The Inhibitory Activity of the AU-Rich RNA Element in the Human Papillomavirus Type 1 Late 3′ Untranslated Region Correlates with Its Affinity for the elav-Like HuR Protein

Abstract

A 57-nucleotide adenosine- and uridine-rich RNA instability element in the human papillomavirus type 1 late 3′ untranslated region termed h1ARE has previously been shown to interact specifically with three nuclear proteins that failed to bind to an inactive mutant RNA. Two of those were identified as the heterogeneous ribonucleoproteins C1 and C2, whereas the third, a 38-kDa, poly(U) binding protein (p38), remained unidentified. Here we show that partially purified p38 reacts with a monoclonal antibody raised against the recently identified elav-like HuR protein, indicating that p38 is the HuR protein. Indeed, recombinant glutathione S-transferase (GST)-HuR protein binds specifically to sites within the h1ARE. Determination of the apparent K_d value of GST-HuR for the h1ARE and the inactive mutant thereof revealed that GST-HuR bound with a more than 50-fold-higher affinity to the wild-type sequence. Therefore, the binding affinity of GST-HuR for the wild-type and mutant h1AREs correlates with their inhibitory activities in transfected cells, strongly suggesting that the HuR protein is involved in the posttranscriptional regulation of human papillomavirus type 1 late-gene expression.

Human papillomaviruses (HPVs) are a group of nonenveloped, double-stranded DNA viruses that infect squamous epithelial cells (25, 54). Approximately 100 different HPV types have been identified to date. We have focused on HPV-1, which infects the cutaneous epithelium and causes benign tumors, e.g., deep plantar warts. The circular HPV genome consists of approximately 8,000 bp and encodes the early-gene products (E1 to E8), which are involved in virus replication, transcription, and host cell transformation (10, 30, 54), and the late-gene products (L1 and L2), which are the structural capsid proteins; it also carries a noncoding region containing the late 3′ untranslated region (UTR) and various cis signals for transcription and replication of the virus genome (Fig. 1A). The life cycle of HPV is dependent on epithelial differentiation (10, 29, 45, 54). Upon infection of epithelial cells in the basal layer of the epithelium, the early genes are expressed and the genome replicates (2, 25, 54). However, the production of the late structural proteins L1 and L2 is restricted to the terminally differentiated cells in the upper layers of the squamous epithelium (10, 29, 45, 54). Consequently, propagation of HPVs in vitro requires culture of HPV-infected cells in a differentiating environment, e.g., organotypic cell cultures or xenografts on nude mice (24, 34, 45).

FIG. 1.

(A) Schematic illustration of the HPV-1 genome and multiply spliced late mRNAs (2). Numbers refer to nucleotide positions in the HPV-1a genomic clone (11). The early (E1 to E7) and late (L1 and L2) ORFs are indicated as grey and white boxes respectively, and the early and late poly(A) signals [p(A)E and p(A)L] are shown as black triangles. The position of the HPV-1 AU-rich element (h1ARE) in the late 3′ UTR is shown. (B) The minimal h1ARE sequence used as the RNA probe named XB is shown. The functionally important sequence motifs are underlined.

cis-acting DNA sequences which regulate late-gene expression of papillomaviruses have been identified (4, 23, 46). In addition, at least three papillomaviruses (bovine papillomavirus type 1, HPV-1, and HPV-16) encode cis-acting RNA elements specific for the late mRNAs, which appear to play important roles in the posttranscriptional regulation of virus late-gene expression (reviewed in references 3 and 42). In the bovine papillomavirus type 1 late 3′ UTR, an unutilized 5′ splice site binds the U1 small nuclear ribonucleoprotein (snRNP) and inhibits late polyadenylation (19, 20). A negative element in the HPV-16 late mRNA 3′ UTR contains multiple 5′ splice site-like sequences (20) and an inhibitory GU-rich sequence that reduces mRNA stability in vitro (13, 20, 27). In addition, negative elements have been identified in the HPV-16 L1- and L2-encoding open reading frames (ORFs) (42, 43, 47). We previously identified an inhibitory AU-rich element (ARE) in the HPV-1 late 3′ UTR region (48). When this sequence was placed downstream of a chloramphenicol acetyltransferase (CAT) reporter gene in the sense orientation, the CAT protein and mRNA levels were reduced in transfection experiments (44). This could be explained in part by a reduction of mRNA stability in transfected cells (44). The minimal inhibitory sequence is a 57-nucleotide AU-rich sequence containing AUUUA and UUUUU motifs termed the HPV-1 ARE (h1ARE). Replacing uracil (U) with cytidine (C) in these motifs inactivated the h1ARE (44). The wild-type and mutant h1AREs were then used as substrates in a UV cross-linking assay to identify cellular proteins that bind preferentially to the wild-type sequence (44, 52, 53), since such factors could potentially be mediating the inhibitory activity of the h1ARE. Three nuclear proteins bound to the wild-type but not to the mutant inactive sequence (44). Two of those were identified as the heterogeneous ribonucleoproteins (hnRNPs) C1 and C2 (44), whereas the third, 38-kDa poly(U) binding protein remained unidentified. hnRNP C1 and A1 proteins bind AREs on cellular mRNAs, but the hnRNP A1 protein did not interact with the h1ARE (44).

