Abstract
Purpose: Cell-free Bmi-1 mRNA is stably detectable in the serum/plasma and is associated with the development and progression of some tumors. Previous methods detecting extracellular Bmi-1 mRNA with RNA extraction are inefficient. This study developed a novel reverse transcription quantitative PCR (RT-qPCR) approach directly applied in serum (RT-qPCR-D) to quantify Bmi-1 mRNA, and assessed its diagnostic and prognostic potential in colorectal cancer.
Experimental Design: The feasibility of the RT-qPCR-D method was first analyzed in 50 serum samples. Then, using the RT-qPCR-D method, Bmi-1 mRNA expression was validated in serum from an independent cohort of patients with 87 normal colonoscopy, 76 hyperplastic polyp, 82 inflammatory bowel disease, 68 adenoma, and 158 colorectal cancer. Receiver operating characteristic (ROC) curves and Cox analyses were used to evaluate its diagnosis and prognosis value, respectively.
Results: In a pilot study, levels of Bmi-1 mRNA were increased in colorectal cancer serum samples detected by RT-qPCR-D and significantly associated with results obtained by RT-qPCR. In a validation cohort, serum Bmi-1 mRNA levels were significantly elevated in the colorectal cancer group and the adenoma group when compared with other groups. The area under ROC curve distinguishing colorectal cancer from benign colorectal diseases was 0.888, with 72.2% sensitivity and 94.9% specificity, which was superior to carcinoembryogenic antigen. Bmi-1 mRNA levels were significantly associated with survival. Cox analysis indicated Bmi-1 mRNA was an independent prognostic factor for overall survival.
Conclusions: Detection of cell-free Bmi-1 mRNA in serum by RT-qPCR-D is a simple and noninvasive approach and may be used for colorectal cancer diagnosis and prognosis. Clin Cancer Res; 21(5); 1225–33. ©2014 AACR.
Early detection and removal of cancerous or precancerous lesions is thought to be of critical importance for reducing the incidence and improve the prognosis of colorectal cancer. Evidence showed that Bmi-1 played an important role in colorectal cancer development and progression, and its mRNA had relatively strong stability in serum, which is a promising candidate as a biomarker for the diagnosis and prognosis of colorectal cancer. In this study, we developed a novel RT-qPCR approach (RT-qPCR-D) to directly quantify Bmi-1 mRNA in serum without RNA extraction, which may simplify the procedure and improve overall assay comparison. On the basis of this method, cell-free Bmi-1 mRNA has been found to be increased in colorectal cancer and its precancerous lesions, and associated with metastasis and poor survival. We revealed that detection of cell-free Bmi-1 mRNA in serum by RT-qPCR-D can be used for the early diagnosis and prognosis of colorectal cancer, which may complement and improve on current colorectal cancer detection strategies.
Introduction
Colorectal cancer is one of the most common malignancies worldwide, with about 1.2 million new cases and 608,700 deaths annually (1). Early detection and removal of cancerous or precancerous lesions is thought to be of critical importance for reducing the incidence and improve the prognosis of colorectal cancer (2). To date, fecal occult blood testing (FOBT) is the most widely used technique in early detection of colorectal cancer. However, this test is prone to produce false results due to colorectal tumors intermittent bleeding (3). Colonoscopy examination can offer high diagnostic accuracy, but it is invasive and may give rise to additional complications, which limits to the second-level investigation. Stool DNA test has been under continuous development during the past several years, while the challenges of high cost and limited sensitivity render it impractical for mass cancer screening (4). Thus, the development of a noninvasive, sensitive, and cost-effective method that can complement and improve on current colorectal cancer screening strategies has become a major challenge.
Polycomb group (PcG) proteins are known as epigenetic gene silencers through chromatin modification involved in multicellular development, stem cell biology, and tumorigenesis (5). As a key member of the PcG complex, B-cell–specific moloney murine leukemia virus integration site 1 (Bmi-1), was originally identified as an oncogene cooperating with c-myc in a murine lymphoma model. It contains a conserved N-terminal RING finger domain and a central helix–turn–helix–turn–helix–turn (HTHTHT) motif, which is necessary for telomerase activity induction and cellular immortalization (6). Like other PcG group members, Bmi-1 functions as a transcriptional repressor that regulates inhibitors of cyclin-dependent kinase inhibitor, p16Ink4a, and a p53 regulator, p14Arf, to prevent premature senescence (7). Recently, Bmi-1 overexpression has been found in a variety of malignant tumors (8–12), and silencing endogenous Bmi-1 expression can cause apoptosis and/or senescence of tumor cells (13), suggesting an important role in tumor cell growth and survival. Notably, several reports showed levels of cell-free Bmi-1 mRNA were increased in plasma/serum of some malignant tumors, such as breast cancer (14), cervical cancer (15), and gastric cancer (16), reflecting the presence of cancer, staging, or prognosis. In light of these observations, we hypothesized detection of circulating Bmi-1 mRNA might provide a noninvasive avenue for cancer diagnosis and prognosis. Nonetheless, until now only little is known about the clinical value of extracellular Bmi-1 mRNA in colorectal cancer.
Previously, we established a reverse transcription quantitative real-time PCR (RT-qPCR) assay to detect circulating Bmi-1 mRNA with RNA extraction (15). In the current study, we developed a novel RT-qPCR approach to directly quantify extracellular Bmi-1 mRNA in serum without RNA extraction, which may simplify the procedure and improve overall assay comparison. Then, using a large, independent cohort of patients with normal colonoscopy, hyperplastic polyp, inflammatory bowel disease (IBD), adenoma, and colorectal cancer, we evaluated the clinical significance of cell-free Bmi-1 mRNA as potential biomarkers for colorectal cancer diagnosis and prognosis.
