HPV 16 E7 alters translesion synthesis signaling

The hierarchically clustered heatmap was plotted using the cluster map function in Python Seaborn module. The row order (cancer) was based on the color bar indicating the increasing rate of TLS genes for each cancer type. The columns (TLS genes) were permuted using the hierarchical clustering method (Weighted Pair Group Method with Arithmetic Mean) to get a structured gene pattern with the dendrogram on the top (https://​docs.​scipy.​org/​doc/​scipy/​reference/​generated/​scipy.​cluster.​hierarchy.​linkage.​html).

Web-based software on www.​cbioportal.​org was used to perform the Kaplan Meyer, gene expression, promoter methylation, and copy number alteration analyses of data in TCGA database. Gene ontology analysis was performed using Gene Ontology enRIchment anaLysis and visuaLizAtion tool (GOrilla) on RNAseq data from the cancer cell line encyclopedia and TCGA databases [25‐28]. The 50 cancers examined in Additional file 1: Fig. S1B are Adrenocortical Carcinoma, Hepatobiliary Cancer, Bladder Urothelial Carcinoma, Mucinous Adenocarcinoma of the Colon and Rectum, Rectal Adenocarcinoma, Colon Adenocarcinoma, Breast Cancer, Anaplastic Astrocytoma, Anaplastic Oligoastrocytoma, Oligoastrocytoma, Astrocytoma, Oligodendroglioma, Glioblastoma Multiforme, Pheochromocytoma, Miscellaneous Neuroepithelial Tumor, Cervical Cancer, Esophagogastric Cancer, Tubular Stomach Adenocarcinoma, Stomach Adenocarcinoma, Diffuse Type Stomach Adenocarcinoma, Mucinous Stomach Adenocarcinoma, Signet Ring Cell Carcinoma of the Stomach, Uveal Melanoma, Head and Neck Squamous Cell Carcinoma, Renal Clear Cell Carcinoma, Chromophobe Renal Cell Carcinoma, Papillary Renal Cell Carcinoma, Hepatocellular Carcinoma, Lung Adenocarcinoma, Lung Squamous Cell Carcinoma, Non-Hodgkin Lymphoma, Acute Myeloid Leukemia, Serous Ovarian Cancer, Pancreatic Cancer, Pleural Mesothelioma, Biphasic Type, Pleural Mesothelioma, Epithelioid Type, Prostate Adenocarcinoma, Melanoma, Cutaneous Melanoma, Soft Tissue Sarcoma, Seminoma, Non-Seminomatous Germ Cell Tumor, Embryonal Carcinoma, Thymoma, Follicular Thyroid Cancer, Papillary Thyroid Cancer, Endometrial Carcinoma, Uterine Endometrioid Carcinoma, Uterine Serous Carcinoma/Papillary Serous Carcinoma, and Uterine Carcinosarcoma/Malignant Mixed Mullerian.

HeLa (ATCC^® CCL-2™) and SiHa (ATCC^® HTB-35™) cells were grown in Dulbecco modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and penicillin–streptomycin. Primary Keratinocytes (HFKs) were derived in-house from anonymized medical waste. HFKs were grown in Keratinocyte Growth Medium 2 (PromoCell, Heidelberg, Germany, C-20011), supplemented with penicillin–streptomycin, calcium (0.5 M, 60 μl in 500 ml medium), and SupplementMix (PromoCell, Heidelberg, Germany). Virus for retroviral transduction of E7 and E7ΔDLYC into HFKs was produced in 293 T cells grown in DMEM and retroviral transduction was carried out as previously described [29]. All cell lines were grown at 37C and 5% CO₂.

