The Global Transcriptional Effects of the Human Papillomavirus E6 Protein in Cervical Carcinoma Cell Lines Are Mediated by the E6AP Ubiquitin Ligase

The function of the human papillomavirus (HPV) E6 protein thatis most clearly linked to carcinogenesis is the targeted degradationof p53, which is dependent on the E6AP ubiquitin ligase. Additionalfunctions have been attributed to E6, including the stimulationof telomerase activity and the targeted degradation of othercellular proteins, but in most cases it is unclear whether theseactivities are also E6AP dependent. While E6 clearly influencesthe transcriptional program of HPV-positive cell lines throughthe inactivation of p53, it has been shown that at least a subsetof its p53-independent functions are also reflected in the transcriptionalprogram. For this study, we have determined the extent to whichE6AP is involved in mediating the set of E6 functions that impacton the global transcriptional program of HPV-positive cell lines.The transcriptional profiles of $~$ 31,000 genes were characterizedfor three cell lines (HeLa, Caski, and SiHa cells) after smallinterfering RNA (siRNA)-mediated silencing of E6 or E6AP. Wefound that E6 and E6AP siRNAs elicited nearly identical alterationsin the transcriptional profile of each cell line. Some of theexpression alterations were apparent secondary effects of p53stabilization, while the basis of most other changes was notreconcilable with previously proposed E6 functions. While expressionchanges of the TERT gene (telomerase catalytic subunit) werenot revealed by the array, telomerase repeat amplification protocolassays showed that both E6 and E6AP knockouts resulted in asuppression of telomerase activity. Together, these resultssuggest that E6AP mediates a broad spectrum of E6 functions,including virtually all functions that impact on the transcriptionalprogram of HPV-positive cell lines.

INTRODUCTION

Top

Introduction

Human papillomaviruses (HPVs) are small DNA tumor viruses thatinfect cutaneous or mucosal epithelial cells. Approximately30 HPV types are sexually transmitted and can be further categorizedaccording to their associations with malignancies (65). Low-riskHPV types (e.g., types 6 and 11) are associated with benignlesions, while DNAs of high-risk HPV types (e.g., types 16,18, 33, and 39) are found in >95% of all cervical cancers,strongly suggesting that HPV is a causative agent of this disease(65). The only two viral proteins that are generally expressedin HPV-positive cervical cancers, E6 and E7, have been shownto be necessary and sufficient to immortalize human foreskinkeratinocytes, further suggesting that they are cooperatingoncoproteins (21, 41). HPV E7 interacts with and inactivatesthe retinoblastoma protein (pRb) and the pRb-related proteinsp107 and p130 (40). E7 inactivation of pRb activates E2F transcriptionfactors, which results in increased transcription of E2F-responsivegenes and S-phase cell cycle progression. While several otherE7-interacting proteins have been identified (including histonedeacetylase [38], p21, TATA-binding protein, and TAF110 [40]),results from animal studies indicate that the major biologicactivities of E7 are related to the inactivation of pRb, p107,and p130 (2).

The activity of the high-risk HPV E6 proteins that is most clearlyassociated with carcinogenesis is the stimulation of the ubiquitin-mediateddegradation of the p53 tumor suppressor protein (49). Both thebinding of E6 to p53 and the ubiquitination of p53 are dependenton the cellular E6AP ubiquitin-protein ligase (23, 48), thefounding member of a family of ubiquitin ligases (or E3 enzymes)that contain a carboxyl-terminal HECT domain (22, 24). However,both biochemical and genetic evidence suggests that the high-riskHPV E6 proteins have functions in addition to the inactivationof p53, some of which may also be related to cellular immortalizationor carcinogenesis. For example, HPV E6-dependent and proteasome-mediateddegradation has been reported for several PDZ domain-containingproteins, including Dlg (13), MAGI (16, 55), MUPP1 (36), andScribble (43). A C-terminal PDZ binding domain, present withinall cancer-associated HPV E6 proteins but not present in non-cancer-associatedHPV E6 proteins, mediates interactions with this set of proteins.The integrity of the C-terminal PDZ binding domain has beenlinked to the ability of E6 to induce epithelial hyperplasiain mice in vivo (44). While Scribble ubiquitination was dependenton E6AP in vitro (43), it has been reported that E6AP does notmediate the degradation of Dlg in vitro (47), suggesting thatE6 may target proteins for degradation by an alternative mechanism.E6 has also been reported to target Myc (19), E6TP1 (11, 12),and the proapoptotic Bak protein (26, 54) for degradation. Theinvolvement of E6AP in Bak and Myc degradation has not beendetermined, while E6TP1 degradation was shown to be E6AP dependentin vitro and in vivo (10). Another activity of the HPV E6 proteinwhich is linked to cellular immortalization is its ability toinduce transcriptional activation of the TERT gene (14, 45,57, 58), which encodes the catalytic subunit of telomerase.TERT activation does not appear to be a downstream effect ofp53 inactivation, since a mutant form of E6 (SAT_8-10) that cannottarget p53 retains the ability to activate telomerase (14, 32).Gewin et al. showed that the ability of HPV16 E6 mutants tointeract with E6AP correlated positively with the ability toactivate telomerase (14), and they very recently identifieda telomerase repressor (NFX1-91) that interacts with HPV E6/E6AP,suggesting that this is a novel target of E6/E6AP (15). E6-interactingproteins that are not thought to be targeted for degradationinclude paxillin (56), E6-BP (6), CBP/p300 (46, 64), and Tyk2(37).

