Cancer Letters 314 (2012) 127–136

Contents lists available at SciVerse ScienceDirect

Cancer Letters

journal homepage: www.elsevier .com/locate /canlet

Vascular endothelial growth factor-D promotes growth,lymphangiogenesis and lymphatic metastasis in gallbladder cancer

Wei Lin a,1, Lei Jiang a,1, Yanling Chen a,⇑, Feifei She b,⇑, Shenghua Han a, Jinhai Zhu a,Liangyi Zhou a, Nanhong Tang a, Xiaoqian Wang a, Xiujin Li a

a Department of Hepatobiliary Surgery, Union Clinical Medical College, Fujian Medical University, Fuzhou 350001, Chinab Department of Pathogenic Microorganism, College of Basic Medical Sciences, Fujian Medical University, Fuzhou 350004, China

a r t i c l e i n f o

Article history:Received 17 July 2011Received in revised form 4 September 2011Accepted 5 September 2011

Keywords:Gallbladder cancerVEGF-DGrowthLymphangiogenesisMetastasis

0304-3835/$ - see front matter � 2011 Elsevier Ireldoi:10.1016/j.canlet.2011.09.004

⇑ Corresponding authors. Tel.: +86 13365910313514063583 (F. She).

E-mail addresses: ylchen@medmail.com.cn (Y. Ch(F. She).

1 These authors are contributed equally to this pap

a b s t r a c t

Lymph node metastasis is a major prognostic factor for patients with gallbladder cancer(GBC), and greater understanding of the molecule mechanism of lymph node metastasisin GBC is needed to improve prognosis. VEGF-D has been implicated in the control of lym-phangiogenesis in many carcinomas, but the biological function of VEGF-D in human GBCremains unclear. In this study, we analyzed the role of the VEGF-D in human GBC cells andaddressed the functional role of VEGF-D using a xenograft mouse model. We examined theexpression of VEGF-D in three human gallbladder cancer cell lines. A lentivirus-based effec-tive VEGF-D siRNA vector was infected into GBC NOZ cells. The effect of VEGF-D siRNA onGBC NOZ cells was investigated by cell proliferation assay and invasion assay. Furthermore,we examined the role of VEGF-D-SiRNA on GBC NOZ cells in the mice of subcutaneous andorthotopic xenograft tumor. Our results are as follows: VEGF-D mRNA and protein wereexpressed in all three GBC cell lines (GBC-SD, NOZ, and SGC-996). We successfully selectedD-3/siRNA as the most effective siRNA to silence VEGF-D expression after four VEGF-D siR-NA plasmid transfection in NOZ cells. VEGF-D mRNA and protein expression were sup-pressed by lentivirus-mediated D-3/siRNA. D-3-RNAi-LV inhibited NOZ cells proliferationand invasion ability in vitro. D-3-RNAi-LV inhibited tumor growth and lymphangiogenesisin the NOZ cell subcutaneous xenograft model. D-3-RNAi-LV inhibited lymphangiogenesisand lymphatic metastasis in the NOZ cell orthotopic xenograft model. Furthermore, D-3-RNAi-LV inhibited tumor ascites and hepatic invasion in the NOZ cell orthotopic xenograftmodel. In conclusion, VEGF-D is involved and plays an important role in GBC progression,suggesting that VEGF-D may be a potential molecular target in the treatment of GBC.

� 2011 Elsevier Ireland Ltd. All rights reserved.

1. Introduction

Gallbladder cancer (GBC), a highly lethal disease, is themost common malignant neoplasm of the biliary tract andthe seventh most common gastrointestinal cancer [1]. GBChas an extremely poor prognosis. The median survival for

and Ltd. All rights reserved.

68 (Y. Chen), +86

en), cylsff@tom.com

er.

suspected carcinomas is 9.2 months, and for incidental car-cinomas is 26.5 months [2]. Surgical resection is the onlypotentially curative therapy for GBC. Unfortunately, onlya minority of patients with GBC are candidates for surgicaltreatment, mostly due to the high proportion of tumor thatis advanced at the time of presentation [3]. Apart fromsurgery, all other treatment approaches are palliative.Adjuvant chemotheraphy and/or radiotherapy also hasnot altered the dismal prognosis and survival of patientswith GBC [4]. Local tumor growth, hepatic invasion andlymph node metastasis are the main prognostic factors inpatients with GBC, especially the extent of lymph node

128 W. Lin et al. / Cancer Letters 314 (2012) 127–136

metastasis. The overall rate of lymph node metastasis inGBC ranges from 54% to 64% [5]. Therefore, greater under-standing of the molecule mechanism of lymph nodemetastasis in GBC is needed to improve prognosis.

Whereas the significance of angiogenesis for tumorgrowth and metastasis has been well understood, themolecular mechanisms regulating tumor-associated lym-phangiogenesis and lymphatic metastasis has only re-cently started to be elucidated [6,7]. With the discoveryof lymphatic vessel-specific markers such as lymphaticvessal hyaluronan receptor-1 (LYVE-1) [8]; the membranemucoprotein, podoplanin [9]; prospero-related homeo-boxl, Prox-1 [10] and monoclonal antibody directedagainst M2A antigen (D2-40) [11], which have providedvaluable tools to discriminate between blood and lympha-tic endothelium, lymphangiogenesis has accordingly be-come a focus of intense interest. It will be a new waythat inhibit lymphangiogenesis and lymphatic metastasisof growing tumors.

Vascular endothelial growth factor (VEGF)-C and -D,two novel members of the VEGF family, are known as lym-phatic endothelial growth factors. The VEGF family in-cludes VEGF-A, VEGF-B, VEGF-C, VEGF-D, VEGF-E, VEGF-Fand placental growth factor (PlGF) [12]. VEGFs have a crit-ical role in pathological conditions by binding to and acti-vating VEGFRs, which includes VEGFR-1, VEGFR-2, andVEGFR-3. VEGF-C and VEGF-D, also known as c-fos-in-duced growth factor [13], are initially synthesized as prep-ropeptides that become proteolytically processed outsidethe cell to produce several polypeptides including a matureform [14,15]. This processing progressively increases theaffinity for its receptors, VEGFR-2 and VEGFR-3, whichare predominantly present on endothelium of blood ves-sels and lymphatic endothelial cells, respectively [16].VEGF-C and VEGF-D have been implicated in the controlof lymphangiogenesis in many carcinomas, such as humanmalignant melanoma [17], squamous cell carcinomas ofthe head and neck region [18], gastric adenocarcinoma[19] and breast cancer [20]. Our previous study [21] alsodemonstrated that expression of VEGF-C and -D correlatedwith evidence of lymphangiogenesis and lymph nodemetastasis in tissue samples of human GBC. However, thebiological function of VEGF-D in human GBC have not beenexamined detailedly in vitro and in vivo so far. Thus, weanalyzed the role of the VEGF-D in human GBC cells andaddressed the functional role of VEGF-D using a xenograftmouse model in this study.

