Abstract
The expression level of the p53 family member, p73, is frequently deregulated in human epithelial cancers, correlating with tumor invasiveness, therapeutic resistance, and poor patient prognosis. However, the question remains whether p73 contributes directly to the process of malignant conversion or whether aberrant p73 expression represents a later selective event to maintain tumor viability. We explored the role of p73 in malignant conversion in a clonal model of epidermal carcinogenesis. Whether sporadic or small interfering RNA (siRNA) induced, loss of p73 in initiated p53+/+ keratinocytes leads to loss of cellular responsiveness to DNA damage by ionizing radiation (IR) and conversion to squamous cell carcinoma (SCC). Reconstitution of TAp73α but not ΔNp73α reduced tumorigenicity in vivo, but did not restore cellular sensitivity to IR, uncoupling p73-mediated DNA damage response from its tumor-suppressive role. These studies provide direct evidence that loss of p73 can contribute to malignant conversion and support a role for TAp73α in tumor suppression of SCC. The results support the activation of TAp73α as a rational mechanism for cancer therapy in solid tumors of the epithelium. [Cancer Res 2007;67(16):7723–30]
Introduction
The p53 protein has been well established as a tumor suppressor involved in cell cycle arrest or apoptotic cell fate following DNA damage and is frequently mutated or inactivated in cancer (1). The p73 and p63 genes are rarely mutated in cancer (2). However, the expression of p73 and p63 proteins is frequently altered in epithelial cancers, suggesting either a causative role in cancer etiology or a selective pressure during tumorigenesis. Imbalances between ΔN (dominant negative) and transcriptionally active (TA) isoforms of the p53 family members in a variety of tumor types fosters debate about their function as tumor suppressors or as oncogenes (3). The ΔN isoforms of p73 and p63 are frequently overexpressed in cancer and are inhibitors of the TA isoforms of all three p53 family members, leading to the loss of p53 family–dependent cell cycle arrest and/or apoptotic response (4–7). Mutant p53 can bind to and inactivate p73 (8, 9), and certain cancers exhibit a decrease or loss of p73 and p63 expression, including mammary adenocarcinomas, squamous cell carcinomas (SCC), select lymphoid malignancies, and bladder cancer (10–18). Deregulation of isoform expression is often associated with increased tumor invasiveness, treatment resistance, and poor patient prognosis (17–20). Extensive cross-talk among the many isoforms of the p53 family members (21) have made it difficult to address the tissue-specific tumor-suppressive or oncogenic roles each individual isoform might play and the stages of tumor progression that are driven by the deregulation of p53 family member expression.
In one mouse model, mice heterozygous for p73 or p63 alone or in combination with each other or p53 (i.e., p53+/−/p63+/−, p53+/−/p73+/−, or p63+/−/p73+/−) developed multiple carcinomas, supporting tumor-suppressive roles for p73 and p63 (22). Heterozygosity of p73 and p63 in mice was associated with adenocarcinomas and SCC, whereas p53 heterozygous mice developed lymphomas and sarcomas, implying that p73 and p63 have tissue specificity for epithelial solid tumors compared with p53. This suggests tissue specificity for each p53 family member in tumor suppression. However, in a separately derived mouse model, p63 heterozygous mice did not develop any tumors, continuing the debate about the involvement of each p53 family member in tumor suppression (23). These carcinogenesis studies focused on mice or cells with total knock-out of individual or combinations of p53 family members. Isoform-specific roles in cancer have not been addressed for many tumor types, including SCC.
Our study probes the role of p73 in malignant conversion to SCC, focusing on p73 isoforms naturally expressed in mouse primary, nontransformed, and initiated keratinocytes. Although p63 has the most striking epithelial developmental phenotype (24), all three p53 family members are expressed in epithelial tissues (14, 25) and in epidermal keratinocytes in cell culture (26, 27). We focused on p73 because functional loss of this family member has been linked to SCC (6) and because conversion from the initiation stage to malignancy in our SCC model was accompanied by nearly total sporadic loss of p73 mRNA and protein expression with only slight reduction in p63 protein expression. The SCC model consists of a sequential (genetically related) lineage of mouse keratinocyte cell lines at stages of carcinogenesis, including nontransformed clone 291, initiated (precursor to SCC) clone 03C, and SCC-producing derivatives 03R, all expressing wtp53 (28–30). Knockdown of p73 expression via small interfering RNA (siRNA) in the p53+/+-initiated keratinocytes triggered malignant conversion to poorly differentiated tumorigenicity in vivo and resistance to ionizing radiation (IR) in vitro, similar to the SCC cells that had spontaneously lost p73 during malignant progression. To address isoform specificity in protection against SCC, we reconstituted the malignant-converted keratinocytes with either TAp73α or ΔNp73α. Reintroduction of TAp73α was sufficient to suppress tumorigenicity in vivo, whereas reintroduction of ΔNp73α was not. We conclude that p73 loss can trigger malignant conversion and resistance to IR in initiated cells, and that TAp73α is capable of tumor suppression in SCC.
