HPV positive neuroendocrine cervical cancer cells are dependent on Myc but not E6/E7 viral oncogenes

Using conditional cell reprogramming, we generated a stable cell culture of an extremely rare and aggressive neuroendocrine cervical cancer. The cultured cells contained HPV-16, formed colonies in soft agar and rapidly produced tumors in immunodeficient mice. The HPV-16 genome was integrated adjacent to the Myc gene, both of which were amplified 40-fold. Analysis of RNA transcripts detected fusion of the HPV/Myc genes, arising from apparent microhomologous recombination. Spectral karyotyping (SKY) and fluorescent-in-situ hybridization (FISH) demonstrated coordinate localization and translocation of the amplified Myc and HPV genes on chromosomes 8 and 21. Similar to the primary tumor, tumor cell cultures expressed very high levels of the Myc protein and, in contrast to all other HPV-positive cervical cancer cell lines, they harbored a gain-of-function mutation in p53 (R273C). Unexpectedly, viral oncogene knockdown had no effect on the growth of the cells, but it did inhibit the proliferation of a conventional HPV-16 positive cervical cancer cell line. Knockdown of Myc, but not the mutant p53, significantly inhibited tumor cell proliferation. On the basis of these data, we propose that the primary driver of transformation in this aggressive cervical cancer is not HPV oncogene expression but rather the overexpression of Myc.

and the viral genome was integrated into host genome adjacent to the Myc gene. RNA transcripts were detected which contained a fusion of the E7/Myc genes and, in contrast of most cervical cancers, the neuroendocrine cells harbored a p53 mutation. Knockdown of Myc, but not viral oncogenes or the mutant p53, significantly inhibited tumor cell proliferation.

Results
Generation of a primary cell culture from a rare and aggressive neuroendocrine cervical cancer. During gynecological examination, a 27-year old female was found to have a 3.5 cm cervical tumor (Supplementary Figure S1). Histology of the tumor revealed a rare large cell neuroendocrine carcinoma of the cervix with high Ki67 staining (93%), suggestive of a highly proliferative malignancy. The tumor stained positively with several neuroendocrine markers including chromogranin A, synaptophysin and somatostatin receptor-2 (Supplemental Table 3). The tumor was initially staged as Ib1 cervical cancer and the patient underwent a radical hysterectomy and lymphadenectomy. At that time of the surgery the primary tumor was 4 cm in diameter, and while the surgical margins were free of tumor, 3 of 16 lymph nodes were positive. She was scheduled for 6 courses of cisplatin and etoposide, which was to be followed by chemo-radiation to the pelvis. However, after the initial day of her 5th course of chemotherapy, the cancer was noted to have progressed rapidly, with at least three leftsided, pelvic masses ranging from 1-3 cm in size. A PET CT scan also revealed three liver metastases. The patient's condition deteriorated rapidly and she died 3 months later.
In order to facilitate the analysis of molecular alterations in this lethal neuroendocrine cancer, we established a stable cell culture (GUMC-395) from a liver metastasis using conditional reprogramming 4,5 . Initially the cells grew as adherent spheres (Fig. 1a), which soon began to spread outward to form a monolayer (Fig. 1b). An optimum growth condition was defined by screening several substrate and medium formulations (Supplemental Table S1). Formation and proliferation of the monolayer was optimal using a collagen-coated substrate with F medium plus Y-27632 (Fig. 1c). A growth curve of the GUMC-395 culture is displayed in Fig. 1d. The population doubling time was approximately 40 hours. Short Tandem Repeat (STR) profiling of the GUMC-395 cell DNA was identical to that of the patient's lymphocyte DNA (Supplementary Table S2).
The transformed phenotype of GUMC-395 was demonstrated using cell invasion and migration assays (Sup plemental Figures S2 and S3) and anchorage-independent colony formation in soft agar (Fig. 1e). Intriguingly, the cells showed unusual migration into surrounding agar, consistent with their invasive behavior. In xenograft assays, measurable tumors were observed as early as three-weeks post injection into immunodeficient mice. Xenograft experiments were performed three times independently, and a total of 10 out of 10 xenograft sites produced tumors. Similar to the primary tumor (Supplemental Table 3), the xenografts were composed of a high percentage of Ki67-expressing cells (Fig. 1g). Strong expression of chromogranin, synaptophysin, and somatostatin receptor 2 by immunofluorescence and RT-PCR confirmed that the xenografts were of neuroendocrine origin ( Fig. 1h to j).
