Patterns of global histone modifications have been suggested to be predictors of clinical outcome in many cancers. However, the role of global histone modification patterns in oral squamous cell carcinoma (OSCC) is unclear.
Patterns of global histone modifications have been suggested to be predictors of clinical outcome in many cancers. However, the role of global histone modification patterns in oral squamous cell carcinoma (OSCC) is unclear.
A retrospective clinicopathologic analysis was undertaken of 186 patients with oral squamous cell carcinoma who received complete ablative surgical treatment. Tissue arrays were made from those paraffin-embedded OSCC samples and examined by immunohistochemistry for histone 3 lysine 4 acetylation (H3K4ac), histone 3 lysine 18 acetylation (H3K18ac), histone 3 lysine 4 trimethylation (H3K4me3), histone 3 lysine 9 trimethylation (H3K9me3), and histone 3 lysine 27 trimethylation (H3K27me3).
A low level of H3K4ac and a high level of H3K27me3 were associated with advanced T status, N status, tumor stage, and perineural invasion. They were also correlated with cancer-specific survival (CSS) and disease-free survival (DFS). The 5-year CSS and DFS in H3K4aclow vs. H3K4achigh were 74.8% versus 92.5% (P = .010), and 51.4% versus 76.2% (P = .001), respectively. The 5-year CSS and DFS in H3K27me3low versus H3K27me3high were 94.7% versus 62.3% (P < .001) and 76.4% versus 32.3% (P < .001), respectively. We also found improved prediction for DFS after combining the H3K4aclow and H3K27me3high profiles and comparing the scores with the other modification patterns (P < .0001).
This research demonstrates the potential prognostic utility of global histone modification analysis for OSCC. Cancer 2013;119:4259–4267. © 2013 American Cancer Society.
Oral squamous cell carcinoma (OSCC) is one of the most distressing diseases worldwide and is characterized by high local-regional recurrence and poor long-term survival rates. The most decisive factors affecting the prognosis of patients with OSCC are the nodal status and the stage of the disease. Despite advancements made in the field of oral cancer early detection and multimodality treatment according to different clinicopathologic prognosticators, there has only been minor improvement in the response rate to current therapeutic strategies, particularly for tumors diagnosed in advanced stages. There are likely substantial prognostic factors at the molecular level that contribute to the biological behavior of the tumors.
Cancer is now considered a disease of genetic and epigenetic alterations. Evidence has shown that the accumulation of a wide range of both genetic and epigenetic alterations in a multistep process promotes tumorigenesis and cancer progression. Genetic alterations are irreversible changes in a DNA sequence that can lead to either oncogene activation or tumor-suppressor gene inactivation. In contrast, epigenetic changes are heritable and potentially reversible modifications in gene expression that occur without alterations to the DNA sequence. In the past decade, epigenetic alterations, which include DNA methylation, the modification of histones, chromatin remodeling, and microRNA expression, have gained considerable attention regarding cancer development and progression. By far the most common and best-studied epigenetic modification is aberrant DNA methylation. However, histone modification also plays a critical role in all chromatin-mediated processes, including transcription, replication, and repair. Histone modification can cause structural changes to chromatin, thus altering genomic function, and recent literature has described histone modifications in cancer biology and their use in the prediction of cancer prognosis and patient survival. Cellular levels of histone modifications may also predict responses to certain chemotherapeutic agents, serving as predictive biomarkers that could be used to make clinical decisions on the choice and course of therapy.
Histone residues are subject to a variety of modifications, such as methylation, acetylation, phosphorylation, ubiquitination, and sumoylation. Unlike other modifications, the methylation of lysines (K) can involve mono-, di-, or trimethylation. Both acetylation (ac) and methylation (me) of lysines can activate or inactivate chromatin, and, subsequently, the transcriptional status of genes. Patterns of global histone modifications have been recently suggested to be predictors of clinical outcome in many cancers.[7, 9, 10] To date, there is no known investigation on the prognostic significance of global histone modifications in OSCC. In this study, we examined the relative levels of 5 modified histones, that is, histone 3 lysine 4 acetylation (H3K4ac), histone 3 lysine 18 acetylation (H3K18ac), histone 3 lysine 4 trimethylation (H3K4me3), histone 3 lysine 9 trimethylation (H3K9me3), and histone 3 lysine 27 trimethylation (H3K27me3), by investigating lysine acetylation and lysine trimethylation using immunohistochemistry. The association of the patterns of the selected histone modifications, clinicopathologic variables, and patient outcomes was also assessed individually and combinatorially.
