c-Myb and its Effector COX-2 as an Indicator Associated with Prognosis and Therapeutic Outcome in Colorectal Cancer

Abstract

Background: One of our previous studies have demonstrated that the cancer suppressor miR-150 regulated the progression of colorectal cancer (CRC) by down-regulating v-myb avian myeloblastosis viral oncogene homolog (c-Myb). The purpose of present study was to evaluate the prognostic value of the expression of c-Myb and its effector, prostaglandin-endoperoxide synthase 2 (COX-2) in patients with CRC.

Methods: We used tissue microarrays (containing 202 CRC tissues and matched adjacent normal tissues) and conducted immunohistochemical analysis and western blotting analysis (containing 3 CRC tissues and matched adjacent normal tissues) to detect the expression of c-Myb and COX-2.

Results: Compared with the adjacent nontumorous tissues, both the expression levels of c-Myb and COX-2 were higher in the cancer tissues. A statistically significant correlation was found between the expression of c-Myb and COX-2. Elevated c-Myb and COX-2 were associated with more advanced tumor invasion and poorer overall survival by univariate analysis. Higher expression levels of both c-Myb and COX-2 were significantly associated with shorter overall survival for stage II and stage III patients with 5-Fu based chemotherapy. Multivariate analysis identified the lymph node involvement, distant metastatic spread and the elevated c-Myb and COX-2 as independent factors of poor prognosis for CRC.

Conclusions: In conclusion, the overexpression of both c-Myb and COX-2 would be of prognostic screening value in patients with CRC.

Introduction

Colorectal cancer (CRC) ranks among the second leading cause of cancer-related mortality in the United States and remains the fourth most common cause of cancer deaths worldwide¹. The burden of CRC in China is also a severe public health issue. In 2015, there were 191 thousand CRC deaths cases according to the data from the National Central Cancer Registry of China ². Although the improvements in early screening such as guaiac fecal occult blood tests (gFOBT), fecal immunochemical test (FIT) and fibercoloscope ameliorate therapeutics management of patients with early stage of this disease, the five-year overall survival rates of individuals with advanced and metastatic stages remain bleak due to the lack of useful and clinically- actionable tumor biomarkers for the prognosis monitoring ^3-5. To date, only serum carcinoembryonic antigen (CEA) measurement is recommend for the over-all post-surgical surveillance of early recurrence in CRC. However, there is no universal standard for the frequency of CEA testing and its high false-positive rate limits its monitoring value. These data highlight the critical and imperative need for identification of effective and sensitive biomarkers for assisting surgeons in developing a therapeutics.

Recent researches emphasized a integrative strategy for molecular subtyping⁶, but only few tissue biomarkers such as microsatellite instability (MSI) status, CpG island methylator phenotype (CIMP) status, RAS mutation, and BRAF mutation are applied in clinical-outcome judgement of CRC. Our previous studies have demonstrated that a tumor- suppressive miRNA, miR-150, regulates the tumorigenesis and progression of CRC by post-transcriptionally reducing the expression level of nuclear proto- oncogene, v-myb avian myeloblastosis viral oncogene homolog (c-Myb) in cultured cells and animal models^7,8. As one of the important transcription factors, c-Myb is frequently elevated in a number of carcinomas including acute myelogenous leukemia ^9,10, salivary adenoid cystic carcinoma¹¹, breast cancer¹², and CRC⁹. This gene participates in a wide variety of biological life processes including cell proliferation, cell cycle progression, and apoptosis, indicating that the malignancy maintenance of CRC cells is c-Myb dependent^13-15. Moreover, recent studies discovered the mediation of chronic inflammatory reaction by c-Myb in the gut^13-15. These findings help us understand why anti- inflammatory therapy is useful in the CRC treatment. Meanwhile, it also reminds us to determine the prognostic value of c-Myb and Inflammatory markers in the post-surgical surveillance of CRC.