The HPV-16 genome is integrated adjacent to the cellular Myc gene. Since HPV is present in squamous carcinoma and adenocarcinoma of the cervix and is postulated to have a role in neuroendocrine cervical cancers, we used general HPV detection primers and HPV type-specific primers to screen the GUMC-395 cultures by PCR (Fig. 2a). The HPV-16 primers generated an appropriate-sized product, which was sequenced and verified to be HPV-16. Evaluation of the viral gene expression in the cells detected spliced viral mRNA for the E6 and E1^E4 (Fig. 2b), reflecting active expression of the HPV viral genome. Further, high expression of stem cell transcription factors, Klf4, Oct4, Sox2 and Myc is consistent with the aggressive and malignant nature of these cells ( Fig. 2c-f). Quantitative PCR detected an average of 50 copies of HPV DNA and 30 copies of Myc per cell (Fig. 2g). Myc gene amplification was further confirmed by CGH analysis (Supplementary Figure S4). Increased Myc protein in GUMC-395 was also confirmed in cell extracts by western blot and in xenograft tumor sections by immunohistochemistry ( Fig. 2h and i). Decreased protein levels of RB and p53 were also observed as consequence of HPV-16 infection (Fig. 2h), consistent with the expression of the E6 and E7 HPV genes.
Rolling Circle Amplification (RCA) failed to amplify HPV DNA from GUMC-395 (data not shown), suggesting that the HPV-16 genome was integrated into the host genome instead of existing as a free episome. In order to define the integration site of HPV, a 3′ RACE protocol (the APOT assay 6 ) was used to detect viral-cellular gene fusion transcripts. As shown in Fig. 3a, in addition to the virus-only transcripts, we found an HPV-Myc fusion transcript. The junction of the fused transcripts was located at a short, but highly homologous region (a stretch of 8 nucleotides: TCC/GTGCAG) between the HPV and Myc sequences, suggesting that the fusion arose from microhomologous recombination between the HPV and Myc genes (Fig. 3b (Fig. 3c). However, interestingly in this same figure the same chromosome spread, a MYC-specific FISH probe revealed that Myc is present in chromosome 8 and is additionally translocated to chromosome 21 (Fig. 3c). Further, as demonstrated in Fig. 3d, the overlapping Myc (red) and HPV (green) fluorescence signals indicate that HPV-16 was integrated into the host genome adjacent to Myc on both of these chromosomes. We further applied spectral karyotyping (SKY) 8 on GUMC-395 to detect and define the chromosomal translocations observed in the karyotyping analysis (Fig. 3e). The SKY analysis (Fig. 3e-h) revealed several translocations including t(1, 6), t(5, 7), t(10, 22) and t (17)(18)(19), in addition to a complex non-homologous translocation involving   While p53 is one of the most frequently mutated tumor suppressor genes in cancers 9 , it is only rarely mutated in cervical cancer since the high-risk HPV E6 proteins target p53 for ubiquitin-dependent degradation. In fact, the p53 mutation occurs in only 5% of cervical cancers 10 and only in HPV negative cervical tumors 11,12 . All HPV-positive cell lines, including Hela, SiHA and Caski, have wild-type p53. We examined the GUMC-395 culture to ascertain whether p53 was wild-type. Surprisingly, we observed one allele of the p53 gene in GUMC-395 contained a C to T mutation at nucleotide 817 resulting in a gain-of-function R273C mutation (Fig. 3i). The other allele of p53 was deleted as determined by CGH (Supplementary Figure S4). This neuroendocrine cell line is the first HPV-positive cell line exclusively harboring a p53 mutation.

Proliferation of GUMC-395 cells is independent of the expression of the viral E6/E7 oncogenes.
It is well established that the proliferation and survival of cervical cancer cell lines require the continued expression of the HPV E6 and E7 genes, which respectively lead to the degradation of the cellular p53 and Rb proteins 13 . For example, established cervical cancer cell lines such as HeLa, SiHa and CaSki [14][15][16][17] , as well as primary cervical cancer lines established directly from fresh biopsies 18 , remain reliant on the HPV E6/E7 genes for continued proliferation. To evaluate if the unusual GUMC-395 cell culture shared this same viral oncogene dependence, we used both siRNA and shRNA to knockdown E6 and E7 expression and evaluate whether cell proliferation was affected. Due to the splicing pattern of the E6/E7 genes in HPV-16, knockdown of E7 will reduce the expression of both the E6 and E7 genes. We documented that the transiently transfected E7 siRNA reduced both E6 and E7 mRNA in a 4-day experiment, and, more importantly, caused a consequential dramatic increase in the level of cellular p53 and Rb proteins in the HPV-16 positive SiHa cell line (Fig. 4a) and in the GUMC-395 cells (Fig. 4b).