Materials and Methods
Study design, patients, and sample processing
We obtained the approval of the local institutional board of ethics and all subjects provided their written informed consent. This study was designed as pilot phase and subsequent validation phase. For pilot study, serum samples were collected from a group of 50 subjects with normal colonoscopy (n = 10), hyperplastic polyp (n = 10), IBD (n = 10), adenoma (n = 10), and colorectal cancer (n = 10) recruited at Shandong Traffic Hospital to determine the feasibility of the RT-qPCR-D method. In the validation phase, changes of cell-free Bmi-1 mRNA in serum were validated in an independent cohort of 500 patients undergoing colonoscopy in Qilu Hospital, Shandong University between April 2007 and March 2009. All individuals were classified into 5 mutually exclusive categories (normal colonoscopy, hyperplastic polyp, IBD, adenoma, and colorectal cancer) based on the histologic results of the colonoscopy. Patients with colorectal cancer were further confirmed by postoperative histopathologic analyses, and followed up at regular intervals until death or January 2014. Their medical records, such as age, gender, tumor location, tumor size, differentiation, lymphovascular invasion, local invasion, regional lymph nodes metastasis, distant metastasis, and carcinoembryogenic antigen (CEA) levels were recorded. Patients with colorectal cancer with incomplete medical records, lost to follow-up, or withdrawal of consent were excluded from this study. Finally, 169 patients with colorectal cancer met the above criteria and enrolled in this study, and the remaining 331 subjects with normal colonoscopy, hyperplastic polyp, IBD, and adenoma were selected as controls. Colorectal cancer was staged by tumor–node–metastasis (TNM) classification of American Joint Committee on Cancer (AJCC), with stage I 27 cases, stage II 61 cases, stage III 48 cases, and stage IV 33 cases. The presence of distant disease of stage IV patients was determined at the time of diagnosis. Advanced adenoma was defined as those had at least one adenomatous polyp with diameter ≥10 mm, ≥20% villous components, or high-grade dysplasia (17, 18). Tumor resection was performed on all patients with colorectal cancer and 15 cases with stage IV disease underwent simultaneous partial hepatectomy or pneumonectomy. Neoadjuvant chemoradiation was given to 35 cases with rectal cancer, and patients with high-risk stage II, stage III, and IV disease received 5-fluorouracil-based adjuvant chemotherapy.
Blood samples were collected by vena puncture from all subjects before colon preparation, colonoscopy examination, or any treatment, such as surgery, radiotherapy, chemotherapy. The sera were separated within 1 hour via centrifugation at 1,600 × g for 10 minutes followed by another centrifugation at 16,000 × g for 10 minutes at 4°C to eliminate any residual cells, and then stored at −80°C until use.
RT-qPCR with extracted RNA
Total RNA was isolated from serum using the TRIzol method (Invitrogen) and measured by NanoDrop spectrophotometer (Thermo Fisher Scientific), as described in our previous study (15). The final concentration of RNA ranged from 3.72 to 16.41 ng/μL. First-strand cDNA was generated from about 100 ng of RNA in triplicates using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems) in final volume of 20 μL according to the manufacturer's instructions. Reverse transcription condition was 25°C for 10 minutes and 37°C for 120 minutes. Then, 5 μL of 10-fold diluted cDNA were added in a qPCR reaction contained 18 μL of Power SYBR Green PCR Master Mix (Applied Biosystems), and 2 μL of specific intron-spanning primers (Supplementary Table S1). The reaction condition was 95°C for 10 minutes, followed by 40 cycles of 95°C for 15 seconds and 60°C for 1 minute. Each sample in duplicates was assessed in ABI PRISM 7500 Sequence Detection System (Applied Biosystems) with melting curve analysis. A negative control containing all components except cDNA template and a positive control from HT29 colorectal cancer cell line containing the gene of interest were performed simultaneously.
RT-qPCR-D
We prepared the 2× preparation buffer containing 2.5% Tween 20 (EMD Chemicals), 50 mmol/L Tris (Sigma-Aldrich), and 1 mmol/L EDTA (Sigma-Aldrich), as described by Asaga and colleagues (19). First, 5 μL of serum was mixed with an equal volume of 2× preparation buffer. Then, 10 μL of the above mixture was reverse transcribed in triplicates in a 20 μL reaction volume. Next, a 1:10 dilution of RT product was centrifuged at 16,000 × g for 5 minutes, and a 5 μL supernatant solution was used as a cDNA template for qPCR. The reagents and reaction conditions were the same as RT-qPCR.
Primary data processing
The cases in which either of the two reference genes (GAPDH and UBC) mRNA were not found were eliminated from the study. Amplification efficiency (AE) is calculated from the slope of the standard curve (Fig. 1A) generated 10-fold serial dilutions of cDNA from HT29 cell line. The quantification cycle (Cq) of each gene was converted into quantity relative to the calibrator, and corrected for AE. The relative expression level in each sample was recorded as the ratio of Bmi-1 gene expression to the geometric mean of two reference genes (GAPDH and UBC) expression (20).
Standard curves and comparison between RT-qPCR-D and RT-qPCR in detecting cell-free mRNA. A, standard curves for GAPDH, UBC, and Bmi-1. B, comparison of GAPDH and UBC reference genes among normal colonoscopy (NC; n = 10), hyperplastic polyp (HP; n = 10), IBD (n = 10), adenoma (Ad; n = 10), and colorectal cancer (CRC; n = 10) detected by RT-qPCR-D or RT-qPCR. C, relative serum levels of cell-free Bmi-1 mRNA among different groups detected by RT-qPCR-D and RT-qPCR. D, correlation analysis between RT-qPCR-D and RT-qPCR in detecting serum cell–free Bmi-1 mRNA levels. Cq value is the threshold cycle of PCR at which fluorescence is detectable. R2 shows the correlation coefficient between Cq value and serial dilutions of sample. E represents the PCR amplification efficiency.
