Evaluation of the correlation of vasculogenic mimicry, ALDH1, KAI1 and microvessel density in the prediction of metastasis and prognosis in colorectal carcinoma

We collected samples from all 204 patients (median age: 59.4 years; range: 31–77 years) who were treated for CRC at the Department of Pathology of the First Affiliated Hospital of Bengbu Medical College, from January 2007 to December 2010, along with 204 samples of the corresponding adjacent normal colorectal mucosa tissues (removed the same patient, from surrounding colorectal mucosa tissue at least 5 cm away from the cancer edge). Patients who had received preoperative chemo- or radio-therapy were excluded. All tissue specimens were obtained with patients writing consent and the study was approved by the ethical committee of the Bengbu Medical College and performed in accordance with the guidelines of the Declaration of Helsinki. We collected the entirely clinicopathological and follow-up data (at 6 months intervals by mail, phone or social application). Overall survival (OS) time was counted from the patients operation date to his/her death date or January 2015 (mean OS: 51.6 months; range: 8–96 months). Grade of tumor differentiation was according to WHO (World Health Organization) standard. Tumor-node –metastasis stage was assessed according to the 7^th edition of the AJCC (American Joint Committee on Cancer). Other characteristics see Table 1.

Immunohistochemistry was conducted according to the guideline of Elivision™ Plus detection kit instructions (Lab Vision, USA). All CRC- and corresponding normal colorectal mucosa tissues were fixed in 10% buffered formalin and embedded in paraffin. Then continuous 4 μm thick tissue sections were cut. All specimens were deparaffinized and dehydrated with xylene and graded alcohol, subsequently washed for 10 min with PBS (phosphate buffer solution, pH 7.2). Endogenous peroxidase activity was quenched by incubation of samples in methanol containing 3% H₂O₂ for 10 min at room temperature (RT), then placed in citrate buffer (pH 6.0) and heated to 95 °C for 30 min for antigen repair. After several washes with PBS, all samples were blocked with goat serum for 20 min at RT, then incubated with mouse monoclonal antibody against human ALDH1 (Abcam, USA) or CD34 (Abcam, USA) or KAI1 (Abcam, USA) for 1 h at 37 °C. Microvessel density (MVD) was determined by the number of small CD34 positive vessels counted. All sections were performed periodic acid-Schiff (PAS) -CD34 dual staining to characterize endothelial cells in glycosylated basement membranes of vessels, as well as vasculogenic-like structures [19]. Furthermore, there was no necrosis or hemorrhage near the VM channels in cancer tissues. All samples were counterstained with hematoxylin, dehydrated, air-dried and mounted. The method was adopted from Weidner et al. with some modifications to assess the MVD of CRC [36]. A modified Yue’s method was used to evaluate the VM of CRC [37]. ALDH1 positive staining was mainly confined in the cytoplasm of cancer cells; KAI1 positive staining was mainly confined in the membrane and cytoplasm of cancer cell. Negative controls were prepared by leaving out primary antibodies from the staining procedure.

Immunotaining findings were explained semi-quantitatively by two independent pathologists who were blind to the clinical, pathological and follow-up data. To avoid the intratumoral heterogeneity of antibodies expression, ten representative areas at high-power-fields (HPF) from different areas of each CRC’s section were detected. immunohistochemistry results were counted according to the extent and intensity. The immunostaining intensity scores were graded as follows: 0, none; 1, weak; 2, moderate; and 3, strong. The immunostaining extent scores were graded as follows: 1, <11%; 2, 11–50%; 3, 51–75%; and 4, >75%. Then, the intensity and extent scores were multiplied to reach a final score that ranged from 0 to 12. The scores ≥3 was considered positive. For samples that were positive for both ALDH1 and KAI1, an average of the final score of each slide was taken.

To assess the contributions of ALDH1, VM, MVD and KAI1 to CRC, the results thereof were immunohistochemically assessed for both CRC and normal colorectal mucosa tissue samples. These data were then compared to the clinicopathological variables. The positive rate of ALDH1 expression in the CRC samples (73.5%, 150/204) was significantly higher than that in the control normal tissues (6.9%, 14/204; P < 0.001; Fig. 1a and b). The positive expression rate of ALDH1 in CRC was positively correlated with tumor invasion, lymph node metastasis, distant metastasis and TNM stage, but not with patients age, gender, tumor size, grade or location (Table 2).

Similar to ALDH1, the positive rate of VM (Small vessel-like lumen in CRC that were PAS-positive but CD34-negative were to be VM. The VM channels pattern included linear, tubular, and network and so on.) was significantly higher in CRC (36.8%, 75/204) than that in the control tissues (0%, 0/204; P < 0.001, Fig. 1c and d). The positive rate of VM in CRC was positively correlated with tumor invasion, LNM, distant metastasis and TNM stage, but not patients age, gender, tumor size, grade or location (Table 2). And the positive staining of MVD scores were found to be significantly correlated with tumor invasion, LNM, distant metastasis and TNM stage in CRC. However, the scores of MVD were no significant association with patient age, gender, tumor size, tumor grade and location (Table 2).

The positive rate of KAI1 expression was significantly lower in CRC tissues (31.9%, 65/204) than that in control normal tissues (98.0%, 200/204; P < 0.001, Fig. 1e and f). The positive rate of KAI1 expression was inversely correlated with tumor grade, invasion, LNM, distant metastasis and TNM stage. No correlation was found between KAI1 expression and patients age, gender, tumor size or location (Table 2).

Follow-up data showed that OS time was significantly shorter in CRC patients with positive expression of ALDH1 (47.1 ± 22.4 months) compared with those with ALDH1-negative (64.3 ± 21.9 months; log-rank = 16.908, P < 0.001; Fig. 2a). Similarly, the OS time of VM-positive patients (34.7 ± 19.0 months) was significantly lower than those of VM-negative patients (61.5 ± 20.0 months; log-rank = 86.416, P < 0.001; Fig. 2b). The OS time of MVD-positive (the mean score of MVD is 21.2 ± 5.8, so MVD score ≥21 is considered positive, MVD score <21 is considered negative) patients (44.9 ± 22.8 months) was significantly shorter than those who were MVD-negative group (59.4 ± 21.9 months; log-rank = 15.610, P < 0.001; Fig. 2c). The OS time of KAI1-positive patients (70.4 ± 16.1 months) was significantly longer than those who were KAI1-negative (42.9 ± 21.2 months; log-rank = 60.613, P < 0.001; Fig. 2d). The combination of KAI1 negative expression and positive expression of ALDH1, VM and MVD had a poorer prognosis than did the reverse combination (log-rank = 97.184, P < 0.001; Fig. 2e). In the univariate analysis, OS time was significantly correlated with clinicopathological variables, including invasion (P = 0.002, log-rank = 9.604), LNM (P < 0.001, log-rank = 19.908), and TNM stage (P < 0.001, log-rank = 53.120) (Table 3)

Multivariate analysis suggested that ALDH1 and KAI1 positive expression, VM, invasion, as well as TNM stage, were independent prognostic indicators for CRC (Table 4).

Spearman correlation coefficient analysis indicated a negative association between the positive expression of KAI1 and that of ALDH1 (r = −0.305, P < 0.001), or VM (r = −0.369, P < 0.001), or MVD (r = −0.458, P < 0.001). Expression of ALDH1 and that of VM (r = 0.181, P = 0.010), and MVD scores (r = 0.242, P < 0.001) were positively correlated, as were VM and MVD scores (r = 0.386, P < 0.001; Table 5).