The experiments reported here demonstrate that HPV oncogene expression is required for the proliferation of primary human cervical cancer cells in culture and provide insight into factors that control tropism of SV40 and SV40-based vectors. Numerous laboratories have demonstrated that established human cervical carcinoma cells lines depend on HPV E6 and E7 for growth or survival (Francis, Schmid, and Howley, 2000
; Goodwin and DiMaio, 2000
; Hwang et al., 1993
; Jiang and Milner, 2002
; Parish et al., 2006
), but these previous experiments did not determine whether this requirement was an intrinsic property of cervical cancer cells or whether it was acquired during the establishment or propagation of permanent cell lines. Because established cancer cell lines, especially HeLa cells, diverge widely from tumors and primary cancer cells (Carlson, Iyer, and Marcotte, 2007
; Sandberg and Ernberg, 2005
), it seemed prudent to examine the requirement of E6/E7 in low passage human cancer cells. Furthermore, E6/E7 expression induces human keratinocytes to become resistant to terminal differentiation caused by high calcium and FBS (Munger et al., 1989
; Pei et al., 1998
; Schlegel et al., 1988
; Sherman and Schlegel, 1996
), the conditions typically used to culture established cell lines. Therefore, these conditions may allow the proliferation of only those cell that express E6 and E7. If this were the case, E6/E7 dependence is not an intrinsic property of cervical cancer cells, but rather an artifact of the cell culture conditions.
To determine whether cervical cancer cells have an intrinsic requirement for HPV E6 and E7, we examined the behavior of low passage cervical cancer cells cultured under conditions that do not select for E6 and E7 expression. We isolated cells from primary cervical cancers, cultured them in serum-free keratinocyte media with low calcium, and, after a minimal number of passages, analyzed their dependence on HPV E6 and E7. We believe that the early passage numbers at which these cells were tested and the conditions used to maintain them provide a better model for studying E6 and E7 dependence than established cell lines or cells immortalized in vitro. Unfortunately, the original tumors from which our primary cervical cancer strains were isolated no longer exist, so we cannot directly compare them to the primary cell strains analyzed here.
Expression of BPV E2 in all four primary cervical carcinoma cell strains repressed E6 and E7 and caused a significant inhibition of DNA synthesis (6- to 20-fold). Cancer cells as early as three passages after isolation displayed this response. E2 expression had only a slight effect on DNA synthesis of HPV-negative, primary HFFs and HFKs, and a DNA binding-defective E2 mutant had a minimal effect on DNA synthesis in the cancer cells. Moreover, growth arrest was abrogated by exogenous copies of E6 and E7. These results demonstrate that the ability of the E2 protein to bind to and repress the E6/E7 promoter is essential for its ability to induce growth arrest in primary human cervical cancer cells. Taken together, our results indicate that these primary human cervical cancer cells are dependent on E6 and E7 as early as three passages after isolation from patients and suggest that some in vivo human tumors are also dependent on E6/E7 for growth. However, even though the earliest passage cells we tested were highly dependent on HPV E6/E7, we have not ruled out the possibility that time in culture affects the sensitivity of cervical cancer cells to E6/E7 repression.
Repression of the E6 and E7 genes by the E2 protein in CVX-104 and CVX-106 cells led to elevated levels of p53 and active, hypophosphorylated p105Rb
. Levels of total p105Rb
were not elevated in the primary cervical cancer cells, unlike the situation in the established cell lines, highlighting a biochemical difference between primary cells and established cancer cell lines. In addition, E6/E7 repression in the primary cells resulted in growth arrest, increased autofluorescence and SA-β-galactosidase activity, and cell enlargement and flattening, all markers of cellular senescence. These data imply that repression of E6 and E7 in these primary cells induces senescence as a result of reactivation of p53 and p105Rb
, as is the case in HeLa cells (DeFilippis et al., 2003
). Senescence in this system is not due to up-regulation of p16ink4a
, which is constitutively up-regulated in proliferating cervical cancer cells (Sano et al., 1998
), including the primary cells studied here (unpublished data). Since senescence is irreversible (Campisi and d'Adda di Fagagna, 2007
), we infer that repression of E6 and E7 in primary cervical carcinoma cells induces permanent growth arrest.
Lambert and colleagues demonstrated that repression of HPV16 E7 in a transgenic mouse model of cervical cancer caused tumor regression (Jabbar et al., 2009
). Our results extend these findings from mice to human cells from invasive cancers that had undergone the decades-long process of carcinogenesis in women. The cellular basis of tumor regression in the mouse system has not been established, but in human cervical cancer cells, HPV repression causes senescence, which, like cell transformation, is under markedly different control in human and mouse cells (Balmain and Harris, 2000
; Chaturvedi et al., 2004
; Dotto, 1998
; Goodwin et al., 2000
; Newbold, 1997
; Rangarajan et al., 2004
). Nevertheless, the dependence on HPV oncogenes in primary human cervical cancer cells, established cervical cancer cell lines (and HPV-associated head-and-neck cancer cell lines (Rampias et al., 2009
)) and transgenic mice indicates that oncogene dependence is a fundamental feature of HPV-induced cancers.
