High-Risk TP53 Mutations Are Associated with Extranodal Extension in Oral Cavity Squamous Cell Carcinoma Free

Oral squamous cell carcinoma (OSCC) can be a devastating disease that affects nearly 30,000 patients each year in the United States alone (1). OSCC demonstrates a high propensity to metastasize to regional nodal basins which generally requires treatment escalation to include multimodality treatment strategies (2–4). Once OSCC has metastasized to lymph nodes patient survival is decreased by nearly 50%, and development of extranodal extension (ENE) is further associated with a negative impact on disease control and patient survival (2, 3, 5). Fifteen years ago, we described the clinical impact of ENE on survival in patients with OSCC and identified it as a prognostic marker for regional recurrence and distant metastasis (2). Since that time, we have witnessed significant improvements in the sensitivity and specificity of anatomic (computed tomography, CT; magnetic resonance imaging, MRI) and metabolic imaging (positron emission tomography, PET), in the delivery of radiotherapy, and an increased availability of conventional and targeted chemotherapeutic agents. In this article, we update our previous analysis for ENE impact on survival, utilizing a larger patient cohort.

Recent advances in next-generation sequencing (NGS) and multiplatform genomic and epigenetic characterization of solid tumors including OSCC have allowed us to better understand the molecular changes underlying OSCC development and progression (6, 7). In addition, this genomic information from OSCCs has led to the development of predictive biomarkers of the efficacy of radiation and/or chemotherapy in the adjuvant (postsurgical) setting (8, 9). Data from our group and others have confirmed the clear association of TP53 mutations with the development and progression of oral squamous cancer as well as SCC from other head and neck sites (HNSCC; refs. 6, 8–11). Moreover, we have demonstrated, in both patient cohorts and preclinical models of HNSCC, that the specific types of TP53 mutations can have a profound impact on tumor development, metastasis and response to treatment (8, 9, 12–15). These novel discoveries are critical to continued refinement in the selection of more effective treatment strategies. To this end, we have developed an evolutionarily based TP53 mutation grading system (named EAp53), which classifies TP53 mutations into low-risk missense, high-risk missense, and “other” (nonsense, splice site, frameshift, and insertion–deletion) mutations, and have demonstrated that HNSCC patients with low-risk TP53 mutations are associated with better treatment outcomes similar to those with WT TP53 when compared with those with high-risk or “other” mutations (8, 9). While pathologic nodal status and the presence of ENE are our most reliable prognostic biomarkers for patients with OSCC, these can only be determined in the postoperative setting from histopathologic analysis of the neck dissection specimen. Because an increasing number of patients with HNSCC are treated nonsurgically, prognostic biomarkers, which correlate with nodal status and presence of ENE, have become critically important. Therefore, in this study, we sought to identify a biological signature associated with ENE development in OSCC based on analysis of the TP53 mutational status in a biopsy of the primary tumor specimen.

In the current study, we aimed to address two translational questions. First, we sought to determine whether ENE is a poor prognostic indicator in the modern era of OSCC treatment. Second, we sought to identify a molecular signature associated with ENE development and aggressive clinical behavior in an effort to potentially triage patients more appropriately for multimodality therapy in the absence of information about ENE that is available only after surgical removal of regional lymph nodes. To answer these questions, we evaluated data from a large institutional OSCC patient cohort as well as OSCC data from The Cancer Genome Atlas (TCGA).