2. Materials and methods

2.1. Cell culture

Three cell lines established from human GBC were incu-bated at 37 �C under 95% air and 5% CO2. The GBC-SD cellline was purchased from Shanghai Institutes for BiologicalSciences in China, and the SGC-996 cell line (from the pri-mary mastoid adenocarcinoma of the gallbladder [22]) wasprovided by the Tumor Cytology Research Unit, MedicalCollege, Tongji University, China. Both the cell lines weremaintained in Dulbecco’s Modified Eagle’s Medium(DMEM) containing 10% (v/v) calf bovine serum. The NOZ

cell line (isolated from ascites derived from a 48-year-oldfemale patient with GBC [23]) was obtained from theHealth Science Research Resources Bank in Japan. TheNOZ cell line was cultured in Williams E medium (GIBCO)supplemented with 10% fetal bovine (FBS).

2.2. Reverse transcription-polymerase chain reaction (RT-PCR)

PCR primers used for amplification of VEGF-D were 50-CCA TCG GTC CAC TAG GTT-30 (forward sequence) and 50-GTG TCT GGG TGA AAT AGC-30 (reverse sequence), and ityielded a 684-bp product. PCR primers used for amplifica-tion of b-actin were 50-GGC ATG GGT CAG AAG GAT TCC-30

(forward sequence) and 50-ATG TCA CGC ACG ATT TCCCGC-30 (reverse sequence), and it yielded a 500-bp product.RT-PCR and the detection of the PCR products were carriedout as described previously [24].

2.3. Western blot analysis

The appropriate diluted primary antibodies, includingpolyclonal goat anti-human VEGF-D antibody (1:1000;Santa Cruz), b-actin (1:1000; Santa Cruz) were used in thisexperiment. Western blot and detection of the blottedproducts were carried out as described previously [24].

2.4. VEGF-D siRNA plasmid construction and transfection

According to the siRNA design guidelines, four suitablesiRNA target sequences (D-1, D-2, D-3, and D-4), from hu-man VEGF-D gene GenBank accession no. NM_004469,were synthesized as follows: D-1 (GGG CTC CAG TAATGA ACA TGG); D-2 (GCC AAT CAT ACA GGT TGT AAG);D-3 (GCT ATG GGA TAG CAA CAA ATG); and D-4 (GCAGTT GCT TTG AGT GCA AAG). A small hairpin RNA (shRNA)of human VEGF-D in a pGPU6 gene transfer vector and thenegative control sequence were constructed by Genephar-ma Co., Ltd. (Shanghai, China). All the plasmids were veri-fied by DNA sequencing. The process of transfection wascarried out as described previously [24]. At the time of72 h after transfection, cells were collected for detectingthe VEGF-D expression by RT-PCR.

2.5. Construction of lentiviral vectors and infection

In order to create better stable transfection for animalexperiment, we used lentiviral-mediated siRNA targetingVEGF-D vector. Lentivirus vectors for human VEGF-D smallhairpin RNA (shRNA) encoding a red fluorescent protein(RFP) sequence was constructed by Genechem Co., Ltd.(Shanghai, China). The lentivirus vectors containingVEGF-D shRNA were constructed by ligating the Age I/Eco I digests of pGCSIL-RFP and the VEGF-D shRNA PCRproduct and were verified by DNA sequencing. ShRNA forthe negative controls was also provided by GenechemCo., Ltd. (Shanghai, China). NOZ cells (5 � 104 cells/well)were plated in 6-well plates with Williams E medium(GIBCO) containing 10% FBS overnight. At the time ofinfection, the cells were at approximately 30% confluency.The appropriate quantity of virus (the multiplicity of

W. Lin et al. / Cancer Letters 314 (2012) 127–136 129

infection (MOI) = 50) was mixed with 1 ml complete med-ium containing polybrene (8 mg/mL) and added to cellsand incubated for 10 h at 37 �C. Then, the cells were incu-bated in fresh complete medium containing 10% FBS for20 h. The infection efficiency was quantified by determin-ing the percentage of cells that were RFP positive in every100 cells with a microscope. The cells were collected forsubsequent studies at the time of 72 h after infection.

2.6. Cell proliferation assay

Cell proliferation was assessed by MTT assay. NOZ cellswere incubated in 96-well plates at a density of8 � 103 cells per well with Williams E medium supple-mented with 10% FBS for 48 h. The proliferative activitywas determined by adding 10 ll of sterile MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide)(5 mg/ml, Sigma) to each well and incubated for 4 h at37 �C. Then 100 ll of dimethylsulphoxide (DMSO; Sigma)was added and thoroughly mixed for 10 min. The opticaldensity (OD) value was obtained by measuring absorbanceat a wavelength of 570 nm. The cell growth rate was calcu-lated by following formula: Cell growth rate = (meanabsorbance in all wells of the treatment group)/(meanabsorbance in all wells of the cells control group) � 100%.Each growth assay was repeated three times.

2.7. Cell in vitro invasion assay

The in vitro cell invasive assay was done in matrigelinvasion chambers (24-well) (BD Biosciences). The topand bottom of the cell invasion chamber were separatedby a polycarbonate filter with pores of 8-lm, which wascoated with matrigel matrix (20 lg/well) above and 5 lgFibronectin (BD) below. Approximately 3 � 104 cells weresuspended in 200 ll of Williams E medium with 2.5% FBSand were seeded onto the top chamber, and 600 ll of cul-ture medium with 10% FBS was added to the lower cham-ber. The invasion chamber was incubated for 48 h at 37 �Cand 5% CO2. The noninvading cells on the upper surface ofthe membrane were removed by gentle scrubbing with acotton swab. Membranes were fixed in a stationary liquid(95% ethanol and 5% acetic acid) for 30 min and stainedwith H&E. The invasive activity of cancer cells was deter-mined by counting the cells with a microscope at 200�magnification. Five random visual fields were counted foreach well, and the average was determined.

