Ecdysoneless Protein Regulates Viral and Cellular mRNA Splicing to Promote Cervical Oncogenesis

Abstract

High-risk human papillomaviruses (HPV), exemplified by HPV16/18, are causally linked to human cancers of the anogenital tract, skin, and upper aerodigestive tract. Previously, we identified Ecdysoneless (ECD) protein, the human homolog of the Drosophila ecdysoneless gene, as a novel HPV16 E6–interacting protein. Here, we show that ECD, through its C-terminal region, selectively binds to high-risk but not to low-risk HPV E6 proteins. We demonstrate that ECD is overexpressed in cervical and head and neck squamous cell carcinoma (HNSCC) cell lines as well as in tumor tissues. Using The Cancer Genome Atlas dataset, we show that ECD mRNA overexpression predicts shorter survival in patients with cervical and HNSCC. We demonstrate that ECD knockdown in cervical cancer cell lines led to impaired oncogenic behavior, and ECD co-overexpression with E7 immortalized primary human keratinocytes. RNA-sequencing analyses of SiHa cells upon ECD knockdown showed to aberrations in E6/E7 RNA splicing, as well as RNA splicing of several HPV oncogenesis–linked cellular genes, including splicing of components of mRNA splicing machinery itself. Taken together, our results support a novel role of ECD in viral and cellular mRNA splicing to support HPV-driven oncogenesis.

Implications: This study links ECD overexpression to poor prognosis and shorter survival in HNSCC and cervical cancers and identifies a critical role of ECD in cervical oncogenesis through regulation of viral and cellular mRNA splicing.

Introduction

Human papillomaviruses (HPV) are double-stranded DNA tumor viruses causally linked to certain human cancers (1). Although the availability of a prophylactic vaccine against HPV16/18 is expected to reduce the incidence of HPV-driven cancers, these cancers still remain a major global health issue due to inadequate vaccine availability and affordability in most countries, and sociocultural barriers to vaccination against HPVs (2, 3). Thus, a better understanding of HPV-driven oncogenesis remains imperative.

The high-risk HPV E6 and E7 oncoproteins are expressed in all HPV-linked cancers (3) from a polycistronic transcript that gives rise to multiple alternatively spliced E6 mRNAs (for example E6*I, E6*II; ref. 4). Although some studies have shown that this splicing is essential for E7 protein translation and (4), other studies have shown that E7 oncoprotein is directly translated from the unspliced bicistronic E6/E7 RNA (5). The role of truncated E6*I protein in cellular transformation is still debatable, as some studies suggest its positive role in transformation (4, 6), whereas others suggest it functions as a tumor suppressor (as reviewed in ref. 5). Although splicing factors SRSF1 and hnRNPA1 are shown to alter E6 splicing (5), the full understanding of mechanisms that regulate the viral and host RNA splicing is still not fully understood.

Using the yeast two-hybrid system, we previously identified Ecdysoneless (ECD), the human homolog of Drosophila ecdysoneless gene as an HPV16 E6–binding protein (7). We then showed that ECD is essential for mammalian development, and it regulates cell-cycle progression through its interaction with retinoblastoma protein (RB), RUVBL1 and PIH1D1 proteins, two components of the R2TP cochaperone complex (8, 9). The R2TP complex, within the HSP90 cochaperone complex called PAQosome facilitates the assembly of ribonucleoproteins, that is, essential for the biogenesis of large protein–protein complexes (10). Drosophila and human ECD proteins bind to PRPF8, an essential scaffolding component of the spliceosome machinery (9, 11–13), suggesting a role of ECD in splicing. We recently showed that ECD is in a complex with proteins involved in mRNA export and splicing (12). Although the crystal structure of ECD is not known, our previous circular dichroism and small-angle X-ray scattering analyses suggested that ECD may function as a scaffold to organize key protein interaction or protein–RNA interactions (14).

ECD is overexpressed in breast (12, 15), pancreatic (16), and gastric cancers (17). We showed that co-overexpression of ECD and mutant RAS in immortal epithelial cells resulted in a fully tumorigenic phenotype (18), consistent with a co-oncogenic role for ECD.

Here, we provide evidence that ECD interacts with high-risk HPV E6 oncoproteins, is overexpressed in cervical and head and neck squamous cell carcinomas (HNSCC), with ECD overexpression an independent predictor of shorter survival in patients. Significantly, ECD knockdown (KD) in cervical cancer cells reduced oncogenic traits, its overexpression together with E7 led to immortalization of human foreskin keratinocytes (HFK). We show that ECD plays a role in the splicing of E6/E7 RNA as well as of cellular RNAs important for HPV-driven oncogenesis. Thus, we present a novel role of ECD in HPV-driven oncogenesis through its role in viral and cellular RNA splicing.