Standard curves and comparison between RT-qPCR-D and RT-qPCR in detecting cell-free mRNA. A, standard curves for GAPDH, UBC, and Bmi-1. B, comparison of GAPDH and UBC reference genes among normal colonoscopy (NC; n = 10), hyperplastic polyp (HP; n = 10), IBD (n = 10), adenoma (Ad; n = 10), and colorectal cancer (CRC; n = 10) detected by RT-qPCR-D or RT-qPCR. C, relative serum levels of cell-free Bmi-1 mRNA among different groups detected by RT-qPCR-D and RT-qPCR. D, correlation analysis between RT-qPCR-D and RT-qPCR in detecting serum cell–free Bmi-1 mRNA levels. Cq value is the threshold cycle of PCR at which fluorescence is detectable. R2 shows the correlation coefficient between Cq value and serial dilutions of sample. E represents the PCR amplification efficiency.
CEA assay
Levels of carcinoembryogenic antigen (CEA) were measured by electrochemiluminescence with Roche Cobas e601 Analyzer (Roche AG), and the upper limits were defined as 5 ng/mL according to the manufacturer's recommendations.
Statistical analysis
The normality of the distribution of data was analyzed using the Kolmogorov–Smirnov test. The Kruskal–Wallis test was applied for global comparisons of CEA or cell-free mRNA levels among multiple groups, and further post hoc multiple comparisons were performed using the Mann–Whitney U test, with a Bonferroni-adjusted significance level of 0.05/10. A χ2test was used for categorical variables comparison. Spearman correlation examined the association between RT-qPCR-D and RT-qPCR. Survival curves were constructed by the Kaplan–Meier method, and compared by the log-rank test. The Cox proportional-hazards regression model was employed to evaluate the independent prognostic factors. Above statistical analyses were performed by SPSS Statistics 17.0 for Windows (IBM Corporation). To illustrate the diagnostic power of serum tumor markers, receiver operating characteristic (ROC) curves were constructed and the areas under the curve (AUC) with 95% CI were calculated and compared using MedCalc 9.3.9.0 (MedCalc). The Youden index (sensitivity+specificity−1) was used to determine the optimal cutoff point of cell-free Bmi-1 mRNA levels (21). Logistic regression model was used to combine CEA and Bmi-1 mRNA, and generated predicted probability value was used for the ROC curve analysis. The diagnostic performance, such as sensitivity and specificity, was also estimated. Statistical significance was defined as two-sided P value of less than 0.05.
Results
Comparison between RT-qPCR-D and RT-qPCR in detecting cell-free mRNA
The standard curves for GAPDH, UBC, and Bmi-1 showed good linearity between Cq values and the log of sample concentrations (Fig. 1A). All the negative controls gave no detectable value, corroborating the lack of any contamination or nonspecific signal. No significant differences in GAPDH and UBC (Fig. 1B) mRNA expression were found among the different groups by using RT-qPCR-D or RT-qPCR (both at P > 0.05), indicating both genes expressed at a constant level in serum, and could be used as suitable internal controls for normalizing target circulating mRNAs. Next, we performed a pilot study to compare the two techniques by detecting cell-free Bmi-1 mRNA. As shown in Fig. 1C, both assays demonstrated levels of Bmi-1 mRNA in serum were significantly increased in colorectal cancer compared with normal colonoscopy, hyperplastic polyp, IBD, and adenoma (all at P < 0.05, respectively). Moreover, there was a significantly linear correlation in Bmi-1 mRNA levels between RT-qPCR-D and RT-qPCR (r = 0.981, P < 0.001; Fig. 1D).
To evaluate the stability of two methods, the coefficient of variation (CV) was determined on the basis of the values obtained from measurement of 3 replicates from 3 samples with low, moderate and high Bmi-1 mRNA concentration selected from pilot phase within a single assay (intra-assay variation) and the same samples performed on 3 independent PCR runs (inter-assay variation). Results showed CVs of intra-assay and inter-assay of RT-qPCR-D were slightly lower than RT-qPCR, but no statistical significance (P > 0.05; Supplementary Table S2).
Quantitative analysis of serum cell–free Bmi-1 mRNA using RT-qPCR-D in validation phase
Of 500 serum samples in validation phase, 29 cases (5.8%) were excluded from the study because either of the two reference genes (GAPDH and UBC) mRNA was not found. GAPDH and UBC mRNAs were detected in serum of 158 of 169 (93.5%) patients with colorectal cancer and 313 of 331 (94.6%) controls, indicating an acceptable integrity of RNA in serum. No significant difference was observed between patients with colorectal cancer and controls (P > 0.05). Of 313 controls, 87 were normal colonoscopy, 76 were hyperplastic polyp, 82 were IBD, 68 were adenoma. Patients with adenoma were subclassified as advanced adenoma group (23 cases) and nonadvanced adenoma group (45 cases). The distribution of gender and age in each group is presented in Table 1 and showed no significant differences (both at P > 0.05).