2.8. Animal experiments

Four-to 6-week-old nu/nu nude mice were purchasedfrom Shanghai SLAC Laboratory Animals Co. (Shanghai, Chi-na) and housed in the Experimental Center of the FujianMedical University. All experiments were carried out inaccordance with institutional guidelines and were approvedby the Ethics Committee of the Medical Faculty of the FujianMedical University. For induction of subcutaneous GBC tu-mors, three groups (blank control group, Con-RNAi-LVgroup and D-3-RNAi-LV group) of NOZ cells (1 � 106) in0.1 ml serum-free Williams E medium were implantedsubcutaneously into the left flank of mice (five mice/group).

Tumor growth was monitored every 3 days and measured intwo dimensions. Tumor volume was calculated using theformula width2 (mm2) � length (mm)/2, where width andlength are the shortest and longest diameters, respectively.After 4 weeks mice were killed and primary tumor as wellas axillary and inguinal lymph nodes were collected. Allsamples were fixed in 10% formalin solution for next immu-nohistochemical analysis. In order to induce orthotopic GBCtumors, we refer to the method which has been describedpreviously [25]. Briefly, nude mice (five mice/group) wereanesthetized and laparotomy was done. Three groups ofNOZ cells (3 � 105) in 30 ll 50% precooling matrigel (BD Bio-sciences) were injected into the gallbladder taking cautionto ensure that the tumor cells were not injected into the bil-iary tract. After waiting for approximately 30 s, the Matrigelsolution was solid. The gallbladder and liver lobes were re-placed and the abdominal wall closed using Vicryl 6/0. After4 weeks mice were killed and abdominal and thoracic cavityas well as all visceral organs were macroscopically exam-ined for tumor dissemination. Later, primary tumor as wellas hilar and mesenteric lymph nodes were excised. All sam-ples were fixed in 10% formalin solution for next immuno-histochemical analysis.

2.9. Immunohistochemical analysis

Immunohistochemical staining was performed usingthe standard immunoper-oxidase staining procedure. Inbrief, serial 4-lm-thick sections were cut from formalin-fixed and paraffin-embedded tumor blocks, dewaxed in xy-lene, rehydrated in alcohol, and then incubated with fresh3% hydrogen peroxide (H2O2) for 20 min at room tempera-ture. After washing with phosphate-buffered saline (PBS),the tissue sections were antigen-retrieved by heating thesections in a microwave for 13 min in a citric acid buffersolution (pH 6.0). Sections were blocked with appropriatenormal serum in PBS. VEGF-D polyclonal goat anti-humanantibody (1:80, Santa Cruz Biotechnology), LYVE-1 poly-clonal goat anti-mouse antibody (1:150, R&D Systems),CD31 polyclonal goat anti-mouse antibody (1:200, SantaCruz Biotechnology) and anti-pan-cytokeratin (CK) anti-body (1:100, Maixin-bio, Fuzhou, China) were diluted andplaced on the sections overnight in humidified boxes at4 �C. The sections were then washed with PBS for 5 minfollowed by an incubation with an UltraSensitive S-P kit(Maixin-bio, Fuzhou, China) according to the manufac-turer’s instructions. After exposure to stable 3,3-diam-inobenzidine for 5–10 min, slides were counterstainedwith hematoxylin, dehydrated, and mounted. For a nega-tive control, PBS was used instead of the primary antibody.The expression of VEGF-D was evaluated by using digitalmicroscopic imaging system and Quantity One image anal-ysis software (Bio-Rad Inc., USA). Three slices were ran-domly selected for each group, and then five fields wereselected for each slice. VEGF-D expression in each groupwas semiquantitatively evaluated based on the mean ofpositive area and optical density in fifteen fields respec-tively. The average of CD31-positive and LYVE-1-positivevessels were assessed according to the method describedby Takammi I [26]. Each slide was first scanned at 100�magnification to determine three ‘‘hot spots’’ defined as

130 W. Lin et al. / Cancer Letters 314 (2012) 127–136

areas with the maximum number of positive vessels. Thepositive vessel density was determined by counting allthe immunostained vessels at a 400� magnification andthe mean number of positive vessels was calculated inthe five selected areas for each case.

2.10. Statistical analysis

All statistical analyses were performed using the SPSS(the Statistical Package for Social Science) software (ver-sion 17.0). Data were analyzed by the v2 test and Student’st-test. Results are expressed as means ± SE and were con-sidered to be statistically different at P < 0.05.

3. Results

3.1. Expression of VEGF-D in human GBC cells

To determine the expression of VEGF-D in human GBC cells, we ana-lyzed three different GBC cell lines (GBC-SD, NOZ, and SGC-996) in our lab-oratory. All three gallbladder cancer cell lines constitutively expressedVEGF-D mRNA by RT-PCR, and the express intensity of VEGF-D mRNA inthree cell lines was relatively consistent (Fig. 1a). In addition, the secretedform 31-kDa VEGF-D protein could be precipitated from the total protein ofthese cell lines using an anti-human VEGF-D antibody (Fig. 1b). The VEGF-DmRNA expression level in three cell lines was relatively consistent, but NOZcells expressed relatively higher levels of VEGF-D protein when comparedto the other cell lines. So we used NOZ cells to next further study.

3.2. Expression of VEGF-D mRNA after siRNA plasmid transfection

In order to find an effective siRNA to silence VEGF-D expression, wedesigned four suitable siRNA target sequences from human VEGF-D geneGenBank accession no. NM_004469 according to the siRNA design guide-lines. DNA sequencing results verified that four VEGF-D siRNA plasmidconstruction were successful. We used RT-PCR to detect the VEGF-Dexpression at the time of 72 h after transfection in NOZ cells. RT-PCR indi-cated a decrease in VEGF-D mRNA levels in siRNA transfected cells, whilemRNA levels of the housekeeping gene b-actin remained relatively un-changed. The D-3/siRNA vector resulted in a greater suppression ofVEGF-D mRNA expression when compared to the other vectors (D-1/siR-NA, D-2/siRNA, and D-4/siRNA), while mock transfection or the D–N/siR-NA vector had no effect on VEGF-D mRNA expression (Fig. 2a). Thecorresponding densitometric ratios in the D-1/siRNA, D-2/siRNA, D-3/siR-NA, D-4/siRNA, and D–N/siRNA transfected groups were 68.7, 48.7, 28.0,68.0 and 86.0, respectively. There were distinct differences between D-3/siRNA and mock transfected siRNA groups (P < 0.05; Fig. 2b). Accordingto the above results, we selected D-3/siRNA for further studies.

