RUNX3/H3K27me3 Co-Expression Defines a Better Prognosis in Surgically Resected Stage I and Postoperative Chemotherapy-Naive Non-Small-Cell Lung Cancer

The purpose of this study is to investigate the significance of RUNX3/H3K27me3 co-expression in surgically resected non-small-cell lung cancer (NSCLC) patients. Using tissue microarray (TMA), immunohistochemistry, fluorescent double immunostaining, and western blotting, 208 NSCLC and 5 benign pulmonary patients were studied of their expression of runt-related transcription factor 3 (RUNX3), trimethylated histone H3 at lysine 27 (H3K27me3), enhancer of zeste homolog 2 (EZH2), and Ki-67. Apoptotic index in cancerous tissue was evaluated via TdT-mediated dUTP-biotin nick end labeling (TUNEL). The correlation between clinicopathologic parameters and overall survival was determined by Cox regression and Kaplan–Meier survival estimates and log-rank test. GEPIA and KM plotter were used for validation of some survival analyses. As a result, together with other regular prognostic factors, RUNX3/H3K27me3 co-expression was found to be closely correlated with better prognosis in either pTNM-I or POCT-naive NSCLC patients, which might partially result from a higher cancerous apoptotic index. In conclusion, RUNX3/H3K27me3 co-expression defined some specific NSCLC population with better prognosis and longer OS and could probably be used as a biomarker in the prediction of better postoperative outcomes.


Introduction
Lung cancer is still the leading cause of cancer-related deaths in both sexes in China and worldwide [1,2], and non-smallcell lung cancer (NSCLC) accounts for nearly 85% of all cases. For decades, diagnosis of lung cancer occurred so late that nearly two-thirds of patients had lost the opportunity for radical resection. Recently, with the prevalence of latest diagnostic techniques like low-dose computed tomography (LDCT) or artificial intelligence (AI) and use of liquid biopsy in lung cancer screening, more and more early-stage cases had been diagnosed and received timely surgical resection [3][4][5][6]. Although many surgically resected patients shared identical histology and pathologic stage, their prognosis varied a lot, with nearly one-third of pTNM-stage I patients suffering from early postoperative relapse or distant metastasis [7,8]. Based on this context, molecular staging with sensitive biomarkers on the purpose of accurate prediction of different likelihood of early relapse or metastasis is necessary. One major discrepancy underlying the different prognoses for patients with the same pathologic stage is their different epigenetic status, mainly recognized as histone modifications or DNA methylation involved in the regulation of various oncogenes or tumor suppressor genes (TSGs), facilitating tumorigenesis and/or progression of different types of human malignancies. One such histone modification, trimethylated histone H3 at lysine 27 (H3K27me3), catalyzed by enhancer of zeste homolog 2 (EZH2) and functioning as a mark for de novo DNA methylation in cancer cells by recruitment of DNA methyltransferases (DNMTs), is usually recognized as a transcription-suppressive histone modification and proven to be involved in cell cycle progression and proliferation regulation, as well as hypermethylation of tumor suppressor molecules like mediating epigenetic silencing of a wellknown TSG-human runt-related transcription factor 3 (RUNX3) in a great subset of cancers [9][10][11][12][13][14].
RUNX3, a remarkable biomarker demonstrated in many solid malignancies, plays a main role of tumor suppression and interacts with other signaling molecules in the context of carcinogenesis in various cancer types including NSCLC [15][16][17][18][19][20][21][22][23]15]. Loss of expression and cytoplasmic mislocalization had been indicated partly as underlying causes of RUNX3 dysfunction in many cancer types.
Although loss of RUNX3 expression is prevailing in human solid malignancies, the role of its association in pathogenesis of respiratory malignancies still requires further elucidation. Our previous studies firstly found that NSCLC patients with higher trimethylated histone H3 at lysine 27 (H3K27me3) level demonstrated a lower probability of postoperative local relapse and determined loss of H3K27me3 expression as an independent risk factor [24,25], and secondly found that NSCLC patients with higher RUNX3 level demonstrated a lower likelihood of postoperative distant metastasis, and the loss of RUNX3 expression predicted worst outcome and shorter overall survival (OS) [26]. In light of both local relapse and distant metastasis contributing to worse outcome in the postoperative setting and the epigenetic regulation that histone modification might have on oncogenes and TSGs, we thus hypothesized that assessment of H3K27me3/RUNX3 co-expression might probably play some role in predicting the surgical outcome of NSCLC patients after radical resection.
In the present study, we sought to investigate the potential interrelationship of expression of RUNX3, H3K27me3 and its methyltransferase EZH2 in NSCLC patients, their correlation with clinicopathologic parameters, and the prognostic significance as well, finding that expression of RUNX3 was closely correlated with H3K27me3 and their co-expression was significantly associated with better prognosis especially in either stage I or postoperative chemotherapy-naive (POCT-naive) NSCLC patients.