The expression of several cellular early-response gene (ERG) mRNAs is controlled at the level of mRNA stability and translation through the action of AREs present in their 3′ UTRs (reviewed in references 5, 6, 9, and 39 to 41). The AREs have been grouped in multiple classes based on their differing responses to extracellular stimuli, as well as mechanistic differences of mRNA decay and the detailed cis-acting structure of the ARE (9). The AREs have been classified as AUUUA motifs containing class I (e.g., c-fos) and class II (e.g., granulocyte-macrophage colony-stimulating factor [GM-CSF]) AREs and non-AUUUA-containing AREs (e.g., c-jun) (9). The issue is further complicated by the identification of other, ARE-unrelated sequences which act as mRNA destabilizers (c-fos and c-myc coding regions) or stabilizers (α-globin 3′ UTR) (reviewed in reference 40). The HPV-1 ARE shows homology to the c-fos ARE class; i.e., it has one to three copies of nonoverlapping AUUUA motifs coupled to a U-rich region, suggesting that c-fos and HPV-1 late-gene expression may be coordinately regulated (42, 44, 48).

Proteins in the 35- to 45-kDa size range that bind specifically to AREs on cellular AREs have been identified. Many of these proteins have been implicated in the regulation of mRNA stability (reviewed in references 9, 39, and 40). The cDNA of a ubiquitously expressed, poly(U) binding, elav (embryonic lethal, abnormal vision)-like, and RNA recognition motif-containing protein named HuR was recently cloned and shown to be expressed in human epithelial cells (33, 35). HuR binds to AREs on various mRNAs (33, 35) and nonpolyadenylated snRNAs (15) in vitro. Here we show that the glutathione S-transferase (GST)-HuR fusion protein binds specifically to the functionally important AUUUA and UUUUU motifs within the HPV-1 ARE and that the binding affinity correlates with the inhibitory activity of the element in human epithelial cells.

MATERIALS AND METHODS

Plasmid constructions.

The following plasmids have been described previously: pCCKH1 and pΔXKb (53); pKSXB, pKSAUM, pKSUM, pKSAUM/UM, pKSCUC, pKSB2, pKSC1, pKSC2, pKSC3, and pKSFOS (44); and pGEX-HuR (33). To generate pT7HuR, a PCR fragment was amplified from pGEX-HuR (33) with oligonucleotides HURSTART (5′-GTCGACACAATGTCTAATGGTTATGGAG-3′) and HURSTOP (5′-GGTACCTTATTTGTGGGACTTGTTGG-3′) and ligated to pBluescript (Stratagene) digested with EcoRV and treated with calf intestinal alkaline phosphatase. To construct pCMV-LacZ, the CAT ORF in pΔKXb was replaced with the lacZ ORF from pCH110 (Pharmacia).

GST-HuR purification.

Purified GST-HuR protein was prepared by using glutathione-Sepharose (GS) beads as specified by the manufacturer (Pharmacia). Briefly, a 200-ml culture of Escherichia coli DH5α, transformed with pGEX-HuR (33), was induced with 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG) for 2 hours. The bacteria were pelleted in a Beckman centrifuge at 7,700 × g. The pellets were resuspended in ice-cold phosphate-buffered saline and lysed by short bursts of sonication followed by incubation in 1% Triton X-100. Debris was pelleted, and GST-HuR was purified from the supernatant by using the GS-beads. The yield of GST-HuR protein was determined by comparison to a bovine serum albumin standard on Coomassie blue-stained sodium dodecyl sulfate-polyacrylamide gels.

In vitro transcription, RNA gel shift assay, and UV cross-linking.

Following linearization of pKSXB, pKSAUM, pKSUM, pKSAUM/UM, pKSCUC, pKSB2, pKSC1, pKSC2, pKSC3, or pKSFOS, in vitro transcription was performed as described previously (53), except that the RNAs were labeled with [α-³²P]CTP. The molarities of the probes were calculated after determination of the specific activity by using Cerenkov β-counting. Competitor RNAs were synthesized in the absence of radiolabeled nucleotide. RNA gel shift assays were performed essentially as described previously (53). Briefly, 25 to 40 fmol (or 4 fmol in the K_d determination experiments) of radiolabeled RNA probe was incubated with 1 μg of GST-HuR protein (or dilutions thereof) in a total volume of 20 μl of binding buffer (60 mM KCl, 10 mM HEPES [pH 7.6], 3 mM MgCl₂, 1 mM dithiothreitol (DTT), 5% glycerol, 5 μg of heparin per μl) for 15 min at room temperature. In competition experiments, GST-HuR was preincubated with the RNA competitors for 5 min before addition of the radiolabeled RNA probe. The complexes were resolved on native 4% polyacrylamide gels (acrylamide/bisacrylamide ratio, 60:1). The gels were analyzed by autoradiography and quantified with a GS-250 molecular imager (Bio-Rad). For K_d determinations, both bound and free RNAs were quantified (free RNA was defined as the signal position in the gel where RNA probe alone migrated). UV cross-linking was performed as described previously (53). Probes used for UV cross-linking were labeled with [α-³²P]UTP.