As one of the downstream effectors of c-Myb, prostaglandin-endoperoxide synthase 2 (COX-2), is a consistent hallmark feature of gut inflammation and CRC ¹⁹. Overexpression of COX-2 has been confirmed to play an important role in epithelial- mesenchymal transition (EMT) and lymph node metastasis²⁰. However, the clinicopathological relevance of upregulated expression levels of c-Myb in combination with COX-2, in determining the prognosis of CRC patients has not yet been studied. The aim of present study was to determine the expression levels of c-Myb and COX-2 in sporadic CRC, and decipher their prognostic usefulness in this malignancy.

Materials and Methods

Patient specimens and clinicopathological data collection

Matched normal and cancerous colorectal tissues from sporadic CRC patients were obtained from 202 patients (Han Chinese) who received colorectal surgery at the Sixth People's Hospital affiliated with Shanghai Jiao Tong University, Shanghai, China, between January 2004 and October 2007. The study was approved by the committee from the Sixth People's Hospital affiliated with Shanghai Jiao Tong University. Written informed consent was obtained from the patients, in accordance with the institutional guidelines. The methods were performed in accordance with the approved guidelines. We excluded cases that were diagnosed as familial adenomatous polyposis or with colitis-associated CRC. The endpoints of interest was overall survival. The deadline of follow-up was 1, April 2011 for all CRC cases. Survivors among the cases at last contact were considered as censored for following overall survival analysis. All the patients' clinicopathological data including age, gender, tumor location, differentiation, tumor size, depth of invasion, lymph node involvement, distant metastatic spread, tumor node metastasis (TNM) stage, 5-fluorouracil (Fu)-based chemotherapy, KI-67 expression, P53 status, and CEA status were obtained from the electronic medical records of the Sixth People's Hospital affiliated with Shanghai Jiao Tong University. The summary of clinicopathological data was demonstrated in table 1.

Table 1.

Clinicopathological variables and the expressions of c-Myb and COX-2 in patients with colorectal cancer

Characteristics	Total	Tumor c-Myb expression				Tumor COX-2 expression
	(n)	Elevated (n)	Decreased (n)	P-value		Elevated (n)	Decreased (n)	P-value
All cases	202	116	86			102	100
Age (years)				0.933				0.946
< 60	57	33	24			29	28
≥ 60	145	83	62			73	72
Gender				0.894				0.195
Female	88	51	37			49	39
Male	114	65	49			53	61
Tumor location				0.095				0.030
Right hemicolon	76	51	25			47	29
Left hemicolon	61	32	29			29	32
Rectum	65	33	32			26	39
Differentiation				0.067				0.150
Well-moderate	184	102	82			90	94
Poor	18	14	4			12	6
Tumor size (cm)				0.812				0.199
< 5	110	64	46			51	59
≥ 5	92	52	40			51	41
Depth of invasion				0.020				0.003
Tis - T2	23	8	15			5	18
T3 - T4	179	108	71			97	82
Lymph node involvement				0.062				0.292
N0	126	66	60			60	66
N1 - N2	76	50	26			42	34
Distant metastatic spread				0.092				0.370
Absence (M0)	191	107	84			95	96
Presence (M1)	11	9	2			7	4
TNM stage
0 - Ⅱ	122	62	60	0.019		56	66	0.107
Ⅲ - Ⅳ	80	54	26			46	34
5-Fu-based chemotherapy				0.654				0.481
Absence	100	59	41			53	47
Presence	102	57	45			49	53
KI-67 expression				0.409				0.153
< 50%	71	38	33			31	40
≥ 50%	131	78	53			71	60
P53 status				0.808		28	36	0.192
Negative	64	36	28			74	64
Positive	138	80	58
CEA status				0.569				0.778
Negative	101	56	45			50	51
Positive	101	60	41			52	49

Abbreviations: TNM, tumor node metastasis

Tissue microarray construction

Formalin-fixed and paraffin-embedded tumor blocks were collected from the archives of the pathological departments at the Sixth People's Hospital affiliated with Shanghai Jiao Tong University. After marked on the hematoxylin and eosin slide, the most representative tumor areas of the paraffin blocks were chosen and 2.0-mm dimension tissue cores were arranged in tissue microarray (TMA) blocks. Then, 4.0 μm TMA sections were cut for immunohistochemistry (IHC) analysis.