Under institutional review board approval, 186 archival paraffin-embedded surgical specimen blocks were obtained retrospectively from the Department of Pathology, Taipei Veterans General Hospital. These specimens were acquired from consecutive patients who had complete ablative surgery of oral squamous cell carcinoma (OSCC) with curative intent from January 2004 to January 2006. OSCC patients with evidence of distant metastases were excluded. Patients who had OSCC of an advanced stage or with adverse prognostic features, such as lymphovascular permeation (LVP), perineural invasion (PNI), or tumor emboli, were given either standard postoperative radiotherapy or concurrent chemoradiotherapy, according to our treatment protocol. From a review of medical records, clinical information was collected, including recurrence, survival, and follow-up data. The American Joint Committee on Cancer system (seventh edition) was applied for staging the tumors. The median follow-up period was 48.4 months (range, 2 to 90 months).
OSCC tissue arrays were constructed as described previously. The array samples were deparaffinized in xylene and rehydrated with serial dilutions of alcohol. The treated sections were then placed in a citrate buffer (pH 6.0) and heated in a microwave for two 5-minute sessions. The samples were then incubated with a primary antibody (1:100 dilution) overnight at 4°C. The primary antibodies used in the experiments included H3K4ac (Cat. No. 07-539; Millipore Corporation, Billerica, MA), H3K18ac (Cat. No. 39587; Active Motif, Inc., Carlsbad, CA), H3K4me3 (Cat. No. 9751; Cell Signaling Technology, Inc., Danvers, MA), H3K9me3 (ab8898; Abcam plc, Cambridge, MA), and H3K27me3 (Cat. No. 07-449; Millipore Corporation, Billerica, MA). The conventional streptavidin peroxidase method (Super Sensitive; BioGenex Laboratories Inc., Fremont, CA) was performed to develop the signal, and the tissues were counterstained with hematoxylin. Immunohistochemistry (IHC) for histone H3 (GTX122148; GeneTex, Inc., Irvine, CA) was performed to normalize the levels of histone marks, and all the samples were strongly positively stained of histone H3. p16INK4A IHC (sc-81157; Santa Cruz Biotechnology, Inc., Dallas, TX) was performed in an additional 80 cases to evaluate the correlation between p16INK4A expression and histone marks. Negative controls were obtained by excluding the primary antibody.
The staining results were examined by 2 observers masked to the patients' clinical information. Another reading by a third observer was needed to reach a consensus when there was a significant discrepancy between the initial readings. After excluding tissue array cores that were lost or fragmented, the cases available for scoring were as follows: H3K4ac, 171; H3K18ac, 152; H3K4me3, 145; H3K9me3, 158; and H3K27me3, 160. Only intermediate or strong nuclear staining was considered positive. The percentage of positive nuclear staining in tumor cells was calculated at 200× magnification using a light microscope and was scaled as follows: scale 0, none or less than 20% of the tumor cells were positive; scale 1, 20%-50% of the tumor cells were positive; scale 2, more than 50% of the tumor cells were positive. To obtain correlation with the patient's clinicopathologic features and survival, the expression categories were dichotomized into low (scale 0) and high (scales 1 and 2) grades.