Characteristics and levels of cell-free Bmi-1 mRNA and CEA of patients
. | NC . | HP . | IBD . | Ad . | CRC . |
---|---|---|---|---|---|
Pilot phase (case) | 10 | 10 | 10 | 10 | 10 |
Gender (male/female) | 4/6 | 5/5 | 6/4 | 5/5 | 5/5 |
Age (y)a | 62 (53–75) | 65 (56–82) | 59 (52–85) | 67 (55–81) | 66 (58–87) |
Bmi-1 mRNA with RT-qPCRa,b | 0.020 (0.006–0.044) | 0.035 (0.013–0.062) | 0.021 (0.010–0.040) | 0.047 (0.021–0.099) | 0.223 (0.129–0.336) |
Bmi-1 mRNA with RT-qPCR-Da,b | 0.022 (0.007–0.045) | 0.034 (0.014–0.065) | 0.020 (0.008–0.046) | 0.053 (0.017–0.095) | 0.223 (0.129–0.336) |
Validation phase (case) | 87 | 76 | 82 | 68 | 158 |
Gender (male/female) | 45/42 | 41/35 | 47/35 | 38/30 | 82/76 |
Age (y)a | 59 (49–81) | 66 (54–86) | 61 (49–85) | 63 (51–83) | 64 (54–89) |
CEA (ng/mL)a,b | 2.68 (0.92–4.56) | 2.44 (0.90–4.47) | 2.24 (1.01–4.70) | 2.56 (0.79–4.97) | 4.56 (2.44–12.6) |
Bmi-1 mRNAa,b | 0.023 (0.009–0.029) | 0.026 (0.016–0.051) | 0.023 (0.012–0.033) | 0.055 (0.024–0.084) | 0.211 (0.081–0.315) |
. | NC . | HP . | IBD . | Ad . | CRC . |
---|---|---|---|---|---|
Pilot phase (case) | 10 | 10 | 10 | 10 | 10 |
Gender (male/female) | 4/6 | 5/5 | 6/4 | 5/5 | 5/5 |
Age (y)a | 62 (53–75) | 65 (56–82) | 59 (52–85) | 67 (55–81) | 66 (58–87) |
Bmi-1 mRNA with RT-qPCRa,b | 0.020 (0.006–0.044) | 0.035 (0.013–0.062) | 0.021 (0.010–0.040) | 0.047 (0.021–0.099) | 0.223 (0.129–0.336) |
Bmi-1 mRNA with RT-qPCR-Da,b | 0.022 (0.007–0.045) | 0.034 (0.014–0.065) | 0.020 (0.008–0.046) | 0.053 (0.017–0.095) | 0.223 (0.129–0.336) |
Validation phase (case) | 87 | 76 | 82 | 68 | 158 |
Gender (male/female) | 45/42 | 41/35 | 47/35 | 38/30 | 82/76 |
Age (y)a | 59 (49–81) | 66 (54–86) | 61 (49–85) | 63 (51–83) | 64 (54–89) |
CEA (ng/mL)a,b | 2.68 (0.92–4.56) | 2.44 (0.90–4.47) | 2.24 (1.01–4.70) | 2.56 (0.79–4.97) | 4.56 (2.44–12.6) |
Bmi-1 mRNAa,b | 0.023 (0.009–0.029) | 0.026 (0.016–0.051) | 0.023 (0.012–0.033) | 0.055 (0.024–0.084) | 0.211 (0.081–0.315) |
aData are presented as median (interquartile range).
bDate are compared among normal colonoscopy (NC), hyperplastic polyp (HP), IBD, adenoma (Ad) and colorectal cancer (CRC) groups using the Kruskal–Wallis test, P < 0.001.
The Kruskal–Wallis test demonstrated that there was a significant difference in Bmi-1 or CEA among patients with normal colonoscopy, hyperplastic polyp, IBD, adenoma, and colorectal cancer (Table 1; both at P < 0.001). As shown in Fig. 2A, the levels of cell-free Bmi-1 mRNA were significantly higher in colorectal cancer compared with normal colonoscopy, hyperplastic polyp, IBD, and adenoma (all at P < 0.001, respectively); Its levels were significantly elevated in the adenoma group compared with the normal colonoscopy, hyperplastic polyp, and IBD groups (all at P < 0.001, respectively). In the adenoma group, levels of cell-free Bmi-1 mRNA in the advanced adenoma subgroup were 0.085 (0.057–0.113), significantly higher than the nonadvanced adenoma subgroup with 0.042 (0.023–0.059) (P < 0.001; Supplementary Fig. S1). In addition, no significant differences were found among the normal colonoscopy, hyperplastic polyp, and IBD groups (all at P > 0.05, respectively). Likewise, CEA levels were significantly increased in colorectal cancer compared with normal colonoscopy, hyperplastic polyp, IBD, and adenoma (Supplementary Fig. S2; all at P < 0.001, respectively), whereas no significant differences were noted among the normal colonoscopy, hyperplastic polyp, IBD, and adenoma groups (Supplementary Fig. S2; all at P > 0.05, respectively).
Diagnosis performance analyses of cell-free Bmi-1 mRNA using RT-qPCR-D. A, quantitative analyses of serum cell–free Bmi-1 mRNA using RT-qPCR-D in normal colonoscopy (NC; n = 87), hyperplastic polyp (HP; n = 76), IBD (n = 82), adenoma (Ad; n = 68), and colorectal cancer (CRC; n = 158). B and C, ROC curves for discriminating colorectal cancer from normal colonoscopy, hyperplastic polyp, IBD, and adenoma using Bmi-1 (B), CEA (C), and Bmi-1 combined with CEA (D). P value is measured by the Mann–Whitney test. Yellow line represents the optimal cutoff value as 0.109 for discriminating colorectal cancer from normal colonoscopy, hyperplastic polyp, IBD, and adenoma.