Materials and Methods
Cell culture and IR treatment. The mouse keratinocyte SCC model and culture conditions have been described (30). The nontransformed 291 keratinocytes exhibit normal characteristics similar to primary epidermal cultures, including the regulation of proliferation and terminal differentiation by extracellular Ca2+ and keratin and differentiation marker expression. The 7,12-dimethylbenz(a)anthracene (DMBA)–initiated 03C cells fail to differentiate in response to extracellular Ca2+, and the derivative 03R cells express aberrant keratins characteristic of SCC (K8, K18) and form poorly differentiated metastatic SCC when transplanted into mice. All three of these derivative cell lines express wtp53 (28, 29). Nontransformed 291 cells and the p53-null NK1 epidermal cells obtained from A. Balmain (University of California at San Francisco Comprehensive Cancer Center, San Francisco, CA) were grown in “low-calcium medium” (final concentration, 0.04 mmol/L Ca2+) composed of EMEM (Invitrogen), supplemented with 5% (v/v) FCS (Atlanta Biologicals) treated with Chelex-100 resin (Bio-Rad) to reduce Ca2+ concentration, 10% (v/v) mouse dermal fibroblast-conditioned medium, 10 ng/mL of epidermal growth factor (EGF; Upstate Biotechnology), and 0.8% (v/v) antibiotic/antimycotic solution (Invitrogen). Initiated 03C and malignant 03R cells were maintained in EMEM lacking EGF or conditioned medium but containing 5% native FCS and 1.4 mmol/L Ca2+. Primary keratinocyte cell lysate was provided by Dr. Kathryn King in Dr. Wendy Weinberg's laboratory (Food and Drug Administration, Bethesda, MD). The 03C si-p73, 03C si-Con, and 03C si-p73 + vector, + ΔNp73α, or + TAp73α cells (see below) were maintained as for 03C cells except for the inclusion of 2.5 μg/mL puromycin to maintain the integration of the siRNA into the genome or 400 μg/mL G418 to maintain the integration of p73 isoforms. For IR treatments, cells in log-phase growth at 50% to 70% confluence were treated with 4 Gy of IR using a 137Cs source irradiator.
siRNA of p73 in 03C cells. The pSilencer vector (Ambion), modified to contain a puromycin selectable marker (termed pSilencerpuro), was used to silence all isoforms of p73 in the 03C cells. Oligos were synthesized by the Oregon Health and Science University (OHSU) oligonucleotide synthesis core to contain the 19-nucleotide siRNA p73 sequence (5′-GGGACTTCAATGAAGGACA-3′) or siRNA control sequence (5′-GTTCACCGTTCAAAGCCAG-3′) and subcloned as instructed by Ambion. The pSilencerpuro vectors were transfected into the 03C cells and clonally selected in puromycin (2.5 μg/mL medium). Silencing of p73 was confirmed by immunoblotting and by qPCR. Two independent si-p73 03C cell clones were derived and designated si-p73-1 and si-p73-2, and independently derived vector control clones were designated si-Con.
Immunoblotting and identification of p53 family isoforms. Whole cell lysate aliquots of 40 μg were separated by 8% SDS-PAGE and transferred to nitrocellulose for immunoblotting as previously reported (31). Proteins were reacted with specific antibodies to p53 (pAb122, also called pAb421 Ab-1 from Oncogene), pan p63 (4A4 monoclonal from Santa Cruz, detects all isoforms), p63α (H-129 from Santa Cruz), ΔNp63 (p40 Ab-1 from Oncogene), p73 (monoclonal 5B1288 from Imgenex, detects all isoforms) or actin (C-2, Santa Cruz) as loading control and visualized by enhanced chemiluminescence. Transfection of Myc-tagged p63 plasmids of the various isoforms (provided by Dr. Frank McKeon, Harvard Medical School, Boston, MA) and Flag-tagged p73 plasmids of the various isoforms (provided by Dr. Ute Moll, Stony Brook, NY) were used as positive controls for size comparison to determine endogenous isoform expression within the SCC model.
Reconstitution of si-p73 cells with TAp73α or ΔNp73α. To reconstitute TAp73α or ΔNp73α in the 03C si-p73 clones while maintaining endogenous mouse p73 knockdown, cells were transfected with pcDNA3 (Invitrogen) plasmids expressing Flag-tagged human p73 isoforms and selected in 400 μg/mL G418. The siRNA sequence used to silence mouse p73 isoforms was (5′-GGGACTTCAATGAAGGACA-3′), immediately preceded in the mouse sequence by an AA dinucleotide (GenBank accession NM_011642, nucleotides 985 to 986, followed by the siRNA target, nucleotides 987–1005). In contrast, the human p73 sequence (GenBank accession NM_005427, nucleotides 712–730) has a cytosine substitution (for mouse thymine) at position 11 and is preceded by a GA dinucleotide (nucleotides 710–711). These sequence differences made the siRNA selective against the endogenous mouse p73 isoforms as confirmed by immunoblotting.
