Disruptive TP53 Mutation Is Associated with Aggressive Disease Characteristics in an Orthotopic Murine Model of Oral Tongue Cancer

Head and neck squamous cell carcinoma (HNSCC) consistently ranks among the 6 most frequently diagnosed cancers in the world (1). In 2010, an estimated 49,260 new head and neck cancers are expected in the United States (2). Currently, no molecular biomarkers are available to aid in decisions about treatment of HNSCC. Although human papillomavirus (HPV) infection has been associated with increased lymphatic metastasis and improved response to chemotherapy and radiation treatment (3), this finding has been associated with squamous cell carcinoma of the oropharynx, and treatment of patients with HPV-positive tumors currently remains the same as that for those with HPV-negative tumors. Another candidate biomarker in HNSCC is mutation in the TP53 gene. The TP53 gene is frequently altered in HNSCC, reported in about 40% of cases (4–6). Several groups have shown that mutation in TP53 correlates with poor response to therapy and shorter survival in patients with HNSCC (7–10).

In 2007, Poeta and colleagues evaluated the TP53 mutation status in 420 patients with HNSCC that had been treated surgically and correlated TP53 mutations with clinical outcomes (7). In their study, they classified TP53 mutation into disruptive and nondisruptive mutations, defining disruptive mutations as either nonsense mutations or missense mutations in the L2–L3 DNA-binding domain of p53 leading to an amino acid change resulting in a residue with a different polarity or charge. All other mutations were classified as nondisruptive. On the basis of this classification, they found that patients with tumors containing mutant p53 were associated with decreased survival compared with those with p53 wild-type tumors and that disruptive mutations were associated with the shortest survival. The study by Poeta and colleagues provides clinical evidence that disruptive TP53 mutation is prognostic for poor outcome in patients treated surgically for HNSCC. Despite strong evidence that disruptive TP53 mutation is associated with worse outcome in patients with HNSCC, it remains unclear whether disruptive TP53 mutations are associated with tumors with more aggressive tumor growth and metastasis specifically. Therefore, we evaluated the association of TP53 mutations in a panel of HNSCC cell lines, with outcomes after these cell lines were assessed in an orthotopic murine xenograft model in the present study.

Preclinical animal xenograft tumor models are widely used to study the growth and spread of cancer as well as to examine the therapeutic efficacy and mechanisms of action of new antitumor agents. Orthotopic animal models have been useful in advancing our understanding of the mechanisms of regional and distant metastatic spread of HNSCC (11) because the implantation of tumor cells into the anatomic site of tumor origin rather than into subcutaneous or other heterotopic implantation sites more closely resembles the host microenvironments present in the clinical setting of HNSCC (12). The orthotopic model has been previously validated, and several HNSCC cell lines have been shown to exhibit tumor invasion and cervical lymph node metastasis that are characteristic of the clinical disease (11). For decades, many groups have used the orthotopic xenograft system to study HNSCC, but most have evaluated 1 or 2 cell lines alone to represent a clinical disease that has very heterogeneous behavior and response to treatment (13–16). A large panel of tumor cell lines should better represent the wide variability in behavior of tumors found in patients with a given cancer, leading to more widely applicable conclusions especially in the study of TP53, as the number and type of TP53 mutations are large as reported by the International Agency for Research on Cancer (17).

In this study, our objective was to characterize the in vivo behavior of 48 established HNSCC cell lines in an orthotopic xenograft nude mouse model. We characterized the baseline tumorigenic potential of a large panel of authenticated, short tandem repeat (STR)-genotyped HNSCC cell lines to determine whether different cell lines of the same tumor histology exhibit a wide spectrum of tumorigenicity and metastatic potential in the orthotopic model, mirroring the variation in tumor behavior observed clinically. We also sequenced the TP53 gene in each of these cell lines to assess whether TP53 mutations were associated with aggressive tumor behavior in the in vivo model, matching previously reported clinical observations.

Male athymic nude mice, ages 8 to 12 weeks, were purchased from the animal production area of the NCI-Frederick Cancer Research and Development Center (Frederick, MD). The mice were housed and maintained in laminar flow cabinets under specific pathogen-free conditions. The facilities were approved by the Association for Assessment and Accreditation of Laboratory Animal Care International in accordance with current regulations and standards of the U.S. Department of Agriculture, the U.S. Department of Health and Human Services, and the NIH. The mice were used in accordance with the Animal Care and Use Guidelines of The University of Texas MD Anderson Cancer Center under a protocol approved by the Institutional Animal Care and Use Committee.