Immunohistochemistry and staining evaluation

The sections were dried for 2 h at the 65 °C hot air oven, de-waxed by dimethylbenzene and rehydrated in gradient alcohol. Antigen was retrieved in citric acid buffer(PH 6.0) with a pressure antigen retrieval procedure for 5 min. After washing the slides by PBS buffer(PH 7.4), endogenous peroxidase activity was interdicted by incubating the sections for 10 min in endogenous peroxidase blocking solution at room temperature. Non-specific background was blocking by incubating with normal nonimmune goat serum at room temperature for 30 min. The sections were then incubated with following primary antibodies: anti-c-Myb (Rabbit polyclonal, 1:400, Abcam, Cambridge, United Kingdom, ab117635) and anti-COX-2 (Rabbit polyclonal, 1:200, Abcam, Cambridge, United Kingdom, ab15191). The slices incubated with PBS buffer (PH7.4) only were used as negative control. After washing with the PBS buffer (PH7.4), the tissue sections were detected and visualized with the Polink-1 HRP DAB detection system (ZSGB-BIO, Beijing, China). Then the reaction was stopped under running water and the sections were counterstained with Mayer's hematoxylin, dehydrated with gradient alcohol, mounted with neutral balsam. The immunoreactivity of each slice was examined by three independent investigators and photographed under the light microscope (Leica DM2000, Germany) with digital camera at high magnification (Leica DFC450C, Germany). The staining immunoreactivity was scored ranged from 0% to 100% (5% as intervals) by calculating the proportion of positive cancerous cells over all areas and the scorers were blinded for each other' score. A cutoff point of each marker was chosen by using median value to dichotomize cases. Scoring above or below the cutoff point was classified as “elevated” or “decreased”, respectively.

Western blotting analysis

Total protein were obtained from three pairs of matched normal and cancerous CRC frozen tissues using RIPA Lysis Buffer (Beyotime, Shanghai, China) according to the specification. Total protein was quantified using a BCA kit (TIANGEN, Beijing, China). SDS-PAGE was performed for 0.5 h at 80 V and then at 120 V for 1.5 h. After electrophoresis, the separated proteins were transferred from the gel to the PVDF membranes (Millipore, Bedford, USA) at 250 mA for 2h at 4°C. Then the PVDF was blocked with skim milk in tris-buffered saline tween (TBST) 1 h at room temperature. Following blocking, the membranes was incubated overnight at 4°C with the primary antibody: anti-c-Myb (Rabbit polyclonal, 1:500, Abcam, Cambridge, United Kingdom, ab117635); anti-COX-2 (Rabbit monoclonal, 1:500, Abcam, Cambridge, United Kingdom, ab62331); anti-beta Actin (Mouse monoclonal, 1:5000, SIGMA, California, USA, A5441). After washing with TBST three times (5 min per wash), the PVDF was incubated with the corresponding secondary antibody (anti-rabbit, 1:5000, LICOR, Nebraska, USA, 926-32223; anti-mouse, 1:5000, LICOR, Nebraska, USA, 926-68072) 1 h at room temperature and analyzed with the Odyssey 3.0 system.

Statistical analysis

All statistical analyses were performed by using the IBM SPSS Statistics 20.0 software (IBM, lnc., Chicago, Illinois, USA) and GraphPad Prism 6 software (GraphPad software, lnc., San Diego, California, USA). For comparison of expression of each marker between the CRC tissues and the matched adjacent normal tissues, a paired Student's t test was conducted. The proportion values of positively stained cells were transformed using the arcsine-square-root transformation to normalize the data²¹. The relationships between c-Myb and COX-2 expression was analyzed using linear or non-linear regression, if appropriate. The associations between the clinicopathological data and the staining of c-Myb and COX-2 were performed using Pearson's Chi-square. The impacts of clinical variables and the staining of c-Myb and COX-2 were estimated by using univariate and multivariate Cox proportional hazards regression model. All tests were double tail and the p values below 0.05 were considered significant.