Therefore, while the cancer-associated HPV E6 proteins appearto be multifunctional, E6AP has been shown to mediate only asubset of these functions (e.g., targeting of p53 and E6TP1).Evidence for E6AP involvement in other functions is either conflicting(e.g., targeting of PDZ domain-containing target proteins),lacking (e.g., targeting of Myc and Bak), or suggestive (telomeraseactivation). In addition, it is unclear whether all of the reportedE6 interactions are biologically meaningful and, if they aremeaningful, at which point in the viral life cycle or immortalization-transformationprocess they are most relevant. Rather than determining whetherE6AP is involved in each of the individual functions attributedto E6, we have addressed the question of E6AP involvement inE6 functions globally by examining alterations in the transcriptionalprofiles of HPV-positive cervical carcinoma cell lines uponsmall interfering RNA (siRNA)-mediated knockdown of either E6or E6AP. We predicted that if all HPV E6 functions were dependenton E6AP, then the set of transcriptional alterations due tothe knockdown of E6 would fall entirely within the set of transcriptionalalterations observed upon knockdown of E6AP. In contrast, ifonly a subset of E6 functions were mediated by E6AP, then theset of transcriptional alterations upon E6 knockdown would onlypartially overlap with the alterations observed upon E6AP knockdown.In both scenarios, the overlapping set of genes would be expectedto include those that are transcriptionally induced, eitherdirectly or indirectly, by p53. While not all functions of E6are likely to be reflected by alterations in the transcriptionalprofile, a previous microarray-based study showed that E6 inducesmany p53-independent transcriptional alterations in human keratinocytes(8). Overall, our results indicate that essentially all HPVE6 functions that impact on the global transcriptional programof HPV-positive cell lines are mediated by E6AP. In addition,a direct analysis of telomerase activity indicated that E6-inducedtelomerase activation is also dependent on E6AP.

MATERIALS AND METHODS

Top

Materials and Methods

siRNAs. siRNAs were designed based on the suggestions of Tuschl andcoworkers (9). All siRNAs were obtained from Dharmacon as purified,annealed duplexes or as single-stranded RNA oligonucleotideswhich were annealed according to the manufacturer's instructions.To minimize possible off-targeting effects, we synthesized siRNAscorresponding to two different regions of each of the targetgenes and confirmed the knockouts by Western blotting and/orreverse transcription-PCR (RT-PCR). An siRNA against green fluorescentprotein (GFP) was used as a control for changes in gene expressionas a result of siRNA transfection. The siRNA target sequenceswere as follows: E6AP-A, 5'-CAACTCCTGCTCTGAGATA; E6AP-B, 5'-GATGTGACTTACTTAACAG;18E6-A, 5'-GAGGTATTTGAATTTGCAT; 18E6-B, 5'-CTAACTAACACTGGGTTAT;16E6-A, 5'-GAGGTATATGACTTTGCTT (27); 16E6-B, 5'-ACCGTTGTGTGATTTGTTA;and GFP, 5'-GGCTACGTCCAGGGAGCGCACC.

Tissue culture, RNA isolation, and Western blotting. HPV type 18 (HPV18)-positive HeLa cervical carcinoma cells andHPV16-positive SiHa and Caski cervical carcinoma cells (AmericanType Culture Collection) were maintained in Dulbecco's modifiedEagle's medium supplemented with 10% fetal bovine serum. Cellswere plated 1 day prior to transfection to obtain 40 to 50%confluence at the time of transfection. Cells were transfectedin 15-cm-diameter plates without antibiotics by the use of 2to 3 µl of Oligofectamine (Invitrogen) per ml and a finalsiRNA concentration of 20 nM. HeLa cells were transfected inthe presence of 10% fetal bovine serum. A Cy3 fluorescentlylabeled siRNA (Ambion) was used to estimate the siRNA transfectionefficiency, which was determined to be nearly 100% for eachcell line. Cells were harvested for RNA and proteins at 48 hposttransfection. mRNAs were isolated by use of a FastTrack2.0 kit (Invitrogen). The integrity and purity of the mRNAswere determined by gel electrophoresis and UV absorbance measurements.A portion of the cells were lysed in NP-40 lysis buffer (100mM Tris [pH 7.5], 100 mM NaCl, 1% NP-40, 1 mM dithiothreitol),and the total protein concentration was determined by use ofthe Bradford assay (Bio-Rad). Equal amounts of proteins wereboiled in sample buffer and resolved by sodium dodecyl sulfate-8%polyacrylamide gel electrophoresis. Proteins were transferredto polyvinylidene difluoride membranes (Millipore), blockedin 5% nonfat milk, probed with a p53 antibody (Ab-6; Oncogene),an E6AP antibody, or a pRb antibody (IF8; Santa Cruz Biotechnology)according to standard procedures, and analyzed by enhanced chemiluminescence(Super Signal West Pico enhanced chemiluminescent substrate;Pierce).