Fractionation of HeLa cell nuclear extract.

Nuclear extracts were prepared from subconfluent HeLa cells by the procedure described by Dignam et al. (14). They were then dialyzed against 20 mM Tris (pH 7.5)–2 mM EDTA–1 mM DTT–50 mM KCl–20% glycerol and loaded onto a 1-ml HiTrap Mono Q column (Pharmacia) as described previously (44). Flowthrough fractions and proteins that eluted with 200 mM KCl were collected. The flowthrough fractions were dialyzed against buffer S (20 mM sodium phosphate [pH 7.0], 2 mM EDTA, 1 mM DTT, 20% glycerol) and then loaded onto an SP column (Pharmacia). Bound proteins were eluted with a linear gradient of 0.05 to 1.0 M KCl. Fractions were dialyzed against buffer S and analysed by UV cross-linking or Western immunoblotting.

Western immunoblotting and antibodies.

Western immunoblotting was performed as described previously (48), except that a mouse monoclonal anti-HuR antibody (MAb 16A5) was used as a primary antibody at a 1:5,000 dilution, and a horseradish peroxidase-conjugated rabbit anti-mouse antibody (Dako Patts) was used as the secondary antibody at a 1:10,000 dilution. MAb 16A5 is a mouse monoclonal anti-peptide antibody against a peptide of the N terminus of HuR, which will be described elsewhere. Proteins were visualized by using enhanced-chemiluminescence reagents (Amersham).

Vaccinia virus T7 expression system, transient transfection, and CAT and β-gal ELISAs.

HeLa cells were seeded in 60-mm plates and infected with recombinant vaccinia virus vTF7-3 (18), as described previously (43). Transfection was performed 1 to 2 h postinfection, and the DNA calcium phosphate coprecipitation procedure was used (22), as described previously (48). Plasmid pCMV-LacZ was included as an internal control. Cells were harvested 24 h posttransfection, and the levels of CAT and β-galactosidase (β-gal) proteins were quantified by CAT and β-gal antigen capture enzyme-linked immunosorbent assays (ELISA; Boehringer GmbH), respectively.

RESULTS

The elav-like HuR protein binds specifically to the h1ARE.

A 38-kDa protein (p38) that binds specifically to the h1ARE (44, 53) (Fig. 1A) displays characteristics similar to those of the newly identified HuR protein (15, 33, 35), i.e., molecular weight, affinity for the c-fos ARE and poly(U), and high expression in the human epithelial HeLa cell line. To investigate if p38 might be HuR, we first tested if partially purified p38 reacted with a MAb against HuR. Nuclear extract was first applied to a Mono Q column. The flowthrough fraction and the fraction eluted with 200 mM KCl (200Q) were collected. UV cross-linking to radiolabeled XB RNA showed that p38 resided in the flowthrough fraction and was undetectable in the 200Q fraction (reference 52 and data not shown). The lack of binding of p38 to the Q column is consistent with the positive charge of HuR at neutral pH (pI 9.23). The flowthrough was then fractionated on an SP column as described in Materials and Methods. The fractions were screened by UV cross-linking to radiolabeled XB RNA. p38 was found primarily in fraction 15 (Fig. 2A). Selected fractions were further analyzed by Western immunoblotting and probed with MAb 16A5 against HuR. Figure 2B shows that a protein of approximately 38 kDa could be detected in the flowthrough fraction from the Q column (lane FT) and in fraction 15 from the SP column, which both contained the 38-kDa h1ARE-binding protein, but not in fractions lacking p38 (e.g., fractions 18 and 200Q), strongly suggesting that p38 is HuR.

FIG. 2.

