Clinical significance and prognostic value of TRIM24 expression in esophageal squamous cell carcinoma

TRIM24 mRNA and protein levels were significantly decreased in ESCC tissues

To explore the clinical significance of TRIM24, we first determined its mRNA expression in 42 ESCC samples by quantitative RT-PCR. The result shows that TRIM24 mRNA level in the cancer tissues is significantly lower (P<0.001, Fig. 1A) than that in the adjacent non-cancerous esophageal tissues (NCET). However, no relationship was found between the mRNA expression and clinical features in ESCC patients (Supplemental Table 1) because the samples size is small. In order to confirm whether TRIM24 protein is also decreased, we employed Western blot to analyze 19 pairs of ESCC tissues and matched NCET (randomly selected from the 42 ESCC). The results demonstrate that the protein level of TRIM24 in ESCC tissues is indeed reduced in about 73.7% (14/19) of these ESCCs (P= 0.006, Fig. 1B and 1C), which is concordant with TRIM24 mRNA in ESCC.

To further investigate the clinical significance of TRIM24, we performed immunohistochemical staining of TRIM24 protein in 213 formalin-fixed paraffin-embedded (FFPE) ESCC specimens. As shown in Figure 2A, TRIM24 protein is predominantly localized in the nuclear compartments, with the strongest staining in the normal esophageal epithelium cell, especially in basal and suprabasal layers. Compared with the NCET tissues, the ESCC tissues exhibit significantly lower expression level of TRIM24 protein (P = 0.002, Fig. 2B), which is concordant with he results of RT-PCR and Western blot analysis on fresh ESCC tissues.

Relationship between TRIM24 and clinicopathological characteristics in ESCC

TRIM24 protein is reduced in ESCC, which implies that it functions as a tumor suppressor in ESCC. To elucidate the clinical significance of the TRIM24 protein in ESCC, we first divided all patients into a low (n=137) or high (n=76) expression group based on the cutoff value of 4 immunoreactive scores (IRS) for TRIM24 staining. Chi-square test was used to analyze the relationship between TRIM24 expression and clinicopathological characteristics. The results demonstrate that 71.6% (78/109) of patients with low TRIM24 expression have lymph node metastasis (LNM) while only 28.4% (31/109) of patients with high expression have LNM (P=0.024, Table 1); 71.4% (70/98) of ones with low expression have pathological tumour-node-metastasis stage (pTNM stage) III while only 28.6% (28/98) of ones with high expression have pTNM stage III (P=0.046, Table 1), suggesting that TRIM24 may inhibit ESCC progression. Moreover, patients with low expression have more postoperative relapse and metastasis than those with high expression (72.4% [97/134] vs 27.6% [37/134], P=0.001, Table 1), implying that TRIM24 may suppress ESCC development and metastasis. No statistical association was found between TRIM24 expression and other clinicopathological characteristics (Table 1).

Low expression of TRIM24 predicts poor prognosis in ESCC patients

Since TRIM24 is associated with LNM, high pTNM stage and postoperative recurrence and metastasis of ESCC patients, we reasoned that TRIM24 is correlated with survival of the patients. As expected, Kaplan–Meier survival analysis shows that patients with high TRIM24 expression have significantly better overall survival (OS) and disease-free survival (DFS) than those with low expression (P<0.001, P<0.001, respectively; Fig. 3A). Survival analysis also display that 5-year OS and 5-year DFS of patients with high TRIM24 level are 64.4% and 57.2%, compared with 35.8% and 28.9% of those with low level, respectively, which hint that TRIM24 is a tumor suppressor in ESCC patients.