Diagnosis performance analyses of cell-free Bmi-1 mRNA using RT-qPCR-D. A, quantitative analyses of serum cell–free Bmi-1 mRNA using RT-qPCR-D in normal colonoscopy (NC; n = 87), hyperplastic polyp (HP; n = 76), IBD (n = 82), adenoma (Ad; n = 68), and colorectal cancer (CRC; n = 158). B and C, ROC curves for discriminating colorectal cancer from normal colonoscopy, hyperplastic polyp, IBD, and adenoma using Bmi-1 (B), CEA (C), and Bmi-1 combined with CEA (D). P value is measured by the Mann–Whitney test. Yellow line represents the optimal cutoff value as 0.109 for discriminating colorectal cancer from normal colonoscopy, hyperplastic polyp, IBD, and adenoma.
Diagnostic performance of cell-free Bmi-1 mRNA for colorectal neoplasia
ROC curves analyses illustrated that the serum levels of cell-free Bmi-1 mRNA were robust in distinguishing colorectal cancer (n = 158) from benign colorectal diseases (n = 313), with an area under the ROC curve (AUC) value of 0.888 (95% CI, 0.856–0.915; Fig. 2B). When the cutoff value was set to the optimal point, 0.109, the sensitivity and specificity was 72.2% and 94.9%, respectively. To further evaluate the diagnostic value of extracellular Bmi-1 mRNA, we also assessed diagnostic performance of CEA for colorectal cancer (Fig. 2C). The AUC for the CEA was 0.691 (95% CI, 0.647–0.732), which was significantly smaller than that for cell-free Bmi-1 mRNA (P < 0.001), indicating that cell-free Bmi-1 mRNA was superior to CEA in discriminating the subjects with or without colorectal cancer. Using the standard cutoff value (5 ng/mL), CEA only generated a sensitivity of 44.4% (64/144) and a specificity of 76.0% (193/254) for the detection of colorectal cancer. Also, we tried to combine CEA with cell-free Bmi-1 mRNA to improve the diagnostic capability for colorectal cancer (Fig. 2D). ROC curves analysis showed the AUC for this combination was 0.901 (95% CI, 0.870–0.926), was significantly larger than that for Bmi-1 mRNA (P = 0.001) or CEA (P < 0.001) only.
When used adenoma (n = 68) as the endpoint for detection compared with normal colonoscopy, hyperplastic polyp, and IBD (n = 245), and cell-free Bmi-1 mRNA only yielded an AUC of 0.722 (95% CI, 0.669–0.771) with 67.6% sensitivity and 75.1% specificity in distinguishing adenoma from others (Supplementary Fig. 3A). Furthermore, cell-free Bmi-1 mRNA yielded an AUC of 0.823 (95% CI, 0.776–0.864) with 78.3% sensitivity and 82.8% specificity in distinguishing patients with advanced adenoma from normal colonoscopy, hyperplastic polyp, IBD, and nonadvanced adenoma (Supplementary Fig. S3B).
Association of cell-free Bmi-1 mRNA with clinicopathologic parameters and survival in patients with colorectal cancer
We analyzed the relationship between cell-free Bmi-1 mRNA levels and clinicopathologic characteristics in colorectal cancer (Table 2). The extracellular Bmi-1 mRNA concentrations were significantly associated with regional lymph nodes metastasis (P = 0.024) and distant metastasis (P = 0.006). While the optimal cutoff value (0.109) of cell-free Bmi-1 mRNA was used to categorize patients with colorectal cancer into high-level (n = 114) or low-level (n = 44) group, χ2 tests revealed there were no differences in clinicopathologic characteristics between patients with colorectal cancer with high versus low Bmi-1 mRNA (all at P > 0.05, respectively).
Associations between cell-free Bmi-1 mRNA levels and clinicopathologic characteristics
. | Cell-free Bmi-1 mRNA levels . | ||||
---|---|---|---|---|---|
Parameters . | Median (interquartile range) . | Pa . | Highc . | Lowc . | Pb . |
Age, y | |||||
<64 | 0.161 (0.072–0.270) | 0.181 | 54 | 23 | 0.580 |
≥64 (median) | 0.218 (0.089–0.337) | 60 | 21 | ||
Gender | 0.759 | 0.514 | |||
Male | 0.214 (0.081–0.315) | 61 | 21 | ||
Female | 0.191 (0.075–0.323) | 53 | 23 | ||
Tumor location | 0.823 | 0.528 | |||
Colon | 0.199 (0.090–0.317) | 71 | 25 | ||
Rectum | 0.218 (0.070–0.315) | 43 | 19 | ||
Tumor size | 0.159 | 0.489 | |||
<4 cm | 0.162 (0.072–0.262) | 68 | 29 | ||
≥4 cm | 0.230 (0.118–0.340) | 46 | 15 | ||
Differentiation | 0.319 | 0.891 | |||
Well | 0.190 (0.065–0.251) | 22 | 10 | ||
Moderate | 0.211 (0.080–0.295) | 62 | 23 | ||
Poor | 0.215 (0.087–0.389) | 30 | 11 | ||
Lymphovascular invasion | 0.756 | 0.600 | |||
No | 0.211 (0.077–0.302) | 70 | 29 | ||
Yes | 0.200 (0.082–0.318) | 44 | 15 | ||
Local invasion | 0.277 | 0.683 | |||
T1–T2 | 0.166 (0.071–0.251) | 30 | 13 | ||
T3–T4 | 0.214 (0.082–0.324) | 84 | 31 | ||
