Serum CCL20 combined with IL-17A as early diagnostic and prognostic biomarkers for human colorectal cancer

The whole study consisted of five phases: the discovery phase, training phase, validation phase, prognostic evaluation and functional phase (Fig. 1). In the discovery phase, serum samples were collected from 3 CRC patients and 3 healthy controls. We conducted a multiplex assay on each of the samples to identify the potential biomarkers. We then proceeded to conduct a pan-cancer screening of inflammatory factors (CCL/CXCL chemokines and interleukins) using data from 18 cancer types from TCGA and identified CCL20 and IL-17A as biomarker candidates. We confirmed the identified targeted biomarkers in the training phase by making use of an independent training cohort of 112 CRC patients and 151 controls (59 HD, 52 colitis and 40 colorectal adenoma patients) by Elisa and immunohistochemistry (IHC). The logistic regression model was adopted to develop the panel of the targeted biomarkers. To further evaluate the diagnostic accuracy of the model, we used receiver-operating characteristic (ROC) curves. Subsequently, in the validation phase, the performance of the logistic model was validated by differentiating CRC patients from the control group in another independent validation cohort including 75 CRC patients and 35 healthy controls, 20 colitis and 35 colorectal adenoma patients.

In the prognostic evaluation phase we evaluated the value of the targeted markers, conducting overall survival analysis using independent samples from the training cohort of 112 CRC serum and 59 tumor samples from CRC patients. Finally, in the functional phase, we used Gene ontology (GO) analyses and Gene set enrichment analyses (GSEA) to evaluate the function of the candidate genes. TCGA database were used to assess the correlation between the candidate genes and the genes related to cancer progression.

We collected serum samples of healthy controls, and untreated colitis, colorectal adenoma, and CRC patients. The blood samples were obtained from the Departments of Oncology and Gastroenterology of the First Affiliated Hospital of Zhengzhou University from January 2013 to December 2015. The healthy controls, colitis, and colorectal adenoma patients were enrolled from the Department of Physical Examination Center and Clinical Laboratory of the First Affiliated Hospital of Zhengzhou University. The paraffin section of tissue samples from CRC patients (n = 59) diagnosed between 2011 and 2013 were obtained from the Pathology Department. Patients were classified according to UICC-TNM classification and were graded according to the WHO classification criteria as well, moderately, or poorly differentiated. All CRC patients were histologically diagnosed. Age- and sex-matched controls were selected (Additional file 1: Table S1). Serum was transferred to 1.5 ml tubes and stored at − 80 °C for further processing. Protocols for this study were approved by the local ethics committee (Ethical approval number: Science-2010-LW-1213), and an informed consent was obtained from each patient and volunteer, with available follow-up information.

Genes with higher expression in CRC tumor tissue, were screened by The Cancer Genome Atlas (TCGA) database (http://​www.​cancaergenome.​nih.​gov), which provided data of 18 cancer types. Moreover, the CRC transcriptional expression data (GDS4718) were downloaded from the Gene Expression Omnibus (GEO) database for the correlation analyses. It contained 19 normal colon mucosa (N) and 19 primary tumor (Ca) samples.

To identify which factor was highly expressed in CRC patients compared to healthy donors, a multiplex assay was used. The levels of cytokines and chemokines in sera were analyzed using a multi-analyze flow assay kit (BioLegend, USA), including 13 human cytokines (Cat. 740001) and 13 human chemokines (Cat. 740003), according to the manufacturer’s instructions.

Cytokine levels were measured using the highly sensitive Elisa kits (R&D Systems, Minneapolis, MN, USA), specific for human cytokines, according to the manufacturer’s instructions. Detailed methods are described in Additional file 2: Methods section.

Immunohistochemistry was performed as previously described [20]. Mouse anti-CCL20 and rabbit anti-human IL-17A (1:300; Abcam, USA) were used as primary antibodies. Detailed methods are described in Additional file 2: Methods section.