To figure out whether TRIM24 expression is an independent prognostic factor in ESCC patients, univariate and multivariate analyses were conducted on TRIM24 and other clinicopathological characteristics. The univariate Cox proportional analysis reveals that the TRIM24 expression (P = 0.001), smoking history (P = 0.014), tumor size (P = 0.001), lymph node metastasis (P <0.001) and pathological TNM stage (P<0.001) are significant prognostic predictors for OS and DFS (Table 2). In addition, tumor invasion depth also is a significant prognostic predictor for OS (P = 0.012). Because tumor differentiation, location, invasive depth and lymph node metastasis are determinants of pTNM staging system (7th version), they were not included in the multivariate Cox proportional analysis. As presented in Table 2, TRIM24 expression is identified as an independent survival predictor for OS (HR, 0.519; 95%CI, 0.341-0.788; P=0.002) and DFS (HR, 0.584; 95%CI, 0.397-0.861; P=0.007), and pTNM staging system is another independent prognostic factor.

Furthermore, the prognostic value of TRIM24 expression was also investigated in stratified analysis. When patients were stratified by tumor size (5 cm as a cutoff), we found that TRIM24 expression is a prognostic factor for OS (P = 0.003) and DFS (P = 0.002) in patients with tumor less than 5 cm (Fig. 3B), but not in those with tumor larger or equal to 5 cm (Supplemental Fig. 1). We also found that TRIM24 expression is a significant predictor for OS and DFS in patients with well-differentiated ESCC (OS, P = 0.003; DFS, P = 0.012;Fig. 3C) or earlier pTNM stage (I-II) (OS, P = 0.021; DFS, P = 0.017; Fig. 3D), but not in those with moderate/poor differentiation or advanced pTNM stage (III) (Supplemental Fig. 1). The above results suggest that TRIM24 mainly serves as a potential prognostic biomarker for ESSS patients in the early development stage of ESCC.

Combined TRIM24 expression with pTNM stages markedly improves prediction of prognosis in ESCC patients

It is well known that TNM staging system is the main tool to predict the prognosis of malignant tumor patients. However, as shown in Figure 4A, TNM staging system inadequately predicts survival outcomes of ESCC patients with stage I and II (OS, P = 0.386; DFS, P = 0.650), which is also observed in another study [25]. Since the above result shows that TRIM24 can provide additional prognostic information, we hypothesized that TRIM24 will improve the prediction of survival in ESCC patients when it is combined with pTNM staging system. Therefore, we established a new model of combined TRIM24 and pTNM stage to calculate a new risk score for every patient. TNM stage was scored as 1 for stage I, 2 for stage II, 3 for stage III, and the expression level of TRIM24 was assigned a risk score of 0 for high expression and 1 for low expression. The final risk score for each patient is the sum of stage score and TRIM24 risk score (Range 1-4). Then all 213 ESCC patients were classified into 4 groups based on their new risk score. As expected, the survival predictive accuracy of the new risk model is much better than that of TNM staging system alone (Supplemental Fig. 2). However, the survival outcomes of Group 1 and Group 2 are not different because the sample size of Group 1 is too small (only 9 patients). In this case, we combined Group 1 and 2 into a new Group 1, and patients with combined risk score 3 and 4 were re-assigned into Group 2 and 3. As illustrated in Figure 4B, the survival outcomes of all 3 groups are significantly different. Furthermore, we employed receiver operating characteristic (ROC) analysis to compare the survival predictions of ESCC patients by pTNM staging system, TRIM24 protein expression and a combination of both. The result shows that the survival prediction of the combined risk model is the best among the three predictors. These results suggest that TRIM24 can significantly improve postoperative prognostic prediction for ESCC patients in the combined risk model.

The relationship between TRIM24 and RARα expressions in ESCC

In this study, our findings indicate that TRIM24 functions as a tumor suppressor in ESCC, which is conflicting with some reports on other cancers [12,26]. However, in transgenic mice models, TRIM24 is demonstrated as a tumor suppressor [16,17,27] and can suppress liver carcinogenesis via interacting with RARα [15].