Flow cytometry. Triplicate plates of cells were grown to 70% confluence and either mock treated or treated with 4 Gy IR and then harvested as described previously (28). For propidium iodide (PI) staining, cells were washed in PBS and resuspended in PI solution (200 μg/mL PI) plus Rnase A (100 μg/mL) in PBS + 0.1% Triton X-100 and then incubated at 37°C for at least 30 min before analysis by the OHSU Cancer Center flow cytometry core. A minimum of 10,000 events were counted, and Modfit cell cycle analysis program was used to analyze the data.
Colony viability assays. Cells were trypsinized 24 h following either mock treatment or treatment with 4 Gy IR and plated to obtain at least 100 colony-forming cells per control 60-mm plate in triplicate. Eleven days after plating, cells were fixed with 100% methanol and stained with 10% (v/v) Giemsa (Sigma Diagnostics). Colonies were photographed with a digital camera and counted using Imagepro Plus software (Media Cybernetics).
qPCR of p53 family downstream genes. Total RNA was collected using the TRIzol reagent (Invitrogen), treated with DNase I (Invitrogen), and purified using the RNeasy kit (Qiagen). The cDNA was generated using avian myeloblastosis virus reverse transcriptase (Roche) using random hexamers (Integrated DNA Technologies). Gene expression data were collected using a 7900HT thermocycler (Applied Biosystems, Inc.) and gene-specific primers for mouse p21 (CCATGTCCAATCCTGGTGATG and CGAAGAGACAACGGCACACTT), mouse Noxa (ACTGTGGTTCTGGCGCAGAT and TGAGCACACTCGTCCTTCAAGT), mouse Puma (GCGGCGGAGACAAGAAGAG and TCCAGGATCCCTGGGTAAGG), mouse Bax (GGGAAGGCCTCCTCTCCTACT and GAGGACTCCAGCCACAAAGATG), mouse Gadd45 (TGGTGACGAACCCACATTCA and ACGGGCACCCACTGATC), mouse p73 (GAGTCACCTGCAGCCTCCAT) and (CACCACCGTGTACCTTGTTCAT), and 18S (CGGCTACCACATCCAAGGAA and CCTGTATTGTTATTTTTCGTCACTACCT) in the presence of SYBR-Green I dye (Applied Biosystems, Inc.). SYBR-Green I fluorescence was analyzed using the ΔΔCT method (Applied Biosystems Incorporated).
Mouse tumorigenesis studies. Newborn BALB/c mice (3 days old) were injected s.c. with 2.5 × 106 cells in 50 μL of EMEM. Tumors were monitored weekly and harvested at 1 cm in diameter, fixed in formalin, and processed for H&E staining or snap frozen in liquid nitrogen for protein analysis.
Results
p53 family member protein expression in keratinocytes undergoing malignant conversion to SCC. The protein expression of the three p53 family members was examined in distinct stages of keratinocyte squamous cell carcinogenesis including genetically related nontransformed 291 cells, initiated 03C cells, and malignant 03R cells (Fig. 1A). Spontaneous conversion from initiation to malignancy corresponded to nearly total loss of p73 protein expression, reduced p63 protein expression, and failure to induce p53 following DNA damage with 4 Gy IR (lanes 6–8). The precursor 03C cells expressed at least three isoforms of p73 protein and at least five isoforms of p63 protein detectable by immunoblotting and were competent to induce p53 in response to IR (lanes 3–5). The 03C cells expressed the same comigrating isoforms of p63 and p73 (with increased expression of ΔNp73α) as their nontransformed parental cell line, 291 (Fig. 1A,, lane 2), an independent p53 null mouse keratinocyte cell line, NK1 (Fig. 1A,, lane 1) and cultured primary keratinocytes (Fig. 1B and reported in ref. 32). The TAp73α isoform and the predominantly expressed ΔNp73α isoform were confirmed by size comparison with known isoforms overexpressed as Flag-tagged proteins via plasmid transfection (Fig. 1C). A third, lower molecular weight p73 isoform is expected to be TAp73β based on migration. In order from greatest to least apparent molecular weight, the predominantly expressed ΔNp63α isoform, the expected TAp63β isoform, and the TAp63γ and ΔNp63γ isoforms, were detected by immunoblotting with αp63-specific, ΔNp63-specific, and pan-p63 antibodies and confirmed by size comparison with myc-tagged isoforms overexpressed via plasmid transfection (Fig. 1D). A fifth p63 isoform detected between the TAp63γ and ΔNp63γ isoforms is expected to be ΔNp63β based on the identities of the six isoforms of p63 reported in ref. 21.