To identify potential CRC biomarkers, we examined the levels of serum inflammatory cytokines and chemokines in CRC patients and HD using a multiplex assay. The serum levels of CCL20, CXCL10, CCL2, IL-9, IL-21, and IL-17A were significantly higher in CRC patients than in HD (fold change ≥ 2, p < 0.05) (Fig. 2a). To determine the highly expressed genes in CRC patients, pan-cancer screening of CCL/CXCL chemokine and interleukin families was performed for tumor tissues from 18 cancer types via TCGA data. We observed that CCL20 and IL-17A expression in colon adenocarcinoma (COAD) and rectal adenocarcinoma (READ) tissues were higher than the expression in other cancer tissues (Fig. 2b and Additional file 3: Fig. S1A). After comparing expression levels between cancer tissues and the corresponding normal tissues, we observed that the levels of both markers were higher in CRC cancer tissues than in normal tissues (Fig. 2c, d). After superimposing results from Fig. 2a, b, we realized that CCL20 and IL-17A may exhibit good performances in predicting the diagnosis and prognosis of CRC patients. As a result, CCL20 and IL-17A levels were identified as candidate CRC biomarkers.

To verify the serum levels of CCL20 and IL-17A in CRC patients, 112 samples from initially treated CRC patients and 151 controls (59 HD, 52 colitis, 40 colorectal adenoma patients) were collected. There were no significant differences in the age and sex of the groups in this cohort (Additional file 1: Table S1). As shown in Fig. 3a, serum CCL20 and IL-17A levels in CRC patients were significantly higher than those in the controls (p < 0.0001). The two markers expression in CRC patients were also higher than that in other cancer patients (Additional file 3: Fig. S1B). Moreover, CCL20 and IL-17A levels were elevated in CRC patients with advanced stages (p < 0.01, p < 0.05; Fig. 3b). Spearman correlation analysis indicated a significant positive correlation between these two biomarkers (r = 0.325, p = 0.025; Fig. 3c). Furthermore, we observed that the levels of the two biomarkers were significantly higher in tumor tissues when compared with the levels in the peritumoral tissues (p < 0.001; Fig. 3d). CCL20 was highly expressed in the cytoplasm of tumor cells, and IL-17A was mainly from infiltrating lymphocytes (Fig. 3d). CCL20 and IL-17A levels were also significantly higher in the advanced stage than in the early stage of CRC (p < 0.05, p < 0.01; Fig. 3e). A similar correlation analysis was performed, and a positive correlation was observed between the protein levels of CCL20 and IL-17A in tumor tissues (r = 0.361, p = 0.0025; Fig. 3f). This correlation was confirmed by the mRNA levels obtained from the GEO (r = 0.720, p < 0.0005; Fig. 3f) and TCGA (r = 0.588, p < 0.0001; Fig. 3f) databases. These data strongly confirmed that expression of CCL20 and IL-17A were higher in CRC patients.

112 CRC samples were collected to evaluate the correlation of serum CCL20 and IL-17A levels with the clinicopathological characteristics of patients. As shown in Additional file 1: Tables S2, S3, higher expression of CCL20 and IL-17A together or respectively were significantly associated with higher T stage, N stage, M stage, overall stage, deep invasion, liver metastasis, larger tumor size, adenocarcinoma type, and CEA values. Our data suggested potential roles of CCL20 and IL-17A in CRC progression.

The efficacy of serum CCL20 and IL-17A in discriminating between CRC patients and controls was verified using the ROC curve. As shown in Fig. 4a, the efficacy of CCL20 only (AUC = 0.936; cut-off value: 89.64; sensitivity: 93.1%; specificity: 86.3%) or IL-17A only (AUC = 0.879; cut-off value: 4.7962; sensitivity: 24.8%; specificity: 98.3%) was better than that of CEA (AUC = 0.601; sensitivity: 24.8%; specificity: 98.3%). After combining CCL20 with IL-17A in a logistic regression model, the screening efficacy (AUC = 0.988; cut-off value: 0.54; positive value: 22.4%; negative predictive value: 95.9%; sensitivity: 96.1%; specificity: 96.5% for the combination; Fig. 4a) was better than their individual efficacies (Combination vs. CCL20: p = 0.015 and combination vs. IL-17A: p = 0.0001). The predicted probability of CRC diagnosis from the stepwise logistic regression model was calculated as follows:

Furthermore, we evaluated the diagnostic performance of the established panel in distinguishing CRC patients with early TNM stage from controls. Similar results were obtained in patients with early stages (I/II). The AUC values for CCL20 and IL-17A were 0.935 (cut-off value: 89.68; sensitivity: 97.1%; specificity: 79.7%) and 0.876 (cut-off value: 4.0969; sensitivity: 80.4%; specificity: 84.7%) respectively. The combination of the two biomarkers robustly increased the AUC to 0.976 (cut-off value: 0.1878; sensitivity: 95.1%; specificity: 93.2%). The differences between them were statistically significant (combination vs. CCL20: p = 0.0431; combination vs. IL-17A: p = 0.003) (Fig. 4b). Then, we evaluated the diagnostic accuracy of the established panel according to the CEA levels. In the high-CEA level (> 5 ng/ml) group, the AUC of the panel was 0.927 (sensitivity: 92.8%; specificity: 91.2%; Fig. 4c). In the low-CEA level (< 5 ng/ml) group, the AUC of the panel was 0.976 (sensitivity: 95.1%; specificity: 94.8%; Fig. 4d). Collectively, our data implied that serum levels of CCL20-IL-17A panels used as a screening protocol may improve the efficacy of early diagnosis in CRC patients.

The model was validated in another cohort consisting of 75 patients with early stage CRC and 90 controls (35 HD, 20 colitis, and 35 colorectal adenoma patients) (Additional file 1: Table S1). As shown in Fig. 5a, the AUC value of the model (AUC = 0.976; sensitivity: 93.3%; specificity: 93.3%) was better than CEA (AUC = 0.596; sensitivity: 46.7%; specificity: 86.7%). Simultaneously, in the high-CEA level (> 5 ng/ml) group, the AUC of the panel was 0.959 (sensitivity: 88.6%; specificity: 91.7%; Fig. 5b). In the low-CEA level (< 5 ng/ml) group, the AUC of the panel was 0.981 (sensitivity: 95.0%; specificity: 96.7%; Fig. 5c). These results confirmed that the serum levels of CCL20-IL-17A panel-based screening protocol may improve the efficacy of early diagnosis in CRC patients.

The overall survival (OS) times of CRC patients with lower serum level of CCL20 were longer compared with those with higher CCL20 expression (p = 0.0288, Fig. 6a). IL-17A expression levels were equally inversely associated with outcomes of CRC patients (p = 0.0193, Fig. 6b). Comparing to the patients with low expression of both biomarkers, patients with high expression of both biomarkers showed a evidently shorter OS (p = 0.0147), whereas patients with only one high-risk biomarker have no significant difference (Fig. 6c). It indicated that the two-biomarkers signature exhibited greater prognostic performance than one-single-biomarker.

As shown in Additional file 1: Table S4, univariate analysis revealed that N stage (hazard ratio = 2.60, p = 0.005), M stage (hazard ratio = 1.99, p = 0.03), liver metastasis (hazard ratio = 2.32, p = 0.04), age (hazard ratio = 1.97, p = 0.04), site of lesion (hazard ratio = 1.52, p = 0.04), overall stage (hazard ratio = 1.69, p = 0.02), serum CCL20 levels (hazard ratio = 2.32, p = 0.02), serum IL-17A levels (hazard ratio = 2.32, p = 0.014), and the combination of CCL20 and IL-17A (hazard ratio = 2.46, p = 0.03) were significantly associated with poor prognosis. Interestingly, a multivariate Cox’s regression analysis revealed that age (hazard ratio = 1.98, p = 0.04), site of lesion (hazard ratio = 3.14, p = 0.02), overall stage (hazard ratio = 1.37, p = 0.03), CCL20 (hazard ratio = 1.12, p = 0.001), IL-17A levels (hazard ratio = 1.22, p = 0.04) and the combination of CCL20 and IL-17A (hazard ratio = 1.45, p = 0.04) were independent prognostic factors for the OS of CRC patients (Additional file 1: Table S4). In summary, these results suggest significantly prognostic values of serum CCL20 and IL-17A levels in CRC patients.

To analyze the function of CCL20 and IL-17A, differential expression genes were found according to the high and low expression groups indicated by the GOs. The main GO categories for differential expression genes were related to functions such as biological regulation, response to stimulus, cell–cell signaling, cell development, cell communication, and cell proliferation (Fig. 7a, b). Further analysis of GSEA, a powerful tool in inferring the biological function, was performed. The results indicated that genes that contributed to positive regulation of the canonical WNT signaling pathway, the epithelial to mesenchymal transition, and epithelial cell migration, were significantly enriched in both CCL20 highly expressed (Fig. 7c) and IL-17A highly expressed samples (Fig. 7d) of CRC patients. Finally, we observed that CCL20 and IL-17A levels were both positively associated with levels of CD44 and MMP3, which are closely related to cancer progression (Fig. 7e, f). Collectively, these results suggest that CCL20 and IL-17A levels play an important role in CRC development.