Lysates were generated and immunoblots were run as previously described [30]. The membranes were then probed using the following antibodies: p-ATM (Cell Signaling Technologies, catalog no. 13050S), ATM (Cell Signaling Technologies, catalog no. 92356S), p-ATR (Cell Signaling Technologies, catalog no. 30632S), ATR (Cell Signaling Technologies, catalog no. 2790S), p-CHK1 (Cell Signaling Technologies, catalog no. 2348S), CHK1 (Cell Signaling Technologies, catalog no. 2360S), ub-PCNA (Cell Signaling Technologies, catalog no. 13439S), p-RPA32 (Cell Signaling Technologies, catalog no. 54762S), RPA32 (Cell Signaling Technologies, catalog no. 52445S), TopBP1 (Santa Cruz Biotechnology, catalog no. sc-271043), POLη (Santa Cruz Biotechnology, catalog no., sc-17770), GAPDH (Santa Cruz Biotechnology, catalog no. sc-47724), Nucleolin (Santa Cruz Biotechnology, catalog no. C23-HRP, sc8031) and Rad6 (Abcam, catalog no. ab31917). After incubation with the corresponding horseradish peroxidase (HRP)-conjugated secondary antibody, cells were visualized using SuperSignal West Femto maximum sensitivity substrate (Thermo Scientific).

HeLa and SiHa cells were seeded at 100,000 cells/well on 6-well plates. 24 h after seeding, the DMEM medium was aspirated and replaced by DMEM media supplemented with 10 mg/L nucleoside stock solution (Biological Industries Israel Beit-Haemek Ltd. 01-343-1D). Cells were grown for 72 h before cell lysates were harvested.

HFKs were seeded on 6-well plates at a density of 100,000 cells/well. After 24 h, the keratinocyte growth medium was aspirated and cells were treated with 0.3 mM hydroxyurea (Thermo Fisher Scientific A10831) or 10 μM cisplatin (Sigma Aldrich 479,306-1G) in keratinocyte growth medium. For the UV treatment, cells were exposed to 5 mJ/cm² UV-C (UV Stratalinker 2400, Stratagene) before fresh keratinocyte growth medium was reapplied. After 4 h cell lysates were harvested.

We have previously reported that the expression of TLS genes is often increased in cervical cancers [19]. To understand the extent that this increased expression occurred more or less frequently than in other cancers, we compared TLS gene expression across 16 tumor types found in the TCGA database [31]. This showed that TLS gene expression more than two standard deviations above the mean (z score > 2) was common in tumors of all types, with an average of 62% of tumors having increased expression of one or more TLS genes (Fig. 1A, Additional file 1: S1A). We also found that increased expression of at least one TLS gene was more common in cervical cancers (77%) than all other tumor types except those occurring in the adrenal gland (85%). Of TLS genes, increased expression of SPRTN, DTL, POLD1, PCNA, and VCP occurred most often in cervical cancers. ZBTB1, REV1, POLI, POLK, and REV3L were least commonly increased. Notably, four of the five genes that are least likely to have elevated expression were TLS polymerases. This is consistent with our prior work demonstrating that HPV16 E6 blocks the induction of POLη in response to replication stress [19]. An analysis of TLS gene expression in 50 different cancer types found a clear correlation between TLS gene expression overall and TLS polymerase expression (Additional file 1: Fig. S1B). In this more expansive analysis, elevated TLS gene expression was similarly more common in cervical cancers (6th out of 50) compared to other cancer types.

Of the two HPV16 oncogenes, HPV16 E7 is more closely linked with replication stress [32]. HPV16 E7 increases replication stress by destabilizing RB-E2F1 complexes and thus dysregulating S-phase entry via E2F1-responsive gene expression [33, 34]. Other RB family members (p130 and p107) are destabilized by HPV E7, which also facilitates S-Phase entry [35], p. 130). To understand the extent that the changes in TLS gene expression could be attributed to increased E2F1 availability, we performed an in silico screen of RNAseq data from 1020 cell lines in the cancer cell line encyclopedia [25]. These cell lines were segregated based on E2F1 expression, allowing us to compare TLS gene expression between the cell lines with and without elevated E2F1 expression. 37 cell lines had elevated E2F1 expression (z-score > 2). This cohort of cell lines was designated “high E2F1 expressing” and compared to the other cell lines to identify differentially expressed genes. We performed gene ontology analysis (via Gene Ontology enRIchment anaLysis and visuaLizAtion tool or GOrilla) on these differentially expressed genes and the CaCx transcriptome after ranking genes by their magnitude of expression (Additional file 1: Fig. S1C). The enriched cellular processes were similar in both cases and included TLS (Fig. 1B). Further analysis found a positive correlation of TLS gene expression with E2F1 expression when the cell lines from the cancer cell line encyclopedia. There was elevated expression of most of these same genes in cervical cancers (Fig. 1C). Because HPV16 E7 expression increases as a function of cervical cancer progression, we determined how the expression of these genes differed in cervical tissues with no evidence of disease, premalignant lesions, and in Stage I-III cervical cancer by interrogating a previously published dataset that we generated by normalizing and combining RNAseq data from multiple publicly available datasets (GSE145976) [19]. The expression of most TLS genes, with the notable exception of TLS polymerases, increased as a function of CaCx progression (Fig. 1C).