The partially purified, 38-kDa h1ARE binding protein is detected with anti-HuR MAb. (A) UV cross-linking of fractions eluted from an SP column (fractions 15 to 19) to radiolabeled RNA XB (Fig. 1), as described in Materials and Methods. Numbers on the left indicate the sizes (in kilodaltons) of the previously characterized proteins that UV cross-link to XB RNA (43, 52). NE, nuclear extract. (B) Western immunoblot analysis of selected SP and Mono Q column fractions with anti-HuR MAb 16A5, as described in Materials and Methods. The migration of the HuR protein is indicated by an arrow. MW, molecular weight markers (in thousands). FT, flowthrough fraction from the Mono Q column; 200Q, fraction eluted from the Mono Q column with 200 mM KCl; NE, nuclear extract. (C) Left: HeLa nuclear extract (20 μg) was UV cross-linked to 4 pmol of XB RNA followed by immunoprecipitation with anti-HuR MAb 16A5 or with MAbs against two unrelated human RNA binding proteins (MAb1 and MAb2). The supernatants (Sup.) from the immunoprecipitations are shown. Lane NE shows UV cross-linking of nuclear extract to HPV-1 XB RNA. Right: HeLa nuclear extract (20 μg) was UV cross-linked to 4 pmol of XB RNA, and the products of four cross-linking reactions were pooled and immunoprecipitated with anti-HuR MAb 16A5 (I) or with sera from nonimmunized mice (P). In lane NE, the products of one reaction of UV cross-linked nuclear extract to HPV-1 XB RNA were loaded. (D) Left: an RNA gel shift assay with nuclear extract and XB RNA was performed in the absence or presence of the anti-HuR MAb 16A5. The arrow indicates the supershift induced by the MAb. Right: An RNA gel shift assay in the absence or presence of MAbs against two unrelated human RNA binding proteins (MAb2 and MAb3).

To verify that p38 is HuR, nuclear extract was UV cross-linked to XB RNA followed by immunoprecipitation with the HuR MAb. The results revealed that p38 is specifically immunoprecipitated with the HuR MAb 16A5 (Fig. 2C). Two MAbs, mAb1 and mAb2, against unrelated human RNA binding proteins, did not immunoprecipitate proteins that were UV cross-linked to the HPV-1 RNA, and serum from nonimmunized mice did not immunoprecipitate them either (Fig. 2C). The HuR MAb also immunoprecipitated the UV-cross-linked, 38-kDa protein in fraction 15 (data not shown). To further verify that the HuR protein is present in the HPV-1 XB RNA-protein complex that forms after incubation of RNA XB in nuclear extract, an RNA gel shift assay was performed with nuclear extract and RNA XB in the absence or presence of the HuR MAb. The results show that the MAb induces a supershift (Fig. 2D, left panel). The HuR MAb also supershifted a complex between XB RNA and the 38-kDa protein in fraction 15 (data not shown). MAbs against unrelated human RNA binding proteins failed to induce supershifts (Fig. 2D, right panel). Taken together, these results demonstrate that the 38-kDa protein is HuR.

To test if the HuR protein interacts with the h1ARE, an RNA gel shift assay was performed with GST-HuR fusion protein and the XB RNA which encompasses the minimal functional h1ARE (Fig. 1B). Radiolabeled XB RNA was synthesized and incubated with 1 μg of GST-HuR protein and threefold serial dilutions thereof. The results revealed that the XB probe shifted position in a GST-HuR concentration-dependent manner (Fig. 3A), demonstrating that GST-HuR binds to the h1ARE. A 1-μg portion of purified GST protein or 1 μg of bovine serum albumin did not bind to XB (Fig. 3B and data not shown). As a further control for specificity, the GST-HuR-XB interaction was analyzed by UV cross-linking of XB RNA to GST-HuR or the GST-polypyrimidine tract binding (PTB) fusion protein. The results revealed that GST-HuR cross-linked efficiently to the XB RNA whereas GST-PTB did not (Fig. 3C). A control experiment with the hepatitis C virus 5′ UTR shows that PTB UV cross-linked, which is in agreement with previously published results (1), whereas GST-HuR did not (Fig. 3C). Further characterization of the RNA-protein interaction revealed that the uridine homoribopolymer poly(U) competed efficiently for GST-HuR binding to XB RNA whereas poly(A), poly(G), and poly(C) did not (Fig. 3D), confirming that GST-HuR has affinity for poly(U) (33). Cold XB RNA competed efficiently (Fig. 3E), demonstrating specificity. Since we have previously shown that the c-fos ARE and the h1ARE compete for the proteins in HeLa nuclear extract that UV cross-link to the h1ARE (e.g., hnRNPC1 and hnRNPC2) (44), we compared the ability of the h1ARE and the c-fos ARE to compete for the GST-HuR protein. Figure 3E shows that the XB RNA and the c-fos ARE competed to the same extent for binding to the GST-HuR protein, demonstrating that GST-HuR binds to c-fos and HPV-1 AREs with similar affinities. In conclusion, recombinant GST-HuR protein interacts specifically with the h1ARE.

FIG. 3.