3.3. Expression of VEGF-D mRNA and protein after D-3/siRNA lentiviralvectors infection

Next, we used lentiviral-mediated D-3/siRNA targeting VEGF-D vectorto create better stable transfection for next experiment. DNA sequencingresults verified that D-3/siRNA lentiviral vectors construction was

Fig. 1. Expression of VEGF-D in human GBC cell lines. (a) Total RNA from humanVEGF-D. b-actin was amplified for a positive control. (b) The VEGF-D proteins inwas used as a positive control.

successful. Transfection efficiency was quantified by counting the cellsunder a fluorescent microscope 72 h after infection, and the efficiencyof two groups (Con-RNAi-LV and D-3-RNAi-LV) were greater than 80%(Fig. 3). The RT-PCR results indicate that endogenous VEGF-D mRNAexpression was significantly inhibited at 72 h after infection in NOZ cells.Compared with the Con-RNAi-LV group, the D-3-RNAi-LV group showedrelatively lower quantities of VEGF-D mRNA (Fig. 4a). The semiquantita-tive analysis revealed that D-3-RNAi-LV suppressed VEGF-D mRNA levelsto 37.91% when normalized to the blank control, and there were signifi-cant differences between blank control group and D-3-RNAi-LV group(P < 0.05; Fig. 4b). In accordance with this, western blotting showed thatVEGF-D protein expression was suppressed in the D-3-RNAi-LV groupcompared with the Con-RNAi-LV group in NOZ cells (Fig. 4c). The corre-sponding densitometric ratios in the D-3-RNAi-LV and Con-RNAi-LVgroups were 23.28 and 97.66, respectively. There were distinct differ-ences between D-3-RNAi-LV and blank control groups (P < 0.05; Fig. 4d).

3.4. VEGF-D promotes NOZ cell proliferation and invasion in vitro

To experimentally address the biological function of VEGF-D in vitro,we analyzed the effect of lentivirus-mediated VEGF-D siRNA on NOZ cellproliferation and invasion by MTT assay and cell in vitro invasion assay,respectively. MTT assay results show that compared to blank controland Con-RNAi-LV groups, the cell growth of the D-3-RNAi-LV group wasslower (Fig. 5a). The cell viability of D-3-RNAi-LV was reduced to 41.3%when compared with blank control group (P < 0.01). The data from thecell in vitro invasion assay indicated that the number of cells that pene-trated the basal membrane from the D-3-RNAi-LV group (35.6 ± 4.0)was significantly less than the blank control group (46.8 ± 5.3; P < 0.01),which was similar to the number of cells from the Con-RNAi-LV group(44.8 ± 3.9; P > 0.05) (Fig. 5b).

3.5. VEGF-D induces tumor growth and increases lymphangiogenesis in themice of subcutaneous xenograft tumor

To then address the impact of VEGF-D expression on GBC tumorbiology in vivo, three groups (blank control group, Con-RNAi-LV groupand D-3-RNAi-LV group) of NOZ cells were subcutaneously xenograftedonto nude mice. As shown in Fig. 6, at the end of the 4-week experimentalperiod, the xenograft tumor sizes are as follows (mean ± SE): (2229.6 ±942.5) mm3 in the blank control group (1909.8 ± 896.1) mm3 in theCon-RNAi-LV group, and (571.0 ± 126.7) mm3 in the D-3-RNAi-LV group.There were distinct differences between D-3-RNAi-LV group and Con-RNAi-LV group (P < 0.01). Compared with Con-RNAi-LV group, delayedtumor growth was significant in the D-3-RNAi-LV group and was evidentfrom 2-week after starting xenograft until the day the mice were sacri-ficed. Next, we analyzed the expression of VEGF-D by immunohistoche-my, and its effect on lymphangiogenesis and angiogenesis was alsodetermined by immunohistochemical analysis using anti-LYVE-1 andanti-CD31 antibodies respectively. Compared with the Con-RNAi-LV andblank control groups, the expression level of VEGF-D was significantlydecreased in the D-3-RNAi-LV group (Fig. 7A). Typical results and quanti-tative data from microlymphatic density (MLVD) and the microvesseldensity (MVD) analysis are also shown in Fig. 7B and C, respectively. Mostlymphatic vessels are located in the peritumoral xenograft tumor, but lit-tle in intratumoral xenograft tumor. A statistically significant decrease inMLVD was noted in D-3-RNAi-LV group tumors (12.0 + 2.0) when com-pared to Con-RNAi-LV (20.6 ± 2.5) or blank control groups tumors

GBC cell lines was analyzed by means of RT-PCR using primers for humanhuman GBC cell lines were detected by western blotting. b-actin protein

Fig. 2. Effects of four siVEGF-D treatments on the mRNA expression of VEGF-D. (a) VEGF-D mRNA expression was analyzed by RT-PCR after treatment withsiRNA. b-actin was used as an internal control. (b) Semiquantitative analysis of VEGF-D mRNA. The densitometric value for each group was normalized tothe internal control and relative expression level with the following equation: % = (normalized value of each siRNA group)/(normalized value of mocktransfected group) � 100% (Mean ± SD, n = 3, and P < 0.05).

Fig. 3. Expression of RFP (400�) in NOZ cells 72 h after infection. (a and b) The Con-RNAi-LV group cells of red light vision and white vision. (c and d) The D-3-RNAi-LV group cells of red light vision and white vision. The red fluorescence (RFP) demonstrates successful infection, and the efficiency was greater than80%. There was no significant difference between the groups (P > 0.05; data not shown).

W. Lin et al. / Cancer Letters 314 (2012) 127–136 131

(20.2 ± 1.7) (P < 0.01). However, there are no statistical significance inMVD between D-3-RNAi-LV group tumors (34.3 ± 6.2) and Con-RNAi-LV(34.6 ± 6.4) or blank control groups tumors (35.4 ± 3.2) (P > 0.05). Thesedata suggest that inhibition of VEGF-D in xenograft tumors by D-3-RNAi-LV decreases tumor growth and lymphangiogenesis.