Three patient cohorts were utilized for the data analysis; all studies were approved by the Institutional Review Board and all studies were conducted in accordance with recognized ethical guidelines (Declaration of Helsinki, CIOMS, Belmont Report and U.S. Common Rule). Waiver of written informed consent was provided as part of the approval process by the Institutional Review Board as is common for retrospectively conducted analyses. A cohort of 238 patients from The University of Texas MD Anderson Cancer Center (UTMDACC) was utilized to confirm the impact of ENE on clinical outcomes. Inclusion criteria included OSCC diagnosis, primary surgical treatment (with or without adjuvant treatment) between July 2009 and December 2015. All patients underwent surgery including resection of primary tumor and simultaneous neck dissection. Exclusion criteria included incomplete treatment, treatment for recurrent disease, and preoperative treatment with systemic therapy. Clinical–pathological information; demographics, tumor characteristics, status of lymph nodes, treatment characteristics, disease recurrence, and survival were collected for all patients. Pathologic data for the status of lymph nodes included the presence or absence of ENE. ENE was defined as consisting of tumor extension outside the nodal capsule and into the surrounding soft tissues. The presence of tumor cells in the capsule of the node was not considered ENE. HPV status was not known for the UTMDACC cohort due to the lack of routine testing for HPV in oral cavity patients and the low incidence of HPV-positive OSCC in our patient population (6%; ref. 16).

We extracted clinical and molecular data from TCGA for 230 HPV-negative OSCC patients in order to confirm the impact of ENE on clinical outcomes. In addition, we mined genomic information of TCGA to discover the molecular markers relevant to ENE. These are obtained from FireBrowse.org website or TCGA (17). Inclusion criteria included OSCC diagnosis and primary surgical treatment (with or without adjuvant treatment). Exclusion criteria included the unavailability of either pathologic or genomic information and presence of HPV expression by RNAseq (17).

Finally, using the UTMDACC cohort, 17 OSCC patient specimens were used to examine differences in mutation status between primary tumors and lymph nodes with ENE. Inclusion criteria included OSCC diagnosis, primary surgical treatment (with or without adjuvant treatment) with simultaneous neck dissection, evidence of at least one pathologically metastatic lymph node with ENE and availability of both primary tumor and paired ENE-positive lymph node tissue.

Statistical analysis for the impact of ENE on disease-free survival (DFS) and overall survival (OS) was performed using Kaplan–Meier analysis (log-rank analysis for statistical significance). DFS time was defined as the interval between the date of the end of primary treatment and the date of the development of local, regional recurrence and distant metastasis after surgery. OS time was calculated from the date of initial examination to the date of death, to the date of last contact, or to the date of the 5-year follow-up. The median follow-up period was 28.4 months (range, 1.6–60 months). Moreover, statistical significance of individual clinical parameters was confirmed using multivariate analysis (IBM SPSS Statistics 22.0, IBM Inc, NY).

Two hundred thirty patients were divided into three groups: pathologically negative nodes (pN0), pathologically positive node without ENE (pN⁺/ENE⁻), and pathologically positive node with ENE (pN⁺/ENE⁺). Fisher exact test was conducted for 5,949 genes to compare mutation frequencies across these groups. The TP53 gene was also analyzed individually and mutations were characterized as “high-risk,” “low-risk,” or “other” based on the previously described, validated Evolutionary Action scoring system (EAp53; refs. 8, 9). Missense mutations were scored using the EAp53 from 0 to 100 with higher scores representing alterations calculated to be more deleterious to protein function. Wild-type TP53 sequences were scored as 0 due to assumed normal function. The threshold used to define mutations as “high-risk” was 75.

Seventeen OSCCs, identified from the UTMDACC tumor bank based on the abovementioned criteria, were examined by targeted NGS. Primary tumors and paired metastatic lymph nodes with ENE were reevaluated by a trained head and neck pathologist (Diana Bell) to confirm the histologic diagnosis and the presence of ENE. Genomic DNA was isolated from frozen primary tumors and formalin-fixed paraffin-embedded (FFPE) tissues of entire lymph node metastasis including the site of ENE. As a control (to remove germline mutations), we used genomic DNA from frozen blood samples. Ten nanograms of genomic DNA was used as input for target DNA library preparation using the Ion AmpliSeq Library Kit, and sequenced in the Ion PGM Sequencer platform. Mutations were called with a custom pipeline in the Ion Reporter software.