Materials and MethodsPlasmids and siRNAs

GST-ECD-1–644 (7), GST-E6AP-37–865, and GST-E6AP-del 391–408 (does not bind E6) were provided by Dr. Peter Howley (Harvard Medical School, Boston, MA; ref. 19). HPV16, 18, 11 or 6 E6 and E6 mutants in pSG5 vector, and Myc-tagged HPV16-E6 construct pEF-Myc-E6 have been described previously (7, 20, 21). The N-terminally FLAG-tagged full-length (1–644) and truncated versions of ECD (amino acids 1–291, 292–492 or 493–644) were generated by PCR amplification and cloning in pcDNA3.1 vector. For siRNA-based knockdown (KD), cells were transfected with 30 nmol/L of control or gene-specific siRNAs (ECD, PRPF8, and HPV16E6) using the DharmaFECT 1 Transfection Reagent (T-2001–03, Dharmacon). The details of siRNAs are included in the Supplementary Table S6.

Cell culture and media

Two independently derived primary HFK lines (cultured from discarded Foreskin samples obtained at Tufts-New England Medical Center when the principal investigator was on the faculty), and designated as 10HFK and 11HFK, were cultured in KGM medium (Keratinocyte Growth Medium BulletKit, Lonza). The TERT-immortalized keratinocyte cell line NTERT1 was kindly provided by Dr. James Rheinwald (Brigham and Women's Hospital, Boston, MA; ref. 22). The human immortalized keratinocyte cell line HaCaT (23) was maintained as described previously. HNSCC cell lines SCC1, SCC10B, SCC11, and HCC38 were obtained from Dr. Thomas Carey (University of Michigan, Ann Arbor, MI; ref. 24) and HPV+ HNSCC cell lines UPCI:SCC90, UPCI:SCC152, and UPCI:SCC154, were obtained from Dr. S. Gollin (University of Pittsburgh Cancer Institute, Pittsburgh, PA; ref. 25). These cell lines and 293T cells were maintained in DMEM with 10% FCS with growth factors as described previously in ref. (12). Human cervical carcinoma cell lines HeLa (HPV18-positive), SiHa (HPV16-positive), and CaSki (HPV16-positive) were obtained from the ATCC and grown at 37°C in a humidified atmosphere with 5% CO2 in α-MEM supplemented with 10% FCS and other usual supplements. All cell lines were Mycoplasma negative based on testing with Mycoplasma PCR Detection Kit (Sigma-Aldrich).

In vitro binding assays of ECD interaction with HPV E6 proteins

The 35S-labeled HPV E6 proteins for binding assays were generated using pSG5 plasmid (pSG5-HPV16 E6-FLAG) constructs as templates in wheat germ or rabbit reticulocyte lysate-based coupled in vitro transcription–translation systems (TNT Wheat Germ or Rabbit Reticulates Lysate System; Promega) in the presence of [35S] cysteine or [35S] methionine, as described previously (21). Aliquots of 35S-labeled proteins were incubated with 1 μg of GST or appropriate GST fusion proteins noncovalently bound to glutathione–Sepharose beads in 300 μL of lysis buffer (100 mmol/L Tris, pH 8.0; 100 mmol/L NaCl; 0.5% Nonidet P-40) for 2 hours at 4°C, and bound proteins were resolved by SDS-PAGE and visualized by fluorography.

Coimmunoprecipitation analysis of ECD association with HPV16 E6

293T cells were plated overnight at 106 per 100-mm dish and transfected with 0.5-μg pCR3.1- FLAG-ECD and 2.5-μg pEF-myc-HPV16E6 plasmids, alone or in combination, using the calcium phosphate method (7, 20, 21). After 48 hours, cell lysates were prepared in NP-40 lysis buffer (100 mmol/L Tris, pH 8.0; 100 mmol/L NaCl; 0.5% NP-40; 1 mmol/L phenylmethylsulfonyl fluoride). The lysates were precleared twice with protein G-agarose and equal aliquots of lysate protein (quantified using the BCA method) were used for immunoprecipitation with anti-FLAG mAb (M2 FLAG antibody; Sigma) followed by immunoblotting with anti-MYC (9E10) or anti-FLAG mAbs. Enhanced chemiluminescence (ECL Plus; Thermo Fisher Scientific) was used for signal detection.