3.6. VEGF-D increases lymphangiogenesis and promotes lymphatic metastasisin the mice of orthotopic xenograft tumor

Scince the positive lymph nodes are not found in the mouse model ofsubcutaneous xenograft tumor, we try to induce the mouse model oforthotopic GBC tumors according to the method which has been de-scribed previously [25]. We have successfully established the mouse

model of orthotopic GBC tumors by using NOZ cells, and also found thathepatoduodenal ligament could seen enlarged lymph nodes, which weremostly positive by detected human cytokeratin positive GBC cells usingan anti-human pancytokeratin antibody (Fig 8c and d). Three groups(blank control group, Con-RNAi-LV group and D-3-RNAi-LV group) ofNOZ cells were orthotopicly xenografted into the gallbladder of nudemice and allowed to form tumors for 28 days. When mice were sacrificed,4 out of 5 mice of Con-RNAi-LV/blank control groups tumors had devel-oped ascites, whereas none of the animals with D-3-RNAi-LV group tu-mors had ascites at the time of autopsy (Fig. 8a and Table 1) (P < 0.05).Hepatic invasions were macroscopically detected in 5 out of 5 livers(100%) from mice of Con-RNAi-LV/blank control groups tumors comparedto 0 out of 5 livers from mice of D-3-RNAi-LV group tumors (Fig. 8b andTable 1) (P < 0.001). We also analyzed the expression of VEGF-D,

Fig. 4. Downregulating VEGF-D mRNA and protein expression by lentivirus-mediated VEGF-D siRNA in NOZ cells. (a) VEGF-D mRNA expression wasanalyzed by RT-PCR after treatment with D-3-RNAi-LV. b-actin was used as an internal control. (b) Semiquantitative analysis of VEGF-D mRNA. Thedensitometric value for each group was normalized to the internal control and relative expression level with the following equation: % = (normalized valueof each lentivirus-mediated siRNA group)/(normalized value of blank control group) � 100% (Mean ± SD, n = 3, and P < 0.05). (c) VEGF-D protein expressionwas semiquantified by western blotting after treatment with D-3-RNAi-LV. b-actin gene was used as an internal control. (d) Semiquantitative analysis ofVEGF-D protein. The densitometric value for each group was normalized to the internal control and relative expression level with the same equationdescribed in b.

Fig. 5. Inhibiton of gallbladder cancer NOZ cell proliferation and invasive ability by lentivirus-mediated VEGF-D siRNA. (a) D-3-RNAi-LV decreased theproliferation of NOZ cells, which were incubated in Williams E medium with 10% FBS for 48 h. Cell proliferation was assessed by a MTT assay, and the MTTassay was completed as described in Section 2. (b) Invasion of NOZ cells was inhibited by infection with D-3-RNAi-LV as measured at 48 h in the Williams Emedium supplemented with 2.5% FBS. Invasive cells were stained with H&E on the lower side of the chamber (200�). ��P < 0.01.

132 W. Lin et al. / Cancer Letters 314 (2012) 127–136

lymphangiogenesis and angiogenesis by immunohistochemical analysis.Compared with the Con-RNAi-LV and blank control groups, the expres-sion level of VEGF-D was significantly decreased in the D-3-RNAi-LVgroup (P < 0.01) (Fig. 9a). A statistically significant decrease in MLVDwas noted in D-3-RNAi-LV group tumors (14.8 ± 2.1) when compared toCon-RNAi-LV (49.1 ± 3.0) or blank control groups tumors (51.5 ± 3.8)(P < 0.01) (Fig. 9b). However, there are no statistical significance inMVD between D-3-RNAi-LV group tumors (55.0 ± 1.6) and Con-RNAi-LV(56.4 ± 2.4) or blank control groups tumors (56.4 ± 1.9) (P > 0.05)(Fig. 9c). Importantly, the inhibition of lymphangiogenesis was paralleledby an inhibition of lymphatic spread. Metastases were found in 0 out of 5(0%) lymph nodes from mice of D-3-RNAi-LV group tumors compared to 5out of 5 (100%) lymph nodes from mice of Con-RNAi-LV/blank controlgroups tumors (P < 0.05) (Fig. 8c, d and Table 1).

4. Discussion

GBC has an extremely poor prognosis because of metas-tasis. Lymphatic spread is a major component in the GBCmetastasis [5], and it is also the most important prognosticfactor in patients with GBC. Thus, the therapeutic methodof inhibiting lymphatic spread would be predicted to con-vey a major clinical benefit in GBC. But how lymphaticmetastasis arises from the primary tumor is obscure. Re-cent studies have demonstrated a principal role of mole-cules in the mechanism of lymphogeneous metastasis,

Fig. 6. Suppression of tumor growth by lentivirus-mediated VEGF-DsiRNA in NOZ cell xenografts. Three groups (blank control group, Con-RNAi-LV group and D-3-RNAi-LV group) of NOZ cells were subcutane-ously xenografted onto nude mice. The size of the primary tumors wasmeasured every 3 days. Mice were sacrificed after 4 weeks. Points, meantumor volume in each group; bars, SE. �Significantly different from Con-RNAi-LV (P < 0.01).

W. Lin et al. / Cancer Letters 314 (2012) 127–136 133

such as VEGF-D and VEGFR-3 [27]. VEGF-D (also known asc-fos-induced growth factor or FIGF), which is similarly toVEGF-C, has been identified as a kind of essential agent forthe growth and establishment of lymphatic vessels in

Fig. 7. Effects of lentivirus-mediated VEGF-D siRNA on lymphangiogenesis andharvested and immunostained with VEGF-D, LYVE-1 and CD-31 antibody to assesD expression in tumor cells. Shown are quantitative evaluation of the results (rigLV and D-3-RNAi-LV tumors (left panel). Bars represent the mean OD ± SEM. (b) Lof the results (right panel) and representative LYVE-1 staining on blank control, CMLVD ± SEM. (c) CD-31-positive microvessel. Shown are quantitative evaluationcontrol, Con-RNAi-LV and D-3-RNAi-LV tumors (left panel). Bars represent the m

many human tumors, such as gastric carcinoma [19], hepa-tocellular carcinoma [27], lung cancer [28], colorectal can-cer [29] and pancreatic cancer [30]. In human GBC, ourprevious study [21] has demonstrated that the proteinexpression of VEGF-D in tumor tissues was significantlyhigher than that in normal gallbladder away from the tu-mor, and it also correlated with evidence of lymphangio-genesis and lymph node metastasis in tissue samples ofhuman GBC. In order to further study VEGF-D, we next ex-plored detailedly their biological function in vitro andin vivo.

To our knowledge, this study is the first to demonstratethe role of VEGF-D in promoting growth, lymphangiogene-sis and lymphatic metastasis in GBC. In this study, we usedRT-PCR and Western blot techniques to analyze the mRNAand protein expression of VEGF-D in three different GBCcell lines (GBC-SD, NOZ, and SGC-996). The mRNA and pro-tein levels of VEGF-D were detected in all the three celllines, which indicated that VEGF�D is one of the importantmolecules in participating in process of GBC metastasis.The mRNA expression levels of VEGF-D was relatively con-sistent in three cell lines, but NOZ cells expressed relativelyhigher levels of VEGF-D protein when compared to theother cell lines. This may be due to different mechanismsof protein translation in the different cells. The resultswere similar to our previous research [24], in which foundNOZ cells expressed relatively higher levels of VEGF-Ccompared to the relatively lower mRNA levels in NOZ cells.

angiogenesis in NOZ cell subcutaneous xenograft tumor. Tumors weres lymphangiogenesis and angiogenesis, as described in Section 2. (a) VEGF-ht panel) and representative VEGF-D staining on blank control, Con-RNAi-YVE-1-positive microlymphatic vessel. Shown are quantitative evaluation

on-RNAi-LV and D-3-RNAi-LV tumors (left panel). Bars represent the meanof the results (right panel) and representative CD-31 staining on blankean MVD ± SEM. �Significantly different from Con-RNAi-LV (P < 0.01).

