The present study included 97 patients with NSCLC who underwent either lobectomy or pneumonectomy at Southwest Hospital, Xinqiao and Daping Hospitals, between February 2005 and February 2008. Among 97 samples, 34 lung adenocarcinoma samples were sliced into two parts, and one part was placed into the liquid nitrogen immediately, this subset was preserved for extracting mRNA and differential gene screening. The other part was formalin-fixed, paraffin-embedded NSCLC tissues that were retrieved from the files of our pathology department. Pathological stage was re-evaluated and determined with the present TNM classification as revised in the WHO 2004 classification criteria. Tissue blocks containing a representative fraction of the tumor and the tumor-lung parenchyma interface were used. Operative tissues were also embedded with paraffin from the 97 patients with NSCLC. The study was approved by the Ethics Committee (Faculty of Medicine, Third Military Medical University).

Specific pathogen-free female C57BL/6 mice (4–6 weeks old) and female BALB/c mice (6 weeks old) were purchased from the Institute of Experimental Animal of Third Military Medical University (Chongqing, China). All animals had free access to standard laboratory mouse feed and water. The present study was conducted in accordance with the national and regional guidelines for the care and use of laboratory animals.

Murine Lewis Lung Cancer (LLC) and L929 (murine fibroblast) cells were purchased from American Type Culture Collection (ATCC; Manassas, VA, USA), and were maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum (FBS) (both from Invitrogen, Carlsbad, CA, USA). Isolation and culture of LECs was performed as previously described (18). Briefly, female BALB/c mice were intraperitoneally injected with emulsified incomplete Freund's adjuvant (Sigma-Aldrich, St. Louis, MO, USA) to induce lymphangiomas. After 2 months of induction, tumors in the peritoneal cavity were removed and mechanically disrupted. LECs were isolated and resuspended in endothelial cell growth supplement (EBM-2; Cambrex BioScience, Wokingham, UK) with 20% FBS, and 50 ng/ml endothelial cell growth supplement (Cambrex BioScience) at 37°C in a humidified atmosphere of 5% CO₂. LECs were used in appropriate experiments or cultivated till the fourth passage/phase.

LECs (6×10⁴ cells/well) were put into a 24-well plate that had been pre-coated with 100 Matrigel (10 mg/ml; Clontech, Palo Alto, CA, USA) and cultured for 24 h. Transwell upper inserts were then placed into the co-cultivation system. The L929, LLC and Si-IGFBP7 LLC cells, respectively, were seeded into the Transwell upper inserts of chambers consisting of polycarbonate membranes (0.4-mm pore size; Millipore, Billerica, MA, USA). The cells in the upper inserts were LECs at 1×10⁴ cells/well, and were co-cultured for 48 h. Cells that were cultured in basal medium were used as a control. Formation of tube-like structures was monitored by microscopic observation at a magnification of ×100 over 6 different fields of each well. They were then photographed to measure the length of tube-like structures as previously described (19).

Serial 5-µm thick sections prepared from formalin-fixed, paraffin-embedded tissues from radical prostatectomy specimens were used for the study. Tissue blocks that contained the maximum amount of tumor and highest Gleason score were selected for each case, in order to ensure that the representative blocks contained cancer of the same Gleason score as the overall score of the case, but recognizing the limitation of sample variation. Slides from these representative blocks were analyzed. Slides were deparaffinized in xylene twice for 5 min and rehydrated through graded ethanol solutions to distilled water. Antigen retrieval was carried out by heating sections in 0.1 mol/l citrate buffer, pH 6.0, in a pressure steamer for 20 min. Endogenous peroxidase activity was inactivated by incubation in 3% H₂O₂ for 15 min. Non-specific binding sites were blocked using protein block (Dako Corp., Carpinteria, CA, USA) for 20 min. The slides were then incubated sequentially with primary antibody (clone podoplanin, prediluted antibody; Abcam Inc., Cambridge, MA, USA), biotinylated secondary antibody, avidin-peroxidase complex and chromogenic substrate diaminobenzidine. Positive and negative controls were run in parallel with each batch and demonstrated that the procedure functioned properly (20).

At least 6 random fields per cross-section were visualized at a magnification of ×20 and used for image analysis that was performed with the NIS-Elements Advanced Research 2.3 imaging software (image pro-plus), which identifies signals by threshold key intensity values. Furthermore, it permits imposing restrictions to the measurements by excluding false positive signals. Briefly, the number of positive cells expressing a particular antibody was calculated as a percent of the region of interest (ROI), as indicated in the individual figure legend. Co-localization of two antibodies was calculated by converting the area occupied by cells positive for the first antibody into a ROI. Then the percent of cells that were positive for the second antibody was calculated within the ROI (21).