To decipher the relationship between TRIM24 and RARα in ESCC, immunostaining was also performed on RARα in 213 ESCC samples. As shown in Figure 2A, RARα is also primarily located in the nucleus of tumor cells, and partly in the cytoplasm, which is consistent with the previous report [28]. The protein expression level in nuclear parts is significantly increased in ESCC tissues compared with the paired NCETs (P< 0.001, Fig. 2C). Patients were classified into high- or low-RARα expression group according to the same cutoff value (4 IRSs) as used in patients classified by TRIM24 protein. However, no significant correlation between RARα and clinicopathological characteristics was found except for gender (P = 0.004, Supplemental Table 2). Moreover, RARα expression also is not associated with OS (P = 0.223, Supplemental Fig. 3A) and DFS (P = 0.770, Supplemental Fig. 3A). Surprisingly, the expression level of RARα protein has no association with TRIM24 expression level according to Pearson’s correlation analysis (data not shown), which suggests that RARα protein may not be correlated with TRIM24 protein in ESCC.

Patients and tissues

The formalin-fixed paraffin-embedded (FFPE) ESCC tissues were obtained from 213 patients who underwent esophagectomy between January 2008 and December 2009 in Sun Yat-Sen University Cancer Center. All patients were included according to as the following criteria: i) Not receiving any chemotherapy and/or radiotherapy before surgery; ii) R0 resection, which means that no tumor tissue at the surgical margins of the resected specimens was found by postoperative pathology examination; iii) Esophageal squamous cell carcinoma diagnosed by pathology; iv) No distant metastases before surgery, including supraclavicular and paraaortic lymph node metastases. The cohort patients consisted of 168 males (78.9%) and 45 females (21.1%), aging from 34 to 88 years (mean age: 60 years old). The follow-up time ranged from 3 to 90 months with a median of 45 months. Overall survival (OS) was defined as from the date of surgery to the date of death or last follow-up, and disease-free survival (DFS) time was from the date of surgery to the date of recurrence, metastasis, or last follow-up. Clinical and pathological tumor staging was performed according to the 7^th edition of the TNM classification of the International Union against Cancer (UICC).

Another forty-two fresh paired ESCC tissues and adjacent noncancerous esophageal tissues (NCET) that were at least 2cm far from the tumor tissues were collected from the patients who had undergone esophagectomy at the Cancer Center from June 2012 to December 2013. Written informed consent was obtained from all the patients before the surgery, in which patients agreed to use their medical record data and tissue samples for the research. This study was reviewed and approved by the medical ethics committee of Sun Yat-Sen University Cancer Center.

Isolation of total RNA and real-time reverse transcription quantitative PCR (RT-qPCR)

The total RNA from tumor tissues and the paired non-tumor samples was extracted by TRIzol reagents following the manufacturer’s instructions (Invitrogen life Technologies, Carlsbad, CA, USA). RNA was evaluated in a NanoDrop^TM 2000 spectrophotometer (ThermoFisher Scientific, Waltham, MA, USA). All RT-qPCR reactions were performed as described previously [37]. Briefly, 1 μg of total RNA was reversely transcribed into cDNA using the Moloney Leukemia Virus Reverse Transcriptase (A5001; Promega, Madison, WI, USA). Quantitative real-time PCR was performed on Roche PCR machine in a total 15 μL volume of reaction mixture consisting of 7.5 μL of GoTaq® qPCR Master Mix (A6001; Promega), 0.25 μL (10μmol/L) of each primer, 1 μL cDNA template, and 6μL nuclease-free water. PCR protocol is as follows: 95 ^oC for 120s, 45 cycles of 95 ^oC for 15s, 60 ^oC for 60s, followed by a dissociation stage. The detection for each sample was repeated in triplicate. To calculate the relative level of TRIM24 mRNA expression in tumor and paired adjacent normal tissues, the comparative CT method (2^-ΔΔCT) was performed and GAPDH was detected as the reference gene. The primers used in the RT-PCR were as follows: TRIM24 forward, 5^’-GGAGTCATTCGTTGCCCAGT-3^’; reverse,5^’-TTCTGCGTTGTCCTCACAGC-3^’; amplification fragment length: 147bp. GAPDH forward, 5^’-CTCCTCCTGTTCGACAGTCAGC-3^’; GAPDH reverse, 5^’-CCCAATACGACCAAATCCGTT-3^’; amplification fragment length: 205bp.