RT-PCR and real-time quantitative PCR. mRNAs (250 to 500 ng) were reverse transcribed by the use ofSuperscript II (Invitrogen) and either an oligo(dT) primer,random primers (Invitrogen), or a gene-specific primer, as suggestedby the manufacturer. Semiquantitative RT-PCRs for 18E6 and E6APwere performed by using 1 µl of a 1:20 dilution of cDNAthat had been synthesized with random primers. The cycling parameterswere 95°C for 2 min and then 35 cycles of 95°C for 15s, 55°C for 20 s, and 75°C for 40 s, with a final extensionstep of 75°C for 2 min. The bands were quantified with QuantityOne software (Bio-Rad). Real-time quantitative PCRs were performedin a 96-well plate with 20-µl reaction volumes containing1 µl of cDNA template that had been synthesized with anoligo(dT) primer or a gene-specific primer by use of a DynamoHS SYBR green qPCR kit (Finnzymes) and an ABI Prism 7700 (AppliedBiosystems) real-time detection system. The reactions were incubatedat 50°C for 2 min and 95°C for 15 min and then for 40cycles of 95°C for 10 s, 56 to 58°C for 20 s, and 72°Cfor 30 s. Data were analyzed with ABI Prism sequence detectionsystem 1.9.1 software (Applied Biosystems). Glyceraldehyde-3-phosphatedehydrogenase (GAPDH) was used as a reference gene, and sampleswere amplified in duplicate. Relative expression was calculatedas , where is and ${Delta}$ ${Delta}$ C_T is .

Microarray hybridization and analyses. For microarray analysis, mRNAs were isolated from mock-transfectedand siRNA-transfected cells. mRNAs (3 µg) were reversetranscribed in the presence of amino-allyl deoxyuridine triphosphates.Cy3 and Cy5 fluorescent dyes (Amersham Biosciences) were thencoupled to the amino-allyl groups. Cy3 (green) was coupled tocDNAs from mock-transfected cells, and Cy5 (red) was coupledto cDNAs from siRNA-transfected cells. The two cDNA probes werecombined and simultaneously hybridized to spotted cDNA microarraysprovided by V. Iyer (University of Texas). Each microarray contained $~$ 49,000 elements, including $~$ 31,000 discrete Unigene clusters(61). The microarrays were scanned with a GenePix 4000 scanner(Axon Instruments), and the accompanying GenePix Pro 4.2 softwarepackage was used to locate and extract data from individualspots on the microarrays. Data filtering was performed by useof the Longhorn Array Database, a relational microarray databaseand analysis tool set (29). The elements (genes) chosen forfurther analysis were selected based on several data qualitystandards, including minimum intensity and pixel consistency.Genes were considered significant if the |log₂R| was $≥$ 0.85, whereR is the red/green normalized ratio. Only genes whose expressionwas similarly altered for both siRNAs against a single targetgene and was not altered in controls were considered significant.

Telomerase assays. Telomerase activities were assayed by a telomerase repeat amplificationprotocol (TRAP) as previously described (35), with slight modifications.Briefly, cells were washed with phosphate-buffered saline, pelleted,washed in a buffer containing cold 10 mM Tris-HCl (pH 7.5),1.5 mM MgCl₂, 10 mM KCl, and 1 mM dithiothreitol, pelleted again,and lysed in lysis buffer (10 mM Tris-HCl [pH 7.5], 1 mM MgCl₂,1 mM EGTA, 5 mM ß-mercaptoethanol, 0.5% CHAPS, 10%glycerol, and 0.1 mM phenylmethylsulfonyl fluoride) at 4°Cfor 30 min. Cell debris was pelleted by centrifugation at 12,000x g, and the protein concentrations in the supernatants weredetermined by use of the Bradford assay (Bio-Rad). TRAP assayswere performed in 0.2-ml PCR tubes containing 50 µl ofa reaction mixture composed of TRAP buffer (20 mM Tris-HCl [pH8.3], 1.5 mM MgCl₂, 63 mM KCl, 0.005% Tween 20, and 1 mM EGTA),a 0.2 mM concentration of each deoxynucleoside triphosphate,0.1 mg of bovine serum albumin/ml, 0.002 µg of TS primer(5'-AATCCGTCGAGCAGAGTT-3')/µl, 2 U of Taq DNA polymerase(Invitrogen), 0.3 µl of [ ${alpha}$ -³²P]dGTP (10 µCi/µl;Perkin-Elmer), and 0.1 or 0.02 µg of protein. After 20min at room temperature, 0.1 µg of CX primer (5'-CCCTTACCCTTACCCTTACCCTAA-3')was added. PCRs were performed for 27 cycles of 30 s at 94°C,30 s at 50°C, and 1 min at 72°C. Ten microliters ofeach reaction mixture was subjected to electrophoresis in 12%nondenaturing polyacrylamide gels (1x Tris-borate-EDTA). Thegels were dried and subjected to autoradiography.