The data were analyzed using the SPSS (Statistical Package for the Social Sciences) statistical program version 17 (SPSS, Inc., Chicago, IL). The χ2 test or Fischer's exact test was used to correlate categorical clinicopathologic variables and the histone modification status of the tumor cells. To facilitate the statistical analysis, we classified the patients into different age groups (≤55 years vs >55 years) as well as groups based on tumor (T) status (T1 + T2 vs T3 + T4) and nodal (N) status (N0 vs N+). The cumulative survival distribution was estimated using the Kaplan-Meier analysis technique. A comparison of the survival curves between the different groups was performed using the log-rank test. Cancer-specific survival (CSS) was defined as the period between the pathologic OSCC diagnosis and the time of death from OSCC, and disease-free survival (DFS) was defined as the time from pathologic OSCC diagnosis to locoregional recurrence or distant metastasis of the tumor. A multivariate analysis was performed according to the Cox proportional hazards regression model with a 95% confidence interval (CI) for various clinical parameters. In all the analyses, P < .05 was considered statistically significant.
There were 158 male and 28 female patients with a median age of 53.8 years (range, 22 to 88 years). Their clinicopathologic features are summarized in Table 1. According to TNM staging, 32 cases were stage I, 75 were stage II, 34 were stage III, and 45 were stage IV. Lymph node metastasis was identified in 49 patients (26.3%), and 137 patients (73.7%) were free of nodal metastases. The anatomical locations of OSCC involvement included the buccal mucosa (37.6%), tongue (43.0%), gingiva (11.3%), lip (4.3%), mouth floor (2.2%), and palate (1.6%). Among those patients for whom pathologic features were available, 7.0% had lymphovascular permeation, and 34.4% had perineural invasion.
|Lymphovascular permeation (LVP)|
|Perineural invasion (PNI)|
Five histone markers (H3K4ac, H3K18ac, H3K4me3, H3K9me3, and H3K27me3) were investigated by immunohistochemical staining, and the representative results for each histone modification are shown in Figure 1. The staining scale distributions of the histone markers plotted against their frequencies in different stages of OSCC are shown in Figure 2. For H3K18ac and H3K27me3, early-stage cases tended to distribute in low-grade staining (scale 0). In H3K27me3, stage IV cases tended to distribute predominantly at scale 2. After statistical analysis, we found the expression of H3K18ac and H3K27me3 was positively correlated with OSCC stage (Table 2). Conversely, H3K4ac was negatively correlated with OSCC stage, as stages III and IV predominantly distributed with low-staining scale. H3K4me3 and H3K9me3 levels were similar in all stages, with nearly even staining graded in all the stages. No statistical correlation was found between histone modification and stage for H3K4me3 and H3K9me3.
|Clinical Characteristics||H3K4ac (n = 171)||H3K18ac (n = 152)||H3K4me3 (n = 145)||H3K9me3 (n = 158)||H3K27me3 (n = 160)|
|≤55||76 (67.9)||36 (32.1)||.371||64 (64.0)||36 (36.0)||.323||65 (67.0)||32 (33.0)||.027a||62 (58.5)||44 (41.5)||.032a||56 (53.8)||48 (46.2)||.689|
|>55||36 (61.0)||23 (39.0)||29 (55.8)||23 (44.2)||23 (47.9)||25 (52.1)||21 (40.0)||31 (59.6)||32 (57.1)||24 (42.9)|