We evaluated 48 HNSCC cell lines. Information about each cell line and appropriate growth media is shown in Supplementary Table S1. Adherent monolayer cultures were maintained on plastic and incubated at 37°C in 5% CO₂ and 95% air. The integrity of all maintained cell lines was clearly established by STR profiling. The integrity of each maintained cell line was clearly established by comparing the result from STR profiling (18) with that reported for the original stock. The cultures were free of Mycoplasma species and were maintained for no longer than 12 weeks after recovery from frozen stocks.

To determine the tumorigenicity and metastatic potential of the 48 HNSCC cell lines, we used the orthotopic nude mouse model of HNSCC (11, 12). Each of the HNSCC cell lines was harvested from subconfluent cultures by trypsinization and washed. HNSCC cells (5 × 10⁴ cells suspended in 30 μL of serum-free medium) were injected into the lateral tongue as described previously (11). For each cell line, 8 to 10 mice were prepared. Twice a week, the mice were weighed to determine weight loss and the tumor was measured in 2 dimensions with microcalipers. Tumor volume was calculated as V = AB²(π/6), where A is the longest dimension of the tumor and B is the dimension of the tumor perpendicular to A. The mice were asphyxiated with CO₂ when they had lost more than 20% of their preinjection body weight or had become moribund. The remaining mice were asphyxiated 60 days after cell injection.

At necropsy, the tongue tumors and cervical lymph nodes were removed. Each tumor was fixed in formalin and embedded in paraffin for immunohistochemistry and hematoxylin and eosin (H&E) staining. The cervical lymph nodes were also embedded in paraffin and sectioned, stained with H&E, and evaluated for metastasis by light microscopy. Each lymph node identified was evaluated by H&N staining of a single cross section through the lymph node. Measurement of tumor volume on day 32 was chosen for comparative purposes to determine the tumorigenicity of cell lines, as it was found that the majority of aggressive cell lines established modest tumors by this date.

Sequencing of TP53 was done on all 48 HNSCC cell lines. Cells from each line were cultured, trypsinized, and pelleted via centrifugation. Genomic DNA was then extracted with the QIAamp DNA Mini Kit (QIAGEN), and exons 2 to 11 of coding regions and splice sites of the TP53 gene were evaluated via direct sequencing. Sequencing was done twice for each cell line, once at the Baylor College of Medicine Human Genome Sequencing Center and once through the sequencing service provided by Beckman Coulter Genomics. When there was a discrepancy between the 2 sequencing results, the tracings were reviewed to arrive at the final sequencing result.

We conducted Western blot analyses to determine expression of p53 in each HNSCC cell line. Cells from each of the 48 HNSCC cell lines were individually plated on 6-cm plates (Costar) and incubated for 24 hours. The medium was then changed in control plates, whereas a second plate for each cell line was treated with 5 μg/mL 5-fluorouracil (Sigma) for 24 hours to elicit a DNA damage response and thereby amplify wild-type p53. Cells were then lysed with radioimmunoprecipitation assay buffer, and the protein fraction was collected and evaluated via Western blot analysis as previously described (13). The membranes were blocked for 1 hour at room temperature with 5% bovine serum albumin in TBS containing 0.1% Tween-20 (TBS-T) and incubated overnight at 4°C with anti-p53 DO-1 (1:1,000; Santa Cruz Biotechnology) in 5% bovine serum albumin in TBS-T. Next, the membranes were washed with TBS-T and incubated for 1 hour at room temperature in species-appropriate fluorescently conjugated secondary antibodies (goat anti-mouse IRDye 800; Invitrogen). The membranes were then analyzed with the SuperSignal West Chemiluminescence System (Pierce Biotechnology) and an Odyssey infrared imaging system (LI-COR Biosciences). Relevant signal intensities were determined with LI-COR imaging software version 3.0. To verify equal protein loading, the membranes were probed with anti-β-actin (1:5,000).

Wild-type TP53 was defined as unperturbed sequence of p53 DNA. Mutations were grouped into disruptive and nondisruptive categories. Disruptive mutations, as described previously (7), were defined as any mutation leading to a stop codon, or missense mutations occurring within the L2–L3 loop of the key DNA-binding domain, leading to substitution with an amino acid of a different polarity or charge group. All other mutations were defined as nondisruptive. We also applied an alternate classification system based on TP53 mutation and p53 protein expression, combined. Null mutations were defined as p53 mutations that showed loss of p53 expression on Western blot analysis in both the basal and 5-fluorouracil–induced conditions. Overexpressed mutations were defined as cell lines with mutant TP53 and high basal levels of p53 expression. Overexpressed mutations were then further classified into disruptive and nondisruptive mutations, as described earlier.