While the reduction of E6/E7 and the resultant increase in p53 and Rb greatly slowed the proliferation of the SiHa cell line (Fig. 4c), there was no significant impact on the growth of GUMC-395 cells (Fig. 4d). This experiment was repeated two additional times, and the same effect was observed. To further evaluate whether the GUMC-395 cells remained resistant to E6/E7 knockdown beyond 4-day time points, we utilized shRNA lentiviruses that targeted E6/E7 expression to generate stable cell cultures. As shown in Fig. 4i, the lentivirus infected cells had greatly reduced expression of E6/E7 but this had no effect on GUMC-395 proliferation (Fig. 4j), even at 14 days. This is the first example of an HPV-positive cervical cancer cell line being independent of E6/E7 expression.
Growing evidence demonstrates that some mutant p53 proteins exhibit a gain-of-function phenotype that can contribute to malignant progression 19 . More specifically, recent studies indicate that the R273C p53 mutation can enhance cell proliferation 20 . To evaluate whether the R273C p53 mutation contributed to the proliferation of GUMC-395, we used siRNA to reduce cellular p53 levels (Fig. 4e). However, despite reducing p53 mRNA expression by approximately 80%, there was no effect on the growth of the GUMC-395 cells or the control SiHa cells (Fig. 4f). In contrast to the lack of effect of p53 knockdown, siRNA against Myc significantly reduced Myc mRNA expression (Fig. 4g) and inhibited the proliferation of GUMC-395 cell cultures in a 4-day experiment (Fig. 4h).
To determine if these findings would be validated with longer-term knockdown, we used shRNA lentivirus to examine cell proliferation up to 12 days. In these extended experiments we confirmed that knockdown of Myc, but not the mutant p53 or viral E6/E7, significantly inhibited tumor cell proliferation ( Fig. 4i and j). Thus, both siRNA and shRNA data indicate that the proliferation of GUMC-395 is independent of E6/E7 expression and resistant to the resultant increase in cellular p53 and Rb. One alternative interpretation of these results is that the presence of the ROCK inhibitor, Y-27632, in the medium of GUMC-395 cells might bypass the requirement for E6/E7 expression. To address this possibility, we transduced SiHa cells in medium containing the same concentration of Y-27632 that is used to propagate the GUMC-395 cells (Supplemental Figure 6). If Y-27632 were able to bypass the inhibitory effects of siRNA to E6/E7 (either via modulating HPV or Myc), then we would expect the SiHa cells to grow at the same rate as control-transduced cells. However, we found that, similar to Fig. 4c, the SiHa cells were indeed inhibited by the siRNA. Y-27632 does not bypass the requirement for continued expression of E6/E7 and we can conclude that the GUMC-395 cell line, in contrast to standard HPV cell lines such as SiHa, does not require E6/E7 protein to proliferate in vitro. While our studies indicate that Myc is essential for HPV positive GUMC-395 cell proliferation, clearly we cannot claim tumor cell specificity since Myc is also required for normal cell proliferation. However, our data does suggest that new approaches for controlling Myc overexpression might be useful in treating this tumor type.

Discussion
The molecular etiology of neuroendocrine cervical cancer is poorly understood due to the lack of well-defined cell lines. We have successfully generated and propagated a culture (GUMC-395) of an HPV-16-positive, large cell neuroendocrine cervical cancer that was metastatic to the liver. GUMC-395 cells expressed the HPV-16 E6 and E7 oncogenes, but unlike all other HPV-positive cervical cancers, their proliferation was independent of these viral proteins. Indeed, it appears that the overexpression of Myc is the predominant driver of transformation in GUMC-395. We presume that the overexpression of the Myc protein in the cell line results from the amplification of the Myc gene. The primary tumor also showed dramatic overexpression of the Myc gene, suggesting that this event was not the consequence of cell line establishment. However, since we did not evaluate Myc gene copy number in the primary tumor, we cannot conclude that the overexpression of Myc protein in the primary tumor was the consequence of gene amplification.