GST-HuR interacts specifically with the h1ARE. (A) RNA gel shift assay with radiolabeled XB RNA in the absence of GST-HuR protein (−) or in the presence of threefold serial dilutions of 1 μg of purified GST-HuR protein. The positions of free and bound RNAs are indicated. (B) RNA gel shift assay with radiolabeled XB RNA and 1 μg of GST-HuR protein (lane HuR) or 1 μg of GST (lane GST). The positions of free and bound RNAs are indicated. −, gel shift in the absence of GST-HuR protein. (C) Left panel: UV cross-linking of 1 μg of GST-HuR (lane HuR) or 1 μg of GST-polypyrimidine tract binding (lane PTB) fusion proteins to 1 pmol of radiolabeled XB RNA (molar ratio of RNA to GST-HuR, 1:15,000). Right panel: UV cross-linking of PTB to XB RNA or hepatitis C virus 5′ UTR (HCV RNA). MW, molecular weight marker (in thousands). (D) RNA gel shift assay with radiolabeled XB RNA and 1 μg of purified GST-HuR protein after preincubation of GST-HuR with poly(A), poly(U), poly(G), or poly(C) homoribopolymer competitors (A, U, G, and C, respectively). The migrations of free and bound RNAs are indicated. (E) RNA gel shift assay with radiolabeled XB RNA and 1 μg of purified GST-HuR protein after preincubation of GST-HuR with a 10-, 3.3-, 1.1-, and 0.38-fold excess of unlabeled h1ARE (XB) or c-fos ARE (FOS) RNA competitors. The positions of free and bound RNAs are indicated. −, no competitor; −HuR, gel shift in the absence of the GST-HuR protein.

The GST-HuR protein binds to multiple sites within the h1ARE.

The h1ARE can be divided into an AU-rich region containing two AUUUA motifs and a U-rich region containing three UUUUU motifs, and both parts are necessary for the inhibitory activity of the h1ARE (44, 53). To determine the regions of the h1ARE to which the HuR protein binds, an RNA gel shift assay was performed with 1 μg of GST-HuR protein and threefold serial dilutions thereof and radiolabeled XB RNA or various radiolabeled, overlapping RNAs spanning the h1ARE (Fig. 4A). Figure 4B shows that GST-HuR binds efficiently to XB, less efficiently to C1 and B2, weakly to C3, and not at all to C2. Therefore, GST-HuR bound to all h1ARE subfragments that contained either AUUUA or UUUUU motifs but not to RNAs lacking these motifs, i.e., C2.

FIG. 4.

The h1ARE contains multiple binding sites for GST-HuR. (A) Schematic illustration of the XB RNA and the overlapping B2, C1, C3, and C2 RNAs. The AUUUA and UUUUU motifs are underlined. (B) RNA gel shift assays with radiolabeled XB, B2, C1, C3, or C2 RNAs and threefold serial dilutions of 1 μg of purified GST-HuR protein. The positions of free and bound RNAs are indicated. (C) Left: RNA gel shift assays with radiolabeled XB RNA and 1 μg of purified GST-HuR protein preincubated with a 10-, 3.3-, 1.1-, and 0.38-fold excess of unlabeled XB, C1, or B2 RNAs as competitors. The positions of free and bound RNAs are indicated. −, no competitor; −HuR, gel shift in the absence of GST-HuR protein. Right: RNA gel shift assays with radiolabeled XB RNA and 1 μg of purified GST-HuR protein preincubated with a 10-, 3.3-, 1.1-, and 0.38-fold excess of unlabeled C1 RNA or a 10-, 3.3-, and 1.1-fold excess of unlabeled C2 RNA as competitors. The positions of free and bound RNAs are indicated. −, no competitor.

To confirm these results, competition experiments in which unlabeled B2, C1, C2, and XB RNAs (Fig. 4A) were preincubated with GST-HuR protein, prior to the addition of radiolabeled XB probe, were performed. Figure 4C (left panel) shows that the B2 and C1 RNAs competed with intermediate efficiency for binding of GST-HuR to the XB RNA. The C2 RNA did not compete at all (Fig. 4C, right panel). This further suggested that both the AUUUA and UUUUU motifs were specific binding sites for the GST-HuR protein and demonstrated that the h1ARE contains multiple binding sites for GST-HuR.

GST-HuR binds specifically to the functionally important AUUUA and UUUUU motifs within the h1ARE.

We have previously demonstrated that the two AUUUA and three UUUUU motifs are functionally important and are required for the inhibitory activity of the h1ARE (44). Base changes of U to C in both the AUUUA and UUUUU motifs, converting them to ACCUA and UUCCU, respectively, inactivated the h1ARE (44). The XB RNA sequence and the substitution mutants thereof are depicted in Fig. 5A. To test if HuR binds to the AUUUA and UUUUU motifs embedded in the h1ARE, competition for binding of GST-HuR to radiolabelled XB RNA was performed with the unlabeled XB, AUM/UM, AUM, UM, and CUC RNAs (Fig. 5A) as competitors. Figure 5B shows that XB competed efficiently, CUC competed slightly less well, AUM and UM competed with intermediate efficiency, and AUM/UM did not compete at all. Note that the concentrations of the AUM competitor are lower than those of the other competitors. The results confirm that the GST-HuR protein binds specifically to the UUUUU and AUUUA motifs within the h1ARE. The XB competitor competes more efficiently than CUC with the radiolabelled probe at ratios of 1.1 and 0.38 (Fig. 5 and data not shown). The results also showed that the A residues surrounding the UUUUU motifs appeared to be important for HuR binding.