Results

Detection of c-Myb and COX-2 expression in CRC tissues and paired adjacent mucosa

Two hundred and two cases with matched normal and neoplastic tissues from CRC patients were included in our present study. Representative IHC staining of c-Myb and COX-2 for CRC tissues and paired adjacent mucosa are represented in Figure 1. Immunoreactivity for c-Myb was detected primarily in the nuclei, while COX-2 was expressed primarily in the cytoplasm. The median staining score of c-Myb was 85% in the CRC tissues and 30% in the matched adjacent normal tissues. Meanwhile, the median staining score of COX-2 was 60% in the cancerous tissues and 20% in the matched noncancerous tissues. In agreement with the reported studies^9,22,23, the expression levels of both c-Myb (p < 0.001) and COX-2 (p < 0.001) were higher in the CRC tissues in comparison with the adjacent normal tissues (Figure 2A & B).

Figure 1.

Representative IHC staining of c-Myb and COX-2 for CRC tissues and paired adjacent mucosa. A) IHC of c-Myb protein in tumorous or adjacent normal tissues. B) IHC of COX-2 protein in tumorous or adjacent normal tissues. C) Low and high expression level of c-Myb protein in tumorous tissues. D) Low and high expression level of COX-2 protein in tumorous tissues. Original magnifications: ×100 or ×200.

Figure 2.

The positive expression rate of c-Myb and COX-2 in CRC. A) IHC detection of c-Myb in 202 paired of CRC and noncancerous tissues (paired Student's t test; ***, p < 0.001); B) IHC detection of COX-2 in 202 paired of CRC and noncancerous tissues (paired Student's t test; ***, p < 0.001). C) Correlation of c-Myb expression with COX-2 expression (linear regression; r = 0.391, p < 0.001). D) Western blot analysis of the protein expression of c-Myb and COX-2 in the CRC tissues, as compared with noncancerous tissues.

Furthermore, linear regression model was examined to explore the association between c-Myb and COX-2 expression. A statistically significant correlation was found between the proportional immunostaining level of c-Myb and the expression of COX-2 (r = 0.391, p < 0.001; Figure 2C). We further validated the result by Western blot. Compared with noncancerous tissues, the protein expression of c-Myb was found to be upregulated in the CRC tissues, accompanied by high expression of COX-2 (Figure 2D).

The associations of c-Myb and COX-2 expression with clinicopathological variables

To explore the relationship of c-Myb and COX-2 expression with clinicopathological variables, the median value of each marker was chosen as the cutoff point to dichotomize the staining scores. With this criterion, 116 cases were classified as high-expression and 86 cases as low-expression for c-Myb. Likewise, 102 cases were classified as high-expression and 100 cases as low-expression for COX-2.

As demonstrated in Table 1, elevated c-Myb expression was significantly associated with more advanced tumor invasion (p = 0.020) and higher TNM stage (p = 0.019). Although the elevated c-Myb expression was slightly associated with tumor differentiation (p = 0.067) and lymph node involvement (p = 0.062), the results didn't reach statistical significance. Elevated COX-2 expression was also significantly correlated with more advanced tumor invasion (p = 0.003) and tumor location (p = 0.030). Among variables including age, gender, tumor differentiation, tumor size, lymph node involvement, distant metastatic spread, TNM stage, 5-Fu-based chemotherapy, KI-67 expression, P53 status and CEA status, there was no association between these variables and the COX-2 expression.

Expression of c-Myb and COX-2 correlated with poor overall survival in colorectal carcinoma

Univariate associations of the expression of each protein with overall survival were listed in Table 2. Poorer tumor differentiation (Hazard ratio (HR), 2.348; 95% confidence interval (95% CI), 1.272 - 4.333; p = 0.006), more depth of invasion (HR, 4.324; 95% CI, 1.365 - 13.698; p = 0.013), positive lymph node involvement (HR, 4.000; 95% CI, 2.545 - 6.286; p < 0.001), positive distant metastatic spread (HR, 3.407; 95% CI, 1.752 - 6.623; p < 0.001), higher TNM stage (HR, 2.637; 95% CI, 1.975 - 3.520; p < 0.001), and positive CEA status (HR, 1.583; 95% CI, 1.019 - 2.458; p = 0.041) demonstrated shorter overall survival of CRC. Cases with c-Myb overexpression, as determined by IHC, had a 2.693-fold (95% CI, 1.640 - 4.421; p < 0.001) more risk for death (Figure 3A). Meanwhile, cases with COX-2 overexpression had a 2.334-fold (95% CI, 1.485 - 3.669; p < 0.001) poorer overall survival (Figure 3D).