Total RNA was isolated from cultured cells using a single-step procedure with the TRIzol reagent (Invitrogen, Gaithersburg, MD, USA). The cells were lysed to extract the total RNA. The quality of total RNA was excellent, as deemed by measuring the 260/280 nm ratio (>2.0) and was then used for further cDNA synthesis and a chip hybridization procedure according to the manufacturer's instructions (HU133 Plus 2; Affymetrix, Inc., Santa Clara, CA, USA). The microarray data were analyzed at the NetAffx Analysis Center within the Affymetrix website, http://www.affymetrix.com/analysis/index.affx. cDNA was synthesized from 5 µg of total RNA with 200 U of MMLV-RT (Promega, Madison, WI, USA) and 500 ng of a 16-mer oligo(dT). All the primer sequences for the PCR reactions are listed in Table I and the cycling conditions for all of the PCRs were 30 cycles of a denaturation period (94°C/30 sec) that was followed by an annealing period (55°C/30 sec), and an extension period (72°C/40 sec), then a final extension period (72°C/10 min), except for β-actin which contained an annealing period of 60°C/30 sec in a thermal cycler (Bio-Rad MJ Research PCR; Bio-Rad, Waltham, MA, USA). PCR products were separated on a 1.5% agarose gel in a 0.5X TBE buffer and were stained with ethidium bromide and visualized under UV light.

Purified total RNA (5 Ag) was labeled and hybridized onto the Affymetrix U133 Plus 2.0 GeneChip oligonucleotide arrays (Affymetrix, Inc.) according to the manufacturer's instructions. Briefly, hybridization signals were scaled in Affymetrix GCOS software (version 1.1.1) using a scaling factor determined by adjusting the global trimmed mean signal intensity value to 500 for each array and importing them into GeneSpring version 6.2 (Silicon Genetics, Redwood City, CA, USA). Signal intensities were then centered to the 50th percentile of each chip, and for each individual gene, to the median intensity of each specific subset first to minimize the possible technical bias and then to the whole sample set. The intensities of any replicate hybridizations were averaged subsequent to further analysis.

Relative gene copy numbers and gene expression was determined by quantitative real-time PCR using a PRISM 7500 Sequence Detection system (Applied Biosystems) and a QuantiTect SYBR-Green PCR kit (Qiagen, Inc. Valencia, CA, USA). The standard curve method was used to calculate target gene copy number in the tumor cDNA sample normalzed to a repetitive element line-1 and normal reference cDNA. The comparative threshold cycle method was used to calculate gene expression normalized to β-actin as a gene reference and normal human lung RNA was used as an RNA reference. Primers were designed using Primer3 (http://frodo.wi.mit.edu/cgibin/primer3/primer3_www.cgi) and were synthesized by Sangon Biotech Co., Ltd. (Shanghai, China). Primers (Invitrogen, Shanghai, China) used for RT-PCR were: IGFBP7 5′-CTGGGTGCTGGTATCTCCTC-3′ (sense), and 5′-TATAGCTCGGCACCTTCACC-3′ (antisense); VEGF-C 5′-TGTAAAACGACGGCCAGT-3′ (sense), and 5′-CAGGAAACAGCTATGACC-3′ (antisense); GAPDH 5′-GCACCGTCAAGGCTGAGAAC-3′ (sense), and 5′-TGGTGGTGAAGACGCCAGT-3′ (antisense).

Gene Ontology (GO) annotation of the identified proteins was obtained using DAVID (version 6.7); (see http://david.abcc.ncifcrf.gov/summary.jsp.) Differentially expressed proteins were analyzed using Ingenuity Pathway Analysis (IPA; Ingenuity Systems; see www.ingenuity.com). Cell Death, Cellular, Development, Hematological Systems Development and Function, Endocrine System, Disorders, Metabolic Disease, Cell-To-Cell Signaling interactions were generated based on information contained in the Ingenuity Pathways Knowledge Base.

Biological replicates were used for western blot analysis of astroglial differentiation. In brief, lysates from whole cell extracts or membrane pellets containing 20 µg of proteins were subjected to gel electrophoresis. Proteins were then transferred to PVDF membranes (Millipore). The blots were blocked in 4% BSA in TBST solution for 30 min at room temperature and were then incubated at 4°C overnight with the primary antibody. Anti-β-actin (1:4,000) and IGFBP7 (1:1,000) were purchased from Abcam, Inc. After incubation with secondary antibodies (1:2,000, Millipore) at room temperature for 1 h, the blot was visualized by ChemiDoc XRS imaging system (Bio-Rad, Hercules, CA, USA).

Transfection with IGFBP7 siRNA (10 pmole) and irrelevant scrambled control siRNA was performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. LLCs at 50% confluency were treated with siRNA for 24 h in medium lacking 20% serum and antibiotics. Inhibition of IGFBP7 protein expression was observed within 24 h and this knockdown was maintained for at least 72 h after removal of the medium containing the siRNA.