Malignant conversion to SCC is accompanied by nearly total loss of p73 and aberrant expression of all three p53 family member proteins. A, immunoblot of p53 family member protein expression in nontransformed 291 cells (lane 2), initiated 03C cells (lanes 3–5), and malignant 03R cells (lanes 6–8). Cells were either mock treated (–) or harvested 3 or 24 h following treatment with 4 Gy IR. A 10-min exposure of p73 in 03R cell lysate (compared with a 2 min exposure for p73 in lanes 1–8) emphasizes the nearly complete loss of total p73 protein expression (top right). An independent p53 null (p53−/−) keratinocyte line NK1 showed the same p73 and p63 isoform expression pattern as 03C (lane 1). B, immunoblot of p53 family member protein expression showing comparable patterns of isoforms in initiated 03C cells and primary keratinocytes (PK). C, determination of p73 isoform expression in 03C keratinocytes by relative size comparison with transiently transfected Flag-tagged TAp73α and ΔNp73α isoforms. Because the Flag-tagged p73 isoforms were overexpressed, a lighter exposure (30 s) is shown than for (A). D, determination of p63 isoform expression in 03C keratinocytes by size comparison with transiently transfected myc-tagged TAp63α, TAp63γ, ΔNp63α, or ΔNp63γ isoforms. Lysates were blotted for p63α or ΔNp63 in addition to blotting with the pan-p63 4A4 antibody that detects all isoforms. A 2-s exposure is shown.
Malignant conversion to SCC is accompanied by nearly total loss of p73 and aberrant expression of all three p53 family member proteins. A, immunoblot of p53 family member protein expression in nontransformed 291 cells (lane 2), initiated 03C cells (lanes 3–5), and malignant 03R cells (lanes 6–8). Cells were either mock treated (–) or harvested 3 or 24 h following treatment with 4 Gy IR. A 10-min exposure of p73 in 03R cell lysate (compared with a 2 min exposure for p73 in lanes 1–8) emphasizes the nearly complete loss of total p73 protein expression (top right). An independent p53 null (p53−/−) keratinocyte line NK1 showed the same p73 and p63 isoform expression pattern as 03C (lane 1). B, immunoblot of p53 family member protein expression showing comparable patterns of isoforms in initiated 03C cells and primary keratinocytes (PK). C, determination of p73 isoform expression in 03C keratinocytes by relative size comparison with transiently transfected Flag-tagged TAp73α and ΔNp73α isoforms. Because the Flag-tagged p73 isoforms were overexpressed, a lighter exposure (30 s) is shown than for (A). D, determination of p63 isoform expression in 03C keratinocytes by size comparison with transiently transfected myc-tagged TAp63α, TAp63γ, ΔNp63α, or ΔNp63γ isoforms. Lysates were blotted for p63α or ΔNp63 in addition to blotting with the pan-p63 4A4 antibody that detects all isoforms. A 2-s exposure is shown.
Initiated cells with knocked down p73 exhibited characteristics of SCC cells in vitro including changes in growth morphology and resistance to IR. The most striking difference in p53 family protein expression levels between initiated 03C and malignant 03R cells was the spontaneous loss of p73. Therefore, we examined whether targeted reduction in total p73 protein expression in initiated 03C cells could lead to similar alterations in cellular growth characteristics, DNA damage response, or malignant conversion. To this end, initiated 03C cell clones were selected for stable expression of siRNA directed against either p73 or a mismatched sequence that did not alter expression of any p53 family members. Data are shown for two independent 03C si-p73 knockdown clones (designated 03C si-p73-1 and si-p73-2) and one control 03C clone (designated 03C si-Con). The siRNA in the knockdown clones was specific to p73 (Fig. 2A,, lanes 3 and 4, top) and did not alter the protein expression of the other p53 family members (Fig. 2A , bottom). Somewhat higher expression levels of the p53 protein in the 03C si-p73 clones than in the 03C cells are not strictly attributable to the loss of p73 protein expression. Quantification of p53 levels in 03C versus 03R cells (lanes 1 and 2) and in 03C si-p73 clones (lanes 3 and 4) versus the vector control 03C si-Con cells (lane 5) relative to actin loading control revealed <30% difference in total p53 expression. Quantitative real-time PCR (qPCR) confirmed the reduction of p73 mRNA expression in the 03C si-p73 cells (data not shown).
Disruption of p73 expression in initiated keratinocytes leads to malignant cell characteristics in vitro. A, immunoblot of p53 family members in 03C si-Con cells, 03C si-p73-1 cells, and 03C si-p73-2 cells compared with the 03C and 03R p53 family expression levels. B, cell morphology of 03C, 03C si-Con, 03R, and 03C si-p73-2 cells grown for 72 h following plating at equal densities. Bar, 50 μm. C, viability assays to determine cellular sensitivity to IR. The graph represents the percent of viable colonies remaining after IR treatment compared with the mock-treated cell type control (average of triplicate culture plates ± SD). D, flow cytometry to determine cellular G1/S arrest following treatment with IR. The graph represents the ratio of G1/S in cells as indicated. Ratios are calculated from the average (± SD) of triplicate analysis.