The Myc proto-oncogene has been shown previously to be deregulated in many types of tumors including breast, colon, small-cell lung carcinomas, osteosarcomas, glioblastomas, melanoma, myeloid leukemias, and cervical cancers (reviewed in ref. 21). Studies have shown that 30% of all HPV-18 integrations occur within the chromosomal band 8q24 near the Myc proto-oncogene, whereas HPV-16 integration occurs near Myc at a much lower rate [22][23][24] . Interestingly, integration of HPV into the Myc region strongly correlates with high levels of Myc expression 7,23 . In earlier studies short overlaps between HPV and genomic sequences were observed 25,26 . A recent study on cervical squamous carcinoma and adenocarcinoma has shown a significant enrichment of microhomologies (MHs) between the human genome and the HPV genome at or near integration breakpoints 7 . Our data suggest that, in the GUMC-395 culture, MH-mediated DNA repair pathways and MH-mediated, break-induced replication might be also be involved in HPV integration into the cellular genome. The GUMC-395 cells now provide the first in vitro platform for analyzing the molecular mechanism of HPV integration and an opportunity to identify potential therapies for this rare, aggressive and rapidly lethal cervical cancer.
It is well documented that p53 mutations are very rare in cervical cancers 12,27,28 and the analysis of cell lines derived from most HPV-positive cervical cancers has shown the p53 gene to be wild-type. p53 mutations have occasionally been shown in HPV-negative cervical tumors, although at a much lower rate compared to other types of cancers 29 . The GUMC-395 cervical cell line described in this study is the first to harbor a p53 mutation. Even more interesting, the mutation is predicted to have a gain-of-function phenotype. It is unclear how this specific mutation might contribute to the current growth properties of the cell line, especially since the presence of an active HPV E6 oncogene greatly reduced the level of p53 protein in the cells. In addition, further reduction of p53 by siRNA also had no effect on cell proliferation. However, we do not know the chronology of the genetic changes during the evolution of this tumor and it is possible that an early mutation in P53 was critical in the initiation of neoplasia. Interestingly, it was shown previously that a mis-sense mutation in p53 (codon 245) was present in an HPV-18 positive small cell neuroendocrine carcinoma of the cervix 30 . The possibility exists that tumor suppressor proteins such as p53 have somewhat different activities in cervical neuroendocrine cells and cervical squamous cells.

Materials and Methods
Ethics Statement. The patient was enrolled into study number 03-C-0277 at the National Cancer Institute.
The study was approved by National Cancer Institute Institutional Review Board. Informed, written consent was obtained from participant prior to any study procedures. All experiments were performed in accordance with protocol relevant guidelines and regulations.
Cell transfection with siRNA and cell infection with shRNA. Cells were transfected according to manufacturer's instructions using Lipofectamine RNAiMAX (Invitrogen) and 9 nM siRNA. To control for non-specific siRNA effects, a pool of non-targeting control siRNAs (ON-TARGET plus non-targeting, Thermo Scientific) was used, also at 9 nM. siRNAs for E7 (GGACAGAGCCCAUUACAAU), p53 and Myc were purchased from Dharmacon. Cell numbers were counted every 24 hours for up to 4 days. Lentiviruses with Myc shRNA, p53 shRNA, E6E7 shRNA or GFP shRNA were purchased from Santa Cruz, GUMC-395 cells were infected according to the manufacturer's protocol. 5 × 10 4 infected cells were plated and numbers of the attached viable cells were counted every 4 days, the curve was plotted as cell numbers versus days after plating.
Spectral Karyotyping (SKY) and Analysis. Preparation of SKY probes, slide pre-treatment, slide denaturation, detection, and imaging have been described previously 31 . Protocols can be accessed at: http://www.riedlab.nci.nih.gov/protocols. Metaphase chromosome suspensions were prepared by treating cells with a hypotonic solution (0.075 M KCl), followed by methanol: acetic acid (3:1, vol/vol) fixation. The suspension was then dropped onto slides using a Thermatron ™ to control humidity. The slides were aged at 37 °C for approximately one week prior to hybridization. Chromosome preparations were hybridized with SKY probes (prepared in-house) for 72 hours. Slides were imaged for SKY analysis using a Leica DMRXE microscope (Leica, Germany) equipped with DAPI and SKY ™ filters (Chroma, Bellows Falls, VT), a Xenon lamp, and Spectracube ™ (Applied Spectral Imaging, Vista, CA). Spectrum-based classification and analysis of the fluorescent images (SKY) was achieved using SkyViewTM software (Applied Spectral Imaging). Approximately 15-20 metaphase spreads were acquired for SKY analysis for each cell line and scored for numerical and structural chromosomal aberrations according to established human chromosome nomenclature rules from ISCN (2009) [2].