Fig. 8. Ascites, hepatic invasion and lymph node metastase were seen in the mice model of orthotopic GBC tumors by using NOZ cells. (A) Presence ofascites in tumor bearing mice. (B) Presence of hepatic invasion in tumor bearing mice. Black arrows mark the invasive tumor in the liver of mice. (C) Lymphnode metastas can be seen in the hepatoduodenal ligament of mice. Black arrow marks enlarged lymph nodes. (D) Cytokeratin positive expression of cancercells in the hepatoduodenal ligament lymph node. Cytokeratin protein was detected using immunohistochemical techniques by an anti-humanpancytokeratin antibody. Black arrows show the CK positive expression of cancer cells in lymph node; red arrow shows the lymphoid follicles.

Table 1Effects of lentivirus-mediated VEGF-D siRNA on ascites, hepatic invasionand lymph node metastase in NOZ cell orthotopic xenograft tumor.

Blankcontrol

Con-RNAi-LV D-3-RNAi-LV

Animals with ascites (%) 4/5(80%) 4/5(80%) 0/5(0%)a

Hepatic invasions (%) 5/5(100%) 5/5(100%) 0/5(0%)a

Lymph node metastasis (%) 5/5(100%) 5/5(100%) 0/5(0%)a

a Significantly different from Con-RNAi-LV (P < 0.05).

134 W. Lin et al. / Cancer Letters 314 (2012) 127–136

Based on the protein level, we used NOZ cells to next fur-ther study.

SiRNA technology is the most widely used technique ofgene silencing, and provides a novel strategy for investigat-ing gene expression and function [31]. Compared with thetraditional genesilencing technology (such as antisense oli-go-nucleotides), SiRNA allows us to target any gene withgreater specificity, stability and efficiency [32–34]. In addi-tion, SiRNA plays an important role in the study of genefunction and has the potential for therapeutic applicationsin human cancer diseases [35,36]. Thus, we used siRNA toassess the effect of suppressing the VEGF-D gene in GBCin vitro and in vivo. In this study, we successfully selectedD-3/siRNA as the most effective siRNA to silence VEGF-Dexpression after four VEGF-D siRNA plasmid transfectionin NOZ cells. VEGF-D mRNA and protein expression wassuppressed by lentivirus-mediated D-3/siRNA. In vitro,

downregulation of VEGF-D by siRNA technology inhibitedthe proliferation and invasion of NOZ cells. Our resultswere similar to previous research, in which Tanaka et al.[37] found transfection with VEGF-D increased in vitro cellgrowth of human gastric carcinoma cells KKLS cells as a re-sult of an autocrine role of VEGF-D in gastric carcinoma.Goodyear et al. [38] also found that the VEGF-D leads to in-creased cell migration and invasion in primary prostatecell lines. The results suggest that VEGF-D can affect thebiological characteristics of tumor cells in GBC.

In our in vivo model of GBC, we similarly observed thatdownregulation of VEGF-D by siRNA inhibited growth ofsubcutaneous xenograft tumor, consistent with ourin vitro study. In addition to the induction of tumorgrowth, downregulation of VEGF-D by siRNA also signifi-cantly decreased lymphangiogenesis in the GBC subcuta-neous xenografts. Our results are consistent with study ofthe function of VEGF-D in other tumor entities [39–41],which indicated that VEGF-D played a key role in aninduction of lymphangiogenesis. For instance, VEGF-Dstimulated lymphangiogenesis when overexpressed inhepatocellular carcinoma [27] and promoted the develop-ment of lymphatics and lymph node metastasis in experi-mental tumors [30]. Since the positive lymph nodes werenot found in the mouse model of subcutaneous xenografttumor, we also induced the mouse model of orthotopicGBC tumors. In order to in line with human tumor natural

Fig. 9. Effects of lentivirus-mediated VEGF-D siRNA on lymphangiogenesis and angiogenesis in NOZ cell orthotopic xenograft tumor. (a) VEGF-D expressionin tumor cells. Shown are quantitative evaluation of the results (left panel) and representative VEGF-D staining on blank control, Con-RNAi-LV and D-3-RNAi-LV tumors (right panel). Bars represent the mean OD ± SEM. (b) LYVE-1-positive microlymphatic vessel. Shown are quantitative evaluation of theresults (left panel) and representative LYVE-1 staining on blank control, Con-RNAi-LV and D-3-RNAi-LV tumors (right panel). Bars represent the meanMLVD ± SEM. (c) CD-31-positive microvessel. Shown are quantitative evaluation of the results (left panel) and representative CD-31 staining on blankcontrol, Con-RNAi-LV and D-3-RNAi-LV tumors (right panel). Bars represent the mean MVD ± SEM. ��Significantly different from Con-RNAi-LV (P < 0.01).

W. Lin et al. / Cancer Letters 314 (2012) 127–136 135

metastasis in vivo, Fidler [42] proposed the tumor cell sus-pension was injected directly into the corresponding organrecipient in animal (i.e, orthotopic transplantation), whichcould improve tumor spontaneous metastasis. In theorthotopic transplantation model, the graft site is betterto the growth of tumor, and the tumor can display prefer-ably their biological characteristics, so it can be similar tothe way the human tumor invasion and metastasis. In thisstudy, we have successfully established the mouse modelof orthotopic GBC tumors by using NOZ cells, and alsofound that hepatoduodenal ligament could seen enlargedlymph nodes. Downregulation of VEGF-D by siRNA can de-crease significantly lymphangiogenesis and lymphaticmetastasis in the mice of orthotopic xenograft tumor. Thisindicated that VEGF-D can induce tumor growth, increaselymphangiogenesis and promote lymphatic metastasis inthe GBC. However, downregulation of VEGF-D by siRNAcannot decrease angiogenesis in the mice of subcutaneousand orthotopic xenograft tumor. Their results are inagreement with previous research, in which [43] VEGF-Dexpression did not increase or even reduce angiogenesisin transgenic approaches. Apart from less lymphangiogen-esis and lymph node metastasis, downregulation ofVEGF-D by siRNA in the experimental GBC tumors reducedascites and hepatic invasions. This might be associatedwith an decreased tumor cells proliferation and invasion.