It was previously shown that human keratinocytes immortalized in vitro
with transfected HPV16 DNA and maintained in the absence of FBS and in low concentrations of calcium are dependent on E6 and E7 for continued growth (Lee et al., 2002
). However, immortalized cells are very different from cervical carcinoma cells. For example, unlike cells derived from tumors, cells immortalized in vitro
by HPV are not tumorigenic in experimental animals (Durst et al., 1987
; Lee et al., 2002
; Pirisi et al., 1987
; Woodworth et al., 1988
). Furthermore, the vast majority of cervical high-risk HPV infections do not progress to cancer, and when progression does occur, it typically takes many years. These findings indicate that E6/E7 expression is not sufficient for tumorigenesis and that additional factors and genetic alterations are required for tumor formation in women. Thus, while it is not surprising that cells recently immortalized with E6/E7 in vitro
retain their dependence on the proximal immortalizing agent, the relevance of this finding to human cancer cells is unclear.
It is striking that cancer cells retain E6/E7 dependence despite the onslaught of mutagenesis engendered by the viral oncoproteins. Remarkably, this dependence on E6/E7 extends to established cervical cancer cell lines such as HeLa, SiHa, CaSki, which have been exposed to decades of stress and mutagenesis during time spent in culture. It is possible that the viral oncogenes are more effective than acquired mutations at inactivating cellular tumor suppressor pathways or activating mitogenic pathways, thereby eliminating selective pressure for mutations in these pathways. However, expression of dominant-negative p53 blocks p53 signaling more completely than expression of HPV E6 (Butz et al., 1995
). Rather, we propose that sustained dependence on the viral oncogenes is a consequence of the multiplicity of pathways that are targeted by the viral oncoproteins. For example, in addition to the p53 pathway, the E6 protein targets telomerase, several PDZ domain-containing proteins, and components of the DNA repair and apoptosis machineries (Howie, Katzenellenbogen, and Galloway, 2009
). Because these multiple pathways are unlikely to be activated or inactivated by mutations in one or a few cellular genes, viral proteins are more effective drivers of cell growth and survival than are random mutations occurring in cellular DNA. This implies that therapies directed against individual cellular pathways targeted by the viral oncoproteins are likely to be less effective than therapies against the viral oncogenes themselves.
We have been able to expand primary cultures with high purity from 20–30% of cervical cancer tumor biopsies. Although this rate of success is higher than in a previously reported attempt to culture cervical carcinoma cells (Ku et al., 1997
), we do not know if cervical carcinomas that fail to grow in culture also depend on E6 and E7. Nevertheless, our data indicate that E6 and E7 dependence is an intrinsic property of cervical carcinoma cells from at least a substantial fraction of patients and is not a trait that is acquired during the establishment and propagation of permanent cell lines. These findings in turn imply that treatments that inhibit the expression or activity of E6 and E7 may have therapeutic benefit in cervical cancer patients. The cancer isolates we studied were from early stage carcinomas; additional studies should determine whether cells isolated from advanced cervical carcinomas are also dependent on E6 and E7 for growth.
Our ability to infect these primary cancer cells with our E2 vector was informed by an understanding of the factors that control SV40 infection. The efficiency of SV40 infection appears to be dictated in large part by levels of cell-surface GM1, which binds directly to the major viral capsid protein, VP1 (Campanero-Rhodes et al., 2007
; Neu et al., 2008
; Tsai et al., 2003
). In our experiments, CVX-102 and HeLa-Sen2 cells, which displayed the highest level of GM1, were efficiently infected with SV40 and Pava, whereas cells with lower levels of GM1, such as SiHa and CaSki cells and most primary cervical cancer cells, were infected poorly unless supplemented with GM1. These results demonstrated that infection by SV40 vectors can be improved by addition of GM1 to human cells.
There are reports that recombinant SV40 vectors and in vitro
assembled SV40 virus-like particles are capable of efficiently transducing many mammalian cell types (Kimchi-Sarfaty et al., 2004
; Lund et al., 2005
; Strayer et al., 2005
). However, by using quantitative measures of early viral gene expression, we found considerable differences in the ability of different cells to be infected by SV40 and SV40-based vectors generated in permissive monkey cells. In fact, seven of nine cell strains tested in this study infected poorly in the absence of GM1 supplementation. Similar differences have been documented by other groups in studies of the mechanism of SV40 infection, which identified GM1 as a cell surface receptor for this virus (Campanero-Rhodes et al., 2007
; Low et al., 2004
; Tsai et al., 2003
). Our demonstration that different cell isolates display marked differences in GM1 cell-surface levels provides an explanation, at least in part, for these differences in infectivity. The addition of GM1 to the culture medium provides a simple and inexpensive method to expand the tropism of SV40 and SV40-based vectors for gene transfer and gene therapy applications.