One hundred and fifty NSCLC specimens, including 75 adenocarcinoma (AC) and 75 squamous cell carcinoma (SCC), were obtained from patients undergoing surgical resection due to lung cancer between July 2008 and February 2016 at the department of Cardiothoracic Surgery, of the Affiliated Hospital of Xuzhou Medical University (Jiangsu, China). All patients had definite pathological diagnosis after surgery. Patients who received adjuvant therapy were excluded. Written informed consent was obtained from all patients before tissue acquisition. Staging was performed after pathological examination of the formalin-fixed specimens according to the current seventh edition of the TNM classification. The clinical and biological data from the patients are listed in Table I. All patients were followed up every 3-6 months for two years post-operatively. After the second year, every patient was followed-up on a yearly basis until death or the last date of follow-up (November 1st, 2017). The median follow up period for the population studied was 54.5 months (range, 20-89 months). This study gained approval from the Ethics Committee of the Affiliated Hospital of Xuzhou Medical University.

Immunostainings were performed as per the standard protocols. Briefly, de-paraffinized in xylene, rehydrated using graded alcohols firstly, 4 µm thick tissue sections, then, were incubated in acitrate buffer at a pH of 6.1 for 15 min to retrieve the antigen. To remove the endogenous peroxidase activity and reduce background non-specific staining, sections were treated with freshly prepared 3% hydrogen peroxide for 10 min and incubated with 10% goat non-immune serum for 20 min. Thenceforth, after incubation with the goat anti-VEGI polyclonal antibody (1:100, cat. no. sc-23185, Santa Cruz Biotechnology, Inc.) or mouse anti-CD31 monoclonal antibody (1:80, 3528, Cell Signaling Technology) at 37˚C for 1 h (h), the sections were incubated with biotin-labeled secondary antibody (Invitrogen) at 37˚C for 20 min, and then incubated with HRP-conjugated streptavidin (Invitrogen) at 37˚C for 30 min. The color development was performed with a DAB Substrate kit (Dako). At last, the sections were counterstained with dehydrated, cleared and mounted hematoxylin. Appropriate negative controls were carried out by excluding the primary antibody and/or replacing it with an irrelevant anti-serum (no data shown).

The evaluation of the immune histochmical staining of VEGI was based on the calculation of the percentage and intensity of the positive cells (per section by two independent pathologists). On the basis of the percentage of the positive tumor cells, the immunoreactivity of the tumor cells for VEGI was scored as follows: 0=no cell stained, 1=1 to 25% of cells stained, 2=26 to 50% of cells stained, 3=51 to 75% of cells stained and 4=more than 75% of cells stained. The intensity of positive tumor cells was graded as follows: 0 for negative staining, 1 for weak staining, 2 for moderate staining, and 3 for strong staining. Scores for the intensity and percentage of positive cells were multiplied. Finally, when the total score was below 2, the classification of the expression of VEGI was considered as negative/weak, while when the total score was above 3, it was classified as moderate/high. He expression of VEGI was negative/weak moderate/high if the total score was ≥3.

The immunohistochemical staining of CD31 was used for microvascular density (MVD) counting as described previously (17). Briefly, the vessel hot spot regions with immune reactivity against CD31 were identified under magnification, x100 and the number of MVD was counted manually in 5 hot spots selected randomly at x400.

All statistical analyses were achieved with SPSS v16.0 software (SPSS, Inc.). Chi-square test was used to assess the correlation between VEGI protein expression and clinicopathological variables. The Student's t-test was evaluated the difference of MVD in different group. Survival curves were constructed by means of the Kaplan-Meier and the differences between the curves were compared by the log-rank test. P<0.05 was considered to indicate a statistically significant difference.

The immunostaining of VEGI protein was represented as particles in the cytoplasm and the cell membrane which was 31.3% (47/150) in NSCLC patients (Fig. 1A and C). However, the VEGI expression was lost or weak in 103 (68.7%) out of 150 NSCLC patients, which was significantly higher in AC, 76.0% (57/75), than in SCC, 61.3% (46/75) (Table I).

To investigate whether VEGI exert an inhibitory effect on tumor-associated blood vessel formation and growth, we estimated the intratumoral microvessel density (MVD) by measuring CD31-positive staining, which was mainly detected in vascular endothelial cell membrane and a lower degree in the cytoplasm. MVD was significantly lower in VEGI-rich group than in the VEGI-poor group (P=0.01; Fig. 2), indicating that VEGI may pose an inhibitory effect on tumor-associated blood vessel formation and growth, at least partially.

The relationship between the VEGI expression and the clinicopathological features was analyzed in 150 NSCLC patients (Table I). Significant correlation between VEGI expression and clinicopathological parameters was not observed. Nevertheless, the down-regulation of VEGI tended to be associated with the histopathological type (P=0.059), lymphovascular invasion type (P=0.065) as well as the lymph node metastasis type (P=0.057), but this trend was likely due to the association with AC.

Subsequently, the association of VEGI with clinicopathological variables classified according to their histology type was investigated (Table II). In AC patients, the down-regulation of VEGI protein was observed in 91.3% (21/23) of cases with lymphovascular invasion, and in 88.6% (31/35) of cases with lymph node metastasis. A strong correlation was found between VEGI expression and lymphovascular invasion (P=0.039) and lymph node metastasis (P=0.017; Table II). No correlation was observed in the squamous carcinoma group.

To further investigate the clinical significance of VEGI expression in lung cancer, the follow-up was performed on the patients. Patients with moderate/high expression of VEGI seemed to have a longer survival period than those with negative/weak VEGI expression, and was correlated with disease-free survival (P=0.021; Fig. 3A). Moreover, analysis of data from a published study including 1926 lung cancer cases, available on the kmplotwebsite (http://www.kmplot.com/analysis/index.php?p=background), revealed that 967 patients who displayed the low expression of VEGI mRNA in lung tumor tissues had shorter disease free time compared with the 959 cases whose VEGI mRNA expression levels were high in lung tumor tissues (Fig. 3B). These data strongly indicate that the expression of VEGI correlates with the prognosis of NSCLC patients.

Then, a further analysis was performed within each NSCLC histopathological type because there was a significant correlation between VEGI expression and lung AC. In AC, patients with negative/weak VEGI expression had poorer overall survival than those with moderate/high VEGI expression (P=0.011; Fig. 3C). In contrast, these effects were not observed in patients with squamous carcinoma (P=0.692; Fig. 3D).