We generated a box plot to compare the overall gene expression among samples and the 3 groups listed above from the RNAseq data available within the head and neck TCGA cohort. Hierarchical clustering as well as principal component analysis (PCA) were used to evaluate sample quality and differences between samples. Genes with the expression value of less than 5 (corresponding to 31 reads) in all samples were filtered out. One-way ANOVA was applied on a gene-by-gene basis to test for difference among the 3 ENE groups. The Benjamini–Hochberg method was applied to the resulting P values, computed from test statistics, to adjust for multiple testing. Pairwise comparisons between different groups were done using Tukey honest significance (HSD) test with 95% family-wise confidence level.

Copy-number alterations were evaluated for each gene. Fisher exact test was conducted for each gene to compare copy number across the groups to test for association between copy number and the ENE groups. We generated three groups for each gene: deletion, normal, and amplification. Fisher exact test was then conducted for this version of copy-number values. Benjamini–Hochberg correction was applied to adjust for multiple testing. With the amplification peaks output by GISTIC (Genomic Identification of Significant Targets in Cancer; www.broadinstitute.org/cancer/cga/gistic), we obtained genes from each peak and retrieved their copy-number values for each patient sample. Then, we considered the genes from the same peak as a group and assigned a copy-number value to each patient. Samples were separated into two groups, a group without gains (−2, −1, 0) and a group with gains (1, 2), and compared the difference across the groups. Deletions were evaluated in a similar manner.

A total of 238 patients with OSCC were analyzed as part of the UTMDACC cohort. One hundred twenty-eight of the 238 patients (53.8%) had negative nodes (pN0), 40 (16.8%) had a pathologically positive node without ENE (pN⁺/ENE⁻), and 70 (29.4%) had a pathologically positive node with ENE (pN⁺/ENE⁺). The UTMDACC patient clinical–pathological information is summarized in Supplementary Table S1. A total of 230 OSCC patients' data were analyzed as part of the TCGA cohort. One hundred thirteen of the 230 patients (49.1%) were pN0, 62 (27.0%) were pN⁺/ENE⁻, and 55 (23.9%) were pN⁺/ENE⁺. TCGA patient clinical–pathological information is summarized in Supplementary Table S2.

To assess the impact of ENE on survival, the survival data of the patients were compared among the three groups, pN0, pN⁺/ENE⁻, and pN⁺/ENE⁺. Kaplan–Meier survival curves of DFS and OS for the UTMDACC cohort are presented in Fig. 1A and B, respectively. The 3-year and 5-year DFS rates are 76.6% and 71.9% for pN0, 70.2% and 62.4% for pN⁺/ENE⁻, and 49.7% and 45.9% for pN⁺/ENE⁺ patients. The 3-year and 5-year OS rates are 83.0% and 75.2% for pN0, 73.0% and 59.0% for pN⁺/ENE⁻, and 47.8% and 45.0% for pN⁺/ENE⁺ patients. These results demonstrate the adverse impact of pN⁺/ENE⁺ compared with pN0 and pN⁺/ENE⁻ on DFS (log-rank test, P < 0.0001 and P = 0.036, respectively) and OS rates (log-rank test, P < 0.0001 and P = 0.043, respectively). Multivariate Cox proportional hazards analysis, including clinical–pathological factors revealed that smoking and ENE were independent predictors of DFS [HR, 0.598 and 2.629; 95% confidence interval (CI) = 0.375–0.954 and 1.650–4.189, P = 0.031 and P < 0.0001, respectively; Supplementary Table S3). With regard to OS, pathologic T stage and presence of ENE showed a statistically significant prognostic value (HR, 1.981 and 2.652; 95% CI, 1.201–3.270 and 1.625–4.327, P = 0.007 and P < 0.0001, respectively; Supplementary Table S3). Therefore, the presence of ENE is the only significant independent predictor of both DFS and OS.

To validate the impact of ENE on survival, we analyzed data from the TCGA patient cohort. The 3-year and 5-year OS rates were 64.0%, 56.4% for pN0, 66.8%, 61.7% for pN⁺/ENE⁻, and both 13.9% for pN⁺/ENE⁺ patients, respectively (Fig. 2). We confirmed the adverse impact on OS of pN⁺/ENE⁺ compared with pN0 and pN⁺/ENE⁻ (log-rank test, P < 0.0001). Multivariate Cox proportional hazards analysis, including clinical–pathological factors and TP53 mutation status (WT/“low-risk” versus “high-risk”/ “other”), revealed that pathologic T stage, and ENE were independently correlated with OS (HR, 2.607 and 3.090; 95% CI = 1.558–4.362 and 1.982–4.819, P < 0.0001 and P < 0.0001, respectively; Supplementary Table S4). DFS rates were not analyzed for this cohort due to the limited recurrence details collected by TCGA.

