell proliferation activated by E7 also induces apoptosis, inhibition of cell apoptosis by HPV16 E6 protein is of paramount importance (10). Since HPV16 E6 and E7 are produced from two mutually exclusive, alternatively spliced mRNAs, a balanced splicing is critical for survival of the HPV16-infected cells. This simultaneous production and relative levels of E6 and E7 are controlled by the ratio of unspliced E6 mRNAs to spliced E7 mRNAs. Cancer cells driven by HPV16 would succumb to apoptosis or would enter a state of senescence if either E6 or E7 levels were altered. Thus, splicing factors that control the ratio of the alternatively spliced HPV16 E6 and E7 mRNAs play a key role in HPV16-caused cancer. Cellular factors that control this splicing event are of major importance not only in the HPV16 replication cycle but also during carcinogenesis. It was previously shown that the levels of RNA binding proteins in HPV16-infected cervical epithelium change during carcinogenesis as well as in response to cell differentiation (22,23). We have shown that the levels of hnRNP A1 increased in premalignant lesions as well as in malignant cervical lesions and cervical cancer (22). In other cancer cells, it has been shown that knockdown of hnRNP A1 induces apoptosis (58), suggesting a general role for hnRNP A1 in maintaining the cancer phenotype of a transformed cell (59). Taken together, these observations support an important role for hnRNP A1 in the HV16 life cycle as well as during HPV16-induced carcinogenesis. However, it is the spliced form of the mRNA that encodes E7 that dominates in HPV16 infected cells. The cellular factor that promotes this splicing event and antagonizes hnRNP A1 remains to be identified.
If hnRNP A1 plays an important role in determining the optimal ratio of HPV16 E6 and E7 mRNAs and proteins, it is important to control the activity of hnRNP A1. This protein can be phosphorylated, sumoylated, and acetylated (46). A previous study suggested that splicing in the HPV16 E6 region is controlled by the epidermal growth factor pathway through hnRNP A1 (38). It was shown that activation of the Erk1/2kinase pathway promoted retention of the E6 intron and production of E6 mRNA. Knockdown of hnRNP A1, but not hnRNP A2, prevented Erk1/2-induced intron retention in the E6 region, which is in line with the results presented here, and suggested that hnRNP A1-mediated splicing inhibition and intron retention is controlled by phosphorylation of hnRNP A1. Phosphorylation of cellular splicing factors has also been shown to play a role in HPV16 late gene expression, as the Akt-kinase phosphorylates hnRNP L to suppress HPV16 late gene expression (60). The Akt-kinase also appears to play a role in the control SR proteins that regulate bovine papillomavirus late gene expression (61). One can speculate that inhibitors of enzymes that posttranslationally modify RNA binding proteins that control HPV16 gene expression by splicing can be used for treatment of premalignant lesions or cancer caused by HPV16.
Finally, we found that hnRNP A1 and A2 have opposite effects on this very vulnerable HPV16 splicing event in the E6 coding region, despite the fact that hnRNP A1 and A2 are two of the most closely related proteins in the hnRNP family of RNA-binding proteins. From a mechanistic point of view, it would be interesting to determine how these two proteins affect splicing of HPV16 E6 and E7 mRNAs in such a diametrically opposed manner. Our results suggest that the reason for the differential effects by hnRNP A1 and A2 on HPV16 mRNA splicing is primarily caused by differences in the less conserved C-terminal portion of the two proteins. Our results also suggested that hnRNP A2 interacts with other, downstream RNA sequences to promote splicing to the downstream 3=-splice SA742 rather than causing intron retention in the E6 coding region. Further experiments are needed to determine how hnRNP A2 promotes splicing to HPV16 SA742, a splice site required for production of the E6^E7 fusion protein of unknown function and for the E1 and E1^E4 mRNAs expressed from the HPV16 early promoter p97.