Telomerase activity assay

Telomerase activity was measured using a nonradioactive modification of the TRAP assay (Telomeric Repeat Amplification Protocol), according to Herbert and colleagues (26) using the TRAPeze telomerase detection kit (cat # S7700; Millipore), replacing the [32P]-labeled TS oligo with a Cy5-conjugated TS oligo (5′-Cy5-AATCCGTCGAGCAGAGTT-3′.

IHC staining of cervical cancer tissue microarray

A tissue microarray (TMA) of formalin-fixed and paraffin-embedded normal/benign cervical tissues from eight subjects and 220 cases of cervical cancer tissues (each is 1.1 mm2) was purchased from Pantomics (catalog number: CXC2281; data on age, tumor type, TNM staging, and grading are provided by the manufacturer). IHC staining with an anti-ECD mAb, previously well-characterized for specificity and sensitivity, was carried out as described previously (15). The expression of ECD in invasive cervical tumor cells was evaluated on the basis of the semiquantitative histochemical score (H-score) that was obtained by multiplying the staining intensity (0 for negative, 1, 2 or 3 for weak, moderate and strong, respectively) by the percentage of stained cells, producing a total range of 0 to 300 (15). H-score ≥135 were considered as medium/high group and ≤135 as low/weak-expressing ECD. Scoring was completed for only samples harboring adequate invasive tumor area >15% of the core surface area.

TCGA database analysis

To explore the ECD mRNA expression in cervical and head and neck cancers, we obtained gene expression transcripts per million (TPM) values from the UCSC Xena database (27). The Cancer Genome Atlas (TCGA) Cervical Cancer set included 303 primary tumor samples, 2 metastatic tumor samples, and 3 normal tissue samples from TCGA (28). The TCGA head and neck cancers set included 520 primary tumor samples, 2 metastatic tumor samples, and 44 normal tissue samples. To investigate the statistical association of ECD expression and the outcomes, we obtained data for Cervical and HNSCC from the cBioPortal for Cancer Genomics. The TCGA cervical dataset comprised 294 cases whereas The TCGA HNSCC comprised 515 cases. RNA-sequencing (RNA-seq) expression level read counts were normalized using the Upper Quartile Fragments per Kilobase to transcript per Million mapped reads (FPKM-UQ) calculation and the ECD expression is represented as log base 2 in box plots compared with adjacent normal tissue. Dichotomization of ECD expression was done using X-tile software in relation to outcome. For TCGA cervical dataset, 593.04 was used as a cut-off point for high and low ECD mRNA expression, whereas for the TCGA head and neck dataset 649.5 was used as a cut-off point for high and low ECD mRNA expression.

Transfections/retroviral infections

The HPV16 E6 or E7 retrovirus–producing PA317 cell lines (provided by Dr. Denise Galloway, Fred Hutchinson Cancer Research Center, Seattle, WA) have been described previously (29). Retroviral supernatants for ECD expression were generated by transient transfection of pMSCV-hygro-ECD (30 μg) into the Platinum-GP (Plat-GP) packaging cell line (Cell BioLabs) and collection of supernatants after 48 hours. HFKs were incubated with retroviral supernatants in the presence of polybrene (10 μg/mL) for 48 hours and the transduced cells were selected in G418 (100 μg/mL for HPV16E6), puromycin (0.5 μg/mL, for HPV16E7) or hygromycin (5 μg/mL for ECD). HeLa and CaSKi cell lines were infected with retroviral supernatants coding for control shRNA or two distinct ECD shRNAs (12). The transduced cells were selected in 0.5 μg/mL of puromycin for 3 days, and expression of endogenous ECD was assessed by Western blotting of whole-cell lysates (8, 12).

RNA extraction and PCR

Total RNA was isolated using the TRizol reagent (Invitrogen) using standard protocols as described previously and cDNA was prepared (12). The qRT-PCR primers used in this study are indicated in Supplementary Table S7. For gene splicing validation, PCR was performed with specifically designed primers for splice isoform detection and products was run on 2% agarose gel.

Western blotting

Western blotting was performed using standard protocols, as described previously (8, 9, 12). Signals were detected using ECL Plus (Thermo Fisher Scientific). Details of antibodies used in this study are indicated in the Supplementary Table S8.

CellTiter-Glo luminescent cell viability assay

2,000 cells per well were plated in six replicates in 96-well plates. Culture medium was changed every two days. Viable cells were quantified at days 0, 2, 4, 8, 10, and 12, using the CellTiter- Glo luminescent cell viability assay (Promega), as done previously (12).