Sequencing information for the TCGA OSCC tumors is summarized in Supplementary Tables S5 and S6. A total of 181 (78.7%) patients exhibited any mutations in the TP53 gene and 113 (49.1%) patients had missense mutations. Forty patients had more than one mutation in TP53 and 21 patients had the combination of missense mutations and other mutations. In these cases, the representative type of mutation for the each patient was defined using the missense mutation. We classified missense mutations into two groups: “low-risk” mutations and “high-risk” mutations with the previously described and validated Evolutionary Action scoring system (8, 9). Consequently, a total of 45 patients had “low-risk” mutations, 68 patients had “high-risk” mutations, and 68 patients had “other” mutations. Eight patients whose tumors had both “low-risk” mutations and “high-risk” mutations were classified as having “high-risk” mutations.

We divided patients into 3 groups: pN0, pN⁺/ENE⁻, and pN⁺/ENE⁺; 49 out of 55 (89.1%) patients with ENE exhibited mutations in TP53. The frequency of TP53 mutation in pN⁺/ENE⁺ (89.1%) was higher compared with both pN0 (72.4%) and pN⁺/ENE⁻ (80.6%). Moreover, when we classified TP53 status into 4 groups (WT, “low-risk” mutations, “high-risk” mutations, and “other” mutations), “high-risk” and “other” mutations were overrepresented in pN⁺/ENE⁺ tumors. Among pN⁺/ENE⁺ patients with mutations, 24 out of 49 (49.0%) had “high-risk” and 20 (40.8%) had “other” mutations (Supplementary Table S5; Fig. 3). The EAp53 scoring system only calculates a score for missense mutations; therefore, nonmissense mutations are classified as “other.”

Chi-square analysis across the 3 patient groups showed TP53 “high-risk” mutations were significantly correlated with pN⁺/ENE⁺ group (P = 0.014). When we classified TP53 status into 2 groups: WT/“low-risk” mutations and “high-risk”/“other” mutations, a further statistical association was seen between “high-risk”/“other” mutations and pN⁺/ENE⁺ group (χ² test, P = 0.001). Multivariate logistic regression analysis, including clinical–pathologic factors, TP53 mutation status, showed that smoking habit, pathologic T stage, and TP53 mutation status (WT/“low-risk” versus “high-risk”/“other”) were independently significant factors of ENE (odds ratio = 2.757, 3.528, and 3.683, 95% CI = 1.172–6.484, 1.655–7.519, and 1.725–7.863, P = 0.020, 0.001, and 0.001, respectively, hit ratio = 75.9%; Table 1).

Kaplan–Meier survival curves for OS according to TP53 mutation type are presented in Supplementary Fig. S1. The 3-year and 5-year OS rates were both 67.7% for patients with WT, both 54.1% for those with “low-risk,” 52.8%, 46.2% for those with “high-risk,” and 45.4%, 36.1% for those with “other” mutations, respectively (log-rank test, WT versus “other” mutations: P = 0.036). When we classified TP53 status into 2 groups: WT/“low-risk” mutations and “high-risk”/“other” mutations, the 3-year and 5-year OS rates were both 60.0% for patients with wild type/“low-risk” mutations, and 48.9% and 40.9% for those with “high-risk”/“other” mutations, respectively (log-rank test. P = NS).

The sequencing results for OSCC primary tumors and paired lymph node metastasis of 17 UTMDACC patients with ENE are shown in Table 2 and Supplementary Table S7. We found that 12 (70.6%) out of 17 patients had a TP53 mutation in the primary tumor (PT) and/or lymph node metastasis with ENE (LN). Eight patients (66.7%, 8/12) had concordant TP53 mutations in both PT and paired LN. Another 4 patients (33.3%, 4/12) were shown to have discordant mutations between the PT and paired LN. TP53 mutation scores were equally distributed when comparing all PT and LN samples.