|Male||93 (64.1)||52 (35.9)||.377||78 (61.4)||49 (38.6)||.894||70 (57.9)||51 (42.1)||.116||72 (52.9)||64 (47.1)||.789||76 (55.1)||62 (44.9)||.963|
|Female||19 (73.1)||7 (26.9)||15 (60.0)||10 (40.0)||18 (75.0)||6 (25.0)||11 (50.0)||11 (50.0)||12 (54.5)||10 (45.5)|
|T1-T2||68 (56.2)||53 (43.8)||< .001a||69 (69.7)||30 (30.3)||.003a||60 (65.2)||32 (34.8)||.141||57 (50.9)||55 (49.1)||.520||68 (61.3)||43 (38.7)||.031a|
|T3-T4||44 (88.0)||6 (12.0)||24 (45.3)||29 (54.7)||28 (52.8)||25 (47.2)||26 (56.6)||20 (43.4)||21 (42.9)||28 (57.1)|
|N0||73 (58.9)||51 (41.1)||.003a||66 (63.5)||38 (36.5)||.396||61 (59.2)||42 (40.8)||.571||60 (50.4)||59 (49.6)||.353||72 (61.0)||46 (39.0)||.01a|
|N+||39 (83.0)||8 (17.0)||27 (56.2)||21 (43.8)||27 (64.3)||15 (35.7)||23 (59.0)||16 (14.0)||16 (38.1)||26 (61.9)|
|I||8 (29.6)||19 (70.4)||< .001a||8 (72.7)||3 (27.3)||.003a||6 (50.0)||6 (50.0)||.340||13 (48.1)||14 (51.9)||.721||17 (73.9)||6 (26.1)||.002a|
|II||42 (59.2)||29 (40.8)||49 (75.4)||16 (24.6)||43 (68.3)||20 (31.7)||33 (50.0)||33 (50.0)||44 (65.7)||23 (34.3)|
|III||26 (76.5)||8 (23.5)||18 (54.5)||15 (45.5)||17 (60.7)||11 (39.3)||19 (61.3)||12 (38.7)||14 (43.8)||18 (56.3)|
|IV||36 (92.3)||3 (7.7)||18 (41.9)||25 (58.1)||22 (52.4)||20 (47.6)||18 (52.9)||16 (47.1)||13 (34.2)||25 (65.8)|
|Pathological Characteristics||H3K4ac (n = 126)||H3K18ac (n = 111)||H3K4me3 (n = 105)||H3K9me3 (n = 118)||H3K27me3 (n = 119)|
|Negative||67 (57.8)||49 (42.2)||.200||56 (56.6)||43 (43.4)||.327||55 (59.1)||38 (40.9)||.090||56 (52.3)||51 (47.7)||.664||58 (52.7)||52 (47.3)||.095|
|Positive||8 (80.0)||2 (20.0)||5 (41.7)||7 (58.3)||4 (33.3)||8 (66.7)||5 (45.5)||6 (54.5)||22 (22.2)||7 (77.8)|
|H3K4ac (n = 129)||H3K18ac (n = 112)||H3K4me3 (n = 106)||H3K9me3 (n = 120)||H3K27me3 (n = 120)|
|Negative||32 (46.4)||37 (53.6)||< .001a||31 (58.5)||22 (41.5)||.651||32 (64.0)||18 (36.0)||.102||37 (56.1)||29 (43.9)||.287||42 (63.6)||24 (36.4)||.001|
|Positive||46 (76.7)||14 (23.3)||32 (54.2)||27 (45.8)||27 (48.2)||29 (51.8)||25 (46.3)||29 (53.7)||18 (33.3)||36 (66.7)|
The patterns of the different histone modifications were correlated with different clinical-pathologic parameters, including age, sex, T status, N status, lymphovascular permeation, and perineural invasion (Table 2). We observed that only H3K4ac, H3K18ac, and H3K27me3 had significant correlations with clinicopathologic factors. The low grade of H3K4ac was significantly associated with T status (P < .001), nodal invasion (N+; P = .003), advanced tumor stage (P < .001), and PNI (P < .001). H3K18ac demonstrated a positive correlation with T status (P = .003) and tumor stage (P = .003), whereas H3K27me3 had a significantly positive correlation with T status (P = .031), N status (P = .01), tumor stage (P = .002), and PNI (P = .001).
We also analyzed the correlation of the expression of different histone marks. The result showed that there was correlation of the expression of H3K4me3 and H3K18ac, H3K9me3 and H3K4ac, H3K9me3 and H3K4me3, and H3K27me3 and H3K18ac. Furthermore, we also analyzed the correlation of p16INK4A with 2 major histone marks found in our study (H3K4ac and H3K27me3) in an additional 80 cases. The results showed both H3K27me3 and H3K4ac had no correlation with p16 INK4A (P = .644 and P = .713, respectively; detailed analysis not shown).