RESULTS

Top

Results

Synthetic double-stranded siRNAs were designed to target themRNAs encoding human E6AP and the HPV type 16 and 18 E6 proteins.Two functional siRNAs were identified for each target. The siRNAsagainst E6AP, designated E6AP-A and E6AP-B, were targeted totwo 21-bp regions within the first 320 bases of the E6AP openreading frame (ORF). Both siRNAs were predicted to target mRNAsencoding all of the known isoforms of E6AP (31, 62). siRNAsagainst HPV16 and HPV18 E6-encoding mRNAs were designed to avoida simultaneous knockdown of HPV E7, as E6 and E7 are translatedfrom an alternatively spliced bicistronic mRNA (7, 50). E6 isexpressed from the unspliced transcript, which includes thecomplete E6 and E7 ORFs, while E7 is translated from a transcriptthat has been spliced to remove an intron that lies entirelywithin the E6 ORF. siRNAs were therefore targeted to the intronregion to specifically knockdown the unspliced E6 and E7 transcripts.This strategy has been shown previously to knockdown E6 expressionwithout affecting E7 expression (5, 27).

Transfection of either the E6AP-A or -B siRNA into three HPV-positivecell lines (HeLa, Caski, and SiHa) resulted in a decrease inE6AP protein levels by 24 h posttransfection, with a maximaldecrease between 48 and 72 h (data for the 48-h time point areshown in Fig. 1A to C). As expected, the p53 protein was stabilizedover the same time course. Although both siRNAs against E6APresulted in a significant decrease in the amount of E6AP proteinand an accumulation of the p53 protein, the E6AP-A siRNA consistentlyresulted in lower levels of E6AP and higher levels of p53 thandid E6AP-B. Both HPV18 siRNAs (18E6-A and -B) resulted in thestabilization of p53 in the HPV18-containing cell line (HeLa),while the HPV16 siRNAs (16E6-A and -B) resulted in the stabilizationof p53 in the HPV16-containing cell lines (Caski and SiHa).As a further control for the specificity of the E6 siRNAs, wedetermined that the HPV18 E6 siRNAs did not stabilize p53 inthe HPV16-containing cell lines and that the HPV16 E6 siRNAsdid not stabilize p53 in HeLa cells (data not shown). Immunoblotsdemonstrating the knockdown of E6 proteins are not shown becauseof the difficulty in detecting the low levels of E6 proteinin these cell lines with available antibodies; however, a semiquantitativeRT-PCR analysis showed that E6-expressing transcripts were reducedapproximately threefold by the E6-A and E6-B siRNAs (Fig. 1D).E6AP transcript levels were reduced four- to sixfold by theE6AP-A and E6AP-B siRNAs. E6AP siRNAs did not affect E6 transcriptlevels, and vice versa (Fig. 1D).

View larger version (30K):
[in this window]
[in a new window]
FIG. 1. siRNAs against E6AP and HPV E6 stabilize p53. Western blots of extracts from HeLa (A), Caski (B), and SiHa (C) cells after transfection with siRNAs (20 nM) against E6AP and HPV E6 are shown. The cells were harvested at 48 h posttransfection, and proteins were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and immunoblotted by use of a rabbit polyclonal antibody against E6AP (upper panels) or a mouse monoclonal antibody against p53 (lower panels). (D) Semiquantitative RT-PCRs confirmed the knockdown of E6AP (upper panel) and HPV E6 (middle panel) transcripts. GAPDH was used as a control (lower panel). (E) Western blot of HeLa cell extracts with a mouse monoclonal antibody against the retinoblastoma protein (pRb) showing no change in protein levels following the knockdown of HPV 18E6 or E6AP. pRb-p, phosphorylated pRb.

Although the expression of E6 and E7 has been shown to be requiredfor the continual growth of HPV-positive cervical carcinomacell lines (18, 59), we did not observe any morphological indicationsof apoptosis or a loss of viability of cells transfected witheither E6 or E6AP siRNAs. In addition, p53 levels returned tonormal by $~$ 7 days posttransfection, and the viability of cellsat late time points was similar to that of cells transfectedwith a control siRNA against GFP. All of the experiments describedhere used a final concentration of 20 nM siRNA, which has beenshown to be within the optimal range for observing siRNA silencingwhile minimizing off-targeting events and nonspecific effects(51). Transfection with a higher concentration (200 nM) of E6or E6AP siRNAs, but not GFP siRNA, did result in cell deathat later time points (5 to 7 days), in agreement with previousreports (5). Cell death was likely due to a more efficient andprolonged inactivation of E6- and E6AP-dependent degradationof p53 at the higher siRNA concentration.