The TCGA data were mined for mutation frequency across all OSCC specimens that met inclusion criteria. The most commonly mutated genes were FAT1 (28%), CDKN2A (27%), NOTCH1 (21%), CASP8 (18%), and PIK3CA (16%). An unbiased analysis of mutational frequency and ENE status did not identify any mutated genes other than TP53 significantly associated with ENE (no gene with FDR of <0.05).

Copy-number analysis was performed to determine if any copy-number alterations were associated with ENE. In order to improve the statistical power, only significant GISTIC regions from the full HNSC TCGA analysis were analyzed (Firebrowse.org). We found that the regions of 11q23.1 and 19p13.3 demonstrated an association between ENE and copy-number loss (adjusted P = 0.0246 and P = 0.02509, respectively; Supplementary Table S8). 11q23.1 contains the gene SDHD and is lost in 55% of ENE⁺ patients compared with 24% of pN0 and 42% of pN⁺/ENE⁻. 19p13.3 contains LKB1/STK11 and is lost in 55% of ENE⁺ cases compared with 26% of pN0 and 35% of pN⁺/ENE⁻. Global gene expression analysis identified 9 genes differentially expressed between the ENE groups (Supplementary Table S9). Hierarchical clustering of the samples based on the expression of these genes identified a cluster enriched for LN metastasis (94% vs. 44%, 29/31 vs. 88/199), but not ENE specifically (Supplementary Fig. S2). No pathways were enriched among these genes.

Despite significant technological and scientific advances over the last 3 decades, survival for patients with advanced OSCC remains poor (Figs. 1 and 2). The data summarized here, confirms that ENE is an important biomarker of OSCC prognosis. Given that these data match quite precisely findings from our previous institutional cohort despite lack of any patient overlap, we consider ENE to be one of the most reliable prognostic markers of OSCC clinical outcomes (2). Our institutional findings are further strengthened by confirmation of ENE importance in the TCGA patient cohort. In fact, clinical outcomes for pN⁺ENE⁺ OSCC patients in the TCGA cohort are even worse, with 2-year survival at approximately 25%. Whether this represents simply a statistical anomaly, selection bias in patients in whom samples were collected for the TCGA, or variation of clinical treatments for advanced OSCC at the many centers that provided specimens is unclear and will need to be further investigated. Inclusion of HPV-positive patients in the TCGA cohort did not alter the overall results and/or conclusions.

As OSCC is primarily treated as a surgical disease, ascertaining pN⁺ and ENE status will continue to represent an important component of standard of care postsurgical adjuvant treatment paradigms. In light of the recent publication by D'Cruz and colleagues on the diagnostic and clinical impact of elective neck dissection for OSCC tumors compared with therapeutic node dissection, it is likely that the majority of patients with OSCC will continue to undergo surgical management, which can reliably generate pN⁺ and ENE status (18). In contrast to OSCC, other subsites (i.e., oropharynx, larynx) within the head and neck are increasingly treated using nonsurgical treatment algorithms (19–23). As a result, pN⁺ and ENE status and their impact on clinical prognosis cannot be reliably and consistently ascertained. Therefore, there is a critical need to define the biological signature associated with pN⁺ and ENE in order to incorporate this prognostic information into nonsurgical treatment algorithms