Survival analysis of the different clinicopathologic variables and histone modification patterns was performed using the log-rank test (Table 3). For clinicopathologic variables, T status, N status, tumor stage, LVP, and PNI were all significantly correlated with 5-year CSS and DFS. For histone modification patterns, only H3K4ac and H3K27me3 showed a significant correlation with survival outcome. The low-scoring H3K4ac and high-scoring H3K27me3 groups exhibited poorer survival than the high-scoring H3K4ac and low-scoring H3K27me3 groups. Five-year CSS/DFS was 92.5%/76.5% in high-scoring H3K4ac and 74.8%/51.4% in low-scoring H3K4ac (P = .010), whereas the 5-year CSS/DFS was 94.7%/76.4% in low-scoring H3K27me3 and 62.3%/32.3% in high-scoring H3K27me3. Other histone markers, including H3K18ac, which was found to be significantly correlated with T status and tumor stage, did not show a correlation with patient survival.
|5-Year Survival (%)||P||5-Year Survival (%)||P|
|T1-T2||88.5||< .001a||64.2||< .001a|
|N0||88.9||< .001a||66.5||< .001a|
|I||92.9||< .001a||68.7||< .001a|
|Negative||84.6||< .001a||64.9||< .001a|
|Negative||93.0||< .001a||74.8||< .001a|
|Low||94.7||< .001a||76.4||< .001a|
Multivariate survival analysis was performed by considering factors that affected patient survival, and this analysis demonstrated that both H3K4ac and H3K27me3 expression had independent prognostic influences on survival (Table 4).
|Variable||HR||(95% CI)||P||HR||(95% CI)||P|
We further analyzed whether a combination of H3K4ac and H3K27me3 was able to predict the patients' outcome more effectively. By combining the H3K4ac (low expression as a score of 1, high expression as a score of 0) and H3K27me3 (high expression as a score of 1, low expression as a score of 0) profiles, we generated a sum of the scores that enabled us to stratify the patients into 3 groups. The results showed the cumulative effect of low-scoring H3K4ac and high-scoring H3K27me3 on the DFS of OSCC patients. Patients with this histone modification pattern had the shortest time to tumor recurrence (5-year DFS, 26.0%; median DFS time, 20.17 months; Fig. 3). Furthermore, the H3K4aclow H3K27me3high pattern was also able to predict OSCC recurrence and survival independently (DFS: hazard ratio, 35.867; P = .001; CSS: hazard ratio, 13.703; P = .030; Table 5).
|HR||(95% CI)||P||HR||(95% CI)||P|
Recent studies have supported the concept that histone modifications can be used as prognostic markers. Seligson et al demonstrated that decreasing levels of H3K4me2 and H3K18ac predict the recurrence of prostate cancer. Several other histone markers were also found to be associated with prostate cancer prognosis, such as H3K4me, H3K9me2, H3K9me3, H3K27me3, and H3- and H4-pan acetylation (H3Ac, H4Ac).[13, 14] Poor clinical outcome in relation to global histone modifications was also demonstrated in different cancers, including breast cancer (moderate to low levels of H3K18ac, H3K9ac, H4K12ac, H3K4me2, H4K20me3, H3K9me3, and H4R3me2),[15, 16] lung cancer (low levels of H3K4me2, H4K20me3, and H3K18ac and high levels of H3K9ac),[9, 17, 18] gastric cancer (high levels of H3K9me3), esophageal cancer (high levels of H3K18ac, H3K27me3, and H4R3me2),[20, 21] kidney cancer (low levels of H3K4me1-3, H3K9me1, H3K27me1-3, and H3K18ac),[22-25] and liver cancer (high expression of H3K4me3 and H3K27me3).[26, 27] However, the significance of histone modification patterns in OSCC has not yet been reported.