Three statistical methods were used to identify the optimal prognostic gene signature: Cox proportional-hazard regression modeling, bootstrapping, and a 2-fold cross-validation. Hierarchical clustering was done to identify major clusters of gene activation and investigate their associations with patient covariates. The method for developing a risk index is similar to that previously described. Pathway analysis was carried out by first mapping genes to the biological process categories of GO and then calculating the significance of overrepresented categories in the selected gene list. Correlations between podoplanin and the vessel numbers as continuous variables were used to determine positive vessel counts with the Spearman rank correlation test. Categorical data was compared by the χ² or Fishers' exact probability test. Distribution was normal or within the Mann Whitney U test if the sample distribution was asymmetrical. The relationship between lymph vessel variables and lymph node status was analyzed by one-way ANOVA, followed by the Neuman Keuls test.

Podoplanin expression mostly presented in the thin-walled structures. Podoplanin was positive in endothelial cell plasma, in thin-walled lymph vessel, as indicated by the brown-yellow color. Podoplanin positive lymph vessels were both located at the adenocarcinoma interstitium (Fig. 1A) and at the tumor boundary (Fig. 1B). In some cases, low density lymphatic vessels (LVD-_low) appear around tumor cells (Fig. 1C). Podoplanin positive stained cells only appear in thin-walled structures like the lymphatic vessel, but does not stain blood vessels, which are indicated by the presence of blood cells (Fig. 1D).

Table I depicts pathologic N1-2 as significantly different with No patients in assessing peritumoral lymphatic vessel density, but not intratumoral lymphatic vessel density (P<0.05). PtLVD was 24.1±8.2/field in clinic pathlogical stage III–IV, which was significantly different with stage I–II (P<0.05), the mean of ptLVD is 18.4±5.6/field. The median ptLVD was 18.7/field, which is separated by LVD-_high and LVD-_low group. The clinicopathological features show that for ptLVD there was no significant difference with age, gender or tumor differentiation. No significant association was found with ptLVD or any of the clinicopathological criteria.

With analyzing association of LVD with clinicopathological feature, the three high ptLVD and low ptLVD lung carcinoma tissues were chosen for microarray screening. Using the criteria described in Materials and methods for microarray data analysis, we found 181 genes which showed a 2-fold difference. Among these, 97 genes are upregulated in high lymphatic vessels and 84 were downregulated. Upregulated and downregulated genes are reported in Table II. For each gene symbol, GenBank ID, fold-change and description are reported. Based on bioinformatics analysis and their respective functions reported in the literature, we selected 10 differentially expressed genes of interest and determined the expression levels by quantitative real-time RT-PCR. RT-PCR among these 10 genes demonstrated IGFBP7 to be the most differentially expressed. Thus, we chose to focus on IGFBP7 in the present study, whereas other candidates are being investigated in additional studies (Fig. 2).

The DAVID Bioinformatics Resource 6.7 was used for annotation of the cellular component. IPA was used to assign identified proteins into different functional groups based on the Ingenuity Pathways Analysis literature database. All the differentially expressed proteins were uploaded to the IPA server. For molecular and cellular functions, the data indicated that many proteins were involved in top 10 biological processes (Fig. 3A). Pathway analysis was also used to analyze different cellular functional networks to determine which were altered in the lymphangiogenesis procedure. In the present study, several networks were grouped by IPA, i.e. there are 20 proteins involved in cell death, cellular, development, hematological system development and function were grouped as the top network (network 1), which had the highest score (47) (Fig. 3B). Additionally, network 2 grouped by IPA has 12 related genes involved in endocrine system, disorders, metabolic disease, cell-to-cell signaling and interaction, and had a score of 25 (Fig. 3C).

In the NSCLC tissue, moderate IGFBPP7 immunoreactivity was present in the cytoplasm of lung cancer cells; a weak immunoreactivity was seen in some ductal cells within the small ductules. IGFBP7 positive substances that were in the shape of brownish-yellow fine particles were mainly located in the cytoplasm of the cancer cells. There was expression of various degrees on the cell membrane, presenting pale-yellow to brownish-yellow (Fig. 4).

The associations of high expression levels of IGFBP7 with clinicopathological parameters are shown in Table III. The positive expression of IGFBP7 is 55.67% (54/97). We observed a significant association between IGFBP7 expression in NSCLC associated with increased lymph node metastasis (P=0.002). Furthermore, ptLVD of the positive IGFBP7 is 23.1±8.5/field, statistically significantly higher than the negative group (16.9±6.0). There were no statistically significant differences with regard to patient age, gender or histological types (Table III).

In the co-culture system, different cells were placed in the upper chambers, while LECs were in the lower chambers. After 48 h of LLC-LECs co-culture, LECs start establishing a complex tube system. After 72 h of culture in conditioned medium, the number of LECs permeating septum in the LLC group was the highest among the four groups. LLCs more strongly enhanced the formation of extensive capillary-like structures in vitro than medium alone or the L929 cells group (P<0.05). Furthermore, compared to the LLC group, the LLC-Si-IGFBP7 group decreased LEC formation of capillary-like structures in vitro (P<0.05) (Fig. 5).