Patients and Tissue Samples.
Archival formalin-fixed, paraffin-embedded tissue sections from 5 normal lung tissues (with a diagnosis of pulmonary bulla) and 208 patients who underwent surgery for NSCLC at the Department of oracic Surgery of Fujian Medical University Cancer Hospital during 2010-2011 were selected. None had received chemo or radiotherapy prior to tissue collection. e histopathologic features of cancerous specimens and TNM staging were determined according to the 8 th version of AJCC guidelines for NSCLC. e final follow-up date was February 5, 2020, and all patients were available of their survival data. Patients' survival data were censored if they were still alive or dead of disease other than lung cancer at the date of surveillance. e study protocol was approved by the Human Ethics Review Committee of Fujian Medical University Cancer Hospital, and signed informed consent was obtained from each patient.

Tissue Microarray Building.
A fresh section was cut from each donor block, stained with hematoxylin and eosin (HE), and used as a guide to select the morphologically mostrepresentative regions of the tumor to sample the individual core needle biopsies. A duplicate of 0.6 mm diameter cores were then punched from tumoral areas of each donor tissue block and introduced into previously prepared recipient paraffin blocks [27], after having made hosting holes in the blocks with a tissue microarrayer (Beecher Instruments, Silver Spring, MD). We constructed 4 recipient blocks with a maximum of 10 × 6 dots. With a microtome, 4 μm sections were cut from the TMA blocks and placed onto 3-aminopropyltriethoxysilane-coated glass slides to generate TMA slides for further analyses. Some sections were stained with HE in a routine manner for histological examination.
2.3. Immunohistochemical Detection of RUNX3, H3K27me3, EZH2, and Ki-67. Immunohistochemistry was performed with the indirect enzyme-labeled antibody method, as described previously [24][25][26]. Information of antibodies used is shown in Table 1. For detection of RUNX3, H3K27me3, and EZH2, TMA sections were deparaffinized with toluene and rehydrated in graded alcohols. After autoclaved for 15 min at 120°C in 10 mM citrate buffer (pH 6.0) for antigen retrieval, endogenous peroxidase was inactivated with 0.3% hydrogen peroxide in methanol for 15 min. e sections were then preincubated with 500 μg/ml normal goat IgG dissolved in 1% BSA in PBS (pH 7.4) for 1 h, reacted with primary antibodies for 16 h, washed with 0.075% Brij 35 in PBS, and then incubated with HRP-conjugated goat antimouse (RUNX3)/rabbit IgG (H3K27me3/EZH2/Ki-67) in 1% BSA in PBS for 1 h. After washing with 0.075% Brij 35 in PBS, the sites of HRP were visualized with DAB and H 2 O 2 . As a negative control, some sections were reacted with normal mouse/rabbit IgG instead of the specific antibodies. For simultaneous fluorescent double immunostaining of H3K27me3 and RUNX3, the sections were incubated with Alexa 546 anti-rabbit IgG and Alexa 488 anti-mouse IgG (both 1 : 500) in darkness for 1 h, then washed with 0.075% Brij 35 in PBS in darkness, and finally observed with 0.5 μg/ ml DAPI for 1 min. e stained slides were analyzed under a laser scanning microscope (LSM 5 PASCAL; Carl Zeiss Inc., Germany).