In this study, VEGF-D mRNA and protein were ex-pressed in all three GBC cell lines (GBC-SD, NOZ, and

SGC-996). We successfully selected D-3/siRNA as the mosteffective siRNA to silence VEGF-D expression after fourVEGF-D siRNA plasmid transfection in NOZ cells. VEGF-DmRNA and protein expression was suppressed by lentivi-rus-mediated D-3/siRNA. D-3-RNAi-LV inhibited NOZ cellsproliferation and invasion ability in vitro. D-3-RNAi-LVinhibited tumor growth and lymphangiogenesis in theNOZ cell subcutaneous xenograft model. D-3-RNAi-LVinhibited lymphangiogenesis and lymphatic metastasis inthe NOZ cell orthotopic xenograft model. Furthermore,D-3-RNAi-LV inhibited tumor ascites and hepatic invasionin the NOZ cell orthotopic xenograft model.

Taken together, our results firmly link the expression ofVEGF-D to lymphangiogenesis and increased lymphaticnode metastasis in GBC xenograft models as well as tothe induction of tumor growth, ascites and hepatic inva-sions, indicating an important role for VEGF-D in GBC pro-gression. All this suggest that VEGF-D is involved in theprogression of GBC and represents a potential moleculartarget in the treatment of GBC.

Acknowledgments

This study was supported by the Key Sci-Tech ResearchFoundation of Fujian Province in China (No. 2009Y0021),and the Key Project of Fujian Medical University in China(No. 09ZD017).

136 W. Lin et al. / Cancer Letters 314 (2012) 127–136

References

[1] J.H. Donohue, A.K. Stewart, H.R. Menck, The National Cancer DataBase report on carcinoma of the gallbladder, 1989–1995, Cancer 83(1998) 2618–2628.

[2] C. Wullstein, G. Woeste, S. Barkhausen, et al., Do complicationsrelated to laparoscopic cholecystectomy influence the prognosis ofgallbladder cancer? Surg. Endosc. 16 (2002) 828–832.

[3] E.C. Lazcano-Ponce, J.F. Miquel, N. Munoz, R. Herrero, C. Ferrecio, I.I.Wistuba, P. Alonso de Ruiz, G. Aristi Urista, F. Nervi, Epidemiologyand molecular pathology of gallbladder cancer, CA Cancer J. Clin. 51(2001) 349–364.

[4] T.C. Chao, Y.Y. Jan, M.F. Chen, Primary carcinoma of the gallbladderassociated with anomalous pancreatobiliary ductal junction, J. Clin.Gastroenterol. 21 (1995) 306–308.

[5] A. Muratore, R. Polastri, L. Capussotti, Radical surgery for gallbladdercancer: current options, Eur. J. Surg. Oncol. 26 (2000) 438–443.

[6] S.A. Stacker, M.G. Achen, L. Jussila, M.E. Baldwin, K. Alitalo,Lymphangiogenesis and cancer metastasis, Nat. Rev. Cancer 2(2002) 573–583.

[7] K. Alitalo, P. Carmeliet, Molecular mechanisms of lymphangiogenesisin health and disease, Cancer Cell 1 (2002) 219–272.

[8] S. Banerji, J. Ni, S.X. Wang, S. Clasper, J. Su, R. Tammi, M. Jones, D.G.Jackson, LYVE-1, a new homologue of the CD44 glycoprotein, is alymph-specific receptor for hyaluronan, J. Cell Biol. 144 (1999) 789–801.

[9] S. Breiteneder-Geleff, A. Soleiman, H. Kowalski, R. Horvat, G. Amann,E. Kriehuber, K. Diem, W. Weninger, E. Tschachler, K. Alitalo, D.Kerjaschki, Angiosarcomas express mixed endothelial phenotypes ofblood and lymphatic capillaries: podoplanin as a specific marker forlymphatic endothelium, Am. J. Pathol. 154 (1999) 385–394.

[10] Tatiana V. Petrova, Taija Makinen, et al., Lymphatic endothelialreprogramming of vascular endothelial cells by the Prox-1homeobox transcription factor, EMBO J. 21 (17) (2002) 4593.

[11] A. Marks, D.R. Sutherland, D. Bailey, et al., Characterization anddistribution of an oncofetal antigen (M2A antigen) expressed ontesticular germ cell tumours, Br. J. Cancer 80 (3–4) (1999) 569.

[12] Z.K. Otrock, J.A. Makarem, A.I. Shamseddine, Vascular endothelialgrowth factor family of ligands and receptors: review, Blood CellsMol. Dis. 38 (2007) 258–268.

[13] M. Orlandini, L. Marconcini, R. Ferruzzi, S. Oliviero, Identification of ac-fos-induced gene that is related to the platelet-derived growthfactor/vascular endothelial growth factor family, Proc. Natl. Acad.Sci. USA 93 (1996) 11675–11680.

[14] V. Joukov, T. Sorsa, V. Kumar, et al., Proteolytic processing regulatesreceptor specificity and activity of VEGF-C, EMBO J. 16 (1997) 3898–3911.

[15] S.A. Stacker, K. Stenvers, C. Caesar, et al., Biosynthesis of vascularendothelial growth factor-D involves proteolytic processing whichgenerates non-covalent homodimers, J. Biol. Chem. 274 (1999)32127–32136.

[16] V. Joukov, K. Pajusola, A. Kaipainen, D. Chilov, I. Lahtinen, E. Kukk, O.Saksela, N. Kalkkinen, K. Alitalo, A novel vascular endothelial growthfactor, VEGF-C, is a ligand for the Flt4 (VEGFR-3) and KDR (VEGFR-2)receptor tyrosine kinases, EMBO J. 15 (1996) 290–298.

[17] S.S. Dadras, B. Lange-A sschenfeldt, et al., Tumor lymphangiogenesispredicts melanoma metastasis to sentinel lymph nodes, Mod. Pathol.18 (9) (2005) 1232–1242.

[18] S.M. Maula, M. Luukkaa, R. Grenman, et al., Intratumoral lymphaticsare essential for the metastatic spread and prognosis in squamous cellcarcinomas of the head and neck region, Cancer Res. 15 (2003) 63. 8.