Table 2.

The effects of clinicopathologic variables on overall survival rate by univariate and multivariate Cox proportional hazard regression analysis

Variables	Univariate			Multivariate
	HR	95% CI	P-value	HR	95% CI	P-value
c-Myb	2.693	1.640 - 4.421	< 0.001	1.841	1.082 - 3.133	0.024
COX-2	2.334	1.485 - 3.669	< 0.001	1.672	1.022 - 2.736	0.041
Age	1.033	0.643 - 1.661	0.892
Gender	0.957	0.621 - 1.476	0.843
Tumor location	0.910	0.698 - 1.185	0.482
Differentiation	2.348	1.272 - 4.333	0.006	1.409	0.749 - 2.651	0.288
Tumor size	1.192	0.775 - 1.833	0.425
Depth of invasion	4.324	1.365 - 13.698	0.013	1.995	0.612 -6.500	0.252
Lymph node involvement	4.000	2.545 - 6.286	< 0.001	3.535	1.127 - 11.084	0.030
Distant metastatic spread	3.407	1.752 - 6.623	< 0.001	2.247	1.014 - 4.978	0.046
TNM stage	2.637	1.975 - 3.520	< 0.001	0.945	0.285 - 3.134	0.927
5-Fu-based chemotherapy	1.038	0.674 - 1.598	0.865
KI-67 expression	1.212	0.762 - 1.928	0.417
TP53 status	1.318	0.814 - 2.135	0.262
CEA status	1.583	1.019 - 2.458	0.041	1.258	0.795 - 1.990	0.326

Abbreviations: TNM, tumor node metastasis; HR, hazard ratio; CI, confidence interval.

Figure 3.

The influence of c-Myb and COX-2 expression for overall survival in CRC. A) Relationship of c-Myb expression to overall survival in all 202 patients (log-rank; p < 0.001). B) Relationship of c-Myb expression to overall survival in patients without 5-Fu-based chemotherapy (log-rank; p = 0.024). C) Relationship of c-Myb expression to overall survival in patients with 5-Fu-based chemotherapy (log-rank; p < 0.001). D) Relationship of COX-2 expression to overall survival in all 202 patients (log-rank; p < 0.001). E) Relationship of COX-2 expression to overall survival in patients without 5-Fu-based chemotherapy (log-rank; p = 0.039). F) Relationship of COX-2 expression to overall survival in patients with 5-Fu-based chemotherapy (log-rank; p = 0.001).

Multivariate Cox proportional hazards regression analysis was performed using all 202 cases with complete clinical data for variables that were significant on univariate model (Table 2). After multivariable adjustment by considering the clinicopathological variables including tumor differentiation, depth of invasion, lymph node involvement, distant metastatic spread, TNM stage, and CEA status, only the lymph node involvement (HR, 3.535; 95% CI, 1.127 - 11.084; p = 0.030), distant metastatic spread (HR, 2.247; 95% CI, 1.014 - 4.978; p = 0.046), and expression of c-Myb and COX-2 remained independently significant. Tumor tissues with elevated c-Myb expression had a 1.841-fold (95% CI, 1.082 - 3.133; p = 0.024) more risk for death. Tumor tissues with COX-2 overexpression had a 1.672-fold (95% CI, 1.022 - 2.736; p < 0.041) poorer overall survival.

To further investigate the influence of c-Myb and COX-2 expression on patients with adjuvant chemotherapy, we conducted subgroup analysis in patients with or without 5-Fu-based chemotherapy. Elevated expression of c-Myb or COX-2 had a worse response to 5-Fu-based chemotherapy (p < 0.001 and p = 0.001, respectively; Figure 3C, 3F) in comparison with patients without chemotherapy (p = 0.024 and p = 0.039, respectively; Figure 3B, 3E).