Validation via RNA Sequencing Expression and Survival
Analyses of Core Genes. e gene expression profiling interactive analysis (GEPIA) (http://cancer-pku.cn) was applied to analyze the data of RNA sequencing expression on the basis of samples from the GTEx projects and TCGA. Kaplan-Meier plotter (http://kmplot.com/analysis/index. php?p�service&cancer�lung) was used to determine the effect of genes (RUNX3 and EZH2) on survival based on EGA, TCGA database, and GEO (Affymetrix microarrays only). Survival within groups was compared by log-rank estimates.

Western
Blotting Analyses of RUNX3, H3K27me3, EZH2, and β-Actin. Western blotting was carried out as detailed previously [24]. In brief, 7 pairs of human NSCLC specimens together with their related normal lung tissue were used in this session, and each procedure was repeated 3 times. e specimens were homogenized, and the lysates were centrifuged. Soluble proteins were separated on 10% SDS-PAGE gel (Daiichi Pure Chemical, Tokyo, Japan) with equal amounts (10 μg) of protein per lane. Separated proteins were electrophoretically transferred onto polyvinylidene difluoride (PVDF) membranes (Millipore Corporation, MA, USA), blocked with 10% nonfat milk in TBS (20 mM Tris buffer, pH 7.6, and 150 mM NaCl) for 1 h and then incubated overnight at 4°C with EZH2, H3K27me3, RUNX3, and β-actin antibodies. As a secondary antibody, HRP-goat antirabbit IgG was reacted for 1 h and then the bands were visualized with DAB, Ni, Co, and H 2 O 2 . e grey ratio value of EZH2/H3K27me3/RUNX3 to β-actin for each specimen was calculated. Individual value of EZH2/H3K27me3/ RUNX3 to β-actin for each patient was expressed as mean ± SEM, and then final value in normal and cancerous lung tissue was compared.

TUNEL Staining for Apoptotic Cells in NSCLCs.
To identify nuclei with DNA strand breaks at a cellular level, TUNEL was performed according to the method of Gavrieli et al. [28], with a slight modification. Detailed procedure was described previously [26]. For statistical analysis, more than 10,000 cancer cells/patient were counted, and the number of TUNEL-positive cells was expressed per 1000 of the total cells (mean ± SEM). Data for different groups were compared for statistical difference using Student's t-test. A P value of <0.05 denoted the presence of a significant difference.

Statistical Analysis.
e best cutoffs of RUNX3 and H3K27me3 were determined via X-tile software program by dichotomizing them into high and low expression subgroups, as described in our previous studies [24][25][26]. e SPSS 24.0 statistical software package (SPSS Inc, Chicago, IL, USA) and GraphPad Prism (Version 8.3.0) were employed for all analyses. e association between tested markers and different clinicopathologic parameters of the patients were evaluated by Pearson's χ 2 or Fisher's exact test as appropriate. Survival functions were estimated using the Kaplan-Meier method and compared using log-rank test.
e Cox proportional hazard model was used to evaluate the association between various markers and patient's survival. Univariate and multivariate analyses were determined by Cox regression. A 2-sided P value less than 0.05 was considered statistically significant.

EZH2
. EZH2 was localized in nuclei of NSCLCs while it was negative in normal lung tissue (Figures 1(j)-1(l)). Staining score of EZH2 also ranged from 0 to 12.
Using an average value of 1.9, we determined the samples into low (IHC score ≤1.9) and high (IHC score >1.9) expression subgroups for EZH2. All 5 normal lung tissues were negative of EZH2 expression. Of 188 NSCLC cases, 63 (34%) were of high while 125 (66%) were of low expression. As shown in Figure 2(d), the staining score of EZH2 was significantly lower in normal lung tissue compared to LUSC or LUAD (0 vs. 2.96 ± 0.47 vs. 1.19 ± 0.23, * * P < 0.01). GEPIA analysis demonstrated that no expression discrepancy in EZH2 existed in LUAD when compared to paired normal lung tissue, while EZH2 expression in LUSC was statistically higher than its paired normal lung tissue (Figure 2(e)).

