Patients who underwent surgery for both primary CRC and liver metastases between December 2011 and April 2018 at the West China Hospital were enrolled into the present study. The inclusion criterion was the availability of patients' primary and metastatic tumor tissues. The exclusion criteria were as follows: Patients who achieved pathological complete response following neoadjuvant chemotherapy, patients who had other primary tumors or patients with a positive margin. A total of 56 patients were included in the present study. Surgical specimens, including primary tumor tissues and matched metastases were collected in April 2018. The present study was approved by the Ethics Committee of the West China Hospital (approval no. 2017-169) and written informed consent was provided by patients or their family members prior to the study start.

Clinical characteristics and serum indicators at the time of hepatic resection were obtained for subsequent analyses. Patients were prospectively followed-up until February 27, 2019. The median follow-up was 20.6 month (1.0–64.2 months). Progression-free survival (PFS) was the time interval between hepatectomy and the first postoperative disease progression or death. Overall survival (OS) was the time interval between the date of hepatectomy and either the date of death or last follow-up.

A total ofsix markers, including CD133, CD44 (CSC markers), Snail, E-cadherin, β-catenin (EMT markers) and Ki-67 (proliferation index) were selected for use in the present study.

Tissue samples were cut into 4-μm-thick sections, deparaffinized in xylene, dehydrated in a graded ethanol series, placed in ethylenediaminetetraacetic acid buffer (pH 8.0) or citrate buffer (pH 6.0) for antigen retrieval and immersed in a 0.3% hydrogen peroxide solution to inhibit endogenous peroxidase activity. Subsequently, tissue sections were incubated with primary antibodies against CD133 (cat. no. MAB4399-I; 1:100; Merck KGaA), CD44 (cat. no. 3570S; 1:100; Cell Signaling Technology, Inc.), Snail (cat. no. ab180714; 1:100; Abcam), E-cadherin (cat. no. 14472S; 1:75; Cell Signaling Technology, Inc.), β-catenin (cat. no. 8480S; 1:75; Cell Signaling Technology, Inc.) and Ki-67 (cat. no. ab15580; 1:400; Abcam) overnight at 4°C. Following the primary incubation, the sections were washed three times with 0.1 M PBS (pH 7.4) and incubated with PV6001 Two-Step Immunohistochemistry Detection Reagent (ZSJQ-BIO) for 1 h at 37°C. The slides were subsequently stained with 3,3′-diaminobenzidine for 3 min at room temperature and counterstained with 0.1% hematoxylin for 3 min at room temperature, washed under running tap water, dehydrated in ethanol and cleared in xylene.

The expression levels of the antigens were blindly assessed by two investigators using light microscope (at ×400 magnification). For CD133 (29), membranous and cytoplasm staining were considered (<10% as low expression vs. ≥10% as high expression). For CD44 (30), only membranous staining was considered (<50% as low expression vs. ≥50% as high expression). For Snail staining (31), nuclear staining was considered(<1% as low expression vs. ≥1% as high expression). E-cadherin (32) staining was defined as detectable immunoreactions in cell membranes. The product of the intensity (0, negative; 1, weak and 2, strong) and percentage (1, 25%; 2, 26–50%; 3, 51–75% and 4, 76–100%) was used as the final score (0-1 as low expression vs. 3–8 as high expression). Ki-67 staining was only expressed in the nuclei (<50% as low expression vs. ≥50% as high expression). Considering the complexity of β-catenin expression and the significance of ectopic β-catenin from the membrane to the nuclei (33), samples were divided into normal and abnormal groups. Abnormal expression was associated with decreased intercellular adhesion and activation of downstream tumor proliferation-related target genes, resulting in tumorigenesis and metastasis (34). Normal expression was defined as positive membranous staining in ≥70% of cells and positive cytoplasmic/nuclear staining in <10% of cells. Abnormal expression was considered in the absence of membranous staining or positive membranous staining in <70% of cells or positive cytoplasmic/nuclear staining positive in ≥10% of cells.

Statistical analysis was performed using SPSS 20.0 (IBM Corp.). Classification data were analyzed using the χ2 and Fisher exact tests. McNemar's test was used for paired data selection, while the rank-sum test was used for nonparametric data and grade variables. Univariate analysis was performed using the Kaplan-Meier method and log-rank test following liver metastasis resection. Multivariate Cox regression analysis was performed to determine the prognostic value. P<0.05 was considered to indicate a statistically significant difference.

A total of 56 patients with primary and secondary CRC were included in the present study. Patient characteristics are summarized in Table I. A total of 46 patients had liver metastases diagnosed at the time of CRC diagnosis or before, of which 16 patients had isolated metastases and 40 patients had multiple lesions (no more than five). A total of 31 patients underwent simultaneous resection of the primary tumor and metastases. A total of 31 patients received neoadjuvant chemotherapy, while 41 patients received adjuvant chemotherapy.

The median follow-up time was 20.6 months, and 18 patients died. At the end of the study, the median OS time following liver metastasectomy was 34.4 months [95% confidence interval (CI), 28.5–40.2 months; Fig. 1A] and the median PFS time was 11.0 months (95% CI, 8.4–13.7 months; Fig. 1B).

Samples with low quality staining were excluded from the analyses. In the 112 tissue samples, the high expression rates of Ki-67, CD133, CD44, Snail, E-cadherin and β-catenin were 40.2, 49.1, 16.1, 61.6, 48.2 and 76.8%, respectively (Fig. 2). The expression levels of the markers were similar between the primary tumors and matched metastases (Table II), and the consistency rates ranged from 51.8–78.6%. Among these, CD44 expression was the most similar between the primary tumors and matched metastases (consistency rate of 78.6%). Consistent with this result, McNemar's test demonstrated no significant difference (P_Ki-67, 0.541; P_CD133, 0.248; P_CD44, 1.000; P_Snail, 0.424; P_E-cadherin, 1.000 and P_β-catenin, 1.000). In addition, the rank-sum test revealed no significant variation trends between the target proteins (_PKi-67, 0.414; P_CD133, 0.178; P_CD44, 1.000; P_Snail, 0.317; P_E-cadherin, 1.000 and P_β-catenin, 1.000).

High CD133 expression in primary tumors was significantly associated with tumor location of the colon (P=0.013), while high CD133 expression in liver metastases was significantly associated with age (P=0.003) and positive lymph nodes (P=0.018). Furthermore, Snail expression in liver metastases was significantly associated with more metastases (P=0.028) and a lower differentiation degree (P=0.012). No other clinical features were significantly associated with the target proteins (Tables SI–VI).

Kaplan-Meier survival analysis demonstrated that the maximum diameter of liver metastases was <5 cm (P=0.023; Fig. 3A) and low Snail expression in metastases was significantly associated with a longer OS time following resection of colorectal liver metastases (P=0.029; Fig. 3B). In addition, N0 stage (P=0.021; Fig. 3C) and low CEA expression (P=0.007; Fig. 3D) were also associated with longer PFS times. Multivariate Cox regression analysis demonstrated that the maximum diameter of liver metastases [hazard ratio (HR), 3.447; 95% CI, 1.154–10.041; P=0.019) and Snail expression in liver metastases (HR, 3.405; 95% CI, 1.225–9.697; P=0.026) can be used as independent prognostic factors for OS following resection of colorectal liver metastases (Table III).