We then defined the extent that the increased expression of TLS genes could be attributed to established determinants of gene expression by analyzing the TCGA database. As expected, copy number alterations showed strong positive with TLS gene expression in cervical cancers. Similarly, promoter methylation negatively correlated with TLS gene expression in these tumors (Fig. 2A). In vitro studies have shown that HPV16 E7 induces two demethylases (KDM6A and KDM6B) resulting in global changes in gene expression [23]. This led us to hypothesize that expression of KDM6A and KDM6B would positively correlate with TLS gene expression. While there were strong correlations between some TLS genes and KDM6A or KDM6B expression, there was an approximately equal frequency of positive and negative correlations (Fig. 2A).

It has also been demonstrated in vitro that HPV16 E7 expression causes increased RRM2 expression. Because RRM2 facilitates de novo synthesis of nucleotides, the increased expression of the protein is believed to help cells tolerate the replication stress caused by HPV16 E7 [20]. As TLS responds to replication stress, we hypothesized that increased RRM2 expression would negatively correlate with TLS gene expression. However, our analysis of the TCGA data did not support this hypothesis (Fig. 2A). RRM2 expression rarely correlated with TLS gene expression in a statistically significant manner. Further, when RRM2 expression significantly correlated with the expression of a TLS gene, it was roughly equally likely to be a positive or a negative correlation. These analyses suggest that KDM6A, KDM6B, and RRM2 help cervical cancer cells tolerate HPV E7-associated replication stress independently of TLS gene expression, highlighting the substantial investment that cells commit to mitigating this stress.

We next determined how often the increased expression of a TLS gene could be attributed to an increased copy number of that gene. This analysis demonstrated that increases in copy number occurred 60% of the times that there was increased expression of the corresponding TLS gene (Fig. 2B). Despite the high overall frequency, there was considerable variation in the influence of copy number alterations on TLS gene expression. Nearly all occurrences of high SPRTN and DTL expression occur in tumors with copy number increases, while the increases in NEXMIF expression rarely occur in tumors that have increased NEXMIF copy number. This suggests that increases in copy number are a major driver of elevated TLS gene expression overall, but that the contribution of copy number increases on TLS gene expression varies among TLS genes. As protomer methylation is a negative correlate of TLS gene expression, we also determined how frequently high promoter methylation (defined as 90–100% of maximal methylation for the gene of interest) occurred in highly expressed TLS genes. As expected, this was rare. High methylation occurred only 0.4% of the times that TLS genes are highly expressed.

We continued to evaluate the expression of TLS genes using two cervical cancer cell lines (HeLa and SiHa). Our published data demonstrate that these cell lines have increased TLS abundance [19]. Combined with our in silico data, this suggests cervical cancer cells utilize the TLS pathway to address replication stress. Thus, we hypothesized that TLS activity could be reduced by lowering replication stress. To test this hypothesis, we grew cervical cancer cells in media with and without additional nucleosides. The addition of nucleosides should remedy replication stress stemming from depleted nucleoside pools and reduce the need for TLS. We harvested whole cell lysates from these cells and conducted immunoblot analysis. ATR is activated via phosphorylation (Thr1989) in response to replication stress [21]. PCNA ubiquitination (Lys 164) is a hallmark of TLS activity [36]. Therefore, antibodies were used to detect total and phosphorylated ATR as well as total and ubiquitinated PCNA or ub-PCNA. We found that the addition of nucleosides to HeLa and SiHa growth media reduced ATR phosphorylation and ub-PCNA abundance (Fig. 3A). Further analysis found that supplemental nucleosides were also able to reduce the abundance of another marker of replication stress (TopBP1) and multiple TLS proteins (POLκ, POLη, RAD18, and RAD6) (Fig. 3B).