Regional lymph nodes metastasis | 0.024 | 0.263 | |||
No | 0.163 (0.071–0.258) | 64 | 29 | ||
Yes | 0.236 (0.115–0.354) | 50 | 15 | ||
Distant metastasis | 0.006 | 0.199 | |||
No | 0.178 (0.072–0.258) | 88 | 38 | ||
Yes | 0.273 (0.139–0.419) | 26 | 6 | ||
CEA levels | 0.399 | 0.224 | |||
<5 ng/mL | 0.201 (0.070–0.275) | 63 | 29 | ||
≥5 ng/mL | 0.213 (0.115–0.348) | 51 | 15 |
. | Cell-free Bmi-1 mRNA levels . | ||||
---|---|---|---|---|---|
Parameters . | Median (interquartile range) . | Pa . | Highc . | Lowc . | Pb . |
Age, y | |||||
<64 | 0.161 (0.072–0.270) | 0.181 | 54 | 23 | 0.580 |
≥64 (median) | 0.218 (0.089–0.337) | 60 | 21 | ||
Gender | 0.759 | 0.514 | |||
Male | 0.214 (0.081–0.315) | 61 | 21 | ||
Female | 0.191 (0.075–0.323) | 53 | 23 | ||
Tumor location | 0.823 | 0.528 | |||
Colon | 0.199 (0.090–0.317) | 71 | 25 | ||
Rectum | 0.218 (0.070–0.315) | 43 | 19 | ||
Tumor size | 0.159 | 0.489 | |||
<4 cm | 0.162 (0.072–0.262) | 68 | 29 | ||
≥4 cm | 0.230 (0.118–0.340) | 46 | 15 | ||
Differentiation | 0.319 | 0.891 | |||
Well | 0.190 (0.065–0.251) | 22 | 10 | ||
Moderate | 0.211 (0.080–0.295) | 62 | 23 | ||
Poor | 0.215 (0.087–0.389) | 30 | 11 | ||
Lymphovascular invasion | 0.756 | 0.600 | |||
No | 0.211 (0.077–0.302) | 70 | 29 | ||
Yes | 0.200 (0.082–0.318) | 44 | 15 | ||
Local invasion | 0.277 | 0.683 | |||
T1–T2 | 0.166 (0.071–0.251) | 30 | 13 | ||
T3–T4 | 0.214 (0.082–0.324) | 84 | 31 | ||
Regional lymph nodes metastasis | 0.024 | 0.263 | |||
No | 0.163 (0.071–0.258) | 64 | 29 | ||
Yes | 0.236 (0.115–0.354) | 50 | 15 | ||
Distant metastasis | 0.006 | 0.199 | |||
No | 0.178 (0.072–0.258) | 88 | 38 | ||
Yes | 0.273 (0.139–0.419) | 26 | 6 | ||
CEA levels | 0.399 | 0.224 | |||
<5 ng/mL | 0.201 (0.070–0.275) | 63 | 29 | ||
≥5 ng/mL | 0.213 (0.115–0.348) | 51 | 15 |
aP value was estimated by the Mann–Whitney U test or the Kruskal–Wallis test.
bP value was estimated by a χ2 test.
cPatients with colorectal cancer were classified as high or low cell-free Bmi-1 mRNA levels based on the optimal cutoff value (0.109).
One-hundred and fifty eight patients with colorectal cancer were followed up with the mean duration of 42.5 (range 5–63) months, overall survival (OS) was calculated from the time of diagnosis to the date of death or study completion, and the cumulative 5-year OS rate was 46.2%. Patients with colorectal cancer with high Bmi-1 mRNA expression had significantly worse cumulative 5-year OS rate compared with patients low Bmi-1 mRNA expression (Fig. 3A; 37.7% vs. 68.2%; P = 0.001). Next, we classified all patients with colorectal cancer as low or high CEA expression based on the standard cutoff value (5 ng/mL) to further observe whether the combination of these two serum biomarkers might improve the predict value of survival in colorectal cancer. Kaplan–Meier survival curve showed patients with both low expression of CEA and Bmi-1 mRNA had the longest survival times, and the reduced survival was then observed for patients with high CEA expression alone, followed by patients with high cell–free Bmi-1 mRNA expression alone, and the shortest survival times were found in patients with both biomarkers highly expressed (Fig. 3B).
Kaplan–Meier curves for OS according to cell-free Bmi-1 mRNA (A) and combined cell-free Bmi-1 mRNA and CEA (B). Patients with colorectal cancer were classified as negative or positive Bmi-1 expression according to the optimal cutoff value (0.109), and negative or positive CEA expression based on the standard cutoff value (5 ng/mL).
Kaplan–Meier curves for OS according to cell-free Bmi-1 mRNA (A) and combined cell-free Bmi-1 mRNA and CEA (B). Patients with colorectal cancer were classified as negative or positive Bmi-1 expression according to the optimal cutoff value (0.109), and negative or positive CEA expression based on the standard cutoff value (5 ng/mL).
Finally, we employed Cox proportional hazard regression model to evaluate the prognostic factors of colorectal cancer (Table 3). Cox analyses showed OS was significantly associated with cell-free Bmi-1 mRNA (P = 0.002), lymphovascular invasion (P < 0.001), local invasion (P = 0.049), regional lymph nodes metastasis (P < 0.001), distant metastases (P < 0.001), and CEA (P = 0.043) at the univariate level. More important, multivariable analysis demonstrated only cell-free Bmi-1 mRNA (P = 0.030), lymphovascular invasion (P = 0.001), lymph nodes metastasis (P < 0.001), distant metastases (P < 0.001), and CEA (P = 0.035) retained independent prognostic value.