One representative unstained slide of an orthotopic tumor from each cell line was chosen for immunohistochemical detection of p53 using CM-1 antibody (1:200; Leica Microsystems, Inc.). Immunohistochemical procedures were done as described previously (13). Immunostained slides were reviewed by 2 investigators (V.C. Sandulache and T.J. Ow) blinded to cell line and p53 status. Slides were scored for the presence of stained tumor cells, greater or less than 50% of tumor cells with positive staining, as well as nuclear and/or cytoplasmic staining. After blinded review, any discrepancies between the 2 reviewers were discussed between the 2 reviewers to arrive at a consensus.

To make a construct expressing mutant p53-P151S, total RNA was isolated from Tu138 cells with TRIzol reagent (Invitrogen) and reverse transcription PCR was done to amplify p53 cDNA with the Iscript cDNA Synthesis Kit (Bio-Rad), using the primers: 5′-BamHI, 5′-cgggatccaagtctagagccaccgtcca and 3′-EcoRI; 5′-GGAATTCTCAGTCTGAGTCAGGCCCTTCTGTC. The PCR products were inserted into the BamHI and EcoRI restriction sites of the pBaBe-puro retroviral vector (Addgene). After verification of the mutant p53 by direct sequencing, the pBaBe-p53-P151S construct was transfected into Phoenix cells with 2 helper plasmids, pCGP and pVSVG, to generate virus. The cell culture supernatant was filtered and used to infect UM-SCC-1 cells to carry a splice-site mutation making it p53 null. The infected cells were selected with puromycin and pooled cells were used in the study.

To decrease the expression of mutant p53 in Tu138 cells, lentiviral vector pLVUHshp53 containing short hairpin RNA (shRNA) that targets p53 was purchased from Addgene, along with the helper plasmids, pMD2.G and pCMV-dR8.2 dvpr. These were transfected into 293T cells. The supernatant was filtered and used to infect Tu138 cells. The infected cells were sorted with flow cytometry using the GFP selection marker expressed by pLVUHshp53.

Pearson correlation coefficients were computed for the relationships between survival time and tumor volume on day 32 and between incidence of cervical lymph node metastasis and tumor volume on day 32, and P values were determined for the 2-tailed test. Survival was analyzed by the Kaplan–Meier method and compared with log-rank tests. The Fisher exact test was used to analyze associations between TP53 mutation status and incidence of cervical lymph node metastasis. The unpaired 2-tailed t test was used to compare the differences in mean tumor volume between groups. Analysis was conducted with GraphPad Prism version 5.01 (GraphPad Software). For all comparisons, P < 0.05 was considered statistically significant.

The mean tumor volumes that were measured twice a week after tumor cell implantation are shown in Supplementary Fig. S1 for each of the 48 HNSCC cell lines. Seven cell lines (UM-SCC-11A, UM-SCC-11B, UM-SCC-19, UM-SCC-22B, UM-SCC-25, MDA1386TU, and MDA1386LN) did not form tumors in any of the mice injected. The remaining 41 (85.4%) HNSCC cell lines formed tumors in the tongues of nude mice with an incidence shown in Table 1. Survival curves for mice injected with each of the 48 HNSCC cell lines are shown in Supplementary Fig. S2. The mean survival time ranged from 20.7 to 60.0 days. We found a strong inverse correlation between mean tumor volume on day 32 and mean survival time (r = −0.969, P < 0.0001; Supplementary Fig. S3A).

To study the metastatic potential of HNSCC cell lines in the orthotopic model, we harvested cervical lymph nodes at necropsy and examined them histologically to identify metastases. We found that 19 (39.6%) of the 48 HNSCC cell lines produced cervical lymph node metastases (Table 1). The incidence (percentage of mice with lymph node metastasis) ranged from 0% to 88%. The HN5, HN31, FaDu, OSC-19, and Detroit562 cell lines developed metastases at the highest rate; cervical disease was identified in 50% or more of the mice injected with each of these cell lines. We found a correlation between the mean tumor volume on day 32 and the incidence of cervical lymph node metastasis (r = 0.581, P < 0.0001; Supplementary Fig. S3B).