Disruption of p73 expression in initiated keratinocytes leads to malignant cell characteristics in vitro. A, immunoblot of p53 family members in 03C si-Con cells, 03C si-p73-1 cells, and 03C si-p73-2 cells compared with the 03C and 03R p53 family expression levels. B, cell morphology of 03C, 03C si-Con, 03R, and 03C si-p73-2 cells grown for 72 h following plating at equal densities. Bar, 50 μm. C, viability assays to determine cellular sensitivity to IR. The graph represents the percent of viable colonies remaining after IR treatment compared with the mock-treated cell type control (average of triplicate culture plates ± SD). D, flow cytometry to determine cellular G1/S arrest following treatment with IR. The graph represents the ratio of G1/S in cells as indicated. Ratios are calculated from the average (± SD) of triplicate analysis.
In culture, the 03C si-p73 clones exhibited morphologic characteristics similar to the 03R SCC cells, including pronounced multilayering of cells compared with the 03C cells (Fig. 2B). We have reported previously that the malignant 03R cells exhibit resistance to IR in vitro characterized by increased cell viability, failure to undergo G1 arrest, and failure to induce p21 protein expression following DNA damage by treatment with 4 Gy IR as compared with initiated 03C cells (28). The malignant cells were also defective in apoptosis following IR treatment, exhibiting 3-fold less caspase-3 activity compared with the initiated cells (data not shown). The 03C si-p73 cells, like the 03R cells, exhibited IR resistance (survival when replated in colony-forming assays), whereas the 03C si-Con cells retained IR sensitivity, forming fewer colonies (Fig. 2C). Furthermore, the 03C si-p73 cells failed to arrest in G1 compared with the 03C si-Con cells (Fig. 2D). Thus, knockdown of p73 caused initiated cells to acquire multiple in vitro characteristics of the malignant 03R cells.
Initiated cells with knocked down p73 undergo malignant conversion to form SCC in vivo. To determine whether the loss of p73 protein expression in initiated p53+/+ cells could contribute to malignant conversion, we s.c. injected the 03C si-p73 cells, 03R cells as positive control for cancerous growth, or 03C or 03C si-Con cells as negative controls into newborn BALB/c mice. The two independent clones of the 03C si-p73 cells formed poorly differentiated SCC (percent of mice with tumors graphed in Fig. 3A) histologically indistinguishable from 03R-derived SCC (Fig. 3B). The 03C si-p73-1 and si-p73-2 clones formed tumors at different rates, and the si-p73-2 clone formed fewer tumors. The si-p73-2 clone also exhibited a slightly higher G1/S ratio than the si-p73-1 clone (Fig. 2D). These differences were not attributable to differences in siRNA efficiency against p73 or alterations in p53 family member expression levels (Fig. 2A). Clonal selection of the 03C si-p73-1 versus si-p73-2 clones may have resulted in genetic variations to account for these phenotypic differences. Even so, the common feature of both clones was specific silencing of p73 and increased tumorigenicity relative to mice injected with 03C or 03C si-Con cells (open diamonds and squares, respectively).
Initiated cells with knocked down p73 undergo malignant conversion to form SCC in vivo. A, newborn mice injected with 03C cells (n = 4), 03C si-Con cells (n = 4), 03C si-p73-1 cells (n = 6), 03C si-p73-2 cells (n = 7), or 03R cells (n = 7) were monitored for tumor growth with the end point being a tumor measuring 1 cm in diameter. Mice injected with 03C si-p73 cells or 03R cells formed tumors (closed data points), whereas mice injected with 03C or 03C si-Con cells did not form tumors (open data points). The Kaplan-Meier method was used to calculate the number of days until tumor appearance in mice injected with each cell line compared with mice injected with 03C si-Con cells. Log-rank statistics were used to test for overall differences among groups, and a P value <0.05 was considered statistically significant. 03R, P = 0.0131; 03C si-p73-1, P = 0.0027; 03C si-p73-2, P = 0.049. B, histologically indistinguishable poorly differentiated SCC tumors harvested from mice injected with 03C si-p73 cells or 03R cells (H&E staining). Bar, 50 μm.
Initiated cells with knocked down p73 undergo malignant conversion to form SCC in vivo. A, newborn mice injected with 03C cells (n = 4), 03C si-Con cells (n = 4), 03C si-p73-1 cells (n = 6), 03C si-p73-2 cells (n = 7), or 03R cells (n = 7) were monitored for tumor growth with the end point being a tumor measuring 1 cm in diameter. Mice injected with 03C si-p73 cells or 03R cells formed tumors (closed data points), whereas mice injected with 03C or 03C si-Con cells did not form tumors (open data points). The Kaplan-Meier method was used to calculate the number of days until tumor appearance in mice injected with each cell line compared with mice injected with 03C si-Con cells. Log-rank statistics were used to test for overall differences among groups, and a P value <0.05 was considered statistically significant. 03R, P = 0.0131; 03C si-p73-1, P = 0.0027; 03C si-p73-2, P = 0.049. B, histologically indistinguishable poorly differentiated SCC tumors harvested from mice injected with 03C si-p73 cells or 03R cells (H&E staining). Bar, 50 μm.