Cox proportional hazards regression model analyses of overall survival in patients with colorectal cancer
. | Univariate analysis . | Multivariate analysis . | ||
---|---|---|---|---|
Parameters . | HR (95% CI) . | P . | HR (95% CI) . | P . |
Age | 1.302 (0.850–1.995) | 0.226 | ||
Gender | 0.915 (0.598–1.401) | 0.683 | ||
Tumor location | 1.135 (0.737–1.748) | 0.565 | ||
Tumor size | 1.481 (0.966–2.270) | 0.223 | ||
Differentiation | 1.004 (0.730–1.381) | 0.979 | ||
Lymphovascular invasion | 2.193 (1.428∼3.367) | <0.001 | 2.073 (1.337∼3.216) | 0.001 |
Local invasion | 1.688 (1.002–2.841) | 0.049 | 1.321 (0.841–2.074) | 0.227 |
Regional lymph nodes metastasis | 3.927 (2.516–6.129) | <0.001 | 2.932 (1.841–4.671) | <0.001 |
Distant metastasis | 4.515 (2.846–7.162) | <0.001 | 3.453 (2.097–5.684) | <0.001 |
CEA | 1.552 (1.014–2.375) | 0.043 | 1.588 (1.032–2.444) | 0.035 |
Bmi-1 mRNA level | 2.533 (1.426–4.499) | 0.002 | 1.932 (1.066–3.501) | 0.030 |
. | Univariate analysis . | Multivariate analysis . | ||
---|---|---|---|---|
Parameters . | HR (95% CI) . | P . | HR (95% CI) . | P . |
Age | 1.302 (0.850–1.995) | 0.226 | ||
Gender | 0.915 (0.598–1.401) | 0.683 | ||
Tumor location | 1.135 (0.737–1.748) | 0.565 | ||
Tumor size | 1.481 (0.966–2.270) | 0.223 | ||
Differentiation | 1.004 (0.730–1.381) | 0.979 | ||
Lymphovascular invasion | 2.193 (1.428∼3.367) | <0.001 | 2.073 (1.337∼3.216) | 0.001 |
Local invasion | 1.688 (1.002–2.841) | 0.049 | 1.321 (0.841–2.074) | 0.227 |
Regional lymph nodes metastasis | 3.927 (2.516–6.129) | <0.001 | 2.932 (1.841–4.671) | <0.001 |
Distant metastasis | 4.515 (2.846–7.162) | <0.001 | 3.453 (2.097–5.684) | <0.001 |
CEA | 1.552 (1.014–2.375) | 0.043 | 1.588 (1.032–2.444) | 0.035 |
Bmi-1 mRNA level | 2.533 (1.426–4.499) | 0.002 | 1.932 (1.066–3.501) | 0.030 |
Discussion
To the best of our knowledge, this is the first report on quantitative assessment of circulating mRNA in serum using RT-qPCR-D without RNA extraction. On the basis of this method, we found significantly elevated levels of cell-free Bmi-1 mRNA in the serum of colorectal neoplasia, and for the first time demonstrated its potential value in early detection of colorectal cancer. Finally, our results showed cell-free Bmi-1 mRNA could be as a prognostic predictor for colorectal cancer.
Despite of high concentration of ribonucleases in peripheral blood, previous studies have confirmed it is feasible to detect circulating mRNA in plasma/serum of tumor patients, especially for colorectal cancer, such as LISCH7 mRNA (22), hTERT mRNA (23), BIRC5 mRNA (24), and so on. Several studies reported these circulating RNA were protected by RNA binding proteins/lipids or embedded into microvesicles (25–27), thus rendering them resistant to the action of ribonucleases. In this study, to directly amplify cell-free mRNAs in serum, samples were first mixed with a preparation buffer, in which Tween-20 can dissociate the lipid-bound RNA from protein (19). Results obtained from RT-qPCR-D showed high association with those detected by RT-qPCR. Moreover, because the direct assay bypassed extraction of circulating RNA and minimized human and mechanical errors, it showed a superior inherent variability and inter-assay reproducibility. Consequently, we think RT-qPCR-D is a robust and reliable method for detection of cell-free mRNAs in serum.
As we know, colorectal cancer development is a multistep process with a series of genetic/epigenetic alterations contributing to normal mucosa–adenoma–cancer sequence (28). As a pivotal transcriptional repressor, Bmi-1 had been reported to be involved in colorectal carcinogenesis by repressing the INK4a/ARF pathway (29). Recently, Sangiorgi and Capecchi (30) found Bmi-1 was expressed in intestinal stem cells, which essential for self-renewal of colorectal cancer–initiating cells and initiation of colorectal cancer. Meanwhile, downregulation of Bmi-1 could inhibit cell self-renewal, restrain cell proliferation, stop cell cycle, and promote cell apoptosis in colorectal cancer cell lines (31, 32). Collectively, these reports suggested that Bmi-1 played an important role in colorectal cancer tumorigenesis. In the current study, elevated Bmi-1 in pilot phase was also observed in validation phase when compared with normal colonoscopy, hyperplastic polyp, IBD, and adenoma, which was in line with the results found in colorectal cancer tissues based on immunohistostaining (33). Interestingly, increased cell-free Bmi-1 mRNA was also found in the serum of advanced adenoma recognized as a precursor to colorectal cancer, which was similar to the phenomena observed in precancerous lesions of some other tumors, such as cervical cancer (15), gastric cancer (9), bronchial squamous cell carcinoma (10), esophageal carcinoma (11), oral squamous cell carcinomas (12), suggesting the dysregulated expression of Bmi-1 might be an early event.