The TP53 gene was analyzed in 48 HNSCC cell lines by direct sequencing of genomic DNA (Table 2). Five (10.4%) of the cell lines (HN30, UM-SCC-6, UM-SCC-17A, UM-SCC-25, and UM-SCC-47) were classified as having wild-type TP53, and the other 43 (89.6%) were found to have mutations in TP53. Eight (16.7%) of the cell lines (HN31, UM-SCC-4, MDA1686, MDA1986LN, FaDu, OSC-19, TR146, and CAL-27) were found to have disruptive TP53 mutations. The other 35 (73.0%) were found to have nondisruptive TP53 mutations; these included 23 missense mutations, 4 deletion mutations, and 8 mutations identified at splice sites (ref. 7; Table 2).

To determine whether there was an association between tumor behavior and TP53 mutation, we categorized mice into groups by TP53 mutation status (wild-type, nondisruptive, or disruptive) and compared survival time between these groups using Kaplan–Meier survival analysis and log-rank tests. The mean survival times for the wild-type TP53 group and the mutant TP53 group were 55.1 and 51.7 days, respectively. Although mice in the wild-type TP53 group survived longer than those in the mutant TP53 group, this difference was not significant [HR, 0.607; 95% confidence interval (CI), 0.360–1.025; P = 0.061; Fig. 1A, and Table 3). We also determined whether there was an association between the mean tumor volume on day 32 and TP53 mutation status (wild-type vs. mutant; Fig. 1C). Although the mean tumor volume seemed smaller in the wild-type group than in the mutant group (9.22 mm³ vs. 16.10 mm³), this difference was not significant (P = 0.073).

We found that the cell lines with disruptive TP53 mutations were significantly more aggressive than those with wild-type TP53 or nondisruptive TP53 mutations. The mean survival time for mice whose tumors had disruptive TP53 mutations was 46.2 days, significantly shorter than those for mice whose tumors had wild-type TP53 (HR, 0.340; 95% CI, 0.192–0.603; P = 0.0002) or nondisruptive TP53 (HR, 0.347; 95% CI, 0.217–0.553; P < 0.0001) mutations (Fig. 1B and Table 3). However, the difference in mean survival time between the wild-type TP53 group and the nondisruptive TP53 mutation group (55.1 vs. 52.9 days) was not significant (HR, 0.682; 95% CI, 0.383–1.214; P = 0.193; Fig. 1B and Table 3). The mean tumor volume on day 32 was also significantly higher in the disruptive TP53 mutation group (26.12 mm³) than in the wild-type TP53 group (9.22 mm³) and the nondisruptive TP53 mutation group (13.83 mm³; P = 0.0003 and P = 0.0001, respectively; Fig. 1D).

Cervical lymph node metastasis was detected in only 1 (2.0%) of the 49 mice in the wild-type group (Table 3). In the mutant TP53 group, 51 (12.1%) of the 421 mice developed cervical lymph node metastasis. We found that 29 (8.4%) of the 345 mice in the nondisruptive TP53 mutation group and 22 (29.0%) of the 76 mice in the disruptive TP53 mutation group developed cervical lymph node metastasis. Thus, mice in the disruptive TP53 mutation group showed a significantly higher incidence of cervical lymph node metastasis than mice in the wild-type group (P = 0.0001) and the nondisruptive TP53 group (P = 0.0001). Mice in the mutant TP53 group also showed a significantly higher incidence of cervical lymph node metastasis than those in the wild-type group (P = 0.030). Although the incidence of cervical lymph node metastasis was also higher for mice in the nondisruptive group than for those in the wild-type TP53 group, this difference was not significant (P = 0.152).

We also conducted Western blot analyses to determine whether mutation in the TP53 gene led to p53 protein stabilization or loss of expression (Table 2 and Supplementary Fig. 4). We found that mutation in TP53 led to overexpression and stabilization of mutant p53 protein in 29 ell lines whereas protein expression was lost in 14 cell lines (which we refer to as null cell lines). Of the 14 null cell lines, 10 carried splice site or deletion mutations. The 14 p53-null cell lines were categorized into a separate group and compared with the wild-type group (5 cell lines) and the overexpressed mutation group (29 cell lines). The overexpressed p53 mutation group was then stratified into nondisruptive (23 cell lines) and disruptive (6 cell lines) groups, as defined previously.