Transwell migration, invasion, and soft agar colony formation assays

HeLa or SiHa cells transfected with scrambled or two distinct siRNA/shRNA against ECD were analyzed for migration and invasion using Transwell chambers (Matrigel-coated for invasion), and for anchorage-independent colony formation assay using soft agar, as previously published (12, 18).

RNA-seq analysis

SiHa cells were transfected with Control siRNA or siRNA against ECD or PRPF8, total RNA was isolated, RNA purity was analyzed on a Bioanalyzer, and RNA-seq libraries were generated using the TruSeq Stranded RNA Library Preparation Kit (Illumina), per the manufacturer's recommended protocol. The libraries were subjected to 75 bp paired-end sequencing on an Illumina NextSeq 500 instrument to generate approximately 30 to 35 M pairs of reads per sample in each of three biological replicates. For differential expression analysis, paired-end reads were aligned to the human genome version hg38 using STAR 2.7.3a guided by Ensembl gene annotations (30), and annotated transcripts were quantified and TPM normalized using Stringtie 2.1.1 (31). Differential expression was assessed by DESeq2 (32) and significantly changed genes were required to have a Benjamini–Hochberg adjusted P value of <0.05 and at least a 2-fold change in expression compared with control samples. Gene Ontology (GO) and Gene Set Enrichment Analysis (GSEA) were performed using WebGestalt (33), whereas P-HIPSTer pathogen-host predictions were obtained via Enrichr (34).

Analysis of mRNA splicing

Spliced Transcripts Alignment to a Reference (STAR) RNA-seq mapper was used for spliced RNA sequence alignment, and splicing was visualized in Sashimi plots with IGV (35). Differential splicing was determined by rMATS (36) 4.0.2 at an FDR < 0.05, with novel splice site detection enabled. To estimate splicing of viral genes, a chimeric sequence database consisting of human genome hg38 and the viral genome HPV16 with annotation from NCBI was used to generate the STAR index, with default parameters. After mapping, reads that aligned to the HPV16 genome were kept and visualized in Sashimi plots with IGV or quantified and normalized by Stringtie. HPV16 intronic reads representing full-length E6 RNA were isolated by removing any reads that overlapped exons.

Statistical analysis

Data analysis was performed using SAS version 9.4 (SAS Institute) and GraphPad prism. SPSS 24.0 statistical software (SPSS Inc.) was used for statistical analysis of ECD mRNA expression. Survival curves were analyzed by Kaplan–Meier analysis with the log-rank test. Cox proportional hazard models were used for multivariate analyses, adjusting for potential confounding influence of pathological covariates as variables on associations with ECD expression. The χ2 test was carried out for inter-relationships between categorical variables. Comparisons between two groups were made using t tests for continuous outcomes. Comparisons among at least three treatment groups were made using one-way ANOVA for continuous outcomes. If the overall tests yielded significant results, post hoc tests with the Tukey method for multiple comparisons were conducted. Difference between two groups was assessed using Welch t test (P < 0.05 was considered statistically significant).

Data availability

RNA-seq reads and processed files are available from NCBI's Gene Expression Omnibus (GEO) under the accession GSE174317.

ResultsECD binds to high-risk HPV E6 oncoproteins

We previously identified ECD in a yeast two-hybrid screen of a human epithelial cell library with HPV16 E6 as a bait (7). Here, we carried out in vitro binding analyses between full-length ECD fused to the C-terminus of GST (GST-ECD) and in vitro-translated, 35S-labeled HPV E6 proteins. GST-E6AP, used as a positive control (19), showed substantial binding to high-risk E6 proteins (HPV16 or 18) but little binding to low-risk E6 proteins (HPV6 or 11; Fig. 1A). Importantly, GST-ECD showed substantial binding to HPV16 E6 and modest binding to HPV18 E6, whereas no binding was seen with low-risk HPV E6 proteins (Fig. 1A). Given the low levels of E6 proteins known to be expressed in available immortalized epithelial cells or cervical cancer cell lines, we performed transient transfection of FLAG-ECD and/or Myc-HPV16 E6 in 293T cells and the lysates were subjected to anti-FLAG IP followed by anti-Myc WB. Anti-Myc (Fig. 1B, top, lanes 1–3) and anti-FLAG WB (Fig. 1B, bottom, lanes 1–3) demonstrated the expression of E6 and ECD proteins. Notably, E6 was detected in anti-FLAG IPs of cells cotransfected with ECD together with E6 (Fig. 1B, lane 6, top). These results established that E6 and ECD can associate in mammalian cells.