Combination of c-Myb and COX-2 in predicting prognosis of CRC

To confirm that the combination of c-Myb and COX-2 may help predict prognosis of CRC patients, we constructed three subgroups based on the expression levels of both genes: subgroup 1, low c-Myb and low COX-2; subgroup 2, (low c-Myb and high COX-2) or (high c-Myb and low COX-2); subgroup 3, high c-Myb and high COX-2. As expected, overexpression of both genes predicted worse overall survival (p < 0.001; Figure 4A). When stratified by 5-Fu-based chemotherapy, the promising prognostic model by combination of c-Myb and COX-2 were yet again confirmed, suggesting that the two markers may help to predict and select the effect of chemotherapy on survival time in patients with CRC (Figure 4B, 4C).

Figure 4.

Combination of c-Myb and COX-2 in predicting prognosis of CRC. A, B and C) Combination of both up-regulated molecules predicted worse overall survival in all patients (A; log-rank; p < 0.001 for subgroup 2 and p < 0.001 for subgroup 3, respectively) or in patients without 5-Fu-based chemotherapy (B; log-rank; p = 0.002 for subgroup 2 and p = 0.004 for subgroup 3, respectively) or in patients with 5-Fu-based chemotherapy (C; log-rank; p < 0.001 for subgroup 2 and p < 0.001 for subgroup 3, respectively). Subgroup 1, low c-Myb and low COX-2; subgroup 2, (low c-Myb and high COX-2) or (high c-Myb and low COX-2); subgroup 3, high c-Myb and high COX-2. Subgroup 1 was chosen as control.

5-Fu-based adjuvant chemotherapy is established routine therapy for CRC patients with stage Ⅱ and stage Ⅲ disease, especially for those with poor prognostic features. When further stratified by TNM stage, the co-overexpression of both genes was still of value for prognostic evaluation. In detail, overexpression of c-Myb and COX-2 was significantly associated with shorter overall survival for stage Ⅱ (p = 0.028; Figure 5A, 5B) and stage Ⅲ (p = 0.003; Figure 5C, 5D) patients with 5-Fu based chemotherapy.

Figure 5.

Combination of c-Myb and COX-2 in predicting prognosis of CRC in stage II and III patients. A and B) Combination of both up-regulated molecules predicted worse overall survival in stage Ⅱ patients without (A; log-rank; p = 0.748 for subgroup 2 and p = 0.193 for subgroup 3, respectively) or with 5-Fu based chemotherapy (B; log-rank; p = 0.412 for subgroup 2 and p = 0.028 for subgroup 3, respectively). C and D) Combination of both up-regulated molecules predicted worse overall survival in stage Ⅲ patients without (C; log-rank; p = 0.029 for subgroup 2 and p = 0.349 for subgroup 3, respectively) or with 5-Fu based chemotherapy (D; log-rank; p < 0.001 for subgroup 2 and p < 0.001 for subgroup 3, respectively). Subgroup 1, low c-Myb and low COX-2; subgroup 2, (low c-Myb and high COX-2) or (high c-Myb and low COX-2); subgroup 3, high c-Myb and high COX-2. Subgroup 1 was chosen as control.

Discussion

Although our previous study identified the oncogenic role of c-Myb in cellular apoptosis, progression and proliferation in cultured cells and an animal model, its clinical significance remains unclear and requires interrogation⁷. In this study, we have firstly developed a prognostic model based on two proteins to enhance and better predict the prognosis of CRC. Our study presented that the proportional immunostaining level of c-Myb was significantly correlated with the expression of COX-2. The expression level of c-Myb in combination with COX-2 can fairly well categorize CRC cases into low- and high-risk phenotypes with difference in overall survival rate. Cases with high c-Myb and COX-2 were involved in worse response to chemotherapy.