Although the mean survival time for the overexpressed group (50.8 days) was significantly shorter than that for the wild-type group (55.1 days; HR, 0.583; 95% CI, 0.342–0.993; P = 0.047), the difference between the null group (52.2 days) and the wild-type group was not significant (HR, 0.628; 95% CI, 0.333–1.181; P = 0.149). However, after stratification of the overexpressed mutation group into disruptive and nondisruptive categories, the mean survival time for disruptive p53 mutation (46.1 days) was significantly shorter than that for the wild-type (HR, 0.306; 95% CI, 0.161–0.580; P = 0.0003), null (HR, 2.409; 95% CI, 1.403–4.138; P = 0.0014), and nondisruptive (HR, 0.335; 95% CI, 0.207–0.609; P = 0.0002) groups. The difference between the mean survival time for the wild-type group and that for the nondisruptive group (52.1 days) again was not significant (HR, 0.660; 95% CI, 0.365–1.194; P = 0.169; Fig. 2A and B, Table 3).

The mean tumor volumes on day 32 for the null group, overexpressed mutation group, and nondisruptive group were not significantly different from those for the wild-type group (P = 0.139, P = 0.091, and P = 0.231, respectively; Fig. 2C and D). However, the mean tumor volume on day 32 for the disruptive TP53 mutation group was significantly higher than those for the wild-type TP53, null, and nondisruptive TP53 mutation groups (P = 0.0003, P = 0.0002, and P = 0.0002, respectively; Fig. 2D).

Cervical lymph node metastasis was detected in 12 (8.7%) of the 138 mice in the null group, 39 (13.8%) of the 283 mice in the overexpressed mutation group, 23 (10.1%) of the 227 mice in the nondisruptive group, and 16 (28.6%) of the 56 mice whose tumors had disruptive TP53 mutations (Table 3). Thus, mice in the disruptive TP53 mutation group showed a significantly higher incidence of cervical lymph node metastasis than those in the wild-type group, the null group, the overexpressed mutation group, and the nondisruptive mutation group (P = 0.0001 for all four). Although the incidence of cervical lymph node metastasis was higher for mice in the p53-null group and the nondisruptive group than for those in the wild-type TP53 group, these differences were not significant (P = 0.189 for null and P = 0.091 for nondisruptive). The difference in incidence of cervical lymph node metastasis between the wild-type group and the overexpressed mutation TP53 group was significant (P = 0.016).

Thirty-eight orthotopic tumors were available for p53 immunohistochemical analysis (Supplementary Table S2 and Supplementary Fig. S5). Among the 4 cell lines that were TP53 wild-type, 3 lines showed positive staining for p53, with a variable staining pattern in more than 50% of tumor cells present. Interestingly, UM-SCC-47, the only line among the panel that is positive for HPV DNA (19) was among these lines. UM-SCC-6 showed no cells positive for p53. Among the 9 cell lines that were absent for p53 after immunoblotting with DO-1, 6 cell lines showed an absence of p53 expression upon immunohistochemical evaluation. HN4 and MDA686LN showed diffuse, intense staining, whereas OSC-19 showed positive staining in a variable pattern in more than 50% of tumor cells. Among the 26 cell lines that showed overexpression of p53 on Western blot analysis, all of the tumors from these cell lines showed expression of p53, with diffuse staining in 24 of these. UM-SCC-17B and UM-SCC-22A showed some variability in p53 staining with more than 50% of tumor cells staining. In general, there appeared to be specific staining patterns that corresponded to TP53 status, that is, a positive but variable pattern was associated with TP53 wild-type status, TP53 missense mutation appeared to be associated with diffuse positive staining, and TP53 nonsense, deletions, and splice mutations appeared to be more often associated with the absence of staining. However, there were discrepancies as noted earlier.

It was noted that the Tu138 cell line was highly aggressive in the orthotopic model, and this cell line harbored a relatively novel mutation, converting proline to serine at codon 151. p53 was overexpressed in this cell line, and the mutation is classified as “nondisruptive” because it is outside of the L2–L3 loop. To test potential gain-of-function properties of the p53-p151s mutant, we orthotopically injected Tu138 lentiviral control, Tu138-lenti shp53, UM-SCC-1-pBabe, and UM-SCC-1-pBabe-p53-p151s into 10 nude mice for each cell line. Forced expression of the mutant p53 in UM-SCC-1 cells resulted in enhanced tumor growth and decreased survival compared with UM-SCC-1-pBabe control cells (Fig. 3A and B), whereas reduced p53-p151s expression in Tu138 cells led to slower tumor growth and longer survival than Tu138 empty vector control cells (Fig. 3C and D). Furthermore, p53-p151s conferred metastatic behavior to the UM-SCC-1 cell line, which was nonmetastatic in transfected cells (Fig. 3E). Similarly, decreased endogenous expression of p53-p151s in Tu138 resulted in decreased incidence of metastasis (Fig. 3F and G).