Western Blotting of EZH2, H3K27me3
, and RUNX3 in NSCLC Tissues. Individual grey ratio for EZH2 (95 kD), H3K27me3 (15 kD), and RUNX3 (50 kD) to β-actin (45 kD) in 7 pairs of NSCLC and related normal lung tissues is shown in Figure 4(a). Each procedure was repeated for 3 times and the grey ratio value was recorded. Individual value for RUNX3, H3K27me3, and EZH2 was calculated and expressed as "mean ± SEM" (Figure 4(b)). Typical photograph of western blot is shown in Figure 4(c), and final comparison of RUNX3/H3K27me3/EZH2 between normal 6 Journal of Oncology and cancerous lung tissue is indicated in Figure 4(d), demonstrating obvious statistical discrepancy on H3K27me3/EZH2 while not on RUNX3. Tables 5 and 6, univariate analysis indicated that factors favorable for longer overall survival (OS) in NSCLC patients were as follows: ECOG PS ≤1 (P < 0.001), no LN involvement (P < 0.001), no mediastinal LN involvement or absence of N2 disease (P < 0.001), no distant metastasis at time of diagnosis (P < 0.001), early TNM staging, i.e., stage I-II (P < 0.001) or stage I (P < 0.001), no pleural involvement (P � 0.012), no lymphatic vessel involvement (P � 0.003), no nerve involvement (P < 0.001), R0 resection (P < 0.001), employment of PORT (P � 0.003), employment of postoperative chemoradiotherapy (P � 0.006), normal serum CEA level at diagnosis (P � 0.001), no postoperative regional relapse (P < 0.001), no postoperative distant metastasis (P < 0.001), higher H3K27me3 expression (P � 0.009), and presence of RUNX3/H3K27me3 co-expression (P � 0.003). No correlation had been observed between better survival and age, gender, smoking status, BMI, histology, T-staging, vascular invasion, postoperative chemotherapy, RUNX3 expression level, nuclear RUNX3, or EZH2 expression (all P > 0.05). Multivariate analyses showed that no distant metastasis at the time of diagnosis (P � 0.027), TNM-I staging (P � 0.036), no postoperative regional relapse (P < 0.001), ECOG PS ≤1 (P < 0.001), and no postoperative distant metastasis (P � 0.044) were the independent prognostic factors for better OS. Kaplan-Meier survival analyses with log-rank tests indicated that no difference had been demonstrated in NSCLC patients with different expression level of RUNX3 (P � 0.2338, Figure 5 (a)  significance was found in LUAD patients with different expression levels, with better outcome in higher level patients, while no survival benefit had been found in LUSC patients with different levels of H3K27me3 expression (all P > 0.05). Also, the situation is similar in EZH2

Survival Analyses. As indicated in
(Figures 6(i)-6(p)), and no survival discrepancy had ever been found in the patient subgroups with different histology or TNM staging related to different EZH2 level (all P > 0.05).
In order to testify the present findings, GEO, EGA, and TCGA data were used to verify the results found concerning  We then set out to analyze the survival curves in patients with concomitant expression of RUNX3/H3K27me3 and found out that, as shown in Figure 9(a), in the 4 subgroups based on different expression status of RUNX3/H3K27me3, NSCLC patients with RUNX3/H3K27me3 co-expression demonstrated the best OS in comparison to that of other subgroups, with either RUNX3 or H3K27me3 expression or neither. On this precondition, we further analyzed the outcome of NSCLC patients either with different pathologic staging or histology and found that patients with RUNX3/ H3K27me3 co-expression exhibited better outcome and longer OS especially in the pTNM-I NSCLC (P � 0.0163, n � 68) and pTNM-I LUSC (P � 0.0081, n � 25) subgroups, while no survival difference had been demonstrated in other subgroups (all P > 0.05, Figures 9(b)-9(j)).
Adjuvant therapy, especially postoperative adjuvant chemotherapy (POCT), was often taken for patients with stage-II and above or in stage-I patients suffering postoperative relapse or metastasis. We wondered about the prediction efficacy that RUNX3/H3K27me3 co-expression might have on these populations and found that no survival difference had been determined in POCT patients (P � 0.3820, Figure 10(a)) based on the different status of RUNX3/H3K27me3 co-expression, while in non-POCT patients (P � 0.0029, Figure 10(b)), this difference was  obvious, and non-POCT NSCLC patients with RUNX3/ H3K27me3 co-expression demonstrated a better prognosis, and this situation was ever true in pTNM-I (P � 0.0456, Figure 10(c)) and non-pTNM-I (P � 0.0244, Figure 10(d)) patients.