Figure 1.Figure 1.Figure 1.

Interaction between HPV E6 Proteins and ECD. A, 35S-labeled E6 proteins were incubated with GST, GST-ECD, GST-E6AP, or GST-E6APmut and binding assay was done. B, 293T cells were transfected with FLAG-ECD or myc-HPV16 E6 alone or in combination and cell lysates were subjected to IP with anti-FLAG antibody and resolved by SDS-PAGE (right). 40 μg aliquots of lysates (left) were subjected to IB with anti-Myc (top) and anti-FLAG (bottom left). C, Schematic representation of deletion mutants. D, 35S-labeled ECD FL or mutant proteins were incubated with GST or GST-E6 and binding experiments were performed. E, 293T cells were transfected with FL-ECD or indicated mutants with or without Myc-16E6 or Myc-16E6 alone. Cell lysates were subjected to IP with anti-FLAG antibody and WB with anti-MYC or anti-FLAG antibodies.

The C-terminal part of ECD interacts with E6

Next, similar binding experiments were carried out with full-length ECD or its fragments (aa 1–291), middle (aa 292–492) and C-terminus (aa 493–644; Fig. 1C). Notably, only the C-terminal fragment showed binding (Fig. 1D). Furthermore, 293T cells transiently transfected with FLAG-tagged FL-ECD or deletion constructs and Myc-E6, followed by IP/WB showed that FL and C-terminal fragment of ECD can associate with E6 (Fig. 1E top right, lanes 14 and 17). These results establish that the C-terminal region of ECD mediates E6-ECD association in vitro as well as in mammalian cells.

HPV16 E6 protein does not target ECD protein for degradation

E6 is known to target several of its associated proteins, such as p53, for degradation through its association with E6AP (19). Next, we transfected SiHa cells with scrambled siRNA or HPV16 E6-specific siRNA (sequences provided in Supplementary Table S6) followed by WB and qRT-PCR (Supplementary Tables S7 and S8). As expected, the levels of p53 and its target protein p21 increased upon E6 KD (Supplementary Fig. S1B). In contrast, ECD mRNA (Supplementary Fig. S1A) or protein (Supplementary Fig. S1B) levels remained unaltered upon E6 KD. These results demonstrate that E6 does not target ECD for degradation.

ECD is overexpressed in cervical cancer and its overexpression predicts shorter patient survival

Next, we investigated the expression of ECD in primary HFK cell lines and their E6/E7 or TERT-immortalized derivatives, as well as in cervical and HNSCC cell lines. Compared with a relatively low ECD expression in primary HFKs and HaCaT, an immortalized cell line (Fig. 2A, lanes 1–3; Fig. 2B lanes 1–4), substantially higher levels of ECD expression was seen in E6+E7-HFKs (Fig. 2A, lanes 4–6), and in hTERT-HFKs (Fig. 2A, lane 7). Higher ECD expression was also seen in cervical cancer cell lines HeLa and SiHa (Fig. 2A, lanes 8 and 9), and in both HPV+ and HPV− HNSCC cell lines (SCC, Fig. 2B, lanes 5–11). Similar to protein, ECD mRNA expression was higher in immortal and cancer cell lines as compared with normal mortal HFKs (Supplementary Fig. 2SA).

Figure 2.Figure 2.Figure 2.

ECD expression in cervical cancers and HNSCC lines (A and B) and cervical tissues (C). A and B, WB of cell lysates from primary HFKs (lanes 1–3), immortalized HFKs by E6+E7 (lanes 4–6) or by TERT (lane 7) and cervical cancer cell lines (lanes 8 and 9), immortal keratinocytes HaCaT (B, lane 4), and HNSCC lines (B, lanes 5–11). C, Representative IHC staining of cervical tissue array for ECD expression in normal (i, ii); adenocarcinoma (iii, iv); adenosquamous carcinoma (v, vi); squamous cell carcinoma (vii and viii). Images, 100× magnification (inset 400×). D, IHC was scored using semi-quantitative H-score and histogram were plotted on the basis of the expression of ECD in different histopathologic types of cervical cancer. The number in the histogram represents number of patients included in the study. E, Kaplan–Meier survival analysis from TCGA data reveal ECD mRNA overexpression correlates with short patient OS (F) as well as disease-specific survival.