In this study, we determined the in vivo growth and metastatic characteristics of 48 HNSCC cell lines using an orthotopic xenograft nude mouse model of oral tongue carcinoma. A wide range of behavior was observed, from cell lines that formed no tumors to those that formed tumors that grew rapidly and consistently metastasized. We examined the TP53 mutation status and tumor behavior and showed that cell lines with disruptive TP53 mutations were associated with shorter survival, faster tumor growth, and increased incidence of cervical lymph node metastasis in the orthotopic model of HNSCC.

Our study of 48 HNSCC cell lines in this in vivo model is the most comprehensive to date. We have shown that both tumor growth rate and incidence of cervical lymph node metastasis vary considerably between cell lines, consistent with the range of behavior observed clinically between tumors in individual HNSCC patients. Although we found some correlation between tumor volume and incidence of cervical lymph node metastasis, some of the cell lines with the highest rates of cervical lymph node metastasis showed only a moderate tumor volume whereas some cell lines with fast tumor growth did not form metastases. These findings suggest that cell characteristics other than proliferation rate contribute to metastasis and that these cell lines can be differentiated by using this orthotopic model. The data presented here provide a valuable resource for future studies of HNSCC tumor progression and metastasis, using the orthotopic xenograft model and HNSCC cells.

We chose to study TP53 mutation because of the fundamental role p53 seems to play in HNSCC (20). After analyzing the TP53 sequence of 48 HNSCC cell lines, we found that only 5 (10.4%) of the cell lines retained the wild-type TP53 sequence; this is a substantially lower proportion than that reported in clinical specimens, which is typically 40% to 60% (7, 9, 21). This observation may suggest that there is a selection for TP53 mutations when immortalized cell lines are established, or perhaps the incidence of TP53 mutation in the clinical setting is underreported, for example, as a result of the inherent contamination of tumor specimens with normal tissue containing wild-type TP53. This latter possibility is consistent with findings of a large-scale exomic sequencing of oral cavity cancers that we are completing (our unpublished observations).

The strongest evidence to date linking TP53 mutation with outcome in HNSCC has been presented by Poeta and colleagues, who found that disruptive mutations were associated with decreased overall survival (7). In our study, we consistently observed that disruptive TP53 mutations were significantly associated with shorter survival, greater tumor volume, and increased incidence of cervical lymph node metastasis in the orthotopic xenograft model. These results indicate that disruptive TP53 mutations may play a critical role in the progression and metastasis of HNSCC and support previous clinical reports concluding that disruptive TP53 mutations are a prognostic biomarker for patients with HNSCC.

Despite our evidence supporting the disruptive/nondisruptive classification system, it is recognized that this categorization has a theoretic structural basis that predicts p53 function based on the location and type of substituted amino acids. We have isolated the p151s mutation, classified as “nondisruptive,” and showed that it has significant gain-of-function activity in our model. It is possible that the disruptive categorization merely selects for mutations that are deleterious, and it seems that a more concrete delineation of p53 mutations based on alteration of function or demonstration of one or several consistent gain-of-function mechanisms is necessary. The fact that TP53-null mutations were relatively benign in our study suggests that gain-of-function activity may contribute significantly to the deleterious effect of TP53 mutations associated with the poorest outcome. A large-scale analysis of TP53 mutations is required to identify and characterize the most deleterious mutations from a mechanistic standpoint.

In our study, we found that most of the insertion, deletion, and splice mutations resulted in loss of p53 expression. Therefore, we used the combination of mutation and lack of protein expression to further define a separate, p53-null group. Although previous clinical studies of ovarian cancer have suggested that patients with tumors containing p53-null mutations are at risk for distant metastases and have a poor prognosis (22, 23), we found that cell lines with null mutations were significantly less aggressive than those in the disruptive TP53 category, suggesting that p53-null mutations are less deleterious than disruptive mutations in HNSCC cells.