So far, the overall sensitivity of CEA, a traditional serological tumor marker incorporated into clinical practice for colorectal cancer, only varied between 43% and 69%. This was also confirmed in our study with a sensitivity of 44.4% and a specificity of 76.0%. Compared with this benchmark, cell-free Bmi-1 mRNA in the current study demonstrated improved sensitivity and specificity, associated with a markedly higher AUC value of 0.888 in distinguishing colorectal cancer from benign colorectal diseases. As there was no significant association between cell-free Bmi-1 mRNA and CEA in serum, we further observed whether combination of these two markers could improve detection capability. ROC curve showed this combination had better diagnostic performance than that for Bmi-1 mRNA or CEA alone. This may be due to the Bmi-1 mRNA similar levels between the low CEA group and high CEA group in patients with colorectal cancer. Thus, we think detection of cell-free Bmi-1 mRNA may be an effective complement to current colorectal cancer detection strategy. Another interesting finding of current study is that cell-free Bmi-1 mRNA showed some diagnosis power for premalignant adenomas. This is important from a colorectal cancer early intervention perspective as removal of adenomatous resulted in a reduction of colorectal cancer incidence.
In this study, we also showed serum Bmi-1 mRNA concentrations were significantly associated with factors of poor clinical outcome, such as lymph nodes metastasis and distant metastases. This was confirmed by Kaplan–Meier survival curves that patients with serum cell–free Bmi-1 mRNA levels above the cutoff value showed significantly reduced OS than patients with levels below the cutoff value. Taking a step further, Cox proportional hazards regressions model analysis illustrated that high expression of serum Bmi-1 mRNA was an independent prognostic factor as classical prognostic factors like lymphovascular invasion, lymph nodes metastasis or distant metastases. Several investigators have reported the similar findings in other types of tumor. Silva and colleagues (14) quantified circulating Bmi-1 mRNA in plasma of breast cancer, and found its levels were elevated in women with poor prognoses, such as negative progesterone receptors and positive p53 staining, and its expression were associated with shorter OS. In our previous studies, cell-free Bmi-1 mRNA was also found as an independent predictor of prognosis in cervical cancer (15). Thus, we speculated the cell-free Bmi-1 mRNA might also serve as a prognostic biomarker for colorectal cancer. Moreover, this study showed the combinatorial Bmi-1 and CEA improved the power of survival prediction.
To provide more accurate quantification data of target gene, some problems needed to be pointed out. Although Bmi-1 mRNA was detected by a robust direct serum assay with low variation, normalization is also important. In this study, we preliminarily selected two reference genes, GAPDH and UBC, which were stably expressed in colorectal cancer tissues as described previously (34). Our results also showed both of them displayed consistent expression in serum and not influenced by colorectal disease status using RT-qPCR as well as RT-qPCR-D. Therefore, GAPDH and UBC were suitable internal controls for our investigations. Another is caused by variability of PCR amplification efficiencies. We circumvented this concern using the formula with actual amplification efficiencies for calculating various genes expression. Thus, the final Bmi-1 mRNA relative levels were represented as a ratio against the geometric average of GAPDH and UBC, instead of 2−ΔCt, a traditional relative expression method that all amplification efficiencies were considered as 100%.
Although the findings of our current assay are promising, there still existed some limitations. First, the origin and release mechanisms of cell-free Bmi-1 mRNA are not fully understood. Recently, some investigators considered these circulating cell-free RNAs were secreted from tissues and represented some signaling molecules for cellular communications (35, 36). Although the expression of serum Bmi-1 mRNA in our study were line with others reported in colorectal cancer tissues, this need be further confirmed in paired tumor and serum samples. Second, because the patients with premalignant adenomas had not been followed up, it was unclear whether those with higher Bmi-1 mRNA levels have an increased risk of developing colorectal cancer. Third, as Bmi-1 mRNA seemed also to be a marker of other preinvasive or invasive neoplasms, this assay might not be a good screening tool for patients with colorectal cancer (37). Thus, we hope to develop other markers combined with Bmi-1 mRNA for specific detection of colorectal cancer in next study.
In conclusion, this study uses an RT-qPCR-D assay to detect serum cell-free mRNA and finds cell-free Bmi-1 mRNA is a potential noninvasive maker for early diagnosis and prognostic prediction of colorectal cancer. Further multicenter studies that include a higher number of patients collected from several hospitals, or even diverse ethnic populations, are required to validate whether it can be incorporated into routine clinical practice.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: X. Zhang, X. Yang, C. Wang
Development of methodology: X. Zhang, X. Yang, X. Liu, C. Wang
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Zhang, X. Yang, Y. Zhang, X. Liu, L. Wang
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Zhang, X. Yang, Y. Zhang, L. Du
Writing, review, and/or revision of the manuscript: X. Zhang, X. Yang, G. Zheng, Y. Yang, L. Wang, C. Wang
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Zhang, X. Yang, Y. Zhang, G. Zheng, C. Wang
Study supervision: X. Zhang, X. Yang, C. Wang
Acknowledgments
The authors thank Prof. Rifai Nader (Departments of Laboratory Medicine and Pathology, Boston Children's Hospital and Harvard Medical School, Boston, MA) and Dr. Yue Liu (Department of Anesthesia, University of California, San Francisco, CA) for critically reviewing the manuscript and providing helpful suggestions in writing the manuscript.
Grant Support
This work was supported by National Natural Science Foundation of China (No. 81301506; 81300297; 81271916), Research Fund for the Doctoral Program of Higher Education of China (No. 20130131120067), Shandong Province Natural Science Foundation (No. ZR2013HQ063); and Shandong Technological Development Project (STDP, 2013GSF11859).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.