Next, we analyzed the expression of ECD in a commercially available cervical cancer TMA (PANTOMICS; CXC2281), using IHC with a well-characterized anti-ECD antibody (15, 16). Although most tumor specimens showed moderate to high ECD staining, little or no staining was observed in normal cervix tissue specimens (representative pictures with strong staining of score 3 are shown in Fig. 2C). A frequency distribution histogram of ECD expression revealed the median H-Score to be 135 and staining below 135 was considered low/weak and 135 to 300 was considered moderate to strong. Both nuclear and cytoplasmic ECD staining were observed (Fig. 2D; Supplementary Table S1), consistent with the presence of ECD in both nucleus and cytoplasm in previous studies (15, 16, 37).

Kaplan–Meier analyses of publicly available TCGA data showed that higher expression of ECD mRNA is associated with shorter overall survival (OS; p = 0.024) and disease-specific (p = 0.033) survival in patients with cervical cancer (Fig. 2E and F). Multivariate cox regression analysis with a model incorporating available clinical parameters, such as Federation of Gynecology and Obstetrics (FIGO for tumor staging and age), showed the ECD overexpression to be an independent predictor of poor patient outcomes, including shorter OS (p = 0.002), disease-free survival (p = 0.005), and progression-free survival (PFS; p = 0.05; Supplementary Tables S2–S4).

Similarly, TCGA Kaplan–Meier analyses in patients with HNSCC showed that ECD overexpression predicts poor clinical outcomes, shorter OS (P = 0.003), disease-specific survival (p = 0.008), and PFS (p = 0.013; Supplementary Fig. S2B and S2C). Multivariate Cox regression analysis showed that ECD overexpression was an independent predictor of poor patient outcome, including shorter OS (p = 0.007), disease-free survival (p = 0.013), and PFS (p = 0.029; Supplementary Tables S2–S4).

Furthermore, although ECD expression showed a trend of slightly higher levels in HPV+ versus the HPV− samples, it was not statistically significant (Supplementary Fig. S3). Overall, our analyses demonstrate that ECD overexpression is associated with shorter survival in cervical patients, as well as patients with HNSCC.

ECD is required to sustain oncogenic traits of cervical cancer cells and ECD cooperates with E7 to immortalize HFKs

Next, we expressed control or ECD shRNAs in HeLa (Fig. 3) and CaSki cells or siRNAs (two independent) in HeLa (Fig. 3), and SiHa (Supplementary Fig. S4) cell lines and then assessed the impact on oncogenic traits. The KD efficiency was confirmed by WB (Fig. 3A; Supplementary Fig. S4A). ECD KD led to a significant impairment in the proliferation rates of HeLa and CaSki cell lines (Fig. 3B; Supplementary Fig. S4B) as well as decreased anchorage-independent growth (Fig. 3C; Supplementary Fig. S4C). ECD KD also resulted in significant reduction in cell migration (Fig. 3D and E; Supplementary Fig. S4D and S4F) and invasion (Fig. 3F; Supplementary Fig. S4E) of both HeLa and SiHa cell lines. Taken together, our results establish a significant role of ECD in sustaining the oncogenic traits of cervical cancer cells.

Figure 3.Figure 3.Figure 3.

ECD KD decreases oncogenic traits of HeLa cell line. Scrambled or two different ECD shRNAs (#1 and #2′ A–C,) or control or siRNA#1 or #2 expressing cells (D–F) were WB (A and D), and then analyzed for cell proliferation (B). C, Anchorage dependence assay followed by staining of colonies with crystal violet, counted, and plotted as histograms cells. Cells were plated on Boyden chambers for assessing their ability to migrate (E) or invade (F). Twenty-four hours later, migrated and invaded cells were stained with propidium iodide, counted, and plotted as histograms. ANOVA analysis indicated that mean number of colonies of control versus ECD KD cells were statistically different (all Tukey-adjusted P < 0.001). *, P ≤ 0.05; **, P ≤ 0.01, ***, P ≤ 0.001; and ns, nonsignificant. Mean ± SE was derived from three independent experiments, each done in triplicates. G and H, ECD cooperates with E7 to immortalize HFKs. G, WB of cell lysates with indicated antibodies; β-actin was used as a loading control. H, Cumulative population doublings (PD) is plotted over times from two independent experiments (Ex-I and Ex-II). I and J, qRT-PCR using specific primers of indicated genes from RNA samples of indicated cells over increasing passages (P). β-Actin was used as an internal control. Fold change with respect to control after normalizing with β-actin was calculated and plotted. Each experiment was repeated three times.