We examined the effects of E6AP and E6 siRNAs on the globaltranscriptional profiles of the HeLa, Caski, and SiHa cell lines48 h after siRNA transfection. Fluorescently labeled (Cy5; red)cDNAs from siRNA-transfected cells were mixed with fluorescentlylabeled (Cy3; green) cDNAs from mock-transfected cells and hybridizedto spotted cDNA microarrays. Microarray analyses of siRNA experimentswere repeated at least four times for each cell line. The resultsdiscussed below are from duplicate transfections hybridizedwith spotted cDNA microarrays from a single printing of thearrays for consistency in data analysis. Data were filteredfor spot intensity and uniformity, and genes that were inducedor repressed $≥$ 1.8-fold were considered significant. In addition,only those genes that were similarly affected by both siRNAsagainst a single target gene were considered significant inour analyses. Cells transfected with the GFP siRNA were usedto control for artifacts due to siRNA transfection and/or microarrayhandling. A small number of mRNAs (<10) were reproduciblyaltered $≥$ 1.8-fold after GFP siRNA transfection into the threecell lines, although none of these genes were affected by theE6 or E6AP siRNAs.

To validate the general approach of combining siRNA transfectionwith microarray analyses, we initially examined microarray elements(i.e., specific genes) for predicted effects of E6AP or E6 siRNAs.For example, E6AP mRNA levels were confirmed to be lower (green)in cells that were treated with E6AP siRNAs, but not in cellstreated with E6 or GFP siRNAs (Fig. 2A). In addition, whilep53 protein levels were induced by both E6 and E6AP siRNAs (Fig.1), p53 transcript levels were not expected to change, sincethe stabilization of p53 occurs posttranslationally, and thiswas also confirmed by semiquantitative RT-PCR (data not shown).In contrast, genes that are known to be directly transcriptionallyactivated by p53 were expected to be induced after the transfectionof either E6 or E6AP siRNAs. As shown in Fig. 2A, the genesencoding p21, HDM2, and p53-inducible nuclear protein 1 (p53INP1)were induced after transfection with either E6 or E6AP siRNAs.A real-time quantitative PCR analysis confirmed the up-regulationof transcript levels for HDM2 and p21 (Fig. 2B). We also examinedgenes predicted to be affected by changes in the expressionof HPV E7 in order to indirectly confirm that our E6 siRNAsdid not also alter E7 functions. HPV E7 interacts with pRb,p107, and p130, activating E2F transcription factors and increasingthe transcription of E2F-responsive genes, such as those encodingcdc6, PCNA, thymidine kinase, minichromosome maintenance proteins,cyclin A, cyclin E1, cdc25A, E2F2, E2F3, and dihydrofolate reductase(reviewed in reference 52). Therefore, if our E6 siRNAs alsoreduced E7 expression, we would expect this set of E2F targetgenes to be down-regulated. No E2F target genes were significantlyup- or down-regulated by either the E6 or E6AP siRNAs (the expressionof cdc6 and dihydrofolate reductase is shown in Fig. 2A). Additionally,E7 has been reported to target pRb for proteasomal degradation(4, 17, 60), although a candidate ubiquitin ligase has not beenidentified. The fact that E6AP siRNAs did not affect E2F-regulatedgenes or pRb protein levels (Fig. 1E) indicates that E6AP doesnot mediate the E7-dependent degradation of pRb. This is consistentwith in vitro experiments that have also failed to show an involvementof E6AP in E7-dependent pRb ubiquitination (unpublished results).

View larger version (26K):
[in this window]
[in a new window]
FIG. 2. Validation of siRNA and microarray approach. (A) Specific genes were expected to be up- or down-regulated or unchanged after E6, E6AP, or GFP siRNA transfection into HeLa, Caski, and SiHa cells. Each column represents an individual experiment (with a separate mRNA sample and array hybridization), and each row represents a single gene. Red and green represent up-regulation and down-regulation, respectively, relative to the expression level in mock-transfected cells. Black represents equal expression relative to mock-transfected cells. (B) Real-time quantitative PCR analysis verifying the up-regulation of HDM2 and p21 and the down-regulation of hTERT. MT, mock transfected.

The overall expression profiles after siRNA transfection werefirst analyzed separately for each of the three HPV-positivecell lines. For effects to be considered significant, individualgenes had to be either induced or repressed $≥$ 1.8-fold by bothsiRNAs against either E6 or E6AP in at least three of the fourexperiments shown in Fig. 3. The total number of genes affectedin each cell line was 75 for HeLa, 38 for Caski, and 17 forSiHa. Together, these represented a total of 99 genes, as thesets were partially overlapping (discussed below). Figure 3illustrates the relative degrees of alteration in mRNA levelsfor the 99 genes across all three cell lines, and Table 1 liststhe complete set of 99 genes by accession number, gene name,and symbol, indicates in which cell line(s) they were affected,and shows whether they are known p53-responsive genes. TableS1 in the supplemental material shows the same information alongwith the functions of the gene products (if known) and the relativeinduction/repression ratio for each gene in all of the experimentsshown in Fig. 3. Interestingly, all of the genes that met thecriteria for being significantly affected were up-regulated(i.e., their relative expression is represented in red in Fig.3). A dye switch control experiment resulted in the same setof genes being up-regulated but appearing green (not shown).The most striking result was that for each cell line, the setof genes affected by the E6 siRNAs was, with only two exceptions,completely concordant with the set of genes affected by theE6AP siRNAs. That is, there were almost no genes that were specificallyaffected by the E6 siRNAs that were not also affected by theE6AP siRNAs, and vice versa. The two exceptions were the genesfor ENTPD7 (ectonucleoside triphosphate diphosphohydrolase 7)and interleukin-11, which were up-regulated by both 18E6 siRNAs,but not by the E6AP siRNAs, in HeLa cells. Overall, with theexception of these two genes in HeLa cells, these results suggestthat nearly the complete spectrum of HPV E6 protein functionsthat are reflected in the transcriptional program of HPV-positivecell lines are mediated by E6AP. Furthermore, the sets of genesaffected by the E6 and E6AP siRNAs were essentially identical,indicating that no natural (E6-independent) functions of E6APare reflected in the transcriptional profiles of these celllines.