Proliferative and Apoptotic Index.
We then tried to further explore the underlying elements that might probably result in the different outcomes or OS in the facet of cellular proliferation and apoptosis from different co-expression status of RUNX3/H3K27me3 and found that no statistical WB grey ratio

Discussion
Our previous findings [24][25][26] demonstrated the intimate connection between high H3K27me3 expression and low regional relapse, high RUNX3 level or nuclear localization of RUNX3, and low distant metastasis and hypothesized that the use of RUNX3/H3K27me3 co-expression would probably sensitize the prediction of both postoperative relapse and metastasis. In the present study, we investigated the prognostic value of RUNX3, H3K27me3, and EZH2 immunohistochemically in 188 surgically resected NSCLC patients, demonstrating that NSCLC patients with a cellular signature of simultaneous expressions RUNX3 and H3K27me3 would have better outcome and longer OS, irrespective of their histology and TNM staging. Further stratification analyses based on pathologic staging indicated that RUNX3/ H3K27me3 co-expression might define some specific earlystage patient group (pTNM-I) so as to achieve good outcome and better survival after radical surgery and thereafter could be used as a good biomarker in the postoperative NSCLC patients. Surprisingly, in connection with postoperative regimens especially POCT, our findings still indicated that the prediction efficacy worked well especially in POCT-naive settings, irrespective of the histology or staging. In order to explain the phenomenon, we looked into the underlying mechanism resulting in the survival difference and we tried to compare the difference in cellular proliferation and apoptosis, finding out that co-expression subgroup demonstrates a higher apoptotic index in comparison to that of non-coexpression, while no difference had been determined in proliferation index. is could partially explain a better prognosis in the co-expression subgroup.   [8]. Early relapse and distant metastasis still remain a major concern in this patient population. Unlike locally advanced disease whose outcome might rely mainly on different comprehensive modalities, issues underlying early postoperative relapse/ metastasis in early-stage patients, especially pTNM-I patients, might majorly contribute to the malignant behavior or genetic nature that tumor itself possesses. So, it is of vital importance to discriminate the "evil tumor" that would probably end up with worse outcome. Our present study concentrated on the basis to determine the factors affecting the outcome of surgically resected NSCLC patients. Consistent with our previous and others' findings, patients with good ECOG PS, no LN involvement, no mediastinal LN involvement, no metastasis at diagnosis, early disease (stage I/II), no pleural involvement, no lymphatic vessel involvement, no nerve involvement, R0 resection, postoperative  radiotherapy, postoperative concurrent chemoradiotherapy, no CEA elevation at diagnosis, and no postoperative relapse or distant metastasis would have longer survival and better prognosis [30]. In addition, both Cox regression and Kaplan-Meier survival analysis with log-rank test confirmed that tumor with a signature of RUNX3/H3K27me3 coexpression exhibited a better prognosis in the survival analysis, especially in pathologic TNM-I population, which might help outline the high-risk population demanding close surveillance.
Postoperative therapy, either POCT or PORT, is often used in the locally advanced or postoperatively relapsed/ metastasized settings. In our present study, about 52.1% (98/188) patients received POCT while around 19.7% (37/ 188) received PORT, and it is indicated by univariate analysis that POCT (P � 0.089) did not help boost up OS in overall population while PORT (P � 0.003) did. ese findings were partially consistent with other reports [31][32][33]. Our further analysis on the effectiveness in the prediction of RUNX3/H3K27me3 co-expression among POCT or non-POCT population found that it worked only in non-POCT cohort, irrespective of the histological types or staging.