FIG. 5.

GST-HuR interacts specifically with the functionally important motifs within the h1ARE. (A) Nucleotide sequence of the XB RNA and the AUM, UM, AUM/UM, and CUC mutant RNAs. The AUUUA and UUUUU motifs are underlined. Substitution mutations of uracil (U) to cytidine (C) are marked with X. (B) RNA gel shift assay with radiolabeled XB RNA and 1 μg of purified GST-HuR protein preincubated with a 10-, 3.3-, 1.1-, and 0.38-fold excess of unlabeled CUC, UM, or AUM/UM mutant RNAs, a 10-, 3.3-, and 1.1-fold excess of XB RNA, or a 7.1-, 2.3-, 0.78-, and 0.27-fold excess of AUM. The positions of free and bound RNAs are indicated.

The binding affinity of GST-HuR for the AUUUA and UUUUU motifs in the h1ARE correlates with the inhibitory activity of the h1ARE in vivo.

Since the GST-HuR protein binds specifically to the functionally important motifs in the h1ARE, we wished to investigate if binding of GST-HuR correlates with the inhibitory activity of the h1ARE. To test this, we determined the K_d of GST-HuR for the wild-type XB RNA and the mutant and functionally defective or inactive AUM, UM, and AUM/UM RNAs (Fig. 5A) (44). An RNA gel shift assay was performed with serially diluted GST-HuR protein (ranging from 1.7 to 750 nM) and radiolabelled XB, AUM, UM, or AUM/UM RNAs (Fig. 5A). Figure 6 shows representative shifts of GST-HuR protein with XB RNA or with the mutant RNAs. All the experiments were performed at least three times, and the mean percentages of free RNA were plotted against the concentration of GST-HuR protein. Figure 6A shows that GST-HuR binds to the wild-type h1ARE (XB) with an apparent K_d of 2.8 ± 0.6 nM, indicating high-affinity binding. Conversion of the two AUUUA motifs to ACCUA, as in AUM, reduced the affinity for GST-HuR approximately threefold (apparent K_d, 8.2 ± 2.5 nM) (Fig. 6B). However, the binding affinity to GST-HuR protein was reduced >10-fold (apparent K_d of 35 ± 13 nM) when the three UUUUU motifs were converted to UUCCU, as in UM (Fig. 6C), demonstrating preferential binding of GST-HuR to sequences containing more than three uridyl residues. The AUM/UM RNA, which encode ACCUA and UUCCU motifs instead of wild-type motifs, bound to GST-HuR with the lowest affinity (apparent K_d, >150 nM) (Fig. 6D), which is in accordance with the cumulative affinity reduction of AUM and UM RNAs (>40-fold lower than for XB RNA). Overall, this ranks the affinity of the GST-HuR protein for the tested RNAs in the following order (from highest to lowest affinity): XB, AUM, UM, and AUM/UM.

FIG. 6.

Affinity determination of GST-HuR for the wild-type h1ARE or h1ARE substitution mutants. The panels show representative RNA gel shift assays with 4 fmol of radiolabeled XB (A), AUM (B), UM (C), or AUM/UM (D) RNAs with 1.5-fold serial dilutions of the GST-HuR protein, starting with 750 nM. The positions of free and bound RNAs are indicated. −, gel shift in the absence of GST-HuR protein. The graphs show the mean values of percent free XB, AUM, UM, or AUM/UM RNA quantified in three independent RNA gel shift experiments plotted against the concentration of serially diluted GST-HuR protein. The 100% values correspond to the concentration of free RNA in the presence of 1.7 nM GST-HuR protein. Free RNA was quantified by using a phosphorimager.

The K_d values determined for the interaction of GST-HuR with the XB, AUM, UM, and AUM/UM RNAs are shown in Table 1, together with the previously determined inhibitory activities of the same sequences in transient-transfection assays (44). The partially inhibitory AUM and UM sequences have a lower binding affinity for GST-HuR than does the wild-type XB RNA (Table 1). Furthermore, the functionally inactive AUM/UM sequence displays the lowest affinity for GST-HuR (Table 1). In conclusion, the binding affinity of GST-HuR for the AUUUA and UUUUU motifs in the h1ARE correlates with the inhibitory activity of the h1ARE and its derivatives in cells.

TABLE 1.

Inhibitory activities and apparent K_d values for wild-type and mutant RNA sequences

RNA	Fold inhibitory activity in HeLa cellsa	Apparent Kd for HuR binding (nM)b
XB	9.0 ± 1.2	2.8 ± 0.6
AUM	3.3 ± 0.8	8.2 ± 2.5
UM	2.8 ± 0.1	35 ± 13
AUM/UM	1.2 ± 0.3	>150

^aData from reference 44. Standard deviations are shown. 