View larger version (38K):
[in this window]
[in a new window]
FIG. 3. Cluster image showing expression profiles of HeLa, Caski, and SiHa cells after transfection with siRNAs against HPV E6 and E6AP. Columns, rows, and colors are the same as those described in the legend to Fig. 2. Gray represents inadequate or missing data (data that did not pass the filter criteria). The numbers assigned to each gene correspond to those used in Table 1 for gene descriptions. The identities of some of the specific genes discussed in the text are indicated to the left. M (top) represents a control experiment in which cDNAs from mock-transfected cells were labeled with Cy5 and hybridized with the same cDNAs labeled with Cy3.

View this table:
[in this window]
[in a new window]
TABLE 1. Genes that were significantly altered by siRNAs^a

Figure 4 illustrates the overlap among the total set of 99 genesaffected by E6 and/or E6AP siRNAs in the three cell lines. Twenty-threeof the 99 genes have been previously shown to be induced, eitherdirectly or indirectly, by wild-type p53. Fifty-six genes wereinduced specifically in HeLa cells that were not significantlyinduced in either the Caski or SiHa cell line, and nine of thesewere known p53-responsive genes. Nineteen genes, two of whichwere known p53-induced genes, were specific to Caski cells,and five genes, two of which were known p53-induced genes, werespecific to SiHa cells. A set of 12 genes was up-regulated inall three cell lines, and 7 of the these genes were known p53-responsivegenes. The other four genes in this set were uncharacterizedexpressed sequence tags or encoded proteins of unknown function.

View larger version (13K):
[in this window]
[in a new window]
FIG. 4. Overlap of affected genes among cell lines. The total number of genes affected by the E6 and E6AP siRNAs for each cell line is indicated in parentheses; the Venn diagram indicates the overlap of these sets of genes among the three cell lines.

The TERT gene, encoding the catalytic subunit of telomerase,was expected to be down-regulated following E6 knockdown, sinceit has been shown that E6 induces the transcription of TERT(57). However, we could not analyze TERT expression alterationson the array because the intensity of the hybridization signalat the TERT microarray element was consistently low and thespot was filtered out during the selection of high-quality data.We therefore directly assayed the telomerase catalytic activityin extracts from HeLa cells that had been transfected with E6or E6AP siRNAs. Using a TRAP assay, we found that the transfectionof either of the siRNAs against HPV18 E6 resulted in a significantdecrease in telomerase activity compared to that in mock-transfectedcells (Fig. 5). As a control for specificity, transfection withHPV16 E6 siRNAs did not affect telomerase activity in HeLa cells(an HPV18 cell line). Additionally, the transfection of eitherthe E6AP-A or -B siRNA resulted in a decrease in telomeraseactivity equivalent to that seen with the HPV18 E6 siRNAs. Thedown-regulation of hTERT transcript levels by both E6 and E6APsiRNAs was verified by real-time quantitative PCR (Fig. 2B).These results indicate that E6AP mediates the ability of E6to activate telomerase, a process that was shown previouslyto be p53 independent (14, 32).

View larger version (93K):
[in this window]
[in a new window]
FIG. 5. Telomerase activity in HeLa cells decreases after transfection with siRNAs against HPV18 E6 or E6AP. TRAP assays were performed on extracts from cells transfected with siRNAs targeting either HPV18 E6, E6AP, or HPV16 E6 (control) or on extracts from mock-transfected cells. Results for duplicate transfections are shown, except for the HPV16 E6 siRNA transfections, which were done with two different HPV16 E6 siRNAs.

DISCUSSION

Top

Discussion

The goal of this study was to determine if a broad spectrumof functions of the high-risk HPV E6 proteins was dependenton the E6AP ubiquitin ligase. This was addressed by comparinggene expression profiles of >30,000 genes after siRNA-mediatedsilencing of E6 or E6AP expression in HPV-positive cell lines.This experimental approach had the advantages of (i) being unbiasedwith respect to the significance of reported E6 functions, (ii)revealing potential effects due to as yet uncharacterized functionsof E6, and (iii) having, as an internal control, the gene expressionalterations that were predicted to accompany the activationof p53. A disadvantage of this approach is that it is likelythat some functions of E6 are not manifested at the level ofgene expression alterations. Since the E6/E6AP complex functionsas a ubiquitin ligase, it would clearly be desirable to alsoexamine alterations in global protein levels upon E6 or E6APsilencing. However, until protein array technology becomes availablethat allows for the surveillance of a significant fraction ofthe proteome (63), gene expression analysis provides the mostcomprehensive and unbiased method for analyzing the responseof HPV-positive cells to the depletion of E6 or E6AP.