Stable reconstitution of malignant si-p73 cells with TAp73α but not ΔNp73α restored cellular morphology comparable to initiated cells. In both the 03R cells and the 03C si-p73 cells, loss of total p73 protein expression correlated with reduced DNA damage responsiveness and tumorigenicity. To address isoform specificity of p73 in malignant conversion, the 03C si-p73 cells were stably transfected with either pcDNA3 vector alone or with vectors containing FLAG-tagged human TAp73α or ΔNp73α cDNA. Data are shown for one representative (out of two individually derived clones tested for each group) 03C si-p73 cell clone + vector, + ΔNp73α, or + TAp73α. Immunoblotting confirmed that physiologically relevant levels of p73 proteins were stably expressed comparable to those in the precursor 03C cells (Fig. 4A). Furthermore, the si-p73 cell clones expressing specific p73 isoforms or empty vector expressed p63 and p53 proteins similarly to the 03C cells. Interestingly, 03C si-p73 cells reconstituted with TAp73α reverted to morphologic characteristics of initiated 03C cells in culture, including a more adherent monolayer growth pattern compared with cells reconstituted with vector only or ΔNp73α (Fig. 4B). Cells retained resistance to DNA damage with IR as determined by flow cytometry and colony-forming efficiency assays regardless of p73 isoform expression (data not shown). This lack of IR response was verified by qPCR evaluation of induction of endogenous p21, noxa, puma, bax, and gadd45, all known to be regulated by p53 family members following DNA damage. Although the 03C si-Con cells exhibited reproducible 2- to 3-fold induction of the tested p53 downstream targets typical of the 03C initiated cells, the 03R and 03C si-p73 cells did not induce any of these genes (Fig. 4C,, top graph). Reconstituting 03C si-p73 cells with TA− or ΔNp73α did not lead to induction of p53 target genes following treatment with IR relative to vector controls (Fig. 4C,, bottom graph; under 2-fold mRNA induction is considered below the level of detection of the assay). The induction of p21 mRNA in the ΔNp73α-expressing cells was within 30% that of vector control and did not correlate with cell cycle arrest (data not shown). However, the 03C si-p73 cells expressing TAp73α expressed higher basal steady-state levels of p53 target mRNA (p21, noxa, and gadd45) compared with si-p73 cells expressing vector only, whereas ΔNp73α did not increase and seemed to marginally repress basal expression of these targets (Fig. 4D).
Stable reconstitution of TAp73α in SCC cells restores cells to premalignant phenotype. A, immunoblot of p53 family protein expression levels in 03C si-p73 cells that were stably reconstituted with either pcDNA3 vector, Flag-tagged human ΔNp73α, or Flag-tagged human TAp73α. B, cell morphology of 03C cells or 03C si-p73 cells + vector, + ΔNp73α, or + TAp73α grown for 72 h after plating at equal densities. Bar, 50 μm. The 03C si-p73 cells + TAp73α divided at a rate similar to the 03C cells and remained in monolayers, whereas cells reconstituted with vector or ΔNp73α formed multilayers like their si-p73 precursors shown in Fig. 2B. C, analysis by qPCR of p53 downstream target mRNA induction (p21, noxa, puma, bax, and gadd45) in 03C si-Con, 03C si-p73, and 03R cells (top) or 03C si-p73 + vector, + ΔNp73α, or + TAp73α (bottom) 24 h following treatment with 4 Gy IR. The graphs represent fold induction over mock-treated cell type control. Bars, SD of triplicate reactions on two separate qPCR plates. D, analysis by qPCR of p53 downstream target mRNA levels in cultured cells. The graph represents mRNA levels in 03C si-p73 cells + ΔNp73α or + TAp73α normalized to Si-p73 + vector. Bars, SD of triplicate reactions on two separate qPCR plates.
Stable reconstitution of TAp73α in SCC cells restores cells to premalignant phenotype. A, immunoblot of p53 family protein expression levels in 03C si-p73 cells that were stably reconstituted with either pcDNA3 vector, Flag-tagged human ΔNp73α, or Flag-tagged human TAp73α. B, cell morphology of 03C cells or 03C si-p73 cells + vector, + ΔNp73α, or + TAp73α grown for 72 h after plating at equal densities. Bar, 50 μm. The 03C si-p73 cells + TAp73α divided at a rate similar to the 03C cells and remained in monolayers, whereas cells reconstituted with vector or ΔNp73α formed multilayers like their si-p73 precursors shown in Fig. 2B. C, analysis by qPCR of p53 downstream target mRNA induction (p21, noxa, puma, bax, and gadd45) in 03C si-Con, 03C si-p73, and 03R cells (top) or 03C si-p73 + vector, + ΔNp73α, or + TAp73α (bottom) 24 h following treatment with 4 Gy IR. The graphs represent fold induction over mock-treated cell type control. Bars, SD of triplicate reactions on two separate qPCR plates. D, analysis by qPCR of p53 downstream target mRNA levels in cultured cells. The graph represents mRNA levels in 03C si-p73 cells + ΔNp73α or + TAp73α normalized to Si-p73 + vector. Bars, SD of triplicate reactions on two separate qPCR plates.