^bMean apparent K_d values from three independent experiments with standard deviations are shown. 

Overexpression of HuR in HeLa cells.

To investigate if the HuR protein counteracts or enhances the effect of the h1ARE in cells, we overexpressed the HuR protein in the vaccinia virus T7 expression system in HeLa cells transfected in triplicate with plasmids pCCKH1 and pΔKXb (53). Plasmid pCCKH1 contains the h1ARE, but plasmid pΔKXb does not (Fig. 7A). The amount of CAT protein produced was quantified and normalized to β-gal levels produced from the pCMV-LacZ internal control plasmid. The results showed that overexpression of HuR did not alter the CAT protein levels produced from either of the two plasmids (Fig. 7B). Western immunoblot analysis shows that higher levels of HuR were produced in the cells cotransfected with the HuR expression plasmid (Fig. 7C). Further experiments are required to determine the role of the HuR protein in the posttranscriptional regulation of HPV-1 late-gene expression.

FIG. 7.

Overexpression of HuR in HeLa cells. (A) Schematic illustration of the plasmids. Plasmid names are indicated on the left. The cytomegalovirus immediate-early promoter (CMV), the CAT ORF, the HPV-1 late 3′ UTR-containing sequences, and the HPV-1 late poly(A) signals (pA1 and pA2) are indicated. Brackets mark the limits of a deletion, and lines between the boxes represent vector sequences. (B) Histogram showing the quantified CAT levels produced from pCCKH1 or pΔKXb normalized to β-gal levels produced from pCMV-LacZ in HeLa cells infected with vTF7-3 in the absence (−HuR) or presence (+HuR) of the HuR-producing plasmid pT7HuR. Fold inhibition represents the CAT levels produced from pΔKXb divided by the CAT levels produced from pCCKH1. A representative experiment performed in triplicate is shown; results are given as mean ± standard deviation. (C) Western immunoblotting on the cell extracts from the transfections with pCCKH1 in the absence (−HuR) or presence (+HuR) of pT7HuR. The HuR protein was detected with MAb 16A5.

DISCUSSION

The elav-like proteins are human homologues of the ELAV gene product initially studied in Drosophila (8). The tissue-specific Hel-N1 protein has been implicated in the regulation of mRNA degradation and translation based on its high affinity for AREs (21, 31). Similarly, the ubiquitously expressed elav-like protein HuR was characterized as having high affinity for cellular AREs (33, 35, 49), and the HuR interacted with AREs in a manner that correlated with mRNA destabilization in vivo. Furthermore, an HSUR1 ARE has been shown to bind to HuR (15). This element reduces mRNA stability when placed in the 5′ end of HSUR RNA or in the 3′ end of a stable β-globin mRNA (15). Thus, the binding of HuR to various AREs in a manner that correlates with the ARE function in cells suggests a role for HuR in the regulation of mRNA stability. Here we show that binding of HuR to the h1ARE correlates with the inhibitory activity of the h1ARE in HeLa cells (Table 1). The h1ARE acts at least partly by reducing mRNA stability (44). Taken together, these results suggest that HuR plays a role in the regulation of HPV-1 late-gene expression.

HuR was recently shown to bind to a 27-nucleotide c-fos ARE sequence which contained AUUUA, AUUUUA, and AUUUUUA motifs (33). Replacement of U’s by G’s or C’s in the AUUUA, AUUUUA, and AUUUUUA motifs resulted in lower affinity for HuR. Most obvious was the affinity loss when the motifs with more than three U residues were mutated (i.e., the AUUUUA and AUUUUUA motifs). The affinity for HuR also appeared to be sensitive to point mutations in the AUUU sequences in the (AUUU)₄CUU motifs in the 5′ end of the nonpolyadenylated HSUR1 RNAs (15). Another study demonstrated high affinity of HuR for (AUUU)₄ motifs and the c-fos ARE but not for (AGGU)₄A motifs (35). Taken together, these results are consistent with the observations made here, i.e., that HuR has affinity for both the AUUUA and UUUUU motifs (Fig. 5 and 6 and Table 1) and that HuR binds with higher affinity to sequences containing five U’s than to the AUUUA motifs (Fig. 5 and 6 and Table 1). The difference in affinity of HuR for the wild-type h1ARE and the AUM or the UM sequence is statistically significant (P < 0.05), strongly suggesting that HuR interactions with both AUUUA and UUUUU motifs are functionally important. The U-rich region within the h1ARE also has certain structural features in common with the HSUR1 ARE. In conclusion, the HuR protein binds specifically to AU-rich sequences on cellular and viral mRNAs.