A recent publication has reported that truncating TP53 mutations, defined as nonsense, splice, and frameshift mutations, and largely represented by the “null” group in our study, are associated with poor prognosis among patients with HNSCC (24). The discrepancy of our data with this literature may be due to several considerations. The most obvious is that mice in our study are not treated for their disease, whereas patients in the study by Lindenbergher-van der Plas and colleagues were treated with either surgery or surgery and postoperative radiation. It may be that specific TP53 mutations contribute greatly to tumor formation and progression, whereas other mutations are more deleterious the setting of radiation therapy. A second consideration is that a significant number of patients in the study had oropharyngeal tumors, and though HPV was evaluated in this study, microenvironmental factors specific to the lymphoid-rich oropharynx may contribute to tumor phenotype and thus alter the role of TP53 mutation. Further mechanistic evaluation is required, and our work represents a model in which these questions can be isolated and examined in future studies.

The categorization of mutations into disruptive and nondisruptive is not the only system that has been correlated with outcome in HNSCC. Another classification scheme has been described that is based on p53 transactivation activity in a yeast-based functional assay (6, 25). The transactivation activity of p53 plays a key role in its tumor suppressor function, and it has been shown that a genetically engineered mouse model carrying a transactivation-deficient TP53 mutation is predisposed to cancer development (26, 27). Perrone and colleagues recently reported that oral cavity tumors with wild-type TP53 or mutations that retain partial transactivation ability showed a higher response rate to induction chemotherapy (8). In the present study, 28 of the cell lines could be categorized by functional status on the basis of transactivation ability. Splice site, deletion, and mutations resulting in a premature stop codon are not classifiable by this system, although the majority would be considered nonfunctional. We classified these 28 cell lines into fully functional, partially functional, and nonfunctional groups. However, only 2 of the 28 cell lines carried mutations that retained partial transactivation ability whereas all others were classified as nonfunctional. As expected, in the orthotopic nude mouse model the cell lines with nonfunctional p53 mutations were associated with shorter survival, greater tumor volume, and increased incidence of cervical lymph node metastasis than the cell lines with wild-type or partially functional p53 mutations (data not shown). Because only a few of our panel of cell lines carried mutations that retained partial transactivation ability, it is difficult to draw conclusions about the role of partially functional mutations in our system from these results. The true impact of partially functional mutations in HNSCC remains to be elucidated, and this could perhaps be further studied in the preclinical model we describe here if more cell lines with partially functional p53 mutations are identified or created.

There are some discrepancies between TP53 mutations identified in the present study and those described in previous reports evaluating the same HNSCC cell lines (Supplementary Table S3). In the current study, all of our cell lines were received from the American Type Culture Collection or requested directly from the laboratory from which they originated, and each line was validated by STR genotyping confirmed against that obtained by the originators (19). We also sequenced TP53 twice in independent laboratories for each line and confirmed p53 protein expression by Western blotting. The inconsistencies between our report and those previously reported may reflect the differences of the procedures to detect TP53 status in HNSCC cells. Alternatively, these discrepancies may be the result of cell line contamination or mis-identification in the previous reports. Discrepancies among published works reporting TP53 status in cell lines have been reviewed, and this unfortunately occurs frequently (17). Although we cannot resolve discrepancies between existing literature and our report, we attest to undertaking rigorous cell line authentication (19) and TP53 evaluation in our study.

In summary, we have shown the utility of the orthotopic xenograft model by characterizing the largest panel of HNSCC cell lines to date. Our results have corroborated clinical reports showing that disruptive TP53 mutations correlate with aggressive tumor characteristics. Furthermore, we have proposed a new approach to classify TP53 mutations by combining TP53 sequence information with p53 protein expression levels. These results suggest that disruptive TP53 mutations may play a key role in the tumor progression and cervical lymph node metastatic potential of HNSCC. Our work should help facilitate further studies examining the cellular and molecular mechanisms of TP53 mutations in HNSCC tumor progression, metastasis, and response to therapy.

No potential conflicts of interest were disclosed.

The authors thank Dr. Karen Muller for her critical editorial review of the manuscript.

This work was supported by the University of Texas MD Anderson Cancer Center PANTHEON program, NIH Specialized Program of Research Excellence Grant P50CA097007, RO1 DE14613 NIH, National Research Science Award Institutional Research Training Grant T32CA60374, and NIH Cancer Center Support Grant CA016672 (CCSG).

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.