Protein extraction and Western blot analysis

19 ESCC and matched NCET tissues were selected randomly from all the fresh samples, and homogenized to extract protein in a RIPA lysis buffer containing 1 mM phenylmethylsulphonyl fluoride (PMSF). Protein concentration was measured by BCA protein assay. After mixing with 4×SDS sample buffer, 50μg of the protein samples were separated in 8% SDS-PAGE and then transferred onto polyvinylidene difluoride (PVDF) membranes. Then membranes were blocked with TBST containing 5% skimmed milk and incubated sequentially with the primary antibodies against TRIM24 (1:1000 dilution) (14208-1-AP; Proteintech, Chicago, IL, USA) and GAPDH (1:1000 dilution) (#5174; Cell Signaling Technology, USA) at 4 ^oC overnight. After three 10-minute washes with TBST, the membranes were incubated with HRP-conjugated secondary antibodies (dilution for 1:5000) (Jackson Immunoresearch Inc, PA, USA) for about 1 hour at room temperature. The immunocomplexes were detected by an enhanced chemiluminescence system (ECL; cell signaling), and the gray value of the protein bands was measured by using Image J software (http://rsb.info.nih.gov/).

Immunohistochemistry

Surgically resected esophageal tumor specimens were fixed with 10% formalin, embedded in paraffin, and sectioned into 4mm-thick slide. The section was immersed twice in xylene for 10 minutes each. After being rehydrated in a series of graded alcohol (100%, 95%, 85%, 75%, 50%) for 5 minutes each, the section was washed with Phosphate Buffered Saline (PBS) three times for 5 min each. Endogenous peroxidase activity was blocked by 3% H₂O₂ solution for 10 minutes, and then the sections were rinsed again with PBS. The sections were then put in 0.01M citrate buffer and heated for 20 min in a microwave to retrieve antigen. When the sections were cool to room temperature, they were blocked with 10% fetal bovine serum (FBS) for 30 min to prevent nonspecific staining. After being washed with PBS, tissue sections were incubated with TRIM24 rabbit polyclonal antibody (1:1000 dilution) (14208-1-AP; Proteintech, Chicago, IL, USA) or RARα rabbit polyclonal antibody (1:150 dilution) (sc-551; Santa Cruz Biotechnology, Santa Cruz, CA, USA) overnight at 4 ^oC. Negative control was established by omitting the primary antibody. After washing, the sections were incubated with HRP polymer conjugated Pika universal antibody (Dako, Glopstrup, Denmark) for about 30minutes at room temperature, followed by diaminobenzidine (DAB) to develop color reaction. Hematoxylin was used for counterstaining, and then the sections were dehydrated in ethanol before mounting.

Two experienced pathologists who were blinded to the clinical information evaluated the immunohistochemistry staining independently. The staining intensity (SI) of TRIM24 protein was defined as 0 (negative), 1 (weak), 2 (moderate), 3 (strong), and the percentage of positive staining (PP) was assigned as 1 (1-25%), 2 (26-50%), 3 (51-75%), 4 (76-100%). The final immunoreactive score (IRS) was calculated by the multiplication of SI and PP, from 0 to 12. The patients were stratified into a low expression group (≤ 4) or a high expression group (> 4) based on the final scores [26].

Statistical analysis

The SPSS 16.0 software package (Chicago, IL, USA) and Graph Pad Prism 6 software (www.graphpad.com) were used for all the statistical analysis in this study, and P value <0.05 was considered statistically significant. The Chi-square test was performed to analyze the correlation between TRIM24 or RARα expression level and the various clinicopathological parameters. Comparison of mRNA and protein expression levels of TRIM24 and protein level of RARα in ESCCs and the paired NCETs was carried out using paired-sample t test. Pearson’s correlation analysis was used to examine whether there is a relationship between the expression level of TRIM24 and RARα. In order to estimate the overall survival (OS) and disease-free survival (DFS), the Kaplan-Meier curves and the log-rank test were used. Univariate and multivariate Cox proportional hazards regression analyses were performed to evaluate the potential prognostic biomarkers and clinical factors.