We have previously been unable to detect proteins that cross-link or complex specifically with the AUUUA motifs within the h1ARE in HeLa cellular extracts (44, 53). Recently, a Mg²⁺-dependent, RNase E-like endo-RNase activity that specifically cleaves after the second U residue in AUUUA sequences has been isolated from eucaryotic cells (50). Thus, it is possible that this activity degrades the AUUUA-containing RNAs during the incubation with cellular extract in vitro, prior to UV cross-linking. Since recombinant HuR binds to the AUUUA motifs within the h1ARE, it would be interesting to investigate if the h1ARE is a target for endo-RNase degradation and if HuR may be involved in protecting the RNA. Although we show that overexpression of HuR in HeLa cells does not reverse the inhibitory effect of the h1ARE at the protein level (Fig. 7B), it is possible that the HuR protein acts locally on the RNA to protect it against endo-RNase attacks at the ARE. Alternatively, the HuR protein may act in concert with the h1ARE binding protein hnRNPC1/C2 to inhibit HPV-1 late-gene expression.

A large number of ARE binding proteins have been identified by UV cross-linking or RNA gel shift assays (9, 39, 40). The cDNA of an AU-rich RNA binding factor (AUF1 or hnRNPD) has been cloned, and the recombinant protein has been shown to bind to various AREs (in, e.g., c-fos, c-myc, β-globin, and β-adrenergic receptor mRNAs [7, 37, 51]). The AUF binding affinity correlates with the destabilizing activity of the ARE (12, 37). In addition, AUF1 has been shown to be a part of the α-globin stability complex through protein-protein interactions with the RNA-binding α-complex proteins 1 and 2 (αCP-1 and αCP-2) (28), indicating that AUF could play a role in protecting the ARE-containing mRNAs from degradation. The cDNA of a 32-kDa protein termed AUH was cloned from neuroblastoma cells, and recombinant AUH was shown to bind specifically to c-fos, interleukin-3, GM-CSF, and c-myc AREs (36). The recombinant AUH protein was shown to act bifunctionally by binding to an AUUUA matrix and catalyzing a hydratase activity simultaneously. Both AUF and AUH bind to the c-fos ARE and may therefore also interact with the h1ARE. Further studies with recombinant ARE binding proteins may elucidate the roles of various AU-rich RNA binding proteins in the regulation of HPV-1 late-gene expression.

HuR binds with similar high affinities to both the c-fos ARE (apparent K_d, 2 nM [33]) and the h1ARE (apparent K_d, 2.8 ± 0.6 [Fig. 3D and Table 1]), and c-fos expression parallels the increase in HPV-1 late-gene expression in the differentiating epithelium (17, 54), indicating that HuR may regulate the expression of both c-fos and HPV-1 late mRNAs. It would be of interest to determine the intracellular concentrations of HuR at various stages of epithelial-cell differentiation and to investigate if the concentrations of HuR that allow specific binding to AREs correlate with the expression of c-fos or HPV-1 late genes in the differentiating squamous epithelium.

It has been suggested that overexpression of HuR protein in tissue culture cells has a stabilizing effect on various mRNAs containing the GM-CSF ARE, the c-fos ARE, or the vascular endothelial growth factor ARE (16, 32, 38). Perhaps we cannot see an effect of the overexpressed HuR here since we worked with human epithelial cells whereas the other studies used murine fibroblasts. Fan et al. used mouse L929 cells to monitor these effects, since the L929 and mouse NIH 3T3 cells have low endogenous levels of HuR protein (16). Similarly, Peng et al. used NIH 3T3 fibroblasts to study the effects of HuR overexpression (38). In contrast, Levy et al. used an antisense HuR-mRNA approach to inhibit the expression of endogenous HuR protein in human 293T cells (32). Alternatively, the recombinant vaccinia virus-expressed HuR protein used here may affect the nucleocytoplasmatic distribution of HuR in the cell and in this way interfere with HuR function. Experiments to determine the effect of HuR on HPV-1 late-gene expression are in progress.

Different papillomaviruses contain different cis-acting regulatory sequences. The BPV-1 late-gene expression is controlled by an unutilized 5′ splice site present in the late 3′ UTR which binds to the U1 snRNP and inhibits late polyadenylation (19, 20). The HPV-16 late 3′ UTR-negative element contains multiple 5′ splice site-like sequences and an inhibitory GU-rich sequence that reduces mRNA stability in vitro (13, 20, 26, 27, 47). Furthermore, we have identified an ARE in the HPV-1 late 3′ UTR which interacts with the hnRNPC1/C2 and HuR proteins and acts by reducing mRNA stability and mRNA utilization (44, 48). Taken together, it appears that the papillomaviruses regulate the switch from early- to late-gene expression by heterologous posttranscriptional mechanisms, which may reflect differences in tropism and the levels of virions produced from cells infected with the different papillomavirus types. Studies of the regulation of papillomavirus late-gene expression may contribute to the development of tissue culture systems for propagation of HPVs and increase our understanding of posttranscriptional gene regulation in eucaryotic systems.