Reconstitution of TAp73α protein expression in SCC cells suppressed tumorigenicity in vivo. Finally, we examined whether the cells reconstituted with TAp73α exhibited a less tumorigenic phenotype than cells expressing vector only or ΔNp73α. Newborn BALB/c mice injected with the 03R cells, the si-p73 + vector cells or the si-p73 + ΔNp73α cells rapidly developed SCC (Fig. 5). However, only one of seven mice injected with 03C si-p73 + TAp73α cells developed a tumor (Fig. 5, shaded triangles). These results were confirmed in a replicate experiment in which three independent 03C si-p73 clones expressing TAp73α collectively developed tumors in only 20% of mice compared with 85% to 90% of mice injected with cells lacking p73 or expressing ΔNp73α (n = 8 mice each group). We conclude that expression of TAp73α, but not ΔNp73α, restored tumor suppression in keratinocytes that had previously undergone malignant conversion induced by the loss of p73. None of the tumors analyzed by immunoblotting had lost p53 or p63 protein isoform expression. However, the small percentage of tumors that did develop from cells reconstituted with TAp73α had lost the ectopic expression of this isoform (data not shown).
Reconstitution of TAp73αprotein expression in SCC cells reduces tumorigenicity in vivo: graph depicting the percentage of mice developing tumors following injection with 03C si-p73 cells + TAp73α (n = 7, shaded triangles), + ΔNp73α (n = 6, shaded squares), + empty vector (n = 6 for both groups, solid data points), or known SCC-producing 03R cells (n = 3, solid diamonds). Only 1:7 of the mice injected with 03C si-p73 + TAp73αdeveloped a tumor (P = 0.0351 using the Kaplan-Meier method and log-rank statistics to analyze differences in time to first appearance of tumor compared with mice injected with 03C si-p73 cells reconstituted with vector only).
Reconstitution of TAp73αprotein expression in SCC cells reduces tumorigenicity in vivo: graph depicting the percentage of mice developing tumors following injection with 03C si-p73 cells + TAp73α (n = 7, shaded triangles), + ΔNp73α (n = 6, shaded squares), + empty vector (n = 6 for both groups, solid data points), or known SCC-producing 03R cells (n = 3, solid diamonds). Only 1:7 of the mice injected with 03C si-p73 + TAp73αdeveloped a tumor (P = 0.0351 using the Kaplan-Meier method and log-rank statistics to analyze differences in time to first appearance of tumor compared with mice injected with 03C si-p73 cells reconstituted with vector only).
Discussion
Because p73 is rarely mutated in human cancer (3), it is not considered a tumor suppressor in the classic Knudson definition (33). However, loss of heterozygosity (LOH) and biallelic methylation (gene silencing) of p73 has been reported in many cancers (see Table 1 in ref. 34). The majority of the early clinical studies to examine p53 family members in cancer focused on expression levels showing that, surprisingly, p73 and p63 were frequently up-regulated. However, loss of p73 and p63 expression was associated with increased tumor invasiveness, poor differentiation, chemoresistance, and poor prognosis in many human cancers regardless of p53 mutation status, including esophageal cancer, bladder cancer, inflammatory breast cancer, thyroid cancer, SCC, and cervical cancer (11, 13, 15–19, 35). In well-differentiated human head and neck squamous cell carcinoma, similar to normal tissue, basal cells were immunopositive for p73, whereas in moderately differentiated carcinomas, p73 was ubiquitously expressed (10), and in poorly differentiated SCCs, immunostaining was negative for p73 (14). As in human tumors, whereas an initial increase in p73 immunostaining accompanied the early stages of DMBA-induced hamster oral SCC (36), the loss of p53 family member expression in p53+/−;p63+/−, p53+/−;p73+/−, or p63+/−;p73+/− mice was associated with increased tumor invasiveness (22). It was shown that 45% of p53+/−;p73+/− mice and 30% of p63+/−;p73+/− mice developed metastatic tumor types, including SCC that metastasized to the heart and lung. Only 5% of tumors from mice heterozygous for p53 alone metastasized. In these studies, expression levels of individual p73 isoforms were not distinguished. Collectively, these findings show that p73 loss is frequently associated with the later, more aggressive stages of carcinogenesis. However, the question remains about whether p73 can act as a tumor suppressor in early stages of conversion to malignancy and whether direct manipulation of p73 can affect tumor progression. The current study shows a tumor-suppressive role for p73 showing that loss of p73 in initiated p53+/+ keratinocytes was sufficient to trigger malignant conversion to poorly differentiated SCC. To address isoform specificity, we showed that recapitulation of TAp73α expression, but not ΔNp73α expression, was sufficient to restore tumor suppression.