Together these data suggest that HPV16 E7-associated replication stress leads to increased TLS protein abundance. To test this, we used primary neonatal human foreskin keratinocytes (HFKs) because HPV infections naturally occur in keratinocytes. A lentiviral system allowed us to express HPV16 E7 at physiological levels and compare changes induced by HPV16 E7 to vector only (LXSN) HFKs. This system is widely used to study HPV oncogene biology as it drives HPV oncogene expression at levels accepted to be below those seen in cell lines derived from cervical cancers, where super enhancer elements drive increased oncogene expression [37]. We confirmed the expression of HPV16 E7 in these cells using RB levels as a surrogate marker of HPV16 E7 (Supplemental Fig. 2). Immunoblot analysis demonstrated that HPV16 E7 HFKs had the expected reduction in RB protein levels. Consistent with reports from other groups that high-risk HPV oncogenes increase replication stress markers [13, 20, 38‐40], p. 5, [22, 41], we found that HPV16 E7 increased the abundance of replication stress (TopBP1, p-RPA32, p-ATR, p-CHK1) and DSB repair (p-ATM) proteins (Fig. 4). HPV16 E7 also increased TLS activity, as indicated by an increase in ub-PCNA abundance. Further, HPV16 E7 increased the amount of POLη protein that we detected (Fig. 4). The ability of HPV16 E7 to induce replication stress has been linked with its ability to bind and destabilize RB-E2F1 complexes. This binding can be abolished by deleting the residues of E7 that facilitate the interaction with RB (HPV16 E7Δ21-24^DLYC) [42]. We will refer to this mutant as HPV16 E7ΔDLYC. To determine the extent that HPV16 E7 increased TLS activity and protein abundance by binding RB, we expressed this mutant in HFKs (HPV16 E7ΔDLYC HFKs). Consistent with the mutation abolishing RB destabilization, HPV16 E7ΔDLYC HFKs had approximately the same amount of Rb as LXSN HFKs (Additional file 1: Fig. S2). HPV16 E7ΔDLYC HFKs have generally increased the abundance/activation of TLS proteins (POLη, ub-PCNA, but not Rad6) and replication stress response markers (p-ATR, p-ATM, TopBP1, p-RPA32, but not p-CHK1) compared to LXSN HFKs (Fig. 4). However, the increase was typically less than the increase seen in HPV16 E7 HFKs.

We next used these cell lines to determine if HPV16 E7 altered cellular responses to replication stress-inducing treatments. Three stressing agents were used, two that caused replication stress by inducing DNA lesions (Cisplatin and UV) and one that depletes nucleoside pools (hydroxyurea or HU). For this analysis, we focused on a single marker of replication stress (TopBP1) and a single marker of TLS (POLη). TopBP1 was chosen because the ability of HPV16 E7 to increase TopBP1 abundance is well established [40, 43, 44]. POLη was chosen because we have previously shown that HPV16 E7 prevented the accumulation of POLη in response to UV damage. In LXSN HFKs, TopBP1 protein levels were increased in response to UV and Cisplatin, but not in response to HU. POLη levels were increased in response to each of the three replication stressing agents in LXSN HFKs (Fig. 5). In contrast, HPV16 E7 prevented the induction of POLη and TopBP1 in response to these stimuli. HPV16 E7ΔDLYC also prevented these increases (Fig. 5).

We have previously demonstrated that increased expression of TLS polymerases was associated with worse outcomes for people with cervical cancer [19]. To determine if the same was true for TLS genes other than the TLS polymerases (POLH, POLI, POLK, REV1, and REV3L), we segregated the cervical cancers in the TCGA database based on whether they did or did not have increased expression of at least one TLS gene. The rationale for leaving these genes out of the analysis is that we have already shown that when grouped they are associated with worse patient outcomes [19]. As expected most cervical cancers had increased expression of at least one TLS gene. However, this was not associated with poor prognoses. Instead, the patient population with increased TLS gene expression had a median survival of 134.23 months (Fig. 6). In contrast, survival statistics were significantly worse for the group of people without this increased expression (median survival of 39.75 months).