[19] J.H. Choi, Y.H. Oh, Y.W. Park, H.K. Baik, Y.Y. Lee, I.S. Kim, Correlationof vascular endothelial growth factor-D expression and VEGFR-3-positive vessel density with lymph node metastasis in gastriccarcinoma, J. Korean Med. Sci. 23 (4) (2008) 592–597.

[20] M. Skobe, T. Hawighorst, D.G. Jackson, R. Prevo, L. Janes, P. Velasco, L.Riccardi, K. Alitalo, K. Claffey, M. Detmar, Induction of tumorlymphangiogenesis by VEGF-C promotes breast cancer metastasis,Nat. Med. 7 (2001) 192–198.

[21] L. Jiang, Y.L. Chen, F.F. She, et al., Expressions of VEGF-C and VEGF-Dand their correlation with lymphangiogenesis and angiogenesis ingallbladder carcinoma, Zhonghua Zhong Liu Za Zhi 32 (2010) 190–195.

[22] Yao-qin Yang, Hui-hong Tao, Hu-chuan Yang, Establishing of humanprimary gallbladder cancer cell line SGC-996, J. Tongji Univ. (Med.Sci.) 24 (6) (2003) 457–459.

[23] S. Homma, S. Hasumura, S. Nagamori, H. Kameda, Establishment andcharacterization of a human gall bladder carcinoma cell line NOZ,Hum. Cell 1 (1988) 95–97.

[24] Yanling Chen, Lei Jiang, Feifei She, et al., Vascular endothelialgrowth factor-C promotes the growth and invasion ofgallbladder cancer via an autocrine mechanism, Mol. Cell.Biochem. 345 (2010) 77–89.

[25] Jan-Hendrik Egberts, Bodo Schniewind, Clemens Schafmayer, et al.,Establishment of a novel orthotopic xenograft model of humangallbladder carcinoma, Clin. Exp. Metast. 24 (2007) 141–148.

[26] I. Takanami, Lymphatic microvessel density using D2–40 isassociated with nodal metastasis in non-small cell lung cancer,Oncol. Rep. 15 (2006) 437–442.

[27] A. Thelen, A. Scholz, C. Benckert, Z. von Marschall, M. Schroder, B.Wiedenmann, P. Neuhaus, S. Rosewicz, S. Jonas, VEGF-D promotestumor growth and lymphatic spread in a mouse model ofhepatocellular carcinoma, Int. J. Cancer 122 (2008) 2471–2481.

[28] T. Niki, S. Iba, M. Tokunou, T. Yamada, Y. Matsuno, S. Hirohashi,Expression of vascular endothelial growth factors A, B, C, and D andtheir relationships to lymph node status in lung adenocarcinoma,Clin. Cancer Res. 6 (2000) 2431–2439.

[29] J.D. White, P.W. Hewett, D. Kosuge, et al., Vascular endothelialgrowth factor-D expression is an independent prognostic markerfor survival in colorectal carcinoma, Cancer Res. 62 (2002)1669–1675.

[30] Zofia von Marschall, Arne Scholz, Steven A. Stacker, et al., Vascularendothelial growth factor-D induces lymphangiogenesis andlymphatic metastasis in models of ductal pancreatic cancer, Int. J.Oncol. 27 (2005) 669–679.

[31] I. Bantounas, L.A. Phylactou, J.B. Uney, RNA interference and the useof small interfering RNA to study gene function in mammaliansystems, J. Mol. Endocrinol. 33 (2004) 545–557.

[32] H. Kawasaki, K. Taira, Short hairpin type of dsRNAs that arecontrolled by tRNA (Val)promoter significantly induce RNAi-mediated gene silencing in the cytoplasm of human cells, NucleicAcids Res. 31 (2003) 700–707.

[33] S.M. Elbashir, J. Harborth, K. Weber, et al., Analysis of gene functionin somatic mammalian cells using small interfering RNAs, Methods26 (2) (2002) 199–213.

[34] S.M. Elbashir, J. Harborth, W. Lendeckel, et al., Duplexes of 21-nucleotide RNAs mediate RNA interference in cultured mammaliancells, Nature 411 (2001) 494–498.

[35] N.J. Caplen, S. Mousses, Short interfering RNA (siRNA)-mediated RNAinterference (RNAi) in human cells, Ann. NY Acad. Sci. 1002 (2003)56–62.

[36] D.M. Dykxhoorn, D. Palliser, J. Lieberman, The silent treatment:siRNAs as small molecule drugs, Gene Ther. 13 (2006) 541–552.

[37] Miwako Tanaka, Yasuhiko Kitadai, Michiyo Kodama, et al., Potentialrole for vascular endothelial growth factor-D as an autocrine factorfor human gastric carcinoma cells, Cancer Sci. 101 (10) (2010) 2121–2127.

[38] S.M. Goodyear, S.B. Kheyfets, F.U. Garcia, M.E. Stearns, Role of theVEGFR3 /VEGFD receptor axis in TGFbeta1 activation of primaryprostate cell lines, Prostate 69 (2009) 982–990.

[39] Y. Yokoyama, D.S. Charnock-Jones, D. Licence, A. Yanaihara, J.M.Hastings, C.M. Holland, M. Emoto, M. Umemoto, T. Sakamoto, S. Sato,H. Mizunuma, S.K. Smith, Vascular endothelial growth factor-D is anindependent prognostic factor in epithelial ovarian carcinoma, Br. J.Cancer 88 (2003) 237–244.

[40] H. Funaki, G. Nashimura, S. Harada, I. Ninomiya, I. Terada, S. Fushida,T. Tani, T. Fujimura, M. Kayahara, K. Shimizu, T. Ohta, K. Miwa,Expression of vascular endothelial growth factor D is associated withlymph node metastasis in human colorectal carcinoma, Oncology 64(2003) 416–422.

[41] S. Onogawa, Y. Kitadai, S. Tanaka, T. Kuwai, S. Kimura, K. Chayama,Expression of VEGF-C and VEGF-D at the invasive edge correlateswith lymph node metastasis and prognosis of patients withcolorectal carcinoma, Cancer Sci. 95 (2004) 32–39.

[42] L.J. Fidler, Critical factors in the biology of human cancer metastasis:twenty-eighth G.H.A. Clowes memorial award lecture, Cancer Res.50 (1990) 6130–6138.

[43] L. Kopfstein, T. Veikkola, V.G. Djonov, V. Baeriswyl, T. Schomber, K.Strittmatter, S.A. Stacker, M.G. Achen, K. Alitalo, G. Christofori,Distinct roles of vascular endothelial growth factor-D inlymphangiogenesis and metastasis, Am. J. Pathol. 170 (2007)1348–1361.