It has been proposed that the common induction of downstream target genes is one method by which the TA isoforms of the p53 family members carry out tumor-suppressive functions. Consistent with >65% homology in their DNA binding domains, the p53, p63, and p73 proteins can transactivate common downstream target genes, including cell cycle arrest response gene p21; DNA damage repair gene gadd45; and apoptotic response genes noxa, puma, and bax (37–41). Intrafamily cooperation has been reported in that p73 and p63 were required for p53 transcription-dependent apoptotic response to DNA damage in mouse embryonic fibroblasts and in the developing central nervous system (42). However, this obligate cooperation for p53-dependent apoptotic response may be tissue specific, given that p73 and p63 were not similarly required in T cells (43). In our study, although the loss of p73 expression was linked to cellular resistance to IR and reconstitution of TAp73α but not ΔNp73α restored tumor suppression, TAp73α was unable to restore IR sensitivity to the SCC cells as detected by cell cycle analysis, colony-forming efficiency assays, and analysis of p53 target gene induction. This implies irreversible changes to cell cycle arrest and apoptotic p53 family member activities accompanying malignant conversion associated with p73 loss. Restoring these activities may require a more precise balance of expression ratios of p53 family members or additional cofactors. Recently, the DNA damage response of p53 was uncoupled from its tumor suppressor function in a mouse model of radiation-induced lymphoma (44). It was shown that p53 mediated an acute DNA damage response immediately following IR, but that this response was not responsible for suppression of tumors. Rather, p53 mediated its tumor-suppressive function through the induction of the p19ARF pathway only in select cells. The tumor-suppressive activities of other p53 family members may also be largely distinct from their DNA damage responsiveness. Further studies of p53 family activities at discrete, progressive stages of tumorigenesis should be informative in this regard, uncovering the mechanisms through which p73 and p63 isoforms exert their tumor-suppressive activities.
Our findings show that direct manipulation of p73 can alter epidermal cell fate, that loss of p73 can trigger conversion to SCC, and that reconstitution of p73 function can suppress tumor formation, supporting the notion that restoring p73 function is a rational target for molecular targeted therapy in SCC. Several studies have shown that the activation of p73 was sufficient to trigger apoptosis of cancer cells and even cause tumor regression in mice regardless of p53 status. This is particularly exciting considering that, although the restoration of p53 function lead to tumor regression in vivo (45, 46), p53 is mutated in 50% of cancers, functionally inactivated in up to 90%, and therefore, difficult to reactivate as a therapeutic target (1). Adenovirus-mediated transfer of p73 to pancreatic, colorectal, breast, lung, liver, and ovarian cancer cells attenuated cell growth even when the cells were resistant to wild-type p53 gene therapy (47–49). Furthermore, reactivation of p73 by a p53-derived apoptotic peptide led to tumor regression in vivo and attenuation of cancer cell growth in vitro, independent of whether p53 was wild type, mutant, or functionally inactivated (50). However, certain specific mutant p53 molecules, altering the conformation of the p53 DNA binding domain, have been shown to bind to and inactivate p73 (8, 9). Thus, the reactivation of p73 was effective for tumor suppression in our study in p53+/+ cells lacking wild-type p53 function (28), but may not be effective in the context of certain p53 mutations. In the current study, we show that the direct specific reconstitution of TAp73α in p53+/+ malignant-converted cells suppresses poorly differentiated SCC in vivo, adding credence to the value of TAp73 activation as a potential therapeutic target in solid tumors of the epithelium.
Overall, our studies show that, whether sporadic or siRNA induced, loss of p73 in initiated p53+/+ keratinocytes was associated with the loss of cellular responsiveness to IR and conversion to poorly differentiated SCC. The direct manipulation of p73 by specific knockdown in initiated cells and the ability to trigger malignant conversion provide evidence for a causative role of p73 in cancer etiology rather than a strictly secondary selective effect during tumorigenesis. Reconstitution of TAp73α but not ΔNp73α restored tumor suppression in vivo, focusing on TAp73α as a tumor-suppressive isoform. These results imply a role for p73 in epithelial cell fate. Further studies of the functional role of p73 in the context of other endogenous family member isoforms will not only increase our understanding of the role of the p53 family in the distinct stages of tumor development, but could lead to better predictions of patient prognosis and more effective treatments for epithelial-derived tumors.
Acknowledgments
Grant support: OHSU Department of Dermatology, CA98577 and CA98893, the OHSU Cancer Center CA69533, an OHSU Tartar Trust Fellowship (J. Johnson), and the NIH Ruth L. Kirschstein National Research Service Award (J. Johnson).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We are grateful to Loa Nowina-Sapinski for assistance with the preparation of this manuscript, and Drs. Amador Albor, Hua Lu, Charles Lopez, and Rosalie Sears for helpful discussion.