Next, we assessed the role of ECD in keratinocyte immortalization using two independent HFKs, primary cells transduced with retroviruses expressing vector, E6, E7, ECD, E6+ECD, E7+ECD or E6+E7 (the last one served as a positive control) and cultured the cells with G418 and/or puromycin in appropriate selection medium. Expression of transduced genes was confirmed by WB (Fig. 3G). As expected, introduction of the vector, E6 alone or E7 alone did not extend the proliferative lifespan of HFKs in two independent experiments, with cells exhibiting senescence (data not shown). Similarly, overexpression of ECD alone or ECD+E6 did not extend the lifespan. However, E6+E7-HFKs (used as positive control) and ECD+E7-HFKs continued to proliferate when passaged at a 1:10 split ratio achieving more than 200 population doublings (PD) before freezing (Fig. 3H; Supplementary Fig. S5A). Morphologically, ECD+E7-HFKs were comparable with E6+E7-HFKs, with epithelial and cuboidal cells of relatively homogeneous shape, that grew in tightly packed colonies (Supplementary Fig. S5B).

Notably, ECD+E7-HFKs showed higher levels of p53 (Fig. 3G; Supplementary Fig. S6A), which is consistent with our previous findings (7). Adriamycin treatment to induce the DNA damage (7) showed a time-dependent increase in the levels of p53 and p21, confirming that ECD+E7-HFKs immortalization does not require the loss of p53 function (Supplementary Fig. S6A). However, both ECD+E7-HFKs lost p16 expression (22) with increasing passages (Supplementary Fig. S6B), and this was associated with enhanced cell proliferation (Supplementary Fig. S6C and S6D) and for colony-forming efficiency (Supplementary Fig. S6E and S6F).

The levels and activity of telomerase reverse transcriptase (TERT) complex components (38) were markedly elevated in ECD+E7-HFKs (Fig. 3I), TERT mRNA progressively increased with passages (Fig. 3J), with elevated telomerase activity (Supplementary Fig. S7, lanes 5 & 7). Further analysis of ECD+E7-HFKs showed these cells were unable to form soft agar colonies, and lacked migration or invasion abilities, indicating that ECD overexpression together with E7 does not induce full transformation of HFK cells (data not shown). Taken together, ECD can cooperate with HPV E7 to induce the immortalization of keratinocytes.

ECD regulates E6/E7 RNA splicing

Our recent proteomics analyses identified several components of mRNA splicing to be in complex with ECD (12), suggesting a role of ECD in RNA splicing. Next, we carried out RNA-seq from SiHa cells expressing control or siRNAs against ECD or PRPF8 (used as positive control) followed by bioinformatics analyses (30–36). By aligning reads to both the human and HPV16 genomes, we were able to examine the effects of ECD KD on cellular and viral transcripts. Next, mapping of the reads to a human HPV16 genome was performed (Supplementary Fig. S8A). The KD efficiencies were further measured by RNAseq (Supplementary Fig. S8B). Our analysis revealed that ECD KD leads to higher E6*I (123 compared with 81 in control) and E6*II (30 compared with 10 in control) (Fig. 4A; Supplementary Table S5). In contrast PRPF8 KD decreases both E6*I (43 vs. 81) and E6*II (5 vs. 10; Fig. 4A; Supplementary Table S5). Although KD of ECD and PRPF8 appeared to decrease the expression of intact E6 (intron-containing) RNA, there was no difference among the three groups (Fig. 4B; Supplementary Table S5). RT-PCR confirmed the increase in the levels of 16E6*I and 16E6*II expression upon ECD KD (Fig. 4C and D) and decrease by PRPF8 KD (Fig. 4E and F).

Figure 4.Figure 4.Figure 4.

ECD regulates E6 intron splicing. A, IGV visualization of HPV E6/E7 reads-distribution along with the HPV reference genome (left) and number of the averaged splice junction reads of E6*I and E6*II (right) in SiHa cells upon control or KD of ECD or PRPF8 (source data Supplementary Table S5). B, Number of the averaged E6 reads from the E6 intron region upon control, ECD KD or PRPF8 KD (source data Supplementary Table S5). SD of triplicate samples. C and E, SiHa cells were treated with control or siRNA against ECD or PRPF8, followed by RT-PCR analyses, and endpoint PCR products were run on 2% agarose gel. Representative images are shown. GAPDH was used as an internal control. Quantification of gel bands were determined by using ImageJ software (D and F) and WB of SiHa cell lysates (G–I) from control or KD of ECD, PRPF8, or DDX39A. GAPDH was used as a loading control.