Strikingly, our overall finding was that in all three HPV-positivecell lines examined, the alterations in gene expression followingE6 silencing were nearly identical to the alterations observedfollowing E6AP silencing. This indicates (i) that the spectrumof E6 functions that are reflected in the transcriptional programis dependent on E6AP and (ii) that essentially no normal (i.e.,E6-independent) functions of E6AP are reflected in the transcriptionalprofiles of the cell lines examined here. Of the set of 99 genesthat were affected in HeLa and/or SiHa and Caski cells, 23 wereeither known direct targets of the transcriptional activationfunction of p53 (e.g., p21, HDM2, TP53INP1, and Maspin) or knownto be induced indirectly after p53 activation (e.g., Notch1and gelsolin).

Our experimental design was based on the presumption that E6siRNAs would not affect E6AP mRNA or protein levels, and viceversa. We showed that this was true at the mRNA level by examiningE6 and E6AP transcripts by RT-PCR, and we showed by immunoblottingthat E6 siRNAs did not affect E6AP protein levels. The difficultyof detecting the E6 protein in HPV-positive cervical carcinomasdue to the low level of protein expression and possibly to theuse of poor immunologic reagents prevented us from demonstratingthat E6 protein levels were not posttranslationally affectedby the silencing of E6AP. Our microarray results are not consistentwith a scenario in which E6 is stabilized by the silencing ofE6AP, since the transcriptional effects of the E6AP knockdownwere nearly identical to the effects of silencing E6. Conversely,if the silencing of E6AP led to a concomitant decrease in E6protein levels, we would have expected our microarray resultsto be similar to those observed, and while our overall conclusion—thatall E6 functions are dependent on E6AP—would have beenthe same, the underlying mechanism would have been quite distinct.Importantly, however, there are no reports in the literatureto suggest that E6AP affects E6 protein expression levels, anda previous study has shown that steady-state levels of functional,epitope-tagged E6 proteins are not influenced, either positivelyor negatively, by their ability to interact with E6AP (28).It therefore seems unlikely that the silencing of E6AP wouldhave affected E6 protein levels in our experiments.

While we did not see overt signs of apoptosis when we transfectedcells with E6 or E6AP siRNAs at a concentration of 20 nM, manyof the induced genes encode apoptosis-related proteins, includingNotch1, gelsolin, and STK17A. At the same time, a small numberof antiapoptotic proteins were also induced (TRAF1, GAS6, andPEA15). It was not possible to directly link the induction ofthe apparent p53-independent genes to other targets or bindingpartners of E6, such as PDZ domain-containing proteins, E6TP1,Bak, Myc, or CBP/p300. It should be noted, however, that neitherMyc nor Bak was stabilized after the E6 or E6AP siRNA knockoutand that neither protein was ubiquitinated in the presence ofE6 or the E6/E6AP complex in vitro under conditions in whichp53 was efficiently recognized (S. Beaudenon, M. L. Kelley,and J. M. Huibregtse, unpublished data). The basis of differencesin observations with Myc (19) and Bak (53) relative to publishedreports is unknown.

All cervical cancer-associated HPV E6 proteins, but not otherhuman and animal papillomavirus E6 proteins, contain a PDZ domainbinding epitope consisting of a T-X-V/L sequence at their Ctermini. The high-risk HPV E6 proteins have been shown to interactwith several PDZ domain-containing proteins and to target themfor degradation, including Dlg (13), Scribble (43), MUPP1 (36),and MAGI (16, 55). Scribble ubiquitination has been shown tobe E6AP dependent (43), although Dlg degradation has been reportedto be E6AP independent (47). While work from our lab indicatesthat Dlg degradation is also E6AP dependent (Beaudenon et al.,unpublished data), it is not clear how any of the observed expressionchanges that we observed upon E6 or E6AP silencing might belinked to the stabilization of any of the known PDZ domain-containingtarget proteins. Scribble and Dlg have been linked to cell growthand differentiation decisions through their influence on epithelialcell polarity in Drosophila (3). Potential gene expression alterationsdue to Scribble or Dlg stabilization may be a reflection ofthe role that cell adhesion and junctional complexes can haveon signaling pathways in epithelial cells (25). It is perhapssignificant in this regard that several of the genes that wereup-regulated after an E6 or E6AP siRNA treatment were involvedin cell adhesion, motility, or extracellular matrix interactions(fibronectin, MCAM, tenascin C, and Jagged 1).

Although there was an overlap of affected genes after the knockdownof HPV E6 and E6AP among these three cell lines, with the majorityof these genes known to be p53 responsive, many of the observedeffects appear to be cell line specific. HPV E6 and E6AP silencingappears to have a greater impact on the transcriptional programof HeLa cells than on that of either Caski or SiHa cells. Thesedifferences among cell lines are not explained by differencesin the transfection efficiencies of the cell lines, as thesewere determined to be nearly 100% for all three lines. Whilewe cannot offer an obvious explanation for differences in thetranscriptional program responses, it is likely that the inherentgenetic differences among these cell lines are the main contributors.In addition, the tissue types from which the original tumorsarose were not the same. HeLa cells were derived from an adenocarcinoma,the SiHa cell line was derived from a squamous cell carcinoma,and Caski was derived from a metastasis of a cervical carcinoma(American Type Culture Collection [www.atcc.org]).