In this study, we demonstrated that global histone modifications, including H3K4ac, H3K18ac, and H3K27me3, play a major role in OSCC progression and patient prognosis. The positive correlation between H3K27me3 and clinicopathologic factors (ie, T status, N status, tumor stage, and perineural invasion) as well as survival was, in general, consistent with those results reported in prostate, esophageal, and liver cancers. Unlike histone lysine acetylation, which is typically associated with the transcriptional activation of genes, the functional effect of histone lysine methylation is varied. Whether specific genes are silenced, inactivated, or activated depends on the number of methylated residues and their position within the histone tail.
Our previous study demonstrated a poorer prognosis and a higher risk of tumor recurrence in OSCCs that lack p16INK4A expression. Other studies have shown that the INK4A-ARF locus, which encodes the key regulators of cell senescence, that is, tumor suppressors p16INK4A and p14ARF, is epigenetically silenced by H3K27me3. The enrichment of H3K27me3 was shown to be an early event in the silencing of p16INK4A during tumor development. Taking the aforementioned studies together, we speculated that high levels of H3K27me3, through repressing p16INK4A, might possibly lead to an aggressive phenotype of cancer. However, there was no correlation between H3K27me3 and p16INK4A in the current study. Furthermore, H3K4ac, which is another major histone mark found in the current study, was also not associated with p16INK4A status. Hence, we consider that both H3K27me3 and H3K4ac could be prognostic factors independent of p16INK4A status. In addition, the experiment conducted by Abbosh et al showed that the removal of H3K27 methylation reversed the platinum-resistant phenotype in cancer cells. This evidence strengthens that H3K27me3 may be a promising therapeutic target for OSCC.
Another finding in the current study was that low levels of H3K4ac were significantly associated with advanced cancer status and poor survival. Several histone deacetylases (HDACs), which remove the acetyl groups from histones, have been discovered. The potent antitumor effects of HDAC inhibitors were first demonstrated in cutaneous T-cell lymphoma and were approved for therapeutic use in other hematologic malignancies.[31-33] We postulated that low levels of H3K4ac were also associated with the overexpression and/or increased activity of HDACs. Thus, OSCC patients with lower levels of H3K4 acetylation with poor clinicopathologic characteristics and survival might benefit from HDAC inhibitor therapy. Additional studies are needed to validate the potential therapeutic use of HDAC inhibitors in OSCC treatment.
The prediction of patient outcome and a guide for OSCC treatment is based mainly on tumor stage, N status, and the existence of pathologic risk factors (ie, perineural invasion, lymphovascular permeation, and tumor emboli). In the current study, we demonstrated that epigenetic heterogeneity corresponded to OSCC prognosis. Certain histone modifications (low levels of H3K4ac, high levels of H3K18ac and H3K27me3) were associated with advanced T status, N status (except H3K18ac), tumor stage, and PNI (except H3K18ac). Unsurprisingly, both H3K4ac and H3K27me3 were correlated with CSS and DFS because of their significant relationship with multiple unfavorable clinicopathologic risk factors. In contrast, the levels of H3K18ac, which were only correlated with tumor size (T status) and stage, had no significant difference concerning CSS and DFS. Multivariate analysis, however, identified that only H3K27me3 level was an independent prognostic factor for CSS, whereas the levels of both H3K4ac and H3K27me3 were independent prognostic factors for DFS. We also found a cumulative effect and improved predictive power (P < .0001) after combining these histone scores and comparing the scores with other clinical or pathologic factors.
From the above results, our current research notably demonstrated the potential to utilize the presence of global histone modifications in predicting OSCC patient outcome. There has been no known report studying global histone modifications in oral cancers, so further study of other histone markers with a larger patient cohort is warranted.
This work was supported by the National Science Council (NSC101-2321-B-010-007 to M.H.Y.), National Health Research Institutes (NHRI-EX100-10037BI; to Muh-Hwa Yang), Taipei Veterans General Hospital (V102C-036, V102E8-002), a grant from the Ministry of Education, Aim for the Top University Plan, and a grant from the Department of Health, Center of Excellence for Cancer Research (DOH101-TD-C-111-007; to Muh-Hwa Yang). This work was assisted in part by the Division of Experimental Surgery of the Department of Surgery, Taipei Veterans General Hospital.
The authors made no disclosures.