Given the controversy that E6*I ORF is essential or not for E7 protein translation (4, 5), we analyzed E7 and retinoblastoma protein (RB) levels in SiHa cells after KD with control, ECD or PRPF8. KD of DDX39A, another ECD binding protein (12) was used as a negative control (Fig. 4GI). Notably, ECD KD resulted in decreased E7 and consequently increased retinoblastoma (as E7 is known to decrease retinoblastoma protein; ref. 39). Furthermore, while PRPF8 and ECD KD had opposite effects on E6/E7 splicing, KD of both proteins showed similar decrease in E7 protein. KD of DDX39A did not affect E7 or retinoblastoma proteins. These results suggest that ECD KD may regulate E7 protein levels independent of its effect on E6/E7 RNA splicing.

ECD regulates splicing of key genes in oncogenic pathways as well as of classical splicing factors

Further analyses showed that while PRPF8 KD altered a higher number of differential splicing genes (DSG; Fig. 5A), a significant number of DSGs in ECD KD cells overlapped with DSGs in PRPF8 KD cells (Fig. 5B). However, nearly 43% of DSGs in ECD KD showed a normal splicing pattern in PPRF8 KD cells (Fig. 5B), suggesting that ECD may function in splicing partially but not completely through PRPF8 (40). ECD KD cells exhibited a marked increase in exon inclusion/retention (38.2%), exon skipping (33.8%), and intron skipping (22.6%) with lower level of intron retention (5.4%) (Fig. 5C). Intriguingly, the highest overlap between siECD and siPRPF8 were on retained introns (Fig. 5D). Taken together, our results demonstrate a novel role of mammalian ECD in regulation of RNA splicing.

Figure 5.Figure 5.Figure 5.

ECD regulates RNA splicing. A, Differential splicing events as determined by rMATS after KD of ECD or PRPF8 as compared with control. B, Overlap of DSGs in cells with KD of ECD or PRPF8. C, Percentage of exon/intron splicing defects in KD of ECD or PRPF8. D, Heat map shows concordance of splicing defects between KD of ECD and PRPF8. Colors represent fraction of genes in each ECD KD category and the overlap with each splicing category in PRPF8 KD. E–G, RT-PCR followed by endpoint PCR products run on 2% agarose gel electrophoresis in SiHa cells upon control or ECD KD. Quantification using ImageJ software. The values presented are normalized with actin (used as control). Sashimi plots were generated using RNA-seq datasets.

Ingenuity pathway analyses (IPA) of DSGs seen upon ECD KD identified alterations in mRNAs of genes involved in several oncogenic pathways, including EIF2, mTOR, and PI3K/AKT signaling pathways (Supplementary Fig. S9). We validated some of these genes using specific primers (Fig. 5F; Supplementary Table S7) from mTOR pathway as it is one of the key pathways activated in HPV+ oncogenesis (41). These included RAC1, which is involved in autocrine STAT3 activation in HPV+ cervical cancer through a virus-driven Rac1-NFκB–IL6 signaling axis (42); RPS6KB1, which is reported to be amplified and spliced in various cancers (43). ECD KD also induced differential splicing of a major tumor suppressor gene TSC2, that serves as a negative regulator of AKT–mTOR signaling (44). HPV16 E6 was shown to interact with TSC2 (45). Similarly, ECD KD altered splicing of STAG1 (46) and EIF4A2 (47), genes associated with various cancers.

Significantly, ECD regulated alternative RNA splicing of four classical splicing factors (Fig. 5G), including SRSF2, a known proto-oncogene to regulate cell proliferation and oncogenesis (48) and SRSF3 (SRp20), gene highly expressed in cervical cancer and is essential for cell growth and transformation maintenance and it regulates HPV RNA splicing (49).

ECD KD leads to broad changes in cellular gene expression, but only a small proportion of these are due to aberrant splicing

Concurrent with analysis of differential splicing, we also assessed the impact of ECD KD on cellular gene expression. These analyses showed that ECD KD resulted in 2,839 differentially expressed genes (DEG) relative to control siRNA, a number considerably higher than seen with the KD of PRPF8 (1,089 DEGs; Fig. 6A). GO analysis and GSEAs of DEGs showed significant enrichment of pathways related to IFN signaling and viral response (Fig. 6B and C). IPA of DEGs further identified pathways involved in oncogenesis (Supplementary Fig. S10).

留言 (0)

沒有登入
gif