The MYB family of the transcription factor genes contain c-Myb, b-Myb and a-Myb²⁴. All members of the MYB family contain a conserved DNA-binding domain²⁴. Compelling studies have shown that c-Myb plays a pivotal role of proto-oncogenes in cancers including CRC²⁵. However, a recent study published by Michal Tichý found a negative association between the expression of c-Myb in CRC tissue and the presence of distant metastases²⁶. Considering that both pro-invasive and anti- metastatic roles of c-Myb were described, we examined the expression of c-Myb at protein level in 202 CRC tissue samples and verified it at mRNA level using public databases from The Cancer Genome Atlas (TCGA) cohort of patients with colon adenocarcinoma (TCGA COAD). In agreement with the reported studies^9,22,27, we found that the expression of c-Myb (p < 0.001), detected by IHC, was significantly higher in the CRC tissues in comparison with the adjacent normal tissues. Meanwhile, elevated expression of c-Myb was significantly associated with more advanced tumor invasion (p = 0.020), higher TNM stage (p = 0.019) and poorer overall survival (p < 0.001). Whereas, there were no significant association between the amount of the c-Myb mRNA and the poor outcome of patients with CRC according to the TCGA COAD (Supplementary table S1). These findings supported the fact that the protein level of the c-Myb, but not the amount of the gene mRNA, may serve as a good prognostic marker.

As a nuclear transcription factor, c-Myb participates in diverse cellular processes by up-regulation of its downstream transcriptional targets including COX-2^28-30. COX-2 inhibits apoptosis and promotes cell proliferation in numerous cancers including breast cancer^31,32, cervical cancer³³, lung cancer³⁴, gastric cancer³⁵, and CRC³⁶. A myriad of literature has confirmed that long-term therapy with non-steroidal anti-inflammatory drugs (NSAIDs) such as aspirin, or selective COX-2 inhibitors like celecoxib, are associated with a reduced risk of CRC^36-39. When it comes to the prognostic value, the COX2 expression by IHC was found to have a significant correlation with tumor stage, despite few studies demonstrated a negative effect^40-42.

Ramsay and his colleagues were the first to demonstrate that human COX-2 promoter contains a high-affinity binding site of c-Myb. Given that both c-Myb and COX-2 drive tumor progression²⁹, we have reason to believe that using c-Myb-COX-2 partner model to predict CRC prognosis may be better than utilizing any of the single indicator. A recent study had found that the combination of c-Myb, v-kit Hardy-Zuckerman 4 feline sarcoma viral oncogene homolog (c-kit), and COX-2 show better clinical significance in predicting prognosis in salivary adenoid cystic carcinoma⁴³. However, the relationship between c-Myb and COX-2 in CRC hasn't been verified in clinic specimens up to now. In this context our findings have important clinical significance and urge us to confirm the prognostic value of c-Myb-COX-2 partner model in CRC. In our study, both univariate and multivariate Cox proportional hazards regression analysis supported that cases with high c-Myb and COX-2 indicated poorer overall survival. When stratified by 5-Fu-based chemotherapy status and TNM stage, the prognostic value by combination of c-Myb and COX-2 were reconfirmed and validated. These findings suggest that the overexpression of both c-Myb and COX-2 is of prognostic screening value in patients with CRC. However, according to the TCGA COAD database, there were no significant association between the mRNA level of both gene and the prognosis of CRC (Supplementary table S2). The reason for this could be that c-Myb and COX2 promotes the progression of the disease mainly at the posttranscriptional level. Studies based on gain-of-functions/loss-of-functions experiments are required to further explain this phenomenon.

In conclusion, our present study demonstrates that both the c-Myb and COX-2 are two independent prognostic biomarkers and the c-Myb-COX-2 model reinforces the function in prognostic assessment in patients with CRC.

Supplementary Material

Supplementary tables.

Abbreviations

CEA

carcinoembryonic antigen

c-Myb

v-myb avian myeloblastosis viral oncogene homolog

COX-2

prostaglandin-endoperoxide synthase 2

CRC

colorectal cancer

HR

hazard ratio

95% CI

95% confidence interval

IHC

immunohistochemistry

NSAIDs

non- steroidal anti-inflammatory drugs

PBS

phosphate buffer saline

TBST

tris-buffered saline tween TMA, tissue microarray

TNM

tumor node metastasis

WB

western blot