Putative tumor suppressor activity of RUNX3 has been presented extensively in many solid epithelial tumors, with loss of expression favoring tumorigenesis and/or prognosis [34]. However, this has been contradicted by other reports of RUNX3 behaving more like an oncogene, with overexpression leading to tumorigenesis [35]. Despite the controversy and inconsistent mechanistic evidence, the preponderance of evidence in the literature supports a role for RUNX3 in tumor biology and prognosis. In our previous study, it is indicated that loss of RUNX3 expression, irrespective of its localization, was an adverse factor related to the OS of NSCLC patients [26]. However, with the prolongation of follow-up time, no benefit had been demonstrated between patients subgroups with different level of RUNX3, as was also validated by TCGA dataset. In addition to its alteration in expression level, RUNX3 protein mislocalization from nucleus to cytoplasm was commonly reckoned as another risk factor leading to worse prognosis in cancer of stomach [36], colorectum [37], and breast [38], while not in lung cancer [26]. e mechanism underlying the translocation of RUNX3 protein in the abovementioned malignancies was thought to partially be correlated to factors like ATBF1 [39], src kinase [40][41][42], and TGF-β [43]. In the present study, however, we found that non-nuclear expression of RUNX3 was closely correlated to low level of EZH2 and H3K27me3 via immunohistochemistry, and enhanced RUNX3 expression was positively correlated with high level of EZH2 and H3K27me3, as was also confirmed via fluorescent double immunostaining and western blot. It was reported by Fujii et al. [9] that EZH2 downregulated RUNX3 by increasing H3K27me3 in gastric, breast, prostate, colon, and pancreatic cancer cell lines. However, Rehman et al. [44] recently reported a significant positive correlation between RUNX3 and EZH2 in Indian patients with esophageal cancer, which was similar to our finding in NSCLC. is paradox results concerning RUNX3 might inevitably point to the debate that whether RUNX3 functions as a tumor suppressor gene or oncogene or even both were depending on specific tumor context. Recent exploration indicated that whether RUNX3 played a role of TSG or oncogene was dependent on the status of p53 [45].
at is to say, in the state of DNA damage or oncogenic stress, RUNX3 would positively regulate p53 and was in turn suppressed by it, and p53 would prevent tumorigenesis by decreasing the activity of some crucial oncogenes and thus retard the onset of malignancy. Upon inactivation of p53, dysregulated RUNX3 started to upregulate MYC aberrantly. us, p53 status was considered as a contextual determinant for whether RUNX3 behaved as a tumor suppressor or oncogene. Our further study in future might take these    issues into account and try to figure out the mechanism underlying the balance of between the roles that RUNX3 might play as an oncogene or TSG. Some limitation in this study should be mentioned. Firstly, the sample size used in our study is relatively small, and a cohort with a larger sample size to test the authenticity would be necessary. Secondly, this is a retrospective study and only overall survival rather than progression-free survival was taken into account, so the results achieved in the present study would need a validation by a prospective study with both PFS and OS in the future. irdly, intense study focusing on the interaction and mechanism of RUNX3/H3K27me3 co-expression should be further undermined.  Figure 11: Cellular dynamic parameters between NSCLC patients with different RUNX3/H3K27me3 co-expression status. (a, b) Immunostaining for Ki-67 in NSCLC tissue with/without RUNX3/H3K27me3 co-expression. (c, d) TUNEL staining for apoptotic cells in NSCLC tissue with/without RUNX3/H3K27me3 co-expression, characterized with chromatin condensation and nuclear fragmentation, accompanied by pyknosis, retraction of pseudopodes, and formation of crescent caps of condensed chromatin at the nuclear periphery (arrows, 400× magnification, scale bar � 20 μm). (e) Ki-67 index in NSCLC tissue with/without RUNX3/H3K27me3 co-expression (P � 0.694). (f ) Apoptotic index in NSCLC tissue with/without RUNX3/H3K27me3 co-expression (P < 0.001).

Conclusion
RUNX3/H3K27me3 co-expression defined some specific pTNM-I NSCLC population with higher proportion of apoptotic index and thus better prognosis and longer OS and could probably be used as a biomarker in the prediction of postoperative relapse and metastasis.