The impact of TP53 mutation on HNSCC biology and response to treatment has been thoroughly documented by our group and others using both preclinical disease models and retrospective patient cohorts (8, 9, 11, 13, 15, 24–26). TP53 mutations can behave in a heterogeneous manner with response to both tumorigenesis and treatment response (8, 9, 14). Most recently, we developed an evolutionarily based scoring system to classify TP53 mutations as: (i) “low-risk” missense mutations, (ii) “high-risk” missense mutations, and (iii) “other” TP53 mutations (e.g., nonsense, splice site, insertion–deletion). Although this classification overlaps some of the older systems (i.e., disruptive vs. nondisruptive), it has been shown to correlate with both prognosis and treatment response in both preclinical models and patient cohorts (8, 9, 14). Here, we demonstrate that “high-risk” TP53 mutations are also associated with pN⁺ENE⁺ status in 2 distinct patient cohorts. Whether “high-risk” TP53 mutations actually drive pN⁺ENE⁺ development cannot be addressed via retrospective analysis and will require additional investigation using existing preclinical disease models. However, if both ENE development and treatment resistance are in fact partially driven by “high-risk” TP53 mutations, it is possible that “high-risk” TP53 mutations may represent a robust biomarker of ENE development and adverse clinical outcomes.

In addition, “other” mutations are likely to be functionally inactivating and could be given a score of 100. In support of this, they performed similarly to the “high-risk” mutations in the data presented here. However, “high-risk” mutations may not be strictly loss-of-function, rather, many of them have been shown to demonstrate gain-of-function properties, and there can be differences between the “high-risk” and “other” groups depending on the patient cohort or in vitro phenotype being studied (8, 9, 11, 13–15). For this reason, the “other” mutations should be considered a separate group and not automatically combined with the “high-risk” group. Additionally, as “other” mutations are loss-of-function, they are considered recessive to “low-risk” or “high-risk” mutations in patients harboring the combination of missense mutations and “other” mutations.

Although Wang and colleagues reported on a gene signature associated with ENE development in OSCC, our analysis did not independently identify a gene expression profile that consistently associates with ENE in the TCGA cohort (27). The institutional patient cohort analysis was primarily focused on TP53 mutations and therefore did not provide gene expression information. The lack of complete concordance between TP53 mutational status of the tumor and corresponding metastatic lymph nodes is likely related to tumor evolution and heterogeneity. It has been shown that different regions of primary tumors can have diverse mutational profiles, likely due to divergent evolution from a single cell into multiple clonal populations in the primary tumor (28). Sampling of the primary tumor away from a clone represented in the metastatic population would account for a discordance as would clonal expansion after metastasis of a clone carrying the mutation found in the lymph node not being found in the primary tumor. A more thorough study, focusing on just this endpoint, should be performed with additional sampling from the primary and nodal regions in order to get a better understanding for the heterogeneity present and whether this will affect the sensitivity of genomic biomarker testing.

Identification of copy-number changes in LKB1 and SDHD generates intriguing hypotheses regarding a role for metabolism in ENE development, which will require dedicated investigation in future studies. We have previously shown that TP53 mutations generate a profound effect on HNSCC tumor metabolism; the presence of high-risk TP53 mutations can potentially link ENE with alterations in the tumor metabolic profile. This gives rise to a metabolic phenotype which supports a more aggressive biological phenotype, but also uncovers potential targeting strategies based on lack of metabolic flexibility. Given that our group has previously shown that TP53-mutant tumor cells exhibit decreased metabolic flexibility and enhanced susceptibility to metabolic inhibition, this may represent a viable therapeutic strategy (12, 14, 29, 30).

No potential conflicts of interest were disclosed.

Conception and design: V.C. Sandulache, J.N. Myers, C.R. Pickering

Development of methodology: V.C. Sandulache, S. Trivedi, M. Zhao, C.R. Pickering

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V.C. Sandulache, C. Michikawa, D. Bell, S. Trivedi, M. Zhao, S.A. Jasser, C.R. Pickering

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V.C. Sandulache, C. Michikawa, P. Kataria, F.O. Gleber-Netto, X. Rao, J. Wang, J.N. Myers, C.R. Pickering

Writing, review, and/or revision of the manuscript: V.C. Sandulache, C. Michikawa, J.N. Myers, C.R. Pickering

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Michikawa, P. Kataria, S. Trivedi, S.A. Jasser, J.N. Myers, C.R. Pickering

Study supervision: J.N. Myers, C.R. Pickering

This work was supported by National Institute of Dental and Craniofacial Research 5R01 DE014613 12 (J.N. Myers).

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.