The stabilization of telomeres or chromosome ends, often bythe activation of telomerase activity, is a hallmark of cancercells (20). It was previously shown that the high-risk HPV E6proteins stimulate telomerase activity and that this stimulationis independent of the effects of E6 on p53 (14, 32). While inductionoccurs at the level of transcriptional activation of the TERTgene (14, 45, 57, 58), TERT gene expression was not clearlyaltered in our microarray experiments after E6 or E6AP silencing.However, the strength of the hybridization signal at the TERTelement on the array was reproducibly very low. This was consistentwith an earlier microarray-based study that also failed to detectE6-dependent up-regulation of TERT on the array, even thoughtelomerase was in fact induced in those experiments (8). Thisprompted us to directly assay the telomerase activity to determinethe effects of E6AP on this function. We found that both E6and E6AP siRNAs led to a significant reduction in telomerasecatalytic activity. The existence of an E6/E6AP target involvedin the regulation of hTERT transcription was previously suggestedbased on correlative data showing that HPV16 E6 mutants thatretained the ability to interact with E6AP also retained theability to activate telomerase, while mutants that did not interactwith E6AP did not activate telomerase (14). In agreement withthese results, it was recently reported that the transductionof HFK cells with a siRNA targeting E6AP, with or without HPV16E6 expression, also decreased telomerase activity (15). Thisprevious report identified an E6/E6AP-interacting protein thatis a known repressor of telomerase transcription, NFX1-91, andshowed that the expression of E6/E6AP stimulated the ubiquitinationand degradation of NFX1-91, suggesting that NFX1-91 is a substrateof E6/E6AP (15). Interestingly, E6 has been shown to be presentat the TERT promoter by the use of chromatin immunoprecipitationassays (58), and Schlegel and coworkers have also found thatE6AP is present at the TERT promoter in E6-expressing cells(37a). There is accumulating evidence that the ubiquitinationof transcriptional activators or core components of the transcriptionmachinery can serve to "license" their activities (42), andit is interesting to speculate that the E6/E6AP complex mayprovide such a function at the TERT promoter.

While E6AP was discovered during the characterization of E6-mediatedp53 inactivation, it is now known that mutations in the geneencoding E6AP (UBE3A) are the cause of Angelman syndrome (AS),a severe form of mental retardation (30, 39). Both alleles ofUBE3A are expressed in virtually all tissues and cell types,with the exception of subregions of the brain where only thematernal allele is expressed (1). AS patients lack expressionof a wild-type maternal UBE3A allele and therefore do not expressE6AP in these brain subregions, which include the hippocampalneurons and Purkinje cells. The lack of ubiquitination of oneor more substrates of E6AP in the brain is hypothesized to bethe cause of AS. The identity of the putative critical AS-relatedsubstrate(s) of E6AP in the brain is unknown, and only a fewnatural (HPV E6-independent) targets of E6AP are known for anycell type (33, 34). We were therefore hopeful that siRNA silencingof E6AP in conjunction with microarray expression analysis mightreveal pathways that are normally controlled or affected byE6AP independently of E6. However, at least with the cell typesexamined here, we did not observe any gene expression alterationsafter E6AP knockdown that were not also seen after E6 knockdown.In other words, none of the normal functions of E6AP appearto be reflected in the transcriptional profile. The transfectionof E6AP siRNAs into an HPV-negative cervical carcinoma cellline (C33-A) also did not reveal any significant alterationsin the gene expression profile of these cells (not shown). Thegeneral approach of utilizing transcriptional profiles to revealpathways affected by E6AP may yet prove useful if it is appliedto cells of a neuronal origin.

ACKNOWLEDGMENTS

This work was supported by a grant from the National CancerInstitute (CA72943 [J.M.H.]) and by a National Institutes ofHealth National Research Service Award (M.L.K.). Microarraysetup was supported by a grant from the National Institute ofAlcohol Abuse and Alcoholism (grant AA13518 of the INIA Project).

We thank Vishwanath Iyer, Jonathan Davies, and Patrick Killionfor providing cDNA arrays and assistance with computer analysis,Sylvie Beaudenon for a critical reading of the manuscript, andRichard Schlegel for valuable discussions and communicationsof unpublished results.

FOOTNOTES

* Corresponding author. Mailing address: Institute for Cellular and Molecular Biology, Section of Molecular Genetics and Microbiology, University of Texas at Austin, Austin, TX 78712. Phone: (512) 232-7700. Fax: (512) 232-3432. E-mail: huibreg{at}mail.utexas.edu.

Supplemental material for this article may be found at http://jvi.asm.org/.

Present address: Ambion, Inc., Austin, TX 78744-1832.

REFERENCES

Top